CUB domain-containing protein 1 controls HGF responses by integrating Src and Met–STAT3 signaling

: Hepatocyte growth factor (HGF) controls various biological responses, including morphogenesis, organ regeneration, and cancer invasion, by activating its receptor, Met. However, the mechanisms that precisely control diverse Met signaling remain unclear. Here, we identified CUB domain-containing protein 1 (CDCP1) as a crucial element of HGF signaling. In MDCK cysts, HGF induced Src activation via CDCP1 upregulation, and CDCP1 ablation abrogated HGF responses, i.e., extended invasive cell protrusions and promoted cell growth/proliferation. Mechanistically, CDCP1 upregulation promoted Src recruitment into lipid rafts to activate Met-associated STAT3. In breast cancer cells in which Met and CDCP1 were co-upregulated, CDCP1 knockdown suppressed HGF-induced cell invasion. Furthermore, in vivo analysis showed that CDCP1 ablation suppressed compensatory renal growth by attenuating Met – STAT3 signaling. These findings demonstrate that CDCP1 plays a crucial role in controlling HGF responses by integrating Src and Met-STAT3 signaling, and provide new insights into the regulatory mechanisms underlying the multifaceted functions of HGF.

The activation of HGF-Met signaling induces complex cellular responses that drive socalled "invasive growth," including morphological changes associated with the epithelialmesenchymal transition (EMT) and promotion of cell growth/proliferation (Gao & Vande Woude, 2005). The HGF-Met pathway is aberrantly upregulated in various cancers (Matsumoto, Umitsu, De Silva, Roy, & Bottaro, 2017), such as breast, esophageal, hepatocellular carcinoma, and non-small cell lung cancers. The invasive-growth phenotypes induced by HGF contribute to the progression of invasion and metastasis, as well as the growth/survival of cancer cells. Thus, HGF-Met signaling has been considered as a promising therapeutic target for a subset of cancers (Comoglio, Trusolino, & Boccaccio, 2018).
The Madin-Darby canine kidney (MDCK) cell line was derived from renal tubule epithelial cells and is a physiologically relevant in vitro model for studying the regulation of HGF functions (O'Brien, Zegers, & Mostov, 2002;Zegers, 2014). When grown in threedimensional 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, Matsumoto, et al., 1991;Montesano, Schaller, & Orci, 1991). During this morphogenesis, MDCK cells lose their epithelial polarity via a partial EMT-like phenotypic change and protrude into the extracellular matrix (ECM) by degrading the basement membrane (O'Brien et al., 2004;Pollack, Runyan, & Mostov, 1998). In addition, HGF promotes cell growth and proliferation, resulting in the formation of multi-cell layered cysts. These observations indicate that HGF can induce invasive growth in MDCK cysts. However, how multiple signaling pathways downstream of Met are controled to properly execute HGF-induced invasive growth responses has not been addressed.
To address the above issues, we tested the hypothesis that there an additional factor(s) that selectively links and/or controls critical pathways required for HGF-induced invasive growth responses. To this end, we re-dissected the HGF-Met signaling pathway using threedimensional cultures of MDCK cysts and identified Src and its membrane-scaffold protein, CUB domain-containing protein 1 (CDCP1), as critical elements of HGF-Met signaling. We then investigated the mechanisms underlying the functions of CDCP1 in MDCK cysts, and confirmed the role of CDCP1 in HGF-induced cancer migration and invasion, using breast cancer MDA-MB231 cells with increased Met and CDCP1 expression. Finally, we verified in vivo roles of CDCP1 in compensatory renal growth using Cdcp1-knockout mice. The results show that CDCP1 plays a crucial role in controlling the HGF-induced invasive growth responses by focally and temporally integrating Src and Met-STAT3 signaling in 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 diameters) and morphological changes with some cysts exhibiting extended cell protrusions (Fig. 1, A-C). To dissect the HGF-induced 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, Matsumoto, Shimura, Tanaka, & Nakamura, 1997), robustly inhibited cell proliferation and morphological changes, confirming that the observed phenomena were dependent on HGF. Treatment with the selective mTOR inhibitors, Torin1 and rapamycin, significantly suppressed cell growth/proliferation but was less effective against the formation of cell protrusions (Fig. 1, A-C). In contrast, treatment with dasatinib (a Src kinase inhibitor) potently suppressed cell proliferation and protrusion extension (Fig. 1, A-C). Furthermore, immunofluorescence analysis revealed that activated Src (pY418) was concentrated at the tip of protruding cells (Fig. 1D), which was visualized with mCherry-GPI, a marker protein of lipid rafts (Fig. 1E). These observations imply that HGF signaling is associated with Src activation in lipid rafts on protruding cells.
To assess the role of Src in HGF signaling, we examined the effects of Src activation on MDCK cysts by expressing a Src protein fused to a modified estrogen receptor (Src-MER) (Kuroiwa, Oneyama, Nada, & Okada, 2011) that could be activated by treatment with hydroxytamoxifen (4-OHT; Fig. S1A). Src activation induced both the formation of multiple cell protrusions and cell proliferation (Fig. S1B). Detergent-resistant membrane (DRM)separation analysis revealed that activated Src-MER accumulated in lipid raft fractions (Fig.  S1C). We thus searched for scaffolding proteins that accommodate and activate Src in these fractions. Src-MER-interacting proteins were isolated from DRM fractions by coimmunoprecipitation and were identified by mass spectrometry (Fig. S1, 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 scaffold of Src (Alvares, Dunn, Brown, Wayner, & Carter, 2008;Wortmann, He, Deryugina, Quigley, & Hooper, 2009) (Fig. S2A). Immunofluorescence analysis revealed that endogenous CDCP1 was transiently concentrated at the tips of protruding cells during 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 invasive growth phenotypes were efficiently abrogated by Cdcp1-knockout in MDCK cells (Fig. 1, G-I and Fig. S2, B-D). These results suggest that the CDCP1 is a crucial component of HGF signaling.

CDCP1 upregulation phenocopies HGF-induced invasive growth by activating Src in lipid rafts
To elucidate the mechanisms whereby CDCP1 regulates HGF signaling, we generated MDCK cells with doxycycline (Dox)-inducible in CDCP1 expression. Dox treatment induced the expression of full-length and cleaved CDCP1 at the plasma membrane (Fig. S3, A and B). DRM-separation analysis showed that a fraction of CDCP1 was distributed in lipid raft fractions (Fig. S3C). Moreover, the induction of CDCP1 expression dramatically activated Src (pY418/Src) (Fig. S3A). Furthermore, activated Src was concentrated in the lipid raft fractions and was associated with CDCP1 ( Fig. S3, C and D). In contrast, a CDCP1 mutant that lacked the Src-binding site (Y734F; CDCP1-YF) failed to activate Src (Fig. S3A). A different CDCP1 mutant lacking the lipid raft localization signal (residues C689G and C690G; CDCP1-CG) could activate Src, but only in non-lipid raft fractions (Fig. S3, A and C). These findings indicate that upregulated CDCP1 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 S1). In the early stages (12−24 h after induction), cells expressing CDCP1 gradually protruded toward the ECM. During the later stages (48 h post-induction), multiple cell protrusions formed randomly and extended aggressively toward the ECM, with cysts forming multiple layers (Fig. 2B). Concurrent with these phenomena, CDCP1 upregulation activated Src in the protruding cells (Fig. 2C) and Src inhibition by dasatinib strongly suppressed CDCP1-induced cell proliferation and protrusion extension ( Fig. 2D and Table S1). In addition, 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 phenocopy HGF-induced invasive growth phenotypes via Src activation in lipid rafts.

CDCP1 activates the STAT3 pathway via lipid rafts
To dissect the signaling pathways operating downstream of the CDCP1-Src axis, we performed DNA-microarray analyses of MDCK cysts expressing either wild-type CDCP1 or the CDCP1-CG mutant. Gene ontology analysis implied that CDCP1 expression participated in regulating signal transduction (Fig. S4A) and Ingenuity Pathway Analysis revealed that the STAT3 pathway was more prominently activated by CDCP1 expression than by CDCP1-CG expression (Fig. S4, B and C, Data file S1). Indeed, the level of phosphorylated STAT3 (pY705) increased significantly in cysts expressing wild-type CDCP1 (Fig. 3A). STAT3 is a author/funder. All rights reserved. No reuse allowed without permission.
In fact, the expression levels of several MMP-encoding genes were significantly increased by CDCP1 upregulation (Fig. 3B). In addition, immunofluorescent staining of laminin revealed that CDCP1 expression disrupted the basement membrane, enabling the extension of cell protrusions (Fig. S5A). This effect was suppressed by treatment with marimastat, a pan-MMP inhibitor (Fig. S5, B and C). Similar suppression of cell protrusion extension by marimastat was also observed in HGF-stimulated MDCK cysts (Fig. S5, D and E). Analysis with DQ-collagen, a fluorescent indicator of collagen degradation, also showed that CDCP1 expression induced ECM degradation (Fig. S5F). These observations suggest that the HGF-induced formation of cell protrusions is associated with ECM rearrangements via STAT3-mediated upregulation of MMPs. Mitogenic genes such as Myc (Myc) and Ccnd1 (Cyclin D1) were also upregulated in cysts expressing CDCP1 (Fig. 3B), indicating that CDCP1-induced cell proliferation is attributable to STAT3 activation.
We further confirmed that STAT3 contributes to HGF signaling 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. S5G). In contrast, treatment of MDCK cysts with STAT3-specific inhibitors suppressed CDCP1-induced cell growth ( Fig.  3C and Table S1), and overexpression of dominant negative STAT3 (STAT3-Y705F) also inhibited CDCP1-induced cellular events (Fig. 3D). Furthermore, STAT3 inhibitors efficiently suppressed cell proliferation (Fig. 3E) and protrusion formation (Fig. 3F) in HGFstimulated MDCK cysts, underscoring the crucial role of STAT3 activation in HGF-induced invasive growth.

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 screen assay showed that various Met-specific inhibitors potently suppressed CDCP1-induced invasive growth ( Fig. 4A and Table S1), suggesting that Met activity facilitated the function of CDCP1. Because activated Met associates with STAT3, we predicted a close association between the CDCP1-Src and Met-STAT3 axes. 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 removing the first CUB domain of CDCP1, but was diminished by deleting 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, Harrington, & Hooper, 2016), it is likely that activated CDCP1 functionally interacts with Met. To test 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 (K365A, R368A, and K369A; Fig. 4D). CDCP1-PR expression induced Src activation to a level similar to that observed with wild-type CDCP1, but failed to activate STAT3 (Fig. 4D). Consistent with these biochemical effects, CDCP1-PR failed to induce the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/789339 doi: bioRxiv preprint formation of cell protrusions (Fig. 4, E and F). These data suggest that an association between activated CDCP1 and Met is required for Src-mediated STAT3 activation and subsequent cellular events (Fig. 4G). In addition, the initiation of cell protrusion induced by HGF treatment or Met overexpression was efficiently suppressed by the loss of CDCP1 (Fig.  S6A). These lines of in vitro evidence suggest a model wherein CDCP1-Src upregulation in lipid rafts is required for Met-mediated STAT3 activation during HGF-induced invasive growth in MDCK cysts (Fig. S6B).

CDCP1 is required for HGF-induced cancer invasion
To further explore the role of the CDCP1-Met interaction in other HGF-dependent cellular responses, we examined whether CDCP1 was involved in HGF-induced invasion of cancer cells. CDCP1 is upregulated in various cancers (He et al., 2016), including malignant triple-negative breast cancers (Turdo et al., 2016). Indeed, highly invasive triple-negative breast cancer MDA-MB231 cells expresses CDCP1 at high levels ( Fig. 5A), while noninvasive luminal A-type breast cancer MCF7 and T47D cells do not express CDCP1. Notably, Met is also upregulated in MDA-MB231 cells, but not in MCF7 or T47D cells (Fig. 5A), and CDCP1 knockdown decreased Met protein expression (Fig. 5B). These findings suggest that CDCP1 functionally interacts with Met, even in cancer cells. Boyden chamber assays revealed that CDCP1 knockdown significantly inhibited HGF-induced cell migration (Fig. 5C) and invasion (Fig. 5D). The inhibition of invasive activity was rescued by re-expression of wild-type CDCP1, but not by the CDCP1-CG or CDCP1-PR mutant. Furthermore, HGFinduced formation of lamellipodia, a membrane structure required for cell migration, was also inhibited by CDCP1 knockdown and rescued only by wild-type CDCP1 (Fig. 5, E and F). These observations demonstrate that CDCP1-Src in lipid rafts plays crucial roles in HGFinduced dynamic cell migration and in the invasion of cancer cells.

CDCP1 contributes to induction of compensatory renal growth
Finally, to verify the aforementioned model in vivo, we investigated the role of the CDCP1-mediated Met-STAT3 pathway in compensatory renal growth. For this study, we generated Cdcp1-knockout mice using the CRISPR-Cas9 system (Fig. S7, A-C). The Cdcp1knockout (Cdcp1 −/− ) mice grew normally and did not show any overt phenotype under normal conditions (Fig. S7, C-E), as reported previously (Spassov, Wong, Wong, Reiter, & Moasser, 2013). Eight weeks after UNX, the remaining kidney was enlarged in wild-type and heterozygous (Cdcp1 +/− ) mice (Fig. 6A). Kidney/body weight ratios in wild-type and Cdcp1 +/− mice were elevated to approximately 136% and 135% compared to those in shamoperated mice, respectively (Fig. 6B). In contrast, the kidney/body weight ratio elevation was significantly smaller in Cdcp1 −/− mice (approximately 121%) (Fig. 6, A and B). Because 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 fluoresceinlabeled specific lectin (LTL-FITC). Wild-type mice exhibited prominent thickening of LTLpositive proximal renal tubules (to approximately 148%), whereas this thickening was suppressed in Cdcp1 −/− mice (approximately 117%) (Fig. 6, C and D). These data confirm that compensatory growth of proximal renal tubules was suppressed in Cdcp1 −/− mice.

CDCP1 induces compensatory renal growth by activating Met-STAT3 signaling
author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/789339 doi: bioRxiv preprint To address the cause of defective renal growth in Cdcp1 −/− mice, we analyzed the remaining kidneys 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. 7A), 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, while the activation of both signals was appreciably attenuated in Cdcp1 −/− renal tubules (Fig.  7, B and C). Activation of CDCP1 (pY734) also occurred in renal tubules in a manner similar to that of Met and STAT3 (Fig. S8A). Notably, activated Met, STAT3, and CDCP1-Src were co-localized in a subset of intracellular small vesicles that were positive for EEA1 (Fig. S8, B-D), a marker of early endosomes, supporting the functional 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 gradually increased and peaked 2 days after UNX (Fig. 7,D and E). However, in Cdcp1 −/− mice, the appearance of Ki67-positive cells was delayed, and the overall number of proliferating cells 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.
We further examined the effects of CDCP1 loss on the surrounding environment of proximal tubules after UNX. Before UNX, collagen IV was observed around the proximal tubules in both wild-type and Cdcp1 −/− mice (Fig. 7F). In wild-type mice, collagen IV gradually degraded 24 h after UNX, but this degradation was attenuated in Cdcp1 −/− mice. Consistently, MMP2 and MMP9 were upregulated and occasionally concentrated in vesiclelike structures in proximal tubules 24 h after UNX ( Fig. 7G and Fig. S8E). However, these phenomena were barely detectable in Cdcp1 −/− proximal tubules. These observations suggest that secretory MMPs remodel extracellular matrix for progression of compensatory renal growth downstream of CDCP1-Src-mediated Met-STAT3 signaling.

Discussion
To address the regulatory mechanisms underlying the multifaceted functions of HGF, i.e., induction of invasive growth phenotypes, we dissected the mechanism of HGF signaling using MDCK cysts and identified Src and its scaffold CDCP1 as critical elements of HGF-Met signaling. Upon stimulation with HGF, CDCP1 concentrated temporally at the tip of protruding cells. Considering that CDCP1 expression is induced by HGF (Gusenbauer, Vlaicu, & Ullrich, 2013) or by epidermal growth factor (EGF)-mediated activation of the MAPK pathway (Adams et al., 2015;Dong et al., 2012), the observed upregulation of CDCP1 might be caused by activation of the HGF-Met-MAPK pathway. We also showed that continuous CDCP1 upregulation induced the formation of multiple cell protrusions and cell growth/proliferation. However, 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. S9A). This finding suggests that the temporal CDCP1 upregulation by growth factors such as HGF might trigger the initial phase of morphogenesis, specifically cell protrusion extension and growth, and that the coordinated author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/789339 doi: bioRxiv preprint negative feedback regulation of CDCP1 might be involved in establishing the compound tubular/acinar epithelial system.
In this study, we demonstrated that activation of the CDCP1-Src complex in lipid rafts was required to induce HGF responses in MDCK cysts. Src activation by CDCP1 in lipid rafts induced STAT3 phosphorylation at Tyr705 (Bromberg et al., 1998;C. L. Yu et al., 1995), leading to 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 STAT3 activation via CDCP1-Src in lipid rafts is crucial for inducing invasive growth phenotype. We also found that a physical association between cleaved/activated CDCP1 and Met was 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 enable efficient STAT3 phosphorylation by Src (Fig. 4G). Hence, this focal integration of the CDCP1-Src axis with the Met-STAT3 complex might be a key event in controlling HGF responses. However, 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, these findings suggest that CDCP1 can also serve as a general regulator of membrane receptor signaling, which involves Src activation. CDCP1 upregulation has been implicated in tumor progression (He et al., 2016;Ikeda et al., 2009;Miyazawa et al., 2010), particularly cancer invasion and metastasis (He et al., 2016;Liu et al., 2011;Uekita & Sakai, 2011;Wright et al., 2016). In this study, we showed that CDCP1 and Met were co-upregulated in highly invasive breast cancer MDA-MB231 cells, but not in non-invasive MCF7 or T47D cells, suggesting that a functional interaction occurred between them. Indeed, CDCP1 knockdown downregulated Met and inhibited HGFdependent invasive activity. Because Src binding sites in CDCP1 and lipid rafts were requited for the promotion of HGF-dependent invasive activity, CDCP1-mediated Src activation in lipid rafts likely contributes to cancer invasion in a manner similar to that in MDCK cysts. Interestingly, we also found that CDCP1 upregulation in MDCK cysts induced the expression of cytokeratin 14 (Fig. S9B), a marker of collective cancer invasion (Cheung, Gabrielson, Werb, & Ewald, 2013). This finding suggests a potential role for the CDCP1-Src axis, even in collective cancer invasion. In support of these observations, The Cancer Genome Atlas (TCGA) database analysis revealed a significant correlation between CDCP1 high /MET high groups and poor prognoses in patients with breast or kidney cancer (Fig.  S10, A and B). Therefore, further analyses of the functions of the CDCP1-Src-Met-STAT3 pathway in a wide range of cancers could reveal new therapeutic targets for treating some malignant cancers.
HGF is immediately upregulated after the loss of kidney mass and plays important roles in compensatory growth by inducing cell proliferation and anti-apoptotic effects (Matsumoto & Nakamura, 2001). In this study, we uncovered the contribution of CDCP1-Src to HGFinduced compensatory renal growth using Cdcp1-deficient mice. We showed that activation of Met and STAT3 in the renal tubule cells following UNX was significantly attenuated in Cdcp1-deficient mice. Furthermore, MMP2/9 upregulation and basement membrane collagen IV degradation (Tan & Liu, 2012) were detected in the renal tubule cells after UNX (Fig.  S8F). These observations are consistent with those of our in vitro study in MDCK cysts, author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/789339 doi: bioRxiv preprint supporting the crucial role of the functional integration of CDCP1-Src with Met-STAT3 pathways during compensatory renal growth. Given that CDCP1 expression 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 acutely amplify the HGF signaling required for regenerative organ growth in critical situations. Furthermore, because STAT3 is an important regulator of the regeneration of other organs such as the liver, muscles, and intestines (Lindemans et al., 2015;Taub, 2004;Tierney et al., 2014), it is possible that the temporal upregulation of CDCP1 widely contributes to the promotion of organ regeneration by inducing focal activation of the Src-STAT3 pathway.
In conclusion, we showed that CDCP1 served as a critical regulator of HGF-induced invasive growth phenotypes by focally integrating Src and the Met-STAT3 pathway (Fig.  7H). 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 regenerative growth and cancer invasion, and might contribute to the development of more promising therapies for malignant cancers that are associated with the co-upregulation of CDCP1 and Met.

Materials and Methods
Cell culture MDCK (type-I), HEK293T, MDA-MB231, MCF7, and T47D cells were cultured in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum (FBS) at 37 °C in a 5% CO 2 atmosphere. Three-dimensional culture was performed using type-I collagen (Cellmatrix Type I-A, Nitta Gelatin) according to the manufacturer's protocol. Type I-A collagen (3 mg/ml) was neutralized 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 collagen gel) were combined with collagen in DMEM supplemented with 5% FBS, and then polymerized at 37°C in a 5% CO 2 atmosphere. The medium was replaced every 2 days. A collagen gel containing 2% type-I DQ-collagen (Molecular Probes) was used for the DQ-collagen degradation assay.

Mice
Cdcp1-knockout mice were generated in a C57BL/6N background using the CRISPR/Cas9 system. Animals were housed in environmentally controlled rooms at the animal experimentation facility of Osaka University. All animal experiments were conducted 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 guide RNA (gRNA) and primer sequencess used for genotyping are shown in Supplementary Table 3 (see also Supplementary Fig. 9a).

Microarray 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 (GO) 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 expression was normalized to expression of the reference gene, GAPDH. The sequences of the primers used for this analysis are listed in Table S2.

Immunofluoresce microscopy
author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/789339 doi: bioRxiv preprint For two-dimensional culture, cells were grown on coverslips coated with type-I collagen, fixed with 4% paraformaldehyde (PFA) and permeabilized with phosphate-buffered saline (PBS) containing 0.03% Triton X-100. For three-dimensional culture, cysts embedded within the collagen matrix were fixed with 4% PFA and permeabilized with PBS containing 0.5% Triton X-100. Permeabilized cells and cysts were blocked with 1% bovine serum albumin (BSA) and incubated with primary antibodies, and then incubated with Alexa Fluor 488/594phalloidin (Molecular Probes). For the kidney sections, kidneys were prepared by perfusion fixation with 4% PFA 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 4ʹ′,6-diamidino-2-phenylindole (DAPI). Immunostained objects were observed under an FV1000 confocal microscope (Olympus). For time-lapse observations 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
MDCK cells were co-transfected with the pSilencer1.0-U6 plasmid containing target-gRNA, and the pMJ920 plasmid (Addgene) that co-expresses Cas9 and EGFP using the MDCK Cell Avalanche Transfection Reagent (EZ Bioscience) per the manufacturer's protocol. Three days after transfection, single EGFP-positive cells were isolated with a FACSAria III flow cytometer (BD Biosciences). Knockout of the Canis CDCP1 gene was confirmed by immunoblotting. The gRNA and primer sequences used for genotyping are shown in Table S3 (see also Fig. S2B).

Migration and invasion assay
BioCoat cell culture permeable supports and Matrigel Invasion Chambers (Corning) were used for the migration and invasion assays, respectively. Cells (0.5 × 10 5 for migration assays and 1 × 10 5 for invasion assays) were seeded on inserts and transferred to chambers author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/789339 doi: bioRxiv preprint containing culture media and 0.1% FBS, with or without 100 ng/ml HGF. After incubation at 37°C for 24 h, migrated or invaded cells were fixed with 4% PFA and then stained with 1% crystal violet. Invaded cells were counted on micrographs; in each experiment, cells were counted on five randomly chosen fields. The migration and invasion assays were repeated at least three times.

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 shamnephrectomized mice were used as controls for UNX mice.

Kaplan-Meier survival analysis
Clinical and RNA-Seq data from the publicly available TCGA dataset from patients with breast cancers and kidney clear cell carcinoma were used for the analysis of CDCP1 and Met expression in tumor samples. Survival curves were estimated using the Kaplan-Meier method, and the obtained survival curves were compared by using the log-rank test.

Statistics and reproducibility
For data analyses, unpaired two-tailed t-tests were performed to determine the P-values. For multiple group comparisons, two-way analysis of variance (ANOVA) was used. A Pvalue of <0.05 was considered to reflect a statistically significant difference. All data and statistics were derived from at least three independent experiments.

Data availability
Supporting microarray data have been deposited in Gene Expression Omnibus under accession code GSE99375. All other supporting data are available from the corresponding author upon request. The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/789339 doi: bioRxiv preprint and actin filaments were stained with Alexa Fluor 594-phalloidin (magenta). (G) Wild-type and CDCP1knockout MDCK cysts were incubated in the presence of HGF for 1 day. Ki67 was visualized with an Alexa Fluor 594-conjugated antibody (magenta), and actin filaments were stained with Alexa Fluor 488-phalloidin. The arrowheads indicate transiently formed protrusions. The scale bars indicate 50 µm. (H) Diameter (µm) of cysts (n = 100). (I) Fraction of the total number of cysts counted (n > 100) with protrusions. The scale bars indicate 10 µm. The mean ratios ± SD were obtained from three independent experiments. **, P < 0.01; ***, P < 0.001; ****, P < 0.0001; NS, not significantly different; ANOVA was calculated compared to HGF-treated cysts.
author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/789339 doi: bioRxiv preprint

Fig. 2. CDCP1 upregulation phenocopies HGF-induced invasive growth by activating Src in lipid rafts.
(A) Schematic representation of the procedure used to analyze TRE-CDCP1-EGFP-harboring MDCK cysts (CDCP1-EGFP cysts). TRE-CDCP1-EGFP-harboring cells were cultured within a collagen matrix for 5 days to allow time for cyst formation, and then CDCP1-EGFP cysts were incubated in the presence of Dox for ~4 days. (B) CDCP1-EGFP cysts embedded within the collagen matrix were incubated in the presence of Dox (1 µg/ml) for the indicated time periods. Actin filaments were stained with Alexa Fluor 594-phalloidin (magenta). Filled arrowheads indicate protruding cells and open arrowheads indicate multi-layered structures. (C) CDCP1-EGFP cysts were incubated in the presence of Dox for 2 days. Activated Src was visualized using an anti-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 for 4 days. (E) CDCP1-YF-EGFP and CDCP1-CG-EGFP cysts embedded within the collagen matrix were incubated in the presence of Dox for 4 days. Actin filaments were stained with Alexa Fluor 594-phalloidin (magenta). The scale bars indicate 50 µm. (F) Fraction of the total number of cysts counted (n > 150) with protrusions. The mean ratios ± SDs were determined from three independent experiments. ***, P < 0.001; ****, P < 0.0001; NS, not significantly different; ANOVA compared to the Dox-treated cysts.
author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/789339 doi: bioRxiv preprint 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 collagen matrix and cultured for 9 days. Cyst lysates were subjected to immunoblotting using the indicated antibodies. (E) CDCP1-PR-EGFP cysts embedded within the collagen matrix were incubated with Dox for 4 days. Actin filaments were stained with Alexa Fluor 594-phalloidin (magenta). The scale bars indicate 50 µm. (F) Fraction of the total number of cysts counted (n > 150) with protrusions. The mean ratios ± SDs 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. author/funder. All rights reserved. No reuse allowed without permission.

Fig. 5. CDCP1 is required for HGF-dependent cancer invasion
(A) Total lysates from MCF7, T47D, and MDA-MB231 cells were subjected to immunoblotting using the indicated antibodies. (B) MDA-MB231 cells treated with or without HGF (100 ng/ml) were transfected with the indicated CDCP1 siRNAs, and total cell lysates were subjected to immunoblotting using the indicated antibodies. (C) The in vitro migration activities of MDA-MB231 cells treated with control and siRNAs were examined by performing transwell assays in the presence or absence of HGF. (D) The in vitro invasion activity of control-and siRNA-treated MDA-MB231 cells was examined by performing Matrigel transwell assays in the presence or absence of HGF. Rescue experiments were also performed by re-expressing wild-type CDCP1-myc, CDCP1-CG-myc, and CDCP1-PR-myc. (E, F) Formation of lamellipodia in control-and siRNA-treated MDA-MB231 cells was examined by immunofluorescent analysis for F-actin in the presence or absence of HGF (left panels). Yellow arrowheads indicate lamellipodia. Rescue experiments were also performed by re-expressing wild-type CDCP1-myc, CDCP1-CG-myc, and CDCP1-PR-myc (upper right panels). The ratios of the length of lamellipodial membrane to the total cell perimeter were calculated from at least 30 cells of each cell type (lower right graph). The scale bars indicate 50 µm. The mean ratios ± SDs were obtained from three independent experiments. *, P < 0.05; **, P < 0.001; ***, P < 0.001; ****, P < 0.0001; NS, not significantly different; twoway ANOVA. author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/789339 doi: bioRxiv preprint Fig. 6. 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 a sham operation. Regenerative growth of the remaining left kidney was analyzed 8 weeks after operation and increases in remaining kidney/body weight ratios were assessed. (B) The mean ratios ± SDs were obtained from 10 mice per group. (C) The remaining kidney was removed from UNX-or sham-operated mice, and proximal tubules were stained with FITC-LTL (green). The scale bars indicate 50 µm. (D) The proximal tubule thickness was determined by the length of FITC-LTL-stained epithelial cell layer (n = 100). The mean ratios ± SDs were obtained from 5 mice per group. *, P < 0.05; **, P < 0.01; ****, P < 0.0001; NS, not significantly different; two-way ANOVA.
author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/789339 doi: bioRxiv preprint Fig. 7. Met-STAT3 signaling is attenuated in Cdcp1-knockout mice. (A) Cdcp1 wild-type (+/+) and homozygous knockout (−/−) mice at 8 weeks of age were subjected to left UNX or a sham operation. Compensatory renal growth of the remaining left kidney was analyzed at the indicated time points after operation and increases in remaining kidney/body weight ratios were assessed. The mean ratios ± SDs were obtained from at least 6 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), MMP9 (G), and Alexa Fluor 594-conjugated secondary antibody (magenta). Proximal tubules were visualized by staining with FITC-LTL (green). The 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 3 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 regulated Met-STAT3 signaling leading to compensatory renal growth through the induction of ECM rearrangement and cell growth/proliferation. author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/789339 doi: bioRxiv preprint Fig. S1. Src associates with CDCP1 in lipid raft fractions. (A) Src-MER-overexpressing MDCK cells were incubated with 10 µM 4-OHT for the indicated time periods. Cell lysates were subjected to immunoblotting using the indicated antibodies. (B) Src-MER-overexpressing cysts embedded within the collagen matrix were incubated with 10 µM 4-OHT for 2 days. Activated Src was visualized with an anti-Src pY418 antibody (green). Actin filaments were stained with Alexa Fluor 594-phalloidin (magenta). The scale bars indicate 50 µm. (C) Src-MER-overexpressing MDCK cells were incubated with 10 µM 4-OHT for the indicated time periods. DRM and non-DRM fractions were separated in a sucrose-density gradient. Aliquots of the fractions were analyzed by immunoblotting with an anti-Src pY418 antibody. (D) DRM fractions from Src-MER-activated cells were subjected to immunoprecipitation with an anti-estrogen receptor (ERα) antibody. Immunoprecipitates were confirmed by silver staining. (E) Schematic flow diagram of process used to identify Src-associated scaffolding proteins. All proteins in a polyacrylamide gel were trypsinized and subjected to mass spectrometry. Proteins were identified using the SwissProt database, and the number of identified proteins is depicted. author/funder. All rights reserved. No reuse allowed without permission.

Supplementary Figures
The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/789339 doi: bioRxiv preprint 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 CDCP1 localization in lipid rafts. Upon Tyr734 phosphorylation 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 the transmembrane domain. (B) Schematic diagram of CRISPR/Cas9-based generation of CDCP1-knockout MDCK cells. The blue text indicates the PAM sequence and the bold letters indicate the start codon. The red arrowhead indicates the cleavage site. (C) Immunoblotting analysis of CDCP1knockout 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 or 6 days. Ki67 was visualized with an Alexa Fluor 488-conjugated antibody (green) and actin filaments were stained with Alexa Fluor 594-phalloidin (magenta). The scale bars indicate 10 µm.
author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/789339 doi: bioRxiv preprint author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/789339 doi: bioRxiv preprint  Supplementary Table 2. Red and green symbols indicate transcript levels that were upregulated and downregulated by CDCP1 overexpression, respectively. Arrows are colored to indicate activation (orange), inhibition (blue), inconsistency with the state of the downstream molecule (yellow), and unknown effects (grey). (C) Heat map representing changes in the expression of STAT3 target genes in CDCP1-EGFP (WT) and CDCP1-CG-EGFP (CG) cysts. author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/789339 doi: bioRxiv preprint author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/789339 doi: bioRxiv preprint The mean ratios ± SDs were obtained from ten mice per group. NS, not significantly different; unpaired twotailed t-test author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/789339 doi: bioRxiv preprint author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/789339 doi: bioRxiv preprint Fig. S10. Co-upregulation of CDCP1 and Met correlates with poor prognosis in patients with breast and kidney cancer. (A, B) Correlations of CDCP1-and Met-expression levels with the prognoses of patients with breast cancer (A) or kidney clear cell carcinoma (B) were estimated using the Kaplan-Meier method, based on the transcriptome dataset from the TCGA project. Statistical significance was calculated by performing a logrank test. author/funder. All rights reserved. No reuse allowed without permission.
Supplementary Data file S1. Upstream regulator analysis using Ingenuity Pathway Analysis software.
author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/789339 doi: bioRxiv preprint