Nanos promotes epigenetic reprograming of the germline by down-regulation of the THAP transcription factor LIN-15B

Nanos RNA-binding proteins are required for germline development in metazoans, but the underlying mechanisms remain poorly understood. We have profiled the transcriptome of primordial germ cells (PGCs) lacking the nanos homologs nos-1 and nos-2 in C. elegans. nos-1nos-2 PGCs fail to silence hundreds of transcripts normally expressed in oocytes. We find that this misregulation is due to both delayed turnover of maternal transcripts and inappropriate transcriptional activation. The latter appears to be an indirect consequence of delayed turnover of the maternally-inherited transcription factor LIN-15B, a synMuvB class transcription factor known to antagonize PRC2 activity. PRC2 is required for chromatin reprogramming in the germline, and the transcriptome of PGCs lacking PRC2 resembles that of nos-1nos-2 PGCs. Loss of maternal LIN-15B restores fertility to nos-1nos-2 mutants. These findings suggest that Nanos promotes germ cell fate by downregulating maternal RNAs and proteins that would otherwise interfere with PRC2-dependent reprogramming of PGC chromatin.


Introduction
In animals, formation of the germline begins during embryogenesis when a few cells (~30 in mice, two in C. elegans) become fated as primordial germ cells (PGCs) -the founder cells of the germline. PGC specification requires the activity of chromatin regulators that induce genome-wide changes in gene expression. For example, in mice, the transcriptional repressor BLIMP1 initiates PGC specification by blocking the expression of a mesodermal program active in neighboring somatic cells Saitou et al., 2005). In C. elegans, the NSD methyltransferase MES-4 and the Polycomb Repressive Complex (PRC2, including MES-2, 3 and 6) cooperate to place active and repressive histone marks on germline and somatic genes, respectively (Gaydos et al., 2012). Despite their critical roles during germ cell development, the BLIMP1 and MES/PRC2 chromatin regulators are not germline-specific factors and also function during the differentiation of somatic lineages (Cui et al., 2006;Gaydos et al., 2012;Seydoux and Braun, 2006). How the activities of these global regulators are modulated in germ cells to promote a germline-specific program is not well understood.
In C. elegans, genetic analyses have shown that MES-dependent activation of germline genes is antagonized in somatic lineages by a group of transcriptional regulators first identified by their effects on vulval development (Curran et al., 2009;Petrella et al., 2011;Unhavaithaya et al., 2002). Among these, components of the DRM (named for its Dp/E2F, pRB, and MuvB subunits) class of transcriptional regulators and LIN-15B, a THAP domain DNA binding protein, have been implicated in the silencing of germline genes in somatic cells Wu et al., 2012). Loss of DRM factors or LIN-15B causes ectopic activation of germline genes in somatic cells leading to growth arrest at elevated temperatures (26˚C). Inactivation of mes-2, mes-3, mes-4 and mes-6 suppresses the ectopic germline gene expression and restores viability to lin-15B mutants at 26˚C . These observations have suggested that DRM factors and LIN-15B antagonizes MES activity in somatic lineages to keep germline genes off . A similar antagonism, but in reverse, has been uncovered in the adult germline between the NSD methyltransferase MES-4 and the DRM transcription factor LIN-54 (Tabuchi et al., 2014). The X chromosome is a major focus of MES repression in C. elegans germline. The X chromosome is silenced throughout germ cell development except in oocytes, which activate the transcription of many X-linked genes (Kelly et al., 2002). mes mutants prematurely activate the transcription of X-linked genes in pregametic germ cells leading to germ cell death (Bender et al., 2006;Gaydos et al., 2012;Seelk et al., 2016). Reducing the function of the synMuvB transcription factor lin-54 in mes-4 mutants restores the expression of X-linked genes closer to wild-type levels (Tabuchi et al., 2014). Therefore in the germline, MES activity antagonizes DRM activity to keep the X chromosome silent. Together, these genetic studies suggest that competition between the MES chromatin modifiers and the DRM/LIN-15B transcription factors balance the transcription of somatic and germline genes in somatic and germline tissues. How this competition is biased during development to ensure appropriate gene expression in each tissue is not known. In this study, we have discovered a link between Nanos and LIN-15B that provides an explanation for how MES activity becomes dominant in the nascent germline ( Figure 1A).
The C. elegans PGCs arise early in embryogenesis from pluripotent progenitors (P blastomeres) that also generate somatic lineages ( Figure 1B). RNA polymerase II activity is repressed in the P lineage until the 100 cell stage when the last P blastomere P 4 divides to generate Z2 and Z3, the two PGCs (Seydoux et al., 1996). RNA polymerase II becomes active in PGCs, but these cells remain relatively transcriptionally quiescent, and exhibit reduced levels of active chromatin marks compared to somatic cells throughout the remainder of embryogenesis (Kelly, 2014). Active marks and robust transcription return after hatching when the L1 larva begins to feed and the PGCs resume proliferation in the somatic gonad (Fukuyama et al., 2006;Kelly, 2014). The mechanisms that maintain PGC chromatin in a silenced state during embryogenesis are not known, but embryos lacking the nanos homologs nos-1 and nos-2 have been reported to display abnormally high levels of the active mark H3meK4 mark in PGCs (Schaner et al., 2003). nos-1nos-2 PGCs initiate proliferation prematurely during embryogenesis and die during the second larval stage (Subramaniam and Seydoux, 1999). Nanos proteins are broadly conserved across metazoans and have been shown to be required for PGC survival in several phyla, from insects to mammals (Asaoka-Taguchi et al., 1999;Beer and Draper, 2013;Deshpande et al., 1999a;Lai et al., 2012a;Tsuda et al., 2003). Nanos proteins are cytoplasmic RNA-binding proteins that regulate gene expression post-transcriptionally by recruiting eLife digest Every new embryo inherits a set of starting instructions from its mother. These instructions are called a 'maternal dowry' and help a fertilized egg through the first few stages of development. Later, the maternal dowry is removed so that the embryo's genetic instructions can take over.
In animals, some of the cells in this early embryo become specialized to produce eggs (technically called oocytes) or sperm. These cells are called germ cells, and they are needed for reproduction. A protein called Nanos helps germ cells become different to other cells, but it is not clear how Nanos has this effect.
Lee et al. studied Nanos in the embryos of the worm Caenorhabditis elegans. In many ways, early development is the same in the worm as in many other animals. By examining worms that did not have Nanos, Lee et al. showed that germ cells without Nanos do not lose their maternal dowry. As a result, the cells still contain a molecule called LIN-15B, which makes other types of cells in the worm. Ultimately, without Nanos, the germ cells do not develop and die leaving the worm sterile.
Germ cells are essential for living things to reproduce and have children. Understanding how they are created can teach scientists a lot about how embryos develop before birth. This could eventually help to boost fertility in endangered species or to treat human sterility. modifiers balances activities that promote somatic (red) and germline (green) gene expression. In somatic lineages, LIN-15B and DRM transcription factors opposes PRC2 to silence germline genes . In pregametic germ cells, PRC2 activates germline genes (with the help of MES-4) and silences somatic genes and X-linked genes, including genes expressed in oocytes (e.g. lin-15B) (Gaydos et al., 2012). In this study, we show that Nanos is required to remove maternally-inherited LIN-15B from PGCs to allow proper PRC2/MES-4 function. (B) Germline cycle in C. elegans. The zygote (stripped red and green) inherits from the oocyte maternal mRNAs that promote the development of somatic (red) and germline (green) cell fates during embryogenesis. P 4 is the germline founder cell that gives rise to Z2 and Z3, the two primordial germ cells. Z2 and Z3 do not divide and remain mostly transcriptionally quiescent during embryogenesis. They resume division and transcription in the first larval stage (L1). (C-D). Transcriptome comparison between PGCs isolated from wild-type and nos-1(gv5)nos-2(RNAi) L1 larvae using SMART-seq libraries (Materials and methods, See Figure 1-figure supplement 1B-C for results with Truseq libraries). (C) Volcano plot showing log2 fold-change in transcript abundance for each gene. The number of genes that were significantly up or downregulated in nos-1(gv5)nos-2(RNAi) (designated as nos-1/2) PGCs are indicated. Figure 1 continued on next page effector complexes that silence and degrade mRNAs in the cytoplasm. Six direct Nanos mRNA targets have been identified to date [Drosophila hunchback, cyclin B and hid (Asaoka-Taguchi et al., 1999;Dalby and Glover, 1993;Kadyrova et al., 2007;Murata and Wharton, 1995;Sato et al., 2007;Wreden et al., 1997), Xenopus VegT (Lai et al., 2012a), and sea urchin CNOT6 and eEF1A (Oulhen et al., 2017;Swartz et al., 2014)]), but none of these targets are sufficient to explain how Nanos activity might affect PGC chromatin. In this study, we characterize the gene expression defects of PGCs lacking nanos activity in C. elegans. Our findings indicate that nanos activity is required in PGCs to silence maternal transcripts inherited from the oocyte. We identify maternal lin-15B as a critical target of Nanos regulation that must be turned-over to establish MES dominance in PGCs.

Figure 1 continued
Dashed line marks the significance cutoff of q = 0.05 (Y axis) above which genes were counted as misexpressed. (D) Bar graphs showing expected and observed number of genes (Y axis) in different expression categories (X axis). Genes were assigned to a particular expression category based on their preferential expression patterns as determined in (Gaydos et al., 2012;Ortiz et al., 2014); Materials and methods and Supplementary file 1). The lists are non-overlapping and include 1694 pregamete genes, 1594 oocyte genes, 2042 sperm genes, and 2684 somatic genes. Oocyte genes include genes required for oogenesis and maternal genes required for embryogenesis. Because genes were categorized based on their preferential gene expression pattern, genes on one list may also be expressed in other tissues. tbx-2, and hnd-1 transcripts were upregulated 5.1-fold, 11-fold and 91-fold, respectively, in nos-1 (gv5)nos-2(RNAi) PGCs (Supplementary file 4). We conclude that nos-1(gv5)nos-2(RNAi) PGCs overexpress oogenic and maternal genes normally expressed in oocytes, and fail to activate pregametic genes normally expressed in PGCs.
Turnover of maternal transcripts is delayed in PGCs lacking nos-1 and nos-2 The overexpressed oocyte-class transcripts in nos-1(gv5)nos-2(RNAi) L1 PGCs could correspond to maternal transcripts that failed to turnover during embryogenesis or to zygotic transcripts synthesized de novo in nos-1(gv5)nos-2(RNAi) PGCs. To distinguish between these possibilities, we first examined the fate of maternal RNAs in PGCs. We isolated PGCs from embryos with fewer than 200 cells, at a time when PGCs are still mostly transcriptionally silent (EMB PGCs) (Schaner et al., 2003;Seydoux and Dunn, 1997). By comparing the EMB PGC transcriptome to a published oocyte transcriptome (Stoeckius et al., 2014), we observed an excellent correlation in relative transcript abundance between oocytes and EMB PGCs (Figure 2-figure supplement 1A), suggesting that the transcriptome of EMB PGCs consists primarily of maternal mRNAs inherited from the oocyte. This finding is consistent with in situ hybridization experiments that showed that many maternal RNAs persist in the embryonic germ lineage at least to the P 4 germline founder cell (Seydoux and Fire, 1994). Next, we compared the transcriptome of EMB PGCs to that of L1 PGCs to identify PGC transcripts whose abundance decline during embryogenesis. We identified 411 down-regulated transcripts, including 197 oocyte transcripts ( Figure 2A and B and Supplementary file 5), consistent with turnover of many maternal mRNAs in PGCs after the 200 cell stage. Strikingly, the amplitude of this turnover was diminished in nos-1(gv5)nos-2(RNAi) mutants: the abundance of the 411 transcripts remained high overall during the transition from EMB PGCs to L1 PGCs in nos-2(RNAi)nos-1(gv5) embryos ( Figure 2C). Furthermore, when comparing wild type and nos-1(gv5)nos-2(RNAi) EMB PGCs, we identified 182 differentially expressed transcripts (11 down-and 171 upregulated), including 71 of oocyte transcripts that were more abundant in nos-1(gv5)nos-2(RNAi) EMB PGCs (Figure 2-figure supplement 1B). Together these findings suggest a defect in maternal mRNA turnover in nos-1(gv5)nos-2(RNAi) PGCs that is already detectable at the 200 cell stage and persists through embryogenesis.
PGCs lacking nos-1 and nos-2 activate the transcription of many genes normally silent in PGCs By the L1 stage, PGCs are transcriptionally active. To explore whether inappropriate transcription also occurs in nos-1nos-2 PGCs by the L1 stage, we examined transcripts that increase in abundance during the transition from EMB and L1 PGCs. We identified 130 such transcripts in wild-type PGCs, including 30% in the pregamete category (39/130, Figure 2B), consistent with PGCs transitioning to a pregamete fate by the L1 stage. In contrast, in nos-1(gv5)nos-2(RNAi) PGCs, many more (510) transcripts increased in abundance during embryogenesis, and these were distributed among all genes categories, including oocyte genes (16%, 84/510 Figure 2E and F and Supplementary file 5). This finding suggests that, unlike wild-type PGCs, nos-1(gv5)nos-2(RNAi) PGCs fail to transition to a pregamete program and instead adopt a hybrid transcriptional profile that includes activation of oocyte genes.
PGCs lacking mes-2 or mes-4 upregulate X-linked genes Failure to silence X-linked genes has been reported for germ cells lacking the chromatin regulators mes-2 and mes-4 (Bender et al., 2006;Gaydos et al., 2012). To directly compare the effect of loss of nos versus mes function in PGCs, we purified PGCs from L1 larvae derived from hermaphrodites where mes-2 or mes-4 was inactivated by RNAi (Materials and methods). As expected, loss of mes-2 and mes-4 led to a significant upregulation of X-linked genes in L1 PGCs ( Figure 4A-B, Figure 4figure supplement 2, and Supplementary file 5 for lists of misregulated genes). To directly compare these changes to those observed in nos-1(gv5)nos-2(RNAi) PGCs, we compared, for each genotype, the log2 fold change over wild-type for X-linked genes and for autosomal oocyte genes. As expected, we observed a strong positive correlation between mes-2 and mes-4 in both gene categories (R = 0.91 and R = 0.76, X-linked and autosomal oocyte genes, respectively) ( Figure 4C and D). We also observed a strong correlation between mes-4(RNAi) and nos-1(gv5)nos-2(RNAi) (R = 0.75, Figure 4E, Figure 4-figure supplement 1) and mes-2(RNAi) and nos-1(gv5)nos-2(RNAi) (R = 0.73, Figure 2 continued expression categories (X axis). The lists of expression categories used here are the same as those in Figure 1B (Supplementary file 1). Oocyte and soma genes are overrepresented among downregulated genes and pregamete genes are overrepresented among upregulated genes. Asterisks indicate significantly more genes than expected (hypergeometric test, p-value<0.001 [**]). (C) Box and whisker plot showing the expression levels (log10) of 411 genes that are downregulated during embryogenesis in wild-type PGCs. Expression of these genes remains high on average in nos-1 (gv5)nos-2(RNAi) PGCs (designated as nos-1/2). Each box extends from the 25th to the 75th percentile, with the median indicated by the horizontal line; whiskers extend from the 2.5th to the 97.5th percentiles. (D) Photomicrograph of embryos hybridized with single molecule fluorescence probes (red) against mex-5, C01G8.1 and Y51F10.2. Wild-type and nos-1(gv5)nos-2(ax3103) embryos were raised at 25˚C and are compared here at the same stage (as determined by the number of DAPI-stained nuclei shown in blue). By the stages shown, all three transcripts have turned over in wild-type (N2), but are still present (red signal) in PGCs in nos-1(gv5)nos-2(ax3103) embryos. At least 10 embryos were examined per probe set in different genotypes shown. (E-F) Transcriptome comparison between PGCs isolated from nos-1(gv5)nos-2(RNAi) embryos and nos-1(gv5)nos-2(RNAi) (designated as nos-1/ 2) L1 larvae. (E) Volcano plot showing log2 fold-change in transcript abundance for each gene. The numbers of genes whose expression were up or downregulated in L1 PGCs compared to embryonic PGCs are indicated. Dashed lines mark the significance cutoff of q = 0.05 above which genes were counted as misexpressed. (F) Bar graphs showing expected and observed number of genes (Y axis) in the different expression categories (X axis). Asterisks indicate significantly more genes than expected (hypergeometric test, p-value<0.001 [**]). DOI: https://doi.org/10.7554/eLife.30201.005 The following source data and figure supplements are available for figure 2: not shown) for X-linked genes. Interestingly, the correlations were weaker for autosomal oocyte genes (R = 0.35, Figure 4F), which tended to be more overexpressed in nos-1(gv5)nos-2(RNAi) L1 PGCs. This finding is consistent with the notion that, while nos-1nos-2 and mes PGCs share a defect in the silencing of X-linked loci, nos-1nos-2 PGCs also have an additional defect in maternal mRNA turn over.

nos-1(gv5)nos-2(ax3103); lin-15B(n744)
The following figure supplements are available for figure 5: PGCs, lin-15B transcript levels rose by~2 fold in nos-1(gv5)nos-2(RNAi) embryos but not in wild-type embryos, suggesting that the lin-15B locus is also inappropriately transcribed in nos-1(gv5)nos-2 (RNAi) L1 PGCs. Unfortunately, we were not able to confirm these RNA-seq observations by in situ hybridization due to the low abundance of lin-15B RNA and its presence in all somatic cells.
To determine whether LIN-15B protein expression is under the control of Nanos activity, we stained embryos derived from nos-1(gv5)nos-2(ax3103) hermaphrodites with the anti-LIN-15B antibody. We found that, in contrast to wild-type, nos-1(gv5)nos-2(ax3103) embryos maintained high LIN-15B levels in embryonic PGCs ( Figure 6B Right panels). nos-1(gv5)nos-2(ax3103) embryos could misregulate LIN-15B by delaying the turnover of maternal LIN-15B or by activating premature zygotic transcription of the lin-15B locus. To distinguish between these possibilities, we created a lin-15B transcriptional reporter by inserting a GFP::H2B fusion at the 5' end of lin-15B locus in an operon configuration to preserve endogenous lin-15B expression ( Figure 6-figure supplement 1 and Supplementary file 6). We crossed nos-1(gv5) males carrying the lin-15B transcriptional reporter to wild-type or nos-1(gv5)nos-2(ax3103) hermaphrodites and examined crossed progenies for GFP expression. In both cases, we observed strong GFP expression in somatic cells, but no expression in PGCs during embryogenesis (data not shown), indicating that zygotic expression of LIN-15B in PGCs is not activated in embryogenesis in either wild-type or nos-1(gv5)nos-2(ax3103) embryos. In wildtype, we first observed zygotic expression of the lin-15B transcriptional reporter in the germline of L4 stage animals ( Figure 6-figure supplement 1), in germ cells that have initiated oogenesis. In contrast, in animals derived from nos-1(gv5)nos-2(ax3103) mothers, zygotic expression of the lin-15B transcriptional reporter could be detected as early as the L1 stage in PGCs and their descendants ( Figure 6C). This expression was maintained until the L2 stage when nos-1nos-2 PGC descendants undergo cell death. We conclude that nos-1nos-2 activity is required both to promote the turnover of maternal LIN-15B in PGCs during embryonic development and to prevent premature zygotic transcription of the lin-15B locus in PGCs in the L1 stage.

Discussion
In this study, we have examined the transcriptome of PGCs lacking Nanos function in C. elegans. We have found that PGCs lacking Nanos overexpress 100 s of transcripts normally expressed in oocytes.
Our observations indicate that this defect is due to both delayed turn-over of maternal mRNAs and inappropriate transcription of genes normally silent in PGCs, including oocyte genes. We identify LIN-15B as a critical maternally-inherited factor that must be turned-over by Nanos in PGCs to prevent inappropriate transcription. We propose that clearing of maternal LIN-15B by Nanos allows the PRC2/MES-4 network of chromatin modifiers to reprogram PGCs away from an inherited (maternal) oocyte program to a pregametic germ cell program.

Nanos activity is required for the timely turnover of maternal mRNAs in PGCs
During oogenesis, oocytes stockpile mRNAs and proteins in preparation for embryogenesis. These include mRNAs and proteins with housekeeping functions as well as factors required to specify embryonic cell fates (somatic and germline). During embryogenesis, maternal products are eventually turned over to make way for zygotic factors (maternal-to-zygotic transition). Our findings suggest that Nanos facilitates this transition in PGCs by accelerating the turnover of maternal mRNAs. Nanos family members are thought to silence mRNAs by interacting with the sequence-specific RNA-binding protein Pumilio and with the CCR4-NOT deadenylase complex, which interferes with translation and can also destabilize RNAs. (Lai et al., 2012a;Suzuki et al., 2012;Swartz et al., 2014;Wharton et al., 1998). In the C. elegans genome, there are eight genes related to Drosophila pumilio. Depletion of five of these (fbf-1, fbf-2, puf-6, puf-7 and puf-8) phenocopies the nos-1nos-2 PGC phenotypes, including failure to incorporate in the somatic gonad, premature proliferation, and eventually cell death (Subramaniam and Seydoux, 1999). These observations suggest that NOS-1 and NOS-2 function with Pumilio-like proteins to repress the translation of certain maternal RNAs. Paradoxically, in sea urchins, Nanos silences the mRNA coding for the CNOT6 deadenylase, which indirectly stabilizes other maternal mRNAs (Swartz et al., 2014). In that system, Nanos was also found to silence eEF1A expression, leading to a transient period of translational quiescence in PGCs (Oulhen et al., 2017). In combination, these effects could promote the turnover of maternal mRNAs and proteins that promote somatic development (e.g. LIN-15B) while preserving germline mRNAs (e.g. mes) whose translation could be reactivated at a later time. In C. elegans, the redundant nanos homologs nos-1 and nos-2 are expressed sequentially in PGCs during the maternal-to-zygotic transition and may have distinct targets. Genetic analyses already have suggested that nos-1 and nos-2 have both shared and unique functions (Kapelle and Reinke, 2011;Mainpal et al., 2015). It will be important to determine whether nos-1 and nos-2 both target lin-15B, and whether they do so directly, by silencing lin-15B mRNA translation, or indirectly, by silencing other factors required for LIN-15B protein translation and/or stability.

In the absence of Nanos, maternal LIN-15B interferes with MESdependent reprogramming of PGC chromatin
Two lines of evidence indicate that LIN-15B is responsible for much of the abnormal gene expression observed in nos-1nos-2 PGCs by the first larval stage. First, elimination of maternal LIN-15B restores fertility to nos-1nos-2 mutants and lessens the upregulation of many misregulated genes ( Figure 6). Second, LIN-15B is a known genetic antagonist of MES function in somatic cells Wang et al., 2005), and PGCs lacking mes activity upregulate many of the same genes upregulated in nos-1nos-2 PGCs. The strongest correlation is seen for genes on the X chromosome ( Figure 4E, Figure 4-figure supplement 1), a well-documented focus of MES-dependent silencing (Bender et al., 2006;Garvin et al., 1998;Gaydos et al., 2012). Together these findings indicate that failure to downregulate maternal LIN-15B interferes with MES-dependent reprogramming of PGC chromatin and is the primary cause of PGC death in Nanos mutants.
The lin-15B locus is on the X chromosome and is ectopically transcribed in nos-1nos-2 PGCs at hatching. These observations raise the possibility that maternal LIN-15B potentiates zygotic lin-15B expression as MES-dependent silencing of the X-chromosome becomes compromised. How LIN-15B opposes MES activity is not known, but another X-linked gene, utx-1, may oppose MES activity directly. utx-1 encodes a de-methylase that removes the silencing mark deposited by the PRC2 complex. Upregulation of utx-1 was shown recently to promote reprogramming of adult germline stem cells into neurons (Seelk et al., 2016). In nos-1nos-2 PGCs, utx-1 is upregulated in a lin-15B-dependent manner, and RNAi of utx-1 partially suppresses nos-1nos-2 sterility (Figure 5-figure supplement 1). Suppression by loss of utx-1 is weaker than that observed when inactivating lin-15B, suggesting that utx-1 is not the only lin-15B target that opposes PRC2. Loss of two other synMuvB genes, lin-35/Rb and dpl-1, also suppresses nos-1nos-2 sterility ( Figure 5A), albeit again less stringently than loss of lin-15B. It will be interesting to determine whether these genes function with, or in parallel to, LIN-15B to oppose PRC2 activity in PGCs.
Recently, nos-2 was shown to function redundantly with xnd-1 to repress histone active marks in PGCs (Mainpal et al., 2015). XND-1 is a chromatin-associated protein expressed in PGCs throughout embryogenesis. An exciting possibility is that XND-1 directly activates MES-dependent remodeling in PGCs. In that context, Nanos could promote germ cell fate simply by eliminating any maternal factors that would interfere with that remodeling. Because PGCs derive from embryonic blastomeres that also give rise to somatic lineages, they inherit many transcripts with somatic functions. In addition to LIN-15B, we have found that Nanos accelerates the turn-over of several maternal mRNAs coding for transcription factors that function in somatic embryonic lineages, including pha-4, hlh-1 and tbx-2 (Supplementary file 4). We speculate that perdurance of these somatic transcription factors contributes to the complex transcriptional profile of nos-1nos-2 PGCs. The primary function of Nanos may be, therefore, to clear the PGCs of any mRNAs that promote somatic development. This interpretation is consistent with previous studies in Drosophila and Xenopus that reported the expression of somatic transcripts in PGCs lacking Nanos (Deshpande et al., 1999b;Hayashi et al., 2004;Kadyrova et al., 2007;Lai et al., 2012b;Oulhen et al., 2017;Swartz et al., 2014).
Nanos functions in an ancient regulatory switch that controls somatic and germline fates throughout development?
Our genetic findings indicate that, in PGCs, Nanos opposes LIN-15B and DRM transcription factors. Studies in Drosophila and mammals have reported that, in somatic cells, the reverse is true: DRM transcription factors silence Nanos. Loss of the DRM subunit lethal (3) malignant brain tumor [l(3) mbt] leads to tumorous growth in Drosophila imaginal disks and ectopic expression of germline genes, including nanos (Janic et al., 2010). Similarly, loss of the retinoblastoma transcription factor (Rb) leads to activation of nanos transcription in mammalian tissue culture cells and in Drosophila wings, which in turn is thought to repress the translation of Rb targets . A complex regulatory feedback loop has also been reported between the LSD1 demethylase and the Nanos partner Pumilio in Drosophila and human bladder carcinoma cells (Miles et al., 2015). Taken together, these observations suggest that Nanos functions in an ancient transcriptional/post-transcriptional regulatory switch that controls gene expression during development. Key questions for the future will be to understand how the switch is activated in the embryonic germline to favor germ cell development (what turns on Nanos expression in PGCs?), how the switch is flipped back during oogenesis to favor somatic development (what activates LIN-15B expression in oocytes and embryos?), and how the switch becomes deregulated in malignancies.

Continued on next page
(100 mg/mL) media for 5-6 hr, induced with 5 mM IPTG for 30 min, plated on NNGM (nematode nutritional growth media)+ampicillin (100 mg/mL)+IPTG (1 mM) plates, and grown overnight at room temperature. Embryos isolated by bleaching gravid hermaphrodites, or synchronized L1s hatched in M9, were put onto RNAi plates. For sterility counts, the progeny of at least six gravid adult hermaphrodites were tested. Adult progenies were scored for empty uteri ('white sterile' phenotype) on a dissecting microscope. For all immunostaining and smFISH experiments shown in Figures 2D,  6A  To verify the efficiency of RNAi treatments used to create sequencing libraries, we scored animals exposed to the same RNAi feeding conditions for maternal-effect sterility. For nos-1(gv5) strain on nos-2 RNAi, sterility was 81 ± 10% at 20˚C and 86 ± 6% at 25˚C; mes-2(RNAi) maternal effect sterility was 51 ± 1.4% and mes-4(RNAi) maternal effect sterility was 95.5 ± 3.5%. To test the efficiency of the double RNAi treatment for nos-1(gv5)nos-2(RNAi);lin15B(RNAi) RNA-seq libraries, we performed two additional controls. First we exposed a nos-2::FLAG strain (Paix et al., 2014) to the same RNAi feeding conditions and stained the embryos with a-FLAG antibody to confirm knock down of nos-2 (4/15 embryos showed weak staining, compared to 15/15 embryos with strong staining in the untreated controls). Second, we exposed a lin-15B::GFP strain (Paix et al., 2014) to the same double RNAi feeding conditions and observed no GFP expression in embryos. nos-1(gv5)nos-2(RNAi);lin15B (RNAi) animals gave 34 ± 19% sterile progenies.

Confocal microscopy
Fluorescence microscopy was performed using a Zeiss Axio Imager with a Yokogawa spinning-disc confocal scanner. Images were taken and stored using Slidebook v6.0 software (Intelligent Imaging Innovations) using a 40x or 63x objective. Embryos were staged by DAPI-stained nuclei in optical Z-sections and multiple Z-sections were taken to include germ cells marked by a-PGL-1 (K76) staining. For images of embryonic PGCs, a single Z-section was extracted at a plane with the widest area of DAPI staining for nuclear signal of LIN-15B, MES-3, and MES-4. For MES-2-GFP, the Z-section was determined based on widest area of GFP signal. Equally normalized images were first taken by Slidebook v6.0, and contrasts of images were equally adjusted between control and experimental sets using Image J.

Germ cell isolation and sorting
RNAi treatments for sorting experiments were done by seeding synchronized L1 (hatched from embryos incubated in M9 overnight) onto RNAi plates and growing them to gravid adults. Additional RNAi or control bacteria were added once to ensure enough food to support development. Early embryos were harvested from gravid adults. These embryos were either used directly to isolate embryonic PGCs or incubated for 12~16 hr in M9 solution until reaching the L1 stage for PGCs isolation. To isolate L1 PGCs from fed animals, the L1s were plated onto RNAi plates for additional 5 hr before processing for PGC isolation. For RNA-seq experiments described in Figure 1 and Figure 2, RNAi treatments were done at 25˚C. For the rest of RNA-seq experiments, RNAi treatments were done at 20˚C. See Supplementary file 7 for sequencing library information.
To isolate PGCs from embryos, cell dissociation was performed as described in Strange et al. (2007) (Strange et al., 2007) with the following modifications: 10 6 embryos were treated in 500 ml chitinase solution (4.2 unit of chitinase (Sigma # C6137) in 1 ml of conditioned egg buffer). After chitinase treatment, embryos were collected by centrifugation at~900 xg for 4 mins at 4˚C and resuspended in 500 ml accumix-egg buffer solution for dissociation (Innovative Cell Techologies, AM105, 1:3 dilution ratio in egg buffer). In the final step, cells were resuspended in chilled egg buffer before sorting using BD FACSAriaII. 65,000~120,000 PGL-1::GFP PGCs were used for RNA isolation.
To isolate PGCs from L1 larvae, total of >5 million L1 divided into~500,000 L1 per reaction were used for cell dissociation as described in Zhang and Kuhn (Zhang and Kuhn, 2013) (www.wormbook.org/chapters/www_cellculture/cellculture.html#sec6-2) with the following modifications: starved and fed (for 5 hr) L1 were incubated with freshly thawed SDS-DTT solution for 2 min and 3 min, respectively, with gentle agitation using a 1000 ml pipette tip. Pronase treatment was performed using 150 ml of 15 mg/ml pronase (Sigma P6911). Pronase treatment was stopped by adding 1000 ml conditioned L-15 medium and spin at 1600xg for 6 min. Cells were resuspended in chilled egg buffer and washed three times to remove debris before sorting using BD FACSAriaII or Beckman Coulter MoFlo sorter.~75,000 sorted cells were pelleted at 1600xg for 5 min, snap frozen and saved in À80˚C for later RNAseq analysis.
To assay the purity of isolated PGCs, aliquot of sorted PGCs were either passed through FACS sorter again to re-analyzed their GFP expression or subjected to GFP positive cell counting under microscope. PGC purity is defined by the percentage of GFP positive and propidium iodide negative in the sorted population. The purity of sorted embryonic PGCs is 95.7 ± 3.8% (N = 3); The purity of sorted L1 PGCs is 94.7 ± 4.7% (N = 10). From purified embryonic cells, we identified 1347 PGC enriched genes (enrichment over somatic blastomeres). We cross-reference our embryonic PGC enriched gene set with other published PGC or germline enriched gene sets. 392/1347 embryonic PGC enriched genes were identified as PGC enriched genes in Spencer et al., 2011 (in which 979 genes with enriched expression in Z2/Z3); 700/1347 were characterized as either germline specific or germline enriched genes in Gaydos et al., 2012. The result is summarized in Supplementary file 8. The reproducibility of sorting/RNAseq procedure is demonstrated by PCA analysis as described in the section of RNAseq library preparation and analysis.

RNA extraction
RNA was extracted from sorted cells using TRIZOL. The aqueous phase was transferred to Zymo-SpinTM IC Column (Zymo research R1013) for concentration and DNase I treatment as described in manual. RNA quality was assayed by Agilent Bioanalyzer using Agilent RNA 6000 Pico Chip. All RNAs used for library preparation had RIN (RNA integrity number)>8.

RNAseq library preparation and analysis
Three different RNA-seq library preparation methods were used for this study: SMART-seq, which uses poly-A selection (Figures 1 and 2), NuGEN Ovation, which uses random priming (Figure 2figure supplement 1, top), and Truseq combined with Ribozero to remove ribosomal RNAs (all other figures). The first two methods allow library construction from <10 ng of total RNA, whereas the later method requires >50 ng total RNA. We compared SMART-seq and Truseq-Ribozero performance on L1 PGCs isolated from wild-type and nos-1(gv5)nos-2(RNAi) and observed identical trends, with an overall higher number of misregulated genes identified with Truseq-Ribozero (Compare Figure 1 (SMART-seq) and Figure 1-figure supplement 1B, C and D (Truseq/Ribozero). For the experiment shown in Figure 2-figure supplement 1 (top panels) where we compared RNA levels between embryonic PGCs and an oocyte library reference, we used Nugen Ovation libraries which avoids any bias due to poly-A selection while allowing library construction from <3 ng of RNA. For all experiments, control and experimental libraries were made using the same method. Supplementary file 5 contains lists of misregulated genes from analyses. Supplementary file 7 lists all the RNA-seq libraries used in this study and the corresponding figures.
SMART-seq libraries: libraries were made from 2 ng of total RNA isolated from sorted PGCs from worms grown at 25˚C. Libraries were constructed using SMART-seq v4 Ultra Low input RNA kit (Clontech, Cat. No. 634888) followed by Low Input Library Prep Kit (Clontech, Cat. No. 634947). The cDNAs were then fragmented using Covaris AFA system at the Johns Hopkins University microarray core and cloned using the Low Input library prep Kit.
NuGEN Ovation libraries: libraries were made from 3 ng of total RNA isolated from sorted cells from worms grown at 25˚C. Libraries were constructed using Nugen Ovation system V2 (#7102-08) followed by Nugen Ultralow library system.
TruSeq libraries: 50 ng of total RNA isolated from sorted PGCs from L1 worms grown at 20˚C were subjected to Ribozero kit (illumina, MRZE706) to remove rRNA. Libraries were constructed using Truseq Library Prep Kit V2.
For correlation plots of gene expression shown in Figure 4C-F, information from different pairs of cuffdiff analyses (WT vs mes-2, WT vs mes-4 and WT vs nos-1/2) was used. Genes with sufficient aligned reads to pass statistical test (OK status in test status from cuffdiff output) were kept, and those without enough alignments (NOTEST, LOWDATA in test status), or other conditions prevent statistical testing were excluded. Values of Log2 fold change were extracted from each cuffdiff output file and list of genes were further consolidated to generated correlation plots. The data process results in different number of genes in selected categories (1173 vs 1250 in X-linked genes, and 1063 vs 1092 in autosomal oocyte genes). However, majority of genes were overlapped between comparisons (1117 for X-linked genes and 1062 for autosomal oocyte genes) In Figure 6F, the area-proportional Venn diagram was created using the VennDiagram R package. For comparisons shown in Figure 2-figure supplement 1A, oocyte transcriptome data was extracted from Stoeckius et al. (2014), and embryonic soma and germ cells expression profiles were from this study (Supplementary file 7). Expression of each gene was log10 transformed, ranked and ordered. Correlations were plotted using custom R codes and can be found in Figure 2figure supplement 1A source code.

Principal component analysis
Principal component Analysis (PCA) was used to evaluate reproducibility of RNA-seq experiments. PCA revealed clustering of biological replicates with the same library preparation procedure as shown in Figure 2-figure supplement 3. In Figure 2-figure supplement 3A, two different sets of libraries (one set was made with NuGEN protocol and the other was made with SMART-seq protocol) were generated using the same RNA and clustered differently, suggesting different library making procedures could introduce biases. Sequence reads were mapped to transcriptome version ce10 using Hisat2. HTseq-count was used to generate raw counts for each gene. The command lines are listed as below.

Gene categories
We defined four gene categories based on expression characteristics reported in published microarray, serial analysis of gene expression (SAGE), and RNAseq data sets that profiled specific tissues or whole worms with or without a germline (Gaydos et al., 2012;Meissner et al., 2009;Ortiz et al., 2014;Reinke et al., 2004;Wang et al., 2009). The oocyte category (1594 genes) and sperm category (2042 genes) are genes with differential enrichment in dissected female gonads from adult fog2(q71) animal compared to dissected male gonads from adult fem-3(q96) animals (Ortiz et al., 2014). The soma category (2684 genes) was obtained by taking genes with SAGE tags in at least one somatic tissue (intestine, muscle, or nerve) as described in Gaydos et al., 2012, and substracting from that list all genes in the oocyte and sperm categories described above. The pregamete category (1694 genes) was constructed by adding the germline-enriched and the germline-specific gene sets from Gaydos et al., 2012 and substracting from that list all genes in the oocyte and sperm categories described above. Germline-enriched genes include genes whose expression is significantly higher in germline based on comparison of adults with and without a germline (Reinke et al., 2004). Germline-specific genes are those with SAGE tags in dissected germlines and not in somatic tissues (intestine, muscle and nerve cells). For gene sets enrichment test, we used total number of 15851 expressed genes with RPKM >0.1 as the cutoff from our PGC RNA-seq experiments to calculate 'expected' values for each category.
Expected value = (No. of significantly changed genes) x (No. of genes in category/15851). Hypergeomatric test was performed to derived p-values (hypergeometric probability), and listed in figure legends.

ATAC-seq library preparation and analysis
ATAC-seq was performed as described in Buenrostro et al., 2015. Experimental pipeline was described as follows: Scheme 1. Experimental procedure for ATAC-seq analysis. DOI: https://doi.org/10.7554/eLife.30201.024 30,000 sorted L1 PGCs were washed with 60 ml cold cell culture grade PBS once and spun at 2000xg for 10 min. Cell nuclei were isolated by resuspending cell pellets in cold lysis buffer (10 mM Tris-Cl pH7.4, 10 mM NaCl, 3 mM MgCl 2 , 0.1% Igepal CA-630) followed by centrifugation at 3500xg for 10 min at 4˚C. The transposition reaction was performed with a 50 ml reaction mixture (25 ml TD, 2.5 ml TDE, 22.5 ml nuclease-free H 2 O. Illumina, Nextera DNA library preparation Kit FC-121-1030) at 37˚C for 30 min. Transposed DNA was purified using Qiagen MinElute kit and saved in À20˚C. qPCR was used to determine appropriate PCR cycle number for PCR amplification as detailed in  cycles of PCR amplification were used. Final cDNA libraries (150 bp to 700 bp) were selected using Agencourt AMPure beads (Beckman-Coulter A63880). Two biological samples for wild type and nos-1(gv5)nos-2(RNAi) were sequenced with Hiseq2500 platform.

Quantitative RT-PCR assay
To verify our analysis pipeline for RNAseq data, quantitative RT-PCR (qRT-PCR) reactions using sequencing libraries as templates were performed. The cDNA libraries were diluted to 1 nM before performing qRT-PCR. Primers for qRT-PCR were listed in key resources table. Enrichment of target mRNAs between wild-type and nos-1/2 was calculated using DDCt with tbb-2 expression then normalized to wild-type control. Fold changes were plotted and significance was calculated by paired t-test.

Technical v biological replicates
Biological replicates refer to experiments performed on independently treated hermaphrodites (in the case of RNA-seq libraries, this refers to worms exposed to independent RNAi treatments  Table  S1