ARTICLE https://doi.org/10.1038/s41467-021-27132-8 OPEN A de novo paradigm for male infertility M. S. Oud 1,42, R. M. Smits2,42, H. E. Smith3,42, F. K. Mastrorosa 3, G. S. Holt3, B. J. Houston4, P. F. de Vries1, B. K. S. Alobaidi 3, L. E. Batty3, H. Ismail3, J. Greenwood5, H. Sheth 6, A. Mikulasova3, G. D. N. Astuti7,8, C. Gilissen 7, K. McEleny9, H. Turner10, J. Coxhead 11, S. Cockell12, D. D. M. Braat 2, K. Fleischer2, K. W. M. D’Hauwers13, E. Schaafsma14, Genetics of Male Infertility Initiative (GEMINI) consortium*, L. Nagirnaja 15, D. F. Conrad 15, C. Friedrich16, S. Kliesch 17, K. I. Aston18, A. Riera-Escamilla19, C. Krausz20, C. Gonzaga-Jauregui21, M. Santibanez-Koref3, D. J. Elliott3, L. E. L. M. Vissers 1, F. Tüttelmann 16, M. K. O’Bryan 4, L. Ramos 2, M. J. Xavier 3,43, G. W. van der Heijden 1,2,43 & J. A. Veltman 3,43 ✉ 1234567890():,; De novo mutations are known to play a prominent role in sporadic disorders with reduced fitness. We hypothesize that de novo mutations play an important role in severe male infertility and explain a portion of the genetic causes of this understudied disorder. To test this hypothesis, we utilize trio-based exome sequencing in a cohort of 185 infertile males and their unaffected parents. Following a systematic analysis, 29 of 145 rare (MAF < 0.1%) protein-altering de novo mutations are classified as possibly causative of the male infertility phenotype. We observed a significant enrichment of loss-of-function de novo mutations in loss-of-function-intolerant genes (p-value = 1.00 × 10−5) in infertile men compared to controls. Additionally, we detected a significant increase in predicted pathogenic de novo missense mutations affecting missense-intolerant genes (p-value = 5.01 × 10−4) in contrast to predicted benign de novo mutations. One gene we identify, RBM5, is an essential regulator of male germ cell pre-mRNA splicing and has been previously implicated in male infertility in mice. In a follow-up study, 6 rare pathogenic missense mutations affecting this gene are observed in a cohort of 2,506 infertile patients, whilst we find no such mutations in a cohort of 5,784 fertile men (p-value = 0.03). Our results provide evidence for the role of de novo mutations in severe male infertility and point to new candidate genes affecting fertility. A full list of author affiliations appears at the end of the paper. NATURE COMMUNICATIONS | (2022)13:154 | https://doi.org/10.1038/s41467-021-27132-8 | www.nature.com/naturecommunications 1 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-021-27132-8 M ale infertility contributes to approximately half of all severe male infertility and addresses the need for further inves- cases of infertility and affects 7% of the male popula- tigation in larger patient–parent trio cohorts to solidify these tion. For the majority of these men the cause remains results. unexplained1. Despite a clear role for genetic causes in male infertility, there is a distinct lack of diagnostically relevant genes and at least 40% of all cases are classified as idiopathic1–4. Pre- Results Discovery of de novo mutations in infertile male trios. In this vious studies in other conditions with reproductive lethality, such as neurodevelopmental disorders, have demonstrated an impor- study, we investigated the role of DNMs in 185 unexplained cases of oligozoospermia (<5 million sperm cells/ml; n = 74) and tant role for de novo mutations (DNMs) in their etiology5. In line with this, recurrent de novo chromosomal abnormalities play an azoospermia (n = 111) by performing whole exome sequencing (WES) in all patients and their parents (see Supplementary important role in male infertility. Both azoospermia factor (AZF) deletions on the Y chromosome as well as an additional X Figs. 1, 2, Supplementary Notes and Data for details on metho- dology and clinical descriptions). In total, we identified and chromosome, resulting in Klinefelter syndrome, occur de novo. Collectively, these de novo events explain up to 25% of all cases of validated 192 rare DNMs (MAF < 0.1%), including 145 protein- altering DNMs. All de novo point mutations were autosomal, nonobstructive azoospermia (NOA)1,4. Interestingly, in 1999 a DNM in the Y-chromosomal gene USP9Y was reported in a man except for one on chromosome X, and all occurred in different genes (Supplementary Data 1). Two rare de novo copy-number with azoospermia6. Until now, however, a systematic analysis of the role of DNMs in male infertility had not been attempted, even variations (CNVs) were also identified affecting a total of 7 genes (Supplementary Fig. 3). None of the 145 protein-altering DNMs though a pilot exome sequencing study in 13 infertile men and their parents was recently published7. This is partly explained by occurred in a gene already known for its involvement in auto- somal dominant human male infertility. This is not unexpected as a lack of basic research in male reproductive health in general4,8, but also by the practical challenges of collecting parental samples only four autosomal dominant genes have so far been linked to isolated male infertility in humans3,9. for this disorder, which is typically diagnosed in adults. In this work, we address this lack of knowledge by analysing exome sequencing data of 185 infertile males and their parents Intolerance analysis of genes with de novo loss-of-function and reporting on our findings of 29 DNM in these men which are mutations. Broadly speaking, across genetic disorders, dom- likely causative for the infertility phenotype, based on variant and inantly acting disease genes are usually intolerant to LoF muta- gene level evidence. We emphasize an enrichment for loss-of- tions, as represented by a high pLI score10 or a low LOEUF function (LoF) DNM in LoF-intolerant genes and missense DNM score11. In our cohort of infertile men, we detected a significant in missense-intolerant genes. We identify a number of promising enrichment in the number of LoF-intolerant genes with a LoF candidate genes for male infertility, including the mRNA splicing DNM (n = 17). No such enrichment was identified in a cohort of gene RBM5, which contains a possibly causative DNM in our trio 1,941 control cases from de novo-db v1.6.112 (median pLI in cohort, and possibly causative heterozygous variants in six patients with male infertility = 0.80, median pLI in controls = additional patients for which parental information is not avail- 3.75 × 10−5, p value = 1.00 × 10−5, N simulations = 100,000) able. This work suggests a potential role for DNM as a cause of (Fig. 1a). Similar results were obtained using the LOEUF scores Fig. 1 Analysis of the intolerance to loss-of-function and missense variation in genes with de novo mutations. a Violin plot with quantile lines showing pLI scores in all genes in gnomAD (red), all genes affected by rare protein-altering loss-of-function (LoF) de novo mutations (DNMs) in a control population (http://de novo-db.gs.washington.edu/de novo-db/) (green) and in all genes with a rare protein-altering LoF DNM in our trio cohort (blue). Using the permutation-based, nonparametric test defined by Lelieveld et al. 64 a significant enrichment of LoF DNMs in LoF-intolerant genes in patient cohort was detected in comparison to the number of LoF in fertile control cohort (DNM LoF mutations in patients n = 17, median pLI in patients with male infertility = 0.80, DNM LoF mutations in controls n = 21, median pLI in controls = 3.75 × 10−5, p value = 1.00 × 10−5, N simulations = 100,000). The black dot indicates median pLI scores. b Violin plot with quantile lines showing the distribution of Z-scores for genes with predicted benign (n = 59) and pathogenic missense DNMs (n = 63) in infertile patients. A significant increase in predicated pathogenic DNMs in missense-intolerant genes was detected compared to benign missense DNM (Two-sided Mann–Whitney U test, p value of 3.44 × 10−4). (***p value < 0.001). 2 NATURE COMMUNICATIONS | (2022)13:154 | https://doi.org/10.1038/s41467-021-27132-8 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-021-27132-8 ARTICLE (median LOEUF in patients with male infertility = 0.34, median was observed for the genes highlighted as likely benign (n = 61, LOEUF in controls = 0.59, p value = 1.00 × 10−5, N simula- PPI enrichment p value = 0.206) or those affected by synon- tions = 100,000) (Supplementary Fig. 4a). This observation indi- ymous DNMs (n = 35, PPI enrichment p value = 0.992, Supple- cates that LoF DNMs likely play an important role in male mentary Fig. 5). This suggests that the proteins affected by infertility, similar to what is known for developmental disorders predicted pathogenic DNMs share common biological functions. and severe intellectual disability13,14. As an example, a hetero- The STRING network analysis also highlighted a central zygous likely pathogenic frameshift DNM was observed in the module of interconnected proteins with a significant enrichment LoF-intolerant gene GREB1L (pLI = 1) of Proband_076. Homo- of genes required for mRNA splicing (Supplementary Fig. 6). The zygous Greb1l knockout mice appear to be embryonic lethal, genes U2AF2, HNRNPL, CDC5L, CWC27, and RBM5 all contain however, typical male infertility phenotypic features such as predicted pathogenic DNMs and likely interact at a protein level abnormal fetal testis morphology and decreased fetal testis volume during the mRNA splicing process. Pre-mRNA splicing allows are observed15. Interestingly, this patient has a reduced testis gene functions to be expanded by creating alternative splice volume and severe oligozoospermia (Supplementary Notes variants of gene products and is highly elaborated within the Table 1). Nonsense and missense mutations in GREB1L in testis25. One of these genes, RBM5 has been previously humans are known to cause renal agenesis16 (OMIM: 617805), not highlighted as an essential regulator of haploid male germ-cell known to be present in our patient. Of note, all previously pre-mRNA splicing and male fertility in mice26. Mice with a reported damaging mutations in GREB1L causing renal agenesis homozygous ENU-induced allele point mutation in RBM5 are either maternally inherited or occurred de novo. This led the present with azoospermia and germ cell development arrest at authors of one of these renal agenesis studies to speculate that round spermatids. Whilst in mice, a homozygous mutation in disruption to GREB1L could cause infertility in males15. A recent RBM5 is required to cause azoospermia, this may not be the case WES study involving a cohort of 285 infertile men also noted in humans as is well-documented for other genes27, including the several patients presenting with pathogenic mutations in genes recently reported male infertility gene SYCP29. Of note, RBM5 is with an associated systemic disease where male fertility is not a tumor suppressor in the lung28, with reduced expression always assessed17. affecting RNA splicing in patients with non-small cell lung We also assessed the damaging effects of the two rare de novo cancer29. HNRNPL is another splicing factor affected by a CNVs by looking at the pLI score of the genes involved. possible pathogenic DNM in our study. One study implicated a Proband_066 presented with a large 656 kb de novo deletion on role for HNRNPL in patients with Sertoli cell-only phenotype30. chromosome 11, spanning 6 genes in total. This deletion partially The remaining three mRNA splicing genes have not yet been overlapped with a deletion reported in 2014 in a patient with implicated in human male infertility. However, mRNA for all cryptorchidism and NOA18. Two genes affected in both patients, three is expressed at medium to high levels in human germ cells QSER1 and CSTF3, are LoF-intolerant with pLI scores of 1 and and all are widely expressed during spermatogenesis31. Specifi- 0.98, respectively. In particular, CSTF3 is highly expressed within cally, CDC5L is a component of the PRP19-CDC5L complex that the testis and is known to be involved in pre-mRNA 3′-end forms an integral part of the spliceosome and is required for cleavage and poylyadenylation19. activating pre-mRNA splicing32, as is CWC2733. U2AF2 plays a role in pre-mRNA splicing and 3′-end processing34. Interestingly, CSTF3, one of the genes affected by a de novo CNV in Missense intolerance in de novo mutation genes. To system- Proband_066, affects the same mRNA pathway18. atically evaluate and predict the likelihood of these DNMs caus- ing male infertility and identify novel candidate disease genes, we assessed the predicted pathogenicity of all DNMs using three DNMs uncovering recessive disease and analysis of maternally prediction methods based on SIFT20, MutationTaster21, and inherited mutations. Whilst DNMs most often cause dominant PolyPhen222 with a minimum of 2 of the 3 showing pathogenicity disease, they can contribute to recessive disease, usually in com- to define a variant as Pathogenic. Using this approach, 84 of 145 bination with an inherited variant on the trans allele. In order to rare protein-altering DNM were predicted to be pathogenic, while look for this, we analysed all DNM genes for the presence of the remaining 61 were predicted to be benign. To further analyse inherited mutations on the other allele in the same patient. In the impact of the variants on the genes affected, we looked at the Proband_060, who carried a DNM in Testis and Ovary Specific missense Z-score of all 122 genes affected by a missense variant, PAZ Domain Containing 1 (TOPAZ1) on the paternal allele, we which indicates the tolerance of genes to missense mutations23. did identify a maternally inherited variant predicted to be We identified no significant enrichment in missense DNMs in pathogenic (Supplementary Fig. 7). TOPAZ1 is a germ cell-spe- missense-intolerant genes in our infertile cohort when compared cific gene which is highly conserved in vertebrates35. Studies in to controls (median Z-score in male infertility patients = 0.83, mice revealed that Topaz1 plays a crucial role in spermatocyte, median Z-score in controls = 1.04, p value = 1, N simulations = but not oocyte, progression through meiosis36. In men, TOPAZ1 100,000) (Supplementary Fig. 4b). Interestingly, however, we is expressed in germ cells in both sexes31,37,38. Analysis of the observed a significantly higher median missense Z-score in genes testicular biopsy of this patient revealed a germ cell arrest in early affected by a missense DNM predicted as pathogenic (median Z- spermiogenesis (Fig. 3). score = 1.21, n = 63) when compared to genes affected by pre- Maternally inherited mutations can also result in dominant dicted benign (median Z-score = 0.98, n = 59) missense DNMs causes of male infertility if not affecting female fertility. We in our cohort (p value = 5.01 × 10−4, Fig. 1b). It should be noted therefore studied all DNM genes for the presence of maternally that the same analysis in controls showed no such significant inherited mutations in the entire cohort and compared this to the difference (Supplementary Fig. 4c). presence of paternally inherited mutations in the same genes. A total of 4 maternally inherited variants predicted to be pathogenic Protein–protein interactions reveal link to mRNA splicing. An were identified in DNM genes (TENM2 (2×), CWC25, and EVC). analysis using the STRING database24, revealed a significant All of these variants, however, were also observed multiple times enrichment of protein interactions amongst the 84 genes affected in an exome dataset from a cohort of 5784 fertile men suggesting by a protein-altering DNM predicted to be pathogenic (PPI that these maternally inherited variants are not causative of male enrichment p value = 2.35 × 10−2, Fig. 2). No such enrichment infertility (Supplementary Data 2). NATURE COMMUNICATIONS | (2022)13:154 | https://doi.org/10.1038/s41467-021-27132-8 | www.nature.com/naturecommunications 3 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-021-27132-8 Fig. 2 Protein–protein interactions predicted for proteins affected by pathogenic de novo mutations. Significantly larger number of interactions were observed in proteins affected by de novo mutations than expected for a similar sized dataset of randomly selected proteins (PPI enrichment p value = 2.35 × 10−2). The central module of the main interaction network (blue dashed circle) contains 5 proteins involved in mRNA splicing (Supplementary Fig. 6). Further analysis in additional cohorts of infertile males. In Fisher’s Exact test, p value < 0.05, Table 2). After applying the addition to all systematic analyses described above, we evaluated Bonferroni correction to counteract the effects of multiple testing, the function of all DNM genes to give each a final pathogenicity however, the only significant enrichment was observed in the classification (Table 1, details in Methods). Of all 192 DNMs, 29 RBM5 gene (adjusted p value = 0.03). In this gene, six infertile affected genes were linked to male reproduction and classified as men were found to carry a rare pathogenic missense mutation, in possibly causative, with a further 50 as unclear. For replication addition to the proband with a de novo missense mutation purposes, only one pilot study including 13 trios was recently (Supplementary Fig. 8, Supplementary Data 7). Importantly, no published in male infertility7. None of the DNM genes reported such predicted pathogenic mutations were identified in men in in this study showed DNMs in our cohort. To further study the fertile cohort. In line with these results, RBM5, already the DNM genes identified in our cohort, we looked for the highlighted above as an essential regulator of male germ cell pre- presence of rare predicted pathogenic mutations in these mRNA splicing and male infertility26, is highly intolerant to genes in exome datasets of infertile men (n = 2,506), in colla- missense mutations (missense Z-score 4.17). boration with members of the International Male Infertility In addition to the comparison between fertile and infertile Genomics Consortium and the Geisinger-Regeneron DiscovEHR men, we investigated whether there was any difference between collaboration39. For comparison, we included an exome the number of predicted pathogenic mutations carried in fertile dataset from a cohort of 11,587 fertile men and women from men compared to fertile women. However, none of the DNM Radboudumc. genes showed a significant difference between the sexes (Two- In the additional infertile cohorts, we identified 17 LoF tailed Fisher’s Exact test with Bonferroni correction adjusted mutations in our DNM LoF-intolerant genes (pLI ≥ 0.9), p values = 1, Supplementary Data 3, 5). although we did not detect a statistical enrichment in the LoF mutations in these genes compared to fertile men (Two-tailed Fisher’s Exact test with Bonferroni correction adjusted p values Phasing of de novo mutations to identify parent of origin. >0.05, Supplementary Data 3, 4). Next, we looked for an Given the predicted impact of these DNMs on spermatogenesis, enrichment of rare predicted pathogenic missense mutations in we were interested in investigating the parental origin of DNMs these cohorts (Table 2 and Supplementary Data 5, 6). A total of in our trio cohort. We were able to phase 29% (n = 59) of all our 11 genes showed an enrichment of pathogenic missense DNMs using a combination of short-read WES and targeted mutations in infertile men compared to fertile men (Two-tailed long-read sequencing (Supplementary Table 1). In agreement 4 NATURE COMMUNICATIONS | (2022)13:154 | https://doi.org/10.1038/s41467-021-27132-8 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-021-27132-8 ARTICLE Fig. 3 Description of control and TOPAZ1 proband testis histology and aberrant acrosome formation. a, b H&E stainings of (a) control and (b) Proband_060 with pathogenic mutations in TOPAZ1 gene. The epithelium of the seminiferous tubules in the TOPAZ1 proband show reduced numbers of germ cells and an absence of elongating spermatids based on the analysis of 150 seminiferous tubules in control and patient. c, d immunofluorescent labeling of DNA (magenta) and the acrosome (green) in control sections (c) and TOPAZ1 proband sections (d). (c) The arrowhead indicates the acrosome in an early round spermatid and the arrows the acrosome in elongating spermatids. Spreading of the acrosome and nuclear elongation are hallmarks of spermatid maturation. (d) No acrosomal spreading (see arrowheads) or nuclear elongation is observed in the TOPAZ1 proband. The asterisk indicates an example of progressive acrosome accumulation without spreading. Scale bar: 40 µm (a, b) and 5 µm (c, d). also be facilitated by the fact that from spermatogonia onwards, Table 1 De novo mutation classification summary. male germ cells form cysts and share mRNAs and proteins44. As such, the interconnectedness of male germ cells, which is essential Possibly Unclear Unlikely Not Total for their survival45, could mask detrimental effects of DNMs causative causative causative occurring during spermatogenesis. Missense 21 38 50 13 122 Frameshift 4 8 1 0 13 Stop gained 1 3 0 0 4 In-frame indels 3 1 1 1 6 Discussion Splice site 0 0 0 11 11 In 2010, we published a pilot study pointing to a de novo para- variant Synonymous 0 0 0 36 36 digm for mental retardation46 (now more appropriately termed TOTAL 29 50 52 61 192 developmental disorders or intellectual disability). This work contributed to the widespread implementation of patient–parent Rare DNMs were classified based on pathogenicity prediction, ACMG classification, number of cases in gnomAD and presence of the exact mutation in verified fertile men of the control WES studies in research and diagnostics for neurodevelopmental cohort, as well as functional data taking into account RNA expression in testis, RNA enrichment disorders47, accelerating disease gene identification and increas- in the testis or involvement in spermatogenesis, protein expression in the testis, model organisms, the protein function in relation to spermatogenesis and interactions with known ing the diagnostic yield for these disorders. The data presented fertility genes. here suggest that a similar benefit could be achieved from trio- based exome sequencing in male infertility. In order to achieve this there is an urgent need to expand on this work as larger with literature40–43, 72% of all DNMs occurred on the paternal studies are essential to identify recurrently DNM genes and fur- allele. Interestingly, phasing of 8 likely causative DNMs showed ther demonstrate the exact contribution of DNMs to male that 6 of these were of paternal origin (75%). This suggests that infertility. Modeling studies recently done for developmental DNMs with a deleterious effect on the future germline can escape disorders showed that more than 350k trios may be required to negative selection in the paternal germline. This may be possible have approximately 80% power to detect all haploinsufficient because the DNM occurred after the developmental window in genes causing this disorder48. Evidently, these numbers can only which the gene is active, or the DNM may have affected a gene in be reached by implementing trio-based exome sequencing as a the gamete’s genome that is critical for somatic cells supporting routine diagnostic test and by sharing these diagnostic data with the (future) germline. Transmission of pathogenic DNMs may the international research community. This research community NATURE COMMUNICATIONS | (2022)13:154 | https://doi.org/10.1038/s41467-021-27132-8 | www.nature.com/naturecommunications 5 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-021-27132-8 Table 2 Rare pathogenic missense mutations in exome data from various cohorts of infertile men and fertile control cohorts. Gene Missense Z-score Total infertile men Fertile Dutch men Burden test infertile vs. (n = 2,506) (n = 5,784) fertile men p value Adjusted p value RBM5 4.17 7 0 0.0002 0.03 HUWE1 8.87 6 0 0.001 0.12 REN 0.80 7 1 0.001 0.21 HIST1H1D −8.06 10 5 0.004 0.59 ABLIM1 1.62 6 1 0.004 0.60 FUS 2.21 4 0 0.01 1 CNOT4 3.49 5 1 0.01 1 CDC5L 2.78 6 2 0.01 1 ZNF629 3.86 6 2 0.01 1 PCDHB1 1.02 11 8 0.01 1 AK3 −1.97 10 7 0.02 1 Genes affected by a rare missense DNM were investigated in additional cohorts of infertile patients and a cohort of verified fertile men to identify other individuals carrying rare missense mutations. A burden test was used to compare the total number of predicted pathogenic missense mutations observed in the infertile vs. fertile men. A two-tailed Fisher’s Exact test was performed with and without Bonferroni correction applied to adjust p values for multiple testing of all 152 genes of interest. (Also see Supplementary Data 5). will also have the enormous task of functionally validating the Radboud diagnostics center as the healthy parent of a child with a severe illness. impact of these DNMs on spermatogenesis. Altogether, this will Although these men fathered a child with intellectual disability, their fertility is expected to be similar to an unselected sample of the male population. not only help to increase the diagnostic yield for men with infertility but will also enhance our fundamental biological understanding of human reproduction and natural selection. In Exome sequencing. WES samples were prepared and enriched following the addition, it will indicate whether male infertility follows a manufacturer’s protocols of either Illumina’s Nextera DNA Exome Capture kit or Twist Bioscience’s Twist Human Core Exome Kit. All sequencing was performed dominant inheritance pattern, and this has impact for disease on the NovaSeq 6000 Sequencing System (Illumina) achieving comparable results transmission. Couples that seek treatment for male infertility covering more than 99% of all exonic regions using either kit (Supplementary should be counseled on the risk of transmitting this condition to Table 2) and an average depth of 72× (Illumina’s Nextera Kit) and 99x (Twist their offspring, something that is now limited to couples receiving Bioscience’s Kit) (see Supplementary Fig. 9 and Supplementary Table 3). Sequenced reads were aligned to Human Reference Genome (GRCh37.p5/hg19) fertility treatment due to Y-chromosome deletions. Male inferti- using BWA-Mem v0.7.1751, Picard52, and GATK v4.1.4.153. The sex, ancestry and lity is also increasingly seen as the most visible symptom of a relatedness of each samples was calculated using peddy54, samples found to have more complex disease with associated comorbidities49. Studying the incorrect sex or were unrelated to the correct samples were excluded from this the long-term health of men with DNMs in specific genes should study. Following best practice recommendations, single nucleotide variations and small indels were identified and quality-filtered using GATK’s HaplotypeCaller help in identifying genotype–phenotype correlations that may obtaining comparable results independently of the kit or the origin of the DNA (see impact more than the fertility of these men. Supplementary Table 3). Afterwards, all variants were further analysed using a custom GATK4-based algorithm to identify and separate high- and low-confidence de novo variants from inherited variants. Briefly, posterior genotype probabilities Methods (GQ) were recalculated for each sample at each variant site using Bayes’ rule to take Cohort of infertile patients and fertile parent trios. We enrolled a total of 185 into account family and population priors53,55. Proband variants absent in parental patients who presented with unexplained (idiopathic) azoospermia (N = 111) or samples with recalculated proband GQ > = 10 and allele count (AC) below 4 or severe to extreme oligozoospermia (with or without asthenozoospermia N = 74) at allele frequency (AF) < 0.1% in all samples, whichever is more stringent, were the Radboudumc outpatient clinic between July 2007 and October 2017 (N = 170) classified as low-confidence DNMs. Variants with recalculated GQs ≥ 20 and the and at the Newcastle upon Tyne Hospitals NHS Foundation Trust (Newcastle, UK) same AC/AF criterion were classified as high confidence DNMs. Afterwards, tagged between January 2018 to January 2020 (n = 15). The reference values and semen variants with coverage <10, variant read percentage <15% and GATK quality scores nomenclature were used according to the WHO guidelines50 (see Supplementary <400 were removed to ensure only the most reliable variants were considered. Note). Clinical evaluation did not lead to an etiologic diagnosis and all patients Sanger sequencing was then used to validate DNMs calls. Ensembl’s Variant Effect were negative for AZF deletions and chromosomal anomalies (see Supplementary Predictor (VEP)56 was used to fully annotate all de novo variants. Notes). The study protocol was approved by the respective Ethics Committees/ Institutional Review Boards (Nijmegen: NL50495.091.14 version 5.0, Newcastle: REC ref. 18/NE/0089) and written informed consent from all patients and their Variant filtration and interpretation. The primary stages in filtering of variants parents was obtained prior to enrollment in the study. We used residual genomic included removing all variants with an allele frequency of >0.1% in the gnomAD DNA extracted from a blood sample taken at the time of evaluation and treatment database to only include rare variants in our analysis. All variants then with <10 at the fertility center. DNA from all proband’s parents was obtained from saliva by reads in the exome data and/or less than 15% of these reads containing the using the Oragene OG-500 kit (DNA Genotek, Ottawa, Canada). mutation were then removed. At this stage, any remaining variants lying outside the exonic regions were then removed. This provided the initial list of 192 rare de Immunofluorescence staining of human testis biopsies. Tissue sections were cut novo variants. All synonymous and non-protein-altering spice site variants were from formalin fixed paraffin embedded (FFPE) testicular biopsies. As staining then removed, leaving a total of 145-protein-altering rare DNMs. Pathogenicity controls testicular biopsies obtained from fertile men after a previous vasectomy prediction was then based on SIFT20, MutationTaster21, and PolyPhen222 and all was used. FFPE sections were prepared for staining following standard protocols. variants were classified according to the American College of Medical Genetics and To detect RBM5, antibody HPA018011 from Atlas Antibodies was used in a 1:500 Genomics (ACMG) and the Association for Molecular Pathology (AMP) 2015 dilution. For detection, a donkey anti-rabbit Alexa488 antibody was used (A-21206, guidelines57. All protein-altering variants predicted to be pathogenic by at least 2 1:1000 dilution), which was applied in combination with lectin coupled to out of 3 prediction models, absent from the fertile male cohort, present in <5 males Alexa568 (L32458, 1:1500 dilution) to detect the acrosome (both Thermo Scien- in the gnomAD database were considered for further functional analysis (n = 84). tific). Slides were counterstained with DAPI. Images were obtained with a Zeiss Maternally inherited mutations present in genes identified as having a protein- Axio Imager ZI fluorescence microscope equipped with the Zen software package. altering DNM were identified in all patients and submitted to the exact same method of filtration and interpretation as described above. Functional analysis was split into six different categories, each category Cohort of verified fertile Dutch parents. We used an anonymized exome dataset provided a score of either 1 or 0 depending on whether they met the threshold for derived from 5784 Dutch men and 5803 Dutch women who had conceived at least that category. These categories included: RNA expression of the gene in the testis, one child as a control cohort for the frequency of rare variants in fertile men and RNA enrichment in the testis or presence in spermatogenesis, protein expression in fertile women. These men and women received routine exome sequencing at the the testis, whether an infertile mouse model already exists for the given gene, the 6 NATURE COMMUNICATIONS | (2022)13:154 | https://doi.org/10.1038/s41467-021-27132-8 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-021-27132-8 ARTICLE protein function in relation to spermatogenesis and finally whether the given gene aberrations and microdeletions of the AZF region were excluded in advance. WES interacts with any known fertility genes. For expression levels retrieved for each was performed as previously described65. The 88 patients diagnosed with male gene of interest from the GTEx database (https://www.gtexportal.org/), an infertility participating in the Geisinger-Regeneron DiscovEHR collaboration were expression of medium (≥10 < 100 TPM) or high (>100 TPM) gave a score of 1 with selected from deidentified EHR information using the ICD-10CM code N46 which low (>2 < 10 TPM) and no expression (<2 TPM) giving a score of 0. RNA refers to “Male Infertility” including oligospermia, azoospermia, other male enrichment was based on elevated expression (tissue enriched, group enriched, or infertility and male infertility unspecified. All patients were sequenced at the tissue enhanced) in the Human Protein Atlas58 or being among the genes up- or Regeneron Genetics Center (RGC) as previously described39. In brief, 1ug of downregulated during spermatogenesis as found in a recent single cell RNA genomic DNA per sample was used for targeted exome capture using the Nim- sequencing study59. Protein expression was retrieved from the Human Protein bleGen VCRome 2.1 or the IDT XGen reagents. Captured libraries were sequenced Atlas58 and interaction with known infertility genes3 was calculated using STRING on the Illumina HiSeq 2500 platform with v4 chemistry using paired-end 75 bp version 1124. The final classification of the genes was then split into Not causative, reads. Exome sequencing was performed such that >85% of the bases were covered Unlikely causative, Unclear and Possibly causative. These classifications were given at 20× or greater. Raw sequence reads were mapped and aligned to the GRCh38/ based on the variant scores out of 6 with: [0 points + not expressed/not detected/ hg38 human genome reference assembly using BWA-mem51, single nucleotide and not present on several occasions = Unlikely causative], [0 points + “Unknown” on indel variants were called using GATK’s HaplotypeCaller53. The Genetics of Male several occasions = Unclear], [1–2 points = Unclear] and [3–6 points = possibly Infertility INitiative (GEMINI) is a multicenter study funded by the United States causative]. NIH. The GEMINI project performed whole-exome sequencing on 1,011 unrelated men diagnosed with spermatogenic failure, the vast majority with unexplained NOA. Sequencing of genomic DNA was performed at the McDonnell Genome CNV analysis. CNV calling was performed on our trio-based exome data with a Institute of Washington University in St. Louis, MO, USA, using an in-house custom GATK4-based pipeline. This workflow exploits the GATK4 sequence read exome targeting reagent capturing 39.1 Mb of exome and 2 × 150 bp paired-end depth normalization60 and a custom R based segmentation and visualization61. sequencing on Illumina HiSeq 4000. Following sample QC, a final cohort of 924 Parental samples from the trios under examination were used as controls for the men were analysed as part of the current study. normalization step. The CNVs detected were annotated using AnnotSV (https:// Genetic variants identified within the 152 candidate genes were extracted from lbgi.fr/AnnotSV/)62. CNVs present in more than 1% of the samples of the Database each exome dataset. Consistent with our filtering method described above variants of Genomic Variants present in more than 10% of the patients were excluded from with <10 reads and/or <15% reads containing the mutation were discarded. To the analysis. The remaining rare deletions and duplications were individually minimize discrepancies between genomic positions and annotation, genomic inspected through the genomic profiles and detailed Log2Ratio plots generated by coordinates were recalculated to the GCRh37/hg19 where necessary and fully the workflow. Only CNVs involving more than 2 exons were further considered to reannotated with VEP56. Following annotation variants from each of the additional minimize the inclusion of false positives, and we selected 2 CNVs present in the case and control cohorts were filtered and processed in an identical manner as probands but absent in their parents for further validation. previously described. Shortly, variants with allele frequency >1% in gnomAD were discarded to focus only on rare variants: Pathogenicity predictions based on SIFT20, Variant validation. Validation of low-quality DNMs was performed using stan- MutationTaster21 and PolyPhen22 were then used to exclude benign variants, all dard Sanger sequencing approach on an Applied Biosystems SeqStudio Genetic remaining variants were classified according to ACMG guidelines56. Like before, all Analyzer (ThermoFisher, MA, USA) to confirm the presence of the mutation in protein-altering variants were considered pathogenic if predicted to be so by at probands and its absence in the parents. Primers for each SNV were designed using least 2 out of 3 prediction models, absent from the fertile cohorts and present in <5 PrimerZ63 (Supplementary Data 8) and PCR reactions were performed using males in the gnomAD database. A similar analysis was done for all variants AmpliTaq 360 DNA Polymerase (ThermoFisher, MA, USA) according to the obtained in the control cohorts. However, just to be clear, for the male control manufacturer’s protocol. cohort we did not exclude variants as pathogenic if they were present in this cohort Validation of CNVs was performed with the whole genome Illumina Infinium itself. CytoSNP-850K v1.1 microarray platform for the larger deletion on chromosome 11 and a gene-specific TaqMan Copy-Number assay designed for NXT2 was exploited Burden testing. Having identified several likely pathogenic rare loss-of-function and to validate the smaller CNV using the Applied Biosystems QuantStudio 7 Flex missense mutations in these 152 genes we performed a gene-based burden test to Real-Time PCR System (ThermoFisher, MA, USA). compare the combined data in all cohorts of infertile men with the control cohort of fertile fathers. The proportion of individuals with pathogenic variants in each of the 152 Functional enrichment. To evaluate the intolerance of each gene for loss-of- genes was statistically evaluated using two-tailed Fisher’s Exact tests, individual p values function (LoF) mutations, we used the probability of LoF intolerance (pLI) score, were corrected using the Bonferroni method corrections to adjust for performing 152 based on data from the Genome Aggregation Database (gnomAD)8 containing consecutive statistical tests and reduced the risk of Type I errors. Similarly, a gene-based genetic data from 141,456 individuals. We computed the likelihood of the observed burden test was performed to compare fertile fathers with fertile mothers from the median pLI score of each gene (LoF in controls) set compared to the expected control cohort of verified fertile parents to investigate whether any of the sexes pre- median pLI based on the method described in Lelieveld et al.64. In short, we dominantly carried a greater number of rare pathogenic mutations. simulated the expected number of recurrently mutated genes by redistributing the observed number of mutations at random over a determined set of genes based on their specific LoF and functional mutation rates, however, in contrast to Lelieveld Phasing analysis to determine parent-of-origin. The origin of DNMs identified et al.64 and Samocha et al.23 before them instead of using the complete set of 18,226 in the exomes of patients was first investigated in the short-read exome data by pLI annotated genes to obtain expected median pLI scores, we used a set of 2766 performing phasing analysis on those variants that contained a parental infor- coding DNMs in 1941 control individuals, downloaded from the de novo-db mative SNP (iSNP) within 150 bp from the DNM. As a next step, all DNMs were version 1.6.1 (http://de novo-db.gs.washington.edu/de novo-db/)12, to correct for target-enriched with long-range PCR and sequenced using the Oxford Nanopore’s the gene-specific mutation rate. The empirical P value was calculated by comparing MinION sequencer (Oxford Nanopore technologies, Oxford Science Park, UK). the observed median pLI to the expected pLI following 100,000 random sampling Target regions were designed to encapsulate both the DNM and a parentally simulations. Case and fertile controls were processed using the exact same filtration informative SNP, from which parent-of-origin, and allele frequencies (percentage and annotation parameters as described above so that each variant detected was read counts associated to a given allele) could be ascertained and DNM pre-/post- evaluated in a comparable manner. The same method was then repeated using the zygosity could be determined. Loss-of-function observed/expected upper bound fraction or LOEUF score, which Primers were designed using Primer366 (version 2.3.6) and GRCh37.p5 based also is an indicator of LoF intolerance. To evaluate the impact of the de novo in-house GUI-wrapped pipeline. All expected fragment sizes were limited to a missense mutations to each gene, we used missense Z-scores calculated by maximum of 12 kb for quality control and enrichment success rate. For those gnomAD10,23 to predict the tolerance of each gene to variation in place of the pLI DNMs with no exome supported iSNPs within a 10 kb distance, primers were scores when applying the Lelieveld et al.64 methodology described above following designed to cover approximately 2.5 kb on either side of the DNM with the 100,000 simulations. The presence of missense mutations in intolerant genes was expectation of finding additional iSNPs in the intronic regions. Long-range PCR compared between predicted pathogenic and benign using a two-tailed target enrichment was carried out using our optimized running conditions of Mann–Whitney U test in our samples and in controls independently. To predict 3 separate supermixes/enzymes (see Supplementary Table 4). Sample fragment the affected protein function and the potential role in disease, we evaluated the sizes were confirmed using gel electrophoresis, and quantities were measured with interactions between the genes with a DNM using STRING version 1124. the Qubit dsDNA HS kit (ThermoFisher Scientific, Waltham, MA, USA), with the best quality supermix enrichment for each given sample/target selected for sequencing, where quality was assessed by cleanest banding in gel electrophoresis Additional cohorts of infertile men. The strongest candidate genes with DNMs and greatest concentration. were further investigated in exome data from four additional cohorts of infertile The long-range PCR target enrichments of >20 ng were prepared for sequencing men. For the Italian cohort of 48 patients with NOA, exome sequencing was with the ONT ligation sequencing kit (SQK-LSK109) following the manufacturer’s carried out as a service by Macrogen Inc. (Republic of Korea) utilizing the Agilent protocol, with adjustments for sample type and yield. Individual sample libraries SureSelect_V6 enrichment and a NovaSeq 6000. The German Male Reproductive were concentrated where necessary at given bead clean-up steps and pooled based Genomics (MERGE) study comprised exome data of 887 men with azoo-, crypto-, on fragment size. Fragment size-based pools were combined prior to flowcell or severe oligozoospermia. Known causes for male infertility like chromosomal loading. Prepared samples were sequenced on the MinION using the FLO-MIN106 NATURE COMMUNICATIONS | (2022)13:154 | https://doi.org/10.1038/s41467-021-27132-8 | www.nature.com/naturecommunications 7 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-021-27132-8 version 9.4 flowcell platform. Flowcells were run until complete pore exhaustion, 19. Grozdanov, P. N., Li, J., Yu, P., Yan, W. & MacDonald, C. C. Cstf2t regulates with minimal refuel of flowcells performed whenever active pore percentages expression of histones and histone-like proteins in male germ cells. Andrology dropped below 70%, achieving an average of 30 billion basecall yields per flowcell 6, 605–615 (2018). and coverage depth per sample of >5000×. 20. Vaser, R., Adusumalli, S., Leng, S. N., Sikic, M. & Ng, P. C. SIFT missense The sequence signal data in multi-fast5 format were basecalled using Guppy67 predictions for genomes. Nat. Protoc. 11, 1–9 (2016). (version 3.4.4, https://nanoporetech.com/), resulting fastq outputs were adapter 21. Schwarz, J. M., Rödelsperger, C., Schuelke, M. & Seelow, D. MutationTaster trimmed and low-quality reads discarded using cutadapt (version 2.5)68. Cleaned evaluates disease-causing potential of sequence alterations. Nat. Methods 7, fastq files were mapped against Human Reference Genome (GRCh37.p5/hg19) 575–576 (2010). using BWA-Mem51 (version 0.7.17), and sample targets were extracted from the 22. Adzhubei, I. A. et al. A method and server for predicting damaging missense resulting BAM file using SAMtools (version 0.1.19)69. Aligned ONT reads were mutations. Nat. Methods 7, 248–249 (2010). phased using an in-house tool, with frequencies and pre/post-zygosity calls 23. Samocha, K. E. et al. A framework for the interpretation of de novo mutation affirmed via IGV and principal component analysis using the available exome in human disease. Nat. Genet. 46, 944–950 (2014). sequence data for probands and parents to support the ONT data. 24. Szklarczyk, D. et al. The STRING database in 2017: quality-controlled protein–protein association networks, made broadly accessible. Nucleic Acids Reporting summary. Further information on research design is available in the Nature Res. 45, D362–D368 (2017). Research Reporting Summary linked to this article. 25. Song, H., Wang, L., Chen, D. & Li, F. The function of Pre-mRNA alternative splicing in mammal spermatogenesis. Int. J. Biol. Sci. 16, 38–48 (2020). 26. O’Bryan, M. K. et al. RBM5 is a male germ cell splicing factor and is required for Data availability spermatid differentiation and male fertility. PLoS Genet. 9, e1003628 (2013). Sequencing data has been deposited in the European Genome-phenome Archive(EGA) 27. Elsea, S. H. & Lucas, R. E. The mousetrap: what we can learn when the mouse under the accession code EGAS00001005417 and will be made available upon reasonable model does not mimic the human disease. ILAR J. 43, 66–79 (2002). request for academic use and within the limitations of the provided informed consent by 28. Jamsai, D. et al. In vivo evidence that RBM5 is a tumour suppressor in the the corresponding author upon acceptance. Every request will be reviewed by the lung. Sci. Rep. 7, 16323 (2017). Newcastle University Male Infertility Genomics Data Access Committee; the researcher 29. Liang, H. et al. Differential Expression of RBM5, EGFR and KRAS mRNA and will need to sign a data access agreement after approval. protein in non-small cell lung cancer tissues. J. Exp. Clin. Cancer Res. 31, 36 (2012). Code availability 30. Li, J. et al. HnRNPL as a key factor in spermatogenesis: Lesson from functional The code for the integrated pipeline used to process sequencing data to detect and call proteomic studies of azoospermia patients with sertoli cell only syndrome. J. rare germline copy-number variants (CNVs) is available at https://github.com/ Proteomics 75, 2879–2891 (2012). AnetaMikulasova/CNVRobot. 31. Wang, M. et al. Single-Cell RNA sequencing analysis reveals sequential cell fate transition during human spermatogenesis. Cell Stem Cell 23, 599–614.e4 (2018). Received: 15 March 2021; Accepted: 2 November 2021; 32. Ajuh, P. Functional analysis of the human CDC5L complex and identification of its components by mass spectrometry. EMBO J. 19, 6569–6581 (2000). 33. Brea-Fernández, A. J. et al. Expanding the clinical and molecular spectrum of the CWC27-related spliceosomopathy. J. Hum. Genet. 64, 1133–1136 (2019). 34. Millevoi, S. et al. An interaction between U2AF 65 and CF Im links the References splicing and 3′ end processing machineries. EMBO J. 25, 4854–4864 (2006). 1. Krausz, C. & Riera-Escamilla, A. Genetics of male infertility. Nat. Rev. Urol. 35. Baillet, A. et al. TOPAZ1, a novel germ cell-specific expressed gene conserved 15, 369–384 (2018). during evolution across vertebrates. PLoS ONE 6, e26950 (2011). 2. Tüttelmann, F., Ruckert, C. & Röpke, A. Disorders of spermatogenesis. Med. 36. Luangpraseuth-Prosper, A. et al. TOPAZ1, a germ cell specific factor, is Genet. 30, 12–20 (2018). essential for male meiotic progression. Dev. Biol. 406, 158–171 (2015). 3. Oud, M. S. et al. A systematic review and standardized clinical validity 37. Guo, F. et al. The transcriptome and DNA methylome landscapes of human assessment of male infertility genes. Hum. Reprod. 34, 932–941 (2019). primordial germ cells. Cell 161, 1437–1452 (2015). 4. Kasak, L. & Laan, M. Monogenic causes of non-obstructive azoospermia: 38. Li, L. et al. Single-cell RNA-seq analysis maps development of human challenges, established knowledge, limitations and perspectives. Hum. Genet. germline cells and gonadal niche interactions. Cell Stem Cell 20, 858–873.e4 140, 135–154 (2021). (2017). 5. Veltman, J. A. & Brunner, H. G. de novo mutations in human genetic disease. 39. Dewey, F. E. et al. Distribution and clinical impact of functional variants in Nat. Rev. Genet. 13, 565–575 (2012). 50,726 whole-exome sequences from the DiscovEHR study. Science. 354, 6. Sun, C. et al. An azoospermic man with a de novo point mutation in the aaf6814 (2016). Y-chromosomal gene USP9Y. Nat. Genet. 23, 429–432 (1999). 40. Francioli, L. C. et al. Genome-wide patterns and properties of de novo 7. Hodžić, A. et al. de novo mutations in idiopathic male infertility—a pilot mutations in humans. Nat. Genet. 47, 822–826 (2015). study. Andrology 9, 212–220 (2021). 41. Rahbari, R. et al. Timing, rates and spectra of human germline mutation. Nat. 8. De Jonge, C. & Barratt, C. L. R. The present crisis in male reproductive health: Genet. 48, 126–133 (2016). an urgent need for a political, social, and research roadmap. Andrology 7, 42. Goldmann, J. M. et al. Parent-of-origin-specific signatures of de novo 762–768 (2019). mutations. Nat. Genet. 48, 935–939 (2016). 9. Schilit, S. L. P. et al. SYCP2 translocation-mediated dysregulation and 43. Jónsson, H. et al. Parental influence on human germline de novo mutations in frameshift variants cause human male infertility. Am. J. Hum. Genet. 106, 1,548 trios from Iceland. Nature 549, 519–522 (2017). 41–57 (2020). 44. Braun, R. E., Behringer, R. R., Peschon, J. J., Brinster, R. L. & Palmiter, R. D. 10. Lek, M. et al. Analysis of protein-coding genetic variation in 60,706 humans. Genetically haploid spermatids are phenotypically diploid. Nature 337, Nature 536, 285–291 (2016). 373–376 (1989). 11. Karczewski, K. J. et al. The mutational constraint spectrum quantified from 45. Greenbaum, M. P., Iwamori, T., Buchold, G. M. & Matzuk, M. M. Germ cell variation in 141,456 humans. Nature 581, 434–443 (2020). intercellular bridges. Cold Spring Harb. Perspect. Biol. 3, a005850–a005850 (2011). 12. denovo-db, Seattle, WA. http://denovo-db.gs.washington.edu (2020). 46. Vissers, L. E. L. M. et al. A de novo paradigm for mental retardation. Nat. 13. Gu, Y. et al. Three intellectual disability-associated de novo mutations in Genet. 42, 1109–1112 (2010). MECP2 identified by trio-WES analysis. BMC Med. Genet. 21, 99 (2020). 47. Vissers, L. E. L. M., Gilissen, C. & Veltman, J. A. Genetic studies in intellectual 14. Fritzen, D. et al. de novo FBXO11 mutations are associated with intellectual disability and related disorders. Nat. Rev. Genet. 17, 9–18 (2016). disability and behavioural anomalies. Hum. Genet. 137, 401–411 (2018). 48. Kaplanis, J. et al. Evidence for 28 genetic disorders discovered by combining 15. De Tomasi, L. et al. Mutations in GREB1L cause bilateral kidney agenesis in healthcare and research data. Nature. 586, 757–762 (2020). humans and mice. Am. J. Hum. Genet. 101, 803–814 (2017). 49. Kasman, A. M., Del Giudice, F. & Eisenberg, M. L. New insights to guide 16. Brophy, P. D. et al. A gene implicated in activation of retinoic acid patient care: the bidirectional relationship between male infertility and male receptor targets is a novel renal agenesis gene in humans. Genetics 207, health. Fertil. Steril 113, 469–477 (2020). 215–228 (2017). 50. World Health Organization. WHO laboratory manual for the examination 17. Alhathal, N. et al. A genomics approach to male infertility. Genet. Med. 22, and processing of human semen. vol. 5th ed. (World Health Organization, 1967–1975 (2020). 2010). 18. Seabra, C. M. et al. A novel Alu-mediated microdeletion at 11p13 removes 51. Li, H. & Durbin, R. Fast and accurate long-read alignment with WT1 in a patient with cryptorchidism and azoospermia. Reprod. Biomed. Burrows–Wheeler transform. Bioinformatics 26, 589–595 (2010). Online 29, 388–391 (2014). 52. Broad Institute. Picard toolkit. https://broadinstitute.github.io/picard/ (2019). 8 NATURE COMMUNICATIONS | (2022)13:154 | https://doi.org/10.1038/s41467-021-27132-8 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-021-27132-8 ARTICLE 53. Auwera, G. A. et al. From FastQ data to high‐confidence variant calls: the (DFG, CRU326) to C.F. and F.T. This project was also supported in part by funding from genome analysis toolkit best practices pipeline. Curr. Protoc. Bioinform. 43, the Australian National Health and Medical Research Council (APP1120356) to 11.10.1–11.10.33 (2013). M.K.O.B., by grants from the National Institutes of Health of the United States of 54. Pedersen, B. S. & Quinlan, A. R. Who’s Who? Detecting and Resolving Sample America (R01HD078641 to D.F.C. and K.I.A., P50HD096723 to D.F.C.) and from the Anomalies in Human DNA Sequencing Studies with Peddy. Am J Hum. Biotechnology and Biological Sciences Research Council (BB/S008039/1) to D.J.E. Genet. 100, 406–413 (2017). 55. Poplin, R. et al. Scaling accurate genetic variant discovery to tens of thousands of samples. bioRxiv 201178. https://doi.org/10.1101/201178 (2018). Author contributions This study was designed by M.S.O., L.E.L.M.V., L.R. and J.A.V. R.M.S., J.G., H.T. and 56. McLaren, W. et al. The ensembl variant effect predictor. Genome Biol 17, 122 G.W.v.d.H. provided all clinical data and performed the TESE histology and cytology (2016). analysis under supervision of L.R., D.D.M.B., E.S., K.F., K.D.H. and K.M. J.C. performed 57. Richards, S. et al. Standards and guidelines for the interpretation of sequence the exome sequencing with support from B.A. and bioinformatics support was provided variants: a joint consensus recommendation of the American College of by M.J.X., G.A., C.G. and S.C. Sanger sequencing was performed by P.F.d.V., H.I., H.E.S., Medical Genetics and Genomics and the Association for Molecular Pathology. L.E.B. and B.K.S.A. M.S.O. and H.E.S. performed the SNV analyses with support from Genet. Med. 17, 405–423 (2015). M.J.X., F.K.M. performed CNV analysis with support from A.M. and M.S.K. and G.S.H. 58. Uhlén, M. et al. Tissue-based map of the human proteome. Science 347, and L.E.B. performed the phasing. D.J.E., H.S., B.J.H. and M.K.O.B. provided support on 1260419 (2015) the functional interpretation of mutations. D.F.C., L.N., C.F., S.K., F.T., K.I.A., A.R.E., 59. Guo, J. et al. The adult human testis transcriptional cell atlas. Cell Res. 28, C.K. and C.G.-J. were involved in the replication study. The first draft of the paper was 1141–1157 (2018). prepared by M.S.O., H.E.S., R.M.S., M.J.X., G.W.v.d.H. and J.A.V.. All authors con- 60. McKenna, A. et al. The Genome Analysis Toolkit: a MapReduce framework tributed to the final paper. for analyzing next-generation DNA sequencing data. Genome Res. 20, 1297–1303 (2010). 61. R Core Team. R: A language and environment for statistical computing. R Competing interests Foundation for Statistical Computing,Vienna, Austria. https://www.R- The authors declare no competing interests. project.org/ (2017). 62. Geoffroy, V. et al. AnnotSV: an integrated tool for structural variations annotation. Bioinformatics 34, 3572–3574 (2018). Additional information 63. Tsai, M.-F. et al. PrimerZ: streamlined primer design for promoters, exons and Supplementary information The online version contains supplementary material human SNPs. Nucleic Acids Res. 35, W63–W65 (2007). available at https://doi.org/10.1038/s41467-021-27132-8. 64. Lelieveld, S. H. et al. Meta-analysis of 2,104 trios provides support for 10 new genes for intellectual disability. Nat. Neurosci. 19, 1194–1196 (2016). Correspondence and requests for materials should be addressed to J. A. Veltman. 65. Wyrwoll, M. J. et al. Bi-allelic mutations in M1AP are a frequent cause of Peer review information Nature Communications thanks the anonymous reviewers for meiotic arrest and severely impaired spermatogenesis leading to male their contribution to the peer review of this work. Peer reviewer reports are available. infertility. Am. J. Hum. Genet. 107, 342–351 (2020). 66. Untergasser, A. et al. Primer3—new capabilities and interfaces. Nucleic Acids Reprints and permission information is available at http://www.nature.com/reprints Res. 40, e115 (2012). 67. Wick, R. R., Judd, L. M. & Holt, K. E. Performance of neural network basecalling Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in tools for Oxford Nanopore sequencing. Genome Biol. 20, 129 (2019). 68. Martin, M. Cutadapt removes adapter sequences from high-throughput published maps and institutional affiliations. sequencing reads. EMBnet.J. 17, 10–12 (2011). 69. Li, H. et al. The Sequence Alignment/Map format and SAMtools. Bioinformatics 25, 2078–2079 (2009). Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Acknowledgements Commons license, and indicate if changes were made. The images or other third party We are grateful for the participation of all patients and their parents in this study. We material in this article are included in the article’s Creative Commons license, unless thank Laurens van de Wiel (Radboudumc), Sebastian Judd-Mole (Monash University), indicated otherwise in a credit line to the material. If material is not included in the Arron Scott and Bryan Hepworth (Newcastle University) for technical support, and article’s Creative Commons license and your intended use is not permitted by statutory Margot J Wyrwoll (University of Münster) for help with handling MERGE samples and regulation or exceeds the permitted use, you will need to obtain permission directly from data. This project was funded by The Netherlands Organization for Scientific Research the copyright holder. To view a copy of this license, visit http://creativecommons.org/ (918-15-667) to J.A.V. as well as an Investigator Award in Science from the Wellcome licenses/by/4.0/. Trust (209451) to J.A.V. a grant from the Catherine van Tussenbroek Foundation to M.S.O. a grant from MERCK to R.S. a UUKi Rutherford Fund Fellowship awarded to B.J.H. and the German Research Foundation Clinical Research Unit “Male Germ Cells” © The Author(s) 2022 1 Department of Human Genetics, Donders Institute for Brain, Cognition and Behaviour, Radboudumc, Nijmegen, The Netherlands. 2Department of Obstetrics and Gynaecology, Radboudumc, Nijmegen, The Netherlands. 3Biosciences Institute, Faculty of Medical Sciences, Newcastle University, Newcastle upon Tyne, UK. 4School of BioSciences, Faculty of Science, The University of Melbourne, Parkville, VIC, Australia. 5Department of Genetic Medicine, The Newcastle upon Tyne Hospitals NHS Foundation Trust, Newcastle upon Tyne, UK. 6Foundation for Research in Genetics and Endocrinology, Institute of Human Genetics, Ahmedabad, India. 7Department of Human Genetics, Radboud Institute for Molecular Life Sciences, Radboudumc, Nijmegen, The Netherlands. 8Division of Human Genetics, Center for Biomedical Research, Faculty of Medicine, Diponegoro University, Semarang, Indonesia. 9Newcastle Fertility Centre, The Newcastle upon Tyne Hospitals NHS Foundation Trust, Newcastle upon Tyne, UK. 10Department of Cellular Pathology, The Newcastle upon Tyne Hospitals NHS Foundation Trust, Newcastle upon Tyne, UK. 11 Genomics Core Facility, Faculty of Medical Sciences, Newcastle University, Newcastle upon Tyne, UK. 12Bioinformatics Support Unit, Faculty of Medical Sciences New, castle University, Newcastle upon Tyne, UK. 13Department of Urology, Radboudumc, Nijmegen, The Netherlands. 14 Department of Pathology, Radboudumc, Nijmegen, The Netherlands. 15Division of Genetics, Oregon National Primate Research Center, Oregon Health & Science University, Beaverton, OR, USA. 16Institute of Reproductive Genetics, University of Münster, Münster, Germany. 17Centre of Reproductive Medicine and Andrology, Department of Clinical and Surgical Andrology, University Hospital Münster, Münster, Germany. 18 Department of Surgery, Division of Urology, University of Utah School of Medicine, Salt Lake City, UT, USA. 19Andrology Department, Fundació Puigvert, Universitat Autònoma de Barcelona, Instituto de Investigaciones Biomédicas Sant Pau (IIB-Sant Pau), Barcelona, Catalonia, Spain. 20 Department of Biomedical, Experimental and Clinical Sciences “Mario Serio”, University of Florence, Florence, Italy. 21Regeneron Genetics NATURE COMMUNICATIONS | (2022)13:154 | https://doi.org/10.1038/s41467-021-27132-8 | www.nature.com/naturecommunications 9 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-021-27132-8 Center, Tarrytown, NY, USA. 42These authors contributed equally: M. S. Oud, R. M. Smits, H. E. Smith. 43These authors jointly supervised this work: M. J. Xavier, G. W. van der Heijden, J. A. Veltman. *A list of authors and their affiliations appears at the end of the paper. ✉email: joris.veltman@newcastle.ac.uk Genetics of Male Infertility Initiative (GEMINI) consortium Donald F. Conrad22, Liina Nagirnaja22, Kenneth I. Aston23, Douglas T. Carrell23, James M. Hotaling23, Timothy G. Jenkins23, Rob McLachlan24,25, Moira K. O’Bryan4, Peter N. Schlegel26, Michael L. Eisenberg27, Jay I. Sandlow28, Emily S. Jungheim29, Kenan R. Omurtag29, Alexandra M. Lopes30,31, Susana Seixas30,31, Filipa Carvalho30,32, Susana Fernandes30,32, Alberto Barros30,32, João Gonçalves33,34, Iris Caetano33, Graça Pinto35, Sónia Correia35, Maris Laan36, Margus Punab37, Ewa Rajpert-De Meyts38, Niels Jørgensen38, Kristian Almstrup38, Csilla G. Krausz39,40 & Keith A. Jarvi41 22 Department of Genetics, Oregon National Primate Research Center, Oregon Health & Science University, Beaverton, OR, USA. 23Andrology and IVF Laboratory, Department of Surgery (Urology), University of Utah School of Medicine, Salt Lake City, UT, USA. 24Hudson Institute of Medical Research and the Department of Obstetrics and Gynaecology, Monash University, Clayton, VIC, Australia. 25Monash IVF and the Hudson Institute of Medical Research, Clayton, VIC, Australia. 26Department of Urology, Weill Cornell Medicine, New York, NY, USA. 27Department of Urology, Stanford University School of Medicine, Stanford, CA, USA. 28Department of Urology, Medical College of Wisconsin, Milwaukee, WI, USA. 29 Washington University in St Louis, School of Medicine, St Louis, MO, USA. 30i3S—Instituto de Investigação e Inovação em Saúde, Universidade do University of Porto, Porto, Portugal. 31IPATIMUP—Instituto de Patologia e Imunologia Molecular da Universidade do Porto, Porto, Portugal. 32 Serviço de Genética, Departamento de Patologia, Faculdade de Medicina da Universidade do Porto, Porto, Portugal. 33Departamento de Genética Humana, Instituto Nacional de Saúde Dr Ricardo Jorge, Lisboa, Portugal. 34ToxOmics, Faculdade de Ciências Médicas, Universidade Nova de Lisboa, Lisboa, Portugal. 35Centro de Medicina Reprodutiva, Maternidade Dr. Alfredo da Costa, Lisboa, Portugal. 36Institute of Biomedicine and Translational Medicine, University of Tartu, Tartu, Estonia. 37Andrology Center, Tartu University Hospital, Tartu, Estonia. 38Department of Growth and Reproduction, Rigshospitalet, University of Copenhagen, Copenhagen, Denmark. 39Department of Experimental and Clinical Biomedical Sciences, University of Florence, Florence, Italy. 40Andrology Department, Fundacio Puigvert, Instituto de Investigaciones Biomédicas Sant Pau (IIB- Sant Pau), Barcelona, Spain. 41Division of Urology, Department of Surgery, Mount Sinai Hospital, University of Toronto, Toronto, ON, Canada. 10 NATURE COMMUNICATIONS | (2022)13:154 | https://doi.org/10.1038/s41467-021-27132-8 | www.nature.com/naturecommunications