Constrained Evolution of Human Immunodeficiency Virus Type 1 Protease during Sequential Therapy with Two Distinct Protease Inhibitors

Abstract

Human immunodeficiency virus type 1 (HIV-1) variants that have developed protease (PR) inhibitor resistance most often display cross-resistance to several molecules within this class of antiretroviral agents. The clinical benefit of the switch to a second PR inhibitor in the presence of such resistant viruses may be questionable. We have examined the evolution of HIV-1 PR genotypes and phenotypes in individuals having failed sequential treatment with two distinct PR inhibitors: saquinavir (SQV) followed by indinavir (IDV). In viruses where typical SQV resistance mutations were detected before the change to IDV, the corresponding mutations were maintained under IDV, while few additional mutations emerged. In viruses where no SQV resistance mutations were detected before the switch to IDV, typical SQV resistance profiles emerged following the introduction of IDV. We conclude that following suboptimal exposure to a first PR inhibitor, the introduction of a second molecule of this class can lead to rapid selection of cross-resistant virus variants that may not be detectable by current genotyping methods at the time of the inhibitor switch. Viruses committed to resistance to the first inhibitor appear to bear the “imprint” of this initial selection and can further adapt to the selective pressure exerted by the second inhibitor following a pathway that preserves most of the initially selected mutations.

Inhibitors of the human immunodeficiency virus type 1 (HIV-1) protease (PR) are among the most active antiretroviral compounds used in the therapy of HIV-1 infection (8). These agents impair the maturation and the infectivity of viral particles and lead to a rapid blocking of virus replication (1, 7, 11, 16, 21, 30). Combinations of antiretroviral agents that include a PR inhibitor can drive the levels of HIV-1 RNA in the plasma below the limit detectable by current quantitation techniques and lead to long-term clinical and immunological improvements (2, 12, 19). Nevertheless, suboptimal therapies that fail to achieve a complete and sustained inhibition of virus replication usually lead to the selection of viral variants with decreased susceptibility to the corresponding agents (6, 10, 15, 18, 23). Mutations in the viral PR that can confer resistance to PR inhibitors have been thoroughly described both in vitro and in vivo (5, 20, 22, 23, 25, 26, 28). Selection for HIV-1 resistance to PR inhibitors is a complex process that can recruit mutations located in different sites of the PR. Although they may differ from one PR inhibitor to another, there seems to be considerable overlap in the mutational pathways leading to resistance to most currently used PR inhibitors (22). Resistance to saquinavir (SQV) is characterized by the selection of mutations L90M and/or G48V (15, 28, 31, 32). Resistance to indinavir (IDV) in most cases involves early selection of mutation V82A (5, 6). However, most initially selected resistance mutations are insufficient to induce high-level resistance, which requires the addition of secondary mutations that are often located outside of the active site of the PR, such as L10I, M36I, M46I, L63P, or A71V. Such secondary mutations, which increase the level of resistance and often correct the loss of fitness that can characterize resistant viruses, are observed during selection of most PR inhibitors (3, 14, 17, 22, 24, 29). As a consequence, resistant variants most often display significant cross-resistance to several inhibitors of the same class (6, 13, 32). In spite of accumulating in vitro data on cross-resistance between PR inhibitors, the clinical significance of cross-resistance is still unclear. In particular, the impact of resistance to one PR inhibitor on the virological response to a second molecule of the same class has not been fully evaluated. We have examined the evolution of HIV-1 genotypes and phenotypes in patients who have received SQV followed by IDV. Since virological failure of SQV has been reported to be often accompanied by the persistent replication of viruses that do not bear detectable resistance mutations in the PR (15, 28), we also wished to determine whether such apparently sensitive viruses would retain full in vivo responsiveness to subsequent therapy by another PR inhibitor.

Eleven patients who had failed sequential PR inhibitor therapy with SQV followed by IDV were retrospectively selected out of a population of 54 patients who had received SQV (hard-gel capsule formulation, 600 mg three times per day) as part of a clinical trial or of an expanded use program at the Antoine Béclère Hospital. Because the aim of the study was an analysis of the evolution of viruses under sequential selective pressure by distinct PR inhibitors rather than an evaluation of the response to such a regimen in treated patients, the criteria for treatment failure and patient selection in this study were arbitrarily defined in order to include viruses that clearly escaped both PR inhibitors. Therefore, treatment failure was defined as a reduction in plasma viremia of less than 1 log₁₀. Of the 54 selected patients, 26 were considered to have failed SQV therapy. Twenty-two of these 26 patients were switched to IDV (800 mg three times per day) with or without concurrent changes in other antiretroviral drugs. Of the 22 patients who were switched to IDV, 11 were considered to have failed IDV therapy and were subsequently studied. Four of these patients (patients C, D, G, and J) (Table 1) had a transient response to IDV characterized by a drop in plasma viremia in excess of 1 log₁₀, but their viremia subsequently rebounded to within 1 log₁₀ of viremia at the time of the switch. For all subjects, HIV-1 PR sequences were amplified from plasma virus by nested reverse transcription (RT)-PCR using the primer pair ProA⁺ (5′ GCT AAT TTT TTA GGG AAG ATC TG 3′) and ProA⁻ (5′ GGC AAA TAC TGG AGT ATT GTA TG 3′) and the pair ProB⁺ (5′ TTT TTA GGG AAG ATC TGG CCT TC 3′) and ProB⁻ (5′ GGA GTA TTG TAT GGA TTT TCA GG 3′). Population-based sequencing of the PR coding region was performed by the dideoxy-termination method on automated ABI sequencers using bulk second-round PCR products as templates and primers PRO1 (sense, 5′ CCC TCT CAG AAG CAG GAG 3′) and PRO2 (antisense, 5′ TGG GCC ATC CAT TCC TGG CTT 3′). Plasma virus loads were determined with the Amplicor kit (Roche).

TABLE 1.

PR mutations in plasma-derived HIV-1 sequences from PR inhibitor-treated subjects

Patient	Timea	PR inhibitor treatment	VLb	Amino acids at selected PR positionc												Other changes
				10	20	36	46	48	54	63	71	73	82	84	90
Group I
A	M−3	>6 Months of SQV	5.1	L	K	M	I	G	I	P	A	G	V	I	M
	M0	Switch to IDV	5.2	L	K	M	I	G	I	P	A	G	V	I	M
	M7		4.2	I	K	M	I	G	I	P	A	S	V	I	M
	M10		4.6	I	K	M	I	G	I	P	A	S	V	I	M
B	M−8	>6 Months of SQV	4.9	I	I	I	M	G	I	P	T	S	V	I	M
	M0	Switch to IDV	4.4	I	I	I	M	G	I	P	T	T	V	I	M
	M3		4.7	I	I	I	I	G	I	P	T	T	V	I	M
	M6		4.7	I	I	I	I	G	I	P	T	T	V	I	M
C	M−8	>6 Months SQV	4.8	L	K	M	M	G	I	P	T	G	V	I	M
	M0	Switch to IDV	5.5	L	K	M	M	G	I	P	T	G	V	I	M	T12S
	M3		4.2	L	K	M	M	G	I	P	V	S>G	V	I	L>M	T12S
	M6		3.5	V	K	M	M	G	I	P	V	S	V	I	M	V13I, I72R, V77I
D	M−3	Start SQV	5.1	L	K	M	M	G	I	L	V	G	V	I	L
	M0	Switch to IDV	5.2	L	K	M	M	G	I	P	V	S>G	V	I	M
	M2		3.8	L	K	I	M	G	I	P	V	S	V	I	M
	M4		4.6	L	K	I	M	G	I	P	V	S	V	I	M
E	M−7	Start SQV	4.5	L	K	M	M	G	I	P	A	G	V	I	L
	M0	Switch to IDV	4.5	L	K	M	M	G	I	P	V	S	V	I	M
	M5		4.5	L	K	M	M	G	I	P	V	S	V	I	M
	M8		5.1	I	K	M	M	G	I	P	V	S	V	I	M
Group II
F	M−1	Start SQV	5.8	I	R	I	M	G	I	C	A	G	V	I	L
	M0	Switch to IDV	5.6	I	R	I	M	G>V	I	P	A	G	V	I	L	E16G, D37E
	M3		5.7	I	R	I	M	V	I	P	A	G	A	I	L	E16G, D37E
	M8		4.8	I	R	I	M	V	I	P	A	G	A	I	L	D37E
G	M−4	Start SQV	5.7	L	K	M	M	G	I	P	A	G	V	I	L
	M0	Switch to IDV	5.5	L	K	M	M	G	I	P	T	G	V	I	L
	M6		4.0	I	K	M	L	G	I	P	T	G	A	I	M
	M9		5.3	I	K	M	L	G	I	P	T	G	A	I	M
Group III
H	M−9	Start SQV	5.0	L	K	M	M	G	I	P	A	G	V	I	L
	M0	Switch to IDV	5.1	L	K	M	M	G	I	P	A	G	V	I	L
	M5		5.5	L	K	M	M	G	I	P	S	G	V	I	M
	M10		4.6	I	K	M	I	G	I	P	S	G	V	I	M	72R, 95F
I	M−8	Start SQV	6.2	L	K	M	M	G	I	P	T	G	V	I	L
	M0	Switch to IDV	5.9	L	K	M	M	G	I	P	T	G	V	I	L
	M3		5.2	L	K	M	M	G	I	P	T	S	V	I	M
J	M−10	Start SQV	5.3	L	K	M	M	G	I	P	A	G	V	I	L
	M0	Switch to IDV	5.4	L	K	M	M	G	I	P	A	G	V	I	L
	M3		4.2	I	K	M	M	G	I	P	T	G	V	I	M
	M5		4.6	I	K	M	M	G	I	P	A	S	V	I	M	V13I
K	M−5	Start SQV	4.9	L	K	M	M	G	I	P	A	G	V	I	L
	M0	Switch to IDV	4.8	L	K	M	M	G	I	P	A	G	V	I	L
	M5		4.1	I	K	M	M	G	I	P	A	S	V	I	M
	M10		5.4	I	K	M	M	G	I	P	V	S	A	I	M	I62V, I93L

^aTimes are indicated in months relative to the switch from SQV to IDV, represented as M₀. 

^bPlasma viral load in log₁₀ HIV-1 RNA copies/milliliter. 

^cOne-letter code for amino acids at the indicated positions in the HIV-1 protease. The positions were found to be frequent sites for amino acid substitutions in HIV-1 variants with decreased susceptibility to PR inhibitors. Amino acid changes often associated with resistance are shown in bold type. 

The evolution of HIV-1 PR sequences in the 11 study patients before and after the switch from SQV to IDV is presented in Table 1. According to this evolution, we subdivided the 11 patients into the following three groups: group I (patients A through E), in which significant genotypic PR resistance profiles resulting from SQV treatment were present at month 0 (M₀), the time of the switch; group II (patients F and G), in which relatively little evolution of PR had occurred before M₀; and group III (patients H through K), in which HIV-1 PR amino acid sequences did not bear detectable resistance mutations at M₀. In group I, patients A, B, and C had received SQV for more than 6 months before the earliest sequence was available. Several months before the switch, all three patients displayed the characteristic SQV-resistance mutation L90M, which is associated with other amino acid changes. Comparable amino acid changes were observed in patients D and E at M₀. After the switch from SQV to IDV, all five viruses retained the L90M mutation with the exception of a relative and transient decrease of that mutation at M₃ in virus populations from patient C. In this group of viruses, a few additional mutations emerged following the switch, involving codons 10, 46, 71, and 73. However, none of the five viruses required the addition of a mutation at codon 82, most often associated with IDV resistance. Therefore, in this group, mutations selected under SQV treatment appeared to induce sufficient cross-resistance to IDV to allow significant virus replication in the presence of IDV in vivo. In the two viruses in group II, only one mutation had been selected during the course of SQV treatment. Compared to standard HIV-1 subtype B sequences, virus from patient F displayed amino acid changes at positions 10, 20, and 36 which are often observed during PR resistance, but which, because the corresponding virus had not been exposed to SQV, were more likely to be a natural polymorphism than the result of SQV selection. In this virus, the SQV-resistance mutation G48V was present as a minority at M₀ and was maintained thereafter. Moreover, the V82A mutation was observed at M₃, a frequent association with the G48V mutation in viruses with high-level resistance to SQV (32). In patient G, the only mutation selected at M₀ was A71V, which was followed by the addition of several other resistance mutations, including V82A. In group III, plasma virus sequences displayed no detectable resistance mutations at the time of the switch to IDV. Careful examination of the sequence chromatograms at M₀ failed to reveal even minor species of mutant genomes. Surprisingly, the resistance mutations that developed in these viruses after the switch were very similar to those described above for patients in group I, which had developed resistance mutations under SQV treatment. In particular, under the selective pressure of IDV, all four viruses in group III developed the L90M mutation, which is associated with remarkably similar combinations of mutations at codons 10, 46, 71, and 73.

To confirm the results of the genotypic analysis reported above, HIV-1 SQV and IDV susceptibility phenotypes were examined sequentially in nine of the patients. We used a rapid recombination method that allows testing of recombinant virus carrying plasma-derived PR sequences (27). Briefly, PR sequences were amplified from plasma by RT-PCR using the same primers as for PR sequencing and were recombined by cotransfection in HeLa cells with a linearized pNL4-3-derived plasmid carrying a full deletion of PR coding sequences. The resulting recombinant virus was tested for susceptibility to SQV and IDV on HeLa-derived indicator P4 cells, as previously described (33). Drug susceptibility was expressed as IC₉₀ values (the drug concentration required to inhibit 90% of HIV replicative events) using the median effect method (4). The results of this phenotyping analysis are presented in Table 2. For patients E, G, H, I, and K, the parental pre-SQV virus was used as the reference to calculate the decrease in drug susceptibility. For patient J, in which the pretherapy virus was not tested, we used the M₀ virus, which was obtained at the time of the switch and was free of detectable resistance mutations, as the reference. For patients A, C, and D, in which genotypic resistance had developed before the first tested plasma sample was obtained, the reference was pNL4-3 virus. In all patients, as shown in Table 2, measurable dual resistance to SQV and IDV developed, with a gradual increase over time in viral resistance to both inhibitors. With the notable exception of virus from patient H, there was a general trend towards a higher resistance to SQV than to IDV. In the four tested group I viruses (from patients A, C, D, and E), which displayed clear SQV resistance genotypes at M₀, phenotypic testing revealed an increase in SQV resistance of 5.8-, 6.4-, 19.2-, and 8-fold, respectively. Viruses from patients A and E displayed an increase in resistance to both SQV and IDV over time, while virus from patient C, in which the L90M mutation was temporarily reduced to a minority at M₃, showed a corresponding decrease of resistance to both inhibitors. Therefore, as noted from the genotypic analysis of viruses of this group, the initial resistance to SQV was in most instances not only maintained, but clearly reinforced by the switch to IDV. In virus from patient G, in which a single A71T mutation emerged at M₀, followed by numerous other resistance mutations, including L90M and V82A (a typical IDV resistance mutation), a parallel increase in the level of resistance to both drugs was measurable after M₀. Finally, as expected from their genotypes, viruses in group III had no significant phenotypic resistance to either of the two drugs at M₀. Following the switch from SQV to IDV, these four viruses displayed a decrease of their susceptibility to both drugs. In patient H, the M₁₀ virus was clearly more resistant to IDV than to SQV. However, in the three other patients (subjects I, J, and K), the increase in viral resistance was significantly more prominent regarding SQV. Again, in these cases, the phenotypic analysis confirmed the genotype data showing that viruses that were selected following the switch from SQV to IDV were in fact mainly SQV-resistant viruses.

TABLE 2.

IC₉₀s of IDV and SQV

Patienta	Timeb	IC90 of IDV (fold increase)c	IC90 of SQV (fold increase)c
A	M−3	72 (2.9)	25 (5)
	M0	100 (4)	29 (5.8)
	M7	199 (8)	63 (12.6)
	M10	ND	ND
C	M−8	109 (4.4)	21 (4.2)
	M0	72 (2.9)	32 (6.4)
	M3	44 (1.8)	19 (3.8)
	M6	ND	ND
D	M−3	66 (2.6)	16 (3.2)
	M0	243 (9.7)	96 (19.2)
	M2	398 (15.9)	ND
	M4	ND	ND
E	M−7	50 (1)	5 (1)
	M0	199 (4)	40 (8)
	M5	199 (4)	41 (8.2)
	M8	250 (5)	63 (12.6)
G	M−4	25 (1)	1.5 (1)
	M0	40 (1.6)	10 (6.7)
	M6	199 (8)	25 (16.7)
	M9	ND	ND
H	M−9	25 (1)	5 (1)
	M0	40 (1.6)	11 (2.2)
	M5	79 (3.2)	10 (2)
	M10	794 (32)	32 (6.4)
I	M−8	19 (1)	3.5 (1)
	M0	20 (1.1)	9 (2.6)
	M3	30 (1.6)	25 (7.1)
J	M−10	ND	ND
	M0	25 (1)	5 (1)
	M3	62 (2.5)	50 (10)
	M5	125 (5)	53 (10.6)
K	M−5	27 (1)	1.5 (1)
	M0	ND	ND
	M5	224 (8.3)	39 (26)
	M10	316 (11.7)	16 (10.7)

^aThe letters designating treated subjects are the same as those used in Table 1. 

^bTime is expressed in months relative to the switch from SQV to IDV. 

^cIC₉₀ values are expressed in nM. ND, not determined. The increase in resistance, shown in parentheses, is expressed relative to the earliest available virus for patients E, G, H, I, and K, and relative to the M₀ virus for patient J. For patients A, C, and D, the increase was calculated relative to wild-type pNL4-3, an infectious molecular clone of HIV-1. 

The results of this study reveal that viruses selected in vivo for SQV resistance display a significant cross-resistance to IDV. As a result of this cross-resistance, the switch to IDV maintained the SQV-selected mutations, while only a small number of additional mutations emerged as the result of IDV selective pressure. The corresponding viruses, bearing the “imprint” of SQV selection, were not required to radically modify their pattern of PR-resistance mutations. In particular, only a few of the viruses selected here needed a mutation of codon 82, most frequently observed during selection for IDV resistance. This can be explained by the fact that the cross-resistance to IDV resulting from the initial SQV selection is sufficient for in vivo resistance to IDV. It could also be explained by a possible relative incompatibility between mutation L90M and V82A in terms of viral replicative fitness. Indeed, it has been recently reported that in virus populations where mutations of codon 82 develop as a result of SQV selection, quasi-species bearing the L90M mutation are often gradually eliminated over time (9). Correspondingly, mutations of codon 82 are significantly more often observed in association with G48V, a mutation able to confer high-level resistance to SQV, than with L90M.

More interestingly, viruses with no detectable resistance mutations at the time of the introduction of IDV (group III viruses) rapidly developed a pattern of genotypic and phenotypic resistance predominantly to SQV, even after removal of that drug. In these cases, it can be assumed that minor virus variants bearing SQV-resistance mutations failed to be selected by SQV, probably due to an insufficient selective pressure provided by the poorly bioavailable formulation of SQV used in this study. Because of the insufficient amount of active drug and also presumably because of a slightly reduced viral fitness in mutant quasispecies compared to that of wild-type virus, these variants failed to significantly take over wild-type virus before the inhibitor switch. After the introduction of IDV, by virtue of the significant SQV-IDV cross-resistance measured here and of the better bioavailability of IDV, the mutant quasi-species were readily selected, again bearing the imprint of SQV selection. Later, these imprinted viruses behaved similarly to the SQV-selected viruses found in the other patients, withholding their mutation selection pathway.

It should be feared that such a scenario can be applied to many instances where PR inhibitors are used sequentially. Indeed, one can assume that in cases of insufficient virological response to a first PR inhibitor, even if resistant variants are not perceptible by current genotyping or phenotyping methods and provided that they display sufficient cross-resistance to the second PR inhibitor, these variants will be selected following the switch and will gradually emerge, leading to therapeutic failure of the second inhibitor. Therefore, it should be stressed that antiretroviral treatments including PR inhibitors should aim at a full suppression of viral replication in vivo, thereby preventing any cryptic selection of resistant and potentially cross-resistant variants.