Lung cancer remains one of the leading causes of cancer mortality worldwide. Despite significant progress in our understanding of tumour biology, with advances in diagnosis and management, the prognosis for patients diagnosed with lung cancer is poor, with 5-year survival below 15%. With the ability to negatively regulate gene expression, microRNAs (miRNAs) have emerged as a unique class of oncogenes and tumour suppressor genes, with demonstrated roles in multiple human malignancies. The underlying hypothesis examined by this thesis was that aberrant expression of specific miRNAs are important for lung carcinogenesis and therefore potentially represent a new class of diagnostic, prognostic and therapeutic targets.
The primary aim of this thesis was to identify novel oncogenic and tumour suppressing miRNAs in primary non-small cell lung cancer (NSCLC) using two genome-wide microarray based approaches to characterise miRNA expression and miRNA genomic copy number (CN) changes. The secondary aim of this thesis was to characterise miRNA expression in clinical subgroups of NSCLC, specifically in relation to asbestos-related lung carcinogenesis; somatic mutations in NSCLC (EGFR, KRAS and FGFR1); and finally identify miRNA signatures for predicting patient prognosis in adenocarcinomas (ACs) and squamous cell carcinomas (SCCs).
Existing genome-wide aCGH data (Human Genome CGH Microarray 44B, Agilent) in 132 primary NSCLC (72 ACs, 60 SCCs) was used to map 474 human miRNAs (miRBase v9.0; http://microrna.sanger.ac.uk), to regions of DNA CN gain or loss. 89 miRNAs mapped to regions of aberrant CN based on their immediately flanking probes or host gene being deleted or amplified in ≥25% of primary tumours using both Analysis of Copy Errors algorithm and fold change (> +1.2) analyses. Three independent NSCLC aCGH datasets confirmed that 22 of these miRNAs showed directionally concordant CN changes. miR-218, encoded within two host genes (SLIT2, 4p15.31 and SLIT3, 5q35.1), in a region of CN loss, was selected for validation as it is independently reported as underexpressed in lung cancer. Decreased expression of mature miR-218 and SLIT2/3 was confirmed ain 39 NSCLCs relative to normal lung tissue and was found to be associated with a history of cigarette smoking.
miRNA expression profiling using the 8x15k Human miRNA microarray (Agilent G4470A) was performed in 170 primary NSCLCs and 88 normal lung tissues samples. This cohort was split into a training cohort of 63 primary NSCLCs (37 AC, 26 SCC) with 63 paired normal lung tissues and test cohort of 107 primary NSCLCs (58 AC, 49 SCC) and 25 unrelated normal lung tissues. Predominant downregulation of miRNAs in NSCLC was demonstrated, with miRNAs whose dysregulation was specific to AC or SCC histology identified. Notably, miR-205 was identified as a lung SCC specific miRNA, highly upregulated in SCCs compared to normal lung tissues (fold change of 29.41). Integration of miRNA and existing mRNA gene expression data identified putative miR-205 target genes and pathways.
A 5-miRNA classifier (miR-375, miR-203, miR-205, miR-630 and miR-223) was identified for distinguishing lung ACs from SCCs, able to correctly classify 83% and 77% of primary NSCLCs in both the TPCH training (n=63) and test (n=107) cohorts, with validation in a publically available external NSCLC cohort (n=104).
Aberrant miRNA expression in asbestos-related lung cancers was observed; however, expression changes were not dose-dependent in relation to asbestos fibre burden. miR-663 was downregulated in asbestos-related lung tumours and expression validated in an external cohort. Integrative analysis of miRNA and mRNA gene expression data allowed for identification of potential miR-663 target genes.
miRNA expression profiles differed between NSCLCs with and without FGFR1 amplification (SCCs) and KRAS mutation (ACs); however, high FDRs and non-significant global permutation tests imply that miRNAs may not be truly differentially expressed. This study identified one downregulated and 23 upregulated miRNAs in EGFR+ compared to EGFR-WT tumours, in a cohort of 93 lung ACs with mixed smoking history. miR-34a demonstrated upregulation in EGFR+ TPCH ACs by both microarrays and qRT-PCR, and confirmed using an independent test set of lung ACs. Integration of miRNA and existing mRNA gene expression data identified potential miR-34a target genes and pathways.
Using a semi-supervised principal component method based on Cox proportional hazards models, 3-miRNA signatures were developed to predict prognosis in primary SCCs (miR-127, miR-377, miR-376a) and ACs (miR-29c, miR-34b, miR-375). These signatures not only performed well as independent predictors of recurrence and survival, but showed improved performance when used in combination with clinical covariates.
In conclusion, this thesis employed a genome-wide approach to characterise miRNAs aberrantly expressed in primary NSCLC, including in clinically important subgroups of NSCLC, identifying putative oncogenic and tumour suppressing miRNAs involved in lung cancer. miRNA-based signatures were identified for predicting prognosis in lung cancer patients; however, require further validation. Although miRNAs are emerging as novel diagnostic and therapeutic agents, the role of miRNAs in lung carcinogenesis is likely to be extensive and complex, highlighting the need for further investigation prior to development of miRNA-based diagnostics or therapeutics.