AmericAn Society for BiochemiStry And moleculAr Biology

C o m p e n d i a

Twenty-five Years of Exploration into Protein Science:

The JBC Celebrates The Protein Society’s

Anniversary Symposium

1 Twenty-five Years of Exploration into Protein Science: The JBCCelebrates The Protein Society’s Anniversary Symposium.H. Smith

2□S

Phylogenetic and Functional Analysis of Histidine ResiduesEssential for pH-dependent Multimerization of von WillebrandFactor. Luke T. Dang, Angie R. Purvis, Ren-Huai Huang,Lisa A. Westfield, and J. Evan Sadler

9□S

Characterization of Prefibrillar Tau Oligomers in Vitro and inAlzheimer Disease. Kristina R. Patterson, Christine Remmers,Yifan Fu, Sarah Brooker, Nicholas M. Kanaan, Laurel Vana,Sarah Ward, Juan F. Reyes, Keith Philibert, Marc J. Glucksman, andLester I. Binder

23 Mechanism of Intracellular cAMP Sensor Epac2 Activation.cAMP-INDUCED CONFORMATIONAL CHANGES IDENTIFIED BY AMIDEHYDROGEN/DEUTERIUM EXCHANGE MASS SPECTROMETRY (DXMS).Sheng Li, Tamara Tsalkova, Mark A. White, Fang C. Mei, Tong Liu,Daphne Wang, Virgil L. Woods, Jr., and Xiaodong Cheng

32□S

Shortened Engineered Human Antibody CH2 Domains.INCREASED STABILITY AND BINDING TO THE HUMAN NEONATALFc RECEPTOR. Rui Gong, Yanping Wang, Yang Feng, Qi Zhao,and Dimiter S. Dimitrov

38□S

An Undecided Coiled Coil. THE LEUCINE ZIPPER OF Nek2 KINASEEXHIBITS ATYPICAL CONFORMATIONAL EXCHANGE DYNAMICS.Rebecca Croasdale, Frank J. Ivins, Fred Muskett, Tina Daviter,David J. Scott, Tara Hardy, Steven J. Smerdon,Andrew M. Fry, and Mark Pfuhl

49□S

Interaction of Actin with Carcinoembryonic Antigen-relatedCell Adhesion Molecule 1 (CEACAM1) Receptor in Liposomes IsCa2�- and Phospholipid-dependent. Rongze Lu,Michiel J. M. Niesen, Weidong Hu, Nagarajan Vaidehi,and John E. Shively

58 Curcumin Modulates Nuclear Factor �B (NF-�B)-mediatedInflammation in Human Tenocytes in Vitro. ROLE OF THEPHOSPHATIDYLINOSITOL 3-KINASE/Akt PATHWAY.Constanze Buhrmann, Ali Mobasheri, Franziska Busch,Constance Aldinger, Ralf Stahlmann, Azadeh Montaseri,and Mehdi Shakibaei

69□S

Crystal Structure of Clustered Regularly Interspaced ShortPalindromic Repeats (CRISPR)-associated Csn2 ProteinRevealed Ca2�-dependent Double-stranded DNA BindingActivity. Ki Hyun Nam, Igor Kurinov, and Ailong Ke

The Journal of Biological ChemistryTABLE OF CONTENTS

2011 COMPENDIA COLLECTION: Twenty-five Years of Exploration intoProtein Science: The JBC Celebrates The Protein Society’s AnniversarySymposium

□S Online version of this article contains supplemental material.

JOURNAL OF BIOLOGICAL CHEMISTRY i

Twenty-five Years of Exploration into Protein Science: TheJBC Celebrates The Protein Society’s AnniversarySymposium*

H. SmithFrom the American Society for Biochemistry and Molecular Biology, Rockville, Maryland 20852

Proteins are special, which is precisely why they are called“proteins.”1 And so it goes, as one is known by the company onekeeps, that protein scientists are special. In the spirit of thesesimple truths, the Journal of Biological Chemistry (JBC) con-gratulates The Protein Society on twenty-five years of extraor-dinary success.The American Society for Biochemistry andMolecular Biol-

ogy (ASBMB) shares The Protein Society’s mission of advanc-ing protein science and supporting the creative researcherswho continue to revolutionize the ways that proteins are stud-ied and used in a variety of scientific, medical, engineering, andtechnological venues. Members of the ASBMB were instru-mental in founding The Protein Society in 1986, and theASBMB has been proud to have played many formal roles inpromoting The Protein Society since its inception.

The ASBMB is particularly pleased to be a sponsor of theTwenty-fifth Anniversary Symposium of the Protein Societythis year in Boston and likewise appreciates The ProteinSociety’s support of ASBMB’s new JBC/Herbert TaborYoung Investigator Award series launched this year.Throughout 2011, the JBC will be offering a series of com-pendia to highlight particular areas of biochemical research,and on the occasion of The Protein Society’s twenty-fifthanniversary, we have compiled the present compendium ofrecent research from the pages of the JBC. The current JBCpapers that are presented here are freely available, and wevery much hope that members of The Protein Society, and allthose who share in celebrating twenty-five years of phenom-enal protein science, will enjoy reading these basic investi-gations of proteins, their structures, and functions. The JBCremains committed to communicating the best of researchfindings in these important areas, and we look forward tosupporting the fascinating “translational” science of pro-teins, encompassing both basic and clinical biomedicaldimensions, in the years to come.

* To cite articles in this collection, use the citation information that appears inthe upper right-hand corner of the first page of the article.

1 See Merriam Webster’s Collegiate Dictionary (10th ed): protein … F[rench]proteine, from late Greek pro� teios, primary, from Greek pro� tos first … 1: anyof numerous naturally occurring extremely complex substances…

© 2011 by The American Society for Biochemistry and Molecular Biology, Inc. Printed in the U.S.A.PROLOGUE

1

Phylogenetic and Functional Analysis of Histidine ResiduesEssential for pH-dependent Multimerization of vonWillebrand Factor*□S �

Received for publication, April 8, 2011, and in revised form, May 13, 2011 Published, JBC Papers in Press, May 17, 2011, DOI 10.1074/jbc.M111.249151

Luke T. Dang, Angie R. Purvis, Ren-Huai Huang, Lisa A. Westfield, and J. Evan Sadler1

From the Department of Medicine and Department of Biochemistry and Molecular Biophysics, Washington University School ofMedicine, St. Louis, Missouri 63110

vonWillebrand factor (VWF) is a multimeric plasma proteinthat mediates platelet adhesion to sites of vascular injury. Thehemostatic function of VWF depends upon the formation ofdisulfide-linkedmultimers,which requires theVWFpropeptide(D1D2 domains) and adjacent D�D3 domains. VWF multimerassembly occurs in the trans-Golgi at pH�6.2 but not at pH 7.4,which suggests that protonation of one or more His residues(pKa �6.0) mediates the pH dependence of multimerization.Alignment of 30 vertebrate VWF sequences identified 13 highlyconserved His residues in the D1D2D�D3 domains, and His-to-Ala mutagenesis identified His395 and His460 in the D2 domainas critical forVWFmultimerization. Replacement ofHis395withLys or Arg prevented multimer assembly, suggesting thatreversible protonation of this His residue is essential. In con-trast, replacement of His460 with Lys or Arg preserved normalmultimer assembly, whereas Leu,Met, and Gln did not, indicat-ing that the function of His460 depends primarily upon the pres-ence of a positive charge. These results suggest that pH sensingby evolutionarily conservedHis residues facilitates the assemblyand packaging of VWF multimers upon arrival in thetrans-Golgi.

von Willebrand factor (VWF)2 is a multimeric hemostaticprotein that mediates platelet adhesion to sites of vascularinjury. VWF is assembled from identical 350-kDa precursorsconsisting of multiple structural domains in the order D1-D2-D�-D3-A1-A2-A3-D4-B1-B2-B3-C1-C2-CK (where CK indi-cates the cystine knot domain). The hemostatic function ofVWF depends upon its assembly intomultimers, and defects inVWF multimer formation are common causes of von Will-ebrand disease, the most prevalent inherited bleeding disorder(1).

The assembly of VWF multimers begins in the endoplasmicreticulum (ER) of endothelial cells, where pro-VWFmonomersform homodimers through “tail-to-tail” intersubunit disulfidebonds between C-terminal cystine knot domains. After trans-port to the Golgi, pro-VWF dimers assemble into large multi-mers through “head-to-head” N-terminal interchain disulfidebonds between D3 domains. Subsequent cleavage of VWF pro-peptide (D1D2) by furin yields mature, multimeric VWF that ispacked into tubular arrays and stored in rod-shaped granulescalled Weibel-Palade bodies (2).The completion of VWF multimer assembly in the Golgi

apparatus poses several challenges to the cell. Disulfidebonds typically form in the ER, where disulfide oxidoreduc-tases and chaperones are devoted to this function. However,the Golgi is relatively hostile to disulfide rearrangementbecause the lower pH inhibits deprotonation of the sulfhy-dryl moiety of Cys residues, which is necessary for disulfiderearrangement. In addition, suitable oxidoreductases andchaperones are thought to be absent from the Golgi. To facil-itate disulfide bond formation under these circumstances,the N-terminal D1D2D�D3 domains of the VWF precursoract as an endogenous pH-dependent oxidoreductase.Deletion of the VWF propeptide prevents multimerization,

but expression of the propeptide and mature subunit on sepa-rate plasmids (in trans) restores efficient multimer assembly(3). In the ER, the VWFpropeptide forms a transient intrachaindisulfide-linked intermediate with the D3 domain that rear-ranges in the Golgi to yield interchain disulfide bonds linkingVWF subunits into multimers (4). In addition, the D1 and D2domains both contain vicinal cysteines in a CGLCmotif similarto the active site of disulfide isomerases, and insertion of anadditional Gly into either motif inhibits VWF multimerization(5). These findings suggest that the VWF propeptide and pro-tein disulfide isomerases employ similar mechanisms to facili-tate disulfide bond formation.VWF multimer assembly and packaging into Weibel-Pal-

ade bodies are both regulated by pH differences between theER and relatively acidic Golgi. Neutralizing the pH of theGolgi by treating cells with ammonium chloride or chloro-quine blocks VWF multimer assembly (6). In a cell-free sys-tem, VWF multimer assembly does not occur at pH 7.4,which is typical of the ER, but proceeds to some extent at pH6.2 and is optimal at pH 5.8, conditions typical of the trans-Golgi and Weibel-Palade body, respectively (7, 8). In addi-tion, neutralizing the intracellular pH disrupts the tubular

* This work was supported, in whole or in part, by National Institutes of HealthGrant HL72917 (to J. E. S.) and by American Heart Association MidwestAffiliate Postdoctoral Fellowship Award 0825817G (to R. H. H.).

� This article was selected as a Paper of the Week.□S The on-line version of this article (available at http://www.jbc.org) contains

supplemental methods, sequences for Gorilla gorilla and Tursiops trunca-tus, Table S1, and Figs. S1–S5.

The nucleotide sequence(s) reported in this paper has been submitted to theDDBJ/GenBankTM/EBI Data Bank with accession number(s)BK007980 –BK008005.

1 To whom correspondence should be addressed: Washington UniversitySchool of Medicine, 660 S. Euclid, Box 8125, St. Louis, MO 63110. Tel.: 314-362-8802; Fax: 314-454-3012; E-mail: esadler@dom.wustl.edu.

2 The abbreviations used are: VWF, von Willebrand factor; ER, endoplasmicreticulum; BHK, baby hamster kidney.

THE JOURNAL OF BIOLOGICAL CHEMISTRY VOL. 286, NO. 29, pp. 25763–25769, July 22, 2011© 2011 by The American Society for Biochemistry and Molecular Biology, Inc. Printed in the U.S.A.

2

packing of VWF multimers and causes rod-shaped Weibel-Palade bodies to become spherical, which prevents theorderly secretion of VWF filaments in response to secreta-gogues (9). In vitro, the VWF propeptide forms noncovalenthomodimers that bind tightly to multimeric VWF at pH 6.4but dissociate at pH 7.4 (10). Purified VWF propeptide anddimers of N-terminal D�D3 fragments self-assemble intotubules at pH 6.2 and disassemble at pH 7.4 (10). Thus, lowpH is necessary to form intersubunit disulfide bonds duringmultimer assembly, after which low pH is also necessary formultimers to condense into tubules. These findings suggestthat the two processes, multimer assembly and tubular pack-ing, may be regulated by distinct sets of pH-sensing residues.Several proteins employ His residues as pH sensors to reg-

ulate their function upon transport to the Golgi. For exam-ple, His69 in the autoinhibitory propeptide of furin becomesprotonated in the trans-Golgi, after which the propeptidedissociates and furin can cleave other substrates (11). Therecycling mannose lectin ERGIC-53 binds cargo in the ERand releases it in the slightly more acidic ER-Golgi interme-diate compartment (ERGIC), and cargo release requires pro-tonation of conserved His178 (12). A recent His-to-Alamutagenesis study of the ligand-receptor binding interfaceshowed that the evolutionarily conserved residues His180 inhuman prolactin and His188 in the prolactin receptor arecritical for pH-dependent ligand binding at the cell surfaceand release in acidic endosomes (13). We have used phylo-genetic comparisons to identify candidate pH-sensing Hisresidues in VWF and homologous gel-forming mucins,which were functionally evaluated by mutagenesis. Theresults indicate that at least two His residues in the D2domain of the VWF propeptide are critical for pH-depen-dent multimer assembly in the Golgi.

EXPERIMENTAL PROCEDURES

Sequences—Vertebrate VWF sequences were identified byusing human VWF cDNA (14–17), genomic DNA (18, 19), andencoded protein sequences for iterative BLASTN and TBLASTN(National Center for Biotechnology Information (NCBI) version2.2.24) searches of GenBankTM sequences including genomeassemblies, high throughput genome sequences, genome surveysequences,wholegenomeshotgunsequences, expressed sequencetags, nonredundant nucleotide sequences, trace archives, andshort read archives. Exonswere identified based on sequence sim-ilarity using the programs SeqMan, MegAlign, and EditSeq (ver-sion 8.1.5, DNASTAR, Inc., Madison, WI) and assembled intocDNA sequences. Splice site prediction with the program Splice-Port (20) was used to support the identification of exon 2, whichencodes a cleaved signal peptide. Gaps were assigned lengthsbasedon theconserved lengthof thecorrespondingexons inVWFgenes of other species. Sequences encoding the D1D2D�D3 seg-ment of themucinsMUC2,MUC5AC,MUC5B,MUC6,MUC19,and FIMB1 were identified by methods similar to those used forVWF. The sequences, supporting evidence, andmethods for phy-logenetic analyses are described in detail in the supplementalmaterial.All sequenceshavebeenscannedagainst thesedatabases,and accession numbers for sequences with significant relatednessto the deposited sequences are included.

Constructs—Mutations were constructed utilizing a Quik-Change XL kit (Stratagene, La Jolla, CA) with oligonucleotidesfrom Integrated DNATechnologies (Coralville, IA). Mutationswere prepared in pSVHVWF1.1, which encodes full-lengthhuman VWF (21), in plasmid pSVH-propeptide-FLAGNT,which encodes the VWF propeptide (amino acid residues1–763) with an N-terminal FLAG tag between the signal pep-tide and domain D1 (4), or in plasmid pSVH-D3-FLAGNT/cMycCT, which encodes theVWFD1D2D�D3domains (residues1–1241)withanN-terminalFLAGtagandaC-terminal c-Myc tag(4). Plasmid pSVH-KSDR-D3-FLAGNT/cMycCT is similar topSVH-D3-FLAGNT/cMycCT except that amino acid residues760RSKR763weremutated to 760KSDR763,whichprevents cleavageby furin (22). Plasmid pSVH-D�D3�pro-c-Myc encodes theVWFsignal peptide (residues 1–22) followed by domains D�D3 (resi-dues764–1220) andac-Myc tag (4). PlasmidpSVH-�proencodesthe VWF signal peptide (residues 1–22) followed by the matureVWF subunit (residues 764–2813).Cell Culture—BHK-fur4 cells that express human furin (23)

were grown in Dulbecco’s modified Eagle’s medium (Invitro-gen) supplemented with 10% heat-inactivated fetal bovineserum and 2 mM glutamine. BHK-fur4-D1D2D�D3-FLAGNT-c-MycCT cells were described previously (4). BHK-fur4-D1D2-KSDR-D�D3-FLAGNT-c-MycCT cells, stably transfected withpSVH-KSDR-D3-FLAGNT/cMycCT, were prepared similarly.BHK cells were transiently transfected with pSVH-pro-

peptide-FLAGNT and pSVH-D�D3�pro-c-Myc constructs,pSVH-VWF constructs, or pSVH-D3-FLAGNT/cMycCT con-structs, using Lipofectamine Plus reagent (Invitrogen) in Opti-MEM (Invitrogen) according to the manufacturer’s instruc-tions. Conditionedmediumwas collected after 48 h and treatedwith 40 mMN-ethylmaleimide and 144 �M phenylmethylsulfo-nyl fluoride.Protein Analysis—The concentration of VWF was deter-

mined by ELISA (24). Falcon Pro-bind ELISA plates (BD Bio-sciences)were coatedwith a 1:1000 dilution of rabbit polyclonalanti-VWF antibody (082; DAKO, Carpinteria, CA). BoundVWF was detected with a 1:5000 dilution of HRP-conjugatedrabbit polyclonal anti-VWF antibody (P226; DAKO) and tetra-methyl-benzidine (Pierce). Plates were read by a Spectra Max-PLUS microplate reader with the SoftMaxPro 4.7.1 software(Molecular Devices, Silicon Valley, CA).Western Blotting—D�D3 samples were heated to 100 °C for 5

min in Laemmli sample buffer and subjected to SDS-PAGE on4–15% gradient gels (Bio-Rad). VWFmultimer gel electropho-resis was performed as described previously (25).Multimer gelswere incubated in 1.34 mM mercaptoethanol in PBS for 15 minto increase the efficiency of protein transfer by electroblottingonto polyvinylidene (PVDF) membranes (Whatman). Blotswere blocked for 1 h in 50 mM Tris-HCl, pH 7.4, 150 mMNaCl,0.1% Tween 20, and 0.5% casein and then incubated overnightat 4 °C in a 1:1000 dilution ofHRP-conjugated rabbit polyclonalanti-VWF (P226; DAKO) and developed using the ECL Plus kit(GE Healthcare).

RESULTS AND DISCUSSION

VWFStructure—Coding sequences forVWFwere assembledfor 30 vertebrates including 20 placental mammals, a marsu-

pH Sensing and VWF Multimerization

3

pial, two birds, a reptile, an amphibian, and five fish (supple-mental material and supplemental Fig. S1). Exon 2 encodes asignal peptide and was identified or could be proposed with ahigh degree of confidence for all amniotes except dolphinbecause of a gap in the dolphin genome assembly. The proposedsecond exons for frog and fish VWF are relatively uncertainbecause they lack support from cDNA sequences or from sim-ilarity to signal peptides forVWF genes of other species. Exon 1is noncoding and could not be identified for 10 species. Theremaining coding sequences are complete except for portionsof six exons (0.4% of 1530 total exons) encoding 199 amino acidresidues (0.08% of total amino acids).The VWF gene structure varies relatively little among the

species studied. All 52 exons present in human VWF are con-served, and exon length is invariant for 23 exons. Among 1504total exons (excluding exons 1–2), only 113 (0.7%) deviate fromthe length of the corresponding human exon, and in 63instances, the difference is one codon. All splice junctions arestandard except for eight (0.53%) that are predicted to employ aGC splice donor (supplemental Table S1), which is similar tothe 0.56% prevalence of noncanonical GC-AG splice sitesreported for a dataset of more than 22,000 mammalian introns(26). The use of these noncanonical splice sites in anole VWFintron 26 and zebrafish VWF intron 44was verified by compar-ison of genomic DNA and cDNA sequences (supplementalTable S1).Exon 28 exhibits more extreme variation in structure. In tet-

rapods and zebrafish, exon 28 varies in length from 1346 to1436 nucleotides and encodes VWF domains A1 and A2. How-ever, exon 28 is split into exons 28a (304–310 nucleotides) and28b (1066–1162 nucleotides) in the three-spined stickleback,Japanese medaka, spotted green pufferfish, and fugu. Thesefour fish are closely related when compared with zebrafish(supplemental Fig. S1), which suggests that a single exon 28 isancestral and the split exon 28 is a derived character.The amino acid sequence of VWF is also highly conserved.

Human VWF (amino acid residues 23–2813) is at least 78%identical to VWF of other placental mammals, 73% identical toopossum VWF, 55–56% identical to bird, reptile, and amphib-ianVWF, and 45–46% identical to fishVWF (supplemental Fig.S2). Human VWF has 233 Cys residues (not including one Cysin the signal peptide), and all are conserved in other vertebrateswith two kinds of exceptions that involve four Cys residues.VWF of theWestern clawed frog lacks two Cys residues cor-

responding to Cys418 and Cys521 in the D2 domain of humanVWF. The simultaneous absence of these twoCys residues sug-gests that they form a disulfide bond in frog VWF, which issupported by the identification of a Cys898–Cys993 disulfidebond between the corresponding residues in the homologousD3 domain of human VWF (19, 27).VWF of all five fish studied is missing two Cys residues that

correspond to Cys1669 andCys1670 of humanVWF and are con-served in other species. These residues are located at the C-ter-minal end of the VWFA2 domain, where they form an unusualvicinal disulfide bond (27). The A2 domain unfolds in responseto hydrodynamic shear stress to expose a cleavage site forADAMTS13, a regulatory metalloprotease that is specific forVWF.The rigidCys1669–Cys1670 disulfide bond is tightly buried

in a hydrophobic pocket and resists the force-dependentunfolding of the A2 domain (28). The absence of the Cys1669–Cys1670 disulfide bond would be expected to decrease the sta-bility of the A2 domain and decrease the force required to ini-tiate unfolding. Therefore, the lack of this disulfide bond in fishVWF may reflect adaptation to distinct hemostatic require-ments. For example, blood circulates in fish at relatively lowvelocities that are likely to generate low shear forces, whichmayrequire a reduced threshold for shear-induced unfolding ofVWF to allow proteolytic cleavage by ADAMTS13. The rela-tionship between structure and force sensing could be evalu-ated directly by comparing the unfolding of fish and mamma-lian A2 domains with laser tweezers (29).Phylogenetic Analysis of Candidate pH Sensors—Histidine

has a pKa value suitable for detecting the difference in pHbetweenER (pH7.4) and the trans-Golgi (pH6.2) and is likely toperform this function during the assembly and storage ofVWF multimers in Weibel-Palade bodies. Intracellular multi-merization and storage of VWF are conserved among vertebrates,which suggests thatpH-sensingHis residues couldbe identifiedbyphylogenetic comparison of VWF sequences. The D1D2D�D3segment of VWF is sufficient for pH-dependent oligomeriza-tion and tubular storage, and relevant His residues shouldreside in these domains. Of the 32 His residues in theD1D2D�D3 domains of human VWF, 9 are conserved in all 30species studied, and 4 others are conserved in at least 28 of 30species (supplemental Fig. S3 and Fig. 1A).VWF and gel-forming mucins belong to a family of evolu-

tionarily related vertebrate proteins that form large polymers(30). Like VWF, the homologous gel-forming mucins MUC2,MUC5AC, MUC5B, MUC6, MUC19, and FIMB1 have N-ter-minal D1D2D�D3 domains (31). MUC2 (32), MUC5AC (33),andMUC19 (34) have been shown to form interchain disulfidebonds between N-terminal D�D3 domains. As in the case ofVWF, these mucins form dimers in the ER that assemble intodisulfide-linked multimers in the Golgi (31), suggesting a con-served role for mucin D1D2D�D3 domains in the pH-depen-dent assembly of multimers.Alignment of human VWF with representative gel-forming

mucins (supplemental Fig. S4) shows that a subset of His resi-dues conserved in VWF is also conserved in mucins (Fig. 1B).His596 is present in all mucins, and His874 is present in allmucins except FIMB1. His395 and His831 occur in all MUC2,MUC5AC, andMUC5B sequences, but not inMUC6,MUC19,or FIMB1.Conservation across VWF and gel-forming mucins suggests

that some of these His residues might participate in pH-sensi-tive biosynthetic steps that are shared by all of these proteins.However, VWF and mucins have structural differences thatmay reflect significant differences in the mechanism of biosyn-thesis and intracellular storage. For example, the D1D2domains of VWF comprise a propeptide that is separated fromthe D�D3 domains by a conserved furin cleavage site, but thiscleavage site is not present in gel-forming mucins (supplemen-tal Fig. S4). Consequently, the D1D2 domains are likely toremain covalently attached to multimeric mucins, as demon-strated for recombinantMUC19 (35). The pH-dependent bind-ing of the VWF propeptide to the mature VWF subunit is

pH Sensing and VWF Multimerization

4

important for storage inWeibel-Palade bodies, but anyHis res-idues that mediate this noncovalent binding in VWF may beunnecessary for mucins if the D1D2 domains and D�D3domains remain covalently linked.Dependence of D�D3 Dimerization on Intracellular pH—The

formation of intersubunit disulfide bonds betweenD3 domainsof VWF subunits occurs in the Golgi and is blocked by raisingthe intracellular pH with ammonium chloride, chloroquine, ormonensin (6). This pH-dependent assembly process can bestudied conveniently with constructs of VWF that are trun-cated after the D3 domain (4, 36). For example, the expressionof VWF D1D2D�D3 domains in BHK cells results in the secre-tion of D�D3 55-kDamonomers and 110-kDa dimers, as well ascleaved 85-kDa VWF propeptide (domains D1D2) (4), andtreatment of cells expressing this construct with chloroquineprevented D�D3 dimerization (Fig. 2A). Chloroquine concen-trations of 30 �M or greater also inhibited cleavage of VWFpropeptide, probably because furin is inactive at elevated pH(11), which resulted in secretion of monomeric 140-kDaD1D2D�D3 that has the same electrophoretic mobility as D�D3dimers.Similar effects were observed upon treatment of cells with

increasing concentrations of ammonium chloride, which pro-gressively inhibited the dimerization of D�D3. At 100 mM

ammonium chloride, propeptide cleavage was inhibited, andthe predominant secreted product appeared to be monomericD1D2D�D3 (Fig. 2B). This interpretation was confirmed byexpressing a variant of D1D2D�D3 in which the furin recogni-tion site was mutated to KSDR to prevent cleavage. For thisconstruct, dimerization was completely inhibited at 100 mM

ammonium chloride, and the secreted product consistedentirely of monomeric 140-kDa D1D2D�D3 (Fig. 2C). Theseresults show that the formation of intersubunit disulfide bondsby VWF D1D2D�D3 exhibits the expected dependence onGolgi pH. This system was employed to evaluate the role ofconserved His residues on pH-dependent disulfide bondformation.

Effects ofHisMutagenesis onD�D3Dimerization—Thedecreasein pH between ER andGolgi would facilitate the reversible proto-nation of a pH-sensing His residue. To prevent the acquisitionof positive charge at acidic pH, His residues were replaced byAla, and the effect of charge removal onD�D3 dimerizationwasassessed in BHK-fur4 cells transiently transfected with the cor-responding D1D2D�D3 constructs (Fig. 3A). Similar results

FIGURE 1. Conservation of His residues in VWF and gel-forming mucins. Human VWF was aligned with VWF (A) or representative gel-forming mucins MUC2,MUC5AC, MUC5B, MUC6, MUC19, and FIMB1 (B) from the vertebrate species at the left of each graph. The positions of His residues in human VWF and theirlocations with domains D1, D2, D�, and D3 are indicated at the top. A square in the grid is colored when the aligned amino acid is His, as in human VWF, and whitewhen it is not His. The color key indicates the percentage of sequence identity at that position from purple (0%) to red (100%).

FIGURE 2. Effects of chloroquine and ammonium chloride on D�D3dimerization. BHK-fur4-D1D2D�D3-FLAGNT-c-MycCT cells (A and B) or BHK-fur4-D1D2-KSDR-D�D3-FLAGNT-c-MycCT cells (C) were treated with the indi-cated concentration of chloroquine (A) for 48 h or NH4Cl for 72 h (B and C).Conditioned medium was analyzed by SDS-PAGE (4 –15%) and Western blot-ting with HRP-conjugated rabbit anti-human VWF antibody, which preferen-tially detects products containing D�D3 domains. The VWF propeptide,domains D1D2, can be seen as a weakly reactive 95-kDa band in some lanes.The mass in kDa of marker proteins is indicated at the left.

pH Sensing and VWF Multimerization

5

were obtained by expressing D1D2 and D�D3 domains on sep-arate plasmids (supplemental Fig. S5).All constructs were expressed efficiently except H1159A, for

which secretion was severely impaired. Dimerization was pre-vented by the substitutions H395A and H460A, whereasdecreased but detectable dimerization was observed for H95A,H596A, and H737A. The remaining constructs assembleddimers with approximately normal efficiency, indicating thatthe corresponding His residues are not essential for intersub-unit disulfide bond formation.Replacement by Ala at five of eight conserved His residues in

the VWF propeptide impaired D�D3 dimerization; one is indomain D1 (His95), and four are in domain D2 (His395, His460,His596, His737). These results suggest that the VWF propeptidecontains pH-sensing His residues, which is consistent with theobservation that recombinant D1D2 exhibits increased nonco-valent self-association at pH 6.2 when compared with pH 7.4(10). Conversely, none of five evaluable His residues in theD�D3 domains was required to form D�D3 dimers.

D�D3 monomers do not self-associate at low pH (10), whichmight suggest thatD�D3domains lack pH-sensingHis residues.However, low pH does promote the binding of D�D3 mono-mers toD1D2. Low pH also inducesD�D3dimers to bindD1D2domains and form large helical tubules similar to the VWFtubules within Weibel-Palade bodies of endothelial cells (10).Thus, it remains possible that pH-sensing residues on D�D3domains mediate interactions with D1D2 that contribute tointersubunit disulfide bond formation, tubular packing inWeibel-Palade bodies, or both processes.For selected His-to-Ala substitutions that impaired dimeriza-

tion, the His residue was replaced by Arg to prevent the loss ofpositive charge at neutral pH, and the effect of this charge stabili-zation was assessed (Fig. 3A). D�D3 dimerization was not pre-served in constructs H395R, H596R, or H737R. However, H460Rformed dimers approximately as efficiently as the wild typeconstruct.Mutagenesis of His460 was extended to include replacement

with nonpolar amino acids of varying sizes (Ala, Leu, Met),positively charged amino acids (Arg, Lys), and the polar aminoacid Gln. Only Lys and Arg restored D�D3 dimerization, sug-gesting that formation of disulfide bonds between D3 domainsdepends on a positive charge at this position (Fig. 3B).Conserved His and VWF Multimer Assembly—Conserved

His residues within the D1D2D�D3 domains were also changedto Ala in the context of full-length VWF. When expressed inBHK cells, VWF assembles into disulfide-linked multimerswith more than 20 bands visible upon SDS-agarose gel electro-phoresis. Bands correspond to multimers with even numbersof subunits ranging from dimers to larger than 40-mers.Whether VWF is encoded by a single plasmid (cis) or byseparate plasmids for the VWF propeptide and mature sub-unit (trans), multimers assemble with similar efficiency (Fig.4). We took advantage of this property by constructing mostD1D2 mutations in the smaller propeptide construct, whichwas coexpressed with the mature subunit. Mutations locatedin the D�D3 domains were constructed and expressed in asingle full-length VWF plasmid.In most cases, mutations at His residues behaved similarly

whether expressed in truncated D1D2D�D3 (Fig. 3) or in full-length VWF (Fig. 4). Multimers composed of VWF H1159Awere not secreted (data not shown), which is comparable withthe results obtained for the same mutation in the truncatedD1D2D�D3 construct. As expected, His residues that were notrequired for dimerization ofD1D2D�D3 (Fig. 3) were not essen-tial in full-length VWF (Fig. 4). In particular, none of five evalu-able His-to-Ala substitutions in the D�D3 domains had a pro-nounced effect upon multimer assembly: H831A, H874A,H1174A, H1221A, and H1226A.Three His-to-Ala mutations produced somewhat different

results when expressed in truncated or full-length VWF. H95Aand H596A allowed the secretion of trace amounts of D�D3dimer (Fig. 3A) but were compatible with the secretion of arange of VWF multimers, although the size distribution wasskewed toward smaller species (Fig. 4). Conversely, the muta-tion H725A did not reduce the production of D�D3 dimer (Fig.3A) but was associated with the secretion of relatively smallVWF multimers when compared with wild type VWF (Fig. 4).

FIGURE 3. Expression of truncated VWF D1D2D�D3 constructs with Hismutations. BHK-fur4 cells were transiently transfected with wild type (WT) ormutant D1D2D�D3 constructs having the indicated amino acid substitutions(A) or with D1D2D�D3 constructs that encode the indicated amino acids atposition 460 (B). Conditioned medium was analyzed by SDS-PAGE and West-ern blotting with HRP-conjugated anti-VWF. Positions corresponding tomonomeric D�D3 and dimeric (D�D3)2 are indicated at the left.

pH Sensing and VWF Multimerization

6

This variation suggests that sequencesC-terminal to the crit-ical D1D2D�D3 region can influence the efficiency of VWFmultimer assembly. Alternatively, the intermediate and vari-able phenotype caused by the substitution H725A may reflectthe presence of the adjacent His726, which is found in VWFfrom 24 of 30 species sequenced (supplemental Fig. S3). If bothHis725 and His726 contribute to pH-dependent changes in theD2 domain, then loss of onemay cause a partial loss of function.Ala substitutions at His395 and His737 markedly decreased

D�D3 dimer formation (Fig. 3A) and VWF multimer assembly(Fig. 4). In addition, neither H395A nor H395R formed multi-mers when expressed in full-length VWF. H737A and H737Ralso had similar defects in D�D3 dimer formation (Fig. 3A) andmultimer assembly (Fig. 4), although less severe than associatedwith mutations at His395. These results indicate that the fixedpositive charge of a guanidinium group in Arg cannot substi-tute functionally for His395 or His737.Amino acid substitutions at His460 had congruent effects on

D�D3 dimerization (Fig. 3) and VWF multimer assembly (Fig.5). As observed for the truncated D1D2D�D3 construct, muta-tion of His460 to Ala, Leu, Met, or Gln profoundly impairedmultimerization, but replacement by Arg or Lys restored mul-timerization at least to the level of wild type VWF. H460K andH460R constructs formedmultimers with equal efficiency (Fig.5), indicating that intersubunit disulfide bond formationdepends on the presence of a positive charge at this position.Results were consistent with the proposed function of con-

served His residues as pH sensors. The conserved His395 andHis460 located in the propeptide play an essential role in VWFmultimerization.Among the residues examined, replacement of His395 or

His460 byAla (that isH395AorH460A, respectively) caused themost profound defect in intersubunit disulfide bond formation,which suggests that these residues contribute to the pH-depen-dent regulation of VWF multimer assembly. Both residues arehighly conserved in VWF (Fig. 1A), but His460 is not conservedin gel-formingmucins (Fig. 1B). The significance of this distinc-

tion is unknown at present but may be related to the absence ofa furin site between the D1D2 and D�D3 domains of mucins(supplemental Fig. S4); His460 may be required for pH-depen-dent interactions between the D1D2 propeptide and matureVWF subunit after cleavage by furin but dispensable formucins. In addition, VWF is stored intracellularly in denselypacked helical tubules, whereas gel-forming mucins do notappear to undergo a similar process of highly ordered conden-sation in secretory granules. Therefore, the pH-dependenttubular packing of VWF may well depend on His residues thatwould not need to be conserved in mucins.Multimer assembly and tubular packing are critical for the

hemostatic function of VWF. Both events depend on low pH inthe trans-Golgi but can be dissociated by appropriately targetedmutations. For example, the von Willebrand disease mutationY87S impairs both processes, preventing multimer assemblyand disrupting tubular packing; VWFY87S dimers are stored inspherical, disorganizedWeibel-Palade bodies (9, 37). However,

FIGURE 4. Multimer assembly by full-length VWF constructs with His mutations. BHK-fur4 cells were transiently transfected to express wild type VWF froma single plasmid (WT cis) or to express the VWF propeptide and mature subunit on separate plasmids (WT trans) as described under “Experimental Procedures.”Trans indicates mutations at His residues in D1D2 domains that were constructed in the VWF propeptide and cotransfected with the mature VWF subunit. Cisindicates mutations that were constructed in full-length VWF. Conditioned medium was analyzed by SDS-agarose gel electrophoresis and Western blottingwith HRP-conjugated rabbit anti-human VWF antibody.

FIGURE 5. Mutational analysis of His460 and VWF multimer assembly.BHK-fur4 cells were transiently transfected to express full-length VWF withthe indicated amino acid residues at position 460. Conditioned medium wasanalyzed by SDS-agarose gel electrophoresis and Western blotting with HRP-conjugated rabbit anti-human VWF antibody.

pH Sensing and VWF Multimerization

7

insertion of an extra Gly into the CGLC motif of domain D1blocks multimer assembly without affecting storage in rod-shapedWeibel-Palade bodies (5).Whether specific pH-sensingHis residues participate in one or both of these processes will beestablished by additional analyses that discriminate betweenthe noncovalent dimerization of the VWF propeptide (D1D2),binding of propeptide to D�D3 domains, disulfide-mediatedmultimerization, and packing of VWF into tubules (4, 10).

REFERENCES1. Sadler, J. E. (1998) Annu. Rev. Biochem. 67, 395–4242. Wagner, D. D. (1990) Annu. Rev. Cell Biol. 6, 217–2463. Wise, R. J., Pittman, D. D., Handin, R. I., Kaufman, R. J., and Orkin, S. H.

(1988) Cell 52, 229–2364. Purvis, A. R., and Sadler, J. E. (2004) J. Biol. Chem. 279, 49982–499885. Mayadas, T. N., andWagner, D. D. (1992) Proc. Natl. Acad. Sci. U.S.A. 89,

3531–35356. Wagner, D. D., Mayadas, T., and Marder, V. J. (1986) J. Cell Biol. 102,

1320–13247. Mayadas, T.N., andWagner,D.D. (1989) J. Biol. Chem.264, 13497–135038. Erent,M.,Meli, A.,Moisoi, N., Babich, V., Hannah,M. J., Skehel, P., Knipe,

L., Zupancic, G., Ogden, D., and Carter, T. (2007) J. Physiol. 583, 195–2129. Michaux, G., Abbitt, K. B., Collinson, L. M., Haberichter, S. L., Norman,

K. E., and Cutler, D. F. (2006) Dev. Cell 10, 223–23210. Huang, R. H., Wang, Y., Roth, R., Yu, X., Purvis, A. R., Heuser, J. E., Egel-

man, E. H., and Sadler, J. E. (2008) Proc. Natl. Acad. Sci. U.S.A. 105,482–487

11. Feliciangeli, S. F., Thomas, L., Scott, G. K., Subbian, E., Hung, C. H., Mol-loy, S. S., Jean, F., Shinde, U., and Thomas, G. (2006) J. Biol. Chem. 281,16108–16116

12. Appenzeller-Herzog, C., Roche, A. C., Nufer, O., and Hauri, H. P. (2004)J. Biol. Chem. 279, 12943–12950

13. Kulkarni, M. V., Tettamanzi, M. C., Murphy, J. W., Keeler, C., Myszka,D. G., Chayen, N. E., Lolis, E. J., and Hodsdon, M. E. (2010) J. Biol. Chem.285, 38524–38533

14. Sadler, J. E., Shelton-Inloes, B. B., Sorace, J.M.,Harlan, J.M., Titani, K., andDavie, E. W. (1985) Proc. Natl. Acad. Sci. U.S.A. 82, 6394–6398

15. Shelton-Inloes, B. B., Titani, K., and Sadler, J. E. (1986) Biochemistry 25,3164–3171

16. Shelton-Inloes, B. B., Broze, G. J., Jr., Miletich, J. P., and Sadler, J. E. (1987)Biochem. Biophys. Res. Commun. 144, 657–665

17. Bonthron, D., Orr, E. C., Mitsock, L. M., Ginsburg, D., Handin, R. I., and

Orkin, S. H. (1986) Nucleic Acids Res. 14, 7125–712718. Mancuso, D. J., Tuley, E. A., Westfield, L. A., Lester-Mancuso, T. L., Le

Beau, M. M., Sorace, J. M., and Sadler, J. E. (1991) Biochemistry 30,253–269

19. Mancuso, D. J., Tuley, E. A., Westfield, L. A., Worrall, N. K., Shelton-Inloes, B. B., Sorace, J.M., Alevy, Y.G., and Sadler, J. E. (1989) J. Biol. Chem.264, 19514–19527

20. Dogan, R. I., Getoor, L., Wilbur, W. J., and Mount, S. M. (2007) NucleicAcids Res. 35,W285–291

21. Matsushita, T., and Sadler, J. E. (1995) J. Biol. Chem. 270, 13406–1341422. Rehemtulla, A., and Kaufman, R. J. (1992) Blood 79, 2349–235523. Bodo, I., Katsumi, A., Tuley, E. A., Eikenboom, J. C., Dong, Z., and Sadler,

J. E. (2001) Blood 98, 2973–297924. Tuley, E. A., Gaucher, C., Jorieux, S., Worrall, N. K., Sadler, J. E., and

Mazurier, C. (1991) Proc. Natl. Acad. Sci. U.S.A. 88, 6377–638125. Raines, G., Aumann, H., Sykes, S., and Street, A. (1990) Thromb. Res. 60,

201–21226. Burset, M., Seledtsov, I. A., and Solovyev, V. V. (2001) Nucleic Acids Res.

29, 255–25927. Marti, T., Rosselet, S. J., Titani, K., and Walsh, K. A. (1987) Biochemistry

26, 8099–810928. Zhang, Q., Zhou, Y. F., Zhang, C. Z., Zhang, X., Lu, C., and Springer, T. A.

(2009) Proc. Natl. Acad. Sci. U.S.A. 106, 9226–923129. Zhang, X., Halvorsen, K., Zhang, C. Z., Wong, W. P., and Springer, T. A.

(2009) Science 324, 1330–133430. Lang, T., Hansson, G. C., and Samuelsson, T. (2007) Proc. Natl. Acad. Sci.

U.S.A. 104, 16209–1621431. Perez-Vilar, J., and Hill, R. L. (1999) J. Biol. Chem. 274, 31751–3175432. Godl, K., Johansson,M. E., Lidell,M. E.,Morgelin,M., Karlsson, H., Olson,

F. J., Gum, J. R., Jr., Kim, Y. S., andHansson,G. C. (2002) J. Biol. Chem. 277,47248–47256

33. Sheehan, J. K., Kirkham, S., Howard,M.,Woodman, P., Kutay, S., Brazeau,C., Buckley, J., andThornton, D. J. (2004) J. Biol. Chem. 279, 15698–15705

34. Perez-Vilar, J., Eckhardt, A. E., and Hill, R. L. (1996) J. Biol. Chem. 271,9845–9850

35. Perez-Vilar, J., Eckhardt, A. E., DeLuca, A., and Hill, R. L. (1998) J. Biol.Chem. 273, 14442–14449

36. Voorberg, J., Fontijn, R., van Mourik, J. A., and Pannekoek, H. (1990)EMBO J. 9, 797–803

37. Rosenberg, J. B., Haberichter, S. L., Jozwiak, M. A., Vokac, E. A., Kroner,P. A., Fahs, S. A., Kawai, Y., and Montgomery, R. R. (2002) Blood 100,1699–1706

pH Sensing and VWF Multimerization

8

Characterization of Prefibrillar Tau Oligomers in Vitro and inAlzheimer Disease*□S

Received for publication, March 8, 2011, and in revised form, May 4, 2011 Published, JBC Papers in Press, May 6, 2011, DOI 10.1074/jbc.M111.237974

Kristina R. Patterson‡1, Christine Remmers‡, Yifan Fu‡, Sarah Brooker‡, Nicholas M. Kanaan§, Laurel Vana‡,Sarah Ward‡, Juan F. Reyes‡, Keith Philibert¶, Marc J. Glucksman¶, and Lester I. Binder‡

From the ‡Department of Cell and Molecular Biology, Feinberg School of Medicine, Northwestern University, Chicago, Illinois60611, the §Division of Translational Science and Molecular Medicine, Michigan State University, Grand Rapids, Michigan 49503,and the ¶Midwest Proteome Center and Department of Biochemistry and Molecular Biology, Rosalind Franklin University ofMedicine and Science/Chicago Medical School, Chicago, Illinois 60064

Neurofibrillary tangles, composed of insoluble aggregates ofthemicrotubule-associated protein Tau, are a pathological hall-mark of Alzheimer disease (AD) and other tauopathies. How-ever, recent evidence indicates that neuronal dysfunction pre-cedes the formation of these insoluble fibrillar deposits,suggesting that earlier prefibrillar Tau aggregates may be neu-rotoxic. To determine the composition of these aggregates, wehave employed a photochemical cross-linking technique toexamine intermolecular interactions of full-length Tau in vitro.Using thismethod, we demonstrate that dimerization is an earlyevent in the Tau aggregation process and that these dimers self-associate to form larger oligomeric aggregates. Moreover, usingthese stabilized Tau aggregates as immunogens, we generated amonoclonal antibody that selectively recognizesTaudimers andhigher order oligomeric aggregates but shows little reactivity toTau filaments in vitro. Immunostaining indicates that thesedimers/oligomers are markedly elevated in AD, appearing inearly pathological inclusions such as neuropil threads and pre-tangle neurons as well as colocalizing with other early markersof Tau pathogenesis. Taken as a whole, the work presentedherein demonstrates the existence of alternative Tau aggregatesthat precede formation of fibrillar Tau pathologies and raisesthe possibility that these hierarchical oligomeric forms of Taumay contribute to neurodegeneration.

Alzheimer disease (AD)2 is characterized by the extracellularaccumulation of plaques composed of amyloid � (A�) and theintracellular accumulation of the microtubule-associated pro-tein Tau into neurofibrillary tangles (NFTs). Unlike A�plaques, the spatial and temporal progression of NFTs posi-tively correlates with the progression of clinical symptoms (1,

2). Additionally, Tau is necessary for A�-induced neurotoxictyin cell culture and transgenic mouse models (3–5). Tauinclusions are found in other tauopathies that lack A�pathology, including Pick’s disease, corticobasal degenera-tion, and progressive supranuclear palsy (6). Notably, muta-tions in the tau gene cause some forms of frontotemporaldementia (7–10), signifying that Tau dysfunction is suffi-cient to cause neurodegeneration.Under physiological conditions, Tau is a highly soluble

microtubule-associated protein with limited secondary struc-ture (11). In AD and other tauopathies, however, Tau becomeshyperphosphorylated and undergoes conformational shiftsthat lead to its self-association into filamentous and non-fila-mentous aggregates (12). Filamentous Tau is highly ordered,possessing �-pleated structure, as is typical of amyloidogenicproteins (13). Although NFT load correlates with neuronal cellloss and the severity of cognitive impairment in AD, whether ornot these filamentous Tau aggregates are neurotoxic remainscontroversial. Overexpression of both wild type and mutantTau induces neurodegeneration in various animal models (14–19); however, memory deficits and cell loss precede detectableNFT-likeTau pathology (20, 21).Moreover, suppression of Tauexpression improves memory function and halts further cellloss yet NFTs persist, suggesting that neurofibrillary pathologyis not sufficient for neurodegeneration (22, 23). Importantly,neurodegeneration occurs in some animalmodels overexpress-ing Tau despite the absence of overt neurofibrillary pathology(19, 24). Thus, NFTs are not required for and may not be theprimary cause of neurotoxicity and cognitive dysfunction. Infact, levels of early Tau multimeric aggregates that precededNFTs correlated with memory deficits in transgenic mice thatoverexpress Tau (25). Collectively, these studies strongly sug-gest that an intermediate Tau aggregate preceding NFT forma-tion may be responsible for neuronal dysfunction observed inAD and other tauopathies.Although prefibrillar Tau aggregates have been isolated

from AD homogenates, the precise composition of theseaggregates is unclear (26–28). Past attempts to characterizethe earliest stages of Tau multimer formation in vitro werecontingent upon disulfide bridge formation (26, 29). How-ever, the formation of disulfide bridges can inhibit aggrega-tion of Tau isoforms containing four microtubule bindingrepeats (MTBRs) (30–32), and aggregates of the four repeatisoforms are associated with many of the neurodegenerative

* This work was supported, in whole or in part, by National Institutes of HealthGrants T32 AF020506 (to K. R. P.), AG09466 (to L. I. B.), National Center forResearch Resources S10 RR19325 (to M. J. G.), and Health Resources andServices Administration C76 HF03610-01-00 (to M. J. G.).

□S The on-line version of this article (available at http://www.jbc.org) containssupplemental Figs. S1–S3.

1 To whom correspondence should be addressed: 300 E. Superior St., Tarry8-754, Chicago, IL 60611. Tel.: 312-503-0824; Fax: 312-503-7912; E-mail:k-patterson@md.northwestern.edu.

2 The abbreviations used are: AD, Alzheimer disease; NFT, neurofibrillary tan-gle; A�, amyloid �; MTBR, microtubule binding repeat; B4M, benzophe-none-4-maleimide, AA, arachidonic acid; TR, Thiazin Red; TOC1, Tau oligo-meric complex-1; SELDI-TOF, surface enhanced laser desorption andionization time of flight.

THE JOURNAL OF BIOLOGICAL CHEMISTRY VOL. 286, NO. 26, pp. 23063–23076, July 1, 2011© 2011 by The American Society for Biochemistry and Molecular Biology, Inc. Printed in the U.S.A.

9

tauopathies (33). In this study, we describe a cross-linkingtechnique that does not rely upon oxidative conditions.Using this method, we clearly demonstrate that Tau dimersare aggregation intermediates in the formation of higherorder oligomers. Supporting the premise that Tau dimer andoligomer formation are important disease-related events, wegenerated a monoclonal antibody that selectively recognizesTau dimers and higher order oligomeric aggregates. Inter-estingly, these oligomeric aggregates are markedly elevatedin AD. Moreover, immunohistochemical studies with thisantibody indicate that Tau oligomerization precedes the for-mation of mature NFT inclusions. Collectively, our findingsdemonstrate that Tau dimers and higher order oligomersform during the earliest stages of AD pathogenesis and mayrepresent a prefibrillar toxic form of aggregated Tau.

EXPERIMENTAL PROCEDURES

Recombinant Tau Protein Expression and Purification

Tau proteins are numbered according to the largest isoformfound in the central nervous system (hTau40), which consists of441 amino acids and contains both alternatively spliced N-ter-minal exons and four MTBRs. HTau40 was used for all experi-ments except where otherwise indicated. The Tau MTBR con-struct consisting of residues 244–372 was generated bycreating a linear PCR product of this region. An EcoRIrestriction site and methionine residue were added N-termi-nal to residue 244 as well as a poly-His tag and NotI restric-tion site C-terminal to residue 372. The linear PCR productwas inserted into the pET-17b vector (Novagen) via EcoRI/NotI digestion followed by T4 ligation. All other Tau con-structs are cloned into vector pT7c and are described else-where (34–37, 86). Tau constructs were expressed inEscherichia coli and purified using TALON metal affinityresin (Clontech) followed by size exclusion chromatographyas previously described (36, 38).

In Vitro Aggregation

Tau—Aggregation was induced with arachidonic acid (AA)as previously described (39). Briefly, Tau proteins (2–16 �M)were incubated at room temperature or at 37 °C for deletionmutants. Aggregation assays were performed in a solution con-taining 50 mM HEPES, pH 7.6, 50 mM KCl, and 5 mM DTT(unless otherwise indicated) in the presence of 37.5–150 �M

peroxidase-free AA (Cayman Chemical). Working solutions ofAA were prepared in 100% ethanol immediately prior to use.

�-Synuclein—Untagged recombinant full-length human�-synuclein was prepared and aggregated as previouslydescribed (40). Briefly, �-synuclein (4 �M) was incubated at37 °C in 10mMHEPES, pH 7.6, 100mMNaCl, and 5mMDTT inthe presence of 200 �M AA for 24 h.A�—Recombinant human A�42 (rPeptide) HFIP (1,1,1,3,3,3-

hexafluoro-2-propanol)-treated stocks were prepared as previ-ously described (41). A�42 oligomers and filaments were gen-erated according to established protocols (41) with theexception that 4 mMHEPES, pH 8.0, was substituted for Ham’sF-12media. In all cases, aggregationwas confirmed using trans-mission electron microscopy (EM). Samples were fixed with2–10% glutaraldehyde, spotted on formvar/carbon-coated cop-

per grids (Electron Microscopy Sciences), and stained with 2%uranyl acetate.

Benzophenone Cross-linking

HTau40 (15 �M) was reacted with a 10-fold molar excess ofbenzophenone-4-maleimide (B4M) cross-linker (Invitrogen) inthe dark for 4 h in 100 mM Tris, pH 7.4, 0.1 mM EGTA, and 0.5mM Tris(2-carboxyethyl)phosphine hydrochloride. The reac-tion was quenched with 5 mM DTT and excess B4M wasremoved using Nanosep centrifugal filter devices (Pall;MWCO 30 kDa). B4M-labeled Tau (8 �M) was aggregatedovernight in 100 mM Tris, pH 7.4, 0.1 mM EGTA, and 5 mM

DTT in the presence of AA as described above. Cross-linkingwas induced with shortwave UV light (254 nm) for 5 minunless otherwise indicated. For sedimentation analysis, sam-ples were centrifuged in a TLS-55 Beckman rotor over a 40%glycerol cushion at 269,000 � g for 30 min at 25 °C. Thesupernatant was collected and the pellet was resuspended in100 mM Tris, pH 7.4.

Electroelution

Cross-linked hTau40 aggregates were concentrated by sedi-mentation (as described above), diluted in Laemmli buffer,heated in a boiling water bath for 10 min, and separated bySDS-PAGE using 4–8% linear gradient polyacrylamide gels.Gels were stainedwith the E-Zinc Reversible Stain Kit (ThermoScientific) according to the manufacturer’s instructions. Indi-vidual bands were isolated and placed inside D-Tube Dialyzers(MWCO 6–8 kDa, Novagen). Samples were electroeluted for5 h at 150 V in 25 mM Tris, 192 mM glycine, and 0.025% SDS.Electroeluted proteins were concentrated to �50 �l volumewith Pall Nanosep centrifugal concentrators prior to incuba-tion in 400 �l of SDS-away (Protea Biosciences) overnight at�20 °C to precipitate the proteins. Samples were centrifuged at18,000 � g for 10 min at 4 °C, the pellets were resuspended in400 �l of SDS-away, recentrifuged, and finally resuspended in50 mM HEPES, pH 7.6, 50 mM KCl, 0.01% Triton X-100. Sam-ples were boiled for 5 min and centrifuged at 16,000 � g for 1min to remove insoluble proteins.

Mass Spectrometry Analysis

Protein samples were subjected to analysis by surfaceenhanced laser desorption and ionization time of flight (SELDI-TOF)mass spectrometry to examine the distribution of massesof protein components. Samples were spotted in three succes-sive 5-�l aliquots onto individual spots of a ProteinChip NP20Array (Bio-Rad) and allowed to partially air dry for �10 min toreduce the volume between applications. The chip was thenwashed three times with 5 �l of distilled water and air-driedprior to the addition of 2.5 �l of 10 mg/ml sinapinic acid(Bio-Rad) in 60% acetonitrile, 0.1% formic acid. Controlsomitting the air drying for multiple applications of the sam-ple as well as omitting the formic acid with matrix crystalli-zation indicated no difference in the distribution of multim-ers (supplemental Fig. S2). Samples were analyzed using aBio-Rad ProteinChip System 4000 Enterprise mass spec-trometer calibrated against the Bio-Rad All-in-1-ProteinStandard. Each spot was divided into 10 partitions with 210

Tau Oligomers in AD

10

shots/partition and the data were collected at a laser energyof 3000 joule with a mass range between 10 and 160 kDaunless otherwise indicated.

Tau Oligomeric Complex-1 (TOC1) Antibody Generation

Mouse monoclonal antibodies were raised against electro-eluted cross-linked Tau dimers. Female Tau null mice (TheJackson Laboratory; stock number 7251) were immunized sub-cutaneously five times with 2–10 �g of cross-linked Tau dimer.For the final two immunizations, 100 �g of cross-linked Tauoligomers (hTau40 incubated in the presence of AA for 15min,cross-linked, and flash frozen) was administered. Once thedesired serum titer was attained, splenocytes were isolated andfused to SP2/o myeloma cells (42). Two weeks after hybridomaselection in hypoxanthine/aminopterin/thymidine medium,positive clones were chosen based on their ability to bind Tauoligomers but not Tau monomer. One cell line, TOC1, wassubcloned two times to ensure monoclonality and hybridomastability. The clone was adapted to serum-free medium, grownin a CELLine CL 1000 Bioreactor (Sartorius), and the antibodypurified by size exclusion chromatography prior to storage inHEPES-saline buffer, pH 7.4, containing 50% glycerol. Isotyp-ing indicated that TOC1 was of the IgM isotype.

Human Brain Homogenates

Frozen frontal cortex samples from control (Braak I-III) andAD brains (Braak V-VI) were obtained from Rush UniversityMedical Center or the University of Miami Medical Center.Samples were homogenized as previously described (43). Fordenaturing conditions, homogenates were diluted in 2� Laem-mli buffer, heated in boiling water for 5 min, and briefly centri-fuged at 8,000 � g for 3 min to remove cellular debris prior tocollection of the supernatant. For non-denaturing conditions,homogenates were briefly centrifuged at 3,000 � g for 10 min.

Immunoblots

Recombinant Tau and human brain homogenate sampleswere separated by SDS-PAGE on 4–15% linear gradient gelsand transferred to nitrocellulose membranes as described pre-viously (43). For dot blots, samples were spotted directly ontonitrocellulose membranes (44). Both Western and dot blotmembranes were blocked with 5% nonfat dry milk in TBS, pH7.4, and incubated in primary antibodies overnight at 4 °C. Fordot blots, 0.05% Triton X-100 was added to the TBS solution.The Tau12 (45), Tau5 (46), Tau7 (47), MOAB-2,3 and TOC1mouse monoclonal antibodies were used at 0.0067, 0.01, 0.04,0.05, and 0.1 �g/ml, respectively. The mouse monoclonal anti-�-synuclein (Chemicon; MAB5320) and rabbit monoclonalanti-�-actin (Cell Signaling Technology; clone 13E5) antibod-ies were diluted to 1:2000 and 1:1000, respectively. The rabbitpolyclonal antibodies R1 (48) and His-probe (Santa Cruz Bio-technology; clone H-15) were used at 0.0075 and 0.5 �g/ml,respectively. After rinsing, the membranes were incubated inperoxidase-conjugated horse anti-mouse secondary antibody(Vector Labs) or peroxidase-conjugated goat anti-rabbit (Vec-

tor Labs) secondary antibody for 1 h at room temperature.Reactivity was visualized using ECL substrate (Pierce). Dotblots were quantified using ImageJ software (National Insti-tutes of Health) and expressed as the ratio of TOC1:Tau12 orTOC1:His probe intensity.

Immunogold Labeling of Recombinant Tau Aggregates

AggregatedTau sampleswere spotted onto 300-mesh formvar/carbon-coated nickel grids (Electron Microscopy Sciences),blocked with 0.1% gelatin, 5% goat serum in TBS, and incubatedwith the His probe (Santa Cruz; 10 �g/ml) or TOC1 (50 �g/ml)primary antibodies in 5% goat serum/TBS.Gridswere then rinsedwith TBS prior to incubation with 10-nm diameter gold-conju-gated anti-rabbit IgG (Sigma) or anti-mouse IgM (ElectronMicroscopySciences) secondaryantibodiesdiluted1:20 in5%goatserum/TBS. Finally, grids were rinsed with �10 TBS to reducenonspecific labeling, rinsed withH2O, and stained with 2% uranylacetate. Samples were not fixed with glutaraldehyde because thisobscured the TOC1 epitope.4 Optimas 6.0 imaging software(Media Cybernetics) was used to identify and measure oligomers(defined as objects �50 nm in length) and filaments (defined asstructures �50 nm in length) and individual gold particles weretallied. To control for differences in the relative quantities of oli-gomers and filaments present in a givenTau aggregation reaction,results were expressed as the ratio of TOC1:His tag labeling of therespective oligomeric or filamentous structures. Three fields fromeach grid were chosen for quantitation.

Immunohistochemistry

Tissue sections (40 �m) of control (Braak stages I and II) andsevere AD (Braak stages V and VI) cases from the entorhinalcortex (n � 4), hippocampus (n � 3; AD only), and superiortemporal gyrus (n � 2) were obtained from the Cognitive Neu-rology and Alzheimer’s Disease Center at Northwestern Uni-versity. Tissue sections were processed as previously described(49). The TOC1 antibody (0.05 �g/ml) was incubated with tis-sue sections overnight at 4 °C. The tissue was incubated in bio-tinylated goat anti-mouse secondary antibody (Vector; diluted1:500) for 2 h followed by incubation in ABC solution (Vector;according to themanufacturer’s instructions) for 1 h. The stain-ing was developed with 3,3�-diaminobenzidine (Sigma). Sec-tions were mounted onto glass slides, dehydrated throughgraded alcohols, cleared in xylenes, and coverslipped with Per-maslip (Alban Scientific).

Immunofluorescence

Tissue sections (as described above) from the entorhinal cor-tex of control (Braak stages I and II;n� 2) and severeAD (Braakstages V and VI; n � 2) cases were processed for immunofluo-rescence using methods similar to those previously described(50). Briefly, sections were incubated with TOC1 (0.2 �g/ml),the rabbit monoclonal antibody pS422 (Epitomics; diluted1:2500), and/orMN423 (0.5�g/ml) (51) overnight at 4 °C. Afterwashing with PBS containing 0.04% Triton X-100, sectionswere incubated in Alexa Fluor 488 goat anti-mouse IgM �chain-specific (Invitrogen, diluted 1:500), Alexa Fluor 594 goat

3 MOAB-2 is a mouse monoclonal antibody produced in the Binder laboratorythat recognizes A�40 and -42; it is an IgG1. 4 K. R. Patterson and L. I. Binder, unpublished observation.

Tau Oligomers in AD

11

anti-rabbit IgG (Invitrogen, diluted 1:500), and/or Alexa Fluor594 goat anti-mouse IgG2b (Invitrogen, diluted 1:500) second-ary antibodies for 2 h at room temperature. Sectionswere coun-terstained with Thiazin Red (0.004%) as indicated. Tissue sec-tions were rinsed in PBS/Triton X-100, mounted onto glassslides, incubated with Sudan Black (0.05%) to eliminate lipofus-cin autofluorescence, and coverslipped using Vectashield mount-ing medium (Vector). Staining was visualized using a LSM 510Meta (Zeiss) laser scanning confocal microscope. All confocalimages were acquired as z-stacks of single optical sections andanalyzed using Zeiss LSM Image Viewer.

Statistics

SigmaStat software (Systat Software, Inc.) was used for allstatistical tests. Comparisons were made using a t test or one-way analysis of variance followed by Student-Newman-Keulspost hoc analysis as indicated. Data were expressed as mean S.E. and significance was set at p values as noted.

RESULTS

Cross-linking of the Longest Isoform of Tau (hTau40) Revealsan Apparent 180-kDa Oligomer—We employed an establishedphotochemical cross-linking method to examine intermolecu-

lar interactions of full-length Tau in vitro (52) (Fig. 1A). TheB4M cross-linker is distinctive in that one end reacts withnative cysteine residues and the other end is UV-photoactivat-able. Thus, B4M can be used to label the cysteine residues priorto aggregation and then UV light can induce cross-linking ofcarbon-hydrogen bonds within 9 Å after the assembly reactionis complete (52). This substantially differs from previous workin which the cysteine residues are cross-linked and then Tau ispolymerized into filamentous aggregates (26). In addition,using B4M eliminates accessibility issues that may arise whenTau is aggregated prior to cross-linking. B4M cross-linking didnot affect the ability of Tau to aggregate (Fig. 1, B andC), allow-ing for the stabilization of Tau in its aggregation-competentconformation.Treatmentwith an anionic inducing agent such asAA reveals

sharp bands at apparent molecular masses of 180 kDa andhigher when analyzed by SDS-PAGE (Fig. 1D). As monomerichTau40 (actual molecular mass 47 kDa) migrates at 60 kDa onSDS-PAGE, this cross-linked product exhibits the apparentmass of a trimer. Importantly, only a small amount of this spe-cies is present in the absence of inducer.Moreover, backgroundlevels of the 180-kDa species are observed in the absence of

FIGURE 1. Cross-linking Tau aggregates with B4M cross-linkers. A, photochemical cross-linking was performed by incubating soluble Tau with B4M.Full-length Tau (hTau40) possesses two native cysteine (Cys) residues located in the 2nd and 3rd MTBRs. After conjugating the maleimide moiety of B4Mto the native cysteines (Cys*), aggregation of Tau was induced using AA and the resultant Tau aggregates were cross-linked with short-wave UV light.Aggregation-competent Tau is depicted in the Alz-50 conformation in which the N terminus comes into close proximity of the MTBR region (35). Redcircles indicate potential cross-linked sites. B, EM of aggregated Tau with no cross-linker and no UV treatment. Scale bar is 200 nm. C, EM of B4M-conjugated Tau UV irradiated for 5 min is morphologically identical to untreated Tau. D, Western blot of cross-linked Tau filaments and controls.Cross-linking of Tau aggregates reveals an apparent 180-kDa multimer, as well as larger cross-linked products. Note the background levels of the180-kDa multimer in the absence of cross-linking. 500 ng of Tau was loaded per lane and blotted with R1. Results are representative of five independentexperiments.

Tau Oligomers in AD

12

cross-linker indicating that this oligomeric species is somewhatSDS stable. To ensure that our resultswere not an artifact of ourreaction conditions, we varied the UV irradiation time, Tauconcentration, and AA concentration (supplemental Fig. S1).Under all conditions tested, the 180-kDamultimer was the pre-dominant cross-linked species.The 180-kDa Oligomeric Species Is Found in AD but Not

Controls—Frontal cortex homogenates from severe AD (BraakV and VI; n � 4) and control cases with no cognitive impair-ment (Braak I–III; n � 4) were analyzed for the presence ofSDS-stable Tau oligomeric aggregates. Interestingly, we foundthe 180-kDa species in all four AD cases examined but not thecontrol cases (Fig. 2). Given that cleavage of Tau occurs in AD(53–55), we probed the homogenates for intact Tau using anti-

bodies to the extreme N and C termini (Tau12 and Tau7,respectively) and found that the 180-kDa aggregates possessboth termini and likely are not composed of cleavage products.When cross-linked recombinant Tau aggregates were run inparallel, the 180-kDa aggregates from brain and recombinantTau samples appeared to migrate similarly (Fig. 2). These datademonstrate that SDS-stable Tau oligomers are disease specificand that the in vitro cross-linking method stabilizes Tau aggre-gates of comparable size.The 180-kDa Cross-linked Species Is a Dimer—Cross-linked

recombinant Tau aggregates were sedimented to remove resid-ual unaggregated Tau, dissociated with 2% SDS, and separatedby SDS-PAGE. Monomeric Tau (�60 kDa on SDS-PAGE) andthe 180-kDa cross-linked product were purified via electroelu-tion. During purification of the 180-kDa cross-linked spe-cies, it was subjected to harsh denaturing conditions usingSDS to dissociate noncovalent interactions. After purifica-tion, the 180-kDa product was analyzed via SDS-PAGE toverify that it still migrated at the same molecular mass (Fig.3A). Variable amounts of monomeric Tau were co-elutedwith the 180-kDa band, signifying that not all of the apparent180-kDa products are cross-linked. Attempts to purify sig-nificant quantities of larger cross-linked species wereunsuccessful.SELDI-TOF MS of the electroeluted monomer revealed a

prominent peak at 47 kDa as predicted (Fig. 3B). Smaller peaksat 94 and 141 kDa corresponding to dimeric and trimeric Tauspecies were also observed. In contrast, SELDI-TOF analysis ofthe purified apparent 180-kDa species revealed a prominentTau dimer peak at 94 kDa in addition to the 47-kDa peak rep-resenting uncross-linked monomer (Fig. 3C). Minor trimer(141 kDa) and tetramer (188 kDa) peaks were also observed(supplemental Fig. S2). It is improbable that the dimeric Tauidentified here is a breakdown product of the purified apparent180-kDa cross-linked product because, if that were the case, wewould expect to see another band that migrates at a molecularmass between the monomeric and 180-kDa species on SDS-PAGE. Thus, the cross-linked product is a dimer that migrates

FIGURE 2. An apparent 180-kDa Tau oligomer is found in Alzheimer dis-ease but not in controls. Whole homogenates (35 �g/lane) of frontal corti-ces obtained from 4 control and 4 AD brains were analyzed for the multimericTau species by Western blotting using polyclonal Tau antibody R1 (total Tau).Recombinant B4M cross-linked hTau40 (R) was run in parallel. In AD cases,multimeric Tau aggregates migrate at a similar Mr to cross-linked recombi-nant Tau indicating that these aggregated species may be comparable innature. Multimeric Tau aggregates were not observed in control cases. Addi-tionally, the 180-kDa multimeric Tau aggregate was visible in AD cases whenprobed with Tau12 (N terminus) and Tau7 (C terminus), indicating that thisspecies is not an amalgam of cleavage products of Tau monomers. Actin wasused as a loading control.

FIGURE 3. SELDI-TOF MS analysis reveals that the apparent 180-kDa cross-linked product is predominantly a dimer. A, cross-linked hTau40 aggregateswere concentrated and separated by SDS-PAGE. Monomer and oligomer bands were extracted and electroeluted and then run on SDS-PAGE followed bystaining with Coomassie Brilliant Blue R-250 to verify separation. B, SELDI-TOF MS analysis of Tau monomer reveals a peak at 47 kDa, the mass of hTau40.Additionally, minor peaks at 94 and 141 kDa were observed. C, SELDI-TOF MS analysis of the apparent 180-kDa cross-linked product reveals the prominent peakas 94 kDa, which corresponds to a dimer. Noticeably, a large Tau monomer peak is also present, likely representative of uncross-linked SDS-stable dimers thathave since become dissociated. Monomer 2 charge peaks are denoted by arrows.

Tau Oligomers in AD

13

anomalously when analyzed by SDS-PAGE most likely due tothe anisotropy of these stable multimers.Tau Dimers Aggregate to Form Oligomers but Not Long

Filaments—Cross-linked Tau aggregates were sedimented toconfirm that these species incorporate into larger aggregates. Inthe absence of inducer, the small amount of dimer present wasfound only in the supernatant (supplemental Fig. S3). In thepresence of AA, all of the cross-linked species sediment indi-cating that they incorporate into larger Tau aggregates. Thisreveals that dimers are intermediates in at least some form ofTau aggregation as they are observed in the supernatant with-out inducer and the pellet after AA is added.Time course experiments were performed to determine

when dimer formation occurs. Tau dimerization reached peaklevels 15 min after the addition of AA (Fig. 4A). Interestingly,only oligomers and the occasional short filament but no longfilaments were observed at the earliest time points. Filamentlength steadily increased over the next several hours; however,a concomitant increase in dimerization was not observed. Thisindicates that dimerization is an early event that precedes fila-ment formation in vitro.

To further ascertain the role of dimer formation, we aggre-gated purified cross-linked Tau dimers in the presence of AA(Fig. 4B). Similar to the early time points, mostly oligomericaggregates were visible in the EM even 24 h after the addition of

inducer. In contrast, assembly of electroelutedmonomeric Tauresulted in the formation of oligomers and long filaments. Littleaggregation of either monomeric or dimeric Tau was observedin the absence of AA.5 Thus, it appears that Tau dimerization isa prenucleation event as AA is required for aggregation. It ispossible that the oligomers composed of Tau dimers are “off-pathway” aggregates that do not incorporate into filaments;however, in the aggregated dimer preparations, a few short fil-aments were observed in addition to the oligomeric aggregates(Fig. 4B), suggesting that aggregated dimers can form filamentsbut that filament elongation likely favors a differentmechanismof addition to the ends of the Tau polymer.Tau Oligomeric Complex-1 (TOC1) Monoclonal Antibody

Recognizes Tau Oligomers—Previous work demonstrated theexistence of Tau oligomers in AD brain homogenates (25, 56);however, the role of these oligomers in disease remains contro-versial. To confirm that Tau dimers and/or higher order oligo-meric species exist in AD, we raised a novel mouse monoclonalantibody, TOC1, against purified recombinant cross-linkedTau dimers. The selectivity of TOC1 for aggregated Tau overunaggregated Tau was confirmed using dot blots (Fig. 5, A andB). Importantly, when operating at saturating levels of anti-body-immunogen binding, TOC1 showed no reactivity towardunaggregated Tau. In contrast, Tau12 reacted with both unag-gregated and aggregated Tau. Furthermore, TOC1 failed toreact with either �-synuclein or A� in the unaggregated, oligo-meric, or filamentous states (Fig. 5A).Next, we assessed the reactivity of TOC1 against purified

cross-linked dimers. When probed using dot blots, TOC1preferentially labeled cross-linked dimers over electroelutedB4M-labeled monomeric Tau isolated from the same aggre-gation reaction (Fig. 5D). These data indicated that the expo-sure to AA alone did not confer the TOC1 epitope and that aminimum of two Tau molecules was required to form theepitope.To further clarify which aggregated species of Tau was rec-

ognized by TOC1, we used immunogold labeling of Tau aggre-gate preparations. Qualitative analyses of the electron micro-graphs revealed that TOC1 preferentially detected Tauoligomers over filaments; however, occasional labeling of theends of filaments was observed (Fig. 6A). To control for con-centration differences between oligomeric and filamentousaggregates, TOC1 immunogold labeling results were comparedwith immunogold labeling of the poly-His tag in the sameaggregate preparation (Fig. 6C). TOC1 was over seven timesmore likely to label oligomers (structures �50 nm in length)than filaments (structures �50 nm in length). Collectively,these data establish that TOC1 is selective for Tau dimers andhigher order oligomers but does not effectively label Taumono-mers or filaments. Interestingly, TOC1 appears to label only asubpopulation of the oligomeric aggregates suggesting thatmore than one Tau oligomer conformation may exist.TOC1 Recognizes a Conformational Epitope of Tau—The

affinity data verifies that non-phosphorylated recombinantTau possesses the necessary amino acid sequences for TOC1binding but, unlike Tau12, the primary structure is not theonly determinant of the interaction between oligomeric Tauand TOC1. To determine the amino acid sequences of Tau

FIGURE 4. Tau dimers associate to form oligomers but not long filaments.A, Tau was cross-linked at defined time points after the addition of AA. Notethat dimer cross-linking saturates 0.25 h after the onset of aggregation. Cor-responding EMs demonstrate that long filaments do not emerge until afterdimer formation has already saturated. B, EMs of purified dimer (4 �M totalTau equivalent to 2 �M dimeric Tau) and monomer (2 �M) 24 h after theaddition of AA demonstrates that dimer aggregation produces mostly oli-gomers and a few short filaments (arrows). Conversely, electroeluted mono-mer forms long filaments in addition to oligomers when induced to aggre-gate with AA. Protein loading was 500 ng/lane blotted with R1. Scale bars in Aand B are 200 nm.

Tau Oligomers in AD

14

that serve as epitopes of TOC1, a recombinant library con-taining internal deletions of hTau40 was utilized (Fig. 7A).As this antibody requires aggregation of Tau for reactivity,deletion mutants were aggregated with AA prior to screen-ing (Fig. 7B). Our results suggest that TOC1 binding is medi-ated by two segments on Tau. The first region lies within theproline-rich region (Gly155–Gln244) and is absolutely neces-sary for TOC1 binding (Fig. 7C). Deletion of a second region(Leu376–Ser421) contained within the C-terminal portion ofthe protein causes �50% reduction in TOC1 binding. Giventhat these two regions are on either side of the MTBR, theresults are consistent with the formation of an antiparalleldimer (Fig. 7D).The deletion of residues Cys291–Arg349 also results in �50%

reduction of TOC1 signal. Because this deletion lacks a largeportion of the MTBR domain, it is not surprising that it aggre-gates poorly in comparison with the other internal deletionmutants (Fig. 7B). Therefore, the presence of this region isnecessary for Tau aggregation and deletion of this regionlikely obstructs the formation of the TOC1 epitope. As anadditional control, we also tested a construct that possessedonly the MTBRs (Gln244–Glu372), and TOC1 showed no

affinity for this construct. This indicates that this antibodydoes not recognize the portion of the Tau aggregates pos-sessing �-pleated sheets (57) in the absence of the rest of theTau molecule. Nonetheless, we cannot entirely rule out thepossibility that the MTBRs comprise part of the epitope.A potential caveat of these experiments is that deletion of

these amino acid sequences could alter the conformation ofTau in such a way as to obscure or prevent the formation of theTOC1 epitope potentially obfuscating our results. Therefore,future experiments will be necessary to refine this preliminarymapping of the TOC1 epitope.TOC1 Immunoreactivity Is Elevated inAD—Using theTOC1

antibody, we examined control and AD brains for the presenceof Tau oligomeric aggregates. Interestingly, Western blot anal-yses of AD homogenates or in vitro cross-linked Tau usingTOC1 were unsuccessful.5 This indicates that even if Taudimers do not dissociate when treated with SDS, the TOC1conformation is not preserved. Thus, frontal cortex homo-genates of control and severe AD cases were probed usingdot blots under non-denaturing conditions. TOC1 demon-strated a marked increase in selectivity for AD samples overcontrols (Fig. 8,A and B). These results not only demonstrate

FIGURE 5. Characterization of the TOC1 monoclonal Ab. A, dot blot demonstrating TOC1 immunoreactivity. TOC1 preferentially labels uncross-linked Tauoligomeric and filamentous aggregates prepared with AA as opposed to unaggregated Tau (�AA). Moreover, TOC1 does not react with either �-synuclein (�S)or A� in the monomeric, oligomeric, or filamentous states. 45 ng/spot was applied to the nitrocellulose. Total Tau was determined using Tau12. B, quantifica-tion of TOC1 and Tau12 immunoreactivity at varying Tau concentrations. Whereas, Tau12 (gray dashed line) demonstrates a high affinity for both unaggregated(E) and aggregated (F) Tau, TOC1 (black solid line) reacts exclusively with Tau aggregates. Each point represents a minimum of three independent measure-ments. C, EMs of Tau oligomers and filaments (Tau Olig Fil), �-synuclein oligomers and filaments (�S Olig Fil), A� oligomers (A� Olig), and A� filaments (A�Fil) confirm the generation of aggregates of the appropriate morphology for each protein. Scale bar is 200 nm. D, TOC1 preferentially reacts with Tau dimers ondot blots. Monomeric and dimeric Tau were isolated from the same aggregation reaction. Also included is an electroeluted monomeric sample that was neverexposed to AA (�AA). 12 ng/spot was applied to the nitrocellulose.

Tau Oligomers in AD

15

that Tau oligomers are elevated in AD but also validate therelevance of the TOC1 conformation to human disease.To establish the spatial and temporal pattern of Tau oli-

gomer formation in AD, TOC1 immunohistochemistry wasperformed on tissue sections from severe AD (Braak V and VI)and age-matched control cases (Braak I and II). Qualitativeexamination of sections from the entorhinal cortex, hippocam-pus, and superior temporal gyrus indicate that TOC1 reactivityfollows the characteristic pattern of Braak staging (1) (Fig. 8). Incontrol cases, TOC1 revealed neuropil threads as well as dif-fusely labeled neurons displaying both apical and basal den-dritic processes (Fig. 8, C–E). These pretangle neurons are oneof the earliest identifiable indicators of Tau deposition. Asanticipated, a substantial increase in TOC1-positive inclusionswas observed with increasing disease severity. In severe ADcases, TOC1 labels much of the classical Tau AD pathology,including neuropil threads, neuritic plaques, and neuronal pre-tangle and NFT-bearing inclusions (Fig. 8, F–K). These dataindicate that Tau oligomerizaton is an early event in AD patho-genesis and that oligomers persist throughout the duration ofthe disease.TOC1 Immunoreactivity Precedes NFT Formation—To fur-

ther delineate the temporal pattern of Tau oligomer formation,we investigated whether TOC1 positive inclusions correlatewith early or late stage markers of tangle evolution in AD. Thephosphorylation of serine 422 (pS422) on the Tau protein is anearly modification in AD pathogenesis (58, 59). Thus, tissuesections from the entorhinal cortex of controls and severe AD

cases were double-stained with TOC1 and pS422 antibodiesusing double label immunofluorescence. Qualitative observa-tions revealed extensive colocalization of TOC1 and pS422 inboth control and AD samples, and, as pathology progressedinto AD, the amount of both TOC1 and pS422 reactivityincreased (Fig. 9, A and B).

Next, we compared TOC1 immunofluorescence with thatof a late stage NFT marker, MN423. MN423 recognizes Tautruncated at Glu391 (60). Qualitative analysis of severe ADcases revealed that TOC1 and MN423 reactivity are rarelyobserved in the same cell, indicating that the TOC1 epitopedisappears prior to the truncation of Tau at residue 391 (Fig.9C).To establish whether or not TOC1 immunoreactivity is asso-

ciated with mature NFTs, we performed immunofluorescencewith TOC1 and Thiazin Red (TR). TR is a fluorescent dye thatrecognizes �-pleated sheet conformation in the neurofibrillarypathology of AD (61). We used the criteria of TR staining todifferentiate between neurons containing seemingly moremature NFTs and those containing granular diffuse aggre-gates of Tau characteristic of early pretangle neurons. Inter-estingly, in areas possessing abundant TR-positive pathol-ogy, TOC1 predominantly labeled dystrophic neurites andneuropil threads but did not colocalize with TR (Fig. 9D).Occasionally, TOC1 labeled cells containing TR positiveNFT pathology; however, TOC1 was primarily confined tothe periphery of the tangles in these instances. Takentogether, these data indicate that Tau oligomer formationrepresents an early event in AD pathogenesis that precedesneurofibrillary degeneration.

DISCUSSION

NFTs, a pathological hallmark of AD, are composed of Tauaggregates in the form of paired-helical filaments and straightfilaments (62, 63). Although the spatiotemporal distribution ofNFTs correlates with neuron loss and cognitive impairment inAD, current evidence suggests that NFTs may not be the pri-mary form of Tau underlying neuronal dysfunction. Forinstance, the amount of neuronal loss greatly surpasses theamount of tangle formation in AD (64). Furthermore, neuronalloss and cognitive deficits precede neurofibrillary pathology intransgenic mouse models (20, 21). Consequently, it has beenproposed that prefibrillar Tau aggregates may be responsiblefor a large part of disease-related neurotoxicity. To this effect,prefibrillar Tau aggregates correlate with cognitive dysfunctionin a tauopathy mouse model, and similar multimeric Tau spe-cies were found in AD (25). In this study, we identify and char-acterize prefibrillar Tau aggregates that may be associated withdisease-related cognitive decline in AD. We demonstrate thatdimerization is an early event in the Tau aggregation processand that dimers are building blocks for prefibrillar oligomericspecies. Furthermore, using a novel antibody, we provide thefirst glimpse of Tau oligomers within the context of early ADpathology.One obstacle to studying Tau aggregation in vitro is that,

unlike A� and �-synuclein, Tau does not aggregate spontane-ously under physiological conditions (65). Anionic inducermolecules such as free fatty acids (e.g.AA) (66), heparin (67), or

FIGURE 6. TOC1 preferentially labels Tau oligomers. A, TOC1 immuno-gold labeling reveals preferential labeling of oligomeric structures. How-ever, labeling of the ends of a few filaments was observed as well (arrow).B, higher magnification of TOC1 immunogold labeling. C, immunogoldlabeling of the poly-His tag reveals abundant labeling of both Tau oligo-mers and filaments. D, quantification of TOC1 immunogold labeling offilaments (�50 nm) and oligomers (�50 nm) relative to His tag immuno-gold labeling of structures of the same category. *, p � 0.01, paired t test.Scale bars are 100 nm.

Tau Oligomers in AD

16

RNA (68) are required to induce Tau polymerization in vitro;however, the physiological inducer of Tau aggregation in situremains unknown. Using AA combined with a photochemicalcross-linking technique, we stabilized Tau multimers thatapproximate those seen in AD when analyzed by SDS-PAGE.Tau multimers (e.g. dimers, trimers, etc.) are described else-where; however, previous work primarily focused on intermo-lecular associations mediated by disulfide bonds (29, 31, 69).Although three-repeat Tau possesses a single cysteine residue,four-repeat Tau has two cysteine residues favoring intramolec-ular over intermolecular disulfide bonds (30, 31). Furthermore,multiple reports affirm that reduction-resistant Taumultimersare present in AD and frontotemporal dementia (25, 70). Here,we utilized a benzophenone cross-linking technique that doesnot depend upon disulfide bridges. This technique has severaladvantages: (a) site-specific labeling of native cysteine residues,

while Tau is in its soluble, extended conformation; (b) lack ofinterference with the Tau aggregation process; and (c) genera-tion of stable, irreversible cross-linking ofTaumultimers allow-ing for downstream applications such as aggregation and tox-icity studies.Although the existence of Tau multimeric aggregates in AD

brain homogenates has been reported (25, 29, 56), the compo-sition of these aggregates remains controversial. Previous stud-ies suggest that Tau dimers are created when inducers such asheparin (29, 31) or thioflavin S (26) are used, and even in theabsence of inducers (29, 71, 72), whereas others suggest Tauoligomers are composed of trimeric subunits (28). However, allof these previous conclusions were based on molecular weightestimations with SDS-PAGE, gel filtration chromatography,and/or mathematical modeling. Because Tau binds anoma-lously to SDS and exists as a non-globular protein,mass estima-

FIGURE 7. Epitope mapping of TOC1. Deletion mutants of hTau40 were assembled with AA, spotted onto nitrocellulose, and probed with TOC1. A, schematicrepresentations of the deletion mutants utilized. B, EM of wild type (WT), �9 –155, �273–305, �144 –273, �291–349, and �321– 441 hTau40 was used to confirmthe presence of aggregates. The morphological characteristics of the other deletion mutants are described elsewhere (30, 36, 37, 53). Scale bar is 500 nm.C, immunoreactivity of TOC1 to deletion mutants of hTau40 expressed as the ratio of TOC1:total Tau. Total Tau was measured with either Tau12 or a His probeantibody. Results are normalized to WT hTau40. Each bar represents the average of three independent experiments with the exception of MTBR (n � 2). Theproposed discontinuous epitope of TOC1 consists of residues 155–244 (Site 1) and residues 376 – 421 (Site 2) and is represented schematically in panel A. *, p �0.05; **, p � 0.01; ***, p � 0.001, one-way analysis of variance; ND, p value not determined. D, the discontinuous epitope of TOC1 and its preferential bindingto dimeric over monomeric Tau are consistent with an antiparallel dimer conformation. The gray shaded box represents the MTBR, and the proposed TOC1epitope is circled.

Tau Oligomers in AD

17

tion using these techniques can be inaccurate (73). Therefore,we utilized SELDI-TOF mass spectrometry to analyze thecross-linked apparent 180-kDa multimer to provide a moredefinitivemeasurement of themolecularmass of this aggregate.Our data clearly demonstrate that this aggregate is in fact adimer.Wedemonstrate thatdimericaggregatesoccur in theabsenceof

cross-linker and even, to some extent, in the absence of inducer;however, B4M cross-linking obviously stabilizes these aggregates.

In AD, Tau is cross-linked by transglutaminases and products oflipidperoxidation suchashydroxynonenal (productofAAperoxi-dation), and these modifications may even promote Tau aggrega-tion by stabilizing AD-associated Tau conformations such asAlz-50 (74–79). Although it is likely that Tau dimerization occursunder physiological conditions, the processmay become dysregu-lated in disease. Formation of stable cross-linksmay be onemech-anism by which the equilibrium shifts away from soluble, mono-meric Tau toward Tau aggregates.

FIGURE 8. TOC1 immunoreactivity is elevated in AD. A, TOC1 preferentially labels Tau in AD brains compared with that from controls in non-denaturing conditions on dot blots (580 ng/spot) indicating that Tau in the TOC1 conformation is more abundant in AD. B, quantification of TOC1labeling in extracts from the frontal cortex of control and AD brains in A represented as TOC1:Tau12 ratios. *, p � 0.01, unpaired t test. C, as expected,the superior temporal gyrus (STG) does not contain TOC1 immunoreactive Tau pathology in control cases (Braak I and II). D, in the entorhinal cortex (EC)of control cases, TOC1 labels early pretangle neurons with staining extending into both apical and basal dendritic processes. In addition, neuropilthreads are labeled with TOC1. E, higher magnification of TOC1-labeled pretangle neuron in the EC from a control case reveals that these neurons do notcontain mature compact NFTs. F, TOC1 labels abundant pathology in the STG of AD cases (Braak V and VI), including neuropil threads, neuritic plaques,as well as pretangle and tangle-bearing neurons. G, in the EC of AD cases, TOC1 labels similar pathology in addition to neurons that have lost theirdendritic processes indicating that they are further along in the process of tangle evolution. H, higher magnification of TOC1 staining of the EC in AD.I, low magnification TOC1 immunostaining of the hippocampus of an AD case. J, TOC1 immunostaining in the CA1 region of the hippocampus in ADreveals flame-shaped inclusions within pyridimal neurons that are characteristic of this region. K, higher magnification of TOC1 immunolabeling of aneuritic plaque and pyramidal neurons in the CA1 region of an AD case. Scale bars are 50 �m.

Tau Oligomers in AD

18

Weshow that dimerization is an early event in theTau aggre-gation process, which precedes the formation of filaments. Infact, aggregation of purified Tau dimers resulted in the forma-tion of oligomers and a few short filaments in contrast to theaggregation ofmonomericTau,which formed longer filaments.Stable filaments can form by a nucleation-elongation reactionin which monomers first form a stable nucleus (nucleation) towhich additional monomers can add (elongation) (80). Thenucleation process is critical and rate-limiting, as lowering thisenergy barrier induces the production of more filaments.Because Tau dimers did not self-associate in the absence of

inducer, our findings suggest that Tau dimer formation pre-cedes nucleation and that some other conformational shiftmust occur prior to oligomer and filament formation. A poten-tial caveat is that the cross-linked Tau dimers were denaturedduring the purification process, and, thus, it is possible that theproteins did not re-nature completely or correctly. Moreover,our results indicate that dimeric Tau is not favored for elonga-tion, at least not in the cross-linked form. It is possible thatdimers that are not cross-linked have the freedom to rearrangeinto a conformation such that elongation could occur. Furtherstudies are required to elucidate the events necessary for Tau

FIGURE 9. TOC1 immunoreactivity colocalizes with early markers of Tau pathology. A and B, laser scanning confocal microscopy was used to determine thedegree of colocalization between TOC1 (green) and pS422 (red) in the entorhinal cortex. Colocalization between the two antibodies was almost complete inboth control (A) and AD (B) cases, indicating that formation of the TOC1 epitope correlates with phosphorylation of Ser422, an early event in AD pathogenesis.C, immunofluorescence was performed using TOC1 (green) and MN423 (red) in severe AD cases. Very little colocalization was observed. D, in severe AD cases,TOC1 and TR did not colocalize, indicating that the TOC1 epitope precedes �-sheet formation characteristic of NFTs. Although occasional inclusions with bothTOC1 and TR exist (arrow), different portions of the cell are labeled by each. Scale bar is 50 �m.

Tau Oligomers in AD

19

nucleation, and to determine whether dimers lead directly tooligomers and subsequent filament formation or whether theseoligomers represent off-pathway aggregates that do not formfilaments.To validate the relevance of these findings to AD, we gen-

erated a novel monoclonal antibody that selectively labelsTau dimers and higher order oligomers. Interestingly, TOC1immunoreactivity was greatly elevated in AD brains overcontrols. TOC1 avidly labeled the hallmark Tau pathology inAD including neuropil threads, neuritic plaques, and neuro-nal inclusions. Moreover, the presence of TOC1 immunore-active inclusions in control cases (Braak stages I and II) sug-gests that dimer/oligomer formation is an early event indisease pathogenesis. Supporting this assertion, immunoflu-orescence in human tissue sections indicated that oligomer-ization closely associates with Ser422 phosphorylation, anearly pathological event in AD (58). Moreover, colocaliza-tion with TR, a marker for fibrillar forms of both Tau and A�,and colocalization with MN423, a late NFT marker (60), wasscarcely observed. These findings confirm that TOC1 doesindeed label prefibrillar pathology. The presence of TOC1positive prefibrillar inclusions in AD supports our in vitroTau aggregation and cross-linking results suggesting thatTau dimers/oligomers may be intermediates in the aggrega-tion process that precede NFT formation in situ.AD is characterized by the presence of both A� and Tau

inclusions; however, Tau is necessary for toxicity of A� in bothcultured cells and transgenicmice (3, 4). Therefore, elucidatingthe mechanism by which Tau converts from a soluble, micro-tubule-bound protein into stable, toxic aggregates appears keyto understanding this disease. Nonetheless, the link betweenA� and Tau aggregation remains enigmatic. Several studieshave established that A� oligomers increase phospholipase A2activity, which in turn elevates AA resulting in neuronal toxic-ity (81–83). Increased activity of cytoplasmic phospholipase A2has been confirmed in AD, generating increased levels of freefatty acids including AA (84). Interestingly, reduction of phos-pholipase A2 ameliorates cognitive deficits in an amyloid pre-cursor protein mouse model of AD (83). In the same model,reduction of endogenousTau also prevented behavioral deficits(4). These findings suggest that AA and Tau may work syner-gistically. Here we provide evidence that AA-stimulated Tauaggregation is relevant to AD, suggesting that the liberation offree fatty acids including AA is a potential link between A� andTau pathology.The discontinuous epitopes of antibodies provide evidence

for conformational changes in Tau during the disease process.For instance, epitope mapping of the Alz-50 antibody revealsthat the N terminus comes into close proximity to the MTBRearly in tangle evolution (35, 85). TOC1 is also a conformation-selective antibody that appears to recognize portions of the pro-line-rich region and C terminus of Tau. It is possible that thisepitope may occur from intramolecular folding events; how-ever, because TOC1 does not recognize monomeric Tau, it ismore likely that this is an intermolecular event consistent withthe formation of anti-parallel dimers given that the twoepitopes are on either side of the MTBRs (see Fig. 7). Thisexpands upon a previous study showing that three-repeat Tau

can dimerize in an antiparallelmanner (72). Our results suggestthat both three- and four-repeat Tau form antiparallel dimers;however, TOC1 labeled only a portion of the Tau aggregates,suggesting that there may be other aggregation-competentconformations to discover. Future studies arewarranted to fullychart the evolution of the conformational shifts of Tau and howthey relate to TOC1 during disease pathogenesis. Even so, ourresults confer significant insights concerning the earliest stagesof Tau aggregation and provide an important foundation forthe future study of potentially neurotoxic Tau oligomers in dis-ease pathogenesis.

Acknowledgments—We thank Dr. Sarah E. Rice, Dr. Kristin Deitrich,and IsabellaUgwu for technical assistance.We also thankDr.NicholeLaPointe for the �-synuclein preparation.

REFERENCES1. Braak, H., and Braak, E. (1991) Acta Neuropathol. 82, 239–2592. Arriagada, P. V., Growdon, J. H., Hedley-Whyte, E. T., and Hyman, B. T.

(1992) Neurology 42, 631–6393. Rapoport, M., Dawson, H. N., Binder, L. I., Vitek, M. P., and Ferreira, A.

(2002) Proc. Natl. Acad. Sci. U.S.A. 99, 6364–63694. Roberson, E. D., Scearce-Levie, K., Palop, J. J., Yan, F., Cheng, I. H., Wu,

T., Gerstein, H., Yu, G. Q., and Mucke, L. (2007) Science 316, 750–7545. Vossel, K. A., Zhang, K., Brodbeck, J., Daub, A. C., Sharma, P., Finkbeiner,

S., Cui, B., and Mucke, L. (2010) Science 330, 1986. Feany, M. B., and Dickson, D. W. (1996) Ann. Neurol. 40, 139–1487. Poorkaj, P., Bird, T. D., Wijsman, E., Nemens, E., Garruto, R. M., Ander-

son, L., Andreadis, A., Wiederholt, W. C., Raskind, M., and Schellenberg,G. D. (1998) Ann. Neurol. 43, 815–825

8. Poorkaj, P., Grossman,M., Steinbart, E., Payami, H., Sadovnick, A., Noch-lin, D., Tabira, T., Trojanowski, J. Q., Borson, S., Galasko, D., Reich, S.,Quinn, B., Schellenberg, G., and Bird, T. D. (2001) Arch. Neurol. 58,383–387

9. Spillantini, M. G., Murrell, J. R., Goedert, M., Farlow, M. R., Klug, A.,and Ghetti, B. (1998) Proc. Natl. Acad. Sci. U.S.A. 95, 7737–7741

10. Hutton, M., Lendon, C. L., Rizzu, P., Baker, M., Froelich, S., Houlden, H.,Pickering-Brown, S., Chakraverty, S., Isaacs, A., Grover, A., Hackett, J.,Adamson, J., Lincoln, S., Dickson, D., Davies, P., Petersen, R. C., Stevens,M., de Graaff, E., Wauters, E., van Baren, J., Hillebrand, M., Joosse, M.,Kwon, J. M., Nowotny, P., Che, L. K., Norton, J., Morris, J. C., Reed, L. A.,Trojanowski, J., Basun, H., Lannfelt, L., Neystat, M., Fahn, S., Dark, F.,Tannenberg, T., Dodd, P. R., Hayward, N., Kwok, J. B., Schofield, P. R.,Andreadis, A., Snowden, J., Craufurd, D., Neary, D., Owen, F., Oostra,B. A., Hardy, J., Goate, A., van Swieten, J., Mann, D., Lynch, T., and Heu-tink, P. (1998) Nature 393, 702–705

11. Schweers, O., Schonbrunn-Hanebeck, E., Marx, A., and Mandelkow, E.(1994) J. Biol. Chem. 269, 24290–24297

12. Iqbal, K., Liu, F., Gong, C.X., AlonsoAdel, C., andGrundke-Iqbal, I. (2009)Acta Neuropathol. 118, 53–69

13. von Bergen, M., Friedhoff, P., Biernat, J., Heberle, J., Mandelkow, E. M.,and Mandelkow, E. (2000) Proc. Natl. Acad. Sci. U.S.A. 97, 5129–5134

14. Lewis, J., McGowan, E., Rockwood, J., Melrose, H., Nacharaju, P., VanSlegtenhorst, M., Gwinn-Hardy, K., Paul Murphy, M., Baker, M., Yu, X.,Duff, K., Hardy, J., Corral, A., Lin,W. L., Yen, S. H., Dickson, D.W., Davies,P., and Hutton, M. (2000) Nat. Genet. 25, 402–405

15. Gotz, J., Chen, F., Barmettler, R., and Nitsch, R. M. (2001) J. Biol. Chem.276, 529–534

16. Tanemura, K., Akagi, T., Murayama, M., Kikuchi, N., Murayama, O., Ha-shikawa, T., Yoshiike, Y., Park, J. M., Matsuda, K., Nakao, S., Sun, X., Sato,S., Yamaguchi, H., andTakashima, A. (2001)Neurobiol. Dis. 8, 1036–1045

17. Tatebayashi, Y.,Miyasaka, T., Chui, D. H., Akagi, T.,Mishima, K., Iwasaki,K., Fujiwara, M., Tanemura, K., Murayama, M., Ishiguro, K., Planel, E.,Sato, S., Hashikawa, T., and Takashima, A. (2002) Proc. Natl. Acad. Sci.

Tau Oligomers in AD

20

U.S.A. 99, 13896–1390118. Allen, B., Ingram, E., Takao,M., Smith,M. J., Jakes, R., Virdee, K., Yoshida,

H., Holzer, M., Craxton, M., Emson, P. C., Atzori, C., Migheli, A.,Crowther, R. A., Ghetti, B., Spillantini, M. G., and Goedert, M. (2002)J. Neurosci. 22, 9340–9351

19. Andorfer, C., Acker, C. M., Kress, Y., Hof, P. R., Duff, K., and Davies, P.(2005) J. Neurosci. 25, 5446–5454

20. Oddo, S., Caccamo,A., Shepherd, J. D.,Murphy,M. P., Golde, T. E., Kayed,R., Metherate, R., Mattson, M. P., Akbari, Y., and LaFerla, F. M. (2003)Neuron 39, 409–421

21. Spires, T. L., Orne, J. D., SantaCruz, K., Pitstick, R., Carlson, G. A.,Ashe, K. H., and Hyman, B. T. (2006) Am. J. Pathol. 168, 1598–1607

22. Santacruz, K., Lewis, J., Spires, T., Paulson, J., Kotilinek, L., Ingelsson,M., Guimaraes, A., DeTure, M., Ramsden, M., McGowan, E., Forster,C., Yue, M., Orne, J., Janus, C., Mariash, A., Kuskowski, M., Hyman, B.,Hutton, M., and Ashe, K. H. (2005) Science 309, 476–481

23. Sydow, A., Van der Jeugd, A., Zheng, F., Ahmed, T., Balschun, D.,Petrova, O., Drexler, D., Zhou, L., Rune, G., Mandelkow, E., D’Hooge,R., Alzheimer, C., and Mandelkow, E. M. (2011) J. Neurosci. 31,2511–2525

24. Wittmann, C. W., Wszolek, M. F., Shulman, J. M., Salvaterra, P. M.,Lewis, J., Hutton, M., and Feany, M. B. (2001) Science 293, 711–714

25. Berger, Z., Roder, H., Hanna, A., Carlson, A., Rangachari, V., Yue, M.,Wszolek, Z., Ashe, K., Knight, J., Dickson, D., Andorfer, C., Rosenberry,T. L., Lewis, J., Hutton,M., and Janus, C. (2007) J. Neurosci. 27, 3650–3662

26. Friedhoff, P., von Bergen, M., Mandelkow, E. M., Davies, P., and Man-delkow, E. (1998) Proc. Natl. Acad. Sci. U.S.A. 95, 15712–15717

27. Congdon, E. E., Kim, S., Bonchak, J., Songrug, T., Matzavinos, A., andKuret, J. (2008) J. Biol. Chem. 283, 13806–13816

28. Lasagna-Reeves, C. A., Castillo-Carranza, D. L., Guerrero-Muoz, M. J.,Jackson, G. R., and Kayed, R. (2010) Biochemistry 49, 10039–10041

29. Sahara, N., Maeda, S., Murayama, M., Suzuki, T., Dohmae, N., Yen, S. H.,and Takashima, A. (2007) Eur. J. Neurosci. 25, 3020–3029

30. Barghorn, S., and Mandelkow, E. (2002) Biochemistry 41, 14885–1489631. Schweers, O., Mandelkow, E. M., Biernat, J., and Mandelkow, E. (1995)

Proc. Natl. Acad. Sci. U.S.A. 92, 8463–846732. Gamblin, T. C., King, M. E., Kuret, J., Berry, R. W., and Binder, L. I. (2000)

Biochemistry 39, 14203–1421033. Lee, V. M., Goedert, M., and Trojanowski, J. Q. (2001) Annu. Rev. Neuro-

sci. 24, 1121–115934. Goedert, M., Spillantini, M. G., Potier, M. C., Ulrich, J., and Crowther,

R. A. (1989) EMBO. J. 8, 393–39935. Carmel, G., Mager, E. M., Binder, L. I., and Kuret, J. (1996) J. Biol. Chem.

271, 32789–3279536. Abraha, A., Ghoshal, N., Gamblin, T. C., Cryns, V., Berry, R. W., Kuret, J.,

and Binder, L. I. (2000) J. Cell Sci. 113, 3737–374537. Gamblin, T. C., Berry, R. W., and Binder, L. I. (2003) Biochemistry 42,

2252–225738. Carmel, G., Leichus, B., Cheng, X., Patterson, S. D., Mirza, U., Chait, B. T.,

Kuret, J. (1994) J. Biol. Chem. 269, 7304–730939. Gamblin, T. C., King, M. E., Dawson, H., Vitek, M. P., Kuret, J., Berry,

R. W., and Binder, L. I. (2000) Biochemistry 39, 6136–614440. Necula, M., Chirita, C. N., and Kuret, J. (2003) J. Biol. Chem. 278,

46674–4668041. Stine, W. B., Jungbauer, L., Yu, C., and LaDu, M. J. (2011) Methods Mol.

Biol. 670, 13–3242. Binder, L. I., Frankfurter, A., and Rebhun, L. I. (1985) J. Cell Biol. 101,

1371–137843. Reyes, J. F., Reynolds, M. R., Horowitz, P. M., Fu, Y., Guillozet-Bon-

gaarts, A. L., Berry, R., and Binder, L. I. (2008) Neurobiol. Dis. 31,198–208

44. LaPointe, N. E., Morfini, G., Pigino, G., Gaisina, I. N., Kozikowski, A. P.,Binder, L. I., and Brady, S. T. (2009) J. Neurosci. Res. 87, 440–451

45. Horowitz, P. M., Patterson, K. R., Guillozet-Bongaarts, A. L., Reynolds,M. R., Carroll, C. A., Weintraub, S. T., Bennett, D. A., Cryns, V. L., Berry,R. W., and Binder, L. I. (2004) J. Neurosci. 24, 7895–7902

46. LoPresti, P., Szuchet, S., Papasozomenos, S. C., Zinkowski, R. P., andBinder, L. I. (1995) Proc. Natl. Acad. Sci. U.S.A. 92, 10369–10373

47. Horowitz, P.M., LaPointe,N., Guillozet-Bongaarts, A. L., Berry, R.W., andBinder, L. I. (2006) Biochemistry 45, 12859–12866

48. Berry, R.W., Sweet, A. P., Clark, F. A., Lagalwar, S., Lapin, B. R., Wang, T.,Topgi, S., Guillozet-Bongaarts, A. L., Cochran, E. J., Bigio, E. H., andBinder, L. I. (2004) J. Neurocytol. 33, 287–295

49. Kanaan, N. M., Kordower, J. H., and Collier, T. J. (2007) J. Comp. Neurol.502, 683–700

50. Garcia-Sierra, F., Ghoshal, N., Quinn, B., Berry, R. W., and Binder, L. I.(2003) J. Alzheimers Dis. 5, 65–77

51. Novak, M., Wischik, C. M., Edwards, P., Pannell, R., and Milstein, C.(1989) Prog. Clin. Biol. Res. 317, 755–761

52. Dorman, G., and Prestwich, G. D. (1994) Biochemistry 33, 5661–567353. Gamblin, T. C., Chen, F., Zambrano, A., Abraha, A., Lagalwar, S., Guil-

lozet, A. L., Lu, M., Fu, Y., Garcia-Sierra, F., LaPointe, N., Miller, R., Berry,R. W., Binder, L. I., and Cryns, V. L. (2003) Proc. Natl. Acad. Sci. U.S.A.100, 10032–10037

54. Novak, M., Kabat, J., and Wischik, C. M. (1993) EMBO J. 12, 365–37055. Guo, H., Albrecht, S., Bourdeau,M., Petzke, T., Bergeron, C., and LeBlanc,

A. C. (2004) Am. J. Pathol. 165, 523–53156. Maeda, S., Sahara, N., Saito, Y., Murayama, S., Ikai, A., and Takashima, A.

(2006) Neurosci. Res. 54, 197–20157. Barghorn, S., Davies, P., and Mandelkow, E. (2004) Biochemistry 43,

1694–170358. Guillozet-Bongaarts, A. L., Cahill, M. E., Cryns, V. L., Reynolds, M. R.,

Berry, R. W., and Binder, L. I. (2006) J. Neurochem. 97, 1005–101459. Mondragon-Rodríguez, S., Basurto-Islas, G., Santa-Maria, I., Mena, R.,

Binder, L. I., Avila, J., Smith, M. A., Perry, G., and García-Sierra, F. (2008)Int. J. Exp. Pathol. 89, 81–90

60. Guillozet-Bongaarts, A. L., Garcia-Sierra, F., Reynolds, M. R., Horowitz,P. M., Fu, Y.,Wang, T., Cahill, M. E., Bigio, E. H., Berry, R.W., and Binder,L. I. (2005) Neurobiol. Aging 26, 1015–1022

61. Luna-Munoz, J., Peralta-Ramirez, J., Chavez-Macías, L., Harrington, C. R.,Wischik, C. M., and Mena, R. (2008) Acta Neuropathol. 116, 507–515

62. Kidd, M. (1963) Nature 197, 192–19363. Kosik, K. S., Joachim, C. L., and Selkoe, D. J. (1986) Proc. Natl. Acad. Sci.

U.S.A. 83, 4044–404864. Gomez-Isla, T., Hollister, R., West, H., Mui, S., Growdon, J. H., Pe-

tersen, R. C., Parisi, J. E., and Hyman, B. T. (1997) Ann. Neurol. 41,17–24

65. Chirita, C. N., Congdon, E. E., Yin, H., and Kuret, J. (2005) Biochemistry44, 5862–5872

66. Wilson, D. M., and Binder, L. I. (1995) J. Biol. Chem. 270, 24306–2431467. Perez, M., Valpuesta, J. M., Medina,M., Montejo de Garcini, E., and Avila,

J. (1996) J. Neurochem. 67, 1183–119068. Kampers, T., Friedhoff, P., Biernat, J., Mandelkow, E. M., andMandelkow,

E. (1996) FEBS. Lett. 399, 344–34969. Guttmann, R. P., Erickson, A. C., and Johnson, G. V. (1995) J. Neurochem.

64, 1209–121570. Maeda, S., Sahara, N., Saito, Y., Murayama, M., Yoshiike, Y., Kim, H.,

Miyasaka, T., Murayama, S., Ikai, A., and Takashima, A. (2007) Biochem-istry 46, 3856–3861

71. Rosenberg, K. J., Ross, J. L., Feinstein, H. E., Feinstein, S. C., and Israel-achvili, J. (2008) Proc. Natl. Acad. Sci. U.S.A. 105, 7445–7450

72. Wille, H., Drewes, G., Biernat, J., Mandelkow, E. M., and Mandelkow, E.(1992) J. Cell Biol. 118, 573–584

73. Cabre, F., Canela, E. I., and Canela, M. A. (1989) J. Chromatogr. 472,347–356

74. Sayre, L. M., Zelasko, D. A., Harris, P. L., Perry, G., Salomon, R. G., andSmith, M. A. (1997) J. Neurochem. 68, 2092–2097

75. Liu,Q., Smith,M.A., Avila, J., DeBernardis, J., Kansal,M., Takeda, A., Zhu,X., Nunomura, A., Honda, K., Moreira, P. I., Oliveira, C. R., Santos, M. S.,Shimohama, S., Aliev, G., de la Torre, J., Ghanbari, H. A., Siedlak, S. L.,Harris, P. L., Sayre, L. M., and Perry, G. (2005) Free Radic. Biol. Med. 38,746–754

76. Appelt, D. M., and Balin, B. J. (1997) Brain Res. 745, 21–3177. Balin, B. J., and Appelt, D. M. (2000) Methods Mol. Med. 32, 395–40478. Dudek, S. M., and Johnson, G. V. (1993) J. Neurochem. 61, 1159–116279. Singer, S. M., Zainelli, G. M., Norlund, M. A., Lee, J. M., andMuma, N. A.

Tau Oligomers in AD

21

(2002) Neurochem. Int. 40, 17–3080. Oosawa, F., andAsakura, S. (1975)Thermodynamics of the Polymerization

of Protein, Academic Press, New York81. Kriem, B., Sponne, I., Fifre, A., Malaplate-Armand, C., Lozac’h-Pillot, K.,

Koziel, V., Yen-Potin, F. T., Bihain, B., Oster, T., Olivier, J. L., and Pillot, T.(2005) FASEB J. 19, 85–87

82. Malaplate-Armand, C., Florent-Bechard, S., Youssef, I., Koziel, V.,Sponne, I., Kriem, B., Leininger-Muller, B., Olivier, J. L., Oster, T., andPillot, T. (2006) Neurobiol Dis. 23, 178–189

83. Sanchez-Mejia, R. O., Newman, J. W., Toh, S., Yu, G. Q., Zhou, Y.,Halabisky, B., Cisse, M., Scearce-Levie, K., Cheng, I. H., Gan, L., Palop,J. J., Bonventre, J. V., and Mucke, L. (2008) Nat. Neurosci. 11,1311–1318

84. Farooqui, A. A., and Horrocks, L. A. (2006) Neuroscientist 12, 245–26085. Jicha, G. A., Bowser, R., Kazam, I. G., and Davies, P. (1997) J. Neurosci. Res.

48, 128–13286. Kanaan, N.M.,Morfini, G., Lapointe, N. E., Pigino, G., Patterson, K. R., Song,

Y., Andreadis, A., Brady, S. T., and Binder, L. I. (2011) J. Neurosci., in press

Tau Oligomers in AD

22

Mechanism of Intracellular cAMP Sensor Epac2 ActivationcAMP-INDUCED CONFORMATIONAL CHANGES IDENTIFIED BY AMIDEHYDROGEN/DEUTERIUM EXCHANGE MASS SPECTROMETRY (DXMS)*�

Received for publication, January 25, 2011, and in revised form, March 10, 2011 Published, JBC Papers in Press, March 17, 2011, DOI 10.1074/jbc.M111.224535

Sheng Li‡1, Tamara Tsalkova§¶1, Mark A. White¶, Fang C. Mei§¶, Tong Liu‡, Daphne Wang‡, Virgil L. Woods, Jr.‡2,and Xiaodong Cheng§¶3

From the §Department of Pharmacology and Toxicology and ¶Sealy Center for Structural Biology and Molecular Biophysics, TheUniversity of Texas Medical Branch, Galveston, Texas 77555-0615 and the ‡Department of Medicine and Biomedical SciencesGraduate program, University of California, San Diego, La Jolla, California 92093-0656

Epac2, a guanine nucleotide exchange factor, regulates a widevariety of intracellular processes in response to second messen-ger cAMP. In this study, we have used peptide amide hydrogen/deuterium exchange mass spectrometry to probe the solutionstructural and conformational dynamics of full-length Epac2 inthe presence and absence of cAMP. The results support amech-anism in which cAMP-induced Epac2 activation is mediated bya major hinge motion centered on the C terminus of the secondcAMP binding domain. This conformational change realignsthe regulatory components of Epac2 away from the catalyticcore, making the later available for effector binding. Further-more, the interface between the first and second cAMP bindingdomains is highly dynamic, providing an explanation of howcAMP gains access to the ligand binding sites that, in the crystalstructure, are seen to be mutually occluded by the other cAMPbinding domain. Moreover, cAMP also induces conformationalchanges at the ionic latch/hairpin structure, which is directlyinvolved inRAP1 binding. These results suggest that in additionto relieving the steric hindrance imposed upon the catalytic lobeby the regulatory lobe, cAMPmay also be an allosteric modula-tor directly affecting the interaction between Epac2 and RAP1.Finally, cAMP binding also induces significant conformationalchanges in the dishevelled/Egl/pleckstrin (DEP) domain, a con-served structural motif that, although missing from the activeEpac2 crystal structure, is important for Epac subcellular target-ing and in vivo functions.

Exchange proteins directly activated by cAMP (Epacs)4 are afamily of novel guanine nucleotide exchange factors (GEFs)

specific for downstream effectors, RAP1 and RAP2 (1, 2). Inmammals, there are two major isoforms of Epac, Epac1 andEpac2, encoded by two independent genes with distinct tissuedistributions. At the cellular level, Epac1 and Epac2 are knownto reside in various discrete subcellular compartments, wherethey mediate a wide range of cellular functions including celladhesion, exocytosis, secretion, differentiation, proliferation,and apoptosis (3, 4). Accumulating evidence also suggests thatthe abilities of Epac to participate in diverse multiprotein com-plexes may dictate their subcellular destinations and in vivofunctions (5–10).At the molecular level, Epac1 and Epac2 are multidomain

proteins with extensive sequence homology. Epac1 and Epac2share aC-terminal catalytic corewith an identical domain orga-nization that consists of a RAS exchange (REM) domain, a RASassociation (RA) domain, and aCDC25homologyGEFdomain.The activity of both Epacs is modulated by the N-terminal reg-ulatory lobe that contains a dishevelled/Egl/pleckstrin (DEP)domain and one or two cAMPbinding (CBD) domains in Epac1or Epac2, respectively.Since the initial discovery of the Epac family of proteins in

1998 (1, 2), significant progress has beenmade toward elucidat-ing the molecular mechanism of Epac activation using a varietyof structural and molecular biophysical approaches, includingthe x-ray crystal structure determinations of the apo-full-length Epac2 protein in 2006 and the ternary complex of anEpac2 deletion construct in-complex with a cAMP analog andRAP1 in 2008 (11, 12). These twocrystal structureshaveprovideda clear “before and after” snapshot of the cAMP-induced activa-tion process in atomic detail. In the apo-Epac2 protein, the accessof downstreameffector to the catalytic lobe is sterically blocked bythe regulatory lobe. cAMP binding triggers a chain of structuralreorganizations that includesa localized“hinge”motionthat reori-ents the autoinhibitory regulatory lobe away from the catalyticcore and leads to the eventual activation of Epac.Despite these advances, our understanding of the mecha-

nism of Epac activation remains incomplete. For example,extensive studies have revealed that Epac proteins exist, in solu-tion, as a dynamic ensemble ofmultiple conformations (13, 14).X-ray crystal structures usually capture only one of the many

* This work was supported, in whole or in part, by National Institutes of HealthGrants GM066170 (to X. C. and V. L. W.) and GM020501, NS070899 GM093325and RR029388 (V. L. W.) and NIEHS Center Grant ES06676 (to X. C.).

� This article was selected as a Paper of the Week.1 Both authors contributed equally to this work.2 To whom correspondence may be addressed: Dept. of Medicine, University

of California San Diego, 9500 Gilman Dr., La Jolla, CA 92093-0656. Tel.:858-534-2180; Fax: 858-534-2606; E-mail: vwoods@ucsd.edu.

3 To whom correspondence may be addressed: Dept. of Pharmacology andToxicology, The University of Texas Medical Branch, 301 University Blvd.,Galveston, TX 77555-0615. Tel.: 409-772-9656; Fax: 409-772-7050; E-mail:xcheng@utmb.edu.

4 The abbreviations used are: Epac, exchange protein directly activated bycAMP; CBD, cAMP binding domain; DXMS, deuterium exchange massspectrometry; DEP, dishevelled, Egl-10, pleckstrin domain; GEF, guaninenucleotide exchange factor; H/D, hydrogen/deuterium; RA, RAS associ-

ation; RAP, RAS-proximate; REM, RAS exchange motif; (Sp)-cAMP, aden-osine-3�, 5�-cyclic monophosphorothioate, (Sp) isomer; GuHCl, guani-dine hydrochloride.

THE JOURNAL OF BIOLOGICAL CHEMISTRY VOL. 286, NO. 20, pp. 17889 –17897, May 20, 2011© 2011 by The American Society for Biochemistry and Molecular Biology, Inc. Printed in the U.S.A.

23

possible low energy conformations that is compatible with thecrystal lattice. Although the apo-Epac2 crystal structure isbased on the full-length protein, the Epac2 protein in the activecomplex lacks the first CBD and DEP domains that are criticalfor Epac functions in vivo (15–18). In addition, the electrondensity for several flexible regions in Epac2 is missing in boththe autoinhibitory and the active structures. It is generallybelieved that structurally dynamic and flexible regions in pro-teins often play critical roles in function. To address these gapsin our understanding, we have probed the cAMP-inducedEpac2 activation using peptide amide hydrogen/deuteriumexchange mass spectrometry (DXMS), an approach comple-mentary to x-ray crystallography, capable of providing solutionstructural information about regions of the full-length Epac2 inboth its apo and its cAMP-bound states, which may be toodynamic to be stabilized and visualized in homogeneous crystals.

EXPERIMENTAL PROCEDURES

Protein Expression and Purification—Recombinant full-length Epac2 was expressed and purified as described previ-ously (14). All proteins were at least 95% pure, as judged bySDS-polyacrylamide gel electrophoresis.Optimization of Protease Digestion Conditions for DXMS

Analysis—To optimize peptide probe coverage, the concentra-tion of denaturing guanidine hydrochloride (GuHCl) was sys-tematically varied prior to acid proteolysis. 5 �l of stock solu-tion of Epac2 (12.5 mg/ml in 10 mM Tris-HCl, pH 7.5, 5 mM

DTT, 1 mM EDTA, 150 mM NaCl) was diluted with 15 �l ofwater and then quenched with 30 �l of 0.8% (v/v) formic acidcontaining various concentrations of GuHCl (0.8, 1.6, 3.2, 6.4M) at 0 °C and frozen at �80 °C as described previously (19).The frozen quenched samples were automatically thawed (20)and then immediately passed over tandem protease columns(porcine pepsin, 66-�l bed volume, followed byAspergillus sai-toi fungal protease type XIII, 66-�l bed volume) (21), with0.05% trifluoroacetic acid (TFA) in water (Solvent A) at a flowrate of 100 �l/min for 4 min. The proteolytic products weredirectly collected by a reverse phase C18 guard column (Vydaccatalog number 218GD51MS) and then desalted by flushing for1 min with 0.05% TFA at a flow rate of 100 �l/min. The chro-matographic separation was achieved on a Vydac C18 column(C18, 50� 1.0mm inner diameter, 5�m)with a linear gradientof 8–48% Solvent B (80% acetonitrile and 0.01% TFA) over 30min at 50 �l/min. The eluant was directed to a Finnigan LCQClassic mass spectrometer with electrospray mass ionizationvoltage set at 5 kV, capillary temperature at 200 °C, and dataacquisition in either MS1 profile mode or data-dependentMS/MS mode. The SEQUEST software program was used toidentify the sequence of the protease-generated peptide ions.Tentative identifications were confirmed using specialized datareduction software (DXMS Explorer, Sierra Analytics Inc.,Modesto, CA) (19). Optimal peptide coverage maps wereobtained employing a final concentration of 2.0 M GuHCl in0.5% formic acid.DXMS Analysis—Deuterated samples were prepared at 0 °C

by diluting 5�l of Epac2 stock solutionwith 15�l of deuteratedbuffer (8.3mMTris, pD 7.2, 1mMDTT, 150mMNaCl) followedby on-exchange incubation for varying times (10, 30, 100, 1,000,

10,000, and 100,000 s) prior to exchange quenching by the addi-tion of 30 �l of quench buffer (0.8% formic acid, 3.2 M GuHCl)at 0 °C and then stored frozen at �80 °C. Deuterated sampleswere then automatically thawed at 0 °C and subjected to proteo-lysis and LC/MS analysis as described above, along with con-trol samples of non-deuterated and equilibrium-deuteratedEpac2 (prepared by incubation of protein in 0.5% formic acid in100% D2O for 12 h at room temperature). The centroids ofisotopic envelopes of non-deuterated, partially deuterated andequilibrium-deuterated peptides were measured using DXMSExplorer and then converted to deuteration levels with correc-tions for back-exchange by the equation: D�MaxD� (m(P)�m(N))/(m(E) � m(N)), where m(P), m(N), and m(E) are thecentroid value of partially deuterated, non-deuterated, andequilibrium deuterated peptides, respectively (22). MaxD is thecalculated maximum deuterium incorporation to the givenpeptide (23). The deuteron incorporation within the sequenceof Epac2 was further sublocalized using differences betweenoverlapping peptides as described previously (24). The H/Dexchange experiments performed with our automated appara-tus typically produce measurements of deuteron incorporationwhere the standard deviation is less than 2% of the mean oftriplicate determinations (25, 26). In the present work, as in ourprevious studies, only changes in deuteration level greater than10% are considered as significant (23, 27).

RESULTS

Protease Fragmentation of Epac2 and Peptide Identification—To maximize the resolution and sequence coverage forEpac2 DXMS analysis, tandem protease digestion using pep-sin andA. saitoi fungal protease type XIII was employed (21).Protease digestion and HPLC separation conditions thatproduced Epac2 fragments of optimal size and distributionfor exchange analysis were established prior to H/Dexchange experiments. Optimal pepsin digestion for Epac2was obtained by quenching deuterated samples to a finalconcentration of 2.0 M GuHCl in 0.5% formic acid). Theseconditions generated 278 unique peptides covering 96% ofthe Epac2 sequence (Fig. 1).DXMS Analysis of Epac2 in the cAMP-free State—Exchange

of full-length apo-Epac2 was performed at 0 °C for various timeintervals ranging from 10 to 100,000 s. The MS isotopic enve-lopes of one specific peptide fragment, Val447–Leu457, prior andsubsequent to functional deuteration are shown in Fig. 2. Incor-poration of deuterium was clearly evident from the increase inoverall mass and complexity of the peptide mass peaks as afunction of deuteration time. The deuteration levels of cAMP-free Epac2 peptide fragments were determined as a function oftime to generate the DXMS profile of apo-Epac2. When therewere two or more overlapping peptide fragments in the sameregion, it was possible to sublocalize the deuterium by subtrac-tion to provide a DXMS profile with improved resolution asillustrated in Fig. 3. Six regions of Epac2, 1–14, 168–173, 458–483, 498–512, 529–541, and 618–629, were found to be morethan 70% deuterated at the earliest time point examined, indic-ative of a highly dynamic region whose amides are readilyexposed to solvent water. On the other hand, regions 186–197,563–570, 762–775, 796–812, 822–857, 896–899, 928–932,

Mechanism of Epac2 Activation

24

943–947, and 982–986 showed very low deuteration levels, lessthan 20% even at the longest time point. These peptides withvery slow exchange rates are in the core of DEP, REM, and

CDC25 homology domains and represent very stable regions ofthe protein.Effect of cAMP Binding on Epac2 Hydrogen Exchange—To

elucidate the conformational changes associated with Epac2activation, the time-dependent amide H/D exchange patternsof Epac2 in the presence and absence of 2 mM cAMP weremeasured and compared. For many of the peptide fragments,the binding of cAMP had no effect on the time-dependentincorporation of deuterium. On the other hand, the ratesand/or the extent of the exchange of several peptides changed,either increasing or decreasing, respectively, in response tocAMP binding over the experimental time course, indicatingthat local environments of these peptide fragments are per-turbed by the binding of cAMP.The exchange difference plot between active and inactive

Epac2 was generated by subtracting the DXMS profile of apo-Epac2 from that of Epac2�cAMP (Fig. 4A) and presented in Fig.5A, which allows ready visualization of the regions in Epac2 thatunderwent changes in exchange in response to the binding ofcAMP. Among the peptides with different rates of deuteriumexchange in the presence and absence of cAMP, a group ofthem showed a decreased exchange rate in the presence ofcAMP, as evidenced by peptides 71–80, 109–116, 117–129,367–376, and 403–417. These regions mostly overlap with thecAMP binding pockets A and B in Epac2. The apparentdecrease in exchange at the cAMPbinding pockets upon cAMPbinding is consistent with the notion that ligand binding ingeneral decreases local protein dynamics and/or provides steric

FIGURE 1. Pepsin digestion maps of Epac2. The peptide fragmentation pattern (indicated by the solid lines) of cAMP-free Epac2 is shown. The secondary structuresof Epac2 are shown above the peptide fragments and colored by domain: yellow, CBD-A; cyan, DEP; green, CBD-B; brown, REM; pink, RA; blue, GEF domain. The samedomain color scheme is used for Figs. 3–5 unless indicated otherwise, as indicated by color-coded legends in the bottom right corners of the figures.

FIGURE 2. Mass spectra of an Epac2 peptide fragment as a function ofdeuteration time. Shown is an isotopic envelope for peptide fragmentVal447–Leu457 before deuteration (ND), at various time points of deuteration,and after full deuteration (FD), displaying increased backbone amide deu-teron incorporation as indicated by the shifting of the isotopic envelope to ahigher m/z ratio.

Mechanism of Epac2 Activation

25

protection of residues directly involved in binding. Althoughour DXMS analysis could not distinguish between these possi-bilities, the coincidence between the observed major region ofsolvent protection and the cAMP binding pocket provides anindependent validation for our DXMS analysis. Changes indeuterium exchange rate in response to cAMP binding werealso observed for several peptide fragments located outside thecAMP binding sites. The majority of these peptide fragments,as represented by peptides 35–43, 224–233, 301–310, 434–448, and 928–932, displayed increased exchange upon cAMPbinding, whereas a few peptide fragments, such as 910–925 and920–927, showed a decreased exchange rate. These observedchanges in solvent accessibility are at regions outside the cAMPbinding pocket.Comparative DXMS Analyses of Epac2 in the Presence of

cAMP and (Sp)-cAMP—To help interpret the observed poten-tial cAMP-induced conformational changes revealed byDXMSanalysis, we also determined the amideH/D exchange profile ofEpac2 in the presence of 500 �M (Sp)-cAMP (Fig. 4B) as thecrystal structure of active Epac2 was recently solved as theEpac2�305�(Sp)-cAMP�RAP1 ternary complex. The exchange

difference plot between Epac2�(Sp)-cAMP and apo-Epac2 wasessentially identical to that between Epac2�cAMP and apo-Epac2 except within a region of the first N-terminal 130amino acids, which corresponds to the cAMP bindingdomain A (CBD-A) of Epac2 (Fig. 5B). The apparent lack ofprotection of the CBD-A domain by (Sp)-cAMP suggestseither that (Sp)-cAMP is not capable of binding to CBD-A inEpac2 or that the binding affinity is too weak under theexperimental conditions employed in this study. This differ-ential binding of cAMP and (Sp)-cAMP provided a windowof opportunity to further dissect the characteristics of thetwo cAMP binding sites in Epac2. The crystal structure ofthe full-length apo-Epac2 shows that cAMP binding toCBD-A and CBD-B is mutually exclusive (11), suggestingthat the events of cAMP binding are interdependent andsequential, i.e. the binding of the first cAMP leads to a con-formational change that allows the binding of a secondcAMP. This observation is further supported by our earlierisothermal titration calorimetry analysis in which the bind-ing isotherm for cAMP derived from the integrated heat datacan be fit best by a model of two sequential binding sites, but

FIGURE 3. Amide H/D exchange profiles of apo-Epac2. The percentages of deuteration levels of each peptide fragment at various time points are shown asa heat map color-coded from blue (�10%) to red (�90%), as indicated at the bottom right of the figure. Each block represents a peptide segment analyzed ateach of the six time points (from top to bottom: 10, 30, 100, 1,000, 10,000, and 100,000 s). Proline residues, and regions with no amide exchange data availableare colored in gray. Three independent DXMS analyses were performed. Standard deviations of deuterium incorporation were 2.5% or less between replicates.

Mechanism of Epac2 Activation

26

not by amodel with two independent sites (28). However, theexact order of cAMP binding (A site first or B site first) has notbeen established. Our current DXMS analyses reveal a condi-tion in which only the CBD-B site is protected by cAMP. Thisobservation suggests that cAMP binding to CBD-B does notrequire the occupancy of the CBD-A, which rules out the A site

first binding sequence. Taken together, our data suggest thatcAMP binding to Epac2 follows a sequential model with bind-ing at the B site proceeding that of the A site. In addition, ourresults further confirm that cAMP binding to CBD-A is notrequired for biochemical activation of Epac2 in vitro because(Sp)-cAMP, at 500 �M, is capable of fully activating Epac2.

FIGURE 4. Summary of hydrogen/deuterium exchange rates of Epac2 in the presence of ligand. A and B, deuteration levels of representative peptidefragments of Epac2�cAMP complex (A) and Epac2�(Sp)-cAMP complex (B) at various time points (from top to bottom: 10, 30, 100, 1,000, 10,000, and 100,000 s)are shown as a pseudo color scale. The percentages of deuteration levels of each peptide fragment at various time points are shown as a heat map color-codedfrom blue (�10%) to red (�90%), as indicated at the bottom right of the figure.

Mechanism of Epac2 Activation

27

DISCUSSION

DXMS has proven to be a powerful approach for studyingprotein dynamics and conformational changes, providinginformation complementary to high-resolution structuraltechniques. Although x-ray crystallography is capable of pro-viding exquisite structural detail at the atomic level, attempts tounderstand protein dynamics and conformational changes

relying solely on crystallography data are limited by constraintsplaced upon protein conformation by the crystal lattice and/orthe fact that highly flexible/dynamic regions of protein areoften refractory to crystallization. In this study, we have usedDXMS to probe the protein dynamics and conformationalchanges associated with Epac2 activation upon binding ofcAMP.

FIGURE 5. Changes in hydrogen/deuterium exchange rates of Epac2 induced by binding of cAMP or (Sp)-cAMP. A and B, differences in deuteration levelsin the free and cAMP-bound Epac2 (A) or in the free and (Sp)-cAMP-bound Epac2 (B) at various time points (from top to bottom: 10, 30, 100, 1,000, 10,000, and100,000 s) are shown in a color-coded bar ranging from blue (�50%) to red (50%), as indicated at the bottom right of the figure.

Mechanism of Epac2 Activation

28

Monitoring amide hydrogen exchange rates provides a directmeasurement of relative backbone solvent accessibilities,which in turn allow assessment of local conformational flexibil-ity or dynamics. Several Epac2 peptides exhibited fast H/Dexchange rates as defined as more than 75% deuteration after

100 s of exposure inD2Oat 0 °C (Fig. 3), indicative of disorderedor highly dynamic regions with solvent-exposed amides.Whenwemapped these peptide segments onto the x-ray crystal struc-ture of apo-Epac2, most of these fragments were located in theloop and turn regions (Fig. 6A). This is consistent with the fact

FIGURE 6. Highly dynamic and flexible regions in Epac2 revealed by DXMS. A, stereo view of the apo-Epac2 model. The yellow C� spheres denote dynamicregions that exchange rapidly. The domains are colored as follows: yellow, CBD-A; cyan, DEP; green, CBD-B; brown, REM; pink, RA; blue, GEF domain. This figureis based on the apo structure (Protein Data Bank (PDB) ID 2BYV) with missing disordered loops modeled using Swiss-Model. Insets show the orientation of theview relative to the complete structure. Regions colored bright blue or red show decreased or increased H/D exchange, respectively, upon cAMP binding. Themodeled cAMP molecules are shown as Corey-Pauling-Koltun-colored sticks. C, close-up of the dynamic loops in the apo-REM domain.

FIGURE 7. cAMP-induced conformational changes in Epac2 revealed by DXMS. Left and right panels, graphic representation of x-ray crystal structures of thefull-length apo-Epac2 (PDB ID: 2BYV; left panel) and Epac2�305�(Sp)-cAMP�RAP1B complex (PDB ID: 3CF6; right panel). Regions with significantly enhanced orreduced amide H/D exchanges as a result of cAMP binding are highlighted in bright red or blue, respectively. The domains are colored as follows: yellow, CBD-A;cyan, DEP; green, CBD-B; brown, REM; pink, RA; blue, GEF domain. The deleted N-terminal domains in the active structure (right) are indicated by the pale yellowarea. The RAP1B molecule is shown as a semitransparent purple surface. Small arrows show the domain motions in passing from the apo to complex form.

Mechanism of Epac2 Activation

29

that loops and turns are usually exposed to solvent and aremoreflexible. Moreover, three of the fastest exchanging fragments,peptides 1–14, 458–483, and 618–629, as revealed by DXMS,match to three out of the four regions in the Epac2 structurethat lack significant electron density, a sign of high flexibilityassociated with structural disorder.Most of the fast-exchanging peptides of Epac2 are clustered

in discrete regions of the Epac2 surface. These dynamic regionsof Epac2 are most likely to have important functional implica-tions. For example, in the autoinhibitory apo-Epac2 structure,the first and second CBDs are arranged in a face-to-face con-figuration so that each blocks the entry of the ligand into thebinding site of the other (11). It is not clear how cAMP gainsaccess to the cAMP binding pocket, particularly in the secondCBD, where it must bind to induce activation of Epac2. OurDXMS analyses show that the interface between CBD-A andCBD-B is highly dynamic (Fig. 6B). Therefore, it is conceivablethat a highly dynamic/flexible domain interface may lead to atransient opening of the two CBDs or a local structural unfold-ing, which allows cAMP to access both of the cAMP bindingpockets. Another highly dynamic region in Epac2 is located atthe hinge/switchboard and the REM domain (Fig. 6C). Thisregion undergoesmajor conformational changes in response tocAMP-induced Epac2 activation that include a bending of thehinge toward the �-core of CBD-B, a capping of the cAMPbinding pocket with the lid formed by the C-terminal�-strandsof CBD-B and the first helix of the REM domain, and the for-mation of a new interface between the CBD-B and REMdomains (12). The observed high flexibility at this region isprobably required to accommodate these dramatic conforma-tional changes.Whenwemapped the changes in the rates of amide hydrogen

exchange in response to cAMP binding along the primarysequence (Fig. 5A) or three-dimensional structures of Epac2(Fig. 7), respectively, two patterns emerged. First, almost all thechanges in solvent accessibility induced by cAMP are located intheN-terminal half of the protein that spans the regulatory lobeof Epac2, including both the CBDs, the DEP domain, and thehinge switchboard. The catalytic lobe, except for a small regionnear the C terminus, remains relatively unaffected by cAMPbinding. Second, the majority of cAMP-induced changes in theamide hydrogen exchange rates are located within or immedi-ately adjacent to the highly dynamic regions of Epac2. Thisclose proximity between regions of high flexibility and areas ofconformational change suggests that protein dynamics mayplay an important role in cAMP-induced Epac2 activation.The largest observed changes in the rates of exchange

induced by binding of cAMP are centered precisely at residues444–448, the hinge of Epac2 (Fig. 5A). This is not surprising asextensive structural and biochemical studies have demon-strated that the hinge/switchboard region is the most crucialstructural component required for Epac activation (11–15, 19).Structural comparison between the crystal structures of apo-Epac2 and the Epac2�305�(Sp)-cAMP�RAP1B complex revealsthat the hinge helix swings closer to the core of the CBD-Bdomain. During this process, the last two turns of the hingehelix dissolve into an extended loop between the hinge helixand the C-terminal �-strands that allows the C-terminal

�-strands and the whole catalytic region to rotate about 90°sideways (11, 12). The apparent increased amide hydrogenexchange rate at the hinge region in response to the binding ofcAMP is consistent with the structural observation of a partialmelting of the hinge helix (Fig. 8A). In addition to the confor-mational changes at the hinge that lead to the exposure of theRAP1 binding surface, the binding of cAMP also produces sig-nificant alterations in solvent accessibility at regions in Epac2that interact directly with RAP1. As shown in Fig. 8B, one helixof the helical hairpin, to which the switch II of RAP1 region ispacked, is highly dynamic in the inactive, apo-Epac2 state. This

FIGURE 8. Details of the cAMP-induced conformational changes in Epac2revealed by DXMS. A, structure of the hinge/switchboard in the (Sp)-cAMP-bound Epac2. The regions highlighted in bright red or blue become respec-tively more or less exposed upon cAMP binding. The domains are colored asfollows: yellow, CBD-A; cyan, DEP; green, CBD-B; brown, REM; pink, RA; blue,GEF. B, the RAP1B binding site. Residues involved in RAP1B binding to theactive Epac2 molecule are shown as purple sticks. The RAP1B molecule wasremoved for clarity. C, the DEP domain of apo-Epac2 with regions of increasedsolvent accessibility due to cAMP binding colored in red.

Mechanism of Epac2 Activation

30

helix and part of the ionic latch loop show decreased amideexchange upon cAMP binding. This cAMP-induced decreasein protein dynamics at the RAP1 binding surfacemay produce afavorable entropic contribution to RAP1 binding, providing afine-tuning mechanism for regulating the Epac2-RAP1 inter-action. The observation of cAMP-induced conformationalchanges right at the RAP1B binding interface suggests thatcAMP, in addition to relieving the steric hindrance imposedupon the catalytic lobe by the regulatory lobe, may also directlyregulate the interaction between Epac2 and RAP1 as an allo-steric modulator.Changes in amide hydrogen exchange in response to cAMP

binding have also been observed within the DEP domain (Fig.8C) including a part of the long helix at the interface betweenthe DEP and CBD-B domains and a nearby short helix loop.Both of these regions became more solvent-exposed in thepresence of cAMP. On the other hand, one exposed loop in theDEP domain became less solvent-accessible upon cAMP bind-ing. These observations suggest that the DEP domain mostlikely undergoes a significant conformational change duringEpac activation, which may include a domain rearrangementbetween the DEP and CBD-B domains. The DEP domain, ini-tially identified in disheveled, EGL-10, and pleckstrin protein(29), is a highly homologous structuralmotif found in a numberof proteins involved in signal transduction, such as regulatorsof G-protein signaling (RGS) and Epac. Consistent with thenotion that DEP domains function as a membrane anchor viainteractions with phospholipids and/or membrane proteins(30, 31), theDEPdomain is known to be necessary for themem-brane targeting of Epac1 (32). Recently published data showthat in addition to the temporal control of Epac1 GEF activity,cAMP also induces the translocation of Epac1 toward theplasma membrane. Importantly, the DEP domain is a neces-sary determinant of plasma membrane translocation that isrequired for the proper cellular function of Epac1 (17).Although it has been shown that the first CBD of Epac2 isrequired for protein localization to the plasma membrane(16), whether the DEP domain and the observed conforma-tional changes in the DEP domain induced by cAMP play arole in Epac2 cellular targeting and translocation, respec-tively, requires further investigation.

REFERENCES1. de Rooij, J., Zwartkruis, F. J., Verheijen, M. H., Cool, R. H., Nijman, S. M.,

Wittinghofer, A., and Bos, J. L. (1998) Nature 396, 474–4772. Kawasaki, H., Springett, G. M., Mochizuki, N., Toki, S., Nakaya, M., Mat-

suda, M., Housman, D. E., and Graybiel, A. M. (1998) Science 282,2275–2279

3. Gloerich, M., and Bos, J. L. (2010) Annu. Rev. Pharmacol. Toxicol. 50,355–375

4. Cheng, X., Ji, Z., Tsalkova, T., and Mei, F. (2008) Acta Biochim. Biophys.Sin. 40, 651–662

5. Dodge-Kafka, K. L., Soughayer, J., Pare, G. C., Carlisle Michel, J. J., Lange-berg, L. K., Kapiloff, M. S., and Scott, J. D. (2005) Nature 437, 574–578

6. Mei, F. C., and Cheng, X. D. (2005)Mol. Biosyst. 1, 325–3317. Nijholt, I.M., Dolga, A.M.,Ostroveanu, A., Luiten, P. G., Schmidt,M., and

Eisel, U. L. (2008) Cell. Signal. 20, 1715–17248. Ozaki, N., Shibasaki, T., Kashima, Y., Miki, T., Takahashi, K., Ueno, H.,

Sunaga, Y., Yano, H., Matsuura, Y., Iwanaga, T., Takai, Y., and Seino, S.(2000) Nat. Cell Biol. 2, 805–811

9. Shibasaki, T., Sunaga, Y., Fujimoto, K., Kashima, Y., and Seino, S. (2004)J. Biol. Chem. 279, 7956–7961

10. Liu, C., Takahashi, M., Li, Y., Dillon, T. J., Kaech, S., and Stork, P. J. (2010)Mol. Cell Biol. 30, 3956–3969

11. Rehmann,H., Das, J., Knipscheer, P.,Wittinghofer, A., and Bos, J. L. (2006)Nature 439, 625–628

12. Rehmann, H., Arias-Palomo, E., Hadders, M. A., Schwede, F., Llorca, O.,and Bos, J. L. (2008) Nature 455, 124–127

13. Rehmann, H., Rueppel, A., Bos, J. L., and Wittinghofer, A. (2003) J. Biol.Chem. 278, 23508–23514

14. Tsalkova, T., Blumenthal, D. K., Mei, F. C., White, M. A., and Cheng, X.(2009) J. Biol. Chem. 284, 23644–23651

15. de Rooij, J., Rehmann,H., vanTriest,M., Cool, R. H.,Wittinghofer, A., andBos, J. L. (2000) J. Biol. Chem. 275, 20829–20836

16. Niimura,M.,Miki, T., Shibasaki, T., Fujimoto,W., Iwanaga, T., and Seino,S. (2009) J. Cell Physiol. 219, 652–658

17. Ponsioen, B., Gloerich, M., Ritsma, L., Rehmann, H., Bos, J. L., and Jalink,K. (2009)Mol. Cell Biol. 29, 2521–2531

18. Mei, F. C., Qiao, J., Tsygankova, O.M.,Meinkoth, J. L., Quilliam, L. A., andCheng, X. (2002) J. Biol. Chem. 277, 11497–11504

19. Brock, M., Fan, F., Mei, F. C., Li, S., Gessner, C., Woods, V. L., Jr., andCheng, X. (2007) J. Biol. Chem. 282, 32256–32263

20. Wong, L., Lieser, S., Chie-Leon, B., Miyashita, O., Aubol, B., Shaffer, J.,Onuchic, J. N., Jennings, P. A., Woods, V. L., Jr., and Adams, J. A. (2004) J.Mol. Biol. 341, 93–106

21. Englander, J. J., Del Mar, C., Li, W., Englander, S. W., Kim, J. S., Stranz,D. D., Hamuro, Y., andWoods, V. L., Jr. (2003) Proc. Natl. Acad. Sci. U.S.A.100, 7057–7062

22. Zhang, Z., and Smith, D. L. (1993) Protein Sci. 2, 522–53123. Hamuro, Y., Anand, G. S., Kim, J. S., Juliano, C., Stranz, D. D., Taylor, S. S.,

and Woods, V. L., Jr. (2004) J. Mol. Biol. 340, 1185–119624. Burns-Hamuro, L. L., Hamuro, Y., Kim, J. S., Sigala, P., Fayos, R., Stranz,

D. D., Jennings, P. A., Taylor, S. S., andWoods, V. L., Jr. (2005) Protein Sci.14, 2982–2992

25. Hsu, S., Kim, Y., Li, S., Durrant, E. S., Pace, R. M., Woods, V. L., Jr., andGentry, M. S. (2009) Biochemistry 48, 9891–9902

26. Hsu, Y. H., Burke, J. E., Stephens, D. L., Deems, R. A., Li, S., Asmus, K. M.,Woods, V. L., Jr., and Dennis, E. A. (2008) J. Biol. Chem. 283, 9820–9827

27. Burke, J. E., Babakhani, A., Gorfe, A. A., Kokotos, G., Li, S., Woods, V. L.,Jr., McCammon, J. A., and Dennis, E. A. (2009) J. Am. Chem. Soc. 131,8083–8091

28. Tsalkova, T., Gribenko, A. V., and Cheng, X. (2011) Assay Drug Dev.Technol. 9, 88–91

29. Ponting, C. P., and Bork, P. (1996) Trends Biochem. Sci. 21, 245–24630. Wong, H. C., Mao, J., Nguyen, J. T., Srinivas, S., Zhang, W., Liu, B., Li, L.,

Wu, D., and Zheng, J. (2000) Nat. Struct. Biol. 7, 1178–118431. Simons, M., Gault, W. J., Gotthardt, D., Rohatgi, R., Klein, T. J., Shao, Y.,

Lee, H. J., Wu, A. L., Fang, Y., Satlin, L. M., Dow, J. T., Chen, J., Zheng, J.,Boutros, M., and Mlodzik, M. (2009) Nat. Cell Biol. 11, 286–294

32. Qiao, J.,Mei, F. C., Popov, V. L., Vergara, L. A., andCheng, X. (2002) J. Biol.Chem. 277, 26581–26586

Mechanism of Epac2 Activation

31

Shortened Engineered Human Antibody CH2 DomainsINCREASED STABILITY AND BINDING TO THE HUMAN NEONATAL Fc RECEPTOR*□S

Received for publication, April 25, 2011, and in revised form, June 2, 2011 Published, JBC Papers in Press, June 13, 2011, DOI 10.1074/jbc.M111.254219

Rui Gong‡, Yanping Wang‡§, Yang Feng‡, Qi Zhao‡, and Dimiter S. Dimitrov‡1

From the ‡Protein Interactions Group, Center for Cancer Research Nanobiology Program, Center for Cancer Research, and§SAIC-Frederick, Inc., NCI-Frederick, National Institutes of Health, Frederick, Maryland 21702

The immunoglobulin (Ig) constant CH2 domain is critical forantibody effector functions. Isolated CH2 domains are promis-ing scaffolds for construction of libraries containing diversebinders that could also confer some effector functions.We haveshown previously that an isolated human CH2 domain is rela-tively unstable to thermally induced unfolding, but its stabilitycan be improved by engineering an additional disulfide bond(Gong, R., Vu, B. K., Feng, Y., Prieto, D. A., Dyba, M. A., Walsh,J. D., Prabakaran, P., Veenstra, T. D., Tarasov, S. G., Ishima, R.,andDimitrov,D. S. (2009) J. Biol. Chem. 284, 14203–14210).Wehave hypothesized that the stability of this engineered antibodydomain could be further increased by removing unstructuredresidues. To test our hypothesis, we removed the seven N-ter-minal residues that are in a random coil as suggested by ouranalysis of the isolated CH2 crystal structure and NMR data.The resulting shortened engineered CH2 (m01s) was highly sol-uble,monomeric, and remarkably stable,with amelting temper-ature (Tm) of 82.6 °C, which is about 10 and 30 °C higher thanthose of the original stabilized CH2 (m01) and CH2, respec-tively. m01s and m01 were more resistant to protease digestionthanCH2.Anewly identified anti-CH2antibody that recognizesa conformational epitope bound to m01s significantly better(>10-fold higher affinity) than to CH2 and slightly better thanto m01. m01s bound to a recombinant soluble human neonatalFc receptor at pH 6.0 more strongly than CH2. These data sug-gest that shortening the m01 N terminus significantly increasesstability without disrupting its conformation and that ourapproach for increasing stability and decreasing size by remov-ing unstructured regions may also apply to other proteins.

Monoclonal antibodies are nowwell established therapeuticsand invaluable tools for biological research (1). A major prob-lem for full-size monoclonal antibodies is their poor penetra-tion into some tissues (e.g. solid tumors) and poor or absentbinding to regions on the surface of somemolecules (e.g. on theHIV envelope glycoprotein) that are accessible by molecules of

smaller size. Antibody fragments, e.g. Fab fragments (�60 kDa)or single-chain Fv fragments (20�30 kDa), are significantlysmaller than full-size antibodies (�150 kDa) and have beenused as imaging reagents and candidate therapeutics. There-fore, discovery of even smaller scaffolds, including engineeredantibody domains, continues to be of major importance in thedevelopment of candidate therapeutics and imaging agents(2–4).The second domain of the heavy chain constant regions,

CH2, is unique among the other antibody domains in that itexhibits very weak carbohydrate-mediated interchain protein-protein interactions, in contrast to the extensive interchaininteractions that occur between the other domains. Theexpression of mouse and human CH2 in bacteria, which doesnot support glycosylation, results in a monomeric domain (5,6). We have proposed that the CH2 domain (CH2 of IgG, IgA,and IgD and CH3 of IgE and IgM) could be used as a scaffoldand could offer additional advantages compared with engi-neered antibody domains based on other domains because itcontains binding sites or portions of binding sites conferringeffector and stability functions (7). Supporting this possibility isthe finding that the half-life of humanCH2 (�70 h) in rabbits ismuch longer than that of CH3 and Fab (�15 h), andCH2mightfunction to trigger the complement system (8, 9).The native CH2 domain has significantly lower thermal sta-

bility compared with other small scaffolds such as the tenthtype III domain of human fibronectin (5, 6, 10), which increasesthe probability of instability when engineering binding to anti-gens and enhanced effector functions. In the quest for a morestable CH2-based scaffold, we found previously that the stabil-ity of an isolated human IgG1 CH2 can be significantlyincreased by engineering an additional disulfide bond betweenthe A and G strands (6). One of the newly developed mutants,denoted m01, exhibited significantly higher stability than wild-type CH2.We have hypothesized that the stability of m01 could be fur-

ther increased by removing unstructured terminal residuessuch as the seven N-terminal residues that are in a random coilas suggested by our analysis of the isolated CH2 crystal struc-ture and NMR data (6, 11). To test our hypothesis, we removedthese residues and characterized the resulting shortened engi-neered CH2 (m01s). m01s was remarkably stable, with a melt-ing temperature (Tm) of 82.6 °C, which is about 10 and 30 °Chigher than those of the original stabilized CH2 (m01) andCH2, respectively. To detect possible conformational changesin m01s compared with CH2 and m01, a novel anti-CH2 anti-body (m01m1) that recognizes a conformational epitope was

* This work was supported by the National Institutes of Health IntramuralAIDS Targeted Antiviral Program (IATAP), the National Institutes of HealthNIAID Intramural Biodefense Program, and the National Institutes ofHealth NCI Center for Cancer Research Intramural Research Program. Thiswork was supported, in whole or in part, by federal funds from NationalInstitutes of Health NCI Contract N01-CO-12400.

□S The on-line version of this article (available at http://www.jbc.org) containssupplemental “Experimental Procedures” and Fig. S1.

1 To whom correspondence should be addressed: CCRNP, CCR, NCI-Frederick,NIH, Bldg. 469, Rm. 150B, Frederick, MD 21702-1201. Tel.: 301-846-1352;Fax: 301-846-5598; E-mail: dimiter.dimitrov@nih.gov.

THE JOURNAL OF BIOLOGICAL CHEMISTRY VOL. 286, NO. 31, pp. 27288 –27293, August 5, 2011Printed in the U.S.A.

32

identified. It bound to m01s significantly better (�10-foldhigher affinity) than toCH2 and slightly better than tom01.Wealso expressed CH2 and m01s on the yeast cell surface andcompared their binding to a soluble human neonatal Fc recep-tor (shFcRn)2 (12). Interestingly, we found that the binding ofm01s was stronger than that of CH2 in a pH-dependentmanner.On the basis of these data, we suggest that m01s could be

used as a scaffold for development of engineered antibodydomains. These results also demonstrate for the first time thatthe stability of constant antibody domains can be furtherincreased by decreasing their size. Importantly, although short-ening the m01 N terminus significantly increases stability, itretains or increases some other properties of CH2 (e.g. bindingto shFcRn). The increase in stability of isolated domains mayresult in an increase in stability of larger antibody fragments,e.g. Fc, and therefore could have implications as a generalmethod for increasing antibody stability. It may also apply toother proteins as a method to increase stability and decreasesize.

EXPERIMENTAL PROCEDURES

m01 Mutant Design and Plasmid Construction—To designthem01mutant, we used the isolatedCH2 crystal structure andNMR data (6, 11). The truncated m01 mutant (denoted m01s)with the absence of seven residues in theN terminuswas clonedinto pComb3X (provided by Dennis Burton, The ScrippsResearch Institute, La Jolla, CA). The clone was verified bydirect sequencing and used for transformation of the Esche-richia coli strain HB2151. m01s was expressed and purifiedsimilarly to wild-type CH2 (6).Size Exclusion Chromatography—Purified m01s was loaded

into a HiLoad 26/60 Superdex 75 HR 10/30 column (GEHealthcare) running on an AKTA BASIC pH/C chromatogra-phy system (GE Healthcare) to assess possible oligomer forma-tion. PBS (pH 7.4) was selected as the mobile phase. Gel filtra-tion of standards consisting of aldolase (158 kDa), bovine serumalbumin (67 kDa), ovalbumin (44 kDa), chymotrypsinogen A(25 kDa), and ribonuclease A (13.7 kDa) was used to define themolecular mass.Circular Dichroism—The secondary structure of m01s was

determined by CD spectroscopy. The purified proteins werediluted in PBS at a final concentration of 0.54 mg/ml, and the2 The abbreviation used is: shFcRn, soluble human neonatal Fc receptor.

FIGURE 1. Design, expression, and estimation of oligomer formation of m01s. A, amino acid sequence alignment of wide-type CH2 (NCBI accession numberJ00228), m01, and m01s. B, comparison of the expression of CH2, m01, and m01s. M, marker. C, size exclusion chromatography of m01s. The inset is a standardcurve. mAU, milli-absorbance units.

Shortened Antibody Constant Domains

33

CD spectra were recorded on Aviv Model 202 CD spectrome-ter. Wavelength spectra were recorded at 25 °C using a 0.1-cmpath length cuvette for native structure measurements. Tomeasure the Tm, m01s was diluted in PBS (pH 7.4) with 0, 3, 3.5,4, and 5M urea, respectively. Thermal stabilities under differentconditions with 0, 3, 3.5, 4, and 5 M urea were measured at 216nm (in the absence of urea) and 225 nm (in the presence of ureaat different concentrations) by recording the CD signal at atemperature range of 25–90 °C with a heating rate of 1 °C/min.The temperature was recorded with an external probe sensor,and the temperature inside the microcuvette was calculated bycalibration; it was �2–3 °C (range of 1.9–4.0 °C for tempera-tures from 20 to 90 °C) lower that the one measured by theexternal sensor.Limited Proteolysis—The purified CH2, m01, and m01s pro-

teins were subjected to trypsin digestion at 1:5 (w/w) trypsin/protein for 0, 1.5, and 3 h at 37 °C. After digestion, loadingbuffer was added immediately to the samples to terminatedigestion, and all samples were examined by SDS-PAGE.Conformational Changes in m01s Detected by Anti-CH2 Fab

m01m1—ELISA was used for comparison of the structures ofCH2, m01, and m01s. Briefly, CH2, m01, m01s, and humanserumalbumin (negative control) were coated on 96-well platesat a concentration of 2 �g/ml. Anti-CH2 Fabm01m1 identifiedfrom a human Fab naïve library (13) by panning against isolatedCH2 expressed in E. coli, which recognized the conformationalepitope of CH2 (data not shown), was added at concentrationsof 0.064–200 �g/ml. HRP-conjugated anti-Fab (Sigma) was

used as the secondary antibody. To confirm the result, a biotin-conjugated commercial mouse anti-human CH2 monoclonalantibody (AbD Serotec) was also used for ELISA at concentra-tions of 0–10 �g/ml. HRP-streptavidin (Sigma) was used fordetection of biotinylated mouse anti-human CH2 antibody.Construction of CH2, m01s, Fc, and CH3 for Yeast Surface

Expression—CH2, m01s, Fc, and CH3 were cloned into thepYD7 vector, a modified version of pCTCON2 (14), allowingexpression of fusion proteins with agglutinin, Aga2p, located atthe C terminus. The clones were verified by direct sequencing.The constructs were transformed into EBY100 cells for surfaceexpression according to the protocol described previously (14).Flow Cytometry Assay—For measurement of the binding of

CH2, m01s, Fc, and CH3 to shFcRn, yeast cells containingpYD7-CH2, pYD7-m01s, pYD7-Fc, and pYD7-CH3 weregrown in SDCAAmedium (20 g/liter dextrose, 6.7 g/liter Difcoyeast nitrogen base w/o amino acid, 5 g/liter Bacto casaminoacids, 5.4 g/liter Na2HPO4, and 8.56 g/liter NaH2PO4�H2O),and expression was induced in SGCAA medium (20 g/litergalactose, 20 g/liter raffinose, 1 g/liter dextrose, 6.7 g/liter Difcoyeast nitrogen base w/o amino acid, 5 g/liter Bacto casaminoacids, 5.4 g Na2HPO4, and 8.56 g/liter NaH2PO4�H2O) accord-ing to published protocols (14). For shFcRn binding, 5 � 105yeast cells were harvested, washed with PBSA (PBS � 0.1%bovine serum albumin at pH 6.0), and resuspended in 50 �l ofPBSA (pH 6.0) containing 100 nM biotin-conjugated shFcRn.The samples was kept on ice for 2 h, and the cells were washedagain with PBSA (pH 6.0) and resuspended in 50 �l of PBSA

FIGURE 2. Stability of m01s. A, CD spectra of m01s measured at 25 °C. m01s displayed a maximum negative peak between 210 and 225 nm, indicating a typical�-helical secondary structure. deg, degrees. B, thermo-induced unfolding of m01s in PBS without urea. The change in mean residue ellipticities was monitoredat 216 nm. No second transition point was observed up to the highest temperature measured (90 °C). C, thermo-induced unfolding of m01s in PBS withdifferent concentrations of urea (3, 3.5, 4, and 5 M). The fraction folded of the protein (ff) was calculated as ff � ([�] � [�M])/([�T] � [�M]), where [�T] and [�M] arethe mean residue ellipticities at 225 nm of the folded state at 25 °C and the unfolded state at 90 °C. The Tm values (68.9, 65.7, 63.6, and 59.3 °C correspond to 3,3.5, 4, and 5 M urea, respectively) from CD were determined by the first derivative (d(fraction folded)/dT) with respect to temperature (T). D, calculation of them01s Tm (82.6 °C) at 0 M urea.

Shortened Antibody Constant Domains

34

(pH 6.0). 1 �l of R-phycoerythrin-streptavidin (PE-streptavi-din; Invitrogen) was added to the resuspended cells. After 30min of incubation on ice, the cells were washed with PBSA (pH6.0) and resuspended in 0.5ml of PBSA (pH6.0) for flow cytom-etry measurement. The mouse anti-human CH2 monoclonalantibody described above and Alexa Fluor 488-conjugated goatanti-mouse IgG (Invitrogen) were used to test the expression ofCH2,m01s, and Fc on the yeast, whereas FITC-conjugated goatanti-human Fc polyclonal antibody (Sigma) was used to test theexpression of CH3 on the yeast. The same samples prepared atpH 7.4 were used as controls.Competition Flow Cytometry Assay—To verify the specificity

of binding of m01s to shFcRn, human IgG1 was used as a com-petitor. 5 � 105 yeast cells expressing m01s were harvested,washed with PBSA (pH 6.0), and resuspended in 50 �l of PBSA(pH 6.0) containing 100 nM biotin-conjugated shFcRn andhuman IgG1 at serial concentrations (0, 125, 250, 500, 1000,2000, and 4000 nM). The sample prepared at pH7.4with 100 nMbiotin-conjugated shFcRn only was used as a control. Bindingwas then analyzed by the same method described above. Themean value was taken to calculate the inhibition efficiency.

RESULTS

Analysis of an IsolatedCH2Crystal Structure andNMRData—We have previously solved the crystal structure of an isolated

�1CH2domain (11). Analysis of theCH2N terminus suggestedthat the first seven residues form a random coil. Our previouslyreported NMR data (6) indicate that the very N-terminal resi-dues exhibit increased flexibility. Therefore, we have hypothe-sized that the disordered N terminus with increased dynamicsmay contribute to a decreased thermal stability and that itsremoval may result in increased stability.Design and Generation of an N-terminal Truncated Engi-

neered CH2 Domain—To test our hypothesis, we used a previ-ously developed engineered CH2 domain (m01) in which anadditional disulfide bond was introduced between the A and Gstrands, resulting in increased stability (6). The first sevenN-terminal residues of m01 were deleted, resulting in a shortervariant of m01 termed m01s (Fig. 1A). m01s was expressed inE. coli at a higher level than that observed for eitherCH2orm01(Fig. 1B). m01s was completely monomeric in PBS at pH 7.4 asdetermined by size exclusion chromatography (Fig. 1C).m01 Is Significantly More Stable than CH2 and m01—The

secondary structure and thermodynamic stability of m01s weremeasured by CD. The CD spectra of m01s showed that it hashigh �-sheet content at 25 °C (Fig. 2A). The �-sheet structurewas gradually disrupted, and the protein unfolded as the tem-perature increased (Fig. 2B). However, even at the highest tem-perature measured (90 °C), part of the protein was still in afolded state, and there was no second transition point (Fig. 2B).Therefore, we used an indirect approach to estimate the melt-ing temperature at which 50% of the protein is in the foldedstate (Tm) using increasing concentrations (3, 3.5, 4, and 5 M) ofurea to decrease the Tm and extrapolate to 0 M urea (Fig. 2C).The sigmoidal curves were fitted by a two-state model that wasused previously (5). Tm was calculated at each urea concentra-tion and fitted to a linear equation according to a previouslydescribedmethod (15):Tm� 82.6� 4.7Uc, whereUc is the ureaconcentration (Fig. 2D). Therefore, at Uc � 0, Tm equals82.6 °C. This value is significantly higher than those previouslymeasured by the same method for CH2 (54.1 °C) and m01(73.4 °C) (6).m01s and m01 Are More Resistant to Protease (Trypsin)

Digestion than CH2 but Are Equally Stable in Human Serum inVitro—To evaluate the stabilities of CH2, m01, and m01sagainst digestion by a typical protease, trypsin was used as an

FIGURE 3. Limited proteolysis of CH2, m01, and m01s by trypsin. After 3 hof digestion, m01 and m01s were partially digested, whereas CH2 was almostcompletely digested. M, marker.

FIGURE 4. Binding of CH2 (f), m01(F), m01s (Œ), and human serum albumin (ƒ) to anti-human CH2 Fab m01m1 selected from a human naïve Fablibrary (A) and commercial mouse anti-human CH2 IgG (B). The EC50 values of Fab m01m1 for CH2, m01, and m01s were �1305, 181, and 129 nM,respectively, whereas the EC50 values of commercial mouse anti-human CH2 IgG for CH2 and m01s were 4.9 and 0.59 nM, respectively. In both cases, m01s couldbe better recognized by the antibodies (Fab and mouse IgG) compared with CH2.

Shortened Antibody Constant Domains

35

example. Significantly larger amounts of CH2 were digestedcompared withm01 andm01s in 3 h (Fig. 3).We also estimatedtheir stability in human serum by incubation at 37 °C for 9 days(see supplemental “Experimental Procedures”). No significantdegradation was observed (supplemental Fig. S1). These datasuggest that the stabilized variants of CH2 aremore resistant toprotease (trypsin) digestion than CH2 but are equally stable inhuman serum in vitro.Conformational Changes in m01s and m01 Demonstrated by

a Newly Identified Anti-CH2 Fab, m01m1—To determinewhether the CH2 engineering resulted in conformationalchanges, we developed a human Fab, m01m1, specific for CH2.Fab m01m1 was identified by panning and screening of a Fablibrary against purified CH2 as described under “ExperimentalProcedures.” It was expressed in E. coli and purified. m01m1bound to CH2 with relatively low affinity (EC50 � 1305 nM)(Fig. 4A). Interestingly, the addition of a disulfide bond in m01and m01s resulted in a significant increase in binding: EC50 �181 and 129 nM, respectively (Fig. 4A). The shortening of m01led to a slight increase in m01m1 affinity (Fig. 4A). Similarresults were obtained with a commercial mouse anti-humanCH2monoclonal antibody that binds a conformational epitope(Fig. 4B). These results indicate that some conformationalepitopes are largely conserved in CH2, m01, and m01s, butthese conformational epitopes are more exposed in the engi-neered CH2 domains than they are in CH2.Binding of shFcRn at pH 6.0 to m01s Is Stronger Than That to

CH2—Recently, we developed a method to produce shFcRn inmammalian cells with high yield (12). We used this shFcRn totest its binding at pH 6.0 to CH2 and m01s expressed on yeastcells (Fig. 5). The fluorescence intensity shift for m01s waslarger than that for CH2, indicating stronger binding ofm01s toshFcRn comparedwith CH2. The largest fluorescence intensityshift in the case of Fc indicated that binding of shFcRn at pH 6.0to Fc on yeast cells was stronger than that tom01s on yeast cells.No fluorescence intensity shift was observed for CH3 at pH 6.0,indicating that CH3makes aminor contribution to the bindingof Fc to shFcRn.

FIGURE 5. Binding of yeast-expressed CH2, m01s, Fc, and CH3 to shFcRnat pH 6 (red) and pH 7.4 (blue). A very slight fluorescence intensity shiftoccurred in the case of CH2, indicating very weak binding to shFcRn. A mod-est fluorescence intensity shift was observed in the case of m01s, indicatingmodest binding to shFcRn. The largest fluorescence intensity shift wasobserved in the case of Fc, indicating strong binding to shFcRn. CH3 did notbind to shFcRn in a pH-dependent manner; there was no observable fluores-cence intensity shift. The expression of CH2, m01s, Fc, and CH3 was testedby the corresponding antibodies. PE-streptavidin was used as negativecontrol.

FIGURE 6. Inhibition of yeast-expressed m01s binding to shFcRn by IgG1. A, fluorescence intensity shift in the presence of IgG1 at different concentrations.B, inhibition curve. Percent inhibition � ((meanmax at pH 6.0 � mean at pH 6.0)/(meanmax at pH 6.0 � mean at pH 7.4)) � 100, whereas meanmax at pH 6.0 is themean value of the fluorescence intensity measured at pH 6.0 in the absence of IgG1, mean at pH 7.4 is the mean value of the fluorescence intensity measuredat pH 7.4 in the absence of IgG1, and mean at pH 6.0 is mean value of the fluorescence intensity measured at pH 6.0 with different IgG1 concentrations. Thebinding decreased with an increase in the IgG1 concentration; IC50 � 804 nM.

Shortened Antibody Constant Domains

36

Binding of m01s to shFcRn Is Inhibited by Human IgG1—Totest the specificity of binding ofm01s to shFcRn at pH 6.0, IgG1was used as a competitor in a flow cytometry assay. With anincrease in the IgG1 concentration, the fluorescence intensityshift decreased (Fig. 6A). The half-maximal binding inhibitionconcentration (IC50) was 804 nM (Fig. 6B), which is consistentwith a previous report (16). Therefore, as expected, m01s andIgG1 share the same binding site on shFcRn; this result alsoindicates that the binding of m01s to shFcRn at pH 6.0 isspecific.

DISCUSSION

The major findings of this study are that shortening an iso-lated unglycosylated engineered human �1 CH2 domain signif-icantly increased its stability without affecting some otherproperties and the identification of a new anti-CH2 Fab,m01m1, which is sensitive to CH2 conformational changesinduced by an additional disulfide bond. The measured m01sTm is the highest achieved for isolated CH2-based domains andcomparable with that of the tenth type III domain of humanfibronectin (10). We have not investigated possible molecularmechanisms that determine the higher stability of m01s com-pared with CH2 and m01. One can speculate that decreaseddynamics at the N terminus is a major factor of the increasedstability. The decreased size could also contribute. Theseresults also indicate that the stability of other proteins could beincreased by removing unstructured regions and decreasingsize. The size dependence of folding and stability is in agree-ment with a previous analysis (17). The increased thermal sta-bility combined with high solubility and expression levels andresistance to proteases are highly desirable properties for a can-didate therapeutic protein based on the m01s scaffold. In addi-tion, shortening CH2 in intact Fc and Ig could also result in anincrease in their stability, and ongoing experiments are testingthis possibility.Naturally occurring additional disulfide bonds are found in

camel single-domain antibody fragments (VHH fragments)(between CDR1 and CDR3) and in the shark IgNAR (Ig newantigen receptor) V domain (between CDR3 and frameworks)(18, 19). The thermostability of a single isolated antibodydomain is typically increased by 10 °C after introduction of anadditional disulfide bond (20). The strategy is similar to thatused in the design of m01. It would be interesting to determinewhether shortening such domains with a naturally occurringsecond disulfide bond and also in general other antibodydomains could result in an increase in their stability.We also found that shorteningm01 (m01s) does not perturb,

to any significant degree, some other properties, includingbinding to shFcRn. Because of the lack of CH3, the binding ofm01s to shFcRn was relatively weak compared with whole Fc.

However, interestingly, there was an increase in binding com-paredwithCH2.We are currently further improving binding ofm01s to shFcRn, which may lead to an extended half-life ofm01s in vivo. However, the removal of N-terminal residuesfrom CH2 may affect binding to Fc� receptors and relatedeffector functions. Further studies are in progress to elucidatethis and other possible effects due to the N-terminal shorten-ing, including binding to complement.These findings could have implications for exploration of the

unfolding mechanisms of antibody domains and for the devel-opment of candidate therapeutic proteins with increased sta-bility and extended half-lives.Whether the observed increase instability against temperature and chemical agents and bindingto shFcRn in vitro will result in increased stability and longhalf-lives in vivo remains to be seen.

Acknowledgments—We thank Dr. Sergey G. Tarasov for helpful dis-cussions, Marzena A. Dyba for technical support, and Dr. K. DaneWittrup for providing plasmid pCTCON2.

REFERENCES1. Dimitrov, D. S. (2010)mAbs 2, 347–3562. Kolmar, H., and Skerra, A. (2008) FEBS J. 275, 26673. Skerra, A. (2007) Curr. Opin. Biotechnol. 18, 295–3044. Nygren, P. A., and Skerra, A. (2004) J. Immunol. Methods 290, 3–285. Feige, M. J., Walter, S., and Buchner, J. (2004) J. Mol. Biol. 344, 107–1186. Gong, R., Vu, B. K., Feng, Y., Prieto, D. A., Dyba, M. A., Walsh, J. D.,

Prabakaran, P., Veenstra, T. D., Tarasov, S. G., Ishima, R., and Dimitrov,D. S. (2009) J. Biol. Chem. 284, 14203–14210

7. Dimitrov, D. S. (2009)mAbs 1, 26–288. Yasmeen, D., Ellerson, J. R., Dorrington, K. J., and Painter, R. H. (1976)

J. Immunol. 116, 518–5269. Ellerson, J. R., Yasmeen, D., Painter, R. H., and Dorrington, K. J. (1972)

FEBS Lett. 24, 318–32210. Hackel, B. J., Kapila, A., and Wittrup, K. D. (2008) J. Mol. Biol. 381,

1238–125211. Prabakaran, P., Vu, B. K., Gan, J., Feng, Y., Dimitrov, D. S., and Ji, X. (2008)

Acta Crystallogr. D Biol. Crystallogr. 64, 1062–106712. Feng, Y., Gong, R., and Dimitrov, D. S. (2011) Protein Expr. Purif., in press13. Zhu, Z., and Dimitrov, D. S. (2009)Methods Mol. Biol. 525, 129–142, xv14. Chao, G., Lau, W. L., Hackel, B. J., Sazinsky, S. L., Lippow, S. M., and

Wittrup, K. D. (2006) Nat. Protoc. 1, 755–76815. Greene, R. F., Jr., and Pace, C. N. (1974) J. Biol. Chem. 249, 5388–539316. Suzuki, T., Ishii-Watabe, A., Tada, M., Kobayashi, T., Kanayasu-Toyoda,

T., Kawanishi, T., and Yamaguchi, T. (2010) J. Immunol. 184, 1968–197617. De Sancho, D., Doshi, U., and Munoz, V. (2009) J. Am. Chem. Soc. 131,

2074–207518. Saerens, D., Pellis, M., Loris, R., Pardon, E., Dumoulin, M., Matagne, A.,

Wyns, L., Muyldermans, S., and Conrath, K. (2005) J. Mol. Biol. 352,597–607

19. Stanfield, R. L., Dooley, H., Flajnik, M. F., andWilson, I. A. (2004) Science305, 1770–1773

20. Hagihara, Y., Mine, S., and Uegaki, K. (2007) J. Biol. Chem. 282,36489–36495

Shortened Antibody Constant Domains

37

An Undecided Coiled CoilTHE LEUCINE ZIPPER OF Nek2 KINASE EXHIBITS ATYPICAL CONFORMATIONAL EXCHANGEDYNAMICS*□S

Received for publication, October 22, 2010, and in revised form, May 14, 2011 Published, JBC Papers in Press, June 13, 2011, DOI 10.1074/jbc.M110.196972

Rebecca Croasdale‡, Frank J. Ivins§, Fred Muskett‡, Tina Daviter¶, David J. Scott�, Tara Hardy‡, Steven J. Smerdon§,Andrew M. Fry‡1, and Mark Pfuhl2

From the ‡Department of Biochemistry, University of Leicester, Leicester LE1 9HN, United Kingdom, the §Medical Research Council,National Institute of Medical Research, Division of Molecular Structure, Mill Hill, London NW7 1AA, United Kingdom, the¶ISMB Biophysics Centre, Birkbeck, University of London, London WC1E 7HX, United Kingdom, and the �NationalCentre for Macromolecular Hydrodynamics, School of Biosciences, University of Nottingham, Sutton Bonington,Leicestershire LE12 5RD, United Kingdom

Leucine zippers are oligomerization domains used in a widerange of proteins. Their structure is based on a highly conservedheptad repeat sequence in which two key positions are occupiedby leucines. The leucine zipper of the cell cycle-regulated Nek2kinase is important for its dimerization and activation. How-ever, the sequence of this leucine zipper is most unusual in thatleucines occupy only one of the two hydrophobic positions. Theother position, depending on the register of the heptad repeat, isoccupied by either acidic or basic residues. Using NMR spec-troscopy, we show that this leucine zipper exists in two confor-mations of almost equal population that exchange with a rate of17 s�1. We propose that the two conformations correspond tothe two possible registers of the heptad repeat. This hypothesisis supported by a cysteinemutant that locks the protein in one ofthe two conformations. NMR spectra of this mutant showed thepredicted 2-fold reduction of peaks in the 15N HSQC spectrumand the complete removal of cross peaks in exchange spectra. Itis possible that interconversion of these two conformationsmaybe triggered by external signals in a manner similar to that pro-posed recently for the microtubule binding domain of dyneinand the HAMP domain. As a result, the leucine zipper of Nek2kinase is the first example where the frameshift of coiled-coilheptad repeats has been directly observed experimentally.

Intracellular signaling pathways that regulate processes suchas cell cycle control rely on formation of specific protein com-plexes at the right time and place. As a result, a wide range ofconserved interaction motifs have evolved among which the

leucine zipper is one of the most common and versatile. Leu-cine zippers were first identified as dimerization domains inbZIP transcription factors with a sequence motif consisting ofleucines repeated every 7 amino acids (1). The relevance of therepeating heptad sequencewas clarifiedwhen it was shown thatleucine zippers assume a coiled-coil fold (2–4). In this struc-ture, the first and fourth residues (i.e. positions A and D in theheptad sequence, ABCDEFG) of each helix point toward eachother and thus form a hydrophobic core. Residues in positionsE and G flank the hydrophobic core residues and are oftenoccupied by charged residues that can form salt bridges. Thelatter are of particular significance in heterodimeric leucinezippers as they help to determine specificity. Residues in posi-tions B, C, and F are usually not of importance as their side-chains point away from the coiled-coil interface. Leucine zip-pers show great versatility as they can exist as dimers, trimers,or tetramers, can be homo- or hetero-oligomers and can formparallel or anti-parallel complexes (5–7).Although leucine zippers have been best characterized in

transcription factors, they also exist in many other signalingproteins including protein kinases (8). Protein kinase activationoften involves a trans-autophosphorylation step that is facili-tated by the physical proximity of two kinase molecules. In thecase of receptor tyrosine kinases, this may be brought about bycrosslinking of two receptors as a result of extracellular ligandbinding. Some cytoplasmic kinases on the other hand containtheir own oligomerization domain. One example is the cellcycle-regulated kinase, Nek2, which consists of an N-terminalcatalytic domain and a C-terminal region that contains multi-ple regulatory motifs, including a leucine zipper (9) (Fig. 1A).For this kinase, oligomerization via the leucine zipper is essen-tial for full activation both in vitro and in vivo, most likely as aresult of it promoting trans-autophosphorylation (10, 11).In a previous study on the role of the Nek2 leucine zipper, we

noted that the sequence, in terms of the distribution of hydro-phobic and charged residues, is somewhat unusual (10) (Fig. 2).However, no structural studies were undertaken at the time.Here, we now show that the Nek2 leucine zipper does indeeddisplay highly atypical biophysical properties with NMR spec-troscopy clearly showing that the leucine zipper exists in twoconformations, which exchange on a relatively slow timescale.This raises the intriguing possibility that the dimerization and,

* This work was supported by grants (to A. M. F.) from the Hope Foundationfor Cancer Research, Cancer Research UK, the Association for InternationalCancer Research, and The Wellcome Trust.

Assignments were deposited with the Biological Magnetic Resonance Data Bank,accession code 17417.

□S The on-line version of this article (available at http://www.jbc.org) containssupplemental Figs. S1–S6.Author’s Choice—Final version full access.

1 To whom correspondence may be addressed: University of Leicester,Department of Biochemistry, Lancaster Road, Leicester LE1 9HN, UK. Tel.:44-116-229-7069; E-mail: amf5@le.ac.uk.

2 To whom correspondence may be addressed: King’s College London,Randall Division for Cell and Molecular Biophysics and CardiovascularDivision, London SE1 1UL, UK. Tel.: 44-0-20-7848-6478; E-mail: Mark.Pfuhl@kcl.ac.uk.

THE JOURNAL OF BIOLOGICAL CHEMISTRY VOL. 286, NO. 31, pp. 27537–27547, August 5, 2011Author’s Choice © 2011 by The American Society for Biochemistry and Molecular Biology, Inc. Printed in the U.S.A.

38

as a result, activation of the Nek2 kinase may be subject tospecific regulation. It is also the first time that the exchange of acoiled-coil domain, indirectly inferred for several completelyunrelated proteins, has been directly observed experimentally.

EXPERIMENTAL PROCEDURES

Sequence Analysis—Domain definitions for Nek2 kinasewere taken from annotations of entry P51955 from the Uniprotdatabase with small modifications for Fig. 1A. Leucine zipperprediction was performed with the 2ZIP server (12). Thecoiled-coil prediction was performed with the COILS server(13). Single helix preference as well as N- and C-caps were cal-culated with the web-based version of AGADIR (14). Patternsearches were performed with the program pattinprot (15).Phosphorylation sites around the Nek2 leucine zipper weretaken from the literature (11, 16). The helical wheel in Fig. 2Bwas initially generated using the EMBOSS (17) application pep-wheel and then adapted to incorporate the two heptad repeats.Limited Proteolysis—SDS-PAGE analysis was performed fol-

lowing a 1:500 trypsin:target (w/w) incubation at 25 °C. Ali-quots were withdrawn at 2, 5, 10, 20, 30, and 60 min. The reac-tion for each time point was immediately halted by the additionof Pefabloc SC (2 mM final) and PSC protector solution (5% v/vfinal) from Roche. Sequencing grade modified trypsin wasobtained from Promega. To identify the protected fragments, a60-min limited digest was performed, as above, and the prod-ucts separated by reverse-phase chromatography using aJASCO HPLC and a 4.5 ml Zorbex stable bond 300 C3 columnat 1 ml/min heated to 55 °C. The column was developed with5–50% acetonitrile, 0.05% TFA (trifluoroacetic acid) pH 1.8 at1% acetonitrile/min. Peaks were collected in 0.5-ml fractionsand monitored at 280, 220, and 210 nm wavelengths. The frac-tions were analyzed by standard electrospray MS procedures.Mass Spectrometry—Protein molecular weight was deter-

mined using a stand-alone syringe pump (Perkin Elmer, FosterCity, CA) coupled to a Platform electrospray mass spectrome-ter (Micromass, Manchester, UK). Samples were desalted on-line using a 2 � 10 mm guard column (Upchurch Scientific,Oak Harbor, WA) packed with 50 micron Poros RII resin (Per-septive Biosystems, Framingham) inserted in place of the sam-ple loop on a rheodyne 7125 valve. Proteins were injected ontothe column in 10% acetonitrile, 0.10% formic acid, washed withthe same solvent and then step-eluted into the mass spectrom-eter in 70% acetonitrile, 0.1% formic acid at a flow rate of 10�l/min. The mass spectrometer was calibrated using myoglo-bin. Standard samples comprised of 100 pmol of protein at aminimum concentration of 1 �M.Construct Generation—All protein expression constructs

were cloned into pETM-11 vectors obtained from the ProteinExpression Laboratory at EMBL Heidelberg. Inserts weregenerated by polymerase chain reaction using 2.5 units ofPlatinum� Pfx DNA Polymerase (Roche) with 200 ng of tem-plate, 500 nM of each primer, 1.2mMdNTPmix, and 1mMMgSO4on a Techne TC-312 thermocycler. Amplification was done byinitial denaturation for 2� at 94 °C followed by 30 cycles of 15�melting at 94 °C, 60� at 60 °C and30� extension at 68 °C. Primersfor constructs LZ0 and LZ5 were LZ05�: GGAGCGCCCATG-GCGCGACAATTAGGAGAG; LZ03�: GGATCCTTATAGC-

AAGCTGTAGTTCTTCACAGATTTTCTGC; LZ55�: GCG-CCCATGGCGGTATTGAGTGAGCTGAAACTG; LZ53�:GGATCCTTAGTCCTCTGCTAGTCTCTCACG, respec-tively. PCR products were purified with QIAquick PCR purifi-cation kit according to the manufacturer’s protocol. PurifiedPCR products and pETM-11 target vector DNA were doubledigested with NcoI and BamHI. The product of the vectordigestion was purified by electrophoresis on a 1% agarose gel(analytical quality, Melford Labs). DNA was extracted fromexcised bands using theQIAquick gel extraction kit. 50 ng of gelpurified digestion product from the vector and 150 ng ofdigested PCR product were mixed and ligated using the rapidDNA ligation kit (Roche). 1/10 of the ligation reaction wastransformed into 100 �l DH5a-T1R chemical competent cells(Invitrogen). Transformed cells were checked for inserts by col-ony PCR.Site-directed Mutagenesis—Point mutations were generated

using the QuikChange kit (Stratagene) using the manufactur-er’s protocol. Mutagenesis primers were C335A: CAGAAAG-AACAGGAGCTTGCAGTTCGTGAGAGACTAG and GTC-TCTCACGAACTGCAAGCTCCTGTTCTTTCTG; K309C:CTGTATTGAGTGAGCTGAAACTGTGTGAAATTCAGT-TACAGGAGCGAGA and TCTCGCTCCTGTAACTGA-ATTTCACACAGTTTCAGCTCACTCAATACAG; E310C:ATTGAGTGAGCTGAAACTGAAGTGTATTCAGTTACA-GGAGCGAGAGC and GCTCTCGCTCCTGTAACTGAA-TACACTTCAGTTTCAGCTCACTCAAT (mutated codonindicated in bold). For all constructs and mutants, small scalecultures were grown for several positive clones, and DNA wasprepared with the Qiagen miniprep kit. Accuracy of vector andinsert was checked by DNA sequencing.Protein Expression and Purification—For protein expression,

miniprep DNA was transformed into BL21* cells (Invitrogen).Transformed cells were grown up in LB medium to OD �0.8when they were induced with 0.75 mM IPTG for 4 h. For 15Nisotope labeling the protocol was modified as suggested (18).Cells were opened using a French Press cell at 1000 psi. Theproteins were purified on fast flow 6 (GE Healthcare) columnsof 2 ml resin, equilibrated as per the manufacturer’s instruc-tions. After loading the samples, the columnswerewashedwith30 ml of wash buffer (20 mM phosphate pH 7.5, 500 mM NaCl,10mM imidazole, 1mM �-ME, 0.02%NaN3) before elutionwith10 ml of elution buffer (as wash buffer, but with 500 mM imida-zole). The eluted protein was incubated with AcTEV prote-ase (Invitrogen) for 2 h at room temperature followed by dialy-sis 3� against 1 liter of fast flow 6 wash buffer. The proteinsolution was applied a second time to the affinity column toremove nonspecific binding proteins. Where required, a finalpolishing step using a Sephadex 16/70 gel filtration column (GEHealthcare) was performed using an AKTA purification sys-tem. Fractions containing the pure protein were pooled andconcentrated in Vivaspin concentrators (Sartorius) with 3 kDamolecular mass cut-off. Protein concentrations were deter-mined using the Qbit fluorescence assay (Invitrogen). Proteinsamples were exchanged into NMR buffer (20 mM sodiumphosphate, 50mMNaCl, pH 7.0, 2mMDTT, 0.02%NaN3) usingPD10 or Nap5 columns (GE Healthcare), which were also usedfor all other experiments. The only exceptions were samples of

The Nek2 Kinase Leucine Zipper

39

disulfide locked LZ5 K309C/C335A, LZ5 E310C/C335A, andLZ0 K309C/C335A for which DTT was omitted from thebuffer.CD spectroscopy. All CD experiments were recorded on a

JASCO700 instrument fittedwith a Peltier temperature controlsystem. Square cuvettes with 0.1 or 1mmpath lengthwere usedwith protein concentrations ranging from 20–200 �M. Spectrawere calibrated using software provided by the manufacturer.Secondary structure content was estimated using homewrittenMathematica (Wolfram Research) macros by comparison tostandard curves for �-helix, �-sheet, and random coil. Tomea-sure thermal unfolding curves, samples were heated at 1 °C/minwhile the CD signal at a constant wavelength of 222 nm wasmeasured. Unfolding curves were fitted to the equation for atwo-state unfolding reaction using a home written Math-ematica macro to extract the melting temperature.Analytical Ultracentrifugation—Analytical ultracentrifuga-

tion (AUC)3 sedimentation velocity experiments were carriedout on a Beckman XL-I centrifuge using an An50-Ti rotor at4 °C and a speed of 42,000 rpm (LZ0) and an An-60 Ti rotor at20 °C and a speed of 60,000 rpm (LZ5 and LZ5 mutants). Scanswere recorded using the interference optical system until nofurther sedimentation occurred. Sample concentrations wereabout 10–200�M in standardNMRbuffers. Protein partial spe-cific volume and buffer density and viscosity were calculatedusing SEDNTERP (19). The experimental data were analyzedusing Sedfit (20) by fitting to the c(s) and c(s,f/f0) models withone discrete component (LZ0) and results for some samplesconfirmed by two-dimensional spectrum analysis, enhancedvan-Holde-Weischet analysis and Genetic Algorithm analysisusing UltraScan (21) confirmed by Monte-Carlo analysis.NMR Spectroscopy—Spectra were recorded at a temperature

of 298 K on Bruker Avance 600 and 800 MHz spectrometersfitted with 5 mm cryoprobes. The HSQC spectrum was used asprovided by the manufacturer with small modifications toincrease safety of the probe. Exchange experimentswere as pre-viously described (22), except modified to increase the disper-sion of the peaks by changing from a 15N-1H view to a 1H-1HNOESY-like view and also by combining both views intoa three-dimensional experiment.4 For qualitative analysisexchange experiments with a mixing time of 64 ms were used,for the quantitative analysis of the exchange rate the followingmixing times were used: 8, 16, 24, 32, 64, 96, and 160 ms. Diag-onal and cross peak intensities for sufficiently well resolved sys-tems of exchanging amide resonances were extracted usingCCPN analysis (23) and fitted to standard equations for slowexchange (22) using a home written Mathematica macro toyield exchange rates and 15N longitudinal relaxation rates. 15Nlongitudinal (R1) and transversal (R2) were also measureddirectly using delays of 16, 48, 96, 192, 288, 384, 512, 704, 880,

1120, and 1440 ms for R1 and 5, 10, 15, 20, 31, 41, 61, 82, 102,133, 154 ms for R2.Sequence-specific assignment of mutant LZ5 K309C/C335A

in non-reducing NMR buffer was based on standard tripleresonance three-dimensional experiments (HNCACB, HN-(CO)CACB, HBHA(CBCACO)NH) recorded on a 0.6 mM15N/13C labeled sample on a Bruker Avance 500 MHz spec-trometer equipped with a cryoprobe combined with a 3D 15Nresolved NOESY spectrum recorded on a 0.8 mM 15N labeledsample on a 700 MHz Bruker Avance spectrometer equippedwith a cryoprobe. The assignment was performed with CCPNanalysis (23).RDCs were measured in the presence of 10 mg/ml of pf1

phage (Hyglos GmbH, Germany) in NMR buffer using a stan-dard IPAP 1H-15N correlation experiment (24). The error asso-ciated with the measured RDC values is � 1.5 Hz.Model Building—A coiled-coil model was generated for the

Nek2 leucine zipper residues 299–341 using a program pro-vided byG.Offer (25). This program generates standard coiled-coils based on the definition of the geometry provided as input.No further energy minimization was performed. Parametersused were pitch � 144 Å, helix radius � 4.7 Å, relative rotationof strands � 210o, residue translation for one residue in thehelix� 1.495Å. The same sequence was used as input formod-els for both heptad repeats. The only difference was the defini-tion of the first A- residue: Leu-306 in HepI and Leu-303 inHepII.RDCs were calculated for the models using PALES (26)

selecting pf1 phage as alignment medium at a concentration of10mg/ml, electrostatic mode, a sodium chloride concentrationof 50 mM and default settings for all other parameters.

RESULTS

The Nek2 Leucine Zipper Resides within a Larger Proteolyti-cally Resistant Fragment—The Nek2A kinase consists of anN-terminal catalytic domain (residues 8–271) followed by aC-terminal regulatory region (residues 272–445) that encom-passes a leucine zipper (residues 305–335) followed immedi-ately by an additional short coiled-coil (residues 340–355) (Fig.1A). To determinewhether the C-terminal region is subdividedinto a particular domain organization, limited proteolysisexperiments were performed on full-length protein (Fig. 1B)and the complete C-terminal regulatory region (Fig. 1C andsupplemental Fig. S1) to identify stable fragments. A �8 kDafragment appeared consistently in both experiments. It wasshown by mass spectrometry to cover not only the leucine zip-per but all of the following coiled-coil and a short section of thelinker connecting it at the N-terminal end to the catalyticdomain (residues 290–360). This 8 kDa fragment was sub-cloned into the pETM-11 expression vector and termed LZ0.For comparison, a series of constructs was generated coveringonly the predicted core leucine zipper. The best behaved ofthese, based on a number of criteria including expression yieldin bacteria, solubility and stability, was termed LZ5 (residues299–343). For all biophysical experiments, LZ0 and LZ5 pro-teins were expressed in bacteria, purified using nickel affinitychromatography followed by removal of the His tag and pol-ished via gel filtration (data not shown).

3 The abbreviations used are: AUC, analytical ultracentrifugation; NOESY,nuclear Overhauser enhancement spectroscopy; HSQC, heteronuclearsingle quantum coherence; TOCSY, total correlation spectroscopy; RDC,residual dipolar couplings; MTBD, microtubule binding domain; HAMP,domain in histidine kinases, adenylyl cyclases, methyl accepting che-motaxis receptors, and phosphatases.

4 F. W. Muskett, R. A. Croasdale, A. M. Fry, and M. Pfuhl, unpublished results.

The Nek2 Kinase Leucine Zipper

40

Sequence Analysis of the Nek2 Leucine Zipper—Meaningfulleucine zipper scores provided by the program 2zip (12) startaround residue 305 and continue to approximately residue 335(Fig. 2A). Coiled-coil scores provided by the COILS software(13) cover the leucine zipper, drop to insignificant levels aroundresidue 335, and then return to a significant level from residue340–355. The leucine zipper is defined by the presence of leu-cine residues every 7 amino acids.However, in the conventionalheptad repeat of a leucine zipper, a second position is also occu-pied by another leucine or equally compatible hydrophobic res-idue such that, in the repeating heptad ABCDEFG, positions Aand D would normally both be occupied by hydrophobic resi-dues. In Nek2, though, one of these hydrophobic positions ismissing, such that the heptad pattern can be positioned in twoways: the leucines can be in the A-position (heptad I) or in theD-position (heptad II). It is traditionally thought that leucineprefers the D-position, but the energetic contribution to coiled-coil stability and the frequency by which leucine is found ineither position in leucine zipper sequences are very similar (27).Regardless of the positions of the heptad repeats in Nek2,charged residues would occupy the second conserved position. Inheptad I, it would be lysine or arginine in the D position, while inheptad II, it would be glutamate in the A position. Curiously,lysine and arginine in general prefer the A position while gluta-mate prefers the D position (27, 28). Even though charged res-idues have been found in the hydrophobic core of other coiled-coils, e.g. myosin, they compromise the stability and make theNek2 leucine zipper a non-ideal coiled-coil. It is interesting tonote the consistent occupation of the normally hydrophobic Aand D positions by charged residues in the leucine zipper ofNek2. Of the 6 A positions in heptad II, 5 are occupied by glu-tamate (the first is occupied by a leucine), while in the case ofheptad I, lysine and arginine each take 3 of the 6 D positions

(Fig. 2). The high degree of conservation of these destabilizingresidues suggests an important function.To determinewhether other proteins contain related leucine

zippers which might allow hetero-oligomerization, similaritysearches were performed with a pattern search program (15)using the pattern LXXR/KEXX repeated five times. No coiled-coil sequences other than that of Nek2 were found in thissearch, even allowing for up to two mismatches. Hence, theprimary function of this motif in Nek2 would appear to be topromote homodimerization rather than heterodimerizationwith a different partner molecule.Circular Dichroism Spectroscopy of the Wild-type Nek2 Leu-

cine Zipper—As a first biophysical approach to understandtheir conformation, CD spectroscopy was performed on theLZ0 and LZ5 polypeptides. The spectra were virtually identicaland typical of coiled-coils with minima at 208 and 222 nm andthe intensity of the 208 nm peak slightly stronger than the 222nm peak (Fig. 3A). The molar ellipticities at 222 nm of approx-imately �22,000 deg mol�1 cm�2 suggested the presence of�70% �-helix, in good agreement with expectation. Meltingcurves showed reasonably cooperative thermal unfolding withrelatively high melting temperatures of 57 °C for LZ5 and aslightly higher value of 66 °C for LZ0 (Fig. 3B). These data showthat the core leucine zipper sequence alone is sufficient toassume a proper fold and that the additional coiled-coil onlyadds extra stability.Analytical Ultracentrifugation—To determine whether the

bacterially expressed LZ0 and LZ5 proteins were dimeric asopposed to higher order oligomers, they were subjected to sed-imentation velocity analytical ultracentrifugation (AUC).Results strongly supported the conclusion that LZ0 and the LZ5proteins predominantly formed dimeric molecules that werestable over the relevant concentration range (tested over

FIGURE 1. Nek2 domain organization and limited proteolysis. A, scheme of full-length Nek2A kinase annotated with functional and structural motifsindicated with their position in the sequence. B, SDS-PAGE and Coomassie Blue analysis of full-length Nek2A kinase subjected to limited proteolysis for thetimes indicated (mins) FL, full-length protein; KD, kinase domain; *, 8 kDa fragment. C, as for B, but using the C-terminal non-catalytic region in the proteolysisassay. CTD, C-terminal domain; *, 8 kDa fragment; Z, nonspecific fragment corresponding to a region of Nek2A lacking tryptic sites.

The Nek2 Kinase Leucine Zipper

41

10–250 �M). As an example, the sedimentation velocity analy-sis is shown for LZ0 (Fig. 4A) with the main peak indicating amolecular weight within 500 Da of the calculated molecularmass of the dimer at 17,468 Da. All parameters are listed inTable 1. Although the peaks are not perfectly symmetrical, itcan be concluded that the leucine zipper of the Nek2 kinase,with or without the extra coiled-coil portion, forms a dimer asthe predominant species. Under the conditions of these exper-iments there was no sign of higher order oligomers.NMR Spectroscopy of the Wild-type Nek2 Leucine Zipper—

Early NMR studies of constructs of the Nek2 leucine zippershowed unusual “twins” of resonances in homonuclear NOESYand TOCSY spectra suggesting the presence of multiple formsof the protein (data not shown). A more detailed analysis ofpotential isoformswas therefore performed using a 15N-labeledsample of LZ5. In an HSQC experiment, automatic peak pick-ing with CCPN analysis (23) gave75 peaks, significantlymorethan the expected �40 peaks, confirming the presence of atleast two forms of the protein (Fig. 5A). Very similar spectrawere obtained for LZ0, although as the quality of the spectrawas much inferior (Fig. 3) all subsequent NMR work was doneon LZ5.To establish if the isoformswere in dynamic equilibrium, two

NMR exchange experiments were recorded (22) via transfer to15N. The resulting two-dimensional spectrum can then take the

shape of an HSQC experiment by frequency labeling the nitro-gen in t1 (Fig. 5B), or the appearance of a NOESY experimentwith frequency labeling of the amide proton in t1 (Fig. 5C).Both experiments clearly demonstrated that a substantial

number of resonances in the Nek2 leucine zipper undergo slowexchange on the chemical shift time scale (Fig. 5, B and C). Tofurther improve the identification of exchange cross peaks,both versions of the exchange experiment were combined intoa three-dimensional version that has the appearance of a three-dimensional 15N NOESY-HSQC. A few representative slicesare shown in Fig. 5D. Using this three-dimensional exchangespectrum it was possible to identify more than 34 exchangingpairs of resonances suggesting that essentially the entire pro-tein is subject to exchange and that only two isoforms of theprotein exist in solution. Comparison of the diagonal peakintensities of selected well-resolved exchange pairs suggeststhat both forms of the protein exist in almost equal abundance.A quantitative analysis of the exchange was performed by

fitting cross and diagonal peak intensities of pairs of exchangingresonances from a series of exchange experiments with differ-ent mixing times to yield the exchange rate (kex) and the 15Nlongitudinal relaxation rate (R1) (22). An example for the fit isshown in Fig. 5E for one of the four exchanging pairs that wereanalyzed. The values obtained for kex, 18.2, 17.4, 16.9, 17.2 s�1,were well within the error of the fitting procedure so that they

FIGURE 2. Sequence analysis of the Nek2 leucine zipper. A, sequence of the Nek2 leucine zipper is shown color coded by properties (green, hydrophobic; red,negatively charged; blue, positively charged; magenta, polar uncharged; yellow, cysteine). Below the sequence, the two heptad repeats, HepI and HepII, areshown with the key hydrophobic positions, A and D, marked in uppercase. The leucine zipper prediction by the program 2ZIP is then shown (Lzip) followed bythe coiled-coil prediction by the program COILS (cc). The next three lines are provided by the program AGADIR: probability for folding as isolated �-helix(agadir) as well as putative N- and C-caps and finally predicted serine phosphorylation sites (S-PO). B, helical wheel plots of the core leucine zipper residues ofNek2. Amino acids shown in boxes are colored as in A. Heptad repeat positions are indicated in circles for the two possible conformations, HepI and HepII.

The Nek2 Kinase Leucine Zipper

42

were averaged to give a single rate constant of 17.4 � 1.7 s�1

assuming a single, cooperative exchange event. 15N R1 valuesobtained from the same fit for the four systems were 1.5, 1.7,1.4, and 1.6 s�1, averaged to 1.6 � 0.3 s�1. Fitting the peakintensities of resolved diagonal peaks directly to an exponentialdecay to obtain the apparent 15N R1 gave values of 14.9, 16.8,15.3, 16.1 s�1 which averaged to 15.8 � 1.6 s�1, approximatelyten times the nominal value. These dynamics values were com-plemented by the direct measurement of 15N R1 and R2 values.For the four well defined resonances R1 values of 14.1, 15.9,14.6, 15.3 s�1 and R2 values of 29.7, 30.8, 29,9 and 31.1 s�1 wereobtained giving averages of 15.0 � 0.7 s�1 and 30.4 � 0.8 s�1,respectively.To ensure that the residues undergoing exchange were not

doing so because they are unstructured, a three-dimensional15N NOESY-HSQC experiment was also recorded and severalrepresentative slices are shown superimposed on the exchangespectrum (Fig. 5D). It is apparent that for the majority of reso-nances with exchange cross peaks there are genuine amide-amide sequential NOE cross peaks, typical for �-helices. Thus,the exchange is not a result of unfolding events.Probing the LeucineZipperConformational Exchange by Site-

directed Mutagenesis—The existence of two conformations inall recombinant versions of the leucine zippermade interpreta-tion of NMR spectra very difficult. This was due, firstly, toextensive overlap of peaks: the chemical shift dispersion of

coiled-coils is notoriously poor and so was made worse by hav-ing two highly similar versions of the same protein in the sam-ple. Furthermore, although the chemical exchange is slow onthe chemical shift time scale, it is very close to the transitionregion toward intermediate exchange. As a result, both longi-tudinal and transversal relaxation rates of protons and nitro-gens are significantly accelerated, making the recordingof complex three-dimensional experiments required forsequence specific assignment and structure calculation vir-tually impossible.It was therefore decided to employ site-directedmutagenesis

to probe the LZ structure and dynamics. For this, we hypothe-sized that the conformational dynamics might result fromexchange between the two alternative heptad repeats describedearlier (see Fig. 2). To test this hypothesis, individual arginine/lysine or glutamate positions in the 2nd or 3rd heptad repeatswere mutated to cysteine. The intrinsic cysteine of the leucine

FIGURE 3. Comparison of the leucine zipper constructs LZ5 (blue) and LZ0(red). A, CD spectrum at T � 298 K, concentration 50 �M, path length 1 mm.B, thermal melting curve measured at � � 222 nm. Melting temperaturesobtained from fits to a two state unfolding equation are 57.8 � 0.2 °C for LZ5and 66.6 � 0.2 °C for LZ0. C, superposition of 15N HSQC experiments recordedat 600 MHz and T � 298K.

FIGURE 4. Sedimentation velocity analytical ultracentrifugation. c(s) dis-tributions for (A) 228 �M LZ0 (B) LZ5 (blue) and LZ5 K309C/C335A at a con-centration of 224 �M in reducing (red) and non-reducing conditions (green).LZ5 wt is shown in blue, LZ5 K309C/C335A in reducing buffer in red and innon-reducing buffer in green. For details on the data analysis see “Experimen-tal Procedures.” For full results see Table 1.

TABLE 1Sedimentation velocity results for the leucine zipper constructsemployed in this study

Construct Sapp S20,w F/F0Monomer

MWApparent

MWS S kDa kDa

LZ0, 228 �M 0.86 1.40 1.7 8.7 17.0LZ0, 29 �M 0.89 1.40 1.7 8.7 17.7LZ5 1.10 1.11 1.5 5.4 10.7LZ5 C335A/K309C reduced 1.07 1.08 1.5 5.4 10.0LZ5 C335A/K309C oxidized 1.05 1.06 1.6 5.4 10.1

The Nek2 Kinase Leucine Zipper

43

zipper, Cys-335, was replaced by alanine to avoid interference.Assuming that theNek2 leucine zipper folds as a parallel coiled-coil, a disulfide bridge can form between cysteines in eitherthe A-position (mutation from glutamate) or the D-position(mutation from lysine/arginine) once the sample is in non-re-ducing buffer. Disulfide bonds in coiled-coils have beenobserved to occur naturally, e.g. in myosin (29) and were shown

to contribute to the stability of the coiled-coil in synthetic leu-cine zippers (30).Based on this hypothesis, one would predict that a glutamate

to cysteine or lysine/arginine to cysteine mutant under non-reducing conditions should exist in only one conformation andshould not exchange. Should it be possible to obtainNMRspec-tra of conformationally “locked” mutants then, at least in gen-

FIGURE 5. NMR spectra of LZ5. A, HSQC spectrum at T � 278K, 600 MHz. B, 15N exchange experiment shown as HSQC (red), mixing time 60 ms, superimposedon HSQC (blue). Connections of exchanging species in the HSQC spectrum via the cross peaks in the exchange spectrum are shown by black boxes for a fewresidues. C, 15N exchange experiment shown as NOESY, mixing time 60 ms. The exchange for a number of residues is illustrated by red squares. D, superpositionof 1H-1H slices of the three-dimensional 15N exchange experiment (64 ms mixing time) with a 15N NOESY-HSQC (100 ms mixing time). Contours for the NOESYare in red, contours for the exchange experiment in blue. E, time dependence of exchange (red, green) and diagonal peak (blue, magenta) intensity of a selectedresidue (boxes) compared with the results of the fits (continuous lines). Fitting errors are shown below.

The Nek2 Kinase Leucine Zipper

44

eral, it should be possible to reconstruct the spectrum of thewild-type protein from the spectra of one pair of cysteinemutants, representing the heptad I or heptad II conformation.Initially, experiments were attempted with an LZ5-E310C/C335A mutant that should lock the protein in the heptad IIconformation with the acidic residues in the A position. How-ever, CD spectra showed only a modest amount of �-helix atlower temperatures, regardless of the oxidation state of thebuffer, while AUC data of this mutant could not be interpretedand 15N HSQC spectra were even more complex than thoseobtained with the wild-type protein (data not shown). All ofthese data suggest that the mutant is unfolded or at best onlypartially folded.We therefore prepared an LZ5-K309C/C335A mutant to

lock the heptad I conformation with the basic residues in the Dposition. A preliminary analysis of this mutant by CD showedthat oxidation of the cysteines to form the expected disulfidebond led to a substantial increase in the �-helix content (Fig.6A). This reflects the fact that a reduced cysteine in position Ddestabilizes the leucine zipper (30). The reduced form also hada much lower stability compared with the oxidized form asillustrated by the thermal unfolding characteristics. The melt-ing temperature of the reduced form was around 27 °C, while

the oxidized form melted at 55 °C (Fig. 6B). While the CDexperiments did not provide information on the proteindynamics, they did suggest that themutated protein folds into aparallel coiled-coil. This was supported by AUC sedimentationvelocity data indicating the existence of a dimer in non-reduc-ing buffer (Table 1). Fig. 4B shows the c(s) distributions for LZ5and the double mutant K309C/C335A in reducing and non-reducing conditions. Again, there is one main peak with fittingparameters suggesting a molecular weight close to that of thedimer over the entire concentration range tested.To analyze the effect of disulfide bond formation on the

exchange dynamics, a HSQC spectrum of the mutant LZ5K309C/C335A was recorded in non-reducing buffer (Fig. 7A).It was apparent that the total number of peaks was significantlyless than in a spectrum of wild-type LZ5. Automatic peak pick-ing produced a total of just over 40 peaks, very close to thepredicted number of 45. This suggests that this sample con-tained only a single conformation of the leucine zipper. A directcomparison of the C335A and K309C/C335A mutants of LZ5revealed the similarity of the spectra (Fig. 7B, supplemental Fig.S2). In the resolved region on the left, it was particularly appar-ent that, as predicted, one peak of an exchanging pair (indicatedby brackets) vanished leaving only its partner behind. It wasapparent that the peaks were much sharper than in the HSQCof the wild-type protein, also hinting at a significant change inthe dynamics of the protein. This observationwas confirmed bya NOESY-type exchange experiment which had no exchangecross peaks at all (Fig. 7C). The good quality of the spectraallowed us to obtain an almost complete backbone and partial

FIGURE 6. CD spectroscopy of the LZ5 K309C/C335A mutant. A, CD spec-trum at T � 278 K. The spectrum of the reduced protein is shown in red, thespectrum of the oxidized protein is shown in blue. B, thermal unfoldingcurves. The measured data is shown in red boxes for the reduced sample andin blue boxes for the oxidized sample. The fitted curves are shown for bothsamples as thin black lines. Melting temperatures obtained from a fit to a twostate unfolding equation are 57.7 � 0.3 °C for the oxidized sample and 24.3 �0.3 °C for the reduced sample.

FIGURE 7. NMR spectra of the LZ5 K309C/C335A mutant. A, 15N HSQC ofLZ5-K309C/C335A, oxidized, T � 298K. B, same as in A superimposed on thespectrum of wild-type LZ5. Two exchange pairs are indicated by brackets inthe well resolved part of the spectrum of the wild-type protein B: 15Nexchange experiment shown in NOESY view of oxidized LZ5 K309C/C335Asuperimposed on the same experiment for wild-type LZ5. In both B and C thespectrum of wild-type LZ5 is red, for LZ5-K309C/C335A blue.

The Nek2 Kinase Leucine Zipper

45

side chain assignment (deposited with the BMRB, accessioncode 17417). This was used to collect an extensive set of sec-ondary structure specific proton-proton distances which sup-port thewell defined helical structure (supplemental Fig. S6). Inaddition, 1H-15N residual dipolar couplings were recorded andcompared with RDC values calculated for models created forboth conformers shown in Fig. 8A. As can be seen in Fig. 8Cthere is good agreement of the experimental data with thosepredicted for theHepImodel (axial componentDa� 9.33 10�4,rhombic componentDr� 1.80 10�4, correlation coefficient r�0.84,Q-value 0.21)while very little agreement is seenwith thosepredicted for HepII (axial component Da� �3.20 10�3, rhom-bic component Dr � �2.06 10�3, correlation coefficient r �0.34,Q-value 37.3). Taken together, these data strongly supportthe hypothesis that the LZ5 K309C/C335A mutant representsone of the conformers observed for thewild type leucine zipper.

DISCUSSION

Unusual slow chemical exchange between two conforma-tions with a rate of 17 s�1 was observed for the Nek2 leucinezipper in solution. This exchange was observed regardless ofconstruct, sample or any other experimental conditions (sup-plemental Fig. S4–6) so that it cannot be dismissed as an arti-fact but has to be seen as a genuine property of the isolateddomain thatmaywell hold true for the full-length kinase in vivo.Examples of dynamics in leucine zippers based on experi-

mental data are few and far between. A modest level of dynam-ics, albeit fast on the chemical shift timescale, was inferred froma comparison of the NMR data with the crystal structure of thefirst well characterized leucine zipper, GCN4 (3, 31) and similarobservations were made in the leucine zipper of c-Jun (32, 33).These were based on the presence of an asymmetric conforma-tion of a central asparagine in the crystal structure while only asingle signal was seen in the NMR spectra suggesting fast aver-aging in solution.Slow chemical exchange between folded conformations for a

protein involving a leucine zipper was only ever observed once,in the case of the tetramerization domain of the Mnt repressor(34). The tetrameric leucine zipper exists in two different con-formations which are generated by a staggered assembly of twocoiled-coil dimers that are packed into a distorted four helicalcoiled-coil. Such a complex assembly is well beyond the capa-bilities of the small dimeric Nek2 leucine zipper.The best explanation for the unusual dynamics of the Nek2

leucine zipper would therefore seem to be the existence of twocoiled-coils based on a shift in the heptad register (Fig. 8A)where the charged side chains in the A/D positions minimizeadverse interactions by extending outside the hydrophobic corewith their charged groups (Fig. 8B). Determination of the struc-ture by experimental methods is very challenging and efforts tocrystallize the protein may well fail as a result of these inherentdynamics. Also standard NMR methods cannot be used toobtain structural information because of the detrimental effectscaused by the exchange rate of 17 s�1. Even though it is slow onthe chemical shift timescale, it is still sufficiently fast to havedetrimental effects inNMR experiments. As a consequence it isimpossible to record triple resonance three-dimensional exper-iments required for complete assignments.

We were therefore forced to use an indirect approach basedon its unusual sequence to investigate the dynamics of theNek2leucine zipper. As this sequence contained only one leucineevery seven amino acids, there were twoways of positioning theheptad (Figs. 2A and 8A). In essence, both heptads satisfied thedefinition of a leucine zipper albeit with charged residues inplaces normally occupied by hydrophobic amino acids. A sim-ilar arrangement has been previously described in the myosincoiled-coil (29), where polar residues in these positions havelong side chains so that the charged group is at least partlysolvent accessible, while the long aliphatic portion can makesome contribution to the hydrophobic core as shown in Fig. 8B.The downside of such an arrangement is unfavorable side chaintorsion angles that compromise the stability of such leucinezippers.To provide experimental proof for the twoheptad hypothesis

we decided to lock the protein in one of the two conformationsusing a disulfide bond. Two mutants were generated in LZ5,K309C to lock heptad I and E310C to lock heptad II, both asdouble mutants with C335A to avoid interference. MutantK309C/C335A in non-reducing buffer performed verymuch asexpected: it stopped the exchange and reduced the number ofpeaks in the HSQC to that expected for a simple leucine zipper.Moreover, the line widths of the peaks were now more consis-tent with proteins of this molecular weight. The fact that thesepeaks did not exactlymatch the peaks in thewild type spectrumcan be explained by the presence of a double mutant and thesignificant changes in environment introduced by the disulfidebond. A similar, though less apparent, effect of the formation ofthe disulfide bond in theK309C/C335Adoublemutantwas alsoseen in LZ0 (supplemental Fig. S6). In addition to supportingour assumption that the K309C/A335C mutant representsthe HepI conformation this result provides indirect evidencefor the distinct charge distribution on the surface of the twoconformers. Their overall shape is virtually identical (Fig.8A) so that not much of a difference could be expected iftheir alignment was purely steric. Using pf1 phage as align-ment medium, however, means that the alignment isstrongly driven by electrostatic interactions, which allows todifferentiate the two different conformers as suggested pre-viously (35, 36).In contrast, mutant E310C/C335A did not give a clear cut

result in agreement with previous results. The disulfide bondfor residue Cys-310 would be in the A position which wasshown to be destabilizing rather than stabilizing unlesslocated right at the N or C terminus (30). As the Nek2 leucinezipper is already less than ideal, the modest destabilizationby a disulfide bond in the A position could compromise theentire structure.While it would be preferable to have at least one mutant to

lock each conformation,we believe that these data are sufficientto provide support for the hypothesis that the slow exchangeseen in theNMRspectra of theNek2 leucine zipper results fromthe interconversion of the two possible heptad arrangements.The actual conformational change would be a simple rotationof each helix by �20o about its long axis. The relationshipbetween degree of structural change required and time scale ofexchange matches well the two examples in the literature: the

The Nek2 Kinase Leucine Zipper

46

rearrangement in the Mnt tetramerization domain is substan-tial which is reflected in the slow exchange rate (1 s�1) whencompared with the Nek2 leucine zipper (17 s�1), which in turnis much slower than the simple rotation of the asparagine side

chain in GCN4 and c-Jun, which happens on a fast timescale(1000 s�1).

The populations of the two conformations as judged by thepeak intensities in the HSQC spectra is not far from 50:50, sug-

FIGURE 8. A, model of the proposed conformations of the Nek2 leucine zipper. The leucine zipper for both conformations is shown with the N terminus at thetop and the C terminus at the bottom. The main chain is shown as a helical scheme, the side chains are shown as stick models. No hydrogens are shown. Sidechains are colored according to their properties as in Fig. 2. Residues selected for easy visibility are labeled to provide guidance. Arrows indicate the position ofArg-116 in HepI and Glu-317 in HepII where the slices shown in B are taken. Residues Lys-309 and Cys-335, which are mutated, are labeled in bold and the sidechains encircled in red. B, packing of charged side chains in the A position of HepI (Arg-316) and the D position in HepII (Glu-317) in a slice through the centerof the coiled-coil as marked by arrows in A. C, validation of the model by measurement of 1H-15N RDC values for mutant K309C/C335A of LZ5. Measured RDCson the x-axis are compared with predicted RDCs (on the y-axis) calculated with PALES for the two models.

The Nek2 Kinase Leucine Zipper

47

gesting that the exchange, at least in vitro, does not serve to shiftthe equilibrium toward one or other of the conformations.Nev-ertheless, the surfaces, in particular the charge distribution, ofboth conformations are very different: heptad I has a highlypositive charge density in the center and negative charges onthe edges, while heptad II has negative charges on the inside andpositive charges on the ridges, indirectly supported by the RDCdata (Fig. 8A). As a result the conformational exchange could bemodulated by the interaction of Nek2 with partner proteins orvia phosphorylation, which has been shown to occur on serines299 and 300 in a cell cycle-dependent manner (16).An externally triggered shift in the register of a heptad repeat

has been suggested recently for two completely different pro-teins, the microtubule binding domain (MTBD) of dynein (37,38) and the HAMP domain that is found in a range of trans-membrane proteins (39). In the case of MTBD it is hypothe-sized that binding to microtubules is the initial trigger and thatthe heptad repeat shift is used to transmit this signal to thedistant AAA domain ring structure to activate the ATPaseactivity. For both these proteins indirect data based mainly onsite directed mutagenesis was used to support this hypothesiseven though only one form of each protein was observed. Thelatter is possibly due to the fact that the heptad repeat sequenceconforms much more to the standard than is the case withNek2. As a result, it is therefore not surprising to see a heptadrepeat shift as a tool for signal transmission within a protein.The leucine zipper of Nek2 kinase is simply a more extremecase than MTBD and HAMP, which makes it the first proteinwhere this shift could be observed experimentally. In eukary-otic cells coiled-coil proteins are often part of highly dynamicstructures and we expect that our findings will stimulate newapproaches to understanding cellular regulation. This willmake the Nek2 leucine zipper a powerful model system for theexperimental study of this novel mechanism for the regulationof protein function.

Acknowledgments—We thank the Protein Expression Laboratory atEMBL Heidelberg for a gift of pETM-11 plasmid, Gerald Offer formaking his coiled-coil modeling program available to us, AndrewAtkinson for help with the IPAP experiment, and Ragini Gosh for helpwith the AUC experiments.

REFERENCES1. Landschulz,W. H., Johnson, P. F., andMcKnight, S. L. (1988) Science 240,

1759–17642. Oas, T. G., McIntosh, L. P., O’Shea, E. K., Dahlquist, F. W., and Kim, P. S.

(1990) Biochemistry 29, 2891–28943. Saudek, V., Pastore, A., Castiglione, M. A., Frank, R., Gausepohl, H., Gib-

son, T., Weih, F., and Roesch, P. (1990) Protein Eng. 4, 3–104. O’Shea, E. K., Klemm, J. D., Kim, P. S., and Alber, T. (1991) Science 254,

539–5445. Alber, T. (1992) Curr. Opin. Genet. Dev. 2, 205–2106. Adamson, J. G., Zhou, N. E., and Hodges, R. S. (1993) Curr. Opin. Biotech-

nol. 4, 428–4377. Mason, J. M., and Arndt, K. M. (2005) ChemBioChem 5, 170–1768. Mukai, H., and Ono, Y. (1994) Biochem. Biophys. Res. Commun. 199,

897–9049. Fry, A. M. (2002) Oncogene 21, 6184–619410. Fry, A. M., Arnaud, L., and Nigg, E. A. (1999) J. Biol. Chem. 274,

16304–1631011. Rellos, P., Ivins, F. J., Baxter, J. E., Pike, A., Nott, T. J., Parkinson, D.M., Das,

S., Howell, S., Fedorov,O., Shen,Q. Y., Fry, A.M., Knapp, S., and Smerdon,S. J. (2007) J. Biol. Chem. 282, 6833–6842

12. Bornberg-Bauer, E., Rivals, E., and Vingron, M. (1998) Nucleic Acids Res.26, 2740–2746

13. Lupas, A., Van Dyke, M., and Stock, J. (1991) Science 252, 1162–116414. Munoz, V., and Serrano, L. (1994) Nat. Struct. Biol. 1, 399–40915. Bucher, P., Karplus, K.,Moeri,N., andHofmann,K. (1996)Comput. Chem.

20, 3–2316. Daub, H., Olsen, J. V., Bairlein,M., Gnad, F., Oppermann, F. S., Korner, R.,

Greff, Z., Keri, G., Stemmann, O., and Mann, M. (2008) Mol. Cell 31,438–448

17. Rice, P., Longden, I., and Bleasby, A. (2000) Trends Genet. 16, 276–27718. Marley, J., Lu, M., and Bracken, C. (2001) J. Biomol. NMR 20, 71–7519. Laue, T. M., Shah, B. D., Ridgeway, T. M., and Pelletier, S. L. (1992) in

Ultracentrifugation in Biochemistry and Polymer Science (Harding, S. E.,Rowe, A. J., andHorton, J. C., eds) pp. 90–125, Royal Society of Chemistry,Cambridge, U.K.

20. Schuck, P. (2000) Biophys. J. 78, 1606–161921. Demeler, B. (2005) UltraScan A Comprehensive Data Analysis Software

Package for Analytical Ultracentrifugation Experiments, In Modern Ana-lytical Ultracentrifugation: Techniques and Methods, (Scott, J., Harding,S. E., and Rowe, A., eds) pp. 210–229, Royal Society of Chemistry (UK)

22. Farrow, N. A., Zhang, O., Forman-Kay, J. D., and Kay, L. E. (1994)J. Biomol. NMR 4, 727–734

23. Vranken, W. F., Boucher, W., Stevens, T. J., Fogh, R. H., Pajon, A., Llinas,M., Ulrich, E. L., Markley, J. L., Ionides, J., and Laue, E. D. (2005) Proteins59, 687–696

24. Ottiger, M., Delaglio, F., and Bax, A. (1998) J. Magn. Reson. 131, 373–37825. Offer, G., and Sessions, R. (1995) J. Mol. Biol. 249, 967–98726. Zweckstetter, M. (2008) Nat. Protoc. 3, 679–69027. Tripet, B., Wagschal, K., Lavigne, P., Mant, C. T., and Hodges, R. S. (2000)

J. Mol. Biol. 300, 377–40228. Straussman, R., Ben-Ya’acov, A., Woolfson, D. N., and Ravid, S. (2007) J.

Mol. Biol. 366, 1232–124229. Blankenfeldt,W., Thoma, N. H.,Wray, J. S., Gautel, M., and Schlichting, I.

(2006) Proc. Natl. Acad. Sci. U.S.A. 103, 17713–1771730. Zhou, N. E., Kay, C. M., and Hodges, R. S. (1993) Biochemistry 32,

3178–318731. Saudek, V., Pastore, A., Morelli, M. A., Frank, R., Gausepohl, H., and Gib-

son, T. (1991) Protein Eng. 4, 519–52932. Junius, F. K., Mackay, J. P., Bubb, W. A., Jensen, S. A., Weiss, A. S., and

King, G. F. (1995) Biochemistry 34, 6164–617433. Junius, F. K., O’Donoghue, S. I., Nilges, M., Weiss, A. S., and King, G. F.

(1996) J. Biol. Chem. 271, 13663–1366734. Nooren, I. M., Kaptein, R., Sauer, R. T., and Boelens, R. (1999)Nat. Struct.

Biol. 6, 755–75935. Zweckstetter, M., Schnell, J. R., and Chou, J. J. (2005) J. Am. Chem. Soc.

127, 11918–1191936. Schnell, J. R., Zhou, G., Zweckstetter, M., Rigby, A. C., and Chou, J. J.

(2005) Protein Sci. 14, 2421–242837. Carter, A. P., Garbarino, J. E., Wilson-Kubalek, E. M., Shipley, W. E., Cho,

C., Milligan, R. A., Vale, R. D., and Gibbons, I. R. (2008) Science 322,1691–1695

38. Gibbons, I. R., Garbarino, J. E., Tan, C. E., Reck-Peterson, S. L., Vale, R. D.,and Carter, A. P. (2005) J. Biol. Chem. 280, 23960–23965

39. Hulko, M., Berndt, F., Gruber, M., Linder, J. U., Truffault, V., Schultz, A.,Martin, J., Schultz, J. E., Lupas, A. N., and Coles, M. (2006) Cell 126,929–940

The Nek2 Kinase Leucine Zipper

48

Interaction of Actin with Carcinoembryonic Antigen-relatedCell Adhesion Molecule 1 (CEACAM1) Receptor in LiposomesIs Ca2�- and Phospholipid-dependent*□S

Received for publication, March 2, 2011, and in revised form, June 7, 2011 Published, JBC Papers in Press, June 13, 2011, DOI 10.1074/jbc.M111.235762

Rongze Lu‡1, Michiel J. M. Niesen§1, Weidong Hu§1, Nagarajan Vaidehi§, and John E. Shively§2

From the ‡Irell and Manella Graduate School of Biological Sciences and the §Department of Immunology, Beckman ResearchInstitute of City of Hope, Duarte, California 91010

The regulation of binding of G-actin to cytoplasmic domainsof cell surface receptors is a common mechanism to controldiverse biological processes. Tomodel the regulation of G-actinbinding to a cell surface receptor we used the cell-cell adhesionmolecule carcinoembryonic antigen-related cell adhesion mol-ecule 1 (CEACAM1-S) in which G-actin binds to its short cyto-plasmic domain (12 amino acids; Chen, C. J., Kirshner, J., Sher-man, M. A., Hu, W., Nguyen, T., and Shively, J. E. (2007) J. Biol.Chem. 282, 5749–5760). A liposome model system demon-strates that G-actin binds to the cytosolic domain peptide ofCEACAM1-S in the presence of negatively charged palmitoyl-oleoyl phosphatidylserine (POPS) liposomes and Ca2�. In con-trast, no binding of G-actin was observed in palmitoyl-oleoylphosphatidylcholine (POPC) liposomes orwhen a key residue inthe peptide, Phe-454, is replaced with Ala. Molecular Dynamicssimulations on CEACAM1-S in an asymmetric phospholipidbilayer showmigration of Ca2� ions to the lipid leaflet contain-ing POPS and reveal two conformations for Phe-454 explainingthe reversible availability of this residue for G-actin binding.NMR transverse relaxation optimized spectroscopic analysis of13C-labeled Phe-454 CEACAM1-S peptide in liposomes plusactin further confirmed the existence of twopeptide conformersand the Ca2� dependence of actin binding. These findingsexplain how a receptor with a short cytoplasmic domain canrecruit a cytosolic protein in a phospholipid and Ca2�-specificmanner. In addition, this model system provides a powerfulapproach that can be applied to study other membrane proteininteractions with their cytosolic targets.

Monomeric G-actin is recruited to the plasma membrane inresponse to a variety of cell receptor activation signals (1–4).Although much is known about subsequent steps, includingactin polymerization, branching, and participation in functionssuch as cell motility, less is known about the regulation of theinitial G-actin recruitment event. Among the many cell surfacereceptors that bindG-actinwhen activated, the so-called “short

form” of CEACAM13 stands out as a rather simple example inthat it is a single-pass transmembrane protein with a cytoplas-mic domain of only 12 amino acids. When a single amino acidphenylalanine 454 in the cytoplasmic domain is mutated to ala-nine (F454A), it no longer binds G-actin in in vitro assays, andwhen transfected intoMCF7 cells that form a lumen with wild-type CEACAM1-S, it no longer forms a lumen in three-dimen-sional culture (5). Intrigued by the ability of such a short stretchof amino acids to convey G-actin binding in response to thehomotypic cell-cell interaction function of CEACAM1, wespeculated that the adjacentmembrane environment and Ca2�

signaling may play a role in regulating the binding, otherwisebinding would be constitutive and irreversible. Given the closeproximity of the cytoplasmic domain of CEACAM1 to the lipidbilayer and the inherent asymmetry of the lipid bilayer withrespect to charge, we propose that negatively charged phospho-lipids would attract Ca2� in response to cell-cell interactionsbecause Ca2� signaling almost always accompanies cell-cellinteractions (6, 7). Furthermore, Ca2� recruited to the nega-tively charged inner leaflet of the plasma membrane wouldeffectively promote the interaction of the cytoplasmic domainof CEACAM1 with G-actin. To test this hypothesis, we gener-ated an in vitro model of the peptide-actin interaction in thecontext of negatively charged liposomes in the presence orabsence of Ca2� and analyzed their interactions by a combina-tion of fluorescent bead analysis, NMR TROSY experiment,molecular dynamics (MD) simulation, and surface plasmonresonance analysis.

EXPERIMENTAL PROCEDURES

Preparation of Liposome-coatedGlass Beads—Palmitoyl-ole-oyl phosphatidylserine (POPS) and palmitoyl-oleoyl phos-phatidylcholine (POPC) powders (Avanti Polar Lipids) werestored in chloroform at 25mg/ml. Five�l of stock phospholipidsolution in a glass tube was vaporized under argon gas to from athin and even membrane on the glass tube. Phospholipid wasdissolved in 100 �l of PBS, vortexed, and subjected to extrusionthrough a 200-nm membrane 11–13 times (Avanti PolarLipids).

* This work was supported, in whole or in part, by National Institutes of HealthGrant CA 84202.

□S The on-line version of this article (available at http://www.jbc.org) containssupplemental Tables S1 and S2, Figs. S1–S6, and an NMR Discussion.

1 These authors contributed equally to this work.2 To whom correspondence should be addressed: Dept. of Immunology,

Beckman Research Institute of City of Hope, 1450 East Duarte Rd., Duarte,CA 91010. Tel.: 626-359-8111 (ext. 62601); E-mail: jshively@coh.org.

3 The abbreviations used are: CEACAM1, carcinoembryonic antigen-relatedcell adhesion molecule1; CEACAM1-S, short isoform of CEACAM1; MD,molecular dynamics; MUA, mercaptoundecanoic acid; POPC, palmitoyl-oleoyl-phosphatidylcholine; POPS, palmitoyl-oleoyl-phosphatidylserine;TROSY, transverse relaxation optimized spectroscopy.

THE JOURNAL OF BIOLOGICAL CHEMISTRY VOL. 286, NO. 31, pp. 27528 –27536, August 5, 2011© 2011 by The American Society for Biochemistry and Molecular Biology, Inc. Printed in the U.S.A.

49

Glass microspheres (4.5 �m; Bangs Laboratories, Inc.) werecleaned in “piranha solution” (30% H2O2 and concentratedH2SO4 1/4 (v/v) for 3 min at 80 °C, use extreme caution!) fol-lowed by extensive washing with water on a clean sintered glassfunnel. Clean glass beads were diluted in water (108 beads/�l)and stored at 4 °C. One to 2 �l of suspended glass beads wereincubated with 100 �l of liposome at 37 °C for 2 h. Liposome-coated glass beads were centrifuged at 4000 rpm for 1 min andwashed three timeswith PBS to remove unboundphospholipid.Glass bead-supported phospholipid membranes were incu-bated with MUA (mercaptoundecanoic acid)-CEACAM1-Speptide (0.2 mg/ml) for 1 h at 37 °C and collected by centrifu-gation. Peptide-liposome-coated glass beads further incubatedwith 1% BSA in PBS for 1 h and washed three times using 1%BSA/PBS.G-actin (100�g/ml) or rhodamine-actin (100�g/ml)in the presence or absence of 1 mM CaCl2 was incubated withglass beads for 1 h. In the last step, FITC-conjugated anti-actinantibody was incubated with glass beads. Flow cytometry wasused to detect the florescence of rhodamine-actin or FITC-conjugated anti-actin antibody on a BD FACS caliber. Singlebeads were selected for analysis based on forward and side scat-ter analysis (�80%).Surface Plasmon Resonance Assay—Surface plasmon reso-

nance assays were performed on a Biacore T100 (BIAcoreAB)instrument. Liposome or liposome-containing CEACAM1-SMUA-peptide was prepared using the same method as above.The HPA sensor chip (GE Healthcare) was washed with 40 mM

�-octylglucoside at a flow rate of 5�l/min for 5min. Liposomescomposed of POPS (2 mg/ml) containing CEACAM1-S MUA-peptide (0.2 mg/ml) in PBS were injected at flow rate of 1�l/min. Unbound lipid was washed by increasing the flow rate.One percent BSA in PBS was injected to block the sensor chipsurface. Actin (0.2 mg/ml) was injected plus or minus 1 mM

CaCl2 at a flow rate of 20 �l/min. 0.5 M EDTA (10 �l) was usedto regenerate the HPA sensor chip surface. Liposomes withoutCEACAM1-S MUA-peptide were immobilized to HPA sensorchip and used as a blank.MD System Preparation—A molecular model of the trans-

membrane and cytoplasmic domain of CEACAM1-S derivedpreviously using homology modeling methods was used in thisstudy (5). The F454A mutant peptide was generated by usingthe side chain coordinates of the C� of Ala-454 in place of Phe-454. The CEACAM1-S peptide was placed in a simulation boxwith preequilibrated POPS lipids (courtesy of Peter Tieleman).The center of the peptide and the lipid bilayer were aligned, andthe peptide was oriented with its first principal axis perpendic-ular to the lipid bilayer. To create five different starting confor-mations the peptide was rotated around its first principal axisby 0, 45, 90, 180, and 270°, respectively.To pack the lipid around the peptide we used the inflategro

package inGROMACS (8). First, the lipid bilayer was expandedby a factor of 4, and lipid molecules within 14 Å of the peptidewere removed to avoid clashes between lipid and peptideatoms. To generate an asymmetric lipid bilayer, the POPS lipidsof the outer leaflet and 50%of the POPS in the inner leaflet wereconverted into POPC. Lipids were iteratively packed aroundthe CEACAM1-S peptide by shrinking the bilayer area by 5%using the inflategro package in GROMACS, with steepest

descent energy minimization, using position restraints to pre-vent the peptide from moving, performed after every iteration.This procedure was repeated until an area per lipid close to theexperimental value of 65.8 A2 (9) was reached.Single-point charge water molecules (10) were added on

both sides of the lipid bilayer, and an appropriate number ofsodium ions were added in the water to neutralize the sys-tem. For each of the five conformations we also generated asystem with 2 additional Ca2� ions that were neutralized by4 Cl� ions. All ions were added by displacing random watermolecules using the genion command in GROMACS. The finalsystems were minimized by steepest descent energy minimiza-tion with a maximum force of 1000 kJ/mol per nm as conver-gence criterion.MD Simulation Parameters—MD simulations on

CEACAM1-S in a lipid bilayer consisting of POPS and POPCwere performed using GROMACS 4.0.5 (11) and the GRO-MOS96 53a6 forcefield (12) extended with Berger lipid param-eters (13). Short range nonbonded interactions were truncatedat 1.2 nm,with the neighbor list updated every 10 fs. To accountfor the cutoff in the van der Waals interactions, long rangedispersion correctionwas applied to energy andpressure terms.Long range electrostaticswere calculated using the smoothpar-ticle mesh Ewald method (14). Bonds were constrained usingthe LINCS algorithm (15) to allow for a time step of 2 fs. Forcontinuity, periodic boundary conditions were applied in everydimension.MD simulations were performed on five starting conforma-

tions of CEACAM1-S in lipid without Ca2� ions and five con-formations of CEACAM1-S in lipid with Ca2� ions. Each sys-tem was equilibrated by performing 100 ps of MD at 310 Kusing a NVT ensemble followed by 5 ns of MD under NPTconditions with a pressure of 1 bar. The velocity-rescaling ther-mostat (16) was used for temperature coupling during theequilibration and a Parrinello-Rahman barostat (17, 18) forpressure coupling. The CEACAM1 peptide was kept in placeduring these equilibration steps using position restraints of1000 kJ/mol per nm2.

After the systems were equilibrated at the correct tempera-ture and pressure, MD simulations of 100 ns were performedfor each of the 10 conformations, using a NVT ensemble with aNose-Hoover thermostat (19). Simulations on the F454Amutant were performed over 40–100 ns for a total of �600 ns.Data were not collected for the first 5 ns of each simulation.K-Means Clustering Analysis—The interatomic distances

between the C�, C�, and C� atoms of the CEACAM1-S peptidewere extracted every 10 ps from theMD trajectories. The first 5ns of each trajectory was left out to avoid not well equilibratedconformations. The distances of all simulations were combinedinto one distance matrix for clustering analysis, leading to atotal of 9510 sampled conformations with each conformationexpressed as 4950 interatomic distances. The distance matrixwas grouped into two clusters using a K-Means clustering algo-rithm. The process was repeated 10 times to verify that theclusters were reproducible; grouping into three clusters wasattempted but was not reproducible.Preparation of Samples for NMR—Fmoc (N-(9-fluorenyl)-

methoxycarbonyl)-Phe-OH-13C9, 15N 98 atom % 13C, 98 atom

Interaction of CEACAM1 and Actin Is Ca2�- and Lipid-dependent

AUGUST 5, 2011 • VOLUME 286 • NUMBER 31 JOURNAL OF BIOLOGICAL CHEMISTRY 2752950

% 15N was purchased from ISOTEC. C18 Sep-Pak cartridgeswere purchased from Waters (Ireland). POPS (porcine brain)and POPC (chicken egg) were purchased from Avanti PolarLipids. Nonmuscle actin (99% pure) was purchased fromCytoskeleton.To remove the residual TFA from the synthesized peptide, a

C18 cartridge was conditioned using 70% acetonitrile solutionfollowed by 1% acetic acid wash. After the peptide solution wasloaded onto the cartridge, the column was washed using 1%acetic acid, then eluted with 70% acetonitrile-water. Sampleswere lyophilized and stored at �80 °C. The peptide concentra-tionwas determinedusing anNMRmethod (20). Bulk solutionsof POPS and POPCwere prepared by dissolving 100mg of eachlipid in 4 ml of chloroform, flushed with argon, and stored at�20 °C.For POPS or POPC liposomes, 288 or 268 �l of bulk solution

was transferred to a 1.5-ml polypropylene tube, and the solventwas evaporated under argon and further dried overnight undervacuum. To the dry lipid film, 400�l of 50mMphosphate buffer(90% D2O, pH 7.0) and 1.5 �l of 166 mM EDTA solution wereadded, vortexed, and sonicated for 30 min or until the lipidsolution became transparent. The 400-�l liposome solutionwas then mixed with 100 �l of 0.5 mM MUA-13C, 15N-Phe-CEACAM1-Speptide in 50mMphosphate buffer (90%D2O, pH7.0), and 1mM tris (2-carboxyethyl)phosphine. The sample wasthen transferred to the NMR tube and flushed with argon gas.For the study of the complex with actin, 460 �l of the MUA-CEACAM1-S peptide-liposome solution was added to 2 mg ofactin powder. After vigorous vortexing, the solution was trans-ferred to an argon-flushed Shigemi NMR tube. For the samplein the presence of Ca2�, the Ca2� was added after the additionof actin at a final Ca2� concentration of 5 mM. For the mixedPOPS/POPC liposome system, 144 �l of POPS and 134 �l ofPOPC were taken from the bulk POPS and POPC solution,respectively, and the two lipids were mixed well before evapo-rating the solvent.NMR Experiments—We used 13C-labeled Phe in the

CEACAM1-S peptide as a probe because the chemical shifts ofthe aromatic ring are well separated from signals of the lipid.Because the CEACAM1-S peptide is inserted into the liposomethrough the MUA-aliphatic group, there is significant linebroadening of the NMR signal from 13C-labeled Phe aromaticring. For 0.1 mM concentration of peptide, the peptide signal isvery weak from constant-time 13C-1H heteronuclear single-quantum coherence experiment (21).Much better signal inten-sity (�2-fold) was obtained from the aromatic ring using con-stant time TROSY experiment (22). The experiments werecarried out at 35 °C on a 600-MHzBruker instrument equippedwith a cryoprobe. The constant time duration, spectrumwidth,and total number of free induction decays for the 13C dimen-sion are 17.6ms, 4 ppm, and 20, respectively. For the 1H dimen-sion, the total number of points and spectrum width are 2048and 14 ppm, respectively. The recycle delay is 1.2 s. The exper-imental time is about 18 h with the number of scans rangingfrom 2200 to 2560. The TROSY data were processed using theNMRPipe software (23) and analyzed using the NMRView pro-gram (24).

RESULTS

Lipid Bilayer Model System—CEACAM1 has two cy-toplasmic domain isoforms, of which the short isoform(CEACAM1-S) is sufficient to induce lumen formation whenexpressed in the breast cancer cell line MCF7 and grown inthree-dimensional culture (Fig. 1A). When CEACAM1-S isexpressed in cells that undergo cell-cell contact, CEACAM1-S

FIGURE 1. CEACAM1-S induces lumen formation and migrates to cell-cellcontacts. A, expression of wild-type CEACAM1-S induces 97% lumen forma-tion in transfected MCF7 cells grown in three-dimensional culture (300 acinicounted; representative acini shown in a), whereas the cytoplasmic domainmutant F454A suppressed lumen formation by �90% (16% lumen formation,300 acini counted; representative acini shown in b). The data are taken fromChen et al. (5) with permission. B, HeLa cells transfected with a CEACAM1-S-GFP fusion protein (5) showing accumulation of CEACAM1-S (green, a) andF-actin (red, b) at the cell surface of single cells making initial cell-cell contactswith almost complete redistribution of their CEACAM1-S between the cells (c)when cell-cell contact is complete.

Interaction of CEACAM1 and Actin Is Ca2�- and Lipid-dependent

27530 JOURNAL OF BIOLOGICAL CHEMISTRY VOLUME 286 • NUMBER 31 • AUGUST 5, 201151

migrates to the cell-cell contact region along with F-actin (Fig.1B). Although we have shown that the initial interaction of theshort cytoplasmic domain is with G-actin (5), the small size ofthis domain and its close proximity to the inner leaflet of theplasma membrane prompted us to investigate the role of themicroenvironment in its interaction with G-actin in a modelsystem. The model comprises a phospholipid bilayer coatedonto 4.5-�mglass beads into which is inserted anMUA versionof the 12-amino cytoplasmic domain of CEACAM1-S (Fig. 2,Aand B). The MUA-peptide spontaneously inserts into the lipidbilayer in a saturable manner that can be quantitated by fluo-rescently tagged anti-peptide antibodies using flow cytomtery(Fig. 2C). The liposome-glass bead displayed peptide is thenincubated with either G-actin (detected by anti-actin antibody

plus fluorescent-tagged secondary antibody) or rhodamine-ac-tin in the presence or absence of Ca2�. Two phospholipid envi-ronments were analyzed: POPS, an abundant negativelycharged phospholipid on the inner leaflet of the plasma mem-brane, and POPC, the most abundant neutral phospholipid onthe outer leaflet. As shown on Fig. 3A, G-actin binds to thepeptide only in POPSwith Ca2�, but not in POPCwith or with-out Ca2� (Fig. 3B). To test the validity of this model systemfurther, we repeated the experiment with twomutated versionsof the peptide, the F454A mutation, which abrogates in vitrobinding toG-actin and in vivo lumen formation, and the K456Amutation, which enhances G-actin binding (5). Consistent withthe previous studies, the F454Amutation abrogated actin bind-ing (Fig. 4A), and the K456A mutation enhanced actin binding(Fig. 4B). These results not only confirm our previous findingsbut also suggest that both the lipid and Ca2� environmentenable the interaction, thus providing a context for the regula-tion of the interaction. They also show that these liposomes canbe used to query the interaction of a transmembrane proteininteracting with intracellular proteins.MD Simulations—To investigate any direct effect Ca2� ions

may have on the conformation of the CEACAM1-S peptide inits native lipid environment we performed a total of 1�s ofMDsimulations on CEACAM1-S embedded in an asymmetric lipidbilayer. To reduce bias caused by the choice of an initial con-formation we have initiated simulations from five unique start-

FIGURE 2. Generation of lipid-embedded cytoplasmic domain ofCEACAM1-S. The cytoplasmic domain of CEACAM1-S has 12 amino acids thatwhen synthesized with an N-acyl-MUA (A) spontaneously inserts into 4.5-�mglass beads coated with POPS or POPC liposomes used in actin binding exper-iments (B) and can be quantitated with fluorescent-tagged anti-peptide anti-bodies (mean fluorescence intensity versus concentration of peptide, C).

FIGURE 3. G-actin binds to CEACAM1-S peptide in POPS liposome coatedglass beads in a Ca2�-specific manner. A, POPS liposome-coated 4.5-�mglass beads � N-acyl-MUA-wild-type peptide � rhodamine G-actin (or �G-actin and � anti-G-actin antibody) � 1 mM Ca2� in PBS plus 1% BSA wereanalyzed by FACS (see right panel for details). Significant actin binding is onlyseen for the fully reconstituted system. B, same experiment as in A exceptPOPC liposome-coated glass beads were used. There was no significant actinbinding even for the fully reconstituted system.

Interaction of CEACAM1 and Actin Is Ca2�- and Lipid-dependent

AUGUST 5, 2011 • VOLUME 286 • NUMBER 31 JOURNAL OF BIOLOGICAL CHEMISTRY 2753152

ing conformations. Each initial conformationwas simulated for100 nswith andwithout the addition of 2 Ca2� ions (simulationbox size, 7 � 7 � 10 nm). During all five of theMD simulationswith Ca2� we observed a migration of Ca2� ions to the POPSlipid head groups, where they may mediate the interactionbetween CEACAM1-S and actin. K-Means clustering of theCEACAM1-S conformation over all MD trajectories revealedtwo distinct conformational clusters (supplemental Fig. S1).The conformations in the distinct clusters differ in the orienta-tion of the cytoplasmic domain from Phe-454 onward withrespect to the transmembrane domain (supplemental Fig. S1).The structure of the cytoplasmic and transmembrane domainsby themselves is invariant between the clusters, thus suggestingthat Phe-454 and its surrounding residues may serve as a hingearound which a conformational change occurs. The two differ-ent clusters exhibit differences in the orientation of Phe-454(Fig. 5,A–C). In one conformation Phe-454 is pointing outwardinto the solvent, whereas in the other conformation Phe-454 isparallel to the lipid bilayer and buried by the surrounding lipidmolecules. In the conformationswhere Phe-454 is buried by thesurrounding lipids we observed a cation-� interaction (25)between Phe-454 and Lys-456 (Fig. 6). This interaction is espe-cially favored when Phe-454 is buried in the lipid bilayer. In thesolvent-exposed conformations the dehydration of the amineion of Lys-456 causes the cation-� interaction to be unfavorableand hence broken (26). The observed cation-� interactionbetween Phe-454 and Lys-456 may be one of the mechanismsbywhich Lys-456 inhibits actin binding. As expected, themuta-tion K456A relieves this inhibition (Fig. 4B). Simulations werealso performed for the F454A mutant embedded in the lipid

bilayer. The peptide conformations from these simulationsclustered into a single conformation resembling the conforma-tion of Phe-454 that points into solvent (supplemental Fig. S2).This shows that Phe-454 is required forCEACAM1-S to exist intwo different conformations.Individual simulations show no transitions from one confor-

mational cluster to the other cluster (supplemental Fig. S1), andtherefore it is speculative to discern whether addition of Ca2�

causes one cluster to be favored over the other. NMR TROSYexperiments were conducted to assess which of the two confor-mations is favored for actin binding and whether Ca2� ionshave an effect on the balance between the two conformations.NMR Analysis of Peptide-Liposomes—The conformation of

Phe-454 in a POPS liposome environmentwas assessed directlyby synthesis of the N-acyl-MUA peptide with 13C-labeled Pheand analysis by NMR TROSY. This approach allowed a cleanseparation of the aromatic signals of Phe-454 from the abun-dant phospholipid signals and had sufficient sensitivity to allowquantitation of peak volumes despite the peak broadening dueto insertion of the peptide in liposomes. Two different environ-ments (further supported in supplemental Fig. S3) wereobserved for Phe-454. Each environment has a distinct set oftwo cross-peaks (Fig. 5, D–F). Although three expected peakswere observed from free peptide in the absence of liposome(data not shown), it is clear that the lipid environment perturbsthe pattern of cross-peaks, increasing the signal line width sig-nificantly, reducing the number of cross-peaks from three totwo (for possible explanations, see supplemental NMR Discus-sion). The addition of actin caused a decrease in the overallintensity of the peaks aswell as a change in the ratio between thepeaks (Fig. 5E and Table 1). The decrease in peak intensity canbe interpreted as binding of actin to one of the conformerswhich, due to an increase in molecular size, causes increasedline broadening so that the signal becomes NMR invisible, theso-called “dark” state (27). The shift in ratio between the twoconformations could indicate that one conformation is pre-ferred by actin. When Ca2� is added, the total peak intensitiesare further decreased by 44% (Fig. 5F and Table 1), and the shiftin ratios is increased, emphasizing a role for Ca2� in enhancingthe binding of CEACAM1-S to actin.The results of the NMR TROSY study not only support our

hypothesis that the phospholipid and Ca2� affect the interac-tion of the CEACAM1-S peptide with actin, but they also sug-gest that two major conformers of Phe-454 exist in this envi-ronment, supporting the findings from the MD simulationstudy. Further NMR TROSY experiments were conducted todetermine the effect of different phospholipids in the liposomeson the Phe-454 interaction with actin in the presence orabsence of Ca2� (supplemental Figs. S3 and S4). In the POPCliposome environment, three conformers were observed, andalthough some peak intensities decreased with the addition ofactin, therewas no significant change in total peak volumeuponthe addition of Ca2� (if anything, it decreased the interaction).This result agrees with the prediction that POPCdoes not favorthe peptide-actin interaction in the presence of Ca2�. In themixed POPC/POPS liposome environment, the results areessentially between the pure POPS and pure POPC environ-ment (supplemental Table S1). The total peak volume is

FIGURE 4. G-actin does not bind F454A mutant and binds more stronglyto K456A mutant of CEACAM1-S. A, same experiment as in Fig. 3A exceptF454A mutated peptide was used. B, Same experiment as in A except K456Amutated peptide was used.

Interaction of CEACAM1 and Actin Is Ca2�- and Lipid-dependent

27532 JOURNAL OF BIOLOGICAL CHEMISTRY VOLUME 286 • NUMBER 31 • AUGUST 5, 201153

decreased by 15% upon the addition of Ca2� to the complex ofactin and peptide.To address whether Ca2� affects the interaction between

G-actin and CEACAM1-S peptide by perturbation of the equi-

librium between two conformations of Phe-454, we carried outan NMR experiment on free peptide in the POPS environmentwith different concentrations of Ca2� in the absence of actin.We found that [Ca2�] per se does not affect the equilibrium

FIGURE 5. Two major conformer ensembles predicted by MD on the CEACAM1-S peptide in an asymmetric POPS/POPC lipid bilayer and two confor-mations of Phe-454 in CEACAM1-S peptide inserted into POPS liposomes in the presence of actin and Ca2� observed by NMR. A–C, MD simulationsreveal two conformations that represent different environments for Phe-454. A, in one conformational ensemble Phe-454 is exposed to solvent surroundingthe lipid bilayer. B, the other ensemble has Phe-454 buried in the surrounding lipid head groups. C, representative (nearest to mean) conformations are shownfor both ensembles aligned and with Phe-454 in a stick representation. D–F, 13C-labeled Phe-454 N-acyl-MUA-peptide (0.1 mM) in POPS (18 mM) liposomes givestwo sets of cross-peaks, C1-a and C1-b, corresponding to conformation 1, and C2-a and C2-b corresponding to conformation 2 (total percentages of eachconformation indicated) as analyzed by two-dimensional NMR TROSY (D). The total percent of conformation 2 increases with the addition of actin (E) andCa2� (F).

Interaction of CEACAM1 and Actin Is Ca2�- and Lipid-dependent

AUGUST 5, 2011 • VOLUME 286 • NUMBER 31 JOURNAL OF BIOLOGICAL CHEMISTRY 2753354

between the two conformers (supplemental Fig. S5 and supple-mental Table S2). Thus, most likely, Ca2� is first recruited tothe negatively charged phospholipid, which in turn, enhancesthe peptide-G-actin interaction. In other work, although G-ac-tin has been shown to interact directly with negatively chargedliposomes (28), the additional effect of inserted peptides has notbeen studied.The NMR analysis also indicates that pH affects the equilib-

rium of the two observed conformations of the peptide embed-ded in the lipid (supplemental Fig. S3,A andB). BecauseCa2� isa better studied signal transducer at the plasma membranecompared with pH, the effects of pH were not studied further.Analysis by Surface Plasmon Resonance—A third approach

to assess the peptide-G-actin interaction in a phospholipidenvironment is the use of surface plasmon resonance where aphospholipid bilayer can be monolayered onto a hydrophobicchip (29). In this approach, a lipid monolayer of POPS wasapplied to an existing hydrophobic substrate, N-acyl-MUA-peptide was inserted into the upper phospholipid layer, and thesubsequent binding of actin was monitored in the presence orabsence of Ca2�. The results demonstrate that N-acyl-MUApeptide binds in a saturable manner to the POPS phospholipid

monolayer and that actin binding requires Ca2� (supplementalFig. S6).

DISCUSSION

Three completely different experimental approaches, to-gether with MD simulations, demonstrate a requirement fornegatively charged phospholipid POPS and Ca2� in the regula-tion of the binding of G-actin to the cytoplasmic domain pep-tide of CEACAM1-S. In the first approach, POPS or POPCliposomes mimicking the inner or outer leaflet of the plasmamembrane, respectively, were coated onto 4.5-�m glass beads,N-acyl-MUA-peptide was inserted into the outer leaflet, andthe binding of actin was monitored in the presence or absenceof Ca2�. This approach has the advantage that it creates a cell-sized lipid surface that allows specification of interacting com-ponents present at either the inside or outside of the cell mem-brane and measurement of interactions by flow cytometryusing fluorescent labeling of components of interest. Thisapproach validated our earlier finding that Phe-454 plays anessential role in actin binding (5). A disadvantage of theapproach is that it provides no direct evidence of the molecularstate of the inserted peptide. To assess this information, we

FIGURE 6. Cation-� interaction observed between Phe-454 and Lys-456 only when Phe-454 is in the lipid head groups. The distributions of C�–C�

distances between Phe-454 and Lys-456 sampled during MD are shown. A, when Phe-454 is solvent-exposed (blue conformation) there is only one peak in thedistribution observed at �9.5 Å, indicating that there is no interaction between Phe-454 and Lys-456. B, when Phe-454 is in the lipid head groups (redconformation) an additional peak in the distribution is observed at �5.2 Å, indicating an interaction between Phe-454 and Lys-456.

TABLE 1NMR peak volumes of 13C-Phe-CEACAM1-S peptide in POPS liposomeAll peak volumes are corrected against factors of number of scans and the dilution due to addition of Ca2� to the system.

Peaka Vpepb Vpep_act

c (Vpep_act � Vpep)/Vpep Vpep_act_Ca2�d (Vpep_act_Ca2� � Vpep)/Vpep

C1-a 1,006 684 �32% 396 �61%C1-b 2,408 1,732 �28% 857 �64%C2-a 147 301 105% 236 61%C2-b 153 326 113% 223 46%Total 3,714 3,043 �18% 1,712 �54%

a Peak symbols (C1-a/b and C2-a/b) are shown in Fig. 5.b Vpep is the peak volume measured from the free CEACAM1-S peptide in the liposome.c Vpep_act is the peak volume measured from the complex of CEACAM1-S and actin.d Vpep_act_Ca2� is the peak volume measured from the complex of CEACAM1-S and actin in the presence of 5 mM Ca2�.

Interaction of CEACAM1 and Actin Is Ca2�- and Lipid-dependent

27534 JOURNAL OF BIOLOGICAL CHEMISTRY VOLUME 286 • NUMBER 31 • AUGUST 5, 201155

performed NMR analysis on the 13C-labeled peptide (at theresidue of interest) inserted into liposomes under a variety ofconditions. This approach demonstrated that Ca2� played amajor role in the peptide-actin interaction and not in the pep-tide-phospholipid interaction. An advantage of this approach isthe use of liposomes that more accurately reflect the lipid envi-ronment in the cell as opposed to the conventional use of pep-tides embedded in micelles.In our model system 1 mM Ca2� was required to effect actin

binding. Although the global concentration of Ca2� in the cyto-plasm during signaling peaks at about 1.4 �M, the local concen-tration of Ca2� in themicroenvironment around the cytoplasmadjacent to the cell membrane would be in themillimolar rangedue to the effect of calcium ion channel proteins and/or cal-cium transporters that produce a spike in the local Ca2� con-centration. Because our liposomes lack the presence of theseCa2� transporters, amillimolar bulk concentration ofCa2�wasused to mimic the local environment in the cell. The differencebetween local and global concentrations of Ca2� can be visual-ized in our molecular simulations shown in Fig. 5, A–C, wherewe have added only two Ca2� ions to the lipid bilayer-cytosolbox (7� 7� 10 nm). This leads to a local Ca2� concentration ofabout 8mM. If we simulated the physiological global concentra-tion of Ca2� inside activated cells we would either have to use abox 103 times larger or assume 10�3 Ca2� ions in our box,assumptions that are illogical. We believe that these consider-ations justify our use of 1 mM Ca2� in our experiments.

A potential disadvantage of the NMR method is that theoverall molecular size (including the lipid bilayer) of the systemis quite large, and thus a millimolar sample concentration isneeded to compensate for the increased line width of NMRpeaks. Thus, we employed a third approach, namely the use ofsurface plasmon resonance that has the advantage of allowingkinetic measurements of the binding of various components toeach other at lower concentrations. This method also con-firmed Ca2� and lipid specific requirements for G-actin bind-ing to lipid-embedded CEACAM1-S peptide. However, thismethod suffers from the lack of structural information. Thus,we performed MD simulations to gain access to changes in theentire peptide embedded in the lipid bilayer at the atomic level.This approach allowed us to provide a theoretical basis for ourfindings and to make predictions that could be validated byfurther experiments. A combination of MD simulations andNMR experiments reveal two reversible conformational statesof Phe-454. The population of these states is not affected byvarying the Ca2� concentration but does change on actin bind-ing. MD simulations predict that in one conformation Phe-454is solvent-exposed whereas in the other conformation Phe-454is buried in the lipid head groups. Furthermore, MD simula-tions show that the conformation with Phe-454 buried in thelipid head groups may be stabilized by a cation-� interaction(25, 30) of Lys-456with Phe-454. This predicted cation-� inter-action, and its role in conformational selection, could explainwhy mutation of Lys-456 affects actin binding in a positivemanner. Finally, we performed MD simulations for the F454Amutant peptide and found that Ala-454 was unable to adopttwo confirmations in the lipid bilayer, strengthening our use of

MD as a method to examine the structural constraints of themodel system.Our findings help explain why many receptor cytoplasmic

domain peptides, especially those as short as the CEACAM1-Speptide, may have highly specific and tunable interactions withcomponents of the cytosol. Previously, receptor aggregationwas proposed as a major mechanism for activation of cytoplas-mic domains in downstream signaling (31, 32). However, thismechanism is inadequate to account for the cooperative effectof the local phospholipid environment and Ca2� signaling. Interms of the downstream interacting protein G-actin, it hasbeen shown previously that it is attracted to negatively chargedphospholipids (28), but this interaction alone is insufficient toconfer membrane site specificity or temporal control by Ca2�

signals. Indeed, for actin to polymerize at the right location andtime at the membrane, it must find a suitable binding partnerthat responds to rapid signals from both the outside (say recep-tor clustering) and inside (say Ca2� signaling). CEACAM1-Sprovides the outside activation step by receptor clusteringduring its cell-cell interactions (31) and the inside activationstep via Ca2�-generated interactions at Phe-454 in the nega-tively charged phospholipid environment. In the case ofCEACAM1-S, the source of the intracellular Ca2� is likely fromother receptors that become activated during cell-cell interac-tions, including G protein-coupled receptors (33, 34). In thisrespect, both CEACAM1 (35–37) and G protein-coupledreceptors (38) reside within the same lipidmicroenvironments,emphasizing their local ability to coordinate signals. Finally, foractin to polymerize, it must be correctly oriented to formthe characteristic double helix of F-actin. In the case ofCEACAM1-S, the polymerization would occur directly adja-cent to the phospholipid bilayer, exactly the situation seen forthe formation of cortical actin. Thus, it is likely that corticalactin formation is dynamic, responding to cell-cell interactionsthat allow local changes in the distribution of receptors at thecell surface (for an example of the redistribution ofCEACAM1-S and actin during cell-cell adhesion, see Fig. 1).In summary, these findings explain how receptors with short

cytoplasmic tails can recruit cytosolic proteins in a phospholip-id- and calcium-specificmanner. In addition, thesemodels pro-vide a powerful approach that can be applied to other mem-brane protein interactions.

Acknowledgment—Preliminary flow studies were performed by KeithLe.

REFERENCES1. Brozinick, J. T., Jr., Berkemeier, B. A., and Elmendorf, J. S. (2007) Curr.

Diabetes Rev. 3, 111–1222. Scita, G., Confalonieri, S., Lappalainen, P., and Suetsugu, S. (2008) Trends

Cell Biol. 18, 52–603. Weed, S. A., and Parsons, J. T. (2001) Oncogene 20, 6418–64344. Huang, Y., and Burkhardt, J. K. (2007) J. Cell Sci. 120, 723–7305. Chen, C. J., Kirshner, J., Sherman, M. A., Hu,W., Nguyen, T., and Shively,

J. E. (2007) J. Biol. Chem. 282, 5749–57606. Ko, K. S., Arora, P. D., Bhide, V., Chen, A., and McCulloch, C. A. (2001)

J. Cell Sci. 114, 1155–11677. Clapham, D. E. (2007) Cell 131, 1047–10588. Kandt, C., Ash, W. L., and Tieleman, D. P. (2007)Methods 41, 475–488

Interaction of CEACAM1 and Actin Is Ca2�- and Lipid-dependent

AUGUST 5, 2011 • VOLUME 286 • NUMBER 31 JOURNAL OF BIOLOGICAL CHEMISTRY 2753556

9. Tieleman, D. P., Forrest, L. R., Sansom, M. S., and Berendsen, H. J. (1998)Biochemistry 37, 17554–17561

10. Berendsen, H. J., Postma, J. P., van Gunsteren, W. F., and Hermans, J.(1981) in Intermolecular Forces, (Pullman, B., ed) pp. 331–342. Reidel,Dordrecht, The Netherlands

11. Van Der Spoel, D., Lindahl, E., Hess, B., Groenhof, G., Mark, A. E., andBerendsen, H. J. (2005) J. Comput. Chem. 26, 1701–1718

12. Oostenbrink, C., Villa, A., Mark, A. E., and van Gunsteren, W. F. (2004)J. Comput. Chem. 25, 1656–1676

13. Berger, O., Edholm, O., and Jahnig, F. (1997) Biophys. J. 72, 2002–201314. Essmann, U., Perera, L., Berkowitz, M. L., Darden, T., Lee, H., and Peder-

sen, L. G. (1995) J. Chem. Phys. 103, 8577–859315. Hess, B., Bekker, H., Berendsen, H. J., and Fraaije, J. G. (1997) J. Comput.

Chem. 18, 1463–147216. Bussi, G., Donadio, D., and Parrinello, M. (2007) J. Chem. Phys. 126,

01410117. Parrinello, M., and Rahman, A. (1981) J. Appl. Phys. 52, 7182–719018. Nose, S., and Klein, M. L. (1983)Mol. Phys. 50, 1055–107619. Nose, S. (1984) J. Chem. Phys. 81, 511–51920. Wider, G., and Dreier, L. (2006) J. Am. Chem. Soc. 128, 2571–257621. Vuister, G. W., and Bax, A. (1992) J. Magn. Reson. 98, 428–43522. Pervushin, K., Riek, R., Wider, G., and Wuthrich, K. (1998) J. Am. Chem.

Soc. 120, 6394–640023. Delaglio, F., Grzesiek, S., Vuister, G. W., Zhu, G., Pfeifer, J., and Bax, A.

(1995) J. Biomol. NMR 6, 277–293

24. Johnson, B. A., and Blevins, R. A. (1994) J. Biomol. NMR 4, 603–61425. Dougherty, D. A. (1996) Science 271, 163–16826. Kumpf, R. A., and Dougherty, D. A. (1993) Science 261, 1708–171027. Fawzi, N. L., Ying, J., Torchia, D. A., and Clore, G. M. (2010) J. Am. Chem.

Soc. 132, 9948–995128. Bouchard, M., Pare, C., Dutasta, J. P., Chauvet, J. P., Gicquaud, C., and

Auger, M. (1998) Biochemistry 37, 3149–315529. Ida, M., Satoh, A., Matsumoto, I., and Kojima-Aikawa, K. (2004)

J. Biochem. 135, 583–58830. Dougherty, D. A. (2007) J. Nutr. 137, 1504S–1508S; discussion

1516S–1517S31. Gray-Owen, S. D., and Blumberg, R. S. (2006) Nat. Rev. Immunol. 6,

433–44632. Muller, M. M., Klaile, E., Vorontsova, O., Singer, B. B., and Obrink, B.

(2009) J. Cell Biol. 187, 569–58133. Werry, T. D., Wilkinson, G. F., and Willars, G. B. (2003) Biochem. J. 374,

281–29634. Bouschet, T., Martin, S., and Henley, J. M. (2008) Trends Pharmacol. Sci.

29, 633–63935. Thorp, E. B., and Gallagher, T. M. (2004) J. Virol. 78, 2682–269236. Chen, C. J., and Shively, J. E. (2004) J. Immunol. 172, 3544–355237. Muenzner, P., Bachmann, V., Kuespert, K., and Hauck, C. R. (2008) Cell

Microbiol. 10, 1074–109238. Fallahi-Sichani, M., and Linderman, J. J. (2009) PLoS One 4, e6604

Interaction of CEACAM1 and Actin Is Ca2�- and Lipid-dependent

27536 JOURNAL OF BIOLOGICAL CHEMISTRY VOLUME 286 • NUMBER 31 • AUGUST 5, 201157

Curcumin Modulates Nuclear Factor �B (NF-�B)-mediatedInflammation in Human Tenocytes in VitroROLE OF THE PHOSPHATIDYLINOSITOL 3-KINASE/Akt PATHWAYReceived for publication, April 29, 2011, and in revised form, May 30, 2011 Published, JBC Papers in Press, June 13, 2011, DOI 10.1074/jbc.M111.256180

Constanze Buhrmann‡, Ali Mobasheri§1, Franziska Busch‡, Constance Aldinger‡, Ralf Stahlmann¶,Azadeh Montaseri�, and Mehdi Shakibaei‡2

From the ‡Musculoskeletal Research Group, Institute of Anatomy, Ludwig-Maximilian-University Munich, 80336 Munich, Germany,the §Musculoskeletal Research Group, Division of Veterinary Medicine, School of Veterinary Medicine and Science, Faculty ofMedicine and Health Sciences, University of Nottingham, Sutton Bonington LE12 5RD, United Kingdom, the ¶Institute of ClinicalPharmacology and Toxicology, Charite–Universitatsmedizin Berlin, 14195 Berlin, Germany, and the �Department of AnatomicalSciences, Medicine Faculty, Medical University of Tabriz, Tabriz, Iran

Inflammatory processes play essential roles in the pathogen-esis of tendinitis and tendinopathy. These events are accompa-nied by catabolic processes initiated by pro-inflammatory cyto-kines such as interleukin-1� (IL-1�) and tumor necrosisfactor-� (TNF-�). Pharmacological treatments for tendinitisare restricted to the use of non-steroidal anti-inflammatorydrugs. Recent studies in various cell models have demonstratedthat curcumin targets the NF-�B signaling pathway. However,its potential for the treatment of tendinitis has not beenexplored. Herein, we used an in vitromodel of human tenocytesto study the mechanism of curcumin action on IL-1�-mediatedinflammatory signaling. Curcumin at concentrations of 5–20�M inhibited IL-1�-induced inflammation and apoptosis in cul-tures of human tenocytes. The anti-inflammatory effects of cur-cumin included down-regulation of gene products that mediatematrix degradation (matrix metalloproteinase-1, -9, and -13),prostanoid production (cyclooxygenase-2), apoptosis (Bax andactivated caspase-3), and stimulation of cell survival (Bcl-2), allknown to be regulated by NF-�B. Furthermore, curcumin sup-pressed IL-1�-induced NF-�B activation via inhibition of phos-phorylation and degradation of inhibitor of �B�, inhibition ofinhibitor of �B-kinase activity, and inhibition of nuclear trans-location of NF-�B. Furthermore, the effects of IL-1� were abro-gated bywortmannin, suggesting a role for the phosphatidylino-sitol 3-kinase (PI-3K) pathway in IL-1� signaling. Curcuminsuppressed IL-1�-induced PI-3K p85/Akt activation and itsassociation with IKK. These results demonstrate, for the firsttime, a potential role for curcumin in treating tendon inflamma-tion through modulation of NF-�B signaling, which involvesPI-3K/Akt and the tendon-specific transcription factor scleraxisin tenocytes.

The global incidence of tendon injuries has increased in con-junction with the rise in aging and inflammatory diseases (1).

The etiology of tendinopathy is considered to bemultifactorial,but mechanical loading, overuse injury (or association witharthritis), adverse effects of quinolone antibiotics or otherdrugs, degeneration, and inflammation that cause tendon inju-ries and rupture are the major causative factors (2–5). Teno-cytes are embedded in an extensive three-dimensional networkof extracellular matrix components consisting predominantlyof collagen type I fibrils (�95% of the total collagen in tendon),other types of collagen (type III and type V), proteoglycans,elastin, and fibronectin (6–8). These specific matrix compo-nents give tendon its resilience and biomechanical stability.Tendon and ligament are dense fibrous connective tissues

with a very limited intrinsic potential for regeneration (7, 9).Therefore, their repair and regeneration poses a complex clin-ical challenge. It is important to understand the cellular andmolecularmechanisms involved in tendondegeneration duringthe early stages of disease pathogenesis to develop new andeffective treatments for tendinopathy. Subtle changes such asthe release of IL-1� or other inflammatory cytokines by infil-trating macrophages/monocytes may occur (10). Moreover,like in other connective tissue injuries, tendon inflammation isaccompanied by the up-regulation of pro-inflammatory cyto-kines such as IL-1�. In vitro studies have shown that IL-1� caninduce inflammatory mediators such as COX-2, prostaglandinE2, and matrix metalloproteinases (MMP),3 all known to beinvolved in tendon matrix degradation (11, 12).IL-1� is a potent pro-inflammatory cytokine that has been

reported to be present in significantly increased quantities inthe synovium where it enhances inflammatory reactions ininjured joints (13, 14). The intracellular signaling pathwaysactivated by IL-1� are responsible for stimulatingMMPexpres-sion andCOX-2production.However, these pathways have notbeen explored in detail in tendon cells. Pro-inflammatory cyto-kines (e.g. IL-1�) induce activation of a central transcriptionfactor known asNF-�B, which is a key regulator of gene expres-sion (15, 16). NF-�B is present in the cytoplasm in its restingstage as a heterotrimer complex consisting of two subunits andan additional inhibitory subunit, I�B� (17). During the activa-tion process, the inhibitory subunit I�B� is phosphorylated at

Author’s Choice—Final version full access.1 Supported by the Biotechnology and Biological Sciences Research Council

(BBSRC) and The Wellcome Trust, UK.2 To whom correspondence should be addressed: Institute of Anatomy, Mus-

culoskeletal Research Group, Ludwig-Maximilian-University Munich,Pettenkoferstrasse 11, D-80336 Munich, Germany. Tel.: 49-89-5160-4827;Fax: 49-89-5160-4828; E-mail: mehdi.shakibaei@med.uni-muenchen.de.

3 The abbreviations used are: MMP, matrix metalloproteinase; IKK, I�B�kinase; SCXA, scleraxis; TRAF1, TNF receptor-associated factor 1.

THE JOURNAL OF BIOLOGICAL CHEMISTRY VOL. 286, NO. 32, pp. 28556 –28566, August 12, 2011Author’s Choice © 2011 by The American Society for Biochemistry and Molecular Biology, Inc. Printed in the U.S.A.

58

Ser-32 and Ser-36 residues by IKK kinase (I�B� kinase) and issubsequently degraded. Once released, subunits of activatedNF-�B translocate to the nucleus and mediate transcription ofvarious inflammatory and catabolic gene products (16, 18).NF-�B activation has been shown to regulate the expression ofmore than 500 different gene products linked with inflamma-tion, tumor cell transformation, survival, proliferation, inva-sion, angiogenesis, metastasis, and chemoresistance (19). Thus,inhibitors of NF-�B activation may have therapeutic potentialand are actively being researched.Non-steroidal anti-inflammatory drugs are commonly pre-

scribed for the treatment of tendinitis (20). However, the use ofnon-steroidal anti-inflammatory drugs is associated withnumerous side effects, which can be quite adverse. Therefore,the search is still on for safer andmore selective pharmacother-apies for tendinopathy. Curcumin (diferuloylmethane) is a nat-urally occurring polyphenol derived from the rhizome of Cur-cuma longa Linn, with the potential for treatment of variousdiseases acting via NF-�B inhibition (21–23). Commerciallyavailable preparations of curcumin contain three major com-ponents: curcumin (77%), demethoxycurcumin (17%), and bis-demethoxycurcumin (3%), altogether referred to as the “cur-cuminoids” (22, 24–28). Recent studies have shown thatcurcumin mediates its effects by modulation of several impor-tant molecular targets, including transcription factors (e.g.NF-�B,AP-1,�-catenin, and peroxisomeproliferator-activatedreceptor-�), enzymes (e.g. COX-2, 5-LOX, and iNOS), pro-in-flammatory cytokines (e.g. TNF-�, IL-1�, and IL-6), and cellsurface adhesion molecules. Because of its ability to modulatethe expression of these targets, the therapeutic potential of cur-cumin for treating cancer, arthritis, diabetes, Crohn disease,cardiovascular diseases, osteoporosis, Alzheimer disease, pso-riasis, and other pathologies is now under investigation (24, 28,29). Furthermore, curcumin has been studied in clinical trialsfor its anti-inflammatory, anti-carcinogenic, and free radicalscavenger properties (22). Phase I clinical trials have indicatedthat human subjects can tolerate curcumin doses as high as8–12 g/day with no adverse side effects (30, 31). Moreover,several aspects of the pharmacological properties and the use ofcurcumin for cancer chemoprevention have been reviewedrecently (32). Although curcumin is a potent inhibitor ofNF-�B, its effects on human tenocytes have not been investi-gated at the cellular or molecular levels.Phosphatidylinositol 3-kinases (PI-3Ks) are a highly con-

served family of kinases that catalyze the 3-position of the ino-sitol ring of phosphoinositides to generate phosphatidylinositol3-phosphate, phosphatidylinositol 3,4-bisphosphate, and phos-phatidylinositol 3,4,5-trisphosphate (33). PI-3K is a heterodi-meric lipid kinase consisting of an 85-kDa regulatory subunitand a 110-kDa catalytic subunit that plays a pivotal role in cellmovement, growth, vesicular trafficking, mitogenesis, and cellsurvival (34, 35). PI-3K is involved in the IL-1� signaling path-way and mediates activation and translocation of NF-�Bthrough targeting IKK-� or phosphorylation of p65, a processthat is inhibited by the PI-3K-specific inhibitor wortmannin(36, 37). Several reports suggest that PI-3K activates proteinkinase B (Akt), one of themain downstream kinases in cells (33,

38). However, the PI-3K/Akt signaling pathway has not yetbeen implicated in the activation of NF-�B in tenocytes.The aim of this study was to exploit an in vitro model of

human tenocytes to study themechanism of curcumin in IL-1�signaling and investigate whether curcumin might antagonizethe catabolic effects of pro-inflammatory cytokines by sup-pressing NF-�B-activation and NF-�B-induced gene expres-sion. We also explored the molecular mechanisms by whichcurcumin suppresses NF-�B activation in tenocytes, a processthat was partly mediated by the PI-3K/Akt signaling pathway.

EXPERIMENTAL PROCEDURES

Antibodies—The following antibodies were purchased fromChemicon International, Inc. (Temecula, CA): polyclonal anti-collagen type I, polyclonal anti-collagen type III, polyclonalanti-decorin antibody, and alkaline phosphatase linked sheepanti-mouse and sheep anti-rabbit secondary antibodies forimmunoblotting. Polyclonal anti-active caspase-3was obtainedfrom R&D Systems. Monoclonal antibody to �-actin, and pro-tein A/G-Sepharose beads were from Sigma. Polyclonal anti-tenomodulin (sc-49325) and anti-phosphatidylinositol 3-ki-nase (PI-3K) p85 antibodies were obtained from Santa CruzBiotechnology (Santa Cruz, CA). Polyclonal anti-scleraxis(SCXA) (ab58655) was obtained from Abcam (Cambridge,UK). Polyclonal anti-active caspase-3 (AF835) was purchasedfrom R&D Systems. Antibodies against phospho-specific I�B�(Ser-32/36) and against anti-phospho-specific p65 (NF-�B)/(Ser-536) were obtained from Cell Technology (Beverly, MA).Anti-I�B kinase (anti-IKK)-� and (anti-IKK)-� antibodies wereobtained from Imgenex (Hamburg, Germany). All antibodieswere used at concentrations and dilutions recommended by themanufacturer (dilutions ranged from 1:100 for immunomor-phological experiments to 1:10,000 for Western blot analysis).Secondary antibodies for immunofluorescence were purchasedfrom Dianova (Hamburg, Germany).Growth Medium and Chemicals—Growth medium (Ham’s

F-12/Dulbecco’s modified Eagle’s medium (50/50) containing10% fetal calf serum (FCS), 25 �g/ml of ascorbic acid, 50 IU/mlof streptomycin, 50 IU/ml of penicillin, 2.5 �g/ml of amphoter-icin B, essential amino acids, and L-glutamine) was obtainedfrom Seromed (Munich, Germany). Trypsin/EDTA (EC3.4.21.4) and wortmannin were purchased from Sigma. Eponwas obtained from Plano (Marburg, Germany). Curcumin witha purity of greater than 95% was purchased from Indsaff (Pun-jab, India). Curcumin was dissolved in dimethyl sulfoxide as astock concentration of 500 �M and stored at �80 °C. Serialdilutions were prepared in culture medium. Curcumin wasdiluted in dimethyl sulfoxide as a 5000 �M concentration andthen further diluted in cell culturemedium. IL-1�was obtainedfrom Acris Antibodies GmbH (Herold, Germany).Tenocyte Isolation and Culture—The peritendineum of

human tendon explants (healthy finger tendon of one malemiddle-aged donor), obtained during tendon-rupture surgery,was carefully removed before culturing in growth medium(Ham’s F-12/Dulbecco’s modified Eagle’s medium (50/50)) forseveral days (39). Tendon samples were derived from patientswith full informed consent and local ethics committeeapproval. After 1–2 weeks, tenocytes continuously migrated

Curcumin Inhibits NF-�B Signaling in Tenocytes

59

from this explant and adhered to Petri dishes. Tendon cellswere trypsinized and expanded in monolayers to gain a suffi-cient number of cells for the experiments described.Experimental Design—Tenocyte monolayer cultures were

washed three times with serum-starvedmedium and incubatedfor 1 h with serum-starved medium (3% FCS). Serum-starvedhuman tenocytes were either left untreated or treated with 10ng/ml of IL-1� alone for the indicated time periods or pre-treated with 5 �M curcumin or 5 �M curcumin for 4 h followedby co-treatment with 10 ng/ml of IL-1� and 5 �M curcumin forthe indicated time periods. For investigation of NF-�B translo-cation and I�B� phosphorylation, primary tenocyte monolayercultures were washed three times with serum-starved mediumand incubated for 1 h with serum-starved medium (3% FCS).The cells were then either treated with 5�M curcumin or 20 nMwortmannin or pre-treated with 5 �M curcumin or 20 nMwort-mannin for 0, 5, 10, 20, 40, and 60 min, followed by stimulationwith 10 ng/ml of IL-1� for 30 min. In separate experiments,tenocytes were preincubated with curcumin or wortmannin atthe indicated concentrations for 4 h followed by co-treatmentwith 10 ng/ml of IL-1� and curcumin or wortmannin for 30min. Nuclear and cytoplasmic extracts were then prepared asdescribed later. These experiments were performed in tripli-cate and the results are provided as mean values from threeindependent experiments.Immunofluorescence Analysis of NF-�B (p65) Localization—

The effect of curcumin on the nuclear translocation of p65 wasinvestigated by an immunocytochemical method as previouslydescribed in detail (28). Briefly, human tenocytes were culturedon glass coverslips and incubated for 24 h. Serum-starvedhuman tenocytes were treated with 10 ng/ml of IL-1� or 5 �M

curcumin, or were pre-treated with 5 �M curcumin for 4 h andthen co-treated with 10 ng/ml of IL-1� and curcumin for 30min in serum-starved medium. Glass coverslips with tenocytemonolayers were rinsed three times in Hanks’ solution beforemethanol fixation for 10min at ambient temperature, and rins-ing with PBS. Cell and nuclear membranes of tenocytes werepermeabilized by treatment with 0.1% Triton X-100 for 1 minon ice. Cells were overlaid with protease-free bovine serumalbumin (BSA) for 10 min at ambient temperature, rinsed withPBS, and incubatedwith primary antibodies (phospho-p65) in ahumid chamber overnight at 4 °C. They were gently washedseveral times with PBS before incubation with rhodamine red-conjugated secondary antibody for 2 h at ambient temperatureand finally washed again three times with Aqua dest. Counter-staining was performed with DAPI to visualize the cell nuclei.Samples were evaluated by light microscopy (Leica, Germany)and photomicrographs were digitally captured and stored.Western Blot Analysis—For Western blot analysis proteins

were extracted from the monolayer cultures with lysis buffer(50mMTris-HCl, pH 7.2, 150mMNaCl, 1% (v/v) Triton X-100,1 mM sodium orthovanadate, 50 mM sodium pyrophosphate,100mM sodium fluoride, 0.01% (v/v) aprotinin, 4 �g/ml of pep-statin A, 10 �g/ml of leupeptin, 1 mM phenylmethylsulfonylfluoride, PMSF) on ice for 30 min as previously described (40).Total protein concentration was measured with the bicincho-nic acid assay system (Uptima,Monlucon, France) using bovineserum albumin as a standard. Samples were further reduced

with 2-mercaptoethanol and equal quantities of protein (500ng/lane), separated under reducing conditions by SDS-PAGEusing 5, 7.5, 10, and 12%gels and transferred onto nitrocellulosemembranes using a transblot apparatus (Bio-Rad). After prein-cubation in blocking buffer (5% skimmed milk powder in PBS,0.1% Tween 20) for 1 h, membranes were incubated with pri-mary antibodies at 4 °C overnight, washed three times withblocking buffer, and then further incubated with alkaline phos-phatase-conjugated secondary antibodies for 2 h at ambienttemperature. After further washing in 0.1 M Tris, pH 9.5, con-taining 0.05 MMgCl2 and 0.1 MNaCl, specific antigen-antibodycomplexes were detected using nitro blue tetrazolium and5-bromo-4-chloro-3-indoylphosphate (p-toluidine salt; Pierce).Transmission Electron Microscopy—Electron microscopy

was performed as previously described (41). Briefly, monolayercultures were fixed for 1 h in Karnovsky fixative followed bypost-fixation in 1% OsO4 solution. After dehydration in anascending alcohol series, pellets were embedded in Epon andcut ultrathin with a Reichert-Jung Ultracut E (Darmstadt, Ger-many). Sections were contrasted with a mixture of 2% uranylacetate/lead citrate and examined with a transmission electronmicroscope (TEM 10, Zeiss, Jena, Germany).Isolation of Tenocyte Cytoplasmic and Nuclear Extracts—

Isolation of tenocyte cytoplasmic and nuclear extracts was per-formed as previously described (29). Briefly, tenocytes weretrypsinized and washed twice in 1 ml of ice-cold PBS. Thesupernatant was carefully removed. The cell pellet was resus-pended in hypotonic lysis buffer containing protease inhibitorsand was incubated on ice for 15 min. Then 12.5 �l of 10% Non-idet P-40 was added and the cell suspension was vigorouslymixed for 15 s. The extracts were centrifuged for 1.5 min. Thesupernatants (cytoplasmic extracts) were frozen at �70 °C. 25�l of ice-cold nuclear extraction buffer were added to the pel-lets and incubated for 30 min with intermittent mixing.Extracts were centrifuged, and the supernatant (nuclearextracts) was transferred to pre-chilled tubes for storage at�70 °C.Immune Complex Kinase Assay—To test the effect of curcu-

min on IL-1�-induced IKK activation, immune complex kinaseassays were performed as previously described (29). The IKKcomplex was immunoprecipitated from whole cell lysates withantibodies against IKK-� and IKK-� and subsequently incu-bated with protein A/G-agarose beads (Pierce). After 2 h ofincubation, the beads were washed with lysis buffer and resus-pended in a kinase assay solution containing 50mMHEPES, pH7.4, 20 mM MgCl2, 2 mM dithiothreitol, 10 �M unlabeled ATP,and 2 mg of substrate GST-I�B� (amino acids 1 to 54), andincubated at 30 °C for 30 min. Phosphorylation of GST-I�B�was assessed using a specific antibody against phospho-specificI�B� (Ser-32/36). To demonstrate the total amounts of IKK-�and IKK-� in each sample, whole cell lysateswere transferred toa nitrocellulose membrane after SDS-PAGE under reducingconditions as described above. Detection of IKK-� and IKK-�was performed by immunoblotting with either anti-IKK-� oranti-IKK-� antibodies.Immunoprecipitation of p65 and p65 Acetylation Assay—To

examine the effect of curcumin on IL-1�-induced acetylation ofp65, tenocytes were pre-treatedwith 5�M curcumin for 4 h and

Curcumin Inhibits NF-�B Signaling in Tenocytes

60

then exposed to 10 ng/ml of IL-1� for the indicated times. Thecells were washed with ice-cold phosphate-buffered saline(PBS), and lysed in a buffer containing 50mMTris-HCl, pH 7.2,150 mMNaCl, 1% (v/v) Triton X-100, 1 mM sodium orthovana-date, 50 mM sodium pyrophosphate, 100 mM sodium fluoride,0.01% (v/v) aprotinin, 4 �g/ml of pepstatin A, 10 �g/ml of leu-peptin, 1 mM phenylmethylsulfonyl fluoride (PMSF) to preparewhole cell lysates.Whole cell extracts were pre-cleared by incu-bating with 25 �l of either normal rabbit IgG serum or normalmouse IgG serum and protein A/G-Sepharose beads, then withprimary antibodies (anti-p65 antibody) appropriately diluted inwash buffer (0.1% Tween 20, 150 mM NaCl, 50 mM Tris-HCl(pH 7.2), 1 mM CaCl2, 1 mM MgCl2, and 1 mM PMSF) for 2 h at4 °C. After 1 h of incubation, immunocomplexes were washedwith lysis buffer, boiled with SDS sample buffer for 5 min,resolved on SDS-PAGE, and subjected toWestern blot analysisusing an anti-acetyl-lysine antibody.Pharmacological Experiments with Wortmannin—Teno-

cytes were grown in growth medium for 24 h. PI-3K inhibitionexperiments were carried out in serum-starvedmedium. Teno-cytes were pre-treated for 1 hwith serum-starvedmedium con-tainingwortmannin (1, 10, and 20 nM) for 1 h, treatedwith 5�M

curcumin for 4 h, and then exposed to 10 ng/ml of IL-1� for 1 h.After these treatments, nuclear extracts were prepared andexamined for NF-�B as described above.Statistical Analysis—The optical density (specific binding) of

each protein band was measured and semiquantitatively ana-lyzed using the Quantity One software package (Bio-Rad). Theresults are shown as the mean � S.D. of a representative exper-iment performed in triplicate.

RESULTS

Cell Culture—This study was undertaken to investigate theeffect of curcumin on the PI-3K/Akt signaling pathway leadingto activation of the transcription factor NF-�B signaling path-way and on NF-�B-regulated gene products in human teno-cytes in an in vitro model of tendinopathy. Tenocytes treatedwith curcumin (5–20�M) orwortmannin (1–40nM) showednosigns of cytotoxic effects or any negative effects on cell viability(data not shown) at the light microscopic and ultrastructurallevels. To examine the effect of curcumin on the NF-�B activa-tion pathway, we used IL-1� because the pathway activated bythis cytokine is relatively well understood.Curcumin Inhibits IL-1�-induced Degenerative Features and

Apoptosis in Tenocytes—Untreated monolayer tenocytesexhibited typically flattened shapeswith small cytoplasmic pro-cesses, largemostly euchromatic nuclei with nucleoli and awellstructured cytoplasm (Fig. 1A,a). Treatment of tenocytemono-layer cultures with 10 ng/ml of IL-1� for 12, 24, and 48 h led todegenerative changes such as multiple vacuoles, swelling ofrough endoplasmic reticulum, and clustering of swollen mito-chondria, condensed heterochromatin in the cell nuclei, andmultiple autophagic cytoplasmic vacuoles. The flattenedmonolayer tenocytes became more and more rounded, losttheir microvilli-like processes, and became apoptotic (Fig. 1A,b–d).Tenocytes that were pretreatedwith curcumin (4 h) and then

co-treated with IL-1� and 5 �M curcumin for 12, 24, and 48 h

showed less severe cellular degeneration at the ultrastructurallevel as early as 12 h after co-treatment (Fig. 1A, e–g). Thetenocytes regained a flattened shape and numerous microvilli-like cytoplasmic processes.Statistical evaluation of the data clearly highlighted changes

in the number of cells with mitochondrial changes before andafter IL-1�-treatment. Co-treatment with curcumin clearlydecreased the number of cells with mitochondrial changes(Fig. 1B).Effect of Curcumin on IL-1�-induced Inhibition of Extracellular

Matrix and Signaling Protein Expression in Tenocytes—Serum-starved human tenocytes were cultured for 24 h and thentreated with 10 ng/ml of IL-1� or 5 �M curcumin or were pre-treated with 5 �M curcumin for 4 h and then co-treated with 10ng/ml of IL-1� and curcumin or left untreated and evaluatedafter 24 h. Western blotting was performed by probing wholecell lysates with antibodies against collagen types I and III,decorin, tenomodulin, and the tendon-specific transcriptionfactor SCXA (Fig. 2). Primary human tenocytes stimulatedwithIL-1� alone showed a significant down-regulation of synthesisof collagen types I, III, decorin, tenomodulin, and SCXAexpression (Fig. 2). In contrast, pre-treatment of tenocytes withcurcumin followed by stimulation with IL-1� resulted in aninhibition of cytokine-induced effects on the above mentionedprotein production (Fig. 2). Synthesis of the housekeeping protein�-actin remained unaffected (Fig. 2).Curcumin Inhibits IL-1�-induced NF-�B-dependent Pro-

inflammatory and Matrix Degradation Gene Products inTenocytes—We examined whether curcumin canmodulate theexpression of IL-1�-induced NF-�B-regulated gene productsinvolved in the inflammatory and degradative processes intenocytes. Primary human tenocytes with or without pre-treat-ment with curcumin were examined for IL-1�-induced geneproducts by Western blot analysis using specific antibodies(Fig. 3A). IL-1� induced the expression of COX-2, MMP-1,MMP-9, and MMP-13, but treatment with curcumin inhibitedthe expression of these proteins in primary tenocytes (Fig. 3A).Effect of Curcumin on IL-1�-inducedNF-�B-dependent Anti-

apoptotic and Pro-apoptotic Gene Products in Tenocytes—It isknown that NF-�B regulates the expression of anti-apoptoticproteins Bcl-2, Bcl-xL, and TRAF1 (TNF receptor-associatedfactor 1) (42, 43).We investigated whether curcumin canmod-ulate the expression of these anti-apoptotic gene products.IL-1�-stimulated primary human tenocytes were examined byWestern blot analysis with or without curcumin pre-treatment(Fig. 3B). As shown in Fig. 3B, IL-1� inhibited the expression ofBcl-2, Bcl-xL, and TRAF1 in a time-dependentmanner. In con-trast, curcumin stimulated the expression of these anti-apo-ptotic proteins (Fig. 3B). To determine whether curcumininhibits the IL-1�-induced pro-apoptotic gene product, Baxand activated caspase-3, in the same cell cultures, tenocyteswere incubated with IL-1� (10 ng/ml) alone for the indicatedtime or preincubated with curcumin (5 �M) for 4 h and thenco-treated with IL-1� (10 ng/ml) for the indicated time. Asshown in Fig. 3C, pre-treatment with curcumin significantlydown-regulated the level of Bax and biologically activecaspase-3 in IL-1�-stimulated cultures compared with humantenocytes stimulated with IL-1� alone.

Curcumin Inhibits NF-�B Signaling in Tenocytes

61

Curcumin Suppresses IL-1�-induced Activation and NuclearTranslocation of NF-�B in a Concentration- and Time-depen-dent Manner in Tenocytes—To evaluate whether curcumininhibits the IL-1�-induced activation and nuclear translocationof NF-�B, nuclear protein extracts from serum-starved teno-cytes were probed for the phosphorylated form of the p65NF-�B subunit after either the cells were stimulated with 10ng/ml of IL-1� for the indicated times or pre-treated with 5 �M

curcumin for the indicated times followed by stimulation with10 ng/ml of IL-1� for 30 min (Fig. 4A).IL-1� induced p65 phosphorylation in the nuclear fraction in

a time-dependent manner (Fig. 4A, left panel). Curcuminblocked IL-1�-induced translocation of p65 to the nucleus in atime-dependent manner (Fig. 4A, right panel). Tenocytes wereeither incubated with curcumin at various concentrations for4 h or preincubated with curcumin at various concentrations

for 4 h followed by 10 ng/ml of IL-1� stimulation for 30 min.The nuclear extracts were probed for phospho-p65 byWesternblot analysis (Fig. 4B). Curcumin inhibited IL-1�-inducedNF-�B activation in a concentration-dependent manner (Fig.4B, right panel).PI-3K Signaling Pathway Is Involved in the IL-1�-mediated

NF-�B Activation in Tenocytes—To examine a possible func-tional relationship between inhibition of the PI-3K pathwayand suppression of NF-�B activation in response to curcumintreatment, tenocytes were either stimulated with 20 nM wort-mannin, a specific inhibitor of PI-3K, for the indicated times orpreincubated with 20 nM wortmannin for the indicated timesfollowed by 10 ng/ml of IL-1� stimulation for 30min. Tenocytenuclear protein extracts were probed for the phosphorylatedform of the p65 NF-�B subunit (Fig. 5A). In separate experi-ments, tenocytes were preincubated with wortmannin at the

FIGURE 1. Effect of curcumin on IL-1�-induced cellular degeneration and apoptosis in tenocytes. A, electron microscopy was performed to demonstratethe effects of curcumin on IL-1�-stimulated tenocytes in monolayer culture. Untreated control tenocytes containing mitochondria, rough endoplasmicreticulum, and many other cell organelles are in panel a. In contrast, stimulation of tenocytes with 10 ng/ml of IL-1� for 12, 24, and 48 h resulted in degenerativechanges of the cells. After 12 h, tenocytes became rounded and the nucleus contained more condensed chromatin (b). After 24 h multiple vacuoles, swellingof rough endoplasmic reticulum, and clustering of swollen mitochondria was visible (c). Longer incubations of 48 h led to the formation of apoptotic bodies andcell lysis (d). However, pre-treatment of IL-1�-stimulated tenocytes with curcumin inhibited the adverse effects of IL-1� (e– g) and after 48 h treatment (g)tenocytes demonstrated large, flattened cells with numerous microvilli-like processes and mitochondria were comparable with control cultures. Bar, 1 �M. B, toquantify cellular degradation and apoptosis in these cultures, 100 cells from 20 microscopic fields were counted. The number of degraded and apoptotic cellswas highest in cultures stimulated with IL-1� alone. In contrast to this, pre-treatment of IL-1�-stimulated cultures with curcumin inhibited the cellulardegradation and apoptotic effects of IL-1� and the number of degraded and apoptotic cells remained significantly lower over the entire culture period (*).

Curcumin Inhibits NF-�B Signaling in Tenocytes

62

indicated concentrations for 4 h followed by co-treatment with10 ng/ml of IL-1� and wortmannin for 30 min (Fig. 5B). Asshown in Fig. 5, A and B, wortmannin substantially inhibitedIL-1�-induced NF-�B activation in a time- and concentration-dependent manner, suggesting that the PI-3K signaling path-way is functionally involved in the process of IL-1�-inducedactivation of NF-�B. Wortmannin alone had no effect on theNF-�B activation (Fig. 5A, left panel).Curcumin Inhibits IL-1�-dependent I�B� Degradation and

Phosphorylation—It is well known that an important prerequi-site for the activation of NF-�B is the phosphorylation and deg-radation of I�B�, the natural blocker of NF-�B (17, 42, 44). Toexamine whether the inhibitory activity of curcumin wasthrough inhibition of I�B� degradation, cells were treated withcurcumin, followed by IL-1�, and subsequently probed forNF-�B expression in the nucleus and I�B� expression in thecytoplasm by Western blot analysis. IL-1� induced I�B� deg-radation in control cells within 10min but not at all in curcum-in-treated cells (Fig. 6A, row I). These results indicate that cur-cumin blocks IL-1�-induced I�B�degradation.Data shown arerepresentative of three independent experiments.Next, we examined whether inhibition of IL-1�-induced

I�B� degradation was through inhibition of I�B� phosphory-lation, the tenocytes were treated with curcumin and then withIL-1� and examined for I�B�phosphorylation in the cytoplasmbyWestern blot analysis. IL-1�-induced I�B-� phosphorylationwas almost completely blocked by curcumin (Fig. 6A, row II).

Curcumin Inhibits IL-1�-induced Nuclear Translocation ofNF-�B (p65)—Pro-inflammatory cytokines induce the phos-phorylation of p65, which is required for NF-�B transcriptionalactivity. The phosphorylation of NF-�B is known to be medi-ated through IKK (28, 45, 46). In addition, cytoplasmic extractswere examined for expression of pan-/phospho-p65 (Fig. 6A,rows III and IV). Western blot analysis confirmed the IL-1�-induced phosphorylation of the cytoplasmic pool of p65 in atime-dependent manner, and p65 phosphorylation could beseen as early as 5 min and increased up to 40 min (Fig. 6A, rowIV). Pre-treatment with curcumin inhibited the IL-1�-inducedphosphorylation of cytoplasmic p65. IL-1� also induced p65phosphorylation in the nuclear fraction in a time-dependentmanner, and curcumin blocked IL-1�-induced translocation ofp65 to the nucleus (Fig. 6B, rows I and II).

The immunocytochemical analysis confirmed the Westernblot findings that curcumin blocked the translocation of p65from the cytoplasm to the nucleus. The p65 subunit of NF-�Bwas localized in the cytoplasm of untreated cells. IL-1� inducednuclear translocation of p65 and curcumin blocked the trans-location (Fig. 7).Effect of Curcumin on IL-1�-induced Acetylation of NF-�B in

Tenocytes—The acetylation of p65 plays an essential role inI�B�-mediated activation ofNF-�B transcriptional activity (47,48). To investigate the effect of curcumin on the acetylation ofp65 by IL-1�, human tenocytes were pretreated with curcuminfor 4 h and then exposed to IL-1� for the indicated times.Whole cell extracts were examined and immunoprecipitatedwith anti-p65 antibody, and Western blot analysis was per-formed using anti-acetyl-lysine antibody. As shown in Fig. 8,IL-1� induced the acetylation of p65 in a time-dependentman-ner, and curcumin completely suppressed the IL-1�-inducedacetylation of p65 (Fig. 8).Effect of Curcumin on IL-1�-induced Activation of IKK—It

has been shown that IKK is required for TNF-induced Phos-phorylation of I�B� (44). We further evaluated the effect ofcurcumin on IL-1�-induced IKK activation, which is requiredfor IL-1�-induced phosphorylation of I�B�. The results fromthe immune complex kinase assay showed that IL-1� inducedthe activation of IKK in a time-dependent manner and thatcurcumin blocked IL-1�-activated IKK (Fig. 9A, lane I). IL-1�or curcumin had no direct effect on the expression of IKK-� orIKK-� proteins (Fig. 9A, lanes II and III).

Next, to examine further whether curcumin blocks IKKactivity directly by binding IKK or indirectly by inhibition of itsactivation, human tenocytes were treated with IL-1� or leftuntreated. Whole cell extracts from untreated cells and IL-1�-stimulated cells were incubated with anti-IKK-� antibody.After precipitation with protein A/G-Sephadex beads, theimmunocomplexes were exposed with various concentrationsof curcumin. As shown in Fig. 8B, the immunocomplex kinaseassay revealed that curcumin directly suppressed the activity ofIKK (Fig. 9B, row I). IL-1� or curcumin had no direct effect onthe expression of IKK-� or IKK-� proteins (Fig. 9B, rows IIand III).Curcumin Blocks IL-1�-inducedAkt Stimulation—The stim-

ulation of IKK-� is required for the phosphorylation of I�B�.Moreover, it has been reported that pro-inflammatory cyto-

FIGURE 2. Effects of curcumin on IL-1�-induced inhibition of extracellularmatrix compounds and signaling proteins expression in tenocytes. Toevaluate the effects of curcumin on IL-1�-stimulated tenogenic inhibition intenocytes, whole cell lysates (500 ng of protein/lane) were probed with anti-bodies to collagen type I (CI), collagen type III (CIII), decorin, tenomodulin(Tnmd), and the tenogenic specific transcription factor SCXA. Serum-starvedhuman tenocytes (1 � 106 cells/ml) were exposed to 10 ng/ml of IL-1� alone,5 �M curcumin alone, or were pre-treated with 5 �M curcumin for 4 h andfollowed either by incubation with IL-1� alone or incubation with IL-1� andcurcumin, or left untreated for 24 h. Each experiment was performed in trip-licate. Expression of the �-actin housekeeping gene was not affected by IL-1�and/or curcumin.

Curcumin Inhibits NF-�B Signaling in Tenocytes

63

kine-mediated stimulation of IKK-� is associated with anupstream protein kinase, Akt (serine-threonine kinase, proteinkinase B), and in turn Akt mediates IKK-� phosphorylation(49). To investigate whether curcumin blocks the IL-1�-in-duced I�B� phosphorylation due to inhibition of Akt, theserum-starved tenocytes were treatedwith IL-1� (10 ng/ml) fordifferent indicated times or pre-treated with curcumin (5 �M)

for 4 h and then co-treated with IL-1� for the indicated times.The whole cell extracts were analyzed by Western blottingusing anti-phospho-specific-Akt antibody. As shown in Fig. 10row I, IL-1� induced activation of Akt in a time-dependentmanner. In contrast, in co-treated tenocytes, curcumin clearlyinhibited the phosphorylation of Akt.Next, we explored whether curcumin affects the association

of Akt with IKK.Whole cell lysates from human tenocytes pre-treated with curcumin and co-treated with IL-1� were immu-

FIGURE 3. Effects of curcumin on IL-1�-induced NF-�B-dependent pro-inflammatory, anti-apoptotic, and pro-apoptotic gene products in tenocytes.A, to evaluate whether curcumin exerts effects on IL-1�-induced NF-�B-dependent expression of pro-inflammatory (COX-2) and matrix-degrading (MMP-1, -9,and -13) gene products, serum-starved tenocytes (1 � 106 cells/ml) were exposed to 10 ng/ml of IL-1� alone, 5 �M curcumin alone, or were pre-treated with 5�M curcumin for 4 h and followed either by incubation with IL-1� alone or incubation with IL-1� and curcumin, or left untreated for 24 h. Each experiment wasperformed in triplicate. Expression of the �-actin housekeeping gene was not affected by IL-1� and/or curcumin. B, to determine whether curcumin treatmentactively stimulates the production of anti-apoptotic gene products (Bcl-2, TRAF1, and Bcl-xL), serum-starved human tenocytes (1 � 106 cells/ml) were exposedto 10 ng/ml of IL-1� alone, 5 �M curcumin alone, or were pre-treated with 5 �M curcumin for 4 h and followed either by incubation with IL-1� alone orincubation with IL-1� and curcumin, or left untreated for 24 h. Synthesis of the �-actin housekeeping gene remained unaffected. C, whole cell lysates ofserum-starved human tenocytes (1 � 106 cells/ml) were exposed to 10 ng/ml of IL-1� alone, 5 �M curcumin alone, or were pre-treated with 5 �M curcumin for4 h and followed either by incubation with IL-1� alone or incubation with IL-1� and curcumin, or left untreated and evaluated with Western blot analysis toexamine the effect on the pro-apoptotic proteins Bax and caspase-3. Expression of the housekeeping protein �-actin remained unaffected.

FIGURE 4. Curcumin suppression of IL-1�-induced NF-�B activation.A, serum-starved human tenocytes (1 � 106 cells/ml) were either stimulatedwith 10 ng/ml of IL-1� for the indicated times or preincubated with 5 �M

curcumin for 0, 5, 10, 20, 40, and 60 min, co-treated with 10 ng/ml of IL-1� for30 min, and then probed for phospho-p65 by Western blot analysis usingantibodies to phospho-specific p65 and poly(ADP-ribose) polymerase (PARP)(control). B, serum-starved human tenocytes were either incubated with cur-cumin at various concentrations (0, 1, 2, 5, 10, and 15 �M) for 4 h or preincu-bated with curcumin at various concentrations for 4 h followed by 10 ng/ml ofIL-1� stimulation for 30 min. The nuclear extracts (500 ng of protein/lane)were probed for phospho-p65 by Western blot analysis using antibodies tophospho-specific p65 and PARP (control). Expression of PARP remained unaf-fected in the nuclear extracts. The results shown are representative of threeindependent experiments.

FIGURE 5. Effect of wortmannin on IL-1�-induced NF-�B activation.A, serum-starved human tenocytes (1 � 106 cells/ml) were either stimulatedwith 20 nM wortmannin for 0, 5, 10, 20, 40, and 60 min or preincubated with 20 nM

wortmannin for 0, 5, 10, 20, 40, and 60 min, co-treated with 10 ng/ml of IL-1� for30 min, and then probed for phospho-p65 by Western blotting using antibodiesto phospho-specific p65 and poly(ADP-ribose) polymerase (PARP) (control). B,serum-starved tenocytes were preincubated with wortmannin at various con-centrations (0, 1, 10, 20, 30, and 40 nM) for 1 h followed by 10 ng/ml of IL-1�stimulation for 30 min. Nuclear extracts (500 ng of protein/lane) were probed forphospho-p65 by Western blotting using antibodies to phospho-specific p65 andPARP (control). Expression of PARP remained unaffected in nuclear extracts. Theresults shown are representative of three independent experiments.

Curcumin Inhibits NF-�B Signaling in Tenocytes

64

noprecipitated with anti-IKK-� antibody followed by Westernblot analysis using anti-Akt antibody. As shown in Fig. 10, rowII, IL-1� induced a combination between IKK-� and Akt in atime-dependent fashion, and this was clearly inhibited by cur-

cumin. Taken together, these results indicate that curcuminblocks IKK-� activation through inhibition of an upstreampro-tein kinase B (Akt).Curcumin Modulates IL-1�-induced NF-�B Activation by

Inhibition of the PI-3K/p85 Signaling Pathway in Tenocytes—To gain a mechanistic insight into the mode of action of curcu-min on IL-1�-stimulated tenocytes we studied the effects ofcurcumin on PI-3K activation using the Akt/IKK assay. Akt is awell known downstream protein kinase B target of PI-3K, andits activation ismainly induced by the phosphorylation of Ser orThr residues (50). Previous work has shown that the PI-3K sig-naling pathway is inhibited by curcumin treatment in MCF7cells (51). Therefore, to investigate whether the effect of curcu-min on the IL-1�-induced Akt phosphorylation is mediatedthrough inhibition of the PI-3K signaling pathway in humantenocytes, the serum-starved tenocytes were treatedwith IL-1�(10 ng/ml) for the times indicated or pre-treatedwith curcumin

FIGURE 6. Effect of curcumin on the IL-1�-induced phosphorylation anddegradation of I�B� and the phosphorylation and translocation of p65.Serum-starved human tenocytes (1 � 106 cells/ml) were either stimulated with10 ng/ml of IL-1� for the indicated times or pre-treated with 5 �M curcumin for4 h, followed by 10 ng/ml of IL-1� stimulation for the indicated times. Cytoplas-mic (A) and nuclear extracts (B) were prepared, fractionated with SDS-PAGE, andelectrotransferred to nitrocellulose membrane. Western blot analysis was per-formed with anti-I�B�, anti-phosphospecific I�B�, anti-p65, and anti-phospho-specific p65 antibodies. The results shown are representative of three indepen-dent experiments. PARP, poly(ADP-ribose) polymerase.

FIGURE 7. Immunocytochemical analysis of p65 localization after treatment with IL-1� as revealed by immunofluorescence microscopy. Tenocytecultures either served as controls (not treated) or were treated with IL-1� alone for 10 min or co-treated with 5 �M curcumin and 10 ng/ml of IL-1� for 40 minbefore immunolabeling with phospho-p65 antibodies and rhodamine-coupled secondary antibodies. Data shown are representative of three independentexperiments. Original magnification, �160.

FIGURE 8. Effect of curcumin on IL-1�-induced acetylation of p65. Serum-starved human tenocytes (1 � 106 cells/ml) were either stimulated with 10ng/ml of IL-1� for the indicated times or pre-treated with 5 �M curcumin for4 h and then exposed to 10 ng/ml of IL-1�. Whole cell extracts were prepared,immunoprecipitated (IP) with anti-p65 antibody, and subjected to Westernblot analysis using anti-acetyl-lysine antibody. The same blots werere-probed with anti-p65 antibody. IB, immunoblot.

Curcumin Inhibits NF-�B Signaling in Tenocytes

65

(5 �M) for 4 h and then co-treated with IL-1�. Whole cellextracts were probed by Western blot analysis using a PI-3K/p85 subunit antibody. As shown in Fig. 11, row I, IL-1�-inducedactivation of PI-3K/p85 in a time-dependent manner. In con-trast, in co-treated tenocytes, curcumin substantially inhibitedthe activation of PI-3K/p85. Consistent with this observation,IL-1� significantly induced Akt phosphorylation, which wasreduced by curcumin in a dose-dependent manner (Fig. 10).Taken together, these findings suggest that curcumin mediatesits anti-inflammatory effects, at least in part, through modula-tion of the PI-3K/Akt signaling pathway in human tenocytes.Expression of the housekeeping protein �-actin was unaffected(Fig. 11, row II).Suppression of PI-3K Signaling Supports the Inhibitory Effect

of Curcumin—Themechanism involved in curcumin-mediatedinhibition of IL-1�-induced NF-�B activation in human teno-cytes was further investigated. It is known that the PI-3K path-way is required for activation of NF-�B by IL-1� and wortman-nin is a blocker of PI-3K signaling (36). Pre-treatment oftenocytes with different concentrations of wortmannin (1, 10,and 20 nM) for 1 h, treated with curcumin (5 �M) for 4 h, andthen treated with IL-1� for 1 h, inhibited the IL-1�-induced

NF-�B activation. The inhibitory effects of wortmannin andcurcumin on IL-1�-stimulated human tenocytes appeared tobe synergistic (Fig. 12, row I).

DISCUSSION

The purpose of this study was to investigate the effects ofcurcumin on the IL-1�-inducedNF-�B activation pathway andNF-�B-regulated gene products that influence inflammationin tendon. Using monolayer-cultured human tenocytes, wefound that curcumin inhibited IL-1�-induced NF-�B activa-tion through suppression of I�B� phosphorylation, I�B� deg-radation, I�B� kinase activity, and NF-�B-dependent geneproducts involved in inflammation (COX-2), in extracellularmatrix degradation (MMPs), apoptosis (Bcl-2, Bcl-xL, andTRFA-1), and activation of apoptosis (i.e. activation of caspase-3). This inhibitionwas correlatedwith suppression of p65 phos-phorylation, p65 nuclear translocation, and p65 acetylation.We also demonstrated that the PI-3K/Akt signaling pathway isactivated in response to IL-1� and suppression of IL-1�-in-duced NF-�B activation by curcumin appears to involve thePI-3K/Akt pathway and its association with IKK.Tendinopathy is accompanied by inflammation and degra-

dation of the tendon extracellular matrix. At a tendon injurysite, pro-inflammatory cytokines such as IL-1� may initiate acascade of events leading to tendon destruction and loss of bio-mechanical structural integrity. Furthermore, besides the up-regulation of inflammatorymediators, we found that IL-1� sig-nificantly down-regulates the expression of collagen types I andIII, decorin, and tenomodulin in tenocytes. Thus, IL-1�-medi-ated suppression of collagen type I and other tendon-specificextracellular matrix compound expression may lead to thereduced deposition of extracellular matrix and consequently, itmight affect normal tissue remodeling and lead to the develop-ment of tendinopathy. The control of IL-1� secretion may becritical for protecting tendons from pathological processes. Infact, human tenocytes express IL-1� receptors so that theligand-receptor signal is transduced via the specific and func-FIGURE 9. A, effects of curcumin treatment on IL-1�-induced I�B kinase acti-

vation. Serum-starved primary human tenocytes were either stimulated with10 ng/ml of IL-1� for the indicated times or pre-treated with 5 �M curcuminfor 4 h and then co-treated with IL-1� (10 ng/ml) for the indicated times.Whole cell extracts were immunoprecipitated with an antibody against I�Bkinase (IKK)-� and analyzed by an immune complex kinase assay. To examinethe effect of curcumin on the expression level of IKK proteins, whole cellextracts (500 ng of protein/lane) were fractionated by SDS-PAGE and exam-ined using Western blot analysis with anti-IKK-� and anti-IKK-� antibodies.Data shown are representative of three independent experiments. B, directeffect of curcumin treatment on IL-1�-induced I�B kinase activation. Serum-starved human tenocytes (1 � 106 cells/ml) were treated with 10 ng/ml ofIL-1�. Whole cell extracts were prepared and immunoprecipitated with anti-IKK-� antibody. The immunocomplex kinase assay was performed in theabsence or presence of curcumin at the indicated concentrations. Datashown are representative of three independent experiments.

FIGURE 10. Effect of curcumin on IL-1�-induced Akt phosphorylation. Serum-starved human tenocytes (1 � 106 cells/ml) were either stimulated with 10ng/ml of IL-1� for the indicated times or pre-treated with 5 �M curcumin for 4 h and exposed to 10 ng/ml of IL-1� for the indicated times. Whole cell extractswere analyzed by Western blot analysis using anti-phospho-specific Akt (row I). Cell extracts were immunoprecipitated (IP) with anti-IKK-� antibody and theimmunoprecipitates were separated (500 ng of protein/lane) by SDS-PAGE and analyzed by immunoblotting (IB) using anti-Akt antibody (row II) or withanti-IKK-� antibody (row III, as a loading control). Results shown are representative of three independent experiments.

FIGURE 11. Effects of curcumin treatment on IL-1�-induced PI-3K/p85.Serum-starved human tenocytes (1 � 106 cells/ml) were either stimulated with10 ng/ml of IL-1� for the indicated times or pre-treated with 5�M curcumin for 4 hand exposed to 10 ng/ml of IL-1� for the indicated times. Whole cell extracts (500ng of protein/lane) were analyzed by Western blot analysis using anti-PI-3K/p85.Expression of the housekeeping protein �-actin remained unaffected. Resultsshown are representative of three independent experiments.

Curcumin Inhibits NF-�B Signaling in Tenocytes

66

tional IL-1� receptor.Moreover, IL-1� activates numerous sig-nal transduction systems through protein kinases and thiscauses induction of genes by activation and suppression of spe-cific transcription factors such as NF-�B (52).We found that curcumin suppresses the activation of NF-�B

in human tenocytes in vitro and inhibits the expression ofNF-�B-regulated gene products, including COX-2, MMPs,Bax, and caspase-3. Curcumin can both stimulate and inhibitapoptotic signaling, and the treatment time as well as the con-centration may determine the effects of curcumin on variouscell types (53). Although curcumin has been shown to inhibitcytokine-induced NF-�B activation in many different pri-mary cells and cell lines of various origins (22, 25, 28, 29,54–57), to the best of our knowledge, this is the first suchreport in human tenocytes.A possible mechanism underlying the inhibition of inducible

NF-�B by curcumin could be its capacity to inhibit the PI-3Kand Akt signaling pathways. Previous studies using other cellshave shown that curcumin inhibits the DNA binding functionofNF-�B through suppression of I�B� phosphorylation (25, 29,58). Thus, down-regulation of upstream signaling proteins,such as PI-3K/Akt, may be involved in curcumin-mediatedactivation of IL-1�-induced NF-�B inhibition in tenocytes. Infact it has been reported that the PI-3K pathway is required foractivation ofNF-�Bby cytokines such as IL-1� (36). Our resultsdemonstrate that wortmannin, a specific inhibitor of the PI-3Kpathway, inhibits NF-�B activation and its translocation in thenucleus in tenocytes. These observations suggest that the PI-3Kpathway may be involved in IL-1� signaling. Furthermore,IL-1�-induced activation of PI-3K/p85 and Akt could be inhib-ited clearly by curcumin in human tenocytes. We found thatinhibition of PI-3K/p85 by curcumin, a process required forAktactivation, inhibits IKK and phosphorylation of both I�B� andp65. We have also shown that curcumin stimulates the expres-sion of several anti-apoptotic proteins that are regulated byNF-�B, including Bcl-2, Bcl-xL, and TRAF1. Curcumin alsoinhibits the pro-apoptotic protein caspase-3, the matrixdegrading MMPs, as well as the inflammatory enzyme COX-2.This is consistent with previous reports that have shown thatNF-�B activation requires the PI-3K/Akt signaling pathway(49, 59, 60) and that curcumin suppresses the expression ofpro-apoptotic proteins (Bax and caspase-3) andmatrix degrad-ing enzymes (MMPs) and mediators of inflammation (COX-2)(61, 62). Several groups have demonstrated that curcumin canalso inhibit the two subunits of PI-3K (p110 and p85) and phos-phorylation of the Akt signaling pathway (63, 64). These find-

ings might explain the anti-inflammatory and anti-apoptoticeffects of curcumin in tenocytes.It is well known that the tendon-specific transcription factor

SCXA is required for expression of tendon-specific extracellu-lar matrix genes (65). We also observed a reduction in collagentype I and SCXA expression in tenocytes after treatment withIL-1�. However, curcumin pre-treatment inhibited the IL-1�-induced down-regulation of collagen type I and SCXA expres-sion. Thus, curcumin stimulated tenocytes, at least in part,through activation of the tenogenic transcription factor scler-axis, enhancing transcription of tendon-associated collagens ina SCXA-dependent fashion.Moreover, curcumin did not exert any toxicity on the cells.

Studies on the phase I clinical trials suggest that curcumin canbe orally administered safely at doses of 0.2–12 g/day with nodose-limiting toxicity, reaching peak serum concentration at1–2 h (0.51 � 0.11 �M at 4000 mg, 0.63 � 0.06 �M at 6000 mg,and 1.77 � 1.87 �M at 8000 mg) and is eliminated within 12 h(30, 31, 66, 67). Recently, a phase II study of this agent hasshown that this compound has biological activity in patientswith pancreatic cancer (68).Overall, our data suggest that curcumin down-regulates

NF-�BandNF-�B-regulated gene products involved in apopto-sis, matrix degradation, and inflammation in human tenocytesin vitro. These effects are mediated, at least in part, throughdown-regulation of PI-3K/Akt signaling. This study providesadditional support for designing anti-inflammatory com-pounds based on curcumin for diseases mediated throughNF-�B activation. Therefore, curcumin might have prophylac-tic potential for the treatment of tendinitis.

Acknowledgments—Katharina Sperling and Ursula Schwikowski aregratefully acknowledged for excellent technical assistance.

REFERENCES1. Maffulli, N.,Wong, J., and Almekinders, L. C. (2003)Clin. Sports Med. 22,

675–6922. Almekinders, L. C. (1998) J. Am. Acad. Orthop. Surg. 6, 157–1643. Riley, G. (2004) Rheumatology 43, 131–1424. Sendzik, J., Shakibaei, M., Schafer-Korting, M., Lode, H., and Stahlmann,

R. (2010) Int. J. Antimicrob. Agents 35, 366–3745. Sendzik, J., Shakibaei, M., Schafer-Korting, M., and Stahlmann, R. (2005)

Toxicology 212, 24–366. Bernard-Beaubois, K., Hecquet, C., Houcine, O., Hayem, G., andAdolphe,

M. (1997) Cell Biol. Toxicol. 13, 103–1137. Kannus, P. (2000) Scand. J. Med. Sci. Sports 10, 312–3208. Rees, S. G., Flannery, C. R., Little, C. B., Hughes, C. E., Caterson, B., and

Dent, C. M. (2000) Biochem. J. 350, 181–1889. Butler, D. L., Juncosa, N., and Dressler, M. R. (2004) Annu. Rev. Biomed.

Eng. 6, 303–32910. Tsuzaki, M., Guyton, G., Garrett, W., Archambault, J. M., Herzog, W.,

Almekinders, L., Bynum, D., Yang, X., and Banes, A. J. (2003) J. Orthop.Res. 21, 256–264

11. Archambault, J., Tsuzaki,M.,Herzog,W., andBanes, A. J. (2002) J. Orthop.Res. 20, 36–39

12. Tsuzaki, M., Bynum, D., Almekinders, L., Yang, X., Faber, J., and Banes,A. J. (2003) J. Cell Biochem. 89, 556–562

13. Gotoh, M., Hamada, K., Yamakawa, H., Nakamura, M., Yamazaki, H.,Ueyama, Y., Tamaoki, N., Inoue, A., and Fukuda, H. (2000) J. Rheumatol.27, 2886–2892

14. Gotoh, M., Hamada, K., Yamakawa, H., Yanagisawa, K., Nakamura, M.,

FIGURE 12. Effects of wortmannin on curcumin-mediated inhibition ofNF-�B activated by IL-1�. Serum-starved human tenocytes (1 � 106 cells/ml) were pre-treated with different concentrations of wortmannin for 1 h,treated with 5 �M curcumin for 4 h, and then exposed to 10 ng/ml of IL-1� for30 min. After these treatments, nuclear extracts were prepared and analyzedfor NF-�B and poly(ADP-ribose) polymerase (PARP) (control) as describedunder “Experimental Procedures.”

Curcumin Inhibits NF-�B Signaling in Tenocytes

67

Yamazaki, H., Inoue, A., and Fukuda, H. (2002) J. Orthop. Res. 20,1365–1371

15. Barnes, P. J., and Karin, M. (1997) N. Engl. J. Med. 336, 1066–107116. Largo, R., Alvarez-Soria, M. A., Díez-Ortego, I., Calvo, E., Sanchez-Per-

naute, O., Egido, J., and Herrero-Beaumont, G. (2003)Osteoarthritis Car-tilage 11, 290–298

17. Kumar, A., Takada, Y., Boriek, A. M., and Aggarwal, B. B. (2004) J. Mol.Med. 82, 434–448

18. Ding,G. J., Fischer, P. A., Boltz, R. C., Schmidt, J. A., Colaianne, J. J., Gough,A., Rubin, R. A., and Miller, D. K. (1998) J. Biol. Chem. 273, 28897–28905

19. Sung, B., Pandey, M. K., Ahn, K. S., Yi, T., Chaturvedi, M. M., Liu, M., andAggarwal, B. B. (2008) Blood 111, 4880–4891

20. Wang, J. H., Iosifidis, M. I., and Fu, F. H. (2006) Clin. Orthop. Relat. Res.443, 320–332

21. Bharti, A. C., andAggarwal, B. B. (2002)Ann. N.Y. Acad. Sci. 973, 392–39522. Bharti, A. C., Donato, N., Singh, S., and Aggarwal, B. B. (2003) Blood 101,

1053–106223. Mukhopadhyay, A., Bueso-Ramos, C., Chatterjee, D., Pantazis, P., and

Aggarwal, B. B. (2001) Oncogene 20, 7597–760924. Aggarwal, B. B., Kumar, A., and Bharti, A. C. (2003) Anticancer Res. 23,

363–39825. Buhrmann, C., Mobasheri, A., Matis, U., and Shakibaei, M. (2010) Arthri-

tis Res. Ther. 12, R12726. Hatcher, H., Planalp, R., Cho, J., Torti, F. M., and Torti, S. V. (2008) Cell

Mol. Life Sci. 65, 1631–165227. Jurenka, J. S. (2009) Altern. Med. Rev. 14, 141–15328. Shakibaei, M., John, T., Schulze-Tanzil, G., Lehmann, I., and Mobasheri,

A. (2007) Biochem. Pharmacol. 73, 1434–144529. Csaki, C., Mobasheri, A., and Shakibaei, M. (2009) Arthritis Res. Ther. 11,

R16530. Cheng, A. L., Hsu, C. H., Lin, J. K., Hsu, M. M., Ho, Y. F., Shen, T. S., Ko,

J. Y., Lin, J. T., Lin, B. R., Ming-Shiang,W., Yu, H. S., Jee, S. H., Chen, G. S.,Chen, T.M., Chen, C. A., Lai, M. K., Pu, Y. S., Pan,M. H.,Wang, Y. J., Tsai,C. C., and Hsieh, C. Y. (2001) Anticancer Res. 21, 2895–2900

31. Sharma, R. A., McLelland, H. R., Hill, K. A., Ireson, C. R., Euden, S. A.,Manson, M.M., Pirmohamed, M., Marnett, L. J., Gescher, A. J., and Stew-ard, W. P. (2001) Clin. Cancer Res. 7, 1894–1900

32. Shehzad, A., Wahid, F., and Lee, Y. S. (2010) Arch. Pharm. 343, 489–49933. Vanhaesebroeck, B., and Alessi, D. R. (2000) Biochem. J. 346, 561–57634. Coffer, P. J., Geijsen, N., M’rabet, L., Schweizer, R. C., Maikoe, T., Raaij-

makers, J. A., Lammers, J.W., and Koenderman, L. (1998) Biochem. J. 329,121–130

35. Heldin, C. H., Ostman, A., and Ronnstrand, L. (1998) Biochim. Biophys.Acta 1378, F79–113

36. Reddy, S. A., Huang, J. H., and Liao, W. S. (1997) J. Biol. Chem. 272,29167–29173

37. Reddy, S. A., Huang, J. H., and Liao, W. S. (2000) J. Immunol. 164,1355–1363

38. Burgering, B. M., and Coffer, P. J. (1995) Nature 376, 599–60239. Schulze-Tanzil, G., Mobasheri, A., Clegg, P. D., Sendzik, J., John, T., and

Shakibaei, M. (2004) Histochem. Cell Biol. 122, 219–22840. Shakibaei, M., Schulze-Tanzil, G., de Souza, P., John, T., Rahmanzadeh,

M., Rahmanzadeh, R., and Merker, H. J. (2001) J. Biol. Chem. 276,

13289–1329441. Shakibaei, M., De Souza, P., and Merker, H. J. (1997) Cell Biol. Int. 21,

115–12542. Aggarwal, B. B. (2004) Cancer Cell 6, 203–20843. Aggarwal, B. B., and Takada, Y. (2005) Cancer Treat. Res. 126, 103–12744. Ghosh, S., and Karin, M. (2002) Cell 109, (suppl.) S81–9645. Ghosh, S., May, M. J., and Kopp, E. B. (1998) Annu. Rev. Immunol. 16,

225–26046. Sizemore, N., Lerner, N., Dombrowski, N., Sakurai, H., and Stark, G. R.

(2002) J. Biol. Chem. 277, 3863–386947. Chen, Z. J., Parent, L., and Maniatis, T. (1996) Cell 84, 853–86248. Stancovski, I., and Baltimore, D. (1997) Cell 91, 299–30249. Ozes, O. N., Mayo, L. D., Gustin, J. A., Pfeffer, S. R., Pfeffer, L. M., and

Donner, D. B. (1999) Nature 401, 82–8550. Alessi, D. R., Andjelkovic, M., Caudwell, B., Cron, P., Morrice, N., Cohen,

P., and Hemmings, B. A. (1996) EMBO J. 15, 6541–655151. Squires, M. S., Hudson, E. A., Howells, L., Sale, S., Houghton, C. E., Jones,

J. L., Fox, L. H., Dickens, M., Prigent, S. A., and Manson, M. M. (2003)Biochem. Pharmacol. 65, 361–376

52. Williams, D. H., Jeffery, L. J., and Murray, E. J. (1992) Biochim. Biophys.Acta 1180, 9–14

53. Chan,W. H.,Wu, H. Y., and Chang,W. H. (2006) Food Chem. Toxicol. 44,1362–1371

54. Jang, M. K., Sohn, D. H., and Ryu, J. H. (2001) Planta Med. 67, 550–55255. Kumar, A., Dhawan, S., Hardegen, N. J., and Aggarwal, B. B. (1998)

Biochem. Pharmacol. 55, 775–78356. Plummer, S.M.,Holloway, K.A.,Manson,M.M.,Munks, R. J., Kaptein, A.,

Farrow, S., and Howells, L. (1999) Oncogene 18, 6013–602057. Sandur, S. K., Deorukhkar, A., Pandey, M. K., Pabon, A. M., Shentu, S.,

Guha, S., Aggarwal, B. B., andKrishnan, S. (2009) Int. J. Radiat. Oncol. Biol.Phys. 75, 534–542

58. Zhang, C., Li, B., Zhang, X., Hazarika, P., Aggarwal, B. B., and Duvic, M.(2010) J. Invest. Dermatol. 130, 2110–2119

59. Kane, L. P., Shapiro, V. S., Stokoe, D., and Weiss, A. (1999) Curr. Biol. 9,601–604

60. Romashkova, J. A., and Makarov, S. S. (1999) Nature 401, 86–9061. Pahl, H. L. (1999) Oncogene 18, 6853–686662. Shishodia, S., and Aggarwal, B. B. (2002) J. Biochem. Mol. Biol. 35, 28–4063. Hussain, A. R., Al-Rasheed, M., Manogaran, P. S., Al-Hussein, K. A., Pla-

tanias, L. C., Al Kuraya, K., and Uddin, S. (2006) Apoptosis 11, 245–25464. Shankar, S., and Srivastava, R. K. (2007) Int. J. Oncol. 30, 905–91865. Schweitzer, R., Chyung, J. H., Murtaugh, L. C., Brent, A. E., Rosen, V.,

Olson, E. N., Lassar, A., and Tabin, C. J. (2001) Development 128,3855–3866

66. Goel, A., Kunnumakkara, A. B., and Aggarwal, B. B. (2008)Biochem. Phar-macol. 75, 787–809

67. Sharma, R. A., Euden, S. A., Platton, S. L., Cooke, D. N., Shafayat, A.,Hewitt, H. R., Marczylo, T. H., Morgan, B., Hemingway, D., Plummer,S. M., Pirmohamed, M., Gescher, A. J., and Steward, W. P. (2004) Clin.Cancer Res. 10, 6847–6854

68. Dhillon, N., Aggarwal, B. B., Newman, R. A., Wolff, R. A., Kunnumakkara,A. B., Abbruzzese, J. L., Ng, C. S., Badmaev, V., and Kurzrock, R. (2008)Clin. Cancer Res. 14, 4491–4499

Curcumin Inhibits NF-�B Signaling in Tenocytes

68

Crystal Structure of Clustered Regularly Interspaced ShortPalindromic Repeats (CRISPR)-associated Csn2 ProteinRevealed Ca2�-dependent Double-stranded DNA BindingActivity*□S �

Received for publication, April 30, 2011, and in revised form, June 15, 2011 Published, JBC Papers in Press, June 21, 2011, DOI 10.1074/jbc.M111.256263

Ki Hyun Nam‡, Igor Kurinov§, and Ailong Ke‡1

From the ‡Department of Molecular Biology and Genetics, Cornell University, Ithaca, New York 14850 and the §NortheasternCollaborative Access Team (NE-CAT), Department of Chemistry and Chemical Biology, Cornell University, Argonne, Illinois 60439

Clustered regularly interspaced short palindromic repeats(CRISPR) and their associated protein genes (cas genes) arewidespread in bacteria and archaea. They form a line of RNA-based immunity to eradicate invading bacteriophages andmali-cious plasmids. A key molecular event during this process is theacquisition of new spacers into the CRISPR loci to guide theselective degradation of the matching foreign genetic elements.Csn2 is a Nmeni subtype-specific cas gene required for newspacer acquisition.Herewe characterize theEnterococcus faeca-lis Csn2 protein as a double-stranded (ds-) DNA-binding pro-tein and report its 2.7 A tetrameric ring structure. The innercircle of the Csn2 tetrameric ring is �26 A wide and populatedwith conserved lysine residues poised for nonspecific interac-tions with ds-DNA. Each Csn2 protomer contains an �/�domain and an �-helical domain; significant hinge motion wasobserved between these two domains. Ca2� was located at stra-tegic positions in the oligomerization interface. We furthershowed that removal of Ca2� ions altered the oligomerizationstate of Csn2, which in turn severely decreased its affinity fords-DNA. In summary, our results provided the first insight intothe function of the Csn2 protein in CRISPR adaptation byrevealing that it is a ds-DNA-binding protein functioning at thequaternary structure level and regulated by Ca2� ions.

Clustered regularly interspaced short palindromic repeats(CRISPR)2 drives the adaptation to harmful invading nucleicacids, such as conjugative plasmids, transposable elements, and

phages, using an RNA-mediated defense mechanism with fun-damental similarities to our innate and adaptive immuneresponses (1–7). Although the details of this defense mecha-nism remain to be determined, two distinct stages have beenrecognized: (i) adaptation upon first exposure to the foreignnucleic acid, whereby some combination of CRISPR-associated(Cas) proteins extracts recognizable features from the genomesof viruses (bacteriophages) and plasmids as proto-spacers thatare subsequently incorporated as spacers at the 5� end of theCRISPR loci; and (ii) interference upon re-exposure to the samenucleic acid whereby a ribonucleoprotein complex comprisedof small guide RNAs derived from the genomic CRISPRs(crRNAs) and different Cas proteins targets foreign nucleicacids for destruction (8–16). CRISPR-Cas defense systemshavebeen identified in 83% of archaeal genomes and 45% of bacterialgenomes thus far sequenced, including important humanpathogens such as Campylobacter jejuni, Clostridium botuli-num, Escherichia coli, Listeria monocytogenes, Mycobacteriumtuberculosis, and Yersinia pestis (8, 17, 18). The significance ofthis pathway for human health is best illustrated in the humanpathogen Staphylococcus epidermidis, where horizontal genetransfer through conjugation and plasmid transformation isprevented by CRISPR-Cas (19).Despite strong interest in understanding the CRISPR adap-

tation process, its detailed molecular mechanisms remain to beelucidated. It was shown that new spacers are integrated at the5�-end (leader end) of the CRISPR cluster (10, 20, 21). Coupledwith a new integration event, loss of repeats elsewhere has beenfrequently observed, suggesting the occurrence of spontaneousrecombination (2, 3, 22). Two of the most conserved core casgenes, cas1 and cas2, have been implicated in the new spaceracquisition process (2, 5). Cas1 has been predicted to act as anintegrase in new spacer acquisition (9, 11). Recently,Pseudomo-nas aeruginosa Cas1 protein has been characterized as ametal-dependent double-stranded (ds-) DNA endonuclease(24), whereas E. coli Cas1 possesses nuclease activity againstsingle-stranded (ss-) and branched DNAs (25). Cas2 genescould be further divided into subgroups in different CRISPRsubtypes. Sulfolobus solfataricus Cas2 protein was character-ized as a metal-dependent endoribonuclease with sequencepreference for U-rich ss-RNA (26). Bacillus halodurans Cas2,however, contains metal-dependent ds-DNA endonuclease

* This work was supported, in whole or in part, by National Institutes of HealthGrant GM-086766 and GM-059604 throughout the Public Health Service(to A. K.). This work was also supported by the National Research Founda-tion of Korea Grant NRF-2010-357-C00106 from the Korean Ministry ofEducation, Science and Technology (to K. H. N.).

� This article was selected as a Paper of the Week.The atomic coordinates and structure factors (code 3S5U) have been deposited in

the Protein Data Bank, Research Collaboratory for Structural Bioinformatics,Rutgers University, New Brunswick, NJ (http://www.rcsb.org/).

□S The on-line version of this article (available at http://www.jbc.org) containssupplemental Figs. S1–S6 and Table 1.

1 To whom correspondence should be addressed. Tel.: 607-255-3945; Fax:607-255-6249; E-mail: ak425@cornell.edu.

2 The abbreviations used are: CRISPR, clustered regularly interspaced shortpalindromic repeats; Cas, CRISPR-associated; Ni-NTA, nickel-nitrilotriaceticacid; SUMO, small ubiquitin-like modifier; SAD, single wavelength anoma-lous dispersion; r.m.s., root mean square.

© 2011 by The American Society for Biochemistry and Molecular Biology, Inc. Printed in the U.S.A.

69

activity in our hands.3 Although the nuclease activities of Cas1and Cas2 could be involved in the new spacer integration in theCRISPR adaptation stage, a convincing biochemical reconstitu-tion of this process has not been demonstrated (3). A less con-served core cas gene, cas4, bears resemblance to the RecB fam-ily of exonucleases and was suggested to play a role in newspacer acquisition (8, 9).Genetic screens further identified subtype-specific cas genes

involved in new spacer acquisition. For example, in Streptococ-cus thermophilus, a Nmeni Cas subtype organism, the casoperon contains only two additional cas genes (csn1 and csn2)besides cas1 and cas2. Although csn1 is required for crRNA-mediated silencing, csn2 was shown by a genetic screen to berequired forCRISPR adaptation (10). Although structuralmod-els and biochemical characterizations are available for the Cas1and Cas2 proteins, little is known about the structure and func-tion of the Csn2 protein. Here we show that the Enterococcusfaecalis Csn2 binds to double-stranded (ds)-DNA and describeits 2.7 Å crystal structure. We conclude that the Csn2 proteinfunctions at the quaternary structure level, by adopting its finalshape through tetrameric ring formation. Tetramerizationleads to a conserved set of lysine residues being presentedtoward the inner circle of the ring for interactions with theds-DNA. The observation of tightly bound Ca2� ions in theCsn2 structure led to further investigations that demonstratedthat Ca2� regulates the Csn2 function by affecting its oligomer-ization state and enabling DNA binding. These results providethe first insight into the role of csn2 in theCRISPR adaptation inthe Nmeni subtype organisms.

EXPERIMENTAL PROCEDURES

Cloning, Expression, and Purification—Full-length csn2 gene(accession number: C7UDU4) from E. faecalis was cloned intoamodified pSUMOvector fusedwith aHis6-taggedN-terminalSUMO protein. Recombinant protein was expressed fromE. coli BL21 star (Novagen) cell after the cell density reachedA600 of 0.8 by the addition of 0.5 mM isopropyl-1-thio-�-D-galactopyranoside at 18 °C for 12 h. The harvested cells wereresuspended in lysis buffer containing 50mMTris-HCl, pH 8.0,and 0.3 M NaCl. After sonication and centrifugation, the super-natant was loaded onto a 5-ml Ni-NTA column (Qiagen) equil-ibrated with lysis buffer plus 2 mM 2-mercaptoethanol andeluted with the same buffer plus 300 mM imidazole. After adialysis to remove imidazole, the N-terminal SUMO tag wascleaved by incubatingwith the SUMOprotease and removed bypassing through a secondNi-NTAcolumn.ResultingCsn2pro-teins were concentrated and further purified on a Superdex 200column (GEHealthcare) equilibratedwith sizing columnbuffercontaining 50mMTris-HCl, pH8.0, 0.2MNaCl, and 2mMDTT.To remove bound nucleic acids, the Csn2 fractionswere pooledand further purified on a Mono Q column in a NaCl gradient(GE Healthcare). The Csn2-bound nucleic acids were thendesalted and concentrated using a Centriprep filter (molecularweight cutoff of 10,000) and cloned into the pJET blunt-endcloning vector (Fermentas).

Oligomeric State Analysis—The oligomeric state of the Csn2protein was analyzed at 4 °C using the analytical Superdex 200column equilibrated in the sizing column buffer. In the Ca2�

dependence analysis, the protein and the elution buffer (10 mM

Tris-HCl, pH 8.0, 200 mM NaCl, and 2 mM DTT) were supple-mented with 20 mM CaCl2, EDTA, or EGTA, respectively.Analysis of the Interaction between Csn2 and ds-DNA—Co-

purifying nucleic acids were separated from the Csn2 proteinon the Mono Q column, concentrated, and analyzed on a 1%(w/v) agarose gel stained with ethidium bromide. To revealtheir identity, DNase I (0.1 �g/ml) and RNase A(0.1 �g/ml)digestionswere carried out in a buffer containing 20mMHEPES(pH 7.5) for 30min at 25 °C. Alkaline hydrolysis was carried outin 100 mM sodium carbonate, pH 8.8, and 2 mM EDTA at 75 °Cfor 5 min. The reaction products were visualized using a GelDoc XR System (Bio-Rad).Electrophoretic Mobile Shift Assay (EMSA)—The ds-DNA-

Csn2 interaction was assayed using 100 ng of ds-DNA contain-ing the clonedCsn2-boundnucleic acid #3 PCR-amplified fromthe pJET cloning vector. Metal-dependent ds-DNA binding ofCsn2 (0–160 �M) was measured in 20 mM HEPES, pH 7.5, and50 mM NaCl and supplemented with 20 mM CaCl2, EDTA, orEGTA. After incubation, the reactionmixture in the Ca2� con-dition was separated on a 6% native Tris-glycine gel, and theEDTA- and EGTA-containing samples were separated on a 6%native Tris-borate-EDTA gel to avoid incompatibility with theTris-glycine buffer. Gels were visualized using ethidium bro-mide staining and analyzed using a Gel Doc XR system(Bio-Rad).Crystallization, Data Collection, and Structure Determi-

nation—The Csn2 crystals were grown at 18 °C using the hang-ing drop vapor diffusion method by mixing 5 mg/ml protein ata 1:1 ratio with the well solution containing 0.1 MMES, pH 6.0,0.1 M calcium acetate, and 6–14% (w/v) PEG 6000. Suitablecrystals for x-ray diffraction grew within 5–7 days. Both thenative and the selenium-methionine derivatized crystals werecryo-protected by soaking the crystals in the well solution sup-plemented with 30% (v/v) ethylene glycol. The native data setwas collected at 100 K at the Macromolecular Diffraction atCornell High Energy Synchrotron Source (MACCHESS) beamline A1. The selenium-methionine single wavelength anoma-lous dispersion (SAD) datasetwas collected through themail-incrystallography service at theNE-CAT beam line 24ID-C at theAdvanced Photon Source (APS). Diffraction data sets wereindexed, integrated, and scaled using HKL2000 (27). Initial setsof phases were obtained from a selenium-methionine SADdataset using the direct method in SHELXC/D/E (28). Refinementof the heavy metal sites and automated model building werecarried out using the PHENIX software suite (29). Structurebuilding and refinement were carried out using the programsCOOT (30) andREFMAC (31), respectively. The final structuremodel was refined using the PHENIX software suite (29). Sim-ulated annealing omit maps were systematically generated tocheck the quality of the model. We further checked the qualityof the model using MOLPROBITY (32). The coordinates andstructure factor have been deposited in the Protein Data Bankwith the accession code 3S5U.3 K. H. Nam, I. Kurinov, and A. Ke, unpublished data.

Ca2�-dependent ds-DNA Binding in Csn2

70

Structural Analysis—The structure-based sequence align-ment was carried out using ClustalW and ESPRIPT (33, 34).The three-dimensional structural similarity between Csn2 andother proteins was identified using the DALI server (35).Molecular contacts, buried surface areas, and temperatureB-factor distribution were analyzed using CCP4 and CNS(36).Surface conservation within the Csn2 family of proteins wascalculated and illustrated using the Consurf server (37) andChimera (23). Figure illustrations were generated usingPyMOL (38).

RESULTSE. faecalis Csn2 Binds ds-DNA—E. faecalis Csn2 protein

recombinantly expressed from E. coli was found to assembleinto a large oligomeric state and interact with nucleic acidstrongly, displaying higher absorbance at UV260 rather thanUV280 after Ni-NTA and size exclusion chromatography (Fig.1A). In the presence of the co-purifying nucleic acids, Csn2migrated as a large oligomeric species on size exclusion chro-matographywith an estimatedmolecularmass of over 660KDa.Removing the co-purifying nucleic acids reduced the average

FIGURE 1. Nucleic acid binding and oligomerization in Csn2. A, elution profile of the Csn2 protein with (peak 1) or without (peak 2) the co-purifying nucleicacids on an analytical Superdex 200 10/300 size exclusion column. Csn2 in peak 2 showed an average size of a pentamer to hexamer formation. Further changesin oligomerization state upon Ca2� binding are shown in Fig. 5D. B, separation of the co-purifying nucleic acids from the Csn2 protein on the Mono Q column.C, analysis of the fractions in panel B using the Coomassie Blue-stained SDS-PAGE gel (upper panel) or ethidium bromide-stained agarose gel (lower panel).Fraction numbers are consistent with those shown in panel B. D, identity of the Csn2-bound nucleic acids examined by incubating with DNase I, RNase A, or thealkaline hydroxyl buffer.

Ca2�-dependent ds-DNA Binding in Csn2

71

molecularmass of theCsn2 to in a broad range of a pentamer orhexamer (Fig. 1A). As shown later, the oligomerization and ds-DNA binding properties of Csn2 were further regulated by thepresence of Ca2�. The co-purifying nucleic acids could beextracted from theCsn2 protein using anion exchange chroma-tography (Fig. 1, B and C) and were shown to be �100–500 bpin size. They were sensitive to DNase I digestion, but not RNaseA digestion nor alkaline hydrolysis treatment that selectivelydegrades RNA (Fig. 1D), suggesting that the Csn2-boundnucleic acids are likely the E. coli endogenous ds-DNA, but notRNA. Consistent with this observation, we showed that theCsn2-bound nucleic acids could be cloned into a blunt-endcloning vector, and the sequencing reads from four such cloneswere shown to match to either the E. coli genomic DNA or theCsn2 expression plasmid (supplemental Table S1). The size dis-tribution of these DNA species was likely the result of physicalshearing from sonication during cell lysis. No sequence homol-ogy was found between these DNA species, indicating thatCsn2 is largely a nonspecific ds-DNA-binding protein, which isconsistent with its structure features described below. How-ever, a more detailed biochemical study will be required toinvestigate whether certain short DNA sequences are preferen-tially bound by Csn2. A further study using EMSA revealedstrong interactions between Csn2 and a ds-DNA substratePCR-amplified to contain one of the sequences of the co-puri-fying DNA (supplemental Table S1, row 3), confirming thatCsn2 interacts strongly with ds-DNA (see below). By contrast,EMSA done at similar conditions using ss-DNA, ss-RNA, andds-RNA substrates did not show appreciable interactions withthe Csn2 protein (data not shown). Analysis of these nucleicacids after Csn2 incubation on a sequencing gel did not revealnuclease or ribonuclease activity in Csn2 either (data notshown).Overall Structure of Csn2—The crystal structure of Csn2

from E. faecalis was solved by single SAD using the selenium-methionine derivatized protein (Table 1). The asymmetric unitof the orthorhombic P212121 space group contains two non-crystallographic symmetry-related tetrameric rings (supple-mental Fig. S1). The twoCsn2 tetramers adopt slightly differentconformations with an r.m.s. deviation of 1.3 Å for all C� atoms(supplemental Fig. S2). Each diamond-shaped tetrameric ringmeasures 70 � 70 Å in width and 50 Å in height (Fig. 2A). EachCsn2 protomer contains a globular �/� domain and anextended�-helical domain extruded from themiddle of the�/�domain. Extensive interactions between the two �/� domainslead to Csn2 dimer formation. Further hydrophobic interac-tions between two such Csn2 dimers at the extended �-helicaldomain lead to the tetrameric ring formation (Fig. 2A). Fourtightly bound Ca2� ions were found at this interface. The innerdiameter of the ring measuring 26 Å at the narrowest regionagrees well with accommodating the binding of a ds-DNA sub-strate through the center. Electrostatic analysis revealed thepresence of a set of positively charged lysine residues populat-ing the inner surface of the ring (Fig. 2B). Both features areconsistent with this region being involved in the binding ofds-DNA in a sequence-nonspecific fashion. As the Cns2 familyof proteins are highly conserved (Fig. 3A, supplemental Fig. 3),

the observed structural features described here are likely sharedby all Csn2 proteins.ATPase-like Scaffold in the Csn2 �/� Domain and a Confor-

mational Hinge—As mentioned above, each Csn2 protomerconsists of two domains: an�/� domain connected to an�-hel-ical domain insertion through a flexible hinge region (Fig. 3B).The �/� domain is composed of a central �-sheet structuresandwiched by the �1-helix on one side and �5 and �6 on theother. Extensive van der Waals interactions between the�-sheet and the three helices give rise to the hydrophobic coreof this domain. The central �-sheet consists of five parallel�-strands (�1–�5) and the C-terminal �3� packed in an anti-parallel fashion. Above the �1-helix side of the central �-sheet,residual electron densities for a parallelly packed, small �-sheetcomprised of �1�, �2�, and �3� were observed in some, but notall, of the eightmolecules in the asymmetric unit, suggesting thepresence of local flexibility in this region. The �-helical domaincomprising three �-helices (�2, �3, and �4) extrudes from themiddle of the�/� globular domain. Lacking a strong hydropho-bic core, its conformation is critically influenced by the oligo-merization interactions and the binding of Ca2� ions (seebelow).A conformational hinge was found between the �/� and the

�-helical domains. Each �/� domain (Pro-11-Va-l62 and Leu-144-Ala-219, excluding a flexible loop betweenTyr-38 andGlu-58) and an�-helical domain (Ala-73–Ala-132), is almost super-imposablewith their counterparts in the two tetrameric rings inthe asymmetric unit, with an r.m.s. deviation less than 1.1 Å

TABLE 1Data collection and refinement statisticsHighest resolution shell is shown in parentheses.

Native SADData collection statisticsBeamline MACCHESS APS

A1 24IDC,Wavelength 0.9770 0.97917Space group P212121 P212121Unit cell parameters (Å)a 104.11 104.05b 140.09 138.73c 148.43 148.29

Resolution (Å) 20.0-2.70 (2.75-2.70) 30.0-3.10 (3.21-3.10)Completeness (%) 98.8 (95.5) 99.8 (99.6)Redundancy 5.5 (3.6) 2.6 (2.5)I/�(I) 25.26 (2.76) 14.63 (1.57)Rmerge (%)a 8.6 (29.5) 3.9 (34.8)

Refinement statisticsResolution (Å) 20.0-2.70Rwork/Rfree(%)b 21.90/27.5B-factor (Averaged)Protein 51.75Ca2� 49.43Water 44.35

r.m.s. deviationsBond lengths (Å) 0.016Bond angles(°) 1.697

Ramachandran plot (%)cfavored 96.3allowed 3.7

a Rmerge � �h �i� I(h,ii) � � I (h) � /�h �iI (h,ii),where I (h,i) is the intensity of theith measurement of reflection h and � I (h) is the mean value of I(h,i) for all imeasurements.

b R � ��Fobs� � �Fcalc�/��Fobs�, where Fobs and Fcalc are the observed and calculatedstructure-factor amplitudes, respectively. Rfree was calculated as Rwork using arandomly selected subset (10%) of unique reflections not used for structurerefinement.

c Categories were defined by MOLPROBITY.

Ca2�-dependent ds-DNA Binding in Csn2

72

(supplemental Fig. S2). The angle between the two domains,however, varied by as much as 5° among eight Csn2 protomersin the asymmetric unit, suggesting the presence of a hingeregion between these two domains (Fig. 3C). Nevertheless, thetwo connecting loops in this hinge region between �2 and �2(Thr-63–Ser-72) and also between �4 and �5 (Leu-133–Thr-143) display similar temperature B-factors as the rest of thestructure due to the binding of Ca2� ions to stabilize the hingeregion (see below).The �/� globular domain in Csn2 bears structural similarity

to a family of ATP-binding proteins, although the sequencehomology is hardly detectable (11% sequence identity).Structural homologs include theClostridiumperfringens cobaltimport ATP-binding protein, the enterobacterial phage T7DNA primase/helicase, the Pyrococcus furiosus DNA double-strand break repair RAD20 ATPase, and the M. tuberculosisRecA (PDB accession codes: 3GFO, 1CR1, 3QKU, and 1MO4,respectively; supplemental Fig. S4). Closer structural analysis,however, suggested that ATP binding would unlikely be thenative function of the Csn2 protein as neither the backbonegeometry nor the key contacting residues at the ATP-bindingsite are conserved in Csn2. Indeed, we were not able to detectinteractions between ATP and Csn2 using isothermal titration

calorimetry analysis (data not shown). The distinct topology inthe�/� domain and the presence of the�-helical domain inser-tion set the Csn2 protein apart from other ATP-binding �/�domain proteins (supplemental Fig. S4).Two Dimerization Interfaces Lead to Csn2 Tetrameric Ring

Formation—The tetrameric ring formation is best described astwo sequential dimerization events, first leading to the dimerformation betweenmolecules A-C and B-D (interface A-C; Fig.4A) and then the dimerization in molecules A-B and C-D(interface A-B; Fig. 4B). Superimposition of each dimer (mole-cules A-B or A-C) yielded similar r.m.s. deviation of 0.5–1.1 Åin C� alignment (supplemental Fig. S2). Both interfaces involvehighly conserved residues among the Csn2 family of proteins(Fig. 3D; supplemental Fig. S3). The A-C interface involvessymmetric interactions to the side of the �1-helix and �-sheetand the top of the �5 and �6 helices between two �/� domains,burying a surface area of�2100Å2 (�16%of the total surface ineach protomer; Fig. 4A and supplemental Fig. S5). Among theinterface residues of molecule A-C, 33% are hydrophobic, 66%are hydrophilic (i.e. H-bonds between Tyr-36–Asp-64, Asp-64–Thr-171, and Asn-168–Tyr-172), and 16% are charged.Interface A-B involves reciprocal interactions at the �2–�4

helices and hinge loop region of molecules A and B, burying a

FIGURE 2. Overall structure of the Csn2 tetrameric ring structure. A, the top-down and the side views (upper and lower panels, respectively) of the Csn2tetrameric ring structure. The four protomers A–D are colored yellow, cyan, teal, and magenta, respectively. The two 2-fold non-crystallographic symmetry axesand the measurement of the ring dimensions are marked. B, the same views to reveal the electrostatic potential on the Csn2 surface. A clear partition of positiveand negative charges to the inner and outer portions of the tetrameric ring, respectively, is observed.

Ca2�-dependent ds-DNA Binding in Csn2

73

FIGURE 3. Structural analysis of the Csn2 protomer. A, sequence alignment among the Csn2 family of proteins from E. faecalis (accession code: C7UDU4),Enterococcus faecium (D4W167), L. monocytogenes (E3YJP9), Streptococcus anginosus (E6J3Q7), S. bovis (E0PEL0), Streptococcus pyogenes serotype M1 (Q99ZV9),and S. thermophilus (Q03JI9). The absolutely conserved residues are boxed in red, and the highly conserved ones are in unfilled boxes and red letters. Residuesinvolved in the binding of Ca2� and ds-DNA (putative) are marked with red and blue asterisks, respectively. B, two views of the monomeric structure of theE. faecalis Csn2 protein. Csn2 consists of an �/� domain and an �-helical domain. Hinge loops between the two domains are colored magenta. A disordereddistal loop in the �/� domain is represented by the magenta dotted line. The two Csn2-bound Ca2� ions are represented in yellow spheres. The two �-sheets anda 310-helix inside the �/� domain are colored in yellow and green, respectively. C, superimposition of the eight Csn2 monomers in the asymmetric unit alongthe �1-helix (in blue). Hinge motion at the hinge region in magenta leads to �5° variation in the orientation of the �-helical domain. Large deviations inequivalent atom positions are marked. D, surface conservation in the Csn2 protomer. Residues are colored from magenta to cyan with descending order ofconservation.

Ca2�-dependent ds-DNA Binding in Csn2

74

surface area of 4200 Å2 (�32% of each subunit; Fig. 4B andsupplemental Fig. S5). The interface residues are 51% hydro-phobic, 49% hydrophilic, and 31% charged. The hydrophobicinteractions include the two anti-parallel leucine/isoleucinezipper �3 helices and the contacts from �3 to �2 and �4. Addi-tional contacts include salt bridges between Lys-90 and Glu-114 and between Glu-116 and Arg-156.Potential ds-DNA-binding Site Inside the Tetrameric Ring—

Electrostatic surface potential analysis revealed a clear segrega-tion of positive and negative charges on the inner and outersurfaces of the Csn2 tetrameric ring, respectively (Fig. 2B). Thecharge distribution is conserved among the Csn2 family of pro-teins (Fig. 3A) and is in line with its putative ds-DNA bindingfunction. The lysine-rich basic patch is particularly interestingbecause itmaymediate nonspecific interactionswith the sugar-phosphate backbone of the ds-DNA (Figs. 2B and 4C). Thispatch is composed of at least seven highly conserved lysine res-idues (Lys-52, Lys-55, Lys-77, Lys-131, Lys-160, Lys-161, andLys-162; Fig. 4D). The presence of a positive change is consist-ent in the place of Lys-55, Lys-77, Lys-160, Lys-161, and Lys-

162, whereas glutamine is occasionally found in the place ofLys-52 andLys-131. In the crystal structure, Lys-77 andLys-131are located in the �-helical domain, and the rest of the Lysresidues are located in the �/� domain. Due to the presence ofa hinge region between these two domains, the exact positionsof the lysine residues differ among the Csn2 protomers. Theversatility in lysine positions and the flexibility in their sidechain conformations are consistent with this lysine-rich patchinside the tetrameric ring contacting the ds-DNA.Ca2� Ions Stabilize theOligomerization Interface—Introduc-

tion of Ca2�, but not other divalent cations, was a prerequisitecondition in the crystallization of the E. faecalis Csn2 protein.Upon structure determination, we located four potential Ca2�

ions in two Ca1 sites and two Ca2 sites in each Csn2 tetramericring (Fig. 5A and supplemental Fig. S6). These sites appeared asstrong peaks in the Fo � Fc simulated annealing omit differencemap counted at the 5 � level and formed octahedral or squarepyramidal coordination with surrounding ligands (Fig. 5, B andC). The temperature B-factor and occupancy refinements sug-gested that these sites were stoichiometrically occupied by

FIGURE 4. Interfaces A-C and A-B that mediate the tetrameric ring formation in the Csn2 protein. A, interface A-C that leads to the dimerization of two �/�domains (colored in yellow and magenta). Side chains of the contacting residues are displayed. B, interface A-B that leads to the dimerization of two �-helicaldomains. An extensive leucine/isoleucine zipper and four Ca2�-binding sites (two visible) are displayed. C, electrostatic potential of a Csn2 protomer. Theinterface A-B may be significantly weakened without the Ca2� ions to shield the strong negative charges at this interface. D, the conserved positive chargesfrom the lysine residues (also in panel C) span a distance of 35 Å along the Csn2 protomer surface. Upon tetramerization, these residues give rise to a positivelycharged inner surface potentially important for ds-DNA binding.

Ca2�-dependent ds-DNA Binding in Csn2

75

Ca2� ions. EGTA chelation of Ca2� fromCsn2 crystals prior todata collection resulted in severe degradation of x-ray diffrac-tion resolution (data not shown). Incubation of equimolaramounts of Mn2� and Ca2� with Csn2 crystals, followed bydata collection at the anomalous edge of Mn2�, did not showstrong anomalous difference peaks indicative of competitivebinding of Mn2� to the Ca2� sites (data not shown). These twocircumstantial evidences further support the presence of Ca2�

in the Csn2 structure.Both Ca2�-binding sites are located at the interface A-B and

mediate interactions between Csn2 protomers by shielding thecharge repulsion among coordinating functional groups. ThetwoCa1 sites are located in themiddle of the interface A-B nearthe �4-helix. Ca2� is octahedrally coordinated by O� of Asp-122 (average distance of 2.42 Å), the main chain carbonyl ofGly-123 (2.26 Å), O� of Glu-128 (2.56 Å) in molecule A, themain chain carbonyl of Ala-132 (2.42Å) inmolecule B (Fig. 5B),

as well as two water (2.42 Å) molecules. The two Ca2 siteslocated at the hinge region near the N-terminal end of the�5-helix are square pyramidally coordinated by the O� of Asp-118 (average distance of 2.42 Å) in molecule A and by the O�and the main chain carbonyl of Glu-138 (2.40 and 2.52 Å,respectively), O� of Asp-142 (2.39 Å) and the O� of Glu-150(2.37 Å) in molecule B (Fig. 5C). The key Ca2� coordinatingresidues including Asp-118, Asp-122, Glu-128, Glu-138, andGlu-150 are highly conserved (Fig. 3A and supplemental Fig.S3), suggesting that Ca2� binding is a conserved feature amongCsn2 proteins. The other two coordinating residues Gly-123and Ala-132 can be substituted by other residues among theCsn2 family because only the main chain carbonyl groups areinvolved in coordination.Ca2� Influences the Oligomerization and ds-DNA Binding

Properties of Csn2—Because our crystal structure revealed thatthe Ca1 and Ca2 sites were strategically positioned at the oligo-

FIGURE 5. Ca binding critically influences the oligomerization and DNA binding properties of the Csn2 protein. A, two views of the Ca1- and Ca2-bindingsites around the interface A-B. In both cases, Ca2� binding requires the participation of residues from both Csn2 protomers (in yellow and cyan). B, Ca1 sitesuperimposed with the composite omit electron density contoured at the 5 � level. Contacting residues from both Csn2 protomers, as well as two orderedwater molecules, are highlighted. C, Ca2 site displayed in a similar fashion. Although most Ca2� chelating residues are from molecule A, Asp-118 from moleculeB makes a critical contact to seal this Ca2�-binding site. This contact also rigidifies the conformation of the otherwise flexible loop connecting the �3 and �4helices in molecule B. D, size exclusion chromatography showing that the presence of Ca2� ions leads to a more defined Csn2 tetramer formation, whereas thecomplete removal of Ca2� ions using EDTA or EGTA leads to a bigger Csn2 oligomer formation. E, EMSA experiments where a titration of Csn2 protein (5–160�M) was incubated with 100 ng of ds-DNA in the presence of 20 mM Ca2�, Mg2�, EDTA, or EGTA. Results showed that the Csn2 protein interacted strongly withthe ds-DNA in the presence of Ca2� ions. This ds-DNA binding activity was decreased to the background level in the presence of Mg2� or Ca2�-chelating EDTAor EGTA buffers.

Ca2�-dependent ds-DNA Binding in Csn2

76

merization interface A-B, we speculated that Ca2� may criti-cally influence the oligomerization state and ds-DNA bindingproperty of theE. faecalisCsn2 protein. Size exclusion chroma-tography was carried out to study the effect of Ca2� binding onthe oligomerization state of Csn2. As shown previously, puri-fied E. faecalisCsn2 proteinmigrated as amixture of oligomerswith an average size of a pentamer or hexamer (Figs. 1A and5D). The addition of EDTA or EGTA to further remove theco-purifying Ca2� from the Csn2 protein encouraged the for-mation of highermolecularweight oligomers, whereas incubat-ing with 20 mM Ca2� promoted the formation of a Csn2tetramer. Therefore, structural and biochemical evidence con-verged in, suggesting that Ca2� plays an important role in pro-moting the formation of a Csn2 tetrameric ring.Next, we studied whether Ca2� binding may influence the

ds-DNA binding property of the Csn2 protein. EMSAs werecarried out in the presence of Ca2�, EDTA, EGTA, or Mg2�

using a ds-DNA substrate PCR-amplified from the clonedCsn2-co-purifying DNA (Fig. 1). The Csn2 protein was foundto interact strongly with this DNA only in the presence of Ca2�

ion, but not in the presence of EGTA or EDTA (Fig. 5E). Inter-estingly, Mg2� was not able to restore the ds-DNA binding byCsn2 either, suggesting again that the Csn2 interacts specifi-cally with the Ca2� ion. Together with the Ca2�-dependentoligomerization study above, these results suggest that Ca2�

binding promotes tetramerization and that the tetrameric ringgives rise to the ds-DNA binding property in the Csn2 protein.

DISCUSSION

Nmeni CRISPR-Cas subtype provides us a unique opportu-nity to study CRISPR adaptation because a genetic screen inS. thermophilus has identified the key players required for newspacer acquisition (10). The identified proteins included thecore cas genes cas1 and cas2 and a Nmeni subtype-specificgene csn2. Although structural models and biochemical dataare available for Cas1 andCas2 proteins, this study provides thefirst set of such data for the Csn2 protein. Our crystal structureclearly reveals that the conserved family of Csn2 proteins func-tions at the quaternary structure level; that is, Csn2 assumes itsultimate shape and charge distribution only after tetramericring formation. The inner diameter of the ring and the align-ment of the positively charged lysine residues inside the ringcoincide with the characterized nonspecific ds-DNA bindingfunction of Csn2 quite well. The crystal structure revealed aninteresting twist of Ca2� regulation that otherwise would likelybe missed in solution studies. We observed the presence of twoCa2� ions; one of them is bound to a critical hinge region ineach Csn2 protomer, and the other shields many positivecharges in the oligomerization interface. Follow-up solutionstudies revealed that Ca2� binding leads to more defined Csn2tetramerization, likely through rigidifying the internal hingeregion. More importantly, alteration to the tetrameric ringstructure through Ca2� chelation also abolishes the ds-DNAbinding function of Csn2. Collectively, the evidence is quitestrong to support that the Csn2 tetramer, influenced by Ca2�

binding, is the functional unit in binding ds-DNA.The Csn2 in the Nmeni subtype is regarded as an essential

protein in the CRISPR-mediated silencing pathway (10). Nev-

ertheless, it is not universally present in all Nmeni subtypeorganisms (8). The rather diverseNmeni subtype can be furtherdivided into two gene clusters; one contains the csn1-cas1-cas2-csn2 cassette (in E. faecalis), and the other contains the csn1-like cas1-cas2-cas4 cassette (in Wolinella succinogenes) (2, 9).This observation seems to suggest that the csn2 and cas4 genesare functional homologs of each other. However, at thesequence level, the Cas4 protein, which was predicted to be aRecB-like nuclease and contains a Zn2�-binding cluster, isquite distinct from the Csn2 protein (9). This might be anotherexample of the presence of mechanistic diversity in theCRISPR-Cas systems.The exact function of Csn2 requires further study. It remains

unclear whether the Csn2 protein is involved in the productionof proto-spacers or the downstream step of new spacer integra-tion. Although our preliminary cloning experiment suggestedthat Csn2 binds to diverse DNA sequences, it cannot be ruledout that certain short sequences are preferentially selected bythe Csn2 protein. This can be further studied by deletion map-ping of the cloned DNA substrates. The results could be quiteinteresting because it has been suggested that the short proto-spacer adjacent motif or the leader sequences at the target siteplay an important role in new spacer acquisition (2, 5).

Acknowledgments—We thank the beam line staff at Advanced Pho-ton Source ID24 and Macromolecular Diffraction at Cornell HighEnergy Synchrotron Source for assistance in data collection andMat-thew Bratkowski for helpful discussions and comments on the manu-script. This work is based upon research conducted at the AdvancedPhoton Source on the Northeastern Collaborative Access Team beamlines, which are supported by Award RR-15301 from the NationalCenter for Research Resources at the National Institutes of Health.Use of the Advanced Photon Source, an Office of Science User Facilityoperated for the United States Department of Energy (U.S. DOE)Office of Science by Argonne National Laboratory, was supported bythe U.S. DOE under Contract DE-AC02-06CH11357.

REFERENCES1. Sorek, R., Kunin, V., and Hugenholtz, P. (2008) Nat. Rev. Microbiol 6,

181–1862. van der Oost, J., Jore, M.M.,Westra, E. R., Lundgren, M., and Brouns, S. J.

(2009) Trends Biochem. Sci. 34, 401–4073. Marraffini, L. A., and Sontheimer, E. J. (2010)Nat. Rev. Genet. 11, 181–1904. Karginov, F. V., and Hannon, G. J. (2010)Mol. Cell 37, 7–195. Deveau, H., Garneau, J. E., and Moineau, S. (2010) Annu. Rev. Microbiol.

64, 475–4936. Horvath, P., and Barrangou, R. (2010) Science 327, 167–1707. Waters, L. S., and Storz, G. (2009) Cell 136, 615–6288. Haft, D. H., Selengut, J., Mongodin, E. F., and Nelson, K. E. (2005) PLoS

Comput. Biol. 1, e609. Makarova, K. S., Grishin, N. V., Shabalina, S. A., Wolf, Y. I., and Koonin,

E. V. (2006) Biol. Direct 1, 710. Barrangou, R., Fremaux, C., Deveau, H., Richards, M., Boyaval, P.,

Moineau, S., Romero, D. A., and Horvath, P. (2007) Science 315,1709–1712

11. Brouns, S. J., Jore, M. M., Lundgren, M., Westra, E. R., Slijkhuis, R. J.,Snijders, A. P., Dickman,M. J., Makarova, K. S., Koonin, E. V., and van derOost, J. (2008) Science 321, 960–964

12. Carte, J.,Wang, R., Li, H., Terns, R.M., and Terns,M. P. (2008)Genes Dev.22, 3489–3496

13. Hale, C. R., Zhao, P., Olson, S., Duff, M. O., Graveley, B. R., Wells, L.,

Ca2�-dependent ds-DNA Binding in Csn2

77

Terns, R. M., and Terns, M. P. (2009) Cell 139, 945–95614. Haurwitz, R. E., Jinek, M., Wiedenheft, B., Zhou, K., and Doudna, J. A.

(2010) Science 329, 1355–135815. Deltcheva, E., Chylinski, K., Sharma,C.M.,Gonzales, K., Chao, Y., Pirzada,

Z. A., Eckert, M. R., Vogel, J., and Charpentier, E. (2011) Nature 471,602–607

16. Perez-Rodriguez, R., Haitjema, C., Huang, Q., Nam, K. H., Bernardis, S.,Ke, A., and DeLisa, M. P. (2011)Mol. Microbiol. 79, 584–599

17. Grissa, I., Vergnaud, G., and Pourcel, C. (2007) Nucleic Acids Res. 35,W52–57

18. Rousseau, C., Gonnet, M., Le Romancer, M., and Nicolas, J. (2009) Bioin-formatics 25, 3317–3318

19. Marraffini, L. A., and Sontheimer, E. J. (2008) Science 322, 1843–184520. Lillestøl, R. K., Redder, P., Garrett, R. A., and Brugger, K. (2006)Archaea 2,

59–7221. van der Ploeg, J. R. (2009)Microbiology 155, 1966–197622. Tyson, G. W., and Banfield, J. F. (2008) Environ. Microbiol. 10, 200–20723. Pettersen, E. F., Goddard, T. D., Huang, C. C., Couch, G. S., Greenblatt,

D. M., Meng, E. C., and Ferrin, T. E. (2004) J. Comput. Chem. 25,1605–1612

24. Wiedenheft, B., Zhou, K., Jinek,M., Coyle, S.M.,Ma,W., andDoudna, J. A.(2009) Structure 17, 904–912

25. Babu, M., Beloglazova, N., Flick, R., Graham, C., Skarina, T., Nocek, B.,Gagarinova, A., Pogoutse, O., Brown, G., Binkowski, A., Phanse, S.,Joachimiak, A., Koonin, E. V., Savchenko, A., Emili, A., Greenblatt, J.,Edwards, A. M., and Yakunin, A. F. (2011)Mol. Microbiol. 79, 484–502

26. Beloglazova,N., Brown,G., Zimmerman,M.D., Proudfoot,M.,Makarova,K. S., Kudritska, M., Kochinyan, S., Wang, S., Chruszcz, M., Minor, W.,Koonin, E. V., Edwards, A. M., Savchenko, A., and Yakunin, A. F. (2008)J. Biol. Chem. 283, 20361–20371

27. Otwinowski, Z., and Minor, W. (1997)Methods Enzymol. 276, 307–32628. Sheldrick, G. M. (2010)Acta Crystallogr. D Biol. Crystallogr. 66, 479–48529. Adams, P.D., Afonine, P. V., Bunkoczi, G., Chen,V. B., Davis, I.W., Echols,

N., Headd, J. J., Hung, L. W., Kapral, G. J., Grosse-Kunstleve, R. W., Mc-Coy, A. J., Moriarty, N. W., Oeffner, R., Read, R. J., Richardson, D. C.,Richardson, J. S., Terwilliger, T. C., and Zwart, P. H. (2010) Acta Crystal-logr. D Biol. Crystallogr. 66, 213–221

30. Emsley, P., and Cowtan, K. (2004)Acta Crystallogr. D Biol. Crystallogr. 60,2126–2132

31. Winn, M. D., Murshudov, G. N., and Papiz, M. Z. (2003) Methods Enzy-mol. 374, 300–321

32. Chen, V. B., Arendall, W. B., 3rd, Headd, J. J., Keedy, D. A., Immormino,R.M., Kapral, G. J., Murray, L.W., Richardson, J. S., and Richardson, D. C.(2010) Acta Crystallogr. D Biol. Crystallogr. 66, 12–21

33. Larkin, M. A., Blackshields, G., Brown, N. P., Chenna, R., McGettigan,P. A., McWilliam, H., Valentin, F., Wallace, I. M., Wilm, A., Lopez, R.,Thompson, J. D., Gibson, T. J., and Higgins, D. G. (2007) Bioinformatics23, 2947–2948

34. Gouet, P., Courcelle, E., Stuart, D. I., and Metoz, F. (1999) Bioinformatics15, 305–308

35. Holm, L., and Rosenstrom, P. (2010) Nucleic Acids Res. 38,W545–W54936. Brunger, A. T., Adams, P. D., Clore, G. M., DeLano, W. L., Gros, P.,

Grosse-Kunstleve, R.W., Jiang, J. S., Kuszewski, J., Nilges,M., Pannu,N. S.,Read, R. J., Rice, L. M., Simonson, T., and Warren, G. L. (1998) ActaCrystallogr. D Biol. Crystallogr. 54, 905–921

37. Ashkenazy, H., Erez, E., Martz, E., Pupko, T., and Ben-Tal, N. (2010) Nu-cleic Acids Res. 38,W529–533

38. DeLano, W. L. (2010) The PyMOL Molecular Graphics System, version1.3r1, Schrodinger, LLC, New York

Ca2�-dependent ds-DNA Binding in Csn2

78