Abstract. The existence of organ-specific HIV-1 populations within infected hosts has been studied for many years; nonetheless results reported by different authors are somewhat discrepant. To tackle this problem, we used a population genetics approach to analyze previously published data from the V3 hypervariable region of the envelope env gene. Our results are compatible with a population subdivision by organs in 95% of individuals analyzed at autopsy. In addition, populations infecting the nervous system and testicles clearly appear as differentiated subsets of the socalled macrophage-tropic variants. Liver and kidney may harbor differentiated populations as well. Although it is widely accepted that organ compartmentalization arises as a consequence of different selective pressures imposed by different organs, a definitive demonstration has not yet been provided. Our analysis of the pattern of synonymous and nonsynonymous nucleotide substitutions provides evidence supporting this hypothesis, without discarding the role of other evolutionary processes. In contrast, positive selection does not seem to be the mechanism responsible for the evolution of patient-specific sequences.Key words. Adaptation, gene flow, metapopulations, organ-driven selection, viral dynamics, viral evolution.Received October 6, 2003. Accepted January 19, 2004 It has been well established that human immunodeficiency virus type 1 (HIV-1) infections begin with a homogeneous population (Karlsson et al. 1998;Zhang et al. 2002), but during the initial acute stage of the disease diversity rapidly rises (McNearney et al. 1992;Pang et al. 1992;Poss et al. 1995) and keeps growing during the nonsymptomatic stage of the disease (Keys et al. 1993). Various organs are already infected at this early stage (Zhang et al. 2002), and assessing how viruses from these organs evolve should help to understand AIDS progression and HIV-1 pathogenesis. A possible scenario is that viruses from different organs could diverge and evolve separately.As an example of this scenario, HIV-1 populations infecting blood and semen show different biological properties (Coombs et al. 1998) and are sometimes compartmentalized (Zhu et al. 1996;Delwart et al. 1998;Curran and Ball 2002). Similarly, variants are unevenly distributed in cervical swabs and blood (Poss et al. 1995), but are rarely restricted to one of them (Sankale et al. 1998;Shaheen et al. 1999). Viruses from blood and lung have also been shown to evolve separately (Itescu et al. 1994), though this seems not to be the norm for all patients studied (Collins et al. 2002). An alternative scenario is that all body sites harbor a common viral pool of which blood-circulating samples are representative. Supporting this possibility, several authors found no compelling evidence for organ-structure in lung, lymphatic nodes, testicles, spleen, kidney or even bone marrow (Ball et al. 1994;van't Wout et al. 1998), between colon and blood (van der Hoek et al. 1998), or between thymus and lymphatic nodes (Alves et al. 2002). Virus...