1 MicroRNA-203 represses selection and expansion of oncogenic 2 Hras transformed tumor initiating cells 3 4 Kent Riemondy1, Xiao-jing Wang2, Enrique C. Torchia3, Dennis R. Roop3 and Rui Yi1,4 5 6 1Department of Molecular, Cellular, and Developmental Biology, University of Colorado, 7 Boulder, Boulder, CO 80309 8 2Department of Pathology, University of Colorado Denver Anschutz Medical Campus, Aurora, 9 CO 80045 10 3Department of Dermatology and Charles C. Gates Center for Regenerative Medicine and Stem 11 Cell Biology, University of Colorado Denver Anschutz Medical Campus, Aurora, CO 80045 12 13 4Correspondence: 14 Rui Yi 15 Department of Molecular, Cellular, and Developmental Biology 16 University of Colorado, Boulder 17 Boulder, CO USA 80303 18 Tel: 303-735-4886 19 Email: [email protected] 20 21 Competing Interests Statement 22 The authors declare no competing interests. 1 23 Abstract 24 In many mouse models of skin cancer, only a few tumors typically form even though many cells 25 competent for tumorigenesis receive the same oncogenic stimuli. These observations suggest an 26 active selection process for tumor-initiating cells. Here we use quantitative mRNA- and miR-Seq 27 to determine the impact of HrasG12V on the transcriptome of keratinocytes. We discover that 28 microRNA-203 is downregulated by HrasG12V. Using a knockout mouse model, we demonstrate 29 that loss of microRNA-203 promotes selection and expansion of tumor-initiating cells. 30 Conversely, restoration of microRNA-203 using an inducible model potently inhibits 31 proliferation of these cells. We comprehensively identify microRNA-203 targets required for 32 Hras-initiated tumorigenesis. These targets include critical regulators of the Ras pathway and 33 essential genes required for cell division. This study establishes a role for the loss of microRNA- 34 203 in promoting selection and expansion of Hras mutated cells and identifies a mechanism 35 through which microRNA-203 antagonizes Hras-mediated tumorigenesis. 36 37 38 39 40 41 42 43 44 45 46 2 47 Introduction 48 Recent efforts in comprehensively sequencing human cancer genomes have confirmed 49 ~140 protein-coding genes that, when mutated, can drive tumorigenesis [1]. When genome 50 sequencing data were utilized to construct the history of cancer cells in breast cancer, it was 51 revealed that a considerable amount of “molecular time” exists between the common ancestors 52 that harbor the great majority of driver mutations and the phenotypically identified cancer cells 53 that compose the bulk of the tumor [2]. In support of these observations, lineage tracing 54 experiments conducted in genetically engineered mouse models revealed only a few clones give 55 rise to tumors whereas a vast majority of mutated cells are unable to sustain tumorigenesis [3,4]. 56 These results suggest that even after the acquisition of key driver mutations in the nascent cancer 57 cells, these cells must still undergo continuous evolution and likely clonal selection before 58 developing into clinically apparent tumors. To begin to understand the molecular basis 59 underlying such selection, we examined papilloma formation driven by oncogenic Hras in the 60 skin, a well characterized model where Hras has been shown to initiate the formation of tumors 61 that clonally evolve [3,5,6]. Oncogenic Ras mutations are some of the most frequently detected 62 driver mutations in human cancer. Among the three Ras genes (H-, K- and N-ras), Hras is 63 commonly mutated in tumors originated from stratified epithelial tissues including squamous cell 64 carcinoma in the skin, head and neck cancer as well as bladder cancer [7–9]. Experimental and 65 genomic sequencing studies have revealed that the vast majority of Ras mutations are missense, 66 point mutations at amino acid residues glycine 12 (G12), glycine 13 (G13) or glutamine 61 67 (Q61) [8]. Structural and biochemical studies have further confirmed that all of these mutations 68 generally interfere with the GTP binding pocket and compromise the GTPase activity of Ras 69 proteins. In turn, these mutations lead to uncontrolled activation of downstream effectors 3 70 including Raf/MEK/ERK and PI(3)K pathways, resulting in sustained cell survival and 71 proliferation observed in human cancers. Because of the prominent role of Ras mutations in 72 human cancer, extensive efforts have been devoted to uncover and subsequently target 73 downstream pathways that are regulated by Ras mutations. However, the immediate impact of 74 Ras mutations on the transcriptome, in particular with regards to microRNAs (miRNAs) remains 75 unclear. 76 miRNAs are a class of small, noncoding RNA species that are involved in virtually all 77 biological processes examined in mammals including mouse and human. These regulatory RNA 78 molecules function by repressing the protein producing ability of mRNA targets through 79 destabilization of mRNAs and inhibition of translation [10]. miRNAs typically target a large 80 number of mRNAs in a dosage- and cell context-dependent manner [11]. As prominent proto- 81 oncogenes, Ras mutations have long been recognized to interact with the miRNA pathway. 82 Indeed, Hras, Kras and Nras all harbor multiple binding sites for the let-7 miRNA, a founding 83 member of miRNAs, in their 3’UTRs [12]. Additionally, impaired miRNA biogenesis in the 84 form of Dicer1 disruption has been shown to be a tumor suppressing mechanism for the 85 development of Kras induced lung cancer in a mouse model [13]. A number of individual 86 miRNAs were also found to function as modifiers for Ras induced tumorigenesis that include 87 miR-21, -29 and miR-17~92 as tumor promoting miRNAs and miR-34, -15/16, and miR-143/145 88 as tumor suppressing miRNAs [14–16]. Collectively, these seminal studies demonstrate 89 unequivocally that the miRNA pathway and individual miRNAs play important roles in Ras- 90 induced tumorigenesis. However, it is unclear how Ras mutations, usually the tumor-initiating 91 drivers, directly alter the landscape of miRNA expression during tumorigenesis. Importantly, it is 92 also unknown whether the changes in miRNA expression play a role in the selection of 4 93 oncogenic Ras transformed cells during tumor initiation. Finally, the lack of a comprehensive 94 survey of high-confidence miRNA targets that may play a role downstream of Ras mutations 95 hinders our mechanistic understanding and limits the potential to develop miRNA based 96 therapeutics. 97 In this study, we utilized our recently improved quantitative miR-Seq techniques to 98 examine the impact of an oncogenic Hras mutation (HrasG12V) on both mRNA and miRNA 99 expression. We discovered that miR-203, the most highly expressed miRNA in the skin [17,18], 100 is downregulated by HrasG12V. Using both knockout (KO) and inducible models, we provide 101 evidence for an important role of miR-203 in restricting expansion of oncogenic Hras 102 transformed cells in vitro and in vivo. We comprehensively surveyed skin-specific targets of 103 miR-203 and identified a number of novel targets that have important implications for Hras- 104 mediated tumorigenesis. Our results suggest that miR-203 plays a tumor-suppressing role in 105 inhibiting selection and expansion of tumor-initiating cells early in tumor development. 106 107 Results: 108 HrasG12V profoundly deregulates the mRNA and miRNA transcriptome in the skin 109 Oncogenic mutation of the Hras gene is one of the initiating drivers in the development 110 of benign papillomas and malignant squamous cell carcinomas in murine skin chemical 111 carcinogenesis. However, the molecular consequences defining the cellular changes that 112 accompany expansion of oncogenic Hras-transformed keratinocytes to initiate papillomas 113 remain elusive. We first investigated the consequences of HrasG12V activation on the mRNA 114 transcriptome using a modified form of PolyA+ RNA-Seq, known as 3P-Seq, or 3seq (Figure 115 1A-C). Compared to traditional RNA-Seq, 3Seq allows both quantification of mRNA transcripts 5 116 and detection of changes in alternative 3’ UTR formation [19]. To examine the immediate 117 impact of HrasG12V on primary skin cells, we used primary keratinocytes isolated from newborn 118 skin and performed 3Seq after HrasG12V transduction. We did not observe widespread shortening 119 or alternative formation of 3’UTRs, which are often ascribed to oncogenic transformation when 120 comparing tumor cell lines to normal cells (data not shown). This is similar to our previous 121 observation that alternative 3’UTR usage is infrequent within the skin lineages [19]. Over 1,100 122 transcripts were differently expressed (two-fold change and FDR <0.05) in keratinocytes 123 expressing HrasG12V, compared to the control (Figure 1C and Figure 1-source data). Gene 124 ontology functional analysis revealed profound deregulation in three core processes by HrasG12V: 125 activation of cellular migration, upregulation of pro-angiogenic pathways, and suppression of the 126 terminal differentiation program (Figure 1D). All of these three processes are identified as 127 hallmarks of human cancer [20]. The observed widespread changes in the transcriptome also 128 endorse the driver role of HrasG12V in skin tumorigenesis. Importantly, transcripts upregulated by 129 HrasG12V in our primary keratinocytes strongly and significantly overlapped with the putative 130 cancer stem cell signatures obtained from murine squamous cell carcinoma (SCC) models [21]. 131 In addition, transcripts upregulated by HrasG12V significantly overlapped with transcripts know to 132 be targets of the c-Fos transcription factor in a genetic model of SCC [22]. Furthermore, known 133 core components of the Hras signaling pathway were also among the differentially detected 134 genes [23] (Figure 1E). These transcriptome data indicate that we have captured the initiating 135 changes induced by oncogenic Hras in the keratinocytes. 136 To define the impact of the oncogenic Hras on the landscape of miRNA, we applied our 137 recently developed, quantitative miRNA-Seq [24] to HrasG12V transformed keratinocytes. Overall, 138 we detected 15 differentially expressed miRNAs upon HrasG12V expression (FDR <0.05, two- 6 139 fold change) (Figure 1F-H).