Sequence and Phylogenetic Analysis of the Untranslated Promoter Regions for HLA Class I Genes

The MHC is the most polymorphic locus in the human genome (1, 2), and it is also the richest segment genome-wide in terms of disease associations, including bacterial, viral, inflammatory, and autoimmune diseases, cancer, neurologic disorders, and drug hypersensitivity (3). Assigning the mechanism for these disease associations is confounded by the extensive polymorphism, dense gene content with strong linkage disequilibrium between pairs of genes, and functional similarity of the molecules encoded by genes within the MHC.

Several studies have indicated that expression levels of certain HLA loci vary in an allotype dependent manner (4–16) and this variation has been shown to associate with certain diseases, such as outcome after HIV infection (4, 17), Crohn’s disease (4), HBV clearance (8), graft-versus-host disease after unrelated hematopoietic cell transplantation (18, 19), Parkinson’s disease (20), systemic lupus erythematosus (21), and certain cancers (22–24). The regulation of HLA class I expression is mediated by a series of core promoter motifs, which are located within 500 bp upstream of the start site and are relatively conserved (25). Binding sites for regulatory proteins further upstream from the core promoter have also been identified (26, 27) and it is likely that others exist, some of which may be polymorphic. Indeed, genetic variants and epigenetic modifications located within the 3′ and 5′ untranslated regions (UTRs) (6, 13, 25, 28, 29), as well as the coding region (14), have been shown to alter HLA allele expression levels, but so far, these mechanisms account for only a small percentage of the expression variation found within the respective genes.

In this study we provide extensive allelic sequence information for the 5′ UTR and upstream promoter regions of the classical HLA class I genes and the homologous region for one nonclassical gene, HLA-E. Based on these sequences, the intralocus nucleotide diversity and the phylogenetic congruence between the 5′ UTR promoter and the gene region sequences for each of the three classical class I loci were determined. Expression variation observed among the 57 HLA class I lineages was examined in relation to promoter sequences. These data serve as a basis for functional studies aimed at identifying the variants that are involved in differential HLA allelic expression levels.

Cell lines, homozygous at the HLA class I loci, from the Tenth International Histocompatibility Workshop were selected for HLA class I promoter sequencing (30). These cell lines represent common alleles that predominantly originate from European American (EA) ancestry. We also selected common African alleles for promoter sequencing from homozygous donors recruited from the Frederick National Laboratory for Cancer Research Donor Program. At least two donors homozygous for a particular allele were selected for study.

Primers specific for each HLA gene interrogated (HLA-A, -B, -C, and -E) were designed using online Primer 3 software (31). The targeted region covering ∼2 kb upstream of the transcription start site was amplified with locus-specific primers (Supplemental Table I) using the following PCR conditions; 95°C for 5 min, 40 cycles of 95°C for 15 s, 60/63°C for 15 s and 72°C for 1 min, finally 72°C for 1 min. An annealing temperature of either 60 or 63°C was used depending on the locus. The purified PCR products were sequenced in both directions with at least two overlapping primer sets using the ABI-3130XL DNA analyzer (Applied Biosystems).

Sequences were analyzed using Sequencher (Gene Codes) and Mutation Surveyor (Soft Genetics) software and aligned with BioEdit (http://www.mbio.ncsu.edu/bioedit/bioedit.html) and Clustal. All donors homozygous for a specific HLA coding region were also homozygous in the 5′ UTR sequence. Previously identified transcription factor binding sites (TFBSs) were highlighted using the BioEdit program.

Putative TFBSs were obtained using the online software Alibaba 2.0 (32) with default settings.

Sequences were deposited to the GenBank database with the accession numbers KY427751–KY427809 (https://www.ncbi.nlm.nih.gov/genbank/).

Healthy EA donors from the Research Donor Program at the Frederick National Laboratory for Cancer Research were used to measure the HLA-B (n = 178) and HLA-E (n = 27) mRNA expression levels using quantitative PCR (qPCR). Primers are located in exon 4 and exons 2–3 for HLA-B and -E, respectively. HLA-B, -E, and the housekeeping gene β-2 microglobulin (β2M) were amplified with the PCR conditions previously described (5, 6). Primer specificity was verified by both melt curve analysis and sequencing of HLA-B and -E PCR amplicons. PCR amplification efficiencies were tested for all common HLA-B (n = 18) and -E (n = 2) allelic lineages. HLA-B and -E expression levels were normalized to β2M and calculated using the 2^−∆∆Ct method. Normalizing HLA-B and -E expression levels by the housekeeping genes GAPDH, 18s, and β-Actin show very consistent results to those using β2M (data not shown).

Primer-BLAST checks and sequencing of PCR amplicons were performed to verify that only the specific targeted lineage corresponding to each given pair of HLA-B and -E primers was amplified. To further confirm accuracy of the HLA-B primers, an additional set of primers within the exon 5–7 region of the HLA-B locus was designed. The PCR amplification efficiencies and HLA-B specificity for the exon 5–7 primer set were shown to be equal across HLA-B allelic lineages.

A comprehensive alignment, including the sequences generated in this study, full length genomic sequences obtained from the IMGT/HLA database (1), and sequences identified by Primer-BLAST searches of nucleotide databases, was constructed for phylogenetic analyses. The completed alignment consisted of up to 5.3 kb, which encompassed promoter (∼2 kb) and coding regions (∼3.3 kb) for 68 alleles from 57 lineages of the three classical HLA class I loci. Phylogenetic trees were constructed using the neighbor-joining method (33) based on nucleotide distances calculated by Kimura’s 2-parameter method (34) across relevant segments of the alignment. Gaps were excluded on a pairwise basis.

Nucleotide diversity for homologous positions within the promoters of HLA-A, -B, and -C was estimated using Kimura’s 2-parameter method (34) and examined by looking at intralocus nucleotide diversity within small (150 bp) windows with sliding window analysis across the alignment as previously described (35). Single representatives from the 57 distinct lineages of HLA class I genes were included in the sliding window analysis.

The estimated mean expression levels of individual HLA-B (178 donors) and HLA-C (256 donors) lineages were calculated using the linear model function from the linear regression R package (36) as previously described for HLA-A mRNA (6). All donors were of European descent. The magnitude and range of expression differences across the three loci are likely related to consequent phenotypic effects at each HLA locus. Although standard units are used to measure intralocus expression differences, comparisons between loci require a normalization procedure to correct for the different experimental measurement procedures developed for each locus. To standardize the measurements across loci, the coefficient of variation (CV) was determined by estimating the ratio of the SD to the mean of expression in a cohort of individuals typed at each of HLA-A, -B, and -C (n = 172). The CV can be either directly estimated from each of the 172 individual measurements of expression at each locus or alternatively, it can be estimated by utilizing mean effects of allelic lineages. In the first approach, expression was experimentally determined by qPCR for every individual at each locus allowing direct calculation of means, variances, and CV. For the second approach, expression was inferred by combining the mean effects of allelic lineages that make up the known genotypes of each individual. The individual’s expression level is thus determined from the standard expression levels of the alleles making up his/her genotype. The latter approach should be less sensitive to the increases in CV due to experimental errors.

Spearman correlations and Mann–Whitney U test were performed using GraphPad Prism software version 6 (GraphPad Software, San Diego, CA).

A region of 1.8–2 kb extending upstream of the translation start site (TSS) was sequenced for a total of 68 alleles from the three HLA classical class I genes: 19 HLA-A, 33 HLA-B, and 16 HLA-C alleles. The 68 alleles group into 57 distinct allelic lineages, comprising 17 of 21 known lineages at HLA-A, 27 of 36 known lineages at HLA-B, and 13 of 16 known lineages at HLA-C. Two or more donors homozygous for each allele were sequenced and covered the common allotypes present within individuals of European and African descent. A previous report provided sequences 859 bp upstream of the TSS for the 16 HLA-C lineages (37), which overlaps with about half of that provided herein, and these 859 bp sequences are identical to those determined in this study for the same segment of DNA.

Variation in TFBSs of the HLA gene core promoter (38–42) was identified between allelic lineages of HLA-C (Fig. 1A), where HLA-C*03, -C*07, and -C*17, which were previously identified as having low expression levels (4), possess variation in at least one of the eight HLA-C core promoter motifs. Relative to consensus, the HLA-C*03 lineage contained a single nucleotide change in the TATA binding site, whereas the HLA-C*07 lineage contained a single nucleotide change within the Enhancer A (ĸB2) binding site and a deletion polymorphism in the ISRE binding site. The HLA-C*17 lineage contained single base changes within both the Enhancer A (ĸB1) and TATA binding sites, and two base changes within the Enhancer A (ĸB2) binding site. Furthermore, the HLA-C*17 lineage possesses a 7 bp deletion located at position −37 to −43 bp upstream of the TSS, and an insertion of 6 bp is observed at position −181 to −187 (Fig. 1A). The potential for these variants to explain the low cell surface expression of HLA-C*03, -C*07, and -C*17 will require experimental examination.

Among HLA-A lineages, the core promoter TFBSs are highly conserved. A single variant located at position −208 bp upstream of the TSS is observed in the Enhancer A (ĸB1) binding site for the HLA-A*30 lineage (Fig. 2A). Variation in the HLA-B core promoter TFBSs is also limited with a few nucleotide changes in the ISRE, S (H)-box, and X1/X2 binding sites (Fig. 3A). In general, however, the HLA-B promoter region as a whole is more polymorphic than that of HLA-A or -C, in keeping with the relative polymorphism across their coding regions.

The nucleotide diversity observed within the MHC region, ranging upwards of several percent for classical MHC genes, has previously been shown to be far higher than the genome average of ∼0.1% (43). This holds true for the regulatory region ∼2 kb upstream of the initiation codon of the classical class I genes, as well, where the intralocus nucleotide diversity across the region for the pool of HLA-A (Fig. 2B), -B (Fig. 3B), and -C (Fig. 1B) lineages is on average ∼1.8, ∼1.9, and ∼0.9% respectively, roughly following the diversity within the coding regions. Divergence across the three loci is also evident, in part due to the presence of insertion elements, which were identified in the University of California Santa Cruz browser (https://genome.ucsc.edu) as long interspersed nuclear elements MLT1E3 and L2 within HLA-A, and L1M4a1 within HLA-B and -C. Surprisingly, the most conserved regions for HLA-B and -C are not those of the core promoter region where multiple well-characterized TFBSs are positioned, but rather positions −700 to −500 bp, which are more conserved in HLA-C than any other 200 bp window examined across the remainder of the promoter region (generally above 1% diversity) (Figs. 1B, 3B). The lowest HLA-B diversity of ∼0.2% is located from positions −1720 to −1620 bp (Fig. 3B) and that for HLA-A of ∼0.5% is located near −500 bp (Fig. 2B).

The phylogenetic tree of allelic lineages for the HLA-A gene constructed from promoter sequences is generally congruent with that found for the remainder of the gene region, indicating that recombination has not scrambled the sequences extensively. Seven promoter clades are identified (Fig. 4A), consisting of A*02/A*34/A*68; A*25/A*26/A*66; A*23/A*24; A*31/A*33; A*29/A*32/A*74; A*01/A*11; and A*03/A*30 lineages. These clades are also recovered for the remainder of the gene (Fig. 4B), apart from the A*34 lineage being located in the A*25/A*26/A*66 clade. The relative position of the A*23/A*24 clade also differs between promoter and gene coding segments. Linking previously published HLA-A mRNA expression levels (6) with the phylogenetic trees indicated that certain lineages with similar or identical promoter regions possess similar mRNA expression levels, such as A*23/A*24 and A*31/A*33 (Fig. 4C), whereas others, such as A*03 and A*30, do not. A*03 and A*30 show a 10-fold difference in average lineage expression levels, which may result in part from the eight variable sites observed between the two lineages within the ∼2 kbp promoter region (Supplemental Table II).

Nine clades were identified within the HLA-B promoter tree (Fig. 5A): B*35/B*51/B*52/B*53/B*58; B*18/B*27/B*37; B*40/B*41/B*45/B*49/B*50; B*15/B*46; B*55/B*56; B*13/B*44/B*47; B*14/B*38/B*39; B*07/B*08/B*42; and B*57 (the latter being the only lineage-specific clade). The tree derived from coding sequences recovers the B*35/B*51/B*52/B*53/B*58 clade with incorporation of the B*57 lineage into the clade (Fig. 5B). The B*15/B*46 and B*55/B*56 pairings also reflect the promoter clades. However, the promoter-based B*40/B*41/B*45/B*49/B*50 clade splits, where B*40/B*41 groups with B*07/B*08/B*42 in the coding region. Likewise, where the promoter forms a B*13/B*44/B*47 clade, B*27/B*44/B*47 form a clade in coding regions and B*13 forms an isolated tree branch. These and other inconsistencies between the promoter and the coding region trees can be explained by infrequent recombination events that have displaced some, but not all, relationships. A parsimonious reconciliation of the trees requiring the fewest recombination events suggests that lineages sharing inconsistency arose after their common ancestral sequence underwent recombination, giving rise to novel associations across promoter and coding regions.

HLA-B mRNA expression levels were determined in 178 EA donors, and unlike allele-specific differential expression at HLA-A and -C, we observed very minor differential expression across alleles (Fig. 5C). A fold change of only 1.2 was observed between the CV expression levels of the highest (B*13) and the lowest (B*52) lineages at HLA-B (Fig. 5C, 5D), whereas the fold change within the HLA-A and -C loci is 3.8 and 2.2, respectively, when comparing the highest and lowest CV for expression levels at each of these loci.

HLA-B mRNA expression levels were measured using a set of qPCR primers that are specific only to HLA-B lineages, as determined by sequencing and online Primer-BLAST software (data not shown). The primer set, which targets exon 4, was shown to amplify all common HLA-B lineages with equal efficiency (Supplemental Fig. 1A), and no correlation was observed between PCR efficiency of the primers for each HLA-B lineage and average expression level of each lineage (Spearman correlation coefficient; R² = 0.22, p = 0.36, Supplemental Fig. 1A). A second set of qPCR primers spanning HLA-B exons 5–7 was generated, which amplified all HLA-B lineages with equal efficiency and validated the data obtained using exon 4 primers (data not shown). A significant positive correlation between HLA-B mRNA expression levels using exon 4 versus exon 5–7 primer sets was observed for 61 EA healthy donors (R² = 0.81, p < 0.0001, Supplemental Fig. 1B). HLA-B mRNA expression levels were also shown to be quite stable in 30 EA donors for whom measurements were repeated 3 y after the initial measurement (R² = 0.72, p < 0.0001, Supplemental Fig. 1C).

The HLA-C phylogenetic trees for the promoter region and the coding region show very similar clade structure (Fig. 6A versus 6B), the two main differences involving the C*03 lineage, which groups with C*01/C*04/C*14 in the coding region, but forms a distinct (though weakly supported) clade in the promoter region, and C*02, which groups weakly with C*15in the coding region, but again forms a distinct weakly supported clade in the promoter region. Similar to HLA-A (6) and HLA-B (Supplemental Fig. 1C), the mRNA expression levels of HLA-C are stable across time, as observed in 68 EA donors whose expression was measured 3 y apart (2012 and again in 2015) (R² = 0.62, p < 0.0001, Supplemental Fig. 1D). C*01 and C*14, which are distinguished by only four nucleotides in the promoter region, are the two lineages with the greatest difference in mRNA expression levels (2 fold) within a promoter region clade (Supplemental Table II). These two lineages differ at position −824, where an Sp1 is predicted for the C*01 lineage (Alibaba 2.0 software (32)), but not the C*14 lineage, which could explain the difference in their expression levels.

It is notable that despite the allele-specific variation in expression level for HLA-A and HLA-C mRNA (5, 6), the extent of promoter region variation between pairs of alleles overall does not correlate with differential mRNA expression levels of those alleles. There is no clear link between the phylogenetic structure as represented by pairwise nucleotide differences in the promoter region with estimated differences in mRNA expression levels observed for HLA-A (Spearman correlation coefficient; R² = 0.003, Fig. 7A), HLA-B (Spearman correlation coefficient; R² = 0.0007, Fig. 7B), and HLA-C (Spearman correlation coefficient; R² = 0.025, Fig. 7C). Note that the nonindependence of pairwise comparisons in Fig. 6 (due to shared phylogeny) indicates that the significance values are not valid. Nevertheless, these data suggest that only very few specific variants in the promoter region are likely to account for differential expression across lineages.

Promoter sequences for the HLA-A, -B and -C genes form distinct locus-specific clades (data not shown) in a manner that largely reflects locus specificity of the coding region. The common ancestral locus that gave rise to MHC-A, -B and -C underwent duplication to give rise to two loci, one of which was the ancestral MHC-A, whereas the other locus underwent duplication at a later date, giving rise to the ancestral MHC-B and -C loci (44). These relationships are reflected in the promoter tree where HLA-A branches outside of an HLA-B and -C clade and indicate that interlocus conversion events have not noticeably contributed to the diversity of the promoter. This absence contrasts to findings in the 3′ UTR where interlocus conversion between HLA-B and HLA-C is strongly implicated in microRNA-induced control of expression (35).

Expression levels of HLA-A, -B, and -C mRNA were measured in 172 EA individuals and tested for correlations between pairs of loci. Very poor correlations in expression levels were detected between the pairs HLA-A and -B (Spearman correlation coefficient; R² = 0.0002, p = 0.87, Fig. 8A), HLA-A and -C (Spearman correlation coefficient; R² = 0.02, p = 0.05, Fig. 8B), and HLA-B and -C (Spearman correlation coefficient; R² = 0.02, p = 0.07, Fig. 8C). These data indicate that unlike the HLA-class II genes, which are regulated by common factors and display strong correlations in expression levels across loci (29, 45), under healthy conditions, the classical class I genes are regulated largely independently, presumably by at least some distinct mechanisms.

The CV for expression of HLA-A, -B, and -C was estimated in these 172 individuals for whom complete three-locus data were available, and used to determine the relative variation in expression for the three loci. Both observed and expected CV, where expected values are based on regression estimates of allelic effects for each genotype, were determined. The CVs increased in order of HLA-B (lowest) to HLA-A (highest) and the ratio relative to HLA-B is 1–2.84 (HLA-C) to 3.5 (HLA-A) for observed, and 1–4 (HLA-C) to 5.6 (HLA-A) for expected expression. Observed expression fully incorporates experimentally determined variation, whereas experimental noise can be minimized using expected variation.

No mRNA expression variation was observed between the only two common HLA-E alleles, HLA-E*01:01 and -E*01:03, as measured by qPCR in 27 individuals (Supplemental Fig. 1E, p = 0.95). The qPCR primers amplified both alleles with virtually the same PCR efficiency of 1.95 and 1.96, respectively. HLA-E*01:01 and -E*01:03 differ in the coding region by only a single nonsynonymous nucleotide change, and in keeping with their high sequence similarity, the ∼1.2 kb sequence upstream of the TSS for these alleles was identical (data not shown).

Differential HLA expression levels have recently been shown to associate with a number of diseases (4, 5, 8, 18, 19, 46), and the sequences provided herein serve as an initial step in identifying novel variants responsible for allele-specific expression variation. Despite substantial variation in the promoter regions of the HLA class I genes, which reflects that observed in the coding region of these loci, a Spearman correlation test revealed no strong correlation between similarity in promoter sequence and expression levels between pairs of alleles, suggesting that only a limited number of promoter region variants are likely to impact allele-specific expression variation.

Three HLA-C lineages, HLA-C*03, -C*07 and -C*17, possess the lowest expression levels and also contain variants within confirmed TFBSs of the core promoter. These three lineages all carry the intact microRNA-148a binding site in their 3′ UTR, which has been shown to reduce HLA-C expression levels (28). It is possible that the disrupted core promoter, along with microRNA regulation of these three alleles, results in their low expression levels relative to other HLA-C alleles, highlighting multiple distinct mechanisms determining differential expression of the HLA-C alleles.

Despite high levels of nucleotide diversity found in the HLA-B promoter region, we observe a low level of HLA-B expression variation across donors and find no evidence for allele-specific mRNA expression levels for HLA-B, further distinguishing HLA-B from HLA-A and -C. The stability of HLA-B mRNA expression levels measured 3 y apart indicates that expression differences across subjects are genuine and are not simply due to measurement fluctuation. The lack of allele-specific expression along with the stability of HLA-B mRNA expression levels suggests that trans, rather than cis-polymorphism may modify HLA-B expression. Although we find no evidence for HLA-B lineage-specific mRNA expression differences, recent reports suggest that cell surface expression levels of HLA-B do vary in a lineage-specific manner to some extent (47, 48), which may result from lineage-specific translational events. Using three Abs (Tu149, B1.23.2, and 22E1), Chappell et al. (48) showed that the surface expression levels of HLA-B*57, -B*27, -B*07, and -B*35 inversely correlated with promiscuity of peptide binding, where B*57 was expressed at the highest level, followed by B*27, B*07, and B*35 in descending order. A second study used the anti-Bw4 Ab clone 0007 to measure the surface expression of Bw4 alleles, including HLA-B*57 and -B*27 (47). The anti-Bw4 Ab, however, indicated that HLA-B*27 has higher expression than -B*57 (47). These studies were performed with different cell types, which may result in a distinct hierarchy of HLA-B allotype expression levels. As differential cell surface expression levels are likely to impact HLA-B restricted T cell responses, it will be important to pursue this line of study more fully.

Differential allele-specific expression levels have been observed at loci within the MHC, including the TNF, LTA, HSP70, ZFP57, and HLA-DRB1 genes (48–50). This regulation is thought to involve a complex interplay between genetic and epigenetic mechanisms (50), which is also likely to be true for HLA-A, where methylation levels are known to associate with differential expression levels of HLA-A alleles (6). Interestingly, methylation does not appear to regulate expression levels of HLA-B or -C.

A highly efficacious vaccine against SIV in rhesus monkeys has been shown to elicit MHC-E–restricted CD8⁺ T cell responses against a broad array of SIV peptides (49, 50). These responses are likely to explain in part the remarkable protection conferred by this vaccine, sparking interest in considering HLA-E in vaccine design as a mediator of protection against HIV-1. There are only two common alleles of HLA-E, -E*01:01 and -E*01:03, which differ by a single amino acid (Gly to Arg at position 107, respectively). HLA-E*01:03 is expressed at a higher level on the cell surface relative to HLA-E*01:01 (15), which may affect their functional activity. We found no evidence for allele-specific expression variation in HLA-E transcript levels and no difference in the sequence ∼1.2 kb upstream of the TSS between these two common lineages, suggesting that any variation in cell surface expression of the allotypes are due to posttranslational modification.

Regulation of HLA class I expression levels appears to occur at both the transcriptional and translational level, depending on the locus. The work described herein provides a foundation for greater understanding of the mechanisms responsible for differential regulation of class I allelic expression, which may lead to the ability to manipulate HLA expression levels in diseases where class I expression levels are known to associate with outcome.

We thank the participants from the Research Donor Program at the Frederick National Laboratory for Cancer Research.

The authors have no financial conflicts of interest.