Franklin Roosevelt's granddaughter was one of 556 children in Washington, DC, bedridden with measles during one week in 1934. Before the introduction of a vaccine in the 1960s, explosive outbreaks occurred in large cities every two years. Influenza, by contrast, strikes every winter in a wave of infection that circles the globe before dying out. What makes patterns of spread in these two RNA viruses so different?

Bryan Grenfell of Penn State University's Center for Infectious Disease Dyanmics (CIDD) says the next step in understanding disease emergence and spread is to link epidemiology with pathogen evolution in an approach he terms phylodynamics.¹ Epidemic patterns are largely determined by the time course of an infection, whether it's measured in days, weeks, or years. Phylogenetic patterns result from an interaction between natural selection mediated by the immune system (and/or drug treatments) and random epidemiological processes. Here's how some common pandemics play out.

Measles is a highly contagious infection lasting an average of 12 days. Infection (or vaccination) confers lifelong immunity to all strains of the virus, although the reason for this is unclear. The observed biannual epidemic cycles occur as the number of susceptible hosts are repeatedly exhausted and replenished with a new generation of schoolchildren. Nevertheless, the battle between the virus and host never escalates because natural selection operates equally on all strains. No single strain of measles predominates, and around the globe a number of strains are able to coexist.

Influenza A also has a short infection period but differs from measles in that there is no cross-immunity between different strains of the virus. Each year random mutation produces a large genetic diversity that doesn't necessarily change the face of the virus to the immune system. Suddenly, a mutation results in a large shift in the shape of the surface molecules, and the virus escapes immune recognition. Natural selection whittles away strains with so-called neutral mutations, and the end result is punctuated evolution and lowstanding genetic diversity.² This is what makes the flu-vaccine business such a guessing game.

Dengue virus has evolved into four clusters, or serotypes, that coexist in a stable equilibrium. Rather than competing (as in influenza) or blending into a homogenous mass (as in measles), these serotypes have established a truce with each other. Humans can be infected with several serotypes at any given time, a condition which leads to the more severe forms of dengue: dengue hemorrhagic fever and dengue shock syndrome. In fact, the presence of one serotype may reinforce the others. Natural selection may preserve a certain level of antigenic dissimilarity so that the virus can outsmart cross-protective antibodies.

HIV provides an opportunity for researchers to look at both between-host and within-host evolution. Because the immune system exerts strong selective pressure on virus populations within the host, the observed phylogenies - like that of influenza A - exhibit lowstanding diversity in the body at any given time. Between-host evolution, on the other hand, is relatively slow because of the long interval between transmission events. HIV strains do not differ in their transmissibility, and the evolutionary history reflects demographic effects rather than natural selection. Consequently, the between-host phylogeny resembles that of measles.

What about microbial diseases? The short answer is that their genomes are larger, and extensive horizontal gene transfer makes a phylogenetic approach questionable, says evolutionary geneticist Daniel Falush at Oxford. Stephan Schuster at CIDD is tackling this issue by combining a phylogenetic approach - which depends on slowly evolving markers such as 16s - with a genome-wide approach that does not depend on cloning of sequences. Consequently, he's able to get an unbiased sample of the population gene pool. "It has been a two-year struggle to convince the field that we now need statistical tools to look at the data," he says.