Functional evolution of a morphogenetic gradient

Bone Morphogenetic Proteins (BMPs) pattern the dorsal-ventral axis of bilaterian embryos; however, their roles in the evolution of body plan are largely unknown. We examined their functional evolution in fly embryos. BMP signaling specifies two extraembryonic tissues, the serosa and amnion, in basal-branching flies such as Megaselia abdita, but only one, the amnioserosa, in Drosophila melanogaster. The BMP signaling dynamics are similar in both species until the beginning of gastrulation, when BMP signaling broadens and intensifies at the edge of the germ rudiment in Megaselia, while remaining static in Drosophila. Here we show that the differences in gradient dynamics and tissue specification result from evolutionary changes in the gene regulatory network that controls the activity of a positive feedback circuit on BMP signaling, involving the tumor necrosis factor alpha homolog eiger. These data illustrate an evolutionary mechanism by which spatiotemporal changes in morphogen gradients can guide tissue complexity. DOI: http://dx.doi.org/10.7554/eLife.20894.001


Introduction
The specification of different cell fates by morphogen gradients has been a longstanding focus within developmental biology. While it is well established that gradients of diffusible morphogens produce complex pattern during development, their role as drivers of morphological evolution has mostly been inferred from theoretical studies, due to the challenge of quantifying and functionally assessing their activities in species outside of select genetic model organisms (Turing, 1952;Kondo and Miura, 2010;Green and Sharpe, 2015;Marcon et al., 2016). Bone Morphogenetic Proteins (BMPs) pattern the embryonic dorsal-ventral axis of bilaterian embryos, raising the question of the role of the BMP gradient in the evolution of body plans (Bier and De Robertis, 2015). To address this question, we compared the functions of embryonic BMP gradients in two fly species that differ in tissue complexity downstream of BMP signaling.
In Drosophila melanogaster embryos, the BMP gradient forms through directed extracellular BMP movement and initiates a positive feedback circuit leading to a bistable pattern of BMP signaling by the end of the blastoderm stage (reviewed in O'Connor et al., 2006;Shilo et al., 2013;Wharton and Serpe, 2013). High levels of BMP signaling, centered on the dorsal midline, specify a single extraembryonic tissue, the amnioserosa. However, in basal-branching flies, including Megaselia abdita (Phoridae), BMP signaling specifies two extraembryonic tissues, the serosa and the amnion (Schmidt-Ott and Kwan, 2016). Previously, we showed that the dynamics of BMP signaling in the blastoderm are similar between Megaselia and Drosophila, but differ in the early gastrula when the Megaselia gradient broadens while the Drosophila gradient remains static (Rafiqi et al., 2012). Here, we show that differences in the control of a positive feedback circuit involving eiger (egr) (Gavin-Smyth et al., 2013) are responsible for the altered dynamics of BMP signaling and amnion specification in Megaselia. We hereby reveal an evolutionary mechanism by which morphogen gradients can alter the complexity of tissue types between species.

Results and discussion BMP signaling during gastrulation is necessary and sufficient for amnion specification
In Megaselia and Drosophila, BMP signaling specifies extraembryonic membranes ( Figure 1A) and can be quantified by staining with an antibody specific to the activated phosphorylated form of Mad (pMad), an essential transcriptional effector of the BMP pathway (Dorfman and Shilo, 2001). During early blastoderm stages in both species BMP signaling is initially low and broadly distributed over the dorsal regions of the embryo but refines into a narrow dorsal stripe of high activity by the onset of gastrulation. However, during early gastrulation in Megaselia, the BMP signaling domain broadens to encompass the edge of the germ rudiment comprising the presumptive amnion, while the domain in Drosophila remains static ( Figure 1B).
In Megaselia, serosa and amnion specification can be visualized with a combination of genetic markers. A homolog of zerknüllt (Mab-zen), which encodes a homeodomain protein, marks and specifies serosa cells in blastoderm (stage 5) and gastrula embryos (stage 6) (Rafiqi et al., 2008). Homologs of hindsight (Mab-hnt) and dorsocross (Mab-doc, Mab-doc2) ( Figure 2A and  (Rafiqi et al., 2012), encompassing both the prospective serosa and amnion. Lastly, a homolog of the TNF alpha gene eiger (Mab-egr) is expressed in the serosa and amnion of gastrulating Megaselia embryos ( Figure 2B-C). Mab-egr expression continues until dorsal closure, but from germ band extension (stage 11) onwards it is expressed only in the amnion cells, which at this stage are polyploid and much larger than the adjacent embryonic cells ( Figure   To test the hypothesis that temporally distinct BMP signaling sequentially specifies the two extraembryonic membranes, we decreased BMP signaling during gastrulation by Mab-dpp knockdown after 50% blastoderm cellularization and monitored the expression of Mab-eve. Knockdown of Mabdpp by injection of double stranded RNA (dsRNA) at the early blastoderm stage completely suppresses both serosa and amnion specification and results in the circumferential expression of Mabeve (Rafiqi et al., 2012). In contrast, in all 13 late Mab-dpp knockdown embryos fixed during early gastrulation, the expression of Mab-zen was not affected; however, in five embryos the repression of Mab-eve in the amnion anlage was incomplete ( Figure 2H-I). Late Mab-dpp knockdown reduced Mab-egr expression in a majority of stage 11/12 embryos (55%, n = 40) whereas late ectopic Mabdpp expression caused an expansion of the Mab-egr domain in at least 35% of the embryos (n = 57) ( Figure 2J-L). Taken together, these data provide evidence that BMP signaling during gastrulation is necessary and sufficient for amnion specification.
Mab-doc promotes amnion development by elevating BMP signaling in the amnion anlage at the onset of gastrulation In Drosophila, both BMP signaling and zen are necessary at the blastoderm stage for the expression of the three doc paralogs and hnt in the amnioserosa anlage, although the essential function of these genes in amnioserosa maintenance becomes apparent only after gastrulation (Yip et al., 1997;Reim et al., 2003). However, in Megaselia, BMP signaling activates Mab-doc and Mab-hnt independently from Mab-zen (Figure 3-figure supplement 1A-I), suggesting these genes could play a role in amnion specification. Following knockdown of Mab-doc/doc2 or Mab-hnt activity by RNAi, we observed confluent expression domains of Mab-zen and Mab-eve during early gastrulation (4/9 and 5/11 embryos, respectively; Figure   The activities of Mab-doc and Mab-hnt could promote the amnion formation in an instructive manner, by activating the amnion gene network of Megaselia, or they might promote the amnion formation in a permissive manner, e.g., by elevating BMP signaling. Overexpression of Mab-doc in Mab-dpp knockdown embryos did not induce any expression of Mab-egr in stage 11/12 embryos (n = 44) ( Figure 3H). Conversely, overexpression of Mab-dpp in Mab-doc/doc2 knockdown embryos resulted in ectopic expression of Mab-egr at stage 11/12 (36%, n = 47) ( Figure 3I). Thus, BMP signaling is sufficient to direct the expression of amnion specific genes in the absence of Mab-doc/ doc2 activity. To confirm that this result was not due to an excessive non-physiological level of Mab-Dpp produced by the injected mRNA, we asked whether the endogenous level of BMP signaling at the dorsal midline in the blastoderm embryo would be sufficient to specify amnion in the absence of both Mab-doc/doc2 and the serosal determinant Mab-zen. Knockdown of Mab-zen partially restored amnion in Mab-doc/doc2 knockdown embryos (Figure 3-figure supplement 1O). Taken together, these data suggest Mab-doc promotes amnion formation permissively.
To directly test whether Mab-doc activity elevates BMP signaling, we quantified the intensity of pMad staining in embryos after Mab-doc/doc2 knockdown. While Mab-doc/doc2 knockdown had little effect on pMad levels during the late blastoderm stage compared to control embryos (Wilcoxon rank sum test, p=0.3697; Figure 3J), in early gastrula stage embryos, knockdown of Mab-doc/doc2 resulted in significantly reduced pMad levels compared to control embryos (Wilcoxon rank sum test, p=0.01165; Figure 3K). In contrast, knockdown of Mab-zen did not alter the average level of pMad at the beginning of gastrulation (Wilcoxon rank sum test, p=0.2367; Figure 3L). The observation that Mab-doc/doc2 is dispensable for amnion cell fate specification but necessary for wild-type levels of BMP signaling at the early gastrula stage strongly support the model that amnion formation is driven by a Mab-doc-dependent elevation of BMP signaling in the amnion anlage at the onset of gastrulation.
Mab-doc-dependent control of Mab-egr expression contributes to a positive feedback circuit that promotes BMP signaling during gastrulation We then explored the mechanism by which Mab-doc promotes BMP signaling at the gastrula stage. Embryos injected centrally with Mab-doc mRNA displayed a local expansion of the pMad domain during gastrulation (15/15) coupled with the frequent depletion of endogenous pMad in adjacent regions (12/15) ( Figure 4A). This result parallels a phenotype observed in Drosophila where injection of mRNA encoding activated BMP receptors into the blastoderm embryo causes an increase in BMP ligand-receptor interactions, coupled with a decrease in BMP ligand-receptor binding in nearby regions (Wang and Ferguson, 2005). These data indicate that a positive feedback circuit downstream of BMP signaling increases local receptor-ligand interactions, and that, due to a limiting amount of BMP ligand, ligand-receptor interactions decrease in nearby regions. Conversely, Megaselia embryos injected with Mab-zen mRNA (n = 11) had a similar pMad domain to injected control embryos (n = 12) and developed a reduced or abnormal amnion (44/51) (Figure 4-figure in Mab-doc/doc2 knockdown embryos (red) at the cellular blastoderm stage (n = 10, control n = 10) (J), at early gastrulation (n = 11, control n = 11) (K), and in Mab-zen knockdown embryos (red) at early gastrulation (n = 10, control n = 17) (L) with representative embryos stained for pMad underneath each plot. DOI: 10.7554/eLife.20894.008 The following figure supplement is available for figure 3:    (Figure 4B), only a few embryos (2/14) displayed a depletion of endogenous pMad in adjacent regions. These data indicate that Mab-egr increases the ability of cells overexpressing Mab-doc to compete for BMP ligands during early gastrulation.
In Drosophila, loss of egr reduces intensity of pMad staining by 50% (Gavin-Smyth et al., 2013). Similarly, we found that, at the onset of gastrulation, pMad levels in Mab-egr knockdown embryos were reduced by about 50% on average (Wilcoxon rank sum test, p=0.00381) ( Figure 4C and Figure 4-figure supplement 2). Moreover, confluent expression domains of Mab-eve and Mab-zen could be observed in such embryos (3/10) ( Figure 4D). As Mab-egr expression extends to the edge of the gastrulating germ rudiment, these observations suggest that Mab-egr promotes amnion specification downstream of Mab-doc/doc2 by elevating BMP signaling during gastrulation in prospective amnion cells.
The pMad gradients of Mab-egr RNAi embryos were on average slightly broader than in Mabdoc/doc2 RNAi embryos ( Figures 3K and 4C), suggesting that Mab-doc/doc2 might control more than one gene with a role in shaping the BMP gradient, similar to Drosophila where the BMP-dependent feedback circuit appears to involve more genes than just egr (Gavin-Smyth et al., 2013). As a potential second Mab-doc/doc2 target we examined the Megaselia ortholog of cv-2 (Figure 4-figure supplement 3), which encodes an extracellular, context-and concentration-dependent modulator of BMP signaling (Conley et al., 2000;Serpe et al., 2008). However, although Mab-cv-2 likely also promotes BMP signaling in Megaselia (Figure 4-figure supplement 4), it appears to function independently of the Mab-doc/doc2-dependent feedback loop (Figure 4-figure supplement 5).
Lastly, we examined the regulation of Mab-egr expression. In Drosophila, egr expression begins at the syncytial blastoderm stage under the control of both BMP signaling and zen, whereas in Megaselia, Mab-egr expression begins at the onset of gastrulation. In Mab-dpp knockdown embryos, Mab-egr expression was completely absent ( Figure 4E-F). In Mab-doc/doc2 knockdown embryos, Mab-egr expression was greatly reduced ( Figure 4G). Mab-doc/doc2/hnt triple knockdown did not further reduce Mab-egr expression during gastrulation (not shown). Mab-zen knockdown embryos displayed only a slight reduction in Mab-egr expression during gastrulation; however, at germ band extension, Mab-zen knockdown embryos displayed an increase in the number of Mab-egr expressing cells due to the transformation of serosa into amnion ( Figure 4H-I). Thus, in Megaselia, Mab-egr is primarily under the control of Mab-doc, not Mab-zen.

Conclusions
While previous work has demonstrated that BMP gradients can form and be stabilized through molecular feedback circuits (Bier and De Robertis, 2015), we have shown here that positive feedback circuits can also be an important target in the evolution of body plans. Specifically, the distinct BMP gradients of Megaselia and Drosophila, which result in the specification of distinct extraembryonic tissue complements, are the result of spatiotemporal changes in an egr-dependent positive feedback circuit ( Figure 4J).
Given that Megaselia has retained distinct serosa and amnion tissues, the function of the BMP gradient in this species might be similar to the ancestral condition in higher flies. What changes to the underlying genetic network during evolution would have been necessary to transform the shape of the Megaselia BMP gradient into that seen in Drosophila? In blastoderm embryos of Megaselia, the BMP gradient establishes the expression of Mab-zen in prospective serosa tissue, and Mab-doc/ doc2 and Mab-hnt in prospective serosa and amnion tissues. While this patterning phase is not sufficient to differentiate between serosa and amnion tissue, it sets the stage for Mab-doc/doc2-dependent Mab-egr expression during gastrulation. The Drosophila gene network that regulates egr expression differs at least in two ways. First, doc (along with hnt) is expressed downstream of zen. Conceptually, this regulatory difference is sufficient to explain the difference of egr expression between the two species during gastrulation, and hence also the difference in BMP signaling and tissue specification. We therefore propose that this change led the evolutionary transition. Once doc was downstream of zen, the latter might have gradually gained control over egr. This scenario is consistent with our observation that even in Megaselia, Mab-zen slightly promotes Mab-egr expression. Second, Drosophila acquired a BMP-independent zen expression domain in the syncytial blastoderm, which is not observed in other dipterans (Goltsev et al., 2007;Rafiqi et al., 2008). The acquisition of this early zen domain could have promoted egr expression in the blastoderm of Drosophila, where egr is part of a zen-dependent network that confers robustness to the BMP gradient (Gavin-Smyth et al., 2013).
Our data suggest that the ancestral function of the positive feedback circuit was to promote amnion specification through BMP signaling. While the identity of regulatory factors of the positive feedback circuit may be evolutionarily labile (in Tribolium doc and hnt appear to be dispensable for amnion specification [Horn and Panfilio, 2016]), we propose that the mechanism of amnion specification through feedback-driven BMP signaling dynamics applies to a wide range of insects, because in Tribolium the pMad domain also gradually shifts to become elevated in the presumptive amnion during early gastrula stages (van der Zee et al., 2006;Nunes da Fonseca et al., 2008). The principle of evolving tissue complexity through changes in positive feedback circuits of morphogen gradients has not yet been documented in other developmental contexts, but it may also apply to unrelated traits, such as eyespots on butterfly wings (Monteiro, 2015).
(University of Chicago) for comments on the manuscript. This work was supported by the National Science Foundation Grant IOS-1121211 to US-O and an award of the University of Chicago Hinds Fund to CWK. The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Additional information
Author contributions CWK, Conception and design, Acquisition of data, Analysis and interpretation of data, Drafting or revising the article; JG-S, ELF, Analysis and interpretation of data, Drafting or revising the article; US-O, Conception and design, Analysis and interpretation of data, Drafting or revising the article