No abstract

Although various methods have been developed for sequencing cytosine modifications, it is still challenging for specific and quantitative sequencing of individual modification at base-resolution. For example, to obtain both true 5-methylcytosine (5mC) and true 5-hydroxymethylcytosine (5hmC) information, the two major epigenetic modifications, it usually requires subtraction of two methods, which increases noise and requires high sequencing depth. Recently, we developed TET-assisted pyridine borane sequencing (TAPS) for bisulfite-free direct sequencing of 5mC and 5hmC. Here we demonstrate that two sister methods, TAPSβ and chemical-assisted pyridine borane sequencing (CAPS), can be effectively used for subtraction-free and specific whole-genome sequencing of 5mC and 5hmC, respectively. We also demonstrate pyridine borane sequencing (PS) for whole-genome profiling of 5-formylcytosine and 5-carboxylcytosine, the further oxidized derivatives of 5mC and 5hmC. This work completes the set of versatile borane reduction chemistry-based methods as a comprehensive toolkit for direct and quantitative sequencing of all four cytosine epigenetic modifications.

We present long-read Tet-assisted pyridine borane sequencing (lrTAPS) for targeted base-resolution sequencing of DNA methylation and hydroxymethylation in regions up to 10 kb from nanogram-level input. Compatible with both Oxford Nanopore and PacBio Single-Molecule Real-Time (SMRT) sequencing, lrTAPS detects methylation with accuracy comparable to short-read Illumina sequencing but with long-range epigenetic phasing. We applied lrTAPS to sequence difficult-to-map regions in mouse embryonic stem cells and to identify distinct methylation events in the integrated hepatitis B virus genome. BackgroundRecent advances in third-generation sequencing methods, including PacBio SMRT sequencing [1-3] and Oxford Nanopore sequencing [4], have enabled long-read and single-molecule sequencing that is distinct from the mainstream short-read Illumina sequencing. These newer sequencing platforms allow unambiguous mapping of repetitive and complex regions of the genome and provide unprecedented opportunities for detecting structural variants, phasing haplotypes, and assembling genomes [5,6]. While Nanopore sequencing still has a high error rate (~10%), the latest SMRT sequencing provides accuracy similar to Illumina sequencing (99.8%) but with an average read length of 13.5 kilobase (kb) compared to~0.3 kb with Illumina [3].Long-read sequencing of DNA modifications, particularly the two abundant modifications-5-methylcytosine (5mC) and 5-hydroxymethylation (5hmC) [7,8], is needed to obtain phased epigenomes that will enable new understanding of the functions of epigenetic modifications, for example allele-specific methylation in genomic imprinting [9] and heterogeneous cancer samples, and diagnosis of brain tumors [10]. Although the SMRT and Nanopore platforms can detect DNA modifications directly, there are major barriers to their application. SMRT sequencing can directly detect DNA modifications using polymerase kinetics information, but requires a minimum of 250× per strand coverage to detect 5mC [11], largely defeating the purpose of long-read sequencing. Several computational methods have been developed to detect base modifications directly from Oxford Nanopore sequencing [12][13][14][15]. However, these approaches require complicated training data from control DNA samples of known methylation status and sophisticated computational analysis, limiting their accuracy to determine 5mC. Moreover, both native SMRT and Oxford Nanopore DNA methylation sequencing require microgram levels of native, unamplified DNA as input. Since