Id1 expression in kidney endothelial cells protects against diabetes‐induced microvascular injury

The inhibitor of differentiation (Id) transcription regulators, which are induced in response to oxidative stress, promote cell proliferation and inhibit senescence. Inhibitor of differentiation 1 (Id1) expression is limited to endothelial cells (EC) in the normal mouse kidney and is required for a normal response to injury. Endothelial dysfunction leads to the development of diabetic nephropathy, and so, we hypothesized that endothelial Id1 may help protect against hyperglycemia‐induced microvascular injury and nephropathy. Here, we tested this hypothesis by using streptozotocin to induce diabetes in Id1 knockout (KO) mice and WT B6;129 littermates and examining the mice at 3 months. Expression of Id1 was observed to be increased 15‐fold in WT kidney EC, and Id1 KO mice exhibited increased mesangial and myofibroblast proliferation, matrix deposition, and albuminuria compared with WT mice. Electron microscopy demonstrated peritubular capillary EC injury and lumen narrowing, and fluorescence microangiography showed a 45% reduction in capillary perfusion area with no reduction in CD31‐stained areas in Id1 KO mice. Microarray analysis of EC isolated from WT and KO control and diabetic mice demonstrated activation of senescence pathways in KO cells. Kidneys from KO diabetic mice showed increased histological expression of senescence markers. In addition, premature senescence in cultured KO EC was also seen in response to oxidative stress. In conclusion, endothelial Id1 upregulation with hyperglycemia protects against microvascular injury and senescence and subsequent nephropathy.

The inhibitor of differentiation (Id) transcription regulators, which are induced in response to oxidative stress, promote cell proliferation and inhibit senescence. Inhibitor of differentiation 1 (Id1) expression is limited to endothelial cells (EC) in the normal mouse kidney and is required for a normal response to injury. Endothelial dysfunction leads to the development of diabetic nephropathy, and so, we hypothesized that endothelial Id1 may help protect against hyperglycemia-induced microvascular injury and nephropathy. Here, we tested this hypothesis by using streptozotocin to induce diabetes in Id1 knockout (KO) mice and WT B6;129 littermates and examining the mice at 3 months. Expression of Id1 was observed to be increased 15-fold in WT kidney EC, and Id1 KO mice exhibited increased mesangial and myofibroblast proliferation, matrix deposition, and albuminuria compared with WT mice. Electron microscopy demonstrated peritubular capillary EC injury and lumen narrowing, and fluorescence microangiography showed a 45% reduction in capillary perfusion area with no reduction in CD31-stained areas in Id1 KO mice. Microarray analysis of EC isolated from WT and KO control and diabetic mice demonstrated activation of senescence pathways in KO cells. Kidneys from KO diabetic mice showed increased histological expression of senescence markers. In addition, premature senescence in cultured KO EC was also seen in response to oxidative stress. In conclusion, endothelial Id1 upregulation with hyperglycemia protects against microvascular injury and senescence and subsequent nephropathy.
Microvascular impairment due to endothelial dysfunction is a major complication of diabetes and has been shown to be an early promoter of nephropathy [1,2]. Hyperglycemia adversely affects key endothelial functions including maintenance of vascular tone, permeability, matrix secretion, and angiogenesis. Extensive research has clarified the mechanisms by which this occurs with a central role of reactive oxygen species (ROS) production by NADPH oxidase (Nox) [3] and mitochondrial superoxide production [4,5] contributing to vascular dysfunction and stimulation of inflammatory and profibrotic molecules, including TGF-b1.
While hyperglycemia-induced metabolic derangements and resulting ROS-mediated endothelial cell (EC) dysfunction have been well documented, it remains unclear how this leads to the development of diabetic nephropathy. In addition to the profibrotic effects of decreased nitric oxide production [6], other proposed mechanisms include endothelial-mesenchymal transition to fibroblasts [7], EC apoptosis [8], or senescence [9], all potentially resulting in microvascular rarefaction and kidney fibrosis.
TGFb can be counteracted by bone morphogenetic protein (BMP) signaling through a related group of receptors that signal through Smad proteins and downstream induction of Id (inhibitor of differentiation) expression. Ids are a family of 4 (Id1-4) related proteins with distinct cell-specific expression and regulation, and function as dominant-negative regulators of profibrotic basic helix-loop-helix (bHLH) transcription factors (reviewed in Ref. [10]). Id expression is decreased in response to TGFb/Smad 2/3 but has a sustained increase in response to BMP/Smad 1/5/8 signaling. Id proteins are induced by oxidative stress [11,12] and promote EC survival and delay onset of cellular senescence [13], suggesting a potentially important role in preventing EC dysfunction.
The key role of inhibitor of differentiation 1 (Id1) expression in promoting EC survival and preventing capillary loss and fibrosis has been shown in lung [14] and colon [15] injury models. Kidney EC express Id1 at relatively high levels compared with other tissues, but whether or not Id1 has a role in preventing kidney microvascular injury remains unknown. Given the prosurvival effects of increased Id1 expression following oxidative stress, we sought to determine the effects of streptozotocin (stz)-induced hyperglycemia on endothelial Id1 expression and the effect of Id1 knockout (KO) on EC response to hyperglycemia, and the resulting effect on the development of nephropathy and the mechanisms by which this may occur. We demonstrate that hyperglycemia and oxidative stress increase endothelial Id1 expression. Id1 KO results in endothelial injury and expression of senescence markers in response to hyperglycemia with resulting microvascular damage and nephropathy.

Id1 expression is increased in WT EC following streptozotocin-induced DM and Id1 KO results in DM-induced EC injury and nephropathy
To determine whether Id1 expression changes following stz-induced diabetes, Id1 was examined by western blot using whole kidney lysates and by immunohistochemistry and double-labeling immunofluorescence in kidney sections from vehicle-injected control mice and at 3 months following stz injection. Compared to control 5-month-old male B6;129 littermates, Id1 protein levels increased up to 15-fold following stz-induced diabetes (Fig. 1A,B). Id1 was detected in peritubular, arteriolar, and glomerular EC in both normal (Fig. 1C) and stz-treated mice (Fig. 1D), with increased expression in glomerular capillary EC (Fig. 1D,F, EC labeled with CD31) in stz treated compared with control mice (Fig. 1C,E). No Id1 expression was detected in other kidney cell types using these techniques.
To determine whether increased Id1 expression has a protective role and prevents hyperglycemia-induced kidney injury, Id1 KO mice in a diabetic nephropathyresistant B6;129 mouse strain were examined and compared with WT mice at 3 months following stz injection (Fig. 2). As previously reported, WT male B6;129 mice displayed mild mesangial matrix expansion by periodic acid-Schiff (PAS) staining (Fig. 2C,D,K) but no other pathological changes by silver staining ( Fig. 2A,B,L) [16]. In contrast, Id1 KO littermates developed mesangial expansion and mesangial and peritubular matrix deposition by PAS (Fig. 2G,H,K) and silver staining (Fig. 2E,F,I,J,L).

Diabetic Id1 KO mice have decreased peritubular capillary perfusion
The effect of Id1 KO on capillary density following hyperglycemia was determined by fluorescence microangiography [17] to examine perfusion, combined with CD31 staining for capillary density (Fig. 4). Differences in microvasculature perfusion and pattern were quantified for both fluorescent labels using AN-GIOTOOL [18]. No significant differences between control WT ( found. By microangiography, KO diabetic mice (Fig. 4D) had a 45% reduction in microvascular area and a 100% increase in lacunarity, indicating greater pattern variability [19] (Fig. 4E,F). No difference in capillary density was detected by CD31 staining (data not shown). This was associated with a large number of poorly perfused CD31-positive glomerular and peritubular capillaries (Fig. 4D).
Gene expression microarrays reveal activation of pro-inflammatory and fibrotic pathways and genes associated with cell senescence Whole-genome microarray analysis was performed to identify molecular pathways responsible for hyperglycemia-induced damage of the renal vasculature in KO mice. EC from both control and diabetic WT and Id1 KO mice (n = 3-4/group) were isolated by intravital fluorescence labeling with a VE-Cadherin antibody followed by cell sorting. Kidney sections from these mice showed labeling of peritubular and glomerular EC in all four groups (Fig. S1). EC RNA was used for Agilent gene expression microarray analysis. Principal component and hierarchical clustering analysis demonstrated very close correlation of gene expression levels within normal WT and KO EC groups and the presence of one partially dissimilar sample from both the WT and KO diabetic mice (Fig. S2). Analysis of differentially expressed genes revealed a large number of significantly upregulated compared with downregulated genes in EC isolated from KO in comparison with WT control mice (Table S1), consistent with Id1 function as a transcription inhibitor. Id proteins inhibit bHLH transcription factors by both protein dimerization and blocking transcription [20]. Multiple HLH factors were significantly increased in KO EC including a compensatory increase in Id4 (Table 1). Pathway analysis using INGENUITYÒ IPA software (Qiagen, Redwood City, CA, USA) revealed numerous significantly upregulated pathways in KO EC from control mice as expected given the large number of differentially expressed genes. The canonical TGFb superfamily/ Smad signaling pathway was one of the most significantly affected pathways (Fig. S3, P = 0.00002), supporting the specificity of the microarray.
Additional comparisons focused on differences between WT and KO diabetic mice. Pathway analysis revealed that EC from diabetic KO mice had significant increases in gene expression associated with actin cytoskeleton activation including paxillin, integrin, and adhesion kinase signaling in comparison with diabetic WT mice. Downregulated gene expression included pathways associated with mitochondrial dysfunction, oxidative phosphorylation, and RXR activation, with a common upstream regulation by PPARa, suggesting that Id1 may be a downstream mediator of PPARa signaling ( Table 2). In response to the PPARa agonist fenofibrate, MyEnd microvascular EC cultures showed a fivefold increase in Id1 protein levels, further supporting this possibility (Fig. S4). These pathways are identical to those identified by gene expression microarray [21] and proteomics analysis [22] of senescent-cultured EC. KO diabetic EC had increased inflammatory gene expression including IL-6 compared with KO control EC. In contrast, no inflammatory genes were increased in WT diabetic EC compared with control EC. Sirtuin 1 expression, a key inhibitor of EC senescent phenotype [23], however, was increased 2.5-fold in these cells. Analysis of microarray results comparing KO and WT EC from nondiabetic mice revealed activation of key DNA damage response-induced senescence pathways in KO cells including p53 and ATM (P = 0.0003) and NF-jB (P = 0.0004). Combining gene expression lists from the Reactome [24] and Molecular Signatures Database (Fridman_Senescence_Up) [25] for cell senescence showed upregulation of 59 genes in KO control and diabetic cells including cell cycle regulatory, IGF pathway, MAPK signaling, and cytoskeletal and pro-inflammatory factor genes that are characteristic of this phenotype (Table S2, P = 0.03 using Fisher's exact test).

Increased expression of senescence-associated bgalactosidase and heterochromatin in Id1 KO diabetic mice
Senescence-associated b-galactosidase (SABG) expression by the Glb1 gene [26] is one of the hallmarks of cell senescence. Glb1 expression was increased 3.5-fold (P = 0.04) by microarray analysis in KO diabetic EC. To determine whether senescent cells accumulate in Id1 KO diabetic kidneys, tissue was stained for SABG and MacroH2A1.1, a histone enriched in senescenceassociated heterochromatin foci [27] that is used as a marker for senescent cells [28]. As shown in Fig. 5, transmission electron microscopy of SABG-stained kidneys showed that X-gal crystals were present in epithelial ( (Fig. 6A). Peritubular EC expression was also detected in KO diabetic kidneys (Fig. 6D) along with a twofold to threefold increase in expression compared with WT diabetic kidneys by western blot (Fig. 6E,F). No MacroH2A1.1 expression was detected by either technique in WT control kidneys.

Increased expression of fibronectin and aSMA in Id1 KO diabetic mice
Expression of the matrix protein fibronectin is another hallmark of senescent cells [29]. Increased microarray expression in EC isolated from KO control and diabetic EC was verified by immunofluorescence of kidney tissue (Fig. 7A,B), demonstrating increased glomerular endothelial cell signal in diabetic KO kidneys and by western blot (Fig. 7C,D), demonstrating a fourfold increase in expression in KO diabetic kidneys.
Paracrine effects of endothelial injury were evaluated by examining pericapillary aSMA expression, a marker for pericyte or myofibroblast activation, by immunohistochemistry. aSMA expression was confined to arterioles in WT diabetic kidneys (Fig. 7E). Increased expression (Fig. 7H) was detected in KO diabetic kidneys in both glomeruli (Fig. 7G) and peritubular cells (Fig. 7F), and small number of cortical peritubular capillaries (Fig. 7G). The association between Id1 and aSMA expression was also examined by immunohistochemistry in two genetic models of diabetes including DBA.2Akita and Lep ob /WiscJ mice (Fig. S5). DBA.2Akita diabetic mice had nearly undetectable Id1 expression compared with DBA.2 WT and Lep ob / WiscJ diabetic mice. Interstitial and glomerular aSMA staining was only detected in DBA.2Akita mice, correlating with results from Id1 KO mice.  increased ROS in KO cells (Fig. 8G) compared with control KO cells (Fig. 8E), an effect inhibited by pretreatment with a Nox inhibitor (Fig. 8D,H). Positive control cells treated with the oxidant menadione are shown in Fig. 8B,F. No differences in mitochondrial superoxide production were detected using MitoSOX with the same experimental conditions (data not shown). To determine whether hyperglycemia-induced ROS results in DNA damage, WT and KO cells were stained for 8-OHdG following treatment with 25 mM glucose for 12 h. Correlating with the ROS assay, the oxidant menadione induced 8-OHdG in both genotypes, while hyperglycemia-induced oxidative DNA damage was only detected in KO cells (Fig. 8I).
The effect of Id1 expression on the oxidative stressinduced DNA damage response and cell senescence was also examined in EC isolated from WT and Id1 KO mice (Fig. 9). H 2 O 2 (50 lM) treatment resulted in increased cH2AX expression (Fig. 9A), a marker of DNA damage, in KO compared with WT EC. To determine the effect of Id1 KO on oxidative stress-induced cell senescence, MacroH2A.1.1 (Fig. 9B) and SABG expression (Fig. 9C) (Fig. 9D). To determine the effect of decreased Id1 levels on oxidative stress-induced p53 levels, the MyEnd EC line was treated with LDN193189, a BMPr1/Smad 1/5/8 inhibitor followed by treatment with H 2 O 2 . Western analysis (Fig. 9E) demonstrated that decreased Id1 expression resulted in increased p53 expression with no additive effect with H 2 O 2 treatment. Treatment with SB431542, an Alk5, Smad 2/3 inhibitor, resulted in increased nuclear Id1 level and suppression of p53. Results from these in vitro assays correlate with activation of p53associated DNA damage response pathways in Id1 KO EC compared with WT EC by microarray analysis (Table S3).

Discussion
This study demonstrates that Id1 protects against hyperglycemia-induced endothelial injury and senescence. Endothelial injury in Id1 KO diabetic mice was shown to be associated with mediators of senescence and the senescence-associated inflammatory phenotype by microarray analysis, and in vivo and in vitro detection of senescence markers. EC injury resulted in nephropathy by decreased microvascular perfusion and increased matrix deposition.
Changes in vessel architecture in response to hyperglycemia including lumen narrowing and basement membrane thickening have been described in multiple organs including the kidney [30][31][32][33]. At a molecular level, cytoskeletal remodeling due to ECM modifications is a key mechanism [34]. Our results indicate that Id1 KO leads to significant decreases in capillary perfusion rather that rarefaction due to loss of EC by endothelial-mesenchymal transition or other mechanisms. We initially hypothesized that endothelial Id1 KO would result in EndMT due to unopposed TGFb and possible sensitization to BMP due to ineffective Smad 1/5/8 signaling as previously demonstrated in Id knockdown epithelial cells [35]. Unlike a previous

study [7]
, we detected very few capillaries or interstitial cells (< 1%) that colabeled with CD31 and aSMA, suggesting this was not a mechanism of endothelial injury. Microarray analysis in this and other studies and histological results suggest that the observed perfusion defects may be due to endothelial cytoskeletal activation and changes in matrix including basement membrane thickening and fibronectin secretion. EM analysis demonstrated marked narrowing of peritubular and glomerular capillary lumens associated with enlarged EC cytoplasm that may contribute to the observed hypoperfusion.
Premature senescence in response to hyperglycemia and other forms of oxidative stress has predominately been studied in cell culture. In addition to irreversible cell cycle arrest, senescence is characterized by morphological changes, persistent DNA damage response, and senescence-associated secretory phenotype, an inflammatory response that is regulated at the transcriptional level by NF-jB [36,37]. Microarray analysis showed a significant increase in gene expression of the NF-jB pathway and interferon-and interleukin-regulated genes in Id1 KO EC. Senescence-associated inflammation contributes to tissue damage and fibrosis in both disease and aging, a mechanism supported by studies showing that deletion of senescent cells in a mouse model of premature aging resulted in reduction of aging-associated phenotypes [38] and reduced glomerulosclerosis in normal aging [39]. Currently, there is no definitive evidence of in vivo EC senescence with kidney aging or injury. Identification of senescent cells, including EC, in vivo is technically challenging due to the lack of reliable markers. X-gal staining for SABG expression has been used to identify senescent EC in atherosclerotic arteries [40] but this technique lacks sensitivity for EC staining in kidney and other tissue sections. Studies have therefore relied upon examining the effects of genetic manipulation of key senescence mediators such as p16INK4a in models of aging and tissue injury [41]. Our study uses a combination of previously characterized changes in microarray gene expression, identification of X-gal crystals using a more sensitive electron microscopy technique [39], and expression of the senescence-associated heterochromatin marker MacroH2A.1.1. that functions upstream of ATM and is critical for persistent DDR and the inflammatory phenotype during senescence [27].
Id1 downregulation in senescent EC has previously been demonstrated in microarray studies [42]. In contrast, induced Id1 expression inhibits senescence [13]. Inhibition of cell senescence by Id1 through repression of CDKN2A (p16INK4a) has been demonstrated in vitro in numerous cell types including EC [13,43]. ETS2, a transcriptional activator of CDKN2A (p16INK4a), is directly antagonized by Id1 [44]. Our microarray results showed fourfold to fivefold increases in ETS1 and 2 in KO EC. Although we did not detect increased CDKN2A levels, increases in CDKN2d (p19INK4d), CDKN2Aip, and CDKN1b (p27Kip1) were demonstrated. CDKN2Aip can bind p53 directly and induces cellular senescence through multiple pathways [45,46] along with these other cell cycle inhibitors [47]. In vitro studies have also identified mechanisms by which Id1 is downregulated with senescence. Id1 expression is decreased by DEC1, an effector of p53 that is significantly increased with cell senescence [48] and by Smurf2-induced polyubiquitination [49]. In addition to inhibition of p53 and p16INK4a expression, other mechanisms by which Id1 inhibits senescence may exist. EC-specific KO of SIRT1, an inhibitor of cell senescence, resulted in increased kidney fibrosis in a folic acid injury model [23]. Interestingly, microarrays showed that SIRT1 was increased only in WT diabetic EC, suggesting a protective mechanism absent in KO cells. Id1 KO mice used in this study were shown to have increased EC Id4 expression by microarray analysis. The significance of this increased expression is uncertain since Id4 protein expression has only been reported in the CNS and tumor cells [50,51]. A compensatory increase due to Id1 KO, however, cannot be ruled out.
Our microarray results showed a sevenfold larger number of upregulated genes in Id1 KO EC from nondiabetic compared with diabetic mice in comparison with WT cells. Moreover, increased senescence-associated gene expression compared with WT cells was predominately found in these cells and not Id1 KO EC from diabetic mice. This finding correlates with the absence of MacroH2A1.1 expression in WT control cells and relative lower changes in expression in KO compared with WT diabetic kidneys. In addition to the relative increased expression of these genes in WT EC from diabetic mice, this discrepancy may also be due to the small sample size and increased heterogeneity of the diabetic EC. Although young Id1 KO mouse kidneys are phenotypically normal, increased expression of senescence-associated genes in these mice may only lead to pathology following cell stress such as hyperglycemia. Low-level oxidative stress has been shown to result in ROS generation necessary for propagation of downstream mitogenic and antiapoptotic signals or premature senescence, while higher level stress results in apoptosis. Levels of stress-induced p53 were mediated by the DNA damage response pathway having a major role in determining which of these final outcomes occur. Previous studies have shown that Id1 and p53 inversely regulate each other, suggesting a mechanism by which Id1 regulates the DNA damage response. To explore this possibility, we used lung EC isolated from both WT and Id1 KO mice and pharmacological treatment of a microvascular EC line to demonstrate that hyperglycemia results in increased ROS in KO cells and that low-level oxidative stress in KO cells and cells with pharmacological decreases in Id1 expression results in an increased DNA damage response, p53 expression, and markers of cell senescence. Lung EC were used in this study since we and others have found that EC isolated from kidney are very sensitive to the isolation procedure and standard cell culture conditions [52]. Isolated cells undergo myofibroblast-like phenotypic changes within 1-2 days of culture or are 'contaminated' with pericytes that are attached to the EC during the purification process. These problems are not seen with isolated lung EC. While differences in gene expression and paracrine growth factor requirements among EC isolated from different organs are noted [53], our experiments focused on response to in vitro oxidative stress only. There is no literature to suggest that this response differs among EC isolated from different organs.
In summary, Id1 expression is increased in EC in stz-induced diabetic mice. Id1 KO results in hyperglycemia-induced EC injury leading to decreased microvascular perfusion and characteristic diabetic pathology. Microarray analysis suggests that Id1 KO mice develop premature EC senescence and activation of inflammatory pathways, an outcome supported by increased senescence and fibrotic marker expression.
In vitro experiments demonstrate that Id1 has a protective function against hyperglycemia-induced ROS and oxidative stress-induced DNA damage that may be mediated by regulation of p53 levels. EC Id1  levels are increased by pharmacological agonists of BMPr1 and PPARa receptors, suggesting that these drugs may be useful in preventing microvascular injury with diabetes through this mechanism.
Diabetes was induced by intraperitoneal stz (Sigma, St. Louis, MO, USA) injection (125 mgÁkg À1 ) for two consecutive days as previously described [54]. Control mice were injected with equal volumes of 0.1 M Na citrate buffer. Blood glucose was monitored weekly. Mice with glucose levels between 300 and 500 were included in the study. All mice were housed in accordance with guidelines from American Association for Laboratory Animal Care. Research protocols and procedures were evaluated and approved by the CAVHS Institutional Animal Care and Use Committee and followed accepted guidelines for laboratory mouse welfare.

Western blot
Protein samples were separated by electrophoresis on 5-20% Tris-glycine gels (Novex, Invitrogen/Thermo Fisher, Waltham, MA, USA). Proteins were transferred to nitrocellulose membranes (Bio-Rad, Hercules, CA, USa) using an XCell miniblot system (Invitrogen). Membranes were blocked with TBST w/5% NFDM for 1 h and incubated overnight at 4°C with primary antibodies. Proteins were detected using HRP-conjugated anti-rabbit secondary antibodies at 1 : 2000 dilution (Cell Signaling #7074, Danvers, MA, USA) and SignalFire TM ECL reagent (Cell Signaling). For quantification of protein levels, autoradiographs were scanned, and densitometry was performed with NIH IMAGEJ software (Bethesda, MD, USA). Results were corrected for variations in the amount of protein loaded on each lane using corresponding GAPDH or actin levels.

Fluorescence microangiography
Mice (n = 3/group) were anesthetized with isoflurane and perfused with FluoSpheres sulfate 0.02 lM yellow-green microspheres (Thermo Fisher) conjugated to low melting point agarose (Lonza, Allendale, NJ, USA) exactly as detailed in the protocol by Kramann et al [17]. Seven-micrometer cryosections were double-labeled with rat anti-mouse CD31 antibody (clone MEC 13.3; BD Biosciences) and detected with Alexa Fluor 647-conjugated goat anti-rat IgG (Cell Signaling Technology #4418). Vascular area and lacunarity were measured in separate fluorescence channels for FluoSpheres and CD31 using ANGIOTOOL software (National Cancer Institute, Bethesda, MD, USA). Fluorescence images were obtained with a Zeiss Axio Imager microscope (White Plains, NY, USA).

Gene expression microarrays
Mouse kidney EC were isolated by Angiocrine Bioscience (New York, NY, USA). Control and 3-month post-stz-treated diabetic WT mice and Id1 À/À mice (n = 3-4/group) were intravitally labeled with VE-Cadherin Alexa Fluor 647 (Novus Bio, clone BV13). Following cardiac perfusion with M199 media, kidneys were minced and digested with collagenase, dispase, and DNase. Dissociated tissues were filtered, and EC were isolated with a FACSJazz cell sorter. FSC and SSC gates were used to identify the cells, and doublets were excluded, resulting in > 95% pure cells as previously described [53]. Cells were sorted directly into TRIzol LE and kept frozen at À80°C until further use.
Microarrays were performed by Phalanx Biotechnology (San Diego, CA, USA). The RNA concentration and purity were checked by OD260/OD280 (≧ 1.8) and OD260/ OD230 (≧ 0.7), and the yield and quality were accessed using an Agilent 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA, USA). Fluorescent aRNA targets were prepared from 0.1 lg total RNA samples using Low Input Quick Amp Labeling Kits (Agilent Technologies). Fluorescent targets were hybridized to Agilent Mouse 8x60K v2 Microarrays (Agilent Technologies). After 16 h of hybridization at 65°C, 10 r.p.m., nonspecific binding targets were washed away with two different washing steps (Wash I 25°C 1 min; Wash II 37°C 1 min), and the slides were scanned using an Agilent G2505C DNA Microarray Scanner (Agilent Technologies). The intensities of each probe were determined using FEATURE EXTRACTION 10.7.3.1 software (Molecular Devices, San Jose, CA, USA). The raw intensity of each spot was loaded into R version 2.12.1 to analyze the data. Control probes were first removed before processing the data. Gene clustering, using PCA (principal components analysis) and clustering analysis, was performed on selected gene lists (chosen using the average linkage algorithm) after data transformation and mean centering. The probes with log 2 ratio ≧ 1 or log 2 ratio ≦ À1 and P value < 0.05 via t-test were defined as differentially expressed genes, and these differentially expressed genes were further used for pathway enrichment analysis.

Urine creatinine and albumin assays
Urine albumin was measured by ELISA (Exocell, Philadelphia, PA, USA) and normalized to urine creatinine measured by Jaffe reaction (Creatinine Companion, Pointe Scientific, Canton, MI, USA).

ROS assay
Cells were grown to 70% confluence on 1% gelatin-coated dishes in low-glucose (4.5 mM) medium, following treatment with either high glucose (25 mM), menadione (50 lM; Sigma), or high glucose and Nox inhibitor VAS2870 (10 lM; EMD Millipore, Taunton, MA, USA) for 2 h. ROS and cell viability were assayed using a Cell ROX Deep Red Flow Cytometry Kit (Thermo Fisher) and analyzed with a BD LSRFortessa TM flow cytometer.

SA-b-galactosidase staining
Cell cultures and tissue slices were fixed for 3 and 25 min, respectively, and stained overnight at 37°C using a commercially available kit (Cell Signaling Technologies). Tissue slices were subsequently postfixed with 4% paraformaldehyde followed by 30% sucrose for cryosections and 4% paraformaldehyde/0.2% glutaraldehyde for EM.

Quantitative PCR
RNA was isolated from cell cultures using a ReliaPrep Kit (Promega, Madison, WI) with DNAse treatment. cDNA was prepared with a High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems, Foster City, CA, USA), and comparative qPCR was performed with TaqMan p16 and 18S (endogenous control) primers and a TaqMan Gene Expression Master Mix Assay (Applied Biosystems). Relative quantification (RQ) values were calculated with WT control EC as a reference sample using Applied Biosystems SDS software.

Statistical analysis
Data were collected from independent experiments and presented as mean AE SD. An unpaired two-tailed Student's ttest was used to compare two series of data. One-way ANOVA for independent samples with a Tukey HSD test for comparison between two groups (www.vassarstats.net/ anova1u.html) was used to compare three or more experimental groups. A P-value < 0.05 was considered significant. Microarray pathways were analyzed with the Fisher's exact test using INGENUITY software. heterogeneity in organ maintenance and regeneration. Dev Cell 26, 204-219. 54 Tesch GH and Allen TJ (2007) Rodent models of streptozotocin-induced diabetic nephropathy. Nephrology 12, 261-266.

Supporting information
Additional supporting information may be found online in the Supporting Information section at the end of the article.