NMR (nuclear magnetic resonance) investigation through the exploitation of paramagnetic effects is passing from an approach limited to few specialists in the field to a generally applicable method that must be considered, especially for the characterization of systems hardly affordable with other techniques. This is mostly due to the fact that paramagnetic data are long range in nature, thus providing information for the structural and dynamic characterization of complex biomolecular architectures in their native environment. On the other hand, this information usually needs to be complemented by data from other sources. Integration of paramagnetic NMR with other techniques, and the development of protocols for a joint analysis of all available data, is fundamental for achieving a comprehensive characterization of complex biological systems. We describe here a few examples of the new possibilities offered by paramagnetic data used in integrated structural approaches.
The interest of structural biologists in paramagnetism-based NMR (nuclear magnetic resonance) has grown over the last few decades because of the potentialities inherent in the long-range nature of the structural restraints it provides and of the possibility to obtain intrinsically unambiguous data. Investigation of systems for which classical techniques cannot be satisfactorily applied started to appear [1–6]. Paramagnetic effects chiefly depend on the electronic structure of the paramagnetic center, and thus on the local structure around it. Paramagnetic molecules can either be organic radicals or complexes containing a paramagnetic metal ion. Different metals impart different behavior on the nuclei, and the extent of the effects is dictated by the molecular structure . Pseudocontact shifts (PCSs) and paramagnetic relaxation enhancements (PREs) strongly depend on the distance (r) of the nuclei from the paramagnetic centers (PCSs as r−3, PREs are r−6). PCSs also depend on the paramagnetic susceptibility anisotropy tensor, which is determined by the details of the structure at the paramagnetic site, and on the angular positions of the nuclei in a common frame centered on the paramagnetic metal. The information on the structure and the mobility of biomolecules that these data contain are thus very precious for the characterization of systems for which different types of data can be hardly obtained, or can provide only partial/low-resolution snapshots, as in large macromolecular systems, systems exploring a large conformational variability, etc.
Of course, in order to measure paramagnetic data for a biomolecule, either the biomolecule is one of the naturally occurring metalloproteins (ca. 40% of the proteasome ), containing either a paramagnetic metal or a diamagnetic metal that can be substituted with a paramagnetic one, or a paramagnetic tag must be attached. To this aim, an extensive library of tags has been developed in the last years, allowing for rigid tagging of proteins through many different strategies [9–12]. The effort devoted to tag development testifies the widespread interest toward paramagnetism-based NMR and largely extends the number of biomolecular systems that can now be studied with this tool.
On the other hand, paramagnetic data alone usually do not suffice to provide a complete picture of the investigated biosystems. Therefore, integration of paramagnetic data with data from other sources (theoretical predictions, ‘diamagnetic' NMR, EPR (electron paramagnetic resonance), X-ray crystallography, small angle X-ray/neutron scattering, cryo-electron microscopy, etc.) and a subsequent joint analysis are fundamental for achieving a comprehensive structural characterization. The development of strategies for an optimized analysis of data from different sources, including paramagnetic NMR, thus represents a strategic step to open new promising perspectives in structural biology.
NMR resonance assignments
The first step in any NMR study is the assignment of the NMR spectra. This is a challenging work for large proteins, due to their intrinsic broader NMR signals and spectral crowding. Paramagnetism can be useful in this respect. The NMR shifts are, in fact, affected by the presence of a paramagnetic metal ion in a way that depends on the structure of the molecule. For nuclei more than few chemical bonds away from the paramagnetic metal ion, and outside its electron spin density distribution, the change in the NMR shifts is the PCS (Figure 1). Therefore, if the structure of the investigated molecule is known, for instance through X-ray crystallography, the PCSs of all protein atoms can be predicted and compared with the experimentally observed values. A correct assignment can thus be obtained in order to achieve a good correspondence between predicted and experimental data. This procedure is complicated by the degeneracy of the possible nuclear positions in agreement with the same PCS value, and by the fact the prediction of the PCSs passes through the knowledge of the susceptibility anisotropy tensor of the paramagnetic molecule, which is usually unknown. The problem of the degeneracy can be solved by acquiring PCS data collected for multiple paramagnetic metal ions alternatively substituted in the same or in different positions . The problem of estimating the five parameters defining the paramagnetic susceptibility anisotropy tensor can be addressed in different ways: (i) the tensor can be derived from general considerations about the metal coordination and, in the case of paramagnetic tags, known within a reasonable accuracy from previous studies; (ii) it can be built from the identification of at least five resonances which can be unambiguously ascribed to the corresponding nuclei. The tensor can then be iteratively refined after each step of resonance assignments achievable from the comparison between predicted and observed PCSs. The program PARAssign [14,15] uses PCSs obtained from several paramagnetic centers attached to the protein to obtain amide and/or methyl assignments, using the Hungarian method for minimal cost assignment, and can be used either when the positions of the paramagnetic centers are known or when they are unknown.
The hyperfine shift.
PCSs can be measured also in protein crystals through magic angle spinning solid state NMR [16,17]. In this case, they encode information not only on the protein structure but also on the arrangement of the protein molecules within the crystal [18,19]. PCSs result very useful for the assignment of the solid state NMR spectra needed to obtain distance restraints used for structure calculations. Spectroscopic crowding and 13C peak linewidths, in fact, in many cases only permit ambiguous assignments. On the other hand, if complemented with PCSs, the relatively few unambiguous assignments can allow for the calculation of a preliminary structure, which can be used to solve a large number of ambiguities in the spectra, extending the assignment of the latter and providing additional restraints in an iterative procedure [17,20,21].
The sizable paramagnetic relaxation induced on 1H nuclei can also assist in the assignment of the resonances of the nuclei close to the paramagnetic center . In a recent example, several resonances of the protein anamorsin [23,24] could be assigned using relaxation-tailored experiments, which also permitted to monitor the transfer of its [Fe2S2]2+ cluster .
NMR can be conveniently used to identify weak ligands of macromolecular targets as candidate drugs, providing at the same time information on the binding affinity and on the structure of the ligand-target complex [29,30]. Binding is revealed by a decrease in the intensity of the ligand NMR signals in the presence of the target, caused by the slower reorientation time of the complex and thus by a larger transverse relaxation rate. A larger increase in the ligand transverse relaxation rates can be achieved when the target protein contains a paramagnetic metal ion, so that PRE contributions are also present. In the case of weak binders of matrix metalloproteins, this effect permitted to decrease the concentration of the protein required to measure an effect by about a factor 5. Furthermore, whereas in diamagnetic samples the intensity of the NMR signals decreases similarly for all ligand nuclei, in paramagnetic samples the decrease is larger for ligand nuclei closer to the paramagnetic metal (Figure 2). Therefore, exploitation of PREs can be a strategy to enhance the sensitivity of ligand screening by NMR, providing at the same time information on the ligand pose . Localization of low affinity protein−ligand binding clefts is also proposed by detection of intermolecular protein PCSs, after covalent attachment of a paramagnetic binding tag to the ligand of interest . This has been shown to be of particular relevance in cases where conformations of the ligand may change during the binding event, for instance in weak protein–carbohydrate interactions [33,34].
The NMR signals of ligand nuclei are differently broadened upon binding to a paramagnetic protein.
Ligand PCSs measured in paramagnetic protein systems can also be used to determine the structures of the protein–ligand complexes, which can be challenging through other techniques for low-affinity complexes. Ligand PCSs measured in fast exchanging systems can, in fact, be used as docking restraints, after rescaling according to the bound fraction [29,35]. The magnetic susceptibility anisotropy tensor, needed for relating the experimental PCSs to the available (X-ray) structure of the protein, can conveniently be estimated from the PCSs measured for the (assigned) protein nuclei [36–38].
PREs as distance restraints
As already anticipated, paramagnetic centers can affect the NMR signals depending on the distance of the nuclei from the paramagnetic centers. Dissolving diamagnetic proteins into solutions containing paramagnetic molecules (either radicals or paramagnetic metal complexes) as co-solutes can thus allow for the identification of the protein residues on the surface of the molecule. Only these residues can, in fact, be approached within a short distance by the paramagnetic center, resulting in a large increase in their nuclear relaxation rate, with a consequent line broadening of the NMR signals [39,40]. Comparison of the NMR spectra acquired in the presence and in the absence of the paramagnetic co-solute can thus allow for the identification of the external residues. Using the same effect, the presence of intermolecular interactions and the molecular binding sites can be revealed by inspection of whether and which nuclear signals are broadened after the addition of a paramagnetic co-solute: the nuclei at the interaction site, in the presence of binding, behave as internal ones, although located in the external part of the isolated molecule (Figure 3). This approach, shown to be effective in solution even in large complexes, has been recently applied also in the solid state . Of note, co-solute paramagnetic molecules represent a tool to speed up data acquisition in NMR spectroscopy by shortening the nuclear relaxation time and thus the necessary recycle delay .
Co-solute paramagnetic molecules can allow detection of binding between two proteins and identification of the binding sites.
The exploitation of the NMR line broadening to identify binding protein surfaces has been proposed also to study the presence of weak interactions between proteins and nanoparticles . Under certain kinetic exchange regimes, the protein signals are exchange-mediated and remain visible even in the presence of binding to nanoparticles: the incorporation of paramagnetic centers into nanoparticles offers a high-sensitivity tool for mapping nanoparticle-binding surfaces on the protein. In addition, isothermal titration calorimetry measurements can be performed to investigate the thermodynamics of the interaction .
If a paramagnetic center is present within the molecule, either naturally or through the attachment of a paramagnetic tag (either a spin label or a paramagnetic metal complex), the observed PREs (typically measurable up to 20–25 Å from the effective paramagnetic centers) can be analyzed taking into account that they correlate with the distance between the paramagnetic center and the corresponding nuclei. This piece of information is precious as structural restraint in protocols aiming at calculating the structure of the investigated system, in conjunction with other types of data. For example, in solid state NMR, where distance restraints are difficult to obtain, PREs originating from the EDTA-Mn2+ complex attached to the B1 domain of streptococcal protein G (GB1) can drive toward the structure of the protein when used within the CS-Rosetta protocol . Analogously, in solution, the PREs measured for the spin-labeled actin in the presence of thymosin β4 permitted to calculate the structure of the complex .
Moreover, PREs can allow for monitoring the presence of conformational variability in solution. In the case of systems composed of two structural states, when the exchange rate is larger than the difference in PRE between the two conformers and slower than the reorientation time of the system, the observed PREs are population-weighted averages of the PREs in the two states. If the PRE of a nucleus in the minor state is significantly larger than the PRE in the major state, detection of the minor state becomes possible . Analogously, PREs can pinpoint the presence of transient, sparsely populated encounter complexes and identify the patches of residues in two interacting proteins that come into short-lived close contact with one another [2,47,48]. In the study of disordered proteins, PREs can indicate the main long-range interactions occurring in the protein mapped in a very small number of conformations, which can be as low as one . Of course, in all cases of conformational variability, either occurring in systems composed of well-folded or disordered domains, the ill-defined nature of the problem [50,51] prevents from recovering the real conformational ensemble, and integration of the information that can be available from different techniques is highly desirable.
Structure assembly, validation and refinement
Ab initio structure calculations of large molecular complexes by solution NMR can be very difficult and time-consuming. Paramagnetism can provide crucial information for driving the global fold of large proteins, which may result very challenging using only chemical shift information and NOE (nuclear Overhauser enhancement)-derived distance restraints, for the insufficient number of detectable long-range contacts. The successful PCS-based approach for the structure determination of the integral seven-helical membrane protein pSRII represents a nice example in this respect . PCSs, being long-range restraints describing the global fold of proteins, can provide the missing information needed to correctly determine large protein structures when included together with backbone chemical shifts into the ROSETTA fragment assembly program . This is particularly true when PCSs referring to multiple paramagnetic metals alternatively located in different positions of the protein surface are available . This approach proved successful even for in-cell studies , allowing for the determination of the structure of the B1 domain of protein G in intact Xenopus laevis oocytes from in-cell concentrations as low as 50 µM .
X-ray structures can be obtained more easily than NMR structures if crystallization is feasible. On the other hand, X-ray structures are obtained in conditions different from the physiological ones, and structural differences can arise due to crystal packing forces that are present in the solid state but not in solution . This problem can be particularly relevant for macromolecules composed of multiple domains connected by flexible linkers and for biomolecular complexes. Therefore, a possible approach to determine the structure of large systems can pass through (i) the validation and refinement of the crystallographic structures of the individual subunits present in the system, and (ii) the assembly of the validated/refined subunits to determine their overall structural arrangement. Paramagnetic data have been proved to be useful in both respects.
PCSs and paramagnetic residual dipolar couplings (RDCs) arising from the partial self-orientation of a paramagnetic molecule induced by the anisotropy of its magnetic susceptibility tensor [7,58] (i.e. in the absence of external orienting devices) are long-range accurate reporters of the structural details of the system. Both of them are related to the structural co-ordinates through the same paramagnetic susceptibility anisotropy tensor . PCSs, depending on the nuclear co-ordinates, can provide an assessment of the overall molecular conformation and an accurate estimate of the anisotropy tensor; RDCs, depending on the orientation of the coupled nuclei vector, are very sensitive to local structural details.
Overall, PCSs and RDCs can be used to validate a crystallographic structure in solution, and possibly to refine it within the indetermination allowed by the uncertainty in the X-ray data, or to pinpoint the regions exhibiting structural differences in solution (Figure 4). This can be done through a joint refinement including both NMR and X-ray data as structural restraints, using the program REFMAC-NMR [59,60]. If a conformation can be found fulfilling simultaneously both types of restraints, this conformation can be regarded as a refined structure valid both in solution and in the solid state; differently, a conformational rearrangement must occur on passing from one to the other state.
X-ray and NMR joint refinement.
Once the structures of the individual subunits have been validated and possibly refined, they can be assembled using the same restraints [13,37,61], possibly complemented by PREs [1,62] and/or data from sources different from NMR, like SAXS (small angle X-ray scattering) or cryo-electron microscopy . Double electron–electron resonance (DEER) experiments [64–67] may also be useful to better define the positions of the paramagnetic metal ions, which may not be exactly known if attached with protein-binding tags [68,69]. The strength of PCSs, collected for multiple paramagnetic metal ions to remove degeneracies, in providing structures through rototranslation of rigid subunits has been demonstrated for several protein–protein complexes [5,70–72]. PREs can also be effective, induced either by paramagnetic metal ions  or by spin labels .
Finally, we can predict that PCSs, even if measured only for nuclei far away from the paramagnetic metal ion, can be used to refine the molecular structure not only at the observed nuclei but also around the paramagnetic site. In fact, the paramagnetic susceptibility anisotropy tensor, which can be easily calculated from the collected PCSs and the protein structure, can be described by the electronic structure and the coordination environment, and can be calculated from the molecular g and zero-field splitting tensors . Therefore, given a structural model around the paramagnetic metal ion, first principles calculations of the g and zero-field splitting tensors, possibly complemented by EPR measurements, allow for the reconstruction of the magnetic susceptibility tensor, and thus for a comparison with the experimental tensor (Figure 5).
The magnetic susceptibility anisotropy tensor calculated from the molecular structure at the metal site and from the experimental PCSs.
Ranking structures in the presence of conformational variability
Conformationally averaged PCSs and RDCs, analogously to the already discussed averaged PREs, contain information on the weight of the different conformers sampled by the system [50,51,76–78]. The ill-posed nature of the problem, however, prevents from retrieving the real conformational ensemble, as an infinite number of different ensembles will be in agreement with the available average restraints. Nevertheless, PCS and RDCs can provide precious information for the characterization of the regions of the conformational space sampled by the system, and in particular to determine the conformations probably preferred or disfavored [50,51]. Using the MaxOcc approach , for instance, PCSs and RDCs, possibly complemented by PREs, DEER and/or SAXS/SANS (small angle neutron scattering) data, allow for a ranking of the sterically possible conformations of the system based on the maximum weight that each conformation can have taking into account all possible ensembles in agreement with the averaged restraints [80–83] (Figure 6).
Maximum Occurrence values (MaxOcc) of sterically possible conformations of the two-domain protein matrix metalloproteinase-1 have been calculated using PCSs and RDCs collected in the presence of paramagnetic lanthanoid(III) ions (orange spheres) attached to the catalytic domain of the protein (shown as ribbon).
We have shown that paramagnetic NMR is a toolbox useful to obtain structure and dynamic information even for systems hardly affordable with other techniques. It can assist, for instance, in the assignment of the NMR spectra, in drug discovery studies, in the definition of the interaction surfaces in protein complexes, in the refinement of crystal structures in solution of single proteins and of protein complexes, and in the characterization of the structural heterogeneity of biomolecular systems, including the detection of lowly populated conformational states. A full exploitation of the information contained in the paramagnetic data requires the use of integrated approaches, where complementary data are provided through other techniques.
Paramagnetic data assist in the assignment of the NMR spectra, the first step in any NMR study.
In drug discovery, paramagnetic data provide information on binding affinity and structure of the ligand-target complex.
Paramagnetic relaxation rates provide information on the protein–protein interaction surfaces and on the presence of lowly populated conformational states.
Paramagnetic data can be used to refine a protein structural model in conjunction with X-ray data, pointing out possible conformational differences between solution and solid state.
Paramagnetic data are quite precious for the characterization of the structural ensemble in systems exploiting interdomain conformational heterogeneity.
double electron–electron resonance
electron paramagnetic resonance
B1 domain of streptococcal protein G
nuclear magnetic resonance
paramagnetic relaxation enhancement
residual dipolar coupling
small angle neutron scattering
small angle X-ray scattering
The support from Fondazione Cassa di Risparmio di Firenze, MIUR PRIN 2012SK7ASN, European Commission project pNMR No. 317127, and Instruct-ERIC, a Landmark ESFRI project, is acknowledged.
The Authors declare that there are no competing interests associated with the manuscript.