A Novel Mechanism by Which Hepatocyte Growth Factor Blocks Tubular Epithelial to Mesenchymal Transition

A Novel Mechanism by which Hepatocyte Growth Factor Blocks Tubular Epithelial to Mesenchymal Transition

Junwei Yang,*† Chunsun Dai,* and Youhua Liu* *Division of Cellular and Molecular Pathology, Department of Pathology, University of Pittsburgh School of Medicine, Pittsburgh, Pennsylvania; and †Division of Nephrology, Department of Medicine, Nanjing Medical University, Nanjing, China.

Hepatocyte growth factor (HGF) is a potent antifibrotic cytokine that blocks tubular epithelial to mesenchymal transition (EMT) induced by TGF-␤1. However, the underlying mechanism remains largely unknown. This study investigated the signaling events that lead to HGF blockade of the TGF-␤1–initiated EMT. Incubation of human kidney epithelial cells HKC with HGF only marginally affected the expression of TGF-␤1 and its type I and type II receptors, suggesting that disruption of TGF-␤1 signaling likely plays a critical role in mediating HGF inhibition of TGF-␤1 action. However, HGF neither affected TGF-␤1–induced Smad-2 phosphorylation and its subsequent nuclear translocation nor influenced the expression of inhibitory Smad-6 and -7 in tubular epithelial cells. HGF specifically induced the expression of Smad transcriptional co-repressor SnoN but not Ski and TG-interacting factor at both mRNA and protein levels in HKC cells. SnoN physically interacted with activated Smad-2 by forming transcriptionally inactive complex and overrode the profibrotic action of TGF-␤1. In vivo, HGF did not affect Smad-2 activation and its nuclear accumulation in tubular epithelium, but it restored SnoN protein abundance in the fibrotic kidney in obstructive nephropathy. Hence, HGF blocks EMT by antagonizing TGF-␤1’s action via upregulating Smad transcriptional co-repressor SnoN expression. These findings not only identify a novel mode of interaction between the signals activated by HGF receptor tyrosine kinase and TGF-␤ receptor serine/threonine kinases but also illustrate the feasibility of confining Smad activity as an effective strategy for blocking renal fibrosis. J Am Soc Nephrol 16: 68–78, 2005. doi: 10.1681/ASN.2003090795

epatocyte growth factor (HGF) has recently emerged this reduction of TGF-␤1 in vivo is the primary mechanism as a key antifibrotic cytokine that displays a potent mediating HGF’s actions or simply reflects a secondary conse- therapeutic potential in preventing tissue fibrosis af- quence associated with reduced fibrotic lesions. The activity of H ␤ ter chronic injury (1,2). The ability of HGF to inhibit tissue TGF- 1 is tightly regulated by multiple levels of controlling fibrotic lesions has been attested extensively in a wide variety mechanisms (17–21). Apart from the modulation of its expres- of organs, including kidney (3–6). Growing evidence demon- sion, TGF-␤1 signal transduction circuit is a subject of regula- strates that administration of HGF protein or its gene prevents tion by other signal inputs under different circumstances the progressive loss of kidney functions in chronic renal dis- (17,19). Along this line, we hypothesized that HGF may antag- eases with diverse causes (7–11). Accordingly, blockade of HGF onize TGF-␤1 action in tubular epithelial cells through inter- signaling by a neutralizing antibody markedly exacerbates the cepting its signal transduction. progression of renal fibrosis and dysfunctions (12,13). These TGF-␤1 signals are transduced by transmembrane type II and studies establish that supplementation of exogenous HGF may type I receptors that contain a cytoplasmic serine/threonine provide an effective strategy for combating chronic renal insuf- kinase domain (18,22). Receptor activation initiates a cascade of ficiency (14–16). signaling transduction events involving intracellular mediator Whereas the antifibrotic capacity of HGF has been increas- Smad proteins (23). TGF-␤1 specifically initiates Smad-2 and -3 ingly recognized, the molecular mechanism underlying its ac- phosphorylation, which in turn bind to the common partner tions remains largely unknown. Earlier studies revealed that Smad-4 and translocate into cell nuclei, where they control the administration of HGF markedly inhibits the expression of transcription of TGF-␤–responsive genes. Activated Smads TGF-␤1 (6,13), a well-characterized profibrotic cytokine, in fi- may also assemble complexes with transcriptional co-repres- brotic tissues in vivo. However, it remains a question whether sors, making Smads transcriptionally inactive (24–26). Thus, Smad transcriptional co-repressors could be an important reg- ulatory component that confines TGF-␤1 signaling (26). Need-

Received September 29, 2003. Accepted September 17, 2004. less to say, disruption of any steps in this cascade of signal transduction processes may potentially disturb TGF-␤1 signal- Published online ahead of print. Publication date available at www.jasn.org. ing and result in blockage of EMT. Address correspondence to: Dr. Youhua Liu, Department of Pathology, Univer- In this study, we systematically dissected the signaling sity of Pittsburgh School of Medicine, S-405 Biomedical Science Tower, 200 Lothrop Street, Pittsburgh, PA 15261. Phone: 412-648-8253; Fax: 412-648-1916; events that control HGF blockade of tubular epithelial to mes- E-mail: [email protected] enchymal transition (EMT) induced by TGF-␤1. Our results

indicate that HGF fails to block Smad activation and nuclear Immunofluorescence Staining translocation. Rather, HGF specifically upregulates Smad tran- Indirect immunofluorescence staining was performed using an es- scriptional co-repressor SnoN expression. Such SnoN induction tablished procedure (7). Cells were incubated with the specific primary leads to the formation of transcriptionally inactive SnoN/Smad antibodies described above, followed by staining with FITC- or cyanine complex and blocks TGF-␤1’s action. These findings define a dye Cy3-conjugated secondary antibodies. For some samples, cells were double stained with 4Ј,6-diamidino-2-phenylindole, HCl to visu- novel molecular pathway by which HGF antagonizes profi- alize the nuclei. For immunostaining of kidney sections, cryosections brotic TGF-␤1 signaling. were stained with anti–Smad-2/3 and anti-SnoN antibodies using the Vector M.O.M. immunodetection kit by the protocol specified by the Materials and Methods manufacturer (Vector Laboratories, Burlingame, CA). The slides were Cell Culture, Cytokine Treatment, and Transient then stained for the proximal tubular marker with fluorescein-conju- Transfection gated lectin from Tetragonolobus purpureas (Sigma). Slides were viewed Human proximal tubular epithelial cells (HKC) were provided by Dr. with a Nikon Eclipse E600 Epi-fluorescence microscope equipped with L. Racusen (Johns Hopkins University, Baltimore, Maryland) (27). Cell a digital camera (Melville, NY). In each experimental setting, immuno- culture and cytokine treatments were carried out according to the fluorescence images were captured with identical light exposure times. procedures described previously (7,28). Briefly, HKC cells were seeded in complete medium that contained 10% FBS at approximately 70% Nuclear Protein Preparation confluence. After an overnight incubation, cells were continuously HKC cells were subjected to various treatments with various growth incubated in serum-free medium for another 24 h, before addition of factors for 30 min except where otherwise indicated. Cell nuclei were cytokines. Recombinant human TGF-␤1 was purchased fromR&D isolated by procedures described previously (32). After being collected Systems (Minneapolis, MN). Recombinant human HGF protein was by centrifugation at 5000 ϫ g for 30 min at 4°C, the nuclei were lysed provided by Genentech Inc. (South San Francisco, CA). Chemical in- with SDS sample buffer and subjected to Western blot analysis. hibitors PD98059, wortmannin, and SC68376 were purchased from Calbiochem (La Jolla, CA). Transient transfection of HKC cells was carried out by using the Lipofectamine 2000 according to the instruc- Immunoprecipitation tions specified by the manufacturer (Invitrogen, Carlsbad, CA). After Immunoprecipitation was carried out by a procedure described else- transfection with equal amounts of HA-tagged SnoN expression plas- where (32). HKC cell lysates were immunoprecipitated overnight at 4°C ␮ ␮ mid (29) (provided by Dr. R. Weinberg, Massachusetts Institute of with 1 g of anti–Smad-2/3, followed by precipitation with 20 lof Technology, Cambridge, Massachusetts) or empty vector pcDNA3 (In- protein A/G Plus-Agarose for3hat4°C. After being washed, the vitrogen), HKC cells were incubated in the absence or presence of 2 immunoprecipitates were boiled for 5 min in SDS sample buffer, fol- ng/ml TGF-␤1 for 2 d. Whole-cell lysates were prepared and subjected lowed by immunoblotting with specific antibodies. to Western blot analyses. The stable cell lines overexpressing Smad-7 (HKCpSmad7) and mock-transfection control (HKCpcDNA3) were estab- Animals lished as described previously (30). Male CD-1 mice that weighed approximately 18 to 22 g were purchased from Harlan Sprague Dawley (Indianapolis, IN). Unilateral Northern Blot Analysis ureteral obstruction (UUO) was performed using an established proce- Total RNA isolation and Northern blot analysis for gene expression dure (6,7). Human HGF expression plasmid (pCMV-HGF) and the were carried out by the procedures described previously (31). The empty vector (pcDNA3) were administered to mice by rapid injection cDNA probe for TGF-␤1 was obtained from American Type Culture of naked plasmid solution through the tail vein, as described previ- collection (ATCC, Manassas, VA). The SnoN cDNA probe was pro- ously (6). Mice received an injection of plasmids before (day Ϫ1) and vided by R. Weinberg. After autoradiography, membranes were after (day 3) UUO, respectively. Groups of mice (n ϭ 5) were killed at stripped and rehybridized with rat glyceraldehyde-3-phosphate dehy- 7 d after UUO, and the kidneys were removed. The therapeutic effects drogenase probe to ensure equal loading of each lane. of HGF gene delivery in this mouse model were confirmed and reported previously (6). Western Blot Analysis The preparation of whole-cell lysates and kidney tissue homogenates and Statistical Analyses Western blot analysis of protein expression were performed according to the Statistical analysis of the data was performed using SigmaStat soft- procedures described previously (7). The primary antibodies were obtained ware (Jandel Scientific, San Rafael, CA). Comparison between groups from following sources: anti–␣-smooth muscle actin (anti–␣-SMA) and anti– was made using one-way ANOVA followed by Student-Newman- extracellular signal-regulated kinase-1 and -2 (anti–Erk-1/2; Sigma, St. Louis, Kuels test. P Ͻ 0.05 was considered significant. MO); anti–phospho-specific Erk-1/2, phospho-specific and total p38 mitogen- activated protein kinase (MAPK), and phospho-specific and total c-Jun N- Results terminal kinase (Cell Signaling Technology, Inc., Beverley, MA); anti–phos- ␤ ␤ Regulation of TGF- 1 and Its Receptors by HGF in Tubular pho-specific and total Smad-2 (Upstate, Charlottesville, VA); anti–TGF- 1 Epithelial Cells (sc-146), anti–TGF-␤ type I receptor (anti-T␤RI; sc-398), anti–TGF-␤ type II To gain insights into the mechanism underlying HGF block- receptor (anti-T␤RII; sc-17792), anti–Smad-7 (sc-7004), anti–Smad-6 (sc-13048), ade of TGF-␤1–initiated tubular EMT, we first investigated the anti–Smad-4 (sc-7066), anti–Smad-2/3 (sc-6032), anti-Sp1 (sc-420), anti–c-Ski ␤ (sc-9140), anti-SnoN (sc-9595), anti-TGIF (sc-9826), and anti-actin (sc-1616; effect of HGF on TGF- 1 expression in tubular epithelial cells. Santa Cruz Biotechnology, Inc., Santa Cruz, CA); and anti-fibronectin (clone As shown in Figure 1A, HGF did not significantly affect ␤ 10) and anti–E-cadherin (clone 36; BD Biosciences Pharmingen, San Jose, CA). TGF- 1 mRNA expression at 1, 6, and 12 h after treatment in Affinity-purified secondary antibodies were purchased from Jackson Immu- HKC cells. At 24 h, moderate inhibition of TGF-␤1 mRNA level noResearch Laboratories, Inc. (West Grove, PA). was noticeable, but it returned to the control level at 48 h. The 70 Journal of the American Society of Nephrology J Am Soc Nephrol 16: 68–78, 2005

Tubular EMT Is Dependent on Intracellular Smad Signaling We next explored the possibility that HGF blocks tubular EMT by interfering TGF-␤1 signal transduction. To this end, we initially attempted to identify the TGF-␤1 signaling pathway that is important for mediating EMT in HKC cells. A compre- hensive screening of the activation profile of various signaling pathways stimulated by TGF-␤1 in tubular epithelial cells was carried out by using phospho-specific antibodies. As shown in Figure 2, A through D, TGF-␤1 stimulated Smad-2, Akt/protein kinase B, and p38 MAPK phosphorylation in a time-dependent manner. However, TGF-␤1 seemed not to activate further Erk- 1/2 (Figure 2D) and c-Jun N-terminal kinase (data not shown) in tubular epithelial cells. To determine which signal pathway initiated by TGF-␤1is critical for mediating EMT, we examined the effects of specific blockade of a particular signal pathway on the phenotypic conversion induced by TGF-␤1. As shown in Figure 2E, blockade of Akt activation by wortmannin and p38 MAPK activation by SC68376 did not affect TGF-␤1–induced suppression of epithelial E-cadherin and de novo expression of ␣-SMA, two hall- marks of tubular EMT. Of note, these chemical inhibitors at the concentrations used were able to specifically block Akt and p38 MAPK activation initiated by TGF-␤1 in HKC cells, as reported Figure 1. Regulation of TGF-␤1 and its receptors by hepatocyte previously (28,30). We next found that disruption of Smad growth factor (HGF) in tubular epithelial cells. (A) Northern signaling by overexpressing inhibitory Smad-7 completely blot analysis demonstrates the steady-state level of TGF-␤1 abolished TGF-␤1–induced tubular EMT. As shown in Figure mRNA in tubular epithelial HKC cells after HGF incubation. 2F, in cell line with mock transfection of pcDNA3, TGF-␤1 HKC cells were treated without or with 20 ng/ml HGF for caused E-cadherin suppression and ␣-SMA and fibronectin various periods of time as indicated. Total cellular RNA sam- induction in a time-dependent manner, whereas TGF-␤1 totally ␤ ples were hybridized with cDNA probes for TGF- 1 and glyc- failed to induce the phenotypic conversion in cell line overex- eraldehyde-3-phosphate dehydrogenase (GAPDH), respec- pressing Smad-7. Of note, overexpression of Smad-7 inhibits tively. (B) Western blot analysis shows the protein abundance Smad-2 phosphorylation in HKC cells (30). Hence, it seems of TGF-␤1 and TGF-␤ type I and type II receptors (T␤RI and likely that TGF-␤1–induced EMT is dependent on intracellular T␤RII) in tubular epithelial cells after various treatments. Whole-cell lysates derived from the HKC cells that were treated Smad signaling. with HGF (20 ng/ml), TGF-␤1 (2 ng/ml), or both for various periods of time as indicated were immunoblotted with specific HGF Neither Blocks Smad-2 Phosphorylation and Its ␤ ␤ ␤ antibodies against TGF- 1, T RI, T RII, and actin, respectively. Association with Smad-4 and Subsequent Nuclear ␤ The size and authenticity of TGF- 1 were confirmed by using Translocation nor Influences Inhibitory Smads Expression ␤ the purified TGF- 1 protein in the adjacent lane (data not Having established that Smad signaling is crucial in mediat- shown). ing TGF-␤1–induced EMT, we reasoned that HGF may block EMT by impairing Smad signal transduction. To test this hy- pothesis, we examined the effect of HGF on major Smad sig- level of active TGF-␤1 in the supernatant of HKC cells was naling events in a step-wise manner. As shown in Figure 3, extremely low or undetectable as measured by a specific ELISA, treatment of HKC cells with TGF-␤1 induced Smad-2 phos- and HGF did not alter the abundance of active TGF-␤1 secreted phorylation and activation. However, preincubation of the cells by HKC cells (data not shown). Furthermore, Western blot of with HGF did not affect TGF-␤1–induced Smad-2 phosphory- whole-cell lysates revealed that HGF did not affect TGF-␤1 lation (Figure 3A). protein abundance at 24 and 48 h after incubation (Figure 1B). We next examined whether HGF disturbs the TGF-␤1–induced We further examined the potential effect of HGF on TGF-␤1 association between Smad-2/3 and Smad-4 by immunoprecipita- receptor expression in tubular epithelial cells. As demonstrated tion. As shown in Figure 3B, after stimulation with TGF-␤1, in Figure 1B, treatment of HKC cells with HGF did not inhibit Smad-4 was clearly present in the complexes precipitated the abundance of T␤RI and T␤RII protein, as illustrated by by Smad-2/3 antibody, indicating that TGF-␤1 induces Smad- Western blot analysis. In fact, a trivial increase in T␤RI level 2/3–Smad-4 complex formation in HKC cells. However, HGF also was noticed in HKC cells after stimulation with either HGF or failed to disrupt the physical association between Smad-2/3 and TGF-␤1. However, it returned to the control level when the cells Smad-4 induced by TGF-␤1 (Figure 3B). were simultaneously incubated with both HGF and TGF-␤1 We further investigated whether HGF blocks activated (Figure 1B). Smad-2 nuclear translocation, because this process has been J Am Soc Nephrol 16: 68–78, 2005 HGF Blocks EMT by Upregulating SnoN 71

Figure 2. Tubular epithelial to mesenchymal transition (EMT) induced by TGF-␤1 is dependent on intracellular Smad signaling. (A through D) TGF-␤1 triggers various signal events in tubular epithelial cells. HKC cells were incubated with 2 ng/ml TGF-␤1 for various durations as indicated, and whole-cell lysates were immunoblotted with either phospho-specific or total Smad-2 (A), Akt (B), p38 mitogen-activated protein kinase (MAPK; C), and extracellular signal-regulated kinase-1 and -2 (Erk-1/2) MAPK (D). (E) Chemical blockade of MAPK and Akt activation does not affect TGF-␤1–induced tubular EMT. HKC cells were pretreated with either various chemical inhibitors or vehicle (DMSO) for 30 min, followed by incubating in the absence or presence of 2 ng/ml TGF-␤1 for 48 h. Specific inhibitors for phosphoinositide 3-kinase (10 nM wortmannin), Mek1 (10 ␮M PD98059), and p38 MAPK (20 ␮M SC68376) were used. (F) Overexpression of inhibitory Smad-7 completely blocks TGF-␤1–induced EMT. Stable cell lines either overexpressing Smad-7 (HKCpSmad-7) or with mock-transfection of empty vector (HKCpcDNA3) were treated with 2 ng/ml TGF-␤1 for various periods of time as indicated. Whole-cell lysates were immunoblotted with antibodies against Smad-7, E-cadherin, ␣-smooth muscle actin (␣-SMA), fibronectin, and actin, respectively. shown to be tightly regulated by HGF in renal interstitial and -7, two inhibitory members of the Smad protein family, at fibroblasts (33). As presented in Figure 3C, HGF did not block different time points in HKC cells. In addition, HGF exhibited activated Smad-2 nuclear translocation in tubular epithelial no effects on common Smad-4 protein expression (Figure 4). cells, as the nuclear abundance of phospho-Smad-2 was virtu- ally identical in HKC cells that were treated either with TGF-␤1 HGF Specifically Induces Smad Transcriptional Co-repressor alone or with TGF-␤1 plus HGF. To confirm this result further, SnoN Expression we used an indirect immunofluorescence staining to visualize We next analyzed the expression of Smad transcriptional the nuclear accumulation of activated Smad-2 in HKC cells co-repressors in HKC cells after treatment with HGF. As shown after various treatments. Consistent with the biochemical assay, in Figure 5, HGF dramatically induced SnoN protein expres- treatment of HKC cells with TGF-␤1 induced marked accumu- sion. This induction started as early as 1 h and peaked at 6 h lation of phospho-Smad-2 in the nuclei (Figure 3E), and HGF after HGF incubation (Figure 5). Under the same conditions, apparently had no effect on this TGF-␤1–initiated Smad-2 nu- HGF had little effect on the expression of Ski, a Smad transcrip- clear translocation (Figure 3G). Collectively, this series of ex- tional co-repressor that shares a high degree of homology with periments demonstrates that HGF blocks TGF-␤1–mediated SnoN. The expression of TGIF, another Smad co-repressor that EMT by a mechanism independent of disruption of Smad-2 is upregulated by HGF through protein stabilization in mesan- phosphorylation, its association with Smad-4, and its nuclear gial cells (34), was extremely low or undetectable under both translocation. basal and HGF-stimulated conditions in HKC cells. Because overexpression of inhibitory Smad-7 completely Figure 5 also shows the localization of SnoN in tubular abolished TGF-␤1–induced tubular EMT (Figure 2), we rea- epithelial cells after HGF treatment. In control HKC cells, im- soned that HGF may antagonize TGF-␤1’s action by upregulat- munofluorescence staining for SnoN protein was essentially ing inhibitory Smads expression. As shown in Figure 4, HGF negative. However, treatment with HGF dramatically induced also did not influence the protein expression of both Smad-6 SnoN protein expression, which was exclusively localized in 72 Journal of the American Society of Nephrology J Am Soc Nephrol 16: 68–78, 2005

Figure 3. HGF does not affect TGF-␤1–induced Smad-2 phosphorylation and its binding to Smad-4 and subsequent nuclear translocation in tubular epithelial cells. HKC cells were pretreated with 20 ng/ml HGF for 30 min, followed by incubation with 2 ng/ml TGF-␤1 for 45 min. (A) Whole-cell lysates were immunoblotted with antibodies against phospho-specific or total Smad-2, respectively. (B) Cell lysates were immunoprecipitated with anti–Smad-2/3 antibody, followed by immunoblotting with antibodies against Smad-4 and Smad-2/3, respectively. (C) Nuclear protein lysates were prepared from HKC cells after various treatments as indicated and immunoblotted with anti–phospho-specific Smad-2 antibody. The blot was also probed with antibody against Sp1 for normalization of the nuclear proteins loaded each lane. n.s. indicates nonspecific band. (D through G) Immuno- fluorescence staining for phospho-specific Smad-2 demonstrates that HGF does not block TGF-␤1–mediated Smad-2 nuclear translocation. (D) Control. (E) TGF-␤1 (2 ng/ml). (F) HGF (20 ng/ml). (G) TGF-␤1 plus HGF.

SnoN Forms Complex with Activated Smad-2 and Overrides the Profibrotic Action of TGF-␤1 To understand how increased SnoN blocks TGF-␤1’s action in tubular epithelial cells, we investigated whether SnoN can physically interact with activated Smad-2 once it has been translocated to the nuclei after TGF-␤1 stimulation. As shown in Figure 6, in HKC cells that were treated with TGF-␤1, phosphorylated Smad-2 (green) was translocated to the nuclei (Fig- ure 6B). After treatment with HGF alone, SnoN (red) was induced and localized in the nuclei (Figure 6C). When HKC cells were treated with TGF-␤1 plus HGF, both phosphorylated Smad-2 (green) and SnoN (red) were co-localized in the nuclei (yellow; Figure 6D). To reveal a potential interaction between Figure 4. HGF does not affect inhibitory Smad expression in activated Smad-2 and SnoN, we investigated Smad/SnoN com- tubular epithelial cells. HKC cells were treated with 20 ng/ml plex formation in tubular epithelial cells by co-immunoprecipi- HGF for various periods of time as indicated. Whole-cell lysates tation. As shown in Figure 6E, when cell lysates were precipi- were immunoblotted with specific antibodies against Smad-4, tated with anti-SnoN antibody, phospho-Smad-2 was Smad-6, Smad-7, and actin, respectively. detectable in the precipitates by Western blotting. There was a profound increase in the association between Smad-2 and SnoN ␤ the nuclei. To study whether SnoN protein induction by HGF is in the cells that were treated with TGF- 1 plus HGF, compared ␤ due to an enhanced protein stability or increased gene expres- with TGF- 1 or HGF treatment alone. This suggests that HGF- sion, we examined the SnoN mRNA levels after HGF treatment induced SnoN physically binds to the activated Smad-2 in a ␤ in HKC cells by Northern blot analysis. As illustrated in Figure TGF- 1–dependent manner, which presumably sequesters ac- 5G, four transcripts of SnoN with the sizes of 6.2, 4.4, 3.2, and tivated Smad-2 from transactivation of its target genes. ␤ 2.1 kb were detected in tubular epithelial cells. These tran- Analysis of the whole-cell lysates revealed that TGF- 1by scripts represent differential splicing variants (35). HGF mark- itself did not induce SnoN expression in HKC cells (Figure 6F). edly increased the steady-state mRNA levels of SnoN in HKC In addition, TGF-␤1 did not affect HGF-induced SnoN expres- cells. Such induction began as early as 0.5 h and peaked at 1 h sion. Identical magnitude of SnoN induction by HGF was ob- after HGF treatment, a kinetics that significantly precedes SnoN served in the absence or presence TGF-␤1 (Figure 6F). protein induction. Hence, SnoN protein induction by HGF is To provide direct evidence for SnoN in mediating HGF in- primarily mediated by an increased gene expression in tubular hibition of TGF-␤1 action, we ectopically transfected SnoN epithelial cells. expression vector into tubular epithelial cells, followed by as- J Am Soc Nephrol 16: 68–78, 2005 HGF Blocks EMT by Upregulating SnoN 73

Figure 5. HGF specifically induces Smad transcriptional co-repressor SnoN expression in tubular epithelial cells. (A) Western blot analysis shows that HGF specifically induces the protein abundance of SnoN but not Ski and TG-interacting factor (TGIF) in tubular epithelial cells. HKC cells were treated with 50 ng/ml HGF for various periods of time as indicated. Whole-cell lysates were immunoblotted with specific antibodies against SnoN, Ski, TGIF, and actin, respectively. (B) Graphical presentation of the relative abundance of SnoN after treatment with HGF at various time points as indicated. Relative levels (fold induction over control) are presented after normalized with actin. (C through F) Immunofluorescence staining shows the localization of SnoN in the control (C and D) and HGF-treated cells (E and F). (C and E) Staining for SnoN. (D and F) Nuclear staining with 4Ј,6-diamidino-2-phenylindole, HCl. (G) Northern blot analysis demonstrates SnoN mRNA induction by HGF in HKC cells. Total RNA isolated from HKC cells that were treated without or with HGF for various periods of time as indicated was blotted with cDNA probes for SnoN and GAPDH, respectively. The sizes of transcripts are indicated on the left. sessing their responsiveness to TGF-␤1 stimulation. As shown HGF gene did not block Smad-2 phosphorylation and activa- in Figure 6G, forced expression of exogenous SnoN dramati- tion. Figure 7B shows quantitative results derived from five cally blocked TGF-␤1–induced E-cadherin suppression and animals per group. This finding is consistent with in vitro data ␣-SMA induction in a dose-dependent manner. However, (Figure 3), suggesting that HGF blockade of renal fibrosis in transfection of empty vector in an identical manner did not vivo is not mediated by inhibiting Smad activation. affect TGF-␤1–initiated tubular EMT. Similar results were ob- We further investigated Smad localization in normal and tained by using an indirect immunofluorescence staining ap- diseased kidney by an indirect immunofluorescence staining. proach (Figure 6, H through M). These results indicate that As demonstrated in Figure 7C, there was abundant Smad-2/3 SnoN can effectively override the profibrotic action of TGF-␤1. staining in renal tubular epithelium of sham-operated kidney, when using specific antibody against total Smad-2/3. However, HGF Does not Affect Smad-2 Phosphorylation and Nuclear Smads were apparently localized in the cytoplasm of tubular Accumulation in Tubular Epithelium In Vivo epithelial cells under normal conditions. In the obstructed kid- Earlier studies showed that administration of either HGF ney at 7 d after UUO, Smad-2/3 were predominantly localized protein or its gene prevents tubular EMT and renal interstitial in the nuclei of renal tubular epithelial cells, suggesting that fibrosis induced by UUO (7). To investigate the potential mech- Smads were activated and translocated into the nuclei. Admin- anism underlying HGF’s beneficial effect in the fibrotic kidney, istration of HGF did not prevent Smads activation and their we studied the effect of HGF on Smad signaling in vivo.As nuclear accumulation in the fibrotic kidney (Figure 7E). shown in Figure 7, there was a dramatic activation of Smad-2, as illustrated by a marked increase in phospho-Smad-2 in the HGF Restores SnoN Expression in the Fibrotic Kidney homogenates of obstructed kidney, suggesting an active In view of the induction of SnoN by HGF in tubular epithelial TGF-␤1 signaling in diseased kidneys. However, despite HGF’s cells in vitro, we examined the SnoN expression in the fibrotic preventing renal EMT and fibrosis in this model, delivery of kidney in vivo. Figure 8 shows SnoN protein expression in the 74 Journal of the American Society of Nephrology J Am Soc Nephrol 16: 68–78, 2005

Figure 6. SnoN binds to activated Smad-2 and overrides the profibrotic action of TGF-␤1 in tubular epithelial cells. (A through D) Double immunofluorescence staining for phospho-specific Smad-2 (green) and SnoN (red) in HKC cells after various treatments for 24 h. (A) Control. (B) TGF-␤1 (2 ng/ml). (C) HGF (50 ng/ml). (D) TGF-␤1 plus HGF. (E) Co-immunoprecipitation demonstrates that HGF induces an association between SnoN and activated Smad-2 in tubular epithelial cells. Cell lysates derived from various groups as indicated were immunoprecipitated with anti-SnoN antibody, followed by immunoblotting with antibodies against phospho-Smad-2 and SnoN, respectively. (F) Whole-cell lysates derived from various groups as indicated were immunoblotted with antibodies against Smad-2, SnoN, and actin, respectively. (G) Ectopic expression of SnoN overrides the profibrotic action of TGF-␤1 in tubular epithelial cells. HKC cells were transiently transfected with increasing amounts of HA-tagged SnoN expression vector or empty pcDNA3 as indicated (␮g/well) and followed by incubation in the absence or presence of 2 ng/ml TGF-␤1 for 2 d. Whole-cell lysates were immunoblotted with antibodies against HA, E-cadherin, ␣-SMA, and actin, respectively. (H through M) Immunofluorescence staining demonstrates the effects of SnoN overexpression on TGF-␤1–mediated E-cadherin suppression and ␣-SMA induction. HKC cells were transfected with either pcDNA3 (I and L) or HA-tagged SnoN expression vector (J and M; 5 ␮g/well), followed by incubation with 2 ng/ml TGF-␤1 for 2 d. Cells were immunostained with antibodies against E-cadherin (H through J) or ␣-SMA (K through M), respectively. (H and K) Control HKC cells. kidney of various treatment groups. In sham-operated kidney, cells remained unsolved. Thus, the aim of this study was to abundant SnoN protein expression was observed by Western dissect the signaling events that lead to HGF blockade of blot analysis of total tissue homogenates (Figure 8A). Immuno- TGF-␤1 action. We demonstrated that HGF blocks TGF-␤1– fluorescence staining displayed that SnoN was primarily local- initiated EMT apparently by a mechanism independent of ized in the nuclei of tubular epithelium. After obstructive injury Smads phosphorylation and their subsequent nuclear translo- for 7 d, intrarenal SnoN protein was dramatically decreased, as cation. Instead, HGF specifically induces Smad transcriptional described previously (26). co-repressor SnoN expression, which in turn binds to the acti- After administration of HGF gene, SnoN expression was vated Smads and consequently sequesters their ability to initi- markedly induced in the obstructed kidney when compared ate gene transcription in tubular epithelial cells. These actions with the pcDNA3 controls (Figure 8). Both Western blot anal- of HGF are largely recapitulated in the fibrotic kidney induced ysis and immunofluorescence staining revealed that HGF es- by obstructive injury in vivo. Therefore, our findings have un- sentially restored the SnoN protein expression in the fibrotic covered a novel molecular pathway by which HGF specifically kidney inhibited by ureteral obstruction. Hence, HGF may elicit overrides the profibrotic action of TGF-␤1. its antifibrotic activity in vivo by confining profibrotic TGF-␤1 HGF is a potent antifibrotic factor that prevents the onset and signaling via restoring SnoN expression. progression of chronic renal disease in a variety of animal models (6,8,14). Although the precise mechanism by which Discussion HGF inhibits tissue fibrosis is not fully understood, its interplay Earlier studies have shown that HGF completely blocks TGF- with TGF-␤1 has been suspected to play a crucial role in me- ␤1–mediated tubular EMT (7), a phenotypic conversion that is diating its antifibrotic action (7,13). TGF-␤1 is shown to inhibit thought to play an imperative role in the generation of the HGF expression in various types of cells, including renal mes- matrix-producing myofibroblasts (36,37). However, how ex- angial and endothelial cells (38). However, whether HGF recip- actly HGF antagonizes the TGF-␤1 action in tubular epithelial rocally suppress TGF-␤1 expression remains largely elusive. In J Am Soc Nephrol 16: 68–78, 2005 HGF Blocks EMT by Upregulating SnoN 75

Figure 7. HGF does not inhibit Smad-2 phosphorylation and its nuclear translocation in renal tubular epithelium after obstructive injury in vivo. (A) Western blot demonstrates that delivery of HGF gene does not prevent Smad-2 phosphorylation in the obstructed kidney. Whole-tissue homogenates from different groups as indicated were immunoblotted with antibodies against phospho-specific and total Smad-2, respectively. Numbers (1, 2, and 3) indicate each individual animal in a given group. (B) Graphical presentation of the abundance of phospho-Smad-2 after normalization with total Smad-2 in different groups. Relative levels (fold induction over sham control) are presented as means Ϯ SEM of five animals per group. *P Ͻ 0.01 versus the sham control. (C through E) Immunofluorescence staining shows the localization of Smad-2/3 in sham and obstructed kidneys. Renal cryosections were immunostained with anti–Smad-2/3 antibody, followed by staining with FITC-conjugated lectin from T. purpureas. (C) Sham. (D) Unilateral ureteral obstruction (UUO) with pcDNA3. (E) UUO with pCMV-HGF.

vivo studies seem to suggest that HGF may inhibit TGF-␤1 tively controlling mechanisms that confine TGF-␤1 signaling expression, because administration of HGF reduces TGF-␤1 along the Smad signal transduction pathway. In the extracellular abundance in the fibrotic kidneys (7,13). However, the cause– compartment, active TGF-␤1 can bind to decorin, a proteoglycan effect relationship of this TGF-␤1 reduction in fibrotic kidney associated with the extracellular matrix, which leads to negative remains ambiguous. The present study indicates that HGF only regulation of TGF-␤1 activity (41). This provides a means of con- marginally affects TGF-␤1 and its receptor expression in tubu- straining TGF-␤1 activity at the prereceptor stage. Smad signaling lar epithelial cells, consistent with a recent report indicating can also be negatively controlled by inhibitory Smad-7 and -6, that HGF fails to modulate TGF-␤1 expression in mesangial which act as functional decoys that compete with the receptor- cells at basal condition (10). Such a trivial effect of HGF on regulated Smads (R-Smads) for binding to the activated type I TGF-␤1 and its receptor expression in opposite manners (Fig- receptor, leading to an attenuated R-Smad phosphorylation and ure 1) hardly accounts for its dramatic antifibrotic action. activation (17,22). This level of regulation takes place in the cyto- Hence, the primary mechanism for mediating HGF activity plasm. Once inside the nuclei, Smad signaling is finally subjected may operate through blocking hyperactive TGF-␤1 signaling. to a third level of negative regulation by Smad transcriptional Because TGF-␤1 could induce itself and its receptor expression co-repressors. It is intriguing that HGF seems not to interfere with (39), it is conceivable that HGF blockade of TGF-␤1 signaling Smad signaling in tubular epithelial cells until the assembly of would indirectly lead to an inhibition of TGF-␤1 induction by Smad transcriptional complexes. Major TGF-␤1 signaling events uncoupling the TGF-␤1–positive feedback loop under patho- are still operative upon HGF treatment (Figure 3). By upregulating logic settings. In this context, we interpret the reduction of transcriptional antagonist SnoN, HGF receptor signaling leads to TGF-␤1 in the diseased kidney after HGF treatment as a con- sequestering the ability of R-Smads to initiate the transactivation sequence associated with reduced fibrotic lesions. of TGF-␤1–target genes. These findings underscore that HGF is HGF and TGF-␤1 transduce their signals by distinct mecha- capable of precisely targeting a final, decisive step in Smad sig- nisms. One would expect a cross-talk in the signaling circuits naling before the transcription of TGF-␤1–responsive genes. Our mediated by HGF receptor tyrosine kinase and TGF-␤ receptor observations also establish a novel mode of interactions between serine-threonine kinases, to make it possible for HGF to counteract the signals activated by HGF receptor tyrosine kinase and TGF-␤ the profibrotic action of TGF-␤1. Such cross-talk could theoreti- receptor serine-threonine kinases ligands. cally take place at any point during the cascade of TGF-␤1/Smad SnoN is abundantly expressed and predominantly localized signaling events (33,40). There are at least three levels of nega- in the nuclei in normal tubular epithelium (Figure 8) (26). 76 Journal of the American Society of Nephrology J Am Soc Nephrol 16: 68–78, 2005

Figure 8. HGF restores SnoN expression in obstructed kidney in vivo. (A) Representative Western blot demonstrates that HGF largely restores SnoN abundances by inducing its expression in the obstructed kidney. Whole-tissue homogenates were immunoblotted with antibodies against SnoN and actin, respectively. Numbers (1, 2, and 3) indicate each individual animal in a given group. (B) Graphical presentation of the relative abundance of SnoN after normalization with actin in different groups. Data are presented as means Ϯ SEM of five animals per group. *P Ͻ 0.01 versus sham; †P Ͻ 0.01 versus pcDNA3 (n ϭ 5). (C through E) Immunofluorescence staining shows the localization of SnoN protein in the obstructed kidney at 7 d after surgery. (C) Sham. (D) UUO with pcDNA3. (E) UUO with pCMV-HGF.

Under physiologic conditions, a high level of SnoN would be In summary, the present study identifies a unique mode of important in preventing the activation of transcription by Smad interaction between the signals activated by HGF and TGF-␤1 proteins that may find their way into the nuclei (24). In this in tubular epithelial cells. It is conceivable that HGF can pre- respect, SnoN functions as a crucial gatekeeper that tightly cisely target the hyperactive Smad signaling through inducing constrains Smad signaling under normal conditions. Consis- Smad transcriptional antagonist SnoN expression. Such specific tently, knockdown of SnoN by siRNA strategy dramatically interception of Smad signaling by HGF may eventually over- sensitizes the tubular epithelial cells to respond to TGF-␤1 ride the profibrotic action of TGF-␤1 in the diseased kidneys, stimulation (26). However, this confining mechanism seems not thereby halting the progression of chronic renal fibrosis. to operate in the fibrotic kidney, because SnoN protein abundance is significantly reduced (26). Consequently, Smad signal would be transduced without any restraint after TGF-␤1 stim- Acknowledgments ulation, leading to an amplified transcriptional activation of This work was supported by National Institutes of Health grants TGF-␤1–responsive genes. DK054922, DK061408, and DK064005 (Y.L.). J.Y. and C.D. were sup- The novel finding that HGF specifically upregulates SnoN ported by postdoctoral fellowships from the American Heart Associa- tion Pennsylvania–Delaware Affiliate. expression provides a mechanistic insight into understanding We thank Dr. R. Weinberg for generously providing plasmid vectors. the interplay between antifibrotic HGF and profibrotic TGF-␤1 signaling. It becomes clear that the opposite effects of HGF and TGF-␤1 in regulating myofibroblast activation and tissue fibrosis are coupled by SnoN expression. By inducing SnoN and References restoring its level in the fibrotic kidney, HGF reinstates SnoN 1. Matsumoto K, Nakamura T: Hepatocyte growth factor: transcriptional repressor activity, leading to blockade of tubu- Renotropic role and potential therapeutics for renal dis- lar EMT and renal fibrosis. Notably, administration of HGF eases. Kidney Int 59: 2023–2038, 2001 could induce SnoN expression and block renal fibrosis even in 2. Liu Y: Hepatocyte growth factor in kidney fibrosis: Ther- apeutic potential and mechanisms of action. Am J Physiol the presence of Smad-2 phosphorylation and nuclear accumu- Renal Physiol 287: F7–F16, 2004 lation (Figure 7). This observation is significant in that it sug- 3. Mizuno S, Kurosawa T, Matsumoto K, Mizuno- gests that the levels of SnoN within the cell but not activated Horikawa Y, Okamoto M, Nakamura T: Hepatocyte ␤ Smad determine an outcome of TGF- 1 response in vivo.In growth factor prevents renal fibrosis and dysfunction in accordance with this, delayed administration of HGF, when the a mouse model of chronic renal disease. J Clin Invest 101: injury was already established, is still effective in reversing 1827–1834, 1998 EMT and blocking renal fibrosis (9,10,15). 4. Ueki T, Kaneda Y, Tsutsui H, Nakanishi K, Sawa Y, Mor- J Am Soc Nephrol 16: 68–78, 2005 HGF Blocks EMT by Upregulating SnoN 77

ishita R, Matsumoto K, Nakamura T, Takahashi H, Oka- TGF-beta signal transduction and mesangial cell fibrogen- moto E, Fujimoto J: Hepatocyte growth factor gene therapy esis. Am J Physiol Renal Physiol 284: F243–F252, 2003 of liver cirrhosis in rats. Nat Med 5: 226–230, 1999 21. Eddy AA: Molecular basis of renal fibrosis. Pediatr Nephrol 5. Azuma H, Takahara S, Matsumoto K, Ichimaru N, Wang 15: 290–301, 2000 JD, Moriyama T, Waaga AM, Kitamura M, Otsuki Y, 22. Piek E, Heldin CH, Ten Dijke P: Specificity, diversity, and Okuyama A, Katsuoka Y, Chandraker A, Sayegh MH, regulation in TGF-beta superfamily signaling. FASEB J 13: Nakamura T: Hepatocyte growth factor prevents the de- 2105–2124, 1999 velopment of chronic allograft nephropathy in rats. JAm 23. ten Dijke P, Miyazono K, Heldin CH: Signaling inputs Soc Nephrol 12: 1280–1292, 2001 converge on nuclear effectors in TGF-beta signaling. Trends 6. Yang J, Dai C, Liu Y: Systemic administration of naked Biochem Sci 25: 64–70, 2000 plasmid encoding hepatocyte growth factor ameliorates 24. Wotton D, Massague J: Smad transcriptional corepressors chronic renal fibrosis in mice. Gene Ther 8: 1470–1479, 2001 in TGF beta family signaling. Curr Top Microbiol Immunol 7. Yang J, Liu Y: Blockage of tubular epithelial to myofibro- 254: 145–164, 2001 blast transition by hepatocyte growth factor prevents renal 25. Frederick JP, Wang XF: Smads “freeze” when they ski. interstitial fibrosis. J Am Soc Nephrol 13: 96–107, 2002 Structure (Camb) 10: 1607–1611, 2002 8. Inoue T, Okada H, Kobayashi T, Watanabe Y, Kanno Y, 26. Yang J, Zhang X, Li Y, Liu Y: Downregulation of Smad Kopp JB, Nishida T, Takigawa M, Ueno M, Nakamura T, transcriptional corepressors SnoN and Ski in the fibrotic ␤ Suzuki H: Hepatocyte growth factor counteracts trans- kidney: An amplification mechanism for TGF- 1 signaling. forming growth factor-␤1, through attenuation of connec- J Am Soc Nephrol 14: 3167–3177, 2003 tive tissue growth factor induction, and prevents renal 27. Racusen LC, Monteil C, Sgrignoli A, Lucskay M, Marouil- fibrogenesis in 5/6 nephrectomized mice. FASEB J 17: 268– lat S, Rhim JG, Morin JP: Cell lines with extended in vitro 270, 2003 growth potential from human renal proximal tubule: Char- 9. Dworkin LD, Gong R, Tolbert E, Centracchio J, Yano N, acterization, response to inducers, and comparison with Zanabi AR, Esparza A, Rifai A: Hepatocyte growth factor established cell lines. J Lab Clin Med 129: 318–329, 1997 28. Li Y, Yang J, Dai C, Wu C, Liu Y: Role for integrin-linked ameliorates progression of interstitial fibrosis in rats with kinase in mediating tubular epithelial to mesenchymal established renal injury. Kidney Int 65: 409–419, 2004 transition and renal interstitial fibrogenesis. J Clin Invest 10. Mizuno S, Nakamura T: Suppressions of chronic glomer- 112: 503–516, 2003 ular injuries and TGF-beta 1 production by HGF in atten- 29. Sun Y, Liu X, Ng-Eaton E, Lodish HF, Weinberg RA: SnoN uation of murine diabetic nephropathy. Am J Physiol Renal and Ski protooncoproteins are rapidly degraded in re- Physiol 286: F134–F143, 2004 sponse to transforming growth factor beta signaling. Proc 11. Dai C, Yang J, Bastacky S, Xia J, Li Y, Liu Y: Intravenous Natl Acad Sci U S A 96: 12442–12447, 1999 administration of hepatocyte growth factor gene amelio- 30. Dai C, Yang J, Liu Y: Transforming growth factor-␤1 po- rates diabetic nephropathy in mice. J Am Soc Nephrol 15: tentiates renal tubular epithelial cell death by a mechanism 2637–2647, 2004 independent of Smad signaling. J Biol Chem 278: 12537– 12. Liu Y, Rajur K, Tolbert E, Dworkin LD: Endogenous hepa- 12545, 2003 tocyte growth factor ameliorates chronic renal injury by 31. Yang J, Shultz RW, Mars WM, Wegner RE, Li Y, Dai C, activating matrix degradation pathways. Kidney Int 58: Nejak K, Liu Y: Disruption of tissue-type plasminogen 2028–2043, 2000 activator gene in mice reduces renal interstitial fibrosis in 13. Mizuno S, Matsumoto K, Kurosawa T, Mizuno-Horikawa obstructive nephropathy. J Clin Invest 110: 1525–1538, 2002 Y, Nakamura T: Reciprocal balance of hepatocyte growth 32. Zhang X, Li Y, Dai C, Yang J, Mundel P, Liu Y: Sp1 and Sp3 factor and transforming growth factor-beta 1 in renal fi- transcription factors synergistically regulate HGF receptor brosis in mice. Kidney Int 57: 937–948, 2000 gene expression in kidney. Am J Physiol Renal Physiol 284: 14. Gao X, Mae H, Ayabe N, Takai T, Oshima K, Hattori M, F82–F94, 2003 Ueki T, Fujimoto J, Tanizawa T: Hepatocyte growth factor 33. Yang J, Dai C, Liu Y: Hepatocyte growth factor suppresses gene therapy retards the progression of chronic obstructive renal interstitial myofibroblast activation and intercepts nephropathy. Kidney Int 62: 1238–1248, 2002 Smad signal transduction. Am J Pathol 163: 621–632, 2003 15. Yang J, Liu Y: Delayed administration of hepatocyte 34. Dai C, Liu Y: Hepatocyte growth factor antagonizes the growth factor reduces renal fibrosis in obstructive ne- profibrotic action of TGF-␤1 in mesangial cells by stabiliz- phropathy. Am J Physiol Renal Physiol 284: F349–F357, 2003 ing Smad transcriptional corepressor TGIF. J Am Soc Neph- 16. Yang J, Dai C, Liu Y: Hepatocyte growth factor gene ther- rol 15: 1402–1412, 2004 apy and angiotensin II blockade synergistically attenuate 35. Nomura N, Sasamoto S, Ishii S, Date T, Matsui M, Ishizaki renal interstitial fibrosis in mice. J Am Soc Nephrol 13: R: Isolation of human cDNA clones of ski and the ski- 2464–2477, 2002 related gene, sno. Nucleic Acids Res 17: 5489–5500, 1989 17. Massague J, Chen YG: Controlling TGF-beta signaling. 36. Liu Y: Epithelial to mesenchymal transition in renal fibro- Genes Dev 14: 627–644, 2000 genesis: Pathologic significance, molecular mechanism, 18. Massague J: How cells read TGF-beta signals. Nat Rev Mol and therapeutic intervention. J Am Soc Nephrol 15: 1–12, Cell Biol 1: 169–178, 2000 2004 19. Bottinger EP, Bitzer M: TGF-␤1 signaling in renal disease. 37. Kalluri R, Neilson EG: Epithelial-mesenchymal transition J Am Soc Nephrol 13: 2600–2610, 2002 and its implications for fibrosis. J Clin Invest 112: 1776– 20. Schnaper HW, Hayashida T, Hubchak SC, Poncelet AC: 1784, 2003 78 Journal of the American Society of Nephrology J Am Soc Nephrol 16: 68–78, 2005

38. Yo Y, Morishita R, Yamamoto K, Tomita N, Kida I, Ha- 40. Lo RS, Wotton D, Massague J: Epidermal growth factor yashi S, Moriguchi A, Kato S, Matsumoto K, Nakamura T, signaling via Ras controls the Smad transcriptional co- Higaki J, Ogihara T: Actions of hepatocyte growth factor as repressor TGIF. EMBO J 20: 128–136, 2001 a local modulator in the kidney: Potential role in patho- 41. Isaka Y, Brees DK, Ikegaya K, Kaneda Y, Imai E, Noble NA, genesis of renal disease. Kidney Int 53: 50–58, 1998 Border WA: Gene therapy by skeletal muscle expression of 39. Yu L, Border WA, Huang Y, Noble NA: TGF-beta isoforms decorin prevents fibrotic disease in rat kidney. Nat Med 2: in renal fibrogenesis. Kidney Int 64: 844–856, 2003 418–423, 1996