Zidovudine (AZT) Monotherapy Selects for the A360V Mutation in the Connection Domain of HIV-1 Reverse Transcriptase

Background We previously demonstrated in vitro that zidovudine (AZT) selects for A371V in the connection domain and Q509L in ribonuclease H (RNase H) domain of HIV-1 reverse transcriptase (RT) which, together with the thymidine analog mutations D67N, K70R and T215F, confer greater than 100-fold AZT resistance. The goal of the current study was to determine whether AZT monotherapy in HIV-1 infected patients also selects the A371V, Q509L or other mutations in the C-terminal domains of HIV-1 RT. Methodology/Principal Findings Full-length RT sequences in plasma obtained pre- and post-therapy were compared in 23 participants who received AZT monotherapy from the AIDS Clinical Trials Group study 175. Five of the 23 participants reached a primary study endpoint. Mutations significantly associated with AZT monotherapy included K70R (p = 0.003) and T215Y (p = 0.013) in the polymerase domain of HIV-1 RT, and A360V (p = 0.041) in the connection domain of HIV-1 RT. HIV-1 drug susceptibility assays demonstrated that A360V, either alone or in combination with thymidine analog mutations, decreased AZT susceptibility in recombinant viruses containing participant-derived full-length RT sequences or site-directed mutant RT. Biochemical studies revealed that A360V enhances the AZT-monophosphate excision activity of purified RT by significantly decreasing the frequency of secondary RNase H cleavage events that reduce the RNA/DNA duplex length and promote template/primer dissociation. Conclusions The A360V mutation in the connection domain of RT was selected in HIV-infected individuals that received AZT monotherapy and contributed to AZT resistance.


Introduction
Zidovudine (39-azido-39-deoxythymidine, AZT) was the first antiviral drug approved by the U.S. Food and Drug Administration for the treatment of HIV infection. AZT is a nucleoside reverse transcriptase (RT) inhibitor (NRTI) that, after metabolism by cellular kinases to its triphosphate form (TP) in cells, competes with the natural substrate TTP for binding and incorporation by HIV-1 RT into the nascent viral DNA. Because AZT lacks a 39-OH on the ribose sugar, its incorporation into viral DNA results in chain-termination.
HIV-1 resistance to AZT was first reported in 1989 [1]. Resistance was conferred by mutations in the polymerase domain of RT that included D67N, K70R, T215Y/F and K219Q. Subsequently, two additional AZT resistance mutations were identified: M41L and L210W [2,3,4]. After the discovery that d4T (29,39-didehydro-29,39-dideoxythymidine) could select the same mutations, the term thymidine analog mutations (TAM) was adopted to reflect their role in resistance to both AZT and d4T. In general, each TAM alone confers between 1.5-to 4-fold resistance, and multiple mutations are required for high-level resistance [5]. The most common combination of mutations selected includes M41L, L210W and T215Y and excludes K70R (TAM-1 pathway). A second pattern includes D67N, K70R, T215F and K219Q/E (TAM-2 pathway) [5]. Biochemical studies have demonstrated that RT containing TAMs shows an increased capacity to unblock AZT-monophosphate (MP) terminated primers in the presence of physiological concentrations of ATP [6,7]. In this regard, structural studies have shown that TAMs enhance the binding and/or placement of ATP in the HIV-1 RT active site [8].
Recent studies have demonstrated that AZT resistance can be increased by mutations in the connection domain of HIV-1 RT. For example, the G333D/E, G335C, N348I, A360V/I, T369V and A371V mutations have all be shown to augment AZT resistance alone or in combination with TAMs [9,10,11,12]. However, all of these mutations were identified in RT sequences from participants exposed to multiple RT inhibitors and it is not known if these connection domain mutations were selected specifically by AZT. For example, Yap et al reported that the N348I mutation is highly associated with TAMs, the lamivudine mutation M184V/I, and the non-nucleoside inhibitor resistance mutations K103N and Y181C/I [13]. Consistent with this observation, mechanistic analyses suggest that this mutation may compensate for the antagonism of TAMs by M184V/I and Y181C [14,15].
We previously carried out in vitro selection experiments by serial passage of HIV-1 in increasing concentrations of AZT to determine whether AZT selects mutations in the connection and/or ribonuclease H (RNase H) domains of RT [16]. Two novel mutations -A371V in the connection domain and Q509L in the RNase H domain -were selected in combination with D67N, K70R and T215F that together conferred greater than 100-fold AZT resistance. The goal of the current study was to determine whether AZT monotherapy in HIV-1 infected participants selects the A371V, Q509L or other mutations in the connection or RNase H domains of RT.

Changes in serum HIV-1 RNA with AZT monotherapy
Pre-therapy and last-on-therapy serum samples were available from 5 participants who reached a primary study endpoint (see Methods) on AZT monotherapy and from 18 participants who did not reach a primary study endpoint on AZT monotherapy. The mean pre-therapy HIV-1 RNA concentration of participants reaching a study endpoint was 33,261 copies/mL (range 3,650-67,134 copies/mL), which was not significantly different from that of participants not reaching a study endpoint (mean 41,002 copies/mL; range 1,439-168,576 copies/mL; p = 0.875). There was a strong trend (p = 0.06) toward higher serum HIV-1 RNA concentration in last-on-therapy samples from participants reaching a primary study endpoint (mean 87,264 copies/mL; range 35,227-216,219 copies/mL) compared with those who did not (mean 37,424 copies/mL; range 826-138,776 copies/mL). Table 1 shows the mutations in HIV-1 RT that emerged in participants who received AZT monotherapy. TAMs in the polymerase domain of RT emerged in 5 of 5 participants reaching a study endpoint compared with 11 of 18 (61%) of participants who did not reach a study endpoint. Although each of the known TAMs was identified in one or more on-treatment samples, only the K70R and T215Y mutations were significantly associated with AZT monotherapy. Of note, the A360V in the connection domain of HIV-1 RT was also significantly associated with AZT monotherapy and was more frequent than the M41L, D67N, L210W or K219E/Q mutations. A360V emerged in 2 of 5 participants who reached a study endpoint and in 4 of 18 participants who did not. Among the 6 on-treatment samples that contained A360V, it was a pure mutant (100%) in 2 samples and a mixture (,50%) in 4 samples, as assessed by population sequencing ( Table 2). Analysis of longitudinal samples in the 4 participants who did not reach study endpoint revealed that A360V generally emerged at the same time as TAMs, although in one participant it was the first mutation to emerge (Table 3). In one of the participants who reached a study endpoint, A360V emerged at the same time as TAMs, however in the other participant it clearly emerged after TAMs ( Table 3). The connection domain mutation A371V was more frequent in laston-therapy samples compared to pre-therapy samples, but this difference was not statistically significant (p = 0.25). The N348I and Q509L mutations were not detected in any of the 23 ontreatment samples.

Phenotypic analyses of HIV-1 containing A360V
To determine whether the A360V mutation conferred AZT resistance, we first cloned full-length RTs from participant samples 4, 11 and 29 into pxxLAI-3D, and selected a single clone for phenotypic analyses. Recombinant virus derived from participant sample 29 did not replicate in cell culture, and was excluded from subsequent analyses. The genotypes of the individual clones and population sequences obtained from the participant serum are shown in Table 2. The 360 V mutation in each of the two replication competent viruses was reverted to A360 by mutagenesis and the paired (A360/360 V) recombinants were tested for susceptibility to AZT (Table 4). Both recombinant viruses with RT sequences derived from on-treatment samples were resistant to AZT. Importantly, when the A360V mutation was reverted to wildtype, statistically significant decreases in AZT resistance were observed for each of the virus pairs (Table 4). We also generated a full-length RT and polymerase domain only bulk plasmid population for sample 3, since individual clones did not grow in cell-culture (

Mechanism by which A360V confers AZT resistance
TAMs in HIV-1 RT confer AZT resistance by enabling the enzyme to excise the chain-terminating AZT-MP moiety from the 39-end of the DNA primer using ATP as a phosphate donor [7]. Previous biochemical studies demonstrated that the N348I and Q509L mutations in HIV-1 RT indirectly increase AZT resistance by decreasing the frequency of secondary RNase H cleavages that reduce the RNA/DNA duplex length of the T/P and diminish the efficiency of AZT-MP excision [13,14,23,24,26]. A similar phenotype has been proposed for the A360V mutation; however, the biochemical analyses reported used RTs that also contained N348I [26]. Therefore, we performed biochemical experiments to probe the mechanism(s) by which A360V decreased AZT susceptibility. In this regard, we first assessed the AZT-MP excision activity of the wildtype, TAM-1 and TAM4-1/A360V enzymes on a well-defined DNA/DNA and RNA/DNA T/P substrate that is routinely used in our laboratory [13,14,24,26]. As described previously [13,14], TAM-1 HIV-1 RT unblocked AZT- Table 2. Mutations and polymorphisms associated with A360V.  MP chain-terminated primers efficiently on both DNA/DNA and RNA/DNA T/Ps ( Figure 1A, 1B). Introduction of the A360V mutation into either wildtype or TAM-1 RT increased the enzyme's excision activity on the RNA/DNA T/P but not on the DNA/DNA T/P ( Figure 1A, 1B). We previously delineated the relationship between AZT-MP excision efficiency and RNase H activity on the RNA/DNA T/P substrate used in these experiments [23,24]. These studies showed that the primary polymerase-dependent RNase H cleavage of RT does not impact the enzyme's AZT-MP excision efficiency, but polymeraseindependent RNase H cleavages that reduce the RNA/DNA duplex length to less than 12 nucleotides abolish AZT-MP excision activity [23,24]. In light of these data, we next evaluated the RNase H activity of the wildtype and mutant RTs that occurred during the ATP-mediated excision reactions described in Figure 1B. The A360V mutation was found to significantly reduce the frequency of a polymerase-independent cleavage event that decreases the RNA/DNA duplex to 10 nucleotides ( Figure 1C, 1D). Taken together, these findings are consistent with the concept that A360V in HIV-1 RT impacts the efficiency of the AZT-MP excision reaction by an RNase H-dependent mechanism.

Discussion
Recent studies show that RT inhibitors select for drug resistance mutations in the connection domain of RT including N348I and the A360V mutation reported here. The drug or drugs responsible for the selection of these connection domain mutations, however, are not well defined. For example, Yap et al reported that failure of AZT and nevirapine therapy was associated with the emergence of N348I, TAMs, and the non-nucleoside RT inhibitor resistance mutations K103N, V108I, Y181C/I and G190A/S [13]. By contrast, von Wyl et al reported that N348I predominantly emerged in participants receiving lamivudine and AZT [15], whereas Hachiya et al reported that N348I emerged on AZTand/or ddI-containing therapies [27]. Of note, the A360V mutation has not been associated with failure of any specific drug combinations [28]. The limitation of these prior studies is that they relied on retrospective analyses of sequences in clinical databases. As such, RT sequences were obtained from participants exposed to multiple RT inhibitors, without pretherapy or longitudinal sequences available, making it difficult to pinpoint the specific drug or drugs responsible for mutation selection.   In the current study, we compared paired pre-and on-therapy full-length RT sequences from 23 participants who had received only AZT monotherapy in ACTG study 175. Our analyses reveal that the A360V mutation in the connection domain emerged with similar or higher frequency as TAMs M41L, D67N, L210W and K219E/Q. Longitudinal analyses also showed that A360V was detected at the same time or after the appearance of TAMs but not before. Importantly, we also show that the A360V mutation confers ,2.0-fold AZT resistance both in the full-length RT sequence context it was selected in vivo and in the xxLAI molecular clone. This is the first study to take a systematic experimental approach to determine whether AZT alone selects for mutations in the C-terminal domains of HIV-1 RT. As a result, both the selection of A360V during AZT monotherapy and resistance to AZT from A360V are unequivocally shown.
We previously reported that in in vitro selection experiments HIV-1 selected for two novel mutations -A371V in the connection domain and Q509L in the RNase H domain -in combination with D67N, K70R and T215F that together conferred greater than 100-fold AZT resistance [16]. The A360V mutation was not selected. In the 23 on-treatment samples sequenced in the current study, we noted a strong trend (p = 0.06) toward selection of A371V with AZT monotherapy. By contrast, the Q509L mutation was not present in any of the sequences. These findings indicate that the drug pressure and other environmental factors that promote resistance selection differ between participants and cell culture and the results of in vitro experiments cannot always be extrapolated in vivo.
Of note, the N348I mutation was not identified in any of the sequences in this study. Although our sample size is small (N = 23), this is surprising given that N348I and A360V mutations decrease AZT susceptibility to the same extent (this study, [13]). Furthermore, the phenotypic mechanisms by which these mutations confer AZT resistance are similar (this study, [13,26]) in that they both diminish the frequency of secondary RNase H cleavage events that reduce the RNA/DNA duplex length of the T/P and diminish the efficiency of AZT-MP excision. Additional virological and biochemical analyses may provide insights into this conundrum.
In summary, this study of full-length RT sequences from paired pre-and post-therapy samples shows that the A360V mutation in the connection domain of RT emerges on AZT monotherapy with similar frequency as TAMs in the polymerase domain and confers ,2-fold resistance to AZT through an excision-promoting molecular mechanism. These findings provide additional insight into the importance of the connection domain of RT in resistance to NRTI and NNRTI.

Participant Samples
All samples were obtained from participants enrolled in the AIDS Clinical Trials Group (ACTG) study 175 (ClinicalTrials.gov number, NCT00000625) [17]. ACTG 175 was a randomized, double-blind, placebo-controlled trial designed to compare AZT or didanosine monotherapy to AZT/didanosine or AZT/ zalcitabine combination therapy. Participants provided written informed consent and the study was approved at each site by an institutional review board. In Text S1, we list the institutions and investigators that participated in ACTG 175. ACTG approval was sought for the current study and was obtained through the appropriate New Works Concept Sheet (NWCS 304).The primary study endpoint was defined by: i) confirmed CD4 count 50% below the average of two pre-treatment counts; ii) AIDS related endpoints defined by the 1987 CDC criteria [18]; or iii) death.
For the current study, samples were selected from participants who were randomized to receive AZT monotherapy, had reached a study endpoint and for whom stored serum samples were available at pre-therapy (week 0) and from at least 1 of 3 longitudinal time points (weeks 8, 20 or 32). Samples were also obtained from participants who were randomized to receive AZT monotherapy but did not reach a study endpoint. HIV-1 RNA in serum was quantified using the Roche Amplicor Monitor HIV-1 Ultrasensitive Assay Kit 1.5 (Roche Diagnostics Corporation, Indianapolis, IN).

Production of infectious recombinant virus containing participant-derived RT sequences
The full-length infectious HIV-1 clone pxxLAI [20] was modified by site-directed mutagenesis to introduce silent mutations into RT at codons 321-323, 358, 417-418, 554 and in integrase at codons 29-31 to generate the unique restriction sites BstBI, MluI, HpaI, NgoMIV and SgrAI, respectively (pxxLAI-3D). Next, an RT defective cloning vector pxxLAI-3DDnp was developed by deletion of RT polymerase domain codons 113 to 315, using restriction sites NsiI to PflmI, as described previously [20]. To clone participant derived full-length RTs into pxxLAI-3DDnp, the restriction enzyme sites BclI and SgrAI were first added to the 59 and 39 ends of genotyping amplicons using the primers: Bclforward (59-GTTTTATCAAAGTAAGACAGTATGATCAGA-TAC-39) and Sgr-reverse (59-CTACCACCGGTGGCAGGTTA -39). PCR products and pxxLAI-3DDnp were digested with BclI and SgrAI to form 1.89 kb full-length RT inserts and a 9.87 kb pxxLAI-3D cloning vector. Vector and full-length RT inserts were ligated with T4 DNA ligase (NEB, Ipswich, MA) and transformed into C2925 cells (NEB, Ipswich, MA). Individual recombinant clones or bulk plasmid populations were isolated and DNA sequenced to confirm similarity with the RT sequences in the serum sample from which they were derived. To clone the DNA polymerase domain of participant derived RTs into pxxLAI-3DDnp, the restriction enzyme sites BclI and BstBI were first added to the 59 and 39 ends of genotyping amplicons using the primers: Bcl-forward (59-GTTTTATCAAAGTAAGACAGTAT-GATCAGATAC-39) and Bst-reverse (59-CTATTAAGT-CTTTCGAAGGGTCATAATAC -39). PCR products and pxxLAI-3DDnp were digested with BclI and BstBI to form 1.08 kb full-length RT inserts and a 10.68 kb pxxLAI-3D cloning vector. Vector and full-length RT inserts were ligated and transformed, as described above. The bulk plasmid population of participant-derived, RT polymerase domain recombinant clones was isolated and DNA sequenced to confirm similarity with the RT sequences in the serum sample from which they were derived.
Site-directed mutagenesis (QuickChangeH II XL Site-Directed Mutagenesis Kit; Agilent, La Jolla, CA) was used to revert the 360 V mutation (codon GTC) to the wildtype amino acid (360 A; codon GCC) in participant-derived full-length recombinant clones. Individual recombinant clones were then isolated and sequenced to confirm reversion. Infectious virus for AZT susceptibility assays was generated and titered as described previously [21].

Production of site-directed mutants containing A360V
The A360V mutation was also introduced by site-directed mutagenesis (QuickChangeH II Site-Directed Mutagenesis Kit; Agilent, La Jolla, CA) into the backbone of wildtype HIV-1 RT and into RTs that contained the D67N/K70R or M41L/L210W/ T215Y (TAM-1) mutations in the p6HRT plasmid. The mutated RT sequences were then cloned into pxxLAI and infectious virus was generated and titered as described previously [21].

Drug susceptibility testing
The AZT susceptibility of participant-derived recombinant virus and site-directed mutants were determined in P4/R5 cells as described previously [21]. Briefly, three-fold dilutions of inhibitor were added to P4/R5 cells in triplicate and cells were infected with an amount of virus that produced ,100 relative units of light (RLU) in no drug control wells. After 48 hours, cells were lysed (Gal-Screen; Tropix/Applied Biosystems, Foster City, CA) and the RLU was measured using a ThermoLabSystems luminometer (Waltham, MA). The concentration of drug required to inhibit viral replication by 50% (EC 50 ) was determined from 3-11 independent experiments.

Biochemical analyses of recombinant HIV-1 RT containing A360V
HIV-1 RTs containing the A360V, TAM-1 or TAM-1/A360V mutations were purified to homogeneity as described previously [22]. The ability of the recombinant enzymes to facilitate the ATP-mediated removal of AZT-monophosphate (MP) from chainterminated DNA/DNA and RNA/DNA template/primers (T/Ps) using assay systems and conditions that have been studied extensively in the Sluis-Cremer laboratory [13,14,23,24]. For these assays a 26-nucleotide DNA primer (pr26; 59-CCTGTTCGGGCGCCACT GCTAGAGAT-39) was 59-radiolabeled with [c-32 P]ATP and chain-terminated with AZT-MP to generate P AZT . P AZT was then annealed to either a 35-nucleotide DNA (T DNA ; 59-AGAATGGAAAATCTCTAGCAGT GGCGC-foCCGAACAG-39) or RNA (T RNA : 59-AGAAUGGAAAAUCU-CUAGCAGUGGCGCCCGAACAG-39) template. Excision assays were carried out by first incubating 20 nM T RNA /P AZT or T DNA /P AZT with 3 mM ATP, 10 mM MgCl 2 , 1 mM dTTP and 10 mM ddCTP in a buffer containing 50 mM Tris-HCl (pH 7.5) and 50 mM KCl. Reactions were initiated by the addition of 200 nM wildtype or mutant RT. Aliquots were removed at defined times and analyzed as described previously [13,14,23,24]. RNase H activity was evaluated using the same AZT-MP chainterminated RNA/DNA T/P substrate described above, except the 59-end of the RNA was 32 P-end-labelled. Assays were carried out using 20 nM T RNA /P AZT , 3 mM ATP and 10 mM MgCl 2 in a buffer containing 50 mM Tris-HCl (pH 7.5) and 50 mM KCl. Reactions were initiated by the addition of 200 nM wildtype or mutant HIV-1 RT.

Statistical Analyses
The statistical significance of HIV-1 RNA levels in pre-therapy and last-on-therapy serum samples was calculated using the Mann-Whitney rank sum test. To identify mutations selected by AZT monotherapy, paired pre-therapy and last-on-therapy sequences from each participant were examined for known NRTI resistance mutations in the polymerase domain using the International AIDS Society -USA (IAS-USA) 2010 resistance table [25] and for novel mutations in the polymerase, connection and the RNase H domains of RT. The statistical significance (p,0.05) of differences in mutant frequency between pre-therapy and last-on-therapy sequences was determined using a two-sided exact McNemar's test. Due to small sample size, statistical significance was determined using N = 23 (5 participants who reached a study endpoint and 18 participants who did not) and pvalues were unadjusted for multiple comparisons. For drug susceptibility data, the concentration of drug required to inhibit viral replication by 50% (EC 50 ) were log 10 transformed and compared for significant differences (p-value,0.05) using the twosample Student's t test.

Supporting Information
Text S1 AIDS Clinical Trials Group Study 175 Protocol Team. (DOCX)