Regulation of expression quantitative trait loci by SVA retrotransposons within the major histocompatibility complex

Abstract

Genomic and transcriptomic studies of expression quantitative trait loci (eQTL) revealed that SINE-VNTR-Alu (SVA) retrotransposon insertion polymorphisms (RIPs) within human genomes markedly affect the co-expression of many coding and noncoding genes by coordinated regulatory processes. This study examined the polymorphic SVA modulation of gene co-expression within the major histocompatibility complex (MHC) genomic region where more than 160 coding genes are involved in innate and adaptive immunity. We characterized the modulation of SVA RIPs utilizing the genomic and transcriptomic sequencing data obtained from whole blood of 1266 individuals in the Parkinson’s Progression Markers Initiative (PPMI) cohort that included an analysis of human leukocyte antigen (HLA)-A regulation in a subpopulation of the cohort. The regulatory properties of eight SVAs located within the class I and class II MHC regions were associated with differential co-expression of 71 different genes within and 75 genes outside the MHC region. Some of the same genes were affected by two or more different SVA. Five SVA are annotated in the human genomic reference sequence GRCh38.p14/hg38, whereas the other three were novel insertions within individuals. We also examined and found distinct structural effects (long and short variants and the CT internal variants) for one of the SVA (R_SVA_24) insertions on the differential expression of the HLA-A gene within a subpopulation (550 individuals) of the PPMI cohort. This is the first time that many HLA and non-HLA genes (multilocus expression units) and splicing mechanisms have been shown to be regulated by eight structurally polymorphic SVA within the MHC genomic region by applying precise statistical analysis of RNA data derived from the blood samples of a human cohort population. This study shows that SVA within the MHC region are important regulators or rheostats of gene co-expression that might have potential roles in diversity, health, and disease.

Impact Statement

The genomic region of the major histocompatibility complex (MHC) harbors more than 160 genes involved in innate and adaptive immunity, inflammatory responses and complement functions. There are more than 18 SINE-VNTR-Alu (SVA) retrotransposons within the MHC region, but little or nothing is known about their function or influence on MHC gene expression. Our study shows that eight of these SVA have regulatory effects on gene expression. The regulatory properties of SVAs located within the class I and class II MHC regions were associated with differential co-expression of 71 different genes within and 75 genes outside the MHC region as multilocus expression units in what appears to be a highly coordinated network of transcriptional activity. Thus, future genetic engineering of these SVA elements could be a way to manage various genes in the MHC and the immune response in a coordinated and systematic way not previously envisioned.

Introduction

Human leukocyte antigen (HLA) class I and class II molecules (antigens) are polymorphic cell-membrane-bound glycoproteins that regulate the immune response by presenting peptides of fragmented proteins to circulating cytotoxic and helper T lymphocytes, respectively. The HLA molecules are investigated continuously because of their importance in the regulation of the innate and adaptive immune system, autoimmunity, infectious diseases, and transplantation outcomes.^1–4 The genomic region that encodes these HLA molecules is known as the major histocompatibility complex (MHC) on the short arm of chromosome 6 at 6p21.3 and it encompasses approximately 160 coding genes within ~ 3–4 MB including three distinct structural regions: class I with the classical and nonclassical HLA class I genes and ∼39 non-HLA genes; class II with the classical and nonclassical HLA class II genes and some proteasome-processing and peptide antigen transportation non-HLA genes; and class III that harbors more than 60 non-HLA genes including those involved in stress response, complement cascade, immune regulation, inflammation, leukocyte maturation, and regulation of T cell development and differentiation.^3–5 Of the many HLA-like genes, 18 HLA class I genes (six protein-coding genes and 12 pseudogenes) and seven MHC class I chain-related (MIC) genes (two protein-coding genes and five pseudogenes) are in the HLA class I region, and 18 HLA class II genes (13 protein-coding genes and five pseudogenes) are in the HLA class II region. The classical HLA class I genes, HLA-A, -B, and -C, and the classical HLA class II genes, HLA-DR, -DQ and -DP, are characterized by their extraordinary large number of polymorphisms, whereas the non-classical HLA class I genes, HLA-E, -F, and -G, are differentiated by their tissue-specific expression and limited polymorphism.^6,7

The MHC genomic region contains numerous coding and noncoding gene sequence duplications, insertions, and deletions and considerable sequence diversity or polymorphisms^8,9 that have accumulated by ancestral descent and recombined into thousands of distinct multilocus haplotypes at variable worldwide population frequencies.^5,10,11 The two most frequent (>1%) European HLA ancestral haplotypes 8.1AH (HLA-A*01:C*07:B*08:DRB1*03:DQB1*02) and 7.1AH (HLA-A*03:C*07:B*07:DRB1*15:DQB1*06)^12,13 were estimated to have diverged from their common ancestors at least 23,500 years ago. ¹⁴ Polymorphic MHC multilocus haplotypes have been associated with Parkinson’s disease (PD)^15–17 and autoimmune diseases such as ankylosing spondylitis, type 1 diabetes mellitus, multiple sclerosis, and rheumatoid arthritis. ¹⁸ Polymorphic frozen blocks ¹ or single nucleotide polymorphism-linkage disequilibrium (SNP-LD) blocks that were identified by LD or long-range haplotype and extended haplotype homozygosity tests ¹⁹ of the MHC genomic regions include a variety of genotyped microsatellites, SNPs, and indels. ² The absence and presence of a variety of structurally dimorphic retroelements within the MHC, such as Alu, SVA, long terminal repeats (LTR), and human endogenous viruses (HERVs), have been used as haplotypic markers and for phylogenetic analyses of population ancestral relationships and diversity.^20–24

The investigation of MHC multilocus haplotypes and disease associations is complicated by the regulatory network of cis-acting multilocus expression units known as haplotype-specific expression quantitative trait loci (eQTL).^2,25,26 These eQTL are controlled by an array of retroelements and DNA transposons, ²⁷ both as binding sites for transcription factors ²⁸ and as sources of regulatory noncoding RNAs.^29,30 In this regard, eQTL studies associate genomic and transcriptomic data together from the same individuals to identify loci that affect mRNA expression by linking SNPs to changes in gene expression, which, in turn can be applied as a useful procedure for uncovering important patterns of gene regulation and genome-wide association study variants.^31,32 Studies using homozygous cell-lines and/or biological samples have demonstrated that the expression of various clusters of genes inside or outside the MHC genomic region can be affected differentially by different haplotypes.^2,33 Lam et al. ²⁶ used haplotypic RNA and DNA-sequencing data to show that haplotype sequence variations represented by eQTL SNP alleles can function as cis-acting regulatory variants for multiple MHC genes, and that these eQTL SNPs were localized especially within four segmental regions containing HLA-A (alpha block), HLA-C (beta block), C4A (gamma block), and HLA-DRB1 (delta block). In a study of eQTL within the genomic sequences of lymphoblastoid cell lines, Spirito et al. ³⁴ found that the chromosomal location 6p21.32, which includes the MHC class I and II regions, was a particularly enriched genomic region where structurally polymorphic transposable elements (TEs) influenced gene expression. Taken together, these pioneering eQTL studies incorporating either HLA haplotypes or SNPs are a powerful new approach to identifying causal genetic mechanisms underlying disease associations both inside and outside the MHC region.

Apart from the protein coding genes, pseudogenes, noncoding transcribed RNA, and small nucleolar transcribed RNAs loci, there are at least 8604 repeat elements, including those known as TEs and/or retroelements, and 723 simple repeats (microsatellites) within the MHC of the human genomic reference sequence GRCh38.p14/hg38. ⁵ The MHC of this genomic reference sequence is represented by the 7.1AH derived mostly from the DNA sequencing data of the immortalized PGF homozygous lymphocytic cell-line. ⁸ Retroelements, which are a class I subcategory of TEs, consist of various subfamilies including HERVs, LTRs, long interspersed elements (LINE 1, LINE 2), short interspersed elements (SINE 1, SINE 2), and SINE-R-VNTR-Alu (SVA). ³⁵ Furthermore, retrotransposon insertion polymorphisms (RIPs) ³⁶ are either absent or present in the genome as insertions or deletions (indels) depending on their state of origin or ancestral state. ³⁷ RIPs or indels have been associated strongly with the regulation of gene expression in health and disease.^27,38 In addition, eQTL of non-LTR retroelements such as SINEs, LINEs, and SVAs and their affect on the regulation of gene expression were identified and described for a Parkinson’s Progression Markers Initiative (PPMI) cohort using whole genome sequence and transcriptome data obtained from the blood of hundreds of individuals. ²⁷

At least 18 SVA polymorphic insertions were described previously within the MHC class I, II, and III regions; some were haplotypic or haplospecific for particular HLA gene alleles that varied in frequency between European, Japanese, and African American populations. ³⁹ For example, the SVA-HF, SVA-HA, and SVA-HC were inserted at a relatively low frequency (<0.2) in European (Caucasian) populations and strongly associated with 7.1AH, but not with the 8.1 haplotype.^20,21

In this study, we identified the eQTLs associated with various SVA RIPs within the MHC genomic region of 1266 individuals within the PPMI cohort. We re-mined and added new transcriptome data to the previous analyses of SVA^27,40 and focused on the presence and roles of eight distinct SVA RIPs within the MHC genomic region and their eQTL effects on HLA and non-HLA genes within (cis) and outside (trans) the MHC. We also examined and found distinct structural effects (long and short variants and the CT internal variants) for one of the SVA (R_SVA_24) insertions on the differential expression of the HLA-A gene within a subpopulation (550 individuals) of the PPMI cohort.

Materials and methods

Datasets

In this study of SVA regulation of HLA and non-HLA genes within the MHC genomic region, we utilized the data of 1266 individuals within the PPMI cohort without assessing any differences between the cases and controls. PPMI cohort data ⁴¹ were downloaded from http://www.ppmi-info.org/data (accessed on 19 January 2021) as previously described.^27,40 The dataset contains whole transcriptome data from blood together with genetic and clinical data of 1266 individuals whose race was reported as white. All subjects, including 678 PD cases and 588 controls, were combined as a single study population for the current SVA and eQTL analyses (Figure 1(A)). The structural variant caller Delly2 (https://github.com/dellytools/delly) ⁴² was used with default settings to genotype structural variants from the PPMI cohort. The structural variants located within the coordinates of the human-specific SVAs and available on the UCSC genome browser (https://genome.ucsc.edu/) were extracted with ±100 bp. Structural variants classified as deletions in these regions were inspected manually to determine if the break points were consistent with the reported coordinates of the reference human-specific SVAs. Gene association analysis of SVA RIPs was performed using logistic regression in PLINK (v1.07) ⁴³ with data from both the control and PD samples, and P-values adjusted for multiple testing (Bonferroni correction). In addition to reference RIPs, non-reference RIPs that presented in an individual’s genome, but not the genome reference, were included in this study as non-reference SVA (NR_SVA). The read length of the whole-genome-sequencing data was 150 bp, and, on average, there were 837 million reads per genome, with the coverage more than 30×. RNA-seq reads were also 150 bp in length, and average number of reads per sample was 31 million.

Figure 1.

Methodology Workflow. (A) Whole blood RNA-seq data obtained from 1266 PPMI individuals were quasi-aligned and counted with Salmon. The counts were normalized and analyzed using DESeq2. The effect of SVA RIP genotypes on the expression of genes inside and outside the MHC genomic region were evaluated using the Matrix eQTL software with inbuilt statistical tests including the false discovery rate (FDR). (B) Whole blood RNA-seq data obtained from 550 individuals of the PPMI cohort were quasi-aligned and counted with Salmon, normalized and analyzed using DESeq2. Differential expression of HLA-A transcripts was analyzed via likelihood ratio tests and multiple comparisons (Kruskal−Wallis), stratified by R_SVA_24 length and CT allele genotype data.

Fastq files of whole-blood RNAseq data were downloaded from the PPMI database and mapped to hg19 (GRCh37) by STAR using GENCODE v19. The Salmon software ⁴⁴ was used to call transcript counts from the reads in conjunction with the index and decoy based on the reference genome hg 38. Salmon-generated quant files were imported into R using tximport function from the tximport package of R. Raw counts were extracted with the DESeqDataSetFromTximport function and normalized using the median-of-ratios method implemented in the DESeq2 package. The counts were divided by sample-specific size factors determined by the median ratio of gene counts relative to the geometric mean per gene. PD cases and controls were combined, and transcript expression signals were tabulated after importing Salmon files in the R format to prepare them for the eQTL analysis. Altogether, 236,186 transcripts previously described were used for the initial analysis in combination with all the identified TEs, including the eight SVA within the MHC genomic region. ²⁷

eQTL analysis

Matrix eQTL was used to calculate the genetic loci of SVA regulating the expression transcript variants. ⁴⁵ The additive linear model was used with covariates, age, and sex, with the threshold of the FDR set at 0.05. During eQTL analysis, local (cis) and distant (trans) quantitative loci were called, but ultimately this depended on whether the loci were within the boundaries of the MHC genomic region ranging between the telomeric HLA-F and the centromeric COL11A2 loci. Matrix eQTL reported effect-size estimates as beta values or slope coefficients. The correction for multiple testing of eQTL was performed using FDR, and only the results that remained significant after FDR correction are reported here. For pairwise comparisons between the genotype, a Wilcoxon test was used, and P-values were challenged with the Bonferroni multiple comparison test.

SVA_24 structural polymorphisms and differential expression of HLA-A

A sample group of 550 individuals from the PPMI cohort was used for this study to determine which SVA structural characteristics affect the expression of HLA-A variants. Whole blood RNA-seq data obtained from 371 PD patients and 179 healthy controls were normalized and filtered for batch effects/quality control using open-source statistical programming platform, R, and the DESeq2 analysis package. ²⁷ Count data were compared between long and short SVAs, as well as CT alleles A and B. Differential expression of HLA-A transcripts was analyzed via likelihood ratio tests and nonparametric multiple comparisons (Kruskal–Wallis), stratified by SVA length (long L, short S or absent A) and CT allele genotype data as outlined in Figure 1(B). Kruskal–Wallis (one-way analysis of variance) tests were carried out using the R package “ggpubr.”

Results

Features of eight SVA RIPs within the MHC genomic region

The genomic locations and main characteristics of 10 SVA RIPs within or flanking the MHC genomic region that were identified within the MHC genomic region of 1266 individuals from the PPMI cohort database are listed in Table 1. These include SVA IDs, aliases, start and end positions within the hg38 genomic reference sequence, the subfamilies, sizes, and the flanking genes. The SVA insertions labeled as R_SVA are located within the hg38 genomic reference sequence at UCSC browser (https://genome.ucsc.edu), whereas the NR_SVA are present in PPMI individuals but are absent from the hg38 genomic reference sequence. Five of the R_SVA locations listed in Table 1 as TE aliases, SVA-HA, SVA-HC, SVA-MIC, SVA-DRB1, and SVA-DPA1 were identified in previous studies.^20,21,39

Table 1.

Ten SVA RIPs within or flanking the MHC genomic region (see Figure 2).

SVA ID	SVA	SVA start and end position	Subfamily	TE width	DNA	MHC block	TE insert	Genes flanking TE indels
	Alias	GRCh38/hg38, chr6		bp	Strand	Region	Site
NR_SVA_377		29731783				Alpha cI	L2b	HLA-F-AS1 intron 1
R_SVA_24	SVA-HA	29932088–29933753	SVA_A	1666	pos	Alpha cI	ERV3-16A3	HCG4B and HLA-A
R_SVA_25	SVA-HC	31243861–31245322	SVA_F	1462	pos	Beta cI	MLT2 C2/ERVL	Psors1 C3/HCG27 and HLA-C
R_SVA_26	SVA-MIC	31453746–31456553	SVA_D	2808	neg	Beta cI	L1M4	MICA and HCP5
NR_SVA_380		32546835				Delta cII	L1M5	DRB5 and DRB6
R_SVA_27	SVA-DRB1	32594194–32596780	SVA_F	2587	neg	Delta cII		DRB1 and DQA1
R_SVA_85	SVA-DPA1	33058946–33060797	SVA-D	1852	neg	Epsilon cII		DOA and DPA1
NR_SVA_381		33062533				Epsilon cII	L1MEc	DOA and DPA1
R_SVA_28		33525379–33526168	SVA_F	790	pos	Outside	MER21B	ZBTB9 and BAK1
NR_SVA_382		33535745				Outside	L1ME5	ZBTB9 and BAK1

SVA: SINE-VNTR-Alu; TE: transposable elements; MHC: major histocompatibility complex.

NB. Some SVA are inserted within ancient TE fragments such as L2, L1MEc, L1M4, L1M5 and ERV3_16A3 internal fragment. SVA TE aliases reported by Kulski et al.^20,21,39 R_SVA is the reference genome GRCh38/hg38, whereas NR_SVA is absent in GRCh38/hg38. The top eight SVA are within the MHC genomic region as indicated by the column showing the MHC block region with their presence in the alpha and beta class I (cI) blocks, and in the delta and epsilon class II (cII) blocks. R_SVA_28 and NR_SVA_382 are inserted “outside” the MHC genomic region on the centromeric side of the extended class II (cII) region (Figure 2).

The locations of the eight SVA retrotransposon polymorphic integrations within the MHC genomic region from the HLA-F gene in class I to the HLA-DPA3 gene in class II are shown in Figure 2. Four of the SVA RIPs are within the MHC class I region and four are within the class II region relative to 297 known coding and noncoding gene loci within these two regions and the MHC class III region. We also included two SVA insertions, R_SVA_28 and NR_SVA_382, located between the ZBTB9 and BAK1 genes that are ~48 kb outside the centromeric end of the MHC class II region for comparison of the effects of MHC and non-MHC SVA insertions on MHC gene expression. The MHC class I genomic region is subdivided into three HLA duplicated and regulatory regions or polymorphic frozen blocks, known as the alpha block with HLA-A, the kappa block with HLA-E, and the beta block with HLA-C, HLA-B, and the two MIC genes, MICA and MICB.

Figure 2.

Gene map of MHC genomic region showing the clusters of coding and noncoding gene loci in the class I, class III, and class II regions. The alpha, kappa, and beta blocks within the class I genomic region and the delta and beta blocks of the class II region are labeled in green boxes with horizontal double arrows showing the spread of genes within each block. The extended class II region is indicated by the white box. Positions of the eight SVA integrations within the MHC genomic region are indicated by labeled boxes over black vertical lines. Two SVA insertions that are located between the ZBTB9 and BAK1 genes ~48 kb from KIFC1 at the centromeric end are shown slightly outside the extended MHC class II region. The gene map corresponds to the genomic coordinates of 29,602,228 (GABBR1) to 33,410,226 (KIFC1) on chromosome 6 in the human genome GRCh38.p13 primary assembly of the NCBI map viewer. White or colored (orange, red, and blue) vertical boxes, gray, and black vertical boxes show protein-coding genes, noncoding RNAs (ncRNAs), and pseudogenes, respectively. Red, green, and blue letters indicate HLA class I, MIC, and class II genes, respectively.

Source: Adapted from Shiina et al.^3,4

Cis eQTL effects and genotype frequencies of SVAs

The SVA genotype frequencies and the number of cis and trans genes and transcripts significantly upregulated or downregulated in the eQTL analysis of 1266 individuals are shown in Table 2. The frequency of SVA insertions within the MHC of individuals varied between 0.116 for NR_SVA_377 and 0.980 for R_SVA_85. The homozygous SVA insertion (genotype AA) frequency is substantially lower than for the heterozygous SVA insertion (genotype AP) frequency for all the SVA integrations within the MHC except for R_SVA_85, which at 99% integration is close to complete fixation with the human population, while the other MHC SVA are still in the process of balancing selection or evolution toward fixity or homozygosity by purifying selection. Both R_SVA_85 and NR_SVA_381 that are inserted between the HLA-DOA and -DPA1 genes within the epsilon block of the MHC class II region influenced fewer genes (five and six, respectively) than the other SVA insertions within either class I or class II regions, which affected between 14 and 34 cis genes. The two SVA insertions, NR_SVA_380 and R_SVA_27 near the HLA-DRB1 gene, modified the transcription of 22 and 23 cis genes, respectively. R_SVA_25, which is located 23 kb telomeric of the HLA-C gene within the beta block of the class I region, coordinately modified the expression of 38 cis genes and 72 RNA variants within and between the different MHC polymorphic blocks including genes in the alpha block (HLA-A, -H, -K, -F), kappa block (HLA-L), beta block (HLA-B, HLA-C, MICA, MICB), and delta block (HLA-DRB5, -DRB1, -DQB1, -DQB2), and non-HLA genes in the intergenic regions located between the blocks and within the class III region (Supplemental Tables 1 and 2).

Table 2.

SVA genotype frequencies and number of cis and trans genes and transcripts affected in eQTL analysis of 1266 PD cases and controls.

SVA ID	n	SVA genotype frequencies				Number of cis and trans genes affected
		AA	AP	PP	AP, PP	cis	cis	trans	trans
						Genes	RNA variants	Genes	RNA variants
NR_SVA_377	1266	0.88	0.11	0.01	0.12	14	28	30	30
R_SVA_24	1231	0.61	0.30	0.20	0.40	24	47	3	3
R_SVA_25	1258	0.83	0.16	0.01	0.17	34	72	39	39
R_SVA_26	1262	0.33	0.47	0.21	0.6	17	31	0	0
NR_SVA_380	1265	0.74	0.2	0.02	0.26	22	40	6	6
R_SVA_27	1261	0.76	0.23	0.01	0.24	23	48	4	4
R_SVA_85	1221	0.02	0.30	0.68	0.98	5	8	2	2
NR_SVA_381	1266	0.64	0.32	0.05	0.36	6	10	0	0
R_SVA_28	1264	0.00	0.03	0.96	0.99	1	1	200	214
NR_SVA_382	1266	0.97	0.03	0.03	0.07	1	1	451	488

SVA: SINE-VNTR-Alu; eQTL: expression quantitative trait loci; PD: Parkinson’s disease; AA: totally absent; AP: absent and present; PP: totally present.

The SVA genotypes are AA, AP, and PP.

Supplemental Table 1 provides all the cis and trans genes and transcripts regulated by the MHC SVA. The type of SVA subfamily is correlated weakly with the frequency of the SVA insertions within individuals. However, the youngest SVA_F subfamily members are at a lower frequency than the older SVA-A and SVA_D subfamily members (Table 1).

The statistically significant effects (P-values, FDRs and beta values) of the MHC class I and class II SVAs on the cis gene transcripts within the class I, class II, and class III regions are shown in Supplementary Table 2. The beta value, generated from linear regression model in Matrix eQTL package, represents the change in mRNA expressions with positive (upregulation) or negative (downregulation) associations.

The location of the class I and class II SVA insertions, respectively, and the relative positions of the genes that are regulated by these SVA are shown in Figures 3 and 4. For example, Figure 3(A) shows that NR_SVA_377 regulated four genes centromeric and nine genes telomeric of its insertion site in the region of the HLA-F AS1 gene within the alpha block. Moreover, the regulation by SVA_377 was limited mostly to the alpha block and its telomeric extended region (11 of 14 genes). In contrast, R_SVA_24 modified the transcription of 23 genes with a broader scope both within and between the blocks ranging from 12 genes in the alpha block to two genes in the delta block (Figure 3(B)). Whereas the class II SVA, NR_SVA_380 and R_SVA_27, had a broader reach on genes from the alpha to the epsilon block including the class III region, the influence of R_SVA_85 and NR_SVA_381 on gene transcription was limited mostly to the delta and epsilon blocks of the class II region (Figure 4(A) to (D)).

Figure 3.

The cis eQTL modulated by the four polymorphic SVA inserted within the MHC class I genomic region. Positions of SVA integration are indicated by labeled, colored boxes over black vertical lines. Horizontal arrows above the SVA boxes indicate the 5′ to 3′ direction of the retrotransposons. Genes modulated by SVA_377, SVA_24, SVA_25, and SVA_26 are indicated by (A) orange/yellow, (B) blue, (C) red, and (D) black vertical arrows, respectively.

Figure 4.

The cis eQTL modulated by the four polymorphic SVA inserted within the MHC class II genomic region. Positions of SVA integration are indicated by labeled, colored boxes over black vertical lines. Horizontal arrows above the SVA boxes indicate the 5′ to 3′ direction of the retrotransposons. Genes modulated by SVA_380, SVA_27, SVA_85, and SVA_381 are indicated by (A) orange/yellow, (B) blue, (C) red, and (D) black vertical arrows, respectively.

There are approximately 62 coding and noncoding non-HLA genes located between the MHC alpha and kappa blocks, and the kappa and beta blocks. SVA_24, SVA_25 and SVA_26 affected 13 of the non-HLA genes (Figure 3), further highlighting the SVA regulatory role to interconnect the multilocus expression of different HLA and non-HLA genes within the MHC. Despite a separation of over one megabase of nucleotide sequence, two of the four SVA in the alpha block (SVA_25 and SVA_26) and two of the four SVA in the delta block (NR_SVA_380 and R_SVA_27) coregulated some of the non-HLA genes in the class III region including the complement factor genes C4A and C4B, the spliceosome RNA helicase gene DDX39B, the leukocyte-specific transcript 1 protein gene LST1, a dimethylarginine dimethylaminohydrolase gene DDAH2 involved in nitrous oxide regulation, the heat shock protein 70 gene HSPA1A, and the lysosomal palmitoyl-protein thioesterase 2 gene PPT2. All these gene products have roles in the inflammatory response and may act to mitigate autoimmune disease. For example, DDX39B is an RNA–DNA helicase with known functions in mRNA splicing and nuclear export that regulates inflammatory responses,^46,47 and interacts with influenza A viral proteins ⁴⁸ and the pattern recognition receptor pathway that affects sensitivity to DNA damaging chemicals. ⁴⁹ LST1 regulates leukocyte abundance in lymphoid organs and inflammatory response in the gut, and the gene exons undergo extensive alternative splicing, giving rise to both membrane-bound (encoded by exon 3) and soluble isoforms. ⁵⁰

The non-HLA genes in the MHC class I region between the kappa and beta blocks impacted by the MHC class I SVA were POLR1H, PPP1R11, PAIP1P1, PRR3, DHX16, MDC1, CCHCR1, POU5F1, and TCF19 (Figure 2). Most of the non-HLA gene transcription levels modified by the MHC class I SVAs are involved with various mRNA and protein activities and processes such as transcription factors, RNA polymerase, ubiquitin-protein ligase, helicases, and RNA and DNA binding (Supplemental Table 3). The MHC class II SVAs had no significant influence on these genes.

SVA effects on trans eQTL

The different gene transcripts and chromosomes that were affected outside (trans) the MHC genomic region by the eight MHC polymorphic SVA are listed in Supplemental Table 4. Six of the eight SVA had eQTL trans effects. Only the regulatory effects of the R_SVA_26 insertion within class I and the NR_SVA_381 insertion within class II seemed to be limited totally to the MHC genomic region. The other SVA insertions within the MHC had trans effects involving all chromosomes except chromosomes 9 and 18. NR_SVA_377 and R_SVA_25 had the greatest number of trans results. For comparison, we included R_SVA_28 and NR_SVA_38 that are located outside the MHC between the ZBTB9 and BAK1 genes 48 kb from KIFC1 at the telomeric end of the extended MHC class II region (Table 2, Figure 2). ZBTB9 enables sequence-specific DNA binding and transcription protein binding activity and is involved in the regulation of transcription by RNA polymerase II. ⁵¹ BAK1 (BCL2 Antagonist/Killer 1) is a member of the BCL2 protein family and functions as a pro-apoptotic regular involved in a wide variety of cellular activities. ⁵² R_SVA_28 and NR_SVA_382 each effected one MHC gene (TAP2 and HLA-DPB2) and 200 and 451 genes, respectively, outside the MHC (Supplemental Table 1). Most SVA trans effects were split between coding and noncoding genes, but a biological significant number of coding changes appear to involve neurological functions and development (Supplemental Table 1).

Of the 85 gene transcripts impacted outside the MHC, 57 were expressed by known coding genes, six novel transcripts, nine long intergenic non-protein coding RNA, four antisense RNA, and nine pseudogenes. Among the coding genes were the five immunity genes, immunoglobulin heavy variable 2-26 (IGHV2-26), immunoglobulin lambda variable 9-49 (IGLV9-49), T cell receptor beta variable 7-3 (TRBV7-3), T cell receptor alpha variable 20 (TRAV20), and T cell receptor alpha variable 22 (TRAV22), and interleukin 1 receptor type 2 (IL1R2). Cytidine/uridine monophosphate kinase 2 is a rheostat for macrophage homeostasis, ubiquitin-specific peptidase 13 regulates HMGB1 DNA binding stability and is implicated in multiple diseases, whereas calnexin is a chaperone protecting protein structure, and the heat shock protein 90 beta family member 1 protects against oxidative stress and has a regulatory role in neurodevelopment. On the contrary, Fas cell surface death receptor leads to programmed death (apoptosis), mitogen-activated protein kinase 1 acts as an essential component of the MAP kinase signal transduction pathway, and the progesterone receptor regulates the expression of target genes in response to the progesterone steroid hormone. Some of the translated transcripts regulated by SVA have neurological functions such as PEST proteolytic signal containing nuclear protein (ubiquitination activity), gamma-aminobutyric acid type A receptor subunit beta2 (encephalopathy), acetylcholinesterase (neurotransmitter), and contactin 1 (glycosylphosphatidylinositol (GPI))-anchored neuronal membrane protein that functions as a cell adhesion molecule and that might be associated with neuropathic pain, demyelinating neuropathies, and with the formation of axon connections in the developing nervous system.

SVA effects on locus-specific differential transcript expression of HLA and non-HLA genes

The SVA differential effects on transcript variants of MHC genes including HLA and non-HLA genes are presented in Supplemental Table 2. SVA affected all the classical class I (HLA-A, -B, -C) and class II (HLA-DR, -DQ, -DP) genes, and some of the coregulated differentially expressed nonclassical HLA genes (HLA-F, -G, -H), HLA class I pseudogenes (HLA-K, -U, -V, -W, -L, -J, -T), and HLA class II pseudogenes (HLA-DRB9, -DRB6). None of the class I or class II SVA significantly impacted on the expression of the nonclassical class I HLA-E gene, although two of them downregulated the HLA-L pseudogene in the kappa block. SVA affected the HERV-16 related lncRNA genes, HCP5 in the beta block ²⁹ and HCP5B in the alpha block, ⁵³ as well as the HCG4B lncRNA in the alpha block and HCG27 lncRNA in the beta block. It is noteworthy that the differentially expressed transcript isoforms were mostly upregulated, but that some were downregulated simultaneously.

Box plots of transcriptional activities based on the SVA genotypes (1, absent–absent; 2, absent–present; 3, present–present) for the PPMI cohort are shown in Figures 5 and 6 and Supplemental Figure 1. Figure 5 shows box plots of regulation of the expression of seven different HLA-F transcript sizes (A to G) and two HLA-F-AS1 transcript sizes (H and I) by R_SVA_24 RIPs; whereas Figure 6 shows box plots of regulation of the expression of nine different HLA class II transcript sizes (A to I) and three TAP2 transcript sizes by NR_SVA_380 RIPs as two SVA regulatory examples. In general, most box plots showed that the number of transcripts for various expressed HLA and non-HLA genes are upregulated due to the presence of the SVA insertion. Most SVA insertion homozygotes are at a significantly lower frequency than the other genotypes (Table 2). Thus, the genes that are upregulated by low frequency SVA homozygotes or heterozygotes might occur in fewer individuals than the larger number of heterozygotes and homozygotes without the SVA insertions, ranging from less than 0.5% and 21% depending on the SVA insertion in the MHC class I region, and less than 1.3% and 97% depending on the SVA insertion in the MHC class II region (Table 2). In contrast, the R_SVA_85 and R_SVA_28 insertion frequencies of >90% suggest that most individuals within the cohort would have SVA upregulated genes, whereas a few homozygous individuals without the SVA insertions might have the same genes downregulated.

Figure 5.

Box plots of regulation of the expression of seven different HLA-F transcript sizes (A to G) and two HLA-F-AS1 transcript sizes (H and I) by R_SVA_24 RIPs as totally absent (AA, n = 745), absent and present (AP, n = 364), or total present (PP, n = 122) on the chromosomes of 1231 individuals. All transcripts were significantly affected as shown by the P-values and FDR and beta values in Supplemental Tables 1 and 2. The panels (A), (B), (D), (E), (G), and (I) show upregulation of expression and panels (C), (F), and (H) show downregulation of gene expression.

Figure 6.

Box plots of regulation of the expression of nine different HLA class II transcript sizes (A to I) and three TAP2 transcript sizes by NR_SVA_380 RIPs as totally absent (AA, n = 935), absent and present (AP, n = 311), or total present (PP, n = 19) on the chromosomes of 1266 individuals. All transcripts were significantly affected as shown by the P-values and FDR and beta values in Supplemental Tables 1 and 2. The panels (A), (B), (D), (E), (F), (K), and (L) show upregulation of expression and panels (C), (G), (H), (I), and (J) show downregulation of gene expression.

Downregulated genes shown in Supplementary Figure 1 for the R_SVA_24 insertion homozygotes included HLA-F-220 and HLA-F-223, HLA-G-207, HLA-A-203, HLA-A-204, HLA-W-207, and POLR1H-203. The downregulated genes for the R_SVA_25 insertion homozygotes were HLA-H-201, HLA-L-205, CCHCR1-201, CCHCR1-203, CCHCR1-209, MICA-214, HCP5-202, HCP5-206, DDX39B-208, DDX39B-214, and HLA-DQB2-232. Thus, these SVAs and others have differential effects at the level of transcription induction and reduction and splicing of introns and exons. Many of the downregulated gene transcripts might not be translated into proteins (Supplemental Table 2).

Effect of R_SVA_24 (alias SVA-HA) structural polymorphisms on differential expression of HLA-A

The R_SVA_24 insertion has a long and short variant and a separate internal variable number of tandem repeats (VNTR) – CT length genotypes that affect the expression of the different HLA-A isoforms (Figure 7). The violin plots shown in Figure 8C to J displayed the absence of a normal distribution, with many containing multiple peaks. The P-values generated by these tests were all highly significant (P = 4.4 × 10⁻² up to 2.22 × 10⁻¹⁶), bar 2. The impact of absent versus short SVA on HLA-A-204, and CT alleles A and B on HLA-A-201 expression, had P-values of 0.24 and 0.64, respectively (Figure 8(C) and (J)). Long and short SVAs, and different CT alleles were associated with differential expression and the impacts of these variations differed in a transcript-dependent manner. For example, the presence of either SVA length (short more so than long) significantly increased HLA-A-202 versus the absence of the SVA (Figure 8(D)). HLA-A-203 expression was greater in the presence of the long SVA than the short (Figure 8(E)), whereas expression of HLA-A-201 was significantly suppressed by either SVA (Figure 8(F)). Following this trend, expression of HLA-A-202 was significantly reduced by the presence of CT allele A versus allele B (Figure 8(H)), while the reverse was observed in HLA-A-203 (Figure 8(I)). Those P-values generated by likelihood ratio tests and multiple comparisons were substantially weaker for CT variants than SVA length (Figure 8(G) to (J) versus (C) to (F)). This might be caused by the large disparity in sample numbers for each data type. From all 550 samples genotyped, 486 had length metadata, whereas only 106 had viable CT allele calls.

Figure 7.

Variable SVA structures and effects on HLA-A gene transcription. (A) SVA structure diagram, highlighting highly polymorphic CT and VNTR regions. CT and VNTR copy numbers were thought to underlie SVA length variation. (B) Nucleotide sequences of CT alleles A and B. Nucleotide differences are boxed in red. (C) Exemplar PCR/gel electrophoresis of SVA showing two bands representing each length variant, long and short, indicated by red and purple arrows. (D and E) Example plots (HLA-A-202) generated for HLA-A expression against SVA CT allele A (light blue) and B (dark blue), and VNTR (length) variation, absent (green), long (red), short (purple). Y-axis transcript numbers (counts). X-axis SVA_24 CT allele A (n = 51) or B (n = 55) (D), SVA_24 genotype, absent (n = 268), long (n = 157), or short (n = 61) (E).

Figure 8.

Kruskal–Wallis multiple comparisons of HLA-A transcript expression relative to SVA length and CT allele. (A) Chromosome 6 ideogram, with location of the HLA-A locus shown in red. (B) Gene models of each HLA-A transcript (data derived from ensembl.org), exons indicated by solid blocks, introns by arrowed lines. Transcription start sites (TSS; left most exon) are unique for each transcript. All shown exons are protein coding. (C to F). Violin plots of HLA-A expression by SVA length genotype, absent (n = 268), long (n = 157), or short (n = 61). (G to J). Violin plots of HLA-A expression by CT allele genotype A (n = 51) or B (n = 55). Y-axis transcript numbers (counts). X-axis SVA_24 genotypes as length or CT allele.

Figure 9 shows the violin plots for the differential expression of HLA-A transcription variants stratified according to R_SVA_24 genotypes, homozygous present (PP), homozygous absent (AA), or heterozygous present–absent (PA). HLA-A-204, HLA-A-202, HLA-A-203 were upregulated significantly, whereas HLA-A-201 was downregulated.

Figure 9.

Expression of HLA-A isoforms. (A) HLA-A-204, (B) HLA-A-202, (C) HLA-A-203, and (D) HLA-A-201, stratified according to SVA_24 genotypes, homozygous present (PP), homozygous absent (AA), or heterozygous present–absent (PA). Y-axis transcript numbers (counts). X-axis SVA_24 genotype, AA, PA, or PP.

Discussion

Recent RNAseq studies of a PPMI cohort have shown that hundreds of SVA RIPs regulate or modify coding and noncoding genes, pseudogenes, and TEs at the transcription level within various genomic regions and across different chromosomes.^{27,40,54–56} Some SVA RIPs were associated with progression markers of PD, such as R_SVA_24 that was significantly associated with changes in the Hoehn and Yar stage (measure of symptom progression) in PD subjects. ⁴⁰ Here, we expanded the previous analyses of SVA_24, SVA_26, and SVA_27^27,40 to include an additional five SVAs and other targeted gene loci, differentially expressed transcripts and spliced variants expressed by the targeted genes within the MHC genomic region. The differentially expressed transcript isoforms were mostly upregulated in accordance with previous observations of SVA effects on gene transcription in the human genome. ²⁷ However, some of the transcripts of the genes that were either upregulated or downregulated might not be translated into functional proteins. For example, according to the Ensembl database, HLA-C-201, -C-203, -C-204 transcripts translate a full-length protein of 261aa-366aa, HLA-C-202 translates a truncated protein of 131aa, whereas the other HLA-C transcripts translate no protein because of retained introns or other unknown modifications. It is unclear what the outcome or purpose of the differential overexpression of the HLA classical genes might be in relation to immunity with the generation of different-sized HLA antigens that might bind and present peptides at the cellular membranes or be secreted as soluble antigens, and how often such molecular and cellular processes occur normally or pathologically. None of the SVA RIPs had any significant effects on the regulation of the HLA-E gene, a gene with two alleles that seems to act independently of the other classical and nonclassical class I HLA genes. ⁵⁷ Similar trends of shared and independent gene regulation were observed for the SVA RIPs in the class II region. Six of the eight SVA RIPs also had eQTL trans effects outside the MHC region. Only the regulatory effects of R_SVA_26 insertion within class I and NR_SVA_381 insertion within class II seemed to be limited to the MHC genomic region. The other SVA insertions within the MHC had trans effects involving all chromosomes except chromosomes 9 and 18. Thus, possible epistatic effects of these SVA on multiple gene interactions both inside and outside the MHC might either debilitate or protect regarding the pathogenesis of PD.

Koks et al. ²⁷ showed that R_SVA_27 significantly upregulated HLA-B, HLA-DRB1, and HLA-DRB5 and that R_SVA_26 also significantly upregulated HLA-B in an additive manner. In the present study, we observed that these two SVA RIPs and the six others regulated clusters of genes both within and outside the MHC in different ways, but occasionally with overlapping or possible additive effects on some of the same genes. Because of the different frequencies for the SVA RIPs within the MHC region of the PPMI cohort, it is likely that overlapping and additive effects of the expressed genes overall were low. Although we did not perform calculations or statistics for these potential additive dosage effects, we did note their coexistence in some of the same individuals. The overlapping and expression of genes regulated by the different MHC SVAs probably depend on the MHC haplotypes within the PPMI cohort for which we have no HLA allelic information at present. The eight SVA RIPs in this study were limited to the alpha and beta blocks of the HLA class I region and the beta and epsilon blocks of the MHC class II region. Three of these SVA RIPs annotated within the reference genome, GRCh38.p14/hg38 (https://genome.ucsc.edu) and were described previously in MHC haplotyped reference cell-lines and in the populations of European Australians, Japanese, and African Americans. ⁵⁴ The R_SVA_24 (alias SVA-HA), R_SVA_25 (alias SVA-HC), and R_SVA_26 (alias SVA-MIC) retrotransposons are inserted at a relatively low frequency in Australian European populations and strongly associated with the ancestral MHC haplotype 7.1AH, but not the 8.1AH.^20,21,39 SVA-HA, which is referred to as R_SVA_24 in this study, was first detected in Australian Europeans, Japanese, and African Americans at a homozygous frequency of about 10% and a strong LD with the HLA-A3/A11/A30, HLA-C7, and HLA-B7 alleles. ³⁹ This is half the homozygous insertion frequency of 19.8% for R_SVA_24 in the PPMI cohort (Table 2). Coincidently, 7.1AH along with other HLA class II alleles were associated with PD. ¹⁵ Although NR_SVA_381 (genotype PP) was the only MHC SVA RIP associated significantly with the risk of PD (data not shown), many other SVA RIPs outside of the MHC genomic region of the PPMI cohort were found to modify the progression of PD.^{27,38,40,54–56} Also, there are many other conserved (fixed) and structurally polymorphic SVA within the MHC than just the eight that we have examined in the PPMI cohort, and they might regulate gene expression in other population groups and disease cohorts. ⁵⁸

The eight MHC SVA insertions described in the present analysis are transcribed in whole blood samples and homozygous lymphoblastoid cell lines (unpublished findings). However, there are at least another 10 SVA within different loci of the human MHC genomic region including at least four that appear to have been fixed in human and chimpanzee genomes.^20,21,39 Two SVA RIPs integrated within the MHC class III region, one within an intron of the STK19 gene centromeric of C4A ⁵⁹ and the other within an intron of the C2 gene, ⁶⁰ were among the first SVA to be discovered and mapped precisely within the human genome. ⁶¹ Also, there are SVA-RIP near HLA-C (SVA-HC) and HLA-B (SVA-B)^20,21,39 that were not identified in the present analysis. We found no evidence here that these other SVA are transcribed or are associated with eQTL of MHC genes. The P values (corrected for multiple testing by the Bonferroni test), FDRs, and beta values that we obtained using an eQTL software computing package in this study is an extension of our previous published findings.^27,40 These statistical measures are reasonably reliable with FDR providing additional corrections for multiple testing. ⁶² However, it is possible that our high level of statistical stringency has lost some of the lower level biologically significant SVA effects on HLA and non-HLA genes. In this regard, our analysis may have underestimated rather than overestimated some of the SVA effects. Also, our study was limited to the blood cells of the PPMI cohort, whereas the SVA effects on the MHC genes are likely to vary in different tissues or cells, population groups, cohorts, disease groups, and individuals with different MHC haplotypes. Since we have not genotyped or haplotyped the individuals in our PPMI cohort, we could not make direct associations of the SVA effects with MHC haplotypes or HLA alleles. In this regard, our study is preliminary, and many other studies are still needed to follow-up and progress our findings.

We also examined in more detail the structural and regulatory effects of R_SVA_24 on HLA-A gene expression than previously reported.^27,40 Size variations in the SVA that were associated with the differential expression of HLA-A transcripts (Figure 9) may have been in part mediated by transcription factor binding to the VNTR region.^40,56 For example, transcription factors YB-1 and CTCF were shown to bind to a polymorphic VNTR in the SLC6A4 gene in a copy number-specific manner. ⁶³ Such in vitro studies have yet to be performed on the MHC SVA sequences. The violin or box plots (Figures 5 to 9) nevertheless further highlight the differential effects of the SVA RIPs such as R_SVA_24 on the splicing mechanisms involved in generating the different HLA transcript sizes that in turn probably affect the HLA antigen structures, membrane attachments, solubility, and binding to beta-2-microglobulins, T cell and killer cell receptors. Three of the four SVA insertions (NR_SVA_377, R_SVA_24, R_SVA_25) within the MHC class I region modified the HLA-A gene expression and transcription variants, whereas none of the four SVA insertion within the class II region had any effect on HLA-A gene expression. This provides further evidence for the hypothesis that SVAs have a differential regulatory role in HLA transcription, and that this role is dependent on the SVA integration site, the SVA variable sequence type, and possibly also the gene allele.

The present analysis of eight SVA RIPs within the MHC genomic region further highlights the importance of these REs in the multigenic regulation of HLA and non-HLA gene expression (eQTL) both within (cis) and outside (trans) the MHC at the level of (1) homozygosity and heterozygosity of the SVA RIPs, and (2) the expressed transcripts as spliced variants of different sizes and exonic representations. Moreover, these SVA RIPs are modulators or regulators of multigenetic expression that integrate the MHC with other genomic regions as part of a complicated and dynamic network or “interactome” in regulating human health and disease.^64,65 The mechanism by which these SVA regulate multiple genes over long distances both within and outside the MHC on various chromosomes is not known. Various biochemical and genetical mechanisms of long-range chromosomal interactions and gene regulation probably have a role including epigenetic processes, chromatin loop formation, nuclear matrix proteins, and RNA,^66,67 and hundreds of genomic anchors that organize chromatin loop domains across the MHC. ⁶⁸ Since the eight MHC SVA insertions that we described in the present analysis are transcribed, it is likely that the SVA RNAs regulate both the multigene networks and alternative splicing outcomes, involving mechanisms such as SVA exon capture^69,70 and/or interaction with distant promoters and enhancers by way of 3D chromatin structures.^65,67 The mechanisms involved in the long-range regulation of multigene expression by SVA insertions is a challenge that still needs considerable research and modeling with coherent and testable hypotheses.