MHC class I and II and HLA disease associations

The major histocompatibility complex (MHC) is the genomic region that encodes the molecules responsible for presenting peptide antigens to T lymphocytes. In humans it is called the human leukocyte antigen (HLA) system. The region covers 0.13% of the human genome and spans approximately 4 megabase pairs on the short arm of chromosome six at position 6p21, within a region that contains more than 250 annotated genes and pseudogenes. This density is remarkable: a small stretch of DNA contains the most polymorphic genes in the human genome, together with genes for antigen processing, complement components and inflammatory mediators. The classical class I and class II regions have extensive patterns of linkage disequilibrium, and a high degree of single nucleotide polymorphism at the HLA genes can differentiate populations worldwide. ^[1]

The functional logic is simple to state and hard to overstate. A T cell receptor cannot see intact antigen. It sees a short peptide held in the groove of an MHC molecule on the surface of another cell. This solves a problem at once: it allows the immune system to survey what is being made inside a cell, not just what is floating outside it, and it makes the response MHC-restricted, so that T cells are activated only when antigen is presented in the correct context. The price of this system is polymorphism: because MHC molecules must bind a vast range of peptides, the genes encoding them are the most variable in the genome, and that variability is what creates the HLA disease associations discussed later.

Class I and class II molecules differ in structure, expression and function. A classical MHC class I molecule is a heterodimer of a heavy chain (alpha chain) non-covalently associated with beta-2 microglobulin. The heavy chain has extracellular domains: alpha-1 and alpha-2 form the peptide-binding groove, and alpha-3 binds the CD8 co-receptor. Class I molecules are expressed on almost all nucleated cells and present peptides derived mainly from proteins degraded in the cytosol. A classical MHC class II molecule is a heterodimer of an alpha chain and a beta chain, both of which contribute to the peptide-binding groove. Class II molecules are expressed constitutively on professional antigen-presenting cells such as dendritic cells, macrophages and B cells, and can be induced on other cells by interferon-gamma. They present peptides derived mainly from proteins degraded in endosomes and lysosomes.

The MHC region also contains non-classical genes. HCP5 is a long non-coding RNA gene located between MICA and MICB within the class I region, about 105 kilobases centromeric of HLA-B. It is a human-specific gene with a promoter that has at least 88% identity with the promoters of classical class I genes, and its expression is regulated by interferon-gamma along with other class I genes. HCP5 illustrates two general principles. First, the MHC contains regulatory elements as well as coding genes. Second, linkage disequilibrium means that a variant in a non-coding gene can act as a marker for a nearby classical HLA allele. Much of what is known about HLA disease associations comes from exactly this kind of marker analysis. ^[1]

MHC polymorphisms influence many critical biological traits and susceptibility to chronic autoimmune diseases including type 1 diabetes, rheumatoid arthritis, celiac disease, psoriasis, ankylosing spondylitis, multiple sclerosis, Graves' disease, schizophrenia, bipolar disorder, inflammatory bowel disease and dermatomyositis. The breadth of this list reflects the central role of antigen presentation in immune regulation. A variant that changes the shape of the peptide-binding groove, the level of surface expression or the stability of the mRNA can shift the balance between tolerance and activation. The same polymorphism that protects against one infection may predispose to autoimmunity, which is why HLA associations are best understood as trade-offs rather than simple defects. ^[1]

The clinical importance of the MHC extends beyond autoimmunity. Different viral infections and cancers are associated with suppression of MHC genomic expression, particularly in the class I and class II regions. This is a form of immune evasion: if a virus or tumour cell reduces MHC expression, it becomes less visible to T cells. The MHC is therefore both a record of past selection and a live target in infection and cancer. Understanding its structure is a prerequisite for interpreting transplantation matching, disease association studies and the growing field of immunomodulatory drugs.

A final definitional point concerns nomenclature. HLA alleles are named by locus (A, B, C, DP, DQ, DR) followed by an asterisk and a numeric designation. The number often reflects the serological specificity, and the colon-separated fields indicate sequence variants. This system allows a clinician to interpret a transplant report or a pharmacogenomic alert without needing to know the underlying nucleotide sequence. The key is to remember that the same allele can have different associations in different populations, because linkage disequilibrium with neighbouring genes varies.

Antigen presentation is a logistics problem. The immune system needs to display samples of proteins from every compartment on the cell surface, in a form that T cells can read. MHC class I and class II solve this problem with separate routes that converge on the same outcome: a peptide-loaded MHC molecule at the plasma membrane.

The class I pathway begins in the cytosol. Proteins, including viral proteins and abnormal self-proteins, are degraded by the proteasome into short peptides. The proteasome can be replaced by the immunoproteasome, which is induced by interferon-gamma and produces peptides with hydrophobic C-terminal residues that favour class I binding. Peptides are then transported into the endoplasmic reticulum by the transporter associated with antigen processing (TAP). In the endoplasmic reticulum, the class I heavy chain associates with beta-2 microglobulin and with chaperones including calreticulin and tapasin. Tapasin stabilises the empty class I molecule and facilitates peptide loading. Once a suitable peptide binds, the complex becomes stable and travels through the Golgi apparatus to the cell surface.

The class II pathway begins in the endosomal compartment. Antigen-presenting cells take up extracellular material by phagocytosis, macropinocytosis or receptor-mediated endocytosis. The class II alpha and beta chains are synthesised in the endoplasmic reticulum and associate with a chaperone called the invariant chain. The invariant chain occupies the peptide-binding groove and directs the class II molecule to the endosomal pathway. In the endosome, proteases degrade the invariant chain, leaving a residual fragment called CLIP in the groove. HLA-DM, a non-classical class II molecule, catalyses the exchange of CLIP for antigenic peptides generated in the endosome. The loaded class II molecule then moves to the surface.

Cross-presentation is the exception that proves the rule. Some dendritic cells can take up exogenous antigen and divert it into the class I pathway, so that CD8+ T cells can be primed against viruses and tumours that do not infect dendritic cells directly. The mechanism is still debated, but the outcome is clear: the immune system can generate cytotoxic responses to antigens that are not made inside the presenting cell. This is important for vaccines that deliver protein antigens and for understanding why some tumours are immunogenic despite not being directly infected.

Class I expression is not simply housekeeping. The class I transactivator NLRC5 induces genes involved in the class I-dependent antigen processing and presentation pathway, activating the promoters of classical and non-classical class I genes, beta-2 microglobulin, the immunoproteasome component PSMB9 and the peptide transporter TAP1. Interferon-gamma can activate class I transcription by upregulating IRF1, which binds interferon-sensitive response elements in the class I promoter. This is why viral infection and inflammation increase class I expression: the cell is being told to display more of what it is making.

Class II expression is more restricted and more regulated. Constitutive expression is largely confined to professional antigen-presenting cells, but interferon-gamma can induce class II on many cell types, including endothelial cells and epithelial cells. This inducible expression is part of the inflammatory programme and can contribute to autoimmune pathology when it allows non-professional cells to present self-antigen to autoreactive T cells.

The pathways also differ in the nature of the peptide repertoire. Class I peptides are derived from proteins that are present in the cytosol, so the repertoire reflects the cell's current biosynthetic activity, including viral proteins, mutated oncoproteins and normal self-proteins. Class II peptides are derived from proteins that have been taken up from the extracellular space, so the repertoire reflects the tissue environment and the pathogens that have been encountered. This division allows the immune system to distinguish between intracellular infection, which requires a cytotoxic response, and extracellular infection, which requires a humoral and helper response.

A practical point for medical students is that the terms "endogenous" and "exogenous" refer to the route of entry into the cell, not to the origin of the protein. A viral protein made inside a cell is endogenous; the same viral protein taken up as a dead particle is exogenous. The immune system uses the route to decide which kind of T cell response to prime. This is why live attenuated vaccines and inactivated vaccines can generate different immune profiles, and why adjuvants that target class I cross-presentation are valuable.

T cell activation requires more than peptide-MHC engagement. The T cell receptor provides specificity, but the co-receptors, adhesion molecules and co-stimulatory signals determine whether the encounter leads to activation, anergy or death. The major co-receptors are CD8, which binds MHC class I, and CD4, which binds MHC class II.

CD8 is a heterodimer of alpha and beta chains (or a homodimer of alpha chains) that binds to the alpha-3 domain of the class I heavy chain. This interaction is outside the peptide-binding groove, so it does not affect peptide specificity. Instead, it increases the avidity of the T cell for the antigen-presenting cell and brings the tyrosine kinase Lck close to the T cell receptor complex. Lck phosphorylates immunoreceptor tyrosine-based activation motifs in the CD3 chains, initiating the signalling cascade that leads to T cell activation. The functional consequence is that CD8+ T cells are specialised for recognising antigen presented on class I molecules, which means they can detect intracellular pathogens and tumours.

CD4 is a monomeric glycoprotein with immunoglobulin-like domains. It binds to the beta-2 domain of the class II molecule. Like CD8, it recruits Lck and amplifies signalling. CD4+ T cells are therefore restricted to class II molecules and are specialised for recognising antigen that has been taken up from the extracellular environment. This restriction is the structural basis of the helper function: CD4+ T cells see what antigen-presenting cells have processed from the tissue environment, and they respond by producing cytokines that help B cells, CD8+ T cells and innate cells.

The functional division between CD8+ and CD4+ T cells is one of the clearest examples of structure determining function. CD8+ T cells become cytotoxic effectors that kill infected cells through perforin and granzyme, and they also produce cytokines such as interferon-gamma and TNF. CD4+ T cells become helper cells that produce a range of cytokines depending on the context. T helper type 1 cells produce interferon-gamma and activate macrophages; T helper type 2 cells produce IL-4, IL-5 and IL-13 and help B cells; T follicular helper cells migrate to germinal centres and support affinity maturation; T helper 17 cells produce IL-17 and recruit neutrophils; and regulatory T cells suppress inappropriate responses.

The interaction between CD4+ and CD8+ T cells is bidirectional. CD4+ T cells provide help for CD8+ T cell memory formation and for B cell class switching. This help can be delivered through cytokines, through co-stimulatory molecules such as CD40 ligand, or through direct cell-cell contact. In viral infections, the balance between CD4+ help and CD8+ effector function often determines whether the response clears the infection or causes immunopathology.

Beyond the classical co-receptors, other receptor families modulate T cell and natural killer cell responses to MHC molecules. The leukocyte immunoglobulin-like receptors (LILRs) bind classical and non-classical MHC class I alleles with different affinities. LILRB1 has lower affinity for some HLA-A alleles, and LILRB2 binds most strongly to HLA-A and weakly to HLA-B, with weakest binding to a subset including HLA-B27 and HLA-B*5701^[1]. Some of these outliers are MHC class I alleles with known disease associations, which has led to the suggestion that LILR recognition might influence susceptibility to and outcome of viral infections or autoimmune diseases^[1].

This "degrees of self" model is useful because it moves beyond the simple idea that an HLA allele is either protective or harmful. An allele that binds an inhibitory receptor strongly may dampen immune responses and increase susceptibility to infection; the same allele may reduce the risk of autoimmunity by raising the threshold for activation. The net effect depends on the infection, the tissue and the genetic background.

The co-receptor interaction is also relevant to therapeutic manipulation. Blocking CD4 or CD8 function can prevent graft rejection or treat autoimmune disease, but it also impairs antiviral immunity. Understanding the structural basis of co-receptor binding helps explain why these therapies have predictable infectious complications and why they are used with careful monitoring.

The association between HLA alleles and disease is one of the oldest and strongest findings in immunogenetics. The MHC region contributes to more autoimmune diseases than any other region of the genome. The list of conditions associated with MHC polymorphisms includes type 1 diabetes, rheumatoid arthritis, celiac disease, psoriasis, ankylosing spondylitis, multiple sclerosis, Graves' disease, schizophrenia, bipolar disorder, inflammatory bowel disease and dermatomyositis. The breadth of this list is a reminder that antigen presentation sits at the centre of immune regulation.

HLA-B27 and ankylosing spondylitis illustrate the complexity of HLA-disease associations. The association is strong enough to be clinically useful as a supportive test, but it is not diagnostic because many HLA-B27 carriers never develop disease. The mechanism is thought to involve presentation of arthritogenic peptides, although the exact peptides remain uncertain. Other hypotheses include misfolding of the B27 heavy chain, which can trigger endoplasmic reticulum stress and inflammatory signalling, and interaction with innate immune receptors. The clinical lesson is that HLA-B27 testing is most useful in the context of typical clinical features and imaging, not as a stand-alone screening test.

HLA-DR3 and DR4 are associated with type 1 diabetes and rheumatoid arthritis. The association with type 1 diabetes is strongest for certain combinations of DR and DQ alleles, reflecting the extensive linkage disequilibrium across the class II region. In rheumatoid arthritis, particular DR4 subtypes are associated with anti-citrullinated peptide antibody-positive disease, suggesting that the relevant mechanism involves presentation of citrullinated self-peptides to CD4+ T cells. This is a good example of how a genetic association can point to a specific molecular hypothesis.

HLA-DQ2 and DQ8 are associated with celiac disease. The DQ2 heterodimer encoded by DQA105 and DQB102 is the principal risk allele, and DQ8 is an alternative permissive haplotype. The mechanism is well characterised: gluten-derived peptides are deamidated by tissue transglutaminase and then presented by DQ2 or DQ8 to CD4+ T cells, which drive the inflammatory response in the small intestine. This is one of the clearest examples of an HLA association with a defined environmental trigger and a defined molecular mechanism. It also explains why almost all patients with celiac disease carry DQ2 or DQ8, and why the absence of these alleles makes the diagnosis very unlikely.

Beyond the classical autoimmune associations, HLA alleles influence infectious disease outcomes. In HIV infection, a meta-analysis found that 46 of the 50 best signals originated from the HLA locus, emphasizing the critical role of the MHC in controlling HIV-1 replication and delayed disease progression. The HCP5 single nucleotide variant rs2395029-G is in strong linkage disequilibrium with HLA-B57:01, and this haplotype is associated with HIV non-progression. However, the association is complicated: in multivariate models that partition out the protective effect of HLA-B57, the HCP5 minor allele is associated with disease acceleration and enhanced viral replication. This contradiction illustrates how linkage disequilibrium can create misleading associations and why fine-mapping is necessary. ^[1]

HLA associations also extend to psoriatic disease. The same HCP5 and HLA-B genotypes are associated with psoriasis and psoriatic arthritis. The HCP5 variant rs2395029 had the highest odds ratio with both psoriasis and psoriatic arthritis for 223 Caucasian individuals, and its effect was independent of the most highly associated single nucleotide polymorphism upstream of HLA-C. The MHC, particularly the class I region, is the only genomic region consistently associated with psoriasis. This is another example of how a non-coding variant can act as a marker for a functional HLA allele. ^[1]

The broader lesson is that HLA-disease associations are population-specific and haplotype-dependent. The same allele may be protective in one context and harmful in another, because the relevant peptide repertoire and the relevant immune response differ. This is why genetic testing for HLA associations must be interpreted with clinical context and why a positive result is rarely sufficient on its own.

Drug hypersensitivity is an immune-mediated adverse reaction to a medication. The clinical spectrum ranges from mild maculopapular rash to severe cutaneous adverse reactions such as Stevens-Johnson syndrome and toxic epidermal necrolysis, and to systemic reactions such as drug reaction with eosinophilia and systemic symptoms. The MHC system is central to many of these reactions because the drug or its metabolite must be presented to T cells in the context of an HLA molecule.

Abacavir is a nucleoside reverse transcriptase inhibitor used in HIV treatment. Hypersensitivity to abacavir is strongly associated with HLA-B57:01. The HCP5 single nucleotide variant rs2395029 acts as a marker for HLA-B57:01 in populations of mainly Caucasian or Hispanic descent, and genome-wide association studies identified it as a potential marker for abacavir-induced hypersensitivity. The mechanism is thought to involve non-covalent binding of abacavir to the peptide-binding groove of HLA-B57:01, which alters the repertoire of self-peptides presented and triggers a T cell response. Screening for HLA-B57:01 before starting abacavir has reduced the incidence of hypersensitivity reactions and is now standard practice in many settings. ^[1]

Flucloxacillin is a penicillinase-resistant penicillin used for staphylococcal infections. It can cause cholestatic liver injury, and the same HCP5 and HLA-B genotypes associated with psoriasis have been found to be a major determinant of flucloxacillin-induced liver injury. The association is weaker than the abacavir association but is clinically important because flucloxacillin is widely prescribed. The mechanism is less well understood than for abacavir, and the risk is low enough that routine screening is not currently recommended.

Carbamazepine is an anticonvulsant and mood stabiliser that can cause Stevens-Johnson syndrome and toxic epidermal necrolysis. The risk is strongly associated with HLA-B15:02 in some Asian populations, particularly Han Chinese, Thai and other Southeast Asian groups. The association is less strong in European populations, which is why screening recommendations vary by ancestry. The mechanism is thought to involve direct interaction of carbamazepine with the HLA-B15:02 molecule, altering peptide presentation.

Allopurinol is a xanthine oxidase inhibitor used for gout and hyperuricaemia. It can cause severe cutaneous adverse reactions, and the risk is associated with HLA-B*58:01, particularly in Han Chinese, Thai and Korean populations. As with carbamazepine, the association is population-specific, and screening is recommended in high-risk groups before starting allopurinol.

The mechanism of HLA-associated drug hypersensitivity is an active area of research. Several models have been proposed. The hapten model suggests that the drug or its metabolite binds covalently to a peptide or to the HLA molecule, creating a neo-antigen. The pharmacological interaction model suggests that the drug binds non-covalently to the HLA molecule or to the T cell receptor, altering the immune synapse without covalent modification. The altered peptide repertoire model suggests that the drug binds inside the peptide-binding groove and changes the repertoire of self-peptides that are presented, so that T cells see peptides that were previously hidden. The abacavir reaction is best explained by the altered repertoire model, whereas other reactions may involve different mechanisms.

The clinical implications are substantial. Pharmacogenomic testing is now recommended before abacavir in many guidelines, and before carbamazepine in populations with a high prevalence of HLA-B*15:02. The cost-effectiveness of screening depends on the prevalence of the allele, the severity of the reaction and the availability of alternative drugs. In populations where the allele is rare, universal screening may not be justified, whereas targeted screening based on ancestry can be cost-effective.

HCP5 is a good example of how non-coding MHC variants can be clinically useful. It is a human-specific long non-coding RNA gene within the class I region, and its single nucleotide variants have been associated with HIV viral load, autoimmune diseases, disease relapse after transplantation and various cancers. For drug hypersensitivity, the HCP5 variant rs2395029 is a marker for HLA-B*57:01 and therefore for abacavir hypersensitivity risk. The variant itself may not be functional; it is simply a genetic signpost that travels with the functional HLA allele.

The broader lesson is that drug hypersensitivity is a model system for understanding HLA biology. It shows how a single amino acid change in the peptide-binding groove can alter the immune response to a small molecule, and it shows how genetic testing can be translated into clinical practice. It also shows the limits of pharmacogenomics: a positive test identifies risk, not certainty, and a negative test does not eliminate risk. Clinical judgement remains essential.

The first and most important pitfall is to mistake a proxy for a cause. The HCP5 variant rs2395029 is a useful genetic marker or proxy for HLA-B57:01 in Europeans, and it was used in exactly that way to show that abacavir hypersensitivity in HIV patients is associated with HLA-B57:01. Investigators concluded that HCP5 had no direct antiviral effect, and that the association of an HCP5 variant with viral control was most likely due to an HLA-B*57:01-related effect or to other functional variants in the haplotype, or both. A marker that travels with a functional allele is not itself the functional allele. Students who read a genome-wide association signal as a causal gene will predict the wrong mechanism and design the wrong experiment.

The second pitfall is to assume that proxy testing always works. When both the HCP5 single nucleotide polymorphism and HLA-B57:01 were genotyped in a set of 1888 samples, the correlation was good, but one HLA-B57:01-positive sample tested negative for the HCP5 variant, and HCP5 could not be amplified in two samples because of a homozygous deletion of the HCP5 gene. A proxy assay therefore has failure modes that a direct HLA genotype does not. This is why genotyping for HLA-B57 is regarded as essential, since the accepted hypothesis is that abacavir alters the repertoire of peptides able to bind MHC, allowing presentation of novel self-peptides that would not bind HLA-B57:01 in the absence of the drug. ^[1]

The third pitfall concerns genomic coordinates. Annotation can be misleading. One assembly placed the HCP5 variant at an extended genomic coordinate spanning 76,805 nucleotides, with the 5-prime end lying within the MICA gene sequence about 76,000 base pairs from the transcription start site and overlapping MICA by 60 per cent, rather than within the promoter region closer to the transcription start site of the 2630 base pair HCP5 variant. Variants reported against such coordinates may sit outside the actual transcript. A concrete example is the HCP5 single nucleotide polymorphism rs6938467 associated with Kawasaki disease in Korean children, which is positioned well outside the HCP5 RNA reference sequence and upstream of the LINC01149 locus. A clinician reading a reported variant should check which transcript it is annotated against.

The fourth pitfall is to ignore linkage disequilibrium in the other direction. It is not only that a proxy may stand in for a functional allele; it is also that many reported associations have not been resolved into functional variants. A 2016 analysis concluded that more studies were needed to clarify whether these variants are functional or simply reflect strong linkage disequilibrium with other variants that are functional but as yet unidentified. At the same time, HCP5 is often in linkage disequilibrium with many genes in the MHC yet can be activated or suppressed independently of most of them. Both statements are true, and together they mean that association data must be interpreted locus by locus rather than by assumption.

The fifth pitfall is to treat expression as uniform across cell types. The original description reported a 2.5 kilobase transcript in human B cells, phytohaemagglutinin-activated lymphocytes, a natural killer-like cell line, normal spleen, hepatocellular carcinoma, neuroblastoma and other non-lymphoid tissue, but not in T cells. HCP5 is widely expressed at low levels but is expressed at higher levels in immune tissues such as spleen, blood and thymus. Absence of signal in one cell type is therefore not evidence that the locus is silent, and presence in a tumour line is not evidence that the locus is oncogenic in that tumour.

The sixth pitfall is to assume that losing the gene must be harmful. The absence of HCP5 from the MHC region of chimpanzees and gorillas, and from some human haplotypes carrying the HLA-B*48:01 allele, appears not to be deleterious or life-threatening, and great apes and humans without the gene seem to live in relatively good health. Deletion of a regulatory non-coding gene may be buffered by redundancy, which is a useful corrective to the assumption that every conserved element is indispensable. ^[1]

The seventh pitfall is to read epigenetic data as causal. Hypomethylation of HCP5 with upregulation of its transcript is a shared feature across several cancerous and non-cancerous diseases, which makes it an attractive candidate biomarker. But a methylation difference in diseased tissue may be a consequence of the disease process rather than a cause. The same caution applies to the finding that HCP5 and PTPRN2 were differentially methylated in the endometrium in a study linking DNA methylation, gene expression and the humoral immune response to influenza. Association with a phenotype is the start of the argument, not the end.

Finally, a structural pitfall: interactions between MHC class I and its receptors are not always between cells. Members of the leukocyte immunoglobulin-like receptor family may interact in cis with MHC class I on the same cell surface, as demonstrated for PIR-B and LILRB1^[1]. A model that assumes all relevant binding happens in trans between two different cells will miss part of the biology. Similarly, HCP5 has been described as a defence-response gene that can also have disease-progression and oncogenic effects, acting like a double-edged sword depending on other regulators in the pathway. Simple labels such as protective or harmful are usually incomplete.

The most developed clinical use is pharmacogenomic screening before abacavir. Abacavir hypersensitivity occurs in approximately 5 to 8 per cent of Caucasian patients, but is much less common in patients of African descent, often cited as around 1 per cent or less. An analysis of 245 HIV patients concluded that HCP5 rs2395029 testing could be as useful as HLA-B57:01 typing to prevent the reaction, and because HCP5 testing is cheaper, less time-consuming and easier to perform, it might confidently replace HLA typing in clinical settings. That conclusion has to be read alongside the later discordance data: in 1888 samples, one HLA-B57:01-positive sample was negative for the HCP5 variant and two samples could not be amplified because of homozygous deletion. The practical synthesis is that a proxy assay is attractive for resource-limited settings, but a direct HLA-B57 genotype remains the reference standard, because genotyping for HLA-B57 is essential to the mechanistic explanation that abacavir alters the peptide repertoire and generates novel self-peptides that trigger a cytotoxic response. A structurally polymorphic lineage insertion within intron 1 of MICB, known as AluMICB, has been proposed as an alternative simple proxy for HLA-B57 and HLA-B48:01. ^[1]

Beyond abacavir, HLA and MHC variation guides interpretation of infectious disease course. The polymorphism designated minus 35C/T lies 35 kilobases upstream of the HLA-C locus, and the minus 35C allele corresponds with increased HLA-C expression, which in turn is associated with delayed onset of AIDS^[1]. HLA-C proteins are more stable in the open conformer form than their HLA-A and HLA-B counterparts and are upregulated after immune cell activation, and LILRA1 and LILRA3 preferentially bind HLA-C open conformers, which offers a further mechanism for MHC class I disease associations during HIV infection^[1]. For hepatitis C, expression of HLA-B27 is associated with spontaneous clearance of the virus^[1]. In dengue, large case-control studies have identified MHC class I alleles with protective effects, and antibody-opsonized dengue has been shown to co-ligate the inhibitory receptor LILRB1 when engaged by Fc gamma receptor, leading to inhibition of Fc gamma receptor signalling^[1]. These examples show why HLA typing informs prognosis and why receptor-level pharmacology is being explored.

In rheumatology, the same molecules inform risk stratification and treatment monitoring. Increased serum levels of functional LILRA3 are a proposed genetic risk factor for rheumatoid arthritis, with serum levels correlating directly with disease severity, and LILRA2, LILRA5, LILRB2 and LILRB3 are prominently expressed in synovial tissue of patients with rheumatoid arthritis, with reduction of LILRA2, LILRB2 and LILRB3 in patients who respond to disease-modifying antirheumatic drugs^[1]. LILRA3 deficiency may be a risk factor for Sjogren's syndrome in some populations, while the functional allele is a suggested risk factor in others^[1]. In ankylosing spondylitis, 68 differentially methylated sites were identified between patients and osteoarthritis controls, and HCP5 and POU5F1 were both hypomethylated in HLA-B27-positive compared with HLA-B27-negative patients, leading to the suggestion that HLA-B*27 might act in part through epigenetic dysregulation.

Dermatology and oncology provide further examples. The nine top-ranking variants in one psoriasis study came from the MHC, and seven remained significant after adjustment for multiple testing, with the same HCP5 variant implicated in psoriasis and in HIV infection. Two HCP5 variants were strongly associated with susceptibility to herpes zoster caused by varicella zoster virus. At least six HCP5 variants, including rs3130907, were associated significantly with lung cancer susceptibility, and inactivation of BRG1 contributed to non-small cell lung cancer aggressiveness by altering nucleosome positioning and downregulating HCP5 and all of the classical and non-classical HLA class I genes. HCP5 is overexpressed in lung adenocarcinoma and correlates with poor prognosis specifically in smokers with EGFR and KRAS mutations, and it is overexpressed in lymph node metastasis of small cell lung cancer, glioma, colorectal cancer, and cancers of the colon, thyroid, cervix and breast. Kawasaki disease susceptibility mapped to the HLA region on chromosome 6p21.3, including HLA-C, HLA-B, MICA and HCP5 variants. Plantar warts were associated with a variant lying between HCP5 and MICB. ^[1]

Metabolic and inflammatory disease also connect to this region. In an epigenome-wide study of adiposity, NLRC5 was highly significantly hypermethylated, and NLRC5 is known to regulate MHC class I gene expression and to limit activation of inflammatory pathways, leading to the suggestion that obesity might suppress the adaptive immune response and induce inflammation that could contribute to insulin resistance. In the same programme of work, NLRC5 and HCP5 together with HLA-B, TAP1, TAP2, PSMB8 and HLA-E were significantly hypermethylated for both general and abdominal adiposity. For the clinician, the unifying message is that MHC variation is used to select drugs safely, to interpret infection outcomes, to monitor rheumatological treatment response, and increasingly to understand why some tumours escape T cell recognition.