Escargot and Scratch regulate neural commitment by antagonizing Notch activity in Drosophila sensory organs Free

During nervous system formation, the Notch (N) signaling pathway regulates neuronal specification at two levels: first, by regulating the number of neural precursor cells; and second, by conditioning sibling cell fates in neural cell lineages (Guruharsha et al., 2012; Jukam and Desplan, 2010; Simpson, 1997). In the latter case, N signaling is differentially activated in pairs of newly divided sibling cells and, as such, restricts the acquisition of a given cell fate to only one daughter cell. The molecules and regulatory interactions involved in triggering such N-dependent asymmetric processes have been investigated in depth, but the factors involved in maintaining N-related cell fate commitment are unknown. Here, we shed light on one mechanism that maintains the N_off state in neural precursor cells in the Drosophila mechanosensory organ (mSO).

Drosophila mSOs are formed by a neuron and three accessory cells. All these cells arise from a primary precursor cell, called pI, which undergoes a stereotyped sequence of oriented asymmetric cell divisions. The pI cell division produces the secondary precursor cells pIIb and pIIa. The pIIa cell gives rise to the outer terminal cells, the socket and the shaft cells. The pIIb cell produces the neural type cells, the glial cell (which disappears as a result of apoptosis) and the tertiary precursor cell pIIIb, which then generates the sheath cell and the neuron (Fichelson and Gho, 2003; Gho et al., 1999; Lai and Orgogozo, 2004). At each of these divisions, each daughter cell acquires a different identity due to the differential activation of the N pathway. This differential activation of the N pathway between daughter cells has been well documented during pI division: one daughter cell, pIIb (the N_off cell), acts as an N signal sender and its sibling, pIIa (the N_on cell), as an N signal receiver (Guo et al., 1996). The bias in the activation of the N pathway is mainly assured by the stereotyped segregation of determinant factors such as Numb and Neuralized (Neur) in one daughter cell (the future pIIb cell) (Guo et al., 1996; Le Borgne and Schweisguth, 2003; Morin and Bellaiche, 2011).

Snail transcription factors are involved in a wide range of cellular functions and are essential in metazoan development (Barrallo-Gimeno and Nieto, 2005; Nieto, 2002; Thiery et al., 2009). Two gene families compose the Snail superfamily: Snail and Scratch. These two families encode transcription factors that contain a divergent N-terminal region and a highly conserved C-terminal region containing four to six zinc fingers that recognize sequence-specific DNA binding motifs similar to the E-box, the core binding site of basic helix-loop-helix (bHLH) transcription factors (Nieto, 2002). In Drosophila, snail (sna), escargot (esg) and worniu (wor) compose the Snail gene family, and scratch (scrt) and two putative genes (scratch-like 1 and scratch-like 2) compose the Scratch family (see Nieto, 2002). Snail, esg, wor and scrt encode factors involved in neurogenesis. For instance, the ventral nerve cord is severely underdeveloped in Drosophila embryos mutant for Snail-related factors (Ashraf and Ip, 2001; Ashraf et al., 1999; Cai et al., 2001). Moreover, sna, esg and wor redundantly control the segregation of cell fate determinants during asymmetric neuroblast divisions (Ashraf and Ip, 2001; Cai et al., 2001). Although these data indicate that Snail/Scratch factors play central roles in neurogenesis, little is known about the molecular basis of their involvement.

A growing body of evidence supports the involvement of transcription factors from the Snail superfamily in synergy with the N pathway in cell identity specification (Cowden and Levine, 2002; Morel et al., 2003). In vertebrates, during development as well as in pathological situations, N signaling and several Snail factors interact in a complex crosstalk during the epithelial-mesenchymal transition (Wang et al., 2010). In addition, direct interaction between the intracellular domain of the N receptor and Snail has been reported in hepatocellular carcinoma cells (Kim et al., 2013; Lim et al., 2011).

The present study shows that Esg and Scrt are specifically expressed in the neural/pIIb sublineage and act redundantly to maintain the binary neural/non-neural fate decision by downregulating N pathway activity. Our results indicate that Esg and Scrt, acting redundantly, are necessary and sufficient for the acquisition of the neural precursor pIIb cell fate. Moreover, these factors repress N pathway activation; they probably bind to the promoter of N target genes and downregulate their transcription. We present evidence suggesting that Snail factors downregulate N target gene transcription by competing with activators as proneural factors. Since Esg and Scrt orthologs are present in other precursor cells, our findings on the interaction between Snail factors and the N pathway are fundamental to understanding precursor cell fate acquisition in other systems.

We studied members of the Drosophila Snail gene superfamily in a screen designed to identify factors involved in neural commitment in the mSO. After individual esg or scrt downregulation, external sensory structures were unaffected (Fig. 1A-C). By contrast, when both esg and scrt were downregulated, mSOs with two bristles were observed in 48% of the flies (n=83, Fig. 1D). Similar double-shaft structures were also observed in esg null clones induced in the heteroallelic scrt^jo11/jo16 background (Fig. 1G). The double-shaft phenotype could result from either a cell transformation or from extra divisions in the outer cell sublineage. In pupae at 24 h after puparium formation (APF), mSOs composed of no more than four cells were observed after esg and scrt downregulation (Fig. 1F,F′,H-H″), as in the control (Fig. 1E). Moreover, except for the apoptosis of glial cells, we never observed other cells in the process of fragmentation (not shown). These observations rule out the possibility that double-shaft mSOs arose from extra divisions of outer cells, and suggest instead that they resulted from a cell fate transformation. Indeed, after esg and scrt downregulation, mSOs were formed exclusively of four outer cells [Pdm1 (Nubbin) positive, Fig. 1F′], two of which were Su(H)-positive socket cells (Fig. 1H′), whereas no inner cells marked by Elav (neurons) or Prospero (Pros) (sheath cells) were observed (Fig. 1F″). Taken together, these observations show that the extra set of external structures observed after downregulation of both esg and scrt results from a cell transformation. This shows that esg and scrt are involved in neural cell fate decisions in the mSO cell lineage.

We next studied whether Esg and Scrt are sufficient to induce neural type cell fates. Overexpression of scrt (HS-scrt) at 16 h APF produced flies with severe bristle loss (Fig. 1I). As shown in Fig. 1J, this phenotype was associated with mSOs formed exclusively of neural type cells (identified by the lack of Pdm1 staining, Fig. 1J′), two of which were neurons (Elav-positive cells, Fig. 1J″). Similar bristle loss phenotypes (Fig. 1K) and mSOs composed exclusively of neural type cells (two Elav-positive cells or one Elav and one Pros cell, Fig. 1L-L″) were also observed when esg was overexpressed (using an HS-Gal4/UAS-esg fly line heat shocked at 16 h APF). These observations show that overexpression of esg or scrt is sufficient to induce a cell transformation whereby all sensory cells adopt a neural identity.

Taken together, our data indicate that esg and scrt, acting redundantly, are necessary and sufficient to maintain neural type fate decisions.

The most likely explanation for the phenotype observed after downregulation of esg and scrt is that pIIb neural precursor cells are transformed into pIIa non-neural precursor cells. To address this possibility, we monitored the expression of tramtrack (ttk), a non-neuronal fate determinant expressed in pIIa cells (Audibert et al., 2005), following esg and scrt downregulation. In non-clonal heteroallelic scrt^jo11/jo16 tissue at 19 h APF, two-cell sensory clusters were composed of pIIb and the Ttk-positive pIIa cell (Fig. 2A,A′, arrowheads). By contrast, two-cell mSOs inside esg^G66B null clones induced in the heteroallelic scrt^jo11/jo16 background were composed of two Ttk-positive pIIa cells (Fig. 2A,A′, asterisks; 14 out of 84 mSOs analyzed). As such, downregulation of esg and scrt leads to a cell transformation whereby the neural precursor cell pIIb acquires a non-neural Ttk-positive pIIa cell fate.

Next, we studied whether this cell transformation was associated with an ectopic activation of the N pathway in the presumptive pIIb cell. To this end, in two-cell sensory clusters, we monitored N receptor activation by detecting nuclear translocation of the N intracellular domain (NICD) (Couturier et al., 2012). In control conditions, NICD was detected in the nucleus of the posterior pIIa cells (Fig. 2B,B′,D), whereas after esg and scrt downregulation NICD was also detected in the anterior cell (Fig. 2C,C′) and at a level comparable with that found in full N_on Ttk-positive pIIa cells (Fig. 2D). This level of NICD in the anterior cell was associated with Ttk staining (not shown). Altogether, these observations show that, after downregulation of esg and scrt, the N pathway is ectopically activated in the anterior presumptive pIIb cells. This led to the expression of non-neuronal fate determinants such as ttk, and the presumptive pIIb cell consequently adopted a pIIa cell fate.

Similarly, the phenotype observed after overexpression of esg or scrt might be explained by a cell transformation of pIIa non-neural precursor cells into pIIb neural precursor cells. We analyzed the expression of pros to identify pIIb cells following esg or scrt overexpression. As shown in Fig. 2E,F, we observed two-cell mSOs where both cells expressed cortical Pros, whereas in the control mSO only one precursor cell, i.e. the neural precursor pIIb cell, expressed Pros. These observations show that the presumptive non-neural precursor pIIa cell acquired a neural precursor pIIb cell identity after overexpression of esg.

We show here that the secondary precursor cell fate decision is sensitive to the level of activity of both Esg and Scrt. Moreover, Esg and Scrt are necessary and sufficient for the acquisition of the neural precursor pIIb cell fate.

We next explored the expression of esg and scrt during the completion of the bristle lineage (from 16 h to 24 h APF). As revealed by specific Esg polyclonal antibodies, a nuclear-located signal was clearly observed in some pIIb cells (3 out of 9 two-cell organs), and this staining vanished at later stages (Fig. 3A,A′). In vivo recordings of the Esg protein-trap fly line (esg^p1986::eGFP) revealed that Esg::GFP was absent in pI cells and was turned on progressively in pIIb cells, 30 min prior to their division (Fig. 3C, Movie 1). This was consistent with the low proportion of Esg-positive pIIb cells observed by immunofluorescence. Thereafter, Esg::GFP was detected in all pIIb descendants: the glial, pIIIb and sheath cells as well as in the neuron (Movie 1). According to the expression pattern obtained with antibodies, we believe that the GFP signal in pIIb descendants is due to the persistence of GFP molecules.

Using anti-Scrt polyclonal antibodies, a signal was clearly observed in the nucleus of pIIb as well as in pIIIb but was undetectable later on (Fig. 3B,B′). Scrt expression was turned on progressively in pIIb, since we observed many two-cell mSOs in which Scrt was absent or barely detectable (4 out of 15, second frame in Fig. 3B). In addition, Scrt was only expressed in neural progenitor cells and was not detected in pIIa. Using a protein-trap scrt::GFP line and an enhancer-trap scrt⁴³⁹-lacZ reporter line we confirmed the expression of Scrt in pIIb (Fig. 3D, Fig. S1). Furthermore, the low proportion of pIIb staining for Scrt suggests that the onset of Scrt expression occurred late in these cells.

We also analyzed the expression of sna and wor, the other two members of the Snail family. Using specific antibodies, we failed to observe wor expression in bristle cells at any period in the lineage (not shown). By contrast, sna was strongly expressed in terminal glial cells prior to entering into apoptosis (Fig. S2).

Taken together, the results show that, in the bristle lineage, Esg and Scrt are specifically co-expressed in late pIIb cells and in early pIIIb (Fig. 3E). This specific expression in both neural precursor cells is compatible with the possibility that esg and scrt control neural precursor cell fate in this lineage.

The neural to non-neural cell transformation induced after esg and scrt downregulation is reminiscent of that observed following overactivation of the N pathway. Conversely, overexpression of esg or scrt mimics N pathway downregulation. This prompted us to test whether esg and scrt function in the same genetic pathway as N in the control of fate determination in bristle cells. We analyzed the penetrance of the double-shaft phenotype observed after depletion of both esg and scrt in conditions where N activity was slightly dysregulated (Fig. 4A). Using the pannier (pnr)-Gal4 driver to express esg RNAi and scrt RNAi in the median region of the notum, we observed around four duplicated organs per notum when pupae were maintained at 30°C (n=77). This value was significantly reduced to two double-double bristles per notum when esg and scrt were downregulated in an N heterozygous background (N^ts-1/N⁺ pupae maintained at 30°C; n=76, P<0.001). Conversely, a significant increase in the number of duplicated organs (around six double-double bristles per notum) was observed when esg and scrt were downregulated in a background in which N activity was slightly increased (HS-N^intra pupae maintained at 30°C; n=63, P<0.001). Since we never observed double bristles under these two mild N dysregulation conditions alone, this indicates that Esg and Scrt are functionally related to N in the control of neural cell fate decisions in mSO cells. Furthermore, and more importantly, our data indicate that Esg and Scrt act on neural cell fate determination by counteracting N activity.

Where in the N pathway do Esg and Scrt act? To analyze whether they act on the reception of the N signal, we overexpressed esg or scrt in the context of ligand-independent N pathway activation by overexpressing the activated form of the N receptor N^intra. Overexpression of esg alone induced mSOs formed almost exclusively of Elav-positive neurons (Fig. 4B-B″), whereas overexpression of N^intra alone induced the formation of mSOs formed exclusively of Pdm1-positive outer cells (Fig. 4C-C″). Strikingly, co-expression of N^intra with Esg induced the formation of mSOs with a low level of Pdm1 (Fig. 4D,D′). Moreover, some of these cell clusters were formed exclusively of small strongly Elav-positive cells (Fig. 4D″). Similar results were obtained using Scrt instead of Esg to counteract N pathway activation (Fig. S3). Altogether, these data show that Esg or Scrt overrides the differentiation towards outer fate that results from enforced N^intra expression. Thus, Esg or Scrt prevents direct activation of the N pathway by N^intra overexpression, suggesting that they counteract N pathway activation by acting downstream of N signal reception.

Overexpression of N^intra and Esg from the beginning of puparation (0 h APF) resulted in similar clusters composed of Elav-positive cells (not shown). This rules out the alternative possibility that the preponderance of esg or scrt over N^intra could be due to a differential delay in the expression of esg, scrt and N^intra. It is relevant to note that mSO cells were not totally transformed to an external or internal fate. We believe that this was due to the mild overexpression procedure rather than to differences in cell sensitivity. Indeed, to avoid any interference of esg and scrt overexpression with pI specification, pupae were shifted after specification and maintained for a short period of time at the restrictive temperature.

To investigate the possibility that Esg binds directly to N^intra and thereby neutralizes its transcriptional action (Kim et al., 2013; Lim et al., 2011), we analyzed the expression of the N reporter construct Gbe-Su(H)-lacZ (Furriols and Bray, 2001). This synthetic promoter is composed exclusively of Su(H) and Grainyhead binding sites. In wing discs, this reporter was strongly expressed in the wing margin (Fig. 4E) and was unaffected after expression of Esg driven by engrailed or patched (Fig. 4F; data not shown). These data indicate that Esg does not interact directly with N^intra to antagonize N signaling. Our results support the idea that the N pathway inhibition induced by Esg impairs N-mediated transcriptional activation and requires native N target promoters.

To test whether Esg and Scrt negatively control N activity by downregulating the transcription of N target genes, we analyzed the effects of these factors on transcription of E(spl)m4, a well-described N target gene, (Bailey and Posakony, 1995; Nellesen et al., 1999).

First, using a DNA-mediated Esg pull-down assay, we studied whether Esg binds to the proximal 305 bp of the E(spl)m4 promoter as bait. This fragment harbors two E-boxes: E1, a canonical binding site for the proneural factors Achaete (Ac) and Scute (Sc), and E2, a canonical Esg binding site (Bailey and Posakony, 1995; Maeder et al., 2007). The data presented in Fig. 5A reveal the efficient retention of Esg by the E(spl)m4 promoter fragment (lane 2), whereas almost no retention was observed when beads were coated with E-box-free promoter fragments (lane 1, ftz promoter). Interestingly, very little binding was detectable when beads were coated with modified E(spl)m4 promoter fragments in which the canonical E-box E2 was replaced by an unrelated sequence (lane 3). These data indicate that Esg binds to the E(spl)m4 promoter through a sequence-specific mechanism.

Second, to test whether Esg can repress N target gene transcription in vivo, we used a construct in which the expression of β-Gal is under the control of the 510 bp proximal region of the E(spl)m4 promoter (Bailey and Posakony, 1995). In the wing pouch of the E(spl)m4::lacZ wing imaginal disc, β-Gal was detected in sensory cells at each side of the anterior wing margin, as well as in groups of chordotonal mSOs located in the hinge (arrowhead and asterisks, respectively, in Fig. 5B) (see also Bailey and Posakony, 1995). As shown in Fig. 5C,C′, this β-Gal staining was absent in sensory cells in the margin (arrow) and in chordotonal organs (bracket) located inside Esg-expressing patches (identified by the expression of GFP). This β-Gal suppression was not associated with an anti-neurogenic phenotype since Senseless-positive precursor cells were still detected in the region of Esg overexpression (not shown). These data show that overexpression of Esg blocks the expression of the E(spl)m4 reporter without affecting the selection of mSO precursor cells. A similar blockage of E(spl)m4-lacZ expression was observed when Scrt was overexpressed, suggesting that Esg and Scrt have similar repressive properties on N target gene expression (Fig. S4).

In an E(spl)m4 mutant reporter line in which both E1 and E2 were mutated [E(spl)-2m::lacZ (Bailey and Posakony, 1995)], β-Gal expression was only maintained in chordotonal sensory cells (Fig. 5D). Interestingly, this expression was not affected by overexpression of Esg (Fig. 5E,E′). Together, these results indicate that Esg prevents transcription of E(spl)m4 by binding to the promoter in a sequence-dependent manner. As such, these results suggest that Esg and Scrt downregulate N pathway activity by acting as transcriptional repressors of N target genes.

Since Esg and Scrt and proneural factors Ac and Sc bind E-boxes in a restricted domain of the E(spl)m4 promoter, we asked whether Esg or Scrt and Ac or Sc compete with each other for the control of E(spl)m4 transcription. To address this question, we studied Esg binding to the proximal 305 bp of the E(spl)m4 promoter in the presence of increasing amounts of Ac. We assayed extracts that consisted of a fixed amount of esg::GFP-expressing embryo extract complemented with a mixture of white embryo and ac-expressing embryo extracts at variable proportions such that the total amount of protein remained constant. We observed that the binding of Esg to beads coated with E(spl)m4 was displaced by increasing the amount of Ac extract (Fig. 5F, compare lanes 2-4). This displacement was almost complete since the level of Esg binding was comparable to the non-specific binding observed (Fig. 5F, compare lanes 4 and 1). These results suggest that Esg and Ac mutually interfere in binding to the E(spl)m4 promoter. As such, the downregulation of E(spl)m4 transcription by Esg is probably due, first, to the displacement of an activator complex brought by proneural factors, such as Ac, and second, to a repressor complex conveyed by Esg as a DNA-binding factor.

We have shown that esg and scrt are involved in the control of neural precursor cell identity by preventing N-mediated activation of target gene expression. To test whether Esg and Scrt also act as cell determinants in the control of neural precursor cell identity, we studied the identity of mSO cells when esg and scrt were downregulated in a context in which N pathway activity was also impaired. Transient inhibition of the N pathway (using N^ts-1) during the period of pI cell division induced mSO composed exclusively of Pros- and Elav-positive internal cells due to transformation of pIIa into pIIb cells (Fig. 6B). However, mSOs were composed exclusively of external cells (no Elav and Pros staining) when esg and scrt were downregulated (Fig. 6C). Interestingly, in pupae in which the N pathway and esg and scrt were simultaneously downregulated, mSO cells acquired an internal precursor cell identity (Fig. 6D). These cells co-expressed Pros and Elav at a lower level than those observed in terminal neurons or sheath cells (compare Fig. 6A-D), a distinctive molecular profile of neural precursor cells in this system. Strikingly, although cells adopted a neural-like fate, no clear distinction between neuron and sheath cells was observed under these conditions. This suggests that Esg and Scrt were also required in pIIIb progeny to implement the neuron sheath identities. Nonetheless, the acquisition of neural precursor cell fate was not impaired when esg, scrt and the N pathway were downregulated. Moreover, these data show that the external cell fate adopted by sensory cells after esg and scrt downregulation requires N signaling. Altogether, our results show that esg and scrt do not determine neural precursor cell fate. Rather, they are required to maintain an N_off state in these cells, an essential condition to implement the neural precursor cell identity.

Using a combination of genetic and cell biology analyses, we show that Esg and Scrt, which belong to the Snail gene superfamily, control neural cell fate commitment in the Drosophila mSO. Both transcription factors are expressed at early stages in the bristle lineage, particularly in the sublineage that gives rise to neural type cells. We show that these factors redundantly control the secondary precursor fate decision and, in particular, they maintain the N_off cell fate, i.e. the neural precursor pIIb fate. In addition, our results indicate that Esg and Scrt counteract the activity of the N pathway. Using the E(spl)m4 promoter as a paradigm of an N target gene, we present data suggesting that Esg and Scrt downregulate the transcription of N target genes by binding to the promoter. In addition, Esg and Scrt seem to act as proneural antagonists by outcompeting proneural transcription activators for binding to N target gene promoters. As such, we propose that Esg and Scrt, by repressing N target gene transcription, lock the neural precursor pIIb identity by absorbing any potential variation in N activity.

Cumulative evidence obtained in both vertebrate and invertebrate models indicates that Snail factors can function redundantly and compensate for each other's loss during essential developmental processes such as neurogenesis, hematopoiesis or chondrogenesis (Ashraf and Ip, 2001; Chen and Gridley, 2013; Pioli and Weis, 2014). We show that when both esg and scrt were downregulated, 48% of flies harbored mSOs composed only of outer cells (the double-shaft phenotype). This phenotype was not observed in flies in which esg or scrt was downregulated individually. To our knowledge, this is the first report showing that Esg and Scrt act redundantly and compensate for one another in the control of cell fate. Since Esg and Scrt are two distantly related members of the Snail superfamily, this redundancy suggests that ancestors of Snail factors were already involved in the control of cell fate identity.

Interestingly, esg is expressed in several stem cell types in Drosophila, including histoblasts, intestinal stem cells and male germ stem cells (Fuse et al., 1994; Micchelli and Perrimon, 2006; Streit et al., 2002). Several studies performed on these cell types have shown that Esg is required to maintain stem cell fate. Indeed, loss of esg drives stem cells to differentiate (Hayashi, 1996; Korzelius et al., 2014). An interesting question is whether scrt is expressed in these stem cells. If so, the phenotype observed in the esg mutant could be partially penetrant due to an scrt-mediated redundancy similar to that described here in neural progenitor cells. Since members of the Snail and Scrt families are also expressed in neural progenitors in vertebrates (Itoh et al., 2013; Vieceli et al., 2013; Zander et al., 2014), our results are likely to be of general relevance and might help improve our understanding of the mechanisms underlying vertebrate neurogenesis.

Downregulation of esg and scrt induced a transformation from N_off pIIb to N_on pIIa cells resulting in double-shaft mSOs. Interestingly, the number of transformations induced under esg and scrt downregulation conditions changed significantly with mild variations in the activity of the N pathway – variations that do not induce changes in cell fate by themselves. As such, we favor the idea that Esg and Scrt act redundantly to control neural precursor pIIb cell identity by downregulating N pathway activity. In agreement with this idea, we observed that Esg or Scrt overrides sustained activation of the N pathway. Furthermore, we show that Esg or Scrt antagonizes transcriptional activity of the E(spl)m4 promoter. E(spl)m4 is part of the E(spl) complex that contains direct N target genes encoding transcription factors that inhibit neural development (Delidakis and Artavanis-Tsakonas, 1992; Lai et al., 2000). This promoter is subject to two modes of transcriptional activation: one mediated by Ac and Sc proneural factors that operate by binding to E-box binding sites and the other mediated by the N pathway through Su(H) binding sites (Bailey and Posakony, 1995). We have shown here that Esg binds specifically to the E(spl)m4 promoter and negatively controls the transcriptional activity of this promoter in vivo. Our data also show that Esg and Ac/Sc compete for binding to the E(spl)m4 promoter. We do not know whether this reflects direct competition for similar E-box binding sites or is an allosteric interference between the factors binding to two different E-boxes. In any case, the result seems to be that Ac/Sc binding to E-box binding sites activates transcription, whereas Esg occupancy represses it. As such, we propose a mechanism for the transcriptional repression exerted by Esg on N target genes: first, it displaces the transcriptional activator brought by proneural factors, such as Ac, and second it brings a transcriptional repressor complex to the promoter (Fig. 7). Since E-box and Su(H) binding sites are frequently associated in promoters of E(spl) genes and other N target genes (Bernard et al., 2010; Cave et al., 2005, 2011; Maeder et al., 2007; Nellesen et al., 1999; Singson et al., 1994), it is expected that the Esg-mediated negative control of E(spl)m4 expression shown in the present work will reveal a more general control of the N signaling response exerted by Snail factors.

It is relevant to note that in vivo mapping of Esg binding sites on DNA using the DamID technique in Drosophila midgut cells has identified a number of N target genes. These include several E(spl) genes, such as E(spl)mα and E(spl)mβ. As such, these results show that Esg, and probably Scrt, binds the endogenous promoters of N target genes (Loza-Coll et al., 2014). Accordingly, transcriptomic studies have shown that dysregulation of Esg expression in Drosophila intestinal stem cells is associated with the upregulation or downregulation of N target genes. In agreement with our data, E(spl)m4 expression was upregulated when esg expression was downregulated (Korzelius et al., 2014). These results support our proposition that Esg or Scrt binds to the promoter of some N target genes and represses their transcription. The fact that a similar mechanism seems to be employed in other systems suggests that this function of Snail family transcription factors is widespread in diverse precursor cell types.

That Snail factors and proneural transcription activators may compete for similar DNA binding sites has already been suggested in a pioneer study on the Esg binding site (Fuse et al., 1994). A comparable mechanism in which Snail factors antagonize proneural activity has been proposed for the loss of bristles phenotype observed in Scutoid mutants, in which Snail and Esg are ectopically expressed in mSO cells. Moreover, consistent with our observations, overexpression of esg suppressed the excess of bristles observed after expression of the proneural factor asense alone (Fuse et al., 1999). Interestingly, it has been suggested that proneural factors and bHLH proteins encoded by E(spl)-HLH genes can compete for binding to E-box sequences resulting in transcriptional repression rather than activation (Jennings et al., 1999). We have shown that Esg and Scrt regulate the transcription of E(spl)m4, which is a Bearded factor rather than a bHLH factor. However, the fact that the promoter of several E(spl) genes harbors E-boxes raises the exciting possibility that Esg and Scrt downregulate N signaling at two different levels: first by downregulating transcription of N target genes, such as E(spl) genes (this work), and second by outcompeting E(spl) factors for common DNA sites. The existence of these potential interactions would depend on the heterogeneity of E-box sequences, on the sequence flanking the consensus core, and on the spatial configuration of E-boxes present in particular promoters (Chang et al., 2015).

N activation in secondary precursor cell determination is a rapid process, occurring during the first 30 minutes after birth (Couturier et al., 2012; Remaud et al., 2008). In addition, the onset of esg and scrt expression occurs in late pIIb cells. Furthermore, when esg, scrt and N signaling were downregulated, sensory cells acquired a neural cell fate. As such, these data indicate that Esg and Scrt are not determining the neural pIIb precursor cell identity per se. Additionally, these results show that cell transformation from N_off pIIb to N_on pIIa cells observed after downregulation of esg and scrt requires functional N signaling. Therefore, our results indicate that Esg and Scrt are involved in the maintenance of the N_off state. We propose that Esg and Scrt fix pIIb identity by absorbing any variation in N activity in the normally N_off pIIb precursor cell. An increasing number of reports indicate that stochastic fluctuations in biochemical reactions are the cause of variations in cellular behavior (Sanders et al., 2009; Süel et al., 2006). Our results suggest that Esg and Scrt, by downregulating N target gene transcription, may act as a buffer against these potential fluctuations in N signaling activity.

The data shown in Fig. 6 raise the possibility that Esg and Scrt might be needed to assure the identity of pIIIb progeny. However, because of their expression pattern, we favor the idea that the role of Esg and Scrt in maintaining cell fate is specific to pIIb. Other cell determinants also appear to act cell specifically in the bristle lineage. In particular, the N_off state of neural progenitor cells seems to be assured by diverse mechanisms. Hamlet, a transcriptional regulator that is involved in the fate decision of the two pIIIb daughter cells, also acts by reducing N signaling. Hamlet appears to modify the chromatin landscape and controls the accessibility of transcriptional activator complexes at the promoter of N target genes (Endo et al., 2011; Moore et al., 2004). Moreover, several reports suggest that N transduction involves different sets of mediators, Su(H) being essential to determine pIIa but not pIIIb cell fate decisions (Le Gall et al., 2008; Nagel et al., 2000; Wang et al., 1997). Although N signaling is involved in each binary decision in the bristle lineage, this pathway does not specify each cell identity. Thus, other factors must come into play. In the case of Esg and Scrt, these factors will act in secondary precursor cell decisions to maintain the N_off state of the pIIb neural precursor cells.

In conclusion, Esg and Scrt act redundantly to block N activity in secondary neural precursor cells. This effect seems to be mediated by downregulation of N target gene transcription. We propose that Esg and Scrt are involved in a cell-specific mechanism that maintains an N_off state in neural precursor cells. Since Esg and Scrt binding sites are known to be present in the promoters of several N target genes, we believe that this mechanism assures N activity damping in a wide variety of cell types.

Fly crosses were carried out at 25°C except where otherwise stated. Using the Gal4/UAS expression system (Brand and Perrimon, 1993) the following constructions were expressed: UAS-histone H2B::RFP (H2B::RFP) (Bellaïche et al., 2001), UAS-esg (gift from Shigeo Hayashi, Riken Center for Developmental Biology, Kobe, Japan), UAS-scrt (F0001921 FlyORF), UAS-scrt^RNAi [No 105201, Vienna Drosophila RNAi Center (VDRC)], and UAS-esg^RNAi (No 9794 and 9793, VDRC). As drivers, we used neuralized ^p72-Gal4 (neur-Gal4), pannier-Gal4 (pnr-Gal4), engrailed-Gal4 (en-Gal4) and HS-Gal4. The protein-trap line esg^P01986::GFP (FlyTrap), scrt::GFP (#318073, VDRC; Sarov et al., 2015) and the enhancer-trap scrt::lacZ (Bloomington Drosophila Stock Center) were used to monitor esg and scrt expression. HS-scrt transgenic flies were generously provided by E. Bier (University of California San Diego, La Jolla, USA).

Somatic clones were obtained by means of the FLP/FRT recombination system (Xu and Rubin, 1993) using Ubx-FLP or HS-FLP lines. The w; FRT40A esg^G66B/CyO line was used to generate esg null clones. Escapers of the strong allelic combination scrt^jo11/scrt^jo16 (gift from E. Bier) were used to analyze the scrt mutant phenotype. Flip-out overexpression clones were obtained using the HS-FLP; tubulin FRT-stop-FRT Gal4; UAS::GFP line. The E(spl)m4::lacZ and mutant form E(spl)-2m::lacZ lines were obtained from J. Posakony (University of California San Diego, La Jolla, USA). For wing disc clones, larvae maintained at 25°C were heat shocked at 24 h and 48 h after egg laying (37°C, 1 h) and dissected at late third instar. For intralineage clones in the wing margin, pupae at 11 h APF were heat shocked twice (37°C at 11 h and 12 h APF) and dissected at 24 h APF.

Transformation of pIIb into pIIa was realized by ectopic expression of the active form of N (N^intra) via an HS promoter. White pupae were collected, kept for 17 h at 25°C and then heat shocked for 1 h at 37°C. We dissected them 1 h 30 min after heat shock. Transformation of pIIa into pIIb was realized by downregulation of N using a thermosensitive allele (N^ts-1). In this case, fly crosses and embryonic and larval development were carried out at 18°C. White male pupae were then collected, kept for 17 h at 25°C and then at 30°C for 1 h before dissection.

The conditional neur-Gal4, tub-Gal80^ts and pnr-Gal4, tub-Gal80^ts were used to overexpress both N^intra and Esg or Scrt. To avoid the lethality induced by overexpression of these factors at earlier stages of pupal development, pupae were maintained at 18°C and shifted to 30°C at 15 h APF.

To test epistasis between esg, scrt and the N pathway, pupae from the cross of females (N^ts-1/N^ts-1;;pnr>Gal4/TM6Tb)× males (+/Y; UAS-RNAi scrt/UAS-RNAi scrt; UAS-RNAi esg/UAS-RNAi esg) were maintained at 18°C from 0 h to 28 h APF (equivalent to 14 h APF at 25°C), shifted to 30°C for 4 h and returned to 18°C for 12 h until dissection (equivalent at 24 h APF at 25°C).

Pupal nota were dissected between 17 h and 35 h APF and processed as previously described (Gho et al., 1996). Antibodies are described in the supplementary Materials and Methods.

Polyclonal anti-Scrt antibodies were immune purified from rabbit serum immunized with a mix of two Scrt peptides (IYRPYSLDDKPAHGYR and YTYEAFFVSDGRSKRK) mixed with Freund's complete adjuvant and boosted three times by injection with Freund's incomplete adjuvant (Covalab).

Images were obtained on an Olympus BX41 fluorescence microscope (40× or 60× oil immersion objective) equipped with a Yokogawa spinning disc and a CoolSnapHQ2 camera driven by Metaview software (Universal Imaging). Images were processed with ImageJ (NIH).

All experiments were performed with the same batch of Drosophila protein extract obtained from daughterless-Gal4/UAS-esg::GFP (da>esg::GFP) embryos taken 20 h after egg laying, aliquoted and stored at −80°C, 1 mg of devitellinized embryos in 5 µl de lysis buffer. DNA templates were generated by PCR using 5′ biotinylated primers. An ftz template was obtained using 5′-GGGAGTTGCGCACTTGCTTG-3′ and 5′-GTGCACGCAACGCTGGTGAG-3′ primers. An E(spl)m4 template was obtained using 5′-AAGGATCCTTTCGAACCGAAACTGTG-3′ and 5′-AGAACCCGAAGCCGAGCAGG-3′, which correspond to the proximal 305 bp of the E(spl)m4 promoter bearing the canonical (E2) E-box ACAGGTG. In the E(spl)m4 mutated promoter the more proximal canonical E-box sequence was replaced by a TAGAATT sequence. Biotinylated DNA was coupled to streptavidin-coated magnetic beads (M280, Dynal Biotechnology), with 0.1 mg beads per 200 ng DNA, overnight at 4°C. The beads were washed three times in B&W buffer (as recommended by Dynal Biotechnology) and streptavidin-immobilized DNA was saturated for 1 h in PBS containing 20% horse serum before incubation for 1 h with the da>esg::GFP embryo protein extract in PBS containing 0.15% Triton X-100. Protein extract and beads were separated according to the manufacturer's instructions and washed four times with 100 mM NaCl, 25 mM NaH₂PO₄. Each fraction was then processed for SDS-PAGE. GFP was revealed using rabbit anti-GFP (1/800; Roche) and horseradish peroxidase coupled to an anti-rabbit antibody (1/1000; Promega).

Competition experiments were performed by incubating beads coated with E(spl)m4 templates with a fixed amount of protein extract from da>esg::GFP embryos complemented with the same volume of a protein extract from da-Gal4/UAS-achaete and white embryos at a ratio of 0:2, 1:1 and 2:0. As such, the total amount of protein in the final protein extract mixture was relatively constant.

We specially thank James W. Posakony for providing E(spl) fly stocks; Yohanns Bellaiche for generous aid; the fly community for fly strains and antibodies; Heather McLean, Fred Bernard and Yohanns Bellaiche for critical reading of the manuscript; and Carine Torset for technical help.