Semiconservative Replication and the Meselson–Stahl Experiment

DNA replication is semiconservative: when a double helix is copied, the two parental strands separate, and each serves as a template for a new complementary strand. The result is two daughter duplexes, each consisting of one original ("parental") strand and one newly synthesized strand, held together by the same Watson-Crick Base Pairing rules that specify the double helix in the first place. No parental strand is destroyed or diluted across many generations; it persists intact, physically paired with a new partner. This was not the only model on the table in the 1950s. Three schemes were logically consistent with a double-stranded template carrying sequence information:

• Conservative replication — the original duplex stays fully intact, and an entirely new duplex is synthesized from scratch, made of two brand-new strands. After replication you would have one all-old molecule and one all-new molecule.
• Semiconservative replication — the duplex unwinds, and each strand templates a new partner. Every daughter molecule is a 50/50 hybrid of old and new.
• Dispersive replication — the parental strands are fragmented and interspersed with new DNA segments, so every daughter strand is a patchwork of old and new material distributed along its length.

All three models correctly predict that the total amount of DNA doubles each generation and that daughter molecules carry the same sequence. What differs is the distribution of parental material among daughter molecules, a difference invisible to sequencing or base-composition analysis but visible to a technique sensitive to mass: density-gradient centrifugation. The genius of Meselson and Stahl's design was recognizing that if you make "old" DNA physically heavier than "new" DNA, using a heavy isotope, the three models make three distinct, quantitative, testable predictions about the density of DNA molecules after one, two, and more rounds of replication.

Step 1 — Labeling parental DNA with a heavy isotope. Meselson and Stahl grew Escherichia coli for many generations in a growth medium in which the only nitrogen source was ¹⁵NH₄Cl, ammonium chloride built from the heavy, non-radioactive isotope ¹⁵N (natural nitrogen is almost entirely the lighter ¹⁴N). Nitrogen is incorporated into the purine and pyrimidine bases of DNA, so after many generations essentially all of the bacterial DNA carried ¹⁵N in its bases, making these molecules measurably denser than ordinary ¹⁴N-DNA. This heavy DNA served as the "parental" pool.

Step 2 — The density shift. The heavy-nitrogen culture was abruptly transferred to fresh medium containing only ¹⁴NH₄Cl (light nitrogen), and the bacteria were allowed to continue growing and dividing. From this moment on, any newly synthesized DNA strand would incorporate light nitrogen, while any strand that already existed at the moment of the switch would remain heavy. Samples of cells were harvested at intervals corresponding to zero, one, two, and further generations of growth (bacterial generation time under their conditions was about 40 minutes).

Step 3 — Extracting and banding the DNA. DNA was extracted from each sample and mixed with a concentrated solution of cesium chloride (CsCl), then spun in an ultracentrifuge at very high speed, around 140,000 × g, for around 20 hours. Under this prolonged centrifugal force, the heavy cesium ions themselves settle into a density gradient, denser at the bottom of the tube, lighter at the top. Each DNA molecule migrates through this self-forming gradient until it reaches the point where the solution density exactly equals its own buoyant density, where it stops and forms a sharp, thin band — this is equilibrium (isopycnic) density-gradient centrifugation. Because ¹⁵N-DNA and ¹⁴N-DNA differ in buoyant density by a small but reliably resolvable amount, molecules of pure heavy DNA, pure light DNA, and heavy-light hybrid DNA band at three distinct, predictable positions in the tube.

Step 4 — Visualizing the bands. The bands were photographed using ultraviolet absorption (DNA absorbs UV strongly at 260 nm), producing a photographic image in which each band appears as a dark line at a specific position, convertible to a density value by comparison against reference markers of known density.

Step 5 — Reading out the predictions of each model. This is where the three replication models become distinguishable:

The critical discriminating observation is generation 1. Conservative replication predicts two distinct bands, one fully heavy (old duplex intact) and one fully light (brand-new duplex) — because in this model, the old duplex never mixes with new material at all. Semiconservative and dispersive replication both predict a single intermediate band, since every daughter molecule contains a 50/50 mixture of old and new material by mass (whether cleanly divided into two whole strands or scattered as patches). Generation 1 data alone already refutes conservative replication. The discriminating observation between semiconservative and dispersive comes at generation 2 and beyond, and from an additional physical test: strand denaturation. If the hybrid DNA from generation 1 is heated to separate its two strands and each single strand is then centrifuged, semiconservative replication predicts the single strands separate cleanly into one fully heavy strand and one fully light strand (two bands, because each strand is internally uniform). Dispersive replication predicts the single strands remain intermediate in density even after separation, because each individual strand is itself a patchwork mosaic of heavy and light segments and cannot cleanly resolve into a pure heavy or pure light strand.

Step 6 — Interpreting the actual gradient photographs. In the real experiment, Meselson and Stahl did not need to guess: the UV-absorption photographs of the centrifuge tubes showed, generation by generation, exactly the pattern semiconservative replication predicts, described in detail in the evidence section below.

The organism and the design. Meselson and Stahl used Escherichia coli growing in liquid culture, chosen for its short generation time (roughly 40 minutes under their conditions) and ease of harvesting many synchronized samples. Bacteria were cultured for at least 14 generations in a defined medium in which ¹⁵NH₄Cl was the sole nitrogen source, ensuring essentially complete labeling of all nitrogenous bases in the DNA with the heavy isotope. This "fully heavy" starting population is the equivalent of Meselson and Stahl's generation 0.

The pulse. At time zero, the heavy culture was rapidly diluted into a large excess of fresh medium containing only ordinary ¹⁴NH₄Cl and growth was allowed to continue. Samples were withdrawn at intervals matched to successive doublings of the bacterial population (measured independently by optical density and cell counts), so each sample corresponded to a known number of completed replication rounds.

The physical separation. DNA from each sample was extracted and subjected to equilibrium density-gradient centrifugation in cesium chloride, spun at roughly 140,000 × g for about 20 hours until the CsCl ions themselves formed a stable, smooth density gradient inside the tube and every DNA molecule had migrated to its equilibrium position. The tubes were then photographed under ultraviolet light (DNA's 260 nm absorbance), producing images with sharp, quantifiable bands whose positions could be converted into buoyant density values, precise enough to resolve differences of a fraction of a percent.

The results, generation by generation:

• Generation 0 (before the shift): a single band at high (heavy) density, as expected — all DNA was made in ¹⁵N medium.
• Generation 1 (after one round of replication in ¹⁴N medium): a single band appeared at a density almost exactly intermediate between fully heavy and fully light DNA. Crucially, there was no heavy band and no light band — only one hybrid band. This single observation already excludes the conservative model, which requires two separate bands (one heavy, one light) at this stage.
• Generation 2: two bands appeared, one at the same intermediate (hybrid) density seen in generation 1, and one at the fully light density. The two bands were present in roughly equal amounts.
• Generation 3 and beyond: the intermediate hybrid band persisted, unchanged in position and at a constant, non-diminishing amount, while the light band grew progressively larger, exactly as expected if each generation's stock of hybrid duplexes contributes one heavy strand and one light strand to two new daughter duplexes, only one of which remains hybrid.

The decisive follow-up: strand separation. To rule out dispersive replication (which also predicts a single intermediate band at generation 1), Meselson and Stahl took the hybrid-density DNA and heated it to denature the duplex into single strands, then centrifuged the single strands separately. The result: the single strands resolved cleanly into two distinct bands, one at the density expected for a pure heavy single strand and one at the density expected for a pure light single strand. This is only possible if each strand, as a whole, is internally uniform — entirely old or entirely new — which is precisely the semiconservative prediction. Dispersive replication, by contrast, predicts that even single strands should remain a density mosaic that cannot cleanly resolve, because each strand itself is stitched together from alternating old and new segments.

What this proved. Together, the intact-duplex banding pattern (ruling out conservative replication) and the denatured-strand banding pattern (ruling out dispersive replication) left semiconservative replication as the only model consistent with every observation. The experiment is often praised for its logical economy: three simple predictions, one clean physical measurement (buoyant density in CsCl), and no ambiguity in interpretation. It also stands as a landmark demonstration that isotope labeling combined with a purely physical separation technique can resolve a mechanistic question about molecules far too small to see directly, a strategy that echoes the logic of the earlier Hershey-Chase Experiment establishing DNA as genetic material.

Example 1 — Predicting band patterns at generation 4. Suppose a culture of fully ¹⁵N-labeled E. coli is shifted into ¹⁴N medium and allowed to undergo exactly four rounds of replication. Assume every cell divides synchronously and no DNA is degraded. Work out what fraction of DNA duplexes are hybrid (heavy/light) versus fully light, and predict the banding pattern. Reasoning: Semiconservative replication means every hybrid duplex, when replicated once more, yields one hybrid daughter (heavy strand + new light strand) and one fully light daughter (old light strand + new light strand). The number of hybrid duplexes therefore stays constant in absolute number across generations (each of N₀ original heavy strands is permanently housed in exactly one hybrid duplex, generation after generation), while the total number of duplexes doubles each generation. Starting with 1 fully heavy duplex (generation 0):

So at generation 4, the density gradient should show two bands: a hybrid band representing 2 out of 16 duplexes (12.5%), and a light band representing 14 out of 16 duplexes (87.5%). The hybrid band never disappears and never grows past 2 duplexes worth of material (it becomes proportionally fainter each generation but never zero), because the two original heavy strands persist indefinitely, always paired with new light partners. This precise asymptotic thinning of the hybrid band, rather than its sudden disappearance, is exactly what Meselson and Stahl observed and reported through several generations.

Example 2 — A clinical/diagnostic analogy: distinguishing replication modes using a modern experiment design. Imagine a hypothetical newly discovered thermophilic archaeon whose replication mode is unknown. You culture it in medium with ¹³C-labeled carbon sources (a heavy stable isotope, analogous to ¹⁵N) for many generations, then shift to ¹²C medium and take samples after 1 and 2 generations, running CsCl-equivalent isopycnic centrifugation (or an analogous mass-based separation, such as high-resolution mass spectrometry of intact strands). Reason through what result would let you conclude the organism uses semiconservative replication rather than some organism-specific dispersive variant sometimes proposed for organisms with unusual replicon structures. Reasoning: 1. If generation 1 shows two separate bands (one heavy, one light) instead of one hybrid band, conservative replication is indicated, and you would stop here. 2. If generation 1 shows a single hybrid band, both semiconservative and dispersive remain possible; you cannot yet distinguish them. 3. Denature the generation-1 hybrid DNA into single strands and re-run the separation. If the single strands resolve into two clean bands (pure heavy, pure light), the replication is semiconservative — each strand as a whole is homogeneously old or new. 4. If instead the single strands still band at an intermediate, smeared position, the replication is dispersive — each individual strand itself contains a patchwork of new and old segments, consistent (for instance) with a replication mechanism involving frequent strand breakage and re-ligation with fresh segments, as later shown to occur naturally only in specialized contexts like some breakage-and-repair models, not in normal chromosomal replication. This stepwise logic — first distinguish conservative from the other two using intact duplexes, then distinguish semiconservative from dispersive using denatured single strands — is the general template for testing replication (or any templated copying process) mode in any organism, and it is exactly the two-stage logic Meselson and Stahl actually used.

Matthew Meselson (b. 1930) was a young researcher working at Caltech, in a scientific environment saturated with the aftershocks of the Watson–Crick structure paper, where questions about mechanism were suddenly answerable rather than purely speculative. Franklin Stahl (1929–2018), his collaborator, brought expertise in bacteriophage genetics from working with Max Delbrück's group, part of the same intellectual lineage that had produced the Luria–Delbrück fluctuation test establishing the random, pre-existing nature of mutations. Meselson and Stahl designed their density-gradient centrifugation technique specifically to make an invisible molecular question visible as a physical measurement, refining ultracentrifugation methods to the point where they could resolve density differences of only about 1%, sensitive enough to separate ¹⁴N-DNA from ¹⁵N-DNA and, remarkably, hybrid molecules from both. The experiment was conducted at Caltech and published in 1958 in the Proceedings of the National Academy of Sciences under the modest title "The replication of DNA in Escherichia coli." It took roughly three years of methodological refinement, developing reliable equilibrium centrifugation and reproducible band photography, before the clean generational data set was obtained. The result was so immediately persuasive and methodologically elegant that it is frequently cited in textbooks and by working biologists as the most beautiful experiment in the history of biology, prized for how a single, well-designed physical measurement could adjudicate decisively among competing molecular hypotheses.

Understanding that replication is semiconservative underlies essentially every modern technique that tracks DNA synthesis or turnover. Molecular diagnostics exploit strand-specific and density-based labeling logic in techniques descended conceptually from Meselson–Stahl, including density-based separation of newly replicated DNA in studies of replication timing across the genome. Cancer biology relies on tracking newly synthesized DNA (using thymidine analogs like BrdU or EdU, direct descendants of the isotope-labeling logic) to measure how fast tumor cells are dividing, informing chemotherapy dosing and prognosis. Antibiotic development targets bacterial DNA replication machinery (gyrase inhibitors like fluoroquinolones, for example) precisely because replication is essential and mechanistically well-understood at the strand level, allowing rational design of drugs that block specific replication steps without harming the human host's very different-shaped replication apparatus. Forensic and ancient DNA work depends on understanding strand integrity and inheritance patterns rooted in the same template-strand logic. And in basic research, the conceptual scaffold of "one strand old, one new" directly motivates the mechanistic questions taken up next: how the double helix is physically unwound at an Origin of Replication to create a Replication Bubble, and how DNA Polymerase enzymes synthesize the new strand exclusively in the 5′-to-3′ direction, reading the old strand as template.

Ask the tutor: How does the replication fork physically unwind the parental duplex to expose the two template strands for semiconservative copying?