Semen quality is affected by HLA class I alleles together with sexually transmitted diseases

Abstract

Background:

The human leukocyte antigen (HLA) locus includes several genes with key roles in antigen presentation and immune response, some of them inclusively found to be associated with non-obstructive azoospermia. Still, HLA connections to other infertility phenotypes such as semen hyperviscosity (SHV), asthenozoospermia (AST) and oligozoospermia (OLI) have been often neglected.

Objectives:

In this work, we aimed to evaluate the association of HLA class I and II genes with SHV, AST and OLI phenotypes while exploring a possible role in an adaptive immune response to sexually transmitted diseases (STD).

Materials and methods:

Whole exome sequencing was performed in a Portuguese cohort of 71 infertility cases and 68 controls, followed by HLA typing using a specific software - HLA*PRG:LA tool. Molecular screenings of seven STD were carried out in a subset of 72 samples (30 cases and 42 controls).

Results:

Statistical tests uncovered three protective alleles: HLA-A*11:01, associated with all forms of male infertility (P=0.0006); HLA-DQB1*03:02 with SHV and OLI (P_SHV=0.0303, P_OLI=0.0153) and HLA-A*29:02 with OLI (P=0.0355), which was found to interfere in sperm number together with HPV (P=0.0313). Five risk alleles were also identified: two linked with SHV (HLA-B*50:01, P=0.0278; and HLA-C*06:02, P=0.0461), another one with both SHV and OLI (HLA-DQA1*05:01, P_SHV=0.0444 and P_OLI=0.0265) and two with OLI (HLA-C*03:03, P=0.0480; and HLA-DQB1*03:01, P=0.0499). Here, HLA-C*03:03 carriers tend to be HPV infected.

Conclusions:

The application of HLA*PRG:LA tool to the study of male infertility, provided novel insights for an HLA correlation with semen quality, namely among SHV and OLI phenotypes. The discovery of an HLA-A*29:02/HPV crosstalk, together with former reports of HLA alleles conferring resistance-susceptibility to diverse human pathogens raises the hypothesis of a mechanistic link between male infertility, HLA polymorphism and host response to STD.

INTRODUCTION

Infertility is a major health issue that affects 10–20% of couples within reproductive age and in which male factors, either alone or in combination with female causes, are known to contribute to its etiology in ~50% of cases ¹. More specifically, male infertility is a multifactorial disorder with a large phenotypic variability in semen quality, including: azoospermia (AZO) and oligozoospermia (OLI), for spermatozoa absence or low counts, respectively; asthenozoospermia (AST), for reduced spermatozoa motility; and semen hyperviscosity (SHV), for the persistence of semen viscous proprieties after its expected liquefaction time ². Although such abnormalities may in some instances be correlated with congenital defects, endocrine dysfunction, urogenital infections and genetic anomalies, in a high proportion of cases the underlying causes for male infertility remain unknown ³.

In the last decade, several genome-wide association studies (GWAS) shed some light into the genetic makeup of this disorder, mainly concerning non-obstructive azoospermia (NOA), by uncovering novel susceptibility genes like PDE3A (phosphodiesterase 3A), SOX5 (SRY-box 5), DMRT1 (doublesex and mab-3 related transcription factor 1) and HLA-DRA (major histocompatibility complex, class II, DR alpha) ^4–6. Aside from these GWAS, some candidate gene approaches centered on HLA classes I and II also reported several associations with specific HLA alleles (HLA-A*26, HLA-A*28, HLA-A*33, HLA-B*18, HLA-B*44, HLA-DPB1*04:01, HLA-DQB1*06:04 and HLA-DRB1*13:02) and idiopathic male infertility ^7–10. However, once again these studies were predominantly focused on NOA patients and in most instances in men of East Asian ancestry.

In overview, the human leukocyte antigen (HLA) locus (6p21.32–22.1) is one of the most peculiar regions of our genome, characterized by high levels of polymorphism and heterozygosity, as well as extended linkage disequilibrium (LD) across different alleles. Briefly, the HLA system contains several genes and pseudogenes that can be divided into class I, II and III ^11,12 (Supplementary Fig. S1A). Even though all three classes are involved in the immune response, HLA class I and II molecules are the ones in charge of antigen presentation to T cells. Classical HLA class I genes, HLA-A, -B, and -C, are expressed in most cells, where they code for distinct α-chains of HLA class I heterodimers and where the β-chain is encoded by β2-microglobulin (B2M; 15q21.1; Supplementary Fig. S1B). On the other hand, classical HLA class II genes are expressed on specific immune cells, such as B cells, activated T cells, macrophages and dendritic cells, and are organized in three families: HLA-DP, -DQ and -DR. Each family comprises at least two genes coding for α- and β-chain molecules (Supplementary Fig. S1). Here, HLA-DR family accommodates a copy number variation (CNV), determining the origin of β-chains in HLA-DRB3, -DRB4 or -DRB5 genes (Supplementary Fig. S1C)¹³.

Importantly, several alleles of HLA class I and II genes have already been correlated with sexually transmitted disease (STD) caused by Neisseria gonorrhoeae (gonorrhea) ^14,15, Chlamydia trachomatis ^14,16, Treponema pallidum (syphilis) ^17,18, human immunodeficiency virus (HIV) ^19,20, human papilloma virus (HPV) ²¹, hepatitis B virus (HBV)²² and hepatitis C virus (HCV) ^23–25. Such microbial agents per se and together with activated immunological processes can lead to anatomical obstruction, compromise normal gland function, affect the availability of diverse seminal components and interfere in spermatozoa viability ^26–30. To be more precise, HIV, HPV and HBV, can be internalized by spermatozoa causing DNA damage and altering sperm number, motility and morphology, while in case of HBV, it can also induce the production of antisperm antibodies (ASA) ^29,31–37. Likewise, C. trachomatis and N. gonorrhoeae infections, as well as other bacteriospermias in general, may reduce semen quality by multiple mechanisms such as genital tract dysfunction, impaired spermatogenesis and bacterial cellular interactions ^38–42.

In this work, we investigated the association of HLA class I and II genes to SHV, AST (two less explored infertility phenotypes) and OLI, by applying for the first time the HLA*PRG:LA tool ⁴³ to a whole exome sequencing (WES) panel generated in a Portuguese cohort of patients and controls. The algorithm implemented through HLA*PRG:LA tool has the advantage over standard alignment methods used for the analysis of next-generation sequencing data, since it compares collected reads against multiple available HLA sequences instead of a single human reference. Thus, this method allows to circumvent the mapping problems arising from high polymorphism levels and from the presence of multiple paralog genes within the HLA locus. Then, for those HLA alleles showing significant associations with male infertility, we screened seven well known pathogens to test the hypothesis of increased susceptibility/resistance to STD.

Overall, our results provide further evidence for a possible association of HLA locus variability with male infertility, which we propose to be connected with an adaptive immune response to sexually transmitted pathogens often underlying asymptomatic urogenital infections or on the other hand, with an autoimmune reaction leading to the production of anti-sperm antibodies.

MATERIAL AND METHODS

Ethical approval

This study has been conducted in accordance with the ethical standards of the involved institutions and with the Helsinki Declaration. All participants provided written informed consent.

Patient recruitment and classification

A total of 139 samples were selected from a previous cohort of Portuguese men undergoing spermiogram analysis, for which peripheral blood and semen were collected ^44,45. Since our infertility patients (N=71; mean age=37.3±4.33) showed several abnormal parameters in spermiogram, samples were stratified into three non-mutually exclusive phenotypes: SHV (semen drops form a thread >2cm long; N=61), AST (rapid progressive motility <25%; N=45) and OLI (sperm counts <20million/mL; N=28) ². Samples that did not fit any of these infertility criteria were considered as controls (N=68; mean age=36.6±4.29). This cohort was excluded for male accessory gland infections (negative results in seminal culture), four sexually transmitted virus (HIV; HBV; HCV; and human T lymphotropic virus - HTLV) and syphilis (T. pallidum) ^44,45. Individuals with known infertility causes, including chromosome anomalies and Yq microdeletions, were not considered in this study.

Whole exome sequencing (WES)

DNA was extracted from peripheral blood leukocytes using Citogene Blood Kit (Citomed). For cases, exonic and UTR regions were captured by SureSelect Human All Exon V5+UTR enrichment kit and paired-end sequenced on an Illumina HiSeq 2000 (Macrogen Inc, Seoul, South Korea). Raw reads were mapped to the human reference genome (GRCh37) using BWA-MEM v.0.7.15 with default parameters ⁴⁶. Picard v.1.138 (http://broadinstitute.github.io/picard/) and Samtools v.1.2 ⁴⁷ were applied for post-alignment processes such as indexing, sorting and marking duplicates. The alignments were further submitted to indel realignment and base quality score recalibration using the Genome Analysis Toolkit (GATK) v. 3.4–0-g7e26428 ⁴⁸.

Control samples were analyzed in the framework of the Genetics of Male Infertility Initiative (GEMINI; http://gemini.wustl.edu/), aiming to map the genetic architecture of severe male fertility. Up to 39.1 Mb of exonic regions were captured using in-house exome targeting reagent and submitted to paired-end sequencing on Illumina HiSeq 4000 at the McDonnell Genome Institute of Washington University (genome.wustl.edu). The mapping and alignment procedures were similar to the approach applied to cases. In summary, raw reads were processed with Picard v.2.10.0 and samtools v.1.6 and mapped to the human reference genome assembly GRChg38 in an alternate contig-aware manner using BWA-MEM v.0.7.17 followed by INDEL recalibration and base quality score recalibration using GATK v.3.6.0. These WES data were extracted from two larger and ongoing works yet to be published.

HLA type inference

HLA typing using WES data was performed with HLA*PRG:LA program ^43,49, a bioinformatics tool published only on bioRxiv (not peer-reviewed). This tool has the advantage of automatically detecting the genome version of inputted BAM files and using a population reference graph (PRG) framework to impute HLA alleles for 10 genes (HLA-A, -B, -C, -DQA1, -DQB1, -DPA1, -DPB1, -DRB1, -DRB3 and -DRB4). Two caveats of HLA*PRG:LA is that the model does not take into account the number of copies of HLA-DRB paralogs nor analyzes HLA-DRB5. Hence, we followed HLA*PRG:LA method recommendations and used the obtained coverage of HLA-DRB3 and -DRB4 together with HLA-DRB1 alleles to infer the CNV: HLA-DRB3, -DRB4, or -DRB5 ^13,43. Moreover, the presence/absence of HLA-DRB5 was confirmed in all individuals by two amplicons. Positive samples were sequenced with BigDye Terminator v3.1 Cycle Sequencing Kit on an ABI 3130 automated sequencer. PCR conditions for HLA-DRB5 are described in Supplementary Table S1.

Molecular screening of sexually transmitted diseases (STD)

A subset of 72 samples (42 controls and 30 cases, distributed by phenotypes as 29 SHV, 20 AST, 16 OLI) was evaluated for seven well-known agents of STD: C. trachomatis, N. gonorrhoeae, Ureaplasma urealyticum, Mycoplasma genitalium, herpes simplex virus −1 and −2 (HSV-1; HSV-2) and HPV. To this end, total DNA was extracted from seminal plasma using QIAamp DNA mini kit (Qiagen) and several duplex and multiplex PCR reactions were carried out using primers reported in the literature ^50–53 (for PCR conditions see Supplementary Table S2). All samples were amplified in triplicate and all PCR reactions were performed in the presence of negative and positive controls for each one of the tested STD. To confirm PCR specificity, obtained amplicons were submitted to Sanger sequencing (BigDye Terminator v3.1 Cycle Sequencing Kit on an ABI 3130 automated sequencer) and inspected using BLAST tool and NCBI database (https://www.ncbi.nlm.nih.gov/).

Statistical analysis

To consider all possible associations to male infertility, four sets of comparisons were carried out in this study – all cases, SHV, AST or OLI.

Hardy–Weinberg equilibrium (HWE)

The HWE was evaluated at gene-level for controls and cases, separately, using PyPop v0.7.0 ⁵⁴ and assuming Guo and Thompson’s test ⁵⁵. Then, Chen’s measure was used to identify those genotypes explaining the observed departures from HWE ⁵⁶.

Association tests for HLA alleles

Identified HLA alleles (4-digit resolution, corresponding the first 2-digits to the allele group and the remaining ones to specific HLA proteins) were divided into common or rare according to their frequency (f) in the control sample (f ≥ 0.01 and f < 0.01, respectively). For common alleles, differences between cases and controls were estimated using Fisher’s exact test implemented in PyHLA package ⁵⁷. Next, a Bonferroni correction was used to adjust P-values for multiple comparisons in which the number of observed common alleles per gene was used as the number of tests. Welch’s ANOVA was also performed for those alleles showing associations with either AST or OLI and nominal P-values < 0.05 (IBM SPSS Statistics v.25). For these tests, the entire cohort was used to evaluate the effect of HLA alleles on quantitative spermiogram parameters: sperm motility (fast forward progression) and sperm number, respectively. Taking into account the limited statistical power of these approaches to detect rare variant associations, an enrichment analysis of HLA alleles with f < 0.01 was assessed by a simplified CAST test ^58,59. Essentially, the number of chromosomes carrying rare alleles was compared between cases and controls using a 2 × 2 contingency table and Fisher’s exact test.

Haplotype analysis

To evaluate the hypothesis of a non-random association of HLA alleles due to local high LD levels, pairwise D’ statistics were calculated using Pypop software ⁵⁴. For allele pairs found to be strongly correlated (D’ > 0.85), the corresponding haplotypes were inferred using the bigdawg R package ⁶⁰, and their association to different infertility phenotypes was addressed by Fisher’s exact tests. Moreover, to assess if those allele pairs contribute to a phenotype in a combined manner or if one of them results from a spurious association caused by LD, we used the statistical genetic approach called “interaction analysis” from PyHLA software ⁵⁷.

Amino acid sharing analysis

To circumvent a possible lack of statistical power due to a reduced sample size or because of multiple comparisons, an amino acid sharing approach was implemented in our analysis. Specifically, association tests at residue level were performed for both common and rare alleles with the bigdawg R package using only HLA regions known to encode antigen recognition sites (exon 2 and 3 for class I and exon 2 for class II genes) ⁶⁰. Contingency tables and χ² tests were first performed for each polymorphic amino acid position, followed by contingency tables and Fisher’s exact test for each residue with significantly associated positions. For variants with expected counts below five (cases or controls), comparisons were done in a single combined category. The results are expressed according to the mature protein numbering.

Copy number variation (CNV) association tests

In the association analysis of HLA-DRB3, -DRB4 and -DRB5, we used two approaches. In the first one, HLA-DRB3, -DRB4, -DRB5 or none of the previous were treated as alternative alleles of a single CNV locus (a 2 × 4 contingency table and χ² test was used). In the second approach, three bi-allelic CNV loci corresponding to the presence or absence of HLA-DRB3, -DRB4 and -DRB5 were considered (three 2 × 2 contingency tables and Fisher’s exact test were employed). In all instances, GraphPad Prism 5 software (GraphPad Software) was used to evaluate statistical significance of association.

Association tests for STD

To evaluate the strength of STD association to male infertility, we used the data from our molecular screening to perform Fisher’s exact tests (2 × 2 contingency tables for cases vs. controls). In addition, Welch’s ANOVA was carried out for sperm number and sperm motility variables. Then, to address whether detected HLA associations could be related with a host response to STD, two statistical approaches were used: Fisher’s exact tests were employed to compare presence/absence of HLA allele with STD; and two-way ANOVAs were carried out to accommodate the interaction between HLA allele, STD and the semen parameters “sperm number” or “sperm motility”. These analyses were performed for STD as a whole and for C. trachomatis and HPV separately. In all instances, IBM SPSS Statistics v.25 was used to evaluate statistical significance.

RESULTS

Fine-mapping of HLA alleles in male infertility

To assess a possible genetic influence of classical HLA class I and II genes into abnormal semen parameters, we analyzed the HLA allelic variation in 71 infertility cases and 68 controls genotyped as part of a whole-exome study (unpublished data). A large number of alleles were found in the full cohort: 113 common (f ≥ 0.01) and 63 rare (f < 0.01) (Supplementary Table S3). Globally, no violation of HWE was observed in our control dataset for HLA-A, -B, -C, -DQA1, -DQB1, -DRB1, -DPA1 and -DPB1 eliciting its usage in the following disease association tests.

Among common HLA alleles, eight showed significant differences (P < 0.05) between cases and controls (Table 1 and Fig. 1). Remarkably, seven of these associations could be correlated with one or two male infertility phenotypes. For instance, HLA-C*03:03 and HLA-C*06:02 were both detected as risk alleles for OLI and SHV, respectively. On the other hand, in HLA-DQB1 gene, whereas HLA-DQB1*03:01 was identified as an OLI risk allele, HLA-DQB1*03:02 was classified as a protective allele for OLI and SHV. The single association that appears to be transversal to male infertility (all cases) is correlated with HLA-A*11:01 allele that is absent from patients, while presenting a 7.4% frequency in controls. Even though this was the only allele reaching statistical significance after Bonferroni correction (P < 0.0029), other alleles displaying nominal P-values < 0.05 were correlated as well with the quantitative variables sperm number or sperm motility (Table 2).

Table 1 -.

Significant case-control associations for common HLA alleles.

Gene	Allele	Frequency
		Controls	All cases	SHV cases	AST cases	OLI cases
HLA-A	11:01	0.074	0	0	0	0
	29:02	0.081	0.028	0.025	0.022	0
HLA-B	50:01	0.015	0.063	0.074	0.056	0.071
HLA-C	03:03	0.022	0.042	0.041	0.056	0.089
	06:02	0.052	0.113	0.123	0.067	0.089
HLA-DQA1	05:01	0.199	0.289	0.312	0.289	0.357
HLA-DQB1	03:01	0.169	0.197	0.197	0.222	0.304
	03:02	0.132	0.063	0.049	0.056	0.018

Significant nominal P-values (P < 0.05) for Fisher’s exact test are underlined.

Significant P-value after adjustment for multiple testing (Bonferroni correction) are shown in bold.

Figure 1 -.

Risk effect analysis for HLA alleles in male infertility. The relative position of HLA genes (central diagram) is shown along with forest plots for alleles presenting significant associations with male infertility (P < 0.05). Each allele plot shows the Odd Ratios (OR) for four comparisons performed: Controls vs. oligozoospermia (OLI); Controls vs. asthenozoospermia (AST); Controls vs. semen hyperviscosity (SHV) and Controls vs. all cases. The dashed black line represents OR = 1. For each comparison OR values and confidence intervals are shown by a dot and a line, respectively, and where color denotes the allele effect: black – none (non-significant); blue – protective (significant); red – risk (significant).

Table 2 -.

Quantitative analysis of HLA alleles associated with male infertility (P<0.05).

Gene	Allele	Welch’s ANOVA P-value
		Sperm number	Sperm motility
HLA-A	11:01	–	0.0097
	29:02	0.0186	–
HLA-C	03:03	0.0360	–
HLA-DQA1	05:01	0.0511	–
HLA-DQB1	03:01	0.0218	–
	03:02	0.2657	–

Significant nominal P-values (P < 0.05) are underlined.

^-not performed

To overcome the low power of standard statistical methods to detect HLA rare allele associations, we performed an approach similar to the burden tests currently recommended for low-frequency variants with f < 0.01 ^58,61. Overall, a significant enrichment of rare alleles was identified for the entire HLA locus and across all four comparisons (Table 3). However, in a later gene-based analysis, the higher burden of rare alleles was found to be mainly attributed to HLA-A and -B, and SHV and/or AST phenotypes. These results also seem to explain the significant HWE departures detected for HLA-A in the full case set and in SHV patients (P = 0.0418 and P = 0.0486, respectively), which are probably connected with the occurrence of several unique genotypes (data not shown).

Table 3 -.

Enrichment analysis of rare HLA alleles among male infertility phenotypes.

Locus	P-value
	All cases vs Controls	SHV cases vs Controls	AST cases vs Controls	OLI cases vs Controls
HLAa	1.0443 × 10−5	8.0640 × 10−6	0.0003	0.0003
HLA-A	0.0179	0.0085	0.0660	0.0570
HLA-B	0.0220	0.0268	0.0418	0.0997
HLA-C	1	1	1	1
HLA-DPA1	1	0.4729	0.3982	0.2917
HLA-DPB1	1	1	1	1
HLA-DQB1	0.6226	0.6041	0.3039	0.2042
HLA-DRB1	0.0814	0.0714	0.1904	0.1219

^aall HLA genes covered in this study except HLA-DRB3, -DRB4 and -DRB5.

Significant nominal P-values (P < 0.05) for Fisher’s exact test are underlined.

Haplotype analysis

The pairwise LD analysis performed for the eight studied HLA genes uncovered nine regions with unusual LD levels (Supplementary Table S4). Among these, three could be connected with previously identified signals of association to male infertility (Table 1). A first one spanning HLA-A and -B (D’_SHV = 0.86), a second comprising HLA-B and -C (D’_SHV = 0.91) and a last one encompassing HLA-DQA1 and -DQB1 genes (D’_SHV =0.87; D’_OLI =0.87). Indeed, HLA-B*50:01 and HLA-C*06:02 were found to co-segregate in one of the most prevalent HLA-B~HLA-C haplotypes in SHV cases (f_controls = 0.007 and f_SHV= 0.066; P = 0.0118 and OR = 9.47). A similar finding was obtained for HLA-DQA1*05:01 and HLA-DQB1*03:01 alleles in the most common OLI haplotype (f_controls = 0.125 and f_OLI= 0.250; P = 0.0298 and OR = 2.33). On the other hand, the associations of HLA-A*11:01 and HLA-DQB1*03:02 (Table 1) to infertility phenotypes appear to be independent and not a by-product of the high LD levels observed for HLA-A~HLA-B and HLA-DQA1~HLA-DQB1 (Supplementary Table S4), given that neither of these alleles were linked to other significantly associated alleles.

In order to assess if co-segregating HLA alleles were both playing a role in SHV and OLI, or if one of them was a spurious association caused by LD, we tested the hypothesis of a combined allele contribution to the disease. According to our results, in both instances, the two alleles are likely to confer an increased infertility risk (HLA-B*50:01 / HLA-C*06:02 pair P = 0.0117 and OR = 7.53; HLA-DQA1*05:01 / HLA-DQB1*03:01 pair P = 0.032 and OR = 3.24).

Amino acid association analysis

To unravel functionally important domains of HLA molecules that otherwise may not emerge as associated to disease, we performed an analysis based on the amino acid variation. This approach led to similar results as the previous tests with HLA-A, -B, -DQA1 and -DQB1 showing associations to different infertility phenotypes (Table 1 and Supplementary Table S5). Interestingly, the protective effect of HLA-A*11:01 and HLA-A*29:02 to male infertility in general and to OLI, respectively, seems to be explained by the occurrence of a tyrosine (p.Y9 in HLA-A*11:01) or threonine (p.T9 in HLA-A*29:02) at residue 9 in prejudice of the phenylalanine (p.F9) (Fig. 2 and Supplementary Tables S5–S10). Concerning HLA-B*50:01 association with SHV, it is probably correlated with the increment of p.E152 in SHV over p.V152 (Fig. 2 and Supplementary Tables S6 and S10). In addition, for HLA-B the myriad of associations observed in AST is likely to be connected with the aforementioned enrichment of rare alleles and to a sharing of specific amino acid residues (Table 3 and Supplementary Tables S5, S8 and S10). Regarding HLA-DQA1*05:01, it appears to be related with the prevalence of p.S75 in OLI cases in contrast to p.I75 in controls (Fig. 2 and Supplementary Tables S9 and S10). Finally, the HLA-DQB1*03:01 allele conferring risk to OLI is possibly connected with p.A13, p.Y26 and p.45E in contrast to p.G13, p.L26 and p.G45 residues that are more common in controls and associated to the protective allele HLA-DQB1*03:02 (Fig. 2 and Supplementary Tables S9 and S10).

Figure 2 -.

Evaluation of amino acid composition of HLA-A, HLA-B, HLA-DQA1 and HLA-DQB1 alleles associated to male infertility (P < 0.05). Frequencies of the different residues encoded by HLA alleles showing significant results for semen hyperviscosity (SHV), asthenozoospemia (AST) or oligozoospermia (OLI). *P < 0.05; **P < 0.01; ***P < 0.001.

Copy number variation of HLA-DRB3, -DRB4 and -DRB5

We started our analysis by treating the CNV as a single locus (Table 4), which uncovered a significant result between OLI and controls (χ² = 10.4559 and P = 0.0151). Then, further tests were performed for the presence or absence of each paralog in a multiple loci approach that disclosed the association of HLA-DRB3 with SHV and OLI (P = 0.0454 and P = 0.0023, respectively) and of HLA-DRB4 with OLI alone (P = 0.0123; Table 4).

Table 4 -.

Association analysis for HLA-DRB3, -DRB4 and -DRB5 copy number polymorphism.

Gene	Frequencies
	Controls	All cases	SHV cases	AST cases	OLI cases
HLA-DRB3	0.375	0.472	0.500	0.511	0.625
HLA-DRB4	0.412	0.310	0.295	0.289	0.214
HLA-DRB5	0.096	0.099	0.098	0.078	0.071
NULLa	0.118	0.120	0.107	0.122	0.089
	P-value
	Gene	All cases vs Controls	SHV cases vs Controls	AST cases vs Controls	OLI cases vs Controls
Single locus		0.3145	0.1835	0.1874	0.0151
Multiple locus	HLA-DRB3	0.1152	0.0454	0.0545	0.0023
	HLA-DRB4	0.0815	0.0525	0.0671	0.0123
	HLA-DRB5	1.0000	1.0000	0.8118	0.7817

^aThe NULL designation corresponds to chromosomes lacking any copy of HLA-DRB3, -DRB4 or -DRB5.

Significant nominal P-values (P < 0.05) for Fisher’s exact test or χ² test are underlined.

STD screening in seminal plasma

To investigate whether HLA associations to male infertility could be related with susceptibility or resistance to STD, we screened seven well-known pathogens in a subset of 30 cases and 42 controls. Among tested STD, only C. trachomatis, N. gonorrhoeae, U. urealyticum and HPV were detected in the seminal plasma either alone or as co-infections (Supplementary Table S11). The most prevalent pathogens were C. trachomatis and HPV, but no differences were observed between cases and controls (C. trachomatis: f_controls = 0.429 vs. f_cases = 0.300; HPV: f_controls = 0.286 vs. f_cases = 0.400). Still, for HPV, previous reported associations with lower sperm numbers ^29,30 were confirmed in this study when this parameter was considered as a quantitative trait (Fig. 3). No similar association was observed for C. trachomatis or STD in general (Fig. 3).

Figure 3 -.

Effect of STD agents in sperm concentration. Group differences between sperm number and STD (all identified pathogens), Chlamydia trachomatis or human papilloma virus (HPV) presence/absence are shown.

Next, we examined whether STD prevalence diverged between carriers and non-carriers of HLA alleles associated to male infertility (nominal P < 0.05). For HLA-A locus, this revealed two times less STD affected subjects in the presence of the protective allele HLA-A*11:01 (P = 0.1096; OR = 0.2341; 95% CI = 0.0421–1.3017; Supplementary Fig. S2). On contrary, for HLA-C*03:03 a two-fold increase in HPV infected individuals was connected with this risk allele (P = 0.0906; OR = 4.6000; 95% CI = 0.7783–27.1885; Supplementary Fig. S2). Finally, statistical models of interaction were taken into account in the analysis of HLA allele, STD and infertility outcome. Notably, these tests showed that HLA-A*29:02 and HPV had an effect in sperm concentration (P = 0.0313), in which the highest sperm counts, even within the normal range, were observed in the presence of HLA-A*29:02 without HPV, differing from any other combination of analyzed variables (Fig. 4).

Figure 4 -.

Interaction analysis of HLA-A*29:02 allele and human papilloma virus (HPV) in sperm concentration. Mean sperm number and 95% confidence intervals are shown for four combinations according to allele presence or absence and to HPV infection status (HPV+ or HPV-). *P < 0.05; **P < 0.01; ***P < 0.001.

DISCUSSION

In this work, we evaluated the possible contribution of classical HLA variability into the etiology of different male infertility phenotypes, SHV, AST and OLI, through the analysis of a WES panel of Portuguese patients and controls. We identified three protective (HLA-A*11:01, HLA-A*29:02 and HLA-DQB1*03:02) and five risk alleles (HLA-B*50:01, HLA-C*03:03, HLA-C*06:02, HLA-DQA1*05:01 and HLA-DQB1*03:01). Notably, none of these have been previously linked to male infertility, neither former described associated alleles were replicated in this study ^{4,7–10,62,63}. This is not entirely unexpected given that the analyzed populations and phenotypes are quite distinct. Indeed, except for the work of Aleksovski et al. that analyzed men from the former Yugoslavia with impaired spermatogenesis (OLI or AZO), all the remaining studies were performed in NOA patients with East-Asian ancestry ^{4,7–10,62,63}. Moreover, in many instances, HLA alleles linked to male infertility are found at contrasting frequencies in different human populations ⁶⁴, indicating that previous findings might have been influenced by the ethnicity of analyzed samples and thus restricted to those populations. Interestingly, as pivotal players in the immune response, HLA molecules are thought to have been shaped by natural selection, and their polymorphism is known to correlate with geography (distance from Africa) and pathogen richness ⁶⁵. In this sense, the selective pressures exerted by a higher burden of infectious agents might have favored HLA diversification, especially in certain continental regions, to recognize a broader repertoire of pathogenic peptides ^65–68.

Despite the absence of a prevailing rule on how HLA classes access antigens, class I molecules are believed to mainly bind intracellular peptides of viral origin, triggering a cytotoxic response by their presentation to CD8+ T cells. Conversely, class II molecules are thought to bind mostly to extracellular antigens of bacterial origin that, once presented to CD4+ T cells, stimulate antibody production by B cells ^11,66,69,70. Therefore, HLA classes I and II would be expected to have dissimilar outcomes in STD like gonorrhea (N. gonorrhoeae) and syphilis (T. pallidum), and infections caused by C. trachomatis, HPV, HSV, HIV, HBV and HCV, which are all accepted risk factors for male infertility ²⁹. However, according to our study results, only class I genes (HLA-A and -C) appear to influence the outcomes of HPV infection and STD in semen quality.

Notably, some of our candidate alleles for male infertility were already associated in the context of other disorders with several of these pathogens. For example, HLA-A*11:01 and HLA-DQB1*03:02 were both described to confer resistance against HIV and in the case of HLA-DQB1*03:02 it was also reported to protect from HBV in chronic hepatitis ^19,20,22. Consistently, in our cohort, HLA-A*11:01 and HLA-DQB1*03:02 were correlated with a lower overall risk to male infertility and to SHV and OLI, respectively. These findings are therefore suggestive of a protective role through an improved immune response against HIV and HBV, but also possibly to other viruses, given that the presence of HIV and HBV was excluded in our samples ⁴⁵. Indeed, this assumption seems to apply to another class I allele, HLA-A*29:02, for which we found evidence of a relationship with HPV infection in sperm counts.

Concerning HLA-DQB1*03:01 allele, identified as a risk factor for OLI, previous studies showed conflicting results: whereas in chronic hepatitis this allele has been correlated in some instances with HCV clearance ^23–25, other authors have reported a link between HLA-DQB1*03:01 and HPV6 infection ²¹. Notably, this HPV subtype, despite being considered as low-risk for cervical cancer, is established to cause genital lesions and to be found in the semen ^71,72. In our cohort, no association of HLA-DQB1*03:01 with HPV was not detected. However, it is important to note that no discrimination of HPV subtypes was done in our STD screening. Given that some HPV subtypes have more severe outcomes in semen quality than others, our molecular analysis might have hampered the detection of any association signal ⁷³.

Less is known about bacterial agents, but for HLA-DQA1*05:01 allele correlated with OLI and SHV, a previous work showed contrasting results among women infected by C. trachomatis and N. gonorrhoeae. While in positive C. trachomatis cases HLA-DQA1*05:01 had an increased frequency, in N. gonorrhoeae patients this allele displayed only reduced counts ^14,29. However, as far as we could evaluate, no association of HLA-DQA1*05:01 with C. trachomatis, nor with N. gonorrhoeae (found in a single case), was detected in our sample.

Noteworthy, our cohort overlaps with the one used for a global characterization of seminal microbiota, in which pooled samples of controls, asthenoteratozoospermia, oligoasthenoteratozoospermia (OAT) and SHV were screened by sequencing 16S ribosomal RNA gene ⁴⁵. Again, according to Monteiro et al., the above association of HLA-DQA1*05:01 to Chlamydia alone seems improbable, given that its relative abundance compared to other bacterial taxa only reaches very low frequencies (f<0.1%). On contrary, Neisseria and other infectious agents known to cause urogenital infections, namely Haemophilus, Pseudomonas, Klebsiella, and Serratia, were found to be augmented in OAT and/or SHV samples ⁴⁵. Although the different methodological approaches and sample subsets used in these studies might have prevented a throughout appreciation of their findings, the complexity of seminal microbiota still suggests the existence of multiple targets for HLA antigenic presentation in the male urogenital system.

Taking into account the above considerations, it is tempting to hypothesize HLA alleles as critical immunological players in STD susceptibility. Supporting this conjecture is the evidence for an activation of the immune system in asymptomatic OAT patients as shown by their infiltrations of HLA-DR positive macrophages in the ejaculate and also the expression of HLA class II genes in spermatozoa of infertile men ^74,75. Nevertheless, we cannot rule out an autoimmunity cause for our findings, since HLA alleles can be associated with ASA production and male infertility as well ^76,77.

In conclusion, we were able to apply for the first time a population reference graph approach implemented through HLA*PRG:LA tool to the study of male infertility. This allowed us to infer HLA genotypes from a panel of WES data as well as to perform a preliminary study of HLA variability in this disease. Even though our cohort may be considered as small sized it was enough to provide novel hints for a possible role of HLA locus in this complex disorder. We demonstrate that both HLA class I and II alleles can be correlated with SHV and OLI phenotypes and that these also appear to be implicated in the susceptibility/resistance to different pathogens known to cause male urogenital infections and to affect reproductive fitness. Future studies in extended WES cohorts, followed by seminal transcriptomics research of human and pathogens will be fundamental to assess the proposed host-pathogen interaction in male infertility.