We have developed two HIV-1 replication fitness assays that use flow cytometry to detect the number of cells infected by a drug-resistant mutant, compared to the number infected by a WT reference strain in the same culture. The SCA utilizes a vector in which env
is deleted and viruses are pseudotyped with MLV envelope. Therefore, viruses undergo a single round of infection with no subsequent rounds. Mutant virus produced after transfection is normalized for p24 capsid protein against WT, and cells are subsequently infected with equal amounts of WT and mutant viruses. The SCA vector has the mouse Thy
gene cloned in place of nef
, resulting in the expression of Thy protein on the surfaces of infected cells. In contrast, the MCA, which we have described previously (21
), utilizes Thy-expressing virus with an intact envelope; viruses therefore undergo several rounds of replication during the course of the experiment.
We used these flow cytometry-based MCAs and SCAs to test a panel of 22 drug-resistant mutants. The most striking finding of this study was that four protease inhibitor-resistant mutants had marked differences in relative fitness by the two assays (Fig. ). The replication deficit of the D30N mutant was moderate (20% reduction relative to WT) in the SCA and more pronounced (68% reduction) in the MCA (Table ). Similarly, the I54L mutant had a fitness similar to that of the wild type in the SCA but a significant reduction in the MCA (Table ). In contrast, the I54M and V82A mutants had greater decreases in fitness in the SCA, 66% and 56%, respectively, and were similar to the WT in the MCA (Fig. ). The G48V and I50V mutants also had significant differences in their fitness values between the two assays; however, their relative fitness compared to that of the wild type in both assays was very poor (Table ).
We hypothesize that the SCA, in which the lack of an envelope gene results in noninfectious progeny, should detect abnormalities in virus entry, reverse transcription, integration, and protein expression but would be insensitive to changes in the virus life cycle after early gene expression, such as the production of infectious virions from infected cells (i.e., burst size), the efficiency of cell-to-cell transmission, or the life span of infected cells. The MCA should reflect all these steps in the virus life cycle and may also reflect compensatory interactions between events that require multiple rounds of infection and those that can be measured by both assays (e.g., increases in cell-to-cell transmission could completely or partially compensate for reductions in protease activity). We believe that discordant results for the protease mutants provide important mechanistic clues as to which specific steps in the HIV replication cycle are affected by the mutation. The protease activity of the D30N, I54L, I54M, and V82A mutants correlates with the relative fitness measured in our SCA, indicating that infectivity is one step of the viral life cycle measured by the SCA (1
). The fitness defects of the D30N and I54L mutants were even more pronounced in the MCA, which could be due to defects in both maturation and steps in virus spread that occur only in the MCA. One possible explanation as to why the I54M and V82A mutants are more fit in the MCA is that defects in particle production and maturation, which would be detected in both assays, are compensated for by improvements in steps of viral replication that are measured only in the MCA. This would result in an overall improvement in fitness in the MCA for the V82A mutant.
To determine whether burst size might explain the fitness differences seen for the protease mutants, two duplication mutants of the PTAP motif of p6 were made. This motif has been shown to help the release of particles from the cell by sequestering TSG101 (22
). The duplications used in this study have been observed in drug-experienced patients and in cytotoxic T-lymphocyte (CTL) escape variants (9
). It is hypothesized that these duplications increase the virus's ability to release particles from the cell surface. In our assays, the fitness of these mutants was not significantly different between the two assays. Therefore, either these mutants do not increase particle production, or burst size is measured in both assays and does not influence only the MCA fitness value.
The SCA has two technical differences from the MCA that could cause the potential variability that was observed in this study. The first is the need to cotransfect producer cell lines with the vector and envelope. Differences in the abilities of the two clones to be comparatively expressed from one transfection to another can introduce variability in the amount of envelope protein that is incorporated into the virion. Even in the face of equal expression, there is also the potential for variability in the amount of envelope protein that is functionally incorporated into virions. A second difference is that the RC is determined by a measurement at a single time point. Since relative fitness is not measured as a change over time, as in the MCA, the assay is much more sensitive to the relative amounts of WT and mutant viruses that are added. These two differences potentially create more variability in the RC values of the SCA than in the MCA values and result in the need to average the results for several different stocks.
For the panel of mutants we tested, some previously published studies gave results similar to ours and some gave dissimilar results. There is no previously published information comparing the fitness of HIV-1 drug-resistant mutants in multiple- versus single-cycle assays. In one previous report, the fitness deficit of the D30N mutant was 40% (40
) compared to the WT in an SCA and was reduced 60% in an independent MCA (36
). Like our studies, this MCA study also showed that the fitness of the L90M mutant was similar to that of the WT and greater than that of the D30N mutant. Our results for the I50V mutant are also consistent with another report in which the fitness of this mutant was reduced 90% relative to that of the wild type as measured in an SCA in combination with the L10F and M46I mutants (44
). Our results are also consistent with our previously published results for NNRTI drug-resistant mutants. The fitness of the K103N mutant is similar to that of the WT, whereas the fitness levels of the V106A, G190S, and P226L mutants are reduced (4
In contrast to our study, the replication deficit of the V82A mutant in an NL4-3 background has been reported to be similar to the WT in an SCA, only 20% reduced, as measured by a chlorophenol red-β-d
-galactopyranoside (CPRG)/β-galactoside assay (35
). However, this study used a different method to measure the infectivity of the mutants and did not compare the WT and mutant in the same culture. Competitions where WT and mutant viruses are in the same culture have been shown to be more sensitive to differences (12
). Therefore, differences in sensitivity may explain why our SCA may have detected a greater reduction in fitness for the V82A mutant. In support of our results, the V82A mutant had reduced fitness in the context of a clinical isolate sequence in the single-cycle Monogram RC assay, again indicating that the sequence backbone may influence the results (7
). The results of the current study and previous work further support the idea that the method used to measure fitness may influence the relative fitness of some mutants. We suspect that these methodological differences can account for at least some of the contradictory fitness values for drug-resistant mutants that have been reported in some previously published studies.
Cong et al. have tested the relative fitness, which was different from our results, of several RT mutants using an MCA (13
). There were three main differences between the two assays used: the vector backbone (NL4-3 versus HXB2), the cell type (PM1 versus MT4 cells), and the method used to detect the proportion of viruses (flow cytometry versus sequencing). We have shown previously that the flow cytometry used in our study and the sequencing used in the Cong study give similar relative ratios of mutant to WT virus (21
), and the studies described here show that the relative fitness in MT4 and PM1 cells is the same. The only explanation we could find for the discordant results is differences in sequence backbones. We and others have previously shown that the background sequence in which a drug-resistant mutant develops can influence its relative fitness, prevalence in patients, and level of drug resistance (20
There are three differences between the SCA and MCA that may or may not explain the differences in fitness seen for the protease mutants. These include the use of MLV envelope to pseudotype the SCA virus stocks, a different transfection reagent, and differences in the length of time the virus particles are in the culture supernatant before they find a target cell. However, we believe these differences are unlikely explanations for the differences seen between the SCA and MCA, since they cannot explain the fact that some mutants had higher relative fitness in the SCA than in the MCA while other mutants had higher relative fitness in the MCA. Therefore, the difference would have to be dependent on the specific mutant being tested. In addition, even though differences in the envelope could impact binding and entry, there is no evidence that protease activity is needed for binding and entry, and the transfecting reagents are present during transfection only and are not present during infection.
The fitness difference between the multiple-cycle and single-cycle assays for the V82A, I54M, I54L, and D30N mutants ranges from 39% to 48%. Is this magnitude of difference biologically relevant? Several studies have looked at the correlation of fitness with clinical measures such as viral load and CD4 count. A study of chronically infected subjects using a whole-virus MCA showed that a 20% decrease in fitness correlates with at least a 0.5 log10
decrease in viral load (8
). Therefore, the mutants identified here, which have about a 40% different in fitness, could impact the viral load by 1 log10
. Another study, with acutely infected subjects, showed that subjects with an RC value of <0.42 measured using an SCA had an average CD4 cell count that was >100 cells/μl higher than that of subjects with an RC value of >0.42 (6
). Therefore, mutants such as the D30N and I54M mutants, which cross the 0.42 threshold, could have an impact on CD4 cell counts. Another study showed that a cutoff of 65% for RC predicted better response, such that subjects with an RC of <65% had a better outcome than subjects with an RC of >65% (14
). Again, the difference in fitness for our mutants crosses this threshold. Other studies, with elite controllers, whose viral load is never >2,000 copies/ml without therapy, show that their viruses have 20% lower fitness than those of progressors (37
). Our group has also shown that NNRTI-resistant mutants, which have a 30 to 70% decrease in fitness compared to highly prevalent mutants, are less prevalent in vivo
). These studies serve as examples that the magnitude of the difference we see between the SCA and the MCA for protease mutants could be clinically important.
More studies are warranted to determine if fitness as measured by an SCA or an MCA correlates better with clinical outcome. Until that is determined, studies of fitness, particularly for protease mutants, warrant the use of both types of assays. Currently, the Phenosense assay from Monogram Biosciences, a drug resistance assay, is the only assay that also provides clinicians with an RC value measured using an SCA. Our work indicates that patient samples containing protease mutants that are tested by this assay may not have an accurate fitness measurement. Therefore, clinicians should use this value cautiously. The magnitude of the difference measured between the SCA and the MCA is large enough to have a biological impact in vivo. Since this work is the most comprehensive comparison of drug-resistant mutants in both assays to date, more studies in different clinical settings are needed to determine its clinical importance.
Our results show that the type of fitness assay that is chosen to analyze the relative fitness of drug-resistant mutants may influence the result. Direct comparison of relative fitness using a multiple-cycle and a single-cycle assay may be a way to determine whether a particular mutant is likely to affect early or late steps in the life cycle. Studies are under way to determine if the early steps of the virus life cycle affect fitness using both the SCA and MCA and if late steps in the life cycle only affect fitness using the MCA.