Missing Pieces in Structure Puzzles: How Hyperpolarized NMR Spectroscopy Can Complement Structural Biology and Biochemistry

Structure determination lies at the heart of many biochemical research programs. However, the “giants”: X‐ray diffraction, electron microscopy, molecular dynamics simulations, and nuclear magnetic resonance, among others, leave quite a few dark spots on the structural pictures drawn of proteins, nucleic acids, membranes, and other biomacromolecules. For example, structural models under physiological conditions or of short‐lived intermediates often remain out of reach of the established experimental methods. This account frames the possibility of including hyperpolarized, that is, dramatically signal‐enhanced NMR in existing workflows to fill these spots with detailed depictions. We highlight how integrating methods based on dissolution dynamic nuclear polarization can provide valuable complementary information about formerly inaccessible conformational spaces for many systems. A particular focus will be on hyperpolarized buffers to facilitate the NMR structure determination of challenging systems.

However, often, NMR cannot adequately describe many important features of biomolecular systems. For example, structures of ligand encounter complexes, [19] folding intermediates, [20,21] or the number and lifetime of any transient states [22,23] or low-quantity targets (e. g., at in-vivo concentrations) cannot be accessed, despite their often vital importance. The reason behind this shortcoming is a lack of NMR sensitivity. To record the characteristics of, for example, an elusive transition state, the NMR signal needs to be acquired within the state's lifetime. However, NMR acquisition times are often on the order of minutes to hours, exceeding the lifetime of most intermediates by far.
This problem is not related to instrumentation issues or pulse sequences but to the intrinsically low energies involved in nuclear magnetism. [24] Considering a simple proton 1 H spin in the most potent commercially available NMR magnet (with a Larmor frequency of 1.2 GHz) [25] at room temperature (300 K), the Zeeman splitting of such an atom is only 7.95 × 10 À 25 J, and the total spin polarization only 0.0096 %. In other words, only one atom out of~10 000 contributes to the NMR signal. The situation is even worse for heteronuclei such as 13 C nuclei (under the same condition) with a polarization of 0.0024 %, where only one atom out of 40 000 produces a signal.
To tackle this issue, so-called hyperpolarization methods have been developed to increase the fraction of used spins in an NMR experiment and improve its sensitivity. [26][27][28][29][30] Various such methods have been devised and tested in recent years, each different in multiple aspects despite their common objective to boost nuclear polarization. This account focuses on hyperpolarization by dissolution dynamic nuclear polarization (DDNP) [31,32] as an apt methodological complement to existing structural biology research programs. We argue that the established "giants", such as X-ray crystallography, electron microscopy, molecular dynamics simulations, and even conventional solid-and liquid-state NMR, are well-suited to resolve the structural dynamics of biomolecular systems under typical invitro conditions but leave quite a few dark spots on the otherwise well-drawn structural depictions. These dark spots can yet be enlightened using DDNP thanks to recent developments in applications to proteins, membranes, and nucleic acids.
In the following, we will first give an overview of how this method works and how it can be embedded in existing workflows. Selected examples of applications follow. Hints towards other hyperpolarization methods will also be given at the end of this account.

How To Use Hyperpolarized NMR To Characterize Aqueous Biological Samples
Dynamic nuclear polarization (DNP) [24] is arguably the most versatile hyperpolarization method. It does not require any particular functional moieties on the target molecule, nor is it based on any particular reaction. Hence, applications to a wide array of systems are conceivable, and integration into existing research programs is possible without the need to modify the target. For DNP, unpaired electrons, for example, in the form of free radicals, are introduced into a sample. [33][34][35][36] Through slightly off-resonance microwave (μw) irradiation, the electron polarization is then transferred to nuclei in the vicinity of the electron spins. Electrons possess a much higher magnetic moment than nuclei (ca. 668 times larger than protons), such that the transfer of their "energies" to nuclei entails a boost in the latter's NMR intensities. The quantum mechanical details of the dynamic polarization transfer upon μw irradiation have been treated in various articles, although some aspects are still discussed. [31,[36][37][38][39][40][41][42][43][44][45][46][47][48][49][50][51][52][53] Since DNP primarily depends on the spin physics of the system, in principle, any NMR-active nucleus can be hyperpolarized; no particular chemical properties are necessary.
Among the DNP methods, currently, DDNP enables hyperpolarized high-field liquid-state NMR detection for a particularly wide array of molecules. (Typical requirements are simply high solubility and longitudinal relaxation times � 1 s or solvent interaction sites, when buffers are used as hyperpolarization vector, vide infra). Hence, this method ideally complements biomolecular NMR studies. The high-field aspect is particularly important in the context of biochemical research, as most NMR studies of proteins or nucleic acids are performed in aqueous solutions using high-field (typically B 0 > 14 T) magnets.

Dissolution DNP
DDNP is a two-stage method for which a sample is hyperpolarized ex-situ in a dedicated DNP apparatus operating at cryogenic temperatures close to 1 K and a magnetic field [68] between about 3 and 11 T -a condition at which the electronnucleus polarization transfer is most efficient. To record highfield solution-state NMR spectra subsequent to DNP at low temperatures (often denoted as the "signal build-up"), the sample is rapidly dissolved with a burst of hot solvent and transferred to a conventional NMR spectrometer for detection. Various implementations of how to dissolve and transfer the sample have been reported for a variety of applications. [69][70][71][72][73] Prototypically though, D 2 O pressurized to > 1 MPa is used for dissolution. In a subsequent step, the sample is pressed through a narrow capillary towards the NMR spectrometer, typically with a chase gas at 0.5-1 MPa. This dissolution and transfer process takes only a few seconds. Significant NMR signal enhancements of over four orders of magnitude can be achieved, translating into a squared reduction in necessary signal acquisition times. Hence, intense NMR signals are regularly recorded by DDNP in less than a second.
Historically, DDNP was developed to boost the NMR signals of small metabolites, such as pyruvate, urea, or glucose, and to monitor their signals by in-vivo imaging. The hyperpolarized molecules light up against the background of the conventional thermal equilibrium image, rendering their localization in the body and/or monitoring of their metabolic conversion possible. Famous examples include the detection and characterization of tumour tissue. [74][75][76]

Integrative DDNP: Towards High-Resolution Structural Biology by Dissolution DNP
DDNP can be used to study a wide variety of molecules under ambient aqueous conditions in any high-field NMR spectrometer. However, when aiming at broad applicability to biomacromolecules, two main problems must be solved. 1) The longitudinal relaxation times of proteins and nucleic acids are typically relatively short, so most of the hyperpolarization is lost during the sample transfer when attempting to shuttle the target molecule itself. 2) Proteins and nucleic acids often denature during dissolution and transfer due to the strong pressure and temperature gradients.
Hence, to enable DDNP of biomolecules, it was recently proposed to hyperpolarize the water [30,[77][78][79][80][81][82] of the buffer, in which the target molecule is dissolved and not the target itself. Due to spontaneous solvent-target proton exchange and nuclear Overhauser effects (NOE), [83,84] mixing of the target and the hyperpolarized aqueous buffer in the NMR spectrometer is sufficient to spontaneously transfer the enhancement from the latter to the former. This hyperpolarization exchange thereby depends on the solvent interaction of the investigated residues of the targets. For stably folded proteins, often many residues are buried in a hydrophobic core, and only a few exposed amino acids exchange sufficiently fast with the hyperpolarized solvent to experience a strong signal boost. In contrast, for intrinsically disordered proteins (IDPs) or RNAs with little secondary structure contents, all residues can be solventexposed, leading to a more homogenous signal enhancement along the primary sequence. This feature of varying selectivity in signal enhancements gives rise to some interesting possibilities (see Section 3.2).
To significantly enhance a residue using exchange with hyperpolarized water, the latter's rate should be faster than the delay between successive detections. For example, if a hyper-Mattia Negroni did his PhD at the University Milano-Bicocca where he studied the properties of molecular rotors in porous materials by solid-state NMR spectroscopy. In 2021 he joined the Kurzbach group working on dissolution dynamic nuclear polarization methods and applications. polarized spectrum is recorded every second (typical recovery delays are on the order of 0.1-1 s), the exchange rate needs to be significantly faster than 1 s À 1 . If this is not the case, the enhancement will be weak. Conversely, if the exchange is too fast such that transverse relaxation times become short, signals might be lost due to line broadening. [30] Luckily, for most exposed amide and imino residues of proteins and nucleic acids, the exchange rates fall into the right regime.
The "hyperpolarized water" approach solves both abovementioned problems at the same time. The relaxation times of water are relatively long, providing hyperpolarization lifetimes up to about 60-120 s, and the target molecule itself does not experience the harsh conditions during dissolution and transfer. The approach is sketched in Figure 1. Many details can also be found in ref. [30].
Importantly, it should be noted that using water as a hyperpolarization vector renders the DDNP largely independent of the NMR detection. Indeed, it was already shown that an extensive array of biomolecules can be studied at superior sensitivity with hyperpolarized water, from small metabolites such as arginine to large folded as well as disordered proteins and nucleic acids. The approach is often coined as HyperW in the literature. [85] Importantly, although coined HyperW/hyperpolarized water for the sake of simplicity, the hyperpolarized water/buffer is typically dissolved with D 2 O or a buffered variant thereof. Hence, the hyperpolarized solution to be mixed with the target molecule mostly consists, upon arrival at the NMR spectrometer, of D 2 O with a small percentage of HDO (typically 2-10 % depending on the setup). The degree of deuteration thereby allows one to select between longer hyperpolarization lifetimes (high deuteration) or stronger total signal enhancements (low deuteration). [30]

Mixing hyperpolarized water with a target solution
For HyperW, a hyperpolarized buffer and a target are typically mixed within an NMR tube already situated in the NMR spectrometer. A solution of the target waits in situ in the tube for the hyperpolarized solution to start detection immediately after mixing. [86] However, achieving quantitative and homogeneous mixing within a narrow 5 mm outer-diameter NMR tube poses a challenge, in particular for viscous samples.
Recently, another solution has been developed, where two solutions can be mixed ex situ in a dedicated "mixing chamber" prior to transfer to the NMR tube. [73] The system is described in Figure 2. The "hybrid sample shuttling system" (HySSS) is located atop a spectrometer and collects the hyperpolarized solution from the DDNP system. The solution is then mixed with the target solution waiting in a dedicated mixing chamber, degassed, and then hydraulically shuttled to the NMR tube. Mixing, degassing, and shuttling together takes less than 2 s. The result is a gas inclusion-free and homogeneously mixed solution. The system has been reported for use in radical scavenging applications. Translation to application in the context of hyperpolarized water/HDO (e. g., mixing water, a ligand, and a target) promises straightforward ease of biomolecular DDNP.
With a setup in place to perform these experiments on a regular basis, one can now envision how to implement HyperW-DDNP with existing workflows in structural biology.

Applications of hyperpolarized water to biomolecular NMR
Historically, the success of NMR spectroscopy can be traced back to its capacity to elucidate the solution-state structures of organic compounds. In synthetic chemistry, NMR is simply the standard method for sample analysis. In most articles, the Figure 1. The HyperW concept. 1) Water is hyperpolarized by microwave irradiation at cryogenic temperatures close to 1 K in a dedicated DNP apparatus (TEMPOL is typically added as the radical source together with a vitrification agent such as glycerol). 2) The water is then dissolved and transferred to a highfield NMR spectrometer for detection at ambient temperature. 3) By mixing the "super buffer" with a target protein within the NMR tube, proton exchange (and NOE albeit much lower efficiency) between solvent and amides, introduces hyperpolarization of the biomolecule. 4) Within ca, 60 s, before the water hyperpolarization decays to zero, rapid 2D or 3D spectroscopy can be recorded, capitalizing on signal enhancements of over two orders of magnitude. Alternatively, a series of 1D spectra can be detected. At t = 0 the protein amide proton signals are strongly enhanced; with time, their signals decay as the hyperpolarized water returns to its thermal equilibrium state. The data are adapted with permission from ref. [92]. Copyright: 2021, Springer Nature. methods and references employed are not even mentioned or cited anymore.
DDNP is far from reaching this level of broad applicability. However, we would like to draw the reader's attention to the possibility of integrating DDNP into existing workflows to add particular pieces of information to the structural puzzles and derive comprehensive pictures of the system under study. Hence, we want to advertise the use of DDNP as a complement to existing workflows rather than aiming at a stand-alone research platform. DDNP by HyperW is well-suited for this task as the target does not need special chemical properties. As will be shown below, even isotope enrichment is not necessary for many applications.
At the time of writing, three main fields of applications emerge that justify the integration of DDNP into the existing methodology of structural biology: 1) While the structures of stable interactants (e. g., apo states, free receptors, or ligands) and of final complexes (e. g., holo states, ligand-bound complexes) in molecular recognition or ligand binding events are typically resolvable by established means, the transition between them often remains entirely unclear (i. e., intermediates, encounter complexes etc.). DNP can provide structural models to depict these transitions and fill the knowledge gap. [87,88] 2) The structure of a protein or nucleic acid is typically determined under in-vitro conditions (in a crystal or at high concentrations in unphysiological buffers) that do not resemble in-vivo environments. Such altered conditions can significantly impact the structure and function of a biomolecule. DDNP can provide the structures of many targets under conditions that are much closer to physiological compared to conventional NMR in thermal equilibrium. [89][90][91] 3) The interactions, conformational transitions, and molecular recognition events of many substrates proceed on the millisecond to low seconds timescale, and it is notoriously difficult to follow their kinetics. With DDNP it is possible to achieve rapid real-time monitoring of such events while maintaining one main advantage of biomolecular NMR, namely residue resolution. [92,93]

Real-time monitoring
Processes that evolve on the high milliseconds to low seconds timescale can typically not be monitored by NMR spectroscopy.
Regarding ligand binding and protein-folding events, it would yet be very informative to track such events in real-time while capitalizing on the high resolution of NMR. This is typically problematic because of two reasons: 1) preparing a sample and starting the experiment often takes longer than the event to be monitored. 2) High intrinsic signal intensities are required for fast sampling, as signal averaging becomes unfeasible. HyperW by DDNP solves both problems at the same time. While the mixing with the arriving hyperpolarized solution (e. g., using the HySSS) is fast enough to start the experiment almost immediately after mixing (the target solution is typically waiting already in the NMR spectrometer), the hyperpolarization-based boost overcomes the need for signal averaging. As a result, real-time NMR with sampling rates > 1 s À 1 can be readily attained for a wide array of targets. [22,87,92,94,95] However, high temporal sampling rates can often only be realized by a series of onedimensional NMR spectra. It should be noted though that so-called ultrafast or single scan detection [96][97][98] can advantageously be combined with dissolution DNP. For this method, 2D spectra are recorded, typically, within a millisecond by pulsed gradient-based spatial encoding of the second dimension such that high temporal sampling can be achieved while retaining higher dimensional detection. This can be particularly useful when aiming at processes taking place on the low seconds timescale. [16,[99][100][101][102][103][104] The challenge to be overcome when using DDNP is sample convection upon injection of the hyperpolarized solution into the NMR tube, which impedes gradient-encoding. Thus, convection needs to settle before the process to be monitored is Figure 2. The hybrid sample shuttling system (HySSS) sits on top of the NMR spectrometer used for detection in DDNP experiments. It collects the sample within a collection chamber, which houses a "vortex breaker". This gadget enables degassing of the hyperpolarized water and simultaneous mixing (within 1 s) with a target solution waiting below the vortex breaker. Subsequently, the mixed sample is shuttled to the NMR spectrometer for detection by using a hydraulic driving liquid. The system is fully automatized, provides gas-inclusion free solutions, and enables controlled and reproducible mixing of the two solutions. Adapted with permission from ref. [73]. Copyright: 2022, American Chemical Society. completed, [69,105] an issue currently limiting applications to molecular recognition events.
Hence, the question of how to achieve residue resolution in a biomacromolecule remains when the conventional multidimensional NMR methods, for example, HSQC or HMQC, cannot be applied. Recently, a solution to this question has been suggested based on the residue dependence of the hyperpolarization flow between buffer and solute. [92] The HyperW method relies primarily on the chemical exchange of water protons with amide protons in a protein backbone or imino protons in a nucleic acid. The exchange replenishes the hyperpolarization between each NMR detection. [106] Hence, the exchange rate has to be faster than the repetition rate of the pulse sequence in order to see the signal of the underlying proton. In other words, the proton hyperpolarization of the target has to be replenished during the recovery delay τ REP . As a result, slowly exchanging residues remain invisible in HyperW experiments, while only faster exchanging ones receive hyperpolarization on the timescale of the pulse sequence. Hence, the number of detected residues can be minimized by tuning the recycling delay τ REP such that only the fastest exchanging residues receive hyperpolarization between each detection. Thus, it becomes possible to render the spectra so sparse that individual resonance can be distinguished even with 1D detection schemes. An analogy to selective isotope labelling [107][108][109][110] of particular amino acids can be drawn, where only signals of labelled residues are recorded. In the case of HyperW, "selective signal enhancement" can be achieved by filtering out the largest share of signals. The result is simultaneous time-and residue-resolved data on short timescales.
Optionally, the hyperpolarized water/HDO can be injected together with a ligand for the target if real-time monitoring of the ligand binding event is desired. Figure 3 shows an example of this approach, one for a protein.
1) The example [92] uses ubiquitin as a model protein dissolved in hyperpolarized water/HDO. By setting a short τ REP of 500 ms between detections of 15 N-edited 1D 1 H spectra to record a sparse set of 1 H N amide protons, the spectra were rendered so sparse that the hyperpolarization of individual amino acids could be followed with a 2 s À 1 sampling rate. Only 13 amino acids received hyperpolarization efficiently during the 500 ms recovery delay. Seven signals could be unambiguously assigned. These acted as site-selective probes to track the real-time dynamics of the protein.
Indeed, the hyperpolarization of individual amino acids could be monitored. Ubiquitin surface residues were enhanced in hyperpolarized physiological saline buffer. The short recycling time of 0.5 s allowed the hyperpolarization of only a hand full of residues, which could even be distinguished with 1D spectroscopy. A reference 2D spectrum enabled resonance assignment in the figure, for the small set of enhanced residues the color code highlights the observed enhancements. The hyperpolarization decay for the resolved residues is shown at the bottom. Seven residues were unambiguously assigned, and their temporal evolution could be tracked. Adapted with permission from ref. [92]. Copyright: 2021, Springer Nature.
2) A second, similar example, [93] was recently published for an RNA aptamer (GSR apt ), which was monitored with a similar approach using a very short recycling time of only~100 ms. A ligand, hypoxanthine, was co-dissolved in hyperpolarized water, and a one-dimensional time series could follow its binding to the signal-enhanced target RNA upon mixing solvent, target, and ligand. Again, resonances that remained above the detection threshold despite the fast pulse sequence recycling were exploited as probes that simultaneously reported residue and temporally resolved data upon ligand binding. It should be noted that real-time monitoring of nonexchanging protons (H α or side chain) is similarly possible by using exchange-relayed NOEs. [84] Concerning the timescales of processes such as molecular recognition or unfolding processes, it should be noted that these can take place on a very wide range, [111,112] from fast picosecond structural vibrations [113] to exchange processes taking tens of minutes. [114] In addition, the pace of biomolecular events can depend drastically on pH and/or temperature. [115] Hence, when considering employing HyperW for real-time detection, the rate of the monitored event should be fitting the experimental design. As a rule of thumb, events with a rate of 0.1 to~5 s À 1 can be characterized using HyperW. The lower limit is mainly due to the lifetime of the hyperpolarization, while the upper limit is due to the time needed to mix the hyperpolarized water/HDO with the target solution. Again, many details can be found in ref. [30].

Rapid correlation spectroscopy
Some species are just not fit for conventional thermal equilibrium NMR. For example, very low sample concentrations or short lifetimes of intermediates prohibit using conventionally employed 2D or 3D NMR spectroscopy. In this case, HyperW by DDNP can be a well-suited complement to existing assays.
The challenge for residue-resolved multidimensional NMR of biomolecules by DDNP lies in the limited lifetime of the water hyperpolarization, typically on the order of a minute. Hence, multidimensional spectra must be recorded very fast. Luckily, the signal enhancement in a hyperpolarized buffer overcompensates the time constraints by outdating the need for signal averaging and enabling 2D (e. g., the most common 1 H, 15 N HSQC) spectra within a few (< 10) seconds -even with intensities superior to lengthy signal-averaged correlation spectra in thermal equilibrium.
Using this approach, it has been shown that short-lived folding intermediates (Figure 4a) can be characterized or exchange processes with minor species [85] enlightened as their signals become drastically amplified.
Interestingly, there is another twist to the HyperW-boosted correlation spectroscopy. Chemical proton exchange typically leads to signal broadening and has spurious effects on the quality of the NMR spectra. With HyperW though, the chemical exchange is turned into an advantage since it boosts the signals of the fast exchanging residues. Even when signals are significantly broadened and remain below the detection threshold in conventional NMR, they are often the most enhanced species in a hyperpolarized buffer.
Conversely, mapping solvent-exposed surfaces of proteins and peptides is also a straightforward possibility. As mentioned above, the water hyperpolarization is transferred mainly by chemical exchange, rendering spectra selective for residues with vivid solvent interaction. Three recent examples highlight applications of this feature (Figure 4b). 1) It has been shown that HyperW selectively enhances surface residues of globular, well-folded proteins [90] in 1 H, 15 N HMQC spectra. This feature can become very handy when surface properties, such as exposed binding sites, are to be studied. For the model protein, ubiquitin, [90] all but three surface residues were shown to be exclusively boosted by amide proton exchange. Three residues, which possessed labile side-chain protons, received hyperpolarization by exchangerelayed NOE pathway in addition. [90] 2) Compacted regions of IDPs, which are better shielded from the solvent than the rest of the protein, will not receive hyperpolarization as efficiently from the buffer as unfolded domains. Hence, these sites can be mapped by recording a 2D spectrum in hyperpolarized water and identifying weak or missing resonances. Thus, it has been shown for the IDP osteopontin [89] that the vital heparin interaction site is efficiently compacted and shielded from the surrounding water. It is only opening and solvent-exposed upon interaction with its target ligand leading to the sudden appearance of the residues within the binding epitope in HyperW spectra (Figure 4b). 3) Membrane binding of small peptides was shown to lead to efficient shielding from the solvent for residues inserted into the membrane. Hence, a similar effect as described above was observed, where only the nonbinding residues remain visible. [117] Interestingly, this was studied with proton-onlydetected NMR (a selective 1 H, 1 H COSY). Isotope enrichment was unnecessary, and the studied peptide could be used for HyperW without any modification. This last point is essential when aiming to include HyperW in existing biochemistry workflows. The target substances are often produced with natural isotopic abundance, and only proton-only spectroscopy can readily be applied.

Computational matching of HyperW features
A very recent development in the use of HyperW is based on its combination with computational methods. The principle is simple: record a HyperW spectrum and use it as a fingerprint of the molecule under study. Then, match the fingerprint with a computational model to verify the latter. This approach yields high-resolution structural models under conditions not accessible by established NMR techniques. Two recent examples can be found in the literature: 1) A transcription factor (the Myc-associated factor X, MAX) was studied at concentrations that are typically too low for conventional thermal equilibrium NMR, namely, 1 μM. [91] With dissolution DNP and HyperW, a 1D 1 H spectrum could yet be acquired under these conditions with a good signal-to-noise ratio. Peaks were detected that remained below the detection threshold in conventional NMR even after extensive signal averaging over one day. In contrast, HyperW could provide a complete spectrum while only requiring a few minutes of NMR machine time (Figure 5a-d).
To obtain the structure of the transcription factor in the next step, the 1D spectrum was then reproduced in silico from a model structure derived from molecular dynamics (MD) simulations. By computing the chemical shifts from an ensemble of MD structures, the HyperW spectrum could be matched and, thus, the modelled conformation confirmed. For the case of MAX, it was thus shown that the protein adopts a The blue spectrum was recorded within a few seconds of initiating the refolding event by diluting an 8 M urea solution with hyperpolarized water/HDO. The grey spectrum shows the fully folded final state as reference. The selectivity of the signal enhancement is shown on the right. Adapted with permission from ref. [88]. Copyright: 2019, the American Chemical Society. b) By using a recycling time of 1 s, ca. 20 surface residues of ubiquitin were selectively enhanced in hyperpolarized water/HDO. The hyperpolarized residues (red) light up in HyperW within a few seconds. The conventional spectrum recorded within a few hours (black) is shown as a reference. Adapted with permission from ref. [90]. Copyright: 2018, John Wiley and Sons. c) Example of rapid correlation spectroscopy to map solvent-exposed and shielded residues. The IDP osteopontin unfolds its compacted core upon binding to heparin. The left panel shows a correlation spectrum obtained within ca. 10 s of binding of the glycines exposed to the solvent by unfolding, which are only visible at this point. The grey spectrum shows a thermal equilibrium spectrum under similar, near-physiological conditions. Only with selective hyperpolarization can residue resolution be achieved, as otherwise the signal overlap is very severe under the experimental conditions. Adapted with permission from ref. [89]. Copyright: 2016, John Wiley and Sons and from ref. [30]. Copyright: 2022, Springer Nature Limited. (Note that using SABRE, Tickner et al. monitored short-lived intermediates, too, during the reaction of pyruvate with H 2 O 2 .) [116] formerly undocumented conformation at high dilutions encountered under physiological conditions. While typical NMR conditions (concentrations close to 1 mM) lead to adopting a well-documented dimer conformation, [14,[118][119][120] a globular folded monomer dominated the conformational space at near-physiological concentrations.
It should be noted though that the computation of 1 H chemical shifts based on protein structures is quite imprecise and can lead to flawed data interpretation when not carefully executed. Indeed, this approach requires appropriate control experiments as well as data interpretation within the precision of the available data. For the case shown in Figure 5, this issue was taken into account through three points: a) Only the envelope of the hyperpolarized spectrum was reconstructed, no individual resonances were interpreted. b) The spectrum was computed by averaging a statistically meaningful set of simulated protein structures (in total 150 MD snapshots). c) A series of control experiments ruled out all other possible data explanations (repetition runs, negative controls).
Further, it should be mentioned that the influence of possible changes in pH/pD upon dissolution in hyperpolarized water has to be taken into account too, for example, again by a number of control experiments (for details see the Supporting Information of ref. [91]).
2) Computational analysis of HyperW data could reveal solvent-target interactions by boosting the direct polarization flow between the solvent shell and the solute. [84] Thus, it was possible to follow direct water-to-target NOEs in real-time. These effects are typically too weak to be observed with conventional NMR but can even become dominating when amplified in hyperpolarized water. In particular, the hydrophobic side-chain protons of arginine and poly-aspartate were monitored after dissolution in hyperpolarized water. The magnetization transfer from the buffer led to a transient intensity modulation of the side-chain signals (Figure 5e-g). By computing the NOE rates and comparing them with the experimentally observed signal intensities, these effects were deciphered, and the contact times and coordination numbers of the water in the solvent shell could be determined. It should be noted that also, in this case, isotope enrichment was not necessary. [84] The radio plot in Figure 5g visualizes the combined results of this approach. It correlates the computed NOE rate, σ L , between solvent and target (here arginine), the observed signal change ɛ I due to the NOE, as well as the simulated average . c) Comparing the hyperpolarized spectrum (pink) with a spectrum computed from the MD data (blue bars) proved the validity of the simulated structure. d) The right panel highlights that conventional thermal equilibrium NMR (purple and black spectra) could not provide a representative spectrum of the MAX monomer either as it would adopt a different fold at high concentrations or as signal intensities were too weak at low concentrations. Adapted with permission from ref. [91]. Copyright: 2022, the American Association for the Advancement of Science. e) In this example of HyperW, arginine was hyperpolarized through NOEs between the water and the nonlabile side-chain protons H β , H γ , and H δ , either directly or through exchange-relay. f) The panel shows how the direct NOE reduces signal intensities during the first 5 seconds after mixing, while the slower solvent-relayed NOE boosts the signals at longer times, before the decay to their thermal equilibrium values within ca. 100 s.
g) The radio plot shows a comparison of the direct signal reduction 1/ɛ i and the solvent residence time, τ, as well as the direct NOE rate σ L . The latter two parameters were extracted from MD simulations. The match between the computed values and the experimentally observed ones again enables an interpretation of the NMR spectra and allows for validation of the NOE pathways computed in silico. Adapted with permission from ref. [84]. Copyright: 2022, Royal Society of Chemistry. contact time τ of water at the sites of H α (red), H β (green), H γ (purple) and H δ (blue). For all protons, a positive correlation of all three parameters can be seen, highlighting the match between the computed relaxation parameters and the experimental hyperpolarized NOEs.
These two examples showcase that HyperW is well-apt to unravel subtle yet vital features that typically remain out of the reach of thermal equilibrium NMR. To this end, it can readily be implemented with other methods, such as MD simulations, to derive structural models and interactions of target molecules.

More Options
It should be noted that other hyperpolarization techniques exist, based on so-called para-hydrogen, which similarly enable the signal enhancements of biologically relevant molecules. Dihydrogen (H 2 ) can adopt two spin isomers para-and orthohydrogen. The former is typically considered the singlet and the latter the triplet state. At room temperature, the ortho isomer is more abundant (3 : 1). [121] At low temperatures, para-hydrogen can yet be significantly overpopulated (usually with the help of a catalyst to speed up the conversion). [122] A polarization of 100 % is achievable < 20 K. Using liquid nitrogen is possible to obtain up to 65 % parahydrogen. [123] Parahydrogen (pH 2 ) can then be used to protonate unsaturated sites, break their symmetry, and create a spin state with a high nuclear polarization (so-called PHIP; para-hydrogen-induced polarization). This can be followed by magnetization transfer to neighbouring nuclei. [124] PHIP has been used for imaging mice [125] and studying prostate tumour cells [126] and invitro metabolism. [127] It has also been demonstrated on acetate, [128] lactate, [129] succinate, [130] glucose, [131] acetylated phenylalanine, tryptophan, tyrosine and acetyltyrosine, [132] γ-aminobutyric acid (GABA; by hydrogenation of trans-amino-crotonic acid). [133] Designing the hydrogenation reaction can be challenging, though, for example, because precursors might tautomerize. [134] Recent solutions include functionalizing the target with an unsaturated auxiliary group (UAG), performing the PHIP, transferring the polarization to the desired nucleus, and removing the UAG (dubbed "side-arm hydrogenation"; PHIP-SAH). [128] Typical UAGs are allyl-, propargyl-and vinyl esters, with the latter showing the best results when properly deuterated. [135] It should be noted that PHIP was also reported to enhance NMR of oligopeptides [136] using propargylglycine or O-propargyl-l-tyrosine as reactive moieties [137] approaching hyperpolarization of larger proteins.
As an alternative that does not require any reaction with an unsaturated moiety, the so-called SABRE (signal amplification by reversible exchange) method employs parahydrogen but avoids the hydrogenation reaction. This is possible thanks to the formation of complexes that mimic the hydrogenated spin system but do not undergo the actual reaction. [138] A metal (typically iridium) [139] complex simultaneously coordinates pH 2 and the target molecule, creating a contact point that triggers a magnetization transfer from the former to the latter.

An alternative at lower magnetic fields: Overhauser DNP
An alternative to DDNP for liquid-state hyperpolarization is Overhauser DNP (ODNP). Radicals such as TEMPOL are dissolved in a liquid-state sample to be subsequently microwave irradiated. The method is conceptionally applicable to any solutionstate sample; however, it is only efficient at low to moderate fields (typically up to 1 T). At high magnetic fields, the DNP efficiency is yet painfully reduced. Hence, attaining the resolution necessary for residue-resolved NMR of larger proteins or nucleic acids is often impossible. Another problem is microwave-induced heating, which can be challenging, particularly for water-based systems. In this regard, a combination of microwave irradiation and dynamic sample cooling under flow has been suggested to enable hyperpolarized water imaging. [60,166]

An alternative for solid samples: LT-MAS DNP
For low-temperature magic angle spinning (LT-MAS) DNP, a solid powder is typically doped with nitroxide-based biradicals and subsequently freeze-quenched to be measured at temperatures between 20 and 100 K. [54,55,[167][168][169] By high-power microwave irradiation (often employing a gyrotron), the NMR signals are amplified, enabling to record MAS NMR spectra in a fraction of the time needed by conventional thermal equilibrium techniques. Sometimes even formerly invisible signals can be lifted above the detection threshold. Thus, it becomes possible to, for example, measure solid-state NMR spectra of entire cells or large membrane-bound receptors with superior sensitivity. However, an often-encountered challenge is the width of the hyperpolarized lines. Signals are broadened due to the presence of the radicals, and recording biomolecular NMR spectra at residue resolution is often hard to achieve.

Conclusion
Many recent examples show the broad applicability of DDNP to biomolecular research, from nucleic acids, peptides, and proteins to membrane vesicles and small metabolites. As all these biomolecules are soluble in water, HyperW by DDNP can be readily used to boost their signals. As a result, a wide array of possible applications emerges, which can complement existing structural biology programs: from structure determination at very low concentrations or short-lived folding intermediates to real-time monitoring of molecular recognition events or characterization of the solvent shell.
Such information is often cumbersome to obtain with established workflows; thus, "black spots" stain the otherwise well-drawn picture of a biomolecular system. DDNP and hyperpolarized NMR can fill these gaps as it is highly complementary to existing approaches based on thermal-equilibrium NMR spectroscopy, EM, and XRD.