Identification of key keratinocyte stem cell regulators in mouse epidermis and their roles in epithelial cancers using computational methods

Abstract

Keratinocyte stem cells (KSCs) isolated from the hair follicle bulge of mice are known to be targets of skin carcinogenesis. KSCs have previously been shown to have properties defining stem cells such as a slowly cycling nature and regenerative activity in vivo, and a high potential for forming large colonies in vitro. Therefore, identifying key regulators of KSCs can help in discovering new pathways in proliferation, carcinogenesis and other skin inherited diseases. We chose here to investigate long non-coding RNAs (lncRNAs) because they are a poorly understood species of regulatory RNA and because what little is known about them is so intriguing.We focus here on understanding the regulation of KSCs by differentially expressed lncRNAs and their nearby protein coding genes (PCG) because this can give important insights in KSC proliferation, differentiation, and potential targets for cancer. Moreover, comparing the lncRNA-mRNA pairs found in mice with their human homologs across different types of cancer from The Cancer Genome Atlas (TCGA) will help us understand the function of lncRNAs and their regulation thereby providing potential gene targets for KSCs and therapeutic targets for skin cancer and other skin diseases. Towards this end, RNASeq was performed following fluorescence activated cell sorting of KSCs (Cd34+/ Cd49f+) and stem cell depleted epidermal cells (Cd34-/ Cd49f+) of C57BL/6 female mice (n=12 individual determinations). Differential gene expression analysis and gene enrichment analysis were performed to identify top significant lncRNAs and PCGs. To understand cis regulation of lncRNAs, we evaluated DE PCGs that were 100 kb upstream and downstream of DE lncRNAs based on their genomic coordinates. A correlation network was constructed based on the Pearson correlational coefficient value (>0.8) calculated between DE lncRA-mRNA pairs. Gene expression and fold change of the human homologs of lncRNA-mRNA pairs from the mouse data were evaluated across different types of cancer tissues from the TCGA database. In the mouse data, 829 lncRNA and 2700 protein coding genes were significant and differentially expressed (DE). The data demonstrated known DE KSC- specific genes such as CD34+, Krt15, Lgr6, and two metallothionine genes, Mt-2 and Mt-6 involved in proliferation of keratinocytes, were near lncRNAs. Our data showed a cluster of six histone-2 genes near an antisense lncRNA, which is overexpressed in non-stem cells but not in stem cells. Such antisense RNAs are known to interact with histone complexes to form chromatin structures. From the TCGA data, the cancer-associated lncRNAs, MALAT1 and PVT1, demonstrated regulation of protein coding genes like CDKN1B, CDH1 and EZH2, which are all of great interest in KSCs. Finally, we developed KSClncdb, a webtool database compiled of key mouse and human lncRNA-mRNA pairs with gene expression and correlational values for researchers and education. Our results identified new mRNAs and lncRNAs in KSCs and skin cancer. We demonstrate here that our investigation and web database provide a useful new tool and novel insights for further exploration of potential regulatory mechanisms underlying stem cell cancer progression.