New Factor Found to Enhance Sperm Fertilization

Sperm exit the testis, acquire fertilizing ability during the epididymal transition, and subsequently fertilize oocytes in the oviductal ampulla of the female reproductive tract. For fertilization to occur, ejaculated sperm must have the capability to migrate from the uterus into the oviduct, release sperm acrosome enzymes, and alter their motility pattern. During the last stages of the fertilization process, sperm bind to the zona pellucida (ZP), the glycoprotein layer surrounding the eggs. The factors responsible for sperm-ZP binding have been a constant search since Lillie first visualized the interaction between sperm and oocyte components more than 100 years ago1.

Mouse ZP is composed of three heavily glycosylated proteins called ZP1, ZP2, and ZP32,3. ZP3 appears to be more important than ZP1 and ZP2 for sperm-ZP binding because ZP3 protein purified from unfertilized eggs inhibits sperm-ZP binding4. Because oocytes from Zp3 knockout (KO) mice lack a ZP5, the physiological functions of ZP3 by itself cannot be analyzed. Further, after digestion of ZP3 with Pronase (a mixture of proteases), the small ZP3 glycopeptides continue to inhibit sperm-ZP binding6, indicating that there is a sperm protein that binds the ZP3 protein-carbohydrate unit. ZP3 contains both N-linked and O-linked oligosaccharides7,8,9. The removal of all N-glycans using Endo-β-N-acetyl-D-glucosaminidase F (Endo-F) and the disruption of all branched antennae that contain galactose, fucose, sialic acid, and N-acetyl glucosamine residues on N-glycans using the mannoside acetylglucosaminyltransferase 1 (Mgat1) KO mice did not affect sperm-ZP binding or fertilization10,11, while the removal of O-glycans from ZP using alkaline β-elimination decreased the number of sperm bound to the ZP10. These results suggest that O-linked oligosaccharides on the ZP are more critical for sperm-ZP binding than N-linked oligosaccharides.

O-linked oligosaccharides in mammals are classified predominantly into four core glycan structures (cores 1–4). So far, core-3-derived and core-4-derived O-glycans have not been detected by mass spectrometry analysis of mouse ZP312,13,14. The physiological functions of core-1-derived and core-2-derived O-glycans for sperm binding have been analyzed using genetically modified mice, but the deletion of T-syn (core 1 β1,3-galactosyltransferase 1) and N-acetylglucosaminyltransferase (C2GnT-L and Gcnt1), which are required for the syntheses of core-1 and core-2-derived O-glycans, failed to affect female fertility and sperm binding number15,16. Thus, the roles of O-linked oligosaccharides on the ZP in sperm binding remain unknown.

Some sperm proteins, such as acrosin and PH-20 (also known as SPAM1), have been suggested as ZP3 receptor candidates using in vitro experiments17. Of these, β1,4-galactosyltransferase (known as GalT and β4GalT-1) was initially the most promising ZP3 receptor because GalT-bound ZP3 O-linked oligosaccharides caused the subsequent acrosome reaction through the activated signaling cascades of pertussis toxin-sensitive heterotrimeric guanine nucleotide-binding protein18,19; however, later studies revealed that sperm lacking GalT could bind to ZP and fertilize eggs20,21. SP56 (also called ZP3 receptor) was another ZP3 receptor candidate, as its alternative name implies22,23,24; however, Sp56 KO sperm could bind to ZP and KO male mice were fertile25, ruling it out as the sole ZP3 receptor. Thus, the candidates suggested by in vitro experiments were mostly dispensable for sperm-ZP binding. However, our team previously identified Calmegin (Clgn) as a testis-specific chaperone, essential for sperm-ZP binding26. Further studies revealed that CLGN is required for the heterodimerization of a disintegrin and metallopeptidase domain (ADAM) 1a and ADAM2, as well as the subsequent maturation of ADAM327,28. Sperm lacking ZP binding ability also exhibit impaired migration through the utero-tubal junction (UTJ)29, indicating that shared mechanisms exist for sperm-ZP binding and UTJ migration. So far, about 20 genes abundantly expressed in the testis and epididymis, such as a testicular isoform of angiotensin-converting enzyme (t-Ace) and Adam3, have been identified as essential factors for both sperm-ZP binding and UTJ migration27,30,31,32,33. Sperm with a knockout of most of these genes have ADAM3 loss, but there are exceptions. For example, our teams reported that ADAM3 persists in sperm lacking post GPI attachment to proteins 1 (Pgap1), lymphocyte antigen 6 complex, locus K (Ly6k), or Ly6/Plaur domain containing 4 (Lypd4), despite the findings that these KO sperm show a UTJ migration defect and impaired ZP binding34,35,36. Specifically, PGAP1 has a function as a GPI inositol-deacylase that removes the palmitate from inositol in the endoplasmic reticulum, suggesting that it is unlikely that PGAP1 is directly involved in the sperm migration into the UTJ and sperm-ZP binding37,38. LY6K, a GPI-anchored protein, interacts with testis-expressed gene 101 (TEX101), which is a substrate of t-ACE in testicular germ cells35,39. The role of LYPD4 in testicular germ cells and sperm in UTJ migration and ZP binding remains unclear, but this protein is present in Adam3 KO sperm36. Thus, we conclude that other sperm factors, rather than ADAM3, which is not conserved in humans, are responsible for sperm-ZP binding and sperm migration through the UTJ.

UDP-N-acetyl-alpha-D-galactosamine: polypeptide N-acetylgalactosaminyltransferase-like 5 (GALNTL5) (also known as pp-GalNAc-T1940, NCBI Reference Sequence: NP_660335.2) was identified as one of the N-acetylgalactosamine (GalNAc) transferase-like proteins required for the synthesis of GalNAc-linked glycans (also known as mucin-type O-glycans)40. Later studies revealed that GALNTL5 is abundantly expressed in human testis, and that GALNTL5 does not have GalNAc transferase enzyme activity in vitro41,42,43. Thus, GALNTL5 functions other than GalNAc transferase activity were analyzed using genetically modified mice; however, the Galntl5 KO phenotype published by Takasaki et al.42 is different from the International Mouse Phenotyping Consortium (IMPC) database. The IMPC database suggests that the male fertility of Galntl5 KO mice is comparable to the control, while Takasaki et al.42 published that a heterozygous null mutation of Galntl5 resulted in infertility due to impaired sperm motility and morphology. To resolve this enigma and establish the physiological functions of GALNTL5 in male fertility, we created four independent lines lacking mouse Galntl5 using CRISPR/Cas9. We reveal that not heterozygous but homozygous null mutations of Galntl5 are almost exclusively male sterile and that GALNTL5 regulates sperm migration into the UTJ and sperm-ZP binding through the binding with GalNAc in the UTJ and on the ZP. We conclude that sperm GALNTL5 is a long thought responsible factor for sperm-ZP binding and sperm migration into the UTJ.

The Galntl5 locus encodes two transcripts

We first examined Galntl5 mRNA expression by PCR using cDNAs from mouse multi-tissues. Two alternative splice variants are transcribed from the genomic sequence coding Galntl5 (Fig. 1a), so we used a primer set to amplify both variants for the multi-tissue expression analysis. Galntl5 mRNA was specifically expressed in the mouse testis (Fig. 1b), and its expression was detected 20 days after birth or later (Fig. 1c), indicating that Galntl5 mRNA begins to be expressed in secondary spermatocytes and later stages. Also, human GALNTL5 is predominantly detected in the testis (Fig. 1d), consistent with the result of the real-time PCR in the previous paper42.

Loss of GALNTL5 causes fertility defects in male mice

To examine the physiological functions of mouse GALNTL5, we generated Galntl5 mutant mice by injecting a guide RNA (gRNA)/Cas9 expressing plasmid into eggs to delete both alternative splice variants (Fig. 1a). By mating F0 mutants with wild-type (WT) mice, we established two mutant lines; enzyme mutation (em) 1 and em2 disrupt 17 and 25 nucleotides in exon 2 of the Galntl5 variants, respectively, leading to frameshift mutations (Fig. 1e and Supplementary Fig. 1a–c). By intercrossing of heterozygous (Het) mutants, we obtained Galntl5 homozygous mutant (KO) mice. To check the disruption of GALNTL5 proteins in KO mice, we performed western blot analysis using extracts of testicular germ cells (TGC) and sperm and antibodies to recognize both termini of GALNTL5 (Supplementary Fig. 1d). We detected the doublet bands corresponding with predicted molecular sizes of variants 1 and 2 (~50 kDa and ~46 kDa) in the TGC extract in the control group, while a single band at ~37 kDa in sperm extracts was detected in the control group only when we used the antibody to recognize the C-terminus (Fig. 1f and Supplementary Fig. 2), indicating that sperm GALNTL5 lacks the N-terminal region. In the extracts of the TGC and sperm from KO mice, these bands disappeared, indicating that the longer TGC variants and the shorter sperm variant of GALNTL5 are disrupted in KO mice (Fig. 1f and Supplementary Fig. 2). To examine the timing when GALNTL5 is processed from ~50 kDa to ~37 kDa, we collected testicular sperm (TS) and sperm from caput, corpus, and cauda epididymides. As shown in Supplementary Fig. 3, when we used the N-terminus antibody, we could not detect the longer forms in epididymal sperm but only in TGC and TS. Using the C-terminus antibody, we barely found a specific signal in the TS due to the poor reactivity of the antibody, but we could detect the longer form in TGC and the shorter form in sperm after the caput epididymis. Based on these data, we speculated that the longer form of GALNTL5 is cleaved after entering the caput epididymis by some protease(s).

To reveal the effects of disruption of GALNTL5 on male fertility, Het and KO male mice were mated with WT females. The fertility of Het males was comparable to WT males. However, females mated with KO males rarely delivered pups, indicating that Galntl5 KO males are nearly sterile [pups/plug: 8.2 ± 1.0 (WT), 9.7 ± 1.5 (em1 Het or em2 Het), 0.4 (em1 KO), 0.2 ± 0.2 (em2 KO)] (Fig. 1g and Supplementary Table 1). Thus, despite creating two independent null alleles, we did not observe haploinsufficiency in our heterozygous mutant mice, unlike a previous paper42.

Galntl5 KO sperm have a defect in passage through the UTJ

As there is no difference in male fertility between both KO mutant lines (Galntl5em1 and Galntl5em2 KO) (Fig. 1g), we used Galntl5em2 mutants for further experiments. To understand why Galntl5 KO males show severe subfertility and sterility, we examined spermatogenesis and sperm motility. The gross morphology and weight of Galntl5 KO testes were comparable to the control [testis weight/body weight: 3.2 ± 0.6 (Het), 3.2 ± 1.1 (KO)] (Fig. 2a, b). We observed spermatogenesis by microscopic observation using Hematoxylin and Periodic acid-Schiff staining, but we could not find an obvious abnormality (Fig. 2c). Furthermore, sperm morphology from Galntl5 Het and KO mice was comparable to WT sperm (Fig. 2d), and we could not detect any defects in sperm motility parameters (Fig. 2e), unlike a previous publication42.

To assess the sperm fertilizing ability, we incubated Galntl5 KO sperm with cumulus-intact and cumulus-free eggs. Galntl5 KO sperm efficiently fertilized cumulus-intact eggs [fertilization rates: 96.9 ± 2.7% (Het), 100% (KO)] (Fig. 3a), indicating that Galntl5 KO sperm show normal fertilizing ability in vitro. Furthermore, when these fertilized eggs were transferred into the oviducts of pseudo-pregnant females, the pups were delivered, similar to controls [control sperm: 86 pups/325 embryos (26%), Galntl5 KO sperm: 101 pups/202 embryos (50%)]. Thus, we conclude that the development of embryos fertilized with Galntl5 KO sperm is normal. However, Galntl5 KO sperm bound infrequently to ZP by insemination of cumulus-free eggs [binding sperm/egg: 14.3 ± 4.3 (Het), 1.3 ± 1.0 (KO)] (Fig. 3b, c). The decrease in sperm binding to ZP does not disrupt fertilization, but previous papers indicate a correlation between sperm ZP-binding and sperm migration through the UTJ26,29,32,33,36,44,45. To evaluate the ability of Galntl5 null sperm to migrate through the UTJ, we observed the behavior of fluorescence-labeled Galntl5 KO sperm in the female reproductive tract four hours after mating. While sperm from the control are observed in the oviduct, Galntl5 KO sperm are abundant in the uterus and are rarely observed in the oviduct (Fig. 3d). Furthermore, when we focused on the UTJ region, Galntl5 KO sperm bound to the UTJ are reduced compared to control sperm (Fig. 3e). Thus, Galntl5 KO sperm have a defect in UTJ binding and passage, leading to severe male fertility defects in vivo.

GALNTL5 is a critical factor for UTJ migration

To reveal why Galntl5 KO sperm show impaired UTJ migration, we examined the UTJ migration-required proteins in the TGC and sperm of Galntl5 KO mice. So far, more than twenty genes expressed in the testis and epididymis have been identified as encoding essential factors for sperm migration into the UTJ, and ADAM3, a sperm membrane protein, commonly disappears in sperm lacking these UTJ migration-related proteins (referred to as the ADAM3-dependent pathway)33. Thus, we first examined ADAM3 in TGC and sperm of Galntl5 KO males. Testicular ADAM3 is detected as a doublet band, and then ADAM3 is reduced in size by protein processing during epididymal transit46. ADAM3 in Galntl5 KO TGC was detected at comparable levels with Galntl5 Het TGC (Fig. 4a); however, the mature form of ADAM3 in Galntl5 KO sperm was slightly decreased (Fig. 4b and Supplementary Fig. 4a). If the mature form of ADAM3 slightly exists in sperm, the impaired UTJ migration of Adam3 KO sperm can be rescued in Adam3 KO males with an Adam3 transgene45. Therefore, we concluded that factors other than ADAM3 cause the impaired UTJ migration of Galntl5 KO sperm. A previous paper showed that disruption of tACE led to an aberrant distribution of ADAM3 in sperm, even if the mature form of ADAM3 was present in tAce KO sperm27. Thus, we examined tACE in Galntl5 KO TGC and sperm, but we failed to detect a difference between Galntl5 Het and KO males (Fig. 4a, c). Furthermore, to reveal the mature form of ADAM3 distribution in sperm, we separated sperm extracts from Galntl5 Het and KO males into two fractions [detergent-enriched region (membrane proteins) and detergent-depleted region (soluble proteins)] using Triton X-114. The previous paper showed that the mature form of ADAM3 mostly disappeared in the detergent-enriched region from tAce KO sperm27, indicating loss of ADAM3 from the sperm membrane. However, the mature form of ADAM3 could be detected in the detergent-enriched region from Galntl5 KO sperm (Fig. 4d). These results indicate that ADAM3 is normally distributed in the Galntl5 KO sperm membrane. From these analyses, we conclude that ADAM3-dependent mechanisms for sperm migration into the UTJ in Galntl5 KO males are normal.

Previous papers showed that Lypd4, Ly6k, and Pgap1 KO males are almost infertile because of impaired sperm migration through the UTJ, although these KO sperm have the mature form of ADAM334,35,36, suggesting there is an ADAM3-independent pathway to pass through the UTJ. Among the three factors, Pgap1 is ubiquitously expressed in the multiple tissues, including the testis, based on the NCBI database (https://www.ncbi.nlm.nih.gov/gene/241062), but it remains unclear whether PGAP1 exists in sperm. Further, previous papers showed that PGAP1 functions as a GPI inositol-deacylase in the endoplasmic reticulum37,38, suggesting that PGAP1 is not directly involved in the sperm migration into the UTJ and sperm-ZP binding. Thus, in this study, we examined LYPD4 and LY6K in Galntl5 KO TGC and sperm. Previous papers showed that LY6K exists in only the testis35 and LYPD4 exists in the testis and sperm36,47. The abundance of LY6K and LYPD4 in Galntl5 KO TGC and sperm was comparable to Galntl5 Het males (Fig. 4a, c). Therefore, we could not find significant differences in the abundance and distribution of known proteins for UTJ migration between Galntl5 Het and KO sperm. To reveal GALNTL5 in KO mice lacking UTJ-migration-related genes, we used Adam3 KO and Lypd4 KO male mice that demonstrate ADAM3-dependent and ADAM3-independent sperm migration33. Testicular GALNTL5 was detected in these KO TGC at a comparable level as in WT TGC (Fig. 4e), but sperm GALNTL5 nearly disappears in these KO sperm (Fig. 4f and Supplementary Fig. 4b). Thus, GALNTL5 is required for both ADAM3-dependent and ADAM3-independent sperm migration.

Galntl5 variant 1 is more important for male fertility

To reveal which Galntl5 variant is more important for male fertility, we newly generated Galntl5 mutant mice by introducing the gRNA/Cas9 protein complex into eggs to delete Galntl5 variant 1 (Fig. 5a). By mating the F0 mutants with WT mice, we established two mutant lines; em3 and em4 have the indel mutation of 8 nt deletion and 3 nt insertion and the indel mutation of 2 nt insertion in exon 2 of the Galntl5 variants, respectively, leading to the disruption of the first Met for Galntl5 variant 1 or the frameshift mutation (Fig. 5b and Supplementary Fig. 1a–c). As we obtained both KO mutant lines (Galntl5em3 and Galntl5em4 KO) by the intercross of Het mutants, we performed western blot analysis using extracts of the TGC and sperm and an antibody to recognize the C-terminus of GALNTL5. Only the upper band corresponding with predicted molecular sizes of variant 1 (~50 kDa) disappeared in the TGC of both KO mice, while the band at ~37 kDa also disappeared in the sperm of both KO mice (Fig. 5c). This result indicates that only variant 1 of GALNTL5 is disrupted in Galntl5em3 and Galntl5em4 KO mice, and the protein from variant 1 becomes the mature form during epididymal maturation. To reveal the effect of the disruption of Galntl5 variant 1 on male fertility, control and Galntl5em3 and Galntl5em4 KO male mice were mated with WT females. The females mated with Galntl5em3 and Galntl5em4 KO males delivered pups infrequently (Fig. 5d and Supplementary Table 1), phenocopying Galntl5em1 and Galntl5em2 KO males (Fig. 1g). Thus, Galntl5 variant 1 is more important for male fertility. Based on the TMHMM analyses, Galntl5 variant 1 is predicted to encode a type II transmembrane protein, while a protein encoded by Galntl5 variant 2 lacks the transmembrane domain (Supplementary Fig. 1e). Further, the predicted protein from Galntl5 variant 1 becomes the mature form lacking the putative transmembrane domain during sperm epididymal transit (Fig. 1f, Supplementary Figs. 1d and 3). To reveal whether sperm GALNTL5 exists on the sperm surface, we treated sperm surface proteins with a biotin labeling kit, and we observed an increase in the molecular weight of sperm GALNTL5 after biotinylation (Fig. 5e). Thus, the protein translated from Galntl5 variant 1 exists on the sperm surface even after truncation during epididymal maturation.

GALNTL5 attaches to the UTJ and ZP via binding with GalNAc

To reveal how GALNTL5 regulates sperm migration into the UTJ and sperm-ZP binding, we examined the physiological function of GALNTL5. As described in the introduction, GALNTL5 contains a conserved domain found in GalNAc transferase (also known as pp-GalNAc-T), suggesting that GALNTL5 may play a role in the glycosylation process in male germ cells. Although the previous publication did not demonstrate glycosyltransferase enzymatic activity of GALNTL5 in vitro41,43, we examined whether the lack of GALNTL5 influences sperm glycosylation by western blot analysis with lectins [PNA (binding to Galβ3GalNAc in O-Glycan), MAL-II (binding to Neu5Acα3Galβ3GalNAc in O-glycan), LSL-N (binding to poly-LacNAc in O-glycan and N-glycan), and Con A (binding to αMan and αGlc in N-glycan)]. We failed to find overt differences in the glycosylation patterns between Galntl5 Het and KO sperm (Fig. 6a). This result indicates that GALNTL5 does not have a function as a glycosyltransferase, corresponding with the previous papers41,43. As shown in Figs. 3b and 3e, Galntl5 KO sperm bind weakly to the UTJ and ZP which contain sugars, such as mannose, N-acetylglucosamine, galactose, and GalNAc48,49. Thus, we examined the possibility that GALNTL5 directly binds to sugars in the UTJ and ZP. Specifically, we generated expression vectors that encode the amino acid sequences of ~50 kDa and ~37 kDa from the C terminus of GALNTL5 (Supplementary Fig. 5a) corresponding with the testicular and the expected sperm molecular sizes, and we subsequently obtained the protein lysates from the culture of cells transfected with these plasmids. Then, we incubated the lysates in sugar-binding gold nanoparticles or a GalNAc-immobilized gel (Fig. 6b, c). In the analysis using the sugar-binding gold nanoparticles, the binding of proteins to sugars leads to the precipitation of these particles, resulting in a decrease in the 530 nm absorbance. When we incubated GALNTL5 of ~50 kDa [testicular GALNTL5 (immature form)] with gold nanoparticles bound to mannose, GalNAc, or glucose, we could observe the decreased absorbance only for GalNAc nanoparticles (Fig. 6b). We further examined the binding of GALNTL5 proteins of ~50 kDa and ~37 kDa [expected sperm GALNTL5 (expected mature form)] and GalNAc using carbohydrate gel-equipped GalNAc-immobilized beads (Fig. 6c). The proteins that did not bind to GalNAc in the column were removed by washing. Proteins bound to GalNAc-immobilized beads were released by adding competing GalNAc; in this experiment, we could detect both immature and expected mature GALNTL5 proteins in the elution buffer (Fig. 6c). These results indicate that sperm GALNTL5 after testicular GALNTL5 is processed during epididymal maturation can also directly bind GalNAc. Furthermore, we examined which region of GALNTL5 is required to bind to GalNAc by narrowing down the amino acid sequence. As shown in Supplementary Fig. 5a, we generated three expression vectors encoding amino acid sequences of ~30, ~20, and ~10 kDa from the C terminal of GALNTL5, and then these vectors were transfected into the culture cells. After obtaining the cell lysates, we incubated the proteins with GalNAc-immobilized gel. We found that GALNTL5 of ~30 kDa and ~20 kDa (containing the predicted glycosyltransferase 2-like domain) can be detected in the elution buffer (Supplementary Fig. 5b). Our results indicate that at least 20 kDa from the C-terminal of GALNTL5 protein is required for GalNAc binding. Furthermore, to examine the GalNAc binding ability of human GALNTL5 (Supplementary Fig. 5c), we incubated human GALNTL5 protein with the GalNAc-immobilized gel. As shown in Supplementary Fig. 5d, we found human GALNTL5 in the elution buffer, indicating that human GALNTL5 also has GalNAc binding ability. Based on these data, we conclude that mouse and human GALNTL5 can directly bind GalNAc.