Protein Engineering vol.9 no.2 pp. 173-180, 1996

FK506-binding protein mutational analysis: defining the active-siteresidue contributions to catalysis and the stability of ligandcomplexes

Maureen T.DeCenzo, Steven T.Park1, Beth PJarrett2,Robert A.Aldape, Olga Futer, Mark A.Murcko andDavid J.Livingston3

Vertex Pharmaceuticals Incorporated, 40 Allston Street, Cambridge,MA 02139-4211, USA

'Present address: New York University School of Medicine, 550 FirstAvenue, New York 10016, NY, USA2Present address: University of Michigan School of Medicine, 1301Catherine St, Ann Arbor 48109, MI, USA

^To whom correspondence should be addressed

The 12 kDa FK506-binding protein FKBP12 is a cis-transpeptidyl-prolyl isomerase that binds the macrolides FK506and rapamycin. We have examined the role of the bindingpocket residues of FKBP12 in protein-ligand interactionsby making conservative substitutions of 12 of these residuesby site-directed mutagenesis. For each mutant FKBP12,we measured the affinity for FK506 and rapamycin and thecatalytic efficiency in the cis-trans peptidyl-prolyl isomerasereaction. The mutation of Trp59 or Phe99 generates anFKBP12 with a significantly lower affinity for FK506than wild-type protein. Tyr26 and Tyr82 mutants areenzymatically active, demonstrating that hydrogen bondingby these residues is not required for catalysis of the cis-trans peptidyl-prolyl isomerase reaction, although thesemutations alter the substrate specificity of the enzyme. Weconclude that hydrophobic interactions in the active sitedominate in the stabilization of FKBP12 binding to macrol-ide ligands and to the twisted-amide peptidyl-prolyl sub-strate intermediate.Keywords: FK506/FKBP/peptidyl-prolyl/PPIase/rapamycin

FKBP12 to membrane receptors, but impart affinities toFKBP12 for binding to intracellular enzymes. In T-lympho-cytes, the binding of FK506 to FKBP12 interrupts cytosolicsignaling events, resulting in the inhibition of early geneexpression induced by ionophores, phorbol esters or anti-T-cell receptor antibodies (Sigal and Dumont, 1992). This signaltransduction pathway is inhibited when the FKBP12-FK506complex binds to calcineurin (K\ 6 nM; Aldape et al., 1992), aCa2+-activated, calmodulin-dependent serine/threonine proteinphosphatase (Liu et al, 1991). We have described previouslythe effects of surface residue mutations on the affinity ofFKBP12-FK506 for calcineurin (Aldape et al, 1992; Futeret al, 1995). FKBP12 also binds rapamycin with high affinity{Kx 0.2 nM). This macrolide exerts an immunosuppressiveeffect by blocking T-cell responses to interleukin-2 receptorsignaling, a late event in T-cell activation (Sigal and Dumont,1992). The target for the FKBP12-rapamycin complex hasbeen identified as TOR2 in yeast (Kunz et al, 1993) andFRAP (or RAFT1) in mammalian cells (Brown et al, 1994;Sabatini et al, 1994), proteins with sequence homology to rP3

kinase. Formation of the ternary FKBP12-rapamycin-TOR2complex arrests the cell cycle.

FKBP 12 also catalyzes the cis-trans isomerization of theprolyl amide bonds of polypeptide substrates (the peptidyl-prolyl isomerase or PPIase reaction; Harding et al, 1989;Siekerka et al, 1989). The substrate specificity for FKBP12catalysis of the PPIase reaction has been reported (Harrisonand Stein, 1990a), and the mechanism of this reaction hasbeen studied by steady-state kinetic methods (Harrison andStein, 1990b; Kofron et al, 1991; Park et al, 1992). ThePPIase activity of FKBP12 may be relevant to its association

IntroductionFKBP12, a 12 kDa multifunctional FK506-binding protein, isan intracellular receptor for the immunosuppressant macrolidesFK506 (tacrolimus; Figure 1) and rapamycin (sirolimus)(Armistead and Harding, 1993; Galat, 1993; Fruman et al,1994). FKBP12 was the first member of the multi-gene FKBPfamily to be identified (DiLella and Craig, 1991; Peartie et al,1992, 1994; Hendrickson et al, 1993). The structure andfunction of FKBP 12 has been investigated intensely since itwas identified in thymus and T-cell extracts (Harding et al,1989; Siekerka et al, 1989). In the absence of macrolideligands, FKBP12 can associate with the ryanodine Ca2+

channel (Jayaraman et al, 1992) and modulate its stabilityand conductance (Mayerleitner et al, 1994). FKBP12 alsoappears to modulate the activity of the phosphatidyl inositolreceptor (TP3R) Ca2+ channel (Cameron et al, 1995). Uponbinding of FK506 or rapamycin, associations of FKBP 12with these transmembrane proteins are disrupted, therebydestabilizing the receptors and altering their activity.

FK506 and rapamycin not only disrupt the binding of

Ha...

CH

CH3O OCH3

Fig. 1. FK5O6 (tacrolimus).

© Oxford University Press 173

M.XDeCenzo et al

Fig. 2. (A) Stereoview of the binding of FK506 to FKBP12. The side chains of residues which have been mutated in this investigation are shown in stickformat. Phe residues are displayed in purple, Tyr residues in yellow; Trp59 is grey, Asp37 is red, His87 is dark blue and Ile90 is orange. FK506 is shown ingreen. The exposed face of FK506, which does not interact with FKBP12, has been removed for clarity. The Co trace of FKBP12 is shown as a blue ribbon.Key hydrogen bonding interactions between FK506 and FKBP12 are shown as white dotted lines. Coordinates of the complex used are from Wilson et al.(1995). (B) Stereoview of the complex of FKBP12 with an Ac-Leu-Pro-Phe-NMe twisted-amide substrate intermediate. The Leu-Pro amide bond is in thehigh-energy twisted conformation. The conformation is representative of those found in a 500 ps MD simulation (Orozco et al., 1993). The side chains ofresidues which have been studied by mutagenesis are shown in stick format, with the same color scheme as in (A). The substrate is shown in green. A criticalwater molecule which forms a bndging hydrogen bond between the twisted amide carbonyl and Tyr82 is shown in a light blue ball-and-stick format. Thehydrogen bonds made by this water molecule are shown as white dotted lines. The Pj Leu side chain is pointing towards the lipophilic pocket composed ofresidues Phe36, His87, Ile90, Ile91, Ile97 and Phe99. Coordinates of the unliganded FKBP12 from Wilson el al. (1995) were used to construct this model.

with the ryanodine receptor and the binding and modulationof IP3R (Cameron et al., 1995).

Three-dimensional structures have been reported forFKBP12 (Moore et al., 1991; Wilson et al., 1995), for thebinary complexes of FKBP12 with FK506 and rapamycin (VanDuyne et al., 1991, 1993; Wilson et al, 1995), and forthe ternary complex of FKBP12-FK506-calcineurin (Griffithet al, 1995). Both hydrophobic interactions and hydrogenbonds from macrolide ligands to protein side-chain and back-bone heteroatoms stabilize these complexes (Figure 2A and

Table I). Nevertheless, structural information alone is insuffi-cient for the determination of the relative or absolute energeticcontributions of each of the hydrophobic and hydrogen bondinginteractions to the stability of these complexes. The bindingand catalytic constants of a set of FKBP12 mutants, however,in concert with the structural data, can aid in the understandingof selected FKBP12 residue interactions with its protein andmacrolide ligands. Here we report a mutational analysis of theactive site of FKBP12 with respect to the ligand-bindingaffinity and catalytic efficiency of the PPIase reaction. We

174

FKBP12 mutagenesis

Table I. CrystaJlographically defined interactions of active-site FKBP12 residues

Residue

Tyr26Phe36Asp37Arg42Phe46Phe48Gln53Glu54Ile56Trp59Tyr82His87

Phe99

FK506 atominteraction

C9 keto oxygen 'hydrogen bond'C9 keto oxygen hydrogen bond'CIO hemiketal hydroxyl hydrogen bondCI5 methoxypipicolinyl ringpipicolinyl ringcarbonyl to C24 hydroxyl via H2Ocarbonyl to C24 hydroxylamine to Cl lactone carbonylserves as a platform for pipicolinyl ringC8 amide oxygen hydrogen bondforms a surface complementary toC10-C14 pyranose methyl regioninteracts with the pipicolinyl ringC9 keto oxygen 'hydrogen bond'

lnterproteinhydrogen bonds

Asp37

Tyr26, Arg42Asp37

Ala95 via H2O

lnterprotein Van der Waalscontacts

Trp59, Phe99Phe99

Tyr26, Phe48, Trp59Phe46, Trp59

Trp59, Tyr82, Phe99

Tyr26, Phe46, Phe48, Ile56, Phe99

Tyr82, Ile90

Tyr26, Phe36, Ile56, Trp59

FKBP12 residues that interact with FK506 and their lnterprotein contacts (Van Duyne et al., 1991, 1993; Wilson et al., 1995) Hydrophobic pocket-formingresidues are Tyr26, Phe36, Phe46, Val55, Ile56, Trp59, His87, Ile90, Ile91, Leu97 and Phe99.

demonstrate the differential effects of conservative active-sitemutations on the binding interactions with macrolide ligandsand PPIase substrates.

Materials and methodsMaterials and chemicalsMedia components for bacteria growth were purchased fromDifco. The DNA purification kit (Magic minipreps) and T4DNA ligase were obtained from Promega. Agarose was pur-chased from Bethesda Research Laboratories. Restrictionenzymes and polynucleotide kinase were acquired from NewEngland Biolabs. Sequenase™ and other sequencing reagentswere purchased from US Biochemical. Phagemid, helper phage,Muta-Gene Kit, Coomassie Blue and protein assay reagentwere obtained from Bio-Rad. Alkaline phosphatase was boughtfrom Boehringer Mannheim, Gene-Clean kits from BiolOland Sep-Pak cartridges from Millipore. The following reagentswere purchased from Sigma: polyethylene glycol, ethidiumbromide, Trizma base, isopropyl-1-thio-P-D-galactopyranoside(IPTG), lysozyme and chymotrypsin. The tetrapeptide sub-strates for the PPIase assay were obtained from BACHEMBiosciences (Philadelphia) and the calcineurin peptide substratewas from Penninsula Laboratories. [35S]ATP and [32P]ATPwere obtained from NEN-DuPont. Oligonucleotide synthesisreagents were acquired from Applied Biosystems. FK506 andrapamycin were kindly supplied by Chugai PharmaceuticalsCo., Ltd.

Bacterial strains and plasmidsFor site-directed mutagenesis, the durung' Escherichia colistrain CJ236 was used for uracil enrichment of single-strandDNA and the Nova Blue strain (NovaGene) was used forthe selection of heteroduplex DNA after extension-ligationreactions. The E.coli strains used for the expression of humanFKBP12 mutants (Park et al., 1992) were XA90, obtainedfrom G.Verdine (Harvard University, Cambridge, MA), andCAG626 (a Ion' strain on an SCI22 background), obtainedfrom C.Gross (University of Wisconsin, Madison, WI). ApKEN2 phagemid (G.Verdine) was used for site-directed

mutagenesis. FKBP12 proteins were expressed in this vector,as described previously (Park et al., 1992).MutagenesisSite-directed mutagenesis of FKBP12 was performed asdescribed previously (Park et al, 1992). A 431 bp EcoKlfragment encoding human FKBP12 cDNA was ligated intophagemid pKEN2. All oligonucleotide-directed mutagenesiswas performed on the pKEN2-FKBP12 constructs using uracilenrichment of single-strand DNA by the modification ofKunkel (Kunkel, 1985; Kunkel et al., 1987) of the methoddescribed originally for M13 mutagenesis (Zoller and Smith,1982, 1983). A two-primer mutagenesis was performed toconstruct the double mutants. All mutants were sequencedusing the dideoxy method (Sanger et al., 1977) from thepromoter through the region surrounding the mutagenesis site,and most were sequenced in their entirety in die coding region.

Mutagenesis primers were synthesized in the (+) codingorientation. Oligonucleotides were synthesized on an AppliedBiosystems 38OA DNA synthesizer and purified by electro-phoresis in polyacrylamide-urea slab gels, followed by extrac-tion from the acrylamide gel by ammonium acetate andpurification on a Sep-Pak cartridge. The following primerswere used to generate FKBP12 mutants: for Y26F, 5'-GTG-GTGCACTTCACCGGGATG-3'; for F36L, 5'-AAGATGGA-AAGAAACTGGATTCCTCCCGG-3'; for D37V, 5'-GGAAA-GAAATTTGTATCCTCCCGTGA-3'; for R42I, 5'-CCTCC-CGTGACATCAACAAGCCCTTT-3'; for F46L, 5'-GTAAC-AAGCCCCTGAAGTTTATGCTAG-3'; for F48L, 5'-AAGC-CCTTTAAGCTGATGCTAGGCAAG-3'; for W59L, 5'-GTG-ATCCGAGGCCTGGAAGAAGGGGTT-3'; for W59A, 5'-G-TGATCCGAGGCGCTGAAGAAGGGGTT-3'; for Y80F, 5'-TATATCTCCAGATTTCGCCTATGGTGC-3'; for Y82F, 5'-CCAGATTATGCCTTCGGTGCCACTGG-3'; for H87V, 5'-GTGCCACTGGGGTACCAGGCATCAT-3'; and for F99L, 5'-CCACTCTCGTCCTGGATGTGGAGCTT-3'. The FKBP12Y26F/Y82F double mutant was made by two-primer muta-genesis with the primers designed for the individual mutants.The construction of the D37V, H87L and I90A mutantshas been described previously (Aldape et al., 1992; Futeret al, 1995).

175

M.T.DeCenzo et al

320 340 360 380 400

nm

Table II. Drug-binding

FKBPI2

Wild type"W59LW59AY26FY80FY82FY26F/Y82FF36LF46LF48LF99LD37VH87L1

I90A*

affinity of FKBPI2

K, FK506(nM)

0.6 ± 0.246.0 ± 5.3

50 ± 3122.0 ± 5.60.35 ±0.190.98 ± 0.4810.0 ± 5.40.89 ± 0.481.30 ± 0.254 0 ± 1.512 ± 8b

350 ± 5924 ± 4

0 25 ± 0.10

active-site mutants

K, rapamycin(nM)

0.26 ± 0 102 ± 2

28 ± 164.5 ± 2 5

1.10 ± 0.250.60 ± 0.2420.0 ± 4.50.22 ± 0.070.34 ± 0.16

3.0 ± 1.27 ± 2b

36.0 ± 5.820 ± 4

0.7 ± 0.2

Fig. 3. Fluorescence spectra of wild-type and W59L FKBP12 at 25°C. Spectra were determined for solutions consisting of 4.2 |iM wild-type FKBP12 with2 JJ.1 DMSO (—); 4.2 nM wild-type FKBPI2 with 9.7 (iM FK506 (- - -); 4.2 |iM W59L mutant with 2 |il DMSO (• • •). A sufficient quenching of thefluorescence signal for the F99L mutant enabled the determination of Kd values for FK506 and rapamycin by fluorescence titration.

Expression and purification of proteinsMutant duplex DNA was transformed into XA90 cells forexpression and plated on LB plates containing 100 ng/mlampicillin. Overnight cultures were started from multiplecolonies and grown at 37°C in LB-ampicillin medium. Forexpression, cultures were grown at 37°C in 2 1 shake flasksor in a 10 1 fermentor (Biostat ED, B.Braun). Culture medium(LB-ampicillin) was inoculated at a density of 1:100 andgrown to an A^ of 0.5—1.0. Protein expression was theninduced by the addition of IPTG at 1 mM final concentration.Cells were grown for 15-19 h post-induction, and harvestedby centrifugation at 4230 g for 15 min at 4°C. Cell pasteswere washed in 0.1 M Tris, pH 7.4, and frozen at -70°C untilpurification. The purification scheme of the wild type andmutant FKBP12 includes DEAE, gel exclusion and hydro-phobic interaction chromatography, and has been described indetail elsewhere (Park et al, 1992). Overall yields from thepurification scheme were s=50% for all FKBP12 mutantsdescribed.

Kinetic and binding equilibria methodsMeasurement of the catalytic efficiency {k^lK^ for the PPIasereaction was performed essentially according to Harrison andStein (1990a), with the modifications described previously(Park et al, 1992). The substrate used for the determinationof kcJKm and the inhibition constant (K,) was Suc-Ala-Leu-Pro-Phe-pNA. k^f/Km and K-, values were measured at enzymeconcentrations such that k^ was at least 5-fold higher thannon-em- Stock solutions of FK506 and rapamycin were prepared

in Me2SO. The final Me2SO concentration in the PPIase assaywas 0.5% (v/v). Reactions were initiated with chymotrypsinat a final concentration of 50 Hg/ml for assays at 15°C or 1mg/ml for reactions at 5°C. Absorbance at 400 nm wasmonitored at 1 s intervals on a Hewlett-Packard 8452Aspectrophotometer with a thermostatted cuvette holder, inter-faced to a Model 300 Hewlett-Packard computer.

Separate A , and Km values were determined using themodified coupled chymotrypsin assay devised by Kofron et al.(1991). Tetrapeptide substrates were dissolved at 150-225 mMin anhydrous trifluoroethanol (TFE)/470 mM LiCl. For a 1 mlreaction volume, 6 mg chymotrypsin were added. To initiatethe reaction, substrate solutions and additional TFE were addedto bring the final cosolvent concentration to a constant value

>n ^ PPIase reaction and drug affinities of FKBPI2 mutants weredetermined with Sue-Ala—Leu-Pro-Ptie-pNA as substrate at 15°C. Thesedata for wild-type FKBP12 and the D37V mutants were reported previouslyin Aldape et al. (1992), and for the H87L and I90A mutants in Futer et al.(1995).''Since the F99L mutant was a poor catalyst of the PPIase reaction (0.2% ofwild type), ligand affinities (A j) were determined by fluorescence titration at15°C using 203 nM injections of FK506 into a 920 nM solution of protein.

of 2.5% (v/v). Assays were performed at 5°C to minimize thenon-enzymatic PPIase rate. Reactions were monitored at oneof several wavelengths, depending on the concentration ofthe peptide substrate and p-nitroaniline (pNA) product. Thewavelengths used (and corresponding £„, values) were 444(1250), 450 (760), 454 (530) and 460 nm (300). Substratespecificities of the mutants were determined using Suc-Ala-Pi-Phe-pNA as the substrate, where P( is Leu, Phe, Val, Ala,Lys or Gly.

The dissociation constants of the F99L mutant for ligandswere determined by fluorescence titration. The measurementof fluorescence spectra was performed on a Perkin-Elmer LS-50 instrument using FLDM software running on an IBM PC-XT. The excitation wavelength was 290 nm (5 nm slit width)and emission scans were obtained over the wavelength range300-380 nm (15 nm slit width). Samples were equilibratedto 5°C prior to collecting the spectra. 1.0 |il injections of a

176

FKBP12 mutagenesis

0.6 mM solution of FK506 or rapamycin in ME2SO weremade into 3 ml of a 0.9 |lM mutant FKBP12 solution in 0.1M Tris, pH 7.8, at 15°C. Five scans were collected andaveraged for each sample (Figure 3). Scans were corrected forthe minor emission of the peptide by subtracting averagedscans of the peptide in buffer at an equivalent concentrationto that used in the titration of FKBP12.

Data analysisFor the standard coupled PPIase assay, absorbance data pointswere fitted to a first-order function using Hewlett-Packard dataanalysis software for the calculation of /:ca/A'm. K, values werecalculated using non-linear fitting to a competitive tight bindingequation (Morrison, 1969) employing KineTic software v 3.0(BioKin Ltd) running on a Macintosh Hex. Separate £cal andKm parameters were calculated by fitting progress curves toan integrated rate equation derived by Kofron et al (1991)using the KineTic software package running on a MacintoshHex. The integrated rate equation includes terms for both thecatalyzed and uncatalyzed isomerization rates. Initial reactionvelocities calculated by this program were then used tocalculate kcM and Km.

The fluorescence titration data were analyzed by fitting toa Scatchard equation using MultiFit software (Day Computing)running on a Macintosh Hex. The fractional change in fluores-cence intensity at 340 nm was used to solve iteratively for thefree drug concentration at each injection point as a functionof Xt, Af, and Ka (the association constant). These data werethen fitted to a non-linear least squares fit utilizing a Marquardtalgorithm (Connelly et al, 1994).

ResultsFK506 and rapamycin bindingSite-directed mutagenesis was used to make conservativemutations of the FKBP12 internal residues that contact FK506through hydrophobic and hydrogen bonding interactions(Figure 2A and Table I). The indole ring of Trp59 FKBP12 isburied by the pipicolinyl ring of FK506 or rapamycin in therespective complexes (Van Duyne et al, 1993; Wilson et al,1995). The mutation of Trp59 to either Leu or Ala results ina 10- to 100-fold decrease in affinity for FK506 and rapamycincompared with wild-type FKBP12 (Table IT). Phe99 is a centralresidue at the base of the binding pocket, interacting with thearomatic residues Phe36, Tyr26 and Trp59, as well as theFK506 pipicolinyl ring (Figure 2A and Table I). It has alsobeen reported to form an unusual type of hydrogen bondinginteraction from the side-chain Cj-H to the FK506 C9 ketooxygen atom (Van Duyne et al, 1991, 1993; Figure 2A). SuchCaTOm-H---O interactions have been observed in many proteins(Burley and Petsko, 1988) and are believed to be energeticallyfavorable, although not to the same degree as a typicalhydrogen bond (Thomas et al, 1982). An F99L mutationyields a protein with a 20- to 30-fold lower affinity for FK506and rapamycin than the parent protein. Because of the lowenzymatic activity of this mutant, the binding affinities (K4)of the macrolide ligands were determined by fluorescencetitration (Figure 3).

Tyr26 forms a hydrogen bond to Asp37, as well as theunusual side-chain C5-H hydrogen bond to the FK506 C9 ketooxygen atom. The mutation of Tyr26 to Phe decreases theaffinity for FK506 and rapamycin by 37- and 17-fold, respect-ively. Tyr82 forms a hydrogen bond with the FK506 C8 amideoxygen via the phenol proton. However, the mutation of Tyr82

does not affect FK506 or rapamycin affinity. His87 stacksagainst Tyr82, and contacts the FK506 or rapamycin C10-C14 pyran ring and the Cll methyl group. The mutation ofHis87 to Leu lowers the affinity for FK506 and rapamycin40- and 77-fold, respectively (Futer et al, 1995). Only minoreffects on FK506 or rapamycin binding are seen in the Y80Fmutant, which is not in contact with the ligands.

Phe36 also forms a weak hydrogen bond to the C9 ketooxygen atom of FK506 via the aromatic Cg-H. This mutantwould no longer be able to form this hydrogen bond, yet theK, values of the F36L mutant for FK506 and rapamycin areunchanged compared with wild-type FKBP12. The mutationof Phe46 or Phe48, residues which do not form hydrogenbonding interactions with FK506, decreases drug-bindingaffinity by 2- and 7-fold, respectively. Phe46 does makeextensive Van der Waals interactions with the pipicolinyl ringof FK506.

Asp37 is a hydrogen bond acceptor for the Tyr26 phenolproton and for the FK506 C10 ketal hydroxyl proton. Inthe FK506 and rapamycin complexes, Asp37 also forms anelectrostatic bond to Arg42 via the guanidino nitrogen proton.The D37V mutant has an affinity for FK506 and rapamycinthat is 580- and 180-fold lower, respectively, than the wild-type protein (Aldape et al, 1992). This is the most dramaticchange that we have observed in drug affinity caused by apoint mutation. Mutations at position Ile90 of FKBP12 haveminimal effects on ligand binding. I90A FKBP12 has K, valuesfor FK506 and rapamycin of 0.25 and 0.70, respectively (Futeret al, 1995). These K, values are comparable with those ofwild-type FKBP12 (K, = 0.6 nM for FK506 and 0.2 nM forrapamycin).

PPIase catalysisAll FKBP12 mutants characterized are active catalysts of thePPIase reaction. Specific activities of FKBP12 variants rangefrom 0.2 to 175.0% of the wild-type protein. However, whereasmost mutations have minor effects on catalytic efficiency,mutations at four residues in the active site cause >2-folddecreases in the second-order rate constant (k^/Km) for thisreaction. Mutations at two hydrophobic FKBP12 residues atthe base of the active site cause the most significant decreasein catalytic efficiency observed (Figure 4). The mutation ofTrp59 to Leu or Ala decreases catalytic efficiency to 13 and3% of the wild-type activity, respectively. The mutation ofPhe99 to Leu decreases the catalytic efficiency to 0.2% of thewild-type activity. Mutations at the other hydrophobic bindingpocket residues (Phe36, Phe46 or Phe48 to Leu) result inFKBP12 variants with PPIase specific activities very similarto the wild-type enzyme.

The effects of mutating the FKBP12 Tyr residues 26 and82 are of interest because these residues both interact withFK506 (Table I). It has been proposed that hydrogen bondinginteractions of substrate with these residues may play animportant role in catalysis (vide infra). However, mutatingeither or both of these residues to Phe generates catalyticallyactive variants (Figure 4), suggesting that hydrogen bondinginteractions with these residues are not essential for substratebinding or catalysis. In contrast, two hydrophilic residues atthe solvent-exposed interface of the binding pocket, Asp37and Arg42, which form hydrogen bonding interactions withligands, were altered in previous investigations (Aldape et al,1992; Futer et al, 1995), and these mutations were observedto cause significant decreases in PPIase activity.

177

M.T.DeCenzo et al

LEU

PHE

VAL

ALA

LYS

1E+07

_ 1E+06

= 1E+05

1E+04

1E+O3

1E+07

V 1E+06

1E+05

1E+04

1E+03

Fig. 4. FKBPI2 and FKBP12 mutant catalytic efficiency and substrate specificity in the PPIase reaction. The tetrapeptide substrates used were Sue-Ala—P|-Pro-Phe-pNA, where the P, residue is indicated in the legend The assays were performed at 15°C at a substrate concentration of 30 |iM and FKBP12concentrations varying from 5 nM to 3.5 |iM.

Table III. Interactions of active-site FKBP12 residues with peptidyl-prolyl substrates observed in computer simulations of the PPIase reaction1

FKBP12residue atom

Peptide atomhydrogen bond

Peptide conformationwith strongest interaction

Also observed inMD simulation?6

Tyr26 ring HPhe36 ring HAsp37 side-chain OGlu54 backbone C=OIle56 backbone NHTyr82 OHPhe99 ring H

PI C=OPI C=OPI amide NP2' amine Nprolyl C=OPI C=OPI C=O

CIS

twisted-amidetwisted-amide = transtwisted-amidetwisted-amidetranstwisted-amide

nonoyesnononoc

yes

"FKBP12 residues that interact with tetrapeptide PPIase substrates, predicted from the adiabatic mapping simulations of the FKBPI2-catalyzed PPIase reactionreported by Fischer et al. (1993). Hydrophobic pocket-forming residues predicted for the substrate P, side chain are Phe36, Tyr82, His87, Ile90, Ile91, Ile97and Phe99. Trp59 interacts with the substrate Phe residue.•"Comparison of the results for FKBP12 residues that interact with tetrapeptide PPIase substrates, predicted from the MD simulations of the FKBPI2-catalyzedPPIase reaction reported by Orozco et al. (1993). In general, there are few similarities in the specific hydrogen bonding interactions seen in the twosimulations (Figure 2B).Tischer et al. (1993) observed a direct hydrogen bonding interaction with the backbone carbonyl of the P2' residue to the Tyr82 OH in their simulation.However, Orozco et al. (1993) suggested that this residue forms either a direct or water-mediated hydrogen bond with the substrate P, carbonyl group.

Substrate specificity profiles for tetrapeptides varying at thePi position were determined for each of the FKBP12 mutantsby measuring kcJKm. The observed relative substrate specifi-city for each mutant was usually very similar to that observedfor wild-type enzyme. The Y82F and H87L mutants areexceptions. These display a lower relative catalytic efficiencythan the wild-type protein for the P, = Phe tetrapeptidesubstrate (Figure 4) and a higher catalytic efficiency forPi = Lys than the wild-type FKBP12. This altered substratespecificity was also observed for a Y26F/Y82F double mutant.A more detailed kinetic analysis of PPIase catalysis by theY82F and Y26F/Y82F mutants was thus undertaken.

Separation of the second-order rate constant into individualKm and k^ constants for the cis-trans reaction was accomp-

lished using a modified PPIase assay at a lower temperature(5°C; Kofron et al, 1991; Park et al, 1992). At this temper-ature, the ratio of the cis.trans form of the substrate is shiftedin the a s direction. This analysis (Table IV) revealed thatneither the Y82F nor the Y26F/Y82F mutations substantiallyalter the Km value for FKBP12-catalyzed isomerization of thePi = Leu, Phe or Gly substrates, suggesting that the bindingof the cis substrate to the protein is unaffected. As we havereported previously (Park et al, 1992), the substrate specificityof FKBP12 catalytic efficiency is predominantly due to a k^effect, which may indicate that a protein-ligand interactionstabilizing the twisted-amide intermediate has been perturbed.The Y82F mutant shows an 8-fold decrease in Jt^, for thePi = Leu substrate versus the parent protein, a similar

178

FKBP12 mutagenesis

Table IV. Kinetic parameters of FKBP12 Tyr mutants in the PPlase reaction

FKBP12 P,

Wild type Km (M)a

*ca (s"1)l^tt/^m <

Y82F Km

Y26F/Y82F ATm

W*m

Leu

1 5X10"4

ib 600(1);M-'s-')c 1.2X106

2.9XKT1

75(1)2.6XI03

7.5X10^48(1)6.4X104

Phe

3.7XI0"4

267 (0.45)7.2XIO5

2.6X10"4

29 (0.39)1.IX103

8.0X10"4

8(0.17)9.9X103

Gly

3.5XI0"4

0.09 (1.5 X10"4)2.6X102

4.6X10"4

18(0.24)3.9X104

4.6X10^0.9 (0.02)1.9X103

Tetrapeptide substrates in the PPlase reaction used are Sue-Ala—P(—Pro-Phe-pNA, where P| is Leu, Phe or Gly.'Km and t ^ values were determined at 5°C by a modified PPlase assayusing the solvent-perturbed equilibria method developed by Kofron et al.(1991) with the modifications described previously (Park et al., 1992).''Relative k^ (to P| = Leu) in parentheses.Ckcu/Km was determined at 5°C by the standard PPlase assay (Aldape et al.,1992; Park el al., 1992).

discrimination in kcal between the P, = Leu versus Phesubstrates, and less discrimination versus the P| = Glysubstrate. The Y26F/Y82F double mutant is more discriminat-ing than the wild-type protein with respect to the kcU valuefor P, = Leu versus Phe, but less discriminating with respectto P, = Gly.

DiscussionBased on an analysis of the magnitude and pH dependenceof steady-state kinetic parameters, activation parameters andsolvent and deuterium isotope effects, a catalytic mechanismfor the PPlase reaction was proposed (Harrison and Stein,1990b; Park et al, 1992). We suggested that the selectivestabilization of an X-Pro twisted-amide intermediate anddesolvation were important rate-accelerating factors in PPlasecatalysis, lowering the free energy of activation of formationof this intermediate by 6.6 kcal/mol relative to the uncatalyzedreaction. Although the crystal structure of an FKBP12^>eptidesubstrate complex has not been reported, the molecular detailsof the PPlase reaction catalyzed by FKBP12 were subsequentlyinvestigated in two independent simulation studies (Fischeretal, 1993; Orozco etai, 1993). Fischer etal. (1993) modeledthe substrate (Ac-Leu-Pro-Phe-NMe) as a type Via prolineturn, assuming that the twisted-amide transition state for thesubstrate is mimicked by the C9 keto carbonyl of FK506 orrapamycin, and that the other interactions seen with theseinhibitors are largely preserved. The adiabatic reaction pathfor the cis-trans interconversion was then mapped out using acombination of molecular mechanics, ab initio, and continuumdielectric methods. It should be noted that the simulations ofFischer et al. (1993) were performed in the gas phase. Assuch, they cannot invoke water-mediated hydrogen bonds,which may play an important role in catalysis. Likewise,Orozco et al. (1993) modeled the twisted form of their substrate(Ac-Ala-Ala-Pro-Phe-NMe) on rapamycin, but then carriedout extensive solvated molecular dynamics (MD) and freeenergy perturbation calculations.

These investigators proposed details of potential hydrogenbonding and Van der Waals interactions between FKBP12 andpeptidyl-prolyl substrate atoms over the time course of theirsimulation of the cis-trans isomerization, and predicted howthe strength of these interactions change as the peptidyl-prolyl

substrate is isomerized (Table III; Figure 2B). Further, FK506and rapamycin have been proposed to mimic the twisted-amide substrate intermediate (Rosen et al., 1990; Schreiber,1992). Careful comparison of the effects of point mutationson /tca/^n, for tetrapeptide substrates with the effects on drug-binding constants can be a useful test of the more detailedPPlase reaction mechanisms and whether the macrolide ligandsindeed behave as transition state mimics (Rosen et al., 1990;Schreiber, 1992).

The structure of the FKBP12-FK506 complex revealed anumber of hydrophobic and hydrogen bonding interactionsbetween the protein and ligand (Table I). The mutation ofTrp59 and Phe99 resulted in dramatic decreases in FK506 andrapamycin binding, consistent with the importance of thehydrophobic interactions of these residues with the substrateprolyl moiety observed in the MD simulations of Orozco et al.(1993). Asp37 hydrogen bonds to the C10 hemiketal hydroxylgroup of FK506, and the mutation of this residue to Val alsodramatically lowers its affinity for macrolide ligands. Themutation of FKBP12 residue Tyr26 to Phe may maintain theunusual Cg-H hydrogen bond to FK506 (to the C9 keto oxygenatom), but nonetheless results in a 50-fold decrease in K, forFK506. Tyr26 makes a hydrogen bond to Asp37, and the Aspside-chain carboxylate forms a salt bridge to Arg42 as well asa hydrogen bond to the C10 hydroxyl group of FK506. Themutation of Tyr26 to Phe may cause a conformational shift ofthe Asp37 side chain, disrupting these interactions. However,the mutation of Tyr82, which forms a hydrogen bond to theFK506 C8 amide oxygen, does not affect the K-, value. Athermodynamic study of the FKBP12 Y82F mutant providedevidence that the enthalpy required to displace two solventmolecules hydrogen bonded to the Tyr82 phenol proton largelyoffsets the free energy of binding of FK506 (Connelly et al.,1994).

The mutagenesis results reported here demonstrate that thePPlase catalytic efficiency of FKBP12 is remarkably insensitiveto mutation in the active site. In fact, every mutant wegenerated is catalytically active. The mutation of Asp37decreases the catalytic efficiency 10-fold (Aldape etal, 1992).However, this mutant is not 'catalytically inactive', as ourpreviously reported data were erroneously characterized(Fischer et al, 1993). Nonetheless, the significant decrease incatalytic efficiency of the Asp37 mutant may reflect the roleof Asp37 in destabilizing the cis form of the substrate bymaking an unfavorable electrostatic interaction with the sub-strate P| carbonyl oxygen (Fischer et al, 1993; Orozco et al,1993), and in stabilizing the twisted-amide intermediate byforming a hydrogen bond to the substrate P| amide nitrogen.

The mutation of Trp59 to Ala and Phe99 to Leu, residueslikely to interact with the substrate prolyl ring, also decreasescatalytic efficiency. The mutation of Tyr26 or the other active-site Phe residues has only minor effects on PPlase activity,suggesting that if these residues hydrogen bond to the substratePi carbonyl oxygen, as proposed by Fischer et al (1993), suchhydrogen bonds neither significantly destabilize the groundstate interaction nor stabilize the twisted-amide intermediate.

In characterizing >20 FKBP12 mutants to date, we havefound few variants to display an altered PPlase substratespecificity at P, relative to the parent protein (Aldape et al,1992; Park et al, 1992). Exceptions to this are the Y82F andH87L single mutants and a Y26F/Y82F double mutant, inwhich we observed altered substrate specificity for a series oftetrapeptide substrates (Figure 4). These residues form stacking

179

M.TJ>eCenzo et al

interactions with each other and, along with Phe36, Ue90,Ile91, De97 and Phe99, form the hydrophobic pocket for thesubstrate P, residue. Tyr82 has been proposed to stabilize thetwisted-amide substrate intermediate via a direct hydrogenbond to the P2 ' carbonyl group (Fischer et al, 1993) or via awater-mediated hydrogen bond to the P] carbonyl oxygen(Orozco era/., 1993).

Then we determined whether the altered substrate specificit-ies observed for the Y82F mutant and the Y26F/Y82F doublemutant were caused by the perturbed binding of the cissubstrate (reflected in K^ or of the twisted-amide intermediate(reflected in Jtcat). A kinetic analysis of the Y82F and Y82F/Y26F mutants (Table IV) indicated that the altered substratespecificity profiles for these mutants were clearly caused byan effect on k^. Km values were indistinguishable withinexperimental error, suggesting that the binding of the cissubstrate to FKBP12 is unaltered. These observations are ingood agreement with the trends in the catalytic parametersreported for an FKBP12 Y82L mutant by Bossard et al. (1994),who also found no effect on Km and a 50-fold decrease in k^.The observed effect on /fc , supports a role for these residues instabilizing the twisted-amide intermediate through interactionswith the substrate P, residue. The altered substrate specificityof the Tyr82 and His87 mutants is consistent with the formationof water bridges by these residues to the twisted-amide formof the substrate observed during the simulations of Orozcoet al. (1993). The mutation of Tyr82 to Phe increases therelative kcal for P, = Phe or Gly substrates. The increaseddiscrimination of /^al for P, = Leu versus Phe in the Y26F/Y82F double mutant versus Y82F probably reflects the roleof Tyr26 in stabilizing the Asp37 side chain, which hydrogenbonds to the backbone amide nitrogen of the P, residue. Thehydrogen bonding of the Tyr26 to the substrate P| carbonyl(via the aromatic C-H group) appears to be less importantbecause a Phe26 residue should also be able to form thishydrogen bond.

Therefore, we can conclude that while hydrogen bondingby backbone amide heteroatoms and the Asp37 carboxylateside chain are rate-accelerating factors in PPIase catalysis byFKBP12, hydrogen bonding by Tyr26, Phe36 or Phe99 doesnot contribute significant stabilization energy to the twisted-amide intermediate relative to the cis peptide substrate. Instead,these data provide additional evidence that the catalysis ofthe PPIase reaction proceeds primarily by desolvation anddestabilization of the ground state, and by stabilization of theintermediate, primarily through Van der Waals interactionswith residues such as Trp59 and Phe99. Such interactions arealso the major contributors to the binding energy of FKBP12complexes for FK506 and rapamycin. Further, the differentialeffects of these binding pocket mutations on catalytic efficiencyversus macrolide-binding affinity lend support to the suggestion(Park et al, 1992; Orozco et al, 1993) that FK506 andrapamycin should not be considered as exact 'twisted-amidesurrogates', as has been proposed by Rosen et al. (1990). Thecharacterization of the critical FKBP12-macrolide interactionsby mutagenic analysis, together with the structural characteriza-tion of the FKBP12-FK506-calcineurin complex (Griffithet al, 1995) and structure-activity of FKBP12-binding com-pounds (Armistead et al, 1995), should facilitate the designof novel FKBP12-binding immunosuppressive compounds.

AcknowledgementsThe authors thank Mark Fleming for oligonucleotide synthesis and Dr PatrickConnelly for his assistance with fluorescence titrations. We are grateful to

180

Drs D.Armistead, C.Lepre, M.Navia, J.Saunders and K.Wilson for helpfuldiscussions and critical reading of the manuscript.

ReferencesAldape.R.A., Futer.O., DeCenzo.M T., Jarreft,B.P., Murcko.M A. and

Livingston.DJ. (1992)7. Biol. Chem., 267, 16029-16032.Armistead.D.M. and Harding.M.W. (1993) Annu. Rep. Med. Chem., 28,

207-215.Armistead.D.M. et al. (1995) Ada Crystallogr., D51, 522-528.Bossard.M.J., Bergsma.D.J., Brandt.M., Livi.G.P, Eng.W.K., Johnson.R.K.

and Levy.M.A. (1994) Biochem. J., 297, 365-372.Brown.E.J., Albers,M.W., ShinJ.B., Ichikawa,K., Keith.C.T., Lane.W.S. and

Schreiber.S.L. (1994) Nature, 369, 756-758.Burley.S.K. and Petsko.G.A. (1988) Adv. Protein Chem., 39, 125-189.Cameron.A.M., SteinerJ.P., Sabatini,D.M., Kaplin.A.I., Walensky.L.D. and

Snyder.S.H. (1995) Proc. Natl Acad. Sci. USA, 92, 1784-1788.Connelly.P. et al (1994) Proc. Natl Acad. Sci. USA, 91, 1964-1968.DiLella,A.G. and Craig.R.L. (1991) Biochemistry; 30, 8512-8517.Fischer.S., Michnick.S. and Karplus.M. (1993) Biochemistry, 32, 13830-13837.Fruman.D.A., Burakoff,S.J. and Bierer.B.E. (1994) FASEB J., 8, 391^HX).Futer.O., DeCenzo,M.T., Aldape.R.A. and Livingston.DJ. (1995) J. Biol.

Chem., 270, 18935-18940.Galat.A. (1993) Eur. J. Biochem., 216, 689-707.Griffith J.P. et al. (1995) Cell, 82, 507-522Harding.M.W., Galat.A., Uehling.D.E. and Schreiber.S.L. (1989) Nature, 341,

758-760.Harrison.R.K. and Stein.R.L. (1990a) Biochemistry, 29, 3813-3816.Harrison,R.K. and Stein.R.L. (1990b) Biochemistry, 29, 1684-1689.Hendrickson.B.A., Zhang.W., Craig.RJ., Jin,Y.-J., Bierer.B.E., Burakoff.S.

and DiLella.AG (1993) Gene, 134, 271-275Jayaraman.T., Brillantes.A.M., Timerman.A.P., Fleischer.S., Erdjument-

Bromage,H.,Tempst,P.andMarks,A (1992)7. Biol. Chem., 267,9474-9477.KofronJ.L., Kuzmic.P., Kishore.V, Colon-Bonilla.E. and Rich.D.H. (1991)

Biochemistry, 30, 6127-6134.Kunkel.T.A. (1985) Proc. Natl Acad. Sci. USA, 82, 488-492.Kunkel.T.A., RobertsJ. and Zakour.R. (1987) Methods Enz\mol., 154, 367-

382.KunzJ., Henriques.R., Schneider,U., Deuter-Reinhard,M., Mowa,N.R. and

Hall.M.N. (1993) Cell, 73, 585-596.LiuJ., FarmerJ.D., Lane.W.S., FriedmanJ., Weissman.I. and Schreiber.S.L.

(1991) Cell, 66, 807-815.Mayerleitner.M., Timerman.A.P., Wiedderecht,G. and Fleischer.S. (1994) Cell

Calcium, 15, 99-108.MooreJ.M., Peattie.D.A., Fitzgibbon,M.J. and ThomsonJ.A. (1991) Nature,

351, 248-250.Morrison J.F. (1969) Biochim. Biophys. Ada, 185, 269-286Orozco.M., Tirado-RivesJ. and Jorgensen.W.L. (1993) Biochemistry, 32,

12864-12874.Park.S.T., Aldape.R.A., Futer.O., DeCenzo.M.T. and Livingston,D.J. (1992)

J. Biol. Chem., 267. 3316-3324.Peattie.D.A., Harding.M.W., Fleming.M.A., DeCenzo.M.T., LippkeJ.A.,

Livingston.D.J. and Benasutti.M. (1992) Proc. Natl Acad. Sci. USA, 89,10974-10978.

Peattie.D.A., Hsaio.K., Benasutti.M. and LipkeJ.A. (1994) Gene, 146, 239-244.

Rosen.M.K., Standaert,R.F, Galat.M., Nakatsuka.M. and Schreiber.S.L. (1990)Science, 248, 863.

Sabatini.D.M., Erdjument-Bromage.H., Lui.M., Tempst.P. and Snyder.S.H.(1994) Cell, 78, 35^3 .

Sanger.F, Nicklen.S. and Coulson,A.R. (1977) Proc. Natl Acad. Sci. USA,74, 5463-5467.

Schreiber.S.L. (1992) Science, 251, 283-287.SiekerkaJJ., Hung.S.H.Y.. Poe.M., Lin.C.S. and Sigal.N.H. (1989) Nature.

341, 755-757.Sigal.N.H. and Dumont.F. (1992) Annu. Rev. Immunol., 10, 519-560.Thomas.K.A., Smith.G.M., Thomas.T.B. and Feldman.RJ. (1982) Proc. Natl

Acad. Sci. USA, 79, 4843^»857.Van Duyne.G.D., Standaert,R.F, Karplus.P.A., Schreiber.S.L. and ClardyJ.

(1991) Science, 252, 839-842.Van Duyne.G.D., Standaert.R.F, Karplus.P.A., Schreiber.S.L. and ClardyJ.

(1993) J. Moi. Biol., 229, 105-124.Wilson.K.P. et al (1995) Ada Crystallogr, D51, 511-521.Zoller.M.J. and Smith.M. (1982) Nucleic Acids Res., 10, 6487-6500.Zoller.M.J. and Smith.M. (1983) Methods Enzymol., 100, 468-500.

Received August 19, 1995; revised October 29, 1995; accepted November14, 1995