S6K links cell fate, cell cycle and nutrient response in C. elegans germline stem/progenitor cells

The nutritional status of an animal has many implications. Consequences of poor nutrition include growth and reproductive deficiencies in organisms ranging from C. elegans to humans (e.g. Bongaarts, 1980; Greer and Brunet, 2009). Dietary restriction can also extend lifespan and reduce susceptibility to age-related diseases, such as diabetes and certain cancers (Colman et al., 2009; Kritchevsky, 1999; Rous, 1914; Tannenbaum and Silverstone, 1953). Recent studies suggest that specific signaling pathways mediate the cellular effects of changes in diet. For example, although dietary restriction can deter tumor proliferation in some models, tumors with elevated PI3K activity are insensitive to growth-inhibitory effects of dietary restriction (Kalaany and Sabatini, 2009). Therefore, understanding the molecular mechanisms that underlie the effects of diet on development, cell proliferation and reproduction has broad implications.

TOR is a serine/threonine kinase and a conserved regulator of cell growth and proliferation in response to nutritional and growth factor cues (reviewed by Hietakangas and Cohen, 2009; Russell et al., 2011; Wang and Proud, 2006; Wang and Proud, 2009; Wullschleger et al., 2006). Interestingly, although TOR acts downstream of insulin/IGF/PI3K signaling in certain contexts, the two pathways can also have independent functions. TOR participates in a complex (TORC1) with the Regulatory associated protein of TOR (RAPTOR) to promote growth when nutrients are plentiful. Two well-characterized TORC1 targets, p70 ribosomal S6 kinase (p70S6K) and the eukaryotic translation initiation factor (eIF4E)-binding protein 4E-BP1, link TORC1 to translational control. Of these, ribosomal protein S6 kinase (S6K) has been most clearly implicated in cell and organismal growth.

Stem cells are important targets for metabolic control, as they must be tightly regulated to properly establish and maintain stem cell pools and tissue homeostasis in response to changing physiological demands (reviewed by Drummond-Barbosa, 2008). The C. elegans germ line is maintained by a pool of proliferating progenitors (stem cells and their progeny) (reviewed by Hansen and Schedl, 2006; Hubbard, 2007; Kimble and Crittenden, 2007). This system offers a genetically tractable framework to study the effects of nutrition on stem cell proliferation and differentiation in the context of a whole animal (Korta and Hubbard, 2010). The somatic distal tip cell (DTC) serves as the niche for germ cells, maintaining the proliferative germ cell fate by producing ligands for the receptor GLP-1 (Notch) on neighboring germ cells. In addition, insulin/IGF-like receptor (IIR) signaling is required for robust larval germline proliferation to generate an appropriate progenitor pool for optimal fecundity (Michaelson et al., 2010). In C. elegans, homologs of TOR pathway genes include let-363 (TOR), daf-15 (RAPTOR) and rsks-1 (S6K). Reduction- or loss-of-function of these genes leads to lifespan extension (Jia et al., 2004; Pan et al., 2007; Selman et al., 2009; Vellai et al., 2003). Furthermore, loss of let-363 or daf-15 leads to larval developmental arrest (Jia et al., 2004; Long et al., 2002) and loss of rsks-1 causes reduced body size and smaller broods (Pan et al., 2007; Selman et al., 2009). Obvious sequence homologs of the TOR inhibitors TSC1/2 (Inoki et al., 2002) and of 4E-BP have yet to be identified in the C. elegans genome, although five genes (ife-1-5) encode eIF4E orthologs (Keiper et al., 2000).

Here we show that rsks-1 is required germline-autonomously for the establishment of the proper number of germline progenitors during development and that this role requires a conserved TOR phosphorylation site. We find that rsks-1 both promotes cell cycle progression and inhibits differentiation. A reduction of TOR or RAPTOR homologs causes a more severe germline defect, and rsks-1 and ife-1 appear to mediate the bulk of these effects. Genetic interactions are consistent with rsks-1 acting similarly to glp-1 (Notch) in that loss of rsks-1 enhances and suppresses phenotypes associated with reduced and elevated glp-1 activity, respectively. Our results are also consistent with rsks-1 acting in parallel with both glp-1 and daf-2 (IIR). Surprisingly, genes that mediate the effects of rsks-1 on longevity do not similarly affect the germ line. Finally, we find that dietary restriction strongly reduces the number of proliferative germ cells in wild type and in glp-1 mutants, but not in rsks-1 mutants.

Strains (supplementary material Table S1) were derived from N2 wild type (Bristol) and handled using standard methods (Brenner, 1974). Unless otherwise indicated, worms were grown on OP50 at 20°C. For plasmids and construction details, see supplementary material Table S2. Plasmids constructed for this study were pGC479, pGC480, pGC520, pGC609, pGC610.

Synchronization by hatch-off, timecourse analysis and DAPI staining were performed as described (Pepper et al., 2003a). Microscopy, developmental staging, determination of nuclei number in the proliferative zone, apoptosis (SYTO 12) analysis, distance to transition zone, mitotic index, S-phase index by 5-ethynyl-2′deoxyuridine (EdU) labeling, and DNA quantification were as described (Michaelson et al., 2010). In addition, some proliferative zone counts were semi-automated using a modified ImageJ plug-in originally written and subsequently modified at our request by Vytas Bindokas at the University of Chicago Integrated Microscopy Core. For proximal tumors, counts of undifferentiated nuclei proximal to the proximal-most differentiated nuclei were obtained manually. For gld-1 RNAi-induced tumors, N2 and rsks-1(sv31) animals were grown at 20°C, synchronized by hatch-off, fed gld-1 RNAi-inducing bacteria, and scored as adults (at 48 hours post-mid L4 at 20°C). For glp-1 proximal tumors, glp-1(ar202) and glp-1(ar202) rsks-1(sv31) animals were grown at 15°C and synchronized. L1 larvae were moved to 25°C and scored at the early adult stage [defined as just after the L4/adult molt (Michaelson et al., 2010)].

Brood size and reproductive period were determined as described (Dillin et al., 2002). For embryo survival (hatched larvae/total eggs) the setup was as in the brood size experiment except that the number of eggs on each plate was counted upon adult transfer, and hatchlings were counted after 2 days. For mating experiments, individual rsks-1(sv31) hermaphrodites were placed on separate plates with three N2 males, transferred every 24 hours, and progeny counted; older males were replaced with young males every 2 days. For sperm counts, animals were DAPI stained at the adult molt (N2, ∼54 hours post-hatch) or as adults [rsks-1(sv31), ∼64 hours post-hatch to compensate for later initiation of sperm production; see supplementary material Fig. S1]. Sperm and spermatocytes were counted; each spermatocyte was scored as four sperm. For developmental timing in supplementary material Fig. S1, N2 and rsks-1(sv31) were synchronized by a 1-hour hatch-off. Every 4 hours (36-80 hours post-hatch), animals were staged (vulval morphology, DIC microscopy), DAPI stained and scored for germ cells at the stages indicated.

Animals were grown to early adult, immobilized in levimasole and mounted on agar pads. Body length was measured on images captured at 50× magnification. Cell size was determined as described previously (Michaelson et al., 2010), except that z-stacks (images captured 0.5 μm apart) were captured at 630×, and germ cell volume was estimated by measuring the largest cell area in the z-stack (Zeiss AxioVision) and calculating cell volume as a sphere (4/3 π r³), with radius extrapolated from the measured area (π r²). This method correlated very well with the method described previously.

RNAi by bacterial feeding was performed as described (Timmons et al., 2001). The empty RNAi expression vector L4440 in HT115 bacteria served as a negative control. For let-363, daf-15 and ife-3, RNAi was performed at 25°C, and progeny from the first ∼3 hours of egg laying (collected ∼12 hours post-L4) were scored at the early adult stage.

Immunohistochemical analysis was carried out as described (Pepper et al., 2003b) using rabbit polyclonal anti-GFP (Abcam ab6556, 1:2000 dilution) and Alexa Fluor 488 goat anti-rabbit IgG (H+L) (Molecular Probes A11034, 1:1000).

OP50 bacteria were grown at 37°C for 16 hours to saturation. The bacterial concentration at saturation was determined by serial dilution on LB plates, and corresponded to 1×10⁹ cfu/ml. Bacteria were resuspended in S Basal and diluted (1×10⁸ and 1×10⁷ cfu/ml) or concentrated (1×10¹⁰ cfu/ml). Bacteria (400 μl) were seeded onto 60 mm NGM plates containing 60 μg/ml ampicillin. S Basal and NGM media were as described (Stiernagle, 2006). After plates were dried in the hood for ∼30 minutes, 6-8 gravid adult worms were spot-bleached in 20% alkaline hypochlorite solution directly on the plates (outside the bacterial lawn) to synchronize progeny and avoid bacterial transfer. Progeny of bleached adults were scored as early adults. At 1×10⁷ cfu/ml, animals arrested at the L1 or dauer stages.

rsks-1, which encodes the sole worm homolog of mammalian p70S6K, is required for normal brood size in C. elegans, suggesting a role in reproduction (Fig. 1A) (Pan et al., 2007). Since many factors can contribute to brood size, we examined several aspects of reproduction in the rsks-1(sv31) null mutant. The most striking phenotype was a failure to establish the proper number of germline progenitors (that is, proliferative germ cells located distal to the position of meiotic entry). Under normal laboratory growth conditions, germline progenitors rapidly accumulate during the third and fourth larval stages (L3 and L4) to establish an adult pool of ∼200 progenitors (Killian and Hubbard, 2005). In stage-matched rsks-1(sv31) or rsks-1(ok1255) mutant animals, this accumulation reached only ∼50% of wild-type levels (Fig. 1B-D; supplementary material Fig. S1A, Table S3). By comparison, deletion of rskn-1, which encodes the homolog of p90S6K, had no effect (Fig. 1C; supplementary material Table S3). Additionally, rsks-1 affects sperm production and embryo viability (supplementary material Fig. S1, Table S3). Together with the progenitor defect, these phenotypes are likely to account for the reduced fecundity. Our subsequent experiments focused on the role of rsks-1 in the regulation of germline progenitors.

Three independent lines of evidence support the conclusion that rsks-1 acts within the germ line to control larval stage accumulation of germline progenitors. First, rsks-1 RNAi caused a reduction in the progenitor pool independent of rrf-1, which is required for optimal somatic RNAi (Fig. 1E; supplementary material Table S3) (Sijen et al., 2001). Second, restoring rsks-1(+) in somatic cells did not rescue the germline defect (Fig. 1F; supplementary material Table S3), although it did rescue body length (Fig. 1G; supplementary material Table S3). Third, expressing rsks-1(+) solely in the germ line largely rescued the germline progenitor defect (Fig. 1F; supplementary material Table S3).

We examined three possible mechanisms by which rsks-1 might promote germline progenitor accumulation: cell survival, cell cycle, and the balance between proliferation and differentiation. We found no SYTO 12-positive progenitor cells (n>65) in the distal germ line of rsks-1(sv31) larvae or adults, although apoptotic SYTO 12-positive female germ cells were detected in the loop region, as expected (Gumienny et al., 1999). Nor did we observe debris in the proliferative region that might indicate other forms of cell death. Therefore, inappropriate cell death is unlikely to account for the failure to expand the germline progenitor pool in the rsks-1 mutant.

When we examined the cell cycle we found that the larval, but not adult, germline progenitor cell cycle was delayed in rsks-1(sv31). Specifically, at the mid-L4 stage, the mitotic and S-phase indexes were significantly reduced in rsks-1(sv31) relative to wild type (Fig. 2A,B), whereas both were unchanged in adults (supplementary material Fig. S2A,B). To determine whether these reduced indexes were accompanied by a specific delay in G1 or G2 we measured DNA content, and found no difference between the distributions of rsks-1(sv31) and wild-type germline progenitor cells within each cell cycle phase (Fig. 2C). These results suggest that rsks-1 promotes the larval but not adult germline progenitor cell cycle.

Our finding that rsks-1 promotes the larval but not adult germ cell cycle provided an opportunity to further investigate the hypothesis that cells in adult germline tumors behave like larval germline progenitors (Michaelson et al., 2010). If adult tumor cells were sensitive to rsks-1, this would support the notion that they are regulated similarly to larval progenitors. First, we looked at tumors caused by the weak gain-of-function (gf) allele glp-1(ar202). These tumors arise from germ cells that fail to differentiate in a timely manner (McGovern et al., 2009; Pepper et al., 2003a). We found that proximal germline tumors in adult glp-1(ar202) rsks-1(sv31) double-mutant animals contained 76% fewer cells (Fig. 2D,E; supplementary material Table S4) and displayed a reduced mitotic index (Fig. 2F) relative to glp-1(ar202) tumors, similar to the difference between rsks-1(sv31) and wild type (Fig. 2A).

We also investigated the effect of loss of rsks-1 on gld-1(RNAi)-derived tumors. gld-1 encodes an RNA-binding tumor suppressor protein that is required for proper germ cell differentiation (Francis et al., 1995a; Francis et al., 1995b; Jones and Schedl, 1995). Reduction of gld-1 causes germ cells to exit the early meiotic pathway and return to the mitotic cell cycle, forming a proximal tumor in adulthood. Despite the difference in tumor etiology, we found that, like glp-1(ar202) proximal tumors, loss of rsks-1 also reduced the number of cells in gld-1(RNAi) tumors by 73% (supplementary material Fig. S2C, Table S4). We conclude that rsks-1 regulates germ cell proliferation in both normal larval germ lines and adult germline tumors, supporting the hypothesis that adult proximal germline tumors share features with larval germline progenitors.

In addition to its role in cell cycle progression, we found that rsks-1 affects the balance between germ cell proliferation and differentiation. We noted that the position of meiotic entry is shifted distally in rsks-1(sv31) animals at all stages examined, reminiscent of mutants with reduced glp-1 (Notch) activity (Michaelson et al., 2010). This phenotype becomes most pronounced in rsks-1 adults (25%, 31% and 47% reduction at the mid-L4, early adult and 24 hours post-L4, respectively; Fig. 3A). This result raised the possibility that, in addition to its role in the cell cycle, rsks-1 could affect differentiation, similar to the dual role reported for the cyclin E ortholog cye-1 (Fox et al., 2011).

To explore this hypothesis, we asked whether loss of rsks-1 enhanced the phenotype of a glp-1 reduction-of-function (rf) mutant to an ‘all meiotic’ Glp-1 phenotype (in which all germ cells have entered meiosis) characteristic of strong loss of glp-1 activity. glp-1(e2141) is a temperature-sensitive rf allele that exhibits reduced germline progenitors at the semi-permissive temperature (20°C; 94.7±3.3 versus 209.0±3.6 in wild type; n=22, 33) and a severe Glp-1 phenotype (all germ cells differentiate) at the restrictive temperature (25°C) (Priess et al., 1987). In the double mutant glp-1(e2141) rsks-1(sv31) at 20°C, we found a striking exacerbation of the penetrance of the Glp-1 ‘all meiotic’ phenotype. Whereas virtually all rsks-1(sv31) or glp-1(e2141) single mutants have at least 50 germline progenitors, 60% of glp-1(e2141) rsks-1(sv31) double mutants display the severe Glp-1 phenotype and the remaining 40% contain only 1-20 progenitors (Fig. 3B,C; supplementary material Fig. S3). We also treated glp-1(e2141) and rrf-1(pk1417); glp-1(e2141) with rsks-1 RNAi, and found that rsks-1 RNAi enhanced the Glp-1 phenotype in both strains (supplementary material Fig. S3), suggesting that the effect is germline autonomous. In addition to enhancing the phenotype caused by a reduction of glp-1 activity, loss of rsks-1 suppressed the penetrance of a glp-1(gf) mutant phenotype: the Pro phenotype (percentage of gonad arms that form proximal tumors at 25°C, as distinct from the number of cells within the tumors, above) was reduced from 100% in glp-1(ar202) to 75% in glp-1(ar202) rsks-1(sv31) (Fig. 3D). Animals without tumors produced viable progeny. Closer examination of the double mutant revealed a correlation between the delay in differentiation and suppression of tumor initiation (Fig. 3D,E). Thus, reducing rsks-1 activity enhances phenotypes caused by reduced glp-1 activity and suppresses phenotypes caused by elevated glp-1 activity. Taken together, we conclude that rsks-1 acts similarly to glp-1 to promote the proliferative fate and/or inhibit the differentiated fate.

We reasoned that if rsks-1 requires the presence of GLP-1 (Notch), then loss of rsks-1 should have no effect when glp-1 activity is removed. Conversely, if rsks-1 does not require GLP-1, then loss of rsks-1 should still affect the germ line in the absence of GLP-1. We employed a triple null mutant gld-2(q497) gld-1(q485); glp-1(q175), in which the vast majority of germ cells are in the proliferative state in the absence of GLP-1 (Hansen et al., 2004). We observed a 29% reduction (Fig. 3F,G; supplementary material Table S5) in the number of progenitors in rsks-1(RNAi) versus control RNAi, which is comparable to the 26% decrease observed in the wild type treated with rsks-1(RNAi) (Fig. 1E; supplementary material Table S3). However, this reduction in cell number was not associated with obvious meiotic entry (Fig. 1E; data not shown), suggesting that the cell cycle-promoting role of rsks-1 acts in parallel with GLP-1 and downstream of GLD-1 and GLD-2. The role of RSKS-1 with respect to meiotic entry is more difficult to place within this pathway. Although rsks-1(RNAi) enhancement of the meiotic entry defect of glp-1(rf) is likely to be germline autonomous (supplementary material Fig. S3) and is therefore unlikely to be upstream of glp-1, it might not cause a sufficiently strong meiotic entry phenotype to reveal a clear epistatic relationship. These results suggest that the two functions of rsks-1 in the proliferative region, i.e. cell cycle and cell fate control, might be separable.

Previously, we showed that daf-2 (IIR) signaling promotes larval germline proliferation and that this effect depends on the activities of daf-18 (PTEN) and daf-16 (FOXO) (Michaelson et al., 2010). We therefore investigated the relationship between the daf-2 pathway and rsks-1. We found that rsks-1 regulation of germ cell accumulation does not depend on daf-16 or daf-18, as neither daf-18(RNAi) nor daf-16(–) suppresses the rsks-1 germline defect (Fig. 4A; supplementary material Fig. S4B, Table S6).

We tested the alternative hypothesis that daf-16 negatively regulates rsks-1. This hypothesis is supported by a report of transcriptional repression of daf-15 (RAPTOR) by daf-16 (Jia et al., 2004). If daf-16 negatively regulates daf-15 upstream of rsks-1, then eliminating daf-16 activity should elevate the level of daf-15 and might ameliorate the effects of a partial loss of rsks-1. We tested this possibility using rsks-1(RNAi) because it reduces, but does not eliminate, rsks-1 activity, and we found no difference in the phenotype in the presence or absence of daf-16 (Fig. 4B; supplementary material Table S6), suggesting that daf-16 does not negatively regulate rsks-1 in this context.

We then considered whether rsks-1 and daf-2 might affect larval germline progenitor accumulation independently. If so, the daf-2(e1370) rsks-1(sv31) double-mutant phenotype should be more severe than that of either single mutant. Indeed, this is what we observed. At both 20°C and after an L3 shift to 25°C, the daf-2 rsks-1 double mutant contained significantly fewer germline progenitors than either single mutant (Fig. 4C; supplementary material Fig. S4A, Table S6).

Finally, based on studies in Drosophila showing that reduction in either insulin signaling or p70S6K activity reduces cell size (reviewed by Hietakangas and Cohen, 2009), we examined germ cell size. We found that whereas cell size was not reduced in the daf-2 mutant, loss of rsks-1 led to a significant reduction in germ cell size (Fig. 4D; supplementary material Table S6). We conclude that rsks-1 and daf-2 have different and likely independent effects on germ cell progenitors (see Discussion).

In other systems, S6K is a direct substrate for the TOR kinase. TOR phosphorylates a highly conserved threonine residue in Drosophila and human S6K (Schalm and Blenis, 2002) (supplementary material Fig. S3A). We generated a transgene that alters this T404 residue. Despite similar germline expression levels between wild-type T404 and the T404A mutant transgenes (supplementary material Fig. S3B,C), the T404A transgene did not rescue the germline phenotype of rsks-1(sv31) (Fig. 5A; supplementary material Table S7). Thus, phosphorylation on T404, presumably by LET-363 (TOR), is necessary for the proper establishment of the germline progenitor pool.

Animals bearing null mutations in let-363 (TOR) or daf-15 (RAPTOR) undergo arrest in L3, precluding direct study of germline progenitor accumulation (Fig. 5B,D). We therefore identified conditions that reduce their activities without interfering with overall development. Although the progeny of mothers fed with either dsRNA starting at mid-L4 undergo arrest (Fig. 5B,D), mothers fed from the L4/adult molt produce ‘escapers’ that develop into adults with 57% fewer germline progenitors relative to controls (supplementary material Fig. S3D, Table S7). In addition, in the rrf-1 mutant that attenuates somatic but not germline RNAi, progeny of animals fed let-363 and daf-15 RNAi starting in mid/late L4 exhibit sterility with a severe (82%) reduction in the number of germline progenitors (Fig. 5B-E; supplementary material Table S7), while fewer than 5% of these progeny arrest. The let-363 and daf-15 RNAi phenotypes are very similar, consistent with TOR and RAPTOR acting together, as they do in other organisms. Moreover, let-363(RNAi) reduced the mitotic index compared with controls (supplementary material Fig. S5E), and caused a statistically significant delay in the G2 phase of the cell cycle (supplementary material Fig. S5F). This G2 delay differs from what we observed in rsks-1(sv31), which slowed the cell cycle but did not alter the distribution of germ cells within each cell cycle phase (Fig. 2C). Since daf-2 (IIR) signaling causes a delay in G2 in a manner dependent on daf-16 (FOXO) (Michaelson et al., 2010), we wondered whether the let-363(RNAi) phenotype could be partially suppressed by loss of daf-16. We found that rrf-1; let-363 RNAi caused the same defect in the presence or absence of daf-16, and loss of daf-16 did not suppress the developmental arrest phenotype of let-363 RNAi. Thus, neither the germline nor the somatic arrest phenotypes of let-363 depend on daf-16.

RNAi depletion of let-363 or daf-15 caused a more severe defect in the germline progenitor pool than loss of rsks-1, suggesting that rsks-1 mediates most but not all of the effects of TOR/RAPTOR on germline progenitor accumulation. An alternate candidate substrate of TOR is 4E-BP. When phosphorylated by TOR, 4E-BP cannot bind and inhibit the activity of eIF4E (Wang and Proud, 2006). Therefore, TOR activity promotes the activity of eIF4E. In the absence of obvious sequence homologs of 4E-BP in C. elegans, we examined five genes that encode eIF4Es (ife-1-5) (Keiper et al., 2000). We first assayed the effects of each gene individually on the size of the proliferative germ cell pool, using viable null alleles, with the exception of the essential ife-3 for which we used RNAi. Loss of ife-1 strongly reduced the number of proliferative germ cells (supplementary material Fig. S5G, Table S7), consistent with its expression in the larval germline proliferative zone (Amiri et al., 2001). Further, germ lines of rsks-1(sv31) ife-1(bn127) double mutants averaged 53 progenitors cells, which is significantly fewer than either single mutant (supplementary material Fig. S5H, Table S7). Since reduction of let-363 by RNAi results in ∼30 proliferative germ cells, ife-1 and rsks-1 could account for much of the let-363 effect on germ cell accumulation. We also saw a small but statistically significant effect from ife-3 and ife-4 on the size of the proliferative germ cell pool (supplementary material Fig. S5G), suggesting that these genes contribute as well.

In C. elegans, lifespan extension conferred by loss of rsks-1 depends on the activity of pha-4 (FOXA), egl-9 and aak-2 (AMPK) (Chen et al., 2009; Mango et al., 2008; Selman et al., 2009). We tested whether the rsks-1 defect in the establishment of germline progenitors also depended on these genes, and we found that it did not. Neither rsks-1 mutant animals subject to pha-4 RNAi from the L1 stage (to avoid embryonic lethality) nor viable escapers from late L4 maternal RNAi displayed any restoration in the number of proliferative germ cells (Fig. 6A; supplementary material Table S8; for escapers 78.5±4.3 versus 79.3±4.1 in the control; n=10, 13). Similarly, loss of egl-9, a dioxygenase that negatively regulates hypoxia inducible factor (HIF-1), did not suppress the rsks-1 mutant germline defect (Fig. 6B; supplementary material Table S8). Next, we examined the rsks-1 germline progenitor phenotype in an aak-2 mutant; aak-2 encodes a C. elegans homolog of the catalytic alpha subunit of AMP-activated protein kinase (AMPK). We saw no suppression of the rsks-1 phenotype upon reduction of aak-2 activity (Fig. 6C; supplementary material Table S8, see also Fig. S6A,B for additional combinations with aak-1). Finally, we saw no difference in the brood sizes of the rsks-1(sv31); aak-2(ok524) double mutant compared with the rsks-1 single mutant [see correction to Selman et al. (Selman et al., 2009)] (supplementary material Fig. S4B). Taken together, we conclude that unlike its role in lifespan control, rsks-1 does not act through pha-4, egl-9 or aak-2 in promoting the accumulation of germline progenitors.

In cell culture assays, TOR-S6K pathway activity is responsive to nutritional signals, especially intracellular amino acids (Wang and Proud, 2009). To determine whether TOR-S6K pathway activity links nutritional signals to germline progenitors in C. elegans, we first examined the effect of dietary restriction (DR) on the germ line in wild type, and then tested its effects in the presence and absence of rsks-1. Bacterial dilution had a striking effect on the number of adult germline progenitors: wild-type animals grown on 1×10⁸ cfu/ml bacteria displayed 76% fewer progenitors than siblings grown on 1×10⁹ cfu/ml (215 versus 52; Fig. 7A; supplementary material Table S9). Consistent with the possibility that rsks-1 mediates this germline response, rsks-1 mutants grown on 1×10⁸ cfu/ml contained only 20% fewer germline progenitors than siblings raised on 1×10⁹ cfu/ml (Fig. 7A; supplementary material Table S9). This attenuated response is not simply a consequence of a reduced starting progenitor pool as glp-1(rf) animals contain 67% fewer progenitors at 1×10⁸ than at 1×10⁹ cfu/ml, similar to the wild-type response. We also examined a genetic model of DR, eat-2(ad1116), in which food intake is limited by decreased pharyngeal pumping (Raizen et al., 1995). The size of the germline progenitor pool in the eat-2 mutant was ∼50% of wild type (supplementary material Fig. S7). Again, this was attenuated (14%, as opposed to ∼50%) in the absence of rsks-1 (supplementary material Fig. S7). Together, these data are consistent with a model in which the effect of DR (via general nutrient limitation) on larval germline progenitors is largely mediated by rsks-1.

We further examined whether DR by bacterial dilution affects the cell cycle and cell fate in germline progenitors. We saw a significant effect on the mitotic index (Fig. 7B), but we did not see a strong effect on cell fate. Despite a severe reduction in cell numbers (Fig. 7A), none of the glp-1(rf) animals raised on 1×10⁸ cfu/ml displayed an ‘all meiotic’ Glp-1 phenotype (n=104 arms). This interaction contrasts with the strong enhancement of this phenotype caused by loss of rsks-1 (Fig. 3B; supplementary material Fig. S3), and suggests that DR predominately influences the progenitor cell cycle.

To test whether dietary amino acid levels regulate germline progenitors and to what extent this occurs through rsks-1, we used a genetic approach. In organisms ranging from humans to worms, intestinal transporters enable amino acid uptake (as either free amino acids or as di-/tri-peptides) from the diet (Daniel et al., 2006). PEPT-1, the C. elegans ortholog of mammalian PEPT1, acts as an oligopeptide transporter (Fei et al., 1998; Meissner et al., 2004). We found that adult pept-1 null mutants contain significantly fewer germline progenitors than wild type. Importantly, similar to DR, additional removal of rsks-1 does not exacerbate the defect, whereas glp-1(rf) does (Fig. 7C; supplementary material Table S9). Further, the mitotic index in pept-1 was modestly lower than in wild type (1.6% versus 2.2%; Fig. 7D) and, similar to DR, but contrasting with glp-1(rf); rsks-1, the severely reduced proliferative zone in glp-1(rf); pept-1(lg1601) nevertheless always contained undifferentiated progenitors (n=147). Lastly, unlike the pept-1 effect on lifespan (Spanier et al., 2010), the pept-1 germline phenotype is not dependent on the nuclear hormone receptor daf-12 (Fig. 7E). Consistent with a model in which insulin/IGF-like signaling and S6K act independently in the germ line, reducing daf-12 also did not suppress the rsks-1 mutant germline phenotype, although it suppressed the daf-2 germline phenotype (Fig. 7E; supplementary material Table S9) (D. Dalfó, D. Michaelson and E.J.A.H., unpublished). We conclude that the effect of dietary amino acid uptake as mediated by pept-1 on the proliferation of germline progenitors primarily influences cell cycle and is likely to act through rsks-1.

Our studies uncover a prominent role for the TOR-S6K pathway in the relationships between germline development, stem cell behavior and nutrition, and establish the C. elegans germ line as a model system with which to further explore these relationships. Our findings indicate that rsks-1 (S6K) is required germline-autonomously for larval accumulation of germline progenitors, promoting both cell cycle progression and the proliferative fate. Although certain aspects of the rsks-1 germline function are similar to the glp-1 (Notch) and the daf-2 (insulin receptor) pathways, rsks-1 acts largely in parallel to both. In addition, germline rsks-1 acts independently of genes known to influence the role of rsks-1 in lifespan control. We found that rsks-1 acts downstream of let-363 (TOR), partially mediating the effects of the TORC1 complex, likely in parallel with eIF4E. Finally, we show that DR, and especially protein deprivation, limits germline progenitor cell accumulation, and that rsks-1 plays a key role in this response. Fig. 7D presents a model of the influences of nutrition on germline progenitors. We propose that nutrient limitation influences cell cycle through rsks-1 but that rsks-1 influences cell fate independently. Distinct phospho substrates of RSKS-1 may mediate these differential effects of rsks-1.

TOR and S6K are responsive to PI3K signaling in other systems, and our previous results implicate the insulin/IGF-like-PI3K signaling pathway in the control of germ cell proliferation in C. elegans. Although our rsks-1 (S6K) studies employed null alleles, and the daf-2 (IIR) studies utilized strong hypomorphs (Michaelson et al., 2010), our results suggest that while reducing daf-2 or rsks-1 leads to fewer adult germline progenitors, the phenotypes are distinct in several respects. For example, although both influence the larval but not adult cell cycle, rsks-1 promotes overall cell cycle progression, whereas daf-2 is required for appropriate G2 progression. Additionally, loss of rsks-1 dramatically enhances the penetrance of the ‘all meiotic’ phenotype of a glp-1(rf) mutant, whereas reduced daf-2 does not, suggesting that S6K (but not IIR) inhibits differentiation. Moreover, rsks-1 and daf-2 have distinct roles in reproduction: at the semi-permissive temperature, daf-2 mutants have normal brood sizes but extended reproductive periods (Dillin et al., 2002), whereas at the restrictive temperature they have severely reduced brood sizes (Michaelson et al., 2010). By contrast, the rsks-1 mutant (which is not temperature sensitive) has a reduced brood size but a normal reproductive period (Fig. 1A). Genetically, rsks-1 does not act through daf-16 (FOXO), whereas daf-2 does, and the daf-2 rsks-1 double-mutant phenotype is more severe than either single mutant. rsks-1 is required for normal cell size whereas daf-2 is not. Thus, similar to the insulin pathway-independent role of TOR in Drosophila germline stem cells (GSCs) (LaFever et al., 2010) (see below), our data are consistent with independent roles for germline S6K and IIR. These model organism studies underscore the importance of examining relationships between these pathways in vivo.

We found that the role of rsks-1 in germline progenitors does not depend on the same genes that influence rsks-1 in lifespan control. Reduced germline progenitor proliferation has been correlated with extended lifespan (Arantes-Oliveira et al., 2002), and the germline phenotype of the rsks-1 mutant might influence longevity. Nevertheless, at least some of the genes required to elicit the lifespan-extension phenotype are not involved in the earlier generation of larval germline progenitors. This distinction further suggests that different targets of S6K are relevant in different phenotypic contexts.

We found that rsks-1 (S6K) acts downstream of let-363 (TOR) and daf-15 (RAPTOR) for proper larval expansion of the germline progenitor pool, but does not relay all of the germline functions of the complex. Consistent with a possible role for eIF4E acting downstream of TOR in parallel with S6K, we identified a prominent role for ife-1, which encodes an eIF4E paralog, in the regulation of germline progenitor accumulation. Additional studies are required to define this branch of the pathway, the cellular mechanisms by which it influences germline progenitor accumulation, and its precise relationship to TOR.

Recently, two laboratories demonstrated a role for TOR signaling in GSCs in the Drosophila ovary. Taken together with our results, the data suggest functional conservation of TOR-S6K signaling in the regulation of proliferation and differentiation. First, in Drosophila, TOR promotes GSC proliferation germline-autonomously, independent of insulin-like signaling through FOXO (LaFever et al., 2010). Similarly, rsks-1 (S6K) acts within the germ line to promote larval germline cell cycle progression independently of daf-16 (FOXO). However, whereas reducing TOR in both C. elegans and in Drosophila GSCs delays G2, loss of C. elegans rsks-1 does not cause any change in the distribution of cells in different stages of the cell cycle. Our results are more similar to findings in the Drosophila wing imaginal disc, where loss of S6K slows overall cell cycle but does not affect the distribution of cells in specific phases of the cell cycle (Montagne et al., 1999). It remains to be determined whether S6K has a similar effect in fly GSCs. Furthermore, rsks-1 promotes the proliferative (versus differentiated) fate in C. elegans. Similarly, TOR signaling influences germ cell differentiation in Drosophila, where, surprisingly, either reduced or elevated TOR activity results in GSC loss (LaFever et al., 2010; Sun et al., 2010). We hypothesize that the influence of TOR-S6K signaling on both the germ cell cycle and differentiation might be conserved between worms and flies. However, elevating TOR activity in the C. elegans germ line is not currently feasible, and no obvious homologs of TSC1 or TSC2 have yet been identified. Future studies are required to address this hypothesis.

The insulin and TOR pathways have been implicated in the effect of diet on stem cell proliferation in several contexts, including Drosophila GSCs (LaFever et al., 2010; Sun et al., 2010), intestinal stem cells (Amcheslavsky et al., 2009; Biteau et al., 2010) and neural stem cells (via relay from TOR in the fat body, which promotes insulin-like peptide production in glial cells to activate neuronal stem cells) (Chell and Brand, 2010; Sousa-Nunes et al., 2011). Thus, although some aspects of the cellular and molecular mechanisms linking diet to stem cells may be context specific, their dependence on the nutrient-responsive insulin-like and TOR signaling pathways is likely to be conserved. These mechanisms are likely to have evolved to link nutritional and metabolic status to stem cell production, which for GSCs ultimately affects fertility and fecundity.

We thank David Michaelson for preliminary data; Josefin Friberg and Ming Sheng (Umeå, Sweden) for sharing observations concerning the reduced fertility of rsks-1 mutants; Agneta Rönnlund (Umeå, Sweden) for help in the isolation of rskn-1(sv40); Vytas Bindokas (Chicago, IL, USA) for modification of an ImageJ plug-in; Di Chen and Pankaj Kapahi (Novato, CA, USA), Richard Roy (Montreal, Quebec, Canada), Brett Keiper (Greenville, NC, USA), Ann Wehman and Jeremy Nance (New York, NY, USA), the C. elegans Gene Knockout Consortium, (Oklahoma City, OK, USA and Vancouver, Canada), the Caenorhabditis Genetics Center (CGC, Minneapolis, MN, USA) and Wormbase (www.wormbase.org) for reagents and information; and Robert Schneider and members of the Hubbard and Nance labs for helpful discussions.