E-cadherin bridges cell polarity and spindle orientation to ensure prostate epithelial integrity and prevent carcinogenesis in vivo

Cell polarity and correct mitotic spindle positioning are essential for the maintenance of a proper prostate epithelial architecture, and disruption of the two biological features occurs at early stages in prostate tumorigenesis. However, whether and how these two epithelial attributes are connected in vivo is largely unknown. We herein report that conditional genetic deletion of E-cadherin, a key component of adherens junctions, in a mouse model results in loss of prostate luminal cell polarity and randomization of spindle orientations. Critically, E-cadherin ablation causes prostatic hyperplasia which progresses to invasive adenocarcinoma. Mechanistically, E-cadherin and the spindle positioning determinant LGN interacts with the PDZ domain of cell polarity protein SCRIB and form a ternary protein complex to bridge cell polarity and cell division orientation. These findings provide a novel mechanism by which E-cadherin acts an anchor to maintain prostate epithelial integrity and to prevent carcinogenesis in vivo.

Luminal cells are the most abundant type of the prostate epithelial cells. Most prostate cancers also display a luminal phenotype. Horizontal cell division of luminal cells allows the surface expansion of the secretory prostate lumen and meanwhile maintains the monolayer and polarized epithelial architecture. Disruption of the epithelial integrity and appearance of multilayer epithelia are early events in prostate adenocarcinoma development. However, the molecular mechanism that ensures the horizontal division in luminal cells remains largely unknown. Here, we generated a genetically engineered mouse model in which E-cadherin, a key component of the adherens junction that serves to connect the lateral plasma membrane of neighboring epithelial cells, was knocked out in the prostate luminal cells. E-cadherin deletion leads to loss of cell polarity and disoriented cell division,

Introduction
The prostate initially arises from embryonic urogenital sinus and undertakes ductal morphogenesis postnatally [1,2]. Murine prostatic epithelia are comprised of an inner single layer of polarized luminal cells, an outer layer of loosely distributed basal cells and a small fraction of scattered neuroendocrine cells [3,4]. Basal and luminal cells in the developing prostate epithelium display distinct cell division modes [5]. Luminal cells undergo symmetrical cell divisions during which the spindle orientation aligns parallel to the epithelial lumen and mother cell divides horizontally to generate two luminal cells. In contrast, basal cells undergo either horizontal symmetrical cell divisions to reproduce themselves or vertical asymmetrical cell divisions to give rise to a basal and a luminal daughter cell [5]. Horizontal cell division is of great importance for not only the surface expansion of prostate secreting lumen but also the maintenance of a monolayer luminal epithelial architecture, loss of which is an early event in prostate adenocarcinoma development. However, the molecular mechanism that ensures the horizontal symmetrical cell division of prostate luminal cells remains largely unknown. Previous work has demonstrated that cell polarity is indispensible for correct cell division orientations. Cell polarity are instructed by three types of asymmetrically distributed polarity protein complexes, the Scribble (SCRIB)/Lethal giant larvae (LGL)/Discs large (DLG) protein complex beneath the basolateral cell membrane, the partitioning defective 3 (PAR3)/PAR6/ atypical protein kinase (aPKC) in the cell apical-basal domain, and the Crumbs/PALS/PATJ protein complex under the apical cell membrane. Intensive studies in Drosophila have demonstrated that distribution cues for the spindle orientation determinants are derived from cell polarity [6]. An evolutionally conserved leucine-glycine-asparagine repeat protein (LGN)/ nuclear and mitotic apparatus (NUMA)/ inhibitory alpha subunit of heterotrimeric G protein (Gαi) complex, which forms a lateral cortical belt to generate forces on spindle astral microtubules through interacting with dynein/dynactin, has been shown to be a major spindle positioning machinery [7]. Apical distribution of polarity protein aPKC phosphorylates LGN to exclude LGN from the apical cortex and determines the planar plane of the cell division [8][9][10][11]. DLG can interact directly with LGN and control its localization to orient the spindle position [12][13][14]. SCRIB or DLG knockdown in the developing Drosophila wing disc epithelium results in spindle orientation defects [15]. Nevertheless, how the polarity cues and horizontal spindle orientation are connected in mammalian systems is not fully understood.
Adherens junctions are closely associated with cell polarity and mitotic spindle positioning. They provide spatial landmarks for the anchorage of polarity proteins such as SCRIB and DLG [16]. Disrupted adherens junctions due to low calcium lead to diffused cell polarity protein localization in mammalian epithelial cells [17]. In neuroepithelial cells of Drosophila, adherens junction defects cause a shift of cell division modes due to deviation of spindle positioning from the planar axis and uncoupling of spindle positioning from asymmetric basal protein localization [18]. In the Drosophila follicular epithelium, adherens junctions prompt the apical position of the midbody to achieve asymmetric cytokinesis [19]. While these studies implicate the importance of adherens junctions in cell polarity and spindle positioning, the evidence for adherens junction proteins, such as E-cadherin, as a bridge between the two events is lacking. Importantly, expression of E-cadherin is frequently lost or downregulated in human prostate cancers [20][21][22]. Whether and how the loss of E-cadherin leads to prostate tumorigenesis remain poorly understood.
In the present study, using a murine prostate-specific E-cadherin knockout model, we find that E-cadherin connects the cell polarity and spindle positioning to ensure the horizontal symmetric division of luminal cells and the integrity of prostate epithelia. Importantly, we show that a cell polarity protein SCRIB is recruited by E-cadherin to form an E-cadherin/SCRIB/LGN complex to guarantee the proper cell division mode and to prevent prostatic carcinogenesis.

Loss of E-cadherin leads to a hyperproliferative phenotype of prostatic luminal cells and development of prostate adenocarcinoma in aged mice
To unravel the role of adherens junctions in prostate epithelial development and homeostasis, we first examined expression patterns of E-cadherin, a key component of adherens junctions, at different developmental stages via triple staining for E-cadherin, the basal prostatic cell-specific marker p63 and the luminal cell marker CK8. We found that E-cadherin was expressed throughout the epithelium at postnatal day 5 (P5), and then more predominantly expressed in the lateral cell membrane of tightly-contacted prostatic luminal cells than scattered basal cells over time (S1 Fig). Using the Cre-LoxP recombination system, we generated a mouse model in which E-cadherin was specifically knocked out in the prostate epithelium by crossing Cdh1 fl/fl mice with probasin-cre transgenic mice. Immunostaining results confirmed an efficient E-cadherin deletion in the prostate of Pcre;Cdh1 fl/fl mice (Fig 1A and 1B). Through immunofluorescent staining of Ki67, a mitotic cell marker, we observed a significant increase of proliferative luminal cells from different lobes of Pcre;Cdh1 fl/fl mice at P5, P10 and p15 or during adult prostate regeneration, compared with their littermate controls (Fig 1C and 1D, S2A and S2B Fig and S1 Table). In contrast, significantly elevated cell proliferation was only detected in p63 + cells from P5 but not P10 or P15 Pcre;Cdh1 fl/fl mouse prostates (Fig 1C and 1D, S2A and S2B Fig and S1 Table). This is possibly due to the minimal expression of E-cadherin in basal cells at P10 and P15. Those data suggest that E-cadherin deficiency leads to a hyperproliferative phenotype of prostatic luminal cells.
To determine whether this hyperproliferation led to prostate cancer initiation, we carefully followed prostates of Pcre;Cdh1 fl/fl mice and their control littermates from 6 weeks to 21 months postnatally. Immunohistochemical staining analysis showed that at 6 weeks, different prostate lobes (anterior, ventral and dorsolateral lobes) of Pcre;Cdh1 fl/fl mice developed multifocal epithelial hyperplasia and multilayered polyp-like structure (S2C Fig). We detected lowgrade murine intraepithelial neoplasia (mPIN) in 4 months old Pcre;Cdh1 fl/fl mice (Fig 2A). By 9 months of age, Pcre;Cdh1 fl/fl mice developed high-grade mPIN, characterized by increased number of atypical cells with nuclear enlargement, prominent nucleoli and an elevated nuclear to cytoplasmic ratio ( Fig 2B). As shown in Fig 2B-2D and S2 Table, more than 60% of 9-month-old and 80% 21-month-old E-cadherin deficient prostatic lumens exhibited a hyperplastic phenotype. Of note, we detected invasive adenocarcinoma in the prostates from 37.5% of 21-month-old Pcre;Cdh1 fl/fl mice manifested by loss of p63 + basal cells, disruption of the basement membrane, as demonstrated by loss of laminin staining and invasion of epithelial cells into the surrounding prostatic stroma (Fig 2E-2G).
We further conducted RNA sequencing on prostates from E-cadherin knockout and control animals. Genes regulating cellular biological processes including cell proliferation, cell  Table). We performed additional qRT-PCR confirmation experiments. As shown in S3D Fig,  genes which act to promote cell proliferation and cell migration were more preferentially expressed in the E-cadherin null prostate epithelial cells, whereas molecules that facilitate cell junctions were significantly downregulated in E-cadherin knockout prostate epithelia. Moreover, we utilized the TCGA human prostate cancer database to compare the transcriptional signature between our E-cadherin knockout mouse model and human prostate cancer (S4 Fig). As shown in S4 Fig, we found that the transcriptional profile of E-cadherin knockout mouse prostates resembled the human prostate cancer. This data further substantiated the relevance of E-cadherin knockout mice as a valuable prostate cancer model. Together, those data provided additional molecular explanations to the hyperproliferative phenotype and invasive adenocarcinoma detected in E-cadherin knockout prostates.
Further immunofluorescent staining confirmed the E-cadherin deletion in actively proliferating prostate epithelial cells from Pcre;Cdh1 fl/fl mice. The vast majority of tumor cells in aged Pcre;Cdh1 fl/fl prostates also displayed an absence of E-cadherin expression, indicating that the tumors were derived from hyperproliferation of E-cadherin knockout cells (Fig 3A and 3B). However, E-cadherin deletion did not result in an epithelial to mesenchymal transition, as we did not observe an enhanced expression of mesenchymal cell marker vimentin in Pcre;Cdh1 fl/fl prostate epithelial cells (Fig 3C). Similar to human prostate cancer samples, we also occasionally detected apoptotic epithelial cells in Pcre;Cdh1 fl/fl mouse prostate tumors ( Fig 3D).

Horizontal division of prostate luminal cells is randomized by E-cadherin deletion during postnatal development and regeneration
Previous studies have shown that planar spindle alignment in epithelia is crucial for the maintenance of mitotic daughter cells within the plane of the tissue, whereas spindle orientation perpendicular to the basement membrane is required for asymmetric cell divisions and epithelial stratification in tissues such as skin [23]. The expansion of a monolayer developing prostate lumen is ensured by the horizontal division of luminal cells [5], disruption of which often causes multilayered polyps-like structure as detected in Pcre;Cdh1 fl/fl prostates. We therefore wondered whether loss of E-cadherin affected proper spindle alignment of dividing luminal cells during postnatal prostate development. To test this possibility, we analyzed the spindle orientation of dividing luminal cells by staining sections of developing prostates with p63 and survivin, an approach used for analysis of cell division modes [5]. We observed a more than 2-fold increase of tilted and vertical cell divisions in luminal cells in P5 E-cadherin knockout prostates (40.0% in Pcre;Cdh1 fl/fl mice versus 14.2% in control mice). Similar findings were obtained from P10 and P15 prostates (Fig 4A, 4C and 4D and S4 Table). In addition, during adult prostate regeneration, E-cadherin deletion resulted in randomization of the luminal cell division plane (Fig 4B  and 4E and S4 Table). Collectively, these findings suggest that E-cadherin is required for a proper mitotic spindle orientation during prostate development and regeneration.

Interference of E-cadherin expression results in spindle dis-orientation in the prostatic RWPE-1 cell line
To directly test whether E-cadherin loss led to defects in spindle positioning and subsequently cell division orientation, we modified a previously reported approach to record the spindle epithelial cells from ventral and dorsolateral prostatic lobes at different development stages or regeneration time points (n = 3, Student's t-test, ÃÃÃ P<0.001, ÃÃ P<0.01, Ã P<0.05,error bars = SEM.).
https://doi.org/10.1371/journal.pgen.1007609.g001 position of an immortalized prostate epithelial cell line RWPE-1 in real time [24,25]. RWPE-1 cells were co-transfected with an H2B-RFP and an α-tubulin-GFP plasmid to mark chromosomes and spindles respectively ( Fig 5A). Knockdown of E-cadherin was achieved by infection  of RWPE-1 cells with a shRNA containing lentiviral construct (Fig 5B). Consistent with the in vivo E-cadherin knockout data, we detected a higher proliferating rate in E-cadherin knockdown RWPE-1 cells compared to control cells (S5 Fig). Control or E-cadherin knockdown cells were plated on retronectin-coated dishes and imaged under a confocal microscopy equipped with a live-cell recording unit (Fig 5C). Four-dimensional movies of dividing RWPE-1 cells were analyzed to measure angles of mitotic spindles relative to the culture dish. In line with previous reports [26,27], we found that mitotic spindles of control RWPE-1 cells were positioned at a wide range of angles during prometaphase, but the majority of spindles (83.5%) exhibited a parallel orientation to the substrate from the late metaphase to telophase (Fig 5D and 5E, S4 Table and S1 Movie). This horizontal position of spindles ensured a cell division cleavage plane perpendicular to the substrate. However, we detected a striking increased percentage of E-cadherin-knockdown cells which could not correctly position their spindles in telophase, thereby caused a randomized cell division cleavage plane (Fig 5D and  5E, S4 Table and S1 Movie). These in vitro experiments substantiated our in vivo observations that adherens junction protein E-cadherin was indispensable for the proper positioning of mitotic spindles in prostate epithelial cells.

Ablation of E-cadherin disrupts cellular distribution and association of key spindle positioning proteins LGN and NUMA
Luminal epithelial cells of a developing prostate divide within the plane of the epithelium by orienting spindle poles towards the lateral membrane. Intensive research has revealed that an evolutionarily conserved LGN/NUMA complex plays an essential role in directing the cortical attachment of spindle astral microtubules to dynein in variety tissues from both invertebrates and vertebrates [9,[27][28][29]. In agreement with previous reports, we found that LGN and NUMA formed two crescents underneath the lateral membrane and in parallel to the basal surface in mitotic wild-type prostate luminal cells (Fig 6A-6D). However, in the absence of Ecadherin, distribution of LGN and NUMA in dividing luminal cells can be detected all around the cell cortex (Fig 6A-6D and S5 Table), suggesting that disoriented mitotic spindle alignment was due to the diffused localization of LGN/NUMA complex. In addition, the proteinprotein interaction between LGN and NUMA was also severely abrogated after E-cadherin deletion ( Fig 6E). As a result, the luminal cells underwent perpendicular or oblique divisions, which disrupted the monolayer epithelial structure and formed the disorganized, multilayered polyps-like structure (Fig 1 and S2 Fig).

Loss of E-cadherin impairs the cell polarity of prostate luminal cells during prostatic development
Cell polarity provides spatial cues to guide spindle orientation [10,12,14,30]. Disruption of cell polarity is often seen during the course of cancer initiation [11,[31][32][33][34][35]. Given the observation that horizontal mitotic spindle positioning of prostate luminal cells was severely disrupted in the Pcre;Cdh1 fl/fl prostate, we therefore wondered whether E-cadherin affected luminal cell polarity. We next investigated the impact of E-cadherin loss on the expression and distribution of PAR complex and SCRIB complex. Utilizing immunofluorescent staining, we detected core components of the PAR complex, PAR3 and aPKC were localized in the apical domain of wild-type prostate luminal cells, but diffusedly distributed in Pcre;Cdh1 fl/fl prostate luminal cells (Fig 6F and 6G and S6A and S6B Fig). Likewise, the normal basolateral distribution of polarity proteins DLG-1 and SCRIB in prostate luminal cells were markedly impaired after Ecadherin deletion (Fig 6H and 6I and S6C and S6D Fig). In addition, the formation of SCRIB/ DLG polarity protein complex was disrupted following E-cadherin deletion (S6E Fig). Together, these results indicated that loss of E-cadherin exerted a devastating influence on the formation and maintenance of prostate luminal cell polarity.

E-cadherin recruits SCRIB to form an E-cadherin/SCRIB/LGN complex to bridge cell polarity and spindle positioning
Previous research has demonstrated that adherens junctions are necessary for the establishment of cell polarity and positioning mitotic spindles symmetrically in Drosophila [18]. Given the aforementioned findings that E-cadherin deletion or knockdown led to cell polarity loss and spindle dis-orientation, we proposed a hypothesis that E-cadherin may serve as a central molecule to bridge cell polarity and spindle positioning. It was reported that in mammalian epithelial cells, restriction of SCRIB to the lateral cell-cell junction is dependent on E-cadherin and that SCRIB reciprocally regulates E-cadherin-mediated cell adhesion [16,36,37]. On the other hand, lateral localization of the SCRIB polarity complex determines the planar orientation of mitotic spindles [14,15,38,39]. Therefore, we then tested whether there were physical interactions between E-cadherin and SCRIB. Co-IP assay showed endogenous interactions between E-cadherin and SCRIB in RWPE-1 cells (Fig 7A and 7B). In addition, we observed that E-cadherin associated with LGN (Fig 7A and 7C). To elucidate whether possible interactions existed between the SCRIB complex and the LGN complex in prostate epithelial cells, we carried out additional co-IP experiments. Intriguingly, we found that endogenous SCRIB interacted with LGN, which can be suppressed by E-cadherin knockdown (Fig 7B and 7D). Furthermore, the association between LGN and E-cadherin was markedly attenuated by SCRIB knockdown (Fig 7E). SCRIB knockdown led to an increase in RWPE1 cell proliferation (S7 Fig). To determine whether the interactions between E-cadherin and SCRIB, as well as LGN with SCRIB were direct, and which domains of SCRIB were responsible for the interactions, we designed and performed GST-pull down experiments with different truncated fragments of SCRIB. As show in Fig 7F and 7G, the fragment containing the PDZ domain of SCRIB directly bound to both the E-cadherin and LGN. Collectively, these data revealed a ternary protein complex of E-cadherin/SCRIB/LGN by which adherens junctions were able to tightly regulate and efficiently connect cell polarity complexes and spindle orientation determinants so to precisely position mitotic spindles in dividing cells (Fig 8).

Discussion
At the current study, we find that genetic deletion of E-cadherin in the mouse prostate results in cell polarity loss, mitotic spindle dis-orientation, and monolayer structure disruption in the luminal epithelium. Importantly, E-cadherin knockout leads to prostatic hyperplasia which progresses to invasive adenocarcinoma in aged mice. Mechanistically, E-cadherin can recruit SCRIB to form a protein complex with LGN by directly binding to the PDZ domain of SCRIB to restrict SCRIB and LGN to the lateral cell membrane. These findings provide a novel mechanism that E-cadherin acts as a vital bridge between cell polarity and spindle orientation for the maintenance of a normal prostate epithelial architecture.
In both Drosophila and mammalian systems, adherens junctions, cell polarity and mitotic spindles orientation are intricately intertwined biological processes. However, how these three cartoon illustrates how the mitotic spindle angle of cultured RWPE-1 cells relative to the retronectin-coated dish bottom is measured.

(D) Representative confocal images of α-tubulin-GFP marked spindles and H2B-RFP labeled chromosomes from control or Ecadherin knockdown RWPE-1 cells in different cell cycle phases (Scale bars are 20μm). (E) Quantification of metaphase and telophase spindle angles indicated that E-cadherin knockdown causes a remarkable increase of RWPE-1cells with vertical or oblique mitotic spindle alignment in the telophase.
https://doi.org/10.1371/journal.pgen.1007609.g005 E-cadherin knock-out leads to disoriented prostate luminal cell division and tumorigenesis in mice fundamental attributes of the epithelium are connected remains elusive. In particular, limited previous studies regarding this question were mostly carried out in cell lines or in invertebrates. The current study was carried out in a genetically engineered mouse model. Our examination of early murine postnatal prostate development and adult regeneration reveals that Ecadherin deletion results in loss of luminal cell polarity and mitotic spindle dis-orientation in vivo. In addition, other than microscopic examination of cell mitosis at limited time points, we used a real-time recording system which allows us to follow the whole process of individual cell division. By this way, we further demonstrate vividly that E-cadherin is required for the horizontal alignment of spindles in the telophase of dividing RWPE1 cells in vitro. Critically, we uncover a central role of adherens junction protein E-cadherin in the coordination of cell polarity and mitotic spindles orientation by forming an E-cadherin/SCRIB/LGN protein complex. E-cadherin knockout leads to a dis-assembly of the SCRIB polarity complex and the LGN/NUMA complex, causing a subsequent randomization of the luminal cell division plane and forming of multilayered, disorganized epithelia, a common feature of the early stage of prostate tumorigenesis.
During the preparation of our manuscript, Nelson, W.J.'s group reported that, based on biochemical analysis and examination of MDCK and U2OS cell lines, E-cadherin instructs cell division orientation by binding to LGN and brings LGN to cell-cell adhesions, [40]. In a related study on MDCK cells, Hart, K.C. et al reported that E-cadherin acts to sense mechanical tension across an epithelial sheet and facilitate polarized cortical distribution of LGN to align cell divisions [41]. However, whether this LGN/E-cadherin complex is important for epithelial tissue integrity and the consequences of disruption of this interaction in vivo is unknown. In addition, whether additional components, such as cell polarity, are required to the intricate regulation of LGN by E-cadherin is not determined. Our current study shed new important lights on these aspects:1) We demonstrated clearly in vivo that E-cadherin knockout causes LGN/NUMA dissociation and dis-localization during luminal cell mitosis, which subsequently leads to luminal cell division plane dis-orientation and serious impairment of normal prostate architecture; 2) We provide direct evidences that cell polarity protein SCRIB is required for the efficient complex-forming between E-cadherin and LGN in prostate epithelial cells, as SCRIB knockdown markedly reduces the association of E-cadherin and LGN. Moreover, we found that both E-cadherin and LGN can directly bind to the fragment containing PDZ domain of SCRIB by pull down assays. Of note, our experiments also provide a novel molecular explanation for previous reports regarding how SCRIB is required for the establishment of E-cadherin-mediated cell-cell adhesion and correct positioning of mitotic spindles [16,36,37].
Defects in precisely controlled cell-cell adhesion and junctions, cell polarity and mitotic spindles position are closely associated with developmental disorders, and may contribute to cancer initiation and progression [34,42,43]. For example, conditional knockout of cell polarity protein SCRIB in mouse prostates has been reported to lead to prostatic intraepithelial neoplasia [44]. This is consistent with our observation that SCRIB knockdown promotes RWPE1 cell proliferation. Decreased expression or genetic loss of adherens junction molecule E-cadherin is frequently found in various cancer types including gastric carcinomas, lung cancer  and breast cancer [45][46][47]. In prostate cancer, E-cadherin down regulation is significantly associated with advanced stages and tumor metastasis [20][21][22]. A recent study using a NKX3.1 cre-ERT2;Cdh1 fl/fl mouse model showed that tamoxifen induced partial E-cadherin deletion in luminal cells from 8-week old mice leads to a short-term of anoikis [48]. While we do also detect some apoptotic cells in hyperplastic lumen of Pcre;Cdh1 fl/fl prostates in this study, we demonstrate that a more efficient prostate epithelium deletion of E-cadherin leads to luminal epithelial cell hyperplasia. Those apoptotic luminal cells may come from actively perpendicular or oblique dividing cells that lose cell-cell contact or cell-basement membrane contact. Apoptotic luminal cells can also be frequently found in human prostate cancer samples. A similar simultaneous increase in both proliferation and apoptosis was reported in PAR3 deletioninduced mammary gland hyperplasia [35]. In addition, our observation that invasive prostate carcinoma develops in old E-cadherin knockout mice is in line with a clinical correlation between E-cadherin deletion or downregulation and human prostate cancer progression [20,21]. Moreover, immunofluorescent staining of developmental prostate or tumor tissues from Pcre;Cdh1 fl/fl mice confirms E-cadherin deletion in most proliferating luminal cells, suggesting that the hyperplasia is not derived from compensatory proliferation of residue E-cadherin expressing cells. Therefore, loss of E-cadherin induced cell polarity and cell division plane deregulation can strongly predispose mice for prostate tumorigenesis. These findings provide evidence for an important role of E-Cadherin not only in anchorage of cell polarity proteins with spindle positioning determinants, but also in prostatic carcinogenesis.

Ethics statement
All mice were maintained and utilized according to the ethical regulations at Ren Ji Hospital. The animal protocols were approved by the Ren Ji Hospital Laboratory Animal Use and Care Committee. CO 2 asphyxiation is used for sacrificing mice.

Experimental animals
Probasin-Cre and Cdh fl/fl mice were introduced from National Cancer Institute (NCI:01XF5) and Jackson laboratory (JAX:005319) respectively. Prostates of male mice were dissected for histology and immunostaining.

Plasmids and lentivirus production
The shRNA against E-cadherin, SCRIB and scramble shRNA were cloned into a lentiviral vector pLVTH. Lentivirus was produced as previously reported [49]. RWPE-1 cells were infected with lentivirus and selected with 3ug/ml puromycin for 2weeks to generate the stable E-cadherin and SCRIB knockdown cell line, respectively. The H2B-RFP (#26001) and a-tubulin-GFP (#64060) plasmids were purchased from Addgene, which were originally deposited by Dr. Beronja S. and Dr. Yang W.

Castration and androgen replacement
Pcre;Cdh1 fl/fl mice and their control littermates were surgically castrated. Three weeks after castration, dihydrotestosterone (MCE, HY-A0120) dissolved in sterile corn oil was given via intraperitoneal injection twice each day (50ug/d) to induce prostate regeneration. Prostate tissues were dissected for section and immunofluorescent staining at different regeneration stages (48 hours or 60 hours post testosterone administration).

Immunofluorescence staining
Mouse prostates were fixed in 4% paraformaldehyde for 20 minutes and dehydrated in 30% sucrose solution overnight. Tissues were then embedded in Optimal Cutting Temperature (O. C.T.) compound and quickly frozen in a -80˚C refrigerator for 10min. Frozen sections were cut at a thickness of 6μm. The sections were washed with PBS and subjected to a heat-induced epitope retrieval step in 0.01M sodium citrate (PH 6.0). Then sections were transferred to a blocking solution (PBS with 0.2% TritonX-100 and 10% donkey serum) for 1 hour at room temperature. Primary antibodies diluted in PBS with 0.2% TritonX-100 and 1% donkey serum were then applied to the section overnight at 4˚C and washed away with PBS three times. Sections were then incubated with secondary antibodies, conjugated to Alexafluo-488, 594 or 546 for 1 hour at room temperature. After thorough washing, sections were mounted with Vector Shield mounting medium containing DAPI.

Immunohistochemical staining
Paraffin-embedded tissue sections were deparaffinized, rehydrated and subjected to a heatinduced epitope retrieval step in 0.01M sodium citrate (PH 6.0). Endogenous peroxidase activity was ablated by using 3% hydrogen peroxide. Sections were blocked, stained with primary antibodies then horseradish peroxidase conjugated secondary antibodies as described above. Then DAB staining was undertaken according to the manufacturer's instructions (DAB Staining kit, GK347010, Gene Tech (Shanghai) Company Limited). Sections were washed under running tapping water for 5min then counterstained with hematoxylin, followed by dehydration and mounting with the neutral balsam mounting medium.

GST-Pull down
Human SCRIB full length was cloned from the cDNA of the RWPE-1 cell line. GST-tagged fragments of human SCRIB were cloned into the pET-49b(+)vector with SpeI and EcoRI restriction enzymes. DNA constructs were transformed into BL21 (DE3) E.coli for following recombinant protein expression. For protein purification, glutathione sepharose beads (GE Healthcare) were added to the recombinant protein containing bacterial lysate and incubated for 1.5 hours at 4˚C. Beads were washed with 500μl cold bacteria lysis buffer twice. RWPE-1 cell lysate was added to the beads and incubated overnight at 4˚C. After thorough wash, protein loading buffer were added to the beads and boiled for 10 minutes at a 100˚C metal bath. The immunoprecipitates were analyzed by SDS-PAGE and immunoblotting with GST, or indicated antibodies.

Immunoprecipitation and immunoblotting
For immunoprecipitation, RWPE-1 cells were washed with ice-cold phosphate-buffer saline and lysed in a lysis buffer (50mM Tris-HCL, 150mM NaCl, 1mM EDTA, 0.25% TritonX-100, PH7.4) supplemented with protease and phosphatase inhibitors (Roche, Penzberg, Germany). Cell lysates were incubated with 1μg primary antibodies for 6 hours at 4˚C.Rabbit or mouse immunoglobulin G (Sigma-Alrich, USA) was utilized as controls. Activated Protein G-Agarose (Roche, 11243233001) were added into the protein lysate and incubated for another 1 hour at 4˚C. Then the immunoprecipitates were washed 3 times with the lysis buffer and denaturalized for 10 minutes at a 100˚C metal bath. The Immunoprecipitation results were detected by immunoblotting assay. Protein samples were separated via SDS-PAGE electrophoresis and then transferred to PVDF membranes. The membranes were blocked with 5% non-fat milk in TBST (TBS containing 0.1% Tween 20) and incubated with primary antibodies at 4˚C overnight. Membranes were then rinsed with TBST for 3 times and incubated with secondary antibodies for 1 hour at room temperature. After washing with TBST 3 times, membranes were exposed with enhanced chemiluminescence substrates (Thermo Scientific). Antibodies used in the study were listed in S6 Table.

Confocal microscopy for images and videos
Lentivirus-transfected RWPE-1 cells were seeded on glass-bottom dishes coated with 0.1μg/μl Retronectin (Takara Bio). Images were acquired every 6-7 minute with a xyzt acquisition mode using an Axio Observer under a Z1 microscope with the LSM 700 scanning module (Zeiss). Cell cultures were maintained at 37˚C and 5% CO2 incubator for live imaging.

RNA extraction and quantitative-PCR analysis
To collect single prostate epithelial cells suspension for flow cytometric sorting of basal, luminal and stromal cells, prostates from 12-week-old Pcre;Cdh1fl/fl mice and control littermates were harvested, minced and incubated with 3ml 1mg/ml Collagenase solution on a shaker at 37˚Cfor 2 hours. The cells were then spin down and further digested with 2ml 0.25%Trypsin/ EDTA at a 37˚C water-bath for another 6minutes. Single cells were obtained by filtering the mixture through a 40μm cell strainer. Percp-lineage, FITC-CD49f, APC-Sca-1 antibodies were applied to the cell suspension and incubated in the dark for 40 minutes. Basal (Lineage -Sca-1 + CD49f + ), luminal epithelial cells (Lineage -Sca-1 -CD49f + ) and stromal cells (Lineage -Sca-1 + CD49f -) were sorted using a BD FACSARIA system. Total RNA were extracted from each cell population using Trizol reagents following instructions from the manufacturer (Life technologies). cDNA was synthesized from the extracted total RNA using PrimeScript RT Reagent Kit(RR037A). qPCR was performed using SYBR Premix Ex Taq (TliRNaseH Plus) (RR420A). Relative transcript abundance was determined by the comparative CT method using Actin as a reference gene. Primers used in RT-PCR were listed in S7 Table.

RNA-seq and data analysis
Single Lineage -Epcam + PIprostate epithelial cell suspension from 3-mo old Ecadherin knockout mice and their control littermates were FACS sorted before processing to total RNA extraction, quality control and library preparation. Libraries were constructed using NEBNext Ultra Directional RNA Library Prep Kit for Illumina (NEB, USA) following manufacturer's recommendations and index codes were added to attribute sequences to each sample. The clustering of the index-coded samples was conducted on a cBot Cluster Generation System using TruSeq PE Cluster Kit v3-cBot-HS (Illumina) according to the manufacturer's recommendations. The libraries were sequenced on an Illumina Hiseq platform and paired-end reads were obtained. The software btrim was utilized to discard low quality reads whose insert size were less than 30, or average quality score within a moving window were less than 20. The clean data were mapped to the Mus musculus reference genome (GRCm38). Gene expression levels and differential expression were analyzed using the Cufflinks and Cuffdiff program, respectively, in the Cufflinks tool suite (v2.2.1) with default parameters. The functional annotation of differentially expressed genes was performed using the database for annotation, visualization and integrated discovery software (DAVID v6.8). The RNA-seq raw data in this paper are available at the GEO web with the accession number GSE115204.

Quantification and statistical analysis
The ImageJ 1.46r software was used to determine the positive stained cells in immunostaining images and measure angles of mitotic spindles relative to the retronectin base. Microsoft Excel and Graph Pad Prism5 were used for data compilation and graphical representation. All bar graphs, line graphs and dot plots are represented as mean ± standard deviation. All statistical analysis was done using a two-tailed Student's T-test and a p-value<0.05 was considered significant and indicated by a star ( Ã ) mark.