EFA6 proteins regulate lumen formation through α-actinin 1

During organogenesis, the coordinated establishment of the apico-basal polarity with the de novo formation of an apical luminal space is fundamental to the emergence of the different types of epithelia. In adult organisms, the aptitude of the internal organs, which are lined with epithelial tissues, to ensure specific functions relies on the preservation of these characteristics. The failure in doing so is associated with a large variety of diseases (Blasky et al., 2015; Datta et al., 2011; Sigurbjörnsdóttir et al., 2014). In particular, these features are often compromised in carcinomas and tumors are formed of non-polarized cell aggregates incapable of collectively organizing a lumen (Martin-Belmonte and Perez-Moreno, 2011; Tanos and Rodriguez-Boulan, 2008; Wang et al., 2012). However, compelling tumoral cells to maintain their normal epithelial phenotype can help them override the power of oncogenes (DuFort et al., 2011; Weaver et al., 1997). Thus, it is important to decipher the molecular programs that instruct epithelial cells to collectively organize around lumens in order to maintain their physiological homeostasis.

For most epithelial tissues, the de novo formation of a lumen is generated by hollowing. In this process, apical vesicles are delivered to a focal point of the cell–cell contact named the apical membrane initiation site (AMIS) to give rise to the apical plasma membrane and a facing hollow cavity. The nascent lumen appears first as a closed elongated space named the pre-apical patch (PAP), which is limited by tight junctions. The PAP will then open and expand through a combination of events: the delivery of vesicular membranes, the repulsion of the apposed membrane by highly charged molecules, the increase of hydrostatic pressure, the coalescence of mini-lumens and, eventually, expansion through cell division. This process is closely synchronized with a profound rearrangement of the actin cytoskeleton into discrete structures essential for the attachment of structural and signaling apical proteins. These proteins will then yield a scaffold to shape the lumen and form an acto-myosin ring in support of the circumferential apical junctional complexes (AJCs), which are made of adherens junctions (AJs) and the apical tight junctions (TJs) that outline the luminal space (Datta et al., 2011; Sigurbjörnsdóttir et al., 2014). The organization and functions of the acto-myosin ring attached to the AJ have been extensively studied (Arnold et al., 2017; Braga, 2016; Grikscheit and Grosse, 2016; Lecuit and Yap, 2015); however, far less is known about the actin cytoskeleton associated to the TJ. Nevertheless, it is likely that both structures are somehow intermingled within the so-called apical perijunctional acto-myosin ring (PAMR) (Ebrahim et al., 2013; Sluysmans et al., 2017). The PAMR is described as a sarcomeric-like belt made of F-actin bundles containing myosin-II, which confers contractile properties, and bundling proteins, such as the non-muscle α-actinins, which stiffen the structure. The balance of both activities is believed to determine the flexibility of this belt, its mechanosensitivity and the tension forces exerted on the cell surface (Foley and Young, 2014; Martin and Goldstein, 2014; Murrell et al., 2015; Röper, 2015). The existence of a central apical acto-myosin network with radial contractility has also been reported (Coravos and Martin, 2016).

The α-actinin (ACTN) family comprises four members, the muscle ACTN2 and ACTN3, and the non-muscle ACTN1 and ACTN4, which are expressed in most other cell types. They share a common primary structure with a N-terminal actin-binding domain (ABD) and a C-terminal calmodulin-like domain (CAMD) separated by a central repeat of four spectrin-like domains (spectrin repeats domain; SRD). ACTN molecules form antiparallel dimers through their rigid SRD allowing for the cross-linking of actin filaments by the ABD positioned on either end. In comparison with the filamin proteins, which orthogonally cross-link actin filaments, non-muscle ACTNs form linear F-actin bundles, which increase the stiffness (Jahed et al., 2014; Stossel et al., 2001). They are believed to contribute to myosin-II-driven contractility by facilitating force transmission (Le Clainche and Carlier, 2008). ACTNs also act as a mechanical linker between actin filaments and cell–cell and cell–extracellular matrix (ECM) adhesion complexes. Besides its structural roles, ACTNs could also serve to couple actin nucleation to assembly at cell–cell contacts (Tang and Brieher, 2012) and contribute to the maturation of cell–ECM focal adhesion by transmitting mechanical forces (Iskratsch et al., 2014; Jahed et al., 2014; Parsons et al., 2010; Ye et al., 2014). Both ACTN1 and ACTN4 were found at the apical acto-myosin ring in association with the AJ (Honda et al., 1998; Tang and Brieher, 2012), whereas a few studies suggested a link with the TJ (Chen et al., 2006; Geiger et al., 1979; Nakatsuji et al., 2008). Nevertheless, the repertoire of ACTN molecules associated with the various F-actin structures is poorly defined and is made even more complex by the existence of ACTN1–ACTN4 heterodimers (Foley and Young, 2013). Moreover, it is not completely clear how ACTNs are recruited to cell–cell contacts. Thus, much remains to be discovered about the role of ACTNs at the AJC as well as their role during luminogenesis.

Small G-proteins of the Rho family and their partners are key regulators in the assembly and maintenance of the PAMR. In particular, RhoA contributes to the constitution of the contractile apical acto-myosin array through the localization and activation of the formin proteins and the activation of the motor protein myosin-II by the Rho-associated coiled-coil kinase effectors ROCK1 and ROCK2 (hereafter denoted ROCK) (Arnold et al., 2017; Quiros and Nusrat, 2014; Sluysmans et al., 2017; Takeichi, 2014). Tension forces are necessary for the establishment and functioning of the AJC, as they support the changes in cell shape occurring during epithelial morphogenesis (Coravos et al., 2017; Lecuit and Yap, 2015; Takeichi, 2014). However, how contractility and its regulators are impacting epithelial cell luminogenesis is an issue of ongoing debate.

Another small G-protein that regulates the cortical cytoskeleton is the ADP ribosylation factor 6 (Arf6). It plays a pivotal role in a wide variety of cellular events including cell surface trafficking, phagocytosis, cell–cell adhesion, and tumor cell migration and invasion (D'Souza-Schorey and Chavrier, 2006; Gillingham and Munro, 2007; Jaworski, 2007; Sabe et al., 2009; Schweitzer et al., 2011). In epithelial cells, Arf6 was initially shown to regulate vesicular trafficking to the apical pole of the cell (Altschuler et al., 1999) and later to impact the turnover of the AJ (Palacios et al., 2001, 2002, 2005), the establishment of the TJ (Klein et al., 2008; Luton et al., 2004), cyst morphogenesis (Tushir et al., 2010) and HGF-induced tubulogenesis (Tushir and D'Souza-Schorey, 2007). Consistent with these observations, the exchange factor for Arf6 (EFA6; also known as PSD) was found to be enriched at the apical pole and at the TJ in fully polarized MDCK cells (Luton et al., 2004). During early epithelial polarization, EFA6 is recruited to the cell–cell contact in a manner that is dependent on E-cadherin engagement, where it contributes to the formation of the TJ by stabilizing the apical acto-myosin ring (Théard et al., 2010). Expression of EFA6B (also known as PSD4) in the mammary tumoral MCF7 cell line restored a normal glandular phenotype, with the formation of lumens delineated by TJs (Zangari et al., 2014). Conversely, knockdown of EFA6B expression drives various mammary cell lines into epithelial-to-mesenchymal transition (Zangari et al., 2014). In breast cancer patients, the loss of EF6AB expression is associated with the claudin-low subtype characterized by the loss of expression of all the TJ components and a poor prognosis (Zangari et al., 2014).

The EFA6 family consists of four isoforms (EFA6A–EFA6D; EFA6C is also known as PSD2, and EFA6D as PSD3) sharing a general structure that comprises a variable N-terminal domain, a catalytic Sec7 domain bearing the nucleotide exchange activity, a PH domain responsible for their plasma membrane localization and a conserved C-terminal region involved in actin cytoskeleton rearrangement (Derrien et al., 2002; Franco et al., 1999; Sakagami, 2008; Sironi et al., 2009). In a previous study, we found that mutations that abolish the nucleotide exchange activity or delete the C-terminal domain abrogated the stimulatory effects of EFA6, indicating that both Arf6 activation and the C-terminal domain are necessary for epithelial polarization (Luton et al., 2004). In addition, complementation experiments demonstrated a finely tuned cooperation between the two signaling pathways associated with the activated Arf6 and with the EFA6 C-terminus (Klein et al., 2008).

In this study, we aimed to determine the signaling pathway associated to EFA6 C-terminus that contributes to its action on epithelial morphogenesis. Our data showed that: (1) ACTN1 is a direct partner of EFA6A C-terminal domain, (2) EFA6A is a crucial regulator of luminogenesis for which ACTN1 is the major effector, (3) together, they stimulate the formation and enlargement of a single lumen with a proper round shape, (4) they act by regulating cortical acto-myosin contractility, and finally (5) ACTN1 is also a partner of the EFA6B isoform in the promotion of lumen formation in the mammary tumoral cells MCF7.

Looking for functional partners of EFA6A in epithelial cells, we performed a two-hybrid screen of an epithelial library using the C-terminus (Cter) of EFA6A as a bait, and identified ACTN1 as the major interacting protein. A similar result was found by Sakagami et al. in a previous screen using a neuronal library (Sakagami et al., 2007). A pulldown assay was used to confirm that the Cter of EFA6A could bind exogenously expressed ACTN1–GFP (Fig. 1A). We then further characterized this interaction by assessing whether the two proteins could bind directly. We found that both GST–ACTN1 and GST–ACTN4 could pulldown full-length His–EFA6A (Fig. 1B), and that the central SRD of ACTN1 binds directly to His–EFA6A (Fig. 1C). In BHK cells, GFP–EFA6A co-immunoprecipitated endogenous ACTN1 (Fig. 1D), and Myc–EFA6A re-localized the endogenous ACTN1 to F-actin-enriched lamellipodia as well as to the cell surface microspikes induced by Myc–EFA6A (Fig. 1E; Derrien et al., 2002; Macia et al., 2008). We conclude that, in vitro, ACTN and EFA6A can bind directly through their respective spectrin and C-terminal domains, and that, in vivo, EFA6A can recruit the endogenous ACTN1 to cortical F-actin structures. In agreement with our previous reports (Luton et al., 2004; Théard et al., 2010; Zangari et al., 2014), during lumen formation mRFP–EFA6A was transiently found to be enriched at the AMIS and the opened PAP before its expression at the apical surface decreased to the low level found in mature cyst (Fig. S1A). We also examined the localization of ACTN1 and ACTN4 in polarized cysts formed by MDCK cells. ACTN1–GFP was diffused within the cytoplasm and enriched at the apex of cell-cell junctions (Fig. S1A). Colocalization of ACTN1–GFP with occludin (Fig. 1Fa–a″) indicated its accumulation at the TJ. In contrast, ACTN4–GFP was not consistently observed at the TJ at above the level of the GFP control (Fig. S1A; Fig. 1Fc–c″). In addition, ACTN1–GFP was found to be enriched at the nascent lumen formed in between two cells and its surrounding TJ (Fig. 1Fb–b″), whereas ACTN4–GFP was found on the newly formed luminal membrane and also was enriched along the cell–cell contact (Fig. 1Fd–d″). Later, in mature cysts, ACTN4–GFP appeared to be more enriched in AJs stained for E-cadherin (Fig. S1B). Given the enrichment of ACTN1 at the TJ, we focused our study on the importance of its role as an effector of EFA6.

EFA6A and ACTN1 are cortical actin regulators and both partially localize to the TJ in polarized epithelial cells; we thus investigated the mechanism of their direct interaction. ACTN1 could serve as a receptor to recruit EFA6A to the TJ. However, EFA6A localization to the plasma membrane was shown to rely on its phosphatidylinositol 4,5-bisphosphate (PIP2)-specific PH domain and its capacity to bind to F-actin (Macia et al., 2008). Furthermore, the EFA6A mutant deleted of its C-terminus, which no longer binds ACTN1, has been shown to localize at the plasma membrane in a similar manner to the full-length protein (Franco et al., 1999; Macia et al., 2008). In contrast, as shown above (Fig. 1E) ACTN1 was re-localized to Myc–EFA6A-induced membrane ruffles, suggesting that ACTN1 can be used as an effector of EFA6A in order to remodel the cortical actin cytoskeleton.

To further analyze the binding of ACTN1 to EFA6A and its functional properties in living cells, we ectopically expressed EFA6A away from the plasma membrane. We fused EFA6A to the mitochondrial-targeting peptide ActA and determined whether EFA6A–ActA was capable of re-localizing the endogenous ACTN1 to the outer membrane of mitochondria, from which it is normally absent. Since ACTN1 binds to the C-terminal domain of EFA6A, we first studied the EFA6ACter fused to mRFP and ActA (named hereafter EFA6ACter–mRFP–ActA). As previously observed by others, depending on the expression rate and construct used, in cells transfected with ActA chimeras the mitochondrial network tended to aggregate around the nucleus (Bubeck et al., 1997; Moeller et al., 2004; Reinhard et al., 1999; Zhang et al., 2009). Nevertheless, all constructs localized to the mitochondria as assessed by monitoring colocalization with the mitochondrial protein Hsp60 (also known as HSPD1) (Fig. S2A). When expressed in BHK cells, EFA6ACter–mRFP–ActA localized to the mitochondria (Fig. S2A) and recruited endogenous ACTN1 (Fig. 2A; magnification in Fig. S2B). In contrast, the control mRFP–ActA construct, which also localized to the mitochondria (Fig. S2A), did not recruit ACTN1 (Fig. 2A). Interestingly, the full-length EFA6A–mRFP–ActA localized to the mitochondria (Fig. S2A) but did not re-localize ACTN1 (Fig. 2A). This observation suggests that the Cter was not available for ACTN1 binding in the full-length EFA6A. Our previous work indicates that EFA6A exists in a closed conformation where the Cter is folded back onto the PH domain, preventing its association with the C-terminus of β-arrestin1 (denoted β-arrestin1Cter), another EFA6ACter ligand (Macia et al., 2012). To test whether the EFA6A–mRFP–ActA was in a locked conformation, we co-expressed β-arrestin1Cter–GFP with EFA6ACter–mRFP–ActA or the full-length EFA6A–mRFP–ActA constructs. Similar to what was observed with ACTN1, β-arrestin1Cter–GFP could only bind to EFA6ACter–mRFP–ActA (Fig. 2B) indicating that the binding of ACTN1 to EFA6A is regulated, and requires the release of its C-terminal domain.

Some F-actin staining was observed colocalized together with EFA6ACter–mRFP–ActA and the endogenous ACTN1 at the mitochondria (Fig. 2A; magnification in Fig. S2B). When the cells were treated with Latrunculin A, F-actin was absent from the mitochondria whereas the endogenous ACTN1 was still efficiently recruited by EFA6Cter–ActA (Fig. 2C). Thus, independently of the presence of F-actin, EFA6A can directly recruit ACTN1, which in turn might function as an effector to organize EFA6A-regulated actin-based structures.

We have previously shown that expression of EFA6A tagged with the G glycoprotein of the vesicular stomatitis virus (VSV-G) (VSV-G–EFA6A) stimulates apical polarity development and TJ formation in MDCK cells (Klein et al., 2008; Luton et al., 2004). To assess the role of ACTN1 as an effector of EFA6A, we analyzed the effects of its downregulation in cells grown in a 3D-culture system in Matrigel. Several siRNAs against ACTN1 were tested for their efficiency to downregulate its expression (Fig. S3A). ACTN1 knockdown was carried out in MDCK cells in which the expression of VSV-G–EFA6A is under the control of the tetracycline (Dox)-repressible transactivator (hereafter denoted MDCK-VSV-G–EFA6A cells) (Luton et al., 2004). Fig. 3A is a representative immunoblot analysis of ACTN1 depletion and VSV-G–EFA6A expression in the presence or absence of doxycycline.

We first examined whether EFA6A overexpression stimulated epithelial polarity in MDCK cells grown in Matrigel over 3 days. Cysts of homogenous size (from 4 to 15 cells) were analyzed for the formation of one or multiple lumens and for their shape. They were also characterized for extension and expansion. Extension refers to the opening of the lumen to all the cells of the aggregates, while expansion refers to the enlargement of the luminal space. Upon VSV-G–EFA6A expression (−Dox) we observed an increase in the formation of cysts displaying a single central lumen (SCL). In addition, the luminal space became more round with an almost doubling (from 23% to 40%) of the SCL, with it displaying a round shape upon VSV-G–EFA6A expression (Fig. 3B, shaded bars and upper right panel). In contrast, depletion of ACTN1 in control conditions (+Dox) reduced the number of cysts with a SCL and severely altered the shape of the lumen such that the lumens adopted an ‘octopus-like’ shape with a small central opening from which closed or barely opened luminal extensions reached in between the cells (Fig. 3B, shaded bars and lower left panel). Depletion of ACTN1 in cells expressing VSV-G–EFA6A also impaired enlargement of the lumens, although the opening was more visible (Fig. 3B, shaded bars and bottom right panel). Thus, ACTN1 is required for the stimulatory effects found for EFA6A on SCL formation and on the rounding of the luminal space of multicellular cysts. However, in all conditions, the cells remained well polarized as indicated by the correct localization of the apical podocalyxin (PDX; also known as PODXL) and basolateral (E-cadherin) markers (Fig. S3B), and by the basal positioning of the nuclei, the general F-actin organization and the apical assembly of the TJ (Fig. 3B). To confirm that the phenotypes induced by the siRNA against ACTN1 (siACTN1) were due to ACTN1 depletion we carried out a rescue experiment. MDCK cells expressing an ACTN1–GFP which is insensitive to the siACTN1 (ACTN1S744–GFP) did not display any defect in lumen formation upon depletion of the endogenous ACTN1 (Fig. 3C).

The loss of cysts with SCLs in the ACTN1-depleted cells was counterbalanced by an increase of cell aggregates with multiple lumens and others with incomplete extension to all of the cells (Fig. 3D, see Materials and Methods). In both cases, the lumens displayed the distorted octopus-like shape. The extension defect was further reflected by a strong increase in the number of aggregates displaying multiple lumens that only opened in between two cells (hereafter called mini-lumens) that were often blocked at the PAP stage with no visible opening (Fig. 3E), which suggests that ACTN1 depletion might favor the early formation of an initial lumen. VSV-G–EFA6A expression did not significantly alter the phenotypic changes imposed upon ACTN1-depletion; however, it stimulated the enlargement and rounding of the mini-lumens blocked at the two-cell stage (Fig. 3E, right panel). Thus, as opposed to what is seen for mature lumens, EFA6A can stimulate the volumetric enlargement of nascent mini-lumens in an ACTN1-independent manner. Taken together, these observations suggest that ACTN1 is dispensable for the formation of the initial lumen in between two cells but is required later on for its extension and enlargement (see Discussion section for further comments). In summary, depletion of ACTN1 blocked both the effects of VSV-G–EFA6A on luminogenesis, that is, the formation of fully extended SCL and its expansion as regular round lumen. This suggests that ACTN1 acts as an effector downstream of EFA6A that is important for luminogenesis.

If EFA6A controls luminogenesis and ACTN1 acts as an effector, then EFA6A depletion should hamper luminogenesis, and when combined with ACTN1 depletion there should have no additional effect. MDCK-VSV-G–EFA6A cells grown in the absence or presence of doxycycline were submitted to siRNA against EFA6A (siEFA6A)-, siACTN1- or simultaneous siEFA6A- and siACTN1-mediated depletion. The efficient knockdown of the indicated proteins and the induction of the expression of VSV-G–EFA6A were analyzed by immunoblotting (Fig. 4A; Fig. S4).

Depletion of the endogenous EFA6A was accompanied by a reduction in the amount of cell aggregates with a SCL demonstrating that EFA6A is required for luminogenesis (Fig. 4B). Expression of the exogenous human VSV-G–EFA6A, which is insensitive to the canine-specific siRNA, rescued the normal phenotype, thus controlling for the specificity of EFA6A knockdown (Fig. 4B). As shown in Fig. 3, ACTN1 depletion hampered luminogenesis and blocked the VSV-G–EFA6A stimulation. However, ACTN1 depletion did not have any additional effect when the endogenous EFA6A was knocked down (Fig. 4B). In addition, the exogenous expression of VSV-G–EFA6A in the double knockdown cells could not rescue the normal phenotype (Fig. 4B). In addition, we analyzed the consequences on the lumen formation of expressing a mutant of EFA6A deleted of its C-terminus (EFA6AΔC), which contains the ACTN1-binding site. Expression of VSV-G–EFA6AΔC impaired normal luminogenesis and caused the formation of lumens with an octopus-like shape (Fig. 4C). Although the C-terminus of EFA6A likely interacts with other molecules, it is important to note that its truncation generated lumens with shapes that were similar to those observed in ACTN1-depleted cells. Taken together, these observations indicate that EFA6A mediates luminogenesis in MDCK cells and that ACTN1 is a crucial effector for this process.

We have reported that the EFA6B isoform is an antagonist of breast cancer development (Zangari et al., 2014). Tumoral MCF7 cells grown in 3D-culture systems form compact aggregates with no lumen (Han et al., 2010; Kenny et al., 2007). The exogenous expression of VSV-G–EFA6B in the tumoral MCF7 cells restores an epithelial phenotype characterized by the appearance of aggregates with extended lumens (although not often a SCL), delineated by functional TJs (Zangari et al., 2014). Thus, we asked whether ACTN1 was required downstream of EFA6B to contribute to luminogenesis in MCF7 cells.

We first verified that ACTN1 could also bind to the EFA6B isoform (Fig. S5). Next, we analyzed the effect of ACTN1 depletion in both control MCF7 and MCF7-VSV-G–EFA6B cells grown in Matrigel. An immunoblot analysis demonstrated the efficient knockdown of ACTN1 in both cell lines (Fig. 5A). We quantified the number of aggregates with extended lumens opened to at least four cells, as opposed to mini-lumens opened in between only two cells. As previously reported, MCF7 control cells do not form cysts with lumens (Han et al., 2010; Kenny et al., 2007; Zangari et al., 2014), whereas exogenous expression of VSV-G–EFA6B induced the formation of cysts with extended lumens (Fig. 5B; Zangari et al., 2014). We observed that ACTN1 depletion blocked the formation of extended lumens induced by EFA6B indicating that, in MCF7 cells, ACTN1 is also a crucial effector of EFA6B-induced lumen formation (Fig. 5B). In addition, although control MCF7 cell aggregates did not form extended lumens, in ∼20% of the aggregates one or several mini-lumens were observed (Fig. 5C,D). Upon ACTN1 depletion, the number of these mini-lumens increased in both the VSV-G–EFA6B-expressing and non-expressing MCF7 cell lines. However, in VSV-G–EFA6B-expressing cells aggregates the luminal space was enlarged while in MCF7 controls cells the mini-lumens were seen as PAPs (Fig. 5C,D). Thus, similar to what we had observed in MDCK cells, ACTN1 depletion facilitates the initial lumen formation in between two cells but then hampers both extension and enlargement.

The PAMR and ventral stress fibers (SFs) are sarcomere-like actin-based structures capable of contractility that have been implicated in epithelial morphogenesis. Contractile actin bundles are formed by aligned fibers cross-linked by a periodic distribution of ACTNs that alternates with myosin-II (Burridge and Wittchen, 2013; Sluysmans et al., 2017). RhoA and its effectors (ROCK and mDia1) have been recognized as major regulators of the contractile properties of the acto-myosin cytoskeletons (Arnold et al., 2017; Quiros and Nusrat, 2014; Sluysmans et al., 2017; Takeichi, 2014). Thus, we tested the possibility that EFA6A and ACTN1 regulate luminogenesis by contributing to the PAMR contractility. First, we used an antibody directed against the phosphorylated regulatory myosin light chain 2 (pMLC, also known as MYL2) as a proxy to evaluate the tension forces in response to modulating EFA6A and ACTN1 expression. Fig. 6B shows a representative immunoblot analysis of ACTN1 depletion and VSV-G–EFA6A expression in the presence or absence of doxycycline. In control cells, the pMLC staining appeared as small intracellular dots, as well as larger dots located in proximity to the apical membrane. Under conditions of EFA6A or ACTN1 depletion, this apical staining disappeared and the pMLC distribution became more diffuse. Upon VSV-G–EFA6A expression the proportion of apical pMLC was increased and its staining was evenly distributed all around the luminal cortex. In ACTN1-depleted cells, the VSV-G–EFA6A-induced apical accumulation of pMLC was abrogated (Fig. 6A). These results suggest that EFA6A can modulate the apical tension forces in an ACTN1-dependent manner.

We investigated the contribution of ACTN1 by first analyzing the effect of its depletion on the RhoA-stimulated ventral SFs in MDCK cells by expressing the N-terminal Myc-tagged constitutively active mutant of RhoA (Myc–RhoAV14) under the control of the inducible Tet-off system. The levels of expression of the proteins in different conditions were analyzed by immunoblotting (Fig. 6C). We monitored the Myc–RhoAV14-induced SF contractility by immunofluorescence. We looked at the F-actin organization and the contractility status by following the distribution of pMLC and paxillin, a marker of the focal adhesions (Fig. 6D). As expected, the expression of Myc–RhoAV14 (−Dox) stimulated the formation of SFs that appeared as thick bundles of parallel F-actin going across the whole cell cluster (Fig. 6Db). In Myc–RhoAV14-expressing cells, pMLC was redistributed all along the SFs (Fig. 6Db′) while in control cells pMLC was enriched at the periphery of the cell cluster and excluded from cell–cell contacts (Fig. 6Da′). Concomitantly, paxillin was re-localized to the periphery of the cell cluster mostly decorating the focal adhesions at the extremity of the SF (Fig. 6Db″). As previously observed by others (Oakes et al., 2012), depletion of ACTN1 in control cells (+Dox) led to the loss of bundles or shortening of radial F-actin bundles resembling transverse arcs (Fig. 6Dc). These structures were stained for pMLC, which remained absent from cell–cell contacts (Fig. 6Dc′). The paxillin signal increased but was still homogeneously distributed within each cell of the cluster similar to what was seen in control cells (Fig. 6Dc″). Depletion of ACTN1 blocked the effects of Myc–RhoAV14, leading to cells displaying a phenotype that was comparable to that of the control cells with respect to all three markers (Fig. 6Dd–d″). We conclude that ACTN1 can balance the contractility status of RhoA-dependent SFs in MDCK cells.

Interfering with the RhoA–ROCK–myosin-II contractility pathway has been shown to alter luminogenesis in MDCK cells (Ferrari et al., 2008; Kim et al., 2015; Rodríguez-Fraticelli et al., 2012). In particular, under permissive conditions, the constitutively active RhoAV14 mutant blocks the initial step of lumen formation (Ferrari et al., 2008). We thus asked whether ACTN1 depletion could prevent the Myc–RhoAV14 inhibitory effect on luminogenesis. In our 3D cell culture conditions, Myc–RhoAV14 expression abrogated lumen formation. The cells became round and loosely attached, forming aggregates almost exclusively without lumens and that displayed an inverted polarity, as judged by the peripheral localization of the apical PDX marker (Fig. 6E,F). ACTN1 depletion partially rescued the formation of lumens (Fig. 6E). We observed aggregates with multilumens or incomplete lumen formation, essentially as mini-lumens (Fig. 6E). The cells adopted a more cuboidal shape and formed more-compact aggregates (Fig. 6F). These results suggest that the knockdown of ACTN1 acts to reduce the strong contractility imposed by Myc–RhoAV14, thus allowing the initiation of lumen formation. This is in agreement with previous studies that show that reduction of the myosin-II contractility stimulates the initial early step of lumen opening at the two-cell stage (Ferrari et al., 2008; Rodríguez-Fraticelli et al., 2012). However, the extension and enlargement of the lumens, which we demonstrate to be dependent on ACTN1, were both still compromised upon reduction of contractility induced by ACTN1 depletion. Thus, our findings support a model whereby the initial stage of the lumen formation is facilitated by the absence of ACTN1, followed by lumen maturation, which is ACTN1 dependent.

Taken together, our data suggest that EFA6A and its effector ACTN1 contribute to SCL formation through the regulation of the cell surface tension forces within the cell aggregate.

In this study, we aimed to define the role of EFA6, a regulator of epithelial polarization, in lumen formation by using a 3D epithelial cell culture system. Given the importance of the reorganization of the actin cytoskeleton during cystogenesis, we were particularly interested in deciphering the signaling pathway associated with the EFA6 C-terminal domain that is capable of remodeling the cortical actin independently of EFA6 GEF activity.

Searching for partners of EFA6ACter, we find that EFA6A binds directly to the non-muscular ACTN member ACTN1. We characterized this interaction in living cells by using a mitochondrial-targeting system, and observed that it is likely regulated by the structural conformation of the EFA6A C-terminus. This observation confirms our previous results obtained with another EFA6Cter ligand, β-arrestin1 (Macia et al., 2012), and implies the existence of a common regulatory mechanism whereby binding to EFA6ACter requires an opening signal. A possible candidate for mediating this signal is Arf6-GTP, the product of EFA6A Sec7-dependent nucleotide exchange activity. ARNO (also known as CYTH2), another Sec7 family guanine nucleotide exchange factor for Arf1, was shown to adopt an auto-inhibited conformation where its short C-terminus downstream of the PH domain interferes with the catalytic Sec7 domain (DiNitto et al., 2007). Arf6-GTP binds to the PH domain of ARNO and stimulates, in synergy with plasma membrane lipids, the nucleotide exchange activity of ARNO on Arf1 (Cohen et al., 2007b; Stalder et al., 2011). We found that Arf6-GTP binds to the PH-C-terminal region of EFA6 (Padovani et al., 2014). Thus, by analogy with ARNO, one could speculate that Arf6-GTP, by binding to the PH domain of EFA6, could release the C-terminus and allow for ACTN1 binding.

We had previously reported that EFA6A is present at and regulates the TJ (Luton et al., 2004; Théard et al., 2010), where we now find ACTN1. Our results support the hypothesis that it is EFA6A that recruits ACTN1. The re-localization experiments performed in the presence of Latrunculin B indicate that the binding between EFA6A and ACTN1 is independent of F-actin. However, F-actin is found around the mitochondria together with EFA6ACter–ActA and ACTN1 suggesting that the EFA6A and ACTN1 couple could organize actin filament structures or even contribute to their nucleation, as proposed for ACTN4 at the AJ (Tang and Brieher, 2012). ARNO was also found to bind directly to ACTN1, through its C-terminal extremity, to modulate neurite extension, suggesting that ACTNs could be a general player in the Arf-regulated actin cytoskeleton (Torii et al., 2012).

Since EFA6A recruits ACTN1, we hypothesized that ACTN1 was acting as its effector. In the past, studying monolayers of MDCK cells, we had found that overexpression of EFA6A accelerates the general program of epithelial polarization, including the assembly of functional TJs (Klein et al., 2008; Luton et al., 2004). Part of the contribution of EFA6A has been attributed to its ability to stabilize the PAMR. Here, by using a 3D cell culture system together with siRNA-mediated depletion, we show that EFA6A is necessary for normal luminogenesis in MDCK cells. Conversely, overexpression of EFA6A increased the proportion of cysts with a SCL and stimulated their enlargement. Thus, we asked whether ACTN1 was acting as an effector to transduce some of the effects of EFA6A on luminogenesis. Indeed, we observed that depletion of ACTN1 blocked the stimulatory effect of EFA6A on both the formation and enlargement of the round SCL. In further support of this idea, depletion of ACTN1 had no additional disruptive impact on cells knocked down for EFA6A. This addition, expression of a mutant of EFA6A deleted of its C-terminus, which contains the ACTN1-binding site, impaired normal luminogenesis. These results demonstrate that ACTN1 is a crucial effector of EFA6A whose function is to promote normal luminogenesis.

To understand the roles of ACTN1, we analyzed in detail the defects induced upon its depletion. In the absence of ACTN1, there was an increase in cysts with multiple lumens, mostly as mini-lumens formed between two cells, or cysts with lumens that failed to extend to all cells. We propose that, in the absence of ACTN1, the initial fusion event to form a PAP is facilitated; however, the subsequent coalescence and extension of lumens to all cells of the aggregate is precluded. In addition, the volumetric growth of the lumens is impaired. In the ACTN1 knockdown cells, the mini-lumens appeared to be essentially blocked at the PAP stage and, in larger cysts, the extended lumens and SCL adopted an octopus-like shape rather than a nice round hollow. Interestingly, in both MDCK and MCF7 cells, in the absence of ACTN1, EFA6 overexpression is capable of rescuing enlargement of the mini-lumens. However, at later stages EFA6A overexpression cannot rescue the enlargement because ACTN1 is required as an effector downstream of EFA6. This implies that the enlargement of the luminal space relies on different molecular machinery at different stages along the process of lumen formation. Thus, ACTN1 is dispensable at the initial stage to create the nascent lumens, but it is essential downstream of EFA6A at later stages for the coalescence, extension and enlargement of the lumens.

ACTN1 is an actin-bundling protein known to regulate the contractility and stiffness of acto-myosin. It competes with myosin-II to bind F-actin in order to maintain a complementary periodicity, which results in linear F-actin bundles with contractile properties. The ratio and distribution of these two proteins determines the overall contractility and rigidity of F-actin cytoskeletons, including that of the PARM (Arnold et al., 2017; Ferrari et al., 2008; Sluysmans et al., 2017). Thus, ACTN1 could act by regulating tension forces at the surface of the developing lumen. Several observations support of this hypothesis. First, EFA6A and ACTN1 depletion reduced the amount of apical pMLC, whereas EFA6A expression stimulated its accumulation at the apical pole in an ACTN1-dependent manner. Second, upon ACTN1 depletion in both MDCK and MCF7 cells, we observed two defects: (1) a stimulatory effect on the initial formation of lumens as indicated by the rapid accumulation of multiple mini-lumens and, (2) at later stages, a defect of maturation (coalescence and extension) that leads to large cysts with several lumens or incomplete lumens. These results are consistent with a role for ACTN1 on contractility or stiffness. Indeed, several independent studies have shown that decreasing contractility through treatment with blebbistatin, ROCK inhibitors, or downregulation of LKB1 (also known as STK11), an upstream regulator of RhoA, stimulates the initial formation of a lumen at the two-cell stage (Cohen et al., 2007a; Ferrari et al., 2008; Rodríguez-Fraticelli et al., 2012; Taniguchi et al., 2015). Less appreciated is the impact of contractility at later stages. We found that similar to what is seen upon ACTN1 depletion, blebbistatin leads to large cysts with multiple lumens (our unpublished results). A similar observation was made by reducing contractility through depletion of LKB1, which first stimulates the initial formation of lumens but then prevents their coalescence into a SCL (Rodríguez-Fraticelli et al., 2012). Third, supplemental evidence for the role of ACTN1 on contractility came from its effects in RhoAV14-expressing cells. The small G protein RhoA, through its action on the apical PAMR, is a general regulator of the assembly and maintenance of the AJ (Lecuit and Yap, 2015) and TJ (Terry et al., 2010). We showed that ACTN1-depletion could counterbalance the RhoAV14-mediated contractility. We also found that ACTN1-depletion rescued the initial formation of lumens inhibited by RhoAV14 expression. Taken together, these results indicate that luminogenesis is regulated by acto-myosin contractility and stiffness, which should be kept low for the initial formation of a mini-lumen, and then be increased to allow for their coalescence and extension to form a SCL in larger multicellular aggregates.

The implementation of tension forces relies on molecular pickets, which are usually transmembrane proteins that anchor the acto-myosin structures. In polarized epithelial cells, the AJ and TJ are anchor points for the PAMR. It is noteworthy that neither EFA6A nor ACTN1 appeared to be essential for the assembly and positioning of the AJ and TJ, nor the establishment and maintenance of the asymmetry along the apico-basal axis. Thus, EFA6A and its effector ACTN1 modulate the contractility by affecting the activity and/or organization of the apical acto-myosin cytoskeleton and not the anchoring junctional complexes.

Not all the functions of ACTN1 appeared to be mediated through regulating contractility. ACTN1 depletion prevented the luminal enlargement, while the inhibition of contractility by inhibiting myosin-II or LKB1 depletion (Rodríguez-Fraticelli et al., 2012) did not affect the luminal enlargement. Enlargement has been proposed to depend on apical membrane transport involving the delivery of highly charged molecules, hydrostatic pressure mediated by ion channels and coalescence of multilumens. In ACTN1-depleted cells, the surface of the apical membrane appears large enough to accommodate a bigger luminal volume. In addition, the highly charged PDX protein is properly delivered and coalescence is not relevant when considering SCLs with an octopus-like shape. However, ACTN1 could help to retain polycystins, which are implicated in intercellular mechanotransduction (Li et al., 2005); Wilson, 2001), as well as ion channels. In fact, many ion channels have been shown to bind or to require ACTNs in order for them to be retained at the cell surface (Cukovic et al., 2001; Lu et al., 2009; Maruoka et al., 2000; Sadeghi et al., 2002; Schnizler et al., 2009; Wyszynski et al., 1997; Ziane et al., 2010). EFA6A was also shown to bind the ion channels TWIK1 and Kir3.4 (Decressac et al., 2004; Gong et al., 2007). Furthermore, it has been reported that ACTN4 regulates the outwards water transport in a process called regulatory volume decrease in response to osmotic swelling (Ando-Akatsuka et al., 2012). In summary, together with EFA6A, the recruitment of ACTN1 to the apical surface might help stabilize ion channels regulating fluid influx and consequently lumen enlargement.

Alternatively, ACTN1 could serve to organize a non-contractile scaffold to help support the spherical architecture of the luminal membrane. To do so, ACTN1 could participate by bundling F-actin filaments into rigid structures and/or could serve to link or even nucleate actin filaments at the AJ or TJ to anchor the luminal actin cytoskeleton. Such a role has been proposed for ACTN4 at the level of the AJ (Tang and Brieher, 2012).

Finally, we examined the role of ACTN1 in the human tumoral mammary cell line MCF7. When grown in 3D culture, these cells fail to assemble TJ, form a lumen or polarize (Han et al., 2010; Kenny et al., 2007; Zangari et al., 2014). Only a small fraction of cell aggregates display one or several mini-lumens. As shown in the past, EFA6B expression stimulated the formation of extended lumens and restored an epithelial-like phenotype. So, in contrast to MDCK cells, the formation of extended lumens in MCF7 is strictly dependent on the expression of VSV-G–EFA6B. Thus, these cells make a good model to study the role of ACTN1 downstream of EFA6B. Indeed, ACTN1 depletion totally blocked lumen formation, demonstrating that ACTN1 is a crucial effector for EFA6B in the induction of luminogenesis. Furthermore, ACTN1-depletion in wild-type MCF7 cells stimulated the formation of multiple mini-lumens that were blocked at the PAP stage. EFA6B expression did not rescue the formation of extended lumens but did enlarge their volume. Thus, in agreement with our observations in MDCK cells, ACTN1 depletion favors the initial formation of lumens in between two cells but prevents their coalescence and enlargement.

In conclusion, we show that ACTN1 is an effector of EFA6A and EFA6B to promote luminogenesis in normal and tumoral cell models. We propose a scenario whereby, at the onset of luminogenesis, the acto-myosin contractility and rigidity must be kept low to relax the sub-membranous actin cytoskeleton where vesicle fusion occurs at the AMIS. At this stage, EFA6 proteins have a stimulatory effect on the formation and the enlargement of the mini-lumens suggesting that it acts through an effector other than ACTN1; perhaps through Arf6, which has already been shown to be involved in luminogensis (Tushir et al., 2010; Zangari et al., 2014). At a later stage, EFA6 proteins recruit ACTN1 to contribute to the coalescence of the mini-lumens, their extension to neighboring cells and enlargement. We do not exclude the possibility that EFA6 proteins could recruit other effectors to mediate its effects. Nevertheless, ACTN1 is certainly a primary effector that might act by modulating the contractility of the acto-myosin ring and thereby mediating its effects on coalescence and extension (by helping to pull the lumens together and breaking through the junctional complexes). Given these results, the EFA6A–ACTN1 pathway might be important for tubulogenesis occurring through the cord hollowing mechanism, where multiple mini-lumens are formed along the tubular structure and subsequently fuse to coalesce in a single lumen, in a ROCK–myosin II-dependent manner (Bernascone et al., 2017; Kim et al., 2015; Sigurbjörnsdóttir et al., 2014).

In summary, we identified and characterized the role of two new regulators of luminogenesis: EFA6A and ACTN1. ACTN1 behaves as an effector to transduce the stimulating effect of EFA6A on the formation of a single and well-expanded central lumen by facilitating the extension and enlargement stages. The EFA6A–ACTN1 couple acts, at least in part, by balancing the contractility of the cortical acto-myosin cytoskeleton. In the tumoral MCF7 cells, this pathway has the capacity to restore the apico-basal polarization, TJ assembly and collective cellular organization into a cyst with a central lumen pointing to new directions for cancer research and therapy.

BHK cells were grown in Glasgow's minimum essential medium (GMEM; Invitrogen, Paris, France) supplemented with 5% heat decomplemented fetal calf serum (FCS; Biowest-Abcys, Nuaillé, France) and penicillin-streptomycin (Invitrogen). Transient transfection was performed by lipofection using JETPEI (Polyplus Transfection, Illkirck, France). MDCK clone II cells expressing VSV-G–EFA6A, VSV-G–EFA6AΔC (Luton et al., 2004) or Myc-RhoAV14 (Jou and Nelson, 1998) under the control of the tetracycline-repressible transactivator were grown in MEM (Sigma-Aldrich, Saint-Quentin-Fallavier, France), 5% decomplemented FCS (Biowest-Abcys), penicillin-streptomycin and 20 ng/ml doxycycline. Expression of EFA6A and RhoAV14 was induced upon removal of doxycycline. Plasmids and siRNA transfections were performed by nucleofection (Nucleofector™; Lonza, Köln, Germany). For stable expression of ACTN1–GFP, ACTN4–GFP and ACTN1S744–GFP, the cells were selected with geneticin (Invitrogen) 2 days after transfection. MCF7 and MCF7-VSV-G–EFA6B (Zangari et al., 2014) were grown in Dulbecco's modified Eagle's medium (DMEM; Sigma-Aldrich) containing 10% decomplemented FCS (Hyclone™, GE Healthcare, France), supplemented with insulin, transferrin, selenium, glutamine, sodium pyruvate, MEM non-essential amino-acids and penicillin-streptomycin (all from Invitrogen). Transient transfection was performed by nucleofection. For 3D cell culture, 10⁴ cells were mixed with 20 µl of 5 mg/ml Matrigel (BD Biosciences, Le Pont de Claix, France) deposited as a drop on a 12 mm glass coverslip. The BHK and MCF7 cells were obtained from the ATCC and authenticated by short tandem repeat profiling by the vendor. The parental MDCK II cell line was from Dr Keith Mostov (University of California, San Francisco, USA). MDCK and derived cell lines were tested for species specificity. Newly thawed cells from frozen stocks were tested for absence of mycoplasma contamination and used for 10 passages before replacement.

Rabbit polyclonal sera specific for EFA6B (HPA034722; Sigma-Aldrich), occludin (71-1500, Invitrogen), phosphorylated (Thr18/Ser19) myosin light chain 2 (3674, Cell Signaling Technology, Leiden, The Netherlands), the PI3K p85 regulatory subunit (ABS234, Millipore, Molsheim, France), Hsp60 (Ab46798, Abcam, Paris, France), GST (27-4577-01, GE Healthcare) were used. Mouse monoclonal antibodies specific for gp135/podocalyxin (3B8; gift from George Ojakian, State University of New York Downstate Medical Center), the VSV-G tag (P5D4; Roche Diagnostics, Mannheim, Germany), actin (AC-40; Sigma-Aldrich), E-cadherin (36; Invitrogen), paxillin (610052, BD Biosciences), the 6× histidine tag (HIS1; Sigma-Aldrich), the Myc tag (9E10 and rat hybridoma 3F10; Roche Diagnostics), GFP (7.1, Roche Diagnostics), ACTN1 (BM75.2; Sigma-Aldrich and H-2; Santa Cruz Biotechnology, Heidelberg, Germany) were used. The rabbit polyclonal anti-EFA6A was as described elsewhere (Sakagami et al., 2007). The horseradish peroxidase (HRP)-coupled and fluorescent probes (secondary antibodies, phalloidin and DAPI) were from Jackson ImmunoResearch Labs (Suffolk, UK) and Molecular Probes (Eugene, Oregon, USA), respectively. Full details of antibodies, including dilutions used are given in Table S1. All other reagents and chemicals were from Sigma-Aldrich.

Constructs for the expression of the following proteins have been described elsewhere: GFP–EFA6A (Decressac et al., 2004), mRFP–EFA6A (Théard et al., 2010), GST–EFA6A, GST–EFA6ACter (Macia et al., 2008), (6xhis)–EFA6A (Macia et al., 2008), βarrestin1Cter–GFP (Scott et al., 2002), the GST–ABD [amino acids (aa) 1–269], GST–SRD (aa 218–749) and GST–CAMD (aa 713–887) fragments of ACTN1 (Fraley et al., 2003), ACTN1–GFP (Rajfur et al., 2002), GST–ACTN4 (Khurana et al., 2011), ACTN4–HA and ACTN4–GFP (Michaud et al., 2006). The GST–ACTN1 construct was prepared by PCR amplification of full-length ACTN1 from pEGFP-ACTN1 and cloned into EcoR1 and Xho1 sites of the pGEX4T1 using the following primers:5′-GATCGATCGAATTCATGGACCATTATGATTCTCAG-3′ and 5′-TGTATCACTCGAGTTAGAGGTCACTCTCGCCGTAC-3′. The siRNA #2225 insensitive ACTN1–GFP was generated by introducing silent mutations at positions 2236, 2237 and 2238 (aa Ser744) using the QuikChange mutagenesis kit (Agilent Technologies, Courtabœuf, France) using the following primers 5′-GCCAAGGGCATCTCGCAGGAGCAGATGAATG-3′ and 5′-CATTCATCTGCTCCTGCGAGATGCCCTTGGC-3′. The resulting plasmid was termed ACTN1S744–GFP. Full-length EFA6A and EFA6ACter were cloned into mRFP-N1-ActA (Benjamin et al., 2010) by PCR amplification at the EcoRI and SacII sites. Primers used to amplify and clone the full-length EFA6A were 5′-GATCGATCGAATTCGCCGCCACCATGCCTCTCAAGTCACCTGTG-3′ and 5′-GATCGATCCCGCGGGGGCTTCCGCCGCCCACTGCC-3′, and those for the EFA6ACter fragment were 5′-GATCGATCGAATTCGCCGCCACCATGTTCTCTGCGCCCCCCTTCC-3′ and 5′-GATCGATCCCGCGGGGGCTTCCGCCGCCCACTGCC-3′. The resulting plasmids were termed EFA6A–mRFP–ActA and EFA6ACter–mRFP–ActA.

The specific and control siRNAs were designed and obtained from Eurogentec (Angers, France) and Sigma-Aldrich. The silencing efficiency of several siRNAs per target was assessed by immunoblotting (Figs S3 and S4). Knockdown of canine EFA6A and ACTN1 expression in MDCK cells was carried out using the siRNA #2661 5′-CCUAUCAGAGGCGGAGCUA-3′ or #1440 5′-CUCUUUCAGUUGUGUGUUU-3′ and siRNA #2225 5′-GCAUCAGCCAGGAGCAAAU-3′ or #594 5′-GGACGACCCACUCACAAAU-3′, respectively. Knockdown of human ACTN1 expression in MCF7 cells was carried out using the siRNA #1789 5′-CCUCAGGAGAUCAAUGGCAAA-3′. Silencing specificity was verified with control siRNAs, rescue experiments (Fig. 3C and Fig. 4B) and additional independent siRNA (Fig. S4).

The induction and purification of GST constructs with glutathione–sepharose CL-4B beads (GE Healthcare) was as previously described (Macia et al., 2008). The N-terminal 6×his–EFA6A was purified on Ni-NTA columns according to the manufacturer's instructions (Qiagen, Courtabœuf, France). For the GST-EFA6A pull-down from cell lysates, MDCK-ACTN1–GFP cells were lysed at 4°C in 0.5% Nonidet P-40, 20 mM Hepes pH 7.4, 125 mM NaCl, 1 mM phenylmethylsulphonyl fluoride (PMSF) and a cocktail of protease inhibitors (Complete™, Roche Diagnostics). The cleared lysates were incubated for 4 h at 4°C with 1.5 µM of the indicated GST-fused proteins and 30 µl of glutathione-sepharose CL-4B beads. After three washes in lysis buffer, the beads were boiled in Laemmli buffer, submitted to SDS-PAGE and the proteins revealed by immunoblotting. For the GST pull-down of purified His–EFA6A, 10 µM of the indicated GST fusion proteins and 10 µM of His–EFA6A were incubated together for 2 h at 4°C, and the experiments carried on as described above. For all GST pulldown experiments, an aliquot of the mixture was analyzed by SDS-PAGE followed by immunoblotting to estimate the total amount of the added proteins. For immunoprecipitation, cells were solubilized in ice-cold Triton X-100 lysis buffer (1% Triton X-100, 150 mM NaCl, 5 mM EDTA, 20 mM Triethanolamine-HCl pH 8.1 and 1 mM PMSF). After centrifugation for 20 min at 16,000 g at 4°C, the supernatants were pre-cleared at 4°C for 10 min, centrifuged for 10 min at 16,000 g and combined with protein A–sepharose and the indicated antibody overnight at 4°C. The beads were then washed three times in washing buffer (1% Triton X-100, 0.2% SDS, 150 mM NaCl, 5 mM EDTA, 8% sucrose, 1 mM PMSF and the cocktail of protease inhibitors) and washed once in washing buffer without detergent. The immunoprecipitates were then resuspended and boiled for 5 min in Laemmli buffer before SDS-PAGE and immunoblot analysis.

Samples were resolved on SDS-PAGE and proteins transferred onto a nitrocellulose membrane. Membrane blocking (30 min at room temperature) and antibody dilutions were performed in PBS with 5% non-fat dried milk. The membranes were incubated overnight at 4°C with the indicated antibody. The proteins were revealed by chemiluminescence (ECL, Amersham, GE Healthcare) using secondary antibodies directly coupled to HRP. The membranes were analyzed with the luminescent image analyzer LAS-3000 (Fujifilm, Saint-Quentin-en-Yvelines, France).

Cells were fixed in 4% PFA, the samples processed as previously described (Luton et al., 2004) and imaged on a confocal microscope (Leica TCS-SP5, Nanterre, France and Zeiss LSM780, Marly-le-Roi, France). Images were processed for presentation using NIH Image and Adobe^® Photoshop^® CS2 software.

Unless specified, all quantifications were from cysts of 4–15 cells obtained after 48 to 72 h growth in Matrigel. Experiments were repeated at least three times in triplicate and a minimum of 100 cysts per experimental replicate were analyzed. Each cell aggregate was scanned by using confocal microscopy to analyze the shape of the lumens and determine the presence of a single lumen or multiple lumens. Cysts were classified within five categories: cysts with a single central opened lumen (SCL), two or more lumens regardless of their shape (multilumen), with an opened lumen but that was not extended to all cells (incomplete lumen), with an optically closed lumen (otherwise named pre-apical patch, PAP), and with no lumen. Within the SCL class, we discriminated those lumens that were well enlarged from those that were barely opened and not expanded displaying an ‘octopus-like’ shape (see Fig. 3B). Note, that cysts with dividing cells or with lumen-containing cells and/or nuclei were not included. In addition, a separate quantification was performed for a lumen opened only in between two cells (mini-lumen) as they fell in the multilumen, incomplete and optically closed lumen categories (see Fig. 3E). Enlargement refers to the increase of the luminal volume. When enlargement was compromised there was an octopus-like lumen. Extension refers to the opening of a lumen (in general starting in-between two cells) to all the cells of the cell aggregate or coalescence of multiple small lumens to eventually form a single lumen. When extension was compromised it led to an incomplete lumen and/or to the presence of multiple lumens including mini-lumens (see Fig. 3D,E).

The experiments were performed at least three times in triplicate, and data from all experiments were combined. Values are mean±s.d. Statistical significance was calculated with a two-tailed Student's t-test. Non-significant difference (N.S.) are P>0.05; *P≤0.05; **P≤0.01; ***P≤0.001.

We thank Drs Hiroyuki Sakagami (Kitasato University, Japan) for the EFA6A-specific anti-serum, James Nelson (Stanford University, USA) for the MDCK-RhoAV14 cells and the ActA–mRFP plasmid, Dr Jeffery Greenwood (Oregon State University, USA) for the plasmids encoding the three fragments of GST–ACTN1 and GFP–ACTN1, Dr Chris Kennedy (University of Ottawa, Canada) for plasmids encoding HA–ACTN4 and GFP–ACTN4, and Dr Hung-Ying Kao (Case Western Reserve University, USA) for plasmid encoding GST–ACTN4. We are grateful to K. L. Singer for critical reading of the manuscript.