Evaluating whole HIV-1 genome sequence for estimation of incidence and migration in a rural South African community

Abstract

Background: South Africa has the largest number of people living with HIV (PLWHIV) in the world, with HIV prevalence and transmission patterns varying greatly between provinces. Transmission between regions is still poorly understood, but phylodynamics of HIV-1 evolution can reveal how many infections are attributable to contacts outside a given community. We analysed whole genome HIV-1 genetic sequences to estimate incidence and the proportion of transmissions between communities in Hlabisa, a rural South African community. Methods: We separately analysed HIV-1 for gag, pol, and env genes sampled from 2,503 PLWHIV. We estimated time-scaled phylogenies by maximum likelihood under a molecular clock model. Phylodynamic models were fitted to time-scaled trees to estimate transmission rates, effective number of infections, incidence through time, and the proportion of infections imported to Hlabisa. We also partitioned time-scaled phylogenies with significantly different distributions of coalescent times. Results: Phylodynamic analyses showed similar trends in epidemic growth rates between 1980 and 1990. Model-based estimates of incidence and effective number of infections were consistent across genes. Parameter estimates with gag were generally smaller than those estimated with pol and env. When estimating the proportions of new infections in Hlabisa from immigration or transmission from external sources, our posterior median estimates were 85% (95% credible interval (CI) = 78%-92%) for gag, 62% (CI = 40%-78%) for pol, and 77% (CI = 58%-90%) for env in 2015. Analysis of phylogenetic partitions by gene showed that most close global reference sequences clustered within a single partition. This suggests local evolving epidemics or potential unmeasured heterogeneity in the population. Conclusions: We estimated consistent epidemic dynamic trends for gag, pol and env genes using phylodynamic models. There was a high probability that new infections were not attributable to endogenous transmission within Hlabisa, suggesting high inter-connectedness between communities in rural South Africa.

Keywords: HIV; phylodynamics.

Figure 1.. Epidemic dynamics in the South African data set for pol gene using phydynR.

Left: New infections per year. Right: Effective number of infections per year. Posterior medians and 95% credible intervals are shown with solid line and shaded area, respectively.

Figure 2.. Transmission rates in the South Africa data set for gag, pol and env genes using phydynR.

Boxplot showing the distribution of transmission rates per year.

Figure 3.. Epidemic dynamics in the South Africa data set for gag, pol and env genes using phydynR.

Left: New infections per year. Right: Effective number of infections per year. Posterior medians and 95% credible intervals are shown with solid line and shaded area, respectively.

Figure 4.. Effective population size through time using skygrowth based on time scaled phylogenetic partitions for gag, pol and env genes.

Posterior medians and 95% credible intervals are shown with solid and dotted lines, respectively.

Figure 5.. A down sampled time-scaled phylogeny based on gag sequences from the Hlabisa, South Africa data set.

Colours of branches correspond to partitions found with treestructure. The histogram on the right shows the distribution of clade membership among CGR and South Africa (ZA KZN: Hlabisa, KwaZulu-Natal) sequences. Most CGR sequences appear in a single partition.