RESULTS

pum is required for primordial germ cell development before oogenesis: The pum mutations cause failure of germline stem cell maintenance during oogenesis (Lin and Spradling 1997; Forbes and Lehmann 1998). To investigate whether pum is also required for germline development before oogenesis, we examined the phenotypes of the ovarette class of pum mutants (pum^ovt) for potential defects in primordial germ cell development. Unlike “classical” maternal efffect pum mutants that can undergo oogenesis to produce embryos defective in posterior patterning, pum^ovt mutants show severe oogenic defects and fail to produce any eggs (Lin and Spradling 1997). Third instar larval ovaries from homozygous pum^ovt mutant and wild-type larvae were stained with anti-VASA antibodies to specifically label germline cells (Hayet al. 1990; see materials and methods). Because pregermline stem cells in the third instar larval ovaries reflect the final stage of primordial germ cell proliferation and migration (see Introduction), any abnormality in number or morphology reflects a defect in primordial germ cell proliferation and development. Wild-type third instar larval ovaries typically contain 54.6 ± 7.4 germ cells located in the medial region of the ovary (Table 1 and Figure 1A). However, ovaries from pum^ovt mutants contain either significantly reduced or increased numbers of germ cells (data from four representative mutants are shown in Table 1). The requirement of pum during primordial germ cell development is further confirmed by dramatic overproliferation of primordial germ cells seen in pum⁴²⁷⁷ and pum¹⁶⁸⁸ mutant ovaries. These observations suggest that pum function is required for the normal proliferation of primordial germ cells before oogenesis.

In addition to abnormal numbers of germ cells, the mutant germ cells also exhibit various morphological defects (Figure 1, B and C). The abnormal size and morphology of pregermline stem cells suggest that the development of primordial germ cells is highly aberrant in pum mutants even though certain germline characteristics, such as VASA expression, are still maintained (Figure 1, B and C). The germline defects are often accompanied by a drastic increase in the size of the ovary, with an increased number of somatic cells (Figure 1, compare B and C to A). These results suggest that pum is also required for the proper formation of the larval ovary.

pum^ovt mutants are also defective in germline stem cell maintenance, cyst formation and oocyte differentiation during oogenesis: The proliferation defects of the primordial germ cells and the abnormal morphology of the resulting pregermline stem cells in the pum^ovt larval ovary suggest that these cells may not be able to function normally as germline stem cells during subseqeunt oogenesis. To test this hypothesis, we examined whether germline defects at the larval stage correlate to subsequent ovarian defects in pupal and adult ovaries.

We first examined the viability of pum^ovt mutants at the pupal stage. The viability difference between the pum^ovt mutants and their heterozygous siblings is within 25% (Figure 1N). This rules out any potentially significant skew in the observed pupal or adult defects caused by selective lethality against pupae with a particular type of oogenic defect.

We then examined the pum⁶⁸⁹⁷ mutant because a high proportion (77%) of pum⁶⁸⁹⁷ larval ovaries contain either underproliferated (54%) or overproliferated (23%) primordial germ cells. Ovaries were isolated from homozygous pum⁶⁸⁹⁷ females at ∼48 hr after pupation. They were double stained with anti-VASA antisera to label germline cells and anti-1B1 antibody to outline somatic cells and label spectrosomes and fusomes, two special structures in the early germline cells (Zaccai and Lipshitz 1996; Deng and Lin 1997). By this stage, pupal ovaries from wild-type siblings have differentiated into an average of 16.7 ± 4.0 ovarioles per ovary (Figure 1D). In mutant ovaries, 1B1 staining of somatic cells shows that a similar number of germaria have also formed (Figure 1E). In contrast to the control ovarioles that always contain a full complement of germline cells, however, 63% of the mutant ovarioles are germlineless (Figure 1F). The remaining mutant ovarioles contain a small number of germline cells. By the adult stage, 86% of the pum⁶⁸⁹⁷ ovaries lack germline (Table 2). These observations indicate that the pum⁶⁸⁹⁷ mutant is severely defective in oogenesis but not in ovariole formation.

Parallel analyses on six other pum^ovt mutants revealed similar oogenic defects, with the proportion of germlineless ovaries varying among the mutants (Table 2). The remaining ovaries contain either developing egg chambers (Figure 1H) or undifferentiated germ cell clusters (Figure 1I), as described previously for the pum²⁰⁰³ mutation (Lin and Spradling 1997). This confirms the previous conclusion that the self-renewing asymmetric division of germline stem cells is disrupted in the pum²⁰⁰³ mutant.

pum^ovt mutations also affect subsequent oogenic events. Developing mutant egg chambers sometimes contain few nurse cells but no oocytes (Figure 1, J–L). This defect also exists in the maternal effect lethal class of pum mutants. Staining of these mutant egg chambers with rhodamine-conjugated phalloidin reveals a reduced number of ring canals (Figure 1J), indicating that the reduced number of nurse cells results from reduced divisions of cystoblasts. Furthermore, these mutant egg chambers show a pronounced unequal ploidy among these nurse cells (Figure 1L), indicating that the endoreplication mechanism in these egg chambers has been affected severely. These defects suggest that pum is required for the proper division of cystoblasts and differentiation of germline cysts.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1.

—Phenotype of pum^ovt mutants. (A) Wild-type third instar larval ovary stained with an anti-VASA antibody. Approximately 55 pregermline stem cells are located in the medial region of the ovary. (B) pum⁶⁸⁹⁷ mutant third instar larval ovary stained for VASA. Germline cells are overproliferated in 23% of the mutant ovaries (see Table 1). However, 54% of the mutant ovaries contain only 19.6 ± 4.4 germ cells. Moreover, these germ cells are grossly abnormal in morphology and size. In addition, the ovary is much larger than the wild-type ovary. (C) pum³²⁰³ mutant third instar larval ovary stained for VASA showing underproliferated and defective germline cells. A total of 55% of the pum³²⁰³ ovaries contain only 5.7 ± 2.4 germ cells. Specifically, 30% of pum³²⁰³ ovaries completely lacked germline cells. (D) A wild-type midpupal ovary stained for VASA; 18 germaria are fully developed and contain a full complement of germ cells. (E and F) A pum⁶⁸⁹⁷ midpupal ovary stained with anti-1B1 and anti-VASA antibodies, respectively. (E) Although individual germaria have formed, (F) many of them do not contain germline cells, as marked by arrows. (G) Wild-type adult ovarioles stained with DAPI showing germaria (g) that have produced a string of developing egg chambers. (H) A VASA-stained pum¹⁶⁸⁸ ovariole containing only two mature egg chambers and a rudimentary germarium (g). (I) A VASA-stained pum⁴²⁷⁷ ovariole containing three clusters of undifferentiated germ cells, presumably because of the loss of asymmetry during the division of three germline stem cells. (J) A pum¹⁶⁸⁸ egg chamber containing three nurse cells double stained with DAPI and rhodamine-conjugated phalloidin to mark ring canals (arrow). (K) A pum⁴²⁷⁷ egg chamber stained with DAPI showing a single nurse cell flanked by extensive and overproliferated interfollicular stalk cells (arrow). (L) A pum⁴²⁷⁷ egg chamber stained with DAPI to visualize the number and relative DNA content of the nurse cells. (M) DAPI-stained pum⁴²⁷⁷ terminal filament containing misplaced somatic cells within the single cell stack (arrow). (N) The percentage of homozygous and heterozygous pum^ovt larvae that survived the pupal stage to eclosion.

In addition to germline defects, somatic defects were also detected at a low frequency in pum^ovt mutants, such as long interfollicular stalks (Figure 1K) and the disruption of single-stack cells in the terminal filament (Figure 1M). Together, these data reveal that multiple germline and somatic processes of oogenesis are disrupted by pum^ovt mutations.

pum is directly involved in germline stem cell division during oogenesis: The severe defects observed in pum mutant priomordial germ cells suggest that these cells are unlikely to undergo normal oogenesis, which precludes the opportunity to analyze whether pum also plays a direct role in germline stem cell division and other oogenic processes. To overcome this problem, we let germ cells develop normally before oogenesis and then removed the pum activity in germline stem cells at the onset of oogenesis by inducing homozygous pum^– germline clones using the FLP-DFS technique (Chou and Perrimon 1992; see materials and methods). Forbes and Lehmann (1998) have shown by pole cell transplantation that pum functions cell autonomously in the germline. Thus, by removing pum from germ cells at the onset of oogenesis, we were able to examine directly whether pum is required during oogenesis.

We induced the homozygous pum mutant germline in pum²⁰⁰³, pum¹⁶⁸⁸, and a maternal effect mutation pum^ET1, because they are strong mutations representing three types of molecular lesions (see below). Homozygous pum^– germline clones were generated in the ovo^D1 background (see materials and methods). ovo^D1 dominantly blocks oogenesis in a cell-autonomous manner, arresting egg chambers uniformly at stage 3 (Figure 2; for staging see King 1970). This defect is distinctively different from either the germlineless or differentiating phenotype of pum^ovt (see above) or the wild-type ovariole. The developmental fate of homozygous pum^– germline stem cell clones depends on whether pum is required during oogenesis. If pum is required during oogenesis, pum^– clones should show corresponding pum^– oogenic defects. Alternatively, if pum is not required during oogenesis, the pum^– clones should undergo oogenesis normally.

For all three pum alleles tested, pum^– germline clones exhibited typical pum^– defects. For example, many pum²⁰⁰³ germline clones produced ovarioles that contain only one to three mature egg chambers but no other germ cells (Figure 2A). These typical differentiated pum²⁰⁰³ ovarioles are distinctively different from the ovo^D1 ovarioles (Figure 2A). The differentiated ovarioles are also seen in pum¹⁶⁸⁸ and pum^ET1 germline clones (Figure 2, B and C). Sometimes, differentiated ovarioles contain four to five egg chambers (Figure 2D), indicating that these pum^– germline stem cells have divided once before entering oogenesis.

A quantitative summary of the clonal analysis is presented in Table 3. In addition, pum²⁰⁰³, pum¹⁶⁸⁸, and pum^ET1 clonal ovarioles contained an average of 4.2 ± 2.0 (n = 17), 2.8 ± 1.7 (n = 21) and 2.2 ± 1.5 (n = 14) egg chambers, respectively. Other defects, such as germlineless germaria (see Figure 1, E and F) and ovarioles with undifferentiated germ cell clusters (see Figure 1I), though more difficult to quantify, also exist in the clone-induced ovaries. These defects indicate that pum activity is directly required in the germline stem cells during oogenesis.

pum appears to be required in somatic cells for oviposition: A striking difference between the pum germline clonal females and their corresponding nonclonal homozygous females is that the eggs produced by the clonal females are often laid, yet eggs produced by their nonclonal counterparts are never laid. This suggests that pum may be required in somatic cells for oviposition. To further test this possibility, we let the germline clonal females lay eggs for 9 days and then dissected the females to count the number of eggs that were still held in the ovary to measure the efficiency of oviposition. Among three alleles tested, pum²⁰⁰³, pum¹⁶⁸⁸, and pum^ET1 clonal females achieved oviposition efficiency of 56, 85, and 78%, respectively (for absolute number of eggs laid by the females, see Table 3). This egg-laying ability is never observed in homozygous pum²⁰⁰³ or pum¹⁶⁸⁸ females. The restoration of the egg-laying ability in females whose soma is no longer deficient in pum suggests that the pum gene is required in somatic cells for oviposition. This function is affected by pum²⁰⁰³ and pum¹⁶⁸⁸ mutations as well as by the maternal effect class of mutations such as pum^ET1.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2.

—Oogenic defects of pum germline clones induced at the third instar larval stage. All the ovaries were dissected at 9–12 days after eclosion and were stained with anti-1B1 antibodies to outline ovarian cells. (A) A pair of pum²⁰⁰³ ovaries. The right ovary contains only arrested ovo^D1 ovarioles (ovo^D), while the left ovary contains multiple homozygous pum²⁰⁰³ clonal ovarioles (three ovarioles are shown and labeled as pum²⁰⁰³). Each pum²⁰⁰³ clonal ovariole contains two to three mature eggs, but no other germ cells. This is in sharp contrast to ovo^D1 ovarioles, which contain numerous egg chambers arrested at stage 3. (B) Several pum¹⁶⁸⁸ clonal ovarioles. One ovariole contains a single mature egg and a rudimentary germarium. (C) A pum^ET1 clonal ovariole (pum^ET1) embedded in a pair of ovo^D1 ovaries. (D) A pum¹⁶⁸⁸ clonal ovary in a 12-day-old female containing five egg chambers, suggesting some residual germline stem cell activity in this ovariole.

pum²⁰⁰³ and pum¹⁶⁸⁸ mutations do not abolish the maternal effect function of pum in embryos: A more surprising observation is that, despite pum²⁰⁰³ or pum¹⁶⁸⁸ mutations causing both germline and somatic defects in oogenesis and oviposition (see above), eggs laid by females containing the homozygous pum²⁰⁰³ or pum¹⁶⁸⁸ germline were sometimes capable of developing into adulthood (Table 3). This is in contrast to eggs laid by females containing the homozygous pum^ET1 germline, which showed typical posterior patterning defects and failed to hatch, as reported previously for maternal effect alleles (Lehmann and Nüsslein-Volhard 1987). The ability of the pum²⁰⁰³ or pum¹⁶⁸⁸ eggs to develop further suggests that pum²⁰⁰³ or pum¹⁶⁸⁸ mutations do not abolish the maternally provided pum activity required for embryogenesis. The pum^– embryos were produced by mating virgin mutant pum²⁰⁰³ or pum¹⁶⁸⁸ clonal females to homozygous pum²⁰⁰³ or pum¹⁶⁸⁸ males, respectively. This rules out any possible paternal contribution to the embryonic pum function.

pum¹⁶⁸⁸ partially complements other pum^ovt and maternal effect pum alleles: Both pum²⁰⁰³ and pum¹⁶⁸⁸ mutations lead to more defects than maternal effect mutations during oogenesis, yet they do not completely affect embryonic development. Conversely, maternal effect mutations have fewer pleiotropic effects during oogenesis, yet they completely block embryogenesis. These differences could suggest that maternal effect mutations are stronger mutations. Alternatively, it is possible that pum is a complex locus encoding several discrete and complementable functions. To test these possibilities, we conducted inter se genetic complementation tests among pum^ovt and maternal effect alleles of pum. Previous complementation analyses had shown that pum²⁰⁰³ and pum³²⁰³ failed to complement several of the maternal effect lethal pum alleles, yet pum¹⁶⁸⁸ partially complements the maternal effect alleles (Lin and Spradling 1997). We further quantified the degree of complementation (Figure 3I). The results reveal that pum¹⁶⁸⁸ partially complements the embryonic lethality of maternal effect alleles pum^ET1, pum^MSC, and pum^ET7; of these, pum^MSC and pum^ET7 are known to disrupt the RNA-binding domain at the C terminus of the PUM protein (Forbes and Lehmann 1998). This complementation leads to the production of viable yet sterile progeny. In addition, pum¹⁶⁸⁸ also partially complements pum²⁰⁰³ and pum³²⁰³ of the pum^ovt class for oogenic defects and embryonic lethality (Figure 3, E and I). However, trans-heterozygotes between pum¹⁶⁸⁸ and all maternal effect alleles tested, despite partial complementation for lethality, showed oogenic defects (Figure 3, B–D) similar to those of the pum^ovt (Figure 3E) or maternal effect class mutants (Figure 3, G and H), but not to those of the wild-type ovary (Figure 3A). For example, ovaries from the trans-heterozygous females are often depleted of germline (Figure 3B, compare to Figures 1, E and F, and 3H) or contain differentiated ovarioles with only a few mature eggs (Figure 3C, compare to Figures 1H and 3G). These data indicate that germline stem cell division defects fail to be complemented in these heteroallelic combinations, even though the oviposition defect of pum¹⁶⁸⁸ has been complemented by the maternal effect alleles and the embryonic lethality of the maternal effect alleles has been complemented by the pum¹⁶⁸⁸ allele.

pum^ovt mutations map within the pum transcription unit: To study the molecular nature of the pum^ovt mutations, we isolated genomic DNA sequences encompassing the PZ-element insertion sites in the pum^ovt mutants (see materials and methods). Molecular mapping of these genomic clones with respect to the cosmid clones spanning the pum locus (Barkeret al. 1992; MacDonald 1992) revealed that all the pum^ovt PZ insertions are located in introns of the pum gene (Figure 4). Among them, the pum¹⁶⁸⁸ insertion is located between exons 3 and 4 while other pum^ovt insertions are clustered within a 505-bp region at the center of a 120-kb intron between exons 8 and 9. All the PZ insertions are oriented in the same direction so that the rosy gene in the PZ element is toward the 5′ side of the pum gene while the lacZ gene is oriented toward the 3′ side. This orientation of PZ insertions in introns has been shown by Horowitz and Berg (1995) to cause aberrant splicing and transcription termination because of the presence of a splice acceptor site in the l(3)s12 sequence near the rosy gene (Figure 4). Consistent with this, we identified a new P-insertional allele, ep(3)1196, carrying a P element lacking the l(3)s12 sequence (Rorth 1996). Even though ep(3)1196 is inserted in the same orientation and location, 185 base pairs 5′ to pum⁴²⁷⁷, it does not cause any detectable lethality or sterility.

pum²⁰⁰³ and pum¹⁶⁸⁸ eliminate two different PUM proteins, each of which is sufficient for maternal effect function, yet both of which are required for oogenesis: The partial complementation between different P alleles suggests that they may differentially affect the pum gene product. Although two splicing variants of the pum transcript have been characterized (Barkeret al. 1992), they differ only in the 5′-untranslated region but share the same predicted ORF (Figure 4). However, previous analysis of the PUM protein from embryonic extracts detected two bands positioned at ∼156 kD among several bands of smaller molecular weight (MacDonald 1992; Murata and Wharton 1995). It was known only that the 156-kD doublet represents the PUM protein. Because the pum^ovt P insertions were located within the pum transcription unit, we performed Western blot analysis to examine how these mutations affect the expression of different bands that are recognized by anti-PUM antibodies. Because pum¹⁶⁸⁸ complements other pum^ovt alleles and potentially causes aberrant splicing at a different exon than other pum^ovt alleles, we tested how pum¹⁶⁸⁸ and other pum^ovt alleles affect the expression of PUM isoforms, using pum²⁰⁰³ and pum⁴²⁷⁷ as representatives of other pum^ovt alleles.

Western blots of Drosophila ovarian extracts prepared from flies homozygous for pum¹⁶⁸⁸, pum²⁰⁰³, and pum⁴²⁷⁷ mutations were probed with four different antibodies made against the PUM protein (MacDonald 1992; Forbes and Lehmann 1998; this study, see materials and methods). Protein prepared from pum^MSC/pum^ET9 and pum^MSC/pum^ET1 trans-heterozygotes, all of which have been proposed to be null mutations (Forbes and Lehmann 1998), were used as negative controls. All antisera tested showed two major bands, the 156- and the 130-kD bands, in wild-type flies (Figure 5). The 156-kD band can be resolved as a doublet upon shorter exposure. This concurs with previous Western blotting analyses (MacDonald 1992). Interestingly, the pum¹⁶⁸⁸ mutation completely eliminates the 156-kD doublet but does not affect the 130-kD band. Conversely, pum²⁰⁰³ and pum⁴²⁷⁷ mutations diminish the 130-kD band but do not appear to affect the 156-kD doublet (Figure 5). These results suggest that both the 156-kD doublet and the 130-kD band are functionally important protein isoforms of the pum gene. Affecting either one of the isoforms leads to defective pum function during preoogenic germline development and oogenesis, but one of the isoforms is sufficient for embryogenesis (see above). The heteroallelic pum^MSC/pum^ET9 and pum^MSC/pum^ET1 combinations eliminated both the 156-kD doublet and the 130-kD isoform (for pum^MSC/pum^ET9, see Figure 5C), suggesting that these three maternal effect mutations are null mutations.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3.

—Phenotypes of pum heteroallelic combinations. (A) Anti-VASA stained pum¹⁶⁸⁸/TM3 ovary shows wild-type morphology. (B) A pair of DAPI-stained pum¹⁶⁸⁸/pum^ET7 ovaries display a typical germlineless morphology (compare to Figure 1, E and F). (C) A pair of anti-1B1-stained pum¹⁶⁸⁸/pum^ET1 ovaries exhibits typical differentiated phenotype, with the few egg chambers in each ovariole developed to maturity (compare to Figure 1H). (D) DAPI-stained pum¹⁶⁸⁸/ pum^MSC ovary also showing differentiated phenotype. (E) DAPI-stained pum¹⁶⁸⁸/pum³²⁰³ ovary displaying a mildly differentiated phenotype. (F) A pair of DAPI-stained pum³²⁰³/pum^MSC ovaries containing both germlineless and differentiated ovarioles. (G) Anti-VASA stained pair of differentiated pum^MSC/pum^ET9 ovaries. (H) DAPI-Stained pair of germlineless pum^MSC/ pum^ET9 ovaries. (I) Histograph showing the number of progeny from heteroallelic combinations normalized to wild-type controls.

• Download figure
• Open in new tab
• Download powerpoint

Figure 4.

—Location of the pum^ovt mutations in the pum locus. The map of pum is adapted from Barker et al. (1992), with transcription proceeding from right to left corresponding to a distal-to-proximal orientation on chromosome 3. Exons are numbered serially from 5′ to 3′. Open-boxed exons denote the 5′-untranslated region in the mRNA, while the black exons encode the ORF. Each of the P-insertion sites is marked by a triangular flag, with the direction of the flag pointing to the rosy (ry) side of the P element. The P element in all the alleles, except for that in ep(3)1196, is the PZ element, whose structure is shown in the lower part of the figure at a different scale. A left-pointing arrow in the l(3)S12 region (shaded area) of the PZ element indicates the splice acceptor site. Note that the pum¹⁶⁸⁸ P insertion is located in intron sequence upstream of exon 4, ∼60 kb distal to the other pum^ovt insertions, which are clustered within a 505-bp region at the center of a 120-kb intron between exons 8 and 9. Among them, the 3′-most alleles, pum⁶⁸⁹⁷ and pum⁷⁰⁹⁸, are 387 bp from pum²⁰⁰³ and have identical insertion sites despite being independently isolated lines. The pum²⁰⁰³ and pum⁴⁸⁰⁶ insertions are separated by a single base pair. pum³²⁰³ and pum⁴²⁷⁷ are located at more 5′ sites, 109 and 118 bp from pum²⁰⁰³ and pum⁴⁸⁰⁶, respectively. The pum RNA probe used for in situ hybridization is derived from exons 4–6 and the beginning of exon 7. The 15-amino-acid peptide used for generating the Ovtpep2 antibody corresponds to a sequence in exon 4.

A novel band of ∼88 kD appears in the homozygous pum²⁰⁰³ and pum⁴²⁷⁷ mutants, but not in pum¹⁶⁸⁸, in maternal effect pum mutants, or in the wild-type flies (see Figure 5; not shown are pum⁴²⁷⁷ data, which are identical to those of pum²⁰⁰³). Thus, this band is likely to be the product of aberrant splicing specific to pum²⁰⁰³, pum⁴²⁷⁷, and other pum^ovt mutations residing in the 120-kb intron. This 88-kD band was detected by antibodies against the C-terminal region of the PUM protein containing the RNA-binding domain downstream of all the pum^ovt insertion sites, suggesting that the aberrant splicing does not cause premature termination of protein synthesis at the pum^ovt insertion sites, as reported previously (Horowitz and Berg 1995). Instead, it suggests a novel splicing activity of the PZ-element insertions that awaits further investigation.

• Download figure
• Open in new tab
• Download powerpoint

Figure 5.

—Different pum^ovt mutations eliminate different PUM protein isoforms. Shown here is a Western blot probed with PUM2#1 antibody (see materials and methods). The 156- and 130-kD isoforms are both present in (A) wild-type adult flies and (B) 0–2 hr embryos and ovaries as well as in the (A) pum²⁰⁰³/TM3 flies. However, the 156-kD isoform is eliminated in pum¹⁶⁸⁸/pum¹⁶⁸⁸ flies, while the 130-kD isoform is greatly reduced in pum²⁰⁰³/pum²⁰⁰³ flies. The same result is seen for the pum⁴²⁷⁷/pum⁴²⁷⁷ flies (data not shown). (B) In pum^MSC/pum^ET9 flies, both isoforms are eliminated. A low, abundant band of 98 kD is also affected in pum¹⁶⁸⁸ and pum²⁰⁰³ mutants. (C) The Western blot from B probed with the Drosophila nonmuscle myosin heavy chain (the zipper gene product, ZIP) as a loading control.

In addition to the 156-kD doublet and the 130-kD isoform, several other weak bands of lower molecular weights are also detected by the anti-PUM antibodies in wild-type flies. Among them, a weak band of ∼98 kD is reduced in both pum¹⁶⁸⁸ and pum²⁰⁰³ mutants (Figure 5), suggesting its possible involvement in pum function. Other bands, however, are not altered or eliminated in the pum mutants and, thus, are unlikely to be pum products.

A pum cDNA transgene containing the known ORF encodes only the known 156-kD PUM protein and rescues the pum^ovt preoogenic and oogenic defects: The 156and 130-kD isoforms of PUM could both derive from the known ORF of pum by posttranslational processing (Figure 4). Alternatively, one of them could be encoded by a novel species of alternatively spliced pum mRNA that is yet to be identified. To discriminate between these possibilities, we introduced into pum¹⁶⁸⁸ and pum²⁰⁰³ mutants a P[nos-pum] transgene that contains a pum cDNA encoding the known ORF driven by the nanos promoter (Barkeret al. 1992). These mutant backgrounds allowed us to determine which PUM isoform this cDNA encodes and which pum mutant defect it can rescue (see materials and methods). This test thus also served the purpose of allowing us to examine which PUM isoform is responsible for a particular developmental process. Western analysis revealed that the missing 156-kD isoform in the homozygous pum¹⁶⁸⁸ mutant is clearly replenished by the P[nos-pum] transgene, while the missing 130-kD isoform in the homozygous pum²⁰⁰³ mutant is not restored (Figure 6A). These results indicate that the known pum ORF encodes only the 156kD isoform.

Although the P[nos-pum] transgene encodes only the 156-kD isoform, it rescues the preoogenic and oogenic defects of both pum²⁰⁰³ and pum¹⁶⁸⁸ mutants. In both homozygous pum²⁰⁰³ and pum¹⁶⁸⁸ mutants carrying the transgene, most ovarioles contain actively dividing germline stem cells that support normal oogenesis, even in 8-day-old mutant females, as is evident by the presence of a progression of wild-type egg chambers in most of their ovarioles (Figure 6, B–D). This indicates the complete rescue of germline and oogenic defects of both pum²⁰⁰³ and pum¹⁶⁸⁸ mutants by the P[nos-pum] transgene. Thus, even though the lack of either the 156- or 130-kD isoform leads to severe defects during zygotic germline development and oogenesis, increasing the expression of the 156-kD isoform alone can compensate for the lack of the 130-kD isoform and rescue the germline and oogenic defects.

In addition to the complete oogenic rescue, 8 out of 15 transgene-carrying pum²⁰⁰³ females produced progeny; 3 out of 11 transgene-carrying pum¹⁶⁸⁸ females also produced progeny (Figure 6E). These observations further support the conclusion that the 156- and the 130-kD PUM isoforms are not functionally distinct. Either isoform is sufficient for the maternal effect function of pum (see above).

Despite the rescue of the female sterility, P[nos-pum] does not rescue the semilethality of pum²⁰⁰³. The viability of homozygous P[nos-pum]; pum²⁰⁰³/pum²⁰⁰³ flies is 14% of that of P[nos-pum]; pum²⁰⁰³/+ flies, which is similar to the viability of pum²⁰⁰³/pum²⁰⁰³ flies (Lin and Spradling 1997). Thus, as expected, the somatic function of pum in supporting viability is not rescued by the germline-specific expression of the P[nos-pum] transgene.

pum is expressed in the soma and germline in the ovaries: Phenotypic analysis reveals a somatic function for pum during oogenesis, yet previous RNA and protein in situ analyses showed only the germline expression of pum (MacDonald 1992; Forbes and Lehmann 1998). To detect the possible somatic expression of pum, we first examined the enhancer trap staining pattern of most pum^ovt mutations (see materials and methods), which revealed that they are specifically expressed in the terminal filament cells (Figure 7A), as was reported previously for pum²⁰⁰³ (Lin and Spradling 1997). To test whether this reflects part of the pum expression pattern, we conducted RNA in situ hybridization to whole-mount wild-type ovaries, using conditions that allow the detection of RNA in both surface somatic cells and inner germline cells (Coxet al. 1998; also see materials and methods). Under conditions that favor the detection in surface cells, pum RNA is easily detectable in the terminal filament cells and epithelial sheath cells in the interfollicular stack region but is barely detectable in the follicle cells (Figure 7B). In the germline, pum mRNA is present in the germarium and later stages of oogenesis, as reported previously (MacDonald 1992). The terminal filament and interfollicular cell expression may reflect the involvement of pum in ovary differentiation and oviposition (see discussion).

Immunostaining using different anti-PUM antibodies consistently revealed that the PUM protein is present in several somatic and germline cell types in the ovaries, including the terminal filament and the invaginating follicle cells in germarial regions IIb and III, as well as postgermarial follicle cells (Figure 7C). In the germline, PUM is present at the highest level in germline stem cells and at lower levels in other germarial germline cells, as described previously (Forbes and Lehmann 1998). The follicle cell expression of PUM may be related to the aging property of follicle cells observed by Forbes and Lehmann (1998).