A full theory that relates protein primary structure to the speed and accuracy of its folding is still not available. The hypothesis of hierarchical protein folding, which posits that the major elements of secondary structure form independently and assemble into a mature fold, has been examined over the years, but has given way to other explanations, which appear to be better compatible with the experimental evidence. One class of current theories of folding relies on the existence of a molten-globule transition state, where a high fraction of amino acid residues is found within helices and strands, and at the same time there is no tightly packed interior core [5]. A contrasting “zipping-and-assembly” model of protein folding does not demand a molten-globule intermediate [6], but suggests that small elements of secondary structure, such as beta-hairpins and alpha-helical turns, form at many independent sites along the chain and then grow by extension (zip) or coalesce (assemble) with other such elements.
Despite the differences between these and other views on protein folding mechanisms, all of them require some extent of secondary structure to emerge early in the process of formation of the native protein structure, typically at a local scale (see also [7, 8]). Studies examining the folding of short peptides, where distinct steps of structure formation could be monitored directly, suggested that the initiation (“nucleation”) of the first alpha-helical turn is the rate-limiting step in the formation of an alpha-helix (on average, it takes about as much time to form the first turn of a helix as all the other turns combined). Studies of beta-sheet stability have similarly indicated that the formation of beta-turns and beta-hairpins is likely to be the rate-limiting step in sheet assembly [5].
From a physical point of view, these turns, twists and hairpins are precisely the types of structural elements that would be induced if a torque force was applied to the longitudinal axis of a long thin cylinder between two points that were fully or partially restricted from rotating. In fact, just one distal point has to be fixed, if the torque is applied at another point that is itself stabilized in space. Torsion of a linear biopolymer is a notion familiar to molecular biologists from studies of the topology of the DNA double helix. It is well known that if the ends of a double-stranded DNA molecule are covalently linked to each other, or if they are restricted in mobility by interaction with other molecules, then the torque applied to the main chain of the molecule will result in negative or positive supercoiling [9]. Sometimes omitted from this account is a more general rule, i.e., that a twist of any string, such as a single-strand linear polymer, will also induce secondary structure. This has been studied more recently with single-stranded DNA [10], and there is no reason why similar forces applied to a polypeptide should not produce qualitatively similar outcomes, i.e., turns and twists of the molecule.
A quantitative physical model of protein torsion and twist, which would take into account the geometry of the chemical bonds and energetically favorable conformations within the protein main chain, the effects of chain elasticity and viscosity of the solution, as well as the molecular interactions of the side chains and the solvent molecules, is beyond the scope of this paper. Here, we would like instead to discuss the intracellular structures and processes that could result in the application of torque to a polypeptide. In order for chain twisting to be a significant component of protein folding in vivo, those twisting forces should be available during the maturation of many classes of proteins, and the process has to be supplied with external energy. Many molecular machines may interact with unfolded, partially folded or misfolded proteins in the cell in an energy-dependent manner, releasing proteins with native three-dimensional structure. These include signal recognition particles, secretion systems, chaperone systems, and protein processing modules. We think, however, that an even more universal device spends energy specifically to introduce a twist of the nascent peptide chain and thereby facilitates their subsequent folding: the ribosome itself.
Stereochemical modeling on partially solved structures of the ribosomal large subunit have predicted 30 years ago that helical twisting of the nascent peptide occurs in the ribosomal exit tunnel [11]. Recent experimental data confirm that a partially helical conformation is attained by certain peptides in the exit tunnel and exit vestibule [12–14].
A rough estimate of energy balance during the ribosomal cycle suggests that the hydrolysis of two GTP molecules and deacylation of an aminoacyl-tRNA bond liberates ~30 kcal/mol of amino acid, only a small fraction of which is consumed for positioning of the incoming charged tRNA that facilitates the peptide bond formation, whereas the rest is thought to dissipate as heat ([15], p. 159). Even if we consider other energy expenditures, such as tRNA translocation or motions of ribosome parts, the energetic needs of twisting the protein chain might be more than covered by the energy surplus in a ribosomal cycle. On the other hand, formation of an alpha helix of the length typical of globular proteins, or of a beta hairpin, may require overcoming a barrier of about only 1-5 kcal/mol of amino acid (calculated based on the data from [16]).
The structural elements of the ribosome that would be able to relay some of this energy into a torque force on the chain are not known. Another crucial question concerns the locations at which the torsion could be applied, though the peptidyl transferase center itself may be a possibility. As for the mechanisms of constraining the torsional mobility of the chain at a downstream site, it is plausible that any of the systems that are able to bind a nascent protein chain, including the aforementioned modules of maturation, sorting, and reactivation of specific subsets of proteins, can play this essential role. In addition, some of those molecular machines may be able to apply their own rotational force to their specific substrates. Cooperation between the ribosome and other machines may be particularly important for the formation of beta-hairpins, which typically bear a twist not fully explained by the current theories and may be too large to fit into the exit vestibule of the ribosome [17] (compare, however, with results in [14]).