|Home | About | Journals | Submit | Contact Us | Français|
Conceived and designed the experiments: LF PRR RL LW. Performed the experiments: LF YR YR WW JW. Analyzed the data: PRR RL YR CG ZZ JC WQ DL LW. Contributed reagents/materials/analysis tools: PRR RL YR. Wrote the paper: LF PRR RL LW.
Cholera, caused by Vibrio cholerae, erupted globally from South Asia in 7 pandemics, but there were also local outbreaks between the 6th (1899–1923) and 7th (1961–present) pandemics. All the above are serotype O1, whereas environmental or invertebrate isolates are antigenically diverse. The pre 7th pandemic isolates mentioned above, and other minor pathogenic clones, are related to the 7th pandemic clone, while the 6th pandemic clone is in the same lineage but more distantly related, and non-pathogenic isolates show no clonal structure. To understand the origins and relationships of the pandemic clones, we sequenced the genomes of a 1937 prepandemic strain and a 6th pandemic isolate, and compared them with the published 7th pandemic genome. We distinguished mutational and recombinational events, and allocated these and other events, to specific branches in the evolutionary tree. There were more mutational than recombinational events, but more genes, and 44 times more base pairs, changed by recombination. We used the mutational single-nucleotide polymorphisms and known isolation dates of the prepandemic and 7th pandemic isolates to estimate the mutation rate, and found it to be 100 fold higher than usually assumed. We then used this to estimate the divergence date of the 6th and 7th pandemic clones to be about 1880. While there is a large margin of error, this is far more realistic than the 10,000–50,000 years ago estimated using the usual assumptions. We conclude that the 2 pandemic clones gained pandemic potential independently, and overall there were 29 insertions or deletions of one or more genes. There were also substantial changes in the major integron, attributed to gain of individual cassettes including copying from within, or loss of blocks of cassettes. The approaches used open up new avenues for analysing the origin and history of other important pathogens.
Cholera was recognised in accounts from ancient India and China , but it was only in 1817 that an epidemic in India spread westwards to become the first of 7 pandemics. The etiological agent of pandemic cholera was the bacterium Vibrio cholerae serotype O1, which was later divided into two biotypes: classical biotype and El Tor biotype. The first six pandemics were caused by the classical biotype, but in the early 1900s a new form of the species emerged that caused mild or no symptoms, and was later put into the El Tor biotype, to distinguish it from the form causing cholera (classical biotype). The El Tor biotype differed by being haemolytic, and positive in the Voges Proskaur reaction that detects a specific fermentation pathway. The 6th pandemic continued until 1923, but from then on was largely confined to Asia (and thus not pandemic). The 7th pandemic, which erupted in 1961, was El Tor in biotype, which was quite unexpected. However there had been several outbreaks of cholera, now known as pre-7th pandemic outbreaks, in the Makassar area, Sulawesi, Indonesia between 1937 and 1957. The organism was El Tor but the disease resembled true cholera in severity and mortality. It differed in not spreading rapidly but remained localised as reported in a review soon after the events . The 7th pandemic (also El Tor) arose in the same area, and is generally thought to have arisen from the Makassar outbreak strain. It spread rapidly and eventually fully replaced the classical biotype 6th pandemic strain, which was present in India throughout the period between the 6th and 7th pandemics and for many years after the 7th pandemic arrived (see also Supporting Introduction in Text S1).
V. cholerae has also been found in water, particularly brackish water, and associated sediments, and also on invertebrates in these habitats. These environmental isolates are phenotypically very diverse with over 200 serotypes reported. They resemble the El Tor biotype, but the term is not used as the biotyping is confined to serotype 1. The fact that the 6th and 7th pandemics are of different biotype, with the 7th pandemic strain more similar to environmental isolates than the 6th, suggests that the 7th pandemic strain did not arise from the 6th pandemic but arose independently.
We have already looked at the relationships of the pathogenic strains using sequence of 26 house keeping genes . This confirmed earlier conclusions from several groups, based on MLEE and sequence data, that the 6th and 7th pandemic clones, the pre 7th pandemic outbreak isolates, and some other pathogenic strains are related. However with 26 genes and 12 isolates an unambiguous tree was obtained, with no reverse or parallel events. It was also clear that the 6th pandemic clone stands apart from the other pathogenic strains (Figure 1). Single base change mutational and multi bases change recombinational events could be distinguished, and this showed that there has been extensive recombination, with 8 of the 26 genes having undergone recombination in either 6th or 7th pandemic lineage since divergence. A 1937 Makassar outbreak isolate (strain M66-2) appeared to be on the direct line to the 7th pandemic. The picture that we have for V. cholerae is of a species that is diverse with as many O antigen forms as E. coli, with one lineage that has adapted to a new niche as an intestinal pathogen of humans. On at least 2 occasions this pathogenic form developed a high level of transmission to become pandemic , .
These circumstances provide an ideal opportunity to study at DNA level the events that enabled the Makassar outbreak strain to become a pandemic strain and enabled the 7th pandemic clone to fully replace the 6th clone. We sequenced the genomes of strain M66-2, and 6th pandemic isolate O395, and compared them with the published genome of 7th pandemic isolate N16961. We describe a method to distinguish mutation from recombination-based changes over a full genome, and this provided the data to assess the rate of mutational change, as the 7th pandemic strain N16961 was isolated in 1971, and the 6th pandemic strain, O395, was isolated in 1965. Comparison of the genomes enables us to document the changes that had occurred during divergence of the three clones. Significant findings include an estimate for the mutation rate that is 100 times higher than generally assumed, a model for change in the major integron, and assessment of mutational, recombination, integron and other evolutionary events to the separate lineages.
The V. cholerae M66-2 and O395 genomes were sequenced and, as expected for Vibrio genomes, both comprise a large and small chromosome. The statistics for the three genomes are shown in Figure 2A and Table S1. The genomes are generally collinear, but in O395 each chromosome has an inversion relative to the others (Figure 3).
As predicted from our analysis of individual gene sequences, we could distinguish recombinant segments with a relatively higher frequency of base substitutions, and developed a formal procedure for identifying recombinant segments in the genome (for details see Supporting Methods in Text S1; Figure 4 and Figure S1). Base changes outside these segments were treated as mutations, and each mutation or recombination event was allocated to the M66-2 lineage, the N16961 lineage, or to the divergence between the joint M66-2/N16961 lineage and the O395 lineage (the O/NM divergence), as shown in Table 1 and Figure 2B. We also showed that recombinant segments were randomly distributed in the M66-2 and N16961 lineages, and those attributed to the O/MN divergence are probably also randomly distributed, but there is some distortion probably because with 32% of the genome affected, some recombinant segments overlap (see Supporting Methods in Text S1). It can be seen that there were far more recombination and mutation events during the O/NM divergence, than in either M66-2 or N16961 lineage, and also that M66-2 had undergone significant change. There is no genome sequence suitable for use as an outgroup to allocate events to the O or MN lineages, as examination of a series of genes shows that there has been extensive recombination, such that individual gene trees are not congruent and not useful for analysis of strain relationships. However for the 368 mutational changes we can ask if that particular base has undergone any change in the genome sequences that we have for comparison, which are in effect a complex array of lineages reassorted by recombination, as if not we can use that information in effect as an outgroup. We took 5 genomes that are outside of the pathogenic lineage under study, and for each of the mutational SNPs in the O/MN divergence, determined the base present in these strains. Of the sites the base present was same in all 5 genomes, and was either the base found in the O395 genome or that in both N16961 and M66-2. On this basis we allocated most of these changes to either the O or MN lineage. For some sites the sequence segment involved was not present in any of the outgroup strains, while for others it was not sufficiently consistent to enable us to determine the lineage. For details see Table S2. In this way we were able to root the tree as shown in Figure 2.
The level of change observed in M66-2 puts it on a side branch to the main lineage to N16961, and this was not predicted from the history, as M66-2 was isolated in 1937, in the first of a series of El Tor outbreaks from 1937–1959, in Makassar, the same region in which the 7th pandemic arose in 1961, and it had appeared from that to be on the direct lineage to the 7th pandemic clone (see Supporting Introduction in Text S1).
Recombination has introduced substantial changes to these genomes with 32.60% replacement during the O/NM divergence and 9.52% during divergence of N16961 and M66-2 (Table 1). The size distribution of recombinant segments is given in Figure S2. The largest recombinant segment is 70 kb with the majority between 1 and 20 kb. V. cholerae can undergo gene transfer by conjugation, transduction and transformation but we have no means of inferring the processes involved. However the sizes are all within the range observed for transduction in E. coli . Some must result from overlap between events but clearly recombination can introduce large segments of DNA. Sequence variation within recombinant segments varies from 0.15% to 17.65% suggesting that all or most donors are V. cholerae. In the 3 lineages that we can distinguish, there are in total 467 intragenic mutations involving 415 genes, and 181 recombination events involving 1446 genes. However, there are 24893 base substitutions in recombinant segments, more than 44 times the number (554) attributed to mutation (Table 1). However some of the substitutions in the recombinant segments will be due to mutation after the recombination event. The number of mutations that would have occurred if the rate had been the same in recombinant segments as elsewhere is 740 (Table 1). If we assume that half of the additional substitutions would have occurred before recombination and thus lost, we have 24800 of the observed substitutions due to recombination and 647 due to mutation, indicating that there have been 38 times as many of the substitutions observed due to recombination than to mutation.
The distribution of pseudogenes is interesting (Figure 2B and Table S3). The common ancestor of the 3 strains must have had very few pseudogenes as there are only 6 common to the three genomes, but there are 53 cases of pseudogene formation during divergence. For pseudogenes we assume that in general the mutation in the lineage involved loss of function, and hence can attribute the pseudogene mutations that occurred during O/NM divergence to either the O or MN lineage. On this basis 30 occurred in the O395 lineage, 3 in the N16961/M66-2 common lineage, and 7 and 13 respectively in N16961 and M66-2 after their divergence (Table S3).
Most of those pseudogenes differ from the parent form only by a single frameshift mutation, or base change to give a stop codon. Also the 6 pseudogenes common to the 3 strains, are identical in all 3, except for VC0254 which has an additional indel in O395. Note that because even very low levels of sequencing error would affect these conclusions, the sequences for the regions involved were verified as discussed in Supporting Methods Text S1. Most pseudogenes are present in only one strain and a priori it is possible that the mutations involved occurred after strain isolation. We therefore looked at the sequence data for MO10, an O139 strain isolated in 1992, the first year in which that O139 variant of the 7th pandemic clone was observed, as it is a descendent of the O1 form of the 7th pandemic clone, although N16961 may not be on the direct lineage. The data suggested that at least 4, and perhaps 6, of the 7 pseudogenes attributed to the N16961 lineage, predated isolation of N16961 (2 genes were not in the current release of the MO10 sequence so no information, and one was a functional gene, so that N16961 pseudogene could have arisen before or after isolation). Details for the genes involved are given in Supporting Information Text S1.
V. cholerae has a very large integron which fits the usual pattern of having at the 5′ end an intI gene followed by an attI site, with a series of cassettes downstream. In common with other Vibrio species it has many more cassettes than in the archetypal drug resistance integrons, and is often called a superintegron. Also the attC sites at the 3′ end of each cassette have a relatively conserved sequence . Both M66-2 and O395 include a major integron that fits this pattern, and is in the same location as in N16961 . The 3 intI genes are identical as are the attI sites, and there are 139, 142 and 179 cassettes respectively in the integrons of M66-2, O395 and N16961, with 168, 188 and 214 genes respectively. V. cholerae attC sites are generally species specific . And using data from several genomes we were able to show that some in the V. cholerae genomes came from other Vibrio species, presumably associated with new cassettes (see Supporting Discussion Text S1 and Figure S4 for details).
There is substantial conservation of cassette order with cassettes generally in blocks with a common cassettes order within the blocks, and block boundaries determined by differences in distribution of the blocks among the 3 strains. The distribution of these blocks, named A–R, is shown in overview in Figure 5, with details for blocks and cassettes given in Figures 5, S3, and S4. Figures 5 and S4 in particular can be usefully used in conjunction. Blocks D, G, H, I, K, M. P, Q and R contain mostly cassettes present in all 3 or the same 2 strains, and in the same order in each strain. These cassettes are presumed to have been in the relevant common ancestor, and are therefore treated as a single occurrence of that cassette, to give a total of 271 cassette locations as shown in Figure S4. The distribution of cassettes at each location is shown in Table S4.
Many cassettes are present at more than one location within a strain or at different locations in different strains, as has been noted before . The sequences are generally identical and it seems that they arose by duplication rather than independent transfer from other strains. There were other cassettes that were near identical for which the same applies, as some divergence by mutation is expected, and for the purposes of this analysis we consider both identical and near identical (less than 1% sequence difference) cassettes to have arisen by duplication within the lineages under study. Treating these near identical cassettes as variants of the same form gives us a total of 207 cassette forms occupying the 271 locations. The relationships of the 207 cassette forms were assessed by using BLASTN with a cutoff e-value of 10−10, which puts 92 of them into 22 groups of 2 or more related cassette forms. The distribution of these cassettes is shown in Table S4. There were an additional 79 forms (Table S4, cassette forms 93–171) that were present in more than one strain and/or location but not allocated to groups because they are identical or near identical within each form. Finally there are 36 cassettes (Table S4, cassette forms 172–207) that are found in only one strain at one location.
Integrons can gain new cassettes at the 5′ end by incorporation at attI . M66-2, N16961 and O395 have blocks (A, B and C) of 40, 9 and 4 cassettes respectively at the 5′ end, that fit the pattern expected for cassettes added sequentially at attI since their divergence. M66-2 and N16961 share the downstream block D, also potentially added cassette by cassette at attI prior to strain divergence (Figure 5). A phylogenetic analysis of the attC sites indicates that many of these cassettes were of foreign origin (details in Supporting Discussion Text S1).
However as discussed above many cassettes are present at more than one location. There are 39 cassettes are present in duplicate, and 6, 3 and 1 with 3, 4 or 5 copies respectively (see Table S4), all most easily explained by copying of pre-existing cassettes in the same integron. The details of this proposal are presented in Figure 5 and Figure S4. For these cassettes, most of the upstream (new) copies are separated as a group from most of the downstream (donor) copies, although there is overlap of the 2 regions (Figure 5B). This pattern is assumed to reflect copying from downstream to upstream locations, as the only known mechanism for adding individual cassettes is at the upstream end of the integron. On this hypothesis 21 of the 40 cassettes in block A, proposed to have been added to M66-2 since divergence from N16961, were copied from downstream cassettes, one (M6) from another cassette in block A, 12 from cassettes still in M66-2, and 8 from cassettes in N16961 and presumably since lost in M66-2.
The original evidence was for excision of cassettes as circular DNA and integration at attI. It was shown that IntI can excise a cassette to give a circular form, and that this circular form can then be incorporated at the attI site, again by IntI, to give a new cassette downstream of attI, effectively transferring a cassette from one site to another. These experiments involved cassettes in different integrons, so gave no indication of the status of the donor integron after “excision”, but a priori did not provide a model for copying of a cassette while leaving the original intact. However under the current model for the action of IntI , , the formation of the translocation intermediate involves only a single strand of the donor cassette, and it is entirely plausible that at the time of its removal as an IntI-associated folded single strand, it would be replaced by repair DNA synthesis on the complementary strand. This would generate the pattern that we observe if this IntI-associated copy were later inserted into the same integron. Indeed our observations provide support for such a model for generation of the IntI-associated recombination intermediate. There are alternative explanations for the copying process, such as copying from one fork to another after DNA replication, or by involvement of the chitin induced DNA transformation between cells of the same clone forming a biofilm on chitin, . The latter paper showed that DNA released by one cell could be taken up by another in the biofilm. The experiments involved homologous recombination but any circular cassettes present would also be released, and if taken up could be incorporated at attI. If both donor and recipient were from the same “colony” in the biofilm, there would be duplication as inferred from the distribution of cassettes.
There are 29 insertions and deletions involving at least one gene, which we refer to as large indels to distinguish them from mutations involving one or few base pairs. They are named B1 to B20, and S1 to S9 respectively in the large and small chromosomes (Table S5), and marked in Figure S1 or Figure 6. Three (B9, B16 and B20) can be identified as deletions and 2 (S2 and B17) as insertions in M66-2, as we have O395 as outgroup, and likewise for the 3 presumed insertions (B2, B3 and B5) in N16961. In addition B8 and S3 are thought to be a deletion and insertion respectively in O395 for reasons given in the Supporting Discussion Text S1. The remaining 19 indels are attributed to events during the O/NM divergence, with 15 present and 4 absent in O395.
The most significant additions are the 2 known islands (VSP-I and VSP-II)  in N16961 and 2 novel islands in O395 (B1 and B4). O395 also has a K139-like and 2 Mu-like phages. The CTXΦ is an indel because of its absence in M66-2, and an extra copy present on the small chromosome in O395, which is identical in sequence to that on the large chromosome. More interestingly, the CTXΦ genomes of O395 and N16961 contain different rstR genes at the ends, and have significant base differences throughout the phage genome except for ctxA, which is identical, and ctxB which differs by only 2 bases (Figure 6), suggesting replacement of the whole or part of the CTXΦ in one of the strains. There is further discussion of the indels in Supporting Discussion Text S1.
With full genome sequences for 3 of the isolates it is possible to estimate the divergence times for the 2 nodes present in Figure 2B. We used only mutational changes identified above to estimate the timeframe for the changes, and only synonymous substitutions as is traditional –, referring to them as synonymous SNPs (Single-Nucleotide Polymorphisms) as at each site there are 2 alternative nucleotide bases. We excluded regions inferred to have undergone recombination, but extrapolated from the number of mutational synonymous SNPs observed to cover the whole genome (see Table 1 and Supporting Methods Text S1). The exclusion of recombinational SNPs has been used before for whole genomes, for example for Y. pestis ,  but the problem of recombination did not arise in these studies (see also Supporting Methods Text S1). The widely used Dykhuizen and Whittam divergence rates ,  gave 10,000 to 50,000 years for divergence of the two pandemic strains and 2,000 to 10,000 years for divergence of M66-2 and N16961 (Table S6). This takes us long before epidemiological records for pandemic cholera and is also not consistent with the Makassar outbreak being a precursor to the 7th pandemic.
However it has recently been shown that species that diverged less than about a million years ago have accumulated up to 10 fold more than the expected number of SNPs, as many of the mutations seen in intraspecies comparisons are later lost due to random genetic drift or purifying selection , . With an accurate record of the history of these strains, and in the absence of any remaining basis for an externally based rate of divergence, we took the split between M66-2 (isolated in 1937) and N16961 (isolated in 1971) as the basis for assessing relative time frames. There have been 14 synonymous mutations in M66-2 since its divergence from N16961 compared with 47 for N16961. To calculate the divergence rate, we assumed the same rate of change from divergence to isolation in 1937 and 1971 respectively, which gave a rate of 0.97 SNPs/genome/year and a divergence date for M66-2 and N16961 in 1923. Applying the same rate of change to their common MN lineage gives an estimated divergence date of 1880 for the 6th and 7th pandemic strains (see Methods for details). Note that as for pseudogenes, the sequences for the regions involved were verified as discussed in Supporting Methods Text S1. These dates are clearly estimates. There will be statistical sampling errors, and these will be significant as the number of synonymous SNPs between N16961 and M66-2 is low, and this was used to calibrate the mutation rate used to estimate both divergence dates, Also there is no reason to expect a constant rate of divergence over such short periods, as many factors will vary between the different lineages and there is not time for these to average out. For example the average generation time will depend markedly on the fraction of time spent in the host and environment, which will vary over time, Also the effective population size will vary and affect retention of the mutations involved. Indeed O395 and N16961 have clearly had different mutation rates as O395, isolated in 1965, has 30% more mutations than N16961 isolated in 1971. However, even as approximations, these estimates fit well with the fact that the El Tor outbreaks first appeared in 1937 and the 6th pandemic started in 1899. They are much more realistic than the 10,000–50,000 years ago as estimated using conventional rates.
We conclude that for the V. cholerae pandemic strains the traditional criteria give divergence times about 100 fold too long. The rate estimates of Ho et al. ,  were based on related species or forms of a species with good fossil data for divergence times, but our conclusion on much more limited data seems realistic given that 1 million years gave a 10 fold increase over the long term rate for accumulation of divergence, and we are dealing with much shorter periods. It seems clear that as for the cases analysed by Ho et al., we must use data appropriate to the timeframe involved rather than assume a constant rate for accumulation of SNPs. It is also probable that for clones within species, there will be variation in rates of divergence between species and even within species for clones in different niches. If we are to have realistic estimates of divergence times we will need to accumulate data for a range of species from different environments. We also caution that this first application of the approach is based on a very small sample and certainly not so precise as to preclude the possibility that pandemics 1 to 6 were due to the same clone as has also been suggested .
A major reason for studying the evolution of V. cholerae is to better understand its pathogenicity and pandemicity. We undertook several analyses focusing on this issue, and while the results were informative, there were no clear “suspects” for the change from causing local outbreaks as exhibited by the Makassar outbreak clone represented by M66-2, and pandemic capacity as exhibited by the 7th pandemic clone, represented by N16961 (for details see Supporting Discussion Text S1, Figure S5, Figure S6 and Table S7).
The 3 genomes we compare represent critical stages in the history of cholera. The 6th and 7th pandemic strains were both major in their time, but 7th has now fully replaced the 6th  and is clearly fitter in evolutionary terms. The Makassar outbreak is seen as being at a critical stage in the development of the 7th pandemic strain. The three strains are closely related such that recombination events were generally well separated in the genome, enabling us to distinguish between mutation and recombination events. The genomes have been very dynamic with substantial changes through mutation, recombination, acquisition of genes in islands, and acquisition of cassettes in the major integron. We now have the catalogue of genetic differences between the 3 strains that must be responsible for differences in pathogenicity and pandemicity, opening up new areas for research.
We have been able to put a realistic timeframe to the divergence of the pandemic strains because 1) we could distinguish between mutation and recombination events and used only the mutational SNPs to calculate divergence rate, and 2) N16961 (7th pandemic) is thought to be derived from a Makassar outbreak of known date and represented by M66-2, which allowed us to infer a divergence rate, that could then be used to infer a date for divergence of the 6th and 7th pandemic strains. V. cholerae is unusual in having a defined date and location for the origin of a major pandemic that allowed this approach. There are other landmarks in the spread of V. cholerae which will allow extension of the approach to further refine our understating of the evolution of pandemic cholera, and suitable circumstances are found in some other species. There are certainly other estimates of divergence times for important human pathogens that are surprising and worth reanalysis, such as E. coli O157:H7 (~40,000 years) , S. enterica serovar Typhi (10,000–43,000 years) , and our own estimate for Shigella (35,000–270,000) .
For most methods and further details of those discussed below, see Supporting Methods Text S1.
V. cholerae strains M66-2 and O395 were sequenced by random shotgun sequencing. For M66-2 the final genome is based on 62,004 reads from an ABI 3730 sequencer. For O395 the final genome is based on 14,210 reads from an ABI 3730 sequencer and 334,497 reads from a Roche 454 GS-FLX System. Open reading frames from 30 amino acids in length were predicted using Glimmer 3.0 and verified manually using the annotation of N16961. Artemis  was used to collate data and facilitate annotation. Pseudogenes were detected by comparing the genome sequences of M66-2, N16961 and O395 using the program Psi_Phi  and verified manually (for details see Supporting Methods Text S1).
The complete nucleotide sequences and annotations of the V. cholerae strain M66-2 and O395 have been deposited at GenBank under the accession number CP001233, CP001234, CP001235 and CP001236 respectively.
The genomes of N16961 and MO10 were downloaded from GenBank (RefSeq accession numbers NC_002505, NC_002506 and AAKF00000000). Alignment of the genomes was done using Multi-LAGAN  and checked visually. Because of the significance that each variation carries in such closely related strains, we verified the sequence accuracy for all mutational SNPs and all pseudogenes (for details see Supporting Methods Text S1 and Table S8).
N16961 was isolated in 1971, and is thought to have diverged from M66-2 in about 1937, when M66-2 was isolated in the first of the El Tor outbreaks in Makassar. There are 14 synonymous SNPs since the MRCA with N16961 compared with 47 for N16961 (Table 1). We assumed that the rate of change per year (r) was the same in both lineages, ie (1937-X) * r=14 and (1971-X) * r=47, where X is the divergence date and r is the number of mutations accumulated per annum since divergence. This is equivalent to r=(47−14)/(1971−1937) or 0.97 per year and X is 1923. There are 39 synonymous SNPs which can be allocated to the MN common lineage. We applied the same rate of mutation to the MN common lineage, but first notionally allocated the 14 synonymous SNPS in the O/NM divergence that could not be allocated to the O or NM lineage, in the proportion of those that could be allocated to the O and MN lineages, to give a final allocation of 42 to the MN lineage and a divergence date of 1880 for the MRCA of N16961 and O395. The number of synonymous SNPS allocated to the O395 lineage on this basis is 129 for a mutation rate of 1.5 of synonymous SNPs per year.
This Supporting Text file includes Supporting Introduction, Supporting Discussion, Supporting Method and Supporting reference.
(0.19 MB PDF)
Plot of mutations, recombination events and indels in the genomes of strains M66-2, N16961 and O395
(6.69 MB PDF)
Histogram of the lengths of the recombinant segments
(0.29 MB PDF)
Comparison of the 3 integrons using artemis comparson tool and dot plots
(0.58 MB PDF)
Relationships of cassettes in the integrons of strains M66-2, N16961 and O395
(0.31 MB PDF)
Function Categories (COGs) of genes with or without mutational changes
(0.05 MB PDF)
Correlation of gene expression level with ratio of synonymous and non-synonymous substitution rates
(0.46 MB PDF)
Distribution of Inter-SNP segment (ISS) lengths for the 6 chromosomes
(0.37 MB PDF)
General features of the M66-2, N16961 and O395 genomes
(0.02 MB PDF)
Allocation of SNPs in the O/MN divergence to either O395 or MN lineage
(0.17 MB PDF)
Pseudogenes allocated to the M66-2, , N16961, MN and O395 lineages
(0.04 MB PDF)
Integron cassette groups
(0.08 MB PDF)
Major indels in the 3 genomes
(0.04 MB PDF)
Substitution rates and divergence times
(0.04 MB PDF)
Genes in recombinant segments with protein product having ks/ka less than 4
(0.10 MB PDF)
Corrections to N16961 genome bases
(0.06 MB PDF)
Ning Wei, Zihua Han, Gang Yuan, Zhirong Ren, Junguo Shen, Li Li and Peng Du are acknowledged for their work in DNA sequencing.
Competing Interests: The authors have declared that no competing interests exist.
Funding: This work was supported by the Tianjin Municipal Special Fund for Science and Technology Innovation Grant 05FZZDSH00800, the National Natural Science Foundation of China (NSFC) Key Programs Grants 30530010 and 20536040, NSFC General Program Grant 30670038, the National 863 program of China Grants 2006AA020703 and 2006AA06Z409, The National Key Science and Technology Project from Ministry of Health of the People's Republic of China Grant 2008ZX10004-009 and the National 973 program of China Grants 2007CB9143010 and 2009CB522603. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.