What the HLA-I!—Classical and Non-classical HLA Class I and Their Potential Roles in Type 1 Diabetes

Stimulated transcriptional regulation of HLA-I genes usually falls under two main modules in the proximal promoter region: (a) the enhancer A (EnhA) and the interferon (IFN)-stimulated response element (ISRE) and (b) the SXY module. NOD-like receptor family CARD domain containing 5 (NLRC5) and signal transducer and activator of transcription 1 (STAT1) are important regulators of HLA gene expression as a part of coordinated immune responses to infections. In human embryonic kidney cells, NLRC5 binds and transactivates HLA class I gene promoters [12]. In vascular smooth muscle cells, the interaction of IFNγ and toll-like receptor 4 affected expression of a large number of STAT1-dependent genes including chemokine adhesion molecules and antiviral/antibacterial genes [13]. Whereas STAT1 regulates HLA-E, it does not appear to regulate HLA-G and HLA-F [14]. The main transcriptional regulators of HLA-G appears to be specificity protein 1 (Sp1), ISRE and SXY, whereas the main regulators of HLA-F are nuclear factor κB (NFκB), interferon regulatory factor 1 (IRF1) and class II, major histocompatibility complex transactivator (CIITA). CIITA also regulates HLA-E expression.

Although this all points to regulation by interferons, the differences highlight a potential for very complex and individualised responses to different infections and modes of interferon activation.

While evidence of aberrant HLA class II expression in the pancreatic islets of patients with type 1 diabetes is increasing [15], hyperexpression of HLA-I antigens in insulin-containing islets (ICIs) is now widely accepted as a defining hallmark of the disease. HLA-I hyperexpression is defined as dramatically elevated expression of HLA-I in all the islet cells (not just β cells), when compared to the surrounding acinar tissue of the same donor or other islets from individuals without type 1 diabetes.

HLA hyperexpression has been confirmed at both the RNA and protein levels in patients with recent-onset disease and in many patients who retain residual ICIs with < 10 years of diagnosis [16••]. Hyperexpression is observed in the ICIs of type 1 diabetes organ donors in the presence or absence of inflammatory infiltrates of the islets (termed ‘insulitis’) [17]. In support of this, studies examining insulitis in newly diagnosed type 1 diabetes donors over the age of 18 years demonstrate that only 25–30% of the ICIs have insulitis, yet all residual ICIs hyperexpress HLA-I [18, 19•]. It is possible that some of these islets have yet to experience insulitis, and thus, HLA-I hyperexpression might represent an earlier stage in the disease pathogenesis. It is hypothesised that HLA-I hyperexpression is primarily driven by the release of diffusible interferons by the β cells themselves [20•], which could derive from the sensing of an infection. Hyperexpression could be exacerbated following immune cell infiltration through the release of cytokines (including IFNγ) from the infiltrating cells. Importantly, once the β cells have been destroyed, the hyperexpression of HLA-I is lost [16••, 17]. The majority of studies in the field of type 1 diabetes have focused on the classical HLA-I (HLA-A, B and C), its accessory protein, β₂M and transporter associated with antigen processing 1 (TAP1), which is involved in class I assembly. However, expression of select non-classical HLA-I molecules have been described in pancreatic islets, and aberrant regulation in the islets of individuals with type 1 diabetes has been reported [16••, 21]. However, much less is known about the function of these non-classical isoforms and the role they may play in type 1 diabetes. This will be the focus of the following review.

HLA-E shares many structural features with classical class I molecules. It is assembled and regulated using common pathways and is ubiquitous throughout tissues, but is the least polymorphic of the HLAs and is transcribed at lower rates. Twenty-seven HLA-E alleles encode 8 proteins, according to the IMGT/HLA Database (http://​hla.​alleles.​org/​ April 2019). Under basal conditions, the HLA-E protein is present predominantly in immune cells and endothelial cells, and cell surface presentation of HLA-E requires the loading of one of a limited range of peptides [10]. Transcription of HLA-E is upregulated by IFNγ, mediated by an upstream STAT1 binding site. HLA-E transcription can be also be induced by CIITA through the SXY regulatory module, but it is not upregulated by NFκB or IRF1 [14]. Two known functional variants of this molecule are HLA-E*01:01 (HLA-E107R) and HLA-E*01:03 (HLA-E107G), differing by a single amino acid at position 107 (Arg/Gly). This change impacts the thermal stabilities of the HLA/peptide complexes and their length of interaction with cognate receptors, as well as expression levels [22]. It has been speculated that individuals carrying HLA-E*01:03 have high levels of the molecule with a greater affinity for viral antigens [23].

As stated above, a limited diversity of self-peptides can occupy the peptide binding groove of HLA-E. This includes the leader sequences of HLA-I molecules HLA-A, HLA-B, HLA-C and HLA-G. When these peptides bind HLA-E, the resulting complex is presented on the cell surface and can interact with inhibitory CD94/NKG2 receptors, located on most natural killer (NK) cells and on a subset of T cells. This is a critical mechanism of tolerance and self-surveillance. In tumour cells, for instance, the loss of classical HLA-I expression would provide a survival advantage, but the subsequent reduction in availability of leader peptides, along with the co-regulation, determines a decrease in HLA-E presentation: this removes signals that are inhibitory for NK cells, licensing them to kill the target tumour cells.

Micro-environmental stresses such as hypoxia and glucose deprivation can lead to upregulation of HLA-E [24]. An environment that impairs the peptide transporter TAP, which can be inhibited by a viral infection or in tumour tissue, can determine a reduction in availability of HLA-I leader peptides and can lead to an alternative repertoire of peptides being presented in the context of HLA-E; this can subsequently impact the outcome of the viral infection or the elimination of damaged or neoplastic cells [24‐26]. In endothelial cells, the pro-inflammatory mediators TNFα, interleukin-1β and IFNγ can upregulate the cell surface expression of HLA-E and induce the release of soluble HLA-E. Upregulation of membrane-bound HLA-E protects activated endothelial cells from NK cell lysis, whereas soluble HLA-E protects bystander cells [10].

In summary, the expression of HLA-E is induced by many of the same stimuli that regulate classical HLA-I, and the critical peptides presented by HLA-E are the leader peptides derived from these. As one of the key functions of HLA-E is to regulate NK cell activity, it is conceivable that the coordinated upregulation of HLA-E with classical HLA-I is designed to provide the target cell with protection from NK-mediated cytotoxicity, while still facilitating the ability of the CD8+ T cells to specifically kill any cells in which they recognise their target peptide bound to the classical HLA-I. In an inflammatory environment where many immune cells have been recruited, this mechanism would result in a shielding of bystander cells from immune attack, while still allowing the targeted destruction of select dysfunctional/virally infected cells (Fig. 1).

HLA-F is highly conserved with low levels of polymorphism in humans: 38 alleles give rise to 6 different full-length protein molecules (according to the IMGT/HLA Database http://​hla.​alleles.​org/​ April 2019). Expression of HLA-F is tightly controlled and tissue-specific, with higher levels in lymphoid cells compared with non-lymphoid cells. Transcription can be upregulated by NFκB through EnhA, IFNγ via the ISRE and CIITA [14]. Although this molecule is known to associate with β₂M and TAP, it is largely intracellular rather than expressed at the surface [27]. Upon immune cell activation, HLA-F can bind HLA-I heavy chains and facilitate their migration to the cell surface in an open-conformation (i.e. not bound to peptides) [28]. These complexes can act as ligands for inhibitory NK cell receptors to modulate the immune system [29]. HLA-F is also a genetic determinant of fecundability [30] and exerts an immunomodulatory function in maternal-foetal tolerance during pregnancy. Extravillous trophoblast cells that invade the maternal endometrial decidua express HLA-F, along with HLA-E and HLA-G, [31]. HLA-F is believed to interact with the inhibitory immunoglobulin-like transcript 2 (ILT2) and 4 (ILT4) receptors present on a variety of immune cells [32], as well as with KIR receptors present on NK cells [29] (Fig. 1). HLA-F also acts as a ligand for inhibitory KIR3DL2 to prevent astrocyte toxicity towards motor neurons in the development of amyotrophic lateral sclerosis [33].

Until recently, efforts to characterise HLA-F associated with β₂M and sequence peptides have been largely unsuccessful [28, 29]. A more recent study, however, successfully engineered β₂M to HLA-F as a single polypeptide in order to solve the first crystal structures of HLA-F and characterise peptide repertoires presented by the molecule. This showed that HLA-F can exist as an open conformer but can also bind a diverse number of peptides (> 2000) between 7 and 30 residues long and be recognised by leukocyte immunoglobulin-like receptor 1 (LIR1). This raises the possibility that HLA-F could be recognised by other immune receptors and elicit different responses depending on their conformation [34••, 35]. Although most immunomodulatory actions of HLA-F are inhibitory, under some circumstances HLA-F open conformers activate primary human NK cells. For example, HLA-F binding to the KIR3DS1 receptor can activate NK cells and elicit an antiviral response to inhibit HIV-1 replication [36].

HLA-G is the most polymorphic of the non-classical isoforms, but its grade of polymorphism is still low (61 alleles encode 19 proteins according to the IMGT/HLA Database http://​hla.​alleles.​org/​ April 2019) compared to classical HLA-I, for which > 4000 alleles encode thousands of proteins. Alternative splicing of the primary mRNA leads to 7 isoforms of HLA-G. G1–G4 isoforms have transmembrane and cytoplasmic domains, meaning they are membranous, whereas G5–G7 isoforms are soluble. G1 and G5 are also able to bind non-covalently to β₂M [11]. Unlike HLA-E and the classical HLA-I molecules, HLA-G is not constitutively expressed in tissues, but it is found in a limited number of cell types, such as cytotrophoblast cells at the maternal-foetal interface of the placenta, corneal, nail matrix, embryonic mesenchymal stem cells and pancreatic islet β cells [37, 38]. HLA-G can bind a repertoire of self-peptides. Regulation of HLA-G transcription is peculiar, as this gene is unresponsive to NFκB, IRF1 and CIITA mediated pathways, although it may be responsive to NLRC5 [12, 14]. Alternative regulatory elements have been described, some examples of which include a heat shock element (HSE), which would respond to stress-induced heat shock proteins (HSP), long interspersed elements (LINEs) and the hypoxia-inducible factor (HIF), involved in cellular responses to oxygen depletion [39]. HLA-G mediates immune responses of NK, B and T cells through interactions with inhibitory receptors including ILT-2, ILT-4, KIR2DL4 and CD160 (Fig. 1). It can inhibit a wide range of immune functions, including the antigen-specific cytolytic function of cytotoxic T cells, alloproliferative response of CD4+ T cells, ongoing proliferation of NK and T cells and maturation of dendritic cells [11, 40]. Hence, HLA-G upregulation in tumours and in the placenta during pregnancy is advantageous for concealment from immune surveillance and to impart tolerance. Interestingly, a recent study showed that HLA-E complexed with an HLA-G leader peptide enrich a population of adaptive NK cells [41]. These cells, characterised by a reduced FεCRγ expression, have an upregulated CD25 expression, increased proliferation and increased antibody-dependent cell-mediated cytotoxicity and prompt a heightened IFNγ response. NK cells exposed to HLA-E molecules presenting HLA-C leader peptides, in contrast, did not show the same effects [41].

Very few studies have directly examined the expression of HLA-E in the pancreas. However, studies utilising the unique type 1 diabetes pancreas resection material from recently diagnosed patients enrolled in the Norwegian DiViD study [42] demonstrated that HLA-E RNA expression was elevated in inflamed islets when compared with islets from control pancreata. HLA-E expression was particularly high in the non-infiltrated islet core of the type 1 diabetes donors when compared to the peri-islet area that contained infiltrating immune cells, which suggests that the islet cells rather than the immune cells have the highest expression [43••]. Elevated HLA-E RNA expression was also observed in 4 further recent-onset T1D organ donors, and elevated protein was confirmed in select islets of one of these donors. Hyperexpression was also observed in exocrine tissue of the same T1D donor [44]. More recent studies of recent-onset T1D donors from the Exeter Archival Diabetes Biobank (EADB) and DiViD have revealed that HLA-E is specifically upregulated in the ICIs. This was confirmed at the RNA and protein level. Intriguingly, HLA-E expression appears higher in α cells but was also observed in β cells. HLA-E levels in insulin-deficient islets are comparable with those of islets in non-diabetic controls (Fig. 2; Richardson, unpublished data). There is limited evidence to suggest that certain HLA-E genotypes (HLA-E107R/G) are associated with age at onset of type 1 diabetes [45].

Upregulation of HLA-F (at both the RNA and protein level) has been observed in the insulin-containing islets of patients with recent-onset type 1 diabetes when compared with islets from non-diabetic controls [16••]. This upregulation is lost when islets are devoid of insulin, suggesting that like the classical HLA-I, the stimulus regulating its expression is derived either from the β cells themselves, or requires the presence of β cells to elicit the release of this factor from other cell types, e.g. infiltrating immune cells. Similar observations were made in donor samples from the nPOD, DiViD and EADB cohorts. These findings were supported further by RNA expression data, which showed that HLA-F was upregulated by 1.71 ± 0.04-fold in islets from patients of the DiViD cohort [16••]. Furthermore, bulk sorted β cells from T1D donors showed increased expression of HLA-F compared to non-diabetic donors [15].

Similar to classical HLA-I in T1D, HLA-F was expressed in both α and β cells. Uniquely, in contrast with conventional HLA-F expression, this molecule was primarily localised to the surface of the cell (Fig. 2; inset), co-localising with classical HLA-I [16••]. The ability of HLA-F to aid the translocation of free HLA-I to the cell surface, where it may bind to exogenous peptides, raises the possibility that cross-presentation of peptides from within the extracellular space could be occurring in α and β cells in T1D. Proteins released from neighbouring β cells may be bound, endocytosed and processed within a cell, and subsequently presented in the context of normal HLA-I, which could have important implications for immune responses at the islet site [16••] (Fig. 1).

Associations between an HLA-G polymorphism and the age of onset of type 1 diabetes have been found [46]. Homozygosity for the deletion of 14 bp of the 3′ untranslated region of HLA-G was associated with an earlier age of type 1 diabetes onset, whereas heterozygotes (carrying only one deletion) had a later age of onset of type 1 diabetes [46]. HLA-G is expressed by pancreatic islet and duct cells, and it is upregulated in response to pro-inflammatory cytokines [37]. In pancreatic endocrine cells, HLA-G is constitutively expressed at low levels as a β₂M-free heavy chain protein and remains mainly intracellular. Interestingly, HLA-G appears to be associated with a subset of insulin-containing granules and can be exported to the β cell surface, not only through the constitutive secretory pathway, but also through the regulated pathway by which insulin is secreted [37]. Numerous autoantigens in islet immunity are components of secretory granules [47‐49]; therefore, insulin exocytosis sites may be the sites where the immunogenic ligands become exposed. As the activation of autoreactive T cells depends upon surface density of antigen/MHC complexes, this may lead to the activation of low-affinity cytotoxic T cells. Consequently, the presence of HLA-G at such granule exocytosis sites could represent a prevention mechanism for aberrant immune activations.

In pancreas tissue from patients with recent-onset type 1 diabetes, upregulation of HLA-G at the protein level can be seen in the islets, in both α and β cells. Expression is predominantly cytoplasmic, but HLA-G is also observed at the cell surface (Fig. 2; Wyatt, unpublished data).