Bioinformatic analysis to identify putative E3 ubiquitin-protein ligases
We used multiple bioinformatics and gene expression approaches to uncover testis-enriched genes that are components of the ubiquitin–proteasome pathway. We searched in-house and published RNAseq data lists [12] and identified four putative RING family E3 ubiquitin ligases (RNF133, RNF148, RNF151, and ZSWIM2). To confirm the expression profiles of Rnf133, Rnf148, Rnf151, and Zswim2, we performed RT-PCR of mouse and human reproductive and non-reproductive tissues, confirming that all four E3 ubiquitin ligases are testis-specific in mice and humans (Fig. 1A left and right panels, Additional file 1: Table S1). To glean insight into the potential spermatogenic cell population(s) expressing Rnf133, Rnf148, Rnf151, and Zswim2, we performed RT-PCR of mouse testes isolated at postnatal day (P) 5, a timepoint enriched for gonocytes transitioning to Type A spermatogonia, P10 (early, Leptotene, spermatocytes), P15 (late, Pachytene, spermatocytes), P20 (early, round, spermatids), and P30, 35, 42, and 60, which all display fully mature elongated spermatids, and either one complete wave, or multiple waves, of spermatogenesis [13]. As shown in Fig. 1A (center panel), Rnf133 and Rnf148 are expressed at day 25, corresponding to the period when round spermatids are transitioning to elongating spermatids, Rnf151 is expressed at day 20, at the round spermatid stage, and Zswim2 is expressed abundantly at day 15, towards the end of meiosis. Based on these results, we suspect that all four genes could function during spermiogenesis or in sperm formation and/or for sperm function. Considering the correspondence of the expression pattern of Rnf133 and Rnf148 in mouse, we compared the domain structures of those two proteins in mouse and human. In silico prediction reveals that both RNF133 and RNF148 have one transmembrane region and one RING finger domain after the transmembrane region and localized in the cytoplasm (Fig. 1B). Typically, E3 transmembrane proteins are localized to the ER, which our immunostaining for recombinantly expressed human RNF133 confirms (Additional file 2: Fig. S1) [8]. In addition, RNF133 and RNF148 share 58.9% and 54.9% identity in mouse and human, respectively, based on pairwise sequence analysis (Fig. 1C).
Generation of testis-specific ubiquitin ligase gene knockout mice
We generated knockout mice using the CRISPR/Cas9 system to examine the function of these proteins in vivo. The two gRNAs for each gene were designed as shown in Fig. 1D and Additional file 3, 4, 5, 6: Figs. S2-4, Table S2. One-cell stage embryos were electroporated with Cas9 protein and sgRNAs that were designed to produce large deletions of the coding region of each gene, and the two-cell stage embryos were then transferred into the oviducts of pseudopregnant females. For each gene, the electroporation resulted in deletions of more than 50% of the open reading frame (ORF) in Rnf133, Rnf148, Rnf151, and Zswim2 (Fig. 1D and Additional file 3, 4, 5: Figs. S2-4), which leads to loss of their functional zinc finger domains. After the deletions in each gene were confirmed by direct sequencing analysis, specific primers for the wild-type (WT) or KO allele were designed and used for genotyping (Fig. 1D, E and Additional file 3, 4, 5: Figs. S2-4). The KO mice containing mutations in each gene did not show any obvious developmental abnormalities or differences in sexual behavior.
Because the Rnf133 and Rnf148 genes were closely linked on mouse chromosome 6 by 11.01 centimorgan and structurally similar (Fig. 1B), we could not interbreed single mutants. To generate double KO (DKO) of Rnf133 and Rnf148, Rnf133 KO female, and Rnf133 heterozygous (HET) male mice were subjected to in vitro fertilization (IVF), and the zygotes from this IVF were electroporated with gRNAs, to target the Rnf148 locus. The founder female mice, whose genotype showed Rnf133 KO and a confirmed deletion in the Rnf148 gene, were intercrossed to obtain subsequent generations of Rnf133/Rnf148 DKO mice. To generate Rnf148/Rnf151 DKO and Rnf148/Rnf151/Zswim2 triple KO (TKO) mice, mutant mice of each gene were intercrossed to obtain the required mutant alleles.
Rnf133 KO and Rnf133/Rnf148 DKO males show subfertility, while Rnf148, Rnf151, and Zswim2 KO males retain their fertility
To assess the functions of RNF133 in vivo, sexually mature Rnf133 HET or KO males were housed with two WT females for 4 months. The average number of offspring per litter was counted. Five Rnf133 heterozygous mutant mating pairs had 8.6 ± 0.7 pups per litter on average, whereas five homozygous KO males showed 2.3 ± 1.6 pups per litter on average (Fig. 2A). These data demonstrate that RNF133 has an important role in male fertility. Meanwhile, the Rnf148 KO, Rnf151 KO, Zswim2 KO, Rnf148/Rnf151 DKO, and Rnf148/Rnf151/Zswim2 TKO males showed comparable litter size with control male mice (Additional file 7: Fig. S5).
To further evaluate the subfertility of Rnf133 KO males, timed-matings with WT females were conducted. We counted 15 (HET) or 20 (KO) copulation plugs in total. Rnf133 HET males produced a successful pregnancy rate (litters/copulation plugs) for 100 ± 0% of copulation plugs, while Rnf133 KO males had only 10.0 ± 10% pregnancy success (Fig. 2A). The Rnf133 KO males also fathered significantly fewer pups/plug (0.15 ± 0.15) compared with HET males (8.4 ± 0.24) (Fig. 2A). This result indicates Rnf133 KO males’ sperm are less likely to successfully fertilize oocytes, resulting in a failed pregnancy and dramatically reduced fecundity per coitus.
Based on the similarity of RNF133 and RNF148 proteins, we hypothesized that RNF148 compensates in the absence of RNF133 in vivo. Mating analysis using Rnf133/Rnf148 DKO male mice revealed that their subfertility remained at a similar level as Rnf133 single KO males (2.5 ± 0.15 pups, Fig. 2A right). In contrast to Rnf133 KO and Rnf133/Rnf148 DKO males, Rnf148, Rnf151, or Zswim2 single KO, Rnf148/Rnf151 DKO, and Rnf148/Rnf151/Zswim2 TKO males showed similar fertility as control males (Additional file 7: Fig. S5).
To characterize the etiology of subfertility of Rnf133 KO males and Rnf133/Rnf148 DKO male mice, we analyzed testes from Rnf133 HET, KO, and Rnf133/Rnf148 DKO mice. Histologically, we found that spermatogenesis was grossly normal in Rnf133 KO and Rnf133/Rnf148 DKO mice in comparison to heterozygous controls (Fig. 2B and Additional file 8: Fig. S6A). The testes of the Rnf133 KO and the Rnf133/Rnf148 DKO were nearly identical to each other and showed a significant increase in sperm nuclei surrounded by excess cytoplasm at stage IX, a phenomenon that we observed considerably less in the control (HET/double HET(DHET)) testes (Fig. 2C). We quantified this occurrence and found that the percentage of sperm nuclei surrounded by excess cytoplasm at stage IX, out of all tubules at stage IX, was 20.5% in the control versus 80.6% in the Rnf133 KO and 89.2% in the Rnf133/Rnf148 DKO.
To define the cause of the fertility defects in the Rnf133 KO mice, we performed IVF with sperm from Rnf133 HET and KO mice and oocytes from WT female mice (Fig. 3A). In contrast to sperm from Rnf133 HET controls which fertilized 83.9 ± 8.2% of oocytes and resulted in 73.1 ± 10.0% 2-cell stage embryos (2cell), the sperm from Rnf133 KO mice only fertilized 6.8 ± 3.0% of oocytes and resulted in 1.4 ± 0.7% 2-cell stage embryos. Similar defects were also observed for in vivo fertilization with WT female mice in which the fertilization rate by Rnf133 KO sperm was significantly lower than corresponding controls: 17.2 ± 4.1% pronuclei positive (PN +) and 18.4 ± 4.4% 2-cell stage embryos (2cell) from Rnf133 KO sperm, versus 73.9 ± 11.7% PN + and 69.2 ± 11.3% 2cell from Rnf133 HET sperm (Fig. 3B). To further understand the origin of this fertilization defect, we examined the number and motility of sperm-derived from the cauda epididymis. The sperm counts from the cauda epididymis between Rnf133 HET and KO mice showed no significant differences (Fig. 3C). However, when we examined cauda epididymal sperm using computer-assisted sperm analysis (CASA), the percentage of motile sperm and amount of progressive motility were significantly decreased in Rnf133 KO mice after 15 min incubation (21.7 ± 3.5% and 14.5 ± 2.7%, respectively) compared to HET (41.1 ± 6.4% and 38.4 ± 6.2%, respectively) (Fig. 3D, E). After 120 min incubation, Rnf133 KO sperm showed no difference in total motile sperm or progressive motility (19.1 ± 5.5% and 11.0 ± 4.1%, respectively) in comparison to Rnf133 HET sperm (21.2 ± 4.2% and 14.2 ± 3.1%, respectively). The velocity parameters (average path velocity: VAP, curvilinear velocity: VCL, straight-line velocity: VSL) were impaired in Rnf133 KO sperm. The decrease in VAP (HET, 167.1 ± 7.7 vs. KO, 131.7 ± 9.3) and VSL (HET, 153.1 ± 7.3 vs. KO, 111.4 ± 9.9) at 15 min of incubation, and VAP (HET, 129.6 ± 3.9 vs. KO, 101.0 ± 9.2) and VCL (HET, 222.0 ± 11.2 vs. KO, 177.9 ± 15.3) at 120 min of incubation were significantly different (Fig. 3F). Likewise, although Rnf133/Rnf148 DKO mice showed no significant difference from Rnf133/Rnf148 double HET (DHET) in testicular size, weight, and caudal epididymal sperm, we found significant decreases in the percentage of motile sperm (15 min, DHET, 64.6 ± 4.5% vs. DKO, 36.9 ± 2.1%; 120 min, DHET, 43.9 ± 2.6%, DKO, 24.8 ± 5.3%) and progressive motility (15 min, DHET, 60.1 ± 4.6% vs. DKO, 30.9 ± 2.4%; 120 min, DHET, 25.0 ± 3.3%, DKO, 13.0 ± 3.9%) (Additional file 8: Fig. S6A-E).
To evaluate the rate of sperm hyperactivation, which is a critical factor for fertilization success, we utilized the machine learning algorithm CASAnova [14]. The percent of hyperactivated motile Rnf133 HET sperm incubated in capacitation medium increased ~ tenfold, from 0.5 ± 0.3% at 15 min to 5.2 ± 1.7% after 120 min, while Rnf133 KO sperm did not show a dramatic change in hyperactivation (0.7 ± 0.4% at 15 min and 1.3 ± 0.4% at 120 min) (Fig. 3G). Likewise, the hyperactivation rate between DHET and Rnf133/Rnf148 DKO sperm was not statistically different at 15 min (DHET, 0.3 ± 0.1% vs. DKO, 0.3 ± 0.2%) but a difference was observed at 120 min (DHET, 3.8 ± 0.8%, DKO, 1.6 ± 0.3%) (Fig. S6F). There was no significant difference between the sperm hyperactivation rates of the Rnf133 KO and the Rnf133/Rnf148 DKO mice (P = 0.69 at 15 min, P = 0.24 at 120 min). Our discovery of compromised motility of sperm and the impaired hyperactivation of sperm from Rnf133 KO and Rnf133/Rnf148 DKO males explains the in vitro and in vivo defects in fertilization and the subfertility in the Rnf133 KO and Rnf133/Rnf148 DKO males. Consistent with no significant difference in fertility (Additional file 7: Fig. S5), sperm from the Rnf148 KO, Rnf151 KO, and Zswim2 KO did not show any significant differences in terms of testis weight and sperm parameters in comparison to controls, except VSL at 120 min for Rnf148 KOs, and VAP and VSL at 120 min for Rnf148/Rnf151 DKOs. However, as mentioned, this did not affect their fertility in natural mating with WT female mice (Additional file 7, 9, 10, 11, 12: Fig. S5, S7-S10).
Rnf133 KO results in aberrant head-neck morphology
To characterize the cause of the impaired motility in Rnf133 KO males, we analyzed the waveform of the flagella of sperm (Fig. 4A). Tracing of sperm flagella after 15-min incubation revealed that Rnf133 KO and Rnf133/148 DKO males had sperm cells with aberrant morphologies such as bending at the head/midpiece and bending at the midpiece/principal piece (Fig. 4B), which are rarely observed in Rnf133 littermate controls. Since we observed a significant decrease in total motile and progressively motile sperm in Rnf133 KO and Rnf133/148 DKO in comparison to controls (Fig. 3D, E and Additional file 8: Fig. S6C, D), we next examined the relationship between aberrant morphology and change in motility. Indeed, when morphology and motility were co-quantified in Rnf133 KO and Rnf133/Rnf148 DKO sperm in comparison to controls, we found statistically significant relationships for most subsets of sperm. As shown in Fig. 4C, there was a statistically significant 1.9- and 2.5-fold decrease in the percentage of normal morphology motile sperm in Rnf133 KO (50 ± 4.6%) and Rnf133/148 DKO (38 ± 4.8%) in comparison to controls (93.7 ± 1.6%). Additionally, a statistically significant 10.1-fold and 12.8-fold increase in the percentage of normal morphology immotile sperm was observed in KO (13.9 ± 3.4%) and DKOs (17.7 ± 3.3%) in comparison to controls (1.4 ± 0.8%), and a statistically significant 67-fold and 82-fold increase in the percentage of abnormal morphology immotile sperm in KOs (25.9 ± 6.6%) and DKOs (31.9 ± 4.3%) in comparison to controls (0.4 ± 0.3%) (all P < 0.001, as calculated through one-way ANOVA). To confirm that the increased immotility does not result from an increase in sperm cell death, we performed a live/dead cell assay and found no significant difference between Rnf133 KO and Rnf133/Rnf148 DKO sperm in comparison to controls (Fig. 4D,E and Additional file 8: Fig. S6G-I). These data suggest that the aberrant shape of Rnf133 KO and Rnf133/148 DKO sperm is the major cause of the reduced motility and hyperactivation in these mutant sperm, although other additional defects such as abnormal sperm head morphology may also be possible.
RNF133 is an E3 ligase that interacts with the ER-localized E2 enzyme UBE2J1
E3 ubiquitin ligases function with E2 ubiquitin-conjugating enzymes in the ubiquitylation process. To identify the E2 protein that interacts with RNF133, we first conducted in silico analysis using our recently published RNAseq data (Fig. 5A,B) [12]. Based on our findings, we decided to focus on Ube2g2, Ube2j1, and Ube2j2, which are all conserved in mouse. As seen in Fig. 5A, B, Ube2g2 Ube2j1, and Ube2j2 are all expressed in testis and testicular germ cells, with Ube2j1 showing the highest levels of expression in testis and testicular germ cells in both humans and mice (Fig. 5A,B). Moreover, UBE2J1 is an ER-localized transmembrane protein and has been reported to have a critical role in ER quality control during spermiogenesis, with Ube2j1 KO mice displaying a similar phenotype to our Rnf133 KO and Rnf133/Rnf148 DKO [9]. Thus, we hypothesized that RNF133 interacts with UBE2J1. Using HEK293 cells, FLAG-tagged mouse or human RNF133 and HA-tagged mouse or human UBE2J1 were co-expressed and subjected to immunoprecipitation analysis to detect their complexes (Fig. 6A and Additional file 13, 14: Fig. S11, S12). Total proteins were extracted from transfected cells, and each protein sample was pulled down using anti-HA beads. Input lysate and pull-down samples were subjected to SDS-PAGE, and immunoblots were incubated with anti-FLAG antibody. We confirmed robust expression of FLAG and HA in each sample transfected with the vectors encoding RNF133-FLAG and/or HA-UBE2J1 (Fig. 6A and Additional file 13, 14: Fig. S11, S12) and found both mouse and human FLAG-tagged RNF133 interact with HA-tagged UBE2J1 (Fig. 6A, and Additional file 13, 14: Fig. S11, S12). Interestingly, although we observed interaction between HA-UBE2C and RNF133-FLAG (Fig. 6A and Additional file 13, 14: Fig. S11, S12), interaction between HA-UBE2J/RNF151-FLAG was not detected (Fig. 6A and Additional file 13, 14: Fig. S11, S12). To verify that the robust interaction between RNF133 and UBE2J1 is not an artifact, we created a human RNF133 construct in which the RING finger domain was deleted (RNF133-mutant) and replicated the co-immunoprecipitation (IP) assay using this human RNF133-mutant and UBE2J1. Although there was a faint interaction between RNF133-mutant and UBE2J1, possibly due to both proteins localizing at the ER membrane and/or because other regions of RNF133 and UBE2J1 interact normally, the intensity of the signal in the immunoprecipitated sample with UBE2J1 was reduced in RNF133-mutant compared to RNF133 wild-type. As a numerical analysis, we calculated a ratio based on the band intensity of the IP lane to the input lane for each RNF133:UBE2J1 interaction and confirmed that human RNF133-WT:UBE2J1 or mouse RNF133-WT:UBE2J1 had an ~ sevenfold or ~ 16-fold IP:input ratio, respectively, compared to human RNF133-mutant:UBE2J1 IP:input ratio (human RNF133-WT, 1.74 vs. mouse RNF133-WT, 3.95 vs. RNF133-mutant, 0.24) (Fig. 6A and Additional file 15: Fig. S13).
In parallel, transfected HEK293 cells were stained with the anti-FLAG antibody for RNF133, and with the anti-HA antibody for UBE2J1. We confirmed the co-localization of both human and mouse RNF133 and UBE2J1 in the ER in vitro (Fig. 6B). These data strongly support the interaction of RNF133 and UBE2J1 during spermiogenesis in mice and humans.
Electron microscopic analysis of Rnf133 KO sperm resembles the Ube2J1 KO phenotype
Since we discovered abnormal sperm morphology with bent neck/flagella in Rnf133 KO mice and confirmed RNF133 interaction with UBE2J1 in vitro, we performed transmission electron microscope (TEM) analysis of epididymal tissue to assess the morphology of KO sperm at the ultrastructural level. Observation of caudal epididymis from Rnf133 KO mice revealed that although Rnf133 KO sperm had morphologically normal nuclei, the cytoplasm around the nuclei and neck was retained and contained organelles such as the endoplasmic reticulum and vacuoles (Fig. 7) Similar abnormalities were shown previously in Ube2j1 KO sperm [9]. These observations robustly suggest that RNF133 and UBE2J1 function together in a ubiquitination pathway to degrade and discard unnecessary proteins from the ER.