CDCP1 controls compensatory renal growth by integrating Src and Met signaling

Compensatory growth of organs after loss of mass and/or function is controlled by hepatocyte growth factor (HGF), but the underlying regulatory mechanisms remain elusive. Here, we show that the transmembrane glycoprotein CUB domain-containing protein 1 (CDCP1) controls compensatory renal growth by acting as a critical regulator of HGF signaling. In MDCK cysts, HGF induces the temporal upregulation of Src and its scaffold CDCP1, and the ablation of CDCP1 abrogates HGF-induced morphological change and cell growth. Mechanistically, upregulated CDCP1 recruits Src into lipid rafts to activate STAT3 associated with the HGF receptor Met, resulting in the upregulation of downstream factors required for HGF functions. In vivo, CDCP1 is upregulated in the renal tubules after unilateral nephrectomy, and the ablation of CDCP1 significantly suppresses compensatory renal growth by attenuating Met-STAT3 signaling. These findings demonstrate that CDCP1 controls HGF-induced compensatory renal growth by integrating Src and Met signaling, and provide new insights into the regulatory mechanisms underlying the multifaceted functions of HGF during regenerative growth of various organs.


Introduction
Controlling of organ size during development and/or regenerative growth is important to maintain organ function, body homeostasis, and health. The kidneys are paired organs that generate urine through the filtration of blood and reabsorption of water and nutrients, and kidney mass is strictly related to total body mass. The renal tubules constitute most of the mass and function, and have a remarkable capacity to undergo regenerative growth. Unilateral nephrectomy (UNX), a surgical procedure to reduce kidney mass, increases fluid flow in the remaining kidney and promotes the subsequent growth and proliferation of tubular epithelial cells to compensate for the increased flow (Fine, 1986). This compensatory renal growth is regulated via the activation of mTOR signaling pathways (Chen et al, 2005;Chen et al, 2009;Chen et al, 2015). However, interfering with the function of this pathway does not completely suppress renal growth, suggesting the potential contribution of one or more additional signaling pathways.
Compensatory renal growth also requires a number of growth factors (Hammerman et al, 1993), among which hepatocyte growth factor (HGF) plays a prominent role (Trusolino et al, 2010). HGF is produced by the surrounding or distal mesenchyme in the remaining kidney immediately after UNX (Ishibashi et al, 1992;Joannidis et al, 1994;Nagaike et al, 1991), inducing the upregulation of its receptor, Met, in the renal tubules (Ishibashi et al, 1992). In addition, HGF promotes dynamic morphogenesis through the induction of epithelialmesenchymal transition (EMT) in epithelial cells during development and regenerative growth of the kidney (Chang-Panesso & Humphreys, 2017;Matsumoto & Nakamura, 2001). HGFmediated morphogenesis requires STAT3 signaling (Boccaccio et al, 1998), which is regulated by Met through endosomal trafficking (Kermorgant & Parker, 2008). However, the molecular mechanisms through which the multifaceted HGF functions are precisely controlled during compensatory renal growth remain elusive.
Madin-Darby Canine Kidney (MDCK) cell-line was derived from renal tubule epithelial cells and is a physiologically relevant in vitro model to study the regulation of HGF functions in the kidney (O'Brien et al, 2002;Zegers, 2014). When grown in three-dimensional cultures, MDCK cells spontaneously form spherical cysts that resemble renal tubules, comprising an epithelial monolayer and lumen. Upon stimulation with HGF, MDCK cysts undergo morphological alterations and form branched tubular structures (Montesano et al, 1991a;Montesano et al, 1991b). During this morphogenesis, MDCK cells lose their epithelial polarity via a partial EMT-like phenotypic change, and protrude into the extracellular matrix (ECM) by penetrating the basement membrane (O'Brien et al, 2004;Pollack et al, 1998). In addition, HGF promotes cell growth and proliferation, resulting in the formation of multi-cell layered cysts. To elucidate the mechanisms underlying HGF-induced phenotypic changes, the roles of multiple signaling axes downstream of Met, such as the Ras-ERK, Akt-mTOR, Src, and STAT3 pathways, have been investigated extensively (Khwaja et al, 1998;Nakamura et al, 2011;O'Brien et al, 2004;Pollack et al, 1998;Santos et al, 1993). However, the molecular mechanism by which these diverse signaling pathways are accurately coordinated by HGF-Met needs to be clarified. Furthermore, the most critical pathway for HGF-induced compensatory renal growth remains undefined.
Here, using three-dimensional cultures of MDCK cysts as a model of renal tubules, we identified Src and its membrane scaffold, CUB domain-containing protein 1 (CDCP1), as critical elements of HGF signaling. We then investigated the mechanisms underlying the functions of the CDCP1-Src axis in MDCK cysts and verified its physiological roles in compensatory renal growth using Cdcp1-knockout mice. The results obtained in these studies demonstrate that CDCP1 plays a crucial role in controlling HGF-induced compensatory renal growth by focally and temporally integrating Src and Met-STAT3 signaling on lipid rafts.

CDCP1 is required for HGF signaling in MDCK cysts
MDCK cells cultured in a collagen matrix formed cyst structures with a luminal space (Fig 1A). HGF treatment immediately promoted cell growth/proliferation (as indicated by increased cyst diameter) and morphological changes with some cysts exhibiting extended cell protrusions (Gao & Vande Woude, 2005;O'Brien et al, 2002) (Fig 1, A-C). Similar HGFinduced effects were observed in primary cultured renal proximal tubule cells (Bowes et al, 1999), indicating that MDCK cysts are suitable for the mechanistic analysis of HGF functions. To dissect the HGF-related signaling pathways, we first examined the effects of various signaling inhibitors on HGF-induced phenomena. Treatment of MDCK cysts with NK4, a specific HGF antagonist (Date et al, 1997), robustly inhibited cell proliferation and morphological changes, confirming that the observed phenomena were dependent on HGF. Treatment of the cysts with Torin1 or rapamycin, selective mTOR inhibitors, significantly suppressed cell growth/proliferation but was less effective against the formation of cell protrusions (Fig 1, A-C). By contrast, treatment with dasatinib, a Src kinase inhibitor, potently suppressed both cell proliferation and protrusion extension (Fig 1, A-C). Furthermore, an immunofluorescence analysis revealed that activated Src (pY418) was concentrated at the tip of protruding cells (Fig 1D) and co-localized with mCherry-GPI, a marker protein of lipid rafts (Fig 1E). These observations imply that HGF signaling is associated with the activation of Src in lipid rafts on protruding cells.
To assess its role in HGF signaling further, we examined the effects of Src activation on MDCK cells by expressing Src-MER, a Src protein fused to a modified estrogen receptor (MER) (Kuroiwa et al, 2011) that could be activated by treatment with hydroxytamoxifen (4-OHT) (Fig EV1A). Src activation induced both the formation of multiple cell protrusions and cell proliferation (Fig EV1B), supporting the involvement of Src activity in HGF signaling. A detergent-resistant membrane (DRM) separation analysis revealed that activated Src-MER was concentrated in lipid raft fractions (Fig EV1C). Because Src alone can only partially localize to lipid rafts (Oneyama et al, 2009), we searched for scaffolding proteins that accommodate and activate Src in these structures. Src-MER interacting proteins were isolated from DRM fractions by co-immunoprecipitation and identified by mass spectrometry (Fig EV1, D and E). Among the candidate proteins identified, we selected the transmembrane glycoprotein CDCP1 for further analysis because it has palmitoylation sites required for lipid raft localization and serves as a membrane adaptor of Src (Alvares et al, 2008;Wortmann et al, 2009) (Fig EV2A). We then validated the roles of CDCP1 in HGF signaling. An immunofluorescence analysis revealed that endogenous CDCP1 was transiently concentrated at the tips of protruding cells during the early stage of morphological changes (Fig 1F), in a manner similar to that of activated Src and mCherry-GPI (Fig 1, D and E). More importantly, HGF-induced cell protrusion extension and cell proliferation were efficiently abrogated by Cdcp1-knockout in MDCK cells (Fig 1, G-I and Fig S2, B-D). These results suggest that the CDCP1-Src axis is a crucial component of HGF signaling.

Upregulation of CDCP1 induces HGF-related phenotypic changes by activating Src in lipid rafts
To elucidate the mechanisms through which CDCP1 regulates HGF signaling, we generated MDCK cells in which the expression of CDCP1-EGFP could be controlled by doxycycline (Dox). Dox treatment induced the expression of full-length and cleaved CDCP1 at the plasma membrane (Fig EV3, A and B). A DRM separation analysis showed that a fraction of CDCP1 was distributed in lipid raft-enriched fractions (Fig EV3C). Moreover, the induction of CDCP1 expression dramatically increased the amount of phosphorylated Src (pY418) without affecting Src protein levels (Fig EV3A), indicating that Src was activated by the upregulation of CDCP1. Furthermore, activated Src was concentrated in the lipid raft fractions and was physically associated with CDCP1 (Fig EV3, C and D). By contrast, a CDCP1 mutant that lacked the Src-binding site (Y734F; CDCP1-YF) failed to activate Src (Fig EV3A). A different CDCP1 mutant lacking the lipid raft localization signal (C689G−C690G; CDCP1-CG) was able to activate Src, but only in non-lipid raft fractions ( Fig  EV3, A and C). These findings corroborate the notion that CDCP1 specifically activates Src in lipid rafts.
Next, we examined the effects of CDCP1 expression on the phenotypes of MDCK cysts (Fig 2A). CDCP1 expression was induced after the completion of cystogenesis and morphological changes were observed (Movie EV1). In the early stages (12−24 h after induction), cells expressing CDCP1 gradually protruded toward the ECM. During the later stages (48 h after induction), multiple cell protrusions formed randomly and extended aggressively toward the ECM, with cysts forming multiple layers (Fig 2B). Concurrent with these phenomena, upregulation of CDCP1 activated Src in the protruding cells (Fig 2C), and inhibition of Src by dasatinib strongly suppressed CDCP1-induced cell proliferation and protrusion extension (Fig 2D and Table EV1). Meanwhile, the upregulation of CDCP1-YF or CDCP1-CG did not significantly induce phenotypic changes (Fig 2, E and F). Based on these results, it is likely that upregulation of CDCP1 can induce HGF-related phenotypic changes, i.e., cell protrusion formation and cell proliferation, in MDCK cysts via the activation of Src in lipid rafts.

CDCP1 activates the STAT3 pathway via lipid rafts
To dissect the signaling pathways downstream of the CDCP1-Src axis, we performed DNA microarray analyses of MDCK cysts expressing either wild-type CDCP1 or the CDCP1-CG mutant. A gene ontology analysis implied that CDCP1 expression is involved in the regulation of signal transduction (Fig EV4A) and Ingenuity Pathway Analysis revealed that the STAT3 pathway was more prominently activated by CDCP1 expression than by CDCP1-CG expression (Fig EV4, B and C, Dataset EV1). Indeed, the level of phosphorylated STAT3 (pY705) increased significantly in cysts expressing wild-type CDCP1 (Fig 3A). STAT3 is a transcription factor that is activated by Src-mediated phosphorylation of Tyr705 (Bromberg et al, 1998) and regulates various cellular events including the upregulation of MMPs (Yu et al, 2009) and mitogenic genes.
A qPCR analysis confirmed that the expression levels of several matrix metalloproteinase (MMP)-encoding genes were significantly increased by upregulation of CDCP1 (Fig 3B). In addition, immunofluorescent staining of laminin revealed that CDCP1 expression disrupted the basement membrane, allowing the extension of cell protrusions ( Fig  EV5A). This effect was suppressed by treatment with marimastat, a pan-MMP inhibitor ( Fig  EV5, B and C). Similar suppression of cell protrusion extension by marimastat was also observed in HGF-stimulated MDCK cysts (Fig EV5, D and E). Analysis with DQ-collagen, a fluorescent indicator of collagen degradation, also showed that CDCP1 expression induced degradation of the ECM (Fig EV5F). These observations suggest that the HGF-induced formation of cell protrusions is associated with ECM rearrangements via STAT3-induced upregulation of MMPs. The expression of mitogenic genes such as Myc (Myc) and Ccnd1 (Cyclin D1) was also upregulated in cysts expressing CDCP1 (Fig 3B), indicating that CDCP1-induced cell proliferation is attributable to STAT3 activation.
We confirmed the contribution of STAT3 to cell protrusion formation and proliferation by specifically perturbing its activity. Activation of STAT3 using the STAT3-MER system induced the formation of cell protrusions and a multi-layered structure (Fig EV5G). By contrast, treatment of MDCK cysts with STAT3-specific inhibitors suppressed CDCP1-induced cell growth (Fig 3C and Table EV1), and overexpression of dominant negative STAT3 (STAT3-Y705F) also inhibited CDCP1-induced cellular events (Fig 3D). Furthermore, STAT3 inhibitors efficiently suppressed cell protrusion formation (Fig 3E) and proliferation (Fig 3F) in HGF-stimulated MDCK cysts, underscoring the crucial role of STAT3 activation in HGF functions.

CDCP1 focally integrates Src and Met-STAT3 signaling
Next, we investigated the functional link between the HGF-Met pathway and the CDCP1-Src axis. An inhibitor screening assay showed that various Met-specific inhibitors potently suppressed CDCP1-induced phenotypic changes (Fig 4A and Table EV1), suggesting that Met activity is involved in the function of CDCP1. Because activated Met associates with STAT3, we predicted a close association between CDCP1-Src and Met-STAT3. To assess physical interactions between these proteins, Met and various CDCP1 mutants that lacked particular extracellular CUB domains were co-expressed in HEK293 cells (Fig 4, B and C). Co-immunoprecipitation assays revealed that CDCP1 interacted with Met. This interaction was enhanced by removal of the first CUB domain of CDCP1 but diminished by deletion of all CUB domains (Fig 4, B and C), indicating that CDCP1 interacts with Met through extracellular domains, and that removal of the first CDCP1 CUB domain is required for efficient association with Met.
Because CDCP1 is activated via proteolytic shedding between the first and second CUB domains (Casar et al, 2012;He et al, 2016), it is likely that activated CDCP1 functionally interacts with Met. To verify this hypothesis, we analyzed the function of a mutant CDCP1 (CDCP1-PR) rendered resistant to proteolytic shedding due to the presence of three point mutations within the protease recognition sites (K365A-R368A-K369A) (Fig 4D). Expression of CDCP1-PR induced the activation of Src to a level similar to that observed with wild-type CDCP1 but failed to activate STAT3 (Fig 4D). Consistent with these biochemical effects, of CDCP1-PR failed to induce the formation of cell protrusions (Fig 4, E and F). These data suggest that an association between cleaved CDCP1 and Met is required for Src-mediated STAT3 activation and subsequent cellular events (Fig 4G). The functional interaction between Met and CDCP1 was further confirmed using Cdcp1-knockout MDCK cysts. The initiation of cell protrusion induced by HGF treatment or overexpression of Met was efficiently suppressed by the loss of CDCP1 (Fig EV6A). Taken together, these lines of in vitro evidence suggest a model in which the upregulation of CDCP1-Src in lipid rafts is required for Met-mediated STAT3 activation during HGF-induced phenotypic changes (Fig EV6B).

Compensatory renal growth is suppressed in Cdcp1-knockout mice
To verify the aforementioned model in vivo, we investigated the role of the CDCP1mediated Met-STAT3 pathway in compensatory renal growth following UNX. For this in vivo study, we generated Cdcp1-knockout mice in a C57/BL6 background using the CRISPR-Cas9 system (Fig EV7, A-C). The Cdcp1-knockout (Cdcp1 −/− ) mice grew normally and did not show any overt phenotype under normal conditions (Fig EV7, C-E), as reported previously (Spassov et al, 2013). Eight weeks after UNX, the remaining kidney was enlarged in wild-type and heterozygous (Cdcp1 +/− ) mice (Fig 5A). Kidney/body weight ratios in wildtype and Cdcp1 +/− mice were elevated to approximately 136% and 135% relative to those in sham-operated mice, respectively. By contrast, the kidney/body weight ratio elevation was significantly smaller in Cdcp1 −/− mice (approximately 121%) (Fig 5B). Since compensatory renal growth is achieved via the expansion of proximal renal tubules (Anderson, 1967;Chen et al, 2015), we visualized the renal tubules using a fluorescein-labeled specific lectin (LTL-FITC). Wild-type mice exhibited prominent thickening of LTL-positive proximal renal tubules (to approximately 148%), whereas this thickening was suppressed in Cdcp1 −/− mice (approximately 117%) (Fig 5, C and D). These data were consistent with the observed elevation of the kidney/body weight ratios (Fig 5B), confirming that compensatory growth of proximal renal tubules is suppressed in Cdcp1 −/− mice.

Met-STAT3 signaling is attenuated in Cdcp1-knockout mice
To address the cause of defective renal growth in Cdcp1 −/− mice, we analyzed the remaining kidney at earlier stages of compensatory growth (within 4 days after UNX) because the expression of HGF and Met is transiently upregulated in renal tissues within 12 h (Ishibashi et al, 1992;Nagaike et al, 1991). The mass of the remaining kidney immediately increased in wild-type mice (Fig 6A), whereas acute enlargement of the kidney tended to be delayed in Cdcp1 −/− mice. Immunofluorescence analysis revealed that activation of Met (pY1234/1235) and STAT3 (pY705) occurred in wild-type renal tubules 12 h after UNX, as reported previously (Kermorgant & Parker, 2008), while the activation of both signals was appreciably attenuated in Cdcp1 -/renal tubules (Fig 6, B and C). Activation of CDCP1 (pY734) also ocurred in renal tubules in a manner similar to that of Met and STAT3 ( Fig  EV8A). Notably, activated Met, STAT3, and CDCP1/Src were co-localized in a subset of intracellular small vesicles that were positive for EEA1 (Fig EV8, B-D), a marker of early endosomes, supporting the intimate interactions among these signaling molecules. These results suggest that the CDCP1-Src axis mediates HGF-Met-STAT3 signaling even in proximal renal tubules.
We also analyzed cell proliferation by immunostaining for Ki67. The ratios of Ki67positive cells in wild-type proximal tubules were gradually elevated and peaked 2 days after UNX (Fig 6, D

and E). However, in Cdcp1
−/− mice, the appearance of Ki67-positive cells was delayed, and the overall number of proliferating cells was decreased. These results suggest that HGF-induced cell proliferation through Met-STAT3 signaling is also attenuated in Cdcp1 −/− mice, thereby retarding the onset of compensatory renal growth.
Finally, we examined the effects of CDCP1 loss on the surrounding environment of proximal tubules after UNX. Proximal tubules are covered by a basement membrane comprised mainly of collagen IV. Before UNX, collagen IV was observed around the proximal tubules in both wild-type and Cdcp1 −/− mice (Fig 6F). In wild-type mice, collagen IV was gradually degraded 24 h after UNX, but this degradation was attenuated in Cdcp1 −/− mice. To identify the molecules involved in basement membrane degradation, we focused our analysis on secretory MMPs, including MMP2 and MMP9, both of which are targets of STAT3 and are expressed in proximal tubules (Tan & Liu, 2012). In wild-type mice, MMP2 and MMP9 were upregulated and occasionally concentrated in vesicle-like structures in proximal tubules 24 h after UNX (Fig 6G and Fig EV8E). However, these phenomena were barely detectable in Cdcp1 −/− proximal tubules. These observations suggest that secretory MMPs contribute to compensatory renal growth downstream of CDCP1-Src-mediated Met-STAT3 signaling.

Discussion
To address the regulatory mechanisms underlying HGF-induced regenerative renal growth, we dissected the HGF signaling using MDCK cysts as a model of renal tubules, and identified Src and its scaffold CDCP1 as critical elements of HGF signaling. Upon stimulation with HGF, CDCP1 was temporally concentrated at the tip of protruding cells. Since previous reports showed that CDCP1 expression is induced by HGF (Gusenbauer et al, 2013) or by epidermal growth factor (EGF)-mediated activation of the MAPK pathway (Adams et al, 2015;Dong et al, 2012), it is likely that the observed upregulation of CDCP1 was caused by activation of the HGF-Met-MAPK pathway. We also found that continuous upregulation of CDCP1 induced the formation of multiple cell protrusions and cell growth/proliferation. Notably, when CDCP1 expression was halted by removing Dox from the media, cells in the protruding cord-like structures recovered epithelial features, resulting in the formation of a luminal structure (Fig EV9A). This finding suggests that the temporal upregulation of CDCP1 by growth factors such as HGF might trigger the initial phase of morphogenesis, specifically cell protrusion extension and growth, and that the coordinated negative feedback regulation of CDCP1 might be involved in establishing the compound tubular/acinar epithelial system. In our model system, CDCP1 was tyrosine-phosphorylated by Src, and trapped and further activated Src particularly in lipid rafts. Although CDCP1-mediated Src activation has been reported in other cell types (Liu et al, 2011;Wortmann et al, 2011), we demonstrated here that localization of the CDCP1-Src complex in lipid rafts is required for HGF-induced phenotypic changes in MDCK cysts. We also found that upregulation of CDCP1 focally activates Src in lipid rafts to induce the phosphorylation of STAT3 at Tyr705 (Bromberg et al, 1998;Yu et al, 1995). Furthermore, activated STAT3 induced the upregulation of MMPs and mitogenic factors such as Myc and Cyclin D1, which promote cell protrusion extension and cell proliferation, respectively. These findings suggest that the activation of STAT3 via CDCP1-Src is one of the major pathways involved in HGF signaling. We also found that a physical association between cleaved/activated CDCP1 and Met is required for STAT3 activation. Because STAT3 interacts directly with activated Met through an SH2 domain (Boccaccio et al, 1998), it is likely that the cleaved form of CDCP1 binds Src and interacts with the Met-STAT3 complex in lipid rafts to allow the efficient phosphorylation of STAT3 by Src (Fig 4G). This focal integration of the CDCP1-Src axis with the Met-STAT3 complex in lipid rafts might be crucial for the activation of STAT3 that is required for HGF function. CDCP1 can also interact with other transmembrane receptors, including HER2, EGFR, integrin β1 and E-cadherin (Alajati et al, 2015;Casar et al, 2014;Law et al, 2013;Law et al, 2016). For example, HER2 associates stably with CDCP1 and forms a heterodimer complex on the plasma membrane; this interaction enhances Src activity in an EGF-independent manner (Alajati et al, 2015). Together with our results presented here, these previous findings indicate that CDCP1 is a more general regulator of membrane receptor signaling, which involves Src activation.
Upregulation of CDCP1 has been implicated in tumor progression (Alajati et al, 2015;Awakura et al, 2008;Dong et al, 2012;Emerling et al, 2013;He et al, 2016;Hooper et al, 2003;Ikeda et al, 2009;Leroy et al, 2015;Miyazawa et al, 2010;Scherl-Mostageer et al, 2001;Turdo et al, 2016;Wright et al, 2016). In some cancer cells, upregulated CDCP1 promotes invasion, metastasis, and tumor growth (Alajati et al, 2015;Casar et al, 2012;Kollmorgen et al, 2013;Lin et al, 2014;Liu et al, 2011;Miyazawa et al, 2010;Turdo et al, 2016;Uekita et al, 2007), although the underlying mechanisms remain elusive. Our findings presented here suggest that CDCP1-mediated activation of the Src-STAT3 pathway contributes to malignant progression by inducing invasive and growth promoting phenotypes, even in cancer cells. Indeed, we found that upregulation of CDCP1in MDCK cysts induced the expression of cytokeratin 14 (Fig EV9B), a promising marker of collective cancer invasion (Cheung et al, 2013), suggesting a potential role of upregulation of the CDCP1-Src axis in this process. Further analyses of the functions of the CDCP1-Src-Met-STAT3 pathway in a wide range of cancer cells could reveal new therapeutic targets for the treatment of some malignant cancers.
HGF is immediately upregulated after the loss of kidney mass and plays important roles in compensatory growth through the induction of cell proliferation and anti-apoptotic effects (Matsumoto & Nakamura, 2001). In this study, we uncovered the contribution of CDCP1-Src to HGF-induced compensatory renal growth using Cdcp1-deficient mice. Immunofluorescence analyses showed that activation of Met and STAT3 in the renal tubule cells following UNX was significantly attenuated in Cdcp1-deficient mice. Furthermore, upregulation of MMP2/9 and degradation of basement membrane collagen IV (Tan & Liu, 2012) were detected in the renal tubule cells after UNX (Fig EV8F). These observations are consistent with those of our in vitro study of MDCK cysts, supporting the crucial role of the functional integration of CDCP1-Src with Met-STAT3 pathways, even during compensatory renal growth. However, the renal growth was not completely suppressed by CDCP1 loss, suggesting that some other CDCP1-related molecules or distinct growth factor signaling pathways might compensate for the function of CDCP1 in vivo. Given that the expression of CDCP1 is temporally upregulated under pathological conditions such as tissue injury, hypoxia, and cancer recurrence (He et al, 2016), it is likely that upregulated CDCP1 might function to acutely amplify the HGF signaling that is required for regenerative organ growth in critical situations. Furthermore, since STAT3 is an important regulator of the regeneration of other organs such as the liver, intestine, and muscle (Lindemans et al, 2015;Taub, 2004;Tierney et al, 2014), it is also possible that the temporal upregulation of CDCP1 contributes more strongly to the promotion of organ regeneration by inducing focal activation of the Src-STAT3 pathway.
In conclusion, we found that CDCP1 serves as a critical regulator of HGF-induced compensatory renal growth by focally integrating Src and the Met-STAT3 pathway (Fig 6H). Our discovery of the CDCP1-Src axis as a new component of HGF-Met signaling provides insights into the regulatory mechanisms underlying the multifaceted functions of HGF during morphogenesis and/or regenerative growth, and might contribute to the development of more promising therapies for malignant cancers that are associated with the upregulation of CDCP1 and Met.

Materials and Methods Cell culture
MDCK type I and HEK293T cells were cultured in DMEM supplemented with 10% foetal bovine serum (FBS) at 37 °C in a 5% CO 2 atmosphere. Three-dimensional culture was performed using collagen type I (Cellmatrix Type I-A, Nitta Gelatin) according to the manufacturer's protocol. Collagen type I-A (3 mg/ml) was neutralised with reconstitution buffer (2.2% NaHCO 3 , 0.05 N NaOH and 200 mM HEPES) and diluted with 5× DMEM (Gibco). MDCK cells (1.5 × 10 5 cells/ml of the collagen gel) were combined with the collagen in DMEM supplemented with 5% FBS, and then polymerised at 37 °C in a 5% CO 2 atmosphere. The medium was replaced every 2 days. For the DQ-collagen degradation assay, a collagen gel containing 2% DQ-collagen type I (Molecular Probes) was used.

Mice
Cdcp1-knockout mice were generated in a C57BL/6N using the CRISPR/Cas9 system. Animals were housed in environmentally controlled rooms at the animal experimentation facility at Osaka University. All animal experiments were carried out according to the guidelines of the Osaka University committee for animal and recombinant DNA experiments and were approved by the Osaka University Institutional Review Board. The sequence of gRNA and primers used in genotyping are listed in Supplementary Table 3 (see also Supplementary Fig. 9a).

Microarray analysis and quantitative real-time PCR
For microarray analysis, total RNA was isolated from MDCK cysts using the Sepasol-RNA Kit (Nacalai Tesque). Microarray analysis was performed on a G2505C Microarray Scanner (Agilent Technologies) using the Canis (V2) Gene Expression Microarray 4×44K (Agilent Technologies). Microarray data were subjected to gene ontology analysis and upstream regulator analysis using the IPA Program (Qiagen, http://www.qiagenbioinfomatics.com/). For quantitative real-time PCR analysis, cDNA was prepared from RNA using the Transcriptor First Strand cDNA Synthesis Kit (Roche) according to the manufacturer's instructions. Real-time PCR was performed on a 7900HT Fast Real-Time PCR System (Applied Biosystems) using the Thunderbird qPCR Mix (Toyobo). Total RNA was normalised to expression of the housekeeping gene GAPDH. The primers used in this analysis are listed in Table EV2.

Immunofluorescent microscopy
For two-dimensional culture, cells were grown on collagen type I-coated coverslips, fixed with 4% paraformaldehyde and then permeabilized with PBS containing 0.03% Triton X-100. For three-dimensional culture, cysts embedded within the collagen matrix were fixed with 4% paraformaldehyde and permeabilized with PBS containing 0.5% Triton X-100. Permeabilized cells and cysts were blocked with 1% BSA and incubated with primary antibodies, and then incubated with Alexa Fluor 488/594-phalloidin (Molecular Probes). For the kidney sections, kidneys were prepared from perfusion fixation with 4% paraformaldehyde and dissected. The fixed kidneys were embedded in OCT compound (Sakura Finetek), sectioned, and mounted on glass slides. The kidney sections were blocked with Blocking One (Nacalai Tesque), incubated with primary antibodies, and then incubated with Alexa Fluor 488/594-conjugated secondary antibodies (Molecular Probes) and the fluorescein-conjugated proximal tubule marker, FITC-LTL (FL-1321, Vector Laboratories). Nuclei were counterstained with DAPI. Immunostained objects were observed under a FV1000 confocal microscope (Olympus). For time-lapse observation of three-dimensional cultures, cysts embedded within in collagen matrix were observed under a FV1200 confocal microscope (Olympus).

CRISPR/Cas9-based generation of CDCP1-knockout MDCK cells
Target-gRNA containing the pSilencer1.0-U6 plasmid, Cas9 and the EGFP coexpressing plasmid (pMJ920; Addgene) were transfected into MDCK cells with MDCK Cell Avalanche Transfection Reagent according to the manufacturer's protocol (EZ Bioscience). Three days after transfection, EGFP-positive single cells were isolated with a FACSAria III Sorter (BD Biosciences). Knockout of the Canis CDCP1 gene was confirmed by immunoblotting. The sequence of the gRNA and the primers used in genotyping are listed in Table EV3 (see also Fig EV2B).

Surgical procedures of mice
Compensatory renal growth was induced by right UNX in male mice (8 weeks of age) under anesthesia as previously described (Chen et al, 2015). Renal growth was evaluated by measuring the weight of remaining kidney and the body. Left kidneys of shamnephrectomised mice were used as controls for UNX mice.

Statistics and reproducibility
For data analyses, unpaired two-tailed t-tests were used to determine the P-values. For multiple group comparisons, two-way ANOVA was used. A P-value less than 0.05 was considered to be significant. All data and statistics were derived from at least three independent experiments.

Data availability
Supporting microarray data have been deposited in the Gene Expression Omnibus (GEO) under accession code GSE99375. All other supporting data are available from the corresponding author upon request.   (C) CDCP1-EGFP cysts were incubated in the presence of Dox (1 µg/ml) for 2 days. Activated Src was visualised with a Src pY418 antibody (magenta). (D) CDCP1-EGFP cysts embedded within the collagen matrix were pretreated with 20 nM dasatinib for 2 h, and then incubated with Dox (1 µg/ml) for 4 days. (E) CDCP1-YF-EGFP and CDCP1-CG-EGFP cysts embedded within the collagen matrix were incubated in the presence of Dox (1 µg/ml) for 4 days. Actin filaments were stained with Alexa Fluor 594-phalloidin (magenta). Scale bars indicate 50 mm. (F) Fraction of the total number of cysts counted (n > 150) with protrusions. The mean ratios ± SD were obtained from three independent experiments. ***, P < 0.001; ****, P < 0.0001; NS, not significantly different; ANOVA compared to the Dox-treated cysts.

Figure 3. STAT3 activation is required for CDCP1-induced phenotypic changes. (A)
CDCP1-myc-and mutant-overexpressing MDCK cells were embedded within the collagen matrix and cultured for 9 days. Cyst lysates were subjected to immunoblotting using the indicated antibodies. (B) CDCP1-EGFP cysts were incubated with Dox (1 µg/ml) for 2 or 4 days, and then subjected to quantitative real-time PCR. Relative mRNA expression levels were calculated by setting the mean value for non-treated cysts to one. The mean ratios ± SD were obtained from three independent experiments. ANOVA calculated compared to nontreated cysts. (C) CDCP1-EGFP cysts were pretreated with the indicated STAT3-specific inhibitors for 2 h and then incubated with Dox (1 µg/ml) for 4 days. Dotted blue indicates the average diameter of non-treated cysts. (D) STAT3-Y705F-overexpressing CDCP1-EGFP cysts were incubated with Dox (1 µg/ml) for 4 days. Actin filaments were stained with Alexa Fluor 594-phalloidin (magenta). Scale bars indicate 50 µm. (e, f) MDCK cysts embedded within the collagen matrix were pretreated with S3i-201 (+ S3i, 50 nM; ++ S3i, 100 nM) or Stattic (+ Sta, 2.5 nM; ++ Sta, 5.0 nM) for 2 hours and incubated in the presence of HGF (50 ng/ml) for 1 day. (E) Diameter (µm) of cysts (n = 100). (F) Fraction of the total number of cysts counted (n > 100) with protrusions. The mean ratios ± SD were obtained from three independent experiments. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001; NS, not significantly different; ANOVA was calculated compared to the HGF-treated cysts. Immunoprecipitates were subjected to immunoblotting using the indicated antibodies (IgG HC, IgG heavy chain). (D) CDCP1-myc-and CDCP1-PR-myc-overexpressing MDCK cells were embedded within the collagen matrix and cultured for 9 days. Cyst lysates were subjected to immunoblotting analysis using the indicated antibodies. (E) CDCP1-PR-EGFP cysts embedded within the collagen matrix were incubated with Dox (1 µg/ml) for 4 days. Actin filaments were stained with Alexa Fluor 594-phalloidin (magenta). Scale bars indicate 50 µm. (F) Fraction of the total number of cysts counted (n > 150) with protrusions. The mean ratios ± SD were obtained from three independent experiments. **, P < 0.001; ****, P < 0.0001; NS, not significantly different; ANOVA was calculated compared to the Dox-treated cysts. (G) Schematic model of the role of CDCP1-Met association in Src-induced STAT3 phosphorylation.

Figure 5. Compensatory renal growth after UNX is suppressed in Cdcp1-knockout mice.
(A) Cdcp1 wild-type (+/+), heterozygous (+/-), and homozygous knockout (-/-) mice at 8 weeks of age were subjected to left UNX or sham operation. Regenerative growth of remaining left kidney was analysed 8 weeks after operation and increases in remaining kidney/body weight ratios were assessed. (B) The mean ratios ± SD were obtained from ten mice per group. (C) The remaining kidney was removed from UNX-or sham-operated mice, and proximal tubules were stained with FITC-LTL (green). Scale bars indicate 50 mm. (D) Proximal tubules thickness was determined by the length of FITC-LTL-stained epithelial cell layer (n = 100). The mean ratios ± SD were obtained from five mice per group. *, P < 0.05; **, P < 0.01; ****, P < 0.0001; NS, not significantly different; two-way ANOVA. Figure 6. Met-STAT3 signalling is attenuated in Cdcp1-knockout mice. (A) Cdcp1 wildtype (+/+) and homozygous knockout (-/-) mice at 8 weeks of age were subjected to left UNX or sham operation. Regenerative growth of remaining left kidney was analysed at the indicated time points after operation and increases in remaining kidney/body weight ratios were assessed. The mean ratios ± SD were obtained from at least six mice per group. (B-D, F, G) The remaining kidney after UNX was subjected to microscopic immunofluorescence analysis with specific antibodies against Met pY1234-1235 (B), STAT3 pY705 (C), Ki67 (D), collagen IV (F), and MMP9 (G) and Alexa Fluor 594-conjugated secondary antibody (magenta). Proximal tubules were visualized by staining with FITC-LTL (green). Scale bars indicate 50 µm. Ki67-positive proximal tubules (%) were estimated by calculating the ratio of Ki67-stained tubules to the total number of tubules (n > 100) (E). The mean ratios ± SD were obtained from three mice per group. *, P < 0.05; **, P < 0.01; ****, P < 0.0001; NS, not significantly different; two-way ANOVA was calculated compared to sham operated control.
(H) Schematic model of HGF-induced adaptive renal regeneration. CDCP1-Src regulates the Met-STAT3 signalling leading to compensatory renal growth through induction of ECM rearrangement and cell growth/proliferation. In the cytoplasmic region, CDCP1 contains a Src association motif around Tyr734 and two palmitoylation sites (Cys residues 689 and 690), which are required for the localization of CDCP1 in lipid rafts. Upon phosphorylation at Tyr734 by Src, CDCP1 is additionally phosphorylated at Tyr762, resulting in direct association with PKCδ, which promotes cell migration. In the extracellular region, CDCP1 contains three CUB domains that are required for protein−protein interactions. CDCP1 also harbors a proteolytic cleavage (shedding) site between the first and second CUB domains and can, thus, be present in either the full-length or cleaved form depending on the cellular context. TM indicates transmembrane domain. (B) Schematic diagram of CRISPR/Cas9-based generation of CDCP1-knockout MDCK cells. Blue indicates the PAM sequence and bold letters indicate the start codon. The red arrowhead indicates the cleavage site. (C) Immunoblotting analysis of CDCP1-knockout MDCK cells. Lysates from wild-type and CDCP1-knockout MDCK cells were subjected to immunoblotting using the indicated antibodies. (D) CDCP1-knockout MDCK cysts were incubated in the presence of HGF (50 ng/ml) for 4 and 6 days. Ki67 was visualised with an Alexa Fluor 488-conjugated antibody (green) and actin filaments were stained with Alexa Fluor 594-phalloidin (magenta). Scale bars indicate 10 µm. In the cytoplasmic region, CDCP1 contains a Src association motif around Tyr734 and two palmitoylation sites (Cys residues 689 and 690), which are required for the localization of CDCP1 in lipid rafts. Upon phosphorylation at Tyr734 by Src, CDCP1 is additionally phosphorylated at Tyr762, resulting in direct association with PKCδ, which promotes cell migration. In the extracellular region, CDCP1 contains three CUB domains that are required for protein−protein interactions. CDCP1 also harbors a proteolytic cleavage (shedding) site between the first and second CUB domains and can, thus, be present in either the full-length or cleaved form depending on the cellular context. TM indicates transmembrane domain. (B) Schematic diagram of CRISPR/Cas9-based generation of CDCP1-knockout MDCK cells. Blue indicates the PAM sequence and bold letters indicate the start codon. The red arrowhead indicates the cleavage site. (C) Immunoblotting analysis of CDCP1-knockout MDCK cells. Lysates from wild-type and CDCP1-knockout MDCK cells were subjected to immunoblotting using the indicated antibodies. (D) CDCP1-knockout MDCK cysts were incubated in the presence of HGF (50 ng/ml) for 4 and 6 days. Ki67 was visualised with an Alexa Fluor 488-conjugated antibody (green) and actin filaments were stained with Alexa Fluor 594-phalloidin (magenta). Scale bars indicate 10 µm. presence of Dox (1 µg/ml) for the indicated time periods. Actin filaments were stained with Alexa Fluor 594-phalloidin (magenta). Scale bars indicate 50 µm. (C) CDCP1-EGFP-and mutant-harboring MDCK cells were incubated in the presence of Dox (1 µg/ml) for 48 h. DRM and non-DRM fractions were separated on a sucrose density gradient. Aliquots of the fractions were subjected to immunoblotting analysis using the indicated antibodies. (D) DRM fractions of CDCP1-myc-overexpressing MDCK cells were subjected to immunoprecipitation with an anti-Src or anti-myc-tag antibody. Immunoprecipitates were subjected to immunoblotting using the indicated antibodies (IgG HC, and IgG heavy chain). pretreated with 20 µM marimastat for 2 h, and then incubated with Dox (1 µg/ml) for 4 days. The basement membrane was visualized using an anti-laminin antibody (magenta). (C) The histogram depicting the percentage of cells with protrusions was calculated by setting the total number of cysts to 100% (n > 150). (D) TRE-CDCP1-mCherry-harbouring MDCK cells were embedded within a 2% DQ-collagencontaining matrix. CDCP1-mCherry cysts were pretreated with 20 µM marimastat for 2 h, and then incubated with Dox (1 µg/ml) for 4 days. DQ fluorescence (green) was visualized under a fluorescent microscope. Scale bars indicate 50 µm. (E, F) MDCK cysts embedded within the collagen matrix were pretreated with marimastat (+ Mari, 10 µM; ++ Mari, 20 µM) for 2 hours, and then incubated in the presence of HGF (50 ng/ml) for 1 day. (E) Diameter of cysts (µm) of cysts (n = 100). Dotted blue line indicates the average diameter of non-treated cysts. (F) Fraction of the total number of cysts counted (n > 100). The mean ratios ± SD were obtained from three independent experiments. *, P < 0.05; **, P < 0.01; **** P < 0.0001, Two-way ANOVA was calculated relative to the HGF-treated cysts. (G) STAT3-MER-overexpressing cysts embedded within the collagen matrix were incubated with 1 µM 4-OHT for 4 days. STAT3-MER was visualized with ERα antibody (green) and actin filaments were stained with Alexa Fluor 594-phalloidin (magenta). Scale bars indicate 50 µm.  and left kidney (E) weight of wild-type (+/+) and Cdcp1 homozygous knockout (-/-) mice at 8 weeks of age. The mean ratios ± SD were obtained from ten mice per group. NS, not significantly different; unpaired two-tailed t-test.