Evidence of a Role for the Q151L Mutation and the Viral Background in Development of Multiple Dideoxynucleoside-Resistant Human Immunodeficiency Virus Type 1

Abstract

The majority of human immunodeficiency virus type 1 (HIV-1)-infected patients treated with zidovudine (AZT) plus zalcitabine (ddC) and didanosine (ddI) develop AZT resistance mediated by mutations such as T215Y and M41L. Only a small proportion of patients develop multiple dideoxynucleoside resistance (MDNR) mediated by the Q151M mutation. To gain insight into the factors responsible for the low frequency of selection of Q151M, we evaluated the replication capabilities of recombinant viruses carrying two possible intermediates (151L or 151K) of the Q151M mutation generated in different reverse transcriptase (RT) genetic backgrounds. The 151L and 151K mutations were introduced by site-directed mutagenesis in RTs from two patient-derived HIV-1 isolates that had either wild type (WT) Q or the Q151M (posttreatment isolate) mutation. For comparison, both mutations were also introduced in a laboratory-adapted HIV-1 strain (HIV-1_HXB2). Analysis of replication capabilities showed that both 151L and 151K were lethal in RT genetic backgrounds of the WT isolate and in HIV-1_HXB2. In contrast, 151L but not 151K allowed virus replication in RT backgrounds of the posttreatment isolate. Three mutations (V35I, S68G, and I178M) were present in the RT background of the posttreatment isolate but not in the WT isolate. Introduction of S68G in the RT of both the WT isolate and HIV-1_HXB2 partially restored replication capacity of recombinants carrying the 151L mutation. The S68G mutation alone did not confer a significant replicative disadvantage in WT viruses. Like HIV-1_151M, HIV-1_151L RT was found to have six- to eightfold resistance to AZT-triphosphate (TP), ddA-TP, and ddC-TP, indicating an MDNR phenotype. However, HIV-1_151L was found to be less fit than HIV-1_151M, which may explain the preferential selection of HIV-1_151M observed in vivo. The demonstrated ability of HIV-1_151L/68G to replicate and the associated MDNR suggest that 151L is a potential intermediate of Q151M. The dependence of HIV-1_151L on other mutations, such as S68G, for replication may explain the low frequency of the Q151M-mediated pathway of resistance.

The nucleoside reverse transcriptase (RT) inhibitors zidovudine (AZT), didanosine (ddI), zalcitabine (ddC), stavudine (d4T), lamivudine (3TC), and abacavir are widely used to treat persons infected with human immunodeficiency virus type 1 (HIV-1) (3). However, specific patterns of mutations in the HIV-1 pol gene usually occur in these patients following treatment and have been associated with resistance and cross-resistance to each of these compounds. For instance, the Met184Val (M184V) mutation is associated with resistance to 3TC, the L74V mutation is associated with resistance to ddI and ddC, the T69D mutation is associated with resistance to ddC, and the T215Y/F, M41L, and K70R mutations are associated with resistance to AZT (21). The presence of these mutations generally results in the loss of the antiretroviral and clinical benefit of these drugs.

The T215Y mutation is the first mutation identified in the majority of patients receiving combination antiretroviral therapy with AZT and ddC-ddI (2). However, in the past few years several reports have shown that HIV-1 carrying a different pattern of mutations is selected in 3 to 16% of patients treated with AZT and ddC-ddI and sporadically in patients receiving other ddI-containing regimens (12, 22–26). These mutations occur at positions 62 (A62V), 75 (V75I), 77 (F77L), 116 (F116Y), and 151 (Q151M) and confer resistance to all currently approved nucleoside analogs including AZT, ddI, ddC, d4T, 3TC, and abacavir (6, 26–28; K. Van Laethem, M. Witvrouw, J. Balzarini, J.-C. Schmit, S. Sprecher, P. Hermans, M. Leal, T. Harrer, L. Ruiz, B. Clotet, M. Van Ranst, J. Desmyter, E. De Clercq, and A.-M. Vandamme, Letter, AIDS 14:469–471, 2000). The emergence of these multiple dideoxynucleoside resistance (MDNR) mutations may compromise the clinical efficacy of this entire class of compounds, thus severely limiting treatment options.

Among the five MDNR mutations, the Q151M mutation plays a pivotal role in the acquisition of the MDNR phenotype. Q151M is the first mutation identified in the majority of patients who develop MDNR and by itself confers low-level resistance to AZT, ddI, ddC, and d4T (11, 12, 22, 27). In contrast, the A62V, V75I, F77L, and F116Y mutations do not affect drug susceptibility by themselves, but their co-occurrence with Q151M results in high-level resistance to AZT, ddI, ddC, and d4T and low cross-resistance to 3TC (6, 11).

Patients treated with AZT plus ddI-ddC develop AZT resistance through acquisition of either the T215Y or Q151M mutation but very rarely through both mutations (12, 22). The lack of coexistence of these two mutations observed in vivo indicates the presence of two different pathways for HIV-1 to develop AZT resistance. Although the viral determinants that may influence selection of either Q151M or T215Y remain undefined, it is known that coexistence of T215Y and Q151M is not constrained since introduction of both mutations in the same virus does not significantly affect replication or enzymatic activity of the RT (15, 28).

The factors responsible for the low frequency of selection of Q151M compared to T215Y are not known. The low frequency of Q151M cannot be explained by the number of mutations required for each amino acid change, since both Q151M (CAG→ATG) and T215Y (ACC→TAC) require two-base transversions and therefore, may preexist at a similar low frequency in the absence of drug pressure. In addition, the replicative fitness of HIV-1_Q151M has been found to be higher than that of HIV-1_T215Y both in the presence and in the absence of selective pressure with AZT (13, 15). Thus, decreased fitness of HIV-1_Q151M does not explain the low frequency of selection of the Q151M mutation compared to T215Y. Therefore, additional factors such as drug susceptibility and replication capabilities of putative intermediates of Q151M may influence selection of the Q151M pathway of resistance.

The possible intermediates that require a single base substitution from the wild type (WT) Q (CAG) to M (ATG) at codon 151 are leucine (L; CTG) or lysine (K; AAG). Little is known about the replication capability and drug susceptibility of HIV-1_151L or HIV-1_151K. A recent report indicated that these intermediates were deleterious when present in a laboratory-adapted (HIV-1_HXB2) genetic background (13). Based on these results, Kosalaraksa et al. postulated that for Q151M to develop, two base-pair transversions are needed concurrently or within a short period of time, which may be a relatively infrequent event (13). However, studies using HIV-1_HXB2 may be limited in their ability to assess how polymorphisms present in the RT genetic background of patient-derived HIV-1 isolates may influence or compensate the deleterious effect of the 151L or 151K mutation. Therefore, to fully evaluate the role of the 151L and 151K mutations in the acquisition of Q151M, additional analysis using patient-derived isolates may be necessary.

In the present study, we investigated whether replication of HIV-1_151L or HIV-1_151K may limit acquisition of the Q151M mutation. We introduced the 151L and 151K mutation in HIV-1 RT backgrounds obtained from a patient who developed the Q151M mutation during antiretroviral therapy. We compared the effect of these mutations in the patient-derived isolates with that seen in a laboratory-adapted HIV-1 strain and found that only the genetic background associated with Q151M allows replication of HIV-1_151L. We also identified a mutation at codon 68 (S68G) that alone can partially restore replication of HIV-1_Q151L in WT viruses. Our results suggest that Q151M may potentially be acquired through a 151L intermediate. The dependence of 151L on other mutations such as S68G may explain the low frequency of selection of the Q151M-mediated pathway of resistance.

MATERIALS AND METHODS

Virus isolates.

HIV-1 isolates L1S and L2S were obtained from an HIV-1-infected patient who developed MDNR after sequential treatment with AZT, ddC, and ddI. They were kindly provided by Anne-Mieke Vandamme (Rega Institute for Medical Research and University Hospital, Katholieke Universiteit, Leuven, Belgium). Detailed information about the patient antiretroviral treatments, as well as the two isolates used, has been previously described (23). Isolate L1S has WT Q at codon 151, while isolate L2S has the Q151M mutation. No other MDNR mutations were present in isolate L2S. L1S and L2S were isolated from samples collected 14 months apart (23).

Cloning of full-length HIV-1 RT from isolates L1S and L2S.

Full-length RT from isolates L1S and L2S was amplified by RT-nested PCR and cloned using the TA cloning kit (Invitrogen; Promega). Briefly, RNA was extracted from culture supernatants using the QIAmp Viral RNA kit (Qiagen). The RT reaction was done for 1 h at 42°C using primer RT2 as described previously (29). After a first round of PCR amplification using primers AV150 and RT2, 4 μl was subjected to a second round of amplification using primers IN3 and IN5 (29). A 1,703-bp PCR product comprising the complete HIV-1 RT was gel purified (QIAquick Gel Extraction; Qiagen) and ligated into the pCR2.1 vector. OneShot competent cells (TOP10F′) were then transformed, and single colonies were screened for the insert by PCR amplification using primers IN3 and IN5 (29). Plasmids containing the whole HIV-1 RT from L1S or L2S (referred to as pL1S or pL2S, respectively) were purified from individual colonies (Qiagen Plasmid Purification Kit) and were used for site-directed mutagenesis.

Primers used for site-directed mutagenesis at codons 151 and 68 of HIV-1 RT.

The following primers were used to generate the 151L or 151K mutation (mutations are underlined in the primer sequences): 5′-CAG TAC AAT GTG CTT CCA CTG GGA TGG AAA GGA TCA CC-3′ (151LF1, sense) and 5′-GG TGA TCC TTT CCA TCC CAG TGG AAG CAC ATT GTA CTG-3′ (151LR1, antisense) for the 151L mutation; and 5′-CAG TAC AAT GTG CTT CCA AAG GGA TGG AAA GGA TCA CC-3′ (151KF1, sense) and 5′-GG TGA TCC TTT CCA TCC CTT TGG AAG CAC ATT GTA CTG-3′ (151KR1, antisense) for the 151K mutation. The primers used to generate mutants at codon 68 (68G) of HIV-1 RT were 5′-GCC ATA AAG AAA AAA GAC GGT ACT AAG TGG AG-3′ (68GF1, sense) and 5′-CT CCA CTT AGT ACC GTC TTT TTT CTT TAT GGC-3′ (68GR1, antisense).

Site-directed mutagenesis.

Mutations at positions 151 and/or 68 were introduced in the pHXB2RIP7-based infectious clone pSUM9 (kindly provided by H. Mitsuya) or in plasmid preparations containing full-length L1S- or L2S-derived RT sequences (plasmids pL1S.18_Q151, pL1S.28_Q151, pL2S.4_M151, and pL2S.16_M151). The following 13 plasmids with mutant RTs were made: pL1S.18_Q151L, pL1S.18_Q151K, pL1S.28_Q151L, pL1S.28_Q151K, pL1S.28_Q151L/S68G, pL1S.28_S68G, pL2S.4_M151L, pL2S.4_M151K, pL2S.16_M151L, pL2S.16_M151K, pHXB2_Q151L, pHXB2_Q151K, and pHXB2_Q151L/S68G.

Site-directed mutagenesis was done using the QuikChange Site-Directed Mutagenesis Kit (Stratagene). To generate mutants at codon 151 in patient-derived RTs, 25 ng of purified plasmids pL1S.18_Q151, pL1S.28_Q151, pL2S.4_M151, or pL2S.16_M151 was added to a cocktail containing 125 ng of forward and reverse mutagenic primers (151LF1-151LR1 or 151KF1-151KR1), 2.5 U of PfuTurbo DNA polymerase, and deoxynucleoside triphosphates (dNTPs). Cycling conditions were 95°C for 30 s and then 12 cycles of 95°C for 30 s, 55°C for 1 min, and 68°C for 12 min. After digestion of the parenteral (nonmutated) DNA template by the DpnI restriction enzyme, Epicurian Coli XL1-Blue supercompetent cells were transformed, and individual colonies were screened for the presence of the 151L or 151K mutation by sequence analysis using primers AV36 and NE(1)35. Plasmid preparations pL1S.28_Q151L/S68G and pL1S.28_Q151/S68G were generated from pL1S.28_Q151L and pL1S.28_Q151, respectively, using the cycling conditions described above and the mutagenic primers 68GF1 and 68GR1. All plasmids containing the desired mutations were purified, and their RT sequence inserts were used to generate recombinant viruses. Mutant RT inserts were all sequenced following mutagenesis and cloning to verify absence of unexpected changes.

The pSUM9 infectious clone was used to generate pHXB2_Q151L and pHXB2_Q151K. Conditions for site-directed mutagenesis were as described above, except the extension at 68°C was increased to 32 min. Following mutagenesis, the whole RT was amplified by PCR using primers IN3 and IN5 and cloned using the TA cloning system. Individual colonies were then screened for the 151L or 151K mutation by sequence analysis. Mutagenic primers 68GF1 and 68GR1 were used to generate pHXB2_Q151L/S68G from pHXB2_Q151L as described above.

Generation of recombinant viruses with cloned RT sequences from isolates L1S, L2S, and HIV-1_HXB2.

Both mutated or nonmutated RT sequences were used to generate recombinant viruses with the RT-deleted HXB2-based proviral molecular clone pHIVΔRTBstEII. Recombinants were generated in MT-4 cells as previously described (10, 29). Briefly, the full-length RT was amplified by PCR from plasmid preparations using primers IN3 and IN5 and purified with the QIAquick PCR purification kit (Qiagen). MT-4 cells (2.5 × 10⁶) were then cotransfected with 1 or 2 μg of purified PCR product and 10 μg of BstEII-linearized pHIVΔRTBstEII (10). Culture supernatants were collected at different time points or when the full cytopathic effect (CPE) was observed, and then they were aliquoted and stored at −70°C. Levels of p24 antigen (Coulter HIV-1 p24 Antigen Assay) and RT activity were quantitated in cell-free culture supernatants. Amino acids 7 to 246 of HIV-1 RT were sequenced in all virus stocks to exclude possible spontaneous reversion of relevant mutations or accumulation of additional mutations.

The following recombinant viruses were obtained using the RT from HIV-1_HXB2: HXB2_Q151 (nonmutated), HXB2_Q151L, HXB2_Q151K, and HXB2_Q151L/S68G. The recombinant viruses generated using the RT from isolate L1S were L1S.28_Q151 (nonmutated), L1S.28_Q151L, L1S.28_Q151K, L1S.28_Q151L/S68G, L1S.28_S68G, L1S.18_Q151L, and L1S.18_Q151K. The recombinant viruses generated using the RT from isolate L2S were L2S.4_M151 (nonmutated), L2S.4_M151L, L2S.4_M151K, L2S.16_M151L, and L2S.16_M151K.

Titration of virus stocks.

To determine the infectious virus titers of each stock, MT-4 cells (30,000 cells/well in six replicates) were added to 96-well flat-bottom microtiter culture plates (Costar) and exposed to serial dilutions of each infectious virus. Cells were examined for CPE at day 5 of culture, and the 50% cell culture infectious dose (CCID₅₀) was determined by the method of Reed and Muench (19).

Analysis of replication kinetics in MT-4 cells.

Inocula of 450 CCID₅₀ was used to infect 4.5 × 10⁵ MT-4 cells (multiplicity of infection [MOI] = 0.001). After incubation for 2 h at 37°C, cells were washed twice with phosphate-buffered saline and resuspended in complete medium at 7.5 × 10⁴ cells/ml. Two-milliliter cultures were done in triplicate using 24-well tissue culture plates (Costar). Supernatants (200 μl) from each culture were collected at different days and then an equal volume of culture media was added. Levels of p24 antigen were quantitated in cell-free culture supernatants and were used to monitor replication kinetics.

Sequence analysis of HIV-1 RT.

Sequence analysis of HIV-1 RT (from nucleotides 2529 to 3333 of HXB2; amino acids 7 to 246) was done in an ABI 373 automated sequencer using primers AV36, AV44, A35, and NE(1)35 (23). The DNAsis program was used to analyze the data and to determine deduced amino acid sequences.

Amino acid frequencies in HIV-1 sequence databases were analyzed using the RT and Protease Sequence Analysis Program developed by Robert Shafer, Duane Jung, and Brad Betts (HRP-ASAP v1.1, Stanford University; http://hivdb.stanford.edu/hiv/).

Analysis of relative replicative fitness in virus mixtures.

Relative replicative fitness was analyzed in growth competition assays as previously described (8). Viruses were adjusted according to their CCID₅₀ values before mixtures were prepared. Briefly, a 300-μl inoculum of the two competing variants mixed at different ratios was used to infect 3.0 × 10⁵ MT-4 cells at an MOI of 0.001 as described above. After 5 to 6 days in culture, 200 μl of the supernatant from 2-ml cultures were used to reinfect a fresh aliquot of 3.0 × 10⁵ MT-4 cells. The relative proportion of the two competing variants was determined both at baseline and in each passage based on the ratios of the specific mutations. Ratios were estimated based on the relative peak heights in electropherograms obtained by automated DNA sequencing as previously described (13). The ratios of leucine and methionine in mixtures of L2S.4_M151L and L2S.4_M151 were determined based on the relative proportions of cytosine and adenosine at the first nucleotide position at codon 151, respectively. Similarly, the ratios of serine and glycine in mixtures of L1S.28_S68 and L1S.28_S68G were determined based on the relative proportions of thymidine and cytosine at the second nucleotide position at codon 68, respectively. To control for the possible spontaneous reversion of mutated codons, HIV-1 RT from cultures done with each virus separately was also sequenced after each viral passage.

Quantitation of RT activity in culture supernatants.

Levels of RT activity were quantitated in 10 μl of cell-free culture supernatants using the ultrasensitive PCR-based Amp-RT assay (9). Quantitation of Amp-RT signals was done by an enzyme-linked immunosorbent assay-based, nonradioactive oligoprobing system using a standard curve of known units of RT activity/milliliter as previously described (5).

RT drug susceptibility testing in recombinant viruses.

The susceptibility to AZT-triphosphate (AZT-TP), ddA-TP (the active form of ddI), and ddC-TP of RTs from recombinant viruses was determined enzymatically by measuring IC₅₀ values using the Amp-RT assay as previously described (6). Testing conditions included duplicate Amp-RT reactions done in the absence or presence of several concentrations of the TP form of the nucleoside analog (6). The concentration of each dNTP in the RT step of Amp-RT was variable depending on the nucleoside analog tested. A concentration of 15 μM dTTP was used for reactions containing AZT-TP, while 5 μM dCTP and 5 μM dATP were used for reactions done with ddC-TP and ddA-TP, respectively. The other three dNTPs were used at 20 μM each. The RT-generated cDNA was detected by PCR amplification as described previously after increasing the dNTP concentrations to 200 μM (6). AZT-TP was obtained from Moravek Biochemicals, Inc. (Brea, Calif.), and ddC-TP and ddA-TP were from Sigma Chemicals (St. Louis, Mo.).

RESULTS

Generation of recombinant viruses carrying the 151L or 151K mutation in a patient-derived WT genetic background and in HIV-1_HXB2.

We first investigated the effect of the 151L or 151K mutations in the WT RT background of isolate L1S. We introduced these mutations in two randomly selected RT clones from this isolate (clones L1S.28 and L1S.18) and determined virus production following cotransfection of MT-4 cells with an RT-deleted proviral molecular clone. The 151L and 151K mutations were also introduced in the RT from HIV-1_HXB2 as a control genetic background.

Table 1 shows the levels of p24 antigen and RT activity observed after 12 days of culture. The infectivity of recombinant viruses measured as CCID₅₀ values per milliliter is also shown. Transfections done with the WT RT clone L1S.28_Q151 resulted in high levels of both RT activity and p24 antigen. A high titer of infectious virus was also seen, indicating that this RT sequence did not compromise replication capability of HIV-1 (Table 1). In contrast, little or no RT activity or p24 antigen was observed in transfections done with either L1S.28_Q151L or L1S.28_Q151K. The low RT levels seen in L1S.28_Q151L and L1S.28_Q151K were not associated with detectable infectious virus in culture supernatants, suggesting that the 151L or 151K mutations profoundly affected the replication capabilities of these viruses. Comparison of the RT sequence from plasmid pL1S.28_Q151 with the sequences of pL1S.28_Q151L or pL1S.28_Q151K showed no mutations other than 151L or 151K (not shown), confirming that the observed lack of productive infection is due to a deleterious effect of the 151L or 151K mutation. Similar results were seen in two other mutants that were generated with an additional RT clone from isolate L1S (clone L1S.18) (Table 1), further supporting the deleterious effect of both 151L and 151K on virus replication. The two RT clones used (L1S.18 and L1S.28) were not identical and differed at four amino acids at codon positions 82 (K or E), 113 (D or N), 202 (I or L), and 220 (K or I). These amino acid changes had no effect in replication of HIV-1_151L or HIV-1_151K, and they may likely represent natural polymorphisms.

TABLE 1.

Levels of p24 antigen, RT activity, and infectious virus titers (CCID₅₀) in culture supernatants obtained following cotransfection of MT-4 cells with an RT-deleted clone and RTs from HIV-1_HXB2 or isolate L1S (clones L1S.28 and L1S.18) having Q, L, or K at codon 151^a

Recombinant virus	p24 antigen (pg/ml)	RT activity (U/ml)	CCID50/ml
HXB2Q151	1,600,000	8.38 × 10−1	390,625
HXB2Q151L	675	1.27 × 10−3	<25
HXB2Q151K	<7.8	2.58 × 10−8	<25
L1S.28Q151	1,410,000	5.78 × 10−1	390,625
L1S.28Q151L	<7.8	2.12 × 10−6	<25
L1S.28Q151K	<7.8	4.62 × 10−6	<25
L1S.18Q151L	<7.8	ND	<25
L1S.18Q151K	<7.8	ND	<25

^aCCID₅₀ values and p24 and RT levels were determined in a single viral stock supernatant collected 12 days after cotransfection except for HXB2_Q151 (10 days). ND, not detected. 

The effect of the 151L and 151K mutation was also determined using the RT genetic background from HIV-1_HXB2. High titers of p24 antigen, RT activity, and infectious virus were seen with a recombinant generated using the WT RT sequence from HIV-1_HXB2 (Table 1). In contrast, low or undetectable levels of p24 antigen and RT activity were seen in HXB2_Q151L and HXB2_Q151K, respectively. Both sequences were not associated with detectable infectious virus (Table 1), demonstrating that HXB2_Q151L and HXB2_Q151K replicate much less efficiently compared with HXB2_Q151. Taken together, our results indicate that the 151L or 151K mutation profoundly decreases the replication capability of HIV-1, suggesting that acquisition of Q151M may be limited by impaired replication of a 151L or 151K intermediate.

The RT genetic background of an isolate carrying the Q151M mutation supports the replication of HIV-1_151L but not HIV-1_151K.

To determine whether the genetic background associated with Q151M could support replication of either HIV-1_151L or HIV-1_151K, we mutated M151 to L or K in two different RT clones from isolate L2S (clones L2S.4 and L2S.16). Table 2 shows that the levels of p24 antigen and RT activity observed in recombinants carrying the 151L mutation (recombinants L2S.4_M151L and L2S.16_M151L) 12 days after cotransfection were similar to those observed in a parenteral virus having the Q151M mutation (recombinant L2S.4_M151), indicating that the genetic background associated with Q151M could support the 151L mutation without compromising replication capabilities. Both recombinant viruses L2S.4_M151L and L2S.16_M151L were able to reinfect MT-4 cells with kinetics of p24 antigen production similar to those of WT viruses or viruses carrying the Q151M mutation, indicating that these two viruses can replicate efficiently (Fig. 1). These findings also indicate that the amino acid changes observed between the two RT clones L2S.4 and L2S.16 (R or K at codon 83, and W or R at codon 153) have no effect on replication capacities and may likely represent natural polymorphisms. In contrast to the 151L mutation, cotransfections done with two different RT clones carrying the 151K mutation (L2S.4_M151K and L2S.16_M151K) resulted in little or no RT activity or p24 antigen, indicating that the 151K mutation was deleterious in these RT backgrounds (Table 2).

TABLE 2.

Levels of p24 antigen and RT activity in culture supernatants obtained following cotransfection of MT-4 cells with an RT-deleted molecular clone and two RT clones from isolate L2S (L2S.4 and L2S.16) having M, L, or K at codon 151^a

Recombinant virus	p24 antigen (pg/ml)	RT activity (U/ml)
L2S.4M151	1,550,000	9.19 × 10−1
L2S.4M151L	149,000	2.49 × 10−1
L2S.4M151K	<7.8	5.22 × 10−8
L2S.16M151L	2,500,000	6.84 × 10−1
L2S.16M151K	<7.8	1.04 × 10−5

^ap24 and RT levels were determined in a single viral stock supernatant collected 12 days after cotransfection. 

FIG. 1.

Replication kinetics of WT HIV-1 (HXB2_Q151 and L1S.28_Q151) and mutants carrying the 151L (L2S.4_M151L and L2S.16_M151L) or 151M (L2S.4_M151) mutation. Virus production was monitored overtime by measuring p24 antigen in supernatants. Mean p24 values from triplicate cultures are shown. Mock, uninfected cells.

HIV-1_151L has a significant replication disadvantage compared to HIV-1_151M.

We next compared replicative fitness of HIV-1_151L and HIV-1_151M to determine whether a low replication capability of HIV-1_151L could explain why viruses carrying the 151M mutation but not 151L are selected in vivo. Replicative fitness was determined by using a competitive HIV-1 replication assay of HIV-1_151L and HIV-1_151M and by measuring virus infectivity/total HIV-1 virion particle ratios. Total HIV-1 virion particles in culture supernatants were expressed as levels of p24 antigen or RT activity.

Virus infectivity/virion particle ratios measured in L2S.4_M151L and L2S.16_M151L were found to be 20- to 73-fold and 125- to 200-fold lower than the mean ratios seen in WT viruses, respectively, suggesting that the 151L mutation decreases replication capabilities in HIV-1 (Table 3). In contrast, no clear evidence of reduced replication capabilities was seen when HIV-1_151L was compared to HIV-1_151M by analysis of virus infectivity/virion particle ratios. Ratios measured in L2S.4_M151L and L2S.16_M151L were either similar to or 3- to 10-fold lower than those of L2S.4_M151. The decreased replication capability of HIV-1_151L compared to that of HIV-1_151M was more evident in a competitive HIV-1 replication assay of L2S.4_M151L and L2S.4_M151 (Fig. 2): L2S.4_M151 completely outgrew L2S.4_M151L after 12 days in culture, even in experiments started with 85% L2S.4_M151L and 15% L2S.4_M151. Cultures done with L2S.4_M151L or L2S.4_M151 alone showed no mutations at codon 151 after 12 days of cultivation, indicating stability of the 151L and 151M mutation (Fig. 2). These results demonstrate that replicative fitness of HIV-1_151L is lower than that of HIV-1_151M and may explain why only viruses carrying the Q151M mutation are observed in vivo (12, 22).

TABLE 3.

Virus infectivity (CCID₅₀)/virion particle ratios^a in recombinant viruses generated with wild-type RTs (HXB2_Q151 and L1S.28_Q151) or RTs carrying the 151M (L2S.4_M151) or 151L (L2S.4_M151L L2S.16_M151L) mutation

Recombinant virus	CCID50/U of RT activity (104)	CCID50/pg of p24 antigen (102)
HXB2Q151	46.6	24.4
L1S.28Q151	67.6	27.7
L2S.4M151	2.47	1.46
L2S.4M151L	0.78	1.31
L2S.16M151L	0.46	0.13

^aVirion particles are expressed as levels of p24 antigen or RT activity measured in a single viral stock supernatant collected 12 days after cotransfection. 

FIG. 2.

Competitive replication assay of HIV-1_151M (L2S.4_M151) and HIV-1_151L (L2S.4_M151L). Two experiments, each initiated at different proportions of the two viruses (85% L2S.4_M151L–15% L2S.4_M151 or 55% L2S.4_M151L–45% L2S.4_M151) are shown. The proportion of L at codon 151 is plotted over time. Day 0 represents proportions in the initial virus mixtures. The results of infections done with L2S.4_M151L and L2S.4_M151 separately are also shown.

The S68G mutation restores replication of HIV-1_151L in both HXB2 and a patient-derived WT genetic background.

We next compared RT sequences of isolates L1S and L2S to identify the genetic changes in isolate L2S that may have allowed replication of HIV-1_151L. Figure 3 shows all mutations present in L1S.18, L1S.28, L2S.4, and L2S.16. A consensus sequence for L1S and L2S is also shown. Each consensus sequence represents the amino acid most frequently observed in seven RT clones from isolates L1S and L2S. Three mutations at codon positions 35 (Val→Ile), 68 (Ser→Gly), and 178 (Ile→Met) were found in L2S.4 and L2S.16 compared to L1S.18 and L1S.28. Analysis of the frequency of each of these amino acid changes in the data base using the HRP-ASAP v1.1 program indicated that V35I and I178M are common mutations identified in 5 to 10% of both treated and untreated HIV-1-infected patients, suggesting that both mutations may be natural polymorphisms. The S68G mutation, however, has been previously found to be associated with Q151M in approximately 50% of patients who develop MDNR mutations (12, 22), suggesting that this mutation might play an important role in replication of viruses having mutations at codon 151.

FIG. 3.

Comparison between deduced amino acid sequences of the RT (amino acids 7 to 246) from two clones from isolate L1S (L1S.28 and L1S.18) or isolate L2S (L2S.4 and L2S.16). Consensus sequences of isolate L1S and L2S are also shown. Sequences are compared with the reference HIV-1_HXB2 strain. Asterisks indicate amino acids changes observed in L2S.4, L2S.16, and consensus L2S, but not in isolate L1S or HIV-1_HXB2.

To examine the effect of S68G on replication of HIV-1_151L, we determined whether the deleterious effect of the 151L mutation seen in the genetic background of L1S.28 could be reverted by introducing the S68G mutation. Figure 4A shows the mean levels of p24 antigen obtained in duplicate transfections done with the double mutant L1S.28_Q151L/S68G. Results of control duplicate transfections done with the nonmutated L1S.28 WT RT (L1S.28_Q151) or RTs carrying the Q151L or S68G mutations alone (L1S.28_Q151L and L1S.28_S68G, respectively) are also shown. Transfections done with L1S.28_Q151L did not result in detectable p24 antigen even after 20 days of culture. In contrast, transfections done with the double mutant Q151L/S68G (recombinant L1S.28_Q151L/S68G) resulted in detectable p24 antigen, although the kinetic of virus production was delayed compared to that of the control nonmutated WT virus or the virus carrying the S68G mutation alone (Fig. 4A). The effect of S68G on replication of viruses carrying the 151L mutation was not limited to L1S.28. Introduction of S68G in HXB2_Q151L also enhanced replication capability of this virus. Figure 4B shows the higher levels of p24 antigen following transfections with HXB2_Q151L/S68G compared to those observed in HXB2_Q151L during the entire culture. These results demonstrate that the S68G mutation increases replication of viruses carrying the 151L mutation and suggest a compensatory role for Q151L.

FIG. 4.

Effect of S68G on replication of HIV-1_151L. WT RTs from either L1S.28 (A) or HIV-1_HXB2 (B) were used to generate site-directed mutants at codon 151 or 68. MT-4 cells were transfected with 1 μg of proviral WT or mutant RT and 10 μg of the RT-deleted proviral molecular clone as indicated in Materials and Methods. Virus production was monitored by measuring p24 antigen levels in culture supernatants. The results are the mean values observed in duplicate transfections.

The S68G mutation is not deleterious in a WT RT genetic background.

To determine whether the S68G mutation alone could affect replicative fitness of WT HIV-1, we compared replicative fitness of L1S.28_S68G with that of L1S.28_S68 by measuring replication kinetics and by using a competitive HIV-1 replication assay. Figure 5A shows similar kinetics of p24 antigen production following reinfection of MT-4 cells with recombinants L1S.28_S68 and L1S.28_S68G, suggesting that the S68G mutation does not affect replication capability of HIV-1. Similar replicative fitness of both viruses was evident in competitive assays containing different mixtures of L1S.28_S68 and L1S.28_S68G (Fig. 5B). The proportion of L1S.28_S68G remained constant after 24 days in culture in all virus mixtures, thus confirming that S68G does not significantly affect replication capabilities in HIV-1. Taken together, these results indicate that the S68G mutation in a WT genetic background does not confer a significant replicative disadvantage or loss of fitness to the virus, suggesting that viruses carrying the S68G mutation would likely preexist in the virus population.

FIG. 5.

Replicative fitness of L1S.28_S68 and L1S.28_S68G. (A) Replication kinetics of L1S.28_S68 and L1S.28_S68G in MT-4 cells. Virus production was monitored overtime by measuring p24 antigen in supernatants. Mean p24 values from triplicate cultures are shown. (B) Competitive replication assay of L1S.28_S68 and L1S.28_S68G. Two experiments, each initiated at different proportion of the two viruses (80% L1S.28_S68G–20% L1S.28_S68, or 40% L1S.28_S68G–60% L1S.28_S68) are shown. Day 0 represents proportions in the initial virus mixtures. Results of infections done with L1S.28_S68 and L1S.28_S68G separately are also shown.

Susceptibility to nucleoside analogs of RTs from recombinant viruses carrying the 151L or 68G mutation.

We next sought to determine whether the 151L or 68G mutations could affect susceptibility to nucleoside analogs. Table 4 shows that RT susceptibility to AZT-TP, ddA-TP, and ddC-TP in a recombinant virus carrying the 151L mutation (L2S.4_M151L) was six- to eightfold lower than that of WT HIV-1 (L1S.18). Interestingly, IC₅₀ values observed in L2S.4_M151L were similar to those of L2S.4_M151 for the three drugs tested, indicating that both 151L and 151M confer similar levels of MDNR.

TABLE 4.

Susceptibility to AZT-TP, ddA-TP, and ddC-TP or RT in recombinant viruses containing WT (L1S.28) or mutant RTs carrying the S68G (L1S.28_S68G), 151M (L2S.4_M151), or 151L (L2S.4_M151L) mutation

Recombinant virusa	IC50, μMb
	AZT-TP	ddA-TP	ddC-TP
L1S.28	0.23	0.46	0.44
L1S.28S68G	0.16	0.28	0.49
L2S.4M151	1.41 (6.1)	2.95 (6.3)	2.87 (6.5)
L2S.4M151L	1.84 (8)	2.93 (6.3)	3.61 (8.2)

^aRecombinants were adjusted by their levels of RT activity (10⁻⁷ U) before IC₅₀ determination. 

^bValues in parenthesis are fold resistance (fold increase in IC₅₀ relative to the IC₅₀ value of L1S.28). 

We also compared RT susceptibility in L1S.28 and L1S.28_S68G. Table 4 shows similar IC₅₀ values for AZT-TP, ddA-TP, and ddC-TP in L1S.28_S68G and L1S.28, indicating that the S68G mutation does not affect the susceptibility to these drugs.

DISCUSSION

We investigated whether decreased replication capabilities of viruses carrying the two intermediates between Q and M at codon 151 might explain the lower frequency of selection of MDNR compared to the classical AZT resistance pathway mediated by the T215Y mutation. We have focused on the 151L and 151K intermediates since preexistence of HIV-1_151L and/or HIV-1_151K could allow emergence of Q151M via a single-base substitution. We also wanted to investigate the effect that the RT genetic background has on the replication of both HIV-1_151L and HIV-1_151K to identify possible viral determinants that might influence selection of the Q151M mutation.

Several reports have shown that viruses carrying intermediates of the T215Y mutation, such as T215S and T215N, replicate efficiently and that these mutations confer little or no fitness cost in the absence of AZT, thus facilitating the development of AZT resistance through this pathway (7, 13, 14). However, in contrast to T215S or T215N, our results indicate that both the 151L and 151K intermediates are lethal in all WT RT genetic backgrounds tested, including those obtained from a WT patient-derived isolate or from HIV-1_HXB2. The decreased replication capabilities of these two intermediates may imply that two base transversions are needed to evolve from the WT Q to M at codon 151, which may be a relatively infrequent event. Similar findings and conclusions were also reported by Kosalaraksa et al. who analyzed the effect of 151L and 151K in the HIV-1_HXB2 genetic background only (13). However, our findings on the effect of the 151L mutation in the genetic background associated with Q151M provide new information that suggests a role for Q151L as a potential intermediate of Q151M. Several observations suggest that Q151L could be a viable intermediate. First, the observed reversion of the deleterious effect of 151L in the 151M-derived RT background indicated that replication of HIV-1_151L is possible. Second, we found that the 151L mutation confers MDNR, which can favor the selection of viruses with this mutation in patients treated with AZT and ddC or ddI. Third, our results showed that the replicative fitness of HIV-1_151L is lower than that of HIV-1_151M, further supporting a role of 151L as a transient intermediate. The higher replicative fitness of HIV-1_Q151M may explain the frequent observation in vivo of Q151M but not Q151L (12, 22–26).

Our data also indicate that replicative fitness of HIV-1_151L depends on the presence of at least one compensatory mutation. We demonstrated the compensatory role of S68G by showing that introduction of this mutation partially restores replication capacity of HIV-1_151L in WT genetic backgrounds. The S68G mutation might therefore provide a “bridge” across the adaptive “valley” caused by the 151L mutation. These findings indicate a dependence of 151L on other mutations such as S68G for replication which may explain the low frequency of the Q151M-mediated pathway of resistance.

Several observations suggest that S68G may represent a preexisting polymorphism. First, the S68G mutation only requires a single base substitution (AGU→GGU), which is a relatively frequent event (4, 17). Second, fitness assays showed that S68G does not confer a significant replicative disadvantage to the virus, indicating that this mutation is near neutrality. Third, HIV-1_S68G has WT susceptibility to AZT, ddC, and ddI, as expected from a natural polymorphism. Fourth, S68G is present in ∼3% of untreated patients, as indicated by our analysis of HIV-1 sequence databases.

The compensatory role of S68G in HIV-1_151L demonstrated in our study in both a patient-derived virus and in HIV-1_HXB2 may provide a mechanism by which other patients acquire Q151M. The frequency of S68G in treated individuals is significantly higher among patients who carry the Q151M mutation than in those having the classical AZT resistance mutations (12, 22). About 50% of patients with Q151M also have S68G, supporting a role of this mutation in the acquisition of Q151M, possibly through 151L. However, the fact that many patients with MDNR phenotypes do not show evidence of a S68G mutation may suggest either a transient role for S68G or the involvement of alternative compensatory mutations (12, 22).

Our findings on the effect of the S68G mutation in replication of HIV-1_151L emphasize the importance of the genetic background in determining the fitness of mutations. The role of the genetic background in replication of viruses carrying different mutations associated with resistance to protease and RT inhibitors has been previously documented. For instance, an isoleucine at codon 10 of the protease has been found to be critical for allowing replication of viruses carrying several protease resistance mutations, and a proline at codon 63 compensates for the deleterious effect of a L90M mutation (16, 20). Similar compensatory effects on virus replication have been observed in the RT for the G190E and L74V mutations or for the Y115W and M230I mutations (1, 18).

Selection of the Q151M or T215Y pathways of resistance may also be influenced by other factors such as virus population size. The higher fitness of HIV-1_Q151M compared to HIV-1_T215Y observed in several studies (13, 15) implies that if just one viable Q151M mutant appears before fixation of T215Y, then Q151M should displace T215Y, thus leading to MDNR. As population size increases, the time required for fixation of T215Y should also increase. A large population size might also result in an increased number of replication-competent HIV-1_151L. Thus, the probability of appearance of Q151M may increase as the effective viral population size increases.

In conclusion, we show evidence that suggest that 151L but not 151K is a potential intermediate of Q151M. We demonstrate that replication capabilities of HIV-1_151L are dependent on the presence of other polymorphisms such as S68G, which may explain the low frequency of selection of MDNR mediated by Q151M.