Survival of the flattest
We have studied how RNA viruses can thrive at high mutation rates, and how they adapt to high mutation pressure. On the basis of computer simulations, we have predicted that strains evolving under high mutation pressure trade replication speed for mutational robustness. Strains thus evolved can outcompete strains with much higher replication speed (or fecundity) but lower robustness to mutations. We have termed this effect survival of the flattest. Several groups have since observed this effect in laboratory populations of RNA-based pathogens, e.g. Codoner et al., PLoS Pathogens 2:e136, 2006; Sanjuan et al., PLoS Genetics 3:e93, 2007.
The broader importance of this work lies in its implications for lethal mutagenesis. According to prevailing thought, RNA viruses operate at their maximally tolerable mutation rate. If this were true, a slight increase in the mutation rate would be sufficient to kill any RNA virus under all circumstances. Instead, according to the survival-of-the-flattest effect, there are likely many cases in which a slight increase in mutation rate will lead to the evolution of a more robust virus, and more severe mutagenesis is necessary to achieve viral extinction.
Selection against protein misfolding
We have argued that mistranslation of seemingly intact genes leads to misfolded, toxic protein species which can cause disease, in particular neurodegenerative disease. Consequently, it is necessary to reassess what kind of mutations are disease-causing. A mutation that makes a gene more susceptible to translation errors but otherwise leaves the correctly translated protein intact may look innocuous but nevertheless cause a severe disease phenotype. Our work possibly explains the origin of certain misfolding diseases, for which much of the disease-associated protein folds properly and for which either many different or no causative genetic mutations have been found.
Evidence for our hypothesis comes from extensive evolutionary analysis of E. coli, yeast, worm, fly, mouse, and human. We observe (i) that the pattern of covariation between rates of evolutionary divergence (dN, dS), codon bias, transition--transversion bias, and expression level is conserved from bacteria to humans, (ii) that there is selection for translational accuracy in all six species, and (iii) that in fly, mouse, and humans, the covariation is strongest in neuronal tissues, even for genes that are not specific to these tissues. These observations implicate mistranslation-induced protein misfolding as a major cost on the cellular metabolism, and suggest that this cost can lead to neurodegenerative diseases in animals. A large-scale simulation of protein evolution, mistranslation, and misfolding establishes that this cost is sufficient to generate the observed patterns of evolutionary divergence.
Theory of lethal mutagenesis
Lethal mutagenesis, the killing of a microbial pathogen with a chemical mutagen, is a potential broad-spectrum antiviral treatment currently under investigation by numerous empirical groups. Yet a theoretical framework to describe lethal mutagenesis is largely nonexistent. Specifically, we don't know what mutagen dosages are required to achieve viral extinction, and what treatment options are likely successful or potentially counterproductive. Moreover, for many years there was confusion in the literature over the relationship between lethal mutagenesis on the one hand and the error catastrophe as proposed by Eigen on the other hand.
We have started developing a comprehensive theory of lethal mutagenesis. We have demonstrated that (i) the extinction threshold, i.e., the mutation rate sufficient to cause viral extinction, depends on both genetic and demographic factors, such as overall virus clearance by the host; (ii) there is no universal extinction threshold, the threshold can vary dramatically even for the same virus strain grown under somewhat different conditions, (iii) because of (ii), extinction can happen at mutation rates much lower or much higher than the one of the error catastrophe, and (iv) if an error catastrophe occurs, it will in general hinder rather than promote lethal mutagenesis.
Stage effect of anti-HIV drugs
The efficacy of drugs against HIV-1 is commonly assessed by the speed with which the viral load in patients declines after initiation of the drug regimen. By this measure, a new class of antiretroviral drugs, integrase inhibitors, are among the most efficient drugs developed so far. We have demonstrated by mathematical modeling that using the rate of the viral-load decline as a measure of drug efficacy can be misleading, because that rate is determined primarily by the stage in the viral life cycle at which the drug acts. If we compare two drugs that act by different mechanisms but have identical efficacies, for example, two drugs that both are capable of complete suppression of viral replication, the viral load decline will be more rapid for the drug acting later in the viral life cycle. The rate of decline is primarily determined by the abundance of cells that contain virus past the stage at which the drug acts. The later the drug acts, the smaller that abundance, and hence the faster the viral load decay. Therefore, we argue that the rate of the viral load decay is a valid measure of drug efficacy only when comparing drugs that act at the same stage of the viral life cycle.