A total of 36 samples were selected from adult patients residing in São Paulo who had previously been diagnosed in two independent unpublished studies in our laboratory as being infected with HIV-1 F1 based on pol subgenomic fragment sequencing. The first study investigated the long-term therapeutic interruption and genetic survey of HIV-1 variants in 137 patient samples collected in 2002 and identified 8 HIV-1 F1 (5.8%) in their group. The second study assessed the genotypic resistance and virus distributions in a cohort of 751 HIV-1 infected patients on antiretroviral therapy between 2006–2007 and found a total of 28 patients (3.7%) infected with sub-subtype F1. All of the 36 samples initially characterized by both projects were submitted for NFLG.

Genomic DNA was extracted from peripheral blood mononuclear cells (PBMCs) using the QIAamp blood kit (Qiagen GmbH, Hilden, Germany) according to the manufacturer's instructions. Extracted gnomic DNA concentration of the samples under investigation were determined by comparing the samples band relative to the band of the comparable intensity in the high DNA mass ladder (Invitrogen, Brazil) using 0.5–1.0% agarose gel. All extracted genomic DNA were submitted for amplification of a house keeping gene (β-globin) with primers PC03 and PCO4 34 to examine DNA integrity and exclude the presence of PCR inhibitors. Proviral DNA was amplified from the purified genomic DNA by PCR using primers and methods described in previous studies 2930. Amplification reactions were done in duplicate to eliminate PCR artifacts, ensuring that sequenced NFLG were not assembled from heterogeneous DNA targets. Both DNA complementary strands were sequenced directly from purified PCR products by using variety of internal primers, fluorescent-dye terminators, and Taq polymerase on an automated sequencer (ABI 3130, Applied Biosystems Inc., Foster City, CA).

The data from each sequenced fragment were edited and initially screened for the presence of recombination patterns by the REGA HIV-1 subtyping tool (v2) 35 and the jumping profile Hidden Markov Model (jpHMM) 36 and further confirmed by using the bootscanning method 37 implemented in SimPlot 3.5.1 for Windows 38 using the following parameters, window size 250 bp, step size 20 bp and the Kimura 2-parameter as a model to estimate nucleotide substitution. The significant threshold for the bootscan was set at 90%. This strategy has allowed us to early identify samples with any recombinant fragments and to exclude them from further experiments and analyses. Only edited subclade F1 fragments were assembled into contiguous sequences on a minimum overlap of 30 bp with a 99–100% minimal mismatch and a consensus sequence was formed by the Sequencher program (Gene Code Corp., Ann Arbor, MI). NFLG consensus sequences were again analyzed for recombination by using the jpHMM web tool. Only jpHMM confirmed pure sub- subtype F1 isolates were aligned with reference sequences representing subtypes A-D, F-H, J and K obtained from the Los Alamos database http://hiv-web.lanl.gov using the CLUSTAL X program 39. Aligned sequences were further refined by manual editing and trimmed to the minimal shared length in the BioEdit Sequence Alignment Editor Program 40. Gaps and ambiguous positions were removed from alignment. The phylogenetic trees were constructed by the maximum likelihood (ML) methods implemented in the program PHYML 41 using the GTR + I + G substitution model and a BIONJ starting tree. Heuristic tree searches under the ML optimality criterion were performed using the NNI branch-swapping algorithm. The approximate likelihood-ratio test (aLRT) based on a Shimodaira-Hasegawa-like procedure was used as a statistical test to calculate branch support. Only aLRT statistical values of >70% were considered significant and displayed at the tree nodes. Trees were plotted using the program MEGA version 4 http://www.megasoftware.net. The mean genetic nucleotide distances within and between sequences were calculated using the maximum composite likelihood model implemented in MEGA version 4.0 software. Comparison of tree topology from subclade F1 subgenomic regions was performed by the recently described algorithm 42. Prediction of HIV-1 coreceptor usage from the env V3 region was determined using the web-based service geno2pheno [coreceptor] http://www.geno2pheno.org which considers all V3 mutational patterns, and not only changes of arginine or lysine at positions 11 or 25.

GenBank accession numbers for the proviral NFLG sequences reported in this study are [02BR082: FJ771006, 02BR170: FJ771007, 06BR564: FJ771008, 06BR579: FJ771009, 07BR844: FJ771010].

Of the 888 subjects in total, 36 were initially assigned as infected with non-recombinant HIV-1 subclade F1 variants using approximately 1200 bp region in the pol gene as depicted in figure 1. Thus, based on the analysis of this small portion of viral genome, subclade F1 is still present in São Paulo and roughly accounts for 4.05% of circulating strains.

To find the true prevalence of this strain, all the 36 samples were subjected to complete genome amplification and only genuine HIV-1 F1 strains were corroborated by further phylogenetic analysis of the complete coding sequences and part of LTR region. Of the 36 patients investigated, chimeric viruses comprised of B/F1 and F1/C based on fragment analysis were detected in 31 patients and were omitted from further analysis. Only 5 sequences inferred from pol region as subclade F1 also holds true for the genome as a whole and, thus, estimated the true prevalence of HIV FI at 0.56%. Inspection of coding regions of the five HIV-1 F1 sequences obtained in this study displayed open and intact reading frames for majority of HIV proteins. To exclude laboratory strains contamination, a BLAST search of GenBank HIV-1 sequences did not reveal any evidence for contamination with strains obtained from our patients. Figure 2 shows the ML tree of 53 complete genome coding sequences, including the five isolates sequenced in this study and 48 reference strains (GenBank and Los Alamos database) representing subtypes A-D, F-H, J and K. The five isolates, namely 02BR082, 02BR170, 06BR564, 06BR579 and 07BR844, fell into subclade F1 reference group (100% aLRT), with an overall average distance of 6.9%. Close inspection of figure 2 shows that strains clustered in subclade F1 are divided further into two major groups, in which the Brazilian F1 sequences, FI.93.FIN9363 and BE.93.VI850 fell into the first group and F1.FR.96.MP411, F1.ES.x.P1146, F1.ES.x.X1093 2, F1.ES.x.X1670 and F1.x.x.MVP 30846 fell into the other. Within the first group that includes the Brazilian F1 sequences, FI.93.FIN9363, BE.93.VI850 and the Argentinean F1.AR.2002.ARE933 isolate, where the node within this cluster had 99% of aLRT statistical support, there is an indication of two major clusters of the Brazilian sequences separated by 90% of aLRT values. To test the stability and branching orders of our sequences, ML trees were independently made from all the full length HIV-1 F1 gag-pol (n = 21) and env sequences (n = 27) available in the database (Figure 3a &3b). The phylogenetic trees from both regions received an overall topological score of 69.3%. Consistent with the results obtained by NFLG, all the Brazilian isolates along with other South American (SAm) sequences formed a distinct cluster supported by aLRT value of greater than 80% in both regions as depicted in figure 3. Furthermore, close looking at the SAm sequences in figure 3a and 3b revealed two main clusters separated from each other by 84% and 99% of aLRT values in the gag-pol and env regions, respectively, and indicated possible intrasubtype recombination. For example, isolate 06BR579 placed the gag-pol region within the 07BR844 and F1.AR.2002.ARE933 cluster in 72% of aLRT values, but the env region grouped within other cluster that included F1.BR.1989.BZ163 and F1.BR.1989.BZ126 (aLRT 94%). The computed topological score of this cluster in both regions was 50% with branch length mismatch of 59.7%. Therefore, the shifting of topological positions into 2 different phylogenetic trees is suggestive evidence of intrasubtype recombination event or other factors such as convergence.

To establish the relationship of our sequences to each other and to the previously published HIV-1 F1 sequences from a variety of geographic regions, another phylogenetic tree was constructed using the C2-C3 env. This region was selected because it contains a larger number of HIV-1 F1 sequences in the database. As shown in figure 4, except for isolate 02BR034 (Accession DQ358812), the sequences from SAm including the new isolates in this study emerge from a common node along with two strains, one isolated in Spain from an Argentinean immigrant (BF1.ES.2002.X1241) and the other from Portuguese women (F1.PT.-.envNTM44 89) with unknown epidemiological link to any SAm country. The Brazilian F1 isolates did not form a rigid cluster but were dispersed among the SAm non-Brazilian sequences and were distinct from the Romanian and African isolates. In contrast, the Brazilian strain 02BR034, which branches well out of the SAm cluster, is positioned in the phylogenetic tree close to other isolates from the DRC. Moreover, the branching patterns indicate that the common ancestor of the SAm cluster was also shared with isolates belonging to DRC.

The mean genetic nucleotide distances of the SAm HIV-1 F1 sequences derived from the C2-C3 region were computed and compared with reference strains from Romania and Central Africa (Angola and DRC). The results showed that the SAm isolates differed from the Romanian and Central African F1 strains by nucleotide distances of 12.8% and 12.3%, respectively. The mean nucleotide variation of 8.5%, 11.2% and 12.3% was observed among F1 strains from SAm, Romania and Central Africa, respectively.

Detailed inspection of the amino acid alignment of the obtained isolates and the Brazilian subtype F1 reference genomes showed that strains 07BR844 and 01BR087 (Accession DQ358801) had 18 bp insertional mutation in the N-terminus of the p6^Gag protein that created a duplication of PTAPP motifs. The same sequences showed concomitant insertion of 6–7 amino acids within the p^6Pol epitope (NSPTRREL) with particularly conserved repeat of SPT amino acids. Analysis of the phenotypic characteristics of the 3^rd (V3) region in the env gene of the new Brazilian subtype F1 isolates described in this study suggests that all, except the isolate 07BR844, were derived from macrophage tropic R5 viruses. The GPG motif at the tip of the V3 loop and the potential N-linked glycosylation sites were highly conserved.

A detailed scrutinization of the partial nucleotide alignment of the 3' LTR regions relative to HXB2 and consensus sequences of other HIV subtypes is shown in figure 5. Conform to the consensus sequence GGGRNNYYCC, two potential NF-κB binding sites were localized in our five new strains. A subclade F1 specific insert of 13–15 nucleotides was observed just downstream of the NF-κB^III binding site. The existence of this insertional nucleotides signature has also been reported in a previous study 43