Re: Homologous Recombination is Very Rare or Absent in Human Influenza A Virus
After the most recent post, it is time for a review.
The paper being discussed on this thread is by a group that is on record as saying there is no recombination in influenza. They tried to back up their earlier comments by using a limited database to find recombinants and parental sequences. By limiting the database to full human sequences generated under an NIAID project, they eliminated many examples of obvious recombination.
Some obvious recombination in swine influenza had been published at Nature Precedings, which the authors acknowledged. The swine examples, especially in PB2 and PA had many examples of long stretches that EXACTLY matched a pair of 1977 swine isolates from Tennessee. The lab that generated the data has said they they never had the 1977 isolates in their lab, yet the sequences were clearly present in the 2003/2004 Canadian swine isolates (as were long stretches of identity with other swine isolates from 1998 and 2002).
In addition to the swine sequences above, there was a series of human flu sequences which also had obvious recombination over long stretches. These 2002 HA sequences from Korea had long stretches of sequences from the early 1990's. The recent paper found no examples of recombination in HA. The obvious examples from Korea were not found because the sequences were not generated under the NIAID sequencing program (although they have been public for many years). Moreover, the parental sequences from the early 1990's also were not generated under the program, and the 1990 sequences were only partial sequences (1000 BP) and had also been public for many years. Thus, the obvious HA recombination was missed because neither the recombinants nor the parental sequencess were in the restricted database analyzed in this paper.
However, the recent paper did find shorter stretches of recombination in five of the eight gene segments, but all were said to be <100 BP. In the abstract the paper acknowledged that the short stretches were statistically significant and for NA, the odds were a billion to 1 that the acquistions (recombinations) were due to chance.
However, the swine paper with the long streches of recombination also looked at the PB1 sequences because the PB1 sequences in the swine were from human flu circulating in the 1990's. That analysis looked at three regions of 120 BP each and showed that the dominant haplotype for each region switched from year to year in a manner more consistant with recombination than sequential changes due to random mutations. In some instances the sequence that was dominant in H3N2 PB1 subsequently became dominant in human H1N2 PB1 (and human H1N2 PB1 is quite different than H3N2 PB1). In other instances, like the HA sequences in Korea, a haplotype "disappeared" from the database, only to reappear a decade later (and although it disappeared from human flu, it was present in swine flu).
These examples explore a very limited number of examples (like the dominant haplotype), so finding sequences that reappear by chance is VERY unlikely. Finding examples by looking at all possible combinations would reduce the significance, but that is not how the examples are found.
Thus, it is easy to see that the sequence from 1993 has a region of identity with 2007 and this region is not limited to a couple of examples, but is found in the dominant haplotype in 1993 (51 examples) as well as 2004 (168 examples), 2005 (67 examples), 2006 (28 examples), 2007 (60 examples).
Thus, the above examples were in the database used in this paper and were in regions >100 BP, which again raises questions about the methodology / analysis used in the paper.
After the most recent post, it is time for a review.
The paper being discussed on this thread is by a group that is on record as saying there is no recombination in influenza. They tried to back up their earlier comments by using a limited database to find recombinants and parental sequences. By limiting the database to full human sequences generated under an NIAID project, they eliminated many examples of obvious recombination.
Some obvious recombination in swine influenza had been published at Nature Precedings, which the authors acknowledged. The swine examples, especially in PB2 and PA had many examples of long stretches that EXACTLY matched a pair of 1977 swine isolates from Tennessee. The lab that generated the data has said they they never had the 1977 isolates in their lab, yet the sequences were clearly present in the 2003/2004 Canadian swine isolates (as were long stretches of identity with other swine isolates from 1998 and 2002).
In addition to the swine sequences above, there was a series of human flu sequences which also had obvious recombination over long stretches. These 2002 HA sequences from Korea had long stretches of sequences from the early 1990's. The recent paper found no examples of recombination in HA. The obvious examples from Korea were not found because the sequences were not generated under the NIAID sequencing program (although they have been public for many years). Moreover, the parental sequences from the early 1990's also were not generated under the program, and the 1990 sequences were only partial sequences (1000 BP) and had also been public for many years. Thus, the obvious HA recombination was missed because neither the recombinants nor the parental sequencess were in the restricted database analyzed in this paper.
However, the recent paper did find shorter stretches of recombination in five of the eight gene segments, but all were said to be <100 BP. In the abstract the paper acknowledged that the short stretches were statistically significant and for NA, the odds were a billion to 1 that the acquistions (recombinations) were due to chance.
However, the swine paper with the long streches of recombination also looked at the PB1 sequences because the PB1 sequences in the swine were from human flu circulating in the 1990's. That analysis looked at three regions of 120 BP each and showed that the dominant haplotype for each region switched from year to year in a manner more consistant with recombination than sequential changes due to random mutations. In some instances the sequence that was dominant in H3N2 PB1 subsequently became dominant in human H1N2 PB1 (and human H1N2 PB1 is quite different than H3N2 PB1). In other instances, like the HA sequences in Korea, a haplotype "disappeared" from the database, only to reappear a decade later (and although it disappeared from human flu, it was present in swine flu).
These examples explore a very limited number of examples (like the dominant haplotype), so finding sequences that reappear by chance is VERY unlikely. Finding examples by looking at all possible combinations would reduce the significance, but that is not how the examples are found.
Thus, it is easy to see that the sequence from 1993 has a region of identity with 2007 and this region is not limited to a couple of examples, but is found in the dominant haplotype in 1993 (51 examples) as well as 2004 (168 examples), 2005 (67 examples), 2006 (28 examples), 2007 (60 examples).
Thus, the above examples were in the database used in this paper and were in regions >100 BP, which again raises questions about the methodology / analysis used in the paper.
Last edited: