ABSTRACT
Stem cell regulation relies on extrinsic signaling from a niche plus intrinsic factors that respond and drive self-renewal within stem cells. A priori, loss of niche signaling and loss of the intrinsic self-renewal factors might be expected to have equivalent stem cell defects. Yet this simple prediction has not been borne out for most stem cells, including C. elegans germline stem cells (GSCs). The central regulators of C. elegans GSCs include extrinsically-acting GLP-1/Notch signaling from the niche, intrinsically-acting RNA binding proteins in the PUF family, termed FBF-1 and FBF-2 (collectively FBF), and intrinsically-acting PUF partner proteins that are direct Notch targets. Abrogation of either GLP-1/Notch signaling or its targets yields an earlier and more severe GSC defect than loss of FBF-1 and FBF-2, suggesting that additional intrinsic regulators must exist. Here, we report that those missing regulators are two additional PUF proteins, PUF-3 and PUF-11. Remarkably, an fbf-1 fbf-2; puf-3 puf-11 quadruple null mutant has a GSC defect virtually identical to that of a glp-1/Notch null mutant. PUF-3 and PUF-11 both affect GSC maintenance; both are expressed in GSCs; and epistasis experiments place them at the same position as FBF within the network. Therefore, action of PUF-3 and PUF-11 explains the milder GSC defect in fbf-1 fbf-2 mutants. We conclude that a “PUF hub”, comprising four PUF proteins and two PUF partners, constitutes the intrinsic self-renewal node of the C. elegans GSC RNA regulatory network. Discovery of this hub underscores the significance of PUF RNA-binding proteins as key regulators of stem cell maintenance.
INTRODUCTION
Stem cell regulatory networks govern the balance between self-renewal and differentiation. Transcription factors are well established stem cell regulators (e.g. BOYER et al. 2005), as are RNA-binding proteins (e.g. WICKENS et al. 2002; YE AND BLELLOCH 2014; GROSS-THEBING et al. 2017). Indeed, the PUF (for Pumilio and FBF) family of RNA-binding proteins promote stem cell self-renewal in multiple tissues across animal phylogeny from planaria to mammals (LIN AND SPRADLING 1997; FORBES AND LEHMANN 1998; CRITTENDEN et al. 2002; SALVETTI et al. 2005; NAUDIN et al. 2017; ZHANG et al. 2017). Genome-wide studies reveal that PUF proteins bind hundreds of RNAs (HAFNER et al. 2010; KERSHNER AND KIMBLE 2010; PRASAD et al. 2016; PORTER et al. 2019), consistent with a central role in the stem cell regulatory network (KERSHNER et al. 2013). Yet the full significance of PUF proteins in self-renewal has been unclear. Are PUF proteins the primary stem-cell intrinsic self-renewal regulators? Or do they play only a supporting role? This question has been difficult to tackle in organisms where only one or two PUF proteins are responsible for many diverse processes, including embryogenesis, germ cell development and neural activities. Here we address the question in nematodes, where the number of genes encoding PUF proteins has expanded during evolution, yielding functional specialization among family members.
We focus here on the role of PUF proteins in regulating self-renewal in the C. elegans germline. Germline stem cells (GSCs) expand the germline tissue from two cells at hatching to ∼2000 cells in adults; they replenish the tissue as germ cells are lost to gametogenesis during reproduction (KIMBLE AND WHITE 1981; CRITTENDEN et al. 2006); and they regenerate the tissue upon feeding after starvation (ANGELO AND VAN GILST 2009; SEIDEL AND KIMBLE 2011). Key regulators of GSC self-renewal include GLP-1/Notch signaling from the niche; its transcriptional targets, lst-1 and sygl-1; and two PUF proteins, FBF-1 and FBF-2 (collectively FBF) (Figure 1A) (AUSTIN AND KIMBLE 1987; CRITTENDEN et al. 2002; KERSHNER et al. 2014).
While FBF-1 and FBF-2 are crucial in GSC control, they cannot account for all the effects of niche signaling. Removal of either the GLP-1/Notch receptor or both lst-1 and sygl-1 has an early and severe germline proliferation (Glp) defect in both sexes: the two GSCs at hatching divide once or twice and then differentiate prematurely as sperm (AUSTIN AND KIMBLE 1987; KERSHNER et al. 2014). Moreover, GLP-1/Notch and its targets drive self-renewal throughout larval development and in adults (AUSTIN AND KIMBLE 1987; KERSHNER et al. 2014). By contrast, GSCs in the fbf-1 fbf-2 double mutant are lost to differentiation much later, at the fourth larval stage (L4s) (CRITTENDEN et al. 2002). Indeed, this delayed L4 GSC defect occurs in mutants raised at 15° and 20°, but at 25°, the effect is even further delayed with GSCs continuing to divide in adults and persisting in an undifferentiated state (MERRITT et al. 2008; SHIN et al. 2017; this work). We designate the less severe germline phenotype of fbf-1 fbf-2 mutants pGlp (“partial Glp”) to distinguish it from the more severe Glp GSC defect seen upon loss of niche signaling. The pGlp defect implies that some other self-renewal factor, “gene X” (Figure 1A), must maintain GSCs in larvae at all temperatures and in adults at 25°.
Here, we report that two additional PUF proteins, PUF-3 and PUF-11, are gene X. When PUF-3 and PUF-11 are removed in an fbf-1 fbf-2 mutant, the GSC defect is essentially the same as that seen upon loss of niche signaling. Additional results solidify the conclusion that PUF-3 and PUF-11 are intrinsic regulators of GSC self-renewal. We conclude that a “PUF hub”, comprising four PUF proteins and two PUF partners, constitutes the intrinsic self-renewal node of the C. elegans GSC RNA regulatory network. Discovery of this hub underscores the significance of PUF RNA-binding proteins as key regulators of stem cell maintenance.
MATERIALS AND METHODS
Nematode strains and maintenance
C. elegans strains were maintained at 20°, unless specified otherwise, on nematode growth medium (NGM) plates spotted with E. coli OP50 (BRENNER 1974). Wild-type was N2 Bristol strain. For a complete list of strains used in this study, see Table S1. puf-11(gk203683) is from the Million Mutations Project (THOMPSON et al. 2013). We also used the following balancers: hT2[qIs48] I;III (SIEGFRIED AND KIMBLE 2002), mIn1[mIs14dpy-10(e128)] II (EDGLEY AND RIDDLE 2001) and nT1[qIs51] IV;V (EDGLEY et al. 2006).
Due to incompatibility between mIn1 and nT1 balancers, we generated a “pseudo (Ψ)-balancer” to maintain quadruple mutant strains. This LG II Ψ-balancer harbors a transgene driving expression of a red fluorescent protein in somatic nuclei, oxTi564 [Peft-3::tdTomato::H2B::unc-54 3’UTR + Cbr-unc-119(+)] (FRØKJÆR-JENSEN et al. 2014) plus a closely linked dpy-10(q1074) deletion. Quadruple mutants were thus maintained as fbf-1 fbf-2/ Ψ-balancer [dpy-10(q1074) oxTi564] II; puf-3 puf-11 IV/nT1[qIs51] IV;V.
CRISPR/Cas9 genome editing
We used RNA-protein complex CRISPR/Cas9 genome editing with a co-conversion strategy (ARRIBERE et al. 2014; PAIX et al. 2015) to generate a number of alleles for this work (see Table S2 for details). For each edit, we prepared an injection mix containing: gene-specific crRNAs (10 µM, IDT-Alt-RTM); dpy-10 or unc-58 co-CRISPR crRNAs (4 µM, IDT-Alt-RTM); tracrRNAs (13.6 µM, IDT-Alt-RTM); gene specific repair oligo (4 µM); dpy-10 or unc-58 repair oligo (1.34 µM); and Cas-9 protein (24.5 µM). Wild-type germlines were injected to make puf-3(q966), puf-11(q971), puf-3(q1058) and puf-11(q1128); EG7866 germlines were injected for dpy-10(q1074). F1 progeny of injected hermaphrodites were screened for desired mutations by PCR and Sanger sequencing. Each allele was outcrossed with wild-type at least twice prior to analysis.
Isolation of puf-3(q801)
The puf-3(q801) 776 nucleotide deletion allele was isolated from a mutagenized library (gift of Maureen Barr), PCR screened to identify homozygotes and outcrossed against wild-type six times before analysis.
RNA interference
For RNA interference (RNAi), we identified clones targeting puf-3, puf-5, puf-6, puf-7, puf-8 and puf-9 from the Ahringer library (FRASER et al. 2000). The library does not include a clone targeting puf-11, so we generated one using the Gibson assembly method (GIBSON 2009). Nucleotides 1-800 of the puf-11 ORF were amplified from C. elegans cDNA and cloned into the plasmid L4440 at the Nco I site. For all RNAi experiments, the L4440 plasmid lacking a gene of interest (“empty” RNAi) served as a control.
To perform RNAi by feeding (TIMMONS AND FIRE 1998), E. coli HT115(DE3) were transformed with RNAi vectors and cultured at 37° overnight in 2xYT media containing 25 μg/μl carbenicillin and 50 μg/μl tetracycline. Bacterial cultures were concentrated and seeded onto NGM plates containing 1mM IPTG, then induced overnight at RT. We then placed mid L4 hermaphrodites on these plates. For experiments shown in Figure 3D,E, we assayed treated animals 48 hours after plating. For all other experiments, treated animals were allowed to lay eggs and their F1 progeny were assayed for defects.
Analysis and quantification of germ cells in adults
We first staged animals to roughly the same stage of adulthood. Specifically, we picked mid-L4s raised at 15°, 20° or 25° for at least one generation and grew them for an additional 18 hours (25°), 24 hours (20°) or 36 hours (15°) to synchronize adults for analysis. Whole worms were then DAPI stained and imaged by compound microscopy. The presence or absence of a progenitor zone (PZ) was assayed by nuclear morphology of germ cells at the distal end of the gonad: those lacking meiotic nuclear morphology and often possessing M-phase or anaphase nuclei were scored as PZ-positive (see also PZ analysis, below); those having a meiotic prophase nuclear morphology were scored as “meiotic” and PZ-negative; arms where all germ cells had differentiated as mature sperm were scored as “sperm” and PZ-negative. To estimate total germ cell number, we counted mature sperm (which have a distinctive, compact DAPI morphology) using the Multipoint Tool in Fiji/ImageJ (SCHINDELIN et al. 2012) and then divided sperm number by four (each germ cell makes four sperm). In some cases, we counted sperm number in one gonadal arm and multiplied that number by two to estimate total sperm number. In cases where all germ cells in an animal had not yet differentiated to sperm, we did not count total germ cells.
Phenotype analyses: Fertility, brood size, and embryo viability
L4 hermaphrodites were placed onto individual plates at 20°. At 6- to 24-hour intervals, each hermaphrodite was moved to a new plate and the embryos were counted to score for fertility and determine brood size. Several days later, hatched progeny on each plate were counted to determine embryo lethality.
Progenitor zone (PZ) analysis
From roughly staged adults as described above, gonads were extruded, DAPI stained, and imaged by compound or confocal microscopy. We examined morphology of germline nuclei to determine PZ size according to convention (CRITTENDEN et al. 2006; SEIDEL AND KIMBLE 2015) (see also Figure 1E). Briefly, DAPI-staining of nuclei in early meiotic prophase adopts a crescent shape. To count number of germ cells in a PZ, we counted the total number of cells in the distal germline that had not entered early meiotic prophase, using Fiji/ImageJ Cell Counter plugin (SCHINDELIN et al. 2012). We also measured the distance to the end of the PZ in germ cell diameters (gcd) from the distal end. To this end, we selected a middle focal plane and counted the number of germline nuclei along each edge of the gonad until the first one with crescent morphology. We averaged the two values from each edge to determine PZ size.
Quantification of germ cells in larvae
To generate roughly synchronous embryos, we allowed gravid hermaphrodites to lay eggs for 2 hours at 20°. At subsequent timepoints corresponding to early L1, early L2, late L2 and early L4, larvae were harvested and germ cell number quantitated. For early L1s, we scored germ cell number in live animals by DIC microscopy. For all remaining samples, we used whole mount staining with the reduction/oxidation method (FINNEY AND RUVKUN 1990; MILLER AND SHAKES 1995). Briefly, samples were fixed in Ruvkun Fixation Buffer with 1% (v/v) paraformaldehyde for 30 minutes. Disulfide linkage reduction was performed, then samples were incubated in blocking solution (PBS with 1% (w/v) bovine serum albumin, 0.5% (v/v) Triton X-100, 1 mM EDTA) for 40 minutes at RT, followed by overnight incubation at 4° with rabbit α-PGL-1 (1:100, gift from Susan Strome, University of California, Santa Cruz) diluted in blocking solution. Secondary Alexa 555 donkey α-rabbit (1:1000, Thermo Fisher Scientific #A31570) antibody was diluted in blocking solution and incubated with samples for at least one hour along with 1 ng/μl DAPI for DNA visualization. Samples were mounted in Vectashield (Vector Laboratories #H1000) on 2% agarose pads then assayed by fluorescent compound microscopy. Number of germ cells in L2s was determined by counting the total number of PGL-1-postitive cells. For early L4s, we counted the number of PGL-1-positive cells in one gonadal arm then multiplied by two to estimate the total number of germ cells per animal.
Immunostaining and DAPI staining
We immunostained gonads as described (CRITTENDEN et al. 2017) with minor modifications. To extrude gonads, we dissected animals in PBS buffer with 0.1% (v/v) Tween-20 (PBStw) and 0.25 mM levamisole. Gonads were fixed in 3% (w/v) paraformaldehyde diluted in PBStw for 10 minutes, then permeabilized in either 0.2% (v/v) Triton-X diluted in PBStw or ice-cold methanol for 10-15 minutes. Next, gonads were incubated for at least 1 hour in blocking solution (0.5% (w/v) bovine serum albumin diluted in PBStw) and incubated overnight at 4° with primary antibodies diluted in blocking solution as follows: mouse α-V5 (1:1000, SV5-Pk1, Bio-Rad #MCA1360); mouse α-SP56 (1:200, gift from Susan Strome, University of California, Santa Cruz). Secondary antibody Alexa 488 donkey α-mouse (1:1000, Thermo Fisher Scientific #A21202) was diluted in blocking solution and incubated with samples for at least one hour. To visualize DNA, DAPI (4′,6-diamidino-2-phenylindole) was included with the secondary antibody at a final concentration of 1 ng/μl. Samples were mounted in Vectashield (Vector Laboratories #H1000) or ProLong Gold (Thermo Fisher Scientific #P36930) before imaging. All steps were performed at room temperature unless otherwise indicated. Where only DNA visualization was required, we skipped all blocking solution steps and simply incubated samples in PBStw with 1 ng/μl DAPI for 15 minutes prior to mounting.
Microscopy
Images in Figure 2C,D, 3D,E, 5B-G and Figure S1A,B and S4B-J were taken using a laser scanning Leica TCS SP8 confocal microscope fitted with both photomultiplier and hybrid detectors and run using LAS software version X. A 63x/1.40 CS2 HC Plan Apochromat oil immersion objective was used. All images were taken with 400 Hz scanning speed and 125-200% zoom. To prepare images for figures, Adobe Photoshop was used to equivalently and linearly adjust intensity among images to be compared. Images in Figure 1F,G, 4A-D and S5 were captured using a Hamamatsu ORCA-Flash4.0 cMOS camera on a Zeiss Axioskop compound microscope equipped with 63x 1.4NA Plan Apochromat oil immersion objective. The fluorescent light source was a Lumencore SOLA Light Engine, and Carl Zeiss filter sets 49 and 38 were used for DAPI and Alexa 488 visualization. The acquisition software was Micromanager (EDELSTEIN et al. 2010; EDELSTEIN et al. 2014). When required, images were combined using the Pairwise Stitching function in Fiji/ImageJ (PREIBISCH et al. 2009). To prepare images for figures, Adobe Photoshop was used to adjust intensity equivalently and linearly among images to be compared.
Fluorescence quantitation
Quantitation of fluorescence in Figures 5H,I was performed with Fiji/ImageJ (SCHINDELIN et al. 2012). In Figure 5H, we performed four independent immunostaining experiments and quantitated a total of at least 27 gonads per genotype. In Figure 5I, we performed two independent experiments and quantitated at least 24 gonads per genotype. From confocal image stacks, we collected raw pixel intensity data from each gonad image by projecting the sum of all Z-slices onto a single plane. A freehand line, 50 pixels wide and 50 μm long that bisected the gonad, was drawn manually using the Plot Profile tool starting at the distal tip of the tissue. We found the mean intensity value of the plot profile of each gonad to get a single value reflecting the amount of protein present. Next, we subtracted any signal representing nonspecific antibody binding for each independent experiment: we calculated the mean intensity value in the respective wild-type control gonads, then subtracted it from each experimental sample. In Figure 5H,I, we report the background subtracted mean and standard error.
Genetic epistasis experiments
To test the relationship between lst-1 sygl-1 and fbf-1 fbf-2; puf-3 puf-11, we used transgenes that ubiquitously express LST-1 and SYGL-1 protein. Because these lst-1(gf) and sygl-1(gf) transgenes cause germline tumors and are sterile, they were maintained on lst-1 or sygl-1 RNAi, respectively, prior to the experiment. Expression of LST-1 or SYGL-1 was then induced by transferring L4s to OP50-seeded NGM plates and passaging for several generations (SHIN et al. 2017). The assay for Figure 6A was performed at 25°. From populations grown for 9 generations on OP50, we plated L4s onto puf-3/11 RNAi. Next, progeny of puf-3/11 RNAi treated animals were staged to 24 hours past L4 and stained with DAPI to assay effects on germline development. Because lst-1 and sygl-1 require fbf-1 fbf-2 for function, neither lst-1(gf) nor sygl-1(gf) formed a germline tumor in these experiments.
To test the relationship between gld-2 gld-1 and fbf-1 fbf-2; puf-3 puf-11 for Figure 6B, we plated mid-L4 staged JK5778 to puf-3/11 RNAi plates at 20°. F1 progeny were staged to 24 hours past L4, then DAPI stained and imaged using compound microscopy to assay germline phenotype.
Statistical analysis
Welch’s ANOVA and Games-Howell post hoc tests were performed to calculate statistical significance for multiple samples. All statistical tests were performed in R and the p-value cut off was 0.05.
Yeast two-hybrid
Modified yeast two-hybrid assays were performed as described (BARTEL AND FIELDS 1997). PUF proteins were fused to the LexA DNA binding domain as follows: cDNA sequences encoding PUF repeats of FBF-1 (amino acids 121-614), FBF-2 (121-632), PUF-3 (88-502), PUF-9 (162-703) and PUF-11 (91-505) were each cloned into the Nde I site of pBTM116 using the Gibson assembly method (GIBSON et al. 2009). We also used full-length LST-1 (1-328) and SYGL-1(1-206) fused to Gal4 activation domain in the pACT2 vector (SHIN et al. 2017). More details about plasmids are available in Table S3. To test for protein-protein interactions between PUFs and LST-1/SYGL-1, activation and binding domain pairs were co-transformed into a L40-ura3 strain (MATa, ura3-52, leu2-3,112, his3Δ200, trp1Δ1, ade2, LYS2::(LexA-op)4 –HIS3, ura3::(LexA-op)8 –LacZ) using the LiOAc method (GIETZ AND SCHIESTL 2007). LacZ reporter activity was measured using the Beta-Glo® Assay system (Promega #E4720), following commercial protocols and yeast-specific methods (HOOK et al. 2005). Luminescence was quantitated using a Biotek Synergy H4 Hybrid plate reader with Gen5 software.
Western blot
For the western blot in Figure 7C, we grew yeast transformants in -Leu -Trp liquid media and prepared samples by boiling yeast in sample buffer (60mM Tris pH 6.8, 25% glycerol, 2% SDS, 0.1% bromophenol blue with 14 mM beta-mercaptoethanol or 100 mM DTT). Analysis was conducted on a 4-15% SDS-PAGE gradient gel (Biorad #456-1083). We probed with primary antibodies overnight at 4° as follows: 1:50,000 mouse anti-HA (HA.11, Covance #MMS-101R), 1:1000 mouse anti-V5 (1:1000, SV5-Pk1, Bio-Rad #MCA1360), or 1:10,000 mouse anti-actin (C4, Millipore #MAB1501). For secondary antibodies, blots were incubated for 1-2 hours at RT with 1:20,000 donkey anti-mouse horseradish peroxidase (Jackson ImmunoResearch #715-035-150). Immunoblots were developed using SuperSignalTM West Pico/Femto Sensitivity substrate (Thermo Scientific #34080, #34095) and a Konica Minolta SRX-101A medical film processor. For final figure preparations, intensity of the blot was linearly adjusted in Adobe Photoshop.
Data and reagent availability
The authors affirm that all data necessary for confirming the conclusions of this article are present within the article, tables, figures, and supplemental material. All strains and plasmids are available upon request or via the Caenorhabditis Genetic Center, supported by the NIH Office of Research Infrastructure Programs (P40 OD010440). All protocols are available upon request.
RESULTS
Solidifying evidence for the existence of gene X
The existence of a missing GSC self-renewal regulator, gene X, was proposed because of striking differences in GSC defects upon removal of either the GLP-1/Notch receptor or its target genes (lst-1 and sygl-1) on the one hand and removal of FBF-1 and FBF-2 on the other (Figure 1B) (AUSTIN AND KIMBLE 1987; CRITTENDEN et al. 2002; KERSHNER et al. 2014). To provide comprehensive data as a critical baseline for this study, we scored GSC defects in key mutants at 15°, 20° and 25° (Figure 1C). As expected, mutants lacking GLP-1/Notch or both its target genes, lst-1 and sygl-1, generated only about four germ cells at each temperature, but fbf-1 fbf-2 double mutants made many more (Figure 1C). In addition, fbf-1 fbf-2 mutants raised at 25° possessed distal germ cells in mitotic metaphase or anaphase, consistent with active divisions (Figure S1A). A previous study showed that these distal cells express the mitotic marker, nuclear REC-8 (see SHIN et al. 2017, Fig S5G). Thus, loss of FBF-1 and FBF-2 has a much less severe and later effect on GSC maintenance than loss of niche signaling or loss of the niche targets, confirming the notion of some missing self-renewal regulator (gene X, Figure 1A).
PUF-3 and PUF-11 proteins are likely the missing self-renewal regulators
To begin our search for the missing GSC regulators, we considered other PUF proteins as logical candidates. Among 10 C. elegans PUF-encoding genes (LIU et al. 2012), PUF-11 piqued our interest because of two similarities with FBF: the PUF-11 protein interacts with LST-1 in a genome-wide yeast two-hybrid screen (BOXEM et al. 2008), and its RNA binding specificity is similar to that of FBF (BERNSTEIN et al. 2005; KOH et al. 2009). Because the C. elegans genome encodes a PUF-11 paralog with nearly identical sequence, called PUF-3 (Figure 1D, Figure S2), we tested both for a role in GSC maintenance. To this end, we used two feeding RNAi clones, one targeting puf-3 (RNAi A) and the other targeting a distinct region of puf-11 (RNAi B) (Figure 1D). Although these RNAi clones target different gene regions, each was expected to deplete both puf-3 and puf-11 because of their ∼90% sequence identity (HUBSTENBERGER et al. 2012).
A previous puf-3/11 RNAi study, performed in wild-type animals, identified a role for PUF-3 and PUF-11 in oogenesis, but not GSC maintenance (HUBSTENBERGER et al. 2012). We therefore sought a more GSC-specific assay and turned to enhancement of the fbf-1 fbf-2 pGlp phenotype at 25°. For this analysis, we scored the presence or absence of a progenitor zone (PZ), the distal region where germ cells have not yet entered into the meiotic cell cycle and continue mitotic divisions (Figure 1E). Whereas all fbf-1 fbf-2 adults on empty vector RNAi possessed a PZ at 25° (Figure 1F,H), most treated with puf-3/11 RNAi lost their PZ to differentiation (Figure 1G). A comparable effect was seen in both sexes (Figure 1H). Strikingly, GSC divisions generated only ∼4 germ cells per gonad in each sex (Figure 1I), as determined by counting the number of mature sperm in adults and dividing by four. RNAi directed against other puf loci (e.g. puf-8) did not enhance the pGlp fbf-1 fbf-2 phenotype (Figure S1B). Therefore, puf-3/11 RNAi enhances the fbf-1 fbf-2 germline phenotype from its partial pGlp to the full Glp typical of GLP-1/Notch mutants. This enhancement indicates that PUF-3 and PUF-11 are likely the missing self-renewal regulators.
puf-3 and puf-11 mutants have no GSC proliferation defects on their own
The puf-3/11 RNAi experiments could not distinguish between the puf-3 and puf-11 genes for effects on GSC maintenance. To test their individual roles, we analyzed puf-3 and puf-11 single mutants: three deletions, puf-3(q801), puf-3(q966) and puf-11(q971), and one nonsense allele, puf-11(gk203683) (Figure 2A, Figure S2). As a measure of general germline function, we scored fertility, number of embryos laid and embryo viability. The single mutants were fertile with brood sizes comparable to wild-type, and their embryos hatched into young larvae, except for puf-11(q971), which had a partially penetrant embryonic lethality (Figure 2B). In addition, we scored PZ lengths as a proxy for effects on GSCs. The PZ lengths, measured with the conventional metric of germ cell diameters (gcd) from the distal end, were roughly the same as wild-type in all puf-3 and puf-11 single mutants (Figure 2B). Therefore, puf-3 and puf-11 single mutants have no major GSC proliferation defects.
We next assessed puf-3 puf-11 double mutants (Figure 2B-F). Germlines had an organization and size comparable to wild-type (Figure 2C, D), but they made no viable embryos (Figure 2B). Male puf-3 puf-11 crossed to feminized fog-1 hermaphrodites yielded ample cross progeny, so puf-3 puf-11 sperm are functional. By contrast, wild-type males crossed to puf-3 puf-11 hermaphrodites failed to make viable progeny, suggesting that puf-3 puf-11 oocytes are defective, consistent with prior studies (HUBSTENBERGER et al. 2012; HUBSTENBERGER et al. 2013). PZ lengths were comparable to wild-type, scored at 20° (Figure 2B), and number of germ cells therein were comparable to wild-type at both 20° and 25° (Figure 2E,F). Therefore, puf-3 puf-11 double mutants have an oogenesis defect, as previously recognized, but no obvious GSC defects.
puf-3 and puf-11 enhancement of fbf-1 fbf-2 partial Glp phenotype
To further explore puf-3 and puf-11 roles in GSC maintenance, we tested for enhancement of the fbf-1 fbf-2 pGlp phenotype using fbf-1 fbf-2; puf triple mutants as well as fbf-1 fbf-2; puf-3 puf-11 quadruple mutants. To score enhancement, we determined total germ cells made in each strain by counting mature sperm number in adults and dividing by four (Figure 3A,B); at 25°, we scored for the persistence of a PZ in adults (Figure 3C,D).
We first assayed triple mutants raised at 20° (Figure 3A, left; Figure S3). The control fbf-1 fbf-2 double mutants made roughly 100 germ cells before GSCs were lost to spermatogenesis, as previously described (CRITTENDEN et al. 2002; LAMONT et al. 2004; this work). That number decreased in both triple mutants, but the extent of pGlp enhancement differed for puf-3 and puf-11. The decrease in germ cell number was small in fbf-1 fbf-2; puf-3 triple mutants and statistically significant for only one allele; the decrease was larger in fbf-1 fbf-2; puf-11 mutants and statistically significant for both alleles. Nonetheless, each puf gene contributed to larval GSC proliferation at 20°.
We also assayed triple mutants raised at 25° (Figure 3A, right; Figure 3B; Figure S3). The control fbf-1 fbf-2 double mutants made more than 100 germ cells and maintained a PZ, as described previously (MERRITT et al. 2008; SHIN et al. 2017; this work). In fbf-1 fbf-2; puf-3 triple mutants, germline size was comparable to fbf-1 fbf-2 double mutants (Figure 3A, right), but many PZs were lost to meiotic entry (Figure 3B). By contrast, fbf-1 fbf-2; puf-11 triple mutants at 25° made far fewer germ cells overall than fbf-1 fbf-2 (Figure 3A, right), and had a fully penetrant PZ loss with all germ cells differentiating as sperm (Figure 3B). Thus, each puf gene enhanced the fbf-1 fbf-2 pGlp defect at 25°, with puf-11 again having a more severe effect than puf-3.
We next assayed fbf-1 fbf-2; puf-3 puf-11 quadruple mutants, this time raised at 15°, 20° or 25°. The two distinct quadruple mutants, fbf-1 fbf-2; puf-3(q966) puf-11(q971) and fbf-1 fbf-2; puf-3(q801) puf-11(gk203683), were remarkably similar at all three temperatures. Adults had tiny germlines composed entirely of mature sperm. Upon quantitation (Figure 3C; Figure S3), quadruples made a total of 4-9 germ cells on average at 15° and 25°, and 11-16 at 20°. This quadruple Glp phenotype is thus comparable to glp-1 and lst-1 sygl-1 null mutants. Thus, PUF-3 and PUF-11 function during larval development to maintain GSC divisions.
We finally asked if PUF-3 and PUF-11 maintain the PZ in fbf-1 fbf-2 double mutant adults at 25°. To address this question, we treated mid-L4 puf-3(q966) puf-11(q971) double mutants with fbf-1/2 RNAi and assayed PZ presence or absence in adults (48 hours later). Most wild-type animals treated with fbf-1/2 RNAi retained a PZ (92%, n=12) (Figure 3D). However, few puf-3 puf-11 double mutants treated with fbf-1/-2 RNAi retained a PZ (7%, n=30) (Figure 3E). Instead, distal-most germ cells entered early meiotic prophase, visualized by a nuclear “crescent” morphology. Thus, PUF-3 and PUF-11 function in 25° fbf-1 fbf-2 adults to maintain a progenitor zone.
GSC maintenance fails during early larval development in quadruple mutants
Up to this point in this work, germ cell counts were performed in adults by counting sperm. This approach quantitates number of cells generated and differentiated, but cannot detect cells that die without differentiation into gametes. Although germ cell death was not seen in glp-1 or lst-1 sygl-1 mutants (AUSTIN AND KIMBLE 1987; KERSHNER et al. 2014), the fbf-1 fbf-2; puf-3 puf-11 quadruple mutant may be different.
To begin, we assessed overall germline sizes at the L4 stage, which were comparable in wild-type, fbf-1 fbf-2 and puf-3 puf-11, but much smaller in quadruple mutants (Figure 4A-D). Next, we counted total germ cell number at specific intervals during larval development. For this experiment, we scored fbf-1 fbf-2; puf-3 puf-11 quadruple mutants and several controls (wild-type, fbf-1 fbf-2 doubles, puf-3 puf-11 doubles and glp-1), all maintained at 20°. PGL-1 staining was used to identify germ cells for counting (except for early L1, which was scored by DIC in live animals). We found that germ cell numbers increased similarly during larval development in wild-type, fbf-1 fbf-2 and puf-3 puf-11 animals, but they did not increase appreciably in glp-1 or the fbf-1 fbf-2; puf-3 puf-11 quadruple mutant (Figure 4E,F). By L4, the quadruple mutants had made a total of 15 germ cells on average (range=12-20, n=5) (Figure 4E,F), similar to the number of germ cells estimated from adult sperm number at the same temperature (Figure 3B,C) and consistent with cell death having little impact on germ cell number. In parallel, we visualized meiotic entry with DAPI staining. By late L2, germ cells in quadruple mutants had entered meiotic prophase, whereas wild-type, fbf-1 fbf-2 and puf-3 puf-11 had not. No morphological sign of germ cell death was seen over the course of these experiments. Together, these findings allay the concern that puf-3 puf-11 mutants might reduce germ cell number by promoting cell death and thus support the conclusion that puf-3 and puf-11 normally promote self-renewal during larval development.
PUF-3 and PUF-11 expression in GSCs
To test whether PUF-3 and PUF-11 proteins are expressed in GSCs, we generated V5 epitope tagged alleles (Figure 5A; Figure S2). Both PUF-3V5 and PUF-11V5 are functional, as assayed by their lack of fbf enhancement (Figure S4A). Upon immunostaining and imaging, PUF-3V5 and PUF-11V5 proteins were observed in the distal germlines of mid-L4 hermaphrodites raised at 20° (Figure 5B-D; Figures S4B-D for full gonad), of adult hermaphrodites raised at 20° (Figure S4E-G) and of mid-L4 males raised at 20° (Figure S4H-J). In addition, both proteins were present more proximally in developing oocytes (Figure S4E-G). As expected for PUF proteins, both PUF-3V5 and PUF-11V5 were cytoplasmic (Figure 5E-G) and localized to perinuclear granules (Figure 5E,F insets), consistent with an RNA regulatory role. No ⍺-V5 staining was seen in wild-type germlines, as expected because they lacked the epitope tag. We quantified the PUF-3V5 and PUF-11V5 signal in the distal gonads of L4s and adults at 20° (Figure 5H,I), subtracting the very low background in wild-type for each. PUF-11 was more abundant than PUF-3 at both stages. Finally, we confirmed expression of both proteins in adult distal germlines at 25° and again quantitated the signal (Figure 5I). We conclude that PUF-3 and PUF-11 are expressed in GSCs.
puf-3 puf-11 placement in GSC regulatory pathway
The notion that puf-3 and puf-11 represent the missing self-renewal regulators, dubbed gene X, predicts their placement in the GSC regulatory pathway (Figure 1A). Their fbf enhancement, reported above, is consistent with puf-3 puf-11 functioning in parallel to fbf-1 fbf-2, but we conducted two additional epistasis experiments to solidify that pathway position. For these experiments, we used RNAi to deplete puf-3 and puf-11, both for ease of genetic manipulation and because GSC defects were comparable after RNAi and in quadruple mutants.
We first investigated the relationship between puf-3 puf-11 and lst-1 sygl-1 (Figure 6A; Figure S6A-D). Previous studies showed that fbf-1 fbf-2 functions either downstream or in parallel to lst-1 sygl-1. This pathway placement was deduced using gain-of-function (gf) mutants of lst-1 and sygl-1. Both lst-1(gf) and sygl-1(gf) make massive germline tumors when fbf-1 and fbf-2 are wild-type, but acquire a pGlp phenotype when fbf-1 and fbf-2 are removed in lst-1(gf); fbf-1 fbf-2 and sygl-1(gf); fbf-1 fbf-2 triple mutants (SHIN et al. 2017). To ask if puf-3 and puf-11 have the same genetic relationship to lst-1 and sygl-1, we treated lst-1(gf); fbf-1 fbf-2 and sygl-1(gf); fbf-1 fbf-2 triple mutants with either empty vector RNAi as a control or puf-3/11 RNAi. The control germlines had a pGlp phenotype (Figure 6A, Figure S6A,C), as shown previously (SHIN et al. 2017). However, with puf-3/11 RNAi, most germlines had a fully Glp phenotype, typical of fbf-1 fbf-2; puf-3 puf-11 quadruple mutants. Their germlines were tiny with only a few sperm (Figure S6B,D), and upon quantitation, only 4-8 germ cells were made on average per animal (Figure 6A). Therefore, puf-3 and puf-11 likely function downstream or in parallel to lst-1 and sygl-1 (Figure 6C).
We next investigated the relationship between puf-3 puf-11 and gld-1 gld-2 (Figure 6B; Figure S6E,F). The gld-1 and gld-2 genes promote meiotic entry and in their absence, the germline becomes tumorous (KADYK AND KIMBLE 1998). Previous studies showed that fbf-1 fbf-2 functions upstream of gld-1 gld-2. Thus, gld-1 gld-2 tumors are epistatic to fbf-1 fbf-2 pGlp (ECKMANN et al. 2004). To ask if puf-3 and puf-11 are similarly upstream of gld-1 gld-2, we treated gld-1 gld-2; fbf-1 fbf-2 quadruple mutants with puf-3/11 RNAi. Indeed, gld-1 gld-2 tumors were still found, demonstrating gld-1 gld-2 epistasis over fbf-1 fbf-2; puf-3/11 RNAi (Figure 6B, Figure S6E,F). We confirmed that puf-3,11 RNAi knockdown was successful by comparison with embryos of wild-type siblings (Figure S5G). Therefore, puf-3 and puf-11 function upstream of gld-1 gld-2 (Figure 6C). Together, these experiments place puf-3 and puf-11 into the GSC regulatory pathway in a position consistent with their proposed identity as gene X.
PUF-3 interacts with SYGL-1 in yeast
Placement of puf-3 puf-11 alongside fbf-1 fbf-2 in the genetic pathway (above), together with their molecular identity as PUF RNA binding proteins, suggests that PUF-3 and PUF-11 may have molecular activities in GSCs similar to FBF. Previous studies showed that FBF-1 and FBF-2 physically interact with Notch targets LST-1 and SYGL-1 (SHIN et al. 2017; QIU et al. 2019). Moreover, the LST-1 and SYGL-1 partnerships with FBF are essential in vivo for GSC self-renewal (HAUPT et al. 2019; C.R. Kanzler and H.J. Shin, unpublished). We therefore considered the possibility that PUF-3 and PUF-11 also form partnerships with LST-1 and SYGL-1. Indeed, a genome-wide interaction screen had already shown that an interaction between PUF-11 and LST-1 in yeast (BOXEM et al. 2008). We therefore focused our yeast two hybrid assays (Figure 7A) on interactions between PUF-3/PUF-11 and SYGL-1. FBF-1 and FBF-2 were tested as positive controls, and PUF-9 as a likely negative control. PUF-3 interacted robustly with SYGL-1, but PUF-9 and PUF-11 did not (Figure 7B). The discrepancy between PUF-3 and PUF-11 was initially confounding given the similarity of the two proteins. However, a Western blot revealed that PUF-11 was poorly expressed in yeast (Figure 7C). We conclude that PUF-3 interacts with SYGL-1 and suggest that PUF-11 likely does as well, due to the near identity between PUF-3 and PUF-11. These data plus those of the genome-wide screen (BOXEM et al. 2008) indicate that PUF-3 and PUF-11 likely interact physically with LST-1 and SYGL-1.
DISCUSSION
Missing self-renewal regulators are PUF-3 and PUF-11
Despite over three decades of research defining the regulatory network that maintains C. elegans germline stem cells, key self-renewal regulators were clearly missing (see Introduction, Figure 1A,B). The argument was simple. Animals lacking GLP-1/Notch signaling from the niche or lacking the two key GLP-1/Notch targets, lst-1 and sygl-1, have a much earlier GSC defect than animals lacking the only other known self-renewal regulators, FBF-1 and FBF-2. Because LST-1 and SYGL-1 proteins interact physically with FBF-1 and FBF-2 (SHIN et al. 2017; HAUPT et al. 2019; QIU et al. 2019) and that interaction is essential for GSC self-renewal (HAUPT et al. 2019), we predicted that the missing regulators might be additional PUF RNA-binding proteins. Indeed, we report here that animals lacking four PUF proteins, PUF-3 and PUF-11 in addition to FBF-1 and FBF-2, exhibit the same early GSC defect as glp-1 null mutants (Figure 7D). The fbf-1 fbf-2; puf-3 puf-11 germ cell number is equivalent to that of glp-1 null at 15° and 25°. At 20°, the quadruple mutant undergoes one additional GSC division, which is a minor difference (one temperature, one cell division). Thus, PUF-3 and PUF-11 are the major missing self-renewal regulators.
In the process of characterizing puf-3 and puf-11 as self-renewal regulators, we confirmed their primary role in oogenesis. Previous studies using RNAi directed against puf-3/11 had identified their function in oogenesis – namely, to produce viable embryos (HUBSTENBERGER et al. 2012). Our work extends that previous study in three ways. Using deletion mutants of each gene, we find that PUF-3 and PUF-11 act redundantly during oogenesis; using tagged versions of each protein, we show that both are expressed in oocytes; and using genetics, we find that PUF-3 and PUF-11 are not required for spermatogenesis. Importantly, while their oogenesis function is exclusive to hermaphrodites, their role in GSC self-renewal is critical in both hermaphrodites and males and thus is gender-independent. Their molecular mechanism of action in both GSCs and oocytes likely revolves around RNA regulation, a common theme among PUF proteins. Regardless, we emphasize that PUF-3 and PUF-11 are the long-sought missing self-renewal regulators.
A “PUF hub” is responsible for GSC self-renewal
Regulatory networks are central to cell fates and tissue patterning across animal phylogeny. For germline fates and patterning, regulatory networks that rely on post-transcriptional regulation have emerged as particularly prominent (e.g. KIMBLE AND CRITTENDEN 2007; SLAIDINA AND LEHMANN 2014; YAMAJI et al. 2017), with key RNA-binding proteins regulating hundreds of RNAs and directing cell fate programs (KERSHNER AND KIMBLE 2010; AOKI et al. 2018; PORTER et al. 2019; this work). This work reveals that PUF RNA-binding proteins are principle intrinsic regulators in the stem cell network. Four PUFs collectively drive GSC self-renewal in hermaphrodites and males, in larvae and adults and at all laboratory growth temperatures (15°, 20° and 25°). The centrality of PUF proteins to the GSC regulatory network is underscored by the fact that GSC defects of fbf-1 fbf-2; puf-3 puf-11 quadruple null mutants are virtually identical to those of glp-1/Notch null and lst-1 sygl-1 null mutants (Figure 7D).
The remarkable phenotypic congruence of mutants in niche signaling, its targets LST-1 and SYGL-1, and four PUF proteins leads us to propose the concept of a “PUF self-renewal hub” in the stem cell regulatory network (Figure 7E). This hub consists of four PUF RNA-binding proteins and two PUF partner proteins, LST-1 and SYGL-1. LST-1 and SYGL-1 were first identified as partners of FBF (SHIN et al. 2017; HAUPT et al. 2019; QIU et al. 2019), but PUF-3 and PUF-11 also likely partner with LST-1 and SYGL-1 (BOXEM et al. 2008; this work). Moreover, these PUF partnerships are essential for GSC self-renewal (HAUPT et al. 2019; C.R. Kanzler and H.J. Shin, unpublished). The PUF hub therefore serves as the principal node for GSC self-renewal in the stem cell regulatory network
The PUF hub seems remarkably simple and is strongly supported by genetic and molecular analyses. However, puzzles remain. For example, mutants in the fog-1 gene, which encodes a CPEB-related RNA-binding protein, also enhance the GSC defect of fbf-1 fbf-2 double mutants, but that enhancement is coupled to a reversal in germline sex and the mechanism remains a mystery (THOMPSON et al. 2005). In contrast, as emphasized here, the “PUF hub” GSC phenotype is not coupled to any effect on germline sex determination, but instead is equivalent to removal of niche signaling (KERSHNER et al. 2014; this work). Most other intrinsic stem cell regulators do not meet this high bar of equivalence to the niche-defective phenotype. Thus, the PUF hub promises to provide a paradigm for understanding self-renewal hubs more broadly.
Redundancy and buffering within the PUF hub
The PUF hub relies on a striking nexus of functional redundancies. PUF-3 and PUF-11 are redundant with FBF during larval development and in adults at 25° (this work); and FBF-1 and FBF-2 are redundant with each other in late larval development at 15° and 20° (CRITTENDEN et al. 2002). Moreover, the two PUF partners, LST-1 and SYGL-1, are functionally redundant (KERSHNER et al. 2014; SHIN et al. 2017). These layers of redundancy, together with our molecular understanding of individual hub proteins, suggests a simple molecular model (Figure 7F). In this model, each PUF protein binds to target RNAs via 3’UTR regulatory elements and also binds to either LST-1 or SYGL-1 to elicit RNA repression. Evidence for this model is particularly strong for the FBFs, whose mode of action has been analyzed most intensively (BERNSTEIN et al. 2005; WANG et al. 2009; SHIN et al. 2017; HAUPT et al. 2019). Data are also strongly suggestive for the nearly identical PUF-3 and PUF-11 proteins: PUF-11 binds to RNA with a sequence specificity similar to that of FBF (KOH et al. 2009); PUF-3/-11 repress expression of reporter RNAs in oocytes (HUBSTENBERGER et al. 2012); PUF-11 interacts with LST-1 in yeast (BOXEM et al. 2008); and PUF-3 interacts with SYGL-1 in yeast (this work). While further analyses are needed, outlines of the hub architecture and definition of its key molecular features are clear.
The extensive redundancy among hub components suggests that components are, to a first approximation, molecularly interchangeable. That interchangeability likely renders the hub robust, namely capable of maintaining stem cells under many conditions (e.g. developmental stage, growth temperature, sex). Although not yet analyzed, the interchangeability may also help stem cells withstand the barrage of environmental inputs and stresses experienced outside the laboratory. A similar phenomenon of functional redundancy of key regulators has been found in other developmental regulatory networks (e.g. Hox genes in animal development, MAD box genes in plant development) and likely lies at the heart of network evolution more broadly (WAGNER 2008).
In addition to functional redundancy, single hub components likely have specialized individual roles. Intensive studies of FBF-1 and FBF-2 reveal numerous individual features (LAMONT et al. 2004; VORONINA et al. 2012; VORONINA 2013; BRENNER AND SCHEDL 2016; PRASAD et al. 2016; WANG et al. 2016; PORTER et al. 2019). FBF-1 and FBF-2 have distinct low penetrance sex determination defects, genetic interactions, expression, subcellular localization, target RNAs and FBF-specific molecular effects on targets. PUF-3 and PUF-11 will also likely possess differences, between each other and also with the FBFs. Understanding the common and unique roles among the members of the hub will be crucial to understanding how the hub is buffered to maintain stem cells under a variety of physiological and environmental conditions.
COMPETING INTERESTS
The authors declare no competing interests.
FUNDING
This work was supported by the National Institutes of Health [GM050942 to M.W.]; J.K. was an Investigator with the Howard Hughes Medical Institute.
ACKNOWLEDGEMENTS
We thank Peggy Kroll-Conner for help generating strains central to this work as well as isolation of the puf-3(q801) deletion, Maureen Barr for sharing her mutagenesis library, and the Million Mutations Project for generating puf-11(gk203683). We thank Susan Strome (University of California, Santa Cruz) for α-PGL-1 and α-SP56 antibodies. We are also grateful to Sarah Crittenden and Brian Carrick for comments on the manuscript, Laura Vanderploeg for help with figures and Carol Pfeffer for help with manuscript preparation. Some strains used in the study were provided by the Caenorhabditis Genetics Center, supported by the NIH Office of Research Infrastructure Programs (P40 OD010440).