|Home | About | Journals | Submit | Contact Us | Français|
HLA-B27 plays a central role in the pathogenesis of many spondyloarthropathies and in particular ankylosing spondylitis. The observation that the HLA-B27 heavy chain has a tendency to misfold has raised the possibility that associated diseases may belong in a rapidly expanding category of protein misfolding disorders. The synthesis of the HLA-B27 heavy chain, assembly with β2m and the loading of peptide cargo, occurs in the endoplasmic reticulum (ER) before transport to the cell surface. The evidence indicates that misfolding occurs in the ER prior to β2m association and peptide optimization and is manifested in the formation of aberrant inter- and intra-chain disulfide bonds and accumulation of heavy chain bound to the chaperone BiP. Enhanced accumulation of misfolded heavy chains during the induction of class I expression by cytokines, can cause ER stress resulting in activation of the unfolded protein response (UPR).
Effects of UPR activation on cytokine production are beginning to emerge and may provide important missing links between HLA-B27 misfolding and spondyloarthritis. In this chapter we will review what has been learned about HLA-B27 misfolding in human cells and in the transgenic rat model of spondyloarthritis-like disease, considering it in the context of other protein misfolding disorders. These studies provide a framework to support much needed translational work assessing HLA-B27 misfolding and UPR activation in patient-derived material, its consequences for disease pathogenesis and ultimately how and where to focus intervention strategies.
Ankylosing spondylitis (AS) is a complex genetic trait with an estimated four to ten genes responsible for the majority of susceptibility.1 Spondyloarthropathies (SpA) comprise several disorders that are more heterogeneous clinically and where genetic susceptibility is likely to be more complex and variable. Defining genetic loci and ultimately genes that influence susceptibility, is an area of intense investigation. Family-based linkage studies using low-density microsatellite markers have been somewhat disappointing.2,3 However, single nucleotide polymorphism (SNP) identification and mapping has provided a detailed framework on which to perform whole genome association studies. This approach has already provided valuable information on genes involved in susceptibility to other complex genetic diseases4 and studies on AS are now emerging.5 In addition to providing markers that will be useful in identifying AS patients earlier in their disease course, it is anticipated that a more complete picture of genetic susceptibility will inform us on pathways that are important in pathogenesis and identify new therapeutic targets.
Unlike most complex genetic diseases, a single gene (HLA-B) plays a dominant role in AS. The B27 allele contributes as much as 40% of the overall genetic load and is a major factor for many other SpA.6,7 Although the role of HLA-B27 has been the focus of intense investigation for over 30 years, none of the postulated mechanisms has been proven or eliminated.8 While it has been tacitly assumed that a single feature of HLA-B27 is responsible, it is conceivable that this is not the case9 and that the answer to the HLA-B27 conundrum will be even more complex than initially anticipated.
A detailed understanding of pathogenesis requires animal models that phenocopy the human condition and are amenable to genetic manipulation and experimentation,10 combined with translational studies of human material that are confirmatory. The development of animal models has been attempted over the years through the generation of HLA-B27 transgenic mice and rats. Transgenic mouse models of SpA have been disappointing for several reasons. Initially, no spontaneous inflammatory disease was observed11 and although attempts to induce disease with infection revealed some differences in susceptibility,12 the SpA phenotype was not observed. Subsequent studies suggested that HLA-B27 transgenic mice developed spontaneous arthritis if you deleted the endogenous gene for (mouse) β2m.13–15 However, the use of a mixed genetic background may have confounded these studies making reproducibility and thus interpretation problematic.16
Controlling for mixed backgrounds is very difficult and eventual genetic drift can result in loss of the phenotype. Inbred strains of animals are the genetic equivalent of a single human and thus it is not surprising that different genetic backgrounds would influence susceptibility. In humans less than 5% of HLA-B27 positive individuals develop SpA. The whole genome association studies mentioned above are being used to identify other human genes that affect susceptibility. It may be possible to exploit strain differences that are important determinants of disease in HLA-B27 transgenic rodents and use similar genetic approaches to identify the responsible genes. While the rodent genes may not be the same as those found in humans, by definition they will be involved in pathways that are important for pathogenesis of HLA-B27-associated disorders and substantial overlap with pathways identified in human genetic studies would be expected.
The production of transgenic rats expressing HLA-B27 and human β2m (B27/hβ2m) that develop spontaneous inflammation resembling SpA signified a major advance in this area.17 This demonstrated that under certain conditions overexpression of HLA-B27 was sufficient to cause disease, providing the first evidence that the gene product itself was involved. The SpA-like phenotype includes gastrointestinal tract inflammation (e.g., colitis), arthritis and other inflammatory lesions in the skin and testicles. The colitis is highly penetrant, while arthritis is less frequent and depends more on the strain of rat used. Although axial inflammation can occur, it does not recapitulate the spinal inflammation and ankylosis seen in humans.18 However, recently Tran et al. have reported that overexpressing additional hβ2m can alter the phenotype in transgenic rats that already overexpress HLA-B27 and hβ2m.19 High copy number B27/hβ2m transgenics with additional hβ2m develop more severe arthritis and significant axial disease with no apparentchange in colitis. Interestingly, rats with low copy number B27/hβ2m that normally do not develop spontaneous disease, develop axial and peripheral arthritis without colitis, when additional hβ2m is overexpressed.
In this article we will focus on one mechanism that may be the basis for the striking relationship between HLA-B27 and spondyloarthritic diseases. We will explain the general concept and consequences of protein misfolding and then provide a detailed assessment of the special case of HLA-B27 misfolding and how it may be linked to disease through an autoinflammatory stimulus.
Extensive polymorphism at the HLA-B locus results in considerable sequence variation in the HLA-B-encoded heavy chain across the human population. Over 900 alleles have been reported to date, encoding over 800 different proteins (www.anthonynolan.com/HIG/). These amino acid differences affect a number of properties of class I heavy chains, including peptide binding specificity and affinity, T-cell recognition (both as a result of and independent from bound peptide), β2m binding affinity, folding and assembly efficiency and chaperone interaction (e.g., tapasin) (reviewed in ref. 20). There are also polymorphisms in the promoter region of HLA-B at the 5′ end of the gene, which could affect baseline expression level and inducibility.21
Features of HLA-B27 that distinguish it from other alleles have provided the basis for several hypotheses concerning its contribution to disease. It is convenient to classify these ideas based on whether they invoke immunological recognition of some form of the heavy chain versus intracellular effects.8 Immunological recognition by antibodies22 or autoreactive T-cells23 supposes that the form(s) of HLA-B27 being recognized are typical for HLA class I complexes.
More recently the detection of other forms of HLA-B27, such as heavy chain homodimers,24 or unusual unfolded monomers,25 has led to ideas about recognition by leukocyte receptors on NK cells and other leukocytes.26–30 In contrast, the tendency of HLA-B27 heavy chains to misfold in the intracellular compartment known as the endoplasmic reticulum (ER)31,32 has led to the notion that intracellular effects of HLA-B27 might be involved in disease pathogenesis. Misfolding was hypothesized to result in activation of an intracellular stress response pathway known as the unfolded protein response (UPR),33 which has been shown to occur in B27/hβ2m transgenic rats.34,35 The consequences of HLA-B27-induced UPR activation will be discussed in detail later in this chapter. Finally, the observation that cell lines transfected with HLA-B27 but not other alleles exhibit increased bacterial survival36,37 could be important for pathogenesis, particularly in reactive arthritis. Recent evidence suggests that bacterial replication is increased38 and that the p38 MAP kinase pathway may be disrupted.39 This most likely represents an intracellular or at least nonantigen-presenting effect of HLA-B27.40 Experiments using site-directed mutants of HLA-B27 show that the biological effect correlates with heavy chain misfolding, but is not associated with acute UPR activation and therefore the molecular mechanism remains to be defined. It will be important to determine whether the expression of heavy chains that misfold is responsible for this effect, since a related phenomenon has been observed for a mutant of surfactant protein-C that misfolds.41 These authors demonstrated that adaptation to chronic ER stress was associated with modification of an NFκB-dependent pathway, reminiscent of what has been observed in HLA-B27-transfected cells.42
In this chapter we will focus on HLA-B27 misfolding, considering it in the context of other proteins that misfold, the cause of misfolding and more importantly, what we have begun to learn about its consequences.
There has been a tendency to assume that only one hypothesis, or one aspect of the immunobiology of HLA-B27, will eventually be tied to its role in pathogenesis. However, this may not be correct, particularly when one considers phenotypic differences in the diseases associated with HLA-B27 such as reactive arthritis, uveitis, AS and other forms of undifferentiated SpA.
The information required for a protein to fold into its native conformation is contained within its primary sequence, yet a great deal of energy is expended to ensure that this occurs efficiently and without error (reviewed in ref. 43). For multi-subunit proteins or those that transport cargo, the process is even more complex, with many opportunities for errors in the formation of stable, properly folded complexes. It has become increasingly apparent over the last decade that many genetic diseases result from protein misfolding, either due to inherent properties of the mutated gene product, or in some cases as a consequence of abnormalities in the cellular pathways that handle misfolded proteins.
HLA class I complexes of heavy chain, β2m and peptide represent an example of a protein (heavy chain) that transports ‘cargo’ (β2m and peptide) to the cell surface. To perform this function and display self-peptides or pathogen-derived cargo to T-cells during an immune response, HLA class I heavy chains must fold properly, bind β2m and then load peptide prior to exiting the ER compartment (reviewed in ref. 44). High stability of the trimolecular complex is critical for efficient transport through the Golgi, a long half-life on the cell surface and ultimately a productive immune response. The stability of HLA class I complexes is critically dependent on early events in the folding and assembly process, including the formation of two intrachain disulfide bonds.45 The α3 domain folds very rapidly and is stabilized by an intradomain disulfide between Cys-203 and Cys-259. The α1 and α2 domains fold more slowly and this is not complete until peptide is stably bound.46 A second disulfide between the α1 and α2 domains (Cys-101-Cys-164) maintains the integrity of the peptide-binding groove47 as the heavy chain/β2m heterodimer interacts with tapasin, ERp57 and the transporter associated with antigen processing (TAP) to form the peptide loading complex (PLC). Although there are allelic differences in the need to interact with tapasin (and thus the PLC), in general this process facilitates the binding of high affinity peptides. For example, HLA-B27 (the B*2705 subtype) is expressed relatively efficiently in tapasin-deficient cells48 and is frequently referred to as a tapasin-independent allele. However, it interacts with tapasin when present and this affects the peptide repertoire.49 It is possible that the ability of HLA-B27 to be expressed at high levels on tapasin-deficient cells may reflect its tendency to fold slowly and be retained in the ER in a peptide-receptive state without tapasin-PLC interaction. This could favor the binding and optimization of available peptides without tapasin-mediated retention.
ERp57 binds to tapasin via a disulfide (ERp57-Cys-57-Cys-95-Tapasin) and plays an important role in disulfide bond isomerization in the heavy chain during class I assembly.50 Recent evidence indicates that formation of the ERp57-tapasin conjugate prevents ERp57-mediated reduction of the a1-a2 interdomain disulfide in the class I heavy chain, thus maintaining the peptide binding groove in a receptive state.45 When tapasin is missing or mutated at Cys-95 and thus unable to form a complex with ERp57, the class I heavy chain a1-a2 disulfide is reduced until suitable peptide cargo can bind. Free ERp57 (or ERp57-calreticulin complexes) appears to catalyze this reduction in the absence of tapasin leading to the concept that tapasin performs its function by sequestering ERp57.
The final stages of peptide binding to HLA class I molecules includes trimming by the ER aminopeptidase associated with antigen processing (ERAP1).51–56 Peptides appear to be nestled into the peptide-binding groove at their C-terminus first with ERAP1-mediated N-terminal trimming to their final size of 8–11 amino acids. In humans, L-RAP or ERAP2 may also play a role in this process. In addition to peptide trimming for presentation by class I molecules, ERAP1 appears to have another role in the immune system. It was discovered independently as aminopeptidase regulator of TNF receptor (TNFR1) shedding (ARTS-1), but also regulates shedding of IL-6 and IL-1 decoy receptors.57–59
The vast majority of proteins are made in the cytosol, or cotranslationally inserted into the ER in the case of membrane bound and secreted proteins. In these two compartments, there are parallel molecular chaperone systems that assist and monitor the folding process to ensure high fidelity production of proteins that can function properly. When protein folding goes awry, due to mutations or polymorphisms that alter the amino acid sequence, or abnormalities in components of the chaperone systems, misfolding can result (reviewed in ref. 43). The consequences of misfolding depend on the site of synthesis of the protein, the nature and severity of the folding defect, the relative importance of the gene product and whether protein quality control (PQC) processes have intervened sufficiently. Many misfolded and even incompletely folded ER proteins can be eliminated efficiently by ER-associated degradation (ERAD) if they have remained in the ER for a sufficient time. Diseases that ensue are typically due to loss-of-function with classic examples being hemophilia (Factor VIII mutations) and hereditary emphysema (α-1-antitrypsin deficiency)60 (Fig. 1). Gain-of-function phenotypes are more common and more varied. Misfolded proteins that accumulate within (e.g., forming aggresomes or inclusion bodies) or outside the cell (e.g., amyloid fibrils) can be toxic either to the involved cell or surrounding cells. Alpha-1-antitrypsin mutations can also cause pathology due to ER retention, aggregation and mitochondrial injury in the liver,61 providing a striking example of phenotypic variation due to cell-specific differences in the handling of misfolded proteins.
The cellular response to ER protein misfolding referred to as the UPR (unfolded protein response), is part of a more global homeostatic response to ER stress.60 The UPR also plays a key role in ER expansion during the differentiation of certain cell types, such as plasma cells that become highly specialized to produce and secrete large amounts of immunoglobulins.62 The UPR can also contribute to the pathogenesis of certain diseases with the most clear-cut examples being situations where UPR-induced apoptosis results in the loss of important cells, such as pancreatic β-cells in the Akita mouse model of diabetes63,64 or neural tissue in Pelizaeus-Merzbacher disease (proteolipid protein 1 in spastic paraplegia).65 Another interesting example may be idiopathic inflammatory myositis. In certain forms of the disease muscle tissue (myocytes) exhibits robust UPR activation.66 This has been associated with caspase-12 activation and it has been postulated that the UPR plays a role in myositis pathogenesis by promoting apoptosis, although additional mechanisms are possible.66,67 Enforced class I upregulation (H-2Kb) via a tetracycline-regulated transgene driven by a muscle-specific promoter can result in an inflammatory phenotype that recapitulates much of the pathology seen with human disease.68 This is interesting and may represent an example of inappropriate expression of a class I protein, perhaps with insufficient concomitant expression of peptides, β2m and/or other chaperones such as tapasin, leading to ER stress and UPR activation.
While gain-and loss-of-function classification schemes are useful, disease pathogenesis is often complex and may result from more than one consequence of protein misfolding. This is best exemplified by α-1-antitrypsin mutations that result in both types of sequelae.
The first indication that HLA-B27 had a tendency to misfold came from mutagenesis experiments where the entire ‘B pocket’ was changed by substituting residues from the HLA-A2 allele (creating a hybrid referred to as B27.A2B).69 Remarkably, this dramatically altered the folding and assembly characteristics of the heavy chain with B27. A2B behaving more like HLA-A2 and other alleles that exhibit rapid folding and assembly kinetics.31 Evidence that the heavy chain was in fact misfolding, came from experiments looking at where it was being degraded. Normally, HLA class I heavy chains are internalized from the cell surface and broken down in lysosomes. However, a proportion of HLA-B27 heavy chains were found to be dislocated from the ER membrane shortly after synthesis and before becoming associated with β2m (and probably peptide) and then degraded in the cytosol by proteasomes. This ERAD pathway is used to eliminate ER-synthesized proteins that misfold and/or fail to assemble rapidly.70 B27.A2B, as well as other naturally occurring HLA alleles that were examined, did not exhibit this behavior, thus tying misfolding to B pocket composition and the slow folding characteristic of HLA-B27.32 ERAD of HLA-B27, but not the other expressed alleles, was also detected in EBV-transformed human B-cells indicating that it occurs when there is only a single copy of the HLA-B27 gene and is not merely a consequence of overexpression.31
Interestingly, the B pocket was also found to have an unexpected dramatic effect on peptide binding affinity, in addition to its predicted effect on peptide binding specificity.31 Since this pocket binds the side chain of the second amino acid in the peptide (P2), the specificity conferred by the HLA-A2-like B pocket was almost identical to what is found in peptides that bind to HLA-A2 (Leu/Met), rather than the Arg P2 specificity of HLA-B27. However, what was surprising was that HLA-B27 required a 30-fold higher concentration of peptide (on average) to achieve the same half-maximal binding as B27.A2B. This suggests that the folding abnormality exhibited by HLA-B27 may be related to peptide binding. In other words, this allele might require more peptide to achieve release from the PLC and exit from the ER. It would follow that in situations where the supply of peptides into the ER is restricted and/or the synthesis of heavy chains is increased, HLA-B27 folding might be disproportionately adversely affected in comparison to other alleles.
Further exploration of events occurring in the ER for HLA-B27 revealed that the heavy chain has a tendency to form disulfide-linked complexes with itself (and possibly other proteins; unpublished observations) and exhibit prolonged association with the ER chaperone BiP (Grp78/Hspa5).32,71,72 These features also map to the B pocket and are not exhibited by B27.A2B or other naturally occurring alleles examined to date. Further mutagenesis experiments have defined two B pocket residues that are key for HLA-B27 misfolding; Glu-45 and Cys-67 (reviewed in ref. 73). The single substitution of Met for Glu at position 45 restores rapid folding to the HLA-B27 molecule and eliminates the formation of disulfide-linked complexes and prolonged BiP binding (misfolding) even in the presence of Cys-67. Furthermore, the single substitution of Ala for Cys at position 67 also prevents misfolding, even when Glu-45 is intact. These observations suggested a model where two features of the HLA-B27 heavy chain might be required for misfolding to be prominent; slow folding and the ability to form aberrant disulfide-linked dimers via Cys-67.73
While Glu-45 and Cys-67 are not unique to HLA-B27, they are very uncommon among other alleles. In addition, there is a Lys residue at position 70 that has been reported to increase the reactivity of the sulfhydryl group on Cys-67,74 although it has not been studied in the context of misfolding. These three residues (Glu-45, Cys-67 and Lys-70) are virtually unique to HLA-B27,75 (www.anthonynolan.com/HIG/) and thus would support the idea that misfolding is extremely uncommon if not unique to this allele.
Additional support for this model comes from the observation that an imposed reduction in folding rate caused by incubating cells at reduced temperature also exacerbates dimer formation and BiP binding to heavy chains.72 In this study, evidence was provided that Cys-164, in addition to Cys-67, was involved in dimer formation. This observation has several possible implications since Cys-164 is involved in the intrachain disulfide bridge between the α1 and α2 domains of the class I heavy chain (Cys-101-Cys-164), which normally forms quite rapidly after heavy chain synthesis and is important for maintaining the integrity of the peptide-binding groove (see above).
The involvement of Cys-164 residue in oxidative dimerization of HLA-B27 heavy chains is potentially important as it suggests two possible scenarios related to HLA-B27 misfolding. First, if the Cys-101-Cys-164 disulfide bridge forms quickly in HLA-B27 as in other alleles, then it must not be completely protected from reduction/isomerization if it is eventually involved in dimerization, since the latter process requires it to reform a disulfide with another HLA-B27 heavy chain. Since protection of the Cys-101-Cys-164 disulfide from reduction is a key function of tapasin-ERp57,45 the formation of dimers via Cys-164 could reflect HLA-B27 not interacting efficiently with this complex in the ER. Alternatively, it may be that the a1-a2 domain disulfide does not form normally in HLA-B27, making Cys-164 available to enter into an interchain disulfide linkage. Additional studies are needed to fully delineate the earliest events in HLA-B27 folding that lead to misfolding and its cellular consequences.
A major advance toward understanding the role of HLA-B27 in SpA was made in the 1990s when Taurog and colleagues first produced transgenic rats overexpressing HLA-B27 and human β2m (hβ2m) (B27/hβ2m).17 High copy number B27 transgenic rats were found to develop a ‘spontaneous’ inflammatory disease that includes gastrointestinal inflammation (colitis), arthritis, alopecia with psoriasis-like skin lesions, dystrophic nails and testicular inflammation.18 These phenotypic features are only partially penetrant and are variable in frequency with the exception of colitis, which occurs in 100% of transgenics. The arthritis is predominantly peripheral, although rats overexpressing additional hβ2m were shown recently to develop more severe arthritis with axial involvement19 (discussed below). While transgenic rats do not provide a precise phenocopy of human disease, B27/hβ2m transgenics with and without extra hβ2m provide reproducible animal models that can be used to investigate pathogenic mechanisms that are likely to have relevance to human disease. Unfortunately rats are not as amenable to experimental manipulation as mice.
For example, targeted gene deletion is not currently possible due to the lack of embryonic rat stem cells. Production of transgenics is more expensive and labor intensive, ex vivo transduction of bone marrow cells with retroviruses is not readily achievable and many antibodies useful to visualize and/or block the function of mouse proteins are not available for rats. Nevertheless, a great deal has been learned about the cellular requirements for disease in high copy B27/hβ2m transgenic rats (reviewed in ref. 18). HLA-B27 must be expressed in the bone marrow compartment for the colitis/peripheral arthritis phenotype to occur and ubiquitous expression is not necessary.76 In addition, the spontaneous inflammatory disease appears to be mediated by CD4 rather than CD8 T-cells.77,78 The presence of gastrointestinal flora is also required, yet normal flora is sufficient to trigger inflammation, especially bacteroides spp.79,80 These findings have provided strong evidence against a role for arthritiogenic (or ‘colitogenic’) peptides playing a central role in pathogenesis, but rather suggest that HLA-B27-expressing cells arising from the bone marrow are either targeted by CD4 T-cells or alternatively serve as a stimulus for these cells to become pathogenic.
Reports that CD4 T-cells can recognize normal and abnormal forms of HLA-B27 have emerged,81 raising the question of whether this might explain the importance of these cells for SpA-like disease in B27/hβ2m transgenic rats and also be involved in human disease. For human studies, CD4 T-cells were raised by stimulation with T2 cells transfected with HLA-B27.81 T2 cells are missing a large region of the major histocompatibility complex (MHC) including TAP1 and TAP2 genes and thus are unable to transport peptides into the ER from the cytosol. They have been reported to express HLA-B27 homodimers,24 although this was not observed in other studies.32,72
Evidence supporting the idea that CD4 T-cells could recognize HLA-B27 came from experiments using a monoclonal antibody (ME1) that recognizes HLA-B27 and could block recognition. When cells with an intact antigen presentation pathway were used, including patient-derived B-cells, HLA-B27 was poorly recognized. In a follow-up study CD4 T-cells from two more AS patients were raised using similar methods.82 These T-cells failed to recognize HLA-B27 on T2 cells, but instead appeared to be reacting to other HLA class I alleles expressed at low levels on these cells, perhaps presenting peptides derived from degraded HLA-B27 heavy chains. In separate studies, double transgenic mice expressing HLA-B27 and a human T-cell receptor (TCR) that recognizes the HLA-B27-restricted NP383-391 flu peptide, developed CD4 as well as CD8 T-cells capable of recognizing this peptide presented by HLA-B27.83 If CD4 T-cells that can recognize HLA-B27 develop in rats and humans, this could have implications for disease. However, these transgenic mice represent an unusual situation where there is forced expression of the same TCR on every T-cell regardless of the costimulatory CD4/8 molecule and thus the question of whether this might occur with TCRs directed against other alleles needs to be addressed. The possibility that CD4 T-cells capable of recognizing HLA-B27 exist in transgenic rats has not, to our knowledge, been examined.
The observation that HLA-B27 had a propensity to misfold, as defined by the formation of disulfide-linked heavy chains and stable BiP binding, was confirmed and extended in B27/hβ2m transgenic rats.71 Using several transgenic lines with variable transgene copy number and phenotype, Tran et al. demonstrated a quantitative correlation between the biochemical features of HLA-B27 misfolding in splenocytes and the development of SpA-like disease (colitis and arthritis). This correlation was further supported by the absence of disease in HLA-B7 (B7/hβ2m) transgenic rats, consistent with the evidence that this allele does not misfold, even when overexpressed.32
One of the consequences of protein misfolding in the ER can be activation of the UPR (reviewed in ref. 84). Some of the earliest cellular events that mark this response are phosphorylation and activation of PERK (PKR-like ER kinase) and IRE1 (inositol requiring-1) and proteolytic cleavage of ATF6 (activating transcription factor-6). Immediate downstream events include IRE1-mediated splicing of the mRNA encoding XBP-1 (X-box binding protein-1), PERK-mediated phosphorylation of eIF2α (eukaryotic initiation factor 2 α) and increased transcription of UPR target genes (e.g., BiP, CHOP and many others). The transcriptional response is mediated by activated (cleaved) ATF6, ATF4 (produced in response to eIF2a phosphorylation) and the gene product translated from the spliced XBP-1 mRNA (XBP-1s), all of which are active transcription factors.
Several reagents used to measure proximal UPR activation (e.g., antibodies to ATF6 and phosphorylated forms of PERK and IRE1) are not available for rats. Furthermore, since the response is transient, it is more convenient to assess induction of mRNAs encoding BiP and CHOP and splicing of XBP-1 transcripts (XBP-1s). Using these markers, we found that spleen and thymus cells isolated from B27/hβ2m transgenic rats (F344 33.3 line) exhibited little or no evidence of UPR activation.34 Similarly, bone marrow (BM)-derived macrophages from premorbid rats showed minimal differences in BiP mRNA (50% or 1.5-fold increase), whereas when BM macrophages were prepared from animals with disease, a robust UPR was observed (5-fold increase in BiP mRNA and up to a 10-fold increase in CHOP). Microarray analyses revealed the UPR to be accompanied by an interferon (IFN) signature raising the question of whether IFN exposure is causing UPR activation via HLA-B27 upregulation. The converse was also possible: the UPR might cause IFN upregulation. It was conceivable that both events were occurring simultaneously.
Subsequent studies have revealed a dual role for IFNs in UPR activation in BM macrophages from B27/hβ2m transgenic rats. First, BM macrophages expressing HLA-B27 that exhibit no UPR at ‘baseline’ (i.e., without stimulation) will activate the UPR in response to IFN-γ treatment.34,35
This is temporally related to heavy chain upregulation and accompanied by a striking increase in the accumulation of BiP-bound heavy chains and disulfide-linked heavy chain complexes.35 In contrast, IFN-γ does not activate the UPR in cells from nontransgenic (wild type) or B7/hβ2m transgenic animals. (It should be noted that there is low-level BiP induction and XBP-1 splicing (<2-fold increase) with IFN-γ treatment of macrophages from these animals, but the response in B27/hβ2m transgenics is substantially higher.34 IFN-γ has been reported to cause ER stress in oligodendrocytes, but this response was also quantitatively small (~2-fold BiP induction) and required prolonged (48 h) stimulation.85 This is not surprising given that cytokines and other exogenous stimuli can upregulate hundreds of proteins, including membrane bound and secreted proteins that are produced in the ER. This low level UPR is likely part of the normal physiologic response that enables cells to handle the increased load. It is worth emphasizing that exacerbated HLA-B27 misfolding and UPR activation occur in the face of IFN-γ-mediated upregulation of multiple components of the class I assembly pathway including TAP1/2, tapasin, proteasome subunits LMP2 and LMP7, ERAP1 and β2m. This implies that HLA-B27 misfolding and ER stress occur despite an increased source of cargo (β2m and peptide) as well as equipment necessary to load the cargo. This seems paradoxical and could indicate that one or more of these components exacerbate HLA-B27 misfolding, although an alternative explanation is that they may merely be insufficient to prevent misfolding.
When splenocytes are treated with IFN-γ, we see only low-level upregulation of HLA-B27 and minimal UPR activation. Examination of inflamed colon tissue reveals evidence for UPR activation, although the magnitude of increases in BiP and CHOP transcripts are smaller (<3-fold) than observed in isolated cells such as BM macrophages (reviewed in ref. 34). Together, these data indicate that UPR activation occurs in cells and inflamed tissues from B27/hβ2m transgenic rats, is specific for HLA-B27 and is temporally related to and strongly correlated with HLA-B27 misfolding. Macrophages are particularly affected by HLA-B27 misfolding in terms of UPR activation, while splenocytes, whole spleen and whole thymus tissue are not.35 These results are consistent with HLA-B27 upregulation being a key component of robust UPR activation.
In preliminary studies we have observed UPR activation in BM-derived dendritic cells (DCs) from B27/hβ2m transgenic rats treated with IFN-γ, but it does not appear to be as robust as in macrophages. However, since additional stimuli can contribute to class I upregulation and we have not exhaustively examined other cell types, our understanding of the extent of UPR activation in these rats remains incomplete.
The second part of the IFN story, is the question of whether IFN expression is upregulated by the UPR. We found low-level induction of the Type I IFN, IFN-β, in BM macrophages undergoing a UPR, either due to HLA-B27 upregulation or in cells treated with pharmacologic agents (tunicamycin or thapsigargin) that cause ER stress,86 consistent with a previous report of low-level induction in tunicamycin-treated fibroblasts.87 IFN-β has well-recognized autocrine effects at low concentrations,88–90 and thus UPR-induced IFN-β may have immunological consequences including a pro-survival effect on macrophages.91 However, perhaps more important is the response observed when macrophages undergoing a UPR are exposed to ligands for pattern recognition receptors (e.g., Toll-like receptors or TLRs). TLR4 and TLR3 agonists such as LPS and dsRNA, that upregulate IFN-β via the TRIF (Toll-like receptor/IL-1 receptor related adaptor protein inducing IFN-β)-dependent pathway, cause robust synergistic IFN-β production in cells exhibiting ER stress. The synergistic response appears to require XBP-1s, but not PERK or ATF6 activation. These results suggest a fundamental relationship between ER stress and innate immune signaling with implications beyond HLA-B27 and disease, as well as a novel function of XBP-1 in the convergence of these important signaling pathways.
Links between the UPR and cytokine induction have been reported in the literature. IL-6 production from plasma cells after activation by LPS or CD40 ligation is influenced by XBP-1, although this effect is considerably delayed and may be secondary to other changes.62 Macrophages loaded with cholesterol exhibit UPR activation and increased production of TNF-α and IL-6, effects that appear to be secondary to NFκB, JNK1/2, p38 and/or Erk1/2 activation.92 Using a microarray-based screening approach, we identified IL-23p19 (the unique subunit of the active IL-23 cytokine), as being synergistically induced by LPS-treatment of cells with an active UPR (reviewed in ref. 86). We have found IL-23p19 upregulation in inflamed tissue and myeloid cells derived from the tissue, in B27/hβ2m transgenic rats. Il-23p19 is upregulated in a temporal and spatial manner that is consistent with it being involved in the development of colon inflammation. In addition, there is robust upregulation of IL-17 in the inflamed colon that localizes to CD4 T-cells in the lamina propria and draining mesenteric lymph nodes (reviewed in ref. 93). These findings are of interest in the context of several recent developments in our understanding of T-cell biology, as well as new evidence for genes involved in susceptibility to AS.5
Upon antigenic stimulation, naïve CD4 T-cells differentiate into T helper (Th) cells with specialized cytokine production profiles and effector functions. The Th1/Th2 paradigm established over 20 years ago was that Th1 cells produce large quantities of IFN-γ and are essential for clearing intracellular pathogens, while Th2 cells produce IL-4, 5 and 13 and are important for clearance of extracellular organisms and robust humoral immunity.94,95 Key cytokines that drive these two pathways are IL-12 (IL-12p70) and IL-4. IL-12 induces Th1 differentiation through STAT4 activation in T-cells and IL-4 promotes Th2 development through STAT6 and GATA-3 activation, promoting more IL-4 production.96 IFN-γ from an initial innate immune response (e.g., activated NK cells) is also important for activating the T-bet transcription factor through STAT1 signalling, which in turn activates Th1-specific genes.
Recently, a third subset of effector CD4 T-cells characterized by IL-17 production (‘Th17’) has been discovered.97–99 Th17 cells may have evolved as another arm of the adaptive immune response for enhanced protection against extracellular bacteria (i.e., Klebsiella pneumoniae), protozoa and fungi (e.g., Pneumocystis carinii) by recruiting neutrophils. However, additional roles for Th17 in immune defense are possible. What has become very clear, is that Th17 cells play a crucial role in chronic inflammation in animal models of human autoimmune/autoinflammatory diseases such as RA, MS,100,101 IBD102,103 and psoriasis104 and there is growing evidence that IL-17 is a crucial pro-inflammatory cytokine in the human disease counterparts. In addition to IL-17, Th17 cells can produce TNF-α and IL-6.100,102 IL-17 can act on several cell types including macrophages, fibroblasts, endothelial cells and epithelial cells, to upregulate TNF-α, IL-6, IL-1, as well as several chemokines and metalloproteases (including MMP-3 which has been shown to be a good biomarker for AS).105–107 Thus, downstream effects of IL-17 are diverse and highly proinflammatory.
Several cytokines play key roles in Th17 development and the balance between Th17 and regulatory T-cells (Treg) in mice. For example, the combination of TGF-β and IL-6 drives naïve CD4 T-cells to become Th17-committed108,109 through induction of the retinoic acid orphan receptor (RORγt) in naïve T-cells, which then leads to upregulation of the IL-23 receptor (IL-23R).110 IL-23 can then act on Th17-competent cells stimulating robust and prolonged IL-17 upregulation111,112 (and reviewed in ref. 113). In addition, TCR stimulation by MHC class II-restricted antigens can induce IL-17 production without IL-23.
In mice it appears that CD4 T-cells producing IFN-γ (Th1) and IL-17 (Th17) are distinct populations, while in humans CD4 T-cells producing IFN-γ and IL-17 (Th1/Th17) have been documented in the gut of humans with Crohn's disease.114 In addition, the factors that regulate Th17 development in humans appear to be different from mice with IL-23 and IL-1β playing a more important role than IL-6.115 In addition to the predominant form of IL-17 (IL-17A or CTLA-8) produced by CD4 T-cells, there is an extended family with five additional IL-17 molecules whose cellular source and regulation need to be further defined.107 Other cells that have been reported to produce IL-17 include CD8 and gamma/delta T-cells, neutrophils and even macrophages and lymphocytes at sites of infection.
Preliminary results linking HLA-B27 misfolding and the UPR to enhanced IL-23 induction in macrophages in response to TLR agonists, together with evidence for activation of the Th17 axis in transgenic rats, suggests a novel paradigm for the development of HLA-B27-associated colitis (Fig. 2). In the gastrointestinal tract, a low level immune response to bacterial colonization could result in increased expression of IFNs (Type I and/or Type II), perhaps via innate immune stimuli (IFN-β) and/or NK cell activation (IFN-γ). This would result in upregulation of class I expression and, in cells expressing HLA-B27, activation of the UPR. Macrophages would then become sensitized to TLR agonists such as LPS and other bacterial products, polarizing them toward increased production of IFN-β and IL-23 and possibly more IL-6. IL-23 would then drive IL-17 production from CD4 T-cells that have become committed to the Th17 lineage. While IL-6 and TGF-β have been shown to be important for the development of Th17 T-cells,99 there is evidence that cells with the capacity to produce IL-17 are normally present in the colon.102,103 Thus, in this unique mucosal environment, increased IL-23 expression could be a sufficient stimulus for chronicgastrointestinal inflammation. This is supported by the observation that IL-23p19 transgenic mice develop widespread inflammation without any other additional stimulus.116 In the HLA-B27 transgenic rats, increased IFN-β expression might serve to promote HLA-B27 upregulation and also activate NK cells.117 It is also possible that unusual forms of HLA-B27 expressed on the cell surface might engage leukocyte receptors and serve as an activating stimulus for NK cells.30
It is interesting to note that IFN-γ can inhibit the Th17 axis.97,98 In the model we propose for the development of colitis, IFNs would play an important role in promoting IL-23 production via class I upregulation and subsequent UPR activation, but could conceivably inhibit the IL-23/IL-17 axis through effects on Th17 development. We and others, have documented IFN-γ overexpression in the inflamed colon,18,34 but since CD4 T-cells with the capacity to produce IL-17 may already exist in this location, IFN-γ may have little effect on their development. Furthermore, we do not know the relative importance of Type I vs Type II IFNs in HLA-B27 upregulation in rats in vivo, nor whether Type I IFNs have the same inhibitory effect on Th17 development, although this might be expected given the overlap in effects of Type I and Type II IFNs. These and other questions, including the relative importance of the Th1 axis in transgenic rats, need to be further addressed.
There is heterogeneity within the HLA-B27 group of alleles referred to as subtypes (www.anthonynolan.com/HIG/). The numerical classification for subtypes is to designate them with an asterisk preceding the number (e.g., B*2701, B*2702, etc.,). More than 30 subtypes have been reported for HLA-B27 and since most occur infrequently, little is known about their association with AS or SpA. While most of the relatively common subtypes (e.g., B*2705, B*2702, B*2704) have been associated with disease, for some time B*2706 and B*2709 have been thought to be exceptions. Since hypotheses explaining how HLA-B27 might cause disease have been driven by our understanding of how it differs from other HLA-B alleles, it should be possible to refine our ideas based on properties of subtypes differentially associated with disease. However, the caveat with this approach is that incomplete or incorrect information about disease associations may lead to incorrect conclusions. There are now new data suggesting that B*2709 may be associated with disease, or at the very least the situation is more complex than previously thought.118 Patients with B*2709 who developed SpA were reported several years ago,119,120 and now there are reports of this subtype in AS patients.121,122 A recent examination of the existing data suggests that B*2709 occurs in these individuals at a greater frequency than by chance alone,118 thus supporting the idea that this subtype may indeed be associated with disease. The occurrence of B*2709 on a distinct haplotype from B*2705 in the same population,123 along with genetic evidence that additional MHC-encoded genes influence susceptibility,3,124 raises the possibility that the offending alleles are not present on the B*2709 haplotype.123 It is well known that most individuals with HLA-B27 (and B*2705 by inference) do not develop AS/SpA and HLA-B27-positive family members of patients with AS are at much higher risk for disease than the general HLA-B27-positive population. One recently proposed hypothesis is that B*2709 arose by a single mutation from B*2705 on a low-risk haplotype and that it may be the low-risk haplotype that is more important for disease predisposition than the immunobiological differences between the B*2705 and B*2709 proteins.118 Additional genetic differences between populations with B*2709 and B*2705 might also contribute.
The case for a lack of association between B*2706 and disease is more compelling, in part because it is present in a much larger and probably more genetically diverse population.125 However, patients with AS and this subtype have also been reported126 with two additional cases described recently.127
Subtype associations (or lack thereof ) need to be extended to larger populations and investigated for the possible existence of MHC haplotypes such as those uncovered in Sardinia.123
Another pitfall of using the genetic association data to drive hypotheses about disease causation is that subtyping of HLA-B27 has traditionally focused on coding sequence variation, with little attention to the promoter region of the gene. Promoter polymorphisms, which are known to exist,21 could have consequences for baseline and inducible HLA-B27 subtype expression.
The phenotype exhibited by high copy B27/hβ2m transgenic rats, where colitis and peripheral arthritis predominate, does not include an important component of AS—axial inflammation and ankylosis. Recently Tran et al. found that overexpressing more hβ2m by introducing an additional 35 copies of the hβ2m transgene altered the phenotype of high copy B27/hβ2m transgenic rats (55 copies of HLA-B27 and 66 copies of hβ2m).19 Rats with 55 copies of HLA-B27 and 101 copies of hβ2m had more frequent and more severe arthritis involving the axial skeleton, while colitis was not affected. In addition, the extra 35 copies of hβ2m were able to induce arthritis in intermediate copy B27/hβ2m transgenic rats (20 copies of HLA-B27 and 15 copies of hβ2m) that normally remain free of any spontaneous disease. Thus, spondylitis was induced by additional hβ2m even in the absence of colitis. These observations are potentially important as they provide a model system that may be relevant to the pathogenesis of axial inflammation.
The mechanism by which additional hβ2m modifies the phenotype of B27/hβ2m transgenic rats is not clear. Based on observations that the additional hβ2m increased the folding kinetics of HLA-B27, reduced the formation of aberrant disulfide linked heavy chain complexes and resulted in a reduction of BiP mRNA expression (~25–30%) in splenocytes, the authors concluded that while HLA-B27 misfolding was still associated with intestinal inflammation, it was not critical to the development of HLA-B27-associated arthropathy. However, this conclusion is premature, since UPR activation was not examined after upregulation of HLA-B27, which we have shown is critical for this response.34,35 In addition, since there is some cell type specificity to HLA-B27-induced UPR activation, it will be important to examine cells that are likely to be relevant to disease pathogenesis. Preliminary experiments suggest that while the additional hβ2m reduces the magnitude of UPR activation when HLA-B27 is upregulated, it does not eliminate it (unpublished observations) and thus the role of HLA-B27 misfolding in the spondyloarthritis phenotype will require further investigation.
It is also important to consider that UPR activation might be a ‘double-edged’ sword in the pathogenesis of inflammatory disease. Its consequences could depend on the magnitude of the response. For example, it is well known that a strong and unresolved UPR can lead to apoptosis.
If UPR activation in macrophages drives an inflammatory process due to abnormal cytokine production, one could envision downstream effects being different if the cells causing the problem are destined to undergo UPR-induced apoptosis. The consequences of inappropriate in vivo UPR activation in the immune system are relatively unexplored and it is also likely that we do not yet appreciate precisely what needs to be examined. Our ability to approach these questions would be aided greatly by the development of a mouse model, where many more tools are available to address these complex issues.
Recent advances in deciphering genetic susceptibility to AS point toward the IL-23 receptor (IL23R) gene.5 This gene encodes a protein that combines with another subunit IL-12Rβ1 to form the active IL-23 receptor expressed on developing Th17 T-cells,113 making them responsive to IL-23. IL23R polymorphisms have also been implicated in susceptibility to Crohn's disease and psoriasis, other diseases that have phenotypic overlaps with spondyloarthritis.128,129 Preliminary data indicating that HLA-B27 misfolding may be a stimulus for activating the IL-23/IL-17 axis, suggests a novel mechanism that may explain, at least in part, the role of HLA-B27 in colitis in transgenic rats. The striking convergence of the human genetic data and results from HLA-B27 transgenic rats provides a compelling argument that this axis needs to be further examined in SpA and AS.
This manuscript has been previously published: Colbert RA, Delay ML, Layh-Schmitt G, Sowders DP. HLA-B27 misfolding and spondyloarthropathies. In: Molecular Mechanisms of Spondyloarathropathies. López-Larrea, C and Díaz-Peña, R ed. Austin and New York: Landes Bioscience and Springer Science and Business Media, 2009 In Press.
Previously published online as a Prion E-publication: http://www.landesbioscience.com/journals/prion/article/8072