THE HUMAN MAJOR HISTOCOMPATIBILITY COMPLEX AS A PARADIGM IN GENOMICS RESEARCH

Abstract

Since its discovery more than 50 years ago, the human Major Histocompatibility Complex (MHC) on chromosome 6p21.3 has been at the forefront of human genetic research. Here, we review from a historical perspective the major advances in our understanding of the nature and consequences of genetic variation which have involved the MHC, as well as highlighting likely future directions. As a consequence of its particular genomic structure, its remarkable polymorphism and its early implication in numerous diseases, the MHC has been considered as a model region for genomics, being the first substantial region to be sequenced and establishing fundamental concepts of linkage disequilibrium, haplotypic structure and meiotic recombination. Recently, the MHC became the first genomic region to be entirely re-sequenced for common haplotypes, while studies mapping gene expression phenotypes across the genome have strongly implicated variation in the MHC. This review shows how the MHC continues to provide new insights and remains in the vanguard of contemporary research in human genomics.

The human major histocompatibility complex (MHC) was discovered more than 50 years ago [1]. It was initially known for its role in transplantation through histocompatibility antigens, hence the other commonly used nomenclature, ‘human leukocyte antigens’ (HLA) [2-5]. Several decades of intensive research has defined the remarkable genomic environment of the MHC and how genetic variation within this region plays a key role in susceptibility to autoimmune, infectious and other diseases. The MHC shows extreme levels of gene density and polymorphism. The nature, coinheritance and functional consequences of its genetic diversity have proved complex. Nonetheless, remarkable insights have been gained and this genomic region has become a paradigm for human genomics. In this review, rather than describing specific MHC associated diseases (for reviews see [6-10]), we recall the historical context and describe the genomic landscape of the MHC and some of the challenges which remain: an ongoing adventure in exploring human genetic diversity that keeps the MHC at the forefront of research in human genomics.

THE MHC: DISCOVERY AND BIOLOGICAL SIGNIFICANCE

Discovery of the MHC

The immunogenetics field was born in 1900 with the discovery of the ABO blood groups by Landsteiner, followed by the Rhesus system in 1940 (Figure 1). Peter Gorer was the first to describe in 1936 a histocompatibility system in mice from his observations of the agglutination of erythrocytes by rabbit immune sera [11]. This research was advanced by George Snell who established that graft rejection in mice was due to incompatibility at the level of certain antigens. The murine MHC was called ‘H2’ in honour of the antigen II discovered by Gorer [12].

Figure 1.

Timeline of research in the MHC and human genome

As for humans, the history of the HLA complex began in 1952 with the princeps observation made by Jean Dausset. He hypothesised that a similar antigenic system to that seen in mouse erythrocytes could exist in humans on the surface of leucocytes, which he demonstrated by showing a massive leucoagglutination by the serum of a poly transfused patient. However, the firm discovery of the first human MHC antigen, MAC (HLA-A2), was made only in 1958 following a classical segregation population analysis using a serum reacting in only a subset of the studied sample [13]. This polymorphic system was then confirmed by the work of Jon van Rood, and of Rose Payne and Walter Bodmer, who defined respectively the antigens 4a and 4b (Bw4 and Bw6), and HLA-A2 and HLA-A3, through studies on multiparous women [2, 14]. An international research effort involving the International Histocompatibility Workshops (IHW) initiated in 1964, led progressively to the characterisation of a gene cluster on chromosome 6 with serologically defined alleles and including HLA-A, HLA-B and HLA-C. The complement system was also mapped to the same genetic region. In the 1970’s, HLA class II alleles were characterised through the identification of mixed lymphocyte reactions. Advances in molecular biology then allowed investigation of the HLA system directly at the level of the genes rather than of their products.

The role of MHC molecules is inherent to their polymorphism

The human MHC, on chromosome 6p21.3, is divided into the classical regions denoted class I and class II, and an intervening region dubbed class III (Figure 2). While the MHC was discovered based on its involvement in alloreactivity, its broader role in the immune response was established by Baruj Benacerraf who demonstrated that the MHC controlled the ability to mobilise an immune response against a particular antigen [15]. Indeed, HLA class I and II genes encode glycoprotein molecules expressed at the cell surface where they present antigenic peptides to CD8 positive and CD4 positive T cells respectively [16]. Their extraordinary genetic polymorphism guarantees the broadest diversity of recognized antigens and therefore reactivity against pathogens, conferring a selective advantage. The affinity between the HLA molecules and the antigenic epitopes is a pivotal feature of this mechanism, highlighted by elucidation of the structure of HLA molecules, established for the first time by Bjorkman in 1987 for HLA-A2 [17], and also for class II molecules [18]. Beyond this function of antigen presentation devoted to HLA molecules, MHC genes are involved in numerous aspects of the immune response and some have also non immune functions, such as olfaction [19, 20]. In particular, genes and polymorphisms in the class III region are involved in the regulation of the humoral immune response [21] and in the inflammatory reaction through genes encoding the tumour necrosis factor (TNF) [22], heat shock proteins or components of the complement cascade.

Figure 2. Genetic variation in the MHC.

Number of polymorphisms per 10 kb across the MHC. Polymorphisms were extracted from dbSNP build 129, microsatellites from dbMHC, copy number variants and indels from the Database of Genomic Variants version 6, and segmental duplications of more than 1 kb from the ucsc table browser [174].

THE GENOMIC LANDSCAPE OF THE MHC: EXTREME GENE DENSITY AND POLYMORPHISM

The first substantial human genomic region to be sequenced

As a consequence of its biological significance, defining the nucleotide sequence of the MHC was a priority and was achieved in 1999, providing important new insights and avenues for investigation. The entire 3.6 Mb was published through the combined efforts of the MHC Sequencing Consortium under the direction of Stephen Beck, Daniel Geraghty, Hidetoshi Inoko and Lee Rowen [23]. Covering about 0.12% of the human genome, this was the first substantial contiguous sequence to be determined, two years before the release of the whole genome draft sequence [24, 25]. Since it was derived from multiple individuals with different HLA types, this was a “mosaic” sequence, as is still the case for the rest of the reference genome assembly (build 36.3), although some complete diploid sequences are now becoming available through re-sequencing [26-29]).

The most dense gene region of the human genome

The MHC Sequencing Consortium established that 224 gene loci were present in the human MHC [23], of which 42% were novel. Within the 1.9 Mb class I region, 18 HLA class I genes are known (the classical genes HLA-A, HLA-B and HLA-C, together with non-classical genes HLA-D, HLA-E, HLA-F, HLA-G and 12 pseudogenes) together with 7 MHC class I-chain related genes (MICA, MICB and 5 pseudogenes). Class I and MIC genes are organised in a three-fold repeated unit. The MHC class II region covers about 0.9 Mb and in the mosaic MHC sequence, 19 class II genes were counted including the classical genes HLA-DP, HLA-DQ and HLA-DR; the non-classical genes HLA-DM and HLA-DO; and 8 pseudogenes. Located between class I and class II regions, the class III is remarkably gene dense, including 62 expressed genes with more than 500 exons over 750 kb, many of which are involved in the immune response [30].

If pseudogenes are included, the average density of genes in the MHC is one gene per 16 kb, making this the most gene dense region of the human genome. There are numerous pseudogenes in the class I and II regions. Both these regions seem to have been duplicated several times, generating new gene family members which then diverged [31, 32]. It is believed that pseudogenes could have played a role in the generation of new alleles by genic conversion. In contrast, the class III region is unique in the near absence of pseudogenes except in the duplicated C4 region and contains one expressed gene at least every 15kb. In some instances, transcripts overlap as for example with TNXB and CYP21A2 genes [33].

Approximately 40% of the genes in the classical MHC are expressed in the immune system and they are physically clustered, reflecting their functional roles [2, 34, 35]. Further studies and sequencing of chromosome 6 [36, 37] revealed that flanking genomic regions showing evolutionary conservation now recognised as the extended class I and extended class II regions. The extended MHC spans 7.6Mb and includes 421 gene loci, of which 60% are thought to be expressed [38].

The most polymorphic human genomic region

Since the discovery of the first human biological polymorphism by Landsteiner, the remarkable nature of human genetic variation has been revealed, ranging from single nucleotide differences to large structural genomic variation spanning thousands of base pairs [39]. These include single nucleotide polymorphisms (SNPs), substitutions and deletion insertion polymorphisms (DIPs) as well as variation involving more than one base such as short tandem repeats (STRs) (microsatellites), large insertion deletions (INDELs), inversions and other repeats (Figure 2). As genetic markers, STRs and SNPs have proved particularly informative. While such variants have been identified in almost all genes, nowhere in the genome can yet compete with the extreme polymorphism found in HLA genes.

In 1980, for each of the four genes described by Jean Dausset in his Nobel lecture, HLA-A, -B, -C and –DR, 8 to 39 codominant alleles were known and the number of their haplotypic combinations was already reaching several millions. Other MHC-linked genes were assumed to also be extremely polymorphic, which led him to emphasise that virtually every human carries a unique combination [1]. By the year 2000, more than 15,000 HLA alleles were described [23]. With 86 SNPs per kb and 1,195 alleles, the HLA-B locus is the most polymorphic gene in the human genome, the number of alleles having doubled over the last four years (IMGT/HLA Database http://www.ebi.ac.uk/imgt/hla/ release 2.24.0) [36, 40]. The second most polymorphic is HLA-A with 733 alleles. A total of 3,371 alleles are now reported for HLA loci and 106 for non classical HLA loci (MICA, MICB, TAP1, TAP2).

Polymorphic sites in classical HLA genes are predominantly located in the second and third exons of the class I loci and in the second exon of HLA-DRB, -DQA, -DQB, -DPA and -DPB. Here, non synonymous single nucleotide substitutions lead to amino acid changes in the domain involved in the binding of antigenic peptides. The distribution of the HLA loci variants is different from the one observed in most functional genes where variants occur more often in introns.

Generation of such extreme polymorphism is attributable to point mutations, recombination and genic conversion [41]. In non coding sequences, variants are more frequent in regions flanking the most polymorphic genes [42] through “genetic hitch-hiking” [2, 43]. Polymorphism of class I and II genes is thought to be maintained by the selective advantage conferred by the heterozygous state allowing a more diverse immune response to infectious pathogens such as HIV-1 [44]. Polymorphism could also be selected because rare alleles provide an advantage on a population scale to escape from infections. Both mechanisms are likely to be overlapping. The extreme polymorphism in the MHC required a special nomenclature where the locus name is followed by an asterisk and 4 digits identifying the allele, the first two reflecting the previous serological nomenclature. Additional digits are added for synonymous changes or non coding variants. This system is applied for class I and class II genes, as well as for TAP or PSMB genes. Prior to the current era of genome-wide association studies and use of SNP microarrays as genetic markers, microsatellites were extensively studied to define disease associations. However there was no standardised nomenclature until the 13^th IHW [45] which resolved aliases for 281 microsatellites with an average density of one per 45 kb. Since then, all microsatellites have been submitted to the dbMHC database (http://www.ncbi.nlm.nih.gov/gv/mhc) [46]. The most recent 14^th IHW successfully mapped 664 microsatellite primer pairs in the current genome assembly [47].

Larger scale but submicroscopic structural genomic polymorphism is also increasingly recognised. It involves variants more than 1 kb in size, including copy number variants, segmental duplications, inversions and translocations. For example, within the MHC (specifically chromosome 6p21.3) a total of two inversions, 63 indels (100bp to 1kb in size) and 181 copy number variants are currently listed in the database of genomic variants (http://projects.tcag.ca/variation/) [48, 49], some being of potential functional importance such as CNV-507 or CNV-505 which overlap the susceptibility loci for sarcoidosis and psoriasis respectively [50]. Two main regions in the MHC are affected by segmental duplications, the HLA-DRB1 region and the complement region, although they are specific to certain allelic combinations or “haplotypes” [51].

A PARADIGM IN THE STUDY OF RECOMBINATION, LINKAGE DISEQUILIBRIUM AND HAPLOTYPIC STRUCTURE

Use of the term “haplotype”

The term “haplotype”, as a contraction of “haploid genotype”, refers to DNA sequence polymorphisms co-inherited at a minimum of two linked loci. It was first used in 1967 in the context of the MHC to describe “the combination of individual antigenic [MHC] determinants that are positively controlled by one allele” [36], although the concept was already applied to the MNS antigen system of human blood groups. It originated from the immunogeneticist Ruggero Ceppellini who observed familial genotype data that could be explained only by the joint inheritance of alleles at closely linked loci [52]. Ceppellini’s contribution to this field began even earlier through his studies of rhesus factor when he introduced the expectation maximization (EM) algorithm that is still used for haplotype reconstruction [53]. The earliest record of “haplotype” in PubMed is found one year later, again in relation to the MHC [54].

Numerous SNP studies have described the existence of DNA sequence segments in the human genome, ranging between 5 and 150 kb and maintained on haplotypes in the absence of genetic recombination. Within the MHC short blocks of less than 0.2 Mb involving regions such as the DR-DQ genes or the ‘complotype’ - a combination of alleles at the C4B, C4A, Bf and C2 complement genes – are described [55]. In addition, combinations of blocks forming single genetic units are found, highlighting the complexity of ‘genetic fixity’ in the MHC such that long conserved sequences representing combinations of multiple blocks are observed at relatively high frequencies within populations [56]. These “conserved extended haplotypes” or “ancestral haplotypes” can extend over 3.2 Mb, spanning the whole classical MHC [56-58].

Haplotypes differ between populations with greater diversity in African populations than in any other ethnic group. The 12^th IHW in 1996 highlighted the marked differences in MHC haplotype frequencies seen both between and within population groups. Among those of Asian ancestry particular haplotypes were very common in Japan such as HLA-A2-B52-DR15-DQ1 but of very different frequency in Singapore and Thailand where HLA-A2-Cw11-B46-DR9/DQ3 was common; in African populations the most common haplotype seen in South Africa HLA-A30-B42-DR1-DQ1 for example, is not found in neighbouring Zimbabwe where HLA-A30-B45-DR1-DQ1 is commonly found. Whether these population differences in haplotype frequencies correlate with variable prevalence of HLA-associated diseases has been little investigated at present. However, it has been suggested that the strikingly lower prevalence of type 1 diabetes in Asians compared with Caucasians might result from a parallel variation of HLA susceptibility allele frequencies in both of these ethnic groups [59, 60].

The first region for detailed study of linkage disequilibrium and recombination hotspots

The concept of linkage disequilibrium (LD) was described for the first time in the MHC by Bodmer, Piazza and Dausset [61]. LD refers to an association or correlation between alleles at two or more linked loci. Geneticists working on susceptibility to autoimmune diseases quickly became familiar with the issue as associations between autoimmune diseases and particular HLA loci are often the consequence of strong LD within the MHC. This facilitated the initial mapping of diseases but has made fine mapping of specific variants very difficult [62].

Studies on the H-2 complex in mice suggested different haplotypes could affect the frequency and location of recombination spots [63]. The analysis was more difficult in humans until high density microsatellite and SNP genotyping data was available and sperm typing became possible. Systematic studies showed that recombination and LD patterns across the MHC are not uniform [64, 65]. Two regions are poor in recombination events and show high LD: between HLA-B and HLA-C, and between HLA-DQA1 and HLA-DRB1. The recombination rate in class II and class III regions of 0.74 and 0.94cM/Mb is consistent with the average value of 0.9 cM/Mb, whereas it is only 0.31% in the class I region between HLA-A and HLA-B [66]. This is consistent with the existence of recombination hotspots [67, 68]. Indeed, random association of alleles in TAP1 and TAP2 genes suggested the existence of a recombination hotspot in this 15kb region [69], which was then formally confirmed in 40 families [64].

Using SSCP (single strand conformation polymorphism) analysis, the TAP2 hotspot was fine mapped to the second intron of the TAP2 gene [70], becoming at that time the most precisely defined region of recombination in the human genome. A combination of microsatellite typing and SSCP analysis, allowed to identify two other hotspots in the same region [68]. A first LD map covering 1.5 Mb, between BRD2 and HLA-B genes, was established in a Swedish population using STRs [71]. Analysis using SNPs and haploid genomes (sperm typing) then allowed high resolution mapping of the TAP2 recombination hotspot [72]. Its size of 1.2kb appeared similar to hotspots outside the MHC. It was more active in female than in male meioses and characterized by reciprocal crossovers. It is one of the best known hotspots in the human genome but interestingly it was not found in Chimpanzee genome [73].

Applied to the entire class II region, sperm typing using SNPs revealed three extended regions with high LD, dramatically decreasing around HLA-DOA, DMB and TAP2 [74]. The first two recombination hotspots could actually be resolved into three (DOA) and two (DMB) hotspots of size ranging from 1 to 2 kb and of activity varying between 0.4 and 140 cM/Mb. Nonetheless, all show a remarkably symmetrical distribution of recombination frequencies. These recombination hotspots hold almost all the recombination occurring within the MHC. LD block sizes of 60-90 kb are comparable to the ones described at the same time on chromosome 5 [75] or in the yeast genome [76], suggesting conservation of the mechanisms involved in hotspot distribution. This research culminated in 2002 with the publication of a high resolution meiotic recombination map covering 3.3 Mb of the MHC, with genotyping of 20,031 single sperm from 12 unrelated donors [77]. Six hotspots were found, accounting for 94% of the recombination events. They were separated by “coldspots” covering 87% of the MHC and matching LD blocks. Recombination rates, estimated at 0.49 cM/Mb in the MHC versus 0.92 cM/Mb in the whole genome, may significantly vary between non MHC identical individuals [78], and the recombination pattern is similar between sexes [77]. Hotspots are present at least every 0.8 Mb but are not necessarily evenly spaced. Between these hotspots, low levels of recombination are regularly observed every 100 kb, suggesting the presence of “warm-spots” delimitating the LD blocks. Sperm typing for the whole genome only became possible in 2008 and confirmed the existence of 1-2 kb hotspot regions with recombination rate heterogeneity between individuals [79]. However, recent research has shown the relationship between recombination hotspots and LD blocks is more complex. Recombination hotspots are in general found flanking LD blocks but this need not be the case.

Genomic architecture in the MHC compared to the rest of the genome

How does this detailed characterisation of genetic variation and LD in the MHC relate to the whole genome? The first integrated haplotype map of the MHC revealed that LD extends over larger physical distances in the MHC than in the rest of the genome, the average length being 31.1 kb versus 22.3 kb for the genome [58, 80]. Nonetheless, although physically larger, these blocks are actually shorter in terms of genetic distances (recombination rate of 0.49 cM/Mb versus 0.81 cM/Mb), leading to similar sizes of 0.012 cM versus 0.017 cM. Haplotypic diversity in MHC appears comparable to the rest of the genome except for classical HLA genes.

Higher resolution maps were then established, initially in population of European ancestry [57], then for HapMap populations of European, African, Chinese and Japanese ancestry by the MHC Consortium [81]. In addition to bringing new insights into the evolutionary dynamics of ancestral haplotypes, these two studies identified haplotype tagging SNPs which can be used as surrogates to genotype the classical HLA loci. These informative tagging SNPs are also used in the forthcoming IMAGEN (International Major Histocompatibility Complex and Autoimmunity Genetics Network) project aimed at refining MHC associations in several autoimmune diseases [9].

Maintenance of LD in the MHC

Little is known about the mechanisms maintaining LD in the MHC. Again, work in this area should provide clues to the rest of the genome. Three mechanisms may be involved. First, some allelic associations may have existed in a population of limited size and would have been present after successive generations without yet reaching equilibrium. Second, preferential allelic associations could involve epistatic mechanisms [82] and confer a selective advantage in a given environment. Third, haplotypes may be subject to differential recombination rates [83]. These mechanisms may co-exist and are overlapping in nature. We know that MHC alleles are subject to selective pressures although some suggest that recombination hotspots could be more significant for the LD pattern than population history, with less sharing of haplotypes between populations than in other genomic regions [83]. Numerous studies have suggested that LD and recombination rate of the same genomic region are variable between haplotypes. Observations in MHC are consistent with this hypothesis [84-87].

Potential mechanisms maintaining LD in the MHC are nicely illustrated by the ‘8.1 haplotype’ (or HLA-A1-B8-DR3 haplotype) which consistently shows the highest level with a half-length of LD above 3.5 Mb [87]. With a frequency of about 6%, this is the most common extended haplotype in Caucasian populations [88]. It may be maintained by positive selection of one particular allele, epistatic selection of a combination of alleles, or the existence of a mechanism inhibiting recombination in this specific haplotype.

The first genomic region entirely re-sequenced for common haplotypes

Sequencing single haplotypes has been a dream for geneticists, with attempts to separate parental haplotypes or to perform long-range allele-specific PCRs [89-91]. The latter is now dramatically improved by the next generation DNA sequencing technologies, with the pyrosequencing methods currently generating up to 450 nucleotide reads [92]. However, theoretical determination of haplotype sequence requires a massive coverage. The application of such technologies is more challenging in the MHC due to the extreme polymorphism of the region, notably gene duplication and structural variation. When whole genome re-sequencing studies have been undertaken, the reconstruction of the sequence MHC region remains one of the most difficult.

Fortunately, the MHC has already been the object of a unique re-sequencing approach performed at the haplotypic level by the MHC Haplotype Consortium led by Stephan Beck, Stephen Sawcer, John Todd and John Trowsdale [93]. The MHC Haplotype Project made use of DNA from individuals entirely homozygous for the MHC, as a consequence of consanguinity. Selected haplotypes were frequent in Caucasian populations and associated with common autoimmune diseases [94-96]. Sequences for the ‘8.1 haplotype’, the ‘18.2 haplotype’ [HLA-A26-B18-Cw5-DR3-DQ2] and the ‘7.1 haplotype’ [HLA-A3-B7-Cw7-DR15] were prioritised with the latter replacing the “mosaic” sequence from 1999 as the reference sequence for the MHC. The sequence of the remaining 5 haplotypes of this project was released in 2008 [96], microsatellites have been mapped to the alternate haplotypes [47] and all annotations are available on the Vega database (http://vega.sanger.ac.uk/).

The project revealed different gene content and polymorphisms on each haplotype, with genetic variation between haplotypes mainly concentrated in the classical class I and class II HLA loci and major sequence differences in the complement region (RCCX module) and HLA-DRB. The approach has been extended to 46 other haplotypes [97]. It provided a better understanding of the 8.1 haplotype structure with an estimated age of 35,500 years matching the population migration to Europe. It also revealed new rare SNPs, notably for the class III region, and provided evidence in favour of the rare-SNP hypothesis for MHC associated diseases. An ongoing project sequencing 200 MHC haplotypes using next generation sequencing technology should further advance our knowledge in this area (www.molgen.mpg.de/~genetic-variation/Projects.html).

THE MHC AND GENETIC SUSCEPTIBILITY TO DISEASE: INSIGHTS FOR THE REST OF THE GENOME

The pre-eminent genomic region associated with disease

Point mutations, deletions and other variants within the MHC are responsible for monogenic diseases showing Mendelian inheritance such as haemochromatosis [98] or congenital adrenal hyperplasia [99]. More significant however was the role played by this region of the human genome in many common multifactorial diseases. Serological typing established clear and dramatic evidence of association by the early 1970s between specific HLA antigens and autoimmune diseases such as ankylosing spondylitis [100-102], psoriasis [103] and celiac disease [104, 105]. The importance of the MHC was subsequently robustly demonstrated using microsatellites, single nucleotide polymorphisms (SNPs) and other genetic markers for a range of diseases, and continues today in the era of genome-wide association studies [106]. Mainly autoimmune diseases are associated with genetic variation in the MHC, including the major risk loci for type 1 diabetes [107-109], rheumatoid arthritis [110-112] and multiple sclerosis [113-115] while other disease associations include with infectious and inflammatory diseases [44, 116] as well as cancer [117-119], smoking behaviour [19] or mating preferences [20].

For the majority of disease associations in the MHC, it is the haplotype which is implicated, as opposed to specific variants. Notably, common ancestral haplotypes may be associated with several diseases [120]. Among them, the 8.1 haplotype is the most remarkable because of its association with a very large number of immune phenotypes [121] and diseases [122], such as type 1 diabetes [123], myasthenia gravis [21], rheumatoid arthritis [124], systemic lupus erythematosus [125], celiac disease [126] or IgA deficiency [127]. Fine mapping such haplotypic associations has proved difficult because of the extent of LD which hampers the identification of the causative variants

How the HLA indirectly contributed to create the first physical and genetic map of the human genome

Thanks to the shared Nobel Prize in 1980, Jean Dausset received in 1984 for his laboratory a huge legacy by an art collector, Helène Anavi. With this unexpected gift, he created in collaboration with Howard Cann and Daniel Cohen, the Centre d’Etude du Polymoprhisme Humain (CEPH) [128], which soon after became Foundation Jean Dausset-CEPH. Inspired by the MHC association studies, Dausset and Cohen decided to look at polymorphism across the whole genome to generate markers that would segregate in families to investigate linkage in multiple diseases. The collection included several large families from numerous collaborators, including the large Utah pedigrees studied by Ray White [129]. This CEPH consortium led to the publication of the first microsatellite genetic and physical maps of the human genome [130-132], an essential prelude to enabling the Human Genome Project [133]. This invaluable resource opened a new area in genetic studies, allowing the identification of hundreds of disease susceptibility genes. Today, the CEPH samples are still intensively studied and played a key role in recent genomic advances such as the International HapMap project [134] or the ongoing ‘1000 genomes’ sequencing project (http://www.1000genomes.org).

GENETICAL GENOMICS AND MODULATION OF GENE EXPRESSION IN THE MHC

It has been recently demonstrated that variation in gene expression in humans is heritable [135-138] and genetic determinants can be defined by linkage or association analysis [139]. Approaches developed in model organisms are now applied in humans to map expression quantitative trait loci (eQTLs) by so-called “genetical genomics” [140]. This is an important tool to identify functional variants in disease susceptibility genes [141], particularly when combined with genome wide disease association studies, as illustrated by recent work on asthma [142] or high-density lipoprotein cholesterol [143]. Given its many disease associations and remarkable genetic polymorphism, as might be expected, the human MHC has been shown to be a major site of association for eQTLs (summarized in Table 1) [137, 138, 143-151].

Table 1.

eQTL mapping studies showing strong evidence of linkage or association with the MHC

Study	Description	Expression phenotypes at MHC loci
Cheung et al. 2003	Lymphoblastoid cell lines (LCLs) from 35 unrelated Utah CEPH samples	HLA-DRB1 and RNF5 among the 40 most variable genes between individuals.
Pastinen et al. 2004	LCLs from 63 unrelated individuals Quantification of allelic transcript ratios at 193 SNPs in 126 genes.	MICA among the 23 genes showing allelic imbalance (ratio of 65:35 in 31 heterozygous samples).
Monks et al. 2004	LCLs from 15 families CEPH families Expression measured for 23,499 genesx in 167 individuals 346 autosomal markers genotyped genome-wide linkage study	33 genes with significant linkage (P < 5×10−6) including 4 clustered MHC genes: HLA-DPB1, HLA-DRB3, HLA-DRB5 and HLA-G.
Cheung et al. 2005	LCLs from57 unrelated individuals analysed from CEPH families cis-association analysis performed with the 374 genes showing evidence of linkage (P < 3.7×10−5) in a genome-wide linkage study by Morley et al. 2004. Genome-wide association using >770,000 SNPs (HapMap genotypes) on the 27 genes showing the strongest evidence of cis-linked determinants	Association of rs6928482 (P = 6.5 × 10−11) at the HLA-DRB2 gene where cis-linkage had been previously found (P < 10−11).
Spielman et al. 2007	LCL from 60 European (CEPH), 41 Han Chinese and 41 Japanese individuals from the HapMap project 4,197 genes tested 2 million SNPs	10,097 showing significant difference at P < 10−5 between European and Asian-derived samples, including 11 MHC genes: HSPA1A, GLN1, LTB, BAT2, NEU1, LTA, HLA-DOB, MSH5, PSMB9, PSMB8, STK19.
Stranger et al. 2007	LCLs from 270 individuals genotyped in the HapMap project: 30 Caucasian trios, 45 unrelated Chinese, 45 unrelated Japanese and 30 Yoruba trios. 2.2 million SNPs genotyped	1,348 genes with cis-1Mb associations (0.001 permutation threshold) including TUBB, HLA-C, HLA-DQA2, HLA-DQA in CEU samples, and even HLA-DRB5 when comparing the 4 populations. 180 genes with trans associations ( > 1 Mb) including rs7575 (on chromosome 22) with MICA in Chinese samples, and multiple SNPs on chromosome 6 with HLA-C between 4 populations.
Dixon et al. 2007	LCLs from 400 children of families with a proband with asthma. 20,599 genes phenotyped 408,273 SNPs	MHC loci with lod score for cis-association > 6 included HLA-DQA1 (lod score 29.6), HLA-DRB1 (23.8), DPA1, DQB1, DQA1, DQA2, HLA-C and HLA-A
Göring et al. 2007	Primary blood lymphocytes from1,240 individuals, most of them being members of30 extended families 18,519 genes phenotyped 432 highly polymorphic microsatellites	Among the 20 highest cis lod scores were HLA-DRB3 (lod score 33.8) and HLA-DRB5 (33.1)
Kwan et al. 2008	LCLs from 60 CEPH founders 17,897 genes phenotyped with Affymetrix Exon 1.0ST arrays HapMap phase II SNPs	324 genes with significant associations, including TAP2 with rs3763355 (P = 1.98 × 10−13) reflecting an alternative splice site resulting in differential termination of the open reading frame. This was validated by RT-PCR.
Emilsson et al. 2008	Blood from 1,002 individuals in 209 families Subcutaneous adipose tissue from 673 individuals in 124 families 23,720 transcripts quantified 1,732 microsatellites genotyped in both cohorts for linkage studies 317,000 SNPs genotyped in a subset of 150 unrelated individuals for association analysis.	1 possible hotspot of multiple associations in trans (FDR 5%) at rs2395309 between HLA-DOA and - DPA1 detected in blood and /or adipose tissue.

These associations are striking but care in interpreting such results in the context of the MHC is appropriate for a number of reasons, notably relating to the technology used in analyses of gene expression. First, the commercial expression microarrays used are designed against the reference sequence of the human genome, so do not take into account genetic and haplotypic diversity which is particularly significant in the MHC. As a consequence, individuals carrying haplotypes other than the reference haplotype (for the MHC, HLA-A3-B7-DR15) may not show expression of certain genes because the designed probes do not match the gene sequence of their particular haplotypic background [152]. This could skew the distribution of the expressed gene and may lead to false positive eQTL associations. The second issue is that the arrays used in almost all human eQTL studies to date only target the 3′ end of the transcripts, hence miss a large number of alternatively spliced gene isoforms. Alternative splicing is common across the genome, including the MHC, and is itself modulated by DNA sequence diversity, further compounding these difficulties [146, 153-157]. Methods quantifying accurately, both qualitatively and quantitatively, all transcripts in a haplotypic specific manner - the “haplo-spliceo-transcriptome” [158] - are required to investigate the genetic basis of gene expression in the human MHC. Further molecular and cellular approaches are needed to identify regulatory variants modulating gene expression within haplotypes. Such efforts are facilitated by current projects to define the regulatory landscape of the human genome such as the ENCODE (ENCyclopedia Of DNA Elements) Project [159] and a variety of approaches based on animal models, assays of chromatin accessibility and modifications, protein-DNA interactions and gene expression [160]. For example, at the TNF gene locus, we recently defined novel regulatory elements based on quantitative chromatin profiling [161] while earlier work resolved specific promoter SNPs modulating affinity of transcription factor binding [162, 163]. A further approach is to consider allele-specific gene expression which occurs relatively commonly and is heritable: this approach has been extended from analysis based on relative transcript abundance to considering RNA polymerase loading, allowing resolution of regulatory variants [144, 162, 164]. For example, using the haplotype specific chromatin immunoprecipitation (haploChIP) approach, differences in gene expression were resolved to LTA rather than TNF within an LD block spanning these two genes [164].

Other studies have combined ChIP with RNA-FISH and chromosomal capture conformation at the HLA-DRB1 locus to reveal a new mechanism of regulation of MHC class II genes involving the insulator CTCF binding factor and a particular chromatin long-distance loop configuration [165]. Some specific associations of chromosome territories with nuclear proteins were invoked [166]. An open chromatin configuration has been described in relation to the timing of DNA replication for the MHC which occurred early in S-phase [167]. The generation of a 2 kb resolution tiling path MHC array [168] has also allowed chromatin profiling and identification of cell-specific genomic anchors which show dynamic changes on transcriptional activation [169]. This MHC array has also allowed analysis of methylation across the MHC to define 90 differentially methylated regions using DNA from liver, placenta, CD+ lymphocytes and sperm. This builds on data in which the MHC was studied as the pilot region for the Epigenome project [170-173], which had revealed that 10% of MHC loci had a cell-specific methylation pattern although the relationship of methylation to haplotypic background currently remains unclear.

CONCLUSIONS

The MHC has intrigued, enthralled and challenged several generations of researchers. It is a remarkable region of the human genome which has indeed proved a paradigm for many different aspects of human genetic research. In the genetics Olympics, the MHC would hold many records notably in terms of extreme genetic variation, gene density, biological significance and disease association. Rightly, much research effort has been focused on trying to dissect the complexities of the MHC and many general lessons have been learnt, notably relating to the nature and inheritance of genetic polymorphism. We have seen how this was the first substantial genomic region to be sequenced and the first to be clearly delineated in terms of its genomic landscape, pioneering our understanding of linkage disequilibrium, recombination and haplotypic structure. Despite this, progress in fine mapping disease associations and resolving specific functional variants remains difficult. It is now clear that both structural and regulatory variants are important and may indeed be operating in tandem. Advances in functional genomics to map and characterise polymorphisms modulating gene expression should help advance this field. However the complexities are many-layered, from sequence level diversity to gene regulation, epigenetic and environmental factors. Effects will be context specific, relating to particular cell or tissue types and indeed developmental stages, as well as particular environmental stimuli or other conditions. The MHC will undoubtedly continue to challenge us and has many secrets yet to be revealed which will be of relevance to the rest of the genome. However an integrated approach, building on current successes, should ensure research into the MHC remains at the avant-garde of human genomics.

Key points.

1. The MHC is the most polymorphic and gene dense region of the human genome

2. Genetic variation in the MHC is associated with more diseases than any other genomic region

3. The MHC has been a model to study linkage disequilibrium and haplotypic structure

4. This is the first region of the genome to be sequenced for common haplotypes

5. Current efforts to map expression quantitative trait loci highlight the importance of genetic variation in the MHC