10032

Biochemistry 1990, 29, 10032-10041

NMR-Derived Model for a Peptide-Antibody Complext?$

Barbara Zilber,s Tali Scherf, Michael Levitt," and Jacob A n g l i s t e r * ~ ~ ~ L Department of Polymer Research, The Weizmann Institute of Science, Rehovot 76100, Israel, and Department of Cell Biology, Stanford University School of Medicine, Stanford, California 94305 Received January 23, 1990; Revised Manuscript Received June 5, I990

ABSTRACT: The TE34 monoclonal antibody against cholera toxin peptide 3 (CTP3; VEVPGSQHIDSQKKA) was sequenced and investigated by two-dimensional transferred NOE difference spectroscopy and molecular modeling. The V, sequence of TE34, which does not bind cholera toxin, shares remarkable homology to that of T E 3 2 and TE33, which are both anti-CTP3 antibodies that bind the toxin. However, due to a shortened heavy chain CDR3, TE34 assumes a radically different combining site structure. The assignment of the combining site interactions to specific peptide residues was completed by use of AcIDSQRKA, a truncated peptide analogue in which lysine- 13 was substituted by arginine, specific deuteration of individual polypeptide chains of the antibody, and a computer model for the Fv fragment of TE34. NMR-derived distance restraints were then applied to the calculated model of the Fv to generate a three-dimensional structure of the T E 3 4 / C T P 3 complex. The combining site was found to be a very hydrophobic cavity composed of seven aromatic residues. Charged residues are found in the periphery of the combining site. The peptide residues H I D S Q K K A form a @turn inside the combining site. The contact area between the peptide and the T E 3 4 antibody is 388 A2, about half of the contact area observed in protein-gntibody complexes.

R e c e n t crystallographic studies of protein-antibody complexes have greatly contributed to our understanding of the recognition of protein antigens by antibodies (Amit et al., 1986; Sheriff et al., 1987; Padlan et al., 1989; Colman et al., 1987). The combining site for protein antigens differs in several respects from that for small ligand molecules. The contact area between antibodies and proteins is about 700 A. All antibody CDRs' participate in binding, and an average of 17 antibody residues interacts with 15 protein residues. These combining sites were found to be rather flat with protuberances and depressions that fit the shape of the antigen. In contrast, antibodies against small molecules were found to contain a pocket or groove with a ligand contact area much smaller than that observed in protein-antibody complexes. Despite interest in understanding the antigenic structure of peptides, complexes between peptide and antibodies have not been studied in detail by X-ray crystallography. The interactions of three antibodies (TE32, TE33, and TE34) with CTP3, a peptide corresponding in sequence to residues 50-64 of the B subunit of cholera toxin (VEVPGSQHIDSQKKA), have been investigated by twodimensional transferred N O E difference spectroscopy on Fab/CTP3 complexes (Levy et al., 1989; Anglister & Zilber, 1990). We found that when TE32 or TE33 interacts with excess CTP3, the exchange between bound and free peptide is fast relative to the spin-lattice relaxation time T I of the Fab
'This work was supported by grants from the Joseph and Ceil Mazer Center for Structural Bioloev. the Kimmelman Center for Biomolecular Structure and Assembly, and the US.-Israel Binational Science Foundation (88-00450). 'The nucleotide sequence data reported in this paper for the TE34, TE33, and TE32 antibodies will appear in EMBL, GenBank, and DDBJ Nucleotide Sequence Databases under Accession Numbers M30458M30459 and M30480-M30483 inclusive. * Correspondence should be addressed to this author. #The Weizmann Institute of Science. I'Stanford University School of Medicine. Incumbent of the Graham and Rhona Beck Career Development Chair.

and peptide protons, and therefore, magnetization transferred between Fab and neighboring peptide protons is further transferred to the free peptide. As a result, the NOESY spectrum of the Fab in the presence of a large peptide excess (excess spectrum) contains extra cross peaks due to interactions between the antibody and the bound peptide and the further exchange between bound and free peptide. These extra cross peaks, the so-called transferred NOE, are not present in the NOESY spectrum when the Fab and the peptide are in a 1:l ratio (saturated spectrum). Thus, by subtracting the saturated spectrum from the excess spectrum, we succeeded in eliminating the numerous cross peaks due to intra-Fab interactions, and the transferred N O E cross peaks due to the antibodypeptide interactions are clearly observed (Anglister et al., 1989). The assignment of the interacting peptide protons was accomplished on the basis of the COSY spectrum of the free peptide. To assign overlapping resonances from different peptide residues, we repeated the experiments with specifically deuterated peptide and followed the disappearance of cross peaks. The assignment to specific type of antibody aromatic amino acids was accomplished by specific deuteration of the antibody. This deuteration was accomplished by feeding
Abbreviations: Ac, acetyl; CDR, Complementarity Determining Region of the antibody molecule; COSY, 2D J-correlated spectroscopy; CTP3, Cholera Toxin Peptide 3; ELISA, enzyme-linked immunosorbent assay; Fab, antibody Fragment made of the Fv, the light chain, and the first heavy chain constant regions; FDR, Frame Determining Region of the antibody molecule, the segments that build the conserved structure of the immunoglobulin fold; Fv, antibody Fragment made of the variable regions of the light and heavy chains forming a single combining site for antigen; NCA, N-acetyl-CTP3 Amide; N M R , nuclear magnetic resonance; NOE, nuclear Overhauser effect; NOESY, 2D N O E spectroscopy; TRNOE, transferred NOE; V,, variable region of the heavy chain; ZD, two dimensional; 2D T R N O E difference spectrum, calculated 2D difference spectrum between the measured NOESY spectrum of the protein saturated with the ligand and that of the protein measured in the presence of a large excess of ligand. Name convention: (a) specific peptide residue, lower-case name (e.g., glutamine-7); (b) specific antibody residue, upper-case name ( e g , Tyr 32L).

0006-2960/90/0429- 10032$02.50/0

0 1990 American Chemical Society

Peptide-Antibody Complex deuterated amino acids to the hybridoma cells that produce the antibodies. While TE32 and TE33 bind cholera toxin in a solid-phase immunosorbent assay (ELISA), TE34 does not bind the toxin (Anglister et al., 1988). The off-rate of CTP3 from TE34 was too slow to measure strong T R N O E cross peaks between the antibody and the peptide (Anglister & Zilber, 1990). Much faster off-rates, resulting in strong TRNOEs, were obtained for two peptide analogues: (a) CTP3 with a C-terminal amide (Ac-VEVPGSQHIDSQKKA-NH,); (b) a truncated version of the peptide (N-acetyl-IDSQKKA). These modifications did not interfere significantly either with the interactions of the unmodified part of the peptide with the antibody or with intramolecular interactions occurring in the epitope recognized by the antibody (Anglister & Zilber, 1990). The combined application of these peptides allowed study of the interactions between the antibody and the whole peptide. Two tyrosine residues and one or more tryptophan and phenylalanine residues were found to interact with histidine-8, isoleucine-9, aspartate-]0, lysine-1 3 and/or lysine-14, and alanine-15 of the peptide. In the bound peptide, we observed interactions of a lysine residue with aspartate- 10 P-protons. While the peptide epitope recognized by TE34 is between histidine-8 and the negatively charged C-terminus, that recognized by both TE32 and TE33 includes residues 3-10 of CTP3. Measurements of the intensity of intramolecular magnetization transfer between a pair of neighboring protons after a short irradiation period of one of them or after a short mixing time in a NOESY experiment are used to evaluate the distance between the protons. Such measurements are extensively used to obtain restraints on proton-proton distances subsequently used to calculate the three-dimensional structure of small proteins (Wuthrich, 1986). Clore and Gronenborn (1983) and Clore et al. (1 986) extended the application of such measurements for transferred N O E in protein-ligand complexes. This analysis distinguishes four regions of chemical exchange. In the case of fast exchange between bound and free ligand, when the ligand off-rate is faster then 10 times the spin-lattice relaxation time of its protons and when the resonances of the bound and free ligand are averaged, the magnitude of the transferred NOE (NH,A) between an antibody proton, A, and a peptide ligand proton, H, is given by N H , A (1 - u)uH,AT, (Clore & Gronenborn, 1983; Clore et al., 1986; Meyer et al., 1988), where uH,A is the cross-relaxation rate between an antibody proton and a neighboring bound peptide proton and ( I is the mole fraction of the free peptide. This simple relationship applies to both one-dimensional and two-dimensional T R N O E experiments. The distance ratios between different pairs of protons in the complex can be determined from uH,A, which is inversely proportional to the sixth power of the distance between the protons. When the chemical shifts of the bound and free peptide resonances are not averaged but the off-rate is still faster than T , and cH,A and the peptide is present in large molar excess, the T R N O E intensity between an antibody proton and a free peptide proton is still approximated by the above equation. In the case of medium off-rate, the efficiency of the TRNOE will depend on the off-rate (Clore & Gronenborn, 1983), and the above equation for NH,A can serve only as an upper limit. The off-rate of Ac-VEVPGSQHIDSQKKA-NH2 bound to TE34 was found to be 480 s- at 37 OC (Anglister & Zilber, 1990). Although not every resonance represents the average of the bound and free peptide, this off-rate is very fast relative to the T , relaxation time and the cross-relaxation rates. Therefore, for a large peptide excess as used in our experiments, the above

Biochemistry,

VOI.29, No. 43, 1990 10033

equation for the T R N O E is expected to be a very good approximation. If on the other hand the exchange between bound and free ligand is slow on the cross-relaxation and spin-lattice relaxation scale, the transferred N O E intensity is vanishingly small. In this study we determined the amino acid sequence TE34 and assigned the specific peptide-antibody interactions to the specific residues involved. Proton-proton distances were evaluated by comparison of cross-peak intensities and subsequently used to model the three-dimensional structure of the CTP3/TE34 complex. This model provides insight into the mode of interaction between peptides and antibodies and contributes to our understanding of the antigenic structures of peptides. MATERIALS AND METHODS The Fab labeling and sample preparation and the procedure for NOESY measurements were previously described (Anglister et al., 1989). Peptide synthesis and purification and a more detailed description of the 2D difference spectra calculations were also given (Anglister & Zilber, 1990). NMR measurements with the NCA peptide were carried out at 37 C and pH 7.15, a 10-fold peptide excess being used. With Ac-IDSQRKA the experiment was run at 42 OC with a 2-fold peptide excess. The binding constants of TE34 for AcIDSQRKA and for Ac-IDSQKQA were measured at room temperature by following antibody fluorescence quenching as described previously (Anglister & Zilber, 1990). p H Titration. The TE34 Fab solution in the presence of a 4-fold excess of the CTP3 peptide was titrated between pH 4.00 and pH 9.70. Tryptophan and phenylalanine residues of the antibody were perdeuterated while tyrosines were deuterated at the C6 positions. Fab concentration was 2 mM. Spectra were measured at 37 OC; 256 scans were collected for each titration point with a spectral width of 6000 Hz and 3-s presaturation of the H D O line with minimal power (55L on the AM 500). The individual resonances of the histidine imidazole protons of the bound and free peptide were clearly observed in the difference spectra calculated for different pH values. pK, values were determined from plots of the derivative of the titration curves versus the pH. mRNA Sequencing. The mRNA sequencing was carried out as previously described (Levy et al., 1989). The universal primers for IgG1 heavy chains and K light chains as defined by Griffiths and Milstein (1985) were used to initiate sequencing. Two additional primers were utilized for both the light chain, 5-CTC CAC TCT GCT GAT-3 and 5-AGA CTG GCC TGG CCT CTG-3, and the heavy chain, 5-AGA GGT TTC CAA AGA-3 and 5-ACC CTT TCC TGG AGC CTG-3. Chain-Specific Labeling. Two Fab fragments were prepared: Fab[H(l)L( l)] was unlabeled, while in Fab[F(2)L(2)] all tryptophan, phenylalanine, and tyrosine residues were predeuterated. Both Fabs were reduced and alkylated, and their chains were separated under denaturing conditions on a Bio-Gel DEAE ion-exchange column at pH 8.3 in 8 M urea and 50 mM glycineTris buffer without salt. One of the chains bound to the column and was eluted with the above buffer containing 300 mM sodium chloride. Since the heavy chain fragment and the whole light chain that constitute the Fab share similar molecular weights, they cannot be distinguished on this basis. To differentiate between the separated chains, whole antibody reduced and alkylated was loaded under the same denaturing conditions. Again, a single chain was bound in the absence of salt and later eluted by salt addition. Polyacrylamide gel electrophoresis identified the bound and

10034 Biochemistry, Vol. 29, No. 43, 1990 nonbound polypeptides as the heavy and light chains, respectively. The solutions of the separated chains were adjusted to pH 7.5 with phosphoric acid and cross-recombined to produce L( 1)H(2)- and L(2)H( I)-labeled Fabs. Reconstitution of the native conformation of the Fab was accomplished at 4 C by extensive dialysis against 10 mM phosphate buffer containing 150 mM sodium chloride, pH 7.5. The reconstituted Fab was then concentrated and dialyzed extensively against phosphate-buffered deuterium oxide, pH 7.15. This procedure significantly improved the yield of reconstituted Fab relative to that previously described (Levy et al., 1989). Molecular Modeling. The method used to build a detailed all-atom model of the dimer formed from the light and heavy chain variable domains (Fv) of the TE34 antibody paralleled that used for TE32 and TE33 previously (Levy et al., 1989). I n the previous paper, we described the modeling method in some detail; here we add some additional details. Two computer programs, ABMOD and ENCAD, are used to calculate atomic coordinates for TE34. ABMOD finds segments of known antibodies that can be used to provide initial coordinates for TE34. ENCAD uses energy minimization to refine the coordinates. Before different antibody structures can be pieced together with ABMOD and ENCAD, all the structures must be in the same coordinate system. This is achieved by use of rigid-body rotation to superimpose all known antibody coordinates onto the structure of the KOL antibody (Marquart et al., 1980), which is taken as a standard. Only the a-carbon coordinates of the well-conserved framework are used for this superposition. ABMOD works by first aligning sequences by a modified method that prevents gaps occurring in regions of secondary structure (Lesk et ai., 1986) and then by using the atomic coordinates of the known antibody to ensure that chain gaps are between residues that are close in space. Both methods have been described before (Levy et al., 1989), and the final alignment of TE34 to McPC 603 and HyHel-5, the two proteins that ABMOD chose as the best sources of initial coordinates, is shown in Figure 2. Initial coordinates are simply copied over from the X-ray structures of McPC 603 and HyHel-5 (after rigid-body rotation as described above). These coordinates suffer from two sources of stereochemical errors. (a) There are missing atoms when there is a longer side chain in the unknown antibody structure than in the known structure from which it is built (for example, residue 7 of TE34 is Thr that is modeled from a Ser in McPC 603; see Figure 2). (b) There are gaps in the chain when the unknown structure has a deletion (for example, at residue 31A; see Figure 2). ENCAD, a robust energy minimization program, is used to correct both these defects and produce a low-energy structure containing all atoms including hydrogen, which is so important for NMR. This program, which has been used in the past to simulate the molecular dynamics of protein in solution (Levitt & Sharon, 1988), uses a detailed all-atom potential energy function, with terms for bond stretching, bond angle bending, bond twisting, van der Waals interactions, and electrostatic forces. Energy parameters and numerical methods are given in Levitt (1983). Refinement of the initial TE34 model first involved building a model with only polar hydrogen atoms; the energy of the minimized structure is -1 593 kcal/mol, which is comparable to that found for TE32 and TE33 (Levy et al., 1989). This minimization was done in vacuo and the effect of the missing solvent compensated for by neutralizing ionized side chains (Glu, Asp, Lys, and Arg) and by holding the C-a coordinates

Zilber et al. to the initial X-ray position by a weak restraint. Once a low-energy model was obtained with only the polar hydrogen, all other hydrogen atoms were added by ENCAD. The same energy minimization method was used to model the peptide-antibody interactions. The total potential energy ( v ) was modified to include a term that restrains the distances between protons observed to interact by NOE. The form of this term is

U,,

= C k ( r i j- rij)4kcal/mol if ru 2 rijC = 0 if r . . < r.?

IJ

where k is used as a force constant in order to translate the NOE restraints into energy values in the energy minimization calculations. It was necessary to use a fourth power term to ensure continuous first and second derivatives at rij = r;. This restraint keeps the distance between atoms i andj, rii, less than the distances deduced from the N O E value, rij. The initial model of the peptide was built by hand with the interactive modeling program FRODO (Jones, 1978). Because of the small size of the peptide and the strong restraint imposed by NOE interaction to the rigid antibody binding site, no attempt was made to explore the full range of peptide conformation that could fit the site. RESULTS Nucleic and Amino Acid Sequences. The mRNA coding for TE34 was sequenced directly. In Figure 1 the rearranged nucleotide sequences of the light and heavy chains of TE34 are compared with those of TE32 and TE33. The amino acid sequence of TE34 is given in Figure 2, again compared with those of TE32 and TE33. The aromatic residues in the light chain CDRs are Tyr 32L, His 93L, Phe 94L, and Trp 96L. The TE34 light chain sequence is typical of group I1 K light chains (Kabat et al., 1987). TE34 uses a MUSJKl J-gene while TE32 and TE33 use the MUSJK4 J-gene. The sequences of the light chain variable domains of several other antibodies are very similar to that of TE34: L-1210 and 36-60 CRI-, which are anti-azophenylarsonate antibodies (Juszczak et al., 1984) with 96.5% and 95% amino acid identity, respectively; BXW- 16, an anti-DNA autoantibody (Kofler et al., 1988) with 96.6% identity; 40-150, an anti-digoxin antibody (Panka et al., 1988) with 95% identity; antidextran antibodies (Akolkar et al., 1987) with 95% identity. The amino acid sequences of TE32 and TE33 are 74% and 75% identical, respectively, with that of TE34. It should be noted that Tyr 32L, although being a CDR residue, is conserved in antibodies possessing K light chain from group I or group 11. The aromatic residues in the CDRs of the TE34 heavy chain are Tyr 27H, Phe 29H, Tyr 32H, Trp 50H, Tyr 53H, and Phe 100cH. The heavy chain sequence coded by the V, gene in TE34 is 86% and 89% identical with the corresponding sequences of TE33 and TE32, respectively (Levy et al., 1989). This indicates that all three anti- CTP3 antibodies use V genes of the same family within group IX heavy chain genes. However, the CDR3 of the TE34 heavy chain is only four residues long, three residues shorter than that of TE32 and TE33. The sequence coded by the VH gene is very homologous with sequences of other antibodies of family IX: DB3, an anti-progesterone antibody (Winter et al., 1985) with 85% identity; L6, an anti-carcinoma-antigen antibody (Liu et al., 1987) with 89% identity; U22B5, an anti-arsonate antibody (Meek et al., 1984), with 84% identity. A11 aromatic residues in the part coded by the VH gene are conserved in those antibodies which are assumed to be coded by genes of the same family. TE34 uses the MUSJHZ J-gene in contrast to TE32

Peptide-Antibody Complex
UGET C H A p l VARIABLE W I O N
1 2 3 4 5 6 7 8 TE-34 GAT G T T GTG ATG ATC CAG ACT CCA CTC ACT T T G TE-33 T C T C C TE-32 T C T C C
8 3 4 S

Biochemistry, Vol. 29, No. 43, 1990 10035

9 M l l ~ ~ l 4 ~ I TCG G T T ACC A T T GGA C M CCA CCC TCC ATC T C T TGC AAG C T C Gl'C G T A AA C T C Gl'C G T A GA
3 6 3 7 1 8 4 l U 4 Q U 4 1 4 7 1

TE-33 TE-32

TE.34 CTG GAC T C T GGA GTC CCT ACC ADD T T C ACT GGC A C T GGA TCA GGG ACA GAT T T C ACA CTg AAA ATC AGC AGA GTG GAG GCT GAG GAT A GA G C G TE-33 A GA G C G TE.32 GA T T T s s p ~ ~ f f i m a m g TE-34 ggs GGA GTt T A T T A T TGC TGG C M TE-33 CTG C m TE.32 CTG C m l
B] s 88 ACG T T C GGT C A A CTC C

z] z
m m i

G T C G A C

AG

T T G AAT TGG T T G T T A CAD AGG CCA GGC CAD T C T CCA AAG CGC CTA ATC T G T C AC C G M C G A T G A T GA A G A AC C G

'

i m io1 lop i w GGA GGC ACC AAG TCG G A TCG G A

IDL io5 io8 107 ice i o CTG G M ATA AAA CGG GCT T T

HEAVY GRAIN VARIABLE REGION

TE-32

3 3 3 4 5 3 3 3 7 3 3 8 a 4 l 4 0 4 4 E . 1 4 7 4 8

m
G G T T

m
A A

TE-33 A A AC TE-32 T T AC

A A

A G

C A

m m m 61 a e sp %I 87 a o m n n n 74 m m n ia ' ~ d 80 m 82 ~ Z A ~ B B I C TE-34 GCA T A T G T T GAC GAC T T C AGO GOA CGG T T T GCC T T C T C T T T G GAA ACC T C T GCC AGC ACT GCC T A T T T G CAG ATC AAC AAC CTC AAC TE-33 A T C T A A C C T A A TE-32 A

a t x f f i SI m BI 90 91 8 p TE-34 A A T GAG GAC ACG GCT ACA T A T T T C T G T TE-33 TE-32

s a
T GT

1 W 1DL TGG GGX C A

105 lo8 lU7 106 lo8 110 111 112 113 AA GGC ACC ACT CTC ACA GTC TCC TCA G ACA G G G c ACA G G G c

FIGURE 1: Deoxyribonucleotide sequence of the TE34 light and heavy chain variable regions as compared to those previously determined for the TE33 and TE32 antibodies. The boxes delineate boundaries of the CDRs. The full sequence is given for TE34, whereas only changes with respect to TE34 are listed for TE32 and TE33. Lower-case letters where they appear indicate some ambiguity in the sequence determination; X appears where we could not determine the nucleotide.

and TE33 which use the MUSJH1 J-gene. The third CDR varies considerably in both length and sequence in all the above antibodies. Assignment of Intra- and Intramolecular Interactions to a Specific Lysine Residue. Two tyrosine residues and at least one phenylalanine residue from TE34 interact with protons of the two lysine residues of CTP3 (Anglister & Zilber, 1990). There is also an intramolecular interaction in the bound peptide between a lysine residue and aspartate-10. To assign these interactions to a specific lysine residue, two CTP3 analogues were prepared: Ac-IDSQRKA and Ac-IDSQKQA. These truncated peptides correspond to residues 9-1 5 of CTP3, and for convenience, we maintain the same residue numbering. Ac-IDSQRKA, in which lysine-1 3 was substituted by arginine, binds to TE34 with an affinity constant of 2 X lo5 M-I in comparison to K, = 2.5 X lo5 M-' measured for Ac-IDSQKKA. This similarity in binding indicates that substitution of lysine- 13 by arginine does not significantly interfere with the peptide-antibody interactions. The binding of the peptide Ac-IDSQKQA in which lysine-14 was replaced by glutamine was too weak to be measured (less than lo3 M-I). Thus, it appears that the positive charge on lysine-14 contributes significantly to the binding constant. Figure 3 presents the 2D T R N O E difference spectrum showing interactions between nonaromatic protons of the TE34 Fab/Ac-IDSQRKA complex. The assignment of cross peaks in this spectrum is based on chemical shift identity with those of the free peptide protons, the COSY spectrum of the peptide, and their observed multiplicity. The resonance at 3.02 ppm has the same chemical shift for the C, protons of lysine as the unmodified peptide and therefore is assigned to lysine-14. The

resonance at 3.22 ppm, which does not appear in the difference spectrum for the TE34/Ac-IDSQKKA complex (Anglister & Zilber, 1990), is assigned to the C6 protons of arginine-13, and the cross peaks at (3.22, 1.87), (3.22, 1-78),and (3.22, 1.64) ppm are due to intraresidue interactions with the 6- and yprotons, respectively. The weak cross peaks at (3.02, 2.66) and (3.02, 2.71) ppm are assigned to intra-bound-peptide interactions between lysine- 14 t-protons and aspartate- 10 P-protons. The strong cross peaks at (3.02, 2.56) and (3.02, 2.40) ppm are due either to interactions of peptide lysine-14 with the antibody or to exchange cross peaks. The strong cross peak at (3.22, 2.81) ppm is similar to that previously observed at (3.01, 2.81) ppm for TE34 interacting with AcIDSQKKA. It is assigned to the arginine- 13 &protons and is due to exchange or interaction with antibody protons. Additional cross peaks which are observed in the TRNOE difference spectrum are due to interresidue interactions of the C,H of serine-1 1 (4.38 ppm) with aspartate-10 C,H (4.62 ppm) and C,H (2.66 and 2.71 ppm) and with isoleucine-9 C,H (4.14 ppm). The interresidue interactions in the bound peptide are indicative of its conformation and provide distance restraints that are input into our modeling procedure. NOESY measurements with the modified peptide (K13 R) proved crucial in assigning the interactions of TE34 phenylalanine and tyrosine residues to a specific lysine residue of the peptide. Figure 4A shows the interactions of phenylalanine residues of the antibody with the Ac-IDSQRKA peptide. Several cross peaks are unambiguously assigned to their interacting peptide protons. The cross peaks at (3.20, 6.63), (1.47, 6.38), and (1.32, 6.58) ppm are assigned respectively to the interactions of arginine-1 3 (C,H), lysine-14

10036 Biochemistry, Vol. 29, No. 43, 1990

Light Chain V 6 h b l e Region
10

Zilber et al.
[------------CDR1------------j 3031A B C D E F
[---CDp.2.-.,
40

20

50

TE34 TE33 TE32 McPC HYS

D V V M I O T P L T L S V T I G O P A S I S C K S S O S L L D - S D G K T ~ L N W L L O R P G O S P K R L I Y L V I V H S N F E Y K L K L T S P S L D O L T S P S L D O R I V H S N E Y T L K L A W Y O O K P G O P P K L L I L G A L l U U S Q s L L N 5 G N 0 K N I L T S A I M A S P E K V T M T S A S - - - - - - - V N M Y Y O K S T W I ) T S K
NO

S N N T I

K L D S R F R F R E . A S

51

SA

52

L1 [----CDR3---]

53

54

L2

60

70

a0

90

100

TE34 TE33 TE32 McPC W 5

G V P T R F D D E V P D R F

T G S G S G T D F T L K I S R V E A E D g G V Y Y C W O G T E ~ P U T F G G G T K L E I K S L F S I F S S L F S V T L S T G S G S G T D r T L T I S S - u u s S Y S T S M T A A E 0 W G R N 57 50

56

58

L3

59

Heavy Chain V a r i a b l e Region

10 20

[-----CDRI-----,
30

40

[-------CDRZ------, 50 5 2 A 0 53

55 A 5 6

TE34 TE33 TE32 McPC HYS

O I O l V O S G P E L R K P G E T V K I S C K A S G Y T ~ T D ~ L K T R T L K S T S F Y E V K E G G V O G S L R L A T F --AS V K 1 S C K A S G U - - P
51
SA

G M N W V K O A P G K G L K W M G ~ I N T - Y S T F - S S - S E R P R E I A A S R N K G N K Y G S G -

T G - O P A V T V T T T E S T N

52

HI

53

54

H2

HB

[----------CDR)---------,

60

TE34 TE33 TE32 McPC Hy5

Y V D D F A A S A S V y H E R F

10 a c a 2 A B c a3 90 R G R F A F S L E T S A S T A Y L O I N N L N N E D T A T Y F K K K K I V R D 0 I L M A R A I Y K G K A T F T A D T S S S u U U

100 A B C

110

C V R P A R S A R S A N Y Y
N
H3

86

57

HE

sa

---

W W W Y

- - ~ D S W G q G T T L T V S S Y V T V Y V T V Y E V A V

DGWGOGTTLT
59

FIGURE 2: Comparison of the amino acid sequence of the variable regions of the heavy and light chains of the TE34 antibody with the known sequences for TE32 and TE33. The full sequence is given for TE34, whereas only changes with respect to TE34 are listed for TE32 and TE33. The aromatic residues in the CDRs of TE34 are typeset in boldface, while the subset of these residues found in this study to interact with the peptide antigen is outlined. Also listed are the sequences of both McPC 603 and HyHel-5, whose three-dimensional structures were used to calculate the structure of TE34 (Segal et al., 1974; Sheriff et al., 1987). McPC is a myeloma antibody that binds phosphocholine while HyHel-5 is an anti-lysozyme monoclonal antibody. The particular sequences used to provide initial coordinates for segments of the TE34 model are underlined. Each segment used for modeling is marked by a two-letter segment name placed at the start of that particular segment. Lower-case letters indicate places of ambiguity in sequence determination.

IO

20

'

a a
30
t
-9

-@
a
70

I
64

L ? '

I
76
70
E4

I
76 70 P P M

61

76

40

5'
40

23
PPM

IO

8Pi --,t

30

FIGURE 3: Difference between the NOESY spectrum of TE34 Fab with a 2-fold excess of acetyl-IDSQRKA and that of the peptidesaturated Fab. The spectrum contains cross peaks due to intramolecular interactions in the bound peptide and possibly intermolecular interactions between the peptide and the antibody. Interacting peptide protons are marked by small letters and numbers denoting their location in the peptide sequences.

2D TRNOE difference spectrum showing interactions of aromatic protons of the antibody (TE34) with Ac-IDSQRKA and with Ac-VEVPGSQHIDSQKKA-NH2(NCA). (A) Interactions of antibody phenylalanine with Ac-IDSQRKA. Antibody tyrosine and tryptophan are perdeuterated while phenylalanine is unlabeled. (B) Interactions of C,H of TE34 tyrosine residues with Ac-IDSQRKA. Antibody tryptophan and phenylalanine are perdeuterated while tyrosine residues are deuterated at the Cs positions. (C) Interactions of tyrosine C8H and CQHof the antibody with the NCA peptide. Antibody tryptophan and phenylalanine residues are perdeuterated while tyrosine residues are unlabeled. Assigned peptide residues are marked by small letters and their location in the sequence; antibody residues are marked by capital letters denoting amino acid type and arbitrary numbers.

FIGURE 4:

(C,H), and alanine-1 5 (CH,) with antibody phenylalanine. The strong cross peaks a t (1.72,6.39), (1.72,6.60), and (1.72, 6.98) ppm are assigned to interactions of lysine-14 &protons with antibody phenylalanine protons. The cross peaks at (1.89, 6-38), (1.84, 6.58), and (1.84, 6.98) ppm could be assigned to interactions of P-protons of either arginine- 13 or lysine- 14

with antibody phenylalanine. Figure 4B shows the interaction of C,1 and C,2 protons of antibody tyrosine with the AcIDSQRKA peptide. One tyrosine residue ( Y I ) interacts with arginine- 13 P- and &protons while a second tyrosine (Y2) interacts with lysine-14 0, y, and 6-protons. The arginine and lysine cross peaks were assigned on the basis of the COSY

Peptide-Antibody Complex
t !
w

Biochemistry, Vol. 29, No. 43, 1990

IO
F,ol5

10037

20
-Yl,kI3

60

70

70

60

70
PPM

60

70

60

F I G U R E 5 : 2D TRNOE difference spectra obtained after specific chain labeling: (A) interactions of the light chain aromatic residues with the NCA peptide; (B) interactions of heavy chain aromatic residues with the NCA peptide; (C) interactions of aromatic residues from both chains with the NCA peptide. Assigned antibody residues are marked by capital letters and arbitrary numbers; peptide protons are marked by small letters and numbers denoting the location in the sequence of the residues to which the protons belong.

spectrum of the modified peptide with and without Fab. The cross peak at (4.15, 6.55) ppm is assigned to interaction of Y1 and alanine-] 5 C,H. interactions of C ,Protons of Antibody Tyrosine Residues with the NCA Peptide. To observe the interactions of C6 as well as C, protons of antibody tyrosine with the N C A peptide (acetyl-VEVPGSQHIDSQKKA-NH,), which contains the complete epitope recognized by TE34, we prepared Fab in which all tryptophan and phenylalanine residues were perdeuterated while tyrosine was unlabeled. A section of the 2D T R N O E difference spectrum obtained for the labeled Fab in complex with NCA is presented in Figure 4C. The interactions assigned to lysine- 14 appear in both panels B and C of Figure 4; however, interactions with arginine-13 in Figure 4B are replaced by interactions with lysine-I3 in Figure 4C. The cross peaks in Figure 4C manifest much stronger intensities due to the much faster off-rate of N C A relative to that of the truncated peptide which contains a negatively charged C-terminus. Figure 4C also illustrates the chemical shift degeneracy of both lysine resonances in the N C A peptide. The assignment of cross peaks to tyrosine Cbprotons is based on their absence in the difference spectrum obtained for the Fab in which tryptophan and phenylalanine are perdeuterated while tyrosine residues are deuterated at the C6 positions (Anglister & Zilber, 1990). The assignment of the peptide protons involved was previously reported. The cross peaks a t (4.70, 7.52) and (4.70, 7.03) ppm are assigned to the interaction of the C,H of the peptide histidine-8 with C,H and C,H, respectively, of tyrosine (Y2). The assignment of these two cross peaks to a single tyrosine residue relies on the specific chain labeling and the calculated model for the antibody Fv (see below). Weak interactions of Y1 and Y2 with the C,H of lysine-I 3 and lysine-I4 are observed, (2.99, 6.60) and (2.99, 7.02) ppm, respectively. The strong cross peak a t (4.32, 6.60) ppm was previously assigned to the interaction of Y 1 with a peptide lysine (Anglister & Zilber, 1990). On the basis of the observed interaction between Y1 and alanine-I5 C,H in Figure 4B, we conclude that the cross peak a t (4.32, 6.60) ppm in Figure 4C is actually due to overlap of two cross peaks, one due to interaction with lysine-14 and one due to interaction with alanine-15. The C-terminal alanine proton has a chemical

shift of 4.15 ppm in Ac-IDSQRKA and 4.33 ppm in AcVEVPGSQHIDSQKKA-NH, due to the conversion of the C-terminal into an amide. The weak cross peaks a t (1.76, 6.73), (1.44, 6.73), (1.44, 6.96), (1.76, 6.96), and (1.44, 7.10) ppm are due to interactions of tyrosine C6 protons of Y 1 and Y2 and/or weak interactions of a third antibody tyrosine residue with the peptide. Assignment of Interactions to a Specific Antibody Polypeptide Chain. In order to assign the peptide-antibody interactions to the specific polypeptide chain involved, two specifically labeled recombined Fab fragments were prepared. In Fab[L( l),H(2)], tryptophan, tyrosine, and phenylalanine residues of the heavy chain are perdeuterated while the aromatic residues of the light chain are unlabeled. In Fab[L(2),H(l)], the aromatic residues of the light chain are perdeuterated while those of the heavy chain are unlabeled. The interactions of the light and heavy chain aromatic residues with the N C A peptide are shown in panels A and B of Figure 5, respectively. Inspection of Figure 5A,B reveals a single phenylalanine and a t least one tyrosine residue of the light chain interacting with the peptide, with the heavy chain contributing one each of tyrosine, phenylalanine, and tryptophan. In Figure 5A the cross peaks a t (1.44, 6.97), (1.76, 6.97), (1.44, 7.07), (1.76, 7.07) ppm were not assigned. These cross peaks appear also in Figure 4B and were attributed to interactions of C, protons of Y1 and Y2 and/or interactions of protons of a third antibody tyrosine with the peptide. According to Figure 5A these protons are of a light chain tyrosine. The sum of the two T R N O E difference spectra in Figure 5A,B is in excellent agreement with the spectrum previously obtained for the native unlabeled Fab (Figure 5C). A Calculated Model f o r the Fv. The three-dimensional structure of the Fv of TE34 was modeled on the basis of the known conformations of the other antibodies. The different segments used in the modeling procedure are given in Figure 2. The conformations of the TE34 segments L1, L2, and H1 were found to be very similar to those of the corresponding segments in TE32 and TE33. The L3 segment in TE34 and in both TE32 and TE33 is similar in sequence and conformation between residues 91 and 93 but differs between residues 94 and 96. The conformation of H 2 (residues 52a-55) in the

10038 Biochemistry, Vol. 29, No. 43, 1990 model of TE34 differs greatly from that of H 2 in TE32 and TE33. This difference probably arises from the application of the HyHel-5 antibody coordinates for TE34 modeling (Padlan et ai., 1989), as opposed to the use of 5539 (Suh et al., 1986) in the modeling of both TE32 and TE33. The H2 segment has the sequences TYSG in TE32 and TE33, PDSG in 5539, TYTG in TE34, and PGSG in HyHel-5. According to recent analysis of the conformations of antibody hypervariable loops (Chothia et al. 1989), Hyhel-5 is more suitable for modeling H2 in TE32, TE33, and TE34. In TE34 the H2 segment is located on the periphery of the combining site and therefore does not affect the structure of its central part. The second CDR of the TE34 heavy chain includes residues 50-52 as framework residues. In contrast to the differences noted for residues 52a-55, the main-chain conformation for residues 50-52 and the side-chain conformations for 50 and 51 are very similar in TE32, TE33, and TE34 and also to the conformation observed in other antibodies. Interestingly, residue Trp 50H occupies a central position in the combining site of all three anti bodies. The H 3 segment, which is part of the third C D R of the heavy chain, contains only one residue in TE34 and four residues in both TE32 and TE33. This truncation totally alters the combining site structure of TE34. In TE32 and TE33 the sequence Ser 96, Trp 1OOa, Tyr 1OOb, not found in TE34, is at the tip of the loop created by CDR3 of the heavy chain, forming a protuberance in the center of the combining site. The empty space formed by shortened CDR3 in TE34 creates a pocket in the center of the combining site, the walls of which are composed of a continuum of aromatic side chains and one hydrophobic nonaromatic residue. All of these, Tyr 32L, Tyr 49L, Leu SOL, Phe 94L, Trp 96L, Tyr 32H, Trp 50H, and Phe IOOcH, are exposed to the solvent. In contrast, the periphery of the combining site is composed mostly of nonaromatic residues. On the basis of our calculated model and the assignment of the N O E interactions to their specific polypeptide chain, we are able to assign peptide-antibody interactions to their specific residues in the TE34 sequence: Tyr 32L interacts with alanine- 15 and lysine- 13; Phe 94L interacts with histidine-8, lysine-I 3, and lysine-14; Tyr 32H interacts with lysine-14 and histidine-8; Trp 50H interacts with histidine-8, isoleucine-9, and aspartate-IO; Phe IOOcH interacts with histidine-8 and alanine-I 5. Tyr 49L is in the periphery of the combining site and could in principle interact with alanine-1 5 and lysine-14. It is possible that the unassigned cross peaks at (1.44, 6.97), (1.76, 6.97), (1.44, 7.07), (1.76, 7.07), and (1.86, 7.07) ppm observed in Figures 4B and 5A are due to interactions of Tyr 49L. Derivation of Restraints on Intermolecular Proton-Proton Distances. To translate cross-peak intensities into restraints on proton-proton distances, one needs a calibration usually obtained by measuring the intensity of a cross peak between a pair of proton separated by a known and fixed distance. To obtain such calibration, we measured the difference spectrum between the NOESY spectrum of Fab in the presence of IO-fold peptide excess and the NOESY spectrum of the uncomplexed Fab (without spectrum). All tryptophan and phenylalanine residues of the antibody were deuterated while tyrosine residues were unlabeled. The aromatic section of the difference spectrum is shown in Figure 6. In addition to the transferred N O E cross peaks, this difference spectrum shows the cross peaks due to intermolecular interactions of antibody protons that changed their chemical shift upon peptide binding. In the section of the NOESY difference spectrum showing

Zilber et al.

YI

8
-

70

70
PPM
FIGURE6:

60

Aromatic section of the 2D difference spectrum between the NOESY spectra of the TE34 Fab with a 10-fold excess of NCA and the uncomplexed Fab. Antibody tryptophan and phenylalanine residues are perdeuterated. The cross peaks marked Y 1 and Y are due to intraresidue interactions between the tyrosine C, and the c 6 protons; h8 are the exchange cross peaks of the peptide histidine C6,H. interactions between aromatic protons, one observes cross peaks due to intraresidue interactions between the tyrosine e,- and e,-protons and the 6,- and d2-protons. These cross peaks are not observed when the C6protons of the antibody tyrosine residues are deuterated. A cross peak a t (5.76,6.68) ppm with its symmetrical counter cross peak is very well resolved, and another pair of cross peaks, (6.85, 7.48) and (6,857.54) ppm, is the result of overlap between two tyrosine residues. The other cross peaks due to intra-tyrosine interactions are too close to the diagonal or overlap the resonances of the CB,Hof the bound and free peptide histidine. The intensity of the resolved cross peaks is then compared to the intensity of the exchange cross peak of Cs,H of the peptide histidine. Since this cross peak appears in all the T R N O E difference spectra and all of them are measured at the same mixing time and with the same molar ratio of peptide excess, it can serve as an internal standard for calibrating the intensities in all TRNOE difference spectra. Through this internal standard one gets the ratio between cross-peak intensities in the T R N O E difference spectrum and the cross peak due to intra-tyrosine interactions and subsequently the ratio between the distances of the corresponding pairs of protons. This analysis indicates that all the cross peaks due to intermolecular interactions observed in the TRNOE difference spectra are due to strong interactions between protons that are less than 3.5 8, apart. Measurements of the T R N O E difference spectra for the TE33 antibody a t mixing times of 60 and 100 ms exclude the possibility that the cross peaks observed in the T R N O E difference spectra at 100 ms are due to spin diffusion. Docking of the Peptide into the Antibody Combining Site. The list of the distance restraints used in the modeling of the TE34/CTP3 complex is given in Table I. C, was used as a pseudoatom for tyrosine C,H and CBHprotons. Since as shown in Figure 5B three tryptophan protons are involved in interactions with the peptide, we expect C, and C,, to be two of these protons. For phenylalanine interactions C, was chosen as a replacement for the pseudoatom. Each residue of the epitope was independently docked into the combining site to conform approximately to the N M R restraints while reasonable lengths for peptide bonds were maintained. The model for the complex was then refined under 35 N M R restraints, which are given in Table I. Initially, only the peptide was

Peptide-Antibody Complex

Biochemistry, Vol. 29, No. 43, 1990 10039

FIGURE 7: Structure of the TE34 complex with CTP3 (residues 8-15) obtained after energy minimization and on the basis of a calculated model for antibody Fv with NMR restraints: light blue, Trp 50H; light green, Tyr 32L, Tyr 49L, and Tyr 32H (top to bottom); pink, Phe 94L and Phe lOOcH (left to right); red, Lys 31 eL; violet, Asp 31L and Asp 31cL. Atoms of residues 8-1 5 of CTP3 are represented by balls: orange, oxygen; light blue, nitrogen; yellow, carbon.

allowed to move, whereas in a further refinement the antibody was allowed to move as well under a weak force field acting to maintain its original conformation. The root-mean-square (rms) deviation between distances in the final model (Figure 7) and the NMR restraints is 0.83 A. The deviations between the model and N M R distances are greatest for interactions involving Trp 50H (1.25 A) and for intrapeptide interactions (1.01 A). When these distances are excluded, the rms deviations for the remaining restraints drops to 0.58 A. When the force constants attributed to the NOE restraints are increased by a factor of 5, the rms deviation between the restraints and the model drops to 0.60 A. Decreasing or increasing this force constant effects the orientation of Tyr 32L, Phe 94L, and Trp 50H; however, these changes have a very minor effect on the calculated conformation of the peptide. The large violations could be explained by possible inaccuracies in the orientation of the side chains of the residues in the CDRs and the relative position of the hypervariable loops (Chothia et al., 1989) and especially of CDR2 of the heavy chain. These inaccuracies are only partially rectified by the relatively small number of restraints obtained by the NOESY experiments. It should be noted that although the interactions of Tyr 49L were not included in our restraints, in the calculated model there are interproton distances between Tyr 49L and alanine- I5 methyl protons and lysine- 14 p- and &protons which are less than 5

A.

DISCUSSION The model for the combining site structure of the TE34/ CTP3 complex, obtained by applying NMR-derived interproton distance restraints to the calculated Fv model, is shown in Figure 7. Only peptide residues 8-1 5 are shown since they represent the full peptide epitope. This segment of the peptide is highly polar, containing five charged residues: histidine-8, aspartate- IO, lysine- 13, lysine- 14, and alanine- 15 at the Cterminus. Despite the polarity of the peptide antigen, most of the antibody surface that it contacts is hydrophobic. The binding of the two molecules is made possible by the special folding of the bound peptide: (a) Formation of a @turn manifested by a hydrogen bond between the isoleucine-9 oxygen and glutamine-12 NH. The formation of a @-turnby the peptide segment IDSQ increases the number of residues

interacting with the central part of the antibody combining site, thus potentially increasing its binding energy. (b) Interactions with the aromatic residues of the antibody primarily via the peptide's backbone and @-protons. (c) Intramolecular electrostatic interactions in the bound peptide between the side chains of aspartate- 10 and lysine- 14, which orient their charges away from the hydrophobic surface of the antibody. There are also additional electrostatic interactions between the peptide and charged residues at the periphery of the combining site: (a) The negatively charged C-terminal carboxylate interacts with the e-NH3+of Lys 31eL. This interaction contributes significantly to the binding energy of CTP3 as manifested by the 2 orders of magnitude decrease in its binding constant when this carboxylate is converted into an amide (Anglister & Zilber, 1990). (b) According to the calculated model, the eNH3+of lysine-13 interacts with the negatively charged y-carboxyl groups of Asp 3 1L and Asp 3 1cL. Lysine-13 is exposed to the solvent and projects toward the periphery of the combining site. Its p- and &protons interact with Tyr 32L and Phe 94L. There are also polar interactions between the t-NH3+ of lysine-14 and the Tyr 32H 0,. Histidine-8 interacts with Phe 100cH, Tyr 32H, and Trp 50H; however, the nitrogen atoms of its imidazole ring are exposed to the solvent. The pKvalues of the bound and free peptide histidine (pK = 6.5) are practically indistinguishable, providing support for the partial exposure to the solvent of histidine-8. The side chains of isoleucine-9, serine- 1 1, and glutamine- 12 project toward the solvent; however, the a- and ,&protons of isoleucine-9 interact with the antibody. The total area of contact between the TE34 antibody and residues 8-1 5 of CTP3 was calculated to be 388 A2. In the models of the three anti-CTP3 antibodies investigated, the surface of the antibody in contact with the peptide consists mainly of aromatic side chains, with five in both TE32 and TE34 and six in TE33. Interestingly, Tyr32H, Tyr32L, and Trp5OH are found to interact with CTP3 in all three antibodies; however, they interact with different peptide residues in the TE34/CTP3 complex. Moreover, these aromatic residues are conserved in antibodies coded by the same V, gene family which differ in their specificities. The major difference in the models of TE32, TE33, and TE34 is the shape of the combining sites of the three antibodies: a shallow groove partly

10040 Biochemistry, Vol. 29, No. 43, 1990

Table I : N M R Distance Constraints Used To Model the TE34 Complex with CTP3 (Residues 8-15) distance constraint distance in the model

Zilber et al. 1989). It remains to be determined whether the aromatic core described is a feature common to a large population of antibodies and whether this core persists in antibodies against other antigenic determinants dominated by polar and charged residues. ACKNOWLEDGMENTS We thank Professor Ruth Arnon and Dr. Chaim Jacob for the hybridomas, Professor Fred Naider for very helpful discussions, Mrs. Rina Levy for excellent technical assistance, Mrs. Ruth Sharon for help with the computer programs, and Dr. Ora Goldberg for peptide synthesis.
Registry No. CTP3, 89 157-2843; N-acetyl-IDSQKKA, 12404227-9.

deviation atom in atom j.l (AIb (AIC (A)d his 8 H a Tyr 32H C t 5.0 5.3 0.3 5.0 his 8 H a Tyr 32H C y 5.8 0.8 his 8 C p 3.5 5.0 1.5 Trp 50H H{, Trp 50H Hq his 8 C p 3.5 3.9 0.4 5.5 6.3 Phe 94L C{ his 8 C p 0.8 Phe 96H Hc, his 8 C p 3.5 4.0 0.5 his 8 H6, 4.0 4.9 Phe 96H He2 0.9 4.4 ser 1 1 H a ile 9 H n 3.0 1.4 2.5 4.4 1.9 Trp 50H Hq ile 9 H a Trp 50H HS; ile 9 Hn 2.5 3.9 1.4 3.0 4.4 1.4 asp I O H a ser 1 1 H a 4.5 asp I O HP, ser 1 1 H a 4.8 0.3 4.5 4.3 asp I O HP2 ser 1 1 H a 0 Trp 50H H(, asp I O H a 2.5 3.8 1.3 Trp 50H Hq 2.5 3.5 asp 10 H a 1 .o 3.5 3.8 Trp 50H Hq asp I O ca 0.3 3.5 4.8 asp I O cg Trp 50H HC2 1.3 4.5 5.1 0.6 asp I O HP, lys I4 Cc 4.5 3.7 asp I O HP, lys 14 Cc 0 lys 13 H a Tyr 32L Cy 4.5 5.3 0.8 4.0 2.9 lqs 13 HP, Tyr 32L Cy 0 4.0 4.0 lqs 13 HB2 Tyr 32L Cy 0 4.6 Tyr 32L C( 1)s 13 Cy 0 5.5 6.0 5.9 I j s 13 C 6 Tyr 32L C ( 0 6.7 6.5 Tyr 32L C c lys 13 Cc 0.2 Phe 94L Cy 4.5 4.8 0.3 1)s 13 HP, 5.8 4.5 1.3 lys 13 HP2 Phe 94L Cy 5.7 Phe 94L Cy 0.2 5.5 lys 13 C 6 5.5 6.5 Phe 94L Cy lys I4 Cy 1 .o 6.0 6.3 lya 14 C y Tyr 32H Cy 0.3 5.5 5.6 0.1 l j s 14 C6 Tyr 32H C { 4.8 lys 14 Cc Tyr 32H Cy 0 5.5 4.5 5.4 ala 15 H a Tyr 32L Cy 0.9 5.5 5.2 Tyr 32L Cy 0 ala 15 cp 4.5 5.0 Phe 96H HC ala 15 CB 0.5 Peptide residues are written with lower-case letters while antibody residues begin with capital letters. H , heavy chain residues; L, light chain residues. bThe given value reflects the maximum distance from the NOE restraints and additional distance due to the use of pseudoatoms (Wuthrich et al., 1983). CThedistances in the final model obtained by energy minimization calculations. dThe deviation between the restraint and the actual distance in the model. When the latter is smaller than the restraint. the deviation is set eaual to zero.

REFERENCES Akolkar, P. N., Sikder, S. K., Bhattacharya, S. B., Liao, J., Gruezo, F., Morrison, S. L., & Kabat, E. A. (1987) J . Immunol. 138, 4472-4479. Amit, A. G., Mariuzza, R. A,, Phillips, S. E. V., & Poljak, R. J. (1986) Science 233, 747-753. Anglister, J., & Zilber, B. (1990) Biochemistry 29, 921-928. Anglister, J., Bond, M. W., Frey, T., Leahy, D., Levitt, M., McConnell, H. M., Rule, G. S., Tomasello, J., & Whittaker, M. (1987) Biochemistry 26, 6058-6064. Anglister, J., Jacob, C., Assulin, O., Ast, G., Pinker, R., & Arnon, R. (1988) Biochemistry 27, 717-724. Anglister, J., Levy, R., & Scherf, T. (1989) Biochemistry 28, 3360-3365. Bilofsky, H. S., & Burks, C. (1988) Nucleic Acids Res. 16, 1861-1 864. Chothia, C., Lesk, A. M., Levitt, M., Amit, A. G., Mariuzza, R. A., Phillips, S. E. V., & Poljak, R. J. (1986) Science 233, 155-758. Chothia, C., Lesk, A. M., Tramontano, A., Levitt, M., Smith-Gill, S. J., Air, G., Sheriff, S., Padlan, E. A., Davies, D., Tulip, W. R., Colman, P. M., Spinelli, S., Alzari, P. M., & Poljak, R. J. (1989) Nature 342, 877-883. Clore, G. M. & Gronenborn, A. M. (1983) J . Magn. Reson. 53, 423-442. 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., Tullcch, P. A., Air, G. M., & Webster, R. G. (1 987) Nature 326, 358-363. Griffiths, G . M., & Milstein, C. (1985) in Hybridoma Technology in the Biosciences and Medicines (Springer, T. A., Ed.) pp 103-1 15, Plenum Publishing Corp., New York. Jacob, C. O., Sela, M., & Arnon, R. (1983) Proc. Natl. Acad. Sci. U.S.A. 80, 761 1-7615. Jones, T. A. (1978) Acta Crystallogr., A34, 827-828. Juszczak, E., Near, R. I., Gefter, M. L., & Margolies, M. N . (1984) J . Immunol. 133, 2603. Kabat, E. A., Wu, T. T., Reid-Miller, M., Perry, H. M., & Gottesman, K. S. (1987) Sequences of Proteins of Immunological Interest, 4th ed., National Institutes of Health, Bethesda, MD. Kofler, R., Strobel, R., Balderas, R. S., Johnson, M. E., Noonan, D. J., Duchosal, M. A., Dixon, F. J., & Theofilopoulos, A. N. (1988) J . Clin. Invest. 82, 852-860. Lesk, A., Levitt, M., & Chothia, C. (1986) Protein Eng. 1 , 17-88. Levitt, M. (1983) J . Mol. Biol. 170, 723-764. Levitt, M., & Sharon, R. (1988) Proc. Natl. Acad. Sci. U.S.A. 85, 7557-1561.

surrounding Trp lOOaH in TE32 and TE33 and a deep cavity in TE34 which lacks Trp 100aH. This cavity is the result of the deletion in TE34 of the sequence Ser 96H, Trp IOOaH, Tyr lOObH which is found in the CDR3 of the heavy chain of both TE33 and TE32. These observations support our previous conclusion that aromatic residues inside the antibody combining site create a general active hydrophobic surface, while other residues, and especially variations in the length of the CDRs, modulate its polarity and shape to fit a specific antigen (Levy et al., 1989). On the basis of our models for the three anti-CTP3 antibodies, a distinction can be made between two regions on the surface created by the six CDRs: (a) the center, which is mostly composed of the exposed side chains of aromatic residues; (b) the periphery, where a few aromatic residues are interspersed throughout predominantly nonaromatic residues. The interactions of our peptide occur mainly with the center, which forms the combining site. In contrast, in antibodyprotein complexes, a considerable fraction of the interactions occurs with peripheral residues, a considerable fraction of the interactions occurs with peripheral residues, although calculations have shown that residues in the center of the combining site contribute the most to binding energy (Novotny et al.,

Biochemistry 1990, 29, 1004 1- 10048 Levy, R., Assulin, O., Scherf, T., Levitt, M., & Anglister, J. ( 1 989) Biochemistry 28, 7 168-7 175. Liu, A. Y., Robinson, R. R., Hellstrom, K. E., Murray, E. D., Jr., Chang, P., & Hellstrom, I. (1987) Proc. Natl. Acad. Sci. U.S.A. 84, 3439-3443. Marquart, M., Deisenhofer, J., Huber, R., & Palm, W. (1980) J . Mol. Biol. 141, 369-391. Meek, K., Jeske, D., Slaoui, M., Leo, O., Urbain, J., & Capra, J. D. (1984) J . Exp. Med. 160, 1070-1086. Meyer, E. F., Clore, G. M., Gronenborn, A. M., & Hansen, H. A. S . (1988) Biochemistry 27, 725-730. Novotny, J., Bruccoleri, R. E., & Saul, F. A. (1989) Biochemistry 28, 4735-4749. Padlan, E. A., Silverton, E. W., Sheriff, S., Cohen, G. H., Smith-Gill, S. J., & Davies, D. R. (1989) Proc. Natl. Acad. Sci. U . S . A .86, 5938-5942. Panka, D. J., Mudgett-Hunter, M., Parks, D. R., Peterson, L. L., Herzenberg, L. A., Haber, E., & Margolies, M. N.

10041

(1988) Proc. Natl. Acad. Sci. U.S.A. 85, 3080-3084. Saul, R., Amzel, L., & Poljak, R. (1978) J . Biol. Chem. 253, 585-597. Segal, D., Padlan, E., Cohen, G. H., Rudikoff, S., Potter, M., & Davies, D. (1974) Proc. Natl. Acad. Sci. U.S.A. 71, 4298-4302. Sheriff, S . , Silverton, E. W., Padlan, E. A., Cohen, G. H., Smith-Gill, S. J., Finzel, B. C., & Davies, D. R. (1987) Proc. Natl. Acad. Sci. U.S.A. 84, 8075-8079. Suh, S. W., Bhat, T. N., Navia, M. A., Cohen, G. H., Rao, D. N., Rudikoff, S., & Davies, D. (1986) Proteins: Struct., Funct., Genet. I , 74-80. Winter, E., Radbruch, A., & Krawinkel, U. (1985) EMBO J. 4, 2861-2867. Wuthrich, K. (1986) N M R of Proteins and Nucleic Acids, Wiley, New York. Wuthrich, K., Billeter, M., & Braun, W. (1983) J . Mol. Biol. 169, 949-961.

NMR Identification of Protein Surfaces Using Paramagnetic Probes

Andrew M. Petros,* Lucian0 Mueller,s and Kenneth D.Kopple* SmithKline Beecham Pharmaceuticals, P.O. Box 1539, King of Prussia, Pennsylvania 19406 Received March 16, 1990; Revised Manuscript Received June 22, 1990

ABSTRACT: Paramagnetic agents produce line broadening and thus cancellation of anti phase cross-peak components in two-dimensional correlated nuclear magnetic resonance spectra. The specificity of this effect was examined to determine its utility for identifying surface residues of proteins. Ubiquitin and hen egg white lysozyme, for which X-ray crystal structures and proton N M R assignments are available, served as 1-oxy and test cases. Two relaxation reagents were employed, 4-hydroxy-2,2,6,6-tetramethylpiperidinylthe gadolinium(II1) diethylenetriaminepentaacetate complex ion. Correlations were sought between reagent-produced decreases of side-chain cross-peak volumes in double-quantum-filtered proton correlation (DQF-COSY) spectra and the solvent-exposed side-chain surface area of the corresponding residues. The lanthanide complex produced strong effects ascribable to association with carboxylate groups but was not otherwise useful in delineating surface residues. The nitroxyl, on the other hand, produced clear distinctions among the Val, Leu, and Ile residues that generally paralleled side-chain exposure in the crystal, although consistent correlations were not observed with residues of other types. Although an instance of possible specific protein-nitroxyl association was noted, the nitroxyl appears to be a tool for identifying hydrophobic surface residues.

F o r most enzymes and for globular proteins in general, surface residues play a critical functional and immunological role. Identification of a protein surface may therefore be of value in several ways. For hormonally active proteins, knowledge of the surface may aid in the design of peptide agonists and antagonists, especially when the functional regions are composed of residues not in a continuous sequence. For enzymes, mapping of surface residues may aid in identifying substrate or inhibitor binding sites, which in turn will facilitate inhibitor design. Furthermore, identification of surface residues may supplement NOE' and coupling constant data in
*To whom correspondence should be addressed at SmithKline Beecham Pharmaceuticals, L-940, P.O. Box 1539, King of Prussia, PA
194060939,

the determination of protein structures by NMR. Several methods have been proposed for using N M R spectroscopy to explore the surface structure of proteins and peptides. The photo-CIDNP technique has been used successfully to look at solvent-exposed aromatic side chains of several proteins (Kaptein et al., 1978; Kaptein, 1982; Stob et al., 1988). Other methods that have been employed or proposed include temperature dependence of amide proton chemical shifts (Kopple et al., 1969), proton-deuteron exchange rates (Wagner & Wuthrich, 1982), pH-rate profiles
Abbreviations: NOE, nuclear Overhauser effect; N M R , nuclear magnetic resonance; 1D, one dimensional; 2D, two dimensional; CIDNP, chemically induced dynamic nuclear polarization; COSY, correlated spectroscopy; DQF, double quantum filtered; HyTEMPO, 4-hydroxy2,2,6,6-tetramethylpiperidinyl-l-oxy; Gd[DTPA], gadolinium(II1) diethylenetriaminepentaacetate complex ion; hew, hen egg white; NOESY, nuclear Overhauser effect spectroscopy; TOCSY, totally correlated spectroscopy; tri-NAG, tri-N-acetylglucosamine.

'

'Present address: Abbott Laboratories, Dept. 47G. Bldg. AP9, Abbott Park, IL 60064. $Present address: Squibb Institute for Medical Research, Room D4159, P.O. Box 4000, Princeton, N J 08543.

0006-2960/90/0429- 10041$02.50/0

0 1990 American Chemical Society