salto/CG13164 is required for sperm head morphogenesis in Drosophila

Abstract

Producing mature spermatozoa is essential for sexual reproduction in metazoans. Spermiogenesis involves dramatic cell morphological changes going from sperm tail elongation and nuclear reshaping to cell membrane remodeling during sperm individualization and release. The sperm manchette plays a critical scaffolding function during nuclear remodeling by linking the nuclear lamina to the cytoskeleton. Here, we describe the role of an uncharacterized protein in Drosophila, salto/CG13164, involved in nuclear shaping and spermatid individualization. Salto has dynamic localization during spermatid differentiation, being progressively relocated from the sperm-nuclear dense body, which is equivalent to the mammalian sperm manchette, to the centriolar adjunct and acrosomal cap during spermiogenesis. salto-null male flies are sterile and exhibit complete spermatid individualization defects. salto-deficient spermatids show coiled spermatid nuclei at late maturation stages and stalled individualization complexes. Our work sheds light on a novel component involved in cytoskeleton-based cell-morphological changes during spermiogenesis.

INTRODUCTION

Both in mammals and in insects, spermiogenesis involves severe cell morphological changes, including sperm tail elongation, nuclear remodeling, and ultimately sperm individualization, also called spermiation (Fabian and Brill, 2012; O’Donnell, 2014). Nuclear elongation is associated with DNA condensation and involves different sets of proteins (Rathke et al., 2007). Whereas protamines and transition proteins are involved in chromatin condensation, nuclear elongation requires proteins of both the inner and outer nuclear membranes and involves perinuclear microtubules (Fabian and Brill, 2012; O’Donnell, 2014). In mammals and in Drosophila, proteins of the LINC complex link the nuclear envelope and nuclear lamina to the microtubule cytoskeleton (Kracklauer et al., 2010; Lehti and Sironen, 2016). Specifically, SUN domain–containing proteins of the inner nuclear membrane are required for nuclear morphogenesis (Keller et al., 2009; Kracklauer et al., 2010; Calvi et al., 2015). These SUN domain proteins interact with KASH domain–containing proteins of the outer nuclear membrane that link cytoplasmic microtubules (for reviews see Kierszenbaum, 2002; Lehti and Sironen, 2016). These perinuclear cytoplasmic microtubules are important for nuclear shaping during spermiogenesis. In Drosophila, the perinuclear microtubules compose the dense complex (DC), which is analogous to the mammalian manchette involved in sperm-head shaping (for review see Fabian and Brill, 2012). During spermatid differentiation, microtubules of the DC concentrate on the fenestrated side of the nuclei. Proteins such as Spag4 (SUN domain protein) and Yuri have been shown to distribute respectively on the inner nuclear membrane and along the microtubules of the DC and to interact functionally during nuclear elongation (Texada et al., 2008; Kracklauer et al., 2010; Sitaram et al., 2012). During this process, Spag4 and Yuri progressively concentrate around the centriolar adjunct, a structure that is formed in the vicinity of the basal body, while it anchors to the nucleus during spermiogenesis. Hence, mutations in Spag4 or Yuri induce sperm nuclear–shaping defects together with disturbed basal body anchoring into the nucleus. The latter defect is independent of the integrity of the KASH domains of the only two KASH domain–containing proteins of Drosophila, Klarsicht and Msp300 (Kracklauer et al., 2010). This suggests that still undiscovered specific components are needed to link SUN domain proteins of the nuclear membrane to the microtubule network and centriolar proteins.

Another important feature of spermiogenesis is the individualization of spermatids, also called spermiation, leading to the elimination of bulk cytoplasmic content and to sperm release (O’Donnell, 2014; Steinhauer, 2015). Individualization takes place once sperm head and tail have elongated completely. In Drosophila, this requires three successive steps: 1) recruitment of a specialized actin-based complex, the individualization complex (IC), 2) migration of this IC associated with membrane remodeling, and 3) elimination of excess material. These two last steps require the activation of a non-apoptotic caspase pathway (Fabian and Brill, 2012). IC migration along the 64 interconnected spermatids relies on the recruitment of specific molecular motors and removes most of the cytoplasm and organelles into a membrane-bound sack, called the cystic bulge, which is discarded as a waste bag when ICs reach the tip of the spermatids (Fabrizio et al., 1998). Disturbing IC assembly or migration is correlated with defective individualization (Noguchi, 2003; Steinhauer, 2015). However, individualization failure can also be the ultimate consequence of various previous defects, including meiosis and flagellar assembly defects (Steinhauer, 2015).

In a two-hybrid screen (J.T. and J.A.L., unpublished data), using the ciliary transition zone protein Chibby (Cby; Enjolras et al., 2012; Vieillard et al., 2016) as bait, we have identified an uncharacterized protein encoded by CG13164, which is highly expressed in the testis (modENCODE, tissue expression data; Brown et al., 2014). This protein was previously described as interacting with syntaxin and named Sip2 (for syntaxin–interacting protein2). Because of a potential confusion between CG13164/sip2 and another gene, SIP2 (septin interacting protein 2/CG9188), and based on the mutant phenotype described hereafter, we renamed this gene salto (French for “somersault”). Although we could not confirm the interaction with Cby, we observed a critical function of this protein in spermiogenesis. We show that Salto is located at several major structures involved in sperm-head architecture and in particular with the sperm manchette/DC and centriolar adjunct. salto-null flies are sterile and show spermiogenesis defects characterized by abnormal nuclear coiling of late spermatids, leading to failure of sperm individualization. Actin cones can form but fail to migrate and elevated caspase activity is observed along the length of the spermatids. Because Salto distribution and mutant phenotype are similar to those observed for Spag4 and Yuri, our work suggests that all three proteins work in a common pathway required for sperm-head morphogenesis. salto hence represents a novel actor involved in spermiogenesis and sperm-head morphogenesis in Drosophila.

RESULTS

salto encodes a testis-specific protein in Drosophila that shows a dynamic distribution during spermatogenesis

We identified Salto in a yeast two-hybrid screen using full-length Chibby protein as bait (see Materials and Methods). All isoforms encoded by salto possess potential coiled-coil domains. Structural analysis indicates that Salto contains a pentapeptide repeat domain (12 repeats of the sequence LQEPN and five of LQD(A/D)(T/N)), found in the uncharacterized YjbI protein from Escherichia coli, and a putative SMC_N domain (Supplemental Figure S1B), found at the N-terminus of SMC proteins involved in chromatin maintenance (Strunnikov and Jessberger, 1999). From the CDART database (Geer et al., 2002), only a few proteins share this domain architecture in the eukaryotic kingdom. Salto is conserved almost exclusively among arthropods and is found in all drosophilids. Outside arthropods, a possible homologue, containing the SMC_N domain but not the pentapeptide repeat domain, could be found only in Ciona (Supplemental Figure S1A). The SMC_N domain is the most conserved part of the protein. The NCBI conserved domain database (Marchler-Bauer et al., 2017) also identifies a putative highly divergent BAR domain, but this homology is weak. Multiple pairwise analysis in comparison with other BAR-domain proteins in Drosophila indicate most similarities with the F-BAR domain containing the protein syndapin (38% similarities, and 19% identities with Drosophila syndapin; accession NP_732563.1).

In Drosophila melanogaster, salto potentially encodes four different transcripts (Figure 1A). We designed a reporter construct expressing the longest salto coding frame fused to GFP and under the control of its own promotor (Figure 1A) to create salto::GFP–expressing transgenic flies. We observed a very weak Salto–GFP signal at the tip of the sensory dendrites in chordotonal organs of embryos, suggesting that salto is expressed at a low level in sensory ciliated neurons (Supplemental Figure S2). Strikingly, salto::GFP expression is strong in the testis. Salto–GFP was detectable in all steps of spermatogenesis, from spermatocytes to fully elongated spermatids (Figure 1, B–H). In spermatocytes, Salto–GFP encages spermatocyte nuclei (Figure 1C, arrowhead). In addition, Salto–GFP forms a halo around centrioles and is located at the primary like cilium above the centrioles (Figure 1C, inset and arrows). Transition zone proteins like Cby also locate at this primary like cilium at this stage (Enjolras et al., 2012; Pratt et al., 2016; Vieillard et al., 2016). At the onset of meiosis, high levels of Salto–GFP are observed at the forming spindle poles and decorate the microtubules (Figure 1D). After meiosis, Salto–GFP decorates a half dome encircling the nuclei and anchored to the basal body (Figure 1E). Such staining was previously described for Yuri and Spag4 as the nuclear “cap” (Kracklauer et al., 2010) and likely corresponds at the round spermatid stage to the fenestrated part of the nucleus surrounded by the microtubule network of the sperm manchette or DC. After the beginning of flagellar elongation, Salto–GFP is relocated both to the centriolar adjunct (labeled with γ-tubulin antibodies; Wilson et al., 1997) and to the acrosome (Figure 1F). When spermatids are fully elongated and before the beginning of individualization, Salto–GFP is mainly observed at the acrosome (Figure 1G). In mature sperm, after individualization, this protein is not detected (unpublished data). We also determined the localization of Salto–GFP after treating testes by colcemid. We observed that Salto–GFP signal at the spindle poles in spermatocytes or at the nuclear cap in early spermatids is strongly reduced when microtubule depolymerization is induced, whereas a distinct Salto–GFP signal along the axoneme is not affected by colcemid. Note that Salto–GFP is enriched at the distal growing tip of the axoneme in this colcemid-treated condition (Supplemental Figure S2, arrowheads). Altogether, these observations indicate that Salto–GFP is dynamically associated with the pericentriolar microtubules during meiosis and with the nuclear-­associated manchette microtubule network required for nuclear shaping during spermiogenesis.

FIGURE 1:

Salto–GFP is dynamically distributed during spermatogenesis. (A) Scheme of the salto locus and of its different transcripts. (B) Whole testis showing that Salto–GFP is present at all stages of spermatogenesis. (C) Spermatocytes express Salto–GFP around the centrioles and at the cilia/TZ (arrows). Salto–GFP is also located around the nuclei (arrowhead). A threefold magnification of the inset (white box) is shown in the middle panel. The right panel shows a representative spermatocyte with centrioles labeled with an anti-Asterless antibody (red); Salto–GFP is found both at the tip of the centrioles (arrow) and around the base (arrowhead). The scheme under the panel represents Salto–GFP localization in spermatocytes in green, nuclei in blue, and centrioles in red. (D) Before the beginning of meiosis, Salto also decorates the spindle poles and microtubules. (E) In round spermatids, Salto–GFP is present as a half ring around the nuclei and surrounds the centriole labeled by γ-tubulin antibody. (F) In late elongated spermatids, Salto–GFP is present both at the acrosome and at the centriolar adjunct. (G) Salto–GFP remains at the acrosome in later stages. Schemes underneath represent Salto–GFP localization in spermatids. B–G, scale bars = 10 µm. (H) Salto–GFP (green) is localized at the centriolar adjunct (labeled by γ-tubulin antibody, red) that forms a ring at the base of the centriole, where it is anchored in the nucleus (blue). Bottom panels are 2.2-fold magnification of the inset shown in the top panel. Scale bars = 2 μm.

sal¹-null mutant males are sterile and show sperm individualization defects

To understand the function of Salto, we generated salto-null mutant flies (sal¹ allele) by CRISPR-Cas9–mediated homologous recombination, replacing all salto coding sequences with the mini-white gene (Figure 2A). We confirmed by sequencing that the entire coding sequence of salto was replaced with the white gene. sal¹ homozygous mutant flies and sal¹ heterozygotes over a deletion that uncovers the salto locus (Df(2R)BSC463) are viable and show no obvious behavioral defects. Females are fertile (unpublished data), but males are sterile and do not produce mature sperm (Figure 2B). Male fertility could be partially rescued by introducing two copies of a salto::GFP–expressing transgene (Figure 2B). This incomplete rescue could be explained either by functional interference of the GFP coding sequence or by reduced Salto–GFP production levels, as even though salto–GFP expression is under the control of the salto promoter, the rescue construct lacks the sal 3′UTR that could be needed to reproduce the full expression characteristics.

FIGURE 2:

sal¹ null mutant males are sterile and show sperm individualization defects. (A) The mini-white gene was inserted into the salto locus by CRISPR-Cas9–induced homologous recombination, eliminating all salto coding sequences. (B) Graph of fertility assays representing the number of progeny for control, rescue, and sal¹ mutant males (n = 11, 13, 25, males respectively). (C) Whole testes labeled for polyglycylated tubulin (gray) and actin (red). Investment cone (IC) migration is altered in sal¹ mutant testes compared with controls. Whereas IC (arrows) are clustered and regularly organized in wild-type testes, their distribution is dispersed in sal¹ mutant testes. The spermatid tails are aberrantly coiled (arrowhead) in the proximal part of the sal¹ mutant testes. (D) Images showing coiled nuclei (Hoechst in magenta) in sal¹ mutant (arrowhead) compare to the long needle-shaped nuclei in control testes (arrowhead). Centrioles are labeled with asterless antibody (green). Scale bars = 10 µm. (E) Quantifications of the status of the nuclei and investment cones (IC) in control (n = 10), sal¹ (n = 9), or rescue (n = 9) testes. Class 1 represents elongated nuclei before IC formation, class 2 is nuclei with assembled IC, class 3 is nuclei with IC starting to migrate, class 4 is migrating IC, and class 5 is disassembling IC in waste bags (see Supplemental Figure S4 for representative images of the categories). Results are presented as average numbers of events observed in testes. Statistical analysis was performed using ANOVA with multiple pairwise comparisons. In each class, statistical significance of variations relative to controls is indicated.

Because we observed no mature sperm in the seminal vesicles of sal¹ mutant males (unpublished data), we looked at their testis organization. Detection of polyglycylated tubulin showed normal elongation, but increased tail coiling of late spermatids inside the testis (Figure 2C). This is suggestive of individualization defects. Individualization requires the formation of investment cones (IC), composed in particular of F-actin (Steinhauer, 2015), which can be detected using fluorescent phalloidin (Figure 2C). Whereas IC were apparently correctly forming at the base of the clustered nuclei, only a few progressing IC could be observed in sal¹ mutant testes (Figure 2C). Strikingly, we observed that nuclei were not properly shaped in late elongated spermatids. Instead of forming long needles, they appeared coiled and twisted (Figure 2D).

To determine whether the sperm tail was correctly formed, we performed an ultrastructural analysis of transverse sections of mutant testes. No ultrastructural defects of the axoneme could be detected (Supplemental Figure S3). However, we observed abnormal morphologies of the mitochondrial derivatives on transverse sections of the spermatids (Supplemental Figure S3). The major derivative often appeared elongated (Supplemental Figure S3) relative to the round mitochondrial derivatives in control testes.

Individualization defects are associated with stalled IC

We quantified the efficiency of IC formation and migration by counting the IC at different steps of progression along the testis, either just assembling (class 2), starting to migrate (class 3), progressing along the testis (class 4), or finally disassembling in the waste bag (class 5). We observed that a larger number of nuclei are not associated with IC (class I) and most IC fail to migrate along the testis in sal¹ testes compared with controls (Figure 2E). IC distribution is fully restored in rescued testes. This clearly indicates that Salto is required for IC assembly and progression. To understand the individualization defects, we also addressed the organization of the ICs. Actin detection by phalloidin showed that ICs form on the nuclei just at the onset of coiling (Supplemental Figure S4A), but disconnect from the curled nuclei and do not migrate. Jar, a myosin VI orthologue, is essential for IC migration and locates at the base of the actin cones (Rogat and Miller, 2002) from the onset of IC migration and all through their migration. We observed that in sal¹ testes, Jar motor protein is localized, as in control flies, at the base of the ICs despite their altered distribution and migration (Supplemental Figure S4B). Hence, sal¹ is not essential for the recruitment of Jar at the leading edge of ICs.

Spermatid individualization also involves nonapoptotic caspase activation. We determined whether caspase activation could be affected in sal¹ mutants. In wild-type cysts, a gradient of active caspase 3 is induced, progressively decreasing from the cystic bulge formed by migrating IC toward the distal end of the flagella (Figure 3A). During IC progression, active caspase is removed from postindividualized spermatid ends. Hence, active caspase 3 is found only distal to the migrating actin cones. We observed that sal¹ mutants display elevated levels of active caspase 3 all along the spermatid cysts (Figure 3, A and B). Cystic bulges are barely formed in comparison with controls (Figure 3, A and B), and even in cysts with migrating ICs, activated caspase 3 is observed both proximal and distal to migrating IC (Figure 3B). This indicates that migrating ICs are ineffective in creating cystic bulges and hence in removing active caspase 3 and cytoplasm in sal¹ testes. This phenotype is rescued by adding two copies of sal::GFP transgene (Figure 3A). To check that elevated levels of active caspase are not due to defects in the distribution of caspase inhibitors, we looked for the localization of Soti and dBruce, two inhibitors of the Cullin complex (Kaplan et al., 2010). These inhibitors form two successive gradients of expression, increasing from the nucleus to the caudal flagellar tip in control testes. In sal¹, both proteins are distributed as in controls (Figure 3C).

FIGURE 3:

salto mutant spermatids exhibit individualization defects with elevated levels of active caspase 3. (A) Testes from wild type, sal¹ mutant, and sal¹ rescued with two copies of salto::GFP were stained with anti-cleaved caspase 3 antibody (CC3, red), polyglycylated tubulin (gray), and Hoechst to visualize the nuclei (blue). The red channel is shown alone on the right. Scale bar = 100 μm. (B) Squashed testes stained for cleaved caspase 3 (CC3, red), actin (green), and nuclei (blue) showing the postindividualized (arrowhead) and preindividualized (arrow) segments of the spermatids. In sal¹ mutants, caspase 3 is active both before (arrowheads) and after (arrows) the IC. Bottom panels show the altered IC distribution in sal¹ testes compared with controls. (C) Visualization of Soti (green) and dBruce (red) proteins at the spermatid tips (arrows and asterisks, respectively) both in controls and in sal¹ mutants. Scale bars = 10 µm.

Hence, our results indicate that sperm individualization defects are likely due to impaired or inefficient IC migration and not altered regulation of the caspase pathway.

Salto is required for sperm nuclear shaping

The most striking defects of sal¹ mutants is the coiling of the nuclei at late stages of differentiation. We first determined the precise onset of nuclear coiling. We observed that coiling only takes place after the beginning of actin recruitment before complete assembly of the actin cones (Figure 4A).

FIGURE 4:

Salto is required for sperm nuclear shaping. (A) Nuclei (white) are coiling at the onset of IC assembly (red) as stained with Hoechst. (B) Transition protein Tpl94D (top panels in green) and protamine Mst35Bb (bottom panels in green) are correctly incorporated into the nuclei (magenta) of both control and sal¹ mutant spermatids. Scale bars = 10 µm. (C) EM sections of sperm heads showing the microtubules of the dense complex (arrows) present on the invaginated side of the nucleus. No differences in the apparent organization of the microtubule network could be observed. Scale bars = 200 nm. (D) EM sections of the centriolar adjunct (arrowhead), which appears irregular in shape and size in sal¹ mutant testes compared with controls. Scale bars = 200 nm. (E) EM sections of nuclei. At intermediate stages and before coiling (top panels), sal¹ nuclei have a rougher membrane (arrowhead) than control nuclei. At late stages (bottom panels), bending of sal¹ mutant nuclei compared with controls is indicated by white arrows.

We checked whether nuclear shaping defects could be associated with altered chromatin remodeling. We did not observe any recruitment defects of Tpl94D, a transition protein incorporated at the histone to protamine transition, nor of the protamine protein Mst35Bb (Rathke et al., 2007). Both are incorporated in sal¹ nuclei as in controls (Figure 4B). By EM, we looked at the structure of the DC and the associated nucleus during spermatid differentiation. We did not observe obvious defects of the DC nor of the fenestrated nuclear membrane. The microtubules of the DC cluster on one side of the nucleus in sal¹ testes and mark nucleus invagination during head elongation (Figure 4C), indicating that nucleus invagination is initiated properly. We only observed variability in the size and shape of the centriolar adjunct that surrounds the basal body. This variability was not observed in control testes (Figure 4D). Also, the nuclear membrane appeared more tortuous in sal¹ testes than in controls, in which the nuclear membrane appears smoother (Figure 4E, arrowheads in top panels for intermediate spermatid stages; arrows in bottom panels for late stages).

We thus addressed whether we could observe functional alterations of known DC components. Several proteins have been described as involved in DC homeostasis in Drosophila (Kracklauer et al., 2010; Texada et al., 2011; Sitaram et al., 2012). Spag4 is a SUN domain containing protein associated with the DC of spermatids in Drosophila. At all maturation steps, there was no alteration of Spag4 localization in salto mutants (Figure 5A). As well, Lis1 is a dynein-associated protein involved in the localization of dynein around the nuclei (Sitaram et al., 2012) and is properly located in sal¹ spermatids (Figure 5B). Two other unrelated proteins, CG4329 and CG12184, that we have characterized as being localized at the dense complex (J.A.L., unpublished results), are also properly distributed in sal¹ mutants (Figure 5B). These results indicate that Salto is not required for the localization of these other components associated with the dense complex in Drosophila.

FIGURE 5:

salto and spag4 functional relationships. (A) Spag4-myc localization in sal¹ mutants during spermatogenesis. Centrioles are stained with γ-tubulin antibody in green and nuclei (Hoechst, in blue). Spag4 is correctly located around the nuclei in spermatocytes and at the dense complex in round and elongated spermatids. It is relocated at the centriolar adjunct in late spermatids. (B) Images of dense complex components: Lis1 (first panels in red), CG4329 (second panels in green), and CG12184 (third panels in green) in controls and sal¹ mutants. Basal bodies (Asterless respectively in green or red) are on rare occasions undocked to the nuclei in sal¹ mutant spermatids (arrows). (C) Quantification of basal body attachment defects in sal¹ early elongating spermatids (n = 100 for controls and 96 for sal¹). (D) Salto–GFP localization in spag4 mutant during spermatogenesis. Salto–GFP is similarly located at all steps of maturation in control or in spag4 mutant spermatids. (E) spag4 mutant spermatids exhibit elevated caspase levels (red) compared with controls. Scale bars = 10 µm.

However, using these cell markers, we observed, on a few occasions, defects of basal body–to–nucleus attachment in sal¹ round spermatids (Figure 5, B and C). Such defects are also observed in spag4 mutant spermatids. Indeed, previously described null mutants for spag4 show phenotypes strikingly similar to sal¹ in the testes, with identical individualization and nuclear-shaping defects (Texada et al., 2008; Kracklauer et al., 2010). We hence looked at the distribution of Salto in the previously described null spag4⁶ mutant flies (Kracklauer et al., 2010), but we did not observe any defects (Figure 5D), indicating that Salto and Spag4 localization do not rely on each other. Interestingly, we noticed that in addition to previously described phenotypes, spag4⁶ mutant flies show elevated caspase 3 activation (Figure 5E), with no cystic bulge or waste bags, as described above in the sal¹ mutant, supporting the hypothesis that both genes are involved in the same biological process. We thus investigated possible genetic interactions between salto and spag4. In sal¹/+; spag4⁶/+ trans heterozygotes, fly fertility was not affected and we observed properly formed nuclei and basal body anchoring. The homozygote sal¹, spag4⁶ mutants show defects similar to those in spag4⁶, indicating that spag4 and salto do not have additive or synergistic actions and thus are likely to work in the same pathway. Altogether, our results indicate that salto and spag4 act in the genetic pathway involved in sperm nuclear shaping and tail-to-nucleus attachment during Drosophila spermiogenesis.

DISCUSSION

Our results show that salto encodes a dynamically located protein with centriolar associated components, the dense body, equivalent to the mammalian nuclear manchette, and the acrosome during spermatogenesis. In agreement with this localization, sal¹ mutants do not exhibit developmental defects, but grow normally, and their only phenotype is male sterility characterized by individualization and sperm head morphogenesis defects.

Both cellular phenotypes were fully rescued by expressing the Salto–GFP fusion protein, even though male fertility was not fully restored. This suggests that reduced fertility of the rescued sal¹ mutant is not linked to the function of Salto in sperm-head shaping and sperm individualization, but could rely on other possible functions of Salto in sperm differentiation that our observations did not reveal.

The Salto C-terminus was isolated in a yeast two-hybrid screen with full-length Cby as bait, and Salto can be found in the same compartment as Cby at the transition zone (Enjolras et al., 2012; Vieillard et al., 2016). However, Cby distribution is not affected in sal¹ mutants, and as well, Salto–GFP distribution is not affected in cby¹ testes (unpublished data). In addition, sal¹ and cby null mutants show completely different phenotypes (Enjolras et al., 2012; Vieillard et al., 2016), indicating that Cby and Salto exert their apparent functions in different pathways.

Salto is required for sperm-head morphogenesis

The most striking defect in sal¹ mutant males is the abnormal coiling of the nuclei. It is important to note that even though we did not fully restore male fertility, nuclear coiling defects were fully rescued by expressing Salto–GFP. Nuclear shaping defects could originate from different impaired pathways. Based on Salto distribution, Salto could be required for proper manchette action to shape the sperm head. Indeed, Salto distribution reflects the microtubule network that cages the nuclei and forms the dense complex. Salto could be required for the integrity of the microtubule network of the sperm manchette. However, using immunofluorescence (unpublished data) or electron microscopy (EM) analysis, we did not identify defects of the microtubule network around the nuclei, nor of other components known to be associated with the dense complex in Drosophila. Therefore, we expect Salto to play a role downstream or parallel to the organization of the microtubule network. Salto possibly harbors a BAR domain, known to be involved in lipid membrane association, and thus Salto could participate in linking the microtubule network to membrane domains (Stanishneva-Konovalova et al., 2016). In agreement with this hypothesis, mutation of the Spag4 protein, which is a nuclear–inner membrane protein, leads to nuclear shaping defects similar to those with sal¹ (Kracklauer et al., 2010) indicating that the tight association of the nuclear membrane to the microtubule network is required for nuclear shaping. Salto could also be involved in the active transport that takes place along sperm manchette microtubules during nuclear reshaping.

It is important to note that Salto is also present at the acrosome during spermatid elongation steps. The incomplete rescue of sal¹ fertility could be due to improper function of Sal–GFP at the acrosome which is required for fertility. In humans, defects in acrosome formation are associated with globozoospermia, characterized by abnormal round spermatid heads (Ray et al., 2017). How acrosome defects contribute to the abnormal shaping of the nuclei is not completely understood. Polarized transport along microtubules is a likely common pathway involved in acrosome and nuclear shaping defects. Among the genes responsible for globozoospermia in humans, the kinesin 14 KIFC1 plays a role in organizing microtubule-associated transport (Zhi et al., 2016). In the lobster, mutations of the orthologue KifC1 induce acrosome malformation and nuclear shaping defects (Ma et al., 2017). In this context, Salto could be involved in microtubule-associated transport, implicated in both acrosome formation and manchette function. The acrosome could be an anchoring platform for microtubules and therefore contribute to normal nuclear shaping. As Salto is also present both at the centriolar adjunct and at the acrosome, it could also be required for linking the microtubule network to both ends of the nuclei. In the absence of Salto, these links would be weakened and the nuclei would coil under the mechanical constraints imposed by microtubule rearrangements. Further work is needed to identify more precisely the partners of Salto during nuclear shaping and to understand how they are involved in this process.

Salto is involved in basal body and centriolar adjunct homeostasis

In the sal¹ mutant, we observed a significant but scarcely penetrant defect in basal body anchoring into the nuclei at the round spermatid stage (4% of cells clearly showed this defect). This indicates that Salto is not essential for basal body anchoring but could play a redundant function with other, yet to be identified, genes required for nuclear basal body anchoring. Strikingly, spag4 mutant spermatids show basal body anchoring defects associated with nuclear coiling as sal¹ mutants. We did not identify genetic interactions in the nuclear anchoring function between salto and spag4, as the double heterozygous mutants show no anchoring defects and double homozygous mutants are similar to spag4 mutants (unpublished data). However, this suggests that Salto could act, like Spag4, to connect the nuclear membrane to the cytoskeleton network. In support of a function of Salto in basal body nuclear anchoring, we observed, by EM, shape variability of the centriolar adjunct, indicating that Salto is needed to organize the basal body/nuclear interface.

Interestingly, the angiotensin-converting enzyme (Ance) is a peptidyl dipeptidase implicated in male fertility. Ance mutants present a defect in nuclear morphology similar to those observed in sal¹. Ance mutant spermatids also show mitochondrial defects, the major and the minor derivatives being misshapen compared with control spermatids (Hurst et al., 2003). We observed similar misshaping of the major mitochondria derivative in sal1 mutants (Supplemental Figure S3), suggesting that Ance and Salto could be involved in convergent pathways. Further studies are required to understand the molecular links between nuclear shaping, basal body anchoring, and mitochondria morphogenesis, but Salto appears to be an important player in these processes.

Origin of the individualization defects in sal¹ mutants

Individualization defects can be a consequence of various impaired pathways (Steinhauer, 2015). We have excluded several of the possible causes, as we did not observe axonemal defects (Supplemental Figure S3) or misorganization of the formation of the IC (Supplemental Figure S4). Defects in individualization can also be a consequence of impaired caspase activation (Arama et al., 2007; Fabian and Brill, 2012). During IC progression, activated caspase forms a gradient from the cystic bulge to the distal cyst tail. We determined that salto mutants present an elevated level of activated caspase from head to tail. We did not observe any distribution defects of the regulators of the caspase pathway, Soti or dBruce, indicating that elevated caspase 3 activation is not associated with reduced levels of inhibition. Because salto and spag4 null mutants show similar profiles of caspase 3 activation, it is likely that this defect is a direct consequence of impaired actin cone progression. Alternatively, mitochondria are directly involved in regulating caspase 3 activation during spermatid individualization, as regulation of caspase 3 activation has been shown to be restricted in the mitochondrial vicinity by a protein (succinyl-CoA synthetase [SCS] b-subunit [A-Sb]) of the mitochondrial outer membrane (Aram et al., 2016). Hence, it is also plausible that the mitochondria-shaping defects observed in sal¹ mutants could be responsible for caspase overactivation.

Nevertheless, individualization defects most likely originate directly from impaired manchette organization and nuclear coiling. Indeed, analysis of mutants for salto, spag4, and yuri and also of the proteasomal subunit prosα6T indicates that coiling of elongated nuclei is always correlated with failed IC progression (Zhong and Belote, 2007; Kracklauer et al., 2010; Texada et al., 2011). We observed that nuclear coiling appears after the onset of IC formation in sal¹ mutants. The signals that trigger IC formation and progression are unknown, but these observations suggest that proteins associated with the sperm manchette/dense complex/centriolar adjunct are all required for IC progression. Altogether, this supports the conclusion of a tight association between sperm manchette organization and IC progression.

In conclusion, we show that Salto plays a critical role in organizing the sperm head. The similarities between spermatid individualization processes and nuclear shaping in Drosophila and mammals suggest that further studies of sperm differentiation in Drosophila may provide new insights into the etiology of some forms of human infertility.

MATERIALS AND METHODS

Fly stocks and maintenance

The following lines were obtained from Bloomington Drosophila Stock Center: vasa::Cas9 (Bl#51323); spag4 mutant: w[*];TI{w[+mW.hs] = TI}spag4⁶/CyO (Bl#29978); reporter gene Spag4::Myc: w[*]; P{w[+mC] = spag4.6xMyc}3/TM6B, Tb[1] (Bl#29981); deficiency that covers the salto locus (Bl#24967); Lis1::cherry for dense complex marker (Bl#57339). For controls, we used w¹¹¹⁸ flies to compare background fluorescence intensity with Salto–GFP. w¹¹¹⁸; sal¹/Cyo were used as controls to compare with sal¹ mutant phenotypes. The flies were raised on standard media between 21° and 25°C.

Yeast two-hybrid screen

Total RNA from wild-type Drosophila testis was extracted using TRIzol reagent (Thermo Fisher Scientific). A testis-specific cDNA library was constructed by Hybrigenics, who further performed a yeast two-hybrid screen using Cby as bait.

Generation of sal¹ allele

The sal¹ allele (CG13164) was generated by CRISPR/Cas9-induced homologous directed repair (Gratz et al., 2014). Two guide RNAs (gRNAs) were selected using the http://tools.flycrispr.molbio.wisc.edu/targetFinder/website: 5′-CCATTTACACTAGACGTAGTTCA-3′ and 5′-CCGTCGA­GCAGTTGATGCTGCAG-3′ (protospacer-adjacent motifs are underlined). Oligos were phosphorylated by T4PNK (New England Biolabs) and annealed. Double-stranded 5′ gRNA and 3′ gRNA were cloned in the BbsI sites of pBFv-U6.2 and pBFv-U6.2B vectors, respectively (Kondo and Ueda, 2013). 5′ gRNA was further subcloned in the EcoRI–NotI sites of pBFv-U6.2B to express the two gRNAs from one vector. The 1.5 kb 5′ homologous arm was amplified by PCR with

• Forward (F)-1 5′-TAATTCGCTAGCGTATTCTCGACAAGATTTGG-3′

• Reverse (R)-2 5′-TAATTCCGTACGTGGTACAAAAATGCTTTCT­G-3′.

The 1.5 kb 3′ homologous arm was amplified by PCR with

• F-3 5′-TAATTAGATCTCGGCTTAATTCGCCCCGC-3′

• R-4 5′-TAATTCCCTAGGAGTTGGTCAACAATGTTAGG-3′.

The resulting 5′ and 3′ homology arms were cloned into the BsiWI–NheI and AvrII–BglII sites, respectively, of pJT38 (pRK2 plasmid [Huang et al., 2008], with an attB cassette). The two vectors (gRNAs and homology arms) were injected into vasa::Cas9 embryos. Flies were crossed to w; Bl/CyO virgin females, and the offspring were screened for red-eyed flies. Homologous recombination was checked by PCR.

Generation of Salto and of CG4329 and CG12184 GFP-expressing transgenes

For each gene, the entire coding sequence and at least 1 kb of upstream regulatory sequences were cloned in frame with GFP in the pJT61 vector (Vieillard et al., 2016). Integration platforms and primers used were for salto::GFP (89E11[III])

• F-5 5′-CGAAGTTATGCTAGCGGATCCTAATTGTCTTAATTATCCGCATTG-3′ and R-6 5′-GAATTAGCGGCCGCGAAATGGCGGGGCGAATT-3′;

for CG4329::GFP (89E11[III])

• F-7 5′-TAATTCGCGGCCGCCGATTCTGATTTGCGCAAT-3′; R-8 5′-TAATTCGCGGCCGCCGATTCTGATTTGCGCAAT-3′;

for CG12184::GFP (53B2[II]), 2 PCR fragments of 2.5 kb with regulatory and coding sequence were cloned into pJT61 using a GIBSON assembly cloning kit (NEB):

• F-9 5′-ATTGGGAATTCGTTAACAGATCTACCACTCATCAGTACGATG-3′; R-10 5′-GCGCTGTTTAGCTTTCCTTTGGTCTCTCATGAGTGGC-3′; F-11 5′-GGGAGCAGCAAGAAATTCAGCCACTCATGAGAGACC-3′; R-12 5′-CGAGCCGCGGCCGCAGATCTT CGATCTTCTCTAGATGATTAC-3′.

The resulting constructs were integrated by PhiC31-mediated recombination at attB platforms. All transgenic lines were obtained from BestGene.

Immunofluorescence on Drosophila testes squashes

Testes from young adult flies (between 0 and 2 d old) were dissected in phosphate-buffered saline (PBS)1X and fixed 15 min in PBS1X/paraformaldehyde 4%. For squash preparation, testes were squashed between coverslip and slide and frozen in liquid nitrogen. The coverslip was removed, and the slide was soaked for 2 min in ethanol 100% at –20°C. For both whole testis and squash immunofluorescence, testes were permeabilized 15 min in PBS1X/Triton 0.1% (PBT) and blocked 2 h in PBT/bovine serum albumin 3%/normal goat serum 5%.

Primary antibodies were incubated in blocking buffer overnight at 4°C. Fixed testes were then washed in PBS1X and incubated 2 h in secondary Alexa Fluor antibodies diluted in PBS1X. Slides were washed in PBS1X and rinsed in ultrapure water. Slides were mounted using Vectashield containing Hoechst (1:1000). Antibodies used were the following: guinea pig anti-Asterless (1/30,000; gift from C. Rogers, University of Arizona, Tucson, AZ; Klebba et al., 2013); mouse Axo49 anti–polyglycylated tubulin (1/100; provided by N. Levilliers, Laboratoire de Biologie Cellulaire 4, CNRS, Université Paris-Sud, Orsay, France); rabbit anti-Tpl94D (1/100) and Mst35Bb (1/200) kindly provided by B. Loppin, LBBE, CNRS UMR 5558 (Kimura and Loppin, 2016); mouse anti–alpha-tubulin DM1a (1/500) (Sigma Aldrich); mouse anti-Myosin VI/jaguar (1/200) (Kellerman and Miller, 1992), kindly provided by Stéphane Vincent (ENSL); rabbit anti-cleaved caspase 3 (CC3) (1/100) (Cell Signaling); guinea pig anti-Soti (1/100) and rabbit anti-dBruce (1/500, kindly provided by E. Arama); mouse anti-LaminC (1/25) (DHSB).

For colcemid treatments, testes were incubated 1 h in 10 μM colcemid before being processed as above.

Most slides were imaged using an IX83 microscope from Olympus equipped with an iXon Ultra 888 EMCDD camera from Andor and IQ3 software from Andor. A PlanApo N Apochromat 10× or a 60 × 1.42NA objective from Olympus was used for all acquisitions.

All images were processed with ImageJ. Figures were created with Adobe Photoshop. Only contrasts and offset were adjusted.

Fertility test

One or 2 d–old males were crossed individually with two w¹¹¹⁸ virgin females. After 5 d, crosses with at least one dead fly were eliminated from the test. Mated flies were discarded from conserved crosses. All flies that hatched out from the crosses were counted.

EM

Samples were processed as described previously (Enjolras et al. 2012; Vieillard et al. 2016).

Sequence alignment

Protein sequences were obtained from the National Center for Biotechnology Information (NCBI). We identified orthologues for Salto/CG13164 (NCBI protein accession number AAF58537) in Drosophilidae: D. simulans, GD25850 (KMY93149); D. yakuba, GE13459 (EDW90806); D. erecta, GG22589 (EDV56349); D. sechellia, GM20369 (EDW47532); and in the bilateria C. savignyi, Uniprot accession number H2YNR4 using orthoDB v9.1 (Zdobnov et al., 2017). Protein with a similar SMN domain were also found in other arthropods: Hyalella azteca (accession n° XP_018027323.1), Cyphomyrmex costatus (accession n° XP_018402034.1), and Amyelois transitella (accession n° XP_013184881.1).

Alignment was produced using PRALINE (PSI-BLAST; Simossis and Heringa, 2005).

A tree was made using the distance method on Seaview v4.7 software (Gouy et al., 2010).

Quantification and statistics

For basal bodies–to–nuclei anchoring, we used CG4329–GFP to visualize the half ring of microtubules around the nuclei (stained with Hoechst) and labeled centrioles with γ-tubulin antibody. The centrioles nonanchored to the nuclei were counted on round spermatids. The experiment was repeated four times and 20–25 cells were counted for each.

Results are represented as scatterplots with the mean and SD on all figures. For Figures 2B and 5C, statistical significance was determined by a two-tailed unpaired Student’s t test. For Figure 2E, statistical significance was determined using analysis of variance (ANOVA) with multiple pairwise comparisons. (Prism 6 software; ns, p > 0.05; *, p ≤ 0.05; **, p ≤ 0.01; ***, p ≤ 0.001; ****, p ≤ 0.0001.)

Abbreviations used:

Cby

Chibby

DC

dense complex

EM

electron microscopy

IC

individualization complex, investment cone