Defective Human Leukocyte Antigen Class I-associated Antigen Presentation Caused by a Novel β2-Microglobulin Loss-of-function in Melanoma Cells*

The major histocompatibility complex class I molecules consist of three subunits, the 45-kDa heavy chain, the 12-kDa β2-microglobulin (β2m), and an ∼8-9-residue antigenic peptide. Without β2m, the major histocompatibility complex class I molecules cannot assemble, thereby abolishing their transport to the cell membrane and the subsequent recognition by antigen-specific T cells. Here we report a case of defective antigen presentation caused by the expression of a β2m with a Cys-to-Trp substitution at position 25 (β2mC25W). This substitution causes misfolding and degradation of β2mC25W but does not result in complete lack of human leukocyte antigen (HLA) class I molecule expression on the surface of melanoma VMM5B cells. Despite HLA class I expression, VMM5B cells are not recognized by HLA class I-restricted, melanoma antigen-specific cytotoxic T lymphocytes even following loading with exogenous peptides or transduction with melanoma antigen-expressing viruses. Lysis of VMM5B cells is restored only following reconstitution with exogenous or endogenous wild-type β2m protein. Together, our results indicate impairment of antigenic peptide presentation because of a dysfunctional β2m and provide a mechanism for the lack of close association between HLA class I expression and susceptibility of tumor cells to cytotoxic T lymphocytes-mediated lysis in malignant diseases.

The human leukocyte antigen (HLA) 4 class I molecules, encoded by the genes located in the major histocompatibility complex, are composed of three subunits including a 45-kDa HLA class I heavy chain (HC), a 12-kDa ␤ 2 -microglobulin (␤ 2 m), and an ϳ8 -9-residue peptide (1). Expression of these molecules on the cell surface requires the stepwise assembly of HCs, ␤ 2 m, and peptides in the endoplasmic reticulum (ER) followed by the transport of the trimeric molecule to the plasma membrane. These processes are dependent on a functional antigen processing machinery (APM), which includes the proteasome subunits, the peptide transporters TAP1 and TAP2, and a number of ER-resident chaperons such as calnexin, calreticulin, ERp57, and tapasin (2,3). ␤ 2 m plays an integral part in the assembly and transport of HLA class I molecules because it stabilizes the HC-␤ 2 m heterodimer through noncovalent protein-protein interactions, thereby allowing binding of endogenous antigenic peptides with the help of TAP and tapasin (4). As a result, the assembled HC-␤ 2 m-peptide trimeric complexes can travel to the cell surface, where they are recognized by HLA class I-restricted, antigen-specific cytotoxic T lymphocytes (CTLs).
The lack of HLA class I molecule expression on the cell surface usually reflects defects in ␤ 2 m protein synthesis caused by mutations in the ␤ 2 m gene, as has been found mostly in malignant cells thus far (5). This abnormality renders tumor cells resistant to tumor antigenspecific CTLs and may have a negative impact on the elimination of tumor cells by host CTLs. The defects underlying ␤ 2 m loss have thus far been found to be structural in nature, involving lack of translation because of small deletions or point mutations in most cases and lack of transcription because of large deletions in a few cases (5). Because of a lack of ␤ 2 m expression, the resulting HLA class I loss cannot be corrected by interferon (IFN-␥), a cytokine that up-regulates the expression of most of the molecules participating in the assembly and transport of HLA class I molecules. On the other hand, a low level of HLA class I expression on cells usually reflects nonstructural defects in the APM components because this abnormality can be corrected by IFN-␥ (6).
In the present study, we have elucidated the mechanism underlying HLA class I down-regulation identified in a melanoma cell line and in the metastasis from which the cell line was derived (7). This HLA class I down-regulation phenotype cannot be corrected by IFN-␥ and was unexpectedly found to be caused by an abnormal full-length ␤ 2 m protein that can neither form stable complexes with HCs nor assist in loading peptides onto the HLA class I peptide binding groove.
in RPMI 1640 medium supplemented with 10% heat-inactivated fetal calf serum in a 5% CO 2 atmosphere at 37°C. A tyrosinase 369 -377D -specific 5 CTL line was derived from patient VMM119 who had been vaccinated with a mixture of four peptides including tyrosinase 369 -377D (YMDGTMSQV) (9).
Peptides, IFN-␥, Wild-type Human ␤ 2 m, and Pharmacological Inhibitors-The HLA class I-associated peptides HER2/neu 369 -377 , KIFGSLAFL, and tyrosinase 369 -377D YMDGTMSQV were synthesized with a free amide N terminus and free acid C terminus by standard Fmoc (N-(9-fluorenyl)methoxycarbonyl) chemistry using a model AMS422 peptide synthesizer (Gilson Co. Inc., Middleton, WI). Peptides were purified to Ͼ98% purity by reverse-phase high pressure liquid chromatography on a C-8 column (Vydac, Hesperia, CA) at the University of Virginia biomolecular core facility. Purity and identity were confirmed using a triple quadrupole mass spectrometer (Finnigan, San Jose, CA). Recombinant human IFN-␥ was purchased from Roche Applied Science. The wild-type human ␤ 2 m was purchased from Sigma. The proteasome inhibitor MG132 and the trans-Golgi-to-secretory vesicles FIGURE 1. HLA class I down-regulation on VMM5B melanoma cells. A, control-(thin line) and IFN-␥-treated (thick line) VMM5A, VMM5B, and FO-1 cells were stained with mAb W6/32 and analyzed by flow cytometry. B, VMM5B cells were treated with citric acid (pH 3.0) before being added with either medium alone (pH 3.0) or the combination of HER2/neu 369 -375 peptide (40 g/ml) with human ␤ 2 m (2.5 g/ml) (␤ 2 m ϩ peptide) and cultured for 3 h at room temperature. Subsequently, cells were harvested, stained with mAb W6/32, and analyzed by flow cytometry. The untreated VMM5B cells (pH 7.4) were used as a negative control. The numbers above each open profile are -fold MFI. Representative results of one of the three experiments are shown. The variability in results between experiments is less than 5%. traffic inhibitor monensin were purchased from Sigma and Alexis Corp. (San Diego, CA), respectively.
Low pH Treatment of Cells and Restabilization of Cell Surface HLA Class I Molecules-Low pH treatment of cells was performed as described previously (21). Briefly, cell pellets containing ϳ1-10 ϫ 10 6 cells were resuspended with 0.5 ml of stripping buffer (0.13 M citric acid, 66 mM Na 2 HPO 4 , 1% bovine serum albumin, pH 3.0) for 2 min on ice and then neutralized immediately with 50 ml of RPMI 1640 medium. After three washes, cells (1 ϫ 10 6 ) were pulsed with 40 g/ml peptide and 2.5 g/ml human ␤ 2 m (Sigma) for 3 h at room temperature and then washed three times with 1% bovine serum albumin/phosphatebuffered saline. Subsequently, cells were subjected to staining and flow cytometric analysis as described.
Flow Cytometry-Cell surface staining and intracellular staining were performed as described (22). Results are presented as -fold increase in mean fluorescence intensity over the control (-fold MFI).
Reverse Transcription-PCR-Total RNA isolation from cells was performed with TRIzol (Invitrogen) according to the manufacturer's instructions. First-strand cDNA synthesis was conducted using the SuperScript TM system (Invitrogen) according to the manufacturer's instructions. The PCR was performed using ␤ 2 m-specific primers, forward 261 (5Ј-CCTGAAGCTGACAGCATTCG-3Ј) and reverse 262 (5Ј-ACCTCCATGATGCTGCTTAC-3Ј); GAPDH-specific primers, GAPDH-F (5Ј-TGAAGGTCGGAGTCAACGGATTTGGT-3Ј) and GAPDH-R (5Ј-CATGTGGGCCATGAGGTCCACCAC-3Ј). PCR products were run on a 1.2% agarose gel (Roche Applied Science) and visualized by ethidium bromide staining. The predicted sizes of PCR products were 401 bp for ␤ 2 m and 983 bp for GAPDH. The PCR products were gel-purified, and their sequences were analyzed by a DNA sequencer (ABI PRISM model 377, Applied Biosystems).
Western Blot Analysis-Western blot analysis of cell lysates with ␤ 2 m-, HLA class I heavy chain-, and APM component-specific antibodies was performed as described (22).
Genomic DNA Extraction and PCR-Genomic DNA extraction was performed with blood and cell culture DNA midi kit (Qiagen Inc., Valencia, CA) according to the manufacturer's instructions. PCR amplification of ␤ 2 m gene exon 2 was conducted with the primers, 491 intron 1-specific (5Ј-CCTGGCAATATTAATGTG-3Ј) and 462 intron 2-specific (5Ј-CATACACAACTTTCAGCAGCT-3Ј). The predicted PCR product size is 362 bp. The PCR products were gel-purified and their sequences analyzed by a DNA sequencer (ABI PRISM model 377, Applied Biosystems).
Loss of Heterozygosity (LOH) Analysis of the ␤ 2 m Gene-LOH analysis of the ␤ 2 m gene was performed as described previously (23) with minor modifications. Briefly, purified genomic DNA (100 ng) was subjected to a PCR amplification using two pairs of primers specific to the two short tandem repeat markers (D15S126 and D15S209) located near the ␤ 2 m gene. The sequences of the primers are as follows: D15S126F, 5Ј-GTGAGCCAAGATGGCACTAC-3Ј; D15S126R, 5Ј-GCCAGC-AATAATGGGAAGTT-3Ј; D15S209F, 5Ј-AAACATAGTGCTCTG-GAGGC-3Ј; and D15S209R, 5Ј-GGGCTAACAACAGTGTCTGC-3Ј. The amplification conditions are as follows: 95°C for 12 min, 94°C for 30 s, 55°C for 30 s, 72°C for 30 s for 10 cycles; 89°C for 30 s, 55°C for 30 s, 72°C for 30 s for 20 cycles; and a final extension at 72°C for 10 min. PCR products were then fractionated on a 4% agarose gel and visualized by ethidium bromide staining. The intensity of the staining was determined by the AlphaImager TM 2200 system (Alpha Innotech Corp., San Leandro, CA). LOH is determined by the following formula: (intensity of tumor allele one/intensity of tumor allele two)/(intensity of normal allele one/intensity of normal allele two) (referred to as the percent LOH index). An LOH index less than 50% is considered significant.
Radiolabeling of Cells, Indirect Immunoprecipitation, and SDS-PAGE-Radiolabeling of cells, indirect immunoprecipitation, and SDS-PAGE were performed as described (24).
Transfection-Cells were transfected with a plasmid encoding the wild-type ␤ 2 m or the mutant ␤ 2 m (␤ 2 m C25W ) utilizing Lipofectaminemediated gene transfer (Invitrogen) according to the manufacturer's instructions. Briefly, pcDNA3.1-b2m, pcDNA3.1-b2m C25W , or the empty plasmid pcDNA3.1-neo (Invitrogen) was mixed with Lipofectamine TM 2000 before being added to melanoma cells grown in monolayers with a 90% confluence. Cells were selected 2 days after transfection in medium containing G418 sulfate (1 mg/ml) (Calbiochem). After ϳ2-3 weeks of selection, G418-resistant clones were picked, expanded, and screened by flow cytometry for HLA class I expression. Positive clones were then further expanded in complete media supplemented with a maintenance dose (0.3 mg/ml) of G418.
Recombinant Vaccinia Virus-Virus encoding full-length human tyrosinase (TyrVac) was constructed as described previously (25). Purified recombinant vaccinia virus stock was titered and tested for proper expression of tyrosinase using specific HLA-A2-restricted CTL (data not shown).
Cytotoxicity Assay in Vitro-Standard 51 Cr release assays were performed to determine CTL recognition of the HLA-A2-restricted epitope from the melanocyte differentiation protein tyrosinase (tyrosinase 369 -377D ). Target cells were prepared by either loading with tyrosinase 369 -377D peptide (40 g/ml) for 1 h at room temperature or by infecting with TyrVac (10 plaque-forming units/cell, 10 6 target cells) in 1 ml of Hanks' balanced salt solution supplemented with 0.1% bovine serum albumin, 1.6 mM MgSO 4 , and 1.8 mM CaCl 2 for 1 h and then 4 ml of RPMI 1640 medium supplemented with 10% fetal bovine serum for 8 h to allow for expression followed by labeling with 100 Ci of Na 51 CrO 4 for 1 h for a standard 4 h 51 Cr release assay. Percent cytotoxicity was calculated by the formula: % specific lysis ϭ ((experimental release Ϫ spontaneous release)/(total release Ϫ experimental release)) ϫ 100.

Marked HLA Class I Down-regulation on VMM5B Melanoma Cells-
Fluorescence-activated cell sorting analysis of cells stained with mAb W6/32 showed that HLA class I molecules were barely detectable on melanoma cells VMM5B as compared with autologous VMM5A melanoma cells (Fig. 1A). The low level of staining of VMM5B cells with mAb W6/32 does not represent nonspecific background staining because no staining was detected when the ␤ 2 m-deficient FO-1 cells (8)  To corroborate the specificity of the staining of VMM5B cells by mAb W6/32, cells were treated with low pH (pH 3.0) and then stained with mAb W6/32, a method used to disintegrate the trimeric HLA class I complex on the cell surface (21). As shown in Fig. 1B, following treatment with pH 3.0, VMM5B cells were not stained by mAb W6/32. These results imply the dissociation of peptides and ␤ 2 m from the HCs because the antigenic determinant recognized by mAb W6/32 requires the association of HCs with ␤ 2 m for its expression (10). This possibility is supported by the restoration of the staining with mAb W6/32 following loading of low pH-treated VMM5B cells with wild-type ␤ 2 m along with a HLA-A2-binding peptide (HER-2/neu 369 -397 , KIFGSLAFL, t1 ⁄ 2 ϭ 481.2 min). This result reflects the reassembly of the trimeric HLA class I complex on the cell surface. It is noteworthy that the level of the restored HLA class I expression is similar to that on untreated cells, remaining at a low level, unlike the full restoration of HLA class I expression on low pH-treated autologous VMM5A cells following loading with exogenous peptide and ␤ 2 m (data not shown). These results suggest that the amount of HCs transported to the plasma membrane is the limiting factor in VMM5B cells. This finding is different from the efficient transport to the cell surface of open-form (peptide-free or low affinity peptide-bound) HCs, which can subsequently be refolded to their native, bioactive conformation following loading with a high affinity peptide, as has been observed in the TAP-deficient T2 cells (3). To determine whether the low level of HLA class I molecules on the membrane of VMM5B cells is caused by a reduced level of all HCs and/or by a defect in ␤ 2 m, HC and ␤ 2 m expression in VMM5A and VMM5B cell lysates were analyzed by Western blotting. As shown in Fig. 2A, HCs were detected in both cell lysates and were markedly up-regulated by IFN-␥. In contrast, ␤ 2 m was barely detectable with ␤ 2 m-specific rabbit polyclonal antibodies in a VMM5B cell lysate following IFN-␥ treatment. ␤ 2 m was not detected in a VMM5B cell lysate with a panel of ␤ 2 m-specific mouse mAb ( Fig. 2A  and data not shown). The latter results may reflect the insufficient sensitivity of the Western blotting technique because ␤ 2 m was detected in VMM5B cells by intracellular staining with the ␤ 2 mspecific mAb L368 (Fig. 2B). These results in conjunction with the lack of up-regulation of HLA class I molecules by IFN-␥ suggest that the abnormal HLA class I phenotype of VMM5B cells is caused by a defect in ␤ 2 m.
Cys-to-Trp substitution in ␤ 2 m Caused by a Point Mutation in the ␤ 2 m Gene in VMM5B Cells-To investigate whether structural mutation(s) underlie the low ␤ 2 m level and its functional abnormalities, we amplified the full-length ␤ 2 m cDNA fragments (Fig. 3A) and performed nucleotide sequencing of the open reading frame of the ␤ 2 m gene in VMM5B cells. As shown in Fig. 3, B and C, a cytosineto-guanine transversion mutation at position 135 in exon 2 was detected. This mutation changes codon 25 from Cys to Trp (␤ 2 m C25W ) (Fig. 4A), abolishing the disulfide bond between residue 25 (Cys-25) and 80 (Cys-80) of the full-length ␤ 2 m protein (B). It is noteworthy that this mutation was not acquired during in vitro culture of the VMM5B cell line because the identical mutated nucleotide was detected in the genomic DNA extracted from the melanoma metastasis from which the cell line had been derived (Fig. 5A). Moreover, the wild-type sequence of the ␤ 2 m gene in VMM5B cells was not detected suggesting that one ␤ 2 m gene copy was lost. This pos- sibility is supported by the LOH identified at two chromosome 15 short tandem repeat sites (D15S126 and D15S209) (23) flanking the ␤ 2 m gene in genomic DNA extracted from VMM5B cells and the corresponding melanoma metastasis (Fig. 5, B and C).
Instability of ␤ 2 m C25W in VMM5B Cells-To test whether loss of a disulfide bond in the mutant ␤ 2 m identified in VMM5B cells caused conformational instability of the full-length protein, we examined the effect of Cys-to-Trp substitution at residue 25 on the stability of ␤ 2 m structure using molecular modeling. The wild-type ␤ 2 m is a ␤-sandwich structure composed of two anti-parallel ␤-pleated sheets connected by a disulfide bond between Cys-25 and Cys-80 (Fig. 6A). Each of the ␤-pleated sheets contains three ␤ strands. In the wild-type Cys-25-Cys-80 hydrophobic core, the linked sulfur atoms exhibit favorable van der Waals contacts with respect to the surrounding atoms 4.43, 4.81, and 3.88 Å for Cys25SG-Val27CG1, Cys25SG-Gln8C␣, and Cys80SG-Val82CG2, respectively (Fig. 6B). However, when Cys-25 is replaced with Trp-25, the bulky indole ring of Trp side chain displays drastic steric clashes with the side chains of the neighboring residues. These clashes occur with all 14 possible Trp-25 side chain rotameric conformations analyzed. Fig. 6C shows one representative conformation in which the ring carbon and nitrogen members are within the unfavorable van der Waals distances with the neighboring atoms 2.63, 1.62, 2.74, and 1.95Å for Trp25CZ3-Tyr66CD1, Trp25CZ2-Val27CG1, Trp25NE1-Val82CG2, and Trp25CD1-Cys80SG, respectively. Because all of these distances are below 3.0Å, the minimum distance between two nonbonded carbon atoms, the free energy is drastically increased, leading to a thermodynamically unstable state of the mutant ␤ 2 m (␤ 2 m C25W ).

Degradation of ␤ 2 m C25W and Lack of Stable HC-␤ 2 m C25W Association in VMM5B
Cells-Next we tested whether ␤ 2 m C25W was degraded by the proteasome in VMM5B cells, especially because it was present at a very low level intracellularly, approximately at a level 29-fold and 41-fold lower than that in VMM5A cells under basal conditions and following incubation with IFN-␥ (300 units/ml), respectively (Fig. 2B). To this end, VMM5B cells were sequentially incubated with IFN-␥ (300 units/ml) for 24 h at 37°C and with the proteasome inhibitor MG132 (5 M) for 12 h at 37°C. The intracellular levels of the steady-state free ␤ 2 m C25W , free HCs, and HC-␤ 2 m C25W complexes in IFN-␥/MG132treated and in IFN-␥-treated VMM5B cells were compared utilizing fluorescence-activated cell sorting analysis of cells intracellularly stained with mAbs. As shown in Fig. 7, A and B, both the ␤ 2 m C25W and HCs were increased ϳ2-fold following incubation of cells with IFN-␥, but the level of HC-␤ 2 m C25W complexes remained unchanged. In the presence of IFN-␥ and MG132, the level of ␤ 2 m C25W increased markedly (5-fold above that in untreated cells) along with an ϳ3-fold increase in HCs, but the level of HC-␤ 2 m C25W complexes still remained unchanged. On the other hand, increased secretion of ␤ 2 m C25W did not appear to play a role in its low intracellular accumulation because the level of ␤ 2 m C25W , HCs, and HC-␤ 2 m C25W complex expression was not increased by the addition, 4 h before harvest, of monensin, an inhibitor of trans-Golgi-to-secretory vesicles traffic, to MG132-treated cells cultured in the presence of IFN-␥. Therefore, ␤ 2 m C25W was removed through proteasome-mediated degradation. Even when it accumulates, the mutant ␤ 2 m cannot form stable complexes with HCs. This conclusion is corroborated by the lack of immunoprecipitation of HC-␤ 2 m C25W complexes with mAb W6/32 from MG132 and IFN-␥treated VMM5B cell lysates (Fig. 7C). Attempts to obtain biochemical evidence for the degradation of ␤ 2 m C25W in pulse-chase experiments failed because the available methods were not sufficiently sensitive to detect the mutant ␤ 2 m in immunoprecipitation ( Fig. 7C and data not  shown).

Restoration of HLA Class I Expression and Peptide Presentation on VMM5B Cells following Exogenous and Endogenous Reconstitution of
Wild-type ␤ 2 m Expression-Despite HLA class I expression, VMM5B cells were resistant to lysis by HLA-A2-restricted, tyrosinase-derived peptide tyrosinase 369 -377D (YMDGTMSQV, t1 ⁄ 2 ϭ 212.6 min)-specific CTLs, even following pulsing target cells with the tyrosinase 369 -377D peptide (Fig. 8, A and D). To determine whether reconstitution of wildtype ␤ 2 m expression is sufficient to restore HLA class I expression and its ability to present peptides to CTLs, we transfected VMM5B cells with a wild-type ␤ 2 m cDNA. As shown in Fig. 8B, expression of HLA class I molecules was restored on stable clones of ␤ 2 m-transfected VMM5B cells (VMM5B.␤2.7G) compared with mock-transfected VMM5B cells (VMM5B.neo). Similarly, the gene products of HLA-A and HLA-B, HLA-C loci were restored, although the latter molecules were expressed at a level ϳ10-fold lower than the former ones (Fig. 8B). To test whether the lack of peptide presentation is caused by the mutated ␤ 2 m (␤ 2 m C25W ) associated with HC, VMM5B.neo cells were treated with low pH to dissociate the HC-␤ 2 m C25W complexes on the cell surface before being loaded with the wild-type ␤ 2 m together with the tyrosinase 369 -377D peptide and tested for susceptibility to lysis by tyrosinase 369 -377D -specific CTLs. VMM5B.␤2.7G cells were treated in a similar fashion and analyzed in parallel as a control. Although acidstripped VMM5B.␤2.7G cells became sensitive to CTL lysis following loading with either the tyrosinase 369 -377D peptide by itself or both, the wild-type ␤ 2 m and the tyrosinase 369 -377D peptide (Fig. 8C, left panel), VMM5B.neo cells were sensitive to CTL lysis only following loading with both the wild-type ␤ 2 m and the tyrosinase 369 -377D peptide (Fig. 8C,  right panel). These results indicate that peptide presentation is restored on VMM5B cells only when both the wild-type ␤ 2 m and the tyrosinase 369 -377D peptide are present. Therefore, the presence of ␤ 2 m C25W does not allow exogenous peptide loading onto HLA class I molecules.
To test whether ␤ 2 m C25W affects endogenous peptide loading, VMM5B.neo cells were transduced with TyrVac, a tyrosinase-expressing vaccinia strain, and analyzed for susceptibility to lysis by HLA-A2restricted, tyrosinase 369 -377D peptide-specific CTLs. As shown in Fig.  8D, TyrVac-infected VMM5B.neo cells were resistant to CTL lysis, whereas TyrVac-infected VMM5B.␤2.7G cells were sensitive. In these experiments, the tyrosinase-expressing DM6 cells were used as a positive control, whereas the untransduced VMM5B, VMM5B.neo, and VMM5B.␤2.7G cells were used as specificity controls. Therefore, in VMM5B cells, the endogenously processed tyrosinase 369 -377D peptide cannot be presented by ␤ 2 m C25W associated with HLA-A2 heavy chains unless the wild-type HLA-A2-␤ 2 m dimer is available in sufficient amount following reconstitution with wild-type ␤ 2 m expression. These   data indicate that although expressed on the cell membrane, HLA class I heavy chains, associated with ␤ 2 m C25W , cannot present peptides.
To exclude the possibility that defects in antigen presentation by VMM5B cells were caused by APM abnormalities, the level of APM component expression in VMM5A and VMM5B cells was compared. No marked difference was detected, except for lower LMP7, LMP10, calnexin, and tapasin expression in VMM5B cells than in VMM5A cells, as measured by Western blotting (Fig. 9) and by flow cytometry (for results, see Ref. 7). However, LMP7, LMP10, calnexin, and tapasin downregulation did not appear to have an impact on antigen presentation because peptide presentation was restored in VMM5B cells transfected with a wild-type ␤ 2 m cDNA (VMM5B.␤2.7G) (Fig. 8D). This observation in conjunction with the failure of exogenous peptide loading onto VMM5B cells, as compared with the wild-type ␤ 2 m-transfected VMM5B.␤2.7G cells, indicates that defective HLA class I-associated antigen presentation in VMM5B cells is caused by a dysfunctional ␤ 2 m.
Restoration of Low Level of HLA Class I Expression on ␤ 2 m-deficient Cells Transfected with a cDNA Construct Encoding the Mutant ␤ 2 m C25W -To exclude the possibility that the observed low level of HLA class I expression on VMM5B cells is a cell line-specific phenomenon, we transfected a ␤ 2 m C25W -expressing cDNA construct into FO-1 cells, a well characterized ␤ 2 m-deficient cell line harboring a large deletion in the ␤ 2 m gene (8). Fourteen stable clones were established and analyzed for HLA class I expression on the cell surface by flow cytometry with mAb W6/32. Ten of them were stained weakly by mAb W6/32 as compared with the mock-transfected (FO-1.neo) and wild-type ␤ 2 m-transfected (FO-1.␤2) controls ( Fig. 10 and data not shown). These results indicate that the mutant ␤ 2 m C25W alone is able to restore low level of HLA class I expression on the cell surface regardless of the differences in the genetic background.

DISCUSSION
In this report, a naturally occurring, full-length mutant ␤ 2 m (␤ 2 m C25W ) in the melanoma cell line VMM5B and the metastasis from which the cell line originated was found to cause HLA class I down-regulation rather than loss because of its impaired association with HCs. Notably, the HC-␤ 2 m C25W dimer does not appear to constitute a conventional peptide-receptive conformation because neither exogenous nor endogenous peptides can be presented on it, as indicated by the lack of induction of lysis by the cognate CTLs. These findings represent the first example of a ␤ 2 m structural abnormality that does not cause total HLA class I loss but causes defects in antigen presentation associated with their down-regulation. Importantly, the ability of the mutant ␤ 2 m C25W to restore a low level of HLA class I expression is not restricted to VMM5B cells because the same phenotype was observed in the allogeneic ␤ 2 m-deficient cell line FO-1 transfected with a ␤ 2 m C25W -expressing construct.
Several examples of single amino acid substitutions in proteins involved in HLA class I-associated antigen presentation, such as TAP1,  have shown that such mutations do not affect the level of protein expression (26). However, this is not the case for ␤ 2 m C25W presented here. The steady-state level of ␤ 2 m C25W in VMM5B cells is ϳ29-fold lower than that of its wild-type counterpart in VMM5A cells. Such a low level of accumulation could be explained by the increased degradation of ␤ 2 m C25W because of the structural instability caused by the Cys-to-Trp substitution at position 25. It is likely that a drastic alteration in the mode of folding of the nascent ␤ 2 m polypeptide chain occurs, resulting in either a folded protein with an altered three-dimensional conformation or an unfolded protein with a thermodynamically unstable structure. The latter possibility is supported by the results of our structural analysis because: (i) the strong steric hindrance by the Trp side chain is likely to interfere with the hydrophobic interaction between the two ␤-pleated sheets, and (ii) the disappearance of the disulfide link between Cys-25 and Cys-80 may not keep ␤ 2 m in an energetically favored conformation. The accumulation of ␤ 2 m C25W in VMM5B cells incubated with a proteasome inhibitor further suggests that ␤ 2 m C25W is unstable and ready for proteasome-mediated degradation. On the other hand, the predicted unfolding effect of the C25W substitution may not occur when Cys is replaced with a small, hydrophobic amino acid, such as alanine. In this situation, no apparent steric hindrance would take place, and the two ␤-pleated sheets may remain closely contacted, resulting in the proper folding of the ␤ 2 m protein. The importance of disulfide bonds in the structure and function of a protein has been indicated in several examples; one is represented by the replacement of Cys at position 634 of the RET (rearranged during transfection) proto-oncogene product. This type of mutation is present in 85% of individuals suffering from multiple endocrine neoplasia 2A (27). mAb W6/32, which recognizes a conformational epitope expressed by the ␤ 2 m-associated HCs, detected a low level of HLA class I molecules expressed on the surface of VMM5B cells and inside them. This phenomenon is likely to result from a stable interaction between a small population of aberrantly folded ␤ 2 m C25W and HCs because the amount of HC-␤ 2 m C25W complexes was not increased significantly in IFN-␥and MG132-treated VMM5B cells despite the marked elevation of individual HC and ␤ 2 m C25W . The inability of the HCs associated with aberrantly folded ␤ 2 m C25W to receive their peptide ligands, even when the cognate peptide has a high binding affinity (e.g. tyrosinase 369 -377D , YMDGTMSQV, t1 ⁄ 2 ϭ 212.6 min), may result from the lack of properly conformed peptide binding grooves on the HC-␤ 2 m C25W heterodimer. This possibility is supported by the lack of refolding on the cell surface of the mouse major histocompatibility complex molecule H-2D d HC associated with an artificially mutated ␤ 2 m (␤ 2 m W60A ) following loading with exogenous peptides (28). In the latter situation, ␤ 2 m W60A acts in a dominant-negative fashion. On the other hand, we do not know at present whether the formation of HC-␤ 2 m C25W complexes occurs at the stage of co-translational translocation of individual HC/calnexin and ␤ 2 m C25W into the ER or at a later stage when each of the mature proteins are present.
It is noteworthy that not all the tumor cells in the metastasis, from which the VMM5B cell line had been derived, carry the described antigen presentation defect because the freshly isolated tumor cells showed a certain degree of susceptibility to tyrosinase-specific CTLs, as reported in our previous study (7). In this regard, the VMM5B cell line is likely to represent a subpopulation of tumor cells in the metastasis that has escaped control by CTLs in the course of the disease. Also worthy of mentioning is that the patient did not receive any type of immunother-apy in the course of the disease. Therefore, the elicited CTL response was spontaneous in nature as opposed to the response introduced by T cell-based immunotherapy. This is also reflected by the location of the identified ␤ 2 m gene mutation (TGC to TGG in exon 2), outside the mutational hot spot in exon 1, which in melanoma has been found mostly in patients exposed to T cell-based immunotherapy-related immune selective pressure (29). Whether this type of mutation will represent a common one among malignant cells exposed to spontaneous CTL immune selective pressure remains to be determined. From a practical viewpoint, the resistance of VMM5B cells to lysis by CTLs despite HLA class I expression on the cell surface represents another example of lack of correlation between HLA class I expression and CTLmediated lysis in malignant cells (30). Understanding the mechanisms underlying this phenomenon as reported here may contribute to defining the molecular basis of tumor escape and optimization of the criteria to select patients to be treated with immunotherapy.