Fifty-two years ago I was venturing to the basement of Cal Tech chemistry with some regularity, looking at nucleic-acid diffraction data using the school's admittedly primitive fiber X-ray facilities. My postdoctoral advisor at the time, Linus Pauling, had been interested in finding the structure of DNA, but Watson and Crick had largely eclipsed that effort. Now, collaborating with Jim Watson, who had returned from Cambridge, I was taken with a challenge put forth in his famed double-helix manuscript.¹

In their 1953 publication on the DNA double helix, Watson and Crick stated: "It is probably impossible to form this structure with ribose, instead of deoxyribose." The reason: The 2' hydroxyl on each ribose would create a Van der Waals clash. But the question remained. Could the molecule form any kind of double helix?

Alexander Rich, and the structure of ApU RNA solved at O.8 ￅngstroms. Courtesy of Alexander Rich

RNA was an open question at the time. The 2' hydroxyl represented a potential bifurcation point such that it might exist with branches, rather than linear strands.² So, Jim and I wrote to colleagues asking for purified RNA samples of any kind. The work necessitated some ingenuity. The RNA samples were moistened to a sticky glob on a slide affixed to a microscope stage, and a rod attached to the objective lens was used to carefully pull out glassy strands of RNA fibers.

I must have used this makeshift micromanipulator hundreds of times, and the X-ray diffraction patterns were always the same: fuzzy, inconclusive.³ In 1954, my fellowship ended and I joined the United States Public Health Service in part to keep from being drafted. But as I started my lab at the National Institutes of Health, I had doubts that I could ever produce compelling discoveries. My inability to get good RNA data wasn't helping.

In 1953 Francis Crick had told me about a new rotating anode built in his lab that produced a high-intensity X-ray beam. He had said that if ever my RNA diffraction patterns started looking interesting, I should come over. Early in 1955, Marianne Grunberg-Manago and Severo Ochoa described the polynucleotide phosphorylase enzyme. Feed this enzyme ribonucleoside diphosphates and it produces RNA chains without using a template.⁴ Fibers from these chains began to yield better diffraction patterns.

So, in mid-1955 I went to England, and the experience was transformative. Crick invited me to stay in his home - an invitation I possibly abused by remaining there more than six months. At breakfast one morning we were talking about a new publication of a glycine polymer for which the authors could not deduce the structure. We talked over toast and jam and decided that with physical models we might be able to solve the structure. We rushed off to the Cavendish lab and were essentially finished and even ready to publish before that pleasant English ritual: afternoon tea.

Our morning's work on a new conformation for polyglycine chains paved the way for deducing the triple stranded structure of collagen, a discovery that provided me with considerable self-assurance, as well as proving to Crick that he could make discoveries beyond the DNA structure. Thus, I ended my extended stay and returned to the NIH rejuvenated. Although we hadn't solved the RNA structure, it did not look as formidable.

David Davies, whom I had known at CalTech, joined me at NIH to work on the RNA problem. Together, we created a copolymer with polynucleotide phosphorylase, mixing adenine and uracil residues. The diffraction patterns were the same as in natural RNA - fuzzy, but suggesting at least that native RNA did indeed exist as linear strands rather than branched. What we didn't know at the time was that many of those strands were doubling up on themselves, forming tiny hairpins that showed through in the diffraction pattern.

While examinations of DNA samples revealed equal base ratios, G to C and A to T, the natural RNA we looked at tended to have unequal base ratios. It was an observation that argued against a native double-helix structure. However, with the synthetic RNA polymer, we could control the base ratios, and perhaps create conditions favorable for helix formation. Our copolymer didn't do this, so we tried something that seemed very implausible. We created separate polymers, polyadenylic acid and polyuridylic acid, and added the two together.

There was no reason to believe that this would produce anything, but as soon as we mixed the two, the viscosity of the solution increased noticeably, and the optical density at 260 mm dropped. And the fibers we pulled offered more resistance. At the same time, Robert Warner in Ochoa's department also reported the drop in optical density on mixing.⁵ While I was more focused on what we would see in the X-ray pattern, what happened in that tiny droplet was arguably more remarkable. Without any energetic input we had watched two negatively charged molecules spontaneously zip together. Though there wasn't even a name for it at the time, Davies and I had observed hybridization.

The fibers we drew produced a well-oriented X-ray diffraction pattern with a characteristically helical distribution. Although it had some similarities to the DNA diffraction pattern, there were significant differences. The molecule was 6 ￅngstroms wider in diameter than B-DNA, and the diffraction pattern was less affected by changes in relative humidity.

We wrote a short note reporting these results and sent it to The Journal of the American Chemical Society early in June 1956.⁶ In this note we concluded for the first time that it was possible for an RNA backbone to assume a configuration not unlike that found in DNA, and suggested that this might be the form in which RNA carries out molecular duplication in viruses known to contain only RNA. Finally, we pointed out that this method for forming a two-stranded helical molecule by simply mixing two substances could be used for a variety of studies.

Some colleagues were shocked. I had stopped Herman Kalckar in the long corridor at NIH to tell him we had discovered that polyribo A and polyribo U combined spontaneously to make a double helix. The Danish biochemist had to hold on to his pipe as he responded, "You mean without an enzyme?" Corralling these long tangles into neatly ordered helices worked against entropy. And though it was completely reproducible, some scientists were still skeptical, but not all. After presenting the material at a meeting several months later, Julian Huxley, the English scientist and writer, came up and congratulated me for having discovered "molecular sex."

As is true today, RNA turned up more surprises. In 1957, together with Gary Felsenfeld we found that, on adding magnesium ions, the double-stranded poly A-poly U helix could take on a third strand and become a triplex by adding poly U.⁷ This addition did not increase the diameter of the helix, and we interpreted it as entering into the major groove where the second uracil residue could make two hydrogen bonds to the adenine of the original duplex using the imidazole nitrogen and the amino group. Two years later Karst Hoogsteen found this type of bonding in a single-crystal X-ray analysis of 9-methyl adenine and 1-methyl thymine, and it is called Hoogsteen pairing.

Work on these molecules continued, but it wasn't until 1962 that it was realized that a well oriented RNA double-helix fiber-diffraction pattern had many similarities to the pattern of the dehydrated A form of DNA.

The spectra of polyribo adenylic acid, and polydeoxy thymidylic acid at a 1:1 mixture showing hypochromism at 260 mﾵ due to formation of an RNA-DNA hybrid helix (from reference 13).

In those early years most of the X-ray diffraction analysis was taken from oriented fibers that have an advantage for visualizing repetitive features of certain helical molecules. They have a profound disadvantage, however, in that they do not produce enough experimental data to fix the position of all the atoms in space. Consequently, the detailed nature of the RNA helix was not clear, as this data could not "prove" a structure.

In the 1960s as co-crystal studies involving purines and pyrimidines accumulated, an interesting pattern began to emerge. Whereas, all the single-crystal structures of complexes containing G and C derivatives formed Watson-Crick base pairs, the A and U (or T) derivatives always crystallized with Hoogsteen base pairing. It seemed that this pairing was favorable, and indeed a Hoogsteen formulation of the DNA double helix was suggested.⁸

In 1973, synthetic nucleic acids were not available. However, my students and I found that we could crystallize helical fragments of RNA. We reported the crystal structures of GpC and ApU, both forming fragments of an RNA double helix and both solved at 0.8 ￅngstrom resolution.^9,10 These structures showed unequivocally that Watson-Crick pairing was found when the molecules organized in an RNA double helix. The results elicited a call from Watson in which he told me that, upon reading preprints of our manuscripts, he'd had his first good night's sleep in twenty years.

The intricate structures of folded RNA molecules enable their many biological roles. Indeed, double-stranded RNA provides a key structural element, the "girders," for building complex structures, as seen in tRNA molecules, ribozymes, or the enormous ribosomal scaffolding. The RNA double helix also appears as a key component in many biological systems. It plays a central role in stimulating the interferon response of virus-infected cells. The wide distribution of protein domains that have specific double-stranded RNA-binding motifs testifies to the ubiquity of the conformation.

Both RNAi and microRNA silencing have at their core short segments of double-stranded RNA that carry out their regulatory roles. These systems are associated with elaborate enzymatic machinery, which tailors precursor RNAs in defined ways and then ultimately cuts out well-defined, short segments of duplex RNA to carry out downstream regulatory activities that rely on sequence-specific hybridization. In this regard, it is interesting that in a book chapter published in 1961, just before the discovery of messenger RNA, I speculated that sequences complementary to the "operative" (i.e., messenger) strand might act to control or regulate protein synthesis by forming segments of double-stranded RNA.¹¹

Most remarkable about the discoveries in our short 1956 letter to The Journal of the American Chemical Society is that these dual features of RNA - hybridization and the formation of double helices - seem so obvious now. The simple hybridization reaction has made possible most of the molecular biological revolution. Nevertheless, further developments along the way were needed to flesh out the picture.

That RNA and DNA would hybridize together in a manner similar to our poly A-poly U helices was no certain matter, as the structural differences between RNA and DNA helical structures cast some doubt on the possibility.

An opportunity to test this arose in 1960 with the availability of short segments of polydeoxythymidylate, which Gobind Khorana and colleagues chemically synthesized ¹² and kindly made available. I fractionated the longer ones and showed that a 1:1 helix formed, with one strand of poly-rA and one strand of poly-dT as demonstrated by hypochromism.¹³ (see figure above) This hybrid helix served as the basis for understanding information transfer between DNA and RNA, as in RNA polymerase or even reverse transcriptase. It still serves as the method for isolating messenger RNA using their poly-A tails with immobilized oligo dTs.

In 1960 Julius Marmur and Paul Doty showed that incubating denatured DNA at a temperature below the melting temperature provided an environment where DNA molecules could hybridize.¹⁴ The interaction showed sequence-specific pairing of complex sequences in contrast to our hybridization of synthetic polynucleotides. This crucial advance was also initially met with skepticism by some researchers who felt it unlikely that the molecules would be able to "find themselves."

One year later, Ben Hall and Sol Spiegelman combined the polynucleotide hybrid helix work with the annealing technique and showed that they could anneal bacteriophage T2 DNA with T2 RNA from infected cells.¹⁵ This put all the elements in place for the rapid employment of hybridization techniques to explore biological systems.

Southern and Northern blotting, PCR, in situ hybridization, microarray analysis, and various sequencing technologies ensued. So many technologies rely on the principle that perhaps it's no surprise that the phenomenon is taken for granted.

The necessary discoveries that underlie today's technologies were, like most worthy scientific endeavors, the products of perseverance and serendipity, and they were rightly the targets of much initial skepticism and scrutiny. Nevertheless, the idea that our tiny publication - just over 500 words, figureless, and barely filling a column in the journal's Letters section - might have played a foundational role in biology and biotechnology's phenomenal progress over the past 50 years is both remarkable and singular.

RNA is not likely finished serving up surprises. The regulatory roles of RNAi, microRNAs, and other species yet to be discovered will expand enormously and may come to dominate many aspects of biology. Understanding the structural proclivities of nucleic acids will continue to be important in deciphering the complex network of interactions that RNA controls. Likewise, our use of hybridization to understand biology and harness it for both biological and nonbiological tasks will continue to develop. Although more than 50 years have passed since chemistry started to inform genetics, perhaps the best discoveries are yet to come.