My goal is to reveal the molecular mechanisms by which evolution has produced the remarkable diversity of genes and proteins in our bodies, each with functions exquisitely adapted to their functions in the cell. My laboratory has forged a synthesis of computational evolutionary techniques with the experimental strategies of molecular biology, biochemistry, and structural biology. We resurrect ancient proteins, test hypotheses about their structures and functions, and determine the precise effects of historical mutations that occurred long ago. In short, we are trying to bring the rigor and mechanistic focus of molecular biology to the fascinating questions of evolutionary biology.

The Evolution of Hormones and Their Receptors
Virtually everything a living cell does is driven by specific interactions between molecules—hormones and their receptors, enzymes and their substrates, transcription factors and their DNA-binding sites. Despite their biological importance, however, little is known about how these tight molecular interactions evolve. My lab studies the steroid hormone receptors, a biomedically important protein family that provides a beautiful model for the evolution of specific molecular interactions.

Steroid receptors mediate the classic biological effects of steroid hormones, including estrogen, testosterone, progesterone, the stress hormone cortisol, and the blood pressure regulator aldosterone. These key regulators of reproductive development and function, metabolism, behavior, and homeostasis are also major players in cancers of the breast and prostate. Each hormone binds its favorite receptor with remarkable specificity and affinity, changing its conformation so the receptor can then regulate expression of an array of target genes. My laboratory's goal is to understand how the steroid hormone receptors in our bodies evolved their diverse hormone specificities. We seek to discover the specific molecular events—gene duplications, individual mutations, and atom-level changes in protein structure—by which ancient hormones and receptors diversified and evolved their specific partnerships hundreds of millions of years ago.

Resurrecting Ancient Genes
We have developed a powerful new strategy called ancestral gene resurrection to test hypotheses about molecular evolution. We begin by sequencing genes and mining genomes to assemble a large database of steroid hormone receptor sequences from diverse present-day taxa. We often work on little-studied species that occupy critical positions in the tree of life, such as octopus, sea slug, sponge, lamprey, skate, and lungfish. We then use computational phylogenetics—the same techniques used to reconstruct the "tree of life" among species—to infer the relationships among genes in the receptor family. Once we know the gene family phylogeny, we can infer the protein sequence at any ancestral node in the tree by working our way from the tips back down the tree. For every possible ancestral sequence, we use a Markov model of evolution to calculate the probability that all the sequence data in present-day species would have evolved; the best estimate of the ancestral sequence is the one with the highest likelihood. We then use biochemistry to synthesize DNA coding for the ancestral protein, express it in cultured cells in the laboratory, and characterize its structure and function experimentally using molecular and biochemical assays and x-ray crystallography.

Mechanisms for Evolving a New Function
Some of our most detailed work has been on two closely related hormone receptors—the glucocorticoid receptor (GR), which is activated by the stress hormone cortisol, and the mineralocorticoid receptor (MR), which is activated primarily by aldosterone and more weakly by cortisol. These receptors evolved when a single receptor gene was duplicated, near the base of the vertebrate lineage, more than 400 million years ago. This ancient receptor was activated by both hormones, indicating that the GR got its specificity by losing the ancestral response to mineralocorticoids. To identify how this shift in function occurred, we first identified when it happened by resurrecting successive ancestral genes moving up the gene family tree. We then introduced into the resurrected ancestral gene each mutation that occurred during this interval and determined its effects on the protein's function.

We found that two substitutions together had a major effect in switching the receptor's preference from aldosterone to cortisol, but neither in isolation contributes to the shift in function; one had no apparent effect on the protein, while the other was extremely deleterious. To understand the biophysical mechanisms by which these mutations changed the receptor's hormone sensitivity, we collaborated with Eric Ortlund (Emory University) to determine the x-ray crystallographic structures of ancestral proteins before and after the shift. These atomic maps represent the first experimental structures of any ancient protein. They revealed that the deleterious mutation introduced a kink in the protein backbone, radically repositioning a helix that forms one side of the ligand-binding pocket and destabilizing the hormone-receptor complex. The other mutation occurred at a site on the repositioned helix; it introduced a polar amino acid that in its ancestral location was far from the hormone and thus had little effect on function. Together, however, the first mutation moves the second site close to an oxygen atom that is only found on cortisol; in this new position, the polar amino acid bonds to that oxygen, restoring stability to the receptor only when it is in complex with cortisol.

Although these two substitutions had a huge effect on the protein's preference for cortisol, they did not yield a protein that, like the modern GR, has zero response to aldosterone. We found that three other mutations during the same period optimized the new function, completing the shift to a cortisol-specific receptor. Remarkably, however, the ancestral protein will tolerate these mutations only if two other historical mutations are introduced first; otherwise the three optimizing changes yield a dead receptor that does not function at all. On their own, the "permissive" mutations have no apparent effect on the protein's function. The structure shows that these changes buttressed parts of the receptor that were destabilized by the three optimizing mutations. Together, the seven historical substitutions we isolated are sufficient to yield the complete GR-like function and illustrate how evolution threaded its way through a complex landscape of mutational possibilities where the order in which steps are taken has a profound effect on the pathways and outcomes that are accessible.

Evolution of Other Ancient Receptors
Our work on GR evolution represents a nearly complete mechanistic account for how a complex protein function evolved in the deep past. We are now applying the same strategy to the entire steroid receptor gene family to characterize its biodiversity and explain how all six steroid hormone receptors evolved their distinct ligand specificities.

For example, we discovered the first steroid receptors in invertebrates and found they are surprisingly diverse in function, indicating that this gene family is far older, more widespread, and evolutionarily labile than previously thought. When we resurrected the single ancestral gene from which the steroid receptor gene family evolved, we found it had the specific functions of an estrogen receptor. Because estrogens are the terminal hormones in a biosynthetic pathway that uses progesterone and testosterone as intermediates, these results indicate that the progesterone receptor and androgen receptor evolved their specificity by recruiting upstream steroids in the pathway for new roles as signaling molecules. By resurrecting successive ancestors on the gene family tree, we are zeroing in on the specific mutations by which these shifts in hormone recognition evolved.