Distinct roles and requirements for Ras pathway signaling in visceral versus somatic muscle founder specification

Embryonic development requires that individual cells within a field acquire new, distinct fates. The Drosophila larval musculature has emerged as an exemplary system for investigating this process, revealing important insights into the acquisition of developmental competence, progressive determination of cell fate, and the integration of multiple signals at the transcriptional level. It has proven to be a particularly effective model for understanding how specific outcomes can be obtained from the activation of the receptor tyrosine kinase (RTK)/Ras/mitogen-activated protein kinase (MAPK) signaling cascade (Halfon et al., 2000), a highly pleiotropic pathway involved in numerous developmental processes and dysregulated in a wide set of developmental disorders and cancers (Imperial et al., 2017; Schlessinger, 2000; Tidyman and Rauen, 2009).

In the somatic (body wall) musculature, which has been studied most extensively, individual syncytial muscle fibers develop via the fusion of two cell types drawn from an initially undifferentiated pool of myoblasts within the stage 10-11 (mid-embryogenesis) mesoderm: a single ‘founder cell’ (FC; itself the product of the asymmetric division of a muscle ‘progenitor’) and one or more ‘fusion competent myoblasts’ (FCMs; Fig. 1) (for a review, see Dobi et al., 2015). Whereas FCMs are uncommitted, FCs are induced with specific identities. FCMs fuse with FCs, with the resulting syncytium maintaining the identity provided by the FC. FC specification is thus a crucial step in muscle development as the FC genetic program controls muscle size, orientation, expression of cell-surface proteins, etc.

Although multiple signaling pathways, including the Wg (Wnt) and Dpp (BMP) pathways, are integrated to specify particular muscle fates, the key event in muscle determination is MAPK activation via RTK/Ras pathway signaling (Buff et al., 1998; Carmena et al., 2002, 1998; Halfon et al., 2000). In the somatic mesoderm, the relevant receptors are the Drosophila EGF and FGF receptor homologs. Cells that are competent to respond to Ras/MAPK signaling are induced as an equivalence group and subsequently restricted by lateral inhibition (mediated by Notch-Delta signaling) to a single muscle progenitor.

These events have been studied in detail at the molecular level in the context of the muscle-identity gene even skipped (eve). A 300 bp transcriptional enhancer directly integrates the inductive Ras/MAPK signaling with a combination of additional signal-derived and tissue-specific transcription factors (TFs) to activate eve expression (Halfon et al., 2000). The Ras/MAPK effector is the ETS-domain TF Pointed (Pnt), which binds the enhancer at up to eight distinct sites (Boisclair Lachance et al., 2018; Halfon et al., 2000). In the absence of induction these sites are bound by the ETS-domain repressor Anterior open (Aop, also known as Yan) (Halfon et al., 2000; Webber et al., 2013; Boisclair Lachance et al., 2018). Recent evidence suggests that Pnt bound at these or other sites may also contribute to repression in the absence of MAPK activation (Webber et al., 2018). Importantly, experiments have shown that induction trumps repression: in the absence of both Pnt and Aop binding, there is no gene activation (Halfon et al., 2000; unpublished data). Ectopic activation of the Ras/MAPK pathway leads to ectopic FC formation in all competent cells, at the expense of FCMs; this has been demonstrated at the level of the receptor tyrosine kinases (activated EGFR and FGFR), of Ras (activated Ras) and of the effector (activated Pnt) (Carmena et al., 1998; Halfon et al., 2000).

We focus here on the circular visceral muscle fibers, which surround the midgut and develop from the trunk visceral mesoderm (for simplicity, we will refer to these simply as ‘visceral muscle’ and ‘visceral mesoderm’, respectively). These muscle fibers appear to develop similarly to the somatic muscles (Fig. 1), with the exception that they are only binucleate and it is unclear whether there is a ‘muscle progenitor’ cell specified prior to visceral FC specification (Martin et al., 2001). As with somatic FCs, visceral FC specification occurs following MAPK activation – here via the single signaling pair of the Anaplastic lymphoma kinase (Alk) receptor tyrosine kinase and its ligand Jelly belly (Jeb) – and, just as for the somatic musculature, ectopic activation of the Ras/MAPK pathway causes presumptive FCMs to be re-specified as FCs (Fig. 2D) (Englund et al., 2003; Lee et al., 2003; Weiss et al., 2001). However, the details of these events have not been established.

We now show that, despite the apparent similarities between somatic and visceral FC specification, fundamental differences exist with respect to the role of Ras/MAPK signaling in specifying the FC fate. Unlike the positive inductive role for MAPK activity in the somatic mesoderm, in the visceral mesoderm MAPK activity is instead required to relieve repression of FC fates, and the transcriptional effectors Pnt and Aop do not appear to play a significant role in this process. Moreover, MAPK activity can be dispensed with entirely in the absence of the FCM-specific TF Lame duck (Lmd) or when repressor-binding sites are mutated in an FC-specific enhancer for the mind bomb 2 (mib2) gene. Thus, unlike in the somatic mesoderm, Ras/MAPK signaling is not essential for visceral FC specification. Our results illustrate how similar-appearing developmental processes can result from different underlying molecular mechanisms and provide fresh insights into how unique responses to Ras pathway signaling are determined within similar cellular and developmental contexts.

In a previous screen for genes that respond differentially to different Ras pathway components, we observed that, despite responding to RTK and Ras activation, the FC gene mib2 is not regulated by the Ras effector Pnt in the visceral mesoderm (Leatherbarrow and Halfon, 2009). Expression of both mib2 RNA and a mib2-lacZ reporter gene driven by an FC-specific enhancer (mib2-FCenhancer) is normal in pnt null mutant embryos (Fig. 2A,C,E; Halfon et al., 2011), and expression of a constitutively active form of Pnt (Pnt^act) has no effect on expression of either the endogenous gene or the reporter (Fig. 2F, Fig. S1D; Leatherbarrow and Halfon, 2009). Similarly, mib2 expression in the visceral mesoderm is normal in embryos mutant for the ETS-domain repressor aop (yan) (Fig. 2G, Fig. S1E), and in response to expression of the constitutively repressing version yan^act (Fig. 2H, Fig. S1F; Halfon et al., 2011). This raised the question of whether this is a mib2-specific regulatory effect, or whether these two Ras effectors, which both play a significant role in somatic FC determination, are not required for visceral FC specification.

To test this, we assessed the expression of additional visceral FC and FCM markers in some or all of the pnt null, aop null, pnt^act and yan^act backgrounds. Expression of the somatic muscle-identity gene even skipped (eve) was used as a control (insets in Fig. 2), as its expression is reduced or expanded in response to pnt and aop loss and gain of function in a well-defined way (Halfon et al., 2000). Expression of the FC markers org-1, kirre [also known as dumbfounded (duf)] and RhoGAP15B all appear normal in a pnt^act background, whereas, as reported previously, expression of all three expands with pan-mesodermal expression of activated Ras (Ras^act) (Fig. 3, Fig. S1; Leatherbarrow and Halfon, 2009; Lee et al., 2003). Double labeling with antibodies to Biniou, a general visceral mesoderm marker (Zaffran et al., 2001), showed that the observed expansion is throughout the visceral mesoderm (Fig. 2D, Fig. 3D,E, insets). This suggests that non-FCs (i.e. FCMs) have been re-specified as FCs, rather than that there has merely been increased FC proliferation. Consistent with this, expression of the FCM marker Lmd behaves in the reciprocal fashion: Lmd expression decreases in the visceral mesoderm with Ras^act expression, but is unaffected in pnt null or pnt^act backgrounds (Fig. 3B,E,H; data not shown; Popichenko et al., 2013). Expression of both Org-1 and RhoGAP15B is also unchanged in the pnt null, aop null, and yan^act backgrounds (Fig. S1). Taken together, our results suggest that although Ras activity is sufficient to induce FC fates throughout the visceral mesoderm, neither pnt nor aop appear to play a significant role in this process.

Although mib2 expression in the visceral mesoderm is not dependent on either pnt or aop, the mib2_FCenhancer enhancer was identified in part based on the presence of ETS-type, putative Pnt binding sites (Philippakis et al., 2006). We showed previously that mutation of a set of seven ETS sequences in this enhancer caused an expansion of reporter gene expression driven by the mutated enhancer (Fig. 4G-I; Halfon et al., 2011). Like the expression observed with activation of the Ras pathway, the expanded reporter gene expression extends throughout the visceral myoblast population as marked by Bin expression (data not shown). Interestingly, reporter gene expression is stronger in the FCs than in the rest of the myoblasts (Fig. 4I, Fig. S2A-A″; data not shown). Activated Ras expression restores full-strength reporter activity similar to what is observed with the wild-type enhancer (Fig. 4J, Fig. S2B-B″; data not shown). However, as expected, the same disparity in reporter expression between FCs and non-FCs as seen in the wild-type background is seen with activated Pnt, which by itself does not lead to expanded mib2 expression (Fig. 4K, Fig. S2C-C″; data not shown).

The expanded reporter gene expression observed upon mutation of the ETS sequences suggested that rather than being required for positively activating mib2 expression – as expected based on analogy to the requirement for Ras pathway signaling mediated by ETS-family TFs in the somatic mesoderm (Halfon et al., 2000) – mib2 is repressed via TF binding at these sites. In order to understand better the nature of this repression, we decided to characterize the mib2 regulatory sequences more thoroughly.

Using sequence conservation with other Drosophila species as a guide, we first tested reporter gene activity with a truncated version of the mib2_FCenhancer containing a 5′ 120 bp deletion (Fig. 4A,B, Fig. S3). The deleted region includes one of the putative Pnt-binding sites previously mutated (‘site 1’; Fig. 4B, Fig. S3), as well as a non-canonical Pnt site suggested by protein-binding microarray experiments (Fig. S3; ‘siteN’; personal communication from Alan Michelson, NHLBI, Bethesda, MD, USA). The resulting mib2_FC626 enhancer has activity identical to the longer mib2_FCenhancer (Fig. 4C-F), responds to Ras^act and Pnt^act ectopic expression in the same manner (Fig. S1E,F; Leatherbarrow and Halfon, 2009), and shows a similar Ras^act-like expansion of reporter gene expression when the remaining six ETS-type sequences are mutated (construct mib2_FC626^ETS; Fig. 4G-I). In contrast, a more extensive 5′ 413 bp deletion (Fig. 4B, arrowhead; Fig. S3, gray arrow) leads to a complete loss of visceral mesoderm activity and only a limited residual expression elsewhere (data not shown). As the mib2_FC626 enhancer behaves in all aspects like the original mib2_FCenhancer, we used this shorter sequence as a template for further characterization of mib2 regulation.

We mutated the six remaining ETS-type sequences individually to determine which putative binding sites were responsible for the expanded reporter gene expression (as sites 5 and 6 are close together, we treated them initially as a single site, 5-6). Expanded reporter gene expression is observed only with the site 5-6 paired mutation (Fig. 4L, Fig. S4). We therefore further dissected this pair to test its component individual sites. Mutation of site 5 leads to expansion of mib2_FC626 enhancer activity throughout the visceral mesoderm (Fig. 4M), whereas site 6 by itself has a barely observable phenotype with expanded expression almost indistinguishable from background (Fig. S2G). The expanded expression observed with the single site 5 mutation is notably weaker than that observed with the paired site 5-6 mutation, which itself has weaker expression than the mib2_FC626^ETS six-site-mutated enhancer (compare Fig. 4I,L,M). Although site 5 therefore appears to be the most crucial site contributing to expanded mib2_FC626 expression, the stronger expression seen when multiple sites are mutated suggests that these other sites are functional as well and contribute to overall enhancer activity. Consistent with their more essential roles, site 5 is the most highly conserved of the six ETS sites in the mib2_FC626 sequence, followed by site 6 (Fig. S3B).

In addition to expanded visceral mesoderm expression, the mib2_FC626^ETS construct drives ectopic reporter gene expression in the midline of the ventral nerve cord (Fig. 4M-O; Halfon et al., 2011). This ectopic midline expression is also observed with the mib2_FC626^site7 construct (Fig. 4N), but not with any of the other single-site enhancer mutations. Similar ectopic mib2 expression is observed in aop null mutant embryos (Fig. 4R), suggesting that although Aop does not regulate mib2 in the visceral mesoderm, it may act via this site to repress mib2 in the nervous system.

In the somatic mesoderm, we showed previously that Ras pathway activity is absolutely required for FC gene expression; for example, in the absence of both pnt and aop expression the eve_MHE FC enhancer is inactive, and mutation of the common ETS-type Pnt- and Aop-binding sites eliminates enhancer activity (Halfon et al., 2000). However, the de-repression of the mib2 visceral FC enhancer seen with ETS site mutation suggested that Ras activity, normally not present in the FCM population into which mib2 reporter gene expression expands, might be dispensable when the mib2 enhancer is de-repressed. To test this, we analyzed the activity of the wild type and mutated mib2_FC626 enhancers in a jeb null background, which lacks Ras signaling in the visceral mesoderm. Whereas the wild-type mib2_FC626 enhancer is inactive in the visceral mesoderm of jeb null embryos (Fig. 5A,C), the mib2_FC626^ETS mutated enhancer is expressed throughout the visceral mesoderm just as in a wild-type background (Fig. 5B,D, arrows). Staining for the activated, di-phosphorylated form of MAPK confirmed that no signaling was present in the jeb null background (Fig. 5F). Unlike in the somatic mesoderm, therefore, Ras signaling is not required for expression of a visceral FC gene in the absence of ETS site-mediated repression.

The expanded expression of an FC gene throughout the visceral mesoderm we observed in the case of the de-repressed mib2_FC626 enhancer is reminiscent of what has been observed upon loss of function of the FCM transcriptional activator Lmd. Popichenko et al. (2013) showed that in lmd null embryos, FC markers such as org-1, hand and kirre expand throughout the visceral mesoderm, with reciprocal loss of FCM genes such as Vrp1. This resembles the phenotype observed with Ras activation. In a similar fashion, lmd mutation leads to the conversion of a small subset of FCMs to adult muscle precursor and pericardial cells (Sellin et al., 2009). However, these phenotypes are in sharp contrast to what is observed for the bulk of the somatic mesoderm, where lmd loss of function has no effect on FC specification (Duan et al., 2001; Ruiz-Gomez et al., 2002). Interestingly, the FCM-specific gene sticks and stones (sns) remains expressed in the lmd visceral mesoderm, suggesting that the observed FCM→FC conversion is incomplete (Estrada et al., 2006; Ruiz-Gomez et al., 2002). Given our results with the mib2 enhancer, we wondered whether lmd loss of function-induced FCM→FC re-specification could also occur in the absence of Ras pathway activity. Therefore, we tested jeb;lmd double-mutant embryos for a range of FC markers including org-1, mib2 and RhoGAP15B, which are expressed in all FCs, as well as Connectin (Con) and wingless (wg), which are expressed in only a subset of FCs (Bilder and Scott, 1998). In all cases, lmd was fully epistatic to jeb: whereas in jeb null embryos no FC markers are expressed, expression in jeb;lmd embryos consistently resembles lmd alone, in most cases with expanded FC expression (Fig. 6). Surprisingly, mib2 and RhoGAP15B, which expand throughout the visceral mesoderm both with Ras activation and upon mutation of the mib2 FC enhancer, do not appear to have expanded expression in the lmd background when assayed by in situ hybridization (Fig. 6B,F). This may be indicative of an incomplete conversion of FCMs to FCs, consistent with the maintenance of sns expression in the FCM region reported previously. Likewise, Wg expression also only expands to a few additional cells and not throughout the entire width of the anterior PS8 visceral mesoderm (Fig. 6S). On the other hand, we found that the mib2_FC626 reporter construct does show fully expanded expression in lmd embryos (Fig. 6B′). As expression driven by the mib2_FC626 enhancer appears identical to endogenous mib2 expression in all other contexts we have examined, it may be that the lack of expanded mib2 expression observed in lmd embryos simply represents a failure of detection, given that, similar to what we saw with the reporter construct in other backgrounds, reporter gene expression in the FCM region is weaker than that in the native FC region (Fig. S2D). Consistent with this interpretation, we see a modest widening of mib2 expression in the jeb;lmd embryos (Fig. 6D). Importantly, regardless of the exact degree of expanded expression caused by lmd loss of function, FC expression of all tested genes is clearly rescued in the jeb;lmd background (Fig. 6D,H,L,P), demonstrating the ability for Ras signaling to be bypassed in the absence of lmd expression.

To ensure that the rescue of FC specification observed in jeb;lmd mutants is not the result of a cryptic Ras signaling pathway activated by loss of lmd, we checked for the presence of activated MAPK in the double-mutant embryos. As expected when jeb is absent, no activated MAPK is observed, regardless of presence or absence of lmd (Fig. 6X,Y).

Both somatic and visceral muscle development require as an initial step the specification of individual muscle founder cells from within the general myoblast pool. Superficially, the process appears alike for both tissues: FCs fail to form in the absence of RTK/Ras/MAPK signaling, and ectopic activation of the Ras pathway causes FCMs to be re-specified as FCs. A striking aspect of our current results is that these seemingly similar events are brought about in a mechanistically opposite fashion in the somatic versus visceral mesoderms. Our work therefore serves to underscore how common developmental outcomes can derive from dramatically different gene regulatory mechanisms.

In the somatic mesoderm, it has been well-established that positive induction via Ras/MAPK signaling is essential for specifying FC fates (Buff et al., 1998; Carmena et al., 2002, 1998; Halfon et al., 2000). In the visceral mesoderm, however, repression has primacy over induction. We demonstrate here that Ras/MAPK signaling acts in presumptive FCs to relieve repression of the FC fate, while Popichenko et al. (2013) previously established that it serves to prevent activation of FCM genes. The primary activator of FCM genes is Lmd, which prior to FC specification is expressed in all visceral myoblasts (Ruiz-Gomez et al., 2002; Popichenko et al., 2013). Ras/MAPK signaling in FCs causes phosphorylation of Lmd followed by its export from the nucleus and its degradation, preventing it from activating FCM-specific genes such as Vrp1 (Popichenko et al., 2013). What happens at FC gene loci, however, had not previously been determined. Our results with the mib2 enhancer demonstrate that FC genes can be activated in the absence of Ras/MAPK signaling, through loss of repressor binding at the enhancer.

The simplest model, taking into account our results and those of Popichenko et al. (2013), would be for Lmd to function as both the FC gene repressor and the FCM gene activator; loss of Lmd binding following Ras/MAPK signaling would thus simultaneously de-repress FC genes while halting activation of FCM genes. Although Lmd is typically viewed as an activator, some evidence suggests that it may also be capable of transcriptional repression (Cunha et al., 2010). However, chromatin immunoprecipitation studies have repeatedly failed to detect appreciable Lmd binding in the mib2 enhancer region (Busser et al., 2012; Cunha et al., 2010), and the mib2 enhancer lacks good candidate Lmd-binding sites – particularly in the crucial site 5-6 region (M.S.H., data not shown) – even when surveyed using a range of binding motifs derived from multiple sources (Busser et al., 2012; Nitta et al., 2015; Zhu et al., 2011).

We therefore favor a basic model in which Lmd serves as an activator of the FC gene repressor, such that loss of Lmd leads to loss of repressor activity in FCs and subsequent expression of FC genes (Fig. 7). In wild-type embryos, the main role of MAPK signaling is thus to cause phosphorylation and degradation of Lmd, whereas in lmd mutant embryos, MAPK signaling becomes irrelevant as Lmd is already absent. The repressor could also be a direct target of MAPK, leading to its rapid displacement upon the onset of MAPK signaling (e.g. similar to what happens with Aop at other loci; Rebay and Rubin, 1995). This would be consistent with the rapid time course of FC fate specification following MAPK activation. We surmise that there are also additional, still unknown FC gene activators, activity of which may or may not be MAPK dependent. Tests of these various refinements to the basic model will require identification of the FC gene repressor.

In the somatic mesoderm, the repressor Tramtrack69 (Ttk69; Ttk) appears to play a role as an lmd-dependent FC gene repressor similar to what we posit here for visceral FC fate repression. ttk69 expression is activated downstream of lmd in FCMs, where it represses the transcription of FC genes (Ciglar et al., 2014). In FCs, Ttk69 is likely post-translationally degraded in a Ras-dependent manner (Ciglar et al., 2014; Li et al., 1997), relieving repression of FC genes concurrent with Ras-dependent relief of Aop-mediated repression and induction via Pnt and/or other activators. However, different mechanisms appear to be at work in the visceral mesoderm. Although ttk69 mutants do display some altered visceral mesoderm gene expression (Ciglar et al., 2014), visceral FCs appear to be correctly specified and FC-specific genes such as mib2 are expressed in the appropriate pattern, without expansion into the FCM field (Ciglar et al., 2014; S.E.P., unpublished observations).

Given the de-repression phenotypes observed on mutation of the ETS sites in the mib2 enhancer, we favor the likelihood that the relevant repressor is a member of the ETS TF family, either by itself or working redundantly with Pnt and/or Aop. Several other ETS-domain proteins exist in Drosophila (Chen et al., 1992), although their expression patterns and mutant phenotypes are for the most part not well defined. A role for additional ETS proteins was also previously suggested for the dorsal somatic mesoderm, where pnt loss of function leads to a partial loss of Eve-expressing FCs, but mutation of ETS binding sites completely eliminates expression driven by the eve_MHE enhancer (Halfon et al., 2000). Although our data argue against an absolute requirement for either Pnt or Aop, we cannot rule out a more limited contribution from these factors, and their requirement for expansion of FC fates (as opposed to generation of normal FC fates) has not been fully tested. Indeed, chromatin immunoprecipitation experiments indicate that Pnt can bind to the mib2 enhancer region, although it is not known in what cell types (Webber et al., 2018), and aop mutants show an effect on mib2 expression in the ventral midline (Fig. 4R).

Although there is considerable evidence demonstrating the requirement for RTK/Ras/MAPK signaling for somatic FC specification, the molecular details on the mechanisms governing MAPK-dependent activation and repression come mainly from studies of a single transcriptional enhancer, eve_MHE (Boisclair Lachance et al., 2018; Halfon et al., 2000; Webber et al., 2018, 2013). It is clear that ETS factor-dependent activation is essential for the activity of this enhancer, as mutation of the major ETS binding sites renders the enhancer non-functional (Halfon et al., 2000). However, recent studies suggest that instead of an abrupt switch between activation and repression due to mutually exclusive enhancer occupancy by Pnt and Aop, there is a more subtle balance between these TFs and their binding to the multiple high- and low-affinity ETS binding sites found in the enhancer (Boisclair Lachance et al., 2018; Webber et al., 2018). Other somatic FC enhancers have not been rigorously tested with respect to ETS-family binding, and it may be that the trade-off between activation and repression differs among them. This would help to explain the results of Buff et al. (1998), who demonstrated that different FCs are specified at different levels of RTK/Ras signaling. One way to achieve such differential sensitivity would be through a mixture of activating versus repressing ETS TFs competing for binding at a range of high- and low-affinity sites. Such a mechanism could provide for exquisitely fine-tuned responsiveness to Ras/MAPK signaling, making this an appealing possibility.

In this vein, we note that our detailed molecular insights for visceral FCs again come mainly from the study of a single enhancer, mib2_FC626. Here, elimination of the major ETS binding sites leads to increased activity, the opposite of the situation with the somatic eve_MHE enhancer. However, preliminary analysis of other visceral FC enhancers suggests that eliminating ETS binding sites can lead to loss of enhancer function, more similar to what is seen in the somatic musculature (Y.Z., unpublished observations). Thus, although our data from the mib2_FC626 enhancer as well as from analysis of lmd mutants clearly establishes de-repression rather than induction as the major role for Ras pathway signaling during visceral FC specification, it may be that the molecular basis for how this signaling is modulated by ETS-family TFs at the enhancer level is complex and balanced individually at each FC gene enhancer. Taken together, these plus other recent results (Boisclair Lachance et al., 2018; Webber et al., 2018) point to an elaborate interplay between Ras signaling, ETS TFs, and subtly tuned binding sites, and highlight the need for detailed molecular studies of a more comprehensive set of both somatic and visceral FC enhancers.

Oregon-R was used as the wild type. Mutant stocks are described in FlyBase (Gramates et al., 2017) and include pnt^Δ88, aop¹, lmd¹ and jeb⁵⁷⁶ (Weiss et al., 2001). mib2_FCenhancer-lacZ has been described by Philippakis et al. (2006) and the rp298-lacZ (FlyBase: kirre^rp298-PZ) line was used to assess kirre (duf) expression (Nose et al., 1998). Ectopic expression was achieved using the Gal4-UAS system (Brand and Perrimon, 1993) and used lines Twi-Gal4 (FlyBase: P{Gal4-twi.G}) (Greig and Akam, 1993), UAS-Ras1^Act (Carmena et al., 1998), UAS-PntP2VP16 (Halfon et al., 2000) and UAS-yan^Act (Rebay and Rubin, 1995). Mutant lines were rebalanced over lacZ-marked balancers to allow for genotyping of embryos.

For all analyses, a minimum of ten embryos were analyzed in detail, with representative images chosen for publication. All reported phenotypes were fully penetrant. Antibody staining was performed using standard Drosophila methods (Müller, 2008). The following primary antibodies were used: mouse anti-β-galactosidase (Promega, Z3783; 1:500), rabbit anti-GFP (Abcam, ab290; 1:10,000), rabbit anti-Bin (gift of Eileen Furlong, EMBL, Heidelberg, Germany; 1:300), rabbit anti-Lmd (gift of Hanh Nguyen, Friedrich-Alexander-University of Erlangen-Nürnberg, Erlangen, Germany; 1:1000), rat anti-Org-1 (gift of Manfred Frasch, Friedrich-Alexander-University of Erlangen-Nürnberg; 1:250), mouse anti-activated MAPK (diphosphorylated ERK1&2, Sigma, M9692; 1:250; fixed in 8% paraformaldehyde), mouse anti-Wg (4D4, Developmental Studies Hybridoma Bank; 1:100), mouse anti-Con (C1.427, Developmental Studies Hybridoma Bank; 1:250). ABC kit (Vector Laboratories) was used for immunohistochemical staining. Differential interference contrast (DIC) microscopy was performed using a Zeiss Axioskop 2 microscope and Openlab (PerkinElmer) software for image capture. The following secondary antibodies were used for fluorescent staining: anti-mouse Alexa Fluor 488 (Molecular Probes, A11029; 1:250), anti-mouse Alexa Fluor 633 (Molecular Probes, A21052; 1:500), anti-rabbit Alexa Fluor 488 (Molecular Probes, A11034; 1:250), anti-rabbit Alexa Fluor 633 (Molecular Probes, A21071; 1:500), anti-rat Alexa Fluor 633 (Molecular Probes, A21094; 1:500). Fluorescent staining was visualized by confocal microscopy using a Leica SP2 confocal microscope. In situ hybridization for detection of mib2 and RhoGAP15B transcripts was as previously described (Leatherbarrow and Halfon, 2009). Color and brightness of acquired images were adjusted using Adobe Photoshop.

The mib2_FC626 reporter was made by digesting the mib2_FCenhancer sequence with BamHI and cloning the resulting 3′ portion into either pattBnucGFPh (for GFP reporter expression) or pattB-nucLacZ (for lacZ reporter expression). Mutagenesis of the mib2 enhancer was performed by overlap-extension PCR (Ho et al., 1989), and sequences were cloned into pattBnucGFPh. Mutated sequences are shown in Fig. S3 (primer sequences available in Table S1). Transgenic flies were generated by Genetic Services (Cambridge, MA, USA) using phiC31-transgenesis and the attP2 landing site. The availability of both GFP and lacZ reporter versions of the wild-type mib2_FC626 enhancer allowed for easy comparison of wild-type and mutant enhancers by crossing a lacZ reporter line to a GFP reporter line and double labeling for both lacZ and GFP.

We thank Manfred Frasch, Eileen Furlong, Alan Michelson, Hanh Nguyen, the Bloomington Drosophila Stock Center (NIH P40OD018537) and the Developmental Studies Hybridoma Bank (created by the NICHD of the NIH and maintained at The University of Iowa, Department of Biology, Iowa City, IA 52242) for fly stocks and antibodies. Elizabeth Brennan, Sam Hasenauer, Qiyun Zhu and Jack Leatherbarrow helped with experiments. Steve Gisselbrecht and Michael Buck provided critical comments on the manuscript.