Dose-Dependent Effects of GLD-2 and GLD-1 on Germline Differentiation and Dedifferentiation in the Absence of PUF-8

PUMILIO/FBF (PUF) proteins have a conserved function in stem cell regulation. Caenorhabditis elegans PUF-8 protein inhibits the translation of target mRNAs by interacting with PUF binding element (PBE) in the 3′ untranslated region (3′ UTR). In this work, an in silico analysis has identified gld-2 [a poly(A) polymerase] as a putative PUF-8 target. Biochemical and reporter analyses showed that PUF-8 specifically binds to a PBE in gld-2 3′ UTR and represses a GFP reporter gene carrying gld-2 3′ UTR in the C. elegans mitotic germ cells. GLD-2 enhances meiotic entry at least in part by activating GLD-1 (a KH motif-containing RNA-binding protein). Our genetic analyses also demonstrated that heterozygous gld-2(+/−) gld-1(+/−) genes in the absence of PUF-8 are competent for meiotic entry (early differentiation), but haplo-insufficient for the meiotic division (terminal differentiation) of spermatocytes. Indeed, the arrested spermatocytes return to mitotic cells via dedifferentiation, which results in germline tumors. Since these regulators are broadly conserved, we thus suggest that similar molecular mechanisms may control differentiation, dedifferentiation, and tumorigenesis in other organisms, including humans.


INTRODUCTION
During development, stem cells must make a number of major fate decisions -the initial decision to either proliferate or differentiate, followed by whether to remain in a differentiating state or revert to being undifferentiated as occurs in regeneration or tumorigenesis. A regulatory network controlling these decisions is vital to the development of all multicellular organisms, including humans. Aberrant regulation can result in either loss of a specific cell type or uncontrolled cell proliferation, leading to tumors. To date, significant progress has been made in stem cell differentiation using multiple model systems. Nevertheless, our understanding of how differentiating cells maintain their state and how they are directed to a desired cell type remains largely deficient.
It is widely recognized that Caenorhabditis elegans germline provides an attractive model system for studying the differentiation of stem cells in vivo. Specifically, C. elegans germline is organized in a simple linear fashion that progresses from germline stem cells (GSCs) at one end to maturing gametes at the other ( Figure 1A). Germ cells progress from GSCs at the distal end, through meiotic prophase as they move proximally to become differentiated gametes (sperm and oocytes) at the proximal end ( Figure 1A). This developmental process requires a battery of RNA regulators (Kimble and Crittenden, 2002; Figure 1B). One of the well-studied families of RNA regulators important for germ cell development is the PUF family of RNAbinding proteins. The PUF protein binds a specific regulatory element in its target mRNA 3 untranslated regions (3 UTRs) and inhibits the expression of its target mRNAs by recruiting translational repressor complexes . These include cytoplasmic Ccr4p-Pop2p-Not deadenylase complex (Goldstrohm et al., 2007) and Ago-eEF1A translational complex (Friend et al., 2012).
The C. elegans has multiple PUF proteins with specialized roles in germline and somatic tissues. Of those, three PUF proteins (FBF-1, FBF-2, and PUF-8) are highly expressed in the C. elegans germline and have critical roles in the maintenance of GSCs and mitotic germ cell fate. Specifically, FBF-1 and FBF-2 (collectively FBF) proteins are 95% identical, and they maintain GSCs by repressing the expression of genes that are associated with germline differentiation, including gld-1 (a KH-motif containing RNA-binding protein) , gld-2 [a poly(A) polymerase] (Millonigg et al., 2014), and gld-3 (a bicaudal-C homolog) (Eckmann et al., 2004; Figure 1B). Another C. elegans PUF protein, PUF-8 (a PUF with a striking similarity to human PUMILIO) controls multiple cellular processes such as proliferation, differentiation, and the sperm-oocyte decision, depending on the genetic context (Datla et al., 2014). It has also been reported that PUF-8 acts as a tumor suppressor by inhibiting GLP-1 (one of two C. elegans Notch receptors) (Racher and Hansen, 2012) and MPK-1 (C. elegans ERK/MAPK homolog) signaling pathways . Notably, many cancer cell lines circumvent PUF-mediated regulation of E2F transcription factor, a known oncogene that is dysregulated or overexpressed in cancer (Miles et al., 2012). Therefore, elucidating the biological function of PUF-8 and its target genes will provide insights into the proliferation and differentiation of stem cells as well as contribute to our understanding of tumorigenesis in other animals, including humans.
In this study, we have identified gld-2 as a direct target of PUF-8 repression in the C. elegans germline. Our genetic functional analyses showed that GLD-2 exhibits distinct functions depending on gene dosage in the absence of PUF-8. Under physiological conditions, two copies (+/+) of wild-type gld-2 gene promote the differentiation of GSCs by working with GLD-1. One dose (+/−) of wild-type gld-1 and gld-2 genes, however, in the absence of PUF-8 promotes the formation of germline tumors via the regression of spermatocytes into mitotic cells (dedifferentiation) by activating MPK-1. Collectively, these findings suggest that a regulatory network involving PUF-8 and its repressing target, GLD-2, can promote either differentiation or dedifferentiation of germ cells through GLD-1 and MPK-1, depending on gene dosage.

In silico Approach
Caenorhabditis elegans PUF-8 is a sequence-specific RNAbinding protein (Opperman et al., 2005). PUF-8 specifically binds to a regulatory element, termed the "PUF-8 binding element (PBE)" in target mRNA 3 UTRs (Opperman et al., 2005; Figure 1C). PUF-8 is most similar to human PUM2 Subramaniam and Seydoux, 2003). Human PUM2 protein also binds to the same binding sequences, called Nanos Response Element (NRE), in target mRNA 3 UTRs (Galgano et al., 2008;Bohn et al., 2018). Increasing evidence has shown that many genes with conserved PBEs were validated as PUF targets in vitro and in vivo (Prasad et al., 2016;Bohn et al., 2018). We thus performed an in silico approach to identify potential PUF-8 target genes from C. elegans whole genomes. Briefly, C. elegans 3 UTR sequences were obtained from BioMart, and we identified 800 genes (3.6%) harboring at least one PBE in their 3 UTRs ( Figure 1D and Supplementary Table S1). To investigate functional themes among the 800 potential PUF-8 targets, we used DAVID and PANTHER tools (Dennis et al., 2003;Huang da et al., 2009) to look for enriched categories of biological processes, as defined in the gene ontology (GO) database (Supplementary Table S2). The most enriched GO terms were related to cellular processes, developmental processes, post-translational modification, and cell cycle. Of those, we have focused on P granule-associated proteins (GO ID: 0043186) that function in reproduction (GO ID: 0000003) and meiosis (GO ID: 0051321) ( Figure 1E and Supplementary Table S3). Notably, gld-2 gene was nominated for a putative PUF-8 target included in our selected GO terms ( Figure 1E). The C. elegans gld-2 gene encodes a poly(A) polymerase that is critical for the germline differentiation (Kadyk and Kimble, 1998;Wang et al., 2002).

PUF-8 Binds a PBE in gld-2 3 UTR
The gld-2 3 UTR (1,099 bp) possess one highly conserved PBE (Figure 2A). To assess PUF-8 binding to the predicted gld-2 PBE, we used yeast three-hybrid assay as previously described ; Figure 2B). The yeast three-hybrid system is a useful tool in analyzing protein-RNA interaction. As previously described , a chimeric protein containing both a DNA-and RNA-binding domain tethers RNA to the promoter of a reporter gene. This protein consists of a LexA/MS2 coat protein fusion. A hybrid RNA binds to the MS2 coat protein via tandem MS2-binding sites. A hybrid RNA carrying the query sequence can bridge the LexA-MS2 and GAL4AD-PUF-8 hybrid proteins if PUF-8 binds, which is not possible if PUF-8 fails to bind. Yeast three-hybrid interactions were monitored by production of β-galactosidase from a lacZ reporter. The results indicate that PUF-8 interacts specifically with wild-type gld-2 PBE (gld-2 PBE wt ) and a positive control, hunchback NRE (hb NRE), in yeast three-hybrid assays ( Figure 2C and Supplementary Figure S1). By contrast, mutant gld-2 PBE (gld-2 PBE mut ) with an altered consensus (UGU changed to ACA) did not interact with PUF-8 ( Figure 2C). The strength of the PUF-8-gld-2 PBE interaction was determined by the expression of a HIS3 reporter As cells move proximally, they enter meiosis (green) and differentiate into either sperm (blue) or oocytes (pink). (B) Key RNA-binding proteins that control a balance between proliferation and differentiation. PUFs proteins (e.g., FBF-1/2) promote germ cell proliferation by inhibiting GLDs (e.g., GLD-1/2/3)-mediated germline differentiation (Kimble and Crittenden, 2002). However, PUF-8 controls both proliferation and differentiation, depending on genetic context (Datla et al., 2014). (C) Consensus sequence of PUF-8 binding element (PBE). (D) Pie chart of potential PUF-8 target genes (800, 3.6%) that contain at least one PBE. (E) Identification of gld-2 as a potential PUF-8 target mRNA involved in three gene ontology (GO) terms.
with upstream LexA operators. HIS3 expression confers growth on media without histidine and with 3-amino 1,2,3-triazol (3-AT) that is a competitive inhibitor of the HIS3-gene product. Notably, no significant growth was observed for gld-2 PBE mut strain, but significant growth of the positive control, hb NRE, and gld-2 PBE wt strains was detected at a 3-AT concentration of 4 and 8 mM, respectively ( Figure 2D). This result indicates that an interaction between PUF-8 and gld-2 PBE wt is strong. A direct interaction between PUF-8 and gld-2 PBE wt was determined by gel retardation assay ( Figure 2E and Supplementary Figure S2). Wild-type gld-2 PBE wt bound specifically to purified PUF-8 in gel shifts, but gld-2 PBE mut with an altered consensus did not interact with PUF-8. The apparent Kd value for PBE is about 110 nM. These results indicate that PUF-8 specifically binds to a PBE wt within gld-2 3 UTR. We also asked if a gld-2 PBE is conserved in another nematode species, Caenorhabditis briggsae gld-2 3 UTR, and in human Gld2 3 UTR. Intriguingly, the C. briggsae gld-2 mRNA has one conserved PBE and the human Gld2 mRNA has two conserved NREs in their 3 UTRs ( Figure 2F). Since these genes are highly conserved, we speculate that human PUM2 might also bind human Gld2 NRE, paving the way for an area of inquiry that warrants further pursuit.

PUF-8 Represses gld-2 mRNA Expression in vivo
PUF-8 expression was determined using a transgenic worm expressing a puf-8 (promoter):GFP:puf-8 cDNA:puf-8 3 UTR transgene (Ariz et al., 2009;Racher and Hansen, 2012; Figure 3A). In adult hermaphrodite germline, the GFP:puf-8 was expressed in the distal mitotic germ cells (Ariz et al., 2009;Racher and Hansen, 2012; Figure 3B). However, in adult male germline, the GFP:puf-8 was expressed in distal mitotic germ cells, spermatocytes, and sperm ( Figure 3C). Similar expression pattern was also observed in L4 staged spermatogenic hermaphrodite germline (Supplementary Figure S3). To test if PUF-8 might repress gld-2 expression in the distal mitotic germ cells, we have also generated a transgenic worm expressing a GFP:gld-2 3 UTR transgene in the germline. 3 UTRs control protein expression temporally and spatially. We fused a GFP reporter to the gld-2 3 UTR that contains a PBE and poly(A) signal sequences ( Figure 3D). GFP expression in the germline was visualized by staining dissected adult hermaphrodite gonads with an anti-GFP polyclonal antibody and DAPI. The GFP:gld-2 3 UTR was expressed at a low level in the distal mitotic germ cells, but was abundant in the differentiating meiotic cells [increased in the transition zone and became abundant in the pachytene ( Figure 3E)] and oogenic cells (data not shown). To ask whether PUF-8 inhibits gld-2 expression via its 3 UTR, we introduced puf-8(q725) putative null mutation [puf-8(−/−)] in GFP:gld-2 3 UTR transgenic worms. GFP expression in the germline was also visualized by staining dissected adult hermaphrodite gonads with an anti-GFP polyclonal antibody and DAPI. The expression levels were quantified using ImageJ software. Interestingly, the puf-8(−/−) mutant germlines siginifcantly accumulated GFP expression in the distal mitotic germ cells (Figures 3F,G). This difference was particularly striking within the distal mitotic germ cells, where GFP was about ∼25-fold higher in puf-8(−/−) mutants than in wild-type worms [puf-8(+/+)]. These data suggest that PUF-8 represses gld-2 mRNA expression via its 3 UTR in the distal mitotic germ cells.

gld-2 Hemizygosity Promotes Germline Tumors in the Absence of PUF-8
To assess the biological function of GLD-2 in the formation of puf-8(−/−) proximal germline tumors, we examined their FIGURE 2 | PUF-8 binds specifically to a PBE in gld-2 3 UTR. (A) A putative PBE in gld-2 3 UTR. Nucleotide sequences of a predicted PBE (see bold letters). Wild-type sequence is followed by its mutant, in which UGU is replaced by ACA. hunchback (hb) NRE (Nanos Response Element) served as a positive control for PUF-8 binding. (B) Schematic of yeast three-hybrid assay. (C) Three-hybrid interactions assayed by β-galactosidase activity. (D) HIS3 reporter activation. Growth was monitored on media lacking histidine and containing different concentration of HIS3 competitor 3-AT. (E) Gel retardation assay. Purified PUF-8 binds gld-2 PBE wt , but does not bind gld-2 PBE mut with an altered consensus as detailed in panel (A). (F) Sequence alignment of gld-2 PBEs from C. elegans, Caenorhabditis briggsae, and humans.

DISCUSSION
Differentiation programs of stem cells depend on gene expression largely regulated at the level of mRNAs. Recently, mRNA regulation has emerged as one of the key mechanisms that control the differentiation of stem cells into terminal cell types during animal development (Shigunov and Dallagiovanna, 2015). In this study, we have identified gld-2 as a PUF-8 target mRNA. Our genetic analyses demonstrated that PUF-8 and its repressing target GLD-2 promote germline differentiation by activating GLD-1 and inhibiting MPK-1 in the C. elegans germline (Figure 7). However, one dose of wild-type gld-2 and gld-1 genes [gld-2(+/−) gld-1(+/−)] in the absence of PUF-8 is insufficient for the terminal differentiation of spermatocytes, and instead promotes the formation of germline tumors via a dedifferentiation-like mechanism by activating MPK-1 (Figure 7). These findings suggest that a regulatory circuit involving PUF-8, GLD-1, GLD-2, and MPK-1 controls the program of germline differentiation or dedifferentiation depending on gene dosage and genetic context (Figure 7).

Gene Dosage Effects on Germ Cell Fate Specification
Among C. elegans PUF proteins, PUF-8 is the most similar to human PUMILIO Datla et al., 2014). PUF-8 controls multiple cellular processes during germline development, depending on genetic context (Datla et al., 2014). During early germline development, PUF-8 and MEX-3 (a KH domain translational regulator) contribute to the maintenance of GSCs by promoting mitotic proliferation (Ariz et al., 2009). However, PUF-8 also inhibits the proliferative fate of germ cells by inhibiting GLP-1/Notch signaling or by functioning parallel to it (Racher and Hansen, 2012). Once germ cells enter meiotic cell cycle, PUF-8 works together with LIP-1 to promote oocyte fate at the expense of sperm fate by repressing MPK-1 activation at permissive temperature (20 • C) (Subramaniam and Seydoux, 2003;Cha et al., 2012). Notably, PUF-8 and LIP-1 also inhibit the formation of germline tumors by promoting the meiotic completion of spermatocytes at restrictive temperature (25 • C). In spite of the well documented diverse functions of PUF-8, only a few PUF-8 targets have been identified to date (Mainpal et al., 2011). In the current study, in silico and biochemical analyses have identified gld-2 as a bona fide direct PUF-8 target mRNA (Figure 2). PUF-8 and GLD-2 have opposite biochemical and biological functions. While PUF-8 inhibits mRNA translation, GLD-2 activates it. GLD-2 also plays multiple roles by interacting with distinct RNA-binding protein partners; namely GLD-3 and RNP-8 (an RRM RNA binding protein). GLD-2-GLD-3 and GLD-2-RNP-8 exist as separate cytoplasmic poly(A) polymerase complexes, and they appear to have distinct RNA-binding specificities (Kim et al., 2010). Functionally, GLD-2-GLD-3 complex promotes meiotic entry and sperm fate, whereas GLD-2-RNP-8 complex specifies an oocyte fate (Kim et al., 2009). In particular, GLD-2-GLD-3 complex promotes meiotic entry by activating the translation of gld-1 mRNAs in the C. elegans germline (Suh et al., 2006). This report demonstrates that while one dose of wild-type gld-2 and gld-1 genes [gld-2(+/−) gld-1(+/−)] promotes meiotic entry, it nevertheless remains insufficient for the meiotic completion of spermatocytes in the absence of PUF-8. This eventually causes spermatocytes to revert back into mitotic cells by activating MPK-1, resulting in germline tumors ( Figure 7B). These RNA regulators (PUF-8, GLD-1, and GLD-2) play critical roles in RNA stability and its translation of numerous genes that are involved in key developmental and cellular processes. In addition, a precise interplay between these regulators at normal expressional levels also establishes a regulatory network for germ cell fate specification and homeostasis. For example, C. elegans FOG-1 is critical for sperm fate specification in the germline (Barton and Kimble, 1990). However, low FOG-1 levels [fog-1(+/−)] promote germline proliferation in the absence of FBF-1 and FBF-2 (C. elegans PUF proteins) (Thompson et al., 2005). Likewise, C. elegans FOG-3 (a homolog of vertebrate TOB/BTG) is also vital for sperm fate specification, but it can either promote or inhibit germline proliferation in a manner that is sensitive to both genetic context and gene dosage (Snow et al., 2013). Since the effects of gene dosage and genetic context on cell fate specification have recently emerged in vertebrate systems (Stefanovic and Puceat, 2007;Wang et al., 2010;Deo et al., 2013;Bankaitis et al., 2018), we thus suggest that gene dose-dependent control of cell fates may be conserved from worms to mammals, including humans.

Gradient-Mediated Cell Fate Decision in vivo
How are germ cell fates determined depending on dosage and genetic context? While it still eludes us, a suggested gradient model for cell fate decision is presented. Germ cell fates may be governed by relative levels of key regulators at a certain time and place. For example, at the distal end of the gonad, a somatic distal tip cell (DTC) provides a GSC niche and signals to the GSCs via the Notch signaling pathway. Notch signaling activates the transcription of target mRNAs, which are highly expressed in the mitotic cells but not in the meiotic cells. Well studied genes include sygl-1 and lst-1 (Kershner et al., 2014). These regulators work together with others (e.g., RNA regulators and cell cycle regulators) to build a regulatory network for mitotic cell fate. As the expression of these regulators is suppressed by a regulatory network for meiotic entry (early differentiation), germ cells enter meiotic cell cycle and their cell fates are maintained during differentiation. Importantly, regulators for mitosis and meiosis antagonize each other and generate an overlap area that may be critical for the mitosis-meiosis decision. Germ cells at meiotic cell cycle are then required to make the spermoocyte decision. Notably, many regulators for the mitosismeiosis decision also play critical roles in the sperm-oocyte decision, including FBFs, PUF-8, GLDs, NOS-3, FOGs, and MPK-1 (Kimble and Crittenden, 2002;Morgan et al., 2013). This finding suggests that a relative level (or ratio) of these regulators at a particular time and place may determine sperm or oocyte fate. In this study, we demonstrated that germ cells also decide whether to remain in a meiotic differentiating state or revert to being undifferentiated. Two key players, PUF-8 and GLD-1, regulate this decision with distinct mechanisms. In oogenic germline, GLD-1 is required for the maintenance of meiotic cell fate by regulating its target mRNAs. Thus, in the gld-1(−/−) oogenic germline, early meiotic cells return into mitotic cell cycle, resulting in germline tumors (Jones and Schedl, 1995). In contrast, PUF-8 inhibits the regression of spermatogenic germ cells into mitotically dividing cells only in spermatogenic germline (Subramaniam and Seydoux, 2003;Cha et al., 2012). It was previously reported that PUF-8 functions redundantly with GLD-1 to promote the meiotic progression of spermatocytes in C. elegans germline (Priti and Subramaniam, 2015). These results indicate that the expressional levels of PUF-8 and GLD-1 may govern the decision between the maintenance of the meiotic differentiating state and the regression to the mitotic undifferentiating state (also known as the differentiation-dedifferentiation decision). We here demonstrate that one dose of wild-type gld-2 and gld-1 genes disrupt the frame of the differentiation-dedifferentiation decision in the absence of PUF-8 (Figure 7). Although it still eludes us what ratio of these regulators influence decisions at a particular time and place, our data show that one dose of wild-type gld-2 and gld-1 genes [gld-2(+/−) gld-1(+/−)] are competent for meiotic entry, but are insufficient for the terminal differentiation of spermatocytes in the absence of PUF-8, resulting in dedifferentiation-mediated germline tumors (Figure 7). Notably, the Drosophila blastoderm and the vertebrate neural tube are typical examples of gradient-mediated cell fate decision spatially (Briscoe and Small, 2015). In both tissues, cell fate decision relies on molecular gradients. First, signaling gradients establish initial conditions. Second, these signaling gradients initiate transcriptional networks, including activators and repressors, to generate patterns of gene expression.
Third, the precise positioning of boundaries temporally and spatially determines commitment to specific cell types. Similarly, regulation of mammalian stem cell proliferation and cell fate decision relies on gradients of signaling molecules and an interplay between activators and repressors in specific tissue compartment boundaries (Du et al., 2015;Tian et al., 2016). This suggests that gradient-mediated cell fate decision may be an evolutionarily conserved mechanism from worms to humans. Collectively, our findings from the simple worm model may provide a novel insight into gradient (and/or gene dose)mediated cell fate decision in mammals, where such in vivo methods are not yet feasible or practical.
Feeding RNA Interference (RNAi) RNA interference experiments were performed by feeding bacteria expressing double-stranded RNAs corresponding to the gene of interest (Kamath et al., 2001;Ashrafi et al., 2003). RNAi bacteria were from C. elegans ORF-RNAi library (Source BioScience, Nottingham, United Kingdom). Synchronized L1 staged worms were placed on RNAi plates (a NGM plate containing 100 µg/mL Ampicillin and 0.5 mM IPTG) and incubated for 4 days at 20 or 25 • C.

EdU (5-Ethynyl-2 -Deoxyuridine) Labeling
To label mitotically cycling cells, worms were incubated with rocking in 0.2 mL M9 buffer (3 g KH 2 PO 4 , 6 g Na 2 HPO 4 , 5 g NaCl, 1 mL 1M MgSO 4 , H 2 O to 1 L) containing 0.1% Tween 20 and 1 mM EdU for 30 min at 20 • C. Gonads were dissected and fixed in 3% paraformaldehyde/0.1M K 2 HPO 4 (pH 7.2) solution for 10-20 min, followed by −20 • C methanol fixation for 10 min. Fixed gonads were blocked in 1× PBST/0.5% BSA solution for 30 min at 20 • C. EdU labeling was performed using the Click-iT EdU Alexa Fluor 488 Imaging Kit (Invitrogen, CA, United States, #C10337), according to the manufacturer's instructions. For costaining with antibodies, EdU-labeled gonads were incubated in the primary antibodies after washing for three times, and subsequently in the secondary antibodies as described above.
Yeast Three-Hybrid, 3-AT, and Gel Retardation Assays Three-hybrid assays were performed as previously described . The sequences for the 3 UTR region of gld-2 were cloned using the pIIIa/MS2-2 vector (provided by Dr. Wickens, University of Wisconsin-Madison). These vectors, containing the target sequences of gld-2 3 UTR, were cotransformed with PUF-8:pACTII vector into YBZ-1 yeast strain. The level of β-galatactosidase was assayed in at least three independent experiments. The strength of PUF protein-RNA interaction was determined by the 3-AT assay. The levels of 3-AT resistance were measured by assaying multiple transformants at four different concentrations of 3-AT, from 1 to 10 mM. Gel retardation assays were performed as previously described .

DATA AVAILABILITY STATEMENT
This manuscript contains previously unpublished data.
The name of the repository and accession number(s) are not available.

AUTHOR CONTRIBUTIONS
YP, SO, FT, and ML performed the experiments. All authors contributed the reagents, materials, and analysis tools. ML and MA designed the experiments, analyzed the data, and wrote the manuscript.

FUNDING
This work was supported in part by the NIH (GM112174-01A1), NIA (AG060373-01), and National Science Foundation (MCB1714264) to ML. The Caenorhabditis Genetic Center (CGC) was supported by the National Institutes of Health -Office of Research Infrastructure Programs (P40 OD010440).