Genome-wide detection of spontaneous chromosomal rearrangements in bacteria

Abstract

Genome rearrangements have important effects on bacterial phenotypes and influence the evolution of bacterial genomes. Conventional strategies for characterizing rearrangements in bacterial genomes rely on comparisons of sequenced genomes from related species. However, the spectra of spontaneous rearrangements in supposedly homogenous and clonal bacterial populations are still poorly characterized. Here we used 454 pyrosequencing technology and a 'split mapping' computational method to identify unique junction sequences caused by spontaneous genome rearrangements in chemostat cultures of Salmonella enterica Var. Typhimurium LT2. We confirmed 22 unique junction sequences with a junction microhomology more than 10 bp and this led to an estimation of 51 true junction sequences, of which 28, 12 and 11 were likely to be formed by deletion, duplication and inversion events, respectively. All experimentally confirmed rearrangements had short inverted (inversions) or direct (deletions and duplications) homologous repeat sequences at the endpoints. This study demonstrates the feasibility of genome wide characterization of spontaneous genome rearrangements in bacteria and the very high steady-state frequency (20-40%) of rearrangements in bacterial populations.

Figure 1. Workflow of detection and verification of SGRs in S. typhimurium.

The SGRs detection and verification procedures in this work are as followings: (i) bacterial cells were grown in a chemostat for five days; (ii) samples were collected at each day and 454 pyrosequencing was performed on genomic DNA prepared from three samples collected at day one, two and three; (iii) reads with ‘split mapping’ signature were mined from the three datasets and further subjected to the confirmatory screening based on the three listed criteria; (iv) A substantial fraction of putative rearrangements were selected for experimental verification using padlock probe hybridization and/or PCR.

Figure 2. Summary of stepwise detection and verification of genome rearrangements.

(A) After initial screening based on the ‘split mapping’ signatures (B) After removal of artifacts and quality score analysis (C) After confirmatory screening based on the three criteria (D) After Padlock Probe and PCR verification (expected true rearrangements).

Figure 3. Size distribution of putative deletions.

The size distribution of putative deletions with sizes less than 50 kb was examined for the three datasets, gen48, gen144 and gen240.

Figure 4. Junction microhomology analysis.

The distribution of overlapping microhomologies at junctions was compared between the three datasets (gen48, gen144 and gen240) and one simulated dataset using 20 million in silico chimeric reads. The observed junction microhomology distribution in the datasets gen48, gen144 and gen 240 were represented by triangles and the simulated distributions were represented by boxplot. (A) Comparison between the dataset gen48 and the simulated dataset (181 rearrangements). (B) Comparison between the dataset gen144 and the simulated dataset (120 rearrangements). (C) Comparison between the dataset gen240 and the simulated dataset (42 rearrangements).

Figure 5. Design rationale for padlock probes.

The two end segments of the padlock probes and the connector sequence were designed to be complementary to three consecutive sequences in the target rearrangement junction sequence. If the two end segments and connector sequences are perfectly hybridized a closed circular molecule can be formed by two ligations. For the wild type sequence, only one ligation can occur leading to a non-circularized molecule.

Figure 6. Deduced frequencies of expected true SGRs.

The frequencies were calculated as the number of expected true SGRs divided by the sequencing coverage for the three datasets gen48, gen144 and gen240, respectively. The distributions of the three rearrangement events (deletion, duplication and inversion) were compared between each pair of the three datasets using chi-square two-sample test.