104 Probing DNA shape on molecular and genomic scales

Abstract

Linear DNA sequence encodes its three-dimensional structure. The resulting DNA shape gives DNA unique physical and chemical properties that are explored by DNA-binding proteins (Rohs et al., 2009). Probing DNA shape, therefore, helps to reveal the origin of specificity in DNA-protein recognition. We have developed a Monte Carlo (MC) method for all-atom DNA structure prediction (Joshi et al., 2007). With key components such as parallelization and functionality to handle chemical modifications added, the revamped MC method can efficiently recover the DNA shape of any DNA sequence. To embrace the genomic era, we have trained a high-throughput (HT) prediction method on 2121 independent MC simulations (Slattery et al., 2011). Applying this method on whole genome DNA sequences of 66 Drosophila individuals, we found that among single nucleotide polymorphisms (SNPs) within the immediate vicinity of in vivo transcription factor (TF) binding sites, low-frequency SNPs tend to change the DNA minor groove width more than high-frequency SNPs. This indicates that shape-changing mutations are detrimental to TF binding and in vivo function. We also examined the dependency of DNase I cleavage events on DNA shape using our HT method. Our results show that the rate of DNase I cleavage closely tracks the width of the minor groove. MC predictions for chemically modified bases further demonstrate that cytosine methylation enhances cleavage directly adjacent to CpG dinucleotides through narrowing of the minor groove. In summary, with tools on hand to address biological questions on both molecular and genomic scales, our work provides new insights into gene regulation and evolution.