Loss of SIRT1 Leads to Kidney Fibrosis via EMT

We next wished to test a possible role of SIRT1 in EMT in a completely different context from tumor metastasis by using kidney epithelial cells and a kidney injury model in mice expressing reduced or elevated levels of SIRT1 (Figure S2A).

Treatment of kidney epithelial HK-2 cells with TGF-β or aristocholic acid, to mimic kidney injury, gave rise to cells with mesenchymal phenotypes, as shown by an increase in cells positive for α-smooth muscle actin (α-SMA) and collagen type I (markers for mesenchymal cells) and a decrease in cells positive for E-cadherin, an epithelial marker (Figures 2A and S2B). Overexpression of SIRT1 in epithelial HK-2 cells led to a repression of EMT, as seen by a reduction in α-SMA and collagen-positive cells. Moreover, preservation of the epithelial phenotype was confirmed by the continued expression of E-cadherin after the treatments with TGF-β or aristocholic acid (Figures 2A and S2B). Conversely, loss of SIRT1 by siRNA in HK-2 kidney epithelial cells increased the EMT, as assessed by an increase in α-SMA and collagen-positive cells and a decrease in E-cadherin-positive cells (Figures 2A and S2B).

Open in a separate window

Figure 2

The Role of SIRT1 in EMT of Kidney Epithelial Cells and Kidney Fibrosis In Vivo

(A) Immunofluorescent staining of HK-2 cells with E-cadherin, α-smooth muscle actin (α-SMA), and collagen type I 48 hr following the treatment with TGF-β 5 ng/ml (63× magnification, DAPI staining of nuclei). Numbers in the right bottom corners of images represent percentage of E-cadherin, α-SMA, or collagen-type-I-positive cells over total number of cells per slide (n = 3 slides per group), shown as average ± SD, *p < 0.05, ANOVA Dunnett test versus Ctrl treated with TGF-β; T1Tg, HK-2 cells overexpressing SIRT1; siT1, HK-2 cells with knockdown for SIRT1; Ctrl, control HK-2 cells.

(B) Changes of serum creatinine over time following AKI (n = 12 for each time point).

(C) Survival following the AKI. *p ≤ 0.05 versus Ctrl KO, #p ≤ 0.05 versus Ctrl Tg, log-rank test.

(D) Blood urea nitrogen (BUN) values over time (n = 12 for each time point).

(E) Masson's trichrome staining showing mesenchymal tissue in blue and α-SMA immunofluorescent staining for myofibroblasts in red (magnification 20×, n = 6 per group).

(F) Interstitial fibrosis score of Masson's trichrome-stained kidney sections at 6 weeks following the AKI (n = 6 per group).

(G) Collagen content in kidneys 6 weeks following the AKI measured by Sircol assay (n = 6 per group).

(H) Masson's trichrome staining of kidneys after 4 days of treatment with TGF-β 100 μg/kg/day intravenously (magnification 5×).

(I) Interstitial fibrosis score of Masson's trichrome-stained kidney slides from (H).

(J) Collagen content in kidneys after 4 days of treatment with TGF-β 100 μg/kg/day intravenously, as determined by Sircol collagen assay. *Significantly different from Ctrl KO, #significantly different from Ctrl Tg, ANOVA Dunnett test, p ≤ 0.05. Error bars represent SD. KSTG, kidney-specific transgenic mice; Ctrl Tg, control for KSTG; KSKO, kidney-specific knockout mice; Ctrl KO, control for KSKO mice.

(K) Percentage of α-SMA-positive cells in kidneys of SIRT1^+/+ mice expressing the cre-recombinase (Ctrl) and KSKO mice detected by flow cytometry.

(L) Percentage of cre-recombinase-positive cells out of α-SMA-positive cells in kidneys of SIRT1^+/+ mice expressing cre-recombinase (Ctrl) and KSKO mice as detected by flow cytometry.

(M) Quantitative RT-PCR analysis of cre-recombinase expression in sorted α-SMA-positive cells from kidneys of SIRT1^+/+ mice expressing cre-recombinase (Ctrl) and KSKO mice.

Bars, SD, *p < 0.05 versus Ctrl, t test. See also Figure S2.

To test the effect of SIRT1 on kidney function in vivo, we studied the acute kidney injury (AKI) model of ischemia reperfusion in kidney-specific knockout mice (KSKO) and kidney-specific transgenic mice (KSTG) mice. KSKO mice eliminate SIRT1 from kidney collective tubular epithelial cells, while KSTG increase SIRT1 expression in these same cells (Figures S2C–S2E). At the baseline, KSTG and KSKO had comparable body weight and levels of serum electrolytes (Figures S2F and S2G). Serum creatinine values, representing kidney function, were increased in KSKO mice before the AKI (Figure 2B), suggesting the requirement of SIRT1 for maintenance of tubular epithelial function even in the absence of injury. After clamping the renal arteries to induce injury, the increase of the serum creatinine was much more pronounced in the absence of SIRT1 from kidney tubules (Figure 2B). There was an improvement in creatinine values in KSKO mice on day 4 following the AKI, since the mice with the worst glomerular function had died by that time (Figure 2C). KSTG showed better kidney function parameters with significantly lower creatinine values as compared to the control animals. Blood urea nitrogen (BUN) paralleled serum creatinine measurements (Figure 2D). Most importantly, overexpression of SIRT1 in kidney tubules protected animals from dying after AKI, while deletion exacerbated mortality (Figure 2C). Histological examination of the injured kidneys demonstrated destruction of the tubular basement membrane and disorientation of epithelial tubular cells following the AKI. These pheno-types were significantly more pronounced in KSKO and reduced in KSTG mice (Figures S2H and S2I).

Since aging impedes the recovery from AKI and predisposes to irreversible damage of kidneys (Abrass et al., 1995), we investigated the effect of SIRT1 on fibrosis in kidneys without the injury and on progression of AKI to tubulointerstitial fibrosis and development of chronic kidney damage (CKD) by following the KSKO and KSTG mice for 6 weeks after the AKI. At baseline, KSKO mice had a minimal increase in kidney fibrosis (Figure S2J). After the AKI, KSKO mice had significantly worse kidney function as detected by the increased serum creatinine (Figure S2K) and BUN (Figure S2L), more kidney fibrosis (Figure 2F) as confirmed by Masson's trichrome staining for connective tissue and α-SMA staining for myofibroblasts (Figure 2E) and increased kidney collagen content (Figure 2G). In contrast, KSTG showed improvement compared to the control mice subjected to kidney injury. In summary, the elimination of SIRT1 predisposes kidneys to fibrosis and progression from AKI to CKD, while overexpression of SIRT1 in kidney tubules improves the recovery after the AKI and slows the progression toward CKD.

Our findings above suggest that SIRT1 represses kidney fibrosis. Since it is known that the TGF-β pathway induces fibrosis (Ledbetter et al., 2000), we tested whether SIRT1 repression of TGF-β was sufficient to explain the observed repression in fibrosis. We treated wild-type, KSKO, and KSTG mice with 100 μg/kg of TGF-β intravenously for 4 days. In mice treated with TGF-β, we recapitulated the effects of SIRT1 seen after AKI. Namely, KSTG mice showed less fibrosis on histology (Figures 2H and 2I) and less collagen (Figure 2J) in kidney tissues, and KSKO mice showed more fibrosis and collagen after the treatment with TGF-β. These results suggest that SIRT1 represses kidney fibrosis by repressing the TGF-β pathway.

To address whether EMT was likely to be involved in fibrosis in wild-type or KSKO kidneys, we FACS sorted α-SMA-positive cells after treatment with TGF-β to cause fibrosis, and found, as expected, 4.8-fold more positive (fibrotic) cells in kidneys of KSKO mice (Figure 2K). Since cre-recombinase expression in KSKO mice is driven by the kidney epithelial collecting tubulespecific Hoxb7cre promoter (Yu et al., 2002), we tested for EMT by FACS sorting the α-SMA-positive cells and testing for cre expression. Consistent with recent reports (Humphreys et al., 2010), almost none of sorted α-SMA-positive cells had cre recombinase (0.02%) in SIRT1^+/+ mice expressing cre, after the TGF-β treatment. In striking contrast, 35.7% of α-SMA-positive cells in KSKO treated with TGF-β expressed cre recombinase (Figure 2L). Quantitative RT-PCR also demonstrated vastly higher cre expression in α-SMA-positive cells of KSKO compared to wild-type (Figure 2M). These findings suggest that EMT is not the mechanism of kidney fibrosis in control animals as recently reported (Humphreys et al., 2010). However, EMT is induced at a very high frequency in fibrotic kidney tissue in the absence of SIRT1.

Cells and Treatment

Immortalized tumorigenic primary human breast epithelial cells, HMLER cells (courtesy by R.A. Weinberg, Whitehead Institute), were maintained as described (Elenbaas et al., 2001). The SIRT1^+/+ and SIRT1^−/− MEFs have been described (Lin et al., 2000). HK-2 cells were from ATCC. Cells were treated with 5 ng/ml TGF-β (Invitrogen) or aristocholic acid (Sigma) every other day for 2–8 days (Yang et al., 2010; Yu et al., 2002).

Plasmids and Cell Transfections

Stable cells were generated using lentivirus or retrovirus. The plasmids expressing mSIRT1 and PRRL-GFP vectors were purchased from Addgene, Smad4-shRNA, and pSM2 vector from Open Biosystems. siRNA-SIRT1 plasmid has been previously described (Picard et al., 2004). Lentiviral vectors were made by pcr subcloning cDNA and ligating into pENTR-D-TOPO (Invitrogen). pENTR-D-TOPO vectors were sequenced and then the cDNA was transferred to pLenti4-TO-V5-DEST (Invitrogen) through recombination. Cells were then virally infected. Transient transfections were performed by using Fugene transfection reagent (Roche).

In Vivo Tumorigenesis and Metastasis Model

Severe combined immunodeficiency (SCID) mice (Charles River Laboratories) were used in these studies, and all the protocols were approved by the Massachusetts Institute of Technology Committee on Animal Care. Mice were anesthetized with isoflurane. Five million cells in PBS, diluted 1:2 with Matrigel (Beckton Dickinson), were injected into one mammary gland per female (Onder et al., 2008). Animals were sacrificed 7.5 or 8 weeks postinjection (depending when the primary tumors reached 2 cm in diameter), or sooner if they became moribund.

Meta-Analysis of Breast Cancer Clinical Trial

We performed meta-analysis of SIRT1 expression in Oncomine database of breast cancers (http://www.oncomine.com). The analysis included 12 trials with 143 control and 1181 breast cancer tissues (Table S1). These 12 trials had 35 comparisons. We used the cutoff value of p = 0.01 for significance of SIRT1 expression change in any direction.

Acute Kidney Injury Model

Four-month-old mice (n = 12 per group) were subjected to ischemia-reperfusion model of AKI by clamping both renal pedicles for 30 min with vascular clips (Roboz) (Yang et al., 2010). Blood was collected throughout the experiment and kidneys at the end. Mice were sacrificed 6 days and 6 weeks following the AKI and experiment was repeated twice.

TGF-β Injections

TGF-β 100 mg/kg/day was injected intravenously to six mice/group for 4 days to induce kidney fibrosis. Mice were sacrificed on day 5 and kidneys were collected for analyses. The experiment was repeated twice.

Renal Function

Serum level of creatinine was measured by alkaline picrate, BUN by urease and potassium by ion-selective electrode method (Antech Diagnostics).

Histology, Histomorphometry, and Immunohistochemistry

Metastasis, kidney, bone, and adipose tissue histology was examined on formalin-fixed, paraffin-embedded sections 3 mm thick stained by hematoxylin-eosin and Masson's trichrome for kidneys (six sections/kidney). The degree of interstitial fibrosis was scored semiqantitatively on a 0–4 scale (0, no lesion; 1, <25%; 2, >25%–50%; 3, >50%–75%; 4, >75% of parenchyma affected by the lesion) (Yang et al., 2010). The collagen content of kidneys was quantified with the Sercol soluble collagen assay kit (Biocolor).

Vectastain kit (Vector Laboratories) and MM7 (Abcam) and large SV40 T antigen (Santa Cruz) antibodies were used to perform MMP7 staining of three consecutive kidney sections and large SV40 T antigen, used to transform HMLE cells, staining of tumor metastasis.

Immunofluorescent Staining

Kidney sections were labeled with antibody to α-SMA (Abcam) and collagen type IV. HMLER cells and HK-2 cells on coverslips were labeled with antibodies to E-cadherin (BD Transduction Laboratories), α-SMA, collagen type I and type IV (Abcam), vimentin (Abcam), and MMP7. HMLER cells were labeled with β-catenin antibody. The slides were exposed to Alexa Fluor 488, 568, and 594 anti-rabbit and anti-mouse secondary antibodies (Invitrogen). The staining was examined with fluorescence microscope (Zeiss AxioImager.M1). HMLER cells labeled with GFP were detected macroscopically after forming metastasis at necroscopy when organs were examined under a Leica MZ 12 fluorescence dissection microscope.

Western Blotting and Immunoprecipitation

Organs and cells were homogenized in RIPA buffer including Complete Protease Inhibitor (Roche). Protein (100 mg) was loaded onto SDS-PAGE gels and immunoblotted with anti-SIRT1 (Upstate), Smad2 (Abcam), Smad4 (Abcam), E-cadherin (Cell Signaling), MMP7 (Abcam), Acetyl-Lysine (Santa Cruz Biotechnology), and tubulin (Abcam) antibodies.

The immunoprecipitations were carried out by using Pierce Direct IP Kit (Thermo Scientific). To determine Smad2 and Smad4 acetylation, anti-Smad2, anti-Smad4, or normal rabbit serum were coupled to agarose beads, and then control, KSKO, KSTG kidney samples, HMLER control, and HMLER SIRT1 KD cells were incubated with the beads. The eluate was blotted with anti-Smad2, anti-Smad4, and anti-Acetyl-Lys antibodies.

Flow Cytometric Analysis and Sorting

Cell suspension from kidneys of control mice with cre-recombinase and KSKO mice after 4 days of treatment with TGF-β 100 mg/kg/day was prepared as described in Kitamoto et al. (2009). Following erythrocyte lysis, cells were resuspended in FACS buffer. Kidney cells were labeled using anti-α-SMA and anti-cre-recombinase primary antibodies and Alexa Fluor 488 and Alexa Fluor 647 secondary antibodies, counted by a FACSCalibur Flow Cytometer (BD Biosciences) and analyzed using FlowJo software (Ashland, OR). α-SMA-positive cells were sorted using FACSAria I (BD Biosciences).

RNA Isolation and Analysis

Total RNA from mouse kidneys and HMLER cells was isolated by using Trizol (QIAGEN). cDNA was synthesized from total RNA by SuperScript III reverse transcriptase (Invitrogen) with random primers. The cDNA was then subjected to real-time quantitative RT-PCR analysis with gene-specific primers in the presence of CYBR green (Bio-Rad). Relative abundance of mRNA was normalized to rpl19.

Chromatin Immunoprecipitation Analysis

Chromatin immunoprecipitation analysis from kidneys and HMLER cells was performed with Magna Chip kit (Millipore) by using Smad4 antibody (Abcam) and immunoglobulin (Ig) G. The primer sequences are in Table S2.

Luciferase Assay

Transcription of Smad2-Smad4 was detected by transfecting MEFs with FM-SIM plasmid comprising of Smad-interacting motif (SIM) and Fast/FoxH1 motif (FM), which interacts with phospho-Smad2, fused to the Gal4 DNA-binding domain (Figure S5A) (Ross et al., 2006; courtesy of C. Hill, Cancer Research, UK). The MMP7 promoter region responsive to Smad4 (–438 to –379) and a larger DNA promoter fragment (–660 to –216) were cloned into pGL3-basic vector luciferase plasmid (Addgene) by using Sac1 and Xhol restriction endonucleases (New England BioLabs). Cells were cotransfected with MMP7 promoter-luciferase reporter and pRL-TK (Renilla luciferase; Promega) and treated with TGF-β 2 ng/ml for 48 hr. Luciferase activity was measured by Dual-Luciferase Reporter Assay System (Promega). The firefly luciferase activity was normalized to Renilla. The experiments were performed in triplicates and were repeated three times.

We are grateful to Caroline Hill for FM-SIM plasmid and Robert A. Weinberg for HMLER cell line. We wish to thank Benjamin Humphreys for advice and helpful discussions. This work is supported by Fulbright Association postdoctoral fellowship to P.S., American Cancer Society, Grant Number: PF-11-258-1-TBG to E.O.W., F32 CA 132358 to E.L.B., Ellison/AFAR postdoctoral research grant to M.B. and grants from the NIH and a gift from the Glenn Medical Foundation to L.G. L.G. is co-chair of the Senior Advisory Board for Sirtris/GSK.