Drosophila piwi mutants exhibit germline stem cell tumors that are sustained by elevated Dpp signaling

Abstract

Drosophila Piwi is the founding member of a gonadal clade of Argonaute proteins that serve as silencing effectors for ~26–32 nucleotide piRNAs [1], and piwi mutants exhibit dramatically rudimentary ovaries [2]. It was proposed that somatic Piwi maintains germline stem cells (GSCs) by promoting Dpp signaling, presumably via cap cells that comprise the somatic niche for GSCs [3–5]. However, we unexpectedly observed that piwi mutants exhibit high-frequency GSC-like tumors that persist throughout adult life. Multiple readouts demonstrated hyperactive Dpp signaling in piwi mutants, including the failure to express the germline differentiation factor bag-of-marbles (bam), and restoration of bam expression relieved piwi GSC-like tumors. Tissue-specific rescue and knockdown experiments indicate that Piwi is not required in cap cells, the source of niche Dpp, but instead in gonadal intermingled cells (ICs, the progenitor cells of escort cells). Adult-specific knockdown of dpp in escort cells substantially rescued piwi tumors, demonstrating that they are driven by excess Dpp signaling. However, the temporal requirement for piwi to restrict GSC numbers was much earlier, in wandering 3^rd instar larvae. Indeed, piwi mutant larval gonads exhibited defective morphology and loss of Bam. Our data indicate that loss of Piwi causes defects in ICs and escort cells, leading to ectopic Dpp signaling and consequent blockage of GSC differentiation.

Results and Discussion

Drosophila ovaries are composed of ~20 ovarioles, each consisting of a germarium and a string of maturing egg chambers. The germarium tip contains 2–3 GSCs, identified by their direct contact with cap cells and the presence of ball-shaped spectrosomes stained by Hu-li-tai-shao (Hts); differentiating germline cells maintain Hts expression in branched fusome structures. Ovarian GSCs receive a localized Dpp signal emanating from neighboring somatic niche cells, resulting in phosphorylation of the transcription factor Mothers against dpp (pMad), which in turn represses expression of the differentiation factor bag of marbles (bam) [6–8]. Asymmetric GSC division retains one daughter in the niche; this cell maintains Dpp pathway activity and GSC identity. The other GSC daughter loses Dpp signaling and consequently activates Bam, which promotes its differentiation as a cystoblast [9].

piwi mutants exhibit small ovaries and their ovarioles generate few eggs, and initial studies concluded that Piwi was essential to maintain ovarian GSCs [2]. It was later determined that somatic expression of Piwi rescues piwi mutant germline defects, leading to a model in which Piwi positively regulates Dpp signaling, presumably in cap cells that comprise the GSC niche [3–5, 10]. We examined this using previously studied piwi alleles as well as a deletion allele of the neighboring loci scar and piwi known as [Δ37] [11], which truncates the Piwi protein at residue 745 (personal communication with Eyal Schejter). We confirmed that piwi[Δ37] homozygotes rescued for scar expression lack Piwi immunoreactivity in ovaries, as is the case for other piwi trans-heterozygous combinations (Figure S1A-D).

The rudimentary size and disorganization of piwi ovaries makes them challenging to analyze, and we were also concerned about the potential loss of agametic structures during dissection. We therefore utilized the conventional approach of teasing ovaries apart, as well as examined whole undissected ovaries. The former technique yields flatter mounts of separated ovariole structures for optimal imaging, whereas the latter method is more cumbersome to interpret as ovarioles are more distributed through the Z-axis of the tissue, but it ensures that no material was lost during preparation. We obtained similar results with both methods.

Based on previous reports [3–5], we expected to see a substantial population of germaria lacking GSCs in newly born piwi[1/Δ37] ovaries. Careful examination revealed several classes of structures. We did indeed identify germaria lacking GSCs, but these were rare. We classified these on the basis of a stack of disc-shaped terminal filament (TF) cell nuclei adjacent to Tj+ cap cells (CCs), followed by an identifiable germarium structure containing Tj+ escort cells (ECs) but lacking GSCs (Figure 1A). These structures sometimes included an egg chamber, but often they did not. Much more common, however, were TF/CC structures that not only lacked associated germline cells, but also lacked ECs (Figure 1B). Although such "orphan" TF/CC clusters lack GSCs, their absence of an intact germarium raised the possibility that they never recruited germline cells during niche formation, as opposed to subsequently lost their GSCs. At present we cannot distinguish these scenarios, but as "GSC loss" and "orphan TF/CC" classes comprised 5.3% and 25.2% of structures, respectively (n=246), they collectively comprise a minority of germaria.

Figure 1.

piwi mutants exhibit predominant germline stem cell-like (GSC-like) tumors and elevated Dpp signaling. (A-C) Partial or whole germaria in <1-day old piwi[1/Δ37] ovaries, stained for Hts (red), Tj (green) and DNA (blue), quantified as noted. (A) Example of germarium lacking GSCs. (B) Partial germarium consisting of “orphan" TFs and CCs. (C) Example of germarium displaying a GSC tumor. (D-F) Egg chambers are abbreviated in piwi mutants and exhibit GSC-like tumors; these phenotypes are rescued by myc-piwi. Select germarium regions are boxed and enlarged in D'-F', highlighting ~3 GSC cells with ball spectrosomes (marked with white circles) in piwi heterozygous (D') and rescued piwi mutants (F'), but GSC-like tumors filling the germaria of piwi mutants (E'). (G) Quantification of spectrosome-containing cells in 7-day old piwi heterozygous and mutant ovarioles. (H-J) 30-day old germaria of the respective genotypes resemble their 7-day old counterparts. (K) Quantification of germline cells expressing the GSC marker pMad in different piwi genotypes. (L, M) Representative staining of pMad in piwi heterozygous (L) and mutant (M) germaria. (N) Quantification of germline cells expressing the GSC/cystoblast marker Dad-GFP in different piwi genotypes. (O, P) Representative Dad-GFP expression in piwi heterozygous (O) and mutant (P) germaria. Error bars in G, K, N represent mean with SD.

Surprisingly, the strong majority of germaria (69.5%, Figure 1C and Figure S1E) exhibited extra spectrosomes indicative of GSC tumors. We confirmed similar results in newlyborn piwi[1/2] mutants (n=245, quantified in Figure S1E). We further verified the GSC tumor phenotype in multiple hetero-allelic combinations of piwi[1], piwi[2], piwi[pz] and piwi[Δ37], and this phenotype was rescued by myc-piwi transgene (Figure1D-G and Figure S1F-J). Moreover, piwi[Δ37] homozygotes rescued for Scar expression also exhibited ectopic GSC-like cells (Figure S1H).

We continued to observe highly supernumerary spectrosome-containing germline cells at 30 days (Figure 1I and Figure S1K), indicating that this was not a transient phenomenon. At this point, piwi mutant ovaries exhibited considerable degeneration and ectopic cell death (Figure S1M, N). Nevertheless, in the remaining germaria we still observed large numbers of spectrosome-containing cells in one month-old piwi mutants, and these phenotypes could be fully rescued (Figure 1H-J, Figure S1L). Altogether, we find clear evidence for GSC-like tumor that persist throughout piwi mutant adult life. Since we were currently unable to distinguish whether "orphan" TF/CC structures are attributable to GSC loss or defective GSC recruitment, we focused our subsequent analyses and quantifications on ovarioles bearing intact germaria.

Ovarian GSCs are maintained by activity of nuclear pMad. In newly-born and 2–3 day old ovaries (n=50 for all genotypes), control piwi heterozygous ovaries contained 3.0±0.1 pMad+ cells (Figure 1K). In contrast piwi[1/2] exhibited 8.4±0.6 and piwi[2/Δ37] had 7.7±0.7 pMad+ germline cells, and this defect was fully rescued by myc-piwi to 3.5±0.1 pMad+ cells (Figure 1K). Representative stains of wild-type and piwi germaria in Figure 1L and M highlight that the ectopic pMad+ cells exhibit ball-shaped spectrosomes, instead of branched fusomes typical of multi-cell cystocytes, attesting to their undifferentiated status. A direct downstream transcriptional target of pMad is Daughters against dpp (Dad), which can be monitored by expression of Dad-GFP. Indeed, we observed ectopic Dad-GFP expression in different piwi mutants, 19.2±0.7 in piwi[1/2] and 20.9±0.8 in piwi[2/Δ37] (n=32, Figure 1N); representative germaria are shown in Figure 1O-P. Overall, piwi mutations appear to elevate levels of Dpp signaling in the majority of undifferentiated germline cells, thus keeping them in GSC-like and/or cystoblast-like stages.

A critical function of Dpp signaling in GSCs is to repress the transcription of bag of marbles (bam). We observed drastic reduction of Bam+ germline cells (Figure 2A-C), from 92% of control germaria (Figure 2A, white outlined regions, n=149) to 4.1% in piwi[1/Δ37] (n=171) and 9.2% in piwi [1/2] mutants (Figure 2B-C, n=185). This was a transcriptional effect, since a bam-GFP reporter that is normally activated in late CBs and other differentiating germline cells (Figure 2D) was largely silent in piwi[2/Δ37] (n=57) and piwi [1/2] (n=48) mutants (Figure 2E and F, white outlined areas). To assess whether loss of Bam is causal for piwi GSC-like tumors, we introduced a hs-bam transgene into piwi mutants. While heat-shock alone was unable to rescue fusome morphology in piwi mutants (Figure 2G-H) "branched” fusome morphology was regained in hs-bam, piwi mutants 18 hour after heat-shock (n=46 germaria, Figure 2I-J, dashed line circles). Consistent with this, induction of hs-Bam also reduced ectopic pMad staining in piwi mutants, even eliminating endogenous staining (Figure 2K, L). We conclude that the lack of bam expression in piwi germline cells is a major cause of their GSC-like tumors.

Figure 2.

Failure of GSC differentiation in piwi mutants is due to loss of Bam expression. (A) piwi[1]/+ heterozygote illustrates normal accumulation of Bam protein in differentiating cystocytes (white lines). (B, C) Almost 10% of piwi[1/2] mutant ovaries initiate Bam expression (B, white lines), but the vast majority fail to express Bam (C). Bam staining is shown separately in (A'-C'). (D) piwi[1]/+ heterozygote illustrates a normal pattern of bam-GFP expression that begins in differentiating cystocytes - germline cells other than GSCs (white circles) and CBs (dashed circle). (E) piwi[1/2] and (F) piwi[2/Δ37] mutants fail to activate bam-GFP in those ectopic GSC-like cells (white line areas). Bam-GFP staining is shown separately in (D'-F'). (G-J) Rescue of piwi mutants by ectopic Bam. (G) piwi mutant exhibits GSC-like tumors and these persist following heat-shock (H). (I-J) In piwi mutants bearing hs-bam, spectrosome morphology changes from ball-shaped before heat-shock (I) to branched-shape after a heat-shock treatment (J, dashed line areas). (K) piwi[1/2]; hs-bam/+ germarium exhibits many ectopic pMad+ cells with ball-shaped spectrosomes without a heat-shock treatment. (L) Following heat-shock, piwi[1/2];hs-bam/+ germarium exhibits pMad-negative cells with branched-shaped fusomes (dashed lines) in the niche. pMad staining is shown separately in (K'-L').

To gain insight into the spatial requirement for Piwi function, we used several cell type-specific Gal4 lines for rescue experiments. Consistent with previous reports that Piwi is required primarily in the soma to maintain overt ovarian morphology [4], expression of Piwi using the germline driver nanos-Gal4 did not rescue piwi tumors (Figure S3A, B). hh-Gal4 is active in the somatic niche, i.e., in terminal filament (TF) and cap cells (Figure S2A). Reporter analysis showed that hh-Gal4 activity is not maintained in all piwi mutant germaria (Figure S2B-D); therefore, we scored only germaria with demonstrated Piwi expression in cap cells (Figure S2E, F). Surprisingly, expression of Piwi using hh-Gal4 did not rescue piwi tumors (Figure S3C), suggesting that Piwi is not required in the cells that are believed to be the normal source of niche Dpp. The tj-Gal4 driver is active in cap cells, escort cells and other follicle cells [12] (except stalk cells, Figure S2G-H). Interestingly, expression of Piwi in Tj+ cells completely suppressed piwi germline tumors and restored egg production (Figure S3D), although these eggs could not hatch, likely due to germline Piwi function [13, 14]. The c587-Gal4 driver has similar specificity to tj-Gal4 in germarial areas, except that it is not active in cap cells (Figure S2J-K). We also observed full rescue of GSC dynamics and ovary morphology using c587-Gal4 to activate Piwi (Figure 3A and Figure S3E), implying that Piwi is predominantly required in escort cells.

Figure 3.

Spatial and temporal function of Dpp signaling and Piwi for normal GSC differentiation. (A) piwi GSC-like tumors are rescued by expression of piwi using c587-Gal4, which is active in escort cells and follicle cells. White circles highlight two pMad+ GSCs, characteristic of wildtype. (B) Adult-specific rescue of piwi mutant germaria by temporally-controlled induction of dpp- RNAi. Inclusion of tub-Gal480[ts] in the background restricts Gal4 activity until shifting to the restrictive temperature (29°C) in adult stages. Ovaries of this genotype remain small due to the developmental defect, but examination of their germaria indicates rescue of normal GSC numbers (white circles) and the presence of differentiating cysts (dashed lines). (C, D) Validation of cell-type and adult-specific knockdown of Piwi. Females of the genotype c587-Gal4, UAS-piwi-RNAi; tub-Gal80[ts] were raised at 18°C until eclosion, then maintained at 18°C (C) or shifted to 29°C (D) for 11 days of adult life prior to immunostaining for Piwi, Tj and Gypsyenvelope proteins. Late-stage egg chambers exhibit normal Piwi and Tj accumulation and lack Gypsy reactivity in control (C-C''), whereas animals shifted to the restrictive temperature specifically lack Piwi in somatic cells and exhibit derepression of Gypsy (D-D''). (E) Knockdown of piwi using cell-type and temporal control demonstrates that GSC-like tumors can be generated by specifically depleting Piwi using c587-Gal4 during late larval stages. (F) Expression of Piwi with cell-type and temporal control in piwi mutants shows that GSC-like tumors can only be substantially rescued when providing Piwi at the third instar using c587-Gal4. See also Figures S2, S3, and S4. Error bars in E, F represent mean with SD.

In reciprocal tests, we asked whether cell-type restricted expression of UAS-piwi-RNAi could induce germline tumors. As with the rescue experiments involving hh-Gal4, we only scored knockdown germaria for which we confirmed Piwi loss in cap cells (Figure S2M, N). We failed to find ectopic GSC-like cells upon induction of constitutive piwi knockdown using hh-Gal4 (4.6±0.3, n=30, p=0.26) even when performing the experiment in a piwi heterozygous background to sensitize the knockdown (Figure S3F). On the other hand, knockdown of piwi using either c587-Gal4 (12.8±7.4, n=30, p<0.0001, Figure 3E) or tj-Gal4 (13.1±6.2, n=30, p<0.0001, Figure S4E) both induced large numbers of ectopic GSC-like cells (Figure S3G, H).

Since Dpp is not normally considered to influence GSC dynamics via escort cells, we sought evidence that Dpp is required for piwi mutant GSC tumors via escort cells. We observed GSC loss in 26% of c587-Gal4 driven dpp-RNAi germaria, which potentially suggested interference of Dpp signaling in niche cells during development (data not shown). To bypass potential developmental effects, we used the tub-Gal80[ts] system to temporally control dpp knockdown. A regimen in which Dpp was silenced for 4 days by shifting adult flies to 29°C rescued GSC-like tumors and promoted germline differentiation in 64.7% piwi mutant germaria (Figure 3B, Figure S3I, n=68). This provided compelling evidence that continuous upregulation of Dpp signaling in escort cells is a primary cause of piwi germline tumors in adult stages.

We subsequently investigated temporally-controlled manipulations of Piwi. Control tub-Gal80[ts]; c587-Gal4>UAS-piwi-RNAi females raised at the permissive temperature showed normal Piwi staining and ovary morphology (Figure 3C and Figure S4A). In contrast, siblings shifted to the restrictive temperature showed specific loss of Piwi protein in somatic cells (Figure 3D and Figure S4B). This was accompanied by de-repression of Gypsy transposons, as evidenced by accumulation of Gypsy envelope protein (Figure 3C'' vs. D''). Surprisingly, this adult-specific regimen for Piwi depletion did not significantly induce ectopic spectrosome-containing germline cells (Figure 3E). We therefore assayed other developmental stages. We observed that c587-Gal4-mediated knockdown of piwi specifically in pupal stages caused a mild increase in GSC-like cells, but restricted knockdown in wandering third instar larvae led to a substantial piwi phenocopy (Figure 3E). We obtained identical results by depleting piwi using temporally-controlled activity of tj-Gal4 (Figure S4C-E). We complemented these tests by resupplying Piwi in piwi mutants at various developmental stages, using tub-Gal80[ts] and c587- Gal4. We obtained substantial rescue when Piwi was induced transiently in wandering third instar larvae, but not at pupal or adult stages (Figure 3F). These experiments established that Piwi is critically required during wandering third instar larval stages.

During niche formation in larval gonads, ICs are essential for the differentiation of primordial germline cells (PGCs, the progenitor cells of adult germline stem cells) residing outside of niches. We confirmed that Tj-positive ICs in piwi gonads are not mixed with germline cells as in the control [15], but instead remain on the periphery of the germline cell mass (Figure 4A,B). Concomitant with this was the failure to upregulate bam-GFP in piwi mutant larval gonads (Figure 4C,D), similar to its behavior during adult stages (Figure 2A-F). Moreover, escort cells derived from IC lineages exhibited defective morphology, since membrane extensions labeled by c587-Gal4>UAS-CD8-GFP were clearly reduced in piwi mutants (Figure 4E, F); escort cell numbers were also reduced (Figure 4G). Altogether, the temporal requirement for Piwi in larval gonad development suggested that its absence leads to inappropriately sustained Dpp signaling in ICs, which in turn impedes the developmental switch of PGC differentiation. This appears to be mediated at least partly by defective differentiation of IC-derived lineages, including of escort cells whose cellular extensions are important for ovarian GSC differentiation [16].

Figure 4.

Somatic defects in piwi mutant larval gonads are associated with defective escort cell morphology. Panels A-D show immunostaining of wandering third instar larval gonads; dashed lines delineate areas containing primordial germline cells. (A, B) Intermingling of somatic cells (marked by Tj) and germline cells (marked by Vasa) occurs in piwi heterozygotes (A) but not in piwi [1/Δ37] mutants (B). Upregulation of the germline differentiation marker bam-GFP occurs in piwi heterozygotes (C, circled) but not in piwi[2/Δ37] mutants (D), and only the former exhibits cells with branched fusome structures (C', circled). (E) Escort cells in adult piwi heterozygous germaria exhibit characteristic cellular extensions, but these are largely missing in piwi [1/Δ37] mutants (F). Note that cap cells (dashed line areas) are labeled in piwi mutants (F) but not in control (E). (G) Loss of piwi also reduces the number of escort cells. Error bar in G represents mean with SD.

In summary, we analyzed multiple heteroallelic backgrounds, genetic rescues, and temporally-controlled manipulations, to show that the predominant effect of piwi loss is elevation of Dpp signaling leading to GSC tumors. We demonstrate this using multiple readouts of GSC fate and Dpp pathway status, and further show that piwi mutant tumors were partially rescued by somatic knockdown of dpp or by re-expression of Bam in the germline. Curiously, while our spatially- and temporally-controlled manipulations demonstrate that piwi GSC tumors are actively sustained in the adult by excess Dpp signaling, the direct requirement for Piwi for normal germline formation is largely restricted to larval stages. Tj-positive ICs do not mix with PGCs in late 3rd instar piwi gonads, which coincides with the failure of PGCs to express the differentiation factor Bam. We propose that piwi adult ovarian phenotypes stem from a failure of ICs to downregulate dpp and/or differentiate into escort cells properly.

Further investigation is needed to establish the molecular mechanism by which Dpp signaling is deregulated. In principle, this might be a direct consequence of defective piRNA-mediated silencing, or an indirect cause of defective escort cell differentiation [16]; these possibilities are not exclusive. Finally, while the majority of piwi ovarioles exhibit GSC tumors, we note a substantial population of aberrant structures consisting of "orphan" TF/CC clusters that are not associated with germline cells (Figure 1). These appear distinct from the rare class of intact ovarioles exhibiting overt GSC loss, and raises the question of whether piwi is also required for effective recruitment of GSCs during niche formation. Additional studies are needed to clarify the developmental impact of Piwi loss during different aspects of ovary development.

Materials and Methods

Drosophila strains

Fly crosses were performed at 25°C unless otherwise stated. We analyzed the following piwi alleles and transgenes: piwi[1], piwi[2], piwi[pz06843] (designated as piwi[pz]), Df(2L)[Δ37] (a small deletion that affects scar and piwi [11], hereafter designated as piwi[Δ37]), myc-piwi and UAS-piwi-RNAi (Bloomington Stock Center #33724). For cell-type specific expression and rescue experiments, we used UASp-HA-piwi driven with c587-Gal4 (from Ting Xie), tj-Gal4 (from Dorothea Godt), act5C-Gal4 (from Ruth Lehmann), hh-Gal4, nos-Gal4 and tub-Gal80[ts] (from Bloomington). Analysis of Dpp signaling utilized Dad-GFP [17] and UAS-dpp-RNAi (Bloomington #31172 and #31767); nearly all tj-Gal4>UAS-dpp-RNAi ovaries lacked GSCs, validating their potency (data not shown). Analysis of Bam function utilized bam-GFP and hsbam. To express Bam in piwi mutants, we subjected piwi[1], hs-bam/piwi[2] flies to two 1 hour heat-shocks at 37°C, separated by a 2 hour recovery period at 25°C; ovaries were analyzed 18 hours after the first heat-shock. For crosses involving tub-Gal80[ts], flies were first cultured at 18°C to desirable developmental stages, and they were then up-shifted to 29°C to activate UAS transgenes in specific cells driven by individual Gal4 lines.

Immunostaining

Ovaries from adult flies or larval gonads were dissected and fixed in PBS containing 4% formaldehyde. Mouse anti-Hts (1:10, DSHB), rabbit anti-GFP (1:2500, Clontech), rat anti-Vasa (1:50, DSHB), rabbit anti-pMad (1:2000, gift of Ed Laufer and Tom Jessel), guinea pig anti-Tj (1:10000, gift of Dorothea Godt), rabbit anti-Caspase 3 (1:200, Cell Signaling), mouse anti-Bam (1:10, DSHB) mouse anti-Piwi (1:10, gift of Haruhiko and Mikiko Siomi) and rabbit anti-Gypsy Env (gift of Alain Pelisson). Alexa Fluor-488, -568 and -647 secondary antibodies were from Molecular Probes and used at 1:500. Apoptosis Detection Kit (Millipore) was used for TUNEL essay. Images were captured with a Leica TCS confocal microscope.

Highlights.

• piwi mutants unexpectedly display high frequency ovarian GSC-like tumors

• The tumor phenotype is due to inhibition of bam triggered by ectopic Dpp activity

• The requirement for Piwi to restrict GSC dynamics occurs mostly during the wandering 3^rd instar larval stage

• Piwi appears to function in progenitor cells of escort cells to promote differentiation of PGCs in gonads