The development of drug resistance is an important factor limiting the effectiveness of antiretroviral therapy for the treatment of HIV-1 infection, and recent guidelines have recommended the routine use of HIV-1 drug resistance testing in a number of clinical settings [1,2]. Genotypic and/or phenotypic tests may be performed, with phenotypic tests being a more direct measurement of HIV-1 susceptibility [1,3]. Recently, a number of investigators have reported a surprisingly high degree of reduced susceptibility to antiretroviral agents in untreated and newly infected individuals [particularly to the non-nucleoside reverse transcriptase inhibitors (NNRTI)] [4–8]. These estimates range from 1 to 26% depending upon the drug class, the geographic location, and the fold-reduction in antiretroviral susceptibility (cut-off) used to define drug susceptibility. Decreased susceptibility may  or may not [10,11] be of immediate clinical relevance, depending upon factors such as the drugs involved and the degree of resistance observed .
Although HIV-1 sequences vary widely around the world , and mutations which potentially contribute to HIV-1 drug resistance occur within both the clade B and non-B strains , little published data are available about the world-wide variation of drug susceptibility in untreated individuals. Furthermore, the relatively small numbers of samples utilized in previously published studies (n < 150) limits somewhat the utility of these data in establishing a normal range of values.
Here, we determined the phenotypic susceptibility of clinically derived HIV-1 recombinant viruses for each of 14 antiviral drugs in over 1000 treatment-naive individuals participating in seven clinical studies in the USA, Germany, Canada, or South Africa. These data were used to approximate the natural variability of antiretroviral susceptibility in untreated individuals from geographically diverse populations.
Materials and methods
Patients and samples
The plasma samples used in this study were obtained from HIV-1 infected individuals in the USA (n = 351; from 10 clinical centres across six states), Germany (n = 306; five sites ), Canada (n = 265; one site) and South Africa (n = 358; one site), predominantly since 1997. Plasma samples were shipped on dry ice to Virco and stored at −70 °C until phenotypic and genotypic analyses were performed.
Phenotypic drug susceptibility analysis was performed by a recombinant assay  (Virco, Mechelen, Belgium) with the modifications described by Pauwels et al. . Briefly, protease and reverse transcriptase (RT)-coding sequences were amplified from patient-derived viral RNA with HIV-1 specific primers. After homologous recombination of amplicons into a proviral clone from which the protease and RT-coding sequences were deleted, the resulting recombinant viruses were harvested, titrated, and used for testing of in vitro susceptibility to antiviral drugs. The results of this analysis are expressed as fold-change values, which reflect the fold-increase in the mean 50% inhibitory concentration (IC50) of a particular drug when it was tested with patient-derived recombinant virus isolates relative to the mean IC50 of the same drug obtained when tested with a wild-type reference virus isolate (strain IIIB).
HIV RNA extraction and genotyping
Viral RNA was extracted from 200 μl samples of patient plasma with the QIAamp viral RNA extraction kit (Qiagen, Hilden, Germany) according to the manufacturers instructions. cDNA encompassing part of the pol gene was produced using Expand RT (Boehringer Mannheim, Mannheim, Germany) as described previously . A 2.2 kb fragment encoding the protease and RT regions was then amplified by nested PCR using primers and conditions described previously . The PCR products were genotyped by dideoxynucleoside-based sequence analysis. Samples were sequenced using the Big Dye terminator kit (Applied Biosystems, Warrington, UK) and resolved on an ABI 377 DNA sequencer as described [18,19]. The results of the genotypic analysis are reported as amino acid changes at positions in the RT gene compared to the gene sequence of wild-type reference strain HXB2; clade comparisons were made on the basis of similarity to ‘prototype’ sequences from HIV-1 clades .
Determination of virtual phenotype
Virtual phenotypes of test isolates were determined by retrieving phenotypes of samples that match the genotypes of the test isolates from the Virco relational database in terms of both the presence and the absence of mutations at a number of sites in the HIV RT and protease . This database comprises > 18 000 HIV-1 isolates with both genotype and drug susceptibility phenotype data. Virtual phenotypes of test isolates were the geometric mean fold-change values for each drug in the matching group . At least 10 genotype-phenotype database matches were required for each drug before a virtual phenotype was accepted.
Antiretroviral susceptibility in treatment-naive HIV-1-infected patients
The average phenotypic susceptibility (measured in a recombinant virus assay) for 14 currently available antiretroviral drugs was determined for > 1000 treatment-naive, HIV-1-positive individuals participating in seven clinical studies that included participants from four countries (USA, Germany, Canada, and South Africa). These data were compared with the average phenotypes determined for genetically wild-type viruses for each of the three drug classes (sample sizes ranging from n = 2669 to n = 8774) derived from the Virco relational database  (Fig. 1). For both treatment-naive and genotypically ‘wild-type’ viruses, the geometric mean level of phenotypic resistance observed was close to 1 for all of the drugs examined, with a range of 0.6–1.7-fold resistance compared to the wild-type reference virus IIIB. The highest fold-resistance values in treatment-naive patients were observed for the NNRTI: delavirdine (DLV; approximately 1.7-fold); nevirapine (NVP; approximately 1.3-fold); and efavirenz (EFV; approximately 1.2-fold). Moreover, there was relatively little variation in mean phenotypic resistance to the drugs from study to study (data not shown).
Phenotypic susceptibility in the drug-naive population approximated a symmetrical log-normal distribution. It is important to note that there were remarkably few ‘outliers’ with greatly decreased drug susceptibility in these studies (possibly representing cases of transmission of resistant isolates), and that these have little effect on the overall distribution, as the values are based upon the entire distribution. However, there were differences in the range of treatment-naive HIV-1 susceptibility with different drugs (for examples see Fig. 2). In particular, the frequency distribution of susceptibility to the NNRTI was much broader than for the nucleoside analogues or protease inhibitors, with the greatest value (> 10-fold) for the mean + 2 standard deviations (approximately 97.5% of samples) observed for DLV. The broader distribution observed for the NNRTI class also implies that this class exhibits the greatest proportion of samples < 50% of the mean value (corresponding to some degree of hypersusceptibility) of the three drug classes.
Determination of new resistance cut-off values for antiretroviral drugs
The upper ‘normal’ ranges (defined here as mean + 2 SD or approximately 97.5%) vary for each drug, being higher for NNRTI and lower for dideoxy-nucleosides and some protease inhibitors. For example, HIV-1 from the large majority of treatment-naive patients exhibited < 8-, < 10-, and < 6-fold decreased susceptibility to NVP, DLV, and EFV, respectively. However, the same HIV-1 variants from these individuals exhibited < 3-fold decreased susceptibility (mean ± 2 SD) to stavudine (d4T), abacavir (ABC), indinavir (IDV), saquinavir (SQV), and amprenavir (APV). New biological cut-off values for individual antiretroviral drugs based on their phenotypic distribution in treatment-naive patients as well as new Virtual Phenotype cut-offs (optimized for specificity and sensitivity using receiver-operator curves to reflect most closely the actual phenotype cut-off) are shown in Table 1. In general, actual phenotype and Virtual Phenotype cut-offs were identical, although improvements in the sensitivity and specificity of categorization were obtained by lowering the Virtual Phenotype cut-offs for didanosine (ddI), zalcitabine (ddC), d4T, SQV and APV (Table 1).
Geographic differences in primary resistance to antiretroviral agents
The effects of geographic location on the range of susceptibility in treatment-naive patients was assessed by stratifying the phenotypic data according to point of sample origin (USA, Germany, Canada, or South Africa). The mean fold-resistance of HIV-1 was similar (± 0.5-fold) for all antiretroviral drugs regardless of geographic location (Fig. 3a). Approximately 97.5% of all isolates (mean + 2 SD) had < 2.5–4.0, < 3.0–5, and < 4–15 fold-resistance to five protease inhibitors, six nucleoside analogues, and three NNRTI, respectively. Similar patterns were observed for each location, with the greatest range observed for the three drugs in the NNRTI class.
Effect of HIV-1 clade on distribution of antiviral drug susceptibility in treatment-naive patients
The inclusion of data from patients from South Africa allows an explicit examination of the effect of viral clade on phenotypic susceptibility. Of the 271 samples from South Africa with available viral phenotype and viral RNA sequence data, 190 were most similar to prototype HIV-1 clade C, 74 to clade B and 7 to clade F. There was little difference in the average or range of phenotypic susceptibility for any antiviral agent tested between clade B and clade C for these individuals (Fig. 3b). The number of clade F viruses was insufficient to generate statistically reliable results, but the values obtained were also similar for most antiviral agents (data not shown).
Prevalence of resistance using a > 4-fold cut-off, and new biological, and Virtual Phenotype cut-offs
The prevalence of samples categorized as having reduced susceptibility to each drug using the previous > 4-fold cut-off and the new biological and Virtual Phenotype cut-offs was determined using 5000 randomly selected HIV-1 samples submitted to Virco for routine drug resistance testing during 1999 and 2000 (Fig. 4a and 4b). Actual phenotypes of these samples had been determined previously. Virtual Phenotypes  were derived from the corresponding genotypic data for each sample for each drug (resulting in approximately 70 000 determinations in total). Overall, the prevalence of actual phenotypic NNRTI reduced susceptibility was lower with the new biological cut-offs (NVP, −6.2%; DLV, −12.5%; EFV, −3.1%), whereas the prevalence of dideoxy-nucleoside reduced susceptibility was higher (ddI, +3.0%; ddC, +2.0%, d4T, +5.4%). Although the normal range of drug sensitivity was greatest for the NNRTI, the new cut-offs had the greatest effect on categorization to the dideoxy-nucleosides, with a 77% increase in the prevalence of reported susceptibility decreases for d4T (Table 1).
Typically, 2.5-fold, 4-fold, or 10-fold decreases in drug susceptibility have been used to define decreased antiretroviral drug susceptibility (see for example [23,24]). In this analysis of a large number of samples originating from three continents, we found that no single arbitrary cut-off could accurately reflect the range of drug susceptibility found in treatment-naive individuals for the currently available antiretroviral agents. There were large differences in the phenotypic distribution from drug to drug, with biological cut-off values (based upon a change in drug susceptibility of 2 SD from the mean) ranging from 2.5-fold for SQV and APV to 10-fold for DLV. Interestingly, there was no evidence for large geographical differences in phenotypic susceptibility in treatment-naive patients, with the variability in phenotypic susceptibility in samples from the USA, Germany, Canada, or South Africa being only ± 0.5-fold. These data represent the largest phenotypic study of non-clade B viruses we are aware of, and indicate that there is little difference in phenotypic susceptibility, at least between clades B and C.
Various criteria related to phenotypic susceptibility [4–8,23,24] and/or the presence or absence of RT- and protease-associated mutations [5–8,25,26] have been used to assess HIV-1 drug resistance in newly infected and untreated individuals. Phenotypic evidence of reduced drug susceptibility and/or genotypic evidence of mutations associated with resistance have suggested that significant numbers of newly infected, treatment-naive patients harbour drug resistant HIV-1. Although large numbers of ‘secondary polymorphisms’ (i.e., mutations not classically linked with drug resistance) have been observed previously in HIV-1 variants from untreated individuals (see for example [7,8,27]), their clinical relevance remains to be established. Moreover, the prevalence of primary resistance has been reported to range widely with geographic location, the techniques used and the reporting methods . For example, the prevalence of drug resistance mutations were reported in 0% of recent seroconverters in British Columbia, Canada , 13% of treatment-naive HIV-1 patients in Spain , 16% of newly infected individuals in the USA (New York and Los Angeles) , and 26% of treatment-naive military personnel with acquisition of HIV-1 infection in the USA or overseas . These data have been interpreted in such a way that transmission of drug-resistant HIV-1 is increasingly perceived as a serious emerging problem, particularly for the NNRTI class. The data presented here should provide a useful guide for defining the phenotypic range beyond which there is evidence of reduced susceptibility, and suggest that different ranges are appropriate for different drugs.
There has been recent debate regarding the relevance of cut-off values currently in use, particularly with respect to the dideoxy-nucleoside drugs. These cut-offs are usually the same value for each drug tested and are determined not by clinical criteria but, for example, by the assay variability seen on repetitive testing of a single wild-type standard virus. The approach presented here uses a reasonably large and diverse sample of HIV-1 isolates to establish the ‘normal’ range of drug susceptibility. Mean and standard deviation values of fold-change in susceptibility, relative to a laboratory wild-type standard, revealed that the patient samples displayed inherently different degrees of susceptibility to each drug. We have recently re-defined the phenotypic cut-off values used in the Antivirogram assay based on the observed natural phenotypic variability of treatment-naive individuals presented here: 2 SD from the mean for each drug was chosen as the cut-off between sensitive (within normal range) and resistant (above normal range). Subsequently, we have determined the impact of these cut-off changes on the apparent degree of nucleoside analogue resistance in 5000 clinical samples submitted for routine resistance testing. It is worth noting that actual and Virtual Phenotypes showed greater concordance using the new cut-offs than with the arbitrary > 4-fold cut-off. As expected, there was a significant increase in the overall prevalence of ddI, ddC and d4T resistance.
It is increasingly clear that the interpretation of phenotypic and genotypic data in the context of drug resistance is extremely complex. For example, d4T resistance may involve at least 15 polymorphisms , lamivudine (3TC) susceptibility may occur in the absence of the classical M184V signature mutation , and mutations such as the M184V, which decreases the effect of zidovudine (ZDV) resistance mutations M41L and T215Y, or the Y181C NVP mutation that restores susceptibility to ZDV in the presence of T215Y may be present . These complexities limit the usefulness of arbitrary cut-offs related to the presumption of drug resistance based on the presence of an arbitrarily selected subgroup of drug resistance-associated mutations.
It is important to note that reductions in antiviral susceptibility outside the normal range of biological variation may not necessarily be of sufficient magnitude to confer loss of antiretroviral activity in vivo, and that this study did not address clinical outcome. Indeed, reductions in plasma viral load can be attributed to combination therapy with an agent even in the face of very high level resistance . Conversely, it is possible that even reductions of susceptibility within the ‘normal range’ may be sufficient to impair optimum antiviral response.
There are some limitations of this study, specifically: large numbers of samples from other locations in Africa, Asia or South America were not available for analysis, a minor proportion of individuals may have been infected with drug resistant isolates, and the data represent only individuals with access to resistance testing or clinical trials. Nevertheless, this data set represents the largest study of drug susceptibility in treatment-naive individuals to date. Several factors argue against a significant proportion of these individuals harbouring transmitted drug-resistant HIV-1: the phenotypic data was log-normally distributed in a symmetrical fashion with few outliers, so the inclusion of a small number of individuals with resistance would have little effect on the results. Furthermore, there was no evidence of significant geographic or clade variation, and drug resistance trends were broadly similar in areas with dissimilar histories of exposure to antiviral agents (particularly South Africa). Finally, the classes of drugs tended to behave similarly in all analyses, with the NNRTI class having the highest range, despite being used in the general population to a lesser extent than the other classes of drugs.
We believe such data from large, diverse populations of HIV-1-infected individuals are critical for defining and standardizing the quantification of drug resistance, as well as for establishing the genotypic profiles that contribute to reduced drug susceptibility outside the ‘normal’ ranges.
The authors thank all participants and organizers of these trials, the Virco employees who performed the genotypic and phenotypic tests, and Biochem Pharma for providing access to the samples from South Africa.
1. Hirsch MS, Brun-Vezinet F, D'Aquila RT. et al
. Antiretroviral drug resistance testing in adult HIV-1 infection: recommendations of an International AIDS Society-USA Panel. JAMA 2000, 283: 2417–2426.
2. EuroGuidelines Group. Clinical and Laboratory Guidelines for use of HIV-1 drug resistance testing as part of treatment management. AIDS 2001, 15: 309–320.
3. Harrigan PR, Cote HC. Clinical utility of testing human immunodeficiency virus for drug resistance. Clin Infect Dis 2000, S2: S117–S122.
4. Yerly S, Kaiser L, Race E, Bru JP, Clavel F, Perrin L. Transmission of antiretroviral-drug-resistant HIV-1 variants. Lancet 1999, 354: 729–733.
5. Wegner SA, Brodine SK, Mascola JR. et al
. Prevalence of genotypic and phenotypic resistance to anti-retroviral drugs in a cohort of therapy-naive HIV-1 infected US military personnel. AIDS 2000, 14: 1009–1015.
6. Salomon H, Wainberg MA, Brenner B. et al
. Prevalence of HIV-1 resistant to antiretroviral drugs in 81 individuals newly infected by sexual contact or injecting drug use. Investigators of the Quebec Primary Infection Study.
AIDS 2000, 14: F17–F23.
7. Little SJ, Daar ES, D'Aquila RT. et al
. Reduced antiretroviral drug susceptibility among patients with primary HIV infection. JAMA 1999, 282: 1142–1149.
8. Boden D, Hurley A, Zhang L. et al
. HIV-1 drug resistance in newly infected individuals. JAMA 1999, 282: 1135–1141.
9. Hecht FM, Grant RM, Petropoulos CJ. et al
. Sexual transmission of an HIV-1 variant resistant to multiple reverse-transcriptase and protease inhibitors. N Engl J Med 1998, 339: 307–311.
10. Harrigan PR, Verbiest W, Larder B. et al
. Impact of moderate decreases in baseline NNRTI susceptibility on response to antiretroviral therapy. Antiviral Ther 2000, 5 (suppl 3): 68.68.
11. Bacheler L, Ploughman L, Hertogs K, Larder B. Impact of baseline NNRTI resistance on the efficacy of efavirenz combination therapy in NNRTI therapy naive patients (study DMP 266-006). Antiviral Ther 2000, 5 (suppl 3): 70.70.
12. Korber BT, Osmanov S, Esparza J, Myers G. The World Health Organization Global Programme on AIDS proposal for standardization of HIV sequence nomenclature. WHO Network for HIV Isolation and Characterization.
AIDS Res Hum Retroviruses 1994, 10: 1355–1358.
13. Cornelissen M, van den Burg R, Zorgdrager F, Lukashov V, Goudsmit J. Pol gene diversity of five human immunodeficiency virus type 1 subtypes: evidence for naturally occurring mutations that contribute to drug resistance, limited recombination patterns, and common ancestry for subtypes B and D. J Virol 1997, 71: 6348–6358.
14. Miller V, Hertogs K, Bloor S, et al. Prevalence of phenotypic and genotypic resistance in recently infected, antiretroviral inexperienced patients: The German HIV-research initiative cohort. Seventh European Conference on Clinical Aspects and Treatment of HIV-Infection.
Lisbon, October 1999 [abstract 381].
15. Kellam P, Larder BA. Recombinant virus assay: a rapid, phenotypic assay for assessment of drug susceptibility of human immunodeficiency virus type 1 isolates. Antimicrob Agents Chemother 1994, 38: 23–30.
16. Pauwels, R. Hertogs K, Larder BA, et al. Second International Workshop HIV Drug Resistance and Treatment Strategies.
Lake Maggiore, Italy, June 1998 [abstract 51].
17. Hertogs K, de Bethune MP, Miller V. et al
. A rapid method for simultaneous detection of phenotypic resistance to inhibitors of protease and reverse transcriptase in recombinant human immunodeficiency virus type 1 isolates from patients treated with antiretroviral drugs. Antimicrob Agents Chemother 1998, 42: 269–276.
18. Larder BA, Kellam P, Kemp SD. Convergent combination therapy can select viable multidrug-resistant HIV-1 in vitro. Nature 1993, 365: 451–453.
19. Larder BA, Kohli A, Kellam P, Kemp SD, Kronick M, Henfrey RD. Quantitative detection of HIV-1 drug resistance mutations by automated DNA sequencing. Nature 1993, 365: 671–673.
20. Kuiken CL, Foley B, Hahn B, et al.
Eds. Human Retroviruses and AIDS 1999: A Compilation and Analysis of Nucleic Acid and Amino Acid Sequences.
Los Alamos:Los Alamos National Laboratory; 1999.
21. Larder BA, Kemp SD, Hertogs K. Quantitative prediction of HIV-1 phenotypic drug resistance from genotypes: the virtual phenotype. Antiviral Ther 2000, 5 (suppl 3): 49.49.
22. Hertogs K, Van Houtte M, Larder B, Miller V. Testing for HIV-1 drug resistance: New developments and clinical implications. Recent Res Devel Agents Chemother 1999, 3: 83–104.
23. Condra JH. Resistance to HIV protease inhibitors. Haemophilia 1998, 4: 610–615.
24. Perez-Elias MJ, Lanier R, Munoz V. et al
. Phenotypic testing predicts virological response in successive protease inhibitor-based regimens. AIDS 2000, 14: F95–F01.
25. Gomez-Cano M, Rubio A, Puig T. et al
. Prevalence of genotypic resistance to nucleoside analogues in antiretroviral-naive and antiretroviral-experienced HIV-infected patients in Spain. AIDS 1998, 12: 1015–1020.
26. Alexander CS, Dong W, Schechter MT. et al
. Prevalence of primary HIV drug resistance among seroconverters during an explosive outbreak of HIV infection among injecting drug users. AIDS 1999, 13: 981–985.
27. Balotta C, Berlusconi A, Pan A. et al
. Prevalence of transmitted nucleoside analogue-resistant HIV-1 strains and pre-existing mutations in pol reverse transcriptase and protease region: outcome after treatment in recently infected individuals. Antiviral Ther 2000, 5 (Suppl 3): 7–14.
28. Little SJ. Transmission and prevalence of HIV resistance among treatment-naive subjects. Antiviral Ther 2000, 5: 33–40.
29. Brodine SK, Shaffer RA, Starkey MJ. et al
. Drug resistance patterns, genetic subtypes, clinical features, and risk factors in military personnel with HIV-1 seroconversion. Ann Intern Med 1999, 131: 502–506.
30. Wang D, Bloor S, Larder BA. The application of neural networks in predicting phenotypic resistance from genotypes for HIV-1 protease inhibitors. Antiviral Ther 2000, 5 (suppl 3): 51–52.
31. Hertogs K, Bloor S, De Vroey V. et al
. A Novel human immunodeficiency virus type 1 reverse transcriptase mutational pattern confers phenotypic lamivudine resistance in the absence of mutation 184. Antimicrob Agents Chemother 2000, 44: 568–573.
32. Larder BA, Kemp SD, Harrigan PR. Potential mechanism for sustained antiretroviral efficacy of AZT-3TC combination therapy. Science 1995, 269: 696–699.