Sequence environment of BMP-dependent activating elements controls transcriptional responses to Dpp signaling in Drosophila

Bone morphogenetic proteins (BMPs) control a vast number of developmental and homeostatic processes (Wu and Hill, 2009). In canonical BMP signaling, BMP ligands bind and activate receptor complexes at the cell membrane, which in turn phosphorylate receptor-associated Smads (R-Smads) (Shi and Massagué, 2003). Subsequently, phosphorylated R-Smads associate with the common-Smad (co-Smad, Smad4 in mammals) and the Smad complex accumulates in the nucleus to bind DNA directly and regulate transcription of target genes (Fig. 1A). In Drosophila, BMP-dependent gene regulation has been analyzed in multiple contexts of fly development, including cases of graded (morphogen) BMP signaling during early embryonic development and larval wing development (Affolter and Basler, 2007; Bier and De Robertis, 2015; Upadhyay et al., 2017). In both cases, a spatial gradient of the Drosophila BMP Decapentaplegic (Dpp) generates a gradient of phosphorylated Mad (Mad, the Drosophila R-Smad), which then activates target gene transcription in a threshold-dependent manner (Ashe and Briscoe, 2006; Hamaratoglu et al., 2014). Besides direct pMad input, proper activation of BMP/pMad targets requires the transcriptional repressor Brinker (Brk), which is coupled to BMP signaling through two key properties: first, Brk distributes in a pattern that is inverse to the gradient of pMad, and, second, Brk directly represses Dpp target genes (Ashe et al., 2000; Campbell and Tomlinson, 1999; Jaźwińska et al., 1999a,b; Minami et al., 1999). Thus, Dpp-target genes integrate input from two opposing gradients: activatory input from pMad and repressive input from Brk. Their differential sensitivity to these two cues define their spatial extent of activation within the morphogen field (reviewed by Affolter and Basler, 2007; Hamaratoglu et al., 2014).

The inverse relation of the pMad and Brk distribution is not restricted to the context of graded Dpp signaling but is evident in most instances of BMP signaling during fly development. In most cases, Dpp signaling directly accounts for the distribution of Brk by negatively regulating its transcription. Dpp-dependent repression of brk transcription requires the repressor Schnurri (Shn) and short DNA sequences, the silencer elements (SEs), present in the regulatory region of brk (Charbonnier et al., 2015; Marty et al., 2000; Müller et al., 2003; Pyrowolakis et al., 2004; Torres-Vazquez et al., 2001; Yao et al., 2008). Upon signal activation, Smad trimers consisting of two pMad and one Medea molecule bind directly to the SE, which comprises three minimal Smad-binding sites (GNC; N, any nucleotide) organized in the consensus GNCGNC(N)₅GTCT (minimal Smad-binding sites in bold; Fig. 1B) (Gao et al., 2005; Pyrowolakis et al., 2004). Within this sequence, the two pMad molecules bind the GNC motifs of the GNCGNC box, whereas Medea binds to the GTCT motif. The SE-bound Smad complex can then recruit nuclear Shn, which mediates brk repression. Binding of Shn to the SE/Smad complex seems not to require direct Shn-DNA contact but rather a very specific conformation of the SE-bound Smad complex, which, in turn, depends on determinants within the SE. Specifically, Shn can only dock to the complex when the spacing between the pMad and Medea sites is precisely five nucleotides (independent of the nature of the nucleotides) and when the Medea-binding block contains a T at the last position (GTCT). Any deviation from these two features results in an SE that is fully able to interact with a Smad trimer but cannot recruit Shn in vitro and is, consequently, fully inactive in transcriptional gene repression in vivo. Thus a simple, yet stringent arrangement of Smad-binding sites in the SE implements BMP-dependent repression of an expanding number of BMP targets, including brk (Beira et al., 2014; Crocker and Erives, 2013; Esteves et al., 2014; Vuilleumier et al., 2010; Walsh and Carroll, 2007).

Gene activation by Smad signaling seems to be more complex with BMP-dependent enhancers containing a variable number of Smad- and Brk-binding sites, as well as binding sites for transcription factors that synergize with BMP signaling to achieve robust and tissue-specific target gene expression (Barrio and de Celis, 2004; Liang et al., 2012; Markstein et al., 2002; Rushlow et al., 2001; Saller and Bienz, 2001; Winter and Campbell, 2004; Zhang et al., 2001). Brk- and Smad-binding sites are either dispersed on such enhancers, or can form clusters. One extreme case of the latter is a motif termed an activating element (AE), originally identified in the regulatory region of Daughters against Dpp (Dad) but later identified in a number of BMP-responsive enhancers (Szuperák et al., 2011; Vuilleumier et al., 2018; Weiss et al., 2010). The motif, GGCGYC(N)₅GTCV (Smad-binding sites in bold; V: G, A or C; Fig. 1B), is very similar to the SE; however, it lacks one of the determinants for Shn recruitment (a T at the last position). Instead, the pMad-binding block (GGCGYC) corresponds also to a Brk-binding motif (GGCGYY) explaining the negative impact of Brk on Dad expression. Besides such core determinants for Brinker and Smad recruitment, it is not clear whether the core AE motif contains additional features. The limited available data – mostly derived from biochemical and cell culture assays – suggest that AEs may come in many variants differing from each other in the linker length and nucleotide environment of the core consensus motif (Esteves et al., 2014; Gao and Laughon, 2007). Although this might indicate a flexibility of the element towards the recruitment of the Smads and Brk, it is equally conceivable that, similar to the SE, AEs might contain sequence determinants that facilitate AE-bound Smads and/or Brk to recruit partners impacting on the element's output. Potential partners may include transcriptional co-activators and/or co-repressors, factors affecting the opposing Brk and Smad inputs in transcriptional output, or even proteins conveying tissue specificity to the AE.

Here, we address this question by studying the regulation of the gene wishful thinking (wit). We demonstrate that BMP signaling activates wit transcription in both the larval wing imaginal disc and the follicular epithelium; however, and in sharp contrast to Dad, BMP responsiveness in the two tissues is mediated by distinct cis-regulatory modules (CRMs). In addition, the two identified CRMs are differentially sensitive to Smad and Brk inputs and are equipped with AE-like motifs that differ from the prototypic AE and from each other. Using a combination of genetic mosaic analysis and reporter assays with chimeric CRMs, we demonstrate that the diversified AEs neither mediate tissue specificity nor do they account for the observed differences in the responsiveness of wit to Brk and Smad input in the two epithelia. Our data are consistent with the sequences of AEs being rather flexible and monotonically mediating BMP responses, with qualitative and quantitative aspects of such responses depending on BMP-independent, activatory sequences within their cognate CRMs.

Wit is a Drosophila BMP type II receptor predominantly expressed in neural cells at neuromuscular junctions to control synaptic size and function as well as in a set of neurosecretory cells to regulate expression of neuropeptide genes (Aberle et al., 2002; Allan et al., 2003; Marqués et al., 2003; Marqués et al., 2002; McCabe et al., 2003; Veverytsa and Allan, 2011; Zheng et al., 2003). Wit is also required for the formation of the anterior pMad gradient in the ovarian follicle cells (FCs; Fig. 1C) and for proper eggshell formation (Marmion et al., 2013; Pyrowolakis et al., 2017). In this context, the transcription of wit is activated by BMP signaling itself in an anterior, wedged-shaped stripe of oocyte-associated FCs (Fig. 1C,D,D′). Additionally, wit is expressed in the developing wing imaginal disc epithelium, although there is no evidence for a contribution of the receptor in transmitting BMP signals in this tissue or, generally, in wing development (Marqués et al., 2002). In a recent transcriptional profiling experiment, we have identified wit as a target of Dpp in the developing wing (Alexander Springhorn, M.J. and G.P., unpublished data), prompting us to re-evaluate its expression in this tissue. Wit is present in medial regions of the wing disc and absent from brk-expressing lateral cells, suggesting positive regulation by the BMP/pMad signaling gradient (Fig. 1E-F′). Indeed, reduction of Dpp signaling by clonal expression of Dad resulted in cell-autonomous loss of Wit in both the follicular epithelium and the wing imaginal disc (Fig. 2A-B″). In reverse, clonal activation of Dpp signaling resulted in strong, ectopic Wit expression (Fig. 2C-D″). Thus, similarly to the follicular epithelium, Wit expression is under positive control of BMP signaling in the wing epithelium. This regulatory relationship is reminiscent of Dad, which also encodes a pathway-inherent component and is regulated by BMP signaling in multiple tissues (Tsuneizumi et al., 1997; Weiss et al., 2010).

Previous work on Dad transcription has identified a BMP-responsive CRM (dad13) that accounts for Dad expression in all tissues tested so far, including the follicular epithelium and the wing imaginal disc (Weiss et al., 2010). To understand whether the molecular principles underlying Dad regulation also apply for wit, we investigated the cis elements accounting for its Dpp-responsiveness in both tissues. Our previous work has identified a ∼1 kb fragment, termed witZ, to be expressed in a wit-like pattern in the follicular epithelium (Marmion et al., 2013). A sub-fragment of witZ, witF (F for follicle cells), comprising ∼400 bp of witZ, is still capable of recapitulating all aspects of wit expression in follicle cells and is considered here as the minimal CRM (Fig. 3A,B). However, neither witF nor its parental witZ activate reporter expression in the wing imaginal disc (Fig. 3B′, inset), suggesting that distinct CRMs account for expression of wit in the wing. Testing a collection of fragments tiling the wit genomic locus by reporter assays in transgenic flies, revealed that the pattern of wit in the wing is recapitulated by a ∼600 bp long, intronic fragment hitherto referred to as witW (W for wing disc; Fig. 3A,C). In addition, witW was found to be completely inactive in the follicular epithelium (Fig. 3C′, inset). Using the same experimental set-up as for endogenous Wit, we could demonstrate that the activities of both witF and witW strictly depend on BMP input (Fig. S1). Thus, and in contrast to Dad, Dpp-dependent expression of wit in the eggshell and the wing is achieved by distinct regulatory modules.

Transcriptional targets of canonical BMP signaling in Drosophila can be activated directly by Smads, de-repressed by Smad-dependent repression of Brk, or by a combination of both (see Introduction and Fig. 1A). To assess the relative contributions of Smads and Brk in the regulation of wit in the wing and follicular epithelium, we compared wit reporter expression in Mad and brk mutant clones. As expected, clonal loss of Mad resulted in a complete loss of reporter expression within the cells of the clone in both epithelia (Fig. 4A-B″). In the egg chamber, loss of brk in posterior FCs resulted in strong, cell-autonomous activation of witF, demonstrating a crucial contribution of the Brk-dependent branch to the BMP-mediated activation of wit (Fig. 4C-C″). Reporter expression levels were equally high in anterior and posterior clones and approximated the levels of the reporter at the endogenous stripe. Given the steep and restricted anterior pMad gradient in the egg chamber, it is unlikely that Smad complexes provide activatory input other than downregulating brk expression. Nevertheless, we directly addressed the epistatic relationship of Smads and Brk by analyzing reporter activity in clones that simultaneously lack Mad and Brk. Posterior Mad/brk double-mutant clones displayed strong upregulation of witF, demonstrating that Smads are not required for the ectopic reporter expression observed in brk mutants (Fig. 4E-E″). Importantly, Mad/brk double-mutant clones cutting across the anterior endogenous stripe of wit expression did not affect witF. Thus, the loss of witF activity observed in single Mad mutants can be completely reversed by genetic removal of brk, demonstrating that all effects of BMP signaling on wit expression in the FC epithelium can be assigned to Brk.

Applying the same experiments for the wing-specific witW, uncovered a different behavior. Lateral clones lacking brk displayed an upregulation of the witW reporter, albeit at levels lower than in the endogenous, medial expression domain (Fig. 4D-D″). At the same time, the reporter activity in medial Mad mutant clones could only be partially restored by the simultaneous removal of Brk (Fig. 4F-F″). Thus, in contrast to the wit regulation in the follicular epithelium but is similar to the regulation of Dad in the wing; activation of witW requires a dual input by activated Smads: de-repression (repression of Brk) and additional, potentially direct, activatory input.

Importantly, all the above regulatory interactions, deduced from mosaic analyses with wit reporters as a read-out, could be confirmed for endogenous Wit expression. Specifically, clonal analysis confirmed that Mad activates whereas Brk represses Wit in both the follicle cells (Fig. S2A-A″,C-C″) and the wing (Fig. S2B-B″,D-D″). As with the witF reporter, epistatic analyses using Mad/brk double-mutant clones demonstrate that the role of Mad in follicular Wit expression is limited to the repression of Brk (Fig. S2E-E″). In contrast, and consistent with the behavior of witW, both Brk-mediated and Brk-independent Mad input is required for Wit expression in the wing disc (Fig. S2F-F″).

Our data so far establish two deviations in the transcriptional regulation of wit. First, in striking contrast to Dad, which utilizes the same CRM (dad13) for BMP-dependent activation in multiple tissues, independent CRMs account for BMP-dependent regulation of wit in different tissues. Second, the two CRMs of wit differentially integrate the activity of the transcription factors of the pathway (Smad and Brk). In order to understand the molecular basis for these differences, we analyzed cis requirements for wit expression. Particularly, we investigated whether the identified CRMs contain BMP-dependent response elements and whether such elements might also implement tissue specificity and/or differential sensitivity to the transcription factors of the pathway. The ‘canonical’ AE, first identified in Dad and subsequently shown to impose BMP responsiveness to a number of enhancers, corresponds to the consensus GGCGYCNNNNNGTCV (where N indicates any nucleotide, Y indicates C or T, and V indicates A, G or C; see Introduction). Whereas neither witF nor witW comprise such AEs, we did note that both fragments contain a highly conserved single cluster of Brk/Mad- and Med-binding sites separated by a variable number of nucleotides (Fig. 5A, Fig. S3). To test directly for a potential contribution in the expression of wit, we introduced deletions in witF and witW that completely remove these elements and tested for reporter activity in transgenic flies. Removing the potential AE in witF (witF^ΔAE) resulted in uniform reporter expression throughout the follicular sheet (Fig. 5B,C). Similarly, the same manipulation in witW (witW^ΔAE) resulted in reporter expression throughout the wing disc epithelium, but at levels that were lower compared with the parental witW (Fig. 5G,H). Importantly, in both cases the introduced deletions resulted in a complete loss of BMP responsiveness: neither brk nor brk/Mad mutant clones had any effect on witF^ΔAE or witW^ΔAE activity (Fig. S4). We next tested whether the differential requirement for Smads and Brk can be matched to sequence requirements within the identified motifs. Our genetic mosaic analysis predicts that only the Brk sites but not Smad-binding motifs within the AE of witF are relevant for reporter activity. Indeed, inactivating putative Med-binding sites (witF^Δmed) had no effect on reporter activity (Fig. 5D). However, converting the Brk/Mad hybrid motif into a Mad-only binding site (witF^Δbrk) resulted in the same strong upregulation of reporter expression as seen with witF^ΔAE (Fig. 5C,E) or with a construct lacking both Brk- and Mad-binding sites (witF^Δbrk/mad; Fig. 5F). In the wing imaginal disc, and consistent with inputs from both transcription factors, inactivation of both Brk- and Mad-binding sites (witW^Δbrk/mad) resulted in a lateral expansion of reporter activity and an overall reduction in expression levels similar to witW^ΔAE (Fig. 5H,I). Medial expression levels, but not the medial restriction, were restored by reinstating Smad input (witW^Δbrk; Fig. 5J). The results of the cis analyses are in full agreement with the findings of the genetic mosaic analyses on the different effects of Smads and Brk on witF and witW. We conclude that BMP responsiveness of the CRMs is mediated by the identified AE-like sequences, which, however, differentially integrate the activities of the transcription factors of the pathway, the activated Smad complex and Brk.

The sequences of the AEs of witF and witW deviate both from the original AE consensus motif and from each other. We sought to investigate whether such differences account for the observed differences in CRM behavior. Removal of the AEs exposed the existence of activators that produce spatially uniform, yet tissue-restricted, expression patterns (see above). This already indicates that sequences other than the AEs are essential for tissue specificity. Nevertheless, we directly tested for a contribution of the AEs to tissue-specific expression of their cognate CRM by generating transgenic reporters in which the two AEs where mutually exchanged (witF^WAE and witW^FAE) (Fig. 5K,L). These reporters were found to be exclusively active in the epithelium defined by the CRM backbones and not the AEs. Thus, sequences other than the AEs dictate tissue specificity, whereas the function of the AEs is restricted to integrating BMP input. Consistent with this notion, replacing the AEs in both CRMs with an SE (see Introduction), resulted in expression patterns that are inverse to the parental reporter expression (i.e. no expression in cells with high pMad levels and high expression in brk-positive cells), without affecting tissue specificity (Fig. 5M,N).

We next addressed whether sequence differences within the AEs account for the differences in Brk and Smad responsiveness of the two CRMs. In the follicle cells, the hybrid construct witF^WAE was active in an anterior, wedge-shaped pattern, indicative of Brk-repression (see above and Fig. 5C). As expected, the expression was lost in Mad mutant clones (Fig. 6A-A″). Additionally, both brk mutant and brk/Mad double-mutant clones posterior to the endogenous stripe resulted in strong ectopic reporter expression (Fig. 6B-C″). At the same time, brk/Mad double-mutant clones overlapping the endogenous expression domain had no effect on reporter expression (Fig. 6C-C″). Thus, although the AE of witW integrates Brk and Smad input in the wing, it is only responsive to Brk when assayed in the context of witF in follicle cells. We observed a similar behavior of the witW AE in the context of witF. BMP-dependent activation of witW was found to be exclusively mediated by the indirect branch of the pathway (activation through repression of Brk). If sequence constrains within the AE of witF prohibit a direct input from the activated Smad complex, then witW^FAE should transform into a ‘Brk-only’-responsive CRM. However, when compared with the low and rather uniform expression of witW^ΔAE, expression of witW^FAE was not only lost in lateral cells but also increased in medial cells, consistent with both lateral repression by Brk and medial activation by Smads, respectively (see above and Fig. 5L). Indeed, witW^FAE displayed the same responses as witW: although witW^FAE was lost in Mad mutant clones (Fig. 6D-D″) and ectopically active in lateral brk mutant clones (Fig. 6E-E″), the simultaneous removal of Mad and Brk in medial clones could not reinstate peak levels of reporter activity (Fig. 6F-F″), indicating a direct role of Smads in reporter activation. Thus, the AE of witF, which responds only to Brk in its native context, integrates both Brk-dependent and Brk-independent Smad inputs when assayed in the witW environment.

The above results indicate that it is not the sequence of the AE but rather its context that dictates which branch of the BMP signaling pathway will be utilized for activation. In the simplest scenario, AEs, although tentatively able to integrate both Smad and Brk input, cannot do so in follicle cells. For instance, Smads might not be able to activate transcription directly because essential co-activators are not available in this tissue and, consequently, BMP-mediated activation of witF is restricted to the Brk-dependent branch. Indeed, limited co-factor availability has been elegantly demonstrated for Brk, which contains interaction motifs for multiple co-repressors allowing it to retain activity in tissues that lack its main partner Groucho (Upadhyai and Campbell, 2013). To address this possibility, we studied the activation of Dad, which requires direct and indirect Smad input in the wing for activation. Expression of Dad in the follicular epithelium has been studied using an enhancer trap and is suggested to depend on Smad activity but not on Brk removal (Chen and Schüpbach, 2006). Indeed, Dad reporters, including a ∼400-bp-long subfragment of dad13, dad13A (considered here as the minimal CRM of Dad), were found to be active in an anterior stripe that, unlike witF, did not appear wedge-shaped but rather coincided with the pMad stripe. In addition, clones lacking Brk had only minimal effects on reporter activity, whereas anterior clones lacking Mad or both Mad and Brk resulted in a complete or almost complete, respectively, loss of reporter expression (Fig. S5A-C″). These findings demonstrate that Smads are fully capable of activating targets independently of Brk in follicle cells and exclude the absence of the activatory branch as an explanation for the Brk-only responses of witF and witF^WAE. Following on from these findings, we addressed the behavior of the AE of Dad when the AE was replaced with witF. Notably, the chimeric witF^dadAE reporter displayed the same wedge-shaped anterior expression stripe as witF (Fig. S5D-F) and was strictly responsive to Brk but not to Smads for activation as judged by clonal epistatic analysis (Fig. S5G-I″). Thus, the prototypic AE of Dad, which within Dad primarily integrates direct Smad activation, is converted to a ‘Brk-only’ element in the context of the basal witF. Lastly, we tested the reverse scenario, namely the behavior of the AE of witF when placed into the context of dad13A lacking its native AE. Consistent with the existence of very weak basal, uniform activity within dad13A, an AE-less dad13A construct (dad13A^ΔAE) shows only weak and ‘patchy’ reporter expression throughout the follicular epithelium (Fig. S6A,B). Remarkably, inserting the AE of witF into this construct to generate a chimeric dad13A^FAE, fully reinstated the anterior stripe of expression, suggesting that the AE of witF is now able to respond to direct activatory Smad input (Fig. S6C). Furthermore, dad13A^FAE displays features and genetic requirements that are typical for dad13A, rather than witF. First, the stripe of anterior expression is straight and not wedge-shaped along the dorsoventral axis. Second, reporter activity is only weakly sensitive to genetic removal of brk (Fig. S6D-D″), but almost completely lost in Mad/brk mutant clones (Fig. S6E-E″). Thus, the AE of witF, which in its native context responds solely to Brk, responds to Brk-independent Smad activatory input to boost anterior expression of an otherwise very weak basal CRM.

In the present study, we establish wit as a transcriptional target of BMP signaling in two different epithelia, the wing imaginal disc and the follicle cells of the developing egg chamber. Given its function as a BMP receptor, wit adds to the small group of genes that are coordinately regulated by BMP signaling and are involved in BMP signal transmission and regulation. Such groups are often referred to as ‘synexpression groups’ to emphasize relationships in regulation and function (Karaulanov et al., 2004; Niehrs and Pollet, 1999). Accordingly, the Drosophila BMP synexpression group includes Dad, wit, brk and pentagone (pent; magu), which are either activated (Dad, wit) or repressed (brk and pent) by BMP in a variety of tissues and developmental stages (Hamaratoglu et al., 2014). The concept of synexpression predicts common strategies of transcriptional regulation; however, the activation of wit reveals substantial differences in comparison to Dad. This might mirror differences in expression and, potentially, requirement of Dad and Wit during fly development. Dad is activated by BMP signaling in all tissues and developmental stages analyzed so far – hence its widespread use as a reliable marker for BMP pathway activation. Moreover, expression of Dad is mediated by a single CRM (dad13), which seems to respond to BMP signaling in all contexts, including the wing imaginal disc and the follicle cells. In contrast, and despite its BMP-dependent activation in wing discs and follicle cells studied here, expression of wit is not always connected to BMP signaling. In contrast to Dad, wit is not expressed in all cells with active BMP signaling; for example, wit is not a BMP target in the early embryonic epidermis or in germline stem cells. In addition, there are instances in which wit transcription is independent of BMP signaling. The prominent neuronal expression of wit, for example, seems not to be induced by BMP signaling (Robin Vuilleumier and Douglas Allan, personal communication). This versatility in wit expression might explain the lack of a ‘universal’ (Dad-like) BMP-dependent CRM and instead necessitates distinct, tissue-restricted CRMs. This might also explain the pronounced differences in BMP input in the regulation of wit and Dad in follicle cells. Whereas Dad is directly activated by Smads in the follicular epithelium, activation of wit in the same tissue is delegated to tissue-specific factors, and the BMP input is exclusively mediated by Brk repression (see model in Fig. 7). In addition, the observed differences might complement quantitative constrains. In this case, the tissue-specific, uniform input would be strong enough to activate wit transcription through witF in follicle cells and BMP's role is de-repression (Brk repression) without providing direct activatory input. In contrast, BMP-independent inputs for Dad activation in the same tissue are extremely weak; hence, BMP signaling is primarily required for direct activation rather than alleviating Brk repression. Between these two extreme scenarios, it is conceivable that cues that direct low level basal activation (probably in witF and Dad in the wing) require both activatory Smad input as well as repressive Brk input to boost expression at regions of high pMad levels and erase expression at the low end of the gradient, respectively (Fig. 7).

At the molecular level, both enhancers of wit comprise similar, yet not identical, AE-like elements that fully account for BMP responsiveness to BMP signaling. Importantly, the sequence requirements within the AEs are fully consistent with our genetic analyses. Specifically, our finding that wit and witF are exclusively regulated by the indirect, Brk-dependent branch of the BMP pathway in follicle cells is fully supported by the mutational analysis, which identifies a clear requirement for the Brk-binding site – but not for the Smad sites – within the AE of witF. Similarly, transcription of wit/witW in the wing integrates both direct and indirect BMP inputs and, indeed, mutations that inactivate either the Brk- or the Smad-binding sites of the AE predictably affect the activity of witW.

The sequences of the identified AEs in witW and witF deviate from the prototypic AE of Dad and from each other; however, these differences, despite being evolutionarily conserved, do not seem to have functional consequences and do not account for the different behavior of the cognate CRMs. Consequently, witF and witW respond to Brk alone or to Brk/Smad inputs, respectively, even when their AEs are swapped. An extreme demonstration of the latter phenomenon is exemplified by the Dad and witF chimeric constructs in the follicular epithelium. The original AE of Dad strongly responds to Smad input in the context of the Dad CRMs; however, placing this element into the witF backbone fully overrides its ability to respond to Smads and converts it into a Brk-only response element. In reverse, the AE of witF, which responds exclusively to Brk in its native context, responds to Brk-independent Smad activation in the context of the Dad CRM. Notably, our results are in agreement with a recent study focusing on evolutionary diversification of wit expression between Drosophila species as illustrated by differences in the width of the anterior wit stripe in D. melanogaster and D. virilis egg chambers (Marmion and Yakoby, 2018). Marmion and Yakoby independently identify Brk and the Brinker-binding site as the mediators of BMP-dependent wit expression. At the same time, they demonstrate that differences in the sequences immediately flanking an otherwise identical Brk-binding motif do not account for the observed differences in the expression patterns between the two Drosophila species. The findings cumulatively suggest that the CRM environment, rather than the BMP-response element itself, dictates how the latter will respond to BMP signaling. Such CRM activatory input(s), although able to impact on and equalize the output of any AE variant, have limitations as they cannot depolarize the activity of the SEs. The nature of these activatory elements, which obviously also implement tissue specificity, as well as their integration with the BMP-response elements, need to be elucidated in future studies. It is unclear whether the underlying mechanisms affect binding of Brk and Smads to their cognate sites or whether the decision is made after their binding to the AE. In any case, our data clearly highlight an unexpected flexibility in the structure of the AE, which needs to be considered when employing such elements for in silico detection of BMP target CRMs and genes.

brkB-GFP and brkB14-GFP were used to visualize brk expression in follicle cells and wing disc, respectively (Charbonnier et al., 2015). The dad13-lacZ fly line has been previously described (Weiss et al., 2010). The following fly lines and chromosomes were used for mosaic analyses: brk^M68FRT18A, mad¹²FRT40A, and corresponding FRT chromosomes carrying ubiGFP constructs or mosaic analysis with a repressible cell marker (MARCM) components. Mad/brk double-mutant clones were generated using [brk^BAC]ubiGFPFRT40A (gift from K. Basler, Institute of Molecular Life Sciences, University of Zurich, Switzerland) in a brk mutant background; [brk^BAC] is a genomic rescue construct of brk inserted on 2L22A (Charbonnier et al., 2015; Schwank et al., 2008). FRT-mediated FLP-out clones were generated using (ywhsFLP; Sp/CyO; act>CD2>Gal4, UASGFP) and Dpp signaling was altered by utilizing UAS-dad (Tsuneizumi et al., 1997) and UAS-tkv^QD (Nellen et al., 1996). Larvae (72-96 h after egg laying) or female flies (3- to 5-days old) were subjected to a 37°C heat shock for 1 h or 7-10 min for the generation of mitotic mutant clones or flip-out clones, respectively. Wing discs and ovaries were dissected 48 h after heat shock treatment. A detailed list of fly stocks used in each panel of this study is provided in Table S1.

PCR was used to amplify genomic sequences from wit and Dad loci including introduction of deletions or point mutations. All reporter fragments were subcloned into the placZattB reporter vector and verified by sequencing. A detailed list of primers used to generate the reporter fragments is provided in Table S2. All constructs were inserted by PhiC31/attB-mediated integration into chromosomal position Chr3L, 68A4 (attP2) (Bischof et al., 2007; Groth, 2004).

Drosophila female ovaries and third instar larvae were dissected and fixed in 4% paraformaldehyde/Schneider's S2 medium for 10 min. After multiple washes with PBSTx (1× PBS and 0.1% Triton X-100), samples were incubated with primary antibodies overnight at 4°C. After washing, secondary antibodies were incubated for 2 h at room temperature. The following primary antibodies were used: mouse anti-Wit (1:10; 23C7, Developmental Studies Hybridoma Bank, DSHB) (Aberle et al., 2002), mouse anti-β-Gal (1:500; Z3781, Promega), rabbit anti-β-Gal (1:500; 55976, MP Biomedicals) and chicken anti-GFP (1:1000; ab13970, Abcam). Alexa fluorophore-conjugated secondary antibodies (1:500; A11031, A11039 and A11036, Molecular Probes) and Hoechst 33342 (1:5000; H3570, Invitrogen) were used. Images were obtained using a Nikon C2 confocal microscope and processed with ImageJ and Adobe Photoshop. All images of larval wing discs and egg chambers are positioned posterior to the right and dorsal up.

We would like to thank Konrad Basler and the Bloomington Drosophila Stock Center (NIH P40OD018537) for fly lines and the staff of the Life Imaging Center (LIC) in the Center for Biological Systems Analysis (ZBSA) of the Albert-Ludwigs-University Freiburg for excellent support in image recording. The anti-Wit antibody developed by the Goodman lab (University of California, Berkeley) was obtained from 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. We are indebted to Nir Yakoby and Rob Marmion for discussions and exchange of data.