Migration coupled with recombination explains disparate HIV-1 anatomical compartmentalization signals

The evolution of HIV-1 in a host is shaped by many evolutionary forces, including recombination of virus genomes and the potential isolation of viruses into different tissues with compartmentalized evolution. Recombination and compartmentalization have opposite effects on viral diversification, with the former causing global mixing and the latter countering it through spatial segregation of the virus population. Therefore, recombination and compartmentalization together give rise to complex evolutionary dynamics that convolutes their individual effects in standard, bifurcating phylogenetic trees. Although there are various theoretical methods available to infer the presence of recombination or compartmentalization individually, there is little knowledge of their combined effect. To study their interaction and whether that could explain the many disparate results that have been described in the HIV-1 literature, we developed an age-structured forward-time evolutionary model that includes compartments, migration, and recombination. By tracking the evolutionary history of individual virus variants in an Ancestral Recombination Graph (ARG) and resolving the ARG into standard bifurcating trees, we reexamined 771 anatomical tissue pairs infected with HIV-1. Remarkably, we found that recombination can make the resulting bifurcating tree appear more compartmentalized than the virus population actually is. However, we found migration between all 771 tissue pairs, typically at 2.8-4.9x10-3 taxa-1 day-1. Thus, while different point mutations may arise in different parts of the body, migration eventually brings these variants together and recombination merges them into a relatively homogeneous cloud of a universally evolving quasispecies. Modelling this process, we explain the many different results previous research found among distinct anatomical tissues. We also show that the popular Slatkin-Maddison test comes to different results about compartmentalization at the same migration rate depending on the sample size. SIGNIFICANCEWhether HIV-1 sampled in different anatomical tissues are compartmentalized or not has been a long-standing quest because it relates to both clinical and biological insight. Previous studies of this question have reported disparate results from the same as well as different tissue comparisons. Here, we consolidate and explain these contrasting results with a new evolutionary model that includes compartments, migration, and recombination. We show that no anatomical tissues are fully compartmentalized, and that the migration of HIV-1 between the tissues allows recombination to homogenize the diversity that may arise in separate tissues of an infected person.