Cracking the genetic code was one of the most significant scientific milestones of the last half of the twentieth century. Today, it would be a simple matter of comparing the sequences of nucleic acids and their corresponding proteins. But in the early 1960s, nucleic acid sequencing had not yet been invented, and although protein sequencing methods were firmly established, the process was laborious. Given the state of sequencing methodology, it is surprising that new and ingenious ways of studying and deciphering the code in its entirety were developed at all. Here we briefly review the major experimental strategies that enabled investigators to crack the code.

First, Marshall Nirenberg and Heinrich Matthaei made a simple but remarkable discovery that set in motion the cracking of the genetic code: they found that they could use the enzyme polynucleotide phosphorylase to synthesize RNA templates that would code for protein polymers in cell extracts of E. coli. To prepare the cell extracts, endogenous mRNA had to be removed, to allow the translation machinery to work on the synthetic RNA templates. To do this, the extracts were preincubated so that the endogenous ribonuclease (RNase) activity would destroy all existing mRNA, and deoxyribonuclease (DNase) was added to the extracts to prevent further mRNA production. With these treatments, protein synthesis in the cell extracts (or “translation extracts”) was dependent on the addition of exogenous synthetic RNA.

Polynucleotide phosphorylase does not require a template; it uses ribonucleoside diphosphates (NDPs) to make random polymers of RNA:

The intracellular role of polynucleotide phosphorylase is to catalyze the reverse reaction to degrade RNA, using inorganic phosphate to yield NDPs. In vitro, the enzyme can be made to synthesize RNA by the addition of excess NDPs, which at sufficient concentration are polymerized. However, because polynucleotide phosphorylase is not template-directed, the sequence of the RNA polymer is random.

Using UDP as the only ribonucleoside diphosphate substrate, polynucleotide phosphorylase catalyzes the synthesis of poly(U) RNA. To determine what type of protein synthesis poly(U) directs, Nirenberg and Matthaei added poly(U) to reaction mixtures with the E. coli translation extracts and amino acid mixtures, but using only one radioactively labeled amino acid (Figure 17-14). At the end of the incubation, the reaction mixtures were treated with acid, which precipitates protein polymers but leaves free amino acids in solution. Precipitates were collected and counted in a liquid scintillation counter, which measures radioactivity in counts per minute (cpm). The results showed that poly(U) directed the synthesis of polyphenylalanine, poly(Phe), and therefore the codon for phenylalanine must be UUU. The other homopolymers were synthesized and tested by similar methods. Poly(A) specified the synthesis of poly(Lys), identifying AAA as a codon for lysine. Poly(C) resulted in the production of poly(Pro), and thus CCC is a codon for proline. Unfortunately, poly(G) forms intramolecular hydrogen-bonded structures that prevent its use in these reactions.

601

The use of polynucleotide phosphorylase was extended to copolymers, RNAs containing more than one type of nucleotide. As an example of this type of analysis, Table 17-2 shows the results obtained by Severo Ochoa’s group, using ADP and CDP in a 5:1 molar ratio. Polynucleotide phosphorylase incorporates NDPs into RNA randomly, so these conditions produced RNA with five times more A than C. All possible codons containing A and C were generated in the random RNA copolymer, in the following distribution (setting AAA at 100): AAA (100), A₂C (60), AC₂ (12), and CCC (0.8). (Note the use of A₂C and AC₂ to indicate that the order of nucleotides is unknown.) Use of this heterogeneous RNA copolymer in translation extracts directed the synthesis of polypeptides containing asparagine, glutamine, histidine, lysine, proline, and threonine. From the relative amounts of these amino acids in the precipitated protein product, the investigators could deduce the respective codon compositions specifying each amino acid. No amino acid was observed at the low frequency expected of a CCC codon (0.8) relative to an AAA codon (100). But proline was observed to be incorporated at a relative frequency of 4.7, which is close to the expected frequency for two codons, CCC and AC₂.

The researchers could infer from this result that proline is specified by two codons having the compositions AC₂ and CCC, reflecting the degenerate nature of the genetic code. The result was supported by experiments using poly(C), which identified CCC as a codon specifying proline. The other codon, AC₂, could have the sequence ACC, CCA, or CAC. Because polynucleotide phosphorylase polymerizes NDPs randomly, one can assign only codon compositions from these results, not codon sequences (except, of course, for codons with only one type of nucleotide).

602

Numerous experiments were performed using random RNA copolymers, and in this way the nucleotide composition of about 40 codons could be assigned to particular amino acids. However, to identify the nucleotide sequence of codons, RNA molecules of defined sequence were required.

Chemical synthesis of short nucleic acids of specific sequence was needed to define the complete genetic code. In 1964, Nirenberg and his coworkers at the National Institutes of Health discovered a novel method to identify an amino acid associated with a short synthetic codon. They found that during protein synthesis in E. coli (using translation extracts and ¹⁴C-labeled amino acids), the nascent protein (the protein being synthesized) stayed attached to ribosomes bound to the mRNA and could be separated from unbound amino acids (Figure 17-15a). However, the [¹⁴C]aminoacyl-tRNA–ribosome–RNA codon complex could not be precipitated by acid treatment, because the complex is dissociated by acid. The researchers developed a filtration method that took advantage of the fact that a nitrocellulose filter binds protein but not RNA (including tRNA, charged or uncharged). More importantly, they found that even RNA polymers as short as a trinucleotide could be used in these experiments. Maxine Singer, also working at the NIH, had developed the capability of synthesizing RNAs of defined sequence. This provided an advance that allowed assignment of several additional codons for amino acids and confirmed assignments made by the precipitation method.

Results from Nirenberg’s initial study, using three different [¹⁴C]aminoacyl-tRNAs and the synthetic trinucleotides UUU, CCC, or AAA, are shown in Figure 17-15b. Similar studies of all possible triplet RNA sequences identified approximately 50 codon assignments for specific amino acids. However, not all triplet RNA sequences induced tight binding of their corresponding [¹⁴C]aminoacyl-tRNA to the ribosome, and another technique was required to completely crack the genetic code.

Another breakthrough came from H. Gobind Khorana, who developed chemical methods to synthesize short polyribonucleotides with defined sequences of two, three, or four nucleotide repeats. These short RNAs were then amplified by polymerases to produce long polymers of defined repeating sequence. Results obtained with these long RNA polymers, combined with the earlier results from trinucleotide-induced ribosome binding and the use of random RNA polymers, identified the amino acids specified by all of the codons. For example, the alternating copolymer (AC)_n contains the alternating codons ACA and CAC. The polypeptide synthesized on this RNA copolymer contains equal amounts of threonine and histidine. Parallel studies using random RNA copolymers identified the nucleotide composition of a histidine codon as AC₂, so CAC must code for histidine and ACA for threonine. Examples of results obtained with di-, tri-, and tetranucleotide sequences are shown in Table 17-3.

603

Repeating trinucleotide sequences yield three types of homopolymeric peptides. For example, for a repeating sequence of UUCs, the mRNA triplets can be read as repeating UUC, or UCU, or CUU codons. Poly(UUC) produced poly(Phe), poly(Lys), and poly(Leu). The ribosome-binding assay showed that UCU encodes serine and CUU encodes leucine. Therefore, the UUC codon can be assigned to phenylalanine. Poly(UAG) yielded only two homopolymers, poly(Ser) and poly(Val), because the UAG reading frame is a string of nonsense codons that yields no product. The repeating tetranucleotide poly(UAUC) produced a repeating tetrapeptide of Tyr-Leu-Ser-Ile. Knowing the codon assignments for the four amino acids in the repeating tetranucleotide also confirmed that mRNA is read in the 5′→3′ direction. If the RNA had been read in the 3′→5′ direction, the repeating tetrapeptide would have been Ile-Ser-Leu-Tyr.

Consolidation of the results from many different experiments permitted the assignment of 61 of the 64 possible codons. The other three were identified as termination codons, in part because of the results obtained with synthetic RNAs in which the codons disrupted the amino acid coding patterns. Meanings for all the triplet codons were firmly established by 1966, and they have been verified numerous times, in many different ways. In retrospect, the experiments of Nirenberg and Khorana that identified codons in translation extracts should not have worked in the absence of initiation codons. Serendipitously, with the high magnesium concentration used in their in vitro experiments, the normal initiation requirement for protein synthesis was relaxed.

604

We now know the sequences of entire genomes of organisms ranging from bacteria to humans, and this enormous body of information confirms that the genetic code determined in the translation extract experiments is indeed interpreted in the same way in all living organisms. Even in the 1960s, however, there was evidence that the genetic code was the same in cells as in cell extracts.

By the mid-1960s, more than 100 mutant proteins resulting from single-nucleotide substitutions had been studied, and in all cases the incorrect amino acid in the mutant, relative to the wild type, could be accounted for by a change in a single nucleotide, based on the genetic code. For example, the disease sickle-cell anemia was known to be caused by a single amino acid change, substituting glutamate for valine in human hemoglobin (see Highlight 2-1). Only one nucleotide change is needed to alter the AGU codon specifying glutamate to the UGU codon for valine: these codons specify the same amino acids in extracts of E. coli as they do in blood cells. An astute demonstration that the genetic code in the cell is the same as the code determined in cell extracts is described in the How We Know section at the end of this chapter.