‍ Epistasis, apparent stability and why SARS-CoV-2 may remain evolutionarily metastable

A changing fitness landscape, not a fixed one

‍

A recent study by Luca Sesta and Richard Neher, first posted as a bioRxiv preprint and now published in Genetics, provides an important perspective on how SARS-CoV-2 (SC-2) evolves.(1) The work does not test my immunological hypothesis and does not identify COVID-19 (C-19) vaccination as the source of selective immune pressure. Nevertheless, several of its central findings intersect strongly with the evolutionary framework I have developed in recent Substack articles, particularly the argument that the present SC-2 situation may be better understood as metastable than as a settled endemic equilibrium.

The usual mental picture of viral evolution is too linear: one mutation improves fitness, a subsequent mutation adds another advantage, and the virus continues to optimize through a succession of small, largely independent steps. Sesta and Neher show why that view has shortcomings. The fitness effect of a mutation depends on the genetic background in which it occurs. A change that is deleterious in one variant may be neutral or advantageous in another. Evolution therefore does not proceed across a fixed fitness landscape; as viruses evolve, they continually alter the genetic backgrounds against which the fitness effects of newly arising mutations are expressed.

This already explains why selection today can modify and diversify the set of evolutionary options available tomorrow. A mutation that is effectively inaccessible at one stage because its fitness cost is too high can become accessible after other substitutions have changed the structural and functional context of Spike (S). This is exactly the kind of mechanism through which sustained immune selection can generate delayed, context-dependent and nonlinear ‒ sometimes ‘surprising’ ‒ downstream effects.

1. Epistasis makes evolutionary effects conditional

Sesta and Neher compare mutational fitness effects across major SC-2 clades and show that most effects remain broadly correlated, but important exceptions occur. Their analysis demonstrates that the selective outcome of a mutation cannot always be inferred from the mutation itself. The surrounding genotype can potentiate, suppress or even reverse its effect. Their pairwise epistatic model explains about half of the variance in the observed shifts of mutational fitness effects, and the authors estimate that each mismatch between variant backgrounds substantially alters mutational effects at typically one to three additional positions. (1) A lineage carrying many background-changing substitutions is therefore not simply accumulating independent advantages; it is progressively altering the context in which future mutations will be ‘judged’ by selection. This helps explain why the behavior of newly emerging viral variants cannot reliably be predicted from the identity of their new mutations alone.

This creates evolutionary path dependence. The route already taken influences the routes that remain available. A mutation that is a dead end in one lineage may become a viable stepping stone in another. In other words, the functional impact of a mutation can be strongly ‒ and sometimes decisively ‒ context-dependent.

2. Saltational evolution can reorganize future possibilities

The study also emphasizes that major SC-2 variants are separated by long phylogenetic branches associated with saltational events involving as many as roughly 50 mutations.(1) In an epistatic system, that matters for more than the number of substitutions acquired. A sufficiently large mutational jump can reposition the virus within the fitness landscape, thereby changing the costs and benefits of subsequent mutations.

This provides a mechanistic reason why evolutionary trajectories may appear to change abruptly. A highly mutated lineage, for example, may create a permissive genetic background in which previously unfavorable mutations become tolerable, compensatory changes become accessible or new immune-escape mutations can be acquired at reduced functional cost.

A recent influenza B study independently illustrates this principle. Schwab and colleagues reconstructed eight decades of haemagglutinin evolution and found complex epistatic networks that generated diverse permissive genetic backbones on which immune escape could emerge with limited replicative fitness cost.(2) The pathogen is different, but the evolutionary principle is directly relevant: selection can first create a permissive background that subsequently accommodates functionally relevant immune-escape mutations without substantial fitness cost.

3. Apparent stability can hide continuing evolutionary reconfiguration

An additional implication deserves particular attention. Because the fitness effect of a mutation depends on the genetic background in which it occurs, ongoing diversification can generate a heterogeneous mixture of outcomes: advantageous in one lineage, approximately neutral in another and deleterious in a third. At the level of the total viral population, these opposing effects may partly offset one another, leaving the observer with the impression of a relatively unchanged epidemiological or phenotypic situation.

My point has been that such apparent stability should not automatically be interpreted as evolutionary equilibrium.
It may instead be compatible with metastability: a state in which extensive evolutionary exploration continues beneath an apparently stable population-level phenotype while successive background changes progressively modify which mutational pathways become inaccessible and which become newly selectable.

The important consequence is that stability in the observable phenotype does not imply stability of the underlying evolutionary landscape.

A sufficiently altered genetic background may eventually change the fitness sign of a previously unfavorable mutation or mutational constellation. A mutation that was previously selected against may become neutral or advantageous, whereas one that was previously neutral or advantageous may become deleterious ‒ all depending on genetic context. Epistasis therefore offers a plausible mechanism by which prolonged apparent stability can be followed by a nonlinear evolutionary phase transition.

This possibility becomes even more relevant when many genetically diversified lineages circulate simultaneously. Different branches can explore different portions of the landscape. Most may yield little net gain or terminate in evolutionary dead ends, while one comparatively rare background acquires a combination that changes what becomes selectable next.
As viral transmission continues, successive and increasingly diversified genetic backgrounds can alter the fitness effects of newly emerging immune-escape mutations. Over time, this creates opportunities for a particular background to confer a sufficiently strong fitness advantage to a selected escape mutation ‒ or combination of mutations ‒ to enable a qualitative evolutionary shift. The population can therefore appear globally stable until one branch crosses a threshold that opens a new evolutionary route.

4. Why this matters for my metastability argument

In my recent Substack articles I have argued that recurrent immune escape, saltational events, continued lineage diversification and apparent convergence toward a limited set of functional solutions are more compatible with a metastable host-virus relationship than with a final endemic equilibrium.(3–6)
By metastability I mean an apparently persistent state maintained by a temporary balance of constraints, yet still capable of substantial reorganization if that balance changes.

Sesta and Neher do not use metastability in this sense, nor do they predict the phase transition I have proposed. Nevertheless, what their findings provide is an evolutionary property that a metastable system would require: the adaptive viral landscape is conditional and deformable. Evolutionary constraints do not need to relax smoothly. A lineage can remain confined within a narrow adaptive corridor and then, after sufficient background change, gain access to an immune-escape mutation whose interaction with the evolved genetic background substantially reduces a previously prohibitive fitness cost.

This creates a direct conceptual bridge between epistasis and nonlinearity. Mutations arise stochastically; selection determines which variants propagate. But once epistasis is important, previous rounds of selection change the selective value of mutations that have not yet occurred.

Selection is therefore capable of producing delayed consequences that are not predictable from the immediate phenotypic effect of the substitutions it initially favors.

Hence, the longer viral transmission continues in the absence of effective herd immunity, the more opportunity there is for epistatic interactions to accumulate and reshape the fitness landscape. This may prolong an apparently metastable phase while simultaneously increasing the number of evolutionary opportunities from which a subsequent phase transition could emerge.

5. Immune selection can have downstream effects beyond immediate escape

This is where the paper intersects most strongly with my thinking. Sesta and Neher explicitly note that the assumption of a constant mutational fitness effect can be violated not only by changing genetic background but also by varying selective pressure exerted by the host immune system.(1) The effective fitness of a mutation therefore depends on at least two moving variables: the viral genetic background and the immune environment in which the virus is being selected.

The consequence is important. Immune selection need not directly select a future phenotype in a single step. It may first enrich intermediate backgrounds that alter epistatic relationships, compensate for fitness costs or make other substitutions tolerable.

In that sense, immune selection can change not only which mutations confer an immediate selective advantage, but also which mutations may become strongly advantageous later.

The influenza B work makes this idea tangible. There, epistatic networks generate permissive backbones that allow immune escape while minimizing replicative fitness cost.(2] In other words, selection can build the genetic context required for an antigenic solution that becomes advantageous only later in evolution. This supports the general evolutionary proposition that repeated immune-driven selection can have nonlinear downstream consequences even when the immediate effects of individual mutations appear modest.

6. RBD, NTD and the coupling of antigenicity with viral function

The spatial organization of the SC-2 interaction network is also noteworthy. Sesta and Neher identify extensive epistatic connectivity in S, including sites in both the receptor-binding domain (RBD) and the N-terminal domain (NTD). Several highly connected RBD positions are involved in ACE2 binding and/or immune escape.(1) The broader implication is that antigenicity, receptor usage and structural constraints cannot be treated as fully independent properties.

For my hypothesis, the NTD component is of particular interest. The paper does not validate my proposed mechanisms involving NTD-directed immunity, glycan remodeling, altered dendritic-cell interactions or downstream effects on virulence. But it does support the more general principle that immune-relevant changes in one part of S can alter the fitness consequences of changes elsewhere. That coupling provides a mechanism by which selection acting on antigenic escape could generate effects that become visible only several evolutionary steps later, a possibility that is consistent with a central prediction of my hypothesis.

7. What these studies do ‒ and do not ‒ establish

Sesta and Neher provide strong evidence that SC-2 possesses a dynamic, epistatic and genetic-background-dependent fitness landscape. Schwab and colleagues provide independent evidence in influenza B that epistatic networks can generate permissive backgrounds for immune escape while limiting replicative fitness costs.(1,2) Together, these studies support the general proposition that viral evolution can be path-dependent, conditional and nonlinear and that evolutionary trajectories can continue to change as long as viral transmission and replication persist across the population.

They do not demonstrate that C-19 vaccination caused the observed SC-2 landscape changes. The SC-2 analysis does not stratify sequences by vaccination status, prior infection or time since immune exposure. Its sequence data therefore represent a heterogeneous real-world mixture of immune histories. Accordingly, these studies cannot be cited as proof that vaccine-induced immune pressure is responsible for the evolutionary trajectory I’ve proposed. Their importance is different: they establish the kind of evolutionary architecture through which sustained viral transmission and selection in a changing immune environment could generate delayed and nonlinear effects.

The analysis that now needs to be done

This leads to a testable question. If large-scale C-19 vaccination has materially contributed to shaping SC-2 evolution, that proposition should be examined directly rather than inferred from aggregate sequence data.

A logical next step would be to perform this type of phylogenetic and epistatic analysis separately on viruses sampled from vaccinated and unvaccinated hosts.

Ideally, the analysis should additionally stratify by prior infection, number and type of vaccine doses, time since the last vaccination, time since previous infection, age and, where feasible, immune status. Longitudinal within-host sampling would be especially valuable for distinguishing mutations generated during prolonged infection from those that subsequently acquire population-level fitness.

The central question would be whether particular genetic backgrounds, epistatic couplings or shifts in mutational fitness effects arise or are preferentially propagated under different immune histories.

One could then ask whether vaccine-primed immunity merely changes the frequency of familiar escape substitutions or whether it systematically favors genetic backgrounds that alter which future adaptive pathways become accessible to the virus.

Sesta and Neher show that SC-2 evolution cannot be understood as movement over a static fitness landscape. Schwab and colleagues show in another respiratory virus that epistasis can create permissive backbones for immune escape.

The next question is therefore not simply whether immunity selects viral mutations, but whether different forms of population immunity select different evolutionary backgrounds and whether some of those backgrounds can open qualitatively new evolutionary routes, thereby enabling a genuine phase transition of the virus.

Until such analyses are performed, C-19 vaccination as a principal source of the above-described evolutionary effects remains unproven. But it is a hypothesis that can be subjected to a direct, quantitative test using the very evolutionary tools that now reveal how strongly future viral options can depend on the genetic and immunological context in which selection occurs.
Such analyses could determine whether, and to what extent, the immune environment created by large-scale C-19 vaccination has contributed to shaping those evolutionary trajectories.

References

(1) Sesta L, Neher RA. Epistasis and the changing fitness landscapes of SARS-CoV-2. Genetics. Published 17 September 2026; iyag227. doi:10.1093/genetics/iyag227. Original preprint: bioRxiv 2026.03.12.711354.

(2) Schwab LSU, Xie R, Reilly E, et al. Epistasis facilitates the long-term antigenic evolution of the influenza B virus hemagglutinin. bioRxiv. Posted 22 July 2026. doi:10.64898/2026.07.22.739994.

(3) Vanden Bossche G. The Evolutionary Legacy of Mass Covid-19 Vaccination. Voice for Science and Solidarity, Substack. 1 July 2026.

(4) Vanden Bossche G. The Myth of Endless Fine-Tuning and Why Some Experts Believe the Endemicity Narrative. Voice for Science and Solidarity, Substack. 2 July 2026.

(5) Vanden Bossche G. The Virus Is Not Running Out of Options‒We Are Running Out of Explanations. Voice for Science and Solidarity, Substack. 7 July 2026.

(6) Vanden Bossche G. An Update on the SARS-CoV-2 Pandemic and the Lessons It May Teach Us. Voice for Science and Solidarity, Substack. 14 July 2026.

‍