K. E. Atkins, T. Antal, R. N. Thompson, K. Lythgoe, R. R. Regoes, S. Hue, C. J. Villabona-Arenas
Treatment-as-prevention is the cornerstone of global HIV control, and the 'Undetectable = Untransmittable' (U=U) message rests on the empirical observation that antiretroviral therapy reduces onward transmission to negligible levels. Yet a mechanistic explanation for why viral suppression prevents transmission, and for the broader quantitative features of HIV transmission, has been lacking. Here we develop a mechanistic framework that resolves three long-standing features simultaneously: the low per-act transmission probability, the narrowing of within-host viral diversity to one or a handful of founder variants, and the plateau of transmission risk at high viral loads. We show that three biologically grounded mechanisms -- short windows of susceptibility in the exposed partner, stage-dependent establishment of systemic infection, and target-cell limitation at the site of infection -- are jointly necessary and sufficient to reconcile these observations. Calibrated to six epidemiological datasets and independently validated against three more, the model derives a transmission rate below 0.05 systemic infections per 100 couple-years follow-up under viral suppression, providing the mechanistic basis for the negligible risk observed in PARTNER1 and the U=U message that rests on it. Recalibration to data from men who have sex with men attributes their elevated transmission rates to longer windows of susceptibility, consistent with the biology of the rectal mucosa, rather than to differences in per-act biology -- reframing how population-level risk differences should be interpreted in prevention planning. The model further predicts four to five transient infections for each systemic infection, consistent with evidence from the STEP vaccine trial and independently supported by HIV-specific immune responses in highly exposed seronegative individuals and transient local viral RNA in tissues from non-human primate challenge studies. Embedding the model in a phylodynamic framework substantially narrows estimates of when transmission occurred and how many viral variants established infection. Together, these findings provide a mechanistic foundation for treatment-as-prevention, reframe the determinants of population-level transmission differences, and point to vaccine strategies aimed at preventing systemic establishment after mucosal exposure.