Paramagnetic Relaxation Enhancement for Detecting and Characterizing Self-Associations of Intrinsically Disordered Proteins

Summary

A protocol for the application of paramagnetic relaxation enhancement NMR spectroscopy to detect weak and transient inter- and intra-molecular interactions in intrinsically disordered proteins is presented.

Abstract

Intrinsically disordered proteins and intrinsically disordered regions within proteins make up a large and functionally significant part of the human proteome. The highly flexible nature of these sequences allows them to form weak, long-range, and transient interactions with diverse biomolecular partners. Specific yet low-affinity interactions promote promiscuous binding and enable a single intrinsically disordered segment to interact with a multitude of target sites. Because of the transient nature of these interactions, they can be difficult to characterize by structural biology methods that rely on proteins to form a single, predominant conformation. Paramagnetic relaxation enhancement NMR is a useful tool for identifying and defining the structural underpinning of weak and transient interactions. A detailed protocol for using paramagnetic relaxation enhancement to characterize the lowly-populated encounter complexes that form between intrinsically disordered proteins and their protein, nucleic acid, or other biomolecular partners is described.

Introduction

Intrinsic disorder (ID) describes proteins (IDPs) or regions within proteins (IDRs) that do not spontaneously fold into stable secondary or tertiary structures but are biologically active. Generally, the function of IDP/IDRs is to facilitate specific yet reversible interactions with biomolecules at physiological conditions¹. Thus, IDPs and IDRs are involved in a range of cellular functions, including recruitment, organization, and stabilization of multi-protein complexes, for example, the assembly and activity of the spliceosome², recruitment and organization of components at sites of DNA damage³, organization, and stabilization of the recruitment of transcription complexes⁴, or of the chromatin remodeler BAF⁵. Additionally, IDPs are found at signaling nexuses where their promiscuity for different binding partners enables them to mediate information transfer through cellular protein networks⁶. Recent work has also revealed a proclivity for IDR regions to self-associate forming biomolecular condensates through the process of liquid-liquid phase separation⁷. Many of the aforementioned functions involving ID are also now thought to involve some aspect of condensate formation⁸. Despite the importance of ID for biomolecular complex assembly, stabilization, scaffolding, and signal transduction, the atomic details of their specific interactions are difficult to identify since IDPs and IDRs are typically not amenable to structural investigations using x-ray crystallography or cryogenic electron microscopy.

Nuclear magnetic resonance (NMR) is an ideal technique for investigating ID as it is not dependent on the presence of rigid or homogenous structural ensembles but reports on the immediate local environment of individual nuclei. The resonant frequency, or chemical shift, of a nucleus in a given molecule is influenced by weak magnetic fields induced by the local electronic distribution, which in turn is dependent on bond lengths, angles, nearness of other nuclei, interactions with binding partners, and other factors⁹. Thus, each nucleus acts as a unique, site-specific structural probe sensitive to changes in its local chemical environment. Despite these advantages, NMR is a bulk technique, and the observed chemical shift is the average of all the environments sampled by a particular nucleus. A range of NMR techniques, many of which are described in this issue, have been developed to recover structural, dynamic, and kinetic information about high energy, lowly-populated biomolecular conformations contained in the averaged chemical shift^10,11. Although transiently populated, identification and quantification of these states are important for determining the details of functional mechanisms¹². For example, in the case of IDPs and IDRs, the conformational ensemble may be biased to preferentially sample conformations that are productive for the formation of encounter complexes with physiological binding partners. Detection of these states, as well as identification of the residue-specific inter- and intra-molecular interactions and dynamics, are important to determine the underlying structural mechanisms of protein function and complex formation.

A protocol for using paramagnetic relaxation enhancement (PRE) NMR to investigate transient, lowly-populated states important for the formation of IDP/IDR-mediated biomolecular complexes is described¹³. This approach is useful for studying the transient protein-protein interactions such as those that promote the assembly of amyloid fibrils from α-synuclein^14,15 or the self-association of FUS¹⁶, as well as to characterize specific protein-protein interactions such as between signaling proteins¹⁷. An example of a self-associating IDP is presented, where specific inter- and intra-molecular interactions result in preferentially compacted states as well as site-specific interactions that drive self-association.

The PRE arises from the magnetic dipolar interaction of a nucleus to a paramagnetic center with an isotropic g-tensor, commonly supplied in the form of an unpaired electron on a nitroxide group or as a paramagnetic metal atom¹⁸ (Figure 1). While atoms with anisotropic g-tensors also produce a PRE effect, analysis of these systems is more difficult due to confounding effects contributed by the pseudo-contact shifts (PCS) or residual dipolar coupling (RDC)^13,19. The strength of the interaction between a nucleus and the paramagnetic center is dependent on the <r^-6> distance between the two. This interaction results in an increase in nuclear relaxation rates, which causes detectable line broadening even for long-range interactions (~10-35 Å), because the magnetic moment of the unpaired electron is so strong^20,21. Detection of transient states with the PRE is possible if the following two conditions are met; (1) the transient interaction is in fast exchange on the NMR timescale (observed chemical shift is a population-weighted average of the exchanging states); and (2) the nuclei to paramagnetic center distance is shorter in the transiently populated state than in the major state¹¹. The transverse PRE is denoted Γ₂ and, for practical purposes, is calculated from the difference in ¹H transverse relaxation rates between a sample containing a paramagnetic center and a diamagnetic control. For an in-depth treatment of the theory of the PRE and related pseudocontact shifts in fast and slow exchange regimes, the reader is referred to the comprehensive reviews by Clore and co-workers^13,22. Here, only the situation where ¹H_N-Γ₂ is in the fast exchange regime is considered, where because of the r^-6-dependence of the PRE, the observed relaxation rate is related to both the distance to which the paramagnetic center approaches the nucleus as well as the amount of time it spends in that conformation. Therefore, transient conformations that do not involve a close approach produce a small PRE while closer interactions, even if short-lived, will produce a larger PRE.

For IDPs, the PRE is used to measure and differentiate the interactions occurring within a single molecule (intramolecular) and between separate molecules (intermolecular). By attaching a paramagnetic center to an NMR visible (e.g., ¹⁵N-labeled) or NMR invisible (e.g., natural abundance ¹⁴N) protein, the source (inter- or intra-molecular) of the PRE may be determined (Figure 2). Site-directed mutagenesis that introduces a cysteine residue is a convenient approach to attach a paramagnetic center (spin-label) to a protein²³. Several types of molecules have been proposed for use as spin labels, including metal chelating (EDTA-based) and free-radical (nitroxide-based)²⁴. Various nitroxide spin labels have been described and are available with different cysteine-reactive chemistries such as methanethiosulphonate, maleimide, and iodoacetamide^25,26 (Figure 1). Inherent flexibility of the tag or of the linker may be problematic for certain analyses, and in these situations, different strategies have been proposed to limit the motion of the tag, such as by adding bulky chemical groups or the use of a second linker to anchor the tag to the protein (two site attachment)^27,28. Additionally, commercially available tags may contain diastereomeric proteins but generally this will not contribute to the observed PRE²⁹. The use of the 3-Maleimido-PROXYL attached to a free cysteine via maleimide chemistry is described since it is readily available, cost-effective, non-reversible, and the reducing agent tris(2-carboxyethyl)phosphine (TCEP) can be maintained in the solution throughout the labeling reaction. Since 3-Maleimido-PROXYL has an isotropic g-tensor, no PCS or RDCs are induced, and the same chemical shift assignments can be used for both the paramagnetic and diamagnetic samples¹³.

The ¹H_N-T₂ is measured using a two time-point strategy (T_a, T_b) that has previously been shown to be as accurate as collecting a full evolution series consisting of 8 to 12 time points³⁰. The first time point (T_a) is set as close to zero as practical, and the optimal length of the second time point is dependent on the magnitude of the largest expected PRE for a given sample and can be estimated from: T_b ~ 1.15/(R_2,dia + Γ₂) where R_2,dia represents the R₂ of the diamagnetic sample¹³. If the magnitude of the largest PREs is unknown, setting T_b to ~ one times the ¹H T₂ of the protein is a good initial estimate and further optimized by adjusting T₂ to improve the signal to noise. This two-point measurement strategy significantly reduces the experimental time required to measure PREs and allows time for more signal averaging, particularly since relatively dilute samples are used to minimize the effects of non-specific contacts between molecules. An HSQC-based pulse sequence is used to measure ¹H_N-T₂ and has been described in detail elsewhere³⁰. For improved sensitivity, the hard pulses of the forward and back INEPT transfers may be replaced with shaped pulses; alternatively, the sequence is readily converted to a TROSY-based readout³¹. Since IDPs typically have much longer transverse relaxation rates resulting in narrower line widths (due to the inherent disorder) than similarly sized globular proteins, long acquisition times in the indirect dimension may be used to improve spectral resolution and alleviate the chemical shift dispersion limitation inherent for IDPs.

PRE is a useful tool for studying protein-protein and protein-nucleic acid interactions, particularly interactions that are transient or lowly populated. A detailed protocol for the preparation of an NMR sample suitable for measuring PREs, including steps for protein purification, site-directed spin labeling, setting up and calibrating the pulse program, processing, and interpreting the NMR data, is provided. Important experimental considerations are noted throughout that may impact data quality and experimental outcome, including sample concentration, selection of the spin-label, and removal of paramagnetic components.

Protocol

General requirements for the protocol: protein purification facilities, UV-Vis spectrometer, high-field NMR spectrometer and operating software, post-processing analysis software including; NMRPipe³², Sparky³³, (or CCPN Analysis³⁴, or NMRViewJ³⁵).

1. Recombinant expression and purification of a protein for PRE measurements

1. Design an expression construct for the protein of interest so that there is a single cysteine residue present. Multiple mutations will be required to introduce a free cysteine at different positions in the protein of interest³⁶.
2. Express and purify a natural abundance (¹⁴N) or ¹⁵N-labeled sample of the protein of interest using an established protocol³⁷.
   NOTE: E. coli expression systems provide cost effective and robust method for recombinant protein expression since isotopic enrichment of ¹⁵N is a minimal requirement for biomolecular heteronuclear NMR spectroscopy. Typical steps are expression in minimal media, chromatographic purification, and removal of affinity purification tag. This protocol assumes a robust expression and purification protocol has been established that can produce sufficient protein of suitable quality for NMR investigations.
   1. Maintain 1 mM reducing agent (DTT or TCEP) in buffers at all purification steps to prevent reaction of the free cysteine and formation of intermolecular disulfide bonds for IDPs.
      NOTE: Some systems may be more tolerant and less aggregation prone to non-reducing conditions depending on the specific characteristics of the protein, as well as the temperature, pH, and buffer system chosen for purification³⁸.
   2. Remove affinity tags used for purification before proceeding since they may non-specifically interact with the protein in unpredictable ways or possibly contain reactive cysteine residues that could inadvertently serve as an unintended attachment site.
   3. Prepare a ¹⁵N labeled reference sample without cysteine mutation(s) mixed with a soluble version of the spin-label to assess the contribution of solvent PREs.

2. Conjugating the 3-Maleimido-PROXYL nitroxide spin label

1. Store or exchange the purified protein into a degassed buffer containing 50 mM Tris pH 7 and 1 mM TCEP; the buffer may also contain up to 8 M urea if needed to aid protein solubility.
   ​Alternatively, rapidly dilute a protein stock solution into at least 10 volume equivalents of degassed 50 mM Tris pH 7 and 1 mM TCEP buffer. Ensure that the protein concentration prior to adding spin-label is at least 100 µM.
2. Add 3-Maleimido proxyl from a stock solution to 20x molar excess of the protein of interest. Protect the sample from light and oxygen and incubate overnight at room temperature or 4 °C; gentle rocking or nutation may improve labeling efficiency.
3. Prepare stock solutions of the spin-label by dissolving 3-Maleimido proxyl powder in 95% ethanol. Aliquots of the stock can be stored at -80 °C for less than 6 months.
4. Critical step: Remove the non-reacted free spin-label to prevent non-specific solvent PREs. Achieve this by gel filtration or (preferably) extensive dialysis of the protein sample. This step will also introduce the protein into a buffer suitable for NMR.
   NOTE: Reducing agents should be prepared fresh, and compatibility between buffer components should be considered; for example, TCEP degrades quickly in phosphate-based buffers, and this combination should be avoided³⁹.
5. Treat all buffers used from this step forward with a chelating resin selective for divalent and transition metals to remove paramagnetic ions or spin-label quenchers. If the protein cannot be stored in an NMR buffer, concentrate the protein to be rapidly diluted into a buffer suitable for NMR.
6. Monitoring the efficiency of spin-label incorporation.
   1. Use Ellman's reagent (5,5-dithio-bis-(2-nitrobenzoic acid) for quantifying free sulfhydryl groups in solution⁴⁰.
      NOTE: Detailed protocols are available from the manufacturer. For the purposes here, it is important to determine incorporation of the spin-label, the concentration of free sulfhydryl groups is compared with the total protein concentration. The percent of free sulfhydryl groups is the percent of molecules that do not have a nitroxide spin label attached.
   2. Monitor the intensity of the peak corresponding to the tagged cysteine residue to judge spin-label incorporation into the protein of interest.
      ​NOTE: This is a rapid and effective approach to determine the degree of spin labeling of the protein. Complete incorporation of the spin-label will result in the disappearance of the peak from the spectrum. With the poor dispersion characteristic of IDPs the peak corresponding to the mutant cysteine residue may not always be readily identified, and thus the use of Ellman's reagent (step 2.6.1) is recommended.

3. Prepare NMR sample for measuring intra- or inter-molecular PRE

1. Prepare sample for measurement of intramolecular PRE
   1. Prepare ¹⁵N isotopically enriched, spin-labeled protein to a concentration of at least 100 µM but not more than 300 µM in a buffer suitable for NMR. Total sample volume (including D₂O) is 500 - 550 µL.
      NOTE: Common NMR buffers include phosphate, acetate, (bi)carbonate, and TRIS. Good's buffers such as MES, HEPES may also be appropriate. Exercise caution when selecting buffers to ensure no cross-reactivity with other solution components.
   2. Ensure that the pH is ~7.2 or lower to minimize the effects of amide proton exchange with water. Keep the concentration of salt as low as possible (typically less than 150 mM), although the primary consideration is to maintain protein stability.
      NOTE: Approaches for conducting NMR experiments in high-salt conditions have been described elsewhere⁴¹.
2. Prepare sample for measurement of intermolecular PRE
   1. Follow this step or step 3.1; they are not performed simultaneously. Prepare ¹⁴N natural abundance, spin-labeled protein in the chosen NMR buffer.
   2. Prepare the protein sample by mixing ¹⁵N isotopically enriched non-spin-labeled protein with 1%-50% ¹⁴N natural abundance spin-labeled protein so that the final concentration is identical to the sample prepared in 3.1.1. The total sample volume (including D₂O) is 500 - 550 µL.
   3. Empirically optimize the ratio of the ¹⁵N and ¹⁴N proteins for each protein studied. The ratios of 1%, 5%, and 20% of ¹⁴N-spin-labeled protein are good starting points.
      NOTE: A buildup of the PRE as a function of added ¹⁴N-spin-labeled protein indicates a specific effect; the observed PRE is sample-specific since it depends on distance and population (as discussed above), and therefore higher ratios of ¹⁴N-spin-labeled protein will be required if the interaction is particularly transient¹⁷.
3. Transfer the NMR (either intra- or inter-molecular) sample to a 5 mm NMR tube that is appropriate for use in high-field magnets using a long-stem (9") glass pipette or micropipette. Ensure that all NMR samples include 5%-10% of D₂O to facilitate field locking.
   ​NOTE: NMR tubes that utilize polymer plugs to reduce the necessary sample volume are not recommended for PRE measurements due to difficulties related to effective sample shimming.

4. Set up NMR spectrometer and experiment specific parameters

1. Exercise extreme caution when working around superconducting, high-field NMR spectrometers.
   NOTE: Hazards include injuries due to the sudden acceleration of metallic objects toward the magnet, interference with implanted medical devices, and asphyxiation due to a sudden release of N₂ and He₂ gas in the event of a magnet quench. The following steps assume that the reader has undergone the required training, is aware of these and other local hazards, and has received approval from the facility manager to operate the NMR spectrometer. When in doubt of a step or instruction, consult with the facility manager or experienced user to prevent potential personal injury or damage to the spectrometer.
2. The following steps assume a commercial NMR spectrometer running a modern version of the acquisition control software. Download the pulse program and parameter files and place them in the appropriate directories.
   NOTE: A pulse program and parameter set suitable for use with a Bruker spectrometer and TopSpin (3.2 or later) are available upon request from the authors.
   1. Critical step: Familiarity with installing non-native NMR pulse programs is assumed; consult with the facility manager or an experienced user if necessary.
3. Place the sample in the magnet, lock on the ²H signal using the Lock command, tune and match the ¹H channel according to facility protocols (the exact procedure will depend on if the probe is equipped with a remote tune and match module).
4. Adjust the shims using the topshim subroutine to optimize solvent signal suppression.
5. Calibrate the ¹H and ¹⁵N 90° pulses using standard methods.
   1. Calibrate the ¹H pulse using the popt program (use pulsecal first to estimate pulse length).
   2. Calibrate the ¹⁵N pulse against a standard sample; make sure this value has been calibrated recently by discussing with a technical director or experienced user.
   3. Alternatively, calibrate the ¹⁵N pulse on the sample by varying one of the 90° pulses of an HMQC experiment until a null signal is achieved.
   4. Determine the correct attenuation for shaped pulses using the shape tool (stdisp) subroutine.
   5. Open the appropriate pulse shape file by clicking on the folder icon. The shaped pulses are found in the pulse parameters section of ACQUPARS.
   6. Load the pulse definition file and click on Analyze Waveform > Integrate Shape. Input the calibrated ¹H 90° hard pulse, desired shaped pulse length, and rotation (90° or 180°).
   7. Calculate the power level of the shaped pulse by adding the change of power level to the attenuation for the calibrated 90° pulse.
6. Record a standard ¹H, ¹⁵N HSQC (hsqcetfpf3gpsi) to optimize sweep width, carrier frequency and check water suppression²⁵.
7. Adjust the sweep width and the number of indirect dimension increments using the sw and td commands or directly in the appropriate dialog boxes. Typically, for collecting PREs, spectral widths are chosen so that the spectrum is not folded.

5. Setup the ¹H_N-T₂ experiment

1. Calibrate the shaped pulses as described above (4.4.5-4.5.7). The shaped pulse parameter files for the PRE experiment are Eburp2.1000 (90° pulse), Reburp.1000, and Iburp2.1000. Enter the calibrated pulse lengths in the pulse parameters section on the ACQUPARS tab.
2. This experiment measures the ¹H_N-T₂ using the two time-delay point approach³⁰.
   1. Set the time delays by editing the vdlist file, the first delay (T_a) is set to 0.01 ms.
   2. Choose the second delay, (T_b) using the relationship to the expected maximum PRE (T_b ~ 1.15/(R_2,dia + Γ₂) where R_2,dia represents the R₂ of the diamagnetic sample¹³. Without prior knowledge of the magnitude of the PRE contribution to the observed relaxation, a good starting point is setting T_b to ~1x ¹H T₂.
   3. Then determine a suitable value by comparing the first increments (processed with the efp command) of the T_a and T_b spectra and adjusting T_b such that the signal decays to between 40%-50% of its initial value.
      NOTE: This approach optimizes the spectral signal-to-noise, a necessary consideration for samples that cannot be highly concentrated (< 50 µM). Suitable values of T_b are sample dependent but typically range from 8 - 40 ms for an average sized protein.
3. Determine the number of complex points to record and number of scans for sufficient signal averaging. Since IDPs have longer ¹⁵N T₂ than folded proteins of comparable size, longer acquisition times in the indirect dimension may be used.
   NOTE: This value is dependent on the specific characteristics of the protein but can be roughly estimated from the ¹⁵N T₂ and optimized by monitoring signal decay in the FID. For the direct dimension, 1024* complex points (13 ppm sweep width, 112.6 ms acquisition time) is sufficient for most samples.
4. Use the command expt to calculate the experiment time and then start the experiment with the command zg.

6. Make a diamagnetic sample by reducing spin-label with ascorbic acid

1. Dissolve sodium ascorbate in the NMR buffer and adjust the pH to match the original NMR buffer.
2. Calculate the concentration of sodium ascorbate stock so that a 10x molar excess of ascorbate over the concentration of the spin-label can be added with the least change of sample volume. For example, for a 100 µM protein sample, a 100 mM stock of ascorbate is appropriate. Reducing the spin-label will require adding 5.5 µL of ascorbic acid stock solution, which is only 1 % of the total sample volume.
3. Add the required amount of ascorbic acid to the NMR tube by placing a droplet below the rim of the tube, cap the tube, carefully invert the tube to mix, and then spin at 200-400 x g for 10-20 s in a hand-cranked centrifuge to settle the sample at the bottom of the tube.
4. Wrap the NMR tube in foil to protect from light and allow the reaction to proceed for at least 3 h.
5. Record ¹H_N-T₂ on the diamagnetic sample using the same parameters used for the paramagnetic sample.
6. Recalibrate the pulses. However, they should not have changed from the paramagnetic measurements; if they are significantly different (> 0.5 µs difference), consider sample quality (e.g., degradation, precipitation).
7. Ensure that all acquisition parameters, including the specified relaxation delays (vdlist), number of dummy scans, number of scans collected, number of complex points collected, acquisition time, sweep widths, and carrier frequencies remain the same for the diamagnetic and paramagnetic samples.

7. Process paramagnetic and diamagnetic spectra

1. Copy the data to local computer or workstation that has NMRPipe and Sparky installed and configured. Make a folder named proc in the experiment data directory that contains the ser file.
2. Copy the NMRPipe scripts fid.com, p3d.com, and nmrproc.com into proc (processing scripts are available upon request from the authors).
   1. Use the fid.com script to convert the Bruker data format (ser) into NMRPipe format.
   2. Use the p3D.com script to split the pseudo3D planes into individual spectra.
   3. Use the nmrproc.com script to read the output of the fid.com script, apply solvent suppression, a window function, append zeros to the raw data (zero fill), apply phase correction, execute a Fourier transformation, trim the data for display and write the processed data to disk. The script will output one file for each relaxation delay recorded (T_a and T_b).
      ​NOTE: Each of these scripts is customizable to optimize the processing for the specific details of each experiment. Tutorials and example data sets are included in the NMRPipe distribution available from the NMRPipe website³². NMRDraw may be used for spectral viewing during processing (e.g., setting proper phase angles etc.). Options available for NMRPipe commands can be viewed using the command nmrPipe -help.

8. Transfer resonance assignments and extract peak heights

1. Change the file header information for each spectrum file (T_a, T_b for both paramagnetic and diamagnetic samples) using the command sethdr [filename] -ndim 2.
2. Use Sparky to extract peak heights³³ following steps 8.3-8.5. Other software packages, including NMRPipe (NMRDraw)³², CCPN Analysis³⁴, and NMRViewJ³⁵ are also appropriate.
3. Read the spectral files into Sparky. At this step the data set will consist of one spectrum for each time point spectra (T_a, T_b), for both the paramagnetic and diamagnetic samples, measured for each position of the spin-label in the protein.
4. Use Sparky to pick peaks (command: F8, then click and drag) and transfer assignments using the transfer peak list tool from a reference peaklist.
   NOTE: Resonance assignments of the protein of interest are necessary for sequence-specific interpretation of the observed PREs³⁶.
   1. Set contours in both paramagnetic and diamagnetic spectra to the same level. Ensure to set the contours so that the spectra collected after the time delay do not purposely exclude peaks but are high enough so that the T_a spectra are not overly noisy.
5. Save the new peak lists for each spectrum and include the measured peak intensity and Sparky calculated signal to noise ratio (command: lt to open peaklist, click options to include intensity and SNR columns, command: save).

9. Extract ¹H_N-T₂ rates for each residue and calculate PRE

1. Import the peak lists into spreadsheet software or a preferred programming language such as Python.
   NOTE: For each spin-label position on the protein, the dataset will consist of four peak lists with associated peak intensities, one each of T_a and T_b for both the paramagnetic and diamagnetic experiments.
2. Calculate ¹H_N R₂ for both the paramagnetic and diamagnetic samples using the equation:
   
   
3. Use the above equation to determine the relaxation rate for each residue for the paramagnetic and diamagnetic samples.
4. Determine the ¹H_N-Γ₂ rate for each residue using the equation:
   
5. Use the Sparky calculated signal to noise ratio (SNR) to compute the uncertainty of the peak height for each residue.
6. Propagate the error using the equation:
   
7. Plot ¹H_N-Γ₂ as a function of residue number using a scatter plot including the error calculated in 9.6.

Representative Results

Intramolecular ¹H_N-Γ₂ PREs were recorded on a self-associating, intrinsically disordered fragment (residues 171-264) derived from the low-complexity domain of the RNA-binding protein EWSR1⁴² (Figure 3). Residues in close sequential proximity to the spin-label attachment point (e.g., residue 178 or 260 in Figure 3) are expected to be significantly broadened and are not detectable in the spectrum. Residues sequentially spaced from the attachment point yet show enhanced Γ₂ were spatially close (10-35 Å) to the spin-label. In the case of EWSR1 171-264, attributing the source of the PRE effect is complicated since it may arise from a combination of inter- and intra- residue contacts and is dependent on the distance from the nucleus to the paramagnetic center, the population of that conformation, and the dynamics of the vector connecting the electron and nuclear spins. Further, the magnitude of PREs arising from intramolecular contacts is not concentration-dependent, while PREs arising from intermolecular contacts depend on concentration as well as the kinetics and dynamics of the association between protein molecules.

A possible interpretation of these data is that the IDP ensemble samples conformations that are more compact than an extended chain. Alternatively, the PREs could arise from intermolecular contacts responsible for the self-association of EWSR1, or the PREs could be from a combination of both intra- and intermolecular contacts. In the case presented here, what remains unknown is how close the residues approach the spin-label or for how long they remain in close proximity. With highly flexible molecules such as EWSR1 171-264, it can be difficult to qualitatively disentangle these parameters. By placing the spin-label at different residue positions, contacts between different parts of the chain may be identified, providing a more accurate interpretation of specific interactions that may be functionally relevant for self-association (Figure 3). Measuring intermolecular PREs (¹⁴N spin-labeled protein mixed with ¹⁵N non-spin labeled protein), employing a mutational strategy of residues with larger than average PREs (e.g., residues 196 or 215, Figure 3), and utilizing other biophysical methods such as dynamic light scattering, size exclusion chromatography, and analytical ultracentrifugation, are useful for characterizing the conformational ensemble of an IDP.

Figure 1: Molecules containing an unpaired electron and various functional groups to facilitate attachment to free cysteine residues that are typically used as paramagnetic relaxation agents. Diamagnetic molecules may be used as controls. (A) 3-Maleimido-2,2,5,5-tetramethyl-1-pyrrolidinyloxy, free radical (3-Maleimido-PROXYL) (B) 3-Carboxy-2,2,5,5-tetramethyl-1-pyrrolidinyloxy, free radical (3-Carboxy-PROXYL) (C) 3-(2-Iodoacetamido)-2,2,5,5-tetramethyl-1-pyrrolidinyloxy, free radical (3-(2 - Iodoacetamido-PROXYL) (D) 1-Oxyl-2,2,5,5-tetramethylpyrrolidin-3-yl) Methyl Methanethiosulfonate (MTSL) (E) (1-Acetoxy-2,2,5,5-tetramethyl-δ-3-pyrroline-3-methyl) Methanethiosulfonate (Acetoxy-MTSL) Please click here to view a larger version of this figure.

Figure 2: Depiction of intra- and intermolecular PRE. (A) Intramolecular PRE, the red circle represents the effective radius of a paramagnetic center attached to a ¹⁵N-labeled protein. The PRE effect decreases with an <r^-6> dependence on distance from the paramagnetic molecule. (B) Intermolecular PRE, the paramagnetic group (red circle), is located on a ¹⁴N (natural abundance) protein (blue) that is invisible to NMR. The effects of the paramagnetic group on the non-NMR active protein are observed as increased relaxation rates when it comes into close contact with the ¹⁵N protein (black). Please click here to view a larger version of this figure.

Figure 3: ¹H_N-Γ₂ rates for residues 171-264 of the intrinsically disordered domain of EWSR1. A serine residue at position (A) 178 or (B) 260 that has been mutated to a cysteine serves as the attachment point for a 3-Maleimido-PROXYL spin-label (red *). Increased relaxation rates occur at the location of the tag, other sites of increased relaxation are indicative of intramolecular interactions. Please click here to view a larger version of this figure.

Discussion

A method for characterizing transient interactions that exist at low populations between intrinsically disordered proteins and various binding partners using PRE has been presented. In the example shown, the protein is self-associating, and thus the PRE may arise from a combination of inter and intramolecular interactions. This method is readily extended to heterogeneous samples where the interactions between two different proteins may be characterized. Complementary information about how different regions of the protein interact is available by placing the spin-label at different positions within the protein. Additionally, by alternating the spin-label between NMR active (¹⁵N) and NMR inactive (¹⁴N) species, the intra- and inter-molecular sources of observed PRE may be differentiated from one another, providing information about encounter complexes. The experiment outlined here can report on encounter complex interactions even if they occur on a microsecond timescale¹³.

Central to this method is the incorporation of a spin-label tag into the protein of interest by attachment to a cysteine residue. Some proteins may contain a native cysteine that is suitable (does not participate in disulfide bonds, is surface exposed) for attaching a spin-label. For IDPs, solvent exposure of cysteine is usually not an issue. In the majority of cases, it is desirable to introduce cysteines as conservative mutations (serine to cysteine or other uncharged polar amino acids to cysteine) using site-directed mutagenesis⁴³. In the example presented, the fragment of EWSR1 does not contain native cysteines and is enriched in serines; thus, devising a mutational strategy was straightforward. Proteins that contain native cysteine(s) present a more complicated case, and care needs to be taken to not disrupt the native function (e.g., break a structurally important disulfide bond)⁴⁴. Further, to incorporate a single cysteine for spin-labeling, the native cysteines must be mutated to a residue that does not react with the spin-label (no mercapto group) and based on its size and other properties, serine is a good substitute for cysteine. If native cysteines need to be mutated, careful characterization of the mutants is required to ensure they maintain native structure and function is essential. Simple ¹H,¹⁵N HSQCs of mutants in comparison to the wildtype protein are powerful indicators of perturbations (even minor) to protein structure, and this approach is also useful for IDPs⁴⁵. Other methods to consider are circular dichroism, analytical ultracentrifugation, or biochemical approaches such as activity assays⁴⁶.

Technical considerations for obtaining reproducible, rigorous, high-quality data include the removal of ionic impurities during the preparation of the NMR sample. This is achieved by passing all solutions over chelating resin prior to use. Further, using a properly degassed buffer is important during the attachment of the nitroxide spin label as the presence of oxygen can reduce the efficiency of labeling. Diamagnetic contamination will contribute to a decrease in the observed PRE; however, the effect is less pronounced on intramolecular PREs and can be reduced by decreasing ΔT¹³. Therefore, it is not necessary to obtain 100% label incorporation to proceed with the experiment, particularly for the qualitative interpretation presented here. If free cysteines from incomplete spin-label attachment are problematic, some mercapto-reactive chemistries (e.g., maleimide) are amenable to maintaining a reducing agent in the sample throughout the experiment²⁶. It is important that the paramagnetic and diamagnetic samples match each other as closely as possible. When reducing the spin-label with ascorbic acid to create a diamagnetic control, consider the dilution factor introduced from titrating in an ascorbic acid stock solution. This dilution can be minimized by maintaining the ascorbic acid stock at least 10x the expected working concentration in the NMR buffer.

There are many software packages available for analyzing NMR data, including NMRPipe³², Sparky³³, CCPN Analysis³⁴, NMRViewJ³⁵, among others. The combination of NMRPipe for spectral processing and Sparky for spectral analysis (peak picking and quantification) was described here due to the ease of use of this combination. NMRPipe is commonly used by many NMR groups for spectral processing, but the NMRPipe suite contains the tools necessary for completing all steps of the analysis, albeit with a significant learning curve. Data may also be processed using the NMR spectrometer control software. Sparky was chosen for spectral analysis because of its ease of use and rapid uptake by novice users. There are several options available for spectral analysis (peak picking and measuring peak heights) that can easily substitute for the functionality of Sparky including CCPN Analysis, or NMRViewJ. Notably, many of these programs have overlapping functionalities and the user is advised to select the combination of programs with which they are most comfortable.

Poor chemical shift dispersion is an inherent problem with IDPs leading to significant resonance overlap and introduction of error into the measurement of peak height. Different strategies have been proposed to alleviate this problem. One of the most straightforward, and the one employed here, is to take advantage of the long transverse relaxation characteristic of IDPs and simply extend the acquisition time in the ¹⁵N (indirect) dimension. Alternatively, the triple resonance HNCO experiment is useful for resolving resonance overlap in IDPs due to the superior dispersion of C' resonances. Both TROSY and HSQC versions of the HNCO for measuring PREs have been proposed and are described elsewhere⁴⁷. However, the improved resolution is not always significant enough to warrant the increased complexity of the experiment, longer time for data collection, and added cost for preparing a suitable sample (enrichment of ¹³C). This is indeed the case for EWSR1 171-264 presented here, where no significant improvement in the number of non-overlapped residues was observed between a TROSY-HNCO and an ¹H, ¹⁵N HSQC collected with long acquisition time in the indirect dimension.

This procedure outlined above focuses on the utility of PRE experiments for characterizing weak interactions that exist within and between intrinsically disordered proteins. The PRE has a much broader utility in biomolecular NMR, including determining long-range structural restraints and quantitative determination of sparsely populated conformational states. For example, Clore and co-workers have pioneered the use of the PRE to detect and quantify transient interactions arising from interactions between individual domains of a single protein⁴⁸ or between the subunits of assembled protein complexes¹⁷. There are many examples of the PRE used to derive long-range distance restraints, including for large proteins⁴⁹, or with novel PRE tags⁵⁰, to help determine the overall fold of a protein⁵¹, as well as in highly paramagnetic systems⁵². Finally, while PCS is beyond the scope of this discussion, they have been applied to important biomolecular problems which have been described elsewhere⁵³. The method presented above is suitable for probing the conformation and interactions of IDPs using PREs and was designed to be accessible for novice users. For more quantitative approaches to the analysis of the PRE, the user is referred to the many excellent articles referenced within^11,24,30,31.

Materials

Name	Company	Catalog Number	Comments
0.45 µm and 0.22 µm syringe filters	Millipore Sigma	SLHVM33RS SLGVR33RS	Filter lysate before first purification step and before size exclusion chromatography.
100 mm Petri Dish	Fisher	FB0875713	Agar plates for bacterial transformation.
14N Ammonium chloride	Sigma Aldrich	576794	Use of 15N in M9 medium will produce an NMR visible protein, 14N will produces an NMR invisible protein
15N Ammonium chloride	Sigma Aldrich	299251	Use of 15N in M9 medium will produce an NMR visible protein, 14N will produces an NMR invisible protein
3 L Fernbach baffled flask	Corning	431523	Bacterial expression culture
3-Maleimido-Proxyl	Sigma Aldrich	253375	Nitroxide spin label
50 mL conical centrifuge tubes	Thermo Fisher	14-432-22	Solution/protein storage
Amicon centrifugal filter	Millipore Sigma	UFC900308	Protein concentration
Ampicillan	Sigma Aldrich	A5354	Antibiotic for a selective marker, exact choice depends on the expression construct plasmid
Analytical balance	Oahus	30061978	Explorer Pro, for weighing reagents
Ascorbic acid	Sigma Aldrich	AX1775	Reduces nitroxide spin label
Autoclave			Sterilize glassware and culture media
Calcium chloride	Sigma Aldrich	C4901	M9 media component
Centrifuge bottles	Thermo Fisher	010-1459	Harvest E. coli cells after recombinant protein expression
Centrifuge, hand-crank	Thomas Scientific	0241C68	Boekel hand-driven, low-speed centrifuge with 15 mL buckets that can accommodate NMR tubes
Chelex 100	Sigma Aldrich	C7901	Remove contaminating paramagnetic compounds from buffer solutions
Computer workstation			Linux or Mac OS compatable with NMR data processing and analysis software packages such as NMRPipe and Sparky
Deuterium oxide	Sigma Aldrich	151882	Needed for NMR lock signal
Dextrose	Sigma Aldrich	D9434	M9 media component
Dibasic Sodium Phosphate	Sigma Aldrich	S5136	M9 media component
Ellman's reagent (5,5-dithio-bis-(2-nitrobenzoic acid)	Thermo Fisher	22582	Quantification of free cystiene residues
High speed centrifuge tubes	Thermo Fisher	3114-0050	Used to clear bacterial lysate.
High-field NMR instrument (600 - 800 MHz)	Bruker		Equiped with a multichannel cryogenic probe and temperature control
IMAC column, HisTrap FF	Cytvia	17528601	Initial fractionation of crude bacterial lysate
Isopropyl B-D-thiogalactoside (IPTG)	Sigma Aldrich	I6758	Induces protein expression for genes under control of lac operator
LB agar	Thermo Fisher	22700025	Items are used for transforming E. coli to express protein of interest, substitions for any of these items with like products is acceptable.
LB broth	Thermo Fisher	12780052
Low-pressure chromatography system	Bio-Rad	7318300	BioRad BioLogic is used for low-pressure chomatograph such as running IMAC columns
Magnesium sulfate	Sigma Aldrich	M7506	M9 media component
Medium pressure chromatography system	Bio-Rad	7880007	BioRad NGC equipped with a multi-wavelength detector, pH and conductivity monitors, and automatic fraction collector
MEM vitamin solution	Sigma Aldrich	M6895	M9 media component
Microfluidizer	Avestin	EmulsiFlex-C3	Provides rapid and efficient bacterial cell lysis
Micropipettes	Thermo Fisher		Calibrated set of micropippetters with properly fitting disposable tips (available from multiple manufacturers e.g. Eppendorf)
Monobasic potassium phosphate	Sigma Aldrich	1551139	M9 media component
NMR pipettes	Sigma Aldrich	255688	To remove sample from NMR tube
NMR sample tube	NewEra	NE-SL5	Suitable for high-field NMR spectrometers
Preparative Centrifuge	Beckman Coulter	Avanti J-HC	Harvest E. coli cells after recombinant protein expression
Round bottom polystyrene centrifuge tubes	Corning	352057	Clear bacterial lysate
Shaking incubator	Eppendorf	S44I200005	Temperature controlled growth of E. coli starter and expression cultures
Sodium chloride	Sigma Aldrich	S5886	M9 media component
Sonicating water bath and vacuum source	Thomas Scientific		Used to degas buffer solutions
Sonicator	Thermo Fisher	FB505110	Used for bacterial cell lysis or shearing bacterial DNA
Spectrophotometer	Implen	OD600 Diluphotometer	Monitor growth of E.coli protein expression cultures
Superdex 200 16/600 size exculsion colum	Cytvia	28989333	Final protein purification step
Topspin software, version 3.2 or later	Bruker		Operating software for the NMR instrument
Transformation competent E. coli cells	Thermo Fisher	C600003	One Shot BL21 Star (DE3) chemically competent E. coli, other strains may be compatable
Tris(2-carboxyethyl)phosphine (TCEP)	ThermoFisher	20490	Reducing agent compatable with some sulfhydryl-reactive conjugations
UV-Vis spectrophotometer	Implen	NP80	Measure protein concentration.
Water bath, temperature controlled	ThermoFisher	FSGPD25	For heat shock step of bacterial transformation
Yeast extract	Sigma Aldrich	Y1625	For supplementing M9 media if required