Recombination confounds the early evolutionary history of human immunodeficiency virus type 1: subtype G is a circulating recombinant form

Abstract

Human immunodeficiency virus type 1 (HIV-1) is classified in nine subtypes (A to D, F, G, H, J, and K), a number of subsubtypes, and several circulating recombinant forms (CRFs). Due to the high level of genetic diversity within HIV-1 and to its worldwide distribution, this classification system is widely used in fields as diverse as vaccine development, evolution, epidemiology, viral fitness, and drug resistance. Here, we demonstrate how the high recombination rates of HIV-1 may confound the study of its evolutionary history and classification. Our data show that subtype G, currently classified as a pure subtype, has in fact a recombinant history, having evolved following recombination between subtypes A and J and a putative subtype G parent. In addition, we find no evidence for recombination within one of the lineages currently classified as a CRF, CRF02_AG. Our analysis indicates that CRF02_AG was the parent of the recombinant subtype G, rather than the two having the opposite evolutionary relationship, as is currently proposed. Our results imply that the current classification of HIV-1 subtypes and CRFs is an artifact of sampling history, rather than reflecting the evolutionary history of the virus. We suggest a reanalysis of all pure subtypes and CRFs in order to better understand how high rates of recombination have influenced HIV-1 evolutionary history.

FIG. 1.

Schematic putative phylogenetic trees of our data set and its classification regarding the monophyly rules defined in Materials and Methods. Rule 1, monophyly of CRF02_AG plus G; rule 2, monophyly of CRF02_AG separately; rule 3, monophyly of subtype G separately. If our hypothesis is confirmed, our output trees should show the pattern of panel b.

FIG. 2.

Recombination analysis of subtype G strains compared to all other pure subtype strains. (a) Similarity (top), bootscanning (middle), and sliding Bayes (bottom) analysis done as described in Materials and Methods, with the gene regions indicated on top and the recombination breakpoints as determined by informative site analysis. (b) ML tree of the genome region between bp 4316 and 5162 as indicated in panel a. (c) ML tree of the genome region between bp 5577 and 6083 as indicated in panel a. The genomic regions illustrated in the tree are indicated in the upper panel. ML trees were generated with PAUP v4b10, as described in Materials and Methods. •, midpoint root of the tree; *, zero branch length test with P < 0.001 and NJ bootstrap support of >70; #, zero branch length test with P < 0.001 but NJ bootstrap support of <70.

FIG. 3.

Recombination analysis of CRF02_AG strains. Similarity (top), bootscanning (middle), and sliding Bayes (bottom) analysis done as described in Materials and Methods, using as subtype reference sequences all pure subtypes including subtype G (a) and all pure subtypes excluding subtype G (b). The recombinant structure as defined in the Los Alamos database is shown on top. The region indicated corresponds to the nonrecombinant region analyzed in the final Bayesian tree (Fig. 4). LTR, long terminal repeat.

FIG. 4.

Recombination analysis of subtype G strains compared to all other pure subtype strains and CRF02_AG (considering CRF02_AG as a putative pure subtype representative of subtype A). (a) Similarity (top), bootscanning (middle), and sliding Bayes (bottom) analysis done as described in Materials and Methods and at the top the proposed recombinant structure. (b) ML tree of the merged genome regions bp 1500 to 2325, 3275 to 5475, and 7275 to 7975 as indicated in panel a. ML trees were generated with PAUP v4b10, as described in Materials and Methods. The phylogenetic tree of the J region is shown in Fig. 1c. •, midpoint root of the tree; *, zero branch length test, P < 0.001 and NJ bootstrap support of >70.

FIG. 5.

Phylogenetic analysis to discriminate the parent from the recombinant in the genome region bp 3500 to 4000. The Bayesian tree shown was one of the trees generated by MrBayes in one of two independent MCMC runs. The support of the clustering of CRF02_AG and subtype G was analyzed using the “monophyly rules” described in Materials and Methods. The paraphyletic clade (here CRF02_AG) can be considered the parent, and the monophyletic clade (here subtype G) can be considered the recombinant.