8/12/2019 bi00145a004

1/14

6884 Biochemistry 1992, 31, 6884-6897

Art i d e sInduced Peptide Conformations in Different Antibody Complexes: MolecularModeling of the Three-Dimensional Structure of Peptide-Antibody ComplexesUsing NMR-Derived Distance Restraintst

Tali Sche rf, Reube n Hiller,$ Fred N aider, Michael Levitt, and Jaco b Anglister 'Department of Structural Biology, The Weizmann Institute of Science, Rehovot 76100 , Israel

Received December 28, 1991; Revised Manuscript Received Ma y 1 1992

ABSTRACT: Intram olecula r interactions in bound cholera toxin peptide (CT P3) in three antibody complexeswere studied by two-dimensional transferre d N O E spectroscopy. These measurements tog ether withpreviously recorded spectra th at show intermolecu lar interactions in these complexes were used to obtainrestraints on interproton distances in two of these complexes (TE32 and TE3 3). The NM R-derived distancerestraints were used to dock the peptide into calculated models for the three-dimensional structu re of theantibody combining site. It was found that T E3 2 and TE 33 recognize a loop comprising the sequenceVP GS QH ID an d a P- turn formed by the sequence VPG S. Th e third antibody, TE34, recognizes a dif ferentepitope within the same peptide and a P-turn formed by the sequence ID SQ . Neither of these two turnswas observed in th e free peptide. The formation of a P-turn in the bound peptide gives a comp act conformationthat maximizes the contact with the antibody and th at has greater conformational f reedom than a-helixor P-sheet secondary structu re. A total of 15 antibody residues ar e involved in peptide contacts in the T E3 3complex, and 73% of th e contact are a in the antibody co mbining site consists of the side chains of aroma ticamino acids. A comparison of the NMR-derived models for CT P3 interacting with T E3 2 and TE3 3 withthe previously derived m odel for TE 34 reveals a relationship between amin o acid sequence and combiningsite structure and function. (a) T he three aromatic residues th at interact with the peptide in TE32 andTE33 complexes, Tyr 32L, Tyr 32H, and Trp 50H re invariant in all light chains sharin g at least 65%identity with TE 33 an d T E3 2 and in al l heavy chains shar ing at least 75% identity with TE 33. AlthoughTE34 differs from TE32 and TE33 in its fine specificity, these aromatic residues are conserved in TE34and interact with its antigen. Therefore, we conclude th at th e role of these three aro matic residues is toparticip ate in nonspecific hydrophobic in teractions with th e antigen. (b) Residues 31, 31c, and 31e ofC D R l of the l ight chain interact with the antigen in all three antibodies that we have studied. The aminoacids in these positions in TE 34 differ from those in TE 32 an d TE 33, an d they a re involved in specific polarinteractions with the antigen. (c) CD R3 of the heavy chain varies considerably both in length and insequence between TE 34 and th e two other anti-CT P3 antibodies. These changes modify the shape of thecombining site and the hydrophobic and polar interactions of CD R3 with the peptide antigen.

Interest in the molecular interactions of antibodies withpeptide antigens has been greatly stimu lated by the potentialto use synthetic peptides in the production of vaccines (Ar-non, 1986; Steward & Howard, 1987). Immunization ofanimals with short flexible peptides elicits the prod uction ofa spectrum of antibodies differing in their affinities to thepeptides and the native proteins from which the peptides werederived. Currently there is only limited structural informationabout antibody-peptide complexes and the conformations ofthe bound peptide. Th e three-dimensional stru cture of anantibody complex with a peptideof myohem erythrin was solvedt This work was supported by a grant from the U S- Isr ael BinationalScience Foundation (88-0045 0) to J.A. and by a grant fro m the MinervaFoundation, Munich, Germany, to M.L.* Correspondence should be addressed to this autho r. He is theincumbent of the Gra ham and Rhona Beck Career Development Chair.Levi Eshcol Postdoctoral Fellow. Present address: Departm ent ofBiochemistry and Biophysics, 338 Anatomy Chemistry Building, Uni-versity of Pennsylvania, Ph iladelphia, P A 19 104-6059.Fulbright Exchange Scholar. Present address: Th e Collegeof Staten1sland.TheCityUniversityofNew York,S t. GeorgeCampus, 130Stuyve-sant Place, Staten Island, NY 10301.

recently by Stanfield et al. (1990), and N M R 1 spectroscopywas used to derive a model for an antibody complex with acholera toxin peptide (Zilber et al., 1990).To increase our understanding of the molecular basis forthe cross-reactivity of anti-peptide antibodies with nativeAbbreviations: CD R, complemen tarity determining region of theantibody molecule; CO SY , 2D J-correlated spectroscopy; CTP 3, cholera

toxin peptide 3; DQF-C OS Y, double-quantum-filtered COSY; EL ISA ,enzyme-linked immunosorbent assay; ESR , electron spin resonance; Fab ,antibody fragment made of the Fv, the light chain, and the first heavychain constant regions; Fv, antibody fragment ma deof th evariable regionsof the light and heavy chains forming a single-combining site for antigen;HL T, heat-labile toxin; NM R, nuclear magnetic resonance; NO E, nuclearOverhauser effect; NOESY, 2D NOE spectroscopy; rms, root meansquare; TFE , trifluoroethanol; TR NO E, transferred NO E; VH, ariableregion of the heavy chain; 2D, two dimensional; 2D TR N O E differencespectrum, calculated 2D difference spectrum between the measuredNO ES Y spectrum of the protein saturated with the ligand and tha t ofthe protein in the presence of a large excess of ligand. Nam e convention:(a) specific peptide residue, lower-case letters and the position in thesequence (e .g., 8111); (b) sp ecific antibod y residue, capit al and low er-case letters and the position in th e sequence (e.g., Tyr 32L); H, heavychain residues; L, light chain residues. Thenu mbering of antibod y residuesis according to Chothia and Lesk (1987).0006-2960/92/043 1-6884$03.00/0 1992 American Chemical Society

8/12/2019 bi00145a004

2/14

Peptide-Antibody Complexesproteins, it is necessary to study the conform ation of the pep-tide when it is free in solution, when it is part of the nativeprotein, and when it is bound to different antibodies. Suchstudies should address the following questions: (1) H ow doesthe conformation of the bound peptide compare to itsconformation in solution and to that of the correspondingepitope in the native protein? (2) W hat a re the implicationsof these conformational comparisons on the efficacy of thevaccine and on the mechanism by which anti-peptide antibodiesare elicited? (3) W hat is the size of the antigenic determinantrecognized by different antibodies, and what are the inter-actions between the antibodies and the peptide antigen? (4)How do peptide-antibody interactions compar e with protein-antibody interactions? Considerable knowledge on the lattertopic has been gained from recent crystallographic studies(Amit et al., 1 986; Sheriff e t al., 1987; Colma n et al., 1987;Padlan et al., 1989).To address these issues, we ar e studying the interactions ofmonoclonal antibodies with the peptide ( CT P3) correspond-ing in sequence to residues 50-64 (V EVP GSQ HIDS QKK A)of the B-subunit of cholera toxin and the heat-labile toxin ofEscherichia coli . In vitro, seru m of animals imm unized withCT P3 partially neutralizes th e biological activity of both thecholera toxin and t he heat-labile tox in of E . coli (Jacob et al.,1983, 1984). Thr ee anti-CTP3 antibodies, TE3 2 and TE33which cross-react with cholera toxin in a solid-phase immu-nosorbent assay (EL ISA ) and TE3 4 which does not bind thetoxin at all, were selected for detailed N M R investigation(Anglister et al., 1988).

In our studies we use the Fab fragment of the antibody.This fragment has a molecular weight of 50 000 and is too bigfor a complete struc ture determination by N M R . However,a knowledge of the three-dimensional struc ture of the wholecomplex may not be necessary to address the above issuesconcerning antibody-antigen interactions and the conforma-tion of the bound antigen. We previously used two-dimensionalTR N O E difference spectroscopy to study the interactions ofthe three antibodies (TE32, TE33, and TE34) with CTP3(Anglister e t al., 1 989; Levy et al., 19 89; Anglister & Zilber,1990). He re we present a T R N O E study of intramolecularinteractions in bound C TP3 in the three antibody complexesand a study of the free peptide conformation. N M R dat a onintermolecular and intrapeptide interactions are translatedinto N M R restraints on interproton distances which are thenused to calculate models for the peptide complexed with TE32and TE33. As a template for docking the peptide, we use acalculated model for the antibody Fv that relies on the three-dimensional structu re of a highly homologous anti-progest-erone antibody (Deverson et al., 1987; Stu ra et al., 1987). Theconformation of the bound peptide in TE32 and TE33complexes is compa red to its conformation in three o ther forms:(a) in complex with the TE34 antibody (Zilber et al., 1990),(b) free in solution, and (c) as a segment of HL T ( Sixma etal., 1991). The amino acid sequences of TE3 2 and TE33 arecompared to those of highly homologous antibodies withdifferent specificities. These comparison s provide insight onthe molecular basis for antibody specificity.MATERIALS AND M ETHODS

Materials. Cholera toxin was purchased from Sigma. Th efollowing deuterated amino acids were obtained from MS Disotopes: ~-tryptophan-2,4,5,6,7-d596.9% D ), L -Chydroxy-phenyl-dd-alanine-2 3 3-d3tyrosine) (96.9% D ), L-Chydroxy-pheny1-2 6-d2-alanine-2-d nd ~-phenyl-ds-alanine-3 3-d2(98% D). A partially deu terated tryptopha n, L-tryptophan-

Biochemistry, V ol. 31, No. 30, 1992 68854,6,7-d3 (60% H in positions 2 and 5 and 90% D in positions4, 6, and 7), was a generous gift from Dr. Mei Whittaker,Stan ford University. In all measurements, F ab fragments ofthe antibodies were used. Antibody labeling, Fa b preparation,and peptide synthesis and purification were as previouslydescribed (A nglister et al., 1989; Anglister & Zilber, 1990).

Synthesis of a Spin-Labeled Peptide TE MP O-VEVPG-SQHIDSQ) . 1-0xy-2 2 6 6-tetramethyl-4-isothiocyanatopip-eridine (ITC -TE MP O) was prepared according to Gaffney(1976). The peptide VEVP GSQ HID SQ (430 mg) wasdissolved in a minimal volume of water (5 mL ), and th e pHwas adjusted to 9.6 by addition of triethylamine (approxim ately100 mL). The peptide solution was added slowly to avigorously stirred solution of I TC -T EM PO (300 mg) in di-methylformamide (2 mL). The pH of the reaction mixturewas maintained at 9.6 by the addition of small aliquots oftriethylamine. Afte r addition of the peptide, the reactionmixture was stirred at room temperature for 24 h. ExcessITC-TEMPO was precipitated by the addition of 70 mL ofwater and removed by centrifugation. The supernatant wasconcentrated by partial evaporation and then loaded on aSephadex G-15 column equilibrated with ammonium car-bonate (10 mM ). The peptide was further purified by HP LCusing a gradient of 2-propanol in an aqueous solution of 10mM am monium carbonate. A purity of at least 98% wasobtained. Th e peptide composition and sequ ence were verifiedby amino acid analysis and the N M R spe ctra of the spin-labeled peptide before an d af ter reduction with sodium dithio-nite.

Measurements of Binding Constants by FluorescenceQuenching. The binding constant of TE3 3 for the spin-labeledpeptide was measured as previously described by followingthe quenching of the Fa b fluorescence upon peptide binding(Anglister et al., 1988). The F ab fluorescence was measureda t room temperature on a Perkin-Elmer M PF -44E fluores-cence spectrophotometer. The excitation wavelength was 283nm with a slit bandwidth of 2 nm, and the emission wasmeasured at 363 nm with a slit bandwidth of 15 nm. Fabconcentration was 10-6 M. The F ab fluorescence wasmeasured as a function of the added concentration of thespin-labeled peptide. The required correction for the non-specific absorption of the spin-labeled peptide at th e absorptionand emission wavelengths of Fab fluorescence was found bymeasuring the decrease in th e fluorescence of the tryptophansolution caused by the addition of the spin-labeled peptide.

Measurement of TE3 3 Binding to C holera Toxin by ESR.ES R spectra were measured on a Varian E- 12 spectrometer.Acalibrationcurve for the peak height versus theconcentrationof spin-labeled peptide was obtained. Spe ctra of 25 pM spin-labeled peptide in the presence of 25 pM TE 33 F ab withoutand with 143 pM cholera toxin were measured (Figure 1,spectra A and B, respectively). Due to a much shortercorrection time the free peptide has a sharp resonance separatedfrom the broad signal of the bound peptide. Theconce ntrationof the free peptide was deduced from the peak height of itssignal marked as free in spectrum (Figure 1).

Sample Preparation fo r N M R Measurements. N O E S Yspectra of TE33 F ab in DzO were measured a t 42 OC and pH7.15. Fab concentration was originally 2.8 mM and wasreduced to 2.2 mM afte r the addition of a 4-fold CTP 3 excess.In 90% H 20 / 10%DzO (v/v) solution, TE33 Fab concentrationwas 3.0 mM aft er addition of a 7-fold peptide excess, and thepH was 5.5. Measurements in 90 H z0 / 1 0 % Dz0 werecarried out at 15OC. TE34 F ab (1.7 mM ) in 90% HzO/ 10%D 2 0 solution containing 300 mM Na Cl was measured at 39

8/12/2019 bi00145a004

3/14

6886 Biochemistry, Vol. 31, No. 30, 1992 Scherf et al.A Bree freei

i'V

F I G U R E A) ES R spectrum of 25 p M T E M P O - V E V P G S Q H I D S Qin the presence of an equimolar concentration of TE33 Fab inphosphate-buffered salineat room temperature and p H 7. Thesignalsof the free and bound spin-labeled peptide are ind icated in the figure.(B) The sam e spectrum after the addit ion of 143 pM cholera toxin.OC and pH 4.1 5 , and a 3-fold peptide excess was used. Sp ectr aof 15 mM free CTP 3 in 90% H20 /10 % D2O (v/v) solutionwere measured at pH 3.95 and 4 C. All solutions werebuffered with 10 mM sodium phosphate.N M R Measurements. All spectra were measured on aBruker AM 500 spectrometer using the phase-sensitive mode.The procedure for difference spectra calculations as well asfor NOESY measurements in D2O solutions was describedpreviously (Anglister et al., 1989; Anglister & Zilber, 1990).In the DQF-COSY (Piantini et al., 1982) measurements ofthe F ab in the presence of peptide excess, the H2 0 line waspresatu rated by selective irradiation durin g the preparationperiod before the 90' excitation pulse was applied. For eachof 400 increments of t we collected 16 scans with 8K datapoints in the t2 domain. In NOE SY measurements of H2 0solutions of the Fab-pe ptide complexes, the carr ier frequencywas set on the H2 0 line and the 1-1 ju m pr et ur n pulsesequence was used for selective excitation (P late au & Gueron,1982). The NO ESY spectra (Macu ra & Ernst, 1980) of theFa b in the presence of peptide excess were recorded a t a mixingtime of 100 ms for the TE33 Fab-peptide complex and at 100and 200 ms for the TE34 Fa b-peptid e complex. Typically,64 or 80 scans of 2K d ata points were collected for 400-440t l increments. For the NO ESY and the DQF-CO SYmeasurements of the CTP 3 peptide, 600 t l values were recordedwith 8K dat a points along t 2 ; 32 scans were collected for eacht l value. The mixing time used in the NOE SY measurementof the peptide was 400 ms. In the measurem ents of the pep-tide spectra, the H20 line was supressed by presaturation.RESULTS

TE3 3 Binding to Cholera Toxin. TE3 3 binding to choleratoxin was determined by using ESR spectroscopy to followthe competition between cholera toxin and the spin-labeledversion of CTP3 (TEMPO -VEV PGS QHID SQ). In thesemeasurements, the resonance of the free peptide appears asa narrow signal while that of bound peptide is broadened du eto the slower rotational correlation time of the complex (seeFigure 1). The binding constant of TE33 Fab for the spin-labeled peptide was calculated from m easureme nts of the freepeptide concentration as a function of the tota l concentrationof the spin-labeled peptide added to 25 pM TE33 Fab, givinga value of 0.9 X lo6M-l. The ESR measurements were usedalso to calibrate the peak height of the bound peptide signalversus its concentration. As an alterna tive and to confirm theESR measurements, the binding constant of TE33 Fab forthe spin-labeled peptide was also determined by a fluorescencequenching procedure (see Materials and Methods) and wasfound to be 1.2 X lo6 M-'. This value is identical to that for

b h,h

v I ,eI s11,sa s6,s 1/k14,a

s6,s

I O

d,d h,h d,sll h,i II

1 - , . , . 188 a5 8 2PPmFIGURE: A section of the NOESY spec t rum of f ree CT P3 (VEV PG-S Q H I D S Q K K A ) i n HzO showing interactions of amide protons withother protons in the peptide. Th e spectrum was measured a t 4 OCa n d p H 3.95.CTP 3 (VEVP GSQ HIDS QKK A) (Anglister et al. , 1988) andis in good agreement with the E SR m easurements. Very highconcentrations of the toxin were necessary to obtain measurablechanges in the ratio between th e concentrations of the boundand free spin-labeled peptide. Addition of 0.143 mM choleratoxin (each molecule contains five B-subunits) resulted in a20% increase in the ratio between the free and the boundpeptide (Figure 1B). The measurements were repeated fivetimes, and the average calculated binding constant of the toxinwas found to be (1 f 0.5) X lo3M-', which is 3 orders ofmagnitude weaker than TE33 binding for the spin-labeledpeptide.

Intramolec ular Interactions in Free CTP3. As shown inFigure 2, the NO ES Y spectrum of CT P3 reveals several in-traresidue interactions, all of the strong sequential [ a i)-N(i+l)] interactions, and several very weak interactionsbetween the amide protons of adjacent residues [ N i ) -N( i+ l) ]. Th e multiple amide-amide interactions are confinedto the sequence between gly5 and lys13 of CT P3 and a re tooweak to reflect any significant population of a helicalconforma tion or the presence of reverse turns. Th e peptideresonances were assigned on he basis of the DQ F-CO SY andthe NOES Y spectra.

Intrapeptide Interactions in the Complexes of TE33 andTE32 with CT P 3 . Figure 3A presents a section of the 2DTR N OE difference spectrum showing TR NO E cross-peaksdue to intramolecular interactions between nonaromaticprotons of CTP 3 bound to TE33. Tryptop han and phenyl-alanine residues of the antibody were perdeutera ted, and ty-

8/12/2019 bi00145a004

4/14

Peptide-Antibody Complexes Biochemistry, Vol. 31, No. 30, 1992 6887

44'LL

l

.1.0

2.0 z- 3.0

J o h t14.0

y , , , , , - , , LIc , )b I4.0 3.0 2 0 I o 4 39 29 I.oPPMFIGURE: (A) Difference between the NOESY spectrum of TE33 Fab with a 4-fold excess of CTP3 and that of the TE33 Fab saturatedwith CTP3 showing intramolecular interac tions of aliphatic protons in the bound peptide. Spectra were measured in phosphate-bufferedDzOat 42 OC and pH 7.15 with 100-ms mixing time. Antibody tyrosine Ca protons were deuterated, and tryptophan and phenylalanine residueswere perdeuterated. (B) The corresponding difference spectrum obtained for TE32 under conditions identical to those in (A).

rosine Ca protons were deute rated to minimize spin diffusion.The two strong cross-peaks at 0.96, 3.72 and 0.96, 3.87 ppmar e assigned to interactions of va13 methyl protons with t he6 protons of pro4. T he first of these peaks overlaps a weakercross-peak due to intraresidue interaction in Val1, and thesecond of these peaks overlap s an even weaker cross-pe ak dueto interaction betw een ser6 CpH an d va13 CH3. Th e cross-peaks at 3.84,2.08 and 3.88,2.08 ppm are due to interactionof ser6 l protons withva P CpH. These two cross-peaks overlapa stronger cross-peak due to intraresidue interaction in pro4.These assignments are based on (a) the COSY spectrum offree CTP3 and of a truncated version of CTP3 comprisingresidues 1-10, (b) disappe aran ce of cross-peaks upon deu-teration of va13, (c) c onsiderable dec rease in intensity on pro4deuteration, and (d) the fa ct tha t all these cross-peaks are notaffected by deuteration of glu2 and gln7. The two cross-peaksat 0.96,2.64 and 0.96,2.74 ppm are assigned to interactionsbetween va13 CH 3 and as plo Cp protons. This assignme nt isbased on (a) chemical shifts identical to those of the corre-sponding free peptide resonances, (b) their m ultiplicity, and(c) disappearance on deuteration of va13. Although distantin sequence, the vaP and asplo protons are adjacent in thebound peptide, indicating loop formation. Previous observationof an antibody Trp proton (W) and Tyr C, protons (Yl)interacting with both val3 CH 3 and asplo 0 protons providesadditional evidencefor this loop (Anglister et al., 198 9). Similarinteractions in the bound peptide were observed for the TE32-CT P3 complex (Figure 3B). Some of thecross-peaks in Figure3 are not symmetrical du e to different resolution in the F1 andF dimensions, difficulties in correcting the phases of thecolumns and rows in the difference spec tra, interference of t lnoise, and baseline distortions.Intramolec ular interactions of the amid e protons of the pep-tide bound to TE33 were studied by transferred NOEmeasurements in 90 H20/10% D2O. In order to observethe am ide protons, it is necessary to slow their excha nge withthe hydrogen atoms of water while maintaining an off-ratesufficient for measurement of TR NO E. At pH 4, which isoptimal for slow am ide proton exchange, th e peptide bindingdecreased considerably. To observe amide protons at a higherpH (pH = 5 . 5 , we had to lower the tem pera ture to 15 OCto counterbalance the increase in amide proton exchange dueto the elevated pH . However, the decrease in the tempera-ture concomitantly decreased the off-rate. Successful N M R

measurements at p H 5.5 had to be carriedout using a truncatedligand, VEV PGS QHI D-am ide. In a previous study (Anglis-ter & Zilber, 1990) we found that this truncated peptide,which retains most of the determinant recognized by TE33,has an off-rate which is an order of magnitude faster thanthat of CTP 3. It was found that this truncation of the peptidedid not interfere with the interactions of the remaining pep-tide residues with the antibody (Anglister & Zilber, 1990).The difference between the NO ES Y spectrum of the Fab inthe presence of excess VEVPGSQHID-NH2 and that of theFab in the presence of equimolar peptide concentration wascalculated a nd compared to the N OE SY spectrum of the Fabin the presence of excess peptide. Due to the broad line widthof the protein am ide protons and the large excess of the pep-tide (7-fold), it was not necessary to calculate a differencespectrum to resolve the TRNOE cross-peaks of the peptideamide protons from background cross-peaks contributed bythe protein.Figure 4 presents a section of the NOESY spectrum ofTE3 3 Fab in the presence of excess VEVP GSQH ID-NH2,showing intramolecular interactions of the amide protons ofthe bound peptide observed via TR NO E. A very strong TR-N O E cross-peak due to interaction between the amide protonsof glys and ser6 is observed. This cross-pea k indicates aformation of a &turn by the sequence VPGS. Som e of theinteractions tha t have been shown in Figure 3-Ser6 rotonsintera cting with a va13 l proton and m ethyl protons, a ser6aproton interactin g with va13 methyl protons-support a bendformation. A weaker cross-peak between the amide protonsof ile9 and asplo indicates th at som e bend formation may alsooccur at the C-terminus of the epitope. In Figure 4, a long-range interaction between the am ide proton of asplo and themethyl protons of va13, manifested by a cross-peak a t 8.56,0.94 ppm, furth er confirms the formation of a loop.Intramo lecular Interactions in the Peptide Bound t o TE34.The off-rate of C TP 3 bound to TE 34 was previously foundto be too slow to measure TR N O E (Anglister & Zilber, 1990).Therefore , we used two versions of the peptide which had fastenough off-rates: (a) HID SQK KA -amide , in which sevenN-term inal residues were truncated and th e C-terminal car-boxyl was modified into an amide, and (b) IDSQRKA, inwhich eight N-term inal residues were truncated and lys13wasreplaced by arg to avoid degeneracy in chemical shifts. In thelatte r probe his8, which is in the de termin ant recognized by

8/12/2019 bi00145a004

5/14

6888 Biochemistry, Vol 31, No. 30, 1992 Scherf et al.

d ,d

e e

i _-.. - _ ~ _ ~ _l _ ---- r18 85 8 2

P P r lFIGURE: N OE SY spectrum of TE3 3 Fab with a 7-fold excess ofa truncated version of the CTP3 peptide (VEVPGSQHID-amide)showing intramolecular interactions of the ami de protons in the boundpeptide. Th e measurement was done in phosphate-buffered 90%H 2 0 / 1 0 % D 2 0 a t p H 5 5 and 15 O with 100-ms mixing t ime.Antibody tyrosine Ca protons were deuterated, and tryptoph an andphenylalanine residues were perdeuterated.TE34 (Anglister & Zilber, 1990), was eliminated, but theC-te rm inal carboxyl was not modified. We have previouslyfound tha t these modifications did not inter fere significantlyeither with the interactions of the unmodified p art of the pep-tide with the antibody or with intramolecular interactionsoccurring in the epitope recognized by the antibod y (Anglis-ter & Zilber, 19 90). Th e binding constant of TE34 for itspeptide antigen did not decrease even when the p H was loweredto 4.2. To enhance the off-rate of IDSQ RKA , measurementswith this peptide were carried ou t at 39 OC. Und er theseconditions the am ide exchange was still slow enough. Ac-cording to the previously obtained NMR-derived model forthe TE3 4 complex with CTP 3, residues IDS Q form a fl-turn(Zilber et al., 1990). Such a turn should manifest a strongT RN OE cross-peak between the am ide protons of serll andgln12. However, the 0.05 ppm chemical shift differencebetween these two resonances in the spectrum of HID SQK KA -amide prevents the observation of a cross-peak sufficientlywell resolved from the diagonal. In ID SQR KA, the chemicalshift difference between these two resonances is 0.13 ppm,and cross-peaks should be observed. A section of the two-dimensional TR N OE difference spectrum obtained for TE34Fab in 90% H20 /10 % DzO containing a 3-fold excess of ID-SQR KA and using 200ms mixing time is shown in Figure 5B.Tryptophan, phenylalanine, and tyrosine residues of TE34were perd eute rated to minimize spin diffusion. W e observedstrong interactions between the amide protons of serll and

N-Acetyl ,h-I

I O

2 0

La30

8 6 8 4 8 2PPm

8 0

E-a8 4

8 4 8 2 80 78F I G U R E: (A) N OE SY spec t rum of TE3 4 Fab wi th a 7- fo ld excessof a truncated version of CTP3 (HIDSQKKA-amide) showing in-tramolecular interactions of the amide protons of the bound peptidewith i ts other protons. Th e spectrum was measured in phosphate-b u ff er ed 9 0% H 2 0 / 1 0 % D 2 0 a t 1 5 OC and pH 5.5 with a mixingt ime of 100 ms. Tryp toph an, phenylalanine, and tyrosine residuesof the antibody were perdeuterated. (B) S e c ti o n o f t h e 2 D T R N O Edifference spectrum of TE3 4 using a 3-fold excess of IDS QR KAshowing interactions between amid e protons of the bound peptide.The spectra were measured in phosphate-buffered 90% HzO/ 10%D 2 0 a t 3 9 O and pH 4 .2 with 200-ms mixing t ime. TE 34 wasdeuterated as in A) .

PPm

gln12 and of ar gI3 and lys14 and a m uch weaker interactionbetween the ami de protons of ile9 and a splo . While th e firsttwo interactions ar e strong in a spectrum measured with 100-ms mixing time, th e interacti on between ile9 and asp10 isconsiderably weaker. In th e previously derived model for TE3 4(Zilber et al., 1990) the protons in these three pairs areseparated by distances shorter than 3.1 A. Figure 5A showsa section of the NOESY spectrum of TE34 in the presenceof a 7-fold excess of HIDSQ KK A-am ide. As the off-r ate ofthis peptide is much faster th an th at of ID SQR KA, we observestronger TRNOE cross-peaks due to intramolecular inter-actions in the bound peptide. Several shor t-range interres-idue interactions were observed. Th e amide protons of ile9,asplo, ser , g1n12, and lys14 inter act with the side chains ofhis8, ile9, aspl o, lysI4, and g1n12, respectively.Intra-Fab Interactions in T E3 3 . Information concerningthe interresidue interactions of residues in the antibody com-bining site region provides valuable insights into the struc tur eof the binding domain and helps support the assignment ofantibody-peptide interactions to specific antibody residues.To obtain this information, we subtracted the 2D NOE

8/12/2019 bi00145a004

6/14

Peptide-Antibody Complexes Biochemistry, Vol. 31, No. 30, 1992 6889mental support for the calculated structure.

Spin Diffusion. In order to rule out the possibility thatspin diffusion effects appeared in the 2D T R N O E differencespectra measure d with a mixing time of 100 ms, we carriedout measure ments with different mixing times of 30 ,60, an d100ms. In these measurem ents we used TE33 Fa b in whichtryptophan and phenylalanine residues were perdeuteratedand tyrosine Ca protons were deutera ted. All cross-peaks dueto the interactions of tyrosine C, and histidine imidazole protonsof the antibody were observed at the shorter mixing time,excluding he possibility of spin diffusion. When it was possibleto calculate the volume integral accurately, the change incross-peak intensity as a function of the mixing time ruled outthe possibility of a d elay in the rise of the N O E cross-peaks.Th e increase in intensity of two cross-peaks was proportionalto the increase in m ixing time from 60 to 100 ms. Th e intensityof all other cross-pea ks ncreased to a considerably esser extent,and two of them becam e slightly weaker a t the longer mixingtime.

NMR Distance Restraints. NMR results from previousstudies on intermolecular interactions in the TE32 and T E3 3complexes (Anglister et al., 1988, 1989; Levy et al., 1989)together with those presented here were translated into in-terproton distance restraints using volume integral measure-ments of the NO ES Y cross-peaks. Th e intensity of thestrongest cross-peak [designated W-v3 in Anglister et al.(1989) and Levy et al. (1989)l in the difference spectrumbetween the N OE SY spectrum of the Fab in the presence ofexcess peptide and that of the Fa b saturated w ith the peptidewas compared to the intensity of the exchange cross-peak ofCa2H of the peptide histidine. Since this cross-peak appe arsin all TRNOE difference spectra, and as all spectra weremeasured under the sa me conditions, at the sam e mixing time,and with the sam e molar ra tio of peptide-antibody, it canserve as an internal standard for calibrating the intensities ofall cross-peaks in all T R N O E difference spectra measured inD2O solutions. To evaluate the distance between the peptideva13 methyl protons and the antibody tryptophan proton, weexamined the difference spectrum between the NOESYspectrum of F ab with excess peptide and th at of the Fa b itselfmeasured for T E3 3 Fab in which tyrosine and phenylalaninewere perdeuterated. In addition to the transferred NO E cross-peaks, this difference spectrum contained cross-peaks due tointraresidue interactions of the two tryptophan residues in thecombining site since their chem ical shifts changed upon pep-tide binding. From th e comparison of the intensity of theW-v3 cross-peak with those of the cross-peaks due to intrares-idue interactions between adjacent protons on the tryptophanindole, we concluded that the W-v3 cross-peak correspondsto a distance of about 3 A. Oth er cross-peaks were attributedto distances shorter than 3.0, 3.5, 4.0, and 4.5 A, dependingon their relative intensity. In the absence of stereospecificresonance assignments, heavy atom s that areclose to the centerof the line connecting the two hydrogen atoms were used aspseudoatoms (Wu thrich et al., 1983), and a correction wasadded to the N O E distance constraint. In a few cases ste-reospecific assignments were adde d af ter p reliminary modelswere calculated using restraints on pseudoatoms. Additionalintrapeptide proton-proton distance restraints were obtainedfrom the NO ES Y spectrum of the TE33 Fab-peptide complexin H2 0 (Figure 4). Under the conditions in which thisspectrum was measured, we could observe transferred N O Ecross-peaks only due to intramolecular interactions in thebound peptide. In the absence of a phenyl group in the pep-tide, we had to use a cross-peak due to the intra y N H 2 protons

t

6.0

.7.0

8.0

LW,YI t

YI. Y3,Y4W

80 7 60PPMFIGURE: Difference between the 2D NOE spectrum of the TE33Fab with excess CTP3 and that of the uncomplexed Fab showingintramolecular interactions in the antibody as well as intermolecularinteractions. Antibody phenylalanine residues are perdeutera ted whiletyrosine residues are deuterated at the Ca positions. Tryptophanresidues have 40% deuterium a t Cn and Cal and 90 deuterium atthe other indole positions. (A) Negative cross-peakscontributed bythe TE33-CTP3 complex. B) Positive cross-peaks contributed bythe noncomplexed Fab. Interacting antibody residues are markedby capital letters and arbitrary numbers. Spectra were measured inphosphate-buffered D20 at 42 OC and pH 7.15 with 100-msmixingtime.spectrum of the uncomplexed Fa b from that of the Fa b withexcess CTP 3 peptide. Cross-peaks in this difference spectrumarise from intramolecular interactions of antibody protonsthat change chemical shift upon peptide binding, intermo-lecular interactions between antibody and peptide, and in-tramolecular interactions in the bound peptide. To obtain adifference spectrum free of the nu merous cross-peaks due tointraresidue interactions, tryptophan and tyrosine residueswere partially deutera ted while phenylalanine residues werecompletely deuterated. Comparison with a spectrumobtainedfor TE3 3 under similar conditions, in which tryptophan andphenylalanine residues were perdeuterated and tyrosineresidues deuterated at the Cal and Ca2 positions, enabled usto assign cross-peaks to their specific amino acid types.Inspection of both positive and n egative cross-peaks allowsus to study interactions occurring either in the Fa b or in theTE33-C TP3 complex , respectively, (panels B and A of Figure6). These spectra show strong interaction of Y 1 with H 1, H2,and W . W e previously found that these protons interact withthe peptide; Y1 corresponds to Tyr 32L, H 1 and H 2 to H is31L, and W to CalH of Trp lOOaH (Levy et al., 1989).Additional interactions between the C, protons of two ty-rosine residues (Y3 and Y4) are also observed (Figure 6B).Neither Y3 nor Y4 inte racts with CTP 3 as judged by a two-dimensional TR N O E difference spectrum showing TE33 ty-rosine residue interactions with C TP 3 (Ang lister et al., 1989).Nevertheless, since their chemical shift changes upon CT P3binding, they a re assumed to be in the com bining site region.Specific chain-labeling experiments (Levy et al., 1989) andinspection of the calculated model suggest that Y3 and Y4ar e Tyr lOObH and Tyr 49L, respectively. N M R observationof the interactions between combining site residues which areproximal in the model for th e antibody Fv provides experi-

8/12/2019 bi00145a004

7/14

6890of the glutam ine residue of the peptide as an internal referencefor distance calibration. All the restra ints used in ourcalculations are due to interactions of am ide protons (8.2-8.6ppm) or side-chain protons with chemical shifts between 0.9and 2.1 ppm. In this range of chemical shift the excitationprofile of the 1-1 ju m p re tu rn pulse reaches its maximumand does not vary to a considerable exte nt. Therefore, nocorrection was applied to the measured cross-peak intensity.The lists of the intermolecular distance restraints as well asthe intra-bound-pep tide distances used in the modeling of theTE33-CTP3 and of the TE32-CTP3 complexes are given inTables I and 11, respectively. All cross-peaks were assignedbefore calculation of the model for the complex, and theassignments did not require any correction after preliminarycalculations. We deriveddistance restraints for all the assignedcross-peaks, and all of them were used in the calculations.

CalculationofAntibody Model s. Models for the antibodiesTE32 and TE33 were built as described before with twoimprovements. (a) T he atomic coordinates of the antibodyDB3, which is highly homologous to these antibodies (Levyet al., 1 989 ), were used to provide initial coordinates for allthe main-chain atom s and th e side-chain atoms of homologousresidues. (b) A new method was used for the missing side-chain atoms, which are taken from any of the other eightantibody structures that have known X-ray structures in theBrookhaven Protein Data Bank: the six Fabs, KOL ( Ma r-quart et al., 1980), NEW (Saul et al., 1978), McPC 603(Satow et al., 1987), 5539 (Suh et al., 1986), HyHel-5 (Sheriffet al., 1987), and H yHel-10 (Padlan e t al., 1989) , and the twoVL dimers, REI (Epp et al., 1975) and RHE (Furey et al.,1983). Defects in the stereochemistry of these models wereeliminated as before by energy minimization using the pro gramENCAD (Levitt, 19 83a). The new models of TE32 and TE 33are bette r than th e models built previously (Levy et al., 198 9)using six antibodies known then (KOL, NEW, McPC 603,5539, RE I, and RH E) in th at they have fewer missing atoms(41 and 47 as opposed to 117 and 126) and lower total energyvalues after energy refinement (-1717 and -1735 kcal/ mo lasopposed to-1486and-1534 kcal/mol). Thenew structuresarev erys imila r to that previouslydescribed (Levy et al., 1989)except for the conformation of a segment of CDR2 of theheavy chain starting from Ile 5 1H . Th e conformation of Tr p50H, which is shown by N M R to interact with the peptide,is similar in the two models. Th e possibility that this segm entof CDR2 was not modeled correctly in the first models wasdiscussed previously (Zil ber et al., 1990). Th e improvementin the second model arises from using a m uch more homologousstructure of C DR 2 in th e calculations.

Calculationof the Models ofpeptide-Antibody Complexes.The NMR-derived distance restraints and the preliminarymodels f a he antibodies Fv were used to calculate a mo-lecular model for the antigenic determinant VPGSQHIDbound to TE3 2 and TE33 . With the powerful moleculargraphics program FRODO(Jones, 1978) on an Evans & Suther-land PS390, this sequence was docked into the TE33 com-bining site by bringing each of its amino acid residues nearcombining site residues shown by N M R to in teract. In allcases reasonable lengths for the peptide bonds were maintained.This initial model of the complex was refined by a combinationof restrained energy minimization and molecular dynamicscalculations. W e used restraints that depend on the fourthpower of the difference between NOE and model distance asdescribed before (Zilber et al., 1990).

For macromolecular systems that have many degrees offreedom, minimization with the distance restraints is not

Biochemistry, Vol. 31, No. 30, 1992 Scherf et al.Table I:with CTP3NMR -Derived Distance Restraints for the TE 33 Complex

distance distancecons traintb in modelC deviationdatom in atom j4 4 4 4val 3 CyzVal 3 H,Val 3 C,1Val4 C,Ipro 4 Cypro 4 C6pro 4 Capro 4 Halpro 4 H62pro 4 C,pro 4 H32pro 4 H32pro 4 H,Ipro 4 H,2pro 4 H72gly 5 c,gly 5 C,gly 5 c,gln 7 H,Igln 7 H,Igln 7 H,Ihis 8 Cphis 8 Ha2his 8 H,Ihis 8 H,lhis 8 Ha2his 8 Ha2his 8 C3his 8 Ha2his 8 phis 8 Cpasp 10 Cpasp 10 C BTyr 32L Hr2Tyr 32L Hc2val 3 C,,val 3 C,Ipro 4 Halpro 4 Ha2d 5 HglY 5 Hser 6 C3ser 6 Hser 6 C3ser 6 Hgln 7 Hile 9 Calile 9 C y ]ile 9 H ,llile 9 H,l2ile 9 H

His 3 1L H,1His 31L H,ITyr 32L H 4Trp lOOaH HalHis 3 1L Ha2His 3 1L Ha2His 3 1L H,1Tyr 32L Hr2Tyr 32L H,2Tyr 32L Hs2Phe 9 6L H,zPhe 96L HtPhe 96L HfPhe 96L Ht2Phe 96L H,Phe 96L He2Phe 96L HITrp SOH H,,2Trp SOH H pTr p SOH H72Trp 50H H pTyr 3 2H H,2Tyr 32H H62Tyr 32H H,2Tyr 32H Ha2Trp lOOaH Hn2Trp lOOaH H pTrp lOOaH H aTrp lOOaH H pTrp lOOaH H,2Trp lOOaH H pTyr 32L H,zTrp lOOaH HalHis 3 1L Ha2His 3 1L H, Iasp 10 Hplasp 10 H02val 3 C,IVal 3 C,Ipro 4 H72pro 4 H32val 3 Hgval 3 C Y Ival 3 C,IglY 5HVal 3 C3asp 10 Hasp 10 Hile 9 Hile 9 Hasp 10 HasD 10 H Val 3 C n

4.03.04.04.04.05.05.04.04.05.03.04.04.04.04.04.04.05.03.03.03.05 . 54.04.04.03.03.06.03.05.06.05.55.54.04.04.04.04.04.02.53.05.04.06.03.06.04.04.53.03.03.55.5

4.42.84.34.04.74.45.04.44.45.63.72.33.94.62.34.04.03.83.63.03.36.04.32.72.83.13.45.53.54.83.05.85.93.54.03.03.73.33.73.02.54.24.36.12.34.54.24.02.52.14.35.3

0.40.00.30.00.70.00.00.40.40.60.70.00.00.60.00.00.00.00.60.00.30.50.30.00.00.10.40.00.50.00.00.30.40.00.00.00.00.00.00.50.00.00.30.10.00.00.20.00.00.00.80.0

Peptide residues are w ritten w ith lower-case letters while antibodyresidues are written with cap ital and low er-case letters; H, heavy chainresidues; L, light chain residues. The value given reflects the maximumdistance from the N O E restrain ts and additional distance due to the useofpseudoatoms(Wiithrichet al., 1983). Thedistancesin thefinalmodelwere obtained by energy minimization and molecular dynamics calcu-lations. Thedeviationisthat betweentherestraint and theactual distancein the model. When the latter is smaller than the restr aint, the deviationis set equal to zero.generally ab le to get to the lowest accessible energy minimum.Early studies (Levitt, 1983b) show that running moleculardynamics between energy minimizations is a very effectiveway to get low energies. Th e protocol used for annealingdynamics sta rted with minimization consisting of four passeseach of 200 steps (here a step is a single evaluation of theenergy and its first derivative). This was followed by tenpasses each consisting of 10 ps of molecular dynamics (10000steps each lasting 0.001 ps) a t a temperature of 300 K followed

8/12/2019 bi00145a004

8/14

Peptide-Antibody Complexes Biochemistry, Vol. 31, No. 30, 1992 68911.OA for the side-chain atoms. The averaged rms deviationbetween the model with the lowest energy and the nine othermodels for each of the residu es of the epitop e is given in Figu re9A. Th e higher dispersion aroun d hiss is probably a result ofthe fact th at the two antibody residues (T rp lOOaH and T yr32H ) that interact with his* were allowed to move. We didnot restrain interaction of ile9 with the antibody, and th erear e only three re straints on ile9 interactions with adja centpeptide residues. As a result the conformation of ile9 s poorlydefined. In contrast, the backbone and the side-chainconform ation of vaP, pro4, gln7,and asplo are very well-defined.

Ten m odels were also calculated in th e sam e way for theTE 32 complex, except that residue 96L was not allowed tomove (Phe in TE3 3 is replaced by Leu in TE32 ). The rmsdeviation between the 33 NMR restraint distances and thedistances in the model with the lowest energy is 0.36 A, andthe total energ y of the restrained interactions is -1 18.6 kcal/mol. Figure 8 shows a comparison between the models forthe bound peptide in TE32 and TE33 , respectively. Due toa smaller number of distance restraints, the conformationalvariability in the T E32 complex is considerably greater tha nthat observed in the TE33 complex; however, the peptideassumes very similar conformations in the two antibodycomplexes. The rm s deviation between the nine models andthe model with the lowest energy is 1.06 A for the main-chainatoms and 1.47A for the side-chain atoms. The statistics foreach of the residues of the epitope is given in Figure 9B.

Description of Structure. Panels A and B of Figure 10illustrate the models obtained for the CT P3 complexes withTE 33 and TE3 2, respectively. Tables I11 and IV sum marizethe contacts in the calculated NM R-deriv ed models betweenCT P3 and TE 33 and TE3 2, respectively. In the TE33 complexthe epitope includes CT P3 residues 3-10 with at least sixresidues directly interacting with the antibody. Th e side chainsof va13, pro4, glys, gln7, and hiss are buried inside th e complexand fill a circular groove around T rp 100aH. The shortestdistance between the two ends of the loop formed by the pep-tide, from the va13 methyl protons to the asplo o protons, is2.4 A. The four N -terminal residues of the epitope (VPGS)form a 8-tu rn conformation w hich is stabilized by a hydrogenbond between t he va13 carbonyl oxy gen and t he am ide protonof ser6 and a bifurcated hydrogen bond between ser6 0, H andthev aP carbonyl. Intramolecular polar interactions are foundbetween the gly5 N H and the gln7 0,1 .

Additional intermolecular polar interactions exist betweentheser6carbonyloxygen and the Tyr 53H O fH, the gln76NH2and th e guanidinium group of Arg 95H, th e gln7 carbonyloxygen and the Tyr 53H OCH, and the asplo Oh and both theAsn 31eL 7NH 2 and the His 31L Hal protons. Stronghydrogen bonds are formed between his8 N,2 and S er 96 Hand between asplo 0 6 2 and both Tyr 32L OcH and Ser 31cL0,H. van der W aals interactions are found between pro4 HY 1and the S er 92L carbonyl oxygen and between hiss H,I andthe T hr 3 1H carbonyl oxygen. va13 has hydrophobic inter-actions with His 3 lL , Tyr 32L, and Tr p 10 0aH; pro4 nteractswith His 3 1L, Tyr 32L, His 93L, Ile 94L, and Ph e 96L; glysCH2 interacts with Ile 94L, Phe 96L, and Trp 50H; ln7interacts with Phe 96L, Trp 50H nd Tr p 100aH; and thehiss imidazole is buried in a hydrophobic pocket formed byTyr 32H and Trp 100aH.

A similar structu re obtained for the TE32-CTP3 complexis shown in Figure 10B. The automatic program modeleddifferently the side chain of Arg 95H in TE32 , and as a resultth e position of the side chain of gln7 had b een changed in th ecalculated model for the antibody-peptide complex. Ac-

Table 11: NMR -Derived Distance Restraints for the TE32 Complexwith CTP3distance distanceconstraintb in model' deviationdatom ia atom i A) A) A)

Val 3 Cy2val 3 C,IVal 3 C,Ipro 4 C,pro 4 C6pro 4 CaPro 4 Ha1pro 4 H62pro 4 C,gln 7 C,gln 7 C,gln 7 C,his 8 c,9his 8 Ha2his 8 H,Ihis 8 H,Ihis 8 Ha2his 8 Ha2his 8 c,4his 8 c,9his 8 c,9

gly 5 Ca

asp 10 C,9asp 10 C,9Tyr 32L H,2Tyr 32L HI?Val 3 C,,val 3 C,lpro 4 Halpro 4 H62ser 6 Cpser 6 Ha

His 31L H,ITyr 32L H,2Trp lOOaH HalHis 31L Ha2His 3 1L Ha2His 31L H,1Tyr 32L Hf2Tyr 32L H,2Tyr 3 2L H,2Trp 50H H,zTr p 50 H H,,2T r p 5 0H H pT r p 5 0H H nTyr 32H H,2Tyr 32H Hf2Tyr 32H H,2Tyr 32H Ha2Trp lOOaH H,,2Trp lOOaH HpTrp lOOaH HpT rp lOOaH H,2Trp lOOaH H nTrp lOOaH HalTyr 32L Hs2His 31L Ha2His 31L H,Iasp 10 H,9lasp 10 H,92Val 3 C,]Val 3 C,]Val 3 C,IVal 3 H,9ser 6 c9

Peptide residues are written with lower-case letters while antibodyresidues are written with capital and lower-case letters; H, heavy chainresidues; L, light chain residues. Th e value given reflects the max imumdistance from the N O E restraints and additional distance due to the useofpseudoatoms (Wuthrich et al., 1983). Thed istancesin the final modelwere obtained by energy m inimization and molecular dynamics calcu-lations. d Thedev iation is that between the restraint and thea ctua l distancein the model. When the latte r is smaller than th e restraint, the deviationis set equal to zero.

Val 3 C,l

4.04.04.04.05.55.54.04.05.05.54.04.04.05.54.04.04.03.03.06.05.06.05.55.54.04.04.04.04.04.05.04.06.0

4.64.73.94.14.14.44.64.35.65.14.14.64.35.93.53.93.43.33.45.24.43.65.65.83.34.53.24.43.33.74.74.56.6

0.60.70.00.10.00.00.60.30.60.00.10.60.30.40.00.00.00.30.40.00.00.00.10.30.00.50.00.40.00.00.00.50.6

by minimization for 600 steps. Note th at th e moleculardynamics tha t followed the energy minimization was continuedfrom th e time step immediately preceding the minimization.This gave a single continuous trajectory lasting 100 ps whichwas sampled by minimization every 10 ps to provide 11different energy minima. All hydrogen atoms are includedin these energy minimization and molecular dynamics cal-culations using energy parameters given in Levitt (1983a).Ten models were calculated for the T E33 complex. In thesecalculations all peptide residues and only a few antibodyresidues (Phe 96L, Tyr 32H, Arg 95H, Ser 96H, and Trp100aH ) were allowed to move. Tables I and I1 also give thedeviations between the N M R restraints and th e distances inthe calculated model with the lowest energy. Th er ms deviationbetween the 52 NM R restraint distances and the distances inthe calculated models for TE 33 is 0.32A, and the total energyof the restrained interactions is -147.3 kcal/mo l. In view ofthe fact that a model rather than a known crystal structurewas used for docking the peptide, th e violations of the distanc erestraints are remarkably low, strongly supporting the highquality of the model and the reliability of the NM R-derive ddistance restraints. The bound peptide conformation in theensemble of the calculated mo dels is shown in Figures 7 an d8. Th erm s deviation between the nine models and the modelwith the lowest energy is 0.67 A for the main-chain atom s and

8/12/2019 bi00145a004

9/14

6892 Biochemistry, Vol. 31, No. 30, 1992 Scherf et al.

I/F I G U R E : Stereoview showing superposition of the five best-fit calculated models for the polypeptide backb one of the segm ent VPG SQ HIDbound to TE33.

IS 8 HIS 8

, / \ T l f \ / GLN 7SER 6 SER 6

HIS 8

GLY 7 G L V 7SER 6 SER 6F I G U R E : Stereoview showing the conformational variabil i ty of the models for the segmen t VPG SQ HID of CTP3 . Th e top figure showsthe models for the TE32 com plex, and the bottom fig ure shows the models for the TE 33 complex. For clarity, only the first six models weredrawn.cordingly, all polar interaction s involving the gIn7 side-chainatoms are missing, except for a hydrogen bond between theOC1nd the gln7 amide proton (of the polypeptide backbone).A new intrapeptide hydrogen bond is found between th e car-bonyl oxygen of pro4 and one of th e gln7 6N H Zprotons. The@-turnconformation found in the epitope of TE33 is alsopredicted in TE3 2 and is stabilized by a hydrogen bond betweenthe N H of ser6 and the c arbon yl oxygen of va13.The total contact a rea between the antibody and the pep-tide epitope is 406 A2 in TE33 and 344 A2 n TE32. The totalcontact a rea is about half of th at observed in antibody-proteincomplexes (S heriff et al., 1987; Amit et al., 1 986) and iscomparable to tha t previously found for TE34 (Zilber et al.,1990). Th e side chains of antibody a romatic residues occupy73%ofthiscontact area. Allantibody CDRsexce pt thesecondCD R of the light chain interact with the peptide. In the pep-t ideboundei ther toTE32 or toTE 33,about halfofthecontactwith the antibody is made by gln7 and hiss, and most in-

tramolecular contacts are made by va13. The peptide con-format ion is stabiliz ed by folding aro und va13, which ac ts asa hydrophobic core.The model obtained for CTP3 bound to TE33 when the

restraints on the amide proton interactions are not taken intoaccount is very similar to th at obtained when all restraints a retaken into account and is similar to that obtained for theTE32 complex.DISCUSSION

The Potential of the TR NOE Technique. This studydemonstrates the power of T R NO E measurements to elucidatethe conformation of bound ligands and their inte ractions withlarge proteins. For proteins exhibiting slow ligand off-ratesresulting in weak T R N O E cross-peaks, slight modification ortruncation of the ligand c an be used to ob tain well-resolved2D T R N O E difference spectra with good signal-to-noise ratios

8/12/2019 bi00145a004

10/14

Peptide-Antibody ComplexesA Brckbonr Alomr B

0.50.02 4 6 8 1..*-- 4,

Biochemistry, Vol 31, No. 30, 1992 6893

Sl d r c h r l n A t o m3.02s.0

iio0.50.02 4 6 8 10

AII Atomr23

0.50.0 12 4 6 8 1

Sl d r c h r l n A tomr3.0

110.0 2 4 8RrsldurFIGURE : Average rms deviation between the model with the lowest energy and the nine other models for backbone atoms, all atoms, andside-chain atoms of the epitope residues. Panels: A) TE33-CTP3 com plex; B) TE32-CTP3 complex.

(Anglister & Zilber, 1990). Transferred N O E measurementsare best applied when the three-dimensional structu re of theprotein is known, and one is interested in studying theinteractions between a protein and its ligand and the boundligand conformation. Howe ver, even in the absenceof a solvedprotein struc ture, as in our case, a m odel for the protein canbe used successfully.Modeling of the antibody combining site was found to bevery successful in a number of cases in which it could becompared to a solved antibod y struc ture (C hothia et al., 198 7,1989; Padlan et al., 1989). The NMR results are used toverify and refine the initial model for the free protein and todock the ligand into the combining site as was done in th epioneering work of Dwek and his co-workers (Dwek et al.,1977) and later by McConnell and his co-workers (Anglisteret al., 1987; The riau lt et al., 1991 ). In very favorable cases,one can obtain enough NM R restraints on intramoleculardistances in the bound ligand to solve its conformationindependently of a calculated model (Clore et al., 1986). Inour studies we did not resolve a large enough number of in-trapeptide N O E distance restraints to permit unequivocal de-termination of the structur e of bound CTP3. Enough datawere obtained to qualitatively describe the conformation ofbound CTP 3 peptide: a loop and a 6-tur n formed by thesequence VPGS in the TE3 2 and T E33 complexes and a 8-turn formed by the sequence IDSQ in the TE 34 complex. Th e3 J ~ ~oupling constants between th e (Y and the amide protonsof the sam e residue can be used to obtain restraints on the r ~dihedral angles along the polypeptide backbone only whenthe chemical shifts of the bound and free ligand are averagedas a result of a fast excha nge rate relative to th e chemical shiftdifferences. The 3 J ~ ~oupling constants for the bound pep-tide can be deduced from m easurements using various ratiosbetween free and bound peptide and extrapolating to zeroconcentration of the free peptide. This is not the case for thethree antibody complexes that we have investigated.Influence of Spi n Diffusion. Experiments carried out using30- and 60-ms mixing times ruled out the possibility thatsome of the cross-peaks hat we observed with a 100-ms mixingtime were due to spin diffusion. This larger mixing time wasnecessary to obtain a good signal-to-noiseratio for many cross-peaks. The absence of interference from spin diffusion in theexperiments that we have carried out could be due to the

following: (a) A ll the restraints were derived from experimentsin which we used extensive deuteration of the aromatic am inoacids of the antibody. In the experiments with TE34 allaroma tic amino acids except histidine were deu terated, andin TE33 tryptophan and phenylalanine were perdeuteratedand tyrosine was deuterated at Ca positions. As the com-bining sites of all thre e antibodie s ar e highly arom atic, manyrelaxation pathways that could lead to interresidue spindiffusion were blocked by antibody deuteration. (b) T herotational correlation time of the Fab was previously foundto be 20 ns at room tempe rature (Anglister et al., 1984). AllT R N O E experiments that were reported previously and thatwere used to derive restraints on intermolecular d istances werecarried out between 37 and 42 OC. The higher temperatureprobably reduces the rotational correlation time by as muchas 40%, mainly due to the reduced viscosity of water, thusfurth er decreasing the efficiency of spin diffusion. (c) Indifference spectroscopy one observes only relatively strongcross-peaks, many of which represent distances shorter than3A. In this range of distances and for the rotational c orrelationtime a t the elevated temperatur e and a m ixing time of 100ms, errors in evaluating interproton distances ar e minimal asshown by Cam pbell and Sykes (1991).Size and Conformation of the Epitopes. There ar e severalcommon features in the NMR-derived models for the three-dimensional structur e of the thre e peptide-antibody complexesthat we have determined and in the structure of anothercomplex solved by crystallography (Stanfield et al., 1990).(a) Th e epitope consists of about eight residues. (b) Thecontac t are a be tween the two molec ules is 350-500 AZ, bout60% of tha t observed in protein-antibody complexes. (c) Thepeptide folds into a &t ur n in all four anti-peptide antibodies.In TE3 2 and TE3 3 this turn is part of a larger loop. Thiscompact folding maximizes the contact a rea between the pep-tide and the antibody and allows greater conformationalfreedom than a-helix or &sheet secondary structure.Mechanism ofA ntibo dy Interactions with Peptide Antigens.In the N M R spectra of free CTP3 we do not see any evidencefor a preferred @-turn onformation in the sequence VPGS orIDSQ . The interactions between the amide protons of gly5and ser6, and serll and glnl*, which are indicative of th e 0-turn in VPGS and IDSQ respectively, are weak and com-parable to interactions between other pairs: ser6 and gln7,

8/12/2019 bi00145a004

11/14

6894 Biochemistry, Vol. 31, No. 30, 1992 Scherf et al.

\U

B

r

i

c

F

FIGURE0: A) Stereoview of the lowest energy structure of CTP 3 bound to the com bining si te of TE33 obtained after energy minimization,restrained molecular dynamics, and simulated annealing calculations using N M R distance restraints and a preliminary model for the antibodyFv: l ight blue (top), yellow, and red, C DR 1, 2, and 3 of the l ight chain, respectively; l ight blue (bottom ), blue, and purple, CD R 1, 2, a n d3 of the heavy ch ain; light gree n, residues 3-10 of CTP 3. B) Corresponding structure of the TE32-C TP3 complex. The colors used representthe same regions as specified in A).gln7 and his8, his8 and ile9, asp loand ser ll, an d gln12 and lys13(Figure 2). It is therefore concluded that when CT P3 is freein solution, it does not fold into a w ell-defined conform ation,probably assuming a rapidly interconverting distribution ofdifferent conformations.At room temperature the off-rate of CTP3 bound to TE 33is of the order of 1 s-1 (Anglister e t al. 1 988), and its bindingconstant is 1.2 X lo6M-l, Its on-rate, therefore, is approx-imately lo6 s-l, which is slower than what is expected froma diffusion-controlled on-rate. The fact th at the on-rate is

about 2 orders of magnitude slower tha n a diffusion-controlledon-rate probably indicates peptide binding by an induced-fitmechanism in which the peptide folds into a conformationthat is dictated by the much less flexible structure of theantibody (Dyson et al., 1988a). Immu nization with peptidesinduces th e proliferation of those B -cells carrying preexistingreceptors into which the peptide can fold and create con tactsthat a re energetically favorable. Our studies demo nstratethat a short peptide can present different epitopes to theimm une system and different bound peptide conformations

8/12/2019 bi00145a004

12/14

Peptide-Antibody Complexes Biochemistry, Vol. 31, No. 30, 1992 6895Table 111: Contact.Areas' between the Peptide and the TE33 Antibody

contact area (in A2 toloop residue val 3 pr o4 gly 5 ser 6 gln 7 his 8 ile 9 asp 10 total residues totalloop

C D R l H i s 3 1-Ser 31CbAsn 31ETyr 32C D R 3 S e r 9 2His 93-1le 94-Phe 96C D R l - T h r 3 1Tyr 32C D R 2 T r p 5 0Tyr 53CDR3 -Arg95-Ser 96Trp lOOAtotal

15 15

14 514720 75

23

1342 61 35

Light Chain 1415158

18Heavy Chain 12223 3117 147

5 1321 40 1320 96 87 13 52

44151527 1011472123 711222 345731 8871887 112

406 4060 Contact areas ar e calculated using a variant of the m ethod of Lee and R ichards (1971) to calculate accessible surface area. In that method, theaccessible surfac e area of an a tom is that of a sph ere of radiu s equal to the radius of the atom plus the radius of a water molecule (taken as 1.4 A)minus the sum of contact areas with all adjacent atoms. These contact areas can overlap and are normalized by the actual total contact area afterallowing for all overlap. Thus, the con tact areas measured h ere sum to give the loss in accessible surfac e area of each ato m or residue. The individual

values indica te how m uch the conta ct of a particu lar pair of residues rpduces the total solvent-accessible surfac e area. Con tact areas smal ler than 3A2 are omitted for greater clarity. Residues that ar e different in DB3 and TE3 3 are marked with an -.Table IV: Contact Areas' between the Peptide and the TE32 A ntibody

contact areat (in Az tolooD residue Val3 pr o4 gly 5 ser 6 gln 7 his 8 ile9 asp 10 totalresidue s total loop

CD Rl His31 12 19+Ser 3 1CbAsn 31ETyr 32 12C D R 3 S e r 9 2 1 4Val 94 26

Light Chain

16Heavy Chain

CDR l -Thr31 14Tyr 32 18C D R 2 T r p 5 0 1 0 3 5Tyr 53 7 8 4CDR3 +Arg95 22 12Tr p lOOA 15 13 34 3-Ser 96 8

9 4019 196 68 20 851442 56

1418 324519 6434865 107total 39 59 10 7 94 90 3 42 344 344

See footnote a to Table 111. See footnote b to Table 111. Energy values show the impo rtance of th e variable polar residue more clearly.

are induced upon binding to different antibodies.Folding of a flexible peptide into a @-turn n the antibodycomplex was also demonstrated in studies of the confor mationof a myohemerythrin peptide free in water and trifluoroeth-

anol solutions (Dyson et al., 1988b) and bound to an antibody(Stanf ield et al., 1990). Although m ultiple weak interactionsbetween the amide protons of adjacent amino acids wereobserved in the sequence DFL EK IGG L (residues 11-19) ofthe peptide in aqueous solution and addition of TF E stabilizeda helical conformation at the C-terminus, no nteractions wereobserved in the N-terminal part (EVVPHKKMHK) of the19-residue peptide in aqueous or TF E solution. The lattersequence includes the epitope that interacts with t he antibodystudied by Stanfield et al. (1990) and which assumes a @-turnin the bound state.Implications for the Cross-Reactivity of Anti-PeptideAntibodies with Native Proteins. Arnon, Sela, and their co-workers have shown that anti- CTP 3 serum partially neutralizes

the biolo gical activity of cholera toxin in vitro (Jacob et al.,1983, 1984). Anti-cholera toxin seru m was found to havemuch stronger activity. In our experiments the monoclon alantibody T E33 failed to neutralize the biological activity ofcholera toxin (Reuben Hiller, unpublished results), and itsbinding to cholera toxin in solution was 3 orders of mag nitu deweaker tha n its binding to the peptide. By followingretardation in size-exclusion HP LC , Spangler (1991) couldnot detect TE3 3 and TE3 2 binding tocholera toxin in solution.However, this method cannot detect the very weak bindingthat we found in our ESR experiments. Previous studies fromvarious groups found that the affinity of anti-peptide antibodiesto the native protein is usually lower than t he affinity to theimmun izing peptide by 2-3 orders of magnitud e (Berzofsky,1985; Darseley at al., 1985; Hira yam a et al., 1985). In viewof the modest binding to the peptide and thevery weak bindingto cholera toxin, it is not surprising that TE33 does notneutralize cholera toxin. Unfortunately, TE 33 was selected

8/12/2019 bi00145a004

13/14

6896 Biochemistry, Vol 31, No. 30, 1992on the basis of its strong cross-reactivity with cholera toxinin a solid-phase assay which apparently does not correlatewell to its binding in solution. The refore , the neutra lizationactivity of anti-CT P3 serum must be a ttributed to antibodiesthat have much stronger binding to cholera toxin than doTE3 2 and TE3 3. The observed difference betweer. the cross-reactivity of TE33 an d TE3 2 and tha t of TE34 with choleratoxin in ELIS A is mostly due to the fact th at T E34 bindingto CTP3 is 2 orders of magnitude weaker if the peptidesC-term inal carboxyl is modified into an amide. Since theTE3 4 epitope in intac t cholera toxin ends with an amide bondto residue 65 of the B -subunit, it is not surprising that T E34interac ts very weakly with the protein. Given the abovefinding, a fur ther search for the ne utralizing activities in othe ranti-CT P3 monoclonal antibodies might prove very worth-while.

Calculations made by Barlow et a l. (1986) showed that, ofthe possible conformations for continuous protein epitopes,loops protruding from the s urface have the highest probabilityof forming the large contact surfaces necessary for highantibody binding affinity . The refore , synth etic peptides cor-responding in sequence to such loops are potentially good im-munogens for obtain ing antibodie s cross-reactive with nativeproteins. The three -dimensional structure of cholera toxinhas not been published. However, Six ma et al. (1991) solvedthe stru cture of the heat-labile enterotoxin of E . coli (H L T )which shares 80% sequence homology with cholera toxin. CT P3corresponds to residues 50-64 in both toxins, and anti-serumagainst CTP 3 recognizes both cholera toxin and HL T. InH L T residues 51-58 (residues 2-9 of CT P3 ) form a flexibleprotru ding loop. This loop connects a p-st ruc ture (residues47-50) and a long a-he lix (residues 59-78). Th e flexibilityof this loop an d its protrusion m ay explain th e results of Jaco bet al. (1983,198 4), whode mon strated that of several peptidescovering the complete sequence of the B-subunit of choleratoxin CTP 3 was the most potent in inducing anti-serum cross-reactive with cholera toxin.

It is interesting to compare th e model for the TE33-boundCT P3 to the recently obtained structure for HL T. Super-position of CTP3 bound to TE33 and the correspondingsequence of H L T shows the following: (a) Th e side chainsof gln7, his*, and asp lo in H L T point tow ard th e solvent andcould interact with an antibody. (b) T her e is a good fit betweenour model and the HLT structure with respect to theconform ation of the side chain of gln7 an d its orientationrelative to the peptide backbone. (c) The orientation of theside chain of ile9 is very similar in our m odel and in HL T. (d)Overall, the segment GS QH ID has sim ilar conformations inour model and in H LT , but for good superposition, the his8imidazole requires about a 90 rotation. (e) There aredifferences in the conformation of the sequence VP, and the@-turn VPGS is not observed in H LT . The r ms deviationbetween the conformation of the antibody-bound epitope andthe corresponding part of H LT is 2.6 A. The ac tual bindingof the antibody to the epitope on H LT may require a slighttilt of the loop to further expose the histidine imidazole toenable its interaction with the hydrophobic pocket in theantibody combining site. Bearing in mind that the epitoperecognized by TE33 and TE32 corresponds to a flexibleprotruding loop in HLT and that the conformation of thesequence G SQ HI D is similar in the antibody-bound peptideand in HL T, one can expect that the flexible loop in H L T andin cholera toxin, if indeed its three-dimensional str uctu re provesto be similar to th at of HL T, can fold itself to the conformationrecognized by TE3 3. As the conformations are not identical,

Scherf et al.such folding is probab ly accom panied by a positive contributionto the free energy of binding resultin g in the observed decreas ein the binding consta nt of TE3 3 to cholera toxin. If in solutionthe peptide adopts a stable conformation similar to that of thecorresponding part of the protein, a higher degree of cross-reactivity w ith the native protein is expected. Su ch antibodieselicited against neutralizing epitope may have strongerneutralizing activity.

Relationship between Antibody Primary Structure andFunctionof the Antibod y Combining Site. A computer searchin antibody sequences reveals th at T yr 3 2L is invaria nt in alllight chain sequences that share more than 65% sequenceidentity with T E33 (94 antibodies were found in this search).Tyr 3 2H an d Tr p 50H are invariant in all heavy chains, sharingmore tha n 75% sequence identity with TE33 (25 antibodies).W e have previously pninted out th at Ty r 32L, Tyr 3 2H, andTrp 50H interact with the peptide antigen in all three anti-peptide antibodies that we have studied although TE34recognizes a peptide segment that is different from thatrecognized by TE3 2 and TE33 (Zilber et al., 1990). Tyr 32LandT yr 32H interact with theantigenalsoin theD1.3complexwith lysozyme (Am it et al., 1986). This conservation indicatestha t a mutation in these thre e arom atic residues in the highlyhomologous antibodies probably destroys the capability ofthe antibodies to bind antigens and/or may have severeimplication on their folding to functional antibodies. Thecentral location of the side chains of these residues in theantibody combining site and the fact that they interac t withdifferent antigens in different antibodies indicate that theirrole is to participate in nonspecific hydrophobic interactionswith the antigen.In all three anti-CTP3 antibodies that we have studied,C D R l of the light chain has the same canonical conformation(Chothia & Lesk, 1987;ChothiaetaL 1989). Thesidechainsof residues 31, 31c, 31e, and 32 interact with the antigen inall three complexes. His 3 lL , Ser 3 IC, and Asn 31e in TE32and T E3 3 are replaced by Asp 31L, Asp 31c, and Lys 31e,respectively, which are involved in electrostatic interactionsin the TE 34 complex. These findings imply that the foldingof the C DR and its length dictate the positions in the sequencethat interact with the antigen. In CD Rs that have the samelength a nd conforma tion, the side chains in these fixed positionshave the potential to interact with the antigen.Th e difference in the specificity between TE 34 and the twoother anti-CTP3 antibodies seems to arise from two mainreasons: (a) Several mutations involving mostly polar inter-actions, for example, the mutations in C D R l of the light chain,enable the specific recognition of the C-terminal carboxyland the positively charged t-amine of lys13. (b) Changes inthe length of CD R3 of the heavy chain cause major changesin the shape of the combining site accompanied by changesin hydrophobic and polar interactions with the antigen. Aminoacid sequence comparison with the highly homologous heavychain reveals that C DR 3 of the heavy chain varies considerablyin its length and amin o acid sequence. Th e deletion of threeCD R3 residues including T rp lOOaH changed the shape ofthe combining site from a circular groove into a deep cavity.ACKNOWLEDGMENT

Wearegreatlyindebted to Prof. Ruth Arnon and Dr. ChaimJacob for the gift of the TE 32, TE33, and TE 34 hybridomas,to Dr. Ian Wilson for allowing us to use the coordinates ofDB3 before their general release, and to Drs. Hol and Sixm afor sending us the coordinates of HL T. W e are thankful forthe use of the a mino acid sequence d ata in the following banks:

8/12/2019 bi00145a004

14/14

Peptide-Antibody Complexes Biochemistry, Vol. 31, No. 30, 1992 6897Swiss protein, Genepept (Genebank), and Protein Identifi-cation Resource.REFERENCESAmit, A. G., Mariuzza, R. A., Phillips, S.E. V., & Poljak, R.Anglister, J., & Zilber, B. 1990) Biochemistry 29 921-928.Anglister, J., Frey, T., & McConnell, H. M . 1984) BiochemistryAnglister, J., Bond, M. W., Frey, T., Leahy, D ., Levitt, M., Mc-Connell, H. M., Rule, G. S., omasello, J., & Whittaker, M.1987) Biochemistry 26 6058.Anglister, J., Jacob, C. O., Assulin, O., Ast, G., Pinker, R., &Arnon, R. 1988) Biochemistry 27 717.Anglister, J., Levy, R., & Scherf, T. 1989) Biochemistry 283360.Arnon, R. 1986) Trends Biochem. Sci. 11 521-524.Barlow, D. J., Edwards, M. S., Thornton, J.M. 1986) NatureBerzofsky, J. 1985) Science 229 932.Campbell, P. A,, & Sykes, B. D. 1991) J Magn. Reson. 93Chothia, C., & Lesk, A. M. 1987) J Mol . Biol. 196,901-917.Chothia, C., Lesk, A. M., Tramontano, A., Levitt, M., Smith-Gill, G. A., Sheriff, s.,Padlan, E. A., Davies, D., Tulip, W.

R., Colman, P. M., Spinelli, S., lzari, P. M., & Poljak, R.J . 1989) Nature 342 877.Clore, G. M., Gronenborn, A. M., Carlson, G., & Meyer, E. F.1986) J Mol . Biol. 190 259-267.Colman, P. M., Laver, W. G., Varghese, J. N., Baker, A. T.,Tulloch, P. A., Air, G. M., & Webster, R. G. 1987) NatureDarsley, M. J., & Rees, A. R. 1985) EMBO J 4 383.Deverson, E., Berek, C., Taussig, M., & Feinstein, A. 1987)Eur. J Immunol. 17 9-13.Dwek, R. A., Wain-Hobson, S., ower, S., ettins, P., Sutton,B., Perkins, S. ., & Givol, D. 1977) Nature 266 31-37.Dyson, H. J., Lerner, R . A., & Wright, P.E. 1988a) Annu. Rev.Biophys. Biophys. Chem. 17 305.Dyson, H. J., Rance, M., Houghten, R. A., Wright, P. E., &Lerner, R. A. 1988b) J Mol. Biol. 201 20.Epp, O. Lattman, E. E., Schiffer, M., Huber, R., & Palm, W.1975) Biochemistry 14 4943.Furey, W., Jr., Wang, B.C., Yoo, C. S.,& Sax, M. 1983) JMol. Biol. 167 661.Gaffney, B. J. 1976) Spin L abeling Theory and Applications(Berliner, L. J., Ed.) pp 184-238, Academic Press, New York.Hirayama, A., Takagaki, Y., & Karush, F. 1985) J Zmmunol.134 3241.

J . 1986) Science 233 747-753.

23 1138-1 142.

322 747.

77-92.

326 358-363.

Jacob, C. O., Sela, M., & Arnon, R. 1983) Proc. Na tl. Acad.Sci. U.S.A.80 761 1.Jacob, C. O., Sela, M., Pines, M., Hurwitz, S.,& Arnon, R.1984) Proc. Natl. Acad. Sci. U .S.A. 81 7893.Jones, T. A. 1978) J . App l. Crystallogr. 11 268.Lee, B., & Richards, F. H. 1971) J Mol . Biol. 55 379.Levitt, M. 1983a) J Mol . Biol. 168 595-620.Levitt, M. 1983b) J Mol . Biol. 168 621-657.Levy, R., Assulin, O., Scherf, T., Levitt, M., & Anglister, J .Macura, S.,& Ernst, R. R. 1980) Mol. Phys. 41 95-117.Marq uart, M., Deisenhofer, J., Hu ber, R., & Palm, W. 1980)J. Mol . Biol. 141 369.Padlan, E. A., Silverton, E. W. , Sheriff,S., ohen, G. H., S mith-Gill, s. J., & Davies, D. R. 1989) Proc. Natl. Acad. Sci.Piantini, U., Smensen, 0 W., & Ernst, R. R. 1982) J Am.Plateau,P., &Gueron,M. 1982) J Am. Chem.Soc.104,7310-Satow, Y. Cohen, G. H., Padlan, E. A., & Davies, D. R. 1987)Saul, F. A., Amzel, L. M., & Poljak, R. J. 1978) J Biol. Chem.Sheriff,& Silverston,E. W., Padlan, E. A., Cohen, G. H., Smith -Gill, s. ., Finzel, B. C., & Davies, D. R. 1987) Proc. Na tl.Acad. Sci. U.S.A. 84 8075.Sixma, T. K., Pronk, S.E., Kalk, K. H., Wartna, E. S., anZante n, B. A. M., Witholt, B., & Hol, W. G. J. 1991) Nature

1989) Biochemistry 28 7168.

U.S.A.86 5938-5942.Chem. SOC 04 6800-6801.7311.J . Mol . Biol. 190 593.253 585-595.

351 371-377.Spangler, B. D. 1991) J Zmmunol. 146 1591-1595.Stanfield, R. L., Fieser, T. M., Lerner, R. A., & Wilson, I. A.Steward, M. W., & Howard, C. R. 1987) Zmmunol. Today 8Stu ra, E. A., Feinstein, A., & Wilson, I. A. 1987) J Mo l. Biol.Suh, S . W., Bhat, T. N., Navia, M. A., Cohen, G. H., Rao, D.

N., Rudikoff, S.,& Davies, D. R. 1986) Proteins: Struct.,Funct., Genet. 1 74.Theriault, T. P., Leahy, D. J., Levitt, M., & McConnell, H. M .1 99 1) J . Mol . Biol. 221 257-270.Wiithrich, K., Billeter, M., & Braun, W. 1983) J Mo l. Biol.Zilber, B., Scherf, T., Levitt, M., & Anglister, J. 1990)

(1990) Science 248 7 12-7 19.51-58.193 229-23 1.

169 949-96 1.Biochemistry 29 10032.