The P2Y₂ receptor promotes Wnt3a- and EGF-induced epithelial tubular formation by IEC6 cells by binding to integrins

Multicellular tubular structures form the common architecture of vertebrate epithelial tubular organs, including salivary glands, lungs, mammary glands, intestines, kidneys and the liver (Gumbiner, 1992). Within the tubular formations, several cells make cylindrical structures in which polarized cells in a monolayer are tightly connected by intercellular junctions, thereby forming internal hollow spaces (Datta et al., 2011; O'Brien et al., 2002). The tubular structures vary in length, diameter and shape, but different types of epithelial cells might use conserved machineries to regulate cellular events such as proliferation, migration and polarization (Lu and Werb, 2008). In order to find common mechanisms involved in tubular formation, an in vitro assay system is necessary. Madin Darby canine kidney (MDCK) cysts are a well-established model for tubulogenesis whereby hepatocyte growth factor (HGF) induces the formation of tubular structures from the cysts in three-dimensional (3D) culture with type I collagen (Debnath and Brugge, 2005; O'Brien et al., 2002; Zegers et al., 2003). However, 3D basement membrane matrices (BMMs) such as Matrigel have been shown to suppress HGF-induced tubular formation of MDCK cells (Santos and Nigam, 1993). By contrast, embryonic epithelial rudiments isolated from lungs, kidneys and salivary glands were able to change morphology and formed tubular structures in 3D Matrigel (Ohtsuka et al., 2001; Steinberg et al., 2005; Zhang et al., 2012). Therefore, a new in vitro approach in which epithelial cells can form tubes in 3D BMMs is necessary for fully understanding tubulogenesis in vivo.

We developed a new in vitro tubular formation assay using rat intestinal epithelial cells (IEC6 cells) in 3D Matrigel (Matsumoto et al., 2014). IEC6 cells formed cysts in 3D Matrigel, and simultaneous stimulation with Wnt3a and EGF induced tubular formation. Using DNA microarray analyses, we found that Wnt3a and EGF induce expression of ADP-ribosylation factor (Arf)-like 4c (Arl4c), which is a member of the small G protein superfamily (Burd et al., 2004). Arl4c expression in IEC6 cells activated Rac1 through Arf6 and inhibited RhoA, which together comprise the small G protein signal cascade, leading to changes in cell morphology (cell extension). These morphological changes in the leading cells of tubular structures induced the nuclear localization of Yes-associated protein (YAP) and of transcriptional co-activator with PDZ-binding motif (TAZ), which are Hippo pathway transcriptional activators and play crucial roles in cell proliferation (Saucedo and Edgar, 2007), resulting in tubular formation. Owing to expression of Arl4c, MDCK cells formed tubular structures in response to HGF in 3D Matrigel culture, and Arl4c expression was also involved in elongation and budding of kidney ureteric buds of mouse embryos, suggesting the existence of a common mechanism regulating tube formation (Matsumoto et al., 2014).

DNA microarray analyses also revealed that P2ry2 [encoding the P2Y₂ receptor (P2Y₂R)] mRNA is expressed by Wnt3a and EGF during tube formation (Matsumoto et al., 2014). P2Y₂R is a nucleotide-activated G-protein-coupled receptor (GPCR) and binds to ATP (purine) and UTP (pyrimidine) (Burnstock, 2006). P2Y₂R is expressed in various tissues, including the heart, lungs, skeletal muscle, intestine, spleen and bone marrow (Moore et al., 2001) and couples phospholipase C (PLC) β1 through Gq/11, resulting in the generation of diacylglycerol and inositol trisphosphate, leading to the activation of protein kinase C (PKC) and the induction of intracellular Ca²⁺, respectively (Burnstock, 2006). Therefore, P2Y₂R regulates various cellular functions, including co-transmission and neuromodulation in nerve cells, exocrine secretion in parotid and lachrymal acinar cells, endocrine secretion in the adrenal cortex and pituitary gland, as well as platelet aggregation, vasodilation and cell proliferation (Burnstock, 2006). However, the involvement of P2Y₂R expression in epithelial tubulogenesis has not been previously studied.

In addition to functioning as an ATP and UTP receptor, P2Y₂R has been reported to bind to integrins α_vβ₃ and α_vβ₅ through an arginine-glycine-aspartic-acid (RGD) domain in the first extracellular loop of P2Y₂R, and to induce cell migration by regulating Rac and Rho activities in 1321N1 astrocytoma cells (Bagchi et al., 2005; Erb et al., 2001; Liao et al., 2007). Activation of P2Y₂R mediated by UTP induces the clustering of integrin α_v, which is associated with a flattened phenotype in U937 promonocytic cells (Chorna et al., 2007). Thus, P2Y₂R is suggested to regulate cellular activities by coordinating integrin functions. Here, we show that Wnt3a and EGF-induced P2Y₂R expression is involved in tubular formation of IEC6 cells by means of P2Y₂R–integrin binding, rather than nucleotide binding.

P2ry2 mRNA and protein (P2Y₂R) levels in IEC6 cells were slightly upregulated by EGF alone but not by Wnt3a alone. The simultaneous Wnt3a and EGF stimulation induced the expression of P2ry2 mRNA and protein synergistically (Fig. 1A). In rats, there are eight P2ry genes, P2ry1, P2ry2, P2ry4, P2ry6, P2ry10, P2ry12, P2ry13 and P2ry14. Among them, we found that P2ry1 and P2ry2 were expressed in IEC6 cells, but other P2ry mRNAs were barely detected (supplementary material Fig. S1A). Wnt3a and EGF dramatically increased P2ry2 but not P2ry1 mRNA expression (supplementary material Fig. S1A). CHIR99021, an inhibitor of glycogen synthase kinase-3 (GSK-3), stabilizes β-catenin, leading to activation of the Wnt/β-catenin-dependent pathway through T-cell factor 4 (Tcf4) (Bennett et al., 2002). Similar to Wnt3a, CHIR99021 also increased P2ry2 mRNA and protein levels in cooperation with EGF (Fig. 1B). Wnt3a- and EGF-induced P2ry2 mRNA expression was reduced by knockdown of β-catenin (supplementary material Fig. S1B) and overexpression of a dominant negative form of Tcf4 (DN-Tcf4) (supplementary material Fig. S1C), indicating that β-catenin and Tcf4 are involved in the induction of P2ry2 expression. To confirm that P2Y₂R induced by Wnt3a and EGF is expressed at the cell surface, IEC6 cells were subjected to a surface biotinylation assay. When cell lysates were precipitated with neutravidin–agarose, P2Y₂R was detected in a Wnt3a- and EGF-dependent manner (Fig. 1C). Thus, Wnt3a and EGF indeed induce P2Y₂R expression at the cell surface.

EGF activates various signaling cascades, including the Ras–Raf–extracellular-signal-regulated kinase 1/2 (ERK1/2, also known as MAPK3 and MAPK1, respectively), phosphoinositide 3-kinase (PI3K)–AKT, and PKC–c-Jun N-terminal kinase (JNK) pathways (Avraham and Yarden, 2011). We found that Wnt3a- and EGF-induced P2ry2 mRNA expression in IEC6 cells was inhibited by PD168393 (an EGFR inhibitor), but not by U0126 (a MEK inhibitor), SP600125 (a JNK inhibitor) or LY294002 (a PI3K inhibitor) (supplementary material Fig. S1D). Transforming growth factor-β (TGF-β) signaling, which is mediated by the TGF-β receptor type I (TGFBR1, also known as Alk5), and the Smad2 and Smad3 transcriptional activators (Miyazono et al., 2004) have also been shown to mediate EGF signaling (Anzano et al., 1982; Uttamsingh et al., 2008). We found that A83-01 (a TGFBR1 inhibitor) and small interfering RNAs (siRNAs) against Smad2 and Smad3 inhibited Wnt3a- and EGF-induced P2ry2 mRNA expression, but not Wnt3a- and EGF-induced Arl4c mRNA expression (supplementary material Fig. S1E,F). As Arl4c expression has been shown to be dependent on MEK and ETS1/2 signaling (Matsumoto et al., 2014), it follows that the mechanisms underlying P2Y₂R expression would be different from Arl4c expression. In addition, Wnt3a and EGF still induced P2ry2 mRNA expression in Arl4c-depleted cells and vice versa, suggesting that Arl4c expression and P2Y₂R expression are independently regulated (supplementary material Fig. S1G).

Knockdown of P2Y₂R did not affect the morphology of IEC6 cysts, but inhibited the elongation and branching of tubes induced by Wnt3a and EGF (Fig. 1D; supplementary material Fig. S2A). Wild-type (WT) P2Y₂R–HA expression rescued the phenotypes induced by siRNA against P2Y₂R, excluding siRNA off-target effects (Fig. 1D; supplementary material Fig. S2A). Exogenously expressed P2Y₂R levels in P2Y₂R–HA-expressing IEC6 cells were higher than endogenous P2Y₂R levels induced by Wnt3a and EGF, and P2Y₂R–HA localized to apical and basolateral membranes of P2Y₂R–HA-expressing IEC6 cells (supplementary material Fig. S2B,C). Cyst formation was not affected by P2Y₂R expression. Notably, EGF alone was able to induce tubular formation of P2Y₂R–HA-expressing IEC6 cells (Fig. 1E), whereas Wnt3a alone did not have the same effect (supplementary material Fig. S2D). Wnt3a and EGF induced the formation of continuous lumens in IEC6 cells, but discontinuous lumens were observed when P2Y₂R–HA-expressing IEC6 cells were treated with Wnt3a and EGF, or EGF alone (Fig. 1D,E), suggesting that spatial and temporal expression of P2Y₂R is necessary for the development of apical and basolateral polarization. IEC6 cells are known to form tubular structures from single cysts when treated with Wnt3a and EGF (Matsumoto et al., 2014). Microscopic examination of fixed points revealed that P2Y₂R–HA-expressing IEC6 cells also form tubular structures from single cysts in response to Wnt3a and EGF, but no tubular structures are formed by the assembly of multiple cysts (supplementary material Fig. S2E). In addition, A83-01, which inhibits P2ry2 mRNA expression, suppressed Wnt3a- and EGF-induced tube elongation and branching (supplementary material Fig. S2F). Therefore, Wnt3a- and EGF-induced P2Y₂R expression is involved in tubule formation.

P2Y₂R functions as a receptor for the nucleotides ATP and UTP, and mediates nucleotide signaling (Burnstock, 2006). Apyrase causes the hydrolysis of ATP and ADP, resulting in the generation of AMP, which does not activate P2 receptors (Komoszyński and Wojtczak, 1996). In addition, suramin is a competitive antagonist for P2Y receptors, including P2Y₂R (Burnstock, 2007). Treatment of IEC6 cells with these reagents led to the suppression of Wnt3a- and EGF-induced tubular formation (Fig. 2A). However, addition of ATPγS, a non-hydrolysable ATP, did not affect the morphology of Wnt3a- or EGF-treated cysts or Wnt3a- and EGF-induced tubular formation (Fig. 2B). Thus, extracellular endogenous nucleotides could be required for tubular formation in our assay conditions, but exogenously added nucleotides did not affect this process. It seems that extracellular endogenous nucleotides are sufficient for contributing to Wnt3a- and EGF-dependent tubular formation.

In IEC6 cells, we found that ATPγS activated ERK1/2 in a dose-dependent manner, and knockdown of P2Y₂R was able to partially suppress ATPγS-dependent ERK1/2 activation (Fig. 2C). Furthermore, WT P2Y₂R–HA expression enhanced ATPγS-dependent ERK1/2 activation (Fig. 2D). A human P2Y₂R mutant (P2Y₂R^R265L) caused a significant decrease in ATP- and UTP-induced P2Y₂R-mediated Ca²⁺ release in 1321N1 cells (Burnstock, 2006; Erb et al., 1995; Hillmann et al., 2009). In IEC6 cells, the expression of rat P2Y₂R^R264L–HA, which corresponds to the human P2Y₂R^R265L, did not enhance basal or ATPγS-dependent ERK1/2 activation, confirming that rat P2Y₂R^R264L is not involved in mediating nucleotide signaling (Fig. 2D; supplementary material Fig. S3A). These results were also confirmed using HEK293T cells (supplementary material Fig. S3B).

P2Y₂R knockdown suppressed Wnt3a- and EGF-dependent tubular formation, and expression of P2Y₂R^R264L in P2Y₂R-depleted IEC6 cells rescued the phenotypes to a level comparable to WT P2Y₂R expression (Fig. 2E). In addition, both P2Y₂R–HA cells and P2Y₂R^R264L–HA-expressing IEC6 cells induced tubular formation by stimulation with EGF alone (Fig. 2F). Taken together, these results suggest that the nucleotide signaling function of P2Y₂R is not required for Wnt3a- and EGF-induced tubular formation, and P2Y₂R must therefore mediate other signaling pathways during tubular formation.

The RGD domain of the first extracellular loop of human P2Y₂R has been suggested to form a complex with integrins (α_vβ₃ and α_vβ₅) in 1321N1 cells (Erb et al., 2001). In this study, we used rat P2Y₂R to analyze its functions in rat IEC6 cells. Rat P2Y₂R contains a Gln-Gly-Asp sequence (QGD, with amino acid ‘D’ at position number 97) instead of the RGD sequence in human P2Y₂R, but the alteration of the arginine residue to the glutamine residue is considered to be a conservative substitution that maintains integrin binding (Erb et al., 2001). Indeed, a proximity ligation assay (PLA) confirmed that rat P2Y₂R–HA interacts with endogenous integrin α_v as well as human P2Y₂R–HA in HeLaS3 cells (supplementary material Fig. S3C).

Human P2Y₂R^D97E does not bind to integrin α_v in 1321N1 cells (Erb et al., 2001). Indeed, PLA revealed that rat P2Y₂R^D97E–HA reduced the ability to form a complex with endogenous integrin α_v in HeLaS3 cells compared with P2Y₂R–HA and P2Y₂R^R264L–HA (Fig. 3A and supplementary material Fig. S3D). P2Y₂R–HA or P2Y₂R^D97E–HA was expressed with integrins α_v and β₃ in HEK293T cells, and cell lysates were immunoprecipitated with anti-HA antibody. The biochemical assay also showed that P2Y₂R–HA forms a complex with integrins and P2Y₂R^D97E–HA decreases integrin-binding activity (Fig. 3B). In addition, P2Y₂R^D97E–HA expression did not enhance ATPγS (in the 2.5–10 μM range)-dependent ERK activation in IEC6 and 293T cells (Fig. 3C and supplementary material Fig. S3E,F), consistent with the finding that P2Y₂R^D97E required a 10⁴-fold higher concentration of UTP in 1321N1 cells (Erb et al., 2001; Qi et al., 2005). P2Y₂R^D97E–HA expression did not rescue the phenotypes induced by P2Y₂R knockdown during Wnt3a- and EGF-induced tubular formation under conditions in which P2Y₂R–HA or P2Y₂R^R264L–HA expression was able to rescue them (Fig. 2E, Fig. 3D). Furthermore, the expression of P2Y₂R–HA or P2Y₂R^R264L–HA but not P2Y₂R^D97E–HA induced tubular formation in the presence of EGF alone (Fig. 2F, Fig. 3E). These results suggest that the integrin binding of P2Y₂R plays an important role in Wnt3a- and EGF-induced tubular formation rather than nucleotide signaling through P2Y₂R.

The RGD domain of fibronectin was originally identified as an integrin-binding domain (Ruoslahti and Pierschbacher, 1987). IEC6 cells were shown to produce high levels of fibronectin compared with other RGD-domain-containing proteins, including vitronectin, TSP-1/2 (also known as THBS1 and THBS2, respectively), tenascin-C, entactin-1 and nephronectin (supplementary material Fig. S4A) (Xu and Mosher, 2011). To examine the roles of the interaction between integrins and fibronectin during tubulogenesis, the cyclic peptide antagonist RGDfV was added to the culture medium (Pfaff et al., 1994). Together with EGF, RGDfV cooperatively induced tubular formation at a concentration of 1–5 μg/ml with the highest activity at 5 μg/ml; however, 10 or 25 μg/ml RGDfV was not able to induce tubulation (Fig. 4A). Under these conditions, RGDfV did not affect basal or EGF-induced P2ry2 mRNA expression (supplementary material Fig. S4B). Therefore, we reasoned that inhibition of integrin and fibronectin binding by appropriate expression of P2Y₂R might be important in the tubular formation of IEC6 cells. Fibronectin was observed in the basal membranes of IEC6 cells in the tubular structures and not detected in the surrounding Matrigel (Fig. 4B). When cells were cultured in two dimensions on Matrigel-coated dishes, fibrous fibronectin structures were also detected at the basal side of IEC6 cells and these connected to paxillin-positive focal adhesions in the central region but not the peripheral region (supplementary material Fig. S4C). This suggests that fibronectin deposits are involved in cell-to-substrate adhesion in the central region and fibronectin-free focal adhesions are motile at the cell edge.

As Matrigel does not contain fibronectin as a component, any effects of fibronectin on tubulogenesis must be based on the production of fibronectin by IEC6 cells. A mild reduction in fibronectin levels mediated by 0.25 nM siRNA induced tubular formation of IEC6 cells in the presence of EGF alone, but a severe reduction using 1 nM siRNA resulted in a failure of formation of the tubular structures (Fig. 4C; supplementary material Fig. S4D). A different siRNA against fibronectin also gave similar results (supplementary material Fig. S4D,E). Furthermore, Wnt3a- and EGF-induced tubular formation was suppressed by knockdown of fibronectin with 1 nM siRNA (Fig. 4D). A fibronectin coat on solidified Matrigel suppressed Wnt3a- and EGF-induced tubular formation, although it did not affect cyst morphology in the presence of either Wnt3a or EGF alone (Fig. 4E; supplementary material Fig. S4F), suggesting that exogenously added fibronectin overcomes the inhibitory effect of P2Y₂R on the binding of integrins and fibronectin. However, laminin-111, which is mostly recognized by integrins through the C-terminal laminin globular (LG) domains of the α chain (Xu and Mosher, 2011), was not able to inhibit Wnt3a- and EGF-induced tubular formation (supplementary material Fig. S4G). Taken together, these results suggest that the interaction of integrins with fibronectin is necessary for tubulogenesis and that the appropriate inhibition of their interaction by P2Y₂R expression induced by Wnt3a and EGF promotes tubular formation.

Binding of P2Y₂R–HA or P2Y₂R^R264L–HA to endogenous integrin α_v in HeLaS3 cells was suppressed by exogenously adding fibronectin (i.e. a fibronectin coat) in a PLA (Fig. 5A). Although the binding activity of P2Y₂R^D97E–HA to integrin α_v was lower, there was no further inhibition upon addition of fibronectin (Fig. 5A), suggesting that P2Y₂R interacts with integrins in both an RGD-domain-dependent and -independent manner. In addition, endogenous integrin β₅ bound to endogenous fibronectin in IEC6 cells (Fig. 5B). PLA revealed that the expression of P2Y₂R–HA or P2Y₂R^R264L–HA, but not P2Y₂R^D97E–HA, suppresses the formation of a complex between integrin β₅ and fibronectin (Fig. 5B). These results imply that the RGD-domain-independent interaction of P2Y₂R with integrins does not affect tubulogenesis. An immunoprecipitation assay confirmed that integrin α_v interacts with fibronectin in HEK293T cells and that ectopically expressed P2Y₂R–HA inhibits their interaction (Fig. 5C). Therefore, the R(Q)GD domain of P2Y₂R, upon stimulation with Wnt3a and EGF, might compete with fibronectin for the binding to integrins.

Our previous results demonstrated that Wnt3a- and EGF-dependent Arl4c expression induces cell morphological changes (cell extension) through the regulation of the small G proteins Rac and Rho, and that it promotes cell proliferation through modulating the nuclear localization of YAP and TAZ, leading to tubular formation (Matsumoto et al., 2014). Therefore, we speculated that P2Y₂R expression could also affect cell behaviors. When IEC6 cells were seeded on Matrigel-coated dishes in two dimensions, IEC6 cells displayed a rounded shape and EGF did not affect the cell morphology dramatically (Fig. 6A). P2Y₂R–HA-expressing IEC6 cells displayed a similar shape to control IEC6 cells, and EGF induced cell elongation (Fig. 6A) as occurs when Arl4c is expressed (Matsumoto et al., 2014). RGDfV cooperatively induced cell elongation with EGF in a dose-dependent manner up to 5 μg/ml, but 25 μg/ml RGDfV did not affect cell morphology (Fig. 6B).

When IEC6 cells were cultured in 3D Matrigel, YAP and TAZ were observed in the cytoplasm of IEC6 cells in cysts, even in the presence of EGF (Fig. 6C). In P2Y₂R–HA-expressing IEC6 cells, EGF induced nuclear localization of YAP and TAZ in the leading regions of tubular structures, where the cells are elongated, but not in the trunk regions, indicating that these leading cells are proliferating (Fig. 6C). These results suggest that P2Y₂R expression induces cell morphological changes and proliferation, which is consistent with our model for Arl4c expression (Matsumoto et al., 2014). Although Arl4c expression suppressed RhoA activity (Matsumoto et al., 2014), P2Y₂R expression actually weakly activated Rho (Fig. 6D). Therefore, P2Y₂R might regulate tubular formation by affecting components that are downstream of RhoA signaling. Treatment with a low concentration (2 μM) of Y27632 augmented tubular formation of P2Y₂R-expressing IEC6 cells in a combination with EGF, but a high concentration (100 μM) of Y27632 inhibited this (Fig. 6E). Thus, the appropriate inhibition of actomyosin contraction is important for tubular formation in IEC6 cells even though RhoA is activated.

In this study, we found that Wnt3a- and EGF-induced P2Y₂R expression is required for tubular formation of IEC6 cells in 3D-BMM culture. In our previous report (Matsumoto et al., 2014), we showed that a combination of Wnt3a and EGF induces Arl4c expression synergistically and that the expression of Arl4c is required for the formation of tubular structures. During the expression of Arl4c mRNA, Ets1, a transcriptional factor of the ERK1/2 pathway, constitutively binds to the 3′-untranslated region of the Arl4c gene. Tcf4 forms a complex with Ets1, and simultaneous stimulation with Wnt3a and EGF promotes the association of β-catenin with a Tcf4 and Ets1 complex, thereby increasing histone acetylation around the Ets1-binding region of the Arl4c gene. However, we were unable to identify the binding regions for Ets-family proteins, Tcf4 or LEF1 on the P2yr2 gene. In addition, although inhibition of EGFR suppressed Wnt3a- and EGF-dependent P2yr2 mRNA expression, inhibition of MEK, JNK and PI3K did not affect it, suggesting that these well-known pathways activated by EGF are not be involved in P2Y₂R expression. Instead, we found that TGF-β signaling through Smad2 and Smad3 mediates Wnt3a- and EGF-dependent P2yr2 mRNA expression. The Alk5 inhibitor A83-01 also inhibited Wnt3a- and EGF-induced tubular formation, and it can be deduced that TGF-β signaling is indeed required for tubulogenesis of IEC6 cells. Therefore, Wnt3a and EGF regulates Arl4c and P2Y₂R expression through different mechanisms.

As examples of crosstalk between EGF and TGF-β signaling, it has been shown that TGF-β potentiates EGF-dependent activation of MEK1 and ERK1/2, thereby enhancing EGF signaling in C3H10T1/2 cells (Yan et al., 2000) and that TGF-β upregulates receptor type protein tyrosine phosphatase-κ, which attenuates EGF-dependent EGFR tyrosine phosphorylation, thereby inhibiting EGF signaling in keratinocytes (Xu et al., 2010). In addition, there is crosstalk between Wnt and TGF-β signaling; for example, it has been shown that TGF-β upregulates Tcf4 and increases Tcf4 transcriptional activity during osteoblast maturation (McCarthy and Centrella, 2010). Although it is currently uncertain how TGF-β signaling is involved in Wnt3a- and EGF-dependent gene expression and tubular formation in IEC6 cells, Wnt3a and EGF signaling could potentiate the quality and quantity of the TGF-β signaling components, such as ligands, receptors and Smads.

P2Y₂R acts as a receptor for ATP and UTP (Burnstock, 2007). Extracellular endogenous nucleotides have been shown to regulate various cellular functions, including proliferation, differentiation, migration and death (Burnstock, 2006). Therefore, ATP-mediated signaling might cooperate with Wnt3a and EGF signaling to form epithelial tubular structures. It has been reported that colon P2Y₂R expression is increased in patients with inflammatory bowel diseases (Grbic et al., 2008). In IEC6 cells, ATP- or UTP-dependent P2Y₂R activation has been shown to inhibit GSK-3 through AKT activation and stabilized microtubules, thereby inducing cell migration (Degagné et al., 2013). These results suggest that inflammation-dependent P2Y₂R expression could be involved in the wound healing process of intestinal regeneration because ATP and UTP are released from luminal bacteria, apoptotic cells and intestinal immune cells. Furthermore, interleukin-1β, which is released from microglial cells and macrophages in neurodegenerative diseases, increases P2Y₂R expression in primary cortical neurons (Kong et al., 2009; Peterson et al., 2013). ATP and UTP promote Gq-dependent Ca²⁺ mobilization and G₁₂-dependent Rho activation through P2Y₂R, leading to neurite extension, which might be involved in neuroprotective responses (Peterson et al., 2013). These nucleotides also promote the processing of the amyloid precursor protein. In addition, β-amyloid upregulates P2Y₂R in microglial cells, which enhances phagocytosis and degradation of β-amyloid (Kim et al., 2012). Thus, P2Y₂R expression in neurons and microglial cells under pro-inflammatory and pro-apoptotic conditions might prevent neurodegeneration. In these scenarios, injury- or stress-dependent P2Y₂R expression responds to extracellular nucleotides released from the neighboring cells and tissues.

As apyrase, which hydrolyzes ATP and ADP, specifically (Komoszyński and Wojtczak, 1996) suppressed Wnt3a and EGF-dependent tubular formation in IEC6 cells, it appears that extracellular endogenous ATP is involved in epithelial morphogenesis. However, it is unlikely that P2Y₂R expression mediated by Wnt3a and EGF regulates extracellular ATP signaling, because expression of a P2Y₂R mutant (P2Y₂R^R264L), which lacks the responsiveness to ATP, was able to enhance tubular formation of IEC6 cells and rescue the phenotypes in P2Y₂R-depleted IEC6 cells. Therefore, extracellular ATP could be involved in Wnt3a- and EGF-dependent tubular formation through nucleotide receptors other than P2Y₂R.

In addition to its function as a nucleotide receptor, P2Y₂R forms a complex with integrins α_vβ₃ and α_vβ₅ (Erb et al., 2001), and this interaction promotes nucleotide-dependent G-protein-mediated Ca²⁺ mobilization and cytoskeletal control (Bagchi et al., 2005; Chorna et al., 2007; Degagné et al., 2013; Erb et al., 2001; Liao et al., 2007). Unlike P2Y₂R^R264L, P2Y₂R^D97E, which lacks integrin-binding activity, was not able to rescue the phenotypes in P2Y₂R-depleted IEC6 cells. As this mutant also decreased the potency of UTP to less than three orders of magnitude (Qi et al., 2005), we cannot exclude the possibility that 1–10 mM ATP activates P2Y₂R^D97E and then regulates tubular formation. However, this possibility is unlikely, because ∼1 μM ATP was released from the colonic mucosa in in vivo conditions, and mechanical stress stimulated the increase of ATP by 10- to 20-fold (Patel et al., 2011). Therefore, the interaction of P2Y₂R with integrins can promote tubular formation independently of extracellular ATP.

Many adhesive proteins, including fibronectin, vitronectin, thrombospondin, tenascin-C and entactin-1, which are present in extracellular matrices, contain the RGD sequence as their adhesion recognition site (Ruoslahti and Pierschbacher, 1987). Among them, fibronectin was most abundantly expressed in IEC6 cells and deposited on a cell surface. Fibronectin is known to play essential roles in MDCK cell tubulogenesis and development of mouse submandibular salivary glands, kidneys and lungs (Jiang et al., 2000; Sakai et al., 2003). Therefore, we speculated that knockdown of fibronectin would affect tubular formation if this was dependent on the binding of integrins to fibronectin. Severe reduction of fibronectin suppressed Wnt3a- and EGF-dependent tubular formation, but its mild depletion instead induced tubular formation in the presence of EGF alone. These results suggest that, although fibronectin is required for IEC6 cell morphogenesis, baseline expression levels of fibronectin suppresses it, and appropriate inhibition of fibronectin expression is able to induce tubular formation without P2Y₂R expression. Therefore, P2Y₂R might compete with fibronectin for binding to integrins, and its expression might induce cell morphological changes and proliferation during tubular formation.

Previous reports have revealed that cell-to-substrate adhesion mediated by integrins is required for epithelial lumen morphogenesis (Datta et al., 2011). As constitutive expression of P2Y₂R in IEC6 cells led to the formation of discontinuous lumens during tubulogenesis, transient or spatially restricted expression of P2Y₂R induced by Wnt3a and EGF could be important for correct cell polarization and lumen morphogenesis. Wnt3a- and EGF-dependent Arl4c expression inhibits RhoA, resulting in cell extension and proliferation during tubulogenesis. P2Y₂R expression also induced cell elongation and proliferation, leading to tubular formation, which is consistent with our model for Arl4c expression (Matsumoto et al., 2014). However, P2Y₂R expression in fact slightly activated RhoA, which is consistent with previous reports that P2Y₂R is involved in RhoA activation in NIH3T3 and 1321N1 cells (Liao et al., 2007; Singh et al., 2010), but P2Y₂R induced cell extension and proliferation. Although the mechanisms of Arl4c- and P2Y₂R-dependent tubular formation are different, P2Y₂R expression could induce cell extension downstream of RhoA signaling. The role of P2Y₂R induced by Wnt3a and EGF could be to inhibit the interaction of integrins with fibronectin, thereby inducing cell morphological changes without RhoA inhibition. As depletion of either Arl4c or P2Y₂R inhibits Wnt3a- and EGF-induced tubular formation in IEC6 cells, expression of both proteins are required for this process. Further studies are necessary in order to fully understand the mechanisms underlying growth-factor-induced epithelial morphogenesis.

pcDNA3-integrin α_v and pcDNA3-integrin β₃ was provided by Dr Tomiyama (Department of Blood Transfusion, Osaka University Hospital, Osaka, Japan). pSecTag2-fibronectin-His and CSII-CMV-MCS-IRES2-Bsd were provided by Drs Sekiguchi (Institute for Protein Research, Osaka university, Osaka, Japan) and Miyoshi (RIKEN-BRC, Tsukuba, Japan) (Miyoshi et al., 1998), respectively. Wnt3a was purified to homogeneity as previously described (Kishida et al., 2004; Komekado et al., 2007). EGF was purchased from R&D Systems (Minneapolis, MN). CHIR99021 was from Stemgent (Cambridge, MA); A83-01 from Tocris Bioscience (Minneapolis, MN, USA); and LY294002 from Cell Signaling Technology (Beverly, MA). ATPγS and Y27632 were from Wako (Osaka, Japan). Apyrase was from New England Biolabs (Ipswich, MA). Suramin, PD168393, and RGDfV were from Calbiochem (San Diego, CA). U0126 was from Promega (Madison, WI). SP600125, UTPγS, and fibronectin were from Sigma-Aldrich (St Louis, MO). Laminin-111 was from BD Biosciences (San Jose, CA). The primary antibodies used in this study are listed in supplementary material Table S1. Other materials and chemicals were from commercial sources.

IEC6 cells were maintained in α minimum essential medium (αMEM) supplemented with 10% fetal bovine serum (FBS), penicillin-streptomycin, 4.5 mg/ml glucose, 10 μg/ml insulin, and 1× non-essential amino acids as described previously (Matsumoto et al., 2014). X293T and HeLaS3 cells were maintained in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% FBS and penicillin-streptomycin, and HEK293T cells were maintained in DMEM with Ham's F12 (1:1) supplemented with 10% FBS and penicillin-streptomycin.

Standard recombinant DNA techniques were used to construct the following plasmids: pcDNA3-rat P2Y₂R-HA, pcDNA3-human P2Y₂R-HA and pCAG-rat P2Y₂R-HA. For construction of cDNA-containing lentiviral vectors, EGFP, rat P2Y₂R–HA, rat P2Y₂R^R264L–HA and rat P2Y₂R^D97E–HA cDNAs were cloned into CSII-CMV-MCS-IRES2-Bsd.

DNA microarray analyses of IEC6 cells in 3D Matrigel culture were performed using gene microarray technology (GeneChip^® Rat Genome 230 2.0 array, Affymetrix, Santa Clara, CA, USA) as described previously (Matsumoto et al., 2014). Data analyses were performed with GeneSpring GX 11 software (Agilent Technologies, Santa Clara, CA), and the change in ratios between the hybridization intensities of Wnt3a- and EGF-treated, and control samples were determined.

The lentiviral vector CSII-CMV-MCS-IRES2-Bsd harboring cDNAs was transfected with the packaging vectors, pCAG-HIV-gp and pCMV-VSV-G-RSV-Rev, into X293T cells using FuGENE HD transfection reagent (Roche Applied Science, Basel, Switzerland). To generate IEC6 cells stably expressing GFP, P2Y₂R–HA, P2Y₂R^R264L–HA or P2Y₂R^D97E–HA, 5×10⁴ parental cells per well in a 12-well plate were treated with lentiviruses and 10 μg/ml polybrene, centrifuged at 1200 g for 30 min and incubated for 24 h. Cells were selected and maintained in the medium containing 2.5 μg/ml Blastcidin S.

Epithelial morphogenesis in 3D Matrigel (Corning, New York, NY) was analyzed as described previously (Matsumoto et al., 2014). Briefly, 40 μl of Matrigel was mounted on a round coverslip and incubated for 30 min at 37°C to solidify the gel. Where necessary, solid Matrigel was coated with 200 μl of PBS containing 40 μg of fibronectin or 300 μg of laminin-111 for 2 h at room temperature. IEC6 cells (4×10⁴ cells) suspended in 1 ml of growth medium containing 2% Matrigel (v/v), 40 ng/ml Wnt3a and/or 5 ng/ml EGF were added to solid Matrigel and incubated for 48, 60 or 72 h. To observe 3D epithelial morphogenesis in Matrigel at fixed points, P2Y₂R–HA-expressing IEC6 cells (5×10³ cells) were seeded sparsely. Cells at fixed points were photographed using phase contrast microscopy at 0, 24, 36, 48 and 72 h after seeding.

Where necessary, 2 μM A83-01, 50 μM apyrase, 25 μM suramin, 100 μM ATPγS, 1, 2.5, 5, 10 or 25 μg/ml RGDfV, or 2, or 100 μM Y27632 was added to the growth medium. The extended structure was defined as a chain- or cord-like structure containing at least four cells that extended from the main multicellular trunks as described previously (Matsumoto et al., 2014). The number of extended structures from the cysts without cellular extensions or that from cysts with cellular extensions from less than three cells were defined as ‘0.1’ for statistical reasons. The number of extended structures from multicellular trunks of cysts was counted.

For 2D culture, glass coverslips were coated with growth medium containing 2% Matrigel (v/v). P2Y₂R–HA-expressing IEC6 cells were grown on a Matrigel-coated dish in a minimum essential medium (MEM) containing 10% FBS.

Cell surface biotinylation was performed as described previously (Yamamoto et al., 2006). IEC6 cells were treated with or without Wnt3a and EGF for 48 h, and then the cells were incubated with ice-cold binding medium [α-MEM containing 20 mM HEPES-NaOH at pH 7.5 and 0.1% bovine serum albumin (BSA)] for 1 h at 4°C. Next, the cells was washed three times with cold PBS containing 1 mM MgCl₂ and 1 mM CaCl₂. The cells were incubated with 0.5 mg/ml sulfo-NHS-biotin (Pierce, Rockford, IL) for 30 min at 4°C. After quenching of excess biotin with 50 mM NH₄Cl in PBS, cells were lysed in 200 μl of TNE buffer (25 mM Tris-HCl pH 7.5, 150 mM NaCl and 5 mM EDTA) containing 0.4% Triton X-100, 0.4% sodium deoxycholate, 2 μg/ml leupeptin, 2 μg/ml aprotinin, and 100 μM phenylmethylsulfonyl fluoride. After centrifugation, the lysates were precipitated using neutravidin–agarose beads (Pierce) and the precipitates were detected with anti-P2Y₂R, anti-LRP6 and anti-β-tubulin antibodies.

HEK293T cells were seeded on 35-mm-diameter dishes coated with poly-D-lysine (Sigma, St Louis, MO), and pCAG-P2Y₂R-HA, CSII-P2Y₂R-HA, CSII-P2Y₂R^D97E-HA, pcDNA3-integrin α_v, pcDNA3-integrin β₃, and/or pSecTag2-fibronectin-His plasmids were transfected into the cells using Lipofectamine 2000 (Invitrogen, Life Technologies, Inc., Carlsbad, MA). At 24 h after transfection, cells were washed once with PBS and lysed in 200 μl lysis buffer [20 mM Tris-HCl pH 7.5, 150 mM NaCl, 2 mM EGTA, 1% NP40 with protease inhibitors (1 mM phenylmethylsulfonylfluoride, 1 µg/ml leupeptin, and 1 µg/ml aprotinin)] for 10 min on ice. After centrifugation, the supernatant was collected and rotated with 1.0 μg of each antibody for 2 h at 4°C, and then rocked with 30 µl of protein-G– or protein-A–Sepharose (50% slurry) for 1 h at 4°C. The beads were then washed three times with lysis buffer and finally suspended in Laemmli's sample buffer.

Cells grown on glass coverslips were fixed for 10 min at room temperature in PBS containing 4% (w/v) paraformaldehyde and permeabilized in PBS containing 0.2% (w/v) Triton X-100 and 2 mg/ml BSA for 10 min. IEC6 cells grown in 3D culture were fixed for 30 min at room temperature in PBS containing 4% (w/v) paraformaldehyde and permeabilized and blocked in PBS containing 0.5% (w/v) Triton X-100 and 40 mg/ml BSA for 30 min. Cells were incubated with primary antibodies for 3 h at room temperature or overnight at 4°C, and with secondary antibodies in accordance with the manufacture's protocol (Molecular Probes, Carlsbad, CA). Samples were viewed and analyzed using LSM510 laser scanning microscope (Carl-Zeiss, Jena, Germany).

Cells grown on glass coverslips were fixed for 10 min at room temperature in PBS containing 4% (w/v) paraformaldehyde and permeabilized in PBS containing 0.2% (w/v) Triton X-100 and 2 mg/ml BSA for 10 min. Slides were then blocked in blocking solution (Olink Bioscience, Uppsala, Sweden) for 30 min and incubated with primary antibodies diluted in blocking solution for 1 h at room temperature. After washing, slides were incubated with Duolink PLA Rabbit MINUS and PLA Mouse PLUS proximity probes (Olink Bioscience) and proximity ligation was performed using the Duolink detection reagent kit (Olink Bioscience) according to the manufacturer's protocol. PLA dots were detected using a LSM510 laser scanning microscope and counted.

Total RNA was isolated from IEC6 cells and quantitative RT-PCR was then performed as described (Hino et al., 2005). Forward and reverse primers were as follows: rat GAPDH, 5′-ATCAACGACCCCTTCA-3′ and 5′-TTTGGCCCCACCCTTC-3′; rat P2ry1, 5′-CGGAGAGGAGAGTTGTCCAG-3′ and 5′-CTGGGATTCGGAAAAACAAA-3′; rat P2ry2, 5′-CTCTACTTTGTCACCACCAGCGC-3′ and 5′-GTGGTCCCATAAGCCGGTTTG-3′; rat P2ry4, 5′-ACTAGGTCCCAGCCCAAGTT-3′ and 5′-GTGTCTGACAATGCCAGGTG-3′; rat P2ry6, 5′-GGGTGGTATGTGGAGTCGTT-3′ and 5′-TAGCAGGCCAGTAAGGCTGT-3′; rat P2ry10, 5′-TCATGCTTTGCTGATCTTGG-3′ and 5′-ACATGAAAACCATCCGCAAT-3′; rat P2ry12, 5′-CTTCGTTCCCTTCCACTTTG-3′ and 5′-CTCAGCATGCTCATCAAGGA-3′; rat P2ry13, 5′-GCTGCTGTGGCATCAAGTAG-3′ and 5′-GGGCAAAGCAGACAAAGAAG-3′; rat P2ry14, 5′-TCTTCATCACAGGGGTCCTC-3′ and 5′-ACGGCAGACACCCTGAATAC-3′; rat Arl4c, 5′-CTGCTGGTCATCGCCAACAA-3′ and 5′-CCTGAAACGCAGGAAGTCTC-3′; rat fibronectin, 5′-GAAAGGCAACCAGCAGAGTC-3′ and 5′-CTGGAGTCAAGCCAGACACA-3′; rat vitronectin, 5′-ACCCTGATTATCCCCGAAAC-3′ and 5′-CAAACACGGCTGACAGAGAA-3′; rat thrombospondin-1, 5′-CCAGTTCAACCAACGTCCTT-3′ and 5′-TTGCGAATGCTGTCCTGTAG-3′; rat throm-bospondin-2, 5′-CCTTTTCAGCATCAGCAACA-3′ and 5′-ACCTTCCAACACCAGGAGTG-3′; rat tenascin-1, 5′-GTTTGGAGACCGCAGAGAAG-3′ and 5′-ACCTGCCGTTCCACTGTATC-3′; rat entactin-1, 5′-AGAGAGGCTTCCCAGAGGTC-3′ and 5′-CACGGCAGGTACTTGGTTTT-3′; rat nephronectin, 5′-CTGGGGACAGTGTCAACCTT-3′ and 5′-GTCAAGAGGAGTTGGGTGGA-3′.

In analyses using siRNAs, the following target sequences were used: randomized control, 5′-CAGTCGCGTTTGCGACTGG-3′; rat P2ry2, 5′-GCAAACAGCTGCCATTCCT-3′; rat β-catenin, 5′-CCATGGAGCC-AGACAGAAA-3′; rat Smad2, 5′-GGAGCTCATTGGAAGACTT-3′; rat Smad3, 5′-GCTTCTTGCCCTGAGGTTT-3′; rat Arl4c, 5′-GCTGTGGG-AACTGAGTAAT-3′; rat fibronectin #1, 5′-GGATCAAAGGGAAAC-ACAG-3′; rat fibronectin #2, 5′-TCTGGGATCAAAGGGAAAC-3′. IEC6 cells were transfected with a mixture of siRNAs against genes of interest, at 0.25−20 nM each, using RNAiMAX (Invitrogen, Carlsbad, CA) and cells were used for experiments at 36–48 h post-transfection.

Rho activity was measured as described previously (Matsumoto et al., 2014). Western blotting data were representative of at least three independent experiments.

The experiments were performed three or four times and the results are expressed as mean±s.d. Statistical analysis was performed using JMP Pro 10. Differences between the data were tested for statistical significance using Welch's t-test or Student's t-test. P<0.05 was considered statistically significant.

We thank Dr T. Yamamichi for technical assistance. We also thank Drs H. Miyoshi, Y. Tomiyama, H. Kato and K. Sekiguchi for donating plasmids.