Animal care, breeding setup, and experimental procedures were approved according to the German law of animal protection and in agreement with the approval of the local institutional animal care committees (Landesamt für Natur, Umwelt und Verbraucherschutz, North Rhine-Westphalia, approval ID: AZ84- 02.04.2013.A429).

Guide RNAs (gRNAs) specific for Pfn3 were designed using the online tool from Feng Zhang’s lab¹ (Hsu et al., 2013). Designed gRNAs were annealed and cloned into pX330-U6-Chimeric_BB-CBh-hSpCas9 (plasmid # 42230 obtained from Addgene) (Cong et al., 2013) as described previously (Ran et al., 2013).

The functionality of gRNAs was checked using E14Tg2a mES cells (kind gift of Christof Niehrs, IMB Mainz, Germany). On gelatinized cell culture dishes with standard ES cell medium without antibiotics, 3 × 10⁵ cells per well were seeded at 37°C and 7.5% CO₂. After 3 h, cells were transfected with Lipofectamine 2000, pX330 containing gRNAs in a ratio of 1:3, according to the manufacturers’ protocol (Thermo Fisher Scientific, Waltham, United States). To remove DNA–Lipofectamine complexes, the medium was changed after 6–8 h to standard ES media.

Superovulation was done by intraperitoneal injections of PMS (pregnant mare’s serum, 5 IU) and hCG (human chorionic gonadotropin, 5 IU) in C57BL/6J female mice and two females mated with one C57BL/6J male. At 0.5 dpc, zygotes from the oviducts were isolated and microinjection was done using an inverted microscope (Leica, Wetzlar, Germany) equipped with micromanipulators (Narishige, Japan) and piezo unit (Eppendorf, Hamburg, Germany). Injection pipettes (PIEZO 8-15-NS, Origio, Charlottesville, United States) were filled with Fluorinert (FC-770, Sigma-Aldrich, Taufkirchen, Germany) for appropriate piezo pulse propagation. Co-injection of Cas9 mRNA (100 ng/μl) (Sigma-Aldrich) and in vitro-transcribed single-guide RNAs (sgRNAs) (50 ng/μl each) were achieved into the cytoplasm of zygotes, as described previously (Yang et al., 2014). The zygotes that survived after microinjection were kept for 3 days in KSOM medium in a CO₂ incubator. Resulting blastocysts were transferred into the uteri of pseudo-pregnant foster recipients. The alleles were registered with mouse genome informatics and received the following IDs: Pfn3^em1Hsc MGI:6384215, Pfn3^em2Hsc MGI:6384216, and Pfn3^em3Hsc MGI:6384217.

Genomic DNA was extracted from mice following the phenol/chloroform method and subjected for PCR. Gene-specific primers were used for PCR. PCR products were sequenced to identify the locus-specific deletions mediated by CRISPR/Cas9 genome editing. Primer list is given in the Supplementary Data (Supplementary Table 1).

For fertility analysis, total five–seven WT and KO male mice aged 10 to 12 weeks were individually housed with sexually mature WT C57BL/6J females in a controlled breeding experiment. Females were observed for the presence of vaginal plugs and pregnancies. The average litter size from pregnant females were calculated.

Pfn3 knockouts were used for gross morphological analyses including body weight, testis weight, epididymis weight, and appearance of testes as compared to heterozygous and wild-type littermates (n = 13 animals/genotype).

Mice epididymal sperm were extracted by multiple incisions of the cauda followed by a swim out for 30–60 min in M2 medium. Using a Neubauer hemocytometer, sperm count was determined (n = 13). A sperm vitality analysis was performed using eosin and nigrosine (E&N) staining, and sperm membrane integrity was accessed by hypo-osmotic swelling test (HOS test); 200 spermatozoa were calculated from three animals per genotype (WT, heterozygous, and knockouts) in a repetition, and results were expressed as percentage of live-to-dead sperm ratio. For all experiments, sexually mature males with an age of 2–6 months were used.

For nuclear morphology analysis, after swim out, sperm cells were fixed and washed three times in 2% paraformaldehyde (PFA). After fixation, sample was diluted in a fixative and spread evenly on a glass slide and allowed to air dry. Slides were counterstained with DAPI (Carl Roth, Karlsruhe, Germany) as described previously (Skinner et al., 2019). Images were taken on the Leica DM5500 B/JVC KY-F75U digital camera. Images were analyzed using the ImageJ plugin “Nuclear morphology analysis v1.15.3” according to the developer’s instructions.

The testes were fixed in Bouin’s solution for 4–24 h, washed with 70% ethanol to remove excess Bouin, embedded in paraffin, and sectioned at 5-μm thickness. Immunohistochemistry (IHC) was done in the Lab Vision PT module (Thermo Fisher Scientific) and Autostainer 480S (Medac, Hamburg, Germany) as published previously (Nettersheim et al., 2013). The primary antibodies against PFN3 (BEG6 kind gift of Prof. Dr. W. Witke), LC3B (ab58610, Abcam), p-mTOR (Cat#2971, Cell Signaling), RAB5 (PA-5-29022), and ARPM1 (27580-1-AP) were used.

For the crude protein extracts, the testes were homogenized into 500–900 μl RIPA buffer (Thermo Fisher Scientific, Waltham, United States) using a Dounce homogenizer (Sartorius, Göttingen, Germany). The homogenized mixture was kept on ice for 15 min followed by 15 min of centrifugation at 4°C and 13,000 rpm. In order to load equal amounts of protein, supernatant was used to measure the protein concentration using the Pierce BCA Protein Assay Kit (Thermo Fisher Scientific, Waltham, United States). Cytoplasmic and nuclear fractions were separated as described previously (Hara et al., 2008). SDS gel followed by western blot was performed as described previously (Nettersheim et al., 2016) with primary antibodies against PFN3 (BEG6 kind gift of Prof. Dr. W. Witke), CFL1, CFL2, and ADF (kind gift of Prof. Dr. W. Witke), and ARPM1 (27580-1-AP, ProteinTech), ATG2A (Cat#PA5-77794, Thermo Fisher Scientific), TRIM27 (12205-1-AP), LC3B (ab58610, Abcam), SQSTM1 (Cat#5114, Cell Signaling), mTOR (Cat#2972, Cell Signaling), p-mTOR (Cat#2971, Cell Signaling), and AMPK (Cat#5831, Cell Signaling) were used.

Protein immunoprecipitation (IP) was performed on whole adult testis lysate. Testis tissue was homogenized in RIPA buffer followed by sonication at high speed for five cycles of 30-s ON/30-s OFF with a Bioruptor^® sonication device, and protein extraction was done as described above. Co-IP was performed using Dynabeads^® M-280 Sheep Anti-Rabbit IgG as described by the manufacturer’s protocol. The captured and eluted proteins were separated by SDS-PAGE and transferred to nitrocellulose membranes in preparation for immunoblot analysis.

RNA was extracted from testes tissue after removal of the tunica albuginea using TRIzol reagent according to manufacturer’s protocol (Life Technologies, Carlsbad, United States). The concentration and purity of isolated RNA was measured by a NanoDrop instrument (Peqlab, Erlangen, Germany). After DNAseI treatment of RNA, cDNA was synthesized using RevertAid First Strand cDNA Synthesis Kit (Fermentas, St. Leon-Rot, Germany). Quantitative reverse transcription polymerase chain reaction (qRT-PCR) was performed on ViiA 7 Real-Time PCR System (Applied Biosystems, distributed by Life Technologies) using Maxima SYBR Green qPCR Master Mix (Life Technologies) as described previously (Jostes et al., 2017). At the end of each PCR run, a melting point analysis was performed. GAPDH was used as reference gene for data normalization.

RNA integrity (RIN) was determined by the sequencing facility (UKB sequencing core facility) using the RNA Nano 6000 Assay Kit with the Agilent Bioanalyzer 2100 system (Agilent Technologies, Santa Clara, CA, United States). RIN values ranged from 8.2 to 10 for all samples. RNA sample quality control and library preparation were performed by the sequencing facility (UKB sequencing core facility), using the QuantSeq 3’-mRNA Library Prep (Lexogen). RNA-seq was performed on the Illumina HiSeq 2500 V4 platform, producing > 10 million, 50-bp 3’-end reads per sample.

All samples were mapped to the mouse genome (GRCm38.89). Mapping was done using HISAT2 2.1 (Kim et al., 2015). Transcript quantification and annotation was done using StringTie 1.3.3 (Pertea et al., 2015). Gene annotation information for the mouse genome was retrieved from the Ensembl FTP server² (GRCm38.89). We used the python script (preDE.py) included in the StringTie package to prepare gene-level count matrices for analysis of differential gene expression.

Differential expression was tested with DESeq2 1.16.1 (Love et al., 2014). Pseudogenes were removed from the count matrices based on “biotype” annotation information extracted from Biomart (R-package biomaRt) (Durinck et al., 2005). Low counts were removed by the independent filtering process implemented in DESeq2 (Bourgon et al., 2010). The adjusted p-value (Benjamini–Hochberg method) cutoff for DE was set at <0.05, and log2 fold change of expression (LFC) cutoff was set at >1.5.

The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found at: https://www.ncbi.nlm.nih.gov/geo/, GSE171068.

Epididymal sperm and testis tissue were washed in phosphate buffered saline (PBS) and fixed in 1.5% glutaraldehyde in 0.1 M cacodylate buffer (pH 7.4). Sperm cells were rinsed in 0.1 M cacodylate buffer and fixed in 2% osmium tetroxide again followed by an additional washing in 0.1 M cacodylate buffer. Afterward, dehydration was performed by an increasing ethanol concentration, terminated by two incubations in propylene oxide for 15 min, and an interim staining in 0.5% uranyl acetate. Samples were stored in propylene oxide and EPON mixture (1:1) overnight and followed by embedding in EPON for 24 h at 70°C. Ultrathin sections were picked up on grids, stained with 3.5% uranyl acetate for 25 min and lead citrate solution for 7 min, and images were taken on the Philips CM 10 TEM. For immunogold labeling, prior to fixation in glutaraldehyde, sections were incubated with PFN3 antibody (1:50) for 1 h. Followed by washing, gold-conjugated anti-rabbit IgG (10 nm) secondary antibody (1:50) was incubated for 1 h. In the negative control, primary antibody was omitted. Finally, the sections were post-stained with uranyl acetate.

Sperm cells were fixed in primary fixative for 30 min at 4°C: 1% glutaraldehyde and 0.4% formaldehyde in 0.1 M sodium cacodylate buffer (pH 7.2). After washing the samples three times in 0.1 M sodium cacodylate buffer for 5 min, they were post-fixed with 1% OsO₄ in 0.1 M sodium cacodylate buffer for 30 min followed by three washing steps. Images were taken using a Verios 460L (FEI, Eindhoven, Netherlands) (Nussdorfer et al., 2018) equipped with a STEM 3 detector.

Immunofluorescence staining was performed on testis sections using peanut agglutinin (PNA)- fluorescein isothiocyanate (FITC) Alexa Fluor 488 conjugate (Molecular Probes, Invitrogen), anti-mouse GM130 (610823, BD Biosciences, United States), and anti-rabbit TGN46 (JF1-024, Thermo Fischer Scientific) to check the acrosome biogenesis and cis- and trans-Golgi structural organization, respectively; 5-μm testis tissue sections were taken on slides. Deparaffinization on Bouin’s fixed paraffin-embedded testis tissue was performed by immersing them in xylol for two times, followed by dehydration steps in 100, 95, and 70% alcohol. After washing with PBS twice, tissue sample was permeabilized in 0.1% Triton X-100 for 5 min at 37°C, followed by blocking in 1% bovine serum albumin (BSA) for 1 h at room temperature. Tissue sections were incubated with diluted PNA-FITC in PBS at room temperature for 30 min; next, slides were washed in PBS for two times. Tissue sections were incubated with GM130 and TGN46 antibodies diluted at 2 μg/ml in 0.1% BSA at 4°C for overnight, followed by PBS washing, and respective secondary antibodies were used for 1 h at room temperature. Tissue sections were mounted with ROTI^®Mount FluorCare DAPI (Carl Roth, Karlsruhe, Germany). Images were taken within 24 h using an LSM 710 (Zeiss, Oberkochen, Germany).

For sperm immunofluorescence, after swim out, sperms were washed two–three times in PBS followed by fixation in 4% paraformaldehyde for 20 min at room temperature. After washing with PBS, spermatozoa were incubated with PNA-FITC and Mito Red (5 nM, 53271; Sigma Aldrich) diluted in PBS at room temperature for 30 min. For manchette staining, germ cell population was isolated from testes as described previously (Li W. et al., 2015), and staining with anti-alpha tubulin antibody (clone DM1A Alexa Flour 488 conjugate; 16-232 Sigma Aldrich) was performed as described previously (Rotgers et al., 2015). Next, spermatozoa were plated onto Superfrost Plus Microscope Slides (Thermo Fisher Scientific), mounted with ROTI^®Mount FluorCare DAPI (Carl Roth, Karlsruhe, Germany). Images were taken within 24 h using an LSM 710 (Zeiss, Oberkochen, Germany).

Epididymal sperm cells after swim out were diluted into PBS (1:10); 400 μl of diluted sperm were loaded on poly-L-lysine (CAS 25988-63-0, Sigma Aldrich)-coated coverslips in a six-well plate and air dried for 30 min at 37°C. After removing the PBS, sperm cells were fixed in 4% PFA followed by quenching with 50 mM NH₄Cl for 15 min. Sperm cells were permeabilized by 0.02% Triton X-100 for 3 min followed by washing with PBS, incubation for 1 h at RT with Phalloidin ATTO647 (1/1,000 in 3% BSA, ab176759, Abcam), and mounting with ProLong Gold Antifade (#P36930, Life Technologies). STED micrographs were acquired using a four-channel easy3D super-resolution STED optics module (Abberior Instruments, Göttingen, Germany) coupled with an Olympus IX73 confocal microscope (Olympus, Tokyo, Japan) and equipped with an UPlanSApo × 100 (1.4 NA) objective (Olympus, Tokyo, Japan) (Mikuličić et al., 2019), available in the LIMES Imaging Facility.

Epididymal sperm were incubated in TYH medium at 37°C for 20 min, after swim out sperm cells were diluted into the TYH medium supplemented with BSA. The diluted cell suspension was placed onto the pre-warmed (37°C) slide, and video was recorded using the Basler Microscopy ace camera (acA 1920-155uc) at 300 frames, streaming video using pylon Viewer (v.5.0.11.10913, Basler AG, Ahrensburg, Germany). Sperm motility was evaluated by using the OpenCASA program as described previously (Alquézar-Baeta et al., 2019).

Sperm were isolated from cauda epididymis and allowed to swim out in M2 medium for 10 min. Capacitation was induced by incubating in HTF medium for 90 min at 37°C, 5% CO₂. To induce acrosome reaction, calcium ionophore A23187 (10 μM, c7522; Sigma Aldrich) was added. After 15 min, sperm were spread on glass slides, air dried, and fixed in methanol for 3 min at RT. Followed by PBS washing, sperm were subjected to Coomassie brilliant blue staining (2% w/v G250) for 3 min as described previously (Liu et al., 2019). Followed by PBS washing two times for 2–3 min each, mounting was performed by using ROTI^®Mount FluorCare (HP 21.1 Carl Roth). Two hundred spermatozoa for each genotype (n = 3) were assessed by a bright field microscope.

The mean of all values for a particular data set has been represented in graphical form. Error bars have been used to denote standard deviation. Student’s T-test (two-tailed unpaired and one-tailed paired) and ANOVA (Tukey’s post hoc) were performed to ascertain the significance of the variability in data. P-value less than 0.05 (p < 0.05^∗, <0.005^∗∗, 0.001^∗∗∗) was considered as statistically significant.

We examined the distribution of PFN3 at ultrastructural level during spermiogenesis with a PFN3 polyclonal antibody and a gold-conjugated secondary antibody using immunoelectron microscopy. With immunogold labeling, we were able to determine localization of PFN3 in sub-domains of cellular compartments at the cis- and trans- Golgi network, the acroplaxome, mitochondria, and the manchette. In the Golgi phase, the electron dense giant acrosomal granule forms from the numerous Golgi-derived proacrosomal vesicles (Supplementary Figure 1A). Gold particles were detected representing PFN3 in the cis- and trans-part of the Golgi network adjacent to the nuclear pole of the round spermatid (Figure 1A and Supplementary Figures 1B,C,E,F, arrows). The cis-part is responsible for organizing and sorting of proteins imported from the endoplasmic reticulum (ER) and transported to the trans-part of the Golgi network where they are modified and exported as proacrosomal vesicles (Supplementary Figure 1E). Immunogold labelling was detected in the acrosomal granule (Figure 1B, circle), a giant structure attached to the middle of the acroplaxome (Figure 1B, white stars) of developing spermatozoa. We detected gold traces in the mitochondria, which are responsible for providing energy for flagellum propelling and germ cell differentiation (Supplementary Figures 1D,E, white arrows). Some of the gold particles were detected in the inner membrane of acrosome–acroplaxome interface, suggesting that PFN3 contributes to the attachment of acrosomal granule (Supplementary Figure 1E, white stars). In the cap phase (Supplementary Figure 1G), we detected PFN3 in the acrosomal vesicles (Figure 1C and Supplementary Figures 1H,I, asterisk), which are responsible for the formation of acrosome. In addition, we detected few gold traces in the nucleus (Figures 1B,C and Supplementary Figures 1H–M), suggesting that PFN3 might play a role in sperm nuclear shaping. Immunogold labelling also showed PFN3 localization in inner and outer membranes at the leading edge of the acrosome (Supplementary Figures 1H,I,K, elbow). Gold traces were also detected in highly specialized structures of elongating spermatozoa such as the manchette (Figure 1D and Supplementary Figures 1J,M, white arrow heads), the acroplaxome marginal ring (Supplementary Figure 1O, double arrows), as well as flagellum formation (Supplementary Figures 1N,Q,T). Cross sections of the flagellum showed gold traces in the mitochondria, which are gathered around the axoneme (Supplementary Figures 1U–W) to form the mitochondrial sheath in the sperm mid-piece (Supplementary Figure 1W). So, PFN3 is detected in the Golgi sub-domains, the acroplaxome–manchette complex, and the mitochondria, suggesting a role for PFN3 in acrosome biogenesis, sperm head shaping, and tail formation. The negative control is given in the Supplementary Data (Supplementary Figures 1Y,Z).

We generated Pfn3-deficient mice by injecting Cas9 mRNA and two sgRNAs into the cytoplasm of fertilized eggs targeting the exon of the gene (Figure 2A). Six pups carrying CRISPR/Cas9-induced mutations were identified (Supplementary Figure 2A) and three mouse lines, Pfn3Δ254, Pfn3Δ41, and Pfn3Δ29 harboring deletions of 254, 41, and 29 bp, respectively, were established by backcrossing with C57BL/6 mice. All deletions result in null alleles since they encode for frame shifts in the PFN3 reading frame leading to premature translational termination (Supplementary Figure 2C). qRT-PCR, western blotting, and IHC confirmed the deletion of Pfn3 in the Pfn3-deficient mice (Supplementary Figure 2D). Western blotting displayed a signal in testis/post-natal testis (day 28). All other tissues tested were negative for PFN3. Similarly, IHC produced a signal for PFN3 in the testis but not in the brain and kidney sections. These results indicate the specificity of the PFN3 antibody (Supplementary Figure 2E). IHC of WT testis section using anti-PFN3 antibody showed a strong expression of PFN3 in the nuclear region of elongated spermatids (Supplementary Figure 2D). In addition, IF using PFN3 antibody on purified germ cells revealed that PFN3 is localized to the cytoplasm of round spermatids. As round spermatids develop into elongating spermatids, PFN3 is localized in a punctuated manner around the acrosomal region (Supplementary Figure 2F). In non-acrosome-reacted spermatozoa, PFN3 is located in the acrosome and the nucleus. In acrosome-reacted spermatozoa, PFN3 can be detected in the head region (Supplementary Figure 2G). Males heterozygous for either Pfn3Δ254, Pfn3Δ41, and Pfn3Δ29 produced an average litter size of 7.6, 7.9, and 7.3, respectively (Figure 2B), which is comparable to wild-type mice with a mean litter size of 8.2 (Biggers et al., 1962). Homozygous Pfn3 male mice are sub-fertile since they produced an average litter size of 1.5, 2, and 4 for Pfn3Δ254, Pfn3Δ41, and Pfn3Δ29, respectively (Figure 2B).

Adult Pfn3-deficient males for either Pfn3Δ254, Pfn3Δ41, or Pfn3Δ29 mated with females normally as vaginal plugs were clearly detectable. Relative weights of the testes and epididymis of Pfn3-deficient mice (Pfn3Δ254) were slightly reduced; however, that reduction was not significant (Figures 2C,D). Histological analysis using hematoxylin and eosin H & E) staining on Pfn3^+/+, Pfn3^+/–, and Pfn3^–/– testes sections showed normal morphology of seminiferous tubules (Figure 2E). This suggests that deletion of Pfn3 left the gross morphology of the testes and epididymis unaffected.

However, Pfn3^+/– and Pfn3^–/– males showed a significant reduction in sperm count (Figure 2F). Next, E&N staining was performed to assess the vitality of spermatozoa. E&N distinguish live (whitish in color, arrow head) from dead sperm (pink in color, arrow in Supplementary Figure 2K). A percentage of live sperm in the range of 60–80% is considered normal, borderline 40–60%, and below 40% is considered abnormal. E&N staining showed that in heterozygous males, the percentage of viable sperm is in the normal range, while in case of Pfn3-deficient mice, the percentage of viable sperm was at borderline (∼40%) (Figure 2G). Further, the HOS test was performed to check the integrity of sperm membrane, where intact (live) sperm displays a swelling of the tail (Supplementary Figure 2M). The percentage of hypo-osmotic reactive sperm for Pfn3^–/– mice was again at borderline (Figure 2H). Male mice for Pfn3Δ41 and Pfn3Δ29 have the same phenotype as Pfn3Δ254 mice (Supplementary Figures 2H–N). In conclusion, loss of Pfn3 impinges not only on sperm quantity but also sperm quality.

We next used geometric morphometric analysis to analyze sperm head shape in detail. DAPI-stained sperm cells showed altered head shape for Pfn3^–/– compared to control (Figure 3A). An analysis of 994 nuclei revealed (334 Pfn3^+/+, 289 Pfn3^+/–, and 371 Pfn3^–/–) that Pfn3-deficient sperm shows alterations in area (Figure 3B), circularity (Figure 3C), length of hook (Figure 3D), bounding width (Figure 3E), and regularity (Figure 3F). Next, clustering was performed to categorize the sperm heads. Of note, 50% of heterozygous sperm nuclei are similar to WT, while 50% have abnormal morphology similar to Pfn3-deficient sperm nuclei (Figure 3G). This finding suggests a gene–dosage effect, which, however, does not seem to affect the mating success of Pfn3^+/– males. Interestingly, 70–80% of Pfn3-deficient sperms showed irregular/round head morphology. Our analysis clearly shows that PFN3 plays a role in shaping of the sperm head during spermiogenesis.

To analyze the swimming properties of Pfn3-deficient sperm, we performed computer-assisted semen analysis (CASA). Compared to WT and Het sperm samples (Table 1) Pfn3^–/– sperm showed significantly reduced progressive and total motility. The other sperm motility parameters such as curvilinear velocity (VCL), straight-line velocity (VSL), and average path velocity (VAP) are reduced, but the difference is not significant when compared to sperm samples of controls. Of note, the observed reduction in sperm motility most likely is the reason of the decreased sperm count in Pfn3-deficient mice.

PFN3 is detected in actin-rich structures such as acrosome–acroplaxome and Golgi complex, suggesting a role in acrosome biogenesis. In order to investigate the development of acrosome, we used PNA-FITC fluorescence labeling on testes sections. In tubules of wild-type (Figure 4A) and heterozygous males (Figure 4B), Golgi-phase spermatids showed developing acrosomes forming a homogenous single cluster on the apical face of cell nuclei. In mutant spermatids, PNA staining was scattered, suggesting a less uniform acrosomal compartment (Figure 4C). This abnormal formation of the acrosome was further observed in the next step, the cap phase. Here, the acrosome forms a cap-like structure covering the anterior half on round spermatids as seen in WT (Figure 4D) and heterozygous (Figure 4E) sections. In Pfn3^–/–, the cap structures were more unevenly distributed in round spermatids (Figure 4F). In the acrosomal phase, as spermatids started to elongate during sperm head remodeling, these defects were more prominent, and the acrosomal content failed to form an arrow-like shape in elongating spermatids of Pfn3^–/– mice (Figure 4I), compared to WT (Figure 4G) and heterozygous (Figure 4H). These results indicate that loss of PFN3 impairs acrosomal biogenesis already at the Golgi phase.

To understand the impaired acrosome biogenesis more in detail, an ultrastructure analysis using transmission electron microscopy (TEM) was performed on developing spermatids. In the Golgi phase, proacrosomal granules originate from the trans-Golgi and fuse to form a single, large acrosome vesicle attaching itself in the middle of the acroplaxome on the nuclear surface as seen in WT (Supplementary Figure 3A) and heterozygous cells (Supplementary Figure 3B). However, sections of Pfn3-deficient testes showed that the proacrosomal vesicle fail to attach in the middle of the acroplaxome and did not form a dense giant vesicle (Supplementary Figure 3C). In the cap phase, the proacrosomal granule starts to develop and flattens over the nucleus, forming the acrosomal cap, displayed in WT (Supplementary Figure 3D) and heterozygous spermatids (Supplementary Figure 3E). However, in Pfn3-deficient testes, the acrosomal vesicle fails to form a continuous cap-like structure and present as a detached granule from the acroplaxome (Supplementary Figure 3F). In the acrosome phase, the cap continues to develop as an arrow-like cover spanning the anterior two-third of the nucleus as indicated in WT (Supplementary Figure 3G) and heterozygous spermatids (Supplementary Figure 3H). Finally, at the end of maturation phase, acrosome formation is completed (Supplementary Figures 3I,J). In Pfn3-deficient sperm, during the acrosomal and maturation phases, the acrosomal granule fails to develop further (Supplementary Figures 3K,L).

In order to see the acrosomal defect in mature sperm, we next analyzed epididymal sperm cells of Pfn3-deficient males using TEM and immunofluorescence staining (PNA-FITC). An ultrastructural analysis and acrosomal labelling revealed malformation of the acrosomal region in epididymal sperms of Pfn3^–/– mice. Sperm cells present with elongating projections (highlighted by arrows), detached acrosome, abnormal acrosomal covering, and impaired removal of cytoplasm (Figure 4J). Similarly, mature sperm stained with PNA showed malformation of the acrosome in addition to abnormal sperm head morphology (Figure 4K). Imaging of 200 spermatozoa revealed that, in Pfn3^–/– male mice, 51–56% of spermatozoa display malformed acrosome compared to heterozygous (18%) and WT (16%) spermatozoa (Figures 4L,M).

As a consequence of the malformed acrosome, we reasoned that the acrosomal reaction (AR) could be impaired. We used the A23187 to induce AR. Acrosomal status was evaluated using Coomassie staining; intact acrosomes were stained dark blue as a crescent-like shape on the top of the sperm head (Figure 5A, white arrows), whereas acrosome-reacted sperm showed that the crescent-like shape (acrosome) on the top of the sperm head was not present (Figure 5A, white arrow heads). The rate of A23187-induced AR was significantly reduced in Pfn3-deficient sperm. Upon exposure to A23187, more than 60% of sperm from Pfn3^+/+ mice and ∼60% of sperm from Pfn3^+/– mice (Figure 5B) underwent acrosome exocytosis, whereas the AR occurred only in 37.5% of sperm from Pfn3^–/– mice (Figure 5B). This result indicates that the observed acrosome malformation impairs the acrosome exocytosis.

In order to investigate the underlying mechanism of impaired acrosome biogenesis in Pfn3-deficient mice, we performed immunofluorescence (IF) staining by using GM130 and TGN46 antibodies, markers for cis- and trans-Golgi network, respectively. GM130 plays a crucial role in vesicle tethering, fusion, and maintaining cis-Golgi structural integrity (Tiwari et al., 2019), while TGN46 is important for formation of exocytic vesicles and secretion from the trans-part of the Golgi network (Huang et al., 2019). IF revealed cis- and trans-Golgi predominantly concentrated at one pole of the Pfn3^+/+ (Figures 6A,D) and Pfn3^+/– (Figures 6B,E) spermatids, whereas the Pfn3^–/– spermatids showed defects and disorganization in cis- (Figure 6C) and trans-Golgi network (Figure 6F). These results indicate that loss of Pfn3 leads to disruption of the Golgi sub-domains causing defects in Golgi-derived proacrosomal vesicles leading to acrosome malformations.

Berruti and Paiardi (2015) published that in addition to the Golgi-derived biosynthetic pathway, the endocytic pathway contributes to acrosome biogenesis. In order to check whether loss of PFN3 affects the endocytic pathway, we performed IHC staining using anti-Rab5 antibody on testis sections for all three genotypes. Rab5 is a marker for early endosomes and a key factor in early endosome transport. The Rab5-mediated endo-lysosomal trafficking pathway is responsible for maturation of early endosomes to late endosomes. Interestingly, we did not observe any difference in the Rab5 staining of Pfn3^–/– testis sections as compared to the Pfn3^+/– and Pfn3^+/+ testis sections (Supplementary Figure 4). This result suggests that in Pfn3-deficient mice, the endocytic pathway is not affected.

Since the function of PFN3 is only beginning to be understood, we set out to identify whether deletion of Pfn3 impinges on global gene expression and performed RNA-sequencing on total RNA isolated from testes of Pfn3^+/+, Pfn3^+/–, and Pfn3^–/– mice. RNA-sequencing identified 38 significantly differentially expressed (DE) genes with log2 fold change > 1.5 in Pfn3^–/– as compared to control (Figure 7A). In total, 27 genes were found to be upregulated and 11 genes to be downregulated in Pfn3^–/– testes compared to wild type (Figure 7B). In order to check that the results observed are not skewed due to a defect in spermiogenesis in the PFN3-deficient mice, we tested for expression levels of marker genes indicative for Leydig cells, Sertoli cells, and spermatogonia. Box plots of marker genes for the different cell types are given in the Supplementary Data (Supplementary Figure 5) and reveal that the overall quantity and development of sperm cells are not affected in Pfn3^–/– mice. Tukey’s multiple-comparison ANOVA was used to check for statistics, and all groups showed non-significant difference. Quantitative real-time PCR was used to validate the results for 13 of the DE genes, which are related to male fertility. Cfl1, Trim27, B3gnt9, Mul1, and Hhatl were upregulated, while Exosc1, Srsf9, Atg2a, Copa, Slc25a36, Sap30, Dpm2, and Qrich1 were downregulated in Pfn3^–/– mice (Figure 7C). These results indicated that deletion of Pfn3 disrupts the expression of genes involved in actin cytoskeletal dynamics, regulation of autophagy, and mitochondrial and Golgi network structural integrity.

The RNA-seq analysis revealed Trim27 as the most upregulated gene in Pfn3-deficient mice. The overexpression of Trim27 leads to activation of AKT signaling (Zhang et al., 2018). AKT in turn is an activator of mTOR (Hahn-Windgassen et al., 2005). Autophagy is regulated by AMPK/mTOR signaling pathways, with AMPK being stimulating and mTOR being repressive (Liang et al., 2018). Next, we were interested to determine whether the signaling pathways downstream of Trim27 responsible for regulation of autophagy are affected in Pfn3-deficient mice. Indeed, immunoblotting showed increased protein levels for mTOR and phospho-mTOR, while the level of phospho-AMPKα was reduced (Figure 8A). Increased level of phospho-mTOR was validated by IHC in Pfn3-deficient mice (Figure 8B). So, these data suggest that loss of Pfn3 leads to an upregulation of Trim27, which results in the activation of mTOR signaling (Hahn-Windgassen et al., 2005; Liang et al., 2018; Zhang et al., 2018), causing a decrease in p-AMPKα resulting in the disruption of autophagy. Furthermore, the autophagic gene Atg2a was expressed at lower levels in Pfn3-deficient mice. Atg2a is involved in the phagophore elongation leading to the formation of the autophagosome (Bozic et al., 2020). The elongation step is completed by the conjugation of LC3B, known as microtubule-associated protein and a widely used marker for autophagosomes (Tang et al., 2017). LC3B is a core protein in the autophagic flux where it functions as an autophagic cargo by interacting with an autophagic substrate SQSTM1/p62 (Tang et al., 2017). It is well established that depletion of Atg2a results in blocking of autophagic flux leading to accumulation of LC3B and SQSTM1 (Bozic et al., 2020). Therefore, we analyzed LC3B and SQSTM1 protein levels. Interestingly, in Pfn3-deficient testes, levels of LC3B were increased as shown by WB and IHC (Figures 8A,C, respectively). We further found an accumulation of SQSTM1 in Pfn3-deficient testes using WB (Figure 8A). Quantification of protein levels (Supplementary Figure 6) revealed an increase in LC3B, P62, p-mTOR, and mTOR levels and reduction of ATG2a and AMPK. So, we hypothesized that deletion of Pfn3 results in upregulation of Trim27, which leads to mTOR-mediated inhibition of autophagy hallmarked by lower levels of Atg2a. As a consequence, autophagic flux stalls, indicated by accumulation of LC3B and SQSTM1. This might cause the disturbance of the acrosome formation in Pfn3-deficient mice.

Next, we used co-immunoprecipitation (Co-IP) to test whether PFN3 and TRIM27 interact. From whole-testis lysate, a PFN3-specific antibody pulled down TRIM27 (Figure 9A, lane 3) and the IP with a TRIM27 antibody was able to capture and elute PFN3 (not shown). The specificity of the assay was confirmed by using protein extraction buffer as a negative control (Figure 9A, lane 2); whole-testis lysate was used as an input control (Figure 9A, lane 1). To validate this interaction, reciprocal antibodies on immunoblot (Figure 9B) showed PFN3 (∼14 kDa) and TRIM27 (∼58 kDa) proteins from input control (lane1), Co-IP (lane 2), flow through (lane 3), and negative control.

Furthermore, the expression of cofilin1 (Cfl1) was upregulated in Pfn3-deficient testes. We decided to check protein levels using WB for the cofilin traditional proteins (CFL1, CFL2, and ADF) known as actin-binding proteins. The level of ADF protein was increased in Pfn3^–/– testes, while CFL1 and CFL2 (Figure 10) protein levels were already increased in heterozygous testes.

PFN3 is detected in a complex with ARPM1, specifically in the sperm nucleus, however, not in the cytoplasm (Hara et al., 2008). In sperm cytoplasm of Pfn3^–/– mice, a moderate signal of ARPM1 can still be detected (Supplementary Figure 7A). We performed WB using anti-ARPM1antibody on proteins isolated from cytoplasmic and nuclear fractions of both testes and sperm. Western blot showed that ARPM1 could not be detected in the nuclear fraction of Pfn3-deficient testes and sperm, while cytoplasmic ARPM1 protein levels in testes are slightly reduced in Pfn3-deficient mice (Supplementary Figure 7A).

These findings were further confirmed by immunohistochemistry; the nucleus of spermatozoa in testes of Pfn3-deficient mice were devoid of ARPM1 (Supplementary Figure 7B). This finding suggests that loss of PFN3 destabilizes the PFN3–ARPM1 complex, leading to loss of ARPM1 in the nuclei of Pfn3-deficient sperm.

Pfn3 binds to actin monomers and plays a role in actin polymerization. Actin is mainly located in the mid-piece of sperm flagellum, suggesting a role in sperm motility, elasticity, and membrane integrity. F-actin is present as a helical structure (Gervasi et al., 2018), which is in parallel with the organization of the mitochondrial sheath in mouse sperm (Amaral et al., 2013). We wanted to see whether the Pfn3 deletion affects the actin cytoskeleton organization in the sperms. Phalloidin ATTO647 fluorescence staining revealed that actin assembly was not altered in the mid-piece of Pfn3^–/– sperm (Supplementary Figure 8).

We found that PFN3 localized in the manchette complex of the developing spermatid. The manchette is a transient structure on the posterior part of the sperm head and is involved in nuclear shaping and protein transport for sperm flagellum formation. The manchette is first detected in step 8, when round spermatids start to elongate by nuclear polarization and the nuclear shape changes from spherical to slightly elongated. After shaping the sperm head, the manchette disappears at step 13 (Okuda et al., 2017).

In order to investigate whether the abnormal head morphology of Pfn3-deficient sperm was due to the alteration in manchette structure or formation, we performed an immunofluorescence staining using alpha tubulin for step-by-step comparison on the germ cell population isolated from testes. Tubulin staining revealed that in WT and heterozygous spermatids (Figures 11A,B), manchette was forming the proper skirt-like structure while the manchette was not properly covering and constricted overly at the posterior region in Pfn3-deficient round spermatids (Figure 11C). This abnormal manchette development is more obvious in the later steps of development (steps 9–13) in Pfn3-deficient spermatids. We conclude that deletion of PFN3 leads to disruption of manchette formation, which in turn contributes to the abnormal shape of the sperm head in Pfn3-deficient sperm.

Since we detected low sperm motility in Pfn3-deficient mice, we analyzed sperm flagella structure. We used Mito Red immunostaining to assess the mitochondrial sheath in the mid-piece of sperm flagella. While Mito Red uniformly stained the mitochondrial sheath in spermatozoa of WT (Figure 12A) and heterozygous (Figure 12B), Pfn3-deficient sperm flagella showed a variety of aberrations spanning from abnormally thick mid-piece (Figure 12C), slightly tapered bent thick neck (Figure 12D), and cytoplasmic droplets with amorphous sperm heads (Figure 12E). This suggests that loss of PFN3 resulted in flagellar deformities due to mitochondrial disorganization.

Next, we performed TEM on mature sperm isolated from cauda epididymis to analyze whether the mid-piece of sperm have normal axonemal and mitochondrial structure. TEM of Pfn3-deficient sperm showed several ultrastructural defects, like plasma membrane not covering uniformly the mitochondrial sheet (Figure 12H), disorganized or vacuolated mitochondria (Figure 12I, asterisk), axonemal fibrous sheet enclosed two or more mitochondria and axonemal flagellar complex (Figure 12J) compared to WT (Figure 12F), and heterozygous sperm ultrastructure (Figure 12G). Scanning electron microscopy (SEM) showed similar sperm abnormalities in Pfn3-deficient mice, thick sperm mid-pieces with distal cytoplasmic droplets (Figures 12M,N) compared to WT (Figure 12K) and heterozygous sperm (Figure 12L). This result correlates with the previous (Mito Red and TEM ultrastructural sperm analysis) findings. These findings showed that sperm from Pfn3-deficient mice display an abnormal morphology of mitochondrial and axonemal fibrous sheet. Statistical analysis revealed that the percentage of deformed sperm flagella in Pfn3-deficient mice (∼55%) compared to WT and heterozygous sperms was significantly higher (Figure 12O). We speculate that these defects contribute to the reduced sperm motility. Altogether, our results indicate that loss of PFN3 located in mitochondria resulted in sperm flagellar defects and further support a cytoplasm removal defect.