Human influenza A and B viruses (A and B viruses) are equally prevalent among human beings, yet most A virus genes evolve two to three times faster than the corresponding genes in B viruses (11
). The high evolutionary rate of A viruses has been attributed to positive selection by the human immune system (2
), and the difference between A and B to an absence of such selection in B viruses (2
). However, a possible alternative explanation of the difference in evolutionary rates, a difference in mutation rates at the molecular level, has not been investigated (27
). To correct this lack, we have compared the mutation rates of the nonstructural (NS) genes of several circulating A and B viruses during replication in Madin-Darby canine kidney (MDCK) cells, which are unaffected by the human immune system.
Previously, Parvin et al. reported a mutation rate for the NS gene of influenza A virus A/WSN/33(H1N1) (WSN) of 1.5 × 10−5
mutations per nucleotide per infectious cycle during the growth of a single plaque in MDCK cells (23
). In their study, Parvin et al. demonstrated that most of the NS mutants they detected replicated with fitness similar to that of the parent virus. This suggests that the MDCK cells did not encourage or discourage the appearance of these mutations, and hence are a suitable medium for the study of mutation rates. We have adopted their method (23
), with some modifications.
Since the 1991-1992 and 1992-1993 influenza seasons, the receptor-binding specificity of the A viruses, both H1N1 and H3N2, has changed (19
). In particular, the binding affinity of H3N2 viruses to MDCK cells has become weaker (22
). To avoid this effect, we have used viruses isolated before this change occurred. All viruses used in this study, A/Aichi/1/87(H1N1) (A/1/87), A/Aichi/12/92(H3N2) (A/12/92), B/Aichi/29/99 (B/29/99), and B/Aichi/44/01 (B/44/01), were directly isolated from clinical samples in MDCK cells.
The mutation rates for the NS genes of A (ANS) and B (BNS) viruses were measured using the procedure illustrated in Fig. . In brief, well-isolated parent plaques were picked when they had reached a diameter of approximately 2 mm at 48 and 71 h postinfection for A and B viruses, respectively. The parent plaques were used to produce a series of descendant plaques grown to similar diameters.
FIG. 1. Measuring the mutation rate during the growth of a single plaque. Plaque formation was performed in confluent monolayers of MDCK cells infected with A and B viruses. A well-isolated parent plaque of approximately 2 mm in diameter was picked at an appropriate (more ...)
The nucleotide sequences of the NS genes in each plaque were determined directly using the reverse transcription-PCR (RT-PCR) technique (Table ) (21
). Mutations were identified by comparing the sequence in randomly selected descendant plaques to the sequence in the corresponding parent plaque using sequence analysis software, GENETYX-MAC ver.11.0 (Genetyx Corporation, Japan). The direct sequencing of RT-PCR products produces an average sequence of many independent copies of the viral cDNA. Although some molecules will be miscopied during the PCR process, they represent a small fraction of the total and do not affect the average sequence (10
). All mutations (Table ) were confirmed by sequencing with the different sense primers listed in Table a total of at least three times.
Analysis of mutation rate for NS genes of human influenza viruses
Mutations observed in NS genes of descendant plaque viruses compared to the parent plaque
Most experiments counting the frequency of mutants in viral plaques find a high variability (16
). Most of this variability arises from replication errors occurring early in the growth of the plaque. Such random events, when they occur, are amplified to produce large clone lines of identical mutants (Fig. ), greatly increasing the variance of the number of mutants detected (18
). With direct sequencing, it is possible to distinguish each replication error. Counts of the number of unique mutant sequences are related to the number of mutation events, as opposed to the number of progeny resulting from them. Such counts vary only as expected from Poisson statistics. Although no mutant clone lines were detected among the samples taken during the main experiment, in a preliminary test with a B/44/01 parent plaque grown for 96 h, a sample of 169 descendants revealed three mutants. Of these, two exhibited the same mutant sequence, indicating they were very likely clones descended from the same mutation event. By contrast, an A/12/92 parent plaque grown for 72 h contained three different mutant sequences in a sample of 162 descendants.
The results from the main sequencing experiment are listed in Tables and . A total of six experiments were performed, three with B viruses (Table , experiments 1 to 3) and three with A viruses (Table , experiments 4 to 6). For B viruses, a total of two mutations were detected among 461,304 nucleotides, giving an observed mutation frequency for BNS of (2 ± √2)/461,304 = 0.4 × 10−5 ± 0.3 × 10−5 mutations/site.
For A viruses, the first two experiments with primary plaques formed at 34°C (Table , experiments 4 and 5) indicated a mutation frequency far lower than that reported by Parvin et al. (23
), who grew their parent plaque at 37°C. To examine the possible effect of temperature on mutation frequency, we remeasured A/1/87 at 37°C. The number of mutations detected was identical to that at 34°C. Taking all three A virus experiments together, the average mutation frequency for
ANS in the parent plaques was (5 ± √5)/352,359 = 1.4 × 10−5
± 0.6 × 10−5
The higher mutation frequency of WSN(H1N1) compared to A/1/87(H1N1) may reflect the peculiar character of WSN, a neurotropic virus, produced by repeated pulmonary and intracerebral infection of mice (9
). There may also have been some problem with the technology available in 1986. In the absence of PCR and simplified sequence technology, Parvin et al. (23
) were unable to repeat the sequence analysis to confirm mutations by reading both strands of the cDNA. For this reason, their reported number of mutations might not be accurate.
To compare the mutation rate per infectious cycle of the A and B viruses, we determined the generation time of each virus in MDCK cells. As shown in Fig. , the mean generation time, when the PFU reach 50% of the final value (1
), was about 7 and 10 hours for A and B viruses, respectively. As the dynamics of plaque growth are complicated and the budding of progeny viruses could not be synchronized, these generation times were adopted as probable values during plaque formation. Since the 2-mm-diameter parent plaques were picked at 48 h (A viruses) and 71 h (B viruses) postinfection, they correspond to approximately seven infectious cycles in both cases. Using these values, average mutation rates for ANS and BNS were calculated as 1.4 × 10−5
mutations/site/7 infectious cycles = 2.0 × 10−6
mutations/site/infectious cycle, and 0.4 × 10−5
mutations/site/7 infectious cycles = 0.6 × 10−6
mutations/site/infectious cycle, respectively.
FIG. 2. Comparison of the generation times of A and B viruses in MDCK cells. Confluent monolayers of MDCK cells (5 × 105 cells) were infected with A/Aichi/1/87 and B/Aichi/29/99 at a multiplicity of infection of 8, kept at room temperature for 1 h, washed (more ...)
It should be noted that some authors interpret “infectious” and “replication” cycle differently (5
). The observed difference in mutation rates is (BNS − ANS) = −(1.4
mutations/site/infectious cycle, giving a z
−1.4 and P
= 0.08. This difference in mutation rate presumably reflects a more accurate polymerase in the B virus than in the A virus.
Finally, we attempted to estimate evolutionary rates from the observed in vitro mutation rates in MDCK cells. Since the parent plaques for the A and B viruses were grown for 48 and 71 h, and since a year is 8,760 h long, if the viruses propagated continuously all through the year, we estimate evolutionary rates for the A and B viruses of 1.4 × 10−5
× 8,760 h = 2.6 × 10−3
mutations/site/year and 0.4 × 10−5
mutations/site/71 h × 8,760 h = 0.5 × 10−3
mutations/site/year, respectively. Evolutionary rates for the NS genes of human influenza A and B viruses have been reported in the range of 1.8 × 10−3
to 2.2 × 10−3
and 0.45 × 10−3
to 0.8 × 10−3
nucleotide substitutions/site/year, respectively, on the basis of nucleotide sequencing of circulating viruses (13
). Although the generation time of viruses and the selection factors on NS gene products in vivo may not be the same as those in MDCK cells, our estimated evolutionary rates are in good agreement with the reported actual rates. Whether this agreement is a consequence of the random fixation of neutral mutations (14
), of some other origin, or purely an accident remains to be solved.
The current study suggests that one of the reasons for the lower evolutionary rate of influenza B viruses could be a lower mutation rate.