Artificial polyploidy improves bacterial single cell genome recovery

Abstract

Background: Single cell genomics (SCG) is a combination of methods whose goal is to decipher the complete genomic sequence from a single cell and has been applied mostly to organisms with smaller genomes, such as bacteria and archaea. Prior single cell studies showed that a significant portion of a genome could be obtained. However, breakages of genomic DNA and amplification bias have made it very challenging to acquire a complete genome with single cells. We investigated an artificial method to induce polyploidy in Bacillus subtilis ATCC 6633 by blocking cell division and have shown that we can significantly improve the performance of genomic sequencing from a single cell.

Methodology/principal findings: We inhibited the bacterial cytoskeleton protein FtsZ in B.subtilis with an FtsZ-inhibiting compound, PC190723, resulting in larger undivided single cells with multiple copies of its genome. qPCR assays of these larger, sorted cells showed higher DNA content, have less amplification bias, and greater genomic recovery than untreated cells.

Significance: The method presented here shows the potential to obtain a nearly complete genome sequence from a single bacterial cell. With millions of uncultured bacterial species in nature, this method holds tremendous promise to provide insight into the genomic novelty of yet-to-be discovered species, and given the temporary effects of artificial polyploidy coupled with the ability to sort and distinguish differences in cell size and genomic DNA content, may allow recovery of specific organisms in addition to their genomes.

Figure 1. B. subtilis response to PC190723 treatment: Increased cell size.

(A) Typical cytographs from treatments illustrate increase in cell size in PC190723-treated B. subtilis versus untreated control and DMSO-treated cells at 60 minutes after treatment. Crosshairs were made at the mode of the untreated control population and applied to the DMSO- and PC190723-treated cytographs. Resultant quadrants give the relative percentage of the population (from 100,000 cells) inhabiting each quadrant. Thus, in quadrant two (Q2) of the control population, 44.7% of the cells are observed, while over 80% are observed in the PC190723-treated population. DMSO has no effect on cell size. Percentages are calculated by omitting the lower left population (A Control: black circle) from each cytograph, which likely contains dead cells and cell debris. (B) By determining ΔQ2 over time, cell size is shown to plateau after 60 minutes. Error bars are 95% CI. (C) Typical light microscope images of sorted untreated control and PC190723-treated cells. PC190723-treated cells are longer than untreated control cells, but the diameter appears unchanged.

Figure 2. B. subtilis response to PC190723 treatment: Increased DNA content.

DNA content was assessed for PC190723 treated (maximum effect from ten replicates) and control cells via six replicates of 50 sorted cells for primer sets A (located near the origin) and B (located near the terminus). Theta or multi-fork replication may account for the discrepant reports from both primer sets. Cells treated for 50 minutes had more DNA content than the respective untreated controls. Prolonged drug exposure may account for the reducing DNA content at 60-min treatment as related to DNA degradation. Error bars are 95% CI.

Figure 3. PC190723-treated cells have increased genomic coverage and less amplification bias.

(A) Genome coverage of mapped reads increases with treatment. Also, LBS correlates with genome coverage (R² = 0.6998, p = 0.1634). Best-fit line determined by least-squares linear regression. (B) Mapped to B. subtilis ATCC 6633 genome, untreated control cells have more gaps (red lines) than inhibited cells. Similar gap placement on the genome suggests amplification bias. Because of increased genomic template in polyploid cells, the number of gaps is greatly reduced (black circles).

Figure 4. Treated cells have better sequencing results.

Genome coverage following de novo genome assemblies similarly increases as LBS decreases due to treatment with PC190723 (R² = 0.6033, p = 0.2233). Best-fit line determined by least-squares linear regression.