A better understanding of the fitness cost of mutations associated with immune escape can have important implications for understanding viral pathogenesis, transmission and vaccine development
[1–9]. To accurately determine fitness of viruses present in vivo, we developed a new PASS fitness assay with unmodified HIV-1 genomes that were identified in vivo. Using this assay, we have precisely determined the impact of mutations in the cognate T/F virus genome. Although much more effort is required to generate T/F IMCs to study viral fitness than to use existing IMCs from unrelated viruses, we have generated over 20 IMCs and identified a number of CTL cell escape mutations through analysis of longitudinal follow-up sequences
. They can serve as ideal candidates for further study of the impact of immune escape mutations on the fitness of cognate viral genomes. We determined the fitness cost of two CTL escape mutations (R355K in Env and T242N in Gag) in the context of other mutations in the cognate viral genomes (TK and NIA, respectively). Both CTL escape mutations were present together with other mutations that were within or outside the CTL epitopes. Interestingly, both mutant viruses were as fit as the T/F virus. The CTL escape mutation T242N alone had a significant fitness cost as previously shown by others
[14–16]. It has been shown that compensatory mutations, within or outside the T cell epitopes, can partially restore the fitness loss caused by T cell escape mutations
[7, 14, 15, 32, 33]. Here, we demonstrate that fitness loss by CTL escape mutations can be completely restored by the mutations in the same epitope in their cognate virus genome. The TK virus with both the I64T mutation in Tat/Rev and the R355K CTL escape mutations in Env represent a naturally selected virus in vivo. Although the TK virus had two mutations, it was as fit as the T/F virus. No fitness loss of the TK virus also explained why it outcompeted the T/F in vivo and dominated the viral population under the CTL selection pressure only 14 days after the date when the T/F virus was inferred. More importantly, this represents another case, like the NIA virus, demonstrating that the overall viral fitness is not impaired by the presence of both a CTL escape mutation and an additional mutation in the T/F viral genome.
These observations may have important implications for understanding pathogenesis and vaccine development. Many mutations that were not associated with immune responses are often identified in the viral genome and their biological functions have not been well elucidated
. Results from other studies and ours indicate that such mutations, especially those reversion mutations, may play an important role in compensating fitness loss caused by the CTL escape mutations, so the virus can survive the unfavorable environment in the presence of immune selection pressures. Alternatively, they may represent adaptations that favor virus replication in the host target cell environment.
In two comparison pairs (T/F verses T242N and NIA verses T242T), no difference in replication rates were observed for the compared viruses in the single passage assay even though the viruses increased exponentially in the cell culture medium. However, the fitness loss by the T242N mutation was clearly demonstrated when the cell free viruses were passaged multiple times as shown in this study and a previous report
. This may be explained by the following reasons. First, the T242N mutation affects the viral replication through interacting with cyclophilin A
. Since the viral entry was not impaired by the T242N mutation, the T242N was marginally less fit than T/F in the single passage assay. The p24 concentration of T242N was only less than two fold lower than that of the wt or T/F virus at the end of the culture when virus replication plateaued as shown in a previous study
 and by our result (Figure
2B). Second, the infectious viruses generated during the first 2 days of culture were far fewer than those in the inoculum (90,000 infectious units (IU)/ml versus 200 IU/ml). Thus, it is unlikely that the majority of the new infections during the four-day culture were initiated by the cell-free viruses newly generated in the single passage culture. Third, cell-to-cell infection was about 100-1000 time more efficient than the cell-free virus in culture
. Thus, the subsequent infection is most likely established through the cell-to-cell infection in the single passage culture. If the efficiency of the cell-to-cell infection was equal for the compared viruses with the identical env gene, the proportion of each virus might not change over time in the culture. As a result, the relative fitness determined in the single passage culture would be similar between the compared viruses. However, by passaging the cell-free viruses to the fresh CD4+ T cells, the cell-free viruses would compete with each other multiple times. Since the more fit virus accounted for a larger portion of the viral population at each passage, they would infect more cells in each subsequent passage as shown in this study and a previous report
. Importantly, the dynamics of proportion changes for each compared virus between the inoculum and the first passage was consistent with the viral replication slope during the subsequent passage(s), suggesting that the initial changes in viral proportion from the inoculum to the first passage represented intrinsic viral fitness differences and could be further confirmed by subsequence passages (Figure
5). Since virus growth was not in exponential stage throughout passaging, previous models that assume constant exponential growth could not be used to determine relative fitness between compared viruses. Thus, we developed a new mathematical model that does not assume constant exponential growth to measure relative fitness of HIV-1 strains after multiple rounds of passages. Taken together, results from this and other studies indicate that multiple passages are needed to better determine the fitness difference between compared viruses if the proportions of viruses change obviously between the inoculum and the first passage, but no difference in replicative slope is observed during the first passage (Figure
3C). However, if the proportions of viruses do not change from the inoculum to the culture throughout the first passage, this indicates that the compared viruses are similarly fit and additional passages may not be necessary (Figure
3D). When the difference in replication slope is clearly demonstrated over time during the first passage as shown in our previous study
, the relative fitness can be determined without further passaging.
Viral fitness is generally determined by either directly sequencing the bulk PCR products or by detecting markers that are incorporated at different locations of the viral genome
[16, 18, 19, 26, 27]. Thus, the impact of recombination on viral fitness analysis during culture has not been fully elucidated. Since the high frequency recombination has been observed in culture
[24, 25], it is likely that recombination has a significant impact on viral fitness assays, although recombination was considered low in a previous study
. The PASS assay can detect multiple sites in the same viral genomes, and the linkage analysis of these sites allows detection of recombination among a limited number of mutations within a small region (≤ 2 kb) in the viral genome. The emergence and gradual increase of the recombinants to as high as 35.4% of the viral population in vitro suggest recombinant viruses are more fit than both parental viruses. Interestingly, all three recombinants (V247I, I64T and R355K) detected in vitro in this study were also identified in vivo (Figure
1C), suggesting that those recombinant viruses are naturally present in HIV-1-infected individuals. The V247I virus was detected as the predominant virus before the CTL escape mutation T242N was detected and then fixed together with the T242N mutation in the viral population. Since V247I was a reversion mutation, the emergence and domination of this virus suggest that it is also more fit than the T/F virus in vivo although this still needs to be experimentally confirmed. The accumulation of more fit viruses with V247I mutation can also readily compensate the fitness cost of the T242N mutation when it is selected later. Two other recombinants (viruses with either I64T or R355K mutation) were only detected in vivo at low frequencies shortly after infection and then quickly replaced by the TK virus (Figure
1C), suggesting they are less fit in vivo under selection pressure. This suggests that although both recombinants are more fit than the T/F and TK viruses in vitro, the TK virus, under the immune selection pressure, has a higher replication advantage over both of the recombinants in vivo.
Although it was possible that the one base difference could be caused by mutation, the odds for one mutation to occur at a particular site in different experiments at various time points was much smaller than recombination between two existing parental viruses. The point mutation rate for HIV has been estimated as ~ 2.2-5.4 × 10-5 per base per replication in vitro[36, 37]. Because we were interested in mutation to a specific nucleotide, we divided the point mutation rate by 3 (0.73 -1.8 × 10-5), and because either of the two possible bases could mutate, the probability of the observed sequence occurring by mutation was 1.4 - 3.6 × 10-5. In contrast, recent estimates of the recombination rate were 1-1.4 × 10-5 per site per generation
[38, 39]. Thus if the recombination target is 13 nucleotides between T/F and NIA, the expected recombination rate is 1.8 × 10-4 per generation. Thus, it is about 5-12 times more likely than a point mutation to a specific nucleotide at one or the other site. In the case of the comparison between T/F and TK, the recombination target is 1258 base long and thus the recombination probability is about 1.8 × 10-2 per generation. This is 500-1200 times more likely to be due to the recombination than the point mutation.
The mutations detected in the recombinant genomes were the same as those selected by CTL or reversion mutations in vivo. There is a possibility that they represented reversion mutations during the in vitro culture. However, analysis of 51 SGA sequences did not show the reversion mutation (I64T) after the T/F virus was passaged six times. In addition, analysis of over 2000 viral genomes from the co-culture of the NIA and T242N viruses (both with the T242N mutation) at passage 4 did not show the wild type base at position 242. This is in good agreement with the in vivo observations, in which the T242N mutation reverted back to the wild type base after months of infection
[8, 32, 40–42]. These results demonstrate that mutations detected in the recombinant genomes were the results of the recombination between parental viruses, not the results of reversion or random mutations during the time period of the assay. HIV-1 is known to be highly recombinogenic
[24, 25, 38, 39, 43]. Results from this study indicate that the recombination frequently occurs between compared viruses in competition fitness assays and should be considered for accurate estimation of viral fitness. If the different nucleotides are few and within 1.5 kb in the compared viral genomes, the impact of recombination on the fitness analysis can be determined by the PASS fitness assay.
There are several advantages of the new PASS fitness assay. First, no modifications in the viral genome are needed. Second, the mutations of interest are directly measured. Third, fitness can be simultaneously compared for multiple viruses. Fourth, recombinant viruses between targeted mutations can be detected and the influence of recombinant viruses on fitness can be accurately evaluated. Fifth, it is highly sensitive for minority variants present at 0.01% - 0.1% in the population
. Sixth, the detection efficiency is equally efficient for compared viruses because of the identical match between the primers and all compared viral genomes. Finally, the viruses representing those in vivo are used to study viral fitness, and mutations evolved from the T/F virus and identified through genetic and immunological analysis are introduced back to the cognate T/F viral genome. Using this assay, we have also found that a neutralizing antibody (nAb) escape mutant identified six months after infection was less fit than the virus without the nAb escape mutation
. Thus, this method can be a useful tool to precisely measure the impact of the mutations on viral fitness in the cognate viral genome and if viral fitness plays an important role in viral set points, transmission and pathogenesis of HIV-1.