Persistence of frequently transmitted drug-resistant HIV-1 variants can be explained by high viral replication capacity

Background In approximately 10% of newly diagnosed individuals in Europe, HIV-1 variants harboring transmitted drug resistance mutations (TDRM) are detected. For some TDRM it has been shown that they revert to wild type while other mutations persist in the absence of therapy. To understand the mechanisms explaining persistence we investigated the in vivo evolution of frequently transmitted HIV-1 variants and their impact on in vitro replicative capacity. Results We selected 31 individuals infected with HIV-1 harboring frequently observed TDRM such as M41L or K103N in reverse transcriptase (RT) or M46L in protease. In all these samples, polymorphisms at non-TDRM positions were present at baseline (median protease: 5, RT: 6). Extensive analysis of viral evolution of protease and RT demonstrated that the majority of TDRM (51/55) persisted for at least a year and even up to eight years in the plasma. During follow-up only limited selection of additional polymorphisms was observed (median: 1). To investigate the impact of frequently observed TDRM on the replication capacity, mutant viruses were constructed with the most frequently encountered TDRM as site-directed mutants in the genetic background of the lab strain HXB2. In addition, viruses containing patient-derived protease or RT harboring similar TDRM were made. The replicative capacity of all viral variants was determined by infecting peripheral blood mononuclear cells and subsequently monitoring virus replication. The majority of site-directed mutations (M46I/M46L in protease and M41L, M41L + T215Y and K103N in RT) decreased viral replicative capacity; only protease mutation L90M did not hamper viral replication. Interestingly, most patient-derived viruses had a higher in vitro replicative capacity than the corresponding site-directed mutant viruses. Conclusions We demonstrate limited in vivo evolution of protease and RT harbouring frequently observed TDRM in the plasma. This is in line with the high in vitro replication capacity of patient-derived viruses harbouring TDRM compared to site-directed mutant viruses harbouring TDRM. As site-directed mutant viruses have a lower replication capacity than the patient-derived viruses with similar mutational patterns, we propose that (baseline) polymorphisms function as compensatory mutations improving viral replication capacity.


Background
The viral enzymes reverse transcriptase (RT) and protease were the first targets of antiretroviral therapy and the most commonly used drug regimens still aim at inhibiting these viral proteins [1]. In resource-rich settings, drug resistance mutations in protease and RT are detected in 10-15% of newly diagnosed HIV patients [2,3].
The majority of transmitted drug-resistant viruses contain limited resistance profiles to single drug classes. Nucleoside RT inhibitor (NRTI) mutations are the most frequently observed transmitted drug resistance mutations (TDRM). Especially thymidine analogue mutations (TAMs) M41L and T215 variants, that have been selected by drugs extensively used in the past, are often observed in newly diagnosed patients [4]. A worrying trend is the increased prevalence of non-nucleoside RT inhibitor (NNRTI) related mutations in newly diagnosed patients [3,5], as single NNRTI mutations, such as the frequently observed K103N mutation, can result in high levels of resistance against first generation NNRTIs [6]. In protease, M46I/L and L90M are the most frequently observed TDRM [2,3]. When present in combination with other protease drug resistance mutations, both M46I/L and L90M are related to reduced susceptibility to several protease inhibitors (PIs) [6].
It is generally acknowledged that most drug resistance mutations decrease the replicative capacity (RC) of HIV-1 [7,8]. As such, in the absence of drugs TDRM can revert to wild type, thereby increasing viral RC. Indeed, follow-up of untreated individuals diagnosed with a drug resistant HIV variant revealed that certain mutations with a detrimental effect on the viral RC, such as M184V in RT, after transmission to a new host often revert rapidly in the plasma [9,10]. In addition, the use of very sensitive assays shows that minority drug resistance mutations are frequently found in untreated individuals, suggestive of reversion after transmission [11,12].
The aim of our study was to gain more insight in the mechanisms causing persistence of drug resistant HIV-1 variants after transmission. Therefore, we investigated the molecular evolution of HIV-1 protease and RT harboring the most frequently observed TDRM in great detail. The majority of TDRM persisted during the followup, and only few additional polymorphisms were selected during this period. Most patient-derived viruses had a higher RC than the corresponding site-directed mutant viruses, indicating that persistence can be explained by a high replication capacity of most transmitted drug resistant HIV-1 variants.

Patients diagnosed with a transmitted drug resistant HIV-1 variant
To investigate the in vivo evolution of transmitted drug resistant HIV variants, we selected 31 patients from four European countries (Belgium, Greece, the Netherlands, Slovenia) who were diagnosed in 2001 to 2008 with an HIV variant harboring a frequently observed TDRM (prevalence >5% in patients diagnosed with HIV-1 harboring TDRM in the SPREAD-programme). Patients were included if a plasma sample was available at one year (10-14 months) after diagnosis if therapy was not yet initiated. If available, a third time point before start of treatment was investigated. Prior negative HIV tests were available for 14 patients, revealing that at least nine patients had been infected for less than two years. The majority of the patients were men having sex with men (MSM), which is the most important route of transmission in Western Europe. The median plasma HIV-RNA in our group of patients was 4.6 log copies/ ml, comparable to the median HIV-RNA observed in the SPREAD-programme in 2002-2006 (4.8 log copies/ml). The median baseline CD4 count was 653 cells/ mm 3 , which is higher than the median observed in the SPREAD programme (343 cells/mm 3 ) [3].
Surveillance studies demonstrated that most transmitted drug resistant HIV-1 variants harbor resistance against a single drug class [3,4]. In line with this observation, only 3/31 of the patients selected for this study had been diagnosed with an HIV-1 variant resistant to multiple drug classes. A total of 55 mutations at positions included in the WHO list for surveillance of transmitted drug resistant HIV-1 [26] were observed in the transmitted viruses at baseline. A single TDRM was detected in 10/16 patients with viruses harboring only NRTI-related TDRM, for the other six patients a profile of two to four TDRM was observed. The vast majority of NRTI-related TDRM were TAM-related mutations. In six of the selected patients viral variants containing a single NNRTI-related TDRM were observed. Six patients were diagnosed with HIV-1 harboring a single PI-related TDRM (Table 1). In addition to TDRM, polymorphisms were present in all baseline sequences. For variants containing RT TDRM, the median number of RT polymorphisms was 7 (range: 4-21) when compared to HXB2 and 6 to consensus B (range: [2][3][4][5][6][7][8][9][10][11][12][13][14][15][16][17][18][19]. Viruses harboring PR resistance mutations had a median  of 6 baseline polymorphisms in protease when compared to HXB2 (range: 4-9) and median of 5 when compared to consensus B (range: 3-8).

In vivo evolution of transmitted drug resistant HIV-1 variants
The vast majority (51/55) of TDRM persisted during the first year of follow-up. For 24/31 patients a third and sometimes a fourth genotypic analysis was performed at a median of 28 months (range: 14-99 months) after the first sample. During this more extensive follow-up period of up to eight years, all resistance mutations present at one year after diagnosis persisted in the plasma (Table 1).
To gain more understanding of in vivo persistence of TDRM, we performed a comprehensive analysis of in vivo viral evolution during the follow-up. Viruses harboring protease drug-resistance mutations selected a median of 1 (range: 0-1) additional polymorphisms in protease during the first year of follow-up. Likewise, viruses harboring drug-resistance mutations in RT selected a median of 1 (range 0-3) additional RT polymorphisms ( Table 2). As a measure of evolution at the nucleotide level, the p-distance between baseline and follow-up sequences was calculated. For the majority of patients, this revealed a very low p-distance between baseline and one year, confirming limited viral evolution. In line with this observation, the dN/dS ratio of the viral populations, which is an indicator of selection, did not change significantly in any patient (Table 2). However, in all transmitted viruses at least one change at a polymorphic site was observed, which is described in Table 2.

Impact of frequently observed TDRMs on in vitro RC
We determined the impact of TDRM on viral RC by introducing frequently observed drug-resistance mutations M46I, M46L or L90M in protease or M41L, M41L + T215Y or K103N in RT in the background of the lab strain HXB2 by site-directed mutagenesis ( Figure 1). Viruses were named according to mutations and origin; the prefix "SDM" indicates site-directed mutagenesis. The RC of all viral variants was determined in primary peripheral blood mononuclear cells (PBMCs), which are natural target cells for HIV. Site-directed mutants HIV-M184V, −I and -T with a known impact on RC were used as controls, and to enable comparison of RC between various experiments [27]. The difference in RC between HIV-WT, −M184V and -M184I has been demonstrated to be biologically relevant in vivo [28,29].
All mutations caused a decrease in RC as compared to HIV-WT, except for mutation L90M in protease. The reduction in RC of the M41L, M41L + T215Y and K103N variants was comparable to each other, and to controls HIV-M184V and -I. M46I and M46L in protease resulted in the most severe reduction of RC ( Figure 1).
In vitro RC of patient-derived HIV-1 variants harboring frequently observed TDRM Subsequently, the RC of frequently observed TDRM was determined in their natural genetic background (Figure 1). We constructed recombinant viruses using patientderived protease containing M46L, M46I or L90M, or patient-derived n-terminus of RT containing M41L or K103N into HXB2. In addition, two more complex transmitted viruses were studied: a protease-variant containing I54V + V82A + L90M and an RT-variant carrying M41L + T69S + L210E + T215S. Patient-derived clones are indicated by the prefix "p", followed by the TDRM.  The RC of p46I and p46L was similar to controls HIV-M184I and -V, indicating a diminished replication. The RT variant pK103N had an RC comparable to HIV-WT and the RC of pL90M was higher than HIV-WT. For M41L, it has been described that V60I and S162A function as compensatory mutations in transmitted HIV-1 variants [30]. We selected a patient-virus with M41L but without the potential compensatory mutations (pM41L). In this genetic background, the viral RC was as low as HIV-M184T and even lower than SDM-M41L. However, in vivo the variant containing this M41L mutation persisted for 8 months without selection of V60I or S162A before the patient initiated therapy (data not shown).
Interestingly, except for the pM41L variant, all patientderived viruses had a higher RC than the corresponding site-directed mutants (Figure 1). The RC of all protease mutation-harboring patient-derived viruses was higher than the corresponding SDM-viruses, and the RC of pL90M and pI54V + V82A + L90M were even higher than WT. In line with these results, the RC of pK103N and pM41L + T69S + L210E + T215S surpassed the RC of the corresponding SDM-viruses to the level of wild type virus. These observations suggest the presence of compensatory mutations in the genetic backbone of patient-derived viruses at the moment of diagnosis that are able to restore viral RC.

Discussion
In this study we strived to explain the in vivo persistence of the majority of TDRM in patients diagnosed with a drug-resistant HIV-1 variant. We selected patients diagnosed with HIV-1 containing limited profiles of TDRM, which are the most frequently transmitted variants as shown by large epidemiological studies [2,4]. In our patients, the vast majority of TDRM persisted for at least a year and up to eight years, confirming observations from previous studies that except for M184V/I, TDRM generally persist for longer than one year [10,[13][14][15][16][17][18][19][20][21][22][23][24][25].
To explore the potential role of viral RC in persistence of TDRM, we investigated the impact of TDRM on the RC. In vitro determination of RC in PBMCs demonstrated that most site-directed mutant viruses harboring 1-2 frequently observed TDRMs had a reduced RC. However, in line with in vivo persistence the majority of patient-derived viruses had a higher RC than the corresponding SDM viruses. This suggests that polymorphisms, which may be present at baseline, can act as compensatory mutations. Our extensive sequence analysis demonstrated limited evolution on polymorphic positions, suggesting that in many transmitted HIV variants harboring TDRM compensatory mutations are already present at diagnosis.
Interestingly, when present as a SDM in the commonly used lab strain HXB2, K103N decreased the RC in our experiments although this NNRTI-related mutation has been described to have a low impact in several [32][33][34] but not all [35] previous studies. This discrepancy may be due to the use of different assays or differences in replication caused by polymorphisms in lab strains. Indeed, the RC of patient-derived K103N was similar to WT virus, indicating that polymorphisms can restore viral RC. This may explain the in vivo persistence of K103N in our and previous studies [10,21].
Several papers have described the impact of some drug resistance mutations on the RC of HIV-1 [16,32,33,35]. To our knowledge, the viral RC of frequently observed protease and RT TDRM has never been compared. Our data reveal that site-directed mutations at position 46 in protease have the most severe impact on RC.
Lack of reversion of the TDRM could be explained by a relatively small viral population size resulting in limited evolution. However, the median plasma HIV-RNA level of the included patients is similar to the HIV-RNA generally observed for newly diagnosed patients in the SPREAD programme [3]. Furthermore, although viral evolution was limited, in all transmitted viral variants changes at polymorphic sites were observed, indicating that replication could result in molecular evolution.
Certain resistance mutations such as M46I in protease have been described to decrease recognition of epitopes by certain HLA types [36]. As a result, also the immune system may affect viral evolution and persistence of TDRM. However, the majority of frequently observed TDRM may not impact or can even enhance recognition of epitopes [36,37] and as such, it is unlikely that the immune system is the major driving force behind persistence of all TDRM.
We previously hypothesized based on an extensive literature study that the lack of reversion is related to the RC of transmitted HIV-1 variants harboring TDRM [9]. The currently described data confirms that TDRM may persist due to a high RC of the transmitted HIV-1 variant. Alternatively, the selection of additional mutations may restore the RC or result in compensatory fixation [30,38]. This important role of polymorphisms was supported by the differential impact of TDRM in the presence of patient-derived genetic background compared to site-directed mutants. For all but one investigated frequently observed TDRM, in vitro RC of patient-derived virus was higher than the corresponding SDM. A striking example is M46L. Although the single presence of M46L in HXB2 causes a large decrease in viral RC, this defect in RC is largely restored when M46L is present in a patient-derived genetic background.
M41L is one of the most frequently observed TDRM, and is an intriguing example emphasizing the impact of the genetic background on RC. As a single mutation, M41L in the background of wild type virus HXB2 decreased the RC. This decrease was even more pronounced in the genetic background of pM41L, which was specifically selected for this study because of the absence of known compensatory mutations V60I and S162A [30]. In sharp contrast, pM41L + T69S + L210E + T215S, the patient-derived virus with an extensive profile containing a M41L in the presence of the compensatory mutation V60I had a similar RC as wild type virus [30].
In addition, compensatory mutations may be observed outside the target gene of the antiviral compound. It has been demonstrated that mutations in gag may help to compensate the reduced protease activity conferred by resistance mutations in the protease itself [39]. Unfortunately sequencing of gag is usually not included in routine genotyping within Europe, impeding investigation of a potentially compensatory role of gag in this study. For RT, compensatory mutations may also be present in the connection domain [40], which again is not included in routine genoptyping.
For only a subset of patients we had laboratory evidence of recent infection. We cannot exclude that patients were initially infected with a viral variant harboring a more extensive resistance profile and that some of these mutations had reverted before the patients were diagnosed. As such, the observed limited evolution of pol may be a result of viral adaptation before diagnosis or may even have taken place in previous hosts. By using a more sensitive sequence method, we might have been able to increase the detection of TDRM in the included patients [11]. However, we have previously used ultra-deep sequencing to investigate the quasispecies in plasma of patients who were newly diagnosed with an HIV-1 variant harboring a single NRTI-related resistance mutation. In most patients we were unable to detect viral minority variants harboring more extensive resistance profiles in the plasma, which may be suggestive of infection with a circulating HIV-1 variant harboring a limited resistance profile [41]. It is not unlikely that onward transmission of highly stable HIV-1 variants harboring limited resistance profiles greatly contributes to the current epidemic of transmitted drug resistant HIV-1 variants. Indeed, phylogenetic studies have demonstrated that onward transmission by untreated patients is a major source of transmission of drug-resistant HIV-1 [42][43][44].
It is of great clinical importance to be able to distinguish whether transmitted drug resistant HIV-1 variants harbor complex but partially reverted resistance profiles or circulating HIV-1 variants containing limited resistance profiles. For the frequently observed NNRTI-resistance mutation K103N, it is well-known that it causes high levels of resistance against all first generation NNRTIs [45,46]. Even when K103N is present as minority variant, it can contribute to therapy failure [11]. Fortunately, the recently approved second-generation NNRTIs remain active against HIV-1 harboring a single K103N [47,48]. In contrast, we have demonstrated that the NRTI-related M41L in RT has limited impact on selection of resistance against currently used NRTIs [49]. M46I/L or L90M as a single TDRM in protease may cause low level resistance to commonly used protease inhibitors such as lopinavir.

Conclusion
In conclusion, we confirmed persistence of the most frequently observed TDRM. All transmitted HIV-1 variants harbored additional polymorphisms, with limited selection of additional mutations. Limited reversion of TDRM is in concordance with the high in vitro RC of patient-derived viruses harboring TDRM. As SDM viruses with the same TDRM as patient-derived viruses have a lower RC in vitro, we propose that polymorphisms that function as compensatory mutations (partially) restoring viral RC explain the in vivo persistence of TDRM. The stability of transmitted drug resistant HIV-1 variants can facilitate onward transmission of these viruses.

In vivo evolution Ethics statement
Ethical requirements differ between countries according to national legislation. In countries where a national surveillance system was established, legally no informed consent was needed. In other countries, approval was obtained by the institutional medical ethical review committees. All data were anonymized at national level.

Patients
Patients from four countries participating in the SPREADprogramme (Belgium, Greece, the Netherlands, Slovenia) were included. For all included patients, a baseline genotypic resistance test performed on a plasma sample within three months after diagnosis of HIV-1 infection revealed at least one mutation on a position associated with transmitted drug resistance as described in the mutation list as recommended by the WHO [26]. Patients were included on the basis of sample availability; a base line sample and a sample one year (10-14 months) later. If available, a sample at later time points were included. All included patients were at least 18 years of age and not exposed to antiretroviral therapy during the study period.

Sequence analysis
Genotypic resistance tests were performed by population sequencing of the viral protease and part of reverse transcriptase using commercially available assays or in-house methods covering at least amino acids 4-99 of protease and amino acids 30-249 of RT. All laboratories collaborated in the quality control program of ESAR to ensure high quality genotypic data [3,4]. HIV-1 subtype was determined using REGA 2.0 [50]. To investigate evolution, the p-distance and the ratio of the proportions of synonymous and nonsynonymous substitutions (dS/dN ratio) were calculated using MEGA 5.05. The p-distance is the proportion of nucleotides between two sequences that has been changed. The dS/dN ratio, a measure of selection pressure [51], was calculated with the Nei-Gojobori method and statistically tested with a Z-test. Baseline patient-derived viral protease genes harboring M46I, M46L, L90M or I54V + V82A + L90M or the Nterminus of RT containing M41L, M41L + T69S + L210E + T215S or K103N were introduced into HXB2 using the same vector system [52].
Clones were obtained and sequence analysis was performed to verify resemblance to population sequences. Subsequently, at least three recombinant virus stocks were generated by Lipofectamine 2000 (Invitrogen) transfection of HEK293T cells according to manufacturer's guidelines. TCID 50 was determined by end-point dilution in MT2 cells, demonstrating similar replication in this T cell line in all cases. A random clone was selected and quantified by p24 ELISA (Aalto Bioreagent, Dublin, Ireland) for the RC analysis.

RC analysis
PBMCs were isolated from HIV-seronegative blood donors by Ficoll-Paque density gradient centrifugation and stored in liquid nitrogen until use. To minimize differences between batches caused by variation between donors, each batch of PBMCs consisted of five combined donors. The RC of the virus panel was determined by infecting 5×10 6 phytohaemagglutinin-stimulated (2 mg/L) donor PBMCs with the equivalent of 40 ng HIV-1 p24 for two hours. Subsequently, cells were washed twice and maintained for 14 days in RPMI1640 with L-glutamine (BioWhittaker), 10% fetal bovine serum (Biochrom AG), 10 mg/L gentamicin (Gibco) and 5 U/ml IL-2. Cell-free supernatant was harvested daily for monitoring of the p24 production. The RC of either the SDM-viruses or the patient-derived viruses was compared to the RC of control viruses (WT, HIV-M184V, −M184I and -M184T). By comparing viruses containing only the mutation(s) or gene of interest in the exact same HIV-WT background, it is possible to determine the impact of these relevant mutation(s) or genes on viral RC. For all viruses, replication curves were performed in four biological replicates divided over two independent experiments. The mean p24 production of two replicates within representative experiments are indicated in Figure 1A-C for protease and 1D-F for RT. Figure 1G represents the median p24 production relative to HIV-WT of all four replicates on day 7 post infection.