DNA Circle-sequencing lowers the single molecule sequencing error threshold and identifies ultrasonication as a source of DNA damage

Abstract

DNA mutation is the ultimate source of all heritable genetic and phenotypic variation. On human timescales, DNA mutation results in the evolution of antibiotic resistance, viral resistance to vaccines, the emergence human cancers, and more. Precise measurement of DNA mutation is therefore desirable to rapidly detect and analyze low frequency mutations in a population of cells. Precise measurement of DNA mutation by high throughput sequencing has been hindered by sequencing error rates, which occur at rates 1x10-2 to 1x10-3 per base sequenced. Previous work on circle sequencing had pushed a putative sequencing error rate down to roughly 2.8x10-4 per base for genomic yeast DNA in the absence of DNA repair enzymes, where the expected number is 4.7x10-9 per base per generation. Through revision of the assay, we are now capable of taking 125ng of genomic DNA and obtaining mutation rate estimates with a sequencing resolution floor of 2x10-7 per base in the absence of repair enzymes, a resolution improvement of over 3 orders of magnitude. In practice, circle-seq recovers some of the mutation spectrum of mismatch-repair deficient E. coli, although some signature of C [->] T and G [->] A errors remain present. Curiously, it calls a mutation rate of 2x10-9 per base per generation, lower than expected, underscoring the difficulty of directly comparing mutation rates per base to MA mutation rates per site per generation. This protocol readily recognizes Covaris ultrasonication as mutagenic in library preps, with a mutation spectrum dominated by G:C [->] A:T, G:C [->] C:G, and G:C [->] T:A errors. The protocol is either detecting in vivo DNA damage that has yet to be repaired, is significantly biased in its ability to detect mutations, or is not yet sensitive enough to detect mutation rates of MMR-E. coli cells. The protocol may have some use in measuring mutation abundance from tissue samples, and in vitro sources of DNA damage. The protocol also highlights some biases inherent to single-stranded DNA (ssDNA) cyclization, including a 10-bp periodicity which echoes the constraints of dsDNA cyclization.