Gag Determinants of Fitness and Drug Susceptibility in Protease Inhibitor-Resistant Human Immunodeficiency Virus Type 1

Abstract

Mutations can accumulate in the protease and gag genes of human immunodeficiency virus in patients who fail therapy with protease inhibitor drugs. Mutations within protease, the drug target, have been extensively studied. Mutations in gag have been less well studied, mostly concentrating on cleavage sites. A retroviral vector system has been adapted to study full-length gag, protease, and reverse transcriptase genes from patient-derived viruses. Patient plasma-derived mutant full-length gag, protease, and gag-protease from a multidrug-resistant virus were studied. Mutant protease alone led to a 95% drop in replication capacity that was completely rescued by coexpressing the full-length coevolved mutant gag gene. Cleavage site mutations have been shown to improve the replication capacity of mutated protease. Strikingly, in this study, the matrix region and part of the capsid region from the coevolved mutant gag gene were sufficient to achieve full recovery of replication capacity due to the mutant protease, without cleavage site mutations. The same region of gag from a second, unrelated, multidrug-resistant clinical isolate also rescued the replication capacity of the original mutant protease, suggesting a common mechanism that evolves with resistance to protease inhibitors. Mutant gag alone conferred reduced susceptibility to all protease inhibitors and acted synergistically when linked to mutant protease. The matrix region and partial capsid region of gag sufficient to rescue replication capacity also conferred resistance to protease inhibitors. Thus, the amino terminus of Gag has a previously unidentified and important function in protease inhibitor susceptibility and replication capacity.

The development of antiretroviral drugs and their use in highly active antiretroviral therapy have led to the ability to effectively control human immunodeficiency virus (HIV) replication in infected patients. The emergence of resistance to these drugs, while reduced when the drugs are used in combination, remains a problem. A recent study has estimated that by 10 years of treatment, nearly 10% of patients will have experienced resistance to the three main classes of drugs that form the basis of highly active antiretroviral therapy. Further, the risk of death within 5 years after extensive triple-class failure is 10% (31). Resistance develops due to several factors, including the high error rate of reverse transcriptase (RT), viral recombination, high viral turnover, and suboptimal drug levels in patients. Mutations emerging in viruses can lead to cross-class resistance.

Resistance to protease inhibitors (PI) occurs by the stepwise accumulation of mutations in protease itself, many of which lead to a reduced replication capacity (RC) of the virus (5-7, 22, 23, 37). Primary, or major, mutations mostly lead to a reduced affinity for the inhibitor through alterations in the enzyme active site. Further secondary, or minor, mutations also arise in protease that have little effect on inhibitor binding and therefore do not convey resistance alone but enhance the activity of the mutant protease (4, 27). Resistance mutations also reduce the RC of the virus due to decreased binding of the natural substrate, Gag. Gag mutations that partially restore the RC are therefore selected both at cleavage sites and elsewhere (9, 12, 24, 35, 39). Gag cleavage site mutations (CSM) have also been shown to confer resistance in the absence of protease mutations, although the clinical significance of this remains unclear (28). The assumption has been that viruses resistant to PI, especially multi- or cross-resistant viruses, have a significantly reduced RC in the absence of drug (7, 22).

Several assay systems have been developed for studying resistance mutations of HIV. These are based on generating a virus containing RT and protease genes with mutations created by site-directed mutagenesis, selected in vitro or in vivo, including those from clinically derived viruses (3, 13, 20, 30, 33, 38). Most of these include only partial gag from the virus of interest typically including gag-protease and RT downstream of an ApaI restriction enzyme site within the nucleocapsid (NC or p7) coding region of gag. The regions included downstream of this ApaI site, specifically, the NC(p7)/p1 and p1/p6 cleavage sites, have been shown in several studies to be linked to PI resistance and RC (9, 14, 20, 21, 28, 34, 35, 39). However, mutations other than Gag CSM and upstream of the ApaI site have also been shown to have an effect in the context of PI resistance mutations (12, 24).

We speculated that Gag mutations outside of the commonly studied CSM might impact on PI resistance and/or RC. We therefore studied full-length gag and protease genes from a clinical isolate with multidrug resistance in order to dissect the more precise impact of gag on drug-resistant protease.

MATERIALS AND METHODS

Clinical HIV samples.

HIV-1-positive plasma samples were chosen after routine resistance testing had identified multiple mutations in protease. Independent PCRs were carried out, and different PCR primers were used to remove any bias introduced from one set of primers. Sequences obtained from these PCR products generated the same sequence. Sequences of multiple clones of PCR products were also the same. Thus, the sequences are representative of the majority virus population.

Resistance vectors.

A retroviral vector system was adapted to study viral RC and drug susceptibility. Briefly, p8.91Ex (15), a modified form of gag-pol expression plasmid pCMV-Δ8.91 (25) that includes a NotI site upstream of gag and unique ApaI and SpeI sites in gag, was further modified to include a unique XmaI site in integrase (bp 4403 of HXB2), creating p8.9NSX, into which gag-protease and RT sequences can be cloned.

The gag, protease, and gag-protease of plasma-derived virus from a highly drug-experienced patient had previously been cloned into a virus vector system, pNL4-3XPSX (17). These were transferred to p8.9NSX by PCR with primers GagNot+ (GCGGCGGCCGCAAGGAGAGAGATGGGTGCG [gag start codon in italics]) and PolXmaR (CATATCCCGGGACTACAGTCTAC) by introducing NotI and XmaI restrictions sites (underlined) and using standard molecular biology techniques.

Resistance test vectors were prepared by transfection of confluent HEK293T cells in a 10-cm plate with 18 μl of Fugene-6 (Roche) in 200 μl of OptiMEM (Invitrogen), 1 μg of pMDG encoding the vesicular stomatitis virus G protein, 1 μg of p8.9NSX-derived gag-pol expression vector, and 1.5 μg of retroviral expression vector pHR-SIN-CSGW (green fluorescent protein [GFP]) or pCSFLW (luciferase) as described previously (1, 2, 36). Virus-containing supernatant was collected at 48 and 72 h posttransfection.

RC.

The RC of resistance test vectors was determined by titration of serial virus dilutions on 293T cells at 48 h after infection. Infection was determined by fluorescence-activated cell sorting for resistance test vectors encoding GFP. Titers were determined at multiplicities of infection of less than 0.2 and within the linear range. Determination of infection of resistance test vectors encoding luciferase was done with Bright-Glo or Steady-Glo (Promega) and a GloMax96 luminometer (Promega) by following the manufacturer's recommendations. Titers were determined by using values that showed a linear relationship to the inoculum.

p24 ELISA.

A p24 enzyme-linked immunosorbent assay (ELISA) was performed as described previously (29), with reagents supplied by Aalto BioReagents Ltd. Resistance vector-containing supernatants were inactivated with Empigen (Sigma) (1% by volume) and heat inactivation at 56°C for 30 min before being titrated against known concentrations of HIV-1 p24 in an ELISA quantified with the AMPAK Detection System (Oxoid Ltd.) by following the manufacturer's recommendations.

Drug susceptibility.

HEK293T cells were transfected as described above; 16 h later, cells were seeded in the presence of serial PI dilutions. Resistance vectors were harvested 24 h later and used to infect fresh target 293T cells. Replication was determined by measuring the luciferase expression of infected target cells at 48 h postinfection and expressed relative to that of no-drug controls. Fifty percent inhibitory concentrations (IC₅₀s) were determined by linear regression analysis. Results are expressed as n-fold changes in the IC₅₀ compared to wild-type (WT) subtype B (p8.9NXS) and are the mean of at least two separate experiments.

Nucleotide sequence accession numbers.

The nucleotide sequences determined and used in this study have been submitted to GenBank and assigned accession numbers FJ224363 and FJ224364.

RESULTS

Patient-derived virus.

To study the relationship between Gag and protease in the context of multidrug resistance, a patient (patient 1) was identified who had experienced extensive antiretroviral therapy. In total, patient 1 had been treated with four PI (indinavir [IDV], lopinavir [LPV], ritonavir, and saquinavir [SQV]) and six RT inhibitors (didanosine, lamivudine, stavudine, zidovudine, efavirenz, and nevirapine) and was failing therapy comprising ritonavir-boosted LPV, didanosine, lamivudine, and nevirapine at the time that the plasma sample was obtained. In order to eliminate the effect of RT mutations, only gag and protease from the patient-derived virus were studied (GenBank accession number FJ224363). Population-based sequencing indicated that this virus had four major protease mutations, L33F, M46I, I54V, and V82A; four minor resistance mutations, L10V, L24I, K55R, and L63P; and six further mutations, including a two-amino-acid insertion, M36QNL (16). The Stanford HIVdb Genotypic Resistance Interpretation Algorithm (http://hivdb.stanford.edu/pages/algs/HIVdb.html) predicts high-level resistance to IDV, LPV, and nelfinavir (NFV); intermediate resistance to atazanavir (ATV), darunavir (DRV), amprenavir (APV), and tipranavir (TPV); and low-level resistance to SQV.

The patient-derived viral Gag protein also has many mutations compared to the consensus HXB2 protein, both at cleavage sites and elsewhere (Fig. 1). The p2-p7 cleavage site has four amino acid changes, although this site has been reported to be the most variable (19). Mutations A431V at the p7/p1 cleavage site and L449V at the p1/p6 cleavage site, which have been previously associated with resistance, are also present (20, 34, 39). Within the 132 amino acids of the matrix protein (p17) are 10 changes, as well as a 2-amino-acid insertion (Q116TQ), in a region that is variable in length across HIV-1 subtypes. There are also numerous mutations in p6. Notably, the PTAPP motif (HXB2 amino acids 455 to 459), which interacts with the cellular ESCRT pathway, resulting in virus budding at the cell membrane, is PSAPP; however, this variant retains the interaction with the ESCRT pathway (10).

FIG. 1.

Schematic representation of HIV Gag protein showing amino acid changes found in patients 1 and 2 (Pt1 and Pt2, respectively). Functional Gag matrix (p17), capsid (p24), p2, nucleocapsid (p7), p1, and p6 proteins are shown with protease cleavage sites indicated. Changes are numbered according to the HXB2 consensus. Changes found in patient 1 are shown above with patient 2 changes up to the SpeI site shown below. The positions of the SpeI and ApaI sites are indicated. Changes within Gag cleavage sites are underlined. Changes identical in patients 1 and 2 are in solid boxes; changes at the same position but to different amino acids are in dotted boxes.

Contribution of Gag to RC.

The gag, protease, and gag-protease portions of the patient 1-derived virus (hereafter termed mutant) were cloned into the gag-pol expression vector p8.9NSX, creating p8Pt1gag, p8Pt1pro, and p8Pt1gagpro, respectively. A panel of resistance vectors encoding GFP (as a marker of infection) was made incorporating either the mutant protease alone, mutant gag alone, or mutant gag-protease together (Fig. 2A). RC was determined by serial dilution, and the number of infectious units per milliliter was calculated. Incorporation of the mutant protease (with WT gag) leads to a significant drop in infectivity, down to 5% of the WT level (Fig. 2B). Combining the mutant protease gene with the mutant gag gene leads to complete recovery of RC to WT levels. Mutant gag alone, with WT protease, has an RC similar to that of the WT, indicating that this is an efficient gag gene with either mutant (PI-resistant) or WT protease (Fig. 2B). In order to reduce any possible effect of transfection efficiency, the RC was also determined in terms of infectious units per nanogram of p24 after quantifying p24 levels by ELISA. Figure 2C shows that expressing the results as infectious units per nanogram of p24 does not significantly alter the results shown in Fig. 2B. Western blot analysis with an anti-p24 antibody did not show a difference between the relative amounts of p24 and p55 in these four viruses (see Fig. S1 in the supplemental material). Thus, the 95% drop in infectivity seen with the protease alone is not due to an inability of the protease to cleave Gag.

FIG. 2.

RC of patient-derived mutant Gag, protease, and Gag-protease with schematic representations. Recombinant resistance vectors containing patient-derived, mutant viral protease, gag-protease, and gag were prepared by cotransfection of 293T cells with three plasmids required to produce infectious particles encoding GFP. Titrations were determined by detection of infected cells expressing GFP by fluorescence-activated cell sorting. Error bars show the standard error of the mean of at least three different dilutions from the same titration. (A) Names of the gag-protease constructs and schematic representations. Protease cleavage sites are shown, and cleavage sites and the p1 and p2 spacer peptides on either side of p7 are depicted by heavy lines. Functional Gag matrix (p17), capsid (p24), nucleocapsid (p7), p6, and protease (Pro) proteins are listed (p1 and p2 are not shown). The WT is shown as white, and the mutant is shown as gray. (B) RC of resistance vectors as infectious units/ml. (C) RC of resistance vectors as infectious units/ng p24 after the p24 concentration was determined by ELISA.

In order to determine the regions of gag responsible for this recovery in RC, chimeric mutant-WT gag constructs with the mutant protease were made with unique ApaI (in the p7 NC coding region [Fig. 1]) and SpeI (in the p24 capsid coding region [Fig. 1]) restriction enzyme sites, creating p8Pt1gpAS and p8Pt1gpSS, respectively (Fig. 3A). Linking p7/p1 and p1/p6 CSM (p8Pt1gpAS from the ApaI site in p7-NC onward) to the mutant protease leads to recovery of RC (Fig. 3B and C). Inclusion of the region to the SpeI site with a further nine amino acid changes from the HXB2 consensus (p8Pt1gpSS from the SpeI site in p24 onward [Fig. 3A]), including three changes at the p2/p7 cleavage site, also rescues RC back to WT levels (Fig. 3B and C).

FIG. 3.

RC of resistance vectors showing the contributions of different regions of gag. Recombinant resistance vectors containing patient-derived, mutant viral protease, gag, and chimeric WT-mutant gag were prepared as described in the legend to Fig. 2. Error bars show the standard error of the mean. (A) Names and schematic representation of constructs as described in the legend to Fig. 2. The positions of the NotI (N), SpeI (S), and ApaI (A) sites are shown. (B) RC of resistance vectors as infectious units/ml. (C) RC of resistance vectors as infectious units/ng p24.

The patient-derived mutant gag-encoded protein contained a number of amino acid changes in the matrix region (Fig. 1). To investigate the impact of these mutations, two further constructs were made. The NotI-to-ApaI and NotI-to-SpeI fragments of mutant gag were cloned into p8Pt1pro, giving p8Pt1gpNA and p8Pt1gpNS, respectively (Fig. 3A). Both constructs rescued the RC of the mutant protease back to WT levels (Fig. 3B and C). Construct p8Pt1gpNS rescued RC without CSM since the p17/p24 cleavage site contains no changes compared to the HXB2 consensus. In order to ensure that the differences observed were due to effects of the mutations on RC, rather than differences in transfection efficiency, the amount of p24 produced in the supernatants was determined by ELISA. As with Fig. 2, expressing RC as infectious units per nanogram of p24 does not change the overall interpretation of the results (Fig. 3C).

All of the gag mutations included in p8Pt1gpNS can be found in consensus WT sequences if all of the subtypes are considered (18). Some of these mutations are also found in molecular clones MJ4 (subtype C) and 94UG114.1.6 (subtype D) (11, 26). MJ4 has the Y79F change in the matrix and the S173T change in the capsid; 94UG114.1.6 has the K30R, K76R, H124N, and N126S changes in the matrix; and both MJ4 and 94UG114.1.6 have the I94V change in the matrix. The NotI-to-SpeI region of gag, encoding the matrix region and part of the capsid region, from MJ4 and 94UG114.1.6, was therefore cloned into p8Pt1pro to determine if the mutations common to these had a role in the rescue of RC (8CPt1pro and 8DPt1pro, respectively, Fig. 4A). However, neither was able to significantly recover the reduced RC of mutant protease (Fig. 4B and C). Adjusting the RC for levels of p24 shows that the WT subtype D and B matrix and partial capsid have almost the same RC when linked to mutant protease. Subtype C matrix and partial capsid increased the RC of the mutant protease by threefold, suggesting that amino acid differences between them may have an effect on RC, although far less than the effect seen with the mutant matrix and partial capsid (Fig. 4C).

FIG. 4.

RC of resistance vectors showing the contributions of different matrix and partial capsid regions to the RC of mutant protease. Recombinant resistance vectors were prepared containing patient-derived mutant viral protease and chimeric gag with the matrix and partial capsid regions from either WT subtype C or D HIV-1 or a second, unrelated multidrug-resistant patient-derived virus. (A and D) Names and schematic representations of constructs as described in the legends to Fig. 2 and 3. WT subtypes C and D are shown in lighter gray with the relevant subtype indicated (panel A). Mutant matrix and partial capsid from a second, unrelated, multidrug-resistant, patient-derived virus are shown in black and labeled 2 (panel D). (B and C) Retroviral test vectors were produced encoding GFP as described in the legend to Fig. 2. (B) RC of resistance vectors as determined by infectious units (i.u.)/ml. (C) RC of resistance vectors as determined by infectious units/ng p24. (E and F) Retroviral test vectors were produced encoding luciferase and titrated. Luciferase activity was determined with SteadyGlo and a GloMax luminometer (both Promega), and a mean was determined by using at least four values within the linear range. (E) RC expressed as relative light units (rlu)/μl. (F) RC expressed as relative light units/ng p24. Error bars show the standard error of the mean. RC is shown as a percentage of WT RC.

In order to determine if a common mechanism that recovers the RC of PI-resistant protease mutations can be mapped to the matrix and partial capsid regions, this region from another multi-PI-resistant patient-derived virus (patient 2, GenBank accession number FJ224364) was linked to the mutant protease (8Pt2Pt1Pro, Fig. 4D). Figure 1 shows the amino acid changes in the matrix and capsid to the SpeI site from this second virus, compared to HXB2. This second sample has the same R76K, T84V, I94V, and N126S mutations as the original mutant (solid line boxed in Fig. 1). Patient 2 also has K30Q rather than K30R and H124S rather than H124N in patient 1 (dotted line boxed in Fig. 1). Figure 4E and F show that significant recovery is achieved from the matrix and partial capsid from the patient 2 subtype B multi-PI-resistant sequence, which is otherwise unrelated to the mutant protease, suggesting that a common mechanism is involved. When corrected for levels of p24, the RC of the mutant protease increases from 3% to 42% of the WT level (a 14-fold increase in RC) by the coexpression of the matrix and partial capsid from patient 2 (Fig. 4F).

The matrix and partial capsid alone for both of the mutant viruses (Pt1 and Pt2, from patients 1 and 2, respectively) studied do not compromise RC in a WT background (data not shown). We therefore conclude that the virus can tolerate mutations in this region, whether coexpressed with mutant or WT protease.

Antiretroviral drug susceptibility.

The susceptibility of the retroviral vectors containing mutant Gag, protease, and Gag-protease was determined (Fig. 5). Expressing mutant gag alone with a WT protease results in an increase in IC₅₀ for all of the PI (Fig. 5A). The change compared to the WT varies from 3.7-fold for TPV to 16.5-fold for IDV. Both APV and ATV show a change of greater than 10-fold, more than seven times (APV) and four times (ATV) the first clinical cutoff (CCO1) used by Virco to define resistance (www.vircolab.com). For LPV, a change in the IC₅₀ of 6.5-fold is nearly 4 times the Virco biological cutoff and over the CCO1 of 6.1-fold. All of the changes determined for p8Pt1gag are more than two times the Virco biological cutoffs, highlighting the degree of cross-resistance in this patient-derived virus conferred by Gag alone.

FIG. 5.

PI susceptibility of mutant Gag, protease (Pro), and Gag-protease constructs. Susceptibility is shown as the change in IC₅₀ compared to that of WT p.8.9NSX; thus, a value of 1 indicates the same susceptibility as the WT. The names and schematic of constructs are shown above each bar chart; the WT is shown as white, and the mutant is shown as gray. Changes are shown within each bar. Error bars represent the standard error of the mean from at least two separate experiments. (A) PI susceptibility of mutant Gag plotted against a linear scale. (B) PI susceptibility of mutant protease plotted against a log scale. (C) PI susceptibility of mutant Gag-protease plotted against a log scale.

As expected, given the known resistance mutations present, mutant protease alone gives high-level resistance to IDV and LPV of more than 200 times the IC₅₀ of the WT, as well as significant resistance to APV (124-fold) and NFV (93-fold), all well above (4 times or more) both the Virco second and Monogram upper clinical cutoffs (www.monogramhiv.com). The protease alone shows some resistance to ATV, 5.3-fold being more than twice the Virco biological cutoff (2.4-fold) and CCO1 (2.5-fold); while for DRV, 7-fold is more than the Virco biological cutoff (2.4-fold) but less than CCO1 (10-fold). Protease only remains susceptible to SQV but is hypersusceptible to TPV (Fig. 5B).

Combining the mutant Gag and protease sequences together results in an increase in the IC₅₀ compared to that of the WT, Gag alone, or protease alone, suggesting a synergistic effect between the enzyme and substrate that have coevolved (Fig. 5). The exception is TPV; linking Gag and protease reverses the hypersusceptibility observed for protease alone (Fig. 5C). It is not possible to determine the impact of linking Gag and protease together for LPV and IDV, since protease alone gives a greater-than-200-fold change in the IC₅₀.

The antiretroviral drug susceptibilities conferred by the matrix and partial capsid regions of Gag capable of rescuing RC (8Pt1gNS) and the remainder of mutant Gag (from the SpeI site forward, 8Pt1gS>) were also determined, both with WT protease (Fig. 6A and B). As observed with full-length Gag alone, both regions confer resistance; however, other than SQV and TPV, the matrix and partial capsid regions give greater resistance than the remainder of Gag, which includes altered cleavage sites. More than half of the reduced susceptibility to APV, ATV, and IDV conferred by Gag can be contributed by the matrix region and part of the capsid region alone. It is interesting that the sum of the changes from the different regions of Gag is close to the change for the whole of Gag for many of the PI drugs (APV, DRV, IDV, LPV, and TPV).

FIG. 6.

PI susceptibility of mutant-derived Gag regions. Susceptibility and PI are shown as described in the legend to Fig. 5. (A) PI susceptibility of mutant matrix and partial capsid, p8Pt1gNS (NotI to SpeI). (B) PI susceptibility of mutant partial capsid to the end of Gag, p8Pt1gS> (SpeI site onward). Pro, protease.

DISCUSSION

We have demonstrated that the gag and protease genes from an HIV-1 isolate derived from a highly drug-experienced patient have coevolved to be highly resistant to PI drugs, with no measurable loss of RC. PI-resistant viruses with reduced RC have been frequently described; compensatory mutations accumulate in protease itself, within the Gag cleavage sites, and elsewhere in Gag, allowing partial recovery of RC, although this recovery is rarely to the same level as in the WT virus (7, 9, 12, 14, 20, 22, 24, 27, 35, 37, 39). We confirm the fitness deficit conferred by key protease resistance mutations, down to 5%. Our observation of complete recovery of RC by full-length Gag, in addition to further reducing PI susceptibility, suggests that analysis of full-length Gag with protease is required to fully understand the biology and impact of PI drug resistance. While this work was undergoing review and modification, Dam et al. described similar effects of full-length Gag contributing to PI resistance, as well as compensating for fitness loss. Our results therefore support those of Dam et al. with respect to full-length Gag (8).

Other groups have studied PI resistance in the context of p7/p1/p6, often using the same ApaI site used here (9, 20, 30, 39). We find that inclusion of this region of Gag leads to significant recovery of RC compared to protease alone. Inclusion of further regions of Gag has some impact, possibly due to the four mutations at the p2/p7 cleavage site. This cleavage site is, however, known to be the most variable, so changes here may simply be polymorphic rather than selected under drug pressure (19). This is in line with previous work and in keeping with the hypothesis that an important factor leading to increased RC are CSM (9, 20, 34, 39). Non-CSM have also been linked to RC of protease-resistant viruses, although the mechanism remains unclear (12, 24). However, in addition to these studies and in contrast to Dam et al. (8), our results demonstrate that the amino-terminal region of the Gag polyprotein can play a key role in RC and drug resistance, since mutant matrix and partial capsid regions can recover the RC of the mutant protease to WT levels, as well as lead to reduced PI susceptibility in the absence of mutations in protease. Thus, both the amino- and carboxy-terminal regions of Gag can lead to recovery of RC due to PI resistance mutations in protease. This suggests that multiple pathways can compensate fitness and mutations at each end of gag will be selected as PI resistance evolves. Recombination within patients is therefore likely to be an important factor in the pathogenesis of PI-resistant viruses.

The mechanism by which mutations in the matrix region and part of the capsid region can rescue RC and confer PI resistance remains unclear; however, it seems to be conserved between isolates. The corresponding region of a second highly drug-resistant but unrelated subtype B virus that shares four mutations with the original mutant also leads to an increase of 14-fold in the RC of the mutant protease, although it does not reach WT levels. This recovery is greater than that conferred by the same region from the WT subtype C or D virus, despite the fact that these share some of the changes found in the matrix and partial capsid of the original mutant. This suggests that changes to the Gag proteins are selected during the evolution of the multi-PI-resistant viruses and are likely to have a structural effect that is functionally related in two otherwise unrelated viruses. These sequences do not have changes to their cleavage sites; thus, the mechanism is independent of CSM. Identifying the specific amino acid changes that are responsible for the recovery of RC will help explore the mechanisms responsible for the effects seen here.

The level of resistance achieved by full-length Gag is more than the Virco biological cutoff for all drugs. This biological cutoff represents a level below which drug susceptibility can be considered normal for treatment-naïve clinical isolates with 97.5% confidence (www.vircolab.com); therefore, Gag does contribute to PI resistance, as also shown by Dam et al. (8). We also show that different regions of Gag contribute to PI resistance (without mutations in the protease itself). The matrix and partial capsid regions (without CSM) confer more resistance to APV, ATV, DRV, IDV, LPV, and NFV than the protease-proximal regions that contain CSM. For many of the PI drugs, the sum of the change for each half of Gag is approximately equal to the total change for full-length Gag, suggesting that the mechanisms for the different regions are independent of each other. In contrast, the coevolved Gag protein and PI-resistant protease act synergistically, suggesting that they function together, as has been shown for the A431V CSM. The A431V CSM has been shown to function due to increased contact of the cleavage site with the enzyme, but only in the presence of resistance mutations in protease (32). However, our results and those of others suggest that CSM can function as PI resistance mutations in the absence of protease mutations, both after in vitro selection (28) and in clinical isolates (our results and those of Dam et al. [8]).

While this work is based on detailed study of one clinical isolate, it demonstrates that Gag-protease should be considered a single functional unit and therefore changes to escape the action of PI and influence RC may be found throughout the whole of Gag-protease. Such seemingly unconnected changes, separated by several hundred base pairs on the gene, have effects on the protein and may indeed be close in the quaternary structure. The net effect is the production of infectious virus at levels similar to those of the WT. Further studies of full-length Gag, as well as regions of Gag, and protease interaction are warranted.