Stem cell-like populations from multiple different malignancies can self-renew, differentiate and regenerate malignant tumours 123456789. When introduced into SCID mice, a single so-called Cancer Stem Cell (CSC) is often sufficient to form a tumour representative of the original malignancy 810. The phenotype of the resultant tumour can vary dramatically between malignancies but almost all CSCs generate tumours with populations of undifferentiated and differentiated cells. Tumours containing high concentrations of undifferentiated stem cells are considered to be highly malignant and differentiated tumours less malignant. We postulate that the differentiation capacity of the stem cell population within a malignancy may ultimately determine tumour grade. We aim to elucidate why stem cells have different differentiation potentials and generate tumours with different grades. Addressing this, we have chosen the embryonal carcinoma (EC) model, the only human stem cell model containing both pluripotent and nullipotent cells 1112. Pluripotent NTera2 EC cells differentiate into teratocarcinomas, three germ layer tumours containing a small proportion of undifferentiated stem cells 13. In contrast, nullipotent 2102Ep EC cells can avoid differentiation during tumourigenesis, generating pure embryonal carcinomas, tumours consisting almost entirely of undifferentiated stem cells 14. Thus this model allows comparative analysis of stem cell populations that generate highly and less malignant tumours through differing differentiation potentials. We postulate that the mechanisms facilitating tumourigenesis without differentiation may represent an avenue for targeting.

Ovarian cancer is the 8^th leading cause of cancer in women in the US and the leading cause of death from gynaecological malignancy in the western world 15. Cancer of the ovary represents about 30% of all cancers of the female genital organs. About 205,000 cases of ovarian cancer are diagnosed worldwide each year 16. Strikingly, stem cell-like populations linked to epithelial ovarian cancer (ovarian serous adenocarcinoma [OSC] is the most common histotype 17; germ cell tumours of the ovary are rare) are anti-apoptotic and chemoresistant, suggesting a role in recurrent disease 1819. Significantly, EC is one of the most highly aggressive forms of ovarian malignancy, and intuitively, CSC-targeting is a potential avenue though which anti-cancer therapeutics can be advanced.

MicroRNAs (miRNAs) are short, non-coding RNAs that influence the transcription or translation of target mRNAs 2021. Primary miRNA transcripts (pri-miRs) are processed through stem-looped pre-miRs to achieve mature miRNAs, a process that is facilitated by the Drosha, Dicer and eIF6 proteins. Mature miRNAs hybridize at multiple locations along their target mRNA, including the seed region of the 3' untranslated region (UTR) 21. In most cases, hybridization suspends these targets within the cell, preventing their translation. This post-transcriptional mechanism influences the timing at which mRNAs are presented for translation. At least one miRNA (miR-373) binds to a site in the promoter of its mRNA targets (E-Cadherin and CSDC2), acting as a positive regulator of transcription 22. Specific miRNA populations have been described in stem cells, CSCs and malignancy in general 202324. Additionally, miRNAs located in specific clusters on chromosomes are often simultaneously synthesized. Simultaneous expression of miRNAs located in clusters along chromosomes 14 and 19 has been linked to ovarian malignancy 2526 while the oncogenic role of the miR-17/92 family is well defined 272829. Aberrant expression of cancer-specific 'onco-miRs' is associated with the targeting of oncogenes and tumour suppressors in several different malignancies 30. Many of these miRNAs have been shown to be vital to cancer cells 30.

In this study, we used realtime miRNA qPCR to individually, quantitatively analyze alterations in expression of all known miRNAs during early differentiation of NTera2 and 2102Ep EC cells. Using NTera2 cells as a model of a functionally differentiating EC cell, we aimed to identify miRNA mechanisms associated with the absence of differentiation associated with 2102Ep cells. Our data identifies key differences in expression levels of miRNA biosynthesis genes and of mature miRNAs, particularly those associated with the miR-17/92 family and clustering to chromosomes 14 and 19. Interestingly, undifferentiated 2102Ep cells express mature miRNAs at increased levels despite a decreased level of expression of Drosha and Dicer. While the malignancy-associated miR-17/92 cluster is prominent in both undifferentiated cell types, 2102Ep-specificity was associated with clustering to malignancy-associated chromosomes 19 and 14. Expanding this analysis we have identified NTera2- and 2102Ep-specific differentiation mechanisms that relate to miRNA biosynthesis and expression levels of mature miRNAs. Subsequently, we demonstrate that these mechanisms are similarly relevant to OSC patient samples. OSCs display up to 85% similarity with 2102Ep cells at the miRNA level. Our data reveals that miRNA regulation in 2102Ep EC cells is highly relevant to OSC samples.

Compromising the ability of CSCs to remain in the undifferentiated state is a potential avenue for anti-cancer therapies. Before this can be achieved, we must identify mechanisms involved. 2102Ep cells are the stem cell population of ECs, malignant tumours that can arise in the ovary 11. In vivo, these cells avoid differentiation to produce highly-aggressive, poorly-differentiated tumours 11. In this study we report that miRNA regulation is associated with this phenotype. In undifferentiated cells, this involves decreased expression of miRNA biosynthesis genes, increased expression of mature miRNAs and expression of miRNAs clustering along chromosomes 14 and 19. When treated with RA, 2102Ep cells avoid differentiation and continue to proliferate. This is associated with consistent levels of expression of miRNA biosynthesis genes and mature miRNAs while expression of miRNAs clustering to chromosomes 14 and 19 is deemphasised. OSC samples displayed decreased expression of miRNA biosynthesis genes, decreased expression of mature miRNAs and prominent clustering to chromosome 14 but not 19. This indicates that miRNA biosynthesis and levels of miRNA expression, particularly from chromosome 14, are tightly regulated both in progenitor cells and in tumour samples.

Mechanistically, our data indicates that undifferentiated 2102Ep EC cells can express more miRNAs at higher levels of expression than NTera2 cells despite their decreased expression of miRNA biosynthesis genes. We have previously described the association of decreased eIF6 expression with high-grade OSC samples from patients with reduced disease-free survival 35. In concordance with this, decreased Dicer and Drosha expression is linked to advanced stage ovarian cancer and increased expression to increased patient survival 41. Interestingly, eIF6 expression was unaltered in EC cells. This indicates the complexity of miRNA biosynthesis regulatory mechanisms in EC stem cells and tumours. This mechanism is clearly linked to higher grade malignancy and its elucidation will be the subject of ongoing analysis. The levels of expression of miRNAs were higher in undifferentiated 2102Ep cells than NTera2 cells. 2102Ep cells express 21 miRNAs in both states that are not expressed by NTera2 cells. Similarly, OSC samples showed biased upregulation of miRNAs compared to non-malignant samples. Thus levels of mature miRNA expression are tightly controlled both in progenitor cells and developed tumours. It is widely reported that specifically regulated miRNA groups commonly occur in clusters on specific chromosomes. Prominent clustering to three particular sites was observed in this study: the miR-17/92 cluster and chromosomes 14 and 19, which have been linked with numerous malignancies. miR-17/92 family clusters are associated with regulation of proliferation, angiogenesis and apoptosis in malignancy 272829. These miRNAs were highly expressed by both undifferentiated cell types and were not prominently 2102Ep specific. Previous associations of chromosome 19 with germ cell tumours and of chromosome 14 with ovarian cancer are particularly striking 2642. miRNAs with 2102Ep specificity prominently clustered to these chromosomes while Group 1 miRNAs did not. miRNAs in these regions may contribute to the 2102Ep phenotype and will be assessed by ongoing analysis.

2102Ep cells avoid differentiation through a mechanism that involves maintained expression of pluripotency master genes Oct4 and Nanog 11. We have identified miRNA regulation mechanisms associated with this phenotype. Group 1 miRNAs behave similarly in each EC cell type and are thus likely to act upsteam of the 2102Ep differentiation lesion. Group 2 miRNAs are altered upon differentiation of NTera2 cells but not in 2102Ep cells, suggesting that their role lies downstream of the 2102Ep differentiation lesion. It is possible that Group 1 miRNAs are involved with initiation of tumourigenesis from EC cells. For example, miR-10a targets HoxA1, a long established marker of undifferentiated EC cells 43. Approximately half of these miRNAs were OSC-specific, indicating that both groups are relevant to tumour biology. This may be reflective of the heterogeneous nature of tumour samples, which contain a spectrum of differentiating cell types. Our data indicates that unaltered expression of Group 2 miRNAs is associated with the ability of 2102Ep cells to remain in the undifferentiated state in the presence of a differentiation signal. Maintenance of these miRNAs may protect these EC cells from differentiation signals in vivo. This is supported by their reported validated targets. For example, differentiation regulators are targeted by miRs-199a (Bmp2) and -206 (MyoD) 4445. The future characterisation and manipulation of this lesion may facilitate generation of lower grade tumours from 2102Ep cells.

The substantial overlap between miRNAs expressed by EC cells and in OSC samples exists despite their different phenotypes. EC is of germ cell origin whilst OSC is of epithelial origin. However, morphologically, EC is composed of primitive epithelial cells, which may explain the similarities reported here. It may also be related to tissue-specific expression or reflect a temporal relationship in terms of degree of dedifferentiation Regulation of miRNA biosynthesis and mature miRNA expression in these diverse samples indicates the importance of these mechanisms to ovarian malignancy generally. More than 80% of tumour-specific miRNAs were expressed in 2102Ep cells. This clearly indicates that miRNA regulation in 2102Ep cells is highly relevant to tumour samples, more relevant than miRNA regulation in tumour samples is to 2102Ep cells. Many of these miRNAs have reported associations with malignancy. Stem cells represent a small proportion of a well-differentiated tumour. In contrast, 2102Ep cells generate a malignant tumour in vivo that is almost completely EC cells 1114, while melanoma contains a high proportion of stem cells 46. Thus it is not surprising that highly aggressive 2102Ep cells are more relevant to tumour samples than NTera2 cells.

In this study we have identified two 2102Ep-specific mechanisms. A group of 21 miRNAs are constantly expressed, half of which are OSC-specific. The functional significance of this overlap is suggested by their validated targets. For example, miR-224 targets apoptosis inhibitor 5 (Api5) while miR-503 suppresses cyclinD1 4748. 2102Ep cells respond to RA treatment via a second specific mechanism that is independent of NTera2 mechanisms and has not been previously demonstrated. Group 3 miRNAs are alternatively regulated in each differentiated cell type. This represents a 2102Ep mechanism that, in response to differentiation, acts in the exact opposite fashion to NTera2 cells. Approximately 62% of Group 3 miRNAs were OSC-specific, the largest overlap observed between EC cells and OSC samples. Group 3 miRNAs represent a key target group for future analysis. It is tempting to postulate that this mechanism may facilitate counteraction of differentiation to some extent, a possibility that will be assessed through ongoing analysis. miR-137 is an interesting example as it is expressed in only differentiated 2102Ep cells and in undifferentiated NTera2 cells and is associated with stemness and malignancy 49. miR-137 is downregulated in OSC samples, indicating complex regulation. The identification of a fourth group of miRNAs is potentially highly relevant to our understanding of tumourigenesis from 2102Ep cells. Group 4 miRNAs are altered upon RA treatment of 2102Ep cells. In contrast, Group 4 miRNAs are not altered in NTera2 cells. This indicates that 2102Ep cells can regulate a specific miRNA response to this differentiation signal. Group 4 miRNAs displayed the lowest overlap with OSC samples. This suggests that Group 4 miRNAs are highly relevant to 2102Ep cells. It is possible that Group 4 miRNAs may act against differentiation to contribute to the high grade phenotype, a possibility that is being actively assessed.

The highly malignant phenotype of 2102Ep EC cells employs a three pronged mechanism of miRNA regulation involving miRNA biosynthesis, levels of mature miRNA expression and alternative expression of miRNAs in response to differentiation. This miRNA regulation is associated with the ability of 2102Ep cells to avoid differentiation to generate high-grade tumours and that is relevant to tumour samples. These miRNAs are either similarly or alternatively expressed during tumourigenesis. As the precise mechanisms of miRNA targeting are still being elucidated, it is possible that miRNAs expressed in 2102Ep cells may play similar or diverse roles in OSCs. Due to their association with high-grade progenitor cells and tumours, Group 3 and 4 miRNAs are of particular relevance to future analysis.

Afp: Alpha fetal protein; CSC: Cancer stem cell; EC: Embryonal carcinoma; Eno3: Enolase 3; FFPE: Formalin-fixed paraffin-embedded; mRNA: Messenger RNA; miRNA: MicroRNA; Ncam1: Neural cell adhesion molecule 1; OSC: Ovarian serous adenocarcinoma; qPCR: Quantitative polymerase chain reaction; RA: Retinoic Acid RT: Reverse transcription; UPGM: Unweighted pair group method; UTR: Untranslated region.

The authors declare that they have no competing interests.

MFG, RJF and SAE designed the experiment, analyzed and interpreted data and were the primary authors of the manuscript. SAE and RJF conducted the experiment and data analysis. RJF (pathology), MFG, YMS, SFV (cancer stem cell biology) and MFG, MYL (stem cell biology) contributed to data interpretation. JKM designed figures and contributed to the preparation of the manuscript. PCS contributed to data analysis and experimental design. SOT and AL conducted experimental work. KD, JL and STA aided experimental design and procedure. KL contributed to technical interpretation of the data. CMM (tumour biology) and OS (pathology) critically analyzed the manuscript. JOL directly supervised the study. All authors read and approved the final manuscript.