Chapter 3.1.1 — Why NMR Structures? 🧲🧬 (detailed summary)

Big picture: what makes NMR special?

NMR is valuable in structural biology because it gives both structure and dynamics in conditions that often resemble a protein’s physiological (solution) environment. Unlike a single “best” structure, an NMR structure is typically an ensemble of conformers that all satisfy the experimental restraints—this naturally aligns with the fact that proteins can be dynamic rather than rigid.

A quick historical anchor (why the method mattered early on)

• The first NMR solution structure of a biological macromolecule was solved in 1984 (Wüthrich and coworkers) for a 57–amino acid proteinase inhibitor IIA from bull seminal plasma.
• NMR’s value was strongly reinforced when the solution NMR structure of rat metallothionein-2 showed that an available crystal structure needed revision—a key early example of NMR revealing that “crystal ≠ solution reality” in some cases.

Where NMR sits among the “big three” structure methods 🧩

The chapter frames structural biology around three main experimental approaches (highlighted in the pie chart on page 2, Fig. 3.1.1):

1. X-ray crystallography
2. Solution NMR
3. Cryo-electron microscopy (cryo-EM)

Key takeaways:

• X-ray and NMR are the two main atomic-level methods in routine use for biomolecules (in crystals vs solution).
• Cryo-EM is increasingly important for very large assemblies (ribosome, microtubules, viruses), and methodological advances have pushed its “routine” applicability down to around ~200 kDa.
• Cryo-EM often has lower resolution than X-ray, but may capture more biologically relevant conformations because it avoids crystal lattice constraints.
• Methods can be combined: atomic structures (X-ray or NMR) of components can be docked into cryo-EM maps to build pseudo-atomic models and identify contact interfaces.

Figure 3.1.2 (page 2) gives a key “size preference” message via the bar chart:

• NMR is preferentially used for low molecular weight systems
• Cryo-EM is preferentially used for very large assemblies

Practical/technical differences: what each method “needs” and what it “sees”

Hydrogens:

• NMR restraints (like distance restraints) often explicitly involve hydrogen positions, so H atoms are typically included in NMR PDB coordinate files.
• X-ray structures usually don’t include hydrogens in coordinate files because diffraction is dominated by heavier atoms and hydrogens are not directly seen in typical data.

Sample requirements (surprisingly similar): For both NMR and X-ray you generally need protein samples that are:

• homogeneous
• stable for days/weeks
• soluble enough
• not irreversibly aggregated at high concentration

Even if you meet these, you still can’t be sure you’ll get:

• crystals that diffract well, or
• NMR spectra good enough for structure determination

NMR as a screening tool for crystallography 🧪➡️🧊

Because broad requirements overlap, NMR has been used to screen protein samples to identify ones more likely to crystallize well.

1D ¹H NMR screening idea

The chapter notes that ¹H 1D spectra can categorize proteins into groups that:

• crystallize with similar success rates,
• but differ in the quality of diffraction the crystals provide.

Better diffraction tends to come from proteins whose ¹H spectra show:

• higher chemical shift dispersion (suggesting a defined tertiary structure),
• narrower linewidths (arguing against conformational exchange/equilibria).

HSQC as the workhorse screen (and why)

A central concept here is the ¹H–¹⁵N HSQC spectrum:

• It gives a 2D map where (ideally) you see one peak per backbone amide (except prolines).
• ¹⁵N labeling is relatively affordable, making HSQC-based screening feasible.
• HSQC can guide rational construct design (domain boundaries, trimming flexible ends/loops) before crystallization trials, improving chances of crystallization success.

Key point: NMR spectral quality ≠ guaranteed crystallography success ⚠️

The chapter is explicit that correlation exists, but it is not stringent.

Two example findings:

• In one study (159 proteins), only about ~20% of high-resolution X-ray structures came from samples with excellent/good HSQC spectra.
• Surprisingly, ~33% of proteins with lower-quality but still “promising” HSQC spectra could still yield successful X-ray structures.

Another study (263 proteins) found:

• only 21/263 (~8%) were amenable to both NMR and X-ray based on HSQC + optimized crystallization trials, but using both methods in parallel increased the total “solvable” targets to 107/263 (~41%), split roughly as:
• 43 NMR-only
• 43 X-ray-only (and the remainder being the overlap/other outcomes described in the text).

Interpretation (as given in the text): strong complementarity. Some proteins are simply better suited to one method than the other even if they look “good” by shared criteria (solubility/monodispersion).

Why would a protein be “NMR-friendly” but “crystal-hostile”?

A major reason proposed: highly flexible or unstructured regions.

• These regions may not wreck NMR spectra as badly,
• but can be extremely adverse to crystallization (crystals often require a consistent, well-packed arrangement).

Construct redesign can sometimes help:

• If unstructured parts are at termini, you may remove them.
• If they’re internal (e.g., a long loop between secondary structure elements), redesign may be difficult or impossible.

How to “read” HSQC spectra as a folding/quality report 📈

Figure 3.1.3 (page 4) is central here: it shows three HSQC patterns that correspond to different folding states.

A simple quantitative metric

You can evaluate HSQC quality by:

• counting the number of well-detectable (intense, well-resolved) backbone amide peaks,
• comparing that count to the number of non-Pro residues in the sequence.

This gives an estimate of how many residues are in folded regions.

What different features mean

• Weak minor peaks and/or broad peaks collapsed near the center indicate regions lacking a defined 3D structure.
• More peaks than expected can indicate multiple conformations that interconvert slowly on the chemical shift timescale (so you see separate sets rather than one averaged peak).

The three HSQC “fingerprints” (Fig. 3.1.3)

• (a) Fully folded protein: sharp, intense peaks; wide chemical shift dispersion; number of backbone amide peaks matches expectation.
• (b) Partially unfolded: reduced dispersion (especially in ¹H); intense peaks become broad and cluster around ~8.5 ppm (random-coil-like); peaks outside are weak.
• (c) Unfolded: almost all backbone amide resonances collapse into a narrow band around ~8.3 ppm; even side-chain NH₂ pairs (Gln/Asn) lose resolution as unfolding increases.

No “scientific” reason to always prefer X-ray or NMR 🧠

The authors argue there is no a priori scientific reason to prefer one technique for a protein target. The choice is often practical/feasibility-based, and the methods are complementary.

Practical considerations mentioned:

• NMR structure determination has historically been less automated than X-ray pipelines (though automation exists and is discussed elsewhere in the book).
• NMR generally requires isotopic labeling, and the needed labeling strategy depends on protein size.

A big conceptual advantage: solution relevance and conformational selection 🌊🧬

A distinct advantage of NMR emphasized here:

• X-ray structures can capture a conformation that is not the dominant one in solution, which is often closer to physiological conditions for soluble proteins.

Example given:

• HPr (histidine-containing phosphocarrier protein) from Enterococcus faecalis: an X-ray structure showed a strained conformation of an important active-loop residue (proposed to be catalytically important), but NMR later indicated this conformation is at most a minority in solution.

Proposed explanation:

• If multiple states exist in solution, crystallization may select one—the most “crystal-prone” conformation (a form of conformational selection imposed by crystallization conditions).

Second illustrative case:

• Matrix metalloproteinase-12: different crystalline environments “freeze” different conformations of flexible regions, while NMR in solution reveals conformational equilibria across timescales.

Core limitation of solution NMR: size 🐘 vs 🧲

The chapter reiterates a key limitation:

• De novo solution NMR structure determination is largely confined to proteins below ~25 kDa.

Why (physical + practical):

1. Slower tumbling for larger proteins increases linewidths, potentially making signals too broad to detect.
2. More residues create more peaks → spectral overlap increases → assignments become difficult.

This is described as an active frontier: pushing the size limit is a major goal in NMR methodology.

Consequences for method choice:

• X-ray is typically the technique of choice for large proteins/complexes at atomic resolution.
• Even larger or non-crystallizable systems can be studied by cryo-EM.
• NMR/X-ray can be combined with cryo-EM by supplying atomic structures for components to interpret the lower-resolution overall map.

Solid-state NMR: a way around the “tumbling” size limit 🧱🧲

A key point near the end of 3.1.1:

• The size limit due to slow tumbling does not apply to solid-state NMR, because molecules are not freely rotating in solution—they’re effectively fixed in orientation.

The chapter notes major recent efforts to develop solid-state NMR methods for various sample types:

• soluble proteins in microcrystalline form,
• protein fibrils,
• membrane proteins in lipid bilayers, etc.

Important nuance:

• In favorable cases, membrane proteins can still be studied by solution NMR if reconstituted in detergent micelles.

Why microcrystals matter:

• Microcrystals can be easier to produce than the high-quality single crystals needed for X-ray diffraction, motivating solid-state NMR as an alternative route for larger proteins.

But also:

• Solid-state NMR is still in heavy methodological development, without fully standardized protocols yet (as described here).

Where solid-state NMR has been especially successful:

• Protein fibrils: solid-state NMR can uniquely provide detailed atomic-level packing information.
• Membrane proteins: a long-standing target for solid-state NMR (decades), and often difficult to crystallize.

Historical and modern examples mentioned:

• Early successes: bacteriorhodopsin and gramicidin as cases where solid-state NMR enabled structural study when other methods struggled.
• Functional conformational cycling: retinal proteins (e.g., rhodopsin) undergoing light-activated state changes in both protein and chromophore.
• M2 proton channel: studied for structure, function, and ligand binding; solid-state NMR has uniquely enabled investigation of ligand binding (drugs/toxins) to functional membrane proteins.

Tiny “stick it in memory” recap 🧠✨

• NMR shines because it links structure + dynamics in solution (often more physiological).
• NMR vs X-ray is often about complementarity, not superiority.
• HSQC spectra are a powerful way to judge folding state, conformational heterogeneity, and sometimes crystallization prospects—but correlation is imperfect.
• Solution NMR size limit (~25 kDa) comes from tumbling/linewidth and overlap/assignment issues; solid-state NMR can bypass tumbling limits and is key for fibrils/membranes.