Evolutionary reversion of live viral vaccines: Can genetic engineering subdue it?

Abstract

Attenuated, live viral vaccines have been extraordinarily successful in protecting against many diseases. The main drawbacks in their development and use have been reliance on an unpredictable method of attenuation and the potential for evolutionary reversion to high virulence. Methods of genetic engineering now provide many safer alternatives to live vaccines, so if live vaccines are to compete with these alternatives in the future, they must either have superior immunogenicity or they must be able to overcome these former disadvantages. Several live vaccine designs that were historically inaccessible are now feasible because of advances in genome synthesis. Some of those methods are addressed here, with an emphasis on whether they enable predictable levels of attenuation and whether they are stable against evolutionary reversion. These new designs overcome many of the former drawbacks and position live vaccines to be competitive with alternatives. Not only do new methods appear to retard evolutionary reversion enough to prevent vaccine-derived epidemics, but it may even be possible to permanently attenuate live vaccines that are transmissible but cannot evolve to higher virulence under prolonged adaptation.

Keywords: R0; synthetic biology; virus.

Figure 1.

Hypothetical, visual model for evolution of an attenuated virus to exceed the epidemic threshold. (A) For each individual vaccinated, there is a distribution of average transmission chain lengths that depends on the number of secondary contacts infected and how many contacts they infect. Thus, the vaccinated individual may transmit to (a) no one (chain length 1), (b) to one secondary contact but they fail to transmit (length 2), (c) to two secondary contacts, one of which does not transmit further (length 2) and one of which transmits to a single contact who does not transmit (length 3), and so on for all possibilities. The average chain length increases with the average R₀ of the virus. (B and C) For the viruses that have attained a particular chain length, there will have been some evolution occurring during that chain, resulting in a distribution of R₀ values increasing with chain length. If the chain is long enough, there may have been enough evolution that some viruses now have $R_0>1$, and if they get into the next patient, an outbreak may occur (depending on other factors). The net probability of evolving a virus that exceeds the epidemic threshold combines the probability of each chain length with the probability that viruses in that patient have $R_0>1$ and escape to the next patient. The net probability of escape will be vanishingly small if chain lengths are short and the virus either never evolves $R_0>1$ or is slow to do so.

Figure 2.

A visual model of permanent attenuation. Fitness during continued evolution of the engineered virus (blue curve) reaches a plateau below that of the wild-type virus (dashed).

Figure 3.

Fitness declines approximately linearly with the number of suboptimal codons in phage T7. All changes were engineered in the viral major capsid gene. Fitness is growth rate, population doublings per hour, a geometric measure of fitness. The leftmost point represents wild type. Used with permission from Bull, Molineux, and Wilke (2012).

Figure 4.

Evolution of attenuating genome rearrangements in phage T7. Blue curves give fitness (growth rate measured as population doublings/hour) across the generations of adaptation; the horizontal dashed line gives the approximate upper limit of fitness as observed in wild-type T7 adapted to the growth conditions. (Left) Two replicates of a genome with an RNAP gene displaced approximately half the genome length. In the replicate that evolved the higher fitness, an early rearrangement returned the RNAP gene to its wild-type location but displaced other genes. Subsequent evolution was gradual and may have reached an asymptote. (Right) The RNAP gene was displaced about 15% of the genome length, so the initial fitness was higher than in the left graph. However, no apparent fitness improvement occurred during adaptation. Both panels are from Cecchini et al. (2013), the left amended with data from Springman et al. (2005). Adapted with permission from Cecchini et al. (2013)