Chapter I
A Mechanism Read Off a Map
The first protein structure, myoglobin, explained very little. Myoglobin stores oxygen, and seeing where the haem sits confirmed that it has a pocket for it, which was not news. The question was whether structures would ever explain chemistry.
Lysozyme answered it. David Phillips's group solved the enzyme at 2 Å in 1965, and then did the decisive experiment: solved it again with a sugar inhibitor bound. The substrate lies in a long groove across the molecule's face. Two acidic residues sit on either side of the bond that gets cut — Glu35 positioned to donate a proton to the leaving oxygen, Asp52 positioned to stabilise the positive charge that develops on the sugar. And the sugar in the fourth subsite cannot fit without being distorted out of its relaxed shape, towards the flattened geometry it must adopt in the transition state.
Everything in that paragraph is visible in the map. An enzyme accelerates a reaction by binding the transition state better than the substrate, and here was a picture of it being done. After lysozyme, "solve the structure" became the standard move for understanding any protein, and the rest of the field is about extending the range of what can be solved.
Chapter II
Three Barriers, Removed in Turn
Crystals. Kurt Wüthrich showed in the 1980s that a structure can be obtained in solution instead. Nuclear magnetic resonance can report which protons are within about 5 Å of each other; collect enough such constraints, assign each resonance to a specific atom in the sequence, and compute the conformations consistent with the list. The method is limited to smaller proteins, and it has an advantage crystallography lacks: it sees motion, and reports a family of conformations rather than one.
Membranes. A membrane protein has a hydrophobic belt that must be in contact with lipid, so it will not dissolve in water, and the detergents that keep it soluble interfere with crystal packing. The consensus was that these proteins could not be crystallised, which was awkward, since they are about a quarter of the proteome and most drug targets. Hartmut Michel found conditions that worked for a bacterial photosynthetic reaction centre, and Johann Deisenhofer and Robert Huber solved it in 1985. The structure shows the pigments down which an electron hops after a photon is absorbed, at spacings that explain the direction and speed of the transfer — a result that belongs as much to wave optics and quantum mechanics as to biology.
Size. By the late 1990s the frontier was the ribosome: two subunits, three RNA chains of several thousand bases, more than fifty proteins, a quarter of a million atoms, and a crystal that diffracts badly. Three groups solved it between 1999 and 2001, in a race described frankly by one of the participants.
Chapter III
A Closer Look: Eighteen Ångströms of Nothing
The ribosome structures settled a question that biochemistry had been unable to touch. The peptidyl transferase centre is where an incoming amino acid is joined to the growing chain — the chemical step at the heart of all protein synthesis. Was it catalysed by one of the ribosome's fifty-odd proteins, or by its RNA?
The argument from the structure is a distance measurement. In the 2.4 Å map of the large subunit, with a transition-state analogue bound in the active site, the nearest atom of any protein side chain is about 18 Å from the site of the reaction. For comparison:
| Distance | What is at that range |
|---|---|
| 1.5 Å | a covalent bond |
| 2.8 Å | a hydrogen bond |
| 3–4 Å | van der Waals contact |
| 18 Å | five water molecules' worth of nothing |
No chemistry reaches 18 Å. Acid–base catalysis requires a proton donor in hydrogen-bonding range; electrostatic stabilisation falls off with distance and, through water, is screened within a few ångströms. Whatever the ribosomal proteins do — and they do stabilise the structure, and assist assembly — they cannot be performing the catalysis. The catalytic site is lined by RNA bases, and the ribosome is therefore a ribozyme.
The consequence reaches back four billion years. Protein synthesis cannot have required proteins to begin with, so the machine that makes proteins is made of the other polymer. Crick, Orgel and Woese had each suggested in the 1960s that RNA came first, on the grounds that it can both carry information and fold; catalytic RNAs were found in the 1980s, which showed it was possible; the ribosome structure showed that the most fundamental process in the cell still works that way.
There is a second lesson in the numbers, about what resolution buys. At 5 Å the ribosome is a shape, and the argument above cannot be made. At 2.4 Å individual bases, bound waters and the analogue's geometry are all placed, and an 18 Å gap becomes a measurement rather than an impression. The difference between the two maps is a factor of four in the number of reflections measured — and about fifteen years of work.
Chapter IV
Pictures Instead of Crystals
The last barrier fell for a mundane reason: better cameras. Electron microscopy of frozen biological molecules had been possible since Jacques Dubochet worked out in the early 1980s how to freeze a sample so fast that the water becomes glass rather than ice, and Joachim Frank had developed the mathematics for taking tens of thousands of images of individual particles lying in random orientations and averaging them into a three-dimensional map. The resulting maps were blurry, and the field was nicknamed blobology.
What changed between 2012 and 2014 was the detector. Direct electron detectors record individual electrons and read out fast enough to split an exposure into frames, so the drift of the specimen during exposure — previously an irreducible blur — can be tracked and corrected. Resolutions crossed 3 Å, and then went further.
The consequences have been larger than an improvement in resolution. No crystal is needed, so complexes that had resisted crystallisation for decades were solved within months, including the spliceosome and many membrane receptors. And because each image is of one particle, a sample containing several conformations can be sorted computationally into separate maps instead of being averaged into an uninterpretable mean — so a machine can be caught in several states of its cycle, which is what molecular machines requires.
What structural biology still cannot handle is the third of the proteome that has no fixed structure at all. Disordered regions are conserved, functional, and invisible to every method in this chapter, and the question of what should replace a list of coordinates for them is open.