Differential In Vitro Kinetics of Drug Resistance Mutation Acquisition in HIV-1 RT of Subtypes B and C

Background HIV-1 subtype B is the most prevalent in developed countries and, consequently, it has been extensively studied. On the other hand, subtype C is the most prevalent worldwide and therefore is a reasonable target for future studies. Here we evaluate the acquisition of resistance and the viability of HIV-1 subtype B and C RT clones from different isolates that were subjected to in vitro selection pressure with zidovudine (ZDV) and lamivudine (3TC). Methods/Principal Findings MT4 cells were infected with chimeric virus pseudotyped with RT from subtype B and C clones, which were previously subjected to serial passage with increasing concentrations of ZDV and 3TC. The samples collected after each passage were analyzed for the presence of resistance mutations and VL. No differences were found between subtypes B and C in viral load and resistance mutations when these viruses were selected with 3TC. However, the route of mutations and the time to rebound of subtype B and C virus were different when subjected to ZDV treatment. In order to confirm the role of the mutations detected, other clones were generated and subjected to in vitro selection. RT subtype B virus isolates tended to acquire different ZDV resistance mutations (Q151M and D67N or T215Y, D67D/N and F214L) compared to subtype C (D67N, K70R, T215I or T215F). Conclusions/Significance This study suggests that different subtypes have a tendency to react differently to antiretroviral drug selection in vitro. Consequently, the acquisition of resistance in patients undergoing antiretroviral therapy can be dependent on the subtypes composing the viral population.


Introduction
Human immunodeficiency virus type 1 (HIV-1) can be segregated into several groups, subtypes, sub-subtypes and circulating recombinant forms (CRF) as a consequence of its genetic diversity [1]. While subtype B predominates in the developed world, other non-B subtypes or CRF are responsible for most of the infections occurring in developing countries [2]. Of note, subtype C is responsible for over 60% of global HIV-1 infections, as this variant can be found in countries with the highest known prevalence in sub-Saharan Africa and in highly populated countries such as India and China [2].
HIV-1 subtypes may have different biological characteristics, and may respond differently to diagnostic, immunologic and therapeutic interventions [6]. In regard to HIV antiretroviral (ARV) treatment, several studies have shown that HIV-1 subtype-specific differences influence the in vitro susceptibility as well as the resistance mutations selected upon treatment with specific drugs [7][8][9]. Furthermore, HIV-1 subtypes may also differ in the rates of mutation selection and fixation during ARV exposure [10].
In this work we studied the in vitro behavior of viruses carrying subtype B or C RT under the selective pressure of two NRTIs, ZDV and 3TC. We could observe a different mutational pattern upon ZDV exposure for subtype B RT variants (TAM-1 pathway; Q151M complex mutations) as compared to subtype C RT variants (TAM-2 pathway). On the other hand, 3TC displays a similar mutational behavior among these two subtypes.

In vivo Recombinant Virus Generation and in vitro Selection (RT B and RT C Virus)
A recombinant virus assay technology was used to generate two identical viruses differing solely in the RT palm-finger region (RT codons . Donor subtype B sequence was obtained from the pNL43 infectious clone. Subtype C sequence belongs to drugnaïve isolate from Brazil [11]. Several differences in the RT gene sequence could be observed between these isolates (V35T, E36A,  T39D, K43R, S48T, V90F, Q102R, D121Y, K122E, D123N,  S134I, I135T, C162S, E169K, K173N, Q174K, D177E, T200A,  Q207E, R211K, and F214L); none of them were previously related to NRTI resistance (Table 1). Samples from RT gene fragment amplification were co-transfected into MT4 cells (CD4+ T lymphocyte lineage, NIH-USA) with the DDNApolimeraseRT HXB2 BstEII-linearized plasmid carrying RT deleted (DRT) HIV-1 HXB2 genomic DNA, generating a chimeric virus by homologue recombination [12]. Before starting the drug selection, sequencing was performed to confirm the absence of drug resistance mutation (DRM) as well as the integrity of the RT gene. At day 7 post-transfection, supernatants were collected, frozen and used for tissue culture infectious dose 50% (TCID 50 ) determination. Recombinant viruses carrying RT from either subtype were evaluated in vitro for the kinetics of the acquisition of mutations leading to 3TC and ZDV resistance. HIV-1 molecular subtype B and C chimeric clones (RT B and RT C ) were used to infect MT4 cells during the selection process with a multiplicity of infection (MOI) of 0.001. Cells were subsequently resuspended in RPMI medium supplemented with 10% FBS and 0.001 mM ZDV or 0.020 mM 3TC graciously donated by NIH. Drug concentration was increased two-fold after each passage until it completely inhibited virus replication. The viral load (VL) was estimated by Taq ManH real time PCR using a quantitative RT-PCR reaction as previously described [13]. The viral RNA was extracted from the culture and cDNA synthesis was performed. The PCR target was located in the U5 region of HIV-1 59 LTR region using HIV-1 specific oligonucleotides and probe [13]. Calibration curves were generated using a HIV-1 subtype B supernatant with a known VL (10 6 , 10 5 , 10 4 , 10 3 , and 10 2 viral particles/mL). At least three dilutions of every sample were assayed, and data sets in which the linear correlation coefficient of the standard curve was less than 0.98 were discarded.
Cloning of RT Sequence into a HXB2 Infectious Clone (RT B9 and RT C9 Virus) Two vectors were used to clone different fragments of the wild type subtype B and C RT sequence into two molecular clones (pHXB2DRT and pHXB2DNRT) following the methodology previously described [14,15]. The plasmid pHXB2DRT contains the complete genome of HXB2 except for a deletion between codons 25 and 554 in the RT gene. Two unique restriction sites, MluNI and NgoMIV, were inserted in this region through an adapter. The second plasmid, pHXB2DNRT, contains the HXB2 genome carrying a deletion of the N-terminal region of the RT connection domain (codon 25 to 315). It also contains two distinct restriction enzyme sites, MluNI and Van91I, which were inserted for cloning purposes.
The NL4-3 (subtype B) clone RT coding fragment was amplified with primers RT2569 and RT22 in the first PCR round (1744 bp), followed by RTball and NgoMIV-INT1rev in the second round (1623 bp) [13]. Therefore, the PCR product will contain the same flanking restriction enzyme sites, MluNI and NgoMIV, as in the vector pHXB2DNRT. To clone the full length Table 1. Summarized data of chimeric virus utilized in in vitro selections. RT, both the plasmid pHXB2DRT (codon 25 to 555) and the PCR product were quantified and digested with 10U of restriction enzyme MluNI at 37uC for 1 h. After digestion, products were repurified and digested with 10 U of restriction enzyme NgoMIV for another 1 h at 37uC. Finally, the vector and the PCR products were ligated with the enzyme T4 DNA ligase (Invitrogen, USA) and digested with the restriction enzyme AspI to avoid re-ligation of original vector. For subtype C RT cloning, a sample from a naïve patient from the south of Brazil (C23) was amplified with primers RT2569 and 39RTAA339 in the first PCR round (967 bp) and RTball and RT21 in the second round (941 bp) as previously described [14]. To clone the PCR fragment, we used the same strategy explained above, changing NgoMIV for Van91I in the PCR fragment digestion step. The ligation reaction was used to transform E. coli JM109 and HXB2 RTpNL 4.3 (RT B9 ) and HXB2 NRTC23 (RT C9 ), positive clones were purified by maxiprep (QIAGEN) for subsequent transfection into MT-4 cells to generate viral stocks. Several differences could be observed between RT B9 and RT C9 (V35T, E36A, T39D, K43R, S48T, Q102K, D121Y, K122E, C162S, K173N, Q174K, D177E, T200A, Q207E, R211K); however, none of these differences were previously associated with NRTI resistance ( Table 1). All viral stocks were titrated through tissue culture infectious dose 50% (TCID 50 ) prior to infection.

In vitro Selection of Subtype B and C RT Infectious Clones (RT B9 and RT C9 )
The initial drug concentration for the in vitro selection process was determined based on EC 50 calculated for each inhibitor in our assays, which were 0.06 mM and 0.40 mM for ZDV and 3TC, respectively. Each chimeric strain was then used to infect 10 6 MT-4 cells in sextuplicate using a MOI 0.002 spinoculation [16] generating 24 independent infections. Viral replication and efficiency of infection were monitored by light microscopy to observe cytopathologic effect (CPE) formation. In all samples, regardless of the viral clone in analysis, the cells were centrifuged after 5-6 days in culture, and 1 mL of the supernatant was used to re-infect fresh MT4 cells. The cells and three aliquots of infection supernatant were frozen at 270uC for further analysis. Drug concentration was maintained or increased two-fold based on cytopathologic effect after each passage. In vitro selection was stopped when drug cytotoxicity signs were evident. All collected aliquots were sequenced and analyzed for drug resistance mutation accumulation.

Sequence Analysis (RT B , RT C , RT B9 and RT C9 )
To check for the presence of drug resistance mutations in each passage, the first 225 codons of RT were amplified using specific primers (RT9-GTACAGTATTAGTAGGACCTACACCTGTC and RT12-ATCAGGATGGAGTTCATAACCCATCCA). The amplicons were purified, sequenced in an automated ABI3100 sequencer (Applied Biosystems) and edited manually using Seqman software (DNASTARH). The genotypic interpretation of antiretroviral drug resistant mutations in RT was carried out through electronic submission to the Stanford database (http:// hivdb.stanford.edu). ARV mutations were scored in all culture passages obtained from RT viral training.

Results
To evaluate if the RT from different HIV-1 subtypes would acquire different resistance mutation patterns in vitro, we subjected recombinant viruses carrying either subtype B or C RT (RT B and RT C respectively) to increasing concentrations of ZDV or 3TC, and analyzed the acquisition of resistance mutations over time. In addition to that, HIV viral load was determined by real time PCR in culture, and correlated to the appearance of NRTI resistance mutations. Figures 1A and 1B depict the results of these analyses for 3TC and ZDV, respectively.
When 3TC was used to select for resistance in viruses carrying RT B or RT C , the VL showed a clear change after 6 passages in increasing drug concentration (42 days, and 0.64 mM of 3TC). In fact, we could detect a 4-log decrease in the VL, which immediately rebounded to original levels and stayed constant all over the remaining passages. In the case of 3TC, both subtypes B and C behaved similarly and we could not observe any major difference in the kinetics. In contrast, the cultures under ZDV selection showed some differences in the VL kinetic between RT C and RT B . The VL rebounded 14 days later in RT C infected cultures when compared with the RT B infected ones.
Several aliquots of culture supernatant were collected at the beginning and during the selection experiment (before VL rebound, in the middle of rebound time, and after VL recovery), viral RNA was isolated and the HIV-1 RT palm-finger region was sequenced. The rebound in 3TC selection was related to the appearance of a unique mutation M184I after 56 days (2.4 mM) regardless the subtype analyzed.
Contrasting to that, subtype B and C isolates followed a different mutational pathway under ZDV selection. The mutation Q151M was detected in subtype B clones after 63 days (0.256 mM), right after rebounding. This mutation was retained after subsequent passages. Additionally, the mutation D67N was incorporated after 77 days (1 mM), when the VL rebounded to original levels before the drug selection. Interestingly, subtype C followed a different route: it accumulated D67N after 77 days (1 mM) before rebounding and K70R after 84 days (2 mM) during the rebound process. The first mutation (D67N) was replaced by T215I at day 91 (2 mM), after the virus reached a VL level comparable to that before selection.
In order to check the reproducibility of these primary data, we cloned NL4-3 subtype B (RT B9 ) and C (RT C9 ) RT sequences into HXB2 background infectious clones. The infectious clones generated were transfected in MT4 cells and the supernatant was harvested after detection of cytopathologic effect (sincytium formation). The recombinant viruses were titrated and used to infect MT4 cells. The established infected cell culture was then selected with increasing concentrations of ZDV and 3TC in six independent experiments for each of the B and C clones.
In accordance to our previous results, 3TC selection yielded similar mutational profiles between RT C9 and RT B9 , but with a slightly difference in kinetics. Mutation M184I appeared between 50-56 days (,5-10 mM) of the selective process and was substituted by M184V or M184M/I/V at later passages with higher concentration of 3TC. The kinetics of M184I acquisition was similar for both RT clones. However, M184V was selected slight faster in RT C9 than RT B9 . The only exception in the mutational profiles was the selection of the non-polymorphic mutation E203K in one replicate of clone RT C9 (Table 2). This mutation fluctuates between 44-96 days and was fixed with 104 days (1310 mM).
Confirming our previous results, ZDV selection generated different mutational patterns between subtypes B and C. Although the selection onset among replicates has temporal differences, the mutational profile is the same in the end. Contrasting to previous results observed in RT B , we could not find the multi-drug resistance mutation Q151M in the early stages of the selection process ( figure 1B and Table 2).

Discussion
The selection of resistance mutations during antiretroviral therapy is associated with a reduction in drug susceptibility and viral fitness. Resistance-related mutations have been conventionally classified as primary or secondary based on their effect on drug susceptibility. While primary mutations reduce drug susceptibility and impact on replicative capacity, secondary mutations do not confer resistance by themselves but can enhance the replicative capacity and resistance levels of viruses carrying primary mutations.
The RT B and RT C recombinant clones showed the same VL levels (replicative capacity) in culture supernatant at the onset of training experiments (data not shown). However, when ZDV was added to the culture, a notable difference was observed in the replicative capacity. In this case, the kinetics of rebound and mutational patterns were distinct among RT B and RT C ( figure 1B). Nevertheless, both clones behaved similarly when the same selection (M184I) was obtained using 3TC ( figure 1A).
Our study has shown that both RT B' and RT C' treated with 3TC selected M184I in 8 weeks, but when selection continues until higher 3TC concentrations, we observed a shift from M184I to M184V (primary mutations). However, in some replicates, this shifting was incomplete and a mixture M184M/I/V was present at the end of the selection process. In addition, RT C ' appears to select M184V faster than RT B' (Table 2), which is in accordance with previous reports [17]. This switch from isoleucine to valine could be due to the fact that M184I has a minor impact in RT processivity despite conferring 3TC resistance. Nevertheless, M184V has a major impact in both 3TC resistance and RT processivity. The selection of this mutation only under high drug concentrations tends to be advantageous for the virus. Moreover M184V brings an advantage for management of the treatment because the M184V-containing enzyme is less processive, decreasing the error of RT and consequently the frequency of mutations throughout the viral genome [18].
One of the RT C' replicates treated with 3TC selected the E203K mutation (Table 2). This change, alongside with other mutations (K43E/N/Q, H208Y, and D218E), have already been associated with NRTI resistance; however, its actual impact in NRTI resistance has not been yet characterized [19,20,21].
Regarding the pattern of mutations selected with ZDV, we observed a difference in the profiles between RT B and RT C . Clone RT B started accumulating the mutation Q151M right after rebounding and this change was retained all over the culture.
Furthermore, an additional mutation, D67N, was incorporated when the VL rebounded to original levels comparable to those before the onset of drug selection ( figure 1B). Q151M is a primary mutation that in vivo can be co-selected together with a group of secondary mutations (A62V, V75I, F77L and F116Y) that confers a cross-resistance with all NRTIs. This complex, located around the catalytic site of the RT, is commonly referred to as ''Q151Mmediated multinucleoside resistance'' (Q151M-MNR) [22,23]. Interestingly, no secondary mutation of Q151M-MNR was significantly evident upon selection with ZDV when the mutation Q151M was selected (RT B ).
The additional selection experiments done in sextuplicate with RT B' selected the T215Y TAM-1 pathway in all cases associated with F214L and D67N (Table 3). Although some researchers characterize D67N as a TAM-2 pathway mutation, this change has also been found in a TAM-1 background, in agreement with our data [24]. In addition, given that D67N does not impact on replicative capacity, this mutation could be selected first and then be replaced by T215Y, which has a major impact on ZDV resistance. Moreover, D67N could help to enhance RT processivity when selected after T215Y [25]. Stürmer M (2004), Ceccherini-Silberstein (2007) and F Puertas MC (2009) observed a negative association of F214L and T215Y, which was related to a decrease in replicative capacity and resistance if compared with viruses carrying only T215Y. However, these two mutations accumulate together in all RT B' six replicates in our in vitro selection with ZDV [26,27,28].
Interestingly, the recombinant virus carrying RT C followed a different route and initially accumulated D67N before rebounding and added K70R during the rebound process. The mutation D67N was replaced by T215I after the virus reached a VL value comparable to the levels before the onset of selection ( figure 1B).
The additional six independent experiments with the RTc' clone treated with ZDV selected the same TAM-2 pathway mutations (Table 2B). Nevertheless, the way that these mutations emerged was different between replicates. Some replicates initiated selection with T215F and others with the D67N and K70R mutations. However, three different final profiles with D67N and K70R mutations were observed in all replicates: T215I or T215F or T215F/I. Essex M. et al. (2009) studied the impact of zidovudine resistance and thymidine analog mutations (TAMs) on subtype C HIV-1 replicative capacity and showed that the 67N and 70R accessory mutations gave an advantage over the WT in subtype C, but not in subtype B. They also showed that the TAM-2 mutant D67N/K70R/T215F had the slowest replication levels between both subtypes [25]. This might explain why D67N and K70R emerged first and then T215I was selected instead of 215F in 66% of replicates. Probably T215I has less impact on replicative capacity than T215F.
It is not known which factors cause the segregation of TAMs into two pathways. Whereas T215Y is one of the TAM-1 mutations firstly selected, that is not the case for T215F in the TAM-2 pathway. This can be probably explained considering the impact of these mutations on viral fitness [29]. Similarly our RT B ' data show that T215Y was the first mutation detected in TAM-1 pathway. However, RT C and RT C' data suggest that D67N and K70R were selected instead of T215F and then directed to TAM-2.  Table 2. Drug resistance mutations found during the in vitro selection of RTB 9 and C 9 with escalating concentration of 3TC.    Our findings show that ZDV selected mutations in different RT subtypes belong to different TAM pathways. Whereas RT B' mutations are related with TAM-1, RT C' is related with TAM-2, in accordance with previous studies [30,31,32,33]. However, other studies did not find this association in subtype C [34]. Still, these later studies were conducted in patients receiving highly active antiretroviral therapy (HAART), in which more than one drug is used. It is well known that drug combinations could influence the pattern of resistance mutations. Follow-up studies with patients infected with different HIV-1 subtypes receiving the same treatment should be done for better understanding of this phenomenon.
Of note, TAM-1 mutations are associated with an increase in phenotypic resistance when compared to TAM-2 ones [36,37]. Furthermore TAM-1 is associated with cross-resistance to didanosine and tenofovir whereas TAM-2 remains susceptible [38]. These data combined with ours suggest that HIV-1 subtype C-infected patients may have larger chances of therapeutic success in ZDV-containing HAART regimens compared to those harboring subtype B.
The selection of drug resistant mutants during antiretroviral therapy could be a consequence of a complex interaction between the effect of a mutation on drug susceptibility and the effect of the mutation on the viral replication potential fitness. In this scenario, the polymorphism carried by different subtypes could influence the mutation pattern selected. This work was not designed to identify polymorphic sites found in subtype C RT that could be responsible for this different behavior in culture under drug pressure. Further experiments need to be done to clarify this point. The data shown here reveal that different subtypes can react differently to antiretroviral drug selection in vitro and suggest different odds of developing ARV resistance among patients infected with different HIV-1 subtypes.