Temporal dynamics and metagenomics of phosphorothioate epigenomes in the human gut microbiome

Abstract

Background: Epigenetic regulation of gene expression and host defense is well established in microbial communities, with dozens of DNA modifications comprising the epigenomes of prokaryotes and bacteriophage. Phosphorothioation (PT) of DNA, in which a chemically-reactive sulfur atom replaces a non-bridging oxygen in the sugar-phosphate backbone, is catalyzed by dnd and ssp gene families widespread in bacteria and archaea. However, little is known about the role of PTs or other microbial epigenetic modifications in the human microbiome. Here we optimized and applied fecal DNA extraction, mass spectrometric, and metagenomics technologies to characterize the landscape and temporal dynamics of gut microbes possessing PT modifications.

Results: Exploiting the nuclease-resistance of PTs, mass spectrometric analysis of limit digests of PT-containing DNA reveals PT dinucleotides as part of genomic consensus sequences, with 16 possible dinucleotide combinations. Analysis of mouse fecal DNA revealed a highly uniform spectrum of 11 PT dinucleotides in all littermates, with PTs estimated to occur in 5-10% of gut microbes. Though at similar levels, PT dinucleotides in fecal DNA from 11 healthy humans possessed signature combinations and levels of individual PTs. Comparison with a widely distributed microbial epigenetic mark, m⁶dA, suggested temporal dynamics consistent with expectations for gut microbial communities based on Taylor's Power Law. Application of PT-seq for site-specific metagenomic analysis of PT-containing bacteria in one fecal donor revealed the larger consensus sequences for the PT dinucleotides in Bacteroidota, Firmicutes, Actinobacteria, and Proteobacteria, which differed from unbiased metagenomics and suggested that the abundance of PT-containing bacteria did not simply mirror the spectrum of gut bacteria. PT-seq further revealed low abundance PT sites not detected as dinucleotides by mass spectrometry, attesting to the complementarity of the technologies.

Conclusions: The results of our studies provide a benchmark for understanding the behavior of an abundant and chemically-reactive epigenetic mark in the human gut microbiome, with implications for inflammatory conditions of the gut.

Fig. 1.

Workflows and optimized technologies for identifying and quantifying PT dinucleotides and mapping PT sites in gut microbes. (A) Purified fecal DNA was analyzed for PT dinucleotide content by LC-MS and subjected to PT-seq for metagenomic analysis of bacterial identity and PT consensus sequence. (B) Workflow for optimized fecal DNA isolation and purification, which improved yield by 10-fold. (C) Workflow for optimized PT-seq. The streptavidin capture step significantly reduced the level of noise and increased the read pileup sensitivity.

Fig. 2.

LC-MS analysis of mouse and human fecal DNA for PT dinucleotides by reveals the presence of PT-containing microbes. (A) Fecal DNA from individual C57BL/6 mice reveals a relatively uniform spectrum of PT dinucleotides with insignificant differences between males and females. Data represent mean ±SD for 20 mice. (B) Fecal PT dinucleotide spectra differ among 11 human donors. Data represent mean ±SD for N=11. (C) Analysis of PT dinucleotides in fecal DNA from donor #5 over 22 months reveals constant proportions of G*A, C*C, and C*A dinucleotides. Data represent mean ±SD for three separate DNA isolations from a single fecal sample. The bars representing PT levels for each of the 3 detected PT dinucleotides are superimposed, not stacked. (D) Overlay of m⁶dA levels on the time course of total PT dinucleotides from panel C. m⁶dA data represent mean ±SD for three separate DNA isolations from a single fecal sample. (E) Taylor’s Power Law (V = amb) analysis of temporal dynamics of PT dinucleotides and m6dA in Donor #5, where the mean abundance of a species (m) in a mixed population fluctuates over time with variance (V) linearly related to m to the power of b. Here a plot of ln(V) = ln(a) + b*(ln(m)) yields b = 1.4. (F) Correlations among the levels of the three PT dinucleotides from Donor #5.

Fig. 3.

The taxonomic composition of microbiome and quantification of PT sites. (A) The abundance of microbiome was estimated by metagenomic sequencing using Kraken2 and Bracken. The phylogenetic composition of microbiome taxa collapsed at the phyla level in fecal sample (Donor #5). (B) The number of different PT modification motif sites identified by ICDS aligned to genomes of human gut microbiome isolates. A total of 26,817 sites were detected, with PTs denoted by a ‘*’.