Genome-wide detection of chromosomal rearrangements, indels, and mutations in circular chromosomes by short read sequencing

Abstract

Whole-genome sequencing (WGS) with new short-read sequencing technologies has recently been applied for genome-wide identification of mutations. Genomic rearrangements have, however, often remained undetected by WGS, and additional analyses are required for their detection. Here, we have applied a combination of WGS and genome copy number analysis, for the identification of mutations that suppress the growth deficiency imposed by excessive initiations from the Escherichia coli origin of replication, oriC. The E. coli chromosome, like the majority of bacterial chromosomes, is circular, and DNA replication is initiated by assembling two replication complexes at the origin, oriC. These complexes then replicate the chromosome bidirectionally toward the terminus, ter. In a population of growing cells, this results in a copy number gradient, so that origin-proximal sequences are more frequent than origin-distal sequences. Major rearrangements in the chromosome are, therefore, readily identified by changes in copy number, i.e., certain sequences become over- or under-represented. Of the eight mutations analyzed in detail here, six were found to affect a single gene only, one was a large chromosomal inversion, and one was a large chromosomal duplication. The latter two mutations could not be detected solely by WGS, validating the present approach for identification of genomic rearrangements. We further suggest the use of copy number analysis in combination with WGS for validation of newly assembled bacterial chromosomes.

Figure 1.

Chromosomal rearrangements detected by read frequency analysis. The number of reads starting at each base is plotted as a function of the relative position on the reference chromosome; oriC is set to 0, and ter is set to ±1 to reflect the bidirectional DNA replication. (Blue rhombus) The average of 1000-bp windows; (red squares) the average of 10.000-bp windows. Any window containing repeat sequences is omitted. The oriC/ter ratio is the ratio of reads per base pair at oriC to reads per base pair at ter. Sequencing template was isolated from: (A) an overnight culture of MG1655; (B) an exponentially growing culture of MG1655; (C) an exponentially growing culture of hsm-7; and (D) an exponentially growing culture of hsm-8. (C,D) Clear deviations from the pattern of the host strain MG1655 (B) indicating an inversion and a duplication, respectively.

Figure 2.

The chromosome of hsm-7 is inverted between two IS5 elements. The read frequency analysis of hsm-7 (Fig. 1C) indicates an inversion between two inverted copies of the insH gene encoding the IS5 transposase and trans-activator. (A) Genetic map of the two copies of insH and their neighboring genes in the wild-type (wt) and in the suggested hsm-7 configuration. The heads of the arrows indicate the approximate locations of PCR primers designed to differentiate between the wt and the hsm-7 configuration by amplifying the insH genes with neighboring sequences. The expected size of each PCR product is indicated. (B) PCR analysis of MG1655 DNA (wt) and hsm-7 DNA (-7) with the primers shown in A. MG1655 shows the expected wild-type fragments of 1604 bp and 1962 bp. hsm-7 shows the 1399-bp and the 2167-bp fragment expected from the inversion.

Figure 3.

The rrnA–rrnE region of the chromosome of hsm-8 is duplicated. The read frequency analysis of hsm-8 (Fig. 1D) indicates a duplication of the rrnA–rrnE region. This duplication can be caused by homologous recombination between sequences repeated in the rrnA and rrnE ribosomal RNA operons creating a rrnE/A chimeric operon and duplicating the entire sequence between rrnA and rrnE. (A) The duplication in hsm-8 compared to wild type (wt). (B) A detailed read frequency analysis was made to narrow down the recombination point. (16) rrs genes for 16S RNA; (t) genes for tRNA; (23) rrl genes for 23S RNA; (5) rrf genes for 5S RNA. The read frequency of a 500-bp window on each side of the rrnA and rrnE operons and at positions with unique sequences (shown by triangles) within the rrn operons in hsm-8 was normalized to the read frequencies of the same positions in wt. The relative read frequency of the 500-bp window to the left of rrnA was set to 1. The read frequency shifts from 1.0 to higher than 2 on the opposite side of the 16S RNA gene. This indicates that the entire part of the rrnA operon downstream from the 16S RNA gene is duplicated. In the rrnE operon the read frequency drops fourfold from the left side to the right side of the operon. This fourfold drop is in agreement with Figure 1D and indicates a long DNA replication time for the rrnE operon. The read frequency at the only measurable point in the rrnE operon is lower than half the read frequency of the left side and higher than the read frequency of the right side and is less conclusive. The combined analysis of rrnA and rrnE indicates that recombination took place within the 16S RNA genes. (C, left) PCR primers were designed to amplify the rrsA, rrsE, and the rrsE/A chimeric genes. (Heads of the arrows) The approximate locations of the primers and the expected PCR product sizes are indicated. (C, right) PCR products obtained from MG1655 (wt) and from hsm-8 (-8). hsm-8 shows the native rrnA and rrnE operons as well as the rrnE/A chimeric operon, whereas MG1655 only shows the native rrnA and rrnE operons.

Figure 4.

Simulation of detection of errors in de novo assembled bacterial sequences. (A) Chromosomal DNA is replicated bidirectionally from oriC (black arrow) indicated by the green line (left arm) and blue line (right arm). Strain W3110 (right) carries an inversion between the ribosomal operons rrnD and rrnE (red arrows) but is otherwise very similar to MG1665 (left). This inversion includes the origin of replication, oriC, and displaces oriC 215 kb to the left compared to MG1655. (Green and blue) The inverted chromosomal arms. Reads generated from DNA template of a stationary phase (B) and an exponentially growing culture of MG1655 (C) were mapped to the genomic sequence of MG1655 (left) and W3110 (right). The number of reads starting at each base pair was calculated for a 10-kb-wide sliding window (red squares), and windows covering repeat sequence were removed. The ideal read frequencies were calculated using an oriC/ter ratio of 2.7 and shown with the same color code as in A.