|Home | About | Journals | Submit | Contact Us | Français|
The relative fitness of a variant, according to population genetics theory, is that variant's relative contribution to successive generations. Most drug-resistant human immunodeficiency virus type 1 (HIV-1) variants have reduced replication fitness, but at least some of these deficits can be compensated for by the accumulation of second-site mutations. HIV-1 replication fitness also appears to influence the likelihood of a drug-resistant mutant emerging during treatment failure and is postulated to influence clinical outcomes. A variety of assays are available to measure HIV-1 replication fitness in cell culture; however, there is no agreement regarding which assays best correlate with clinical outcomes. A major limitation is that there is no high-throughput assay that incorporates an internal reference strain as a control and utilizes intact virus isolates. Some retrospective studies have demonstrated statistically significant correlations between HIV-1 replication fitness and clinical outcomes in some patient populations. However, different studies disagree as to which clinical outcomes are most closely associated with fitness. This may be in part due to assay design, sample size limitations, and differences in patient populations. In addition, the strength of the correlations between fitness and clinical outcomes is modest, suggesting that, at present, it would be difficult to utilize these assays for clinical management.
Human immunodeficiency virus type 1 (HIV-1) is a retrovirus that infects primarily CD4-expressing T cells. Primary HIV-1 infection occurs with a single variant initially. However, the rapid replication and turnover of HIV-1 (79, 177) and the high mutation rate of the viral reverse transcriptase (110, 147) result in high mutation frequencies that quickly lead to the production of a population of genetically distinct but related variants called a quasispecies (53). If the effective size of the replicating HIV-1 population is large, the variant that predominates in an HIV-1 quasispecies at any one point in time should be the variant that is most fit under the selective pressures that exist (31).
According to population genetics, fitness is defined as a variant's ability to contribute to successive generations (reviewed in reference 54). By this definition, HIV-1 variants with high levels of fitness should have a selective advantage over other less-fit variants in clinical infections. It has also been postulated that HIV-1 variants with reduced fitness may be less pathogenic, leading to improved clinical outcomes. HIV-1 fitness has been studied primarily using in vitro cell culture model systems in which the rate of replication is measured, and it is not known how closely this measure of fitness correlates with viral fitness in patients or clinical outcomes. For the sake of simplicity, we will refer to all such assays as replication fitness assays, recognizing that these assays are varied in their design and do not fully reflect the selective forces impacting viral fitness during clinical HIV-1 infection.
The degree to which HIV-1 replication fitness correlates with clinical outcome continues to be a controversial issue. We believe that this controversy stems in part from a lack of consensus on what the best approaches are to measure fitness and which clinical outcomes are most impacted by fitness. In addition, the labor-intensive nature of most assays used to quantify fitness limits the sample sizes of studies correlating fitness with clinical outcomes. This review will summarize and put into perspective the types of assays used to measure HIV-1 replication fitness, the effects of specific drug resistance mutations on fitness, and the studies done to date evaluating the correlation between in vitro measures of HIV-1 replication fitness and clinical outcome.
The published literature contains data on a multitude of cell culture assays that have been used to measure HIV-1 replication fitness. This variety makes it difficult to comprehensively list the different approaches to measuring fitness and limits the comparisons that can be made between studies from different groups of investigators. One common feature of all assays is that they compare the replication of a test variant to that of a reference strain. The test variant is either a site-directed mutant of a laboratory strain of HIV-1 or an isolate derived from a patient sample, usually peripheral blood (Fig. (Fig.1a1a).
We have proposed further categorizing cell culture fitness assays on the basis of five major features: whether replication of the test strains and replication of the reference strains are compared in parallel infections or directly in a single culture (the latter is referred to as a growth competition assay); whether replication is measured over a single virus life cycle, using pseudotyped virus, or over multiple cycles; whether the test virus is an isolate obtained from a clinical specimen (“whole virus”) versus a recombinant virus containing only a portion of the clinical viral sequence; whether virus is detected directly by assaying a viral gene or protein or indirectly through the use of a reporter gene; and whether the assay utilizes cell lines versus primary human cells (Table (Table1).1). These features of fitness assays are summarized in more detail below.
Differences between growth competition assays and parallel infections are illustrated in Fig. Fig.1b.1b. Growth competition assays are generally preferable to parallel infections for measuring replication fitness since the replication of the test strain and replication of the reference strain are compared in the same culture, eliminating potential artifacts resulting from differences in culture conditions of the test and reference strains. As the total virus population expands, the prevalence of the test variant in a growth competition assay should decrease over time relative to the reference strain if the test variant has lower fitness. If the relative proportions of the test and reference strains are compared at more than one time point after infection, growth competition assays are also relatively insensitive to the specific method used to determine virus inoculum, which is another significant advantage over parallel infections (more detail on methods to quantify virus inoculum is provided below). A third advantage of growth competition assays is that they can detect small differences in replication fitness that are not identified in parallel infections (33, 137).
An important potential disadvantage of growth competition assays is that a quantitative assay that can distinguish the test and reference strains must be available (Fig. (Fig.1b).1b). Quantitation of the relative amounts of test and reference strains usually adds significantly to the complexity and cost of the assay. In addition, the question of viral recombination must be considered, since the production of recombinant progeny viruses could alter the apparent replication fitness of the test strain if it differs genetically from the reference strain at more than one nucleotide position. In order to minimize recombination between the test and reference strains in a growth competition assay, the experimental design should minimize dual infection of cells, which is a prerequisite for retroviral recombination (80), by using a low multiplicity of infection initially and limiting the duration of infection.
Single-cycle assays are typically conducted by deleting the envelope gene from an HIV-1 vector and then producing pseudotyped virus by transfecting an appropriate cell line with the env-deleted HIV vector together with a plasmid expressing a gene product that can serve as an envelope (see Fig. Fig.22 for a depiction of the Monogram Biosciences replication capacity [RC] assay, which is a single-cycle assay utilizing murine leukemia virus [MLV] env-pseudotyped virus). The MLV envelope and vesicular stomatitis virus G protein have each been used to pseudotype HIV-1 variants in single-cycle replication fitness assays (15, 184). Such pseudotyped virions can infect susceptible cells but cannot produce infectious progeny due to the fact that the genome still carries the env deletion. Infection with pseudotyped virus is thus limited to a single round of replication. Single-cycle infections have the advantage of a shorter time frame, typically 24 to 72 h, compared to several days to weeks for a multiple-cycle assay. Multiple-cycle assays have the theoretical advantage of greater sensitivity because differences between the two variants can be amplified over many life cycles, although the two types of assays have not been extensively compared.
Whole-virus assays are performed using intact isolates cultured directly from patient samples. In contrast, recombinant-virus assays require the amplification of a region of interest by PCR, for example, protease and reverse transcriptase, and the subsequent cloning of that region into an HIV-1 vector encoding the remaining viral genome from a laboratory strain (Fig. (Fig.1a).1a). The recombinant virus may be derived from a single clone of the PCR amplicon, or a pool of recombinant viruses can be produced from the bulk cloning of all amplicons and transfection of the pooled viral vectors.
The major advantage of the recombinant-virus assay is that it does not require the isolation of infectious HIV-1 from the patient, which adds significantly to the time and cost of the assay. In addition, viral vectors used in the recombinant-virus assay can be modified to express a reporter gene, such as luciferase, which can be used to detect viral infection. Recombinant assays allow one to look at the effects of a particular gene segment on fitness but have the disadvantage of not taking into account the possible modulating effects of other gene segments located outside the region of amplification.
Whole-virus assays have the advantage of looking at the replication of the complete viral population present in peripheral blood and thus are likely to be a more accurate measure of viral fitness in a patient. However, whole-virus assays are more difficult in that they require target cells that can propagate all variant types contained within the isolate, usually primary human peripheral blood mononuclear cells (PBMCs); in addition, the methods to detect virus growth are limited to those that directly assay viral genes or gene products.
Studies have compared whole-virus and recombinant-virus assays in relatively small numbers of clinical samples in order to determine the relative impact of different gene segments on overall replication. Using virus isolates without drug resistance mutations in pol, overall fitness in a whole-virus assay correlated most closely with that of a recombinant-virus assay in which the env gene was amplified from the patient specimen (142). In contrast, in one isolate with drug resistance mutations, the pol gene appeared to make a major contribution to viral fitness (142). In addition, studies have demonstrated interactions among gag, protease, and reverse transcriptase that affect viral fitness (19, 156). Thus, the correlation between the two assay types will likely vary and will depend on the gene segment included in the recombinant-virus assay, whether the patient population is treatment-naive or -experienced, and, if treatment failure is occurring, the duration of virologic failure.
Fitness assays can also vary in the methodologies used to detect and quantify the test and reference strains (Fig. (Fig.1b).1b). The major distinguishing feature is whether an assay measures a viral gene or gene product directly or whether a reporter gene is used as a surrogate measure of viral replication. Some examples of measuring virus directly include quantitation of a viral protein, such as p24, by enzyme-linked immunosorbent assay (106, 156, 163); measurement of reverse transcriptase activity (164, 175); and quantitation of proviral DNA (116). In addition, in growth competition assays, the relative amounts of test and reference strains can be quantified by sequence analysis of proviral DNA or viral RNA (65, 69, 78, 86, 117, 154, 155, 169, 174), heteroduplex tracking assays (HTAs) (140), or allele-specific real-time PCR (33, 142, 176). It should be noted that the latter methods do not provide information on the extent of viral replication.
The most common method to directly quantify virus growth is p24 antigen concentration in the culture supernatant. This approach has been used to quantify the relative replication rates of different variants in parallel infections (27, 163). p24 antigen concentration can also be used to evaluate the relative expansion of virus during a growth competition experiment in which the relative proportions of test and reference strains are measured by a PCR-based method that does not directly measure virus replication (174).
A number of different methods have been used to quantify the relative proportion of test and reference variants in a growth competition assay. HTAs of the env gene have been used in whole-virus growth competition assays (7, 116, 140, 164). Hybridization of a PCR-amplified product to a labeled nucleic acid probe results in a heteroduplex that has a different electrophoretic mobility from that of the homoduplex if there are sufficient sequence differences between the amplicon and the probe. The mobility of the heteroduplex is affected more by insertions and deletions than by base substitutions. Therefore, if the test and reference strains are sufficiently divergent in their sequences, HTAs can be used to quantify their relative proportions in a growth competition assay over time. The env region is particularly amenable to being assayed by HTA because it often exhibits substantial sequence divergence and insertions/deletions among HIV strains. This method is less labor-intensive than other methods that rely on PCR amplification of the test and reference strains but requires that the two variants have substantial divergence in their nucleotide sequences.
Sequence analysis has also been used to quantify the relative amounts of the two variants in a growth competition assay either by direct sequence analysis of the bulk PCR product (65, 69, 78, 86, 117, 154, 155, 174) or by analysis of individual clones derived from the PCR amplicon (119). One disadvantage of bulk sequencing is that differences in the surrounding sequence may influence the assay's sensitivity for detecting a minority variant at a given codon. This is less of a concern when site-directed mutants that differ by only one or a few codons are compared but may impact reproducibility when clinical isolates are studied. In addition, quantitation of the relative amounts of mutants by analysis of the sequencing electropherogram from the bulk amplicon has a limited linear range. Clonal sequence analysis does not have these limitations if large numbers of clones are assayed, but this approach is significantly more labor-intensive and costly.
Real-time PCR has also been used to quantify the relative amounts of test and reference strains in a growth competition assay by allele-specific amplification of viral nucleic acid at a codon where the two variants differ (33, 142, 176). Real-time PCR has an advantage over other molecular assays used to quantify the relative proportion of the test and reference strains because of its high throughput and wide linear range of detection. However, surrounding genetic variation, which can influence the efficiency of primer or probe hybridization to the target sequence, can significantly influence the performance characteristics of allele-specific real-time PCR. Therefore, if real-time PCR is used to detect a specific viral mutation, the assay must be optimized and validated for each codon of interest.
Examples of reporter genes used to detect viral replication are luciferase (133), green fluorescent protein (GFP) (184), hisD from Salmonella enterica serovar Typhimurium, human placental alkaline phosphatase (PLAP) (103), and the mouse Thy1.1 and Thy1.2 alleles (57). A major advantage of the enzymatic assay to quantify luciferase is its high sensitivity and wide dynamic range of quantitation; a disadvantage is that different variants cannot be distinguished, and therefore, a growth competition assay is not feasible. The advantage of the paired reporter genes hisD/PLAP and Thy1.1/Thy1.2, which are detected by real-time PCR and flow cytometry, respectively, is that each gene in the pair can be cloned into a different vector, allowing the test and reference strains to be distinguished in growth competition assays (57, 103). A further potential advantage of the Thy1.1/Thy1.2 flow cytometry-based assay is that the number of cells infected by each variant is quantified, allowing an assessment of both virus growth and the relative proportions of test and reference strains (57). However, reporter genes are surrogate measures of viral replication, and linkage of the HIV-1 genome and the reporter gene is necessary for an accurate interpretation of the data; this linkage could potentially be disrupted if recombination between the test and reference strains that express different reporter genes were to occur. This theoretical limitation has not posed a problem under the conditions reported in previously published assays, which utilized a low virus inoculum to initiate infection (57, 103).
The specific target cells used in fitness assays may also affect the apparent fitness of a mutant. The best evidence for an impact of cell type on fitness has been obtained for HIV-1 mutants that are resistant to nucleoside analog inhibitors of the viral reverse transcriptase. For example, the replication fitness of M184V, which is resistant to lamivudine, was reduced in primary human PBMCs but was indistinguishable from that of wild-type HIV-1 in a lymphoid cell line (10). Purified reverse transcriptase with the M184V mutation showed reduced processivity of polymerization that was accentuated at low nucleotide concentrations (10). Since PBMCs are known to have lower concentrations of nucleotides than T-cell lines, this finding offers a potential biochemical mechanism for the differences in fitness of this mutant in primary cells versus cell lines.
A systematic study of HIV-1 with K65R or different numbers of thymidine analog mutations (TAMs) showed that the replication of each of these variants was less efficient in primary human macrophages than in T-cell lines (129). It is interesting that a mutant virus with the four TAMs D67N, K70R, T215Y, and K219Q actually had a replication advantage compared to wild-type virus in unstimulated PBMCs; this difference was not observed when the PBMCs were stimulated before infection (26). Similar to the M184V mutant, the differences in replication fitness observed under these conditions were associated with differences in processivity under limiting nucleotide concentrations. Thus, it seems likely that differences in the replication fitness of nucleoside resistance mutants are more likely to be observed in cell types in which nucleotide pools are limited.
In contrast, no such impact of cell type was observed when the replication fitness of HIV-1 mutants that are resistant to nonnucleoside reverse transcriptase inhibitors (NNRTIs) was studied (4, 69, 94, 174). These findings are consistent with the fact that these mutants do not appear to affect the processivity of polymerization or nucleotide affinity (51, 69, 174). We are not aware of published studies comparing the replication fitness of mutants that are resistant to protease inhibitors in different cell types. However, based on the known mechanisms for drug resistance of this class of mutant, one would not expect to see the pronounced effects of cell type on replication fitness that are seen with nucleoside-resistant mutants.
Because of the difficulty in maintaining primary human PBMCs in culture for extended periods of time, most assays of replication fitness utilize T-cell lines rather than primary cells. Thus, these assays may overestimate the replication efficiencies of isolates containing nucleoside resistance mutations.
In addition to these major differences in assay design, methods to determine the amount of virus used to infect a culture-based fitness assay may differ. The method used to determine virus inoculum can have a significant impact on the apparent fitness of a strain in parallel infections, single-cycle assays, or growth competition assays if the relative proportions of test and reference variants at a single time point after infection are compared to the proportions in the original inoculum. One study rigorously evaluated the correlation between different measures of virus input using a number of different clinical isolates and found that the reverse transcriptase activity of the virus stock correlated better with infectious titer than p24 antigen or quantitative viral RNA assays (115). It is not clear whether this analysis included strains containing drug resistance mutations in reverse transcriptase, which might alter the relationship between reverse transcriptase activity and virus infectivity, so these findings may not necessarily apply to drug-resistant strains of HIV-1.
The only commercially available fitness assay is a parallel-infection, recombinant-virus, single-cycle RC assay developed by Monogram Biosciences, Inc. (formerly ViroLogic) (15). This assay, which is a variation on Monogram's phenotypic drug susceptibility assay (133), compares test and reference strains in the absence of drug (Fig. (Fig.2).2). An amplicon spanning the p7-p1-p6 cleavage sites in Gag, the entire protease, and part of the reverse transcriptase (codons 1 to 313) is obtained from patient plasma and cloned in bulk into an HIV-1 vector derived from the pNL4-3 infectious molecular clone containing a luciferase reporter gene that replaces part of env (133). Luciferase is expressed upon establishment of infection and is a measure of viral infection (133). MLV envelope is supplied in trans during the production of the virus stock, yielding pseudotyped virus. Since the RC assay is a single-cycle assay and since luciferase detection can be automated, it has a high throughput. Luciferase quantitation also has a wide linear range. The potential disadvantages of this assay are that it measures the fitness contributions of the 3′ end of gag, protease, and part of reverse transcriptase only and uses parallel infections, which may be less sensitive and more susceptible to variations in virus input than growth competition assays. The use of a murine retroviral envelope instead of HIV-1 could also possibly affect the relative fitness of mutants that are resistant to protease and/or reverse transcriptase inhibitors, although this seems unlikely. The Monogram RC assay has been the most widely used fitness assay in clinical cohorts, and results correlating RC with clinical outcome will be summarized later in this review.
There are several other types of assays developed by individual research laboratories that have been used to quantify the relative fitness of site-directed mutants or clinical isolates. Here, we will briefly describe several different assays to illustrate the types of approaches that have been taken, recognizing that this is not an exhaustive list.
An assay used by several laboratories is a multiple-cycle recombinant-virus growth competition assay in which the relative proportion of test and reference strains is measured using either bulk or clonal sequence analysis (4, 65, 69, 78, 86, 117, 119, 154, 155, 174). This type of assay is labor-intensive and difficult to apply to large numbers of clinical samples but has been commonly used to characterize the relative fitness of site-directed drug-resistant mutants of HIV-1 in protease and reverse transcriptase. Depending on the laboratory and the specific mutants studied, either T-lymphocyte lines or primary human PBMCs have been used as target cells. A recombinant-virus growth competition assay in which test and reference strains are quantified using real-time PCR of the hisD and PLAP reporter genes has been used to study the relative fitnesses of mutants that are resistant to the nucleoside analog zidovudine (81, 105) and the fusion inhibitor enfuvirtide (104). A potential advantage of this assay is the greater dynamic range and reduced labor of real-time PCR.
A whole-virus, multiple-cycle growth competition assay in primary human PBMCs in which the relative proportion of test and reference viruses is measured using an HTA has been used to study the effects of env on fitness and disease progression and to evaluate the relative fitnesses of different subtypes of HIV-1 (6, 11, 140, 164). Because this growth competition assay utilizes an intact viral isolate and primary human PBMCs, it likely provides the closest approximation to HIV-1 replication fitness in patients compared to other currently available assays, but its labor-intensive nature limits its use for large-scale studies correlating fitness with clinical outcomes.
Some laboratories have utilized multiple-cycle, whole-virus, parallel infections to measure growth kinetics in primary human PBMCs (27, 163). This type of assay offers the advantages of using primary cells and an intact virus isolate as well as improved throughput compared to growth competition assays. If replication fitness is estimated using the slope of virus expansion after initial infection, this will markedly reduce the influence of virus inoculum on apparent replication fitness. However, a theoretical concern is assay-to-assay variation, since there is no internal control as in a growth competition assay. These assays have demonstrated associations of replication fitness with viral load and will be discussed in more detail later in this review.
Another recently reported assay developed by Tibotec and Virco is a high-throughput assay to measure replication rates in parallel infections. Eight serial dilutions of virus are made, with six time points tested per sample, and the measurements are normalized for viral input by multiplying by the dilution. Viral replication is measured using a cell line containing an enhanced GFP (EGFP) reporter gene under the control of an HIV long terminal repeat (LTR). Infection with HIV results in the activation of LTR and EGFP expression. The log10 of the product of fluorescence and viral dilution is linear with respect to time, and the slope of this line is the replication rate, which is independent of virus inoculum (146a). The very high throughput of this assay would make it very amenable to use in a clinical setting, although this use of the assay has not yet been reported.
Some recent publications have described fitness assays that utilize flow cytometry as a method to determine virus growth and variant proportions. The potential advantages of such assays are higher throughput and the ability to monitor the spread of infection at a cellular level. Zhang and coworkers designed a recombinant-virus single-cycle assay in which the GFP gene replaces env in the infectious HIV-1 molecular clone pNL4-3; test and reference strains were then compared in parallel infections (184). The use of red fluorescent proteins (DsRed2) and EGFPs as markers in a multiple-cycle recombinant-virus growth competition assay in primary human PBMCs has also been described and was used to characterize the relative fitness of mutants that are resistant to the fusion inhibitor enfuvirtide (125).
A flow cytometry-based recombinant-virus multiple-cycle growth competition assay has also been developed in which the test and reference viral vectors contain the mouse Thy1.1 or Thy1.2 allele in place of nef. The Thy gene products are expressed on the surface of infected cells early during viral infection and can be detected by commercially available fluorescence-labeled monoclonal antibodies (57). There was an excellent correlation of relative fitness of different drug-resistant mutants, as measured by flow cytometry, compared to direct and clonal sequence analysis, indicating that the linkage between the resistance mutations and the reporter gene was not disrupted by recombination (57). Such dual-marker flow cytometry-based assays are exciting in that they have the potential to allow the use of growth competition assays to measure replication fitness in larger-scale clinical trials.
An interesting modification to the cell culture growth competition assays described above is the rapid cell turnover assay (172). In contrast to standard cell culture assays in which half the cultured cells are typically replaced every 4 to 7 days, 90% of the cultured cells are replaced with fresh uninfected cells every 2 days, a time interval similar to the life span of T cells in patients. Under these conditions, it has been shown that the fitness of the highly pathogenic simian immunodeficiency virus clone Mne170 was higher than that of its parental clone CL8, whereas under normal culture conditions, they were indistinguishable. The result obtained under rapid cell turnover conditions is more consistent with the relative pathogenicity of the two clones in animals, suggesting that rapid cell turnover conditions may be a better predictor of viral pathogenicity in clinical infection. This interesting finding needs to be confirmed with more extensive studies of HIV-1 isolates.
Although more labor-intensive and much less commonly used, HIV-1 replication fitness has also been measured in SCID-hu mice that are reconstituted with human peripheral blood leukocytes and then infected with patient isolates. This methodology has been used to look at the properties of isolates with resistance mutations in protease and reverse transcriptase (91a, 135, 159). An interesting feature of this assay is that CD4+ T-cell depletion can be evaluated, suggesting that this assay may be able to evaluate some aspects of virus pathogenicity that are not captured in traditional cell culture-based assays of replication fitness. The disadvantages are that it is much more labor-intensive and costly than cell culture assays, and therefore, the ability to correlate viral replication fitness with clinical outcomes in significant numbers of clinical samples is extremely limited. Nonetheless, this assay has the potential to be a very interesting research tool to better understand the relationship between HIV-1 replication efficiency and pathogenicity.
Measurements of viral fitness have also been made by assaying the prevalence of TAMs in viral quasispecies present in plasma from patients not receiving antiretroviral therapy (71, 72). Allele-specific PCR assays or clonal sequence analysis was used to quantify the relative proportions of the different drug-resistant variants over time, and mathematical modeling was used to determine the fitness gain of the revertant virus over the mutant. An important advantage of this approach is that it takes into account selective pressures present in patients. However, this approach assumes that the specific mutation being assayed has the same effect in each viral genome and does not take into account the potential modulating effects of other mutations on viral fitness. For example, if the prevalence of T215Y in a patient did not decrease after the discontinuation of the drug due to a selective advantage conferred by a nonresistance mutation at another codon, this fitness value could erroneously be attributed to T215Y. In addition, the replication properties of the “wild-type” reference strain that overgrows the mutant will differ from patient to patient. Therefore, one must use caution in attributing fitness values obtained from such studies solely to the resistance mutation(s) being assayed.
Although cell culture-based fitness assays can provide qualitative information regarding the replication fitness of one variant relative to that of another, determining whether such in vitro measures of fitness correlate with clinical outcome requires reliable methods to quantitate the degree to which fitness is altered relative to the reference strain. Reliable quantitation will also be required if fitness assays are to be used in clinical practice. Unfortunately, there has been no clear consensus on how to quantify relative fitness, and publications differ in their mathematical definitions of a fitness coefficient. This lack of consensus has led to confusion in the literature and has further limited the ability to compare results from different studies. We will discuss methods to quantify fitness for multiple-cycle and single-cycle assays separately, since investigators have taken different approaches for these two types of assays.
According to population genetics, the relative fitness of a variant, defined as 1 + s, is the relative contribution of that variant to the next generation; s is defined as the selection coefficient. If a mutant is more fit than the reference strain, 1 + s will be greater than 1, and successive generations will have an increasing proportion of progeny derived from that mutant. If a mutant is less fit than the reference strain, 1 + s will be less than 1, and the prevalence of that mutant will decline over time relative to the reference strain.
A confusing aspect of the literature on fitness is that the value of 1 + s has sometimes been defined differently (reviewed in more detail in reference 181). Using viral dynamic models, Wu and coworkers attempted to clarify the different definitions of fitness, proposing standard nomenclature that distinguishes different approaches to measuring fitness and providing approaches to perform statistical evaluations (181). In these viral dynamic models, it is assumed that only two variants are present (i.e., no recombinants) and that there is a constant number of target cells over time (i.e., a large excess of susceptible target cells). Three parameters were defined, each of which takes a different approach to quantifying fitness: “log relative fitness,” “log fitness ratio,” and “production rate ratio.” The log relative fitness value, d, is equal to the natural log of 1 + s, where s is the selection coefficient defined by population genetics theory (Table (Table2).2). d is equivalent to the difference between the net growth rates of the mutant and reference strains (Table (Table2).2). In contrast, the log fitness ratio, r, is the ratio of the net growth rates of the mutant and reference strains. The production rate ratio, p, differs from the two previous parameters in that it is the ratio of the production rates of the mutant and reference strains and assumes that the life span of infected cells does not contribute to fitness. p is equivalent to 1 plus the selection coefficient defined by Maree and coworkers (112), which differs from the selection coefficient defined by population genetics theory.
An important point is that d and r compare the growth rates of the mutant and the wild type at the same time point and therefore do not need to be corrected if the culture is diluted during the course of the growth competition assay. In addition, p must incorporate the total expansion of the two viral variants; use of relative proportions of variants without accounting for the degree of viral expansion will give incorrect values. A publicly available website provides calculators and guidance on the use and statistical analysis of these different fitness parameters (http://www.urmc.rochester.edu/bstools/vfitness/virusfitness.htm).
The Monogram Biosciences assay expresses RC of a test strain as a percentage of the wild-type reference virus (14, 15, 41). All values are normalized for the efficiency of the transfections used to produce the virus stocks. For example, if infection with a recombinant virus derived from a clinical sample leads to half of the luciferase activity of the wild-type reference strain, and if the test and reference virus stocks had equal transfection efficiencies and were used in equal volumes in the infection, the RC value of the clinical isolate would be reported as 50%. There are no published data on how this method of normalizing virus inoculum compares to other methods.
A major focus of early literature on HIV-1 replication fitness was an evaluation of the impact of specific drug resistance mutations on viral fitness. The majority of these studies utilized site-directed mutants of laboratory strains, although some studies also evaluated clinical isolates (either whole virus or recombinant viruses containing a genome segment of the clinical isolate). This continues to be an important area of study as new drugs and combination therapies are introduced into clinical practice. In general, nearly all the drug resistance mutations studied adversely impact HIV-1 replication fitness to some extent, although different mutations vary in the magnitudes of their effect. Since replication fitness is proposed to influence the prevalence of an HIV-1 variant in patients and may also impact clinical outcome, an understanding of the relative impact of different drug-resistant mutations may provide important, clinically relevant information. The following section will describe what is known about the replication fitness of specific drug resistance HIV-1 variants conferring resistance to drugs that are currently FDA approved, organized according to the drug class that is affected. Because of the variations in assay design and approaches to calculating fitness values summarized above, much of this discussion will be qualitative, with comparisons among mutants limited primarily to those that were studied with the same methods (see Table Table33 for a summary of the effects of the single drug resistance mutations discussed below on fitness).
Nucleoside and nucleotide reverse transcriptase inhibitors (nRTIs) are competitive inhibitors of nucleotides that are normally incorporated during the synthesis of the viral genome. nRTIs bind to the polymerase active site of reverse transcriptase and competitively inhibit the synthesis of proviral DNA by acting as chain terminators during synthesis (for a review of this class of drugs, see reference 171). Several nRTI resistance mutations can confer cross-resistance to more than one nRTI.
The replication fitness deficit conferred by the M184V mutation in reverse transcriptase, which causes high-level resistance to the cytosine analogs lamivudine and emtricitabine, has been the subject of extensive study (reviewed in reference 132). M184V occurs frequently in viral isolates from patients failing lamivudine or emtricitabine therapy (63, 153). M184V has a fitness similar to that of the wild type in most T-cell lines where nucleotide concentrations are high (10) but reduced fitness in primary cells that have limited nucleotide pools, such as PBMCs and macrophages (3, 10, 129). Of note is that there are some more-recent studies that have found reduced fitness of this mutant in cell lines (35, 48).
A multitude of biochemical abnormalities have been proposed to explain the reduced fitness conferred by the M184V mutation, including decreased processivity of polymerization that is aggravated by low nucleotide concentrations, decreased initiation of minus-strand DNA synthesis from , decreased plus-strand DNA synthesis from the HIV RNA polypurine tract sequence, reduced binding affinity for nucleotides, improved fidelity of nucleotide incorporation, and slowed rates of RNase H cleavage (10, 48, 49, 173, 181a).
There has been a great deal of focus in the literature on the reduced replication fitness of M184V and its proposed clinical implications. Early observations of viral load rebound during lamivudine monotherapy and lamivudine-containing regimens demonstrated that plasma HIV-1 RNA concentrations remained lower than pretherapy levels despite the uniform presence of M184V in plasma viruses (132, 153). However, other studies have shown that lamivudine does exert some antiretroviral effect on the M184V mutant, suggesting that the incomplete viral rebounds that occur may represent residual antiviral activity rather than selection for a mutant with reduced fitness (139). Studies of selective lamivudine treatment interruptions demonstrate increases in viral load before reversion of the M184V mutant, also supporting the hypothesis that lamivudine retains antiretroviral activity against M184V (28).
In summary, the M184V mutant is clearly reduced in replication fitness, but these reductions appear to be no greater than those of many other drug-resistant mutants when measured in cell culture assays. Although the M184V mutant might be attenuated in ways that are not detected by standard replication fitness assays, many of the clinical observations supporting this hypothesis can also be explained by the persistent antiviral activity of lamivudine. It thus seems less likely that M184V has unique properties that make it substantially less pathogenic than other drug-resistant mutants.
Using growth competition assays in cell lines with site-directed mutants, the TAMs T215Y and M41L and the M41L/T215Y double mutant were all less fit than the wild type (78). K70R, which is usually the first TAM to emerge during zidovudine therapy, was substantially more fit in this assay than T215Y (78). This finding suggests that the reduced fitness of T215Y may explain the delay in this mutant's emergence relative to K70R, although the fact that T215Y requires a two-nucleotide change also likely contributes to its delayed emergence.
A quadruple TAM mutant (D67N/K70R/T215Y/K219Q) is unusual in that it has been shown to have a replication advantage over the wild type in PBMCs that were stimulated before infection but not if they were stimulated after infection, suggesting that this improved fitness relative to wild-type virus was dependent on low-nucleotide pools (26). This observation was supported by biochemical studies that demonstrated that the quadruple TAM mutant had increased processivity of polymerization relative to that of the wild type in the presence of low nucleotide concentrations (26).
Observations of patients failing zidovudine-containing regimens have identified two preferred combinations, or pathways, of TAMs: M41L/L210W/T215Y (TAM-1) and D67N/K70R/T215F/K219Q (TAM-2) (77, 111). The TAM-1 pathway is more common than TAM-2 in clinical isolates and differs from TAM-2 in that it causes cross-resistance to tenofovir. Using a recombinant-virus growth competition assay in a cell line, T215F was less fit than T215Y, which may explain why T215F rarely occurs by itself and is less common than the T215Y mutation (81). L210W reduced the fitness of D67N/K70R/K219Q in the presence or absence of drug but increased the fitness of M41L/T215Y in the presence of drug, which may explain why this mutation is seen only as part of the TAM-1 complex (81).
Studies of patients have demonstrated that the T215Y mutant can revert to intermediate mutations, such as T215S, T215D, or T215C, in the absence of drug. A study of a patient newly infected with a T215Y-containing strain of HIV-1 demonstrated the gradual replacement of T215Y by T215S (72), indicating that the revertant mutant was more fit than T215Y in this patient. Using viral dynamic modeling in which the ratio of their production rates was measured, the fitness of T215S was estimated to be 0.4 to 2.3% better than that of T215Y (72). In another individual recently infected with a drug-resistant isolate containing M41L and a mixture of T215Y/T215D/T215S, T215D and T215S were estimated to be 10 to 25% and 1% more fit, respectively, than T215Y (71). The improved fitness of the T215D/T215S mutants relative to T215Y was also observed in another study of site-directed mutants (66).
L74V, which confers resistance to both didanosine and abacavir, has reduced fitness in studies of site-directed mutants relative to wild-type virus and K70R using growth competition assays in primary human PBMCs and a cell line (35, 154). It has been shown that L74V in the context of the NL4-3 laboratory strain has an 11% loss of fitness compared to that of the wild type or K70R (154). Biochemical studies have shown that this replication decrease is associated with a reduction in the incorporation of nucleotides compared to the wild type and a decrease in the initiation of minus-strand and plus-strand DNA synthesis from the and polypurine tract primers (47, 49). A reduced ability of L74V viruses to synthesize proviral DNA was demonstrated in cell culture (61).
K65R, which confers resistance to the nucleotide analog tenofovir, has been shown to have a reduction in fitness when introduced into laboratory strains, as well as in the context of patient sequences, as measured by the Monogram Biosciences RC assay (35, 46, 179). Similar findings were also obtained using whole isolates obtained from patients failing tenofovir therapy (175). One study using single-cycle parallel cultures found a decreased fitness of K65R in primary macrophages but not in a cell line or primary human PBMCs (129). The replication defect of K65R correlates with a decrease in processivity of polymerization and in the polymerization rate, kpol, of purified reverse transcriptase (46, 179).
The Q151M complex (A62V, V75I, F77L, F116Y, and Q151M) confers resistance to all FDA-approved nRTIs except tenofovir (87) and yet develops much less frequently than T215Y (reviewed in reference 171). Both the single Q151M mutant and the full Q151M complex were first shown in parallel infections (in cell lines and PBMCs) to have a replication fitness similar to that of the wild type (106), but later competition experiments showed that the full Q151M complex was more fit than the Q151M single mutant and that both were actually more fit than the wild type, which is inconsistent with their low frequency in patients (93). Of note is that the mutation Q151M, like T215Y, requires a two-nucleotide change from the wild-type sequence. It is interesting that the replication fitness of 151L and 151K, the intermediate mutations leading to 151M, are much lower than that of the wild type, suggesting that the genetic barrier to developing Q151M is due in part to the low fitness of the intermediate mutants (64, 93).
NNRTIs are a structurally diverse group of compounds that bind to a hydrophobic pocket adjacent to the polymerase active site of reverse transcriptase. They inhibit reverse transcriptase by causing an allosteric change in the enzyme that interferes with polymerization. There are currently three FDA-approved NNRTIs: nevirapine, delavirdine, and efavirenz (for a review of NNRTIs, see reference 89). Efavirenz is widely used as part of first-line combination regimens for treatment-naive patients, and nevirapine is frequently used for the prevention of mother-to-child transmission during pregnancy (44, 45).
These drugs are potent inhibitors of HIV-1 replication, but NNRTI-resistant variants of HIV-1 develop rapidly if virologic suppression is not complete. The most commonly seen resistance mutations in clinical isolates from patients failing nevirapine and delavirdine therapy are K103N and Y181C; K103N is most common during efavirenz failure (9, 43, 145). However, several mutations, such as G190S, P236L, and V106A, are equally resistant to one or more of these drugs but occur uncommonly in clinical isolates. K103N is postulated to develop commonly in clinical isolates because it has minimally reduced replication fitness in the absence of drug (33, 57, 69, 94). Although one group of investigators has reported that Y181C has a slight fitness advantage over the wild type (84), two other groups have found that Y181C is modestly less fit than K103N and the wild type (4, 33). In contrast to K103N and Y181C, uncommonly occurring NNRTI resistance mutations confer significantly greater replication deficits (4, 33, 69, 94, 174). Some studies in which fitness assays were performed in the presence of efavirenz have shown that these replication defects relative to K103N persist in the presence of drug (69, 94, 174).
Less-fit NNRTI-resistant mutants have greater reductions in rates of RNase H cleavage, suggesting that this biochemical abnormality contributes to their reduced replication fitness (4, 5, 69, 174). In addition, recent studies suggest that reduced initiation from the primer also contributes to the reduced replication efficiency of at least some NNRTI-resistant mutants (174; for a more detailed review of the biochemical effects of NNRTI resistance mutations on reverse transcriptase function, see reference 52). When NNRTI-resistant mutants have been compared directly, the magnitude of their fitness deficit generally correlates with the magnitude of their biochemical effects on reverse transcriptase function.
Studies evaluating how long NNRTI resistance mutations persist after discontinuation of an NNRTI have also been used to estimate relative fitness. For example, NNRTI resistance mutations such as K103N and Y181C, which have minimal reductions in fitness, can persist for up to a year after the withdrawal of therapy (88).
More than one NNRTI resistance mutation may occur, particularly with more prolonged virologic failure, raising the question of how these mutations interact to affect HIV-1 replication fitness. One study of nevirapine-resistant mutants demonstrated that triple mutants were generally less fit than double mutants, which were less fit than single mutants (33). However, the fitness values of specific single and double mutants did not always predict the relative fitness of the corresponding triple mutant, suggesting that the relative fitness of NNRTI-resistant mutant combinations is not merely a function of their number. In addition, some quadruple mutants were more fit than the triple and some of the double mutants (33). The fact that the number of NNRTI resistance mutations does not always predict replication fitness is also supported by studies of secondary mutations conferring resistance to efavirenz that develop after the emergence of K103N (94). Those studies indicate that which specific secondary mutation is present has a significant impact on the relative fitness of the double mutant and that the relative fitness of the double mutants generally correlates with their frequency in clinical samples (94).
Thus, there is evidence from studies of both nRTI and NNRTI resistance mutations that relative replication fitness, as measured in a cell culture assay in the absence of drug, generally correlates with mutant prevalence during virologic failure in patients. This finding is compatible with predictions based on theoretical models of viral dynamics (31). A prediction of this model is that HIV-1 replication fitness in the absence of drug influences the relative frequency of randomly generated mutants before therapy initiation and, thus, the likelihood that they will emerge during virologic failure (31). Studies with some poorly fit NNRTI-resistant mutants demonstrate that the selective disadvantage relative to K103N persists even in the presence of drug. These studies provide support for the concept that relative fitness correlates with mutant prevalence in patients and suggest that this information may be helpful during preclinical drug development, particularly in the search for newer-generation NNRTIs that do not select for K103N or Y181C.
Protease inhibitors act as competitive inhibitors of protease cleavage, which is required for the posttranslational processing of the gag and pol gene products (reviewed in reference 85). They are widely used in combination with nRTIs as initial therapy for treatment-naive patients and as part of combination regimens for treatment-experienced patients (reviewed in reference 20). High-level resistance to protease inhibitors usually requires several mutations in the protease gene. Protease resistance mutations are divided into two groups: major and minor mutations (87). It should be noted at the outset that the classification into major and minor mutations is not always clear-cut, and some mutations have been placed into both categories, depending on the specific protease inhibitor. Nonetheless, this classification scheme has proven useful for many of the protease inhibitor resistance mutations.
Major mutations (also referred to as primary mutations) are generally those that are acquired early in failure or which directly affect drug binding or drug susceptibility. Examples of major mutations are D30N, I50V/L, V82A/F/T/S, I84V, and L90M. Some of these major mutations confer resistance primarily to specific protease inhibitors, such as D30N for nelfinavir and I50L for atazanavir. These major mutations are generally located in the active site of the HIV-1 protease.
Minor mutations (also referred to as secondary mutations) usually develop after the emergence of major mutations and by themselves do not increase the level of drug resistance (87). Examples of minor mutations are L10I/R/V, A71V/T, and V77I. Minor mutations are generally located away from the active site of protease. Some minor mutations may be present in treatment-naive patients, although usually fewer than three minor resistance mutations are observed in this patient population (42, 95).
The development of HIV-1 resistance to the protease inhibitor nelfinavir occurs by the acquisition of the primary resistance mutation D30N or L90M (128). D30N is more resistant to nelfinavir than L90M but is susceptible to other protease inhibitors, unlike L90M. The fitness of D30N is quite impaired and is substantially lower than that of either the wild type or L90M; L90M is only modestly impaired in replication (57, 119, 130). Thus, the higher degree of nelfinavir resistance appears to drive the preferential selection for D30N, whereas the relatively preserved fitness of L90M appears to favor its emergence despite its lower level of drug resistance. D30N was also found to be significantly impaired in its ability to infect SCID-hu mice transplanted with human thymus (91a). The D30N mutant protease has reduced catalytic efficiency, which presumably explains its decreased fitness (30). In another study, D30N virus demonstrated processing of Gag and Gag-Pol proteins that was substantially reduced compared to that of L90M, consistent with their relative replication fitness (161). D30N and L90M rarely occur together in clinical isolates, most likely due to the fact that the double mutant is profoundly reduced in replication fitness (130, 161).
G48V, which is a major mutation conferring resistance to saquinavir, was shown to have decreased fitness in a single-cycle assay (109). I50L, which confers resistance to atazanavir, also has reduced fitness in parallel infections (34). The I50V mutation, which is associated with resistance to amprenavir, lopinavir, and darunavir, showed a marked reduction in fitness in a single-cycle assay; I84V, which is more broadly cross-resistant, showed a similar reduction in fitness in the same assay (138). Another group did not identify a clear reduction in the replication of the I84V mutant but did demonstrate impaired proteolytic processing by this mutant (36). V82T, which is one of the substitutions at codon 82 that confer broad cross-resistance, was also significantly reduced in fitness compared to that of the wild-type virus (118). V82A, which confers low-level but broad cross-resistance, did not demonstrate significantly reduced fitness in a single-cycle assay but did have a replicative advantage compared to wild-type virus in the presence of drug (109). Consistent with these results from site-directed mutants are studies of serial clinical isolates obtained from patients failing protease inhibitor therapy that have correlated the development of reduced fitness in the Monogram RC assay with the emergence of V82A, I84V, or L90M (14). In addition, clinical isolates containing V82A had reduced replication in the SCID-hu mouse model (135).
In summary, most major protease inhibitor resistance mutations confer some replication defect, although in some cases, this defect is modest. For those mutants that have been studied, these replication deficits appear to be associated with reduced catalytic efficiency, consistent with their location at or near the active site of protease. The greatest deficits appear to be due to those mutations, such as D30N and G48V, which confer higher-level resistance that is restricted to one or a few protease inhibitors. Other primary mutations that confer lower-level, broader cross-resistance appear to result in smaller replication deficits, although these two types of mutants have not been directly compared in a single study.
In contrast to major mutations, minor protease inhibitor resistance mutations by themselves generally have little or no adverse effect on HIV-1 replication fitness. For example, little or no reductions in fitness were seen for L63P, M46I (118), and L10F (138). In addition, at least some minor resistance mutations can completely or partially improve the replication deficits conferred by major resistance mutations. Some of these interactions can be quite specific; for example, L63P can compensate for the reduced fitness of L90M but not D30N (119). A71V can also compensate for the reduced fitness of D30N (130).
It should be noted, however, that some combinations of major and minor resistance mutations actually decrease fitness, also suggesting that these mutational interactions are specific (138). A recently reported study in which a large number of major and minor mutations (31 single and 42 double mutants) were evaluated for infectivity and fitness with the Monogram RC assay indicates that most minor mutations can compensate for the fitness deficit conferred by most major mutations. This finding suggests that the mechanism(s) for compensation are probably nonspecific for double mutants; more complicated combinations of mutations were not studied (78a).
Consistent with most of these studies of site-directed mutants are studies of clinical isolates obtained from patients failing protease inhibitor therapy that demonstrated an initial loss of fitness and increase in resistance associated with the emergence of major mutations, followed by progressive increases in fitness with the accumulation of minor mutations (14, 126). However, the improvements in fitness were not always complete, suggesting that HIV-1 protease may be limited in its ability to become both highly resistant and highly fit (14). It has been shown that recombinant viruses containing protease genes from clinical isolates tend to evolve in cell culture in the absence of drug only if the initial replication fitness, as measured by the Monogram RC assay, is substantially reduced (166). The paths to improved fitness involved either a reversion of the major resistance mutation(s) or the acquisition of compensatory mutations. Recombinant viruses that replicated similarly to or better than the wild type did not evolve, suggesting that there were no available evolutionary pathways to further increase fitness (166).
For some combinations of protease inhibitor resistance mutations present in clinical samples, the reversion of any one mutation in the combination results in virus that is less fit than virus with all the mutations intact, indicating that the virus would have to go through a fitness trough in order to revert (167). This finding could explain the persistence of what appear to be unfit mutants in patients after the discontinuation of therapy.
The fitness studies demonstrating compensation by minor resistance mutations are also supported by biochemical analyses. For example, one study demonstrated that M46I and L63P by themselves augment the catalytic efficiency of HIV-1 protease and compensate for the reductions in catalytic efficiency conferred by the major mutations V82A and I84V (152). When both M36I and A71V are introduced into a protease containing D30N, the triple mutant demonstrated improved catalytic efficiency compared to the single and double mutants as well as wild-type protease (30).
Hypersusceptibility to protease inhibitors, defined as a 50% inhibitory concentration of ≤0.4-fold compared to the wild type, can occur in treatment-experienced patients and some treatment-naive patients and is associated with reduced viral replication fitness (23, 120, 144, 185). The genotypes responsible for the correlation of hypersusceptibility with replication fitness are still not fully understood, and more studies are needed to determine if the same mutations are responsible for both hypersusceptibility and reduced fitness.
Enfuvirtide (T-20) is the first FDA-approved drug belonging to the growing class of entry inhibitors. This class of drugs targets the step in the viral life cycle when the virion attaches and fuses to the cell membrane. Enfuvirtide is a 36-amino-acid peptide that mimics the structure of the heptad repeat 2 (HR-2) domain of gp41; enfuvirtide binds to the HR-1 domain of gp41, thus disrupting an intramolecular interaction required for the fusion of the viral envelope with the cell membrane (75). When HIV-1 is passaged in the presence of enfuvirtide in cell culture, mutations at amino acids 36 to 38 of HR-1 emerge, which confer resistance to enfuvirtide (146). In patients failing enfuvirtide-containing regimens, mutations at codons 36 to 45 and the double mutations N42T/N43K and V38A/N42T emerge and have been shown to confer resistance (122, 158). The resistance mutations selected by drug in cell culture (37T, 38 M, and 36S/38 M) are less fit than the wild type in the absence of drug (104). The mutation combinations selected in patients were also shown to be less fit than the wild type, with a hierarchy as follows: wild type > N42T > V38A > N42T/N43K = N42T/N43S > V38A/N42D = V38A/N42T > V38E/N42S. The reduced fitness conferred by enfuvirtide resistance mutations in HR-1 is most likely due to their impairment in infectivity and delayed fusion kinetics (143). Of interest is that one study also found that some of these enfuvirtide-resistant mutants conferred enhanced sensitivity to neutralizing antibodies, suggesting another way in which the acquisition of enfuvirtide resistance may confer a selective disadvantage to the virus.
HIV-1 isolates from patients who are enfuvirtide naive have a range of susceptibilities to enfuvirtide. The reduced susceptibility observed in some isolates from enfuvirtide-naive patients is not thought to be clinically significant, since no reduction in the virologic response to enfuvirtide in patients whose isolates demonstrate this property could be demonstrated (121). Interestingly, envelope sequences from enfuvirtide-naive patients that had reduced susceptibility to enfuvirtide but that did not have known resistance mutations had preserved fitness, using a growth competition assay in PBMCs, in contrast to resistant mutants that develop during treatment failure (125).
We have already described some interactions among nRTI resistance mutations that typically cluster together during the failure of a specific nRTI. For example, L210W has different effects on fitness depending on with which TAM pathway it is associated (81). A mutant with the TAMs M41L and T215Y is more fit than M41L alone, offering an explanation as to why M41L is rarely observed without T215Y (78). The mutations that are part of the Q151M complex interact to improve the fitness of the Q151M single mutant, also providing an explanation for the preferred association of these mutations (93).
In addition, there are examples of interactions among nRTI resistance mutations that confer resistance to different nucleoside analogs. L74V and K65R, which are rarely found together in clinical isolates, have markedly antagonistic effects on viral fitness that correlate with reductions in the ability to utilize normal nucleotides during DNA synthesis (47). Similarly, K65R and M184V also have antagonistic effects on fitness that are associated with abnormalities in nucleotide incorporation (46). The combination of K65R and the Q151M complex reduced fitness compared with either mutant in isolation and reduced the catalytic efficiency of reverse transcriptase. In contrast, the Q151M complex did not reduce the fitness of M184V, suggesting that the combinations of nRTIs prescribed to the patient may influence whether the Q151M complex develops (66a). A recent report has demonstrated that the reduced fitness of M184V is improved when placed in combination with the TAM-2 cluster of mutations; no such effect is seen when M184V is combined with TAM-1 mutations (35). The two clusters of TAM mutations had opposite effects when combined with K70R (35).
A recent report, which documented the co-occurrence of L74V, Q151M, and a deletion at codon 70 (Δ70) in a patient isolate, confirmed (using site-directed mutagenesis) that Δ70 improved the fitness of L74V and Q151M so that each double mutant was more fit than the wild type. The positive impact of Δ70 on fitness in this context was also confirmed by studies of a recombinant virus containing the patient's viral reverse transcriptase sequence (81a).
Serine insertions between codons 69 and 70 of reverse transcriptase can also confer multidrug resistance. The replication fitness of an isolate containing this insertion at codon 69 is highly dependent on the presence of T215Y (137). These data indicate that this insertion may not have an adverse effect on fitness provided that it occurs in combination with another specific nRTI resistance mutation(s). Such studies also have the potential to provide information that could assist in designing combination nRTI regimens that delay the onset of resistance, for example, using nRTIs that select for resistance mutations that have antagonistic fitness (or resistance) interactions and that avoid those that have synergistic effects to increase fitness (or resistance).
Accumulating data indicate that NNRTI and nRTI resistance mutations can interact to affect viral fitness and that these interactions may influence how frequently a certain NNRTI-nRTI resistance mutation combination emerges. For example, the highly efavirenz-resistant mutant G190E has markedly reduced replication fitness (82, 127). A study in which the G190E mutant was passaged in cell culture in the presence of an NNRTI, but without nRTI selection pressure, demonstrated the acquisition of the nRTI resistance mutation L74V (92). Further studies have demonstrated that the L74V mutation substantially improves the replication fitness of G190E both in laboratory strains and in clinical reverse transcriptase backbones (82). Studies using the single-cycle Monogram RC assay also indicated that L74V improved the replication fitness of the NNRTI-resistant G190S mutant (82).
A recently presented study also demonstrated that the L74V mutation substantially improves the fitness of the NNRTI-resistant double mutant K101E/G190S (174a). It was also observed that the NNRTI-resistant double mutant K103N/L100I was frequently associated with L74V in clinical samples, even in patients who did not receive didanosine or abacavir at the time that the triple combination was selected for; the frequent association of these three mutations was also observed in another study (2, 41a). In cell culture, L74V increased the fitness of K103N/L100I (94). Thus, it appears that L74V is able to compensate for the fitness deficits of a number of different poorly fit, NNRTI-resistant variants. This raises the question of whether nRTIs that select for L74V, if combined with efavirenz, may promote the rapid emergence of highly resistant, otherwise poorly fit NNRTI-resistant variants. Although this phenomenon has not been definitively demonstrated, it is very interesting that a number of the mutants emerging during early virologic failure of the combination of didanosine, tenofovir, and efavirenz contained either L74V/G190E or L74V/K103N/L100I, NNRTI resistance mutations that are normally rare during therapy with efavirenz (97, 136).
There is a recent report indicating that the TAMs M41L and T215Y also compensate for the reduced replication fitness of K101E/G190S, although unlike L74V, these TAMs reduce the magnitude of efavirenz resistance of the K101E/G190S mutant (174a). Those studies raise the question of whether interactions between NNRTI and nRTI resistance mutations to affect viral replication fitness (or resistance) may affect which drug-resistant mutants emerge or the timing of their emergence.
As noted above, Q151L, which is likely an intermediate of the Q151M multi-nRTI resistance mutation, has a profound replication deficit and was lethal when placed in the reverse transcriptase backbone of both HXB-2 and a pretherapy isolate from a patient that ultimately developed the Q151M variant (64). Of interest is that the profound replication deficit of Q151L was markedly improved when placed in the reverse transcriptase backbone of the posttherapy isolate from this patient. When the S68G mutation, which was present only in the posttherapy isolate, was introduced into HXB-2 and pretherapy reverse transcriptase sequences containing Q151L, the ability of the corresponding recombinant virus to replicate was partially restored. This may explain the association between S68G and Q151M observed in one study of clinical samples (151). This finding also suggests that sequence context may partially explain the infrequent occurrence of Q151M and influence which virus isolates develop that mutation.
The S68G mutation is also associated with K65R in clinical samples, although clonal analyses indicate that these mutations are not always on the same genome (148, 180). A recently reported study demonstrated that S68G partially compensates for the fitness deficit of K65R, as does the A62V mutation (161a).
As-yet-undefined polymorphisms present in clinical reverse transcriptase sequences can also modulate the effects of resistance mutations on HIV-1 replication fitness. A double serine insertion at codon 69 (69insSS), when present in the clinical reverse transcriptase backbone in which it evolved, conferred improved replication fitness compared to a laboratory isolate with the same insertion as well as the same clinical reverse transcriptase backbone without the insertion (141). A study of the NNRTI resistance mutations K103N and P236L, which confer preserved and reduced replication fitness, respectively, demonstrated that most recombinant viruses containing clinical reverse transcriptase sequences with these mutations replicated as expected in parallel growth kinetic assays (56). However, there was one reverse transcriptase sequence in which the relative replication efficiency of K103N and P236L variants appeared to be reversed, suggesting that polymorphisms in reverse transcriptase can also modulate the fitness effects of NNRTI resistance mutations.
The interactions between major and minor protease resistance mutations provide clear examples of how the effect of drug resistance mutations on fitness can be modulated (see “Minor protease inhibitor resistance mutations” above). In many instances, the selection for minor mutations after the emergence of major protease resistance mutations occurs in association with increases in viral replication fitness, implicating that compensatory fitness mechanisms are an important reason for the selection of minor protease resistance mutations (14, 126). Some, but probably not all, of these mutational interactions are specific. For example, the reduced fitness of D30N can be rescued by the accumulation of minor resistance mutations such as A71V and N88D, whereas the reduced fitness of L90M is rescued by L63P (119, 130). Biochemical studies also support the ability of M36I and A71V to rescue the reduced catalytic efficiency of protease containing D30N (30).
There is also evidence that mutations within protease other than those classified as resistance mutations can modulate the effects of major protease resistance mutations on viral replication fitness. For example, D30N is observed much less frequently relative to L90M in subtype C isolates compared with subtype B. Interestingly, D30N appears to have a greater adverse impact on fitness in a subtype C reverse transcriptase backbone than in a subtype B backbone, suggesting that reduced fitness makes the selection of D30N less likely in a subtype C background (70, 76).
Patients who fail protease therapy and harbor subtype F isolates do not acquire D30N and accumulate L90M only infrequently (25). This appears to be due to the fact that accumulation of L90M in subtype F is dependent on first acquiring L89M. Subtype F HIV-1 with L90M alone is highly unfit relative to subtype F virus containing either L89M/L90M or L89M alone. These data suggest that the genetic background can alter protease resistance mutation frequency by influencing HIV-1 replication fitness.
One study has demonstrated that some env-recombinant viruses derived from patients failing therapy with enfuvirtide had minimal reductions in fitness, even though they carried enfuvirtide resistance mutations known to reduce the replication fitness of HIV-1 (123). This finding provides another example of how background sequence may influence a drug resistance mutation's effect on fitness. Because of the small number of cloned viruses studied, it is not clear how often this phenomenon occurs with enfuvirtide-resistant variants.
The replication capacity of some drug-resistant HIV-1 isolates is higher than that of recombinant viruses derived from laboratory strains that contain the protease and reverse transcriptase regions of the corresponding clinical isolates (19, 156). These findings indicate that protease and reverse transcriptase sequences do not contain the sole determinants of HIV-1 replication fitness in drug-resistant isolates and that mutations outside these regions can modulate the effects of drug resistance mutations on fitness. The following paragraphs describe specific examples of extragenic compensation.
It has been shown that the passage of HIV-1 in the presence of a protease inhibitor results in drug resistance mutations in protease but also mutations at the sites where protease cleaves Gag to form p1 and p6 or nucleocapsid (p7) and p1 (55). These cleavage site mutations improved the efficiency of cleavage by protease inhibitor-resistant proteases and were necessary for optimal virus replication, suggesting that they were selected for because of their compensatory effects on fitness (55). One study observed gag cleavage site mutations in HIV-1 isolates from patients during virologic failure on protease inhibitors, although these mutations were not present in the majority of protease inhibitor-resistant HIV-1 isolates (108). Another study, which focused on the p7/p1 and p1/p6 cleavage sites, found mutations in 60% of treatment-experienced patients and in only 10% of those who were treatment naive (170). Cleavage site mutations were seen primarily in samples with two or more major resistance mutations in protease, suggesting that cleavage site mutations are selected during therapy (170).
In an analysis of a clinical sequence database, the L449F and P453L mutations at the p1/p6 cleavage site in Gag were frequently observed in combination with the major protease inhibitor resistance mutation I50V, although the two cleavage site mutations were not observed together in individual clones derived from patient samples (107). These mutations each partially improved the replication fitness of HIV-1 strains containing the I50V mutation and increased the degree of protease inhibitor resistance; no effect on protease inhibitor resistance was seen when these mutations were introduced into a drug-sensitive laboratory strain (107). In the absence of these mutations, the I50V mutant accumulated uncleaved p1/p6 protein; this abnormality was partially corrected by the addition of one of the cleavage site mutations (107). Those studies also confirmed that purified protease containing the I50V mutation cleaved a mutant p1/p6 cleavage site more efficiently than a wild-type cleavage site (107). Positive and negative associations between specific cleavage site and major protease-inhibitor resistance mutations were also observed in another study of treatment-experienced patients (170).
Other studies of clinical gag and pol sequences also demonstrated that cleavage site mutations can partially, but not fully, compensate for the reduced fitness of protease inhibitor-resistant mutants and that these partial improvements in fitness correlate with partial improvements in the cleavage of the Gag polyprotein (108). It is possible that there may be evolutionary barriers to the accumulation of mutations at all cleavage sites. For example, mutations at the p66/p51 cleavage site dramatically reduced viral fitness and levels of virion-associated reverse transcriptase in one study (1).
Another study demonstrated the occurrence of amino acid insertions near Gag cleavage sites (162). These insertions reduced the fitness of drug-sensitive virus but improved the fitness of protease inhibitor-resistant viruses. These differences in fitness were attributable to changes in cleavage efficiency by HIV-1 protease (162). Another small study showed that mutations in the cleavage sites between p2 and nucleocapsid and between p6 and protease can develop in patients on protease inhibitor therapy and are important for improving the replication capacity of HIV-1 variants that are resistant to protease inhibitors (92a).
The PTAPP motif is a conserved proline-rich region in the amino terminus of the Gag p6 protein. p6 plays a role in efficient virus release and the incorporation of Vpr into virions (46). The PTAPP motif is duplicated in some isolates from heavily nucleoside-experienced patients (83, 96, 114, 131, 162). Infectious clones containing a three-amino- acid duplication in p6, APPAPP, demonstrated reduced p6 processing and increased packaging of reverse transcriptase (131). Interestingly, these clones demonstrated a significant growth advantage over wild-type virus in the presence of nucleoside inhibitors, suggesting that the augmented incorporation of reverse transcriptase allowed the virus to become resistant to nucleoside analogs (131).
In the absence of drug, the fitness of isolates containing duplications in the PTAPP region without other resistance mutations is reduced relative to that of the wild type (162). However, the introduction of these duplications improved the replication of protease inhibitor-resistant strains (162). Studies using Western blot analysis of virions have shown that insertions in the PTAPP region impair Gag processing if the protease sequences contain no drug resistance mutations but improve the processing of protease inhibitor-resistant strains (162). Presumably, the proximity of these insertions to the p6 cleavage site affects the efficiency of cleavage.
Mutations in gag that were not associated with cleavage sites or the PTAPP motif were also observed to accumulate when a laboratory strain was passaged in the presence of escalating concentrations of protease inhibitors (68). Two of these mutations, L75R and H219Q, improved the replication fitness of the protease inhibitor-resistant mutant L10F/V32I/M46I/I54M/A71V/I84V in the absence of drug (68). These mutations in gag were also essential for the mutant virus' replication in the presence of protease inhibitors. The L75R and H219Q mutations also improved the replication fitness of the wild-type laboratory strain NL4-3 but did not allow replication in the presence of protease inhibitors (68). It is of interest that the H219Q mutation is in the cyclophilin A binding loop and that the replication advantage conferred by it was influenced by the cyclophilin A concentration of different cell types (67), suggesting that gag-cyclophilin A interactions are important for HIV-1 replication fitness. There are several additional examples of mutations outside of protease cleavage sites that improve the fitness of protease inhibitor-resistant viruses in the absence of cleavage site mutations, indicating that the coevolution of gag and protease is very complex (124).
An example of extragenic compensation unrelated to drug resistance is the development of mutations in gag that compensate for viruses that have deletions in the 5′ untranslated region (5′-UTR) (98, 99, 149). The 5′-UTR has several elements that are important for the HIV-1 viral life cycle (32), one of which is the dimerization initiation sequence (DIS) (8). The DIS is important for viral genomic RNA packaging into the virion because it promotes annealing between the two RNA genomes and also interacts with the p24 gag protein that plays a role in the packaging of viral genomes (17). Deletion of the DIS results in decreased viral infectivity, which can be restored by mutations in gag (V35I in matrix, I91T in capsid, T12I in p2, and T24I in nucleocapsid) (98, 99). Another 5′-UTR element, stem-loop 3, has also been shown to be important for genomic RNA packaging and dimerization. The decrease in fitness observed by the deletion of this stem-loop can be compensated for by an A11V mutation in p2 or an I12V mutation in nucleocapsid (149). It has been shown that deletions in sequences downstream of the DIS can also be compensated for by mutations in the 5′-UTR as well as in nucleocapsid and the primer binding site (100). Although this example does not involve drug resistance mutations, it illustrates that compensatory mutations can occur in unexpected regions of the genome.
Thus, there are substantial data demonstrating that viral mutations, either within the same gene product (intragenic) or outside that gene product (extragenic), can modulate the effects of drug resistance mutations on HIV-1 replication fitness. Studying the effects of these interactions is important because one can potentially obtain information on the underlying mechanisms of fitness compensation and better understand the interactions between different domains of a drug target or between different gene products. In addition, if replication fitness does impact clinical outcome, an understanding of which mutations modulate the effects of drug resistance mutations on fitness could be relevant to monitoring HIV-1-infected patients failing therapy.
Some studies have evaluated the relative contributions of different regions of the HIV-1 genome to replication fitness in cell culture. This information is interesting from a pathogenic perspective but also provides potentially important data for designing a clinically relevant recombinant virus assay to assess the fitness of HIV-1 isolates in different patient populations. The relative fitnesses of whole-virus isolates of HIV-1 obtained from untreated patients and recombinant viruses containing the corresponding env sequences were very similar in a multiple-cycle growth competition assay in PBMCs, suggesting that env has a dominant influence on the replication fitness of HIV-1 in untreated patients (116, 142). In one of these studies, which evaluated two different strains of HIV-1, the relative fitnesses of the whole virus and its corresponding env-recombinant virus also correlated with affinity for CD4 and CCR5 receptors (116). It is interesting that in the one treatment-experienced patient assessed, the correlation between the replication of whole virus and that of the env-recombinant virus was poor, suggesting that resistance mutations in protease and reverse transcriptase have a greater influence on replication fitness than env during treatment failure (142). Another study found that differences in LTRs in different HIV-1 subtypes were associated with differences in replication fitness that were cell type specific, suggesting that noncoding regions of the viral genome may also influence replication fitness (168, 169). A third study of untreated patients found that the replication rate, as measured in multiple-cycle, whole-virus, parallel infections, was significantly associated with fitness as measured by the Monogram RC assay, although the magnitude of the correlation was only moderate (r2 = 0.53; P = 0.007) (27). This study suggests that protease and reverse transcriptase sequences also contribute to relative fitness.
More studies that directly compare the relative contributions of different gene segments in a large number of isolates are needed to definitively determine the relative importance of different genomic regions in influencing fitness. It does stand to reason, however, that regions other than protease and reverse transcriptase could play a dominant role in HIV-1 fitness in untreated patients, raising questions about the clinical utility of recombinant virus assays that were designed primarily to assess drug resistance to measure HIV-1 fitness in this patient population. Although protease and reverse transcriptase sequences likely have dominant effects on fitness during virologic failure, the documented effects of extragenic compensatory mutations suggest that pol-based recombinant virus assays may become less predictive of fitness as the duration of virologic failure increases.
There is good evidence that HIV-1 replication fitness of a mutant influences its likelihood of developing during treatment failure. One important question is whether fitness also influences the likelihood of a mutant being transmitted from one person to another. Another critical question is whether the transmission of mutants with reduced fitness leads to improved clinical outcomes. Since patients are initially infected with one or a few strains, it is reasonable to postulate that unfit mutants could persist after the establishment of infection and influence clinical outcomes. The plasma HIV-1 RNA concentration is a marker of virus burden and presumably could be influenced by how efficiently an HIV-1 variant replicates. Thus, it seem reasonable to postulate that initial infection with poorly fit mutants may result in lower viral load set points after primary infection. Viral load is a major predictor of the rate of CD4+ T-cell decline, and both viral load and CD4+ T-cell count are important predictors of disease progression. Thus, reduced replication fitness may also be associated with slower disease progression. Some investigators have postulated that certain drug-resistant variants affect primarily CD4+ T-cell depletion without substantial effects on viral load. It is not clear how a reduced rate of replication would selectively affect CD4+ T-cell depletion; therefore, traditional assays for fitness that measure replication efficiency would likely not be ideal to assess this aspect of viral fitness, although it is certainly possible that reduced replication efficiency is one feature of mutants with reduced pathogenicity.
Mutants with reduced replication fitness can also be selected for during failure of antiretroviral treatment regimens. Although these mutants clearly have a relative selection advantage over wild-type virus in the presence of drug, their absolute rate of replication in the presence of drug is likely lower than that of the wild-type virus before initiation of therapy. Therefore, selection for an unfit variant might result in viral loads during treatment failure that do not fully return to pretherapy levels. In addition, since different drug-resistant mutants have different degrees of impairment in fitness, patients failing similar treatment regimens could have different viral load and CD4+ T-cell trajectories depending on the relative replication fitness of the drug-resistant mutant that is selected. Evidence for and evidence against an association between HIV-1 replication fitness and these clinical outcomes are summarized below.
It should be noted that the establishment of HIV-1 infection by a given variant is a complex process that is influenced by a number of factors: the likelihood that transmission from the source patient will occur, the prevalence of the mutant in genital secretions, the ability of that mutant to establish infection in dendritic cells, and the ability of that mutant to replicate and establish infection in CCR5+/CD4+ T cells. Thus, the fact that a specific mutant occurs less frequently than expected in recently infected patients may be due to alterations in any or all of these factors. For the sake of simplicity, we will refer to this process as “transmission,” recognizing that this term does not reflect the full complexity of the process by which HIV-1 infection is established.
There is growing evidence that drug-resistant strains of HIV-1, which usually have impaired replication fitness as measured in cell culture assays, are underrepresented in primary HIV-1 infection. Surveillance studies have demonstrated that approximately 10 to 20% of recently infected patients carry drug-resistant strains of HIV-1 (74, 101, 157, 183). In addition, approximately 8 to 10% of drug-resistant isolates are identified in treatment-naive, chronically infected patients (178). Retrospective studies of chronically infected patients that could serve as potential transmitters of HIV-1 infection indicate that the prevalence of drug-resistant mutants is substantially higher than in contemporary cohorts of recently infected patients (24, 182). This evidence has been interpreted to indicate that reduced replication fitness correlates with reduced transmission efficiency, although these studies did not directly measure replication fitness.
There is some evidence that specific drug-resistant mutants may be less transmissible than others. For example, one study found that isolates with M184V or major protease inhibitor resistance mutations were underrepresented in patients with primary drug-resistant HIV-1 infection compared to patients with chronic infections (165). In another study, the reverse transcriptase mutations M184I/V and T215F/Y and the protease mutations M46I/L were specifically underrepresented in recently infected patients compared to potential transmitters. In contrast, V118I, Y181I/C, and K219E/Q in reverse transcriptase and I84V and L90M in protease were overrepresented in recently infected patients and therefore appeared to be more efficiently transmitted. Other mutations such as Q151M and Y188C/H/L in reverse transcriptase and D30N, G48V, and V82A/F/S/T in protease were not detected in any seroconverters in that study, despite their presence in potential transmitters (165).
None of those studies directly correlated replication fitness, as measured in a cell culture assay, with likelihood of transmission, so a correlation between replication fitness and transmission efficiency can only be inferred. Some direct support for this concept is provided by a study that utilized the Monogram Biosciences RC assay, which measures the impact of the 3′ end of gag, protease, and part of reverse transcriptase on replication fitness and found that increased fitness was associated with an increased risk of mother-to-child transmission, even when accounting for other confounding variables (60).
Consistent with that study is one that found that HIV-1 group O and HIV-2, which occur at a low frequency worldwide relative to group M HIV-1 isolates, also have reduced replication fitness relative to group M isolates in a multiple-cycle whole-virus assay in PBMCs (6). However, another study demonstrated that subtype C, which is the most common subtype globally, actually has reduced replication fitness compared to subtype B in PBMCs (6, 11). This finding is not consistent with the hypothesis that reduced replication fitness in cell culture correlates with reduced transmission efficiency. Of note is that some studies suggested that replication in dendritic cells may be a better predictor of transmission efficiency (50); no differences between the replication of subtype B and C isolates were observed in dendritic cells, consistent with this theory (11). The authors of that study argued that the reduced replication fitness in PBMCs correlates with reduced pathogenicity rather than transmission efficiency (see below). This property would theoretically lead to longer asymptomatic stages of clinical infection and increased opportunities for subtype C transmissions. Clearly, more studies are needed to resolve the question of whether reduced replication fitness in cell culture correlates with reduced transmission efficiency and whether replication in PBMCs is the best in vitro assay to predict transmission efficiency.
Drug-resistant mutants selected during therapy in chronically infected patients are often rapidly overgrown by less drug-resistant variants once therapy is discontinued. In contrast, once infection with a drug-resistant variant is established in a newly infected patient, it can persist as the dominant member of the HIV-1 quasispecies for several years in the absence of drug selection pressure (21, 22). This finding is likely due to the fact that initial infection is usually established by a single strain of HIV-1; therefore, there is presumably no preexisting more-fit, drug-sensitive mutant during primary infection that can rapidly emerge when drug selection is removed.
In contrast, significant evolution of HIV-1 at epitopes targeted by cytotoxic T-lymphocyte (CTL) responses has been observed as early as a few weeks after initial HIV-1 infection (90, 102). These CTL escape mutants are not as well recognized by the host immune response and indicate that HIV-specific cellular immune responses exert strong selective pressures on HIV-1 evolution. Development of escape mutants can also be associated with more rapid disease progression (73). The primary target of these immune responses appears to be the viral envelope (102); thus, studies focusing on pol would not detect these evolutionary changes. It is interesting that not all potential CTL epitopes recognized by the patient's immune response evolve, suggesting that CTL escape mutants at these epitopes may have a reduced selective advantage for other reasons (90, 102). The hypothesis that CTL escape mutants have reduced fitness is supported by studies of simian immunodeficiency virus in which cloned CTL escape mutants obtained from a macaque showed slowed growth in cell culture and reversion to the CTL-susceptible wild-type sequence after inoculation into macaques lacking the major histocompatibility class I determinants necessary for the recognition of the mutant epitopes (62). Studies of escape mutants in the Gag p24 protein demonstrated that these mutations do confer substantial reductions in replication fitness, as measured in a multiple-cycle cell culture assay (117, 134). It is interesting that these escape mutants, when they occur, are seen primarily in patients with HLA types that are associated with the successful control of HIV viremia. These p24 escape mutants are usually associated with compensatory mutations in gag (117). Thus, studies of replication fitness in patients with primary HIV-1 infection, particularly those who are not infected with drug-resistant virus or treated with antiretroviral therapy, need to consider the significant variation and changes in fitness that are occurring in gag and env.
A cross-sectional study of 191 acutely and recently (less than 1 year) HIV-1-infected patients found that viral replication fitness, using the Monogram RC assay, varied widely (13). Nineteen percent of isolates demonstrated genotypic evidence of resistance to either protease or reverse transcriptase inhibitors. There was a statistically significant decrease in fitness of HIV-1 isolates with mutations conferring resistance to protease inhibitors (P = 0.01) (Fig. (Fig.3).3). Although isolates with mutations conferring resistance to nucleoside analogs and NNRTIs tended to have lower replication fitness than drug-sensitive strains, these differences were not statistically significant (Fig. (Fig.3).3). Of interest is that only 6% of the variation in RC values was attributable to drug resistance mutations.
This study demonstrated a statistically significant inverse correlation between replication fitness and CD4+ T-cell count at the first patient evaluation, although the strength of the correlation was modest (Spearman's ρ of −0.29; P < 0.0001) (Fig. (Fig.4).4). Surprisingly, no correlation was observed between replication fitness and viral load at baseline (13). Single-step regression tree analysis demonstrated that a threshold RC value of 42% best predicted the baseline CD4+ T-cell count, although the 95% confidence intervals were large (12% to 93%). Using this threshold, patients infected with HIV-1 who had an RC value of ≤42% had an average CD4+ T-cell count of 663 cells/μl versus 512 cells/μl for viruses with an RC value of >42%.
A study of 243 acutely or recently infected patients beginning a combination antiretroviral regimen found no correlation between replication fitness measured by the Monogram RC assay and baseline viral load, baseline CD4+ T-cell count, or viral load responses to antiretroviral therapy (12). Low viral replication capacity at the time of initiation of therapy did predict improved CD4+ T-cell responses but only after 12 months of therapy (12). The reason for the delayed increase in the CD4+ T-cell count is not clear.
Some isolates obtained during early infection have been shown to have greater fitness than a wild-type reference strain, as measured in a whole-virus assay; these properties were not as pronounced in a recombinant-virus assay using protease and reverse transcriptase sequences (156). One concerning case was reported in which a patient was newly infected with a dual-tropic, multidrug-resistant HIV-1 isolate and had an unusually rapid progression to AIDS (113). A pol-recombinant virus derived from this patient replicated more efficiently than the wild type in the Monogram Biosciences RC assay (113). Intact biological clones isolated from this patient also replicated more rapidly than the wild type in growth kinetic assays and were highly cytopathic in human PBMCs (123a). Although such rapid clinical progression is uncommon, it has been reported; thus, a cause-and-effect relationship between the increased replication fitness of this isolate and accelerated clinical progression cannot be definitively established.
One study of HIV-1-infected patients with unusually prolonged durations of asymptomatic infection (“long-term nonprogressors”) found reduced viral replication fitness in some patients, as measured in multiple-cycle, whole-virus, parallel infections (29). However, patients in this study also demonstrated evidence of robust HIV-specific cellular immune responses and high titers of neutralizing antibodies that could account for their less severe disease (29). Another study found isolates with reduced HIV-1 replication fitness in four of seven long-term nonprogressors that persisted over the course of several years, as measured by a multiple-cycle, whole-virus, parallel infection assay (18). Replication fitness in this study also was associated with the plasma HIV-1 RNA concentration. Such defects in replication were not observed in patients with high viral load that did have clinical progression (18).
Another study, which used a multiple-cycle whole-virus growth competition assay in PBMCs also found a correlation between replication fitness and disease progression in a small number of patients (140). Two other studies demonstrated good correlations between HIV-1 replication fitness, as measured in multiple-cycle whole-virus parallel infections in PBMCs, and plasma HIV-1 RNA concentration before the initiation of antiretroviral therapy (27, 163). In one of those studies, patients with the CCR5Δ32 allele, known to have a beneficial impact on disease progression, or syncytium-inducing virus, which is associated with more rapid disease progression, were excluded so as to avoid confounding variables in their analysis (27). Those investigators were unable to find a correlation between CD8+-mediated cellular immune responses and plasma HIV-1 RNA concentrations, suggesting that fitness was the primary influence on the viral load set point (27). The association between fitness and disease progression was stronger for the whole-virus assay (r2 = 0.71; P < 0.001) (Fig. (Fig.5A)5A) than for the Monogram RC recombinant virus assay (r2 = 0.44; P = 0.019) (Fig. (Fig.5B),5B), although a significant correlation between viral load and fitness, as measured by the latter assay, was still observed (27).
A large hemophilia cohort was studied to evaluate the relationship between HIV-1 replication fitness, as measured using the Monogram RC assay, and surrogate markers of clinical outcome (37). Approximately half of the patients were on therapy, but only five had been treated with a potent combination antiretroviral regimen. Fitness was significantly correlated with viral load and inversely correlated with baseline CD4+ T-cell counts in the 128 patients who had these assays performed, although the strength of the association was only modest (R2 value of 0.189 and P value of 0.03 for viral load and R2 value of −0.199 and P value of 0.02 for CD4+ T-cell count) (Fig. 6a and b, respectively). There was also a trend for replication fitness to be inversely correlated with change in the CD4+ T-cell count over time when corrected for baseline viral load and CD4+ T-cell count, although the magnitude of the effect was modest (relative hazard = 1.07; P = 0.081). Patients whose virus had an RC value in the lowest quartile were also at greater risk for clinical progression to AIDS (Fig. (Fig.77).
A retrospective study of 10 patients using a multiple-cycle whole-virus growth competition assay in PBMCs showed that viral replication fitness increased over time (164). Increases in viral fitness were associated with increasing genetic diversity in env. Syncytium-inducing CXCR4-tropic (X4) viruses, which develop late in infection and are associated with a worse clinical outcome, had greater replication fitness than non-syncytium-inducing CCR5-tropic (R5) viruses. However, the increases in fitness that were observed over time were not due solely to a change in viral coreceptor utilization (164).
An important observation in HIV-1-infected patients failing their antiretroviral therapy was that discontinuing treatment led to further increases in plasma HIV-1 RNA concentrations and declines in CD4+ T-cell counts, indicating that these apparently failing regimens were still providing some clinical benefit (41). These changes in viral load and CD4+ T-cell count were also associated with a loss of drug resistance mutations and a gain in HIV-1 replication fitness, as measured by the Monogram single-cycle recombinant-virus RC assay, raising the question of whether selection for a more-fit variant led to rises in plasma viremia (41).
It is interesting that these rebounds in viremia and declines in CD4+ T-cell count appear to depend on the class of drug being interrupted. In a nonrandomized study, patients who selectively interrupted protease inhibitors or NNRTIs had stable viral loads through 24 weeks of observation, whereas those who discontinued nucleoside analogs experienced a rapid rise within 2 weeks of discontinuation (38). Approximately one-third of patients discontinuing protease inhibitors showed a loss of protease inhibitor-resistant variants 12 to 36 weeks after interruption; in these patients, viral fitness increased approximately twofold after the time of reversion (interquartile range, 1.5 to 4.3) (38). The loss of nucleoside resistance mutations was incomplete and involved primarily the lamivudine-resistant M184V mutation. However, the loss of M184V occurred well after the rise in viral load, suggesting that the residual antiviral activity of nucleosides, rather than an increase in viral fitness, was the cause of the viral rebound (38). Similar observations were made in nonrandomized studies of lamivudine or lamivudine-zidovudine interruption (28, 59). Selective interruption of the fusion inhibitor enfuvirtide leads to prompt but limited increases in viral load (~0.1 to 0.2 log10 copies/ml), indicating that this drug also has some limited efficacy despite the presence of drug-resistant variants (39). These increases in viral load were not temporally associated with the loss of resistance mutations, suggesting that improvements in viral fitness were not driving these viral rebounds.
Another study found that the fitness of virus (measured using whole-virus parallel infections in CD8-depleted PBMCs) before treatment interruption did correlate with the ability to spontaneously control viremia (P = 0.014) (163). In that study, fitness also correlated with the viral load set point before the initiation of therapy (r2 = 0.26; P = 0.02) but not the change in viral load (Fig. (Fig.8).8). Thus, the association between fitness and the ability to control viremia during treatment interruption is likely due to the association between fitness and pretherapy viral load levels. Those investigators also found that viral diversity, viral replication fitness, and neutralizing antibody activity also correlated with the ability to control viremia during treatment interruption, indicating that a number of confounding variables may complicate the assessment of the role of replication fitness in outcomes (91). It should be noted that structured treatment interruptions have not been shown to be beneficial clinically (16, 58, 150). These observations are interesting in that they provide additional support for a correlation between replication fitness and viral load in the absence of antiretroviral therapy.
One study of patients failing their antiretroviral regimens demonstrated that replication fitness, measured using the Monogram RC assay, correlated with plasma HIV-1 viral load (40). Viral replication capacity was also reduced in patients who had preserved CD4+ T-cell counts despite virologic failure on antiretroviral therapy compared to patients who experienced CD4+ T-cell declines (12% versus 22%; P = 0.04) (160). More studies are needed to determine to what extent replication fitness predicts outcome during treatment failure and whether an assay for replication fitness could be useful in predicting the viral load and CD4+ T-cell trajectories at the time of first virologic failure. If it was found useful, such a test might allow more judicious use of antiretroviral regimens and preserve future treatment options.
Based on the studies summarized above, it is difficult to conclude with certainty which type(s) of fitness assay would be best suited for future studies of clinical outcome. Certainly, it is essential to have an assay with high throughput before substantial sample sizes can be evaluated. This factor in part explains the dominance of the Monogram RC assay (a recombinant-virus single-cycle assay) in the literature evaluating the clinical significance of fitness and the smaller sample sizes of studies in which growth competition or whole-virus assays are used. Good correlations were found between whole-virus parallel assays or a whole-virus growth competition assay and clinical outcome in untreated patients, suggesting that whole-virus assays may potentially provide greater clinical predictive potential than recombinant-virus assays, assuming that the problem of low throughput could be addressed. This observation is consistent with the discussion above that regions other than pol and the 3′ end of gag contribute to HIV-1 replication fitness and are the subject of selective pressures during clinical infection. Thus, recombinant-virus assays may be more appropriate for treatment-experienced patients failing an antiretroviral regimen in which major selective forces are acting on pol. However, none of the assays studied has yet demonstrated a sufficiently robust association with clinical outcomes that would warrant routine clinical use at present.
In general, HIV-1 drug-resistant mutants reduce replication fitness, but different mutants can vary in their degree of fitness impairment. Fitness, along with the level of drug resistance, appears to influence the likelihood of a mutant emerging during therapy. Fitness is also often associated with how long a mutant persists in the absence of therapy. There appears to be a correlation between reduced HIV replication fitness and either clinical outcome or surrogates of clinical outcome (CD4+ T-cell counts and viral load), although not all studies found correlations of fitness with both viral load and CD4+ T-cell count. Those studies are retrospective and have not always accounted for other confounding variables that can influence outcome. In addition, the larger studies evaluating the clinical significance of replication fitness have been performed primarily using the Monogram RC assay, which measures contributions of the 3′ end of gag, protease, and part of reverse transcriptase only. Statistically significant clinical correlations have clearly been found using this assay, but the modest strength of the associations raises questions as to how an RC value from an individual patient would be interpreted. Further study is needed to determine whether measurement of fitness will be a useful monitoring tool in addition to CD4+ T-cell count, viral load, and resistance testing.
This work was supported in part by R01-AI-041387 and R01-AI-065217.
We thank Hulin Wu for his careful review of the section on mathematical approaches to quantifying replication fitness.