A new paper  published in Nature experimentally explores the mechanisms that may generally underpin Dollo's law. With a flurry of recent papers, which exclusively rely on phylogenetic approaches (often demonstrably flawed) to challenge one of the soundest principles of character evolution , this original approach to an age-old problem is a stand-out.
The group, led by Joe Thornton at the University of Oregon, previously dissected the genetic basis of a shift in function of the glucocorticoid receptor (a DNA transcription factor), using statistical phylogenetics, functional assays, and X-ray crystallography . They found that a once-promiscuous binding receptor--activated by aldosterone, deoxycorticosterone, and to the lesser extent, cortisol--in the common ancestor of jawed vertebrates, became highly cortisol-specific in the lineage leading to ray-finned fishes and tetrapods.
The key substitutions took place in the receptor ligand-binding domain, with two changes necessary and sufficient to shift preference, and several other changes that restrict receptor sensitivity further and stabilize its structure . The experimental reversal of these same changes in the derived receptor surprisingly renders it useless. It cannot activate transcription in the presence of any ligand. In other words, the forward changes cannot be reversed through the same direct path, because it is unsuccessful in the derived background.
Bridgham and colleagues have now identified the group of mutations that comprise this background, the order in which the groups of changes would have to proceed for a reversal of function to take place, and reveal the underlying mechanism.
The initial sequence of mutations fixed in ligand preference switching is largely unconstrained. Any number of them could improve cortisol binding. However, once a particular evolutionary path is taken, it is unlikely that selection can replay the highly specific sequence of events to revert to the ancestral state.
The process is akin to an evolutionary ratchet, making the receptor intolerant of ancestral features present immediately preceding the present form.
The paper convincingly shows that the initial stages of the reversal must follow a very particular sequence, as well as take steps that do not yield a measurable advantage in either the derived or ancestral function. The reversal paths would thus have to pass through adaptive troughs, even if the selective environment favored the ancestral form. Paraphrasing Günther Wagner in the accompanying news commentary, the receptor's path effectively burns bridges that it just went over, thus making the reverse path exceedingly implausible.
 Bridgham, J. T., Ortlund, E. A. & J. W. Thornton. 2006. An epistatic ratchet constrains the direction of glucocorticoid receptor evolution. Nature 461, 515-519.
 Gould, S. J. 1970. Dollo on Dollo's law: irreversibility and the status of evolutionary laws. J. Hist. Biol. 3:189–212.
 Bridgham, J. T., Carroll, S. M. & J. W. Thornton. 2006. Evolution of hormone-receptor complexity by molecular exploitation. Science 312: 97–101; Ortlund, E. A., Bridgham, J. T., Redinbo, M. R. & J. W. Thornton. 2007. Crystal structure of an ancient protein: evolution by conformational epistasis. Science 317:1544–1548.