Reader Comments

Post a new comment on this article

Some comments

Posted by JimDoval on 11 Dec 2012 at 20:13 GMT

Dear authors,

Thank you first for this valuable paper dealing with an issue has never been addressed before to my knowledge.
I would have some comments I hope you'll find constructive, at least interesting. I separated my comments in several points:

(i) I was first surprised by the Pc variants of the sequences that happened to be different to what is claimed in the text:
- IVS2 (AM991977) is claimed to be PcW although it displays the so-called hybrid PcH1 variant that is of intermediate strength but determines the same IntI1 allele. However, while there is no promoter interferences between PcW and PintI1, Guerin et al., 2010 detected some with PcH1 that yielded into slightly reduced integrase expression level. Moreover, the IntI1 of IVS2 displays another mutation, namely Gly91Val that does not appear in the other IntI1 investigated and therefore might be a potential bias. Likewise, even though it has been shown not to interfere with the PintI1 activity, the P2 promoter is present in IVS2 while it is not in another, which may also be another source of experimental bias.
- IVS3 (JX259274) is described as being an hybrid promoter, the PcH2 from what I could see; here again the IntI1 displays an usual Thr5Ile mutation.

Therefore I wondered whether there was any pelicular reason for the choice of PcH1+P2 instead of PcW and PcH2 rather to PcH1?

(ii) Likewise, the chosen integrons also differ in the number and the nature of the gene cassettes. Here again I wondered why you did not choose isogenic constructs to ensure the gene cassettes could not somehow interfere with the fitness phenotypes.

(iii) Did you check that the evolved IVS integrons with truncated/inactivated integrases still contained the same arrays (and with no mutations)? Did you notice some variations in the Pc promoter, PintI1 promoter or in the attI1 region that are also likely to greatly reduce the shuffle of gene cassettes? Were the evolved strains safe for any genetic rearrangements (PFGE patterns) ?
I remember that early reports dealing with integrons reported a potential toxicity effect associated to the overexpression of intI1, probably for the reasons you evoke in the discussion, do you think this could, at least in part, account for your results ?

(iv) You discuss the high frequency of inactivated/disrupted integrase in nature, although the quoted papers deals with chromosomal integrons ("superintegrons"). I would suggest to perform an overview of the truncated intI1 among the thousands class 1 integrons sequences available in the public database (although most of them do not display the whole intI1); I performed such analysis several years ago and could not significant differences between the (truncated+mutated)/total ratios between PcS-integrons and PcW-integrons (55-60% in both cases).

Sorry for this long comment and thanks again for this interesting report!

No competing interests declared.

RE: Some comments

pjj000 replied to JimDoval on 15 Dec 2012 at 11:15 GMT

Dear Dr. Doval,
Thank you for your interest in our work. Here’s our response to points 1-4 above:

1. We’d like to point out that we did not ”choose” strains with different promoters for these studies. Our original hypothesis was that ”integrons are selectively neutral” and that this may in part be an explanation for their successful spread. The different promoter variants were analyzed after we made the unexpected discovery that newly acquired integrons dramatically reduced fitness in the absence of antibotic selection.

Your comments on the Pc variants: You are right about the published accession number (the original report on the Salmonella integron), and we are grateful that you made us aware of this. We have off-course re-analyzed our own sequence data of both IVS2 as well as the donor strain and our claims in the paper stand firm. IVS2 has a weak promoter identical to the one published by Jove and co-workers, 2010 PloS Genetics. As soon as we get the accession numbers we will make a minorl correction tin the paper. As you will see in the published sequence, the Gly91Val mutation is not there either.

2. Our data suggest that the expression of gene cassettes have little, if any, effects on bacterial fitness. We initially chose class-1 integrons with different cassette arrays to see if these had any impact on host fitness. Isogenicity in our study was assured between the competitors in head to head competition experiments where the only difference was the presence/absence of integrons. Thus we could measure the biological cost of carrying three different class-1 integrons.

As shown in Table S3, the evolved integrons all revealed unchanged kanamycin MIC’s indicating that the cassette expression levels were unchanged. Moreover the fitness effects of the frameshift mutations were in the same ranges as the experiments with insertionally inactivated integrase. We think these evidence strongly support the non-functional integrases being the mechanistic explanation for the observed fitness amelioration.

Rearrangements in the evolved strains cannot be assured. However these strains were not used in competitions. If we did just that it would be difficult to separate fitness ameliorations from other adaptive mutations (adaptation to the growth conditions). The evolved integrons were in fact transferred back into the ancestral genetic backgrounds where genetic rearrangements are unlikely. In the back transfers no change in size of the integrons, or differences in resistance profiles, were observed.

3. If you in your over-expression argument mean that we could have inserted the integrons into a region where the integrase was overexpressed we think this is unlikely. Unpublished data showed that the cassette arrays were very stable and we could only occasionally detect circularized cassettes using an inverse PCR approach.

4. Yes, many of the truncated integrases are present in chromosomal super-integrons, but far from all. Moreover, as mentioned in the paper, the most important data that actually support our hypothesis on a costly integrase are those who demonstrate a link between non-functional integrases and absence of LexA-regulation. We believe this adds generality to our data.

No competing interests declared.