Hh signalling is essential for somatic stem cell maintenance in the Drosophila testis niche

Stem cells are specified and maintained by specific microenvironments called niches. The niches in the fly testis and ovary harbour germline stem cell (GSC) and somatic stem cell populations, the sizes and proliferation rates of which need to be coordinated. In the testis, the somatic cyst stem cells (CySCs) give rise to cyst cells (CyCs) that ensheath the differentiating germ cells. Both stem cell pools are arranged around a group of somatic cells termed the hub (Fig. 1A), which is the main source of niche signals for both lineages (Losick et al., 2011).

Germline and somatic stem cells divide asymmetrically. Daughter cells retaining contact with the hub keep their stem cell fate, whereas sibling cells that lose contact commit to differentiation. The differentiating germline daughter cells continue to divide, while the newly born CyCs stop dividing and exit the niche (Cheng et al., 2011; Hardy et al., 1979).

GSC stemness is determined via BMP pathway activation in both ovary and testis (Losick et al., 2011; Michel et al., 2011). Hub cells also secrete the cytokine-like ligand Upd, which mainly acts on the somatic cells (Flaherty et al., 2010; Issigonis et al., 2009; Leatherman and Dinardo, 2008; Leatherman and Dinardo, 2010; Singh et al., 2010). Despite differences in detail, the overall organisation of the testis GSC niche resembles that in the ovary (Decotto and Spradling, 2005; Lopez-Onieva et al., 2008).

In the ovary, Hedgehog (Hh) produced by the cap cells controls maintenance of the somatic follicle stem cells (FSCs) (Forbes et al., 1996; Margolis and Spradling, 1995; Zhang and Kalderon, 2001). Here we show that Hh is also an essential niche signal for somatic stem cells in the testis, implying a higher degree of organisational homology than previously appreciated.

smo³ (Chen and Struhl, 1998), ptc^IIw (Strutt et al., 2001), hh^ts2 (Ma et al., 1993), hh::Gal4 (Tanimoto et al., 2000), c587::Gal4 (Zhu and Xie, 2003), nos::Gal4VP16 (Van Doren et al., 1998), hh::lacZ=hh^P30 (Lee et al., 1992), ptc::lacZ=ptc^10.8L (Chen and Struhl, 1996), UAS::Hh (Burke et al., 1999), UAS::RedStinger (Bloomington 8545), ci::GFP (Quinones-Coello et al., 2007), and the MARCM strains (Lee and Luo, 1999) have been described previously. Clones were induced by three 1-hour heat shocks at 37°C at 1-hour intervals.

Testes were imaged as described (Michel et al., 2011). Additionally, the following antibodies were used: anti-Hh (Panakova et al., 2005) 1:100, anti-Tj (Li et al., 2003) 1:1000, anti-β-Gal (Promega) 1:1000, anti-Zfh1 (Gilboa and Lehmann, 2006) 1:5000, anti-phospho-histone H3 (Ser10) (Upstate) 1:2000, and anti-BrdU488 (BD) 1:200. Anti-Ptc (1:100), anti-Smo (1:10) and anti-Eya (1:25) were obtained from the Developmental Studies Hybridoma Bank, University of Iowa.

Data were analysed using R (www.R-project.org). Significance was calculated using Fisher's exact test or ANOVA followed by Tukey's HSD or Kruskal-Wallis tests.

Hh signals via its receptor Patched (Ptc) and the GPCR-like signal transducer Smoothened (Smo) to activate the transcription factor Cubitus interruptus (Ci) (Ayers and Therond, 2010; Ingham et al., 2011). An hh::lacZ enhancer trap that reflects Hh expression in the larval testis and adult ovary (Forbes et al., 1996) and an hh::Gal4 construct are both active in the adult hub (Fig. 1B; supplementary material Fig. S1A). However, the range of endogenous Hh protein appears to be limited (Fig. 1C).

Hh sensitivity is often controlled via Ci expression (Eaton and Kornberg, 1990; Sun and Deng, 2007). In the testis Ci::GFP (supplementary material Fig. S1B) is active in the CySCs (Dinardo et al., 2011) (Fig. 1D). CySCs also express Smo (Fig. 1E). However, we cannot exclude the possibility that these components are also present in the germline.

Since ptc is itself a transcriptional Hh target, Ptc expression (Fig. 1F) as well as its vesicular distribution in CySCs (Fig. 1G) suggest that Hh signalling is active in these cells (Chen and Struhl, 1996), consistent with the previously reported ptc::Gal4 pattern (Schulz et al., 2002; Tazuke et al., 2002).

Thus, in the adult testis, Hh appears to signal from the hub to the adjacent somatic cells.

We generated cells homozygous for the amorphic alleles smo³ or ptc^IIw (Chen and Struhl, 1998; Strutt et al., 2001) using the MARCM system (Lee and Luo, 1999). Loss of Smo blocks Hh signalling, whereas ptc loss-of-function mutations activate the pathway (Ingham et al., 2011). Since, under normal conditions, CySCs are the only dividing somatic cells (Cheng et al., 2011; Hardy et al., 1979), all marked cells must be the product of a stem cell mitosis. CySCs were marked by the Jak/Stat target Zfh1, which is, however, also present in the second tier of somatic nuclei (Leatherman and Dinardo, 2008).

We generated control as well as smo and ptc mutant clones (Fig. 2A-D) with comparable efficiency (supplementary material Fig. S2). The fraction of testes with detectable clones increased between 1 and 2 days after clone induction (ACI). Since CySC cell-cycle length must be coupled to that of the GSCs, CySCs divide roughly once per day (Sheng and Matunis, 2011; Wallenfang et al., 2006). The lag between induction and detectability is presumably caused by the variable time before individual CySCs exit G2 phase and the time required for the Gal80 protein to decay after mitosis (Lee and Luo, 1999).

Compared with controls, the fraction of Zfh1-positive smo³ cells was already reduced at 2 days ACI (1.8±3.6% versus 12.8±8.1% in controls, P<0.001, ANOVA; Fig. 2E). By 3 days ACI, these cells had largely disappeared, although Zfh1-negative smo³ cells were still present further out (Fig. 2C). Assuming equal induction rates for smo³ and control CySCs, we fitted the ratios of Zfh1-positive smo³ cells to control cells versus time with an exponential decay with a half-life of ~9 hours (supplementary material Fig. S3).

By contrast, the fraction of ptc^IIw cells among the Zfh1-positive pool increased relative to controls between 2 and 3 days ACI (2 days ACI, 17.5±9.7% versus 9.4±9.2% in controls, not significant; 3 days ACI, 34.1±15.4% versus 18.1±13.4%, P<0.001, ANOVA; Fig. 2F). Thus, by 3 days ACI, 67.6±18.6% of ptc^IIw cells were Zfh1 positive in comparison with 35.1±20.7% of control clones (P<0.001, t-test; supplementary material Fig. S4A). This was not caused exclusively by the accumulation of additional ptc CySCs next to the hub. In controls, the average distance between the centre of Zfh1-positive cells and the hub did not change between 2 and 3 days ACI (12.3±3.1 μm versus 11.1±2.8 μm, P>0.05, ANOVA; supplementary material Fig. S4B). These distances are characteristic for CySCs, which are expected to lie between the first and second tier of germline cells found at 8.3±1.0 μm and 13.1±3.0 μm, respectively. By contrast, the average distance from the hub of Zfh1-positive ptc^IIw mutant cells at 3 days ACI (20.6±6.3 μm) was increased relative to both mutant cells at 2 days ACI (16.5±5.8 μm, P<0.001) and controls at 3 days ACI (P<0.001).

We therefore examined proliferation of the ptc cells by phospho-histone H3 (pH3) staining. In neutral control clones, 0.5±0.9% (3 days ACI) to 1.1±1.1% (2 days ACI) of all homozygous cells were pH3 positive (Fig. 2G; supplementary material Fig. S5). As expected (Cheng et al., 2011; Hardy et al., 1979), all somatic mitoses occurred adjacent to the hub. By contrast, at both time points, ~2.5% of all ptc mutant cells (2 days ACI, 2.5±1.2%; 3 days ACI, 2.5±0.5%) were pH3 positive (ptc versus control at 3 days ACI, P<0.05, t-test; supplementary material Fig. S5). About one-third of these excess mitoses (27/71) occurred away from the hub (Fig. 2H). These results implicate Hh in CySC proliferation, resembling the situation in the ovary, where Hh signalling drives FSC proliferation (Forbes et al., 1996; Hartman et al., 2010; Zhang and Kalderon, 2001).

Control CySCs marked by CD8-GFP could be identified by their processes extending towards the hub (Fig. 3A). smo³ mutant cells were missing from the hub at 3 days ACI (Fig. 3B), but could be seen ensheathing germline cysts (Fig. 3B), suggesting that they had differentiated into functional CyCs.

Consistently, by 3 days ACI, almost all smo³ cells expressed the CySC differentiation marker Eyes absent (Eya) (Fabrizio et al., 2003) (Fig. 3C), whereas control clones retained a sizeable Eya-negative population (smo clones, 1.4% Eya negative, n=139 cells; controls, 22.1% Eya negative, n=417 cells; P=5.5×10⁻¹¹, Fisher's exact test; Fig. 3D). The average number of homozygous cells per testis was about threefold higher in controls than in smo³ clones (19.4±8.7 versus 6.6±3.9, P<0.001, t-test).

Homozygous ptc^IIw CySCs were able to retain contact with the hub. However, other mutant cells ensheathed germline cysts (Fig. 3E) or expressed Eya (Fig. 3F). Both the number of homozygous cells per testis (31.6±13.2 versus 13.3±9.7, P<0.01) and the size of the Eya-negative fraction (28.1±12.5% versus 13.8±16.8%, P<0.05) were increased for ptc mutant cells relative to controls (Fig. 3D). Eya-negative ptc mutant cells were also found further away from the hub than corresponding controls (Fig. 3D), mirroring the expansion of the corresponding Zfh1-positive population (Fig. 2D,F). Thus, activation of the Hh pathway promotes CySC fate but does not prevent differentiation.

To address whether systemic loss of Hh affected niche function, we turned to the temperature-sensitive allele hh^ts2 (Fig. 4A,B), which behaves similar to the wild-type allele at 18°C, but as an amorphic hh mutant at 30°C (Ma et al., 1993). The average number of Zfh1-positive cells did not differ between testes from homozygous hh^ts2 males raised at 18°C relative to their heterozygous siblings (P>0.05, ANOVA) or males held at permissive conditions for 13 days (Fig. 4C). By contrast, CySC number was reduced in hh^ts2 males relative to heterozygous controls after 3 (Fig. 4A,B), 6 or 10 days at 30°C (P<0.001 for 3 and 6 days, P<0.05 at 10 days, ANOVA; Fig. 4C). However, CySCs did not completely disappear from the hub when Hh signalling was equally affected in all cells, and there was no significant difference in BrdU uptake between Zfh1-positive cells from hh^ts2 homozygous versus heterozygous testes (hh^ts2, 45±22%, n=7; control, 49±14%, n=9; P>0.05, t-test; supplementary material Fig. S6A,B).

We therefore propose that competition between the somatic stem cells (Rhiner et al., 2009) is largely, although not exclusively, responsible for the loss of homozygous mutant smo CySCs and the proliferative advantage of individual ptc mutant CySCs in an otherwise heterozygous testis.

GSC number was reduced in hh^ts2 testes following incubation at 30°C relative to heterozygous controls kept at the same restrictive conditions and to homozygous hh^ts2 males left at the permissive temperature of 18°C (Fig. 4D-F). Hub maintenance, however, was not affected in hh^ts2 males (supplementary material Fig. S7).

We induced smo and ptc germline clones marked by nlsGFP (Fig. 4G,H). Importantly, smo³ and ptc^IIw mutant GSCs were maintained at the same level as control clones (Kruskal-Wallis test, P>0.05 for all comparisons; Fig. 4I). Thus, Hh signalling is not required in the germline, and the observed reduction in GSC number caused by systemic inactivation of Hh is presumably mediated via the somatic cells.

In summary, Hh provides a niche signal for the maintenance and proliferation of the somatic stem cells of the testis. CySCs that are unable to transduce the Hh signal are lost through differentiation, whereas pathway overactivation causes overproliferation. Hh signalling thereby resembles Jak/Stat signalling via Upd. Partial redundancy between these pathways might explain why neither depletion of Stat activity nor loss of Hh signalling causes complete CySC loss (Leatherman and Dinardo, 2010).

We have shown here that loss of Hh signalling in smo mutant cells blocks expression of the Jak/Stat target Zfh1 (Leatherman and Dinardo, 2008), whereas mutation of ptc expands the Zfh1-positive pool. Overexpression of Zfh1 or another Jak/Stat target, Chinmo, is sufficient to induce CySC-like behaviour in somatic cells irrespective of their distance from the hub (Flaherty et al., 2010; Leatherman and Dinardo, 2010). By contrast, Hh overexpression in the hub using the hh::Gal4 driver only caused a moderate increase in the number of Zfh1-positive cells relative to a GFP control. Ectopic Hh overexpression in somatic cells under c587::Gal4 control increased this number further (Fig. 4J,K). However, unlike in somatic cells with constitutively active Jak/Stat signalling (Leatherman and Dinardo, 2008), the additional Zfh1-positive cells remained largely confined to the testis tip, although their average range was increased threefold (Fig. 4J,L). Thus, Hh appears to promote stem cell proliferation, in part, also independently of competition.

It is tempting to speculate that further stem cell expansion is limited by Upd range. Consistently, cells with an ectopically activated Jak/Stat pathway remain undifferentiated (Leatherman and Dinardo, 2008), whereas ptc cells can still differentiate. Future experiments will need to formally address the epistasis between these pathways. However, our observations already show that Hh signalling influences expression of the bona fide Upd target gene zfh1, and therefore presumably acts upstream, or in parallel to, Upd in maintaining CySC fate.

In addition, the reduction in GSC number following somatic stem cell loss implies cross-regulation between the different stem cell populations that presumably involves additional signalling cascades, such as the EGF pathway (Gilboa and Lehmann, 2006; Sarkar et al., 2007).

In recent years, research has focused on the differences between the male and female gonadal niches (Fuller and Spradling, 2007). We would instead emphasize the similarities: in both cases, Jak/Stat signalling is responsible for the maintenance and activity of cells that contribute to the GSC niche, and Hh signalling promotes the proliferation of stem cells that provide somatic cells ensheathing germline cysts. In the testis, both functions are fulfilled by the CySCs (Leatherman and Dinardo, 2010), whereas in the ovary the former task is fulfilled by the postmitotic escort stem cells/escort cells (Lopez-Onieva et al., 2008; Morris and Spradling, 2011) and the latter by the FSCs (Margolis and Spradling, 1995). Finally, male desert hedgehog (Dhh) knockout mice are sterile (Bitgood et al., 1996; Clark et al., 2000). Dhh is expressed in the Sertoli cells and is thought to primarily act on the somatic Leydig cells (Yao et al., 2002). However, the signalling microenvironment of the vertebrate spermatogonial niche is, as yet, not fully defined. Future experiments will need to clarify whether these similarities reflect convergence or an ancestral Hh function in the metazoan gonad.

We thank D. Allan, C. Dahmann, S. DiNardo, S. Eaton, D. Godt, P. Lasko, R. Lehmann, G. Weidinger, T. Xie, the Bloomington Stock Center and the Developmental Studies Hybridoma Bank for providing flies and reagents; R. Pérez-Palencia for expert technical assistance; H. Grandel for critical reading of the manuscript; and members of the Junior European Drosophila Investigators (JEDI) initiative for discussions.