Mouse Homologue of Human HLA-DO Does Not Preempt Autoimmunity but Controls Murine Gammaherpesvirus MHV68

The immune system has to maintain a delicate balance between efficient defense against pathogens and self-damaging inflammation. During an infection, pathogen-derived peptides are presented on the surface of professional APCs by MHC class II (MHCII) molecules to CD4⁺ T cells to initiate the adaptive immune response. Peptide loading of MHCII molecules occurs in the endosomes and lysosomes of APCs and is catalyzed by human HLA-DM (H2-M), a nonclassical MHCII-like heterodimer (1). The final MHCII peptide array, however, is fine-tuned by another MHCII-like heterodimer, human HLA-DO (H2-O), which binds to H2-M and inhibits its function (2–4).

H2-O is an obligate αβ heterodimer, which in mice is encoded by the H2-Oa (Oa) and H2-Ob (Ob) genes (5). Biochemical, structural, and functional analyses all supported the role of H2-O as a negative regulator of immune responses (2, 3, 6–8). Moreover, supporting this idea, H2-O–deficient mice and rare humans carrying loss-of-function alleles of DO generate potent neutralizing Ab responses against several viruses (9, 10).

Negative regulation of immunity by H2-O/DO led to prediction of its role as a guardian against autoimmunity, suggesting that autoimmune reactions should be found in H2-O/DO–deficient organisms. However, experimental evidence for autoimmunity in H2-O–deficient mice was not entirely compelling [the appearance of IgG2a anti-nuclear Abs (ANA) (4) was not confirmed by others (3, 9)]; whereas a report of an enhanced development of experimental autoimmune encephalomyelitis (EAE) in B6.Oa^−/− mice (11) had an unusually low disease score in the wild-type (WT) control mice. However, the possibility still remained that a role for H2-O in preventing autoimmunity could be more definitively shown in mice from autoimmune-prone backgrounds.

Experimental mouse models of autoimmunity have been developed for numerous human diseases, including organ-specific and systemic models. Spontaneous models of autoimmunity are of particular interest, as they recapitulate many features of human diseases and are controlled by multiple (known and unknown) genes. Two broadly used models, the NOD mouse model of type 1 diabetes (T1D) and the C57BL/6J.NZM (New Zealand mixed [NZM] B6J [B6.NZM]) model of systemic lupus erythematosus (SLE) (12), display these characteristics. Other models of autoimmunity rely on induction of disease via immunization, cell transfer, or other experimental manipulations. One example is an induction of EAE in B6 mice immunized with myelin oligodendrocyte glycoprotein (MOG) peptide in the presence of pertussis toxin (13).

To conclusively and definitively establish a role (or a lack thereof) for H2-O in autoimmunity, we used these three mouse models to test whether H2-O-deficiency would result in accelerated and/or exacerbated disease phenotypes. We found no significant enhancement of the disease phenotype in H2-O–deficient mice in all three models. Instead, we found that H2-O–deficient mice had an increased susceptibility to γ-herpesvirus MHV68. Therefore, we have to conclude that the function of H2-O is not in prevention of autoimmunity but is likely in controlling chronic viral infections.

NOD/ShiLtJ (NOD), C57BL/6J (B6), and B6;NZM-Sle1^NZM2410/Aeg Sle2^NZM2410/Aeg Sle3^NZM2410/Aeg/LmoJ (B6.NZM) mice were purchased from The Jackson Laboratory. B6.Ob^−/− produced by us (9) were maintained at the University of Chicago, at Rutgers University, and at the Medical College of Wisconsin. B6.Oa^−/− mice (3) were maintained at Rutgers University. NOD.Ob^−/− mice were produced at the Transgenic and Embryonic Stem Cell Core of the University of Chicago using CRISPR/Cas9 technology. The 5′-CTGTAGATGTCACCAAGCTC-3′ guide targeting exon 3 was used to produce NOD.Ob^−/− mice. Mice from one line (#19), which have a 22-bp deletion in the exon 3, were used in the studies. The founder of the line mouse #19 was crossed to NOD mice to ensure the germline transmission of the knockout (KO) allele. Heterozygous female mice were crossed to homozygous males to generate homozygous and heterozygous mice inheriting the same microbiota from their mothers.

B6.NZM mice were crossed to B6.Ob^−/−, and the resulting B6.NZM.Ob^−/+ mice were intercrossed to obtain B6.NZM.Ob^−/− and B6.NZM.Ob^+/+ mice. PCRs specific for the NZM SLE (Sle) 1/2/3 loci mapped to chromosomes 1, 4, and 7 were done with primers at proximal, intermediate, and distal locations of each of the loci using the following primers: proximal Sle1 forward, 5′-GTGTCTGCCTTTGCACCTTT-3′ and proximal Sle1 reverse, 5′-CTGCTGTCTTTCCATCCACA-3′; intermediate Sle1 forward, 5′-TCCACAGAACTGTCCCTCAA-3′ and intermediate Sle1 reverse, 5′-ATACACTCACACCACCCCGT-3′; distal Sle1 forward, 5′CTGACCTCCACACGACCC-3′ and distal Sle1 reverse, 5′-GCTTGGGAAACTGGATGAAA-3′; proximal Sle2 forward: 5′-TGGCCAACCTCTGTGCTTCC-3′ and proximal Sle2 reverse, 5′-ACAGTTGTCCTCTGACATCC-3′; intermediate Sle2 forward, 5′-GGCTTTGCAATGCTATGCAT-3′ and intermediate Sle2 reverse, 5′-TGGCAGGAGGTATGACAGAA-3′; distal Sle2 forward, 5′-GCTTGCTTTAGGAGTGTGCC-3′ and distal Sle2 reverse, 5′-TATTTGCTCTCCATTTCCCC-3′; proximal Sle3 forward, 5′-CCAGACCATCTGATCCAGATC-3′ and proximal Sle3 reverse, 5′-GGAGGTTGCAGTGAATTCAAG-3′; intermediate Sle3 forward, 5′-CCCACCAGAGATCACCAAGT-3′ and intermediate Sle3 reverse, 5′-CACAATGAAGGCTGAAAGCA-3′; distal Sle3 forward, 5′-CACTTGGGGAACGTCAAGAC-3′ and distal Sle3 reverse, 5′-TGTAGACCATAGCCCATAAGCC-3′. Only PCRs with distal and proximal primers are shown in Supplemental Fig. 1A.

Between 6 and 7 wk of age, mice were intranasally inoculated with 1000 PFU of MHV68 (Washington University Medical School) diluted in sterile serum-free DMEM (15 µl/mouse) under light anesthesia. MHV68 viral stock was prepared and titrated on NIH 3T12 cells. The spleens were harvested from euthanized mock-treated and MHV68-infected animals at 16 d postinfection. Mice were euthanized by CO₂ inhalation from a compressed gas source in a non-overcrowded chamber.

The studies described in this study have been reviewed and approved by the Animal Care and Use Committees at the University of Chicago, Rutgers University, and the Medical College of Wisconsin, which are accredited by the Association for Assessment and Accreditation of Laboratory Animal Care.

Pellets of 1 × 10⁷ RBC-depleted splenocytes were lysed in 20 mM Tris-HCl, 130 mM NaCl (pH 7.4) containing 1% Triton X-100, and protease inhibitor mixture (Roche LifeScience) for 30 min on ice. Following centrifugation to eliminate nuclei and cellular debris, lysates were incubated at 98°C for 10 min in Laemmli buffer and resolved on a 15% SDS-PAGE gel (Bio-Rad Laboratories). After electrotransfer, PVDF membranes were blocked in 5% nonfat dry milk and incubated with a rabbit serum to the cytoplasmic tails of Mβ (R.Mb/c; see below) or Oβ [R.Ob/c (14)] or β-actin (clone AC-74; Sigma-Aldrich). The membranes were washed and probed with HRP-conjugated anti-rabbit Ig secondary Abs (Jackson ImmunoResearch). After extensive washing, blots were developed with chemiluminescent peroxidase substrate and visualized by an LAS-3000 Image Reader (FujiFilm).

Polyclonal Abs to the Mβ cytoplasmic tail (R.Mb/c) were produced in rabbits after multiple immunizations with a peptide encoding the Mβ cytoplasmic tail (Mβ/c; YTPLSGSTYPEGRH) conjugated to keyhole limpet hemocyanin. Mβ/c-specific Abs were purified from the sera by affinity chromatography with an Mβ/c peptide column.

The MOG_35–55/CFA Emulsion PTX Kit for EAE induction was purchased from Hooke Laboratories and was used according to the manufacturer protocol. Male and female mice between the ages of 9 and 12 wk were used in these studies. Mice were monitored for 28 d after EAE induction using the company’s scoring guide. In experiments performed at Rutgers University, litters of B6.Oa^−/− and WT B6J pups or B6.Ob^−/− and WT B6J pups were mixed and cross-fostered in two groups on B6 dams to normalize the microbiota. For experiments at the University of Chicago, either Ob^+/+ littermates of Ob^−/− mice (sharing the microbiota source) or B6J mice purchased from The Jackson Laboratory were used as control.

To measure T cell activation, RBC-depleted splenocytes were stained with anti-CD4 (eBioscience), anti-CD8a (Invitrogen), anti-CD62L (BioLegend), anti-CD44 (BioLegend), and anti-CD69 (Invitrogen) Abs in the presence of Fc Block (BD Biosciences). To compare B cell subpopulations, the same samples were stained with anti-B220 (BioLegend), anti-CD93 (BioLegend), anti-CD23 (BioLegend), anti-IgM (Invitrogen), anti-CD21/35 (BioLegend), and anti-CD1d (Invitrogen) Abs in the presence of Fc Block. To compare T follicular helper cells (Tfh), T follicular regulatory cells, and T regulatory cell populations, the samples were stained with anti-CD4 (BioLegend), anti-CXCR5 (BioLegend), anti–PD-1 (Invitrogen), anti-Foxp3 (Invitrogen), and anti-Bcl6 (Invitrogen). All stainings mentioned above were preceded by incubating cells with Fc Block (BD Biosciences). For intracellular staining, cells were permeabilized using Foxp3/Transcription Factor Staining Buffer (Invitrogen). Dead cells were gated out by staining with propidium iodide (Sigma-Aldrich) or Zombie Aqua (BioLegend). For MHV68 studies, the following Abs were used: anti-CD3, anti-CD4, anti-CD95, anti–PD-1, anti-B220, anti-GL7, anti-IgD, anti–IRF-4, anti-CD19, anti-CD11b, anti-CD11c (all from BioLegend), and anti-CXCR5 (BD Pharmingen). Intracellular staining for MHV68 studies was performed using a BD Cytofix/Cytoperm Kit (Thermo Fisher Scientific). Data were acquired using LSRFortessa or LSR II flow cytometers (Becton, Dickinson & Company). Data analysis was performed using FlowJo software (Becton, Dickinson & Company).

Kidney samples embedded in OCT (Tissue Tek; Sakura) were cut with a cryostat into 8-µm-thick sections, which were transferred to microscope slides. Slides were fixed in −20°C acetone, dried, and stained with tetramethylrhodamine-coupled anti-mouse IgG (Jackson ImmunoResearch) or anti-mouse FITC-labeled C3a Ab (MP Biomedicals) in FACS buffer (1% FBS, 0.02% NaN₃ in PBS). After washes in FACS buffer, slides were wetted with 50% glycerol and covered with coverglass (IMEB). The slides were stored at 4°C until imaging using a DMLB microscope (Leica Microsystems) equipped with a SPOT camera (Diagnostic Instruments).

HEp-2 slides (Bio-Rad Laboratories) were incubated with serum samples diluted at 1:100 in FACS buffer and counterstained with tetramethylrhodamine-labeled donkey anti-mouse IgG (Jackson ImmunoResearch) diluted at 1:100 in FACS buffer. After washes in FACS buffer, slides were wetted with 50% glycerol and covered with coverglass (IMEB). The imaging and scoring were performed on the Leica DMLB fluorescent microscope.

Mouse genomic DNA was extracted from toe or ear samples using DirectPCR Lysis Reagent (Viagen Biotech) according to the manufacturer’s protocol and subjected to PCR. The PCR products were separated on 5% acrylamide gel (Bio-Rad Laboratories). For genomic analysis of the NOD.Ob KO allele, the PCR reactions were treated with ExoSAP-IT (Thermo Fisher Scientific) before submitting for sequencing.

Diabetes development was monitored by weekly testing of urine glucose with Diastix Reagent Strips (Bayer, Elkhart, IN).

Insulitis was scored on ≥100 islets per pancreas using 5-µm, H&E-stained sections with 40-µm intervals and scored as follows: 0, no visible infiltration; 1, peri-insulitis; 2, insulitis with <50% islet infiltration; and 3, insulitis with >50% islet infiltration. Kidney pathology was scored using 5-µm, PAS-stained sections with 40-µm intervals and scored as follows by a renal pathologist: 0, no visible change; 1, focal mesangial proliferative changes; 2, diffuse mesangial proliferative disease; 3, diffuse proliferative glomerulonephritis (GN) with focal crescents; and 4, diffuse GN with crescents in >50% of glomeruli.

WT MHV68 (15) and MHV68.LANA:βlac (16) viruses were grown and titrated on 3T12 fibroblasts. Mice were infected intranasally with 1 × 10³ PFUs of WT MHV68 or i.p. with 1 × 10⁶ PFUs of WT MHV68 or MHV68.LANA:βlac (17). Peritoneal cells were collected by lavage at 3 d postinfection and subjected to flow analyses.

Frequency of virally infected nucleated cells (cells harboring viral DNA) was determined by limiting dilution PCR analysis; the frequency of ex vivo reactivation to identify cells capable of producing infectious virus was determined by limiting dilution assay as previously described (18). Briefly, to determine the frequency of cells harboring viral DNA, splenocytes were pooled from all mice within each experimental group (3–5 mice per group), and six serial 3-fold dilutions were subjected to a nested PCR (12 replicates per dilution) using primers against viral genome. To determine the frequency of cells reactivating virus ex vivo, serial 2-fold dilutions of splenocytes harvested from infected mice were plated onto monolayers of mouse embryonic fibroblasts (MEF) immediately following harvest, at 24 replicates per dilution. MHV68 was allowed to reactivate from primary cells, and virus was further amplified within the same well via subsequent replication in MEF. At 21 d postplating, all replicates and dilutions were scored in a binary fashion for the presence of live fibroblasts (no viral reactivation/replication) or absence of such (as a result of cytopathic effect driven by lytic replication). Because primary MEF were used to amplify the virus, the sensitivity of limiting dilution reactivation assay was below a single PFU of MHV68 defined using a cell-line plaque assay. Because the end point of viral amplification in MEF was measured, the limiting dilution reactivation assay was not susceptible to variability of titers released from primary cells upon viral reactivation ex vivo.

Developed in the 1970s, the NOD mouse model of T1D has been used extensively for understanding the genetic contributions to T1D (19). T1D development in mice from this strain is dependent on multiple insulin-dependent diabetes (Idd) genetic loci (20). The major contributor to T1D is Idd1, which contains the MHC genes on chromosome 17. The unique NOD MHC haplotype H2^g7 is critical for development of diabetes in this strain (21). To test whether H2-O deficiency would accelerate diabetes development in NOD mice, we used a CRISPR/Cas9 approach to produce H2-O–deficient NOD mice (Fig. 1A). Targeting of the Ob gene led to the loss of Oβ protein but did not affect the levels of H2-M (Fig. 1B). Diabetes development in NOD mice is a sexually dimorphic trait more penetrant in females (22, 23); thus, diabetes incidence was tracked in female mice. Female NOD.Ob^+/− and NOD.Ob^−/− littermates had no difference in onset of overt diabetes or in overall disease penetrance (Fig. 2A). Infiltration of the pancreatic islets by immune cells is a prerequisite to diabetes development (23). To determine whether insulitis was exacerbated in H2-O–deficient mice, H&E-stained sections from 13-wk-old NOD.Ob^−/− and NOD.Ob^−/+ mice were scored (Fig. 2B, 2C). The differences in the distribution of insulitis scores observed between H2-O–deficient and H2-O–sufficient mice were not significant. If anything, NOD.Ob^−/− mice had mildly reduced islet infiltration compared with NOD.Ob^+/− mice. Taken together, these data demonstrate that H2-O does not influence the progression of T1D in NOD mice.

To address the contribution of H2-O to a systemic autoimmune disease, the B6.NZM tricongenic model was used. B6.NZM mice resulted from the transfer of three SLE (Sle) loci identified in New Zealand white and New Zealand black onto the B6J background (24). The hybrid NZM mouse is highly susceptible to an SLE-like disease, which includes production of ANA and glomerulonephritis (12, 24). Similar to NOD mice, SLE development in B6.NZM mice is a sexually dimorphic trait and is more penetrant in females (12). Mice from the B6.NZM strain are overall B6J genetically, with three loci, Sle1, Sle2, and Sle3, present on chromosomes 1, 4, and 7, respectively, derived from NZM mice (12, 25, 26). These mice possess the H2^b haplotype of MHC, like that of the conventional B6 mice (12). To generate H2-O–deficient B6.NZM mice, we crossed B6.NZM to B6.Ob^−/− and confirmed that these mice lacked Oβ, and thus H2-O, via Western blotting (Fig. 1B), and confirmed they had Sle1-3 loci of the NZM background via PCR (Supplemental Fig. 1A).

Female Ob^+/+ and Ob^−/− B6.NZM mice were studied for three major autoimmune phenotypes at 8 mo of age: 1) ANA titers (Fig. 3A), 2) kidney glomerulonephritis (Fig. 3B, Supplemental Fig. 1B), and 3) kidney IgG and C3 complement deposition (Fig. 3C, 3D). Measurement of these hallmarks of SLE-like autoimmune disease in multiple organ systems showed no differences between H2-O–deficient and –sufficient mice. Furthermore, no differences in splenic immune cell activation, a known and measurable precursor for SLE disease in B6.NZM mice (27–29), were observed. The frequencies of splenic T regulatory cells, Tfh, and T follicular regulatory cells (Fig. 3E, 3F; Supplemental Fig. 1C) were similar between Ob^−/− and Ob^+/+ B6.NZM mice. Furthermore, no significant differences were noted in frequencies of CD69⁺-activated CD4⁺ (Fig. 3G) or in frequencies of activated and effector/effector memory CD8⁺ T cells (Supplemental Fig. 1C). The absolute numbers of the CD4⁺ T cell subsets from (Fig. 3E–G, were also similar between the WT and the KO animals (Supplemental Fig. 1D). Follicular B cells and marginal zone B cells were present at similar frequencies (Fig. 3H, 3I) and numbers (Supplemental Fig. 1E) in the spleens of Ob^−/− and Ob^+/+ B6.NZM mice. Of all parameters measured, only splenic CD4⁺ effector/effector memory T cells were slightly but significantly elevated in Ob^−/− compared with Ob^+/+ B6.NZM mice (Fig. 3J, Supplemental Fig, 1F). This difference was not observed between age-matched B6.Ob^−/− and B6.Ob^+/+ mice (Fig. 3J) and thus was attributable to the H2-O deficiency on the NZM background. In the absence of any evidence for augmented autoimmune pathology, however, the physiological relevance of this elevation in CD4⁺-activated memory T cells remains unclear. Overall, our data indicate that H2-O deficiency does not exacerbate SLE-like pathology in the B6.NZM model of autoimmunity.

In addition to models of spontaneous autoimmunity, we also tested whether H2-O deficiency would contribute to an acute onset of induced autoimmunity. EAE is a mouse model of multiple sclerosis and is induced by immunization of B6 mice (susceptible strain in this case) with a MOG peptide emulsified in CFA, followed by the injection of pertussis toxin. MOG peptide is presented to T cells in the context of I-A^b MHCII molecules, thus allowing for the use of B6.Ob^−/− and B6.Oa^−/− mice in testing for the role of H2-O in development of the inflammation in the CNS. Autoimmune infiltration leads to an ascending paralysis beginning in the tail ∼9 d postimmunization that progresses to the hind limb and, in some cases, forelimb paralysis several days after symptom onset. To test whether H2-O–deficient mice were differentially susceptible to EAE compared with H2-O–sufficient mice, EAE was induced by immunization in B6.Oa^-/-, B6.Ob^−/−, and control B6J mice. Immunized mice were monitored for the next 28 d for the symptoms of neuroinflammation. No statistically significant differences in disease development in H2-O–deficient (B6.Ob^−/− mice) compared with H2-O–sufficient mice were observed when the data from all experimental mice from the University of Chicago and Rutgers University (both males and females) were combined (Fig. 4A, 4B) or when only female mice were compared (Fig. 4C, 4D). Similarly, B6.Oa^−/− mice did not develop EAE disease with kinetics or severity that was different from their controls (Fig. 4E, 4F). A recent study (11) found that B6.Oa^−/− mice were more susceptible to EAE than control B6J animals. However, that conclusion was based on a single experiment, in which control B6J mice developed very mild EAE disease with an average score of 1, which is far lower than the disease scores published by other groups (30–32) and also observed in our study. Thus, in this inducible model of autoimmunity, H2-O deficiency does not contribute to either disease development or severity.

Thus, far we have not found a support for any substantial involvement of H2-O in prevention of autoimmunity. Moreover, our own findings in mice (9) and relevant human data (10) suggested that the loss of H2-O/HLA-DO function could be beneficial for promotion of strong neutralizing antiviral Ab responses. However, because H2-O is highly conserved throughout mammalian evolution (33, 34), its function must be important. To search for such a function, we broadened the scope of testing an importance of H2-O in viral infections.

For that, we turned to the mouse γ-herpesvirus 68 (MHV68; also known as murid herpesvirus 4), a natural rodent pathogen that shares high levels of genetic conservation with EBV and Kaposi sarcoma–associated herpesvirus (KSHV) (15, 35, 36) infecting humans. Like EBV, MHV68 initially infects naive B cells and drives both infected and bystander B cells to become germinal center (GC) B cells (37–39). Virally induced GC B cells require Tfh cells (40, 41) and harbor both virus-specific and virus-nonspecific B cells (40, 42, 43). These GC B cells transition to memory B cells that support life-long latent infection or to plasma cells, in which the viral “latent to lytic” switch occurs (44, 45). After a brief acute lytic replication in a naive host, MHV68 establishes latency in several organs, including the spleen (46, 47). Viral latency in the spleen peaks at 14 to 18 d postinfection, with most of the latent virus being present in the GC B cells (48, 49).

To determine whether H2-O has a role in controlling γ-herpesvirus infection, B6.Ob^−/− and B6.Ob^+/+ mice were infected with MHV68, and parameters of viral latency and reactivation were determined at 16 d postinfection. To compare the frequencies of latently infected splenic cells in B6.Ob^−/− and B6.Ob^+/+ mice, we performed a limiting dilution PCR assay amplifying the MHV68 genome from pooled cells in multiple replicates. The frequency of latently infected cells in a given sample was determined by a number of cells added per well needed to achieve 62.5% MHV68 positivity of all replicates. Splenocytes from infected B6.Ob^−/− mice contained approximately six times more latently infected cells than B6.Ob^+/+ splenocytes (Fig. 5A). Similarly, for the ex vivo reactivation assay, splenocytes were added to microplate wells containing MEFs, and the cytopathic effect of the virus (loss of viable MEFs) was scored after 3 wk of coculture. The frequency of infected cells was determined similarly by the PCR assay. The viral reactivation was five times higher in B6.Ob^−/− mice than in B6.Ob^+/+ mice (Fig. 5B). Furthermore, flow cytometry analyses of individual spleens demonstrated that, compared with B6.Ob^+/+ mice, B6.Ob^−/− mice had significantly increased frequencies of GC B cells, Tfh cells, and class-switched plasma cells (Fig. 5C–E; gating strategy is shown in Supplemental Fig. 3), reflecting the γ-herpesvirus–driven B cell differentiation. Thus, functional H2-O attenuates the establishment of chronic MHV68 infection. One possible caveat was a higher level of infection of susceptible cells in H2-O–deficient animals. That possibility was ruled out; frequencies of infected cells in the peritoneum of H2-O–sufficient and –deficient mice infected with the virus via i.p. injection were similar among the cells expressing H2-O (dendritic cells, B cells) and macrophages that do not express H2-O (Supplemental Fig. 4).

By interacting with H2-M/HLA-DM, H2-O/HLA-DO edits the repertoire of peptides presented by MHCII to T cells. Indeed, small but significant differences were found in MHCII peptidomes between H2-O/DO–deficient and DO-sufficient cells, with the latter containing many peptides that were absent from or present at much lower frequency in the H2-O/DO–deficient cells (11, 50). The data suggested a potential predisposition to autoimmunity in the absence of H2-O/DO due to an aberrant presentation of some high-affinity self-peptides by MHCII or due to the loss of some low-affinity tolerogenic peptides. Some findings of autoimmune phenotype (4) were not confirmed by others (3, 9), whereas other findings were obtained using control mice developing an underwhelming disease (11). Because we were not convinced that prevention of autoimmunity is the major function of H2-O/DO, we aimed at testing this hypothesis in three distinct disease models in which H2-O deficiency was introduced in animals already prone to autoimmunity. That should have increased the chances to detect higher susceptibility of H2-O–deficient animals to autoimmunity. These models included spontaneous autoimmunity (T1D in NOD mice and SLE in B6.NZM mice), as well as induced autoimmunity (EAE in B6, B6.Oa^−/−, and B6.Ob^−/− mice). In sum, except for a small increase in CD4⁺ T cells with memory phenotype in B6.NZM.Ob^−/− mice compared with B6.NZM mice (model of SLE), which had no visible effect on any other parameters of pathology reflecting disease severity, we found no evidence in favor of H2-O being a negative regulator of autoimmunity. This small increase might be a result of interaction between Ob and an unknown gene mapped in one of the three SLE loci, which may be interesting to dissect in the future.

Interestingly, overexpression of human DO in NOD mouse dendritic cells led to the amelioration of T1D development (51). That was caused by the propensity of human DO to suppress mouse H2-M function and, hence, presentation of diabetogenic peptides. Apparently, endogenous mouse H2-O that coevolved with H2-M was not capable of reaching such a level of suppression of H2-M function. This experiment underlies the importance of DO–DM interactions in editing self-peptides presented by MHCII. However, at the steady state, even if the loss of H2-O–mediated MHCII peptidome editing abrogates the removal of self-peptides that can drive autoimmune disease, other major mechanisms of self-tolerance (negative selection, lack of costimulatory signals, T regulatory cells, etc.) must be sufficient to prevent activation of autoreactive T and B cells.

Because our data do not support a role for H2-O as a protector from autoimmune disease development, we posit that the key to understanding the functions of H2-O/HLA-DO should be found by studying the involvement of these molecules in resistance to pathogens. However, this idea seemingly does not sit well with our own previous findings that H2-O–deficient mice produced potent neutralizing Ab responses against mouse mammary tumor virus, a mouse retrovirus, and that some rare individuals with loss-of-function DOA and DOB alleles clear hepatitis C virus and hepatitis B virus (9, 10). Moreover, the presence of some gain-of-function alleles of DOA and DOB in humans correlated with persistence of hepatitis C virus and hepatitis B virus (9, 10). Therefore, the presence of functional H2-O/DO genes appears to benefit some pathogens, yet these genes are highly conserved among all mammals (33). One reasonable explanation for H2-O/DO conservation throughout mammalian evolution could be in its possible role in protection from some highly prevalent pathogens. Such pathogens are likely to be intracellular, targeting MHCII-positive cells that can present an array of pathogen-derived peptides. Indeed, it has been found that mice of I/St strain carrying a defective Ob allele were uniquely sensitive to Mycobacterium tuberculosis compared with B6 and A/Sn mice carrying functional alleles of Ob (9, 52–54). Although M. tuberculosis can cause an Ab response (55), this response cannot provide pathogen clearance and could even become detrimental by enhancing pathogen uptake through FcR-mediated phagocytosis (56). In the current study, we have identified one such microorganism, γ-herpesvirus MHV68, a prototype of cancer-associated human γ-herpesviruses, such as EBV and KSHV (57), which is controlled by H2-O.

γ-Ηerpesviruses are highly prevalent in multiple mammalian species. These viruses infect B cells and rely on robust B cell differentiation reaction to establish a life-long infection of memory B cells (58). Although B cells are usurped by the virus as hosts of latent infection and viral reactivation, they are required as APCs (59) for inducing virus-specific CD8⁺ T cells (60–62). CD4⁺ T cells have a dual role during γ-herpesvirus infection. Whereas CD4⁺ T cells are proviral during establishment of latency [as they are required for GC reactions induced by the virus (41)], they also exert an antiviral function controlling persistent virus by providing help to virus-specific CD8⁺ T cells (63). We found that MHV68-mediated GC B cell and Tfh cell expansion was significantly increased in H2-O–deficient mice compared with H2-O–sufficient mice (Fig. 5C–5E), coinciding with a rise in the number of latently infected splenic cells (Fig. 5A). Although the virus is capable of infecting both B cells and myeloid cells, B cells are the most likely contributors to the overall rise in the latently infected cells, as the ratio of B cells to myeloid cells in the spleen is ∼40–70:1. These results indicate that proper expression of H2-O, most likely in B cells, puts the pressure on the virus during the establishment of latency, restricting the overall pathogen burden.

Most adults worldwide are chronically infected with EBV and the level of seroprevalence for KSHV reaches 50% in Sub-Saharan Africa and ∼10% in central Europe or North America (57). These viruses can cause severe illnesses in immunocompromised patients; however, they are well controlled by the immune system in most immunocompetent hosts. Our data suggest that the control of γ-herpesviruses is dependent on the presence and proper functioning of the HLA-DO. In addition, because H2-O expression is not restricted to B cells, as it is also highly expressed in dendritic cells (14) and thymic epithelial cells (64), it is possible that H2-O may also contribute to control of chronic viruses infecting these cells.

The precise mechanism by which H2-O/DO control the establishment of viral latency is yet to be determined, but one can speculate that it is based on reduction of DM-dependent presentation of high-affinity peptides, leading to downregulation of GC reactions triggered by the CD4⁺ T cells.

We thank members of the Golovkina and Chervonsky laboratories for discussions and Louis Osorio for technical assistance.

The authors have no financial conflicts of interest.