Human hepatitis B viral infection outcomes are linked to naturally occurring variants of HLA-DOA that have altered function

Abstract

Human leukocyte antigen (HLA) molecules of the Major Histocompatibility complex class II (MHCII) bind and present pathogen-derived peptides for CD4 T cell activation. Peptide loading of MHCII in the endosomes of cells is controlled by the interplay of the non-classical MHCII molecules, HLA-DM (DM) and HLA-DO (DO). DM catalyzes peptide loading while DO, a MHCII substrate mimic, prevents DM from interacting with MHCII resulting in an altered MHCII-peptide repertoire and increased MHCII-CLIP. Although the two genes encoding DO (DOA and DOB) are considered non-polymorphic, there are rare natural variants. Our previous work identified DOB variants that altered DO function. Here we show that natural variation in the DOA gene also impacts DO function. Using the 1000 Genomes Project database, we show that ~98% of individuals express the canonical DOA*0101 allele, and remaining individuals mostly express DOA*0102, which we found was a gain-of-function allele. Analysis of twenty-five natural occurring DOα variants, that included the common alleles, identified three null variants and one variant with reduced and nine with increased ability to modulate DM activity. Unexpectedly, several of the variants produced reduced DO protein levels, yet efficiently inhibited DM activity. Finally, analysis of associated SNPs genetically linked the DOA*0102 common allele, a gain-of-function variant, with human Hepatitis B viral persistence. In contrast, we found that the DOα F114L null allele was linked with viral clearance. Collectively, these studies show that natural variation occurring in the human DOA gene impacts DO function and can be linked to specific outcomes of viral infections.

Introduction

Antigen processing and presentation is one of the key steps in the initiation and propagation of adaptive immune responses. This pathway results in the display pathogen-derived peptides at the cell surface bound to Major Histocompatibility Complex (MHC) molecules that are recognized by T cell receptors (TCRs) expressed on the surface of T cells, ultimately leading to T cell activation and the initiation of an adaptive immune response. MHC class II (MHCII) antigen processing and presentation results in the presentation of peptides from pathogen that transit through or persist in the endosomal compartments of antigen presenting cells (APCs) (1, 2). The resulting MHCII-peptide complexes are specifically recognized by TCRs on CD4 T cells which provide help by secreting cytokines and other molecules that direct the other cells of the immune system to develop the appropriate immune response to the invading pathogen. For example, CD4 T cells assist in activating antigen specific B cells which can ultimately result in the generation of pathogen-neutralizing antibody (Ab) responses.

The cellular and molecular pathways by which MHCII acquires peptide cargo have been examined in detail (3). Newly synthesized MHCII αβ heterodimers, associate with the invariant chain (Ii) during assembly in the ER. Ii occupies the peptide-binding groove and prevents premature peptide binding. Ii also targets MHCII-Ii complexes via the cell surfaces to late endosomal and lysosomal compartments (4) where Ii is degraded by resident proteases leaving only remnants of Ii, or class II-associated Ii peptides (CLIP), in the MHCII groove (5, 6). Exchange of CLIP for peptide from pathogenic (and self) proteins is catalyzed by HLA-DM (DM) (7-9). Once loaded, MHCII-peptide complexes traffic to the cell surface for presentation to and activation of CD4 T cells. Previous biochemical and crystallographic studies have shown that DM forms stable complexes with peptide-free, empty MHCII (i.e. a short-lived transition state) and that this interaction is disrupted when high-affinity peptides occupy the MHCII binding groove (10-17). Thus, when mixtures of low and high affinity peptides are present, as would be the case in the endosomes of MHCII expressing cells, DM favors the loading of high stability MHCII-peptide complexes, a process that is called peptide editing (15, 18, 19).

Peptide loading of MHCII by DM is finely tuned by HLA-DO (DO). DO, like DM, is a non-classical MHCII-like αβ constitutive heterodimer. Studies have shown that DO acts as a structural mimic of MHCII and binds to DM, inhibiting its ability to productively interact with MHCII (2). DO expression in B cells results in a broader repertoire of peptides presented compared to cells lacking DO (20). Presumably, the DO-dependent peptides are lower affinity peptides that would be exchanged if DM was fully active. Thus, the functional interplay between DM and DO determines which MHCII-peptide complexes are presented at the cell surface. However, there has been scant biological evidence showing that the modulation of the MHCII-peptide repertoire by DO impacts the adaptive immune response to pathogens. Our recent studies showing that H2-O, the mouse homologue of DO controls the neutralizing antibody response to mouse mammary tumor virus (MMTV) provided the first experimental evidence that this might indeed be the case (21).

While many mouse strains are susceptible to MMTV infection, I/LnJ mice are resistant to MMTV infection (22). MMTV resistance in I/LnJ mice is mediated by the development of a strong adaptive immune response, and specifically, a neutralizing Ab response (23). Positional cloning identified the β chain of H2-O (H2-Ob) as the gene conferring resistance to MMTV (21). Infection of H2-O deficient mice with MMTV resulted in a neutralizing Ab response to the virus, demonstrating that H2-Ob controls the immune response to mouse retroviruses (21). Although virus-resistant I/LnJ mice produce H2-Oβ, it produces a non-functional H2-O (21). Therefore, the functional H2-O found in MMTV susceptible mice blocks the development of a neutralizing Ab response. This finding led us to ask whether gene variants of human DOB could also contribute to the neutralizing Ab response to human viruses such as Hepatitis C virus (HCV). HCV resistance is also associated with the development of a neutralizing antibody response (24-26). Although DOB is mostly non-polymorphic, our previous study functionally characterized DOB variants and identified 5 loss- and 2 gain-of-function variants in terms of their ability to alter DM activity (21). One gain-of-function DOβ variant (G77V) was genetically linked to HCV persistence in humans. Thus, our data linked the lack of a neutralizing Ab response (HCV persistence) with a DOβ variant with increased DM inhibitory function. Collectively, our data identified a previously unknown role for DO (H2-O) in the control of persistent viral infections in mice and humans.

The DOα chain contributes the majority of contacts between DM and DO (2, 27, 28) suggesting that genetic variation in DOA may also confer altered DO function. Similar to DOB, the DOA gene is also mostly non-polymorphic and over 99% of all individuals express the 4 known common alleles (29). Using an in vitro cell line transfection-based screen to measure DO function, we showed that three of the four common DOA alleles, when paired with the most common DOB allele, had altered function. Two were gain-of-function variants whereas one was a loss of function allele and made no detectable protein due to a frameshift mutation. Using bioinformatic approaches we determined the naturally occurring haplotypic combinations of the DOA and DOB common allele genes. Functional analyses of the haplotypic pairs showed that polymorphisms in the DOA common alleles altered the ability of DO to modify DM activity however, changes in the DOB common alleles did not. Extension of the functional analysis to 21 rare DOA missense variants showed that 2 variants produced no detectable heterodimeric DO protein, 7 variants were gain-of-function mutations, and one variant had reduced function, despite producing more DO protein. Interestingly, a SNP in one of the common alleles, DOA*0102, a gain-of-function variant, was shown to be linked to an allele of HLA-DP that was previously associated with persistent Hepatitis B virus (HBV) infection. Conversely, a SNP in one of the DOα null alleles was linked to a HLA-DP allelic SNP that was associated with resistance to HBV. Thus, a specific DOA allele generated a DO protein with enhanced function that may be detrimental to HBV immunity, whereas a non-functional allele of DOα produced no protein and thus might have been beneficial with respect to HBV infection outcomes.

Materials & Methods

Cell Line

HeLa cells transfected with the class II transactivator (CIITA) (30) were grown in DMEM with 5% FBS at 37°C and 5% CO₂. CIITA expression allowed for the expression of the Ii, MHCII, and DM but no detectable DO protein (21).

Functional Analysis of DOA Variants

DOA variants were synthesized and cloned into pCDH-EF1-MCS-IRES-mRuby (System Biosciences, Inc.) by GeneScript. Common DOB alleles were synthesized and cloned into pEF1a-MCS-IRES-AcGFP1 (Clonetech Laboratories) also by GeneScript as previously described (21). HeLa.CIITA cells were seeded in 6 well plates and 12-24 hours later were transfected with 1 μg of pCDH-EF1-DOA-MCS-IRES-mRuby and 1 μg of pEF1a-DOB-MCS-IRES-AcGFP1 mixed with Lipofectamine 2000 (ThermoFischer). HeLa.CIITA cells transfected with pCDH-EF1-DOA*0101-MCS-IRES-mRuby and pEF1a-MCS-IRES-AcGFP1 (empty vector) as well as pCDH-EF1-MCS-IRES-mRuby (empty vector) and pEF1a-DOB*0101-MCS-IRES-AcGFP1 were used as controls. Transfected cells were harvested 72 hours later and analyzed by flow cytometry, immunoprecipitation, and/or western blotting.

FACS Analysis

MHCII was measured using an antibody specific for DR, DP, and DQ that was conjugated to PE-Vio770 (clone REA332; Miltenyi Biotech). MHCII-CLIP was measured using the monoclonal antibody CerCLIP.1 (31) and DM and DO levels were measured using the heterodimer specific monoclonal antibodies MaP.DM1 (32) and Mags.DO5 (33), respectively. The MHCII-CLIP, DM and DO specific antibodies were purified from bioreactor supernatants using standard Protein G chromatography and conjugated with Alexa 647 (Mags.DO5) or biotin (CerCLIP.1 and MaP.DM1).

Transiently transfected HeLa.CIITA cells were harvested and split for extracellular staining (MHCII-CLIP and MHCII) or for intracellular staining (DM and DO). For extracellular staining, HeLa.CIITA cells were blocked with normal mouse serum, incubated with antibodies specific for MHCII-CLIP (CerCLIP.1-biotin) and MHCII at 4°C for 30 min and washed. The cells were then incubated with streptavidin-Alexa 647 (Invitrogen) at 4°C for 30 min, washed, and analyzed by flow cytometry. DAPI was added prior to analysis for dead cell exclusion. For intracellular staining for DM and DO, transfected HeLa.CIITA cells were fixed and permeabilized using a Cytofix/Cytoperm kit (BD Biosciences) for 30 min at 4°C. Cells were washed and blocked with normal mouse serum followed by incubation with antibodies specific for DM (MaP.DM1-biotin) and DO (Mags.DO5-Alexa 647) for 30 min at RT. After washing, cells were incubated with streptavidin-PE-Cy7 for 30 min at RT, washed, and analyzed. HeLa.CIITA were transfected individually with pCDH-EF1-DOA*0101-MCS-IRES-mRuby or only pEF1a-DOB*0101-MCS-IRES-AcGFP1 were used as compensation controls for mRuby and AcGFP fluorescence. Data was acquired using a custom 5 laser BD LSRII cytometer and analyzed using FlowJo software (BD Biosciences). Background fluorescence for MHCII-CLIP and DO expression was removed by subtracting the MHCII-CLIP and DO gMFI values of either the EV-mRuby/DOB*0101-AcGFP or the EV-mRuby/EV-AcGFP controls from each sample.

Cell Sorting

HeLa.CIITA cells were transfected as described previously and harvested 72 hours later. mRuby+ AcGFP+ cells were sorted, counted, and frozen for subsequent biochemical analysis.

Western Blotting and Quantification

Transiently transfected HeLa.CIITA cells were lysed for 30 min on ice in 20 mM Tris-HCl, 130 mM NaCl pH 8.0 with 1% Triton X-100 and cOmplete Mini, EDTA-free Protease Inhibitor Cocktail Tablets (Roche Life Science). Nuclear material was removed by centrifugation and supernatants were mixed with Laemmli sample buffer containing 20 mM DTT and incubated at 95°C for 5 min. Protein samples were separated on 10%−20% gradient SDS-PAGE gels (Criterion TGX; Bio-Rad) and transferred to polyvinylidene fluoride membrane (Millipore). For analysis of DO immunoprecipitations, Laemmli sample buffer (without DTT) was added to the washed Protein G bead pellets, the samples were heated at 95°C for 5 min and analyzed by SDS-PAGE as above, prior to transfer to polyvinylidene fluoride membranes for western blot analysis. The membranes were blocked with 5% powdered milk for 30 min at room temperature and the resulting membranes were incubated with an affinity purified polyclonal rabbit antibody specific for the cytoplasmic tail of DOβ (R.DOB/c; (34)) or a rabbit monoclonal antibody specific for the cytoplasmic tail of DMβ (clone EPR7981; Abcam). After washing, the primary Abs were detected by the addition of HRP-conjugated donkey or goat anti-rabbit antibodies (Jackson ImmunoResearch) followed by development with chemiluminescent peroxidase substrate (Pierce Biotechnology) and exposure to film.

To determine the relative amount of DO and DM that were recovered after co-immunoprecipitation, data were normalized to correct for differences in co-transfection efficiencies, using the following two strategies. For some experiments the individual DOβ and DMβ band intensity values obtained after immunoprecipitation of DO were normalized based on the percentage of mRuby and AcGFP double positive cells in each sample (as determined by flow cytometry prior cell lysis). For other experiments, normalization was performed prior to cell lysis. The same number of mRuby and AcGFP double positive cells in each sample as determined by flow cytometry were lysed prior to immunoprecipitation. In this case the resulting uncorrected DOβ and DMβ band intensity values obtained were used for calculations. In both cases, the resulting values for DOβ and DMβ were subsequently normalized to the values obtained after expression of DOA*0101/DOB*0101 in HeLa.CIITA cells to allow for comparison across independent experiments. Bands were quantified using ImageJ software.

Immunoprecipitations

HeLa.CIITA cells transiently transfected with plasmids encoding DOB*0101 and the various DOA variants were lysed as described above. After removal of nuclear material by centrifugation, supernatants were precleared with 40 μL of Protein G-Sepharose (GE Healthcare Life Sciences) and 5 μg of mouse IgG with rotating for 30 min at 4°C. The samples were centrifuged, and the precleared supernatants were added to tubes containing 5 μg of Mags.DO5 and 30 μL of Protein G-Sepharose to immunoprecipitate the DO-DM complex. After rotation at 4°C for 2 hr, the beads were washed 3-times with 20 mM Tris-HCl, 130 mM NaCl pH 8.0 containing 0.1% Triton X-100.

For the DO-depletion studies, lysates from matched numbers of sorted cells were lysed and precleared as above. After preclearing the lysates were split in two and 5 μg of Mags.DO5 (DO precipitate) or 5 μg of mouse IgG (negative control precipitate) and 30 μL of Protein G-Sepharose were added to the two tubes. The samples were rotated for 1 hr, and each DO or control precipitation was repeated with the resulting supernatants two more times (for 3 total sequential precipitates) for 1 hr each. The resulting Protein G bead pellets were washed 3-times with 20 mM Tris-HCl, 130 mM NaCl pH 8.0 containing 0.1% Triton X-100 prior to analysis by SDS-PAGE and western blotting.

Analysis of DO Allotypes and LD Analysis

The association between DOA and DOB allelic variants was analyzed by retrieving all rsSNPs defining the corresponding alleles stated by the HLA-IMGT database and then querying their pairwise combinations using Haploview 4.2. These queries were performed on the corresponding region Chr 6: 3300000-32000000 (GRCh38p12) of the human genome deposited in the 1000 Genomes Project dataset (35). Furthermore, a Python script was written to retrieve the haplotype frequencies and their combinations at the individual level using the pLink and HaploView software packages. The script is available at https://github.com/e-morrison/dofreqs.

The LDpair tool from the LDlink Suite (36) was used to determine linkage disequilibrium between SNPs.

Statistical Analyses

For Figs. 3C, 3D, 5A, 5B, and 5C, the average plus and minus two standard deviations were calculated from the measured levels of DOA*0101 combined with each of the five common DOB alleles (gray zones on bar graphs). In other experiments, significance was determined by performing Student’s unpaired t tests using Prism Graph Pad.

FIGURE 3.

Analysis of DOA and DOB allelic variations and their natural haplotype combinations.

(A) Schematic of the DOB common alleles relative to conical DOB*0101 gene. DOB contains the following domains as indicated: a signal sequence (SS), a MHCII beta-like (beta-like), an immunoglobulin-like (Ig), a transmembrane (TM), and a cytoplasmic tail (Ct). The four other common alleles have single missense mutations that result in single amino acid changes which are shown. Numbering indicates amino acid number starting before (negative) or after signal sequence cleavage and the amino acid encoded by the DOB*0101 gene is shown in the rectangle domains followed by an arrow showing amino acid changes for DOB*0102, DOB*0103, DOB*0104, and DOB*0105 alleles.

(B) Association of DOA and DOB allelic variants into haplotypes and their corresponding frequencies shown as a pie charts for all individuals (ALL) covered by the 1000GP, and the corresponding subpopulations (AFR = African; AMR: American; EAS: East Asians; EUR: European; SAS: South Asian). Enlarged pie pieces show frequencies of DOA*0102. The table on the right side depicts the color code used for each haplotype in the pie charts and the frequency of homozygous (homo) and heterozygous allotypes containing DOA allelic variants different than the canonical DOA*0101 is given for all individuals and each subpopulation. Combinations that were not found in the 1000GP database are left blank. The following SNPs were used to search 1000GP DOA*0102, (rs11575906); DOA*0103 (rs41542323), DOA*0104 (rs41541116), DOB*0102 (rs2071554), DOB*0103 (rs2621330), DOB*0104 (rs2070121), and DOB*0105 (rs11575907). The two SNPs resulting in substitutions for the DOA*0103 and DOA*0104 allelic variants in the IMGT-HLA are not found in the 1000GP dataset and are not shown.

(C) To determine if natural DOA-DOB haplotype combinations described in (B) resulted in DO proteins with altered function, HeLa.CIITA cells were transiently transfected with vectors encoding DOA*0101 and DOB*0101, DOB*0102, DOB*0103, DOB*0104, or DOB*0105 or DOA*0102 and DOB*0101, DOB*0102, DOB*0104, or DOB*0105 and the gMFI values for MHCII, DM, MHCII-CLIP and DO levels in mRuby+ AcGFP+ cells were determined as in Fig. 1. Background gMFI for MHCII-CLIP and DO was eliminated by subtracting the gMFI value of the EV-mRuby/EV-AcGFP control from each sample. Data from 5 independent experiments were normalized to the values obtained for DOA*0101/DOB*0101 combination (black bars). The gray zones correspond to the average of DOA*0101 with each of the five common DOB alleles plus or minus two standard deviations (MHCII (1.00 ± 0.04), DM (0.86 ± 0.16), MHCII-CLIP (1.02 ± 0.05), DO (0.96 ± 0.11)). Blue bars indicate haplotype combinations with reduced DO levels.

(D) DO proteins resulting from DOA*0102 paired with four DOB allotypes have altered function. MHCII-CLIP:DO ratios for each of the natural DOA-DOB combinations were calculated and normalized to that of the DOA*0101/DOB*0101 combination (black bar). The gray zone corresponds to the average of DOA*0101 with each of the five common DOB alleles plus or minus two standard deviations (1.09 ± 0.19). Blue bars indicate haplotype combinations with function.

FIGURE 5.

Functional Analysis of naturally occurring DOα variants.

(A and B) HeLa.CIITA cells were transiently transfected with a vector encoding DOB*0101 and vectors encoding each of the individual 21 DOA variants. Cells were analyzed by flow cytometry 72 hours later and the gMFI values for MHCII and MHCII-CLIP (A) and DM and DO (B) levels in mRuby+ AcGFP+ cells were determined as in Fig. 1. Background gMFI for MHCII-CLIP and DO was eliminated by subtracting the gMFI value of the EV-mRuby/DOB*0101-AcGFP control from each sample. Data from 4-5 independent experiments were normalized to the values obtained for DOA*0101/DOB*0101 combination (black bars). The gray zones correspond to the average of DOA*0101 with each of the five common DOB alleles plus or minus two standard deviations: MHCII (1.03 ± 0.06), DM (0.97 ± 0.09), MHCII-CLIP (0.90 ± 0.15), DO (0.95 ± 0.11).

(C) MHCII-CLIP:DO ratio for each of the DOA variants combined with DOB*0101. The ratios were normalized to that of the DOA*0101/DOB*0101 combination (black bar) form 4-5 independent experiments. Constructs transfected are indicated below the bar graphs. The gray zone corresponds to the average plus or minus two standard deviations of the values obtained from DOA*0101 combined with each of the five common DOB alleles (1.10 ± 0.19). DOα variants with altered function (increased-blue bars, decreased-red bars) were defined as possessing a MHCII-CLIP:DO ratio that fell outside this range.

Results

Assay Measuring DO Function

DM induces the dissociation of CLIP from the MHCII peptide binding groove (8, 31). DO has been shown to be a substrate mimic of MHCII thereby preventing DM from functionally interacting with MHCII-CLIP (2). DO expression in cells resulted in increased MHCII-CLIP levels (34). Thus, the ability of DO to functionally interact with and inhibit DM activity can be monitored by measuring the level of cell surface MHCII-CLIP (21, 33, 34, 37). Therefore, we took advantage of the direct relationship between DO expression levels and MHCII-CLIP levels to determine if amino acid changes encoded by DOA gene variants resulted in DO proteins with altered function. For these studies, we established and used an in vitro assay to monitor surface MHCII-CLIP levels after expression of DOA gene variants co-expressed with the most commonly expressed DOB allele, DOB*0101. For this assay, DOA and DOB expression constructs were co-transfected into DO-negative HeLa cells that had been previously stably transfected with the class II major histocompatibility complex transactivator (HeLa.CIITA) (30). Previous studies have shown that CIITA drives the expression of MHCII, DM, Ii, and DOA, but not DOB in B cell lines (38). However, others have shown that expression of CIITA in B cell lines and HeLa can also turn on DOB expression (30, 39). In our hands, HeLa.CIITA cells expressed MHCII, DM, Ii but only (very) low levels of DOA mRNA and no DOB mRNA providing a recipient cell in which we could express and functionally test DOA and DOB gene variants (21). To allow for the identification of cells that had been co-transfected with both genes, the DOA gene was followed by an IRES sequence and the mRuby gene and the DOB gene was followed by an IRES sequence and the AcGFP gene. Seventy-two hours after transfection, one half of the cells were surface stained with antibodies specific for total MHCII and MHCII-CLIP (31) and the other half were fixed and permeabilized to allow for intracellular staining for DO and DM using heterodimer-specific Abs (32, 33) and the levels of these proteins were measured by flow cytometry. Only mRuby and AcGFP double expressing cells that marked the co-expression of DOα and DOβ were analyzed (Fig. 1). The mRuby⁺ AcGFP⁺ cells were gated for MHCII high cells before measuring MHCII-CLIP levels (Fig. 1A) and for the measurement of DO cells were first gated for DM (Fig. 1B) to ensure that any HeLa cells that had lost CIITA were excluded from the analyses. HeLa.CIITA cells transfected with vectors encoding only mRuby and AcGFP (i.e. empty vectors), with the vector encoding only DOA (DOA*0101-mRuby) or the vector encoding only DOB (DOB*0101-AcGFP) resulted in no detectable DO protein and low levels of MHCII-CLIP (Fig. 1). However, when HeLa.CIITA cells were transfected with vectors encoding both the DOA and DOB genes, DO protein was produced and resulted in increased MHCII-CLIP levels (Fig. 1). DM and MHCII levels were similar in all transfection conditions (Fig. 1). These control experiments showed that DM function was inhibited and MHCII-CLIP levels increased only after expression of both chains of the DO αβ heterodimer in HeLa.CIITA cells. These experiments validated the use of the in vitro transient transfection system to measure if natural DOA gene variation in humans results in DO proteins with altered function.

FIGURE 1.

Approach used to measure DO function.

(A) HeLa.CIITA cells (MHCII⁺, Ii⁺ and DM⁺) were transiently transfected with an empty mRuby vector (EV-mRuby) and empty AcGFP vector (EV-AcGFP), a vector expressing DOA*0101 mRuby and EV-AcGFP, EV-mRuby and a vector expressing DOB*0101-AcGFP, or DOA*0101-mRuby and DOB*0101-AcGFP. Cells were analyzed 72 hours later by flow cytometry after staining for cell surface MHCII and MHCII-CLIP expression. Live cells were gated for the expression of both mRuby which reports DOA expression and AcGFP which reports DOB expression as shown. MHCII levels in the mRuby+ AcGFP+ cells are shown (left) and after gating for the MHCII⁺ cells, MHCII-CLIP levels are shown (right). Geometric mean fluorescent intensity (gMFI) values for MHCII and MHCII-CLIP are shown on each histogram.

(B) HeLa.CIITA cells transfected as in (A) were fixed, permeabilized and stained for intracellular DM and DO prior to analysis by flow cytometry. Fixed cells were gated on AcGFP⁺ mRuby⁺ cells (fixation of cells results in decreased mRuby and AcGFP fluorescence) and the DM levels determined (left). After gating for DM⁺ cells, the level of DO was determined (right). gMFI values for DM and DO are shown on each histogram. DO expression and increased MHCII-CLIP levels were observed only after transfection of vectors expressing both DOA and DOB. Data is representative of five similar experiments.

Three of four DOA Common Alleles have altered Function.

DOA, like DOB and the DMA and DMB genes, is considered to be a non-polymorphic gene (40). Currently, there are four common DOA alleles, DOA*0101, *0102, *0103, and *0104 according to the Immuno Polymorphism Database- International ImMunoGeneTics Project (PDL-IMGT) for HLA (41). DOA*0101 is the most common allele and is present in ~98% of the population. DOA*0102 (R80C), DOA*0103 (L74V) and DOA*0104 (P11 del FS) account for ~1.7% of the rest of the DOA alleles found in humans (Fig 2A). The frameshift mutation in DOA*0104 is inserted at amino acid residue 11 and results in the addition of 25 amino acids before ending in an early stop codon. Therefore DOA*0104 is predicted to produce a frameshift protein that terminates after only 36 amino acids (Fig. 2A and 4) resulting in a null protein.

FIGURE 2.

Three of the four common DOA Alleles have altered function.

(A) Schematic of the DOA common alleles relative to the DOA*0101 gene. DOA contains the following domains as indicated: a signal sequence (SS), a MHCII alpha-like (alpha-like), an immunoglobulin-like (Ig), a transmembrane (TM), and a cytoplasmic tail (Ct). The other three common alleles have single missense mutations that result in single amino acid changes which are shown. Numbering indicates amino acid number starting after signal sequence cleavage and the amino acid encoded by the DOA*0101 gene is shown in the rectangle domains followed by an arrow showing amino acid changes for DOA*0102, DOA*0103, and DOA*0104 alleles. For DOA*0104 a frameshift mutation results in the addition of 25 random amino acids prior to a stop codon (see Fig. 4).

(B) HeLa.CIITA cells were transiently transfected with a vector encoding DOB*0101-AcGFP (the most common DOB allele) together with a control empty vector (EV-mRuby), or vectors encoding DOA*0101, DOA*0102, DOA*0103, or DOA*0104. Seventy-two hours later, the transiently transfected cells were stained as in Figure 1 prior to analysis by flow cytometry. Histogram overlays on left show representative data for MHCII and DM and on right for MHCII-CLIP and DO after gating as indicated for MHCII⁺ and DM⁺, respectively, in cells expressing both the DOA and DOB expression constructs (mRuby⁺ AcGFP⁺). Shaded histograms are levels obtained for unstained cells as a negative control. Colors used in histogram overlays for each DOA common allele and empty vector control (EV-mRuby) are shown on right. gMFI values for MHCII, DM, MHCII-CLIP, and DO are shown on each histogram for each corresponding combination, color-coded as on right. Data are representative of one of four to five independent experiments.

(C) Quantification of independent experiments shown in (B). Background gMFI for MHCII-CLIP and DO was eliminated by subtracting the gMFI value of the EV-mRuby/DOB*0101-AcGFP control from each sample. gMFI values measured for MHCII, DM, MHCII-CLIP, and DO were normalized to the value obtained after transfection of DOA*0101 (with DOB*0101) to allow for analysis across independent experiments (black bars).

(D) The relative function of each DOA allele was determined by calculating the ratio of MHCII-CLIP to DO from the values obtained in (C) to correct for different DO protein levels relative to the value obtained for DOA*0101 (with DOB*0101). Blue bars indicate alleles that have increased function relative to DOA*0101 (with DOB*0101; black bar). Data were combined from 4-5 individual experiments. Symbols and error bars represent the mean ± SD. Significance was calculated using an unpaired, two-tailed Student’s t test. *, p=0.0147; ****, p<0.0001.

FIGURE 4.

Amino acid alignment of DOα common alleles and variants analyzed.

(A) Alignment of the four common DOα alleles to DOA*0101. Dashes indicate similarity and the dots indicate a truncated protein (for DOA*0104). Missense and the stop mutation for the 21 DOA variants selected for the screen are indicated below the amino acid sequences for the common variants. DOα alleles that form DO proteins with decreased function relative to DOA*0101/DOB*0101 are colored red, those with increased function in blue and those in black have normal function (see Fig. 5). DOA protein domains are marked by colored frames as indicated.

(B) The structure of the DM-DO complex (2) with the side-chains of amino acids mutated in the DOA variants investigated in this study shown as spheres. The backbone cartoon of DMA is shown in light green and of DMB in dark green. DOA, carrying the mutated residues, is depicted in dark orange and DOB in light orange. As in (A), residues with enhanced function are colored blue, those leading to reduced function in red and mutations without an effect on functionality are shown in grey. Mutations occurring in the transmembrane region are show as the crystal structure did not contain this domain. The Figure was created by PyMol (56) utilizing the PDB file 4I0P.

To determine if DOA*0102, *0103 or *0104 produced DO proteins with altered function compared to the most common DOA allele, DOA*0101, HeLa.CIITA cells were transfected with each of the individual four common DOA alleles and DOB*0101. Note that unless otherwise specifically stated, the individual DOA alleles and variants analyzed in this paper were co-expressed in HeLa.CIITA cells with DOB*0101, the most frequently expressed DOB allele. Seventy-two hours post transfection, the cells were stained and analyzed by flow cytometry as described above. Representative flow cytometry plots obtained for MHCII, MHCII-CLIP, DM and DO are shown in Figure 2B. To allow for quantification of the results obtained from multiple independent experiments, the levels obtained from at least four independent experiments were normalized to the levels obtained after expression of DOA*0101 (Fig. 2C). The expression of the four DOA common alleles had no impact on the levels of MHCII and DM levels were only minimally impacted. MHCII-CLIP levels were uniformly high for DOA*0101, *0102 and *0103, despite DOA*0102 and *0103 expressing only half the amount of DO protein when compared to DOA*0101 (Fig. 2B and 2C). The DO antibody used to measure DO levels (Mags.DO5) recognizes a conformational epitope on the DO heterodimer, however the epitope is on the DOβ chain (28, 33). Thus, mutations in DOα are unlikely to impact Mags.DO5 recognition of DO proteins produced after transfection of DOB*0101 with DOA gene variants. As expected, DOA*0104 expressed no detectable DO protein due to the frameshift mutation and, thus is a null allele. As expected, expression of the DOA*0104 null allele resulted in no increase in MHCII-CLIP (Fig. 2B and 2C).

MHCII-CLIP levels directly correlate with DO protein expression levels (21, 33, 37, 42). Therefore, to determine if the three DOA alleles that produced DO protein (DOA*0101, *0102 and *0103) had differential function, the ratio of MHCII-CLIP to DO was determined. Compared to DOA*0101, expression of DOA*0102 and *0103 resulted in increased MHCII-CLIP:DO ratios showing that these two DO proteins had enhanced function (Fig. 2D). In our previous work, we identified two rare DOB gene variants with increased function that had amino acid mutations that mapped to the MHCII beta-like, membrane distal domain (21). Like the two DOβ variants, the polymorphisms in DOA*0102 and *0103 similarly mapped to the MHCII alpha-like membrane distal domain (Fig. 2A). Collectively these analyses show that the DOA*0102, *0103 and *0104 DOA common alleles either produce no DO protein or DO proteins with altered function when compared to DOA*0101.

Functional Analysis of Common DOA-DOB Combinations

The HLA genetic locus has strong linkage disequilibrium and, as such, specific DOA and DOB alleles are associated as pairs within individual haplotypes. This led us to ask if naturally occurring DO allotypes may have altered function since our studies above demonstrated functional differences in the DOA common alleles. Using rsSNPs (43) present in the individual DOA and the five DOB common alleles (Fig. 3A) (40, 41), the 1000 Genomes Project (1000GP) (35) publicly available sequence data was parsed to determine the naturally occurring DO haplotypes. We have previously used a similar approach to define the naturally occurring DM haplotypes (44). DOA*0103 and DOA*0104 were present in less than 1% of the population and thus, there was not sufficient data to determine their contributions to the naturally occurring haplotypes. DOA*0101 was found in combination with all five common DOB alleles whereas DOA*0102 was found with all but DOB*0103 (Fig. 3B). Frequencies for each combination were compiled from the 1000GP database for all individuals (ALL) and for individuals in each of the following subpopulations: African (AFR), American (AMR), East Asian (EAS), European (EUR), and South Asian (SAS) (Fig. 3B). In line with the commonly accepted idea that DO is non-polymorphic, 65.6% of ALL individuals were either homozygous or heterozygous for DOA*0101/DOB*0101 and 32.8% of individuals had DOA*0101 paired with one of the other DOB common alleles. Thus, DOA*0101 paired with any of the common DOB alleles accounted for 98.4% of all DO haplotypes. DOA*0102 was found in combination with all DOB common alleles except for DOB*0103, with DOA*0102/DOB*0101 found in 1.08% of the 1.72% of ALL individuals that have DOA*0102 in combination with one of the DOB alleles. While DOA*0101 was found in all subpopulations, DOA*0102 was found in all but the AFR population.

We next determined if the naturally occurring DOA*0102/DOB common allele natural haplotypic combinations resulted in DO proteins with altered function by transfecting vectors expressing DOA*0102 and the four individual DOB common alleles into HeLa.CIITA cells. Seventy-two hours later MHCII protein levels were determined by flow cytometry. As a control, we also co-expressed DOA*0101 with the five common DOB alleles since we previously reported that the five DOB common alleles all functioned similarly (21). Figure 3C show a summary of the expression levels obtained for MHCII, MHCII-CLIP, DO and DM combined from at least four independent experiments after transfection of the various combinations of DOA and DOB alleles into HeLa.CIITA cells. To facilitate comparison across independent experiments, results were normalized the DOA*0101 and DOB*0101 control for each experiment. As we previously reported, all five DOB common alleles expressed similar levels of MHCII pathway proteins when co-transfected with DOA*0101 (21). When DOA*0102 was paired with the four DOB common alleles to generate the naturally occurring haplotypic pairs, MHCII, MHCII-CLIP and DM levels were also similar to the levels obtained for the DOA*0101/DOB*0101 pair, however, for all naturally occurring DOA*0102 haplotypic pairs, DO protein levels were about half that observed for the DOA*0101/DOB*0101 pair. Calculation of the relative functional activity of DO by determination of the MHCII-CLIP:DO ratio showed that, when compared to DOA*0101 paired with any of the DOB common alleles, all four naturally occurring DOA*0102/DOB haplotypic pairs produced DO proteins with increased function (Fig. 3D). Interestingly, specific subpopulations of individuals had unique expression patterns for DOA*0102 paired with the different DOB alleles (Fig. 3B). DOA*0102/DOB*0101 and DOA*0102/DOB*0102 heterozygotes were found in all the subpopulations except AFR, however, DOA*0102/DOB*0101 homozygotes were found only in SAS. Similarly, DOA*0102/DOB*0104 and DOA*0102/DOB*0105 heterozygotes were found only in EUR and AMR, respectively.

Functional Analysis of naturally occurring DOA Variants

In our previous work we examined over 60,000 exomes from the ExAC database (45) for DOA allelic variants. These analyses identified 107 different allelic variants, including the DOA*0102, *0103 and *0104 common alleles. The variants had frequencies ranging from 8.24 × 10⁻⁶ to 0.012 (21). Of the 107 DOA variants, 12 were predicted to be null due to frameshift mutations, mutations that generated premature stop codons or mutations that disrupted the start codon. The other 95 variants had single missense mutations. Twenty of the more abundant 95 DOA variants with single missense mutations and one allele encoding a nonsense mutation were chosen for functional analyses (Supplemental Table 1). The missense variants chosen for analyses had single amino acid substitutions that were distributed throughout the protein (Fig. 4). As above, function was determined by co-transfection of the individual DOA gene variants with DOB*0101 (most common DOβ allele) into HeLa.CIITA cells and three days later MHCII, MHCII-CLIP, DM and DO levels were determined by flow cytometry. The results of these analyses are presented as a compilation of the levels obtained for these proteins from four or more independent experiments normalized to the levels obtained after co-expression of DOA*0101/DOB*0101 in HeLa.CIITA (Fig. 5). Since expression of DOA*0101 with each of the five DOB common alleles did not significantly alter MHCII, MHCII-CLIP, DO or DM expression levels in transfected cells (Fig. 3C; (21)), we considered any DOA variant with MHCII pathway protein levels more than two standard deviations above or below the average obtained for DOA*0101 co-expressed with the 5 common DOB alleles as an allele with altered protein levels. Many of the 21 DOA variants when expressed individually with DOB*0101 in HeLa.CIITA cells produced DO proteins that had altered MHCII-CLIP and DO levels (Fig. 5B) while DM and MHCII levels remained unchanged (Fig. 5A). As predicted, HeLa.CIITA cells expressing the Q58stop nonsense DOα variant had no detectable DO protein and as a result, no increase in MHCII-CLIP was observed (Figs 5B). Surprisingly, HeLa.CIITA cells expressing the DOα F114L missense mutation also did not produce any detectable DO protein and consequently also did not show increased MHCII-CLIP (Fig. 5B). HeLa.CIITA cells expressing seven of the 21 DOα alleles (V92M, R97W, P103T, R124H, G126R, R147C, and H150R) had reduced DO protein levels but yet all but one (R147C) had normal MHCII-CLIP levels (Fig. 5B). The R147C variant had a slight but significant reduction in MHCII-CLIP. The phenotype for these seven variants was similar to what was observed for HeLa.CIITA expressing DOA*0102 when paired with four DOB common alleles (Fig. 3) and previously for the DOβ R70G and G77V variants when paired with DOB*0101 (21). Finally, HeLa.CIITA cells expressing the DOα I67M allele exhibited significantly higher DO expression but also had normal MHCII-CLIP.

The function of the DO proteins produced after transfection of the individual DOα variants with DOB*0101 into HeLa.CIITA cells was quantified by calculating the ratio of the protein levels for MHCII-CLIP to DO levels (Fig. 5C). Fifty-two percent of the 21 DOα variants had MHCII-CLIP:DO ratios that fell within 2 standard deviations of ratios obtained after expression of the of DOA*0101 with 5 common DOB alleles and, thus, were considered variants with normal function. Of the remaining alleles, two DOα alleles (Q58Stop and F114L) were null alleles as they produced no detectable DO protein after expression in HeLa CIITA cells, and thus, a MHCII-CLIP:DO ratio could not be calculated. After expression in HeLa.CIITA cells, the DOα I67M allele resulted in significantly higher DO protein levels that resulted in a MHCII-CLIP:DO ratio that was about half of that obtained for the controls. Finally, the remaining 7 DOα variants (V92M, R97W, P103T, R124H, G126R, R147C, and H150R) had MHCII-CLIP:DO ratios that were 1.5 to 2-fold higher than controls showing that these variants produced DO proteins with increased activity.

Biochemical Analysis of Alleles with Altered Function

Our studies identified 10 DOα variants that produced DO proteins with either reduced or enhanced ability to modify the peptide loading of MHCII molecules by DM. DO is a MHCII substrate mimic and binds to DM, preventing DM from interacting with MHCII and mediating peptide loading. Thus, we hypothesized that the DO proteins with altered function might interact either more (increased function) or less (decreased function) efficiently with DM. To test this idea, the DO heterodimer and any co-associated DM molecules were captured from HeLa.CIITA cell lysates expressing the individual 10 DOα variants with altered activity followed by western blotting for any co-associated DM (Fig. 6A). As a positive control, HeLa.CIITA cells expressing DOA*0101/DOB*0101 were used and mock transfected cells were used as a negative control. Western blotting of total cell lysates for DOβ and DMβ confirmed the successful expression of DO for each variant, as well as similar DM expression levels. Assembled DO αβ heterodimers were immunoprecipitated from all cells that expressed the individual DOα variants except for the DOA*0104 and DOα F114L variants (Fig. 5B) confirming the flow cytometric results showing that these two null alleles that did not produce a functional heterodimer (Fig. 2B, 2C). All the remaining DOα variants that resulted in an assembled DO αβ heterodimer were found complexed with DM (Fig. 6A).

FIGURE 6.

Biochemical analysis of DO proteins produced by the DOα variant alleles.

(A) HeLa.CIITA cells were transiently transfected with vectors encoding DOB*0101 and DOA*0101 or one of the DOα variants with altered function (DOA*0102, DOA*0103, DOA*0104, I67M, V92M, R97W, P103T, F114L, R124H, G126R, R147C, and H150R; labeled across the top of the blots). Mock transfected HeLa.CIITA cells (Mock) were used as a negative control. Cells were harvested 72 hours later, lysed, and DO was immunoprecipitated with the DO heterodimer specific antibody Mags.DO5 (33). The resulting precipitated proteins (α-DO-IP) were separated by SDS-PAGE, transferred to membranes and western blotted to determine DOβ (R.DOB/c (34) or DMα protein (clone EPR7981) levels. Western blot analyses of the lysates (Total) used for the immunoprecipitations are included to confirm protein expression (bottom).

(B and C) Quantification of the amount of DOβ (B) and co-associated DMβ (C) protein detected in each α-DO immunoprecipitation. Values were either normalized to correct for transfection efficiency based on the number of mRuby+ AcGFP+ cells in each sample. Values were then normalized to the amount of DOβ or DMβ protein obtained after transfection of DOA*0101/DOB*0101 (black bar) to allow for comparison across four independent experiments.

(D) Ratio of the amount DOβ to DMβ obtained for each α-DO immunoprecipitation combined for the 4 independent experiments. Values from (B and C) were normalized to the ratio obtained after transfection of DOA*0101/DOB*0101 (black bar) to allow for comparison across the four independent experiments.

(E-G) HeLa.CIITA cells were transiently transfected with vectors encoding DOB*0101 and DOA*0101, one of a few select DOα variants with altered function (DOA*0102, DOA*0103, DOA*0104, and H150R) or as negative controls EV-mRuby only or no vectors (mock). Cells were harvested 72 hours later, and mRuby+ AcGFP+ cells were purified by cell sorting, except for the mock transfection. (E) Western blotting of lysates verified DO and DM expression in the cells. (F) Equal numbers of cells were lysed, and the resulting lysates were split into two equal aliquots. DO was depleted by three sequential anti-DO immunoprecipitations with Mags.DO5 (33) and from one aliquot and as a negative control, three sequential immunoprecipitations were performed using mouse IgG. The resulting precipitated proteins were separated by SDS-PAGE, transferred to membranes and blotted to confirm DO depletion by blotting for DOβ (R.DOB/c (34)). (G) Control (C) and DO-depleted (IP) supernatants from the third immunoprecipitation were probed for DM (DMβ; clone EPR7981) to determine if any DM remained after DO removal.

To determine the relative amount of DM co-precipitated with the DO heterodimer from cells expressing the individual DOα variants, the amount of DO (DOβ) and DM (DMβ) in the immunoprecipitations was quantitated across five independent experiments and compared to the amount of protein immunoprecipitated after expression of DOA*0101/DOB*0101. The resulting values were also normalized to correct for differences in transfection efficiencies using the expression of the reporter molecules, mRuby and AcGFP (see Materials and Methods for details). Relative to the amount of DO immunoprecipitated for the control, reduced levels of DO were recovered for DOA*0102, *0103 and DOα V92M, R97W, P103T, R124H, G126R, R147C, and H150R, whereas DOα I67M had increased DO levels (Fig. 6B). These results correlated well with the DO levels measured by flow cytometry (Fig. 2B, 2C and 5B). Unexpectedly, the amount of DM associated with DO was reduced for all DOα variants with altered function, including those with enhanced function and DOα I167M that had increased DO levels (Fig. 6C).

We had expected to find more DM associated with DO in the DOα variants with enhanced function. Therefore, we next estimated how much DM was recovered bound to DO in each immunoprecipitation relative to the amount of recovered DO. With the exception of the DOα I67M and R147C variants, the remaining DOα variants had DO:DM ratios that were similar to the control (DOA*0101/DOB*0101) (Fig. 6D). This indicated that, although the missense mutations in the DOα variants resulted in less overall assembled DO, the resulting DO heterodimer interacted normally with DM, at least as measured using co-immunoprecipitation. Counterintuitively, this also indicates that there should be more free DM (DM not bound to DO) in HeLa.CIITA cells expressing these DOα variants resulting in lower MHCII-CLIP levels. However, these DOα variants had MHCII-CLIP levels that were similar to controls (Fig. 5B). The DOα R147C variant had a reduced DO:DM ratio, which indicates that more DM coprecipitated with this variant relative to the amount of DO in the complex. Surprisingly, the DOα I67M variant which had increased DO levels, had reduced DM levels. DO must assemble and associate with DM to exit the ER (46), thus, reduced DM associated with the DOα I67M variant suggests that it may traffic from the ER somewhat independent of DM, as has been reported for the previously described DOα P11V DO variant (28).

Finding less DO associated with DM after the expression of the DOα variants with altered function in HeLa.CIITA cells suggested that these variants should have increased free DM. If so, these variants should show lower MHCII-CLIP, yet we observed MHCII-CLIP levels that were similar to the control cells. Studies in EBV-transformed human B cell lines and primary human B cells showed that ~50% of DM is DO associated (47, 48). Thus, a possible explanation for our observed result is that expression of DO in HeLa.CIITA cells by transient transfection resulted in overexpression of DO that quantitatively bound up all DM in the cells, resulting in high MHCII-CLIP levels. To test this idea, we expressed a selection of DOα variants (DOA*0101, DOA*0102, DOA*0103, DOA*0104, and H150R) in HeLa.CIITA cells and sort purified mRuby and AcGFP double expressing, DO-positive. As a control, we also purified mRuby and AcGFP double expressing HeLa.CIITA cells that had been transfected with the empty mRuby vector (no DOA*0101) with DOB*0101 (i.e. AcGFP+) and also included (unsorted) cells that had been mock transfected without any vectors. Equal numbers of cells were lysed and divided into two equal aliquots. One aliquot was depleted of DO by three sequential immunoprecipitations with a DO-heterodimer specific mAb and the other aliquot mock depleted using a non-specific Ab. We then determined if any DM remained in the DO-depleted lysates by western blotting for DM. Western blotting of the lysates prior to depletion verified DO and DM expression (Fig. 6E) and western blotting of the three sequential anti-DO immunoprecipitated for DO demonstrated successful depletion of DO from the lysates (Fig. 6F). Finally, western blotting of the DO- and control-depleted lysates showed DM remaining in the lysates, even in the DO-depleted samples (Fig. 6G). Indeed, DO depletion only removed a small proportion of the DM from the lysates (compare the amount of DM in the control compared to the DO immunoprecipitation) indicating that only a small proportion of the total DM is complexed with DO. These studies indicate that a large pool of free DM (i.e. not bound to DO) is present after transient transfection of DO into HeLa.CIITA cells. It is unclear how DOα variants with altered function inhibit DM resulting in high levels of MHCII-CLIP, however the explanation is not due to a lack of free DM in the cells.

DOA variants linked to human Hepatitis B viral infection

Evidence is accumulating that DO function has implications for the immune responses against viruses (21, 49). In mice, a neutralizing antibody response to mouse retroviruses is generated in a strain of mice with a loss of function allele of H2-O and also in H2-O-deficient mice (21). Conversely, a gain of function allele of DOβ (G77V) in humans was genetically linked to individuals with persistent HCV infection (21). Seven of the DOα variants discovered in the screen described above exhibited a similar enhanced function phenotype to DOβ G77V. Therefore, we searched for possible linkages of these DOA alleles to the outcomes of viral infections in humans.

HLA-DP is one of the polymorphic classical MHCII molecules that have been previously associated with the positive and negative outcomes of viral infections (42). In individuals from some Asian populations the ‘C’ and ‘T’ SNP rs3077 in DPA1, the gene that encodes for the α chain of HLA-DP, had been previously associated with HBV persistence and Ab mediated HBV clearance, respectively (50). Therefore, we asked if any of the DOA alleles that resulted in DO proteins with altered function were in linkage disequilibrium with the specific alleles of DPA1. Of the DOA alleles identified with altered function, DOA*0102, *0103, and *0104, and DOα F114L, and H150R were associated with known coding region SNPs (Supplemental Table 1). Using the LDpair platform tool (36), we found that the rs11575906 SNP in the DOA*0102 allele was in linkage disequilibrium with the ‘C’ allele of the rs3077 SNP in DPA1 (Table 1). Conversely, the rs34987694 SNP in DOα F114L was in linkage disequilibrium with the ‘T’ allele of DPA1 (Table 1). The finding that the DOA and DPA1 rsSNPs were in linkage disequilibrium indicates that the specific alleles of DOA are as likely as the specific alleles of DPA1 to impact the immune response to HBV. Importantly, our data, therefore, supports that in the case of HBV, the more functional DOA*0102 allele may be detrimental to viral immunity, perhaps by impeding the generation of a neutralizing antibody response that contributes to the development of viral persistence. Conversely, the non-functional allele DOα F114L may be beneficial by promoting a neutralizing antibody response and viral clearance. However, in the case of the null allele, individuals may need to be homozygous for a beneficial response. Collectively, these data support that naturally occurring variation in DOA contributes to the control of human viral immune responses.

Table 1:

DOA variants linked to certain MHCII genes associated with specific outcomes of HBV infection

Variant Phenotype (SNP)	Linked to: Gene (SNP)	Linkage Disequilibrium D’	Associated Viral Phenotype
DOA*0102 Gain of Function (rs11575906 )	HLA-DPA1 (rs3077 - ‘C’ allele)	1 (p<0.0001)	HBV Persistence
DOα F114L Loss of Function (rs34987694 )	HLA-DPA1 (rs3077 - ‘T’ allele)	1 (P<0.0001)	Ab-mediated HBV clearance

Discussion

Peptide loading of MHCII molecules is mediated by DM and modulated by DO. We asked if naturally occurring variation in the DOα subunit of DO impacts function and found several DOA missense gene variants with altered function. In comparison to DO formed by the pairing of DOA*0101 and DOB*0101, the two most common alleles of each subunit, two of the other DOA common alleles, DOA*0102 (R80C) and DOA*0103 (L74V) were found to be more functional. The only other common allele, DOA*0104 (P11 del FS), was shown to be a functionally null allele as no protein was produced. Naturally occurring haplotypic combinations of the DOA and DOB common alleles were identified from the 1000GP database and analyzed for functional differences. This analysis showed that DOA*0102 paired with any of the DOB common alleles, resulted in DO proteins with increased function. An analysis of 21 rare DOA gene variants with single missense mutations located throughout the DOα coding region showed that 10 variants (48%) produced DO proteins that had altered function. Seven variants had enhanced function, 1 had reduced function and 2 did not produce detectable protein. Counterintuitively, all DOA variants with altered function that produced an assembled DO protein interacted normally with DM and robustly inhibited MHCII peptide loading despite low DO protein levels for most of the variants. Finally, SNPs in two of the DOA gene variants were found to be in linkage disequilibrium with DPA1, a classical MHCII allele that was previous associated with the immune response to HBV. These data support the notion that natural variation in the DOA gene may influence the outcomes of chronic viral infections.

Of the four known DOA common alleles, 98% of all individuals have at least one DOA*0101 allele (35). The other three DOA common alleles, are actually quite uncommon and only DOA*0102 was found in greater than 1% of any population (35). DOA*0103 and DOA*0104 should perhaps be considered rare variants and not common alleles. Indeed, the DOα F114L null variant is found at a higher frequency (0.45%) in the 1000GP database than either DOA*0103 or DOA*0104 (35). In contrast, although DOB*0101 is the predominant allele and is found in approximately 74% of individuals, the other four common alleles accounted for the remaining 25% of total DOB allelic variation. Thus, the DOA gene is less polymorphic than DOB which may indicate that amino acid substitutions in DOβ are better tolerated than changes in DOα. This idea is supported by our previous study showing that all five DOB common alleles produced DO proteins that functioned similarly when paired with DOA*0101 (21). However, when the DOB common alleles were paired with DOA*0102 in combinations found as naturally occurring haplotypes, all DOB-DOA*0102 combinations had enhanced function. Thus, changes in DOα altered function whereas changes in the common DOβ alleles did not. It is not obvious why changes in DOα might be less well tolerated but it is possible that structural elements in DOα are more prone to local perturbations that lead to altered αβ heterodimer formation or disruption of DM-DO interactions.

In addition to examining the functional consequences of polymorphisms in the common DOA alleles, we also analyzed DO proteins resulting from the paring of 21 other naturally occurring DOA variants (when paired with DOB*0101). The DOα variants chosen for analyses were alleles that were present at higher frequencies from the list of 107 DOA variants we had previously identified (21). The 21 variants had single amino acid changes that were located throughout the DOA coding region and in each of the individual domains of DOα, except for the cytoplasmic tail (Fig. 4). Variants that had missense mutations in DOα signal sequence and transmembrane domain were well tolerated and did not impact DO activity. Mutations that impacted DO function were located throughout the MHCII alpha-like and Ig domains, indicating that these domains both contributed to DO activity. The MHCII alpha-like domain preferentially associates with the MHCII beta-like domain of DOβ, indicating that mutations in this region are likely important for DO heterodimer. Indeed, substitution of a methionine for isoleucine at DOα 67 (DOα I67M) located in the helix (Fig. 4B) resulted in increased DO protein. This suggests that a mutation in the MHCII alpha-like domain produced a DO heterodimer that was more stable and may traffic from the ER partially independent of DM, similar to the previously described DOα P11V variant (28). The Ig domain of DOα associates closely with the Ig domain of DOβ, as well as that of DMα ((2) and Fig 4B). Seven of the ten variants analyzed with amino acid substitutions in the Ig domain had altered function and one produced a null protein (DOα F114L). The mutations were located throughout the Ig domain and did not cluster in one particular region of the domain (Fig. 4B). Many of the contacts made with DMα are located in this contact region, defined as Interface II, which makes up about one third of the buried surface area of the DM-DO complex (2) suggestion that changes in the DOα domain may disrupt DM-DO interactions. However, the DO proteins produced by these variants maintained interaction with DM and thus, as many of the mutations identified in this region resulted in reduced DO expression, suggesting they destabilized the DO heterodimer.

DO is a MHCII substrate mimic that binds to DM thereby preventing DM from interacting with MHCII, facilitating CLIP dissociation and catalyzing peptide loading (2, 34). As a consequence, CLIP remains bound to MHCII resulting in higher levels of MHCII-CLIP in DO expressing cells (34). Moreover, DO protein levels directly correlate with MHCII-CLIP levels (21, 33, 37, 42). The DM-DO crystal structure shows a 1:1 stoichiometry for the DM-DO complex. Therefore, it was unexpected that the DOα variants identified in this screen as having altered function expressed approximately half the amount of DO protein as compared to the control (DOA*0101/DOB*0101) yet had the similar levels of MHCII-CLIP. Approximately 50% of DM is free (not bound to DO) in primary APCs (47, 48). Thus, it was possible that expression of DOα and DOβ in HeLa.CIITA by transient transfection resulted in over-expression of DO such that no free DM remained in the cells. However, several lines of evidence support that this was not the case. First, if all DO was complexed with DM (i.e. DM-DO), then the same amount of DM should have been recovered for all DOα variants. However, biochemical analysis showed that less DM was found bound to DO for the DOα variants that had produced less DO protein (Fig. 6C). Secondly, we previously showed that MHCII-CLIP levels increased in HeLa.CIITA cells that expressed the canonical DOα and DOβ alleles as DO protein levels increased (21), as would be predicted from published literature (33, 34, 37). Our previous study also identified multiple DOB allelic variants that behaved as would be predicted. Several of these DOB variants expressed in HeLa.CIITA made reduced DO protein which resulted in correspondingly lower MHCII-CLIP levels (21). Finally, the depletion of DO from DO expressing HeLa.CITTA cell lysates showed that most of the DM remained free of DO in these cells (Fig. 6G). What remains unclear, however is how the DOα gain-of-function variants are able to inhibit DM to the same extent as DOA*0101, despite reduced DO protein expression. Nonetheless, it is clear that reduced DO expression results in less DM binding (Fig. 6C) and that free DM remains in these cells but yet the free DM did not result in efficient CLIP removal from MHCII. The mechanism for how equivalent DM inhibition occurred despite lower DO levels remains unknown. This suggests that DO function may be more complex than previously appreciated.

Uncovering the precise mechanisms by which the DO proteins with altered activity function will require the development of an experimental system that allows for the analysis of DOA variants in which all the components of the MHCII antigen processing pathway are present at endogenous levels. Such a system could be achieved by using Crispr-cas9 mediated gene editing to create EBV B cell lines expressing individual DOA that have altered activity. Furthermore, the DOA variants with altered function identified in this paper are rare variants and homozygous individuals would be uncommon. Using this system, it would also be possible to create cells in which one chromosome was expressing the canonical DOA*0101 allele and the other chromosome expressing one of the rare DOA gene variants identified in this screen as having altered function. This system could also be used to evaluate previously described alleles of DOB, DMA and DMB with altered function (21, 44, 51, 52).

The role of the classical MHCII genes in anti-viral immunity are well defined and are usually attributed to the efficient presentation of viral-derived peptides on specific MHCII alleles that results in a robust CD4 T cell immune response targeted against the virus. It is well appreciated that other genetic modifiers are encoded in the MHC locus. However, the discovery of these modifiers has been hampered by the density of MHC genes (>140), the strong linkage disequilibrium across the MHC, and the effects of multiple HLA loci. Our previous studies in which we characterized rare, naturally occurring DOB gene variants, identified a gain-of-function allele (DOβ G77V) (21), with a phenotype similar to the gain-of-function DOα alleles identified in this study. In that study, we showed that DOβ G77V was associated with a SNP that was in linkage disequilibrium with a SNP located in the DQA2-DQB2 locus which had been previously associated with Hepatitis B and C virus persistence (21, 53). The link of viral persistence to the DQA2-DQB2 locus was assumed to be due to restriction of peptides presented by this classical MHC class II allele, leading to viral persistence. However, our studies showed that viral persistence phenotype could instead be due to altered DO function mediated by the linked DOβ G77V allele. Or in other words, the MHC encoded DOβ G77V allele was as likely to be the genetic modifier that leads to the viral persistent phenotype as the specific DQ allele. In this study, we similarly identified variant DOA alleles with altered function and linked them to the outcome hepatitis B viral infection. A more functional allele DOA*0102 allele was associated with a SNP which was in linkage disequilibrium with the ‘C’ allele of the DPA1 allele which had previously been associated with Hepatitis B Virus persistence (50). Conversely, the ‘T’ allele of the same DPA1 allele was associated with Hepatitis B clearance and the generation of an early neutralizing antibody response (50, 54, 55). Remarkably, our studies showed that this DPA1 ‘T’ allele was linked to the DOα F114L loss of function variant. Consequently, as for DOβ G77V, our studies here also showed that two DOA alleles with altered function were as likely to be the genetic modifiers that leads to the viral clearance or persistent phenotypes as the specific DPA1 allele that encodes for a classical MHCII presentation molecule.

Most strains of mice are susceptible to the mouse retrovirus, mammary tumor virus (MMTV). Our recent studies showed that loss of H2-O function allowed mice to become resistant to MMTV. The mechanism for resistance was shown to be via the production of a neutralizing antibody response (21). While the mouse-based studies allowed us to directly link the neutralizing antibody response to a non-functional H2-Oβ protein (mouse DOβ), our studies thus far in humans are based on genetics and, thus, are correlative. Nevertheless, collectively our discovery of DOA and DOB functional gene variants together with our mouse-based studies showing control of the neutralizing antibody response to retroviruses by H2-O supports that naturally occurring variation in the DO/H2-O genes contributes to the outcome of chronic viral infection in mice and humans.

Key Points:

1. Three of the four HLA-DOA common alleles have differential impact on DM function.

2. Fifty-two percent of all DOA variants tested displayed altered function.

3. Two DOA variants were genetically linked to the outcome of Hepatitis B infection.