MicroRNA-192 in diabetic kidney glomeruli and itsfunction in TGF-�-induced collagen expressionvia inhibition of E-box repressorsMitsuo Kato*, Jane Zhang†‡, Mei Wang*, Linda Lanting*, Hang Yuan*, John J. Rossi†‡, and Rama Natarajan*‡§

*Gonda Diabetes Center, †Division of Molecular Biology, ‡Graduate School of Biological Sciences, Beckman Research Institute of the City of Hope,Duarte, CA 91010

Communicated by Arthur D. Riggs, Beckman Research Institute of the City of Hope, Duarte, CA, December 18, 2006 (received for review November 30, 2006)

Key features of diabetic nephropathy (DN) include the accumula-tion of extracellular matrix proteins such as collagen 1-� 1 and -2(Col1a1 and -2). Transforming growth factor �1 (TGF-�), a keyregulator of these extracellular matrix genes, is increased in mes-angial cells (MC) in DN. By microarray profiling, we noted thatTGF-� increased Col1a2 mRNA in mouse MC (MMC) but alsodecreased mRNA levels of an E-box repressor, �EF1. TGF-� treat-ment or short hairpin RNAs targeting �EF1 increased enhanceractivity of upstream E-box elements in the Col1a2 gene. TGF-� alsodecreased the expression of Smad-interacting protein 1 (SIP1),another E-box repressor similar to �EF1. Interestingly, we notedthat SIP1 is a target of microRNA-192 (miR-192), a key miR highlyexpressed in the kidney. miR-192 levels also were increased byTGF-� in MMC. TGF-� treatment or transfection with miR-192decreased endogenous SIP1 expression as well as reporter activityof a SIP1 3� UTR-containing luciferase construct in MMC. Con-versely, a miR-192 inhibitor enhanced the luciferase activity, con-firming SIP1 to be a miR-192 target. Furthermore, miR-192 syner-gized with �EF1 short hairpin RNAs to increase Col1a2 E-box-lucactivity. Importantly, the in vivo relevance was noted by theobservation that miR-192 levels were enhanced significantly inglomeruli isolated from streptozotocin-injected diabetic mice aswell as diabetic db/db mice relative to corresponding nondiabeticcontrols, in parallel with increased TGF-� and Col1a2 levels.These results uncover a role for miRs in the kidney and DN incontrolling TGF-�-induced Col1a2 expression by down-regulatingE-box repressors.

diabetic nephropathy � mesangial cells � small noncoding RNA �transforming growth factor �1

D iabetic nephropathy (DN) is the most common cause ofkidney failure in patients with diabetes mellitus. The major

characteristics of DN include glomerular basement-membranethickening, mesangial expansion and hypertrophy, and an accu-mulation of extracellular matrix (ECM) proteins (1). Evidenceshows that transforming growth factor �1 (TGF-�) levels areincreased under diabetic conditions in renal cells, includingmesangial cells (MC), can up-regulate ECM proteins such ascollagens (2, 3), and also can promote MC survival and oxidantstress (4).

To date, Smad transcription factors have been shown to be themajor effectors of TGF-� signaling (5, 6). Collagen 1-� 1 and -2(Col1a1 and -2) and other ECM genes are regulated in MC byTGF-� via Smads (7, 8). The regulation of collagen by TGF-� inMC also is mediated by mitogen-activated protein kinases(MAPKs) such as p38 and ERKs (9–11). However, the molec-ular mechanisms by which TGF-� regulates ECM genes still arenot understood fully. The collagen gene has E-box elements inthe far upstream enhancer region (12, 13). An E-box repressor,�EF1, is a key inhibitor of E-cadherin (14) and E2-box tran-scription factors such as Nkx2.5 (12). Moreover, it is a knownrepressor of collagen type 1 and type 2 genes in other cells (12,13), but its role in MC is not known.

During microarray profiling, we observed that TGF-� leads toa marked down-regulation of �EF1 expression in MC. In thisarticle, we demonstrate that this effect can lead to increasedcollagen expression via relief of repression at the E-box elementsin the collagen gene. Furthermore, we found that TGF-� alsodown-regulates the expression of Smad-interacting protein 1(SIP1), another E-box repressor belonging to the same family as�EF1 (15).

MicroRNAs (miRs) are short noncoding RNAs of �22 nt thatrecently have been shown to play important roles in mammaliangene expression (16–19). They induce posttranscriptional generepression by blocking protein translation (by binding to the 3�UTR of their target genes) or by inducing mRNA degradationand therefore have the potential to play central roles in physi-ological and pathological conditions. The field of miRs now isexploding with researchers racing to discover their gene targetsand disease relevance. Recent evidence suggests that miRs mayregulate the expression of key genes relevant to cancer andpotentially other diseases (17, 19). However, to date nothing isknown regarding miRs in the context of diabetic complicationssuch as DN. A recent report showed that at least 5 miRs(miR-192, -194, -204, -215, and -216) are highly and quiteexclusively expressed in the kidney (20), but their functions ortargets are not yet known. In this article, we identify SIP1 as atarget of miR-192. We also demonstrate a previously uncharac-terized mechanism for TGF-�-mediated collagen regulationinvolving a cross-talk between E-box repressors (�EF1 and SIP1)and miRs (miR-192) that could be relevant to the pathogenesisof diseases such as DN.

Results and DiscussionDown-Regulation of �EF1 by TGF-� in MMC. To uncover previouslyuncharacterized pathways regulated by TGF-�, we profiledgenes regulated by TGF-� in mouse MC (MMC) by using TGF-�BMP signaling pathway microarrays (Fig. 1A). Although, asanticipated, TGF-� could increase the expression of key genesassociated with the pathogenesis of DN, such as Col1a1, Col1a2,and plasminogen activator inhibitor 1, several other previouslyunidentified genes with varying expression levels also were

Author contributions: M.K. and R.N. designed research; M.K., J.Z., M.W., L.L., and H.Y.performed research; M.K., J.Z., M.W., and J.J.R. contributed new reagents/analytic tools;M.K., J.Z., J.J.R., and R.N. analyzed data; and M.K. and R.N. wrote the paper.

The authors declare no conflict of interest.

Abbreviations: DN, diabetic nephropathy; MC, mesangial cells; MMC, mouse MC; SIP1,Smad-interacting protein 1; miR, microRNA; ECM, extracellular matrix; Col1a1 and -2,collagen 1-� 1 and 2; shRNA, short hairpin RNA; qPCR, quantitative PCR; AS, antisense; STZ,streptozotocin.

§To whom correspondence should be addressed at: Gonda Diabetes Center, BeckmanResearch Institute of the City of Hope, 1500 East Duarte Road, Duarte, CA 91010. E-mail:rnatarajan@coh.org.

This article contains supporting information online at www.pnas.org/cgi/content/full/0611192104/DC1.

© 2007 by The National Academy of Sciences of the USA

3432–3437 � PNAS � February 27, 2007 � vol. 104 � no. 9 www.pnas.org�cgi�doi�10.1073�pnas.0611192104

detected (Fig. 1 A and B). One of the target genes down-regulated by TGF-� in the microarray was �EF1 (Fig. 1 A, C, andD), a repressor of E-box transcription factors (21). Follow-upverification of the microarray data demonstrated that serumdepletion decreased Col1a2 expression but increased �EF1mRNA levels, and these events were reversed by TGF-� treat-ment (Fig. 1 B and C). Significant decrease in �EF1 mRNA levelsoccurred 6 h after TGF-� treatment (Fig. 1C), whereas asignificant increase in Col1a2 was observed by 24 h (Fig. 1B).Similar to �EF1 mRNA, �EF1 protein levels also were differ-entially regulated by serum depletion and TGF-� (Fig. 1D).Because E-boxes are located in the far upstream enhancer regionof collagen genes (12, 13), these results suggest that TGF-� mayinduce collagen expression by decreasing �EF1 expression andthus relieving �EF1-mediated repression at the E-box elements.

�EF1 Negatively Regulates E-Box Enhancer of Col1a2 Gene. We nextperformed chromatin immunoprecipitation (ChIP) analysis andconfirmed that �EF1 occupancy at the E-boxes in the upstreamenhancer region of Col1a2 gene [schematically shown in Fig. 2A,details provided in supporting information (SI) Fig. 6] is mark-edly enhanced in serum-depleted MMC and is reversed byTGF-� (Fig. 2B). TGF-� treatment significantly decreased �EF1occupancy at the collagen enhancer in serum-depleted cells evenby 6 h, and this effect was sustained up to 24 h.

We cloned the upstream Col1a2 enhancer E-box elements intoa luciferase reporter vector and noted that TGF-� significantlyincreased luciferase activity of this reporter in MMC (Fig. 2C).

To determine whether this effect of TGF-� is caused by itsdown-regulation of �EF1, we next examined whether shorthairpin RNA (shRNA)-mediated knockdown of �EF1 couldmimic the effects of TGF-�. We observed significant decrease ofendogenous �EF1 by transfection with �EF1 shRNA (90% of

Fig. 1. Profiling genes regulated by TGF-� in MMC. (A) Representative OligoGEArray profiling results. Intensities of several spots were up-regulated ordown-regulated by treating serum-depleted MMC with TGF-� (10 ng/ml for24 h) (Top). Close-up magnified view of regions containing Col1a2 and �EF1genes are shown in Middle and Bottom and demonstrate that TGF-�-treatedMMC have marked increase in the intensities of Col1a2 spots but decrease in�EF1 spots (each region has four replicate spots). (B and C) Verification ofCol1a2 and �EF1 expression levels by real-time qPCR. Col1a2 gene was de-creased by serum depletion (SD) relative to cells growing in serum (NT, notreatment), whereas TGF-� treatment significantly increased its expression by24 h but not at 6 h (B). Conversely, �EF1 expression was significantly increasedby serum depletion, and this was reversed by TGF-�, which down-regulated�EF-1 by 6 and 24 h. *, P � 0.05; #, P � 0.01. TGF6, TGF-� for 6 h; TGF24, TGF-�for 24 h. (D) Western blot analysis of �EF1 protein levels. Similar to the mRNAdata, serum depletion increased �EF1 protein expression, whereas TGF-� (24h) reversed (decreased) the expression.

Fig. 2. �EF1 negatively regulates the enhancer region of Co1a2 gene. (A)Schematic diagram of the far upstream region of Col1a2 gene. Three E-boxesare noted in the enhancer region located 16-kb upstream of the Col1a2 genetranscription start site. This region was amplified by PCR using primers indi-cated by arrows. Details of the enhancer sequence are shown in SI Fig. 6. (B)ChIP assay to evaluate �EF1 occupancy at the enhancer region containingE-boxes in the Col1a2 gene. After IP with �EF1 antibody, target ChIPed DNAamplification (with primers shown in A) was noted only in serum-depletedMMC (Upper). Preimmune IgG served as negative control. DNA without IP wasused as input control. Serum depletion increased �EF1 expression and itsCol1a2 gene enhancer occupancy. Bar graph (Lower) shows �EF1 bindingquantified by real-time qPCR. Relative amount was calculated by the ratio ofChIPed DNA/input control DNA. A 6-fold increase in �EF1 occupancy wasobserved in serum-depleted cells (SD) compared with serum-treated controls(NT). TGF-� treatment gradually decreased the amount of �EF1 at the pro-moter from 6 to 24 h (TGF6 and TGF24). *, P � 0.05; #, P � 0.01. (C) The sameenhancer DNA fragment with E-boxes was cloned into the pGL3-promoterdriven by SV40 promoter. TGF-� treatment increased luciferase activity (*, P �0.05). (D) �EF1 shRNA increased transcription from the Col1a2 promoter E-boxelements. �EF1 shRNA cotransfected with the reporter construct pGL3-promoter-Col1a2-Ebox caused a significant increase in luciferase activity (*,P � 0.05). (E) Down-regulation of SIP1 in response to TGF-�. SIP1 mRNAexpression levels were quantified by real-time qPCR. Similar to �EF1, SIP1expression was increased by serum depletion (SD), but TGF-� treatmentreversed this by decreasing expression from 6 h posttreatment up to 24 h (#,P � 0.01). (F) Scheme showing that TGF-� inhibition of �EF1, a repressor of theE-box in the Col1a2 promoter, can increase collagen gene expression. Othernegative factors or E-box repressors (SIP1) also may cooperate to regulate theenhancer elements in the Col1a2 gene.

Kato et al. PNAS � February 27, 2007 � vol. 104 � no. 9 � 3433

MED

ICA

LSC

IEN

CES

mRNA and 70% of protein) in 48 h (SI Fig. 7). �EF1 shRNAsignificantly increased luciferase activity of the vector containingCol1a2 gene E-box elements (Fig. 2D) in MMC, similar to theeffect of TGF-� (Fig. 2C). This observation further supports ourhypothesis that �EF1 is a negative regulator of the Col1a2 geneenhancer and that TGF-�-mediated down-regulation of �EF1 isa key mechanism for up-regulating the collagen gene. Theseresults suggest that, apart from the activation of positive factorsSmads and MAPKs during collagen expression, TGF-�-mediated repression of the negative regulator �EF1 may be apreviously uncharacterized mechanism by which TGF-� in-creases collagen expression in MCs under conditions such asdiabetes associated with enhanced TGF-� levels. Furthermore,�EF1 also can bind to Smad proteins (22) and act as a repressorassociated with CtBP (22, 23). Thus, the down-regulation of�EF1 by TGF-� also may activate the Smad pathway. �EF1therefore may affect not only the far upstream E-box enhancerbut also Smad-responsive elements (7) present in the proximalpromoter of the Col1a2 gene.

SIP1, Another E-Box Repressor, also Is Negatively Regulated by TGF-�.Interestingly, the induction of the luciferase activity by �EF1shRNA, although significant, was relatively modest (Fig. 2D).We therefore hypothesized that other negative factors mayassociate with E-box elements in the Col1a2 gene and cooperatewith �EF1. One potential candidate is a gene belonging to thesame family as �EF1, namely, SIP1, an E-box repressor (alsocalled ZEB2 or Zfhx1b) and closely related to �EF1 (also calledZEB1 or Zfhx1a) (15). Both �EF1 and SIP1 can interact with thecorepressor CtBP (22, 23). We therefore tested whether TGF-�can regulate SIP1 in a manner similar to �EF1. Fig. 2E confirmsthat serum depletion of MMC significantly induced SIP1 ex-pression, and TGF-� treatment reversed this in the same patternas with �EF1 (Fig. 1C). These results support the concept thatSIP1 may indeed be a coregulator of E-box elements along with�EF1 in the Col1a2 enhancer (Fig. 2F).

SIP1 3� UTR Is a Potential Target of Kidney-Specific miR-192 andmiR-215. Importantly, further support for SIP1 as a candidate inthis context was obtained by our observation that the 3� UTR ofthe SIP1 gene is a potential target site of two miRs (miR-192 andmiR-215) that were shown to be highly expressed in the kidney(20). Thus, by computational miR target predictions from miRdatabases [Memorial Sloan–Kettering Cancer Center, NewYork, NY (http://cbio.mskcc.org) and MIRANDA and Tar-getScan at Wellcome Trust Sanger Institute, Cambridge, U.K.(http://microrna.sanger.ac.uk/index.shtml)], we determined thatSIP1 is a potential target of miR-192 and miR-215 because asequence in its 3� UTR has a perfect match to the seed regionof these miRs (SI Fig. 8).

miR-192 Is Expressed in MMC and Up-Regulated by TGF-�. To test thepotential functions of the renal miR-192 and miR-215 in regu-lating SIP1 expression, we first examined the expression of thesemiRs in mouse kidney tissues and MMC by a real-time quanti-tative PCR (qPCR) method (24). We first examined miR-194because it is highly expressed in the kidney and liver (20, 25) andconfirmed that the qPCR expression pattern was similar to thatfound in Northern blot analyses (high levels in liver and kidneybut not in spleen tissues) and thus verified the reliability of ourqPCR method to monitor miR expression (Fig. 3A). (Dissocia-tion curves and amplification curves of miR-194 standards andkidney samples are shown in SI Figs. 9 and 10.) We nextexamined the expression of miR-192 and miR-215. miR-192 washighly expressed in spleen, kidney, and MMC and relatively lessexpressed in liver (Fig. 3B). miR-215 was expressed almostexclusively in the kidney (Fig. 3C). miR-192 was the only miRthat was expressed in MMC (Fig. 3B). Interestingly, the expres-

sion of miR-192 was decreased by serum depletion and increasedby TGF-� (Fig. 3E), a pattern that clearly is opposite to that ofits target SIP1 regulation (Fig. 2E). Both miR-192 and miR-215have similar sequences and are likely to target the same genes.However, unlike miR-192, the expression of miR-215 was in-duced by serum depletion and decreased by TGF-� (Fig. 3F).miR-194 was up-regulated by serum depletion but remainedunchanged by TGF-� treatment (Fig. 3D). These patterns aredifferent from that of miR-192. Therefore, SIP1 is most likelyregulated by miR-192 but not miR-215 or miR-194.

The molecular mechanisms by which miR-192 is induced byTGF-� needs further investigation. However, a highly conservedupstream ets-1 protooncogene binding site has been reported inthe miR-192 promoter (20). In response to serum, ets-1 forms aternary complex with serum response factor. TGF-� also inducesets-1 expression (26). Our data showing that serum depletiondecreased whereas TGF-� increased miR-192 expression (Fig.3E) suggests that this regulation may be mediated by the ets-1site in the miR-192 promoter. ets-1 also is essential for thedevelopment of kidneys, integrity of glomeruli, and expression ofmatrix metalloproteinase (27). Thus, miR-192 may be induced byTGF-� by means of ets-1 under diabetic conditions.

SIP1 3� UTR Is a Bona Fide Target of miR-192 but Not miR-215. In thenext step, we performed experiments to confirm that SIP1 isindeed a target of miR-192. The sequence of the miR targetregion in the SIP1 3� UTR is conserved among human, mouse,and rat (Fig. 4A and SI Fig. 8). We constructed a reporter vectorcontaining the SIP1 3� UTR immediately downstream of theluciferase gene but upstream of poly(A) signal (Fig. 4 A and B).

Fig. 3. Expression of miRs in mouse tissues and MMC and regulation byTGF-�. (A) Verification of miR-194 expression in mouse tissues by using real-time qPCR. miR-194 expression was high in mouse liver and kidney tissue butundetected in the spleen. (B) miR-192 was highly expressed in mouse kidney,spleen, and MMC, with lower levels in liver. (C) miR-215 was expressedexclusively in the kidney with very little in other tissues or MMC. (D) miR-194expression was very low in MMC, but serum depletion and TGF-� treatmentboth increased its expression. (E) Serum depletion decreased miR-192 in MMC,whereas TGF-� increased miR-192 expression from 6 to 24 h (*, P � 0.05). (F)Serum depletion increased miR-215 expression, which was reversed by TGF-�.

3434 � www.pnas.org�cgi�doi�10.1073�pnas.0611192104 Kato et al.

Double-stranded RNA oligonucleotides (dsRNA) mimickingmature miR-192 (Mimic) were cotransfected with this reporterin MMC. miR-192 Mimic significantly inhibited SIP1 3� UTRvector luciferase activity (by �90% relative to that with negativecontrol Mimic) but not in the two control constructs, namely,control without 3� UTR (Fig. 4B, first pair of bars) or that withthe SIP1 3� UTR in the antisense (AS) orientation (Fig. 4B, third

pair of bars). These results confirmed that the SIP1 3� UTRsequence is recognized by miR-192 and that SIP1 is a true targetof miR-192. On the other hand, miR-215 Mimic could notdecrease luciferase activity of this same SIP1 sense 3� UTRconstruct (SI Fig. 11). These results, coupled with earlier data(Fig. 3 E and F) showing a different expression pattern ofmiR-215 relative to miR-192 in MMC treated with TGF-�,indicate that miR-192, but not miR-215, recognizes the 3� UTRsequence in the SIP1 gene and thereby can decrease SIP1expression. Furthermore, because TGF-� increased miR-192expression (Fig. 3E), we tested whether TGF-� can decreaseluciferase activity of the SIP1 3� UTR reporter. Fig. 4C confirmsthat this is indeed the case. TGF-� had no effect in control or ASvector-treated cells.

Because miR-192 Mimic could inhibit SIP1 3� UTR luciferasereporter activity, we next examined whether a miR-192 inhibitorcan have the opposite effect in this system (Fig. 4D). A 5 nMconcentration of miR-192 inhibitor (2�-O-methyl-modified RNAoligonucleotides designed AS to the mature miR) demonstrateda 20-fold increase of luciferase activity that peaked at 10 nM.These results suggest that the miR-192 inhibitor, by arrestingmiR-192 function, can increase reporter activity. The increase inreporter activity induced by miR-192 inhibitor also was inhibitedby TGF-� (Fig. 4E), suggesting that TGF-�-mediated increase inmiR-192 expression can compete with and overwhelm the effectsof the miR-192 inhibitor. These results further confirm that, inresponse to TGF-�, miR-192 can negatively regulate SIP1 geneby targeting its 3� UTR sequence.

We then confirmed that exogenous miR-192 (Mimic) canreduce endogenous SIP1 mRNA expression (Fig. 4F), suggestingthat miR-192 reduces the expression of its target SIP1 most likelyvia mRNA degradation. Mammalian miRs generally function bybinding to their complementary sequences in the 3� UTRs oftarget genes, thereby inhibiting translation of these targets.Although these miRs reduce protein expression (translationalinhibition), it now is evident that this also can occur by mRNAdegradation and destabilization (28, 29). Recent investigationsindicate that miRs can recruit their target mRNAs to P-bodies(that concentrate enzymes involved in mRNA degradation) andinhibit protein translation by preventing ribosomal initiation orby slow degradation of mRNAs (30). One report revealed thatmiR-125b promotes mRNA deadenylation and enhanced deg-radation of its target (31). Thus, miRs can reduce target proteinor mRNA expression or both. These reports support our datademonstrating that miR-192 decreases SIP1 mRNA levels andthat SIP1 mRNA reduction by TGF-� coincides with an increasein miR-192 expression.

Cooperation of miR-192 and �EF1 shRNA in Up-Regulating Col1a2. Thenext step was to correlate these observations to collagen regu-lation. Because miR-192 decreased SIP1 expression, the effect ofmiR-192 on Col1a2-Ebox-luciferase reporter activity in MMCwas examined. Fig. 4G shows that miR-192 alone slightly in-creased reporter activity of this Col1a2 upstream promoterconstruct. �EF1 shRNA caused a modest but significant increasein reporter activity in MMC, as shown earlier in Fig. 2D.However, interestingly, when miR-192 and �EF1 shRNA weretransfected together in MMC, there was a significant synergisticenhancement of reporter activity, which supports our hypothesisthat both SIP1 and �EF1 induce repression at this E-box site andthat a knockdown of both induces a much greater increase inCol1a2 promoter activation than either one alone (Fig. 4G).

Increased Expression of miR-192 in Diabetic Kidney Glomeruli. Im-portantly, we next examined the in vivo relevance in mousemodels of type 1 [streptozotocin (STZ)-injected] and type 2(db/db) diabetes. Sieved renal glomeruli from the diabetic micehad significantly higher levels of miR-192 (Fig. 5 C and F) in

Fig. 4. Regulation of SIP1 gene by miR-192 via the 3� UTR. (A) Alignment ofhsa-miR-192 and mmu-miR-192 with human SIP1 3� UTR and mouse SIP1 3� UTRbased on software from the Memorial Sloan–Kettering Cancer Center (http://cbio.mskcc.org). Seven nucleotides in the 5� region of miR-192 (human andmouse) contain a perfect match with the 3� UTR sequence of the human andmouse SIP1 genes. Numbers next to the SIP1 3� UTR sequence are relative tothe start sites of the 3� UTR. Sequences of these regions were 100% conservedin human, mouse, and rat (SI Fig. 8). (B) miR-192-mediated inhibition ofluciferase reporter activity in vector containing the SIP1 3� UTR sequence. TheSIP1 3� UTR luciferase reporter (0.5 �g) was cotransfected with either themiR-192 Mimic (�miR-192 Mimic) or the negative control mimic (NC Mimic) inMMC. A significant decrease in luciferase activity with miR-192 mimic wasobserved with the SIP1 3� UTR-S (sense) construct (*, P � 0.05) but no changewith the control constructs (without SIP1 3� UTR, control) or with SIP1 3� UTRAS construct (AS). (C) TGF-� treatment decreased the reporter activity ofLuc-SIP1 3� UTR construct (*, P � 0.05). No change in luciferase activity was seenwith control plasmid without 3� UTR or plasmid with AS SIP1 3� UTR. (D)miR-192 inhibitor increased luciferase activity of Luc-SIP1 3� UTR construct (#,P � 0.01). (E) TGF-� treatment reversed the induction of luciferase activity bythe miR-192 inhibitor (*, P � 0.05). MMC were transfected with either Luc-SIP13� UTR or miR-192 inhibitor (5 nM and 10 nM) and treated with TGF-�. (F)miR-192 decreased SIP1 expression. TCMK-1 cells were transfected with miR-192 Mimic and SIP1 mRNA expression quantified by real-time qPCR (*, P �0.05). (G) Cooperation between miR-192 and �EF1 shRNA at the E-box region.miR-192 alone slightly increased Col1a2-Ebox-luc activity. �EF1 shRNA causeda significant increase of reporter activity (*, P � 0.05). Cotransfection ofmiR-192 and �EF1 shRNA lead to significant enhancement of reporter activityover their individual contributions (#, P � 0.01).

Kato et al. PNAS � February 27, 2007 � vol. 104 � no. 9 � 3435

MED

ICA

LSC

IEN

CES

parallel with increased levels of TGF-� (Fig. 5 A and D) andCol1a2 (Fig. 5 B and E) relative to the respective control mice.As mentioned earlier, profibrotic TGF-� levels are increasedunder diabetic conditions in renal cells, including MC, and canup-regulate ECM proteins such as collagens (2, 3). Our resultsshowing parallel increases in TGF-�, collagen, and miR-192 indiabetic glomeruli suggest that TGF-�-induced miR-192 is re-sponsible, at least in part, for increased Col1a2 expression notonly in vitro in MC but also in vivo. Because miR-192 wasincreased in tissues from both type 1 and type 2 diabetic mice,hyperglycemia may be a common factor.

Together, we have discovered a mechanism for TGF-�-mediatedcollagen regulation that involves cross-talk between E-box repres-sors and miRs. Thus, TGF-� induced down-regulation of SIP1 (viamiR-192) and �EF1 (via as-yet-unknown mechanisms) can coop-erate to enhance Col1a2 expression via derepression at E-boxelements (Fig. 5G). This finding could be related to kidney dys-function under pathological conditions such as diabetes and alsotubulointerstitial disease, in general, where TGF-� levels are en-hanced. Our observations provide the first functional role for a miRexpressed in the kidney. Small noncoding RNAs and miRs such asmiR-192 and their inhibitors may be targets for diseases such as DNand other diabetic complications.

MethodsCell Culture. Primary MMC were isolated and cultured as describedin ref. 4. Mouse kidney epithelial TCMK-1 cells (gift from Y.Matsuno and A. D. Riggs, City of Hope National Medical Center)were cultured in DMEM (Irvine Scientific, Santa Ana, CA) sup-

plemented with 10% FBS. Recombinant human TGF-�1 was fromR & D Systems (Minneapolis, MN).

Mouse Models. All animal studies were conducted in accordancewith a protocol approved by the Institutional Animal CareCommittee. Induction of diabetes by STZ injections in C57BL/6mice was carried out as described in ref. 4. Mice were used 7weeks after the onset of diabetes. db/db mice and genetic controldb/� mice (10 weeks old) were obtained from The JacksonLaboratory (Bar Harbor, ME). Glomeruli were sieved fromrenal cortical tissue as described in ref. 4.

Microarray Experiments. Oligo GEArray Mouse TGF-� BMPSignaling Pathway Microarrays (OMM-035) were used for ex-pression profiling (SuperArray Bioscience Corporation, Freder-ick, MD) in conjunction with the TrueLabeling-AMP linearRNA amplification kit, according to manufacturer’s protocols.Expression profiles from array experiments were analyzed byusing GEArray expression analysis suite (SuperArray).

Real-Time qPCR. Real-time qPCR was performed by using SYBRGreen PCR Master Mix and the 7300 real-time PCR system(Applied Biosystems, Foster City, CA) as reported in ref. 4. PCRprimer sequences were Col1a2, 5�-CAGAACATCACCTAC-CACTGCAA-3� and 5�-TTCAACATCGTTGGAACCCTG-3�;�EF1, 5�-CATTTGATTGAGCACATGCG-3� and 5�-AGCG-GTGATTCATGTGTTGAG-3�; SIP1, 5�-CCCTTCTGCGA-CATAAATACGA-3� and 5�-TGTGATTCATGTGCTGC-GAGT-3�; and TGF-�, 5�-CAACGCCATCTATGAGA-AAACC-3� and 5�-AAGCCCTGTATTCCGTCTCC-3�.

Real-time qPCR also was used to detect the miRs, as reportedin ref. 24. In brief, 0.5 �g of total RNA was used for cDNAsynthesis primed by particular miR-specific primer with a tailsequence recognized by the universal primer. cDNA was am-plified by using the miR-specific reverse and universal primers.Amplification of a single fragment was confirmed by a dissoci-ation curve (SI Fig. 9), and good correlation between standardsand threshold-cycle values was observed (SI Fig. 10). Temper-ature for annealing was optimized depending on the amplifica-tion of true targets. 5S RNA served as internal control (Ambion,Austin, TX). Sequences of miR-specific primers and reverseprimers were miR-194, 5�-CATGATCAGCTGGGCCAA-GATCCACATGGAGT-3� and 5�-TGTAACAGCAACTCCA-3�; miR-192, 5�-CATGATCAGCTGGGCCAAGATGTCAAT-TCATA-3� and 5�-CTGACCTATGAATTG-3�; and miR-215,5�-CATGATCAGCTGGGCCAAGAGTCTGTCAATTC-3�and 5�-ATGACCTATGAATTG-3�.

Western Blot Analysis. Western blot analysis was performed asdescribed in ref. 4. Antibodies against �EF1 and actin were fromSanta Cruz Biotechnology (Santa Cruz, CA) and Cell Signaling(Beverly, MA), respectively.

ChIP Assays. ChIP assays were performed according to ourprotocols (32). Briefly, MMC were serum-depleted and treatedwith TGF-� for 24 h and then fixed with formaldehyde. Thecross-linked chromatin was sheared and immunoprecipitated byusing antibodies against �EF1 or IgG (negative control).‘‘ChIPed’’ DNA was purified and used as template for real-timeqPCR by using primers spanning E-box regions (Fig. 2 A). ChIPPCR primer sequences were 5�-CTAAATGCACAGTTCCTC-CATGTGTTTAG-3� and 5�-GGTGAAGTGTCCCAC-CAAATGGCACACC-3�. An aliquot of the formaldehyde-fixedDNA also was reverse-cross-linked and used as input control.Amplified PCR fragment was cloned into pCR3.1 (Invitrogen,Carlsbad, CA) to confirm the sequence (SI Fig. 6).

Fig. 5. Expression of TGF-�, Col1a2, and miR-192 levels in glomeruli frommouse models of type 1 and type 2 diabetes. Total RNAs from normal anddiabetic mouse kidney glomeruli were extracted, and the expression levels ofTGF-�1, Col1a2, and miR-192 examined by real-time qPCR. (A–C) RNA fromtype 1 diabetic model (streptozotocin-treated) and control mice. (D–F) RNAfrom type 2 diabetic model (db/db) and control (db/�) mice. Significantincrease (#, P � 0.01) of the expression levels of TGF-�1, Col1a2, and miR-192was observed in diabetic glomeruli relative to healthy control. (G) A previouslyuncharacterized mechanism for TGF-�-induced Col1a2 gene expression. Dia-betic condition increases TGF-� levels in kidney glomeruli and MC. TGF-�relieves repression at the E-box region of Col1a2 gene by reducing theexpression of two E-box repressors, �EF1 (by unknown mechanisms) and SIP1(by increasing miR-192 expression, possibly via Ets-1), to increase Col1a2 geneexpression.

3436 � www.pnas.org�cgi�doi�10.1073�pnas.0611192104 Kato et al.

Plasmids. To construct the Col12a-Ebox-luc plasmid, the samefragment used in the ChIP assay was excised from the pCR3.1vector and cloned into the pGL3-promoter (Promega, Madison,WI). For SIP1 3� UTR reporters, luciferase cDNA was excised frompGL3-basic (Promega) by digestion with HindIII and XbaI restric-tion enzymes. The fragment then was cloned into HindIII–XbaI sitein pcDNA3.1/Zeo(�) (Invitrogen) to yield pcDNA3.1/luc3 vector.The XbaI site-tagged potential target sequences in the 3� UTR ofSIP1 gene (sense, 5�-CTAGTTTTGCATTCATTTAATTTTAG-GTCAA-3�; AS, 5�-CTAGTTGACCTAAAATTAAATGAAT-GCAAAA-3�) were cloned into the XbaI site of pcDNA3.1/luc3.Plasmid containing sense SIP1 3� UTR sequence was used as areporter (pcDNA3.1/luc-SIP1 3�S), whereas the plasmid containingthe 3� UTR sequence in the AS orientation was used as a negativecontrol (pcDNA3.1/luc-SIP1 3�AS).

Luciferase Assays. MMC were transfected with luciferase reporterplasmids by using FuGENE 6 (Roche, Indianapolis, IN) andtreated with TGF-� as reported in ref. 4. For miR cotransfec-tions, XtremeGENE (Roche) was used according to the man-ufacturer’s protocols. After 24 h, luciferase activities weremeasured as described in ref. 4. For miR and shRNA experi-ments, cells were treated for 48 h.

shRNA Against �EF1. A U6 promoter-derived shRNA expressionvector targeting �EF1 was constructed by using our protocol(33). Briefly, U6 promoter region was amplified by PCR using5� U6 primer and �EF1 target reverse primer 5�-GGACTC-GAGAAAAAACCAGCAGACCAGACAGTATTACTAC-ACAAATAATACTGTCTGGTCTGCTGGCGGTGTTT-CGTCCTTTCC-3� (the target sequence is underlined). Theamplified fragment was cloned into pCR3.1EGFP (33). A plas-mid expressing scrambled shRNA (pCR3.1EGFP-Scr) was usedas control (33).

miRNAs. Oligonucleotides representing the miR Mimics, the nega-tive control 1 for Mimics, the miR-192 inhibitor (2�-O-methyl-modified RNA oligonucleotides designed AS to the mature miR),and the negative control 1 for the inhibitor all were obtained fromDharmacon (Lafayette, CO). Negative control sequences are basedon Caenorhabditis elegans miRNA cel-miR-67.

We thank all members of the R.N. laboratory for helpful discussions.This work was supported by grants from the National Institutes of Health(National Institute of Diabetes and Digestive and Kidney Diseases) andthe Juvenile Diabetes Research Foundation (to R.N.).

1. Ziyadeh FN (1993) Am J Kidney Dis 22:736–744.2. Reeves WB, Andreoli TE (2000) Proc Natl Acad Sci USA 97:7667–7669.3. Sharma K, Ziyadeh FN (1995) Diabetes 44:1139–1146.4. Kato M, Yuan H, Xu Z-G, Lanting L, Li S-L, Wang M, Hu MC-T, Reddy M,

Natarajan R (2006) J Am Soc Nephrol 17:3325–3335.5. Roberts AB, McCune BK, Sporn MB (1992) Kidney Int 41:557–559.6. Yang YC, Piek E, Zavadil J, Liang D, Xie D, Heyer J, Pavlidis P, Kucherlapati

R, Roberts AB, Bottinger EP (2003) Proc Natl Acad Sci USA 100:10269–10274.7. Poncelet AC, Schnaper HW (2001) J Biol Chem 276:6983–6992.8. Tsuchida K, Zhu Y, Siva S, Dunn SR, Sharma K (2003) Kidney Int 63:2000–

2009.9. Kim YS, Xu ZG, Reddy MA, Li SL, Lanting L, Sharma K, Adler SG, Natarajan

R (2005) J Am Soc Nephrol 16:352–362.10. Chin BY, Mohsenin A, Li SX, Choi AM, Choi ME (2001) Am J Physiol

280:F495–F504.11. Hayashida T, Poncelet AC, Hubchak SC, Schnaper HW (1999) Kidney Int

56:1710–1720.12. Ponticos M, Partridge T, Black CM, Abraham DJ, Bou-Gharios G (2004) Mol

Cell Biol 24:6151–6161.13. Murray D, Precht P, Balakir R, Horton WE, Jr (2000) J Biol Chem 275:3610–

3618.14. Eger A, Aigner K, Sonderegger S, Dampier B, Oehler S, Schreiber M, Berx G,

Cano A, Beug H, Foisner R (2005) Oncogene 24:2375–2385.15. Verschueren K, Remacle JE, Collart C, Kraft H, Baker BS, Tylzanowski P,

Nelles L, Wuytens G, Su MT, Bodmer R, et al. (1999) J Biol Chem 274:20489–20498.

16. Ambros V (2004) Nature 431:350–355.

17. He L, Hannon GJ (2004) Nat Rev Genet 5:522–531.18. Pasquinelli AE (2006) Dev Cell 10:419–424.19. Esquela-Kerscher A, Slack FJ (2006) Nat Rev Cancer 6:259–269.20. Sun Y, Koo S, White N, Peralta E, Esau C, Dean NM, Perera RJ (2004) Nucleic

Acids Res 32:e188.21. Sekido R, Murai K, Funahashi J, Kamachi Y, Fujisawa-Sehara A, Nabeshima

Y, Kondoh H (1994) Mol Cell Biol 14:5692–5700.22. Postigo AA (2003) EMBO J 22:2443–2452.23. Furusawa T, Moribe H, Kondoh H, Higashi Y (1999) Mol Cell Biol 19:8581–

8590.24. Raymond CK, Roberts BS, Garrett-Engele P, Lim LP, Johnson JM (2005) RNA

11:1737–1744.25. Krutzfeldt J, Rajewsky N, Braich R, Rajeev KG, Tuschl T, Manoharan M,

Stoffel M (2005) Nature 438:685–689.26. Liu S, Liang Y, Huang H, Wang L, Li Y, Li J, Li X, Wang H (2005) Cell Physiol

Biochem 16:207–216.27. Razzaque MS, Naito T, Taguchi T (2001) Nephron 89:1–4.28. Yekta S, Shih IH, Bartel DP (2004) Science 304:594–596.29. Bagga S, Bracht J, Hunter S, Massirer K, Holtz J, Eachus R, Pasquinelli AE

(2005) Cell 122:553–563.30. Rossi JJ (2005) Nat Cell Biol 7:643–644.31. Wu L, Fan J, Belasco JG (2006) Proc Natl Acad Sci USA 103:4034–4039.32. Miao F, Gonzalo IG, Lanting L, Natarajan R (2004) J Biol Chem 279:18091–

18097.33. Li SL, Dwarakanath RS, Cai Q, Lanting L, Natarajan R (2005) J Lipid Res

46:220–229.

Kato et al. PNAS � February 27, 2007 � vol. 104 � no. 9 � 3437

MED

ICA

LSC

IEN

CES