Reduced fecundity is the cost of cheating in RNA virus ϕ6

John J. Dennehy 0 Paul E. Turner () 0 0 Department of Ecology and Evolutionary Biology, Yale University , PO Box 208106, New Haven, CT 06520 , USA Co-infection by multiple viruses affords opportunities for the evolution of cheating strategies to use intracellular resources. Cheating may be costly, however, when viruses infect cells alone. We previously allowed the RNA bacteriophage /6 to evolve for 250 generations in replicated environments allowing coinfection of Pseudomonas phaseolicola bacteria. Derived genotypes showed great capacity to compete during co-infection, but suffered reduced performance in solo infections. Thus, the evolved viruses appear to be cheaters that sacrifice between-host fitness for within-host fitness. It is unknown, however, which stage of the lytic growth cycle is linked to the cost of cheating. Here, we examine the cost through burst assays, where lytic infection can be separated into three discrete phases (analogous to phage life history): dispersal stage, latent period (juvenile stage), and burst (adult stage). We compared growth of a representative cheater and its ancestor in environments where the cost occurs. The cost of cheating was shown to be reduced fecundity, because cheaters feature a significantly smaller burst size (progeny produced per infected cell) when infecting on their own. Interestingly, latent period (average burst time) of the evolved virus was much longer than that of the ancestor, indicating the cost does not follow a life history trade-off between timing of reproduction and lifetime fecundity. Our data suggest that interference competition allows high fitness of derived cheaters in mixed infections, and we discuss preferential encapsidation as one possible mechanism. 1. INTRODUCTION All interactions between individuals that involve the generation, acquisition and use of a common pool of resources are potentially subject to evolutionary cheating (Crespi 2001; Velicer 2003). Cheaters can benefit by using more, but producing less, fitness-enhancing products than their competitors (Travisano & Velicer 2004). However, cheating strategies may be costly for selfish individuals who are vulnerable when victims are not available. For example, in the slime mould Dictyostelium discoideum, where individuals unite to form fruiting bodies composed of somatic stalk and germinative spore cells, cheaters produce only spore and no stalk cells (Buss 1982; Strassmann et al. 2000). In the absence of ordinary genotypes, cheating is costly, because stalkless fruiting bodies hinder spore dispersal from nutrient-poor environments (Buss 1982). In viruses, defective genotypes lacking one or more essential genes can arise spontaneously in laboratory culture, or can co-evolve alongside ordinary viruses in the wild (Roux et al. 1991; Qiu & Scholthof 2001; L opezFerber et al. 2003). These viruses may cheat by replicating at the expense of co-infecting ordinary (helper) viruses that provide the needed proteins. Hence, the selfish behaviour has led certain virus cheaters to be referred to as defectiveinterfering (DI) particles (Cole & Baltimore 1973; Roux et al. 1991). The fitness of defective viruses is zero, however, when infecting the host alone, because they cannot produce all the needed proteins. Full-length (non-defective) viruses can also cheat during co-infection. For example, umbraviruses cheat by usurping proteins provided by co-infecting luteoviruses, in order to be vector-transmitted by aphids between host plants (Falk et al. 1999). Here, we explore the cost of cheating in full-length genotypes of the RNA virus /6. Previously, an experiment allowed populations of bacteriophage (phage) /6 to evolve for 250 generations (50 serial passages) in an environment permitting high multiplicities of infection (MOI; Turner & Chao 1998). MOI is the ratio of infecting viruses to host cells. At elevated MOI, the probability of co-infection (i.e. multiple virus genotypes replicating within the same host cell) is high, producing strong selection for viral genotypes that thrive under such conditions. At the end of the experiment, the fitness of the evolved genotypes was measured relative to the ancestor (stored in the freezer). The results showed that the evolved viruses possessed a large fitness advantage only in the presence of co-infection with unlike genotypes (Turner & Chao 1998). One interpretation is that the derived viruses are cheaters, which possess traits that promote the selfish use of limited intracellular resources. We demonstrated the phenotypic (overall fitness) cost of cheating by showing that the cheaters compete poorly when co-infecting cells with identical cheater genotypes (Turner & Chao 1999). Here, we examine the cost of cheating in phage /6 more closely through classic burst assays that measure components of the lytic infection cycle (Stent 1963). The infection cycle in phage /6 has been well studied (Mindich 1999), and can be partitioned into three phases (figure 1a, Appendix A), comprising the various life-history stages of a virus (Abedon 1989; Bull et al. 2004). Phase I (dispersal/ diffusion) starts with release of phage from the lysed host, and ends when a phage infects a new host. The rate at which phages infect new hosts is termed the attachment rate. Phase II (juvenile) begins with infection and ends when the phage progeny mature inside the host, a time span termed the latent period. Phase III (adult) begins when the virus offspring mature, and ends with cell lysis, when many progeny are released from the infected cell. The total number of progeny released per infected cell is termed the burst size. The growth parameters associated with each phase can be measured easily using standard microbiological techniques (Stent 1963; Chao et al. 2002). Although phage /6 is increasingly used as a model for evolution in segmented RNA viruses (Chao 1994; Turner 2003), growth differences between wild-type and evolved strains of the virus have never been closely analysed. Moreover, examination of juvenile and adult stages of phage life history for the derived cheater virus and its ancestor is useful, because the associated growth parameters (attachment rate, latent period, exponential growth, and burst) yield plausible hypotheses to account for the phenotypic cost that arises when cheaters infect cells alone. In particular, the most compelling hypothesis to explain the fitness differences of a derived cheater and its ancestor might be the virus equivalent of a life-history trade-off between timing of reproduction and lifetime fecundity (Stearns 1992; Bull et al. 2004). Here, selection may favour genotypes that produce relatively few offspring at an early age. Rapid reproducers could gain an advantage in generation number (and hence, offspring number) per unit time, relative to more fecund longer-lived genotypes. This trade-off is, perhaps, more relevant for microbes that often feature expanding populations than for macroscopic org (...truncated)