The rise and fall of K65R in a Portuguese HIV-1 Drug Resistance database,despite continuously increasing use of tenofovir§

Kristof Theys a,*, Jurgen Vercauteren a, Ana B. Abecasis a,b, Pieter Libin a, Koen Deforche a,Anne-Mieke Vandamme a, Ricardo Camacho b,c

a Rega Institute for Medical Research, Katholieke Universiteit Leuven, Minderbroedersstraat 10, 3000 Leuven, Belgiumb Centro Hospitalar de Lisboa Ocidental, Lisbon, Portugalc Instituto de Higiene e Medicina Tropical, Portugal

Infection, Genetics and Evolution 9 (2009) 683–688

A R T I C L E I N F O

Article history:

Received 17 June 2008

Received in revised form 9 September 2008

Accepted 26 October 2008

Available online 6 November 2008

Keywords:

HIV

K65R

Antiviral resistance

Incidence

Guidelines

Tenofovir

A B S T R A C T

Objective: The overall prevalence of the K65R mutation in HIV-1 reverse transcriptase has increased in

treatment-experienced patients, mostly attributed to the increasing use of tenofovir (TDF). A number of

TDF-based regimens are associated with high rate of early virological failure. In this study, we evaluated

the impact of these combinations on K65R selection over time.

Methods: Treatment-experienced patients who had a genotypic resistance test at time of failure in the

period 2002–2005 in a hospital in Lisbon were included. Incidence of K65R was calculated and compared

with proportions of failing therapies in the respective year of sampling including TDF or didanosine plus

stavudine and their respective frequency of K65R selection were analyzed using classical statistics and

Bayesian Network Learning.

Results: The overall rate of K65R in the database was consistent with earlier reports. The incidence of

K65R increased in the period 2002–2004 but showed a sharp, significant decrease in 2005, despite a

continuous increase of patients failing TDF-based regimens. The proportion of patients failing certain not-

recommended regimens decreased sharply in 2005 and therefore correlated well with K65R incidence

trend.

Conclusions: This study suggests that the use of certain TDF-based regimens caused the increase in K65R

incidence over the last years.

� 2008 Elsevier B.V. All rights reserved.

Contents lists available at ScienceDirect

Infection, Genetics and Evolution

journa l homepage: www.e lsev ier .com/ locate /meegid

1. Introduction

The K65R mutation in HIV-1 reverse transcriptase (RT) is rarelyselected in treated patients. Nevertheless, in vitro and in vivo datareveal a selection by tenofovir (TDF), abacavir (ABC), didanosine(ddI), zalcitabine (ddC) and, in rare cases, stavudine (d4T) (Winterset al., 1997; Garcia-Lerma et al., 2003; Miller, 2004; Trotta et al.,2006). By itself, the mutation confers a variable degree ofphenotypic resistance to TDF, ABC, ddI, lamivudine (3TC) andemtricitabine (FTC), but an increased susceptibility to zidovudine(AZT) and, controversially, d4T (Ruane and Luber, 2004; Wainberget al., 2005). Individuals with pre-existing thymidine analogmutations (TAMs) or receiving thymidine analogs (TA) are less

§ XV International HIV Drug Resistance workshop, 13–17 June 2006, Sitges, Spain.

Abstract 121: rise and fall of the RT K65R incidence in the Portuguese resistance

database.* Corresponding author. Tel.: +32 016332151; fax: +32 016332131.

E-mail address: kristof.theys@uz.kuleuven.ac.be (K. Theys).

1567-1348/$ – see front matter � 2008 Elsevier B.V. All rights reserved.

doi:10.1016/j.meegid.2008.10.013

likely to develop the mutation (Winston et al., 2004). In contrast,K65R has been positively associated with the Q151M-associatedmultinucleoside resistance pathway (Valer et al., 2004; Trottaet al., 2006).

Despite the wide use of K65R-selecting drugs, the overallprevalence of K65R in treatment-experienced patients is quite low,presumably due to the negative impact on replication capacity(Miller, 2004; Winston et al., 2004). However, several groupsreported an increasing prevalence of K65R, ranging from 0.6% in1999 to 3.3% in 2003 and up to 11.5% in 2004 (Kagan et al., 2004;Valer et al., 2004; Winston et al., 2004). This trend is mostlyattributed to the increasing use of TDF in clinical practice since itsapproval in 2001. TDF is associated with a significantly higher riskof selection of K65R than ABC or ddI.

A number of TDF-based regimens are associated with a highrate of early virological failure or at least with decreasing CD4counts (Negredo et al., 2004; Maitland et al., 2005). Thefollowing combinations are no longer recommended in theclinical management of HIV infection: TDF + ddI,TDF + ddI + NNRTI, TDF + ddI + 3TC/FTC and TDF + ABC + 3TC/

K. Theys et al. / Infection, Genetics and Evolution 9 (2009) 683–688684

FTC (Jemsek, Abstract 51 at CROI 2002; Winston et al., 2004;Gallant et al., 2005; Torti et al., 2005). Instead, TDF + 3TC/FTC + EFV and TDF + 3TC/FTC plus a boosted protease inhibitor(PI/r) are now currently recommended TDF-based combinations(Wirden et al., 2005; Gallant et al., 2006). The aim of this studywas to evaluate the impact of the use of these combinations onthe selection of K65R over time.

2. Methods

This study is a retrospective analysis of data from a PortugueseDrug Resistance database on HIV-infected patients followed in 22hospitals located over the whole Portuguese mainland and theMadeira Archipelago. The data available to us were genotypes andtreatment history of all patients. No information was available onviral load, CD4 evolution, drug adherence or successful therapycombinations. Since January of 2001, European guidelinesrecommend resistance testing in case of treatment failure. From2001 on, the implementation of routine resistance testing fortreatment failure was constant and the vast majority of Portuguesesamples were tested at Hospital Egas Moniz in Lisbon, the majorreference laboratory. The analysis was restricted to patients with aresistance test while on treatment. Therapy failure was conse-quently considered as the need to perform resistance testing, whileon treatment with a viral load high enough to allow successfulsequencing. All available genotypes, sampled while on NRTItreatment, from the time period 2002–2005 were included. Viralisolates from a patient, sampled after the first occurrence of K65Rwere excluded from the analysis. The incidence of K65R in a timeperiod was defined as the number of unique drug-experiencedpatients with an incident case of K65R in that time period over thetotal number of unique drug-experienced patients with a least oneresistance test in the same time period. Since the patients includedwere those failing on an RTI-based regimen, all of which includedan NRTI, and since K65R is selected by and confers resistance to abroad number of NRTIs, the denominator represents an acceptablerisk population and the proposed definition can be considered

Fig. 1. Incidence of RT mutation K65R in time period 2002–2005 (striped line) and freq

incidence rather than prevalence. Incidences were modelled overtime using an univariate Poisson regression model.

Univariate Fisher exact test was used to examine thesignificance of the association between therapy combinationsand K65R emergence, and odds ratios were calculated to quantifydirection and magnitude of the association. Multivariate logisticregression was used with presence of K65R as response variableand RTI treatment history, PI experience and presence of TAMs asindependent variables. Pearson correlation coefficients werecalculated to identify relationships between K65R emergenceand trends in failing therapy combinations, by plotting K65Rincidence against the proportion of therapy failures on a particularcombination in time. Data were analyzed using R. BayesianNetworks (BNs) were constructed to investigate the direct andindirect correlations with K65R. A BN is a probabilistic model thatdescribes statistical independencies between variables. Depen-dencies are visualized in a directed acyclic graph, an arc betweentwo variables encodes an unconditional dependency, whichrepresents a direct influence. A detailed description of the use ofBNs in the context of HIV-1 antiviral resistance is describedelsewhere (Deforche et al., 2006a,b). Fisher exact testing (withBonferroni correction) was used to confirm the Bayesian associa-tions.

3. Results

A total of 1617 treatment-experienced sequences (from 1336patients) were available for analysis. In 60 patients, the K65Rmutation was present, providing an overall rate of 4.49% [3.61–5.88], which is consistent with earlier studies (Valer et al., 2004;Winston et al., 2004). Most isolates were on a first-line (34%) orsecond-line regimen (28%) and no significant difference was foundbetween K65R patients and non-K65R patients. The number ofK65R cases was 3 in 2002, 14 in 2003, 31 in 2004 and 12 in 2005.The incidence of K65R increased in the period 2002–2004 andshowed a sharp decrease in 2005 (Fig. 1), despite a continuousincrease of the number of patients failing TDF-based therapy,

uency of therapy failing in respective year of sampling including TDF (dotted line).

K. Theys et al. / Infection, Genetics and Evolution 9 (2009) 683–688 685

ranging from 1.1% in 2002 to 32.3% in 2005 (Fig. 1). A Poissonregression model, including number of K65R cases and acontinuous time variable (expressed in years) was fitted to thedata. Residual deviance, which is a measure of how much of thedata deviates from the model, was used to score the models’goodness of fit. Comparing the residual variance with a x2-distribution with two degrees of freedom, we observed that therewas no significant deviation (P = 0.32). Thus our assumption thatthe Poisson regression model adequately fits the data is valid.

K65R was identified in 42 (70%) patients failing TDF-basedcombinations, but the mutation was also present in 18 patientsnever exposed to TDF. Of the latter, 10 failed on ddI + d4T-basedregimens, 2 on AZT + 3TC + PI and 1 on respectively ddI + FT-C + EFV, ddC + NVP + PI, ddI + AZT, ABC + 3TC + PI, AZT + 3TC + NVPand 3TC + EFV + PI at time of resistance testing. The latter two andone patient on AZT + 3TC + PI had a previous history ofddI + d4T + NNRTI. Of the 16 cases of K65R on TDF + 3TC/FTC + NNRTI, 3 had a previous history of ddI + d4t + NNRTI useand 1 of TDF + ddI + EFV use. The number of TDF-based orddI + d4T-based failing regimens selecting K65R and their respec-tive rate of selection are summarized in Table 1, accounting for 87%(52/60) of the identified K65R cases (the dual combinationsTDF + ddI and TDF + ABC were always combined either with anNNRTI or a PI).

The regimens with the highest selection rates for K65R (seeTable 1) were TDF + ddI + NNRTI, TDF + ddI + 3TC/FTC andTDF + ABC + NNRTI, which is in agreement with earlier reports.The high rate of NNRTI-including regimens contrasted sharplywith the much lower rate of the respective PI-including regimen.Including a PI seems to have a protective role, which has beenpreviously reported (Ruane and Luber, 2004). Surprisingly, of thethree patients failing on TDF + 3TC/FTC + NNRTI + PI two of themselected K65R. However, both K65R cases had a previous history ofddI + d4T + NNRTI and one of them also previous TDF experience. Areasonable explanation for the protective role is the increasedgenetic barrier of the combination. Moreover, a PI is administratedmore frequently after sequential failure with NRTIs, implicating anincreased risk of previous TAM selection. Given their antagonisticrole in K65R selection, the protective effect could be due to thepresence of TAMS. The multivariate logistic regression, however,suggested that for TDF-based regimens, the inclusion of a PI wasprotective for K65R selection, independently of the (antagonistic)

Table 1Failing therapy combinations in 2002–2005 and their rate of K65R selection.

Number of therapy failures Number of K65R ca

TDF-including 290 42

TDF + ddI + PI 28 2

TDF + ddI + 3TC/FTC 9 4

TDF + ddI + NNRTI 33 15

TDF + ABC + PI 6 0

TDF + ABC + 3TC/FTC + PI 2 0

TDF + ABC + NNRTI 4 3

TDF + 3TC/FTC + NNRTI + PI 3 2

TDF + 3TC/FTC + NNRTI 68 16

TDF + 3TC/FTC + PI 35 0

TDF + NNRTI + PI 10 0

Other TDF regimens 92 0

ddI + d4T-including 219 10

ddI + d4T + PI 83 1

ddI + d4T 9 1

ddI + d4T + 3TC/FTC 3 0

ddI + d4T + NNRTI 90 8

Other ddI + d4T regimens 34 0

Summary of therapy failures in the period 2002–2005. The number of therapy failures, the

60 K65R cases identified are given. Significance of the association is given by P-values

effect of TAMs (results not shown). Given the high rate of NNRTI-including regimens, the rate of EFV and NVP in K65R selection wascompared. However, a significant difference could not be found(results not shown).

For each of the analyzed regimens, the trend in therapy failuresfor the time period 2002–2005 is shown in Fig. 2. Given a time periodand therapy combination, proportion is calculated as the number offailures under the respective regimens over the total number oftherapy failures within that time period. To test for a relationshipbetween K65R emergence and trend in therapy failures, K65Rincidence was plotted against the combinations that weresignificantly selecting K65R, and R2-values were derived from theplot. The highest R2-values were found for the combinationsTDF + ddI + NNRTIs (R2 = 0.93) and TDF + ddI + 3TC/FTC (R2 = 0.64).A less strong correlation was found for ddI + d4T + NNRTI (R2 = 0.44).The number of patients failing on TDF-based regimens and morespecific on TDF + ddI + PI, TDF + 3TC/FTC + PI, TDF + NNRTI + PI andTDF + 3TC/FTC + NNRTI increased continuously over time, probablyreflecting the increase of TDF use in clinical practice and the use ofthose recommended regimens. Their trends do not correlate withthe drop in K65R incidence in 2005. Although TDF + 3TC/FTC + NNRTI significantly selects for K65R, the correlation withK65R incidence was rather weak (R2 = 0.18).

Inclusion of TAs has been shown to prevent or delay thedevelopment of K65R. An increased use of TAs (in combination withTDF) could explain the decreased K65R selection in 2005. However,plotting therapy failures including TAs (in combination with TDF)did not correlate with K65R incidence trend (results not shown).

Time trends of failures (shown in Fig. 2), on regimenssignificantly associated with K65R selection (shown in Table 1),are correlating with the trend in K65R selection as shown in Fig. 1.The total number of K65R cases evolved from 3 in 2002, 14 in 2003,31 in 2004 and 12 in 2005. Combinations TDF + ddI + NNRTI,ddI + d4T + NNRTI and TDF + 3TC/FTC + NNRTI accounted for mostof the identified K65R cases (see Table 1). In 2004, nine cases wereselected by TDF + ddI + NNRTI and TDF + 3TC/FTC + NNRTI respec-tively, while in 2005 TDF + ddI + NNRTI selected one case of K65Rand TDF + 3TC/FTC + NNRTI selected five cases. K65R selection byddI + d4T + NNRTI decreased from five cases in 2004 to none in2005. All three cases of K65R on TDF + ABC + NNRTI were selectedin 2005. These results clearly indicate that the drop in K65Rincidence is due to a sharp decrease in selection of K65R by

ses Rate of selection (%) % of K65R cases Odds ratio P-value

14.5 70.0 10.5 <0.0001

7.1 3.0 1.8 0.3182

44.4 6.7 19.9 0.0003

45.5 25.0 25.4 <0.0001

75.0 5.0 73.0 <0.0001

66.7 3.0 46.3 0.0051

23.5 26.7 9.4 <0.0001

4.6 16.7 1 1

1.2 1.7 0.3 0.2527

11.1 1.7 2.67 0.3135

8.9 13.3 2.5 0.0266

number of K65R cases selected, the frequency of selection and the proportion of the

and odds ratios quantify the strength of the association.

Fig. 2. The number of therapy failures including the respective drug combination over the number of therapy failures in that year, expressed as percentages. Therapy

combinations that were found significantly associated with K65R selection, indicated in Table 1, are marked with an asterisk.

K. Theys et al. / Infection, Genetics and Evolution 9 (2009) 683–688686

TDF + ddI + NNRTI and ddI + d4T + NNRTI, probably due to theprogressive abandon of those combinations in clinical practice.

Bayesian Network analysis for prediction of K65R presencerevealed a direct positive association of K65R with the use of TDFand RT mutations S68G, L100I, Y181C and an antagonisticassociation with T215Y (P < 0.001). The presence of K65R haspreviously been associated with mutation Q151M. In our network,the directly associated mutation S68G is linked with mutationQ151M (Trotta et al., 2006). The combination of mutation K65Rand S68G is seen in 19 patients and the triple combination ofmutation K65R, S68G and Q151M is detected in 11 patients. Thesemutational patterns were not significant associated with particulartherapy combinations. The combination of K65R and Q151Mwithout S68G is not observed. It was suggested that the mutationS68G has a positive effect on replication capacity in the context ofthe Q151M complex (Garcia-Lerma et al., 2003). On the other hand,S68G could also have a role in the selection of K65R, explaining thepresence of S68G without Q151M (Roge et al., 2003). Theassociation with T215Y indicated a strong antagonism withK65R, which is in agreement with several other reports (Wirdenet al., 2005). The graphic representation of the network is availableas supplementary material.

4. Discussion

The K65R mutation is associated with broad cross-resistance toNRTIs, seriously compromising combination therapy whichusually includes a NRTI. Understanding trends in K56R emergenceis therefore important for researchers and clinicians treating HIVpatients. Even though K65R was initially described as the markermutation for TDF resistance, significant selection by ddI and ABChas also been observed. Winston et al. (2004) reported that the useof TDF, ddI or ABC alone was not significantly associated withincreased emergence of K65R. The combinations TDF + ddI(+ABC)were however associated with increased risk of selection. AlsoMacArthur et al. (2003) reported that the use of TDF + ddI plus anyother agent was associated with an increased incidence of theK65R mutation, compared with TDF + ABC and ABC + d4T. Wain-berg et al. (2005) concluded that mainly TDF + ddI and to a lesser

degree TDF + ABC was associated with K65R selection. von Wylet al. (2008) suggested that combinations including NNRTI and/orddI with TDF favor the selection of K65R. Regimens with ddI + d4Tas backbone have also been associated with an increased risk ofK65R selection (Roge et al., 2003).

A preliminary report by Camacho et al. (2006) already stressedthe new finding of a decreasing trend in K65R incidence, despitecontinuous increase of TDF failures. This study was intended toidentify the factors associated with the rise and fall in K65Rincidence trend. A significant association of K65R selection wasobserved with the backbones TDF + DDI and TDF + ABC plus NNRTIor 3TC/FTC and TDF + 3TC/FTC + NNRTI. K65R selection was notsignificantly different among TDF + ABC compared with TDF + ddIregimens, however, our study may be lacking the power to find thesmall differences mentioned by several authors (MacArthur et al.,2003; Winston et al., 2004; Wainberg et al., 2005). AlsoddI + d4T + NNRTI was significantly linked with K65R emergence,confirming previous findings. However, this combination dis-played lower selection rates compared with the TDF-basedregimens. The number of failing therapies including these regi-mens displayed a sharp decrease in 2005, and consequentlycorrelated well with K65R incidence trend, except for TDF + 3TC/FTC + NNRTI. The strongest correlation was observed for TDF + ddIcombinations while ddI + d4T + NNRTI showed a weaker relation-ship. Trends in failing therapies are probably due to changinginternational guidelines and resulting prescription behavior.Accordingly, these trends are expressed in the rate of K65Rselection, decreasing from 31 cases in 2004 to 12 in 2005. Amajority of identified K65R cases was selected during eitherTDF + 3TC/FTC + NNRTI or TDF + ddI + NNRTI treatment. How, theirtrend in therapy failing show different correlation patterns withK65R incidence. While a majority of K65R cases in 2003 and 2004failed on regimens no longer recommended, 5 out of 12 isolatesidentified in 2005 failed on TDF + 3TC/FTC + NNRTI, a recommendcombination. K65R presence was observed in eight patients failinga regimen not including TDF or ddI + d4T. These patients wereeither failing while on ddI- (two cases), ABC- (one case) or ddC-based (one case) regimens or had previous experience withddI + d4T + NNRTI (three cases).

K. Theys et al. / Infection, Genetics and Evolution 9 (2009) 683–688 687

The different effect of a TDF-backbone added with eitherNNRTI or PI on rate of K65R development has been described(Leon et al., 2005; Camacho et al., 2006; von Wyl et al., 2008).Since PIs require multiple mutations, compared to singlemutations needed for NNRTIs to become fully resistant, failingtherapy including NNRTI seems more likely. Bayesian Networkanalysis identified positive associations of K65R presence withRT mutations S68G, L100I and 181C, while T215Y displayed anantagonistic association with K65R. Associations of thesemutations and possible roles have been previously documentedand suggested (Miller, 2004; Camacho et al., 2006; von Wyl et al.,2008; Deforche et al., 2008). Interactions between NNRTImutations and K65R can also provide an explanation for theincreased rate of K65R selection in presence of a NNRTI. Deforcheet al. (2008) previously reported a synergistic fitness interactionbetween K65R and Y181C.

In our analysis, we included genotypes that were resultingfrom a request for resistance testing while the patient was ontherapy, and consequently for whom the viral load was highenough to obtain a sequence. Therefore we can safely assumethat the patient was failing therapy, even though virologicalresponse data were missing. Information on drug prescriptionand number of therapy successes is very hard to obtain, and wasnot available to us. Calculations of K65R incidence andproportions of patients failing therapy are therefore not basedon the total number of patients prescribed with the regimen inquestion, but are with respect to patients failing therapy and thusat risk of developing resistance. However, even though notoptimal, our strategy results in meaningful data. Failing patientsare an appropriate risk population, since continuing viralreplication is needed to develop resistance. Moreover, we arecapturing K65R at the time when it first arises, thus minimizingthe risk of overlooking K65R in patients with a fully suppressingregimen after their failure with K65R. It is therefore important tonotice that this study does not draw any conclusions about thepotency of certain combinations, since the number of therapysuccesses is unknown, but rather aims to provide an explanationfor the observed K65R incidence trend. Although TDF + 3TC/FTC + EFV, which is currently a recommended first-line TDF-based regimen, accounts for 26% of the observed K65R cases, itsselection rate is rather low and its effectiveness has been provenin clinical trials. This retrospective study is further limited by thelow number of isolates with K65R and the lack of longitudinalevaluation of resistance results.

A decrease in the incidence of RT mutation K65R was observedin 2005, despite the continuously increased use of TDF in clinicalpractice. Although TDF clearly selects for K65R, our study suggeststhat it is not only the increasing use of TDF which has caused theincrease in K65R incidence over the last years. Our findings areconsistent with the scientific literature suggesting that the use ofcertain TDF-based regimens, no longer recommended in today’sclinical practice, display an increased risk of K65R selection. Trendsin the distribution of therapy failures most likely reflect changingguidelines for treatment of HIV infection. The initial increase ofK65R incidence correlates well with the observation that themajority of therapy failures in that time period consisted ofcurrently non-recommended regimens. The decreasing incidencetrend coincides with the decreased proportion of therapy failureson non-recommended regimens and with an increase in failures ofrecommended combinations, which display much lower selectionrates. In the future it is expected that the incidence of K65R willincrease again but slowly, reflecting the recommended use ofTDF + 3TC/FTC + NNRTI (with or without PI/r) as TDF combination.Therefore these possible trends must be monitored in futurestudies.

Acknowledgements

Kristof Theys is supported by a Ph.D. grant of the Institute forthe Promotion of Innovation through Science and Technology inFlanders (IWT Vlaanderen). This work was supported by Virolab(EU IST STREP Project 027446), by FWO-Vlaanderen grantG.0266.04, by the IUAP grant P6/41 and by the KatholiekeUniversiteit Leuven through grant OT/04/43. The authors aregrateful for the training they received at the ‘‘13th InternationalBioinformatics Workshop on Virus Evolution and MolecularEpidemiology’’, Lisbon, Portugal, September 2007 (http://www.re-ga.kuleuven.be/cev/workshop/).

Kamal Mansinho, Francisco Antunes, Jose Vera, Antonio Dinis,Teresa Branco, Isabel Germano, Eugenio Teofilo, Lino Rosado, MariaJoao Aguas, Paula Fonseca, Domitılia Faria, Telo Faria, Antonio deSousa, Luis Tavares, Antonio Silva Graca, Cristina Guerreiro, FaustoRoxo, Graca Amaro, Isabel Mota, Margarida Bentes de Jesus, JosePocas, Teresa Cabral are acknowledged.

Appendix A. Supplementary data

Supplementary data associated with this article can be found, in

the online version, at doi:10.1016/j.meegid.2008.10.013.

References

Camacho, R., Theys, K., Abecasis, A., Deforche, K., Carvalho, A.P., Cabanas, J., Gomes,P., Vandamme, A.M., 2006. Rise and fall of the RT K65R incidence in thePortuguese resistance database. Abstracts of the XV International Drug Resis-tance Workshop, June 13–17, 2006, Sitges, Spain.

Deforche, K., Camacho, R., Grossman, Z., Silander, T., Soares, M.A., Moreau, Y.,Shafer, R.W., Van Laethem, K., Carvalho, A.P., Wynhoven, B., Cane, P., Snoeck,J., Clarke, J., Sirivichayakul, S., Ariyoshi, K., Holguin, A., Rudich, H., Rodrigues,R., Bouzas, M.B., Cahn, P., Brigido, L.F., Soriano, V., Sugiura, W., Phanuphak, P.,Morris, L., Weber, J., Pillay, D., Tanuri, A., Harrigan, P.R., Shapiro, J.M., Katzen-stein, D.A., Kantor, R., Vandamme, A.M., 2006a. Bayesian Network analysis ofresistance pathways against protease inhibitors. Infect. Genet. Evol. 7, 382–390.

Deforche, K., Silander, K., Camacho, R., Grossman, Z., Soares, M.A., Van Laethem, K.,Kantor, R., Moreau, Y., Vandamme, A.M., nonB Workgroup, 2006b. Analysis ofHIV-1 pol sequences using Bayesian Networks: implications for drug resistance.Bioinformatics 22, 2975–2979.

Deforche, K., Camacho, R.J., Grossman, Z., Soares, M.A., Van Laethem, K., Katzenstein,D.A., Harrigan, P.R., Kantor, R., Vandamme, A.M., non-B Workgroup, 2008.Bayesian Network analyses of resistance pathways against efavirenz and nevir-apine. AIDS 22, 2107–2115.

Gallant, J.E., Rodriguez, A.E., Weinberg, W.G., Young, B., Berger, D.S., Lim, M.L., Liao,Q., Ross, L., Johnson, J., Shaefer, M.S., 2005. ESS300009 Study. Early virologicnonresponse to tenofovir, abacavir, and lamivudine in HIV-infected antiretro-viral-naive subjects. J. Infect. Dis. 192, 1921–1930.

Gallant, J.E., DeJesus, E., Arribas, J.R., Pozniak, A.L., Gazzard, B., Campo, R.E., Lu, B.,McColl, D., Chuck, S., Enejosa, J., Toole, J.J., Cheng, A.K., 2006. Tenofovir DF,emtricitabine and efavirenz vs zidovudine. Lamivudine and efavirenz for HIV.Study 934 Group N. Engl. J. Med. 354, 251–260.

Garcia-Lerma, J.G., MacInnes, H., Bennett, D., Reid, P., Nidtha, S., Weinstock, H.,Kaplan, J.E., Heneine, W., 2003. A novel genetic pathway of human immuno-deficiency virus type 1 resistance to stavudine mediated by the K65R mutation.J. Virol. 77, 5685–5693.

Kagan, R.M., Merigan, T.C., Winters, M.A., Heseltine, P.N., 2004. Increasing preva-lence of HIV-1 reverse transcriptase mutation K65R correlates with tenofovirutilization. Antivir. Ther. 9, 827–828.

Leon, A., Mallolas, J., Martinez, E., De Lazzari, E., Pumarola, T., Larrousse, M.,Milincovic, A., Lonca, M., Blanco, J.L., Laguno, M., Biglia, A., Gatell, J.M., 2005.High rate of virological failure in maintenance antiretroviral therapy withdidanosine and tenofovir. AIDS 14, 1695–1697.

Negredo, E., Molto, J., Burger, D., Viciana, P., Ribera, E., Paredes, R., Juan, M., Ruiz, L.,Puig, J., Pruvost, A., Grassi, J., Masmitja, E., Clotet, B., 2004. Unexpected CD4 cellcount decline in patients receiving didanosine and tenofovir-based regimensdespite undetectable viral load. AIDS 18, 459–463.

MacArthur, R.D., Crane, L.R., Alvarez, D., Fairfax, M., Richmond, D., Curtis, G., 2003.Factors associated with selection of the K65R mutation: a retrospective chartreview. Antivir. Ther. 8 (Suppl. 1), 237.

Maitland, D., Moyle, G., Hand, J., Mandalia, S., Boffito, M., Nelson, M., Gazzard, B.,2005. Early virologic failure in HIV-1 infected subjects on didanosine/tenofovir/efavirenz: 12-week results from a randomized trial. AIDS 19, 1183–1188.

Miller, M.D., 2004. K65R TAMS and tenofovir. AIDS Rev. 6, 22–33.

K. Theys et al. / Infection, Genetics and Evolution 9 (2009) 683–688688

Roge, B.T., Katzenstein, T.L., Obel, N., Nielsen, H., Kirk, O., Pedersen, C., Mathiesen, L.,Lundgren, J., Gerstoft, J., 2003. K65R with and without S68: a new resistanceprofile in vivo detected in most patients failing abacavir, didanosine andstavudine. Antivir. Ther. 8, 173–182.

Ruane, P.J., Luber, A.D., 2004. K65R-associated virologic failure in HIV-infectedpatients receiving tenofovir-containing triple nucleoside/nucleotide reversetranscriptase inhibitor regimens. MedGenMed 6, 31.

Torti, C., Quiros-Roldon, E., Regazzi, M., Ladisa, N., Micheli, V., Orani, A., Patroni, A.,Caputo, S.L., Tirelli, V., Di Giambenedetto, S., Cologni, G., Costarelli, S., Garguilo,F., Manca, N., Carosi, G., 2005. Early virological failure after tenofovir + dida-nosine + efavirenz combination in HIV-positive patients upon starting antire-troviral therapy. RADAR Study Group of MASTER cohort Antivir. Ther. 10, 505–513.

Trotta, M.P., Bonfigli, S., Ceccherini-Silberstein, F., Bellagamba, R., D’Arrigo, R., Soldani,F., Zaccarelli, M., Concetta Bellocchi, M., Lorenzini, P., Marconi, P., Boumis, E.,Forbici, F., Comandini, U.V., Tozzi, V., Narciso, P., Federico Perno, C., Antinori, A.,2006. Clinical and genotypic correlates of mutation K65R in HIV-infected patientsfailing regimens not including tenofovir. J. Med. Virol. 78, 535–541.

Valer, L., Martin-Carbonero, L., deMendoza, C., Corral, A., Soriano, V., 2004. Pre-dictors of selection of K65R: tenofovir use and lack of thymidine analoguemutations. AIDS 18, 2094–2096.

von Wyl, V., Yerly, S., Boni, J., Burgisser, P., Klimkait, T., Battegay, M., Bernasconi, E.,Cavassini, M., Furrer, H., Hirschel, B., Vernazza, P., Rickenbach, M., Ledergerber,B., Gunthard, H.F., 2008. Factors associated with the emergence of K65R inpatients with HIV-1 infection treated with combination antiretroviral therapycontaining tenofovir. The Swiss HIV Cohort Study Clin. Infect. Dis. 46, 1299–1309.

Wainberg, M.A., Brenner, B.G., Turner, D., 2005. Changing patterns in the selection ofviral mutations among patients receiving nucleoside and nucleotide drugcombinations directed against human immunodeficiency virus type 1 reversetranscriptase. Antimicrob. Agents Chemother. 41, 92–99.

Winston, A., Pozniak, A., Mandalia, S., Gazzard, B., Pillay, D., Nelson, M., 2004. Whichnucleoside and nucleotide backbone combinations select for the K65R mutationin HIV-1 reverse transcriptase. AIDS 18, 949–957.

Winters, M.A., Shafer, R.W., Jellinger, R.A., Mamtora, G., Gingeras, T., Merigan, T.C.,1997. Human immunodeficiency virus type 1 reverse transcriptase genotypeand drug susceptibility changes in infected individuals receiving dideoxyino-sine monotherapy for 1 to 2 years. Antimicrob. Agents Chemother. 41, 757–762.

Wirden, M., Marcelin, A.G., Simon, A., Kirstetter, M., Tubiana, R., Valantin, M.A., Paris,L., Bonmarchand, M., Conan, F., Kalkias, L., Katlama, C., Calvez, V., 2005.Resistance mutations before and after tenofovir regimen in HIV-1 infectedpatients. J. Med. Virol. 76, 297–301.