MicroRNA miR-133 Represses HERG K+ Channel Expression Contributing to QT Prolongation in Diabetic Hearts*

We have previously found that the ether-a-go-go related gene (ERG), a long QT syndrome gene encoding a key K+ channel (IKr) in cardiac cells, is severely depressed in its expression at the protein level but not at the mRNA level in diabetic subjects. The mechanisms underlying the disparate alterations of ERG protein and mRNA, however, remained unknown. We report here a remarkable overexpression of miR-133 in hearts from a rabbit model of diabetes, and in parallel the expression of serum response factor (SRF), which is known to be a transactivator of miR-133, was also robustly increased. Delivery of exogenous miR-133 into the rabbit myocytes and cell lines produced post-transcriptional repression of ERG, down-regulating ERG protein level without altering its transcript level and caused substantial depression of IKr, an effect abrogated by the miR-133 antisense inhibitor. Functional inhibition or gene silencing of SRF down-regulated miR-133 expression and increased IKr density. Repression of ERG by miR-133 likely underlies the differential changes of ERG protein and transcript thereby depression of IKr, and contributes to repolarization slowing thereby QT prolongation and the associated arrhythmias, in diabetic hearts. Our study provided the first evidence for the pathological role of miR-133 in adult hearts and thus expanded our understanding of the cellular function and pathophysiological roles of miRNAs.

Abnormal QT interval prolongation is a prominent electrical disorder and has been proposed a predictor of mortality in patients with diabetes mellitus (DM), 3 presumably because it is associated with an increased risk of sudden cardiac death consequent to lethal ventricular arrhythmias (1)(2)(3)(4)(5)(6)(7)(8). Our recent study revealed that the long QT syndrome gene, human ethera-go-go-related gene (HERG) encoding the channel responsible for rapid delayed rectifier K ϩ current (I Kr ), is significantly down-regulated in its expression in diabetic hearts and this down-regulation contributes critically to diabetic repolarization slowing and QT prolongation (9,10). Strikingly, HERG expressions at transcriptional and post-transcriptional levels diverge in diabetic hearts, with its protein levels being reduced by some 60% while the mRNA levels remaining essentially unaltered (10). These disparate changes indicate that HERG expression is impaired mainly at the post-transcriptional level; however, it remained unclear what are the determinants for the differential regulations of HERG expression at protein and transcript levels.
MicroRNAs (miRNAs) are endogenous ϳ22-nucleotide non-coding RNAs that anneal to inexactly complementary sequences in the 3Ј-untranslated regions of target mRNAs of protein-coding genes to regulate gene expression. The major characteristics of miRNA actions is to specify translational repression without affecting the levels of the targeted mRNA (11,12). Among Ͼ300 miRNAs identified thus far, miR-1 and miR-133 are known to specifically express in adult cardiac and skeletal muscle tissues (13,14). Recent studies revealed that miR-1 and miR-133 play critical roles in regulating myogenesis. Increasing expression of miR-1 and miR-133 has been found in neonatal hearts and substantially higher levels are maintained in adult cardiac tissues (14), suggesting that in addition to regulating myogenesis, they may also possess other cellular functions in adult cardiac cells. However, our current understanding of the function of these miRNAs is still limited to developmental regulation and their possible roles in other cellular processes have not yet been explored.
We proposed that the muscle-specific miRNAs miR-1/miR-133 are able to repress HERG translation while keeping its mRNA unaffected and their levels are up-regulated in diabetic hearts, which causes the disparate changes of HERG protein and mRNA levels. This study was designed to test this hypothesis.

Preparation of Rabbit Model of DM-Male New Zealand
White rabbits weighing 1.6ϳ2.0 kg (Charles River Canada Inc.) were used and the procedures for development of alloxan-induced DM model were the same as previously described in * This work was supported in part by the Canada Diabetes Association and detail (9,10). The QT measurements and simultaneously recorded RR intervals were used to derive heart rate corrected QT intervals. Incidences of ventricular tachycardia and ventricular fibrillation were determined. All procedures are in accordance with the guidelines set by the Animal Ethics Committee of the Montreal Heart Institute and of Harbin Medical University. Isolation of Rabbit Ventricular Myocytes and Cell Culture-Myocytes were isolated from rabbit left ventricular endocardium via enzymatic digestion of the whole heart on a Langendorff apparatus with the procedures similar to previously described (9,10). The freshly isolated myocytes were stored either in the extracellular solution for patch clamp recordings or in 199 Medium as detailed elsewhere (9,15).
Whole-cell Patch Clamp Recording-Patch clamp recording of I Kr currents has been described in detail elsewhere (9,10).
Synthesis of miRNAs and Anti-miRNA Antisense Inhibitors and Their Mutant Constructs-miR-1 and miR-133 and their respective mutant constructs were synthesized by Integrated DNA Technologies, Inc. as detailed elsewhere (16) (also see supplemental Fig. 1). The mutant miRNAs each had eight nucleotides mismatches at the 5Ј-end, which disrupts their binding to the target sites and thus turns the miRNAs into negative controls (11)(12)(13)(14)16).
Construction of Chimeric miRNA-Target Site-Luciferase Reporter Vectors-To construct reporter vectors bearing miRNA-target sites, we synthesized (by Invitrogen) fragments containing the exact target sites for miR-1 and miR-133 or the mutated target sites, HERG cDNA, and inserted these fragments into the multiple cloning sites downstream the luciferase gene (HindIII and SpeI sites) in the pMIR-REPORT TM luciferase miRNA expression reporter vector (Ambion, Inc.), as detailed elsewhere (16).
Cell Culture-SKBr3 (human breast cancer cell line) and HEK293 (human embryonic kidney cell line) were purchased from ATCC (Manassas, VA). The cells were cultured as described previously (17).
Transfection and Luciferase Assay-The transfection procedures for cell lines and rabbit cardiac myocytes in primary culture, and luciferase activity assays were the same as described in detail elsewhere (16,17). Before transfection, cells were starved to synchronize growth by incubating in serum-and antibioticfree medium for 12 h.
Quantification of mRNA and miRNA Levels-The procedures for quantification of HERG and SRF transcripts by conventional TaqMan real-time RT-PCR were the same as described previously (16).
The mirVana TM qRT-PCR miRNA detection kit (Ambion), a quantitative reverse transcription (qRT)-PCR kit, was used in conjunction with real-time PCR with SYBR Green I for quantification of miR-1 and miR-133 (miR-133a ϩ miR-133b) tran-scripts (16). The total RNA samples were isolated with Ambion's mirVana miRNA isolation kit from SKBr3 cells, HEK293 cells, rabbit hearts, and human hearts. Fold variations in expression of miR-133 between RNA samples were calculated after normalization to 5s rRNA. Human tissues were obtained from the Second Affiliated Hospital of Harbin Medical University under the procedures approved by the Ethnic Committee for Use of Human Samples of the Harbin Medical University and from the Réseau de tissus pour études biologiques (RETEB) tissue bank under the procedures approved by the Human Research Ethics Committee of the Montreal Heart Institute. The criteria for inclusion of the tissues in our study were the patients that did not have primary heart problems at the time of death.
Western Blot-The procedures for semi-quantification of ERG and SRF protein levels were the same as described in detail elsewhere (9,10,(15)(16)(17). Membrane protein samples were extracted from left ventricular wall of rabbits and SKBr3 cells. The goat polyclonal antibodies against ERG and SRF were both purchased from Santa Cruz Biotechnology Inc.
Data Analysis-Group data are expressed as mean Ϯ S.E. Statistical comparisons (performed using analysis of variance followed by Dunnett's method) were carried out using Microsoft Excel. A two-tailed p Ͻ 0.05 was taken to indicate a statistically significant difference.

Overexpression of miR-1 and miR-133 and Down-regulation of ERG Protein Level in Diabetic
Hearts-Both miR-1 and miR-133 were expressed in rabbit hearts; however, miR-133 was ϳ10 times more abundant than that of miR-1. The levels of both miR-1 and miR-133 were found some 2.2-and 3-fold higher, respectively, in the ventricular RNA samples from rabbits with DM than those from healthy control animals. Up-regulation of the muscle-specific miRNAs was also observed in the ventricular samples from DM patients (Fig. 1).
We also reproduced the observations reported in our previous study (9,10), i.e. the protein level of the rabbit ERG (rbERG) was significantly lower in diabetic hearts than in healthy hearts despite that the transcript level remained unchanged. We further demonstrated the same disparity between HERG protein and mRNA expression levels in the hearts from DM patients (Fig. 1). Note that the molecular masses of ERG in rabbit (155 and 135 kDa) and human (140 and 120 kDa) were somewhat different presumably due to different glycosylations in different species; the larger band represents the mature glycosylated form and the smaller band represents the non-glycosylated form of ERG protein (16). The size rbERG is consistent with our previous finding (9,10) and that of HERG is identical to the results reported by Jones et al. (18).
Post-transcriptional Repression of HERG Expression by miR-133-HERG and rbERG share 91% homology in their sequences. We identified multiple putative target sites for miR-133 in rbERG and in HERG based on complementarity: at least six nucleotides exactly matching the 2-10 nucleotides from the 5Ј-end of miR-133 (supplemental Fig. 1). These sites may cooperate to confer the regulation by miR-133. Neither HERG nor rbERG contains any sites with more than five complementary nucleotides to miR-1.
To verify that HERG and rbERG are the cognate targets of miR-133 for post-transcriptional repression, we first inserted HERG cDNA into the 3Ј-untranslated region of a luciferase reporter plasmid containing a constitutively active promoter to determine the effects of miR-133 on reporter expression. Cotransfection of miR-133 and the chimeric luciferase-HERG vector into HEK293 cells consistently demonstrated smaller luciferase activities relative to transfection of the chimeric plasmid alone, but co-transfection of the mutant miR-133 (M-miR-133) failed to produce any effects ( Fig. 2A). HEK293 cells were used for luciferase reporter assays because these cells do not express endogenous ERG protein and miR-1/miR-133 (supplemental Fig. 2). Co-application of miR-133 with its antisense inhibitor AMO-133 eliminated the silencing effect on luciferase reporter activities (16,19,20). As an additional negative control, application of miR-1 failed to affect luciferase reporter activity.
The uptake and activities of transfected miRNAs was confirmed by using miR-1 and miR-133 standards in which the complementary sequences of miR-1 and miR-133 were cloned downstream of luciferase gene in the pMIR-REPORT plasmid (Fig. 2B).
We determined the effects of miR-133 on endogenous expression of HERG at the protein level by Western blot with SKBr3 membrane protein samples. SKBr3 was used because it is a human cell line that expresses endogenous HERG (17) but does not express the muscle-specific miR-1 or miR-133 (supplemental Fig. 2). Our data showed that transfection of miR-133 reduced HERG protein level down to ϳ10% of control value, and as a negative control the mutant miR-133 did not cause any appreciable changes (Fig. 2C). Co-application of AMO-133 nearly abolished the effects of miR-133, verifying the specificity of the miR-133 action. Moreover, transfection of miR-1 failed to

FIGURE 1. Up-regulation of the muscle-specific microRNAs miR-1 and miR-133 and down-regulation of ERG (ether-a-go-go-related gene) in rabbit hearts of diabetes model (rbERG, n ‫؍‬ 5 hearts for each group) and in human hearts (HERG, n ‫؍‬ 6 hearts for each group) from subjects with DM.
A and B, increases in mRNA levels of miR-1 and miR-133. C and D, downregulation of rbERG and HERG protein levels in DM hearts. AP, pretreated with antigenic peptide; control (Ctl), age-matched and sham-operated control rabbits or patients with healthy hearts. *, p Ͻ 0.05 versus control. GAPDH, glyceraldehyde-3-phosphate dehydrogenase. affect HERG protein level. By comparison, miR-133 produced virtually no effects on HERG mRNA level (Fig. 2D), indicating that miR-133 does not affect HERG mRNA stability.
The functional significance of ERG regulation by miR-133 was explored by whole-cell patch clamp studies of I Kr in isolated ventricular myocytes in primary culture. I Kr density in the myocytes from DM hearts or in the myocytes from healthy control heart transfected with miR-133 was severely diminished (Fig.  2E). The depression induced by DM was partially reversed by AMO-133 and that induced by exogenous miR-133 was abolished by AMO-133. AMO-133 slightly enhanced I Kr in control cells, presumably by eliminating the repressive effects of basal endogenous miR-133. As a negative control, miR-1 failed to affect I Kr .
Potential Role of SRF in miR-133 Overexpression-It has been shown that expression of miR1/miR-133 is dependent upon binding of SRF to their promoter regions (13,14), an important transcriptional factor in cardiac cells (21)(22)(23)(24). SRF protein level was found significantly increased in diabetic hearts relative to healthy hearts and so was SRF transcript level (Fig. 3A). Incubation of the DM myocytes in primary culture with distamycin A, which has been shown to selectively inhibit binding of SRF to its cis-element (25), largely reversed the increases in miR-1/ miR-133 expression (Fig. 3B). This effect was further confirmed by silencing of SRF using the siRNA directed against SRF (SRF-siRNA) (Fig. 3C). Moreover, in cells isolated from DM rabbits, SRF-siRNA, but not the negative control siRNA, increased I Kr density (Fig. 3D). The siRNA and distamycin A both slightly increased I Kr density in healthy control cells, presumably by inhibiting basal SRF. The efficiency of the SRF-siRNA in silencing SRF expression at mRNA level was verified (Fig. 3E). Unfortunately, the primary culture did not allow for sufficient quantity of protein samples for Western blot analysis of SRF protein levels or HERG protein levels.

DISCUSSION
miRNA-mediated gene regulation is now considered a fundamental layer of genetic programs that operates at the posttranscriptional level. However, despite our ability to identify miRNAs, regulatory targets have not been established or even confidently predicted for any of the vertebrate miRNAs, which has hampered progress toward elucidating the functions of miRNAs. Our current understanding of the functions of miR-NAs primarily relies on their tissue-specific or developmental stage-specific expression patterns as well as their evolutionary conservation and is thus largely limited to biogenesis and oncogenesis. Target finding and function discovery are two major challenges to researchers in miRNA research. The present study revealed the ability of a miRNA to regulate ion channel expression and the possible role in electrical remodeling in diabetic myocardium. It thus expanded our understanding of the cellular function and pathophysiological roles of miRNAs in a whole, reconsolidating the view that miRNAs likely have widespread functions in the cells.
Our study provides an explanation for the observed discrepancy between changes of HERG/rbERG expression at protein and mRNA levels. In our recent study on QT prolongation of diabetic hearts, we demonstrated that I Kr density and ERG pro-tein level were remarkably diminished, being the major factor for QT prolongation in diabetic rabbits, while ERG mRNA level was unaffected (9). Reduction of I Kr due to expression repression of HERG by miR-133 is expected to result in repolarization slowing thereby QT prolongation. In our recent study, we found that miR-133 repressed KCNQ1 (16), a channel protein responsible for the slow delayed rectifier K ϩ current (I Ks ) in cardiac cells. However, whether I Ks has significant contribution to diabetic QT prolongation is still an open question and our previous studies suggest a minimal role of I Ks (9, 10). Nonetheless, our study points to an important role of miR-133 in abnormal QT prolongation in diabetes and maybe in other pathological conditions as well. Our data also indicate that the cardiacspecific miR-1 is not responsible for the down-regulation of I Kr in DM heart.
It is important to note here that the phenomenon of disparate changes of ERG expression at protein and mRNA levels  ) and 36 h after, patch clamp recordings were performed. I Kr in display was elicited by a 2-s depolarizing voltage step to a test potential of ϩ10 mV from a holding potential of Ϫ60 mV (see supplemental Fig. 3 for the full range of voltages tested). DA, distamycin A (100 nM); n ϭ 7 cells for each group except for DA group (n ϭ 5 cells). E, verification of the efficiency of SRF-siRNA in silencing SRF in left ventricular myocytes isolated from DM rabbits, determined by quantitative real-time RT-PCR methods. *, p Ͻ 0.05 versus control; ϩ, p Ͻ 0.05 versus siRNA alone; n ϭ 3 samples for each group. GAPDH, glyceraldehyde-3-phosphate dehydrogenase.
have also been observed in failing heart and ischemic myocardium. For example, several studies found that I Kr current density was significantly diminished in myocytes from failing hearts that is also electrophysiologically characterized by repolarization slowing and QT prolongation similar to diabetic hearts, despite that the mRNA level of HERG was barely altered under these conditions (24 -29). Whether these disparate changes of ERG protein and mRNA in failing hearts and ischemic myocardium are consequent to up-regulation of miR-133 expression is worthy of detailed studies.
Our study also provides evidence for the potential role of SRF in miR-133 overexpression in DM myocytes. The SRF-siRNA not only nullifies the increase in miR-133 but also rescues depressed I Kr in DM. Whether SRF inhibition or knockdown could have beneficial effects on diabetic QT prolongation merits future investigations.