Factors affecting HLA expression: A review

2.1 CIITA and NLRC5

The principal HLA gene promoter region is in the 5′ untranslated region (5′UTR), up to 300 bp upstream of the transcription initiation sequence (Handunnetthi, Ramagopalan, Ebers, & Knight, 2010; Wong et al., 2014). All MHC genes share this highly conserved regulatory module consisting of a typical “SXY” box known as the enhanceosome (Elsen et al., 1998; Meissner, Liu, et al., 2012; Gobin et al., 2015. see Figure 1). In addition, HLA class I genes contain the cytokine-inducible promoter site EnhA and the IFNγ-stimulated response element, ISRE (Van Den Elsen et al., 2004; Meissner, Liu, et al., 2012; Neerincx et al., 2013). The enhanceosome also permits binding of thyroid-stimulating hormone (TSH) which acts to repress RNA transcription (Howcroft, Raval, Weissman, Gegonne, & Singer, 2003).

Studies in patients with bare lymphocyte syndrome (BLS) have shown that critical deficiencies in any of the proteins binding to this promoter region will result in failure of HLA expression (Elsen et al., 1998), and the predominant moderators of HLA transcription are the NOD-like receptor (NLR) proteins NLRC5 and CIITA which act through the enhanceosome (Meissner, Li, et al., 2012; Meissner, Liu, et al., 2012).

In studies of BLS, the class II transactivator (CIITA) was identified as an essential component of HLA class II mRNA transcription (Muhlethaler-Mottet, Otten, Steimle, & Mach, 1997). CIITA expression was only detectable in cells expressing HLA class II, and levels of CIITA and HLA class II were correlated. Whilst CIITA appears to have some influence on class I expression, Kobayashi and van den Elsen (2012) were able to show that HLA class I is specifically induced by the protein known as “NOD-, LRR- and CARD-containing 5” (NLRC5) acting through the SXY module in the class I leader sequence in a dose-dependent manner.

Both CIITA (found on chromosome 16p13) and NLRC5 (chromosome 16q13) are themselves highly regulated and different promoter regions control basal and INFɣ-induced upregulation of these proteins (Kobayashi & Elsen, 2012; Wong et al., 2014). It therefore appears that cytokines such as INFɣ not only influence HLA expression directly through the EnhA, ISRE and CRE elements but also act by upregulating the control proteins NLRC5 and CIITA.

HLA class II expression is constitutive in some cell types (APC) and induced via activation and upregulation in others. There are at least five different promoters for CIITA and together these induce alternative splicing of CIITA exon 1 sequences (Devaiah & Singer, 2013; Muhlethaler-Mottet et al., 1997). These splice variants give rise to at least three isoforms which are cell type specific (Devaiah & Singer, 2013; Muczynski, Ekle, Coder, & Anderson, 2003). CIITA isoform pIII binds to the enhanceosome in B cells and activated T cells, whilst isoform pI controls HLA class II expression in dendritic cells. IFNγ initiates splicing to the pIV isoform in non-bone marrow-derived cells such as endothelial cells (Muczynski et al., 2003; Neerincx et al., 2013; Van Den Elsen et al., 2004).

To prevent an uncontrolled inflammatory response following HLA upregulation, CIITA itself is rapidly degraded by the ubiquitin–proteasome pathway giving it a half-life of around 30 min (Devaiah & Singer, 2013). In attempts to control the immune response, some viruses such as EBV have been shown to encode CIITA repressors that block HLA class II upregulation (Dendrou et al., 2018).

Interestingly, extra villous trophoblasts (EVT) do not express NLRC5 or CIITA but do present the classical class I HLA-C proteins in addition to the nonclassical HLA-E and HLA-G. HLA-C possesses a distinct RFX sequence at the NLRC5 binding region which creates a unique recognition site for the ELF3 promotor (see Figure 1). ELF3 is known to be upregulated in trophoblasts, and this alternate RFX binding sequence may be responsible for both the reduced constitutive expression of HLA-C by altered NLRC5 binding in normal cells, and the continued expression of the protein in trophoblasts where NLRC5 is absent (Ferreira, Meissner, Tilburgs, & Strominger, 2017; Johnson, Wright, Li, & Anderson, 2018).

2.2 Promoter polymorphisms

Sequence polymorphisms in the transcription promoter region of HLA genes affect the assembly of the SXY module and overall HLA transcription. Experiments showed that the absence of two bases between the S and X boxes in the 5′UTR for HLA-DQB1*03:01, at position −178, results in significantly higher levels of a reporter molecule than in HLA-DQB1*03:02 where the bases were present (Andersen et al., 1991; Handunnetthi et al., 2010). Ferstl et al. (2004) describe polymorphisms resulting in ten different variations of the SXY promoter region of HLA-DQB1 genes. These variants result in both allele-specific and tissue-specific changes in HLA-DQ expression, with a two-and-a-half-fold increase of HLA-DQB1*03:01 gene products over DQB1*06:02, for example.

Additional polymorphisms in the enhanceosome, particularly in the Y box, affect cytokine-induced upregulation of HLA-DR genes (Sindwani & Singal, 2001), and Belicha-Villanueva and Mcevoy (2009) describe variations in MHC transcription as a result of polymorphisms in the EnhA site immediately upstream of the SXY module in the 5′UTR (Figure 1).

Further upstream from the S box, Fernandez (2003) has used the real-time reverse transcriptase–polymerase chain reaction (real-time RT-PCR) to map increased expression of HLA-DQA1 mRNA to a hypervariable region between 240-bp and 200-bp upstream of exon 1.

Polymorphisms in the transcription factor binding sites for HLA-C are associated with alleles demonstrating low surface expression. Mutations found in the EnhA region affect HLA-C*07 expression, TATA affect HLA-C*03, and insertions and deletions between these sites alter HLA-C*17 protein expression (Ramsuran et al., 2017).

Unlike HLA-A and HLA-C, HLA-B mRNA levels show less variation in expression between alleles, suggesting 5’UTR polymorphism does not significantly affect mRNA production for this locus (Ramsuran et al., 2017). On average, mRNA levels for resting cells varied by around 1.2-fold for different HLA-B alleles compared with 3.8-fold for HLA-A and 2.2-fold for HLA-C.

Additional promoter polymorphisms include a vitamin D response element (VDRE) in the 5′UTR of HLA-DR which is disrupted in many alleles. The VDRE is fully functional in HLA-DRB1*15 alleles, and it is interesting to note that HLA-DRB1*15 and geographical latitude (affecting vitamin D synthesis) are both strongly associated with the disease multiple sclerosis (Handunnetthi et al., 2010).

2.3 Long-range promoters

Bettens et al. (2014) detected an association between HLA-C expression and the extended HLA haplotype, suggesting additional control mechanisms beyond the immediate promoter region. The Bettens group described one particular HLA-B*49:01 – C*07:01 haplotype which consistently demonstrated a markedly elevated HLA-C mRNA level. Similarly for HLA class II, Handunnetthi et al.'s (2010) work describes repeats and inverted repeats of the SXY module several kilobases upstream of HLA class II genes. It is thought that these repeat sequences are involved in histone acetylation, initiating and stabilizing DNA unwinding to allow transcription promoter binding. Histone acetylation prevents the reformation of the DNA helix structure and enables the enzyme DNA polymerase II to initiate gene transcription (see Figure 1) (Venkatesh & Workman, 2015). In addition, Li et al. (2018) describe a region 1,300-bp upstream of the HLA-C start codon which may be involved in enhancing or repressing HLA promoters.

2.4 Splice variants

The HLA protein is transcribed after splicing together up to eight exons from the gene sequence. Alternative splicing of the HLA gene has been shown to result in longer or shorter mRNA sequences and variation in the final protein structure (see Figure 2). Dai et al. (2014) have demonstrated splice variants leading to the deletion of exon 3 from the HLA-A24 mRNA. This deletion resulted in the formation of homodimers and heterodimers with wild-type HLA-A24. The lack of exon 3 would prevent antigen presentation and also disrupt much of the HLA-specific antibody binding, which is commonly directed at protein within the binding cleft, coded for by exons 2 and 3.

Splice variants have been implicated in HLA null alleles by skipping exon two with no resultant protein (Belicha-villanueva & Mcevoy, 2009) and by skipping exon five as soluble isoforms (Dai et al., 2014). Dunn et al. (2011) describe a silent mutation (G-A) at position 705 in exon four of HLA-A*01 that creates an alternative splice site and a resultant exon which is 87bp shorter than the wild-type. Whilst the splice site does create an in-frame reading sequence, the resultant protein is likely misfolded and therefore not expressed. Interestingly, low levels of full-length protein were still detectable on some cells within this individual, showing, as with Dai's work, that the choice of splice variant is not absolute.

Splice variants affect HLA class I (Belicha-villanueva & Mcevoy, 2009; Tijssen, Sistermans, & Joosten, 2000) and class II gene transcription (Handunnetthi et al., 2010; Hoarau, Cesari, Caillens, Cadet, & Pabion, 2004) and a number of these result in an elongated cytoplasmic tail (Belicha-villanueva & Mcevoy, 2009; Hoarau et al., 2004). Whilst the effect of the longer tail is unknown, in the nonclassical class I molecule HLA-G splice variants create either membrane bound or secreted isoforms that are essential for effective trophoblast implantation (Carosella, Rouas-Freiss, Roux, Moreau, & LeMaoult, 2015).

In most of these cases, it is likely that the resultant protein, if any, will be misfolded and will not present peptide to T cells, reducing the ability for these cells to contribute to immune surveillance. In the transplant setting, these variants will almost invariably result in a reduction in HLA mismatched antigen presentation to the recipient immune system. The effect of soluble HLA, for example through exon five splicing, has yet to be elucidated in transplantation.

2.5 Alternative start codons

Li et al. (2018) have found alternative mRNA transcripts due to polymorphisms in exon one of the HLA-C allele. These result in the use of alternative start codons and the generation of an exon of varying length, sometimes even including part of the intron one sequence. The selection of transcript depends on both the HLA-C allele and the tissue in which it is expressed and ultimately influences surface expression of the protein. This is most notable in NK cell maturation, with higher HLA-C levels on educated NK cells. In total, seven start codons were identified including one within the start region of exon two, which could result in the production of intracellular HLA-C (Johnson et al., 2018; Li et al., 2018).

2.6 Alternate polyadenylation sites

Up to 70% of human genes use alternative polyadenylation sites (PAS) at the 3′ end of the gene to create a longer or shorter product, with immune genes in particular switching to the shorter form on infection and upregulation (Kulkarni et al., 2017). Variation in HLA-A expression has been shown to be dependent on two such PAS, the nearer being conserved but the more distant is disrupted by sequence variation in some alleles.

Mutations that disrupt this distal PAS result in a shorter mRNA transcript as only the nearer polyadenylation site is used, but if both sites are active the cell can utilize either. Whilst similar amounts of mRNA are produced regardless of the PAS selected, there appears to be a translation inhibitory region between the two sites resulting in lower amounts of HLA protein from the longer sequence.

Different HLA-A genes are under control of the proximal PAS only, the distal PAS or both. In resting cells, HLA-A*03 uses the distal PAS resulting in lower expression, but switches to the proximal PAS on upregulation by some pathogens, enabling greater surface expression of the protein and increased presentation of pathogenic peptides. HLA-A*01 and A*11 are only able to utilize the proximal PAS and do not switch to the resultant low expression PAS when not activated (Kulkarni et al., 2017). This alternate PAS usage does not occur for HLA-B and HLA-C which only possess the more conserved distal polyadenylation site, suggesting that HLA-A upregulation is more greatly influenced by infection or other stimulation.

A combination of alternate splicing and the use of multiple polyadenylation sites has also been noted for HLA class II transcription where two possible splice sites and three polyadenylation sites were found in the 3′UTR of HLA-DQA1 (Hoarau et al., 2004). The authors detected up to four different mRNA transcripts for each of the HLA-DQA1*01 alleles and six transcripts, using all combinations of splice and polyadenylation sites, for DQA1*02, 03, 04 and 05.

2.7 DNA methylation

DNA methylation, mediated by the DNA methyltransferase enzymes, allows permanent or temporary silencing of genes, and in humans, this commonly occurs across cytosine–guanine dinucleotide (CpG) methylation sites. Point mutations or single nucleotide polymorphisms (SNPs) that create or remove a CpG site can therefore lead to variations in DNA methylation between subjects (Bird, 2002; Morimoto et al., 2004; Nie, 2001; Rakyan et al., 2004). Indeed, HLA expression has been shown to be altered by DNA methylation of CpG sites in the promoter region, within exons and introns, and also across the CIITA promoter itself.

For class I HLA genes, methylation can be locus and allele specific. Despite similar numbers of potential methylation sites, only HLA-A expression is affected by methylation. Ramsuran et al. (2015) demonstrated an inverse correlation between the amount of DNA methylation in the HLA-A gene and mRNA levels. Methylation was concentrated around the promoter region despite numerous possible methylation sites within exons two and three. The team showed that for the high expression allele HLA-A*24, only one CpG site was methylated, whereas the low expresser HLA-A*03 was methylated at more than one site. Ramsuran's group demonstrated that up to 6% of sequencing reads were methylated at some CpG sites for HLA-A*03, whereas <0.5% of reads were methylated for the one CpG site for HLA-A*24.

Using 5′-Aza-CdR treatment, which reverses DNA methylation, Ramsuran's group demonstrated that mRNA expression increased more when 5′-Aza-CdR was used in genes which tended to have more initial methylation (e.g., HLA-A*03). Interestingly, the team also found that despite possessing similar numbers of possible methylation sites, HLA-B and HLA-C were not methylated, and the addition of 5′-Aza-CdR had little effect on mRNA levels for these loci. The authors postulate that upstream regulatory elements and long-range determinants of chromosome structure may affect locus-specific methylation (Ramsuran et al., 2015).

Unlike HLA class I, where only the promoter region is normally affected by methylation, differentially methylated regions (DMR) within exon two of the HLA-DRB1 gene may be an additional risk factor in the disease multiple sclerosis (MS; Kular et al., 2018). Exon two of the DRB1*15:01 allele is hypomethylated in MS sufferers compared to DRB1*15:01-positive individuals without MS. Kular's team suggest that DRB1 protein expression is downregulated by methylation of up to 19 CpG sites in exon two, affecting expression between nonactivated and activated cells, but the hypomethylation of DRB1*15:01 in affected individuals permits increased expression of this antigen even when cells are not activated.

Following a similar pathway shown in HLA-A, reduced methylation within the promoter region is the prime switch for HLA-G expression in human embryonic stem cells (with further post-transcriptional modification affected by microRNA binding). Methylation of the HLA-G promoter will permit permanent silencing in other cells where its expression does not normally occur (Verloes et al., 2017).

Mutations and cancers can cause unusual DNA methylation, and this has been shown to be one reason for the lack of HLA-DR expression in both T-cell and myeloid leukaemia patients (Morimoto et al., 2004). Methylation of the CIITA-PIV splice variant prevents IFNγ-induced expression of HLA-DR and therefore abrogates a normal antitumour response. Neither addition of the methyltransferase inhibitor 5’-Aza-CdR, nor IFNγ alone enabled a normal HLA class II antitumour response, but these cells required both components to restore HLA-DR upregulation.

DNA methylation has therefore been shown to have an impact in the promoter region for HLA-A and HLA-G, intraexon for HLA class II and across CIITA in some cancers.

2.8 Post-transcriptional regulation

MicroRNAs (miRNA) are involved in the post-transcriptional regulation of many genes, including accelerated degradation of HLA-specific mRNA (Handunnetthi et al., 2010; Kulkarni et al., 2011). Increased levels of HLA-C are associated with better HIV control and reduced serum viral load, and polymorphisms in the 3’UTR result in differential expression of HLA-C on the cell surface (Kulkarni et al., 2011). This agrees with McCutcheon et al. (1995) who had previously suggested that low expression of HLA-C is correlated with increased mRNA degradation associated with DNA polymorphisms in this same region.

HLA polymorphisms in the 3′UTR affect a binding site for the specific miRNA known as has-miR-148. mRNA transcripts containing this has-miR-148 binding site undergo increased post-transcriptional degradation resulting in significantly lower surface expression of HLA-C protein (Kaur et al., 2017; Kulkarni et al., 2011). Whilst HLA-A also possesses the has-miR-148 binding site, it is nonpolymorphic in this locus. However, polymorphisms in the has-miR-148 gene itself lead to higher or lower expression of this miRNA which may then influence HLA-A and HLA-C regulation and alter overall HLA levels in some individuals (Petersdorf & O'hUigin, 2018).

In the 3′UTR for the class II gene HLA-DP, a single A/G polymorphism at position 496 results in an almost twofold higher expression of HLA-DP in the G/G homozygous group, than all others (Thomas et al., 2012). This higher expression correlated with greater persistence of HBV infection, whereas lower expression resulted in a faster recovery. The polymorphism is also linked to an increased risk of GVHD in HLA-DPB1 mismatched haematopoietic stem cell transplants where the recipient carries the higher expressed allele (Klasberg et al., 2019; Petersdorf et al., 2015).

Using in silico predictions, Shieh et al. (2018) have shown the low expressing variant (A at position 496) to be associated with potential binding sites for a higher number of miRNAs than the 496-G polymorphism. Interestingly, the 496 polymorphism can be used to separate HLA-DP into two highly divergent clades. These have different miRNA binding sites within the first and second introns and have been shown to contain 174 fixed polymorphisms in the 3′UTR that define the two clades, although other than 496 A/G, how these affect expression or function is not yet known (Klasberg et al., 2019; Shieh et al., 2018).

Similarly, a 14-bp indel in the 3′UTR of the nonclassical class I HLA-G gene has been shown to affect stability of the HLA-G mRNA, with the insertion resulting in a lower amount of mRNA. HLA-G possesses variations in the 3′UTR which affects the binding site for miR-148 and additionally miR-152 (Carosella et al., 2015; Verloes et al., 2017).

2.9 Physical stability and protein expression

The conformational stability of HLA molecules is in part determined by the strength of binding of peptide as it is loaded in the endoplasmic reticulum. “Empty” HLA is rapidly internalized, and HLA molecules which bind peptide strongly are more stable and are thought to remain longer on the cell surface, leading to accumulation of this protein as new molecules are synthesized (see Figure 3). During infection, stable HLA could lead to greater antigen presentation to T cells, but may also contribute to autoimmunity such as coeliac disease and narcolepsy (Dendrou et al., 2018; Rizvi et al., 2014).

Rizvi et al. (2014) show how a SNP in the peptide binding groove of the HLA-B*44 gene results in a molecule where peptide loading is either Tapasin dependent (HLA-B*44:02) or independent (B*44:05). Tapasin-independent loading tends to result in faster shuttling of loaded HLA to the cell surface, whereas HLA molecules requiring Tapasin-mediated peptide loading can result in delayed, but potentially far more stable surface expression.

The passive binding of peptide to the HLA-B*44:05 molecule results in a lower affinity peptide: HLA association and therefore a greater likelihood of peptide dissociation on the cell surface. Loss of peptide initiates HLA internalization and unless reloaded with new peptide leads to degradation (ubiquitination) within the cell (see Figure 3). Tapasin mediates binding of peptides with a much stronger affinity for the HLA molecule and significantly reduced spontaneous peptide dissociation. The prolonged surface expression of stable HLA-B*44:02 molecules with continued shuttling of new molecules to the cell surface enables higher overall cell surface expression of this B*44:02 molecule compared to B*44:05.

Rizvi et al. (2014) noted a general trend of HLA-B molecules carrying the Bw6 motif to have less dependence on Tapasin loading than those expressing Bw4, although whether HLA-Bw6 proteins are generally less populous on the surface of cells requires more investigation.

Similarly, HLA-B molecules which are less promiscuous in their peptide repertoire (e.g., HLA-B*57, B*27:05) usually bind those peptides more strongly and are more dependent on Tapasin-mediated loading. They also appear to be more stable and are therefore expressed at higher densities than those with a less stringent peptide requirement (e.g., B*35:01; Chappell et al., 2015).

Conversely, Kaur et al. (2017) suggest that differences in the binding cleft for HLA-C affect its surface stability through greater promiscuity for peptide. This permits a greater availability of possible peptides, leading to faster loading and shuttling to the cell surface, resulting in higher surface expression. They show that HLA-C*05 alleles, with a wide and easily accessible binding cleft, can hold approximately three times as many distinct peptides and are found in greater abundance on the surface, compared with HLA-C*07 which has a narrow cleft. These differences between HLA-B and HLA-C stability may be locus specific.

2.10 Active internalization of HLA

Total cell surface HLA expression is governed by the interplay between new protein synthesis and the active rate of HLA endocytosis (Muller, Schumacher, & Kyewski, 1993; Tse & Pernis, 1984). When HLA is eventually internalized, there may be specific control over whether ubiquitination or recycling back to the surface occurs (Cho & Roche, 2013; Muller et al., 1993; Reid & Watts, 1990).

In a mouse model, MHC class I and class II internalization is a selective process which varies between cell types (Tse & Pernis, 1984), and the rate of recycling within antigen-presenting cells is organ specific (Muller et al., 1993). Similarly in a human lymphoblastoid cell line, the majority of internalized HLA class I and class II is rapidly recycled back to the cell surface, although this may not reflect normal cell function (Reid & Watts, 1990).

Accelerated endocytosis of intact HLA–peptide appears to be an active and specific process which for HLA class I is determined by the amino acid structure of the cytoplasmic tail (Vega & Strominger, 1989). The amino acid motif YSQA (residues 320–323 of exon seven) is similar to tyrosine-based endocytosis mechanisms described in other cell expression systems (Piguet, Schwartz, Le Gall, & Trono, 1999). Once internalized, the protein is either degraded or recycled back to the cell surface, for example in dendritic cells where recycling enables a rapid response to stimuli and also the opportunity to load alternative peptides in sites of local inflammation (Villadangos, Schnorrer, & Wilson, 2005).

Piguet et al. (1999) and Lin et al. (2007) have shown an increase in HLA endocytosis and degradation in lentiviral and cytomegalovirus (CMV) infections, reducing the half-life of HLA class I on the surface from 60 min to as little as 10. Similarly, in antigen-presenting cells, CMV infection increases HLA class II ubiquitination where it is degraded in lysosomes (Lin et al., 2007).

The HLA class I viral pathway may hijack normal cell machinery as virus-driven endocytosis is still dependent on specific amino acids within the cytoplasmic tail, particularly the tyrosine at position 320 described above. This motif is present in all HLA-A and HLA-B molecules, but HLA-C possesses a cysteine at position 320 instead of tyrosine. HLA-C is unaffected by viral-induced endocytosis, and this may explain the crucial role that HLA-C plays in protection from HIV and other viral infections (Apps et al., 2013; Thomas et al., 2009).

In immature dendritic cells, newly synthesized HLA class II proteins are held within endosomes. Dendritic cell maturation allows rapid movement of this HLA class II to the cell surface, and there is evidence that the class II cytoplasmic tail is modified to reduce or prevent ubiquitination, a unique process to control surface expression (Shin et al., 2006; Villadangos et al., 2005). Cho and Roche (2013) suggest that whilst this may not reduce HLA class II internalization, it allows the protein to recycle back to the cell surface rather than be directed to degradation in the lysosomes. The different cytoplasmic tail amino acid structure appears to direct the class II molecules deeper into the lysosomes for degradation, or remains near the cell surface in multivesicular bodies for peptide reloading and re-expression, resulting in a highly variable half-life of between 15 and 50 hr (Thibodeau, Moulefera, & Balthazard, 2018).

So whilst there are a host of mechanisms governing the production of mRNA and subsequent processing into intact peptide-loaded HLA molecules, the amount of surface HLA is a balance between de novo production, HLA locus-specific internalization and whether the HLA molecule is recycled back to the cell surface or directed to lysosomes for destruction. Cell activation appears to reduce the availability of proteins involved in HLA ubiquitination, allowing more HLA to recycle back to the surface, adding to newly produced HLA protein and increasing overall surface density.

In endothelial cells, HLA class II expression as a result of cell activation is short lived in the absence of continual cytokine stimulation (Cross et al., 2016), and these cytokines may be required to promote recycling back to the cell surface rather than degradation (Cho & Roche, 2013). HLA cycling is therefore locus specific and dependent on the cytoplasmic tail. Polymorphisms in this region will affect the rate of endocytosis and whether internalization leads to recycling or ubiquitination.

In a review of the relationship between HLA and disease, Dendrou et al. (2018) suggest variations in HLA expression caused by many of the mechanisms described above lead to either autoimmunity through insufficient HLA expression and ineffective deletion of autoreactive T cells in the thymus (in type I diabetes for example), or resistance to disease by variation in the presentation of foreign peptides or a greater density of HLA proteins.