Genetic Analysis of a UNAIDS HIV Type 1 from Brazil Revealed an Unexpected Recombination Pattern

Editor: The World Health Organization (WHO) Network for HIV Isolation and Characterization (now named UNAIDS) was established by the WHO Global Programme on AIDS to systematically isolate and characterize HIV strains from different parts of the world, and to obtain information and reagents that would facilitate HIV vaccine development.1,2 The laboratory network collected specimens from WHO-sponsored vaccine evaluation sites in Brazil, Rwanda, Thailand, and Uganda. This collection contains a significant number of intersubtype recombinants with diverse recombination patterns.3,4 Among the sequences of HIV-1 collected in Brazil by this network, we identified the sequences of a recombinant strain that have conflicting recombination patterns in bootscanning analysis. The isolate in question is 92BR023, which was isolated from an asymptomatic heterosexual male from Proto Alegre, Brazil.2 The 92BR023 isolate was genotyped previously, and six sequences derived from the viral genome have been reported.2–6 The sequence fragments included two overlapping regions in gag covering nucleotide positions 859–1587 and 1407–2131 (HXB2 numbering),3,5 two regions in pol spanning positions 2265–3440 and 4230–5064,4,6 and two overlapping sequences in env covering 7032–7310 and 7050–7400 (Fig. 1).7 Based on the genotyping data, it was shown that this isolate is a B/C intersubtype recombinant with a subtype C gag, a subtype C pol, and a subtype B env. When we analyzed these sequences of 92BR023 obtained from the Los Alamos HIV Sequence Database by bootscanning, however, a more complex recombination pattern was observed and one of the sequences appeared to generate conflicting results. FIG. 1. Bootscanning of the 92BR023 viral genome reveals numerous recombination breakpoints. Five GenBank entries associated with the 92BR023 HIV-1 (accession numbers were shown on top of each analysis) were compared with a subtype A ({"type":"entrez-nucleotide","attrs":{"text":"AY521631","term_id":"46946843","term_text":"AY521631"}} ... In the bootscanning analysis (SimPlot version 3.5.1),8 we used a subtype B and a subtype C HIV-1 from Brazil as references, and a subtype A and a subtype K virus from other regions as the outgroup.9 Comparing the two sequences within the gag of 92BR023, the fragment covering the matrix (MA) and capsid (CA) genes belong to subtype C, whereas the part consisting of CA and nucleocapsid (NC) is subtype B (Fig. 1). This is not consistent with previous reports indicating that 92BR023 has a subtype C gag. It is possible that a recombination breakpoint is present in the gag gene, but out of the 181-nt overlapping sequences of the two fragments, there were 12 mismatches. The genetic distance of 0.066 (12/181) is similar to the genetic distance between subtypes B and C gag, suggesting that there might be errors in the HIV-1 sequences. We then analyzed the pol and env of 92BR023 and identified additional recombination breakpoints. Two recombination breakpoints were identified in the bootscanning analysis of the regions covering protease (PR) and reverse transcriptase (RT) genes (Fig. 1). These breakpoints resulted in a short subtype B sequence within the subtype C pol. One potential breakpoint was observed in the integrase (IN) gene. In addition, two recombination breakpoints were identified in the short sequence covering C2–C3 of gp120. None of these recombination breakpoints in pol and env has been reported previously. The results indicate that 92BR023 appears to be a more complex B/C recombinant than the one originally suggested.3,4 Here, we determined the sequence of the full-length genome of 92BR023 to map out the recombination patterns of the virus and to confirm the accuracy of the GenBank entries associated with the isolate. We isolated RNA directly from the virus stock obtained from the AIDS Reagent Program and converted the RNA to cDNA using SuperScript III reverse transcriptase (Invitrogen). The cDNA was amplified using the FastStart High Fidelity PCR System (Roche) in four overlapping fragments covering the full-length genome of 92BR023. The PCR products were sequenced with overlapping primers, and the resulting sequence contigs were assembled with the Staden Package (PCR and sequencing primer sequences are available on request).10 Every nucleotide was identified by at least two sequence contigs to ensure the accuracy of the DNA sequence. The assembled viral sequences were aligned with the reference and outgroup sequences using Clustal X (version 1.8.3).11 Bootscanning was carried out for the 92BR023 sequences with the same references and outgroups described above. We found that the 92BR023 is indeed a B/C recombinant, but the recombination pattern is quite complex (Fig. 1). Notably, the gag of the recombinant is subtype C and does not seem to have recombined with another subtype. This finding ruled out the possibility that a recombination breakpoint is present in gag and illustrated that the GenBank entry, {"type":"entrez-nucleotide","attrs":{"text":"AY090758","term_id":"22654346","term_text":"AY090758"}}AY090758, does not match the sequences of the 92BR023. We compared {"type":"entrez-nucleotide","attrs":{"text":"AY090758","term_id":"22654346","term_text":"AY090758"}}AY090758 with the same region from the full-length 92BR023 sequence and found that the sequences shared only 89.5% similarity, thus confirming that the GenBank entry is incorrect (Table 1). Table 1. Nucleotide Similarity Between 92BR023 and GenBank Entries In the pol, we identified two short subtype B regions within the subtype C sequence (Fig. 1). The recombination breakpoints in the RT genes are consistent with the bootscanning result from the GenBank entry {"type":"entrez-nucleotide","attrs":{"text":"AF009376","term_id":"2338303","term_text":"AF009376"}}AF009376. We also demonstrated the presence of a recombination breakpoint at the start of the IN gene, which confirms the bootscanning analysis of {"type":"entrez-nucleotide","attrs":{"text":"AF458237","term_id":"22750454","term_text":"AF458237"}}AF458237. Two separate studies indicated that 92BR023 has a subtype B env.2,5 Both studies were based on the results from genotyping the short V3 region and phylogenetic analysis using the C2–V3 or C2–C3 region of the env. Our analysis showed that 92BR023 has a subtype B V3, but it also has several recombination breakpoints in the env (Fig. 1). Neither study identified the observed recombination in the C2–C3 region in the phylogenetic analyses, probably because phylogenetic inference is a not a sensitive method to detect recombination, especially when the region of interest is relatively short (300–400 bp). Finally, we verified the accuracy of the GenBank entries associated with 92BR023 by comparing the entries with the corresponding regions of the full-length sequence. Except for {"type":"entrez-nucleotide","attrs":{"text":"AY090758","term_id":"22654346","term_text":"AY090758"}}AY090758, which was described previously, all five sequences shared a high degree of similarity with the full-length sequence of 92BR023 (Table 1), with one to five nucleotide mismatches probably generated from PCR and sequencing errors or due to the fact that our sequence was obtained from virus passage different from those in previous studies. Altogether, we presented the full-length sequence of 92BR023 and mapped the recombination pattern of the virus. We also showed that one of the GenBank entries associated with 92BR023 does not match the sequence of the virus. The full-length sequence of 92BR023 highlights the complexity of HIV-1 B/C intersubtype recombinants in Brazil and will help to elucidate the biological and antigenic properties of HIV-1 B/C recombinants.