Paradoxical Effects of Endoplasmic Reticulum Aminopeptidase 1 Deficiency on HLA-B27 and its Role as an Epistatic Modifier in Experimental Spondyloarthritis

Abstract

Objective.

Examine the functional basis for epistasis between endoplasmic reticulum (ER) aminopeptidase 1 (ERAP1) and HLA-B27 in experimental spondyloarthritis.

Methods.

ERAP1 knockout rats were created using genome editing and bred with HLA-B27/human β₂m (HLA-B27 Tg) and HLA-B7 Tg rats. Effects of ERAP1 deficiency on HLA allotypes were determined using immunoprecipitation and immunoblotting, flow cytometry, allogeneic T cell proliferation assays, and gene expression analyses. Animals were examined for clinical features of disease, and tissue assessed by histology.

Results.

ERAP1 deficiency increased the ratio of folded to unfolded (β₂m-free) HLA-B27 heavy chains, while having the opposite effect on HLA-B7, and reduced HLA-B27 misfolding, while cell surface free HLA-B27 heavy chain dimers and monomers increased. Effects of ERAP1 deficiency persisted during upregulation of HLA-B27, and reduced ER stress. ERAP1 deficiency reduced the prevalence of arthritis in HLA-B27 Tg rats by two thirds without reducing gastrointestinal inflammation. Dendritic cell abnormalities caused by HLA-B27, including reduced allogeneic T cell stimulation and loss of CD103+/MHC class II+ cells were not rescued by ERAP1 deficiency, while excess Il23a upregulation was mitigated.

Conclusions.

ERAP1 deficiency reduces HLA-B27 misfolding and improves folding while having opposing effects on HLA-B7. Partial protection of HLA-B27 Tg rats from spondyloarthritis is consistent with genetic evidence that loss-of-function and/or reduced expression of ERAP1 reduces the risk of ankylosing spondylitis. Functional studies support the concept that effects of ERAP1 on HLA-B27 and spondyloarthritis may be a consequence of how peptides affect the biology of this allotype rather than their role as antigenic determinants.

INTRODUCTION

Common variants in endoplasmic reticulum (ER) aminopeptidase 1 (ERAP1) are associated with the human leukocyte antigen (HLA) class I linked diseases ankylosing spondylitis (1), Behçet’s (2), and psoriasis (3). In each case the ERAP1 association is conditional on the presence of HLA risk alleles indicating epistasis (2–4). ERAP1 trims peptides in the ER to optimize their binding to major histocompatibility complex (MHC) class I heavy chains (HCs) (5). Peptide-HC-β₂m complexes are displayed on the cell surface for recognition by T and NK cells. Consequently, the influence of ERAP1 on peptides as antigenic determinants is thought to underly epistasis (2, 4).

HLA-B27 is a major risk factor for ankylosing spondylitis and related spondyloarthritides, yet its role in pathogenesis remains unclear (6). Arthritogenic peptides remain elusive, and in experimental models of spondyloarthritis, including HLA-B27 transgenic rats, CD8+ T cell recognition of HLA-B27 is not required (7). Other mechanisms based on aberrant properties of HLA-B27 have been proposed. For example, cell surface HCs may trigger innate immune cells (8), aberrant intracellular HC folding and trafficking could lead to activation of stress response pathways (6, 9, 10), and altered BMP/TGFβ signaling could play a role (11). Furthermore, these mechanisms may contribute to alterations in gut microbiota that promote disease development and/or progression (6, 12).

Several non-synonymous ERAP1 variants exist as haplotypes encoding distinct allotypes (13), where reduced aminopeptidase activity is associated with protection from ankylosing spondylitis (4, 14). Consistent with reduced function being protective, increased ERAP1 mRNA and protein expression are associated with increased risk (15). Interestingly, the haplotype conferring protection for ankylosing spondylitis in HLA-B27 carriers is associated with greater risk for Behçet’s disease in HLA-B51 carriers (2, 13), suggesting a complex relationship between ERAP1 function and HLA class I alleles.

ERAP1 deficiency disrupts the production of optimal MHC-binding peptides leading to reduced cell surface class I expression, diminished CD8+ T cell (16) and increased innate immune responses (17), and a marked increase in the average length of peptides presented by MHC class I (18), with a variable effect on individual peptide epitopes (5). For HLA-B27, the effect of reduced ERAP1 activity on cell surface expression of folded and unfolded HCs depends on the study (19–23), while effects of ERAP1 deficiency on HLA-B27 misfolding have not been studied.

Here, we used genome editing to investigate functional interactions between ERAP1 and HLA-B27 and their consequences for experimental spondyloarthritis. We show for the first time that ERAP1 deficiency improves HLA-B27 folding, reduces misfolding, and partially protects HLA-B27 Tg rats from arthritis, despite increasing cell surface expression of aberrant forms of HLA-B27. Reduced misfolding is surprising given the importance of peptide availability in maintaining the fidelity of HLA class I assembly (24). Our results implicate loss of ERAP1 function in protection from experimental spondyloarthritis through a mechanism distinct from the production of antigenic determinants presented by HLA-B27.

MATERIALS AND METHODS

Animals.

HLA-B27 (B*27:05)/human β₂m (hβ₂m)-transgenic (HLA-B27 Tg) Lewis rats (33–3), and HLA-B7 (B*07:02)/hβ₂m-transgenic (HLA-B7 Tg) Lewis rats (120–4) were used (25). ERAP1-edited Sprague-Dawley (SD) rats were generated by Transposagen (Hera BioLabs, Lexington, KY) using TALEN-mediated mutagenesis (Supplemental Methods and Supplemental Figure 1). Allogeneic T cells were isolated from Dark Agouti (DA) rats. Animals were maintained in AAALAC-approved facilities on the NIH campus (Bethesda, MD) and all experiments were approved by the ACUC.

Antibodies and reagents.

HC10 (mouse IgG2a) recognizes HLA-B and C, β₂m-free (unfolded) HCs (26); 3B10.7 (rat IgG2a) recognizes HLA-B in immunoblots (27); W6/32 (mouse IgG2a) recognizes folded (β₂m-associated, conformational epitope) HLA class I (28); ME1 (mouse IgG1), recognizes HLA-B27, -B7, -B42, -B67, and -Bw22 (29); MARB4 (mouse IgG2a), recognizes a subset of HLA-B27 molecules that includes complexes containing long peptides and β₂m-free HCs (30, 31). W6/32 can recognize rat class I bound to bovine or human β₂m, and was not used for flow cytometry. Immunoblotting was performed with 3B10.7, which is specific for human HCs, anti-ERAP1 (Novus Biologicals), anti-GAPDH (Santa Cruz Biotechnology), and anti-BiP (R&D Systems). Anti-CD103 (OX62) microbeads (Miltenyi Biotec) were used for isolation of splenic DCs; anti-CD45RA (OX33) (BD Bioscience), anti-CD11b/c (OX42) (BD Bioscience), anti-CD161 (3.2.3) (Biolegend), anti-CD8 (OX8) (LSBio), and anti-mouse IgG microbeads (Miltenyi Biotec) were used for negative selection of CD4⁺ Tcells. HC10 and mouse IgG2a isotype control were conjugated to APC (Columbia BioSciences). ME1 and MARB4 (and respective isotype control antibodies) were conjugated to Pacific Blue (Life Technologies) and APC/Cy7 (Novus Biologicals), respectively. The following reagents were used: interferon-γ (IFNγ) (Peprotech); methyl methanethiosulfonate (MMTS), lipopolysaccharide (LPS) and, collagenase D (Sigma-Aldrich); Nycodenz AG (Thermo Fisher Scientific) and CellTrace Violet Cell Proliferation kit (Thermo Fisher Scientific); and Lympholyte Cell Separation Media (Cedarlane).

Generation and isolation of myeloid cell populations.

Bone marrow (BM)-derived macrophages (BMDM) were derived as described previously (25) from euthanized 6–8 week-old rats. White blood cells were cultured in high-glucose Dulbecco’s modified Eagle’s medium (Thermo Fisher Scientific) supplemented with fetal calf serum (10%) (FCS) (Thermo Fisher Scientific), penicillin (100 U/mL) (Thermo Fisher Scientific), streptomycin (100 μg/ml) (Thermo Fisher Scientific), with macrophage colony-stimulating factor (20 ng/mL) (Peprotech) for 7 days at 37°C in a 5% CO₂ humified atmosphere prior to experiments.

Dendritic cells (DCs) were derived from non-adherent BM cells with mouse Flt3L (300 ng/mL) (PeproTech) in complete RPMI 1640 supplemented with FCS (5%), L-glutamine (2 mM), penicillin (100 units/mL), streptomycin (100 μg/mL) and 2-mercaptoethanol (50 μM) (Thermo Fisher Scientific) as described (32). Media was changed every two days, and cells were analyzed after 7 days.

For allogeneic T cell proliferation studies, low density splenic DCs were isolated by centrifugation on a 14.5% Nycodenz gradient as described (33), and then positively selected using anti-CD103 (OX62) (Invitrogen) antibody, goat anti-mouse IgG–conjugated microbeads, and magnetic separation columns employing an AutoMACS Pro cell separator (Miltenyi Biotec). Isolated DCs were then cultured overnight in RPMI 1640 Medium 10 (Gibco) supplemented with rat granulocyte-monocyte colony stimulating factor (100 ng/mL) (Peprotech). CD4⁺ T cells were isolated from DA rat lymph nodes by negative selection using a mixture of anti-CD45RA (OX33), anti-CD11b/c (OX42), anti-CD161 (3.2.3), and anti-CD8α (OX8).

Immunoprecipitation and immunoblotting.

Immunoprecipitations were performed as previously described (22). Cells were treated with MMTS (10 mM) to prevent spontaneous formation or loss of sulfhydryl bonds, lysed in the presence of proteases and MMTS, and then centrifuged to remove nuclei and debris. Supernatants were pre-cleared with protein A-sepharose and normal mouse serum for 1 hour, and incubated with purified antibodies and additional protein A-sepharose overnight. Immuoprecipitates were washed as described (22) and then analyzed. Immunoprecipitation of cell surface proteins was also performed as described (22) using HC10 or mouse IgG2a (control). Cells were washed extensively with PBS to remove non-binding antibody, and then lysed as described above followed by immunoprecipitation.

Samples were analyzed by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) under reducing or non-reducing conditions on 4–20% Tris-Glycine gels (Bio-Rad) and immunoblotting as described (25). Following transfer, 3B10.7 (hybridoma supernatant) or anti-BiP (R&D Systems) was added and membranes incubated overnight at 4°C. After washing, blots were incubated with HRP-conjugated secondary antibody for 1 hr, then bound antibody was visualized using Pierce ECL Western Blotting Substrate (Thermo Fisher Scientific). Immunoblots were visualized and quantified as described previously (22) using a ChemiDoc Imager with Image Lab Software (Bio-Rad). Exposures were adjusted to avoid saturation and enable band quantification in the linear response range. Replicates were imaged from the same gel.

Flow cytometry.

Cells were harvested, washed and incubated in PBS with 5% normal mouse serum for 1 hour at 4°C to block non-specific antibody binding, then washed and incubated with fluorescent probe-conjugated specific antibodies or IgG controls for 1 hour at 4°C. Cells were then washed 4 times before analysis using a FACSCanto II Cell Analyzer (BD Biosciences).

Gene expression analysis.

Real time quantitative PCR (qPCR) was performed on total RNA using the iScript cDNA Synthesis kit (Bio-Rad) and SsoFast EvaGreen Supermix on an iCycler Thermo Cycler (BioRad). Expression of Hspa5 (BiP) and Ddit3 (CHOP) target genes was normalized to housekeeping genes (Actb, Hprt1, and Ppia). XBP1 mRNA splicing was measured as described previously by quantifying relative amounts of spliced and unspliced PCR products on 4% agarose gels (34). Oligonucleotide primer sequences are available upon request.

Mixed lymphocyte culture.

Purified CD4⁺T cells labeled with CellTrace Violet (Thermo Fisher Scientific) (1X10⁵ cells/well) were co-cultured with varying numbers of irradiated (2,000 rad) DCs, in round-bottom 96-well culture plates, in a final volume of 200 μl, at 37°C, 5% CO₂ for 5 days. T cell proliferation was assessed by dye dilution using flow cytometry.

Clinical evaluation.

Arthritis was scored as follows: 0=normal, no sign of joint swelling or compromised mobility; 0.5=thickened Achilles tendon or toe swelling; 1=definite ankle swelling; 1.5=increased ankle swelling; 2=severe ankle swelling and/or involvement of distal paw. Fecal material was scored as follows: 0=hard, dry; 0.5=semi-soft; 1=soft and sticky; 2=mushy; 3=watery. Scoring was performed by animal caretakers blinded to genotype.

Histologic scoring.

Hematoxylin and eosin-stained tissue sections from paraffin-embedded colon samples were scored in a blinded fashion by two independent observers using an established scoring system (35). Samples with initial discordant scores (differing by >1 unit) were re-randomized and re-scored, which eliminated discrepancies.

Computed tomography.

Formalin-fixed joints were subject to high-resolution micro computed tomography using the NIH Mouse Imaging Facility.

Statistical analysis.

Statistical analysis included Student’s t-test performed using GraphPad Prism 9.0 with P values less than 0.05 considered significant.

RESULTS

Effects of ERAP1 deficiency on HLA-B allotypes.

To assess how ERAP1-deficiency affects different HLA-B allotypes, rats lacking ERAP1 protein expression were generated (Supplemental Methods and Supplemental Fig. 1) and bred to produce HLA-B27 Tg and HLA-B7 Tg rats with three ERAP1 genotypes (+/+, +/−, and −/−). ERAP1 is undetectable in ERAP1−/− rats, and expressed at about 50% of wild type levels in heterozygous (+/−) rats (Supplemental Fig. 1C). We next assessed the relative proportion of folded, unfolded, and misfolded forms of HLA-B allotypes using antibodies that recognize unfolded and β₂m-free HLA-B HCs (HC10), and folded β₂m-associated HCs (W6/32) by immunoprecipitation and immunoblotting with 3B10.7 (22, 27, 36). The terms ‘free’ and ‘unfolded’ are used synonymously when referring to HCs recognized by HC10. HC10 also recognizes disulfide-linked or ‘misfolded’ forms of HLA-B.

In ERAP1-deficient BMDMs there is less unfolded HLA-B27 compared with ERAP1+/+ cells (Fig. 1A). In contrast, ERAP1-deficiency had the opposite effect on HLA-B7, with greater accumulation of unfolded HCs (Fig. 1A). There was a trend toward more folded HLA-B27 and a ~25% increase in folded HLA-B7 in the absence of ERAP1 (Supplemental Fig. 2). Comparing the ratio of folded to unfolded HCs, it is clear that complete lack of ERAP1 increases the proportion of folded HLA-B27 HCs, while it has the opposite effect on HLA-B7 (Fig. 1B). It should be noted that HLA-B27 and B7 differ dramatically in the proportion of folded to unfolded HCs (Fig. 1B) even in the presence of ERAP1, with the greater proportion of unfolded HLA-B27 appearing similar to our previous observations comparing HLA-B27 to HLA-B18 and HLA-B51 (22). Differential effects of ERAP1-deficiency on the proportion of folded/unfolded HLA class I are consistent with an allotype-specific role for this aminopeptidase (14, 21, 22).

Figure 1.

Effects of ERAP1 deficiency on folded and unfolded forms of HLA-B27 and HLA-B7. A, HLA-B allotypes were immunoprecipitated (IP) from whole cell lysates of BMDMs with HC10, separated by SDS-PAGE, and visualized and quantified by immunoblotting. Representative HC10 IPs are shown, with W6/32 IPs shown in Supplementary Fig. 2. B, The ratio of folded to free HC for HLA allotypes and each ERAP1 genotype. C, HC10 IPs from whole cell lysates were separated by non-reducing SDS-PAGE to visualize HLA-B27 dimers and oligomers. D, Relative expression of dimers and dimer to monomer ratio. E, Representative HC10 IP immunoblotted for BiP. F, BiP quantification. Quantitative results represent averages from duplicate gels from 2–3 independent experiments with standard error of the mean (SEM) shown by error bars. O, indicates oligomers; M, monomers; IP, immunoprecipitating antibody; WB, immunoblotting antibody. In D and F, the values of ERAP1+/+ samples were set to 1 with ERAP1+/− and ERAP1−/− values expressed relative to ERAP1+/+. In C, the M band is saturated to better display D and O bands. When determining the ratio in D unsaturated exposure of the M band was used. (*, p < 0.05; **, p < 0.01; *** p < 0.005)

A proportion of HLA-B27 HCs misfold into disulfide-linked dimers and oligomers that can be visualized using non-reducing SDS-PAGE (27). ERAP1 deficiency also diminishes the accumulation of misfolded forms of HLA-B27 (Fig. 1C), expressed as total and when normalized to monomers (Fig. 1D). Co-precipitation of the ER chaperone BiP, which binds to some folding intermediates and misfolded proteins, with HLA-B27 was reduced by about 50% (Fig. 1E,F). HLA-B7 does not dimerize significantly and only binds BiP transiently (25). Thus, ERAP1 deficiency diminishes HLA-B27 misfolding.

Effects of ERAP1 deficiency on cell surface expression of HLA-B allotypes were assessed by flow cytometry and immunoprecipitation of cell surface complexes (22). In the absence of ERAP1, there are consistent increases in HC10-staining for free HLA-B27 and HLA-B7 HCs (Fig. 2A). However, no significant differences were found with ME1, an antibody that recognizes folded HLA-B27 and HLA-B7 (29, 31) (Fig. 2A). There is a large increase MARB4 staining, an antibody that recognizes a unique HLA-B27 epitope that exists on some β₂m-free HCs (31) including disulfide-linked dimers (27), and complexes containing longer peptides (30), consistent with HLA-B27-peptide sequencing studies of cells with reduced ERAP1 expression (37, 38). Immunoprecipitation of cell surface HLA-B27 with HC10 revealed an approximately 60% increase in disulfide-linked HLA-B27 dimers in the absence of ERAP1 (Fig. 2B,C), confirming our previous results showing more free HLA-B27 HCs on the cell surface of ERAP1-knockdown U937 cells (22). However, cell surface HC10-reactive monomers are increased to an even greater extent in the absence of ERAP1 (Fig. 2B–D), resulting in a 30% reduction in the dimer to monomer ratio (Fig. 2B). The increase in free monomers and disulfide-linked dimers on the cell surface shown here may seem inconsistent with the decrease in free monomers and disulfide-linked HLA-B27 HCs in whole cell lysates noted above (Fig. 1). However, free HLA-B27 HCs on the cell surface are only a fraction of the total cellular free HC (Supplemental Fig. 3). Taken together, these results demonstrate that loss of ERAP1 increases the proportion of folded HLA-B27 HCs, while it has the opposite effect on HLA-B7. Moreover, ERAP1 deficiency diminishes the overall accumulation of misfolded and BiP-bound forms of HLA-B27 HCs, while increasing cell surface free HCs, including dimers.

Figure 2.

Effects of ERAP1 deficiency on cell surface forms of HLA-B27. A, BMDMs from HLA-B27 Tg and HLA-B7 Tg rats were stained with antibodies as indicated. Results are expressed as fold change in mean fluorescence intensity normalized to ERAP1+/+ cells. B,C, Cell surface HLA-B27 was immunoprecipitated, subjected to SDS-PAGE under non-reducing conditions, and visualized by immunoblotting. B, Representative immunoblot of non-reducing gel (NR), showing dimers (D) and monomers (M), with quantification of cell surface HLA-B27 dimers and monomers from multiple immunoblots shown in bar graphs. Left graph, relative expression of HLA-B27 dimers in ERAP1−/− relative to ERAP+/+ cells. Right graph, average dimer to monomer ratio as determined in each immunoprecipitate. Results in A and B represent the mean from 3 animals, with IPs performed in duplicate, while C represents duplicates from 2 animals. Error bars represent the SEM. (*, p < 0.05).

Effects of ERAP1 deficiency on HLA-B allotypes during IFN-γ-induced upregulation.

To examine how ERAP1 deficiency affects HLA-B allotypes during upregulation, BMDMs treated with IFNγ were harvested at various time points. In the absence of ERAP1 the ratio of folded/unfolded HLA-B27 HCs is consistently greater up to 24 hours (Fig. 3A, Supplemental Fig. 4A), in contrast to HLA-B7 where the ratio of folded/unfolded HCs is substantially lower (Supplemental Fig. 5A,B). ERAP1 deficiency also reduces the accumulation of disulfide-linked forms of HLA-B27 (Supplemental Fig. 4A), the ratio of dimers to monomers (Fig. 3B), and the amount of BiP bound to HLA-B27 HCs (Fig. 3C and Supplemental Fig. 4B). We have previously demonstrated that the unfolded protein response, as measured by upregulation of BiP Hspa5 (BiP) and Ddit3 (CHOP) mRNAs and increased XBP1 mRNA splicing, occurs in HLA-B27 but not HLA-B7-expressing or non-transgenic rat macrophages during HLA class I upregulation (36). ERAP1 deficiency mitigates this response in HLA-B27-expressing cells (Fig. 3D), although it is not prevented.

Figure 3.

HLA-B27 folding, misfolding, and UPR activation during upregulation. BMDMs from HLA-B27 Tg rats with the ERAP1 genotypes shown were treated with IFNγ (100 ng/ml) and harvested at the times indicated. Cell lysates were used for immunoprecipitation or RNA isolation. A-C, the ratio of folded (W6/32 IPs) to free (unfolded and misfolded) (HC10 IPs) HLA-B27 HCs (A), ratio of dimers to monomers (HC10 IPs) (B), and the relative amount of BiP co-precipitating with unfolded/misfolded HLA-B27 HCs were determined as for Fig. 1B, D, and E. D, BiP (Hspa5) and CHOP (Ddit3) expression was measured using qPCR normalized to 3 housekeeping genes, and XBP1 splicing determined by RT-PCR and agarose gel electrophoresis and expressed as the percentage of spliced XBP1 mRNA compared to total (spliced plus unspliced). Results represent the mean of 3 independent experiments, performed in duplicate, with error bars indicated the SEM. (*, p < 0.05 and **, p < 0.01 comparing ERAP1+/+ to ERAP1−/− at the same time point).

ERAP1 deficiency protects HLA-B27 Tg rats from arthritis.

HLA-B27 Tg Lewis rats (33–3 line) develop gastrointestinal inflammation at 1–2 months of age, while arthritis and scrotal swelling from epididymoorchitis are infrequent and rarely seen prior to 6–8 months of age. The SD background introduced by breeding HLA-B27 Tg (and HLA-B7 Tg) Lewis rats with ERAP1−/− SD rats is permissive for spondyloarthritis-like disease (39) and increases the frequency of arthritis and epididymoorchitis. Cohorts of HLA-B27 Tg, HLA-B7 Tg, and wild type rats with all three ERAP1 genotypes (+/+, +/−, and −/−) were produced, and stool consistency, clinical arthritis, and scrotal swelling (in males) monitored until approximately 6 months of age. Arthritis develops almost exclusively in HLA-B27 Tg males (Supplemental Table 1). Notably, ERAP1 deficiency reduces the frequency of arthritis in HLA-B27-Tg rats from 32% (12/37; ERAP1+/+) to 12% (4/34; ERAP1−/−) (p=0.03) (Fig. 4A, Supplemental Table 1), with no significant difference in ERAP1+/− HLA-B27 Tg rats (26%; 10/38 vs. 32%; 12/37). Clinical arthritis scores in HLA-B27 Tg rats lacking ERAP1 expression were comparable to those in animals with ERAP1 (1.7 +/−0.58 for ERAP1+/+ vs. 1.5 +/−0.58 for ERAP1−/−), as was the age of onset of arthritis (4 +/−0.5 months for ERAP1+/+ vs. 4.4 +/−0.5 months for ERAP1−/−). Examples of swollen joints with representative histology are shown in Fig. 4D. Micro CT scans of affected joints suggest less bone erosion in the absence of ERAP1 (Fig. 4D), although the limited number of ERAP1-deficient rats with arthritis precluded statistical analysis. Epididymoorchitis, which is clinically apparent as scrotal swelling (39, 40), exhibits inflammatory cell infiltrates on histology (Supplemental Fig. 6). Epididymoorchitis typically correlates with arthritis in HLA-B27 Tg rats (40), and is also significantly reduced by ERAP1-deficiency (Fig. 4B).

Figure 4.

Effects of ERAP1 deficiency on experimental spondyloarthritis in HLA-B27 Tg rats. A,B, cumulative frequency of peripheral arthritis (A) and scrotal swelling (B) in males. Numbers to the right indicate the number of affected animals over the total number in each cohort with the ERAP1 genotype shown. C, colon histology scores from male and female rats euthanized at approximately 2, 4, and 6 months of age. ERAP1 genpotype (black, shaded, and open bars) as indicated in A. Dotted line indicates the average score for unaffected non-transgenic rats from this experiment. For C, 5–6 rats were assessed for each genotype. D, Examples of normal and arthritic joints. Hind limb from healthy HLA-B7 Tg (B7+) rat (left column); affected HLA-B27 Tg (B27+) ERAP1+/+ rat (middle column), and affected HLA-B27 Tg rat with ERAP1−/− genotype (right column). Joint photography (top row); H&E staining of tissue sections (second row); 3D micro-CT (third row); and longitudinal sections from CT (bottom row). Numbers indicate scores assigned by clinical assessment prior to euthanasia. Red arrows indicate erosions.

ERAP1 deficiency had no effect on stool scores in HLA-B27 Tg rats (frequency of abnormal score 10% [6/60] in ERAP1+/+ vs. 10% [6/61] in ERAP1−/−; average score 0.7 +/−0.26 in ERAP1+/+ vs. 0.6 +/−0.22 for ERAP1−/−). Despite the lack of difference in clinical manifestations, histology scores are higher in ERAP1-deficient HLA-B27 Tg rats (Fig. 4C). Thus, while ERAP1-deficiency partially protects HLA-B27 Tg rats from arthritis and epididymoorchitis, colitis may be slightly worse.

Effects of ERAP1 deficiency on HLA-B27-mediated myeloid cell abnormalities.

HLA-B27 expression causes myeloid cell abnormalities that may contribute to the pathogenesis of experimental SpA (32–34, 41). For example, DCs expressing HLA-B27 are poor at stimulating allogeneic T cell proliferation (33), and loss of a CD103+ sub-population implicated in inducing self-tolerance to intestinal antigens and controlling colitis has been reported (32, 41–43). In addition, HLA-B27-mediated unfolded protein response (UPR) activation contributes to Il23a/IL-23p19 overexpression which may promote Th17 development and activation (34, 44). We therefore asked whether ERAP1-deficiency modifies these effects of HLA-B27. Loss of ERAP1 does not restore defective allogeneic T cell proliferation seen in HLA-B27-expressing DCs (Fig. 5A), nor does it restore the CD103+/RT1B+ (MHC class II+) DC population that is reduced by HLA-B27 expression (Fig. 5B). However, ERAP1-deficiency does reduce Il23a upregulation seen in IFNγ-primed HLA-B27-expressing macrophages stimulated with LPS (Fig. 5C), consistent with attenuated UPR activation (Fig. 3D) (34). Thus, effects of ERAP1-deficiency on macrophages raise the possibility that mitigating HLA-B27 misfolding and its consequences contributes to protection from arthritis in HLA-B27 Tg rats. Further studies of these and other cells types will be needed to determine the degree to which IL-23 production may be limited by ERAP1-deficiency, and whether this can account for the reduced frequence of arthritis.

Figure 5.

Effects of ERAP1 deficiency on HLA-B27 mediated myeloid cell abnormalities. A, splenic DCs isolated from rats with the genotypes indicated were incubated with CD4⁺ T cells (1X10⁵ per well) isolated from the lymph nodes of DA rats and labelled with CellTrace Violet. After 5 days, percent proliferating cells was determined by dye dilution using flow cytometry. Results represent the mean from 2 rats performed in triplicate, with error bars representing the SEM. B, DCs were derived in vitro from bone marrow in the presence of Flt3L, and stained with OX62 (anti-CD103) and OX6 (anti-RT1B) to identify CD103+/MHC class II+ cells. Results represent the mean from 3 experiments, performed in triplicate. C, BMDMs from HLA-B27 Tg rats with the ERAP1 genotypes indicated were treated with IFNγ (100 ng/mL) for 24 hours and then stimulated with LPS (10 ng/mL) for the times indicated. Expression of Il23a (encoding IL-23p19) was determined by qPCR. Results represent the mean from 3 rats performed in triplicate, with error bars showing the SEM. (*, p < 0.05)

DISCUSSION

We investigated the epistatic interaction between ERAP1 and HLA-B27, and show that complete loss of ERAP1 confers partial protection from HLA-B27-induced peripheral arthritis and epididymoorchitis without reducing gastrointestinal inflammation. To our knowledge this is the first direct evidence that altering ERAP1 function can protect from arthritis, and is consistent with genetic evidence that human ERAP1 variants conferring reduced activity (4) and/or expression (15) are associated with reduced disease risk in HLA-B27 carriers. This establishes an animal model to explore how reducing ERAP1 function can suppress disease-causing properties of HLA-B27, and supports the potential for targeting this pathway therapeutically.

Our studies demonstrate that ERAP1 deficiency impacts the immunobiology of HLA-B27 in addition to known effects on the peptides that serve as antigenic determinants. Loss of ERAP1 reduces HLA-B27 misfolding as evidenced by a reduction in disulfide-linked and BiP-bound HCs and attenuation of ER stress during upregulation of HLA-B27, and also promotes folding resulting in an increase in folded compared to free HCs. This suggests that ERAP1 deficiency increases the pool of peptides favorable for binding to HLA-B27, which then promotes folding and egress of HCs from the ER, reducing the accumulation of misfolded complexes. Given the prevailing evidence that ERAP1 deficiency decreases the overall availability of peptides optimal for binding to MHC class I proteins, the effect of ERAP1 loss-of-function on HLA-B27 seems paradoxical. Indeed, previous studies have shown that when ER peptide supply is shut off such as in peptide transporter-deficient cells, class I HC misfolding is strikingly increased and they are eliminated by ER-associated degradation (24). Our studies show that in contrast to HLA-B27, loss of ERAP1 impairs HLA-B7 folding, consistent with a reduction in peptides suitable for this allotype, further supporting the paradoxical effect on HLA-B27. These results are also consistent with studies from our lab and others, indicating that misfolding, intracellular dimerization, and the slow assembly phenotype is influenced by peptide binding specificity (27, 45, 46). The paradoxical effect of ERAP1 could be explained if peptides longer than the canonical 8–9 residues are optimal for HLA-B27. Consistent with this idea, extensive studies of the HLA-B27 peptidome have revealed a small percentage of 8-mers relative to other allotypes (38). In addition, ERAP1 has a preference for engaging peptides with hydrophobic C-termini and basic mid-regions, resulting in a shift in the pool of HLA-B27 peptides towards C-terminal Lys and Arg and acidic mid-regions, which are lower affinity ligands for this allotype (38). Thus, peptidome studies strongly support the notion that the net result of ERAP1 activity is to reduce the availability of ligands best suited for HLA-B27 (38), and the paradox of increased folding and reduced misfolding of this allele in the absence of ERAP1 can be best explained by preservation of suitable ligands.

The increase in unfolded forms of HLA-B27 on the cell surface in the absence of ERAP1 may seem at odds with the greater ratio of folded to unfolded HLA-B27. However, only a small proportion of total cell surface HLA-B27 is unfolded (ref. (22) and Suppl. Fig. 3B), and thus it does not significantly impact the overall ratio of folded to unfolded HCs in cell lysates. The presence of more unfolded HLA-B27 on the cell surface is most likely a consequence of the relative instability of complexes containing longer peptides that dissociate resulting in ‘empty’ (β₂m-free) HCs (47). Moreover, free cell surface HLA-B27 HCs are susceptible to endosomal recycling and dimerization (48). Indeed, intracellular dimers form in the ER immediately after HC synthesis, while cell surface dimer formation is considerably delayed (27, 48). Thus, the discordant effect of ERAP1 deficiency on intracellular/ER and cell surface dimers is consistent with their formation via different mechanisms and in different cellular compartments (27). The modest increase in folded HLA-B27 on the cell surface may not account entirely for the overall reduction in misfolded forms of HLA-B27. We (MB and colleagues) have reported previously that HLA-B27 can traffic to a cellular compartment that contains ER chaperones, but is devoid of components of the peptide loading complex (10), raising the possibility of further editing that might limit the appearance of HLA-B27 on the cell surface. Further studies will be needed to explore whether the absence of ERAP1 affects HLA-B27 trafficking in this pathway.

We asked whether loss of ERAP1 impacts the phenotype of HLA-B27 Tg rats, taking advantage of an increase in prevalence of arthritis and epididymoorchitis seen with a mixed Lewis-SD background. There was a two thirds reduction in both arthritis and epididymoorchitis, whereas gastrointestinal inflammation was slightly worse histologically, without a clinically apparent difference. In a variation of the model used here, normally healthy rats with lower HLA-B27 and hβ₂m transgene copy number numbers (21–3 line) develop axial and peripheral arthritis and epididymoorchitis (without gastrointestinal inflammation) when carrying additional copies of hβ₂m from the 283–2 line (21–3 X 283–2) (49). In a previous study using the 21–3 X 283–2 model, we (JDT and colleagues) reported that in the absence of ERAP1 the prevalence of arthritis dropped from 67% (4/6) to 43% (6/14), but this apparent difference was not statistically significant (38) possibly due to sample size. Thus, these two studies support the concept that ERAP1 loss-of-function can reduce the prevalence of experimental spondyloarthritis, while not conferring complete protection. This is consistent with data from humans where protective ERAP1 alleles that are common in the population, and therefore exist in a significant proportion of individuals with HLA-B27-positive axial spondylitis, are not sufficient to confer complete protection.

Several mechanisms could explain the protective effect of ERAP1 deficiency on HLA-B27-induced disease. First, previous studies have shown that ERAP1 deficiency substantially alters the HLA-B27 peptidome, although no peptides specific to animals developing arthritis have been identified (38). While ERAP1 deficiency can diminish CD8+ T cell responses (16), these cells are not required for spondyloarthritis in HLA-B27 Tg rats (7), suggesting that protection is not mediated through arthritogenic peptides. Second, cell surface HLA-B27 HCs have been implicated in spondyloarthritis through innate immune receptor recognition and triggering of IL-17 production (8). We show that free HLA-B27 HCs are actually increased on the cell surface in the absence of ERAP1, making it unlikely that this is a protective mechanism. Finally, we examined effects on myeloid cells, and found that ERAP1-deficiency does not improve the ability of HLA-B27-expressing DCs to stimulate T cell proliferation, nor does it restore the CD103+ population (32, 41). Also known as conventional type 1 DCs, this population appears to include XCR1+ DCs (43), which together are involved in maintaining tolerance to intestinal bacteria (50). Interestingly, CD103-depleted DCs promote Th17/Th1 polarization and colitis (42) in animal models. In contrast, we show that ERAP1-deficiency reduces excess Il23a induction linked to HLA-B27 misfolding and UPR activation (34), raising the possibility that this could be involved in protection. It is worth noting that overexpressing an ER-targeted HLA-B27-binding peptide, which replaces the majority of naturally presented ligands, has a similar effect on disease (39). At the time this was considered as evidence that the specificity of peptides bound to HLA-B27 was critical for arthritis. However, subsequent studies demonstrated that CD8+ T cells were not required for disease (7). Our current studies suggest that overproduction of an HLA-B27-binding peptide might have an effect similar to ERAP1 deficiency, alleviating aberrant intracellular effects of HLA-B27. Further experiments will be necessary to determine precisely how ERAP1 deficiency mitigates disease-causing effects of HLA-B27.

There are several limitations to our study. First, experimental spondyloarthritis in rats requires multiple copies of HLA-B27 and hβ₂m transgenes and does not precisely phenocopy ankylosing spondylitis in humans. However, comparable overexpression of HLA-B7 is innocuous, clearly indicating specificity for a unique aspect of HLA-B27. Second, rodents lack ERAP2, which is also associated with ankylosing spondylitis in humans (51). The protective ERAP2 variant results in a functional knockout of ERAP2 protein expression (51), and thus loss-of-function of both ERAP1 and ERAP2 are associated with protection from ankylosing spondylitis in the context of HLA-B27, a situation recapitulated by the rat model. The lack of ERAP2 in rats suggests the model might already be biased toward protection for spondyloarthritis, which could contribute to the requirement for HLA-B27 overexpression. In this context it is striking that further limiting the production of peptides that are optimal for most class I alleles by eliminating ERAP1 in the absence of ERAP2, improves HLA-B27 folding and reduces misfolding. Finally, the precise cell type (or types) responsible for initiating inflammatory disease in spondyloarthritis remains unknown. We focused our studies on myeloid cells (e.g. monocyte-derived macrophages), as they are clearly involved in rats and almost certainly important in human disease. However, we cannot rule out the possibility that ERAP1 deficiency has different effects on HLA-B27 in other cell types that are relevant for disease.

In summary, our studies provide compelling evidence that epistasis between HLA-B27 and ERAP1 can be modeled in rodents. We show unexpected beneficial effects of ERAP1 deficiency on HLA-B27 folding that are associated with partial protection from spondyloarthritis. This underscores the importance of ERAP1 on the fidelity of HLA class I folding and assembly, and suggests that the effect of ERAP1 on HLA-B27-induced disease may be unrelated to peptides as antigenic determinants.

Supplementary Material

fS1

Supplemental Figure 1. ERAP1 deficient rats. A, Coding sequence of rat ERAP1 exon 1 showing 29 nucleotides (red) that were deleted resulting in 3 in-frame stop codons (bolded and underlined). B, genotyping of ERAP1+/−, ERAP1−/−, and ERAP1+/+ DNA showing distinct band resulting from 29 nucleotide deletion. C, Rat splenocytes and BMDMs from rats with each ERAP1 genotype treated without (−) or with IFNγ (+) for 16 h were lysed, and post-nuclear supernatants subjected to SDS-PAGE and immunoblotting with anti-ERAP1 and GAPDH antibodies.

fS2

Supplemental Figure 2. Effects of ERAP1 deficiency on folded forms of HLA-B27 and HLA-B7. HLA-B allotypes were immunoprecipitated from BMDMs with W6/32, separated by SDS-PAGE, and visualized and quantified by immunoblotting. Representative W6/32 IPs are shown here, with HC10 IPs from the same experiments shown in Fig. 1. The relative expression of folded HLA-B allotypes for each ERAP1 genotype is shown. Quantitative results represent averages from duplicate gels from 2–3 independent experiments with standard error of the mean (SEM) shown by error bars. The values for ERAP1+/+ samples were set to 1. (*, p < 0.05)

fS3

Supplemental Figure 3. Effects of ERAP1 deficiency on cell surface expression of unfolded HLA-B27. Whole cell lysate from a duplicate sample from experiment described in Fig. 2B,C was subjected to immunoprecipitation with HC10. Immunoblot represents “Total” HLA-B27 immunoprecipitated from whole cell lysate, while left immunoblot shows cell surface material (“Surface”) immunoprecipitated from equivalent number of cells.

fS4

Supplemental Figure 4. Effects of ERAP1 deficiency on HLA-B27 folding and misfolding during upregulation. A, BMDMs from HLA-B27 Tg rats with the ERAP1 genotypes shown were treated with IFNγ (100 ng/ml), harvested at the times indicated, and cell lysates used for immunoprecipitation. These supplemental representative images accompany the quantitative results shown in Fig. 3. Left, top; HC10 immunoprecipitates subjected to SDS-PAGE under non-reducing conditions allowing visualization of dimers (D) and oligomers (O) as well as monomers (M). Right, top; W6/32 immunoprecipitates and non-reducing conditions. Bottom, left, and Bottom right; equivalent to top, except SDS-PAGE was performed under reducing conditions. B, representative HC10 and W6/32 immunoprecipitates separated by SDS-PAGE under reducing conditions and immunoblotted for BiP. Quantitative results are shown in Fig. 3B.

fS5

Supplemental Figure 5. Effects of ERAP1 deficiency on HLA-B7 folding during upregulation. A, BMDMs from HLA-B7 Tg rats with the ERAP1 genotypes shown were treated with IFNγ (100 ng/ml) and harvested at the times indicated (in the same experiments shown in Supplemental Fig. 4). Cell lysates were used for immunoprecipitation. Left, top; HC10 immunoprecipitates subjected to SDS-PAGE under non-reducing conditions. Right, top; W6/32 immunoprecipitates and non-reducing conditions. Bottom, left, and Bottom right; equivalent to top, except SDS-PAGE was performed under reducing conditions. B, quantitative results showing ratio of folded to free HCs for HLA-B7 (as shown for HLA-B27 in Fig. 3A). The results represent 2–3 independent experiments. (*, p < 0.05)

fS6

Supplemental Figure 6. H&E stained sections of testicles from non-transgenic WT rat (left) and HLA-B27 Tg rat with scrotal swelling (right) showing inflammatory cell infliltrate.

tS1

Supplemental Table 1. Frequency of arthritis and scrotal swelling in HLA-B27 Tg, HLA-B7 Tg, and WT rats with ERAP1 genotypes as indicated. Table shows the number of affected animals, the total in the cohort, and the percentage (affected/total).