Expression of the Drosophila gonadal gene: alternative promoters control the germ-line expression of monocistronic and bicistronic gene transcripts

Sex determination in Drosophila is achieved through a hierarchy of genetic interactions (reviewed in Baker and Belote, 1983; Cline, 1985; Nothiger and Steinmann-Zwicky, 1985; Baker el al. 1987; Wolfner, 1988; Hodgkin, 1989). The X:A chromosome ratio serves as the primary signal to set the activity state of the master regulatory gene Sex-lethal (Sxl) (Cline, 1978; Cline, 1983). Around the blastoderm stage of embryonic development, this ratio is assessed in individual cells with Sxl activated in female cells (X:A=1) and left inactive in male cells (X:A=0.5) (Baker and Belote, 1983; Sanchez and Nbthiger, 1983; Gergen, 1987). Sxl function in chromosomal females controls a network of genes that regulate different aspects of sexual development, including dosage compensation and sexual differentiation in the soma and germ line. The activities of these genes, in turn, control the terminal differentiation genes responsible for sexually dimorphic development.

Limited information is currently available on the control of germ-line sex determination and differendation. Studies on the ovo locus demonstrated that ovo is required for maintenance of the early female germ line; however, mutations in the gene failed to affect the male germ line (Oliver et al. 1987). The authors suggested that ovo may represent one of the genes functioning early, and perhaps primarily, in germ line sex determination. A recent report by Steinmann-Zwicky and colleagues (1989) demonstrated that both genetic and inductive signals are required for germ-cell differentiation. That is, the X: A ratio represents a cell-autonomous genetic signal in that transplanted germ cells develop according to their chromosomal ratio in ovaries, with XX cells undergoing oogenesis and XY cells forming spermatocytes or undifferentiated germ cells. A somatic inductive signal is also involved in that both transplanted XY and XX germ cells enter the spermatogenic pathway in testes. This event could result from the presence of a male-determining signal that induces all germ cells to initiate spermatogenesis or the absence of a female-determining signal required for XX cells to undergo oogenesis. Both the genetic and somatic inductive signals work through the Sxl gene to determine sex in the germ line. Normal Sxl function is essential for the female pathway of germ-cell development (Schiipbach, 1985; Steinmann-Zwicky et al. 1989) whereas a Sxl constitutive mutation overrules the somatic induction of XX germ-cell development in the male environment (Steinmann-Zwicky et al. 1989).

Nothiger and Steinmann-Zwicky (1985) suggested that a set of germ-line regulatory genes, analogous to those controlling somatic sex determination, may exist downstream of Sxl and function in controlling sexual differentiation in the germ line. Potential candidates for these regulatory sequences include otu (King et al. 1986) and sans fille (Oliver et al. 1988) (also called Hz [Steinmann-Zwicky, 1988]), two genes known to function in female germ-line development, and tra-2 (Baker and Ridge, 1980; Belote and Baker, 1982) and maleless (Bachiller and Sanchez, 1986), two genes required for male germ-line development. Such germ-line control genes might, in turn, regulate additional genes expressed during gametogenesis, resulting in male- and female-specific germ-line products in the differentiated sperm and egg.

The identification of additional germ-line expressed genes, and their subsequent genetic and molecular characterizations, should yield important information on germ-cell differentiation. Several germ-line expressed Drosophila genes have been identified and characterized through molecular screens (Schafer, 1986; DiBenedetto et al. 1987), analysis of transcriptionally complex regions (Yanicostas et al. 1989), and investigation of tissue-specific expression of members of multigene families such as the a- and /5-tubulins (Bo and Wensink, 1989; Michiels et al. 1989). In most instances, these genes are expressed sex-specifically in either the male or female germ line. Through the detailed analysis of a transcriptionally complex region of chromosome 3, we have discovered gonadal (gdl), a gene that is differentially expressed in the male and female germ lines (Schulz and Butler, 1989; Schulz et al. 1989). Here we report our further investigation of the germ-line expression of gdl. Our current studies indicate that, first, gdl is transcribed and translated early during germ-cell differentiation; second, gdl^M transcripts present in the male germ line may serve as bicistronic mRNAs; third, gdl gene expression may be controlled at both the transcriptional and translational levels; and fourth, the transcriptional control elements responsible for gdl male and female germ-line expression are separable. These results suggest that gdl may serve as an important tool in the further analysis of germ-cell sex determination and differentiation in Drosophila.

The gdl^M 1200 cDNA was isolated in a screen of an adult male cDNA library as previously described (Schulz and Butler, 1989). The EcoRI DNA insert containing the full-length cDNA was excised and subcloned into the pBS RNA expression vector (Stratagene) using standard protocols (Maniatis et al. 1982). A PstI fragment including nucleotides +1 to + 159 of the gdl^M 1200 cDNA was deleted to generate a full-length gdl^F 1000 cDNA cloned in pBS. cRNAs were synthesized, in the presence of the cap analog GpppG, by T3 or T7 polymerase transcription of the linearized gdl^M 1200-pBS or gdr 1000-pBS DNA templates as recommended by the enzyme supplier (Stratagene). Subsaturating amounts (0.1 to 0.25 ;zg) of the cRNAs were translated in a rabbit reticulocyte lysate (Promega Biotec) supplemented with [³H]leucine (Dupont NEN, 143 Ci mmole^- ) using the protocol supplied by Promega. The translation products were separated on 15 % polyacrylamide-SDS gels containing 10% glycerol (Laemmli, 1970); ⁴C-labeled low molecular weight protein standards (BRL) were used as size markers. Gels were processed in ENHANCE (Dupont NEN), dried, and exposed to Kodak X-OMAT AR film. After development, the gdl translation products were quantitated by densitométrie scanning of the autoradiograms using an LKB 2202 Ultroscan densitometer.

The five gdl-lacZ gene fusions were constructed in the CaSpeR transformation vector (Pirrotta, 1988). gdl(193)lacZ was constructed by inserting a lacZ gene cassette, derived from pMC1871 (Casadaban et al. 1983), into the unique BamHl site of the 1.8 gdZ-CaSpeR transposon described in Schulz and Butler (1989). This generates a fusion gene with the lacZ coding sequence placed in frame with the gdl ORF193 coding sequence. gdl(193)lacZ-ΔP was constructed by deleting a 487 bp Psti fragment from gdl(193)lacZ. This corresponds to the deletion of gdl sequences between -328 and +159 relative to the gdl^M transcription initiation site. gdl(193)lacZ-ΔP/E was constructed by deleting a 630 bp EcoRI fragment from gdl(193)lacZ-ΔP. This corresponds to the deletion of gdl sequences between +815 and +1444 relative to the gdl^M transcription initiation site. gdl(39)lacZ was generated through a multistep plasmid construction. A 783 bp EcoRI-BvulI fragment (coordinates —600 to +183) was directionally cloned into the EcoRI and Smal sites of pBS. This plasmid was then linearized at the polylinker BamHl site adjacent to the Smal site and treated with Bal 31 nuclease. DNA ends were repaired by SI nuclease and Klenow treatment and BamHl linkers were added using the procedure of Perbal (1984). Shortened EcoRI-BamHI fragments were inserted back into pBS and several BamHl 3’ ends were sequenced by the dideoxy method (Sanger et al. 1977) to identify subclones that still maintained the ORF39 initiator AUG while allowing the in frame insertion of the lacZ cassette. A 732-bp EcoRI-BamHI fragment containing 12 bp of ORF39 was then subcloned into CaSpeR followed by the insertion of the lacZ coding sequence in the BamHl site. gdl(193)lacZ-Δ39 was also generated through a multistep plasmid construction. Starting with the gdl(193)lacZ fusion gene cloned in pBS, exonuclease 111 was used to delete gdl sequences —328 to +130 relative to the gdl^M transcription initiation site. Subsequently, the DNA ends were repaired, Bc/I linkers were added and the plasmid was self-ligated. A Bc/I-EcoRI fragment, containing gdl sequences +130 to +810 and the lacZ gene cassette, was then excised from the plasmid. This DNA was directionally cloned into the BamHl and EcoRI sites of a pBS plasmid containing gdl sequences from —328 to +118; this latter plasmid had been previously prepared by Bal 31 nuclease digestion. Subsequently, a Bc/I-EcoRI fragment, containinggdl sequences from -328 to + 118 and +130 to +810 and the lacZ gene cassette, was subcloned into the CaSpeR transformation vector. A final step involved reinserting a 630 bp EcoRI fragment into this latter construct so as to include the gdl large intron, 3’-exon, and first polyadenylation signal. The resulting gdl(193)lacZ-Δ39 plasmid was identical to gdl(193)lacZ with the exception of the deletion of 12 bp between +118 and +130 that includes the initiating AUG of ORF39.

Germ-line transformation was performed essentially as described by Rubin and Spradling (1982). y w^67c23 embryos were injected with 300 pg ml∼¹ of the gdl-lac Z-CaSpeR transposons and 50μgml^-1 of the helper plasmid wings-clipped (Karess and Rubin, 1984) in 5 min KC1,0.1 mM NaPO₄. Because of the presence of a truncated white gene in the CaSpeR transformation vector, transformants were identified by w⁺ selection. At least two and as many as 16 independent transformant lines were obtained for each injected transposon. All lines were assayed for β-galactosidase activity in the gonads as described by Glaser et al. (1986). The quantitation of β-galactosidase activity in gonads of transformant flies was based on the time period required for the initial accumulation of the histochemical reaction product. Selected lines were further assayed by Northern and in situ hybridization analyses as previously described (Schulz and Butler, 1989; Schulz et al. 1989). Autoradiographs of Northern blots were quantitated by densitométrie measurement of the amount of signal in the gdl-lacZ fusion RNA band relative to the endogenous gdl^M transcripts in RNA from adult males or gdl^p transcripts in RNA from adult females. Fly stocks carrying the yw^67c23 chromosome and the chromosome 2 and 3 balancers CyO and TM3, Sb Ser, respectively, were used to determine the linkage of integrated transposons. The chromosomal markers and balancer chromosomes have been described by Lindsley and Grell (1968).

The gdl gene is a member of an overlapping gene cluster that maps to the 71CD region of chromosome 3 (Schulz and Butler, 1989; Schulz et al. 1989). The organization of the gene cluster and the sex-specific gdl transcripts are depicted in Fig. 1. At its 5’ end, gdl is overlapped by z600, a gene that is expressed dorsally during early embryogenesis (Schulz and Miksch, 1989). At its 3’ end, gdl is overlapped by Eip28/29, a gene that is ecdysone-inducible in Drosophila cell lines (Savakis et al. 1984; Bieber, 1986; Cherbas et al. 1986). All three genes are transcribed in the same direction and map within a 4.5 kb region.

In the male and female germ lines, gdl is differentially expressed due to alternative promoter usage (Schulz and Butler, 1989). Expression in the gdl^M mode leads to a 1200-/1500-nucleotide RNA pair in males, whereas expression in the gdl^p mode results in a 1000-/ 1300-nucleotide RNA pair in females. The 1200- and 1000-nucleotide RNAs and the 1500- and 1300-nucleotide RNAs are identical, except for the presence of an additional 165 to 215 nucleotides at the 5’ end of the gdl^M transcripts. This sequence difference affects the gene’s coding capacity. A common open reading frame (ORF) of 193 codons (ORF193) is present in all four transcripts; this sequence could encode for a polypeptide of ∼22.5K (x10³M_r). The additional sequences in the gdl^M transcripts results in an additional ORF of 39 codons (ORF39) that could yield a predicted polypeptide of ∼4.2K. Both initiator AUG codons reside in a sequence context - ORF39: GUUAt/GU; ORF193: AAGAUGG - that may be considered translationally favorable (Kozak, 1986a; Kozak, 1987a; Cavener, 1987).

To identify the in vitro protein products of the gdl gene, we synthesized gdl cRNAs followed by their translation in a reticulocyte lysate. Figure 2 shows the [³H]leucine-labeled translation products of full-length gdl^M and gdl^p cRNAs. Two polypeptides of ∼4.4 and 24K were derived from the gdl™ 1200 RNA (Fig. 2, lane 2), a result consistent with the translation of ORF39 and ORF193, respectively. The two polypeptides were labeled in a ratio of 5:1 which, based on the leucine composition of the two ORFs, indicated that ORF139 was translated with a two- to three-fold greater efficiency than ORF39. A single polypeptide of 24K was derived from the gdl^p RNA (Fig. 2, lane 3), a result consistent with the translation of ORF193 and the absence of ORF39. Overall, these studies revealed that the gdl^p and gdl^M transcripts can serve as monocistronic and bicistronic mRNAs in vitro.

Previous studies demonstrated that all sequences necessary for proper gdl sex-specific expression are contained within a 1.8 kb DNA region (Schulz and Butler, 1989). To further characterize the sequences required for gdl^Mandgdl^p transcription and translation in the germ line, we constructed a series of gdl-lacZ gene fusions that were cloned in the P element vector CaSpeR. These transposons were introduced into the Drosophila germ line, and transformants were assayed for sex-specific gdl-lacZ expression.

Five gdl-lacZ gene fusions are illustrated in Fig. 3. To generate gdl(193)lacZ, we started with the functional 1.8kb Bell fragment. This DNA contains a truncated gdl gene, with 328bp/492bp of 5’ flanking sequence relative to the gdl™/gdl^p transcription initiation sites, 1445 bp of genic sequence, and the first polyadenylation signal. A lacZ gene cassette, encoding the bacterial enzyme β-galactosidase, was inserted in frame into ORF193. lacZ is a useful expression marker because it facilitates the study of gdl-lacZ fusion gene transcription by Northern and in situ hybridization analyses and it allows the analysis of the spatial and temporal pattern of fusion gene transcription and translation by histochemical staining for -galactosidase activity. The proper expression of this construct should yield fusion RNAs of —4.3 kb in males and —4.1 kb in females. The gdl(193)lacZ-ΔP construct represents a 5’ deletion of gdl(193)lacZ; the deletion removes the gdl^M initiation site while maintaining the multiple gdr initiation sites. The expression of this construct should generate a fusion RNA of —4.1 kb. gdl(193)lacZ-AP/E is a 3’ deletion of gdl(193)lacZ-ΔP. In this construct, the large gdl intron, the 3’ terminal exon, and the polyadenylation signal have been removed. The expression of this DNA should result in a fusion RNA of greater than 3.7 kb, the size depending on the location of the nearest functional polyadenylation signal, which is predicted to reside within the downstream white gene of CaSpeR. The gdl(39)lacZ construct is a gene fusion that contains 600bp of gdl^M 5’ flanking sequence, 120bp of 5’ noncoding sequence, and a portion of ORF39 followed by an in frame lacZ gene. The DNA is oriented in the CaSpeR vector so that the 3’ end of the lacZ sequence is adjacent to a polyadenylation signal of the 3’ P element terminal repeat. Thus a fusion RNA of —3.3 kb would be predicted with the expression of this construct. The final construct shown is gdl(193)lacZ-A39, a gene fusion that is identical to gdl(193)lacZ except that it lacks a 12 bp region that includes the initiator AUG codon of ORF39. Thus the proper expression of this construct should yield monocistronic fusion RNAs of ∼4.3 kb in males and —4.1 kb in females.

Using Northern hybridization analysis, germ-line transformants harboring the different transposons were assayed for gdl-lacZ fusion gene expression. The sex-specific expression of gdl(193)lacZ in the 193Z-134 transformant line is demonstrated in Fig. 4A. RNA was isolated from male third instar larvae, adult males and adult females. To address tissue-specific expression in the gonads, RNAs isolated from dissected testes, ovaries, and the resulting carcasses were analyzed as well. A blot of these RNAs was hybridized with a gdl cRNA probe that detects the endogenous gdl^M and gdl^p transcripts and the gdl-lacZ fusion RNAs derived from the integrated transposon. Three transcripts of the predicted sizes were detected in male larvae (Fig. 4A, lane 3), adult males (Fig. 4A, lane 6), and adult females (Fig. 4A, lane 9), in the testes of larvae (Fig. 4A, lane 1) and adult males (Fig. 4A, lane 4), and in ovaries of adult females (Fig. 4A, lane 7). RNA derived from the resulting larval (Fig. 4A, lane 2), adult male (Fig. 4A, lane 5), and adult female (Fig. 4A, lane 8) carcasses failed to show any detectable transcripts. Thus the fusion gene was expressed with the proper tissue specificity in that transcripts were detected only in the gonads and not in other non-germ-line tissues. These results also indicated that the fusion gene is transcribed early in male germ-cell differentiation, since third instar larval testes contain solely premeiotic stem cells, spermatogonia and spermatocytes (Kemphues et al. 1982).

The results of the Northern analysis of RNAs from transformed flies harboring the other gdl-lacZ gene fusions are shown in Fig. 4B. In this experiment, we assayed for the differential expression of the transposons in adult males versus adult females. In the AP-193Z-32 transformant expressing the gdl(193)lacZ-AP transposon, the gdl cRNA probe detected the endogenous gene transcripts in male (Fig. 4B, lane 1) and female (Fig. 4B, lane 2) RNA. A gdl-lacZ fusion RNA of —4.1 kb was detected solely in the female RNA. The lack of expression in males was consistent with the deletion of the gdl^M 5’ flanking sequence and the gdl^M transcription initiation site. Expression in females indicated that gdl¹’ regulatory sequences are not deleted in this construct but reside internally within the gdl gene. A comparison of the expression of the gdl(193)lacZ and gdl(193)lacZ-ΔP transposons thus revealed that the gdl^M and gdl¹’ transcriptional control elements are separable.

An analysis of the ΔP/E-193Z-3 transformant expressing the gdl(193)lacZ-ΔP/E transposon allowed for the further delimitation of the gdl^F element. This construct differs from gdl(193)lacZ-ΔP by the deletion of the large gdl intron and terminal exon. The gdl-lacZ transposon is still expressed in a sex-specific manner, with a fusion RNA of —4.0 kb detected in female (Fig. 4B, lane 4) but not male (Fig. 4B, lane 3) RNA. Again, these results were consistent with the presence of a gdl^p regulatory element within the gdl gene.

An analysis of the 39Z-91 germ-line transformant indicated that the gdl(39)lacZ fusion gene is expressed sex-specifically in males. In this case, a lacZ cRNA probe was used to probe male (Fig. 4B, lane 5) and female (Fig. 4B, lane 6) RNA; a fusion RNA of —3.3 kb was detected solely in males. This result was consistent with the presence of gdl^M 5’ flanking sequence and the gdl™ transcription initiation site, and with the absence of the gdl^F transcription initiation sites and downstream gene sequences that possess gdl transcriptional control activity.

Finally, the analysis of the Δ39-139Z-11 transformant revealed that the gdl(193)lacZ-Δ39 transposon is expressed in both males and females, despite the 12 bp deletion that includes the initiator AUG of ORF39. Thus this small deletion of gdZ^M-specific sequences had no qualitative affect on transposon transcription; an apparent affect of the deletion on gdl-lacZ fusion RNA translation will be addressed below.

A summary of the different gdl-lacZ germ-line transformants is given in Table 1. The table lists the five different transposons, names of the multiple independent transformant lines, transposon chromosome linkage, and expression characteristics of the gdl-lacZ fusion genes. Results of the Northern analysis are presented as the ratio, in percentage, of the level of the gdl-lacZ fusion RNA relative to the level of the endogenous gdl RNAs in either adult males or females. The presence of the endogenous gdl transcripts in the transformant RNA samples serves as an internal control that allows for the comparison of the expression of the various transposons in the different transformant lines.

Four points can be made based on the Northern data. First, all six lines harboring the gdl(39)lacZ transposon express the fusion gene in males but not females while all of the lines harboring either the gdl(193)lacZ-ΔP or gdl(193)lacZ-ΔP/E express the fusion gene in females but not males. These results indicate that the gdl^M and gdl^F regulatory elements are separable with the gdl^M control sequences delimited to a 460 bp region between -328 and +132 relative to the gdl^M initiation site and the gdl^F control sequences delimited to a 655 bp region between -5 and +650 relative to the 5’-most gdl^F initiation site. Second, with the exception of the ΔP/E-193Z-3 transformant line, none of the transformants accumulated fusion RNAs to levels comparable to that of the endogenous gdl gene transcripts. Thus, while we have clearly included those sequences required for proper sex-specific expression of gdl, the various constructs may be missing an important quantitative gdl regulatory element. Alternatively, the gdl-lacZ fusion

RNAs may be less stable than endogenous gene transcripts leading to their lower relative accumulation in transformant gonads. Third, considerable variation is seen in the expression of a given transposon between different transformant lines. Presumably, this represents a positional effect on transposon expression based on the site of transposon integration. Fourth, considerable quantitative differences are observed in the expression of the different transposons. Again, this result may be due to the presence or absence of gdl quantitative elements in the constructs and/or the differential stability of the resulting gdl-lacZ fusion RNAs.

Upon establishing the gdl-lacZ transformant lines, we were interested in determining the translational characteristics of the fusion RNAs and identifying the germline cells that express the fusion genes. Specifically, we wanted to determine if both ORF39 and ORF193 were translated in the transformants. A positive result would implicate that the gdl^M transcripts served as bicistronic mRNAs in vivo. Prior to assaying the various transformant lines for β-galactosidase activity, we analyzed the transcription of the gdl(193)lacZ and gdl(39)lacZ gene fusions in the male germ line. We felt it was important to assess the spatial accumulation of the ORF193-/acZ and ORF39-/acZ fusion RNAs since the two transcripts differed significantly in their structures and thus could be subject to differential turnover in male germ cells. Using a lacZ probe, we determined the distribution of the fusion RNAs in adult testes of the 193Z-134 and 39Z-91 transformant lines by in situ hybridization. Figures 5A and 5C are bright-field photomicrographs of cross-sections through these tissues while Figs 5B and 5D are dark-field photomicrographs showing the accumulation of the fusion RNAs. A similar gdl-lacZ fusion RNA localization pattern was observed in the gdl(193)lacZ (Fig. 5B) and gdl(39)lacZ (Fig. 5D) tissue sections. In both cases, strong hybridization signals were observed in apical and lateral regions containing spermatocyte cysts. Thus we concluded that the spatial expression characteristics of the two fusion RNAs were comparable.

Gonads derived from the various transformant lines were then assayed for β-galactosidase activity so as to monitor the expression of the fusion genes in the germ line. The tissue-specific expression of the gdl-lacZ gene fusions is demonstrated in Fig. 6. Consistent with the results of the Northern analysis, β-galactosidase activity was detected in gdl(193)lacZ testes (Fig. 6A) and ovaries (Fig. 6B), gdl(193)lacZ-ΔP (Fig. 6D) and gdl(193)lacZ-ΔP/E (Fig. 6F) ovaries, gdl(39)lacZ testes (Fig. 6G), and gdl(193)lacZ-Δ39 testes (Fig. 61) and ovaries (Fig. 6J). The staining of the gdl(193)lacZ and gdl(39)lacZ testes thus indicated that both ORF39 and ORF193 are translationally active in the male germ line. In the positively staining transformant testes, β- galactosidase activity was detected throughout most of the tissue in regions that contain spermatocyte cysts and bundles of maturing sperm (Cooper, 1950). In the positively staining transformant ovaries, β-galactosidase activity was detected in developing egg chambers. Activity was not detected in gdl(193)lacZ-ΔP (Fig. 6C) and gdl(193)lacZ-ΔP/E (Fig. 6E) testes or gdl(39)lacZ ovaries (Fig. 6H).

A summary of the expression of the different transposons, as monitored by β-galactosidase activity in transformant gonads, is given in Table 1. We have indicated a descriptive quantitation of activity based on the initial time of accumulation of histochemical staining product in the transformant tissues (Glaser et al. 1986). The transformant lines expressing the gdl(39)lacZ and gdl(193)lacZ-Δ39 transposons showed rapid (within 30 min) staining in testes while transformant lines expressing the gdl(193)lacZ transposon showed a reproducibly delayed initial period of staining in testes (2-4 h or 4-8 h). This result was observed despite the lower levels of the gdl-lacZ fusion RNAs in gdl(39)lacZ and gdl(193)lacZ-Δ39 transformants as compared to RNA levels measured in the gdl(193)lacZ transformants. A point to note at this time is that the first two transposons generate monicistronic ORF39-ZacZ and ORF193-ZacZ transcripts while the latter transposon generates a bicistronic ORF39/ ORF193-lacZ transcript in the male germ line. Essentially no differences were observed in the quantitation of β-galactosidase activity in ovaries of transformant flies that were expressing the gdl-lacZ transposons.

To identify the germ cells expressing the gdl(193)lacZ fusion gene, we analyzed β-galactosidase activity in third instar larval testes and dissected adult ovaries of the 193Z-134 transformant line. As noted previously, third instar larval testes contain solely premeiotic stem cells, spermatogonia and spermatocytes (Kemphues et al. 1982). β-galactosidase activity was present in the spermatocyte cysts that occupy most of the testis but was absent from the apical region, a domain that may correspond to the stem and spermatogonia cells, and from some adhering somatic fat body cells (Fig. 7A). This analysis indicated that gdl is transcribed and translated early in the male germ line, with β-galactosidase present premeiotically in spermatocytes and post-meiotically in maturing sperm. In the female germ line, β-galactosidase activity was observed at high levels in stage 6 through 14 egg chambers (Fig. 7B). With longer periods of staining, enzymatic activity was detected in earlier stage egg chambers as well. Figure 7C demonstrates that the fusion genes are expressed in germ cells, but not somatic cells, of the female germ line. That is, β-galactosidase was observed solely in nurse cells and oocytes of the stage 10b and 13 egg chambers shown, whereas activity was not detected in the somatically derived follicle cells.

The gdl gene is differentially expressed in the male and female germ lines of Drosophila. An unusual gdl sequence organization provides for a complex regulation of gene expression at both the transcriptional and translational levels. Alternative promoters and separable transcriptional control elements are responsible for the sex-specific expression of the gene. The sequences responsible for male germ-line expression are located within a 460 bp region between -328 and +132 relative to the gdl™ initiation site, whereas female germ-line expression elements reside within a 655 bp region between -5 and +650 relative to the 5’-most gdl^F initiation site (Figs 3 and 4). A consequence of the alternative promoter utilization is the generation of gdl^F and gdl™ transcripts that possess either one or two potential coding regions, respectively. In this study, we demonstrated that these transcripts can serve as mono-cistronic and bicistronic mRNAs. The in vitro translation experiments revealed that a gdl^F transcript yields a 24K protein while a gdl™ transcript yields 4.4K and 24K proteins (Fig. 2). The analysis of germ-line transformants harboring either a gdl(ORF39)-lacZ or gJ/(ORF193)-lacZ fusion gene revealed that both initiating AUG codons are utilized in generating galactosidase activity in the male germ line (Figs 6 and 7). These results infer that the gdl™ transcripts serve as bicistronic mRNAs in vivo as well.

The bicistronic nature of the gdl™ transcripts is not unique; examples do exist of mRNAs (mostly viral) that initiate translation at a site or sites in addition to the AUG codon nearest the 5’ end. One mechanism involves initiation at a downstream AUG as a result of ‘leaky scanning’ (Kozak, 1989), whereby some ribosomes migrate past a weak initiator codon until they find an AUG in a translationally favorable context. Several viral mRNAs have been identified that produce two overlapping proteins that are initiated at the first and second AUG codons (reviewed in Kozak, 1986b). Downstream initiation may also occur even when the first AUG is in a strong sequence context (reviewed in Kozak, 1989). In this case, the strong initiator codon is usually followed by an in-frame terminator codon which results in the presence of a small ORF in the mRNA. It is postulated that after the translation of the minicistron, the 40S subunit remains associated with the mRNA and continues its migration along the message (Peabody and Berg, 1986). When the next acceptable AUG is reached, the 40S and 60S ribosomal subunits reassociate and reinitiate translation of the mRNA at the second initiation site. Such reinitiation is usually inefficient in that the presence of the upstream ORF usually decreases the translational efficiency of the downstream ORF (reviewed in Kozak, 1989). A third possibility exists in which a ribosome would translate only one ORF from a bicistronic mRNA with some transcripts translated beginning with the first AUG, generating the first ORF gene product, and other transcripts translated beginning with the second AUG, generating a downstream ORF gene product. Such internal initiation would constitute a translational event inconsistent with the current scanning model for translation (Kozak, 1989).

Our experimental results do not allow us to conclude which mechanism is used in the translation of the bicistronicgd/^M mRNAs. However, based on transcript structure considerations, the synthesis of the 4.4K and 24K gdl proteins is potentially consistent with translation of the gdl^M mRNA by the termination-reinitiation mechanism. ORF39 constitutes a minicistron possessing an AUG initiator codon in a good sequence context. ORF193, which has an AUG initiator codon in a stronger sequence context, is located only 68 nt downstream of ORF39, generating structural features that should be favorable for reinitiation of translation (Kozak, 1987b). Additionally, deletion of the minicistron from a gdl^M transcript, as is the case in the expression of the gdl(193)lacZ-Δ39 transposon, results in a consistently higher level of β-galactosidase activity in gdl(193)lacZ-Δ39 transformant testes as compared to that observed in gdl(193)lacZ transformant testes (Table 1). That is, gdl(193)lacZ-Δ39 testes stain much more rapidly than do gdl(193)lacZ testes despite the higher levels of gdl-lacZ fusion RNA in the latter transformants. Since the two gdl-lacZ fusion RNAs differ solely by the presence or absence of a 12 bp region that includes the initiator AUG of ORF39, we tentatively conclude that the monocistronic ORF193-lacZ transcript is translated more efficiently in the gdl(193)lacZ-A39 transformants than the bicistronic ORF39/ORF193-lacZ transcript is in the gdl(193)lacZ transformants.

For gdl and other viral and cellular polycistronic mRNAs, the biological significance of the multiple ORFs remains an important question. One must consider the potential of translational control being a means of regulating the expression of the specific gene in vivo. For gdl, it is possible that the presence of ORF39 decreases the translational efficiency of ORF193 in the male germ line, resulting in a lower level of the gene product than that found in the female germ line. Hypothetically, this could serve as a mechanism for producing a sex-specific function of the gdl ORF193 gene product.

Analysis of transformants harboring gdl-lacZ fusion genes allows us to address other aspects of gdl gene expression in the germ line. First, the gdl^F-lacZ fusion RNAs are translated during oogenesis (Figs 6 and 7). The possibility existed that the fusion transcripts might be stored during egg chamber development, maternally contributed to the egg, and then translated during embryogenesis. This raises the implication that gdr transcripts are translated in egg chambers and that the gdl^v gene product may serve some function in oogenesis. Second, β-galactosidase activity is detected in both premeiotic and postmeiotic male germ cells, indicating that the gene is expressed early during male germ-line differentiation. Again, the possibility existed that the fusion RNAs might be stored during the early stages of spermatogenesis, with translation occurring later during sperm maturation. Such is the case for the male germ line-specific mst(3)gl-9 gene that is transcribed premeiotically but translated postmeiotically, three days later in sperm development (Kuhn et al. 1988). Third, since both ORF39 and ORF193 were translationally active in the germ-line transformation experiments, one or two gdl gene products may function in the early male germ line. Clearly, the generation and utilization of antibodies directed against both the ORF39 and ORF193 gene products should give us a better comprehension of the complex post-transcriptional events in the expression of the gdl gene.

Our studies oí gdl demonstrated it to be an intriguing gene because of its unusual structural organization, restricted male and female germ line expression, and complex transcriptional and translational regulation. As commented previously, the further characterization of this gene should yield important information on germ-cell differentiation. Additional transformation experiments should allow us to map more precisely the separable germ-line control elements. The delimitation of the gdl”‘ and gdl^F control sequences will enable us to screen for and characterize genes encoding rrans-acting factors that regulate gdl gene expression. It seems plausible that these genes might represent previously or newly identified components of the germ-line sex determination hierarchy. Furthermore, we now possess transformant lines that express the gdl-lacZ fusion genes sex-specifically in either the male or female germ lines. These transposons may be valuable probes for assessing the status of the germ line in the various mutants affecting germ-cell differentiation. An analysis of gdl^M-lacZ transposon expression in Sxl∼, snf∼, or otu∼ XX germ cells might, for example, yield a sensitive indication of cells entering the spermatogenic pathway as a result of germ-cell transformation. Additionally, the possibility exists that an identical gdl gene product functions in both the male and female germ lines. Characterization of this protein might yield insights into processes that are common to both male and female germ-cell differentiation. Alternatively, gdl expression in the male germ line might be subject to translational control because of the bicistronic gdl^M mRNAs. Such an occurrence could represent yet another means, in addition to the alternative splicing seen with genes controlling somatic sexual differentiation, of generating a sex-specific gene function during Drosophila sex determination.

We are grateful to David Pritchard and Julia Hsi of our lab for constructing plasmid gdl(193)lacZ-CaSpeR and its derivatives and to Teresa Joseph of The University of Texas M. D. Anderson Cancer Center Histology Core Facility for her preparation of the testes sections. S.G. was supported by a National Institutes of Health postdoctoral training grant (HD 07325). This research was supported by a grant to R.A.S. from the National Science Foundation (DCB-8709846).