Genetic Variation as the Raw Material of Evolution

Genetic variation refers to differences in DNA sequence among individuals within a population — differences that can, in principle, be transmitted from parent to offspring. Not all variation is relevant to evolution: a suntan or a weightlifter's muscles are phenotypic differences but are not heritable in the genetic sense, so they do not count as raw material for evolutionary change. What matters is variation at the level of the genotype — differences in alleles, the alternative versions of a gene or DNA sequence at a given locus — because only genotypic differences reliably reappear in offspring. Three technical terms recur throughout population genetics. A locus is a specific position on a chromosome. An allele is one of the alternative sequences that can occupy a locus. A population is polymorphic at a locus if more than one allele exists there at an appreciable frequency (conventionally, the most common allele has a frequency below 0.99, so that genetic drift alone is unlikely to explain the polymorphism by chance). Genetic variation is typically summarized using two complementary statistics: percent polymorphic loci (the fraction of surveyed loci with more than one allele in the population) and heterozygosity (the average probability that two alleles at a locus, drawn at random from the population, differ from each other). Both statistics can be computed from allozyme electrophoresis, microsatellite genotyping, or, most powerfully today, whole-genome resequencing. Critically, the origin of variation is decoupled from its evolutionary fate. Mutation and recombination generate variants without any reference to whether those variants will be helpful, harmful, or neutral in the current environment. This is the principle of undirected mutation: an allele that happens to confer resistance to a pesticide does not arise because the pesticide is present; it arises through ordinary replication error, and only afterward — if the pesticide happens to be applied — does that pre-existing variant get favored by selection. This sequence (variation first, selection second) is what separates Darwinian evolution from Lamarckian ideas of directed, need-based inheritance, and it was rigorously demonstrated, not merely assumed, by mid-20th-century experiments described below.

Heritable variation enters populations through a limited toolkit of molecular and cellular processes. Understanding each one mechanistically clarifies both how much variation is generated and why it appears random with respect to fitness.

1. Mutation — the ultimate source of new alleles. Mutation is any change in DNA sequence, and it is the only process that creates genuinely new alleles (all other processes reshuffle existing ones). Point mutations arise mainly from errors during DNA replication that escape proofreading and mismatch repair, from spontaneous chemical damage (e.g., deamination of cytosine to uracil, depurination), and from the action of mutagens (UV light, reactive oxygen species, certain chemicals). Larger-scale mutations include insertions, deletions, duplications, inversions, translocations, and whole-genome duplications (polyploidy), which can create entire duplicate genes free to diverge in function. Mutation rates are low per base pair per generation — roughly 10⁻⁸ to 10⁻⁹ per nucleotide site per generation in many eukaryotes — but because genomes contain billions of base pairs and populations contain many individuals, the number of new mutations entering a population every generation can be substantial. In humans, each newborn carries on the order of 60–100 new point mutations not present in either parent, based on whole-genome trio sequencing studies.

2. Recombination — reshuffling existing variation. During meiosis, homologous chromosomes pair and undergo crossing-over, physically exchanging segments of DNA. This breaks up the specific combinations of alleles a parent inherited and creates chromosomes with novel combinations to pass to offspring. Recombination does not change allele frequencies by itself, but it changes which alleles are linked together, generating new multi-locus genotypes and allowing beneficial alleles at different loci to be combined into the same individual (or freeing a beneficial allele from a nearby deleterious one — reducing what population geneticists call Hill–Robertson interference).

3. Independent assortment. During meiosis I, each pair of homologous chromosomes orients and separates independently of every other pair. For an organism with n chromosome pairs, this alone generates 2ⁿ possible gamete genotype combinations from a single individual's chromosome set (2²³ ≈ 8.4 million in humans), even before considering recombination within chromosomes.

4. Random fertilization. In sexual populations, which sperm fertilizes which egg is effectively random with respect to genotype, multiplying the combinatorial possibilities from independent assortment and recombination in both parents. The result is that full siblings (except identical twins) are essentially never genetically identical.

5. Gene flow (migration). Movement of individuals — or their gametes, as in pollen dispersal — between populations introduces alleles that may be rare or entirely absent locally, increasing variation within a recipient population while typically reducing genetic differentiation between populations. It is essential to note what does not generate heritable variation: environmentally induced phenotypic changes (developmental plasticity, learned behavior, somatic changes) are not passed to offspring via DNA sequence and so are excluded from this accounting, except in the more limited and still actively researched cases of transgenerational epigenetic inheritance, which does not change DNA sequence itself.

A second, complementary way to see the mechanism is at the level of a single chromosome undergoing meiosis, showing exactly where new combinations of alleles are physically produced.

The single most important experimental question underlying this topic is: are mutations directed by the environment (arising because they are needed), or do they arise spontaneously and randomly, with the environment merely revealing which pre-existing variants survive? This was not obvious a priori, and it was settled by a landmark experiment in bacterial genetics.

The Luria–Delbrück fluctuation test (1943). Salvador Luria and Max Delbrück studied Escherichia coli populations challenged with bacteriophage T1, a virus that normally kills the bacteria. Occasionally, phage-resistant bacterial colonies arose. Two hypotheses were on the table. Under the "acquired immunity" (Lamarckian-style) hypothesis, resistance would arise in response to exposure to the phage, at roughly the same low rate in every independent culture, because each culture has the same opportunity to acquire resistance upon contact with the virus — predicting a Poisson distribution of resistant colony counts with variance ≈ mean. Under the "mutation" hypothesis, resistance mutations would arise spontaneously and randomly at any point during growth, before ever encountering the phage; if a mutation happened to occur early in a culture's growth, its descendants would multiply, producing a "jackpot" of many resistant cells in that culture, while most cultures with no early mutation would show few or none — predicting a highly skewed distribution with variance far exceeding the mean. Luria and Delbrück grew many independent small cultures of E. coli, then plated each entire culture on phage-laden agar and counted resistant colonies. The result: enormous variance among cultures — some cultures showed zero resistant colonies, while a few showed hundreds, far more than Poisson variation would allow. This "jackpot" pattern statistically matched the mutation hypothesis and rejected the acquired-immunity hypothesis. The paper's statistical treatment (variance/mean ratio, later called the Luria–Delbrück distribution) became a foundational tool in population and molecular genetics. Luria and Delbrück shared the 1969 Nobel Prize in Physiology or Medicine (with Alfred Hershey) for this and related work founding molecular biology of viruses.

The Lederberg replica-plating experiment (1952). Joshua and Esther Lederberg extended this logic with an elegant, direct visual demonstration. They grew a bacterial culture without any phage or antibiotic present, then used a velvet-covered stamp to transfer ("replica-plate") the exact same spatial pattern of colonies onto several selective plates (e.g., containing phage or antibiotic). Colonies that grew at the same position on multiple replica plates must have already been resistant on the original, non-selective master plate — before ever being exposed to the selective agent. This proved directly that resistant mutants pre-existed in the population before selection was applied, ruling out induced/directed mutation and confirming that selection merely reveals pre-existing variation rather than creating it. The Lederbergs' work, alongside Luria and Delbrück's, is the empirical backbone of the claim that mutation is random with respect to fitness — one of the load-bearing premises of the entire theory of evolution by natural selection. Joshua Lederberg received the 1958 Nobel Prize in Physiology or Medicine in part for this and other bacterial genetics work. Together, these experiments established a causal order that this whole article depends on: variation (mutation) arises first, independent of need; selection (survival on phage plates) acts second, sorting among variants that already exist. Modern whole-genome sequencing of natural and experimental populations (e.g., the Lenski long-term E. coli evolution experiment, ongoing since 1988) has repeatedly confirmed and extended this picture at the level of individual base-pair substitutions.

Example 1 — A dihybrid cross generating new variation through independent assortment and recombination. Consider a plant heterozygous at two loci carried on different chromosomes: seed color (Y = yellow, dominant; y = green, recessive) and seed shape (R = round, dominant; r = wrinkled, recessive). Cross two double heterozygotes, YyRr × YyRr — this is essentially Mendel's own dihybrid cross. Because the two loci assort independently (different chromosomes, or far apart on the same one), each parent produces four gamete types in equal proportion: YR, Yr, yR, yr — exactly the 2ⁿ rule for n = 2 loci. A standard Punnett square across 4 × 4 = 16 combinations yields the classic phenotypic ratio:

Notice that two of these four phenotypic classes — yellow/wrinkled and green/round — did not exist in either parent's own phenotype in the same combination as the original grandparental lines that founded this cross (imagine the grandparents were pure-breeding yellow-round and green-wrinkled). Independent assortment alone, with no new mutation at all, generated novel heritable combinations of existing alleles at a predictable 9:3:3:1 ratio. This is the mechanistic essence of why sexual reproduction is such a powerful generator of standing variation even though it creates no new alleles: it constantly remixes existing ones into new combinations, some of which may be more or less fit than either parental combination — supplying selection with new genotypes to test every generation.

Example 2 — Estimating heterozygosity from real allele-frequency data. Suppose a population geneticist samples 200 individuals of a beetle species and genotypes them at a microsatellite locus, finding three alleles with frequencies:

Assuming Hardy–Weinberg proportions (no selection, migration, drift, or non-random mating at this locus in this generation), expected heterozygosity (H_e) is calculated as one minus the sum of squared allele frequencies: H_e = 1 − (p₁² + p₂² + p₃²) H_e = 1 − (0.36 + 0.09 + 0.01) H_e = 1 − 0.46 = 0.54 So we expect 54% of individuals to be heterozygous at this locus — a substantial reservoir of standing variation compressed into a single number. Compare this to a second, geographically isolated population of the same beetle where a founder event left only two alleles, at frequencies 0.95 and 0.05: H_e = 1 − (0.9025 + 0.0025) = 1 − 0.905 = 0.095 The founder population has less than one-fifth the heterozygosity of the source population (0.095 vs. 0.54). This is exactly the signature expected after a population bottleneck: rare alleles are disproportionately lost when only a few founders establish a new population, heterozygosity collapses, and — crucially for evolutionary potential — that population now has far less raw material at this locus for selection to act on, making it more vulnerable if the environment later changes in a way that would have favored the now-missing A₃ allele. This calculation is the direct quantitative link between the qualitative sources of variation described above and their consequence for a population's evolutionary future, foreshadowing the fuller treatment of Allele and Genotype Frequencies.

Understanding heritable variation as evolution's raw material is not an abstract exercise — it underlies life-and-death decisions in medicine, agriculture, and conservation. In antibiotic and antiviral resistance management, clinicians and public health agencies design combination therapies (e.g., multi-drug regimens for HIV or tuberculosis) precisely because resistance mutations pre-exist at low frequency before treatment begins, exactly as the Lederbergs demonstrated; hitting a pathogen with several drugs simultaneously makes it statistically unlikely that any single pre-existing lineage already carries mutations conferring resistance to all of them at once. In cancer biology, tumors are understood as evolving cell populations: pre-existing genetic heterogeneity among tumor cells, generated by ongoing mutation, is what allows some clones to survive chemotherapy and drive relapse — directly analogous to phage resistance in bacteria. In agriculture and plant breeding, breeders deliberately maintain or introduce genetic variation — through wild relative crosses, mutagenesis screens, or marker-assisted selection — because a genetically uniform crop has no raw material to respond to a new pathogen or changing climate; the Irish Potato Famine (1845–1852) is a historical case where near genetic uniformity of the planted potato clone left the crop with essentially no standing resistance variation against Phytophthora infestans. In conservation biology, measuring heterozygosity and allelic diversity (exactly as in the worked example above) is now routine for assessing extinction risk in small, bottlenecked populations — famously in cheetahs, whose extremely low genome-wide heterozygosity (documented via skin-graft and later molecular studies) is linked to reduced disease resistance and reproductive fitness. In synthetic biology and biotechnology, directed evolution techniques (for which Frances Arnold shared the 2018 Nobel Prize in Chemistry) explicitly harness mutation and recombination in the laboratory, generating variant protein libraries and then applying artificial selection — an engineered mirror of the natural process this article describes.

Ask the tutor: How does the rate of new mutation entering a population set an upper limit on how fast that population can adapt to a rapidly changing environment?