The sites in the I-Ak histocompatibility molecule photoaffinity labeled by an immunogenic lysozyme peptide.

The class II histocompatibilty molecule I-Ak was photoaffinity labeled by NH2- and COOH-terminal photoreactive conjugates of an immunogenic hen egg white lysozyme (HEL) peptide. The labeled alpha and beta chains were digested with protease from Staphylococcus aureus strain V-8 (protease V-8) and/or trypsin, and the proteolytic fragments were separated by high performance liquid chromatography (HPLC) (peptide mapping). Reproducible peptide maps containing a major labeled component were obtained from the three conjugates reported here whose photoreactive group was attached via short spacers of limited flexibility. The COOH-terminal conjugate N-acetyl HEL-(49-61)-iodo-4-azidosalicyloyl thioester (compound 1) labeled hydrophilic tryptic digest fragments on both chains of I-Ak. The labeled digest fragments were homogeneous in reverse-phase and anion-exchange HPLC, indicating that the photoaffinity labeling was site-specific. Conversely, the NH2-terminal conjugate iodo-4-azidosalicyloyl HEL-(46-61) (compound 2: IASA-(46-61)) labeled exceptionally hydrophobic sequences on both chains of I-Ak. The labeling was also site-specific because reverse-phase HPLC of primary digests with protease V-8 and secondary digests with trypsin showed single major labeled components. The labeling of I-Ak by IASA-(46-61) was fully inhibitible by HEL-(46-61). In contrast, IASA attached to the smallest immunogenic peptide 52-61 (compound 3) labeled a distinctly different hydrophilic tryptic fragment. The site of the I-Ak molecule that was photoaffinity labeled by IASA-(46-61) (compound 2) was determined. IASA-(46-61) labeled selectively at Pro-118 of a primary alpha chain fragment most likely encompassing residues 115-134. It labeled Thr-121 of a primary beta chain fragment most likely encompassing residues 109-138. We also obtained evidence that IASA-(46-61) occupied the antigen-specific site; the conjugate stimulated a T-cell hybridoma that recognizes the sequence 52-61 and also competed for the binding of this smaller peptide to I-Ak. Thus, peptides that bind to the allele-specific binding site and are long enough to extend beyond it can interact with a hydrophobic area of class II molecules. This area is formed by sequences of the first halves of the second domain of both alpha and beta chains.

We also obtained evidence that IASA-(46-61) occupied the antigen-specific site; the conjugate stimulated a T-cell hybridoma that recognizes the sequence 52-6 1 and also competed for the binding of this smaller peptide to I-Ak. Thus, peptides that bind to the allele-specific binding site and are long enough to extend beyond it can interact with a hydrophobic area of class II molecules. This area is formed by sequences of the first halves of the second domain of both a and fi chains.
* These investigations were supported by grants from the National Institutes of Health, the American Cancer Society, and the Monsanto-Washington University agreement. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. Class 11 major histocompatibility complex molecules (MHC) ' are intimately involved in the activation of CD4positive T-lymphocytes (1). Their function is to bind specifically protein antigen or antigen-derived fragments, in correlation with the underlying immune response pattern (2, 3). The three-dimensional structure of class II MHC molecules is not known. According to a model proposed by Brown et al. (4), the antigen binding site is formed by the first domains of the 01 and @ chain in a structure similar to the antigen binding site of class I MHC molecules (5, 6). In this model, the NH,terminal halves of the first domains of the (Y and p chains form a ,&pleated sheet platform, and the COOH-terminal halves form two anti-parallel ol-helices. The binding site is an extended groove whose floor is formed by the P-pleated sheet platform and whose sides are formed by the anti-parallel LYhelices. This hypothetical antigen binding site therefore has a geometry similar to the class I binding site, which is approximately 25  The second issue is whether class II MHC molecules can bind immunogenic peptides in more than one configuration. If so, an expanded spectrum of T-cell reactivities could be elicited against single immunogenic structures (discussed in Ref. 9). Both of these issues can be addressed by determining antigen contact points on class II MHC molecules. In this study we explored the possibility of determining such antigen contact points via photoaffnity labeling. The I-Ak class II MHC molecule specifically binds the immunogenic hen egg white lysozyme peptide (HEL)-46-61 (2, 10) and its NHZ-terminal truncated homologues (HEL)-49-61 (2, 10) and (HEL)-52-61 (11). The interaction of these peptides with the I-Ak molecule can also be studied by photoaffinity labeling (12), which creates a covalent bond between the photoreactive peptide conjugate and the class II molecule. In the present study we analyzed the photoaffinity labeling of I-Ak by different photoreactive 4-azidobenzoyl derivatives of (HEL)-46-61.
The UV light-induced intermediate nitrenes have a very short life span (13) and are able to react with virtually any amino acid residue in proteins (14). The rapid reaction time of the photoreactive peptide conjugates coupled with the extremely slow dissociation of immunogenic peptides from class II MHC molecules (15)  As suggested, the main product probably is the 3-iodo-4-azidosalicyloyl derivative and the minor product the 5-iodo-4-azidosalicyloyl isomer (24). We also prepared the homologous (IASA)-52-61, using peptide 52-61 of HEL. The main product eluted from the HPLC column at 29 min with a minor product at 30 min. In some experiments I-Ak on cell membranes was photoaffinity labeled as described nreviouslv (12). The dried gels were exposed with an amplifying slreen to -Kodak XAR-5 x-rai film at -80 'C. The labeled bands were cut out from the gels, rehydrated in 250 mM Tris buffer, pH 8.0, containing 8.3% glycerol and 0.3% SDS for 3 h at 4 "C. The labeled a and j3 chains were eluted by electrophoresis using a disc electrophoresis apparatus with a constant current of 3 mA/disc overnight at 4 "C. The anodic end of the discs was plugged with glass wool and connected to a closed dialysis membrane tubing ( in amounts proportional to their recognition of (HEL)-52-61-I-Ak complex (1, 11). I-Ak molecules purified from CH27 cells were incubated with either (HEL)-52-61 (11) or (HEL)-34-45 (9), as described (18) conjugates of (HEL)-46-61 and (HEL)-49-61-Cys and analyzed the labeled peptides after digestion of the separated (Y and /3 chains with trypsin or protease V-8. In this paper we will examine our initial results with the three compounds shown in Fig. 1 and in particular with compound 2 ((IASA)-46-61).
The photoreactive IASA group was directly conjugated at the COOH-terminal cysteine of Ac-49-61-Cys (compound 1) or at the NH2 terminus of either (HEL)-46-61 (compound 2) or (HEL)-52-61 (compound 3). With these three compounds, the tryptic peptide maps of the labeled (Y and /3 chain revealed resolved single major labeled components (Figs. 1 and 2). The fragment from the @ chain labeled by compound 2 eluted considerably later (46 min) from the reverse-phase column than the one labeled by compound 1 (21 min) (Fig. l), indicating that it was more hydrophobic. Indeed, peptide maps of unlabeled I-Ak showed very few fragments eluting in the last third of tryptic peptide maps (see Refs. 19,20). Furthermore, we failed to obtain a tryptic peptide map of the (Y chain labeled by compound 2, suggesting that an even more hydrophobic digest fragment was labeled which was no longer amenable to reverse-phase chromatography. In contrast to compound 2, a similar conjugate, but shorter, of the minimal immunogenic peptide from residues 52-61 (11) labeled a different, more hydrophilic, tryptic peptide fragment (Fig. lC, dotted line).
The tryptic peptide map of the (Y chain labeled by compound 1 showed a major labeled component eluting at 22.0 min ( Fig.  2A). To assess the homogeneity of this fragment, it was rechromatographed on anion-exchange (DEAE) HPLC. Again a single major labeled fraction was observed at 21 min (Fig. 2B). Likewise, rechromatography on DEAE HPLC of the major labeled p chain fragment (Fig. 1B) revealed a single major component eluting at 27 min (Fig. 2C). This later elution indicated that the labeled p chain fragment was more acidic than the labeled (Y chain fragment. The observation that the major labeled fragments of both chains were homo-8 16  In the remainder of this study we will focus on compound 2. The site of labeling of I-Ak was entirely inhibited by about a lOOO-fold molar excess of unlabeled compound 2 and by about a 5000-fold molar excess of (HEL)-46-61.
The photoaffinity labeling of I-Ak by compound 2 was further studied with protease V-8 peptide maps. In the case of the LY chain a single major labeled component eluted at 48 min and in the case of the p chain at 41 min (Fig. 3, A and B). The homogeneity of these labeled peptides was assessed by secondary tryptic peptide maps. (The high degree of hydrophobicity of these labeled fragments excluded their analysis by ion-exchange chromatography.) The secondary fragment of the LY chain eluted again as a major single component at 52 min and that of the p chain at 47 min (Fig. 3, C and D). These results indicate that both primary labeled digest fragments are homogeneous and contain tryptic cleavage site(s).
We assessed whether compound 2 labeled the membrane spanning domains of the I-Ak molecule. I-Ak was photoaffinity labeled on cells, and the separated chains were subjected to primary protease V-8 and secondary tryptic digestion. In contrast to purified I-Ak in detergent, in which both chains were labeled, I-Ak on cells was selectively labeled on the LY chain (12).* The peptide map of the I-Ak (Y chain labeled on the membrane was identical to that of I-Ak labeled in detergent (not shown). This result suggests that the I-Ak (Y chains on the membrane and in detergent were labeled at the same site, therefore excluding the transmembrane domain as the labeled site. Identification of the Site Labeled by Compound 2 ((IASA)-46-61)--We examined whether the tryptic cleavage sites of the a and @ chains labeled by compound 2 were argininets) and/or lysine(s). The labeled fragments from protease V-8 digestion were treated with endoprotease Arg-C or Lys-C, which selectively cleaves COOH-terminal to arginine or lysine, respectively, and were rechromatographed under iden-'1. F. Luescher and E. R. Unanue, manuscript submitted for publication.
tical conditions (C-4 reverse-phase HPLC). In the case of the (Y chain, the secondary endoprotease Lys-C peptide map showed a new labeled fragment eluting at 51 min (Fig. 4A). Based on the high specificity of this enzyme (21), this result indicated the presence of a lysine residue in the primary fragment. Similarly, treatment with clostripain or endoprotease Arg-C from mouse submaxillary glands also resulted in a secondary digest fragment (Fig. 4.4). Clostripain treatment gave similar results (data not shown). Since both enzymes are specific for arginine bonds (22,23), these results indicate the presence of an arginine residue in the primary protease V-8labeled peptide.
In the case of the @ chain, peptide maps of the labeled primary protease V-8 digest fragment secondarily digested with trypsin suggested the presence of multiple basic amino  (Fig. 4B). Secondary peptide maps with protease Lys-C and Arg-C showed two new labeled peptides, suggesting the presence of 2 lysines and arginines in the primary digest fragment (Fig. 4B). The relative amount of the secondary digest products in these peptide maps varied in different experiments.
The earlier eluting fragment in the protease Arg-C peptide map (44 min) was the major peptide when clostripain rather than protease Arg-C from mouse submaxillary gland was used.
We next examined whether the labeled protease V-8 digest fragments contained cysteine or tryptophan, infrequent amino acids in I-Ak (Fig. 9). Cysteine labeled I-Ak was incubated with compound 2 and digested with protease V-8. Double counting of the peptide maps showed that the photoaffinity labeled fragment of the (Y chain did not contain [35S] cysteine (Fig. 5, A and B). Interestingly, the same peptide maps were obtained whether the labeled (Y chain was reduced and alkylated prior to digestion or whether the digest products were reduced and incubated with 5,5'-dithiobis(2-nitrobenzoic) acid which gives typical shifts of thiol-containing peptides in reverse-phase HPLC (24) (not shown). It thus appears that the thiol function in this sequence became irreversibly "blocked," during the UV irradiation which is known to destroy cysteine structures (25). The same experiment on the p chain was inconclusive since multiple 'S-containing fragments eluted in the region of the lz51-labeled component (not shown). The latter lz51-labeled component was isolated and then treated with trypsin. The secondary tryptic peptide map on this 'Y-labeled component revealed that the fragment contained [35S]cysteine (Fig. 5, C and D).
Protease V-8 peptide maps showed that the photoaffinity labeled products of both chains contained 13H]tryptophan (Fig. 6). Secondary tryptic peptide maps further indicated that the products of the (Y but not the /3 chain contained tryptophan (Fig. 7). The photoaffinity labeled site and tryptophan had therefore to be located on the same side of the tryptic cleavage site(s) in the case of the (Y but not p chain. I-Ak was then intrinsically labeled with [3H]proline and photoaffinity labeled with (IASA)-46-61. Primary peptide maps with protease V-8 showed proline to be present in the labeled digest fragments of both chains (not shown), which were subjected to amino acid sequencing. In the case of the (Y Fraction 41 (Fig. 4B) of the primary protease V-8 digest of the p chain was digested with trypsin and the digest analyzed by reverse-phase (C-4) HPLC. C shows the "'1 and D the 3H elution patterns. Note that in D the sH radioactivity did not coelute with iz51 radioactivity. chain, [3H]proline eluted in the 4th and 5th cycle and the 1251photoaffinity-labeled residue in the 4th cycle (Fig. 8, A and  B). In the case of the fi chain, 3H-proline eluted at the 17th cycle and the 'Y-labeled residue at the 13th cycle (Fig. 8, C  and D). The findings are compatible with the photoaffinity labeled sites being located in the primary (Y chain fragment at residues 115-134 and in the primary /I chain fragments at residue 109-138. The (Y chain fragment contains proline at positions 4 and 5 (Pro-118 and -119) and the /3 chain fragment at position 17 (Pro-125) (Fig. 9). (Assuming that the trypsin digestion was complete, the secondary peptides likely include residues 115-127 for the (Y chain and 109-127 for the /3 chain.) Interactions of Compound 2 with Z-Ah-Dose-response curves showed that compound 2 ((IASA)-46-61) was slightly less stimulatory than (HEL)-46-61 (Fig. 10). Since the activation of this T-cell hybridoma depends on the recognition of the I-Ak. (HEL)-46-61 complex, these results indicate that the photoreactive groups did not substantially alter this in- The '251-labeled (Y and p chains were obtained from a primary protease V-8 digestion after chromatography (as in Fig. 4, A and B). A  teraction (11). We also examined whether unlabeled compound 2 competed effectively for the binding of bound; and this binding was entirely abolished by lo-fold excess of compound 2.) DISCUSSION We are in the process of determining the contact points of the HEL peptides 46-61 and 49-61 for the I-Ak MHC molecule by photoaffinity labeling. The identification of photoaffinity labeled sites requires that the photoaffinity labeling be sitespecific and that the labeled proteins can be fragmented in a defined and reproducible manner, preserving the photocrosslinked bond. The findings of this first study indicate that photoaffinity labeled sites on the I-Ak molecule can, indeed, be determined.
Site-specific labeling that leads to resolved peptide maps, however, requires that the photoreactive groups be conjugated to the peptide via short spacers of limited flexibility.
Our other studies3 using photoconjugates containing cysteines resulted in unresolved peptide maps perhaps as a result of a longer and more flexible molecule. In this study we have also been able to identify the site of labeling of the (Y and p chains by an NH*-terminal conjugate of HEL 46-61 (compound 2 in the text).
The photoreactive IASA group conjugated directly at the NH*-terminal amino group of (HEL)-46-61 (compound 2), and the COOH-terminal conjugate, compound 1, yielded resolved, labeled peptide maps (Figs. l-3). However, compound 2 labeled exceptionally hydrophobic protease V-8 as well as tryptic fragments. The hydrophobicity of the labeled digest fragments was not accounted for by the photoconjugation. (HEL)-46-61 itself is a tryptic digest fragment that contains two protease V-8 cleavage sites (Fig. 1); the protease V-8 digest fragment will therefore be labeled by a The indicated concentrations of p&tide or conjugates were added to a constant number of fixed CH-27 cells and 3A9 cells. The supernatants were assayed after 24 h for interleukin-2 by [3H]thymidine incorporation by the interleukin-2-dependent CTLL cells. The bars indicate standard deviation of the triplicate culture.

drophilic
"tag" is expected to give only a small delay in the elution of the labeled fragments.
Reverse-phase peptide maps of I-Ak labeled by (IASA)-46-61 indicated that the labeled primary protease V-8 digest fragments of both chains contained tryptic cleavage sites (Fig.  4). In the case of the cy chain, digestion of the V-8 fragment with endoprotease Arg-C and Lys-C showed new labeled fragments, indicating the presence of an arginine and a lysine residue (Fig. 4A). In the case of the p chain, the same experiment showed two new digest fragments in both peptide maps. Because the 125I label was found at only one position during sequencing, these results suggest the presence of 2 lysine and 2 arginine residues in the primary protease V-8 digest fragment (Fig. 4B). Assuming that protease V-8 cleaved at all acidic residues, the following generated peptides are compatible with these findings: I-Ak a-76-89 and -115-134 and I-Ak p-123-138 and -195-238 (Fig. 9). In the case of the (Y chain the former peptide contains three and the latter eight hydrophobic amino acids. In view of the late elution of the labeled protease V-8 peptide (Fig. 4A), the latter sequence seems to be more likely to contain the labeled site(s). This hypothesis was further supported by the failure to obtain a primary tryptic peptide map of the labeled cy chain. The corresponding tryptic peptide that encompasses residues 99-127 contains 15 hydrophobic amino acids and may indeed not be amenable to reverse-phase HPLC (Fig. 9).
Peptide maps on [35S]cysteine intrinsically labeled I-Ak showed that only the labeled peptide derived from the /3 chain contained cysteine (Fig. 5). In contradiction to these findings, however, the predicted labeled /3 chain peptide, including residues 123-138, did not contain cysteine (Fig. 9). It is, however, conceivable that protease V-8 failed to cleave at Asp-122, resulting in the primary labeled protease V-8 peptide being 109-138. Protease V-8 has been reported to cleave sluggishly when aspartic acid residues are followed by bulky hydrophobic residues, such as is found in the present sequence (Asp-Phe-Tyr) (26). Furthermore, a primary V-8 peptide spanning residues 109-138 better accounts for the secondary tryptic peptide maps. The secondary digest /? chain peptide derived from this larger peptide spans residues 109-127 and contains eight hydrophobic amino acids. It, rather than the alternative hydrophilic tryptic peptide spanning residues 123-127, would be expected to elute at 47 min (Figs. 4B and 7C).
Primary protease V-8 peptide maps of I-Ak labeled intrinsically with tryptophan showed that the labeled digest products of both chains contained tryptophan (Fig. 6). This finding is compatible with the primary labeled protease V-8 fragments being I-Ak a-115-134 and I-Ak p-109-138. More importantly, the secondary tryptic digest product of the LY chain, but not the /I chain, contained tryptophan (Fig. 7). Tryptophan in the predicted primary B chain fragment is located between tryptic cleavage sites and is therefore expected to be absent in the secondary peptide. Conversely, in the case of the (Y chain, tryptophan is located in the hydrophobic sequence NH*terminal of the tryptic cleavage sites and is therefore expected to be present in the secondary peptide. In fact, there are no other protease V-8 peptides that contain tryptophan and tryptic cleavage sites in a manner compatible with these peptide maps. The predicted labeled protease V-8 cy chain peptide 115-I34 contains proline at positions 4 and 5 (Pro-118 and -119) and the p chain fragment 109-138 at position 17 (Pro-125). Sequencing of V-8 protease peptides prepared from [3H]proline intrinsically labeled I-Ak molecules localized proline at the predicted sites (Fig. 8). These sequencing experiments further showed lZ51 at the 4th residue of the LY chain and at the 13th residue of the /3 chain, indicating that the photoaffinity labeled residues were I-A" Pro+118 and I-Ak Thr-p121.
Our interpretation is that (IASA)-46-61 folds into the allele-specific peptide-binding site at the critical stretch from residues 52 to 61 and that the peptide from residues 46 to 51 extends beyond the peptide-binding site to a nearby area rich in hydrophobic residues in the (Y* and /3* domains. Several observations support this interpretation. First, the finding that (IASA)-46-61 stimulated a T-cell hybridoma similarly to the unmodified peptide ( Fig. 1) argues that the 52-61 residues of the peptide must be specifically bound to the allele-combining site. This was also confirmed by showing that it competed for the binding of labeled 52-61 or an unrelated lysozyme peptide. We had previously established that (HEL)-46-61 can be NH*-terminally truncated to the minimal immunogenic structure 52-61 (27). The residues that contact I-Ak are Asp-52, Ile-58, and Arg-61. The photoreactive group is spaced by 7 residues from Asp-52. Second, a shorter compound such as compound 3, (IASA)-52-61, failed to label these hydrophobic sites on I-Ak, indicating that their labeling requires a spacer. Third, a less hydrophobic conjugate ABA-Met-46-61 only partially labeled these hydrophobic sequences but also labeled more hydrophilic ones (data not reported). Fourth, the ability of (IASA)-46-61 or (IASA)-52-61 to label other alleles of class II molecules correlated with the labeling of this hydrophobic site, suggesting allele nonspecific hydrophobic interactions with the photoreactive groups (12). The potential of the conjugated photoreactive groups to undergo hydrophobic interactions was indeed observed by reversephase HPLC; ABA-Met-46-61 eluted 4 min later from a C-18 column than (HEL)-46-61, and (IASA)-46-61 was even 10 min later.
The presence of a nonpolymorphic, hydrophobic domain located in the vicinity of the allele-specific antigen binding site of class II molecules constitutes a major difference from class I molecules. In class I molecules, the beginning of the constant (third) domain forms a "loop" at the outer side of the molecule, connecting the helical part of the second domain with P-pleated sheet structures in the constant domain (5). This class I constant domain is hydrophilic, containing mainly charged amino acids. In contrast, the sequences of the corresponding constant (second) domains of class II molecules show a considerably different hydrophobicity pattern, containing mainly uncharged, hydrophobic amino acids. Furthermore, in contrast to class I, class II molecules have two such sequences which, according to the model, exit the antigen binding site on opposite sites (4). It is tempting to speculate