Viral Evolution

Viruses evolve faster than any other biological entity, and the consequences run through the whole of clinical virology: immune escape, vaccine mismatch, antiviral resistance, and the host switching that drives emergence. Two features set viral evolution apart. The first is an extraordinarily high mutation rate, greatest in the RNA viruses, whose polymerases make about one error per genome copied. The second is population size: a single infected person may carry more viral particles than there are people on Earth, so even rare variants are generated many times over each day. Variation on this scale, sorted by selection and chance, is why a virus is rarely a fixed target. This article covers where that variation comes from and how it is shaped; the molecular machinery of replication that underlies it is treated in the Foundational Virology topic’s Virus Replication article.

Sources of genetic variation

All variation begins with mutation, the ultimate source of genetic novelty. For RNA viruses it is the defining evolutionary feature, because the RNA-dependent RNA polymerase lacks the proofreading and repair machinery that guards cellular deoxyribonucleic acid (DNA) replication. The rates differ by orders of magnitude across the genome types, and they sit in a consistent order: retroviruses and RNA viruses mutate fastest, DNA viruses slowest, with single-stranded DNA viruses intermediate.

Genome type	Replication fidelity	Pace of evolution
RNA viruses	Low: the polymerase has no proofreading	Fast: about one change per thousand nucleotides each year
Retroviruses (reverse transcriptase)	Low: no proofreading	Among the fastest of all
Single-stranded DNA viruses	Intermediate	Intermediate
Double-stranded DNA viruses	High: proofreading and repair	Slow: on the order of 100,000 times slower than RNA viruses

For a typical RNA virus this works out at roughly one mutation in every genome each time it is copied, so a large infected population contains essentially every single-point variant at all times. A useful distinction sits behind these figures. The rate at which errors are made, before selection has acted, is not the same as the rate at which changes actually become fixed in the population over time, which already reflects selection and is the figure read off sequence data. The order of the genome types is consistent on both measures: the fast-mutating RNA and retroviruses evolve so quickly that sequences sampled weeks apart already differ, while the proofreading double-stranded DNA viruses barely change over decades.

The second source of variation is recombination, the joining of genetic material from two viruses that have co-infected the same cell. In RNA viruses it usually happens by copy-choice: the polymerase switches templates mid-synthesis and produces a hybrid genome. How often this occurs is largely set by genome structure. It is frequent in retroviruses, which package two RNA molecules per virion, and in some positive-sense RNA viruses such as the coronaviruses, where up to a quarter of the progeny of a co-infected cell can be recombinant. It is rare in the negative-sense RNA viruses, whose genomes are bound tightly into ribonucleoprotein and cannot easily switch templates. Recombination matters clinically because it can combine immune-escape, resistance and host-range changes in a single step, and can create genuinely new viruses.

The third source, reassortment, is recombination’s special case for segmented genomes: when two strains co-infect a cell, progeny can package a mixture of whole segments from each parent. It is very frequent in influenza A virus, and it is the molecular basis of the abrupt antigenic change discussed below.

Mutation rate, genome size and the error threshold

Mutation rate and genome size are inversely related across all of biology, and viruses occupy the fast-mutating, small-genome end of that line. The relationship is not incidental. A higher mutation rate generates more variation, which speeds adaptation, but it also produces more lethal and deleterious mutations, and beyond a point the genome can no longer be copied faithfully enough to preserve itself. RNA viruses appear to sit close to the maximum mutation rate they can tolerate, which in turn caps how large their genomes can be. The constraint is captured by Eigen’s paradox: a higher-fidelity polymerase would need to be larger and more complex, but the longer genome encoding it would accumulate too many errors to establish in the first place, a chicken-and-egg limit on viral complexity. The coronaviruses are the notable exception, having evolved a proofreading exoribonuclease that lowers their error rate and lets them carry the largest known RNA genomes, a point of more than academic interest given how many emerging viruses are coronaviruses.

This ceiling can be exploited therapeutically. Lethal mutagenesis pushes a virus over its error threshold by raising the mutation rate with a mutagenic nucleoside until fit genomes can no longer be regenerated and the population collapses; the classic experimental agents are 5-fluorouracil and ribavirin, and the principle underlies modern mutagenic antivirals used against RNA viruses. The mechanism and the agents are treated in the Antivirals topic.

The quasispecies concept

Because RNA viruses replicate with so many errors, an infected host does not carry a single genome but a swarm of related variants, often called a mutant cloud. The quasispecies concept, developed by Manfred Eigen, makes a stronger and more specific claim than simply noting this diversity. It holds that the variants are generated from one another so rapidly that they are not independent, so natural selection acts on the whole population rather than on any single fittest variant, and the population evolves to maximise its average fitness. One striking prediction is “survival of the flattest”: a cluster of variants with modest but similar fitness can outcompete a population built around a single high-fitness variant surrounded by poor neighbours, a form of mutational robustness.

How far real viral populations behave as true quasispecies is debated, and the concept should not be applied to large DNA viruses, whose error rates are far too low. Its lasting clinical value is the picture it gives: the diverse population already contains rare drug-resistant and immune-escape variants before the drug or the antibody is ever encountered, so resistance and escape can be selected almost immediately rather than waiting for a new mutation to appear. This is why monotherapy fails against HIV and why a single antibody rarely holds a fast-mutating virus for long.

Selection, drift and transmission bottlenecks

Variation is raw material; what survives is decided by selection and chance. Two forms of natural selection dominate. Purifying (negative) selection removes deleterious mutations and is the strongest force acting on RNA and single-stranded DNA viruses, whose compact, multifunctional genomes leave little room for error; most random mutations are harmful, with mutagenesis of vesicular stomatitis virus giving roughly 40% lethal, 29% deleterious, 27% neutral and only 4% beneficial. Positive (diversifying) selection fixes the minority of advantageous changes: immune escape, drug resistance and host adaptation. The balance between the two is measured by the ratio of nonsynonymous to synonymous substitutions (dN/dS), which falls below one under purifying selection and rises above one where positive selection drives change, for instance in the antibody-targeted surface proteins.

Chance acts through genetic drift, whose strength depends on the effective population size. Within a host the population is enormous, so selection is efficient, but transmission imposes a severe bottleneck, and for HIV and influenza a new infection is often founded by a single virus particle. Such bottlenecks slash diversity and let mildly deleterious variants rise by luck rather than fitness, and over successive transmissions the irreversible accumulation of such mutations in small populations (Muller’s ratchet) would degrade fitness were it not offset by the large populations reached within each new host. The width of the bottleneck tracks the route: vertical transmission tends to pass a wider sample of the population than sexual transmission.

Rates of evolution, the molecular clock and phylogenetics

Because viruses accumulate substitutions at a measurable and roughly constant rate, they carry a molecular clock. RNA viruses change at around one substitution per thousand sites per year (~10⁻³), some five orders of magnitude faster than large DNA viruses, which makes them “measurably evolving populations”: sequences sampled only weeks or months apart already differ. Tip dating, which uses the known sampling date of each sequence, turns this into a powerful tool, and reading viral phylogenies underpins modern molecular epidemiology.

A phylogenetic tree records shared ancestry, and two patterns are especially useful. Incongruent trees on either side of a breakpoint reveal recombination, because different parts of the genome have different histories. And comparing a virus tree with its host tree separates two modes of long-term evolution: congruence indicates codivergence, the virus and host speciating together over long periods (as the herpesviruses have with vertebrates), while incongruence signals cross-species transmission, the host jumping that drives emergence. In outbreaks, the same methods date the most recent common ancestor, reconstruct how a virus has spread, and track whether drug-resistance or escape variants are establishing. The detailed application of these tools to outbreak response is covered in the Outbreaks, Surveillance and Pandemic Preparedness article.

Antigenic variation: drift and shift

The clinical face of viral evolution is antigenic change, the way a virus alters its surface so that prior immunity no longer recognises it. Influenza A is the classic example and shows two distinct mechanisms, which map directly onto the sources of variation above.

Feature	Antigenic drift	Antigenic shift
Mechanism	Point mutation under positive selection	Reassortment of genome segments
Genetic change	Small, incremental	Large, abrupt
Result	Gradual escape from existing immunity	A new subtype the population has not seen
Consequence	Seasonal epidemics; annual vaccine update	Potential pandemic

Drift is the steady accumulation of point mutations in the surface proteins, selected because they evade circulating antibody, and it is why the seasonal influenza vaccine must be reformulated each year. Shift is the sudden acquisition of a new surface protein by reassortment when two influenza strains co-infect one host, producing a subtype against which the population has little immunity and which can therefore cause a pandemic. The deeper machinery of antigenic variation, and the influenza-specific detail, belong to the Classification and Nomenclature of Viruses material and the influenza profile; the point here is that drift is mutation plus selection, and shift is reassortment.

Evolution as the engine of emergence

Evolution is the thread that connects this article to the rest of the topic. Cross-species transmission accounts for most emerging viral infections, and RNA viruses host-jump far more readily than DNA viruses precisely because their fast evolution lets them adapt to a new host before the host can respond. Most jumps are not successful: the great majority of spillovers are transient infections that fail to transmit onward and die out, because the virus must clear several evolutionary and ecological barriers to establish. Success is more likely between closely related host species, which share similar cell receptors and habitat, so receptor compatibility is the central molecular barrier to host range. The same properties that make a virus evolve quickly under drug or immune pressure, high mutation rate, large populations and short generation times, are what make it a candidate to emerge. The drivers that bring virus and new host together are the subject of the Drivers of Emergence, Spillover and One Health article.

References and recommended reading

• Geoghegan JL, Holmes EC. Virus Evolution. In: Fields Virology, 7th edition (Fundamentals), Chapter 1. Wolters Kluwer; 2023. The current comprehensive reference for the processes of virus evolution on which this article principally draws.
• Burrell CJ, Howard CR, Murphy FA. Virus Replication and Mutation, Reassortment and Recombination. In: Fenner and White’s Medical Virology, 5th edition. Academic Press / Elsevier; 2017. A concise foundational account of the sources of viral variation.