A fitness cost is usually incurred by CD8 T cell escape mutations; this is most clearly demonstrated when EM virus reverts to the fitter WT upon transmission to MHC mismatched hosts 
. Rapid reversion of EM KP9 virus occurred in Mane-A*10
negative macaques infected with our SHIVmn229
viral stock (). This biological isolate is 11.2% WT by qRT-PCR (9.1% by cloning and sequencing 44 clones) since it was derived from an infectious SHIVHXB2
clone originally passaged in Mane-A*10
+ pigtail macaques 
. In our first analysis, we compared reversion to WT of our stock virus to that seen with two different passaged virus innocula. In each case, naïve Mane-A*10
negative pigtail macaques were infected with plasma and PBMC derived from further in vivo
passages of SHIVmn229
containing 3–30 fold lower proportions of WT KP9 virus 
. We also compared the rates of reversion at another Gag epitope, AF9, for which we have also developed a sensitive qRT-PCR for 
Reversion to WT for different viruses and percentages of WT in the inocula.
Outgrowth of WT virus following inoculation with the EM SHIVmn229 stock (containing 89% EM virus and 11% WT virus) is almost identical and very rapid in all Mane-A*10 negative animals studied (, ). WT virus grows most rapidly in the early days of infection while the target cells are not yet depleted, to dominate the viral population by the second week of infection. The K165R EM virus decays to very minor or undetectable level by day 56 of infection in all animals.
Reversion of KP9 and AF9 mutant viruses.
To investigate whether the rapid outgrowth of WT virus in the setting of substantial levels of WT virus in the inoculum could be generalized to other Gag epitopes, we studied rates of reversion at the AF9 epitope. We inoculated 2 naïve Mane-A*17
negative pigtail macaques with a passaged virus containing ~50% WT and 50% EM virus at AF9 
. Very similar rapid outgrowth of WT virus was observed using separate specific qRT-PCRs for WT virus or the 6-bp deletion AF9 mutation (). Again, WT virus grows at a rapid rate over the first week and is the dominant species over EM virus within 2 weeks of inoculation. Thus, where both WT and EM viruses are present in substantial quantities in the virus inoculum, the WT virus very rapidly outgrows the EM virus upon transfer to MHC-mismatched hosts.
Outgrowth of WT virus delayed and slowed with lower levels of WT in inoculum
The rapid outgrowth of WT virus in the previous experiments suggested sufficient quantities of WT virus were present in the inoculum to co-infect the host and then rapidly out-compete the EM virus. We therefore sought to elucidate the impact of much smaller amounts of WT KP9 on reversion at KP9. We first chose to transfer plasma and cells from Mane-A*10
+ animals infected for lengthy periods of time with SHIVmn229
(essentially further in vivo
passages of the original SHIVHXB2
stock). For these virus transfer experiments, we selected donor Mane-A*10
+ animals that were previously vaccinated with SIV Gag-expressing DNA and recombinant Fowlpoxvirus vaccines 
. The donor animals generated KP9-specific CD8 T cell responses after vaccination, which are further boosted after virus challenge. The KP9-specific responses in the donor animals select the EM virus and further reduce levels of WT virus. By qRT-PCR, the passaged viruses chosen had either only 4% WT virus or 0.34% WT virus at KP9 using our qRT-PCR (, and ).
The viral transfer was successful, resulting in an infection of all animals studied. The appearance of WT virus was delayed in recipients of both in vivo passaged SHIVmn229, which contained lower levels of WT virus (). Although high levels of EM virus were detected within 6 days of transfer with the KP9 qRT-PCR, very low levels of WT virus were detected only by day 8–11 after transfer. Further, even after the detection of WT virus, this variant did not expand dramatically to high levels as seen with the original SHIVmn229 stock or the passaged AF9 mutant virus. The WT virus took 63–75 days to exceed EM virus in the case of the 4% WT stock, and WT virus levels never exceeded EM virus levels to 75 days of follow up in the case of the 0.34% WT stock.
The kinetics of reversion to WT were compared by estimating the time taken for 50% WT virus to be reached, using linear interpolation of the log-transformed EM and WT viral loads. In the animals infected with 4% WT virus, it took 63–75 days before WT virus reached 50% of the total virus levels. In the case of the animals infected with the 0.34% WT stock, WT virus levels never exceeded EM virus levels to 75 days of follow up. Thus, lower initial WT level was associated with an increased delay in WT virus outgrowth.
Infection with escape mutant SIVmac239
Our analyses of biologic isolates of X4-tropic SHIVmn229
(for KP9) and R5-tropic SHIVSF162P3
(for the AF9 epitope) strongly suggested that the levels of WT virus in the inoculum have a major bearing on the time needed for outgrowth of the WT virus. However, it is difficult to completely exclude that these uncloned viral stocks contain quasi-species with mutations at other sites or compensatory mutations (although no clear pattern of such mutations were seen during intensive cloning and sequencing). To avoid these potential confounders we constructed a molecular clone of the K165R KP9 EM virus within SIVmac239
. This enabled us to evaluate rates of reversion using a separate R5-tropic virus in a very tightly controlled manner. We chose to infect naïve Mane-A*10
negative pigtail macaques using plasmid DNA, an approach we and others have previously used successfully, both with attenuated and WT viruses 
. Using clonal proviral DNA to initiate the infection eliminates any possibility of generating alternate viral quasispecies in vitro
prior to in vivo
To first determine if infection with a mix of WT and EM SIVmac239
conforms to the same general principles observed with SHIVmn229
we inoculated 2 naïve Mane-A*10
negative pigtail macaques using a 90
10 mix of EM
WT plasmid DNA. Using our qRT-PCR assay to detect WT or EM virus, both WT and EM virus grew readily for the first 2 weeks but WT virus subsequently rapidly outgrew the EM virus (). The EM virus slowly decayed later in infection. This pattern of outgrowth of WT virus using a 90
WT mix of SIV was almost identical to that observed with the original 89
We next evaluated the generation and outgrowth of revertant WT virus following inoculation with pure clonal K165R EM SIVmac239 (). Again we inoculated proviral DNA and an infection was readily initiated in all 3 Mane-A*10 negative animals studied. We observed an 8–10 day delay in the appearance of the WT virus which took ≥63 days to exceed levels of the EM virus. The patterns of growth of WT virus and decay of EM virus following infection with 0% WT SIVmac239 were strikingly similar to those observed with the passaged SHIVmn229 inocula with ≤4% WT virus. A comparison of levels of WT and EM viruses across the 6 strains used is shown in , with the viruses grouped according to whether they have ≥ or <10% WT in the inoculum. The grey shading in highlights the consistent 8–10 day delay in appearance of the WT virus in the low WT virus inocula and rapid outgrowth of WT virus in the high WT virus inocula.
Modelling reversion of WT virus
The experimental analysis above demonstrates that the rate of reversion from WT to EM virus is linked to levels of WT virus in the infecting inoculum. This is not unexpected since, all other factors being equal, a halving of the initial proportion of WT virus would be expected to require one additional doubling time before the WT virus reached 50% of the initial inoculum. However, as illustrated in , the observed effects of reducing the proportion of WT virus are much stronger than expected by this factor alone. In order to understand how the initial WT proportion affects the subsequent dynamics of infection, we modelled the dynamics of WT and EM virus following infection.
Dependence of reversion dynamics on the percentage of WT in the inoculum.
From previous work we expect that target cell number (the number of uninfected CD4+ T cells available for infection) is important to viral growth and the rate of reversion 
. Therefore we expect that early in infection reversion will be extremely rapid, but this will slow towards the peak of infection and in the chronic stage because there are fewer target cells to infect. We analyzed the time taken to reach 50% WT virus using a simple ‘fixed reversion rate’ model (dashed line in ). This model illustrates the increase in time to reach 50% WT that we would expect solely from the decrease of WT fraction using a difference in replicative capacities of WT and EM of 3.5×10−4
. We also analyzed a model that takes into account the effects of decreasing viral growth with reduced target cell number (solid line in ) using Equation 2, for the same difference in replicative capacities of WT and EM. For an initial WT content between 50 and 10%, the time required for WT to reach 50% of total viral load grows relatively slowly and almost linearly, as expected just from the decrease in WT fraction. However, as the proportion of WT decreases further, this time suddenly increases much more rapidly than expected.
One prediction of this model is that not only will it take longer to reach 50% WT when target cell dynamics are taken into account, but that the reversion rate itself at 50% WT will be slower with lower WT fraction in the inoculum. Indeed, the reversion rate observed at 50% WT is significantly correlated to the fraction of WT in the inoculum (, , Spearman correlation, p
0.788). The solid line shows the dependence of reversion rate on initial WT fraction predicted by the model Equation 2.
Analysis of the dynamics of WT and EM virus over time explains this effect (). When high initial WT proportion is present, the WT virus outgrows the EM virus in the early phase of infection, when the pool of target cells for the virus is still nearly complete and both viruses are still in an exponential growth phase (solid red line in ). Reducing WT virus levels slightly have little effect (just delaying the time to 50% slightly, because of the additional time required for the WT virus to grow). However, once the WT proportion gets below a certain level, it will not reach the 50% level before the peak of infection, and before the extensive depletion of CD4+ T cells (solid green line in ). The depletion of target cells slows the growth of both viruses, but importantly also slows the rate at which WT virus overtakes EM virus. If WT virus has not already reached 50% by the time of peak viral load, its rate of progress towards the 50% level is slowed dramatically, and it takes much longer to outgrow EM. The same pattern is observed experimentally, as shown by the examples of animals with initial 11.2% WT () or with 0.34% WT ().
Biological implications of slow reversion to WT
Slow reversion to WT virus implies that the infecting virus population is dominated by the less fit EM virus for longer periods of time in comparison to animals with fast reversion. In theory, this could result in lower viral loads and less pathogenic infections. Goepfert and colleagues recently showed HIV-1 strains transmitted with multiple Gag CD8 T cell escape mutations resulted in overall lower viral loads 
. However, the direct comparison to otherwise similar virus strains is difficult in humans. Since we had a large series of unvaccinated animals infected with the original SHIVmn229
(which reverts rapidly) from previous infection studies 
and now multiple animals infected with a further passage of SHIVmn229
with less WT virus at KP9 (that reverts slowly), we compared the virologic outcome of infection with both viruses. Both viruses grew effectively and exponentially during acute infection (). However, animals infected with the passaged viruses had much lower content of WT at peak and set point viral loads (average 90% WT at peak in SHIVmn229
stock, versus only 5% WT at peak in passaged SHIVmn229
, and 100% vs. 48% on average respectively at the set point). The increased proportion of WT virus at peak viral load in animals infected with SHIVmn229
stock is associated with an increased peak viral load in the animals (): median peak viral load for SHIVmn229
stock-infected animals was 1.24×108
(95%CI from 9.6×107
), and for passaged SHIVmn229
it was 2.08×106
(95%CI from 1.9×106
, Mann-Whitney p
0.009). Set point viral load was also significantly higher in animals infected with stock virus (): for SHIVmn229
stock it was 9.4×105
(95%CI from 7.3×105
), and for the passaged virus it was 2.3×104
(95%CI from 2.0×103
, with Mann-Whitney p
0.0072). Thus, the slow reversion of WT virus in animals infected with a low proportion of WT virus has direct implications for the virological outcome of infection.
Reduced pathogenicity of passaged SHIVmn229.