Impact of Nelfinavir Resistance Mutations on In Vitro Phenotype, Fitness, and Replication Capacity of Human Immunodeficiency Virus Type 1 with Subtype B and C Proteases

Abstract

Human immunodeficiency virus type 1 subtype B and C proteases were manipulated to contain 90M, 88D, or 89L, and their in vitro biological properties were studied. We showed that D30N has significantly more impact in subtype C than in subtype B counterparts, accounting for the reported low prevalence of this mutation in patients failing nelfinavir-based regimens.

Drugs designed to inhibit reverse transcriptase and protease (PR) of human immunodeficiency virus type 1 (HIV-1) have been implemented in the last decade, and their combination regimens have improved the treatment of infected individuals (13, 17). However, not all patients respond to highly active antiretroviral treatment and eventually develop drug resistance through the accumulation of several patterns of drug resistance mutations. Such mutations have been extensively characterized for subtype B isolates (9, 18; see the updated list in the Los Alamos National Laboratory HIV sequence database at www.hiv.lanl.gov), and their biological relevance and phenotypic impact in non-B subtypes need to be clarified.

The mutation D30N is a unique amino acid substitution found in the PR from isolates of patients failing nelfinavir (NFV)-containing regimens. Recently, Cane et al. (3) studied 58 HIV-1 subtype C-infected individuals treated with several combinations of PR inhibitor (PI)-containing regimens and found the D30N mutation in only one patient, together with the N88D mutation, although NFV was widely used in this patient group. This phenomenon was also observed in patients infected with CRF01(A/E) failing a PI-containing regimen in Japan, where the mutations D30N, A71V, and N88D were found exclusively in subtype B-infected individuals (2). Previous studies conducted with HIV-1 subtype B-infected patients indicated, however, that around 50% of patients receiving NFV fail therapy due to the development of the D30N mutation, whereas the remaining accumulate an alternative mutation pattern containing L90M as the primary mutation (14). This study explores this uneven distribution of D30N and L90M PR mutations between B and C subtype strains and shows the different in vitro behaviors of B and C isolates carrying these mutations.

Clone pNL4-3 (1) was used as the prototypic subtype B strain, and pNL43-C6 (7) was used as the subtype C prototypic clone. Clone C6 is a modified pNL4-3 infectious clone where the six amino acid differences between the subtype C and B consensus sequences (I15V, M36I, R41K, H69K, L89M, and I93L) were introduced through site-directed mutagenesis (7). Prototypic pNL43 and C6 PR genes were cloned in the pCR4 TOPO vector (Invitrogen) by following the manufacturer's protocol. Clones were then subjected to site-directed mutagenesis in order to modify the target codons by a PCR mutagenesis procedure described by the manufacturer (QuikChange kit; Stratagene). Using this technique, we constructed the clones C^D30N, C^D30N/N88D, C^D30N/M89L, C^L90M, B^D30N, B^D30N/N88D, and B^L90M. Primer sequences used in those mutageneses are available upon request. Prototypic and mutant PR clones were inserted in a PR-deleted HXB2 backbone by a recombinant virus assay technology as previously described (8, 11). Cotransfections were always conducted with the same amounts of PR fragment and vector DNA (5 and 1 μg, respectively) in three independent experiments. B^L90M and B^D30N/N88D generated detectable recombinant viruses 6 days posttransfection (pt), 1 day after B^wt, whereas B^D30N and B^D30N/N88D were capable of generating detectable viral particles only after 10 days pt. In contrast, C^D30N, C^D30N/N88D, and C^L90M viruses appeared in the cell supernatant at the same time as the parental clone, C^wt. This kinetics of appearance was reproducible in all three experiments.

RNA from viruses was extracted from transfection cell culture supernatants and reverse transcribed into cDNAs, and the products were used in PCRs targeting the HIV-1 PR gene as previously described (7). PR genes were then resequenced to confirm mutations. To check for possible adaptation of PR cleavage sites with a particular PR genotype (4, 5, 16), the sites located in the gag region were amplified with primers P24INF (5′-GTCCAAAATGCGAACCCAGATTGTAA-3′) and PNL43R (5′-CGCTGCCAAAGAGTGATCTGAGGGAA-3′). The gag fragment had 528 nucleotides and contained CA/p2, p2/NC, NC/p1, p1/TFP, and p1/p6 cleavage sites. PCR fragments were sequenced in both directions in an ABI 310 automated sequencer (Applied Biosystems, Foster City, Calif.), with the same primers used in the amplifications. The recovered virus from subtype B transfections did not accumulate any secondary mutation in either genomic region. In contrast, C^D30N (but not C^wt, C^D30N/N88D, or C^L90M) accumulated a new mutation solely in the PR at position 83 (N83T), encoded by a G-to-T transversion at the first nucleotide of the 83rd PR codon. Of note, the mutation was observed when this experiment was repeated three times. To explore the reason for lack of recovery of C virus carrying solely the D30N mutation, the polymorphism L89M found in the subtype C consensus was reverted by site-directed mutagenesis. This specific position was chosen due to its nearness to positions 88 and 30. This subtype C signature was reverted in the C^D30N clone, generating C^D30N/M89L. This new construct was transfected, generating a productive infection in MT4 cells comparable to that for the mutated C clones. The viruses present in these two culture supernatants were isolated and sequenced, and no secondary mutations (including N83T) could be found.

A chimeric virus containing the HXB2/wt pNL43 PR was also generated as a susceptible reference, and 50% inhibitory concentrations (IC₅₀) for all mutated viruses were determined in phenotyping assays for all Food and Drug Administration-approved PIs, amprenavir, indinavir, lopinavir, ritonavir, saquinavir, and nelfinavir. Phenotyping results shown are the means of three independent assays using the MTT (3-[4,5-dimethylthiazol-2-yl]-2,5-diphenyltetrazolium bromide)-based cell viability assay as previously described (8). All statistical methods to calculate the IC₅₀ of the clones and isolates were performed with the Analyze-it statistics package, version 1.62, for Microsoft Excel and Sigmaplot software. Briefly, a Hill's three-parameter nonlinear regression was performed to obtain the sigmoid curve of viable cells and the IC₅₀ value of each drug for each virus tested. Table 1 shows the mean IC₅₀ values ± standard deviations (SD) for three independent replicates. The level of resistance to NFV induced by L90M is lower than that induced by D30N regardless of the subtype. However, the isolate B^D30N had a higher level of resistance to NFV than C^D30N/N83T. There was a noticeable hypersusceptibility to amprenavir in clones carrying N88D regardless of their subtype. The phenomenon was also observed when ritonavir was tested in C^D30N/N88D and C^D30N/N83T clones, but not in the B^D30N/N88D counterpart. Similar to what has been shown in previous work (7) there is a noticeable hypersusceptibility to lopinavir in all subtype C clones.

TABLE 1.

Phenotypic analysis of B and C clones and reference HXB2\NL-43 PR^a

Clone	IC50 (μM) ± SDb (fold resistancec) of:
	APV	SQV	IDV	RTV	NFV	LPV
HXB2\NL-43	78.75 ± 5.31	3.07 ± 0.81	8.5 ± 1.9	25.25 ± 3.9	41.75 ± 9	28 ± 2.1
BD30N	88.00 ± 6.61 (1.1)	1.89 ± 1.01 (0.6)	6.35 ± 0.58 (0.74)	23.5 ± 5.2 (0.94)	1,743 ± 87.69 (41.7)	20.96 ± 7.5 (0.75)
BD30N/N88D	36.25 ± 4.42 (0.46)	4.02 ± 1.02 (1.3)	8.55 ± 1.1 (1)	12.75 ± 5.1 (0.51)	520.2 ± 34 (12.4)	31 ± 7 (1.1)
BL90M	95.5 ± 10.72 (1.21)	13.75 ± 3.5 (4.47)	32 ± 3.2 (3.7)	86.25 ± 9.6 (3.45)	178 ± 6.8 (4.28)	27 ± 4 (0.96)
CD30N/N83T	27.75 ± 3.096 (0.35)	1.72 ± 0.51 (0.56)	8.27 ± 0.75 (0.9)	2.55 ± 1.2 (0.102)	1,447.5 ± 50 (34.6)	2.27 ± 1 (0.08)
CD30N/N88D	32.375 ± 4.151 (0.41)	2.1 ± 0.2 (0.7)	8.9 ± 1.5 (1.05)	2.05 ± 0.12 (0.082)	636.5 ± 21 (15.2)	2.3 ± 0.5 (0.08)
CL90M	122.75 ± 13.961 (1.5)	14.25 ± 2.6 (4.6)	42.2 ± 4.1 (4.97)	92.5 ± 3.8 (3.7)	241 ± 82 (5.7)	3.15 ± 0.7 (0.11)

^aPI abbreviations: APV, amprenavir; SQV, saquinavir; IDV, indinavir; RTV, ritonavir; LPV, lopinavir.

^bSD of six independent experiments.

^cResistance compared to that for HXB2\NL-43 PR. Resistance values significantly different from that for the HXB2\NL-43 PR reference clone (t test with P < 0.05) are in boldface.

Viral replication capacities (RC) were evaluated by following the protocols previously described (6, 10). Briefly, 1,000 50% tissue culture infectious doses were used to infect 10⁶ MT4 cells (multiplicity of infection [MOI] = 0.001). Cell culture supernatants were collected every day for nine consecutive days. The amount of virus produced in the cell supernatants was evaluated by determining p24 antigen content. For RC estimation, the time in days to reach a level of 5 × 10⁵ pg of p24 antigen/ml in culture (replication constant K) was calculated by using Hill′s three-parameter nonlinear regression modeling. Relative RC were determined from ratios of resistant-strain K (K_rs) to wild-type strain K (K_wt) and were expressed as percentages of the latter. For the evaluation of K_wt, each drug-sensitive clone from either subtype C or B was used as a reference for its respective mutated clones. The K_wt and K_rs used in the determination of the relative RC of HIV isolates were the mean K values for the duplicate experiments with each virus. Figure 1 depicts the results of these experiments. We observed similar RC in subtype B clones carrying NFV resistance mutations (RC of 91.9 and 90.3% of that for B^wt for B^L90M and B^D30N/N88D, respectively), except for clone B^D30N, which had a delayed-growth-curve profile (RC = 86.5%). Of note, all B clones were first detected in the cell supernatant at the same time (3 days postinfection). On the other hand, the RC for subtype C clones after infection corroborate the data obtained in the transfections. The clone with lowest RC, as evidenced by its delayed virus growth, was C^D30N/N83T (RC of 62% of that for C^wt). The C^L90M clone had an intermediate RC (78.9%), while the C^D30N/M89L and C^D30N/N88D clones both replicated with the same efficiency (RC = 72.3 and 72.4%, respectively). There was an evident reversion of the impairment of the D30N mutation in the subtype C PR when the M89L reversion mutation was introduced, similar to what was noticed with the C^D30N/N88D clone (Fig. 1).

FIG. 1.

Virus growth kinetics after infection of MT4 cells (MOI = 0.001) with recombinant infectious virus of subtypes B (A) and C (B). The level of p24 was measured as described in the text. Results shown are means of six independent experiments, and bars represent SD.

The relative viral fitness of two recombinant viruses was analyzed by following the methodology previously described by García-Lerma et al. (6). The viral lysate from two different viruses was used to infect 10⁶ MT4 cells in exponential growth with the same MOI of 0.001 for each isolate. Every 4 days, 200 μl of viral supernatant was used to infect 10⁶ fresh MT4 cells. Five passages were done, and aliquots were collected. The relative amount of each virus in the culture was evaluated by reverse transcriptase PCR, followed by a single dideoxynucleotide incorporation assay using the Big Dye kit (ABI, Alameda, Calif.) as previously described by García-Lerma et al. (6). Briefly, the methodology is based on the measurement of viral relative proportions based on the chromatogram peak areas of each variant, where each one is marked by extension with a dideoxynucleotide of a different fluorescent color. For each competition assay point we calculated the ratio of one virus to the other one (R_N) by dividing the peak area for one virus by the peak area for the other one. The ratio at time zero (R₀) was calculated at the conclusion of the mixture of the two competing viruses at the beginning of the assay (10). Relative fitness was determined by determining the adaptive coefficient s, which is the slope of the plotted curve. MT4 cells were also separately infected with each virus to assess possible spontaneous reversion of the mutated codons. The following competition experiments were conducted: clone C^wt versus C^D30N, B^wt versus B^D30N, C^L90M versus C^D30N, B^L90M versus B^D30N, and B^D30N versus C^D30N. Figure 2 shows the R₀/R_N curves for these experiments, where we could observe a lower fitness (relative adaptive coefficient s) of clone C^D30N than of C^wt and C^L90M. The same phenomenon was observed when subtype B clones were analyzed. However, the differences in coefficients for B^D30N versus those for B^wt and to B^L90M were lower than those for the equivalent comparisons between subtype C clones (Fig. 2).

FIG. 2.

Fitness vectors and corresponding fitness values for diverse virus competitions: C^D30N versus B^D30N, B^L90M versus B^D30N, B^D30N versus B^WT, C^D30N versus C^WT, and C^L90M versus C^D30N. In each competing passage, the ratio between both viruses (R_N) was compared to the ratio for the initial mixtures (R₀). The value in each passage was used to derive the fitness vectors by linear regression. Details of the procedure are described in the text. The slopes of fitness vectors (s) for the HIV-1 clone at the right in each pair are indicated in the plotting area.

This work has explored the uneven distribution of D30N and L90M PR mutations in B and C subtype strains and showed in vitro evidence that D30N can drastically impair subtype C virus replication. This fact is corroborated by a lower RC for the C^D30N clone than for C^L90M and C^wt, and all C clones except C^wt showed a lower RC when matched with equivalent subtype B clones. RC from subtype B clones were not as drastically impacted by NFV-related mutations as those from subtype C clones. In addition, only C^D30N virus carrying a spontaneous mutation (N83T) incorporated upon culture could be obtained. This mutation has been previously associated with PI treatment in B isolates (4, 19). However, the role of this substitution in PI resistance is largely unknown. We speculate that this mutation was selected to restore the fitness of the C^D30N clone, and this might be a favorable in vitro genetic route, as we observed independently three times. Clone B^D30N, however, did not accumulate secondary mutations in the PR gene.

Our study further explored the cause of D30N′s unusual behavior in a subtype C backbone, implicating PR polymorphism L89M in this phenotype. We speculate that the methionine found in C PR (but not in B) at this position disturbs the stability of this protease, which carries the D30N substitution, as this polymorphism involves the only amino acid that differs between C and B PR in PR C terminal and that spatially neighbors residues 90, 88, and 30. The reversion of L89M to M89L in subtype C PR isolates carrying D30N clearly recovered the viral RC of this variant, similarly to what happened with the addition of N88D (see Fig. 1), without incorporation of secondary mutations such as N83T or N88D.

We have showed that the mutation D30N confers a higher level of resistance to NFV (42-fold for B^D30N and 34-fold for C^D30N/N83T) than L90M (4.2-fold for B^L90M and 5.7-fold for C^L90M) regardless of the subtype analyzed (Table 1). This contrasts with the findings reported by Perrin et al. (15), who observed only a twofold difference in NFV resistance level between virus with the D30N and L90M mutations when included separately in subtype B clones. Of note, B^D30N viruses have a level of resistance to NFV comparable to that of the C^D30N virus. The inclusion of the N88D mutation in B^D30N and C^D30N clones reduced the IC₅₀ of NFV and induced in the clones a hypersusceptibility to amprenavir (P < 0.001; t test) compared to that for the NL43 reference strain, as previously reported for subtype B (20). However, we could find a specific hypersusceptibility to ritonavir when C^D30N/N83T and C^D30N/N88D clones were tested. In this case, IC₅₀s for these two clones were shown to be approximately 10 times lower than that for the reference strain (P < 0.003; t test). Although reaching higher NFV resistance levels in the phenotyping assay, the viruses of both subtypes carrying D30N substitutions were less fit than their counterparts carrying the NFV resistance mutation L90M. The virus C^D30N/N83T was slightly less fit than its correlate B^D30N, reflecting the importance of the emerging N83T substitution for the rescue of fitness of the C^D30N clone. The fitness data, taken together with the RC results, explain the preferential route of NFV resistance in subtype C clinical isolates through L90M and not D30N after NFV failure (3). This pathway is less costly to viral fitness and does not require any compensatory mutation in that subtype. This might be a specific phenotype from subtypes C and CRF01_AE and should not be generalized to all non-B strains. Although the results shown here explain the low prevalence of D30N mutations in subtype C strains, we have to bear in mind that the subtype C isolates studied here are really hybrid viruses because they contain a subtype B backbone, and this can be a limitation to the data presented here. This knowledge, however, may influence future therapeutic strategies and interpretation of genotypic resistance in areas where these variants prevail.