Ninety-two men of Han Chinese ethnicity who had been diagnosed with primary infertility and AZS were recruited from Suzhou Municipal Hospital. Details regarding the history of infertility and related information were collected for all patients, and samples of peripheral whole blood were isolated for WES analyses. Clinical assessment indicated that the probands discussed in this study were able to achieve a normal erection and exhibited intact ejaculatory function, proper external development of the male genitalia, appropriate secondary sexual characteristics and bilateral testicular size, and hormone levels within the expected range. All patients were karyotypically normal (46; XY) with no evidence of Y chromosome microdeletions.

DNA for sequencing of the protein-coding genes was prepared from the whole-blood samples using a TIANamp Blood DNA Kit (TIANGEN Biotech, Beijing, China), according to the provided instructions. Exon capture and sequencing were performed with an AlExome Enrichment kit V1 (iGeneTech, Beijing, China) and a Hiseq2000 platform (Illumina, CA, USA), respectively. The reads were aligned to the hg19 (GRCh37) human reference genome using default Burrows-Wheeler Aligner (http://bio-bwa.sourceforge.net/) parameters. Genomic variant calling was performed with the Genome Analysis Toolkit HaplotypeCaller (http://www.broadinstitute.org/gatk/), after which ANNOVAR was used for filtering and annotation (https://annovar.openbioinformatics.org/en/latest/).

Variants with a gnomAD (http://gnomad-sg.org/) allele frequency > 1% were excluded, as were variants in upstream, downstream, or intronic regions. Any frameshift, nonsense, key splice-site, or potentially deleterious missense variants identified using PolyPhen-2, Mutation Taster, and SIFT were retained for further evaluation. Any genes harboring two potentially deleterious missense or loss-of-function mutations were retained, and these candidate genes were cross- referenced with genes known to be enriched in the testis and associated with AZS phenotypes. Positive variant candidates were additionally investigated via Sanger sequencing.

The target genomic regions of interest were amplified with appropriate primers (Supplementary Table S1) and PhantaTM Super-Fidelity DNA Polymerase (Vazyme, P501) via PCR with the following settings: 95°C for 2 min; 35 cycles of 95°C for 10 s, 58°C for 15 s, 72°C for 8 min, and 72°C for 10 min. An ABI Prism Big Dye Terminator Cycle Sequencing Ready Reaction kit was then used for direct sequencing of these amplicons using an ABI 3100 Genetic Analyzer (Applied Biosystems, CA, USA), and the resultant PCR products (214/264 bp) were analyzed via by 2% agarose gel electrophoresis. DNA sequence alignment was performed with SnapGene (v 3.2.1).

Semen samples were collected from study participants after sexual abstinence for 2-7 days and analyzed after liquefaction at 37°C for 30 min. Liquefaction was completed within an hour. The semen parameters were analyzed by a computer-assisted analysis system (CASA) device using IVOS software (version 12, Hamilton-Thorne Biosciences). Up to 10 sequels, 10s long were acquired for each sample. The sample volumes, sperm concentrations, and sperm motility were assessed according to the WHO guidelines (20). H&E staining was performed to evaluate sperm morphology. A minimum of 200 spermatozoa per patient were analyzed to assess the frequency of morphologically abnormal cells as per WHO guidelines.

As described elsewhere (21), human sperm cells (10⁷ cells/ml) were incubated in capacitation medium (Ham’s F-10) supplemented with 3 mg/ml BSA for 3 h at 37°C in 5% CO_2. The semen parameters were then analyzed by CASA. A minimum of 100 sperm cells were analyzed each time. The proportion of hyperactivated (HA) spermatozoa in each sample was determined using the SORT function of the CASA instrument. In human spermatozoa, HA motility is defined by curvilinear velocity (VCL) > 100 μm/s, linearity (LIN) < 60%, and lateral head displacement (ALH) > 5 μm.

Western immunoblotting was performed as previously described (22) with some modifications. Initially, the spermatozoa were lysed with buffer containing 7 M urea, 2M thiourea, 2% (w/v) DTT), and 1% (v/v) protease inhibitors (Pierce Biotechnology). The lysates were centrifuged and proteins in the supernatants were separated by SDS-PAGE and transferred to PVDF membranes. The blots were blocked with 5% skim milk in TBST at room temperature for 2 h and were then incubated overnight at 4°C with appropriate primary antibodies (Supplementary Table S2). The protein bands were then detected using the SuperSignalWest Femto Chemiluminescent Substrate system (Thermo Scientific).

The spermatozoa were rinsed three times with PBS, spread on microscope slides, and allowed to air dry. The cells were then fixed with 1% paraformaldehyde for 10 min, rinsed three times using PBST (10 minutes per wash), blocked with 1% BSA for 1 h, and probed overnight using appropriate primary antibodies at 4°C. Samples were then probed with secondary antibodies for 2 h, nuclei were counterstained for 5 min using Hoechst 33342, rinsed with PBS, mounted with Immu-Mount or VectaShield, and imaged with an ORCA Flash4.0 digital monochrome camera (Hamamatsu Photonics) on a Leica DM5500B microscope (Leica Microsystems).

Transmission electron microscopy (TEM) was performed by initial fixation of spermatozoa in 2.5% phosphate-buffered glutaraldehyde. The cells were then rinsed three times with 0.1 mol/L phosphate buffer (PB, pH 7.2), treated for 60-90 min with 1% osmium tetroxide in 0.1 mol/L PB at 4°C, and dehydrated in an ethanol gradient ( 50, 70, 80, 95, and 100%), followed by 100% acetone. The samples were then treated overnight with 1:1 acetone and SPI-Chem resin containing dodecyl succinic anhydride, N-methylacetamide, SPI-Pon 812, and DMP-30 at 37°C and embedded in Epon 812. Ultrathin sections were cut and stained with lead citrate and uranyl acetate before TEM imaging (TECNAI-10, Philips) at an accelerating voltage of 80 kV.

HEK293T cells were co-transfected with the PcDNA3.1-KIF9-Flag and pcDNA3.1-HYDIN-HA or pcDNA3.1-SPAG6\16-HA constructs. The cells were then lysed in IP lysis buffer (20 mM Tris, pH 7.4, 2 mM EGTA, 1% NP-40) containing a protease inhibitor cocktail (Roche, 04693132001) for 30 min on ice. The lysates were then centrifuged for 15 min at 13000 g and the supernatants were incubated overnight with anti-FLAG at 4°C followed by incubation for 2 h with protein A-Sepharose (GE, 17-1279-03) at 4°C. The precipitates were then rinsed twice in IP buffer and mixed with 1% SDS sample buffer via before separation on SDS-PAGE and immunoblotting as described above.

In this study, 92 Chinese males with AZS underwent WES analyses. Two unrelated members of this cohort exhibiting abnormal sperm motility were found to harbor homozygous KIF9 variants, as confirmed via by Sanger sequencing (Figures 1A–C). One of the homozygous KIF9 variants was a frameshift mutation (c.1433delinsA: p.N478Tfs*39) detected in proband A021 II-1, a member of the consanguineous family A021. Both of the patient’s parents were heterozygous carriers for the variant (Figure 1A). The second homozygous KIF9 variant detected in proband II-1 from family A062 was a stop-gain variant (c.1861A>T: p.K621X) in the C-terminus of KIF9. This patient’s mother was a heterozygous carrier for this variant (Figure 1A). Analysis of the prevalence of these variants by gnomAD showed that they were largely absent in the general population, indicating that they were rare variants (Table 1). The two variants were thus identified as deleterious bi-allelic KIF9 variants present in males with AZS.

The nonsense-mediated decay process facilitates the selective degradation of mRNAs harboring premature stop codons (23, 24). If not degraded, these mRNAs generally produce truncated proteins that result in either damaging gain-of-function phenotypes or dominant-negative characteristics (23, 24). To understand the potential pathogenicity of the detected variants, western immunoblotting and immunofluorescence were used to assess the abundance and localization of KIF9 within spermatozoa. This showed the presence of a KIF9-specific protein band at ~100 kDa in control spermatozoa that was absent in the sperm of the two probands (Figure 1D), indicating a loss of KIF9 expression. Immunofluorescent staining also confirmed that KIF9 was present in control spermatozoa but was undetectable in the sperm flagella from the two probands (Figure 1E). These findings thus indicated that an absence of KIF9 expression in the probands harboring the identified bi-allelic KIF9 variants.

Routine semenological analyses were next performed in accordance with the WHO guidelines. This showed that the semen from both probands was normal with respect to color, pH, volume, and smell (Table 2). Both samples contained sperm concentrations and motility levels above the reference value (Table 2). However, the percentages of forward motile sperm in the samples from the probands were just 13.4% and 18.6%, respectively (normal: ≥32%; 25), consistent with an AZS phenotype (Table 2). Both probands produced sufficient numbers of morphologically normal sperm (65% and 55%, respectively). Further, sperm cells of A021 II-1 were incubated in capacitation medium for 3 h, and the percentage of hyperactive cells was evaluated. The data showed no sperm hyperactivation (Table 2). These results suggest that KIF9-deficient spermatozoa from the two AZS patients were morphologically normal but exhibited impaired forward movement. KIF9 functions as a protein that can specifically regulate flagellar beating.

To determine whether assisted reproductive technologies could overcome the AZS-related infertility observed in the two patients, both impacted couples (which included F1: II- 1 and F2: II- 1) underwent IVF treatment at our hospital. One round of IVF for the couple including the A021 patient led to the retrieval of 8 MII oocytes which achieved a 62.5% fertilization rate. All 8 of these embryos underwent cleavage and developed to the blastocyst stage. The couple carried one of these embryos to term, giving birth in 2021 (Table 3). In total, 15 MII oocytes were retrieved for A062, with a 60% fertilization rate and 9 embryos that developed into blastocysts. After undergoing an embryo transfer, the wife gave birth to a child in 2021 (Table 3). These results were consistent with findings in mice in which Kif9-deficient sperm could still fertilize eggs, albeit at lower efficiency (18).

To clarify the mechanisms whereby KIF9 deficiency contributed to the observed AZS phenotypes in humans, the structural characteristics of the spermatozoa from the two probands were next analyzed. H&E staining showed that the spermatozoa appeared similar to controls, consistent with the results of the semenological analyses (Figure 2A). Immunofluorescent staining for the flagellum (α-tubulin), acrosome (PNA), MS (TOMM20), and FS (AKAP4) confirmed the presence of these components with no abnormalities in both probands (Figures 2B–D).

As the deletion of axonemal proteins frequently disrupts the structural integrity of the axoneme (6, 11, 26), TEM was next used to assess the ultrastructural characteristics of the flagella (Figure 3). Midpiece sections from both a normal control donor and the two KIF9-deficient probands showed mitochondria forming an outer layer with a central axoneme surrounded by ODFs. Moreover, the axoneme, ODF, and FS were clearly visible in samples from both control donors and probands, with the membrane wrapping around the filaments in the principal piece. These intact structures were in line with previous observation on sperm from Kif9-deficient mice.

KIF9/KLP1 have been reported to localize to the axonemal CP in both mice and Chlamydomonas, associating with the CP protein HYDIN (18, 27). To test whether similar interactions also occurred in human cells, HA-HYDIN and FLAG-KIF9 expression plasmids were generated and the two genes were co-expressed in HEK293T cells. Subsequent immunoprecipitation analysis confirmed the interaction of these two proteins (Figure 4A). HA-tagged plasmids encoding SPAG6 and SPAG16, which are also CM components, were also generated but KIF9 failed to interact with either of these proteins (Figure 4A).

Knockdown of HYDIN in Chlamydomonas results in a marked drop in KLP1 expression (27), and Hydin-KO mice also lack KIF9 in the spermatozoa (18). As human spermatozoa lacking HYDIN were unavailable, immunofluorescent staining was instead used to detect KIF9 in sperm deficient in SPAG6, which lack a CP, as evident in ultrastructural cross-sections prepared from the sperm flagella (11, 12). No KIF9 staining was detectable in these SPAG6-deficient sperm (Figure 4B). Moreover, KIF9 staining was performed in sperm lacking DNAH10, which retain the CP but exhibit MMAF phenotypes (10). KIF9 expression was observed in the tails of the DNAH10-deficient spermatozoa even though they were abnormally short and curly (Figure 4B). In addition, we detected the level of HYDIN in the KIF9-deficient spermatozoa, and no difference was detected between the samples (Figure 1D). These analyses indicated that KIF9 is potentially associated with HYDIN and localized to the axonemal CP in humans.

In cilia, KIF9 deficiency has been linked to altered proximal-to-distal radial spoke and inner dynein arm patterning (19). Given the high degree of structural conservation observed between respiratory cilia and sperm flagella, RSPH3 and DNALI1 expression was next evaluated in the spermatozoa from the two probands. These proteins were selected as RSPH3 is a radial spoke protein (28), whereas the DNALI1 is found in association with the inner dynein arm (29). No significant changes in the lengths or distributions of radial spokes or inner dynein arms were observed between the control and KIF9-deficient spermatozoa (Figures 5A, B). Although it is possible that measurements of flagellar length could be inaccurate due to their extended length and tendency to curl. Western immunoblotting confirmed that there were no significant reductions in DNALI1 or RSPH3 levels in KIF9-deficient spermatozoa (Figure 5C), indicating that the loss of KIF9 has no adverse impact on the patterning of the radial spoke or inner dynein arm domains of human flagella.