Lack of evidence for a role of PIWIL1 variants in human male infertility

Dear editor

Infertility affects one in six couples, half of which is explained by a male factor (Tüttelmann et al., 2018). While thousands of genes are involved in spermatogenesis, there is a lack of diagnostically relevant genes. Critical evaluation of newly reported candidate genes is important before incorporating these into the diagnostic work-up. As part of a recently performed clinical validity assessment (Oud et al., 2019), an effort by the International Male Infertility Genomics Consortium (IMIGC, http://imigc.org), a closer look was taken at the quality of the evidence described in this journal by Gou et al., 2017 for involvement of Piwi Like RNA-Mediated Gene Silencing 1 (PIWIL1; also known as HIWI) in human male infertility.

The mouse homologue of PIWIL1, Piwil1, plays a crucial role in producing the piRNA pool, and by extension shaping the mRNA pool, in haploid germ cells (Deng and Lin, 2002; Gou et al., 2014). Gou et al. (2017) expanded on Piwi1’s regulatory repertoire by showing its role in the remodeling of spermatid chromatin through a protein-protein interaction with RNF8. The authors also studied the role of PIWIL1 in human male infertility by investigating the conserved destruction box (D-box) region which was previously shown to be involved in ubiquitin-mediated PIWIL1 degradation in late spermatogenesis (Zhao et al., 2013). Sanger sequencing of the D-box element of PIWIL1 in a cohort of 413 idiopathic azoospermia patients revealed several heterozygous mutations in three patients (Figure S1A).

While we do not dispute their mouse data, the human sequencing data are much less convincing with questionable Sanger peaks indicating mutations shown in Figure 1B of Gou et al. The mutations described in the human patients, if true, point to a mutational mechanism, which, to our knowledge, has never been reported (see Figure S1A and B).

Three years since this high impact publication, no other group has reported a replication of these findings. Therefore, we tested whether variants in the D-box of PIWIL1 are associated with severe spermatogenic failure, by combining genomic data from IMIGC members.

To investigate whether the PIWIL1 D-box region is a mutational hotspot in patients with severe spermatogenic failure, we studied the frequency of D-box mutations in patients (n=2,740) suffering from azoospermia (n=1,950) or severe oligozoospermia (n=790). Although we obtained high-quality data for the D-box region in all patients, no sequencing variants affected the D-box region (Figure S1C). If D-box variants were to explain 0.7% of our azoospermia cases, as reported in Guo et al., the binomial probability of sampling 0 D-box mutant cases in 1,950 azoospermic men is very small under standard assumptions about human mutation (p=6.70x10^-07). If one were to assume the expected D-box mutation rate to be the same in azoospermia and oligozoospermia, the combined sampling probability for 0/2740 cases is even smaller (p=2.11x10^-09).

To test whether PIWIL1 variants in other domains intolerant to variation may cause infertility (Figure SID), we screened the same set of 2,740 patients as well as a control group of 3,347 men who conceived normally, for rare non-synonymous variation (<1% allele frequency in any subpopulation of gnomAD, Figure S1E-G). In both groups, we detected 19 rare non-synonymous variants in PIWIL1, which did not cluster to one of the variation intolerant domains (Figure S1E-G). There was no significance difference in the number of rare variants between the patient group and the control group (p=0.65; OR 1.22; 95% CI 0.65–2.31). Also, there was no significant clustering of rare missense variants in PIWIL1 (see Supplementary Methods; patients: qualifying variants = 14, |X_i–X_j|=754.31, δ_g=3.98x10⁵, p=0.28; controls: qualifying variants = 17, |X_i–X_j|=974.46, δ_g=1.68x10⁵, p=0.82)(Lelieveld et al., 2017), indicating that there are no spatially clustered gain-of-function or dominant-negative variants in critical domains of PIWIL1. Exome sequencing data from 185 patient-parent trios did not reveal de novo mutations in PIWIL1.

In four of 2,740 patients, we identified a heterozygous loss-of-function (LoF) variant in PIWIL1 (Figure SIE). Based on the expected versus observed number of LoF variants found in 141,456 individuals of gnomAD version 2.1.1 (Lek et al., 2016), PIWIL1 is not predicted to be intolerant to LoF variation (pLi=0.0; oe=0.57 [lower and upper bound = 0.42–0.77]) and is, thus, unlikely to cause disease whenhaploinsufficient. This point is further strengthened by the discovery of two heterozygous LoF variants in fertile male controls (Figure S1F). There was no significant difference between the number of LoF variants in patient and control groups (p=0.51; OR 2.44; 95% CI 0.45–13.35). These findings are consistent with observations in mice that heterozygous knock-outs are fertile (Deng and Lin, 2002).

There are several potential explanations for the previously reported results on the D-box region. Possibly, the quality of the DNA or sequencing was suboptimal or genomic rearrangements in studied individuals confounded their results. The authors provided us with the primers and PCR conditions used for Sanger sequencing. In silico PCR predictions and analysis of alignable scaffold-discrepant positions due to an alternative locus for PIWIL1 in the most recent genome build GRCh38/hg38, did not reveal possible confounding factors (data not shown). However, the presence of a different, as yet unrecognized alternate haplotype of PIWIL1 in the population sampled by Gou et al., could potentially explain the unusual patterns of D-box variation summarized in Figure S1A. Validation of the original primer sequences resulted in high-quality Sanger peaks in 14 of our own patients and 2 controls. DNA of the samples reported by Gou et al. was unfortunately not available. It thus remains unclear whether the experimental design or the sequencing methods used had any effect on the aberrant results.

In conclusion, we express our concerns about the claimed role of the PIWIL1 D-box region mutations in human azoospermia as reported by Gou et al. Our data on 2,740 infertile men do not provide any evidence that variation in the D-box, or anywhere else in the PIWIL1 gene is causally linked to human male infertility. While that does not exclude a potential role for this gene in the disease, it is clear that PIWIL1 is not a frequently mutated male infertility gene.