Ramachandran redux

Protein molecules are indispensable to life processes, ranging from catalysis of reactions to transport, signaling, and shaping of cells (1). Despite their intricate architecture, revealed in thousands of 3D structures stored in the Protein Data Bank (2), protein structures rest on a surprisingly small set of principles (1). Perhaps most fundamental of all is the fact that the amide bond (Fig. 1A) is planar, so that only two dihedral angles, denoted by Φ and Ψ (Fig. 1A), define the conformation of the bond linking adjacent amino acids. Following leads from their studies of the structure of collagen, the predominant protein in humans, the crystallographer G. N. Ramachandran and his colleagues first used a 2D diagram to depict the geometry of a dipeptide—two amino acids together with the intervening amide bond (3). They plotted values of one of the angles along the x axis and the second along the y axis, as shown in Fig. 1B. Using the few peptide structures then available, they could see that the angles clustered in only a few sections of the map. Model building led them to conclude that most values of the two angles were inaccessible owing to collisions between atoms of the backbone (Fig. 1B). It is hard to overstate the seminal impact of this representation, which has guided the thinking of protein chemists and structural biologists for many decades, as pointed out by Porter and Rose (4). The article in PNAS by Porter and Rose (4) reports on a unique and quite surprising aspect of the structure of proteins. They have refined the classic Ramachandran plot to introduce the effect of hydrogen bonds (H-bonds) explicitly and are thus able to identify a disfavored bridge region linking the extended β/polyproline II (PPII) domain with the α_R-basin.

Fig. 1.

(A) Definition of Φ and Ψ angles. (B) A contoured population map of the protein coil library (www.roselab.jhu.edu/coil), adapted from Perskie et al. (6). Populated regions lie largely within sterically allowed boundaries for the classic alanyl dipeptide; outliers correspond to glycines. The abridged region (in semitransparent gray) covers a heavily populated basin that corresponds to the i+2 residue of either a type I or type II′ turn. Consequently, the population of type I and type II′ turns would be depleted under unfolding conditions, in which intramolecular hydrogen bonds are disfavored.

H-bonds play a vital role in protein architecture (1). The electrons of the hydrogen atoms linked to oxygen and nitrogen, as in water or the peptide backbone, are polarized and form a bond that tethers the atoms together, in combinations such as O-H•••O or C = O•••H-N. Although individually weak, the total contribution of H-bonds is significant: they explain the stability of ice, for example, and the unusual properties of liquid water. In proteins H-bonds play a critical role in providing specificity to α-helices and β-sheets. Although a given H-bond is easy to break, unsatisfied H-bonds can be extremely costly in energy. Earlier calculations by Rose's group pointed out that many cases in which unsatisfied H-bonds have been reported in structures within the Protein Data Bank are likely to be artifactual (5). The original Ramachandran plot distinguished three major clusters of Φ and Ψ angle values, representing three predominant folds of the backbone in proteins: the α-helix, the PPII structure, and extended sheet-like β-structure (Fig. 1). In addition, there are bridge regions defined by the area around Φ = −80 and Ψ = 30. These bridges consist mostly of turns that enable compact structures to form. Under unfolding conditions, this region is disfavored owing to steric clashes if hydration of solvent molecules is considered; however, under native conditions, Porter and Rose find that this “forbidden” region can be occupied.

What insights do we gain from this exercise? Structures with angles in the domain indicated in the shaded region of Fig. 1B share the property that they are predicted to be solvent inaccessible (or to cause steric clashes if an H-bond is formed with solvent), imposing a severe energetic penalty that leads the region to be forbidden in simulations (4). However, when they compare this region with structures in the data bank, they find that it is occupied to a significant extent. Why? The answer they provide is that in this region intramolecular H-bonds from residues further away in the sequence replace solvent hydration interactions and remove the unfavorable steric barrier. This effect makes occupation of the otherwise forbidden region possible. The region forms a natural bridge linking the PPII domain to the α-basin below. Given the extreme difficulty in determing Φ and Ψ angles in unfolded proteins, the existence of the revised Ramachandran plot with abridged bridge is hard to prove experimentally. Nevertheless the revised plot points to the importance of steric restriction and/or hydration effects in unfolded proteins. The authors propose a low-energy pathway between two basins that converts the left-handed PPII structure to the right-handed α-helix, via an intermediate structure known as a γ-turn. This route facilitates switching the handedness of the backbone geometry from left to right without the ancillary cost of breaking H-bonds. They describe the motion as a molecular hand-off, a low-barrier corridor between unfolded and folded structures as proteins acquire their native states.

The study by Porter and Rose relies on traditional interactions that stabilize protein structure: Lennard-Jones potentials (steric clashes) and H-bonds. A new force, n→π* interaction, emerges from work by Raines and colleagues (7), also recognized by Fufezan (8). According to molecular orbital theory, hyperconjugation connects a variety of seemingly unrelated phenomena, including the preference for the staggered structure of ethane, the gauche effect, the anomeric effect, and the reactivity of organic molecules in many reactions (9, 10). The n→π* interaction is defined by n→π* hyperconjugation; H-bonding reflects a different type of hyperconjugation, that is, n→σ*. In considering the electron orbitals within the peptide backbone, Raines and colleagues discern a favorable interaction between electrons on one C = O group and the π* orbital of its neighbor (Fig. 2). The γ- and other turn intermediates between the PPII and α-helical domains envisaged by Porter and Rose are predicted to have weaker n→π* interactions relative to those in α and PPII (Fig. 2). The switch from n→π*(in PPII) to n→σ* interactions (in turn intermediates) and thence to n-π* plus n→σ* interactions (in α-helix) can be thought of as one manifestation of orbital rearrangements that accompany a low-energy barrier. Switching from n→π* interactions (in PPII) to n→σ* interactions (in turn intermediates) can be thought of as one manifestation of the orbital accompanied by a relatively low-energy barrier. Development of a consistent picture of the peptide backbone merging these two perspectives should yield further insights into the structures of the backbone in proteins and provide us a deeper understanding of underlying forces for different conformational states, as well as the transitions from one to the other upon changing conditions.

Fig. 2.

Two regions of the Ramachandran plot that are predicted to allow significant n→π* interactions by natural bond orbital analysis. Left: The n→π* interaction in PPII; Right: the n→π* natural orbital overlap in an α-helix. Drawing courtesy of R. Raines and A. Choudhary, Department of Chemistry, University of Wisconsin, Madison, WI.

What implications do these considerations have for protein energetics and conformation of the backbone in unfolded proteins (11)? The clear part of the emerging picture is the following: n→π* interactions favor α and PPII; dipole–dipole interactions favor β and PPII; longer-range backbone H-bonds favor α and β conformations. More challenging questions remain to be answered: how to predict the net enthalpy including both favorable contributions and unfavorable penalties from different forces and to quantify residue- and conformation-specific hydration enthalpy and associated entropy changes (12). Given recent developments in detailed folding simulations (13), such a prospect may not be far off if the missing pieces can be incorporated into improved force fields.