Virologic failure was associated with rapid selection of multiple mutations at positions 36, 38, 40, 42, 43, 44, or 45 within the HR1 region of HIV-1 gp41 with no particular mutation pattern, except for 2 patients for whom only the N43D mutation was selected. Interestingly, most mutations were already present in the first sample tested on ENF. In 4 cases (patients 1, 3, 5, and 9), however, a novel mutation was detected after the other(s). In 3 of these cases, the later mutation to emerge was localized at position 42.
Surprisingly, a high number of mixtures of wild-type and mutant strains, as observed in early samples, persisted over time at all codons at which mutations occurred or at a single or several positions. In patient 6, the V/A mixture at codon 38 persisted throughout the 11 months of viral replication under therapy; during this time, position 36 changed from mixture to pure mutant. In patient 10, 3 of the 4 detected mutations persisted as mixtures during at least 8 months of ENF treatment, whereas a pure mutant 38A was found in the first sample tested on ENF.
In some cases, mutants were no longer detected while the patient viruses were still replicating under ENF pressure (patients 3, 5, 7, and 10). In patient 3, the mutation Q40H disappeared between months 1 and 3, when a mixture at position 42 was detected. The mutation L44M was also no longer detected at month 3 but re-emerged at month 5. The passage from mixture to wild type occurred after viral failure on ENF for patient 7 (at position 36) and patient 10 (at positions 42 and 44). These unexpected evolutions were confirmed by re-examination of the electropherograms. It is possible, however, that discrimination between mixtures and pure mutant or wild type would have been different if amplification and sequencing had been performed several times from the same samples.
Disappearance of Resistance Mutations After Enfuvirtide Withdrawal
After removal of ENF in the treatment regimen, mutations within the HR1 region disappeared within 6 months in 8 of 9 patients. For patient 9, N42N/T was still present at his death after 7 months without ENF. All mutations reversed to wild type simultaneously or within a 1- or 2-month interval. The longest delays before disappearance of all mutations occurred in the cases in which the mutations did not disappear simultaneously. Mixtures did not tend to reverse to wild type faster than pure mutants, although in the latter cases, it may be possible that minority wild-type variants were still present in the population but not detected, contributing to more rapid disappearance of mutants than in detected mixtures.
Sequencing of Individual Variants
One striking finding of this study was the high number of mixtures of wild-type and mutated codons at positions associated with ENF resistance within the HR1 region. The distribution and relative proportion of each mutation on each individual genome thus needed to be determined. We therefore studied the viral quasispecies of patient 3 by limiting dilution at baseline and in 4 samples after ENF viral escape over the 6-month treatment period during which wild-type and mutated codons coexisted (Table 2). A total of 71 single genomes were sequenced; at least 15 sequences from each sample were analyzed, except for the last sample, for which only 5 sequences were obtained. Eight sequences contained a deletion of approximately 100 bases. At months 0, 2, and 4, the deletion was localized at the 5′-terminus, whereas month 1 and 2 variants were found to harbor a deletion at the 3′-terminus. The six 5′-deleted sequences were no longer considered. Thus, the results from the 65 studied sequences only are shown in Table 2. The 63 entire sequences carried 2 insertions compared with the reference HXB2 gp41: one between residues 4 and 5 and another between codons 104 and 105.
As expected, at baseline, viral sequences showed wild-type aa at the 4 positions at which mutations were later selected by ENF. Surprisingly, even under ENF selection pressure and during the 6-month period of follow-up, although resistance mutations were selected, between 10% and 20% of quasispecies still carried wild-type codons at these positions.
During the first 2 months of viral escape, we observed a high degree of quasispecies variability. Strains with different mutations at position 38 were selected as well as 1 strain carrying mutated residues at positions 40 and 44 (see Table 2). Most of these variants disappeared after the second month, when genomes bearing V38A and N42T mutations became predominant, accounting for 56% of the strains at month 2 and 80% at month 6. If these 2 mutations seemed to have a replication advantage under ENF selection over other combinations of existing mutations at these positions, selection of other mutated residues within and outside the HR1 region still continued to occur, however. Heterogeneity increased after this bottlenecking effect on quasispecies evolution. The 4 clones with V38A and N42T at month 6 were different from each other even when analyzed at the aa level.
In the phylogenetic tree (Fig. 4), the variants with the same aa pattern at the 4 ENF target positions rather than those from the same time point tended to cluster together. The same phylogenetic analysis was done after removing the 36-to-45 region of gp41 from all these isolated variants; it again showed that most of the sequences of a same group based on the aa at positions 38, 40, 42, 44 formed a cluster, except for the sequences of the QQNL group, which were separated (not shown). Another exception was a subgroup of wild-type sequences (group VQNL) from M4 and M6 plus 1 sequence from M1, which constituted an outgroup. They are also separated in Figure 4 with the complete gp41 ectodomain sequence. Viral sequences from population sequencing of the post-ENF samples are closely related to clones carrying the same pattern of mutations: the month 6 and 12 post-ENF populations with wild-type ENF target residues are more closely related to M0 wild-type clones, whereas M4 and M6 on-ENF wild-type clones form a distinct group as noted previously. In contrast, the M2 post-ENF population, still containing strains with mutations at the 4 ENF target positions, clusters with the V38A and N42T double mutants from different time points of ENF therapy.
Phenotyping and Determination of Coreceptor Use
Phenotypic sensitivity to ENF was performed on longitudinal samples from 8 patients using a newly developed recombinant assay, PHENOSCRIPT (Table 3). Baseline samples from 3 patients could not be amplified by PCR, whereas the recombinant virus assay (RVA) of the PHENOSCRIPT analysis was negative for 4 other samples. Thus, the baseline sensitivity of the strains from only 4 of the 8 patients could be determined, with a mean of 0.032 μg/mL ± 0.035 (range: <0.005 for patient 10 to 0.076 for patient 4).
For a single set of patients (patients 1, 2, 8, 9 and 10; see Table 3A), increased median inhibitory concentration (IC50) values were, as expected, associated with the emergence of mutations within the gp41 ectodomain during ENF therapy, with a return to a wild-type genotype and a sensitive phenotype after stopping ENF. For a second set of patients (patients 3, 4, and 5), a direct correlation between the evolution of the IC50 and the number of mutations was not observed (see Table 3B). For these patients, the gp41 from PCR products obtained for the RVA, spanning the gp120-to-gp41 extracellular region of the envelope, was directly sequenced. On the same samples, important differences in the viral population were detected when different amplification reactions with different sets of primers were performed (see Table 3B). The coexistence of several variants with different aa at a given position, as often observed in this study, has certainly contributed to this difference in amplification efficiency. To resolve such mismatches, it would be necessary to repeat the PCR and sequencing several times with each sample, which is not routinely practicable.
In 7 of 8 patients, only R5 tropic viruses were found. Patient 2 is the only patient who harbored dual/mixed tropic strains. In no case did a switch in coreceptor use occur during ENF therapy or after.
Other Changes Outside the 36-to-45 gp41 Region
We then analyzed the rest of the sequenced region (see Fig. 1) to see whether other mutations could be selected by and contribute to resistance to ENF. The only mutations that were not found in baseline samples of our patients and disappeared after stopping ENF are N126K and E143D, which are located in the HR2 region. E137K and S138A developed in more than 1 patient during ENF treatment but were also found in 1 ENF-naive patient (E137K in patient 11 and S138A in patient 1). N126K was associated with N43D (patient 2); N42N/T (patient 5); a V38V/E plus N43D combination (patient 11); and a V38A, N42S/T, N43N/K, and L44L/M (patient 10) combination, respectively, in HR1. E143D was associated with N43D in patient 8 and a V38A plus N42S/T combination in patient 10, but only at the end of treatment (24 months) in this case. E137K/Q emerged in association with N43D in patient 8 and an N42N/T plus N43N/D combination in patient 9. In 1 of these patients, S138S/A (patient 9) also appeared during therapy. S138A was detected at the end of treatment (M33) in patient 11 and after ENF removal in patient 2 (not shown). In each case, the E137K/Q mutation was initially present. The associated mutations in HR1 were a single N43D mutation (patient 2) or in combination with a V38E mutation present in mixtures with wild type (patient 11).
Mutations in the aa 36-to-45 region of gp41 are considered the major determinants of resistance to ENF.2-7,11,20-22 Our results confirm this finding. In all our patients, mutations in this region developed during ENF administration and were lost when ENF was discontinued. Emergence of these mutations followed the evolution of viral load. The patients in this study were heavily pretreated and showed treatment failure to the 3 other classes of antiretroviral drugs available. Under these conditions, treatment failure to ENF could also be expected. All the patients responded to ENF with a prompt decrease in HIV-1 RNA once ENF was introduced. After 4 weeks of treatment, however, the viral load returned to baseline for most patients, which was associated with the appearance of mutations in the aa 36-to-45 region. A link between the presence of these mutations and phenotypic resistance was also demonstrated in our study. Yet, several data suggest that the mechanisms of resistance to ENF might be different from those described for other drugs acting on the polymerase gene and that information from a genotypic resistance assay based on the key gp41 aa involved in the reduced susceptibility to ENF could be insufficient for a valuable interpretation.
Regarding the ENF gp41 target region, some particular evolutions were observed. The pattern of mutations was different for each patient, except for 2 patients harboring a single mutation. Among the variable number of mutations detected in our small patient population, the combination of 2 mutations was the most frequent. The mutations often emerged simultaneously. Mixtures of mutant and wild type persisted at most of the mutated positions during the whole period of treatment and even after removal of ENF. This occurred in patients receiving ENF for up to 33 months and with a high viral replication rate. Such evolutions have already been described.5,11 Mixtures with wild-type strains were still seen at the end of treatment in 8 of 17 patients treated for 21 to 96 weeks in the study of Xu et al11 and in 1 of 4 patients treated for 80 weeks in the study of Poveda et al.5 Moreover, Xu et al11 reported that 58% of the 119 substitutions detected were present as mixtures. These observations differ from what is known for the other antiretroviral drugs, which are divided into 2 categories: those for which 1 mutation is sufficient to induce full resistance (eg, NNRTIs) and those for which the level of resistance increases with the number of mutations. In this latter category, the different mutations, such as the thymidine analog mutations or the mutations conferring resistance to PIs, are acquired progressively and the variants with an increasing number of mutations are selected over those with fewer mutations. During ENF treatment, the nature of the mutations sometimes changes with time in the same patient, an observation described in other studies,4,5,11,21,22 suggesting independence of the ENF mutations. Furthermore, the phylogenetic analysis in a single patient showing a clustering of the sequences with the same pattern at the ENF target positions, even after removing the 36-to-45 region of gp41, tends to suggest that each population comes from a distinct baseline variant and evolves independently, at least in the first months of treatment. The persistence of mixtures despite the maintenance of a selection pressure on a highly replicating virus would thus be attributable to the possible coexistence of several variants, with each carrying at least 1 resistance mutation. Selection of wild-type virus from a mixture under the continued pressure of ENF on replicating viruses has also been found. In these cases, the reproducibility of the amplification step for minority variants is a likely explanation for some of the unexpected events. There is also evidence that even if mutant genomes could always be present, their proportion can decrease with time while under the therapeutic pressure. This is at least the case for patient 3, for whom the isolation of the Q40H and L44M variants was unsuccessful after M1, although the persistence of a low proportion of these mutants is likely, given the detection of L44M by population sequencing. Another example is patient 5, for whom the G36D mutation was detected until month 2 and not at 3 later time points analyzed. Other studies4,11,21,22 have already described fading of mutations with single mutants replaced by others with a different residue at the same position or a substitution at another position,4,21,22 or variants with multiple mutations evolving toward single mutants.11,21
A possible explanation for the persistence of wild-type virus could be a suboptimal drug level; however, the ENF dosage is not available to evaluate such a possibility. Nevertheless, an adherence problem is unlikely, given that these patients were enrolled in a clinical trial with close monitoring.
Analysis of variants from samples of a single patient with a complex pattern of mutations and several longitudinal samples available showed that a great diversity of populations coexists in the first months of therapy. With time, we observed a tendency toward selection of some of them harboring more resistance mutations, although this is less clear when mixed populations are sequenced. This process seems rather slow, however; in particular, sequences with wild-type aa at all the major positions persisted during the entire treatment, at least in patient 3. In this case, several mutations are well tolerated at position 38 but V38A was progressively selected. This could be associated with a higher initial prevalence of V38A or improved resistance conferred by the addition of N42T. The Q40H and L44M mutations were shared by the same genome independent of the other changes at positions 38 and 42. Thus, the fact that the mutations at residues 40, 44, and 45 emerge only in patients with a combination of 3 or 4 mutations in the mixed-virus population does not mean that they can only develop after other mutations. Moreover, they have been detected as a single mutation in other reports.2,4,7 It could be hypothesized that Q40H and L44M do not confer a high level of resistance because their rate tended to decrease with time on-ENF, but these mutations, alone or in combination, have been associated with a substantial decrease in susceptibility in other published cases.4,7,21 The absence of selection of the variants with V38E, as detected at month 1 in 26% of the clones from patient 3, is also not in accordance with the results of others. Mink et al20 found that the presence of a V38E mutation was associated with an increase in resistance to ENF of more than 1000-fold in the pNL4-to-3 strain background and more than 30-fold when the same mutation was introduced in 2 baseline primary isolates, whereas the V38A substitution “only” conferred a decrease in susceptibility of 16-fold in the pNL4-to-3 strain. Whether this difference is linked to genetic variations outside the aa 36-to-45 region or to the fact that the V38E mutation cannot be associated with the N42T mutation in contrast to the V38A mutation is not known. Another possibility is that variants carrying V38E have a reduced fitness compared with wild-type strains.23
The circulation of different quasispecies in different proportions also has implications at the technical level. The detection of each strain in a mixture is variable according to the sensitivity of the technique as well as from a single experiment to another with the same protocol for rare variants. This problem has led to discordances between genotypic and phenotypic results. Given the characteristic evolution of resistance mutations to ENF discussed previously, such technical problems could be encountered more frequently with entry inhibitors than with RT and PR inhibitors. This needs to be taken into account when interpreting data, not only for the correlation between genotype and phenotype but for the genotype only. Discrepancies with mixtures have been reported previously by comparing population sequencing and clonal analysis.24 The apparent reversion of a mutation can only reflect a decrease in its proportion, because it can be detected by sequencing of another amplification from later samples.
Disagreements between genotype and phenotype suggest that variations in regions of gp41 or gp120 other than HR1 are induced during ENF treatment and could contribute to its resistance. Other studies have identified mutations outside the target region that could influence the potency of ENF, such as N126K10 and S138A11 in the HR2 region. No consistent data have been obtained from our results or those of others in vivo about the nature of other positions involved in ENF resistance. The N126K mutation seems to be induced by ENF and was unselected when it was stopped. It has been shown from a single patient that the combination of V38A and N126K leads to an ENF-dependent hyperfusogenic HIV-1 envelope.10 We found V38A in only 2 of the 4 patients harboring an N126K mutation, however. The role of N126K has also been emphasized using variants selected by serial passage in cells expressing membrane-anchored gp41 peptide encompassing the ENF sequence.12 It was shown that N126K without mutations in the 36-to-45 region was responsible for a decrease in susceptibility to ENF. The impaired fitness conferred by this mutation might explain why it is not detected in ENF-naive individuals and in treated patients without compensatory mutations. Substitutions E137K/Q, S138A, and N143D also tended to emerge in response to ENF but less frequently. Furthermore, E137K/Q and S138A can be seen as polymorphisms and often seem to be associated. It has been shown that the tryptophane-rich C-terminal sequence of ENF can bind to the IFIMIVG sequence of the membrane-spanning domain of gp41, contributing to its HIV-1 fusion inhibitory activity.25 This sequence was conserved in all our patients before or after ENF administration, except in a single pretherapeutic sample from 1 patient (patient 11), in which the first isoleucine was replaced by a leucine (not shown). We did not analyze the gp120 sequence, but data from others suggest that ENF could bind to this protein with important consequences in its activity, which might lead to adaptive changes under ENF pressure.25-27
The genetic background of the HIV-1 envelope has also been proposed to influence the wide range of baseline sensitivity to ENF observed at the population level.4,9 The susceptibility to ENF before treatment might represent an average of coexisting sensitive and insensitive variants. It is not known whether these viruses are naturally resistant because of the presence of sequence variations somewhere in the envelope gene or because of host characteristics. Nevertheless, these variants could be maintained under the ENF selective pressure without developing resistance mutations, which would furthermore explain the persistence of wild-type strains despite continued therapy. Our results from phylogenetic analysis tend to suggest that gp41 wild-type strains persisting during ENF treatment could arise from a minority variant or archived virus at baseline, whereas dominant replicating variants present before therapy re-emerge when the therapeutic pressure is removed. The genetic context of the wild-type gp41 variants present before therapy and of those selected during therapy could be better adapted in the absence and presence of ENF, respectively. It has been shown that the gp120 quasispecies emerging during treatment were maintained after ENF withdrawal, contributing to a tendency toward an increase in fitness during and after ENF therapy compared with baseline.17 Thus, we cannot exclude the possibility that pre-ENF gp41 sequences found in posttherapy samples are associated with gp120 sequences appearing during treatment, such that post-ENF variants would be different from those present at baseline. In addition, it remains unknown whether wild-type strains persisting on treatment were phenotypically sensitive to ENF.
The phenotypic assay used in this study does not enable us to determine the baseline susceptibility to ENF for all our patients. The 4 values obtained vary by more than 10-fold but were all lower than 0.1 μg/mL. The patient with the most sensitive viral population (patient 10) also showed the most durable response to ENF. Of note, for this patient's virus, a polymorphic mutation, N42S, was found. This polymorphism has already been implicated as having better susceptibility to ENF, which thus seems to be confirmed in vivo in our study.
In conclusion, although it would be necessary to have tests to guide therapy with fusion inhibitors, the complexity of the entry process makes that objective more complicated than with RT or PR inhibitors. Intervention of other viral and host factors and the presence of a high number of quasispecies have technical and interpretation implications that hamper the ability of such tests to predict virologic outcome. There is a need for better understanding of how entry inhibitors act and what envelope regions are contributing to their efficacy before valuable genotypic assays can be used. In most of the studies conducted to date, including ours, the patients were in an advanced disease stage and heavily pretreated, with few therapeutic options, such that the suboptimal efficiency of ENF was expected. Because ENF is now more widely used in patients, new data concerning the evolution and relation of the induced mutations might be obtained.
The authors thank Marie Hélène Teillon-Beranger and Marc Carteron for technical assistance with DNA sequencing and Katharina Skrabal for helpful discussion regarding the manuscript.
1. Wild CT, Shugars DC, Greenwell TK, et al. Peptides corresponding to a predictive alpha-helical domain of human immunodeficiency virus type 1 gp41 are potent inhibitors of virus infection. Proc Natl Acad Sci USA
2. Marcelin A-G, Reynes J, Yerly S, et al. Characterization of genotypic determinants in HR-1 and HR-2 gp41 domains in individuals with persistent HIV viraemia under T-20. AIDS
3. Menzo S, Castagna A, Monachetti A, et al. Genotype and phenotype patterns of human immunodeficiency virus type 1 resistance to enfuvirtide during long-term treatment. Antimicrob Agents Chemother
4. Sista PR, Melby T, Davison D, et al. Characterization of determinants of genotypic and phenotypic resistance to enfuvirtide in baseline and on-treatment HIV-1
5. Poveda E, Rodes B, Labernardière B, et al. Evolution of genotypic and phenotypic resistance to enfuvirtide in HIV-infected patients experiencing prolonged virologic failure. J Med Virol
6. Wei X, Decker JM, Liu H, et al. Emergence of resistant human immunodeficiency virus type 1 in patients receiving fusion inhibitor (T-20) monotherapy. Antimicrob Agents Chemother
7. Melby T, Sista P, DeMasi R, et al. Characterization of envelope glycoprotein gp41 genotype and phenotype susceptibility to enfuvirtide at baseline and on treatment in the phase III clinical trials TORO-1 and TORO-2. AIDS Res Hum Retroviruses
8. Derdeyn CA, Decker JM, Sfakianos JN, et al. Sensitivity of human immunodeficiency virus type 1 to fusion inhibitors targeted to the gp41 first heptad repeat involves distinct regions of gp41 and is consistently modulated by gp120 interactions with the coreceptor. J Virol
9. Labrosse B, Labernardière J-L, Dam E, et al. Baseline susceptibility of primary human immunodeficiency virus type 1 to entry inhibitors. J Virol
10. Baldwin CE, Sanders RW, Deng Y, et al. Emergence of a drug-dependent human immunodeficiency virus type 1 variant during therapy with the T20 fusion inhibitor. J Virol
11. Xu L, Pozniak A, Wildfire A, et al. Emergence and evolution of enfuvirtide resistance
following long-term therapy involves heptad repeat 2 mutations within gp41. Antimicrob Agents Chemother
12. Lohrengel S, Hermann F, Hagman I, et al. Determinants of human immunodeficiency virus type 1 resistance to membrane-anchored gp41-derived peptides. J Virol
13. Heil ML, Decker JM, Sfakianos JN, et al. Determinants of human immunodeficiency virus type 1 baseline susceptibility to the fusion inhibitors enfuvirtide and T-649 reside outside the peptide interaction site. J Virol
14. Lazzarin A, Clotet B, Cooper D, et al. Efficacy of enfuvirtide in patients infected with drug-resistant HIV-1
in Europe and Australia. N Engl J Med
15. Palmer S, Kearney M, Maldarelli F, et al. Multiple, linked human immunodeficiency virus type 1 drug resistance mutations in treatment-experienced patients are missed by standard genotype analysis. J Clin Microbiol
16. Kumar S, Tamura K, Nei M. MEGA3: integrated software for molecular genetics analysis and sequence alignment. Brief Bioinform
17. Labrosse B, Morand-Joubert L, Goubard A, et al. Role of the envelope genetic context in the development of enfuvirtide resistance
in human immunodeficiency virus type-1 infected patients. J Virol
18. Coakley E, Petropoulos CJ, Whitcomb JM. Assessing chemokine co-receptor usage in HIV. Curr Opin Infect Dis
19. Agence Nationale de Recherche sur le SIDA et les Hépatites. HIV-1
genotypic drug resistance interpretation's algorithms. 2002. Available at: http://www.hivfrenchresistance.org
. Accessed September 2002.
20. Mink M, Mosier SM, Janumpalli S, et al. Impact of human immunodeficiency virus type 1 gp41 amino acid substitutions selected during enfuvirtide treatment on gp41 binding and antiviral potency of enfuvirtide in vitro. J Virol
21. Lu J, Deeks SG, Hoh R, et al. Rapid emergence of enfuvirtide resistance
-infected patients. Results of a clonal analysis. J Acquir Immune Defic Syndr
22. Cabrera C, Marfil S, Garcia E, et al. Genetic evolution of gp41 reveals a highly exclusive relationship between codons 36, 38 and 43 in gp41 under long-term enfuvirtide-containing salvage regimen. AIDS
23. Lu J, Sista P, Giguel F, et al. Relative replicative fitness of human immunodeficiency virus type 1 mutants resistant to enfuvirtide (T-20). J Virol
24. Marcial M, Lu J, Deeks SG, et al. Performance of human immunodeficiency virus type 1 gp41 assays for detecting enfuvirtide (T-20) resistance mutations. J Clin Microbiol
25. Liu S, Lu H, Niu J, et al. Different from the HIV fusion inhibitor C34, the anti-HIV drug Fuzeon (T-20) inhibits HIV-1
entry by targeting multiple sites in gp41 and gp120. J Biol Chem
26. Yuan W, Craig S, Si Z, et al. CD4-induced T-20 binding to human immunodeficiency virus type 1 gp120 blocks with the CXCR4 coreceptor. J Virol
27. Platt EJ, Durnin JP, Kabat D. Kinetic factors control efficiencies of cell entry, efficacies of entry inhibitors, and mechanisms of adaptation of human immunodeficiency virus. J Virol
Keywords:© 2007 Lippincott Williams & Wilkins, Inc.
enfuvirtide resistance; HIV-1; persistence of wild-type and mutant mixtures