Protein NMR Spectroscopy: Practical Techniques and Applications

By Gordon Roberts

Nuclear Magnetic Resonance (NMR) spectroscopy, a physical phenomenon based upon the magnetic properties of certain atomic nuclei, has found a wide range of applications in life sciences over recent decades. The dramatic advances in NMR techniques have led to corresponding advances in the ability of NMR to study structure, dynamics and interactions of biological macromolecules in solution under close to physiological conditions. This volume focuses on the use of NMR to study proteins.

NMR can be used to determine detailed three-dimensional structures of proteins in solution. Furthermore, it provides information about conformational or chemical exchange, internal mobility and dynamics at timescales varying from pcoseconds to seconds. It is the primary technique used to obtain information on intrinsically disordered (unfolded) proteins, since these proteins will not crystallize easily. NMR is also a very powerful method for the study of interactions of protein with other molecules, whether small molecules (including drugs), nuclear acids or other proteins.

This up-to-date volume covers NMR techiniques and their application to proteins, with a focus on practical details. This book will provide a newcomer to NMR with the practical guidance in order to carry out successful experiments with proteins and to analyze the resulting spectra. Those who are familiar with the chemical applications of NMR will also find is useful in understanding the special requirements of protien NMR.

• Language: English
• Publisher: Wiley
• Release date: Jun 9, 2011
• ISBN: 9781119972822

Book preview

List of Contributors

Igor Barsukov, NMR Centre for Structural Biology, The University of Liverpool, School of Biological Sciences, Biosciences Building, Crown Street, Liverpool L69 7ZB, United Kingdom

Pau Bernadó, Institute for Research in Biomedicine, c/Baldiri Reixac 10, 08028-Barcelona, Spain

Martin Blackledge, Institut de Biologie Structurale, UMR 5075 CEA-CNRS-UJF, 41 Rue Jules Horowitz, Grenoble 38027, France

Paul Driscoll, Division of Molecular Structure, MRC National Institute for Medical Research, The Ridgeway, Mill Hill, London NW7 1AA, United Kingdom

Alex Grishaev, Laboratory of Chemical Physics, National Institute of Diabetes and Digestive and Kidney Diseases, National Institutes of Health, 5 Memorial Drive, Bethesda, MD 20892

Stephan Grzesiek, Biozentrum, University of Basel, Klingelbergstrasse 50/70, CH-4056 Basel, Switzerland

Peter Güntert, Institut für Biophysikalische Chemie, BMRZ, J.W. Goethe-Universität Max-von-Laue-Str. 9, 60438 Frankfurt am Main, Germany

Malene Ringkjøbing Jensen, Institut de Biologie Structurale, UMR 5075 CEA-CNRS-UJF, 41 Rue Jules Horowitz, Grenoble 38027, France

Masatsune Kainosho, Center for Structural Biology, Graduate School of Science, Nagoya University, Chikusa Ku, Furo Cho, Nagoya, Aichi 4648602, Japan

Peter H.J. Keizers, Leiden Institute of Chemistry, Leiden University, Gorlaeus Laboratories, P.O. Box 9502, 2300 RA Leiden, The Netherlands

Lu-Yun Lian, NMR Centre for Structural Biology, The University of Liverpool, School of Biological Sciences, Biosciences Building, Crown Street, Liverpool L69 7ZB, United Kingdom

Phineus Markwick, Howard Hughes Medical Institute, 9500 Gilman Drive, La Jolla, California 92093-0378, USA

Frederick W. Muskett, Henry Wellcome Laboratories of Structural Biology, Department of Biochemistry, University of Leicester, PO Box 138, Lancaster Road, Leicester LE1 9HN, United Kingdom

Gabrielle Nodet, Institut de Biologie Structurale, UMR 5075 CEA-CNRS-UJF, 41 Rue Jules Horowitz, Grenoble 38027, France

Valéry Ozenne, Institut de Biologie Structurale, UMR 5075 CEA-CNRS-UJF, 41 Rue Jules Horowitz, Grenoble 38027, France

Gordon Roberts, Henry Wellcome Laboratories of Structural Biology, Department of Biochemistry, University of Leicester, PO Box 138, Lancaster Road, Leicester LE1 9HN, United Kingdom

Loic Salmon, Institut de Biologie Structurale, UMR 5075 CEA-CNRS-UJF, 41 Rue Jules Horowitz, Grenoble 38027, France

Yang Shen, Laboratory of Chemical Physics, National Institute of Diabetes and Digestive and Kidney Diseases, National Institutes of Health, 5 Memorial Drive, Bethesda, MD 20892

Mitsuhiro Takeda, Center for Structural Biology, Graduate School of Science, Nagoya University, Chikusa Ku, Furo Cho, Nagoya, Aichi 4648602, Japan

Nico Tjandra, Laboratory of Molecular Biophysics, National Heart, Lung, and Blood Institute, National Institutes of Health, 50 South Drive, Bethesda, Maryland 20892, USA

Marcellus Ubbink, Leiden Institute of Chemistry, Leiden University, Gorlaeus Laboratories, PO Box 9502, 2300 RA Leiden, The Netherlands

Geerten Vuister, Protein Biophysics, Institute of Molecules and Materials, Radboud University Nijmegen, PO Box 9101, 6500 HB Nijmegen, The Netherlands; and Henry Wellcome Laboratories of Structural Biology, Department of Biochemistry, University of Leicester, PO Box 138, Lancaster Road, Leicester LE1 9HN, United Kingdom

Introduction

Lu-Yun Lian and Gordon Roberts

The nuclear magnetic resonance (NMR) method is one of the principal techniques used to obtain physical, chemical, electronic and three-dimensional structural information about molecules in solution, whether small molecules, proteins, nucleic acids, or carbohydrates. NMR is a physical phenomenon based upon the magnetic properties of certain atomic nuclei. When exposed to a very strong magnetic field (2–21.1 Tesla) these nuclei align with this field. During an NMR experiment, the alignment is perturbed using a radiofrequency signal (typically a few hundred megahertz). When the radio transmitter is turned off, the nucleus returns to equilibrium and in the process re-emits radio waves. The usefulness of this technique in biochemistry results largely from the fact that nuclei of the same element in different chemical and magnetic environments give rise to distinct spectral lines. This means that each NMR-active atom in a large molecule such as a protein can be observed and can provide information on structure, conformation, ionisation state, pKa, and dynamics. The nuclei which are most relevant to the study of biological macromolecules are shown in Table 1. The proton ( ) is the most sensitive nucleus for NMR detection. For biological studies, and are now just as important, although enrichment with these stable isotopes is necessary.

Table 1 Properties of nuclei of interest in NMR studies of proteins.

The first published NMR spectrum of a biological macromolecule was the 40 MHz spectrum of pancreatic ribonuclease reported in 1957. Since then, the significant milestones for NMR include:

Fourier Transform NMR in the late 1960s;

the development of two-dimensional NMR in the early 1970s;

the development of INEPT/HMQC pulse sequences in the late 1970s;

the application of NMR to solve the full three-dimensional structure of a protein in solution in the early 1980s;

the introduction, in the late 1980s, of three- and four- multidimensional heteronuclear experiments for use with / isotopically labelled proteins, followed in the late 1990s by the TROSY experiments (which require protein deuteration in addition).

Each stage of these developments has been accompanied by improvements in the spectrometer hardware. In particular, the increases in magnetic field strengths, improved probeheads such as cryogenically-cooled probeheads and better electronics have together led to very substantial improvements in resolution and sensitivity. In addition, continuous advances in molecular biology and sample preparation have allowed these NMR-based improvements to be exploited, particularly in the speed with which samples of significant quantities can be produced in a cost-effective way and the ease with which stable isotope enrichment can be accomplished. Finally, the data analysis is now more streamlined and in the case of very high-quality data, the structure determination process, from resonance assignment to structure calculation, can be automated. These developments are important in order for NMR to remain a mainstream technique for high-resolution structure determination and to make significant contributions in structural biology.

For structural biology, NMR is unique in that it can be used for studies of macromolecules in both the solution and solid states and it is, furthermore, the only method that can provide information on dynamics at the atomic level. This book focuses on the use of NMR to study protein structure and interactions in solution, with the aim of providing a practical guide to users of the method. The book attempts to deal with methods, approaches and issues commonly encountered in the everyday use of NMR in structural biology. No attempt is made to provide a description of the fundamental physics of NMR, but in some chapters it is necessary to detail the theoretical aspects of the methodology in order that the methods can be appropriately applied. A full discussion of the fundamental basis of the wide range of solution NMR experiments used in structural biology can be found in [1] and other valuable introductions to modern NMR spectroscopy include [2, 3].

The success of any application of NMR depends on the correct sample preparation, the appropriate use of parameters for data acquisition and processing; these are covered in Chapter 1. Once initial data has been collected to assess if a protein system is suitable for NMR studies, the next step will depend upon whether the objective is a determination of the three-dimensional structure of a protein or its complex, or a more limited specific objective such as screening for ligand binding or the determination of pKa values. For all but the smallest proteins, isotope-labelling will be required (Chapter 2), and to go beyond purely qualitative experiments resonance assignments (Chapter 3) will be essential. Structure determination involves the acquisition and treatment of structural restraints (Chapter 4) and the use of these to obtain structural ensembles (Chapter 5). Chapter 6 describes the additional information on protein structure or complex formation which can be obtained when the protein contains a paramagnetic species – either naturally occurring or introduced specifically for the purpose. Chapter 7 describes different approaches to the study of the binding of small molecules, ranging from screening to full structure determination of the complex; this requires an understanding of the theoretical and practical aspects of the effects of chemical exchange in NMR, which is also important in many other areas of biological NMR. Chapter 8 provides a comprehensive description of the use of NMR to study macromolecular complexes; this is a challenging area and the chapter outlines the problems and approaches which can be taken to overcome these challenges. Chapter 9 focuses on the structural studies of intrinsically disordered proteins. The widespread existence and significance of these proteins are becoming increasingly recognised and NMR is currently the best method to provide detailed information on their conformational distributions.

NMR is uniquely suited for the characterisation of biomolecular dynamics. Since so many nuclei can be detected simultaneously, NMR can provide a comprehensive description of the internal motions and conformational fluctuations at atomic resolution, and NMR methods have been developed to quantify motions that occur at a wide range of timescales, from picoseconds to days and months. At the same time, consideration of dynamics and the averaging processes to which they lead is an essential part of the use of NMR to obtain structural information. As a result, several chapters in this book deal with methods for obtaining dynamic information from NMR. For additional information the reader is also directed to the following reviews [4–7].

Over the last few years there have been significant developments in the application of solid-state NMR techniques as a tool for determining the high-resolution structures of proteins, ranging from microcrystalline soluble proteins to protein fibrils and membrane proteins. It is now possible to assign the spectra of proteins larger than 100 amino acids using , –labelling [8]. However, as yet this remains an area for the expert and it is not covered in detail in this book (although some of the methods for isotope-labelling described in Chapter 2 will also be relevant to solid-state studies). For useful reviews, the reader is directed to [9, 10].

Note Added in Proof

Several valuable relevant reviews have appeared while this book was in production. In particular, two useful qualitative introductions to biomacromolecular NMR for the newcomer to the field would serve as valuable initial reading [11, 12]. Clore [13] has reviewed the use of relaxation methods (see Chapters 6 and 7) to observe species with low population, Wishart [14] has reviewed the use of chemical shifts in structure determination (see Chapters 4 and 5), and Dominguez et al. [15] have reviewed the use of NMR in the study of protein-RNA complexes (see Chapter 8).

References

1. Cavanagh, J., Fairbrother, W.J., Palmer, A.G. III et al. (2007) Protein NMR Spectroscopy: Principles and Practice, 2nd edn, Academic Press, San Diego.

2. Keeler, J. (2005) Understanding NMR Spectroscopy, John Wiley & Sons, Ltd, Chichester.

3. Levitt, M.H. (2001) Spin Dynamics: Basis of Nuclear Magnetic Resonance, John Wiley & Sons, Ltd, Chichester.

4. Mittermaier, A.K. and Kay, L.E. (2009) Observing biological dynamics at atomic resolution using NMR. TIBS, 34, 601–611.

5. Baldwin, A.J. and Kay, L.E. (2009) NMR spectroscopy brings invisible protein states into focus. Nature Chem. Biol., 5, 808–814.

6. Jarymowycz, V.A. and Stone, M.J. (2006) Fast time scale dynamics of protein backbones: NMR relaxation methods, applications, and functional consequences. Chem. Revs., 106, 1624–1671.

7. Igumenova, T.I., Frederick, K.K. and Wand, A.J. (2006) Characterization of the fast dynamics of protein amino acid side chains using NMR relaxation in solution. Chem. Revs., 106, 1672–1699.

8. Schuetz, A. et al. (2010) Protocols for the sequential solid-state NMR spectroscopic assignment of a uniformly labeled 25kDa protein: HET-s(1-227). ChemBioChem., 11, 1543–1551.

9. Renault, M., Cukkemane, A. and Baldus, M. (2010) Solid-state NMR spectroscopy on complex biomolecules. Angew. Chem. Int. Ed., 49, 8346–8357.

10. McDermott, A. (2009) Structure and dynamics of membrane proteins by magic angle spinning solid-state NMR. Ann. Rev. Biophys., 38, 385–403.

11. Kwan, A.H., Mobli, Gooley, P.R. et al. (2011) Macromolecular NMR spectroscopy for the non-spectroscopist. FEBS Journal, 278, 687–703.

12. Bieri, M., Kwan, A.H., Mobli, M. et al. (2011) Macromolecular NMR spectroscopy for the non-spectroscopist: beyond macromolecular solution structure determination. FEBS Journal, 278, 704–715.

13. Clore, G.M. (2011) Exploring sparsely populated states of macromolecules by diamagnetic and paramagnetic NMR relaxation. Protein Sci., 20, 229–246.

14. Wishart, D.S. (2011) Interpreting protein chemical shift data. Prog. Nucl. Magn. Reson. Spectrosc., 58, 1–61.

15. Dominguez, C., Schubert, M., Duss, O. et al. (2011) Structure determination and dynamics of protein–RNA complexes by NMR spectroscopy. Prog. Nucl. Magn. Reson. Spectrosc., 58, 62–87.

Chapter 1

Sample Preparation, Data Collection and Processing

Frederick W. Muskett

Purgamentum init, exit purgamentum

1.1 Introduction

The power of NMR spectroscopy for the analysis of biological macromolecules is undisputed. During the last two decades, the development of spectrometers and the experiments they perform, software, and the molecular biological techniques for the expression and purification of proteins have progressed at a formidable rate. Enrichment of molecules in the three major isotopes used in NMR ( , and ) is now commonplace and the cost is no longer prohibitive. The software used to analyse the plethora of data we can generate makes spectral assignment and the extraction of data straightforward for all but the most challenging systems. With all these developments it is easy to forget some of the more fundamental requirements for obtaining good quality NMR data, namely a good sample and a well-set-up NMR experiment.

1.2 Sample Preparation

The first, and possibly one of the most important, steps before embarking on an NMR-based project is the preparation of the sample. Spending some time optimising sample conditions for concentration, ionic strength, pH and temperature before collecting large amounts of data will pay dividends, particularly if the sample is difficult and expensive to produce. Ideally, the optimised sample will not only give the best possible NMR data, but will also have long-term stability as, assuming the project requires backbone and side-chain assignments, the total acquisition time required can be in the order of several weeks.

The following sections will outline the general requirements of a biological sample that is to be used to record NMR data. The assumption has been made that full resonance assignment is required; however, these guidelines could, and probably should, be applied to all samples regardless of the intention of the experiments.

1.2.1 Initial Considerations

This optimisation can be performed in the NMR spectrometer but much can be done using other biophysical techniques such as circular dichroism or fluorescence spectroscopy. These methods require much lower concentrations and do not require isotopic enrichment. The effects of buffer composition on secondary structure content and the melting temperature of the sample give a useful starting point. Once the initial conditions have been determined, final optimisation in the spectrometer can begin. If the sample is a protein, although much can be learned from a simple one-dimensional proton experiment, by far the most useful experiment is the -edited HSQC. This type of experiment removes a great deal of resonance overlap, allowing the user to see in much more detail the effects of varying pH, ionic strength and temperature.

NMR has intrinsically poor sensitivity and, as a result, the concentration of the sample needs to be in the millimolar range. For a conventional room temperature probe ideally the sample concentration needs to be ≥1 mM but can be as low as 0.5 mM. With the development of cryogenically cooled probes this concentration can be reduced to ≥0.2 mM and given the right sample can be as low as 0.05 mM (depending on the experiments performed). Whilst the sample can be exchanged into the buffer intended for NMR experiments in the last step of purification (usually gel filtration chromatography) it is rarely at the concentration required. The two main methods used to increase concentration are lyophilisation with subsequent re-suspension in a lower volume and ultra-filtration, or a combination of the two. Unfortunately, whether a sample will survive either method cannot be predicted; in the end one must simply try and see what happens. However, lyophilisation is generally considered the more dangerous of the two. The number and type of disposable ultra-filtration devices on the market is large, each with their own characteristics regarding compatibility with a particular sample and the effective volumes with which they can be used; again, try and see what happens. As a final step, either passing the sample though a 0.2 μm filter or centrifuging in a benchtop micro-centrifuge, to remove any insoluble material or dust, will greatly help sample homogeneity.

With modern solvent suppression techniques that can effectively eliminate the 110 M protons from the water signal, dissolving the sample in would, at first, no longer seem to be required. However, such samples are still important in recording the experiments designed to allow assignment of protein side-chain resonances and for -edited nuclear Overhauser effect (NOE) experiments. Even the most efficient solvent suppression techniques still leave residual solvent signal and at the same time suppress or distort the signals of interest in that area of the spectrum. In addition, the use of allows one to carry out experiments in which the coherences are recorded in the directly detected dimension where they have the highest resolution. These two advantages alone outweigh the effort required to transfer the sample into . Methods for exchanging the solvent for are the same as for concentrating samples, either lyophilisation or repeated concentration and dilution with . Alternatively, if the sample is unlikely to survive those methods, the sample can be passed down a short de-salting gel-filtration column that has been pre-equilibrated in .

1.2.2 Additives

As many NMR experiments require hours or days to complete, addition of anti-microbial agents is highly recommended. Sodium azide at a concentration of 0.02 % w/v is an almost universal method; however, in the rare cases where the azide ion interacts with the sample (e.g. some cytochromes) micromolar concentrations of an antibiotic such as ampicillin or chloramphenicol can be substituted. EDTA or AEBSF™ are frequently added to NMR samples at a concentration of 0.1–5 mM in order to reduce proteolysis. However, these compounds have nonexchangeable protons that can interfere with the spectrum of the sample. Excessive use of these compounds is best avoided, a better approach being to improve the purification protocol. If the protein sample contains free cysteines reducing agents such as DTT or TCEP™ are required to stop the protein forming dimers or multimers, which can result in precipitation. In addition, degassing the sample can help; however samples inevitably re-dissolve oxygen during subsequent sample manipulations or during the course of the NMR experiment unless special care is taken to seal the NMR tubes.

1.2.3 Sample Conditions

Although the primary choice of buffer must be that which promotes long-term stability of the sample, some buffer salts are more convenient for NMR than others. As buffer concentrations are typically between 10–50 mM, any covalently bonded protons in the buffer will give rise to sharp and obtrusive signals in the spectrum. This has resulted in phosphate buffer being the primary choice if its buffering range is appropriate to your sample and if it does not interact with your protein – though many proteins bind ligands containing phosphate groups, from ATP to phosphoproteins and DNA, and may bind inorganic phosphate weakly. Otherwise, many of the more common buffer salts are available with deuterium replacing the nonlabile protons. When selecting a pH it should be borne in mind that the exchange rates of amide protons are such that pH values between 3 and 7 are most conducive to observing the signals arising from these groups.

For many biological samples, the addition of salts (typically sodium chloride) to the buffer increases solubility and decreases aggregation. Unfortunately, in NMR the dielectric losses at high ionic strength (greater than 150 mM) are severe, particularly in cryogenically cooled probes and at high magnetic fields. Such losses can be dramatic, as degrading the signal-to-noise ratio twofold results in a fourfold increase in acquisition time to achieve comparable spectra. Recently, alternatives to traditional buffer systems have be proposed, such as dipolar ions [1], low conductivity buffers [2] or the use of ‘solubilising salts’ [3]. The general applicability of these alternatives is yet to be realised; however, they should be investigated if the sample has low solubility and/or requires high ionic strength for stability.

The final parameter to consider in optimising the sample conditions is the temperature at which the experiments are to be performed. As the temperature increases, the correlation time of the molecules decreases and so the resonances become narrower. In addition, varying the temperature will lead to changes in the chemical shift of temperature sensitive groups and may help to resolve any resonance overlaps. Typically, NMR experiments are performed in the temperature range of 293–308 K but if you have determined the thermal stability of the sample in advance (i.e. by CD spectroscopy), you will have a better idea of the attainable upper limit.

1.2.4 Special Cases

There are two types of sample that require extra attention: integral membrane proteins and samples intended for ligand titrations.

The use of solution NMR methods to study membrane proteins, although not mainstream, is now feasible. The solubilisation of integral membrane proteins in detergent micelles is relatively straightforward and the procedures for preparing membrane protein samples are essentially as described. There is much debate about which detergents are best for preparing NMR samples and it is apparent that no one detergent will suit all proteins. As a result, screening of several different detergent types at different concentrations is required. In addition, significant improvement of the spectrum can be achieved by using sample temperatures significantly higher than those used for soluble proteins (>310 K). The reader is referred to some excellent reviews on sample optimisation [4–6].

The study of ligand interactions via NMR is a well-established technique, enabling the identification of a specific binding site and the determination of kinetic information (see Chapter 7). The usual method for obtaining this information is to run successive spectra with increasing concentrations of ligand whilst observing the spectral changes that result. However, there is a danger that addition of the ligand will lead to changes in the sample conditions other than those due to ligand binding, resulting in artefactual/artificial changes in the spectrum. It can be difficult to obtain identical buffer conditions on mixing different proportions of two or more samples even if they have been dialysed against the same buffer but in separate dialysis tubes, as many biological macromolecules have a high affinity for electrolytes. In addition, if the ligand is a small molecule it may be difficult to solubilise and impossible to dialyse into the same buffer as the macromolecule.

The major concern in mixing two samples together or adding a ligand to a sample is a change in pH, as this alone can result in considerable chemical shift changes in the molecule of interest, as any ionisable groups change state. Fortunately, this property of ionisable groups can be used to monitor the pH of the NMR sample. Addition of a small molecule (e.g. imidazole) to the sample can be used to monitor the sample's pH as additions of ligand are made. Any pH shift will become immediately apparent as the chemical shift of the imidazole resonances will change. A number of these molecules have recently been characterised and the reader is referred to this article for more details [7].

In order to minimise the number of manipulations during a titration experiment, and so reduce the likelihood of systematic errors, the following procedure is recommended (see also Chapter 8 for a more detailed description). Two samples should be prepared, one of the biological macromolecule alone and the other with the macromolecule and the ligand at the concentration of its maximum titre – estimated, for example, by a biochemical assay. A spectrum is recorded of each sample, giving the initial and end point of the titration series. Assuming 600 μl sample volumes and that the titration series is to be incremented in steps of 0.1 molar equivalents of ligand, 60 μl of each sample is removed and mixed with the other so giving ratios of 1 : 9 and 9 : 1. Spectra are again recorded, so giving the next highest and lowest points in the titration series. This procedure is repeated until both samples are of equal concentration of macromolecule and ligand, at which a spectrum of each is recorded and compared. At this point, the spectra should, of course, be identical unless an error was made. The advantages of this method are that the concentration of the macromolecule never changes due to dilution by the addition of ligand and also, because there is a clearly defined end-point, any errors, or sudden changes in the sample condition, will not go unnoticed.

1.2.5 NMR Sample Tubes

For high-resolution NMR experiments, the homogeneity of the magnetic field in which the sample sits must of course be very high. This is most easily achieved with a small sample volume, but on the other hand, the sensitivity – always limiting in an NMR experiment – sets a limit to how small a sample can be used. The sample is usually housed in a 5 mm diameter tube; the optimal volume of sample for the highest signal-to-noise ratio and magnetic field homogeneity (and hence the optimal resonance lineshape) will be dependent on the probe design and will vary between manufacturer and between probe generations from the same manufacturer. This information should be provided by the manufacturer but is usually of the order of 600 μl in a conventional 5 mm NMR tube.

However, the use of ‘standard’ 5 mm tubes is now in decline due to the development of a reduced volume symmetrical microtube by Shigemi Inc., commonly referred to as a ‘Shigemi tube’. These tubes have a plug of susceptibility matched glass at the bottom of the tube with the equivalent in the form of a plunger that forms the top of the sample. The advantage of these tubes is that the sample volume need only match the susceptible volume of the transmitter/receiver coils of the probe. The glass plugs eliminate the change in magnetic susceptibility of the solvent/glass or solvent/air interface and hence the ‘end effects’ associated with short samples observed in standard NMR tubes. Therefore, sample volumes can be reduced to 300–350 μl. The temptation to reduce the sample volume even further should be resisted as the resonance lineshape, particularly of the residual water, deteriorates rapidly.

One drawback to the use of these tubes is that samples tend to degas over a period of hours at temperatures much above 298 K. This results in air bubbles forming at the top of the sample, which has disastrous effects on the sample homogeneity and, as a result, on the residual water resonance. Therefore, the sample should be pre-incubated at the required temperature prior to insertion of the sample in the magnet. (Alternatively, the sample homogeneity should be checked several hours after inserting the sample in the magnet and before starting a long NMR experiment.)

If conventional 5 mm tubes are to be used, selecting the correct grade is important. Although the intrinsic linewidth of a high molecular weight biological sample is high (i.e. for human ubiquitin, Mr ≈ 8.5 KDa, amide resonances range between 10 and 15 Hz) using cheap tubes limits the attainable field homogeneity. However, due to these larger intrinsic linewidths extremely high precision, and therefore expensive, NMR tubes are not really required and those graded for use in 500 MHz and above will suffice.

Whichever style of NMR tube is selected, none should be used directly from the box. The chemicals used during their manufacture will result in intense signals contaminating the spectra of the sample. Soaking in commercially available laboratory detergents (e.g. Decon 90™) is usually sufficient for both new and used tubes. However, if a sample has precipitated, the residue adhering to the glass can be particularly stubborn and soaking in a strong mineral acid may be required. However, considering the cost of making the sample in the first place, perhaps such tubes should be consigned to the glass bin. A variety of NMR tube cleaners are commercially available (e.g. Sigma-Aldrich or GPE Scientific Ltd); these are extremely useful for ensuring the tubes are thoroughly rinsed after soaking in either detergent or acid. If the sample is to be dissolved in , the tubes may be soaked in to exchange the residual water bound to the glass. Once rinsed, tubes are best dried in a stream of dry nitrogen gas as heating the tubes can cause distortions which will adversely affect field homogeneity.

1.2.5.1 3 mm Tubes

As discussed previously, the ionic strength of a sample can have profound effects on the signal-to-noise ratio of the observed spectrum. If the sample cannot tolerate low levels of salt then a final option is to use 3 mm NMR tubes, even if the probe is designed to be used with 5 mm tubes. In effect, the sample adds a resistance to the receiver coils of the probe; therefore reducing the volume of sample reduces this additional resistance and so does not degrade the overall sensitivity of the probe. Figure 1.1 shows the effect of increasing ionic strength on the signal-to-noise ratio observed in a 600 MHz cryogenically cooled probe when using a standard 5 mm, 3 mm or Shigemi tube.

Figure 1.1 The effect of ionic strength on the signal-to-noise ratio observed in a 600 MHz cryogenically cooled probehead. Signal-to-noise was estimated using the signal of a 0.5 mM solution of DSS dissolved in²H2O. The ionic strength of the samples was obtained using potassium chloride

At low ionic strength (i.e. <75 mM) signal-to-noise is dominated by the total amount of sample in the susceptible volume of the transmitter/receiver coils of the probe. Once the ionic strength reaches 100 mM, the signal-to-noise ratio for the 3 mm and 5 mm tubes is equivalent, even though there is only 30 % of the sample in the 3 mm tube. Shigemi tubes give higher signal-to-noise ratios until extremely high ionic strengths are reached (>400 mM), due to a combination of sample volume and the physical dimensions of the sample.

1.3 Data Collection

The initial four steps required to obtain an NMR spectrum are identical regardless of the type of experiment to be carried out. These are: to establish field/frequency lock, to tune and match the probe, to optimise the magnetic field homogeneity (shimming) and to calibrate the radio frequency pulse lengths. It is recommended that the user follows these four operations every time a sample is used, even if it has been used frequently or recently, as changes may indicate a problem, either with the sample itself or with the spectrometer. Depending on the age of the spectrometer, these steps may now be either fully or partially automatic. However, knowledge of how to perform these steps manually is essential if the sample is challenging or if better than merely ‘good’ data is required.

The assumption is made throughout this section that the NMR spectrometer to be used is well maintained, i.e. all heteronuclear pulses are properly calibrated, the VT unit is calibrated and the magnetic field homogeneity is optimised. Additionally, no attempt has been made to discuss which experiments should be collected or the theory behind them. Such a discussion is well beyond the scope of this chapter and the reader is directed to later chapters in this volume and to the following excellent discussions [8–11].

1.3.1 Locking

Once the sample is safely in the magnet and its temperature has equilibrated the first step is to establish a ‘field-frequency lock’. Even though the magnetic field produced by the superconducting magnet is extremely stable, there is some drift that will adversely affect the lineshape of the spectrum. The function of locking is to ensure that the magnetic field the sample experiences is constant. The use of deuterium as the locking nucleus is now universal and in the case of samples dissolved in H2O, should be added at a level of 5–10 %.

To lock, the magnetic field is first adjusted to bring the deuterium signal on resonance with the lock frequency of the spectrometer. Care should be taken that too much power is not used on the lock transmitter or the magnetic field stability will be poor and shimming will be unresponsive. Optimal lock transmitter power can be found by increasing the lock power stepwise, and observing the lock signal. Once the signal becomes ‘saturated’, indicated by a rise and then drop in lock level when the power is increased, lock power should be reduced by several dB such that a stable lock signal is obtained. Lock phase should also be adjusted to give the highest signal, ensuring that the lock circuit is at its most sensitive, thus providing optimal stability to the system.

1.3.2 Tuning

Putting a biological sample into the transmit/receive coils of an NMR probe affects the tuned circuit of the coil used to excite the sample and to detect the NMR signal. Therefore, the probe needs to be tuned back to the correct resonance frequency and matched to the correct impedance. The tuning of the observe coils (usually proton) is strongly sample dependent and proper tuning is essential to obtain the highest sensitivity. This circuit should be tuned and matched to each sample and the procedure repeated if, for example during a titration experiment, additions are made to the sample. Conversely, the indirect or heteronuclear coils of the lock circuit are reasonably insensitive to the sample and do not need routine tuning. Modern spectrometers have a visual feedback (the precise nature of which is manufacturer dependent) to facilitate the tuning and matching procedure as each requires the physical adjustment of capacitors in the circuit. As the two adjustments tend to interact with each other this process can be troublesome. However, assuming the tuning of the circuit is not too far off, if an iterative approach is adopted, i.e. the match is first fully optimised, and then the tune and so on until no improvement is achieved, the process should be relatively straightforward.

1.3.3 Shimming

The superconducting magnet of the spectrometer alone cannot produce a magnetic field homogenous enough for NMR experiments. Therefore, additional room-temperature electromagnetic coils, known as ‘shim-coils’, are placed around the sample and are used to cancel out the residual inhomogeneities of the main magnetic field. These inhomogeneities are either inherent to the magnet itself or are introduced when the sample is put into the bore of the magnet. The process of correcting these inhomogeneities is known as shimming. Shimming can be carried out either manually or semi-automatically using gradient based techniques. The affect of adjusting a shim coil can be observed in one of two ways, by observing the Free Induction Decay (FID) or the lock level. As most biological samples will at some time be dissolved in a nondeuterated solvent it is usual to shim these samples using the lock level. The intense signal from the water decays much more rapidly than expected due to the phenomenon known as radiation damping, consequently shimming on the FID is particularly unproductive for these samples.

Modern spectrometers may have up to 40 shim coils named according to the field profile they generate. Due to the field profile of the shims, the higher order shims interact with the lower order shims such that overall shimming is not a first order process. As a consequence, the manual shimming of an NMR magnet has become shrouded in mystery and folklore and there are probably as many protocols to obtain homogeneity as there are NMR spectroscopists. Coupled with the rapid development of automatic gradient-based shimming methods, manual shimming is slowly being ousted to the point that many structural biologists cannot manually shim their samples. But, however robust the gradient-based methods are they can, and do, fail to shim satisfactorily and then the spectroscopist must intervene. The following texts give extensive descriptions of the shimming process and protocols that the reader might try [8, 12, 13]. As almost all studies of biological macromolecules depend extensively upon 2D and 3D experiments, these samples are never spun, and unlike those for small molecules, these shimming protocols do not include the use of sample spinning.

Biological samples usually have linewidths of 10 Hz or more and this should be taken into account during the shimming process. There is little to be gained in spending hours shimming a sample in an attempt to reduce the solute linewidths by a few fractions of a hertz. However, the presence of a very strong solvent signal in samples dissolved in H2O makes reaching ‘good’ shimming much more critical. For these samples, the lineshape of the residual water resonance is important, as a poorly shimmed sample will have long tails from the intense solvent peak obscuring solute signals. In addition, the spectrum will have a distorted baseline that can be particularly detrimental in multidimensional experiments.

An easy way to assess the quality of the shimming in an H2O sample is to acquire a one-dimensional pre-saturation experiment. After optimisation of the transmitter offset to maximise solvent signal suppression the solvent signal should be easily suppressed using a pre-saturation duration of 1.5 sec and a maximum B1 field of 100 Hz, with the residual solvent signal being no more than 150 Hz wide at its base. Such a residual signal should be achievable even on a cryogenically cooled probe where radiation damping is even more of a problem. Failure to achieve such a ‘lineshape’ will result in poor quality spectra. Although pre-saturation is rarely used in modern NMR experiments it is nevertheless a good test of shimming, as failure to adequately suppress the solvent by this method will not be compensated for by more sophisticated methods (e.g. watergate [14]).

1.3.4 Calibrating Pulses

All but the most basic NMR experiments used to study biological samples are multipulse experiments that depend on applying pulses with the correct flip-angle. Even the simplest ‘pulse and acquire’ experiment benefits from a properly calibrated 90° pulse as then maximum intensity is obtained. The heteronuclear (i.e. , and ) pulse widths are insensitive to the sample and once calibrated need only checking occasionally (e.g. every six months or so, unless there has been a change in hardware) and do not need recalibrating for each sample. Calibration of heteronuclear pulses is described in detail in [15]. However, proton pulse widths are sensitive to the sample, particularly ionic strength, and should be calibrated each time the sample is used or if anything is added to the sample. Calibration of proton pulse widths is a straightforward process providing a few simple rules are followed: place the signals of interest in the centre of the spectrum and allow a long enough relaxation delay for the signals to relax fully. Normal practice is to increase the flip angle until the first signal maximum, or the first or second signal null is found. A coarse calibration is performed first, i.e. linearly increasing the pulse width in 4 μsec steps, then when an approximate value is found a finer-grained calibration can be performed. The null methods are more precise as the maxima are rather broad, and if the 360° null is chosen a shorter relaxation delay can be used. Once the 360° null is found, simply dividing by four gives the 90° pulse width. If the sample is dissolved in H2O then it is most practical to use this resonance. However, even if the receiver gain is set to its lowest value, ADC overflow will occur, and the null methods are recommended.

Pulses designed to selectively excite the solvent resonance are commonly encountered in many NMR experiments optimised for biological samples. The spin-lattice relaxation time of H2O is much longer than those of the macromolecule and the water proton resonance is therefore partially saturated at the end of the pulse sequence. Chemical exchange and spin diffusion between protons on the molecule of interest (in proteins the HN and Hα spins are affected most) and the partially saturated water can lead to a partial saturation of the protons of the macromolecule. This can be avoided by selectively returning the water magnetisation back to the z-axis; so called ‘water-flip-back’.

Such selective excitation can be achieved either by simply reducing the radio frequency (RF) field strength and/or by increasing the length of the pulse (referred to as selective pulses or soft pulses) or by the use of a shaped pulse. Here, both the amplitude and the phase are varied during the period of the pulse to achieve the desired excitation profile (commonly used shapes are gaussian and sinc shapes).¹ Whether selective or shaped pulses are used their excitation profile should be such that only the solvent peak is excited and not the resonances of interest. Modern spectrometer software will calculate the approximate power of such pulses based on the calibrated 90° pulse width supplied by the user. Fine tuning of this power is required by the user to achieve optimum solvent suppression. In addition to the power of the pulse, fine tuning of the transmitter offset may be required.

1.3.5 Acquisition Parameters

Once the initial four steps are complete, attention is now shifted to the acquisition parameters of the NMR experiments themselves. Many of these parameters are sample-dependent and therefore no hard and fast rules can be supplied. However, once the acquisition parameters are optimised they can be applied on subsequent occasions. Modern spectrometer software allows the user to write macros for common commands to facilitate spectrometer set-up and data processing; these can be easily adapted to loading experimental parameters, thus avoiding mistakes and hence wasting long hours of spectrometer time (and sample life). The parameters that are applicable to the vast majority of experiments (from 1D to 4D experiments) are considered below and from a practical point of view rather than a theoretical one. The author has avoided the use of manufacturer specific terminology to avoid confusion.

The transmitter offset is usually positioned such that the signals of interest are centred. If the spectrum contains an intense solvent peak it is usual to position the transmitter offset on this resonance. Apart from the solvent suppression method employed, this makes the use of convolution functions (see data processing, below) more convenient and reduces the possibility of spectral artefacts arising from such an intense signal. Inspection of the BMRB database (Biological Magnetic Resonance Data Bank, http://www.bmrbwiscedu/) shows that the vast majority of proton resonances found in biological macromolecules (protein, DNA and RNA) are between 14 and −1.0 ppm. However, it should never be assumed that this spectral range will contain all the resonances of interest. For example, paramagnetic systems have much larger spectral widths, sometimes in the order of 100 ppm (see Chapter 6). For such samples, acquisition of all the proton resonances in a single spectrum is impossible.

Initially, a relatively wide spectral width should be acquired to avoid aliasing or folding in signals that are shifted outside the ‘normal’ range as a result of bound ligands. In addition, modern digital filters are extremely efficient and resonances that are significantly outside the set spectral width will be filtered out and so never observed. Similarly, the heteronuclear chemical shift ranges are well known (e.g. 90–140 ppm for backbone amide nitrogens and 5–85 ppm for aliphatic side-chain carbons) but should be checked for unusually shifted resonances. Arginine side-chain guanidine nitrogen signals are commonly observed in -HSQC type experiments, usually folded into the spectrum from 70 ppm; less commonly lysine side-chain amino nitrogens are observed (folded from 20 ppm).

A common practice is to reduce the spectral width for indirect heteronuclear dimensions (particularly ) and deliberately fold resonances in, so allowing a higher resolution with the same total acquisition time. If the initial incremental delay for the aliased dimension is set to 1/(2SW) where SW is the spectral width (and quadrature is achieved with the States-TPPI method), then resonances that have been folded or aliased an odd number of times will have opposite phase to those that are not aliased (or have been aliased an even number of times). However, care should be taken when folding spectra lest signals be folded on top of others and so increase the resonance overlap (or signal cancellation) in already crowded spectra.

The long correlation times of macromolecules results in the signals relaxing relatively quickly, and as a consequence acquisition times are quite short (i.e. ≤100 ms). There is a temptation to increase the acquisition time of the FID in the misconception that this will enhance the resolution of the spectrum. If data is recorded after the signal has decayed, all that is then measured is noise and, after Fourier transformation, the signal-to-noise ratio of the resulting spectrum is degraded. Determination of the optimal acquisition time is best performed experimentally. If a 1D spectrum with good signal-to-noise is acquired it can be compared to subsequent spectra either acquired with a shorter acquisition time or with the FID truncated during processing. A judgement must be made as the point where the gain in signal-to-noise is outweighed by the loss in resolution.

Acquisition times in the indirect dimension(s) of 2-, 3- and 4- dimensional experiments are set via the number of acquisitions, or increments, and are in turn related to the spectral widths in these dimensions. The relationship of AQ = N/(2SW) (where AQ is the total acquisition time, N is the number of increments and SW the spectral width) is well known. Theoretically, optimal resolution requires an acquisition time up to 3/R2, and the maximum signal-to-noise is obtained at 1.26/R2, where R2 is the transverse relaxation rate [16, 19], but this is rarely possible to achieve, and so a balance must be struck. This balance is additionally complicated by the constraints of the phase cycle of the NMR experiment itself. This sets a minimum number of scans per increment and only integer multiples of a phase cycle should be used.

As a general guide for two-dimensional homonuclear experiments, acquisition times in the indirect dimension of 40–60 msec should give adequate resolution. For heteronuclear three-dimensional experiments, both total acquisition time constraints (experiments usually require 3–4 days) and relaxation limit the proton acquisition time to 20 msec. For dimensions, acquisition times ≥30 msec should be used and for acquisition times are usually limited to 9 msec so as not to resolve one bond carbon couplings. If carbonyl resonances are being recorded, this can be extended to 25 msec to give better resolution in this crowded region. However, when the experiments make use of constant-time periods [17–19] it is usual