Pathogen evolution poses a challenge to public health interventions, and therefore has been extensively studied theoretically and empirically (for example, ref. 1). The efficacy of vaccines may be compromised owing to escape variants that are able to overcome vaccine-induced immunity and cause recurrent epidemics. Pathogen virulence (usually modelled as pathogen-induced death) can also evolve over time. However, in most theoretical models, the evolution of pathogen virulence and antigenic escape have been considered in isolation. The ongoing COVID-19 pandemic demonstrates the importance of being able to predict both adaptations together2. Writing in Nature Ecology & Evolution, McLeod and Gandon3 fill this gap in our understanding of evolution of multiple phenotypic traits with a theoretical model that allows for simultaneous adaptation of the virulence of a pathogen and its ability to escape vaccine-induced immunity.

McLeod and Gandon3 study how vaccines with various protective effects and life-history trade-offs interact with changes in the population of naive and vaccinated hosts to drive the evolution of virulence and vaccine escape. Previous studies of the long-term effects of vaccination on virulence evolution have assumed that the system has settled at an endemic equilibrium before a new variant could emerge and substitute the original strain4. This eco-evolutionary feedback process creates a time-scale separation that simplifies the transition from population genetics to ecological dynamics, but which only rarely holds in the real world. McLeod and Gandon3 elucidate transient contemporary evolution of virulence and immune escape by combining epidemiological modelling with a population genetics approach, which allowed them to track strain and allele frequencies as well as linkage disequilibrium (the nonrandom association of alleles at two or more loci5) between alleles for the two traits.

The virulence–transmission trade-off is one of the cornerstones of pathogen evolutionary theory, and states that the transmission and virulence of a pathogen are not independent, but positively correlated6. It has also been supported by empirical evidence7. In an SIR (susceptible, infectious and/or recovered) model typical of a virus infecting the human upper respiratory tract, McLeod and Gandon3 considered the virulence–transmission trade-off in the context of three main protective effects of the vaccine. These were (1) reducing transmission either by blocking infection or reducing infectiousness; (2) reducing virulence; and (3) reducing infection duration by increasing recovery rate. Previous models predicted that the virulence-reducing component of the vaccine promotes selection for higher virulence, because it removes the cost of increased mortality without affecting the benefit of increased transmission through the virulence–transmission life-history trade-off4,8. However, McLeod and Gandon3 show that if the virulence of a pathogen is positively correlated with transmission rate via a trade-off, then vaccines that reduce virulence mortality generate negative epistasis between (higher) virulence and escape alleles (Fig. 1). This results selection for strains that carry either mutation, but not both. As the authors point out, such a competition between vaccine escape and virulence could have important implications for vaccine development — particularly of cross-protective vaccines, which would be efficient against a broad range of pathogen strains. As the evolution of virulence would most probably be influenced by the type of vaccine protection (because vaccines targeting within-host pathogen growth are predicted to lead to increased levels of virulence), investigating the effects of cross-protective vaccines in the context of contemporary evolution of both virulence and immune escape is of the utmost importance.

Fig. 1: Evolutionary outcome depends on the type of life-history trade-off and protection conferred by a vaccine.
figure 1

This figure graphically represents data from table 1 of ref. 5. a, When virulence is positively linked with transmission, vaccines that reduce pathogen virulence support the evolution of either vaccine escape or increased virulence, but not both in combination. By contrast, vaccines that block infection, or reduce transmission or duration of infection, support the evolution of both traits together. b, When virulence is negatively related to recovery, the outcome of selection by vaccines that reduce virulence is again either for vaccine escape or for higher virulence, but not both. Both traits are selected for when the vaccine reduces transmission. Vaccines that increase recovery (reduce the duration of infection; not included in this figure) can support the mutually exclusive evolution of vaccine escape or higher virulence, or the joint evolution of both traits, depending on the population structure.

Full size image

By contrast, when vaccines reduce transmission or infection duration there is positive epistasis between virulence and escape alleles, so strains that carry both mutations are favoured by evolution (the least desirable public health outcome). Interestingly, the authors also show that, under frequent recombination, sequential fixation of traits occurs; a vaccine-escape allele reaches a quasi-fixated state followed by fixation of a (higher) virulence allele. Therefore, if vaccines could be updated regularly to re-establish high population immunity, the pathogen would keep adapting to vaccine escape rather than evolving higher virulence. Unfortunately, timely updates of existing conventional vaccines remain a challenge: for instance, the vaccine against seasonal influenza virus takes at least six months to be produced in large enough quantities to be distributed among the public. New technologies in vaccine manufacturing could overcome these issues, as has recently been demonstrated by successful deployment of mRNA vaccines against COVID-19 (refs. 9,10).

The authors also looked at other life-history trade-offs that affect pathogen evolution. A decrease in recovery rate with increasing virulence has been observed in myxoma virus in Australian rabbits11 and Mycoplasma gallisepticum infections in North American house finches12. Under this virulence–recovery trade-off, hosts are expected to recover more quickly from less-virulent strains than from highly virulent ones13. McLeod and Gandon3 confronted their model with the virulence–recovery trade-off (Fig. 1) and found that evolutionary outcomes were aligned with those predicted by the virulence–transmission trade-off for all types of vaccines, except for two cases: vaccines that reduce transmission (which did not generate epistasis) and vaccines that increase recovery (which produced either positive or negative epistasis, depending upon population structure and interactions between recovery and the recovery-increasing component of the vaccine).

Predicting the outcomes of evolution at multiple sites in the pathogen genome is challenging and will most probably be affected by additional aspects of population structure, such as waning immunity or different social contact and vaccination rates in different age groups. In the context of the ongoing pandemic of COVID-19, a worrisome question is whether a highly virulent or transmissible strain capable of vaccine escape can emerge. A recent study2 has shown that repeated epidemic waves caused by immune escape can drive selection for higher virulence in pathogens. Whether introducing vaccines targeting different aspects of the pathogen life cycle could exacerbate the negative evolutionary effects is an important question, and the multilocus approach presents a different, and perhaps more realistic, perspective from which to investigate it. The COVID-19 pandemic has also triggered an enormous effort to predict the course of SARS-CoV-2 evolution and antigenic escape from variant analysis of sequence data14. Leveraging such phylodynamic analysis together with epidemiological dynamics may offer an empirical test of McLeod and Gandon’s theoretical conclusions about immune escape and virulence.