|Home | About | Journals | Submit | Contact Us | Français|
Hepatocyte growth factor (HGF) plays central roles in tubulogenesis and metastasis[1-4]. HGF treatment of Madin-Darby canine kidney (MDCK) cells grown as cysts in three-dimensional culture induces tubulogenesis[5, 6], which like most tubulogenic processes, proceeds through distinct intermediate phases. Identification of genes associated with these phases is central to understanding the molecular mechanisms of tubulogensis; however due to inefficient, asynchronous tubule formation, isolating such genes has been unfeasible. Here we developed a synchronous, efficient tubulogenesis system and used time-course transcriptional profiling to identify genes temporally regulated in developmental intermediates. Knockdown (KD) of tensin 4 (TNS4), a particularly highly upregulated gene, leads to a decrease in formation of extensions and tubules, two sequential intermediates in tubulogensis. Exogenous expression of TNS4 marks invasive cells in an epithelial sheet. A mutation in the SH2 domain of TNS4 prevents the transition from extension formation to invasive migration during tubule formation and leads to increased basal activation of STAT3. Exogenous expression of a constitutively active STAT3 mimics the defect by the mutation. Our study highlights the role of TNS4-STAT3 axis in epithelial sheet invasion and tubulogenesis.
Through screening of 168 culture conditions (not shown), we modified the tubulogenesis protocol of Montesano . To induce tubules from an epithelial sheet, a Madin-Darby canine kidney (MDCK) cell monolayer is grown on the top of a collagen matrix together with MRC-5 fibroblasts embedded in collagen. A filter separates the two cell types. We refer to this protocol as 2.5D tubulogenesis (2.5D) because MDCK cells are grown on the top of an extracellular matrix (ECM) gel, rather than embedded as in conventional three-dimensional (3D) culture (Fig. S1A). HGF secreted from MRC5 induces tubulogensis in 3D. Indeed, when we added neutralizing anti-HGF antibodies to 2.5D, tubulogenesis was abrogated as reported in 3D (Fig. S1B). However, in contrast to 3D, HGF alone was unable to induce tubulogenesis in 2.5D (Fig. S1B). Thus, HGF is necessary, but insufficient for tubulogenesis in 2.5D, at least in the assay conditions used.
Using time course image analysis, we defined four morphological intermediates during 2.5D: 1) multicellular apical protrusion, 2) extension, 3) bud/tubule initiation, and 4) tubule elongation (Fig. 1). 1). After co-culture, clusters of ≥2 MDCK cells elongate and their apical membranes protrude toward the luminal side. We term this Multicellular Apical Protrusion (MAP) (Fig. 1B and Fig. 2B). 2). While cells remain in the monolayer, many cells form F-actin-rich extensions, projecting from their basal surface into ECM (Fig. 1C). A MAP is often accompanied by extension (Fig. S2A). 3). Groups of cells begin to invade the matrix as tubular structures (Fig. 1D). 4). Tubules elongate further, with well-polarized cells lining a single lumen, as indicated by apical membrane-facing phalloidin staining (Fig. 1E and Movie S1). The leading edges of growing tubules often consist of multiple cells with long protrusions (Fig. S2B). Unlike 3D, we did not detect chain migration. Instead, invading cells migrate collectively and maintain a continuous lumen during migration, similar to vascular sprouting, in vitro mammary gland branching, and in vivo zebrafish kidney morphogenesis[10-12]. However, unlike the mammary gland branching model, leading cells in 2.5D have extensions (Fig. S2B, asterisk). We occasionally observed cell clusters detached from tubules and scattered individual cells free in the ECM.
Importantly, 2.5D displays highly synchronized developmental intermediates (Fig. S3A). Extensions appear at ~72-96h after co-culture, while tubule initiation begins after extensions retract. These intermediates are temporally well separated. This synchronicity facilitated time-course microarrays to identify genes whose temporal regulation is associated with extensions and/or tubules. Four biological replicates were used for each stage (monolayer, extension, tubule), yielding 12 microarrays. After normalization and processing of datasets, two comparisons (monolayer vs. extension and monolayer vs. tubule) were made. Differentially expressed genes with P values <0.01 and fold changes >3.5 in at least one of the comparisons were considered. The resultant 113 genes were clustered on the basis of expression profiles using k-means algorithms (Fig. S3B). Of 113 genes, 48 showed greater expression changes in extension rather than tubule (Clusters A/C, Fig. S3B).
Among extension stage up-regulated genes, we focused on TNS4, (aka cten), a protein involved in cell migration and metastasis[14, 15]. TNS4 binds to integrin β1, which is required for branching morphogenesis in various systems including MDCK and developing kidney[16, 17]. However, it is unknown if TNS4 is involved in tubulogenesis and how TNS4 contributes to invasive growth. Quantitative PCR (qRT-PCR) confirmed that TNS4 is induced at the extension stage compared to monolayer (Fig. S3C).
To test if TNS4 is required for tubulogenesis, we blocked induction of TNS4 by RNAi. Five shRNAs were constructed in pLKO.1-puro and transfected into MDCK expressing EGFP-tagged TNS4 (GFP-TNS4). Immunoblotting confirmed three effective shRNAs (Fig. S4A). These were nucleofected into MDCK cells, which were then selected with puromycin. Pooled cells exhibited KD of endogenous transcript of TNS4 by 78±8.7%, as determined by qRT-PCR. KD of TNS4 (TNS4-KD) prominently decreased tubule number, compared to control cells expressing an irrelevant shRNA (Fig. S4B). We also infected MDCK cells with the effective shRNA-expressing lentiviruses individually, which produced similar tubulogenesis defects (not shown).
To validate that depletion of endogenous TNS4 causes reduced tubule formation, we introduced silent mutations into TNS4 to generate an RNAi resistant TNS4 (RNAiR-TNS4) and examined whether expression of RNAiR-TNS4 can rescue the tubule formation defect in TNS4-KD cells. Expressed RNAiR-TNS4 was immune to shRNA targeting TNS4 (Fig. S4C). Exogenously expressed RNAiR-TNS4 produced a 2.5±0.32-fold increase in total transcript level of TNS4, which is near the transcript level of endogenous TNS4 induced in the extension stage. Tubulogenesis in TNS4-KD cells was increased by expression of RNAiR-TNS4, but not vector alone (Fig. S4D). Thus, RNAiR-TNS4 can rescue the tubulogenesis defect produced by KD of endogenous TNS4.
To understand how TNS4 KD leads to decreased tubule formation, we created control cells marked with mCherry and TNS4-KD cells marked with EGFP, and monitored morphological changes in a mosaic monolayer. Both cells polarized as a monolayer without discernible defect (Fig. 2A). Compared to control cells, more TNS4-KD cells showed MAP, but this was not significant (Fig. 2B).
Next, the contribution of TNS4 to extension formation was scored using the ratio of the number of TNS4-KD cells projecting extensions to total number of TNS4-KD cells; this value was normalized to that of the control cells. TNS4-KD showed ~6.6-fold reduction in extensions, compared to control (Fig. 2C). For tubules, we plated equal numbers of control and TNS4-KD cells and counted the number of tubules consisting of TNS4-KD cells only, control cells only, or a mixture of cells. In total 300 tubules, 79±6.6% had a mixture of cells types, 19±7% had control cells only, and 2±1% had TNS4-KD-cells only, indicating that TNS4-KD cells were underrepresented in tubules compared to control (Fig. 2D). TNS4-KD cells were also found less frequently in the mixed tubules (not shown). These findings suggest TNS4 is required for extension formation and invasion into the ECM.
The reduced presence of KD cells in tubules prompted us to test whether forced expression of TNS4 would promote initial invasion. We nucleofected GFP-TNS4, and GFP+ cells were sorted by FACS; thus sorted cells express GFP-TNS4 with varying degrees, which facilitates cell tracing. Immunoblotting confirmed the expected size of GFP-TNS4 (Fig. 3A). We determined if the expression level of TNS4 correlates with extension formation and tubule initiation. Extensions were highlighted with beta-catenin, a basolateral marker and phalloidin, while tubules were labeled with ezrin, a sub-apical marker and Hoescht. En face (XY) confocal sections taken at levels representing the monolayer (rear) or the leading edge (front) of the projection, from the extension stage and the tubule stage were observed. Interestingly, higher levels of GFP-TNS4 were often found in cells containing extensions (Fig. 3B). Furthermore, cells expressing GFP-TNS4 were located at buds, suggesting that GFP-TNS4 expression provides an advantage in initiation of tubule invasion.
TNS4 protein contains an SH2 domain. A conserved arginine in the βB5 strand of this domain is critical for binding to phosphotyrosines in interaction partners; mutation of this arginine (R474A) in TNS4 abolishes binding to SH2 domain partners. To disrupt the SH2 domain, we nucleofected GFP-TNS4 (R474A), and GFP+ cells were pooled by FACS. Cells expressing GFP-TNS4 (R474A) showed increased extension formation at ~72-96hrs after co-culture. Strikingly, these cells cannot invade ECM. Instead, they were arrested in the extension stage up to 168hrs during tubulogenesis and showed very limited bud initiation (Fig. 3C). These results highlight that the functional SH2 domain is required for the transition from extension to tubule/bud initiation.
TNS4 induction is required for EGF-driven mammary cell scattering. We tested if TNS4 is induced by HGF. MDCK monolayers on a collagen matrix were treated with 25ng/ml HGF basolaterally for 24hrs. qRT-PCR confirmed that HGF induced a 3.88±0.39-fold increase in total transcript level of TNS4, suggesting HGF alone can induce TNS4.
We therefore tested how the phenotypes observed in 2.5D are interpreted in the HGF-induced 3D tubulogenesis. To form cysts, a single cell suspension of MDCK cells expressing either GFP-TNS4 or GFP-TNS4 (R474A), was grown in Matrigel and cysts then stimulated with HGF in 1:1 mixture of collagen and Matrigel to induce tubule formation. In response to HGF, cysts develop chains of cells, which asynchronously lead to cords and then tubules (Fig. 4A, +HGF). Expression of GFP-TNS4 or GFP-TNS4 (R474A) did not disrupt polarity, though cysts expressing GFP-TNS4 (R474A) were significantly larger (Fig. 4A, unstimulated). 48hrs after HGF application, both cysts expressing GFP-TNS4 and GFP-TNS4 (R474A) displayed invasive growth. However, GFP-TNS4 (R474A) strongly attenuated chain, chord, and tubule formation (Fig. 4A, +HGF). GFP-TNS4 (R474A) exhibited significantly increased extension formation at 12hrs after HGF treatment, compared to wild type TNS4 (not shown). Thus, a functional SH2 domain of TNS4 is necessary for the transition from extension to invasive epithelial migration to form tubules in 3D, similar to 2.5D.
Three principal downstream signaling pathways in 3D tubulogenesis have been identified: ERK1/2, STAT3, and AKT. We tested if GFP-TNS4 (R474A) affects the activation of these proteins. As determined by the phosphorylation status of these proteins in lysates from cysts and tubules (Fig. 4B), expression of GFP-TNS4 (R474A) gave no perturbation of ERK; phosphorylation of Akt was elevated by a borderline amount (P=0.05).
In contrast, GFP-TNS4 (R474A) showed significantly increased activation of STAT3, even without HGF treatment. Surprisingly, HGF application restored STAT3 activation to a level similar to control cells. Though it is puzzling how GFP-TNS4 (R474A) suppresses the increased basal activity of STAT3 after HGF treatment, a plausible explanation is that the correct timing of activation of STAT3 is required for tubulogenesis. To determine if the hyperactivation of STAT3 at the wrong time contributes to the inhibition of tubule formation in 3D by GFP-TNS4 (R474A), we generated stable cell lines expressing a constitutively active form of STAT3 (STAT3-C). Expression of STAT3-C caused ellipsoidal and spherical cysts with multiple lumens (Fig. 4C). Strikingly, these cells did not invade the ECM in response to HGF, similar to cysts expressing GFP-TNS4 (R474A). STAT3-C exerted similar defects in 2.5D (Fig. 4D). This suggests that untimely hyperactivation of STAT3 prevents HGF-induced invasive migration during tubulogenesis. It has been reported that temporal regulation of ERK1/2 and STAT1 is required for tubulogenesis[9, 23]. Together with these findings, our results highlight the importance of temporal regulation of these signaling for proper tubulognensis. Moreover, we have tested whether STAT3-C can induce TNS4 but STAT3-C does not induce TNS4 in our cell lines (not shown).
Our screened gene set includes multiple genes previously known to be involved in migration and tubulogenesis such as CXCR4[24, 25], DUSP6, PLAUR, SNAI2, CD44, and TINAG, as well as genes which have not been studied yet. The latter are novel candidates to analyze invasive cell migration from an epithelial sheet.
Previous work has shown that TNS4 may displace TNS3 from integrin β1; bound TNS4 releases stress fibers from integrin β1 as TNS4 lacks an actin-binding domain. This stress fiber disconnection promotes cell migration. This may be one mechanism of how TNS4 contributes to tubulogenesis. On the other hand, our observation that a mutation in the SH2 domain locks cells in the extension stage provides an additional view of the molecular mechanism of TNS4 in tubulogenesis. TNS4 may recruit other protein(s) through the SH2 domain, which leads to invasive migration during tubulogenesis. Furthermore, GFP-TNS4 (R474A) causes increased basal activation of STAT3, suggesting that inappropriate, untimely over-activation of STAT3 may inhibit HGF-induced tubulogenesis. Indeed, expression of STAT3-C phenocopies the tubule formation defects of GFP-TNS4 (R474A). Although it is unknown how this mutation in the SH2 domain leads to increased STAT3 activation, our study suggests that this aberrant activation of STAT3 blocks HGF-induced tubule formation.
Tubulogenesis occurs by different mechanisms in various developing systems[1, 30]. A central goal is to uncover common principles underlying this apparent diversity. A strength of the MDCK systems is that depending on the culture configuration, MDCK cells can form tubules via several cellular mechanisms, including chain cell migration from cysts (3D) and deformation of a sheet of cells migrate into the ECM (2.5D); these have parallels in various in vivo systems. We found that TNS4 is required for both systems, suggesting that TNS4 represents a common principle underlying multiple cellular mechanisms. Moreover, our results have uncovered an unexpected TNS4-STAT3 axis, which plays a central role in HGF-induced tubulogenesis.
We are grateful to Dr. Ross Metzger for critical reading and members of the Mostov lab for discussion. This work was supported by an NIH Ruth L. Kirschstein postdoctoral fellowship (DK082115) to S.-H.K., an NIH Institutional Research Training Grants (DK07762) to S.-H.K., a German Research Society Fellowship (DFG;Ne-897/1-1) to P.I.N., and NIH grants R01DK074398 and P01AI53194 to K.M.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.