The fatty acid elongase NOA is necessary for viability and has a somatic role in Drosophila sperm development

Male germ cell development in Drosophila is dependent on the close contact of the germ cells with the surrounding somatic cyst cells. This contact is established at the beginning of the process in the tip of the testis tubes. A single germ cell, a spermatogonium, becomes enclosed by two somatic cyst cells, thus forming a cyst. Within this cyst, germ cells synchronously go through the mitotic amplification phase, growth phase, meiosis and through differentiation until individualization. Several signals have been identified during the early phase of spermatogenesis that either are initiated in the cyst cells and influence the surrounded germ cells or are initiated in the germ cells and influence the cyst cells (reviewed by Renkawitz-Pohl et al., 2005). During spermatocyte maturation the transcription factors eya (eyes absent) and so (sine oculis) are required in the somatic cyst cells for correct development of the spermatocytes (Fabrizio et al., 2003). How this signal is transmitted to the developing spermatocytes remains unclear, however. Even less is known about the germ cell-cyst cell contact after meiosis. Although the two cyst cells are not distinguishable until shortly after meiosis, they then clearly behave differently. One of them differentiates into a head cyst cell into which all 64 spermatid heads are anchored, the other develops into a tail cyst cell that surrounds the spermatid tails of 1.8 mm length (Lindsley and Tokuyasu, 1980). The head cyst cell finally is engulfed by cells of the terminal epithelium to allow coiling of the spermatid bundles towards the testis base (Tokuyasu et al., 1972).

In an enhancer trap screen for genes expressed in cyst cells we identified the essential gene neighbor of abl (noa), which encodes the long-chain fatty acid elongase. In addition to expression in various somatic tissues, in the testes noa is expressed in cyst cells and in cells of the terminal epithelium. noa expression is observed in the second half of sperm development, after the encased germ cells have completed meiosis. It is found in both the head and the tail cyst cells. Its function is required for male fertility since an RNAi construct selectively expressed in the cyst cells leads to male sterility.

By analyzing a collection of embryonic lethal enhancer trap lines (Spradling et al., 1999) for expression patterns in adult testes we identified two non-complementing P-element-insertion lines from genomic region 73B with β-galactosidase expression in the cyst cells, namely l(3)01895 and l(3)04106. A third line, l(3)02281, was described in FlyBase (Drysdale et al., 2005) as non-complementing and found to exhibit the same expression pattern (for an overview see Fig. 1A). In remobilization experiments with the insertion lines l(3)01895 and l(3)04106 wild-type revertant flies were observed, indicating that the observed lethality and the P-element insertion are causally related.

The enhancer trap pattern is shown in Fig. 1A. Owing to the cloning strategy the lacZ gene in the P-element vector contains a nuclear localization signal. Thus, β-galactosidase activity is observed in the nuclei of the expressing cells. In adult testes (Fig. 1A) β-galactosidase expression is seen at the tip of the testis tubes (arrow), in scattered nuclei along the tubes that are located along the elongating spermatid tails – and thus represent tail cyst cells (see diagram in Fig. 1) – and in a larger accumulation at the end of the tubes in the terminal epithelium in front of the seminal vesicles (arrowhead). As will be shown later, these nuclei belong to head cyst cells as well as to cells of the terminal epithelium. In the larval testes β-galactosidase activity is only observed in a small number of nuclei at one end of the gonad (Fig. 1B, arrow). Since the larval testes only contain germ cell stages up to the spermatocyte stage, this proves that the scattered nuclei in adult testes are from cyst cells around postmeiotic stages of germ cell development. To demonstrate clearly that both head and tail cyst cell show β-galactosidase activity isolated cysts were stained with Hoechst 33258. A bundle of elongating spermatids can be traced, on which two β-galactosidase-positive nuclei can be seen (arrow, arrowhead for the two cyst cell nuclei and double arrow for spermatid nuclei, Fig. 1C,D).

We cloned P-element flanking genomic sequences in plasmid rescue experiments and with these as a probe identified three incomplete but overlapping cDNA clones from a testis cDNA library. From their analysis the gene structure was derived as depicted in Fig. 1E. The gene contains five exons, spans a genomic region of 12.5 kb (Fig. 1E) and is located in polytene region 73B on chromosome 3L. It is flanked at the 5′ end by the gene for an alpha subunit of a G protein (Gf alpha) (Quan et al., 1993) and at the 3′ end by the gene Abl (Segev et al., 1988). We, therefore, named the gene neighbor of abl (noa). In addition, the snRNA:U12:73B gene resides within the large second intron of noa. The three testis cDNAs add up to 2385 nucleotides and contain a potential polyadenylation signal roughly 300 nt before the 3′ end of the cDNAs, indicating that longer transcripts should exist. Using the EST sequence in the flybase database (Drysdale et al., 2005), the 5′ end that extends furthest (LD 10431) was defined as the transcription start point of noa. The three P-element integration sites were determined by sequencing and reside 18 nucleotides (nt) in front of the transcription start site in the case of l(3)04106, 80 nt into the first exon in the case of l(3)01895 and 360 nt into the first intron in the case of l(3)02281, respectively (Fig. 1E).

In northern blotting experiments two transcripts were detected (Fig. 2A left), which were found to be present in both sexes. The relative amounts of the two RNA species vary, however, between the sexes. Females contain more of the shorter transcript, which seems to be predominantly of ovarian origin (see Fig. 4B), whereas males always contain higher amounts of the larger transcript. The short transcript in males is mainly present in testes but less abundant than the larger transcript and thus is underrepresented in the RNA of whole flies (see Fig. 2A). The smaller species of 2.3 kb length corresponds to the size expected if the polyadenylation signal within the cDNA sequence is used. In some experiments this RNA species is resolved into two distinct bands; this heterogeneity could either be due to different degrees of polyadenylation and/or to the presence of a 60 nt longer variant, detected in head RNA (see legend to Fig. 1E). Three additional canonical polyadenylation signals can be found in the genomic sequence approx. 700 bp, 870 bp and 980 bp, respectively, downstream from the first polyadenylation signal mentioned above. A radioactively labelled PCR product from the 700 bp interval (see Fig. 1E) selectively hybridizes to the longer RNA species of 3 kb thus proving that one of the further 3′ located polyadenylation signals is also used (Fig. 2A right).

To determine the exact size of the noa transcripts an RNaseH experiment was performed (Fig. 2B). Owing to incomplete cleavage, intermediates are formed that can be accounted for by the following processes. The 3 kb RNA is initially reduced by 100 nt, which correlates well with polyadenylation signal 2 (see above). The RNA fragment at 2.5 kb results from an internal A-rich region within the longest transcript (A′ in Fig. 1E) that binds oligo(dT) and thus is cleaved by RNaseH. The 2.3 kb RNA, which is generated from polyadenylation signal 1, is also initially reduced by about 100 nt. The shortest fragment at 1.4 kb finally results from cleavage at an A rich region at nt 1.372 of all transcripts (A in Fig. 1E, Fig. 2B left). Further digestion with RNaseH leads to almost complete cleavage to the 1.4 kb fragment, again supporting the above interpretation (Fig. 2B right).

The open reading frame extends from the second exon into the fifth and encodes a protein of 316 amino acids. It is common to all RNAs from the gene. In a computer search for known motifs the deduced protein shows similarity to a family of long chain fatty acid elongases. It contains all motifs that are found in fatty acid elongase (ELO) proteins (Fig. 3).

Searching the database with the NOA amino acid sequence reveals 20 similar proteins in the Drosophila genome, which reside in nine chromosomal regions (55E-F, 68A-B, 73B, 82B-C, 84F, 85E-F, 94B-C, 94D-E, 95D-E; supplementary material Table S1). For five of the genes no EST clones have been found so far. The ELO protein family is conserved among eukaryotic organisms and always consists of several members: yeast (3), Caenorhabditis (9), Anopheles (17), mouse (7), human (7), Arabidopsis (4), Xenopus (5), zebrafish (14) and Dictyostelium (8). The individual proteins differ in their N- and C-terminal ends, but are highly similar in their core region surrounding the histidine motif (see Fig. 3).

Among the Drosophila ELO proteins, NOA is the least conserved with a mean value of 18.3% amino acid identity to the other ELO proteins, which is in the same range as the conservation to the three yeast proteins SUR4 (also known as Elo3), FEN1 (Elo2) and Elo1 (19% identity). By contrast, the other Drosophila ELO proteins share 36.4% amino acid identity (mean value). In vertebrates the very long chain fatty acid elongase ELOVL6 shows the highest identity to NOA [45-46% in mouse (NP_569717), human (NP_076995) and in zebrafish (AAH59459); Fig. 3]. This value far exceeds the identities observed among the Drosophila proteins (see above), suggesting that NOA is the Drosophila homologue of ELOVL6.

Using a balancer chromosome with an actin-GFP transgene we determined the lethal phase of the P alleles by screening for non-fluorescent individuals. In line l(3)02281 homozygous third instar larvae were never observed. In this line a second P element resides at polytene region 61D. Thus, deleterious effects of both integrations could result in a more severe phenotype and could obscure dose-dependent quantitative effects in homozygotes. This is unlikely, however, since the lethal phenotype of l(3)02281 is not complemented in crosses with a deletion in 73B (see Materials and Methods) or with the two other P-element lines for noa (see below) indicating that the lethality observed in l(3)02281 is caused by the insertion in noa. In addition, lethality could be rescued by the introduction of a wild-type copy of the noa gene (see below).

Homozygous l(3)01895 individuals reached larval third instar and in the case of allele l(3)04106 some homozygous individuals even survived to adulthood. These adult flies were severely impaired in motility and had a drastically reduced lifespan. Males proved to be sterile, but their testes showed no sign of a spermatogenesis defect and they contained motile sperm. Thus, their extremely impaired motility and vitality must be the reason for the observed sterility. Northern analyses demonstrated that all noa transcripts are synthesized in the mutant third instar larvae but at reduced levels compared with the wild type (Fig. 4A). More importantly, the extent of reduction parallels the effect on fitness and viability, i.e. RNA from homozygous l(3)04106 larvae contained much higher levels of noa transcripts than RNA from l(3)01895 larvae both in northern hybridizations (Fig. 4A) and in in situ hybridization experiments on larval tissue (data not shown). Compared with the constant levels of ribosomal protein L9 mRNA the amount of the 3.0 kb noa transcript was reduced to about half in l(3)04106, whereas the 2.3 kb transcript was almost undetectable. The latter is probably due to the fact that the 2.3 kb RNA is not synthesized in larval ovaries. In the case of l(3)01895 even the 3.0 kb transcript was barely visible. Thus, the function of noa must be dose dependent. The analysis of viability in various allelic combinations supports the notion of a dose-dependent function of noa. The flies were just as affected as the occasional survivors mentioned above (data not shown).

To unequivocally correlate noa function with the observed phenotypes a rescue construct was generated (see Fig. 1E) and transformed lines carrying the noa transgene on the second or the first chromosome were established. To test whether this additional wild-type copy of the gene can restore viability, transgenic flies were crossed to the three different P-element-induced noa alleles. With seven of eight different integration lines homozygous l(3)02281, carrying one copy of the transgene, survived to adulthood, although these flies still showed a reduced lifespan. Adult flies with a normal lifespan are observed when four copies of the noa transgene are introduced. This result is in agreement with our earlier interpretation that the encoded gene product has a dose-dependent effect and also indicates that the transgene does not support full wild-type activity.

As described above, two RNA species are synthesized from the noa gene (3.0 and 2.3 kb; Fig. 2A). In northern experiments these RNAs were shown to be present in both sexes, mainly in the gonads, but also to a lesser extent in the remaining parts of adult flies (Fig. 4B). Whereas the amount of the shorter transcript exceeds that of the larger transcript by far in ovaries, in testes – as in all other tissues – the larger transcript was still the predominant form of the noa RNA. In staged embryos the small RNA, of 2.3 kb, was found to be very prominent in 1-hour-old embryos but then quickly declined, in contrast to the increasing amounts of the 3 kb noa transcript (data not shown). It was also predominant in eggs laid by virgin females. Thus, the maternal store mainly consists of the small noa transcript. Apart from this difference both transcripts were found throughout development without significant qualitative or quantitative changes (data not shown).

To identify the cell types expressing noa three independent experiments were performed: (1) analysis of the β-galactosidase pattern generated by the enhancer trap function in the P-element insertion lines; (2) in situ hybridizations with noa cDNAs; and (3) analysis of the expression pattern of noa fusion genes in transgenic flies.

Whole-mount in situ hybridizations on embryos showed expression in the syncytial blastoderm (Fig. 5C), again demonstrating that noa transcripts are maternally provided (see above) since at this stage the embryonic genome is still transcriptionally inactive. In early cellular blastoderm, noa transcripts localized to the apical region of the newly forming cells (Fig. 5D,E black arrow) but the pole cells did not contain noa transcripts at all (Fig. 5E, white arrow). Later, transcript levels gradually decreased. They could be found at low levels in the involuting mesoderm as well as in the invaginating foregut and hindgut. During organogenesis, transcripts were again detectable and accumulated in the CNS (Fig. 5F) as well as in segmental groups of cells from the peripheral nervous system (PNS; Fig. 5G). The gene is still active in the CNS of third larval instar as judged by the persistence of β-galactosidase activity from a noa-lacZ reporter fusion (Fig. 5I). In addition, transcripts were present at low levels in all imaginal discs (Fig. 5H). High expression levels were observed in the larval fat body by means of the enhancer trap function (see Fig. 1B and Fig. 8C).

In the adult fly noa was found to be transcribed in a variety of tissues such as regions with sensory organs, the fat body, the malpighian tubules, parts of the intestine (data not shown) and in specific cells of the gonads. In the ovary, hybridization occurred in the nurse cells as well as in the follicular epithelium that surrounds the growing follicle (Fig. 5B). In early stages of oogenesis germ cells showed β-galactosidase activity from the enhancer trap in all regions of the germarium prior to follicle formation (Fig. 5A). A clear cell-type-specific noa expression could be observed in adult testes. Transcripts accumulated around elongating spermatid bundles (Fig. 7E top). This is consistent with the observation that β-galactosidase expression in the enhancer trap lines occurs selectively in the cyst cells that surround each synchronously developing group of germ cells in postmeiotic stages (see below, Fig. 1A, Fig. 8A). Expression was also seen in the region of the terminal epithelium (Fig. 1A, Fig. 8A). In addition, a weak but reproducible expression was observed in the tip of the testis tubes (Fig. 7A).

A noa-lacZ reporter fusion containing the first 20 amino acids of the noa protein and a full length noa-GFP fusion protein displayed basically the same pattern as seen in the enhancer trap line and in in situ experiments (e.g. Fig. 7F and data not shown), indicating that the enhancer trap function faithfully represents the expression pattern of noa.

To determine the amount of 5′ flanking sequences necessary for noa regulation, promoter fusions to the lacZ gene were generated (Fig. 6). 204 nt and 493 nt of 5′ flanking sequences together with 148 nt of the 5′ UTR can no longer maintain noa expression when fused to a synthetic AUG in front of the lacZ gene (Fig. 6 top). If the entire noa 5′ UTR (212 nt) with the translation start site is included in the construct, however, correct expression is resumed (Fig. 6 middle). This may reflect a higher translation rate from the gene-specific translation start site or the presence of additional promoter elements in the 5′ UTR.

Independent of the extent of 5′ sequences in the gene fusion the 3′ UTR is also necessary for correct noa expression (Fig. 6 bottom). A 1.1 kb segment including the first polyadenylation signal 930 nt after the stop codon increased overall expression. Addition of the sequences to the last possible polyadenylation signal, approx. 2 kb after the stop codon, increased the expression level further. Altogether expression was increased several fold compared to the construct with the hsp70 3′ UTR. Furthermore, expression in imaginal discs, which was undetectable with the shorter 3′ UTR, could now be observed (data not shown). This must mean that regulatory elements reside in both parts of the 3′ UTR and that the shorter transcripts also contain some of these elements. Whether the increase in expression level is due to higher transcriptional activity or to increased stability of the RNA or even to altered translational efficiency has not been determined.

To identify the cell types expressing noa in the testis, in situ hybridizations were performed. Transcripts were detected close to elongated spermatid bundles (Fig. 7E top). In in situ hybridizations with a germ-cell-specific probe Mst87F (Kuhn et al., 1988a) that labels the entire bundle (Fig. 7E bottom), noa transcripts were seen at the periphery of the spermatid bundles, i.e. in the cyst cells. The staining along the elongating bundles must reflect noa activity in the tail cyst cells, whereas an intense staining in the region of the terminal epithelium could reflect activity in the head cyst cells and/or in the cells of the terminal epithelium itself (data not shown). This cell-type-specific expression was also observed using the enhancer trap properties of the residing P elements where lacZ expression was detected in nuclei of only those cyst cells that surround postmeiotic stages of sperm development (see also below, Fig. 1A,B and Fig. 8A,C). A noa-lacZ fusion and a noa-GFP fusion exhibited the same expression pattern (e.g. Fig. 7F). Thus, cyst cell expression of noa starts after the surrounded germ cells have completed meiosis and differentiation begins. Earlier cyst cell stages are free of noa RNA and protein.

In situ hybridizations in cyst cells are at the limits of detection because of the very thin cell body. Therefore, since the enhancer trap activity proved to faithfully reproduce the gene activity of noa, we used the enhancer trap function to clearly demonstrate expression in both cyst cells. Fig. 7C shows an isolated cyst containing early elongating spermatids surrounded by two cells with β-galactosidase-positive nuclei. Hoechst 33258 staining, to locate the spermatid nuclei, permits the identification of the head (arrowhead) and the tail cyst cell (arrow, Fig. 7D).

The cells that exhibit β-galactosidase activity in the testis tip were analysed by antibody staining. An antibody against Fasciclin III was used to identify the cells of the hub (Fig. 7B, red). Anti-β-galactosidase antibody then revealed that the lacZ-positive cells are clustered around the hub in a loose arrangement (Fig. 7B, green). As judged by their number (up to 20), counted in stacks using the confocal microscope, these cells should represent the cyst progenitor cells and not the stem cells of which there should be five to seven.

Since cyst cells are postulated to communicate with germ cells the question arises of whether the development of cyst cells is coordinated with germ cell development, i.e. whether expression of noa is dependent on the development of the encased germ cells. This can be tested if enhancer trap alleles are crossed into mutants that block spermatogenesis (benign gonial cell neoplasm; bgcn) (Gateff, 1981) or do not contain any germ cells at all (tudor; tud) (Boswell and Mahowald, 1985) and alterations in the β-galactosidase pattern are monitored in comparison with the wild-type situation (Fig. 8A,B). In bgcn mutant testes in which spermatogenesis stops during the mitotic proliferation phase and postmeiotic stages never develop β-galactosidase activity is seen in a tightly packed ring of nuclei towards the end of the testis tube (Fig. 8E). These nuclei are too close together to be cyst cell nuclei that surround germ cells. Rather they are nuclei of the terminal epithelium (the extent of which is shown in Fig. 8A,E,G,H in the region between the arrowhead and the arrow). This can be seen very clearly in tudor mutant testes which do not contain any germ cells at all (Fig. 8G). Here the seminal vesicle and the region of the terminal epithelium are well separated and β-galactosidase-positive nuclei are seen at the beginning of the terminal epithelium (black arrowhead in Fig. 8G). Thus, noa expression in cyst cells seems to be dependent on signals emanating from the germ cells or on the differentiation state of the cyst cells themselves. The β-galactosidase-positive nuclei at the tip of the testis (double arrow in Fig. 8A,C,E) remain. By contrast, in the enhancer trap line l(3)04708 the cyst cell nuclei are labelled at the cyst cell progenitor stage and throughout the proliferation phase (Fig. 8B,D; M.A.S., unpublished). This enhancer trap line still shows β-galactosidase-positive nuclei in the tip region of the testes when crossed into a bgcn mutant background (Fig. 8F). In the tud mutant testes β-galactosidase activity is observed in a region above that of the terminal epithelium, demonstrating that this activity is independent from a germline signal and that cyst cells are still present in agametic testes. In addition, the number of stained nuclei proves that the cyst progenitor cells have undergone mitoses even though cysts cannot be formed.

To specifically affect noa transcripts in the cyst cells we generated a PPY-GAL4 line with the promoter of the cyst-cell-specific phosphatase PPY (Armstrong et al., 1995). This was used to activate a noa RNAi construct. Since at 25°C fertility was only partially affected, cultures were shifted to 29°C to enhance Gal4 expression levels. At this elevated temperature fertility is generally affected and OreR control males show almost no sperm motility whereas sperm development itself appears completely normal. By contrast, sperm cell development was specifically affected in the males expressing the PPY-GAL4 driven noa RNAi construct (Fig. 9). Many testes were smaller and contained fewer germ cells or cysts than the wild-type control (Fig. 9A,B). The number of individualisation cones (ICs) was drastically reduced. If ICs had formed they frequently appeared dispersed and overall morphology was distorted (Fig. 9F). In bundles with ICs next to their nuclei frequently some nuclei were dislocated and had no IC (Fig. 9D) whereas in wild-type testes the occasional mislocated nucleus would be accompanied by an IC, as in the rest of the bundle (Fig. 9C). Many elongated spermatid nuclei were observed scattered around in the massively enlarged region of the terminal epithelium. In addition, no individualized sperm could be found in the seminal vesicles (Fig. 9B,D). In general, the effect of RNAi was very variable, ranging from a severe phenotype (Fig. 9A) to a weak phenotype leaving testis morphology relatively normal (Fig. 9F). In an analogous experiment with the T80-GAL4 driver line, which induces expression in imaginal discs, complete lethality resulted, proving the functionality of the noa RNAi construct. Phosphatase PPY was shown to be weakly expressed also in spermatocytes (Armstrong et al., 1995). To formally rule out that the observed effects could result from RNA degradation in spermatocytes we raised the flies at 18°C, shifted them to 29°C and analyzed germ cell morphology 4-5 days later when the early postmeiotic stages had developed into individualizing sperm. Under these conditions the results were the same, thus proving that postmeiotic cyst cell expression is the cause for the defect.

The first elongase described in the mammalian system (CIG30) (Tvrdik et al., 1997) was found to be involved in the recruitment of brown adipose tissue, and its transcription was increased upon cold treatment. In addition, a male-specific elongase in Drosophila has been described recently that showed higher transcript levels at 21°C compared with 24°C and 29°C (elo68α) (Chertemps et al., 2005). This prompted us to test whether noa transcript levels are also temperature dependent. Newly eclosed adult flies were kept at 10°C for 1 or 3 days, before northern analyses were performed. There was no increase in any of the three transcripts at 29°C compared with the situation at 24°C. Transcript levels were also not increased in third instar larvae after the same cold treatment. In addition, if adult flies or third instar larvae were kept for 4 days at 21, 24 or 29°C, transcript levels again remained unaltered (data not shown).

We have presented the cloning and expression pattern of noa, a gene encoding a long chain fatty acid elongase. P-element-induced noa mutant alleles are lethal or semilethal, depending on the level of noa expression in these alleles. As demonstrated by rescue experiments and precise excision of the residing P elements, loss or even only reduction of noa expression is the cause of lethality. Thus, noa is an essential gene and its function is dose dependent.

In Drosophila as in mammals and yeast the majority of fatty acids are composed of species with a maximum chain length between 14 and 18 carbon atoms. A minority, the so called `very long chain fatty acids' consist of 20 or more carbon atoms. They result from further elongation by the very long chain fatty acid elongases (elo vl) and participate in many biological processes since they are the substrate for sphingolipid formation. Sphingolipids are important for cell growth (Wells and Lester, 1983; Hanada et al., 1992), contribute to the epidermal water barrier (Wertz, 1992) and play a role in cell recognition and adhesion (Hakomori and Igarashi, 1995). In addition, sphingolipids and their degradation products are involved in signal transduction (Hannun, 1996; Spiegel and Merril, 1996; Testi, 1996) and formation of lipid rafts (Simons and Ikonen, 1997).

Data relating to the function of long chain fatty acid elongases have been gathered mostly for yeast [three genes, reviewed by Tehlivets et al. (Tehlivets et al., 2007)]. For mouse and human, seven genes have been reported so far (reviewed by Leonard et al., 2004). The first reported gene in mouse, Elovl3 (also known as Cig30) (Tvrdik et al., 1997) plays a role in recruitment of brown adipose tissue upon cold stimulation. In addition, it is important for the formation of neutral lipids that are necessary for normal skin function (Westerberg et al., 2004). In the human, a mutation in ELOVL4 causes Stargart-like macular dystrophy, a dominant form of juvenile macular degeneration (Edwards et al., 2001; Zhang et al., 2001). A high expression in testes is found for ELOVL5 (also known as HELO1) (Leonard et al., 2000) as well as for ELOVL6 (see below). In comparison with Drosophila (and other insects) the number of genes found in mouse and man is rather low. Comparable heterogeneity is generated in this system, however, by a greater variability in expression at the transcript and the protein level for most of the genes. For ELOVL5, for example, ten different transcripts are found, from which seven protein isoforms could be generated (AceView at NCBI).

NOA is the homologue of the mammalian ELOVL6, which is involved in the addition of 2-carbon units to C12-C16 fatty acids (LCE) (Moon et al., 2001). Elovl6 is highly expressed in liver, brown adipose tissue and also in testes. For this gene, too, five different transcripts have been found, which could generate four different protein variants. Tissue-specific expression, however, has not been described.

In Drosophila not much is known about long chain fatty acid elongases, although 20 genes can be identified by their homology to the three yeast genes Elo1, FEN1 (Elo2) and SUR4 (Elo3). For five of these genes the expression is unclear since EST clones have not been identified yet. The large number of elongases in the Drosophila genome could permit highly cell-type-specific expression. Thus, the various genes could cover different and complementing needs for long chain fatty acids. In this case reduction of activity for any of the genes should result in a phenotype. Alternatively, the large number of ELO genes could be expressed in largely overlapping patterns and one gene might be able to compensate for the loss of another gene. Recently one of these elongases, elo68α, has been reported to be expressed male specifically in cells of the reproductive tract (Chertemps et al., 2005).

Knockdown of noa activity in imaginal discs resulted in lethality proving that compensation by other elongases does not occur in this tissue. Also, reduction of noa activity in general leads to reduced viability or even lethality. Apparently none of the other elongases identified in the Drosophila genome can compensate for the decrease in noa function in any of the affected tissues, even though several genes are expressed in the same tissues as judged by the EST clones. Thus, noa should have a function that is distinct from that of other elongases in the Drosophila genome. Such a unique function may also be reflected by the low level of similarity to all other Drosophila elongases. By contrast, when the noa RNAi construct was selectively driven in the follicle cells, no phenotype was observed. Thus, in the follicle cells other fatty acid elongases must be able to compensate for the loss of noa activity.

The noa 3′ UTR is unusually long (900 nt to polyadenylation signal 1 and 1.6 kb in the case of polyadenylation signal 2). This it is about half the size of the entire mRNA and suggests that regulatory elements reside within this gene segment. In accordance with this assumption, addition of the 3′ UTR to lacZ fusions enhanced the overall expression level dramatically and allowed expression in imaginal discs to be observed. A very long 3′ UTR was also reported for the NOA homologue in mouse ELOVL6 (also known as FACE) (Moon et al., 2001; Matsuzaka et al., 2002) and for ELOVL2 (also known as SSC2), which shows particularly high expression in testes (Tvrdik et al., 2000). By contrast, ELOVL1 (SSC1) contains a very short 3′ UTR (Tvrdik et al., 2000). Sequence analyses gave no indication for an involvement of micro RNAs in the regulation via the long 3′ UTR.

In the male gonad noa expression is mainly observed in the cyst cells and in cells of the terminal epithelium. Cyst cell expression is restricted to the postmeiotic stages, when spermatid differentiation takes place. In earlier analyses it was demonstrated that the germ cells affected cyst cell fate during early spermatogenesis (Gönczy and DiNardo, 1996). In agametic testes, cyst cell proliferation continued and some cyst cells adopted hub cell fate. In agreement with these data, germ cell maturation seems necessary for correct onset of noa expression in the surrounding cyst cells. Its `postmeiotic' expression is prevented if germ cells are either not present or never reach the corresponding maturation state. Only expression in the terminal epithelium can be observed. In the tumorous testes of the bgcn mutant it cannot be excluded that the β-galactosidase-positive nuclei are composed of cyst cell nuclei and nuclei of the terminal epithelium. In that case, however, cyst cells must have survived the degeneration process of the undifferentiated germ cells and become concentrated in the area of the terminal epithelium. By contrast, in l(3)04708 flies cyst-cell-specific expression is maintained during the proliferation phase. In the case of the mutant bgcn this could be due to the fact that expression is initiated early, when germ cell development is still correct. However, it is also initiated when germ cells are not present as in tud testes. Therefore, this activity is independent of germ cell development.

When activity of the specifically testis-expressed elongase elo68α in Drosophila was reduced to 1%, a 50% decrease of vaccenyl acetate levels was reported, but no effect on fertility or viability resulted (Chertemps et al., 2005). By contrast, a cyst cell-specific knockdown of noa did result in male sterility. The defect is observed at the end of spermatogenesis when sperm should individualize and enter the seminal vesicle. Instead they seem to separate before maturation and then disintegrate. If indeed noa function in cyst cells is part of a signal to the germ cells it could be possible that noa knockdown results in too low a signal to permit engulfment by cells of the terminal epithelium. Guidance cues might be missing and final maturation of sperm is prevented. Future experiments are aimed at identifying other components of this potential signalling pathway in which noa plays a role.

A collection of embryonic-lethal enhancer trap lines (Spradling et al., 1999) induced by insertion of the PZ element (Mlodzig and Hiromi, 1992) was kindly provided by H. Jäckle. In this collection we identified l(3)01895 and l(3)04106 with cyst-cell-specific β-galactosidase expression.

Enhancer trap line l(3)02281 was described in FlyBase as noncomplementing to l(3)01895 and l(3)04106. We found that l(3)02281 contains a second P insertion in polytene region 61B3. It is located in the short intergenic region between the divergently transcribed genes E(bx) and mthl14.

Strain l(3)02281 as well as the deficiency strains Df(3L)st8P, In(3L)P, ry^*/TM6B and Df(3L)st4, gl[2]e[4]/TM6 were obtained from Bloomington Stock Center. bgcn mutant flies were kindly provided by E. Gateff and tudor mutant flies by R. Boswell.

All fly cultures and crosses were kept on standard fly medium at 25°C or 18°C.

All methods followed standard protocols (Sambrook et al., 1989).

Poly(A)⁺ RNA was isolated directly from tissues and developmental stages using oligo(dT)-coated beads (Dynal and Novagen). RNAs were separated in formaldehyde-agarose gels, transferred to Nylon membrane (Hybond-N, Amersham) and crosslinked by UV light. Hybridization with α-³²P-labeled in vitro transcripts was performed as described previously (Kuhn et al., 1991). RNaseH experiments were performed as described in Hollmann et al. (Hollmann et al., 2002).

Plasmid rescue of P element flanking genomic sequences was performed as described in Steller and Pirrotta (Steller and Pirrotta, 1985). A 14.5 kb long genomic clone was isolated by screening a bacteriophage λ library (Stratagene) with the genomic plasmid rescue clone. cDNA clones were obtained by screening a custom made testis cDNA library (Stratagene) with genomic fragments. Sequencing of genomic and cDNA clones was done by the dideoxy method (Sanger et al., 1977). Database searches were performed with the BLAST program (Altschul et al., 1990) using the WWW server at NCBI and the WWW server of the Berkeley Drosophila Genome Project (BDGP). Alignments were done with the ClustalW program (Higgins et al., 1994) at the EMBL-EBI server and processed with the Jalview program (Clamp et al., 2004). Sequences were deposited under the accession nos. AF279257, AF279258 and AH009765. Sequence information on more than 50 ESTs was derived from BDGP.

For the rescue construct, a genomic clone was combined with an RT-PCR-derived fragment and one of the cDNA clones to include 727 bp upstream of the transcription start site. The 3′ part of the gene was added from a genomic clone that extends 685 bp beyond the last polyadenylation signal. The combined fragments were cloned into pCasPer4 (Pirrotta, 1988).

For promoter lacZ fusions, PCR products of the desired length were amplified with primers containing suitable restriction sites (XbaI and KpnI for pWATGlac1 and KpnI for pWlac2). For constructs excluding part of the noa 5′ UTR pWATGlac1 (Kuhn et al., 1988b) was used, which contains a synthetic ATG codon. The lacZ vector pWlac2 (J. Weinert and U. Schäfer (Max-Planck-Institute for Biophysical Chemistry, Göttingen, Germany), personal communication) was used for fusions of the noa 5′ UTR (including the noa translation start site) to lacZ. To test the role of the 3′ UTR, PCR-amplified 3′ UTR fragments were ligated into a BamHI site at the 3′ end of the lacZ cassette in the existing pWlac2 constructs.

For a noa-GFP fusion construct a PCR fragment containing the noa promoter and the coding region was amplified using the rescue construct (see above) as a template. This fragment was cloned in frame to the GFP cassette in the pUAST-green vector (U. Renner, M.H. and M.A.S., unpublished). The UAS sequences of the vector remain in front of the gene's promoter and have no influence on GFP expression even if a GAL4 driver is crossed into the transgenic flies (M.H., unpublished).

For the RNAi construct the noa coding region was amplified and cloned back to back separated by a 400 bp GFP stuffer fragment into pUAST (Brand and Perrimon, 1993), as in Piccin et al. (Piccin et al., 2001). To drive the noa RNAi construct specifically in cyst cells, the PPY-Gal4 line was established. A 1.5 kb promoter fragment of PPY (Armstrong et al., 1995) was amplified, cloned into the vector p221-Gal4. All constructs were injected into w¹¹¹⁸ embryos according to the standard protocol (Spradling, 1986).

Whole-mount in situ hybridizations with digoxigenin-labelled DNA probes were performed following the protocol of Tautz and Pfeifle (Tautz and Pfeifle, 1989) with modifications as described by Hollmann et al. (Hollmann et al., 2002).

β-Galactosidase staining was performed according to Glaser et al. (Glaser et al., 1986), using glutaraldehyde as fixative. To reduce nonspecific background staining 0.25% Triton X-100 was added to the staining solution. Isolated cysts were counterstained under the coverslip with 2 μg/ml Hoechst 33258 using a filter paper to soak off the buffer and replace it with staining solution.

Testes of PPY-GAL4/noa-RNAi and OregonR wild-type flies as control were dissected in PBS, fixed for 20 minutes with 4% formaldehyde in PBS and washed twice in PBT. Staining for nuclei (1 μg/ml Hoechst 33258) and individualization cones (0.25 μg/ml TRITC-phalloidin) was for 10 minutes followed by three washes for 10 minutes each in PBS. Testes were mounted in PBS-glycerine (1:1) and analyzed by epifluorescence microscopy (Zeiss Axiophot).

For antibody stainings, testes were fixed for 20 minutes with 4% formaldehyde in PBS and then washed twice in PBT. Primary and secondary antibody incubations were performed overnight at 4°C with several washes in PBT at room temperature. Rabbit anti-β-galactosidase (Cappel) was used at 1:500; mouse anti-Fasciclin 1:50; Alexa Fluor 488-conjugated goat anti-rabbit (Molecular Probes) 1:400; Cy3-conjugated sheep anti-mouse (Sigma) 1:200. Testes were mounted in PBS-glycerine (4:1) and analyzed by epifluorescence microscopy (Zeiss Axiophot) and by confocal microscopy (Leica TCS-SP).

We gratefully acknowledge the skilful technical assistance of Tanja Wilhelm, Nicole Schleinschok and Christine Otto in different parts of the work. We appreciate the supply of fly stocks from the Bloomington Stock Center, Herbert Jäckle, Elisabeth Gateff and Robert E. Boswell. We appreciate the help of Harald Rühling and Markus Maniak with the confocal microscopy and the gift of antibodies from Ralph Rübsam and Jürgen Büning. We thank Karl Heinz Glätzer for valuable advice on in situ hybridizations, Ulrich Schäfer for careful reading of the manuscript and Anke Eberhardt for excellent secretarial help. This work was supported by the Deutsche Forschungsgemeinschaft (SFB 271).