Comprehensive Analysis Of A:T>T:A Mutational Signatures In Human Cancers

Maude ARDIN, International Agency for Research on Cancer, France
BOUAOUN L. 2 , BYRNES G. 2 , HERCEG Z. 3 , HOLLSTEIN M. 1,4 , ALEXANDROV L. 5 , ZAVADIL J. 1 , OLIVIER M. 1

1 Molecular Mechanisms and Biomarkers Group, International Agency for Research on Cancer, Lyon, France
2 Biostatistics Group, International Agency for Research on Cancer, Lyon, France
3 Epigenetics Group, International Agency for Research on Cancer, Lyon, France
4 Faculty of Medicine and Health, University of Leeds, Leeds, UK
5 Center for Nonlinear Studies, Los Alamos National Laboratory, Los Alamos, New Mexico, USA

Purpose
Exposure to the genotoxic compound Aristolochic Acid (AA), an IARC Group 1 carcinogen,  leads to renal disease and urological and hepatobilliary cancers. AA-related tumors carry specific types of mutations caused by AA mutagenesis which are characterized by a high prevalence of A:T>T:A transversions in a 5'-Py-A-Pu-3' sequence context with transcriptional strand-bias. The presence of this mutational signature may thus inform of exposure to AA. However, the specificity of the signature towards AA has not been fully demonstrated. In order to address this question and to characterize the general landscape of A:T>T:A mutations in human cancers, we performed a systematic screen of A:T>T:A-based mutational signatures present in publicly available cancer genome data.
 
Method
Somatic mutation data from TCGA, ICGC and COSMIC repositories were analysed for mutational signatures using the non-negative matrix factorization (NMF) algorithm applied solely to A:T>T:A mutations. Tumors from patients with documented AA exposure were analysed as positive controls. Signature comparisons were performed using the cosine similarity method.
 
Results
From 253,292 samples available in public repositories, we identified 1,814 samples passing a minimum threshold number of A:T>T:A mutations. In these samples and 44 samples from AA-exposed cases, we identified six A:T>T:A-based mutational signatures. Four of these signatures matched with the A:T>T:A profiles of previously published signatures, including one that matched both signature 22 (AA) and the A:T>T:A component of signature 4 (smoking). The remaining two signatures, one observed in some esophageal tumors and the other in a subset of skin cancer, were new.
 
Conclusions
Our preliminary analyses revealed that the sequence context of A:T>T:A mutations caused by AA is not specific to AA, and that other distinct A:T>T:A signatures contribute to the mutation load in specific cancer types, for which the causative agents or processes remain to be identified.
 
Funding sources
International Agency for Research on Cancer