Evolution of H1N1pdm HA Gene in Taiwan
In this work we analyzed 147 Taiwanese H1N1pdm viruses to portray the evolutionary dynamics of its historical debut. We collected 77 samples for the 2009/2010 season and 70 samples for the 2010/2011 season, and we divided the two seasons by including the two cases in May of 2010 into the first season. Because no data were collected for March and April 2010 and for June and July 2010, the two May cases might just as well have been assigned to the beginning of the second season instead. The overall statistics discussed here, however, would most likely not have been affected either way. We detected 51 HA sites showing changes in the first pdm season in Taiwan. Moving into the second pdm season, only 22 of the 51 HA mutations had maintained such changes. The other 29 HA sites were found to have recovered their amino acids to what was originally observed in A/California/04/2009. Nevertheless, 49 new HA sites showed amino acid substitutions that had not been observed in the first pdm season, bringing the number of HA sites showing amino acid changes in the second season (relative to the original 2009 California strain) to 71.
Pan et al. 
analyzed H1N1pdm strains deposited to GenBank between April 1 and December 31, 2009, reported only one dominant HA mutation S220T and described its location near the receptor-binding domain (RBD). They did not observe the other four major HA mutations that we have reported in . Potdar et al. 
reported an Indian study involving 13 H1N1pdm isolates from May to September 2009. They found the HA sequences dominated by four mutations, namely P100S, T214A, S220T, and I338V. These results are consistent with our findings, except that Potdar et al. did not notice any E391K mutation. Despite approximately the same time-frame between the two studies (May-September 2009 for India, and June-September 2009 for Taiwan), only six mutations were found in common between 16 mutations detected from 13 Indian viruses and 26 mutations from 39 Taiwanese viruses (shown in ). Ilyicheva et al. 
performed a sequence analysis of 23 Russian H1N1pdm strains and reported 20 HA mutation sites. Graham et al. 
performed HA sequence analysis of 235 Canadian H1N1pdm strains, of which 210 were collected from May to December 2009. Taking together the Indian and Taiwanese cases spanning approximately the same period of time in the first H1N1pdm season, these mutations were detected from viruses isolated in four distinct geographical locations, representing an H1N1pdm HA mutation spectrum around the globe. Most of the sporadic mutations did not occur at sites that are common across these four locations, suggesting that a geographical variation did exist in the early diversification of the viral genome.
Various HA mutations in the first H1N1pdm season from four different geographical locations.
All primary HA mutations observed in the first Taiwanese H1N1pdm season remained abundant in the second season. These include P100S, S220T and I338V. Although T214A substitution was also found dominated throughout the two seasons, its appearance was less frequent (~50%) in the end-of-season months December 2010 to February 2011. also notes other dynamic patterns of HA mutations, including the early emerging and re-emerging mutations of types I.b and I.c, and second-season emerging types II.a, IIb and II.c. The possible association between these dynamic patterns and HA evolution, particularly at a stage after the so-called early diversification of this new virus, is an interesting question to address. Many of these mutations were still present with over 50% occurrence at the season-ending month. The late-breaking mutations (T14I, D114N and I233V of type II.c, and R222K, V266L and K300E of type I.c) were seen at approximately the same frequencies of 11–28% in January 2011, which had nearly doubled to 30–45% in February 2011. Although their prevalence was not as obvious as that of other mutations, our data suggest they may re-appear in the subsequent seasons.
As limited Taiwanese H1N1pdm viruses were isolated and investigated after our sampling period, we gathered 8,876 H1N1pdm HA sequences up to April 2012 from National Center for Biotechnology Information (NCBI) and analyzed the dynamics of mentioned HA mutations. As shown in , the temporal appearance of T214A was in general declining yet in an oscillatory manner after March 2011. For example, T214A was only seen in 2 out of 15 HA sequences in February 2012, and none among 5 cases in March 2012. also supports our finding for the two late emerging/re-emerging type II.c mutations D114N and I233V, and two type I.c mutations R222K and V266L. Their dynamic patterns after March 2011 seem to resemble what was observed in T214A. Interestingly, the remaining type I.c mutation K300E and type II.c mutation T14I are hardly seen in these NCBI HA sequences. In contrast to the generally declining T214A, the other two type I.b mutations E391K and S468N and two type II.b mutations S160G and S202T after the second season appear to be stabilizing in the virus population. Continuous monitoring these mutations in more seasons to come is important to better understand the HA evolutionary spectrum of this virus.
HA mutation dynamics of Taiwanese H1N1pdm viruses versus publicly available H1N1pdm viruses.
Evolution of H1N1pdm NA Gene in Taiwan
We mentioned that only 22 out of 51 HA mutations (43.1%) detected in the first season showed up again at least once in the second pdm season in Taiwan. Of all 71 HA mutations observed in the second season, 49 mutations (69.0%) had not appeared at all in the first season. Such diversification for giving up old and acquiring new mutations across seasons was even more noticeable for NA, in which only 25% of the mutations (7 out of 28) from the first season survived in the second season, and 85.4% (41 out of 48) of the mutations found in the second season were newly emerged. A number of the newly emerged second-season mutations were short lived and had disappeared completely in the final months of the season, including the three type II.a mutations M15I, N189S, and I365T. The three type II.b mutations N44S, V241I, and N369K began in October 2010 and appeared to persist until the season’s end, although they never reached 100% peak as did V106I and N248D. Recall that the previously mentioned HA mutations of L8M, S160G, S202T, and S468N were also found to emerge or re-emerge in October 2010. Nevertheless, similar to a number of type I.c and type II.c HA mutations observed in the final two months of the second season (T14I, D114N, R222K, I233V, V266L and K300E), we also found that the two type II.c NA mutations S299A and I374V emerged only in January and February 2011. Such co-incidence in the evolutionary dynamics of HA and NA suggests that a fitness or co-evolution occurs between the two genes, which may play an important role in shaping the viral genome for many seasons to come.
We gathered 6,017 H1N1pdm NA sequences from NCBI to follow up those mentioned NA mutations in Taiwan. describes their temporal dynamics for the entire three H1N1pdm seasons from April 2009 to April 2012. Other than that V106I and N248D are fixed in the population as expected, those three type II.b mutations (N44S, V241I and N369K) that showed higher prevalence in 2010–2011 season seem to appear in a oscillatory manner. Note that a number of 2011–2012 months are missing in because of no NA sequences are available in NCBI. Nevertheless, the sample counts of the final four months in (August/October of 2011, and January/April of 2012) are only three, two, three and one, respectively. More samples are needed in order to better describe NA evolution in terms of amino acid transitioning.
NA mutation dynamics of Taiwanese H1N1pdm viruses versus publicly available H1N1pdm viruses.
Mutation Rates of Taiwanese H1N1pdm Viruses in 2009–2011
A nationwide molecular surveillance of H1N1pdm genomes in Canada 
revealed HA and NA genetic diversity at 1.98×10−3
amino acid substitution per protein site, respectively. This study’s sampling covered a period of only eight months (from the emergence of H1N1pdm to December 5, 2009) and the NA proteins seemed to display slightly more substitutions than did the HA proteins. Our rates of amino acid substitution for HA and NA in the first season (a 12-month period) were 9.29×10−3
, respectively. To make these comparable with Canadian’s study, a sub-sampling from June to November 2009 was chosen and mutation rates were re-computed as 8.90×10−3
for HA and 4.68×10−3
for NA. Not only did our HA proteins display far more diversity than NA, our study also found much higher mutation rates.
Earlier we discussed the difference in HA mutating sites between Taiwanese and Canadian studies over the same sampling period (). We noted that P100S, T214A, S220T, and I338V occurred in almost all 39 Taiwanese H1N1pdm viruses sampled. For the 210 Canadian viruses collected within the same time period, however, only 72 (34.3%) were observed to contain S220T; no P100S, T214A, or I338V changes were found. Two other major mutations, K2E and Q310H, were detected in 55 (26.2%) of the Canadian HA proteins but not at all in Taiwanese strains. However, the overall HA mutation rate for Taiwanese viruses was apparently higher than that of the Canadian ones. Such geographical disparity may explain the differences in mutation rates described here.
Furuse et al. 
compared the evolutionary rates among seasonal H1N1 (1918–1957 and 1977–2009) virus, swine H1 virus, and 2009 H1N1pdm virus; their results indicated that the rate of H1N1pdm was much lower than that of the others. However, the pdm data analyzed had been sampled over a period of less than ten months, up to the end of the first pandemic season. This sampling limitation may have led to an unreliable estimation of the correlation coefficient for describing the evolutionary trend. The Canadian study mentioned above was also based on an eight-month period only, and many genetic variants were still evolving, at least for the first pandemic season. Nevertheless, our second-season evolutionary statistics indicated that the amino acid mutation rates of H1N1pdm HA and NA are elevated than they were in the first H1N1pdm season. For the mutation frequencies shown in , the Taiwanese H1N1 viruses showed 5.24 substitutions per 566-aa HA segment in the first season, increasing to 8.26 (a 57.6% boost) in the second season. NA displayed a lower mutation frequency than HA in the first season, with 2.45 substitutions per 469-aa segment, but soared to 5.16 (a 111% boost) in the second season. Ongoing surveillance data obtained from various geographical locations and subsequent seasons will enable more accurate descriptions of the evolutionary dynamics of this novel H1N1pdm virus.
It is mentioned that the way these clinical samples were collected did not take into consideration the demographic factors such as gender, age or geographical location. Neither did we gather vaccination history from the patients. As a result, these data are not suitable for revealing correlations between these factors and amino acid mutations. A large-scale, island-wide study by Taiwanese CDC 
revealed that the major affected groups were shifted to older individuals of higher age-specific case fatality ratios (CFRs) from May 2009 to April 2011 in Taiwan. They also discussed the possibility that the shift could be attributed to the vaccination program in which adults aged 18–64 were the shortfall in influenza vaccination. How such shifting of CFRs relates to underlying genome variations, however, remains to be investigated.
In summary, we revealed amino acids transitioning of the two surface glycoproteins of H1N1pdm viruses, particularly on how these mutations shifted in 2010/2011 season after the H1N1pdm’s debut in 2009/2010. We found 17.7% of HA and 14.7% of NA sites had their amino acids mutated based on A/California/4/2009. Many of these mutations were transient, demonstrating how the viral genome has been shaped dynamically. Among those mutations that appeared more frequently (>5% incidence in all 147 viruses from June 2009 to February 2011), many were new after August 2010 which were not seen throughout the first pandemic season ( and ). Furthermore, some late-breaking mutations are found to have statistical correlation to disease severity. Although a number of mutations were made to the antigenic sites, HI tests showed no titer changes for these Taiwanese strains. There was only one recent isolate in February 2011 which contains the well-known resistance marker H275Y in NA, suggesting an overall susceptibility for Taiwanese isolates to NAIs.