Low Constitutive Cell Surface Expression of HLA-B Is Caused by a Posttranslational Mechanism Involving Glu¹⁸⁰ and Arg²³⁹

The classical HLA class I molecules, that is, HLA-A, -B, and -C, bind and present peptides on the surface of a wide variety of cells. The peptides may originate from the cell’s own proteome or from an intracellular pathogen, for example, a virus. The HLA–peptide complex is monitored by cytotoxic T lymphocytes that recognize foreign peptides and induce apoptosis in the cells that present them. Cancer cells can also be identified and removed because of the mutated or aberrantly expressed peptides they may present. HLA class I molecules consist of an extremely polymorphic transmembrane H chain forming the peptide-binding groove and a non–covalently associated L chain β₂-microglobulin (β₂m). β₂m is essential for peptide loading (1) and for maintaining the stability of the HLA–peptide complex (2, 3). Cells that are β₂m deficient express no HLA-A and -B on the surface (4).

Quantitative differences in expression levels are of clinical importance. Reduced HLA expression is commonly found in cancer cells as well as in cells infected by certain intracellular pathogens, and this can render them undetectable for the immune system (5–8). Also, the incidence of severe graft-versus-host disease and nonrelapse mortality in HLA-C–mismatched allogeneic bone marrow transplantation correlates with the expression level of the mismatched patient HLA allomorph on T cells (9). Even minor differences of up to 3-fold in the normal cell surface expression of the various allomorphs may be of importance for immune responses as shown by Apps et al. (10). They found a correlation between the normal cell surface expression levels of HLA-C on T cells and progression of HIV infection (10).

We have recently found that several cell types in the human body vary widely in their expression of the individual classical HLA class I Ags. Thus, we found that expression of HLA-B was often low or absent on many types of human multipotent stem cells and on some differentiated cell types, whereas HLA-A expression was high on most cells (11, 12). In mesenchymal stem cells, we found a 17- to 40-fold lower protein expression of HLA-B when compared with HLA-A (11). These differences clearly exceed those found between HLA-C allomorphs in T cells and may have important implications for the immune response. They are not caused by regulation of transcription, as the mRNA levels of HLA-A, -B, and -C were similar despite very different cell surface expression (11).

By using the HEK293T cell line as a model system, we found that the difference in expression is primarily coding sequence (CDS)–dependent and therefore most likely related to structural differences. Expression studies of transfected HEK293T cells with chimeric HLA-A2 and -B8 constructs revealed that the C-terminal α2 domain and the α3 domain were especially important for the differential expression levels (13).

In this study, we analyzed the amino acid differences between HLA-A2 and -B8 in the C-terminal part of the α2 and the entire α3 domain using a combination of chimeric constructs and point mutations. We identified two amino acid differences in particular that each on its own greatly altered the cell surface expression levels. Gln¹⁸⁰ and Gly²³⁹ were found to be important for the high expression levels of HLA-A. The two amino acids are located close together in the tertiary structure, and they could possibly represent a binding site for an (as of yet) undescribed molecule that discriminates between HLA-A and -B allomorphs or impact the association with chaperones or β₂m in the Ag-processing pathway.

Full-length CDS of HLA-A*02:01:01:01 was PCR amplified with forward primer (A2 forward) 5′-ATGGCCGTCATGGCGC-3′ and reverse primer (A2 reverse) 5′-TCACACTTTACAAGCTGTGAGAG-3′ using cDNA from a telomerase-immortalized human mesenchymal stem cell line hMSC-Tert4 (14) as the template. Full-length CDS of HLA-B*08:01:01 was PCR amplified with forward primer (B8 forward) 5′-ATGCTGGTCATGGCGC-3′ and reverse primer (B8 reverse) 5′-TCAAGCTGTGAGAGACACATCA-3′ using cDNA from human adipose–derived MSC-661 cells (11) as the template. Full-length CDS of HLA-B*27:02 was PCR amplified with forward primer 5′-ATGCGGGTCACGGCG-3′ and reverse primer 5′-TCAAGCTGTGAGAGACACATCA-3′ using cDNA from hMSC-Tert4 (14) as the template. Full-length CDS of β₂m was PCR amplified with forward primer (β₂m forward) 5′-ATGTCTCGCTCCGTGG-3′ and reverse primer (β₂m reverse) 5′-TTACATGTCTCGATCCCACTTAACT-3′ using cDNA from HEK293T cells. Full-length CDS of DsRED2 was PCR amplified with forward primer 5′-ATGGCCTCCTCCGAGAACG-3′ and reverse primer 5′-CTACAGGAACAGGTGGTGGCG-3′ using the pEF1α–internal ribosome entry site 2 (IRES2)-DsRed-Express2 vector as the template (Clontech). Amplified PCR products were cloned into the mammalian pEF6/V5-His TOPO TA expression vector (Life Technologies) and verified using BigDye terminator v3.1 (Life Technologies) sequencing technology.

The IRES2-AcGFP1 cassette was amplified from pEF1α-IRES2-AcGFP1 (Clontech) with primers containing overhangs with restriction enzyme sites for NotI (forward primer) and XbaI (reverse primer). The forward primer sequence was 5′-TTATCCAATAGCGGCCGCGCCCCTCTCCCTCCCC-3′ and the reverse primer sequence was 5′-CATCGTCTAGACTCACTTGTACAGCTCATCCATGC-3′. The parts that anneal to either IRES2 or AcGFP1 are underlined. Both the PCR-amplified IRES2-AcGFP1 cassette and the pEF6/V5-His TOPO TA expression vector containing the gene to be transfected were incubated with NotI and XbaI restriction enzymes (New England BioLabs), purified using the Illustra GFX PCR DNA and gel band purification kit (GE Healthcare Life Sciences), and ligated using the Quick Ligation kit (New England BioLabs) downstream of the gene of interest.

The HLA-A2 or -B8 with β₂m full CDSs in the same pEF6/V5-His TOPO TA expression vector was constructed by amplifying HLA-A2 or -B8 CDS from a pEF6/V5-His TOPO TA expression vector using HLA-A2 (A2 forward) or -B8 (B8 forward) forward primer, respectively (shown above), together with a vector reverse primer (5′-TATAGACGTCTTAACGGGAA-3′) that also functioned as an anchor for a 5′ anchor present in the forward primer for the IRES2-β₂m cassette. The IRES2 forward primer with a 5′ anchor was 5′-AAGGGCAATTCTGCAGATATGCCCCTCTCCCTCCCC-3′ (the 5′ anchor part of the primer is underlined). The β₂m reverse primer is shown above (β₂m reverse) . HLA-A2 or -B8 was fused with the IRES2-β₂m cassette in a subsequent PCR using either HLA-A2 or -B8 forward primers in combination with the β₂m reverse primer and either the HLA-A2 or -B8 CDS together with the IRES2-β₂m cassette as template and then cloned into the pEF6/V5-His TOPO TA expression vector.

The HLA-A2_1–176/B8_177–338 constructs with position 177 Asp→Glu or position 180 Glu→Gln substitutions were created using the previously described HLA-A2_1–176/B8_177–338 construct as the template (13). The point mutation was introduced by amplifying a leading fragment ending just before the codon to be changed in a region where HLA-A2 and -B8 are identical, as well as a trailing fragment that overlaps the leading fragment with 20 nt. The trailing fragment was generated using a forward primer that introduced the desired codon change, and the two fragments were fused in a subsequent PCR before being cloned into the pEF6/V5-His TOPO TA vector. The HLA-A2 forward primer (A2 forward) for the leading fragment is shown above and the reverse primer sequence was 5′-TCCTTCCCGTTCTCCAGGT-3′. The HLA-B8 reverse primer (B8 reverse) is shown above, and the forward primer for the position 177 Asp→Glu construct was 5′-CCTGGAGAACGGGAAGGAGACGCTGGAGCGCGCGGACCCC-3′, and for the position 180 Glu→Gln construct was 5′-CCTGGAGAACGGGAAGGACACGCTGCAGCGCGCGGACCCC-3′. The IRES2-AcGFP1 cassette was inserted in the plasmids as previously described.

The point mutations at positions 239 and 280 were created by fusing two fragments with a 20-bp overlap wherein the mutation was introduced in the primer overhangs. For the HLA-A2 leading fragment, the forward primer (A2 forward) is shown above, and the reverse primer for the position 239 Gly→Arg substitution was 5′-TGGAAGGTTCTATCCCCTGCAGGCCTGGTCTCCACGAGCT-3′, and for the position 280 Pro→Ser substitution it was 5′-GGGATGGTGGACTGGGAAGACGGCTCCCATCTCAGGGTGA-3′. For the HLA-A2 trailing fragment the reverse primer (A2 reverse) is shown above, and the forward primer for the position 239 Gly→Arg fragment was 5′-GCAGGGGATAGAACCTTCCAGAAGTGGGCGGCTGTGGTGG-3′, and for the position 280 Pro→Ser fragment it was 5′-TCTTCCCAGTCCACCATCCCCATCGTGGGCATCATTGCTG-3′. For the HLA-B8 leading fragment the forward primer (B8 forward) is shown above, and the reverse primer for the position 239 Arg→Gly substitution was 5′-TGGAAGGTTCCATCTCCTGCTGGTCTGGTCTCCACAAGCT-3′, and for the position 280 Ser→Pro substitution it was 5′-GGACGGTGGGCTGGGAAGACGGCTCCCATCTCAGGGTGAG-3′. For the HLA-B8 trailing fragment the reverse primer (B8 reverse) is shown above, and the forward primer for the position 239 Arg→Gly substitution was 5′-GCAGGAGATGGAACCTTCCAGAAGTGGGCAGCTGTGGTGG-3′, and for the position 280 Ser→Pro substitution it was 5′-GTCTTCCCAGCCCACCGTCCCCATCGTGGGCATTGTTGCT-3′. Gene construct analysis was facilitated by the use of the CLC Genomics workbench software (CLC bio, Qiagen).

HLA-A2 with -B27 and HLA-B8 with -A3 chimeric constructs were generated by PCR amplifying the two fragments that were to be fused using a reverse primer for the leading fragment that was complementary to the forward primer of the trailing fragment. The two fragments were fused in-frame in a second PCR, and the amplified product was cloned into the mammalian pEF6/V5-His TOPO TA expression vector and verified using BigDye terminator v3.1 (Life Technologies) sequencing technology. The forward primer used for leading HLA-A*02 fragments was 5′-ATGGCCGTCATGGCGC-3′, and for leading HLA-B*08 fragments was 5′-ATGCTGGTCATGGCGC-3′. The reverse primer used for the leading fragments A2_1–176 and B8_1–176 was 5′-TCCAGGTAGGCTCT-3′. The reverse primers used for trailing HLA-A3 and -B7 fragments were 5′-TCACACTTTACAAGCTGTGAGAG-3′ and 5′-TCAAGCTGTGAGAGACACATCA-3′, respectively. The forward primer used for the trailing fragments A3_177–341 and B27_177–338 was 5′-AGAGCCTACCTGGA-3′.

Total RNA was isolated using the RNeasy Plus mini RNA purification kit (Qiagen) and quantified by NanoDrop spectrophotometry (Thermo Fisher Scientific). The cDNA was synthesized using the RevertAid H Minus first strand cDNA synthesis kit with random hexamer primers (Thermo Fisher Scientific).

The HEK293T cell line (15) was grown in DMEM with l-glutamine (Life Technologies) supplemented with 10% (v/v) FBS (Life Technologies), 100 μg/ml streptomycin, and 100 U/ml penicillin (Life Technologies).

Cells were transfected using jetPEI (Polyplus-transfection SA) according to the manufacturer’s protocol in a 24-well cell plate (TPP Techno Plastic Products) and maintained in medium with 100 μg/ml blasticidin (InvivoGen). The cells transfected with constructs containing a reporter for transfection were selected with blasticidin for 3 wk before being analyzed by flow cytometry. Cloning of constructs that did not contain a reporter for transfection was performed as previously described (13).

Three replicates of untransfected and transfected HEK293T cells were grown in DMEM with l-glutamine, 10% FBS, and antibiotics (Invitrogen) to near confluency, detached and washed three times in cold PBS (Invitrogen), centrifuged at 250 × g for 5 min at 4°C, and lysed in RIPA buffer (Sigma-Aldrich) containing a protease inhibitor mixture (Sigma-Aldrich). Samples were frozen at −80°C overnight, sheared five times with a 21-gauge needle/syringe and pelleted at 8000 × g for 10 min. Supernatants were transferred to new tubes and the protein concentration was quantified by NanoDrop spectrophotometry using the DC protein assay kit 2 (Bio-Rad) according to the manufacturer’s instructions. Twenty micrograms of protein whole-cell lysates was denatured at 95°C during 10 min and loaded together with SeeBlue Plus2 (Invitrogen) prestained molecular size markers onto 4–12% Bis-Tris NuPAGE gels (Invitrogen) and run under denaturing conditions with NuPAGE MOPS-SDS running buffer (Invitrogen). Proteins were blotted onto polyvinylidene difluoride membranes by electrotransfer at 4°C overnight with NuPAGE transfer buffer (Invitrogen). The polyvinylidene difluoride membranes were blocked with 5% milk (Bio-Rad) in TBST buffer and incubated at 4°C overnight with anti-tapasin (Abcam, catalog no. ab196764), anti-tapasin–related protein (Abcam, catalog no. ab57411), anti-TAP1 (Biorbyt, catalog no. orb247305), anti-TAP2 (Abcam, catalog no. ab180611), and anti-actin (Cell Signaling Technology, catalog no. 4967) Abs at saturating conditions, followed by incubation with either anti-mouse IgG (Cell Signaling Technology, catalog no. 7076) or anti-rabbit IgG (Cell Signaling Technology, catalog no. 7074) HRP-conjugated Abs during 1 h at room temperature. Immunoreactivity was detected with an ECL Plus Western blotting reagent kit (Pierce ECL Western blotting substrate; Thermo Fisher Scientific) according to the manufacturer’s instructions. Immunoblots were performed in triplicate for each Ab.

Cells were washed twice with PBS (pH 7.4; Life Technologies), and a total of ∼1 × 10⁶ cells were resuspended in 100 μl of FACS buffer (PBS [pH 7.4] containing 2% FBS and 2 mM EDTA). The cells were incubated with the primary mouse mAb at saturating conditions, as determined by titration, for 30 min at 4°C. For quantitative allotype-specific HLA expression measurement, unconjugated anti–HLA-A3 clone GAP.A3 (eBioscience, catalog no. CUST00579) and its isotype control IgG2a (eBioscience, catalog no. 16-4724-82) were used for the HLA-A3 quantification. Unconjugated anti–HLA-B7 clone BB7.1 (AbD Serotec/Bio-Rad, catalog no. MCA986) and its isotype control IgG1 (AbD Serotec/Bio-Rad, catalog no. MCA928) were used for the HLA-B7 quantification. Unconjugated anti–HLA-A2 clone BB7.2 (BD Biosciences, catalog no. 551230) and its isotype control IgG2b clone MPC-11 (BD Biosciences, catalog no. 557351) were used for the HLA-A2 and A2_1–176/B8_177–338 constructs with and without substitutions, as well as the A2_1–176/B27_177–338 construct. Anti–HLA-B8 biotin-conjugated (One Lambda/Thermo Fisher Scientific, catalog no. BIH0536A) and its isotype control IgG2b (BD Biosciences, catalog no. 557351) were used for the HLA-B8 and B8_1–176/A2_177–341 with and without substitutions, as well as the B8_1–176/A3_177–341 construct. Unconjugated anti–HLA-B27 (Merck Millipore, catalog no. MAB1285) and its isotype control IgG2a (eBioscience, catalog no. 16-4724-82) were used for the HLA-B27 construct. Screening for positive clones for the constructs without a reporter for transfection was done as previously described (13).

For indirect immunofluorescence, the cells transfected with constructs without reporter for transfection were further incubated with the polyclonal FITC-conjugated goat F(ab′)₂ anti-mouse Igs (Dako, catalog no. F0479), and cells transfected with constructs containing an AcGFP1 reporter for transfection were incubated with Alexa Fluor 647–conjugated goat F(ab′)₂ anti-mouse IgG (Thermo Fisher Scientific, catalog no. A21237). The secondary Ab was added at saturating conditions for 30 min at 4°C and washed twice in FACS buffer. The cells were analyzed on a CyAn ADP analyzer (Beckman Coulter). The FACS data were analyzed using Summit 4.3 (Beckman Coulter). For quantitative flow cytometry, the indirect immunofluorescence assay QIFIKIT (Dako) with beads carrying known numbers of murine IgG molecules was used to generate a standard curve according to the manufacturer’s protocol.

Statistical analyses were performed using GraphPad Prism 4 software. A Kruskal–Wallis test was used to test differences between medians for multiple groups. A p value <0.05 was considered statistically significant. When significant, the Mann–Whitney U test was used to test for differences between individual constructs. Significant p values are marked with an asterisk next to the uncorrected p value.

Expression of HLA-A- and -B allomorphs were studied in HEK293T cells by quantitative flow cytometry using allotype-specific Abs titrated to yield maximum fluorescence after staining with a fluorochrome-coupled secondary Ab (Fig. 1). Comparison with fluorescence signals from a set of beads with known numbers of murine IgG molecules allowed estimation of the number of HLA molecules on the cell surface. HEK293T is homozygous for HLA-A*03;B*07 but expresses A3 (median of 32,520 molecules per cell) to a much higher level than B7 (median of 6529 molecules per cell), thus resembling the expression pattern found in many cell types outside the hematopoietic system. This feature and the availability of allotype-specific Abs and the permissiveness of transfection led us to choose HEK293T as a model system.

First, we wanted to see the effect of transfection on the endogenous HLA expression and at the same time to eliminate the possibility that the low cell surface expression levels of HLA-B relative to -A were caused by competition for limited supplies of β₂m. A β₂m expression vector was designed with an IRES2-AcGFP1 cassette that served as the transfection reporter inserted downstream of the β₂m gene. The vector contained a blasticidin resistance gene. After 3 wk of culture in the presence of blasticidin, the expression of endogenous HLA-A3 and -B7 was measured by quantitative flow cytometry during overexpression of β₂m after gating on transfected cells (Figs. 2, 3A, 3B). The effect of transfection itself was measured using the same vector except that β₂m was replaced by DsRed2 (Fig. 3A, 3B). We observed only a slightly higher expression of HLA-A3 (1.1-fold) and -B7 (1.3-fold) when transfected with the β₂m-IRES2-AcGFP1 expression vector compared with untransfected cells (Fig. 3A, 3B). The increases, however, were not higher (HLA-A3, p = 0.13; HLA-B7, p = 0.39) than those observed for cells transfected with the control vector (Fig. 3A, 3B), implying a minor effect of transfection, but not β₂m overexpression, on endogenous HLA expression.

HEK293T cells transfected with expression vectors containing either HLA-A2-IRES2-AcGFP1 or -B8-IRES2-AcGFP1 showed a similar expression pattern, with HLA-A2 highly expressed (33,160 molecules per cell) and -B8 expressed at much lower densities (2,563 molecules per cell, p = 0.0012, Fig. 3C). The effect of overexpressing β₂m on the cell surface expression of the transgenic HLA-A and -B molecules was also studied. For this purpose, HEK293T cells were transfected with expression vectors containing either HLA-A2-IRES2-β₂m-IRES2-AcGFP1 or -B8-IRES2-β₂m-IRES2-AcGFP1 (Fig. 3C). Overexpression of β₂m had an effect comparable to that found for the endogenous HLA-A3 and -B7 expression and resulted in a 1.6-fold upregulation of HLA-A2 and 1.3-fold for HLA-B8 (Fig. 3C). Taken together, these data demonstrated that overexpression of β₂m had a limited effect on both endogenous and transgenic HLA expression levels, and that the level of endogenous β₂m was not the limiting factor accounting for the low HLA-B expression in HEK293T cells.

A possible explanation for the difference between HLA-A2 and -B8 cell surface expression in the HEK293T cell model could be intrinsic changes in the expression of key chaperones involved in the MHC class I Ag presentation pathway. Thus, we analyzed the expression of TAP1 and TAP2 as well as tapasin and tapasin-related protein by immunoblotting. Fig. 4 shows that in all cases examined, all molecules were expressed and no significant variation in the amounts of the given proteins were observable between untransfected cells and HLA-A2– or -B8–transfected HEK293T cells.

We have previously reported that differences between HLA-A2 and -B8 in the C-terminal part of the α2 domain and the α3 domain account for most of the discrepancy between the cell surface expression levels (13). There are 12 aa differences between HLA-A2 and -B8 but only 8 aa differences between HLA-A3 and HLA-B8 in this region, as outlined in Table I. To determine the importance of the individual differences, a chimeric construct (B8_1–176/A3_177–341) was made where the C-terminal part of the HLA-B8 molecule (aa 177–338) was exchanged with the corresponding segment of HLA-A3 (Fig. 5). HLA-A3 shares five amino acids (Pro¹⁸⁴, Pro¹⁹³, Ile¹⁹⁴, Gly²⁰⁷, and Gly²⁵³) with HLA-B8 in this region that differ from those in HLA-A2 (Table I). The plasmid was stably transfected into HEK293T cells, positive clones were isolated, and the cell surface expression was measured by quantitative flow cytometry using HLA-B8–specific Abs and compared with a B8_1–176/A2_177–341 plasmid construct. We found no significant difference in expression between the two chimeric constructs (Fig. 6A).

Likewise, an expression construct was made where the C-terminal part of HLA-A2 (aa 177–338) was exchanged with the corresponding part of HLA-B27 for comparison with an HLA-A2_1–176/B8_177–338 chimeric construct (Fig. 6B). The cell surface expression of the HLA-A2_1–176/B27_177–338 construct was much higher (8.6-fold) than that of the HLA-A2_1–176/B8_177–338 construct. HLA-B27 shares two amino acids with HLA-A2 that differ from HLA-B8 in the exchanged sequence (Table I). The huge difference in cell surface expression between the two chimeric constructs can thus be assigned to one of the two amino acid differences at positions 177 and 180 or both.

To determine the individual contributions from the amino acids at positions 177 and 180 to the expression of HLA, two variants of the HLA-A2_1–176/B8_177–338 construct were made changing the HLA-B8 amino acid to the corresponding amino acid in HLA-A2 (Fig. 7A). There was a significant difference in expression of the HLA-A2_1–176/B8_177–338 position 180 Glu→Gln construct having a 2.4-fold higher expression than the HLA-A2_1–176/B8_177–338 construct (Fig. 7A). We observed no significant difference in expression when introducing the position 177 Asp→Glu point mutation. This demonstrates that Gln¹⁸⁰, but not Glu¹⁷⁷, is important for high cell surface HLA expression. To investigate whether the expression of a naturally occurring HLA-B allomorph having Gln¹⁸⁰ and Arg²³⁹ would more closely resemble the expression of HLA-A, HEK293T cells were transfected with wild-type HLA-B*27:02 and cell surface expression was measured (Fig. 7B). The expression of HLA-B27 was markedly lower than that for HLA-A2 (5.3-fold) but significantly higher than that for HLA-B8 (2.4-fold), confirming that Gln¹⁸⁰ is correlated with higher cell surface expression levels but insufficient to raise the level to that of HLA-A2.

Of the five remaining amino acid differences (Thr/Ala at position 182, Met/Val at position 189, Gly/Arg at position 239, Ser/Pro at position 280, and Ile/Val at position 282) between HLA-A2 and -B8 in the terminal part of the α2 domain and in the α3 domain that were not addressed by the previous experiments, two are not conservative substitutions: positions 239 Gly/Arg and 280 Ser/Pro (Fig. 5, Table I). To investigate the importance of the amino acids at these two positions for cell surface HLA expression, constructs with point mutations were made and transfected into HEK293T cells. The expression of the HLA-A2 position 239 Gly→Arg construct was 5.7-fold lower than that for wild-type HLA-A2 (Fig. 8A). The opposite was observed for the HLA-B8 position 239 Arg→Gly construct, as it had a 3.3-fold higher expression than did wild-type HLA-B8 (Fig. 8B). In summary, having glycine at position 239 resulted in higher cell surface expression levels of HLA. In contrast, making the reciprocal exchange of the amino acids Pro/Ser at position 280 between HLA-A2 and -B8 had no significant effect on the HLA expression levels, showing that this amino acid is not important for cell surface HLA expression (Fig. 8).

A three-dimensional model based on the published crystal structure of HLA-A2 revealed that Gln¹⁸⁰ and Gly²³⁹ are both partly surface exposed and positioned relatively close together, with a distance of 13.6 Å measured at the backbone (Fig. 9A, 9B). A similar distance was found for HLA-B8 (13.2 Å), and Glu¹⁸⁰ changes the charge from neutral to positive at the exposed area (Fig. 9C, 9D). The large side chain of HLA-B8 (arginine) at position 239 resides in a pocket formed by Tyr²⁰⁹, Thr¹⁷⁸, Arg¹⁸¹, and Asp¹⁸³ under formation of three hydrogen bonds between the nitrogen atoms of the guanidinium group and the carbonyl groups of Thr¹⁷⁸ and Arg¹⁸¹ and the carboxyl group of Asp¹⁸³, respectively (Fig. 9F, 9G). Interestingly, a very similar pocket exists in HLA-A2, but it is empty (Fig. 9E). The amino acids that were not found to influence cell surface expression levels were all surface exposed on the same side of the molecule (Fig. 9A–D).

We have previously shown that several human stem cell types and bone marrow–derived T lymphocytes constitutively express HLA-A allomorphs but only have a very weak or even absent cell surface expression of HLA-B (11, 12). Individual regulation of HLA class I allomorphs could have important functional implications because low expression levels may interfere with the functions of NK cells and allomorph-restricted T cells and may compromise an efficient immune response due to the diminished number of HLA molecules presenting pathogenic, nonself, or cancer-derived peptides.

In the present study, we investigated the contribution of individual amino acid differences between HLA-A and -B in the terminal part of the α2 domain and the α3 domain that could influence the cell surface HLA expression levels. There was no significant effect of interconverting either of the polymorphic amino acids in positions 177 and 280. Likewise, there was no significant effect of exchange en bloc for the amino acids in positions 184, 193, 194, 207, and 253. The interpretation of the exchange is complicated by the three amino acids that HLA-A3 does not share with either HLA-A2 or -B8 (Leu²⁷⁶, Leu²⁹⁴, and Thr³²¹), but the comparable cell surface expression of the two constructs indicates that the differing amino acids are not involved in the differential expression of HLA-A2 and -B8.

In contrast, we showed that Gln¹⁸⁰ and Gly²³⁹ are associated with high cell surface expression levels and indeed explain most of the difference in surface expression of HLA-A2 and -B8. The remaining three amino acids that differ between HLA-A2 and -B8 at positions 182, 189, and 282 are considered conservative substitutions using the Grantham’s amino acid distance matrix and were not investigated (16).

Gln¹⁸⁰ and Gly²³⁹, both associated with high expression, were found to be conserved in all available HLA-A sequences, with the full-length extracellular domain (460 sequences) available from the IMGT/HLA database (https://www.ebi.ac.uk/ipd/imgt/hla/) (17). In contrast, the amino acids associated with low expression are well represented among HLA-B allomorphs. Thus, >99% of all typed United States bone marrow donors (National Marrow Donor Program) carry Arg²³⁹-positive B alleles on both chromosomes. In fact, of all available sequences with full-length extracellular domains (676 sequences), Arg²³⁹ is conserved in all but three rare allomorphs. One exception is Gly²³⁹ containing HLA-B*73:01, which is probably derived by a recombinational event involving an HLA-A gene (18). The others are a single report of Lys²³⁹ (HLA-B*40:301) (19) and one report of a single nucleotide variation of HLA-B*51:01:10 introducing Gly²³⁹ designated HLA-B*51:30 (20). Glu¹⁸⁰, associated with low expression, is found in 19% of known full-length alleles constituting combined allele frequencies of 14 (n = 185,391), 22 (n = 416,581), and 31% (n = 1,242,890) in Asians, African Americans, and whites, respectively (21). HLA-B alleles combining both low expression variants (Arg²³⁹ and Glu¹⁸⁰) are thus quite common and carried by ∼53% of whites. Moreover, even the HLA-B*27:02 allomorph, which lacks Glu¹⁸⁰, was found to be poorly expressed compared with HLA-A2. It is therefore an intriguing possibility that low constitutive cell surface expression in some cell types is a general feature of HLA-B and is not restricted to the allomorphs examined in the present study.

A valid concern is that exchanging domains or single amino acids between different HLA allomorphs could disrupt the epitope recognized by the allotype-specific mAbs. However, the epitope for the HLA-A2 mAb used in this study has been mapped and found to be in the α1 and α2 domains (Trp¹⁰⁷, Lys¹²⁷, Gly¹⁶², and Arg¹⁶⁹), not involving the terminal part of the α2 domain or the α3 domain (22). The epitope for the HLA-B8 mAb has not been determined; however, HLA-B7, -B27, and -B8 have identical amino acid compositions in the α3 domain, and because the anti–HLA-B8 mAb discriminates between these three tissue types, the epitope must necessarily be situated within the more polymorphic α1 and/or α2 domains, too. Also, the observation of increased binding of the anti–HLA-B8 Ab to cells transfected with the B8_1–176/A2_177–341 or B8_1–176/A3_177–341 chimeric expression constructs compared with wild-type HLA-B8 strongly supports that the epitope is situated in the α1 and/or α2 domains, which are the only parts of these fusion proteins that are derived from HLA-B8.

Another necessary consideration is the possibility that the fusion proteins may introduce problems with proper folding. However, this is not likely the case because we, with the exception of a few surface-exposed amino acids in the terminal part of the α2 domain, only exchanged entire functional domains between HLA-A2, -A3, -B27, and -B8. Moreover, the observation that all B8-A2 and -A3 fusion proteins and point-mutated versions of B8 were expressed at a level similar to or higher than that of the wild-type HLA-B8 molecule argues against artificially introduced misfolding. In certain cell types, HLA-B8 is, in fact, capable of achieving cell surface expression levels comparable to those of HLA-A2, as we previously have reported in B lymphocytes (12). It is possible, however, that the differential expression of HLA-A2 and -B8 in HEK293T and certain normal cells results from a different requirement for chaperones or other molecules involved in folding and transport, and that the high expression of HLA-B8 in B lymphocytes is assisted by a higher concentration or a different repertoire of chaperones in this cell type. In this context, misfolding could be a means of regulating cell surface expression levels, although not energy efficient. It has been suggested that >30% of newly synthesized peptides are quickly degraded, suggesting that regulation by degradation is plausible (23), although the subject is controversial (24). Tapasin is a chaperone that is known to affect cell surface expression of HLA allomorphs differently. HLA-A2 and -B27 and to some extent HLA-A3 and -B7 are not dependent on tapasin for cell surface expression, whereas many other allomorphs are tapasin-dependent (25). However, HLA-B8 is only weakly dependent on tapasin, and the tapasin-deficient cell line 721.220 has a similar expression of HLA-A2 and -B8 (26). Moreover, we have shown that tapasin is indeed present in the HEK293T cells used in this study, and the low expression of the tapasin-independent HLA-B7 and -B27 allomorphs makes tapasin dependency an unlikely explanation for the observed differential cell surface expression of HLA-A and -B allomorphs.

Restricted peptide availability in the endoplasmic reticulum is another factor that could affect surface expression of allomorphs differently. We have previously shown that exchanging the α1 and α2 domains that form the peptide-binding groove between HLA-A2 and -B8 affects surface expression levels in HEK293T cells. The cell surface expression of HLA-A2 was 2.6-fold higher than that for the HLA-B8_1–176/A2_177–341 construct, and there was a 2.1-fold lower expression of HLA-B8 compared with HLA-A2_1–176/B8_177–338 (13). Different peptide preferences of HLA-A and -B combined with restricting supplies of peptides available in the endoplasmic reticulum of the model cell line could be involved. However, these effects are too small to account for the 15-fold difference in cell surface expression levels observed between HLA-A2 and -B8 in transfected HEK293T cells and hence are unable to explain the effects of exchanging single amino acids in the terminal part of the α2 domain and the α3 domain.

The amino acids at positions 180 and 239 that are highly important for cell surface expression levels of HLA-A2 and -B8 are located close together in the tertiary structure. In HLA-A2 the distance is 13.6 Å, and in HLA-B8 13.2 Å was measured between the α carbons of the two amino acids in the backbone. Furthermore, they are both exposed on the surface of the molecule. The surface area of various protein–protein interactions have been reported to be in the range from 381 to 3393 Ų (27). The short distance between the two highly important amino acids at positions 189 and 239 easily allows them to be within a normal protein–protein interaction area. An intriguing possibility is that the two amino acids at positions 180 and 239 could be involved in binding to a molecule that discriminates between HLA-A2 and -B8. In the case of Arg²³⁹, the fact that it bridges the α2 and α3 domains when docking into a pocket probably under the formation of three stabilizing hydrogen bonds also could argue for a direct impact on the folding of the α-chain. Both putative mechanisms warrant further investigation.

In summary, our data demonstrate a hitherto unrecognized posttranslational regulatory mechanism in the processing of HLA molecules within the classical HLA class I Ag presentation pathway in HEK293T cells that discriminates between HLA-A2 and -B8 and probably other A and B allomorphs as well. The amino acids at positions 180 and 239 are essential for constitutive surface expression, and the Gly/Arg dimorphism at position 239 is conserved among almost all known HLA-A and -B allomorphs, respectively. Although the mechanisms are still elusive, the evolutionary conservation of this difference in all human populations may imply that HLA-A and -B exhibit undisclosed discrete functions in the immune system.

We thank Knud Erik Vejen Nielsen for skillful technical contributions and the staff from the Molecular Biology section of the Department of Clinical Immunology for assistance with HLA typing.

The authors have no financial conflicts of interest.