Characterisation of the Ugandan HIV epidemic with full-length genome sequence data from 1986 to 2016

Abstract

Presented here are two large Ugandan HIV genome datasets, one from the modern period (by the MRC/UVRI & LSHTM group and PANGEA), and another generated from stored 1986 serum samples using target-capture next generation sequencing. Uganda uniquely has two HIV subtypes at similar proportions, and although subtype A1 is older than subtype D, both subtypes are well established in all cohorts and risk groups, which has facilitated comparison of the two. Previously, HIV sequence data from East Africa has typically been short gene sequences, thus the full-length genome data here has presented an opportunity to compare the evolutionary histories of the two subtypes across the genome, and carry out an examination of inter-subtype recombination patterns. The majority of the modern HIV genomes are unique recombinant forms (URFs), representing a large number of independent superinfection events, which is consistent with the size and age of the epidemic. There are wide scale patterns of recombination along the genome, which are described. Specifically, the region of envelope from C2 of gp120 to the transmembrane region of gp41 is almost always found intact since disruption of these protein interactions is expected to be highly detrimental. Re-discovered serum samples from 1986 yielded 109 full-length HIV ’historical’ genomes. The subtype distribution is shown to significantly change over time: subtype D fell from 67% in 1986 to 17% in the modern PANGEA sample. Furthermore, co-receptor tropism (CXCR4 or CCR5) was predicted with geno2pheno and a significant difference between the historical subtypes was observed: 63% of subtype D genomes are X4 tropic (known to be associated with faster progression to AIDS) whilst 0% of A1 sequences are X4 tropic. Therefore, co-receptor tropism may have reduced the effective reproductive number of subtype D by reducing the duration of potential onward exposure (due to faster time to death) compared with A1, and can explain a drop in subtype D prevalence over time. Finally, BEAST1 methods are applied to reconstruct the demographic histories of the two subtypes over time using gag, pol, and env gene data, and place the subtypes in their wider East African context. These findings characterise a highly diverse and complex epidemic in Uganda that has shifted from predominantly sub-type D to predominantly subtype A1 between 1986 and 2016, whilst pervasive and ongoing recombination has generated a wide variety of URFs.