**3. Discussion**

InSTIs are a preferred third agen<sup>t</sup> in NRTI-based regimen. Consequently, some patients on an InSTI-containing regimen will or have previously failed therapies containing PIs and RTIs, especially in low- and middle-income countries. Thus, it is important to understand the interaction and development of drug resistance mutations in the context of patientderived full-length *pol* gene. In this study, we used a full-length HIV-1 *pol* gene SGS assay to demonstrate the genetic linkage of drug resistance mutations throughout the *pol* gene. We showed that drug resistance mutations in IN are linked to those in PR and RT, and that different combinations of InSTI resistance mutations can develop concurrently linked to the same PR and RT drug resistance mutations. This is consistent with another study showing the linkage of EVG resistance mutations in IN to drug resistance mutations in PR and RT [15].

Our data also revealed the simultaneous presence of mutations at all three major InSTI drug resistance positions (Q148, Y143 and N155) during treatment failure, albeit on different genomes, which is also consistent with previous findings [19–23]. In this study, the major InSTI resistance mutations were always linked to accessory mutations. This may be due to a longer period between initiation of RAL therapy and first sampling, which was at least 4 months. It is well established that major RAL resistance mutations can appear rapidly (sometimes within a month) after initiation of RAL treatment, with accessory mutations developing subsequently to compensate for fitness loss and/or to increase drug resistance [19–29]. The genetic linkage of major and accessory mutations suggests that the accessory mutations compensate for fitness loss and/or increase drug resistance of a viral variant through cis-acting mechanisms. This has been confirmed for the Q148H + G140S double mutant, in which it was demonstrated that the catalytic properties of IN were greatly impaired by the single mutants. However, the double mutant could fully restore the catalytic properties of IN only when the two mutations were present on the same IN polypeptide [28].

Our findings also shed light on the complexities of the intrapatient evolution of the Y143 resistance pathway, and the effects of accessory mutations linked to this pathway. Previous studies sugges<sup>t</sup> that the accessory mutations linked to the Y143 mutations play a positive role in IN activity and/or RAL resistance [21,24,30]. In this study, we observed that different accessory mutations combined with a particular substitution at position Y143 differentially influenced the levels of susceptibility to RAL. For example, we observed the development of the Y143C resistance mutation, linked to three different accessory mutations (E92Q, T97A and G163R), differentially decreased RAL susceptibility from 19- to 183-fold, as well as viral fitness by 8 to 35% compared to wild-type virus. All three of the accessory mutations are found in the vicinity of the catalytic active site of *IN* [31]; therefore, it is envisioned that they could be affecting susceptibility and replicative fitness by directly influencing the structure of the active site.

The differential effect on RAL susceptibility and viral fitness also extended to different substitutions at primary resistance position 143 (R, C or G), linked to the same accessory mutation (G163R). This finding is contrary to another study which showed that the RAL susceptibilities of both the Y143R and Y143C mutants were similar [30]. However, in this study, the Y143 mutants were not linked to the G163R accessory mutation, which may partially explain the different observations. These data sugges<sup>t</sup> that accessory mutations linked to Y143 mutations affect both the replicative fitness and InSTI susceptibility of the mutant viruses, and that a balance between the two could play a role in determining the development and evolution of resistance. Consistent with this observation, the Y143G + G163R mutant, that was present as a minority variant after RAL treatment was stopped, emerged as the dominant viral variant upon the re-initiation of RAL therapy. This variant exhibited a significantly higher RAL EC50 than the Y143C + G163R mutant, and had a significantly higher replicative fitness compared to the Y143R + G163R and Y143C + G163R mutants. The replicative fitness of Y143G + G163R was in fact the highest of all RAL resistant variants identified in this patient, and was comparable to that of

wild-type virus. It is therefore likely that the subsequent outgrowth of this viral variant after the re-initiation of RAL therapy may be due to its higher replicative fitness compared to the other RAL resistant viruses present in the patient. The late and rare development of the Y143G substitution could be due to a high genetic barrier. The wild-type codon for position 143 in IN in this patient was TAC. One nucleotide change was required for the Y to C substitution (TAC to TGC), whilst two nucleotide changes were needed to generate the G substitution (TAC to GGC) or R substitution (TAC to CGC). The second change required for the Y to R substitution (T to C) is a transition which occurs at a higher frequency compared to the T to G transversion required for the Y to G substitution. Thus, although more advantageous to the virus in terms of fitness, the Y143G substitution is likely to occur less frequently compared to Y143R/C substitutions. This is supported by analysis of data available from the Stanford HIV Drug Resistance Database, which showed no instances of Y143G mutation in InSTI-experienced subtype B infected patients (Figure 1B). This also illustrates that continuous selective drug pressure during a failing regimen will force the virus to continue evolving towards a fitter resistant virus, that is then more likely to persist in the absence of drug pressure [32,33].

A limitation of the study is that the analysis, although in depth, is from one patient. Future studies will focus on investigating this phenomenon in a large group of patients, and include investigation of resistance to second generation InSTIs, such as DTG and use of full-length genome clones. Nonetheless, the data show that the coevolved *PR* and *RT* genes affect the susceptibility and replicative fitness of an *IN* gene harbouring InSTI resistance mutations by up to three-fold. This concurs with another study, that showed an effect on viral fitness and susceptibility to EFV and RAL for certain combinations of NNRTI and InSTI resistance mutations [34]. On the other hand, other studies have shown that mutations in PR and RT have little effect on the susceptibility to InSTIs, but can reduce viral replicative fitness of a resistant *IN* gene [35]. Different experimental approaches, such as the use of site-directed mutants compared to patient-derived fragments, or differences in the combination of resistance mutations and/or accessory mutations, could explain the contradictory outcomes. Therefore, further comprehensive studies coupling biological and biochemical investigations are required to elucidate the interactions between mutations in full-length HIV-1 *pol* gene and their effects on susceptibility and viral fitness, including that to the new InSTIs DTG, BIC and CAB. However, taken together, these data sugges<sup>t</sup> that analysis of only part of the HIV-1 genome is probably not sufficient to gauge the true dynamics in the evolution and extent of drug resistance in the era of cART, as shown by recent studies linking InSTI resistance to env and cPPT regions [33,36,37]. The use of assays encompassing full-length *pol* gene or more of the viral genome may provide useful insights into drug resistance mechanisms, and help devise better treatment strategies as well as improve the prediction of the emergence of drug resistance and subsequent treatment failure.

#### **4. Materials and Methods**

## *4.1. Clinical Samples*

The plasma samples used in this study were obtained from a patient attending the Mortimer Market Clinic, UCLH, who was infected with subtype B virus and initiated on RAL salvage therapy (600 mg daily) in September 2007. They continued RAL in combination with darunavir/ritonavir (DRV/r) and etravirine (ETR) until February 2008, when the patient experienced virological failure. The patient was then switched onto therapy containing tenofovir (TDF) and lamivudine (3TC), but experienced virological failure again 2 months later (April 2008). RAL treatment was then re-started in combination with TDF/emtricitabine (FTC), DRV/r and ETR in September 2008. Six samples were obtained, and these were: pre-RAL therapy (preRAL); 2, 4 and 5 months on RAL (2RAL, 4RAL and 5RAL, respectively); 4 months after RAL was stopped (4post); and 0.5 months after RAL was re-started (reRAL) [23]. Informed consent was obtained in the context of routine resistance testing as a part of clinical protocol.
