The Quest for Psoriasis Autoantigens: Genetics Meets Immunology in the Melanocyte

In a 1979 study of the effects of cyclosporin A on rheumatoid arthritis, Mueller and Herrmann (1979) reported profound improvements in the skin lesions of “control” subjects with psoriatic arthritis. This key observation focused attention on the potential role of T cells in psoriasis, because lymphocytes are relatively specific targets of cyclosporin A (Schreiber and Crabtree, 1992). Although the role of T cells in the pathogenesis of psoriasis is now widely accepted based on a wide array of genetic, immunologic, and pharmacological evidence (Greb et al., 2016), the mechanism by which the immune system is triggered in psoriasis has remained a puzzle. In this issue, two articles are devoted to the nature of autoantigens in psoriasis. One of them, from Jörg Prinz in Munich (Prinz, 2017), provides context for a recent study from his laboratory (Arakawa et al., 2015), asserting that skin melanocytes can proteolytically process a protein naturally expressed by melanocytes, called ADAMTSL5 (A Disintegrin and Metalloproteinase with Thrombospondin Motif-Like 5). This process ultimately results in the formation of a nonameric peptide with the sequence “VRSRRCLRL,” which is capable of forming a ternary complex with the HLA-Class I molecule HLA-C*0602 and a specific T-cell receptor (TCR) derived from pathogenic CD8+ T cells from a psoriasis skin lesion. This clone was identified as “pathogenic” based on its expansion in a psoriatic lesion (Kim et al., 2012).

Similar to immunoglobulin genes, TCR genes contribute to antigen specificity in immune responses via somatic recombination of constant (C), variable (V), diversity (D), and joining (J) regions, with additional variability provided by error-prone DNA recombination and the interaction of two TCR subunits (either α and β, or γ and δ) (Qi et al., 2014). The particular heterodimeric TCR identified by Prinz et al. comprises the TCR α and β chains Vα3S1 and Vβ13S1. Once identified, Prinz et al. engineered the expression of this specific heterodimeric TCR in hybridoma cells capable of yielding a fluorescent readout and then, through a synthesis of multiple strategies, showed that it reacts to VRSRRCLRL (Arakawa et al., 2015). On the basis of Prinz’s findings, Yin et al. in Wuhan, China, assessed T-cell responses to VRSRRCLRL in peripheral blood mononuclear cells (PBMCs) from six patients with progressive psoriasis, and studied the interactions of this peptide with HLA-C*0602 at 2.5 Å resolution (Wei et al., 2017). They demonstrated that amino acid residues at positions 2, 7, and 9 of the VRSRRCLR peptide are “buried” in the antigen binding groove of HLA-C*0602 (arginine, leucine, and leucine, respectively), whereas the other residues present a largely positively charged surface that is highly “exposed” to the TCR (see Figure 1 of Wei et al., 2017).

A distinctive feature that has emerged from the study of HLA associations in psoriasis is that the HLA-C*0602 allele, when considered as a whole, explains the genetic association better than variation in any individual amino acids in the peptide-binding groove (Okada et al., 2014). This is not the case for other diseases such as HIV (International HIV Controllers Study et al., 2010) and rheumatoid arthritis (Raychaudhuri et al., 2012), in which the effects of specific amino acid positions in the HLA peptide-binding groove appear to dominate over the effect of any “classical” HLA allele. The structural studies of Wei et al (2017) provide a possible explanation for this distinction, by suggesting that certain residues of the VRSRRCLRL peptide provide secure and specific binding to the HLA-C*0602 peptide binding groove, whereas others provide a stable interaction surface with a disease-relevant TCR.

Although peptides provide the critical interface between HLA molecules on the one hand, and the TCR on the other, these peptides must be generated from intact proteins. The 2015 study from the Prinz lab (Arakawa et al., 2015) provided the key insight that the ADAMTSL5 protein containing this peptide could be processed by melanocytes to yield the immunoreactive VRSRRCLRL peptide. Prinz (2017) notes that this finding is critically important because HLA Class I molecules present antigens generated from intracellular proteins, and therefore the immune response should be directed against the producing cell. This observation was backed up by the demonstration of CD8+ T cells interacting with MART-1+ melanocytes in psoriasis lesions, albeit not in a cytotoxic fashion (Arakawa et al., 2015). The importance of HLA Class I antigen presentation in psoriasis is further emphasized by the existence of epistatic (gene-gene) interactions between the ERAP1 and ERAP2 genes encoding the endoplasmic reticulum endopeptidase (ERAP1/2) on the one hand, and specific HLA Class I alleles on the other. This occurs not only in psoriasis, but also in the other two major HLA Class I-associated diseases: Behçet’s disease (HLA-B51-associated) and ankylosing spondylitis (HLA-B27-associated) (Lopez de Castro et al., 2016). ERAP1/2 “trims” peptides of 12 to 14 amino acids generated by the proteasome to an optimal length of 8 to 10 amino acids required for HLA Class I binding. Presumably, melanocytes are capable of proteasomal digestion of ADAMTSL5 to yield these longer peptides, which are then further “trimmed” by ERAP1 and/or ERAP2 to yield VRSRRCLRL.

Although the Prinz lab found that cell types other than melanocytes (i.e., keratinocytes and fibroblasts) could present VRSRRCLRL in the context of HLA-C*0602 when the peptide was added exogenously, they noted that keratinocytes and fibroblasts were incapable of endogenously activating the Vα3S1/Vβ13S1 TCR, even when they were transfected with HLA-C*0602 (Arakawa et al., 2015). This interesting observation leaves open the question of whether keratinocytes and fibroblasts express ADAMTSL5. Using three different commercially available polyclonal antibodies from rabbit, goat, and chicken to stain normal and psoriatic skin, Krueger et al. recently reported increased ADAMTSL5 reactivity in lesional keratinocytes, dendritic cells, and T cells in addition to melanocytes (Bonifacio et al., 2016). However, in a subsequent report from the same laboratory utilizing only the chicken antibody, expression was largely confined to dendritic cells, macrophages, and some T cells in the dermis (Fuentes-Duculan et al., 2017). Transcriptome analyses from our laboratory found that ADAMTSL5 mRNA is expressed at low but detectable levels in immortalized keratinocytes (0.5–0.8% of the ribosomal “control gene” RPLP0; Lambert et al., 2017; Stoll et al., 2016) but at even lower levels in normal and psoriatic skin (0.1% and 0.06% of RPLP0, respectively; Li et al., 2014). Using a different “control gene,” Prinz et al. reported that ADAMTSL5 mRNA was expressed at higher levels in melanocytes and melanoma cells than in psoriatic skin (Arakawa et al., 2015). Together, these findings suggest that keratinocytes are not a major source of ADAMTSL5. In any case, ADAMTSL5 is a secreted protein and hence could be produced by one cell type, and taken up by another, such as antigen-presenting cells and T cells (Fuentes-Duculan et al., 2017). However, as noted by Prinz, this scenario would not be compatible with HLA Class I antigen presentation in the context of our current understanding of this process.

The limited specificity of HLA-disease associations has lent an air of mystery to HLA studies of psoriasis and other immune-mediated diseases since their inception in the 1970s. In other words, there is far from a necessary and sufficient relationship between a specific HLA type and disease status. For instance, only approximately 41% of psoriasis cases in our case-control sample are HLA-C*0602 positive (i.e., carriers of at least one copy of HLA-C*0602) (Okada et al., 2014), and only approximately 10% of HLA-C*0602 carriers develop psoriasis.¹ This lack of a “necessary and sufficient” relationship extends to the interaction of HLA-C*0602 protein with the VRSRRCLRL peptide. Thus, although Arakawa et al. (2015) found that VRSRRCLRL stimulated IFN-γ and/or IL-17 production by PBMCs in 26 of 47 patients, approximately the same proportion of the peptide-responsive patients were HLA-C*0602 negative as were HLA-C*0602 positive. Moreover, in both the German and Chinese studies, not all of the HLA-C*0602-positive patients were responsive to the VRSRRCLRL peptide (3 of 5 in the Chinese study, vs. 11 of 32 in the German study). It is possible that HLA-C molecules other than HLA-C*0602 may present ADAMTSL5 to pathogenic T cells, and conversely, that some HLA-C*0602-positive patients present peptides other than VRSRRCLRL to one or more responsive TCRs, whether they be Vα3S1/Vβ13S1 or otherwise.

Of particular note, although the VRSRRCLRL peptide could increase the percentage of CD8+ T cells expressing IL-17 and/or IFN-γ in PBMCs in a substantial proportion of psoriatic individuals, this phenomenon was not observed in any of 11 normal controls, irrespective of their HLA-C*0602 status (Arakawa et al., 2015). Because these findings depend on very small differences in the percentages of responding cells, they deserve confirmation in much larger numbers of cases and controls whose HLA-C*0602 status is known. Nevertheless, the results suggest that disease-related factors over and above HLA-C*0602 status, whether genetic, physiologic, or both, impact the PBMC response to VRSRRCLRL. What are these factors? First of all, it is important to note that the genetic evidence for major histocompatibility complex-related genetic association in psoriasis is not limited to HLA-C. We and others have identified residual associations between psoriasis and specific amino acid residues in HLA-A and HLA-B, and to a lesser extent in HLA-DR, even after controlling for the major genetic effect of HLA-C (Feng et al., 2009; Okada et al., 2014). Moreover, we found that amino acid 45 of HLA-B appears to be an important determinant for development of psoriatic arthritis in patients with purely cutaneous psoriasis, an effect that persists after controlling for known psoriatic arthritis–versus purely cutaneous psoriasis–assisted HLA-B alleles (Okada et al., 2014). Even outside the major histocompatibility complex, bioinformatic analysis of the numerous (over 60) psoriasis genetic signals identified over the past decade strongly supports the notion that “the psoriasis genome is an immunome” (Tsoi et al., 2017). More specifically, there is evidence for epistasis between HLA-C*0602 and genetic variation not only in ERAP1/2, but also in the LCE3 genes mapping to the epidermal differentiation complex, in both European-origin and Chinese populations (Tsoi et al., 2015; Yin et al., 2013). Although the identities of the genes responsible for the identified psoriasis genetic signals largely remain to be identified, it seems likely that their effects may help to explain why despite stimulation by a relevant peptide antigen (VRSRRCLRL) in the presence of HLA-C*0602, normal individuals do not appear to manifest a CD8+ T-cell response to the putative autoantigen (Arakawa et al., 2015).

The complexity of the HLA-antigen-TCR interaction in psoriasis does not end with genetics or with the details of ADAMTSL5 expression, processing, and antigen presentation. Previous studies have revealed several other attractive candidate psoriasis autoantigens, including “hyperproliferative” keratin peptides cross-reactive with streptococcal M-protein (Sigurdardottir et al., 2013); the innate host defense protein LL-37 (Lande et al., 2014); other streptococcal cross-reactive proteins including ezrin, maspin, peroxiredoxin 2, heat shock protein 27, and keratin 6 (Besgen et al., 2010). Other candidate psoriasis antigens, such as streptococcal proteoglycans (Baker et al., 2006), are not human proteins but rather microbial products. Nevertheless, such candidate microbial products are plausible psoriasis “autoantigens” by virtue of their ubiquity and persistence within the human host, even though they are not of human genetic origin. Other important candidates are lipid antigens generated by the phospholipase A2 encoded by PLA2G4D and presented by the HLA Class I-like molecule CD1a (Cheung et al., 2016). Although all these candidates are intriguing, here we focus on the first two because they both contain peptides having structural similarity to VRSRRCLRL (Wei et al., 2017), and they both manifest differential responses between psoriasis cases and controls when used to stimulate PBMCs.

Guttate psoriasis is strongly associated with streptococcal pharyngitis (Sigurdardottir et al., 2013) and with HLA-C*0602 (Mallon et al., 2000). M proteins are filamentous components of the streptococcal outer membrane, of which more than 80 serotypes are known (Fischetti, 1991). All M proteins have conserved amino acids with significant homology to keratins, particularly keratins K16 and K17 (Gudmundsdottir et al., 1999; McFadden et al., 1991; Sigmundsdottir et al., 1997). These keratins are strongly upregulated in psoriasis lesions but are either not expressed in normal skin or only at low levels at sites of predilection (i.e., elbows, knees, scalp) (Leigh et al., 1995). CD8+ T cells from HLA-C*0602-positive patients respond to peptide sequences common to K17 and M protein, whereas nonpsoriatic HLA-C*0602-positive controls only respond to M protein peptides (Johnston et al., 2004). Moreover, skin-homing (CLA+) T cells were enriched 10-fold for K17/M protein cross-reactivity (Johnston et al., 2004). The other VRSRRCLRL-related candidate autoantigen is LL37, an antimicrobial peptide produced by neutrophils and keratinocytes (Kahlenberg and Kaplan, 2013), which complexes with nucleic acids to promote activation of plasmacytoid dendritic cells (Lande et al., 2007). Reminiscent of the results for the VRSRRCLRL peptide (Arakawa et al., 2015), PBMCs responded to LL-37 with T-cell proliferation in nearly half (24 of 52, or 46%) of patients with psoriasis, as opposed to 0 of 35 healthy controls (Lande et al., 2014). Unlike ADAMTSL5, LL-37 appears to be markedly overexpressed by keratinocytes in psoriasis, yet both antigens appear in antigen-presenting cells and T cells (Fuentes-Duculan et al., 2017). This suggests that both melanocytes and keratinocytes may be capable of provoking a skin-focused T-cell response compatibly driven by antigen-presenting cells with the phenotypic manifestations of psoriasis (Kim and Krueger, 2017). However, if the same or similar candidate antigens are secreted and circulating between cells (rather than being produced by the target cells of CD8+ T-cell attack), we could expect an antigen-specific expansion of CD4+ T cells, a topic that merits further investigation.

Although the expansion of CD4+ and CD8+ T cells with specific TCR rearrangements has been noted for over 20 years in psoriasis (Kim et al., 2012), a great deal remains to be learned about the antigen(s) driving these expansions and the role of genetic variants in controlling their activation in normal versus psoriatic individuals. With the advent of improved methods for single-cell transcriptome analysis, which can capture the paired TCR αβ and γδ subunits in hundreds to thousands of T cells from a single psoriasis plaque (Redmond et al., 2016), we can expect that there will be a number of exciting insights into this long-standing puzzle in the years to come.

Clinical Implications.

• A “ternary complex” of HLA, antigen, and T-cell receptor interact to promote “immunologic memory.”

• HLA-C is an HLA Class I protein that presents peptides derived from intracellular proteins to specific T-cell receptors on CD8+ T-cells. HLA-C*0602 is a particular “version” of HLA-C that expressed by HLA-C*0602, which is an allele of the HLA-C gene.

• Melanocytes express and process ADAMTSL5 to a specific peptide that is presented to CD8+ T cells in the context of HLA-C*0602.