Small Organic Compounds Enhance Antigen Loading of Class II Major Histocompatibility Complex Proteins by Targeting the Polymorphic P1 Pocket*♦

Major histocompatibility complex (MHC) molecules are a key element of the cellular immune response. Encoded by the MHC they are a family of highly polymorphic peptide receptors presenting peptide antigens for the surveillance by T cells. We have shown that certain organic compounds can amplify immune responses by catalyzing the peptide loading of human class II MHC molecules HLA-DR. Here we show now that they achieve this by interacting with a defined binding site of the HLA-DR peptide receptor. Screening of a compound library revealed a set of adamantane derivatives that strongly accelerated the peptide loading rate. The effect was evident only for an allelic subset and strictly correlated with the presence of glycine at the dimorphic position β86 of the HLA-DR molecule. The residue forms the floor of the conserved pocket P1, located in the peptide binding site of MHC molecule. Apparently, transient occupation of this pocket by the organic compound stabilizes the peptide-receptive conformation permitting rapid antigen loading. This interaction appeared restricted to the larger Glyβ86 pocket and allowed striking enhancements of T cell responses for antigens presented by these “adamantyl-susceptible” MHC molecules. As catalysts of antigen loading, compounds targeting P1 may be useful molecular tools to amplify the immune response. The observation, however, that the ligand repertoire can be affected through polymorphic sites form the outside may also imply that environmental factors could induce allergic or autoimmune reactions in an allele-selective manner.

Major histocompatibility complex (MHC) molecules are a key element of the cellular immune response. Encoded by the MHC they are a family of highly polymorphic peptide receptors presenting peptide antigens for the surveillance by T cells. We have shown that certain organic compounds can amplify immune responses by catalyzing the peptide loading of human class II MHC molecules HLA-DR. Here we show now that they achieve this by interacting with a defined binding site of the HLA-DR peptide receptor. Screening of a compound library revealed a set of adamantane derivatives that strongly accelerated the peptide loading rate. The effect was evident only for an allelic subset and strictly correlated with the presence of glycine at the dimorphic position ␤86 of the HLA-DR molecule. The residue forms the floor of the conserved pocket P1, located in the peptide binding site of MHC molecule. Apparently, transient occupation of this pocket by the organic compound stabilizes the peptide-receptive conformation permitting rapid antigen loading. This interaction appeared restricted to the larger Gly ␤86 pocket and allowed striking enhancements of T cell responses for antigens presented by these "adamantyl-susceptible" MHC molecules. As catalysts of antigen loading, compounds targeting P1 may be useful molecular tools to amplify the immune response. The observation, however, that the ligand repertoire can be affected through polymorphic sites form the outside may also imply that environmental factors could induce allergic or autoimmune reactions in an allele-selective manner.
The ligands of class II MHC 6 molecules are generated mostly from exogenous protein sources (1). After internalization, they are fragmented by proteases and loaded onto class II MHC molecules in a process catalyzed by the chaperone HLA-DM (2,3). Processing and loading take place in a dedicated endosomal compartment (MCII vesicle) (4). After transport to the cell surface the peptide⅐MHC complexes are presented to CD4ϩ T cells, which upon recognition induce the antigen-specific immune response.
On the cell surface some of these peptides are displayed for up to 100 h (5). A considerable fraction of the peptide⅐MHC molecules, however, dissociates during presentation, so that also "empty" MHC molecules are always present on class II MHC expressing cells. They are particularly abundant on immature dendritic cells, where they seem to play a role in cell membrane-associated antigen processing by capturing of extracellular antigens (6,7). In general, however, the ligand composition of MHC molecules should reflect the protein content of the cell rather than that of the environment. Uncontrolled antigen capture by empty MHC molecules on the cell surface therefore has to be avoided. Presumably as a safeguard mechanism class II MHC molecules inactivate rapidly after dissociation of the ligand. They convert into a "non-receptive" state (8), which is characterized by the inability of the MHC molecule to bind new peptide ligands.
While, in principle, this inactivation is reversible, reconversion into the peptide-receptive state is extremely slow. In previous studies we have shown that certain organic compounds, such as aliphatic alcohols and phenol derivatives, have an intrinsic capacity to accelerate the peptide loading of class II MHC molecules (9). Although their catalytic activity is rather weak, the effect is mediated nonetheless by a defined mechanism (10). While we could establish that these organic compounds are able to re-induce the peptide-receptive state, the molecular basis of this transition remained unknown. A screening of a library of ϳ20.000 organic compounds has now revealed a number of new catalytic molecules that accelerate peptide loading at substantially higher rates. Most importantly, however, the analysis of their catalytic activity revealed a striking allele selectivity, which allowed identifying their putative binding site.
Peptide Loading/Release of Soluble HLA-DR Molecules-Detection of the peptide⅐MHC complex was carried out by ELISA using a monoclonal ␣-HLA-DR capture antibody (L243, ATCC) and Eu 3ϩ -labeled streptavidin (DELFIA, Wallac) as described (10). Peptide loading was determined with 1 M HLA-DR and 100 g/ml biotinylated peptide (phosphate-buffered saline, pH 7.4, 37°C, 1 h). Curve fit for kinetic binding experiments was carried out as hyperbola regression (f(x) ϭ (ax)/(b ϩ x)) using the SigmaPlot 8.0 software. For ligand release experiments 1.6 M HLA-DR1 molecules were loaded overnight with biotinylated IC106-120 peptide, then diluted 1:3 and incubated with 200 g/ml HA306-318 in the absence or presence of 1 mM AdEtOH.
Peptide Loading of Cell Surface MHC Molecules-5 ϫ 10 4 HLA-DR-expressing cells/well were incubated at 37°C in Dulbecco's modified Eagle's medium, 5% fetal calf serum in a 96 well U-bottom plate with biotinylated peptide in the presence of catalytic compounds. After 4 h cells were washed and stained with streptavidin-PE alone or double-stained with ␣-HLA-DR-PE/streptavidin-APC and analyzed by FACS on a FACSCalibur or FACSCanto instrument (BD Biosciences). Cells were gated using a life-gate after staining with propidium iodide.
T Cell Assay-5 ϫ 10 4 HLA-DR-expressing cells/well were loaded for 4 h with peptides in the presence of catalytic compounds as described above. Cells were then washed, and 5 ϫ 10 4 T cells were added to the culture, which was then incubated for 24 h in Dulbecco's modified Eagle's medium, 5% fetal calf serum in 96-well U-bottom plates. The T cell response was determined in a secondary assay with CTL-L cells (ATCC) as described previously (9).
Docking and Energy Minimization-AdCaPy was docked by hand, with the support of the "Dock" module in Sybyl 6.92 software (Tripos Inc.), into the P1 pocket of the x-ray structures HA306-318⅐DRB1*0101 and of MBP86-100⅐DRB1*1501 (Protein Data Bank entry codes 1DLH and 1BX2, respectively). The minimizations were made with MOE (Chemical Computing Group Inc.) using the MMFF94x force field. The default protonation state and the charges were calculated. Water molecules in the x-ray structures were removed. All atoms within 9 Å from the AdCaPy ligand were minimized to an energy root mean square gradient of 0.05. The reported root mean square deviation values were calculated by MOE.

Adamantane Derivatives Are Effective Catalysts for Peptide
Loading of HLA-DR1 Molecules-The catalytic activity of organic molecules can be determined in vitro in a peptide loading assay with soluble class II MHC molecules (10). In this assay the amount of biotinylated peptide loaded onto the MHC molecule is determined by ELISA using HLA-DR-specific capture antibodies together with Eur3ϩ-streptavidin. DRB1*0101 is an allelic variant of a human class II MHC proteins present in ϳ20% of the Caucasian population. 7 In a high throughput screening a soluble version of this MHC molecule was used to test the influence of ϳ20.000 synthetic compounds on the loading with the high affinity peptide HA306-318 (data not shown). Less than 0.5% of the compounds showed any loading enhance-  DECEMBER 15, 2006 • VOLUME 281 • NUMBER 50 ment, but molecules carrying the bulky adamantyl group were frequently found to be active. Some examples are shown in Fig.  1. Within this set highest activity was detected for AdEtOH and AdCaPy. Nearly the same activity was determined for a adamantyl compound substituted with benzene-sulfonamide (AdBeSA) followed by AdCDME and AdPr. Although the chemical nature of the side chain had apparently some influence, the catalytic activity was largely determined by the adamantyl group itself. The introduction of additional methyl substitutions at the adamantyl core structure could completely abrogate its activity (3M-AdCDME).

Peptide Loading of MHC II by Small Molecules Targeting P1
We have shown before that the interaction of the MHC molecule with organic compounds is fully reversible and based on the induction of a peptide receptive state (10). The same applies also for the activity of adamantyl compounds. AdEtOH can therefore efficiently trigger the exchange of MHC-bound peptides with free peptides of higher affinity ( Fig. 2A). IC106-120 was removed from soluble DRB1*0101 in less than 6 h, while almost no exchange is observed in the absence of the catalyst. No removal of IC106-120 was evident in the absence of free HA306-318, indicating that IC106-120 ligand was indeed replaced by the high affinity peptide.
The induction of the receptive state was also evident in enhanced peptide loading (Fig. 2B). Without any catalysts more than 20 h were needed to reach half-maximal loading of DRB1*0101 with the HA306-318 peptide. The presence of AdEtOH or AdCaPy, however, reduced t1 ⁄ 2 to ϳ30 min. The enhancement of the on-rate was clearly dose-dependent (Fig.  2C) and up to a concentration of 1 mM compound did not show saturation (Fig. 2D).
Allele Selectivity of Adamantyl Compounds-Catalytic organic compounds can be used not only with recombinant MHC proteins but also with living cells to mediate loading with peptides and T cell antigens (9,10,15). As shown in Fig.  3A, AdEtOH and AdCaPy enhanced the loading of two cell surface HLA-DR molecules DRB1*0101 and DRB1*0401 with HA306-318. Loading was in fact much more effective than with pCP, an aromatic compound previously shown to exhibit some catalytic activity on HLA-DR molecules (10). No loading was detected on fibroblast cells that do  not express any class II MHC molecules, demonstrating HLA-DR specificity of the adamantyl-mediated loading enhancement.
Extension of the experiments to cells expressing other allelic variants of HLA-DR, however, produced unexpected results (Fig. 3B). MGAR cells are Epstein-Barr virus-transformed B cells which naturally express the HLA-DR variants DRB1*1501 and DRB5*0101 and IC106-120 and MBP86-100 are known to bind to these cells. When loading MGAR cells with the two peptides enhancement by AdEtOH and AdCaPy was indeed observed for IC106-120 (left panel). Notably, however, the two adamantyl compounds completely failed to enhance loading with MBP86-100, while pCP was able to increase the loading of both peptides (middle panel). The effect was not due to the MBP86-100 peptide, since loading of DRB1*1602-expressing RML cells could be improved by AdEtOH (right panel).
To dissect this phenomenon, experiments were repeated with fibroblast cells each expressing only one of the two HLA-DR molecules present on MGAR cells (Fig. 3C). IC106-120 binds with higher affinity to DRB5*0101, since virtually no binding was detected on DRB1*1501 expressing cells. MBP86-100, on the other hand, preferentially binds to DRB1*1501 and only weakly to DRB5*0101. pCP enhanced the binding of both peptides to the respective HLA-DR molecule but AdEtOH enhanced only the binding of IC106-120 on DRB5*0101. No enhancement was detected for the binding of MBP86-100 to DRB1*1501, suggesting that DRB1*1501 was "non-susceptible" to adamantyl-mediated catalysis.
Adamantyl Susceptibility Correlates with Residue Gly ␤86 -The result of the previous experiment implicated that adamantyl compounds exhibit catalytic activity only on a subset of HLA-DR molecules. A comparison of catalytic activity with the allelic variations in fact suggested a correlation of susceptibility with a well known dimorphism at position 86 of the ␤-chain (16). In HLA-DR molecules this position is represented either by glycine (Gly ␤86 ) or valine (Val ␤86 ) (17). All MHC molecules susceptible to adamantyl compounds expressed glycine at this position, while DRB1*1501, the only non-susceptible molecule identified so far, expressed valine at ␤86 (Table 1).
Allele-selective Enhancement of the CD4ϩ T Cell Response-The major function of class II MHC molecules is to present peptide antigens to CD4ϩ T cells. Improved antigen loading by organic compounds therefore directly increases the sensitivity of the antigen-specific CD4ϩ T cell response (9, 10, 15). On DRB1*0101 the presence of AdEtOH during antigen loading can lead to shifts of the dose-response curves of almost 2 orders of magnitude (Fig. 5A). Without any catalyst the threshold concentration for the in vitro T cell response against the influenza virusderived HA306-318 antigen was slightly above 10 ng/ml peptide. 50 M AdEtOH reduced the detection limit to 1 ng/ml peptide, and 250 M AdEtOH further shifted the threshold to almost 0.1 ng/ml.
The effect by AdEtOH, however, is evident only when the antigen is presented by a susceptible HLA-DR molecule. As shown in Fig. 5B, enhancement is evident for the HA306-318-specific CD4ϩ T cell response when presented by Gly ␤86 -expressing DRB1*0101 and DRB1*0401 molecules. The same applies also for the MBP86-100-specific response restricted by DRB1*1502. Abso-  DECEMBER 15, 2006 • VOLUME 281 • NUMBER 50

JOURNAL OF BIOLOGICAL CHEMISTRY 38539
lutely no enhancement is observed when the antigen is presented by DRB1*1501. The two DRB1*15 variants differ only in the substitution at ␤86 (DRB1*1502, Gly ␤86 ; DRB1*1501, Val ␤86 ) and both are able to present the antigen to the 2E12 T cell hybridoma. Due to the allele selectivity of AdEtOH, however, enhancement is evident only when the T cell recognizes the antigen on the susceptible DRB1*1502 molecule.

DISCUSSION
In this study we showed that low molecular weight compounds exist, which are able to enhance the immune response in an allele-specific way. By targeting a polymorphic binding site on the class II MHC molecule they induce a conformational transition that allows the rapid ligand exchange. The presence of these compounds during antigen loading can therefore result in dramatic enhancements of antigen-specific T cell responses. The effect is limited to immune responses restricted by "susceptible" MHC molecules. Sitedirected mutagenesis revealed that this susceptibility correlates with the presence of a glycine residue at position ␤86 of the class II MHC molecule. The residue is located in the peptide binding site and forms the floor of the conserved pocket P1 (18) (Fig. 6A). At least for the adamantyl compounds P1 therefore seems to be the target site to exhibit their catalytic effect.
In complex with the peptide, P1 accommodates an anchor side chain of the peptide ligand (19 -21). Residue ␤86 restricts the depth of P1 (18,22), so that Val ␤86 pockets can bind only small aliphatic anchor side chains (Ile, Leu, Val, and Met), whereas deeper Gly ␤86 -containing pockets accommodate also larger aromatic residues (Phe, Tyr, and Trp) (Fig. 6B). Compared with the smaller pCP molecule, which enhances peptide loading irrespective of ␤86, the spherical adamatyl group requires significantly more space (Fig. 6C). Allele-selective enhancement by adamantyl compounds might therefore be due to the increased space requirements. Computational docking and energy minimization calculations indicated that only the Gly ␤86 -containig P1 pocket is in fact large enough for the bulky adamantyl group. While in a simulation the adamantyl group of AdCaPy was pushed out of the shallow Val ␤86 -P1 pocket of DRB1*1501 by 1.7 Å (data not shown), it remained stably inside the Gly ␤86 pocket of DRB1*0101 (Fig.  6D). The van der Waals radius of the adamantyl group is only slightly bigger than the inner surface of a Gly ␤86 -P1 pocket harboring an aromatic anchor side chain. Energy minimization of the DRB1*0101⅐AdCaPy complex resulted in a root mean square deviation of 0.34 Å for the backbone of both ␣-helices lining the P1 pocket and of only 0.56 Å for the side chains surrounding the docked adamantyl group (Fig. 6E).
Thus, ligand-exchange catalysis by adamantyl compounds seems indeed to require the occupation of P1. The catalytic effect of organic compounds is mediated by the induction of the peptide-receptive state (10), and other groups have shown that mutant DRB1*0101 molecules, which express permanently filled P1 pockets due to a Gly ␤86 3 Tyr substitution, are in fact mostly in the receptive state (23). Hence, occupation of P1 by the adamantyl group seems to have the same effect except that transient binding allows replacement by a peptide to form the stable ligand complex. In a mechanistic model, the effect could be explained by assuming that the receptive form is correlated with an open P1 pocket while the non-receptive state is linked to a collapsed P1 pocket (Fig. 7). Based on this assumption occupation of P1 by adamantyl compounds should stabilize the receptive state, resulting in fast loading rates as the result of an increased pool of peptide-accessible receptive MHC molecules.
Similar interactions might also be the basis of the catalysis by other organic compounds such as pCP. Invariance of their catalytic activity with regard to ␤86, however, leaves it open whether P1 is in fact the target. Although previous studies suggested that H-bond donor groups are crucial for certain organic compounds (9,10), this does apparently not apply for adamantyl derivatives. The structural dissimilarity of the substituents of AdEtOH (hydroxy-ethyl group) and AdCaPy (substituted pyrazole ring) rather suggests that the hydrophobic adamantyl group by itself is largely responsible for the effect.
In contrast to noble metal complexes (24), which seem to strip peptides from class II MHC molecules by irreversibly inactivating the receptor complex, the interaction with organic compounds is reversible (10). This applies also for some other organic compounds, recently identified to bind to the chaperone HLA-DM (25). Similar to the HLA-DR-targeting compounds, they are also able to accelerate class II MHC loading, but their practical use for cell loading appears limited, since they require acidic pH and the presence of the chaperone. Although the development of catalytic organic compounds is still at the beginning, they are likely to become potent amplifiers of the immune response. As "MHC loading enhancer" (MLE), they represent useful tools to improve antigen recognition in a variety of diagnostic and therapeutic settings. This includes the peptide loading of immobilized or multimeric MHC molecules (e.g. as HLA-DR tetramer (26)) or the transfer of antigens onto APC for in vitro T cell assays or immune therapies. Since they could enhance also the loading rate of a peptide or protein vaccine in vivo, they might even become useful vaccine additives.
Adamantyl groups are present in a number of prescription drugs.
Most prominent examples are "rimantadine" (1-(1-adamantyl)ethanamine) and "amantadine" (adamantane-1-amine), which are both used to treat influenza A virus infections. Their mechanism is not fully understood, but involvement of MLE activity seems unlikely, since in particular "amantadine" has low ligandexchange capacity (data not shown). While in this case MLE activity is probably not involved in their primary effect, catalytic antigen loading could still be a cause of unwanted side effects. Autoimmune diseases such as type 1 diabetes, rheumatoid arthritis, and multiple sclerosis (MS) are induced by the "accidental" recognition of self-antigens. The addition of the weak catalyst pCP to preparations of crude spinal cord homogenate can activate encephalitogenic T cells specific for the self-antigen MBP86-100 (15). MBP86-100 has been implicated with human MS and is known to induce an MS-like autoimmune disease in mice, and as shown here, AdEtOH amplified the T cell response against this epitope even more effectively. Most importantly, however, enhancement of the in vitro autoimmune response is evident only on susceptible HLA-DR molecules.
In principle, environmental factors acting similar to AdEtOH could provoke the induction of autoimmune diseases in an allele-selective manner. It is in fact a typical characteristic of most autoimmune diseases that the prevalence is correlated with particular allelic variants of class II MHC (27). In some cases the susceptibility could be even linked to polymorphic P1-like pockets. This applies for type 1 diabetes, which is correlated with position ␤57 in pocket P9 (28), for rheumatoid , and the right panel shows an overlay of the two pockets. The pockets are occupied by side chains of the peptide ligand (tyrosine for the Gly ␤86 -P1 pocket, valine for the Val ␤86 -P1 pocket). Images were generated from crystal structures of HA306-318⅐DRB1*0101 and of MBP86-100⅐DRB1*1501 (Protein Data Bank entry codes 1DLH and 1BX2, respectively). C, space-filling structures of pCP, AdEtOH, and AdCaPy. D, structure of the AdCaPy⅐DRB1*0101 complex after computational energy minimization. Key residues forming the inner surface of the P1 pocket are presented as ball-and-sticks; the AdCaPy ligand is presented in space-fill mode. E, superposition of the AdCaPy⅐DRB1*0101 complex before and after minimization. Red surface and orange ball-and-sticks represent the complex before minimization, and the green surface and light green balland-sticks indicate the position after minimization. arthritis, where the amino acid composition of the P4 pocket seems to be crucial (29,30) and, at least for a cohort of Australian patients, for MS, where susceptibility correlated with Val ␤86 of the P1 pocket (31). All these pockets accommodate anchor residues and participate in peptide selection but, as shown here for P1, might also represent specific target sites for ligand-exchange catalysts. Active compounds could derive from a variety of sources including drugs, chemicals, metabolites, and other biological/environmental toxins. A vast number of organic compounds has been implicated with the induction of autoimmune diseases (32), and it remains to be seen whether any of these molecules can function as MLE compound.
In summary, our experiments revealed that low molecular weight compounds exist that influence the immune response in an allele-specific way. By transiently occupying a dimorphic pocket on the HLA-DR molecule they seem to prevent the rapid inactivation of the complex so that free peptides can reach the vacant peptide binding site. Other natural, synthetic, or environmental compounds might exist that utilize pockets like P1 even more effectively. Their identification and the characterization of their target sites should produce not only a novel set of molecular tools to improve antigen loading, it might also provide new insights on the induction of autoimmune reactions and on the molecular mechanism of antigen loading. FIGURE 7. Mechanistic model. Peptide loading of class II MHC molecules is a multistep process, in which the conversion of the non-receptive state into the unstable peptide-receptive state is rate-limiting (33,34). In this model the peptide-receptive conformation needed for antigen loading is correlated with an open P1 pocket, while the pocket is collapsed in the non-receptive state. A filled P1 pocket should therefore result in the stabilization of the receptive state. Based on this transient occupation of P1 by the adamantyl compound should increase the number of peptide-accessible MHC molecules by preventing the re-conversion into the non-receptive state (indicated by a collapsed peptide binding site).