Reader Comments

Post a new comment on this article

Recombination between Typhi and Paratyphi A is a confounding factor here, which could be addressed using dN/dS instead

Posted by kholt on 09 Dec 2008 at 15:25 GMT

It is entirely possible that secreted effector proteins have evolved under different selective constraints in host-adapted vs host-generalist serovars of Salmonella, but the analysis of D(N) alone doesn't demonstrate this. To detect different selective pressures, it would be much more informative to use the usual dN/dS approach. This is particularly crucial in the present study, since we already know the serovars are not equally divergent.

It was reported in 2007 that approximately a quarter of the Typhi and Paratyphi A genomes have been recombined relatively recently in their evolutionary history. The central observation of this study was that, while most Salmonella enterica serovars showed pairwise nucleotide divergence of ~1.1% across their genomes, the distribution of nucleotide divergence between Typhi and Paratyphi A was different: about 3/4 of the genome looked like any other pair of serovars, i.e. 1.1% divergence, while the other quarter was much less divergent, ~0.1%. The authors then used models to demonstrate that the most likely explanation for this is convergence due to recombination between Typhi and Paratyphi A, rather than recent divergence of the serovars. The paper is available in PubMed Central:
"A bimodal pattern of relatedness between the Salmonella Paratyphi A and Typhi genomes: Convergence or divergence by homologous recombination?"
Xavier Didelot, Mark Achtman, Julian Parkhill, Nicholas R. Thomson and Daniel Falush
Genome Res. 2007 January; 17(1): 61–68
PMCID: PMC1716267

The recombination between Typhi and Paratyphi A needs to be taken into account when interpreting the data presented in the present study. The authors here focus only on D(N), the measure of non-synonymous SNPs between different pairs of serovars... low D(N) between Typhi and Paratyphi A for a particular gene is interpreted as a different evolutionary constraint in these human-adapted serovars compared to the host-generalist serovars considered. However, most of the genes discussed in this paper have particularly low levels of divergence between Typhi and Paratyphi A, suggesting they were exchanged via recombination - e.g. 0-0.6% in sipD, sseC, sseD, sseF. In contrast sifA has 2.1% nucleotide divergence, suggesting it was not recombined between Typhi and Paratyphi A, which is probably why it has a higher D(N). Without normalising to D(S), i.e. calculating dN/dS, it is impossible to say whether the similarities observed between Typhi and Paratyphi A genes is due to selective pressures associated with their host-restriction, or simply chance recombination. The latter would be an example of convergent evolution, which may or may not have provided a selective advantage for the recombinant strain. For example, the nucleotide sequence for sipD is identical in Typhi and Paratyphi A, which says nothing about selection on non-synonymous SNPs, especially given the background information about recombination. Furthermore, dN/dS between Typhimurium and Enteritidis sequences of sipD is 0.30, whereas dN/dS between Typhi and Enteritidis is similar at 0.33, and Typhi-Typhimurium is 0.35 (calculated using the online dN/dS tool SNAP at

RE: Recombination between Typhi and Paratyphi A is a confounding factor here, which could be addressed using dN/dS instead

dipa replied to kholt on 31 Dec 2008 at 10:12 GMT

Dn/Ds Comparison
Many thanks to Kholt for the interesting comment entitled “Recombination between Typhi and Paratyphi A is a confounding factor here, which could be addressed using dN/dS instead. This issue was raised by couple of reviewers before publication and we were able to convince them with following arguments favouring our approach:
1. One should be cautious while interpreting Dn/Ds ratio especially when analyzing closely related bacterial strains. It has been shown convincingly (see the reference below) that if two genomes are more closely related, the Dn/Ds ratio tends to be more and therefore, when one compares two closely related strains, time since divergence of the two strains needs to be considered. Unfortunately, we don’t have any information about the time of such divergence in case of Salmonella. Therefore, plain Dn/Ds ratio might not give clear insights in our case of host specificity in Salmonella and we looked at differential evolution instead of positive selection.
Reference: Comparisons of Dn/Ds are time dependent for closely related bacterial genomes. Journal of Theoretical Biology (2006) 239: 226-235.
2. Another study shows that high Dn/Ds ratios among closely related bacterial species can be statistical artifact.
Reference: Microevolutionary genomics of bacteria, Theoretical Population Biology (2002) 61: 435–447.
3. Moreover, Dn/Ds ratio less than one can not completely rule out positive selection. That implies that positive selection can’t be detected using this approach (of Dn/Ds).
4. Differentially evolved genes that we have identified in our study encode proteins that form the translocons (but not basal apparatus) of type three secretion systems (TTSS) and some of their effector proteins. All these proteins directly interact with the host system. Thus, it is very likely that the evolution of the genes encoding these proteins is influenced by the host system.

5. In support of our observations, sseF and sseC have a unique genetic polymorphism in human adapted serovars that is absent in other serovars
Reference: Genetic determinants and polymorphisms specific for human-adapted serovars of Salmonella enterica that cause enteric fever. Journal of Clinical Microbiology (2006)44: 2007-2018.

6. We have proved statistically that evolution of the ‘differentially evolved genes’ are significantly different from rest of the genome and the pathogenicity islands.