Protein Structure and Interactions from Nuclear Magnetic Resonance Spectroscopy

Email a friend
Contact the editor about this article
How to cite

Protein Structure and Interactions from Nuclear Magnetic Resonance Spectroscopy

Ranajeet Ghose, Graduate Center of the City University of New York, New York, USA

Published online: January 2017

DOI: 10.1002/9780470015902.a0003100.pub3

Full Article on Wiley Online Library

Abstract
Images
References

Abstract

Nuclear magnetic resonance (NMR) spectroscopy is a biophysical technique that facilitates determination of protein structures and interactions in solution. NMR experiments are used to identify spectral signatures corresponding to specific atoms on individual protein residues and determine their order in the primary sequence. Distance, angular and orientational relationships are subsequently measured and utilised in computational protocols to obtain high‐resolution structural models. Structure determination has been aided by advances in biotechnology allowing the generation of samples optimally labelled with NMR‐active isotopes; by improvements in the design of NMR instrumentation to allow spectral signatures to be recorded with high sensitivity; by the development of efficient techniques to manipulate nuclear spins at the quantum level and by the creation of computer algorithms that allow the rapid processing and manipulation of NMR data and the incorporation of novel information available from NMR experiments into strategies to obtain the atomic models of proteins and their complexes.

Key Concepts

Solution NMR is a spectroscopic technique that enables the determination of the three‐dimensional structure of proteins and their interactions under near‐physiological conditions.
The abundance of chemical and structural information available from NMR spectra derives from the ability to selectively manipulate specific nuclei within the molecules placed in a large static magnetic field by customisable sequences of radio‐frequency pulses.
NMR signals report on the variety of chemical environments experienced by nuclei in proteins through unique spectral signatures.
The repertoire of information‐rich NMR experiments has been significantly expanded by the ability to enrich biomolecules with NMR‐active nuclei by taking advantage of advances in biotechnology.
Nuclear spins in proteins form magnetically coupled networks, with characteristic signatures in NMR spectra that can be manipulated to extract angular, distance and orientational information to enable the generation of three‐dimensional structural models.
Changes in the spectral signatures of individual nuclei that make up each component of a complex in the presence of the other component/s provide information about the mode of interaction among the constituents in the assembled state. In some cases, where three‐dimensional structures of individual constituents are available, the spectral changes or perturbations may be utilised in novel computational strategies to obtain three‐dimensional structural models of the complexes.

Keywords: NMR; proteins; three‐dimensional structure; biomolecular interactions; isotope labelling

	Figure 1. One‐dimensional to multidimensional Fourier transform nuclear magnetic resonance (NMR). (a) Representation of a free‐induction decay (FID) time‐domain signal that upon Fourier transformation yields the resonance frequencies of all the constituent nuclei, that is a spectrum. (b) A typical 2D experimental scheme includes a preparation period, an evolution period (the delay t₁ is incremented in discrete steps) and a detection period (t₂) during which the FID is detected. The last two periods usually sandwich a mixing period [e.g. in a NOESY (nuclear Overhauser effect spectroscopy) experiment the mixing period is usually 150–300 ms and allows the dipolar coupling between two protons to enable magnetisation exchange before detection]. Fourier transformation of the detected signal in each of the evolution (indirect) and detection (direct) domains leads to a 2D spectrum with two frequency (chemical shift) domains – δ₁ and δ₂ (expressed in parts per million, ppm). This approach can be generalised to additional dimensions, for example a 3D spectrum has two indirect evolution periods. The 2D spectrum of a hypothetical molecule containing three nuclei (A, B and C) has a 1D NMR spectrum that is displayed along the diagonal of a 2D spectrum (red, blue and green circles). Correlation of the resonance frequencies (chemical shifts, δ_A, δ_B and δ_C) of the atoms leads to off‐diagonal cross‐peaks each representing a correlation, either through‐bond (COSY or TOCSY experiment) or through‐space (NOESY experiment), between the nuclei. In the example shown here, nucleus C is coupled to both nucleus A, and to nucleus B (as evident from the cyan cross‐peaks at δ_A,δ_C and δ_C,δ_A and the yellow cross‐peaks at δ_B,δ_C and δ_C,δ_B); nuclei A and B are not coupled to each other hence there are no cross‐peaks at δ_A,δ_B or δ_B,δ_A. (c) A homonuclear ¹H–¹H 2D NOESY spectrum of a protein showing through‐space correlations. (d) 2D ¹⁵N, ¹H heteronuclear correlation spectrum (HSQC) of a protein displaying the protein amide ¹⁵N and attached ¹H_N chemical shifts. Note that some side chains, for example the NH₂ moieties of certain amino acids (e.g. glutamine and asparagine; two cross‐peaks at each ¹H chemical shift for the same ¹⁵N chemical shift) and the indole NH moiety of tryptophan also yield cross‐peaks in an HSQC spectrum. (e) Advantages of the transition from two to three dimensions can be appreciated by the fact that the red cross‐peak in the 2D spectrum appears to be a single peak but resolves into two separate peaks (light and dark orange) along the third dimension.
	Figure 2. Selective labelling of proteins. (a) A 2D H(N)CO (in red, bottom; similar to a conventional 3D HNCO spectrum except that the amide N chemical shift is not recorded) spectrum for a protein selectively ¹³C′‐labelled only on the phenylalanine residues in a uniformly ¹⁵N‐labelled background. This labelling is generated by supplementing the growth medium (M9 with ¹⁵NH₄Cl, unlabelled glucose) with phenylalanine ¹³C‐labelled at the carbonyl C′ position. The top panel shows a corresponding 2D H(N)CO spectrum of uniformly ¹³C, ¹⁵N labelled protein (in black). Note that not all amino acid residues can be labelled in this simple way due to substantial cross‐talk between their respective biosynthesis pathways that leads to the label being ‘scrambled’. (b,c) Position‐specific labelling can be achieved by supplementing the growth media with precursors of specific amino acids labelled in a particular manner. This scheme exploits the fact that certain chemical moieties on the precursors are incorporated at defined positions in the side chains of the amino acid residues they generate. (b) Specifically labelled α‐ketobutyric acid results in the incorporation of the label at the δ₁ position of isoleucine (red) residues. (c) Specific labels placed on α‐ketoisovaleric acid lead to the transfer of the labels to the δ₁/δ₂ and γ₁/γ₂ positions of leucine and valine residues, respectively. All other positions are ¹²C and ²H. Examples of the resulting ¹³C,¹H HMQC (heteronuclear multiple quantum correlation) spectra of the resultant samples are shown. While ¹³C,¹H HMQC spectra provide similar information as ¹³C,¹H HSQC spectra (discussed in the text), they can be recorded with higher sensitivity when using this particular labelling scheme for methyl groups.
	Figure 3. Establishing intra‐ and interresidue correlations in proteins. The 3D HNCACB experiment (a) correlates the chemical shifts of the ¹³C_α and ¹³C_β nuclei of a residue (i) and the ¹³C_α and ¹³C_β nuclei of the previous (i − 1) residue with the chemical shifts of its own (i) amide ¹⁵N and ¹H_N nuclei. The 3D CBCA(CO)NH (b) experiment correlates the chemical shifts of the ¹⁵N and ¹H_N nuclei of a given residue (i) with the ¹³C_α and ¹³C_β shifts of the preceding (i − 1) residue. In this example (c), two sets of HNCACB/CBCA(CO)NH pairs are shown that yield (i − 2)/(i − 1) and (i − 1)/i correlations. The ¹³C_α and ¹³C_β resonances are shown in dark blue and red (signifying their opposite phases, when one is positive in sign, the other is negative) for the HNCACB experiments, respectively, and both are shown in blue for the CBCA(CO)NH experiment. The J‐couplings that allow transfer of magnetisation include ¹J(CH) ∼ 140 Hz, ¹J(NH_N) ∼ 92 Hz, ¹J(CC) ∼ 35 Hz, ¹J(CC′) ∼ 55 Hz, ¹J(NC′) ∼ 15 Hz, ¹J(NC_α) ∼ 7–11 Hz and ²J(NC_α) ∼ 4–9 Hz. The flow of magnetisation is indicated by the red arrows and is bidirectional for the HNCACB experiment (a so‐called out‐and‐back experiment) and unidirectional in the CBCA(CO)NH experiment (a so‐called straight‐through experiment). Each strip plot is extracted from the 3D data set at the plane corresponding to the chemical shift of the amide ¹⁵N nucleus indicated at the top of each strip.
	Figure 4. Sequence‐specific assignment. Four strip plots showing the sequential connectivity between four successive residues (i − 1)/(i)/(i + 1)/(i + 2) through their ¹³C_α and ¹³C_β resonances, the so‐called HNCACB walk. Note that the (i + 2) residue is a glycine and is therefore missing a ¹³C_β resonance.
	Figure 5. Characteristic chemical shifts for amino acid residues. The mean (represented by circles) and standard deviations (represented by horizontal error bars) for the backbone and side‐chain ¹³C (a) and ¹H nuclei (b) extracted from the Biological Magnetic Resonance Bank (BMRB) are shown. The cysteine C_β position has a large standard deviation as it includes both the oxidised (in a disulfide bond) and the reduced (SH) forms of the amino acid.
	Figure 6. Examples of the types of constraints obtained from NMR experiments. (a) Karplus curves relating measured three‐bond scalar couplings (³J) to dihedral angles. The black, red and blue curves represent the influences of the backbone dihedral angle φ, the backbone dihedral angle Ψ, the side‐chain dihedral angle χ₁ on the ³J(H_NH_α), ³J(H_α,iN_i+1) and ³J(H_αH_β2) coupling constants, respectively. The parametric equations representing the relationships are given by ³J(H_NH_α) = 7.97 cos²(φ − 60°) − 1.26 cos (φ − 60°) + 0.63 Hz; ³J(H_α,iN_i+1) = −1.0 cos²(Ψ − 120°) + 0.65 cos (Ψ − 120°) − 0.15 Hz; ³J(H_αH_β2) = 9.5 cos²(χ₁ − 120°) − 1.6 cos (χ₁ − 120°) + 1.8 Hz. A qualitative evaluation of the values of ³J(H_NH_α) can be used to ascertain whether a particular residue is in a helix or in an extended structure like a β‐sheet given that these values are small in the former (α‐helix, φ = −57°, ³J = 2.84 Hz, 3₁₀‐helix, φ = −49°, ³J = 1.88 Hz) and quite large in the latter (parallel β‐sheet, φ = −119°, ³J = 9.86 Hz; antiparallel β‐sheet, φ = −139°, ³J = 8.95 Hz). These equations are only one of several parametrisations of the Karplus curves for a given ³J coupling; in this case ³J(H_NH_α) is from Vogeli et al. , ³J(H_α,iN_i+1) is from Lohr et al. and ³J(H_αH_β2) is from Demarco et al. . Other forms of the Karplus curves have also been suggested (Schmidt, ). (b) Measurement of residual dipolar couplings (RDCs) from coupled ¹⁵N, ¹H HSQC spectra. Two sets of coupled HSQC spectra are shown, in isotropic solution (right), and in a medium that leads to weak alignment of the molecule (left). Residual dipolar couplings (D) shown here are all negative in sign leading to decreased splitting of the resonances (J + D) in the partially oriented state. (c) The cross‐peak (red) obtained in a long‐range 2D H(N)CO experiment that optimises magnetisation transfer based on a three‐bond J coupling ( $^{h 3} J_{N C^{'}}$ ) between N and C′ nuclei across a hydrogen bond. The cross‐peak appears at the chemical shifts of the nuclei corresponding to the hydrogen‐bonded ¹³C′ and ¹H_N nuclei (right panel). Also shown for comparison is an overlay of the corresponding region of a conventional 2D H(N)CO spectrum (optimised for the one‐bond NC′ coupling, $^{1} J_{N C^{'}}$ ) that show the resonances (blue) at the chemical shifts of the backbone ¹³C′(i − 1) and ¹H_N (i) nuclei, there is no peak corresponding to the hydrogen‐bonded pair in this spectrum. A comparison between these two sets of spectra allows the elucidation of the residues that participate in a hydrogen bond.
	Figure 7. A flowchart illustrating the steps in determining protein structures by NMR. Following the preparation of isotope‐labelled samples and data collection on the NMR spectrometer, a variety of constraints (distance, angular and orientational; the last is usually not used in the initial rounds) are used as inputs into a calculation protocol termed molecular dynamics simulated annealing. The simulated annealing calculation uses a target function that models the potential energies of chemical bonds, nonbonded energies such as van der Waals forces and angular relationships, in addition to the experimental constraints. Validation is performed at each step in a cycle to check for violations of experimental constraints by the cluster of structures generated; incorrect assignments (these are most commonly incorrectly assigned NOESY cross‐peaks) are rectified; and further assignments are obtained to extract additional constraints; additional experiments are performed if deemed to be necessary. The iterative cycle continues until all possible experimental data are utilised in the calculation. The final structural ensemble is validated for local and global geometry and structural characteristics. The final step consists of deposition of the atomic coordinates into the PDB and the resonance assignments into the BMRB databases.
	Figure 8. Structure refinement. Illustration of the improvement in the root‐mean‐squared deviation (RMSD) in a cluster of structures as more and more experimental restraints are incorporated into the structure calculation protocol during the iterative refinement procedure.Reproduced with permission from Shekhtman et al. © John Wiley and Sons.
	Figure 9. NMR spectra to assess molecular interactions and stability. (a) Representative chemical shift changes in ¹⁵N,¹H HSQC spectra of a protein in the presence of a ligand (blue). The spectra of the apo (ligand‐free) protein are shown in black. The boxed resonance does not change position in the presence of the ligand suggesting that the corresponding region is not involved in binding. The protein/ligand interaction illustrated here is in the fast exchange regime resulting in the gradual shift of the resonances from the chemical shift position of the free state to that of the bound state with increasing titrant concentration. The region for which resonances show chemical shift changes in the presence of the ligand are mapped onto the structure of the protein revealing the binding site. (b) Gradual chemical shift changes (Δδ) with increasing ligand concentration for the fast exchange regime. The different curves represent titration profiles different resonances. The chemical shift changes or perturbations are usually estimated from a 2D correlation spectrum, for example a 2D ¹⁵N,¹H HSQC by measuring the changes in heteronuclear (e.g. ¹⁵N) and proton positions during the titration course. The Δδ(j) at each (jth) titration point can then be represented as Δδ(j) = [(δ_H,j − δ_H,free)² + K(δ_X,j –δ_X,free)²]^1/2, where δ_H,j and δ_X,j represent the proton and heteronuclear chemical shifts for the jth titration point, respectively; δ_H,free and δ_X,free represent the corresponding chemical shifts of the proton and the attached heteronucleus in the ligand‐free state; K is a sensitivity factor that normalises the chemical shift responses of proton and heteronucleus to changes in local environment. Fits (denoted by the straight lines) of the experimental Δδ (j) (circles) against ligand concentration to standard binding isotherms (1‐site binding in this case) yields the apparent binding affinity (K_D). (c) Generation of a double tyrosine‐to‐alanine mutant in the core region of a helical protein leads to a significant loss in chemical shift dispersion (red) compared to the wild‐type spectrum (blue) suggesting that the protein unfolds. This represents an example of NMR spectra being used to assess the structural effects of mutations and this approach can be used as a protein quality control step when performing structure‐based mutations to test binding or activity (in enzymes).

Article Title:	Protein Structure and Interactions from Nuclear Magnetic Resonance Spectroscopy
* Name:
* E-mail:
* Note: