HLA–B27 Alters the Response to Tumor Necrosis Factor α and Promotes Osteoclastogenesis in Bone Marrow Monocytes From HLA–B27–Transgenic Rats




To determine whether HLA–B27 expression alters the response of bone marrow monocytes from HLA–B27/human β2-microglobulin–transgenic (B27-Tg) rats to tumor necrosis factor α (TNFα) and, if so, whether this affects the cells involved in bone homeostasis.


Bone marrow monocytes were treated with RANKL or with TNFα to promote osteoclast formation. Osteoclasts were quantified by counting. Gene expression was measured using quantitative polymerase chain reaction analysis, and protein was detected by enzyme-linked immunosorbent assay, immunoblotting, or immunofluorescence. Effects of endogenously produced cytokines on osteoclast formation were determined with neutralizing antibodies.


TNFα treatment enhanced osteoclast formation 2.5-fold in HLA–B27–expressing cells as compared to wild-type or to HLA–B7/human β2-microglobulin–expressing monocytes. TNFα induced ∼4-fold up-regulation of HLA–B27, which was associated with the accumulation of misfolded heavy chains, binding of the endoplasmic reticulum (ER) chaperone BiP, and activation of an ER stress response, which was not seen with HLA–B7. No differences were seen with RANKL-induced osteoclastogenesis. Enhanced interleukin-1α (IL-1α) production from ER-stressed bone marrow monocytes from B27-Tg rats was found to be necessary and sufficient for enhanced osteoclast formation. However, bone marrow monocytes from B27-Tg rats also produced more interferon-β (IFNβ), which attenuated the effect of IL-1α on osteoclast formation.


HLA–B27–induced ER stress alters the response of bone marrow monocytes from B27-Tg rats to TNFα, which is associated with enhanced production of IL-1α and IFNβ, cytokines that exhibit opposing effects on osteoclast formation. The altered response of cells expressing HLA–B27 to proinflammatory cytokines suggests that this class I major histocompatibility complex allele may contribute to the pathogenesis of spondyloarthritis and its unique phenotype through downstream effects involving alterations in bone homeostasis.

The role of HLA–B27 in the pathogenesis of spondyloarthritis remains poorly understood ([1]). Most hypotheses focus on HLA–B27 as an upstream factor that is envisioned as triggering an aberrant immune response, either adaptive or innate, resulting in chronic inflammation. For example, peptide-loaded HLA–B27/β2-microglobulin (β2m) complexes could be targeted by autoreactive CD8+ T cells ([2]), or β2m-free heavy-chain homodimers might trigger the survival and activation of CD4+ Th17 T cells ([3]). HLA–B27 and human β2m expression in transgenic (B27-Tg) rats has been linked to dendritic cell (DC) dysfunction ([4]) and increased apoptosis ([5, 6]), which could contribute to enhanced DC-mediated activation of CD4+ Th17 T cells ([7]) and/or failure to induce tolerance. In addition, the propensity of HLA–B27 heavy chains to misfold and generate endoplasmic reticulum (ER) stress ([8, 9]) initiates intracellular signals through the unfolded protein response (UPR) that converge with Toll-like receptor (TLR) signaling pathways to promote cytokine production, including interleukin-23 (IL-23) ([10-12]). Systemic IL-23 expression has recently been shown to activate an unusual population of enthesis-resident T cells in mice that produce proinflammatory cytokines, such as IL-17 and IL-22, providing an explanation for some of the unique anatomic features of the spondyloarthritis phenotype ([13]). The relative contribution of these mechanisms to human disease remains an area of intensive investigation.

Despite evidence implicating HLA–B27 as an upstream factor in spondyloarthritis, its tendency to misfold and generate ER stress, particularly when up-regulated ([8, 9, 14]), raises the possibility that HLA–B27 may have a downstream impact on disease pathogenesis by altering the response of cells to proinflammatory cytokines or other factors.

One aspect of spondyloarthritis that is poorly understood is the excessive bone loss prominent in vertebral bodies, juxtaposed with pathologic bone formation that occurs along the spine of many patients with ankylosing spondylitis (AS) ([15-17]). One of the several features of spondyloarthritis that is reproduced in B27-Tg rats is trabecular bone loss in the vertebral bodies and long bones ([18]). Since bone integrity is regulated by osteoclasts and osteoblasts, we sought to determine whether HLA–B27 expression could influence osteoclast formation in response to RANKL and tumor necrosis factor α (TNFα), which are known stimulators of osteoclastogenesis.



Rats were housed in Association for Assessment and Accreditation of Laboratory Animal Care–approved facilities on the Bethesda campus of the National Institutes of Health. All animal experiments were preapproved by the Animal Care and Use Committee at the National Institute of Arthritis and Musculoskeletal and Skin Diseases. Lewis HLA–B*27:05 and human β2m–transgenic (B27-Tg) rats were generated by backcrossing the 33-3 transgene locus from dark agouti (DA) B27-Tg rats onto the Lewis background for >10 generations. The 33-3 transgene locus contains 55 copies of HLA–B27 and 66 copies of human β2m ([19]). Lewis HLA–B*07:02 and human β2m–transgenic (B7-Tg) rat breeders carrying the 120-4 transgene locus were obtained from Joel Taurog (University of Texas Southwestern Medical Center, Dallas, TX). The 120-4 locus contains 26 copies of HLA–B7 and 5 copies of human β2m ([20]). Hemizygous B27-Tg rats and homozygous B7-Tg rats (i.e., 52 copies of HLA–B7 and 10 copies of human β2m) were used in this study. Transgene-negative littermates or Lewis rats purchased from Taconic were used as wild-type (WT) rats.

Isolation of bone marrow–derived monocytes and assessment of osteoclast development

Bone marrow was obtained from the tibias and femurs of euthanized 6–9-week-old rats. Mononuclear cells were isolated using Lympholyte-rat (Cedarlane), washed twice, and resuspended at 5 × 106 cells/ml in high-glucose Dulbecco's modified Eagle's medium (Invitrogen) containing 10% fetal calf serum (Invitrogen), 10–8M 1α,25-dihydroxyvitamin D3 (Enzo Life Sciences), and 20 ng/ml of rat macrophage colony-stimulating factor (M-CSF; PeproTech). Cells were cultured for 2 days at 37°C in a humidified atmosphere containing 5% CO2.

To evaluate the effects of RANKL and TNFα on osteoclast differentiation, nonadherent bone marrow monocytes were collected, adjusted to 106 cells/ml in media, and seeded at 105 cells/well in 96-well plates containing M-CSF (20 ng/ml final concentration) to examine the effects of RANKL and TNFα on osteoclast differentiation. Cytokines were obtained from PeproTech and were used at the final concentrations indicated in the illustrations. Bone marrow monocytes were treated for 3 days with RANKL or for 5 days with TNFα, which preliminary experiments showed were optimal conditions for osteoclast formation. Mature osteoclasts were visualized by staining with tartrate-resistant acid phosphatase (TRAP; Kamiya Biomedical), photographed, and quantitated by counting the number of TRAP-positive cells with ≥3 nuclei, as normalized to the area.


Neutralizing antibodies were goat anti-rat IL-1α (AF500; R&D Systems), hamster anti-rat IL-1β (4-7012; eBioscience), and rabbit anti-rat interferon-β (IFNβ) (CL 9143AP; Cedarlane). All antibodies were used at a final concentration of 1 μg/ml during the entire 5 days of TNFα-induced osteoclast development. Thapsigargin (Sigma-Aldrich) was used at a final concentration of 300 nM and tunicamycin (Sigma-Aldrich) at 10 μg/ml.

RNA isolation and quantitation

For evaluation of gene expression, 106 cells per sample were collected by centrifugation (400g for 5 minutes), lysed in TRIzol (Life Technologies), and stored at −80°C. RNA was isolated according to the manufacturer's protocol. Complementary DNA (cDNA) was generated using iScript (Bio-Rad), and quantitative polymerase chain reaction (qPCR) measurements of specific genes were carried out using a MyiQ cycler with the SsoFast EvaGreen Supermix (Bio-Rad) with primers (500 nM) and cDNA corresponding to 20–50 ng of starting RNA. Three housekeeping genes were used for normalization of target gene expression levels: peptidylprolyl isomerase, hypoxanthine guanine phosphoribosyltransferase 1, and TATA box binding protein. Fold changes were calculated by comparing the normalized Ct values for the treated samples to the normalized Ct values for the untreated samples using qBasePlus software (Biogazelle). For determination of Xbp1 messenger RNA (mRNA) splicing, reverse transcription–PCR products were separated on 3% agarose gels and were visualized with ethidium bromide (Bio-Rad) under ultraviolet light. Table 1 lists the primer sequences that were used.

Table 1. Sequences of primers used in the present analyses*
GeneForward (5′→3′)Reverse (5′→3′)

Immunoblotting and immunoprecipitation

Bone marrow monocytes (2.5 × 107 cells) were harvested and resuspended in 1 ml of ice-cold phosphate buffered saline with 10 mM methyl methanethiosulfonate (Thermo Scientific), on ice for 20 minutes, to prevent spontaneous sulfhydryl bond formation and reduction. Cells were then spun down and lysed (500 μl/107 cells) for 30 minutes on ice in 20 mM Tris, pH 7.8, 100 mM NaCl, 10 mM EDTA, 1% Triton X-100, and Halt protease inhibitor cocktail (Thermo Scientific). Nuclei were removed by centrifugation at 13,000g for 5 minutes, and supernatants were collected and used for immunoprecipitation or sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE) and immunoblotting.

For SDS-PAGE, lysates were diluted 1:1 with Tris–glycine SDS sample buffer (Novex; Invitrogen) with or without NuPAGE sample reducing agent (Invitrogen), and subjected to electrophoresis on 4–20% Tris–glycine gels (Novex) followed by immunoblotting. The following primary antibodies were used: HC10 to detect HLA class I heavy chains, rabbit anti-BiP (ab21685; Abcam), and mouse anti-GAPDH (sc-32233; Santa Cruz Biotechnology). Appropriate species-specific secondary antibodies were alkaline phosphatase–conjugated (SouthernBiotech), and BCIP/nitroblue tetrazolium substrate (Thermo Scientific) was used to visualize proteins. HLA–B27 immunoprecipitation from 500 μl of cell lysate was performed as described previously ([21]). Immunoprecipitates were subjected to SDS-PAGE and immunoblotting as described above, except that HLA class I heavy chains were detected using 3B10.7 antibody ([22]).


Immunofluorescence staining was used to identify BiP and HLA–B27–expressing cells during osteoclast development. Four days after the addition of RANKL or TNFα to bone marrow monocytes, cells were fixed with 4% formaldehyde, washed twice with BD Perm/Wash buffer (BD Biosciences), and then incubated in the same buffer for 15 minutes at 4°C to permeabilize the cells. Primary antibodies (anti-BiP; ab21685 or HC10 [23]) were added at a concentration of 1 μg/ml, and cells were incubated for an additional hour at 4°C. Cells were then washed 3 times with Perm/Wash buffer and incubated in the dark for 30 minutes at 4°C with secondary antibodies (Alexa Fluor 594–labeled chicken anti-rabbit IgG [Invitrogen] or fluorescein isothiocyanate–labeled chicken anti-mouse IgG [LifeSpan Biosciences]) diluted 1:100 in Perm/Wash buffer. Cells were washed again and visualized with an Axiovert Zeiss 40 CFL fluorescence microscope at 20× magnification.

Statistical analysis

Results are expressed as the mean ± SD, except where indicated otherwise. Statistical analysis was performed using Student's t-test. P values less than 0.05 were considered significant. For ratios, the mean and 95% confidence intervals are shown.


Osteoclast formation promoted by HLA–B27 expression

To determine whether HLA–B27 has an influence on osteoclast development, bone marrow monocytes from WT and B27-Tg rats were differentiated with RANKL or TNFα. We found no difference in the number of osteoclasts that formed in the presence of RANKL (Figures 1A and B), although we observed a trend toward larger osteoclasts with more nuclei in the presence of HLA–B27 (Figure 1A) and HLA–B7 (results not shown). In contrast, TNFα treatment consistently resulted in greater osteoclast formation in bone marrow monocytes derived from B27-Tg animals (Figures 1C–F). Representative experiments are shown in Figures 1C–E, with the average fold change from several experiments shown in Figure 1F. The average increase in osteoclastogenesis in HLA–B27–expressing cells treated with TNFα was 2.5-fold (Figure 1F). Experiments with bone marrow monocytes derived from B7-Tg rats used as a control for HLA class I overexpression revealed no difference in osteoclast formation compared with WT rats (Figures 1E and F), indicating the effect is specific for HLA–B27. Osteoclasts formed from bone marrow monocytes derived from WT and B27-Tg rats were active and exhibited similar resorption patterns on calcium phosphate–coated slides, indicating that they were functional (data not shown).

Figure 1.

Augmentation of tumor necrosis factor α (TNFα)–induced osteoclast formation by HLA–B27. A–D, Examination of tartrate-resistant acid phosphatase (TRAP) staining of bone marrow monocytes and numbers of osteoclasts. Bone marrow monocytes from wild-type (WT) rats and from HLA–B27/human β2-microglobulin–transgenic (B27-Tg [B27]) rats were treated with macrophage colony-stimulating factor and increasing concentrations of either RANKL for 3 days (A and B) or TNFα for 5 days (C and D). A representative TRAP staining experiment (A and C) is shown for each treatment. Original magnification × 5. Osteoclasts (B and D) were defined as TRAP-positive cells with ≥3 nuclei per unit area. Values are the mean ± SD of triplicate wells from a representative experiment. = P < 0.05 versus WT rats. E, Comparison of TNFα-induced osteoclast formation in bone marrow monocytes from WT rats, HLA–B7/human β2m–transgenic (B7-Tg [B7]) rats, and B27-Tg rats. Results are from a representative experiment. Each data point represents a single rat. Values are the mean ± SD of 5 rats per group. P < 0.05 for the comparison of B27-Tg rats versus B7-Tg rats or versus WT rats. F, Fold increase in the number of osteoclasts (OCs) induced by TNFα (30 ng/ml) in bone marrow monocytes from B7-Tg rats and B27-Tg rats as compared to those from WT rats. Values are the mean and 95% confidence intervals of 3 (B7-Tg rats) and 6 (B27-Tg rats) independent experiments. Color figure can be viewed in the online issue, which is available at http://onlinelibrary.wiley.com/doi/10.1002/art.38001/abstract.

The roles of IL-1α and IFNβ in HLA–B27–induced osteoclast formation

IL-1α and IFNβ have been shown to regulate osteoclast formation. IFNβ is a potent inhibitor ([24]), while IL-1α promotes TNFα-induced osteoclastogenesis ([25]). To determine whether these cytokines were involved in mediating the effect of HLA–B27 expression on TNFα-induced osteoclastogenesis, each cytokine was blocked with a neutralizing antibody. Blocking IL-1α completely inhibited the effect of HLA–B27 on TNFα-induced osteoclastogenesis, whereas an IL-1β–neutralizing antibody had no effect (Figure 2A). The addition of HC10, which recognizes HLA–B27 homodimers as well as free heavy chains, also had no effect on osteoclast formation (Figure 2A).

Figure 2.

Interleukin-1α (IL-1α)–dependent enhancement of TNFα-induced osteoclast formation in HLA–B27–expressing cells and inhibition by interferon-β (IFNβ). A, Numbers of osteoclasts in bone marrow monocytes from WT and B27-Tg rats following TNFα and neutralizing antibody treatments. Neutralizing antibodies (1 μg/ml) against IL-1α (α–IL-1α), IFNβ (α-IFNβ), IL-1β (α–IL-1β), or HLA–B class I heavy chains (HC10) were added to bone marrow monocytes from WT and B27-Tg rats during TNFα-induced (30 ng/ml) osteoclastogenesis. Osteoclast formation was quantified as described in Figure 1. B, Effect of IL-1α (0.1 ng/ml) on TNFα-induced (30 ng/ml) osteoclast formation in bone marrow monocytes from WT rats. C, Analysis of IL-1α production in culture supernatants from TNFα-treated (30 ng/ml for 3 days) bone marrow monocytes from WT and B27-Tg rats, as measured by enzyme-linked immunosorbent assay. D and E, Expression of Il1a (D) and Ifnb1 (E) mRNA in bone marrow monocytes from WT and B27-Tg rats. Cells were treated with TNFα (30 ng/ml) for the indicated times and mRNA expression was measured. Results are representative of a minimum of 3 independent experiments, each performed in triplicate. Values are the mean ± SD. = P < 0.05 versus WT rats. See Figure 1 for other definitions.

To test whether additional IL-1α was sufficient to promote osteoclastogenesis, IL-1α was added to TNFα-treated bone marrow monocytes from WT rats, where it was found to enhance osteoclast formation ∼2-fold (Figure 2B). HLA–B27–expressing cells produced more immunoreactive IL-1α protein (Figure 2C) and exhibited greater induction of Il1a mRNA in response to TNFα (Figure 2D), a finding consistent with this cytokine being responsible for the effect of HLA–B27 on osteoclastogenesis.

In contrast to the results with IL-1α, neutralizing IFNβ further enhanced osteoclastogenesis in cultures of HLA–B27–expressing cells (Figure 2A). This effect was associated with a several-fold induction of Ifnb1 mRNA in HLA–B27–expressing cells following exposure to TNFα, whereas WT and HLA–B7–expressing bone marrow monocytes exhibited only a small increase (Figure 2E). We have not yet found a reliable enzyme-linked immunosorbent assay for rat IFNβ, and thus were unable to quantitate protein production in bone marrow monocyte cultures. Taken together with the results for IL-1α, these data suggest that HLA–B27–expressing cells produce more IL-1α and IFNβ, with opposing effects on osteoclastogenesis, although under these experimental conditions, the net result is greater osteoclast formation.

TNFα-induced ER stress in HLA–B27–expressing bone marrow monocytes

To explore differences between bone marrow monocytes from B27-Tg and WT rats that might contribute to greater TNFα-induced osteoclastogenesis, we compared the effects of TNFα and RANKL on HLA–B27 expression and misfolding. TNFα treatment resulted in 4-fold up-regulation of HLA–B27 mRNA (data not shown), with a comparable increase in heavy chain expression (Figures 1A, B, and E), while RANKL had no effect (Figures 3A, B, and E). TNFα-induced HLA–B27 heavy chains accumulated as disulfide-linked dimers and multimers, as well as conventional monomers (Figure 3A). There was also an increase in coimmunoprecipitation of BiP with HLA–B27 heavy chains (Figure 3B).

Figure 3.

HLA–B27 expression, misfolding, and activation of the unfolded protein response in cultures of bone marrow monocytes from WT and B27-Tg rats. Cells were treated with macrophage colony-stimulating factor (20 ng/ml) in the presence or absence of RANKL (100 ng/ml) or TNFα (30 ng/ml) for 3 days (A, B, and E) or for 20 hours (C and D) and were then harvested. A, Cell lysates were subjected to nonreducing sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE), and HLA–B27 heavy chains were visualized by immunoblotting with the antibody HC10. The lowest band represents monomers. The higher molecular weight material constitutes dimers and multimers. B, HLA–B27 heavy chains were immunoprecipitated (IP) using HC10, subjected to SDS-PAGE under reducing conditions, and then visualized by immunoblotting (IB) with 3B10.7 antibodies (α-HLA). Coprecipitating BiP was visualized by immunoblotting with antibodies (α-BiP). C, BiP mRNA expression relative to untreated cells is shown. Values are the mean ± SD of 3 experiments. = P < 0.05 versus WT rats. D, Xbp1 mRNA splicing in untreated and RANKL- or TNFα-treated cells is shown. Tunicamycin (TM)–treated cells served as a positive control. The spliced form of Xbp1 mRNA (Xbp1s) runs slightly faster than the unspliced form (Xbp1u) due to the removal of 26 nucleotides by activated inositol-requiring enzyme 1α. E, Immunoblots of cell lysates were subjected to SDS-PAGE under reducing conditions, and BiP, HLA–B27, and GAPDH (loading control) were visualized with specific antibodies (α-BiP, α-HLA [HC10], and α-GAPDH, respectively). See Figure 1 for other definitions.

The accumulation of disulfide-linked HC10–reactive forms of HLA–B27 with enhanced BiP binding in response to TNFα is indicative of heavy-chain misfolding and is reminiscent of previous results demonstrating that IFNγ up-regulation of HLA–B27 exacerbates misfolding in macrophages ([9]). This prompted us to look for evidence of ER stress. HLA–B27–expressing bone marrow monocytes exhibited increased Bip mRNA and protein and Xbp1 mRNA splicing (Figures 3C–E) when treated with TNFα, but not RANKL. These data demonstrate that HLA–B27–expressing bone marrow monocytes exhibit ER stress and activate the UPR when treated with TNFα, but not RANKL.

ER stress–up-regulated IL-1α and IFNβ.

To determine whether ER stress alone is sufficient to induce IL-1α and IFNβ expression, bone marrow monocytes from WT rats were treated with thapsigargin, an inhibitor of sarcoplasmic/endoplasmic reticulum Ca2+ ATPase (SERCA), which prevents autophagosome–lysosome fusion and induces ER stress ([26]). UPR activation resulting in Bip mRNA up-regulation and Xbp1 splicing was readily apparent within 3 hours of treatment of the cells (Figures 4A and B) and was accompanied by strong up-regulation of Il1a and Ifnb1 mRNAs (Figures 4C and D).

Figure 4.

Endoplasmic reticulum (ER) stress–induced Il1a and Ifnb1 mRNA in bone marrow monocytes from wild-type (WT) rats. Bone marrow monocytes derived from WT rats were treated with thapsigargin (TG; 300 ng/ml) for 3 hours or 6 hours. A, Expression of BiP mRNA relative to untreated cells. B, Xbp1 mRNA splicing. C and D, Expression of Il1a (C) and Ifnb1 (D) mRNA relative to untreated cells.

Cellular localization of TNFα-induced HLA–B27 and BiP

To determine whether bone marrow monocytes (osteoclast precursors) or mature osteoclasts were exhibiting effects of TNFα on HLA–B27 expression and UPR activation, bone marrow monocytes that had been treated with RANKL or TNFα were assessed by immunofluorescence microscopy for BiP and HLA class I heavy-chain expression. Increased BiP and HLA class I heavy-chain staining was noted only with TNFα, and not RANKL, treatment and was most prominent in small round/oval cells with single nuclei (bone marrow monocytes), rather than large osteoclasts with multiple nuclei (Figure 5). In contrast, RANKL had no effect on HLA–B27 or BiP staining in cells from B27-Tg rats (Figure 5). These data are consistent with the immunoblot analyses shown above (Figure 3) and further indicate that the predominant effect of TNFα on HLA–B27–induced ER stress is in the osteoclast precursors.

Figure 5.

HLA–B27 and BiP expression in bone marrow monocytes from HLA–B27–transgenic rats, as determined by immunofluorescence. Bone marrow monocytes were treated for 3 days with macrophage colony-stimulating factor (M-CSF; 20 ng/ml) alone (NT), with M-CSF plus RANKL (100 ng/ml), or with M-CSF plus tumor necrosis factor α (TNFα; 30 ng/ml). Cells were fixed with formaldehyde, permeabilized, and then stained with antibodies against HC10 (α-HLA) or BiP (α-BiP). Fluorescein isothiocyanate– or Texas Red–labeled secondary antibodies were used to detect HC10 and anti-BiP, respectively. White arrows indicate monocytes showing increased HC10 and BiP staining with TNFα treatment; light blue arrows indicate an osteoclast. Original magnification × 20.


The propensity of HLA–B27 to misfold, generate ER stress, and activate the UPR has the potential to exert biologic effects in many different cell types that express class I major histocompatibility complex (MHC) molecules. We previously documented the effects of HLA–B27 misfolding in rat macrophages, where UPR activation after up-regulation of HLA–B27 with IFNγ promoted the increased expression of IL-23 and IFNβ in response to TLR agonists ([1, 8, 10]). In earlier work, we found no biochemical evidence, such as heavy-chain oligomerization or prolonged BiP binding, to suggest that HLA–B7 misfolds ([1]). Furthermore, in those studies, up-regulation of HLA–B7 did not activate the UPR, even with heavy-chain expression comparable to that of HLA–B27 ([8, 10]). Since these earlier comparisons were made with IFNγ-treated macrophages, we also examined HLA–B7–expressing monocytes treated for 20 hours with TNFα (e.g., as in Figures 2D and E) and found no UPR activation (data not shown). Thus, comparable overexpression in rat cells of another HLA–B allele that is closely related to HLA–B27 but does not misfold does not appear to cause ER stress, at least under the conditions examined to date.

Here, we show that HLA–B27–expressing rat monocytes cultured with M-CSF and stimulated with TNFα alone also exhibit ER stress and UPR activation linked to HLA–B27 misfolding, with enhanced expression of IL-1α and IFNβ, while these effects are not seen with HLA–B7. More importantly, the autocrine/paracrine effects of IL-1α result in enhanced differentiation of monocytes into osteoclasts. Interestingly, the effect of TNFα is almost doubled when IFNβ is neutralized (Figure 2A). Taken together, these results indicate that HLA–B27–expressing rat monocytes produce more pro- and anti-osteoclastogenic cytokines when stimulated with TNFα, with IL-1α promoting the effects of TNFα and IFNβ attenuating the stimulatory effect of IL-1α. The counterregulatory effect of IFNβ is insufficient to prevent the pro-osteoclastogenic effect of IL-1α, at least under the experimental conditions used in the current study.

ER stress induction of IFNβ has been shown to occur via the UPR transcription factor X-box binding protein 1 (XBP-1) ([10]). The active transcription factor produced from the spliced form of Xbp1 mRNA binds to an enhancer to promote transcription of the Ifnb1 gene ([27]). The greatest effect of ER stress on Ifnb1 mRNA induction is seen when macrophages are exposed to TLR-4/3 agonists that already induce IFNβ, such as lipopolysaccharide and double-stranded RNA ([10, 27]). However, even in the absence of TLR agonists, HLA–B27–expressing rat macrophages exhibit low-level (∼3-fold) Ifnb1 mRNA up-regulation when stimulated with IFNγ ([1]). Our current results extend these previous findings in an important way by showing that TNFα can act as an inducer of low-level IFNβ expression in monocytes experiencing ER stress due to HLA–B27 expression. This response is fundamentally different from that seen in WT or HLA–B7–expressing cells, where Ifnb1 mRNA is increased only 2–4-fold in response to TNFα (Figure 2E). Moreover, our results establish that even a low level of IFNβ produced from HLA–B27–expressing cells can be biologically significant as an inhibitor of osteoclastogenesis.

The effects of ER stress on the production of IL-1α have not, to our knowledge, been reported previously. We had observed Il1a mRNA induction in a pilot experiment using tunicamycin (data not shown), which activates the UPR by inhibiting protein glycosylation. We show here that thapsigargin has a similar effect on Il1a mRNA, indicating that Il1a induction is a consequence of ER stress and is unlikely to be due to some other effect of HLA–B27 expression. We have not further investigated the mechanism of Il1a up-regulation, although enhanced ER stress–induced NF-κB and/or activator protein 1 activation could contribute. Interestingly, ER stress has been shown to activate the NLRP3 inflammasome and increase IL-1β production by a mechanism that is independent of classic UPR activation ([28]). However, in contrast to the IL-1α findings, we did not see increased IL-1β production by HLA–B27–expressing rat monocytes (data not shown), so inflammasome activation seems unlikely to be playing a role.

Opposite effects of IL-1α and IFNβ on osteoclast development have been documented previously ([24, 25]). Kobayashi et al ([25]) demonstrated that IL-1α was required for the formation of fully functional osteoclasts induced by TNFα in a RANKL/RANK-independent mechanism. The inhibitory effect of IFNβ on osteoclast formation was demonstrated in the context of RANKL-induced osteoclast formation, where small amounts of IFNβ induced by RANKL and signaling through an autocrine/paracrine loop inhibited c-Fos expression, which is essential for osteoclast formation ([24]). We have not further investigated the mechanism by which IFNβ inhibits TNFα-induced osteoclastogenesis in HLA–B27–expressing rat bone marrow monocytes, but since TNFα also induces c-Fos, it is plausible to expect that the mechanism would be similar. It is important to note that in our system, TNFα-induced IFNβ does not appear to be limiting osteoclast formation in bone marrow monocytes from WT rats, since there was little difference with versus without anti-IFNβ (Figure 2A) and little IFNβ mRNA induction in WT or HLA–B7–expressing cells (Figure 2E).

Trabecular bone loss resulting in osteopenia and osteoporosis is a significant problem in the spondyloarthritides (SpA), particularly AS ([17, 29-31]). Bone loss often begins early in the disease process, prior to physical immobilization ([15]). Osteoclastic bone resorption occurs in the sacroiliac and zygapophyseal joints of patients with AS ([32, 33]) and is even present in those with longstanding disease, where osteoclasts can be found adjacent to fibrous tissue ([34]). It is most severe in the spine, where it increases the risk of vertebral fractures, and in many individuals, it is anatomically juxtaposed to areas of abnormal bone formation, such as syndesmophytes. TNFα inhibition significantly increases bone mineral density in SpA ([35]) and AS ([36]), implicating this cytokine in bone loss, while the effects of TNF inhibitors on syndesmophyte growth remain unclear. Indeed, whether or not trabecular bone loss is coupled with bone formation is the subject of much debate ([37]).

B27-Tg rats carrying the 33-3 transgene locus and serving as a model for SpA-like disease also exhibit substantial loss of trabecular bone in the vertebral bodies and large bones of the leg, as well as reduced bone strength ([18]). Bone loss in this model occurs primarily as a consequence of increased bone resorption, rather than reduced bone formation ([38]), and is associated with increased RANKL expression relative to that of osteoprotegerin ([39]). TNFα also contributes to the inflammatory disease phenotype in B27-Tg rats, although a role in bone loss has not been established. Our results raise the possibility that HLA–B27 expression contributes to TNFα-induced bone loss in this model and in human SpA, including AS, by promoting the local production of IL-1α.

Since these transgenic rats have multiple copies of the HLA–B27 transgene, they overexpress HLA–B27 heavy-chain relative to that in human cells. Although results with HLA–B7 indicate that heavy-chain misfolding is required to generate ER stress in this model, it may still occur more readily than in humans because of its greater expression. It is noteworthy that Feng et al ([14]) recently showed that peripheral blood mononuclear cells (PBMCs) from HLA–B27–positive AS patients, but not healthy controls, generate sufficient ER stress to activate the UPR when treated with IFNγ. In previous studies, HLA–B27–expressing PBMC-derived macrophages from AS patients did not exhibit UPR activation when treated with IFNγ ([40, 41]). This may have been due to a lack of ([40]), or insufficient ([41]), up-regulation of HLA–B in macrophages derived ex vivo. It is also possible that monocytes and macrophages respond differently to cytokine treatments. Indeed, monocyte-to-macrophage differentiation is itself associated with UPR activation, which in turn, protects cells against further ER stress ([42]). It will be important to better characterize the response of PBMCs and isolated monocytes derived from HLA–B27–positive individuals to cytokines that up-regulate class I MHC molecules, including TNFα.

In conclusion, we have demonstrated that HLA–B27 alters the response of rat bone marrow–derived monocytes to TNFα and promotes osteoclastogenesis by causing ER stress and enhancing IL-1α production. This also increases the production of the anti-osteoclastogenic cytokine IFNβ. While the net effect under the conditions studied here was to promote osteoclastogenesis, this balance could be altered in vivo by other mediators of inflammation. These findings suggest that this unusual class I MHC molecule may exert downstream effects in the pathogenesis of spondyloarthritis, with implications for bone remodeling and the unique spondyloarthritis phenotype.


All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Dr. Colbert had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study conception and design. Layh-Schmitt, Yang, Kwon, Colbert.

Acquisition of data. Layh-Schmitt, Yang, Kwon, Colbert.

Analysis and interpretation of data. Layh-Schmitt, Colbert.


We thank Kristina Zaal and Evelyn Ralston (National Institute of Arthritis and Musculoskeletal and Skin Diseases, National Institutes of Health, Bethesda, MD) for their help with the immunofluorescence microscopy.