P2.01-022 A PIWI-Interacting RNAs Co-Expression Networks as a Prognostic Factor in Lung Cancer

Abstract

PIWI-interacting RNAs (piRNAs) are small (24-32 nucleotides) non-coding RNAs. Their functions, widely conserved across species, are associated to epigenetic control of gene expression and maintenance of genomic stability by the repression of mobile elements. In humans, >23,000 piRNAs are known, showing tissue-specific expression patterns. While the aberrant expression of individual piRNAs has been identified in some cancer types, the role of piRNA co-expression networks in the development of lung tumors and their utility as molecular markers remains unexplored. By analyzing over 7000 piRNA transcriptomes from human tumors and non-malignant tissues, we have identified lung cancer (LC) specific expression networks associated with clinically-relevant tumor features and patient prognosis. We developed a custom small-RNA sequence analysis pipeline to generate >7,000 human piRNA transcriptomes. piRNA expression baseline was deduced from 6,378 piRNA transcriptomes (non-malignant/tumors) from 11 organ sites. In lungs, we analyzed 1,082 tumors and 209 non-malignant samples from two cohorts: BC Cancer Agency (BCCA) and The Cancer Genome Atlas (TCGA). Network analysis was performed using the weighted gene co-expression network analysis (WCGNA). We evaluated tumor aggressiveness by considering correlation to several clinical parameters, including stage, number of mutations, nodal/distant metastasis, and overall/disease-free survival. piRNA survival signatures were identified using a Cox Proportional Hazard model. A subset of piRNA showed robust expression in somatic tissues. Expressed piRNAs display organ-specific patterns and mainly map to coding transcripts, suggesting a role in regulation of gene expression. In lungs, 204 piRNAs were consistently expressed in both LC cohorts. Tumor piRNA expression profiles are markedly different from their non-malignant counterparts (133 piRNAs were differentially expressed). The patterns differ between the adenocarcinoma and squamous cell carcinoma, and were influenced by smoking status. Network-based analysis identified piRNA expression changes in two modules of piRNAs are associated with aggressiveness tumor features, such as increased number of mutations, tumor size and nodal metastasis. Finally, combined expression of piRNAs define signatures associated with patient overall and recurrence free survival. We provide evidence of somatic, tissue-specific human piRNA expression. In lungs, aberrant expression patterns are associated with well-established etiological factors of cancer and seem to contribute to lung cancer subtype-specific biology. We discover that specific piRNA-based expression patterns characterize aggressive lung tumors and also exhibit prognostic value. The unique expression patterns of piRNAs offer an opportunity to better understand lung cancer-specific biology as well as develop novel prognostic markers for clinical application.