Resistance and cross-resistance with saquinavir and other HIV protease inhibitors

Introduction The addition of the HIV protease inhibitors to the growing list of antiretrovirals, and the increasing acceptance of polypharmacy in the treatment of HIV infection, has resulted in a marked improvement in clinical response to therapy such that commentators are optimistically suggesting that HIV infection may soon be regarded as a treatable chronic infection. Indeed, with 10 or more drugs now available or imminent, several consecutive double or triple therapy regimens can be used, employing a completely new set of agents at each change. This strategy could potentially provide several years of successful therapy, during which time additional drugs may well become available to provide ongoing treatment. In order to optimize the sequential use of the available drugs, it is important to sequence them in such a way as to minimize the risk of drugs used early in treatment restricting the subsequent use of the related drugs by the induction of cross-resistance. Thus, an understanding of the factors involved in the emergence of viral drug resistance and cross-resistance will help the design of individual treatment regimens with potential long-term value. In this review the general principles of viral drug resistance will be discussed. This will then be exemplified through our current understanding of resistance and cross-resistance among the inhibitors of HIV protease with particular emphasis on saquinavir (SQV). Principles of viral drug resistance Effects of viral replication and selection pressure When a virus has its replication restricted by the presence of an inhibitor of one of the essential life-cycle processes, then any chance variant of the virus that is less sensitive to the effects of the inhibitor will tend to outgrow the sensitive wild-type and hence create a ‘resistant’ mutant virus strain. The point of mutation is usually, although not always, in the target enzyme or receptor. Thus, resistance mutation is a method which the virus uses to overcome attempts to stop its replication. Clearly, two factors are essential to generate ‘resistance’: (i) the virus must be actively replicating so that a specific mutation can be generated and allowed to outgrow the wild-type; and (ii) the antiviral agent must be present in sufficient concentration to exert inhibitory pressure on the virus to allow selective replication of the ‘resistant’ mutant (i.e., there must be ‘selection pressure’). The word ‘resistant’ must be interpreted with caution. A mutant organism that is only twofold less sensitive to the drug than wild-type may, over time, outgrow the wild-type. Hence, resistance is not absolute, but involves a degree of loss of drug sensitivity, which may be modest and of uncertain clinical relevance. Thus, ‘decreased drug sensitivity’ is more appropriate than ‘resistance’, but the latter will be used here for brevity. There are two effects of increasing the concentration of an antiviral agent pertinent to resistance (Fig. 1a). Increased antiviral drug concentration will (i) increase selection pressure, thereby enhancing the chance of selecting resistant virus, and (ii) decrease the viral replication rate, thereby reducing the chance of selecting resistant virus. Thus the two effects work in opposition. The chance of selecting resistant virus is zero at zero drug concentration (no selection pressure), and theoretically zero at infinite drug concentration (no viral replication). Between these two limits the chance of selecting for resistance is finite, as indicated by the curve in Fig. 1. The shape of this curve is unknown and meaningless until the scales on the axes are unequivocally defined (often depicted as bell-shaped). It is seen that very high drug concentrations are ideal both to reduce viral replication and to minimize the chance of resistance. However, from relatively low values an increase in drug concentration could result in an increase in the chance of resistance. The practical way to achieve very low viral turnover without increasing selection pressure is to use drugs in combination (Fig. 1b). At zero concentration of study drug there will be reduced viral replication (load) due to the antiviral effect of the other component(s) in the combination. Their contribution will also bring viral replication to zero at a lower (finite) concentration of study drug. The chance of resistance is now modified to curve 2 and at a given concentration the chance of resistance emerging is reduced from X to Y (Fig. 1b).Fig. 1: . Chance of emergence of resistance (qualitative considerations). (a) Decrease in viral turnover and increase in selection pressure as drug concentration increases. The axes are without scales; thus, there is no implication that the responses are linear or equivalent. A curve is added indicating the chance of emergence of resistance. Again because there is no scale, no specific shape is implied (other than the zero points at zero and infinite drug concentration). (b) Additional curves that would apply if the drug under examination were added to a fixed concentration of another antiviral agent. This second drug would lower the viral turnover throughout the dose–response of the test drug and hence lower the chance of resistance emerging.Effects of viral fitness and degree of resistance Two further factors are important in determining the chance of the emergence of viral resistance: (i) the replication ‘fitness’ of the mutant virus — the less fit the virus is, the less readily it will compete with and outgrow the wild-type virus (all mutants must be to some degree less fit than the wild-type or they would be the wild-type); and (ii) the degree of resistance achieved by the mutation — the greater the advantage the virus gains in terms of reduced drug sensitivity the more readily it will compete with wild-type in the presence of the antiviral drug. The virus is more likely to evolve a reduced ability to replicate if the trade-off is a more marked reduction in drug sensitivity. These factors will alter the shape of the curve (Fig. 1a; for fixed scales) affecting the rate at which resistance will arise for a given viral turnover (load) and drug concentration. Both of these factors are specific, not only for each drug but also for each mutation, because each single mutation will have its particular effect on both viral fitness and reduced drug sensitivity. Thus, in cases where a number of mutations arise in response to one drug, the curve (Fig. 1a) will be distinct for each mutation. However, mutations may have interdependent effects on viral fitness and hence the curves may also be interdependent. Viral fitness: nature's experiment HIV reverse transcriptase is unable to correct errors. Its error rate is 10−4 to 10−5 per base per cycle 1. When this is coupled to an estimated viral turnover in an untreated HIV-infected individual of 109 or more virions per day 2,3, then it can been estimated 4 that every possible point mutation in HIV is produced 104to 105 times per day in each HIV-infected individual. Thus, any mutation that has a comparable replication fitness to the wild-type virus will exist alongside the wild-type virus creating a number of variants within the wild-type population. Conversely, mutations that substantially disable the virus will not accumulate and will be absent in the wild-type population. Thus, the presence of a particular mutation as a variant in the wild-type population will indicate the relative ease with which it can be selected by drug pressure. Recent studies have allowed us to access the results of ‘nature's experiment’ as it pertains to the fitness of variants in HIV protease. During 1996, eight reports were published 5–12 giving sequence data on the HIV protease gene in proviral DNA or viral RNA from individuals who had not been previously exposed to protease inhibitors. Collectively, the data provide a substantial database on samples from over 200 patients and well over 1000 protease sequences. A selection of the data pertaining to the variants of importance to protease inhibitor resistance is shown in Table 1.Table 1: . Incidence (%) of variants in wild-type HIV protease.The key mutations of resistance to SQV, V48 and M90, have zero incidence in the wild-type population, and the key resistance mutations to nelfinavir and 141W94 (also termed VX-478), N30 and V50, respectively 13,14, have very low incidence. The incidence of A82, the primary mutation for resistance to ritonavir 10 is also very low, although other non-consensus amino acids are seen at this position. Of the second and third phase mutations in the ordered accumulation of resistance to ritonavir 10 some are absent (R20, V54, V84), whereas others (I36, I46, V/T71) represent a significant proportion of the wild-type population. The emergence of resistance to indinavir is unusual in that it involves variable patterns of multiple substitutions amongst at least 11 amino-acid residues 11. No single mutation is present in all resistant isolates, although I/L46 or A/F/T82, or both, appear to be the most common. Both of these are found to some degree in the wild-type population. Of the other nine mutations, I/V10, P63, V/M64 and V/T71 all appear as variants in wild-type populations, whereas R20, I24, V54, V84 and M90 do not. Any three or four mutations can be selected to give resistance. Thus, there is the potential for resistance to arise relatively readily with the possibility of one or more mutations 7 already being present at detectable levels in pretreatment virus. Resistance: degree of loss of sensitivity As described above, the trade-off for selecting a poorly replicating mutant is the degree to which drug sensitivity is lost. For SQV the emergence of resistance in vitro is always accompanied by an initial G48V mutation giving about eightfold loss in drug sensitivity 15–17. This is followed by L90M, affording a further loss of about fivefold (i.e., 40-fold overall) 15,17. A single M90 mutation cloned into an NL4-3 or HXB2 background gives less than twofold loss in sensitivity 10,18. In the clinic, L90M is the predominant mutation in response to SQV treatment 5,19–22; the G48V mutant and the double mutant are rare. A recent phenotypic study of about 30 paired (baseline and posttreatment) isolates from SQV-treated patients showed that, consistent with in vitro findings, the loss in drug sensitivity resulting from L90M with or without G48V was modest (Fig. 2). The median inhibitory concentration for SQV rose from 7.5 nmol/l pretreatment to 30 nmol/l at 48–56 weeks post-treatment. Thus, for both L90M and G48V mutations, the advantage gained by the virus in terms of decreased drug sensitivity is modest. Hence, there may be little incentive for the virus to make the changes and adopt the resultant disadvantage for replication fitness, which would seem to be particularly true for G48V. Despite the relatively greater decrease in drug sensitivity associated with this change, its incidence in the clinic is lower than L90M suggesting the G48V mutation is more disabling to the virus (Table 2).Table 2: . Incidence of resistance mutation in HIV protease from patients treated with saquinavir (SQV) in study NV14256.Fig. 2: . The effect of L90M and G48V mutations on viral sensitivity to saquinavir (SQV). Matched pairs (n = 43) of viral isolates were taken pretreatment and then either at 24 or 40–56 weeks from patients treated with SQV monotherapy or SQV plus zalcitabine in study NV14256. (The selection of patient isolates for this phenotypic analysis was biased towards those carrying mutations at positions 48 and 90.) Fold change in median inhibitory concentration (IC50) for virus sensitivity to SQV is shown relative to each individual pretreatment isolate (baseline).Modest decrease in viral sensitivity to SQV and the necessity for disadvantageous mutations are consistent with the relatively low incidence of resistance observed with SQV in vivo. However, the relative ease with which G48V appears in vitro cautions against over-interpretation of in vitro data on resistance selection. There are clearly more challenges to viral replication in vivo than in vitro, and the drug concentrations that can be achieved and constantly maintained in vitro may be well in excess of those found in vivo. Thus, the critical question on the chance that resistance will emerge with a particular drug is not how many mutations are needed, but how well the resulting virus replicates. The critical question after emergence of resistance mutations is the degree of loss of drug sensitivity and whether it will affect clinical outcome. A modest decrease in drug sensitivity may well leave the virus susceptible to the drug concentrations achieved and the patient continuing to benefit from combination therapy. Resistance mutations versus compensatory mutations Natural selection has created HIV protease, or any other enzyme, with a finely balanced structure for optimum function. A change at just one amino-acid residue to achieve resistance to a drug will almost inevitably result in some loss of function. However, the enzyme may be able to readjust by making compensatory changes at other amino-acid positions to restore, at least in part, the original functionality. Thus, on drug treatment these compensatory changes (mutations) will arise alongside, or even predetermine, the emergence of the true resistance mutations. This phenomenon is clearly exemplified in vitro. Rose et al. 23 have shown that G48V and L90M mutations on an NL4-3 clone background reduce the activity of HIV protease by 99%. The addition of A71T increases the protease activity by 16-fold while having little effect on the Ki for SQV (from 26 to 21 nmol/l). In addition, they showed that the triple mutant G48V, A71T and V82A virus on an NL4-3 background had a viability less than 2% of wild-type virus. However, addition of an L10I mutation allowed the virus to grow well. Similarly, and consistently, Tisdale et al. 17 generated a triple mutant G48V, I84V, L90M virus in vitro in the presence of SQV. On subsequent viral passage in the absence of drug, an additional A71V mutation arose. The addition of both V71 and V84 mutations did not significantly alter the sensitivity of virus to SQV over the G48V, L90M double mutant. These additional mutations presumably aided viral growth. The compensatory roles of I10, P/Q/T63 and V/T71 for L90M with or without G48V can be demonstrated from a phenotypic study of clinical isolates from study NV14256 24 (Fig. 3). The isolates containing L90M plus I10, P/Q/T63 or T/V71 were no less sensitive to SQV than isolates containing L90M with consensus residues L10, L63 and A71 (Fig. 3a). Similarly, the addition of the non-consensus residues alone or in combination did not further decrease the sensitivity of V48, M90 double mutant virus (Fig. 3b) or have any effect on the sensitivity to SQV of ‘wild-type’ (G48, L90) virus (Fig. 3c). Nevertheless, in study NV14256 for SQV monotherapy (but not in zalcitabine or combination therapy), a statistically significant increase in I10, P/Q/T63 and T/V71 mutations was observed along with M90, and an association of I10 and T/V71 with M9021.Fig. 3: . Non-consensus amino acids at positions 10, 63, and 71 of HIV protease do not contribute to decrease viral sensitivity to saquinavir (SQV). The sensitivity to SQV of post-treatment viral isolates from patients in study NV14256 is shown. Sensitivity is plotted for each individual isolate against its genotype at positions 10, 63 and 71 showing either consensus sequence (-) or change (X) at each position. The horizontal dotted lines show the mean ± 95% confidence limits for sensitivity to SQV of pretreatment isolates. (a) Isolates with G48 plus M90; (b) isolates with V48 plus M90; (c) isolates with G48plus L90 wild-type (WT) at these positions. It can be seen that non-consensus mutations at positions 10, 63 and 71 alone or in combination do not further decrease sensitivity to SQV when found in wild-type, L90M or G48V plus L90M mutant virus. It can also be seen (consistent with Fig. 2) that the presence of mutations at position 48/90 does not automatically imply a virus will have a sensitivity to SQV outside normal baseline sensitivities. IC50, Median inhibitory concentration.The variants I10, I36, P63 and T/V71 all arise at polymorphic sites in wild-type virus (Table 1). Jacobsen et al. 5 observed that L90M mutations accumulate with a higher frequency in response to SQV in the protease genes containing these minor, non-consensus residues. Thus, data collectively suggest that I10, P/Q/T63, V/T71and possibly I36 and (rarely) V84 are compensatory mutations to the true resistance mutations to SQV (G48V and L90M). A similar analysis on isolates from nelfinavir-treated patients reported by Patick et al. 13 identified D30N as the specific resistance mutation, giving a five to 100-fold decrease in drug sensitivity, while additional mutations E35D, M36I, M46I, A71T/V, V77I and N88D were seen as compensatory, not affecting viral drug sensitivity. Recent evidence 25 suggests that V84, T82, and combinations of I36/V54 or R70/T82 give resistance to ritonavir but impair replication. Subsequent mutations restore replication efficiency. For indinavir this division is less clear. Condra et al. 11 report that progressive accumulation of any of the 11 mutations linked statistically with treatment give rise to a progressive decrease in viral drug sensitivity (Table 3), although I46 or A82 may be essential for the emergence of resistance.Table 3: . Association between the number of protease amino-acid substitutions and the level of resistance to indinavir.Cross-resistance A priori, it might be expected that drugs that select for the least number of mutations in a target protein would be the least likely to generate virus with reduced sensitivity to other drugs directed at the same target, especially if the mutations were not only few but particular to that drug. From the limited current data on phenotypic cross-resistance with clinical isolates from patients treated with protease inhibitors, this generality would appear to apply. Nelfinavir has perhaps only one characteristic resistance mutation, D30N. This has not been found to arise with any of the other protease inhibitors currently approved or in clinical trial. Preliminary studies in six isolates containing the D30N mutation revealed no loss of sensitivity to other protease inhibitors 26. SQV selects two characteristic resistance mutations, G48V and L90M, the former being unique, the latter seen at low incidence and late occurrence with other protease inhibitors, suggesting it may be a compensatory change in these cases. Fourteen isolates carrying L90M with or without G48V have been studied. Eight showed no loss of sensitivity to any protease inhibitor other than SQV itself (indeed, four had less than a fourfold loss of sensitivity to SQV) and only two showed decreased sensitivity to all other (three or four) protease inhibitors tested (indinavir, ritonavir, 141W94 and nelfinavir). Ritonavir 10 gives rise to a large number of mutations in protease which accumulate in a relatively ordered way, commencing with A82 usually followed by V54, V/T71 I/L36, V84, and a number of less common changes including I/L/V46 and M90 (in < 10% of resistant virus). Three to five mutations may accumulate within about 6 months of treatment. A82 and I46 are critical mutations for the emergence of resistance to indinavir and M90 for resistance to SQV. Cross-resistance studies with ritonavir-resistant isolates 10 suggest frequent cross-resistance to indinavir (as might be expected), but at a modest level. In addition, surprisingly, cross-resistance is also seen with nelfinavir, which shows a slightly greater decrease in drug sensitivity than for indinavir. However, there was no significant loss of sensitivity to SQV or 141W94. Indinavir selects at least 11 different mutations, although A82 and I46 may provide the critical core, and four to eight mutations can accumulate within 60 weeks 11. Nineteen drug-resistant isolates have been studied 11, 13 of which have the corresponding pretreatment isolate allowing proper comparison of genotypic and phenotypic change. Complete cross-resistance was found with ritonavir and XM412. The degree of loss of drug sensitivity to these two drugs was in some cases apparently greater than to indinavir itself. However, the true degree of loss of sensitivity in most cases could not be judged due to the choice of a 95% inhibitory concentration as the antiviral endpoint, limiting the dynamic range over which reduced sensitivity could be measured. There was approximately 80% incidence of loss of sensitivity to 141W94 (VX-478) and approximately 60% to SQV. Thus, superficially, the observations fit the expectations. However, there are several provisos and unanswered questions. In all these studies, four- or fivefold change in drug sensitivity was taken as significant (based largely on the confidence level of the reproducibility of drug sensitivity assays). This level may either underestimate or exaggerate the cross-resistance problem in terms of clinical relevance. There are many unanswered questions when trying to rationalize genotypic and phenotypic changes. There is no genotypic between isolates that are and those that are is there any mutants containing L90M have no greater loss in sensitivity to SQV than those is there a ritonavir-resistant isolates are to nelfinavir when they D30N or the I46, V84 mutations that have been generated in virus in vitro. It is also that changes at polymorphic sites are to some degree by most or all protease inhibitors either as compensatory changes or resistance mutations I10, I36, these may or may not cross-resistance when studied in clinical isolates in vitro, they may the virus population to relatively emergence of resistance to subsequent protease inhibitor therapy. For the increase in and selected by nelfinavir, or these mutations plus selected by indinavir may to the emergence of L90M on subsequent treatment with SQV 11 or However, these are variants and because they are only less fit than the consensus will in population with the least selective pressure. Hence, any population by one drug may have only effects on subsequent which itself will exert pressure to generate these changes. These can only be by clinical studies including between protease inhibitors and protease inhibitor studies are critical in order to make optimum clinical of these new and several studies are currently in The of genotypic response seen with each drug is such that studies will to a substantial population to be reports on just one or two individuals may be There is increasing clinical in combination therapy containing at least one protease and combinations of two or more protease inhibitors may be rate of resistance emergence might be expected where two inhibitors select for the same primary mutations ritonavir and indinavir for From a resistance the ideal drug combination would be those selecting different key mutation patterns SQV and nelfinavir). The pertinent question on resistance and cross-resistance for the patient and is, a patient is drug are the of resistance emerging to drug X and also to drug The as can be estimated from in vitro studies with isolates, is from the of two (i) the chance of resistant to treatment with drug and (ii) the chance of resistance to drug X resulting in loss of sensitivity to drug Thus, the drugs with which to treatment would be those with both a low incidence of resistance coupled with a low chance of cross-resistance. In study NV14256 in which clinical benefit of SQV in both and was RNA samples were for the HIV protease approximately of SQV treatment in combination with only of virus the L90M G48V was not and post-treatment pairs of isolates carrying L90M from this and other studies were tested for sensitivity to other protease inhibitors (indinavir, ritonavir, nelfinavir and Of only two of 10 showed reduced sensitivity to other protease inhibitors. these two the chance of loss of sensitivity to other protease inhibitors from SQV plus zalcitabine for about is (i.e., about Thus, 95% of patients be sensitive to subsequent protease inhibitor treatment. In study the clinical the reduction in viral by the SQV plus zalcitabine combination was modest with SQV that more in viral reduce the incidence of resistance further and hence the risk of cross-resistance. A mutation in HIV protease will arise less readily in patients treated with an HIV protease inhibitor if the mutation results in a substantial in viral replication fitness, and if the resulting decrease in viral drug sensitivity is giving little advantage in for the loss of of viral fitness can be from a study of the incidence of variants in the pretreatment viral those variants with a fitness similar to the wild-type virus will exist alongside the wild-type, while those that substantially disable the virus will not be The key mutations for SQV resistance (G48V and are absent from pretreatment viral and give only modest in viral drug sensitivity. This is consistent with the relatively low incidence of emergence of resistance to SQV at or higher levels The key mutations to ritonavir and nelfinavir have all been observed in isolates at low compensatory mutations may subsequent and in addition to the primary mutations. These appear to arise in order to restore, at least in part, viral The mutations seen to this tend to arise at sites that are already polymorphic in the untreated virus population. For SQV and nelfinavir these additional mutations do not appear to further decrease viral sensitivity to these In for indinavir 11 and ritonavir 10 they give rise to a progressive decrease in drug sensitivity with of mutations. cross-resistance is most likely to arise when a drug multiple mutations that substantially with the mutation of other protease inhibitors. data from studies with clinical isolates the order of frequency and of cross-resistance as indinavir ritonavir SQV This is consistent with the although there are several unanswered questions in the of genotype and cross-resistance