SoxNeuro and Shavenbaby act cooperatively to shape denticles in the embryonic epidermis of Drosophila

Information from signaling pathways must be interpreted and modulated for the proper execution of cellular events during development. Much of developmental biology research has focused on the signaling pathways that govern cell fate specification. However, this approach reveals only what the upstream signals are, not what factors integrate and implement this activity to orchestrate downstream morphogenetic events. Understanding the bridge from signal to pattern will ultimately reveal how cells take proper form during development.

Pattern in developing tissues is generated by highly organized cell movements and morphogenetic rearrangements (Pilot and Lecuit, 2005). For example, differentiation of epithelial cells often features cell shape changes that are driven by the actin-based cytoskeleton. Such processes include the formation of denticles, hairs or sensory bristles in Drosophila (Appel et al., 1993) and the stereocilia of the vertebrate inner ear (Tilney et al., 1992). To understand how these developmental events are fine-tuned requires an in vivo system where signaling pathways and their output can be readily analyzed. Denticle formation in the ventral epidermis of the Drosophila embryo provides an excellent system in which to answer such questions (reviewed by Bejsovec, 2013). Denticles are hook-like structures produced by ventral epidermal cells during late embryogenesis (Fig. 1A). They form in a stereotyped pattern as actin-based protrusions that extend from the apical side of the cell and become coated with chitinous cuticle. These provide traction for the crawling larva after the embryo hatches. Denticles are organized in segmentally repeating units known as belts. Each belt is composed of six rows of cells, where each row produces a denticle of specific size, shape and polarity (Fig. 1B,C). The denticle belts are separated by regions of ‘naked cuticle’, which are secreted by cells that do not produce any actin-based protrusions. The naked cuticle region is specified by high-level Wg signaling (Bejsovec and Martinez Arias, 1991; Noordermeer et al., 1992), which mediates repression of the transcription factor shavenbaby (svb; ovo – FlyBase) (Payre et al., 1999) in a striped pattern (Fig. 1D-F). Svb is a zinc-finger transcription factor (Mevel-Ninio et al., 1995; Andrews et al., 2000; Delon et al., 2003) that activates a multitude of genes involved in actin organization, extracellular matrix secretion, or cuticle deposition (Chanut-Delalande et al., 2006). Svb has been considered the primary effector of denticle morphogenesis, integrating inputs from multiple signaling pathways and converting them to structural output.

The cytoskeletal processes that form denticles begin with polarized actin filaments that accumulate and acquire shape over a period of ∼2 h (Dickinson and Thatcher, 1997; Price et al., 2006; Walters et al., 2006; Bejsovec and Chao, 2012). During this time, proteins involved in cytoskeletal reorganization, many of which are transcriptionally regulated by Svb, begin to localize to the denticles. For example, Svb directs the expression of forked, which encodes an actin-bundling protein that functions in denticle and dorsal hair formation (Chanut-Delalande et al., 2006) as well as in adult bristle formation (Grieshaber and Petersen, 1999). The genes singed (Cant et al., 1994) and shavenoid (Ren et al., 2006) are also controlled by Svb, and function during embryogenesis in actin dynamics and denticle formation. Other proteins that function in cytoskeletal dynamics, such as Diaphanous, Enabled and Arp2/3, are not influenced by Svb activity but localize to denticles and aid in their construction. Human homologs of these factors promote an analogous process to construct stereocilia in the developing ear; their loss of function results in physical defects that can lead to hearing impairment and deafness (reviewed by Chanut-Delalande et al., 2012).

Here we explore how the morphological diversity of denticles within the belt is generated. Wg activity is required for the diversification of denticles, in a process that is less clearly understood than the Wg-mediated specification of naked cuticle. In the absence of wg gene function, not only is the naked cuticle component of the pattern lost, but also the denticles produced are reduced to a single type of denticle, similar to those in row 5 (Nusslein-Volhard et al., 1984, 1985; Bejsovec and Wieschaus, 1993). This diversification cannot be explained by Svb modulation alone. Although svb is necessary and sufficient for denticle formation, ectopically expressed svb is unable to recapitulate the denticle morphologies characteristic of the wild-type belt (Payre et al., 1999). This suggested that additional factors are required for the formation of specific denticle shapes. We have discovered that a second transcription factor, encoded by SoxNeuro (SoxN), is necessary and sufficient for denticle formation, and its ectopic co-expression with svb rescues denticle morphology to more closely approximate wild-type structures. These observations, along with the previously described antagonistic relationship of SoxN with the Wg pathway (Chao et al., 2007), suggest that SoxN might act at the interface of signaling pathway and effector output during denticle morphogenesis.

Wg signaling generates denticle diversity in the ventral epidermis between 4 and 6 h after egg laying, and then specifies naked cuticle from 6 to ∼10 h (Bejsovec and Martinez Arias, 1991). We have recently found that Wg is also required at later developmental stages for proper denticle morphogenesis (Bejsovec and Chao, 2012). These findings suggested that Wg activity in denticle-producing cells might titrate Svb levels, where different threshold levels of Svb would direct the diverse morphologies within a belt. To test this idea, we obtained several well-characterized svb enhancer constructs that drive expression in ventral denticle-producing cells as well as in dorsal trichome-forming cells (McGregor et al., 2007; Frankel et al., 2010). The 7.3 ‘proximal’ enhancer, which is located close to the svb transcription start site (McGregor et al., 2007), drives expression in all of the ventral denticle-producing cells (Fig. 1D,E), whereas enhancers located more distal to the svb start site drive expression in subsets of the overall svb domain. The DG3 enhancer promotes expression in a single row of cells at the posterior edge of each belt (Fig. 2A), in roughly the position where row 5 denticles will form in the mature epidermis. This expression domain expanded slightly when Wg activity was partially reduced in wg^PE2 mutant embryos (Fig. 2B). The wg^PE2 missense allele produces a protein that is distributed normally across the segment and is sufficient to generate denticle diversity, but cannot bind the Frizzled 2 receptor strongly enough to specify naked cuticle (Hays et al., 1997; Moline et al., 2000). Expansion of DG3 enhancer-driven expression was greater but still not uniform when wg activity was removed completely in wg^ts mutant embryos at restrictive temperature (Fig. 2C). By contrast, the E3 enhancer drives two stripes of expression in each denticle-producing belt: an anterior stripe that roughly prefigures where rows 1 and 2 will form, and a more posterior stripe at row 5 (Fig. 2D). This more complex pattern of ventral expression suggests that there might be multiple inputs into the activity of this enhancer. Indeed, we found that the two stripes of E3-driven expression respond differently to varying levels of Wg signaling. Reduced Wg activity in wg^PE2 embryos allows the anterior stripe to expand in an anterior direction (Fig. 2E) into cells that produce diverse denticle types at a position where naked cuticle would form in a wild-type embryo. We observe a similar effect when wg^ts embryos are cultured at permissive temperature for the first 5 h of development (Fig. S1A,B), which generates denticle diversity without naked cuticle specification (Bejsovec and Martinez Arias, 1991). The opposite effect was seen when wg function was eliminated more completely; the anterior E3 stripe was unchanged while the posterior stripe expanded in width to match the anterior stripe, both in wg^ts mutants cultured at restrictive temperature (Fig. 2F) and in wg^CX4 null mutants (Fig. S1C). Thus, the DG3 and E3 promoter elements of svb are sensitive to Wg activity level.

The differential responsiveness of svb enhancers led us to test whether different levels of Svb might specify distinct denticle shapes. We therefore developed an assay for denticle formation. The wg-Gal4 line drives ectopic UAS transgene expression in a row of one to two cells that would be fated to produce naked cuticle (Fig. 1F). Thus, any denticles formed in this ectopic location are the direct result of transgene activity. wg-Gal4-driven UAS-svb (wg>svb) produced a stripe of ectopic denticles that are long and thin, lacking the widened base that is characteristic of denticles within wild-type belts (Fig. 3A). This defective morphology has been observed previously when UAS-svb was driven with a variety of Gal4 drivers (Payre et al., 1999). We measured the length and width of these ectopic denticles and compared them with wild-type denticle measurements: wg>svb denticles consistently fell below the normal width of any denticle in the wild-type belt (Fig. 1C, Fig. 3B). This observation suggests that either higher levels of Svb than are produced by this system are required for proper denticle shape, or that additional factors are required.

To differentiate between these two possibilities, we generated fly lines that carry multiple copies of the UAS-svb transgene. Driving multiple UAS-svb transgenes with wg-Gal4 produced denticles that are on average 13.9% longer than those produced by a single transgene (Fig. 3C,D). The additional length of these wg>svb;svb denticles shows that Gal4 levels are not limiting for driving both UAS transgenes at high levels. However, these longer ectopic denticles were thin and lacked the widened base (Fig. 3D), and thus were similar in shape to those produced by a single transgene. We conclude that Svb transcription factor output controls denticle length, but alone is not sufficient to produce normal denticle shape.

Because higher levels of Svb did not rescue ectopic denticle morphology, we suspected that wg-Gal4-expressing cells might lack additional factor(s) necessary for proper denticle morphogenesis. We had previously isolated mutations in SoxN as suppressors of wg loss-of-function phenotypes, and we demonstrated that this Sox class transcription factor plays a role in downregulating Wg pathway activity (Chao et al., 2007). SoxN is expressed in the bands of cells that produce denticle belts, and its activity is required for establishing the proper expression domain of svb (Overton et al., 2007). There is also a feedback loop whereby Svb is necessary for the maintenance of SoxN expression (Overton et al., 2007). When we examined SoxN in our denticle assay system we found that, like svb, it was able to promote ectopic denticle formation. wg>SoxN ectopic denticles were fewer in number and were shorter than those produced by ectopic svb (Fig. 3E,F). On average, the wg>SoxN ectopic denticles also had wider bases than wg>svb denticles, raising the possibility that SoxN activity controls base width whereas Svb controls length. Thus, the combined activities of the transcription factors might be necessary for shaping denticles properly.

To test this hypothesis, we co-expressed UAS-svb and SoxN with wg-Gal4 in our denticle formation assay. Expression of both transcription factors together dramatically rescued the morphology of ectopic denticles, particularly the widened base that is highly characteristic of most wild-type denticles (Fig. 3G). Comparing length and base width values with those of wild-type denticles revealed that these ectopic denticles fell within the range of wild-type denticle measurements (Fig. 3H). Furthermore, these ectopic denticles exhibited diverse morphologies. The majority were similar to row 2 denticles, with a minority resembling the denticles of rows 1 or 3.

Because our denticle formation assay involved the expression of transgenes in the endogenous wg expression domain, it was possible that high-level Wg signaling in these cells might somehow block the ability of Svb to direct proper denticle morphogenesis. Since SoxN downregulates Wg pathway activity, the co-expression of SoxN might then reduce Wg activity, removing its putative block on Svb and indirectly altering denticle shape. To test this possibility, we used a different Gal4 driver line to drive expression outside of the wg expression domain. The engrailed (en)-Gal4 transgene is expressed in the two rows of epidermal cells posterior to the wg-expressing cells: the row of naked cuticle cells just anterior to the denticle belts and the first row of denticle-producing cells (DiNardo et al., 1985). Driving UAS-svb in the en domain produced ectopic denticles similar in shape and size to those produced with wg-Gal4 (Fig. S2A). Likewise, en-Gal4-driven co-expression of svb and SoxN rescued denticle morphology, just as wg-Gal4-driven co-expression did (Fig. S2B). Because distance from Wg production did not alter ectopic denticle morphology, it is likely that the ability of SoxN to rescue morphology is at least partially independent of its ability to suppress Wg activity.

To further explore the interplay of Svb and SoxN in driving denticle morphogenesis, we examined loss-of-function mutant phenotypes. SoxN null mutations result in a slight decrease in the denticle belt expanse and some defects in denticle morphology (Chao et al., 2007; Overton et al., 2007) (Fig. 4A). By contrast, svb loss-of-function mutations show a more severe loss of both ventral denticles (Fig. 4B) and dorsal trichomes (Payre et al., 1999). Some denticles remain in the most posterior rows, namely rows 5 and 6, of each belt; these remaining denticles have been described as ‘atrophied’ or ‘blunted’ (Payre et al., 1999). Coincidentally, immunolocalization of SoxN in wild-type embryos suggests that it is present at higher levels in rows 5 and 6 of the denticle-producing cells (Overton et al., 2007). Thus, SoxN activity might account for the denticles remaining in svb mutants. If so, loss of both svb and SoxN together would be predicted to block denticle formation completely. We found this to be the case: svb;SoxN double-mutant embryos produce uniformly naked cuticle (Fig. 4C), suggesting that the atrophied structures displayed by svb mutants result from SoxN activity.

An alternative explanation for this all-naked phenotype is that Wg signaling might be hyperactivated in the absence of svb and SoxN. High and uniform expression of wg throughout the epidermis results in a comparable loss of all ventral denticles, although such embryos also show reduced body size and severe head defects (Noordermeer et al., 1992), which we do not observe. We examined the expression of en, a known Wg target gene (DiNardo et al., 1988; Ingham et al., 1988) that expands in expression in embryos with hyperactivated Wg signaling (Noordermeer et al., 1992; Pai et al., 1997). We found that en expression did not expand in svb;SoxN double mutants (Fig. 4F,G), as compared with their heterozygous siblings (Fig. 4D,E). Thus, the naked cuticle phenotype shown by svb;SoxN double mutants was not due to hyperactivity of Wg signaling, but rather to a downstream requirement for both transcription factors in driving denticle formation.

We have shown that SoxN is sufficient to produce denticles when ectopically expressed and appears to be responsible for the atrophied denticles present in svb mutants. However, because of the feedback interactions between SoxN and svb, we wondered to what extent SoxN could promote denticle formation independently of svb. We drove high and uniform expression of UAS-SoxN in the epidermis of svb mutants using the E22c-Gal4 line. E22c>SoxN in svb mutants resulted in the production of ventral denticles with the atrophied morphology (Fig. 5A,B). These atrophied denticles were arranged in a segmental pattern reminiscent of denticle belts, with fewer denticles in the region where naked cuticle would normally form. This suggests the presence of additional pattern information that is independent of Svb and SoxN. We also observed the restoration of some trichomes on the dorsal surface (Fig. 5C,D). This was surprising because dorsal trichomes had been definitively connected with svb activity (Sucena and Stern, 2000) and dorsal epidermal cells do not normally express SoxN (Cremazy et al., 2000). Thus, SoxN appears able to substitute for Svb in dorsal element formation, suggesting that svb and SoxN have some shared target genes.

We set out to identify genes that might be subject to Svb and SoxN co-regulation. We first tested whether any known Svb target genes were also responsive to SoxN. A major class of Svb targets are the Zona Pellucida domain (ZPD)-containing proteins that associate with the apical extracellular matrix of cuticle (Fernandes et al., 2010). Each ZPD protein localizes to a specific compartment within the developing denticle, with non-redundant functions in sculpting denticle shape. The well-studied ZPD gene miniature (m) was activated robustly by ectopic expression of svb (Chanut-Delalande et al., 2006) (Fig. 6A). We found that ectopic expression of SoxN did not activate ectopic m expression (Fig. 6B), nor did it contribute to the strength of m expression when co-expressed with svb (Fig. 6C). Therefore, SoxN does not appear to play a role in m regulation. Furthermore, this indicated that any Svb that is coincidentally induced by ectopic SoxN through the feedback loop is not sufficient to drive m target gene activation.

Next we tested ZPD candidates that were found to have residual expression in svb mutants, such as nyobe (nyo). In svb mutants, nyo expression levels were reduced substantially, but some expression remained in the posterior rows of denticle-producing cells (Fernandes et al., 2010). We found that expressing either svb or SoxN alone was insufficient to activate detectable expression of nyo (Fig. 6D,E). Even doubling the amount of ectopic svb was insufficient to activate nyo (Fig. S3). By contrast, co-expression of svb and SoxN together activated moderate nyo expression (Fig. 6F). Thus, nyo requires input from both Svb and SoxN for proper expression.

In examining other ZPD gene products, we discovered two Svb targets that can be activated by either Svb or SoxN. The gene dusky-like (dyl) was activated when svb was expressed ectopically (Fig. 6G), but was also detected strongly when SoxN alone was expressed ectopically (Fig. 6H). There was no discernible difference in dyl expression when both factors were ectopically co-expressed (Fig. 6I). By contrast, the gene neo was activated weakly when either svb or SoxN alone was expressed (Fig. 6J,K), but was expressed strongly when both genes were expressed together (Fig. 6L). Thus, we have discovered four classes of Svb downstream targets: those that are activated exclusively by Svb; those that require input from both Svb and SoxN to be activated; those that are activated by either Svb or SoxN with apparently equal efficacy; and those that can be activated by either Svb or SoxN but require both for optimal expression (Fig. 6M).

We next tested whether SoxN could direct the expression of epidermal target genes independently of Svb. We used an image search tool in Fly Express (http://www.flyexpress.net/) to find genes that exhibit a pattern of expression similar to that of SoxN. We identified two uncharacterized genes, CG16885 and CG30101, that met this criterion. These genes are expressed solely in the ventral epidermis, in cells that prefigure the denticle belts. Ectopic expression of SoxN, but not of svb, activated CG16885 expression (Fig. 7A,C), indicating that it is a downstream target gene of SoxN. CG16885 had been previously tested and eliminated as a potential svb target via in situ hybridization, as its expression did not change in svb loss-of-function embryos (Kondo et al., 2010) nor in embryos in which svb was knocked down with RNA interference (Fig. 7E). Thus, CG16885 expression is not regulated by Svb. A second gene, CG30101, had not been previously tested for regulation by Svb. We found that ectopic expression of SoxN was sufficient to activate CG30101 (Fig. 7B) and that ectopic expression of svb failed to activate CG30101 expression above background levels (Fig. 7D). As with CG16885, driving uniform embryonic epidermal expression of svb RNAi produced no discernible change in CG30101 expression (Fig. 7F), indicating that Svb does not regulate CG30101. Under the same conditions, svb RNAi phenocopies svb null embryos and results in the robust loss of m expression (Fig. 7G). Thus, we have identified two epidermal target genes that can be activated by SoxN without input from Svb.

Consistent with these results, both CG16885 and CG30101 require input from SoxN for their normal expression during development. We hand-selected SoxN homozygous mutant embryos by absence of GFP signal from a tagged balancer chromosome, and performed in situ hybridization to detect CG16885 or CG30101 expression. The expression of m was also tested as a control. In SoxN mutants, m expression was not affected in the dorsal epidermis, although there appeared to be somewhat weaker expression than normal in the ventral epidermis (Fig. 7J,M). This is likely to be due to the feedback loop whereby SoxN boosts svb expression in ventral, but not dorsal, cells. Ventral CG16885 expression was significantly reduced in SoxN mutants compared with their heterozygous siblings (Fig. 7H,K). Expression of CG30101 was also diminished in SoxN mutant embryos, although not as dramatically as CG16885 (Fig. 7I,L). Together, these results suggest that SoxN is necessary and sufficient for the proper expression of CG16885 and CG30101 during development of the embryonic epidermis.

The transcriptional regulation of these SoxN target genes and SoxN-Svb co-targets is likely to be mediated directly. Genome-wide binding patterns and expression profiling of SoxN using DamID and ChIP techniques (Ferrero et al., 2014; Carl and Russell, 2015) have shown that SoxN can bind in regulatory regions of nyo, CG16885 and CG30101. Likewise, Svb binding sites have been identified in enhancers that control downstream Svb target genes, including nyo (Menoret et al., 2013). Locating functional binding sites for Svb and SoxN in nyo regulatory regions will help us to understand how these two transcription factors might cooperatively activate the target gene. We performed a computational search for SoxN binding consensus sequences in nyo, as well as in CG16885 and CG30101, and identified a number of predicted binding sites (Fig. 8A, asterisks; Table S1). In nyo, none of the putative SoxN binding sites that we found falls within the previously identified enhancer region (Fig. 8A, orange bar). All are in introns upstream or downstream of the Svb-binding enhancer element, although one site is located within 1 kb of this enhancer. Since the previously defined region may represent a minimal core enhancer, additional flanking sequences might be required to achieve the robust output required during development or during times of variation in environmental conditions, such as temperature fluctuation (Ludwig et al., 2011). Therefore, it is possible that this enhancer element may extend to the nearby predicted SoxN binding site. Alternatively, several independent regulatory elements, some of which contain SoxN binding sites, might be integrated to control expression of nyo.

Like nyo, the sequences surrounding and within both CG16885 and CG30101 contain a number of computationally predicted SoxN binding sites (Fig. 8A, Table S1). The CG16885 sequence contains several sites in the region immediately upstream of the transcription start site, as well as within the intron and second exon. The sites found in CG30101 all reside in exons, except for one located upstream of the start site. Several of the predicted SoxN binding sites in nyo, CG16885, and CG30101 are strongly conserved in other Drosophila species (Fig. 8A, green asterisks), suggesting that these sites might be functional in the observed transcriptional control. We do not yet know whether the CG16885 and CG30101 gene products play a role in shaping denticles, but both show similarities to cuticular proteins in a number of other insects. As with the ZPD family of proteins (Jazwinska and Affolter, 2004), the protein sequences of these SoxN targets predict a signaling peptide at the N-terminus, which may localize each to the membrane or to the extracellular matrix. No other predicted domains were detected by BLAST (Altschul et al., 1990) in either protein. Further investigation of this novel class of proteins might provide mechanistic insight into denticle morphogenesis.

The Drosophila embryonic cuticle provides a powerful system in which to study the genetic control of tissue patterning. Elegant studies have revealed that regulation of Svb controls dorsal and ventral cuticle patterning in drosophilids (reviewed by Delon and Payre, 2004; Stern and Frankel, 2013) through its activation of numerous effector molecules (Chanut-Delalande et al., 2012). Yet whether and how Svb activity might control the row-specific shapes of ventral denticles has been unclear. The group of ZPD proteins regulated by Svb, some of which we have shown here are also regulated by SoxN, are targeted to specific apical regions and are required to maintain contact between the cell membrane and the cuticle (Fernandes et al., 2010). Although these studies noted no difference in ZPD protein levels among the rows of denticle-producing cells, it remains possible that subtle differences, below the level of detection, might play some role in denticle diversity. We found that the svb E3 enhancer element shows differential responsiveness to Wg signaling, whereby low levels of signaling in the wg^PE2 mutant correlated with both expansion of the E3 anterior stripe and increased denticle diversity. This suggests that svb expression, and hence its transcriptional targets, may be exquisitely sensitive to graded levels of Wg activity in the wild-type embryonic epidermis. Various ZPD proteins have been shown to act in cell shape remodeling in other systems (Roch et al., 2003; Sapio et al., 2005), where a balance of the precise levels of each protein may be critical to the final shape adopted. Our finding that Svb and SoxN differentially activate ZPD genes, and other targets, raises the possibility that each row of cells within a belt might experience slightly different levels of the gene products that control cell shape.

Indeed, our experiments suggest that Svb and SoxN may be responsible for two different aspects of denticle construction. We have shown that increased levels of Svb produce longer denticles, without altering their overall shape. Thus, Svb might activate a battery of genes that promote lengthening of the actin-based protrusions that form denticles. Conversely, expression of SoxN alone produced denticles with a short, blunted structure. Only when SoxN was co-expressed with svb did the ectopic denticles acquire a widened base that more closely matched the wild-type shape. These results lead us to propose a model whereby Svb controls genes that specify denticle length whereas SoxN controls genes that regulate base circumference (Fig. 8B). Varying the levels of each transcription factor with respect to the other could account for the varying length versus width that characterizes each denticle row in the wild-type belt. However, there might be an upper limit to the response in denticle-producing cells. In wg loss-of-function mutants, both Svb and SoxN are expressed at very high, uniform levels (Payre et al., 1999; Overton et al., 2007). This extreme overproduction of Svb and SoxN corresponds with a lawn of denticles (Fig. 2C) where almost all have the size and shape of row 5 denticles, which are the large denticles at the posterior of the wild-type belt (Fig. 1B).

In addition to Svb and SoxN, one or more as yet uncharacterized factors might modulate denticle size and shape, since we observe some segmental denticle diversity when SoxN is expressed uniformly in a svb mutant embryo. We propose that the independent or combined effects of Svb and SoxN, and possibly other factors, set a specific expression level for a given target gene, and that this precise level of target gene output within a denticle row contributes to the final shape of the denticle produced. The set point for Svb and SoxN levels might be achieved by response to the graded distribution of secreted Wg, and/or of other signals within each epidermal segment, to produce the diverse denticle morphologies in each belt. Consistent with this idea, svb in situ hybridization images give the impression of higher mRNA accumulation in the more posterior rows of denticle-secreting cells (Payre et al., 1999), and the same effect has been noted for SoxN antibody staining (Overton et al., 2007). These posterior row cells, which produce the largest denticles, are the furthest from the stripe of Wg-producing cells in each segment (Fig. 1F) and so would experience the least repression from Wg signal transduction.

A precise balance of protein levels might be a common theme for Sox class transcription factors, and may be relevant to cell shape control in other systems. The activity of the SoxN mammalian homolog Sox2 has been shown to depend on specific levels of the protein in embryonal carcinoma cells and embryonic stem cells (Boer et al., 2007), as well as in cochlear cells (Kempfle et al., 2016), which produce the stereocilia. Further characterization of the role of SoxN in shaping Drosophila denticles might help elucidate the forces that shape stereocilia in the human inner ear, and that go awry in some hereditary forms of deafness.

w¹¹¹⁸ flies were used as wild-type controls, to match the genetic background of transgenic stocks. The UAS-svb, 7.3-Gal4, and svb-lacZ lines were gifts from F. Payre (Paul Sabatier University, Toulouse, France) and D. Stern (HHMI Janelia Research Campus, USA). The wg^ts and wg^PE2 alleles are described by Baker (1988) and Bejsovec and Wieschaus (1995), respectively. UAS-SoxN was a gift from S. Russell. The SoxN^NC14 allele was generated by EMS mutagenesis (Chao et al., 2007). The UAS-svbRNAi line was obtained from the Vienna Drosophila Resource Center. All other mutations, balancer chromosomes and Gal4 lines were obtained from the Bloomington Stock Center. Flies were reared on cornmeal-agar-molasses and embryos were collected on apple juice agar plates; all were cultured at 25°C. Cuticle preparations were performed as described (Jones and Bejsovec, 2005). In the case of svb² and SoxN^NC14, hatching rates were calculated after outcrossing to remove balancer chromosomes. Cuticles of all progeny, i.e. unhatched embryos as well as hatched larvae, were examined.

Cuticle images were captured with a SPOT camera (Diagnostic Instruments) and were processed using FIJI software (Schindelin et al., 2012). For wild-type embryos, denticle belts in abdominal segments 4 and 5 were used for measurements. For both wild-type and ectopic denticles, only denticles within 75 µm of the most ventral region were used. At least five different animals were used to collect measurements, with over 70 denticles scored for each plot. Width was measured at the base of each denticle and length was measured as the distance from the midpoint of the base to the tip.

Embryos were collected and aged to the specified developmental time. For immunostaining, embryos were dechorionated in bleach and fixed for 20 min in 4% formaldehyde in PEM buffer (0.1 M PIPES, 1 mM EDTA, 2 mM MgSO₄, pH 6.9). For actin filament visualization, vitelline membranes were removed by vigorously shaking in a 1:1 mixture of ethanol and heptane and washing three times in fresh ethanol. Rhodamine phalloidin (Molecular Probes/ThermoFisher) and anti-β-galactosidase antibody (Promega, Z3781) were both used at 1:500. Anti-En antibody (Developmental Studies Hybridoma Bank, University of Iowa) was used at 1:50 and anti-GFP antibody (EMD Millipore, MAB3580) was used at 1:500 on embryos that were devitellinized using methanol instead of ethanol. Secondary antibodies (Jackson ImmunoResearch, 115-165-003, 111-225-144, 111-165-144) were used at 1:500. Images were captured on a Zeiss 510 confocal microscope. To examine cuticles, eggs were allowed to develop for 24 h at 25°C, dechorionated with bleach and then mounted in Hoyer's medium. Images were captured with SPOT camera (Diagnostic Instruments) on a Zeiss Axioplan microscope, and were processed with SPOT imaging.

Hybridizations were performed as described (Wilk et al., 2010). Antisense probes were made from cloned PCR products of genomic DNA in the range 700 bp to 1 kb. Primers encompassing the genomic region to be probed contain different restriction sites on each primer (Table S2). Vectors were linearized for in vitro transcription to produce antisense probes. The antisense probe against m was made from a 930 bp BamHI fragment of the cDNA clone RE53556, obtained from the Drosophila Genomics Resource Center (stock no. 9261).

Potential SoxN binding sites were identified by entering target gene sequences from FlyBase (Gramates et al., 2017) into Find Individual Motif Occurrences (FIMO) (Grant et al., 2011). The SoxN motif was downloaded from the TRANSFAC database, generated from HT-SELEX data (Nitta et al., 2015). Matches were filtered out with a threshold of P<0.001.

We thank D. Stern, F. Payre, S. Russell, and the Bloomington and Vienna Stock Centers for fly stocks. We are grateful to members of the A.B. lab for comments on the manuscript, and to the dedicated people at FlyBase for their hard work. Special thanks to R. Gordan and A. George for help with computational analyses.