The genetic subtypes of human immunodeficiency virus type 1 (HIV-1) display differences in immunologie reactivity1 and possibly transmissibility.2 In addition to the high replication rate of HIV-1 and the infidelity of its reverse transcriptase, interstrain recombination may increase the diversity of phenotypes this virus generates. The discoveries that more than 10% of the HIV-1 strains sequenced to date are intersubtype recombinant3 and that the subtype E viruses prevalent in Thailand are probably subtype A/E recombinant4,5 suggest that recombination between viruses belonging to distinct subtypes occurs to an appreciable extent and that the resulting recombinant progeny are transmissible.Cross-immunity between HIV-1 antigens with divergent primary sequences has been a desired goal of preventive and therapeutic vaccine design. Here we report our finding that the Zairian strain used to make an inactivated, gpl20-depleted, therapeutic HIV-1 immunogen6,7 is subtype G/subtype A recombinant. This strain, Z321, was isolated from a previously frozen serum sample taken in 1976 from a 26-year-old Zairian woman who died of an AIDS-like illness in 1978.8 The virus was propagated initially in peripheral blood lymphocytes (PBLs) and later in the transformed T cell line HUT 78 for the purpose of manufacturing the therapeutic immunogen. We amplified (by polymerase chain reaction [PCR]) short segments of the Z321 provirus in DNA extracted from Z321 (passage 6 of vaccine productionj-infected CEM 3 and HUT 78 cells (Fig. 1). Amplified products were subcloned and recombinant DNA templates sequenced as described elsewhere.9 Phylogenetic analysis (PHYLIP 3.5) of individual PCR clones revealed a significant association of gag clones for subtype G ( 11 of 11 ) and env clones for subtype A (12 of 17 by method 1, 11 of 17 by method 2; see Fig. 1) (p < 0.001, Fisher exact test). This result raised the question as to whether distinct subtype G and A viruses or recombinational hybrid viruses existed in our Z321-infected cell cultures. Recombination between subtype A and G target sequences during PCR is unlikely to explain our result since this phenomenon would be expected to result in A/G mosaicism within a minority of the PCR products in any one amplification and would therefore not be expected to skew gag clones to G and env clones to A. Because the sequences of most of the PCR primers used were derived from the subtype B isolate HIV-1sf2> preferential amplification efficiency of our gag primers for subtype G sequences and our env primers for subtype A sequences is also unlikely. This was confirmed by theoretical measures of primability and stability of primer matches for consensus A and consensus G targets (Amplify 1.2 computer program, data not shown). All multiple clones derived from any particular amplicon were for the most part identical, and in a few cases nearly identical. Because of the multiple short segments amplified and sequenced, these considerations make more likely the presence of an intersubtype recombinant virus rather t...
