yuri gagarin is required for actin, tubulin and basal body functions in Drosophila spermatogenesis

A unique feature of the genus Drosophila is the formation of unusually long sperm tails. Sperm lengths of millimeters are common within this group, with the 1.8 mm sperm of D. melanogaster being fairly typical. This marked expansion in sperm length reflects an unusual aspect of spermatogenesis in these organisms: in contrast to other species in which an intraflagellar transport system is used for growth of the sperm flagellum (Scholey, 2006), Drosophila sperm axonemes are assembled in syncytial cysts by a mechanism that does not require, and is not limited by, this system (Han et al., 2003; Sarpal et al., 2003). This unusual sperm axoneme development and the resulting expansion of sperm tail length have led to distinctive features of spermatogenesis not found in other species. In D. bifurca, a special `sperm roller' has evolved to package its 6-centimeter-long gametes (Joly et al., 2003). In D. melanogaster, a highly evolved individualization process that generates 64 individual sperm from an elongate cyst containing 64 syncytial spermatids has been identified and studied (Noguchi and Miller, 2003; Tokuyasu et al., 1972a). The distinctive molecular mechanisms needed for this process include a motile filamentous actin system (the investment, or actin, cones) that traverses the entire length of the sperm tails, removing excess cytoplasm and investing each sperm in its own plasma membrane. A specialized microtubule-rich structure (the dense complex) is also associated with the sperm nuclei and functions to position the basal body and also possibly to strengthen the nuclei as they undergo extreme condensation (A. D. Tates, Cytodifferentiation during spermatogenesis in Drosophila melanogaster, PhD thesis, Rijksuniversiteit Leiden, The Netherlands, 1971) (Tokuyasu, 1974).

We have identified a locus, yuri gagarin (yuri), that we show here has multiple roles in the generation of elongate individualized sperm. The gene is only highly conserved in the genus Drosophila, suggesting specialized roles in these organisms. Interestingly, yuri was initially identified through its function in another specialized organ system of insects and arthropods: the chordotonal organs. These are complex mechanosensory structures with roles in proprioception and graviperception. The first mutation at the locus, yuri^c263, was identified in a screen for mutants affecting gravitaxis. Altered gravitaxis was shown to result from perturbed expression of yuri in subsets of chordotonal neurons (Armstrong et al., 2006). The molecular functions of the locus identified here suggest that yuri mediates specialized actin- and microtubule-related activities in Drosophila tissues.

In addition to the cDNA (GH14032) encoding a ∼30 kDa protein that we used previously (Armstrong et al., 2006), we identified 11 further yuri ESTs/cDNAs from adult testis, ovary, S2 cells and embryos through FlyBase. Sequencing of these new cDNAs established that three major transcript classes are generated from yuri (Fig. 1). Two promoters are used, with the medium transcripts initiated at the proximal promoter and the short and long classes from the distal promoter. However, all isoforms begin at one of two closely positioned ATGs. The short transcript class encodes the ∼30 kDa protein identified previously. The medium class encodes isoforms of 64-65 kDa that extend ∼400 amino acids further at the C-terminus. The long class, encoding proteins of 101-107 kDa, extends an additional ∼300 amino acids C-terminally. The short yuri isoform is novel, with only a single recognizable motif (a polyproline stretch). However, the two longer forms contain coiled-coil motifs with weak similarity (∼20% identity) to those in many fibrillar proteins that dimerize, such as myosin heavy chain and CLIP-190. The strongest match is to the coiled-coil of Sticky, the Drosophila citron kinase (Sweeney et al., 2008).

yuri is unique in the D. melanogaster genome, once the weak similarities to coiled-coil regions are disregarded. Thus, to avoid spurious similarities, the shortest yuri isoform was used to find yuri orthologs in other organisms. Significant matches were found in all 11 sequenced Drosophila genomes (Drosophila 12 Genomes Consortium, 2007), but none was identified in other evolutionary orders or other insects, including the closest Dipteran relatives, the Culicidae (mosquitoes) (Fig. 2). Sequence conservation within the Drosophila genus was high (91-37% sequence identity, 93-57% similarity) across the entire ∼100 kDa isoform of D. melanogaster. yuri therefore appears to be a Drosophila-specific gene. Most species have one yuri gene, but two related genes are present in D. pseudoobscura and D. persimilis.

To investigate Yuri expression, we generated antibodies against the sequences common to all isoforms (see Materials and Methods). Immunoblots of yuri⁺ embryos and embryos lacking yuri established the specificity of our antisera and their ability to detect the three predicted Yuri isoform classes (Fig. 3A). These blots also demonstrated that only the short Yuri isoform is maternally loaded into the embryo, with the longer isoforms appearing later in embryogenesis (Fig. 3A,B). In later stages, all three isoform classes are ubiquitously expressed (Fig. 3C). The ∼65 kDa class is most abundant in most situations, although in testis and thorax the other isoforms are also highly expressed (Fig. 3C). The existence of at least two isoforms for both the ∼100 kDa and ∼65 kDa classes was confirmed by these experiments. Additional bands were sometimes present that probably represent specific degradation products, as they were largely missing in embryos lacking yuri (Fig. 3A) and in the yuri^F64 mutant (Fig. 3C) (see below).

The yuri^c263 mutation from our gravitaxic screen is an insertion of P{GawB} just upstream of the transcription start site for the medium length transcripts. We generated further mutations by imprecise excision of P{GawB} and of a second transposon, KG03019 (Roseman et al., 1995), inserted three residues downstream of the yuri^c263 P element. Three excisions that delete the relevant transposon and adjacent genomic DNA were identified. One of these is lethal (yuri^LE1), but the deletion extends upstream into an adjacent gene (cullin3; guftagu) known to affect viability (Mistry et al., 2004). In yuri^L5, a short region of yuri upstream sequence is deleted, causing reduced expression of all Yuri isoforms. Nevertheless, homozygous yuri^L5 animals are viable with no obvious phenotype. Only one deletion, yuri^F64, removes transcribed sequences from the locus. Most of the 5′ UTR of the ∼65 kDa isoforms is deleted, with only ten residues upstream of the first initiator ATG remaining (Fig. 1A). The yuri^F64 deletion lead to complete loss of ∼65 kDa isoforms in all tissues and stages examined (Fig. 3C). The ∼100 kDa isoforms remained strongly expressed, but expression of the ∼30 kDa isoform was decreased in several tissues and undetectable in the testis (Fig. 3C).

Homozygous yuri^F64 mutants (yuri^F64) are viable with normal external morphology. However, yuri^F64 males are completely sterile, whereas females are fertile (data not shown). Flies heterozygous for yuri^F64 and deficiency Df(2L)do1, which deletes yuri, were also male sterile and female fertile. The testis phenotype (see below) was identical in yuri^F64 homozygotes and hemizygotes (data not shown), demonstrating that it results from the effects of the yuri^F64 mutation on the yuri locus. In order to determine whether yuri^F64 affects overall viability, the survival of yuri^F64 homozygous progeny versus heterozygous progeny (yuri^F64/CyO Roi) was quantitated for a cross of yuri^F64 females with heterozygous (yuri^F64/CyO Roi) males. Of 649 progeny, 51% were yuri^F64 homozygotes, indicating that yuri^F64 has no effects on survival to adulthood.

The Drosophila testis contains a stem cell system at its apical tip from which spermatogonial cells are budded off to proceed through spermatogenesis. A somatic stem cell system is also present that produces so-called cyst cells. A pair of cyst cells encases the division products of each spermatogonial cell throughout spermatogenesis and post-meiotic spermiogenesis. Each spermatogonial cell generates a cyst of 64 spermatids, linked by cytoplasmic bridges, which undergoes dramatic elongation. At completion, each cyst has a highly elongate cytoplasm (∼1.8 mm in length) with the 64 condensed nuclei positioned at the seminal vesicle end and 64 axonemes extending from the nuclei along the length of the cyst towards the apical tip. Two giant mitochondrial derivatives, generated by fusion of the mitochondria within each post-meiotic spermatid, extend along the length of each axoneme. The later stages of spermiogenesis involve a specialized process termed individualization (see below) in which the 64 syncytial spermatids are converted into 64 individual sperm. Finally, a coiling process retracts the sperm down to the entrance to the seminal vesicle.

Highly elongate spermatid cysts were present in yuri^F64 testes, some of which were attempting to coil, but it was unclear whether mature sperm were formed. To address this question, we introduced a don juan-GFP fusion construct into the yuri^F64 background. Don Juan protein is produced in the giant sperm tail mitochondria and persists into mature sperm. Don Juan-GFP (Santel et al., 1997) provides a marker for late spermiogenesis (Civetta, 1999; Gao et al., 2003). We examined 8-day-old virgin males, which should have large quantities of sperm in the seminal vesicles. In yuri^F64/CyORoi heterozygotes carrying don juan-GFP, the seminal vesicles were full of fluorescent sperm and the basal testis carried masses of fluorescent coiling sperm (Fig. 4A). In yuri^F64, no fluorescence was detectable in the seminal vesicles and the basal testis contained curled structures, thicker than individual sperm with aberrant coiling (Fig. 4B). Squashes of seminal vesicles confirmed the presence of motile sperm in the controls and their complete absence in the mutant (data not shown).

Phase-contrast examination of testis squashes revealed no defects in spermatogenesis up to the post-meiotic stages; `onion stage' spermatids appeared normal. The structures undergoing abortive coiling in yuri<sup>F64</sup> testes were full-width spermatid cysts, indicating a failure of individualization. Individualization begins after formation of a cone of F-actin around the attachment site of each axoneme to the sperm nucleus, with the flat edge of the cone facing up the length of the sperm tail. All 64 cones within a cyst then travel in unison up the testis. In their wake they leave individual axonemes, each encased in a plasma membrane, and, ahead of the set, excess cytoplasm and organelles are pushed up the testis to be discarded as a `waste bag'. The actin cones are the only significant F-actin structures in the testis and are easily visualized with rhodamine-phalloidin (Fabrizio et al., 1998). Whereas in control (yuri^F64/CyO Roi or w¹¹¹⁸) testes, multiple sets of actin cones and waste bags were detected, the yuri^F64 testes contained neither (Fig. 4). Instead, elongated `sleeves' of actin were seen around the periphery of some spermatid cysts. These appeared as solid tubes in normal fluorescence imaging (Fig. 4B″), but as hollow structures in confocal sections (Fig. 5A). We established that these sleeves are actually present in the somatic cyst cells surrounding the cysts, rather than in the cysts themselves, by use of GFP `exon trap' insertions (Kelso et al., 2004) that express GFP in the cyst cells (Materials and Methods). In the yuri^F64 background, the actin sleeve staining and GFP in the cyst cells precisely overlapped (Fig. 5A). Having identified these sleeves in yuri^F64, we discovered similar structures present at a lower frequency in control testes (Fig. 5B). In controls, these sleeves are always in the basal regions of the testis where sperm coiling takes place, whereas in yuri^F64 they form throughout the testis. We address the significance of these structures in the Discussion. The major conclusion here is that in yuri^F64 no actin cone sets or F-actin structures of any kind are present in the germline cysts proper.

The formation of the F-actin cones of individualization has been studied previously (Fabrizio et al., 1998; Lindsley and Tokuyasu, 1980; Noguchi et al., 2006). Initially, actin fibers accrete along the lengths of the condensed sperm nuclei in the basal testis. The actin then moves to form cones, flaring off the apical ends of the nuclei before release to move up the axonemes. Nuclei in all stages of this process are present in the basal region of wild-type testes. In yuri^F64, although the nuclear sets were seen to descend to this level and undergo some condensation, they were clearly more disorganized, with individual nuclei trailing behind, apparently detached from the main cluster. In some late-stage clusters, almost all the nuclei were distorted in shape, some in a helical or circular configuration (Fig. 5D). No nuclei ever condensed to the tight bundles seen in controls (Fig. 5C). Furthermore, no well-formed sets of cones were ever detected, although a little F-actin accumulated around some nuclei (Fig. 5E,F). Interestingly, small under-developed cones were occasionally found singly or in clusters in this region. Some of them were apparently mobile, as they appeared at some distance from any nuclei (Fig. 5G).

Our antisera, which detect all Yuri isoforms, were used to examine Yuri localization in control testes. Yuri was present at all stages of germ cell development, peaking around meiosis, with most staining being cytoplasmic and diffuse (Fig. 6A). However, in addition, a striking and dynamic pattern of Yuri association with the post-meiotic spermatid nuclei was seen as they condensed during elongation (Fig. 6B-F). While the nuclei were still round, Yuri was seen to accumulate as a cap over one hemisphere of each nucleus. As the nuclei became ellipsoid, the Yuri staining transformed into a stripe along the nuclear long axis and a dot at the apical nuclear tip. In the final stages of nuclear maturation, first the stripe disappeared and then the dot was also lost.

Tokuyasu has described the ultrastructural changes to the nuclei during elongation (Tokuyasu, 1974). Part of the nuclear membrane is fenestrated with nuclear pores during this process. Initially, this region forms a cap over one hemisphere of the round post-meiotic nucleus, with dense material aggregating over this region between the nuclear membrane and adjacent endoplasmic reticulum. As the nuclei elongate, this cap and associated material transform to a stripe along the long axis of the nucleus. More of the dense material accumulates along with microtubules, with the whole complex sinking inwards to form a groove filled with dense cytoplasm and a microtubule bundle (collectively the `dense complex') that runs the length of the nucleus. The nuclei are actually horseshoe-shaped in cross-section at this stage. In the final stages of nuclear maturation, the dense complex is dispersed and the nuclei regain a circular cross-section. The dense complex is thought to provide structural rigidity to the nuclei during the elongation process (Tokuyasu, 1974). Early after meiosis, the single centriole of each spermatid embeds into the spherical nuclear membrane at the center of the dense complex and then converts into the basal body. During elongation, the basal body moves to the apical tip of the nucleus, immediately adjacent to the stripe of dense complex (Fig. 6B).

The pattern of Yuri localization on the spermatid nuclei was strikingly similar to that of the dense complex and associated basal body. To position Yuri relative to these structures, we co-stained for γ-tubulin, Centrosomin (Cnn) and β-tubulin. γ-tubulin is a component of the centriolar adjunct (CA) (Wilson et al., 1997), a torus-shaped structure around the middle of the basal body during elongation (Fig. 7C) (Tokuyasu, 1975). Centrosomin, a centriole component, is present early in the transformation to the basal body but is subsequently lost (Li et al., 1998). β-tubulin is a general marker for microtubules. γ-tubulin/Yuri co-staining established that the basal body is at the center of the Yuri cap in round spermatids (Fig. 7A), providing evidence that the Yuri cap corresponds to the accumulating dense complex. No round spermatid nuclei that co-stained for the Yuri cap and Centrosomin were detected, suggesting that Centrosomin is lost before significant Yuri accumulation. We were not able to detect a stripe of microtubules along the nuclei by staining for β-tubulin. Very high general cytoplasmic staining and/or possibly the burying of the appropriate epitope could underlie this failure.

Although γ-tubulin staining showed an apical dot on the elongating nuclei, interestingly, the Yuri dot and the γ-tubulin dot did not coincide. The Yuri dot, which at high magnification has a bell shape (Fig. 6F), was sandwiched between the dot of γ-tubulin staining and the nuclear membrane. Thus, Yuri is probably not part of the basal body per se but lies between the basal body and the nuclear membrane. EM analysis has established that the basal body is embedded into a ∼0.5 μm indentation in the nuclear membrane (Tokuyasu, 1975) at this stage. It seems likely that the Yuri dot is the residuum of the initial dense-complex cap that was always beneath the insertion point of the basal body, and that Yuri continues to fill the space between the membrane and the basal body during nuclear elongation.

Given the complete failure of actin cone formation in yuri^F64, we also examined the relationship between Yuri and F-actin localization during spermiogenesis. We determined that the cap of dense complex at the round spermatid stage contains not only Yuri, but also F-actin (Fig. 7D). Furthermore, the actin staining extended around the basal body. F-actin continued to colocalize with Yuri in the stripe and dot pattern as the nuclei elongated (Fig. 7E). We also established that Yuri is a component of the F-actin cones used in individualization (Fig. 7F). Yuri immunostaining was seen throughout the large cones moving up the testes, and in cross-sections Yuri appeared concentrated in the inner cone regions, whereas actin was more peripheral.

The yuri^F64 mutation does not eliminate all isoforms of Yuri. Nevertheless, we determined that in yuri^F64 the association of Yuri with the dense complex is completely lost, and all elements of the nuclear staining pattern – the cap, stripe and dot – are missing (Fig. 8A). Thus, the isoforms that are absent in yuri^F64 are essential for protein function at these sites. The absence of Yuri from the dense complex allowed us to determine whether Yuri is necessary for the association of other components with this structure. In yuri^F64, all elements of F-actin nuclear staining from the round spermatid stage onwards were lost (Fig. 8C). Yuri is therefore required for the initial accumulation and subsequent maintenance of F-actin within the dense complex. Similarly, γ-tubulin staining was never observed on the early round nuclei or at the later elongate stages (Fig. 8B), demonstrating that Yuri is required for attachment to, or possibly formation of, the CA of the basal body.

This absence of the CA raised the issue of whether basal bodies are present at all on the spermatid nuclei in yuri^F64. To address this question, a GFP-fusion construct for the PACT domain of the Drosophila Pericentrin-like protein (dPLP; Cp309) (Martinez-Campos et al., 2004) was introduced into the yuri^F64 background. The PACT domain of both mammalian pericentrin and dPLP provides targeting to the centrosomes/centrioles. In the Drosophila testis, GFP-PACT is an excellent fluorescent marker for the basal body (Martinez-Campos et al., 2004). In control cysts (w¹¹¹⁸ or w^–; yuri^F64/CyO Roi), small cylinders of GFP-PACT staining demonstrate the presence of the basal bodies tightly clustered at the apical tips of condensing nuclei (Fig. 9A). GFP-PACT-marked basal bodies were also present on condensing nuclei in yuri^F64. However, they were not tightly localized at the apical tips but scattered along the nuclei. Indeed, in many clusters, a fraction of the basal bodies were actually at the rostral rather than apical nuclear tips (Fig. 9B). Quantitation of the GFP-PACT fluorescence associated with control or yuri^F64 nuclear clusters (using Metamorph software) indicated that yuri^F64 does not affect the level of GFP-PACT binding to the basal bodies. In the final stages of nuclear condensation, the GFP-PACT fluorescence was lost from control nuclei. Similarly, although the nuclei never fully condense in yuri^F64, GFP-PACT was ultimately lost from these nuclei too.

Previous work has implicated cytoplasmic dynein and the related protein Dynactin in the formation of the dense complex (Li et al., 2004). Like Yuri, dynein heavy chain accumulates in the hemispherical cap on round spermatid nuclei but, in contrast to Yuri and the components examined here, its nuclear positioning is transient and it is not detectable in the dense-complex stripe during nuclear elongation. This brief association has a role in basal body functioning, however, because in a null mutant for the 14 kDa dynein light chain (Dlc90F), dynein heavy chain does not accumulate on the nuclei and, later, some nuclei lack a CA as judged by γ-tubulin staining.

In Dlc90F⁰⁵⁰⁹⁰, an RNA-null in the testis (Caggese et al., 2001), the nuclear localization pattern of Yuri was found to be dramatically altered. The initial hemispherical cap of Yuri and the later stripe were highly attenuated and in some cases barely detectable (Fig. 8D,E). However, the bell-shaped dot of Yuri was now present at the base of the basal body, even in round spermatids (Fig. 8D). Furthermore, in both round and elongating nuclei, a second dot of Yuri was present (Fig. 8D,E). Co-staining with γ-tubulin demonstrated that this dot is the region of the basal body distal to the CA (Fig. 8F).

As the centrioles mature into basal bodies, a transition in protein composition occurs: Centrosomin is lost (Li et al., 1998) and the protein Uncoordinated (Unc) now becomes associated with these structures (Baker et al., 2004). Mutations in cnn or unc affect basal body function and produce abnormalities in axoneme structure. Given the loss of the CA and the aberrant positioning of the core basal bodies in yuri^F64, we examined axoneme structure by TEM. This analysis also confirmed the complete failure of individualization in yuri^F64 (Fig. 10C). In contrast to controls (Fig. 10A), the 64 axonemal `triads' – the axonemes and their major and minor mitochondrial derivatives (MDs) – all shared a single cytoplasm. Furthermore, terminal differentiation of the minor MDs was imperfect. In controls, this derivative undergoes dramatic expansion/disruption during individualization (Tokuyasu et al., 1972a) and collapses to a tiny structure in mature sperm (Fig. 10A). In the most developed cysts in yuri^F64, the minor MD was less condensed than normal (Fig. 10C).

We examined axoneme structure in younger elongating cysts. Gross axonemal structure (the typical `9+2' arrangement) was normal in yuri^F64. Of more than 750 studied, only two damaged axonemes were found, showing breaks in the outer circle of nine doublets (Fig. 10D). However, rarely, aberrant arrangements of axoneme-MD triads were found. These included: single axonemes with two major or two minor MDs, as judged by the presence/absence of a paracrystalline body, a marker for the major MD (Fig. 10D); sharing of a major or minor MD between two axonemes (Fig. 10E); major MDs with two or more paracrystalline bodies (Fig. 10E); and major MDs undergoing the expansion typically associated with the minor MD during individualization (Fig. 10D,E).

Although the spermatids in elongating cysts are syncytial, the links between them are narrow cytoplasmic bridges and the overall shape of each individual `cell' is distinguishable in EM cross-sections. Each `cell' typically contains a single axoneme-MD triad, although `fused' cells with two-eight triads have been detected in wild-type cysts (Stanley et al., 1972). The triad abnormalities in yuri<sup>F64</sup> were largely within `cells' that contained multiple triads (Fig. 10E). However, our analysis of testis squashes provided no evidence that these arose as a result of cytokinesis defects in meiosis (see above).

We also examined sperm tails of yuri^F64 heterozygotes in two genetic backgrounds (w^–; yuri^F64/CyO Roi and w^–; yuri^F64/+). Surprisingly, in both backgrounds, almost all mature cysts (∼90%) had a few imperfectly individualized triads (Fig. 10B), with a few cysts in which <50% of the sperm tails were still in syncytial cytoplasm. Thus, although fertile, yuri^F64 heterozygotes clearly have individualization defects. Defects in triad development similar to those seen in yuri^F64 homozygotes were also detected and, unexpectedly, were somewhat more prevalent in the heterozygotes, with ∼30% of the cysts in one testis showing these defects. More-pronounced axonemal abnormalities were also detected: in addition to broken sets of outer doublets, axonemes with no central doublet were present (Fig. 10G,I).

In both the yuri^F64 homozygotes and heterozygotes, occasional examples were found of adjacent axonemes with their outer arms pointing in opposite orientations (Fig. 10D). Such cysts always had the correct number of axoneme profiles (64), and one cyst with 32 axonemes in one orientation and 32 in the other was found (data not shown). We could therefore exclude the possibility that these two orientations represented axonemes that were folded back on themselves. Two alternative explanations remained: either a fraction of the basal bodies have an altered chirality, or some of the 64 sperm nuclei migrate to the wrong end of the elongating cyst so that their associated axonemes extend along the cyst in the wrong direction. The latter explanation proved to be the case. Upon inspection, clusters of condensed sperm nuclei (ranging from one or two to ⩾20 nuclei) with attached basal bodies were detected at the apical end of elongated cysts, close to the stem cell tip (Fig. 9C,D). For the yuri^F64 homozygote, four out of 40 testes examined showed this defect; for the heterozygotes, two out of nine testes had these mispositioned nuclei.

Our initial yuri mutant (yuri^c263) was identified by its altered gravitaxic responses. Further studies indicated that these changed responses arise from altered chordotonal neuron function, but provided no information as to whether yuri is uniquely expressed in these neurons (Armstrong et al., 2006). Studies here reveal that yuri is expressed ubiquitously, indicating that yuri is not dedicated to gravitaxic responses but rather that the yuri^c263 mutation disrupts yuri expression in a manner that specifically affects this function.

All isoforms of Yuri are expressed ubiquitously and yuri^F64 removes the major ∼65 kDa isoform(s) from all tissues studied. Surprisingly, the only obvious developmental defect is male sterility. In the yuri^F64 testis, the 30 kDa isoform is also missing, whereas in other tissues this isoform is less affected. Thus, the yuri^F64 male sterility reflects either unique roles for ∼65 kDa isoforms, or the unique loss of the 30 kDa isoform. That loss of the ∼65 kDa isoforms has no effects in other tissues might indicate redundancy with the ∼100 kDa isoforms. The importance of the 30 kDa isoform is demonstrated by a consideration of the ovary (Fig. 3C). The ∼100 kDa Yuri isoforms are not normally present in the ovary, so that in yuri^F64 the 30 kDa protein is the only isoform detectable in the tissue. Nevertheless, oogenesis and early embryogenesis proceed normally.

Although the major defects seen in yuri^F64 homozygotes are largely absent from heterozygotes, some minor, incompletely penetrant defects (particularly in axoneme structure) are more prevalent in the heterozygous than the homozygous condition. Because yuri^F64 causes loss of particular isoforms, the normal stoichiometric balance between isoforms is disrupted in both homozygotes and heterozygotes, but it is disrupted differently in the two situations. Thus, given that two classes of Yuri isoforms contain coiled-coil regions, altered dimerization or protein interactions that have more severe consequences for axoneme assembly might be produced uniquely in the heterozygote.

The various elements of the yuri^F64 testis phenotype provide clues as to the molecular functions of the protein. One clear implication is that Yuri regulates F-actin function. We show here for the first time that F-actin is associated with the dense complex on spermatid nuclei and that in yuri^F64, F-actin never accumulates on the nuclei, suggesting an initiating role for Yuri in dense-complex formation. Yuri is also a component of the actin cones that mediate sperm individualization and is required for their formation. The actin cones are formed by a two-step process (Noguchi et al., 2006). Initially, parallel actin fibers are formed around the nuclei and then an actin meshwork is added at each apical nuclear tip. Given the absence of actin cone initiation in yuri^F64, it seems likely that Yuri has an early role in F-actin deposition here too.

The aberrant F-actin sleeves formed in the somatic cyst cells in yuri^F64 led us to identify related actin sleeves around actively coiling sperm in control testes. Sperm coiling is executed within the confines of the head cyst cell, which completely engulfs the apical region of the cyst (Tokuyasu et al., 1972b). Elaborate microvilli, full of 50 Å filaments, project from the head cyst cell onto the cyst walls and Tokuyasu and colleagues suggest that coiling largely represents the collapse of the intrinsically helical sperm tails into a flat pile of gyres as a result of contraction and shape change within the head cyst cell. We propose that the actin sleeves in control testes are related to the 50 Å filaments seen by Tokuyasu et al. and that in yuri^F64, F-actin structures form at inappropriate positions in association with abortive coiling.

In addition to regulating actin function, Yuri is implicated in microtubule/tubulin action. The stripe of dense complex along the elongating nuclei accretes a bundle of microtubules that are thought to provide structural rigidity to the nuclei. Although we were not able to image these microtubules, in yuri^F64 many late-stage nuclei lose their rigidity and collapse into helical twirls, suggesting that the microtubules are no longer present. The presence of Yuri in the dense complex is also intimately associated with proper positioning, formation and functioning of the basal body. When Yuri is not present at this site, (1) the basal bodies are scattered along the nuclei, or even mispositioned at the rostral nuclear tips, (2) the CA element of the basal body is missing and (3) the axonemes show defects similar to those of other mutations (cnn and unc) that affect basal body function. Nevertheless, our findings for the GFP-PACT marker indicate that dPLP is recruited normally to the basal bodies in yuri^F64. Interestingly, in mammalian systems, interaction between γ-tubulin and pericentrin is thought to underlie the targeting of γ-tubulin to centrosomes/centrioles (Young et al., 2000). dPLP is therefore implicated in promoting the presence of γ-tubulin and of the CA on the sperm basal body. Our evidence here that in yuri^F64, dPLP is on the basal bodies but γ-tubulin is not, suggests a role for Yuri in the interaction of these two proteins.

At the end of elongation, prior to individualization, the nucleus-basal body association is altered so that the axoneme and sperm head are locked in a permanent configuration relative to one another (Lindsley and Tokuyasu, 1980; Tokuyasu, 1975). This change involves disappearance of the CA and movement of the basal body to lie in a shallow groove on one side of the nucleus. Predictably, the CA components γ-tubulin and Unc are lost from the nuclei at this stage (Baker et al., 2004). We show here that both the Yuri dot and the GFP-PACT marker also disappear at this point. The basal body present on mature sperm is clearly stripped of many ancillary proteins.

Although Yuri appears to anchor tubulin structures, including the basal body, to the nuclear membrane, our findings for the dynein light chain mutant suggest that the initial positioning of Yuri on the nuclear membrane is determined by dynein transport, presumably along microtubules. In the dynein light chain mutant, Yuri localization is dramatically altered, with Yuri now primarily associated with the basal body – a novel association not seen in the wild type. The implication must be that an activity of dynein is required to prevent an interaction of Yuri with the basal body.

The opposing orientations of some adjacent axonemes in yuri^F64 reflects the unexpected positioning of sperm nuclei at the wrong ends of elongated cysts. Contacts that normally hold the nuclei together in tight alignment appear to be missing in yuri^F64, and this could permit loose nuclei to migrate to the wrong location. Axonemes with opposite orientations in a single cyst have been reported for mutations in the Drosophila parkin homolog (Riparbelli and Callaini, 2007). Although these investigators did not report a search for nuclei at the wrong ends of cysts, they did note occasional actin cones pointing in the wrong direction – a finding that suggests the same underlying cause for the two axoneme orientations in both their case and ours.

The finding that different mutations of yuri affect processes as disparate as gravitaxis and spermatogenesis is initially surprising. However, together with sperm, mechanoreceptor neurons, such as those affected by yuri^c263, are the only cell types in Drosophila that possess cilia, and genes that affect ciliary function have been shown to affect both mechanosensory organs and spermatogenesis. Mutations in touch insensitive larva B (tilB) are defective in hearing and touch perception as a result of defects in the chordotonal organs (Eberl et al., 2000). Mutations in unc affect both the chordotonal organs and the external sense organ (eso) class of mechanoreceptors (Eberl et al., 2000). Mutations at both loci are also male sterile because they encode proteins with roles in cilia. TilB is a conserved ciliary protein with a leucine-rich region and a coiled-coil domain (Kavlie et al., 2007) and Unc is associated with the basal bodies in sperm and mechanosensory neurons (Baker et al., 2004). Unc, like γ-tubulin, is a component of the CA and, like Yuri, is insect-specific and contains coiled-coil regions (Baker et al., 2004).

These examples suggest that the yuri function affected in yuri^c263 might be a role in positioning the ciliary basal bodies of the chordotonal neurons, a role comparable to that identified here in spermiogenesis. Furthermore, the intriguing possibility of molecular interactions between Yuri and Unc is suggested. The proteins are physically close at the basal body and their only distinguishing features are coiled-coil domains that presumably facilitate protein-protein interactions. It seems possible that these two proteins have evolved to fulfil specialized roles associated with anchoring the basal bodies that could entail heterodimerization.

The entire coding region of the 30 kDa Yuri isoform from clone GH14032 was amplified by PCR, cloned in Topo vector pCR2.1 (Invitrogen) and sequenced, then recloned into the EcoRI and SalI sites of expression vector pET28a (Novagen). The recombinant His-tagged protein was purified by Ni²⁺ chromatography (Novagen) and used to raise antibodies in chickens (Aves Labs). Recombinant Yuri protein cross-linked to NHS-activated Sepharose 4 Fast Flow (Amersham) was used for affinity purification. For immunoblots, samples were solubilized in SDS sample buffer, run on 12.5% polyacrylamide gels and blotted to Immobilon (Millipore) filters. Bands reacting with the affinity-purified antibody were detected with horseradish-peroxidase-conjugated rabbit anti-chicken antibodies (Sigma) and the West Dura reagent (Pierce).

For fertility testing, ⩾20 individual males or virgin females were placed with three w¹¹¹⁸ partners in food vials for 7 days, after which adults were removed. The original vials were checked for the presence of larvae, pupae and adults for a further 15 days. Although eggs were laid, yuri^F64 homozygous and hemizygous males never produced any viable progeny. A stock with deficiency Df(2L)do1, which removes yuri-containing region 35B1-35D2, balanced over a CyO-GFP balancer (Rudolph et al., 1999), was generated from crosses of stocks 3212 [Df(2L)do1, pr¹ cn¹/In(2LR)Gla, wg^Gla-1 DNApol-γ35²] and 5702 [w¹; noc^Sco/CyO, P{GAL4-Hsp70.PB}TR1, P{UAS-GFP.Y}TR1] from the Bloomington Stock Center. To generate embryos homozygous for Df(2L)do1 or homozygous for CyO-GFP, eggs were collected from the Df(2L)do1/CyO-GFP stock and left >24 hours to ensure that viable embryos hatched. Fluorescent and non-fluorescent embryos were collected separately. Third chromosomes carrying (1) a don juan-GFP construct (Santel et al., 1997) or (2) a GFP-PACT construct (Martinez-Campos et al., 2004) or (3) Flytrap lines ZCL0931, ZCL2183, ZCL2155 and G0147 (Kelso et al., 2004) were introduced into a w^–; yuri^F64 background. Mutation ms(3)05090 at the Dlc90F gene (Caggese et al., 2001) was from the Bloomington Stock Center.

The following yuri ESTs/cDNAs were sequenced: adult head GH14032; adult testis AT03435, AT15480, AT15149, AT19027 and AT25733; adult ovary GM26781 and GM25777; S2 cell line SD06513 and SD11641; embryo RE12523 and RE13793. Two clones from the testis, AT15149 and AT15480, end at the same 5′ residue at a point between exons 3 and 5. The region immediately 5′ to this point scores poorly in analyses designed to detect promoters. Given that these two cDNAs were prepared from the same RNA, we assume they have an incomplete 5′ terminus. However, their 5′-most sequence, which is not present in other cDNAs, is part of an intron between exons 3 and 5 (Fig. 1). These clones thus either (1) provide evidence for the variable presence of an additional exon (labeled 4 in Fig. 1), the 5′ boundary of which is not defined or (2) represent incompletely spliced transcripts. A paralog search in D. melanogaster was performed using the Yuri `PD' isoform (FlyBase) sequences and the BLASTP service at FlyBase. Because protein data sets are not available for all sequenced insect species, and to avoid spurious matches to coiled-coil domains, ortholog searches of translated DNA sequences were conducted with TBLASTN, using the 239-residue protein encoded by the GH14032 cDNA as the query and both the `nr/nt' NCBI database and the 21 insect genome sequences searchable at FlyBase as target data sets. The sequence identity computation for the ∼100 kDa Yuri protein in the 12 sequenced Drosophila sequences (Drosophila 12 Genomes Consortium, 2007) was performed using BLASTP on the GLEANR consensus protein data sets. Coiled-coil predictions were made using the COILS program at http://www.ch.embnet.org/software/COILS_form.html. Default settings were used in searches.

The P insertion yuri^c263 (Armstrong et al., 2006) and SUPor-P insertion KG03019 (Roseman et al., 1995) were mobilized with the Δ2-3 transposase at 99B (Robertson et al., 1988). Standard genetic schemes generated stocks of viable excisions. For lethal excisions, lines with GFP-marked balancers were prepared. Excisions were characterized by PCR. Precise deletion endpoints were determined by sequencing.

Testes were dissected in ice-cold phosphate-buffered saline pH 7.2 (PBS), fixed with 3% paraformaldehyde in PBS for 10 minutes and permeabilized by four washes in BBX (PBS + 0.3% Triton X-100 + 0.1% BSA) for 10 minutes. They were then incubated overnight at 4°C with rotation in BBX + 2% goat serum and one or more of the following antibodies: 1:100 affinity-purified Yuri antibody; 1:500 mouse monoclonal anti-γ-tubulin GTU-88 (Sigma); 1:200 rabbit polyclonal anti-Centrosomin antibody R19 (gift of T. Kaufman, Indiana University, Bloomington, IN); 1:50 mouse monoclonal anti-β-tubulin E7 (Developmental Studies Hybridoma Bank). After two washes each in BBX and BBX + 2% goat serum, appropriate Alexa Fluor-conjugated secondary antibodies (Invitrogen) were added at 1:500 in BBX + 2% goat serum and incubated for 2 hours. Rhodamine- or Alexa Fluor-conjugated phalloidin (Invitrogen) at 1:50 dilution was included with the secondary antibody as appropriate. After four washes with BBX, testes were mounted with 1:2000 Hoechst 33342 (Invitrogen) in 50% glycerol. Images were collected on a Zeiss Axioplan or on Zeiss LSM 410 and 510 confocal microscopes and processed with Metamorph (Molecular Devices) or Zeiss software.

TEM analysis was as described previously (Tokuyasu et al., 1972a), with minor modifications. Sections were cut at 700Å and stained with uranyl acetate and lead citrate. JEOL 1010, JEOL 1230 and Hitachi H-7500 electron microscopes were used.

We thank Dr R. P. Munjaal for contributions to the early phases of this work. We thank Dr Chris Bazinet for the dj-GFP line; Dr David Caprette for EM help; Dr James Fabrizio for Flytrap lines he characterized as expressing GFP in the cyst cells; Dr Thomas Kaufman for Centrosomin antibody; Dr Tatsuhiko Noguchi for critical insight into the actin sleeves in the yuri^F64; Dr Jordan Raff for the GFP-PACT line; Dr Kiyoteru Tokuyasu for helpful discussions on the nuclear localization of Yuri. We are grateful to Kenneth Dunner, Jr, Deborah Townley and Dr Wenhua Guo of the High Resolution EM Facility at MD Anderson, the Integrated Microscopy Core at Baylor College of Medicine and the Smalley Institute of Nano Science and Technology at Rice. respectively, for their assistance with EM work. The help of Rice undergraduates, in particular Summer Bell, Faraz Sultan and Anita Shankar, is gratefully acknowledged. These studies were supported by NIH grant RO1 HD 39766, grant C-1119 from the Welch Foundation of Texas and NASA grant NCC2-1356.