Rhou maintains the epithelial architecture and facilitates differentiation of the foregut endoderm

The definitive endoderm (DE) is a transient embryonic epithelial cell layer that forms the linings of the digestive and respiratory systems as well as organs for gas exchange, digestion and absorption, and removal of metabolic end products. The progenitors of the anterior DE, which underlie the head folds, are derived from epiblast cells anterior to the distal tip of the mid- to late-streak stage embryo (Lawson et al., 1986; Tam et al., 2007). Cells of the anterior DE form the foregut (Franklin et al., 2008; Tremblay and Zaret, 2005), which contributes to the lung, stomach, oesophageal and tracheal epithelia, as well as the liver, pancreas, thyroid and thymus (Zorn and Wells, 2009).

In a transcriptome analysis of the foregut endoderm of early-somite stage mouse embryos, we identified Rhou (also known as Wrch1), which codes for a Rho GTPase with 70% sequence similarity to Cdc42, among the genes that are preferentially expressed in the endoderm of the anterior intestinal portal. Rhou activity has been shown to influence the actin cytoskeleton, cell shape and behaviour in cultured cells. In fibroblasts, Rhou overexpression reduces the number of lamellipodia and the actin content of stress fibres (Saras et al., 2004) but enhances filopodia formation (Saras et al., 2004; Tao et al., 2001) and cell motility (Ory et al., 2007). In HeLa cells, Rhou localizes to focal adhesions and its overexpression promotes the disassembly of focal adhesions and increases cell motility (Chuang et al., 2007). Consistent with the cell culture results, in Xenopus embryos, in which Rhou is expressed in migrating cranial neural crest cells, Rhou knockdown impairs cell migration (Fort et al., 2011). In MDCK (canine kidney epithelial) cells, overexpression of Rhou disrupts tight junction formation and the distribution of F-actin, and depletion of Rhou disrupts lumen formation in cysts (Brady et al., 2009). The findings of these studies suggest that Rhou plays a role in regulating epithelial cell shape and structure and modulating cellular behaviour.

We have investigated the expression and function of Rhou in the development of the foregut endoderm in the mouse. We generated embryonic stem (ES) cell lines in which Rhou activity was stably knocked down. Analysis of differentiation of the Rhou-knockdown cells in vitro revealed a defect in endodermal lineage differentiation. In ES cell-derived embryos, knockdown of Rhou altered the epithelial architecture and impaired the differentiation of the foregut endoderm. Focal electroporation of Rhou small hairpin RNA (shRNA) constructs into the endoderm pinpointed a cell autonomous role for Rhou in maintaining epithelial structure. Our findings highlight a requirement for Rhou activity in maintaining the F-actin-related cytoskeletal organization of the foregut endoderm, which is required for maintaining proper tissue structure and facilitating cell differentiation, and present a mechanistic paradigm that the transition of epithelial architecture constitutes a critical step in the process of controlling the formation of organ buds.

ARC/s embryos at 4- to 5-somite stages were collected and the anterior intestinal portal was isolated by digestion with trypsin and pancreatin and mechanical dissection. Ectoderm and mesoderm tissues were dissected from the head folds and the heart, and combined as the non-endoderm tissues for comparison of gene expression profiles. RNA was extracted from 15-20 pooled samples of endoderm and ectoderm plus mesoderm cells and subjected to two rounds of linear amplification and labelling according to standard methods (Affymetrix). The amplified products were hybridized to Affymetrix mouse MOE430 GeneChips. The experiment was performed twice. RNA amplification labelling, hybridization and scanning were performed by the Australian Genome Research Facility. GC-RMA data normalization and further analysis were performed using GenePattern software.

Embryos were collected, fixed in 4% paraformaldehyde (PFA), dehydrated through a methanol series and stored at –20°C. Automated whole-mount in situ hybridization was carried out on an InSituPro machine (AbiMed) using a published protocol (Wilkinson and Nieto, 1993) modified as described previously (Loebel et al., 2004) and stained with BM purple (Roche). Antisense riboprobes were generated from linearized plasmids for Apom and Pax9, or plasmid inserts amplified with M13 forward and reverse primers for Rhou (IMAGE clone 3964150), Igfbp5 (IMAGE clone 2648602), Cldn4 (IMAGE clone 876728). A probe for Pyy was generated by amplifying a 3′ cDNA fragment by RT-PCR from mouse embryo RNA in two rounds of amplification incorporating a T7 promoter sequence as described (Bildsoe et al., 2009). Primer sequences for Pyy riboprobe fragment generation were (5′-3′): PyyF, TGCTCATCTTGCTTCGGAAGCTGTA; PyyR, TGCGAAATTTGCTTTTTATTTAGGGA; PyyT7R, TAATACGACTCACTATAGGGTGCGAAATTTGCTTTTTATTTAGGGA. Digoxygenin-labelled riboprobes were prepared using Ampliscribe Kits (Epicentre).

Heterozygous Dkk1 knockout (Mukhopadhyay et al., 2001), Ctnnb1-bfc (Nolan et al., 2000) and Lrp6-Gw mice (Bogani et al., 2004) were maintained and genotyped as described (Fossat et al., 2011). ROSA-lacZ mice were generated by crossing together ROSA26R and CMV-Cre mice and selecting for lacZ-expressing offspring.

Oligonucleotides encoding Rhou-specific shRNA molecules were designed using the on-line shRNA Explorer tool (www.genelink.com). We used two shRNA constructs (Kunath et al., 2003) containing oligonucleotide sequences against the 3′UTR (Rhou-809, Rhou-3130; Table 1). In addition, an shRNA construct targeted to the 3′UTR of Sox17 (Table 1), plus shRNA vector-only and non-targeting shRNA (Sigma Mission Control) were also prepared. The shRNA constructs were electroporated into R1-ES cells and selected for G418 resistance. The degree of knockdown of gene expression for individual clones was assessed by quantitative real-time RT-PCR using primers listed in Table 2.

Two protocols of activin A-directed ES cell differentiation were used. First, ES cells were differentiated as embryoid bodies (EBs) for 2 days followed by 5 days in serum-free media containing 100 ng/ml activin A (Kubo et al., 2004). To test the effects of JNK inhibition, EBs were differentiated in the presence of 10 ng/ml JNK inhibitor 1 (Merck). Second, EBs were generated in differentiation medium for 2.5 days and plated into methylcellulose in IMDM containing 30 ng/ml activin A for 3.5 days followed by plating on matrigel for 4 days. Cells were harvested at the end of the differentiation procedure for RNA extraction.

ES cell-derived embryos were generated by injection of ES cells into tetraploid blastocysts or aggregation of ES cells with tetraploid morulae (Nagy et al., 1990). Two-cell embryos, collected from superovulated ARC/S female mice, were electro-fused (CF-150, BLS Hungary) and cultured until the 4-cell stage for aggregation or the blastocyst stage for microinjection. For generating diploid chimeras, ROSA26-lacZ males were mated with super-ovulated C57Bl/6 females, embryos collected at the morula stage and cultured until the blastocyst stage for injection. Chimeric blastocysts were transferred to pseudopregnant ARC/s female mice. Embryos were collected 6-7 days after transfer [equivalent to embryonic day (E)8.5-9.5].

DNA constructs encoding GFP-RHOU (Ory et al., 2007), tdTomato-RHOU or shRNAs were introduced into the foreguts of embryos by electroporation as described previously (Khoo et al., 2007). Embryos were electroporated at E8.5 and cultured for 4 hours (GFP-RHOU group) or electroporated at E7.75 (early head-fold stage) and cultured for 24 hours (shRNA group). Embryos were photographed, then fixed in 4% PFA and processed for immunofluorescence. The shRNA constructs were prepared by cloning double stranded oligonucleotides encoding shRNAs (identical to those used in ES cells) into pRNAT-H1.1, which contained a GFP reporter. A non-targeting control was made in the same vector, using the shRNA sequence from the Mission non-targeting control construct (Sigma). A td-Tomato-RHOU construct was prepared by excising the GFP from GFP-RHOU and replacing it with a PCR product containing the complete open reading frame of tdTomato.

HepG2 human hepatocellular liver carcinoma cells, NIH3T3 cells and NIH3T3-derived lines containing the pLNCX retroviral vector directing constitutive expression of β-galactosidase, Wnt1, Wnt3a, Wnt4, Wnt5a, Wnt7a or Wnt11 (Kispert et al., 1998) were maintained in DMEM with 10% foetal calf serum (FCS). For testing the effects of recombinant WNTs, HepG2 cells were cultured for 6 hours in serum-free DMEM containing 200 ng/ml recombinant mouse Wnt3a, Wnt5a, human WNT7A (R&D Systems) or bovine serum albumin (BSA). Cells were transfected with plasmids pEGFP-N2 (control), pCMVSox17-IRES-eGFP (Pfister et al., 2011) or pβcateninS45A-GFP (Johnson et al., 2009) using Fugene6 (Roche) or Lipofectamine LTX (Invitrogen). Transfection with Stealth siRNA against CTNNB1 (50822) or negative control siRNAs (Invitrogen) was with Lipofectamine RNAiMAX (Invitrogen). Transfected cells were grown for 24 hours and processed for immunofluorescence, or grown for 24 hours (NIH3T3) or 48 hours (HepG2) and harvested for RNA extraction (siRNA transfected cells) or processed for flow sorting to isolate the GFP expressing cells in a FACSVantage cell sorter (plasmid transfection experiments).

RNA was extracted using RNeasy Mini or Micro Kits (Qiagen). First strand cDNA was generated using Superscript III (Invitrogen) and diluted 1:4. RT-PCR was performed using Biomix Taq (Bioline) on a Hybaid multi-block thermocycler for 25-35 cycles. Quantitative real-time RT-PCR (qRT-PCR) was performed using Platinum Taq (Invitrogen) and Sybr Green on a Corbett RotorGene 6000. Primer sequences and annealing temperatures are given in Table 2.

Embryos and embryoid bodies were fixed overnight at 4°C in 4% PFA, permeated with sucrose/PBS and infiltrated with a 1:2 mixture of 30% sucrose in PBS and O.C.T. compound, frozen, sectioned (7 μm) and stored at –80°C. Transfected HepG2 cells grown on coverslips were fixed in 4% PFA for 5 minutes and washed in PBS before phalloidin staining. The following primary antibodies were used: rabbit anti-ZO-1 (Invitrogen 40-2200, 6 μg/ml); rat anti-E-cadherin (Invitrogen 13-1900, 6 μg/ml); rat anti-fibronectin (Abcam ab23750, 5 μg/ml); rabbit-anti GFP (Invitrogen A11122) mouse anti-GFP (ABCAM), rabbit anti-dsRed (Clontech) and mouse anti-acetylated α-tubulin (Abcam ab24610, 4 μg/ml). Secondary antibodies used were: Cy2 donkey anti-mouse IgG (Jackson ImmunoResearch, 4.2 μg/ml), Alexa Fluor 633-conjugated goat anti-rabbit IgG (Invitrogen, 10 μg/ml) and Alexa Fluor 488-conjugated goat anti-rat IgG (Invitrogen, 6 μg/ml). F-actin was detected with Alexa Fluor 546-conjugated phalloidin (Invitrogen).

Immunofluorescence was imaged with a Leica TCS SP2 confocal microscope using a 100× objective. Pixel intensity and cellular dimensions were measured on unmodified images using ImageJ. Contrast and levels were adjusted for display with Adobe Photoshop.

PFA-fixed embryos were embedded in epoxy resin sectioned and viewed with a Philips CM10 transmission electron microscope (Miranda-Saksena et al., 2000).

Embryos were collected in PB1 medium (Kinder et al., 2000), rinsed in PBS and fixed for at least 2 hours in glutaraldehyde solution (0.02% Igepal CA630, 0.01% sodium deoxycholate, 0.2% glutaraldehyde, 5 mM EGTA, 2 mM MgCl₂), washed in Lac Z washing buffer (Watson et al., 2008) briefly before being incubated in Lac Z staining solution (Watson et al., 2008) at 37°C for colour development. The embryos were then washed in PBS and fixed in 4% PFA.

Fixed embryos were processed for wax histology by dehydrating through a graded ethanol series and embedding in paraffin wax. Embedded specimens were sectioned, transferred to SuperFrost slides (Menzel) and stained with Nuclear Fast Red or Haematoxylin and Eosin.

Microarray analysis was carried out on RNA extracted from dissected foregut endoderm and combined headfold and heart tissues (ectoderm and mesoderm; see Fig. S1A,B in the supplementary material). Differential expression analysis revealed that the foregut endoderm samples were enriched with genes with known endoderm-specific expression, confirming the efficacy of the isolation strategy (see Fig. S1C in the supplementary material). To identify Rho-related small GTPases in epithelial tissues, we searched for genes encoding Rho GTPases that displayed preferential expression in the endoderm. One such gene, Rhou, was found to be strongly upregulated in the endoderm (see Fig. S1C in the supplementary material).

Rhou expression was first detected by in situ hybridization in the anterior DE of early-somite embryos. At the 3-somite stage, Rhou expression was strongest in the rim of the foregut pocket (Fig. 1A,B), where the liver progenitors reside (Tremblay and Zaret, 2005). At E8.5 (8 somites), Rhou was expressed throughout the foregut except the regions anterior to the first branchial arch and the ventral lip of the anterior intestinal portal (AIP) (Fig. 1C,Ci-iii). Rhou expression was strong in the columnar epithelium on the lateral wall and the floor of the foregut but weak in the thin squamous epithelium of the roof (Fig. 1Ci,ii). Unlike in Xenopus embryos (Fort et al., 2011), Rhou expression was not detected in migrating cranial neural crest cells. At E9.5, Rhou was expressed in the pharyngeal pouches (Fig. 1D,Di) and the endoderm on the lateral wall of the caudal segment of the AIP. Rhou was weakly expressed in the thickened epithelium on the floor of the foregut where the thyroid primordium emerges (Fig. 1Di) and was not detected in the liver bud (Fig. 1Dii). By E10.0, Rhou expression was weak in the foregut (not shown). The timing and domain of expression of Rhou in the foregut endoderm suggest an association with the development or maintenance of the columnar epithelium of the foregut endoderm.

Rhou was originally identified by its upregulation by Wnt1 in cultured cells (Tao et al., 2001) and might, therefore, be involved in mediating the effects of Wnt signalling on endoderm development (Hansson et al., 2009; Lickert et al., 2005). To examine the effects of elevated Wnt signalling activity in embryos we compared Rhou expression in wild-type E8.5 mouse embryos with embryos harbouring mutations that enhance canonical Wnt signalling activity: the Batface (Bfc) allele of Ctnnb1 and the Gwazi (Gw) allele of Lrp6 (Fossat et al., 2011). Upregulation of Rhou was observed in Ctnnb1b^bfc/bfc and Lrp6^Gw/+ embryos (Fig. 2A,B). Rhou expression in Lrp6^Gw/Gw embryos was variable, consistent with the incomplete penetrance of the mutant phenotype (Fossat et al., 2011). We also observed increased Rhou expression in Dkk1^+/– and Dkk1^–/– embryos by qRT-PCR, and in the foregut and adjacent tissues of Dkk1^–/– embryos by in situ hybridization (Fig. 2C-G).

To explore the regulation of Rhou by Wnt signalling further, we examined the effects of expressing various Wnt factors on Rhou expression in NIH-3T3 cells. Rhou was induced most strongly by Wnt1 and Wnt3a and was above basal levels in the Wnt7a-expressing cells (Fig. 2H). Expression of Dkk1, a canonical Wnt target, was strongly enhanced in cell lines expressing Wnt1 and Wnt7a (Fig. 2H). These ligands primarily activate canonical Wnt pathways. Consistent with the involvement of canonical Wnt signalling, transfection of NIH3T3 cells with a construct encoding phosphorylation-resistant β-catenin point mutant (S45A) resulted in an approximately threefold upregulation of Rhou over the control transfected cells (Fig. 2I).

To test the response of RHOU to WNT signalling in cells that are more closely related to endoderm, HepG2 cells were cultured with recombinant mouse Wnt3a, Wnt5a and human WNT7A. RHOU was only upregulated by Wnt3a, whereas AXIN2, a canonical WNT target, was significantly upregulated by Wnt3a and WNT7A (Fig. 2J,K). Consistent with this, transfection of HepG2 cells with β-catenin-S45A-GFP also resulted in upregulation of RHOU (Fig. 2L). Sox17, a transcription factor crucial for DE development (Kanai-Azuma et al., 2002; Pfister et al., 2011) inhibits the transcriptional response of Rhou to canonical Wnt signalling (Jia et al., 2010; Zorn et al., 1999). Transfection of a Sox17-IRES-GFP expression construct reduced RHOU expression to one-third of the GFP-only control (Fig. 2M). However, knockdown of β-catenin did not result in downregulation of RHOU (Fig. 2N-P), suggesting that Sox17 regulates Rhou expression via β-catenin-independent means in HepG2 cells and that although the level of canonical WNT signalling activity can influence RHOU expression, Rhou is not a direct β-catenin target.

To investigate the requirement for Rhou in the differentiation of endoderm lineages, we generated Rhou knockdown (KD) ES cell lines. Two KD ES cell lines, Rhou809-9 and Rhou3130-3, harbouring different shRNAs that displayed 71% and 59% reduction in Rhou expression, respectively, compared with vector controls as measured by qRT-PCR were selected for this study.

In a preliminary experiment, in which a vector control and one KD ES cell line (Rhou3130-3) were differentiated as EBs for 15 days, the Rhou KD line showed a delayed and less robust upregulation of genes associated with mesendoderm progenitors, endoderm and mesoderm lineages [T, Mixl1, Flk1 (Kdr – Mouse Genome Informatics), Mug1], but maintained prolonged expression of the pluripotency marker Pou5f1 (see Fig. S2A,C,D in the supplementary material).

Culturing ES cells in the presence of activin A can enhance the differentiation of endodermal derivatives (see Fig. S2B,D in the supplementary material) (D'Amour et al., 2005; Kubo et al., 2004; Tada et al., 2005). To test whether knockdown of Rhou in ES cells impacts the differentiation of endodermal lineage in response to activin induction, we examined the expression of endoderm markers by RT-PCR (see Fig. S3A in the supplementary material). The strongest differences between KD and control cell lines were observed for genes that mark the hepatic (Hhex, Ttr, Mug1) or pancreatic lineages (Pdx1, Iapp). In several cases, the relative degree of downregulation of markers (notably Foxa2, Pdx1 and Ttr) in the KD cells correlated with the level of Rhou knockdown in the differentiated cells. Ttr is a target of AP1 transcription factor complexes, which contain the transcription factor c-Jun (Qian et al., 1995) and RHOU has been shown to influence AP-1 activity that is downstream of JNK (Zhang et al., 2011). Differentiation of EBs lacking JNK1 and JNK2 activity (Xu and Davis, 2010) or EBs in the presence of JNK inhibitor reduced the expression of some endodermal lineage markers (see Fig. S4 in the supplementary material). To study further the RHOU-JNK-AP1/c-Jun cascade, we examined other targets of AP1/c-Jun and found that Nrp1, Wnt5a and Dkk1 were downregulated in Rhou KD cells (see Fig. S3B in the supplementary material). In contrast to Dkk1, another Wnt target gene, Ccnd1, did not change its expression in KD cells. Hhex (see Fig. S3A in the supplementary material), a direct target of bone morphogenetic protein (Bmp) signalling was downregulated in Rhou KD EBs, but another Bmp target, Id2, was not (see Fig. S3B in the supplementary material). There was no change in the expression of targets of nodal signalling (Lefty1, Nodal) or fibroblast growth factor (Fgf) signalling (Spry2, Spred; see Fig. S3B in the supplementary material). Our data suggest that Rhou knockdown affects the JNK-AP1/c-Jun activity and some aspects of the Wnt and Bmp signalling cascades.

To test further the differentiation potency of the Rhou-KD ES cells, we cultured EBs in a lower concentration of activin A (30 ng/ml) on matrigel, which promotes both mesoderm and endoderm differentiation (Kubo et al., 2004). Rhou-deficient ES cells upregulated genes associated with cardiac mesoderm more robustly than did control cells (see Fig. S5 in the supplementary material). KD cells differentiated under these conditions generated more colonies (percentage of total number scored) with rhythmic beating activity than did control cells (control 1: 27.8±28; Rhou809-9: 97.6±2.4; Rhou3130-3: 84.2±15.9; mean ± s.e.m.), suggesting an enhancement of mesoderm differentiation. Overall, differentiation studies performed under two different conditions showed that Rhou KD adversely affects the ability of ES cells to differentiate into endoderm derivatives but facilitates mesoderm differentiation.

To analyse the role of Rhou in the development of the foregut endoderm in vivo, we studied the phenotype of embryos derived from Rhou KD ES cells by tetraploid complementation (Kunath et al., 2003; Nagy et al., 1990). To verify the efficacy of our techniques, we established that: (1) tetraploid embryos implanted but failed to develop; (2) embryos that were generated by injection of diploid ES cells harbouring a non-targeting shRNA construct into tetraploid blastocysts developed into morphologically normal embryos at E8.5 and E9.5; (3) in morphologically normal embryos generated from tetraploid hosts and wild-type ES cells, the ES cell contribution was very high; (4) ES-cell derived Sox17-knockdown embryos phenocopied Sox17-null mutant embryos; and (5) the tetraploid host cells did not contribute significantly to the foregut of the ES-derived chimera (see Fig. S6 in the supplementary material). The efficiency of generation of embryos was similar for the control and two KD cell lines (see Table S1 in the supplementary material).

Rhou expression was markedly reduced in KD embryos (Fig. 3A-C), indicating the shRNA knockdown was sustained during post-implantation development. Both lines of KD embryos displayed abnormalities, including a failure to turn, closely packed somites and a bulbous allantois that failed to connect with the chorion (see Fig. S6I-N in the supplementary material). The foregut of KD embryos was dorsoventrally flattened (in 22/26 Rhou809-9 and 28/33 Rhou3130-3 embryos; Fig. 3E,F). By contrast, majority of control embryos (25/28) showed a normal-shaped foregut (Fig. 3D). These phenotypes were not specific to the chosen cell lines. Embryos generated from one additional line each that harboured the Rhou809 or Rhou3130 shRNA construct also displayed similar phenotypes (see Fig. S7A,B in the supplementary material).

Histological examination of Rhou KD embryos revealed that the ventral and lateral endoderm epithelium in the foregut was irregular in thickness. In the most severely affected regions, cells appeared to pile up on each other, revealed by the positions of the nuclei and contrasting with the orderly cell organization in control embryos (Fig. 3G-O). The apical-basal positions of nuclei in the endoderm cell did not differ between control and KD embryos. KD endoderm cells were shorter in height and the nuclei were rounder (see Fig. S8 in the supplementary material). In both control and Rhou KD embryos, a multilayered thyroid primordium emerged from the floor of the foregut. However, the primordium of the Rhou KD embryos contained more cells, had an abnormal shape and elongated further away from the floor of the gut (Fig. 3P-R). These findings show that Rhou activity is required for proper epithelial morphogenesis of the endoderm.

The disruption of the columnar epithelial architecture of the foregut endoderm in Rhou KD embryos suggests an inability of the endoderm cells to maintain proper physical connection with the neighbouring cells. To address this, we examined the organization of tight and adherens junctions. When viewed by TEM, electron-dense tight junctions were found in the sub-apical lateral cell membranes of the foregut endoderm of control and Rhou KD embryos (Fig. 4A-C). Consistent with this, Cldn4 was expressed in both control and KD foregut endoderm (Fig. 4D-F) and ZO-1 (Tjp1 – Mouse Genome Informatics) was also properly localized (Fig. 4G-L). E-cadherin distribution in the adherens junctions was similar in control and KD embryos (Fig. 4J-L). These findings indicate that Rhou-deficient foregut endoderm cells can maintain the proper intercellular junctions required for epithelial organization.

TEM examination of the DE of KD embryos revealed a paucity of microvilli at the apical cell surfaces of endoderm cells (Fig. 4O-Q). Because microvilli contain a core of actin microfilaments (Vicente-Manzanares and Sanchez-Madrid, 2000), we tested whether Rhou and F-actin could potentially interact by examining their localization. A GFP-RHOU fusion protein (Ory et al., 2007) expressed in HepG2 cells was colocalized with F-actin in the cortical cytoplasm (Fig. 4M), and was preferentially localized in the F-actin-enriched apical domain of the DE cells in the foregut (Fig. 4N). We then examined the distribution of F-actin in the DE of KD embryos derived from the two KD cell lines. In both types of KD embryos, there was a marked reduction in F-actin in the apical region of the endoderm cell (Fig. 4R-W; see Fig. S9A-C in the supplementary material). This observation was supported by the reduction in the mean apical:basal ratio of pixel intensity of phalloidin staining in the endoderm cells of KD embryos. In addition, acetylated α-tubulin was dispersed away from the apical region of the KD endoderm cells (see Fig. S9D,E in the supplementary material). These findings show that less F-actin and acetylated α-tubulin is localized to the apical domain of Rhou-deficient cells.

We also examined F-actin, ZO-1 and E-cadherin localization in the endoderm-like cells on the surface of EBs cultured with activin A (see Fig. S10 in the supplementary material). We observed reduced content of apical F-actin in the superficial cells and widespread ZO-1 localization to deeper cells in the Rhou KD EBs. Knockdown of Rhou activity, therefore, affects the localization of the F-actin and tight junctions in the endoderm-like cells in the EBs.

To test whether the impact of Rhou deficiency was cell-autonomous, individual to small groups of cells in the foregut endoderm were subject to focal electroporation of Rhou shRNA constructs (Rhou809GFP and Rhou3130GFP) and a non-targeting control construct (NTC-GFP), all with a GFP reporter to enable identification of the electroporated cells. After 24 hours of culture, the majority (21/26) of NTC-GFP-expressing endoderm cells of the control group embryos retained their proper position in the epithelium and maintained connection with the apical surface of the epithelium (Fig. 5A-B′). By contrast, endoderm cells harbouring shRNAs were often sequestered to the basal side of the epithelium (Rhou809GFP: 13/26 cells, χ²=5.43, P<0.025; Rhou3130GFP: 27/52 cells, χ²=11.8, P<0.001; both significantly more than the control: 5/26 cells; Fig. 5C-E′). The cells that were sequestered basally remained within the E-cadherin-expressing epithelium (Fig. 5A′-E′). The yielding of apical positioning by Rhou809GFP-expressing cells was counteracted by co-electroporation of a construct encoding tdTomato-RHOU (Fig. 5F), resulting in only 4/22 cells being sequestered below apical surface, which is comparable to the control. Measurement of apical:basal ratios of phalloidin fluorescence revealed that individual KD endoderm cells that were able to maintained an apical presence already displayed a reduction in apical F-actin content (Fig. 5G). In diploid chimeras generated from lacZ-expressing host embryos and unlabelled Rhou809-9 ES cells, Rhou KD cells were able to colonize the endoderm extensively (Fig. 5H,I). In these diploid chimeras, the endoderm occupied by the KD (lacZ-negative) cells displayed an irregular appearance similar to that of the endoderm of the ES-tetraploid chimeras (Fig. 5H,I insets). Where the ES cell contribution was low, clusters of Rhou KD (lacZ-negative) cells tended to be localized basally in the epithelium (Fig. 5J,K). These data show that Rhou knockdown has little effect on endoderm potency of the ES cells but impairs ability of endoderm cells to maintain their proper position within the epithelium and this is associated with the disruption in F-actin distribution.

To establish whether the disruption of epithelial morphogenesis by Rhou knockdown affects DE differentiation in the embryo, endoderm marker expression was analysed by in situ hybridization. Because Rhou KD embryos were malformed by E9.5 and few survived beyond E10 (see Fig. S6M,N in the supplementary material), the present study focussed on E8.5-9.5 embryos. Igfbp5 and Pax9, which were expressed in the foregut in control embryos (Fig. 6A,D), were markedly reduced in KD embryos (Fig. 6B,C,E,F). Pyy (Hou et al., 2007), which was expressed in the posterior-ventral foregut endoderm and at the prospective site of the pancreatic primordium in control embryos (Fig. 6G), was reduced to small patches (Fig. 6H,I) or nearly absent (Fig. 6H-K) in Rhou KD embryos (and in embryos generated from two other KD cell lines, see Fig. S7C-E in the supplementary material). Apom, which was expressed in the liver bud of control embryos, (Fig. 6L) was weakly expressed in KD embryos (Fig. 6M,N). Apom-expressing structures reminiscent of the liver bud were formed in the Rhou3103-3 KD embryos, whereas in the Rhou809-9 KD embryos, the Apom-expressing cells were confined to a thin layer of cells in the bud-like structure (Fig. 6L-N, insets). Disruption of the epithelial morphology of the endoderm progenitors during early development of anterior intestinal portal therefore has an adverse impact on liver cell differentiation.

In this study, we show that depletion of Rhou leads to altered F-actin distribution and disrupts epithelial morphogenesis of the endoderm. Reduced Rhou activity affects cellular morphology and cytoskeletal organization in cultured cells (Aspenstrom et al., 2004; Brady et al., 2009; Saras et al., 2004). Of particular note, Rhou knockdown in MDCK cells disrupts epithelial cyst formation in a three-dimensional matrix (Brady et al., 2009). These findings suggest that Rhou activity is essential for the maintenance of cellular and tissue architecture.

Between E8.0 and E8.5 the lateral and rostral parts of the anterior definitive endoderm converge in the midline to form the foregut pocket (Franklin et al., 2008; Tremblay and Zaret, 2005). Concurrently, the endoderm in the peripheral prospective foregut pocket changes from squamous to columnar morphology and F-actin accumulates apically (Fig. 7A). Rhou expression encompasses the sites where liver progenitor cells are localized (Tremblay and Zaret, 2005). After the anterior intestinal portal is formed, Rhou expression is maintained in the ventral and lateral foregut endoderm. Rhou is not expressed in the dorsal foregut endoderm, which remains squamous and lacks apical-basal F-actin polarization, and also not in the emerging liver primordium where the epithelium has transformed into a multilayered cell sheet (Fig. 7A). F-actin is a key structural component of the cytoskeleton and is crucial for the maintenance of intercellular junctions, microvilli and cell-substrate interactions (Harris et al., 2009; Ofek et al., 2009). The function of Rhou in maintaining epithelial morphology and F-actin and α-tubulin distribution is probably cell-autonomous as only cells with Rhou KD lose their ability to maintain their apical position in the epithelium and the phenotypic changes do not affect the neighbouring cells (Fig. 7B). F-actin is also responsible for cellular stiffness and resistance to compressive forces (Jonas and Duschl, 2010; Oberleithner et al., 2009). Loss of apical F-actin in endoderm cells (Fig. 7B) might reduce the resistance of the cell to the compressive forces from surrounding cells and results in the cell being forced out of the apical/luminal domain. Overall, our findings suggest that Rhou affects the morphology of the endoderm cells via its action on the distribution of the actin and α-tubulin-enriched cytoskeleton.

The primordia of endoderm-derived organs are formed by budding of the epithelium. Budding of the thyroid, liver, lung and pancreas primordia begins by a thickening of the simple columnar epithelium, followed by pseudostratification and outgrowing from the foregut (Bort et al., 2006; Fagman et al., 2003). Bud formation from an epithelial tissue involves changes in cell shape including apical constriction, cytoskeletal rearrangement and epithelial sheet folding (Fagman and Nilsson, 2010; Pilot and Lecuit, 2005). These processes are associated with the formation of the thyroid diverticulum (Fagman and Nilsson, 2010; Hilfer et al., 1977). Rhou is downregulated in the ventral pharyngeal endoderm as the thyroid primordium forms, and knockdown of Rhou in embryos does not prevent the formation of organ buds. Rather, the thyroid primordium appeared enlarged in Rhou KD embryos, suggesting that the precocious loss of Rhou function expedites the morphogenesis of an organ bud.

The re-positioning of cells harbouring Rhou shRNAs towards the basal side of the epithelium (Fig. 7B) is reminiscent of the initiation of epithelial invagination. Our finding suggests a mechanism of the initiation of organ bud formation from the endoderm (Fig. 7C). As Rhou is downregulated in the presumptive organ bud, cells lose apical F-actin and are consequently sequestered into the basal part of the epithelium, resulting in the formation of a multilayered cell sheet. The cellular mass might expand, by recruitment of more Rhou-downregulated cells from the adjacent endoderm as well as the proliferation of cells in the multilayered sheet (Fagman and Nilsson, 2010), to form the organ bud (Fig. 7C). It is, therefore, possible that the timing and the domain of downregulation of Rhou in the ventral foregut endoderm might determine the schedule and site of organ bud formation during gut development.

Canonical Wnt signalling contributes to the regulation of Rhou in the embryo, but Rhou is not a direct transcriptional target of β-catenin, and is instead probably regulated via a JNK-mediated pathway (Schiavone et al., 2009; Tao et al., 2001). The results of this and other studies on the molecular mechanism of Rhou action highlight a likely connection between Wnt signalling, epithelial morphogenesis and the formation of organ primordia via Rhou-mediated changes to the cytoskeleton that affect cell shape and mechanical properties (Fig. 7D). Rhou activates JNK, as indicated by c-Jun phosphorylation (Chuang et al., 2007; Tao et al., 2001) and activity of the AP-1 transcription factor complex of which c-Jun forms a part (Zhang et al., 2011). Our ES cell differentiation data show that Rhou knockdown reduces the expression of targets of c-Jun/AP-1, including Wnt5a. JNK activity is required for normal expression of some endoderm lineage markers in differentiating EBs (see Fig. S4 in the supplementary material). JNK also acts via transcription-independent mechanisms to mediate the effects of Wnt on planar cell polarity/convergent extension during Xenopus development (Yamanaka et al., 2002) and regulation of epithelial morphogenesis, including the control of actin polymerization (Bogoyevitch and Kobe, 2006; Xia and Karin, 2004). This supports a connection between Wnt signalling and the regulation of the activity of components of the PCP/JNK pathway (Fig. 7D) and constitutes a novel mechanism whereby the transcriptional regulation of Rhou acts as the conduit for Wnt signalling activity to control cell and tissue morphogenesis and, subsequently, cell differentiation and organ bud formation.

We thank Andreas Kispert and Andy McMahon for the Wnt-expressing cell lines; Heiko Lickert for providing the RasGAP shRNA vector; Anne Blangy for the GFP-RHOU clone; Duncan Sparrow for the Apom cDNA and Rulang Jiang for the Pax9 cDNA; Angelyn Hor for histology; Evelyn Trounson for assistance with cell culture; Ross Boadle and Laurence Cantrill for assistance with microscopy; Sanda Lum and Maggie Wang for cell sorting; and staff of the CMRI BioServices Unit for animal care. Animal experimentation was approved by the CMRI/CHW Animal Care and Ethics Committee.