Developmental stage-dependent transcriptional regulatory pathways control neuroblast lineage progression

Neuroblast lineage progression is a fundamental biological process in which neural stem cells asymmetrically divide, exit the cell cycle and differentiate into distinct neurons. Defects in neuroblast lineage progression cause abnormal neural circuits and neurological disorders (Ming and Song, 2011; Zhao et al., 2008). Despite recent progress in understanding the transcriptional control of neurogenesis (Hsieh, 2012), the full inventory of transcription factors (TFs) and regulatory networks underlying neuroblast development remains largely unknown.

The C. elegans Q neuroblasts divide three times to generate three distinct neurons (oxygen sensory, mechanosensory and interneuron) and two apoptotic cells in L1 larvae (Fig. 1A) (Sulston and Horvitz, 1977). The anterior and posterior daughters of Q neuroblasts, Q.a and Q.p, employ distinct cellular mechanisms in their asymmetric divisions (Ou et al., 2010). A polarization of myosin-based contractility is the primary driver for the production of two distinct daughter cells in Q.a division (Ou et al., 2010), and the myosin polarization also occurs in Drosophila neuroblast asymmetric division (Cabernard et al., 2010). In Q.p division, asymmetric daughter cell sizes and fates arise from the initial displacement of the mitotic spindle towards one side of the cell, which is similar to what occurs in the first division of the C. elegans embryo (Gönczy, 2008; Ou et al., 2010). Pioneering C. elegans genetics uncovered that a POU domain TF, UNC-86, acts to modify latent reiterative cell lineages of the Q neuroblast; Q.a divides normally but Q.p repeats the division pattern of the mother neuroblast in unc-86 mutants (Chalfie et al., 1981). Additional transcriptional regulatory pathways that determine the distinct Q cell division patterns (e.g. Q.a division) have yet to be identified.

Neuroblast division needs to be precisely regulated. In Q neuroblasts, Q.a only divides once whereas Q.p divides twice (Sulston and Horvitz, 1977). The Hippo signaling pathway consists of a core kinase cascade that controls organ size and the development of human cancers. Upon activation in Drosophila, Hippo kinase phosphorylates and activates Warts kinase, which inactivates Yorkie by phosphorylation; Yorkie is a transcriptional coactivator of a TEA domain (TEAD) family TF, Scalloped, which induces gene transcription to promote cell proliferation and inhibit apoptosis (Pan, 2010; Zhao et al., 2011). In C. elegans, CST-1/2, WTS-1, YAP-1 and EGL-44 are homologs to Hippo, Warts, Yorkie and Scalloped, respectively, and they play diverse roles in life span, development and neuronal fate determination (Cai et al., 2009; Iwasa et al., 2013; Kang et al., 2009; Wu et al., 2001). However, none of them has been shown to regulate C. elegans cell cycle progression. Prior genetic studies reported that EGL-44 and EGL-46 (a homolog of human insulinoma-associated protein) control the fate of the touch cell FLP and that EGL-46 is essential for cell cycle exit in Q.ap and Q.paa cells (Wu et al., 2001), but the function of EGL-44/Scalloped in C. elegans neuroblast proliferation remains unclear.

Neuronal differentiation involves numerous cell type-specific transcriptional regulatory cascades that determine neuronal fate (Hobert, 2011). A LIM domain TF, MEC-3, positively regulates the expression of mechanosensory genes in one of three Q cell progenies (Way and Chalfie, 1988). Equally intriguing is the inhibition of the touch fate in non-mechanosensory neurons. The C. elegans C2H2-type zinc-finger TF PAG-3 is the homolog of the Drosophila Senseless proteins and represses touch neuron-specific genes in BDU interneurons (Cameron et al., 2002; Jia et al., 1996; Jia et al., 1997). However, little is known about the repression of the touch fate in Q cell progenies that do not function as the mechanosensory neuron.

This study performed large-scale forward genetic screens, through which we identified three evolutionarily conserved TFs essential for Q neuroblast lineage progression. Using live cell imaging, genetic and biochemical approaches, we showed that they form three transcriptional regulatory pathways that function sequentially to ensure proper asymmetric cell division, cell cycle exit and cell fate determination in Q neuroblast development.

C. elegans strains were raised on NGM plates seeded with the Escherichia coli strain OP50 at 20°C. Strains are listed in supplementary material Table S1. PCR products and plasmid constructs are listed in supplementary material Tables S2 and S3. Integrated Pmec-4::gfp(zdIs5), Pgcy-32::mCherry(casIs35) and Pgcy-32::gfp(casIs36) transgenes were used to visualize Q neuroblast progenies, AVM/PVM and AQR/PQR. EMS mutagenesis was carried out with a single or with two transgenic markers to identify mutations with an ectopic gain or loss of these neurons. We isolated mutant alleles of known genes in Q neuroblast lineage progression; cnt-2(cas4), pig-1(cas5) and unc-86(cas34, cas60, cas61, cas63, cas144, cas148, cas152). Unknown mutations were mapped using snip-SNP techniques and complementation testing (for details of mapping see supplementary material Fig. S1A) and candidate genes in these regions were sequenced to identify mutations.

cas27, cas46 and cas137 mutations caused an extra AQR-like or PQR-like neuron phenotype, and they were determined to affect a single gene by complementary tests. Snip-SNP mapping located cas46 mutations between IV: 5.19 and IV: 7.41. Sequencing of the ham-1 gene in this region revealed nucleotide changes in cas27 and cas137 and a deletion in cas46 (Fig. 1C,D; supplementary material Fig. S1A).

cas3, cas6, cas19, cas58 and cas140 mutations caused one extra AQR-like or PQR-like neuron phenotype. These mutants showed a weak egg-laying defect phenotype and they ectopically expressed Pmec-4::gfp in one or two FLP cells. Snip-SNP mapping located this gene to II: -0.96 and II: 0.12. We sequenced three egl genes in this region and found mutations in egl-44 (Fig. 1C,D; supplementary material Fig. S1A).

cas16, cas18, cas24, cas25, cas36 and cas133 caused one extra AQR-like or PQR-like neuron phenotype. The mutations were localized to V: 0.11 and V: 0.55 where egl-46 is located. Sequencing revealed a mutation in the egl-46 coding region of each allele (Fig. 1C,D; supplementary material Fig. S1A).

cas8, cas10, cas11, cas12 and cas22 mutations caused one extra AVM-like or PVM-like neuron and the Egl phenotype with 100% penetrance, and they were determined to affect a single gene by complementary tests. Snip-SNP mapping located this gene between X: -4.93 and X: -4.38. Microinjection of a single fosmid, WRM068bH06, that carries the DNA sequence spanning part of this region partially rescued both the Egl and extra neuron phenotypes. Sequencing results revealed a Q- or W-to-stop codon mutation in egl-13 of each allele (Fig. 1C,D; supplementary material Fig. S1A).

Four types of mutation occurred in these alleles: (1) nonsense mutations in egl-46 (cas16, cas18, cas24, cas25, cas26) and egl-13 (cas8, cas10, cas11, csa12, cas22); (2) missense mutations in ham-1 (cas27, cas137), egl-44 (cas3, cas6, cas58, cas140) and egl-46 (cas133); (3) a deletion in ham-1(cas46) removed the last 381-414 amino acid and 3′UTR; and (4) a substitution of the consensus 5′ donor splice site of intron 2 in egl-44(cas19).

Our screens and those of others repeatedly identified mutations that changed the same nucleotides in egl-44 and egl-46. Arginine 140 in EGL-44 was mutated to glutamine in cas6, n998, n1087, while cas16 and cas36 or cas18 and cas24 had identical mutations in egl-46 (glutamine 101 or 168 to a stop codon), although they were isolated from different strains (cas16 and cas18 from Pgcy-32::mCherry, cas24 and cas36 from Pmec-4::gfp and Pgcy-32::mCherry double markers).

We did not note any obvious difference in penetrance in the QL and QR lineages of ham-1, egl-44 or egl-46 mutants. In egl-13 mutants, the QL and QR lineages had an identical phenotype but with different penetrance, and their data were distinguished (Fig. 6E,F).

We confirmed that these mutants were correctly cloned. All of the canonical alleles of these genes have the same phenotypes as our newly isolated mutant alleles (supplementary material Fig. S1B). GFP- or mCherry-tagged translational fusion constructs of these genes successfully rescued the corresponding mutant phenotypes (supplementary material Fig. S1C).

C. elegans L1 larvae were anesthetized with 0.1 mM levamisole (Sigma) in M9 buffer, mounted on 3% agarose pads, and maintained at 20°C. Our imaging system includes an Axio Observer Z1 microscope (Carl Zeiss MicroImaging) equipped with a 100×, 1.45 N.A. objective, an EM CCD camera (Andor iXon+ DU-897D-C00-#BV-500), and the 488-nm and 568-nm lines of a Sapphire CW CDRH USB laser system or an argon and krypton laser attached to a spinning disk confocal scan head (Yokogawa CSU-X1 Spinning Disk Unit). Time-lapse images were acquired with exposure time of 300 mseconds every 30 or 60 seconds by Focus Image software (developed by Mr Xiang Zhang at the Institute of Biophysics, Chinese Academy of Sciences). We used ImageJ software (NIH) to process images. In all images shown, the anterior of Q cells or C. elegans is to the left.

Quantification of spindle position and daughter cell size ratio followed our published protocol (Ou et al., 2010). Spindle positioning was quantified by the ratio of the distance of the anterior and posterior centrosomes to the anterior and posterior cell poles (Pa, Pp). Daughter cell size ratio was calculated by Q.aa/Q.ap or Q.pp/Q.pa. Quantification of GFP fluorescence intensity of PIG-1::GFP in wild type (WT) and mutants followed our published methods (Ou and Vale, 2009). The fluorescence intensity of PIG-1 was calculated as fluorescence divided by area using ImageJ. Student’s t-test or χ² analysis was used to examine statistical differences in daughter Q cell size ratio or extra neuron phenotypes in WT and mutants.

Integrated Genome Browser 6.7.3 was used to analyze HAM-1 binding peaks at the pig-1 promoter. The CHIP-Seq data were downloaded from the modENCODE consortium (www.modencode.org).

Sequences encoding EGL-44 and EGL-46 full-length proteins or truncations were cloned into a DNA activation domain (AD) vector (pGADT7, Clontech) and a binding domain (BD) vector (pGBKT7, Clontech) by the in-fusion cloning technique. Plasmids were verified by DNA sequencing. The resulting AD and BD constructs were transformed into the yeast strain AH109 using the Yeastmaker Yeast Transformation System 2 (Clontech). Transformants carrying both AD and BD vectors were initially selected on synthetic complete agar -Leu -Trp medium. To screen protein-protein interaction, a single clone was streaked on the interaction selection media (synthetic complete agar -Leu -Trp -Ade -His) to score growth.

E. coli BL21 (DE3) was used to express GST, GST-EGL-44 or His-EGL-46 with induction by 0.3 mM IPTG for 12 hours at 16°C. Bacterial culture was collected and disrupted in lysis buffer (PBS pH 7.32, with 140 mM NaCl, 2.7 mM KCl, 10 mM Na₂HPO₄ and 1.8 mM KH₂PO₄) using a microfluidizer. Bacterial lysate containing His-EGL-46 was incubated with GST lysate or GST-EGL-44 for 30 minutes at 4°C, and glutathione-agarose beads 4B (GE Healthcare) were added to immobilize GST or GST-EGL-44 for another hour at 4°C. The beads were washed three times with PBS. The bound proteins were eluted with elution buffer (10 mM reduced glutathione in PBS) and the eluents were subjected to western blotting with anti-His and anti-GST antibodies (Zhonghan Golden Bridge).

To identify factors that control Q neuroblast lineage progression, we carried out GFP- and mCherry-based genetic screens. We used a mechanosensory neuron-specific reporter construct (Pmec-4::gfp) and an oxygen sensory neuron-specific reporter construct (Pgcy-32::mCherry) to isolate mutations that alter the numbers of both neuronal types (Fig. 1B). The mec-4 promoter drives GFP expression in six mechanosensory neurons, two of which, AVM and PVM (A/PVM), are derived from Q.p asymmetric divisions. The gcy-32 promoter drives mCherry expression in four oxygen sensory neurons, two of which, AQR and PQR (A/PQR), are produced by the Q.a lineage. Genetic screens using Pmec-4::gfp have been carried out previously (Cordes et al., 2006; Singhvi et al., 2011); however, screens based on Pgcy-32::mCherry that are likely to uncover specific factors for Q.a asymmetric division have not been reported. We generated a double-fluorescence labeling strain in which four Q neuroblast progenies can be visualized, and we screened for mutants with an ectopic gain or loss of these neurons (see Materials and methods) (Fig. 1B). After examining 66,700 haploid genomes treated by ethyl methane sulfonate (EMS), we isolated over 50 mutants with abnormal neuron numbers, including mutant alleles of four genes, cnt-2, egl-46, pig-1 and unc-86, known to regulate Q cell development (Chalfie et al., 1981; Cordes et al., 2006; Singhvi et al., 2011; Wu et al., 2001).

We identified 13 mutants that did not map to the known loci. Using genetic techniques, we showed that these mutants caused molecular lesions in three evolutionarily conserved TFs (Fig. 1C,D; supplementary material Fig. S1, Fig. S2A; see Materials and methods; BLAST E-values range from 6.5e^-21 to 2.7e^-71): HAM-1 is a homolog of human storkhead box 1 (STOX1) (Frank et al., 2005); EGL-44 is a homolog of human transcriptional enhancer factor TEF5 (TEAD3) (Wu et al., 2001); and EGL-13 is a homolog of the human TF SOX5 (Hanna-Rose and Han, 1999). The functions of these TFs in neuroblast lineage progression are not well understood. Our study addresses the interaction between EGL-44 and EGL-46, and we isolated six alleles of egl-46 (Fig. 1C,D).

We next used live cell imaging analysis to determine which cellular processes were defective in these mutants. Among existing mutants, we focused on ham-1(cas46), egl-44(cas6), egl-46(cas36) and egl-13(cas11) because these alleles include a deletion (cas46), the most severe extra neuron phenotype (cas6), or the earliest stop codon (cas36 and cas11).

HAM-1 was previously shown to control C. elegans embryonic but not larval neuroblast asymmetric cell divisions (Frank et al., 2005; Guenther and Garriga, 1996). Although human STOX1/HAM-1 is abundantly expressed in the brain (van Dijk et al., 2010), its neuronal function is largely unknown. We confirmed that Q.p development is independent of ham-1 as no extra A/PVM were generated in ham-1 mutants (Fig. 1C). However, ham-1 mutants generated extra A/PQR neurons (Fig. 1C), indicating that HAM-1 specifically regulates Q.a development.

We studied Q cell asymmetric divisions in ham-1 and other TF mutants by visualizing the dynamics of the GFP-tagged centrosome, mCherry-labeled chromosome and plasma membrane. We found that only Q.a asymmetric division was defective in ham-1 mutants (Fig. 2A,E). In WT animals, Q.a positioned its spindle in the cell center, and an anterior accumulation of myosin during cytokinesis was the likely cause of the generation of a small Q.aa and a large Q.ap (Fig. 2A,C) (Ou et al., 2010). In ham-1 mutants, 62% of QR.a (n=13) properly positioned their spindles, but myosin was evenly distributed in the contractile ring, producing two equal daughter cells; for the remaining 38% of QR.a, the spindle was shifted towards the posterior and myosin symmetrically distributed, producing a large QR.aa and a small QR.ap (Fig. 2A-D; supplementary material Fig. S2B,C).

HAM-1 contains a predicted winged helix DNA-binding motif (Fig. 1D), suggesting that it might function as a TF inside the nucleus. However, prior immunofluorescence studies reported that HAM-1 is asymmetrically localized on the cortex in neuroblasts (Frank et al., 2005; Guenther and Garriga, 1996). We examined the expression and localization of HAM-1 in Q cells using a GFP reporter. The modENCODE consortium constructed an integrant line expressing a GFP-tagged HAM-1 protein under the control of the ham-1 promoter (Pham-1). The Pham-1::ham-1::gfp transgene was functional as it reduced the extra A/PQR phenotype from 31% (n=196) to 9% (n=203) in ham-1 mutants (supplementary material Fig. S1C). By quantifying daughter cell size asymmetry, we found that the transgene partially rescued the asymmetric cell division defects in ham-1 mutants (P<0.01, Student’s t-test; the distribution of the raw data in shown in supplementary material Fig. S1D), which correlated with the partial rescue of the extra neuron phenotype. Using this reporter, HAM-1::GFP fluorescence was visible in both Q.a and Q.p as well as in their neighboring cells (supplementary material Movie 1). Time-lapse imaging analysis showed that HAM-1::GFP was restricted to interphase nuclei as with other TFs (e.g. EGL-44, EGL-46 and EGL-13, supplementary material Fig. S3B) and that HAM-1::GFP was evenly distributed in the cytoplasm of dividing Q.a (n=14) and Q.p (n=10) cells (Fig. 3A; supplementary material Movie 1). The dynamic distribution of HAM-1 suggested that it might function as a TF in Q cells.

To understand how HAM-1 regulates Q.a asymmetric division, we searched for its transcriptional regulatory target. The modENCODE consortium used chromatin immunoprecipitation coupled with high-throughput DNA sequencing (CHIP-seq) to search for the genome-wide binding sites of HAM-1 in C. elegans L1 larvae. From the original datasets, we uncovered that HAM-1 is bound to the promoter region of the pig-1 gene (-266 to -42 bp relative to the ATG; Fig. 3B). PIG-1 is a member of the conserved PAR-1/Kin1/SAD-1 family of serine/threonine kinases that regulate polarity and asymmetric cell division, and the inhibition of pig-1 phenocopied the ham-1 extra neuron phenotype in the Q.a cell (Cordes et al., 2006). Our previous imaging analysis showed that myosin polarization during Q.a cytokinesis was disrupted in pig-1 mutants (Ou et al., 2010). Furthermore, the prior genetic analysis suggested that pig-1 acted downstream of ham-1 (Cordes et al., 2006). Thus, pig-1 is a likely functional target of HAM-1.

We first examined whether HAM-1 promoted pig-1 expression in Q cells. We constructed a strain expressing Ppig-1::pig-1::gfp and showed that the transgene was functional because it reduced extra A/PQR neurons in pig-1(gm344) mutants from 19% (n=154) to 2% (n=174). After crossing the transgene into ham-1 mutants, we compared the expression of Ppig-1::pig-1::gfp in WT and ham-1 mutants. The PIG-1::GFP fluorescence was visible in all the WT Q.a cells (n=18), but was significantly reduced in ham-1 mutants (Fig. 4A,B). PIG-1::GFP fluorescence did not change in neighboring seam cells of ham-1 mutants (Fig. 4A,B, asterisks), suggesting that we could use PIG-1::GFP in seam cells as a fiduciary marker to further quantify PIG-1::GFP changes in Q cells. We measured the fluorescence intensity ratio between Q.a and seam cells within the same animal. PIG-1::GFP in Q.a was 3.3-fold brighter than that of the seam cell (n=18) in WT animals, but this intensity ratio was reduced to 1.4 in ham-1 mutants (n=24; P<0.005, Student’s t-test) (Fig. 4C). Thus, HAM-1 positively regulates pig-1 expression in Q.a.

To study the significance of the HAM-1 binding site in the pig-1 promoter, we deleted it from a functional Ppig-1::pig-1::gfp plasmid. This completely abolished pig-1::gfp expression in the Q.a cell (100%, four independent transgenic lines, 10-15 worms were examined from each line; Fig. 4D, right). As a negative control, the deletion of another region in the pig-1 promoter (-633 to -407 from the ATG) did not change PIG-1::GFP fluorescence in three independent transgenic lines (Fig. 4D, middle). Thus, the HAM-1 binding site in the pig-1 promoter is essential for pig-1 expression.

To validate the role of HAM-1 in the transcriptional regulation of pig-1 further, we examined whether the expression of pig-1 by HAM-1-independent promoters, such as Pegl-17, Pgcy-32 or Pmec-7, could rescue the defects in Q.a asymmetric division in ham-1 mutants. We introduced a transgene that expresses pig-1::gfp under the control of Pegl-17, a promoter active throughout Q cell development, into ham-1 mutants. By quantifying A/PQR neuron numbers, we found that the Pegl-17::pig-1::gfp transgene partially reduced the occurrence of extra A/PQR from 31% to 17% in ham-1 mutants (P<0.01 by χ² analysis; Fig. 4E), suggesting that bypassing HAM-1 regulation of the pig-1 promoter partially rescued asymmetric division defects in ham-1 mutants. The rescue depended on the developmental stage. We expressed pig-1::gfp under the Pgcy-32 or Pmec-7 promoter, neither of which is active until Q cell differentiation, and did not observe any rescue of the extra A/PQR phenotype (Fig. 4E).

Taken together, HAM-1 positively regulates pig-1 expression during Q.a asymmetric division, possibly through the pig-1 promoter region.

Prior studies reported that EGL-44 and EGL-46 repress touch cell fate in FLP cells and that EGL-46 regulates cell cycle exit in Q cell lineages (Desai and Horvitz, 1989; Wu et al., 2001). However, whether EGL-44 is involved in cell cycle regulation has not been determined. We uncovered multiple alleles of egl-44 and egl-46 with the ectopic gain of A/PQR neurons (Fig. 1C,D). Time-lapse recording showed that EGL-44 controls cell cycle exit in Q.ap and Q.paa cells (Fig. 5A; supplementary material Movies 2, 3). Using QR.ap as an example, QR.ap differentiated into AQR in WT animals (Fig. 5A); however, QR.ap had one extra round of division at 158±19 minutes after birth in egl-44 mutants (n=9) or 158±34 minutes after birth in egl-46 animals (n=6) (Fig. 5A). Both QR.ap daughter cells differentiated into neurons expressing oxygen sensory neuron-specific genes, resulting in extra A/PQR in egl-44 or egl-46 mutants (Fig. 1C, Fig. 5A; supplementary material Movies 2, 3).

We next examined the expression pattern of egl-44 and egl-46. egl-44 was expressed throughout the Q cell lineage, whereas egl-46 was only expressed in a subset of this lineage. Initially, egl-46 was not expressed in the Q neuroblast, and in the Q.a lineage it started to be expressed in the Q.a cell, whereas in the Q.p lineage its expression was restricted to Q.paa and was absent from Q.pap (supplementary material Fig. S3A) (Wu et al., 2001). Cell cycle exit in egl-46 mutants is only defective in Q.ap or Q.paa cells, indicating that EGL-46 is specifically expressed in cells where it is essential.

Since egl-44 and egl-46 mutants have the same phenotype in terms of cell cycle regulation and FLP cell differentiation, we examined whether they physically bind to each other. Yeast two-hybrid assays detected the interaction between full-length EGL-44 and EGL-46 (Fig. 5B,C). We narrowed down the minimal regions for their association; the EGL-44 N-terminal fragment, including the TEAD domain, was sufficient to bind to the EGL-46 C-terminal fragment, including the zinc-finger domain (Fig. 5B,C). We used protein pull-down assays with glutathione S-transferase (GST) fusion proteins to show that His-tagged EGL-46 bound to GST-tagged EGL-44 but not to the GST protein alone (Fig. 5D). Thus, EGL-44 and EGL-46 might function in a transcriptional regulatory protein complex to promote cell cycle exit.

EGL-13 belongs to an SRY box (Sox)-containing gene family (Fig. 1D). EGL-13 was initially identified from defects in the connection between the uterus and the vulva in C. elegans (Cinar et al., 2003; Hanna-Rose and Han, 1999). SOX5 and SOX6, which are vertebrate homologs of EGL-13, are involved in chondrogenesis and in the cell cycle progression of neural progenitors in the chick spinal cord; however, little is known about their function in neural fate determination. We found that egl-13 mutants generated extra A/PVM (Fig. 1C) but lost neurons in the Q.a lineage (AQR or PQR was absent in 22% or 42%, respectively, of egl-13 animals; n=77).

We examined the neuronal differentiation of Q.ap cells in egl-13 mutants. For example, QL.ap normally differentiates to PQR by expressing oxygen sensory neuron-specific genes such as Pgcy-32::mCherry. In egl-13 mutants, QL.ap did not express Pgcy-32::mCherry but ectopically expressed Pmec-4::gfp, a marker characteristic of mechanosensory neurons (Fig. 6A,B), indicating that QL.ap was likely to be programmed from an oxygen sensory neuron to a touch neuron. Consistently, QL.ap expressed both Pgcy-32::mCherry and Pmec-4::gfp in 53.3% of egl-13 mutants (n=45), which might be intermediates of both neuronal fates. We examined additional neuronal markers to confirm the neuronal fate transition. In A/PVM, mec-3, mec-4, mec-7 and mec-18 encode touch neuron-specific TF, channel protein, tubulin and CoA synthetase, respectively. No PQR expressed these genes in WT animals, whereas 75-88% of PQR neurons ectopically expressed them in egl-13 mutants (Fig. 6C; supplementary material Fig. S4A). In A/PQR, gcy-32, gcy-36, tax-4 and glb-5 encode membrane receptors or signaling molecules specifically for oxygen sensation. A/PQR neurons are ciliated, and the ciliogenesis depends on the action of intraflagellar transport proteins (e.g. OSM-6). We compared the expression of A/PQR-specific genes in WT and egl-13 animals. All the WT PQR neurons expressed these markers, but PQR neurons expressed them in 6-73% of egl-13 mutants (Fig. 6C; supplementary material Fig. S4B). EGL-13 thus determines the neuronal fate in Q cell lineages.

MEC-3 is a key TF for mec gene expression, and the ectopic expression of mec genes in Q.ap of egl-13 mutants requires MEC-3 because Q cell progenies do not express mec genes in egl-13; mec-3 double mutants (supplementary material Fig. S4C, right). We also found that egl-13 single mutants and egl-13; mec-3 double mutants did not differ in the loss of A/PQR (supplementary material Fig. S4C), indicating that the loss of A/PQR in egl-13 worms does not require mec-3. EGL-13 might repress the touch fate by inhibiting mec-3 and support A/PQR fate independently of mec-3.

We studied the expression pattern of egl-13 in Q cell lineages. Using a transcriptional fusion reporter with gfp, we found that egl-13 started to be expressed after Q.a divisions and that GFP fluorescence reached the maximum level in Q.ap upon differentiation (supplementary material Fig. S4D, Movie 4). By contrast, the GFP signal was barely detected in the Q.p lineage, and the mCherry fluorescence in Q cells did not change during Q cell development (supplementary material Fig. S4D, Movie 4). egl-13 was continuously expressed in A/PQR during larval development (supplementary material Fig. S3C, e.g. L3 larval stage), indicating that EGL-13 can be involved in neuronal fate initiation and maintenance.

To address whether EGL-13 is sufficient for neuronal fate determination, we ectopically expressed it in A/PVM, where EGL-13 is normally absent. Two promoters were used for EGL-13::GFP expression; Pegl-17 was active throughout the entire Q cell lineage, whereas Pmec-7 was only switched on in A/PVM upon differentiation (Fig. 6D). Both transgenes could induce ectopic A/PQR neurons (Fig. 6E). For instance, extra AQR neurons were produced in 40% of transgenic animals expressing Pmec-7::egl-13::gfp. Since the Pmec-7 promoter was stronger than Pegl-17 during differentiation, Pmec-7::egl-13::gfp induced a higher percentage of extra A/PQR neurons than Pegl-17::egl-13::gfp (Fig. 6E). In addition, both transgenes inhibited the touch fate in A/PVM. In WT animals, 100% of A/PVM expressed the Pmec-4::gfp reporter. The Pmec-7::egl-13::gfp transgene significantly reduced the number of GFP-positive A/PVM, and 71% of AVM and 94% of PVM lost their GFP fluorescence from Pmec-4::gfp (Fig. 6F). The Pegl-17::egl-13::gfp transgene also inhibited the PVM fate (Fig. 6F).

Collectively, these data demonstrated that egl-13 is necessary and sufficient to determine the neuronal fate in the Q cell lineage.

We performed double-mutant analysis to investigate interactions between the TFs. Single-mutant analysis showed that ham-1 controlled Q.a asymmetric division, whereas egl-44/egl-46 regulated Q.ap cell cycle exit (supplementary material Fig. S5A). We found that 7.8% of ham-1; egl-44 (n=51) and 20.5% of ham-1; egl-46 (n=34) double mutants generated four PQR-like neurons (supplementary material Fig. S5B), indicating that the disruption of asymmetric cell division in ham-1 and of cell cycle exit in egl-44/46 could be additive. EGL-13 determined the neuronal property of Q.ap (supplementary material Fig. S5A). We found that 12% of ham-1; egl-13 double mutants (n=80) produced three PVM-like neurons, two of which could arise from additive defects in Q.a asymmetric division of ham-1 and of daughter cell differentiation in egl-13 (supplementary material Fig. S5B). Similarly, 4.2% of egl-44; egl-13 double mutants (n=120) generated three PVM-like neurons, probably resulting from the failures of QL.ap cell cycle exit in egl-44 and of QL.ap daughter cell differentiation in egl-13 (supplementary material Fig. S5B). The double-mutant analysis suggested that three transcriptional regulatory pathways additively control Q neuroblast development.

To explore the conserved function of the TFs, we expressed human TEF5/EGL-44 and SOX5/EGL-13 in C. elegans egl-44 and egl-13 mutants, respectively. The extra A/PQR phenotype was reduced by TEF5 from 92% to 30% in egl-44 worms (Fig. 5E) and the extra A/PVM phenotype was also reduced by SOX5 in egl-13 mutants (Fig. 6F). The ectopic expression of SOX5 in mechanosensory neurons inhibited the touch fate of A/PVM (Fig. 6F). Furthermore, the yeast two-hybrid assay detected the direct interaction of human TEF5 and C. elegans EGL-46 (Fig. 5F). These results suggest that human TEF5 and SOX5 could be functional homologs of C. elegans EGL-44 and EGL-13.

This study identified three transcriptional regulatory pathways that sequentially control asymmetric cell divisions, cell cycle exit and differentiation in C. elegans Q neuroblast development (supplementary material Fig. S5A). We suggest that homologs of these TFs might have conserved functions in vertebrate neural development.

Our work first demonstrated that HAM-1 specifically controls Q.a but not Q.p division, providing molecular insight into their distinct cell division patterns. We propose that HAM-1 and PIG-1 regulate myosin polarization and the asymmetric segregation of cell fate determinants during the Q.a cell division. In ham-1 or pig-1 mutants, the Q.a cell divided symmetrically and the equal segregation of neural fate determinants transformed the normally apoptotic daughter cell into its sister cell. Alternatively, HAM-1 and PIG-1 might independently control asymmetric cell divisions and the asymmetric segregation of fate determinants. Consistent with this model, Q.a daughter cell size asymmetry was 100% disrupted in ham-1 or pig-1 mutants, but extra A/PQR neurons were generated in only 31% of ham-1 or 19% of pig-1 mutants, indicating that daughter cell sizes do not fully specify neuronal fates. To generate extra neurons, the HAM-1/PIG-1 pathway must cooperate with others pathways (e.g. an inhibition of apoptosis) to allow daughter cell survival and differentiation.

pig-1 might act differently in Q.a and Q.p development. Although more ectopic neurons are generated from the Q.p lineage (42% extra A/PVM) than those from the Q.a lineage (19% extra A/PQR) in pig-1 mutants (Cordes et al., 2006), Q.a and Q.p asymmetric cell divisions are 100% converted to symmetric division in pig-1 mutants (Ou et al., 2010). These observations are consistent with the notion that PIG-1 may play independent roles in asymmetric cell division and in the asymmetric segregation of fate determinants (Cordes et al., 2006). For neuronal fate determination, PIG-1 might be more essential in the Q.p lineage than in the Q.a lineage, as pig-1 mutants produce more ectopic A/PVM than ectopic A/PQR. For asymmetric cell division, PIG-1 might be more crucial for Q.a asymmetric division than Q.p division, as the PIG-1 protein level is reduced in both Q.a and Q.p of ham-1 mutants but only Q.a asymmetric division is altered.

HAM-1 might have different functions in embryonic and postembryonic neuroblast lineages. Immunofluorescence studies reported that HAM-1 is asymmetrically localized on the cortex in embryonic neuroblasts and that it is not expressed in larvae (Frank et al., 2005; Guenther and Garriga, 1996). The absence of HAM-1 signal by immunostaining in larvae might be due to inefficient antibody penetration owing to the thick cuticle of C. elegans. Using a functional Pham-1::ham-1::gfp strain, we did not detect any asymmetric localization of HAM-1 throughout the Q.a and Q.p cell cycle (Fig. 3A; supplementary material Movie 1). Furthermore, our CHIP-seq data analysis and functional studies demonstrated that HAM-1 could function as a TF to promote pig-1 expression in the Q.a cell.

HAM-1 might transcribe other genes for Q.a divisions. In pig-1 mutants, all the Q.a division was converted into symmetric division. In ham-1 mutants, 68% of Q.a divided symmetrically, while the other 32% made a large Q.aa and a small Q.ap by completely reversing the polarity (Fig. 2A,D), which might be caused by the loss of other HAM-1 targets than PIG-1. Consistently, the introduction of pig-1 under the control of the egl-17 promoter into Q cells only partially reduced the extra A/PQR of ham-1 mutants (Fig. 4E). It might be that multiple TFs regulate pig-1 expression, as although deletion of the HAM-1 binding site in the pig-1 promoter completely abolished pig-1 expression throughout the entire animal, pig-1 expression was only reduced in the Q cells of ham-1 mutants (Fig. 4A-D). pig-1 expression was reduced in Q.p of ham-1 mutants but Q.p divisions were normal (Fig. 4A, Fig. 2B,D), suggesting that Q.p division might require less PIG-1 than Q.a division.

This study links C. elegans Hippo components to cell cycle regulation. EGL-44, human TEF5 and Drosophila Scalloped are homologous TFs, but our work showed that EGL-44 promotes cell cycle exit (Fig. 5A), which is opposite to the function of TEF5 and Scalloped in promoting cell proliferation. Thus, Hippo components might have modulated functions across species. The physical association of EGL-44 and EGL-46 implied that the vertebrate EGL-46 homolog INSM1 might be a previously unknown component in the Hippo signaling pathway. INSM1 plays a pan-neurogenic role in promoting basal progenitor formation in the neocortex (Farkas et al., 2008), and inhibition of INSM1 reduces the radial thickness of the cortical plate, whereas its ectopic expression allows neuroepithelial cells to undergo self-amplification (Farkas et al., 2008), indicating its potential function in cell cycle regulation. The Drosophila homolog of EGL-46, Nerfin-1, is expressed in neuroblasts and regulates early axon guidance in the central nervous system (Kuzin et al., 2005; Stivers et al., 2000), but its function in neuroblast divisions has yet to be determined.

Extra A/PQR neurons were generated in egl-44/egl-46 and unc-86 mutants. In egl-44/egl-46 mutants, an extra round of Q.ap cell division gave rise to extra A/PQR neurons. In unc-86 mutants, extra A/PQR neurons were produced by the reiteration of the Q neuroblast lineage in the Q.p cell (Chalfie et al., 1981). We did not find any changes in egl-44/egl-46 expression in unc-86 mutants (supplementary material Fig. S3D). EGL-44/EGL-46 determined the touch fate of FLP neurons (Wu et al., 2001) but not in Q cell lineages (Fig. 5A; supplementary material Fig. S5A), and we found that egl-13 was expressed in two AQR neurons of egl-44 mutants (supplementary material Fig. S3E).

Our study has uncovered a novel function of EGL-13/SOX5 family TFs in neuronal fate determination. Vertebrate EGL-13/SOX5 regulates the differentiation of prechondrocytes into chrondroblasts (Smits et al., 2004). In the nervous system, SOX5 controls the timing of cell cycle exit by opposing the activity of Wnt/β-catenin in the chicken spinal cord (Martinez-Morales et al., 2010). Mammalian SOX5 regulates the pace of differentiation of corticofugal neurons by fine-tuning the identity of the various closely related subtypes (Lai et al., 2008). This work has shown that EGL-13 represses the touch fate in non-mechanosensory neurons of the Q cell lineage (Fig. 6). The fate of touch neurons can be inhibited by different transcriptional regulatory mechanisms in different lineages. In C. elegans pag-3 mutants, the mechanosensory genes of ALM touch neurons are ectopically expressed in lineally related BDU interneurons (Cameron et al., 2002; Jia et al., 1996; Jia et al., 1997). Taken together, our study demonstrates that three transcriptional regulatory pathways function at distinct development stages to ensure neuroblast lineage progression in C. elegans. The identification of additional transcriptional targets will further advance our understanding of neuroblast development.

We thank Dr G. Garriga and the Caenorhabditis Genetics Center for strains and Drs G. Garriga, W. Zhong, T. Xie and D. Xue for discussion.