Artigo Acesso aberto Revisado por pares

ROCK 2 inhibition triggers the collective invasion of colorectal adenocarcinomas

2019; Springer Nature; Volume: 38; Issue: 14 Linguagem: Inglês

10.15252/embj.201899299

ISSN

1460-2075

Autores

Fotine Libanje, Joël Raingeaud, Rui Luan, Zoé ap Thomas, Olivier Zajac, Joel Paulo Russomano Veiga, Laëtitia Marisa, Julien Adam, Valérie Boige, David Malka, Diane Goèré, Alan Hall, Jean-Yves Soazec, Friedrich Prall, Maximiliano Gelli, Peggy Dartigues, Fanny Jaulin,

Tópico(s)

Cellular Mechanics and Interactions

Resumo

Article18 June 2019free access Transparent process ROCK2 inhibition triggers the collective invasion of colorectal adenocarcinomas Fotine Libanje INSERM U-981, Gustave Roussy, Villejuif, France Search for more papers by this author Joel Raingeaud INSERM U-981, Gustave Roussy, Villejuif, France Search for more papers by this author Rui Luan INSERM U-981, Gustave Roussy, Villejuif, France Search for more papers by this author ZoéAp Thomas INSERM U-981, Gustave Roussy, Villejuif, France Search for more papers by this author Olivier Zajac INSERM U-981, Gustave Roussy, Villejuif, France Search for more papers by this author Joel Veiga Cell Biology Program, Memorial Sloan-Kettering Cancer Center, New York, NY, USA Search for more papers by this author Laetitia Marisa Programme “Cartes d'Identité des Tumeurs”, Ligue Nationale Contre le Cancer, Paris, France Search for more papers by this author Julien Adam Pathology Department, Gustave Roussy, Villejuif, France Search for more papers by this author Valerie Boige Digestive Cancer Unit, Gustave Roussy, Villejuif, France Search for more papers by this author David Malka Digestive Cancer Unit, Gustave Roussy, Villejuif, France Search for more papers by this author Diane Goéré Digestive Cancer Unit, Gustave Roussy, Villejuif, France Search for more papers by this author Alan Hall Cell Biology Program, Memorial Sloan-Kettering Cancer Center, New York, NY, USA Search for more papers by this author Jean-Yves Soazec Pathology Department, Gustave Roussy, Villejuif, France Search for more papers by this author Friedrich Prall Institute of Pathology, University Medicine of Rostock, Rostock, Germany Search for more papers by this author Maximiliano Gelli Digestive Cancer Unit, Gustave Roussy, Villejuif, France Search for more papers by this author Peggy Dartigues Pathology Department, Gustave Roussy, Villejuif, France Search for more papers by this author Fanny Jaulin Corresponding Author [email protected] orcid.org/0000-0002-5110-1800 INSERM U-981, Gustave Roussy, Villejuif, France Search for more papers by this author Fotine Libanje INSERM U-981, Gustave Roussy, Villejuif, France Search for more papers by this author Joel Raingeaud INSERM U-981, Gustave Roussy, Villejuif, France Search for more papers by this author Rui Luan INSERM U-981, Gustave Roussy, Villejuif, France Search for more papers by this author ZoéAp Thomas INSERM U-981, Gustave Roussy, Villejuif, France Search for more papers by this author Olivier Zajac INSERM U-981, Gustave Roussy, Villejuif, France Search for more papers by this author Joel Veiga Cell Biology Program, Memorial Sloan-Kettering Cancer Center, New York, NY, USA Search for more papers by this author Laetitia Marisa Programme “Cartes d'Identité des Tumeurs”, Ligue Nationale Contre le Cancer, Paris, France Search for more papers by this author Julien Adam Pathology Department, Gustave Roussy, Villejuif, France Search for more papers by this author Valerie Boige Digestive Cancer Unit, Gustave Roussy, Villejuif, France Search for more papers by this author David Malka Digestive Cancer Unit, Gustave Roussy, Villejuif, France Search for more papers by this author Diane Goéré Digestive Cancer Unit, Gustave Roussy, Villejuif, France Search for more papers by this author Alan Hall Cell Biology Program, Memorial Sloan-Kettering Cancer Center, New York, NY, USA Search for more papers by this author Jean-Yves Soazec Pathology Department, Gustave Roussy, Villejuif, France Search for more papers by this author Friedrich Prall Institute of Pathology, University Medicine of Rostock, Rostock, Germany Search for more papers by this author Maximiliano Gelli Digestive Cancer Unit, Gustave Roussy, Villejuif, France Search for more papers by this author Peggy Dartigues Pathology Department, Gustave Roussy, Villejuif, France Search for more papers by this author Fanny Jaulin Corresponding Author [email protected] orcid.org/0000-0002-5110-1800 INSERM U-981, Gustave Roussy, Villejuif, France Search for more papers by this author Author Information Fotine Libanje1, Joel Raingeaud1, Rui Luan1, ZoéAp Thomas1, Olivier Zajac1,†, Joel Veiga2,†, Laetitia Marisa3, Julien Adam4, Valerie Boige5, David Malka5, Diane Goéré5, Alan Hall2, Jean-Yves Soazec4, Friedrich Prall6, Maximiliano Gelli5, Peggy Dartigues4 and Fanny Jaulin *,1 1INSERM U-981, Gustave Roussy, Villejuif, France 2Cell Biology Program, Memorial Sloan-Kettering Cancer Center, New York, NY, USA 3Programme “Cartes d'Identité des Tumeurs”, Ligue Nationale Contre le Cancer, Paris, France 4Pathology Department, Gustave Roussy, Villejuif, France 5Digestive Cancer Unit, Gustave Roussy, Villejuif, France 6Institute of Pathology, University Medicine of Rostock, Rostock, Germany †Present address: Department of Translational Research, Curie Institute, Paris, France †Present address: Imagine Institute, Paris, France *Corresponding author. Tel: +33 1 4211 5068; E-mail: [email protected] EMBO J (2019)38:e99299https://doi.org/10.15252/embj.201899299 PDFDownload PDF of article text and main figures. Peer ReviewDownload a summary of the editorial decision process including editorial decision letters, reviewer comments and author responses to feedback. ToolsAdd to favoritesDownload CitationsTrack CitationsPermissions ShareFacebookTwitterLinked InMendeleyWechatReddit Figures & Info Abstract The metastatic progression of cancer is a multi-step process initiated by the local invasion of the peritumoral stroma. To identify the mechanisms underlying colorectal carcinoma (CRC) invasion, we collected live human primary cancer specimens at the time of surgery and monitored them ex vivo. This revealed that conventional adenocarcinomas undergo collective invasion while retaining their epithelial glandular architecture with an inward apical pole delineating a luminal cavity. To identify the underlying mechanisms, we used microscopy-based assays on 3D organotypic cultures of Caco-2 cysts as a model system. We performed two siRNA screens targeting Rho-GTPases effectors and guanine nucleotide exchange factors. These screens revealed that ROCK2 inhibition triggers the initial leader/follower polarization of the CRC cell cohorts and induces collective invasion. We further identified FARP2 as the Rac1 GEF necessary for CRC collective invasion. However, FARP2 activation is not sufficient to trigger leader cell formation and the concomitant inhibition of Myosin-II is required to induce invasion downstream of ROCK2 inhibition. Our results contrast with ROCK pro-invasive function in other cancers, stressing that the molecular mechanism of metastatic spread likely depends on tumour types and invasion mode. Synopsis Metastatic progression is initiated by local invasion of the peritumoral stroma, but the mechanisms underlying cancer cell migration remain unclear. Here, histological analyses of human primary colorectal carcinomas (CRC) combined with screenings of live ex vivo 3D cultures determine the mode of dissemination and identify the Rho-GTPase signalling effectors involved. Conventional CRC cells invade collectively as differentiated epithelial glands with retained apico-basolateral polarity. ROCK2 kinase inhibition triggers leader-cell formation and dissemination of patient-derived xenograft explants. ROCK2 blocks guanine exchange factor FARP2 recruitment to the apical junctional complex. FARP2 activation and Myosin-II inhibition cooperate in collective invasion. Introduction With 90% of cancer patients succumbing from their metastases, there is a pressing need to understand the mechanisms of cancer cell dissemination (Ferlay et al 2015). Worsening patient prognosis and impacting medical treatments, the transition from in situ to invasive tumours is a crucial step in the metastatic progression that is triggered by the acquisition of migratory properties. Cancer cell invasion has long been considered as a single-cell process (Valastyan & Weinberg, 2011) as observed in leukaemia, in loosely organized tissues like sarcomas, or in highly cohesive carcinomas that have undergone an epithelial-to-mesenchymal transition (EMT) (Friedl & Alexander, 2011). Activated by specific transcription factors (EMT-TFs), Twist, Zeb, Snail and Slug, the EMT programme promotes mesenchymal cell invasion, but also increases stemness and survival, all contributing to cancer development and metastatic progression (Thiery et al, 2009; Puisieux et al, 2014). Recently, mice models of breast and pancreas cancers have shown that EMT activation contributes to chemoresistance but is dispensable for metastases formation (Fischer et al, 2015; Zheng et al, 2015a). The participation of EMT-TFs and individualized dedifferentiated single cells to cancer metastatic dissemination is therefore highly debated (Nieto et al, 2016; Aiello et al, 2017; Krebs et al, 2017; Ye et al, 2017). Cancer cell clusters can also invade as cohesive groups, a process called collective invasion. First observed from primary tumour explants over 20 years ago (Friedl et al, 1995), the collective invasion of cancer cells has been confirmed over the past decade by intravital microscopy in a variety of experimental model systems (Friedl, 2004; Alexander et al, 2008; Giampieri et al, 2009) and the systematic prospective analysis of primary tumour explants (Zajac et al, 2018). Yet, the cellular and molecular mechanisms underlying this process have been poorly investigated and most of our knowledge is based on the collective migration driving the embryonic development of model organisms. The front-rear polarization of the moving cohort translates into distinct cells, the leaders and the followers, cooperating to tract the collective (Haas & Gilmour, 2006; Wang et al, 2010). The leaders, protrusive cells at the front, respond to extracellular guiding cues and adhere to the matrix to generate traction forces. They are either defined by their closest position relative to the chemokine gradient, such as in drosophila border cells or mammalian vascular sprouting (Duchek et al, 2001) or by the expression of a transcriptional programme pre-existing to the migration process such as in zebrafish lateral line (Haas & Gilmour, 2006; Aman & Piotrowski, 2008) and in breast and lung carcinoma models (Cheung et al, 2013; Westcott et al, 2015; Konen et al, 2017). As central regulators of the cytoskeleton, the small GTPases of the Rho family and their regulation by guanine nucleotide exchange factors (GEFs) and GTPase activating proteins (GAPs) play important functions in all modes of cell migration (Ridley, 2015). The balance between endogenous RhoA and Rac1 activities regulates the switch between the mesenchymal and the amoeboid mode of single-cell invasion (Sahai & Marshall, 2003; Wolf et al, 2003; Sanz-Moreno et al, 2008). The GEFs and GAPs balancing the activation of these small GTPases vary with cell types, and many of them have been shown to participate in invasion (Tsuji et al, 2002; Ohta et al, 2006; Even-Ram et al, 2007; Sanz-Moreno et al, 2008). Rho-GTPases activities were also shown to be crucial in leader cell and at the cell–cell junctions of collectively moving cells. Tissue-specific GEFs control the polarized activation of either Rac1 or Cdc42 in leader cells in vitro and in vivo (Osmani et al, 2006, 2010; Bianco et al, 2007; Migeotte et al, 2010; Wang et al, 2010; Ellenbroek et al, 2012; Cai et al, 2014; Omelchenko et al, 2014; Westcott et al, 2015; Yamaguchi et al, 2015). In contrast, the collective invasion of tumour spheres with inverted polarity (TSIPs) in hypermethylated colorectal carcinomas is independent of Rac1 and protruding leader cells (Zajac et al, 2018). In follower cells, the level of RhoA-GTP is tightly controlled at cell–cell contacts in order to relax the junctions while maintaining the communication and the cohesion of the group during the migration (Omelchenko & Hall, 2012; Omelchenko et al, 2014; Reffay et al, 2014; Zaritsky et al, 2017). With 1 million new cases worldwide per year, colorectal carcinoma (CRC) is an important public health issue. CRCs develop through a series of genetic, genomic and epigenetic events along two distinct pathways: the CpG islands methylator phenotype (CIMP) generates microsatellite instability (MSI-high) or not (MSS or MSI-low) and gives rise to different histological subtypes, including mucinous and micropapillary CRCs (Yamane et al, 2014). The chromosomal instability pathway (CIN) is initiated by APC loss of function and leads to numerous chromosomes losses and amplifications (Pino & Chung, 2010). This neoplastic transformation is associated with the most common histological form of CRCs, named “Lieberkuhnian” (LBK) or “not otherwise specified” (NOS). Large-scale transcriptomic analyses have defined the consensus molecular subtypes (CMSs) of CRCs, and one of the four groups is associated with poor patient prognosis and mesenchymal signature (Guinney et al, 2015). However, immunostaining of patient specimens and data collected from patient-derived xenografts support that the upregulation of mesenchymal genes actually occurs in the stroma rather than in the tumour through EMT activation (Calon et al, 2015; Matano et al, 2015; McCorry et al, 2018). Histological studies proved that tumour budding (clusters of 5 tumour cells or less) at the invasive front is an independent prognostic factor for CRC patients survival (Prall et al, 2005). Yet, the 3D reconstruction of tissue sections revealed that virtually all tumour buds are connected to the main tumour mass and it is impossible to assess whether they represent a reservoir of migratory cells moving as small unit or releasing individuals (Bronsert et al, 2014; Enderle-Ammour et al, 2017). Xenografted mice models of CRC have shown that silencing Zeb1 reduces metastases formation (Spaderna et al, 2008; Wellner et al, 2009). In contrast, transgenic mice expressing Fascin, a WNT/β-catenin target gene, show that collective invasion also mediates CRC spread (Vignjevic et al, 2007). Therefore, the restriction of histological approaches based on fixed 2D cancer specimens and molecular analyses that cannot distinguish between tumour and stromal compartment have been unable to resolve the invasive behaviour of human CRCs. In this study, we monitored live cancer specimens ex vivo and identified that NOS colorectal adenocarcinomas predominantly undergo collective invasion in the form of differentiated epithelial glands. We then investigated how Rho-GTPases signalling triggers the formation of leader cells to promote the migration of these differentiated neoplastic cell cohorts. Results Conventional colorectal adenocarcinomas undergo collective invasion To determine the mode of invasion involved in the early step of conventional (NOS) colorectal adenocarcinoma dissemination, we first analysed formalin-fixed paraffin-embedded (FFPE) surgical specimens from 16 human primary tumours that have invaded the submucosa (NOS, stage pT1, see Fig EV1A for patients, Fig EV1B for tumour characteristics and Fig 1Ai for a representative example). E-cadherin localized at cell–cell contact of both normal and transformed epithelial cell sheets (Fig 1Aii and iii and Fig EV1C). This staining highlighted the epithelial glandular organization of the neoplastic tissue, including the invasive front, with cohesive cancer cells surrounding a small luminal space (Fig 1Aii and iii). Between these neoplastic glands, stromal cells display a robust vimentin staining (Vim(+), Fig 1Aii and iii). Although we do not exclude that some Vim(+) cells could be CRC cells that have completely lost E-cadherin expression and localize among the normal stromal cells, most of the tumour is organized as a cohesive tissue with E-cadherin-based junctions. This architecture suggested that CRCs may maintain their differentiated features and apico-basolateral polarity during invasion. In support to this, immunostaining revealed the polarized localization of the apical marker villin at the plasma membrane facing the luminal cavity of normal and transformed epithelial glands (Fig 1B, arrowheads). The cell–cell adhesion molecule EpCam is excluded from the apical membrane and rather localizes at the basolateral compartment in contact with adjacent cancer cells and the basal lamina (Fig 1B). Histological assessment by pathologists revealed that in 87% of the patients (14/16), more than 75% of the tumour surface organized as glandular structure (Figs 1C and EV1B and C). This shows that tumour cells at the invasive front of pT1 colorectal adenocarcinomas maintain their cohesion and epithelial identity, preferentially organizing as glandular structures in the peritumoral stroma. Click here to expand this figure. Figure EV1. Colorectal adenocarcinoma cells display cell-cell junctions and glandular organisation in peritumoral stroma The molecular characteristics and classification of primary tumours from CRC patients are annotated according to their location, histotypes (ADENO: adenocarcinoma) and TNM stage. Representative images of haematoxylin/eosin/saffron (HES) staining of CRC primary tumours from the 16 patients described in (A). The insets show the entire specimen and the red box the region displayed in the figure. Representative images of CRC primary tumour described in (A) stained using haematoxylin/eosin/saffron (HES), anti-E-cadherin or anti-vimentin and showing different tissue architectures. Download figure Download PowerPoint Figure 1. Colorectal adenocarcinomas organize as cohesive and polarized epithelial glands Representative specimen of colorectal (CRC) primary tumour stained with haematoxylin/eosin/saffron (HES), or antibodies against E-cadherin or vimentin. (i) The blue, orange and pink dotted lines highlight the normal mucosa, the submucosa and the muscularis propria, respectively. Red dotted line highlights the neoplastic tissue. Black arrowheads indicate the direction of invasion. Boxed regions ii and iii show high magnification of normal colonic glands (ii) and the CRC invasive front (iii). Scale bar: 2 mm and 500 μm. Representative images of histological sections of normal colon and primary CRC stained for EpCam and Villin. Boxed regions i, ii and iii are high magnifications of the luminal cavity of normal colonic gland (i) and colorectal carcinoma glands (ii and iii). Arrow heads point to the apical pole enriched in villin. Scale bars: 50 μm. Graph presenting the percentage of the tumour area displaying a glandular architecture from a cohort of 16 patients (see Fig EV1). Download figure Download PowerPoint The presence of differentiated neoplastic cell cohorts at the invasive front could either result from single-cell invasion and sequential EMT and MET activation or from the collective migration of transformed tissues. To assess the dynamic invasive behaviour of colorectal adenocarcinoma, we monitored live tumour specimens by videomicroscopy. We retrieved primary tumour and metastases explants for 10 patients (Fig EV2A) the day of the cytoreductive surgery and immediately embedded them into tridimensional (3D) gel made of extracellular matrices (ECM). Four days after recovery, we performed time-lapse imaging during 48 h and stained for actin and ezrin at end point. CRC histotype assessment indicated that 2 out of 10 patients had mucinous CRC, correlating with the inverted apico-basolateral polarity of the explants we observed ex vivo and consistent with our previous study (Zajac et al, 2018) (Figs EV2A and EV2B, patients #E and #H, excluded from the present study focusing on NOS adenocarcinomas). For all patients with adenocarcinoma (8/8), all tumour explants remained cohesive and most displayed one or multiple lumens (77.7 ± 6.3% and 81.6 ± 4.9% visualized by DIC or ezrin staining, respectively, Figs 2A and B, EV2C and D). We very rarely saw the detachment of single cells (not shown) and rather observed evident collective behaviour with cell groups protruding into the collagen-I gel (Fig 2C and D and Movie EV1). An average of 14 ± 4% of the explants harboured prominent actin-rich protrusions (Fig 2A arrows and 2C), contrasting with the smooth periphery of the non-invasive ones (Fig 2A (arrowheads), Fig EV2E and Movie EV2). The proportion of protruding explants varies between patients, ranging from 0 to 35%, but we did not detect a significant difference between primary tumours and metastases (Fig 2C). The polarized morphology of the invading gland, with protruding cells harbouring the features of leaders, is not an ex vivo artefact from experimental procedures. Indeed, microscopic study of colorectal adenocarcinoma specimens obtained from the invasive margins by serial semi-thin sections, complemented by electron microscopy, showed that the highly organized state of differentiated neoplastic glands is disrupted focally in “tubular invasion poles” which appears to be the leading edge of invading glands (Fig 2Ei). In the fully formed parts of the neoplastic glands, the epithelia are polarized and abut a basal membrane resembling normal crypts (Fig 2Ei arrowheads and Fig 2Eii). However, at the invasion pole, the basal membrane is missing and “leader” epithelial cells form villiform cytoplasmic extensions in direct contact with the extracellular matrix (Fig 2Ei arrows and Fig 2Eiii). Taken together, the results we obtained from the analyses of fixed and live human primary cancer specimens revealed that NOS CRC maintain their apico-basolateral polarity and epithelial architecture as they collectively invade into the submucosa in the form of neoplastic glands. The focal formation of protruding leader cells seems to be the starting points for this process. Click here to expand this figure. Figure EV2. Colorectal adenocarcinoma invade into the peritumoral stroma as cohesive glands The molecular characteristics and classification of primary tumours or metastases from CRC patients are annotated according to their location (primary tumour (colon or rectum) or metastases (liver peritoneum or pancreas), histotypes (ADENO: adenocarcinoma, MUC: mucinous), KRAS mutation (N.A. non-available, wt: wild type, mut: mutated) and TNM stage. Representative images of haematoxylin/eosin/saffron (HES) staining of the 10 specimens of CRC primary tumours or metastases described in (A). The insets show the entire specimen and the red box the region displayed in the figure. Graph representing the percentage of explants displaying lumens visualized by immunofluorescence (ezrin) for seven patients (one sample was lost during the experimentation) including four patients with primary tumours and three patients with metastases. At least 50 explants were analysed per patient. The error bar represents the standard error of the mean (means ± SEM). P values were calculated using unpaired t-test (ns: non-significant). Quantification of lumens based on ezrin staining (no lumen, single lumen or multiple lumens) from the patients described in (A) (the sample from patient B was lost during the staining). The results are expressed as percentage of each phenotype. Over 50 explants were counted per patient. Time-lapse sequences of an adenocarcinoma explant monitored by DIC microscopy for 2 days (corresponding to Movie EV2). Arrowheads point to non-protruding cells, and white stars point to lumens. No significant protrusive activity was detected in the course of the recording. Scale bars: 20 μm. Download figure Download PowerPoint Figure 2. Colorectal adenocarcinomas explants undergo collective invasion A. Representative confocal images of non-invasive or invasive human colorectal cancer explants collected from 8 patients with NOS adenocarcinoma (Fig EV2A and B). The explants were fixed 4 days after recovery and stained for the lumen (ezrin), F-actin (Phalloidin) and nuclei (DAPI). Boxed regions i, ii, iii and iv are displayed at high magnification. Arrowheads point to non-protruding cells, arrows point to protruding cells, and white stars show off-centred nuclei. Scale bars: 20 μm. B, C. Graphs representing the percentage of adenocarcinoma explants displaying lumens (B) or protrusions (C) from eight patients. Evaluation performed from DIC observation of live tissues, at least 50 explants were analysed per patients. The error bar represents the standard error of the mean (means ± SEM). P values were calculated using unpaired t-test (ns: non-significant). D. Time-lapse sequences of an adenocarcinoma explant undergoing collective invasion into collagen-I gel monitored by DIC microscopy over 2 days (corresponding to Movie EV1). Arrowheads point to non-protruding cells, arrows point to protruding cells, and white stars point to lumens. Scale bars: 50 μm. E. Images of an invading gland of a colorectal NOS adenocarcinoma, at the light microscopic level (i, semi-thin section, 100× objective), and as seen by electron microscopy (ii, 1,000× magnification; and iii, 4,000× magnification). Note that the bulk of the invading neoplastic gland is in a state of glandular organization (regions denoted by arrowheads in the left panel): epithelia (E, middle panel) are polarized to form a lumen (L) and are attached to a basal membrane (asterisk) by hemidesmosomes; the gland is surrounded by fibroblasts (F). However, this organization is focally disrupted at the tubular invasion pole (arrows in the left panel) where neoplastic epithelial cells have lost polarization and directly contact the surrounding stroma as the basal membrane is missing (right panel). Download figure Download PowerPoint ROCK inhibition triggers collective invasion from Caco-2 cysts To investigate the signalling pathways regulating adenocarcinoma collective invasion, we first sought an in vitro experimental model system that recapitulates the features we observed from tumour explants. We grew the CIN CRC cell line Caco-2 in 3D ECM gel. Immunostaining for the apical marker prominin-1 and F-actin revealed that Caco-2 cysts are formed by a polarized monolayer of cells surrounding a central lumen as seen in patients’ specimens. Over 90% of Caco-2 cysts harbour a smooth periphery due to the absence of protruding cells (Fig 3A, left panel, Movie EV3). As such, Caco-2 cysts grown in 3D matrices represent a pertinent organotypic model of non-invasive NOS adenocarcinoma, allowing gain-of-function studies to decipher the mechanisms triggering glandular collective invasion. To this end, we performed a siRNA-based screen targeting the RhoGTPase signalling pathway, central to all modes of cell motility. We transfected Caco-2 cells using a siRNA library targeting the 98 known human RhoGTPase effectors. Accounting for functional redundancy between homologs, we co-depleted the closest pair of proteins (Table EV1). We verified that the co-transfection of two siRNAs does not hinder the depletion of each protein (data not shown). The screen yielded three hits: Caco2 cysts transfected with siPARD6A+PDE6D developed a multi-lumen phenotype (consistent with previous report; Durgan et al, 2011) and siPLXNA1+siPLD1 gave defective lumens (not shown). Caco-2 cysts co-depleted for ROCK1 and ROCK2 displayed a supracellular polarization: a subset of protruding cells, usually two neighbours, get off-centred but remained attached to the cyst (Fig 3A, middle panels, arrows). Their nuclei lost the alignment with the monolayer and engaged into protrusion (Fig 3A, middle panel, stars). In a validation round, Z-stacks acquisition of Caco-2 cysts transfected with siROCKs further demonstrated that the bulging cells formed actin-rich protrusions towards the ECM gel (Fig 3A (right panels, arrows) and Fig EV3A), while the rest of the cohort did not (Fig 3A, arrowheads). These protruding morphology and supracellular polarization of Caco-2 cysts transfected with siRNA targeting ROCK resemble the collective invasion pattern observed form the CRC explants (Fig 2A). To confirm this result and exclude an off-target effect of the siRNAs, we used alternative reagents. First, we transduced Caco-2 cysts using shRNAs against ROCK1 and ROCK2 (Fig EV3B). This induced a 10-fold increase in the number of cysts with protruding cells, raising from 6 ± 1% to 60 ± 1%, demonstrating the phenotype was specific to ROCK depletion (Fig 3B and C). We then used pharmacological inhibitors Y27632 and H1152 to determine whether ROCK kinase activity was sufficient to control this phenotype. Treatment with Y27632 and H1152, respectively, increased the number of protruding cysts from 7 ± 2% to 56 ± 3% and 63 ± 2% (Fig 3B and C). Using live imaging, we confirmed that the “leader/follower” polarization of Caco-2 cysts treated with ROCK inhibitors Y27632 and H1152 resulted from the neoformation of protrusions, as observed from patient explants (Fig EV3C). E-cadherin staining confirmed that all cells of the cohort remained cohesive and that the luminal cavity was conserved (Fig 3B). Figure 3. ROCK inhibition induces the collective invasion of Caco-2 cysts Representative images of Caco-2 cysts transfected with control siRNA (siGlo) or siRNAs targeting ROCK1 and ROCK2, fixed and stained for prominin-1, F-actin (Phalloidin) and nuclei (Hoechst). Confocal Z-sections of cysts are displayed. Arrowheads show non-protruding cells. Arrows point to protruding cells. White stars mark nuclei that are off-centred relative to the cyst's monolayer. Scale bars: 20 μm. Representative confocal images of Caco-2 cysts treated with ROCK inhibitors Y27632 and H1152, non-treated (NT) or transduced with ROCK1 and ROCK2 shRNAs (shROCK1+2). The cysts were fixed 2 days after invasion and stained for E-cadherin, F-actin (Phalloidin) and nuclei (DAPI). Boxed regions i and ii are displayed at high magnification. White arrowheads point to non-protruding cells, green arrowheads point to the apical pole, arrows point to protruding cel

Referência(s)