An exclusive cellular and molecular network governs intestinal smooth muscle cell differentiation in vertebrates

Smooth muscle cells (SMCs) constitute a vital proportion of various organs, including those of the gastrointestinal (GI) tract, urogenital tract, respiratory tract and vascular system. Despite their crucial contribution to organ function, little is known about the ontogeny and genetic developmental programs that drive SMC differentiation in vertebrates. A key challenge to studying the mechanisms of SMC development and differentiation arises from the complex origin of SMCs from seemingly multiple and sometime unknown cell types (Kumar and Owens, 2003). Current concepts describe most SMCs as arising from the condensation of surrounding, vaguely defined mesenchyme under the control of local environmental cues. In coordination with the different cell types present in the developing organs, mesenchyme initially forms early-synthetic SMCs that later develop into mature contractile SMCs (Gabella, 2002). A complex SMC lineage is the intestinal SMCs (iSMCs), which is found around the enteric endoderm-derived epithelium. iSMCs are indispensable for proper gut organogenesis as they contribute to vilification and provide the contractility necessary for intestine functionality (Shyer et al., 2013). Defects in their development are apparent in human congenital disorders such as visceral myopathy.

Lateral plate mesoderm (LPM) is a highly dynamic mesoderm field composed of bilateral stripes of cells appearing in post-gastrula embryos. The LPM is patterned early into distinct regions that will give rise to precursors of kidney, heart, endothelium, hematopoietic and limb cell fates (Davidson and Zon, 2004; Gering et al., 2003; Mosimann et al., 2015). Although previous work has suggested that iSMCs arise from the lateral plate mesoderm (LPM), genetic demonstration for this origin is still missing in a vertebrate model (Roberts et al., 1998). Currently lacking is a cellular and molecular concept of how the bilateral precursor stripes form the smooth muscle layer surrounding the endoderm-derived gut tube, and whether these cells indeed derive from the LPM. How the possibly LPM-derived iSMC precursors induce and regulate their migration to converge on and surround the gut tube also remains unknown. In the past, early events of LPM and gut morphogenesis have been well described, taking advantage of the zebrafish model system (Horne-Badovinac et al., 2003; Stainier, 2005). The anatomical conservation and relative simplicity of its intestine have made the zebrafish an ideal vertebrate model for studying early gut development and endodermal differentiation (Bagnat et al., 2007; Horne-Badovinac et al., 2003; Wallace et al., 2005; Yin et al., 2010), and the initial characterization of iSMCs (Georgijevic et al., 2007; Wallace et al., 2005; Whitesell et al., 2014).

Organogenesis requires a highly coordinated series of molecular and cellular events. Among the different categories of molecules involved in organ formation and cell fate control, miRNAs represent a sophisticated level of gene regulation that coordinates a broad spectrum of biological processes, from development to cancer (Kloosterman and Plasterk, 2006). miRNAs are endogenous ∼22-nucleotide RNAs that control protein expression through translational repression of mRNAs. In cooperation with transcription factors, miRNAs can establish autoregulatory feedback loops and feed-forward loops, reaching high levels of complexity in the regulation of gene expression and subsequently of biological processes (Tsang et al., 2007).

Here, combining genetic, pharmacological and bioinformatics approaches, we characterize cellular and molecular events occurring during LPM differentiation and intestinal SMC development in zebrafish. Using genetic lineage tracing, we demonstrate that iSMCs arise from the LPM in a stepwise process. We show that a TGFβ- and Zeb1a-mediated migration of hand2-positive LPM cells around the gut endoderm drives commitment of epithelial LPM into mesenchymal iSMC progenitors. TGFβ/Alk5 signaling also leads to the expression of miR-145 that is required to switch off the migrating signature of the LPM and to downregulate translation of the Forkhead transcription factor gene foxo1a, a novel component of LPM and iSMC differentiation. Together, our data uncover a sequence of unique molecular events that govern distinct steps during the emergence and differentiation of iSMCs from migrating LPM in vertebrates. Understanding of how iSMCs develop is key to targeting smooth muscle cell-related pathologies and to improve prognostic and therapeutic approaches.

Previous reports indicated that zebrafish embryos mutant for the LPM-expressed transcription factor gene hand2 (heart and neural crest derivatives expressed 2) (Yelon et al., 2000) completely lack iSMCs (Santoro et al., 2009). To investigate how LPM emergence and differentiation are related to iSMC formation, we combined different approaches.

We first tracked LPM derivatives in a BAC-based reporter transgenic line Tg(hand2:EGFP)^pd24 based on the endogenous hand2 cis-regulatory elements that also express in the presumptive posterior LPM from early somitogenesis onwards [Tg(hand2:EGFP)^pd24] (Yin et al., 2010). Using confocal microscopy of transverse embryo cross-sections, we examined EGFP expression between somites 7 and 13, a region in which the enteric endoderm is located at the midline (i.e. above the yolk extension; Fig. 1A and Fig. S1A). By 24 h post-fertilization (hpf), hand2-expressing cells in zebrafish embryos form bilateral mesodermal sheets spanning the entire anterior-posterior (A-P) extent of the trunk. At this time point, this remaining undifferentiated LPM is located lateral to the gut and is composed of polarized proliferating epithelial cells (Horne-Badovinac et al., 2003; Yin et al., 2010). By 30 hpf, these hand2-expressing epithelial sheets started to cover the dorsal region of the gut endoderm. By 36 hpf, the LPM had enfolded the region underneath the endoderm through a process reminiscent of mesenchymalization. By 48 hpf, the gut tube was completely surrounded by hand2-expressing cells. From 60 hpf onwards, these hand2-positive cells expressed acta2 (α-smooth muscle actin) and tagln (transgelin or sm22a-b). These genes are the earliest known markers of committed smooth muscle progenitor cells in vertebrates and remain expressed in differentiated SMCs (Georgijevic et al., 2007; Solway et al., 1995; Santoro et al., 2009). By 96 hpf, iSMCs were fully differentiated in contractile longitudinal and circular smooth muscle fibers, and promoted peristaltic movement of the gut in preparation for the onset of exogenous feeding (Wallace et al., 2005).

To further characterize the morphogenesis of the hand2-expressing LPM, we tested expression of epithelial markers, such as aPKC (atypical protein kinase C), and markers of mesenchymalization, such as N-cadherin, in the LPM from 24 hpf onwards (Fig. 1B,C). These results revealed that hand2-expressing bilateral LPM cells express both markers of epithelial and mesenchymal cells as early as 24 hpf. Our data support the possibility that the LPM cells acquire the feature of a collective migrating epithelial mesenchyme, a common event during embryonic developmental and tissue repair (Rørth, 2012).

By 72 hpf, a subpopulation of hand2-expressing LPM cells start to express the SMC marker Tagln. As shown in Fig. 1C, all the Tagln-positive cells are also positive for hand2 expression, supporting the conclusion that all the iSMCs originate from LPM/hand2+ cells (Fig. 1D). Tg(hand2:EGFP)^pd24 also exhibited EGFP-positive cells located in the enteric submucosa that were negative for Tagln but positive for Hu, a marker specific for neurons (Fig. S1B). As hand2 is also expressed in neural crest derivatives and is required for the development of neural crest-derived neurons (Olden et al., 2008; Reichenbach et al., 2008), we concluded these cells are enteric neurons. Taken together, our observations confirm and extend previous reports that hand2-expressing bilateral LPM cells give rise to the iSMC layer surrounding the developing gut tube.

As a second and independent approach to link iSMCs to an LPM origin, we performed Cre/lox-mediated lineage tracing in the Tg(drl:creERT2) line, which uniquely expresses tamoxifen-inducible Cre recombinase in all presumptive LPM precursors already during late epiboly (Mosimann et al., 2015). We crossed drl:creERT2 with the ubiquitous GFP-to-mCherry loxP lineage trace transgene ubi:Switch (Mosimann and Zon, 2011) and induced Cre activity at late epiboly/tailbud stages, when drl transgene expression is confined to presumptive LPM cells. We detected lineage-labeled precursor iSMCs at 72 hpf and iSMCs around the gut along the entire length of the trunk, concomitant with the expected LPM-derived lineage labeling of the pronephric duct and endothelial cells (Fig. 2A). Lineage-labeled cells surrounding the gut co-stained with the iSMC marker Tagln as early as 72 hpf (Fig. 2B and Fig. S2B). We found lineage-labeled iSMCs in all embryos treated with 4-OH at 1 ss (n=31) (Fig. 2C). In all embryos tested, we observed different grades of switching efficiency, ranging from a few iSMC labeled (class I) cells to complete lineage labeling of all gut-surrounding iSMCs (class III). The variability and efficiency corresponds to the ubiquitous ubi:Switch recombination capacity in controls (Fig. S2A) and in our previous ubi:Switch characterizations (Felker et al., 2016). Taken together, our genetic lineage tracing results demonstrate that initially drl-expressing and subsequently hand2-expressing LPM cells form mesenchymal cells that later on become iSMCs. Altogether, our data show that the LPM gives rise to iSMCs in zebrafish and support the notion that the signaling and genetic pathways driving the emergence and differentiation of the LPM might also underlie iSMC formation.

To specifically track the development and maturation of iSMCs, we next derived two independent transgenic zebrafish reporter lines with fluorescent markers under the control of the acta2 and tagln minimal cis-regulatory elements (Fig. S3A,B; see Materials and Methods for details). Although reporter expression in these lines differed in intensity and specificity, both Tg(acta2:mCherry)^uto5 and Tg(tagln:CAAX-EGFP)^uto37 embryos exhibit fluorescent marker expression in immature iSMCs beginning at 60-72 hpf. By 96 hpf and through adulthood, both reporter lines mark mature and contractile iSMCs covering the entire intestine and swim bladder (Fig. S3A,B). Our new acta2 and tagln transgenic reporters are therefore bona fide reporter lines for immature and mature iSMCs.

We next used our acta2 and tagln reporter lines as readouts to screen for signaling pathways that drive iSMC formation using a panel of established chemical inhibitors (Table S2). Chemical inhibition from 20 hpf of the TGFβ type I receptors by SB431542 and LY364947 selectively impaired iSMC development (Fig. 3A,B and data not shown). We further confirmed the role of TGFβ in iSMCs by analyzing ltbp3 morphants that were previously shown to specifically phenocopy Alk5 inhibition (Zhou et al., 2011). Both pharmacological and genetic perturbation of TGFβ signaling disrupted iSMC differentiation in vivo without interfering with overall gut endoderm specification and morphology (Fig. 3A,B and Fig. S3C). To confirm these data, we then evaluated iSMCs differentiation markers in Tg(hsp70:caALK5), in which heat shock triggers constitutive Alk5 activity and signaling (Zhou et al., 2011). Heat-shock-induced expression of constitutively active Alk5 increased acta2, tagln and myh11 expression, further supporting the role of TGFβ signaling in promoting iSMC mainly through Alk5 receptor (Fig. S3D).

One of the key functions of TGFβ signaling during development is to promote cell migration and invasion (Lim and Thiery, 2012; Zhang et al., 2014). We consequently hypothesized that TGFβ could also control the migration of hand2-positive LPM. In accordance with this hypothesis, LPM ventral migration was severely yet specifically impaired after SB431542 treatment and ltbp3 knockdown (KD) at 48 hpf (Fig. 3C), whereas the total number of LPM/hand2+ cells did not change significantly in Tg(hand2:EGFP)^pd24 embryos upon TGFβ inhibition (Fig. 3D and Fig. S3E). These data support a new role for TGFβ signaling in LPM-to-iSMC differentiation by promoting initial LPM migration.

To elucidate the downstream targets of TGFβ that might drive LPM migration and differentiation in iSMC, we analyzed transcriptomic data to identify: (1) genes induced by TGFβ – specifically and differentially expressed between human alveolar basal epithelial cells (A549) after 72 h of TGFβ induction and untreated cells (Sartor et al., 2010); (2) genes expressed in intestinal mesenchyme – specifically and differentially expressed between the mesenchymal and epithelial fraction of mouse intestine (Li et al., 2007) (Fig. S3F). Among those resulting genes, we focused our attention on zeb1a (zinc finger E-Box binding homeobox 1) and foxo1a (forkhead box protein O1), two transcription factor-encoding genes whose roles during the development of the GI tract remain unknown.

Zeb1a is a potent mediator of cell migration and invasion of tissues downstream of TGFβ signaling (Lamouille et al., 2014; Zhang et al., 2014). Accordingly, a specific role for Zeb1a during vascular SMC differentiation has been well established (Nishimura et al., 2006). However, a potential role for ZEB family members in iSMC development has not yet been determined. Therefore, we investigated whether ZEB1 is required for iSMC formation in zebrafish development using our two reporter transgenic lines. We silenced zeb1a in Tg(acta2:mCherry)^uto5 and Tg(hand2:EGFP)^pd24 embryos; injections of two independent zeb1a morpholinos (translation and splicing blocking) both abrogated iSMC development without affecting gut or endoderm development and morphology (Fig. 4A,B and Fig. S4A-C). To understand whether this defect was due to impaired LPM migration, we analyzed LPM morphology 48 hpf after silencing zeb1a. In zeb1a-impaired embryos, the LPM does not complete its migration and fails to cover the ventral region of the gut endoderm (Fig. 4C). We did not observe any significant differences in hand2 expression levels compared with controls (Fig. S4D) nor in LPM/hand2+ cell number (Fig. S4E). We also collected hand2+ cells from zeb1a knockdown embryos by FACS and analyzed a set of genes associated with mesenchymal migration by qPCR. Compared with controls, silencing of zeb1a markedly increased the expression of epithelial markers, including cdh1 (E-cadherin) and oclna (occludin A), in the hand2-positive cell population. Such molecular features resemble the retention of the compact tight epithelial structure, possibly explaining the migration defects observed before (Fig. 4D).

Altogether, these data support a specific role for TGFβ signaling and zeb1a in driving LPM migration around the gut, a key step towards iSMC commitment. Once lateral-to-medial hand2-positive LPM migration has occurred, mesenchymal cells that now surround the endoderm start to differentiate into iSMCs.

Among the potentially TGFβ-regulated target genes in the intestinal mesenchyme and expressed in the LPM, we also identified foxo1a. Foxo1 belongs to the Forkhead family of transcription factors and regulates myogenic growth and differentiation, maintenance of stemness, and metabolism (Eijkelenboom and Burgering, 2013; Sanchez et al., 2014). A role for foxo1a in iSMC development has not been described previously. To investigate at which step of LPM-to-iSMC differentiation foxo1a might act, we knocked down foxo1a in Tg(acta2:mCherry)^uto5 embryos with both a translational and splice-blocking morpholinos. In addition, we used AS1842856, a specific chemical inhibitor of Foxo1 activity (Nagashima et al., 2010) (Fig. S5A,B). Although foxo1a knockdown did not affect overall embryonic development (nor overall body morphology and gut endoderm morphology or differentiation), it impaired iSMC cell number and marker expression (Fig. S5A,C,D). We then evaluated whether foxo1a was required in the LPM. We found that both genetic and pharmacological inhibition of foxo1a reduced LPM/hand2+ cell number (Fig. 5A,B and Fig. S5E) and LPM proliferation (Fig. 5C). Nonetheless, foxo1a knockdown did not alter LPM migration (Fig. 5A) or the expression of genes associated with EMT and migration compared with controls (Fig. S5F). These data indicate that, complementary to our findings on zeb1a function, foxo1a is dispensable for LPM migration but it is required for LPM proliferation and maintenance.

To further understand the role of foxo1a in the LPM-to-iSMC differentiation, we performed foxo1a overexpression analysis and looked at the LPM differentiation state by measuring hand2 expression levels as an indicator of the LPM versus iSMC differentiation state. Overexpression of foxo1a stimulated hand2 expression in the embryo (Fig. 5D,E), impaired SMC marker expression and iSMC differentiation (Fig. 5D,F), and affected LPM cell number or proliferation (Fig. 5D and data not shown). These data propose foxo1a as a potent previously unrecognized molecular regulator of LPM during early zebrafish iSMC development. Altogether, our data reveal that Zeb1a and Foxo1a each control distinct roles in differentiating hand2-positive LPM (migration versus cell number/proliferation) towards forming functional iSMCs.

We next addressed the spatial and temporal expression of zeb1a and foxo1a in zebrafish, in particular if they are selectively expressed in LPM. We performed whole-mount in situ hybridization for zeb1a and foxo1a mRNA from 24 to 48 hpf stages (Fig. S6A,B). zeb1a is expressed mainly in a region surrounding the gut, possibly mesenchymal tissue. foxo1a expression is evident as early as 24 hpf in a bilateral region similar to the LPM stripes and in the gut region. Later on, foxo1a is also expressed in the endoderm as demonstrated by qPCR on endodermal TgBAC(cldn15la-GFP)^pd1034-sorted cells (Alvers et al., 2014; data not shown).

We next sought to explain the loss-of-function as well as gain-of-function phenotypes of these genes in LPM and iSMCs differentiation. We addressed how the complementary functions of zeb1a and foxo1a are temporally regulated and tuned, and whether a microRNA-based mechanism could be involved. miR-145 is one of the most enriched microRNAs in SMCs where it contributes to the acquisition of the SMC fate and contractile state (Albinsson and Swärd, 2013; Boettger et al., 2009; Cordes et al., 2009; Elia et al., 2009; Xin et al., 2009). Previous work has found that miR-145 expression is also regulated by TGFβ in vascular SMCs in vitro (Long and Miano, 2011). Therefore, we analyzed the expression of miR-145 in developing zebrafish embryos and observed that its expression begins at the onset of iSMC maturation (∼72 hpf) (Fig. 6A). miR-145 was also strongly upregulated in Tg(hsp70:caALK5) embryos after heat shock, whereas chemical or genetic blockade of TGFβ signaling reduced miR-145 expression (Fig. 6B,C). These data indicate that mir-145 is also regulated by TGFβ signaling in iSMCs in vivo and are consistent with a conserved role for TGFβ signaling in miR-145 regulation in both vascular and visceral SMCs (Long and Miano, 2011).

In zebrafish, miR-145 seems highly and selectively expressed in intestinal SMCs (Wienholds et al., 2005; Zeng and Childs, 2012). Previous studies have shown that alterations in miR-145 expression affect overall intestinal maturation (Zeng et al., 2009). To study the role of miR-145 in iSMCs in more detail, we injected low doses of a miR-145 dicer-blocking morpholino, sufficient to significantly reduce mature miR-145 levels without altering endoderm differentiation and overall embryo morphology (Fig. S7A-C). Such miR-145 KD embryos displayed fewer iSMCs in uneven layers around the gut (Fig. 6D). These embryos exhibited only a slight reduction in iSMC marker expression (Fig. S7D) and iSMC number (Fig. 5F). iSMCs in miR-145-impaired embryos showed an altered morphology that was typical of undifferentiated and synthetic SMCs being less stretched and more rounded compared with controls (Fig. 6E) (McHugh, 1996). Crucially, miR-145 knockdown embryos showed severe contractility defects in iSMCs, including deficiencies in swim bladder inflation and gut peristalsis (Fig. S7C and Fig. 6G).

Since miRNAs function by binding and degrading target mRNAs (Bartel, 2009) and by regulating their translation, we sought to identify which protein-coding genes are targets of miR-145 during iSMCs development. We filtered our list of 487 genes induced by TGFβ and expressed in the embryonic intestinal mesenchyme (Fig. S3F) for the presence of a miR-145 binding site. We obtained a list of 41 putative miR-145 targets conserved in human and mouse, containing several genes that had previously been confirmed to be miR-145 targets (Table S3). Among them we found foxo1a, also predicted to be a target in zebrafish. Another gene was zeb2, which has recently been shown to be a direct target of miR-145 (Ren et al., 2014). Within the ZEB gene family in zebrafish, zeb1a has a predicted miR-145 target site. Combined, our data reveal that our identified iSMC regulators foxo1a and zeb1a are potential targets of the SMC-controlling microRNA miR-145.

To test whether zeb1a and foxo1a transcripts are physiologically relevant targets of miR-145 during zebrafish SMC differentiation, we used complementary approaches. We first probed the ability of zebrafish miR-145 to directly bind zeb1a and foxo1a 3′ UTR by luciferase experiments. To achieve this, we cloned the 3′ UTR of both genes into a luciferase reporter vector and performed reporter assays in HEK-293 cells expressing a zebrafish miR-145 mimic or a scramble mimic as negative control. Luciferase expression from the reporter with the wild-type 3′ UTR of zeb1a was significantly repressed but was rescued after mutation of miR-145 binding sites (Fig. 6H and Fig. S7E). We obtained analogous results with the 3′ UTR of the foxo1a gene (Fig. 6H and Fig. S7E). Next, given the unavailability of antibodies to measure Zeb1a and Foxo1a protein levels, we measured the relative abundance of endogenous zeb1a and foxo1a transcripts in control and experimentally manipulated embryos by quantitative PCR (Fig. 6I). Injection of miR-145 morpholino resulted in a ∼2-fold increase in zeb1a and foxo1a expression levels. These data demonstrate that endogenous zeb1a and foxo1a transcript levels change in response to decreased miR-145 activity. Finally, to address the consequence of miR-145-dependent downregulation of zeb1a or foxo1a during iSMC differentiation, we specifically blocked the miR-145-mediated downregulation of zeb1a and foxo1a in live embryos using target protector technology (Staton, 2011). Injections of zeb1a or foxo1a target protectors (zeb1a-TP or foxo1a-TP) in zebrafish embryos specifically impaired iSMC differentiation. foxo1a-TP injection reduced the number of iSMCs (Fig. 6J) whereas zeb1a-TP injection affected iSMC contractility (Fig. 6K). Strikingly, co-injection of foxo1a-TP and zeb1a-TP phenocopied miR-145 knockdown embryos, including fewer hand2-positive iSMCs with disorganized layer architecture (Fig. S7F), indicating that miR-145-mediated targeting of zeb1a and foxo1a mRNA are both required to complete iSMC differentiation and maturation.

We hypothesized that miR-145 is required for differentiation of iSMCs after migration and to allow immature iSMCs to become peristaltic/mature iSMCs. We measured the mesenchymal state of iSMCs in miR-145 knockdown embryos by analyzing the ratio of cdh1 versus cdh2 (N-cadherin) expression. iSMCs with miR-145 knockdown exhibited severe downregulation of cdh1 and, concomitantly, significant upregulation of cdh2 (Fig. 6L,M). Using luciferase assays, we next determined that miR-145 negatively regulated other target genes known to mediate migration, including podxl, fscn1a (fascin actin-bundling protein 1A) and fli1a (Feng et al., 2014; Larsson et al., 2009; Lin et al., 2014) (Fig. 5N). Interestingly, we found that alk5b was also a bona fide target of miR-145 (Fig. 6N), suggesting the existence of a negative-feedback loop between miR-145 and the TGFβ pathway that is responsible for miR-145 induction.

Altogether, these data suggest that miR-145 is required for iSMC maturation and for the acquisition of contractile properties downstream of initial iSMC fate commitment and LPM mesenchymalization and migration. In addition, our results propose that TGFβ-zeb1a and foxo1a regulate LPM morphogenesis and the initial step of LPM-to-iSMC differentiation. The miR-145 expression driven by TGFβ signaling is then required in immature hand2-positive iSMCs to: (1) switch off the mesenchymal program governed by Foxo1a and the migration programs controlled by Zeb1a; and (2) to promote maturation of iSMCs into contractile and fully differentiated SMCs.

Despite their biological and clinical importance, the origin and differentiation of gastrointestinal SMC have been scarcely investigated to date, in particular compared with studies of vascular SMC or endoderm development. Here, using the zebrafish model system, we have studied the developmental origin of vertebrate iSMCs and have identified a genetic program responsible for iSMC differentiation and maturation.

Our data provide evidence that identifies the LPM as the lineage that gives rise to SMCs in the GI tract of zebrafish embryos by combining reporter transgene imaging and genetic lineage-tracing experiments using the LPM-expressed drl:creERT2 (Mosimann et al., 2015). Our lineage-tracing results provide the first genetic confirmation in vertebrate that smooth muscle cells in the gut region are derived from lateral mesodermal organ precursors. These findings are consistent with and extend previous cell culture and transplantation experiments performed in Xenopus and chick, respectively, that provided the first indications that the LPM gives rise to iSMCs (Chalmers and Slack, 2000; Roberts et al., 1998). iSMC formation happens notably later than the medial migration and differentiation of other LPM-derived lineages, including the bilateral precursors for cardiovascular, hematopoietic and renal cell fates that functionally remodel prior to 24 hpf in zebrafish. The absence of obvious defects in the other LPM-derived lineages after TGFβ/zeb1a and foxo1a modulations suggests that these genes are active only in the iSMC-fated LPM population, or that compensatory mechanisms exist in other lineages. Curiously, the sole posterior phenotype of hand2 mutations in zebrafish is the lack of iSMCs, suggesting a dedicated role for hand2 in the posterior LPM stripe that is fated to form intestinal smooth muscle.

We identified TGFβ as a crucial regulator of LPM-to-iSMC differentiation that sustains LPM ventral migration around the endoderm. The TGFβ superfamily consists of several different protein families, including TGFβ proteins, bone morphogenetic proteins (BMPs), activins, Nodal and many others. Our data suggest that a key role in LPM-to-iSMC differentiation is played by the TGFβ type I receptor Alk5, which is targeted by both the inhibitors we used in this study (SB431542 and LY364947). In addition, previous work has also shown that ltbp3 inhibition phenocopies the effect of LY364947 treatments in zebrafish hearts (Zhou et al., 2011). Furthermore, chemical inhibition of BMP signaling does not affect iSMCs in zebrafish (Table S2), indicating once again a specific role for TGFβ proteins. However, more-detailed genetic studies are needed to understand the precise receptors and ligands involved in this process and to exclude the involvement of other signaling molecules.

Despite being a mesodermal tissue, LPM has been described as a polarized epithelium (before 30 hpf) by expression and apical localization of aPKC (Horne-Badovinac et al., 2003). We now show that markers of mesenchymalization (e.g. N-cadherin) are also already present at this developmental stages, questioning the nature of undifferentiated LPM as bona fide epithelium. Later on during development, LPM/hand2+ cells migrate around the gut to give rise to iSMC precursors (48 hpf) in a process that we found to be dependent on Alk5/TGFβ signaling. We reasoned that an important role for TGFβ/zeb1a could be to promote the acquisition of migratory phenotype for LPM. In particular, LPM migration could be driven by a TGFβ-induced partial EMT process. Indeed, unlike canonical EMT, which transforms epithelial layers into individual motile mesenchymal cells, LPM migrates as a cohesive layer of mesenchymal cells. The LPM thus retains at the same time epithelial features such as cell-cell contacts and a supracellular organization, and mesenchymal features such as migration and the ability of ECM remodeling (Yin et al., 2010).

Interestingly, we also found that the migration program in the differentiating LPM could be switched off by miR-145, a microRNA that has already been shown to modulate EMT acting as a tumor suppressor gene in other contexts. In particular, being able to directly bind the 3′ UTRs of oct4 and zeb2 transcripts, miR-145 has been considered as a regulator of invasion and stem cell properties in prostate and lung cancer (Hu et al., 2014; Ren et al., 2014). Our data show that miR-145 regulates iSMC development and differentiation in similar manner by regulating LPM migration and proliferation and homeostasis via zeb1a and foxo1a repression, respectively. miR-145 expression is controlled by TGFβ as master regulator of migration, invasion and EMT, and that miR-145 in turn represses several TGFβ downstream target genes. This interplay establishes an autoregulatory negative-feedback loop that spatiotemporally demarcates LPM migration. Other work showed that miR-145 regulates, and is regulated by, TGFβ signaling in other cell types (Long and Miano, 2011; Zhao et al., 2015), reinforcing the existence of such a feedback loop. Nonetheless, we noticed that miR-145 expression occurs later than initial TGFβ activation, suggesting the existence of a regulatory mechanism that keeps miR-145 transcriptionally silent until its action is needed. More-detailed insights are required into the genetic and epigenetic mechanisms of miR-145 transcriptional regulation in the smooth muscle field and cancer. Besides its role in cancer progression, miR-145 has been found as one of the most enriched miRNAs in vascular smooth muscle cells (vSMCs), where miR-145 is required for vSMC maturation and further regulation of their plasticity and contractility (Albinsson and Swärd, 2013; Boettger et al., 2009; Chivukula et al., 2014; Cordes et al., 2009; Elia et al., 2009; Xin et al., 2009). Many miR-145 target genes have been shown to be involved in these processes; yet, our newly found connection to zeb1a and foxo1a in iSMCs also suggests that these two novel players might be involved in the regulation of smooth muscle cell plasticity.

By analyzing the direct targets of miR-145, we identified Foxo1a as a potent and unforeseen player in intestinal smooth muscle differentiation. Forkhead box O (FOXO) transcription factors are involved in widespread regulation of the cell cycle, apoptosis and metabolism (Eijkelenboom and Burgering, 2013). Support for a role for Foxo1 in smooth muscle cell differentiation also arises from work on mesodermal precursor cells derived from mouse Foxo1^−/− embryonic stem cells (ESCs) that fail to form vascular smooth muscle cells (Park et al., 2009). In vitro ESC differentiation models revealed that Foxo1 activity plays a key role in progenitor cell and stem cell maintenance: Foxo1 is an essential component of the cellular control mechanism that maintains pluripotency in human embryonic stem cells (hESCs) through direct control of OCT4 and SOX2 gene expression by occupation and activation of their respective promoters (Zhang et al., 2011). In the same model system, Xu and co-workers have reported that expression of miR-145 is low in self-renewing hESCs but highly upregulated during differentiation via direct binding and repression of OCT4, SOX2 and KLF4 (Xu et al., 2009). Here, we demonstrate that foxo1a expression is enriched in the hand2+ zebrafish LPM and its absence impairs LPM patterning and differentiation. Furthermore, our data reveal that foxo1a overexpression maintains the undifferentiated/embryonic state of LPM as hand2-positive tissue. We propose a model where miR-145 expression is required to drive mesoderm lineage-restricted differentiation into SMCs by repressing expression of Foxo1. A role for foxo1a in endoderm-derived tissues is conceivable during development, although this function must be unrelated to its regulation by miRNA-145. Overall, we report here that Foxo1 is a direct target of miR-145, which in turn supports the previously unforeseen link between miR145 and stemness via Foxo1.

In summary, we have genetically established that the iSMCs are a cell fate of the LPM, and we have uncovered a new molecular pathway that promotes the coordinated cellular events that drive the LPM towards iSMC differentiation during vertebrate development (Fig. 7). In particular, we have found that miR-145, zeb1a and foxo1a are interconnected key players during iSMC differentiation in zebrafish. Our findings propose a new regulatory pathway through which TGFβ/Alk5 input commits the hand2-positive LPM stripes towards forming iSMC precursors by tuning a tissue-specific mesenchymalization process via zeb1a and miR-145 expression. In particular, miR-145 provides Alk5 signaling with a broadly acting tool to influence the downstream post-transcriptional dynamics of mesenchymalization. In parallel, we have identified foxo1a as an LPM-expressed gene involved in iSMC differentiation that is also regulated by the Alk5 and miR-145 signaling. Alteration in these developmental processes can result in genetic disorders, such as visceral myopathy. Our work provides a new molecular framework from which to analyze these molecular players for their prognostic and therapeutic potential in human gastrointestinal genetic diseases and cancers arising from dedifferentiated iSMCs (Spoelstra et al., 2006; Wangler et al., 2014; Yamamoto and Oda, 2015).

Zebrafish were handled according to established protocols and maintained under standard laboratory conditions. The Tg(hsp70l:Hsa.TGFBR1_T204D-HA,cryaa:Cerulean)^fb6Tg [referred to as Tg(hsp70:caALK5)], TgBAC(hand2:EGFP)^pd24, Tg(Xla.Eef1a1:GFP)^s854, TgBAC(cldn15la-GFP)^pd1034, Tg(-6.4drl:creERT2) and ubi:Switch lines have been described previously (Mosimann et al., 2015; Mosimann and Zon, 2011; Ober et al., 2006; Rohr et al., 2006; Yin et al., 2010; Zhou et al., 2011; Alvers et al., 2014). The generation of the Tg(acta2:mCherry)^uto5 and Tg(tagln:EGFP)^uto37 lines is described below. Following fertilization, embryos were collected and grown in the presence of 0.003% 1-phenyl-2-thiourea (PTU, Sigma-Aldrich) to prevent the formation of melanin pigment.

We analyzed the list of transcription factors represented by JASPAR positional weight matrices (Table S1). For acta2, the AVID alignment tool from VISTA has been used to directly align the region spanning from 2 kb upstream of the TSS to the end of the first intron of ACTA2 in zebrafish, human and mouse. We located the predicted binding sites in the D. rerio genome for the above-mentioned transcription factors using a log-likelihood ratio score, with the background nucleotide frequencies computed over the entire intergenic fraction of the D. rerio genome. The cutoff score was set to 66% of the best possible score for the PWM or an absolute score greater than 9. The Tol2-based acta2:mCherry and tagln:EGFP-CAAX constructs were assembled using the Tol2 Kit and a three-fragment gateway recombination cloning strategy (Kwan et al., 2007). For 5′ entry cloning, ∼350 bp of the acta2 promoter was amplified from the genomic DNA of wild-type zebrafish by PCR with the following primers containing appropriate attB4 and attB1r sites: 5′-GGGGACAACTTTGTATAGAAAAGTTGGCCATTCCTTCTCAGGTGTGG-3′ and 5′-GGGGACTGCTTTTTTGTACAAACTTGGGCACTTACCCTGACAGTGC-3′, respectively. The PCR product was then cloned into pDONRP4-P1R by BP reaction to obtain p5E-acta2. For middle entry cloning, the zebrafish acta2 first intron was amplified with the following primers containing appropriate attB1 and attB2 sites: 5′-GGGGACAAGTTTGTACAAAAAAGCAGGCTACCTAGCTTCTCTCACCTCC-3′ and 5′-GGGGACCACTTTGTACAAGAAAGCTGGGTTTTCAGCTCGGATATCCTTTCTTACTCC-3′, respectively, and cloned into pDONR221 by BP reaction. The 3′ entry clone was p3E-mCherrypA. Entry vectors were assembled in the pDestTol2pA2 vector by LR reaction to create the pDestTol2-acta2-mCherry-pA vector. For the tagln gene, ClustalW alignment was used to align the region spanning 2 kb upstream of the TSS of tagln in four different fish species (zebrafish, Tetraodon, stickleback and medaka). This multiple alignment was used as input to calculate the log-likelihood ratio score of the transcription factor binding represented by JASPAR positional weight matrices. The score cutoff was set to 50% of the best possible score for the PWM. For generation of the tagln:CAAX-EGFP construct, the 2 kb tagln promoter was amplified from the genomic DNA of wild-type zebrafish with the following primers containing appropriate attB4 and attB1 sites: 5′-GGGGACAACTTTGTATAGAAAAGTTGAGACGACAGAATAGAGAGGGCGGTGT-3′ and 5′-GGGGACTGCTTTTTTGTACAAACTTGCAGCAGCTTTATGTTCAGCACGG-3′, respectively. The PCR product was then cloned into pDONRP4-P1R by BP reaction to obtain p5E-tagln. pME-EGFP-CAAX was used as a middle element, and the 3′ element was p3E-polyA. Entry vectors were assembled with the vector pDestTol2pA2 by LR reaction to create the vector pDestTol2-tagln-EGFPCAAX-pA. The vectors were mixed with mRNA for Tol2 transposase and microinjected into one-cell stage wild-type embryos. Injected embryos were raised to adulthood, and founders were screened for red fluorescence in SMCs. The transgenic fish line names Tg(acta2:mCherry)^uto5 and Tg(tagln:CAAX-EGFP)^uto37 were approved by the Zebrafish Nomenclature Committee of the ZFIN (http://zfin.org).

Immunofluorescence was performed as previously described (Santoro et al., 2009). Briefly, embryos were fixed in 4% paraformaldehyde at 4°C overnight and washed three times in PBS. For immunofluorescence on sections, embryos were embedded in 4% low-melting agarose (Sigma-Aldrich). Sections (250 μm) were obtained using a vibratome (VT1000 S, Leica), permeabilized with 1% BSA, 1% DMSO and 0.3% Triton X-100 in PBS for 30 min at room temperature, and then incubated with primary antibody at 4°C overnight. After washing in PBS-T (0.1% Triton X-100 and 1% BSA in PBS), the sections were incubated with secondary antibodies (Alexa Fluor, Life Technologies) and Hoechst 33342 (Life Technologies) for 4 h at room temperature. The sections were washed in PBS-T, followed by PBS, then mounted on slides with Vectashield (Vector Labs). For whole-mount immunofluorescence, the fixed embryos were permeabilized in 1% DMSO and 1% Triton X-100 for 30 min at room temperature and then blocked in 4% BSA and 0.3% Triton X-100 in PBS for 4 h at room temperature. Embryos were incubated with the primary antibody at 4°C overnight, washed and incubated with secondary antibodies for 2 h at room temperature. After the washes, the embryos were embedded in 4% low-melting agarose and sectioned at the vibratome. The sections were mounted on slides with Vectashield. A polyclonal anti-transgelin antibody was produced using the C-terminal sequence (Santoro et al., 2009). For neuronal staining, the monoclonal antibody anti-Hu was used (1:50; mAB 16A11, Molecular Probes). For LPM staining, antibody anti-N-cadherin (1:200, Genetex) and aPKC (1:200, SantaCruz) were used. For actin staining, the sections were permeabilized and incubated with fluorescein isothiocyanate-labeled (1:1000 for 2 h at room temperature; Sigma-Aldrich) or tetramethylrhodamine B isothiocyanate-labeled (1:500 for 2 h at room temperature; Sigma-Aldrich) phalloidin after the washes.

Images were acquired with a TCSII SP5X confocal microscope, a MZ16 FA stereomicroscope equipped with a DCF300FY camera (Leica) or a AZ100 stereomicroscope equipped with an AxioCam MRm camera (Zeiss). The LAS AF and Zen software suites were used for analysis and image processing. Whole-embryo confocal images were acquired using the tile scan and automated mosaic merge functions of Leica LAS AF software. Digital micrograph images were contrast balanced, color matched, cropped and rotated using Photoshop 7 (Adobe).

Cell-tracing experiments were performed essentially as previously described (Felker et al., 2016; Mosimann and Zon, 2011). Briefly, embryos from Tg(-6.4drl:creERT2) (Mosimann et al., 2015) and ubi:Switch line intercross were treated with fresh 10 µM 4-OH tamoxifen (H7904, Sigma-Aldrich) in DMSO at the one-somite stage, with subsequent thorough washing of the embryos in untreated E3 medium at 24 hpf. At the indicated time points, embryos were fixed and processed for confocal analyses.

The in situ hybridization probes were designed with an oligonucleotide-based method. An oligonucleotide pair (including T7 promoter) was used to amplify target region (CDS or 3′UTR) from zebrafish cDNA, followed by in vitro transcription including DIG-labeled NTPs (Roche). Afterwards, RNA was precipitated with lithium chloride, washed with 75% ethanol and dissolved in DEPC water. RNA quality was checked on a MOPS gel. For the zeb1a in situ hybridization probe, the following primers were used: GAGGAGTGCGTCAGTGATGAGG and TAATACGACTCACTATAGGCAGGTGCTCCTTCAGGTGATGC (rev with T7). For the foxo1a in situ hybridization probe the following primers were used: GTGGAGCTAAATTGCAAGGACG and TAATACGACTCACTATAGGCGTGTAAACTCTCTGTACACCG (rev with T7).

Embryos were disaggregated into single cells as previously described (Mugoni et al., 2013). A FACSCalibur flow cytometer (BD Biosciences) and the Cell Quest software were used to measure the percentage of fluorescent cells. A FACS ARIA III sorter (BD Biosciences) was used to isolate single cells for subsequent RNA extraction.

Chemicals for zebrafish treatments were dissolved in DMSO. Zebrafish embryos were treated with the following drugs: SB431548 (50 μM; Sigma-Aldrich); AS1842856 (100 nM; Calbiochem); LY364947 (50 μM; Sigma-Aldrich); purmorphamine (10-100 μM; Calbiochem); cyclopamine (50 μM; Calbiochem); dorsomorphin (10-100 μM; Sigma-Aldrich); LDN193189 (250 nM-1 μM; Sigma-Aldrich); GM6001 (50-200 μM; Merck Millipore); SU1498 (5-100 μM; Calbiochem); SU5416 (10-100 μM; Sigma-Aldrich); L-NAME (100-500 μM; Sigma-Aldrich); SNAP (100-500 μM; Sigma-Aldrich); and PDGFR tyrosine kinase inhibitor V 521234 (1-100 μM; Calbiochem). The treatments were administered from 20 to 72 hpf. Chemicals were refreshed daily.

Gene knockdown experiments were performed by microinjecting morpholinos (Table S4) into zebrafish embryos at the one-cell stage. Morpholinos were synthetized from GeneTools and dissolved in nuclease-free water. The primers for testing the efficacy of the zeb1a morpholino were designed using the zebrafish zeb1a sequence (GenBank accession number: XM_001344071.6) and are as follows: zeb1a_ex2_Fw, 5′-GCGACCTCAGATTCAGATG-3′; zeb1a_ex3_Rv, 5′-TGACCCTTATTTCTCGTATTAAAG-3′; and zeb1a_in2_Rv, 5′-CTATGTGATTGTGCCTGATG-3′. The primers for testing knockdown by the foxo1a morpholino were designed for zebrafish foxo1a (GenBank accession number NM_001077257.2) and are as follows: foxo1a_ex2_Fw, 5′-GGGAAAAGTGGAAAGTCTCC-3′; foxo1a_ex3_Rv, 5′-TGTGTGGGTGAGAAAGAGTG-3′; and foxo1a_in2 _Rv, 5′-TGAATGTGGCCTGAATGAG-3′. As a control, β-actin was detected with the following primers: β-actin_Fw, 5′-GTATCCACGAGACCACCTTCA-3′; and β-actin_Rv, 5′-GAGGAGGGCAAAGTGGTAAAC-3′.

Heat-shock experiments on Tg(hsp70l:Hsa.TGFBR1_T204D-HA,cryaa:Cerulean)^fb6Tg were performed essentially as previously described (Zhou et al., 2011) by administering a 37°C heat shock for 1 h to transgenic and clutch mate controls. For miR-145 analyses, embryos were heat shocked at 48 hpf and 72 hpf, and RNA from the trunk was extracted after 6 and 24 h, respectively. For coding gene analyses, embryos were heat shocked at 48 hpf and RNA from trunk was extracted after 24 h.

Data from a previous study (Sartor et al., 2010) were analyzed to obtain a list of genes differentially expressed between A459 cells after 72 h of TGFβ induction and untreated cells. Using limma (Smyth, 2005) and a false discovery rate (FDR) of 0.01, 1725 upregulated probes and 1444 downregulated probes corresponding to 1010 and 981 unique genes, respectively, were obtained. Similarly, data from Li et al. (2007) were analyzed to obtain a list of genes differentially expressed between the mesenchymal and epithelial fractions of mouse intestine. Using limma and an FDR cutoff of 0.01, we found that 9272 probes were upregulated in the mesenchymal fraction and 3595 were downregulated, corresponding to 5380 and 2384 unique genes, respectively.

The miR-145 target predictions were based on the latest TargetScan release (6.2). In particular, we used the mouse orthologs of the human annotations for mouse predictions and the annotated zebrafish UTRs for zebrafish predictions (Ulitsky et al., 2012). Gene overlaps and comparisons between different species were based on the Homologene (build66) orthology database.

Embryos were anesthetized with 0.04 mg/ml tricaine (Sigma-Aldrich), mounted in 3% methyl cellulose (Sigma-Aldrich), and allowed to adapt for 5 min before recording. Each embryo was recorded for 1 min with an MZ16 FA stereomicroscope equipped with a DCF300FY camera (Leica). The frequency and amplitude of peristaltic movements were compared between controls and injected embryos. Forty embryos per group were analyzed in two independent experiments.

Luciferase reporter vectors containing the 3′ UTR of the indicated miR-145 target genes were generated by PCR amplification of the 3′ UTR from zebrafish genomic DNA and subsequent cloning into the Firefly luciferase reporter pMIR-REPORT vector (Ambion). When indicated, the 3′ UTRs were mutagenized or deleted at the miR-145 recognition site using the QuikChange Site-Directed Mutagenesis kit (Stratagene), according to the manufacturer's instructions with the primers listed below. A total of 5×10⁴ HEK293 cells was co-transfected with 50 ng of the pMIR-REPORT (Ambion) Firefly luciferase constructs containing the 3′ UTRs of the indicated miR-145 potential target genes and 20 ng of pRL-TK Renilla luciferase normalization control (Promega) using Lipofectamine 2000 (Invitrogen Life Technologies). Lysates were collected 48 h after transfection, and Firefly and Renilla luciferase activities were measured with a Dual-Luciferase Reporter System (Promega). The foxo1a 3′ UTR was amplified with the following primers: foxo1a_3′UTR_Fw, 5′-GTGGAGCTAAATTGCAAGGAC-3′; and foxo1a_3′UTR_Rv, 5′-TTAACCACGCCCCTCTTATG-3′. miR-145 binding sites were mutated in foxo1a 3′ UTR using the following primers: foxo1a_Mut1_Fw, 5′-GGGAAGAAGCCCGGGTGAGCGGGAATCGCTG-3′; foxo1a_Mut1_Rv, 5′-CAGCGATTCCCGCTCACCCGGGCTTCTTCCC-3′; foxo1a_Mut2_Fw, 5′-GTAAATCGGAGAGATCCCGGGTTCGACGTTTTTAC-3′; and foxo1a_Mut2_Rv, 5′-GTAAAAACGTCGAACCCGGGATCTCTCCGATTTAC-3′.

The zeb1a 3′ UTR was amplified with the following primers: zeb1a_3′UTR_Fw, 5′-CTTACAGGGGTGATTCTCATG-3′; and zeb1a_3′UTR_Rv, 5′-AACGACTGACACGTTACACAC-3′. miR-145 binding sites were deleted in the zeb1a 3′ UTR using the following primers: zeb1a_Mut1_Fw, 5′-CAAATTTATGCGTATTCCCGGGTGCTGCACGATATTGG-3′; zeb1a_Mut1_Rv, 5′-CCAATATCGTGCAGCACCCGGGAATACGCATAAATTTG-3′; zeb1a_Mut2_Fw, 5′-CTTTTCACAATCTTCAGTGTTTGTCATTTGATCCCGGGAGAGTTTCTCACGTGTTGTTTGATT-3′; and zeb1a_Mut2_Rv, 5′-AATCAAACAACACGTGAGAAACTCTCCCGGGATCAAATGACAAACACTGAAGATTGTGAAAAG-3′.

RNA was isolated with TRIzol reagent (Invitrogen Life Technologies), and cDNA was made with a RT High Capacity kit (Applied Biosystems), according to the manufacturer's protocol. qRT-PCR was performed with an ABI 7900HT Fast Real-Time PCR System (Applied Biosystems) using Platinum qPCR SuperMix-UDG with ROX (Invitrogen Life Technologies). The following genes were analyzed: acta2 (NM_212620.1); tagln (NM_001045467.1); myh11 (NM_001024448.1); foxa3 (NM_131299.1); foxo1a (NM_001077257.2); zeb1a (XM_001344071.6); hand2 (NM_131626.2); E-cadherin (NM_131820.1); N-cadherin (NM_131081.2); occludin A (NM_212832.2); twist1a (NM_130984.2); twist1b (NM_001017820.1); snai1a (NM_131066.1); snai1b (NM_130989.3); and snai2 (NM_001008581.1). The β-actin gene (actb) was included as a control housekeeping gene (NM_131031.1 and NM_181601.4). Specific primers were designed with the dedicated UPL on-line tool (Roche) and are provided in Table S5. Data were analyzed using the ΔΔCt method with ABI software, version 2.1 (Applied Biosystems). For microRNA analyses, RNA was extracted using the TRIzol reagent (Invitrogen Life Technologies). qRT-PCR for microRNA detection was performed with the indicated TaqMan microRNA assays (Applied Biosystems) on 10 ng of total RNA according to the manufacturer's instructions. qRT-PCR was conducted using gene-specific primers on a 7900HT Fast Real-Time PCR System (Applied Biosystems). Quantitative normalization was performed for the expression of the RNU6 small nucleolar RNA. Data analysis was performed using the ΔΔCt method with the ABI software, version 2.1 (Applied Biosystems).

Total RNA (20 μg) isolated as above was resolved by 12.5% (w/v) TBE-urea-polyacrylamide gel electrophoresis and transferred to a Hybond N+ membrane (GE Healthcare Life Sciences). The filter was hybridized overnight at 45°C with a specific miR-145 digoxigenin-labeled LNA detection probe (Exiqon), washed and visualized with a specific DIG antibody (1:10,000) using the DIG Nucleic Acid Detection kit (all from Roche). The filter was then stripped and re-probed overnight at 45°C using a specific U6 digoxigenin-labeled LNA detection probe (Exiqon).

Phosphohistone H3 (Ser10, Cell Signaling) immunofluorescence was used to evaluate cell proliferation. The staining was performed on cross-sections of the gut of Tg(hand2:EGFP)^pd24 at 48 hpf. Ph3/hand2 double-positive cells and hand2 single-positive cells were counted in a minimum of three distinct sections per embryo in eight individual animals. The ratios are represented normalized to controls.

The complete zebrafish foxo1a CDS was amplified by PCR from cDNA using the primers: foxo1a_Fw, 5′-GTACCATGGCTGACGCAG-3′ and foxo1a_Rv, 5′-CTACCCAGACACCCAGCTG-3′. Purified PCR product was cloned in pCS2+ vector. foxo1a mRNA was synthesized using the mMessage Machine kit (Ambion) following the manufacturer's instructions. Wild-type embryos were injected at the one-cell stage with 100 pg of foxo1a mRNA. We also included a control mRNA encoding the fluorescent protein mCherry (100 pg) in each injection.

All experiments were performed at least three independent times for each condition, and the error bars represent the mean±s.d. of the mean unless otherwise stated. Statistical significance was evaluated by Student's test or one-way ANOVA-Dunnett's post-hoc test as appropriate, and significance is reported as *P<0.05, **P<0.01, ***P<0.001 and ****P<0.0001.

We thank Vanessa Barone for cloning the acta2 promoter and generating the Tg(acta2:mCherry)^uto5 line, Dr Caroline E. Burns for providing the Tg(hsp70:caALK5) line, Dr M. Bagnat for providing Tg(claudin:GFP) line, Anastasia Felker for assistance with genetic lineage tracing, Xiaowen Chen for assistance with rescue experiments and members of the Santoro lab for critical reading of the manuscript.