Re: Sequences from Tamiflu Resistant Immunocompromised Patients Released
Hmm, there's a subtlety here I hadn't appreciated previously, and I'd like it if you could clarify matters. When you say that recombination is widespread, do you mean that a single patient gets infected by two different flu strains which then recombine, or (bearing in mind that flu transmits as a population rather than a single sequence) do you mean that there is recombination between the various sequences comprising a single infection?
Toy model here, let's have a single strain A which splits into substrains B and C. Some time after B and C have split, a given polymorphism X appears in both of them. Three possibilities:
1) The mutation X occurs twice independently in substrains B and C
2a) Mutation X occurred once on branch B. Subsequently a patient became simultaneously infected with substrains B and C, and the polymorphism recombined onto C.
2b) Mutation X occurred once in A, the common ancestor of B and C. It was not detected because it was at low levels and didn't show up in the consensus sequence. After B and C have diverged, each of them still retains a low proportion of sequences carrying polymorphism X. Subsequently X is able to separately spread to fixation within B and C, because polymorphisms are freely mobile within each individual population.
If I understand you rightly, you're saying that 2b is the major scenario. That would have significant and testable implications. In particular:
* Since the mutation is deduced to have occurred in A, all other branches from A are at risk of acquiring polymorphism X, whereas strains branching off earlier are at much less risk of acquiring X. Thus you can predict which subclades are more likely to acquire a given change.
* If you could sequence all the individual virions from a given individual, scenario 2b looks like it should make fairly strong predictions about the distribution of polymorphisms between the individual virions. For example, if you had three polymorphisms with a minor allele frequency of 1/10, then 1/1000 of the virions should have all three polymorphisms. This is not what you'd predict from a random mutation hypothesis.
In genetic terms, what scenario 2b would mean is that within a single infected individual, flu should be modelled as a panmictic, freely recombining population. Transmission between patients is equivalent to a severe population bottleneck, with a subset of the flu population being passed on. Once transmission has occurred, there is then a geographic barrier to further "interbreeding" between the daughter populations, in that they're now existing in separate people! If someone gets infected by two strains, it's the equivalent of breaking one of those geographic barriers and re-merging two populations, which may or may not subsequently be compatible enough to recombine with each other. Dual infection seems likely to be rare (but more likely in epidemic / pandemic conditions).
Can you confirm that I've understood your model correctly? Panmixis between the individual sequences present in a single infection is a lot easier to grasp than the nebulous "everything recombines with everything" vibe I get from many of the posts here (not necessarily yours).
The dual infections are quite common and I can give you a specific example with real plaque purification (cloning) to show that recombination really does happen and multiple species can be generated involving the exchange of SNPs.
It was a rather dramatic example of many of the points I have been making, and some of the results have been described
http://precedings.nature.com/documents/459/version/4
However, the data that would make you a "believer" in recombination in influenza has been block by politics (of the worst kind), but the data are quite real.
The story began when I was collaborating with NAMRU-3 on H5N1 outbreaks in Egypt. They had been only sequencing HA and I suggested that they expand the sequencing effort and the highest prioity would be NA. The timing was good becasue it was the end of 2006 and one of the first NA sequences generated was for the Ghrabiya cluster, and the NA sequence showed tamiflu resistance. It wasn't the common H274Y that has been discussed here, but was N294S, which produced somewhat weaker resistance, 10-20 fold instead of 300-1000 fold seen in H274Y. However, 10-20 fold was more than high enough for treatment problems becasue most treatments are borderline, so a requirement for 10-20X more Tamiflu would render that approved treatment level useless.
However, the remarkable aspect of the resistance was the fact that N294S was present BEFORE treatment raising concerns that the polymorphism was on a fit H5N1 and was widespread in birds. As a result, NAMRU-3, which usually focused on human cases, was getting more poultry samples. A fairly extensive seach failed to find N294S in birds or patients, but on Feb 15, a Gharbiya-like sequence was identified in a chicken in Gharbiya, but it didn't have N294S
http://www.ncbi.nlm.nih.gov/nuccore...uence.Sequence_ResultsPanel.Sequence_RVDocSum
However, the NA sequence was a mixture and one position was represented with a Y and that position was polymorphic in the H5N1 sequences in Egypt (other isolates had both "C" and "T" at that position), including two other chicken isolates from Gharbiya that had 11 differences in the NA sequence.
The same was true for the HA sequence
http://www.ncbi.nlm.nih.gov/nuccore...uence.Sequence_ResultsPanel.Sequence_RVDocSum
which had three positions that were represented with a "R" and all three positions were polymorphic with multiple isolates haveing A or G at all three positions, including the same two isolates from Gharbiya collected at the same time (and these two sequences, which were identical to each other had 22 differences the Gharbiya HA sequences).
However, although the three Gharbiya chicken sequences fell into two distinct groups (11 NA differences and 22 HA difference) both had picked up a synonymous NA polymorphism, G743A, which had never been seen in any prior H5N1 in Egypt. The acquistion of the same polymorphism at the same place and same time on two very distinct backgrounds (both backgrounds have been indentified earlier, based on NA and HA sequneces), was a strong signal for recombination.
In addition, the multiple mixed signals in the HA and NA sequences raised the possibility that the Tamifu resistance (N294S) was lurking below the detection level, and could be found by cloning and sequencing the clones.
Consquently the sample from the chicken with the mixture that was related to the Gharbiya cluster was plaque purified and 44 NA sequences were generated. The clones showed that the chicken was indeed infected with two major species that were in a 3:1 ratio. The dominant sequence was closely related to the Gharbiya cluster sequence, but none had N294S, and all had G743A. The minor species matched the other two chicken sequences and also didn't have N294S (as expected), but all of the minor species clones also had G743A, demonstrating that the same change had indeed been acquired by two different backgrounds, and not only were in the same place at the same time, but were also in the same chicken.
Moreover, a few of the sequences were recombinants, with 9 or 10 of the 11 positions matching one parental sequence and 1 or 2 matching the other. Thus, the plaque purified clones demonstrated recombination involving the exchange of one or two SNPs.
However, the data also said quite a bit about what was seen in a mixed sample. Even though the two NA sequences had differences at 11 positions, only 1 of the 11 showed up as a mixed signal in the "consensus" sequence. One of the plaque purified clones was still a mixture, which appeared to be a 50/50 mixture, becasue sequencing of that sample gave mixed signals at all 11 positions. However, in the uncloned sample, 10 of 11 positions were "clean" even though 1/4 of the sequences in the sample were the "other" sequence which was not present in the "consensus".
Thus, there was clear recombination in the Gharbya chicken with the Gharbiya cluster sequence, and the vast majority of the mixed positions (10/11) did not give a mixed signal in the "consensus" sequence.
As seen in the first link above, the G743A synonymous polymorphism subsequently appeared on multiple H5N1 backgrounds in Egypt, as well as very different clade 2.2 H5N1 backgrounds in Russia, Kuwait, Ghana, Ivory Coast, all at the same time (ealry 2007).
Recombination involving the acquisition of SNPs is quite real in infleunza and quite common.
