by Huntington Potter
Both electron micrograph pictures to the right show two EcoR1-linearized plasmid DNA molecules in the midst of undergoing genetic recombination. Since David Dressler and I published it in 1977, the photo and our accompanying diagram have been re-printed in many biochemistry and genetics textbooks as the physical proof of the hypothetical recombination intermediate that Robin Holliday proposed in 1964 to explain genetic recombination in fungi . The micrograph strongly contributed to the general acceptance of the Holliday Model and its intermediate, commonly referred to as Holliday Junction, as explaining the process of recombination in all organisms.
As shown in the diagram (Fig. 2), in this model, the first step of recombination occurs when one single strand of one DNA double helix is nicked and extended by DNA polymerase to form an extruded single strand. This single strand can then invade the homologous double helix to form a D-loop (A – C). After further nicking, extrusion, DNA digestion, and re-ligation (D – E), the Holliday recombination intermediate (F) is formed. Here two exchanged single strands are shared between the two parental DNA molecules in an intermediate structure with apparent two-fold rotational symmetry.
The great strength and acceptance of the Holiday recombination intermediate is based on its ability to explain a number of curious features in the genetic recombination data that had been gleaned from the study of certain fungi, such as Neurospora crassa. The first such feature was the presence of areas of heteroduplex DNA, shown in the diagram as regions where a white strand is paired to a complementary black strand. Such heteroduplex DNA must have been generated during recombination because it could be detected genetically, i.e.,when phenotypically distinguishable alleles occupied the heteroduplex DNA region and, after the first round of mitotic DNA replication after recombination, emerged in a different frequency than they had entered. For example, the two daughter double helices in ‘H’ contain a region in which there are three white strands and only two black ones, and the adjacent area contains the original number of two black and two white strands but in a different order. Neurospora has the unusual feature that it retains all products of recombination and the first mitotic cell division in the order of their production in a linear pod. If the parental cells contain alleles that happen to be adjacent to a recombination site, odd outcomes can be detected, for example, using the colored DNA strands in the diagram as though they were markers, Bl, Bl, Wh, Wh, Wh, Wh, Wh, Wh, Wh or Bl, Bl, Wh, Wh, Bl, Bl, Wh, Wh instead of the usual non-recombinant Bl, Bl, Bl, Bl, Wh, Wh,Wh, Wh. Such unusual daughter genotypes would arise when the DNA molecules in ‘H’ undergo their first mitotic division.
The symmetry of the intermediate that is so apparent in the electron micrograph and in the rotated 'open' intermediate (G) in the diagram shown in Fig. 2 is the second essential feature that makes the recombination intermediate such a powerful centerpiece of Holliday’s proposed mechanism for genetic recombination. That is, Holliday's proposed recombination intermediate could be broken apart, or matured by cleavage of the two recombining DNA double helixes via two symmetrical alternatives: the cleavage could be from left to right in the diagram to yield the left products in 'H' or from top to bottom to yield the right daughter product in 'H'. In one case the external allelic markers (originally A and B on one input DNA molecule and a, b in the other) retain their parental configuration (H left). In the alternative maturation step, the markers are recombined (H right). Thus, after a recombination event, genetically revealed by the formation of heteroduplex DNA, there is an equal chance that the flanking parental alleles will be either recombined or left in their original parental linkage. This too was known from the study of Neurospora. Thus the formation and maturation of the Holliday intermediate succeeded in satisfying and explaining all of the genetic observations.
The simplicity of the Holliday recombination intermediate and the model that utilizes it was attractive but it was not the only recombination model discussed in the 1960s and early 70s, and there was no genetic way to prove which model was correct. That proof required the use of different techniques and led to our discovery of structures such as the one shown in the electron micrograph, in which DNA molecules were captured in the process of genetic recombination, using the Holliday intermediate.
The origin of the molecule pictured above, which proved the existence and universality of the Holliday recombination intermediate, and by inference, the validity of the Holliday model, is an interesting story of good planning, much hard work, and some luck, as is true of most scientific discoveries. The story begins in 1973 in the laboratory of David Dressler in the Biochemistry and Molecular Biology Department of Harvard University. As a new graduate student, I needed to pick a dissertation topic and we were having a discussion about what that might be. I had already worked to some extent on DNA replication, which was David Dressler's forte, for he had discovered the greater-than-one-genome-length single stranded DNA tail that led to the 'rolling circle' model for DNA replication used in a number of phages, such as ΦX174 — see STC's recent post — and in Xenopus laevis ribosomal DNA amplification, and then found that replication of the linear phage T7 DNA proceeded via two growing forks moving in opposite directions from a single origin. This area was fruitful and important, but I was more interested in the process of DNA recombination, about which almost nothing was known mechanistically. I had read numerous articles especially in the phage lambda literature about genetic recombination, thinking that it might be possible to deduce the mechanism from the genetic data. but was frustrated. In our discussion, David suggested that identifying a recombination intermediate might be extremely valuable, as it would give an indication of the mechanism by which the intermediate was formed and then matured, as replication intermediates had provided similar insights into the process of DNA replication. At that time, neither of us knew about Robin Holliday and his theoretical intermediate and model for genetic recombination. David specifically suggested that we use what was then the extremely novel process of recombinant DNA manipulation to introduce a single restriction site into the genome of the ΦX174 virus so as to be able to cleave double-stranded circles that were undergoing recombination and see if we could catch them in the intermediate stage of the process.
That seemed to me to be a good idea but I then suggested a modification, which was to utilize instead a small plasmid that already had a single restriction site in its genome. With this simple approach, I then began to isolate plasmid DNA from bacteria and examine the individual circular molecules in the electron microscope. Naturally, when the DNA was combined with formamide and cytochrome C and spread in a protein monolayer on an aqueous surface before being picked up on the electron microscope grid, the individual DNA circles were very often overlapping each other, and such overlaps were of course indistinguishable from what we might expect a figure 8 recombination intermediate to look like. Therefore the DNA preparation of plasmids had to be sufficiently diluted before spreading so that overlaps would be extremely rare. Needless to say, this increased greatly the amount of time I had to spend on the electron microscope.
As originally planned, I also cleaved the plasmid DNA preparation with the restriction enzyme EcoR1 before spreading for the electron microscope. As there was one EcoR1 site per genome, this converted the double-stranded DNA circles to linear DNA molecules. If any were in the process of DNA recombination with each other, the proposed figure 8 intermediate molecule would be converted to an X, which we termed a Chi (χ) form. Again keeping the DNA concentration extremely low reduced greatly the chances that two linear double-stranded molecules would overlap each other at random and generate an X shape. I easily found such χ-forms at a low, but reproducible frequency in linearized plasmid DNA preparations. But were they real recombination intermediates?
A necessary feature of a genetic recombination intermediate is that the two DNA molecules must be interacting at a point of DNA homology. We could test this requirement in our χ-forms by measuring carefully the lengths of the DNA arms coming out from the apparent crossover. If the arms were of random length, then the X would represent the random overlap of two linear molecules. However, if the arms were composed of two pairs of equal length, then the crossover site was at a region of homology because it occurred at the same distance in each molecule from a known genetic marker (the EcoR1 cleavage site). Each molecule could have a different combination of arm lengths, but it should always be symmetric — and they were!
The fact that we found any recombining χ-forms at all was due to our luck to have available the right reagent: a very concentrated preparation of the EcoR1 restriction enzyme. At that time, only a few, low-activity, restriction enzymes were available commercially and most labs had to make their own. I and a post-doc in the next lab made EcoR1 for everyone, and our prep, by some luck, turned out to be extremely concentrated and powerful: a DNA digestion that might take several hours or overnight with other preparations could be accomplished in 30 seconds! This high activity turned out to be essential for our recombination experiment because once the DNA molecules were linearized, the crossover point could easily 'branch migrate' to the end of the molecules by continued strand exchange with essentially no energy barrier, and release two linear molecules. Only by digesting for a mere 30 seconds and immediately placing the DNA on ice, could we trap the χ-forms before they disappeared!
Finally, to complete the proof that we had found the elusive Holliday recombination intermediate, we had to show that the fine structure of the crossover point of the χ-forms we observed was indeed as predicted by Holliday (see ‘G’ in the diagram), and that they were formed during recombination. The first problem was solved by partially denaturing the molecules in a slightly elevated formamide concentration during spreading for the electron microscope so that the single stands in the crossover could be distinguished as in the micrograph shown in Fig. 1. To complete the second half of the proof, we showed that such χ-forms were only present in recombination proficient E.coli but were absent from plasmid preparations from recombination deficient (recA−) E. coli.
The fact that we found in bacteria the physical recombination intermediate that Robin Holliday predicted for fungi served not only to establish that the fundamental mechanism of genetic recombination utilized this intermediate, but it also unified the field, for recombination evidently used the same intermediate in both prokaryotes and eukaryotes.
Huntington Potter, Ph.D. is Professor of Neurology and Director of Rocky Mountain Alzheimer's Disease Center, University of Colorado School of Medicine, and Director of the Alzheimer's Disease Program at the Linda Crnic Institute for Down Syndrome.
 Holliday R. 1964. A mechanism for gene conversion in fungi. Genet Res, 5, 282 – 304 (not in PubMed)