Rab11 maintains connections between germline stem cells and niche cells in the Drosophila ovary

Drosophila oogenesis is an excellent system in which to study stem cell maintenance and differentiation because all of the steps unfold in well-defined compartments. The initial steps occur within the germarium, which is divided along its anterior-posterior axis into three morphologically distinct regions (Fig. 1A). Two to three germline stem cells (GSCs) are attached by adherens junctions to niche cap cells at the extreme anterior end of germarial region 1(Song et al., 2002; Kirilly and Xie, 2007). The cap cells and other neighboring niche cells continuously secrete Dpp and Gbb,short-range TGF-β-like signaling molecules that maintain GSC identity through repression of bam transcription(Xie and Spradling, 1998; Xie and Spradling, 2000; Song et al., 2004). Each GSC divides along its anterior-posterior axis to produce another GSC, which remains attached to the cap cells, and a posterior cystoblast, which is displaced from the niche and free to differentiate. The axis of GSC division is determined by the membrane- and Spectrin-rich fusome, which anchors one pole of the mitotic spindle to the anterior cortex of the GSC(McGrail and Hays, 1997; Deng and Lin, 1997; de Cuevas and Spradling, 1998). A small amount of fusome is donated to the cystoblast, where it guides four stereotypic rounds of incomplete cell division resulting in a germline cyst of 16 cells interconnected by cytoplasmic bridges, called ring canals(Huynh and St Johnston, 2004). In germarial region 2A, Orb, BicD and other cell-fate determinants become enriched in the cell with the most fusome material, committing it to differentiate as the oocyte, while each of the other 15 cells adopts a nurse cell fate (Huynh and St Johnston,2004). The oocyte is positioned at the posterior end of the germline cyst in region 2B through E-cadherin (encoded by shotgun -FlyBase)-mediated adhesion to neighboring somatic follicle cells(Gonzales-Reyes and St Johnston,1998; Godt and Tepass,1998). Finally, in germarial region 3, the germline cyst is encased in an epithelium of somatic follicle cells thereby forming the stage 1 egg chamber, the basic unit of all subsequent steps of oogenesis(Huynh and St Johnston,2004).

Fly culture and crosses were carried out according to standard procedures(Ashburner, 1989). The wild-type stock was w, or w His2AV::GFP(Morin et al., 2001). The rab11 deletion(rab11^ΔFRT) was made by inducing recombination (Parks et al.,2004) between the FRT insertions (FRT5377 and FRT1994,respectively) of stocks f05377 and d01994 (Harvard Medical School Exelixis Collection). The resulting deletion, which removes the rab11 promoter and first two exons, was initially identified by non-complementation with rab11²¹⁴⁸ (Dollar et al., 2002) and subsequently confirmed by PCR. The rab11^ΔFRT allele complements a lethal allele of rtet, which lies just upstream of rab11 and close to the FRT insertion of f05377, and produces no protein(Fig. 1I). Homozygous mutant clones were generated by crossing w; rab11-null/FRT5377,Hrb98DE::GFP or w; rab11⁺/FRT5377,Hrb98DE::GFP controls to y w hsp::FLP. The FRT5377,Hrb98DE::GFP chromosome was made by recombining the Hrb98DE::GFPtransgene from line ZCL058 (Morin et al.,2001; Kelso et al.,2004) onto the FRT5377-containing chromosome and was verified by PCR. For most experiments, clones were induced in 2- to 5-day-old adults by heat shocking for 1 hour at 37°C on two consecutive days and examined 8 or more days ACI, thus ensuring that all examined rab11-null cells were derived from mutant GSCs; germline cysts normally clear the germarium within∼6 days (Song et al.,2002; Xie and Spradling,1998). For half-life determination, a single large group of 2- to 3-day-old adults were heat shocked twice, 8 hours apart, at 37°C for 1 hour and the number of mutant GSCs and germline cysts were counted 4, 8 or 12 days ACI. Homozygous rab11-null and rab11⁺control clones were identified by their lack of GFP staining. The fully functional rab11::GFP transgene is identical to that described by Dollar et al. (Dollar et al.,2002), except for the omission of the N-terminal His tag.

Ovaries were fixed and immunostained as previously described(Dollar et al., 2002), except that electron microscopy-grade formaldehyde was substituted for paraformaldehyde in the fixative. Primary antibodies were used at the following concentrations: rat anti-Rab11 (1:500)(Dollar et al., 2002); rabbit anti-Rab11 (1:250) (Satoh et al.,2005); E-cadherin (1:40; Hybridoma Bank); GFP (1:250; Invitrogen);α-Spectrin (1:10; Hybridoma Bank); Hts (1b1) (1:4; Hybridoma Bank); Orb(6H4) (1:20; Hybridoma Bank); Vasa (1:5000)(Williamson and Lehman, 1996);HtsRC (1:4; Hybridoma Bank); and BamC (1:500)(McKearin and Ohlstein, 1995). Secondary antibodies were purchased from The Jackson Laboratory and used at the manufacturer's recommended concentrations. Stained ovaries were mounted in 4% n-propyl gallate (Sigma) in 90% glycerol, 10% PBS. Images were collected on Olympus 3L Spinning Disc or Zeiss Meta 510 laser-scanning confocal microscopes.

Our first clue that Rab11 plays important roles in early oogenesis in Drosophila came from immunostaining experiments that revealed strong expression of endogenous Rab11 and a fully functional Rab11::GFP in GSCs,cystoblasts and young (2-4- and 8-cell) germline cysts(Fig. 1). Strikingly, the proteins were concentrated as discrete dots on the fusome(Fig. 1E-L), which electron microscopy and photobleaching studies have shown is highly vesicular and rapidly exchanged with other membrane stores(Mahowald, 1972; Snapp et al., 2004). Triple-stain experiments showed that some of these dots also contained E-cadherin (Fig. 1B-E), which has been shown to transit though Rab11-positive recycling endosomes (REs) en route to the plasma membrane in some cells(Lock and Stow, 2005; Langevin et al., 2005). High-magnification images showed that the Rab11 (and, more rarely, E-cadherin)dots were often nestled into cavities within the fusome(Fig. 1D,E). Such Rab11-harboring cavities were visible in the fusomes of all examined GSCs,cystoblasts and young germline cysts, not only in the ovary but also in the testes (Fig. 1; data not shown). In view of the well-described enrichment of Rab11 in REs(Dollar et al., 2002; Emery et al., 2005; Lock and Stow, 2005; Riggs et al., 2003), we propose that these Rab11- and E-cadherin-harboring cavities are REs and will hereafter refer to them as FREs (fusome-associated REs).

Previous studies of hypomorphic rab11 alleles revealed a role for the gene in polarizing anterior-posterior axis of the mid-stage oocyte(Dollar et al., 2002; Jankovics et al., 2001). To investigate the role of Rab11 during early oogenesis, we set out to examine a rab11-null allele. The one null allele available at the start of these studies proved to be tightly linked to a second site cell-lethal mutation, so we made a new one using the FRT-flipase method(Parks et al., 2004). This new allele, called rab11^ΔFRT,deletes the rab11 promoter and the first two exons of the gene, and produces no detectable protein (Fig. 1I).

Because rab11^ΔFRT is homozygous lethal, we used the FRT-FLP system(Xu and Rubin, 1993) to generate homozygous rab11-null clones that were marked by loss of nuclear (n) GFP. Consistent with a role for Rab11 in the maintenance of GSCs,we recovered a disproportionately small number of rab11-null GSCs compared with rab11-null germline cysts. To determine the half-life of rab11-null GSCs, we calculated the percentage of GSCs that were rab11-null as a function of days after clone induction (ACI). As a control, we made identical calculations for marked clones carrying the wild-type rab11 allele. Such studies revealed a half-life of 4.0 days for rab11- null GSCs, or ∼4-fold less than wild type(Table 1). We also made clones with the rab11²¹⁴⁸ hypomorphic allele and calculated a near wild-type half-life of 15.9 days(Table 1). This was the expected result as this allele, which contains a P-element insertion in the first intron, produces apparently normal amounts of Rab11 protein during early oogenesis (Dollar et al.,2002). We conclude from these data that rab11 is required to maintain GSC identity.

Consistent with previous findings that lost GSCs can be replaced(Kai and Spradling, 2003; Kai and Spradling, 2004), many of the germaria that had lost a rab11-null GSC contained a full complement (two or three) of wild-type GSCs. One apparent replacement event is shown in Fig. 2A,B, in which a wild-type GSC is dividing along an axis parallel to the niche and just anterior to a displaced rab11-null GSC.

To determine whether the observed defects in GSC maintenance reflect a requirement for Rab11 in E-cadherin trafficking, we compared the distribution of E-cadherin in wild-type and rab11-null GSCs. In contrast to wild-type GSCs (Fig. 2C,D,white arrows), we found little or no E-cadherin along the anterior surface(i.e. at the GSC-cap cell interface) of rab11-null GSCs(n=9) 8-10 days ACI (Fig. 2D, yellow arrow). Similar analyses of germaria 2.2 days ACI revealed reduced or no accumulation of E-cadherin along the anterior cortex of 16 of 22 rab11-null GSCs examined(Fig. 2E). Consistent with the idea that such reductions reflect a loss of adherens junctions, we saw similar strong reductions of Armadillo (β-catenin) (data not shown). Concomitant with its reduction along the anterior surface of the GSC, increased amounts of E-cadherin (seen as discrete dots) were detected on the fusomes/FREs of rab11- null GSCs (Fig. 2E, yellow arrow). Thus, whereas wild-type GSCs contained an average of 0.16 dots of E-cadherin per fusome (n=31), rab11-null GSCs contained an average of 1.6 dots per fusome(n=17) (Table 1). We conclude that rab11 is required for the maintenance of adherens junctions between cap cells and GSCs and propose that such maintenance involves the trafficking of intracellular E-cadherin, and possibly other cargoes, from the FRE to the anterior surface of the GSC.

Although the simplest interpretation of the above data is that Rab11 maintains GSC identity through E-cadherin trafficking, we cannot rule out the possibility that the primary role of Rab11 is that of recycling Dpp or other signals required for GSC maintenance, and that the observed defects in E-cadherin trafficking are a secondary effect of insufficient signaling. To test this idea, we immunostained mosaic germaria for Bam, whose expression is negatively regulated by Dpp (Xie and Spradling, 1998; Xie and Spradling, 2000; Song et al.,2004). Such studies revealed a normal pattern of Bam expression;Bam was not detected in rab11-null GSCs or cystoblasts, but was detected in young (2- to 8-cell) germline cysts(Fig. 2G). These data argue strongly against the idea that the primary role of rab11 is that of facilitating Dpp signaling, in which case rab11-null GSCs would be expected to move out of the niche only after they have activated Bam. We conclude from these findings that Rab11 does not affect GSC maintenance or E-cadherin trafficking through regulation of Dpp or other signals that maintain GSC identity via Bam repression.

The rab11-null clones also revealed a role for Rab11 in the subcellular localization and behavior of the fusome. In wild-type GSCs(Fig. 2F, white arrowhead), the fusome is anchored to the anterior cortex and spreads out along the mitotic spindle such that about one-third of it extends into, and is ultimately donated to, the cystoblast (Deng and Lin,1997). By contrast, the fusome of rab11- null GSCs(Fig. 2F, yellow arrow) was not anchored to the anterior cortex and, although it spread out along the mitotic spindle during cell division, it was often splayed and generally much less organized. We suspect such splaying results from the detachment of the fusome from the cell cortex, but cannot rule out a more direct role for Rab11 in fusome segregation. Although it is not known how the fusome is attached to the anterior cortex of the GSC, it is likely to involve association with intracellular domains of the adherens junctions(Song et al., 2002). If so,Rab11, the fusome/FRE and the adherens junctions might comprise a tripartite feedback loop whereby each reinforces the subcellular localization/behavior of the other. Specifically, we propose that the association of Rab11 with the fusome/FRE polarizes the trafficking of E-cadherin by Rab11 toward the GSC-cap cell interface, in turn reinforcing GSC-cap cell junctions, fusome localization and asymmetric cell division.

All rab11-null germline cysts arrested development by stage 6 and were of two phenotypic classes. The rarer (∼10%), more severely affected class arrested development in region 1 of the germarium, often contained less than 16 cells, and had little or no fusome(Fig. 3A,B,D, yellow outlines). Given the splayed fusome phenotype of dividing rab11-null GSCs described above, we speculate that this early arrest reflects a role for Rab11 in faithful segregation of the fusome to daughter cystoblasts. Consistent with this idea, mutations in α-Spectrin and hu li tai shao,which encode components of the fusome, cause a similar early arrest of cyst development (Lin et al., 1994; de Cuevas et al., 1996).

The less affected class of rab11-null germline cysts elaborated a normal fusome (Fig. 3A, white dashes), but contained clumped ring canals(Fig. 3F,G) and arrested development at ∼stage 6. Clumped ring canals have also been reported for sec5, sec6 and rab6 mutations and have been interpreted to reflect a requirement for these genes in bulk membrane trafficking to the cell surface (Murthy and Schwarz,2003; Murthy et al.,2005; Coutelis and Ephrussi,2007). A similar requirement for Rab11 is likely as many of the nuclei of rab11-null germline cysts were clumped together or otherwise poorly spaced (not shown). These cysts also exhibited defects in oocyte positioning. Thus, whereas the oocyte is positioned at the posterior end of wild-type germline cysts in germarial region 2B(Fig. 3E, white arrowhead), the oocytes of rab11-null germline cysts were often in the center(Fig. 3E′, yellow arrow). Previous studies (Godt and Tepass,1998; Gonzales-Reyes and St Johnston, 1998) have shown that oocyte positioning is dependent on enriched accumulation of E-cadherin along the posterior surface of the oocyte. Consistent with a role for Rab11 in such enrichment, we observed reduced accumulation of E-cadherin along the posterior surface of rab11-null oocytes (Fig. 3D, yellow arrow)compared with wild-type oocytes (Fig. 3C,D, white arrowheads) in region 2B and region 3 germline cysts. Nevertheless, it is difficult to conclude whether the observed defects in oocyte positioning in the rab11-null germline cysts reflects a role for Rab11 in E-cadherin trafficking, in bulk membrane trafficking, or both.

Our studies indicate that Rab11 maintains GSC identity through polarized trafficking of E-cadherin and, possibly, other cargoes that reinforce essential GSC-niche contacts. Our studies further indicate that Rab11 is required for fusome localization and asymmetric GSC division and suggest a feedback linkage between these events and E-cadherin trafficking. Although Rab11 has been implicated in the trafficking of E-cadherin in other cells, we know of no other cases in which such trafficking has been correlated with a biological response. It will be of interest to determine whether Rab11 is required for the maintenance of stem cells in other systems and whether such maintenance involves E-cadherin trafficking or the trafficking of other adhesion molecules. It will also be of interest to determine the role of Rab11 in other E-cadherin-dependent cell behaviors, particularly as Rab11, at least in Drosophila, is expressed in only a small subset of E-cadherin-expressing cells (J.X. and R.S.C., unpublished).

We thank Sui Zhang for excellent technical assistance; Cristin Gustafson with help in the construction of the rab11-null allele; Vicki Corbin for comments on the manuscript; Don Reddy, Ting Xie, Lynn Cooley, Ruth Lehman,Hugo Bellen, the Bloomington Stock Center and Doug Dimlich for antibodies and fly stocks; and David Moore for excellent help with confocal microscopy. This work was support by NIH grant R01 GM068022-01 to R.S.C.