LAUSR

HIV-1 subtypes differ, among other things, in their clinical manifestations and the speed in which they spread. In particular, the frequency of subtype C is increasing relative to subtype A and D. We aim to investigate whether HIV-1 subtype A, C and D differ in their per-pathogen virulence and to what extend this can explain the difference in spread between these subtypes. We use data from the Hormonal Contraception and HIV-1 Genital Shedding and Disease Progression among Women with Primary HIV Infection (GS) Study. For each study participant, we determine the set-point viral load value, CD4+ T cell level after primary infection and CD4+ T cell decline. Based on both the CD4+ T cell count after primary infection and CD4+ T cell decline, we estimate the time until AIDS for each individual. We then obtain our newly introduced measure of virulence as the inverse of the estimated time until AIDS. This new measure of virulence has an improved correlation with the set-point viral load compared to the decline of CD4+ T cells alone. After fitting a model to the measured virulence and set-point viral load values, we tested if this relation varies per subtype. We found that subtype C has a significantly higher per-pathogen virulence than subtype A. Based on an evolutionary model, we then hypothesize that differences in the primary length of infection period cause the observed variation in the speed of spread of the subtypes. Author summary HIV-1 subtype C is currently spreading relatively fast in various parts of the world. Data from a study that followed many women infected with different HIV-1 subtypes (A, C and D) before they started treatment shows that neither their viral load nor the disease duration are increased for subtype C compared to subtype A and D. Thus, it seems that subtype C does not have a transmission advantage, neither per contact nor due to longer infection, making the observed relative rise in subtype C a puzzle. We used the same data to test if subtype C has optimized its potential to spread by decreasing the disease duration per unit of viral load (per-pathogen virulence) compared to subtype A and D. However, we find that subtype C has a significantly higher per-pathogen virulence than subtype A and D. This result makes the rise of subtype C even more counter-intuitive. In a last step, we develop an evolutionary model, in which we synthesize all our empirical results. With this model we can show that the most likely explanation for the global spread of subtype C is a difference in the duration of primary infection between the subtypes.