A novel role for PRL in regulating epithelial cell density by inducing apoptosis at confluence

Epithelial cells are tightly connected to each other through cell-cell junctions and form a sheet-like structure. The cell density of an epithelial sheet is exquisitely homeostatic, which is essential for its coordinated function in shaping and maintaining the body of multicellular organisms. One of the classical mechanisms that maintain optimum cell density is contact inhibition of proliferation or, in short, contact inhibition, whereby cells stop dividing when they come in contact with each other at confluence (Abercrombie and Heaysman, 1954; Eagle and Levine, 1967). The mechanism of contact inhibition has been extensively studied, and the implication of various molecules affecting diverse cellular functions, such as cell adhesion, polarization and proliferation, has been reported previously (McClatchey and Yap, 2012). In particular, the importance of Hippo signaling is recognized in transmitting the information of cell-cell contact to the nucleus, which is crucial for contact inhibition. In brief, mechanical cues arising from cell adhesion and the cytoskeleton can mediate the downstream activation of large tumor suppressor (LATS), directly or indirectly through mammalian Ste20-like kinase (MST) or mitogen-activated protein kinase kinase kinase kinase (MAP4K) (Ma et al., 2019). Activated LATS then phosphorylates the transcriptional co-activator Yes-associated protein (YAP) and prevents its nuclear translocation, and the subsequent transcriptional activation of growth-promoting target genes (Zhao et al., 2007, 2008). Alternative mechanisms of maintaining optimal cell density, through extruding the apoptotic or live cells from the epithelial sheet, have also been demonstrated. Apoptotic cell extrusion was originally reported in chick embryonic epithelium and in the cultures of Madin–Darby canine kidney (MDCK) cells, which show typical characteristics of epithelial cells (Rosenblatt et al., 2001), and overcrowding-induced live cell extrusion was first identified in the epithelium of Drosophila melanogaster and MDCK cell cultures (Marinari et al., 2012; Eisenhoffer et al., 2012). In both cases, the extrusion of cells is preceded by the production of sphingosine-1 phosphate (S1P), which then activates S1P receptor 2 (S1P₂) on the neighboring cells, and induces Rho-associated coiled-coil containing protein kinase (ROCK)-dependent actomyosin contraction, to expel cells from the epithelial sheet (Gu et al., 2011; Eisenhoffer et al., 2012). Furthermore, at high cell density, the activation of Piezo-1, a stretch-activated calcium channel, triggers the overcrowding-induced live cell extrusion (Eisenhoffer et al., 2012). On the other hand, what drives cell death and how it is regulated in apoptotic cell extrusion are not understood well.

Phosphatase of regenerating liver (PRL, also known as protein tyrosine phosphatase 4) is a family of prenylated proteins anchored to the plasma membrane, and has three members (PRL1, PRL2 and PRL3, also known as PTP4A1, PTP4A2 and PTP4A3, respectively) in mammals (Diamond et al., 1994; Zeng et al., 1998). Saha et al. (2001) reported that PRL3 is a highly overexpressed gene in the metastases of all colorectal cancers examined, and a number of later studies reported a link between overexpression of PRL and malignant progression of several types of cancers (Bessette et al., 2008). Several studies using cultured cancer cells showed that PRL overexpression can promote cell proliferation, migration and invasion (Bessette et al., 2008), suggesting a role in promoting cancer progression. On the other hand, some studies have reported tumor-suppressive actions of PRL. Prl3 is a direct target gene of p53 tumor suppressor in primary mouse fibroblasts, and its overexpression negatively regulates cell cycle progression (Basak et al., 2008). In D. melanogaster, overexpression of PRL-1, a sole PRL ortholog, counteracts the oncogenic effect of Src (Pagarigan et al., 2013). These observations suggest that, in non-transformed cells, PRL probably has a function other than promoting cancer malignancy. To examine the normal physiological functions of PRL, gene knockout (KO) mice have been generated. Prl1-KO and Prl3-KO mice developed normally and showed no obvious phenotypic alterations (Bai et al., 2016; Zimmerman et al., 2013), whereas Prl2-KO mice showed mild changes in the phenotype, such as growth retardation and reduced fertility (Dong et al., 2012, 2014). Double KO mice for Prl1 and Prl2 were embryonic lethal (Bai et al., 2016). These observations suggest that PRLs are essential for normal embryonic development, but the functional details are not clear. In our previous study, we established MDCK epithelial cells with doxycycline (Dox)-inducible PRL-expression, which showed unique properties when stimulated with growth factors or under acidic conditions (Kojima et al., 2019; Funato et al., 2020). However, under normal culture conditions, PRL expression did not affect the basic properties of cells, such as proliferation rate, overall morphology and cell-cell adhesion. As is often the case with in vitro culture experiments, cells were seeded and cultured relatively sparsely in these studies. Therefore, in the present study, we performed detailed analyses of cells after they reached confluence. From a series of experiments, we found that PRL regulates the cell density in epithelial sheets by promoting apoptosis, mediated by the translational elevation of E-cadherin expression and downstream activation of the TGF-β pathway. Moreover, using zebrafish, we found that PRL plays a critical role in the control of epithelial cell density during convergent extension, a critical process in the early development of vertebrate embryos.

To examine the effect of the expression of PRL in epithelial cells, we used two independent MDCK cell lines that can express GFP-tagged PRL3 (PRL #1 and PRL #2) when induced with Dox (Kojima et al., 2019). We cultured the cells after they reached confluence and observed changes in morphology. As shown in Fig. 1A, culture of the control cells showed the sporadic formation of round and swollen structures, which was completely suppressed upon PRL expression. Similar results were observed when double the number of cells were seeded (Fig. 1B). We also observed the formation of these swollen structures when the cells were cultured without Dox, thereby confirming the effect of PRL expression (Fig. 1A). We tested the effects of expressing C104S, a mutant PRL3 that lacks both phosphatase activity and Mg²⁺ transporter Cyclin M (CNNM)-inhibiting activity, which made it functionally inactive in promoting tumor development (Guo et al., 2004; Kozlov et al., 2004; Funato et al., 2014). The results showed the formation of swollen structures as in control cells, suggesting the importance of functional PRL in this process (Fig. 1A). Furthermore, we investigated the sheet structure in detail by staining filamentous actin (F-actin) with phalloidin, and constructed three-dimensional images through confocal microscopy. The interior of the swollen structure was empty, as shown in the images of horizontal sections (Fig. 1C). Vertical sections, reconstructed by stacking the images of the horizontal sections, showed that each cell sheet was single layered, but some regions in the cultured sheets were found to be detached from the surface of glass coverslips. From these observations, we confirmed that the round and swollen structures observed were ‘domes’ containing a fluid-filled internal space, as described previously (Leighton et al., 1969).

Domes are formed in the confluent culture of MDCK cells due to the transport of salt and water across the epithelial monolayer, which accumulate between the epithelial sheet and culture dish (Cereijido et al., 1981). As the creation of a dome requires the formation of a functional epithelial barrier, establishment of apicobasal polarity and high cell density in the culture (Misfeldt et al., 1976; Cereijido et al., 1980), we next examined whether these requirements were met in the PRL-expressing cells. First, we evaluated the functional epithelial barrier by measuring transepithelial electrical resistance (TEER), a widely accepted quantitative parameter to measure the integrity of tight junctions across the epithelial sheet (Srinivasan et al., 2015). The TEER values of the control cells and PRL-expressing cells were 83.6±4.1 Ωcm² and 317.9±14.8 Ωcm² (mean±s.d.), respectively, which established the high TEER of PRL-expressing cells, and suggested the presence of a functional epithelial barrier. Next, to test the integrity of apicobasal polarity, we performed immunostaining for tight junction marker zonula occludens-1 (ZO-1, also known as TJP1) (Stevenson et al., 1986), adherens junction markers E-cadherin (Takeichi, 1977) and β-catenin (McCrea et al., 1991), and basolateral marker Na⁺/K⁺ ATPase (Skou, 1957) (Fig. S1), and found no alteration in the expression pattern of any of them. This suggested that the establishment of polarized epithelium was not affected by the status of PRL expression. Therefore, we focused on the cell density of each culture.

We stained the nuclei in each cultured cell sheet to determine cell density. We found that PRL expression led to lower cell density, whereas the expression of the C104S mutant was ineffective in this regard (Fig. 2A). These results suggest that PRL is important for lowering the cell density of epithelial sheets. By changing the number of cells used for seeding, we found that each cell culture has its own upper limit of epithelial cell density (Fig. 2B). Notably, the maximum cell density was significantly decreased with the induction of PRL expression. When we compared the proliferation status of cells at 24 h and 48 h after Dox treatment, we found that the percentage of proliferating cells was not suppressed in PRL-expressing cells compared to control cells (Fig. S2A,B). We next examined through a cytotoxicity assay whether PRL expression promoted cell death. The results showed that when the cells were seeded at high density, PRL expression enhanced the release of lactate dehydrogenase (LDH), a cytosolic enzyme, from the damaged cells (Fig. 2C). Interestingly, LDH release in the PRL-expressing cells was critically dependent on cell density, as it decreased significantly at lower cell density. We then determined the extent of cell death by immunostaining with cleaved caspase-3. Again, expression of PRL at high cell density greatly stimulated cell death (Fig. 2D,F), whereas it was suppressed at low cell density (Fig. 2E,F), thus proving the density-dependent induction of cell death by PRL. To corroborate this finding, we seeded cells in a silicon chamber, and stretched the chamber to artificially reduce epithelial cell density (Fig. 2G). Cell death was drastically suppressed when epithelial cell density was lowered by stretching the chamber (Fig. 2H,I), thereby demonstrating the importance of cell density in PRL-induced cell death. Consistently, we found that the number of cells surrounding cleaved caspase-3⁺ cells was significantly larger than those surrounding cleaved caspase-3⁻ cells, suggesting that cells tend to die in more confluent areas within the culture sheet of PRL-expressing cells (Fig. S2C). Next, to examine the morphological features of dying PRL-expressing cells, we stained the cells for F-actin and cleaved caspase-3, and observed them at high magnification. Most of the cleaved caspase-3⁺ cells demonstrated typical characteristics of apoptosis, such as chromatin condensation and cell shrinkage (Fig. S2D), thereby confirming the stimulation of apoptosis upon PRL expression. Besides, in PRL-expressing cells, a small number of cleaved caspase-3⁺ cells were still present in the epithelial layer, whereas most of them were found ‘extruding’ from the epithelium (Fig. S2E). On the other hand, live (cleaved caspase-3⁻) cells extruding from the epithelium were rarely found (Fig. S2E). Collectively, these results indicate that cells die in the epithelium, and this is immediately followed by apoptotic extrusion.

Next, to confirm the importance of endogenous PRL in density-dependent cell death, we knocked down endogenous PRL1 and PRL2, the two most abundant PRL isoforms expressed in MDCK cells, with small interfering (si)RNA (Kojima et al., 2019). For this, we analyzed the cells fixed at 48 h after Dox treatment, when the rate of apoptosis is significantly higher in control cells. Transfection with two different siRNAs for PRL1 or PRL2 significantly reduced PRL1 or PRL2, respectively (Fig. 3A). In fact, knocking down either PRL1 or PRL2 reduced apoptosis significantly (Fig. 3B,C), and increased cell density concomitantly (Fig. 3B,D). It should be noted that the effect of knocking down PRL2, which is expressed more than PRL1 in MDCK cells (Fig. 3A), was more drastic. However, double knockdown of PRL1 and PRL2 seemed to damage the cells, and they became sparse (Fig. S3A), hindering further analyses. Overall, we concluded that endogenous PRL also regulates epithelial cell density, probably through induction of apoptosis. Also, we found an increase in the total expression level of endogenous PRL proteins at high cell density compared to that at low cell density (Fig. S3B), consistent with the physiological role of PRL in maintaining epithelial cell density at normal levels.

As shown in Fig. S1, immunostaining of PRL-expressing cells with anti-E-cadherin antibodies showed a brighter staining signal. As expected, immunoblotting with anti-E-cadherin antibodies showed a significant increase in E-cadherin expression (Fig. 4A). Several studies have reported that increased expression of E-cadherin induces cell death (Lu et al., 2014; Akieda et al., 2019). We determined the effect of E-cadherin knockdown on density-dependent cell death. As shown in Fig. 4B, transfection with E-cadherin siRNA targeting two different sequences, suppressed E-cadherin expression. Immunostaining with anti-cleaved caspase-3 antibody showed a significant suppression of apoptosis in PRL-expressing cells (Fig. 4C,D). Further, we also examined the effect of direct overexpression of E-cadherin on cell death at high density using MDCK cells stably expressing GFP-tagged E-cadherin established in a previous study (Hoshino et al., 2004). We found that these cells constitutively expressed E-cadherin at similar levels to PRL-expressing cells (Fig. S4A). Such high and constitutive expression of E-cadherin might induce cell death just after reaching confluence, so we quantified the death of cells, which are not only attached to the culture dishes but also in the supernatant, by luminescent assay. The result showed that overexpression of E-cadherin solely can significantly increase the activity of caspase-3/7 (Fig. S4B), thereby demonstrating the importance of E-cadherin as a downstream mediator of PRL-induced apoptosis.

To understand how PRL increases E-cadherin expression, we first measured the levels of E-cadherin mRNA by qPCR but found no significant increase in PRL-expressing cells (Fig. S4C). Stability of the E-cadherin protein was also not increased by PRL expression (Fig. S4D,E). Therefore, we speculated that the translation of E-cadherin mRNA was probably enhanced in PRL-expressing cells. Autoradiography of total proteins, with 30 min radiolabeling using [³⁵S]methionine and [³⁵S]cysteine, showed comparable labeling in control and the PRL-expressing cells (Fig. 4E), suggesting no significant difference in global protein synthesis. Immunoprecipitation of each radiolabeled lysate with anti-E-cadherin antibodies (Fig. 4F), showed a fivefold increase in the newly synthesized E-cadherin in PRL-expressing cells (Fig. 4G), indicating a significant enhancement of E-cadherin translation in PRL-expressing cells.

We next sought to understand what drives apoptosis downstream of E-cadherin. Several studies have reported E-cadherin-dependent activation of the TGF-β pathway (Andl et al., 2006; Akieda et al., 2019), which, in turn, triggers cell death downstream. In the canonical TGF-β pathway, ligand binding stimulates TGF-β receptor type II (TGFBR2) to complex with the type I receptor [TGFBR1, also known as anaplastic lymphoma kinase 5 (ALK5)], which allows TGFBR2 to phosphorylate the kinase domain of ALK5. Activated ALK5 propagates the signal further by phosphorylating the receptor-regulated effector proteins (R-Smads, consisting of SMAD2 and SMAD3). Phosphorylated R-Smads then recruit co-Smad (SMAD4) to form a trimolecular complex, which translocates into the nucleus and stimulates the transcription of the target genes (Shi and Massagué, 2003). To examine the activation status of the TGF-β pathway in PRL-expressing cells, we investigated the localization of SMAD2 by immunostaining. We found a ∼twofold increase in the nuclear to cytoplasmic SMAD2 signal ratio, which is comparable to the levels after TGF-β stimulation (Fig. 5A,B). The importance of E-cadherin in the nuclear translocation of SMAD2 was confirmed by siRNA-mediated knockdown of E-cadherin, which resulted in a partial, but significant, reduction in the signal ratio (Fig. 5C,D). When ALK5 was inhibited in PRL-expressing cells using two chemical inhibitors, ALK5 inhibitor II and SB525334, both of which are known to suppress TGF-β signaling (Ichida et al., 2009; Laping et al., 2007), apoptosis was partially, but significantly, reduced (Fig. 5E,F), suggesting the involvement of ALK5 in the induction of apoptosis in PRL-expressing cells. In addition, siRNA-mediated knockdown of SMAD4, a component essential for the activation of TGF-β signaling (Liu et al., 1997), had similar effects (Fig. 5G-I). Next, we sought to examine how E-cadherin might activate the TGF-β pathway. Some previous studies have reported that E-cadherin interacts with TGFBR2 to activate the TGF-β pathway (Andl et al., 2006; Rudini et al., 2008). Therefore, we tested whether analogous interaction occurs in PRL-expressing cells. The immunoprecipitation of endogenous E-cadherin and subsequent immunoblotting with anti-TGFBR2 antibody showed weak, but significant, co-immunoprecipitation signal of TGFBR2 with E-cadherin using the PRL-expressing cell lysate (Fig. 5J,K). Overall, we found that activation of the TGF-β pathway downstream of E-cadherin contributes to drive apoptosis in PRL-expressing cells.

To examine the in vivo role of PRL in controlling epithelial cell density, we studied its effects on the embryonic development of zebrafish, in which changes in the epithelial cell density can be observed. Zebrafish has five isoforms of PRL genes: ptp4a1, ptp4a2a, ptp4a2b, ptp4a3a and ptp4a3b. As RNA-seq showed that the expression levels of ptp4a3a and ptp4a3b were remarkably low at the shield stage [6 h post-fertilization (hpf)], and at the tail-bud-stage (10 hpf) (Fig. 6A), we decided to suppress the expression of the remaining three isoforms, ptp4a1, ptp4a2a and ptp4a2b, with translation-blocking antisense morpholino oligonucleotides (MOs). As there are two splicing variants for ptp4a2b, variant 1 and variant 2, we designed MOs specifically targeting each of them. To verify the effectiveness of each MO, we synthesized four EGFP-fused mRNA constructs containing MO target sites for the above genes (Fig. S5A). Co-injection of each MO with the respective EGFP-fused mRNA construct resulted in almost complete suppression of GFP signal (Fig. S5B), confirming the suppression of mRNA translation. When each of the MOs was injected into one-cell-stage zebrafish embryos, those injected with ptp4a2b variant 1 MO showed mild shortening of the anterior-posterior axis at 24 hpf (Fig. S5C,D). When injected with the mixture of all four MOs (PRL MOs), most embryos showed a clear phenotype, with severe shortening of the anterior-posterior axis and a tortuous notochord (Fig. 6B). Short anterior-posterior axis is a typical phenotype caused by defects in convergent extension, a developmental process involving the reorganization of the embryonic epithelium (Wallingford et al., 2002). In this process, epithelial cells on the lateral sides of the embryo move towards the dorsal midline and intercalate with each other, and concomitantly the epithelial sheet extends along the anterior-posterior axis, dynamically changing the epithelial cell density at the dorsal side of the embryo (Tada and Heisenberg, 2012). In situ hybridization with ntla and dlx3b, markers for the notochord and neural plate border, respectively (Schulte-Merker et al., 1992; Akimenko et al., 1994), showed a short and wide notochord, with a broader neural plate border, indicating a defect in convergent extension (Fig. 6C). As progression of convergent extension requires appropriate cell density (Tada and Heisenberg, 2012), we speculated that suppression of PRL increases the cell density of the epithelium, which in turn affects convergent extension. To visualize cells in the embryonic epithelium, we used transgenic zebrafish that express EGFP in the cells of the outermost periderm layer, and compared the cell density at 10 hpf when convergent extension completes. Embryos reached the tail-bud stage at 10 hpf, irrespective of the expression status of PRL, implying no overt effect on normal embryonic development. Epithelial cell density on the dorsal side of the embryos injected with PRL MOs increased slightly but significantly (Fig. 6D,E). Immunostaining for anti-cleaved caspase-3 showed a significant reduction in the number of positive cells in these embryos (Fig. 6F; Fig. S5E). Altogether, these data implicate the role of PRL in regulating the epithelial cell density during convergent extension in zebrafish embryos, as in MDCK cell culture.

Moreover, to examine the effect of overexpression of PRL, we also performed an mRNA injection experiment. We chose ptp4a2b variant 1 for mRNA injection as the expression of ptp4a2b was reported to be highest at one-cell-stage embryo (White et al., 2017), and the sole injection of ptp4a2b variant 1 MO into the embryo showed mild shortening of the anterior-posterior axis (Fig. S5C). After the injection of the mRNA, we stained the embryo with anti-E-cadherin antibody to quantify E-cadherin levels. The result showed a significant increase in E-cadherin levels in ptp4a2b variant 1 mRNA-injected embryo compared to control emerald-luciferase mRNA-injected embryo (Fig. S6A). In addition, we also stained the embryo with anti-cleaved caspase-3 antibody and confirmed the increase in cell death in the periderm of ptp4a2b variant 1 mRNA-injected embryo (Fig. S6B). Collectively, PRL overexpression in zebrafish embryo increased both E-cadherin expression and apoptosis within the epithelial layer, as with the culture of MDCK cells.

In this study, we showed that PRL regulates epithelial cell density by inducing apoptosis at high cell density by upregulating the translation of E-cadherin, and subsequent activation of the TGF-β pathway. Our observations are consistent with previous studies that have reported E-cadherin-dependent activation of the TGF-β pathway (Andl et al., 2006; Akieda et al., 2019) and cell death downstream of E-cadherin and the TGF-β pathway (Lu et al., 2014; Akieda et al., 2019; Zhang et al., 2017). Although several studies have shown the extrusion of dying cells from the epithelial sheet (Rosenblatt et al., 2001; Gu et al., 2011), what drives cell death and how it is regulated in apoptotic cell extrusion remain largely unknown. Our study showed the importance of PRL in this process, and revealed the sequence of events involved in the induction of apoptosis downstream of PRL.

PRL is overexpressed in cancers and promotes cancer malignancy. Numerous studies have reported the increased expression of PRL, particularly at the advanced stages of cancer compared to early stages (Saha et al., 2001; Polato et al., 2005; Radke et al., 2006; Dai et al., 2009; Mayinuer et al., 2013). Our study identifies PRL as a regulator of epithelial cell density, and this may be a distinct function of PRL, particularly in non-transformed epithelia or at early stages of cancer. Such context-dependent multiple functions have also been reported for E-cadherin and the TGF-β pathway. Although E-cadherin is widely known for its role in normal epithelia by physically joining cells and facilitating other juxtacrine signaling events (Gottardi et al., 2001), some studies suggest that E-cadherin expression in the late stage of cancer promotes invasion and distant metastasis by enhancing cancer cell survival in circulation, as well as by promoting collective cell migration (Cheung et al., 2013; Padmanaban et al., 2019). Moreover, expression of E-cadherin during mesenchymal-to-epithelial transition (MET) at sites of metastases has been reported to promote the formation of metastatic tumors (Yao et al., 2011). In the case of the TGF-β pathway, E-cadherin has tumor suppressive effects in normal cells and in early stage cancers, through inhibition of cell proliferation, induction of apoptosis and inhibition of cell immortalization (Neel et al., 2012), but it promotes malignant progression in advanced stages of cancer, probably due to genetic and epigenetic changes in tumor cells, such as mutations in tumor suppressor p53 or loss of SMAD4 (Adorno et al., 2009; Zhang et al., 2010).

One of the important questions that remains unanswered is the mechanism of PRL-induced upregulation of E-cadherin translation. It was recently reported that overexpression of Human antigen R (HuR), an RNA-binding protein (RBP), and suppression of CUG-binding protein 1 (CUGBP1), another RBP, both upregulate E-cadherin translation in Caco-2 cells (Yu et al., 2016). Both HuR and CUGBP1 interact directly with the 3′-untranslated region (UTR) of E-cadherin mRNA; HuR prevents the translocation of mRNA to processing bodies, thereby increasing its accessibility to translation machinery, leading to upregulation of translation, whereas CUGBP1 promotes the translocation process and suppresses translation. Besides, accumulating evidence suggests that microRNAs play a general role in translation (Valinezhad Orang et al., 2014). Several studies report that microRNA-9 and microRNA-495 target E-cadherin mRNA, and ultimately affect the protein level of E-cadherin (Ma et al., 2010; Hwang-Verslues et al., 2011; Xu et al., 2017). However, whether they can regulate the expression of E-cadherin at the translation level remains poorly characterized. Overall, regulation of E-cadherin translation is still elusive. Generally, the predominant regulation of mRNA translation is exerted at the initiation step (Sonenberg and Hinnebusch, 2009), which is normally facilitated by several eukaryotic translation initiation factors (eIFs) that are partly regulated by mammalian target of rapamycin (mTOR) signaling (Ma and Blenis, 2009). Although mTOR is regarded as a critical regulator of global protein translation, several transcriptome-wide analyses of cells treated with mTOR inhibitors have revealed that the translation of fewer than 600 mRNAs can be considered as extremely sensitive to mTOR (Hsieh et al., 2012; Larsson et al., 2012; Thoreen et al., 2012). This suggests that the regulation of the translation by mTOR is confined to specific mRNA populations. In our previous study, we reported that overexpression of PRL stimulates mTOR signaling via inhibition of the Mg²⁺ transporter CNNM (Funato et al., 2014). Taken together, evidence suggests that PRL may promote E-cadherin translation by activating mTOR signaling. Several other mechanisms, including unwinding of secondary structures of the 5′-UTR by eIFs and binding of RBPs and/or poly(A)-binding proteins (PABPs) to the 3′-UTR, are also known to facilitate the translation of specific mRNA populations (Mignone et al., 2002; Sonenberg and Hinnebusch, 2009; Matoulkova et al., 2012; Burgess and Gray, 2010). Future studies, focusing on these factors responsible for the regulation of translation initiation, may help in understanding the mechanism of upregulation of E-cadherin translation by PRL.

PRL shows phosphatase activity towards certain substrates, such as ezrin, phosphatidylinositol (4,5) bisphosphate, integrin β1, fizzy and cell division cycle 20 related 1 (FZR1), and PTEN (Forte et al., 2008; McParland et al., 2011; Tian et al., 2012; Zhang et al., 2019; Li et al., 2020). It also binds to CNNM Mg²⁺ transporters to regulate the level of intracellular Mg²⁺ (Funato et al., 2014; Hardy et al., 2015; Gulerez et al., 2016), which is crucial for a variety of cellular signaling (Romani, 2011). In this study, dome formation was suppressed in the confluent cultures of cells expressing PRL, whereas cells that express the C104S mutant PRL, which has no phosphatase activity and no CNNM Mg²⁺ transporter-binding ability, were able to form domes, like control cells (Fig. 1A). These results suggest the importance of at least one, if not both, of these two functions of PRL in regulating epithelial cell density. Studies using a very recently identified PRL C104D mutant, which lacks phosphatase activity but has CNNM-binding ability (Kozlov et al., 2020), may identify the specific action of PRL that is essential to modulate the translation of E-cadherin. Future studies of PRL-mediated cell death at high cell density can be expected not only to reveal the details of the cell-death-mediated regulatory mechanism of epithelial cell density, but also broaden the understanding of the role of PRL in promoting cancer malignancy.

MDCK cells were provided by Dr Yasuyuki Fujita (Kyoto University, Japan) and Dr Mihoko Kajita (Ritsumeikan University, Japan). MDCK cells stably expressing GFP-tagged E-cadherin were supplied by Dr Yoshimi Takai and Dr Kiyohito Mizutani (Kobe University, Japan). MDCK-derived cell lines that express GFP-tagged PRL3 (PRL #1, PRL #2 and PRL C104S) in a Dox-inducible manner were established in our previous study (Kojima et al., 2019). The cells were routinely maintained in Dulbecco's modified Eagle's medium (Nissui, 05919) supplemented with 10% fetal bovine serum (FBS) and antibiotics. All cell lines were tested for mycoplasma infection by staining cells with DAPI (Roche, 10236276001), and were authenticated by observing the presence of typical morphological and proliferative characteristics as epithelial cells. For knockdown experiments, 3×10⁵ cells were seeded in 35-mm dishes and cultured for 12 h. Cells were then transfected with siRNAs twice with a 24 h interval using Lipofectamine RNAiMAX (Invitrogen, 13778150). After 12 h, cells were trypsinized, re-seeded (1×10⁶ cells) in other 35-mm dishes and cultured in the presence of Dox (2 µg/ml). The siRNA target sequences were as follows: PRL1 siRNA #1, 5′-GCAACUUCUGUAUUUGGAGAAGUAU-3′; PRL1 siRNA #2, 5′-CCAACCAAUGCGACCUUAAACAAAU-3′; PRL2 siRNA #1, 5′-GGUUCGAGUUUGUGAUGCUACAUAU-3′; PRL2 siRNA #2, 5′-GAGUGACAACUUUGGUUCGAGUUUG-3′; E-cadherin siRNA #1, 5′-GCAUGAUGUUCACUAUCAA-3′; E-cadherin siRNA #2, 5′-CUCUCAUGCCGUAUCUUCU-3′; SMAD4 siRNA #1, 5′-GACUUUGAGGGACAGCCAU-3′; SMAD4 siRNA #2, 5′-GUGUUUGUGCAGAGUUACU-3′; and negative control Lo GC siRNA (Invitrogen, 12935200).

A rabbit anti-PRL antibody was generated in our previous study (Funato et al., 2014). Mouse anti-ZO-1 antibody was a gift from Dr Masahiko Itoh and Dr Mikio Furuse (Furuse et al., 1994). Other commercially available primary antibodies used in this study were as follows: mouse monoclonal antibodies for β-actin (clone 2D4H5; Proteintech, 66009-1-Ig), β-catenin (clone 14; BD Bioscience, 610154), Na⁺/K⁺ ATPase (clone C464.6; Sigma-Aldrich, 05-369), 5-bromo-2′-deoxy-uridine (BrdU) (clone BMG6H8; Roche, 11299964001) and E-cadherin (clone 36; BD Bioscience, 610182) for immunoprecipitation analyses, immunoblotting analyses, staining MDCK cells and for staining zebrafish embryos, SMAD4 (clone B-8; Santa Cruz Biotechnology, sc-7966) and TGFBR2 (clone 2D5H7; Proteintech, 66636-1-Ig) for immunoblotting; rabbit monoclonal antibodies for SMAD2 (clone D43B4; Cell Signaling Technology, 5339), cleaved caspase-3 (clone 5A1E; Cell Signaling Technology, 9664) for staining MDCK cells and cleaved caspase-3 (clone C92-605; BD Bioscience, 559565) for staining zebrafish embryos; rat monoclonal antibodies for E-cadherin (clone ECCD2; Invitrogen, 13-1900) for immunoblotting analyses; and chicken polyclonal antibodies for GFP (Abcam, ab13970). The secondary antibodies used in this study were as follows: Alexa Fluor 568-conjugated anti-rabbit IgG (Invitrogen, A-11036) and Alexa Fluor 568-conjugated anti-mouse IgG (Invitrogen, A-11031) for immunofluorescence staining, and alkaline phosphatase (AP)-conjugated anti-rabbit IgG (Promega, S3731), AP-conjugated anti-rat IgG (Promega, S3831), AP-conjugated anti-mouse IgG (Promega, S3721), horseradish peroxidase (HRP)-conjugated anti-chicken IgY (Promega, G1351) and Mouse TrueBlot ULTRA: Anti-Mouse Ig HRP (Rockland, 18-8817-31) for immunoblotting analyses. Alexa Fluor 647-conjugated Phalloidin (Invitrogen, A-22287) was used for staining F-actin. Antibody dilutions are listed in Table S1. The inhibitors used in this study were as follows: cycloheximide (Sigma-Aldrich, 01810), ALK5 inhibitor II (Cayman Chemical, 14794), and SB525334 (Cayman Chemical, 16281). Recombinant Human TGF-β1 (R&D Systems, 240-B) was used for the stimulation of MDCK cells.

A total of 3.75×10⁴ cells (equivalent to 1×10⁶ cells in 35-mm dishes) were seeded on a membrane of Transwell inserts (Corning, 3413) and cultured for 12 h, treated with Dox (2 µg/ml) when the culture was about to attain confluence, and further cultured for 48 h. Electrical resistance, across the epithelial sheet formed on the membrane, was measured using a Millicell-ERS device (Millipore). Resistance of the blank membrane without cells was subtracted to obtain the actual resistance across the epithelial sheet. TEER was calculated by multiplying the resistance by the area of the membrane (0.33 cm²) (Ωcm²). The mean±s.d. was obtained from three independent experiments.

A total of 3.98×10⁵ cells (equivalent to 1×10⁶ cells in 35-mm dishes) were seeded on glass coverslips placed in a 12-well plate (Thermo Fisher Scientific, 150628) and cultured for 12 h, treated with Dox (2 µg/ml) when the culture was about to attain confluence, and further cultured for 24 or 48 h. Their proliferation status was determined using a BrdU Labeling and Detection Kit II (Roche, 11299964001), according to the manufacturer's instructions. Briefly, cells were labeled with 10 µM BrdU for 30 min at 37°C, fixed with 70% ethanol for 20 min at −30°C, washed three times with PBS and incubated with anti-BrdU antibody for 30 min at 37°C. After this, they were washed three times with PBS and then incubated with Alexa Fluor 568-conjugated anti-mouse IgG antibody and DAPI for 30 min at room temperature. Then, the cells were washed three times with PBS, and the glass coverslips were mounted on glass slides and observed under a confocal laser scanning microscope (FLUOVIEW FV1000, Olympus).

Cells (1×10⁶ or 2.5×10⁵) were seeded in 35-mm dishes containing glass coverslips and cultured for 12 h. They were treated with Dox (2 μg/ml) when the culture was about to attain confluence, and cultured for 24 or 48 h. These cultured cells were then stained with anti-ZO-1 antibody, anti-β-catenin antibody, anti-Na⁺/K⁺ ATPase antibody or anti-Smad2 antibody, as described previously (Itoh et al., 1997; Hirata et al., 2014; Nallet-Staub et al., 2015), with a slight modification for anti-Smad2 staining. For this, cells were permeabilized with 0.1% Triton X-100 for 10 min at room temperature. For staining with anti-E-cadherin antibody, cells were fixed with 3.7% formaldehyde for 10 min on ice, washed three times with PBS, permeabilized and blocked with PBS containing 0.5% Triton X-100 and 5% normal goat serum for 1 h at room temperature. For staining with anti-cleaved caspase-3 antibody (Cell Signaling Technology), cells were fixed with 3.7% formaldehyde for 15 min at room temperature, washed three times with PBS, permeabilized and blocked with PBS containing 0.3% Triton X-100 and 5% bovine serum albumin for 1 h at room temperature. Thereafter, the cells were incubated overnight with respective primary antibodies at 4°C, followed by incubation with fluorophore-conjugated secondary antibodies, rhodamine-conjugated phalloidin (Wako, 165-21641) and DAPI for 30 min (for cleaved caspase-3 staining) or 1 h at room temperature. After washing the cells to remove excess antibody or stain, the glass coverslips were mounted on glass slides and observed under a confocal laser scanning microscope (FLUOVIEW FV1000).

Cells (1×10⁶) were seeded in 35-mm dishes and cultured for 12 h, treated with Dox (2 µg/ml) when the culture was about to attain confluence, and further cultured for 48 h. Cytotoxicity was determined by measuring the activity of LDH released from the cytosol of damaged cells using a cytotoxicity LDH assay kit (Dojindo, CK12-05) according to the manufacturer's instructions. Briefly, culture medium was collected from dishes and centrifuged at 15,000 rpm (17,360 g) for 10 min to remove cell debris. The resulting supernatant was used to measure the activity of LDH released from the cells. To calculate the percentage of LDH release, the total LDH activity of the culture cells was also measured. For this, cells were lysed with 0.1% NP-40 in PBS, and the lysate was centrifuged at 15,000 rpm (17,360 g) for 10 min. The resulting supernatant was used to measure total LDH activity. Finally, the percentage of LDH release was calculated by taking the ratio of the medium LDH activity to the total LDH activity.

Polydimethylsiloxane (PDMS) membranes at the bottom of the wells of the silicon chamber (Nepa Gene, NST-CH-4W) were coated with 0.01% poly-L-lysine. In each well, 2.9×10⁵ cells (equivalent to 1×10⁶ cells in 35-mm dishes) were seeded, cultured for 12 h and treated with Dox (2 µg/ml) when the culture was about to attain confluence. The silicon chamber was then stretched by 20% and the culture continued for 24 h. The cells were fixed with 3.7% formaldehyde for 15 min at room temperature, and then the PDMS membranes were excised and used for immunofluorescence staining with anti-cleaved caspase-3 antibody.

A total of 3.64×10⁴ cells (equivalent to 1×10⁶ cells in 35-mm dishes) were seeded in a white flat-bottomed 96-well plate (Thermo Fisher Scientific, 136101) and cultured for 12 h. The cell culture was then treated with Dox (2 μg/ml), just before the cell culture attained confluence, and cultured for 24 h. The activity of caspase-3/7 was determined using the Caspase-Glo3/7 Assay System (Promega, G8090) according to the manufacturer's instructions. Briefly, cells were treated with Caspase-Glo3/7 Reagent and incubated in the dark for 1 h. The luminescent signal generated upon the cleavage of the substrate by caspase-3/7 released from dying cells was recorded using an LD 400 microplate reader (Beckman Coulter).

From the total RNA, extracted from cultured cells using RNAiso Plus (Takara, 9109), cDNA was synthesized and analyzed by real-time PCR using a Luna Universal One-Step RT-qPCR Kit (New England Biolabs, E3005). A MiniOpticon instrument (Bio-Rad) was used for monitoring fluorescence emitted by amplified DNA. Specific amplification during PCR was verified by agarose gel electrophoresis of the final PCR product from each experiment. The primer sets used were as follows: dog E-cadherin, 5′-AAGCGGCCTCTACAACTTCA-3′ and 5′-AACTGGGAAATGTGAGCACC-3′; and dog GAPDH, 5′-AACATCATCCCTGCTTCCAC-3′ and 5′-GACCACCTGGTCCTCAGTGT-3′.

For radiolabeling, 6.4×10⁶ cells (equivalent to 1×10⁶ cells in 35-mm dishes) were seeded in 10-cm dishes. After 12 h, when the culture was about to attain confluence, cells were treated with Dox (2 μg/ml) and cultured for a further 48 h. The cells were preincubated in methionine- and cysteine-free medium (Gibco, 21013024), supplemented with glutamine and PBS-dialyzed 10% FBS for 30 min at 37°C, before labeling with 200 μCi/ml of EasyTag EXPRESS³⁵S Protein Labeling Mix (Perkin Elmer, NEG772, 11 mCi/ml) for 30 min. The cells were then lysed in RIPA buffer [150 mM NaCl, 50 mM Tris-HCl (pH 7.5), 1% deoxycholate, 0.1% SDS and 1% Triton X-100] containing protease inhibitor cocktail (Roche, 04693159001). For immunoprecipitation, the total protein concentration of each lysate was estimated by Coomassie Brilliant Blue (CBB) staining. Each lysate sample was incubated with mouse normal IgG (3.8 μg/ml)-bound agarose beads for 1 h at 4°C to remove non-specifically precipitated proteins (preclearing). An equal amount of total protein from precleared lysates was subjected to immunoprecipitation with 10 μg of anti-E-cadherin antibody, and the precipitated proteins were separated by SDS-PAGE, immunoblotted with the same anti-E-cadherin antibody and visualized by autoradiography.

For the preparation of cell lysates for TGFBR2-immunoblotting experiments, 16.5×10⁶ cells (equivalent to 1×10⁶ cells in 35-mm dishes) were seeded in 15-cm dishes. After 12 h, when the culture was about to attain confluence, cells were treated with Dox (2 μg/ml) and cultured for a further 48 h. The cells were then lysed in lysis buffer [100 mM NaCl, 25 mM HEPES (pH 7.5), 10% glycerol, 5 mM EDTA and 1% Triton X-100] containing protease inhibitor phenylmethylsulfonyl fluoride (1 mM). For immunoprecipitation of E-cadherin, the total protein concentration of each cell lysate was estimated by CBB staining. Each lysate sample was incubated with mouse normal IgG (3.8 μg/ml)-bound agarose beads for 1 h at 4°C for preclearing. An equal amount of total protein from precleared lysate samples was subjected to immunoprecipitation with 10 μg of anti-E-cadherin antibody, and the precipitated proteins were separated by SDS-PAGE and immunoblotted with the same anti-E-cadherin antibody and anti-TGFBR2 antibody.

Zebrafish were raised and maintained according to internationally recognized guidelines (Westerfield, 2000). For this study, wild-type strain AB was used, along with a previously established transgenic line that expresses EGFP in the periderm, Tg [krt4p:gal4; UAS:eGFP] [a gift from Dr Kawakami and Dr Wada (Wada et al., 2013)]. All experiments and animal care were carried out according to the institutional and national guidelines and regulations (Osaka University, Osaka, Permit 20005). Embryos and larvae were staged as described previously (Kimmel et al., 1995).

Zebrafish embryos were collected at the shield stage and tail-bud stage, which correspond to 6 hpf and 10 hpf, respectively. Total RNA was extracted using a miRNeasy Mini Kit (Qiagen, 217004) according to the manufacturer's instructions. Libraries for sequencing were prepared using a TruSeq stranded mRNA library prep kit (Illumina) according to the manufacturer's instructions. Paired-end sequencing on NovaSeq 6000 (Illumina) yielded 101-bp paired-end reads. Illumina Casava1.8.2 software was used for base calling. Sequenced reads were mapped to the zebrafish reference genome assembly (GRCz11) using TopHat v2.0.13 in combination with Bowtie2 ver. 2.2.3 and SAMtools ver. 0.1.19. The fragments per kilobase of exon per million mapped reads (FPKMs) were calculated using Cufflinks version 2.2.1.

For creating various knockdown embryos of zebrafish, the following antisense MOs (Gene Tools) were used: ptp4a1 MO, 5′-GGGCAGGTCTATTCATACGAGCCAT-3′; ptp4a2a MO, 5′-GGGCGATTCATTTTGACACTCAAAT-3′; ptp4a2b variant 1 MO, 5′-CGTCCCATTCAGGAGAAGAGTTTCT-3′; ptp4a2b variant 2 MO, 5′-TCAACAGCGGCAGGGCGGTTCAT-3′; and control MO, 5′-CCTCTTACCTCAGTTACAATTTAT-3′. For all injections, 1.7 ng of each MO was injected into one-cell-stage embryos.

For mRNA microinjection, template plasmids were constructed. Briefly, total mRNA was extracted from one-cell-staged zebrafish embryos using TRIzol Reagent (Invitrogen, 15596018), and cDNA was synthesized using ReverTraAce (Toyobo, TRT-101). Approximately 100-bp-long 5′-UTR and 5′-ORF regions spanning the translation initiation site of the genes, namely, ptp4a1, ptp4a2a, ptp4a2b variant 1 and ptp4a2b variant 2 (GenBank accession numbers NM_001007775, NM_001005583, NM_001270540 and NM_001024098, respectively), which also contain the target sequence for each of the above four MOs, were amplified by PCR using zebrafish cDNA as template (schematically illustrated in Fig. S5A). The primer sets used were as follows: zebrafish ptp4a1, 5′-GGGCTCGAGAGATTTTCAAGCTCTATTGCAAAG-3′ and 5′-GGGCTGCAGATTGTGGGTAATGAGGAATCTCA-3′; zebrafish ptp4a2a, 5′-GGGCTCGAGCCTGTAAAACAATCGGAAGTTGTTG-3′ and 5′-GGGCTGCAGCCTCATGCATTCATAAGTGATCTCA-3′; and zebrafish ptp4a2b variant 1 and 2, 5′-GGGCTCGAGCTTCCTACAGATTTTTCTCCCTGTA-3′ for variant 1 and 5′-GGGCTCGAGTCTGAGCCAAAGATTTTTGAAAAGA-3′ for variant 2, and 5′-GGGCTGCAGATAGGAAATCTCAACAGCGGC-3′ for both variant 1 and 2. EGFP tag was transferred from the pEGFP-N1 vector into the pCS2 vector to construct pCS2-EGFP, and then each amplified PCR product was inserted in-frame into the C terminus of EGFP. The plasmid constructs were then linearized and transcribed in vitro using an mMESSAGE mMACHINE SP6 Transcription Kit (Thermo Fisher Scientific, AM1340) according to the manufacturer's instructions. Proper synthesis of each mRNA was verified by agarose gel electrophoresis. For all co-injection experiments with MOs, 100 pg of synthesized mRNA was used.

For PRL-overexpressing experiments, full-length coding region of the gene ptp4a2b variant 1 (GenBank accession number NM_001270540) was amplified by PCR using zebrafish cDNA as template. The primer sets used were as follows: zebrafish ptp4a2b variant 1, 5′-GGATCCATGGGACGTCTTGCCAACAT-3′ and 5′-GAATTCCTACTGGATACAGCAGTTGTGAC-3′. The amplified PCR product was inserted into the BamHI and EcoRI sites of the pCS2 vector. For control experiments, Emerald-Luciferase (TOYOBO) was PCR amplified using Luciferase-based reporter plasmid OTM:ELuc-CP (Akieda et al., 2019) as template. The primer sets used were as follows: 5′-TCTTTTTGCAGGATCCGCCACCATGGAGAGAGAG-3′ and 5′-GTTCTAGAGGCTCGAGTTACAGCTTAGAAGCCTTCTCCA-3′. The amplified PCR product was directly inserted into the BamHI and XhoI sites of pCS2p+ vector using an In-Fusion HD Cloning Kit (Takara Bio, 639644). The plasmid constructs were then linearized and transcribed in vitro using an mMESSAGE mMACHINE SP6 Transcription Kit (Thermo Fisher Scientific, AM1340) according to the manufacturer's instructions. Proper synthesis of the mRNA was verified by agarose gel electrophoresis. For PRL-overexpressing experiments, 800 pg of synthesized mRNA was used for injection.

Embryos resulting from crossing the wild-type strain with the strain showing periderm-specific GFP expression (Tg [krt4p:gal4; UAS:eGFP]) were injected with control MO or a combination of four MOs (PRL MOs: ptp4a1 MO, ptp4a2a MO, ptp4a2b variant 1 MO and ptp4a2b variant 2 MO). For PRL-overexpressing experiments, the embryos were injected with emerald-luciferase mRNA or ptp4a2b variant 1 mRNA. GFP⁺ embryos were then selected at the tail-bud stage and fixed with 4% paraformaldehyde (PFA) overnight at 4°C. The embryos were manually dechorionated with forceps, washed four times with 0.5% Triton X-100 in PBS, stained with DAPI and anti-cleaved caspase-3 antibody for quantification of cell death, washed three times with 0.5% Triton X-100 in PBS and observed under a confocal laser scanning microscope (FLUOVIEW FV3000, Olympus). Images of horizontal sections taken at successive focal planes were stacked to generate three-dimensional images. For the determination of cell density, the peridermal cells visible in the dorsal part of the embryos were counted and nuclei per unit area were then calculated. For the quantification of cell death, the number of cells positive for both cleaved caspase-3 and GFP were counted.

Embryos were fixed at the tail-bud stage with 4% PFA overnight at 4°C. The dechorionated embryos were then washed four times with 0.5% Triton X-100 in PBS and blocked with 0.1% Triton X-100 in PBS containing 10% FBS and 1% DMSO for 1 h. Embryos were incubated overnight with anti-cleaved caspse-3 antibody or anti-E-cadherin antibody diluted in 0.1% Triton X-100 in PBS containing 1% FBS and 1% DMSO at 4°C, and washed three times with 0.5% Triton X-100 in PBS. They were incubated overnight with Alexa Fluor 568-conjugated secondary antibody and DAPI diluted in 0.1% Triton X-100 in PBS containing 1% DMSO at 4°C. Images of stained embryos were then acquired using a fluorescence stereo microscope (M205 FA, Leica).

Whole-mount in situ hybridization was performed as described previously (Shimizu et al., 2014). The images were acquired using a stereo microscope (M205A, Leica).

Sample size (n) for each experiment was determined based on past studies. All statistical analyses were performed using GraphPad Prism 6 software (GraphPad Software) and the data are presented as mean±s.e.m. P-values were calculated by either Student's two-tailed t-test or one- or two-way ANOVA with post-hoc test, as described in the figure legends.

We thank Drs Yasuyuki Fujita (Kyoto University) and Mihoko Kajita (Ritsumeikan University) for sharing MDCK cells; Drs Masahiko Ito (Dokkyo Medical University) and Mikio Furuse (National Institute for Physiological Sciences) for the anti-ZO-1 antibody; Drs Hironori Wada (Kitasato University) and Koichi Kawakami (National Institute of Genetics), and the National BioResource Project, for providing transgenic zebrafish; Dr Kazuto Nunomura (Osaka University) for technical support with TEER measurements; Mr Juqi Zou (Osaka University) for technical support with in situ hybridization experiments; and the Genome Information Research Center at the Research Institute for Microbial Diseases (Osaka University) for support with RNA-seq and data analysis.

The peer review history is available online at https://journals.biologists.com/jcs/article-lookup/doi/10.1242/jcs.258550.