HLA-DP, HLA-DQ, and HLA-DR Have Different Requirements for Invariant Chain and HLA-DM*

The MHC is central to the adaptive immune response. The human MHC class II is encoded by three different isotypes, HLA-DR, -DQ, and -DP, each being highly polymorphic. In contrast to HLA-DR, the intracellular assembly and trafficking of HLA-DP molecules have not been studied extensively. However, different HLA-DP variants can be either protective or risk factors for infectious diseases (e.g. hepatitis B), immune dysfunction (e.g. berylliosis), and autoimmunity (e.g. myasthenia gravis). Here, we establish a system to analyze the chaperone requirements for HLA-DP and to compare the assembly and trafficking of HLA-DP, -DQ, and -DR directly. Unlike HLA-DR1, HLA-DQ5 and HLA-DP4 can form SDS-stable dimers supported by invariant chain (Ii) in the absence of HLA-DM. Uniquely, HLA-DP also forms dimers in the presence of HLA-DM alone. In model antigen-presenting cells, SDS-stable HLA-DP complexes are resistant to treatments that prevent formation of SDS-stable HLA-DR complexes. The unexpected properties of HLA-DP molecules may help explain why they bind to a more restricted range of peptides than other human MHC class II proteins and frequently present viral peptides.

MHC class II molecules play an important role in the immune system. They are essential in the defense against infection and are a main consideration in transplantation medicine. In addition to presenting antigenic peptides from predominantly extracellular sources to CD4 ϩ T cells, MHC class II molecules also mediate the thymic selection of helper T cells. MHC class II molecules consist of an ␣ and ␤ chain and are transported to endosomal-lysosomal compartments by the invariant chain (Ii). 2 The Ii is degraded until only a small fragment, dubbed CLIP, remains bound in the peptidebinding groove. Lysosomal pH and the class II-like molecule HLA-DM promote the exchange of the CLIP fragment for more stably binding antigenic peptides (1).
In humans, MHC class II molecules are encoded by three different loci, HLA-DR, -DQ, and -DP, which display ϳ70% similarity to each other. Polymorphism is a notable feature of MHC class II genes. For HLA-DR, most variability comes from DRB, with Ͼ700 known alleles at population level, whereas there are only three DRA variants. In contrast, both chains of HLA-DQ and -DP are polymorphic (2). For HLA-DP, however, only a few alleles are prevalent, most notably the heterodimer DPA1*0103/DPB1*0401 (DP401) (3).
Despite the essential function of MHC class II molecules in immune defense against pathogens, some alleles are frequently linked to immune diseases. For example, HLA-DR1 and DR4 predispose for rheumatoid arthritis, type 1 diabetes, and systemic lupus erythematosus, whereas DR2 confers susceptibility to multiple sclerosis. Similarly, DQ2 and DQ8 are linked to celiac disease (4,5). The role of HLA-DP in immune dysfunction has been less well defined. However, DP0201 is a risk factor for the autoimmune disease myasthenia gravis in the Japanese (6), and DP alleles with a glutamic acid at position 69 are associated with berylliosis, a hard metal lung disease (7). Although presentation of intracellular antigens by MHC class II molecules is considered atypical, HLA-DP4 gene products frequently present viral peptides, for example from HIV envelope protein, rabies virus, and hepatitis B virus envelope protein (8,9). DR1 (DRA, DRB1*0101) was the first MHC class II molecule to be crystallized (10), and HLA-DR is the most intensively studied MHC class II isotype. Efficient peptide presentation by HLA-DR is well recognized to depend on both Ii and DM. Indeed, biochemical studies suggested that HLA-DR alleles that bind inefficiently to the Ii CLIP fragment are more likely to induce an autoimmune response, for example in rheumatoid arthritis (11). Weak affinity of the Ii for DQ has also been associated with juvenile dermatomyositis (12). Structural information has been obtained for some HLA-DQ molecules involved in autoimmune disease; for example, crystal structures of the DQ8-insulin peptide complex (13) and the DQ2-gluten peptide complex have been solved (14). Although SDS-stable DQ molecules have been visualized (15), and Ii supports assembly of the DQ-like H-2A protein in the mouse (16), the relative contributions of DM and Ii in the acquisition of stable DQ␣␤ dimers have not been fully explored. The first crystal structure of an HLA-DP protein, HLA-DP2, has been recently published, in complex with a self-peptide from the DP␣ chain (17). HLA-DP molecules bind a limited set of peptides (18), but the relative lack of molecular and biochemical studies on HLA-DP means that exactly how it acquires peptides is unclear.
The organization and expression of the MHC, particularly of DP-like genes, vary greatly among mammals, making comparative study of DP function in model animals difficult. In mice, which lack functional DP paralogs, I-E and I-A are con-sidered the operative homologs of DR and DQ, respectively. Unlike HLA-DQ and DR, HLA-DPB1 sequences from humans, macaques, and great apes group into distinct lineages, suggesting that DP evolution has occurred after speciation (19).
To overcome the limitations of animal models with respect to HLA-DP biochemistry, we have employed a human cell culture system to compare the assembly and trafficking of DP with DQ and DR directly. Notably, in an identical cellular environment, DR, DQ, and DP have different requirements for Ii and DM. Our results suggest that trafficking and peptide loading of different MHC class II molecules can be modulated by tuning the level of DM and Ii in APCs, and our data have implications for the role of HLA-DP in autoimmune disease.

EXPERIMENTAL PROCEDURES
Cell Lines and Antibodies-Human cervical carcinoma HeLa cells and human melanoma MelJuso cells were maintained in minimum Eagle's medium or DMEM (Invitrogen), respectively, supplemented with 8% fetal calf serum (Sigma), 2 mM GlutaMAX, 100 units/ml penicillin, and 100 g/ml streptomycin (Invitrogen). Daudi cells (DPA1*010301, 020101 and DPB1*020102, 0802) were maintained in RPMI 1640 medium with the above supplements. MelJuso cells were typed by the National Health Service Blood and Transplant Unit (Newcastle, UK) to be homozygous for DPB1*1301.
The mAbs 1B5 against DR␣ and HC10 against MHC class I, and the polyclonal antisera against DR␤, DQ, and DP were a gift from Prof. J. Neefjes (Netherlands Cancer Institute, Amsterdam, The Netherlands). The polyclonal DP serum predominantly recognizes the DP␤ chain and only weakly recognizes the DP␣ chain when DP␣␤ are co-expressed. The mAb against the Ii (PIN.1) and mAbs HL40 (anti-DR␤/DP␤), HL37 (anti-DQ␤), and KUL/05 (anti-class II␤) were purchased from Abcam. The anti-DQ mAbs L2 and SPV-L3 were a kind gift from Prof. J. Robinson (Newcastle, UK).
Transfections-Transfections were done with Lipofectamine 2000 (Invitrogen) or FuGENE HD (Roche Applied Science) according to the manufacturers ' instructions. For Lipofectamine transfection, subconfluent cells in 6-cm dishes were washed with Hanks' balanced salt solution and Opti-MEM and transfected with 1 g of DNA for 6 h in the presence of OptiMEM serum-free medium. After 6 h, the cells were washed and placed back in normal growth medium. For FuGENE HD transfections, the transfection mix was added to the medium. The cells were analyzed 24 h after transfection, and expression of all chains was confirmed by Western blotting. Where indicated, cells were incubated with leupeptin (15 M), NH 4 Cl (20 mM), or vehicle control 1 h before transfection until lysis.
Immunostaining-Cells grown on coverslips were fixed in 4% paraformaldehyde in PBS for 10 min and were either left untreated, or were incubated in 0.2% Triton X-100 in PBS to permeabilize the cells. After blocking in 0.2% BSA in PBS, the cells were incubated with primary antibodies for 1 h, washed in 0.2% BSA/PBS, and incubated with fluorescently labeled secondary antibodies for 1 h (Alexa Fluor 488; Invitrogen). The nuclei were stained with DAPI before mounting the coverslips with Vectashield (Vector Laboratories). Images were taken on an Axio imager.M1 with OpenLab software.
Immunoprecipitations and Western Blotting-Cells and transfectants were lysed on ice with lysis buffer (1% Triton X-100, 50 mM Tris-HCl, pH 8.0, 150 mM NaCl, 5 mM EDTA, protease inhibitor mixture (1 g/ml antipain, chymostatin, leupeptin, pepstatin A)). Postnuclear lysates were incubated with protein A-Sepharose beads (Sigma/Amersham Biosciences) and antibodies for 1-2 h at 4°C. After extensive washing of the beads, immunoprecipitated proteins were eluted by boiling in sample buffer and analyzed by 12% SDS-PAGE. Proteins were transferred to polyvinylidene difluoride membranes (Millipore) at 150 mA for 2 h. The membranes were blocked in Tris-buffered saline Tween (TBST) with 8% milk, followed by incubation with primary antibody. After washing three times with TBST, the membranes were incubated with HRP-conjugated secondary antibodies (DAKO), washed, and visualized by ECL (Amersham Biosciences) and exposure to film (Kodak). Protein markers were from Bio-Rad.
SDS Stability Assay-Lysates were 1:1 diluted with 2 ϫ reducing sample buffer (4% SDS, 20% glycerol, 120 mM Tris-HCl-pH 6.8, bromphenol blue, 100 mM DTT). Half of the sample was left at room temperature, and half of the sample was boiled for 5 min, before analysis by SDS-PAGE and Western blotting as described above. These experiments were repeated 2-5 times with reproducible results.

RESULTS
Reconstitution of HLA-DR, -DQ, and -DP-We and other investigators have shown that different non-APCs can be reconstituted with a functional MHC class II compartment by transfection of HLA-DR molecules, the Ii and DM (20 -22). The advantages of this system are that folding, assembly, and trafficking of different MHC class II isotypes can be compared in the same cell line and that the contribution of individual proteins (Ii, DM, and class II ␣␤ chains) can be examined in the absence of competing endogenous proteins. To assess whether the different MHC class II isotypes, HLA-DR, -DQ, and -DP would all assemble in non-APCs, we transfected HeLa cells with different combinations of ␣, ␤, and Ii constructs. Lysates were subjected to immunoprecipitation with 1B5 (DR), HL37 (DQ), or HL40 (DP). 1B5 efficiently immunoprecipitated the DR␣ chain (Fig. 1A, lanes 3 and 4) and coimmunoprecipitated DR␤ and the Ii ( Ii Is Required for Endosomal-Lysosomal Deposition of HLA-DP, -DQ, and -DR-Previous studies have shown that endosomal-lysosomal deposition of HLA-DR is critically dependent on the Ii (23). To determine whether HLA-DQ and -DP are equally dependent on the Ii for intracellular localization we used immunofluorescence microscopy (Fig. 2). HeLa cells were transfected with ␤ chain only, or in combination with the ␣ chain, Ii, or DM. The cells were fixed and either left unpermeabilized (ϪTx, to demonstrate cell surface deposition), or were permeabilized (ϩTx, to show intracellular distribution) before immunostaining with HL40 (DR and DP) or HL37 (DQ). Single ␤ chains showed a typical ER staining (Fig.  2, A-C, ϩTx) and were not detected at the plasma membrane (ϪTx). When the ␣ chain was co-transfected with the ␤ chain, HLA-DR, -DQ, and -DP were still localized mainly in the endoplasmic reticulum (Fig. 2, A-C, ϩTx); however, cell surface expression was observed on nonpermeabilized cells (Fig. 2, A-C, ϪTx). This indicates that a proportion of ␣␤ dimers was able to reach the cell surface. Co-expression of the Ii with ␣␤ dimers, however, resulted in a punctate staining indicative of endosomal-lysosomal localization (Fig. 2, A-C, ϩTx), similar to that seen in professional APCs. Quantification of repeat experiments showed that of ϳ70 cells counted, 80% (DP), 93% (DR), and 86% (DQ) of ␣␤ϩIiϩDM-expressing cells were positive for the appearance of endosomal-lysosomal structures, as expected from previous experiments with DM (20). DP, DQ, and DR were all able to rescue the known "swollen endosomal-lysosomal vesicle" phenotype that arose when cells expressed Ii alone (supplemental Fig. 1) (24). In the presence of the Ii, MHC class II molecules were readily observed at the cell surface (Fig. 2, A-C, ϪTx). Additional expression of DM did not change the intracellular and cell surface expression of MHC class II molecules (Fig. 2, right  panels). Thus, MHC class II ␣␤ dimers of all subtypes require the Ii to reach the endosomal-lysosomal system. Cell surface expression of HLA-DR, -DQ, and -DP in the absence of the Ii is consistent with previous observations (23,(25)(26)(27). The result is not likely to be an artifact of overexpression because expression levels were similar to or lower than those found in APCs (20), and the ␤ chain alone did not reach the cell surface. Thus, not only DR, but also DQ and DP interacted physically ( Fig. 1) and functionally ( Fig. 2) with Ii.
Differences in the Formation of SDS-stable HLA -DP, -DQ, and -DR Dimers-The acquisition of a stable binding peptide results in a conformational change that renders HLA-DR␣␤ dimers resistant to dissociation in SDS-containing sample buffer at room temperature (28). To compare the assembly of HLA-DR (DR1) with HLA-DQ (DQ5) and -DP (DP4), we made use of this well established SDS stability assay. We cotransfected MHC class II ␣ and ␤ chains with different combinations of Ii and DM. The expression levels of Ii and DM were confirmed by Western blotting (not shown). To analyze DP, DQ, and DR, lysates in SDS-containing sample buffer were left at room temperature or were boiled. As expected, formation of SDS-stable DR␣␤ dimers required expression of both the Ii and DM (using the anti-DR␣ mAb 1B5; Fig. 3A, lanes 5 and 6). Expression of either the Ii or DM alone was not sufficient to induce the formation of SDS-stable DR dimers (Fig. 3A, lanes 3 and 4 and lanes 7 and 8; see also Fig. 5 in 20).
In contrast, the requirements for HLA-DQ to become SDSstable were different from those for HLA-DR. Co-expression of the Ii alone resulted in SDS-stable DQ␣␤ dimers (using a polyclonal anti-DQ serum) (Fig. 3B, lanes 3 and 4). Additional expression of DM increased the amount of SDS-stable DQ␣␤ dimers (Fig. 3B, lanes 5 and 6), indicating that DM promoted peptide loading of HLA-DQ. Co-expression of DM in the absence of the Ii resulted in a negligible amount of dimers (Fig.  3B, lanes 7 and 8), probably because DQ␣␤ did not intersect with lysosomal loading compartments without the Ii (Fig. 2B). To confirm that DQ could form SDS-stable dimers with coexpression of the Ii only, we used two additional anti-DQ mAbs, L2 and SPV-L3. Both L2 and SPV-L3 clearly detected DQ dimers when co-expressed with the Ii (Fig. 3C, lanes 3 and 4 and lanes 7 and 8) but did not recognize DQ␣␤ dimers when only ␣ and ␤ chains were expressed (Fig. 3C, lanes 1 and  2 and lanes 5 and 6). The blot was co-probed with HC10 (detecting MHC class I; cI.I) to confirm equal loading of the samples.
In stark contrast to HLA-DR1 and -DQ5, HLA-DP4 was not dependent on either the Ii or DM to form SDS-stable dimers (using a polyclonal DP antiserum; Fig. 3D, lanes 1 and  2). Ii alone facilitated the formation of slightly more compact ␣␤ dimers (Fig. 3D, compare lanes 2 and 4). DM alone also facilitated SDS-stable DP dimer formation (Fig. 3D, lanes 7  and 8), indicating that the targeting of DP to the lysosomal system was not required for its stability. Co-expression of Ii and DM further increased the amount of SDS-stable DP dimers (Fig. 3D, lanes 5 and 6). Thus, either Ii or DM alone may facilitate the transition of HLA-DP to a more compact dimer state. Note that the amount of monomeric DQ and DP detected before and after boiling remained similar, whereas the amount of DR␣ detected after boiling increased. This may be because the DP and DQ antibodies detect monomers less efficiently than 1B5.
The DP serum also detected a background band at 50 kDa (*, Fig. 3D), the intensity of which varied between experiments. However, the presence of the background band did not influence the recovery of stable DP␣␤ dimers. As shown in Fig. 3E, when the background band was absent, stable DP dimers formed between the DP␣␤ chains alone, and these became more compact in the presence of Ii. This experiment also shows that DP␤ chains alone did not form SDS-stable dimers. In addition, incubation of semipermeabilized HLA-DP transfectants with a specific DP-binding peptide resulted in an increased recovery of stable dimer, showing that DP molecules could be stabilized by peptide in this system (supplemental Fig. 2

and supplemental Experimental Procedures).
To see whether the unusual stability of DP4 was shared by other DP molecules, we tested DPB1*1701, a beryllium disease-associated allele. This allele also gave SDS-stable dimers in the absence of the Ii and DM (Fig. 4A, lane 2). The amount of stable DPB1*1701 ␣␤ dimers increased with co-expression of the Ii, further demonstrating that DP stability was unusual in that it did not require DM (Fig. 4A, lane 6).
The observed stability of DP␣␤ might reflect epitope(s) specifically detected by the anti-DP serum. To examine whether a different antibody could detect SDS-stable DP dimers, we used the DP-reactive mAb KUL/05 (29). KUL/05 recognized DP␤ monomers and DP dimers in DP␣␤ transfectants, demonstrating that the detection of DP stability was not an antibody-specific phenomenon (Fig. 4B, lanes 3 and 4). The amount of dimers increased when DP␣␤ was co-transfected with the Ii and DM, as expected (Fig. 4B, lanes 5 and 6). The DP dimers in transfectants were similar to those in MelJuso, a cell line that expresses MHC class II molecules endogenously (Fig. 4C, lanes 5-8). Note that the recognition of DP monomers decreased in cells expressing endogenous DP (Fig. 4C, lanes 7 and 8) but that both dimers and monomers could be detected by the DP antiserum after immunoprecipitation from MelJuso with the HL40 mAb (Fig. 4C,   lanes 11 and 12). Taken together, these data demonstrate that a pool of HLA-DP␣␤ molecules, unlike HLA-DR1 and -DQ5, can assemble in the absence of the Ii and become SDS-stable without intersecting the lysosomal pathway.
Oxidative Assembly of HLA-DR, -DQ, and DP in the ER-The differences in the SDS stability of DR, DQ, and DP could result from differences in the folding and intrinsic stability of the class II molecules. Having observed that oxidative protein folding is important for heterodimeric (and Ii-independent) assembly of DM (20), we compared the disulfide-dependent protein oxidation of HLA-DR, -DQ, and -DP (Fig. 5). Lysates from HLA-DR, -DQ, and -DP transfectants were analyzed by Western blotting after nonreducing SDS-PAGE to preserve intra-and intermolecular disulfide bonds. The "nonreducing"  complexes observed in these experiments were obtained after boiling in sample buffer ϪDTT and are therefore not the same as those observed in Figs. 3 and 4, which are obtained without boiling in sample buffer ϩDTT.
All ␤ chains migrated faster under nonreducing conditions (Fig. 5, B-D), indicating the presence of intramolecular disul-fide bonds. This was regardless of whether ␤ was co-expressed with ␣ chains or accessory proteins. In contrast, the ␣ chains migrated similarly under reducing and nonreducing conditions (Fig. 5, A, C, and D, compare R with NR), although note that the DQ and DP antisera only weakly recognized the ␣ chains, especially under nonreducing conditions. The difference between ␣ and ␤ chains in migration under nonreducing conditions is explained by the presence of two long range disulfide bonds in ␤ chains versus one in ␣ chains, which makes the ␤ proteins more compact under nonreducing/denaturing conditions. Despite these general similarities for DR, DQ, and DP, there are some marked differences. Unlike the DQ␤ monomers, the DR␤ and DP␤ monomers existed in two distinct oxidation states (Fig. 5, B-D, top panels, bands 1 and 2 for DR and DP). This may be explained by the presence of additional cysteine residues in DR␤ and DP␤. The DR␣ chain also appeared as a doublet, but this did not reflect different oxidation states, as the pattern was essentially the same under reducing conditions (Fig. 5A, lanes 3-5). Rather, the DR␣ doublet is most likely due to two differently glycosylated pools (30), similar to that seen for DM␣ (20).
The nonreducing gels also revealed differences in disulfidelinked complexes across the three MHC class II molecules. In contrast to DR and DP, DQ hardly formed any disulfidelinked complexes, which may reflect the absence of any unpaired cysteine residues in this molecule (Fig. 5C). The disulfide-linked DR␣ and DR␤ dimers and higher molecular mass complexes gradually diminished with further reconstitution (Fig. 5, A and B, top panels). All dimers and high molecular mass complexes disappeared under reducing conditions (data not shown; see also Fig. 3, A-D, lanes 1, 3, 5, and 7). DP␣ and ␤ formed complexes at ϳ50 kDa (Fig. 5D): a lower one when DP␤ was expressed alone (lane 2, band 3), and an additional higher one when DP␣ was co-expressed (lanes 3-5, band 4). Therefore, band 3 probably represented DP␤ disulfide-linked dimers, whereas band 4 could represent DP␣␤ disulfidelinked dimers, as they were only present when DP␣ was coexpressed. The presence of band 4 in DP␣␤ transfectants (and in combination with Ii and DM) correlated with the requirements for the SDS stability of DP (Fig. 3D) and raised the possibility that a disulfide-bonded complex between DP␣␤ might be responsible for DP stability in the absence of the Ii.
To exclude that disulfide-linked DP␣␤ dimers accounted for the SDS stability of DP in DP␣␤ transfectants, we made use of a single cysteine mutant of DP␤, C211A. In contrast to wild-type DP␤, this mutant did not generate band 4 when co-expressed with DP␣ (Fig. 6A, lanes 3 and 4). This mutant, however, was still able to form SDS-stable dimers when DP␣ and DP␤ C211A were co-expressed (Fig. 6B, lanes 9 and 10). To further confirm the molecular nature of the DP␣␤ dimer, we constructed a DP␤ mutant that lacked the conserved Cys 15 -Cys 77 disulfide bond. The Cys 15 -Cys 77 mutant did not form SDS-stable complexes with the DP␣ chain (Fig. 6B, lanes  5 and 6 and lanes 11 and 12), demonstrating that disulfide stabilization of the peptide binding site was required for ␣␤ complex formation, and additionally confirming the specificity of the DP antiserum. The higher expression levels of the C211A mutant suggest that this C-terminal cysteine may be involved in degradation of orphan DP␤ chains (31) or required for interaction with cytosolic/membrane components for DP transport (32); this will be explored in subsequent work. Taken together, these experiments show that the DP SDS-stable dimers observed in Fig. 3 were not the result of misoxidized proteins.
HLA-DP SDS-stable Dimers are Leupeptin-insensitive-Having shown that the Ii is not absolutely required for the stability of properly folded DP, we wanted to establish whether DP stability is acquired in the endosomal-lysosomal system. Formation of SDS-stable HLA-DR dimers can be prevented by treatment of cells with the cysteine/serine protease inhibitor leupeptin (33). Leupeptin prevents the complete degradation of the Ii and affects the generation of peptides by leupeptin-sensitive proteases (34). To see whether the Ii/DMindependent formation of SDS-stable DP␣␤ dimers is sensitive to leupeptin, cells were treated with leupeptin and transfected with different combinations of DR␣␤, DP␣␤, Ii, and DM. After 24 h, lysates from the transfectants were subjected to the SDS stability assay. As expected, DR␣␤ in the absence of the Ii did not gain SDS stability (Fig. 7A, lanes 1-4). When cells co-expressed the Ii and DM, DR clearly formed SDSstable dimers (Fig. 7A, lanes 5 and 6), which were almost completely abrogated by leupeptin treatment (Fig. 7A, lanes 7  and 8). DP dimers were easily detected in DP␣␤ transfectants, and these were not disrupted by leupeptin treatment (Fig. 7B, lanes [5][6][7][8]. As expected, DP␤ alone did not form dimers (Fig.  7B, lanes 1-4). Our results show that unlike DR, DP molecules can acquire SDS stability by an Ii/DM-independent pathway that is insensitive to leupeptin.
To examine whether leupeptin-insensitive DP complexes existed in APCs, we investigated the behavior of DR and DP in a melanoma cell type (MelJuso) and a lymphoma cell line (Daudi) that endogenously express MHC class II molecules. Remarkably, leupeptin or NH 4 Cl (which neutralizes lysosomal pH) treatment left DP dimers intact (Fig. 8, B and D), whereas DR1 dimers from the same lysates were lost (Fig. 8, A and C). Thus, unlike DR1, a pool of DP molecules can acquire stability outside the classical endosomal-lysosomal pathway in both transfectants and in professional APC types.

DISCUSSION
In this paper, we have directly compared the assembly and stability requirements for HLA-DP, -DQ, and -DR for the first time. We show that HLA-DR, -DQ, and -DP differ markedly in their requirements for the invariant chain and DM, despite having ϳ70% amino acid sequence similarity. Our results show that HLA-DR, -DP, and -DQ all require the Ii for endosomal-lysosomal targeting (Fig. 2) but not for stability (Fig. 3). Although ␣␤ complexes are ER-localized in the absence of  DM or Ii, at least a portion of these ␣␤ complexes are folded and can exit the ER, bypassing the endosomal-lysosomal system. The concept that the Ii is not required for the quality control of DP, DQ, or DR per se is supported by work in the mouse, where residual MHC class II molecules appear at the cell surface in the absence of the Ii (25) or when class II is transfected in the absence of the Ii (26,27). The effect of Ii deficiency in mice is also allotype-specific; for example, the BALB/c Ii knock-out has a mild phenotype and develops functional CD4 ϩ T cells (35). In the absence of functional Ii, H-2b does not assemble properly in spleen cells, but H-2k is unaffected (36). In H-2k mice, loss of DM has an effect on E(k) but not A(k) class II molecules (37), supporting the notion that the need for the MHC class II chaperones is alleledependent. In vivo, Ii gene expression is not absolutely coordinated with MHC class II synthesis (38), and there are circumstances where deregulation of Ii may occur, for example during HIV infection, where Ii is a target of Nef (39). Our results therefore raise the possibility that expression levels of Ii could be exploited to manipulate the relative levels of stable MHC class II molecules, either by pathogens or for therapeutic benefit.
Although there are no published studies on the effect of Ii and DM on DP antigen presentation, our finding that the Ii is sufficient for DQ␣␤ to attain a stable SDS conformation is supported by Ettinger et al. (15), who suggest that DQ0602 may present antigen in a DM-independent fashion. We demonstrate here that the Ii alone is required for recognition of SDS-stable DQ5 dimers by L2 and SPVL3 (Fig. 3C). The observation that DQ5 (and DP) can be stabilized in the absence of DM (Fig. 3) has implications for the function of HLA-DO. DO regulates DM activity in a pH-dependent manner (40). It will be interesting to test whether DO can selectively adjust DP or DQ peptide binding, resulting in different relative expression of HLA-DR, DQ, and DP on different APCs.
For MHC class I molecules, it has been suggested that the C-terminal cysteine residue of HLA-B7 can influence recognition by NK cells (41). We noted that DP␤ chains have an unpaired Cys 211 residue that is not shared with DR and DQ ␤ chains, leading us to investigate whether Cys 211 could also influence conformation at a spatially distant site. The DP␤ Cys 211 residue is not required for DP SDS stability, although Cys 211 does increase the propensity of DP␤ to form disulfidelinked complexes (Figs. 5 and 6). In contrast, the Cys 15 -Cys 17 disulfide bond, which anchors the peptide-binding domain, is required for the stable assembly of DP␤ with its cognate ␣ chain (Fig. 6).
SDS stability is a well documented readout for functional, peptide-loaded HLA-DR complexes (28). Here, we have shown unexpectedly that HLA-DP4 does not need Ii to acquire this property (Figs. 3, 6, and 7) and that this is not allelespecific, antibody-specific (Fig. 4), or cell type-specific (Fig. 8). Given that DR1 molecules can be stabilized in vitro with short peptides (42), it will be important to establish whether the stable DP␣␤ complexes seen in different conditions are "empty, " loaded with peptides, or a mix of the two. Although our experiments show that DP in semipermeabilized transfectants can be stabilized by a specific antigenic peptide (supplemental Fig. 2), it remains possible that an unknown accessory factor, such as an ER chaperone, might help to stabilize empty DP complexes. In vitro assays have shown that DP can certainly bind to CLIP fragments (43), and known HLA-DP peptide-binding motifs differ from those of (ER-loaded) MHC class I molecules, so DP is not likely to compete for classical class I-binding peptides (44). However, peptide elution studies have demonstrated that HLA-DP2 is naturally loaded with ER protein-derived peptides, including ERp57 (PDIA3) and Grp94 (endoplasmin) (45), suggesting that in vivo some DP molecules loaded with ER peptides reach the cell surface. It will also be important to establish whether peptides in the ER can compete with Ii to load DP␣␤ in APCs or whether peptides from viral and ER proteins are actually obtained at the cell surface or during DP recycling. The possibility that proteases at the plasma membrane, or in the extracellular matrix, play a role in DP peptide loading in vivo deserves further exploration. Another possibility is that the relative Ii independence of DP makes it more accessible to peptides during autophagy (46), which might also explain why DP cross-presents viral antigens.
One of the first reports about DP (then named SB) function was on the presentation of viral, rather than bacterial, antigens, namely from herpes simplex and influenza viruses (47). It was further demonstrated that SB/HLA-DP, when transfected into murine fibroblasts, could present influenza viral peptides to DP-restricted human T cells (48). There are a growing number of examples of viral peptides that bind to DP molecules and elicit CD4 ϩ T cell responses, particularly for DP4, the allele used in this study. Of particular note is the differential association between DP and chronic hepatitis B in Asian populations (49). It will be informative to compare peptide binding and assembly of different susceptibility haplotypes with the behavior of the protective DP4 examined in our study. Further exploration of the molecular details of DP conformation and stability may shed light on why some autoimmune diseases are DP-linked and whether unique therapeutic routes for peptide delivery to DP can be exploited.