Combover interacts with the axonemal component Rsp3 and is required for Drosophila sperm individualization

Proper differentiation of germline cells into eggs and sperm is essential for the perpetuation of a species. Sperm cells in mammals and Drosophila melanogaster develop from germline stem cells that produce mitotic spermatogonia, which ultimately undergo meiosis and terminal spermatid differentiation. Drosophila has long been a powerful model organism for studying spermiogenesis, which begins after meiosis, when the syncytial spermatids derived from a single gonialblast (the differentiating stem cell daughter; Fig. 1A) begin morphological changes required for their differentiation, including mitochondrial differentiation, flagellar elongation, nuclear compaction and acrosome formation (reviewed by Fabian and Brill, 2012; Fuller, 1993; Hermo et al., 2010a,b). In Drosophila, after meiosis is completed, mitochondria aggregate around the basal body on one side of the nucleus. Subsequently, the mitochondria fuse to form the ‘Nebenkern’ (Tokuyasu, 1975). As differentiation proceeds, spermatids form acrosomes, their nuclei remodel and compact, and sperm tails (flagella) elongate (Fig. 1A,B). During spermatid elongation, both axonemal microtubules (MTs) and the mitochondrial Nebenkern extend to form the flagellum, composed of a central axoneme flanked by major and minor mitochondrial derivatives (Fuller, 1993). Drosophila sperm axonemes are structurally similar to other axonemes, containing a central MT pair ringed by nine outer MT doublets (Fabian and Brill, 2012).

In Drosophila, as in mammals, incomplete cytokinesis during sperm development leads to cytoplasmic sharing between sister spermatids (Greenbaum et al., 2011). Following terminal differentiation, the inter-spermatid cytoplasmic bridges are dissolved and the spermatids' cytoplasmic contents are removed (Fabrizio et al., 1998; Tokuyasu et al., 1972a). This process in Drosophila, called individualization, is carried out by the actin-rich individualization complex (IC) that first forms adjacent to the needle-shaped spermatid nuclei, which by this point are clustered in the basal testis (Fig. 1B,C). The IC is composed of 64 actin cones, one for each spermatid nucleus of the germline cyst (Fabrizio et al., 1998; Tokuyasu et al., 1972a), and it shares similarities with actin comets found on endocytic vesicles (Bazinet and Rollins, 2003). As individualization proceeds, the actin cones of the IC move synchronously away from the nuclei toward the apical domain of the testis (Fig. 1B), traversing the spermatid tails (Fig. 1C, arrowheads) until they reach the end of the flagella (Noguchi and Miller, 2003). Each moving IC generates a growing cystic bulge that contains the extruded cytoplasmic contents of the individualizing cyst. When the IC and associated cystic bulge reach the end of the flagella, they form a waste bag (WB) full of discarded organelles (Fig. 1C, arrow), degradation of which may occur by an apoptosis-like program (Arama et al., 2003; Huh et al., 2004; Tokuyasu et al., 1972a). As a result of individualization, each streamlined spermatozoon resides within its own plasma membrane, most of its cytoplasm has been extruded, and it no longer maintains connections to its sister spermatids (Tokuyasu et al., 1972a). Once this process has been completed, mature sperm are coiled in preparation for release into the seminal vesicle, and WBs and abnormal sperm are eliminated within the base of the testis (Tokuyasu et al., 1972b).

Elegant prior studies have provided a detailed view of the polarized IC structure. At the leading edge of the actin cones, F-actin filaments form a meshwork under the control of the Arp2/3 actin-nucleating complex (Noguchi et al., 2008; Rogat and Miller, 2002), whereas at their rear, F-actin is organized into parallel bundles by the activity of the actin-bundling proteins Quail/Villin, Chickadee/Profilin and Singed/Fascin (Noguchi et al., 2006, 2008). IC activities are thought to be segregated, with the front meshwork being responsible for cytoplasmic extrusion and the rear bundles responsible for movement along the sperm tails. During individualization, cones accumulate actin (Noguchi et al., 2006, 2008), and actin polymerization, but not myosin motor activity, is essential for IC progression (Mermall et al., 2005; Noguchi et al., 2006, 2008; Noguchi and Miller, 2003; Tokuyasu et al., 1972a; reviewed by Steinhauer, 2015).

We previously identified combover (cmb), a gene encoding two major protein isoforms that are predicted to be largely intrinsically disordered (Mészáros et al., 2018) and lack any known domains, as a Rho kinase substrate that acts as a planar polarity effector (PPE) to affect Drosophila wing hair (trichome) formation (Fagan et al., 2014). During pupal development, each cell of the fly wing forms an actin-rich hair, surrounded by membrane, that points towards the distal tip of the wing (reviewed by Adler and Wallingford, 2017; Butler and Wallingford, 2017; Maung and Jenny, 2011). Polarized wing hair formation is governed by the non-canonical Wnt-planar cell polarity (PCP) pathway (reviewed by Adler and Wallingford, 2017; Butler and Wallingford, 2017; Maung and Jenny, 2011), during which core PCP proteins including Frizzled (Fz), Dishevelled (Dsh), Van Gogh (Vang), Prickle (Pk), Diego (Dgo), and Flamingo (Starry night) localize asymmetrically across proximal-distal wing cell borders. PCP-generated asymmetry leads to proximal enrichment of the PPE proteins Inturned, Fuzzy, Fritz and Multiple wing hairs (Mwh), which prevent ectopic wing hair formation (Adler et al., 2004; Strutt and Warrington, 2008). Additionally, Mwh prevents the formation of multiple prehairs, and it bundles actin to restrict hair formation to a single hair during trichome outgrowth (Lu et al., 2015). Although cmb mutants show no wing phenotype, Cmb overexpression causes formation of a multiple hair cell (MHC) phenotype that is enhanced by removal of one gene dose of Rok or the PPE genes, including mwh (Fagan et al., 2014). Indeed, the two major isoforms of Cmb, PA and PB, physically interact with Mwh, and the MHC phenotype observed in mwh single mutants is partially suppressed in mwh cmb double mutants. It was therefore suggested that Cmb promotes actin-based wing hair formation under control of Rok, a function that is antagonized by Mwh (Fagan et al., 2014).

We noticed that cmb mutants are male sterile. Here, we find that this is due to a requirement for cmb in the germline, in the actin-based process of spermatid individualization. Detailed histochemical and electron microscopic analyses of cmb mutant testes show that early cyst formation is normal. Furthermore, ICs form normally and start to move but do not remain properly aligned, and, although they initially recruit actin at normal levels, they fail to maintain and accumulate actin during IC progression. Despite evidence that Cmb affects actin-related processes (wing hair formation in the pupa and individualization of sperm), Cmb does not directly interact with actin in co-immunoprecipitation (CoIP) or co-sedimentation assays. Furthermore, we show that the role of cmb during individualization is independent of and distinct from PCP genes. Significantly, Cmb interacts with the axonemal component Radial spoke protein 3 (Rsp3/CG32392), knockdown of which causes individualization phenotypes highly similar to those of cmb mutants, suggesting that Cmb coordinates IC movement with the axoneme during biogenesis of functional sperm.

cmb mutant males were completely sterile, as individual homozygous cmb^KO (cmb^−/−) males (Fagan et al., 2014) crossed to w¹¹¹⁸ females reproducibly did not yield progeny, in contrast to wild-type or heterozygous males, which gave rise to an average of 67±19 and 63±24 offspring per male, respectively (Fig. 1D). To investigate the cause of the male sterility, testes from cmb mutants were examined cytologically for each stage of spermatogenesis. α-Spectrin staining revealed normal round spectrosomes in the germline stem cells and gonialblasts (compare arrowheads in sibling controls with homozygous cmb^KO mutants in Fig. 2A and 2B, respectively), as well as normal branched fusomes in mitotic spermatogonia (Fig. 2A,B, arrows) of cmb mutants, suggesting that initial germline divisions are normal. Consistent with this, ring canals connecting the spermatogonia and spermatocytes appeared normal (Fig. S1). Phase-contrast imaging of squashed testes was used to visualize spermatids following meiosis (Fig. 2C,D). Each phase-light nucleus (Fig. 2D, arrowhead) was accompanied by a phase-dark mitochondrial Nebenkern of similar size (Fig. 2D, arrow) in cmb^KO spermatids, indicating normal meiotic cytokinesis. Following meiosis, spermatid elongation was normal in homozygous cmb^KO mutants, as assessed by the presence of normal elongation complexes at the ends of the spermatid tails, decorated with α-Spectrin (Fig. 2E,F; Ghosh-Roy et al., 2004), and compacted bundled spermatid nuclei were visible in the basal region of cmb^KO mutant testes (e.g. Fig. 2I,J, DAPI). Several other aspects of spermiogenesis in cmb mutants also proceeded normally, including cytoplasmic caspase activation (Fig. 2G,H) and initial formation of ICs around the bundled spermatid nuclei (Fig. 2I,J, arrowheads; see also below and Fig. 5A-C). However, in cmb mutants the seminal vesicles were empty, indicating the absence of mature sperm (compare heterozygote in Fig. 2K with homozygous mutant in 2L).

Despite apparently normal IC formation and caspase activation in cmb^KO mutants, the advancement of spermatid individualization was highly perturbed. In wild-type or heterozygous cmb testes, the actin cones of each IC progressed synchronously towards the end of the sperm tails, ultimately forming WB structures in the apical domain of the testis (Fig. 3A; normal ICs and WBs quantified in Fig. 3I, as described in Ben-David et al., 2015; see Fig. 1B,C for schematics). In contrast, in cmb^KO mutants, ICs became scattered and disorganized as they progressed along the spermatid tails away from the bundled nuclei (Fig. 3B). Whereas wild type and heterozygous sibling controls showed 8-11 normally progressing ICs and WBs per testis on average (see also Ben-David et al., 2015), cmb^KO mutants averaged fewer than two normally progressing ICs per testis (Fig. 3I). Together, our data demonstrate that male germline cells proceed through spermatogenesis normally until the individualization stage in cmb mutants. Individualization is initiated, with ICs forming and caspases being activated, but IC movement is abnormal. The individualization defect most likely explains the male sterility of the mutant.

The cmb gene encodes four transcripts. The three longer transcripts differ only slightly, with cmb-RA retaining an intron that is removed in cmb-RC and cmb-RD, and all three encode very similar protein isoforms (Fagan et al., 2014). The shortest transcript, cmb-RB, encodes a short Cmb-PB protein isoform that is fully contained within the long isoform Cmb-PA, and both of these isoforms cause similar MHC phenotypes when overexpressed in the wing (see also Fig. S3C for schematic; Fagan et al., 2014). In order to test whether Cmb is required in the germline or the soma and to confirm that the sterility phenotype is indeed due to loss of cmb, we performed rescue experiments specifically re-expressing the Cmb-PA and Cmb-PB isoforms in soma or germline exclusively using traffic jam-GAL4 (tj>) and bag of marbles-GAL4 (bam>), respectively (Siddall and Hime, 2017; White-Cooper, 2012). Expression of the PA isoform in the male germline resulted in robust rescue of both the fertility and individualization phenotypes of cmb mutants (Fig. 3C,F,I-K), whereas expression in the somatic cyst cells did not rescue (individualization phenotype quantified in Fig. 3I). In contrast, expression of the PB isoform only partially rescued fertility: although 20 of 25 single male crosses produced offspring, their number was much lower than in the case of rescue with Cmb-PA (Fig. 3J,K). Consistent with this, quantification of ICs showed that individualization was only partially rescued (Fig. 3D,I) and few mature sperm were observable in the seminal vesicles (Fig. 3G). Expression of each transgene was confirmed by immunofluorescence (Fig. S2A-C). Note that overexpression of Cmb-PA or Cmb-PB in wild-type males did not cause any defects and thus the remaining fertility defects in bam>cmb-PB testes were not due to a gain-of-function or dominant-negative effect (not shown).

Cmb is a phosphorylation target of Rho kinase in vitro (Fagan et al., 2014). We were unable to assess directly a phenotype of Rho kinase (Rok) mutants in the male germline as Rok is on the X chromosome, making it impossible to generate germline clones, and neither RNAi-mediated knockdown nor the expression of a kinase-dead form of Rok resulted in a spermatogenesis defect. In addition, overexpressing a dominant-negative RhoA (Rho1), the GTPase that activates Rok (Winter et al., 2001), resulted in scattered spermatid nuclei and highly abnormal ICs, likely as a consequence of pleiotropic, earlier defects (Fig. S2F). Nevertheless, mutation of the five phospho-acceptor serines/threonines to alanines (including an additional sixth, neighboring threonine; Fagan et al., 2014) abrogated the rescuing activity of Cmb-PA (Fig. 3E,H-J; Fig. S2C). Altogether, we conclude that the longer PA isoform is essential for spermatid individualization and suggest that its phosphorylation is important in this process.

Electron microscopy confirmed the individualization phenotype of cmb^KO mutants. In pre-individualization spermatids, mitochondrial derivatives and axonemes were normal (Fig. 4A,B). However, in contrast to the fully individualized spermatids devoid of cytoplasm seen in wild-type control testes (Fig. 4C,E), cmb mutant spermatids at the individualization stage, recognized by the dense crystalline array within the major mitochondrial derivatives, were still surrounded by cytoplasm (Fig. 4D,F). Defects were also observed in reduction of the minor mitochondrial derivatives, which appeared enlarged relative to wild type at comparable stages (Fig. 4C,D, arrows), a characteristic of individualization failure (Tokuyasu et al., 1972a). Axonemes remained normal in cmb mutant spermatids at the individualization stage (Fig. 4F, compare with control in 4E). Despite the high penetrance of individualization defects in cmb mutants, some cysts displayed a mixture of both improperly and properly individualized spermatids (Fig. 4G, enlarged in 4H), a situation not found in wild type (Tokuyasu et al., 1972a). Altogether, our data suggest that spermatid individualization is the primary defect in cmb mutant testes.

In addition to being scattered, cmb mutant ICs showed structural defects. Actin cones still associated with spermatid nuclei were not obviously aberrant and contained normal levels of actin (Fig. 5A,B; actin intensity quantified in Fig. 5C; see also schematics in Fig. 1B,C), suggesting that ICs initially formed normally. However, once they had moved away from the nuclei, most ICs in the cmb mutant were abnormal, and WBs containing the expunged cytoplasm were rarely found (quantified in Fig. 5D). Abnormalities among the cmb mutant ICs included reduced phalloidin staining, suggesting lower actin levels, and scattered actin cones (compare heterozygote in Fig. 5E with mutant in 5F and 5G). In some instances, cmb mutant actin cones appeared to lack their front flat edges, showing an abnormal needle-shaped morphology (Fig. 5H). In other cases, cmb mutant cones were shorter than normal and appeared to lack their rear tails (Fig. 5I). Elongated, needle-shaped actin cones are reminiscent of those observed in Arp3 (a member of the Arp2/3 complex required for the formation of branched actin structures) mutants, whereas shorter ICs are similar to those observed in mutants of the actin-bundling genes fascin (singed) and villin (quail) (Noguchi et al., 2008).

To characterize further the actin cones in the cmb mutant, we examined other cone-associated proteins. Myosin VI (Jar) localizes to the front edges of the cones, where it is thought to be involved in organizing the meshwork of actin filaments (Fig. 6A-A″; Noguchi et al., 2006, 2008). cmb mutant cones still contained Myosin VI, even when phalloidin staining was strongly reduced (Fig. 6B-B″). Similarly, the actin-bundling proteins Quail and Singed, which localize to the rear tails of the actin cones (Fig. 6C-C″,E-E″; Noguchi et al., 2008), were still present in roughly the right position in cmb mutant cones, even when cone morphology was abnormal (Fig. 6D-D″,F-F″). Cones associated with spermatid nuclei were organized normally, with a rear domain marked by Singed protein and a front domain free of Singed staining (compare heterozygote in Fig. 6G-G″ with mutant in 6H-H″). Thus, the main defect in cmb mutants is abnormal actin accumulation and/or maintenance in individualization cones as the ICs progress along the spermatid tails.

As actin is aberrant in ICs of cmb mutants, we tested whether Cmb could bind actin in CoIP and co-sedimentation assays. As seen in Fig. S3A, neither GFP-tagged Cmb-PA nor Cmb-PB was able to co-immunoprecipitate V5-tagged actin from transfected HEK293 cells. We then tested purified, overlapping GST-tagged subfragments of Cmb-PB that had been used previously to map Rok phosphorylation sites (see schematic in Fig. S3C) (Fagan et al., 2014) for direct binding to muscle actin. The C-terminal GST-SX fragment was not soluble after a 150,000 g preclearing centrifugation, and neither GST-Cmb-ES nor GST-Cmb-BB co-sedimented with actin (compare supernatant and pellet fractions with the αActinin positive control in Fig. S3B). Thus, our experiments indicate that Cmb likely does not directly interact with actin.

Two-hybrid analyses have shown that Cmb can interact with the Radial spoke protein Rsp3 (CG32392) and Pka-R1 (CG42341), the type 1 regulatory subunit of Protein kinase A, two proteins that also have been shown to interact with each other in the same study (Hu et al., 2017; Thurmond et al., 2019). Rsp3 dimers form an assembly scaffold for the radial spokes connecting the outer MT doublets to the central MT pair of axonemes (Huang et al., 1981; Luck et al., 1977; Porter and Sale, 2000; Wirschell et al., 2008; Yang et al., 2006). Rsp3 also is a Pka-anchoring protein (AKAP) in Chlamydomonas and vertebrates and is known to interact with Pka (Gaillard et al., 2001, 2006; Jivan et al., 2009). Indeed, we found that Rsp3 and Cmb-PA co-immunoprecipitated from HEK293 cells (Fig. S4).

Intriguingly, RNAi-mediated knockdown of rsp3 with two independent, non-overlapping RNAi constructs caused IC phenotypes similar to those of cmb mutants (compare control in Fig. 7A with knockdowns in 7B,C). Specifically, ICs were scattered, and the number of normal ICs and WBs were reduced by more than 50% (quantification in Fig. 7E). Similarly, as with cmb mutants, the percentage of total ICs that had abnormal morphology was dramatically elevated compared with controls (Fig. 7F). As in cmb mutants (Fig. 2L), no mature sperm were found in seminal vesicles (compare control in Fig. 7G with 7H,I). Furthermore, one of the two available RNAi lines targeting Pka-R1 also reduced ICs and WBs, although with weaker expressivity (Fig. 7D,E; it did not increase the number of abnormal ICs to a statistically significant level; Fig. 7F). In that case though, mature sperm formed (Fig. 7J). Similar results were obtained upon co-expression of Dcr2 to enhance the knockdown efficiency (not shown). Importantly, acetylated (stable) α-tubulin staining was normal in elongated spermatids in all knockdown genotypes (Fig. 7K-N) and cmb mutants (Fig. 7O). In light of the known function of Rsp3, the similarity between the phenotypes suggest that Cmb, Rsp3 and Pka-R1 link the ICs to the axonemes in a manner that is required for proper IC progression and stability.

Because cmb acts as an effector downstream of the PCP pathway in the wing (Fagan et al., 2014) and PCP genes are known to affect spermiogenesis in mammals (Chen and Cheng, 2016; Chen et al., 2016), we tested whether PCP genes were expressed in testes and were required for fertility or spermatid individualization. By qPCR, core PCP genes frizzled (fz), diego (dgo) and prickle (also known as spiny leg) were expressed at much lower levels in testes compared with imaginal discs, where they are known to be required (Fig. S5A) (Adler and Wallingford, 2017; Butler and Wallingford, 2017; Maung and Jenny, 2011). In contrast, Van Gogh (Vang), Rok and mwh were expressed at comparable levels, whereas dishevelled (dsh) and cmb were expressed at much higher levels in testes than in imaginal discs (Fig. S5A,B). None of the PCP genes tested showed reduced fertility in single male crosses, except, to our surprise, the PCP-specific, viable allele dsh¹ (Fig. S6A), a phenotype that, to our knowledge, has not been reported previously (note though that the number of offspring of fertile dsh¹ males was not different from wild type; Fig. S6B). The reduced fertility of dsh¹ males was partially rescued by expression of a dsh cDNA under control of its endogenous promoter (Fig. S6C,D). We then assessed individualization in testes directly. Mutants of fz, Vang, and the pk¹ and sple¹ alleles of the pk gene showed 8-11 normal ICs per testis on average (Fig. 8, quantified in 8I), and fewer than 10% of ICs in each mutant had abnormal morphology (Fig. 8J), consistent with the range of normal ICs seen in various control genotypes (Ben-David et al., 2015). The dgo mutant showed fewer normal ICs compared with controls (Fig. 8C,I), but no increase in the number of abnormal ICs (Fig. 8J). Thus, unlike cmb, the core PCP genes fz, Vang, pk and dgo are not required for IC progression. In the dsh¹ mutant, testes were short (Fig. 8B), and 21% of ICs that were present were abnormal (Fig. 8I,J; note that the number of normal ICs and WBs, but not the percentage of normal ICs, could be partially rescued; Fig. S6E,F), resembling the normal WBs found at the apical ends of spermatid tails in wild type. This suggests that dsh is required for spermatid elongation rather than individualization per se. Likewise, mwh, the PPE gene in the wing, was not required for male fertility or spermatid individualization (Fig. 8H-J; Fig. S6A,B). Although cmb antagonizes mwh during wing hair formation, as reflected by the suppression of the mwh MHC phenotype in mwh cmb double mutants, mwh does not alter the fertility of cmb males (Fig. S6A,B). Thus, cmb acts independently of the PCP pathway in the testis and likely downstream of distinct signaling pathways in different developmental contexts.

Originally, Cmb was identified in a genome-wide screen for Rho kinase (Rok) substrates (Fagan et al., 2014). Rok stimulates the actin cytoskeleton in multiple ways. Activated by RhoA, Rok phosphorylates Myosin II regulatory light chain and Myosin phosphatase, both leading to an increase of actin/myosin contractility (reviewed by Amano et al., 2010; Rath and Olson, 2012; Riento and Ridley, 2003). In addition, Rok's phosphorylation of Lim kinase results in inhibition of cofilin and concomitant stabilization of actin. As mentioned above, we were unable to directly assess a function of RhoA and Rok during spermatogenesis. However, we found that, in contrast to wild type, a version of Cmb in which the five Rok phosphorylation sites identified on Cmb in vitro (Fagan et al., 2014) were mutated to alanine fails to rescue the phenotype of cmb mutants (Fig. 3E), even though it was expressed in a similar pattern to the wild-type transgene in the developing testes (Fig. S2). It is thus appealing to speculate that Rok may activate Cmb during spermatogenesis. This is distinct from the wing, where an inhibitory relationship was suggested by Rok loss of function dominantly enhancing the MHC phenotype caused by Cmb overexpression (Fagan et al., 2014). Therefore, different functions of Cmb in distinct tissues may be controlled differently by Rok. Alternatively, Rok affects wing hair formation in multiple ways (Winter et al., 2001), and thus the genetic interaction detected in the wing may result from integration of several effects on more than one Rok phosphorylation target.

Orthologs of the PCP proteins Dsh and Vang are emerging as important regulators of mammalian spermatogenesis. For instance, knockdown of PCP proteins Vangl2 or Dvl1-3 in rat leads to defects in transport and maturation of sperm across the seminiferous epithelium due to disrupted actin and microtubule cytoskeletons in somatic Sertoli cells, which results at least in part from altered expression of cytoskeletal regulatory proteins (Chen and Cheng, 2016; Chen et al., 2016; Chen et al., 2018; Li et al., 2019). To our knowledge, no male fertility phenotype has been reported for PCP genes in Drosophila. As Cmb physically and genetically interacts with PCP effectors (Fagan et al., 2014), we tested whether PCP genes show a testis phenotype. We found that dsh¹ and dgo³⁸⁰ alleles show a mildly reduced number of normal ICs but only dsh¹ mutants show reduced male fertility. Interestingly, though, the phenotype of dsh¹ mutant males, shorter testes, differs from the reduced actin content and IC scattering found in cmb mutants. The dsh phenotype is thus likely mechanistically distinct from cmb and possibly also from classical PCP signaling, as Fz, the key PCP receptor, is expressed at much lower levels in testes than in imaginal discs and fz mutants show no obvious spermatogenesis defects (Fig. S5, Fig. 8). Our data do not distinguish whether Dsh acts in the soma or germline. PCP proteins also regulate ciliogenesis in numerous vertebrate cell types (Adler and Wallingford, 2017; Mirvis et al., 2018). The short testes observed in dsh¹ mutants may suggest a role in biogenesis of the axoneme, a primary cilium, but future work will be required to determine how Dsh affects Drosophila spermatogenesis in detail.

Because Cmb was identified as affecting actin organization in response to PCP signaling in the wing (Fagan et al., 2014), we hypothesized that the male sterility of cmb mutants could be due to an actin defect in sperm. Indeed, our results demonstrate that Cmb is essential in the male germline for the actin-dependent process of spermatid individualization. In the cmb mutant, ICs organize themselves normally with roughly normal actin levels around the compacted basal spermatid nuclei (Fig. 5A-C). However, as they progress, they present with numerous defects, the most obvious of which is the scattering of actin cones within individual ICs. Close inspection reveals structural defects within the cones as well, including strikingly reduced actin accumulation in both the cone fronts and the rear bundles. However, in cmb mutant cones, the rear bundling proteins Quail and Singed (Noguchi et al., 2008), as well as Myosin VI, an important component of the cone fronts (Rogat and Miller, 2002), localize fairly normally (Fig. 6). Therefore, we conclude that cmb is important for structural integrity of the entire cone during IC migration rather than for cone polarization.

Several models can be envisaged for Cmb function, the simplest being that Cmb could influence actin recruitment to and/or stability within the ICs by interacting directly with actin. This is, however, unlikely, as we do not detect actin-binding activity in either of two in vitro assays, although it is possible that Cmb requires a co-factor to interact with actin. Interestingly, two-hybrid assays and our CoIPs have shown that Cmb can interact with the Drosophila homolog of Rsp3, a crucial component of the radial spokes extending from each outer MT doublet towards the central MT pair (Hu et al., 2017). Radial spokes are important for regulation of the flagellar waveform beating movement, and Chlamydomonas Rsp3 mutants (pf14) were identified by their paralyzed flagella containing a regular 9+2 complement of MTs but lacking radial spokes (Luck et al., 1977; Porter and Sale, 2000; Witman et al., 1978; Yang et al., 2006). In Drosophila, similar to cmb mutants, germline Rsp3 knockdown results in highly penetrant IC defects, a phenotype that recently has been confirmed in rsp3 mutants (Wang et al., 2019). We were unable to examine flagellar motility in cmb mutants and rsp3 knockdown animals owing to the complete absence of mature sperm (Fig. 7). Despite a reported interaction with Pka-R1 and individualization defects upon its knockdown, we were unable to confirm the physical interaction by CoIPs (not shown). We cannot discern at present whether this is biologically meaningful or due to technical reasons.

The similarities between the loss of cmb and the rsp3 knockdowns strongly suggest that Cmb helps to coordinate IC movement and stability along the axoneme tracks. To the best of our knowledge, no such link has yet been identified. Although expected, the requirement of the axonemes for IC movement is controversial. No individualization defects are found in Fragile X-related mutants (Fmr1) even though their axonemes progressively lose the central MT pair during sperm maturation (Zhang et al., 2004). Likewise, kl-3 and kl-5 (Y chromosome dynein heavy chain genes) mutants lack axonemal outer dynein arms but show normal individualization (Timakov and Zhang, 2000; Zhang and Stankiewicz, 1998), although in those mutants sperm degenerate later. In contrast, individualization defects are found in sperm with reduced axonemal polyglycylation upon knockdown of the glycylase TTLL3b. Loss of polyglycylation on axonemal MTs is accompanied by severe, but variable, progressive loss of MT structures (Rogowski et al., 2009). Although we cannot exclude radial spoke-independent functions of Rsp3 (Yan et al., 2015), it is thus plausible that ICs need the axonemal MTs as tracks to move along the flagella and may become unstable when they are aberrant. At this time, we do not know whether Cmb plays an active role in the actin cones or rather a permissive role on the axonemes for IC progression. Intriguingly, we detect overexpressed FLAG-tagged Cmb-PA decorating the axonemes in elongation-stage cysts (Fig. S2D). However, despite the ability of the transgene to rescue the individualization defects of cmb mutants (Fig. 3), we were not able to evaluate Cmb localization in individualization-stage cysts, likely because of the reduced expression from the bam-GAL4 driver in later stages.

Overall, we show that the PPE gene cmb is required in the germline for proper IC stability and movement along axonemes during sperm individualization. This function is independent of the PCP pathway, but likely requires phosphorylation by Rok. Although the actin-rich IC structure is apparently unique to Drosophila sperm, an analogous actin- and microtubule-dependent process of cytoplasmic extrusion likely occurs during spermiation in mammals (Li et al., 2018, 2017; Qian et al., 2014). Furthermore, the sperm axoneme has long been an accessible primary cilium model, and the interplay between cilia and actin (and membrane) dynamics is just beginning to be appreciated (Adler and Wallingford, 2017; Inaba and Mizuno, 2016; Mirvis et al., 2018).

cmb^KO, UAS-cmb-RA and UAS-cmb-RB have been previously described (Fagan et al., 2014). pTFW (UAS)-cmb-RA6A was made by cloning the mutated fragment of pGEX-4T3-ES (Fagan et al., 2014) into pENTR_3C_cmb_RA and transferring the insert into pTFW via Gateway cloning (Invitrogen). w¹¹¹⁸, tj-GAL4, UAS-Rho1.N19^1.3 (BL#7327) and Rok (TRiP GL00209) and CG32392/rsp3 (TRiP HMJ23928) RNAi lines were from Bloomington Stock Center. RNAi lines targeting Rok (KK107802), rsp3/CG32392 (KK108322; VDRC#100592), Pka-R1 RNAi (KK101172; VDRC#103720 and GD11157; VDRC#26329) were from the Vienna Drosophila Resource Center. dsh¹, a PCP-specific allele of dsh, dgo³⁸⁰, Vang⁶, pk¹, sple¹, fz^P21, fz^R52 and mwh¹ (a kind gift of Dr P. Adler, University of Virginia) were described previously (Fagan et al., 2014; Jenny et al., 2005). UASp-Venus:Rok^K116A and bam-GAL4-VP16 were gifts of Drs J. Zallen (Memorial Sloan Kettering Cancer Center, New York, NY, USA) and Y. Yamashita (Life Science Institute, University of Michigan, MI, USA), respectively.

For sterility tests, single 3- to 5-day-old males were mated with two virgin females of w¹¹¹⁸ for 3 days on standard fly food supplemented with live yeast grains before parents were discarded. Crosses were scored by whether or not adult offspring eclosed and by counting the number of progeny.

Testes were dissected from up to 1-week-old males in PBS and then transferred to fixation buffer overnight (2.5% gluteraldehyde in 0.1 M sodium cacodylate buffer). Samples were then post-fixed with 1% osmium tetroxide followed by 2% uranyl acetate, dehydrated through a graded series of ethanol and embedded in LX112 resin (LADD Research Industries). Ultrathin sections were cut on a Leica Ultracut UC7, stained with uranyl acetate followed by lead citrate and viewed on a JEOL 1200EX transmission electron microscope at 80 kV.

CoIPs were carried out as previously published, transfecting HEK293 cells with 3 µg of each plasmid per 6 cm plate (Fagan et al., 2014). pCDNA3-Act:4-V5 expressing C. elegans act-4 was a kind gift of Dr H. Bülow, Einstein. GST fusion proteins were expressed as described (Jenny et al., 2003, 2005), except that Escherichia coli were lysed in a lysis buffer consisting of 20 mM Tris pH 7.6, 50 mM KCl, 2 mM MgCl₂, 1 mM EDTA, 1 mM DTT, and protease inhibitors. Wash buffer additionally contained 0.1% Triton X-100. Proteins were dialyzed against 20 mM Tris pH 7.6, 50 mM KCl, 2 mM MgCl₂, 1 mM DTT, 0.1% Triton X-100, 10% glycerol. Actin-pelleting assays were performed using the Actin Binding Protein Kit as indicated by the manufacturer (Cytoskeleton) using 10 µl of the indicated protein (16-30 µg). Note that GST-SX, although soluble after centrifugation at 10,000 g is not soluble after the 1 h spin at 150,000 g used to preclear the proteins prior to the actin-binding assay.

Rsp3 was amplified from GH13213 using primers CG32392_for_Sal (GTCGACGCCACCATGCCAGAGACGGAGCAGCAG) and CG32392_rev_Sal (GTCGACTTAGAGATGAGGAGCCGGTGG) using Cloneamp HiFi polymerase (Takara) and cloned into pCR8-GW (Invitrogen). The insert was then transferred into pCS3-Myc6 (Villefranc et al., 2007) using a Clonase LR reaction. Transfections and CoIPs were carried out as published (Fagan et al., 2014) using HEK293 cells (ATCC CRL-1573).

For qPCR, testes without accessory glands (but mostly including seminal vesicles) of 20 males were dissected in PBS, washed three times with PBS and then lysed in 0.8 ml Trizol (Invitrogen), according to instructions by the manufacturer. For comparison, RNA of a mix of eye, wing and leg discs from third instar larvae was prepared. Reverse transcription was performed with a Superscript IV first-strand synthesis system (Thermo Fisher Scientific) according to the instructions of the manufacturer. Real-time PCR reactions were performed using the Master cycler Eppendorf realplex² with Power SYBR Green PCR Master Mix (Thermo Fisher Scientific). An equal amount of cDNA was mixed with 5 pmol primers and power green SYBR master mix to make 25 µl of reaction volume. Following 15 min at 95°C for enzyme activation, cycling conditions were: 94°C for 15 s, 30 s at 60°C, 45 s at 72°C for 45 cycles. Melting curve program was 95°C for 15 s, 60°C for 15 s and 95°C for 15 s. Relative expression levels are plotted using the ΔCT method normalizing to the average expression of RpL11 and Gapdh2 (Schmittgen and Livak, 2008). Primers used for qPCR are listed in Table S1.

For whole-mount staining, testes from 0- to 5-day-old males reared at 23°C were dissected in PBS, fixed in 5% formaldehyde in PBX (PBS+0.1% Triton X-100) for 20 min, and washed in PBX. For immunofluorescence, a 1 h block was performed in 5% normal donkey serum (Jackson ImmunoResearch) +1% Triton X-100 before overnight incubation with primary antibodies. Testes were stained with TRITC-phalloidin (Sigma-Aldrich, P1951) together with secondary antibodies and mounted in DAPI-Fluoromount G (Southern Biotech). IC quantification was performed as described by Ben-David et al. (2015).

For immunostaining of actin-cone proteins, ring canals, and acetylated α-tubulin, testes were squashed on slides and frozen in liquid N₂ before fixation and staining, as described by Sitaram et al. (2014).

Primary antibodies were mouse anti-α-Spectrin [Developmental Studies Hybridoma Bank (DSHB), 3A9, 1:25], rabbit anti-cleaved caspase-3 (Cell Signaling Technology, 9664T, 1:100), mouse anti-Myosin VI (gift of K. Miller, Washington University, St. Louis, MO, USA; 3C7, 1:20), mouse anti-Quail (DSHB, 6B9, 1:50), mouse anti-Sn (DSHB, sn7C, undiluted), mouse anti-phospho-Tyrosine (Abcam, ab10321, 1:200), mouse anti-acetylated α-tubulin (Sigma-Aldrich, 6-11B-1, 1:100), and mouse anti-FLAG (Sigma-Aldrich, M2, 1:500). Donkey secondary antibodies (Jackson ImmunoResearch 711-545-152 or Molecular Probes A21202) were used at 1:200.

Testes were imaged using an Olympus IX-81 motorized inverted microscope with XM-10 monochrome camera (lenses: 10×/0.3 NA, 20×/0.75 NA, 40×oil/1.3 NA, 60× oil/1.35 NA), a Zeiss LSM510 confocal microscope (lenses: 10×/0.3 NA, 20×/0.8 NA, 40×/0.75 NA) or a Zeiss LSM800 confocal microscope (lenses: 10×/0.3 NA, 20×/0.8 NA, 40× oil/1.3 NA). Intensity of ICs was quantified from epifluorescence images of phalloidin-stained testes using Fiji/ImageJ as follows: regions of interest of five ICs still associated with nuclei from five testes per genotype were selected and average fluorescence was tabulated. Quantifying integrated density gave qualitatively similar results.

For phase-contrast imaging of post-meiotic spermatids, testes were squashed onto slides in PBS and imaged live with phase contrast within 30 min using an Olympus IX-71 inverted microscope with DS-U2 Nikon camera and NIS Elements software (20× lens/0.40 NA).

We are grateful to our colleagues for sharing reagents, including P. Adler, J. Zallen, Y. Yamashita, and K. Miller, to Leslie Gunther of the Einstein AIF for help with EM (National Institutes of Health grant P30CA013330), and to Dr Mimi Kim (Einstein Department of Epidemiology & Population Health) for advice on statistics. We thank the Bloomington Drosophila Stock Center (National Institutes of Health P40OD018537), Vienna Drosophila Resource Center, and Developmental Studies Hybridoma Bank (created by the NICHD of the NIH and maintained at The University of Iowa, Department of Biology, Iowa City, IA) for fly stocks and reagents. We are grateful to Steinhauer lab students Tzvi Fishkin and Aryeh Korman, who helped generate preliminary data for the project. We thank Drs J. Secombe, J. Treisman, H. Bülow and J. Rollins for their comments on this manuscript.