🧲 NMR Spectroscopy for Binding Studies

CSP (Chemical Shift Perturbation)

Record ¹⁵N-HSQC (Heteronuclear Single Quantum Coherence) spectra at increasing ligand concentrations. Peaks shift progressively in fast exchange. The magnitude and direction of shifts yield K_D via titration fitting. ¹⁵N labeling required for 2D experiments; backbone assignment needed to map the binding site to specific residues but not for K_D determination.

STD-NMR

Saturation Transfer Difference — selectively saturate protein resonances; saturation transfers to bound ligand via spin diffusion. The difference spectrum reveals which ligand protons contact the protein. No labeling required. Works at low protein concentration (1–50 µM) with large ligand excess (typically ~100×, to ensure unsaturated ligand replaces the departing one). For quantitative epitope mapping or K_D estimation, use initial slopes of build-up curves rather than STD_max — a rapidly off-rating ligand can rebind while still partially saturated, inflating apparent STD at longer saturation times. For simple binding-vs-no-binding screening this is less critical. The gold standard for fragment screening in FBDD. STD and WaterLOGSY are prone to false positives from compound aggregation, non-specific binding, and detergent-like effects — counter-screen with apo (protein-free) controls and orthogonal competition experiments. See also rdSTD (reduced-dataset STD) for faster acquisition.

WaterLOGSY

Water-Ligand Observed via Gradient SpectroscopY — exploits magnetization transfer from bulk water through the protein–ligand complex. Binding ligands show opposite sign compared to non-binders. Complementary to STD with a different magnetization pathway. No labeling, low protein, qualitative binding readout.

CPMG Relaxation Dispersion

Carr-Purcell-Meiboom-Gill experiment measures how R_2,eff varies with refocusing pulse frequency. Detects µs–ms exchange and extracts k_ex, populations, and |Δω| at a single static field. A second static field (or supplementary CEST/R_1ρ data) is needed to remove sign/correlation ambiguity in Δω, since R_ex scales with B₀². Together these methods provide kinetic rates and characterize invisible excited states.

trNOE (Transferred Nuclear Overhauser Effect)

Small ligands tumble fast (weak/positive NOE); when bound to a large protein, they experience slow tumbling (strong NOE with negative cross-relaxation rate, giving NOESY cross-peaks with same sign as diagonal). In fast exchange, bound-state NOEs dominate even at low bound fraction. Reveals the 3D conformation of the ligand in the binding pocket — unique to NMR.

PRE (Paramagnetic Relaxation Enhancement)

Attach a paramagnetic spin label to the protein. The unpaired electron causes distance-dependent relaxation enhancement (∝ 1/r⁶), useful out to ~20–25 Å for nitroxide spin labels (e.g. MTSL) and further still for lanthanide tags (Gd³⁺, Dy³⁺) — far exceeding the ~5 Å range of NOEs. Because of r⁻⁶ averaging weighted toward short distances, PRE is uniquely sensitive to sparsely populated states such as transient encounter complexes, and maps binding interfaces with long-range distance restraints.

Parameters

• Δδ_H — change in ¹H chemical shift (ppm)
• Δδ_N — change in ¹⁵N chemical shift (ppm)
• α — normalizes the much larger ¹⁵N shift range against the smaller ¹H range so both contribute meaningfully to the Euclidean distance. Values are derived either from gyromagnetic ratios (theoretical) or from BMRB shift-range statistics (empirical)

Common α Values

• ~0.10 — magnitude of |γ_N/γ_H| (γ ratio, theoretical); rarely used in practice because it under-weights ¹⁵N relative to its empirical shift dispersion
• 0.14 — Williamson (2013) recommendation based on BMRB statistical analysis of backbone amide shift distributions; most widely used for standard backbone CSP calculations
• 0.17 — Mulder et al. (1999) value derived from backbone amide pairs; seen in older literature
• 0.2 — commonly used empirical value (Farmer et al. 1996)

Global Fitting

Fit multiple residues simultaneously sharing a single K_D, with individual Δδ_max per residue. This is analogous to global kinetic fitting of multiple SPR sensorgrams — the same principle of shared parameters with residue/concentration-specific observables. Consistency of K_D across residues validates a single binding site.

Simplified Form

When [L]₀ >> [P]₀ (ligand in large excess), the equation simplifies to a hyperbolic saturation curve:

Δδ_obs = Δδ_max × [L]₀ / (K_D + [L]₀)

Reliable K_D range: ~1 µM – 10 mM (fast exchange regime).

🧲 Choose NMR when you need…

• Atomic-resolution binding site mapping — the only solution-state method providing true atomic-resolution mapping (requires ¹⁵N labeling; HDX-MS offers peptide-level resolution without labeling)
• Fragment screening — STD/WaterLOGSY are gold-standard for FBDD
• Weak binder detection — excels where SPR/BLI struggle (K_D > 100 µM)
• Bound conformation — trNOE gives 3D structure of ligand in the binding pocket
• Aggregation/quality control — immediately reveals compound aggregation or protein unfolding
• Invisible excited states — CPMG detects transiently populated conformations invisible to other methods

✨ Choose SPR / 💡 BLI / 🌡️ ITC when you need…

• High-affinity K_D — SPR/BLI measure down to pM; NMR limited to > ~1 µM
• Direct kinetics (k_on, k_off) — SPR/BLI provide real-time sensorgrams
• Thermodynamics (ΔH, ΔS) — ITC is the gold standard
• High throughput — SPR/BLI screen hundreds of analytes in hours
• Large proteins — no practical size limit for SPR/BLI/ITC
• Low sample consumption — SPR/BLI require µg quantities vs mg for NMR

Sample Requirements

Protein-observed experiments (CSP, CPMG) require 0.1–0.5 mM protein (~200–500 µL), typically ¹⁵N-labeled. Ligand-observed experiments (STD, WaterLOGSY) need only 1–50 µM protein (unlabeled) with 20–100× ligand excess. Budget several mg of protein for a full CSP titration series. Buffer constraints: samples require 5–10% D₂O for field-frequency lock; use low-proton buffers (phosphate, Tris-d11, HEPES-d18) to minimize background. DMSO tolerance is typically ≤5% v/v for protein-observed experiments and up to ~10% for ligand-observed methods — exceed this and protein signal degrades markedly.

Molecular Weight Limits

Standard experiments work up to ~40 kDa. TROSY (Transverse Relaxation-Optimized SpectroscopY)-based experiments (requires ¹⁵N labeling; perdeuteration strongly recommended for proteins >30 kDa) extend the range to ~100 kDa or beyond. For systems above ~100 kDa (ribosomes, proteasome, GPCR complexes), methyl-TROSY with ILV/ILVA labelling (selective ¹³CH₃ labelling of Ile, Leu, Val, Ala methyl groups in an otherwise perdeuterated background) is the dominant approach, exploiting the favourable relaxation properties of methyl groups. Ligand-observed methods (STD, WaterLOGSY, trNOE) have no practical protein size limit — only the ligand needs to be NMR-visible.

Instrumentation & Cost

High-field NMR spectrometers (600–900 MHz) cost $1–5M. Cryoprobes increase sensitivity 3–4× (per scan; equivalent to ~10–16× time savings) but add cost. Single-field CPMG gives k_ex, populations, and |Δω|; a second field (or CEST/R_1ρ) is needed only to resolve sign/correlation ambiguity in Δω. Most binding studies are performed at academic NMR facilities or pharma core labs. Expect $50–200/hour instrument time.