Dr. Tom Burr, a statistician from Los Alamos National Laboratory, addressed the question of whether it is genetically plausible that the HIV epidemic could have been generated from a single cross-species transfer. Does the fact that there are so many distinct clades of HIV represent many introductions from a common SIV source, or a single introduction that diversified widely as it spread?
Dr. Burr presented the results of an analysis of phylogenetic trees of HIV and feline immunodeficiency virus (FIV) which demonstrated some of the peculiarities of HIV's evolution. The genetic tree of FIV appears to branch down a central line, but the HIV-1 group M tree radiates from a central node, suggesting a single transfer event. The Los Alamos group used a software package called Treevolve (developed at the University of Oxford) to simulate the genealogy of a sample given a variety of values for rates of partner change, new HIV cases per year, and time to AIDS death, based on UNAIDS data gathered between 1980 and 1998. They found 4 possible patterns in these models, only 1 of which was consistent with their HIV phylogenetic tree. The distribution of subtypes in the phylogenetic tree that was created when they modeled an epidemic spread of HIV from a single time point was the only one that resembled the phylogenetic tree currently considered to be the most accurate representation of HIV evolution. This suggested to Dr. Burr that a point source of the HIV epidemic was plausible, which might support the OPV theory. However, as discussed below, even if multiple introductions of SIV occurred in humans, it is equally plausible that only one lineage might result in an epidemic, with other introductions petering out spontaneously.
Three groups of HIV-1 exist today: group M, group O, and group N. While group M is pandemic, groups O and N are not and remain closely restricted to a single location. HIV-2 consists of 1 clear group with a separate path of descent from the form of SIV found in sooty mangabeys in West Africa. This genetic tree suggests 4 separate transfers to humans, a small number of events given the potentially frequent contact with chimps in Central Africa, and if the dating of chimp to human virus transfer back to the 18th century or even earlier is accurate. This discrepancy is one that requires further investigation and, of course, further refinement of the techniques of molecular epidemiology. However, as noted above, many zoonotic introductions might have occurred with dead-end infections, with only a few introductions leading to established human transmissions.
In the opinion of Daniel Low-Beer, an epidemiologist from the University of Oxford, a star-like geographic pattern of infection would need to have been established early in the epidemic; otherwise, chance extinction events in one village could have eliminated the new virus from the human population quickly. He estimates that at least 60% of SIV transfers were dead-end infections of this sort, which may have flared and died in remote locations.
Sir Robert May of University of Oxford, one of the world's foremost mathematical biologists, suggested that it was plausible for many viral transfers to have remained confined in locations where local custom did not encourage mixing between populations in different villages. By his calculations, if the virus spread equally within a village and outside a village, this would result in the extinction of the virus within a village, and a ripple effect that would result in lower and lower prevalence as the virus traveled from its original "hearth." The virus would establish a foothold in a new village only to die out because of low rates of partner change and would be transferred out to other villages if inhabitants took partners in other places, and so on. This effect would result in an initial peak followed by a slow decline over several decades -- estimated at between 30 and 40 years if the average number of individuals infected by each case was only slightly more than 1 -- before the virus moved into geographically concentrated populations with high rates of sexual partner change, as found in the urban conditions of Kinshasa in the 1970s, for example. Very small changes in sexual activity at the high partner end of the spectrum will have a dramatic effect on the basic reproductive rate and may be difficult to detect by standard epidemiological techniques, such as questionnaires and interviews. This theoretical model might explain the difficulty in finding SIV-infected humans representing transfer events in cross-sectional studies, while also explaining how an epidemic might arise as low-level rural infection, barely exceeding the basic reproductive rate (R0), meets a dense social network in the urban environment.