Ttll9^−/− mice sperm flagella show shortening of doublet 7, reduction of doublet 5 polyglutamylation and a stall in beating

Eukaryotic flagella and cilia are microtubule (MT)-based organelles that have a common structure called the axoneme. Motile axonemes are complex molecular machineries composed of dyneins and regulatory components attached to scaffolding MTs (Inaba, 2007, 2011). Axonemal MTs are arranged in so-called 9+2 structures, comprising nine outer doublet MTs and two central singlet MTs. Although the outer doublets are often referred to as rotationally symmetrical, several structural heterogeneities are also reported (Heuser et al., 2012).

One of the tubulin post-translational modifications (PTMs), polyglutamylation, shows an interdoublet heterogeneity, where doublets 1, 5, 6 and 9 have higher levels of modification than other doublets in mammalian sperm flagella (Fouquet et al., 1996; Prigent et al., 1996; Kann et al., 2003). Polyglutamylation is a unique PTM by which variable lengths of glutamate side-chains are attached to the C-terminal tail (CTT) of α- and β-tubulins (Eddé et al., 1990). Because tubulin PTMs can affect the structural and chemical properties of MTs (Wloga and Gaertig, 2010; Konno et al., 2012; Magiera and Janke, 2014), establishing proper polyglutamylation patterns is essential for correct sperm flagellar structures and/or functions. However, mechanisms for establishing the heterogeneous polyglutamylation pattern and the importance of the interdoublet heterogeneity are almost completely unknown.

Polyglutamylation is catalyzed by a subset of tubulin tyrosine ligase-like proteins (TTLLs) (Janke et al., 2005; Ikegami et al., 2006; van Dijk et al., 2007). The importance of the PTMs of motile flagella and cilia has been reported previously in various models, such as mouse (Ikegami et al., 2010; Lee et al., 2013), ependymal cells (Bosch Grau et al., 2013), Tetrahymena (Wloga et al., 2010; Suryavanshi et al., 2010) and Chlamydomonas (Kubo et al., 2010, 2012). Recently, a few reports on transgenic mouse models have also underlined the importance of polyglutamylation in spermatozoa. For example, severely shortened flagella have been found in Ttll1-knockout mice (Ikegami et al., 2010), and subfertility has been reported in Stamp^tm/tm mice, which have a truncation in the non-catalytic region of Ttll5, due to structural defects of sperm flagella and reduced motility (Lee et al., 2013).

Here, we have investigated the sperm structure and motility of transgenic mice lacking polyglutamylase Ttll9 (Ttll9^−/−). Ttll9^−/− causes infertility in male mice owing to reduced sperm count and defective sperm motility. We demonstrate that Ttll9^−/− sperm axonemes lose doublet 7 at distal principal pieces and show reduced tubulin polyglutamylation, with a reduction in polyglutamylation of doublet 5 being the most remarkable. We also show that reduced motility of Ttll9^−/− spermatozoa is caused by frequent stalls of flagella, which is indicative of defective switching in the bending direction. This tendency of the stall patterns seems to be caused, at least partly, by the reduction of polyglutamylation on doublet 5, probably at a position involved in the switching between bending directions. These results indicate that the establishment of heterogeneous polyglutamylation by Ttll9 is essential for both normal structure and beating pattern of murine sperm flagella.

We examined the expression pattern of Ttll9 with reverse-transcriptase (RT)-PCR and found strong expression in testes of wild-type mice (Fig. 1A). The Ttll9 transcript was not detected in brain, probably owing to limited expression in that tissue (Bosch Grau et al., 2013; Zeisel et al., 2015). To identify the area of expression of Ttll9 in wild-type testes, we performed in situ hybridization on cross sections (Fig. 1B). Ttll9 expression was detected inside of seminiferous tubules with a lack of signal at the most peripheral regions, similar to that of many genes involved in spermatogenesis and/or sperm function.

To understand the importance of polyglutamylation in mammalian sperm function, we generated tubulin polyglutamylase Ttll9-deficient (Ttll9^−/−) mice, where a stop codon introduced by a frameshift resulted in premature termination of translation (Fig. S1A–D). The Ttll9^−/− mice did not show a coughing-like phenotype or hydrocephalus, indicating no deficiency in respiratory and ependymal cilia. They also showed neither apparent polydactylism nor polycystic kidneys. Although the external morphology of Ttll9^−/− mice seemed normal (Fig. 1C), and their body weight was comparable to that of wild-type mice (Fig. 1D), Ttll9^−/− males failed to sire pups (Fig. 1E), despite normal mating behavior and ability to deposit a plug.

To reveal the cause of male infertility, we observed the testes of Ttll9^−/− males. Visual inspection of the testicular phenotype of Ttll9^−/− males revealed that the external morphology and the size of wild-type and Ttll9^−/− testes were comparable (Fig. 2A,B). Cross sections of Ttll9^−/− testes also showed no apparent histochemical defects in their gross structures (Fig. 2C). These observations indicate that the morphogenesis of testes and early spermatogenesis are not severely affected in Ttll9^−/− mice. To evaluate the effect of Ttll9 knockout on polyglutamylation, we performed immunoblotting with antibodies against polyglutamate side-chain (polyE) and α-tubulin (12G10) (Fig. 1D,E). The blots showed a reduction of tubulin polyglutamylation in Ttll9^−/− testes.

We then analyzed epididymal spermatozoa and found that the sperm count in Ttll9^−/− cauda epididymides was significantly decreased (Fig. 2F). The polyglutamylation level of Ttll9^−/− spermatozoa was also reduced (Fig. 2G,H). Tubulins seem to be the only major group of proteins to be abundantly polyglutamylated in murine spermatozoa (Fig. S1E). By using light microscopy, we observed that the head and tail of Ttll9^−/− spermatozoa were often detached (Fig. 2I). Nevertheless, some spermatozoa showed a normal appearance with a hook-shaped head and elongated flagellum.

To examine the ultrastructure of Ttll9^−/− spermatozoa, we performed transmission electron microscopy (TEM) analysis. TEM analysis showed that the lumens of seminiferous tubules (Fig. 3A) and cauda epididymides (Fig. 3B) of Ttll9^−/− males contained fewer spermatozoa than did wild-type males. Cell debris was often observed in the lumen of Ttll9^−/− epididymides, but most Ttll9^−/− flagella in cauda epididymides showed normal axonemal structures at midpieces (Fig. 3C). At distal principal pieces, however, the frequent loss of a single doublet, doublet 7, was observed (Fig. 3C, arrow). This demonstrates that doublet 7 was shortened in Ttll9^−/− sperm flagella. In some cases, an electron-dense mass was found in the place where doublet 7 was originally present (Fig. S2A, arrow). Occasionally, a possible breakage of a doublet was observed in longitudinal sections (Fig. S2B). The shortening of the doublet 7 was never observed in Ttll9^−/− testes (Fig. 3A). Minor abnormalities included ectopic and/or excess doublet(s) or outer dense fibers (ODFs) and a distorted fibrous sheath (Fig. S2C). The ultrastructure of respiratory cilia was normal in Ttll9^−/− tracheae (Fig. S2D). The amount of VDAC3 was not altered in Ttll9^−/− spermatozoa (Fig. S2E), although VDAC3 deficiency is reported to result in the doublet 7 loss (Sampson et al., 2001).

We further analyzed the position where the doublet 7 loss occurs along the flagellum. In mouse spermatozoa, nine ODFs at a proximal flagellum tapered and decreased in number with fixed order toward the endpiece, where no ODFs remained (Fig. 3D). We noticed that the loss of doublet 7 always occurred posterior to the point of ODF 7 termination (Fig. 3D,E). At the most distal region where no ODF was present, doublet 7 was missing in more than half of the Ttll9^−/− axonemes (Fig. 3E).

Outer doublet MTs in murine sperm axonemes are heterogeneous in polyglutamylation – the PTM level of doublets 1, 5, 6 and 9 is higher than that of the others (Fouquet et al., 1996). We conducted immunogold electron microscopy analysis to assess the effect of Ttll9 loss on the heterogeneous modification pattern (Fig. 3F; Fig. S2F,G). Cross sections of proximal midpieces, where most Ttll9^−/− axonemes have normal structure, were analyzed. Using an antibody against α-tubulin (DM1A), we found that the mean particle numbers on each axoneme were comparable in wild-type and Ttll9^−/− flagella (Fig. S2H). The particle numbers on each MT set were almost uniform, although appearing to be slightly decreased on doublet 7 in Ttll9^−/− sperm flagella compared to wild-type samples (Fig. S2J).

Using the polyE antibody, we found that the reduction of particle numbers in Ttll9^−/− axonemes was not significant but did show a trend towards decreasing (Fig. S2I), probably because we investigated axonemes with normal ultrastructures, which are expected to have a less severe reduction of the PTM. Among ten MT sets (nine outer doublets and the central pair), doublet 5 showed the largest reduction in the amount of polyglutamylation (Fig. 3F). The largest reduction of particle number on doublet 5 was reproducibly detected in three independent experiments, whereas other doublets showed large variances (Fig. 3G). Statistical analysis revealed that the reduction in particle number on doublet 5 was a significant outlier (P<0.05, Grubbs' test, one tailed).

We performed motility analysis to evaluate the effect of Ttll9 knockout on sperm flagellar motility with high-speed charge-coupled device (CCD) camera recordings. Trajectories of Ttll9^−/− spermatozoa showed reduced percent motility (percentage of motile spermatozoa) and progressive motility compared with those of wild-type spermatozoa (Fig. 4A–C; Movie 1). We then focused on the flagellar waveforms of motile Ttll9^−/− spermatozoa. At higher magnification, wild-type flagella showed smooth wave propagation, whereas Ttll9^−/− flagella exhibited irregular beats with intervening stalls that continued for a variable time period (Movie 2).

Because murine spermatozoa have a hook-shaped head, the direction of flagellar bending can be defined as pro-hook and anti-hook bends (Ishijima et al., 2002). To relate the axonemal geometry with beating direction, we performed TEM analysis. The head of the murine spermatozoa protrudes posteriorly at the anti-hook side, and one can observe the protrusion and the cross section of the proximal flagellum at the same time in a single spermatozoa. TEM analysis revealed that doublet 1 is at the pro-hook side and doublets 5 and 6 are at the anti-hook side in murine spermatozoa (Fig. S2K). We conducted curvature analysis with flagellar waveform traces to assess the characteristics of the stall, and the relationship between the stall and the axonemal architecture (Fig. 4D; Fig. S3). Beat patterns were classified into four categories – normal, stalls after pro- and anti-hook bend (pro-hook stall and anti-hook stall, respectively), and double stall with pro- and anti-hook stall at the same time in a single flagellum. Fig. 4D shows superpositions of representative waveforms (top) and corresponding curvature plots (bottom) of the four types of beat patterns. Two thirds of the Ttll9^−/− flagella showed one of the three stall patterns and, among them, anti-hook stall was the most frequent (Fig. 4E).

In the present study, we reveal that the loss of Ttll9 causes a unique set of abnormalities related to the heterogeneity of axonemal structures and polyglutamylation in murine sperm flagella – i.e. the shortening of doublet 7 and the reduction of the level of polyglutamylation preferentially on doublet 5. Our data suggest that the abnormal motility of Ttll9^−/− spermatozoa seems to be explained by the reduction of polyglutamylation on doublet 5.

The selective shortening of doublet 7 is a unique phenotype of Ttll9^−/− sperm flagella. This was never observed in Ttll9^−/− testes (Fig. 3A), indicating that the cause of the shortening is not defects in axoneme assembly. A possible cause of the specific shortening of the doublet 7 is severing of the doublet. Indeed, we observed evidence for possible severing of a doublet (Fig. S2B). The extra doublet observed can sometimes emerge when severed doublet 7 is dislocated proximally. Another cause is the depolymerization or degradation of MTs. An amorphous electron-dense mass found in the place of doublet 7 could be the remnant of the depolymerized or degraded doublet. Thus, these mechanisms – severing and depolymerization and/or degradation – could underlie the shortening of doublet 7 in Ttll9^−/− sperm flagella. Severing and depolymerization seems to be prevented by mechanical support from the associated ODF 7 in Ttll9^−/− spermatozoa, because the loss of doublet 7 always occurred distal to the level of the ODF 7 termination.

Destabilization of axonemal MTs caused by loss of a polyglutamylase has been reported in other models (Pathak et al., 2011; Lee et al., 2012). In contrast, Kubo et al. (2015) have reported that reduced polyglutamylation through the loss of the TTLL9–FAP234 complex stabilizes Chlamydomonas flagella by affecting the turnover rate of tubulin at the tips. The inconsistency might be explained by the fact that sperm flagella do not have the cytoplasmic pool of tubulin required for turnover, and their stability is likely to be determined by different mechanisms from those of Chlamydomonas. The selective loss of doublet 7 implies that the role of polyglutamylation in sperm flagella is complex and that outer doublet MTs are more heterogeneous than generally considered. Because doublet 7 is one of the MT subsets that have lower polyglutamylation in mammalian spermatozoa, even mild reduction of the PTM might have a larger effect. Nonetheless, we did not observe the loss of the other axonemal MTs with lower polyglutamylation levels (i.e. doublets 2, 3, 4 and 8), indicating the involvement of other factors.

Interestingly, the loss of doublet 7 has been reported previously for spermatozoa of VDAC3-deficient mice (Sampson et al., 2001). VDAC3 is a mitochondrial outer membrane protein but is also a component of ODFs in bovine spermatozoa (Hinsch et al., 2004), and some other VDACs are also thought to be present in mammalian sperm flagella (Liu et al., 2009). VDAC3 does not have glutamate-rich regions, which can be targets for polyglutamylation (van Dijk et al., 2008). Thus, it might not be possible that VDAC3 is polyglutamylated by TTLL9 and that the phenocopy is caused by the loss of polyglutamylation on VDAC3. Indeed, tubulins are the only major group of proteins to be abundantly polyglutamylated in spermatozoa (Fig. S1E). Tubulins bind to VDACs through their negatively charged CTTs (Rostovtseva et al., 2008). Therefore, reduced polyglutamylation is expected to weaken the interaction between VDACs and tubulins, which might be the cause of similar structural phenotypes between Vdac3 and Ttll9 mutants. Future research on the exact localization of VDACs and their interactions with tubulins in flagella would provide useful information to reveal the molecular mechanisms underlying selective loss of doublet 7 in Vdac3- and Ttll9-deficient mice.

The other heterogeneous effect of the loss of TTLL9 on outer doublet MTs is the greatest reduction of polyglutamylation in doublet 5, suggesting some selectivity of TTLL9. Doublet 5 of sea urchin sperm flagella and a corresponding doublet of Chlamydomonas flagella (doublet 1) have been reported to be structurally different from other doublets (Lin et al., 2012). This structural uniqueness of doublet 5 is likely to be conserved in murine sperm flagella (Lindemann et al., 1992). The structural heterogeneity might allow preferential, if not exclusive, recruitment of TTLL9 to doublet 5. This also indicates the possibility of other TTLLs that have selectivity for a specific doublet. Structural heterogeneity is not limited to doublet 5. For example, I1 inner arm dyneins on doublets 3 and 4 are structurally different from those of other doublets in Chlamydomonas flagella (Heuser et al., 2012). It is tempting to speculate that structural heterogeneity can affect the affinity of tubulin-modifying enzymes to specific MT subsets in sperm axonemes. The idea of ‘division of labor’ among TTLLs is also supported by the fact that Stamp^tm/tm mice show specific loss of doublet 4 in sperm flagella (Lee et al., 2013). No obvious ciliary phenotypes in other tissues suggest that TTLL9 has, if any, only a minor function, or that its function is compensated by other TTLLs there. It will be interesting to investigate whether other TTLLs have a similar role in other cell types in future.

The polyglutamylation levels in the testes, whole sperm and proximal part of sperm flagella of Ttll9^−/− mice were 50–70% of those in wild type. The polyglutamylation level on doublet 5 in Ttll9^−/− mice was also ∼60% of that in wild type. An argument that such a reduction could have an impact on flagellar structure and motility is possible. A unique property of polyglutamylation is that even its moderate reduction can cause remarkable effects on the microtubular system – e.g. motility defects of the cilia in Tetrahymena (Suryavanshi et al., 2010). Therefore, it is not surprising that the partial reduction of polyglutamylation caused by Ttll9 loss can have significant effects on axonemal structures and motility. It is also important to note that the severe reduction of the PTM inhibits normal flagellogenesis of mouse spermatozoa (Ikegami et al., 2010). Therefore, ‘hypomorphic’ mutants of polyglutamylation, like Ttll9^−/− mice, offer a valuable opportunity to study polyglutamylation in mammalian sperm flagella.

Ttll9 deficiency strongly affected the beating pattern of murine sperm flagella, in contrast with the Chlamydomonas mutant lacking functional TTLL9, which shows an almost normal beating pattern (Kubo et al., 2010, 2012). We often observed active beats of Ttll9^−/− spermatozoa at distal flagellum that showed a stall at the proximal region (Movie 2). This seems to contradict the observation that doublet 7 is frequently missing at distal flagellum. Because Stamp^tm/tm spermatozoa that lack doublet 4 are reportedly motile (Lee et al., 2013), one possible explanation for the motile distal flagellum in Ttll9^−/− spermatozoa is that the loss of a single doublet can be compensated by other doublets. However, a more convincing explanation is that the motile distal flagellum retains all doublets because about half of flagella should still have nine doublets, even at the end piece (Fig. 3E). Longitudinal heterogeneity of polyglutamylation might also be associated with stalls at proximal flagella. In murine spermatozoa, the polyglutamylation level is highest at the flagellar base and decreases toward the tip (Fouquet et al., 1996). Therefore, the effect of reduced polyglutamylation must be more severe at proximal flagella.

Several studies have reported the importance of tubulin polyglutamylation in ciliary and flagellar motility. Coordinated activation of axonemal dyneins and cyclic interdoublet sliding are essential for smooth bending propagation along flagella and cilia (Satir, 1985). Because reduced polyglutamylation alters the dynein–MT interaction (Suryavanshi et al., 2010; Kubo et al., 2010; Sirajuddin et al., 2014; Alper et al., 2014), reduced polyglutamylation in Ttll9^−/− axonemes is likely to be the cause of stalls. According to one of the promising models for flagellar motility, the geometric clutch model, mechanical stress imposed on doublets plays a key role for normal beating (Lindemann and Lesich, 2010, 2015). The model assumes that flagellar bending generates forces that are transverse to the outer doublets (t-forces), which would pry interacting doublet pairs apart to cease sliding. Kubo et al. (2010) suggest that the reduced polyglutamylation could weaken the interaction between the C-terminal tails of tubulins and the positively charged stalk tips of dyneins. The weakened affinity should alter the threshold of the geometric clutch mechanism and might cause the frequent stalls.

Frequent anti-hook stalls observed in Ttll9^−/− sperm flagella seem to be caused by the failure of bend switching after the anti-hook bend. This can be explained, at least partly, by the geometry of the axoneme in murine spermatozoa. Doublet 1 is located at the hook side of the head, and doublets 5 and 6 are at the anti-hook side in murine sperm flagella (Fig. S2K). Dynein arms protrude clockwise, when seen from the base (Gibbons, 1963). According to the ultrastructural geometry and the switch point hypothesis of ciliary beating (Satir and Matusoka, 1989), the pro-hook bend is directed towards doublets 5 and 6 through the activity of dyneins on doublets 6–9, and the anti-hook bend is directed towards doublet 1 through the activity of dyneins on doublets 1–4 (Fig. 4F). A possible explanation is a defect in switching signals from the central apparatus to doublet 5. Supporting this scenario, interestingly, mice with a defective central apparatus protein, hydin, show frequent stalls of ependymal cilia (Lechtreck et al., 2008), which is similar to what is observed in Ttll9^−/− sperm flagella. Because the C2b projection on the central apparatus is missing in the hydin-deficient cilia, the C2b projection is thought to be essential for switching of bending directions. Because the central pair in Metazoa does not rotate, the C2b projection is always located close to, and has possible interactions with, doublets 4 and 5 (Fig. 4F), where the most severe effect of Ttll9^−/− loss is expected. Although the C2b projection is present in Ttll9^−/− flagella, defects in the function or mechanical property of doublet 5 downstream of C2b signaling likely mimic some features of C2b loss, that is, frequent stalling in Ttll9^−/− flagella. An alternative explanation is the failure of pro-hook bend initiation as a result of a defective doublet 7. Because doublet 7 is one of nine doublets that is expected to work during pro-hook bending (Fig. 4F), its deficiency could impair the initiation of pro-hook bend. A shortened doublet 7 would perturb any feedback system from distal to proximal flagellum (Hayashi and Shingyoji, 2008).

Our data suggest that TTLL9 is involved in the establishment of heterogeneity among doublet MTs, which is essential for structural and functional integrity of murine sperm flagella. Although structural and biochemical heterogeneity among individual doublets has not been highlighted in modelling ciliary and flagellar beating (Lindemann and Lesich, 2010), our results emphasize the importance of interdoublet heterogeneities in normal axonemal structures and functions. Interestingly, the phenotype of Ttll9^−/− sperm flagella is quite different from that observed in the flagella of Ttll9-deficient Chlamydomonas, where no severe structural and motility defects are seen. The inconsistency seems to reflect the differences in polyglutamylation patterns, accessory structures and mechanisms to regulate axonemal motility, highlighting the importance of comparative studies on various cilia and flagella.

Wild-type (WT) mice of the C57BL/6N strain were purchased for breeding from CLEA Japan (Tokyo, Japan). All experiments were performed in accordance with guidelines issued by the Institutional Animal Care and Use Committees of Hamamatsu University, School of Medicine.

For Ttll9 gene disruption, exons 4–5 were replaced with a neo cassette by homologous recombination in RENKA embryonic stem cells in the C57BL/6N background (Mishina and Sakimura, 2007). PCR-based genotyping was performed on genomic DNA extracted from mouse tails. The following primers were used: Ttll9^−/− primer, 5′-GACGTGCTACTTCCATTTGTC-3′; wild-type primer, 5′-CTCTAGAGAGCTCCAACACTT-3′; and common primer, 5′-GCACCTTAGGAAGTAGTTGAG-3′. Expected PCR products were 547 bp for the wild-type allele and 443 bp for the Ttll9^−/− allele. For in vivo fertilization assays, Ttll9^−/− and wild-type males at 8–9 weeks of age were mated one-to-one with 8-week-old C57BL/6N females for a period of no longer than 2 months.

Total RNA was extracted with Sepasol-RNA I Super (Nacalai Tesque, Kyoto, Japan) from various tissues and reverse-transcribed with ReverTra Ace (Toyobo, Osaka, Japan). The following primers were used to examine the expression of Ttll9: full-length Ttll9 (accession number NM_001083618), forward (Ttll9-F) 5′-ATGTCGCGACAGAAGAATC-3′, reverse (Ttll9-R) 5′-TCAGCTAGGGGCTTTCC-3′; as an internal control Gapdh (accession number GU214026), forward 5′-TGCCCCCATGTTTGTGATG-3′, reverse 5′-TGTGGTCATGAGCCCTTCC-3′. Wild-type and Ttll9^−/− transcripts were examined with the following primers: exons 1–15, forward Ttll9-F, reverse Ttll9-R; exons 1–3, forward Ttll9-F, reverse 5′-TTTGACTTCCACCCATCC-3′; exons 4–5, forward 5′-GAAGGCGAATGGGATTTC-3′, reverse 5′-AGGCTTCATGATCCAGGTG-3′; exons 6–15, forward 5′-TCATGGACTGGAGGAAGG-3′, reverse Ttll9-R.

Wild-type mouse testes were fixed in 4% paraformaldehyde and cryopreserved with sucrose in PBS. Testes were then embedded in optimal cutting temperature compound (SAKURA Finetek, Tokyo, Japan) and sliced into 20-µm-thick sections. Full-length Ttll9 was inserted into the TOPO vector (Life Technologies) and linearized for the synthesis of a DIG-labeled probe. In vitro transcription and in situ hybridization were performed as previously described (Mukai et al., 2009).

Testes excised from 8-week-old Ttll9^−/− mice and wild-type mice were snap-frozen in powdered dry ice and sliced into 10-µm-thick sections with a cryostat (CM1950 Cryostat, Leica Microsystems, Wetzlar, Germany). The histological sections were stained with Mayer's hematoxylin and eosin (Sakura Finetek). Histological images were obtained with a Leica LMD6000 laser microdissection microscope (Leica Microsystems) with a 10× objective lens (Leica HCX PL FLUOTAR 10×/0.30) and Hitachi HV-D20 CCD camera (Hitachi Kokusai Electric, Tokyo, Japan) and Leica Laser Microdissection software v6.5 (Leica Microsystems).

The number of spermatozoa in the cauda epididymides of 8-week-old mice was counted based on an established method (Wang, 2003). Briefly, cauda epididymides were weighed and minced in 2 ml of PBS. After a 15-min incubation at 37°C, a homogeneous sperm suspension was obtained after pipetting up and down. The sperm suspension was then diluted with PBS, and the spermatozoa were immobilized by heating at 60°C for 1–2 min. The number of spermatozoa was estimated by loading 10 µl of sperm suspension onto a Neubauer hemocytometer, and normalized using the weight of the cauda epididymides.

A pair of cauda epididymis from a single mouse were minced in 2 ml of PBS and incubated for 15 min at 37°C. The sperm suspension was centrifuged, and the sperm pellet was retrieved and washed twice with PBS. SDS-PAGE sample buffer was added to the sperm pellet, the mixture was sonicated, and the supernatant was collected after centrifugation at 200,000 g for 10 min at 4°C. Proteins were separated in 10% acrylamide gel and then transferred onto a PVDF membrane. PVDF membranes were first treated with blocking solution 1 (7.5% skimmed milk, 0.1% Tween-20 in TBS) at room temperature for 1 h and then incubated with anti-polyglutamate side-chain antibody (polyE, 1:3000), anti-α-tubulin antibody (12G10; 1:2000; Developmental Studies Hybridoma Bank), or anti-VDAC3 antibody (1:1000; Proteintech) in blocking solution 1 for 1 h. After several washes with 0.1% Tween-20 in TBS, the membrane was incubated for 1 h with goat anti-rabbit-IgG antibody horseradish peroxidase (HRP) conjugate (1:10,000; Jackson ImmunoResearch Laboratory) for polyE and anti-VDAC3 antibodies, or goat anti-mouse-IgG antibody HRP conjugate (1:10,000, Jackson ImmunoResearch Laboratory). The membrane was then washed several times with 0.1% Tween-20 in TBS and treated with Amersham ECL western blotting detection reagents (GE Healthcare, Little Chalfont, UK). Chemiluminescence was detected with a luminescent image analyzer, LAS-3000 mini (Fujifilm, Tokyo, Japan). For quantitative analysis, whole testis or sperm proteins from three wild-type and three Ttll9^−/− individuals were blotted on the same membrane to ensure the same conditions during immunostaining. Signal intensity of the original 16-bit TIFF images was quantified with ImageJ software (National Institute of Health).

Testes or cauda epididymides were fixed with 2% glutaraldehyde in 0.067 M phosphate buffer (pH 7.4) for 2 h at 4°C. After several washes with phosphate buffer, they were post-fixed with 1% osmium tetroxide in 0.067 M phosphate buffer for 2 h, dehydrated with an ascending series of ethanol and propylene oxide, and embedded in Quetol-812 (Nisshin EM Corporation, Tokyo, Japan). Ultrathin sections of 80–100 nm in thickness were cut with an ultramicrotome (Ultracut UCT, Leica Mycrosystems). Following consecutive staining with uranyl acetate for 5 min and lead stain solution (Sigma-Aldrich) for 3 min, carbon was deposited with the JEOL JEE-4X vacuum evaporator (JEOL, Tokyo, Japan). The sections were observed with a JEOL 1220 electron microscope (JEOL) at 80 kV. Electron micrographs were taken with Bioscan Camera Model 792 and DigitalMicrograph software (Gatan, CA, USA).

A previously described method was modified and used for immunoelectron microscopy (Kann and Fouquet, 1989). Cauda epididymides were fixed with 1% glutaraldehyde and 0.1 M phosphate buffer (pH 7.4) for 2 h at 4°C. After several washes with 0.1 M phosphate buffer, cauda epididymides were dehydrated with a series of ethanol solutions. The specimen was then moved into LR White and solidified by heating at 55°C for 24 h. The blocks were sliced into 80- to100-nm thick sections, and the sections were attached to nickel grids. These sections were treated with blocking solution 2 (150 mM NaCl, 0.1% BSA, 10 mM glycine in 20 mM Tris-HCl, pH 7.8) for 1 h, rinsed with TBS and treated with primary antibody DM1A (1:1000) or polyE (1:1000) diluted with blocking solution 2 for 1 h. After several rinses with TBS, sections on the nickel grids were treated with goat anti-mouse-IgG 10-nm gold for DM1A (1:50) or goat anti-rabbit-IgG 10-nm gold for polyE (1:50) (BBI solutions, Cardiff, UK) for 1 h. Sections on the nickel grids were then rinsed with TBS and distilled water several times, and stained with uranyl acetate for 30 s and lead stain solution for 10 s. Carbon deposition and analysis of the sections were conducted as described above for TEM. For quantitative analysis of the heterogeneous polyglutamylation patterns, the number of the gold particles on ultrathin sections of cauda epididymides from wild-type and a Ttll9^−/− mice were counted. The results of three experiments with independent sperm samples were standardized, and mean values were acquired from the three experiments.

Cauda epididymides were excised from Ttll9^−/− and wild-type mice, minced in modified Tyrode's albumin lactate pyruvate (mTALP) medium, which had been pre-equilibrated under 5% CO₂ at 37°C for at least 1 h. The fragments of cauda epididymis were incubated in a CO₂ incubator for 10 min to allow sperm to swim out. The medium with spermatozoa was suspended, loaded onto a Leja^® Standard Count 2 Chamber Slide (#SC 20-01-02-B, Leja, Nieuw-Vennep, The Netherlands) and recorded at room temperature at 200 or 500 frame per second (fps) with a digital high-speed camera HAS-L1 (DITECT, Tokyo, Japan) attached to an Eclipse TE2000-U (Nikon, Tokyo, Japan) or Leica DMI3000B (Leica Microsystems) microscope. The objective lenses used were either Plan Fluor 4×/0.13 and Plan Fluor 10×/030 on Eclipse TE2000-U or N PLAN 5×/0.12 and HC PL APO 10×/0.40 on Leica DMI3000B. For trajectory superposition, sequential images at 200 fps obtained as above were downsized to 10 fps. Images for 5 s at 10 fps (50 images in total) were superposed with ImageJ software with Color FootPrint (http://www.jaist.ac.jp/ms/labs/hiratsuka/images/0/09/Color_FootPrint.txt) to show sperm swimming trajectories as a spectrum of colors.

For curvature analysis, original images were downsized to 50 fps using ImageJ. Only spermatozoa showing no rotation for at least more than 0.5 s were selected. Cell motility analysis software, Bohboh (Bohbohsoft, Tokyo, Japan), was used to trace flagella and calculate their curvatures (Baba and Mogami, 1985). Flagella of the selected spermatozoa were traced automatically in each frame. Traces were visually checked and manually corrected when needed. To superpose flagellar traces, the flagellar base was considered the reference point, and the imaginary line between the flagellar base and tip was used as the reference line (see also Fig. S3A). After flagellar traces were arranged as described above, curvatures along a flagellum were calculated and plotted as a function of the distance from the base of the flagellum. Flagellar curvature was defined as the inverse of the radius of the osculating circle at a given point on a flagellum. Curvature plots for 0.5 s at every 20 ms were considered superpositions of the flagellar traces (Fig. S3B,C). Beat direction was determined by the hook-shaped heads of the spermatozoa by adopting the principles of Ishijima et al. (2002) – the pro-hook bend follows the direction that the hook is pointing, and the anti-hook bend was opposite to that of the pro-hook bend. To distinguish between pro- and anti-hook bends, the anti-hook bend curvature was represented as a negative value (Fig. S3B,C). We classified beating patterns into four classes from the viewpoint of stall patterns: (1) normal beat, curvature at any point on a flagellum takes both positive and negative values; (2) pro-hook stall, a given point on a flagellum maintains a pro-hook bend state (only positive value in the curvature within a 0.5 s time frame); (3) anti-hook stall, the opposite of pro-hook stall (only negative values in the curvature); (4) double stall, pro-hook and anti-hook stalls occur at the same time on a single flagellum.

Brightness and contrast of the images presented here were adjusted with Adobe PhotoShop Elements 12 (Adobe Systems). Charts were made with Excel^® 2013 (Microsoft) and modified with Adobe Illustrator CS6 (Adobe Systems) to maintain the original values and proportions. Movie Maker^® (Microsoft) was used to edit the movie files.

We thank members of our laboratory for critical reading of the manuscript. We also grateful for Showbu Sato, Kenji Nakamura, Mineo Matsumoto, Reiko Tsuchiya and Shouko Takamatsu at the Mitsubishi Kagaku Institute of Life Sciences for technical support. The monoclonal antibody 12G10, originally developed by J. Frankel and E. M. Nelsen was obtained from the Developmental Studies Hybridoma Bank, created by the National Institute of Child Health and Human Development of the National Institutes of Health and maintained at the Department of Biology, University of Iowa, Iowa City, IA