Perspective

No biochemistry textbook is complete without a φ,ψ-plot of the alanine dipeptide, or more precisely, the compound Cα-CO-NH−CαHR−CO-NH-Cα, which has two degrees of backbone freedom like a dipeptide (Ramachandran et al. 1963). This plot ranks alongside the double helix and the α-helix among fundamentals of structural biochemistry. The plot is a compact and accessible representation of a profound idea, one that has thoroughly conditioned our thinking about the structure of proteins.

Sadly, G.N. Ramachandran (GNR, as he was known in India) died on April 7 at the age of 79. His long-time colleague and friend, C. Ramakrishnan, has written an obituary for Protein Science. Interested readers should also see the recent perspective by Richard Lavery (2000) who was a postdoc with Ramachandran in the mid 1970s. My remarks here are limited to the φ,ψ-plot and its implications.

The early 1950s were an exciting time in structural biochemistry. In 1952, J.D. Bernal visited India and urged Ramachandran to work on the structure of collagen (see Sarma 1998 for an account of this meeting). The Pauling-Corey-Branson model of the α-helix (Pauling et al. 1951) had just been published and was followed almost immediately by Perutz's dramatic experimental confirmation (Perutz 1951).

Ramachandran pursued Bernal's suggestion enthusiastically, and less than two years later, the paper describing the Ramachandran-Kartha triple-stranded, coiled-coil structure was published (Ramachandran and Kartha 1954). The structure of DNA had just appeared the preceding year (Watson and Crick 1953). Certain nonbonded distances were too short in the Ramachandran-Kartha collagen structure, as noted by Rich and Crick (1955), but given the model-building facilities available to Ramachandran and Kartha at the time, this is hardly surprising. As Sarma (1998) points out,

"Computers had not arrived in Indian science, and they even lacked sophisticated model-building facilities in Madras. In fact, Kartha measured the bond distances in their crude models using pieces of string or the ribs of coconut leaves!"

Ramachandran appears to have taken these criticisms much to heart, and his response was a testament to scientific creativity of the first order. What started in criticism emerged, some years later, as an exhaustive representation of dipeptide stereochemistry—the famous φ,ψ-plot. The full impact of the plot was not immediately apparent. It soon became so after John Edsall invited Ramachandran to contribute a review to Advances in Protein Chemistry (Ramachandran and Sasisekharan 1968). In this remarkable review, Ramachandran and Sasisekharan anticipated many directions the field would take for years to come. The fact that the φ,ψ-plot was based only on the hard sphere (i.e., the repulsive part of the Lennard-Jones potential) had led some to underestimate the generality of this work. In this regard, Fred Richards (1977) commented dryly,

"For chemically bonded atoms the distribution is not spherically symmetric nor are the properties of such atoms isotropic. In spite of all this, the use of the hard sphere model has a venerable history and an enviable record in explaining a variety of different observable properties. As applied specifically to proteins, the work of G.N. Ramachandran and his colleagues has provided much of our present thinking about permissible peptide chain conformations."

It is worth noting that similar stereochemical ideas can also be applied to the analysis of RNA conformation, despite the fact that the monomer unit (i.e., a mononucleotide) in this case has greater backbone freedom than a dipeptide. Pioneering work of Sasisekharan and Lakshminarayanan (1969) and Sundaralingam (1969) has been pursued by several groups, most recently by Duarte and Pyle (1998) and Murthy et al. (1999).

The φ,ψ-plot is a model of physical reality, and its validity needed to be tested by experiment. That test was passed with flying colors as an increasing number of experimentally determined protein structures was solved. Now, of course, it is theory that is used to validate experiments, not the reverse, in programs like PROCHECK (Laskowski et al. 1993). Like Kepler's laws, the theory accounts for the data satisfactorily so that is has become synonymous with reality.

Use of the φ,ψ-plot for validation of experimental structures is so commonplace that it tends to overshadow some of the deeper implications of the plot. For residues other than glycine or proline, sterically allowed conformers fall almost exclusively within two discrete islands, one near φ,ψ = −60°, −40°, the other near φ,ψ = −120°, +130°. Repetition of the backbone dihedral angles from the first island results in an α-helix, whereas repetition of values from the other island results in a β-strand. At the level of the dipeptide, protein structure is essentially digital, because the two islands are discrete.

It has been thought that the conformation of each φ,ψ-pair in a polyalanine peptide is independent of its neighbors (Flory 1969). If so, a chain of length N, in which each residue can occupy either of two islands, can visit 2^N conformers. While formally true within a persistence length, most mixed conformers are scarcely populated, because the chain tends to clash with itself whenever it adopts them (Pappu et al. 2000). Consequently, almost all segments are either extended or helical; sterics inhibit structural hybrids. This conclusion is borne out by the familiar observation that known protein structures consist of isodirectional segments—either helices or strands—interconnected by turns and loops. Chimeric segments fashioned partly from helix and partly from strand are seldom seen. At root, this dichotomy originates in sterics, and it is built into the covalent backbone structure at the level of the alanine dipeptide.

The steric dichotomy between helix and strand populations in proteins is one of nature's deep organizing principles. Many biological phenomena at all levels are discrete, ranging from the hydrophobic effect (e.g., oil vs. water) to genetic phenotypes (brown eyes vs. blue eyes). Such examples of digital assortment, in which phenomena self-classify into distinct, unmixed states, are fundamental to reliable recognition, both microscopic and macroscopic. Perhaps the most far-reaching implication of the Watson-Crick structure (Watson and Crick 1953) is the realization that DNA is a digital molecule. Protein molecules are digital too: Helix-favoring conditions are necessarily strand-disfavoring, and conversely. Above all, we owe this discovery to G.N. Ramachandran and the stereochemical analysis of the alanine dipeptide.