Functional analysis of tryptophans alpha 62 and beta 120 on HLA-DM.

In the endocytic pathway of antigen-presenting cells, HLA-DM catalyzes the exchange between class II-associated invariant chain peptide (CLIP) and antigenic peptides onto major histocompatibility complex class II molecules. At low pH of lysosomal compartments, both HLA-DM and HLA-DR undergo conformational changes, and it was recently postulated that two partially exposed tryptophans on HLA-DM might be involved in the interaction between the two molecules. To define contact regions on HLA-DM, we have conducted site-directed mutagenesis on those two hydrophobic residues. The HLA-DM alphaW62A,betaW120A (DM(W62A/W120A)) double mutant was expressed in HLA-DR(+) HeLa cells expressing invariant chain, and the activity of this DM molecule was assessed. Flow cytometry analysis of cell surface DR-CLIP complexes revealed that DM(W62A/W120A) removes CLIP as efficiently as its wild-type counterpart. DM(W62A/W120A) was found in the endocytic pathway by immunofluorescence, and DM-DR complexes were immunoprecipitated from these cells at pH 5. Finally, mutations alphaW62A and betaW120A on HLA-DM did not affect the association with HLA-DO. The complex egresses the endoplasmic reticulum and accumulates in endocytic vesicles. Moreover, DO and DM(W62A/)W120A were co-immunoprecipitated at pH 7. We conclude that the alpha62 and beta120 tryptophan residues are not required for the activity of DM, nor are they directly implicated in the interaction with DR or DO.

Major histocompatibility complex class II molecules are primordial for activation of CD4ϩ T cells and for immunological response against pathogens (1). Three major histocompatibility complex II ␣/␤ heterodimers associate with a trimer of invariant chain (Ii) 1 to form a nanomeric complex in the endoplasmic reticulum (ER) (2). This association allows the proper folding and trafficking of the major histocompatibility complex II molecules (3)(4)(5). Gradual proteolysis of Ii occurs in the endocytic compartments by the sequential action of proteases including cathepsins (6). A residual peptide of Ii (CLIP) remains in the peptide groove of the class II molecule and stabilizes the heterodimer by preventing collapse of the groove (7,8). For most allotypes, CLIP must be actively removed to allow binding of antigenic peptides. HLA-DM, a nonclassical intracellular class II molecule, plays a central role in the efficiency of antigen presentation as it catalyzes CLIP release from HLA-DR and stabilizes the class II in a chaperone-like fashion (9 -14). In fact, mice lacking the H2-Ma gene express a high level of surface class II-CLIP complexes (15)(16)(17).
The precise molecular mechanism of action of HLA-DM remains to be established. X-ray diffraction studies on HLA-DM crystals revealed a closed peptide-binding groove and an overall quaternary structure similar to classical class II molecules (18,19). Time-resolved fluorescence anisotropy and far-UV circular dichroism spectrum analysis using soluble HLA-DM revealed that its structure is quite rigid and that it is not subjected to gross pH-dependent conformational change. However, many other experiments by the same groups suggest that protonation in the endocytic pathway results in minor, reversible structural changes exposing hydrophobic regions of the DM heterodimer (20,21). For example, fluorescence spectroscopy studies revealed that hydrophobic tryptophan residues are buried in native DM and would become more solvent-exposed at endosomal pH (20). Accordingly, Ullrich et al. (21) used 8-anilino-1-naphthalenesulfonic acid (ANS), a fluorescent dye binding to hydrophobic protein patches, to demonstrate subtle pH-induced change in DM. Since the interaction of DM with DR reduces ANS binding to both molecules, it was postulated that the surface of contact comprises these pH-sensitive regions (21,22).
On the basis of these results, Wiley and co-workers (19) proposed that two partially exposed tryptophans (␣62, ␤120), located on the same lateral surface of DM, could be critical for DR binding. The model predicts a major interaction between tryptophan ␣62 of DM and the protruding phenylalanine ␣51 of DR, resulting in the breakage of multiple hydrogen bonds at the end of the groove and in the release of unstably associated peptides. At least part of this model was recently confirmed when mutation of DR␣F51 was shown to abolish the interaction with DM (23).
To confirm the possible interaction between DR␣F51 and DM␣W62, we have mutated the exposed tryptophan on HLA-DM and tested the ability of this mutant to release CLIP from HLA-DR. Our results show that this mutation on DM does not affect DR contact, CLIP release, or the localization of the protein within the cell. Similar conclusions were drawn after mutating the second exposed tryptophan at position ␤120. Finally, those amino acids substitutions on DM did not affect its interaction with HLA-DO, an intracellular nonclassical class II molecule that binds to DM and regulates its activity. DM␤ cDNA has been described previously (25). DM␣ and DM␤ cDNAs from pBluescript were subcloned successively on the same pBudCE4-A vector (as previously described) (25). The Ii p35 cDNA was obtained by mutagenesis of the second ATG codon of Ii (kindly provided by Rafick Sékaly) and subcloned in the BamHI site of SR␣puro. Details will be provided elsewhere.
DM␣ and DM␤ Mutagenesis-Mutations into the DM␣ and DM␤ cDNA sequences were introduced by PCR overlap extension (26). Briefly, 5Ј PCR products were generated from pBS DM␣ and pBS DM␤ using mutagenic primers (DM␣W62AEcoRI, 5Ј-CTG AGC CGC GTC AGC GAA TTC GGG-3Ј; DM␤W120ANspI, 5Ј-GAA GCC CGC CAC ATA GCA TGC CAG CAT-3Ј) as well as the universal (DM␣) or reverse (DM␤) primers. The 3Ј PCR product was generated using the complementary mutagenic primers. The two overlapping PCR products were mixed, and a final PCR was performed using the flanking primers. Fragments were subsequently subcloned into the PstI-SalI sites of pBSDM␣ and SalI-HindIII of pBSDM␤, thereby replacing the wild-type fragment with the PCR product containing the appropriate mutation. The nucleotide sequence was confirmed by DNA sequencing using T7 polymerase (Amersham Biosciences). Mutant cDNAs and wild-type were introduced into pBudCE4-A as NotI-XhoI (pBSDM␣) and SalI-XbaI (pBSDM␤) fragments.
Flow Cytometry Analysis-Intracellular staining was done as previously described (25). Briefly, saponin-treated cells were incubated with the appropriate primary antibody, and then they were incubated with goat anti-mouse IgG (H ϩ L) coupled to Alexa Fluor ® 488. Cells were then analyzed by flow cytometry on a FACScalibur (Becton Dickinson, Mississauga, Canada).
Fluorescence Microscopy-10 4 cells were plated on coverslips in 24-well plates and cultured for 3 days before staining (25). Cells were analyzed by fluorescence microscopy on a Zeiss axioscope microscope (Carl Zeiss, Thornwood, NY). Photographs were taken with a Zeiss microscope camera MC 100 on Eastman Kodak Co. elite chrome 400 films.
Immunoprecipitations and Western Blotting-Cells (4 ϫ 10 6 ) were trypsinized, washed in phosphate-buffered saline, and lysed in 1% Triton X-100 at pH 7 or in CHAPS 1% at pH 5 as described previously (29). After centrifugation, supernatants were harvested and incubated with protein G coupled to CNBr-activated Sepharose 4B (Amersham Biosciences) and bound to the appropriate antibody (L243 mAb or mouse serum anti-DO␤). Following washes in lysis buffer, samples were resuspended in nonreducing buffers and subjected to SDS-PAGE. After transfer to nylon membranes (Amersham Biosciences), proteins were blotted with the rabbit anti-DO␣ or anti-DM␤ sera. Secondary antibody (peroxidase-coupled goat anti-rabbit; BIO/CAN Scientific) was used at a 1:1000 dilution for 2 h, and the signal was detected using ECL (Amersham Biosciences).

RESULTS
The crystal structure of DM reveals two partially exposed tryptophan residues that are thought to interact with HLA-DR and to participate in CLIP removal (19). Tryptophans ␣62 and ␤120 are located on the same lateral face of HLA-DM in the ␣1 and ␤2 domains, respectively (Fig. 1). These amino acids are highly conserved throughout evolution, emphasizing their potential importance.
In an effort to define the binding site for HLA-DR and to gain insights into the mechanism of action of HLA-DM, we mutated those two bulky aromatic tryptophans to small alanines and tested in vivo the ability of mutant proteins to remove CLIP from HLA-DR. Various combinations of mutated and wild-type DM ␣␤ cDNAs were transfected in HeLa cells expressing DR1 and Ii. DR1 is known to be dependent on DM for the release of CLIP at acidic pH (30). We obtained four stable cell lines: DR1 Ii DM, DR1 Ii DM W62A , DR1 Ii DM W120A , and DR1 Ii DM W62A/W120A . Expression of DR was monitored at the cell surface, while DM and Ii were analyzed in permeabilized cells. As measured by flow cytometry, all cell lines express high levels of the various transfected molecules (Fig. 2). The proper folding and intracellular sorting of DM was assessed by immunofluorescence microscopy. Fig. 3 shows that wild-type HLA-DM (B) accumulates in intracellular vesicles that are also positive for the lysosomal marker Lamp-1 (G). A tyrosine-based motif on the cytoplasmic tail of DM␤ is responsible for its accumulation in endocytic compartments (31)(32)(33). The three mutated forms of DM were also found in peripheral vesicles (Fig. 3, C-E) and co-localized with Lamp-1 (Fig. 3, H-J). Control cells lacking DM showed a weak background using the DM-specific antibody but definite Lamp-1-positive compartments (Fig. 3, A, F, and arrows). Together with the fact that the MaP.DM1 conformational antibody efficiently recognizes the DM mutants, these results show that the overall structure of DM is not affected by replacement of tryptophans ␣62 and ␤120.
The activity of these mutant forms of DM was verified by monitoring the levels of DR-CLIP complexes at the cell surface of transfected cells. As demonstrated by many groups, introduction of HLA-DM in deficient cell lines favors the intracellular exchange of CLIP for more stable peptides (24,34,35). Consequently, CLIP expression is low at the surface of DM ϩ cells. The cell surface expression of CLIP was measured using the Cer-CLIP monoclonal antibody that is specific for the N terminus of CLIP bound to class II molecules and which does not recognize the intact invariant chain (24). As shown in Fig.  4A, control DM Ϫ cells (HeLa DR1 Ii) express high levels of DR-CLIP complexes at their surface. On the other hand, cells expressing wild-type or mutant forms of DM do not express significant levels of CLIP (Fig. 4, B-E). These results suggest that DR-Ii complexes are sorted to the endocytic pathway, where Ii is degraded until a last fragment, CLIP, is actively removed from the peptide binding groove by wild-type or mutant forms of HLA-DM. To confirm the proper interaction between the DM mutants and DR, co-immunoprecipitations were carried out at acidic pH in the CHAPS detergent (36). Fig. 5 shows that DM can be co-immunoprecipitated by the DR-specific antibody only in cells expressing both molecules (Fig. 5,  left three lanes). Also, the three mutant forms of DM could all be efficiently co-immunoprecipitated with HLA-DR (Fig. 5). The stronger DM␤ signal obtained by Western blotting on the HeLa DR1 Ii DM W62A/W120A samples most probably reflects the higher level of HLA-DR expression on these cells (mean fluorescence value ϭ 789) as compared with the other DM ϩ transfectants (mean fluorescence values ϭ 248 -374) (Fig. 2). Taken together, these results show that the mutations ␣W62A and ␤W120A do not affect the folding, sorting, and activity of HLA-DM, nor its ability to interact strongly with HLA-DR.
Although these tryptophan residues are not directly involved in the binding to HLA-DR, they might very well be crucial for the binding to HLA-DO. The latter is mostly expressed in B cells and is a nonclassical class II molecule that modulates the activity of DM (37)(38)(39). The mode of action of DO is not determined, but it clearly affects the peptidic repertoire when present in physiological amounts in an APC (40 -42).
To further characterize the DO-DM contact regions, wildtype and double mutant (DM W62A/W120A ) DM molecules were stably transfected in HeLa DO ϩ cells. Expression of both proteins was confirmed by intracellular staining and flow cytometry (Fig. 6). Since HLA-DO is absolutely dependent on DM association to egress the ER (43) and gain access to the endocytic pathway, we first verified the ability of mutant HLA-DM to direct the transport of DO. Immunofluorescence microscopy showed that HLA-DO reaches the endocytic pathway in the presence of either the wild-type or the DM W62A/W120A molecule (Fig. 7). As opposed to the diffuse ER-like pattern of expression observed for DO in DM Ϫ cells (Fig. 7b), defined HLA-DOcontaining vesicles can be seen in DM ϩ cells (Fig. 7, d and f). The co-localization between DO and DM W62A/W120A in the endocytic pathway (Fig. 7, e and f) strongly argues for an efficient interaction between the two molecules, allowing DO to egress the ER.
The interaction between DO and DM was confirmed by co-immunoprecipitation experiments. Cells were lysed in 1% Triton X-100, and DO was immunoprecipitated using a polyclonal antibody against the cytoplasmic tail of the ␤ chain. Western blotting with antibodies specific for DM or DO showed that wild-type DM and the DM W62A/W120A double mutant were both efficiently co-immunoprecipitated with DO (Fig. 7B). Altogether, these results suggest that ␣62 and ␤120 tryptophans on HLA-DM are dispensable for the interaction with DR or DO. DISCUSSION Understanding the molecular mechanism by which HLA-DM and HLA-DM/DO catalyze peptide exchange will allow the development of new approaches for manipulating antigen loading and presentation. With this in mind, we have undertaken the mapping of amino acids involved in the interactions between these molecules. So far, the group of Mellins has characterized regions of HLA-DR that interact with HLA-DM (23,44). Random mutagenesis on HLA-DR allowed the identification of critical residues on a lateral face encompassing both the ␣1 and ␤2 domains. From these experiments, it was proposed that DM releases unstable peptides through its "lever effect" on amino acids DR␣40 and ␣51, the last one being critical for stability around the P1 pocket of the groove (23). A role for DM in destabilizing those P1 anchors was also proposed from the results of Chou and Sadegh-Nasseri who showed that mutation of DR1␤G86Y rendered the class II rigid (as in a peptide-bound conformation) and reduced the binding to DM (45).
Based on the crystal structure of HLA-DM, the group of Wiley had already proposed that the contact between DM Trp␣62 and DR Phe␣51 could destabilize the P1 pocket and liberate an unstable peptide. Much experimental evidence points to a critical role of such hydrophobic residues in the contact between DM and DR. For example, based on studies measuring the binding of ANS, it was concluded that DR-CLIP complexes display a larger hydrophobic surface than DR molecules associated with stable peptides. Also, there is a prefer-ential association of the two molecules at the acidic pH of endocytic vesicles, where both molecules would expose hydrophobic residues (21). The presence of tryptophan residues in those contact regions was deduced by their preferential binding of ANS as well as from spectroscopy studies measuring variations in fluorescence emission between exposed and buried residues (21,45).
Two of 11 tryptophans are partially exposed and located on the same lateral surface on the crystal structure of DM. These residues may be part of the hydrophobic patches that become accessible to ANS at acidic pH but that are buried in the DM-DR interface (19). While DR Phe␣51 could interact with DM Trp␣62, tryptophan DM ␤120 may contact those hydrophobic residues identified by Mellins and co-workers (23) in the ␤2 domain of DR and which may serve to increase the affinity for DM. However, our results presented here do not support such a model. Mutation of the two tryptophans on DM did not disturb the activity, the sorting, or the conformation of the protein. Still, the possibility remains that those mutations could finely tune the specificity of DM and influence the peptide repertoire of the class II molecules.
The fact that random mutagenesis on HLA-DR allowed the identification of Phe␣51, Leu␤184, and Val␤186 residues certainly suggests the involvement of hydrophobic residues on DM as well (23). However, replacement of residue Glu␤187 for a lysine in the ␤2 domain also decreased DM binding in these studies. The importance of this charged residue on DR prompted us to evaluate the potential role of positively charged DM Lys␣109, Lys␣115, Arg␤93, and Arg␤95 residues. However, the simultaneous mutation of these residues did not inhibit CLIP release by DM in our system (data not shown).
Altogether, our results suggest that the DM/DR interaction relies on other hydrophobic residues in the above described region or that it may implicate another interface of DM. Indeed, Fremont et al. (18) identified two other hydrophobic regions that are located on distinct faces of the H2-M heterodimer. Site-directed mutagenesis in the corresponding regions of DM should help delimiting contact sites with DR.
Finally, our results from immunofluorescence and co-immunoprecipitation studies revealed that tryptophans ␣62 and ␤120 on DM are not necessary to make contact with HLA-DO. Although DO also undergoes conformational change following acidification of the environment (46), it first interacts with DM in the ER, and the strong association is resistant to lysis in 1% Triton X-100 (43). These observations suggest that the nature of the interactions is likely to differ between DR-DM and DO-DM. The existence of a three-molecule complex between DM, DO, and DR suggests the presence of two distinct functional interfaces on DM, and random mutagenesis is under way to delineate the contact regions with HLA-DR and DO (47). FIG. 7. HLA-DM mutants bind to HLA-DO. A, immunofluorescence microscopy analysis was performed on permeabilized cells doublestained using a rabbit anti-DM␤ (a, c, and e) and a mouse anti-DO␤ (b, d, and f) sera. Secondary antibodies were Alexa 488-coupled goat antirabbit and a biotinylated GAM followed by Texas Red-coupled streptavidin. a and b, HeLa DO; c and d, HeLa DO DM; e and f, HeLa DO DM W62A/W120A . B, cells were lysed in 1% Triton X-100 at pH 7, and co-immunoprecipitations were performed using a DO␤-specific mouse serum. Samples were analyzed by Western blotting using the DM␤specific or DO␣-specific rabbit sera.