Epistasis refers to the influence of one genomic mutation or variant on the phenotypic effects of another mutation or variant.  Based on available evidence and theory, this phenomenon has a major influence on evolutionary trajectories for organisms of all sorts.  The role of epistasis has been studied primarily in the context of adaptive evolutionary change.  In a recent paper (2014), Gong and Bloom attempt to determine the relative frequencies of epistatic interactions in adaptive versus stochastic evolution, i.e. evolution driven by selection as opposed to evolution resulting from random processes without a significant selective ‘pressure.’  Gong and Bloom perform this comparison by analyzing homologous nucleoprotein (NP) genes in human and swine influenza A viruses.  The authors argue that the human viruses are subject to substantially more intense selection than the swine viruses since domestic swine are much shorter lived and their viruses are not as likely to be subjected to immune memory responses.

In the introduction, Gong and Bloom describe two primary mechanisms for the fixation of epistatically interacting amino acid substitutions during selection-driven evolution.  First, a mutation that potentially confers a net fitness advantage but is associated with negative phenotypic consequences can be made more beneficial by a subsequent mutation that somehow ameliorates the detrimental effects.  Such subsequent mutations are said to represent a form of epistatic compensation.  Second, a mutation with neutral or negative fitness effects that persists for a period due to stochastic effects can sometimes permit an adaptive mutation that would not have been beneficial in the absence of the prior mutation.  These latter mutations are said to engage in permissive epistatic interactions.

Of importance for the current focus, the influenza virus NP is under strong stabilizing selection in both human and swine hosts since the functionality of NP is essential for packaging and transcription of influenza viral RNA.  In previous work, Gong et al. (2013) showed that several human influenza NP mutations that became fixed during NP evolution despite destabilizing the structure of the protein occurred at sites associated with more than an average number of CTL epitopes and were therefore presumably escape variants. These destabilizing mutations were able to become fixed only in the presence of accompanying stabilizing mutations.

In the present study, the authors compared phylogenetic trees for human and swine NP gene lineages derived, respectively, from human and swine viruses that were derived from a common ancestor that was closely related to the human and swine viruses that caused concurrent pandemics in humans and swine in 1918.  These human and swine NP gene lineages corresponded to viruses that circulated exclusively in, respectively, humans or pigs since 1918.  Analysis of the mutational trajectories revealed many more mutations in the human NP gene lineage (40 amino acid replacements over 44 years) than in the swine NP gene lineage (18 amino acid replacements over 55 years).

The differences in the numbers of mutations in the two lineages were attributed to two main factors: the differential in the intensity of selection mediated by cytotoxic T lymphocytes (CTL) with swine mounting less CTL-based immune selection than humans, and 2) the differential in the intensity of antibody-mediated selection on virion surface proteins (i.e., hemagglutinin and neuraminidase) and associated genetic hitchhiking with, again, swine mounting less immune selection than humans. In this context, hitchhiking refers to the process whereby antibody-mediated selection for a particular hemagglutinin or neuraminidase gene sequences coincidentally leads to increased prevalence of particular NP gene sequences (including particular mutations) that happen to be found in the same viral genomes as the favored hemagglutinin or neuraminidase genes.

Analysis of human and swine NP amino acid sequences revealed comparable numbers of CTL epitopes.  The authors then assessed the relative frequencies of amino acid substitutions in human versus swine NP CTL epitopes.  They found that for the human NP homolog fixed mutations were found in more than the average number of CTL epitopes compared with average amino acid positions.  In the swine NP homolog, fixed mutations were found in more than the average number of CTL epitopes compared with average amino acid positions.  Furthermore, mutations fixed in the human influenza NP gene were associated with more CTL epitopes than were mutations fixed in pig influenza NP.

In the last figure of the study, Gong and Bloom show that 3 of 34 fixed NP mutations in human influenza strains exhibit epistatic interactions while none of 18 fixed NP mutations in swine influenza strains exhibit epistatic interactions.  This difference lacks statistical significance, but additional evidence demonstrates that the epistatically interacting mutations in human NP are all in CTL epitopes and are in more CTL epitopes than are the average NP site or NP sites at which amino acid replacements occurred during the evolutionary trajectory.

Gong and Bloom indicate that their previous studies of influenza NP gene evolution suggest that the permissive epistatic mechanism is more important than the compensatory mechanism in accounting for the evolutionary trajectory this gene.  Nevertheless, both mechanisms are more frequent in the context of evolution based on selection than in the context of evolution based stochastic processes.

From the medical perspective, what makes influenza vaccine development especially challenging is the continuing generation of escape variants that evade either CTL- or antibody-mediated immune mechanisms.  This study offers evidence that this process may be critically dependent on epistatic interactions, at least for the NP gene and gene product.  In other words, many mutations that lead to the inability of CTL to recognize the NP antigen of a given influenza virus with ensuing viral escape would cause overly negative net effects to be maintained in the influenza A genome in the absence of associated permissive or compensatory epistatic mutations.  Similar effects are plausibly involved in the evolution of other influenza proteins, such as the hemagglutinin and the neuraminidase, the main targets of humoral immunity.


Gong LI, Bloom JD. Epistatically interacting substitutions are enriched during adaptive protein evolution. PLoS Genet. 2014 May 8;10(5):e1004328. doi: 10.1371/journal.pgen.1004328. eCollection 2014 May. PubMed PMID: 24811236; PubMed Central PMCID: PMC4014419.

Gong LI, Suchard MA, Bloom JD. Stability-mediated epistasis constrains the evolution of an influenza protein. Elife. 2013 May 14;2:e00631. doi:10.7554/eLife.00631. PubMed PMID: 23682315; PubMed Central PMCID: PMC3654441.