Hyperpolarizing DNA Nucleobases via NMR Signal Amplification by Reversible Exchange

The present work investigates the potential for enhancing the NMR signals of DNA nucleobases by parahydrogen-based hyperpolarization. Signal amplification by reversible exchange (SABRE) and SABRE in Shield Enables Alignment Transfer to Heteronuclei (SABRE-SHEATH) of selected DNA nucleobases is demonstrated with the enhancement (ε) of 1H, 15N, and/or 13C spins in 3-methyladenine, cytosine, and 6-O-guanine. Solutions of the standard SABRE homogenous catalyst Ir(1,5-cyclooctadeine)(1,3-bis(2,4,6-trimethylphenyl)imidazolium)Cl (“IrIMes”) and a given nucleobase in deuterated ethanol/water solutions yielded low 1H ε values (≤10), likely reflecting weak catalyst binding. However, we achieved natural-abundance enhancement of 15N signals for 3-methyladenine of ~3300 and ~1900 for the imidazole ring nitrogen atoms. 1H and 15N 3-methyladenine studies revealed that methylation of adenine affords preferential binding of the imidazole ring over the pyrimidine ring. Interestingly, signal enhancements (ε~240) of both 15N atoms for doubly labelled cytosine reveal the preferential binding of specific tautomer(s), thus giving insight into the matching of polarization-transfer and tautomerization time scales. 13C enhancements of up to nearly 50-fold were also obtained for this cytosine isotopomer. These efforts may enable the future investigation of processes underlying cellular function and/or dysfunction, including how DNA nucleobase tautomerization influences mismatching in base-pairing.


Introduction
With an energy difference between nuclear spin states that is much less than the thermal energy, kT, NMR accesses only~10 −3 % of available molecules (e.g.,~0.0032% for 1 H spins at B 0 = 9.4 T), and thus suffers from poor detection sensitivity. Hyperpolarization methods [1] such as dynamic nuclear polarization (DNP) [2,3], spin-exchange optical pumping (SEOP) [4,5], and parahydrogen-induced polarization (PHIP) [2,6,7] produce highly non-Boltzmann spin populations to greatly enhance NMR signals. With hyperpolarization, the rapid NMR signal acquisition of less-sensitive and low-natural-abundance nuclei (e.g., 15 N, 13 C, 129 Xe, etc.) is greatly facilitated, making it feasible to perform the NMR/MRI of low-concentration species. Among these hyperpolarization techniques, PHIP-based approaches [6][7][8] are attractive because they can be conducted rapidly with low operational In the presence of p-H2 and substrate, a pre-catalyst [e.g., Ir(1,5-cyclooctadeine)(1,3bis(2,4,6-trimethylphenyl)imidazolium)Cl, Ir(COD)(IMes)Cl)] first undergoes a 4-to 6-coordinate transformation induced by the initial exposure to p-H2, giving rise to the standard "IrIMes" SABRE catalyst [17]. The formation of a transient hexa-coordinate center, wherein the symmetry of the nascent hydride spins is broken, allows spin order to be transferred to reversibly ligating substrates through the J-coupling network, particularly within a mixing field (Bmixing field) that roughly matches the frequency difference between source (parahydrogen-derived hydrides) and target spins (protons or other spin−1/2 nuclei of the exchangeable substrate) to the magnitude of the J coupling. SABRE can thus produce solution-phase hyperpolarization for various applications, including the creation of biologically-friendly hyperpolarized spectral probes for NMR, and metabolic contrast agents for MRI [13,14,[18][19][20][21][22][23][24]. However, despite its advantages, a persistent challenge of the methodology has been the limited range of applicable substrates due to the necessity of requiring an sp-and sp 2-hybridized lone pair to interact with the catalyst's Ir-metal center, initially limiting SABRE mostly to nitrogen heterocycles [2,9,11,12,16]. While a number of approaches have recently been developed to greatly widen SABRE's scope [18][19][20] (including biologically relevant carboxylic acid derivatives like α-ketoglutarate and pyruvate) [21,22,[25][26][27][28], the myriad biologically relevant NHCs nevertheless make this structural motif of continued interest for SABRE targets for both fundamental studies and envisioned applications [23].
Of the NHCs studied thus far, derivatives of purine comprise the most widely occurring NHC family in nature and consist of fused pyrimidine and imidazole rings [29][30][31]. Moreover, derivatives of purine and pyrimidine, such as the nucleobases adenine, thymine, guanine, and cytosine, comprise DNA and participate in specific hydrogen bonding to complete canonical (A:T/G:C) Watson-Crick base pairing [32] (Figure 2). The resulting In the presence of p-H 2 and substrate, a pre-catalyst [e.g., Ir(1,5-cyclooctadeine)(1,3-bis(2,4,6trimethylphenyl)imidazolium)Cl, Ir(COD)(IMes)Cl)] first undergoes a 4-to 6-coordinate transformation induced by the initial exposure to p-H 2 , giving rise to the standard "IrIMes" SABRE catalyst [17]. The formation of a transient hexa-coordinate center, wherein the symmetry of the nascent hydride spins is broken, allows spin order to be transferred to reversibly ligating substrates through the J-coupling network, particularly within a mixing field (B mixing field ) that roughly matches the frequency difference between source (parahydrogen-derived hydrides) and target spins (protons or other spin−1/2 nuclei of the exchangeable substrate) to the magnitude of the J coupling. SABRE can thus produce solution-phase hyperpolarization for various applications, including the creation of biologically-friendly hyperpolarized spectral probes for NMR, and metabolic contrast agents for MRI [13,14,[18][19][20][21][22][23][24]. However, despite its advantages, a persistent challenge of the methodology has been the limited range of applicable substrates due to the necessity of requiring an spand sp 2hybridized lone pair to interact with the catalyst's Ir-metal center, initially limiting SABRE mostly to nitrogen heterocycles [2,9,11,12,16]. While a number of approaches have recently been developed to greatly widen SABRE's scope [18][19][20] (including biologically relevant carboxylic acid derivatives like α-ketoglutarate and pyruvate) [21,22,[25][26][27][28], the myriad biologically relevant NHCs nevertheless make this structural motif of continued interest for SABRE targets for both fundamental studies and envisioned applications [23].
Of the NHCs studied thus far, derivatives of purine comprise the most widely occurring NHC family in nature and consist of fused pyrimidine and imidazole rings [29][30][31]. Moreover, derivatives of purine and pyrimidine, such as the nucleobases adenine, thymine, guanine, and cytosine, comprise DNA and participate in specific hydrogen bonding to complete canonical (A:T/G:C) Watson-Crick base pairing [32] (Figure 2). The resulting primary nucleic acid sequence gives rise to secondary structures composed of supercoils that result from protein complexes wound about a DNA helix [33,34]. These structures, along with other modifications (e.g., methylation, histone modification, etc.), are responsible for epigenetic gene regulation [34,35], which has far-reaching consequences in both normal (e.g., cell differentiation) and pathogenic development (e.g., cancer) [34,36,37]. This type of control primarily dictates which segments of the genetic sequence are available for reading (thereby acting as on-and-off switches [38]), depending on the shape, spacing, and composition of the helix and associated further levels of the structure. With RNA these effects are further compounded by the additional functions of RNA, ranging from signaling to catalysis [39,40].
Molecules 2023, 28, x FOR PEER REVIEW 3 of 16 primary nucleic acid sequence gives rise to secondary structures composed of supercoils that result from protein complexes wound about a DNA helix [33,34]. These structures, along with other modifications (e.g., methylation, histone modification, etc.), are responsible for epigenetic gene regulation [34,35], which has far-reaching consequences in both normal (e.g., cell differentiation) and pathogenic development (e.g., cancer) [34,36,37]. This type of control primarily dictates which segments of the genetic sequence are available for reading (thereby acting as on-and-off switches [38]), depending on the shape, spacing, and composition of the helix and associated further levels of the structure. With RNA these effects are further compounded by the additional functions of RNA, ranging from signaling to catalysis [39,40]. Interestingly, nucleobase tautomers play a key role in epigenetic control, as they can alter the secondary level of the DNA's structure [41,42]. For example, tautomerization of a nucleobase can lead to distortions in the shape of (or spacing between) DNA strands, which are then targeted by DNA repair enzymes. In some cases, an enzyme may fail to initiate the repair, thus epigenetically leading to silencing or changes in the gene's expression [42], with implications for cancer and other diseases-depending on the location of the tautomerization [35]. Thus, any method that can probe epigenetic interactions has potential value in both clinical and research settings [35,37].
Due to the ubiquity of the nucleobases that comprise DNA and RNA, along with the myriad of other functions that nucleobases perform (e.g., cell signaling [43][44][45] and ATP [46]), such compounds are ripe for investigation by methods that may be able to sensitively and non-invasively report on nucleobase interactions in living systems or biological media. Such developments would be particularly relevant if those methods could complement the capabilities of other bioanalytical techniques under development (e.g., Refs. [43,47]). Given that the detection of nucleobase-containing biomarkers of disease have opened new areas of research, there should be natural interest in the investigation of rapid and inexpensive parahydrogen-based methods for hyperpolarizing nucleobases and their use in potential applications; however, to date there have only been a few such studies reported. In one type of approach, the unique resonances of catalyst-bound hyperpolarized hydrides [48][49][50] are used to detect the presence of specific nucleobases (and their tautomers) with high sensitivity, including in complex biological media [51,52]. Hövener et al. reported 1 H hyperpolarization of adenine and adenosine as part of efforts to demonstrate continuous polarization in MRI applications [53]. More recently, as part of a larger Interestingly, nucleobase tautomers play a key role in epigenetic control, as they can alter the secondary level of the DNA's structure [41,42]. For example, tautomerization of a nucleobase can lead to distortions in the shape of (or spacing between) DNA strands, which are then targeted by DNA repair enzymes. In some cases, an enzyme may fail to initiate the repair, thus epigenetically leading to silencing or changes in the gene's expression [42], with implications for cancer and other diseases-depending on the location of the tautomerization [35]. Thus, any method that can probe epigenetic interactions has potential value in both clinical and research settings [35,37].
Due to the ubiquity of the nucleobases that comprise DNA and RNA, along with the myriad of other functions that nucleobases perform (e.g., cell signaling [43][44][45] and ATP [46]), such compounds are ripe for investigation by methods that may be able to sensitively and non-invasively report on nucleobase interactions in living systems or biological media. Such developments would be particularly relevant if those methods could complement the capabilities of other bioanalytical techniques under development (e.g., Refs. [43,47]). Given that the detection of nucleobase-containing biomarkers of disease have opened new areas of research, there should be natural interest in the investigation of rapid and inexpensive parahydrogen-based methods for hyperpolarizing nucleobases and their use in potential applications; however, to date there have only been a few such studies reported. In one type of approach, the unique resonances of catalyst-bound hyperpolarized hydrides [48][49][50] are used to detect the presence of specific nucleobases (and their tautomers) with high sensitivity, including in complex biological media [51,52]. Hövener et al. reported 1 H hyperpolarization of adenine and adenosine as part of efforts to demonstrate continuous polarization in MRI applications [53]. More recently, as part of a larger effort to demonstrate SABRE enhancement in a large number of different molecules, Colell et al. [12] showed 15 N enhancement of 15 N-labelled adenine (ε~200) as an example where enamine-imine tautomers are sensitive to SABRE-SHEATH [12]; this sensitivity may be exploitable in potential biological applications. Finally, work has also been performed to optimize 1 H and 15 N hyperpolarization of pyrimidine, which is the framework for cytosine and thymine [54]. In the present work, we explore the applicability of SABRE and SABRE- SHEATH to various nucleobases, motivated by the desire to expand the current scope of these approaches to support the future development of new techniques for studying various biological systems and diseases-both for cellular studies and ultimate potential in vivo applications.

Materials and Methods
The homogenous (Ir(COD)(IMes)Cl, MW = 639.67 g·mol −1 ) pre-catalyst was synthesized as previously described [17]. Each NMR solution consisted of 4 mM catalyst and 40 mM substrate in 600 µL solution of either 100% C 2 D 5 OD (3-methyladenine) or 92% C 2 D 5 OD: 8% D 2 O (cytosine and 6-O-methylguanine); D 2 O was necessary to increase the solubility of the latter nucleobases. Experimental setups at SIUC and Vanderbilt were described previously [11,16,19,55]. Each sample solution was transferred to a 5 mm O.D. NMR tube affixed with a 0.25-inch O.D. Teflon tube, sealed with a wye-connector, and activated at elevated temperatures (60-70 • C) for at least 10 min prior to acquisition. The experimental setup is schematically shown in Figure 3. effort to demonstrate SABRE enhancement in a large number of different molecules, Colell et al. [12] showed 15 N enhancement of 15 N-labelled adenine (ε~200) as an example where enamine-imine tautomers are sensitive to SABRE-SHEATH [12]; this sensitivity may be exploitable in potential biological applications. Finally, work has also been performed to optimize 1 H and 15 N hyperpolarization of pyrimidine, which is the framework for cytosine and thymine [54]. In the present work, we explore the applicability of SABRE and SABRE-SHEATH to various nucleobases, motivated by the desire to expand the current scope of these approaches to support the future development of new techniques for studying various biological systems and diseases-both for cellular studies and ultimate potential in vivo applications.

Materials and Methods
The homogenous (Ir(COD)(IMes)Cl, MW = 639.67 g·mol −1 ) pre-catalyst was synthesized as previously described [17]. Each NMR solution consisted of 4 mM catalyst and 40 mM substrate in 600 µ L solution of either 100% C2D5OD (3-methyladenine) or 92% C2D5OD: 8% D2O (cytosine and 6-O-methylguanine); D2O was necessary to increase the solubility of the latter nucleobases. Experimental setups at SIUC and Vanderbilt were described previously [11,16,19,55]. Each sample solution was transferred to a 5 mm O.D. NMR tube affixed with a 0.25-inch O.D. Teflon tube, sealed with a wye-connector, and activated at elevated temperatures (60-70 °C) for at least 10 min prior to acquisition. The experimental setup is schematically shown in Figure 3. Hydrogen gas (H2) from a cylinder (with initial p-H2 fraction of 25%) is passed over a catalyst bed of FeO(OH) cooled in liquid nitrogen for ~50% conversion to p-H2. A mass flow controller (MFC) at 150 mL min −1 allows for the precise regulation of p-H2 bubbling rate within the NMR tube; pressure was typically maintained at 75 psi. 1 H SABRE and 15 N/ 13 C SABRE-SHEATH experiments were performed by bubbling p-H2 through the sample while placed in either the NMR magnet's fringe field (~5 mT) or a mu-metal magnetic shield (~1 µ T), respectively, prior to rapid manual transfer of the sample to the 9.4 T magnet for high-field detection.
Parahydrogen generators used in this work provided either: p-H2 enrichment of ~50% (at 75 psi and 150 mL min −1 bubbling rate or at 75 psi and 110 mL min −1 bubbling rate at SIUC and Vanderbilt, respectively); or, ~90% p-H2 (for some heteronuclear studies at Vanderbilt). Most 15 N/ 13 C SABRE-SHEATH experiments were performed at Vanderbilt University using a µ -metal shield that was degaussed manually with a Variac and degaussing Hydrogen gas (H 2 ) from a cylinder (with initial p-H 2 fraction of 25%) is passed over a catalyst bed of FeO(OH) cooled in liquid nitrogen for~50% conversion to p-H 2 . A mass flow controller (MFC) at 150 mL min −1 allows for the precise regulation of p-H 2 bubbling rate within the NMR tube; pressure was typically maintained at 75 psi. 1 H SABRE and 15 N/ 13 C SABRE-SHEATH experiments were performed by bubbling p-H 2 through the sample while placed in either the NMR magnet's fringe field (~5 mT) or a mu-metal magnetic shield (~1 µT), respectively, prior to rapid manual transfer of the sample to the 9.4 T magnet for high-field detection.
Parahydrogen generators used in this work provided either: p-H 2 enrichment of~50% (at 75 psi and 150 mL min −1 bubbling rate or at 75 psi and 110 mL min −1 bubbling rate at SIUC and Vanderbilt, respectively); or,~90% p-H 2 (for some heteronuclear studies at Vanderbilt). Most 15 N/ 13 C SABRE-SHEATH experiments were performed at Vanderbilt University using a µ-metal shield that was degaussed manually with a Variac and degaussing coil prior to use [56]; all of those experiments used medium-wall NMR tubes. All NMR experiments were performed on either an Agilent 400 MHz DD2 spectrometer with a wide-bore actively shielded (Oxford) magnet (SIUC) or a Bruker AVANCE III 400 MHz spectrometer with a narrow-bore actively shielded magnet (Vanderbilt). Single-scan acquisitions with 10 • -pulses were used to acquire both SABRE-enhanced 1 H spectra and 1 H spectra from thermally polarized samples. Pulses of 90 • of 18 µs ( 15 N) or 10 µs ( 13 C) and 1 scan were used to acquire 15 N and 13 C SABRE-SHEATH spectra. 15 N thermally polarized reference spectra were acquired with a standard 8.65 M 15 N 2 -imidazole solution (in D 2 O) using a 90 • -pulse of 18 µs and 250 s pre-acquisition delay time, and one scan. 13 C thermally polarized reference spectra were acquired with the activated cytosine solution at 70 • C using a 90 • -pulse of 18 µs, 250 s delay time, and one scan.

1 H SABRE of DNA Nucleobases and Ethanol
Here we probe the efficacy of 1 H SABRE enhancement for the DNA nucleobases 3-methyladenine, 6-O-methylguanine, and cytosine (others were attempted but lacked sufficient solubility under our conditions). We begin by showing 1 H NMR spectra from thermally polarized and SABRE enhanced 3-methyladenine in 100% C 2 D 5 OD at 70 • C ( Figure 4a). Interestingly, both aromatic hydrogens (labelled 1 H A and 1 H B ) are hyperpolarized (albeit weakly), but with 1 H A exhibiting slightly larger (factor of~1.5) enhancement over 1 H B , suggestive of preferential binding of the imidazole ring over the pyrimidine ring to the catalyst. Naively, such differential enhancement might have also been explained by differences in relaxation. However, 1 H A exhibits a shorter T 1 time constant of~5 s in comparison to 1 H B (~8 s, albeit measured at high field instead of the mixing field; Figure 4d; at milli-tesla and micro-tesla fields, the relaxation rates may be significantly different [57], although here we expect the qualitative trend to be the same). The faster rate of hyperpolarization decay for 1 H A presumably reflects greater intermolecular interactions with the catalyst itself (which can act as a relaxation agent) [58], as well as possible contributions from hydrogen-bonding interactions with C 2 D 5 OD in the complex (discussed below).
The present 3-methyladenine enhancement pattern is the reverse of that observed by Hövener et al. in adenine, which included a larger enhancement for the pyrimidine 1 H B resonance (by a factor of approximately two) compared to that of the imidazole 1 H A spin [53]. Such behavior might be rationalized by the fact that 3-methyladenine has one fewer available sp 2 -hybridized nitrogen atoms on the pyrimidine ring to bind to the catalyst; on the other hand, Hövener et al. also observed a larger enhancement for the imidazole resonance in adenosine despite the increased sterics likely caused by the sugar moiety [53]. More generally, binding with the 5-membered ring might be expected to be preferred because of the reduced sterics/stronger binding of 5-membered rings (e.g., imidazole) compared to 6-membered rings (e.g., pyridine). For example, high 15 N polarization values in excess of 50% were reported for methylated imidazole with -CH 3 in the orthoposition in the case of metronidazole [13,14,59], but no detectable 15 N polarization (P 15N < 0.01%) was seen for methylated pyridine in the ortho-position in the case of 2-picoline [60]. 1 H SABRE of 3-methyladenine as a function of mixing field strength ( Figure 4b) shows a maximum enhancement for both signals at~5.2 mT, which was then selected as the mixing field for the remaining 1 H SABRE experiments presented here. 1 H SABRE of 3-methyladenine as a function of temperature (60-70 • C, Figure 4c) shows an increase in enhancement with temperature, which can be attributed to the better matching of exchange rates of 3-methlyadenine and p-H 2 with the catalyst, once the temperature was raised sufficiently to help mitigate the otherwise poor solubility of the substrate in 100% C 2 D 5 OD below 60 • C. Figure 4a also reveals a weak 1 H SABRE of the resonance from residual -OH (slightly shifted due to temperature drop during p-H 2 bubbling) and -CHD (residual 1 H) of the bulk solvent, C 2 D 5 OD. There have been only a few previous studies of hyperpolarized alcohol solvents via SABRE (e.g., Ref. [61]). Currently, there are three proposed mechanisms for hyperpolarization of solvent alcohol molecules: (1) hyperpolarization through direct coordination to the metal center; (2) A SABRE-RELAY [18] type mechanism (particularly in a slightly acidic environment), wherein a hydrogen ion transfers to free substrate (post hyperpolarization), is hyperpolarized from spin coupling to the substrate protons, and then exchanges with the alcohol group on the solvent; and (3) a solvent molecule hydrogenbonds to a free nitrogen of a (e.g., catalyst-bound) substrate, allowing the spin order to transfer via the extended scalar coupling network. Given that 1 H A shows greater enhancement over 1 H B , the imidazole ring of 3-methyladenine contains two nitrogen atoms, and only a single Ir-hydride signal is present (−22 ppm), we suggest that the ethanol solvent molecules may become hyperpolarized predominantly through hydrogenbonding to 3-methyladenine while bound at the metal-center ( Figure 5); direct binding may also contribute, though it is expected that SABRE-RELAY is unlikely to be a significant contributor because of the relatively low 1 H enhancements observed for the substrate.    and only a single Ir-hydride signal is present (−22 ppm), we suggest that the ethanol solvent molecules may become hyperpolarized predominantly through hydrogen-bonding to 3-methyladenine while bound at the metal-center ( Figure 5); direct binding may also contribute, though it is expected that SABRE-RELAY is unlikely to be a significant contributor because of the relatively low 1 H enhancements observed for the substrate.  (Figure 4e), respectively, which is substantially longer than the 1 H's of 3-methyladenine. For residual 1 H of C2D5OD, the contribution from intra-and intermolecular dipole-dipole interactions to T1 relaxation are minimized since the gyromagnetic ratio of 2 H is ~6.5 times smaller than 1 H. The shorter T1 for the residual -OH resonance likely reflects the greater participation in exchange.
Expanding the scope of 1 H SABRE to other DNA nucleobases, we also studied cytosine ( Figure 6a) and 6-O-methylguanine (Figure 6b). Small but clear 1 H SABRE enhancements were observed for 1 H resonances of both of these substrates. The weakness of the effects can partly be attributed to the need for 8% D2O to help solubilize the substrates, even at elevated temperatures; the use of aqueous media is challenging for SABRE due to the poor solubility of H2 in water and a reduced substrate exchange rate with the catalyst [62][63][64][65][66][67][68], thus often resulting in lower SABRE enhancements.
Analogous to 3-methyladenine, 6-O-methylguanine appears to show preferential catalyst binding of the imidazole ring, likely due to less steric hindrance. Hydride resonances showed generally weaker signals compared to those obtained with 3-methyladenine (Figure 4a), particularly with cytosine, where virtually no detectable hydride resonances were observed. The weaker hydride signals are generally consistent with lower SABRE enhancements. Moreover, whereas 3-methyladenine gave rise to a single hydride peak (indicating magnetic equivalence of the complex's two hydride sites on the NMR time scale), the multiple (weak) hydride signals observed with 6-O-methylguanine (comprising an asymmetric, antiphase doublet and a weak absorptive singlet) likely indicate the simultaneous presence of at least two different hydrogenated metal complexes (i.e., having different ligands) in significant concentration (see, for example, Refs. [20,22]). Note also the lack of SABRE enhancement of residual solvent resonances. Tautomerization has been suggested to result in a decrease in 1 H enhancements [69]. Indeed, tautomerization, along with elevated temperatures [70,71] and the weak binding of these substrates, likely con-  (Figure 4e), respectively, which is substantially longer than the 1 H's of 3-methyladenine. For residual 1 H of C 2 D 5 OD, the contribution from intra-and intermolecular dipole-dipole interactions to T 1 relaxation are minimized since the gyromagnetic ratio of 2 H is~6.5 times smaller than 1 H. The shorter T 1 for the residual -OH resonance likely reflects the greater participation in exchange.
Expanding the scope of 1 H SABRE to other DNA nucleobases, we also studied cytosine ( Figure 6a) and 6-O-methylguanine (Figure 6b). Small but clear 1 H SABRE enhancements were observed for 1 H resonances of both of these substrates. The weakness of the effects can partly be attributed to the need for 8% D 2 O to help solubilize the substrates, even at elevated temperatures; the use of aqueous media is challenging for SABRE due to the poor solubility of H 2 in water and a reduced substrate exchange rate with the catalyst [62][63][64][65][66][67][68], thus often resulting in lower SABRE enhancements.

15 N and 13 C SABRE-SHEATH of Nucleobases
While 1 H SABRE represents a rapid screening method for molecules of interest (and was used as such here), 1 H hyperpolarization can suffer rapid T1 decay due to strong interand intramolecular interactions. Non-quadrupolar heteronuclei (e.g., 15 N and 13 C) suffer weaker dipolar interactions and thus generally allow greater accumulation and retention  Analogous to 3-methyladenine, 6-O-methylguanine appears to show preferential catalyst binding of the imidazole ring, likely due to less steric hindrance. Hydride resonances showed generally weaker signals compared to those obtained with 3-methyladenine (Figure 4a), particularly with cytosine, where virtually no detectable hydride resonances were observed. The weaker hydride signals are generally consistent with lower SABRE enhancements. Moreover, whereas 3-methyladenine gave rise to a single hydride peak (indicating magnetic equivalence of the complex's two hydride sites on the NMR time scale), the multiple (weak) hydride signals observed with 6-Omethylguanine (comprising an asymmetric, antiphase doublet and a weak absorptive singlet) likely indicate the simultaneous presence of at least two different hydrogenated metal complexes (i.e., having different ligands) in significant concentration (see, for example, Refs. [20,22]). Note also the lack of SABRE enhancement of residual solvent resonances. Tautomerization has been suggested to result in a decrease in 1 H enhancements [69]. Indeed, tautomerization, along with elevated temperatures [70,71] and the weak binding of these substrates, likely contributes to the weakness of the 1 H SABRE effects observed for this set of molecules. Nonetheless, we have shown successful 1 H SABRE for three nucleobases, including the DNA base cytosine and the modified base 6-O-methylguanine for the first time.

15 N and 13 C SABRE-SHEATH of Nucleobases
While 1 H SABRE represents a rapid screening method for molecules of interest (and was used as such here), 1 H hyperpolarization can suffer rapid T 1 decay due to strong interand intramolecular interactions. Non-quadrupolar heteronuclei (e.g., 15 N and 13 C) suffer weaker dipolar interactions and thus generally allow greater accumulation and retention of hyperpolarization because of longer T 1 values. Moreover, accessing such nuclei allows for nuclei-specific investigations into biological processes. 15 N and 13 C SABRE-SHEATH of DNA nucleobases may thus prove to be more useful in evaluating tautomerization dynamics-in addition to potentially enabling future applications. Here we begin by showing natural-abundance 15 N (0.365%) SABRE-SHEATH of 3-methyladenine at 70 • C in 100% C 2 D 5 OD (Figure 7a). SABRE-SHEATH shows the enhancement of two adjacent 15 N NMR signals from two nitrogen atoms (sites/resonances "A" and "B") that integrate to ã 2:1 ratio for 15 N A compared to 15 N B , which when compared to the thermally polarized signal from 15 N 2 -imidazole, corresponds to enhancement values of ε~3300 (corresponding to 15 N polarization just above 1%) and ε~1900, respectively. The~230 ppm chemical shifts of these two resonances are consistent with the values expected for the N1, N3, and N7 positions of adenine, with the N9 and -NH 2 positions expected~160 ppm and~78 ppm, respectively [72][73][74]. However, no enhancements are observed for the NMR signals in those lower ranges. Moreover, given that no enhancement would be expected for the N3 site in 3-methyladenine (because of steric inhibition of catalyst binding), the enhanced sites are likely associated with the N1 (pyrimidine ring) and N7 (imidazole ring) positions. Moreover, to be consistent with the 1 H SABRE results in Figure 4a (where the larger 1 H enhancement was observed on the imidazole proton), we tentatively assign the moredownfield peak (N A , with the greater SABRE enhancement) in Figure 7a to N7, and the more-upfield peak (N B ) to N1. The absence of any signal attributable to the N9 site would be consistent with the prevalence of a tautomer wherein the N9 position is protonated (as it is commonly depicted).
This interpretation of the 15 N results would be consistent with the 1 H SABRE results for 3-methyladenine and C 2 D 5 OD (Figure 4), as well as the notion of preferential binding of the imidazole ring to the catalyst. Figure 7b shows a 15 N T 1 decay curve, wherein a somewhat rapid decay in signal (T 1 = 18.4 ± 1.8 s) is observed. These results represent the first natural abundance 15   This interpretation of the 15 N results would be consistent with the 1 H SABRE results for 3-methyladenine and C2D5OD (Figure 4), as well as the notion of preferential binding of the imidazole ring to the catalyst. Figure 7b shows a 15 N T1 decay curve, wherein a somewhat rapid decay in signal (T1 = 18.4 ± 1.8 s) is observed. These results represent the first natural abundance 15 N enhancement (spin concentration: ~150 µ M) of a nucleobase via SABRE-SHEATH.
We also investigated the 15 N SABRE-SHEATH of doubly-labelled cytosine (2-13 C; 1,3-15 N2). The enhancement of cytosine 15 N signals from a 40 mM solution (92% C2D5OD: 8% D2O) at 70 °C is shown in Figure 8a, following polarization transfer in a mixing field of ~1 µ T for 10 s; a short bubbling time was used to prevent a rapid decrease in temperature that could cause the substrate to crash out of the solution. In the figure, we see a modest selectivity in relative enhancement (1:0.75) for 15 NA (206 ppm) to 15 NB (141 ppm) compared to expectations based on a thermally polarized spectrum from a sample containing only labelled cytosine (i.e., without catalyst) with relative integrals of 1:1 (not shown); the absolute enhancements were ~240-fold and ~200-fold, respectively. A T1 measurement (performed as part of a different experimental run; discussed further below) saw an evengreater variance in the intensity of the two sites (Figure 8a inset). We also investigated the 15 N SABRE-SHEATH of doubly-labelled cytosine (2-13 C; 1,3-15 N 2 ). The enhancement of cytosine 15 N signals from a 40 mM solution (92% C 2 D 5 OD: 8% D 2 O) at 70 • C is shown in Figure 8a, following polarization transfer in a mixing field of 1 µT for 10 s; a short bubbling time was used to prevent a rapid decrease in temperature that could cause the substrate to crash out of the solution. In the figure, we see a modest selectivity in relative enhancement (1:0.75) for 15 N A (206 ppm) to 15 N B (141 ppm) compared to expectations based on a thermally polarized spectrum from a sample containing only labelled cytosine (i.e., without catalyst) with relative integrals of 1:1 (not shown); the absolute enhancements were~240-fold and~200-fold, respectively. A T 1 measurement (performed as part of a different experimental run; discussed further below) saw an evengreater variance in the intensity of the two sites (Figure 8a inset).
Such deviation may suggest the preferential binding of specific tautomers of cytosine ( Figure 8). For both 15 N atoms to become hyperpolarized, tautomerization of the labile 1 H about the ketone must take place to generate both sp 2 -hybridized 15 N atoms. Thus, tautomers (1), (2), and/or (4) (Figure 8) can potentially bind to the catalyst. We suggest that the larger ε of 15 N A over 15 N B is due to preferential binding of (1) and/or (4) to 15 N A because of steric hindrance from the adjacent ketone and amine to 15 N B , thus reducing interactions of 15 N B with the catalyst. This is further supported by our weak but non-zero 1 H SABRE for the adjacent 1 H (Figure 6). The signal at~206 ppm (N A ) shows what appears to be two overlapping doublets that are not well-resolved, each with a J-coupling of~8 Hz that likely arises from J CN for that 15 N coupled to the labeled 13 C site. Interestingly, the signal at~141 ppm (N B ) shows two unique triplets (split by a presumed J CN for that site of~14.5 Hz). Previous work by Shchepin et al. on the sensitivity of 15 N of imidazole to pH revealed a~30 ppm up-field shift in the 15 N resonance as pH is varied from 1 to 12, thus reporting on protonated and unprotonated 15 N [75]. The finer splitting arises from J NH ; the above J assignments are consistent with thermally polarized 1 H-decoupled 15 N experiments (not shown), where the decoupling collapsed the~3 Hz splittings but not the larger (unequal) splittings. The 15 N relaxation measurements performed at 9.4 T (Figure 8a inset) showed that the N A site had a similar high-field T 1 (20.6 ± 4.2 s) as that of the enhanced 15 N site in 3-methyladenine. However, the N B site's hyperpolarization decayed much more quickly (T 1 = 4.3 ± 0.5 s), suggesting the contribution of additional relaxation mechanisms for this spin (e.g., greater contributions from exchange) that could serve to exacerbate the selective 15 N SABRE enhancement. Such deviation may suggest the preferential binding of specific tautomers of cy ( Figure 8). For both 15 N atoms to become hyperpolarized, tautomerization of the lab about the ketone must take place to generate both sp 2 -hybridized 15 N atoms. Thus, mers (1), (2), and/or (4) (Figure 8) can potentially bind to the catalyst. We suggest th larger ε of 15 NA over 15 NB is due to preferential binding of (1) and/or (4) to 15 NA beca steric hindrance from the adjacent ketone and amine to 15 NB, thus reducing interacti 15 NB with the catalyst. This is further supported by our weak but non-zero 1 H SABR the adjacent 1 H (Figure 6). The signal at ~206 ppm (NA) shows what appears to b overlapping doublets that are not well-resolved, each with a J-coupling of ~8 Hz that arises from JCN for that 15 N coupled to the labeled 13 C site. Interestingly, the signal a ppm (NB) shows two unique triplets (split by a presumed JCN for that site of ~14.5 Finally, we present cytosine 13 C enhancements via SABRE-SHEATH. Figure 9a shows a 13 C SABRE-SHEATH spectrum from doubly-labeled cytosine obtained under conditions of 1 H decoupling, compared to a thermally polarized reference spectrum from the same sample. A close-up of these spectra is shown in Figure 9c. There, the spectra manifest a non-first-order doublet of doublets arising from the two J CN couplings to N A and N B . Although the four peaks exhibit equal intensities for the 13 C thermally polarized spectrum, the SABRE-SHEATH spectrum shows a highly asymmetric pattern of enhancement, which likely arises from unequal efficiencies of polarization transfer through the level anti-crossing regime achieved while the sample was in the magnetic shield. Taken together, the average enhancement (integrating over all four peaks) is~21-fold. In another series of experiments, 13 C SABRE-SHEATH spectra were obtained from doubly-labeled cytosine, but without 1 H decoupling (Figure 9b,d). For these experiments, symmetric multiplets were observed (Figure 9d) with average effective splittings of~6-8 Hz, reflecting both J CN and J CH contributions. The larger 13 C enhancements observed (up to nearly~50-fold) likely reflect the improved experimental efficiency of polarization transfer and are likely unrelated to the decoupling condition. A series of such spectra were obtained with a variable delay at 9.4 T prior to acquisition, allowing the high-field 13 C hyperpolarization lifetime to be measured for the 13 C cytosine resonance (T 1 = 22.9 ± 0.8 s).
hough the four peaks exhibit equal intensities for the 13 C thermally polarized spec the SABRE-SHEATH spectrum shows a highly asymmetric pattern of enhance which likely arises from unequal efficiencies of polarization transfer through the anti-crossing regime achieved while the sample was in the magnetic shield. Tak gether, the average enhancement (integrating over all four peaks) is ~21-fold. In an series of experiments, 13 C SABRE-SHEATH spectra were obtained from doubly-la cytosine, but without 1 H decoupling (Figure 9b,d). For these experiments, symmetri tiplets were observed (Figure 9d) with average effective splittings of ~6-8 Hz, refl both JCN and JCH contributions. The larger 13 C enhancements observed (up to nearl fold) likely reflect the improved experimental efficiency of polarization transfer an likely unrelated to the decoupling condition. A series of such spectra were obtained a variable delay at 9.4 T prior to acquisition, allowing the high-field 13 C hyperpolari lifetime to be measured for the 13 C cytosine resonance (T1 = 22.9 ± 0.8 s).  , top). Close-ups of the cytosin nances from these spectra, both taken under conditions of broadband  H-decoupling, are sh (c); enhancement (ε) ~21-fold. (b) The hyperpolarization decay curve for the integrated cytosi bonyl 13 C resonance, obtained from a second set of SABRE-SHEATH experiments (measured T; each point is a separate experiment with variable delay time at high field prior to acqui yielding T1 of 22.9 ± 0.8 s. Data points were taken from spectra obtained without 1 H decoupl example of which is shown in (d); ε~48-fold; that spectrum also has a very weak, broad peak ppm (not shown) that is tentatively assigned to cytosine bound to the catalyst. The hyperpolarization decay curve for the integrated cytosine carbonyl 13 C resonance, obtained from a second set of SABRE-SHEATH experiments (measured at 9.4 T; each point is a separate experiment with variable delay time at high field prior to acquisition), yielding T 1 of 22.9 ± 0.8 s. Data points were taken from spectra obtained without 1 H decoupling, an example of which is shown in (d); ε~48-fold; that spectrum also has a very weak, broad peak at 150 ppm (not shown) that is tentatively assigned to cytosine bound to the catalyst.
A summary of the measurements obtained in this work-SABRE enhancements and T 1 relaxation measurements-is contained below in Table 1. Taken together, the hyperpolarization lifetimes of the heteronuclei measured for these nucleobases is shorter than may be expected. However, when contemplating the rapid T 1 decays for hyperpolarized 15 N and 13 C signals of the present nucleobases, it should be noted that chemical shift anisotropy (CSA) is likely to be a significant contributor to T 1 relaxation of 15 N and 13 C in these systems. Because CSA relaxation is a function of the square of field strength, SABRE-hyperpolarized DNA nucleobases may indeed prove useful at lower field strengths (e.g., clinical MRI or benchtop NMR), wherein greater enhancements and longer T 1 time constants are likely.
It is also worth noting that the order-of-magnitude lower 15 N SABRE-SHEATH enhancements for (doubly-labeled) cytosine compared to (naturally abundant) 3-methyladenine are qualitatively consistent with the 1 H SABRE results for these molecules (Figures 4 and 6). Such weaker 15 N enhancements would also be much more difficult to observe (and hence optimize) with naturally abundant cytosine. Moreover, it may be difficult to achieve large 13 C SABRE-SHEATH enhancements in nucleobases without the 15 N labeling of the nitrogen sites, because abundant quadrupolar 14 N spins can greatly accelerate 13 C relaxation in the magnetic shield due to scalar relaxation of the second kind [76].

Conclusions
In this study, we have expanded the accessibility of SABRE and SABRE-SHEATH of DNA nucleobases by demonstrating NMR signal enhancement of: 1 H spins in 3methyladenine, cytosine, and 6-O-guanine; 15 N spins in 3-methyladenine and cytosine; and 13 C spins in cytosine. 1 H SABRE generally exhibited weak enhancements for each studied nucleobase. However, 1 H SABRE studies of 3-methyladenine revealed that simple chemical modification (compared to previous work on adenine) gave rise to preferential binding of the imidazole ring over the pyrimidine ring. Moreover, 1 H SABRE also showed hyperpolarization of the solvent molecules (residual protons of deuterated ethanol), ostensibly through hydrogen-bonding to 3-methyladenine, with T 1 values approaching~60 s; the proposed mechanism will be explored in more detail with future studies. Preferential binding of the imidazole ring is further supported by the natural-abundance 15 N SABRE-SHEATH enhancement of 3-methyladenine, upwards of ε~3300-the first natural-abundance result for a DNA nucleobase. Finally, 15 N and 13 C enhancements of cytosine (upwards of ε~240 and ε~50, respectively) via SABRE-SHEATH suggest a balance between tautomerization and catalyst binding kinetics. It is likely that significantly larger enhancements could be achieved by utilizing more-recently optimized hyperpolarizer platforms and conditions [22], greater p-H 2 enrichment (in some cases), and alternative pulse sequences designed to improve SABRE-SHEATH efficiency [77]. Thus, while DNA nucleobases may represent a family of challenging SABRE substrates, these results help pave the way for a variety of envisioned biological studies in the future, including efforts to further the fundamental understanding of the interplay of nucleobase tautomerization, base pairing, and disease. For example, hyperpolarized nucleobases may be useful in cellular or in vivo cell-signaling studies. It may also become possible to probe how certain genetic sequences may be more vulnerable to tautomerization (i.e., via mutagenic exposure) in cellular/cell-lysate studies (e.g., by hyperpolarizing one or more nucleotides within a short DNA sequence and exposing it to a tautomerizing agent). Moreover, mismatches in base-pairing could potentially be investigated by unzipping the strand at elevated temperatures, hyperpolarizing the nucleobases, and allowing the strand to anneal upon rapid cooling. Thus, polarization transfer to pairing nucleobases may allow the study of mutagenesis on a molecular level. Indeed, a clear understanding of interactions between SABRE-sensitive and insensitive nucleobases, followed by their various derivatives including nucleotides, may eventually culminate into the hyperpolarization of single-and double-stranded DNA.