Rho signalling restriction by the RhoGAP Stard13 integrates growth and morphogenesis in the pancreas

Coordination of tissue growth and morphogenesis is crucial for generating adult functional organs (Hogan and Kolodziej, 2002; Lu and Werb, 2008). How the two processes are integrated during the development of many organs is still a major unresolved question. During pancreatic development, branching morphogenesis coincides with growth and differentiation (Jensen, 2004; Pan and Wright, 2011; Puri and Hebrok, 2010; Spagnoli, 2007). As different pancreatic cell types become specified, they organise themselves into discrete domains along the axis of the branches, delineating, for instance, a multipotent progenitor cell domain at the distal tips of the branching epithelium (Zhou et al., 2007). Importantly, recent observations have shown that the initial number of these progenitors pre-determines the final pancreas organ size (Stanger et al., 2007). Thus, establishment of a proper progenitor pool size during branching phase [between embryonic stage (E)12 and E14 in the mouse embryo] is crucial for normal pancreas formation and function, including digestion and blood sugar regulation. However, the molecular mechanisms underlying the coordinated action of branching and progenitor proliferation in the developing pancreas are largely unknown.

Small Rho GTPases play a fundamental role in the control of morphogenesis and cell proliferation of several epithelial organs (Van Aelst and Symons, 2002; Etienne-Manneville and Hall, 2002; Kesavan et al., 2009). GTPases act as molecular switches, cycling between active GTP-bound and inactive GDP-bound states, a process that is tightly regulated by distinct classes of proteins, including the Rho GTPases-activating proteins (GAPs) (Tcherkezian and Lamarche-Vane, 2007). Specifically, GAPs inactivate GTPases by accelerating their intrinsic GTPase activity thereby converting them into the inactive GDP-bound form (Sordella et al., 2003; Tcherkezian and Lamarche-Vane, 2007). Because of the ubiquitous distribution of small GTPases, a restricted expression of GAPs appears to be fundamental for the precise spatial-temporal regulation of their activity.

Here, we have identified the RhoGAP Stard13 [also called Shirin in Xenopus embryos (Spagnoli and Brivanlou, 2006) and deleted liver cancer (DLC) 2 in humans (Leung et al., 2005)], as a tissue-specific GTPase regulator of mammalian pancreas development. Genetic ablation of Stard13 disrupts epithelial branching and distal tip-domain morphogenesis, resulting in hampered proliferation of pancreatic progenitors and subsequent organ hypoplasia. We show that Stard13 acts by regulating Rho signalling spatially and temporally during pancreas development. Finally, our results suggest a reciprocal interaction between the Rho-actin and mitogen-activated protein kinase (MAPK) signalling pathways to regulate progenitor cell proliferation locally. Together, these results identify Stard13 as a molecular integrator of growth and morphogenesis that acts by restricting Rho-actin activity in the pancreas.

A gene-targeting vector for creating a floxed allele of the mouse Stard13 gene was generated using bacterial homologous recombination. The standard two-loxP site strategy was adapted to the use of bacterial artificial chromosomes (BACs) as targeting vectors (Valenzuela et al., 2003). The BAC-containing Escherichia coli carrying the entire Stard13 gene locus (BAC clone RP23-11K10) was identified by the alignment of BAC end sequences in the NCBI database. Chimeric mice were generated by injection of four independently targeted embryonic stem (ES) cells into blastocysts, and the allele was passed through the germline. The Stard13^{flox(Hygro)/+} mice were crossed to a germline Flp-deleter mouse strain (Rodriguez et al., 2000) to remove the Hygro cassette. Removal of the Hygro cassette was confirmed by PCR and Southern blot (data not shown). Further information on the targeting vector, targeting of ES cells, generation of Stard13^flox mice and genotyping can be found in supplementary material Fig. S1. Stard13^flox/flox mice were interbred with CMV-Cre mice (Schwenk et al., 1995) to generate mice carrying a germline-deleted allele of Stard13 (Stard13^Δ; supplementary material Fig. S1) and with Pdx1-Cre transgenic mice for conditional ablation in the pancreas (Gu et al., 2002). All animal experimentation was conducted in accordance with the local ethics committee for animal care.

For RNA isolation, embryonic tissues were dissected and snap-frozen on dry ice. Subsequently, RNA was extracted with RNAzol (Biozol) according to manufacturer’s instructions and total RNA was used for reverse transcription using random hexamers and oligodT and the SuperScript III First-Strand Synthesis System (Invitrogen) according to instructions. Real-time PCR reactions were carried out using the SYBR Green Master Mix (Roche) on ABI StepOne Plus system. Succinate dehydrogenase (SDHA) and ribosomal protein 36B4 were used as reference genes. All the values were normalised to the reference genes and calculated using the software REST (Pfaffl et al., 2002). Data were determined in triplets. See supplementary material Table S1 for primer sequences.

Mouse embryos and pancreata were fixed in 4% paraformaldehyde at 4°C from 2 hours to overnight. Subsequently, samples were equilibrated in 20% sucrose solution and embedded in OCT compound (Sakura). In situ hybridisation on cryostat sections was carried out as described by Schaeren-Wiemers and Gerfin-Moser (Schaeren-Wiemers and Gerfin-Moser, 1993). Cryosections (10 μm) were incubated with TSA (Perkin Elmer) blocking buffer for 1 hour at room temperature and afterwards with primary antibodies at the appropriate dilution (supplementary material Table S2). Immunostainings were analysed with Zeiss AxioObserver, Zeiss LSM 700 and LeicaSPE laser scanning microscopes. For counting, pancreatic tissue of at least three wild-type (WT) and three Stard13^PA-deleted embryos were cut into serial sections and stained cells were counted on every third sections. E-cadherin (cadherin 1)⁺ pancreatic epithelium was measured using AxioVision software (Zeiss). Quantification of immunohistochemical markers focused on the dorsal pancreas. Quantification of the fluorescence intensity of the F-actin staining was measured using ImageJ software on confocal images. The ROI tool was used to measure the Integrated Density values (sum of the values of the pixels in the selection) in Pdx1⁺ area of 5-10 optical sections from a minimum of three different experiments. For morphometric analyses, E18.5 pancreata were fixed in 4% paraformaldehyde and paraffin embedded. Each pancreas was sectioned at three different levels and for each level three sections (4 μm) were collected on a slide. Total pancreatic area identified was quantified using a Scan Scope microscope, analysed by Image Scope viewer (Aperio Technologies, USA) and expressed in μm². The average cell surface was determined on at least five pancreata for each genotype. All results are expressed as mean±s.e.m. and significance of differences between groups was evaluated with Student’s t-test.

E12.5 pancreata were fixed in phosphate-buffered 2.5% glutaraldehyde for 4 hours at 4°C and postfixed in 1% OsO₄. Samples were dehydrated and embedded in epoxy resin. Thin sections were counterstained with uranil acetate and lead citrate and examined with a Hitachi H-7100FA microscope.

Dorsal pancreatic buds were microdissected from mouse embryos at E11.5 and cultured on glass-bottom dishes (Matek) pre-coated with 50 μg/ml sterile bovine fibronectin (Invitrogen) in Basal Medium Eagle (BME) supplemented with 10% foetal bovine serum (Horb and Slack, 2000; Petzold and Spagnoli, 2012). Cultures were maintained for up to 6 days at 37°C in 5% CO₂. The day of plating is referred to as day 0. For the Rho inhibition assay, membrane-permeable C3 transferase (Cytoskeleton) was added at a final concentration of 2.5 μg/ml to the culture medium on day 1 and replaced every 24 hours. The Rho/SRF pathway inhibitor CCG-1423 (Millipore) was added at the final concentration of 5 μM to the culture medium on day 1 and replaced every 24 hours. In the extracellular signal-regulated kinase (ERK) inhibition assay, PD0325901 (Selleck) was added at a final concentration of 2 μM. Explants were fixed in 4% paraformaldehyde, stained as whole-mounts and analysed using a Zeiss LSM 700 confocal laser scanning microscope. For phospho-histone H3 (pHH3) counting, the E-cadherin⁺ volume of pancreatic explants of at least three WT and three Stard13^PA-deleted embryos was measured on acquired confocal z-stacks using Huygens software (Scientific Volume Imaging).

For Rho-GTP pull-down assay, dissected E17.5 pancreata were snap frozen in liquid nitrogen. After genotyping, pancreata were lysed and ∼300 μg total protein extract was incubated with 25 μg Rhotekin-RDB beads according to manufacturer’s instructions (Cytoskeleton). One-tenth of the total lysate (∼25-30 μg) was used for total Rho western blot. Immunoblots were incubated with anti-Rho antibody (1:500; Cytoskeleton) and analysed using the LI-COR Odyssey system.

Glucose tolerance test was carried out on three-month-old male animals that had been fasting overnight, before the day of experimentation (time point –1). Animals were injected intraperitoneally with glucose (2 g/kg body weight). Glucose levels were measured from blood collected from the tail immediately before the glucose challenge (time point 0) and 15, 30, 60 and 120 minutes after the glucose injection using a blood glucose meter (Contour, Bayer).

All results are expressed as mean ± s.e.m. The significance of differences between groups was evaluated with Student’s t-test. P<0.05 was considered statistically significant.

In situ hybridisation analysis revealed an early expression of Stard13 in the developing pancreas from E10.5 onwards in the mouse embryo (Fig. 1A). Notably, at the onset of branching, Stard13 showed a regionalised expression pattern, being enriched at the distal tips of the epithelial branches (Fig. 1A), which are known to contain multipotent progenitors capable of generating all pancreatic cell types (Horb and Slack, 2000; Zhou et al., 2007).

To examine whether Stard13 regulates pancreas morphogenesis and tip-domain organisation, we generated a conditional Stard13 (Stard13^flox) mutant allele (supplementary material Fig. S1) and intercrossed these mice with Pdx1-cre transgenic mice for conditional gene ablation in all pancreatic cell types during embryonic development (Gu et al., 2002). In wild-type (WT) pancreas, branching morphogenesis started normally and well-formed primary branches were visible from E12.5 onwards (Fig. 1B). By contrast, in Stard13^flox/flox; Pdx1-Cre [from here on referred to as Stard13^PA-deleted (Stard13-pancreas deleted)] embryos, the pancreatic tissue failed to form branching structures, displaying a globular shape with internal cavities, though pancreatic fate was normally specified, as judged by Pdx1 expression (Fig. 1B). Because the Stard13 transcript was enriched at the distal tips of the branching epithelium, we investigated whether its ablation affects the typical tip-domain organisation of the pancreatic epithelium (Zhou et al., 2007). To this aim, we analysed the expression of carboxypeptidase A1 (Cpa1), which is confined to WT tip domains and marks tip progenitor cells between E12.5 and E14.5 (Zhou et al., 2007) (Fig. 1B). Cpa1⁺ cells were also detected in Stard13^PA-deleted pancreas, but they displayed random distribution throughout the tissue, and tip structures were not properly formed (Fig. 1B). Subsequently, at E16.5, mature pancreatic tubular networks and acinar structures were evident in the WT, whereas Stard13^PA-deleted ducts and acini appeared as malformed large cellular aggregates (Fig. 1B). These results indicate that Stard13 controls morphogenesis and establishment of proper tip domain within the pancreatic branches during embryonic development.

In the pancreas, organ growth relies on proliferation of progenitor cells and coincides with the time of branching morphogenesis (Seymour et al., 2007; Stanger et al., 2007). Indeed, at the onset of branching, between E11.5 and E12.5, proliferation rate undergoes a 20-fold increase in the WT pancreas epithelium (our unpublished data). We therefore examined whether aberrant morphogenesis in Stard13^PA-deleted embryonic pancreas affects pancreas growth. Gross examination of neonatal pancreata revealed severe pancreatic hypoplasia in Stard13^PA-deleted mice, which was also assessed by morphometry (Fig. 1C,D; supplementary material Fig. S1). Postnatally, Stard13^PA-deleted animals were growth retarded and displayed impaired glucose tolerance (supplementary material Fig. S2). We found that Stard13 was expressed not only throughout pancreatic development but also in the endocrine islets after birth (Fig. 1; data not shown). We performed a glucose tolerance test (GTT) on WT and mutant animals of same age. Although basal blood glucose levels appeared normal, the curve was characterised by a slower and delayed return to basal levels after the glucose challenges in Stard13^PA-deleted animals. These results suggest a role for the RhoGAP Stard13 in β-cell function during adulthood.

We then explored whether alterations in cell proliferation and/or apoptosis during embryogenesis are responsible for pancreatic size reduction in Stard13^PA-deleted mice. Very few apoptotic cells were detected within normal pancreatic epithelium during embryonic stages E12.5-16.5 and apoptosis was not increased in the pancreas of Stard13^PA-deleted embryos, as determined by TUNEL assay and caspase 3 immunostaining (data not shown). By contrast, immunohistochemical measurements of proliferation using the mitosis-specific marker phospho-histone H3 (pHH3) revealed a 27% average decrease in cell proliferation in Stard13^PA-deleted E12.5 pancreas compared with WT (Fig. 1E). Importantly, reduced cell proliferation was observed at E12.5 in the mutant pancreas compared with WT, whereas there was no difference between mutant and WT pancreas at E11.5 (Fig. 1E).

To examine whether Stard13 ablation specifically affects proliferation of progenitor cells, we analysed the percentage of proliferating cells that express the progenitor marker Cpa1. In the Stard13^PA-deleted pancreatic epithelium, a strong reduction in the number of pHH3⁺Cpa1⁺ pancreatic cells was observed (40% reduction; Fig. 1F), whereas the percentage of pHH3⁺Cpa1^– cells was similar between WT and mutant pancreata (only 13% reduction; Fig. 1F). Similar results were obtained upon bromodeoxyuridine (BrdU) in vivo labelling at E12.5 and quantification of double BrdU⁺Cpa1⁺ cells (data not shown). These measurements indicated that the 27% proliferation reduction is mainly due to reduced cell proliferation of the Cpa1⁺ cell compartment (Fig. 1F). Consistent with their decreased proliferative activity, Cpa1⁺ cells were reduced in Stard13^PA-deleted embryos at later stages (supplementary material Fig. S3J). Moreover, we found that ablation of Stard13 did not affect specification of any pancreatic cell type in particular and that the relative numbers of differentiated cell types (e.g. endocrine and exocrine cells between E12.5 and E16.5) were similar in both WT and Stard13^PA-deleted embryos (supplementary material Fig. S3), ruling out the possibility of accelerated differentiation of progenitor cells. Thus, our finding indicates that Stard13 ablation hampers proliferation of pancreatic cells and, in particular, the Cpa1⁺ progenitor pool at early embryonic stages, ultimately resulting in organ hypoplasia. Moreover, these results suggest that Stard13 control on morphogenesis is required to maintain proliferating Cpa1⁺ cells at the branching stage and allows their expansion.

We next sought to gain insight into the cellular defects at the origin of altered pancreas morphogenesis in the absence of the RhoGAP Stard13. Branching morphogenesis involves the restructuring of the pancreatic epithelium into a complex and highly organised tubular network, starting with the transition from a non-polarised cell mass to polarised epithelial monolayers between E11.5 and E12.5 in the mouse embryo (Hick et al., 2009; Jensen, 2004; Kesavan et al., 2009; Villasenor et al., 2010). We examined cell morphology and epithelial polarity, including cell-cell and cell-extracellular matrix (ECM) adhesions, and cytoskeleton organisation. In E12.5 WT pancreas, epithelial cells displayed a columnar polarised shape with basal nuclei and constricted apical pole and were radially oriented around common lumens to form branched tubular structures (Fig. 2A-C). Polarised cells showed E-cadherin confined to the basolateral membrane, F-actin, mucin 1 and atypical protein kinase C (aPKC) isoform ζ at the apical surface, and laminin and integrins at the basal lamina (Fig. 2A,B; Fig. 5; supplementary material Fig. S4). Notably, the Stard13^PA-deleted epithelium did not undergo similar epithelial remodelling: the epithelium stayed stratified, displaying cells cuboidal in shape that were arranged in a disorderly fashion and randomly oriented to surround microlumens (Fig. 2A-C). Ultrastructural analysis corroborated the clear differences in epithelial organisation between WT and Stard13^PA-deleted epithelia (Fig. 2C). Moreover, quantitative measurement of cell outlines defined an increased circularity index in the mutant compared with WT (Fig. 2D), which is consistent with absence or delay of cell-shape change (from cuboidal to columnar) in Stard13^PA-deleted epithelium. Mutant epithelial cells characterised by a pronounced rounded shape were evident already at E11.5 (Fig. 2D).

Laminin and integrins (α3-, α6- and β1-integrins) exhibited unchanged levels and normal basal distribution in the Stard13^PA-deleted epithelium relative to WT (Fig. 2A; Fig. 5). By contrast, we found that the major cytoskeletal components, such as F-actin and myosin (activated phospho-myosin II), not only accumulated to high levels but also were irregularly distributed throughout the cytoplasm in mutant cells (Fig. 2B,E). Polarised distribution of the actomyosin network normally forms cable-like structures that span multiple cells and their coordinated apical contraction results in multicellular structures, which are named rosettes in various epithelial tissues undergoing morphogenesis (Zallen and Blankenship, 2008). Upon closer analysis of E11.5 and E12.5 WT pancreas epithelium, we detected similar multicellular ‘rosette-like’ aggregates that displayed asymmetric enrichment of F-actin and phospho-myosin II on the apical side and that constricted their shared interfaces to form microlumens (Fig. 2E). By contrast, Stard13^PA-deleted epithelial cells occasionally clustered together at a common interface, lacking localised actomyosin distribution and coordinated apical constriction (Fig. 2E). Consequently, fewer higher-order rosettes were formed and did not properly resolve (Fig. 2E).

Collectively, these results suggest alterations in actomyosin cytoskeletal organisation as the source of defective pancreatic epithelium remodelling, including defects in cell-shape changes, cell arrangement and microlumen connection, in Stard13 mutants.

Although Stard13 has been shown to have GAP activity for RhoA in vitro (Leung et al., 2005), its activity in vivo has remained unknown. To elucidate the molecular mechanism of Stard13 function, we tested whether it regulates Rho and whether its ablation leads to elevated levels of active Rho in the developing pancreas. We performed two independent assays that are based on the use of the Rho-binding domain of the Rho effector protein rhotekin fused to glutathione-s-transferase (RDB-GST) as substrate (Malliri et al., 2002). Notably, these assays are designed to detect specifically the Rho proteins in their active GTP-bound conformation and not merely their expression. First, active GTP-bound Rho was detected by RDB-GST pull-down assay in Stard13^PA-deleted embryonic pancreas, whereas no detectable amounts were found in WT pancreas (Fig. 3A). Second, to visualise GTP-bound Rho proteins we used an immunolocalisation assay (Cascone et al., 2003) on cultures of E11.5 pancreas (Fig. 3C) (Horb and Slack, 2000; Petzold and Spagnoli, 2012; Puri and Hebrok, 2007). Indeed, cultured WT pancreas explants recapitulated in vivo early pancreatic morphogenetic and differentiation events, providing a valuable ex vivo model to analyse branching and tubulogenesis (Hick et al., 2009; Horb and Slack, 2000; Kesavan et al., 2009) (Fig. 3B; supplementary material Fig. S5). Strikingly, Stard13^PA-deleted pancreatic explants cultured in the same conditions as WT displayed smaller size and failed to form branching structures, reproducing ex vivo the phenotype observed in vivo in the mutant pancreas (Fig. 1; Fig. 3B), and showed clusters of cells exhibiting active Rho (Fig. 3C). Taken together, these results indicate that the RhoGAP Stard13 is required to restrain the levels of active Rho in the developing pancreas.

Because ablation of Stard13 resulted in noticeable active Rho in the pancreas, we reasoned that constraining Rho activity might rescue tissue architecture in Stard13^PA-deleted pancreas. To address this possibility, we took advantage of the established pancreatic explant culture system and performed pharmacological inhibition of Rho signalling using a membrane-permeable version of the enzyme C3 ribosyltransferase (hereafter termed C3), which inactivates all Rho proteins, but not Cdc42 or Rac (Bishop and Hall, 2000). Upon exposure to C3, Stard13^PA-deleted pancreatic explants started to form tubular aggregates that branched and expanded in overall size compared with untreated Stard13^PA-deleted explants (Fig. 3D), implying that specific Rho inhibition is able to rescue both epithelial and proliferation defects, to some extent. In line with this, we also observed that treatment of WT pancreatic explants with lysophosphatidic acid (LPA), an activator of endogenous Rho (Malliri et al., 2002), led to rudimentary aggregates, which mimicked the Stard13^PA-deleted phenotype (Fig. 4; supplementary material Fig. S5).

In the absence of Stard13, we found accumulation of and altered distribution of F-actin and activated myosin II in vivo (Fig. 2B,E) as well as ex vivo in pancreatic explants (Fig. 4A). F-actin content started to increase at E11.5 and progressively accumulated over time, as judged by fluorescence intensity measurements (Fig. 4B). To address whether uninhibited Rho activity is responsible for the aberrant actomyosin network in Stard13^PA-deleted epithelium, we examined F-actin distribution in explant cultures that were treated with C3. Importantly, reduction of active Rho levels in C3-treated Stard13 mutant pancreas organ cultures not only rescued branching morphogenesis but also restored asymmetric F-actin distribution at the apical cell surface (Fig. 4A) and F-actin levels to near those of WT (Fig. 4C). By contrast, in mutant and LPA-treated WT explants F-actin fluorescence intensity increased and its distinct apical distribution was lost, being detectable all around the cell periphery, similar to the Stard13 mutant in vivo phenotype (Fig. 2; Fig. 4A-C). Accumulation and aberrant F-actin network distribution are consistent with the abnormal apical width and rounded cell shape of the mutant epithelial cells (Kuure et al., 2010) (Fig. 2). Taken together, these findings indicate that Stard13 controls epithelial remodelling and morphogenesis by regulating the Rho/actin signalling axis within the pancreatic epithelium.

In branching morphogenesis, surrounding mesenchymal cues are known to play a crucial role in promoting cell proliferation in the epithelium and guiding the direction of the outgrowth (Hogan and Kolodziej, 2002; Horb and Slack, 2000; Lu and Werb, 2008; Metzger et al., 2008; Watanabe and Costantini, 2004). Once pancreatic tip domains are formed, they possibly require a special supporting ‘environment’ for fast-proliferating progenitors to expand. Previous studies have shown that MAPK signalling factors, such as epidermal growth factor (EGF) and fibroblast growth factor (FGF) molecules, are crucial mediators of epithelial-mesenchymal interaction during pancreatic development (Bhushan et al., 2001; Edlund, 2002). In order to determine whether MAPK signalling might play a role in sustaining pancreatic local proliferation, for example at the branch tips, we tested the levels of phosphorylated (active form) p44/p42 MAPK (here referred to as ERK1/2; also known as Mapk3/Mapk1). Importantly, we found that WT pancreatic cells located at the branch tips displayed significantly higher levels of pERK1/2 than did cells in the trunk both in vivo and ex vivo in organ cultures (Fig. 5A,B). To address whether this differential distribution of pERK1/2 is required for the growth of epithelial branches, we pharmacologically inhibited the kinase activities of ERKs by using the selective compound PD0325901 (Bain et al., 2007) in WT pancreatic explants (Fig. 5C). Strikingly, samples cultured for two days in the presence of PD0325901 were smaller than untreated WT pancreatic epithelium, and displayed a smooth surface with no signs of branching (Fig. 5B,C).

Accordingly, we detected a reduction in cell proliferation in PD0325901-treated WT pancreatic cultures, exhibiting 50% and 60% decrease in the average number of pHH3⁺ cells after 24-hour and 48-hour exposure, respectively (Fig. 5F). However, these defects were not accompanied by obvious and measurable changes in F-actin cytoskeleton organisation (data not shown), suggesting that ERK controls primarily proliferation.

Because epithelial tip structures did not form in the absence of Stard13, we investigated next the possibility that Stard13^PA-deleted pancreatic cells display reduced pERK activity or loss of its spatial distribution. Both immunostaining and western blot analyses showed that the levels of pERK1/2 proteins were substantially reduced in embryonic pancreas upon Stard13 ablation, whereas total ERK, PI3K-Akt and FAK (Ptk2) pathways were unaffected (Fig. 5D,G,H). Importantly, specific inhibition of Rho by C3 rescued branch-tip formation in Stard13^PA-deleted pancreatic explants, restoring pERK activation and cell proliferation to levels similar to WT (Fig. 3; Fig. 5E-G). These results suggest that proliferation of progenitors at the newly formed tips is supported by local activation of the ERK signalling.

The integrin-FAK signalling axis is known for regulating proliferation through the control of ERK (MAPK) pathway (Legate et al., 2009). However, the levels of FAK-Y397 phosphorylation, integrin signalling and, in general, cell-ECM adhesion were unchanged in the absence of Stard13 in the pancreatic epithelium (Fig. 5H-J). Thus, this mechanism does not appear to contribute to ERK regulation in pancreatic progenitor proliferation in the absence of Stard13.

Next, we sought to establish whether cell-shape changes and actin cytoskeleton might directly influence proliferation in the pancreatic epithelium. Actin polymerisation and F-actin filaments accumulation can control gene expression through regulation of MAL/SRF transcriptional activity (Descot et al., 2009; Posern and Treisman, 2006). Strikingly, we found induction of several MAL/SRF downstream effectors, including Srf, Ctgf and Vcl, in the Stard13^PA-deleted embryonic pancreas (Fig. 6A). By contrast, mesenchymal MAL/SRF targets, such as Acta2, or TCF-dependent SRF targets, such as Fos, were not detectable (Descot et al., 2009) (data not shown). Among the induced MAL/SRF targets, we found some with known anti-proliferative function, such as Mig6 (also know as Erffi1) and Zfp36 (Descot et al., 2009). Intriguingly, both Mig6 and Zfp36 are known negative regulators of EGF/MAPK signalling, acting at different levels of the cascade (Descot et al., 2009; Ferby et al., 2006), and they were expressed at very low, almost undetectable, levels in WT embryonic pancreas, but readily accumulated in mutant cells (Fig. 6A,B). To address directly whether SRF transcriptional activation downstream of Rho and actin polymerisation is responsible of the proliferation defects in the Stard13^PA-deleted pancreatic epithelium, we inhibited Rho transcriptional signalling using the Rho/SRF pathway inhibitor CCG-1423 (Evelyn et al., 2007). Treatment of Stard13^PA-deleted pancreatic explants with the compound CCG-1423 partially restored the formation of primary branches that expanded in overall size compared with untreated Stard13^PA-deleted explants (Fig. 6C,D). These results imply that selective disruption of Rho/SRF transcriptional activity is able to rescue proliferation defects in the absence of the RhoGAP Stard13, even though its effect is more modest than that of upstream functional inhibition of Rho signalling, for instance through C3-ribosyltransferase treatment (Figs 3, 4, 5).

As Ctgf is a common target for activated MAL/SRF and Hippo/Yap signalling pathways and as Yap is modulated by high tissue stiffness and by Rho activity (Dupont et al., 2011; Wada et al., 2011), we also examined the activity status and subcellular localisation of the two mammalian components of the Hippo organ-growth pathway, Yap and phosphorylated Yap (p-Yap), in the embryonic pancreas. Both Yap and p-Yap levels and distribution were unchanged in the Stard13^PA-deleted tissue compared with WT (Fig. 6E). These observations indicate that, despite active Rho and F-actin fibre accumulation in the absence of Stard13, there is no Yap modulation in the mutant pancreatic epithelium, in contrast to what has been reported in mesenchymal cells (Dupont et al., 2011; Wada et al., 2011). By contrast, our data suggest that F-actin accumulation converges into the MAPK cascade through SRF targets, providing a mechanistic explanation for the reduced pERK levels in Stard13^PA-deleted pancreatic epithelium (Fig. 6).

Collectively, our findings indicate that the RhoGAP Stard13 temporally and spatially regulates Rho activity in the developing pancreas to ensure that morphogenesis and establishment of tissue architecture are coordinated with organ growth (Fig. 6F).

Precise spatiotemporal regulation of Rho activity is a prerequisite for proper assembly and tension of the actomyosin cytoskeleton (Van Aelst and Symons, 2002; Etienne-Manneville and Hall, 2002). Accordingly, uninhibited Rho activity in Stard13^PA-deleted embryos has a profound impact on the organisation of the actomyosin network, which in turn hampers epithelial remodelling events, including cell-shape changes, cell arrangement in rosette-like structures and microlumen connection, in the developing pancreas. These results unveil a novel role for actin cytoskeletal dynamics at the onset of pancreatic branching morphogenesis. Hence, it is likely that Rho-dependent actomyosin contraction plays distinct roles regulating (1) concerted apical constriction and (2) subsequent resolution of the ‘rosette-like’ structures into elongating tubules. The formation of multicellular rosettes provides an efficient mechanism for rearrangement of cells into a single epithelial layer (Zallen and Blankenship, 2008). Similarly to other elongating epithelia, we have found multicellular ‘rosette-like’ structures in the pancreatic epithelium at the time at which epithelial remodelling starts, suggesting that rosette arrangements might contribute to the transition from stratified to monolayered epithelium in the pancreas too.

Our findings suggest that Rho activity in the developing pancreas is unique and not redundant with Cdc42 GTPase, which is instead required for initiating microlumen formation and apical-basolateral cell polarity (Kesavan et al., 2009). Establishment of cell polarity is regulated by proper sorting and delivery of proteins to different membranes and by conserved signalling complexes, such as the Par, Crumbs and Scribble complexes (Nelson, 2009). Cdc42 is known to establish a functional and mature apical surface by interacting with the Par3-Par6-aPKC polarity complex protein (Van Aelst and Symons, 2002; Etienne-Manneville and Hall, 2002; Kesavan et al., 2009). Upon Cdc42 ablation, pancreatic epithelium fails to generate multicellular common apical surfaces, forming instead autocellular lumens, ultimately resulting in a fragmented epithelium without tubes (Kesavan et al., 2009). By contrast, misregulation of Rho signalling in Stard13^PA-deleted pancreas does not result in autocellular lumen formation, and apical-basolateral (e.g. PKCζ/mucin/laminin) polarity is established in Stard13 mutant epithelium, despite the morphogenetic defects. Taken together, these findings suggest an opposite activity of Rho and Cdc42 GTPases in the developing pancreas that might account for the different pancreatic phenotypes of Cdc42 and Stard13 mutants: fragmented epithelium in the former and globular stratified epithelium in the latter. However, this does not exclude potential crosstalk among different GTPases during pancreatic morphogenesis, as previously described in other contexts (Yamada and Nelson, 2007; Yu et al., 2008).

The mechanisms that regulate pancreatic organ size remain poorly understood. Recent observations have shown that the number of pancreatic progenitors is established early in development by an intrinsic programme and this would dictate final organ size (Stanger et al., 2007; Zhou et al., 2007). At the onset of pancreatic morphogenesis, distal tips of the branching epithelium contain fast-proliferating progenitors and, possibly, define a special supporting ‘niche’, as shown in other epithelial organs (Hogan and Kolodziej, 2002; Horb and Slack, 2000; Watanabe and Costantini, 2004; Zhou et al., 2007). Interestingly, ablation of Stard13, expression of which is enriched at the distal tips, results in pancreatic hypoplasia. Based on our findings, we propose Stard13 as an intrinsic regulator of the number of pancreatic progenitors through the establishment of the pancreatic tip domain. From a mechanistic point of view, our data define a reciprocal interaction between actin-MAL/SRF signalling and MAPK signalling to locally regulate progenitor-cell proliferation in the pancreas (Fig. 6F). As such, the initial progenitor pool can expand and reach its proper final size, which predetermines pancreas organ size in the adult (Stanger et al., 2007; Zhou et al., 2007). By contrast, in the absence of Stard13, constitutively active Rho/actin signalling not only has disruptive effects on tip morphogenesis but also locally reduces active pERK, through MAL/SRF transcriptional regulation, eventually resulting in organ hypoplasia. The temporal succession of events proposed in our model is in line with the fact that aberrant cell shape and F-actin accumulation are the earliest detectable defects in the mutant pancreas, whereas reduction in cell proliferation becomes obvious at E12.5. Importantly, we also show that selective inhibition of downstream Rho/SRF transcriptional signalling rescues the Stard13 mutant cell proliferation defects, to some extent. Nevertheless, upstream inhibition of the Rho signalling using the C3 exotoxin molecule results in a more effective rescue of the Stard13^PA-deleted defects than does transcriptional disruption of the pathway using CCG1423. These observations reinforce the notion that cell shape and cytoskeleton are at the origin of the phenotypic defects in the Stard13 mutant, being not completely eliminated by a downstream rescue at the transcriptional level. Also, these results suggest possible mechanical effects of cell shape and actomyosin cytoskeleton on cell division in the developing pancreas that might be independent of the SRF/MAPK pathway.

Overall, this study presents an unprecedented mechanism for integrating morphogenesis and growth by combining localised control of actin cytoskeleton and growth factor signalling. Full understanding of the relationship between pancreatic progenitor cells and tissue architecture will help in defining the signals that are necessary for maintenance of the progenitor pool or differentiation into a particular lineage. Better understanding of these concepts will have practical implications for any possible cell-based therapy of diabetes based on expansion of pancreatic progenitors and generation of insulin producing β-cells from stem cells or progenitor cells.

We thank Ali H. Brivanlou for invaluable input into the study and contribution to the initiation of the work. We thank all members of the Spagnoli laboratory for discussion. We are grateful to C. Birchmeier, W. Birchmeier, M. Schober and M. E. Torres-Padilla for helpful comments on the manuscript; D. A. Melton for the Pdx1-Cre transgenic strain; to M. Sander for anti-Ngn3 antibody; M. Wegner for anti-Sox9 antibody; and A. Sprinkel for anti-Ptf1a antibody.