Intersubtype BF Recombinants of HIV-1 in a Population of
Injecting Drug Users in Argentina
Alex Espinosa, MS,* Moira Vignoles, MS,† Manuel Go´mez Carrillo, PhD,† Haynes Sheppard, PhD,*
Richard Donovan, PhD,* Liliana Martı´nez Peralta, PhD,† Diana Rossi,‡ Graciela Radulich,§
Horacio Salomo´n, PhD,† and Mercedes Weissenbacher, PhD†
Summary:The presence of recombinant intersubtypes of HIV-1 in Argentina has been reported since the mid-1990s. In this study, se-quences of a region of thegag,pol, andvpugenes of HIV-1 were analyzed in samples of 21 injection drug users (IDUs) residing in the suburbs of the city of Buenos Aires. Genomic characterization and identification of recombination sites were made comparing the 3 re-gions with reference isolation sequences of subtypes B, F, C, A, and B/F recombinants: CRF12_BF and non-CRF12_BF sequences. Sub-type assignment of the analyzed segments was phylogenetically con-firmed. All the samples turned out to be BF recombinants in at least 1 of the 3 studied genes. Twelve samples (57%) had the same pattern as the Argentinean CRF12_BF, whereas in the rest, the pattern differed in at least 1 of the 3 genes. The relation of these fragments to the CRF12_BF was phylogenetically verified. These results indicate the predominance of BF recombinants and the presence of a high percent-age of sequences closely related to the CRF12_BF in the IDU popu-lation in Argentina and suggest a possible association between viral variants and the transmission route.
Key Words:Argentinean injecting drug user, intersubtype recombi-nant, circulating recombinant form
(J Acquir Immune Defic Syndr2004;36:630–636)
T
he molecular epidemiology of HIV-1 has become increas-ingly important as viral subtypes are becoming more dis-persed worldwide. There are currently at least 9 circulating genetic subtypes of HIV-1 (A–K) within group M,1with sub-type B being predominant in Europe, the Americas, and Aus-tralia.2Although the dominant subtype in South America is B, there is evidence of other subtypes circulating through differ-ent regions of the contindiffer-ent, making the epidemic much more complex.3–6Recombinants between the B and F subtypes of HIV-1 were described in Brazil7and Argentina.8,9 More re-cently, Carr et al10showed that BF recombinants were tempo-rally and geographically widespread in South America and es-tablished a new circulating recombinant form (CRF12_BF) and other recombinants in a heterosexual population from Ar-gentina, Uruguay, and Bolivia.The HIV-1 subtype distribution pattern related to sexual behavior in Argentina was described by partial viral character-ization in men who have sex with men (MSM) and heterosex-ual populations.5,11A high percentage of subtype B sequences was found in MSM in contrast to a high percentage of subtype F found in the heterosexual population. These results suggest a close relation between different risk factors and subtype distribu-tion in the Argentinean epidemic during the same period of time. Spreading of HIV-1 recombinant forms in injection drug user (IDU) populations was described in other regions of the world. In some of these epidemics, an explosive dissemination of these genetic variants among the IDU population was seen. In Vietnam and China, genetic diversity of AE recombinants among IDUs was lower than that of persons infected by the sexual route, providing evidence of the recent introduction of these variants in the IDU population.12–14
In our study, we examined the subtypes and recombi-nation patterns of 3 genomic regions of HIV-1 in a group of IDUs from Argentina and determined their relation to the CRF12_BF previously described in a heterosexual population.
METHODS
Study Design
Study subjects were 75 HIV-1–positive IDUs (61 male and 14 female) from the Buenos Aires city surroundings. The samples were collected from June 2000 to March 2001; how-ever, we used only those plasma specimens for which we were Received for publication July 10, 2003; accepted September 22, 2003.
From the *Viral and Rickettsial Disease Laboratory, California Department of Health Services, Richmond, CA; †Centro Nacional de Referencia para el SIDA, Departamento de Microbiologı´a, Facultad de Medicina, Univer-sidad de Buenos Aires, Buenos Aires, Argentina ; ‡Intercambios Asocia-cio´n Civil, Buenos Aires, Argentina; and §AsociaAsocia-cio´n El Retoño, Buenos Aires, Argentina.
Alex Espinosa was supported by an appointment to the Emerging Infectious Diseases Fellowship Program administered by the Association of Public Health Laboratories and funded by the Centers for Disease Control and Prevention. This work was supported in part by National Institutes of Health cooperative agreement U01-AI46725 and with funds granted by UNAIDS, Pan American Health Organization/World Health Organiza-tion, and the Spanish Agency for International Cooperation.
Reprints: Manuel Go´mez Carrillo, Centro Nacional de Referencia para el SIDA, Facultad de Medicina, Universidad de Buenos Aires. Paraguay 2155 Piso 11 (C1121ABG), Capital Federal, Buenos Aires, Argentina (e-mail: mcarrill@fmed.uba.ar).
able to obtain polymerase chain reaction (PCR) products for all 3 regions of interest (ie,gag,pol,vpu). This resulted in only 21 of the 75 specimens being used in the phylogenetic analysis. Written informed consent was obtained from all the study par-ticipants. Along with each specimen, a questionnaire was filled out by the patients regarding sociodemographic parameters, including (among others) gender, age, drug injection fre-quency, and needle sharing. These parameters were associated with the viral characterization done in this study.
Viral Load
Plasma HIV-1 viral load was determined by RNA quan-tification by means of the reverse transcriptase (RT) PCR as-say using the Roche Amplicor HIV-1 Monitor test, version 1.5 (Roche Molecular Systems, Branchburg, NJ) with a limit of detection of 400 copies/mL.
Detection of Early HIV-1 Infection
To determine early HIV-1 infection in the IDU plasma samples, a sensitive/less sensitive enzyme immunoassay (EIA) testing strategy (detuned assay) was used as described by Janssen et al15with the Vironostika HIV-1 MicroElisa Sys-tem (Organon Teknika, Durham, NC). This assay aimed to correlate the time of infection with the subtype characteriza-tion so as to estimate the subtypes currently being transmitted in this population.
RNA Extraction and Reverse Transcription
HIV RNA isolation from plasma was done using the QIAamp Viral RNA Miniprep Kit (Qiagen, Valencia, CA). The reverse transcription ofgagandvpuwas done using the ThermoScript RT-PCR System (Life Technologies, Rockville,
MD) with primer MKenvN (5⬘
CTGCCAATCAGGGAAG-TAGCCTTGTGT 3⬘) forvpuand primer G01 (5⬘ AGGGGTC-GTTGCCAAAGA 3⬘) forgag. The cycling conditions were as follows: 60°C for 60 minutes and 85°C for 5 minutes. RNA isolation from plasma and reverse transcription of polwere done using the ViroSeq system (Applied Biosystems, Foster City, CA) according to the directions of the manufacturer.
Polymerase Chain Reaction
The RT-PCR ofgagandvpuwas followed by a nested PCR. A first-round PCR was run using primer sets MKenvN
and MKenvA (5⬘
GGCTTAGGCATCTCCTATGGCAG-GAAGAA 3⬘) forvpuor G00 (5⬘ GACTAGCGGAGGCTA-GAAG 3⬘) and G01 forgagamplification. The lengths of the first-round PCR products ofvpuandgagwere 3190 base pairs (bp) and 1467 bp, respectively. A second-round PCR was performed using primer sets ACC7 (5⬘ CTATGGCAGGAA-GAAGCGGAGA 3⬘) and ZM140E (5⬘ GGGGTCAACTTTA-CACATGGCTTT3⬘) forvpuor G05 (5⬘
TGTTGGCTCTG-GTCTGCTCT 3⬘) and G20 (5⬘
GTATGGGCAAGCAGG-GAGCTAGAA 3⬘) forgagamplification. The lengths of the second-round PCR products ofvpuandgagwere 602 bp and
1214 bp, respectively. The cycling conditions for the first- and second-round PCRs forgagwere as follows: 94°C for 10 min-utes, followed by 35 cycles of 94°C for 1 minute, and 55°C for 1 minute and 72°C for 6 minutes, followed by a final extension at 72°C for 15 minutes. The cycling conditions for the first-and second-round PCRs forvpuwere as follows: 94°C for 10 minutes, followed by 35 cycles of 94°C for 1 minute, and 55°C for 1 minute and 72°C for 1 minute, followed by a final exten-sion at 72°C for 15 minutes.
DNA Sequencing
Subtype characterization of the studied population was performed by sequencing a region of the gag,pol, andvpu genes. The PCR products were purified with a Centricon-1 col-umn (Amicon, Danvers, MA) and then used as a template for direct sequencing on an automated ABI Prism 377 DNA sequencer using the ABI Prism Big Dye Terminator Cycle Se-quencing Ready Reaction Kit (Applied Biosystems). In addi-tion to the primers used in the second-round PCR
amplifica-tion, 7 others—ZIF (5⬘
TGGGTCACAGTCTATTATGGG-GTACCT 3⬘), ES32 (5⬘
CTGCTTTGGTATAGGATCTTG-3⬘), ES34 (5⬘ GCCTGAGCATCTGATGCAC 3⬘), JL99
(5⬘ TTTAGCATCTGATGCACAAAATAG 3⬘), JL100
(5⬘ GGGGTCTGTGGGTACACAGGCATGTGT 3⬘), ZER
(5⬘ GGGCTGGGATCTGTGGGCACACAGGCA 3⬘), and
E18 (5⬘ TTGTGGGTCACAGTCTATTATGG 3⬘)—were
used to determine the vpu sequence, and 3 others—G25 (5⬘ATTGCTTCAGCCAAAACTCTTGC 3⬘), G55 (5⬘
ATTT-CTCCCACTGGGATAGGTGG 3⬘), and G60 (5⬘
CAGC-CAAAATTACCCTATAGTGCAG 3⬘)—were used for gag
sequencing. The cycling conditions forvpuandgag sequenc-ing were as follows: 25 cycles of 96°C for 10 seconds, 50°C for 5 seconds, and 60°C for 4 minutes. Sequencing products were purified with the DyeEx Spin Kit (Qiagen) before loading the gel. Polsequences were obtained using the ViroSeq system (Applied Biosystems).
Phylogenetic Analysis
Nucleotide sequences* obtained from all primers ofgag, pol, orvpuwere aligned with the reference strain HXB2; sub-sequently, consensus sequences were constructed for each pa-*The nucleotide sequences reported in this study have been submitted to the GenBank sequence database under the following accession numbers: AY140107, AY140108, AY140109, AY140110, AY140111, AY140112, AY140113, AY140114, AY140115, AY140116, AY140117, AY140118, AY140119, AY140120, AY140121, AY140122, AY140123, AY140124, AY140125, AY140126, AY140127, AY140128, AY140129, AY140130, AY140131, AY140132, AY140133, AY140134, AY140135, AY140136, AY140137, AY140138, AY140139, AY140140, AY140141, AY140142, AY140143, AY140144, AY140145, AY140146, AY140147, AY140148, AY140149, AY140150, AY140151, AY140152, AY140153, AY140154, AY140155, AY140156, AY140157, AY140158, AY140159, AY140160, AY140161, AY140162, AY140163, AY140164, AY140165, AY140166, AY140167, AY140168, and AY140169.
FIGURE 1.Phylogenetic analysis of 21 injection drug user samples from Argentina. a, Neighbor-joining tree with bootstrapping ofgag,pol, andvpuassembled sequences. Branches of a cluster containing samples related to the CRF12_BF are drawn in bold. Significant bootstrap values are placed next to the nodes. The genetic distance corresponding to the length of the branches is shown in the bottom line. b, Genetic map of HIV-1. Shaded areas indicate each assembled segment used in the phylogenetic analysis (⽧, CRF12_BF;◊, BF recombinant non-CRFs).
tient sample using the Sequencher version 4.1.2 (Gene Codes Corporation, Ann Arbor, MI). Each consensus sequence was then screened by the BLAST 2.0 HIV-1 subtyping program (National Center for Biotechnology Information, Bethesda, MD) to search for sequence similarities to previously reported reference strains in the database and characterized by subtyp-ing each of the patient samples. Boot scannsubtyp-ing analysis by means of SimPlot version 2.5 (Stuart Ray, http://www. med.jhu.edu/deptmed/sray/download) and visual inspection of alignments were used to identify breakpoints. After
identifica-tion of the breakpoints, subregions of the alignment were ana-lyzed by neighbor-joining with bootstrapping to confirm the subtype assignment.
Nucleotide sequences from vpu,gag, and polregions were assembled and aligned with reference sequences belong-ing to subtype A (SE7253 and SE7535), subtype B (MN, WR27, and RL42), subtype C (BW15B03 and BW1626), subtype F (VI850, BR020, and FIN9363) as well as with 4 CRF12_BF (ARMA159, ARMA185, URTR23, and URTR35) and 2 B/F recombinant non-CRF12_BF (ARMA070 and
FIGURE 2.Subtype structure of non-CRF12_BF sequences in three HIV-1 genome regions:gag(a),pol(b) andvpu(c). Subtype F (white) and subtype B (gray).Subtype structure was determined by visual inspection of aligments and confirmed by phylogenetic analysis. Bootstrap value to confirm the subtype assignment of each segment is based on phylogenetic trees (not shown) and placed in the diagram. Subtype structure of the CRF12_BF prototype (ARMA159) is placed at the top of each gene diagram and the relative position of segments to HXB2 reference strain are established by a ruler.
ARMA062) using ClustalX (Thompson, JD et al, 1997) and visually corrected with BioEdit, version 5.0.9 (T. A. Hall, 1999). Phylogenetic trees were constructed by neighbor-joining using the Kimura 2-parameter model with the MEGA version 2.1 program (Kumar, S et al, 2001). Bootstrap analysis was done to assess the stability of the nodes.
RESULTS
The median plasma HIV RNA load in the study popula-tion was 167,000 copies/mL. Serologic analysis using the de-tuned assay demonstrated that none of the patients in this study had evidence of recent seroconversion.
The neighbor-joining tree of all 21gag,pol, andvpu as-sembled sequences with reference strains showed that all the samples were BF recombinants with a dissimilar distribution (Fig. 1). In spite of the low bootstrapping values, different clusters could be seen in the B/F group. One cluster that in-cluded 12 samples more related to the CRF12_BF was identi-fied. The analysis of the detailed subtype structure of the samples was performed by visual inspection of the alignments, and the subtype assignment was confirmed by phylogenetic trees. With this analysis, we were able to observe that those sequences that grouped with the CRF12_BF showed a differ-ent mosaic pattern compared with those that grouped with the non-CRF12_BF sequences (data not shown). Structure analy-sis of non-CRF12–related samples is shown in Figure 2. Each subtype segment had a bootstrap value ranging between 70% and 100%. Most of the different mosaic patterns were found in thegaggene, in which 7 samples (AR13, AR38, AR50, AR57, AR59, AR60, and AR65) had a diverse BF mosaic pattern in comparison with the CRF12_BF prototype, which is subtype F in this analyzed region (see Fig. 2a). The subtype structure in thepol region revealed that 4 samples (AR27, AR38, AR40, and AR57) had a different mosaic pattern compared with the CRF12_BF pro-totype (see Fig. 2b), whereas only 2 samples (AR65 and AR60) had differences in thevpugene structure (see Fig. 2c).
Based on these results, we further analyzed the se-quences in each gene region. For this, a phylogenetic analysis was done as described previously. Only segments with the same subtype structure as the CRF12_BF were included in each tree. The phylogenetic analysis of each sequenced region revealed the presence of sequences related to the CRF12_BF in gag,pol, and vpugenes (Fig. 3). Bootstrap values for each “CRF cluster” were 95% ingag(see Fig. 3a), 70% inpol(see Fig. 3b), and 77% invpu(see Fig. 3c). A total of 12 samples (AR01, AR03, AR19, AR26, AR28, AR36, AR45, AR48, AR53, AR66, AR74, and AR75), representing 57% of the ana-lyzed samples had similar and related sequences to the CRF12_BF in the 3 sequenced genes.
In the IDU population, the HIV-1 seroprevalence was 46%.16The average age was 31.4 years. The subjects, 17 men and 4 women, reported that 99% had used injected cocaine, 80.0% had 1 or more injections per week, 80.0% had shared
needles, 35% had group sex, and 20% had exchanged sex for drugs. Of the 21 IDUs in the study population, 11 answered the question related to therapy. Of these 11 IDUs, 3 were under-going antiretroviral therapy (27.3%). No significant differ-ences were found when analyzing the association between HIV viral characterization (being or not being closely related to the CRF12_BF) and these sociodemographic parameters (data not shown).
DISCUSSION
In Argentina, a total of 25,811 AIDS cases were reported as of 2001,17with 40% of them related to injecting drug use in both genders (14% for women and 26% for men). The com-mon use of injected cocaine in the Southern Cone of South America was described by others.18,19Since 1990, the main HIV transmission route in Argentina has been the use of illegal drug injection, reaching a peak of almost 50% of the new cases reported in 1996. This tendency was reversed during 2001 by the increasing rate of heterosexual transmission, which reached 33%.
Previous molecular epidemiologic studies with samples from Argentina showed that HIV-1 diversity in this country is complex. Considering the multiple HIV-1 recombinant forms described in the literature and the evidence of the different tribution of HIV-1 among vulnerable populations, early dis-semination of these genetic forms in the Argentinean epidemic is apparent. A previous study in the pediatric population re-vealed that transmission of BF recombinants in Argentina has predominated since the 1980s20 and provided evidence that similar BF recombinants (but not identical to the CRF12_ BF) have been spread for at least 15 years in the heterosexual popu-lation in this country.
In our study, a total of 21 plasma samples belonging to HIV-1–positive IDUs were included with the aim of isolating viral RNA and characterizing genomic segments ofgag, pol, andvpugenes by automated sequencing.
We found that 100% of the 21 studied subjects carried BF intersubtype recombinant virus. On the basis of this find-ing, BF recombinants appear to predominate in this popula-tion. Sequencing results from the vpu,pol, andgag regions, however, showed that these recombinants were not all identi-cal but that 12 samples (57%) had the same recombination pat-tern as the CRF12_BF. Our results correlate with the findings of Carr et al10and Thomson et al,9,21which showed high fre-quencies of BF recombinants in South America. In the hetero-sexually infected population studied by Carr et al,10 the CRF12_BF did not represent the predominant genetic form. Our findings show a high prevalence of the CRF12_BF-related sequences in the IDU population, but full-length genome analysis must be performed to be conclusive. Together with the previous studies, our results support the possibility of a common recombinant ancestor, with subsequent recombina-tion giving rise to the suite of BF recombinants now seen in the
population. In this complex scenario, the IDU population could be associated as a spreading source of BF recombinants among the heterosexual population in the region, but the origin of the CRF12_BF is not yet clear. This study suggests the possible association between viral variants and the transmission route.
HIV diversity might have an influence on the efficien-cy of laboratory techniques used for the monitoring of pa-tients22–24as well as on vaccine development. This study ana-lyzes the molecular pattern of HIV-1 in an IDU population from Argentina and places additional emphasis on the impor-tance of knowledge of subtypes in the epidemiologic evalua-tion of the spread of HIV-1 in this country.
REFERENCES
1. Robertson DL, Anderson JP, Bradac JA, et al. HIV-1 nomenclature pro-posal.Science. 2000;288:55–56.
2. Kuiken C, Thakallapalli R, Esklid A, et al. Genetic analysis reveals epi-demiologic patterns in the spread of human immunodeficiency virus.Am J Epidemiol. 2000;152:814–822.
3. Russell KL, Carcamo C, Watts DM, et al. Emerging genetic diversity of HIV-1 in South America.AIDS. 2000;14:1785–1791.
4. Marquina S, Leitner T, Rabinovich RD, et al. Coexistence of subtypes B, F, and as B/F env recombinant of HIV type 1 in Buenos Aires Argentina.
AIDS Res Hum Retroviruses. 1996;12:1651–1654.
5. Masciotra S, Livellara B, Belloso W, et al. Evidence of a high frequency of HIV-1 subtype F infections in a heterosexual population in Buenos Aires, Argentina.AIDS Res Hum Retroviruses. 2000;16:1007–1014. 6. Velarde-Dunois KG, Guimaraes ML, La Fuente C, et al. Molecular
char-acterization of human immunodeficiency virus type 1-infected individu-als from Bolivia reveindividu-als the presence of two distinct genetic subtypes B and F.AIDS Res Hum Retroviruses. 2000;16:1921–1926.
7. Ramos A, Tanuri A, Schechter M, et al. Dual and recombinant infections: an integral part of the HIV-1 epidemic in Brazil.Emerg Infect Dis. 1999; 5:65–74.
8. Fernandez-Medina D, Jansson M, Rabinovich RD, et al. Identification of human immunodeficiency virus type 1 subtypes B and F B/F recombinant and dual infection with these subtypes in Argentina.Scand J Infect Dis. 1999;31:235–242.
9. Thomson MM, Villahermosa ML, Vazquez-de-Parga E, et al. Widespread circulation of a B/F intersubtype recombinant form among HIV-1-infected individuals in Buenos Aires, Argentina.AIDS. 2000;14:897–899. 10. Carr JK, Avila M, Gomez Carrillo M. et al. Diverse BF recombinants have spread widely since the introduction of HIV-1 into South America.AIDS. 2001;15(Suppl):F41–F47.
FIGURE 3.Phylogenetic analysis of three regions of the HIV-1 genomes from Argentina. Neighbor-joining phylogenetic trees were built for each region:gag(a),pol(b) and vpu(c), and significant bootstrap values are placed next to the nodes. The genetic distance corresponding to the length of the branches is shown in the bottom line. Bold lines indicate the clusters defining the CRF12_BF.⽧CRF12_BF◊BF Recombinants non CRFs
11. Avila MM, Pando MA, Carrion G, et al. Two HIV-1 epidemics in Argen-tina: different genetic subtypes associated with different risk groups.
J Acquir Immune Defic Syndr.2002;29:422–426.
12. Liitsola K, Tashkinova I, Laukkanen T, et al. HIV-1 genetic subtype A/B recombinant strain causing an explosive epidemic in injecting drug users in Kaliningrad.AIDS. 1998;12:1907–1919.
13. Piyasirisilp S, McCutchan FE, Carr JK, et al. A recent outbreak of human immunodeficiency virus type 1 infection in southern China was initiated by two highly homogeneous, geographically separated strains, circulating recombinant form AE and a novel BC recombinant.J Virol. 2000;74: 11286–11295.
14. Kato K, Kusagawa S, Motomura K, et al. Closely related HIV-1 CRF01_AE variant among injecting drug users in northern Vietnam: evi-dence of HIV spread across the Vietnam-China border.AIDS Res Hum Retroviruses.2001;17:113–123.
15. Janssen RS, Satten GA, Stramer SL, et al. New testing strategy to detect early HIV-1 infection for use in incidence estimates and for clinical and prevention purposes.JAMA. 1998;280:42–48.
16. Weissenbacher M, Martı´nez Peralta L, Rossi D, et al. Coinfections with HIV, HBV, HCV, HTLV-I, HTLV-II in injection drug users from Buenos Aires, Argentina [session 45, poster 362]. Presented at the First IAS Con-ference on HIV Pathogenesis and Treatment, Buenos Aires, July 2001. 17. National Ministry of Health, Buenos Aires, Argentina. Boletı´n sobre el
SIDA en la Argentina. Buenos Aires: National Ministry of Health; 2002.
18. Libonatti O, Lima E, Peruga A, et al. Role of drug injection in the spread of HIV in Argentina and Brazil.Int J STD AIDS. 1993;4:135–141. 19. The status and trends of the HIV/AIDS epidemics in the world. Barcelona
Map Symposium Report. Presented at the XIV International AIDS Con-ference, Barcelona, July 2002.
20. Gomez Carrillo M, Avila M, Hierholzer J, et al. Mother-to-child HIV type 1 transmission in Argentina: BF recombinants have predominated in in-fected children since the mid-1980s.AIDS Res Hum Retroviruses. 2002; 18:477–483.
21. Thomson MM, Delgado E, Herrero I, et al. Diversity of mosaic struc-tures and common ancestry of human immunodeficiency virus type 1 BF intersubtype recombinant viruses from Argentina revealed by analy-sis of near full-length genome sequences.J Gen Virol. 2002;83:107– 119.
22. Kijak GH, Rubio AE, Quarleri JF, et al. HIV type 1 genetic diversity is a major obstacle for antiretroviral drug resistance hybridization-based as-says.AIDS Res Hum Retroviruses. 2001;17:1415–1421.
23. Kijak GH, Carobene MG, Salomon H. A highly prevalent polymorphism at codon 72 of HIV-1 reverse transcriptase in Argentina prevents hybrid-ization reaction at codon 74 in the LiPA genotyping test.J Virol Methods. 2001;94:87–95.
24. Mercado JM, Di Dio R, Pradal M. Genetic diversity of HIV-1 subtype F from Brazil: failure of HIV-1 viral load testing based on molecular biol-ogy amplification methods.AIDS. 1999;13:2183–2185.
