X-ray crystallography

Concept

X-ray crystallography

T-093Home BU-306Threads structure · energy

Statement

Determining structure from diffraction.

Why it matters

protein-folding explains why a polypeptide adopts a specific, thermodynamically favoured three-dimensional native state, but it does not by itself tell you what that structure actually looks like at atomic resolution; x-ray crystallography is the classical experimental technique that answers exactly that question, and essentially every atomic-resolution protein structure referenced elsewhere in this unit — the active site geometry underlying michaelis-menten-kinetics, the conformational change underlying allosteric-regulation, the mechanical cycle of molecular-motors — was originally determined, or subsequently confirmed, using this method.

It also matters historically and methodologically as the technique through which the double-helical structure of DNA and the first atomic-resolution protein structures were determined, establishing structural biology as a field capable of directly visualising the molecular machinery that biochemistry and physiology had, until then, only been able to infer indirectly.

Hypotheses

The sample to be studied can be coaxed into forming a well-ordered crystal, a regular, repeating three-dimensional lattice of many identical, identically oriented copies of the molecule.Diffraction from a single molecule alone would be far too weak to detect; the crystal's ordered, repeating lattice causes scattered X-rays from many equivalent molecules to interfere constructively in specific directions, amplifying an otherwise undetectably faint signal into discrete, measurable diffraction spots. Producing a suitable, well-ordered crystal is, in practice, frequently the single hardest and most unpredictable step of the entire technique (Fails without). X-rays are scattered primarily by the electrons of the atoms in the crystal, and the scattering pattern's geometry is governed by Bragg's law, relating the crystal lattice spacing to the angles at which constructive interference (diffraction spots) occurs.Because scattering intensity depends on electron density, heavier atoms (with more electrons) scatter more strongly than light atoms; this is why hydrogen atoms, having only a single electron, are typically difficult or impossible to locate directly at the resolutions routinely achievable by this technique, and their positions are instead usually inferred from the surrounding heavy-atom geometry and standard chemical bonding constraints. Reconstructing the electron density map from the measured diffraction pattern requires knowing both the amplitude and the phase of each diffracted beam, but only the amplitude (intensity) is directly measurable; the phase information is lost in the raw measurement (the phase problem).This is the technique's central mathematical obstacle, distinct from the physical difficulty of growing a crystal: several independent methods (isomorphous replacement using heavy-atom derivatives, anomalous scattering, and molecular replacement using a known, related structure as a starting model) have been developed specifically to recover the missing phase information indirectly.

Proof

n\lambda = 2d\sin\theta \qquad (\text{Bragg's law})

X-rays reflecting from successive, parallel planes of atoms spaced a distance \(d\) apart in the crystal lattice interfere constructively only at specific angles \(\theta\) satisfying this relation, for integer \(n\) and X-ray wavelength \(\lambda\) (cited, established result, developed in bragg-law); this is the geometric principle converting the crystal's internal atomic spacing into a discrete, measurable pattern of diffraction spots at a detector. B

\text{A well-ordered protein crystal is mounted and rotated in an X-ray beam, and the intensity and position of each resulting diffraction spot is recorded across many crystal orientations.}

Each recorded spot's position encodes information about the crystal's lattice geometry (via Step 1), while its intensity encodes information about the electron-density distribution within the repeating unit; collecting spots across a full rotation of the crystal is necessary to sample enough of the diffraction pattern to reconstruct the full three-dimensional structure. A

\rho(\mathbf r) = \frac1V\sum_{\mathbf h} F(\mathbf h)\, e^{-2\pi i \mathbf h\cdot\mathbf r}

The electron density \(\rho(\mathbf r)\) at any point in the crystal's repeating unit is obtained, in principle, as a Fourier transform of the full set of structure factors \(F(\mathbf h)\) (one per diffraction spot), each of which has both an amplitude (measured directly, Step 2) and a phase (not directly measurable, Hypotheses' phase problem); recovering enough phase information to evaluate this sum is the technique's central mathematical challenge. B

\text{Phases are recovered indirectly, commonly via isomorphous replacement (comparing diffraction from the native crystal against a heavy-atom-derivative crystal), anomalous scattering, or molecular replacement (using a known, structurally related model).}

Each method supplies the missing phase information by a different route: heavy-atom derivatives introduce a known, locatable perturbation to the diffraction pattern whose effect on phase can be calculated; molecular replacement instead starts from a related structure's already-known phases as an initial estimate, refined against the new data. Once approximate phases are available for enough reflections, Step 3's Fourier sum can be evaluated to produce an initial electron density map. B

\text{An atomic model (a specific arrangement of atoms, including the known polypeptide sequence for a protein) is built into the electron density map and iteratively refined to best fit the observed diffraction data.}

Because the initial electron density map (Step 4) is generally of imperfect quality, the process of fitting and refining an atomic model, then recalculating and improving the phases and density map from that model, and repeating, converges iteratively toward a final structure that best explains the full set of measured diffraction intensities. A

Result

\text{Crystal} \xrightarrow{\text{X-ray diffraction, Bragg's law}} \text{intensities} \xrightarrow{\text{phase recovery}} \rho(\mathbf r) \xrightarrow{\text{model building/refinement}} \text{atomic structure}

Reading. A well-ordered crystal converts an otherwise undetectable single-molecule scattering event into a measurable diffraction pattern; recovering the pattern's lost phase information by one of several indirect methods allows the electron density, and from it an atomic-resolution structural model, to be reconstructed.

Scope. Requires a diffraction-quality crystal (Hypotheses, often the limiting practical constraint) and sufficient diffraction resolution to resolve individual atoms or atomic groups reliably; flexible or disordered regions of a molecule, which do not adopt a single, consistent conformation across the crystal's many copies, typically appear as weak or missing electron density even when the rest of the structure is well resolved.

Corollaries & converses

protein-folding's native state, the specific thermodynamically favoured conformation a polypeptide adopts, is precisely the structure this technique visualises directly at atomic resolution, converting a general thermodynamic principle into a concrete, atom-by-atom picture for any given protein successfully crystallised.
allosteric-regulation's claim that ligand binding at one site produces a conformational change propagating to a distant active site is most directly and convincingly demonstrated by comparing crystal structures of the same protein solved with and without the regulatory ligand bound, a standard, widely used application of this technique.
Converse: a region of a solved crystal structure showing weak, ambiguous, or entirely missing electron density is itself evidence that this region is conformationally flexible or disordered in the crystal, rather than evidence of a defect in the experiment or analysis (Common errors); a genuinely rigid, well-ordered region is expected to produce clear, interpretable density.

Fails without

Drop a well-ordered crystal (Hypotheses): without sufficiently regular, repeating molecular packing, scattered X-rays from different molecules in the sample do not interfere constructively in the specific, discrete directions Bragg's law (Step 1) predicts; the diffraction pattern degrades from sharp, measurable spots toward diffuse, largely uninterpretable scattering, and no atomic-resolution structure can be extracted — the reason growing a diffraction-quality crystal is so often the actual bottleneck of an entire structure-determination project.
Drop phase recovery (Step 4): without any method to estimate the missing phase information, Step 3's Fourier sum cannot be evaluated even with perfectly measured diffraction intensities in hand; the amplitude data alone, however precisely measured, is mathematically insufficient on its own to reconstruct the electron density map.

Common errors

Assuming a crystal structure shows the molecule's single, uniquely "correct" conformation; in solution, and to some extent within the crystal itself, many proteins sample a range of conformations, and the crystal structure specifically represents whichever conformation was compatible with forming a well-ordered lattice, which can differ subtly (or occasionally substantially) from the dominant solution conformation.
Treating missing or weak electron density in part of a structure as an experimental failure rather than genuine biological information; it is frequently direct evidence of local conformational flexibility or disorder in that region (Corollaries' converse), a real and informative feature of the molecule, not an artefact to be dismissed.
Confusing resolution (a measure of how finely diffraction data can distinguish closely spaced features, reported in Ångstroms, with a smaller number indicating higher, more detailed resolution) with accuracy; a correctly refined lower-resolution structure can still be a broadly accurate model, while a poorly refined higher-resolution dataset can still contain significant modelling errors.
Assuming hydrogen atoms are directly visualised in a typical protein crystal structure; per the Hypotheses, hydrogen's single electron scatters too weakly to be reliably located at the resolutions routinely achieved for macromolecules, and hydrogen positions are usually added computationally based on known bonding geometry rather than observed directly in the electron density.

Discussion

William Henry Bragg and William Lawrence Bragg (father and son) established the foundational law relating diffraction angle to crystal lattice spacing in the early 1910s, work recognised with a Nobel Prize and developed fully in bragg-law. Structural biology's most famous early application of the technique came in 1953, when James Watson and Francis Crick proposed the double-helical structure of DNA, drawing directly on X-ray diffraction data (notably Rosalind Franklin's) revealing the helical geometry; the first atomic-resolution protein structures (myoglobin and haemoglobin) followed shortly after, through the work of John Kendrew and Max Perutz.

Cryo-electron microscopy has, in recent years, become a major complementary (and for some classes of sample, preferred) alternative to X-ray crystallography, particularly for large complexes or membrane proteins that resist forming well-ordered crystals at all; unlike crystallography, cryo-EM does not require a crystal, instead reconstructing structure computationally from many individual images of separately frozen molecules, though it faces its own distinct resolution and sample-preparation challenges. Nuclear magnetic resonance spectroscopy offers a further alternative, capable of studying proteins in solution rather than in crystalline form, at the cost of generally being limited to smaller proteins.

Common misconception: that a published crystal structure is a direct photograph of the molecule. It is, more precisely, a computationally refined atomic model built to best explain the measured diffraction pattern (Step 5), and like any model fit to data, it carries an associated uncertainty (reflected in resolution and refinement statistics) rather than being a literal, unmediated image.

Worked examples

\text{An enzyme is crystallised both alone and in complex with a substrate analogue bound at its active site.}

Comparing the two resulting electron density maps and refined atomic models (Step 5) directly reveals which active-site residues contact the substrate analogue, and any conformational shift induced by binding — exactly the kind of structural detail michaelis-menten-kinetics' rate parameters describe only in aggregate, kinetic terms, without specifying which atoms are actually involved. A

\text{The two structures show a measurable shift in the position of a loop near the active site upon substrate-analogue binding, while a distant regulatory site simultaneously changes conformation.}

This direct, atomic-resolution observation of coupled conformational change between two spatially separated sites is precisely the kind of structural evidence used to support an allosteric mechanism (allosteric-regulation), converting a purely kinetic or biochemical inference into a visualised, physical structural change. A

\text{Apo (unbound) structure vs. substrate-bound structure} \ \Rightarrow\ \text{directly visualised conformational change}

Reading. Solving the same protein's structure under two different binding conditions and directly comparing the two atomic models is one of the most common and most persuasive applications of this technique, converting an indirect kinetic or biochemical inference into a direct structural observation.

Scope. The identical comparative-structure approach applies broadly across structural biology, wherever two or more distinct functional states of the same molecule (different ligands, different oligomeric states, different post-translational modifications) can each be separately crystallised.

Problems

A researcher obtains diffraction data of good quality and high resolution but cannot solve the phase problem using any available method, and no related structure is available for molecular replacement. Can the atomic structure be determined? Explain, using Step 3–4.
Solution
No, not from this dataset alone. Step 3 shows the electron density map requires both amplitude and phase for every measured reflection, but only amplitude is directly measured (Hypotheses' phase problem); without phases obtained by isomorphous replacement, anomalous scattering, or molecular replacement (Step 4), the Fourier sum needed to compute \(\rho(\mathbf r)\) cannot be evaluated, however precise and high-resolution the measured intensities are. The researcher would need to pursue one of the phase-recovery methods (e.g. preparing a heavy-atom derivative crystal) before structure determination could proceed.
Two crystal structures of the same protein, solved independently in two different laboratories using different crystallisation conditions, show a flexible surface loop in slightly different conformations, while the protein's core is essentially identical between the two structures. Explain this observation, using the Corollaries' converse and Common errors.
Solution
This is consistent with the loop being a genuinely flexible region of the protein rather than reflecting an error in either structure. Different crystallisation conditions can trap different, individually valid conformations of a flexible region (Common errors' point about many proteins sampling a range of conformations), while the rigid, well-ordered core, being less conformationally variable, is captured essentially identically regardless of crystallisation condition — exactly the pattern the Corollaries' converse predicts for a genuinely flexible region.
A protein of interest resists forming any crystal despite extensive attempts. Suggest one alternative structural technique mentioned in the Discussion, and state one advantage it offers for this specific case.
Solution
Cryo-electron microscopy (Discussion) does not require the sample to be crystallised at all; structure is instead reconstructed computationally from many individual images of separately frozen molecules, making it a suitable alternative specifically for proteins (including many large complexes and membrane proteins) that resist forming the well-ordered, diffraction-quality crystals required by conventional X-ray crystallography.