Definition of microRNAs that repress expression of the tumor suppressor gene FOXO1 in endometrial cancer.

Endometrial cancer is the most common uterine cancer, accounting for 6% of all cancers in women in the United States. Endometrioid adenocarcinoma (type I) or endometrioid endometrial cancer (EEC) is the most common histologic subtype that represents 75-80% of all endometrial cancers ( 1 - 4 ). The remaining endometrial cancers consist predominantly of highly aggressive type II endometrial cancers, including uterine papillary serous (<10%) and clear cell carcinomas (4%) ( 2 ). The aetiology of endometrial cancer remains unclear although unopposed oestrogen exposure, complex hyperplasia with atypia, and treatment with tamoxifen during breast cancer therapy are recognized risk factors for endometrial cancer. Hysterectomy and bilateral salpingo-oophorectomy is still the primary and most effective treatment for patients with localized disease. Although adjuvant radiation therapy may reduce loco-regional recurrence in patients with stage I disease, the associated toxicity and morbidity is significant. Patients with inoperable disease that are not candidates for radiation therapy may be treated with progestational agents ( 5 ). A combination of cisplatin or paclitaxel with doxorubicin has been reported to improve overall survival in a subset of patients with stage III or IV disease compared to radiation therapy ( 5 ). However, prognosis remains poor for the 15% of patients that develop persistent or recurrent tumours, or whom present with advanced disease. Thus, there is an urgent need for new therapeutic targets and strategies, both of which may be realized through an increased understanding of the molecular mechanisms governing endometrial tumourigenesis. Deregulation of the PTEN/phosphoinositide 3-kinase (PI3K)/PKB (also called Akt) signalling pathway is a hallmark of endometrial cancer ( 6 ). PTEN is inactivated in more than 80% of endometrioid carcinomas, leading to unrestrained PKB signalling. Loss of PTEN is likely to be an early event in endometrial tumorigenesis as it is found in 55% of precancerous endometrial lesions ( 3 , 7 , 8 ). In addition, 40-80% of type II endometrial cancers display HER2 amplification, and 36% of endometrioid carcinomas have mutations in the PI3KCA gene, both of which enhance the activity of the PI3K/PKB signalling pathway ( 6 , 8 ). Key effectors of PI3K deregulation are the FOXO family of transcription factors, consisting of FOXO1, FOXO3a, and FOXO4, which are direct downstream phosphorylation targets of the protein kinase PKB and the related kinase SGK1 ( 9 - 11 ). PKB-dependent phosphorylation of FOXO proteins results in their nuclear exclusion, leading to loss of trans -activation of FOXO target genes, many of which are important for cell cycle regulation, apoptosis, and differentiation. Chemosensitization of endometrial cancer cells in response to PKB inhibition has been shown to require FOXO1 activity ( 12 ). FOXO1 has also been shown to be a crucial regulator of progesterone-dependent differentiation of the normal human endometrium, a process termed decidualization, and the subsequent menstrual shedding ( 13 , Takano, 2007 #503). The central role of FOXO1 in endometrial differentiation was highlighted by a recent mircoarray study demonstrating that FOXO1 knockdown in primary endometrial cells using small interfering RNA (siRNA) perturbs the expression of 15% of all transcripts regulated during decidualization ( 14 ). Importantly, FOXO1 was shown to simultaneously regulate the induction of CDKN1C , which encodes for the cyclin-dependent inhibitor p57 Kip2 involved in G1 cell cycle arrest, and the repression of several genes involved in DNA replication (eg. MCM5 ), G2/M transition (eg. CCNB1, CCNB2, CDC2, BIRC5 , and BRIP1 ), and mitosis (eg. PRC1, NUSAP1, CENPF, SPBC25 and ASPM ). MicroRNAs (miRNAs or miRs) are a class of small non-coding RNAs that regulate gene expression by facilitating mRNA degradation or translational inhibition. They play important roles in development, cellular differentiation, proliferation, cell-cycle control, and cell death ( 15 , 16 ) and have been implicated in a variety of human diseases, including cancer ( 16 , 17 ). Recent evidence suggests that the FOXO family of transcription factors are targets for regulation by miRs. For example, aberrant miR-182 expression has been implicated in melanoma metastasis through down-regulation of FOXO3a ( 18 ). In addition, a number of studies reported altered patterns of miR expression in endometrioid endometrial cancer and identified a subset of miRs that correlated with disease stage and recurrence ( 19 , 20 ). Using an in silico approach, we identified several cross-species conserved miRs predicted to target the 3′UTR of FOXO1. In both clinical samples and endometrial cancer cell lines, we investigated the role of miRs in the deregulation of FOXO1 expression and activity in endometrial cancer.

In silico prediction of FOXO1 targeting miRNAs was performed using the PicTar, TargetScan, miRanda, and miRNA targets (Memorial Sloan-Kettering Computational Biology Centre), miRBase, and RNAReg algorithms ( 21 - 24 ). Venn diagram analysis was then performed ( Figure 3A ) and used to identify miRNAs which displayed both inter-species conservation and were identified by multiple algorithms. Human endometrial cancer cell lines HEC-1B and Ishiwaka originated from the American Type Culture Collection and were acquired from Cancer Research UK (London, UK), where they were tested and authenticated. These procedures include cross species checks, DNA authentication and quarantine. Cell lines used in the present study were in culture for less than 6 months and maintained in DMEM/F12 media (Sigma) containing 10% foetal calf serum (FCS) and 2 mM glutamine (10% CO2; 37°C). The participating Local Research and Ethics Committees approved the study and patient consent was obtained before tissue collection. Snap-frozen endometrial cancer samples were obtained from patients undergoing hysterectomy without preoperative chemotherapy or radiation and histologically validated for type and grade. Normal endometrial samples were obtained from premenopausal women awaiting in vitro fertilization treatment. All of the samples were frozen in liquid nitrogen immediately after resection and stored at −80°C until use. Paraffin-embedded, formalin-fixed endometrial specimens were immunostained for FOXO1 using the Universal LSAB Plus Kits (DAKO, Ely, UK) as described previously ( 25 ) using primary antibodies against FOXO1 (Cell Signalling Technology Inc.; 1:50 dilution). Histological and immunohistochemical assessments were performed by two independent pathologists. Every tumour was given a score reflecting the average intensity of the staining (no staining=0; low staining=1; medium staining=2; strong staining=3). Protein was extracted and SDS-PAGE performed as previously described ( 26 ). FOXO1 (C-20), GAPDH (6C5), p27 Kip1 (SX18F7) and β-tubulin (H-40) antibodies were from Santa Cruz Biotechnology (Autogen Bioclear, Wiltshire, UK). Western blot quantification was performed using ImageJ software (Image Processing and Analysis in Java). All western blot experiments were performed in triplicate. Total RNA was extracted from cell lines using TRI reagent (Ambion, Austin, TX, USA) as previously described ( 27 ) and used for qRT-PCR analysis without enrichment for miRs ( 28 ). cDNA synthesis and qRT-PCR was performed using miRVana qRT-PCR miRNA detection kit and primer sets (Ambion) according to manufacturers instructions. Samples were normalized to 5S (Ambion) and quantified accordingly using standard curves. For FOXO1 mRNA quantification, the following gene-specific primer pairs were used: L19-sense (5′-GCGGAAGGGTACAGCCAAT-3′) and L19-antisense (5′-GCAGCCGGCGCAAA-3′); FOXO1-sense (5′-TGGACATGCTCAGCAGACATC-3′) and FOXO1-antisense (5′-TTGGGTCAGGCGGTTCA-3′); L19 , a non-regulated ribosomal housekeeping gene was used as an internal control to normalise input cDNA. qRT-PCR was performed using the Applied Biosystems HT-7900 using SYBR Green Mastermix (Applied Biosystems, Brackley, UK). Each analysis was performed in three experimental replicates with three technical replicates within each experiment and standard error of the mean determined (SEM). Northern blot analysis was performed essentially on pooled RNA extracted from 7 normal and 10 malignant endometrium samples using Trizol reagent (Invitrogen). 5′-DIG-labelled LNA modified probes against hsa-miR-182, hsa-miR-96, and hsa-miR-194 (which was predicted by qRT-PCR not to change) were used for hybridization (Exiqon). Probe sequences were hsa-miR-96 5′-AGCAAAAATGTGCTAGTGCCAAA-3′; hsa-miR-182 5′-AGTGTGAGTTCTACCATTGCCAAA-3′; hsa-miR-194 5′-TCCACATGGAGTTGCTGTTACA-3′. Northern blot was performed essentially as described ( 29 ). Perfect HybPlus was used for hybridization (68°C; Sigma-Aldrich). Membranes were then developed using the Block and Wash buffer Set (Roche) and CDP-Star reagent (Roche). RNA loading was determined by staining of the tRNAs with ethidium bromide. Cells were seeded on 6-well plates 24 h prior to transfection (Invitrogen, Paisley, UK). miR mimeitcs and inhibitors were transfected into cells at a final concentrations of 60nM using Oligofectamine (Invitrogen) in serum-free conditions. miRNA transfection conditions were optimized to allow for comparative mimetic levels in HEC-1B compared to miRNA levels in Ishikawa cells ( Supplemental Figure S2B ). Cells were incubated with miRNA inhibitors, synthetic pre/anti-miRNA mimetics, or appropriate scramble controls (all from Ambion) for 4 h in OPTI-MEM media before addition of normal growth media. The cells were then assayed 48 h after transfection. Propidium iodide staining was performed as previously described ( 30 , 31 ). All values are presented as mean±SEM where appropriate. Statistical significance between two groups was determined by use of a two-tailed t -test, and values of p<0.05 were considered significant and indicated by (*) and very significant, values of p<0.001 were indicated by (**). To test for differences between tissue types and FOXO1 staining one-way-analysis of variance (ANOVA) was performed followed by Dunnett’s 2-tailed t-test, and the mean difference considered significant at the p<0.05 level ( Fig. S1 ). All statistical analysis was performed SPSS v.16.

The levels of FOXO1 protein expression were first investigated in 9 normal endometrial tissues, 73 hyperplasia cases, and 52 endometrioid endometrial carcinomas by immunohistochemical analysis ( supplemental Fig. S1 ). FOXO1 staining was positive in all normal endometrial tissue samples (100%; 9/9). In normal endometrium, FOXO1 staining was strong in all (9/9), and was detected predominantly in both the cytoplasm and nucleus of epithelial cells, with significant levels of FOXO1 expression also observed in the stromal cells (Fig. 1A, 1B and S1 ). In contrast, FOXO1 staining was undetectable or at low levels in all the endometroid endometrial cancer samples (100%; 52/52) (Fig. 1A, 1B and S1 ). In hyperplastic endometrium, the majority of FOXO1 staining was either weak or undetectable (95%; 69/73). Notably, in both the hyperplasia and cancer samples FOXO1 expression, in contrast to normal endometrium, was restricted to the epithelial cells ( Fig. 1A ). The intensity of the FOXO1 staining in endometrial cancer tissue was on average significantly lower than the intensity observed for hyperplasia samples and for normal endometrium ( Fig. 1C ). To confirm these findings, protein lysates extracted from 7 normal and 10 malignant endometrial samples were analysed by Western blotting ( Fig. 2A ). Consistent with the immunohistochemistry data, the results demonstrated that FOXO1 was expressed at relatively high levels in the normal endometrial tissues, whereas in endometrial cancer samples the levels were much lower or beyond detection ( Fig. 2A and 2B ). The same samples were also analyzed for FOXO1 mRNA expression and, in agreement with other studies ( 4 , 32 - 34 ), the abundance of FOXO1 transcripts was lower in malignant compared to normal cycling endometrium ( Fig. 2C ). However, the down-regulation at mRNA level was much less pronounced than at protein level, suggesting that post-transcriptional mechanisms are involved in inhibiting FOXO1 expression in endometrial cancer. By combining the results of several miRNA target prediction programmes, we first identified a panel of highly conserved miRs with the potential to target the 3′-UTR of FOXO1 transcripts ( Figure 3A and 3B ) and then examined their expression levels in the same 10 malignant and 7 normal endometrial samples by qRT-PCR analysis. As shown in Figure 3C , the levels of several of these in silico predicted miRs, including miR-9, -27, -96, -128, -153, 182, -183, and -186, were significantly up-regulated in endometrial cancer compared with normal endometrium. Notably, the abundance of some miR species, such as, miR-27, -128 and -186 was more than 10-fold higher in the cancer samples. In order to further confirm the difference in miR expression between normal endometrium and endometrial cancer we used Northern blot analysis to examine miR expression level ( Figure 3C ). In agreement with the qRT-PCR data, both miR-96 and miR-182 showed a higher level of expression in the endometrial cancer samples. As an internal control we examined the expression of miR-194, which showed no significant difference in expression by qRT-PCR; miR-192 was detected at a similar level in both normal and malignant tissue by Northern blot analysis ( Figure 3C ). These data suggest that the loss of FOXO1 expression upon malignant transformation of human endometrium coincides with a strong induction of several miRs, suggesting a causal link between these phenomena. To explore the mechanisms that perturbed FOXO1 expression in endometrial cancer, we made use of two well-characterised endometrial cancer cell lines, HEC-1B and Ishikawa. Western blot analysis confirmed that HEC-1B, but not Ishikawa cells abundantly express FOXO1 ( Fig. 4A ) as previously reported ( 33 , 34 ). In agreement, Ishikawa cells also express much lower levels of p27 Kip1 , a FOXO1 target in endometrial cells. As was the case for normal and malignant endometrial biopsies, qRT-PCR analysis demonstrated lower FOXO1 transcript levels in Ishikawa cells ( Fig. 4B ), albeit much less so than would have inferred by the Western blot analysis. The discrepancy between FOXO1 mRNA and protein levels prompted an analysis of putative FOXO1 miRs in HEC-1B and Ishikawa cells. Strikingly, qRT-PCR analysis demonstrated that several microRNAs, including miR-27, -96, -128, -153, 182, -183, and -186, were expressed at much higher levels in Ishikawa compared with HEC-1B cells ( Fig. 4C ). Others, including miR-135, miR-142-3p and miR-194, were expressed at comparable levels in both cell lines whereas miR-486 and miR-9 were more abundant in HEC-1B cells ( Fig. 4C ). Thus, loss of FOXO1 expression in endometrial cancer in vivo and in Ishikawa cells correlated with increased expression of a number of the same miRs that included miR-9, -27, -96, -128, -153, 182, -183, and -186. To test the repressive potential of these miRs, synthetic mimetics were transfected individually into HEC-1B cells and endogenous FOXO1 expression levels monitored by Western blot analysis. In addition, cells were also transfected with a scramble mimetic and a mimetic of miR-29a, which was not predicted to target FOXO1. Overexpression of several miRs (miR-9, -27a, -96, -153, -183 and -186) effectively down-regulated FOXO1 expression ( Fig. 5A ). In contrast, miR-128 (despite its strong induction in endometrial cancer), -152, -486, and the miR-29a control had little effect on FOXO1 expression. In the reverse experiment, we used synthetic hairpin miRNA inhibitors to silence the activity of specific miRs in Ishikawa cells, and monitored the expression of FOXO1 and its target p27 Kip1 . Transfection of some individual miR inhibitors, targeting miR-9, -27a, -96, -153, -183 and -186, elicited a small but reproducible induction in FOXO1 and p27 Kip1 levels ( Fig. 5B ). In contrast, FOXO1 levels were effectively unchanged upon transfection of anti-miR scramble control or a miRNA not predicted to target FOXO1, miR-29a. These observations indicate that targeting of individual miRs is insufficient to fully restore FOXO1 expression in endometrial cancer cells. However, transfection of individual inhibitors did not restore FOXO1 to the levels observed in HEC-1B cells. In order to further examine this effect 2X (anti-miR-153, -183), 6X (anti-miR-9, -27, -96, -153, -183 and -186), or 8X (anti-miR-9, -27, -96, -128, -153, -182, -183 and -186) were co-transfected into Ishikawa cells. Co-transfection of miR-inhibitors dramatically enhanced FOXO1 levels in Ishikawa cells, and was accompanied by the induction of the FOXO1 target p27 Kip1 ( Fig. 6A ). During the development of endometrial cancer the repression of FOXO1 by miR may be hypothesized to confer a proliferative advantage, or to allow cells to escape apoptosis. As such, we next determined the effect of repressing the miR expression in Ishikawa cells. Cell cycle analysis of propidium iodine stained cells by flow cytometry demonstrated that transfection of the 6X anti-miRs was effective in inducing cell cycle arrest and apoptosis in Ishikawa cells ( Fig. 6B ). In order to demonstrate the requirement of FOXO1 induction for anti-miR induced cell death, the 6X anti-miR panel was co-transfected with a siRNA targeting the coding domain of FOXO1, causing a repression in FOXO1 levels ( Figure 6C ). Under these circumstances a reduction in anti-miR induced cell death was observed, consistent with FOXO1 induction being critical for this effect ( Figure 6C ). Whilst HEC-1B cells already express high levels of FOXO1 we next determined whether the transfection of the 6X anti-miR pool would result in cell death ( Figure 6D ). In fact, only a moderate effect was observed on FOXO1 expression level, consistent with low endogenous levels of the miRs observed in the qRT-PCR analysis, and moreover, no significant change in the cell cycle distribution was observed ( Figure 6E ). Together, these observations suggest that a number of miRs may repress endometrial FOXO1 expression, thereby promoting proliferation and survival of endometrial cancer cells.

In this study, we identified and characterised a defined set of FOXO1-targeting miRs that may have a role in endometrial tumorigenesis. The expression of the tumour suppressor FOXO1 is strongly repressed in endometrial cancer samples compared to normal cycling endometrium. Using a combination of miR target prediction programmes, we identified a panel of miRs that may theoretically target the 3′-UTR of FOXO1 transcripts. Comparative analysis of these miRs and FOXO1 in normal endometrium and endometrioid endometrial cancer revealed that loss of FOXO1 expression inversely correlated with significant up-regulation of several of the in silico -predicted targeting miRs, including miR-9, -27, -96, -128, -153, 182, -183, and -186. However, the comparative expression of miRs in endometrial cancer versus normal, and HEC-1B versus Ishikawa cells were not always in agreement. In particular, miR-9 was increased in endometrial cancer tissue, but lower in the HEC-1B compared to Ishikawa cells, whilst miR-486 shows no change in tissue samples, but is higher in HEC-1B cells compared to Ishikawa. This is likely a reflection of the limitations of comparing cancer tissues and cancer cell lines respectively. In addition, whilst HEC-1B and Ishikawa cells display different FOXO1 expression levels, both are still transformed cell lines and thus not an exact model of normal and malignant endometrium. However, HEC-1B did display a miR profile more akin to normal endometrium than cancer, and this cell line was used to examine the ability of miRs to repress endogenous FOXO1 expression. The relatively higher levels of FOXO1 expression in HEC-1B cells is consistent with the fact that HEC-1B proliferates slower compared with other endometrioid endometrial cancer cell lines (eg. Ishikawa and ECC1) ( 33 ). Interestingly, not all predicted miRs inhibited endogenous FOXO1 protein expression in HEC-1B cells. In particular, miR-128, which was expressed at a much higher level in Ishikawa cells, had no significant effect on the expression level of FOXO1, suggesting that reintroduction of miR-128 alone is not sufficient to inhibit FOXO1 expression, or that FOXO1 is not a direct target of miR-128 in this instance. However, these data further narrow the set of putative FOXO1-targeting miR to miR-9, -27, -96, -153, 182, -183, and -186, all of which were also overexpressed in endometrial cancer. FOXO1 expression is substantially lower in Ishikawa compared to HEC-1B cells ( 34 ) and, as is the case in vivo , coincided with increased expression of several putative FOXO1 targeting miRs. Yet, knock-down of individual miRs has only marginal effects on FOXO1 expression in Ishikawa cells in contrast to the overexpression experiments in HEC-1B cells. This is, however, predictable as elevated levels of other FOXO1-targeting miRs will compensate for the loss of one miR species. Similarly in vivo , some but not all FOXO1-targeting miRs may be critical to endometrial tumorigenesis depending on their relative expression levels in the endometrial cancer cells. Co-overexpression of these miRNAs with overlapping function in endometrial cancer may have significant implications in vivo , and may result in the loss of cell cycle and cell death control. The fact that some but not all FOXO1-targeting miRNAs have a predominant role in repressing FOXO1 expression could be related to their relative expression levels in the endometrial cancer cells. Consistent with this postulation, the 6X miRs (miR-9, -27, -96, -128, -153, -183, and -186) that cooperated in FOXO1 repression in Ishikawa cells were also amongst the most up-regulated species in endometrioid endometrial cancer compared to normal endometrium. Interestingly, the 3′UTR seed sites of a number of the miRs implicated in FOXO1 regulation, in particular miR-27 and miR-128, and miR-182 and miR-96, were proximally located, possibly resulting in co-operative repression. Several of the miRs that repressed FOXO1 most effectively were located in the 5′ region of the UTR, which has previously been reported to be indicative of higher site activity. It is also intriguing to note that miR-183, miR-96, and miR-182 are located proximally in genome (chr7:129201768-129201845, chr7:129201981-129202090, and chr7:129197459-129197568 respectively). Moreover, a previous study identifying the proximal promoters of a large number of miRNAs ( 35 ), showed that miR-96, -183, and -182 share the same transcription start site (chr7: 129207158), suggesting these miRNA may be co-ordinately deregulated, and therefore be of particular importance during endometrial tumorigenesis. A recent report also supports our findings by describing the targeting of FOXO1 by miR-27a, -96, and -182 in breast cancer cells ( 36 ). However, article does not determine whether the deregulation of these miRNA observed in breast cancer cell liens was also observed in breast cancer tissue. As such, our findings represent the first demonstration of FOXO1 targeting miRNA in vivo . In addition to the miR-mediated repression, other mechanisms may contribute to loss of FOXO1 expression in endometrial cancer. Previously we reported that the FOXO1 promoter is methylated in some but not all endometrial cancers ( 33 ). Further, a recent report proposed that the low levels of FOXO1 protein observed in Ishikawa cells are due to the upregulation of Skp2, an oncogenic subunit of the Skp1/Cul1/F-box protein ubiquitin complex that promotes ubiquitination and degradation of phosphorylated FOXO1 ( 37 ). Consistent with this, Skp2 is overexpressed in endometrioid endometrial cancer ( 34 , 38 ). However, treatment with the proteasomal inhibitor is insufficient to restore FOXO1 expression in Ishikawa cells ( 33 ), indicating that enhanced proteosomal degradation is not the primary pathway that account for loss of this transcription factor in endometrial cancer. FOXO1 has emerged as a major regulator of progesterone-dependent differentiation of the human endometrium and subsequent apoptosis associated with menstrual shedding ( 14 ). In addition, FOXO1 has been implicated in safeguarding genomic stability of the endometrium during the rapid waves of intense tissue remodelling by regulating the expression of DNA repair genes, such as GADD45A . Consequently, several studies have functionally linked the decrease or complete loss of FOXO1 expression in endometrioid endometrial cancer to uncontrolled cell proliferation, impaired apoptosis and increased susceptibility to genotoxic stress ( 33 , 34 , 39 ). Our results showed that re-expression of FOXO1 in Ishikawa cells upon knock-down of 6 miRs was accompanied by cell cycle arrest and cell death, emphasizing the cardinal role of FOXO1 in cell cycle control and cellular death response in endometrial cells. The extent to which deregulation of the miR machinery can be implicated in benign proliferative disorders of the endometrium, such as endometriosis, is currently under investigation. The corollary of these observations suggest that miR-based gene therapy may provide a novel approach for the treatment of endometrial hyperplasia and cancer ( 40 - 42 ). In summary, we identify a group of FOXO1-targeting miRNAs that are up-regulated in endometrial cancer. The ability of this group of miRs to promote FOXO1 repression may precipitate in a key role in endometrial tumorigenesis through bypassing cell cycle and cell death control.