Biallelic mutations in RNA-binding protein ADAD2 cause spermiogenic failure and non-obstructive azoospermia in humans

Abstract STUDY QUESTION What are some pathogenic mutations for non-obstructive azoospermia (NOA) and their effects on spermatogenesis? SUMMARY ANSWER Biallelic missense and frameshift mutations in ADAD2 disrupt the differentiation of round spermatids to spermatozoa causing azoospermia in humans and mice. WHAT IS KNOWN ALREADY NOA is the most severe cause of male infertility characterized by an absence of sperm in the ejaculate due to impairment of spermatogenesis. In mice, the lack of the RNA-binding protein ADAD2 leads to a complete absence of sperm in epididymides due to failure of spemiogenesis, but the spermatogenic effects of ADAD2 mutations in human NOA-associated infertility require functional verification. STUDY DESIGN, SIZE, DURATION Six infertile male patients from three unrelated families were diagnosed with NOA at local hospitals in Pakistan based on infertility history, sex hormone levels, two semen analyses and scrotal ultrasound. Testicular biopsies were performed in two of the six patients. Adad2 mutant mice (Adad2Mut/Mut) carrying mutations similar to those found in NOA patients were generated using the CRISPR/Cas9 genome editing tool. Reproductive phenotypes of Adad2Mut/Mut mice were verified at 2 months of age. Round spermatids from the littermates of wild-type (WT) and Adad2Mut/Mut mice were randomly selected and injected into stimulated WT oocytes. This round spermatid injection (ROSI) procedure was conducted with three biological replicates and >400 ROSI-derived zygotes were evaluated. The fertility of the ROSI-derived progeny was evaluated for three months in four Adad2WT/Mut male mice and six Adad2WT/Mut female mice. A total of 120 Adad2Mut/Mut, Adad2WT/Mut, and WT mice were used in this study. The entire study was conducted over 3 years. PARTICIPANTS/MATERIALS, SETTING, METHODS Whole-exome sequencing was performed to detect potentially pathogenic mutations in the six NOA-affected patients. The pathogenicity of the identified ADAD2 mutations was assessed and validated in human testicular tissues and in mouse models recapitulating the mutations in the NOA patients using quantitative PCR, western blotting, hematoxylin-eosin staining, Periodic acid-Schiff staining, and immunofluorescence. Round spermatids of WT and Adad2Mut/Mut mice were collected by fluorescence-activated cell sorting and injected into stimulated WT oocytes. The development of ROSI-derived offspring was evaluated in the embryonic and postnatal stages. MAIN RESULTS AND THE ROLE OF CHANCE Three recessive mutations were identified in ADAD2 (MT1: c.G829T, p.G277C; MT2: c.G1192A, p.D398N; MT3: c.917_918del, p.Q306Rfs*43) in patients from three unrelated Pakistani families. MT1 and MT2 dramatically reduced the testicular expression of ADAD2, likely causing spermiogenesis failure in the NOA patients. Immunofluorescence analysis of the Adad2Mut/Mut male mice with the corresponding MT3 mutation showed instability and premature degradation of the ADAD2 protein, resulting in the spermiogenesis deficiency phenotype. Through ROSI, the Adad2Mut/Mut mice could produce pups with comparable embryonic development (46.7% in Adad2Mut/Mut versus 50% in WT) and birth rates (21.45 ± 10.43% in Adad2Mut/Mut versus 27.5 ± 3.536% in WT, P = 0.5044) to WT mice. The Adad2WT/Mut progeny from ROSI (17 pups in total via three ROSI replicates) did not show overt developmental defects and had normal fertility. LARGE SCALE DATA N/A. LIMITATIONS, REASONS FOR CAUTION This is a preliminary report suggesting that ROSI can be an effective treatment for infertile Adad2Mut/Mut mice. Further assisted reproductive attempts need to be carefully examined in humans during clinical trials. WIDER IMPLICATIONS OF THE FINDINGS Our work provides functional evidence that mutations in the ADAD2 gene are deleterious and cause consistent spermiogenic defects in both humans and mice. In addition, preliminary results show that ROSI can help Adad2Mut/Mut to produce biological progeny. These findings provide valuable clues for genetic counselling on the ADAD2 mutants-associated infertility in human males. STUDY FUNDING/COMPETING INTEREST(S) This work was supported by the National Natural Science Foundation of China (32000587, U21A20204, and 32061143006), and the National Key Research and Developmental Program of China (2019YFA0802600 and 2021YFC2700202). This work was also supported by Institute of Health and Medicine, Hefei Comprehensive National Science Center, Hefei, China. The authors declare no competing interests.


Introduction
Spermatogenesis is a highly coordinated process that involves spermatogonial proliferation and differentiation, meiotic division of spermatocytes and post-meiotic differentiation from round spermatids to spermatozoa (also termed spermiogenesis) (de Kretser et al., 1998). Impairment of any step in spermatogenesis can cause non-obstructive azoospermia (NOA), which accounts for 20% of infertility in men (Jiang et al., 2022). Although genetic anomalies have been identified in about 25% of NOA cases (Krausz and Riera-Escamilla, 2018), mutations in only 14 genes have been verified to cause NOA (Sudhakar et al., 2021;Jiang et al., 2022).
Round spermatid injection (ROSI), an ART, involves the injection of round spermatids (derived from testicular biopsies) into the recipient's oocytes (Tesarik et al., 1995). In NOA patients with round spermatids as the most mature germ cells in the testes, ROSI is considered the last resort for the production of biological offspring (Tesarik et al., 1995;Antinori et al., 1997;Barak et al., 1998;Gianaroli et al., 1999;Tanaka et al., 2015Tanaka et al., , 2018. ADAD2 is a double-stranded RNA-binding protein that is expressed exclusively in the testis (Wang et al., 2015;Snyder et al., 2020). Male mice lacking ADAD2 (herein referred to as Adad2 ko ) are infertile with a complete absence of sperm in the epididymides due to defective spermiogenesis (Snyder et al., 2020;Chukrallah et al., 2022). Previous studies have suggested that a homozygous stop-gain mutation in ADAD2 (c.1186C>T, p.Gln396Ter) and compound heterozygous mutations (Hg19: chr16:84012049-84224913del and c.82dupC, p.Gln28ProfsTer136) in ADAD2 might be associated with incomplete spermatogonial arrest (Krausz et al., 2020). Patients with severe asthenoteratozoospermia carrying a homozygous frameshift mutation (c.17_18del, p.Gln6Argfs*3) or a homozygous missense mutation (c. 1381C>T, p.Arg461Trp) in ADAD2 showed morphological deformities in both the sperm head and flagellum assembly (Dai et al., 2023;Tian et al., 2023), suggesting that ADAD2 may affect human spermatogenesis. However, the reproductive phenotype of these ADAD2 mutations identified in humans differ from those in Adad2 ko mice, and the pathological effects of these ADAD2 mutations have not been verified in mouse models. Therefore, the functional role of ADAD2 in human spermatogenesis and testicular spermatogenic defects due to ADAD2 mutations in infertile men require further exploration.
In this study, we identified three ADAD2 mutations (MT1, MT2, and MT3) in six NOA patients from Pakistan. Patients harboring biallelic MT1 or MT2 mutation had significantly reduced levels of testicular ADAD2 protein and defects in spermiogenesis, which is consistent with the observation in Adad2 ko mice. Male Adad2 Mut/Mut mice corresponding to MT3 in men had similar infertile phenotypes. Importantly, although the round spermatids of Adad2 Mut/Mut displayed aberrant chromatin organization, ROSI helped them produce fertile offspring. Thus, our study provides direct in vivo and ex vivo evidence that biallelic mutations in ADAD2 cause NOA in humans, and ROSI is a feasible treatment for the Adad2 Mut/Mut mice which have mutational efficiency similar to that of ADAD2-mutated NOA patients.

Clinical samples
In this study, we recruited six infertile men from three unrelated families who were diagnosed with idiopathic NOA. Four patients, including two brothers in Family 1 and two brothers in Family 3, were born to consanguineous parents. All the patients had normal height and secondary sexual characteristics but failed to produce offspring even after trying to conceive during >6 years of marriage (Table 1). All the patients had normal karyotypes (46, XY) and no Y-chromosome microdeletions (Table 1). Serum levels of reproductive hormones including FSH, luteinizing hormone (LH), testosterone, and estradiol were in the normal range, as measured by local laboratories (Table 1). Testicular shape and outline were normal, while a reduction in the bilateral testicular volume was observed in five patients (Table 1). Semen analysis was performed twice with the 12-week interval between the analyses for each patient, in accordance with the WHO guidelines

WHAT DOES THIS MEAN FOR PATIENTS?
Non-obstructive azoospermia (NOA) is characterized by impaired spermatogenesis and is the most severe form of male infertility. Approximately 25% of NOA cases are attributed to genetic anomalies, but only mutations in a small number of genes have been validated as pathogenic in NOA patients.
Our study focused on six infertile patients with NOA who were carriers of three ADAD2 mutations from different and unrelated families. Functional evidence has shown that these mutations cause premature degradation of the ADAD2 protein and are associated with failure of round spermatids to differentiate into spermatozoa during spermatogenesis.
Despite the lack of spermatozoa in the testes, Adad2 Mut/Mut mice with mutation efficiencies similar to that of human patients were able to produce healthy and fertile offspring after round spermatid injection. Therefore, our study provides a preliminary insight for the genetic counselling of couples with ADAD2-mutation-associated male infertility.
(World Health Organization, 2021); all six men ejaculated the normal volume of semen, which contained no sperm (Table 1).
The controls in this study were men who had been diagnosed with obstructive azoospermia (OA) with no obvious defects in spermatogenesis. Written informed consent was received from all participants prior to the onset of the study. The study was approved by the institutional ethics committee of the University of Science and Technology of China (USTC) with the approval number 2019-KY-168.

Whole-exome sequencing and mutation selection
Whole-exome sequencing (WES) and mutation selection were performed as reported previously (Zhang et al., 2020;Fan et al., 2021). Briefly, total genomic DNA was isolated from the peripheral blood samples. Whole-exome capture and sequencing were performed using the AlExome Enrichment Kit V1 (iGeneTech, Beijing, China) and Hiseq2000 platform (Illumina, San Diego, CA, USA), following standard procedures. The reads were aligned to the human genome reference assembly (hg19) using Burrows-Wheeler Aligner with default parameters. PCR duplicates were removed by the Picard software (http://picard.sourceforge.net/). DNA mutation sequences were analyzed using Genome Analysis Toolkit HaplotypeCaller (http://www.broadinstitute.org/gatk/). Mutations that met the following filtration criteria were subjected to further analyses to consider the following: (i) mutations that could alter protein sequence; (ii) mutations with minor allele frequency <0.01 in the 1000 Genomes, ESP6500, ExAC, and Genome Aggregation Database, and were absent as homozygous or compound heterozygous in our in-house WES data sets from 578 fertile men; (iii) nonsense, frameshift and splice mutations, and missense deleterious mutations as predicted by at least half of the used software: Sorting Intolerant From Tolerant (SIFT), PolyPhen2 HDIV, MutationTaster, MutationAssessor, fathmm_MKL, GERPþþ, and SiPhy; (iv) mutations in genes that are expressed in the testis; (v) mutations following recessive inheritance patterns including autosome recessive, compound heterozygous and sex-linked recessive patterns; and (v) mutations in the spermatogenesis-related genes predicted by the SpermatogenesisOnline database (Zhang et al., 2013) or verified in an animal model. For the consanguineous families (Families 1 and 3), mutations associated with recessive inheritance patterns and located in Regions of Homozygosity (RoHs) were prioritized, while compound heterozygous mutations were preferred from the non-consanguineous family (Family 2). The homozygosity mapping analysis was performed using Bcftools (Narasimhan et al., 2016

Histological analysis
Human and mouse testicular tissues were fixed overnight in Bouin's solution, rinsed with 50% ethanol for 5 min, and then dehydrated in an ethanol gradient (70%, 80%, 90%, and 100%, 20 min for each). The tissues were then transferred into xylene twice for 15 min each time and were immersed three times in Generation of Adad2 Mut/Mut mice Adad2 Mut/Mut mice carrying the mutation analogous to that identified in the NOA patients (ADAD2 c.917_918del) were generated using CRISPR/Cas9 genome editing tools. A single guide RNA (sgRNA) targeting the mouse Adad2 genomic sequence close to the corresponding mutation site was designed using the webserver: http://crispor.tefor.net/. The sgRNA and Cas9 protein were transferred into C57BL/6 zygotes by electroporation. The founder mice mutations were confirmed by Sanger genomic DNA sequencing. The founder heterozygous Adad2 mutant mice were bred with 8-week-old C57BL/6 WT mice (GemPharmatech, Nanjing, China) to produce heterozygous F1 mice. Homozygotes were obtained by the inter-crossing of heterozygous mice from the third backcross. The primers used for the generation and genotyping of Adad2-mutant mice are listed in Supplementary Table  S1. All animals were housed in a specific-pathogen-free animal facility with a 12h:12h light:dark cycle. All animal studies were conducted following the guidelines of the Institutional Animal Care Committee of USTC (approval number USTCACUC1301021) and the Institutional Animal Care and Use Committees of Nanjing Medical University (IACUC-2009002).

Epidydimal sperm count
The unilateral epididymides of WT or Adad2 Mut/Mut mice were removed and cut into small pieces that were transferred to an Eppendorf tube containing 1 ml of 1Â PBS (68.5 mM NaCl, 1.3 mM KCl, 5.0 mM Na 2 HPO 4, and 0.9 mM KH 2 PO 4 ). After incubation for 30 min at 37 C, sperm were released into the PBS. Then, 10 ml of the PBS sperm suspension (diluted 10 times in 1Â PBS) were placed on a hemocytometer and the sperm samples were counted under the microscope. All the mouse samples had three replicates which were averaged to obtain the final value.

RNA extraction and qPCR
Total RNA was extracted from the testes as we described previously (Zhang et al., 2020;Gong et al., 2022). Briefly, total RNA was extracted with TRIzol (Accurate Biology AG21101, Hunan, China) and the cDNAs were synthesized with the PrimeScript RT reagent kit with gDNA Eraser (TaKaRa RR047A, Kusatsu, Japan). The concentration and purity of cDNA were measured using a NanoDrop 1000 Spectrophotometer (Thermo Fisher Scientific). The qPCR was performed with FastStart Universal SYBR Green Master (ROX) (Roche 04913850001, Basel, Switzerland) on a Step One Real-Time PCR System (Applied Biosystems, Thermo Fisher Scientific). The qPCR reactions were performed under the following conditions: 10 min at 95 C, 40 cycles of 10 s at 95 C and 30 s at 60 C. The qPCR data were analyzed by the 2 ÀDDCt method (Schmittgen and Livak, 2008); Actb (NM_007393.5) was the internal control. The primer sequences used are listed in Supplementary Table S1.

Immunofluorescent staining of paraffin-embedded sections
Mouse or human testes were fixed in 4% paraformaldehyde overnight, then embedded in paraffin and cut into 5-lm sections. For immunofluorescence-staining, slides were dewaxed, rehydrated, permeabilized, and then transferred to citrate-based antigen retrieval solution (0.3% trisodium citrate dihydrate and 0.04% citric acid monohydrate in ddH 2 O). Afterward, the slides were heated at 96 C for 20 min and blocked with BDT solution (3% bovine serum albumin and 10% normal donkey serum in 1Â TBST composed of 50 mM Tris (pH 7.4), 150 mM NaCl, and 0.1% Tween-20). The tissue slides were then incubated with primary antibodies, followed by secondary antibodies, and mounted with VECTASHIELD Antifade Mounting Medium (Vector Laboratories H-1000, San Francisco, CA, USA) containing 5 lg/ml Hoechst 33342 (Invitrogen H1399, Thermo Fisher Scientific). Images were captured using an Olympus BX53 Microscope (Tokyo, Japan) with cellSens imaging software. The Image-Pro Plus software (MEDIA CYBERNETICS, USA) was used for HP1a-positive foci counting. The image acquisition parameters are listed in Supplementary  Table S2. The antibodies used are listed in Supplementary Table  S3.

Western blotting
Human testicular proteins were extracted from tissue lysates prepared using TRIzol (Accurate Biology AG21101, Hunan, China) as we reported previously . The 1 ml of TRIzol was added to human testicular tissue, and the sample homogenization was performed on ice. The lysate was then added to 200 ml of chloroform and mixed thoroughly before centrifugation at 12 000Âg for 15 min at 4 C. After centrifugation, the upper aqueous phase was removed and 300 ll of 100% ethanol was added to the interphase and organic phase, before thorough mixing and centrifugation at 2000Â g for 5 min at 4 C. The supernatant was precipitated with 1.5 ml of isopropanol by centrifugation at 12 000Â g for 10 min at 4 C. The new supernatant was discarded, and 2 ml of cleaner solution (0.3 M guanidine hydrochloride in 95% ethanol) was added to the precipitate. After incubation for 20 min, the mixture was centrifuged at 7500Â g for 5 min at 4 C. The precipitate was then thoroughly mixed with 2 ml of 100% ethanol and again centrifuged at 7500Â g for 5 min at 4 C. This final precipitate contained the human testicular proteins, which were dissolved in 200 ll of 1% SDS and heated at 50 C for 30 min and denatured at 100 C for 10 min for subsequent western blotting.
Protein extracts from mice testes were prepared in lysis buffer containing 50 mM Tris (pH 7.5), 150 mM NaCl, 0.5% Triton X-100, 2.5 mM EDTA, and 1Â phenylmethylsulfonyl fluoride protease inhibitor (Thermo Scientific 36978, Waltham, MA, USA). The extracts were sonicated and then centrifuged. After heat denaturation, the proteins were separated by SDS-PAGE and then transferred onto 0.45-lm nitrocellulose membranes (GE Healthcare Life Sciences 10600002, Pittsburgh, PA, USA). The membranes were blocked in 5% nonfat milk and incubated with primary antibodies followed by incubation with secondary antibodies. Finally, the blots were developed for chemiluminescence (GE Healthcare Life Sciences ImageQuant LAS 4000). The antibodies used are listed in Supplementary Table S3.
The blots were quantified using Image J software (National Institutes of Health).

Flow cytometry analysis for haploid cells isolation
Testes from C57BL/6J WT (littermates of Adad2 Mut/Mut mice) and Adad2 Mut/Mut male mice (8-10 weeks old) were isolated. After removal of the tunica, the seminiferous tubules were cut into small pieces, placed in Dulbecco's Modified Eagle Medium (DMEM) (Gibco 11965092, Thermo Fisher Scientific), and digested with 1 mg/ml collagenase (Sigma-Aldrich C4-BIOC) for 15 min, followed by 0.25% trypsin-EDTA (Gibco 25200072) digestion for 10 min. The germ cell suspension was incubated with Hoechst 33342 (Invitrogen, 62249) at a concentration of 5 lg/ml for 15 min at 37 C and sorted on the BD FACSAria Fusion SOP system (BD Biosciences, USA) equipped with BD FACSDiva software (BD Biosciences). Two-way sorting was performed using a 100 lm nozzle size. The flow sorting rate was 3000 events/s. Hoechst dye was excited using a 355-nm laser, and the dye's wide emission spectrum was detected in Hoechst Blue (450/50 nm band-pass filter). Forward Scatter (FSC-A) and Side Scatter (SSC-A) were detected using a 488-nm laser. Haploid cells (1N) were collected based on the peaks of the cell population at 1N DNA content (Bastos et al., 2005;Barroca et al., 2009;Gaysinskaya et al., 2014). The sorted population of haploid cells was collected in 4 ml of DMEM in 15 ml tubes and centrifuged at 200Â g for 10 min. Then haploid round spermatid cells (1N) were selected from the obtained cell pellet under a microscope for subsequent oocyte injection.

Round spermatid injection and embryo transfer
The procedure of oocyte stimulation has been described elsewhere (Li et al., 2012). Before microinjection, the MII oocytes were pre-stimulated in Ca 2þ -free CZB medium (EasyCheck M0000, Nanjing, China) containing 10 mM SrCl 2 (Sigma-Aldrich 439665) for 10 min, followed by a 5 min exposure to M2 medium containing 5 mg/ml cytochalasin B (Sigma-Aldrich C6762). An individual round spermatid cell was injected into a pre-stimulated oocyte with a Piezo-driven pipette. After injection, embryos were activated in Ca 2þ -free CZB medium containing 10 mM SrCl 2 at 37 C under 5% CO 2 for 3-5 h. Finally, the injected embryos were transferred into fresh KSOM medium (Sigma-Aldrich MR-101-D) for in vitro culture. Different stages of embryos were collected at 40 (two-cell stage), 54 (four-cell stage), 84 (morula), and 96 (blastocyst) hours post-HCG injection for counting and imaging. Two-cell stage embryos were collected and transferred into the oviduct of CD1 pseudopregnant females. Full-term pups were delivered naturally.

Statistical analysis
GraphPad Prism software (GraphPad Software, San Diego, CA, USA) was used to perform the statistical analyses. Data are presented as mean § SD in Figs 3C, E and 4B, D, F-G, and Supplementary Figs 3B-D and 4A, B. Student's t-test was used for all statistical analyses. The data with P-values <0.05 were considered significant.

Results
Identification of ADAD2 mutations in three unrelated Pakistani families WES was performed to identify potential candidate genes associated with NOA in the three unrelated Pakistani families, including all six infertile patients, their siblings and parent(s). The obtained mutations were subsequently screened by a series of criteria ( Supplementary Fig. S1 and Table S4). Consequently, recessive mutations in ADAD2 were identified in the three Pakistani families (Fig. 1A). In Family 1, Patients IV-1 and IV-6 harbored a homozygous missense mutation in ADAD2 (MT1: c.G829T, p.Gly277Cys). In Family 2, Patients III-3 and III-4 harbored ADAD2 compound heterozygous missense mutations (MT1: c.G829T, p.Gly277Cys and MT2: c.G1192A, p.Asp398Asn). In Family 3, a homozygous 2-bp-deletion in ADAD2 (MT3: c.917_918del, p.Gln306Argfs*43) was identified in Patients IV-1 and IV-2. The allele frequencies of all these three mutations are below 0.01% in human populations ( Table 2). The MT1 and MT2 mutations were predicted to be deleterious by SIFT, PolyPhen2, MutationTaster, and fathmm-MKL tools ( Table 2). The identified ADAD2 mutations were further verified by Sanger sequencing and found to be co-segregated with male infertility in the respective families (Fig. 1A).
ADAD2 (also known as TENRL, GenBank: NM_001145400.2) is an RNA-binding protein that is specifically expressed in the testes of humans and mice. The three ADAD2 mutations were located in its adenosine deaminase domain, which is highly conserved across species (Fig. 1B). Furthermore, both MT1 and MT2 were predicted to alter the conformation of the adenosine deaminase domain (Fig. 1C). The MT3 mutation introduced a premature stop codon and was predicted to cause protein truncation or nonsense mediated mRNA decay. ADAD2 is reported to be essential for mouse spermatogenesis as Adad2 ko male mice are infertile due to spermiogenesis failure (Snyder et al., 2020). Thus, we speculated that these identified mutations in ADAD2 are likely the cause of male infertility in the three families.
ADAD2 missense mutations dramatically reduced ADAD2 protein levels in families 1 and 2 To investigate changes in the expression level of ADAD2 protein due to ADAD2 mutation, we performed western blotting, and a protein band of the expected size was detected in testicular samples from a control OA man who had normal spermatogenesis. Patient IV-1 from Family 1 (homozygous for MT1) and Patient III-3 from Family 2 (with compound heterozygous MT1 and MT2) had significantly lower levels of ADAD2 protein, seen as only trace amounts of the protein in these patients ( Fig. 2A). Furthermore, immunofluorescence staining was barely able to detect ADAD2 in IV-1 and III-3 testicular samples while the protein was present in the spermatocyte cytoplasm from the OA control (Fig. 2B). Thus, the ADAD2 MT1 and MT2 mutations largely reduced the level of ADAD2 protein in the testes of the affected patients.

Defects in spermiogenesis due to ADAD2 mutations
To further assess the impact of the ADAD2 mutations on spermatogenesis, testicular biopsies of the IV-I (Family 1) and III-3 (Family 2) patients were assessed by H&E staining. Histological analysis revealed the presence of spermatogonia, spermatocytes, and round spermatids in the seminiferous tubules of IV-I and III-3, which were comparable with that of the control OA patient (Fig. 2C). However, few spermatozoa could be seen in the patients' samples (Fig. 2C), suggesting defects in spermiogenesis in both patients. Considering that the spermiogenic defects observed in the IV-I and III-3 patients were highly similar to those in Adad2 KO mice (Snyder et al., 2020), we considered that the MT1 and MT2 missense mutations were deleterious and were likely the genetic cause of NOA in Families 1 and 2.
Adad2 Mut/Mut male mice are infertile and mimic the spermiogenic defects of ADAD2 mutationcarrying men Considering that the ADAD2 protein level was observed to be markedly reduced in the testes of the patients carrying MT1 and MT2 ( Fig. 2A), and pathological histology showed defective spermiogenesis (Fig. 2C) consistent with the phenotype of Adad2 KO mice (Snyder et al., 2020), we hypothesized that MT1 and MT2 were associated with male infertility. As the two patients carrying MT3 (c.917_918del, p.Gln306Argfs*43) in Family 3 did not agree to the testicular biopsy, the pathogenicity of the MT3 mutation could not be verified in patient samples, and we therefore generated a mouse model (Adad2 Mut/ Mut ) carrying a frameshift mutation (c.851insA, p.Pro287Serfs*41) that was corresponded to the human MT3 mutation (Supplementary Fig.  S2A and C). The predicted truncated ADAD2 protein in the patients carrying the MT3 mutation and Adad2 Mut/Mut mouse contains both the whole double-stranded RNA-binding (DBRM) domain and the shortened adenosine deaminase domain containing similar numbers of residues (Supplementary Fig. S2B).
Similar to the ADAD2-mutation-carrying patients, the Adad2 Mut/Mut male mice were infertile ( Supplementary Fig. S3D) with significant reductions in testicular size ( Supplementary  Fig. S3A and B) and an absence of spermatozoa in the epididymides ( Fig. 3A; Supplementary Fig. S3C). H&E staining revealed that the seminiferous tubules of both WT and Adad2 Mut/ Mut mice contained preletotene spermatocytes and multiple layers of round spermatids at Stage VII-VIII (Fig. 3A). However, unlike the WT mice which had a layer of spermatozoa lining in the lumen, only a few spermatozoa were observed in the Adad2 Mut/Mut mice (Fig. 3A), suggesting that spermiogenesis was compromised in the Adad2 Mut/Mut mice. We then used PAS staining to investigate the spermiogenesis defects in more detail. Mouse seminiferous tubules can be divided into 12 stages (Russell et al., 1990). No obvious morphological defects in the Stage I-VIII tubules were seen in the Adad2 Mut/Mut mice ( Supplementary Fig. S3E). From Stage IX, the round spermatids in WT mice began to elongate. However, this process was largely delayed in the Adad2 Mut/Mut mice, with most of the spermatids having insufficiently flattened nuclei at Stage XII, and even spermatids without elongation were also visible ( Supplementary Fig.  S3E). The number of elongating spermatids gradually decreased from Stage XII, and consequently, few spermatozoa were observed in the lumens of Stage VII-VIII tubules in the Adad2 Mut/Mut mice ( Supplementary Fig. S3E). These findings indicate that the Adad2 Mut/Mut mice have defective spermiogenesis due to ADAD2 mutations and suggest that similar mutations (MT3) in ADAD2 could cause human male infertility.
The absence of the ADAD2 protein in Adad2 ko mice has been shown to result in aberrant chromatin organization in meiotic spermatocytes and post-meiotic spermatids, characterized by increased heterochromatin marking at H3K9me3 and heterochromatin protein 1a (HP1a), and reduced euchromatin marking at H3K4me2 (Chukrallah et al., 2022). To investigate whether Adad2 mutations affect the chromatin status in Adad2 Mut/Mut mice, we performed the same chromatin marking analysis in the WT and Adad2 Mut/Mut mice. We found that compared with the WT mice, the level of H3K9me3 marking was not affected but the level of H3K4me2 marking was significantly reduced in the testes of the Adad2 Mut/Mut mice ( Fig. 3B and C). In addition, consistent with the reported Adad2 KO mice, immunostaining of HP1a showed multiple HP1a foci in the nuclei of round spermatids from the Adad2 Mut/Mut mice that were not seen in the WT mice ( Fig. 3D and E). This suggests that defective heterochromatin distribution may be responsible for the aberrant differentiation of round spermatids in the Adad2 Mut/Mut mice.

Reduction of ADAD2 protein in the testes of Adad2 Mut/Mut male mice
To explore the effects of Adad2 Mut/Mut mutations in mice, qPCR analysis was performed, and the relative expression of Adad2 mRNA between the testes of 3-week-old WT and Adad2 Mut/Mut littermates showed that Adad2 expression was reduced by 40% in Adad2 Mut/Mut (Supplementary Fig. S4A). Moreover, immunofluorescence staining was used to examine whether the mutation led to the translation of the truncated ADAD2 protein. In WT mice, the ADAD2 protein was detected in Stage IV pachytene spermatocytes, where it was well-dispersed in the cytoplasm. The immunofluorescence signal of the protein subsequently aggregated to form perinuclear granules in the cytoplasm of mid-and latepachytene spermatocytes from Stage V to IX tubules (Fig. 3F). Bright perinuclear granules were observed in diplotene spermatocytes at Stages X and XI as well as secondary spermatocytes at Stage XII (Fig. 3F) and foci-like ADAD2 was still present in the cytoplasm of round spermatids in early steps in Stage I-VI tubules ( Fig. 3F) but was completely absent during the following stages (Fig. 3F). The staging of the seminiferous tubules in the Adad2 Mut/Mut mice was difficult to determine using the acrosome marker peanut agglutinin (PNA), as the transition from round to elongated spermatids was largely delayed ( Supplementary  Fig. S3E). Therefore, we roughly classified the staging in the Adad2 Mut/Mut mice using the marker for DNA double-strand breaks, cH 2 AX, to distinguish the different phases of meiotic spermatocytes (Shi et al., 2019) and Hoechst 33342 to examine the type of spermatogonia (Russell et al., 1990). Cytoplasmic ADAD2 signals were detected in pachytene spermatocytes but were much weaker than those in the WT in Stage IV-VI tubules which contained intermediate and type-B spermatogonia (Fig. 3G); the exposure time of the ADAD2 signal was the same as that used for the WT. Stage IX-XI tubules were characterized by nuclei filled with cH 2 AX (leptotene/zygotene cells) along the basal membrane, and ADAD2 signals were completely absent from pachytene or diplotene spermatocytes (Fig. 3G). The ADAD2 signals also disappeared from Metaphase I and secondary spermatocytes in Stage XII tubules (Fig. 3G). No ADAD2 staining was found in round spermatids from any of the tubules (Fig. 3G). Given that the level of Adad2 mRNA was close to the background level at diplotene ( Supplementary Fig. S4B) (Chen et al., 2018), the absence of the ADAD2 protein in Adad2 Mut/Mut diplotene spermatocytes suggests Figure 2. Effects of ADAD2 mutations on ADAD2 protein expression and spermatogenesis in patients. (A) Western blot (WB) analysis of ADAD2 expression in testicular tissues of an obstructive azoospermia (OA) male control and men harboring ADAD2 variants. ACTB was used as an internal control. (B) Immunofluorescence staining of ADAD2 (red) and synaptonemal complex protein 3 (SYCP3) (green) in human testicular section of an OA control and men harboring ADAD2 variants. Magnified views of cytoplasmic localization pattern of ADAD2 are shown in the upper right corner of the overlay images. Scale bar, 50 lm. (C) Hematoxylin and eosin (H&E) of testicular cross-sections from an OA control and infertile patients carrying ADAD2 variants. The red arrows point to round spermatids which were differentiated according to previous reports (Clermont, 1963;Nihi et al., 2017). Scale bars, 50 lm. Adad2 Mut/Mut littermates (G) using ADAD2, acrosome marker peanut agglutinin (PNA), and DNA double stranded breaks marker cH2AX. DNA was counterstained with Hoechst. Staging of testicular sections was based on previous reports (Russell et al., 1990;Nakata et al., 2015;Shi et al., 2019;Gao et al., 2020). White arrows indicate the angle of acrosomes. Magnified views of the ADAD2 signal in spermatogenic cells are shown on the right. In, intermediate spermatogonia; B, type B spermatogonia; pL, preleptotene; L, leptotene; Z, zygotene; P, pachytene; D, diplotene; MI, metaphase I; sSC, secondary spermatocyte; Rst, round spermatid; Est, elongating spermatid (in Stage IX-XI tubules) or elongated spermatid (in Stage XII and I-III tubules); Sz, spermatozoa. Scale bars, 50 and 12.5 lm (enlarged box).
that a truncated ADAD2 protein (which would be less stable) was produced but degraded rapidly.
ROSI helped produce healthy offspring from Adad2 Mut/Mut male mice Since the Adad2 Mut/Mut mice model was observed to effectively represent the spermatogenic failure in the ADAD2 mutation-harboring patients, we explored potential therapeutic approaches for ADAD2-related NOA in the Adad2 Mut/Mut mice. As elongated spermatids or spermatozoa were extremely rare in the Adad2 Mut/ Mut mice, we attempted ROSI with round spermatids in stimulated WT oocytes (Fig. 4A). Although the round spermatids of the Adad2 Mut/Mut mice exhibited a significantly lower fertilization rate (87.7 § 1.06% in WT versus 30.85 § 2.76% in Adad2 Mut/Mut ) (Fig. 4B), Birth rate equals to the number of full-term pups delivered divided by the number of two-cell stage embryos transferred. Data were presented as mean § SD. NS, no significance (unpaired t test). n, rounds of ROSI repeated using WT and Adad2 Mut/Mut spermatids, respectively. RS, round spermatid; SO, stimulated oocyte. (G) Fertility test of ROSI-derived progeny. Four WT, four Adad2 WT/Mut male mice, and six Adad2 WT/Mut female mice were involved in the fertility test. Each male mouse was mated with two females for 3 months. Data were presented as mean § SD. NS, no significance (unpaired t test). n, the number of litters produced by each group.
the ex vivo development from zygote to blastocyst stage was comparable between the WT and Adad2 Mut/Mut spermatids ( Fig. 4C and D). We next transplanted the two-cell stage embryos derived from the ROSI into pseudo-pregnant females. Inspiringly, full-term viable Adad2 WT/Mut pups were delivered, and their birth rate was close to that obtained from the WT round spermatids ( Fig. 4E and F). Furthermore, these heterozygous offspring ( Supplementary Fig. S5) did not show any overt developmental defects and had normal fertility as adults (Fig. 4G). Taken together, our results suggested that ROSI could be a potential treatment for male infertility due to Adad2 mutations.

Discussion
In this study, we identified three recessive ADAD2 mutations in six idiopathic NOA-affected men from three unrelated Pakistani families (Fig. 1A). The ADAD2 mutations were predicted to be pathogenic from the WES data combined with strict mutation screening criteria (Supplementary Fig. S1 and Table S4). MT1/ MT2 significantly reduced the testicular levels of the ADAD2 protein and produced defects in spermiogenesis similar to those seen in Adad2-null mice (Fig. 2). In addition, the Adad2 Mut/Mut mouse model harboring MT3 had no sperm in the epididymides, suggesting a correspondence with the spermatogenic deficiency of ADAD2-mutant patients (Fig. 3A). The good symptomatic consistency between the patients and the mouse model suggests a causal relationship between ADAD2 mutations and human NOA. The ADAD2 protein is mainly expressed in spermatocytes in humans (Fig. 2B) and mice (Snyder et al., 2020). The ADAD2 protein levels were observed to be dramatically lower in patients carrying MT1 or MT2 mutations ( Fig. 2A), leading to defects in spermiogenesis (Fig. 2C), while spermatogonial proliferation and differentiation were not affected. The MT3 mutation (recapitulated by Adad2 Mut/Mut mice) potentially results in a truncated ADAD2 protein, predicted to be 60% shorter than the full-length protein and rapidly degraded before the diplonema phase. Spermatogenic analyses of the Adad2 Mut/Mut mice revealed a reduction in testicular size (Supplementary Fig. S3A and B), defective differentiation of round spermatids to elongated spermatids ( Supplementary Fig. S3E), and a complete absence of spermatozoa in the epididymides ( Fig. 3A; Supplementary Fig. S3C); this corresponds with the findings in Adad2 ko mice (Snyder et al., 2020). Therefore, the evidence of the comparable spermatogenic defects between patients with the MT1, MT2, and MT3 mutations, the Adad2 Mut/Mut mice mimicking the MT3 mutation, and the reported findings on the Adad2 ko mouse indicate that mutations in ADAD2 lead to a loss-of-function effect. Taken together, we suggest that ADAD2 is dispensable for spermatogonia development in humans but has a conserved role in mammalian spermiogenesis.
Adad2 Mut/Mut mice with multiple HP1a foci showed defective heterochromatin remodeling in round spermatids ( Fig. 3D and E) together with significantly reduced levels of H3K4me2 marking ( Fig. 3B and C). H3K4me2 is required for sperm-specific protamination (Godmann et al., 2007). Reduced H3K4me2 may impede histone differentiation to protamine resulting in malformed elongated spermatids (Chukrallah et al., 2022). Thus, we hypothesize that elongating/elongated spermatids in Adad2 Mut/Mut mice were eliminated due to faulty chromatin remodeling and morphological abnormalities that originated in round spermatids. However, further studies are required to clearly understand the molecular mechanism of ADAD2 mutation-induced spermatogenic failure.
In Adad2 Mut/Mut mice, we observed very few elongated spermatids or spermatozoa, suggesting that the frameshift or missense variants of ADAD2, located in the adenosine deaminase domain, are associated with a good prognosis for testicular sperm extraction, and that the retrieved sperm might be recovered and used for ICSI. However, a recent report described a patient with asthenoteratozoospermia carrying a homozygous missense mutation of ADAD2 (c.1381C>T, p.R461W), located in the adenosine deaminase domain, the same domain as our MT1 (MT1: c.G829T, p.G277C) and MT2 (c.G1192A, p.D398N), who failed to cause pregnancy even after ICSI treatment (Dai et al., 2023), suggesting that the poor-quality spermatozoa from the ADAD2 mutant patient were not sufficient to father a biological child. Therefore, ROSI might be an alternative therapy for patients harboring ADAD2 mutations. Although round spermatids from Adad2 Mut/Mut mice had a lower fertilization rate (Fig. 4B), comparable ex vivo (Fig. 4C and D) and in vivo ( Fig. 4E and F) embryogenesis capacities were observed between WT and Adad2 Mut/Mut -derived zygotes, suggesting that ADAD2 protein is involved in spermiogenesis and fertilization but not in early embryogenesis. The Adad2 WT/Mut progeny derived from ROSI were viable, healthy and had normal fertility (Fig. 4G), indicating that ROSI can be a feasible treatment for infertile Adad2 Mut/Mut mice. In the clinic, a ROSI-related concern is whether the inadequate replacement of histones by protamine in round spermatids would lead to epigenetic modification(s) in the paternal genome, adversely affecting the next generation (Siklenka et al., 2015). Promisingly, in 2018, a tracking survey of 90 babies born from ROSI showed no significant differences in physical and cognitive development during the first two years of life compared with naturally conceived infants (Tanaka et al., 2018). A recent study on the fetuses and placentae of embryonic day 11.5 mice also showed that the overall transcriptomic profiles and general methylation levels were similar between ROSI and ICSI-produced pups (Zhu et al., 2021). These findings enhance our understanding of ROSI and provide valuable clues for the clinical application of ROSI.
ADAD2 variants appear to be more frequent in Asian populations ( Supplementary Fig. S6), especially the frameshift variant which has a frequency of 1 in 1420 ( Supplementary Fig. S6). We propose that this could be due to a founder effect since ADAD2 variants originated from a subset of ancestors of the Pashtun ethnic group. Notably, ADAD2 is specifically expressed in the testes and ADAD2 variants are deleterious in men but not in women. Hence, ADAD2 mutations can be silent in the population in a heterozygous state in men or homozygous in women and may thus be less subject to selection pressure. The cultural practices of endogamy, polygyny and consanguineous marriage would allow the passage and expansion of this variant over generations within specific populations, leading to relatively high risks of male infertility, as seen in the Pakistani population. Apart from the ADAD2 gene, several pathogenic variants associated with male infertility in other genes have also been reported, many of which appear to be especially prevalent in specific populations. These include the homozygous frameshift variant c.676dup in M1AP which was identified in multiple non-Finnish Europeans. This mutation is relatively prevalent in European populations and most likely originated from a founder mutation . Identifying the existence of population-specific disease alleles is not only valuable for estimating the recurrence risks of related diseases in specific populations, but also provides precious resources for understanding the genetic causes of such diseases.
In summary, we identified a function of ADAD2 in human spermiogenesis and verified a causal relationship between ADAD2 mutations in human NOA. Inspiringly, ROSI helped produce healthy offspring from infertile Adad2 Mut/Mut mice that had abnormal heterochromatin organization in round spermatids. Our work provides preliminary clues for genetic counselling of ADAD2 mutation-associated male infertility.

Supplementary data
Supplementary data are available at Human Reproduction Open online.

Data availability
The data used to support the findings of this study are available from the corresponding author upon request.