FOXO1 regulates expression of a microRNA cluster on X chromosome

Introduction

Activation of the signal transduction cascade that proceeds through phosphoinisitol-3-phosphate kinase and protein kinase B (also known as AKT) is a frequent theme in signaling by various growth factors and a common subject of oncogenic alteration in cancer cells [1, 2]. A detailed knowledge of the consequences of activation of this pathway is essential for a better understanding of the function of a normal cell and may provide insights into the nature and possible mitigation of various disease states, including cancer, diabetes, and many others. Activation of AKT effects cellular changes through two broad groups of mechanisms. It modulates a large cohort of biochemical processes in a manner independent of de novo RNA synthesis. It also affects the levels of various cellular components through controlling the functions of transcription factors. Arguably, the best characterized AKT targets among the transcription factors are the proteins from the FOXO family [3]. AKT phosphorylates FOXO proteins on a series of highly conserved sites, leading to nuclear exclusion and inactivation of these molecules [4]. It is commonly accepted that many of the hallmark effects of AKT activation are directly attributable to FOXO inhibition, and that reactivation of FOXO influences cell response to therapies that target PI3K pathway [5]. This highlights the need to identify and investigate the targets of these transcription factors, especially in the view of the role of PI3K-AKT-FOXO connection in cancer, ageing and diabetes.

Traditionally, the focus of such research was on protein-coding genes, including the loci that control cell proliferation and survival. However, there is a growing appreciation of the role that non-protein coding RNAs play in the control of cellular processes. While some of these RNAs are direct components of signal transduction machinery [6], the best studied class are the so-called “micro RNAs” (miRNAs), which control translation of respective target mRNAs, and through that - a multitude of processes, such as cancer, senescence and differentiation [7-10]. In mammalian cells, typical miRNAs are first produced as imperfect hairpins within longer transcripts. Subsequent stepwise nucleolysis yields an imperfect duplex approximately 22 nucleotides in length, and one strand of this duplex is eventually incorporated into a translation silencing complex, where it serves as the determinant of specificity. Potentially, a single miRNA may target multiple mRNAs, which share with it short stretches of partial homology. There are also reports of miRNAs serving as components of multimolecular complexes that target gene expression at a promoter level (e.g.[11]).

It is increasingly clear that for most, if not all, of the mammalian transcription factors the understanding of the biological role remains incomplete until their effects on the miRNA expression are taken into account [12]. Furthermore, miRNAs are relatively well preserved in various biological samples and could be used as convenient reporters for the biological processes that regulate their expression (e.g. [13-15]). These considerations prompted us to search for the miRNAs, which are controlled by FOXO proteins in human cells.

Results

We have engineered a derivative of human embryonic kidney cell line HEK-293T for the expression of FOXO1-AAA-ER, a regulated form of human FOXO1 (also known as FKHR). The expressed protein contains T24A S256A S319A mutations, which make it insensitive to inhibition by AKT [4]. Furthermore, the protein is produced as a fusion with the ligand-binding domain of mouse estrogen receptor (ER). Therefore, the entire protein remains sequestered in the cytoplasm, until a suitable ligand is added. An additional mutation ensures that the ER fragment retains affinity to some artificial ligands (e.g. 4-hydroxitamoxifen or 4HT), but not to the natural estrogens [19]. Overall, the system provides tight regulation of FOXO1 activity under regular cell culture conditions in the presence of fetal bovine serum, which might contain Akt-activating factors or estrogen.

The miRNA pool from the FOXO1-AAA-ER -expressing cells, with or without 4HT treatment, was profiled using two alternative techniques: hybridization to miRNA-specific microarrays and high-throughput sequencing. Both methods yielded similar results (Fig 1A and 1B): a group of miRNAs, which originate from the same region of X chromosome (Figure 1C), were prominently induced upon activation of FOXO1. The miRNAs included miR-506, miR-507, miR-508, miR-513a-1, miR-513a-2 (highly homologous miR-513a-1 and miR-513a-2 were indistinguishable by array hybridization.), miR-513b and miR-513c. Fold induction could not be estimated for miR-513c in the microarray experiment: the signal was detectable upon induction, but was undetectable in uninduced control. For the same reason the fold induction of miR-507 could not been estimated in the high-throughput sequencing experiment. In these experiments, none of the miRNAs was induced by 4HT treatment in HEK-293T cells transduced with an empty vector in the absence of FOXO1-AAA-ER (data not shown).

We used miR-506 as a representative member of the miRNA cluster. Quantitative RT-PCR analysis confirmed that activation of FOXO1 caused prominent induction of miR-506 in as little as 4 hours (Figure 2). The failure of 4HT to induce the changes in HEK-293 cells, which harbored an empty vector, confirmed that induction was dependent on the function of the expressed FOXO1 protein (Figure 3C).

Although there are reports of miRNA genes transcribed by RNA polymerase III [20], most miRNA appear to be transcribed by RNA polymerase II [21] and are regulated by RNA pol II-associated transcription factors, akin to protein-coding genes [22]. Induction of miR-506 and miR-507 expression was greatly attenuated by low doses of α-amanitin (Figure 3A and B), which selectively targets RNA pol II-dependent transcription. If FOXO1 is indeed a direct regulator of the miRNA transcription, induction should occur even in the absence of de novo protein synthesis. Accordingly, miR-506 was readily inducible in the presence of high doses of cycloheximide, a potent inhibitor of translation (Figure 3C). In fact, the apparent level of miR-506 was even higher in the presence of cycloheximide due to slightly decreased levels of the internal control (RNU6B) in these conditions. Overall, our data suggest that FOXO1 directly activates RNA polII-dependent transcription of the miRNA cluster.

If the miRNAs are regulated by endogenous FOXO activity, one may expect their expression to be sensitive to the status of AKT in the same cells. In cancer cell lines, AKT activity is expected to be high, thus limiting activity of endogenous FOXO proteins. In order to verify whether expression of the miRNA cluster in cancer cells is suppressed by the PI3K/AKT axis, we used LY294002, a specific inhibitor of PI3K, in two human carcinoma cell lines: LNCaP and MCF7. In both cases, LY294002 treatment was accompanied by an increase in miR-506 expression, as measured by quantitative RT-PCR (Figure 4).

Discussion

Our findings demonstrate that expression of a cluster of miRNAs on chromosome X is susceptible to changes in the activity of FOXO1 transcription factor. The phenomenon depends on transcription by RNA polymerase II, but not de novo protein synthesis, suggesting that FOXO1 acts as a classical transcription factor and an immediate regulator of the miRNA expression. At present, the full structure of the corresponding transcripts is unknown. The miRNA may be generated from a single primary transcript or from multiple RNAs, each with its separate promoter. In our experiments we did not observe induction of miR-514b (data not shown), which is located approximately 10 kbps upstream of miR-508. Thus, it is tempting to speculate that at least one of the relevant promoters is located within that region. However, this does not necessarily mean that the relevant FOXO-binding sites have to be located within that area: there are ample examples of genes being affected by the DNA binding events at considerable distances [23, 24]. In fact, a recent study has indicated that FOXO3, which is functionally similar to FOXO1, may transactivate its target genes from distant enhancers via DNA looping [25].

Previously characterized FOXO targets include genes that encode growth suppressors, detoxifying enzymes, as well as regulators of metabolism [26], reflecting the broad range of biological roles of these transcription factors [27, 28]. Indeed, suppression of FOXO activity has been implicated in both tumor-promoting and cytotoxic effects of constitutive activation of PI3K/Akt pathway. The functions of the molecules from the miR-506 cluster are still unknown. One recent report has claimed that overexpressed miRNAs from this cluster, acting in concert with other miRNAs from the adjacent loci, contribute to cell survival and oncogenic transformation in melanoma [29], but no mechanistic explanation was given. So far, we were unable to see such an elevated expression of miR-506 in several melanoma cell lines (RS, ESK, unpublished). Also in melanoma, miRNAs from this cluster were found up-regulated in primary tumor cell lines in comparison to cultured melanocytes, yet downregulated, as invasive potential of melanoma cells increases [30]. There are further reports that downregulation of this cluster marks early relapse in advanced stage ovarian cancer patients [31] and expression of miR-506 inhibits NRAS expression and suppress growth and tumorigenesis in a lung cancer model [32]. Such a pattern of behavior would correspond to the cluster being a suppressor of oncogenic pathways and being negatively controlled by PI3K/Akt pathway, which is a key oncogenic force and is frequently upregulated in later-stage and metastatic cancers (e.g. [33]).

Interestingly, it was reported earlier that mouse homologue of FOXO3 binds DNA in the vicinity of a miRNA cluster consisting of miR-106b, miR-93 and miR-25; and the basal activity of this transcription factor may suppress the levels of these miRNA through an indirect mechanism [34, 35]. However, none of our profiling experiments revealed any appreciable changes in these three miRNA (data not shown). The discrepancy may reflect the difference between FOXO family members or between the cells used in these experiments (e.g. human vs. mouse origin).

It is important to note that bioinformatic prediction of individual miRNA targets is not very accurate[36]. However, a single miRNA is likely to regulate a biological process by targeting more than one involved gene, and co-expressed clusters of miRNAs have been reported to cooperatively control the members of the same biochemical pathway or even of the same protein complex [37]. Hence, the biological function of a miRNA cluster may be implied from the list of pathways, whose components are significantly overrepresented among the predicted miRNA targets. To this end, we used the tools at DIANA LAB web site (http://diana.cslab.ece.ntua.gr/) [38] to identify the pathways from the Kyoto Encyclopedia of Genes and Genomes (KeGG), which are significantly overrepresented among the targets of the miRNA cluster (Supplementary Table S1). Targets were predicted using TargetScan [39], and their list may imply a tumor-suppressive function for the miRNAs. On the list are some anti-apoptotic factors (NFKB1, BCL2L1, etc.). Prominently featured are the activators and the members of the canonical PI3K pathway (IGF1, IGF1R, PDFGRA, IRS2, PIK3CA, PIK3R3, AKT2, etc.). Also abundant among the predicted targets are various activators, intermediaries and downstream effectors of the MAP kinase cascade (e.g. KRAS, RRAS, SOS1, SOS2, MAP3K1, MAP3K3, MAP3K4, MAP2K7, MAPK1, MAPK7, MAPK14, ETS1, ETS2 and many more). PAK1, which acts as a liaison between the PI3K and MAPK pathways [40, 41], is on the list as well. Not surprisingly, the biological processes where these biochemical pathways play prominent roles are predicted to be affected by the miRNA cluster (Supplementary Table S1). These include the geneses of multiple types of cancers; regulation of cell adhesion, migration and cytoskeleton organization; signaling by various cytokines and growth factors; and some of the metabolic and immune functions. In general, this is consistent with the known biological consequences of activation of FOXO factors in mammalian cells [42]. Interestingly, FOXO1 and FOXO3 are also on the list of predicted targets, suggesting a possibility of a feedback loop that may control the abundance of these proteins. Such transcription factor - miRNA circuits appear to be a general theme in the control of mammalian gene expression [12].

Our findings point to a previously unknown facet of the PI3K-Akt-FOXO signaling. Considering the widely acknowledged involvement of this pathway in cancer, diabetes and many other pathological conditions [42], a further study of the identified miRNAs is likely to yield insights into the molecular underpinning and possible biomarkers of these phenomena.

Materials and Methods

Supplementary Materials

Supplemental File