puzzle · intermediate · partly written
DNA replication in 1957: where do the old atoms go?
A replay of the DNA copying puzzle with only the evidence of 1957: three models of what happens to the parent molecule, and no experiment in print that could tell them apart.
Also called: Meselson-Stahl replay, conservative semi-conservative dispersive
Where the history is disputed (both readings, no verdict) more
Who conceived the density-label idea. A Harvard departmental news item (reported) says Meselson asked, after a lecture, about labelling with heavy water, and that Meselson and Stahl met at Woods Hole in 1954; Stahl was at Caltech from 1955. The Embryo Project encyclopedia, which summarises Holmes's histories, dates the start of their joint work differently and credits Meselson, Stahl and Vinograd together with the cesium chloride method. Both agree Meselson was the source of the idea; Stahl's part is not in his own words in anything opened, and Vinograd's share is in no source opened.
Who held 'conservative', and when the three names appeared. Delbruck and Stent's chapter (opened) defines the conservative class as mechanisms in which 'among the daughter duplexes produced by one or more replications there is one which is entirely parental and the rest entirely new', and puts Bloch's scheme and Butler's scheme in that class, together with replication through an intermediary such as protein or RNA; the semi-conservative class is described as 'the original suggestion of Watson and Crick' and the dispersive class names Delbruck's own scheme. The chapter does not name Stent as holding the conservative view. The Embryo Project's page on the categorisation names Stent and Bloch (1955), while its page on the debate names Stent only and says Bloch's mechanism fits none of the three; on the placing of Bloch the chapter itself is the primary record and the two Embryo Project pages disagree with it and with each other. Dating: the volume's preface says the symposium was held on 19-22 June 1956 and the volume prints 'Copyright 1957, The Johns Hopkins Press', so the chapter was given in 1956 and printed in 1957. The 1956 date in some tellings is the talk and the 1957 date in the Meselson-Stahl reference list is the printing; both are right about different events.
What excluded the dispersive model. Many tellings say the two-band pattern at the second generation. The 1958 paper itself names the heat step: heated hybrid DNA split into two species at about half the molecular weight, and the authors write that Delbruck's scheme is thereby ruled out.
Was the earlier work inconclusive? One secondary calls Levinthal's phage work inconclusive; Levinthal's own text supports the Watson-Crick scheme for the large piece of phage DNA and says the detailed predictions cannot be checked by these experiments. Was Taylor 1957 the first proof? Meselson and Stahl cite it only in a list of work bearing on the question, and Taylor wrote in 1983 that too little was known about chromosome structure to be sure chains were segregating. 1234567
Contested
What the 1958 paper itself said it had not shown more
The authors write that it has not been shown that the molecular subunits are single polynucleotide chains, or even that the DNA molecules studied correspond to single DNA molecules of the Watson-Crick structure, and that salmon-sperm DNA did not halve on heating, which they say may bear on whether the Watson-Crick model is sufficient. They use the word 'subunits', not 'strands'. A tidy retelling that says the experiment proved strand-by-strand copying overstates the paper. 3
Established
The observation
DNA is two complementary chains wound in a double helix, and each cell division gives each daughter cell a full set of it. So the molecule doubles. What nobody can see is what happens to the parent molecule when it does: whether it survives intact, is shared out between the two daughters, or is broken up and the pieces spread through both. 89
Reported, not yet replicated
What was known then
The structure suggests a copying mechanism 1953
Watson and Crick's paper of April 1953 says the specific pairing of the bases immediately suggests a possible copying mechanism. 8
Reported, not yet replicated
The copying scheme, hedged 1953
A month later Watson and Crick described a scheme in which the chains unwind and separate and each serves as a template for a new partner chain. They called it speculative and asked what makes the pair of chains unwind and separate. The paper does not use the name 'semi-conservative'. (The paper was read from a poor scan, so wording is approximate.) 9
Reported, not yet replicated
The unwinding objection, in geometry 1954
Delbruck (1954) argued that if the two chains of a helix part and each makes a new partner, the two daughter duplexes come out wound around each other, which he called the principal difficulty. He proposed instead breakage and reunion of the chains, with breaks needed at about every half turn, and gave a label test: with breaks, parental label is shared statistically equally between the daughter duplexes at each replication, while without breaks one daughter would receive all of the label and the other none. He wrote that at present it did not seem possible to discuss the bearing of this implication on Stent's phage-mortality experiments, because of multiplicity reactivation. He also estimated that exchange of label at open chain ends could account for the observed 30 to 50 per cent transmission of parental phosphorus-32 to progeny phage. (Re-read against the paper, 2026-11-17.) 1010
“The principal difficulty of this mechanism lies in the fact that the two chains are wound around each other in a large number of turns” On the replication of desoxyribonucleic acid (DNA), p. 784
“At present it does not seem possible to discuss the bearing of this implication” On the replication of desoxyribonucleic acid (DNA), p. 788
Reported, not yet replicated
The three classes named in print 1957
Delbruck and Stent's chapter, printed in 1957 from a talk given in June 1956, sorts the proposed copying schemes by what they predict for the parent's atoms: conservative (one daughter duplex wholly old, the rest wholly new), semi-conservative (each daughter duplex has one old chain and one new) and dispersive (old and new material spread through all chains). It calls an unambiguous experimental test of the distribution of parental atoms a way to eliminate 'a considerable fraction of potential replication mechanisms', and says that if the new material of each generation could be labelled differently one could read off the ages of the two chains in a duplex. It adds that recombination in phage, which happens about as often as replication, can blur the distribution of parental atoms there. 7
“Unambiguous experimental tests of the actual distribution of the atoms of the parental DNA molecule could thus eliminate a considerable fraction of potential replication mechanisms” On the mechanism of DNA replication, section Distribution of parental atoms, pp. 708-711
Reported, not yet replicated
Radioactive phosphorus in phage 1956
Levinthal (1956) followed phosphorus-32 from parent phage into progeny. Roughly 40 per cent of the phage DNA behaves as one large piece that gives two particles each with about half its atoms, which he judged probably consistent with the Crick and Watson scheme; but only part of the label transfers, the label can be examined only after a whole growth cycle, and he wrote that the detailed predictions cannot be checked by these experiments. 4
Reported, not yet replicated
DNA made in a test tube 1956
A 1956 note (read by the chain as a citation record only) reports that an extract of E. coli makes a DNA-like polymer when DNA primer is added. It tests the chemistry of copying, not what becomes of the parent chains. 11
Reported, not yet replicated
Chromosomes in plant roots 1957
Taylor, Woods and Hughes (15 January 1957) grew Vicia faba roots in tritium-labelled thymidine. After one cycle in label both chromatids were labelled; after a second cycle without label, one of two. They proposed that DNA is synthesised in two complementary units, analogous to the Watson-Crick model. It resolves chromatids, not molecules, and a chromatid may hold many. 5
Reported, not yet replicated
A way to separate by density, with no tag yet 1957
Meselson, Stahl and Vinograd (communicated 27 May 1957) showed that cesium chloride centrifuged near equilibrium forms its own density gradient and macromolecules band where the solution density matches their own; band width falls with molecular weight. T4 DNA banded near 1.70 g/cm3. The paper contains no isotope labelling and no replication: at the freeze there is a method and no tag. 12
Reported, not yet replicated
Your turn
Commit before you read on. What you write stays in this browser and is sent nowhere.
I have committed. Show what was proposed, and what happened.
Hypotheses of the time
Semi-conservative: each daughter keeps one old chain
The chains separate and each templates a new partner, so each daughter helix has one parental and one new chain. 95417
Held by. Watson and Crick 1953; a 'transfer twist' variant attributed to Platt 1955 (reported only).
Outcome. right
Why reasonable. It follows directly from complementary base pairing; Taylor's 1957 chromatid result fitted it at the level of whole chromosomes; Levinthal judged his large phage piece probably consistent with it. Its weakness was the unwinding problem, which the authors themselves posed.
“This is the original suggestion of Watson and Crick and certainly the one most obviously suggested by their structure.” On the mechanism of DNA replication, Semi-conservative mechanisms
Reported, not yet replicated
Dispersive: break and rejoin
Parental material is split into short pieces, about five nucleotides on average, that end up in both daughters, so label dilutes and never segregates. 10
Held by. Delbruck 1954.
Outcome. wrong
Why reasonable. It needs no unwinding, which nobody had seen a helix do; Delbruck's geometry (one interlock per ten-nucleotide turn) was sound; and he argued that the 30 to 50 per cent transfer of phosphorus-32 in phage fits exchange at open ends. He also wrote down the label test that would later be run.
Reported, not yet replicated
Conservative: the parent helix stays whole
The parent duplex is untouched and a wholly new duplex is made, so there are heavy-only and light-only daughters and no hybrid. 147
Held by. Delbruck and Stent 1957 define the class and list schemes of Bloch and of Butler, and replication through an intermediary such as protein or RNA, as conservative; they do not name Stent as holding it. The Embryo Project's pages differ on whether Stent and Bloch held it (secondary, reported). Levinthal cites Stent and Fuerst 1955 as suggesting information passes to a non-DNA component.
Outcome. wrong
Why reasonable. It needs no strand separation, and an information-carrying intermediate was a live enough idea in 1956 that Levinthal addressed it directly: he cited Stent and Fuerst's suggestion that information passes to a non-DNA component and set it aside, writing that "the evidence does not, at present, seem sufficiently compelling to require consideration of this extra complication" (p. 403). So this class drew no support from him; he read his own data as probably complementary replication in the Crick and Watson manner.
“The mechanisms proposed by Bloch (3) and by Butler (9), in which the parental duplex as a whole serves as a direct template for the production of a daughter duplex, likewise fall into this class.” On the mechanism of DNA replication, Conservative mechanisms
Reported, not yet replicated
Copying without separating the chains
Variants in which the strands are not parted: Gamow's spiral, Arley's protein template, Watson's four-stranded models of 1954 to 1955. 1
Held by. Gamow, Arley, Watson (all reported via the Embryo Project; no primary text opened).
Outcome. unresolved
Why reasonable. They dodge the unwinding objection. This record opened no primary text for them, so their predictions are not stated here.
Reported, not yet replicated
What broke it open
Heavy nitrogen plus the cesium chloride gradient, 1958 1958
Meselson and Stahl (communicated 14 May 1958) grew E. coli for 14 generations in medium with 15NH4Cl (96.5 per cent 15N), then switched to 14N. DNA from cells in cesium chloride at density 1.71 g/cm3 (44,770 rpm, about 140,000 g, 25 C, 20 h) banded by ultraviolet absorption, with the heavy and light bands 0.014 g/cm3 apart. Up to one generation after the switch a half-labelled band accumulated and at one generation only that band was present; at two generations half-labelled and light DNA were present in equal amounts, the hybrid band lying 50 +/- 2 per cent of the way between the heavy and light bands. No fully heavy or fully light daughters at one generation excluded the conservative model. 3
Established
The heat step 1958
Hybrid DNA heated at 100 C for 30 minutes split into two species at about half the apparent molecular weight, whose density difference (0.015 g/cm3) matched the label. The authors concluded that the dispersive scheme proposed by Delbruck is thereby ruled out, and wrote that it had not been shown that the subunits are single polynucleotide chains. 3
Established
The answer
DNA replication is semi-conservative: each daughter duplex has one parental and one newly made chain. The 1958 result showed this at the level of 'subunits' conserved through replication, and excluded the conservative and dispersive models. Later work on the shape of the replicating chromosome (Cairns, 1962 and 1963) was not opened for this record. Caveats stated by the authors: density banding says nothing about whether a subunit is one polynucleotide chain; apparent molecular weights are wrong if cesium chloride or water binds preferentially; and band width mixes density heterogeneity with molecular weight. 3
Established
Why it was hard
A radioactive label counts atoms and the atoms decay; it cannot show whether old atoms stayed together. The three models agreed that DNA doubles and differed only on the parent's fate, so any experiment that measured amounts, rates or total label could not tell them apart. A stable, bulk-readable tag, and a sorter that reads each molecule's content at set generations, was the missing pair. In 1957 the sorter existed (July) and the tag had not been put on the same system. The general lesson: when rival mechanisms predict the same totals, change what is tagged until they predict different distributions. 10412
Our inference
The move
Status quo before
What forced it
The trick
Make old and new differ in a bulk physical property (mass, from heavy nitrogen) and let the molecules sort at equilibrium; read the answer from band positions at set generations after a switch from heavy to light medium. 3
Established
Why it worked
Density is additive in heavy atoms, nothing is destroyed, and each model predicts a different picture at each generation: the conservative model gives heavy and light with no hybrid at one generation, semi-conservative gives one hybrid band, and the dispersive model keeps one band that drifts lighter. 310
Our inference
Still open
Were the 1958 subunits single chains?
The authors said it had not been shown. The record author did not open the later work that settled this. 3
See: How were the conserved subunits of the 1958 DNA experiment shown to be single polynucleotide chains? not yet written
Unverified: from our own report, source not opened
Who conceived the experiment, and what did each author contribute?
Secondary accounts agree that Meselson was the source of the idea; Stahl's and Vinograd's parts are not described in their own words in anything opened. 21
See: Who conceived the density-label experiment on DNA replication, and what did each author contribute? not yet written
Contested
Still open here
- Holmes's 2001 book was not opened (publisher-held; no legal full-text copy found). What the Atlas takes from him is second-hand through the Embryo Project's page on the book; see people/frederic-lawrence-holmes.
- Cairns 1962 and 1963, and the work that showed the subunits are single chains, were not opened (publisher sites and PubMed Central page and PDF downloads refused automated access from this machine).
- Levinthal 1956 and Taylor, Woods and Hughes 1957 have since been read in the browser from their page scans (2026-11-18); the obstacle at the time was that the scans carry no machine-readable text and the PNAS site refused access.
- Kornberg's 1956 note was seen as a citation record only.
- Delbruck and Stent cite a 'Delbruck, 1955' in their reference list for the paper that is Delbruck 1954 (PNAS 40:783); the chapter's own list is the source of that slip, and nothing here depends on it.
Linked from
- Equilibrium density-gradient centrifugation
- Semi-conservative replication
- Frederic Lawrence Holmes
- How were the conserved subunits of the 1958 DNA experiment shown to be single polynucleotide chains? not yet written
- Who conceived the density-label experiment on DNA replication, and what did each author contribute? not yet written
Sources
- Categorization of conservative, semi-conservative and dispersive DNA replication theories, 1953-1957 (Embryo Project Encyclopedia). Embryo Project Encyclopedia, Arizona State University. web, not peer reviewed; opened 2026-11-17 all records citing this
- Matthew Meselson's beautiful experiment turns fifty (Harvard MCB news). Harvard University, Department of Molecular and Cellular Biology (2008). web, not peer reviewed; opened 2026-11-17 all records citing this
- Meselson, M., Stahl, F. W.. The replication of DNA in Escherichia coli. Proceedings of the National Academy of Sciences USA 44:671-682 (communicated by M. Delbruck, 14 May 1958) (1958). paper; opened 2026-11-17 all records citing this
- Levinthal, C.. The mechanism of DNA replication and genetic recombination in phage. Proceedings of the National Academy of Sciences USA 42:394-404 (1956). paper; opened 2026-11-18 all records citing this
- Taylor, J. H., Woods, P. S., Hughes, W. L.. The organization and duplication of chromosomes as revealed by autoradiographic studies using tritium-labeled thymidine. Proceedings of the National Academy of Sciences USA 43:122-128 (15 January 1957) (1957). paper; opened 2026-11-18 all records citing this
- Taylor, J. H.. Citation Classic commentary on Taylor, Woods and Hughes 1957 (Taylor's 1983 recollection). Current Contents (Citation Classics) (1984). web, not peer reviewed; opened 2026-11-17 all records citing this
- Delbruck, M., Stent, G. S.. On the mechanism of DNA replication. In McElroy, W. D. and Glass, B. (eds), A Symposium on the Chemical Basis of Heredity, Johns Hopkins Press, pp. 699-736 (1957). paper, not peer reviewed; opened 2026-11-17 all records citing this
- Watson, J. D., Crick, F. H. C.. Molecular structure of nucleic acids: a structure for deoxyribose nucleic acid. Nature 171:737-738 (1953). paper; opened 2026-11-17 all records citing this
- Watson, J. D., Crick, F. H. C.. Genetical implications of the structure of deoxyribonucleic acid. Nature 171:964-967, 30 May 1953 (1953). paper; opened 2026-11-17 all records citing this
- Delbruck, M.. On the replication of desoxyribonucleic acid (DNA). Proceedings of the National Academy of Sciences USA 40:783-788 (communicated 18 May 1954) (1954). paper; opened 2026-11-17 all records citing this
- Bessman, M. J., Lehman, I. R., Simms, E. S., Kornberg, A.. Enzymatic synthesis of deoxyribonucleic acid. Biochimica et Biophysica Acta 21:197-198 (title and content as reported by the research chain) (1956). paper; opened 2026-11-17 all records citing this
- Meselson, M., Stahl, F. W., Vinograd, J.. Equilibrium sedimentation of macromolecules in density gradients. Proceedings of the National Academy of Sciences USA 43:581-588 (communicated 27 May 1957) (1957). paper; opened 2026-11-17 all records citing this
Where it sits on the map
History of this record (8)
This record has been changed 8 times, on the dates below. The note kept against each change is written for the people who maintain the atlas, in their own shorthand, so it is not shown here; it is in the record itself.
- Freeze date
- 1957
- Status
- partly written
- Last updated
- 2026-11-19
- Record
- content/solved/how-dna-copies-itself.json