MiR-27a Functions as a Tumor Suppressor in Acute Leukemia by Regulating 14-3-3θ

MicroRNAs (miRs) play major roles in normal hematopoietic differentiation and hematopoietic malignancies. In this work, we report that miR-27a, and its coordinately expressed cluster (miR-23a∼miR-27a∼miR-24-2), was down-regulated in acute leukemia cell lines and primary samples compared to hematopoietic stem-progenitor cells (HSPCs). Decreased miR-23a cluster expression in some acute leukemia cell lines was mediated by c-MYC. Replacement of miR-27a in acute leukemia cell lines inhibited cell growth due, at least in part, to increased cellular apoptosis. We identified a member of the anti-apoptotic 14-3-3 family of proteins, which support cell survival by interacting with and negatively regulating pro-apoptotic proteins such as Bax and Bad, as a target of miR-27a. Specifically, miR-27a regulated 14-3-3θ at both the mRNA and protein levels. These data indicate that miR-27a contributes a tumor suppressor-like activity in acute leukemia cells via regulation of apoptosis, and that miR-27a and 14-3-3θ may be potential therapeutic targets.


Introduction
MicroRNAs (miRs) are ,22 nt, non-coding RNA molecules that play roles in most cellular processes, including apoptosis [1]. Dysregulation of miRs has been implicated in many disease states, prominently including cancer, where there are significant differences in miR expression profiles of cancer cells versus normal cells from the tissues of origin [2]. MiRs negatively regulate gene expression by binding to partially complementary sites (mediating mainly translational inhibition or mRNA destabilization) or, less frequently, fully complementary sites (mediating mainly mRNA degradation) [3,4]. These target sites can be located in the 39UTR, the coding region (CDS), and/or the 59UTR of target gene mRNA [3,4]. In cancer, a miR can function as either a tumor suppressor gene (e.g., miR-15a,16-1, let-7 family) or an oncogene (e.g., miR-155, miR-17,92) [5].
We and others have found that miRs play major regulatory roles in normal hematopoietic differentiation, evidenced by the discovery of a small set of hematopoietic stem-progenitor cell (HSPC)-expressed miRs (HE-miRs) which post-transcriptionally regulate specific mRNAs involved in hematopoiesis [6][7][8][9]. MiR-23a and miR-24 were identified as enriched in CD34+ HSPCs [6], and miR-24 has a well-defined role in as a regulator of normal erythropoiesis via targeting of human activin receptor type1, ALK4 [10]. Additionally, miR-23a and miR-24 both act as tumor suppressors and are down-regulated in multiple cancers. MiR-23a and miR-24-2 are coordinately transcribed and generally expressed at similar levels with a co-localized third miR, miR-27a. This ''miR-23a cluster'' (miR-23a,miR-27a,miR-24-2) is located on human chromosome 19p13.2, [11], ( Figure S1) and is expressed from its own upstream promoter, located in the 2600 to +36 bp region, which includes a GC-rich region and a transcription start site (0 to 124 bp) [12,13]. There is also a ''miR-23b cluster'' (miR-23b,miR-27b,miR-24-1), which is located on chromosome 9 in C9orf3. These 2 paralogous clusters are under different regulatory controls and hence are differentially expressed in cells [13], yet both the miR-23a and miR-23b clusters are under direct negative regulation by c-MYC [14]. Expression of all 3 miR-23a cluster members is down-regulated in acute promyelocytic leukemia (APML), colorectal cancer, oral squamous cell carcinoma, and prostate cancer, though there are rare cases where miR-23a cluster member expression is not correlated [13].
While the expression and effects of miR-23a and miR-24 are well-described in cancer and hematopoiesis, less is known regarding miR-27a. MiR-27a, in conjunction with RUNX1, regulates normal megakaryocytic differentiation [9]. MiR-27a is up-regulated in estrogen receptor-positive breast cancers (regulating ZBTB10) and gastric cancers [13], and down-regulated in malignant melanoma and the aforementioned cancers where cluster expression is decreased [13]. In this study, we determined the expression of miR-27a in CD34+ HSPCs and its expression Isolation of RNA and measurement of miR levels RNA was isolated using the miRNeasy kit (Qiagen, Valencia, CA) according to the manufacturer's protocol. Levels of mature miRs were measured in a 7900 Real-Time PCR System (Applied Biosystems, Inc, Carlsbad, CA) using TaqMan qRT-PCR (Applied Biosystems); U18 was chosen as an endogenous control based on its consistent expression across acute leukemia cell lines and primary samples. The expression of each miR in each sample was normalized to the average level of that miR in CD34+ HSPCs and expressed as fold difference (2 2DDCt 61 standard error of the mean, SEM). MiR levels were also measured by microarray analysis; RNA was isolated as above and 500 ng RNA per sample was hybridized to each GeneChip miRNA 2.0 Array (Affymetrix, Santa Clara, CA) which contained 46,228 probes for the miRs of 71 organisms. mRNA microarray expression analysis was performed by hybridizing 100 ng RNA to each GeneChip Human Gene 1.0 ST Array (Affymetrix), containing 764,885 probes representing 28,869 annotated genes. cRNA synthesis, labeling, hybridization to arrays, washing, staining, and signal amplification were performed according to manufacturer's instructions in the core facilities of the University of Maryland School of Medicine. Data analysis was performed using GeneSpring GX (Agilent Technologies, Palo Alto, CA); both miR and mRNA normalizations used the robust multi-chip average (RMA) method, generating log 2 transformed intensity values. Correlations of miR expression levels were determined by Pearson correlation; r.0.300 and p,0.05 were considered significant.
Determination of the effect of c-MYC on miR-23a cluster expression P493B cells were a generous gift from Chi van Dang (University of Pennsylvania) [14]. Cells were treated with doxycycline (Dox) for 48 hr, harvested, and total RNA isolated as above. MiR-23a cluster member expression levels were measured via qRT-PCR and normalized to levels in untreated cells (2 2DDCt 6SEM). Levels of miR-27a and c-MYC mRNA expression in acute leukemia cell lines and primary samples were measured via microarray and normalized as above. Correlations between cellular miR-27a and c-MYC levels were determined by Pearson correlation (r,20.3000). Cell lines were treated with 25 mM or 50 mM of the pharmacologic c-MYC inhibitor 10058-F4 (Calbiochem, Gibbstown, NJ) or with DMSO (vehicle) for 48 hr. Expression levels of miR-23a and miR-27a in total cellular RNA were assessed by TaqMan qRT-PCR, as above. c-MYC was knocked down in K562, MOLT16, and Karpas45 cells via transfection with 50 nM or 100 nM of the siGENOME SMART pool c-MYC siRNA (Dharmacon, Lafayette, CO) or 100 nM of control siRNA. After 72 hr, total RNA was harvested, and miR-23a and miR-27a expression levels analyzed via TaqMan qRT-PCR. Protein lysates were isolated to demonstrate down-regulation of c-MYC protein via Western blot (c-MYC antibody S-40, Santa Cruz Biotechnology (SCB), Santa Cruz, CA) with GAPDH (S-137179, SCB) as loading control.

Enforced expression of miR-27a
Expression of miR-27a was enforced using a dual promoter lentiviral vector (FUGW backbone) as described [16]. Cells were transduced with control virus (FUGW) or miR-27a-expressing virus (FUGW/miR-27a) at a multiplicity of infection (MOI) = 5 for 72 hr. Cells were plated in methylcellulose-containing colonyforming assay medium (StemCell Technologies, Vancouver, BC) in triplicate at 400 cells/ml/plate, or in suspension (1 cell/well) in a microtiter plate in a limiting dilution strategy. GFP+ colonies were selected via fluorescence microscopy and then expanded in suspension cultures for 7-10 days, and analyzed for GFP via fluorescence activated cell sorting (FACS) and miR-27a (qRT-PCR) expression.

Determination of cell viability and apoptosis
K562 and REH cell viability (by FACS side scatter vs. forward scatter) of FUGW-and FUGW/miR-27a-transduced colonies was measured weekly after isolation from colonies in methylcellulose cultures. Cell viability was also measured via propidium iodide (PI) staining; FUGW/miR-27a-or FUGW-transduced cells (10 6 cells/ ml media), derived from a clonal population and subsequently expanded for 7 days, were stained with 2 ml PI (500 mg/ml) and FACS-analyzed. Percent apoptosis in cells transduced and isolated as above was determined via correlated AnnexinV and 7-Aminoactinomycin D (7AAD) staining; ,10 6 cells were stained with AnnexinV and 7AAD (BD Biosciences Franklin Lake, NJ, according to manufacturer's protocol) and FACS-analyzed. Percent apoptosis was taken as the percent AnnexinV + /7AAD 2 signals and compared by the Student's t-test (n.3). Correlation between PI+ staining and GFP+ expression in Molt16 cells was determined by FACS analysis on cells isolated by limiting dilution. FACS was also used to determine the change in GFP expression over time in Molt16 cells.
Tet-inducible miR-27a construct pTRIPZ (Open Biosystems, Huntsville, AL), a Tet-On vector, was used to make K562 stable cell lines that express miR-27a inducibly. The DNA fragment encoding pre-miR27a was amplified from human CD34+ HSPC genomic DNA by PCR using the following primers: 59TACCTCGAGCTGAGCTCTGCCAC CGAGGAT and 39 TCAGAATTCGAGG CCAGGCAGCAG-GATGGC. This PCR product was cloned into the pTRIPZ vector using XhoI and EcoRI sites. The construct (TG-27a) and control (TG) vector were packaged into lentivirus according to the manufacturer's protocol; stable cell lines were also made according to the manufacturer's protocol. The effect of Dox on cell growth 48 hrs after treatment was determined with the CellTiter 96H AQueous One Solution Cell Proliferation Assay (Promega, Madison, WI).

Luciferase assays
Predicted miR-27a and miR-24 binding sites from the 39UTR or CDS of 14-3-3h, 14-3-3b, or 14-3-3f, along with ,50-100 surrounding bases from the genomic sequence on each side, were cloned into our luciferase (Luc) modified pcDNA3.1 plasmid, using EcoRI and XhoI as described [6] (Table S1). Because of difficulties cloning the miR-24 predicted binding sites in the CDS region, commercially synthesized oligonucleotides corresponding to the predicted binding regions flanked by 22 nt and with the precut cloning sites for EcoRI and XhoI on either side (Integrated DNATechnologies, Coralville, IA) were obtained. The sequence for each strand was as follows: Strand 1: 59-AATTCAGAAAGC-CAAACTCGCTGAGCAGGCTGAGCGATATGA TGA-TATGGC TGCAGCCATGAAGGCAG TC-39. Strand 2: 59-TCGAGACTGCCTT CATGGCTGCAG CCATATCATCA-TATCGCTCAGCCTGCTCAGCGAGTTTGGC TTTCTG-39 (Underlined bold bases correspond to ''pre-cut'' enzyme cloning sites). The 2 strands were annealed together following the Invitrogen annealing protocol. The annealed duplex was ligated into the Luc reporter plasmid. Deletion mutants were made via PCR with the QuickChange mutation kit (Stratagene, La Jolla, CA) according to the manufacturer's protocol (Table S1). 500 ng of the Luc plasmid or the comparable deletion mutant, was cotransfected (using Lipofectamine 2000, Invitrogen), with or without artificial miR-23a, miR-27a, or miR-24 mimics (25 nM, Dharmacon), into HEK293T cells. A b-galactosidase (b-gal) plasmid (50 ng) was co-transfected for normalization [6]. Cells were lysed 24 hr after transfection, and Luc and b-gal activity measured as described [6]. Significant differences in normalized Luc activity in the presence of the miR vs. normalized Luc activity with plasmid alone were determined by Student's t-test (n$3).

MiR-23a cluster expression levels are low in 90% of precursor-B ALL and T cell ALL cell lines and primary samples
In normal human CD34+ HSPCs, miR-23a, miR-27a, and miR-24 were expressed at high levels (Ct,29 by qRT-PCR), similar to the U18 endogenous positive control ( Figure S2A). However, in 86% of precursor B-cell acute lymphocytic leukemia (pre-B-ALL),100% of T-cell acute lymphocytic leukemia (T-ALL) cell lines and primary samples (18 of 21 and 11 of 11 samples respectively) and 42% of acute myeloid leukemia (AML) cell lines and primary samples (8 of 19 samples), miR-27a was expressed at levels $2 fold lower than in CD34+ cells ( Figure 1A-C). Similar trends were observed for miR-23a ( Figure S3A-C) and miR-24 ( Figure S3D-F). Consistent with these qRT-PCR results, miR-27a expression levels determined by microarray in a larger number of leukemia samples were significantly lower in pre-B-ALL and T-ALL cases than in CD34+ HSPCs; however, there were no significant differences in miR-27a, miR-23a, or miR-24 levels between CD34+ HSPCs and AML cases (Figure 2A-C). Expression levels of mature miR-27a, miR-23a and miR-24 were highly correlated ( Figure S2B-G), and miR-27b and miR-23b expression levels were significantly lower than miR-27a and miR-23a expression in CD34+ HSPCs in all but one acute leukemia tested ( Figure S4A, B), confirming a prior report that the miR-23b cluster is not expressed in hematopoietic cells [17].

Expression of miR-27a is regulated by c-MYC
MiR-23a and miR-23b expression levels were previously found to be directly negatively regulated by c-MYC [13]. To confirm that c-MYC regulation extended to the entire miR-23a cluster, we used the c-MYC-over-expressing cell line P-493B, a human B-cell lymphoma line engineered with a tet-off system [13]. MiR-23a, miR-27a and miR-24 expression levels were increased concordantly by Dox treatment of these cells ( Figure 3A). Consistent with these results, treatment with a pharmacologic c-MYC inhibitor (10058-F4) increased miR-23a and miR-27a expression in REH, KOPN8, and SUPB15 human pre-B ALL cell lines by 2-to 60fold in a dose-dependent fashion ( Figure 3B, C). However, miR-27a in K562 cells was not affected by the c-MYC inhibitor ( Figure 3C). Knock down of c-MYC in K562 cells did not affect miR-27a or miR-23a expression ( Figure 3D, E). However, we observed a 1.3-2-fold increase in miR-27a and miR-23a expression levels in Molt16 and Karpas 45 cells upon c-MYC knock down ( Figure 3D, E; Figure S5A; attempts to knock down c-MYC expression in REH, KOPN8, and SUPB15 cells were unsuccessful due to low transfection efficiencies). Finally, we observed a significant inverse correlation between c-MYC mRNA and mature miR-27a levels in human acute leukemias ( Figure S5B-D); a similar inverse correlation between c-MYC and miR-23a and miR-24 was also observed ( Figure S5E, F).

MiR-27a slows growth and increases apoptosis in acute leukemia cells
We examined functional effects of ''replacing'' miR-27a expression in K562 cells by lentiviral transduction. Cells transduced with FUGW/miR-27a (GFP+) had a distinct growth disadvantage compared to control-transduced cells; in a mixed population of GFP+ and GFP2 cells, the GFP+ population was reduced after 4 weeks ( Figure S6A). In multiple transduced K562 cell clones, the magnitude of growth inhibition correlated with GFP and miR-27a expression levels, and FUGW/miR-27a- Figure 1. Expression levels of mature miR-27a in AML, pre-B-ALL, and T-ALL cell lines and primary samples. Total RNA was isolated from human AML (A), pre-B-ALL (B), and T-ALL (C) cell lines (black bars) and primary cases (primary patient cases identified with unique numbers, white bars). MiR-27a expression levels were normalized to mean l evels in no rmal human CD34+ HSPCs (2 2DDCt = 0.975660.022, black line). A Student's t-test was used to determine the significance of the difference in mean miR-27a expression (6SEM) between each leukemia samples and normal CD34+ HSPCs; significance is indicated as p,0.05 (*). n$3 independent experiments in all cases except for the 10 primary AMLs which were analyzed in triplicate due to limited sample quantity. doi:10.1371/journal.pone.0050895.g001 MiR-27a Regulates 14-3-3h and Apoptosis PLOS ONE | www.plosone.org transduced K562 clones grew ,2-6-fold slower in suspension cultures than controls ( Figure 4A). Furthermore, no FUGW/miR-27a-transduced K562 clones could be passaged beyond ,5-6 weeks post-transduction. To confirm these results, we engineered a tet-inducible miR-27a expression plasmid (TG-27a). K562 cells stably transduced with this plasmid expressed detectable miR-27a only upon Dox treatment ( Figure 4B), and the extent of growth inhibition was dependent on Dox dose and consequent miR-27a expression level ( Figure 4C).
As K562 myeloid leukemia cells can generate a clonal population from a single cell, we assessed multiple stable FUGW/miR-27a and FUGW colonies. Viability of clonal populations from FUGW/miR-27a-transduced K562 colonies decreased over time as compared to cells grown from FUGWtransduced colonies which maintained viability ( Figure 4D). Concurrently, there were greater numbers of PI+ cells in FUGW/miR-27a-transduced clones than in control-transduced clones ( Figure 4E). There were also higher percentages of AnnexinV+/7AAD-cells in clones from FUGW/miR-27a-transduced K562 cells, as well as in FUGW/miR-27a clonal populations from other AML cell lines (TF1 and HL60), than in controls ( Figure 4F). We also investigated the functional effects of miR-27a in clonal populations derived from pre-B-ALL and T-ALL cell lines. FUGW/miR-27a-transduced REH pre-B-ALL clones had lower viability over time compared to FUGWtransduced clones, similar to our observations in K562 cells ( Figure 5A). FUGW/miR-27a REH clones also had higher percentages of AnnexinV+/7AAD-cells ( Figure 5B, Figure S6B). For FUGW/miR-27a-transduced Molt16 T-ALL cells, only a single 100% GFP+ clonal population was isolated, and could not be passaged beyond 2 weeks. The populations of FUGW/miR-27a-transduced Molt16 cells that grew were mixed GFP+/GFPpopulations; those populations with .70% GFP+ cells grew slowly (,1 passage/week), their percentage of PI+ cells correlated with GFP (i.e. miR-27a) expression ( Figure 5C, Table S2), and %GFP+ cells decreased over time, indicating a growth disadvantage ( Figure 5D). We were unable to isolate individual FUGW/miR-27a clones in Karpas 45 T-ALL cells, whereas control FUGWtransduced Karpas 45 clones grew normally.
To determine the miR binding sites necessary for regulation of 14-3-3b protein expression, we cloned the 2 miR-27a sites (individually, 14-3-3bL-1537 and 14-3-3bL-2061) and 4 of the 5 miR-24 sites (14-3-3bL-39UTR and 14-3-3bL-CDS) into our Luc reporter vector (Table S3, Figure S7B-G). Co-transfection of miR-27a plus 14-3-3bL-1537 or 14-3-3bL-2061 resulted in no significant Luc reduction compared to miR-23a or plasmid alone ( Figure 7E). In contrast, co-transfection of miR-24 plus 14-3-3bL-39UTR reduced Luc expression compared to miR-23a or plasmid alone. Deletion of the seed complement in site 1 had no effect on Luc expression; however, Luc inhibition was partially reversed in the presence of miR-24 upon deletion of the seed complement in both sites ( Figure 7F). Co-transfection of miR-24 plus 14-3-3bL-CDS resulted in significantly reduced Luc expression compared to miR-23a or plasmid only ( Figure 7F). Deletion of the seed bases in site 1 completely reversed inhibition of Luc activity; however, deletion of the seed in both sites resulted in some (p.0.05) retained Luc reduction in the presence of both miR-24 and miR-23a; this inhibition may be non-specific. The last miR-24 site was the least thermodynamically strong of the predicted binding sites, and as 3 of the 4 sites previously cloned demonstrated miR-24 binding, we did not clone this last predicted miR-24 site.14-3-3f and 14-3-3c each have multiple predicted miR-23a cluster binding sites (Table S3, Figure S7H-M). However, there was no significant

C) Untransduced K562 cells (WT), K562 cells transduced with a control tet-inducible plasmid, and K562 cells transduced with the TG-27a construct were treated with increasing doses of Dox (as indicated). Cell growth of control-or tet-inducible construct-transduced cells was normalized to that of untransduced K562 cells. (D) Viability of FUGW and FUGW/
miR-27a transduced stable K562 clones was measured at indicated times after isolation from methylcellulose colonies. Each data point represents 1 clonal population (i.e. cells grown up from 1 colony and followed over time). Statistical differences in viability over time in FUGW and FUGW/miR-27a K562 colonies were determined by a 1-way ANOVA test (p,0.0001*). (E) FUGW and FUGW/miR-27a K562 clonal populations were stained with PI and FACS-analyzed. Each data point on the graph represents percent PI of an individual clonal population. The mean (6SEM) of FUGW clones (n = 4) was compared to FUGW/miR-27a clones (n = 7), and significant differences were assessed via student's t-test (p,0.05*). (F) K562, TF1, and HL60 AML cells were stained with AnnexinV and 7AAD and FACS-analyzed. Each data point represents the % AnnexinV + /7AAD 2 population within an individual clone. Significant differences between FUGW and FUGW/miR-27a clones were compared as above. doi:10.1371/journal.pone.0050895.g004 inverse correlation between 14-3-3f mRNA expression and either miR-27a or miR-24 (data not shown); the same was true for 14-3-3c and miR-27a, with the exception of a slight inverse correlation in T-ALL samples, and for 14-3-3c and miR-24 (data not shown). K562 cells transfected with the miR-23a cluster mimics, alone or in combination, did not have decreased endogenous 14-3-3f or 14-3-3c protein expression compared to control ( Figure 8A, B). We cloned predicted miR-27a and miR-24 sites of 14-3-3f into our Luc reporter plasmid, and detected little or no reduction of Luc activity upon co-transfection of these plasmids with miR-27a or miR-24 ( Figure 8C). In addition, we examined the effect of the miR-23a cluster members on 14-3-3e protein expression (no predicted miR-23a cluster binding sites) in K562 cells; as expected, none of these miRs had a significant effect on 14-3-3e protein expression by Western blotting (data not shown).

Discussion
We found that miR-27a, like its coordinately expressed miRs, was expressed at high levels in normal human CD34+ HSPCs. MiR-27a expression strongly correlated with expression of miR-23a and miR-24 in CD34+ HSPCs, consistent with coordinated regulation of other miR clusters. The miR-23a cluster has been reported as either up-or down-regulated in disease, suggesting tissue specific regulation [13], and cooperative functioning of this cluster in physiology and pathophysiology [13]. Although there is a paralogous miR-23b cluster, it is expressed at very low levels in hematopoietic cells [17], as confirmed by our studies. Specifically, miR-27a was expressed $4-fold higher in CD34+ HSPCs.
In contrast to high expression of the miR-23a cluster in CD34+ HSPCs, thought to include the cells of origin of acute leukemias [23][24][25], expression of miR-27a was significantly decreased in 64-66% of acute leukemia cell lines and primary samples examined by qRT-PCR or microarray. 97% of samples with low miR-27a had significantly down-regulated miR-23a and miR-24. This downregulation was present in .80% of pre-B-ALL and T-ALL cell lines and primary samples tested, but in fewer AML samples, consistent with previous reports that the miR-23a cluster is more highly expressed in AML than ALL [26][27][28]. A recent study demonstrating that down-regulation of miR-27a expression in childhood pre-B-ALL correlates with disease relapse, and that patients achieving complete remission had high levels of miR-27a expression complement our findings [29]. This study also reported that patients expressing high levels of miR-27a had a higher relapse-free survival rate of than patients with low levels [29]. In addition, a recent analysis of .4,000 normal and neoplastic human samples reported that the miR-23a and miR-23b clusters MiR-27a Regulates 14-3-3h and Apoptosis PLOS ONE | www.plosone.org are among those genetically deleted in human cancers, including lung cancer, prostate cancer, and CLL [30].
We observed an inverse correlation between levels of c-MYC mRNA and miR-27a, miR-23a, and miR-24 expression in normal CD34+ HSPCs and acute leukemias, suggesting that c-MYC contributes to the regulation of the miR-23a cluster in the hematopoietic system. Gao et al [14] showed that c-MYC downregulated expression of miR-23a and miR-23b and bound to a cognate site upstream of the miR-23b locus in a B lymphoma cell line, others have provided evidence for a feedback loop involving direct negative regulation of c-MYC by miR-24 [13,31], and there is growing evidence for regulation of miRs by c-MYC as well as regulation of c-MYC by miRs in cancers including leukemia [32,33]. Using a conditional c-MYC expressing lymphoma cell line, we showed that miR-27a and miR-24 were coordinately regulated, along with miR-23a, by c-MYC. However, although our profiling data demonstrates coordinate regulation of all 3 cluster members, due to the fact that miR-24-2 and miR-24-1 have the same sequence, we cannot definitively say that miR-24-2 is the only contributor to this expression change in P-493B cells. In addition, a small molecule inhibitor of c-MYC, 10058-F4, which disrupts the interaction between c-MYC and its heterodimeric partner, MAX, thus preventing the complex from binding to a canonical E-Box DNA binding site, up-regulated miR-23a and miR-27a expression in 3 of 4 acute leukemia cell lines tested. One such site is present ,1000 bp upstream of the miR-23a cluster [11,[34][35][36]; ( Figure S1). Thus, evidence suggests that the miR-23a cluster is transcriptionally regulated by c-MYC and is not genetically deleted in the majority of acute leukemias. The dose of 10058-F4 used in our studies (25-50 uM) was chosen based on previously published experiments; functional studies in leukemia cell lines used 100 mM to inhibit c-MYC, and 60 mM of 10058-F4 HEK293T cells were co-transfected with 14-3-3hL-27a or 14-3-3hL-27a-del (500 ng, white bars) alone and with 25 nM of miR-23a (grey bars) or miR-27a (black bars) (top). Normalized mean Luc expression (6SEM) is represented on the graphs. (F) The 14-3-3h ORF rescues miR-27a-induced cell death. FUGW and FUGW/miR-27a transduced K562 clones were transfected with control plasmid (black bars) or 2 mg 14-3-3h ORF (white bars), and stained with PI. FUGW/miR-27a transduced clones expressing the 14-3-3h ORF had significantly less PI+ cells compared to those clones transfected with control vector. Significant differences were determined by Student's t-test (n$3, p,0.05*). doi:10.1371/journal.pone.0050895.g006 MiR-27a Regulates 14-3-3h and Apoptosis PLOS ONE | www.plosone.org was used to demonstrate that c-MYC inhibition increased levels of let-7a, let-7b, and miR-98 expression, mimicking the effects of siRNA knock down of c-MYC [37][38][39]. We verified the results using 10058-F4 by siRNA-mediated c-MYC knock-down in 3 cell lines, which resulted in a 2-fold increase in both miR-23a and miR-27a in 2 of 3 lines. Failure of either siRNA knock-down of c-MYC or 10058-F4 to affect miR-23a and miR-27a expression in K562 cells suggests additional mechanisms of miR-23a cluster regulation. Defined sites of DNA methylation are present in a 500 bp region upstream of the miR-23a cluster in K562 cells, suggesting possible epigenetic regulation in these cells [40]. Additionally, the miR-23a promoter contains a response element for the promyelocytic leukemia gene-retinoic acid receptor alpha (PML-RARA) fusion protein ( Figure S1), and chromosomal immunoprecipitation studies have identified involvement of PML, RARA and retinoid-X receptor alpha in regulation of this cluster [41]. Interestingly, qRT-PCR and microarray expression data indicated that miR-23a cluster member expression was not decreased in HL60, a myeloid cell line with some morphologic features of promyelocytic leukemia that lacks PML-RARA [41].
Based on the differential expression of miR-27a in leukemias versus CD34+ HSPCs, we hypothesized that miR-27a might have tumor suppressor functionality. To test this hypothesis, we ''replaced'' miR-27a in human acute leukemia cell lines. Constitutive expression of miR-27a at levels comparable to endogenous CD34+ HSPC levels resulted in immediate growth inhibition and cell death in AML, pre-B-ALL, and T-ALL cell lines. None of these cell lines survived for more than ,15 population doublings upon miR-27a replacement; in neither of the 2 T-ALL cells lines tested were we able to successfully isolate clonal populations of miR-27a-expressing cells.
We observed a moderate inverse correlation between mature miR-27a expression and 14-3-3h mRNA levels across the acute leukemias. Additionally, via reporter assays and Western blots, we demonstrated direct down-regulation of 14-3-3h by miR-27a, dependent on a 39UTR binding site. Lack of higher correlation between miR-27a and 14-3-3h mRNA expression may reflect a predominantly translational inhibitory mechanism [3,4]. To demonstrate a direct relationship between miR-27a-induced cell death/apoptosis and down-regulation of 14-3-3h, we overexpressed the 14-3-3h ORF (to render it ''insensitive'' to miR-27a). The ORF was able to partially rescue miR-27a-induced cell death increase, indicating that 14-3-3h contributes to miR-27a function, and that there are likely additional miR-27a target genes that play a role in miR-27a-induced cell death. In contrast, we were surprised to find that miR-27a does not regulate 14-3-3b, by Western blotting or Luc assays, despite 2 predicted binding sites in its 39UTR. However, we found miR-24 down-regulated 14-3-3b expression via predicted sites within the CDS and the 39UTR. When 14-3-3b is decreased, pro-apoptotic proteins, such as BAD, are released and translocate to the mitochondria to induce apoptotic signaling [18,50]. Apoptosis due to miR-24 mediated down-regulation of BCL2 has been reported and reduction of 14-3-3b further contributes to apoptosis. Regulation of 14-3-3b by a CDS miR site emphasizes the need to take the CDS into account when assessing miR targets. Similarly, Nanog, Oct4, and Sox2 have confirmed miR target sites in their CDSs [51]. A recent report [31] demonstrated that multiple miR-24-regulated genes have ''seedless'', but otherwise highly complementary binding sites. We were surprised to observe that neither miR-27a nor miR-24 regulated 14-3-3f or 14-3-3c, despite multiple predicted binding sites. This observation highlights the high false positive rate of miR target prediction algorithms. As expected, expression of 14-3-3e protein, our negative control containing no predicted miR-27a sites, was not affected by the miR-23a cluster.
In summary, we confirmed that miR-27a expression is downregulated in multiple acute leukemia cell lines and primary patient samples. Enforced expression (''replacement'') of miR-27a resulted in decreased cell growth and increased cell death due, at least in part, to an increase in cellular apoptosis. MiR-27a-induced apoptosis occurred via specific down-regulation of 14-3-3h, confirmed by the fact that functional effects were ''rescued'' by miR-27a-resistant 14-3-3h ORF. Further, binding and downregulation of 14-3-3b by miR-24, which was also significantly down-regulated in acute leukemias, indicates that a second miR-23a cluster member may have a similar functional effect, although we did not conduct ORF rescue experiments with miR-24 and 14-3-3b. Thus, miR-27a plays a tumor suppressor-like role in acute leukemias by regulating apoptosis, and miR-27a and 14-3-3h are potential target molecules for development of leukemia therapeutics. Figure S1 Schematics and mature sequences of the miR-23a and miR-23b cluster loci: The miR-23a cluster is located on the negative strand of human chromosome 19 (19p13.2), and has a defined transcription start site at 2600 to +32 bp upstream of the open reading frame. Within this region, there are also canonical c-MYC and PML-RARA binding sites defined (11)(12)(13)(14)41). The cluster is intergenic and lies between the genes ZSWIM4 and NANOS3 both of which are on the positive strand; miR-181c and miR-181d are located between the miR-23a cluster and NANOS3, also on the positive strand. Cluster organization is syntenic to both mouse and rat. The miR-23b cluster is located on human chromosome 9 (9q22.32) on the positive strand. The cluster is intragenic, located within the C9orf3 gene and is transcribed from its own promoter approximately 31 Kb upstream. The sequences of miR-23a and miR-23b differ by only one base (position 19, U vs. A) as do miR-27a and miR-27b (position 19, C vs. U); miR-24-2 and miR-24-1 have identical mature sequences. (TIF) Figure S2 MiR-23a cluster expression in normal CD34+ HSPCs. (A) Mature miR-23a (n = 33), miR-27a (n = 36), miR-24 (n = 24) and U18 control (n = 30) levels were determined using specific TaqMan qRT-PCR and presented as the raw Ct value. (B)-(G) Expression of miR-27a correlated with miR-23a and miR-24 expression in normal CD34+ HSPCs and acute leukemia cell lines and primary samples, as determined by qRT-PCR (B)-(D) and microarray analysis (E)-(G). Mature miR-27a, miR-23a and miR-24 expression levels were measured via qRT-PCR and are expressed as raw Ct values on the graphs. Correlation of miR-27a to miR-23a (B) (n = 108), miR-27a to miR-24 (C) (n = 99), and miR-23a to miR-24 (D) (n = 99) were both positive and significant (Pearson r.0.900, p,0.05*). Expression levels of mature miR-27a, miR-23a, and miR-24 were measured by microarray analysis and expressed as their normalized array intensities on the graphs.