MicroRNA-21 Is a Downstream Effector of AKT That Mediates Its Antiapoptotic Effects via Suppression of Fas Ligand*

MicroRNA-21 (miR-21) is highly up-regulated during hypertrophic and cancerous cell growth. In contrast, we found that it declines in cardiac myocytes upon exposure to hypoxia. Thus, the objective was to explore its role during hypoxia. We show that miR-21 not only regulates phosphatase and tensin homologue deleted on chromosome 10 (PTEN), but also targets Fas ligand (FasL). During prolonged hypoxia, down-regulation of miR-21 proved necessary and sufficient for enhancing expression of both proteins. We demonstrate here for the first time that miR-21 is positively regulated via an AKT-dependent pathway, which is depressed during prolonged hypoxia. Accordingly, hypoxia-induced down-regulation of miR-21 and up-regulation of FasL and PTEN were reversed by activated AKT and reproduced by a dominant negative mutant, wortmannin, or PTEN. Moreover, the antiapoptotic function of AKT partly required miR-21, which was sufficient for inhibition of caspase-8 activity and mitochondrial damage. In consensus, overexpression of miR-21 in a transgenic mouse heart resulted in suppression of ischemia-induced up-regulation of PTEN and FasL expression, an increase in phospho-AKT, a smaller infarct size, and ameliorated heart failure. Thus, we have identified a unique aspect of the function of AKT by which it inhibits apoptosis through miR-21-dependent suppression of FasL.

MicroRNA (miRNA) 3 are molecules approximately twenty ribonucleotides long that specifically target mRNA through partial complementarity and, thereby, inhibit translation and/or induce their degradation. miR-21 is one of the most commonly and dramatically up-regulated miRNA in many cancers (1,2) and has been implicated in the inhibition of programmed cell death (2). Some of its validated targets include tropomyosin 1 (3), PTEN (2,4,5), programmed cell death 4 (Pdcd4) (6,7), TAp63 isoform of p53 family, and LRRFIP1, an inhibitor of NFB signaling (8). Similarly, miR-21 is one of the most highly and consistently up-regulated miRNA during car-diac hypertrophy (9 -12). Thum et al. (13) show that miR-21 is predominantly up-regulated in the myofibroblasts where it targets sprouty1 and enhances their survival and, thereby, fibrosis in the heart. Similarly, Roy et al. (14) show that miR-21 is elevated in the myofibroblast-infiltrated area 7 days after ischemia/ reperfusion and suppresses metalloprotease-2 via targeting PTEN. More recently, studies have shown that miR-21 exerts an antiapoptotic function in cardiac myocytes via inhibiting PDCD4 (15) and reduces infarct size via local viral delivery to the heart (16). However, the signaling pathway that regulates miR-21 has not been identified.
Two of the molecules that play a major role in ischemic injury of the heart include PTEN and FasL. PTEN is a major negative regulator of AKT (17) whose activity is modulated by its abundance, oxidation, or phosphorylation (18). It is also targeted by miR-21, which provides a specific post-transcriptional mechanism for regulating its expression (2,4,5). PTEN has been regarded as the Achilles' heel of the myocardium (19), whose knockdown reduces infarction following myocardial ischemia (20). On the other hand, FasL is the main activator of the extrinsic apoptotic pathway in cells and in the heart during failure (21) and ischemia (22) and is responsible for ϳ64% of the apoptosis contributing to myocardial infarction (22). Although both PTEN and FasL play essential roles in myocyte apoptosis, the mechanism of their regulation remains largely unknown. In this report, we show for the first time that both PTEN and FasL are negatively regulated by an AKT-miR-21-dependent pathway that is deactivated during ischemia and is rescued by overexpression of miR-21.

DNA Constructs Cloned into Recombinant Adenovirus-A
320-bp sequence encompassing the stem-loop of miR-21 was amplified from mouse genomic DNA by PCR using the following primers: 5Ј-CCTGCCTGAGCACCTCGTGC-3Ј and 5Ј-GACTGTGACGACTACCCCAA-3Ј and cloned into recombinant adenovirus downstream of a cytomegalovirus (CMV) promoter (23). For a control, a scrambled sequence (5Ј-GAA-CCGAGCCCACCAGCGAGC-3Ј) replaced the mature miRNA sequence within its stem-loop structure (23). This was as a control in all experiments. The miR-21 eraser was synthesized in the form of a tandem repeat antisense sequence of mature miR-21 terminating in (T) 6 . This construct was cloned into recombinant adenovirus downstream of a U6 promoter. Human PTEN cDNA (accession number NM_000314), purchased from Origene, and short hairpin RNA targeting PTEN, synthesized in the form of a hairpin-forming oligonucleotide corresponding to bases 1471-1491 of Mus musculus PTEN (accession number NM_008960), were cloned into recombinant adenovirus downstream of a CMV and U6 promoter, respectively. Akt1/PKB␣ cDNA (activated) and dominant negative, kinase-defective, dnAkt1 cDNA in pUSEamp from Upstate (Millipore), were cloned into recombinant adenovirus downstream of a CMV promoter.
Construction of the miR-21 Transgenic Mouse-The miR-21 transgene was constructed using the miR-21 DNA described above, cloned downstream of the ␣-myosin heavy chain promoter and upstream of an SV40 polyadenylation signal. The transgenic mouse was generated at the Transgenic Core Service, University of Medicine and Dentistry of New Jersey by Dr. Gassan Yehia. All required animal protocols were approved by the Institutional Animal Care and Use Committee at the New Jersey Medical School.
Culturing Cardiac Myocyte and Treatments-Cardiac myocytes were prepared as previously described (24). Briefly, hearts were isolated from 1-to 2-day-old Sprague-Dawley rats. After dissociation the cardiac myocytes were differentially separated by Percoll gradient centrifugation followed by a differential preplating step for further separation of non-cardiac cells. Myocytes were then plated in Dulbecco's modified essential medium/Ham's F-12 (1:1) supplemented with 10% fetal bovine serum. Twenty-four hours after plating, the medium was changed to serum-free, and the cells are infected with recombinant adenoviruses at a multiplicity of infection of 10 -20 particles/cell. Myofibroblast that were separated from the cardiac myocytes were isolated from the same Percoll gradient and cultured in Dulbecco's modified essential medium/Ham's F-12 (1:1) supplemented with 10% fetal bovine serum.
SW-480 colon cancer cell line was purchased from ATCC (catalogue number CCL-228) and propagated in Leibovitz's L-15 medium supplemented with 2.0 mM L-glutamine and 10% fetal bovine serum, in a CO 2 -free incubator. The medium was changed to Dulbecco's modified essential medium/Ham's F-12 (1:1) just before transfer to a hypoxic chamber. An MCF-7 breast adenocarcinoma cell line was purchased for ATCC (catalogue number HTB-22) and propagated in Eagle's minimum essential medium supplemented with 10% fetal bovine serum and 0.1 mg/ml bovine insulin, as recommended by the manufacturer.
Construction of Adenoviruses-Recombinant adenoviruses were constructed, propagated, and titered, as previously described by Dr. Frank Graham (25). Briefly, pBHGlox⌬E1,3Cre (Microbix), including the ⌬E1 adenoviral genome, was co-transfected with the pDC shuttle vector containing the gene of interest, into 293 cells using Lipofectamine (Invitrogen). Through homologous recombination, the test genes integrate into the E1-deleted adenoviral genome. The viruses were propagated on 293 cells and purified using CsCl 2 banding followed by dialysis against 20 mM Tris-buffered saline with 2% glycerol. Tittering was performed on 293 cells overlaid with Dulbecco's modified Eagle's medium plus 5% equine serum and 0.5% agarose.
Northern Blotting-This was carried out as described previously (23).
Immunocytochemistry-Myocytes were plated in fibronectin-coated glass slides for 24 h before infecting them with adenovirus. After the desired period of time slides were fixed in 3% paraformaldehyde plus 0.3% Triton X-100 in CB buffer (10 mM PIPES, 150 mM NaCl, 5 mM EGTA, 5 mM MgCl 2 , 5 mM glucose, pH 6.1) for 5 min at 25°C followed by 3% paraformaldehyde in CB buffer for 20 min at 25°C. The cells were then immunolabeled with anti-FasL (Santa Cruz Biotechnology, catalogue number sc-834) and anti-MHC (MF-20, Developmental Studies Hybridoma Bank, University of Iowa, Iowa City, IA) in Trisbuffered saline with 1% bovine serum albumin. Slides were mounted using Prolong Gold anti-fade plus 4,6-diamidino-2phenylindole (Invitrogen).
Luciferase Assay-A concatemere of miR-21-predicted target sequence within the FasL 3Ј-UTR (GGCCCATTTGAC-TGACTGATAAGCTA) ϫ3 and a mutant lacking the complementarity with miR-21 seed sequence (GGAGACCCACATT-GCGACTATA) ϫ3 were cloned downstream of the luciferase gene driven by CMV promoter, generating Luc.FasL and Luc. Mut vectors, respectively. Cultured neonatal myocytes were transfected with these constructs using Lipofectamine (Invitrogen) in conjunction with plasmids expressing miR-21 (CMV.miR-21) or a nonsense stem-loop. The cells were harvested after 24 h, and luciferase activity was assayed using an Lmax multiwall luminometer (Molecular Devices).
Exposing Myocytes to Hypoxia-To mimic ischemic conditions, cells in culture medium lacking serum were subjected to hypoxia in a "sealed hypoxia chamber" (Billups-Rothenberg Inc.). The chamber was filled with a gas mixture of 95% nitrogen and 4.8% Ϯ 0.2% CO 2 (Inhalation Therapy) at 7 p.s.i./12,000 kPa filing pressure for 15 min and sealed. The chamber was then placed in a 37°C incubator for the required period.
Monitoring Mitochondrial Membrane Potential-Mitochondrial membrane potential was monitored using JC-1 cationic dye (Molecular Probes, Invitrogen), as recommended by the manufacturer. Briefly, the cells were incubated with JC-1 (0.35 g/ml) for 20 min at 37°C. They were then washed with 1ϫ phosphate-buffered saline and imaged live.
Myocardial Ischemia and Ischemia/Reperfusion-Mice 14-16 weeks of age were anesthetized by an intraperitoneal injection of pentobarbital sodium (60 mg/kg). Under mechanical ventilation via tracheal intubation and through a left third intercostal thoracotomy, the pericardial sac was opened and an 8-0 nylon suture was passed under the left anterior descending coronary artery 2-3 mm from the tip of the left auricle. Then non-traumatic silicone tubing was placed on top of the vessel, and a knot was tied on top of the tubing to occlude the left anterior descending coronary artery. At this point, if permanent occlusion is the objective, the chest cage is then closed in layers and the pneumothorax reduced. On the other hand, for ischemia/reperfusion, the left anterior descending coronary artery was occluded for 45 min only, after which the knot was released for a 16-h reperfusion period. The loosened ligature was left in place for subsequent occlusion prior to infusion of Evans blue dye, which was used to assess the area at risk at the conclusion of the reperfusion period. The coronary occlusion and reperfusion were verified by visual inspection and by electrocardiograph monitoring. Throughout the procedure, a 36.8 -37°C rectal temperature was maintained.
Triphenyltetrazolium Chloride Staining-At the conclusion of the ischemia/reperfusion protocol, the animals were anesthetized, the chest cage was opened, and the coronary artery was re-occluded at the same site, using the ligature that was loosened in situ. Evans Blue dye (0.5%) was perfused for a few minutes into the coronary artery above the site of ligation before the hearts were excised and sliced into a 1-mm-thick cross-section. The sections were fixed with formaldehyde and stained with triphenyltetrazolium chloride (1%) solution for 15 min at 37°C. Both sides of each slice were photographed using a camera-mounted dissecting microscope. ImageJ software was then used to assess the percent infarct area relative to the area at risk.
Echocardiography-Mice were anesthetized with 2.5% avertin (0.010 -0.015 ml/g body weight) administered by intraperitoneal injection. Transthoracic echocardiography (Sequoia C256, Acuson, Mountain View, CA) was performed using a 13-MHz linear ultrasound transducer. The chest was shaved. Mice were placed on a warm saline bag in a shallow left lateral position, and warm coupling gel was applied to the chest. Electrocardiographic leads were attached to each limb using needle electrodes. Twodimensional images and M-mode tracing (sweep speed ϭ 100 -200 mm/s) were recorded from the parasternal short-axis view at the mid papillary muscle level. Care was taken not to apply too much pressure to the chest wall. The images were recorded on videotape, and freeze frames were printed on a Sony color printer, scanned into a PC using Photoshop (Adobe). The images were then analyzed by using the NIH ImageJ program. M-mode measurements of LV internal diameter (LVID) and wall thickness were made from three consecutive beats and averaged using the leading edge-to-leading edge convention adopted by the American Society of Echocardiography. End-diastolic measurements are taken at the time of the apparent maximal LV diastolic dimension. End-systolic measurements are made at the time of the most anterior systolic excursion of the posterior wall. LVEF is calculated by the cubed methods as follows: LVEF ϭ LVID d Ϫ LVID s /LVIDd, where d indicates diastolic and s indicates systolic. LV percent fractional shortening (LVFS) is calculated as LVFS% ϭ [(LVEDD Ϫ LVESD)/LVEDD]/100. Heart rate is calculated from the period between two consecutive electrocardiogram tracings.
Hemodynamic Measurements-Mice were anesthetized as described above, and a 1.4-French (Millar Instruments) catheter tip micromanometer catheter was inserted through the right carotid artery into the aorta and then into the LV where pressures, dp/dt, and Ϫdp/dt were recorded.
Statistics-For comparing two experimental groups, an unpaired, two-tailed, Student t test (Excel software) was used for calculating the probability values. p Ͻ 0.05 was considered significant.

RESULTS
Hypoxia Induces Down-regulation of miR-21-Because miR-21 has been reported to target a plethora of antiapoptotic genes, we decided to examine its expression levels in cardiac myocytes exposed to hypoxia and in the ischemic myocardium. Fig. 1a   FIGURE 1. Down-regulation of miR-21 in cardiac myocytes exposed to hypoxia. a, myocytes were exposed to hypoxia for increasing intervals, as indicated. The treatments were staggered to allow synchronized extraction of total RNA and Northern blot analysis. The autoradiogram signals for miR-21 were quantified and normalized to that of U6 (n ϭ 4). The numbers were averaged and plotted as relative values to the control adjusted to 1. Error bars represent ϮS.E. and *, p Ͻ 0.01 versus normoxia (0 time point). b, 16-week-old C57bl/6 mice were subjected to coronary artery occlusion (CAO) for the intervals indicated. For each time point cardiac tissue from the ischemic and remote regions, and a sham-operated heart, were dissected, and total RNA was extracted and used for Northern blot analysis. The autoradiogram signals for miR-21 were quantified and normalized to that of U6 (n ϭ 4). The numbers were averaged and plotted as relative values to the sham adjusted to 1. Error bars represent ϮS.E. and *, p Ͻ 0.05 versus sham. c, myofibroblasts, MCF-7, and SW480 cells were exposed to hypoxia for the indicated periods. RNA was extracted and analyzed by Northern blotting. The numbers below the autoradiogram are the relative levels of miR-21/U6 signal (n ϭ 2).
shows that miR-21 declined by 40 Ϯ 7% within 1 h and by 75 Ϯ 8% within 12-24 h of exposure to hypoxia. Meanwhile, there was no significant change in miR-1 levels (Fig. 1a). Similarly, within 6 h of coronary artery occlusion in mice, miR-21 was reduced by 40% in the ischemic region (Fig. 1b). In contrast, though, it increased by 1.8-fold in the perinfarct zone, as expected in myocytes undergoing hypertrophy. At this early stage of ischemia there was no significant inflammatory cell infiltration or myofibroblast proliferation detected in the ischemic region. We also examined the response of other cell types to hypoxia. Interestingly, colon cancer (SW480) and breast cancer (MCF-7) cell lines exhibited a slight increase in miR-21 after 48 h of hypoxia (Fig. 1c). However, fibroblasts isolated from the neonatal rat heart (myofibroblasts) exhibited a reduction in miR-21 within the same time frame. These differences between cell types may reflect a contrast in their tolerance to hypoxia, cancer cells being the most resilient. The results also suggest that miR-21-targeted genes may play a role in a normal cell's response to hypoxia. miR-21 Targets FasL and PTEN during Hypoxia-PTEN is a validated miR-21 target (2,4,5). Using TargetScan v5.1 and Microcosm target prediction software, we identified FasL as another miR-21-predicted target, which has not been previously experi-mentally validated. Fig. 2a shows the alignment between miR-21 and FasL 3Ј-UTR of both the mouse and human genes. Inclusion of this sequence within the 3Ј-UTR of a luciferase gene rendered it responsive to hypoxia and to overexpression of miR-21, which induced an increase or a decrease in its expression, respectively, as demonstrated in Fig. 2b. However, mutations within the seed sequence's binding site abrogated these responses. This suggests that miR-21 directly targets and inhibits FasL.
We also tested the effect of miR-21 on the expression of the endogenous protein. Interestingly, overexpression of miR-21 did not inhibit basal levels of either PTEN or FasL in cardiac myocytes but did completely suppress their up-regulation after exposure to hypoxia (Fig. 2c). In contrast, knockdown of miR-21 by an antisense construct (miR-21-eraser) induced up-regulation of both genes. This suggests that miR-21 is saturating during normoxia but limiting/reduced during hypoxia, relative to its target mRNAs. This is in contrast to the results observed with the luciferase reporter assay in Fig. 2b, where the cells were supplemented with both an exogenous target (luciferase reporter) and miR-21. There have been contradictory reports regarding the localization of FasL to cardiac myocytes (26 -28). Also, because primary myocyte cultures usually contain Ͻ10% of non-myocytes, the results of Western blot analysis could be ambiguous. Thus, to address this issue we co-immunostained myocytes for FasL (green) and ␣-myosin heavy chain (␣MHC, red). The results show that FasL is expressed during normoxia, becomes more abundant during hypoxia, and is strictly localized to the junctions between myocytes (Fig. 2d). Overexpression of miR-21 produced dramatic reduction in junctional FasL, in addition to its dispersion to internal vesicles. Notably, non-cardiac cells (Fig. 2d, marked by an asterisk) had little detectable FasL.
AKT Regulates miR-21, PTEN, and FasL Expression-AKT is an established anti-apoptotic gene that has been implicated in the negative regulation of FasL (29,30). Therefore, we questioned whether AKT regulates miR-21 or its targets in cardiac myocytes. To address this, we overexpressed constitutively active AKT (caAKT) in the myocytes before exposure to hypoxia. Similar to overexpression of miR-21, this resulted in an effective down-regulation of junctional FasL and its dispersion to internal vesicles (Fig. 3a). In contrast, overexpression of PTEN, a negative regulator of AKT, induced up-regulation of FasL during normoxia, equivalent to its induction by hypoxia FIGURE 2. miR-21 targets FasL. a, an alignment between miR-21 and the 3Ј-UTR of human (hsa) and mouse (mmu) FasL 3Ј-UTR. The " " represents a Watson-Crick base pair, and the ":" represents a wobble base pair. b, the miR-21-targeted site in the mouse FasL 3Ј-UTR (Luc-FasL) and a nonsense mutant (Luc-FasL-mt) were separately cloned into the 3Ј-UTR of the luciferase gene downstream of a cytomegalovirus promoter, as indicated. These constructs were virally delivered to myocytes in the presence or absence of a control, or Ad.miR-21, virus where indicated by the plus sign, for 24 h (n ϭ 6). Cells were then exposed to hypoxia or normoxia for an additional 24 h, as indicated, before total protein was extracted and analyzed for luciferase activity. The results were averaged and plotted as luciferase activity/g of protein.
Error bars represent ϮS.D. and *, p Ͻ 0.05 versus control normoxia. **, p Ͻ 0.01 versus control hypoxia. c, myocytes were treated with Ad.miR-21, Ad.miR-21 eraser, or a control virus for 24 h before exposing them to hypoxia for an additional 24 h, as indicated with the plus sign. Total cellular protein was extracted and analyzed by Western blotting for the molecules listed on the left of each panel (n ϭ 6). d, myocytes were treated with Ad.miR-21, Ad.miR-21-eraser, or a control virus, in the presence or absence of 24 h hypoxia, as indicated. Cells were fixed and immunostained with anti-FasL (green) and anti-␣MHC (red), and nuclei were stained with 4,6-diamidino-2-phenylindole (DAPI, blue) (n ϭ 3). The asterisk marks non-cardiac, ␣MHC-negative cells. Arrowheads mark FasL-positive vesicles. (Fig. 3, a and b). In consensus, knockdown of PTEN (Ad.siPTEN) resulted in up-regulation of phospho-AKT (Fig.  3c) and inhibited hypoxia-induced FasL (Fig. 3a). To determine whether the effect of PTEN was mediated through inhibition of AKT, we delivered it to the cells in the presence of caAKT. The results show that the effect of caAKT was dominant and counteracted that of PTEN (Fig. 3, a and b). Surprisingly, the results also revealed that caAKT inhibited the expression of the exogenously delivered PTEN (Fig. 3b). This suggested that AKT inhibition of both FasL and PTEN might be mediated through up-regulation of miR-21. Indeed, knockdown of miR-21 resulted in reversal of caAKT inhibition of FasL expression (Fig.  3a). Moreover, Northern blot analysis confirmed that caAKT induced up-regulation of miR-21 and rescued its down-regula-tion during hypoxia, which was abrogated upon pretreatment of the cells with Ad.miR-21-eraser (antisense miR-21) (Fig. 3, d  and f). This suggests that caAKT suppresses FasL and PTEN during hypoxia through an miR-21-dependent mechanism.
The AKT pathway is commonly regarded in terms of its activation but rarely its inhibition. However, we speculated that if activated AKT inhibits hypoxia-induced FasL and PTEN and increases miR-21, it is likely that inhibition of endogenous AKT would be responsible for the reverse effects seen during hypoxia. Although determining the effect of knockdown of endogenous AKT by short silencing RNA on miR-21 and its targets was complicated by the rapid ensue of cell death, its inhibition by overexpression of a dominant negative mutant (Ad.dnAKT), PTEN, or treatment of the cells with wortmannin FIGURE 3. AKT regulates FasL through a miR-21-dependent mechanism. a, myocytes were treated with Ad.caAKT, Ad.miR-21-eraser, Ad.PTEN, Ad.siPTEN, Ad.dnAKT, wortmannin, hypoxia, or Ad.control separately or in various combinations, for 24 h, as indicated. Cells were then fixed and immunostained with anti-FasL (green) and anti-␣MHC (red), and nuclei were stained with 4,6-diamidino-2-phenylindole (DAPI, blue) (n ϭ 3). The asterisk marks non-cardiac, ␣MHCnegative cells. Arrowheads mark FasL-positive vesicles. b, myocytes were infected with Ad.PTEN, Ad.caAKT, or a control virus, separately or in combination for 24 h, as indicated by the plus sign. Total protein was extracted and analyzed by Western blotting for the molecules listed on the left of each panel (n ϭ 2). c, myocytes were treated with an adenovirus expressing short hairpin RNA targeting PTEN (Ad.siPTEN) or a control virus for 48 h before total protein was extracted and analyzed by Western blotting for the molecules listed on the left of each panel (n ϭ 2). d, myocytes were treated with Ad.caAKT or a control virus for 24 h, followed by exposure to hypoxia or normoxia for an additional 24-h period, as indicated by plus signs. In parallel, similar treatments were applied to cells that were preincubated with Ad.miR-21-eraser for 24 h. Total RNA was then extracted and analyzed by Northern blotting for the molecules listed on the left of each panel (n ϭ 3). e, myocytes were treated with 5 M wortmannin, Ad.dnAKT, or a control virus for 24 h. Total RNA was then extracted and analyzed by Northern blotting for the molecules listed on the left of each panel (n ϭ 2). f, the autoradiogram signals of miR-21 in d and e were quantified and normalized to that of U6. Results were averaged and graphed as -fold change in miR-21 relative to the normoxia control signal. Error bars represent ϮS.D. and *, p Ͻ 0.01 versus control normoxia. g, myocytes were exposed to hypoxia for the indicated intervals. Protein and RNA were extracted from parallel sets and analyzed by Western blotting and Northern blotting (lower two panels) for the molecules listed on the left of each panel (n ϭ 2).
induced up-regulation of junctional FasL (Fig. 3a) and downregulation of miR-21 (Fig. 3, e and f). This indicated that inhibition of AKT is sufficient for inducing down-regulation of miR-21 and may, thus, be the mediating signal during hypoxia. Fig. 3g shows that exposure of myocytes to hypoxia results in a biphasic effect on phospho-AKT and miR-21 levels. During short intervals (15 min) of hypoxia, AKT phosphorylation is enhanced and is associated with an increase in PTEN phosphorylation that deactivates PTEN and an increase in miR-21. However, longer periods of hypoxia induced a reduction in phospho-AKT, associated with an increase in total PTEN and a decrease in its phosphorylation, and reduced miR-21 levels. The data suggest that long term hypoxia induces inhibition of AKT, which mediates the down-regulation of miR-21 and upregulation of FasL and PTEN. This may be initiated by dephosphorylation and activation of PTEN, which in turn deactivates the AKT pathway. The data also suggest that down-regulation of miR-21 during hypoxia is mediated through inhibition of AKT.
miR-21 Enhances AKT Phosphorylation-Because miR-21 suppresses the expression of PTEN it would, accordingly, be expected to enhance AKT phosphorylation. Thus, to confirm its inhibition of PTEN and also determine if this effect has any functional significance, we tested the consequence of overexpressing miR-21 on AKT phosphorylation. Under quiescent conditions, miR-21 had no influence on the phosphorylation levels of AKT, which reconciles with its lack of inhibition of PTEN under the same conditions (see Fig. 2c). In contrast, when it was added before stimulating the cells with an activator of the AKT pathway, such as insulin, we observed a 5-fold increase in phospho-threonine 308 and a 20-fold increase in phospho-serine 473 following 1 h of insulin treatment (Fig. 4, a  and d). In contrast, the miR-21 eraser inhibited AKT phosphorylation by 40% (Fig. 4, b and d). Insulin was selected here for two reasons: one, it is a robust activator of AKT and, second, insulin resistance has been attributed to an increase in PTEN (31)(32)(33)(34)(35)(36). As seen in Fig. 4a, insulin stimulation was accompanied by a 1.9-fold increase in PTEN that was completely suppressed by overexpression of miR-21. To determine if the effect of miR-21 was indeed mediated through suppression of PTEN, we supplied the cells with miR-21 in the presence of exogenous PTEN. This treatment resulted in complete abrogation of miR-21-enhanced AKT phosphorylation (Fig. 4, c and d). Thus, by suppressing PTEN, miR-21 enhances AKT activity, which, in turn, induces up-regulation of miR-21 (Fig. 3c), creating a positive feedback loop. miR-21 Is Necessary for AKT-mediated Inhibition of Apoptosis-Because miR-21 is required for AKT-mediated inhibition of FasL during hypoxia, we questioned whether it is necessary or sufficient for mediating the antiapoptotic effects of this kinase. To address this we treated myocytes with caAKT in the presence or absence of miR-21-eraser, or independently with miR-21, before exposing them to hypoxia. After 24 h, live cells were loaded with JC-1 dye and immediately imaged. Fig. 5a shows that the JC-1 dye predominantly aggregates in the healthy mitochondria and emits a red florescence, whereas the excess monomeric form remains in the cytosol, emitting green florescence (Fig. 5a). During hypoxia the mitochondrial outer membrane is damaged, and this is reflected by a loss in the red JC-1 aggregates. Both overexpression of caAKT and miR-21 prevented mitochondrial damage by 80 Ϯ 20% and 50 Ϯ 13%, respectively, whereas knockdown of miR-21 partly reversed the effect of caAKT (35 Ϯ 15%, Fig. 5, a and b). Interestingly, though, neither miR-21-eraser nor dnAKT were sufficient for inducing mitochondrial damage when applied during normoxia, despite the fact that they were both sufficient for inducing up-regulation of FasL.
Caspase-8 is a direct downstream mediator of the Fas-induced death signal, thus, to confirm the antiapoptotic effects of miR-21 and its contribution to AKT function, we measured the activity of this caspase. Interestingly, unlike its effect on mitochondrial damage, miR-21 completely inhibited hypoxia-induced the activity of caspase-8, whereas the miR-21-eraser completely abolished AKT inhibition (Fig. 5c). This indicated that probably both the intrinsic and extrinsic apoptosis pathways are involved in mitochondrial injury during hypoxia, where miR-21 has a major role in inhibiting the latter only through suppression of FasL. On the other hand, AKT, through suppression of FasL, as well as BAD and Bax, has the potential to inhibit both pathways. Hence, miR-21 partly mediates the antiapoptotic effects of AKT.  (n ϭ 3). b, myocytes were treated with Ad.miR-21 eraser or a control virus for 24 h before stimulation with insulin for 1 h, as indicated by the plus signs. Total protein was extracted and analyzed by Western blotting for the molecules listed on the left of each panel (n ϭ 3). c, myocytes were treated with Ad.PTEN, Ad.miR-21, or a control virus, separately or in combination, for 24 h before treatment with insulin for 1 h, as indicated by the plus signs. Total protein was extracted and analyzed by Western blotting for the molecules listed on the left of each panel (n ϭ 3). d, the signal for phospho-threonine 308 and phospho-serine 473 in each of the panels was quantified and normalized to total AKT. The results were averaged, and the -fold change of insulin stimulated in the presence of miR-21, miR-21 eraser, or PTEN versus insulin stimulated in the presence of the control virus, was graphed on a log scale. Error bars represent ϮS.D. and *, p Ͻ 0.001 versus insulin stimulated in the presence of the control virus.

Overexpression of miR-21 Retards Ischemic Damage and
Heart Failure-To investigate the role of miR-21 in the heart in vivo, we generated a cardiac-specific transgenic mouse model overexpressing miR-21 (miR-21-TG). Two lines were obtained with ϳ10and 20-fold overexpression of the transgene relative to the wild-type littermates (Fig. 6a). These animals show no overt phenotype, and all physical and functional parameters of the heart are comparable to the wildtype mice before or after short term transverse aortic constriction. To examine whether miR-21 could exert its antiapoptotic effect in vivo in the adult myocytes, we subjected the mice to 45 min of ischemia followed by 16 h of reperfusion (I/R). Fig. 6b shows the results of Evans blue and triphenyltetrazolium chloride staining in cardiac sections from the miR-21-Tg and wild hearts after I/R. The infarct zone/area at risk was significantly smaller in the transgenic model (Fig. 6c). This is consistent with suppression of junctional FasL and, thus, cell death by the overexpressed miR-21.
We also investigated the effect of the transgene on long term ischemic injury in the heart. After 4 weeks of left coronary artery occlusion, cardiac function and physical parameters were assessed by echocardiography and hemodynamic measurements. Heart sections revealed a less dilated left ventricular chamber in the miR-21-Tg versus the wild-type mice (Fig. 6d) that was confirmed by echocardiography measurements of the left ventricular end diastolic and end systolic dimensions (Fig.  6f, LVEDD and LVESD). This, in addition to lower left ventricular end diastolic and systolic pressure (Fig. 6f, LVEDP and  LVSP), and higher ejection fraction (EF(%)) in the miR-21-TG, demonstrates reduced signs of cardiac failure. In agreement, there was no detectable lung congestion or increase in lung weight in the transgenic mice (Fig. 6e). Moreover, there was less collagen deposition, which is a reflection of reduced myocyte death, inflammatory cell infiltration, and myofibroblast proliferation (Fig. 6g), in addition to inhibition of ischemia-induced up-regulation of PTEN, FasL, and caspase-6, and down-regulation of phospho-AKT, in the left ventricle of the miR-21-TG (Fig. 6h). We also confirmed that FasL was localized to the intercalated discs in the ischemic heart, which was undetectable in the transgenic model (Fig. 6i). Thus, the results demonstrate that the miR-21 transgene effectively inhibits PTEN and FasL during myocardial ischemia and ameliorates signs of cardiac failure.

DISCUSSION
miR-21 is one of the most commonly and highly up-regulated miRNA in cancer and cardiovascular diseases and, thus, one of the most studied. A preponderance of reports shows that it targets antiapoptotic and tumor suppressor genes and is involved in promoting cardiac myocyte and cancer cell survival. Its antiapoptotic effect has been predominantly attributed to inhibition of PTEN and PDCD4 expression (2, 4 -7). Whereas up-regulation of miR-21 in various pathological conditions has been the prevailing finding, we discovered herein that miR-21 is downregulated in the ischemic heart, consistent with the results of a recent report (16). We observed a similar down-regulation in isolated cardiac myocytes and myofibroblasts after prolonged hypoxia, albeit at different rates. On the other hand, brief periods of hypoxia (15 min) induced up-regulation of miR-21 in myocytes, in concordance with its increase seen during ischemia preconditioning (37). In contrast, MCF-7 and SW-480 cancer cell lines exhibited a slight increase in miR-21 that did not decline for up to 48 h of hypoxia. This indicated that the pathway that induced down-regulation of miR-21 in myocytes is refractory to hypoxia in cancer cells. We found this to be an AKT-dependent pathway, whose inhibition during hypoxia may be initiated by dephosphorylation and activation of PTEN. This explains why cancer cells that are frequently deficient in PTEN might be insensitive to hypoxia-induced inhibition of AKT.
Thus, in this report we have identified AKT as a positive upstream regulator of miR-21, experimentally validated that FasL is a direct target of miR-21, and showed that AKT regulates PTEN and FasL expression in cardiac myocytes. We also show that this pathway is deactivated during ischemia or hypoxia, which is required for down-regula-tion of miR-21 and, thereby, up-regulation of its targets (Fig. 7). Indeed, there are several notable parallelisms between previously known functions of AKT and miR-21 that support this finding. For example, similar to miR-21, AKT activity is up-regulated in many cancers and cardiac hypertrophy and confers a survival function unto the cells. More significantly, AKT has been implicated in the negative regulation of FasL in smooth muscle (30) and cancer cells (38), and reduction of PDCD4 expression in cancer cells (39,40), both of which are now validated targets of miR-21. Thus, AKT-induced up-regulation of miR-21 provides a mechanism whereby AKT inhibits these molecules. As evidence, we show that AKT-mediated suppression of FasL during hypoxia in cardiac myocytes was reversed upon knockdown of miR-21 (Fig. 3a), as was its effect on caspase-8 activity (Fig. 5c). However, its effect on mitochondrial integrity was only partly reversed (Fig. 5a). This suggests that up-regulation of miR-21 and suppression of FasL only partly contribute to the antiapoptotic effects of AKT. Our recently published data support this conclusion, because it shows that, in addition to up-regulation of miR-21, AKT induces down-regulation of miR-199a and up-regulation of hypoxia-inducible factor-1alpha (Hif-1␣) and Sirt-1 (41). It is important to note that during short term hypoxia Hif-1␣ exerts protective effects, whereas during prolonged hypoxia it is involved in a proapoptotic function that involves stabilization of nuclear p53, which in turn induces an increase in FasL mRNA (42). We have also previously shown that this could be suppressed by overexpression of miR-199a, which inhibits the expression of Hif-1␣ and, thus, p53 (43).
PTEN is an established negative regulator of AKT, but a reciprocal effect has not been shown. Upon overexpression of caAKT we found that it not only inhibited FasL but also suppressed PTEN. This is consistent with AKT inducing up-regulation of miR-21, which in turn inhibits PTEN. Conversely, PTEN may inhibit miR-21, via inhibiting AKT. This reciprocal relation between a miRNA and its target has been previously described for several miRNA target pairs, including miR-200 family and ZEB1 (44), miR-9 and REST (45), miR-145 and OCT4 (46), and let-7 and lin-28 (47), where the miRNA target is itself a negative regulator of the miRNA, creating a double negative feedback loop that could potentially augment their effects. In our transgenic model, overexpression of miR-21 had no impact on physical or functional parameters of the heart during normal conditions. Similarly, overexpression of miR-21 in cultured myocytes had no effect on basal levels of its targets, PTEN or FasL, during resting conditions. This indicated that its levels are saturating relative to its targets under these settings. On the other hand, its function was uncovered following its down-regulation by an antisense construct or during ischemia, which proved necessary for up-regulation of its targets (Figs. 2c and 6 (b-f)). Thus, the question remains regarding its role during cardiac hypertrophy or in cancer, in which it is highly up-regulated. One explanation is that it is an adaptive mechanism against the gradual development of ischemia as mass outgrows vascular supply during enhanced tissue growth. Alternatively, the increase in miR-21 parallels the increase in its target mRNAs, as a consequence of enhanced global transcription associated with induction of growth. Another possibility is that miR-21 might be targeting other genes under those conditions. Indeed, Thum et al. reported that miR-21 is predominantly up-regulated in the myofibroblast during cardiac hypertrophy or failure, where it promotes cell survival and fibrosis through inhibition of sprouty1 (13). Consequently, knockdown of miR-21 induced myofibroblast apoptosis and reduced fibrosis during cardiac failure. On the other hand, our data show that overexpression of miR-21 could reduce fibrosis during ischemic heart disease or failure by decreasing myocyte cell death and, thus, inflammatory cell infiltration, and fibroblast proliferation.
In conclusion, we have outlined a unique aspect of the AKT survival pathway: one in which it regulates the extrinsic apoptotic pathway in cardiac myocytes via miR-21-dependent sup-FIGURE 6. miR-21 protects the heart against ischemic injury. a, a transgenic mouse model was generated with a 320-bp sequence encompassing the stem-loop of mouse miR-21 downstream of the ␣-myosin heavy chain promoter (␣MHC). Two lines were obtained and diagnosed by Northern blot analysis for miR-21 expression in the hearts of 10-week-old mice. b, miR-21-Tg and wild-type littermate mice were subjected to 45-min ischemia followed by 16-h reperfusion. Hearts were then perfused with Evans blue dye, fixed, sectioned, and stained with triphenyltetrazolium chloride. Shown are images from both sides of a representative section from each of the mice. c, the % area of risk (right) and % infarct zone/area at risk (left) were measured, averaged, and plotted (n ϭ 6). Error bars represent ϮS.E. and *, p Ͻ 0.05 versus wild type. d, 20-week-old transgenic and wild-type littermates were subjected to complete left coronary artery occlusion (CAO) or a sham operation for 4 weeks. The hearts were then isolated and sectioned to reveal the extent of left ventricular chamber dilatation (n ϭ 6, each). e, lung weight and heart weight/tibial length (HW/TL) were calculated and graphed. Error bars represent ϮS.D. and *, p Ͻ 0.01 versus sham. f, before sacrifice, cardiac wall and chamber dimensions, and functions, were assessed by echocardiography and hemodynamic measurements. *, p Ͻ 0.05 versus sham of matching genotype. # , p Ͻ 0.05 versus WT-CAO. g, hearts were fixed in formaldehyde, sectioned, and stained with Sirius Red for detection of collagen (red). The collagen was quantified (n ϭ 3, 3 sections each) and graphed as -fold increase in CAO hearts versus sham-operated ones. Error bars represent ϮS.D. and *, p Ͻ 0.001 versus sham, and # , p Ͻ 0.001 versus wild type-CAO. h, total protein was extracted from similarly treated mice groups and analyzed by Western blotting for the molecules listed on the left of each panel. i, 20-week-old miR-21-TG and wild-type littermates were subjected to CAO for 16 h. The hearts were isolated and fixed in formaldehyde and immunostained for FasL (purple). pression of FasL. Thus, in general, the discovery of miRNA, and their functions, is introducing a new dimension to our existing knowledge of signaling molecules and pathways that remains to be explored and exploited for more precise therapeutic targeting.