The binding of peptides to MHC class II molecules is mediated in part by a conserved array of intermolecular hydrogen bonds. We have evaluated the consequences of disrupting the hydrogen bond between β-His-81 of the class II molecule and bound peptide. These studies revealed that peptide dissociation rates were accelerated by factors ranging to 200-fold. The sensitivity of a peptide to loss of the hydrogen bond is inversely correlated with the inherent kinetic stability of the peptide-MHC complex. The same relationship has been observed between inherent kinetic stability and the susceptibility to DM. Given that the rate enhancement observed for MHC class II I-Ad protein mutated at position 81 in the β-chain is comparable with DM-catalyzed rates for other class II molecules, we suggest that DM could function by stabilizing a peptide-MHC intermediate in which one or more hydrogen bonds between the peptide and MHC, such as that contributed by the β-His-81 hydrogen bond, are disrupted.

Successful crystallization of MHC class II molecules has provided detailed insight into the structural elements that control their ability to bind peptides. Both polymorphic and conserved amino acids contact the bound peptide (reviewed in Ref. 1). In addition to pocket interactions, binding of peptide to the class II molecule is characterized by a latticework of hydrogen bonds from genetically conserved amino acids in the class II protein to the main chain of the peptide. The importance of these hydrogen bonds in peptide binding to class II molecules is not yet clear. In class II crystal structures, a mixture of peptides gives a well-defined electron density along the entire binding groove, whereas related class I-peptide structures are disordered in the central region. These results support the importance of binding interactions along the entire length of a peptide.

The contribution of hydrogen bonds between the α helices and peptide main chain in peptide acquisition by class II molecules has not yet been experimentally addressed. We have shown previously that substitution at either of two sites in the β-chain of I-Ad (His-81 or Asn-82) that contribute hydrogen bonds to the peptide main chain had dramatic consequences on class II intracellular trafficking (2, 3). Our recent studies have examined the fate of one of these molecules, MHC class II I-Ad protein mutated at position 81 in the β-chain (81βH),3 when it is associated with invariant chain (Ii) within APCs (4). After endosomal localization, the 81βH molecule is rapidly degraded. Protease susceptibility was shown to be a due to failure of the variant class II molecule to successfully acquire peptides after rapid dissociation of class II associated Ii-derived peptide (CLIP). We speculated that this single amino acid change led to a global deficiency in peptide acquisition. Subsequent steady state binding assays using in vitro-translated class II molecules indicated that a loss of the potential to form one to two hydrogen bonds leads to defects in the assembly of class II-peptide complexes (5).

In the present study, we have examined the contribution of a single hydrogen bond at the periphery of the peptide-binding site to the kinetic stability of peptide-class II complexes. This is an area of particular interest in light of the molecular models put forth to explain the mechanism involved in DM-mediated dissociation of peptides from the class II molecule. DM has been shown to enhance the dissociation of peptides from class II molecules and the subsequent loading with new peptide (6, 7, 8, 9, 10, 11). It has been suggested that such a catalytic function might be accomplished by the binding of DM to an “open” transitional state of class II molecules. We speculated that such an open conformational state might represent destabilized hydrogen bonds. In the experiments reported here, we tested whether destabilization of a single hydrogen bond would accelerate the dissociation kinetics of peptide from the class II molecule.

Peptides were synthesized with standard fast-fluorenylmethoxycarbonyl chemistry, labeled on the N terminus with fluorescein, and purified by reverse-phase HPLC as described previously (12).

I-Ad protein was isolated from detergent lysates of 1–2 × 1010 cells expressing I-Ad or 81βH as described previously (12), using an Ab affinity column (MKD6, Ref. 13). The yield of isolated class II molecules is ∼300 μg per 2 × 1010 cells; purity is >90%.

I-Ad wild-type (WT) or 81βH protein (0.2 μM) and an excess of peptide (10 μM) were incubated in 0.2 mM dodecyl maltoside/100 mM citrate/PBS at pH 5.3 and 37°C for 1–4 h (81βH) or 16–24 h (WT). Control experiments demonstrated that the time of incubation of 81βH with Eα peptide (1.5 h vs 15 h) did not significantly change the dissociation t1/2 (1.8 h vs 2.2 h, respectively). Unbound peptide was removed by rapid size exclusion (Sephadex G50-SF) at 4°C, pH 5.3. The reaction mixture was separated by high performance size exclusion chromatography using a 30-cm TSK3000SWXL column (Tosohaas, Montgomeryville, PA) and a fluorescence detector. The initial amount of labeled peptide bound to MHC was measured as the peak height fluorescence of the peptide/MHC fraction. After subsequent incubations at 37°C, the relative peak height at each time was used as a measure of peptide that remained bound to the protein. t1/2 of dissociation were obtained from single exponential fits to the dissociation data.

Fig. 1 is a schematic representation of the I-Ad structure, showing the potential hydrogen bonds between the class II α helices and the peptide main chain. Our studies focus on a hydrogen bond near the amino terminus of the bound peptide, specifically residue 81 in β. This residue was conservatively mutated (His→Asn) to disrupt the potential to form a single hydrogen bond to the peptide main chain, yielding mutant molecules termed 81βH. Several pieces of data suggest that when in its peptide-bound state, the variant class II molecule does not generally have an altered conformation/structure. First, the His→Asn change occurs at a solvent exposed site and is not predicted to have secondary affects on the conformation (4). Second, when expressed without Ii, 81βH molecules react with a wide panel of Abs that interact with the class II α helices (data not shown). Third, the 81βH molecule displays normal kinetics of transport from the endoplasmic reticulum to the Golgi (2), reflecting an absence of selective recognition by “quality control” mechanisms to retain misfolded proteins. We conclude that the change at 81β has a very limited effect on the overall conformational features of class II and is distinct from WT primarily in its ability to contribute a hydrogen bond to the amino terminal region of the bound peptide.

FIGURE 1.

Schematic representation of the I-Ad-peptide complex. This representation is adapted from Ref. 16 and shows the α- and β-chain α helices in yellow; the peptide backbone is represented in CPK colors. The potential hydrogen bonds between the class II α helices and the peptide main chain are also highlighted. The β-His-81 side chain is highlighted in red.

FIGURE 1.

Schematic representation of the I-Ad-peptide complex. This representation is adapted from Ref. 16 and shows the α- and β-chain α helices in yellow; the peptide backbone is represented in CPK colors. The potential hydrogen bonds between the class II α helices and the peptide main chain are also highlighted. The β-His-81 side chain is highlighted in red.

Close modal

Our previous studies (4, 5) suggested that the 81βH molecules are deficient in assembly with peptide. We sought to determine whether this deficiency could be accounted for by enhanced dissociation rates. The relative kinetic stabilities of the 81βH molecule with peptides known to bind to I-Ad were compared with the same complexes formed with WT I-Ad (Table I). Included were two peptides (cystatin C (CysC) residues 40–55 and Eα residues 52–67) isolated in high yield from the I-Ad expressed on B cells, which were estimated to have very high affinity binding to I-Ad (14). Also included was the Ii-derived CLIP peptide, because of its high affinity for I-Ad (12, 15), its importance in class II biogenesis, and its sensitivity to DM-mediated release. Finally, we included the influenza hemagglutinin-derived peptide (HA126–138), recently cocrystallized with I-Ad (16). L cell fibroblasts, constructed by transfection to express WT I-Ad or 81βH molecules, were used as a source of class II molecules. The class II molecules were expressed without Ii to diminish endosomal localization and to allow surface expression of the 81βH molecule.

Table I.

Peptides analyzed for binding to I-Ad and 81βH

PeptideSequenceSource
CLIP KPVSQMRMATPLLMR Residues 85–99 of murine Ii 
   
CLIP (P → A) KPVSQMRMATALLMR CLIP (85–99) residue 95 Pro to Ala 
   
CysC DAYHSRAIQVVRARK Residues 40–55 of cystatin C 
   
Eα ASFEAQGALANIAVDK Residues 52–67 of I-Ed α-chain 
   
HA HNTNGVTAASSHE Residues 126–138 of influenza hemagglutinin 
PeptideSequenceSource
CLIP KPVSQMRMATPLLMR Residues 85–99 of murine Ii 
   
CLIP (P → A) KPVSQMRMATALLMR CLIP (85–99) residue 95 Pro to Ala 
   
CysC DAYHSRAIQVVRARK Residues 40–55 of cystatin C 
   
Eα ASFEAQGALANIAVDK Residues 52–67 of I-Ed α-chain 
   
HA HNTNGVTAASSHE Residues 126–138 of influenza hemagglutinin 

The dissociation kinetics of peptides bound to either WT I-Ad molecules or the 81βH molecule with the single disrupted hydrogen bond interaction were compared (Fig. 2). A striking and profound loss in the stability of peptide-class II interactions is apparent for each of the peptides studied. Kinetic stability, expressed as t1/2 of dissociation, was calculated for each peptide and is shown in each panel of Fig. 2. Even peptides with extremely long half-lives on the class II I-Ad molecule showed dramatic losses in binding stability to the 81βH molecule. For example, the naturally processed self peptide CysC has an exceptionally long half-life on I-Ad, with a dissociation t1/2 of 190 h; on 81βH, this t1/2 is reduced to 14 h (Fig. 2,A). The HA antigenic peptide displays moderate stability on I-Ad, with a t1/2 of 45 h; this t1/2 is reduced to <1 h on the 81βH molecule (Fig. 2 C). Collectively, these studies reveal that loss of the potential to form a single hydrogen bond between the class II α helices and the peptide main chain profoundly accelerates the dissociation rate of peptides from the class II molecule.

FIGURE 2.

The impact of a single hydrogen bond loss on peptide dissociation kinetics. Purified WT I-Ad (○) or 81βH (•) class II molecules were allowed to bind fluorescein-labeled peptides at pH 5.3. The curves are single exponential fits to the dissociation data (r = 0.966–0.998). The dissociation t1/2 values are the average of three values obtained from single exponential fits to independent dissociation reactions; the reported errors are the SEM. Inset, dissociation curves of 81βH: peptides on shorter time scales.

FIGURE 2.

The impact of a single hydrogen bond loss on peptide dissociation kinetics. Purified WT I-Ad (○) or 81βH (•) class II molecules were allowed to bind fluorescein-labeled peptides at pH 5.3. The curves are single exponential fits to the dissociation data (r = 0.966–0.998). The dissociation t1/2 values are the average of three values obtained from single exponential fits to independent dissociation reactions; the reported errors are the SEM. Inset, dissociation curves of 81βH: peptides on shorter time scales.

Close modal

One particularly interesting pattern that emerged from the preceding analyses became apparent when we compared the magnitude of the effects of loss of the β-His-81 hydrogen bond on the stability of the binding of different peptides. For each peptide, we calculated an enhanced dissociation value, which is a measure of the acceleration of the dissociation rate due to the disruption of the hydrogen bond. This value was calculated by dividing the t1/2 dissociation rate observed with WT molecules by that displayed for the same peptide from the 81βH molecules. These values, presented graphically in Fig. 3, indicated that although all peptides tested showed dramatically enhanced dissociation rates when the hydrogen bond contributed by β-81 was disrupted, the magnitude of the enhancement varied among the peptides. An inverse correlation is observed between the inherent stability of the peptide-class II complex and the magnitude of the acceleration caused by disruption of 81βH. For example, a peptide that shows high stability on WT class II, such as CysC (t1/2 190 h), has only a 14-fold rate enhancement by destabilization of the β-His-81 hydrogen bond. The CLIP peptide, which has a relatively short dissociation t1/2 (13 h), displays an enhancement of dissociation of 170-fold (t1/2 0.08 h). Thus, loss of the hydrogen bond contributed by β-His-81 accelerates the dissociation rate by a factor that can be predicted from the inherent stability of an individual peptide-class II complex. Overall, the kinetic data presented here support a model in which destabilization of the hydrogen bond contributed by β-His-81 accelerates the formation of a shared, naturally occurring intermediate or transition state that is involved in spontaneous dissociation of all peptides from the class II molecule.

FIGURE 3.

Inverse correlation between inherent peptide-class II stability and sensitivity to loss of potential for the β-81 hydrogen bond. The top panel shows the dissociation t1/2 for WT I-Ad class II molecules with four peptides. The bottom panel shows the rate acceleration factor for 81βH for the same peptides. This factor is derived by dividing the dissociation t1/2 from WT by that displayed by 81βH, and thus represents the relative increase in dissociation rate due to destabilization of the β-81 hydrogen bond.

FIGURE 3.

Inverse correlation between inherent peptide-class II stability and sensitivity to loss of potential for the β-81 hydrogen bond. The top panel shows the dissociation t1/2 for WT I-Ad class II molecules with four peptides. The bottom panel shows the rate acceleration factor for 81βH for the same peptides. This factor is derived by dividing the dissociation t1/2 from WT by that displayed by 81βH, and thus represents the relative increase in dissociation rate due to destabilization of the β-81 hydrogen bond.

Close modal

One issue that is important in interpreting these data is whether the lack of the hydrogen bond contributed by β-His-81 causes a general change in the overall mode of binding of the peptides. To gain evidence that the complexes formed between WT I-Ad and 81βH are similar with regard to peptide register, we took advantage of a CLIP variant shown previously to have an improved P6 pocket interaction and to display an enhanced stability with I-Ad. The CLIP peptide was modified at amino acid 95 so that Pro was replaced by Ala (“P→A”), a variant for which molecular modeling and binding studies have demonstrated an improved P6 pocket interaction with I-Ad (17). Indeed, when this CLIP P→A-substituted peptide was tested for its kinetic stability on I-Ad, we found that the t1/2 was extended by a factor of 8, increasing its t1/2 from 13 h to 106 h (Fig. 4). The binding of CLIP P→A showed a similar gain in stability on 81βH from a t1/2 of 0.08 h for CLIP to almost 2 h for CLIP P→A. These data argue strongly that the register of the CLIP peptide in I-Ad is the same as that when it is bound to 81βH. The following observations also support the conclusion that the initial complexes formed between the WT and hydrogen bond-deficient I-Ad molecule are similar. The degree of pH sensitivity of a particular complex is the same for WT I-Ad and 81βH. For example, Eα dissociation is pH insensitive from both WT and 81βH, whereas CLIP dissociation is enhanced ∼5-fold at a pH of 5 compared with a pH of 7.4 for both class II molecules (data not shown). Also, the binding of peptides by 81βH is selective and is limited to peptides that can stably bind to I-Ad, arguing that the peptide-binding motifs used by the two molecules are likely the same (data not shown). Finally, the dissociation curves displayed by both WT class II and 81βH are monophasic (Fig. 2) and independent of the time allowed for complex formation, arguing that there is not significant heterogeneity in the peptide complexes formed by 81βH. These data collectively suggest that the peptide complexes initially formed between WT class II molecules and 81βH are similar with regard to peptide register and in overall conformation.

FIGURE 4.

Alteration of P6 pocket interactions increases the kinetic stability of CLIP on both WT I-Ad and 81βH. A variant (CLIP85–101) peptide was synthesized with Pro at position 95 substituted with Ala (CLIP P→A) to enhance side chain interactions with the P6 pocket of I-Ad. The dissociation t1/2 of the WT CLIP peptide and the CLIP P→A variant on WT I-Ad molecules (top panel) and 81βH (bottom panel) is shown.

FIGURE 4.

Alteration of P6 pocket interactions increases the kinetic stability of CLIP on both WT I-Ad and 81βH. A variant (CLIP85–101) peptide was synthesized with Pro at position 95 substituted with Ala (CLIP P→A) to enhance side chain interactions with the P6 pocket of I-Ad. The dissociation t1/2 of the WT CLIP peptide and the CLIP P→A variant on WT I-Ad molecules (top panel) and 81βH (bottom panel) is shown.

Close modal

The studies reported here demonstrate a profound contribution of the hydrogen bond between MHC class II β-His-81 and the peptide main chain toward the overall stability of interactions between the peptide and MHC class II molecules. All of the peptides analyzed showed a dramatic acceleration of their dissociation rates from the 81βH molecule. This effect is not predicted by an uncoupled or strictly additive model of class II-peptide interactions. In the crystal structures of two peptide-IAd complexes, 12–13 potential peptide-MHC hydrogen bonds of comparable lengths were identified. The best estimate for the energy contribution of solvent-exposed hydrogen bonds, such as the one between the β-His-81 side chain and the peptide backbone, is on the order of 0–2 kcal mol-1 (18). Assuming that the disruption of the β-His-81 hydrogen bond affects the binding affinity to the same degree as the dissociation rate, the 10- to 100-fold increase in dissociation rates that was observed for the 81βH molecule would correspond to a free energy change on the order of 2 kcal mol−1, because a factor of 10 change in affinity is equivalent to a free energy change of ∼1.4 kcal. If all 12–13 hydrogen bonds were of similar energy, hydrogen bonding alone would contribute 20–30 kcal mol−1 toward the binding of peptide to MHC molecules. However, estimates for peptide-MHC affinity are of the order of micromolar to nanomolar (19, 20, 21), a free energy of ∼7–11 kcal mol−1, ∼1 × 1010 smaller than that predicted by the binding energy summed from individual hydrogen bonds. Our results are thus not consistent with a strictly additive model in which the binding energy is uncoupled and equally contributed by the 12–13 individual hydrogen bonds.

The results presented here support a cooperative model in which individual bonds between class II and peptide are dependent upon the integrity of neighboring interactions. Destabilization of one bond will cause a rapid destabilization of adjacent, although chemically distinct, bonds. A cooperative model for peptide dissociation from MHC class II proteins is also consistent with our observation that the susceptibility of any given peptide to the 81βH mutation is inversely proportional to the stability of the peptide on WT I-Ad. If loss of the β-His-81 hydrogen bond decreased the binding energy by some constant value, each of the peptides would have been destabilized by the same amount. If, however, peptide dissociation is cooperative, loss of the β-His-81 hydrogen bond would destabilize other intermolecular interactions, and peptides with less favorable binding energies will be more affected. It has been shown that the susceptibility of peptide-class II interactions to DM-promoted dissociation is also related to the inherent stability of the peptide-class II complexes, with low stability peptides showing exquisite sensitivity, whereas high stability peptides are relatively resistant to DM (9, 11, 22, 23). Given that the enhanced rates observed for 81βH are comparable with DM-catalyzed rates for other class II molecules, we suggest that DM could function by stabilizing a peptide-MHC intermediate in which a hydrogen bond between the peptide and MHC, such as that contributed by β-His-81 hydrogen bond, is disrupted. Our findings thus support the recent speculation by Mosyak et al. (24) that the primary activity of DM may be the destabilization of one or two hydrogen bonds at the amino terminus of the peptide. This mechanism, in which a cooperative, progressive disruption of binding interactions precedes dissociation, engenders MHC class II complexes with an exquisite sensitivity to editing based on peptide kinetic stability. Although DM is thought to achieve this editing by changing the MHC conformation, our results for 81βH suggest that physiologically relevant enhanced peptide dissociation rates can be achieved through the disruption of a single, solvent exposed hydrogen bond in the absence of any initial global changes in MHC conformation.

We thank Patrick Noud for technical assistance with this work. We are also grateful to Jim Miller, Yair Argon, and Janis Burkardt for critical review of this manuscript.

1

This work was supported by National Science Foundation Grant MCB-9722374 (to C.B.) and by grants from the National Institutes of Health (R01AI34359 and P01DK 49799) and the Juvenile Diabetes Foundation (to A.J.S.). B.J.M. was supported by National Institutes of Health National Research Service Award Training Grant 5T32GM08268 from the National Institute of General Medical Science.

3

Abbreviations used in this paper: 81βH, MHC class II I-Ad protein mutated at position 81 in the β-chain; Ii, invariant chain; CLIP, class II associated Ii-derived peptide; WT, wild type; CysC, cystatin C.

1
Jardetzky, T..
1996
. Crystal structures of MHC class I and class II molecules. M. Browning, and A. McMichael, eds.
HLA and MHC Genes, Molecules, and Function
251
-276. BIOS Scientific Publishers, Oxford, U.K.
2
Tan, L. J., S. Ceman, J. Rodriques-Paris, T. Steck, A. J. Sant.
1997
. Late events in the biogenesis of MHC class II molecules are controlled by the 80–82 segment of the class II β-chain.
Eur. J. Immunol.
27
:
1479
3
Chervonsky, A. C., L. Gordon, A. J. Sant.
1994
. A segment of the MHC class II β-chain plays a critical role in targeting class II molecules to the endocytic pathway.
Int. Immunol.
6
:
973
4
Ceman, S., S. Wu, T. Jardetzky, A. J. Sant.
1998
. Alteration of a single hydrogen bond between class II and peptide results in rapid degradation of class II after Ii removal.
J. Exp. Med.
188
:
2139
5
Wolf, P. R., P. Roos, H. L. Ploegh, and A. J. Sant. 1999. Deviant trafficking of I-Ad mutant molecules is reflected in their peptide binding properties. Eur. J. Immunol. In press.
6
Denzin, L. K., P. Cresswell.
1995
. HLA-DM induces CLIP dissociation from MHC class II αβ dimers and facilitates peptide loading.
Cell
82
:
155
7
Kropshofer, H., A. B. Vogt, G. Moldenhauer, J. Hammer, J. S. Blum, G. J. Hammerling.
1996
. Editing of the HLA-DR-peptide repertoire by HLA-DM.
EMBO J.
15
:
6144
8
Sloan, V. S., P. Cameron, G. Porter, M. Gammon, M. Amaya, E. Mellins, D. Zaller.
1995
. Mediation by HLA-DM of dissociation of peptides from HLA-DR.
Nature
375
:
802
9
Vogt, A. B., H. Kropshofer, G. J. Hammerling.
1997
. How HLA-DM affects the peptide repertoire bound to HLA-DR molecules.
Hum. Immunol.
54
:
170
10
Sherman, M. A., D. A. Weber, P. E. Jensen.
1995
. DM enhances peptide binding to class II MHC by release of invariant chain-derived peptide.
Immunity
3
:
197
11
Weber, S. A., B. D. Evavold, P. E. Jensen.
1996
. Enhanced dissociation of HLA-DR-bound peptides in the presence of HLA-DM.
Science
274
:
618
12
Liang, M. N., C. Beeson, K. Mason, H. M. McConnell.
1995
. Kinetics of the reactions between the invariant chain (85–99) peptide and proteins of the murine class II MHC.
Int. Immunol.
7
:
1397
13
Kappler, J., B. Skidmore, J. White, P. Marrack.
1981
. Antigen-inducible, H-2-restricted, interleukin-2-producing T cell hybridomas.
J. Exp. Med.
153
:
1198
14
Hunt, D., H. Michel, T. Dickinson, J. Shabanowitz, A. Cox, K. Sakaguchi, E. Appella, H. Grey, A. Sette.
1992
. Peptides presented to the immune system by the murine class II major histocompatibility complex molecule I-Ad.
Science
256
:
1817
15
Bangia, N., T. H. Watts.
1995
. Evidence for invariant chain 85–101 (CLIP) binding in the antigen binding site of MHC class II molecules.
Int. Immunol.
7
:
1585
16
Scott, C. A., P. A. Peterson, L. Teyton, I. A. Wilson.
1998
. Crystal structures of two I-Ad-peptide complexes reveal that high affinity can be achieved without large anchor residues.
Immunity
8
:
319
17
Liang, M. N., C. Lee, Y. Xia, H. M. McConnell.
1996
. Molecular modeling and design of invariant chain peptides with altered dissociation kinetics from class II MHC.
Biochemistry
35
:
14734
18
Fersht, A. R..
1988
. Relationships between apparent binding energies measured in site-directed mutagenesis experiments and energetics of binding and catalysis.
Biochemistry
27
:
1577
19
Beeson, C., H. McConnell.
1995
. Reactions of peptides with class II proteins of the major histocompatibility complex.
J. Am. Chem. Soc.
117
:
10429
20
Babbitt, B., P. Allen, G. Matsueda, E. Haber, E. Unanue.
1985
. The binding of immunogenic peptides to Ia histocompatibility molecules.
Nature
317
:
359
21
Buus, S., A. Sette, S. Colon, D. Jenis, H. Grey.
1986
. Isolation and characterization of antigen-Ia complexes in T cell recognition.
Cell
47
:
1071
22
van Ham, S. M., U. Gruneberg, G. Malcherek, I. Broker, A. Melms, J. Trowsdale.
1996
. Human histocompatibility leukocyte antigen (HLA)-DM edits peptides presented by HLA-DR according to their ligand binding motifs.
J. Exp. Med.
184
:
2019
23
Vogt, A. B., H. Kropshofer, G. Moldenhauer, G. J. Hammerling.
1996
. Kinetic analysis of peptide loading onto HLA-DR molecules mediated by HLA-DM.
Proc. Natl. Acad. Sci. USA
93
:
9724
24
Mosyak, L., D. M. Zaller, D. C. Wiley.
1998
. The structure of HLA-DM, the peptide exchange catalyst that loads antigen onto class II MHC molecules during antigen presentation.
Immunity
9
:
377