Although every AUG (and GUG) triplet potentially can initiate protein synthesis, normally this does not occur indiscriminately, because certain structural features of mRNA in the neighbourhood of the initiator codon are needed for binding ribosomes and for the specific initiation of protein synthesis. It has been proposed (Shine & Dalgarno, 1974;Steitz & Jakes, 1975) that a key event in this specific recognition in prokaryotic cells is the interaction between the 3'-terminal sequence of the 16S ribosomal-RNA and a purine-rich sequence preceding the initiator triplet by a somewhat variable distance. This type of recognition signal by itself may be adequate in certain cases, where several contiguous bases of the leader sequence of mRNA can hydrogen-bond to the 3'-end of 16S rRNA. However, in most cases this hypothesis is less than adequate.
In this paper it is shown that a characteristic and important feature of prokaryotic initiation sites is a "hairpin sequence" found between the initiation codon and the sequence which is believed to base-pair with 168 rRNA. Most prokaryotic mRNAs sequenced to date can form a hairpin structure adjacent to the 5' -side of the initiating codon. The 3 ' -end of 168 rRNA most frequently can base pair to the repeated 5'-shoulder of these hairpins. The "initiation hairpins" of bacterial origin are generally longer than those of bacteriophages, and they contain an alternative Shine-Dalgarno sequence in their non-hydrogen-bonded loops. It is suggested that in the process of initiation, the hairpin sequence is accomodated in a special cleft of the 30S ribosome, so that the initiation codon resides on the surface of the ribosome, just at the entrance to this cleft, where interaction with initiating-tRNA would take place. The Shine-Dalgarno interactions assist in recognition, orientation and stabilization of the hairpin sequence interaction.