MicroRNA-223 Regulates Cyclin E Activity by Modulating Expression of F-box and WD-40 Domain Protein 7*

F-box and WD-40 domain protein 7 (Fbw7) provides substrate specificity for the Skp1-Cullin1-F-box protein (SCF) ubiquitin ligase complex that targets multiple oncoproteins for degradation, including cyclin E, c-Myc, c-Jun, Notch, and mammalian target of rapamycin (mTOR). Fbw7 is a bona fide tumor suppressor, and loss-of-function mutations in FBXW7 have been identified in diverse human tumors. Although much is known about targets of the Fbw7 ubiquitin ligase pathway, relatively little is known about the regulation of Fbw7 expression. We identified a panel of candidate microRNA regulators of Fbw7 expression within a study of gene expression alterations in primary erythroblasts obtained from cyclin ET74A T393A knock-in mice, which have markedly dysregulated cyclin E expression. We found that overexpression of miR-223, in particular, significantly reduces FBXW7 mRNA levels, increases endogenous cyclin E protein and activity levels, and increases genomic instability. We next confirmed that miR-223 targets the FBXW7 3′-untranslated region. We then found that reduced miR-223 expression in primary mouse embryonic fibroblasts leads to increased Fbw7 expression and decreased cyclin E activity. Finally, we found that miR-223 expression is responsive to acute alterations in cyclin E regulation by the Fbw7 pathway. Together, our data indicate that miR-223 regulates Fbw7 expression and provide the first evidence that activity of the SCFFbw7 ubiquitin ligase can be modulated directly by the microRNA pathway.

The SCF Fbw7 ubiquitin ligase regulates the stability of multiple oncoprotein substrates, including cyclin E, c-Myc, Notch, c-Jun, and mammalian target of rapamycin (mTOR) (1)(2)(3)(4)(5)(6)(7)(8). Recognition of substrates by SCF Fbw7 is promoted by phosphorylation of these proteins at a conserved phosphoepitope, termed the Cdc4 phospho-degron (CPD), 3 which directly interacts with the WD-40 domain of Fbw7 (also known as human Cdc4 or hCdc4) (9,10). Loss-of-function mutations in Fbxw7 cause both hematopoietic and solid organ tumors in mouse models (11)(12)(13). Additionally, Fbw7 is frequently mutated in diverse human cancers, with most mutations introducing amino acid substitutions at the WD-40 domain (14). Thus, Fbw7 is a tumor suppressor, and the abundance of hotspot mutations in human cancers targeting the Fbw7 substrate-binding domain supports the notion that impaired substrate degradation mediates the oncogenicity associated with Fbw7 loss.
Three Fbw7 isoforms with distinct subcellular localizations result from alternative splicing of transcripts made from the FBXW7 gene (15). Currently, little is known about how Fbw7 expression is regulated. The cytoplasmic isoform of Fbw7 (Fbw7␤) is induced by p53, whereas expression of the nuclear (␣) isoform is repressed by UV radiation (16,17). Additionally, impaired mitochondrial oxidative function induces the Drosophila FBXW7 homolog, ago, by a p53-dependent mechanism (18). Other than these reports, most current understanding of how Fbw7-mediated substrate degradation is regulated centers on mechanisms by which Fbw7-substrate interactions are driven (14).
In the case of cyclin E, its degradation via the SCF Fbw7 is directed by phosphorylation within two distinct CPDs based around threonines 62 and 380 (in human cyclin E, Thr-74 and Thr-39 in murine cyclin E (7,8,19,20)). Fbw7 is necessary for the maintenance of cyclin E periodicity in normal cell cycles. In the absence of a functional Fbw7 pathway, caused either by mutating cyclin E at its CPDs or by deleting FBXW7, cyclin E activity is pathologically sustained throughout the cell cycle, leading to aberrant S-phase progression, delayed mitotic exit, and genome instability (21)(22)(23)(24). In vivo, impaired Fbw7-mediated cyclin E degradation causes increased cell proliferation and impaired maturation in a cell type-specific manner as well as increased tumorigenesis (21,(25)(26)(27). The developmental abnormalities associated with impaired Fbw7-mediated regulation in a cyclin E T74A T393A knock-in mouse model are especially pronounced in erythroid cell precursors, which exhibit hyperproliferation, increased apoptosis, and dysplastic morphologies (25).
In this study, we profiled alterations in microRNA expression in erythroblasts obtained from cyclin E T74A T393A knock-in mice and identified down-regulated expression of a number of candidate regulators of Fbxw7 gene expression. We found that two microRNAs, miR-25 and miR-223, when overexpressed, reduce FBXW7 mRNA levels and the activation of a reporter linked to the 3Ј-untranslated region (3Ј-UTR) of the FBXW7 gene. However, only miR-223 significantly increases endoge-nous cyclin E protein and activity levels and increases genomic instability. Conversely, reduced miR-223 expression in primary mouse embryonic fibroblasts results in increased Fbw7 protein expression and decreased cyclin E activity. Finally, we found that miR-223 expression responds to alterations in cyclin E protein levels as regulated by the Fbw7 pathway, suggesting that miR-223 regulates Fbw7 expression as part of a homeostatic response to acute fluctuations in cyclin E activity.

EXPERIMENTAL PROCEDURES
Mice-The cyclin E T74A T393A knock-in mouse has been previously described (25). Wild-type littermate controls (129Sv background) were used for comparison. Mice were housed in barrier facilities, and the Northwestern University Animal Care and Use Committee (ACUC) approved all animal studies. Euthanasia was performed by CO 2 asphyxiation according to ACUC guidelines. Murine bone marrow cells were harvested by grinding femurs and tibia using a mortar/pestle, suspending cells in 2% fetal bovine serum (FBS), and filtering suspensions through 70-micron nylon cell strainers.
Cell Culture-Mouse embryo fibroblasts were obtained by standard methods at days 12-14 postcoitum. (25) and used within two passages. HCT116 and K562 cells were obtained from ATCC. FBXW7-null HCT116 cells were provided by Drs. Jon Grim and Bruce Clurman, Fred Hutchinson Cancer Research Center. All cells were grown in ATCC-recommended media supplemented with 10% fetal calf serum.
Vectors, Transfections, and Retroviral Transduction-The pBabe-cyclin E (T380A) retroviral construct was previously described (23). A retroviral construct encoding human FBXW7 was made by subcloning the FBXW7␣ reading frame into pRetroX-IRES-DsRedExpress (Clontech), allowing transduced cells to be sorted based on fluorescence signal. The retroviral construct expressing miR-223 and green fluorescent protein (GFP) was obtained from Dr. Chang-Zheng Chen (Stanford University) via Addgene (30). Viral supernatants were prepared by transfecting Phoenix cells using the calcium phosphate precipitation method, and target cells were transduced with filtered supernatants with 4 g/ml Polybrene. All microRNAs and antagomir transfections, including co-transfections with plasmid DNA, were performed by electroporation (Amaxa Nucleofector, Lonza Cologne AG) of 150 pmol pre-miRNAs/ 1 ϫ 10 6 cells (Ambion). Transfection efficiency was monitored using a co-transfected GFP-expressing plasmid. For reporter assays, the 3Ј-UTR of FBXW7 was cloned into the pRL-TK vector (Promega) for luciferase expression and into pFLAG-FBXW7␣ (4) for immunoblot. Mutations in the predicted target sites for either miR-25 or miR-223 were introduced into the 3Ј-UTR of FBXW7 via oligonucleotide-directed mutagenesis, which created three consecutive nucleotide substitutions at the center of each 8-nucleotide miRNA targeting site.
Isolation of RNA and Quantitative RT-PCR-For mRNA assays, RNA was isolated using RNeasy mini kit (Qiagen) and reverse transcribed using the AffinityScript quantitative PCR cDNA synthesis kit (Agilent Technologies). TaqMan probes for cyclin E1 and FBXW7 detection were purchased from Applied Biosystems. Expression levels of target transcripts were normalized to 18 S RNA. For miRNA analysis, total RNA was isolated using the mirVana miRNA isolation kit (Ambion) and reverse transcribed using the TaqMan microRNA reverse transcription kit (Applied Biosystems). TaqMan rodent microRNA Megaplex assays were used according to the manufacturer's instruction. Individual assays for microRNA detection were purchased from Applied Biosystems. Expression levels of miRNAs were normalized to U6 or sno234. All quantitative real-time PCRs were performed using the ABI 7900HT Fast real-time PCR system (Applied Biosystems).
Flow Cytometry and Cell Cycle Analysis-CD71 high /Ter-119 high erythroblasts were isolated by FACS on a MoFlo cell sorter (Dako), as were cells transfected or transduced with fluorescent markers. For cell cycle studies, asynchronous mouse embryonic fibroblasts (MEFs) or synchronized MEFs were trypsinized at the indicated times, fixed in 70% ethanol, washed, and incubated in PBS containing propidium iodide and RNase A. Analyses were performed on a BD Biosciences LSR2 and using the Multicycle software. Statistical analyses of cell cycle alterations (one-way analysis of variance) or cell cycle phasespecific gene expression changes (Student's t test) were performed using the Prism 5 software (GraphPad).
Pulse-Chase-HCT116s (in 6-cm dishes) were transduced with the indicated vectors, starved for 30 min in met-free Dulbecco's modified Eagle's medium (Invitrogen Gibco) containing 5% dialyzed FCS, and labeled with 1 ml of trans-35 S label (400 mCi/ml, MP Biomedical) for 30 min. Dishes were washed, "chased," and at the indicated time points, washed with cold PBS and then stored at Ϫ80 degrees prior to lysis and immunoprecipitation.
Kinase Assays-For kinase assays using purified histone subunit H1 (Sigma) as substrate, human cyclin E was immunoprecipitated from extracts normalized for total protein (Bradford protein assay, Bio-Rad) with the HE111 monoclonal antibody (Santa Cruz Biotechnology), and mouse cyclin E was immunoprecipitated using the affinity-purified polyclonal antibody. Buffer conditions for measuring cyclin E kinase activity in vitro are previously described (29). Quantitation of autoradiographs was performed using PhosphorImager and ImageJ software.
Micronucleation Assay-Cultured HCT116 cells were trypsinized and washed in PBS, and nuclei were fixed in 1:3 acetic acid:methanol and dropped onto slides at an approximate density of 5 ϫ 10 4 nuclei/slide. Nuclei were hybridized with a pancentromeric, FITC-labeled FISH probe (ID Labs) according to the manufacturer's protocol and counterstained (VECTASH-IELD with DAPI, Vector Laboratories). Using fluorescence microscopy, micronuclei were counted as described. Statistical significance was determined using one-way analysis of variance calculations (Prism 5).
Luciferase Assays-For the luciferase assays, 1 ϫ 10 6 K562 or HCT116 cells were co-transfected in 6-well plates with 700 ng of Renilla reporter plasmid, 100 ng of pGL3c vector, expressing firefly luciferase (Promega), and 150 pmol of each pre-miRNA. Two days after transfection, cell luminescence was measured using a Dual-Luciferase reporter assay kit (Promega) and a Moonlight 2010 luminometer (Analytic Luminescence Laboratory). Renilla reporter activities were normalized based on firefly luciferase expression (31).

A Screen for microRNAs Dysregulated in Cyclin E T74A T393A Erythroblasts Reveals Candidate Regulators of Fbw7
Expression-To identify alterations in gene expression associated with the defective maturation phenotype in cyclin E T74A T393A erythroid cells, we profiled global gene expression changes associated with dysregulated cyclin E, including for non-coding RNAs, for which we employed a multiplex real-time PCR (RT-PCR) assay. This multiplex assay allowed us to compare the expression of up to 518 microRNAs in primary erythroblasts of cyclin E T74A T393A mice with wild-type, littermate controls. We found 63 microRNAs whose expression was quantifiable in both the wild-type and the knock-in erythroblasts and determined to be at least 2-fold lower in the knock-in cells ( Fig. 1A and supplemental Fig. 1). We used two target prediction tools, TargetScan (32) and PicTar (33), to help elucidate the functional significance of misregulating these microRNAs, by identifying their potential mRNA targets.
We noted that seven microRNAs down-regulated in the cyclin E knock-in erythroblasts are known to regulate inhibitors of cyclin/CDK activity: miR-222, which targets p27 Kip1 and p57 Kip2 (34 -36), five members (miR-17, miR-19a, miR-19b, miR-20a, and miR-92a) of the miR-17-92a cluster, which target both p21 Cip1 and p130 (37,38), and miR-26a, which targets retinoblastoma (39) (supplemental Fig. 1). We then hypothesized that among the microRNAs that are down-regulated in response to high cyclin E expression in primary erythroblasts would be one or more that regulate expression of Fbw7 itself. We identified four candidate microRNA regulators of Fbw7 from among those whose expression was significantly altered in the knock-in cells based on target sites predicted to lie within the FBXW7 3Ј-UTR: miR-25, miR-27a, miR-92a, and miR-223.
We first verified that these microRNAs were down-regulated in cyclin E T74A T393A erythroblasts, using separate quantitative RT-PCR assays (Fig. 1B). Next, we tested the ability of these microRNAs, when overexpressed individually in human proerythroblast K562 cells by transient transfection of synthesized, precursor microRNAs (pre-miRs), to repress the activity of a bioluminescent reporter linked to the 3Ј-UTR of FBXW7. We found that miR-25, miR-27a, and miR-223 effectively repressed reporter activity, whereas miR-92a did not (Fig. 1C).
Predicted MicroRNA Binding Sites Are Confirmed within the FBXW7 3Ј-UTR-We further characterized the functional consequences of microRNA-mediated down-regulation of the Fbw7 ubiquitin ligase pathway in HCT116 cells. We chose this cell line because it has been used extensively to study the Fbw7 pathway, with Fbw7-null HCT116 cells available (22,40), and because we achieve uniformly high (Ͼ90%) transfection efficiencies with these cells. Using the luciferase reporter construct containing the FBXW7 3Ј-UTR, we confirmed that expression of miR-25, miR-27a, and miR-223 repressed reporter activity in HCT116 cells, comparable with our results in K562 cells (supplemental Fig. 2). We then introduced mutations into the predicted binding sites for miR-25 and miR-223. Because miR-27 overexpression elicited neither down-regulation of FBXW7 mRNA levels nor a significant effect on Fbw7 regulation of cyclin E (Figs. 2C and 3A), we chose not to further characterize its binding sites within the FBXW7 3Ј-UTR.
We confirmed both the predicted binding sites for miR-25 and miR-223 within the FBXW7 3Ј-UTR by showing that mutation of each of these sites relieved repression of reporter activity by the corresponding miRNA but not the other (Fig. 2, A and B). Thus, miR-25 and miR-223 are able to bind sites within the FBXW7 3Ј-UTR.
We found that overexpression of either miR-25 or miR-223 caused substantial down-regulation of FBXW7 mRNA levels, whereas miR-27a expression only modestly reduced FBXW7 message (Fig. 2C). We then measured Fbw7 protein abundance in cells transfected with miR-25, miR-27a, or miR-223. Because A, schematic of screen for dysregulated microRNAs in sorted basophilic erythroblast cells of the indicated immunophenotypes, isolated from two adult, homozygous cyclin E T74A T393A mice and two wild-type, littermate controls. The screen utilized a multiplex RT-PCR-based assay allowing the expression of 518 unique microRNAs to be measured via card array format. B, down-regulated expression of four candidate microRNAs with predicted target sites within the FBXW7 3Ј-UTR was independently verified, using individual quantitative RT-PCR assays. Displayed is a representative comparison of microRNA expression in primary erythroblasts isolated from two wild-type (light bars) and cyclin E T74A T393A mice. Error bars indicate standard deviations with triplicate measurements. C, candidate microRNA regulators of Fbw7 expression were validated in a Dual-Luminescence reporter assay performed in K562 cells, utilizing a reporter construct containing the full-length 3Ј-UTR of FBXW7 downstream of the Renilla luciferase reading frame. Normalized Renilla luciferase activities for the four candidate microRNAs, transfected as pre-miRs, are shown as ratios to the normalized activity found with expression of a control miR sequence. Error bars indicate standard deviations with separate transfections.  (40), we utilized a reporter construct, which encodes the FBXW7␣ cDNA with an N-terminal epitope tag and the fulllength 3Ј-UTR of FBXW7. We found that miR-223 overexpression led to significant reduction in ectopic Fbw7␣ abundance. Despite repressing the activity of the FBXW7 3Ј-UTR reporter and mRNA levels, miR-25 did not significantly reduce Fbw7 protein abundance (Fig. 2D). This discrepancy either could be due to an indirect effect of miR-25 overexpression on FBXW7␣ translation and/or protein stability or it may reflect that miR-25 shows less robust targeting of FBXW7 mRNA than miR-223 in our assay.
MiR-223 Overexpression Causes Dysregulation of Cyclin E Activity and Abundance-The most readily identifiable abnormality in HCT116 cells with reduced Fbw7 expression is significantly increased total and phosphorylated cyclin E and cyclin E kinase activity (40). Because the Fbw7 pathway targets active, phosphorylated cyclin E for ubiquitin-mediated proteolysis, measuring both total and phosphorylated cyclin E protein and cyclin E kinase activity provides a readout of Fbw7 function (40). We found that overexpression of miR-223 in HCT116 cells produced the most significant increases in both total and Thr-380-phosphorylated cyclin E and cyclin E kinase activity, indicative of stabilization of the catalytically active, Cdk2-bound fraction of cyclin E (Fig. 3A, left panel). In contrast, overexpression of miR-25 or miR-27 did not result in statistically significant increases in cyclin E kinase activity (p values calculated from data obtained from multiple independent experiments). Cyclin E mRNA levels were not significantly altered by the overexpressed microRNAs (Fig. 3A,  right panel), consistent with the idea that the observed increase in cyclin E abundance and activity with miR-223 overexpression is associated with stabilization of the catalytically active protein fraction. We next measured cyclin E abundance and activity in Fbw7-null HCT116 cells transfected to overexpress miR-25 and miR-223 and found that, unlike the parental cells, there was no further increase in either cyclin E protein or kinase activity with expression of either microRNA (Fig. 3B). This finding further supports the hypothesis that overexpression of miR-223 increases cyclin E activity specifically by inhibiting the Fbw7 pathway. To demonstrate directly that cyclin E degradation is impaired by increased miR-223 expression, we measured endogenous cyclin E protein half-life by pulse-chase in HCT116 cells transduced with either control or miR-223-expressing retroviruses. We found a 2-fold increase in endogenous cyclin E half-life in miR-223-expressing HCT116 cells, consistent with down-regulation of Fbw7-mediated degradation (Fig. 3C).
Previous studies have identified genomic instability as a hallmark of dysregulated cyclin E kinase activity (23,41). Moreover, Fbw7-null HCT116 cells show evidence of genomic instability, which is dependent on cyclin E (22). We employed a micronucleation assay to measure genome instability in cells expressing miR-25 and miR-223. Quantitative micronucleation assays consistently have been found to be concordant with results of chromosome enumerator and metaphase analyses for scoring cyclin E-associated genome instability (22,23,42). We found that miR-223 expression resulted in the greatest amount of micronuclei formation in HCT116 cells when compared with that seen with either miR-25 or the control microRNA, consistent with our finding that miR-223 produces the largest increase in cyclin E kinase activity (Fig. 3D).
Reducing miR-223 Levels Down-regulates Cyclin E Activity in Primary Fibroblasts-To further address the role microRNAs play in the physiologic regulation of Fbw7 expression and activity, we studied the consequences of reducing miR-25 and miR-223 levels in cells. Because HCT116 cells express very low levels of endogenous miR-223, we performed these experiments in primary MEFS, which we found expressed higher levels of miR-223 and readily detectable miR-25 as well (Fig. 4A).
Another advantage to using primary MEFs is that they can be synchronized in G 0 -phase to study miR-25 and miR-223 levels in relationship to Fbw7 expression during cell cycle progression. MEFs were made quiescent by serum starvation and then released into serum using a method that yields populations of cells in G 1 -phase (12 h after release), early S-phase (18 h), and late S/early G 2 -phase (24 h) (25). We identified a significant decrease in miR-223 levels from G 0 -to G 1 -phase, which is associated with an increase in Fbxw7 mRNA levels during the same transition. In contrast, miR-25 levels remained relatively constant during cell cycle progression (Fig.  4B). These data are consistent with the hypothesis that endogenous miR-223 modulates Fbw7 expression. To test this directly, we utilized complementary, single-stranded RNA analogs ("antagomirs") to down-regulate the expression of endogenous microRNAs. Using transfected antagomirs, we were able to reduce endogenous levels of miR-25 and miR-223 by ϳ50% (Fig. 4C). We then utilized the reporter construct for measuring Fbw7 protein in antagomir-expressing MEFs. We found that expression of antagomirs targeting endogenous
Next, we studied the effect of miR-25 and miR-223 reduction in MEFs on endogenous cyclin E expression and activity. In MEFs, due to the high abundance of p21 Cip1 bound to cyclin E-Cdk2, which renders a large pool of cyclin E insensitive to Fbw7, we measured changes in cyclin E kinase activity to discern changes in the catalytically active, Cdk2-bound cyclin E fraction that is susceptible to the Fbw7 pathway (20,21). With expression of an antagomir targeting miR-223 (anti-miR-223), we identified a moderate but statistically significant (40%) reduction in endogenous cyclin E kinase activity and less of an effect with down-regulation of miR-25, which did not achieve statistical significance (Fig. 4D, left panel). We also observed a modest increase in the number of anti-miR-223-expressing cells in G 1 -phase and significantly decreased frequencies of cells in S-and G 2 /M-phases when compared with control anti-miR-and anti-miR-25-expressing cells (Fig. 4D, right panel). These cell cycle distribution data are consistent with a decrease in cyclin E kinase activity because acute reduction in cyclin E is known to prolong G 1 -phase (43). Together, data from these experiments support the concept that endogenous miR-223 can regulate Fbw7 expression sufficiently to potentiate cyclin E activity in primary fibroblasts.

MiR-223 Levels Are Responsive to Acute Changes in Cyclin E and Fbw7
Expression-In considering the candidates from our screen of cyclin E T74A T393A erythroblasts for Fbw7-regulating microRNAs, we hypothesized that expression of a physiologically relevant regulator of Fbw7 would itself be responsive to acute alterations in cyclin E levels as a means of maintaining homeostasis during normal cell cycle progression. Thus, we measured the levels of all four candidate microRNAs, which were originally nominated in our screen of cyclin E T74A T393A erythroblasts, in K562 cells stably transduced either with an Fbw7-resistant cyclin E (T380A) mutant (23) or with a control vector. Only miR-223 levels decreased significantly in response to dysregulated cyclin E expression, in contrast to miR-25, miR-27a, and miR-92a. Thus, the down-regulated expression of these other microRNAs in cyclin E T74A T393A erythroblasts is likely due to an indirect consequence of expression of the knock-in allele rather than a direct result of increased cyclin E activity. A, left panel, HCT116 cells were transfected to express the indicated microRNAs, and endogenous total and threonine 380-phosphoryated cyclin E (T380-P Cyclin E) protein and activity levels were assayed by immunoblotting and H1 kinase assays. In this experiment, to demonstrate internal consistency with our transfection results, miR-27a and miR-27b were expressed separately; these are encoded by distinct pre-miRs, which are processed to yield miRs that differ only by a single nucleotide. Endogenous cyclin E kinase activities were quantified using a PhosphorImager and are indicated, and p values were calculated from three independent experiments (Student's t test, n.s., not significant). Right panel, expression of cyclin E1 in pre-miR-transfected HCT116s was measured using quantitative real-time PCR. B, parental and Fbw7-null HCT116 cells were transfected and endogenous cyclin E protein and activity measured as in A. We verified equivalent transfection efficiencies of 90% in both parental and Fbw7-null HCT116s. C, HCT116 cells were transduced with GFP-expressing retroviral vectors encoding control sequence or miR-223. Fluorescence-activated cell sorts were performed followed by metabolic labeling and endogenous cyclin E half-life determination. Both primary data and quantitation of S 35 -labeled cyclin E protein decay with calculated half-lives are shown. D, micronucleation was measured in HCT116 cells fixed 72 h following transfection with the indicated pre-miRs. Micronuclei were compared among the cell populations after counting 1000 cells/sample in two separate experiments and expressed as ratios to normal nuclei. Error bars indicating standard deviations and p values are displayed. NOVEMBER 5, 2010 • VOLUME 285 • NUMBER 45

JOURNAL OF BIOLOGICAL CHEMISTRY 34443
Conversely, we overexpressed Fbw7␣ in K562 cells and found that miR-223 levels increased in response to enforced Fbw7-mediated cyclin E destruction, whereas expression of the other microRNAs in our panel of candidates did not change (Fig. 5B). Together, these data are consistent with a model in which miR-223 acts as a rheostat that regulates

miR-223 Regulation of Fbw7 and Cyclin E
cyclin E activity through its ability to modulate Fbw7 expression. Thus, we propose that miR-223 is part of a feedback loop connecting cyclin E activity to the regulation of Fbw7 expression (Fig. 5C).

DISCUSSION
A growing body of evidence highlights the importance of microRNAs in regulating the expression of mediators of cell cycle progression (34 -39, 44 -47). A theme emerging from these studies is that microRNAs participate in feedback or feedforward circuits to provide bistability for key transition points in the cell cycle (44 -46). The expression of cyclin E is under partial control of a microRNA-regulated feed-forward loop, whereby miR-15a, which represses cyclin E expression, is itself negatively regulated by a protein kinase C ␣-dependent mechanism (46).
We observed that a number of microRNAs known to regulate inhibitors of cyclin/CDK activity, including p21 and p27, were down-regulated in cyclin E T74A T393A erythroblasts. When compared with wild-type erythroblasts, the knock-in cells express higher levels of these CDK inhibitors. 4 These data are consistent with the hypothesis that dysregulated cyclin E activity provokes a physiologic response, directed at least in part by multiple microRNA-regulated mechanisms aimed at repressing inappropriately high cyclin E activity levels. We hypothesized that microRNAs regulating Fbw7, which regulates the destruction of catalytically active cyclin E, would similarly be down-regulated as part of the physiologic response to dysregulated cyclin E activity in cyclin E T74A T393A erythroblasts. In knock-in erythroblasts, however, the alanine substitutions at the cyclin E CPDs render the mutant cyclin E unresponsive to increased Fbw7 expression.
In this study, we found that miR-223 overexpression causes increased cyclin E abundance, stability, and activity and cyclin E-associated genome instability. Although overexpression of miR-25, another candidate Fbw7 regulator identified in our screen, elicited potent repression of the FBXW7 3Ј-UTR reporter activity and reduced FBXW7 mRNA levels, its effects in most of our assays of Fbw7 function did not achieve statistical significance. The regulation of cyclin E by miR-223 is Fbw7-dependent, as evidenced by the lack of impact of miR-223 overexpression on cyclin E abundance and activity in Fbw7-null cells. In contrast, acute depletion of miR-223 in primary cells leads to decreased cyclin E activity. Recently, miR-223 has been identified as an E2F transcriptional target (48); therefore, E2F activity likely links the effect of acute alterations in cyclin E activity to changes in miR-223 expression in the feedback circuit we propose (Fig. 5C). Indeed, this circuit should reinforce another negative feedback mechanism that ties cyclin E activation to its destruction via Cdk2 autophosphorylation within the cyclin E CPDs, which drives the association of cyclin E with the SCF Fbw7 ubiquitin ligase (29,49).
Another possible utility for microRNA regulation of SCF Fbw7 activity would be to provide cells with a means of 'fine-tuning' individual Fbw7-mediated substrate degradation pathways in response to various stimuli. Importantly, Fbw7-dependent deg-radation of many of its proto-oncogenic substrates utilizes the same kinase, glycogen synthase kinase 3, to drive the substate-Fbw7 interaction by phosphorylating CPDs (4 -6, 20). Thus, regulation of individual degradation pathways could be directed at the SCF Fbw7 rather than toward substrate phosphorylation. We speculate that miR-223-mediated down-regulation of Fbw7 synthesis could establish windows of opportunity for individual protein substrates of the SCF Fbw7 , which are synthesized themselves through distinct gene expression programs, to accumulate. Importantly, our experiments have focused solely upon cyclin E, downstream of Fbw7 modulation. Further work will be necessary to develop a more comprehensive understanding of the physiologic relevance of microRNA-mediated regulation of Fbw7 and its various substrates in different cell-and tissue-specific contexts.
Notably, miR-223, which has been implicated as part of the myeloid cell differentiation program (50 -52), is overexpressed in solid tumors, including ovarian and bladder cancers (53,54). Although FBXW7 is one of several identified targets of this microRNA (51,55,56), the potency of Fbw7 as a tumor suppressor and the frequency at which it is mutated in human cancers suggests that its down-regulation through miR-223 overexpression could be an important event in the development of some cancers. For example, a recent report has described increased expression of miR-223 in some T-cell acute lymphoblastic leukemia cases, a leukemia subtype in which FBXW7 loss-of-function mutations are frequently found (14,57). Possibly, miR-223 dysregulation could provide a means for impairing the tumor suppressor activities of Fbw7 in cell transformation without direct mutation of FBXW7. Further study of miR-223 in animal models and of the relationship between miR-223 expression and that of Fbw7 and its substrates in primary tumors will clarify its relevance in regulating the tumor suppressor functions of the SCF Fbw7 ubiquitin ligase.