The Quest for Better Understanding of HLA-Disease Association: Scenes from a Road Less Travelled By

Abstract

Dozens of human diseases and health traits are significantly more common among individuals carrying particular human leukocyte antigens (HLA) alleles. The underlying mechanism of this phenomenon, commonly referred to as “HLA-disease association”, has been the subject of a decades-long debate. The prevailing hypotheses implicate an auto-aggressive immune response due to aberrant presentation of self-, self-mimicking-, or altered self-antigens by HLA molecules. However, the identity of such putative antigens remains elusive in the vast majority of HLA-associated diseases. Moreover, antigen presentation-based hypotheses are difficult to reconcile with epidemiologic data and functional characteristics of HLA molecules. To provide better answers to these inconsistencies an alternative theory involving allele-based, antigen presentation-independent mechanism is proposed here. Recent research findings in rheumatoid arthritis, an emblematic HLA-associated disease, lend support to the proposed theory.

“Two roads diverged in a wood, and I -

I took the one less traveled by,

And that has made all the difference.”

~ Robert Frost (1874–1963)

Introduction

Glycoproteins encoded by genes of the major histocompatibility complex (MHC) – known in humans as HLA (human leukocyte antigens) – specialize in presentation of short peptides to T lymphocytes and play a key role in the body’s immune defense. However, it has long been observed that some allelic variants of certain HLA genes appear to betray their assigned duty and, paradoxically, facilitate certain diseases, many of which involve immune-mediated tissue damage. How does the MHC, a critical immune-policing mechanism become a wrongdoer is unclear. Here, I briefly review some of the prevailing hypotheses concerning the mechanism of MHC-disease association, and discuss an emerging new paradigm.

HLA Molecules

HLA is a large genetic region that encompasses over 200 highly polymorphic genes that encode for class I and class II antigens, as well as other immune system proteins. Class-I and class-II HLA gene products are expressed on the cell surface. Their best-characterized role is to present antigenic peptides to T lymphocytes. In order to selectively target invading microorganisms and other foreign antigens, the immune system must discriminate between self and non-self. To that end, a process called “MHC restriction” (Zinkernagel & Doherty, 1997) has evolved. This restriction is accomplished by an intricate thymic selection process, which eliminates self-reactive T cell clones, while preserving those that recognize foreign peptide antigens presented by self-MHC. At the end of this selection, distinct CD8 and CD4 T lymphocyte clonal populations are produced with restricted recognition of peptides presented only by self class I or class II MHC molecules, respectively. Class I molecules specialize in presentation of endogenous antigens to cytotoxic T lymphocytes, while class II molecules present exogenous antigens to helper T lymphocytes.

Many of the key principles concerning antigen presentation, including MHC restriction, had been reasonably well defined by the mid-1970s, but the structural basis of antigen presentation by MHC molecules had remained elusive. A series of X-ray crystallography studies published between the late-1980s and early-1990s (Bjorkman et al., 1987a; Bjorkman et al., 1987b; Brown et al., 1988; Brown et al., 1993) catapulted the field forward. In a seminal study, Bjorkman et al. resolved the tri-dimensional structure of a human class I HLA molecule, HLA-A2, (Bjorkman et al., 1987a; Bjorkman et al., 1987b). That study revealed that the α1 and α2 domains of the HLA class I heavy chain fold to make up an oval-shaped groove that accommodates 8–10 amino acid-long peptides. The peptide-binding groove is flanked by two parallel α helical structures on its sides. Interestingly, despite a substantial evolutionary distance between class I and class II HLA genes, and the fact that in class I HLA molecules the peptide-binding groove is coded by a single gene, whereas in class II molecules it is formed jointly by the products of two distinct genes, the tri-dimensional conformations of class I and class II molecules are remarkably similar (Brown et al., 1993).

HLA-Disease Association

Population studies carried out over the last several decades have identified a long list of human diseases that are significantly more common among individuals that carry particular HLA alleles. For example, more than 90% of Caucasian patients with ankylosing spondylitis carry particular class I HLA alleles (e.g. HLA-B*27:02, HLA-B*27:05) (Reveille, 2006). Narcolepsy, a brain disorder characterized by sleep abnormality and falling attacks (cataplexy), is an illustrative example of HLA class II-associated disease, in which 90–100% of Caucasian patients carry the DQB1*06:02 allele (Mignot et al., 1994). In type I diabetes mellitus, more than 90% of patients carry either HLA-DRB1*03/DQB1*02:01, or HLA-DRB1*04/DQB1*03:02 gene haplotypes, compared to only 40% of controls (Erlich et al., 2008). Rheumatoid arthritis (RA) is another emblematic HLA class II-associated disease (reviewed in Holoshitz, 2010). Approximately 90% of Caucasian seropositive RA patients carry one or two HLA-DRB1 alleles (e.g. DRB1*04:01, DRB1*04:04, DRB1*04:05, DRB1*01:01, and a few others) that code for a sequence motif in the DRβ chain called “shared epitope” (SE).

The mechanism underlying HLA-disease association is unclear. This question has captivated the imagination of immunologists for decades, and generated many hypotheses, which fall into two general categories: 1) Those that blame ‘mistaken identity’, in which an HLA allele appears to associate with the disease, although the actual culprit belongs to a different locus in the haplotype, or associates through linkage disequilibrium, and 2) Those that implicate immune reactivity to self-antigens due to aberrant T cell repertoire selection (Nepom & Kwok, 1998; Ridgway & Fathman, 1998), immune cross-reactivity with foreign antigens (Oldstone, 1998) or immune attack on “altered self” antigens (Yin et al., 2013).

HLA haplotypes – combinations of genes at adjacent loci that tend to be inherited together – can indeed explain some misinterpreted associations, as exemplified by narcolepsy, where an apparent association with HLA-DRB1*15 was later found to be actually attributed to an HLA-DQ allele in the haplotype, HLA-DQB1*0602 (Ellis et al., 1997). Linkage disequilibrium means a non-random association of alleles at two or more loci. This phenomenon has been implicated in hereditary hemochromatosis, where an apparent association with HLA-A alleles was later determined to be actually due to two point mutations in a non-classical class I HLA gene, HFE, that are found in linkage disequilibrium with HLA-A3 and HLA-A29 alleles (Ajioka et al., 1997; Cardoso et al., 2002). However, compelling as these examples may be, both haplotype-based association and linkage disequilibrium are rather uncommon events in HLA-disease association. In the majority of diseases, the data support a role for the associated HLA alleles themselves. Importantly, neither haplotype-based association nor linkage disequilibrium identifies a bona fide mechanistic basis for HLA-disease association; they merely direct the blame at another gene. As such, these explanations do not provide testable hypotheses to explain the pathogenic mechanism of HLA molecules in disease etiology.

The three other hypotheses listed in the second category above implicate, in one way or another, antigen presentation by HLA molecules. They postulate that the basis of HLA-disease association is immune response to putative self or foreign antigens, either in their native or modified form. Given the known role of MHC molecules in antigen presentation, these hypotheses have resonated well among immunologists. Unfortunately however, despite its plausibility, presentation of specific antigens as a mechanism of HLA-disease association is difficult to reconcile with a considerable body of scientific data.

First, inconsistent with antigen-specific immune recognition, the cause-effect fidelity between HLA alleles and particular diseases is often challenged. For example HLA-DRB1*04:01 best known for its association with RA, associates with type-1 diabetes as well (Tait et al., 1995). Similarly, HLA-DQB1*06:02 increases disease risk in both narcolepsy (Mignot et al., 1994) and multiple sclerosis (Serjeantson et al., 1992); HLA-DQB1*03:02 associates with both type-1 diabetes (Sabbah et al., 1999) and celiac disease (Setty et al., 2008). Thus, an individual HLA allele can promiscuously associate with distinct diseases that do not share pathogeneses, target tissues or putative antigens. Second, contrary to what could be expected from an antigen specific-based mechanism, allele-associated diseases can demonstrate species non-specificity. For example, HLA-DRB1*04:01, which associates with human RA, confers susceptibility to inflammatory arthritis in mice as well (Taneja et al., 2007), and the SE, a sequence motif found in the vast majority of RA patients, is also associated with susceptibility to inflammatory arthritis in dogs (Ollier et al., 2001). Such ‘trans-species susceptibility’ is difficult to reconcile with MHC-restricted antigen presentation. Third, lymphocyte clonality, commonly expected in antigen-specific immune responses, has not been yet convincingly demonstrated in HLA-associated diseases, and in the majority of such diseases the identity of the putative target antigen is unknown, despite decades-long research effort. Fourth, significant HLA-allele-based associations have been observed in conditions that do not involve antigen recognition, or any immune-based pathogenesis. For example, the most significant class II HLA-disease association documented to date has been found in narcolepsy (Nishino et al., 2010), a neurotransmitter-mediated brain disorder that is not known to involve antigen presentation. Moreover, MHC associations have been shown to exist with certain traits, such as cognition (Shepherd et al., 2004), that do not involve any known immune basis, let alone antigen presentation. Finally, antigen presentation-based hypotheses cannot easily explain allele-dose impact on disease severity, or offer a plausible explanation for allele-dose effects on concordance rates in monozygotic twins (Jawaheer et al., 1994).

Thus, HLA-restricted antigen presentation, a common thread in most of the prevailing hypotheses addressing HLA-disease association, appears to be inharmonious with many observations concerning the biology, epidemiology and evolution of HLA molecules.

The MHC Cusp Theory – A Road Less Travelled By

Seeking better answers to the question of HLA-disease association, we have recently put forward a heterodox, MHC Cusp theory (de Almeida & Holoshitz, 2011). At the focus of this theory is a polymorphic region on HLA molecules whose tri-dimensional cusp-like shape appears to have been preserved in the entire MHC family. The theory proposes that the MHC codes for allele-specific ligands in the cusp region, which interact with non-MHC receptors and activate various pathways. Aberrations in those pathways could cause MHC-associated diseases.

The rationale behind the MHC Cusp theory is based on structural and functional evidence. As discussed above, the crystal structures of class I and class II HLA molecules demonstrate remarkable tri-dimensional conservation despite their considerable evolutionary distance. One of the most conspicuous features in the MHC fold is a prominent cusp-shaped prominence in the α2 domain of class I HLA molecules, and its tri-dimensionally-equivalent β1 domain in class II HLA molecules. The cusp in both molecules involves allele-diversity regions. Similarly shaped structures have been preserved in the entire MHC gene product family, irrespective of whether or not they can present antigens (Rudolph et al., 2006). Importantly, cusp regions in several MHC molecules have been found to act as ligands that perform non-antigen presentation functions. For example, ligands for natural killer cell receptors have been identified in the cusp region of both classical and non-classical (HLA-E) class I HLA molecules (Boyington et al., 2000; Kaiser et al., 2008). Similarly, in HFE (an empty-grooved HLA class I-like molecule), a transferrin receptor-binding ligand has been mapped to the cusp region (Bennett et al., 2000). Additionally, in M10 (an empty-grooved mouse class I-like MHC molecule), the cusp region has been proposed as a pheromone receptor-binding site (Olson et al., 2005), and the SE, located in the HLA-DR cusp-region, has been recently identified as a calreticulin (CRT)-binding, signal transduction ligand (Ling et al., 2007b).

Thus, the cusp region, whose peculiar tri-dimensional shape has been carefully conserved through MHC evolution independent of antigen presentation, is a hub for signal transduction ligands that interact with a variety of non-MHC receptors and activate important biologic functions. In class II and classical class I HLA molecules, the cusp contains allelic hypervariable regions. As discussed above, antigen presentation per se cannot explain HLA-disease association. Accordingly, rather than assigning the blame exclusively to the antigen presentation function of HLA molecules, the MHC Cusp theory posits that HLA molecules may promote diseases due to their auxiliary allele-specific, yet antigen presentation-independent, biologic effects. Recent findings concerning the functional role of the SE in RA lend support to the theory.

The SE: an HLA-DRB1-Coded Arthritogenic Cusp Ligand

RA is a chronic inflammatory disease, which afflicts millions of individuals worldwide. The disease is multi-systemic, but its best-characterized aspect is chronic joint inflammation and articular bone destruction, mediated by activated osteoclasts (OCs) (Gravallese et al., 2000). The etiology of RA is unclear, although genetic factors, especially the HLA-DRB1 locus (Stastny, 1978), play a major role. Genotypic analyses have revealed that susceptibility to RA is closely associated with a number of HLA-DRB1 alleles that code for SE – a five amino acid sequence motif in residues 70–74 of the HLA-DRβ1 chain (Gregersen et al., 1987). The more common SE-coding alleles (and their coded amino acid sequence motif) include: HLA-DRB1*04:01 (QKRAA); HLA-DRB1*04:04, HLA-DRB1*04:05, HLA-DRB1*04:08, HLA-DRB1*01:01, and HLA-DRB1*01:02 (QRRAA); HLA-DRB1*10:01 (RRRAA). Importantly, SE-coding HLA-DRB1 alleles not only confer disease susceptibility, they are also associated with earlier onset of arthritis and more severe bone damage (Gonzalez-Gay et al., 2002). Another intriguing feature of SE-RA association is that the extent of bone damage correlates positively with the number of SE-coding HLA-DRB1 alleles (Weyand & Goronzy, 1994).

The prevalent hypotheses concerning the mechanism of SE-RA association postulate that the SE motif may permit presentation of arthritogenic antigens (Wucherpfennig & Strominger, 1995), or interfere with T cell repertoire selection with resultant emergence of T cell clones that recognize such antigens (Bhayani & Hedrick, 1991). However, data supporting a direct pathogenic role for antigen-specific immune responses in RA are inconclusive. Additionally, the cause-effect fidelity between SE and RA is not stringent, as several non-RA human diseases and experimental animal models of autoimmunity have been shown to associate with SE-coding alleles as well (discussed by De Almeida & Holoshitz, 2011). Furthermore, as discussed above, antigen presentation-based mechanisms are difficult to reconcile with SE allele-dose effects.

The SE is a 5 amino acid sequence motif in the third allelic hypervariable region, near the tip of the HLA-DR β1 domain cusp. Searching for an allele-based - yet antigen presentation-independent - mechanistic basis for SE-RA association, in recent years my laboratory has examined whether the SE acts as a signal transduction ligand, consistent with its cusp location. The decision to take this road came about rather fortuitously. Studying the cytolytic activity of human γδ T cell clones against EBV-transformed B cell lines, we were surprised to discover that these highly potent cytolytic cells, which can kill indiscriminately all malignant or transformed cells, were selectively unable to kill SE-positive targets. The enigmatic resistance was later found to be due to higher constitutive levels of nitric oxide (NO) in SE-positive targets (Ling et al., 2006). That seminal finding compelled us to explore the role of SE-coding HLA-DRB1 alleles using cDNA transfection, followed by studies with synthetic peptides and recombinant proteins expressing the SE sequence motif (Ling et al., 2007a). In other studies we have identified cell surface CRT – a known innate immune system receptor – as the SE-binding receptor (Ling et al., 2007b), and mapped the SE binding site on that receptor (Ling et al., 2010). SE signaling effects and identification of CRT as the signal transducing receptor have been reviewed in detail elsewhere (de Almeida et al., 2011; Holoshitz et al., 2010; Holoshitz & Ling, 2007).

More recently we demonstrated that engagement of cell surface CRT by the SE ligand activates lineage-dependent functional consequences. For example, in CD8⁺CD11c⁺ dendritic cells, the SE ligand inhibits indoleamine 2, 3 deoxygenase, an enzyme known to play an important role in regulatory T (Treg) cell activation. In CD8⁻CD11c⁺ dendritic cells, the SE triggers production of IL-6 and IL-23, cytokines known to be involved in activation and expansion of IL-17-producing T (Th17) cells. The end result of these two complementing effects is a potent SE-activated Th17 polarization, both in vitro and in vivo (De Almeida et al., 2010).

Th17 cells are believed to play an important pathogenic role in RA (Shahrara et al., 2008) and can activate OCs (Sato et al., 2006). Given that, and the known association between the SE and erosive disease, we have undertaken to investigate whether the SE has a direct pro-osteoclastogenic effect. The rationale for this study was further strengthened by the fact that the SE is a potent activator of NO and reactive oxygen species (ROS) (Ling et al., 2006; Ling et al., 2007a), signaling molecules which have been previously shown to affect OC activation (Lee et al., 2005; Rahnert et al., 2008). We have discovered that the SE ligand facilitates OC differentiation and functional activation. Importantly, when administered in vivo, the SE ligand enhanced erosive bone damage in collagen-induced arthritis in mice (Holoshitz et al., 2013).

We have recently put the SE ligand hypothesis to a higher level of scrutiny by studying biostable small molecular SE-mimetic ligands. The rationale behind this endeavor was driven by a search for biostable compounds that would more closely resemble the tri-dimensional conformation of the SE in its physiologic HLA-DR context, and would be more effective in vivo compared to the linear 15-mer peptidic SE ligands used in prior studies. A library of such compounds has been recently synthesized using a backbone cyclization method (Naveh et al., 2012). A prototypic SE-mimetic member of that library, containing the SE primary sequence motif QKRAA, was characterized both in vitro and in vivo. This SE-mimetic compound interacted strongly with the SE receptor CRT, potently activated NO and ROS production and markedly facilitated OC differentiation and function in vitro, with potencies 100,000 to 1,000,000-fold higher than those of a linear SE peptidic ligand. When administered in nanogram doses to mice with collagen-induced arthritis, the SE-mimetic ligand enhanced Th17 cell expansion, increased arthritis severity, enhanced OC abundance in synovial tissues and facilitated bone destruction (Fu et al., 2013, in press).

The majority of individuals carrying SE-coding HLA-DRB1 are perfectly healthy, consistent with the consensus in the field that the SE does not cause RA, but rather increases the risk of disease development. The low RA concordance rates among identical twins (Silman et al., 1993) and epidemiologic surveys (Gabriel et al., 1999) together suggest that in addition to genetic factors, environmental influences play critical roles in RA disease onset. In this context, it is worth mentioning that: 1) Anti-citrullinated protein antibodies, a useful diagnostic marker in RA, are closely associated with SE-coding HLA-DRB1 alleles (Klareskog et al., 2008); 2) Biochemical analyses have revealed over-abundance of citrullinated proteins in RA (Yamada et al., 2005); 3) There is strong synergistic effect between SE positivity and cigarette smoking on RA susceptibility (Karlson et al., 2010); 4) Cigarette smoking has been proposed as a potential cause of protein citrullination (Makrygiannakis et al., 2008). Given all that, it is intriguing that we have recently demonstrated that citrullinated CRT, which is over-abundant in RA synovial tissues, displays higher affinity for the SE ligand and transduces more potent SE-activated signaling (Ling et al., 2013). The pathogenic significance of these findings is unclear, but if confirmed in an experimental disease model, these results could provide a mechanistic model of gene-environment interaction in RA.

Thus, using the SE as a test case, our signal transduction data, receptor binding analyses, cell differentiation and activation studies, and in vivo experiments, collectively support the premise of the MHC Cusp theory, by suggesting that in addition to their role in antigen presentation, HLA molecules can contribute to the pathogenesis of autoimmune diseases through allele-coded ligands that activate aberrant signaling events with resultant pathology.

Evolutionary Perspectives

Bearing in mind that teleological arguments are acceptable only when used as a basis to formulate testable hypotheses, in the remaining paragraphs I venture to address the question “why”. In other words, what has been the “purpose” of the substantial energy invested in order to preserve the tri-dimensional structure of the cusp region through evolution, and why has this endeavor produced, paradoxically, detrimental health outcomes?

My proposed answer to the first question is that a protuberant α helical structure may be critically needed to allow effective physical interaction between cusp ligands and their cognate receptors. In this context, it is worth commenting that the modern MHC is likely a descendent of archaic self/non-self-discrimination systems, which, in turn, likely evolved from basic cell-cell contact systems. The theory proposed here argues that the cusp antigen presentation-independent functions may have been preserved through evolution due to their impact on fundamental biologic functions, such as natural killer cell regulation (in classical and non-classical class I HLA molecules), iron metabolism (HFE), mating (M10), or immune auxiliary functions (HLA class II molecules).

Consistent with recent findings concerning the functional role of the SE, it could be argued that a trans-species polymorphic region in class II HLA molecules has been preserved in its cusp-like shape throughout evolution because of its important role in immune auxiliary function, such as T cell polarization. It is also conceivable that at least some cusp region effects may have been evolutionarily conserved thanks to their ability to productively mimic more archaic innate immune stimulants, such as bacterial products. At this juncture it is worth pointing out that the signaling and functional effects of the SE described by us are intriguingly similar to previously reported effects of muramyl dipeptide.

N-acetylmuramic acid L-alanine D-isoglutamine (LD-MDP), a building block of bacterial cell wall, is the minimal structure required for the adjuvanticity of mycobacteria in complete Freund’s adjuvant (Ellouz et al., 1974). Relevant to the focus of this review, MDP demonstrates many functional similarities to the SE ligand (Table 1). For example, similar to the SE, MDP activates production of IL-6 and TNFα (Sanceau et al., 1990; van der Meer et al., 2009), increases NO (Palacios et al., 1992) and ROS (El-Khoury et al., 2011) production, activates OCs (Dewhirst, 1982; Dziak et al., 1982), enhances Th17 differentiation (Manni et al., 2011) and, importantly, facilitates inflammatory arthritis in rodents (Koga et al., 1986; Koga et al., 1980; Kohashi et al., 1986; Zidek & Masek, 1985). Finally, MDP has been reported to bind to the SE receptor, CRT (Chen et al., 2004). Using a cutting-edge biolayer interferometry technique, we have recently confirmed that MDP and CRT do interact physically. Moreover, virtual receptor-ligand docking simulations predicted that the SE ligand and MDP interact with overlapping binding sites on CRT (Figure 1). It is therefore tempting to propose that the SE ligand mimics the adjuvant effects of MDP. Could such mimicry be an underlying mechanisms by which bacterial adjuvants can trigger autoimmunity (Rose, 2010)? Obviously, this question needs to be examined experimentally.

Table 1.

Functional similarities between the SE ligand and MDP

	Reference
Function	SE ligand	MDP
Th17 polarization	(De Almeida et al., 2010), (Holoshitz et al., 2013)	(Manni et al., 2011)
OC activation	(Holoshitz et al., 2013)	(Dewhirst, 1982),(Dziak et al., 1982)
TNFα, IL-6	(De Almeida et al., 2010)	(van der Meer et al., 2009), (Sanceau et al., 1990)
NO	(Ling et al., 2006)	(Palacios et al., 1992)
ROS	(Ling et al., 2006)	(El-Khoury et al., 2011)
CRT binding	(Ling et al., 2007b), (Ling et al., 2010)	(Chen et al., 2004)
Arthritogenicity	(Holoshitz et al., 2013)	(Koga et al., 1986; Koga et al., 1980; Kohashi et al., 1986; Zidek & Masek, 1985)

Figure 1. Virtual docking of MDP and SE ligands on CRT.

Ligands are shown as colored sticks; CRT surface is in light gray; CRT residues in close proximity to ligands are shown as dark-gray sticks. As can be seen, the two SE ligands, 15-mer linear peptide 65–79*0401 (A) and the SE-mimetic cQKRAA (B), as well as LD-MDP (C), are predicted to occupy overlapping binding sites on CRT, in significant proximity to several key CRT residues.

So, how has it happened that cusp ligands, an evolutionarily conserved fundamental health-promoting mechanism, ended up facilitating disease development? Two mutually non-exclusive answers can be proposed: 1) Cusp-associated diseases are unavoidable outcomes, considering that the hypervariable cusp region represents an MHC mutational “hot spot”. Because most autoimmune diseases appear late in life, they have little impact on genetic fitness; 2) As discussed above, disease development in genetically susceptible individuals is often triggered by environmental exposure. Thus, modern-era environmental pollutants, rather than evolution’s flawed design, should be blamed. Our own data suggest that environmentally induced posttranslational modifications of cusp-binding receptors might lead to higher affinities of receptor-ligand interactions, and potentiate signal transduction. It is conceivable that environmentally-induced posttranslational modifications could occur at the ligand side as well. Consistent with our data, the Cusp theory proposes that such posttranslational modification events in the receptor-ligand junction might increase the intensity of cusp-triggered signal beyond a certain threshold and, in individuals with the right constellation of genetic risk factors, could trigger disease onset.

In closing, the MHC Cusp theory offers a plausible mechanistic basis for HLA-disease association. It should be cautioned, however that alluring as the MHC Cusp road may seem, the long journey to validating this theory has just begun. Only time will tell if taking the road less travelled by will have made all the difference.