PART II
POPULATION AND EVOLUTIONARY DYNAMICS OF VIRUSES AND THEIR MICROBIAL HOSTS
CHAPTER THREE
Population Dynamics of Viruses and Microbes
3.1 ON MEASUREMENTS AND MODELS
In the mid-late 1970s, Bruce Levin and colleagues at Emory University embarked on a program to examine microbial and viral population dynamics. Their approach: to combine strains of bacteria and viruses in a flask, chemostat, or other vessel and to record the change in number of each population with time. Example experimental variables included resource input, strain type, and/or turnover time.
Their program appears straightforward in retrospect, but it is important to recognize three key innovative aspects: (i) the decision to measure the dynamics of the populations by repeated sampling over time; (ii) the integration of mathematical models with experimental work; (iii) the choice to study multiple interacting populations, rather than a single population. The seminal publications in the early stage of this program were those by Levin et al. (1977) and Chao et al. (1977). In contrast, microbiological studies at that time tended to measure the relative change in a population from the start to the finish of an experiment. Even today, many studies focus on a single population arising from a clonal isolate without the use of any mathematical models or theory whatsoever. The approaches could not have been more distinct.
The approach of biologists Bruce Levin and Lin Chao, in collaboration with mathematician Frank Stewart, built on general principles in ecological theory and specific models for population dynamics of viruses and hosts. The first modern model of virus-host population dynamics in which the hosts were themselves microorganisms was developed by Alan Campbell (1961). Campbellās model, like the work of Levin and colleagues, sought to describe the change in population densities arising from interactions among virulent phage and their bacterial hosts. The basic mechanisms, as laid out in Chapter 2, included the reproduction of host cells, the infection and lysis of host cells by viruses, and the decay of viruses. Campbell, like Levin and colleagues, incorporated the use of explicit delays between infection and lysis. Indeed, the system was more specific than many of the āgeneralā models available at the time for the study of predator-prey interactions. The specificity helped reveal not only how populations might interact in principle but also how they interact in practice.
The Levin āschoolā continues to work at the interface of experiments and models. The approach was not only embraced at Emory but shared by many, particularly those involved in a new Gordon Research Conference in Microbial Population Biology that held its first meeting in 1985, with Bruce Levin as chair. Rather than recapitulating the entire history from the late 1970s to the present, I will illustrate through a series of models some of the principles introduced by Levin and colleagues, and subsequent authors, and will describe a few new twists as well. What Levin and colleagues found in those early papers, elaborated on by many since, is that viruses can rapidly decrease host populationsādriving them well below densities found in the absence of virusesāand that host and viral populations can oscillate. They also found that population dynamics does not tell the whole story, but Iām getting ahead of myselfāone story at a time. For now, Iāll discuss just population dynamics, and in the next chapter, Iāll continue with the rest of this story: what happens when populations evolve.
3.2 VIRUSES AND THE āCONTROLā OF MICROBIAL POPULATIONS
3.2.1 ECOLOGY AND CONTROL
How do interactions among organisms and between organisms and nutrients unfold in a complex environment? The study of ecology is preoccupied with this question. Unsurprisingly, there is no single answer, even if certain mechanisms do recur: top-down and bottom-up control. The difference between these two mechanisms can be illustrated with an example. Consider a laboratory flask containing a population of microorganisms and growth medium and from which predators and other pathogens of this microbial population have been excluded. One would expect the density of microbes to be limited by available nutrients in the medium. This is termed bottom-up control. Similarly, consider the scenario in which the resource is replete with nutrients, and a predator of the microbe is added. In this case, the density of microbes might be limited by the predators rather than by the available nutrients. This is termed top-down control. These paradigms provide useful limit cases for interpreting potential outcomes, even if the reality is that many systems exhibit responses that lie on the continuum between these two limits (Carpenter et al. 1985; Hunter and Price 1992; Pace et al. 1999).
How, then, do viruses affect microbes and microbial communities? The obvious answer is that they negatively affect individual microbial cells and, by extension, decrease the total number of microbes in an environment. Figure 3.1 includes the results of multiple experimental manipulations of host organisms with and without viruses. In each case, the density of hosts drops markedly with the addition of viruses. It would seem that viruses, like predators, have the potential to control a microbial population from the ātop.ā Evaluating this potential requires consideration of the drivers, that is, the conditions that can lead to virus drawdown of host populations. Moreover, early dynamics do not always tell the whole story. The next few sections will introduce models that provide a conceptual and mathematical framework for understanding how viral infections change the dynamics of microbial populations and nutrientsāpotentially switching control from bottom up to top down.
3.2.2 MODELS OF VIRUS-HOST INTERACTIONS WITH EXPLICIT RESOURCES
The original experiments of Levin and colleagues (Levin et al. 1977; Chao et al. 1977) examined the effect of viruses on microbial population dynamics and concomitant changes in viral population dynamics. These interactions took place in a chemostat, a continuous-flow reactor in which fresh media with resource (nutrient) concentration J0 is continually introduced at a rate Ļ (Figure 3.2). To maintain a constant volume, an equivalent amount of material is removed from the chemostat at the same rate. This material is a subsample of the chemostat contents and so can contain a mixture of resources, hosts, and viruses. The chemostat is usually stirred or otherwise mixed and can include sampling ports for evaluating the current state of the contents or for adding new populations or other media.
Inside the chemostat vessel, there are direct interactions between viruses (V) and hosts (N), direct interactions between hosts and the growth media (R), and likely many indirect interactions as a result of feedback processes. How much of the detail of what is occurring inside the chemostat vessel should be considered in the model? This is a recurring question that arises when biologists and theorists gather, and matters turn to a research problem of common interest. Sometimes, the biologist is the one asking, usually when the theorist has planned to omit some seemingly crucial element from the model. Sometimes, the theorist is the one asking, usually when the biologist has insisted on a more ārealisticā representation. The tension is real, and there is no one correct answer. Here, the guiding credo is to āmake everything as simple as possible, but not simplerā.1 The choice of a chemostat helps in one significant respect. Every entity in the chemostat is assumed to sample the same average environment. As a consequence, the representation of dynamics here begins by considering the change in the average concentrat...