High levels of discordance between sequencing and serological subtyping in a predominantly non-B subtype HIV-1 infected cohort

Samples from 457 randomly selected HIV-1 infected patients attending King's College Hospital were analysed using a subtype specific enzyme immunoassay. All serotyped non-Bs that provided unambiguous sequence and for which sufficient sample was available (n = 100), which included three serotyped subtype B samples were further analysed by env sequencing and subtyping using neighbour joining phylogenetic analysis, the NCBI Retrovirus Genotyping tool and the Los Alamos BLAST search tool. Of the serotyped viruses, 45% (n = 204) samples were subtype B. Specifically serotyped non-B strains (n = 130) accounted for 28% of the total, of which the largest proportion were subtype C (n = 66). Twenty-seven samples (6%) were classified as non-B, 9% (n = 40) were multiply-reactive and 12% were non-reactive (n = 56). Of the 100 samples subtyped by sequencing the majority were subtype C (n = 32), followed by subtype A (n = 20). There was little concordance between the two methods. Although a 100% match was found among the serotyped and sequenced non-B viruses (n = 13), only 16 of the sequenced subtype C specimens matched the 29 obtained by serotyping. Of the 20 multiply-reactive samples analysed by serotyping, only 1 sample consisted of a subtype mixture by sequencing. Of the 14 serologically non-reactive samples analysed, all were successfully sequenced, with subtype B strains (57%) the most common. Sequencing 15 samples in both env and pol regions revealed differences in subtype assignment for the same sample in some cases. Only 1/6 env subtype A and 4/5 env subtype C samples were concordant in pol sequence subtype. Differences were also found in subtyping by the different methods used. The overall agreement between the three methods was 89%. Four out of 11 samples agreed between the phylogenetic and Los Alamos methods, 1/11 between phylogenetic and BLAST and 2/11 between Los Alamos and BLAST.