Last week saw a momentous advance in molecular neuroscience. A research group from Portland’s Vollum Institute published in Nature several structures of the glycine receptor. This important brain protein is, among other things, a target for alcohol and other anesthetic drugs; my collaborator Erik and I have studied it in our own labs. It’s a story that not only influences our field, but also represents (I think) some interesting features of molecular neuroscience today.
Good receptor: Erik calls this protein Glyra, and I’ll stick to that shorthand. (To me it sounds like a good witch in the Wizard of Oz). Briefly, Glyra is a large protein embedded in the membranes around nerve cells, particularly in the spinal cord but also in brain. Its full name comes from its role in binding glycine, a small but important chemical (neurotransmitter) passed between cells to control electrical signaling. When Glyra binds glycine, it opens a tunnel in the protein’s core to allow ions (specifically chloride) to flow across the membrane into the cell. Generally, this influx of negative charge makes it harder for the nerve cell to fire; for this reason, Glyra is sometimes called an inhibitory receptor, as it can suppress overall electrical signaling. Drugs like alcohol enhance Glyra function, and therefore suppress nerve activity even more than usual, causing sedation or anesthesia.
High impact: The structure of a molecule like Glyra has enormous potential value. As an important target for drugs, it could aid in designing new anesthetics or other pharmaceuticals, and could help explain why some individuals respond differently to drugs than others. Glyra is closely related to several other neurotransmitter receptors in the brain, so it could also provide a framework to understand a large family of important molecules. Accordingly, various parts of this protein have been subject to functional, biochemical, physiological, and structural studies for decades.
Unresolved: However, this important receptor has so far been elusive. Proteins like Glyra need to be embedded in fatty lipid layers, like the ones that surround brain cells, in order to keep their shape and function (hence they are called membrane proteins); this requirement can make them especially difficult to produce and stabilize. The structure of a simpler relative of Glyra, found in bacteria, was determined to atomic resolution (more on this under a note on resolution, below) in 2008; a few other relatives were later visualized from cyanobacteria and worms. An artificial chimera protein, half-bacterial and half-human, provided further insights last year; Glyra’s first functional mammalian relatives, modified receptors for the neurotransmitters GABA and serotonin, were published in summer 2014.
Crystal clear: It’s worth noting that all these structures still represent chemically modified, albeit functional, proteins. The traditional method for determining 3-D molecular structure, X-ray crystallography, requires growing a stable crystal containing millions of identical molecules in a rigid lattice. Crystallization often requires removal of a protein’s more unstable ends, or even some middle sections; sometimes, unrelated proteins such as antibodies are added to stabilize the lattice. Crystals are grown in very different conditions than a protein would encounter in the body, and they must be frozen in liquid nitrogen and blasted with X-rays to determine the protein structure. So there is always a good chance the data will not exactly represent the protein’s most relevant form.
Moment in time: Perhaps more important, a crystal structure is a snapshot in a molecule’s life. Like most biological entities, proteins are constantly in motion: Glyra’s chloride tunnel can open and close in less than a microsecond, depending on its activation by neurotransmitters. Drugs like alcohol are thought to bind to one shape of the protein, but not another. So to truly understand a protein like Glyra, we need to see it in multiple forms. Computer simulations—like those developed in Erik’s group—can predict some molecular transitions, but need to be validated by experimental structures. Unfortunately, crystallization conditions often lock a protein in one form; finding another form of the same molecule can take years of trial and error. Previously, only two of Glyra’s relatives—one from cyanobacteria, one from worms—had been determined in more than one functional state.
Elucidating electrons: Understandably, alternative methods of structure determination are gaining in appeal. Yan Xu and colleagues at Pittsburgh recently used spectroscopy to visualize about 35 % of Glyra. Modern microscopes may offer an even more powerful approach. Traditionally, microscopy has been limited to visualizing objects larger than ~2,500 Ångströms (Å)—one 4,000th of a millimeter, but 25 times the size of a typical protein—according the Abbe diffraction limit. But in principle, illuminating a sample with an electron beam instead of visible light could improve resolution to the scale of individual atoms. Microscope samples can be prepared more easily than crystals, but they still need to be frozen to withstand an electron beam; hence, the approach is commonly called cryoEM (cryogenic electron microscopy). Nigel Unwin established much of what we know about the Glyra family by imaging one of its relatives, extracted from fish, to 4 Å resolution using cryoEM in 2004.
A note on resolution: In principle, the resolution of a structure is the closest two objects can be, without becoming indistinguishable from one another. To see every atom in a protein (33,985 in Glyra), we would like a structure resolved to around the size of an atom-atom bond, about 1 Å. So far, membrane protein structures are nowhere near that good; but even at ~3 Å, we can identify with good confidence each amino acid, the discrete building blocks that make up proteins (Glyra has just 2,105 of these). At lower resolution, as in the 3.5 Å serotonin receptor structure, some amino acids cannot be defined. The 4 Å Unwin structure, enormously informative for its time, may even have misplaced some of its helical segments by a full turn—hardly surprising when, at this resolution, a helix is more or less a solid tube.
Breaking barriers: So it was a coup when the advent of direct electron detectors improved the theoretical limit for cryoEM to ~2 Å. Deconvoluting the millions of individual molecules in a microscopy sample is still a major computational problem, but recent advances in processing power and software have made it more tractable. Ever-expanding public repositories of X-ray structure data also provide valuable templates to identify the individual atoms in a microscope image. Two years ago, my former mentor David Julius finally broke the resolution barrier for these molecules with a cryoEM structure of the capsaicin receptor, a distant relative of Glyra that binds the spicy chemical in hot chiles; and the pace of publications has since been steadily increasing.
Resolution revolution: It is difficult to overstate the impact of cryoEM on the structural biology today. X-ray expert Greg Petsko recently predicted that crystallography would be supplanted by cryoEM (and, perhaps, even better methods) within a decade; this month’s Nature commentary by Ewen Callaway invoked Gill Scott-Heron to express a similar urgency (The revolution will not be crystallized). Universities around the world are now scrambling to obtain new-generation microscopes and cryoEM specialists—the scotch whisky of scientists, suddenly in enormous demand—to operate them.
Evolving view: In the case of Glyra, Eric Gouaux and colleagues first prepared crystals of the protein and tried determining the structure by X-ray diffraction; however, the crystals were fragile and only resolved to 4.3 Å. Switching gears, they were able to fit their low-resolution X-ray model into independent images collected by cryoEM, and eventually solved the structure to 3.9 Å. As expected, the structure contains five copies of the same gene product, arranged like petals around the central tunnel; half the protein is embedded in the membrane, while the other half protrudes into the extracellular space to catch neurotransmitters and other drugs. The real revelation is in the exact positioning of ~80% of the protein’s amino acids, where tiny changes are linked to convulsive disorders, alcoholism, and possibly even autism. At a conference this week, I ran into a colleague who published some of the first models in this protein family; in light of the new Glyra data, he described his own previous work as like looking at cave paintings.
Triple threat: What’s more, due to the ease of cryoEM sample preparation (compared to crystallization), the Gouaux group was able to visualize Glyra in three different forms by adding the drugs strychnine and ivermectin, as well as the natural activator glycine. The minor but critical differences between the three resulting structures populate a landscape between receptor opening, closing, and desensitizing. Their video depiction is beautiful (and, unlike the main article, free).
Data for all: In keeping with best practices in the field, the Gouaux group deposited both the atomic coordinates and their cryoEM maps in freely accessible databases; anyone can now view and manipulate their structures, alongside at least some measures of quality. It remains to be verified that the structures represent at least some of the functional states, and whether there there are even more intermediates. A promising observation is that the structures are largely consistent with previous data from lower organisms: as predicted, structure and function seem to be largely conserved across evolution. These data are sure to provide valuable starting models for homology modeling, and the transitions and fine details will need to be examined by computational simulations. But these efforts now have a new universe of references to draw on.
Seeing drugs: These structures also offer new possibilities, as well as questions, regarding drug design. In one of the Glyra structures, we can see the poison strychnine binding to the protein outside the cell; in another structure, ivermectin—an antiparasitic drug that may also help treat alcoholism—binds within the cell membrane. Yet, these images are still too blurry to identify most small molecules. Even the neurotransmitter, glycine, is unresolved; alcohol, the drug my lab mostly studies, would be far too small to see. Only ~80 % of amino acids in each of the Glyra structures could be entirely resolved; the rest will need to be interpolated for simulation studies. An interesting general question—potentially urgent for these new data—remains as to whether the resolution of a cryoEM structure is exactly comparable to the standards set by the X-ray community, or if the 3.9 Å reported could reflect unexpected inconsistencies.
Imperfect match: It may also be important that these Glyra structures still represent an enormously simplified model system. The Gouaux group used a version of Glyra derived from a small fish: 92 % similar to the human form, it may still exhibit important differences. The structures are also of an unusually symmetrical protein, containing five identical Glyra gene products; the common form in the body is a more complex combination of slightly different subunits. Perhaps most importantly, the authors used a mutant form of Glyra with over 400 amino acids deleted—20 % of the protein, including nearly the entire portion that should extend inside the cell. Further studies will be required to determine the relevance, if any, of these changes.
Rumor mill: We all knew this paper was coming, and from whom. Gouaux is a giant in the field, having already published over 200 X-ray structures, including many membrane proteins. For at least two years, I’ve been hearing at conferences and among colleagues that he had a structure of Glyra; the paper was submitted in March and published 6 months later, a relatively fast turnaround these days. After so much anticipation, it’s a relief to finally have the details of these structures. There’s no question they will further feed the fervor for cryoEM; they may also shift the focus of computational and functional researchers (like Erik and me) to more human-like model systems. They certainly provide the best templates yet for modeling this family of molecules and their chemical interactions. In many ways, these data are a signpost for our evolving understanding of the molecular workings of the nervous system and neuroactive drugs.
It’s exciting these days to work in a field so rapidly improving in both fine detail and large-scale perspective. Sometimes it’s hard to believe the human brain will ever truly understand itself; but weeks like this make it seem a little more feasible.