Proc. Natl. Acad. Sci. USAVol. 93, pp. 5191-5196, May 1996Immunology

Crystal structure at 2.6-A resolution of human macrophagemigration inhibitory factor

(protein structure/cytokine/sepsis/glucocorticoid)

HONG-WEI SUN*, JURGEN BERNHAGENtt, RICHARD BUCALAt, AND ELIAS LOLIS*§

*Department of Pharmacology, Yale University, New Haven, CT 06510; and tThe Picower Institute for Medical Research, Manhasset, NY 11030

Communicated by Anthony Cerami, The Picower Institute for Medical Research, Manhasset, NY January 24, 1996 (received for reviewJanuary 16, 1996)

ABSTRACT Macrophage migration inhibitory factor(MIF) was the first cytokine to be described, but for 30 yearsits role in the immune response remained enigmatic. In recentstudies, MIF has been found to be a novel pituitary hormoneand the first protein identified to be released from immunecells on glucocorticoid stimulation. Once secreted, MIF coun-terregulates the immunosuppressive effects of steroids andthus acts as a critical component of the immune system tocontrol both local and systemic immune responses. We reportherein the x-ray crystal structure of human MIF to 2.6-Aresolution. The protein is a trimer of identical subunits. Eachmonomer contains two antiparallel a-helices that packagainst a four-stranded a-sheet. The monomer has an addi-tional two ,8-strands that interact with the a-sheets of adja-cent subunits to form the interface between monomers. Thethree ,B-sheets are arranged to form a barrel containing asolvent-accessible channel that runs through the center of theprotein along a molecular 3-fold axis. Electrostatic potentialmaps reveal that the channel has a positive potential, sug-gesting that it binds negatively charged molecules. The eluci-dated structure for MIF is unique among cytokines or hor-monal mediators, and suggests that this counterregulator ofglucocorticoid action participates in novel ligand-receptorinteractions.

Macrophage migration inhibitory factor (MIF) is a protein of115 amino acids that is expressed in a wide variety of tissuesincluding the cells of the immune system, anterior pituitary,liver, developing eye lens, and brain (1-4). MIF is consideredto be the first lymphokine discovered and was first describedover 30 years ago as a soluble factor produced by activated Tcells that inhibits the migration of macrophages (5, 6). It hasa long association with delayed-type hypersensitivity reactions(5), but, more recently, MIF has been found to be a novelanterior pituitary hormone and a critical mediator of septicshock (1). The corticotrophic cells of the anterior pituitarycontain large stores of preformed MIF in secretory granulesthat are released into the circulation by stress or infectiousstimuli such as lipopolysaccharide (7, 8). The role of MIF insepsis is supported by studies that show that the injection ofMIF potentiates lipopolysaccharide-induced death, and thatthe administration of anti-MIF antibodies fully protects ani-mals from lethal endotoxemia. MIF is also the first protein tobe identified that is secreted directly from immune cells onglucocorticoid stimulation (8). Once released, MIF "over-rides" the anti-inflammatory effects of steroids on cytokine[tumor necrosis factor-a, interleukin 1B (IL-1f3), IL-8] pro-duction and thus appears to act physiologically as a unique,glucocorticoid-induced counterregulatory hormone. Thecounterregulatory activity of MIF also has been demonstratedto be critical in the pathogenesis of septic shock, as the

coadministration of MIF with dexamethasone to endotoxemicmice can completely block the protective effect of dexameth-asone in this experimental model (8).MIF also may have functions outside the immune system.

For instance, a developmental role for MIF has been proposedbased on the expression of MIF mRNA in the eye lens duringthe differentiation of epithelial cells (4). MIF may play a rolein the hepatic detoxification system as one study reportsglutathione S-transferase (GST) activity for native MIF (9);however, this activity for MIF remains controversial. Morerecently, purified recombinant MIF has been found to catalyzea tautomerization reaction suggesting that an enzymatic reac-tion may underlie some of its biological properties (10).As part of an effort to understand the function of this unique

hormonal mediator that possesses glucocorticoid counterregu-latory and enzymatic activities, we have determined the crystalstructure of human MIF at 2.6-A resolution. The three-dimensional structure presented here can be used for furtherinvestigation of the receptor-binding and catalytic sites of thismultifunctional protein.

MATERIALS AND METHODSExpression and Purification of Recombinant MIF. Recom-

binant MIF was expressed in Escherichia coli using the T7polymerase-based pET-llb plasmid as previously described(44). Briefly, a 1-liter cell growth was grown to an OD600 of 0.7and induced by addition of isopropyl f-D-thiogalactopyrano-side to a final concentration of 0.4 mM. After a 4-hr growth,cells were harvested, resuspended in 40 ml of 20 mM Tris/20mM NaCl (pH 7.4), and lysed using a French press. Cell debriswas removed by centrifugation at 35,000 x g and the super-natant was applied to two ion-exchange resins (Mono Q andMono S). MIF was not retained by either column but retentionof bacterial proteins resulted in 95% purity. The remainingcontaminants were removed by gel-filtration chromatography(Superdex 200), although this was not essential for crystalli-zation. The yield was "40 mg of pure MIF per liter of growth.Amino acid and N-terminal analyses indicate that the N-terminal methionine is posttranslationally removed (data notshown).

Selenomethionine was incorporated into MIF using themethionine auxotroph B834 (DE3) (Novagen) as described(44). Briefly, a clone transformed with pETllb-MIF wasgrown in media containing M9 salts (11), 2 mM MgSO4, 1.2%dextrose, 0.1 mM CaCl2, 0.1% casamino acids, 1 mg/ml

Abbreviations: MIF, macrophage migration inhibitory factor; IL, in-terleukin; GST, glutathione S-transferase.Data deposition: The atomic coordinates and structure factors havebeen deposited in the Protein Data Bank, Chemistry Department,Brookhaven National Laboratory, Upton, NY 11973 (reference1MIF).*Present address: Department of Biochemistry and Cell Biology,Fraunhofer Institute/IGB, D-70569 Stuttgart, Germany.§To whom reprint requests should be addressed.

5191

The publication costs of this article were defrayed in part by page chargepayment. This article must therefore be hereby marked "advertisement" inaccordance with 18 U.S.C. §1734 solely to indicate this fact.

Dow

nloa

ded

by g

uest

on

Oct

ober

29,

202

0

Proc. Natl. Acad. Sci. USA 93 (1996)

thiamine, and 100 ,g/ml ampicillin. Cells were diluted (1:100)into media in which the 0.1% casamino acids had beenreplaced with 40 ,g/ml of each amino acid (excluding methi-onine and cysteine) and 40 gg/ml of selenomethionine. Thecells were induced with 0.4 mM isopropyl f3-D-thiogalactopy-ranoside at an OD600 of 0.85 and allowed to grow for another15 hr. Selenomethionyl-substituted MIF was purified as above.

Crystallization and Data Collection. MIF was concentratedto 15 mg/ml before crystallization. Diffraction quality crystalswere grown using the hanging drop vapor diffusion method asdescribed (12, 44). Equal volumes (2-5 ,ul) of protein andreservoir solution (2.0 M ammonium sulfate/2% PEG 400/0.1M Hepes, pH 7.5) were mixed and allowed to equilibrate atroom temperature. Crystals grew overnight and typicallyreached their final size by 2 weeks. Larger crystals could begrown by addition of 2-5 ,tl of additional protein, whichinitially would dissolve the crystals, but lead to larger crystalsovernight. Crystals with sizes of 0.6 mm x 0.4 mm x 0.2 mmwere used for most heavy atom soaks and data collection.These crystals belong to space group P3121 with unit celldimensions a - b - 96.70 A and c - 106.14 A. Selenomethionyl-substituted MIF crystals were grown under identical condi-tions. These crystals were generally smaller than the wild-typecrystals but diffracted to higher resolution.

All crystals were mounted in quartz capillary tubes, dried ofmother liquor, but surrounded on both sides with solutionfrom the reservoir and sealed with wax. X-ray data werecollected on a Siemens model X-1000 detector with a Rigakumodel RU200 rotating anode x-ray generator operating at 50kV and 60 mA producing Cu Ka x-rays focused with adouble-mirror (0.2-mm focus) system. Data were collected as0.25° w-scans at room temperature, processed with XDS, andscaled with XSCALE (13).

Structure Determination and Refinement. The three-dimensional structure was solved by multiple isomorphousreplacement. Useful heavy atom derivatives were obtained byadding heavy atom solutions (ethylmercury phosphate,K2PtCl4, and KAuCl4) directly to the crystals. Derivative datasets were scaled to the native data using the PHASIT package(W. Furey, University of Pittsburgh). Isomorphous differencePatterson maps and the direct-methods program SHELXS (14)were used to determine the three heavy atom positions for thegold derivative. Single isomorphous replacement phases werecalculated and used in cross difference Fourier maps to locatethe heavy atom sites in each of the other derivatives. Six sitesare present in the platinum and ethylmercury phosphatederivatives, and nine selenium sites are present in the sel-enomethionyl-substituted MIF crystal. The positions and oc-cupancies for these sites were refined using programs in thePHASIT package. Table 1 presents a summary of the diffractiondata and phasing statistics. The heavy atom sites for eachderivative are shown in Table 2. An initial electron density mapwas calculated to 3.5-A resolution in space groups P3121 orP3221. Right-handed helices were clearly visible in the mapcalculated with P3121. A 2.8-A electron density map wascalculated with these final phases in space group P3121 andfound to be clear and easily interpretable (Fig. 1). Solventflattening increased the figure of merit from 0.79 to 0.88. Askeleton representation of the density revealed the trimericarrangement and overall fold of the molecule (15). Theprogram 0 was used to trace the backbone, create a polyala-nine chain, and build the initial model of one entire subunit.Due to the unusual intertwining of polypeptide chains amongthe three subunits, the (Fo,Se-Fo,native) electron density of theselenium sites was used to confirm that the chain was tracedcorrectly. This was achieved by the convenient location of thethree methionine residues that were used as landmarks: me-thionines are near theN terminus (Met-2), near the C terminus(Met-101), and near the middle of the polypeptide chain(Met-47). To build the entire trimer, the first subunit was

Table 1. Statistics

Parameters Native [Se-Met] EMP KAuCl4 K2PtCl4Number of crystals 2 1 1 1 1Concentration (mM) 1.4 1 5Soaking time (hours) 16 144 36Resolution (A) 2.6 2.6 3.2 3.2 3.2Measured 67,901 41,692 25,950 25,637 25,702Unique reflections 14,907 13,376 8,283 9,289 9,684Completion (%) 82 73 81 88 93Rmerge* (%) 7.1 5.1 5.7 5.9 7.6Mean isomorphous

differencet (%) 9.8 10.5 18.3 19.6Phasing power: 2.31 1.77 3.91 1.49Mean figure of merit 0.79

[Se-Met], selenomethionyl-substituted MIF; EMP, ethylmercuryphosphate.*1hij(Ih,i - IhI/2hXiIh,i), where Ih,i is a scaled intensity for the ithobservation of reflection h, and Ih is the mean value.

tllFpH - Fpl/EFp, where FPH and Fp are the scaled derivative andnative structure factor amplitudes, respectively.tPhasing power: root-mean-square heavy-atom structure factor/root-mean-square lack of closure.

rotated into the density for the second and third subunits.Inspection of electron density maps and manual rebuilding ofthese subunits preceded crystallographic refinement. The elec-tron density was well-defined for all residues of the threesubunits in the asymmetric unit. The model was refined bysimulated annealing with an initial temperature of 3000 K andno local symmetry restraints against 2.8-A data using X-PLOR3.1 (16). The model was inspected with 2F.-F, and FO-F, omitmaps, manually rebuilt where appropriate, and submitted fortwo more rounds of refinement (against 2.6-A data) andmanual rebuilding. The final model consists of residues 2-115(Met-1 is posttranslationally cleaved) for each subunit and 24water molecules. The R-factor is 20.0% at 2.6-A resolutionwith a single overall temperature factor and root-mean-squaredeviations of 0.012 A and 1.80 on bond lengths and angles,

Table 2. Refined heavy atom positions

Refined heavyderivative

[Se-Met]-MIF

EMP

KAuCl4

K2PtCl4

RelativeSite occupancy

1 1.002 0.683 0.924 0.995 0.896 0.957 0.978 0.719 0.631 0.502 0.753 0.694 1.005 0.856 0.641 0.992 0.983 0.971 1.182 0.593 0.914 0.425 0.676 0.40

x

0.4670.8370.4250.8560.2540.5010.4050.3260.4840.3250.7270.6760.2440.4510.7120.4610.3930.4810.5220.4910.8390.4670.3030.384

y

0.8510.4320.0130.3520.8190.9020.7910.7800.9680.8690.5640.2960.7900.0310.3240.8360.7930.9030.0010.9180.4850.8340.7180.785

z

0.0140.0380.0410.0760.0730.0860.1000.0960.0810.1570.1130.1060.0890.0570.0610.0110.0850.0850.1260.0880.0730.0020.1130.092

[Se-Met]-MIF, selenomethionyl-substituted MIF; EMP, ethylmer-cury phosphate.

5192 Immunology: Sun et al.

Dow

nloa

ded

by g

uest

on

Oct

ober

29,

202

0

Proc. Natl. Acad. Sci. USA 93 (1996) 5193

FIG. 1. Section of the electron density map superimposed on the final refined MIF coordinates. (A) Initial MIR map calculated at 2.8-Aresolution showing part of the ,3-sheet. The four ,8-strands are from a single subunit. The carbonyl oxygens appear as bumps in the backbone density.The quality of the map displayed here is typical of the rest of the map, which has no breaks along the polypeptide and accounts for all residuesin the mature protein. (B) Final 2Fo-Fc map calculated from experimental data and calculated phases at 2.6A resolution. The maps in are contouredat 1.5 a above the mean value of the electron density.

respectively. The correctness of the model also is supported bythe chemically and sterically reasonable environments of theheavy-atom positions. The gold sites occupy positions nearHis-63 and Met-102 of each subunit. Three of the six platinumatoms share the same site, while the other three are found nearMet-3. Three mercury atoms are complexed to Met-48 andCys-57, and the other three mercury atoms in the ethylmercuryphosphate derivative are found near Cys-81 from each subunit.A Ramachandran plot indicates that there is no violation ofaccepted backbone torsion angles.

Searches, Surface Areas, and Electrostatic Potential Cal-culations. The FASTA and BLASTA programs from the GeneticsComputer Group (Madison, WI) software package were usedfor sequence homology searches. To determine whether thereis any sequence relationship between the two halves of the MIFsequence, the PILEUP program from the Genetics ComputerGroup (gap weight of 3.0 and gap length weight of 0.1) wasused. DALI was used to search the structural database forhomologous protein structures (17). Accessible surface areas(18) were calculated within INSIGHTII (Biosym Technologies,San Diego) using a probe radius of 1.4 A (19). To determinethe surface area buried on trimerization, the accessible surfacearea of the MIF trimer was subtracted from the sum of thesurface areas of each monomer. An electrostatic potential mapwas calculated with MAPROP (from G. J. Kleywegt and T. A.Jones, Uppsala, Sweden). Lys, Arg, and His side chains weregiven a charge of + 1, and Glu and Asp side chains were givena charge of -1.

RESULTS AND DISCUSSIONStructure Description. MIF forms a trimer with dimensions

of 35 A x 50 A x 50 A. The trimeric arrangement of MIF isconsistent with the prediction from a preliminary crystallo-graphic investigation (44) but differs from other reports on thesubunit structure of MIF based on gel-filtration or analyticalultracentrifugation experiments that indicate MIF is either a

monomer or dimer (3, 9, 20). The MIF trimer is an a/13structure with six a-helices surrounding three 13-sheets thatcompletely wrap around to form a barrel with open endsforming a solvent channel (Fig. 2). The trimeric arrangementis due to an intertwining of loops containing short 13-strandsamong the subunits. Although the physiologically relevantsubunit structure for MIF remains to be determined, thesubunit interactions observed in the crystal are based on theinsertion of residues from one subunit into the other (asopposed to packing due to mass action of distinct, globularsubunits) and suggest that the MIF trimer may also form invivo.The MIF monomer contains two antiparallel a-helices and

six 13-strands, four of which form a mixed 1-sheet within the

FIG. 2. Architecture of the MIF trimer. The view is along themolecular 3-fold axis. Strand and helix assignments are as shown in Fig.3A. The a-helices are represented by cylinders and the 13-strands byarrows indicating the direction of the sequence. The three monomersare colored differently. Figs. 2 and 3C were generated with SETOR(S. V. Evans, University of British Columbia, Vancouver, Canada).

Immunology: Sun et al.

Dow

nloa

ded

by g

uest

on

Oct

ober

29,

202

0

Proc. Natl. Acad. Sci. USA 93 (1996)

same subunit. The four-stranded sheet arises from two 13-a-f3motifs, each forming parallel ,3-strands that join together in anantiparallel arrangement. The topology of the secondary struc-ture of the entire monomer has a 13a131313a13a3 motif. Secondarystructure analyses based on circular dichroism predicted ana/X3 structure, which is in good agreement with the crystalstructure (20, 21). The primary sequence with correspondingsecondary structure assignments is shown in Fig. 3A. Thetertiary structure of the MIF monomer is somewhat reminis-cent of the IL-8 dimer in that two a-helices pack against a1-sheet (22). However, the arrangement and topology of thesecondary structure of MIF are entirely different from that ofIL-8. Like the IL-8 dimer that has 2-fold symmetry, thesecondary and tertiary structures of MIF display pseudo 2-foldsymmetry (Fig. 3 B and C); however, an analysis of the aminoacid sequence of MIF does not reveal any homology betweenthe two halves of the protein.The four-stranded 13-sheet of each monomer is extended by

two very short 13-strands contributed by adjacent subunits. Thetwo outer strands, 132 and 135, of the four-stranded 13-sheet ofa monomer are formed by residues 38-44 and 96-102, respec-tively. Strand 13 (residues 48-50) of an adjacent subunitinteracts with 12. As might be expected for such a short strand,only two hydrogen bonds are made between backbone atoms.An additional hydrogen bonding interaction between mono-mers involves the Ns2 atom of His-40 and OEl of Gln-45.Residues Thr-23, Gln-35, Glu-50, and Ser-51 also contribute tothe subunit interface in this region. At the other end of the13-sheet, residues 96-102 (13s) interact with the short strand 136(residues 107-109) of the other subunit leading to the appear-ance of a six-stranded 3-sheet per monomer (Fig. 4). Thetrimer is also stabilized by subunit-subunit interactions be-tween one of the helices and the C terminus: Asn-73, Arg-74,Ser-77, Lys-78, and Cys-81 form a hydrophilic surface of thehelix that interacts with Asn-111, Ser-112, and Thr-113 of anadjacent subunit. The surface area that is buried on formationof the trimer produced by all of these interactions is -1300 A2per monomer.

A 10 20 30 40 50 6

A-I I A-04 A-02 C-n3

FIG. 4. Hydrogen bonding pattern among main-chain atoms of thefour-stranded 13-sheet and the two small ,B-strands contributed byadjacent MIF subunits. Each /3-strand is clearly marked. The threesubunits are indicated by A, B, and C.

The hydrophobic core of the MIF monomer consists ofinteractions between the 13-sheet and the amphipathic a-heli-ces. A striking feature of the hydrophobic surfaces of thehelices is that they contain most of the leucine residues of theprotein (8 of 11). However, these leucines are not arranged ina typical leucine zipper motif. The hydrophobic side chains ofhelix A (residues 19-31) generally extend toward helix B butadopt X angles that lead to interactions with 13-sheet residues.The hydrophobic surface of helix B (residues 70-89) faces the

80

02 03 f4

90 100 1 10

P 06

B

FIG. 3. Primary sequence, secondary structure, and pseudo 2-fold symmetry of the MIF molflmer. (A) Amino acid sequence and secondarystructure assignment of MIF. Regions of secondary structure were based on dihedral angles and hydrogen bonding patterns of backbone atoms.Arrows beneath the sequence indicate /3-strands and rectangles represent a-helices. Each secondary structure unit is labeled. (B) Topology diagramof the secondary structure of the MIF monomer. a-Helices are shown as rectangles and /3-sheets as arrows. The N and C termini are indicated.The topology is pseudo 2-fold symmetric about the point indicated by the ellipse. (C) Ribbon structure of the MIF monomer. The pseudo 2-foldaxis is perpendicular to the plane of the figure and passes between the two internal /3-strands.

5194 Immunology: Sun et al.

MPMFIVNTNVPRASVPDGFLSELTQQLAQATGKPPQYIAVHVVPDQLMAFGGSSEPCALCSLHSIGKIGGAQNRSYSKLLCGLLAERLRISPDRVYINYYDMNAASVGWNNSTFA:::- ==,J>- -=> C=>

Dow

nloa

ded

by g

uest

on

Oct

ober

29,

202

0

Proc. Natl. Acad. Sci. USA 93 (1996) 5195

(3-sheet. Although the subunit interface is not as hydrophobicas the interior of the monomer, a cluster of aromatic residues(Tyr-37, Phe-50, Tyr-96, Trp-109, and Phe-114) is present andforms a hydrophobic patch on the surface.The MIF polypeptide sequence has three cysteines, two of

which (Cys-57 and Cys-60) have been reported to be involvedin a single disulfide (23). The present structure indicates thatall of the cysteines are present as free thiols. The distancesbetween sulfur atoms among cysteines within the trimer rangefrom -7.5 A to 33 A.The Channel. The barrel of MIF has a solvent channel that

runs through the center of the protein and that is coincidentwith the molecular 3-fold axis (Fig. SA). The channel has anodd shape that can best be described as two large funnels thatempty into the center of the protein (Fig. SB). The diameterof the channel ranges from 3-4 A at its most narrow point toabout 15 A at the open ends of the funnels. The neck of thesmaller funnel is defined by three equivalent tyrosine residues(Tyr-100), whereas the neck of the second funnel, which has adiameter of about S A, is formed by Val-43 and Gln-46. Thechannel is lined predominantly by hydrophilic atoms contrib-uted by 24 residues from each monomer. Fourteen of theseresidues protrude into the channel from }3-strand positions.Unaccounted electron density is present in the barrel, partic-ularly around Asn-7, His-41, and Gln-46, which may be due topartially ordered water molecules or sulfate ions. One end ofthe channel is lined by three negative charges from Asp-101 ofeach subunit, and the other end of the channel is surroundedby charges from Arg-12 and Asp-45 from each subunit. Cal-

FIG. 5. Solvent-accessible channel of MIF. (A) CPK model of theMIF trimer showing the channel along the 3-fold axis. (B) Side viewof the solvent-accessible surface (18) of MIF. The surface wascalculated with a probe radius of 1.4 A (19) using INSIGHTII software(Biosym Technologies). The atoms and surface area are coloredaccording to atom type: carbon, green; nitrogen, blue; oxygen, red;sulfur, yellow.

culation of a qualitative electrostatic potential map reveals anasymmetric distribution of the field. The most prominentfeature is that the solvent channel belongs to a region ofpositive potential, indicating that this channel may interactwith negatively charged moieties. It is tempting to speculatethat this unique solvent channel and the positive electrostaticfield also serve an important functional role.There are many protein structures that contain cavities and

short channels, but few proteins have channels that span theentire length of the molecule (24, 25). The most obvious rolefor channels in proteins is to provide a pathway for themovement of ions or molecules through the protein. Amongproteins that have structurally well-characterized channels aregroEL, the (3-subunit of DNA polymerase III, porin, malto-porin, toxin--y, and 14-3-3 protein. In groEL, the large centralcavity with a diameter of about 47 A and a length of 146 Abinds nonnative folding intermediates for subsequent ATP-dependent polypeptide chain folding (26). In the DNA poly-merase III (3-subunit, the ring structure containing a hole witha diameter of 35 A encircles DNA and allows long stretches ofDNA to be replicated without dissociation of the polymerase(27). Porin and maltoporin are transmembrane proteins inwhich a solvent channel facilitates the diffusion of ligandsacross the membrane (28, 29). For toxin- y, the functional rolefor the narrow, 5-7 A channel (based on van der Waal atomicspheres) is more speculative, but it is proposed to be involvedin the passage of small ions through the membrane (30). The14-3-3 protein dimer contains a negatively charged channelthat is 20-A deep and is proposed to be the binding site forprotein kinases (31). MIF has not been reported to haveactivities that involve the movement of molecules through theprotein nor has a receptor been identified. However, recom-binant MIF does bind small molecules such as glutathione(refs. 9 and 20; M. Swope, H.-W.S., and E.L., unpublishedobservations). Before the cloning and characterization of MIF,a protein described as MIF was reported to bind gangliosidesand carbohydrates (32) but this property has yet to be verifiedwith recombinant protein. MIF also catalyzes the tautomer-ization of D-dopachrome to 5,6-dihydroxyindole-2-carboxylicacid (10). Whether any of these molecules bind within ortraverse through the solvent channel remains to be determinedbut it is interesting to note that each possesses negativelycharged groups that could be stabilized within the positiveelectrostatic potential of the solvent channel. The physiolog-ical role of these ligands in the biological activity of MIF isunknown, but the present data will invite further investigationinto the structural interactions of MIF with various smallmolecules.

Relationship to Other Proteins. A search for structuralhomology using the algorithm implemented in DALI (17) didnot identify any proteins that are similar to the monomer ortrimer of MIF. A sequence homology search revealed thatD-dopachrome tautomerase, an enzyme of 118 amino acidsthat catalyzes a similar reaction as MIF, has 27% identity withhuman MIF (10, 33). Display of the identical residues on thestructure of MIF reveals that these residues are distributedthroughout the structure with three notable exceptions. Noidentical residues are found on (i) (3-sheet positions that pointinto the solvent channel, (ii) strand (3, or (iii) strand (6. Thelatter two regions are involved in subunit-subunit interactions.The absence of identical residues within the solvent channelsuggests that the channel is not a likely binding site for thesubstrate D-dopachrome.The report of a GST activity for MIF prompted further

analysis of the relationship between MIF and GST. Thisshowed that: (i) MIF shares 25-35% sequence identity withinthe first 30 amino acids of GSTs; (ii) a hydroxyl-containingactive site residue of GSTs (serine for 0-class GST and tyrosinefor a-, ir-, and ,u-class GSTs) is replaced by a threonine at theequivalent position of MIF; and (iii) polyclonal antibodies

Immunology: Sun et aL

Dow

nloa

ded

by g

uest

on

Oct

ober

29,

202

0

Proc. Natl. Acad. Sci. USA 93 (1996)

raised to a liver MIF preparation cross-react with 0-class GST(9, 34). Nevertheless, the hypothesis that MIF is related to GSTis controversial and has been challenged by the lack ofsignificant sequence homology along the entire polypeptide ofMIF, the large difference in size between MIF (13 kDa) andGST (26 kDa) (35), and the absence of GST activity in purifiedand otherwise biologically active recombinant MIF (21, 36).Members of the GST family of enzymes are dimers of

identical subunits with two distinct domains (37). The struc-tures of proteins from the a, ,u, wr, and oa class of GST have beendetermined (38-41). The MIF structure reported here indi-cates that there is no structural similarity in the overallstructures of MIF and GST. The structures corresponding tothe N-terminal regions have a qualitatively similar secondaryand topological structure consisting of a (3-strand/loop/a-helixmotif. The loop in GST contains residues that interact withglutathione and is much shorter than the corresponding loopof MIF. Glutathione binds to a pocket adjacent to this loop andis surrounded by protein atoms, many of which are contributedby residues from the first 26 amino acids (39, 40, 42). In MIF,the corresponding loops (from each subunit) are found at oneend of the trimer and are fully accessible to the solvent. Ifglutathione were to interact with this loop, it would not be ableto form the constellation of interactions with other proteinatoms as occurs in the case of GST. We therefore predict thatglutathione does not bind at this location in MIF.

Concluding Remarks. The present work elucidates thestructure of MIF, a unique mediator with the ability to bindsmall molecules and catalyze chemical reactions. The structurereveals several atypical features for a cytokine. First, thetrimeric arrangement of MIF produces a novel a/03-barrel fold.The barrel that is formed is not a typical (-barrel that arisesfrom a single /3-sheet but is a result of three 13-sheets that wrapcompletely around (43). Second, these three (3-sheets arearranged to form a solvent-accessible channel that runsthrough the middle of the protein. The channel may be thepotential binding site for small molecule ligands such asglutathione, gangliosides, or dopachrome. The biological func-tion of ligand binding and the role of this channel remain to bedetermined. The three-dimensional structure now makes pos-sible the design and analysis of mutants to test the role of thechannel and other structural elements in the unique biologicalactivities of MIF.

We thank Gregory A. Petsko for helpful suggestions and AnthonyCerami and Robert F. Tilton for critical reading of the manuscript. Weare grateful to the Center for Structural Biology for use of their x-rayfacilities. This work was supported by grants from Connecticut Inno-vations (to E.L.) and the National Institutes of Health Grant A135931(R.B.).

1. Bernhagen, J., Calandra, T., Mitchell, R. A., Martin, S. B.,Tracey, K. J., Voelter, W., Manogue, K. R., Cerami, A. & Bucala,R. (1993) Nature (London) 365, 756-759.

2. Calandra, T., Bernhagen, J., Mitchell, R. A. & Bucala, R. (1994)J. Exp. Med 179, 1895-1902.

3. Galat, A., Riviere, S., Bouet, F. & Menez, A. (1994) Eur. J.Biochem. 224, 417-421.

4. Wistow, G., Shaughness, M., Lee, D., Hodin, J. & Zelenka, P.(1993) Proc. Natl. Acad. Sci. USA 90, 1272-1275.

5. David, J. R. (1966) Proc. Natl. Acad. Sci. USA 56, 72-77.6. Bloom, B. R. & Bennett, B. (1966) Science 153, 80-82.7. Nishino, T., Bernhagen, J., Shliki, H., Calandra, T., Dohl, K. &

Bucala, R. (1995) Mol. Med. 1, 781-788.8. Calandra, T., Bernhagen, J., Metz, C. N., Spiegel, L., Bacher, M.,

Donnelly, T., Cerami, A. & Bucala, R. (1995) Nature (London)377, 68-71.

9. Blocki, F., Schlievert, P. & Wackett, L. (1992) Nature (London)360, 269-270.

10. Rosengren, E., Bucala, R., Arnan, P., Jacobsson, L., Odh, G.,Metz, C. N. & Rorsman, H. (1996) Mol. Med. 2, 143-149.

11. Sambrook, J., Fritsch, E. F. & Maniatis, T. (1989) MolecularCloning: A Laboratory Manual (Cold Spring Harbor LaboratoryPress, Plainview, NY), 2nd Ed.

12. Suzuki, M., Murata, E. & Tanaka, I. (1994) J. Mol. Biol. 235,1141-1143.

13. Kabsch, W. (1993) J. Appl. Crystallogr. 26, 795-800.14. Sheldrick, G. M. (1990) Acta Crystallogr. A 46, 467-473.15. Jones, T. A., Zou, J. Y., Cowan, S. W. & Kjeldgaard, M. (1991)

Acta Crystallogr. A 47, 110-119.16. Brunger, A. T. (1992) XPLOR Manual (The Howard Hughes

Medical Institute and Department of Biophysics and Biochem-istry, Yale University, New Haven, CT), version 3.1.

17. Holm, L. & Sander, C. (1993) J. Mol. Biol. 233, 123-138.18. Connolly, M. L. (1983) Science 221, 709-813.19. Richards, F. M. (1974) J. Mol. Biol. 82, 1-14.20. Nishihara, J., Kuriyama, T., Sakai, M., Nishi, S., Ohki, S.-Y. &

Hikichi, K. (1995) Biochim. Biophys. Acta 1247, 159-162.21. Bernhagen, J., Mitchell, R. A., Calandra, T., Voelter, W., Ce-

rami, A. & Bucala, R. (1994) Biochemistry 33, 14144-14155.22. Baldwin, E. T., Weber, I. T., St. Charles, R., Xuan, J.-C., Apella,

E., Yamanda, M., Matsushima, K., Edwards, B. F. P., Clore,G. M., Gronenborn, A. M. & Wlodower, A. (1991) Proc. Natl.Acad. Sci. USA 88, 502-506.

23. Bernhagen, J., Kapurniotu, A., Stoeva, S., Voelter, W. & Bucala,R. (1995) in Peptides 1994, ed. Maia, H. L. S. (ESCOM, Leiden),pp. 572-573.

24. Hubbard, S., Gross, K. & Argos, P. (1994) Protein Eng. 7,613-626.

25. Kisljuk, O., Kachalova, G. & Lanina, N. (1994) J. Mol. Graphics196, 305-307.

26. Braig, K., Otwinowski, Z., Hegde, R., Boisvert, D. C.,Joachimiak, A., Horwich, A. L. & Sigler, P. B. (1994) Nature(London) 371, 578-586.

27. Kong, X. P., Onrust, R., O'Donnell, M. & Kuriyan, J. (1992) Cell69, 425-437.

28. Schirmer, T., Keller, T. A., Wang, Y.-F. & Rosenbusch, J. P.(1995) Science 267, 512-514.

29. Weiss, M. S. & Schulz, G. E. (1992) J. Mol. Biol. 227, 493-509.30. Bilwes, A., Rees, B., Moras, D., Menez, R. & Menez, A. (1994)

J. Mol. Biol. 239, 1122-1136.31. Xiao, B., Smerdon, S. J., Jones, D. H., Dodson, G. G., Soneji, Y.,

Altken, A. & Gamblin, S. J. (1995) Nature (London) 376, 188-191.

32. Liu, D. Y., David, J. R. & Remold, H. G. (1982) Nature (London)296, 78-80.

33. Zhang, M., Aman, P., Grubb, A., Panagiopoulos, I., Hindemith,A., Rosengren, E. & Rorsman, H. (1995) FEBS Lett. 373,203-206.

34. Blocki, F. A., Ellis, L. B. & Wackett, L. P. (1993) Protein Sci. 2,2095-2102.

35. Pearson, W. R. (1994) Protein Sci. 3, 525-527.36. Sakai, M., Nishihara, J., Hibiya, Y., Koyama, Y. & Nishi, S. (1994)

Biochem. Mol. Biol. Intl. 33, 439-446.37. Gulick, A. M. & Fahl, W. E. (1995) Pharmacol. Ther. 66,237-257.38. Sinning, I., Kleywegt, G. J., Cowan, S. W., Reinemer, P., Dirr,

H. W., Huber, R., Gilliland, G. L., Armstrong, R. N., Ji, X.,Board, P. G., Olin, B., Mannervik, B. & Jones, T. A. (1993) J.Mol. Biol. 232, 192-212.

39. Ji, X., Zhang, P., Armstrong, R. N. & Gilliland, G. L. (1992)Biochemistry 31, 10169-84.

40. Reinemer, P., Dirr, H. W., Ladenstein, R., Schaffer, J., Gallay, 0.& Huber, R. (1991) EMBO J. 10, 1997-2005.

41. Ji, X., von Rosenvinge, E. C., Johnson, W. W., Tomarev, S. I.,Piatigorsky, J., Armstrong, R. N. & Gilliland, G. L. (1995) Bio-chemistry 34, 5317-28.

42. Reinemer, P., Dirr, H. W., Ladenstein, R. & Huber, R. (1992) J.Mol. Biol. 277, 214-226.

43. Lasters, I., Wodak, S. J., Alard, P. & van Cutsem, E. (1988) Proc.Natl. Acad. Sci. USA 85, 3338-3342.

44. Sun, H. W., Swope, M., Cinquina, C., Bedarkar, S., Bernhagen,J., Bucala, R. & Lolis, E. (1996) Protein Eng., in press.

5196 Immunology: Sun et al.

Dow

nloa

ded

by g

uest

on

Oct

ober

29,

202

0