Selective microbial genomic DNA isolation using restriction endonucleases

Abstract

To improve the metagenomic analysis of complex microbiomes, we have repurposed restriction endonucleases as methyl specific DNA binding proteins. As an example, we use DpnI immobilized on magnetic beads. The ten minute extraction technique allows specific binding of genomes containing the DpnI Gm6ATC motif common in the genomic DNA of many bacteria including γ-proteobacteria. Using synthetic genome mixtures, we demonstrate 80% recovery of Escherichia coli genomic DNA even when only femtogram quantities are spiked into 10 µg of human DNA background. Binding is very specific with less than 0.5% of human DNA bound. Next Generation Sequencing of input and enriched synthetic mixtures results in over 100-fold enrichment of target genomes relative to human and plant DNA. We also show comparable enrichment when sequencing complex microbiomes such as those from creek water and human saliva. The technique can be broadened to other restriction enzymes allowing for the selective enrichment of trace and unculturable organisms from complex microbiomes and the stratification of organisms according to restriction enzyme enrichment.

Figure 1. Analysis of biotinylated DpnI.

(A) pUC19 was incubated with DpnI, DpnI-biotin or commercially sourced DpnI in the presence or absence of 10 mM magnesium chloride. The digested fragments were separated on a 1.5% agarose gel. (B) A FAM-labeled DNA duplex containing one G^m6ATC site was incubated with increasing amounts of DpnI or DpnI-biotin (0 to 1200 ng). The reactions were separated on a 20% TBE gel and analyzed with fluorescence imaging. (C) An unmethylated 651 bp DNA fragment and a Dam-methylated 477 bp DNA fragment were combined and incubated with increasing amounts of immobilized DpnI-biotin (80–180 µl). DNA was eluted using GTC and desalted. All fractions were separated on a 3% agarose gel.

Figure 2. Efficiency and range of DpnI pull-down.

(A) Immobilized DpnI was incubated with a mixture of E. coli and human DNA for varying amounts of time. 40% of E. coli DNA binding occurs on less than one minute. Less than 0.2% of human DNA binds to DpnI. (B) Immobilized DpnI was incubated with E. coli DNA in buffers with pH of 4, 8 or 10. Almost all E.coli DNA was recovered in the range of pH tested. (C) A fixed amount of human DNA (1 µg) was mixed with decreasing levels of E. coli DNA and then incubated with immobilized DpnI. Approximately 80% of E. coli DNA is recovered down to levels of 10 fg. All data shown is the average of three experiments. (D) A fixed amount of E. coli DNA (1 ng) was mixed with increasing amounts of human DNA and then incubated with immobilized DpnI. There is a slight decrease in the recovery of E. coli DNA with increasing amounts of human DNA. However, even when human DNA is present at 10,000x, DpnI recovers over 70% of E. coli DNA.

Figure 3. DpnI enriches prokaryotic DNA as determined by NGS.

(A) NGS reads from the input, bound and unbound fractions of the synthetic mix. Reads from the input map overwhelmingly to human and rice, with less than 10% mapping to prokaryotes in the synthetic mixture. Less than 10% of the reads from the bound fraction map to human with the majority mapping to E. coli. (B) DamMT+ genomes are enriched 30 to 70-fold versus their input levels, and 300 to 800-fold versus human DNA. (C) Genomes that lack DamMT are enriched when compared to human and rice.

Figure 4. NGS coverage maps for E. coli from input (A) and bound (B) fractions of the synthetic mixture.

Reads were mapped to E. coli O157:H7 EDL933 and binned into 1000 bp bins. (A) The average depth of coverage is 0.5 for E. coli in the input fraction (green), with 62% of the genome covered. (B) For the bound fraction (blue), the average depth of coverage increases to 60 and 99.5% of the genome is covered. The input fraction (green) is also plotted here for comparison to the bound fraction at the same scale.

Figure 5. NGS data of saliva samples.

(A) Donut plots depicting relative abundance of identifying reads for microbial and human genomes in input, bound and unbound samples. (B, C) Pairwise plots of sample fractions versus input. (B) Bound vs. input. (C) Unbound vs. input. Plotted points are identifying reads of genera. To facilitate direct visual comparison between samples reads were normalized to 10 M total.

Figure 6. NGS data of creek samples.

Pairwise plots of sample fractions versus input. (A) Bound vs. input. (B) Unbound vs. input. Plotted points are identifying reads of genera arbitrarily normalized to 10 M reads total.