Systematic identification and quantification of phase variation in commensal and pathogenic Escherichia coli

Abstract

Bacteria have been shown to generate constant genetic variation in a process termed phase variation. We present a tool based on whole genome sequencing that allows detection and quantification of coexisting genotypes mediated by genomic inversions in bacterial cultures. We tested our method on widely used strains of Escherichia coli, and detected stable and reproducible phase variation in several invertible loci. These are shown here to be responsible for maintaining constant variation in populations grown from a single colony. Applying this tool on other bacterial strains can shed light on how pathogens adjust to hostile environments by diversifying their genomes.

Figure 1

Phase variation caused by inversion. (A) Two genotypes (blue and green circles) are consistently and reproducibly prevalent whenever a single bacterium is grown to a population in a phenomenon termed phase variation. The two genotypes are distinguishable by a genomic inversion - a mutation which occurs when a fragment of DNA residing between two inverted repeats (IRs) is detached from the chromosome, and is then reattached in a reverse manner, resulting in a switch between the two strands. The two phenotypes may differ, for example, if a promoter located inside the fragment changes orientation and alters the transcription (gray arrow) of genes outside the inverted segment. (B) Phase variation in the fim operon. A DNA segment (shaded area) containing the fimA promoter can switch between two phases: an ON phase, where the promoter is correctly oriented, and the fim operon is expressed, and an OFF stage, where it is silenced. The OFF state also destabilizes the DNA recombinase fimE, probably by transcribing its antisense.

Figure 2

Whole genome sequencing and detection of inversions. (A) In the WGS process, sequenced genome is shredded into inserts approximately 500 bp long. Each insert is sequenced from both ends (paired ends), resulting in a pair of approximately 100 bp reads. Each read is mapped independently to the reference genome, and the gap size between the insert’s edges is determined for each pair. The gap size of each read is then plotted against the read’s genomic location. As long as the actual genome is identical to the reference genome, we expect a 'ribbon' formation around 500 bp (gray diamonds). (B) Experimental paired-end data exhibiting the ribbon formation. (C) When the sequenced genome deviates from the reference genome by an inversion (represented by gray shading), inserts whose reads lie on both sides of the inversion’s edge display a unique pattern that we term a 'funnel' (two symmetric diagonal lines composed of abnormally aligned reads). (D) Experimental paired-end data exhibiting a funnel around an inversion (blue diamonds represent plus strand paired with plus strand and green diamonds represent minus strand paired with minus strand). Note that only abnormal gap size reads are shown. (E) Results of the systematic inversion detection algorithm for two strains of E. coli. Exact genomic coordinates are available in Table S1 in Additional file 1.

Figure 3

MGY e14 phase variation. (A) ORF analysis of the phage e14 invertible locus. The invertase pinE resides next to the inverted locus (represented by a shaded rectangle). In the reverse orientation stfE is attached to ycfK, producing a longer ORF than in the forward variant (fusion of the red and green segments). ORFs in all figures were inferred using SnapGene® software (from GSL Biotech, Chicago, IL, USA). (B) Gap size distribution plotted against chromosomal position, centered on the e14 invertible locus. Two formations coexist at the same locus: a ribbon formation of normal reads (gray), and a funnel formation of abnormal reads (blue and green). The relative abundance of each formation represents the relative fraction of each genotype in the bacterial population. The IRs flanking the inversion are marked by orange rectangles (C) PCR confirmation of the coexistence of two genotypes. PCR was conducted on a single MGY colony with two sets of primers. Extracted genomic DNA was used as template for both sets (see Additional file 1 for description of primers). Each band represents the existence of one orientation in the population.

Figure 4

Complex phase variation in EPEC. (A) Two overlapping 'funnel' formations indicate a complex structure of PV. A large inversion (around 2,200 bp) and a smaller inversion (around 1,800 bp) coincide within the same module. (B) Sequence analysis revealed three homologous inverted repeats in the locus (green arrows), which allow for the two inversions. Further analysis indicated four possible variants. Each variant can mutate into two of the other variants by any of the two inversions. (C) While the large inversion retains stable proportions in all clones, the small inversion is unstable and displays great variance between samples. Error bars represent standard deviation between five independently sequenced and analyzed single colonies.

Figure 5

Detection of inversions of different sizes. (A) A summary of all inversion detection techniques presented in this paper and the conditions in which they are applicable. Small inversions will be evident as a sequence of SNPs or by a concentration of soft trimmed reads, while large inversions flanked by oversized IRs can be discovered by mate-pair WGS or by coverage trends. (B) Funnel detection in mate-pair data: gap size against genomic location plots centered on both ends of a mega-inversion. Mate-pair WGS with 2 kbp insert size reveals a funnel pattern in the boundaries of a suspected inverted segment. This funnel is not seen when using a 500 bp insert size. (C) PCR confirmation of the inversion. The wild-type (wt) and mutated strains were compared, using two sets of primers forward (F) and reverse (R), corresponding to both orientations. (D) Inversion detection by coverage trends. Coverage plots of the entire chromosome of the KLY mutant depict the average coverage of a genomic area against its location. Top: mapping to the reference genome reveals a 700 kbp disruption in the coverage trend caused by the mega-inversion. Bottom: mapping to a revised reference genome incorporating the mega-inversion negates the disruption. The origin of replication (ori) and replication terminus (ter) are indicated by arrows.