Resistance Mutations outside the Integrase Coding Region Have an Effect on Human Immunodeficiency Virus Replicative Fitness but Do Not Affect Its Susceptibility to Integrase Strand Transfer Inhibitors

Most studies describing phenotypic resistance to integrase strand transfer inhibitors have analyzed viruses carrying only patient-derived HIV-1 integrase genes (INT-recombinant viruses). However, to date, many of the patients on INSTI-based treatment regimes, such as raltegravir (RAL), elvitegravir (EVG), and dolutegravir (DTG) are infected with multidrug-resistant HIV-1 strains. Here we analyzed the effect of drug resistance mutations in Gag (p2/NCp7/p1/p6), protease (PR), reverse transcriptase (RT), and integrase (IN) coding regions on susceptibility to INSTIs and viral replicative fitness using a novel HIV-1 phenotyping assay. Initial characterization based on site-directed mutant INSTI-resistant viruses confirmed the effect of a series of INSTI mutations on reduced susceptibility to EVG and RAL and viral replicative fitness (0.6% to 99% relative to the HIV-1NL4-3 control). Two sets of recombinant viruses containing a 3,428-bp gag-p2/NCp7/p1/p6/pol-PR/RT/IN (p2-INT) or a 1,088 bp integrase (INT) patient-derived fragment were constructed from plasma samples obtained from 27 virologic failure patients participating in a 48-week dose-ranging study of elvitegravir, GS-US-183-0105. A strong correlation was observed when susceptibility to EVG and RAL was assayed using p2-INT- vs. INT-recombinant viruses (Pearson coefficient correlation 0.869 and 0.918, P<0.0001 for EVG and RAL, respectively), demonstrating that mutations in the protease and RT have limited effect on susceptibility to these INSTIs. On the other hand, the replicative fitness of viruses harboring drug resistance mutations in PR, RT, and IN was generally impaired compared to viruses carrying only INSTI-resistance mutations. Thus, in the absence of drug pressure, drug resistance mutations in the PR and RT contribute to decrease the replicative fitness of the virus already impaired by mutations in the integrase. The use of recombinant viruses containing most or all HIV-1 regions targeted by antiretroviral drugs might be essential to understand the collective effect of epistatic interactions in multidrug-resistant viruses.


Introduction
Productive infection with human immunodeficiency virus type 1 (HIV-1) requires three key steps in the replication of the virus, i.e., reverse transcription of viral genomic RNA into viral cDNA by the viral reverse transcriptase (RT), integration of viral cDNA into host cell genome using the viral integrase (IN), and cleavage of newly synthesized viral polypeptide by the viral protease (PR) into individual viral proteins during new virion assembly [1]. Although all three steps were initially considered as targets for antiretroviral drugs, HIV-1 integrase was the last viral enzyme to emerge as a clinically validated alternative to block HIV-1 replication [2]. Since the late 1990s, several different chemical scaffolds have been studied for their ability to inhibit HIV-1 integration [3,4], leading to the approval of the first HIV-1 integrase strand transfer   [6] has recently been approved in combination with a pharmacokinetic enhancer (cobicistat) and two nucleos(t)ide analog RT inhibitors (emtricitabine and tenofovir) for the treatment of antiretroviral-naïve HIV-infected individuals (QUAD, Stribild, Gilead Sciences) [7]. Dolutegravir (DTG, S/ GSK1349572, ViiV Healthcare), a second-generation INSTI [8], recently completed a phase III clinical trial [9] and is waiting approval to treat HIV infection.
In this study we have used an HIV-1 phenotypic assay (VIRALARTS TM HIV), based on the construction of p2-INT (gag-p2/NCp7/p1/p6/pol-PR/RT/IN) recombinant viruses [49], to (i) characterize the susceptibility of INSTI-resistance viruses to RAL, EVG and twenty additional antiretroviral drugs and (ii) evaluate the role of drug-resistance mutations in PR, RT, and IN coding regions on HIV-1 replicative fitness. We showed that although mutations in PR and RT have a limited effect on susceptibility to INSTIs they affect the ability of the virus to replicate in the presence and absence of RAL and EVG.

Construction of Site-directed Mutant (SDM) Viruses Carrying INSTI-associated Mutations
Fourteen SDM viruses containing single, dual-or triplemutations associated with resistance to INSTI were constructed as previously described [15]. Briefly, the Apa1-Sal1 fragment of the HIV-1 pol gene in vector pUNV5-HisB was mutated using the QuikChangeH Site-Directed Mutagenesis Kit (Stratagene; La Jolla, CA) and transformed into PIR1 E. coli cells (Invitrogen; Carlsbad, CA). Plasmid DNA was purified (Qiagen; Valencia, CA), restriction digested with Apa1-Sal1, and the mutated pol fragment ligated to an HXB2 proviral vector and transformed into XL10-Gold Escherichia coli cells. Plasmid DNA was then used to construct 39Gag(p2/NCp7/p1/p6)/PR/RT/INT-recombinant viruses in a HIV-1 NL4-3 backbone as described below.

Clinical Specimens
Plasma samples were obtained from twenty-seven patients experiencing virologic failure while participating in a 48-week dose-ranging study of elvitegravir (EVG), Study GS-US-183-0105 [16] (Table 1). Written informed consent was obtained from the patients before participation in the study as previously described [15,16].

RT-PCR Amplification and Nucleotide Sequence Analysis of gag-p2/NCp7/p1/p6/pol-PR/RT/IN-coding Sequences
Plasma viral RNA was purified from pelleted virus particles by centrifuging one milliliter of plasma at 14,000g660 min at 4uC, removing 860 ml of cell-free supernatant and resuspending the pellet in the remaining 140 ml, to finally extract viral RNA using QIAamp Viral RNA Mini kit (Qiagen; Valencia, CA). Viral RNA was reverse-transcribed using AccuScript High Fidelity Reverse Transcriptase (Stratagene Agilent; Santa Clara, CA) and the corresponding antisense external primer in 20 ml reaction mixture containing 1 mM dNTPs, 10 mM DTT and 10 units of RNAse inhibitor. A 3,428 nt HIV-1 genomic region encoding the Gag proteins p2, p7, p1 and p6, and the protease, reverse transcriptase, and integrase enzymes was then PCR amplified using a series of external and nested primers with defined cycling conditions [49]. External PCR reactions were carried out in a 50-ml mixture containing 0.

Construction of gag-p2/NCp7/p1/p6/pol-PR/RT/IN -and INT-recombinant Viruses
Two sets of infectious recombinant viruses were constructed from each clinical specimen in a HIV-1 NL4-3 backbone using a novel yeast-based cloning technology as described [49]. Briefly, PCR products spanning the 39 end of gag (p2/p7/p1/p6) and the entire pol gene (PR/RT/IN; p2-INT; 3,428 nt) or the integrasecoding region only (INT; 1,088-nt) were introduced via yeast homologous recombination into pRECnfl-TRP?p2-INT/URA3 or pRECnfl-TRP?INT/URA3 vectors, respectively, containing a near-full length HIV-1 genome with the yeast uracil biosynthesis (URA3) gene replacing the respective p2-INT or INT HIV-1 coding sequences (Fig. 1). Following yeast transformation, vector DNA was purified from the entire number of yeast colonies (typically 200 to 500 individual colonies) and used to transform Electrocomp TOP10 bacteria (Invitrogen). Plasmid DNA from the entire bacteria preparation -to guarantee the continuity of the viral population that may have existed in vivo -was purified from 10 ml of bacteria culture (QIAprep Spin Miniprep Kit, Qiagen) and used to introduce the patient-derived HIV-1 sequences into a pNL4-3-hRluc vector expressing the human Renilla luciferase gene (hRluc) [50] as described [49]. Four micrograms of the resulting plasmid were transfected into HEK293T cells using GenDrill TM (BamaGenH Bioscience; Gaithersburg, MD). Cell culture supernatant was harvested 48 hours post-transfection, clarified by centrifugation at 7006g, filtered through a 0.45 mm steriflip filter (Millipore; Billerica, MA), aliquoted, and stored at 280uC until further use. Tissue culture dose for 50% infectivity (TCID 50 ) was determined in triplicate for each serially diluted virus using the Reed and Muench method [51] and viral titers expressed as infectious units per milliliter (IU/ml). The p2-INT  and INT regions from the p2-INT-and INT-recombinant viruses, respectively, were sequenced as described above to verify the identity and genotype of each constructed virus.

Drug Susceptibility Based in an MT-2 Assay
Susceptibility of the fourteen SDM viruses to EVG and RAL was quantified using cell-viability of single viral infections [15]. Briefly, the multiplicity of infection (MOI) of the viruses was normalized to obtain a signal-to-noise ratio of 4 to 7 for uninfected versus infected cells in the absence of drug (chemiluminescent detection, Cell Titer-Glo, Promega). MT-2 cells were infected for three hours then incubated in triplicate drug dilution series for five days at 37uC. Cell viability data was converted to percentage of cell death and drug concentrations required to inhibit virus replication by 50% (EC 50 ) were calculated by curve fitting (GraphPad Prism v.5.01, GraphPad Software, La Jolla, CA). Fold change (FC) resistance values were calculated by dividing the mean EC 50 of the query virus (HIV-1 SDM ) by the mean EC 50 of the control (HIV-1 HXB2 ).

HIV-1 Replicative Fitness Determination Using Viral Growth Kinetics Analysis
The ability of fourteen SDM viruses, twenty-seven p2-INT-, and twenty-seven INT-recombinant viruses, plus the HIV-1 NL4-3 wild-type control, to replicate in the absence or presence of drug pressure (EVG or RAL) was determined by measuring viral growth kinetics as described [49,52]. Briefly, 3610 6 MT-4 cells were infected in triplicate at a MOI of 0.01 IU/cell in one ml of culture medium and incubated for 2 hrs at 37uC in 5% CO 2 .
HIV-infected cells were then washed two times with 16PBS, then split to be cultured in triplicate wells of a 24-well plate (1610 6 cells/well). Culture supernatant was assayed using a reverse transcriptase assay on days 0, 3, 4, 5, 6, 7, and 10 post-infection as described [53]. Viral replication was quantified using the slope of the growth curves and performing linear regression analysis derived from the equation log(y) = mt+log(h), where y is virus quantity (cpm), t is time in days, and h is the y-intercept (day 0). All slope values for each virus were used to calculate the mean, standard deviation, and 10 th & 90 th percentiles. Differences in the mean values were evaluated using a One Way Analysis of Variance test and the significance difference from the reference HIV-1 NL4-3 virus calculated using the Bonferroni's Multiple Comparison Test (GraphPad Prism v.5.01, GraphPad Software).

HIV-1 Replicative Fitness Determination Using Growth Competition Experiments
Dual infection/competition experiments were carried out as previously described [52,53,54]. Briefly, query (HIV-1 SDM-or p2-INT-or INT ) and control (HIV-1 NL4-3 ) viruses tagged with hRluc and fluc2, respectively, were competed in a 1:1 initial proportion using a MOI of 0.01 IU/cell. One ml of the virus mixture was incubated with 6610 5 MT-4 cells for 2 h at 37uC, 5% CO 2 in the presence and absence of 0.01 nM EVG or 1 nM RAL, corresponding to the EC 50 values for these INSTIs using HIV-1 NL4-3 as determined by VIRALARTS TM HIV. Cells were subsequently washed three times with 16 PBS and cultured in triplicate in a 96-well plate (2610 5 cells/well) with the correspondent amount of drug. At day 5 post-infection, the final proportion of the two viruses in each competition was quantified using the Dual-GloH Luciferase Assay System (Promega, Madison, WI) in a multiwell plate reader (Victor Vumultilabel reader, PerkinElmer) after normalizing to viral production in the HIV-1 monoinfections as described [53,54,55]. Replicative fitness for each virus was calculated and expressed as a percentage of the replicative fitness of the reference virus HIV-1 NL4-3 .

Statistical Analyses
Descriptive results are expressed as median values and interquartile ranges. Pearson correlation coefficient was used to determine the strength of association between categorical variables. All differences with a P value of ,0.05 were considered statistically significant. As described above, differences in the mean of the slope values for the viral growth kinetics curves were determined using a One Way Analysis of Variance test and the significance difference from the reference HIV-1 NL4-3 virus calculated using the Bonferroni's Multiple Comparison Test. All statistical analyses were performed using GraphPad Prism v.5.01 (GraphPad Software, La Jolla, CA) unless otherwise specified.

Susceptibility to EVG and RAL of p2-INT-recombinant Viruses Carrying Mutations Associated with Resistance to INSTIs
We recently developed an HIV-1 phenotyping assay (VIR-ALARTS TM HIV) based on the introduction of patient-derived p2-INT amplicons into an HIV-1 NL4-3 backbone using a yeast-based cloning system [49,56]. This assay was validated using 21 antiretroviral drugs, including RAL [49]; however, EVG was not part of the original characterization. Thus, here we first tested the ability of VIRALARTS TM HIV to quantify susceptibility to EVG and RAL by comparing our results with data obtained with an MT-2 assay. For that, p2-INT fragments were PCR amplified from 14 plasmids containing single, dual or triple INSTI resistance mutations introduced by site-directed mutagenesis, and then used to construct p2-INT-recombinant viruses. Drug susceptibility was assessed by measuring luciferase expression in MT-4 cells (original VIRALARTS TM HIV) and CXCR4 cells, and cell viability in MT-2 cells. Overall, susceptibility to EVG and RAL was similar in all three cell systems ( Fig. 2A), with strong statistically significant correlations (r values ranging from 0.866 to 0.961, P,0.0001, Pearson coefficient correlation), particularly between data obtained with MT-4 and MT-2 cells (Fig. 2B). As expected, a few mutations conferred resistant to EVG but retained susceptibility to RAL, e.g., T66I, G140S, and S147G, while other mutations reduced susceptibility to both INSTIs, e.g., Q148K/R/H, N155H ( Fig. 2A). All viruses were susceptible to d4T and NVP, antiretroviral drugs used as controls (data not shown). Interestingly, the wild type reference control virus (HIV-1 NL4-3 ) was particularly susceptible to EVG using MT-4 cells in this system (EC 50 values ranging from 0.009 to 0.02 nM, data not shown). Thus, any reduction in susceptibility due to the presence of INSTI resistance mutations seems to be magnified in this system producing higher than previously reported EC 50 fold-change values ( Fig. 2A).

Replicative Fitness of p2-INT-recombinant Viruses Carrying Mutations Associated with Resistance to INSTIs in the Absence and Presence of EVG or RAL
One of the advantages of VIRALARTS TM HIV is that, unlike other HIV-1 phenotyping assays based in single-cycle replication, it produces replication-competent p2-INT-recombinant viruses and supports the analysis of replication kinetics during multiple rounds of cell-culture infections. Here we used two different but complementary approaches to quantify the ability of the 14 p2-INT-recombinant viruses, carrying single, dual or triple INSTI resistance mutations to replicate in the presence and absence of EVG and RAL. A first glimpse of the replicative fitness of these viruses was obtained using classical viral growth kinetics in MT-4 cells in the absence of drug as compared to the reference HIV-1 NL4-3 strain (Fig. 3A). Statistical analysis of the slope of the growth curves showed that the replicate fitness of viruses containing the single mutations E138K, S147G, and G140S and the triple mutant E138K+S47G+Q148R was similar to the wild type HIV-1 NL4-3 (Fig. 3B). On the other hand, 10 INSTI-resistant viruses showed an impairment in replication compared to the wild type control, with the Q148R virus being the less fit (Fig. 3B).
Since in vitro growth competition experiments are considered the gold standard method to measure viral fitness [21,22] we used dual infections to quantify the ability of the 14 INSTI-resistant viruses to replicate in the absence and presence of drug pressure (EVG and RAL). Viruses carrying the Q148R or Q148K single mutations showed a marked decrease in replicative fitness, i.e., 0.6% and 4.2% relative to that of the HIV-1 NL4-3 control virus, respectively (Fig. 3C). Interestingly, the virus with a different positive charged amino acid in the same position (Q148H) had a 9-to 60-fold higher replicative fitness than either the Q148R or Q148K viruses (36% of the HIV-1 NL4-3 , Fig. 3C). Moreover, mutations E138K (84%) and S147G (95%), which seem to have a minimal effect on INSTI resistance ( Fig. 2A) and viral fitness contributed to increase the effect of the Q148R mutation, as observed in the triple mutant E138K+S147G+Q148R (36% of the HIV-1 NL4-3 , Fig. 3C). The rest of the p2-INT viruses carrying single or double INSTI resistance mutations showed a range of replicative fitness values from 68% (N155H) to 99% (E92Q) relative to the HIV-1 NL4-3 control (Fig. 3C). No statistically significant correlation was observed between the viral replication slope values (viral growth kinetics) and the % replicative fitness calculated using growth competition (data not shown), driven mainly by discrepancies with the results from viruses E92Q, E92Q+N155H, and E138K+S147G+Q148R (Figs. 3B and 3C).
Finally, each of the 14 INSTI-resistant viruses was competed against the HIV-1 NL4-3 control virus in the presence of EVG (0.01 nM) or RAL (1 nM). The ratio between the replicative fitness of the virus in the presence and absence of drug pressure serves as an indirect measurement of the level of resistance, and the effect on virus replication, provided by certain mutation(s). For example, a ratio of 1 indicates that the mutant virus replicates similarly to the wild type control virus both in the presence and absence of drug pressure (i.e., mutation has a small effect in replicative fitness); however, a positive ratio underlines the ability of the virus to replicate in the presence of the drug and/or the detrimental effect of the mutation(s) on fitness in the absence of drug pressure. Accordingly, INSTI-resistant viruses with replicative fitness values ranging from 68% to 99% of the HIV-1 NL4-3 control in the absence of drug had ratios of drug/no drug replicative fitness in the range of 0.96 to 1.56 (Fig. 3D). The ratio increased to 26 and 36 for viruses carrying mutations Q148H/K and E138K+S147G+Q148R, both with a replicative fitness value of 36%. The two viruses with the lowest replicative fitness in the absence of drug, i.e., Q148K (4.2%) and Q148R (0.6%) showed a substantial recovery on replication capacity in the presence of the INSTIs with ratios of drug/no drug replicative fitness ranging from 106to 1766, respectively (Fig. 3D). Interestingly, the ratio of replicative fitness of the E138K virus, susceptible to both INSTIs, was reduced in the presence of EVG (0.36) but not with RAL (1.26, Fig. 3D).

Antiretroviral Drug Susceptibility of Multidrug-resistant Viruses Obtained from Patients Participating in a Phase II Clinical Trial of Elvitegravir
Patient-derived PCR products from 27 HIV-infected individuals participating in a 48-week dose-ranging study of elvitegravir, GS-US-183-0105 [16] were used to construct p2-INT-recombinant viruses and their susceptibility to 22 antiretroviral drugs, including EVG and RAL, was assessed using VIRALARTS TM HIV [49]. Viruses were divided in groups based on the presence of amino acid substitutions in three positions associated with major INSTI mutations (i.e., E92, N155, and Q148) and secondary mutations ( Table 1). Table S1 includes a full list of amino acid substitutions in the PR, RT, and IN coding regions obtained by population sequencing of plasma samples. As expected from highly treatmentexperienced HIV-infected individuals, most of them carried viruses with multiple primary mutations conferring resistance to many protease, RT, and IN inhibitors. These HIV-1 genotypes were corroborated by the drug susceptibility assay and in the case of INSTIs, mutations Q148R/H, N155H+Q148R, and E92Q+N155H conferred the highest level of resistance to EVG and RAL ( Table 1).

Comparison of the Susceptibility to EVG and RAL between Multidrug-resistant p2-INT-and INTrecombinant Viruses
As described above, to date most of the INSTI susceptibility data has been obtained using INT-recombinant viruses constructed only with patient-derived HIV-1 integrase amplicons [12,25,26,30,31,33]. Since our HIV-1 phenotyping assay uses recombinant viruses carrying longer patient-derived amplicons (gag-p2/NCp7/p1/p6/pol-PR/RT/IN) we compared the susceptibility to EVG and RAL using both p2-INT-and INTrecombinant viruses, with the first set of viruses carrying the 39end of Gag, PR, and RT coding regions, in addition to the integrase, from the 27 patients (Fig. 1). A strong statistically correlation was observed between the EC 50 values calculated with both sets of viruses, i.e., r values of 0.869 and 0.918 (P,0.0001, Pearson coefficient correlation) for EVG and RAL, respectively (Fig. 4).

Replicative Fitness of Multidrug-resistant p2-INT-and INT-recombinant Viruses in the Absence and Presence of EVG or RAL
Multiple drug-resistance mutations in the PR and RT coding regions have been associated with impairing the ability of the virus to replicate in the absence of drug pressure [21,22]. Thus, although susceptibility to INSTI may not be affected by mutations in the PR and RT (Fig. 4), detrimental mutations in these HIV-1 genomic regions may work together with mutations in the integrase to affect the overall replicative fitness of the virus. We first compared the viral growth kinetics of the p2-INT-and INTrecombinant viruses in the absence of drug (Fig. 5A). All 27 p2-INT-recombinant viruses, many of them harboring numerous mutations in the PR, RT, and IN coding regions (Table 1 and  Table S1), showed different levels of impairment on their replication kinetics compared to the HIV-1 NL4-3 control (Fig. 5B). In contrast, not all INT-recombinant viruses, constructed only from patient-derived HIV-1 IN amplicons, showed a decrease in replicative fitness. The replication kinetics of viruses from patients 08-186 (E92Q), 08-196 (E92Q), 08-239 (G140S+Q148H), and 08-193 (T66A) was not different to that of the wild type control virus (Fig. 5B). Moreover, while the rest of the INT-recombinant viruses showed a decrease in replicative fitness, most of their viral replication slopes were higher than those calculated for the p2-INT-recombinant (Fig. 5C).
As with the INSTI-resistant viruses constructed from sitedirected mutants, we used growth competition experiments to determine the ability of the p2-INT and INT-recombinant viruses to replicate in the absence and presence of drug pressure, in this case EVG. A variety of replicative fitness values were observed in both sets of viruses, ranging from 2% to 149% (p2-INTrecombinant viruses) and 4% to 116% (INT-recombinant viruses) relative to the HIV-1 NL4-3 control, in the absence of drug pressure (Fig. 6A). A marked reduction in replicative fitness (below 50%) was observed in 15/27 and 9/27 p2-INT-and INT-recombinant viruses, respectively; mostly related to viruses carrying mutations at positions Q148 and N155 in the integrase coding region (Fig. 6A). Differences in replicative fitness between p2-INT-and INT-recombinant viruses were patient-dependent and guided mainly by the number and type of drug resistance mutations in the PR and RT coding regions (Table 1 and Table S1). Finally, only viruses with highly impaired replicative fitness in the absence of drug pressure (e.g., those carrying N155H and/or Q148R mutations) showed a significant recovery of fitness in the presence

Discussion
Multiple studies have analyzed the role of resistance mutations in the HIV-1 integrase coding region on susceptibility to INSTI and their effect on viral replication capacity [3,4,10,11]. However, all these studies have used site-directed mutant viruses [13,15,19,26,29,30,31,32,37,38], IN-recombinant viruses based only on patient-derived HIV-1 integrase amplicons [12,25,26,30,31,33], or quantified the dynamics of HIV-1 integrase mutations in vivo [27,34,40]. Here we used a novel  Three main pathways of resistance to RAL have been associated with primary mutations Q148R/H/K, N155H, or Y143C/R/H, either alone or in combination with other IN mutations [3,11]. EVG can select several INSTI resistance mutations in vitro [12,13,14] but E92Q, Q148R/H/K, and N155H mutations have been the most common EVG resistance mutations identified in vivo, with other mutations also emerging during clinical studies, e.g., T66I/A/K, S147G, etc. [4]. Therefore, we first tested our system by constructing p2-INTrecombinant viruses carrying these and other single, dual, or triple INSTI resistance mutations introduced by site-directed mutagenesis [15]. Susceptibility to EVG and RAL quantified with our HIV-1 phenotyping assay (VIRALARTS TM HIV) correlated with results from a standard drug resistance test based on MT-2 cells, with mutations Q148H/K/R and N155H conferring the highest level of resistance to both RAL and EVG. This is in agreement with previous reports of cross-resistance between these two INSTIs, mainly related to mutations at these two positions [3,4,10,11]. In addition, since VIRALARTS TM HIV is a multiple replication cycle assay, we were able to study the effect of these mutations on viral replication fitness by measuring viral growth kinetics in single infections or growth competition experiments. Nearly all amino acid substitutions associated with INSTI resistance had an effect on HIV-1 replicative fitness. As previously described [29,31,33,36,37], mutations associated with higher loss of susceptibility to INSTIs, such as Q148H/R/K and N155H had a major effect on the ability of the virus to replicate in the absence of drug pressure. Moreover, only the highly resistant viruses with low replicative fitness in the absence of drug (i.e., Q148R, Q148K, Q148H, and E138K+S147G+Q148R) showed a significant increase in their capacity to replicate in the presence of INSTIs. It is possible that higher INSTI concentrations (i.e., .EC 50 values) could lead to a more marked recovery in fitness for viruses with moderate resistance to INSTI (e.g., T66I, E92Q, or G140S). In summary, these experiments not only corroborated previous studies but also validated our assay to be used with patient-derived HIV-1 fragments to evaluate the potential epistatic relationship between drug resistance mutations in the HIV-1 pol gene.
Interestingly, although most of the replicative fitness values determined by viral growth kinetics or growth competition experiments shared the same trend, we were not able to calculate a statistically significant correlation between the fitness estimated by these two complementary methodologies. In the case of the p2-INT-recombinant viruses constructed from the site-directed mutants, the discrepancies could be attributed to three particular viruses: E92Q, E92Q+N155H, and E138K+S147G+Q148R. It is possible that (i) intrinsic differences between the two tests, (ii) the method used to determine TCID 50 values, and/or (iii) the fact viral growth kinetics were performed in triplicate while growth competitions experiments were performed only once, could have contributed to the differences in fitness observed for some viruses. Nevertheless, and due to the fact that in vitro growth competition experiments are considered the gold standard method to measure viral fitness [21,22], here we believe that the data generated by the competitions is more reliable than the results obtained with viral growth kinetics.
The role of drug resistance mutations in the PR and RT coding regions on susceptibility to INSTIs and their contribution, together with INSTI resistance mutations, to overall HIV-1 replicative fitness is not fully understood. Drug susceptibility studies of protease and RT inhibitors are usually performed separately from those evaluating resistance to INSTIs, typically by constructing independent PR/RT-and IN-recombinant viruses, respectively. This makes difficult, if not impossible, to analyze the natural interaction among drug resistance mutations in the three HIV-1 enzymes. For example, several studies have shown that mutations outside the protease and the polymerase domain of the RT coding region have an effect on susceptibility to PIs and RTIs, respectively. Mutations downstream of the Gag protease cleavage site p24(CA)/p2 have been associated with reduced susceptibility to PIs [44,57], while amino acid substitutions in the connection [46,58] and RNase H [59] domains of the RT have been shown to have an effect on NRTI and NNRTI resistance. In addition, HIV-1 integrase seems to affect nuclear import and viral maturation while its interaction with the RT may affect virus replication [60]. More important for this study, a few specific polymorphisms in the integrase-coding region (e.g., M154L, V165I, G163R, and T206S) seem to be associated with RT resistance mutations in antiretroviral-experienced individuals [61]. Thus, it is only logic to predict that drug resistance mutations in PR, RT, and IN play a role in overall HIV-1 replicative fitness, perhaps affecting the ability of the virus to respond to certain combination(s) of antiretroviral drugs.
A recent genotypic study [62] showed that development of INSTI-resistance mutations is not restricted by drug-resistance mutations in the PR and/or RT coding regions; however, Buzon et al [48] described that although mutations outside the integrase coding region may not affect susceptibility to INSTIs, mutations within the PR and RT rescued the replication capacity of viruses with INSTI resistance mutations, suggesting the existence of epistatic effects on HIV-1 replicative fitness. Epistasis in RNA viruses, particularly positive gene interactions in HIV-1, has been associated with mutations affecting fitness [63,64]. Here we compared the susceptibility of two sets of recombinant viruses, constructed from patient-derived gag-p2/NCp7/p1/p6/pol-PR/ RT/IN or IN fragments, to RAL and EVG and demonstrated that mutations in the protease and RT have limited effect on susceptibility to these INSTIs. The p2-INT-recombinant viruses engineered from 27 highly treatment-experienced patients carried a multitude of drug resistance mutations, in different combinations, which generated diverse levels of drug susceptibility and yet none of them significantly affected the sensitivity of the virus to RAL and EVG. On the other hand, the replicative fitness of the p2-INT-recombinant viruses was different, and in most cases reduced, when compared to that of the INT-recombinant viruses. Epistasis is positive when it enhances the fitness predicted from individual effects of deleterious (e.g., drug resistance) mutations and is negative when it decreases fitness [65]. Unlike the positive epistasis previously reported in HIV-1 [48,63,66], i.e., mutations in other genomic regions restore viral fitness impaired by mutations in a given gene, we show that drug resistance mutations in the PR and RT contribute to decrease the replicative fitness of the virus already impaired by mutations in the integrase. This negative epistasis indicates that HIV-1 is highly sensitive to the combinatory effects of deleterious mutations in different genes, as described for HIV-1 protease [67] and gp120 [66]. However, it is important to note that in this study the negative epistasis associated with a reduction in viral fitness refers to an environment characterized by the absence of drug pressure. The otherwise deleterious (drug resistance) mutations confer a distinctive advantage in the presence of antiretroviral drugs, increasing the replicative fitness of the virus in this new environment and consequently changing the ''sign'' of the epistasis to positive [63]. Thus, epistatic interactions and their role in HIV-1 replicative fitness is complicated and depends on environmental factors such as the absence or presence of drugs [68].
In summary, our novel HIV-1 phenotyping assay based on patient-derived gag-p2/NCp7/p1/p6/pol-PR/RT/IN fragments was not only able to simultaneously quantify susceptibility to protease, RT, and integrase inhibitors but to also measure the effect of multiple drug resistance mutations on viral replicative fitness. Mutations in the PR and RT-coding regions do not seem to have a significant effect on susceptibility to INSTIs; however, all HIV-1 genes are involved in modulating viral replicative fitness, particularly those regions carrying drug resistance mutations. Therefore, the use of recombinant viruses containing most or all HIV-1 regions targeted by antiretroviral drugs seems to resemble better the characteristics of the actual virus circulating in vivo and might be key in future studies aimed at understanding the potential collective effect of epistatic interactions in multidrug resistant viruses. Table S1 HIV-1 genotype for the 27 HIV-infected individuals participating in the GS-US-183-0105 study of elvitegravir. (DOCX)