[13C6,D8]2-deoxyglucose phosphorylation by hexokinase shows selectivity for the β-anomer

A non-radioactive 2-deoxyglucose (2DG) analog has been developed here for hyperpolarized magnetic resonance investigations. The analog, [13C6,D8]2DG, showed 13% polarization in solution (27,000-fold signal enhancement at the C1 site), following a dissolution-DNP hyperpolarization process. The phosphorylation of this analog by yeast hexokinase (yHK) was monitored in real-time with a temporal resolution of 1 s. We show that yHK selectively utilizes the β anomer of the 2DG analog, thus revealing a surprising anomeric specificity of this reaction. Such anomeric selectivity was not observed for the reaction of yHK or bacterial glucokinase with a hyperpolarized glucose analog. yHK is highly similar to the human HK-2, which is overexpressed in malignancy. Thus, the current finding may shed a new light on a fundamental enzyme activity which is utilized in the most widespread molecular imaging technology for cancer detection – positron-emission tomography with 18F-2DG.

The hexokinase (HK) family of enzymes carry out the first step in glycolysis and therefore are of key importance in cellular metabolism 1 . HKs play an important role in malignancy: one of the earliest observations in malignant transformation is an increased glycolytic flux 2 , which is largely attributed to changes in the activity of HKs. For example, in rapidly growing tumor cells, HK expression is markedly elevated and most of the enzyme is localized to the mitochondrial membrane 3,4 , which provides HK with increased ATP availability resulting in increased activity 5 .
HK isoenzymes vary in function and in organ expression. In health, HK-4 (glucokinase, GK) is important in glucose sensing in the pancreas and in the liver 6 while HK-1 and HK-2 mainly function in other tissues such as muscle and brain. Increased or decreased HK expression has been found to correlate with the clinical outcome in a number of disease states [7][8][9][10][11][12][13][14][15][16][17][18][19][20] . Elevated HK-2 expression is correlated with higher histological grade in hepatocellular carcinoma (HCC) 7 as during the process of HCC tumorigenesis, normal GK expression is silenced, and HK-1 and HK-2 are overexpressed 8,9 . Since the affinity of HK-1 and HK-2 for glucose is higher than that of GK, their expression results in increased glucose utilization which supports HCC tumorigenesis. HK-2 expression was shown to be correlated with the survival of patients with HCC, gastric and colorectal cancer, and WHO grade IV glioblastoma [10][11][12][13]20 . Aggressive types of cancer including pancreatic ductal adenocarcinoma 14 , renal cell carcinoma 15 , medulloblastoma 16 , and mouse models of lung and breast cancer 17 are characterized by high levels of HK-2. Other non-malignant diseases with increased HK-2 expression in the liver include fatty liver disease 18 , and hepatitis C virus infection 19 .
[ 18 F]Fluoro-2-deoxy-D-glucose (FDG) is the most widely used tracer in positron emission tomography (PET) for detecting, staging, and monitoring of various malignancies 21 . It was previously suggested that the increased HK-2 activity forms the basis for the utility of FDG-PET imaging of malignant tumors 5 . Nuclear magnetic resonance (NMR) has been used for the study of glucose and 2-deoxyglucose (2DG) metabolism [22][23][24][25] . In 13 C-NMR, as opposed to PET, the chemical evolution of 2DG to 2DG-6-phosphate (2DG6P) can be discerned. This property could be useful for differentiating the effects of glucose transporters expression from those of HK expression in vivo. For the above reasons, glucose and 2DG imaging has been a desired target in magnetic resonance imaging, which does not involve ionizing radiation. MR investigations of glucose and 2DG have been carried out using the gluco-CEST approach 22,26,27 , thermal equilibrium 13 C-NMR with carbon-13 labeled substrates 28 , and deuterium metabolic imaging (DMI) 29 .
The phosphorylation of [ 13 c 6 ,D 8 ]2DG is detectable in real-time by hyperpolarized NMR in the reaction with yHK. Next, we investigated the ability to follow the phosphorylation of [ 13 C 6 ,D 8 ]2DG with yHK and with bGK. To this end, hyperpolarized [ 13 C 6 ,D 8 ]2DG was directly injected to an NMR tube containing the enzymatic reaction buffer solution (Methods) and the yHK or bGK enzyme. In the reactions with yHK, hyperpolarized [ 13 C 6 ,D 8 ]2DG6P was observed immediately -already in the first spectrum acquired after the arrival of the hyperpolarized medium into the NMR tube (within 1 s), as can be seen in Fig. 2A (Fig. 2D, middle panel). The low intensity multiplet observed at a chemical shift close to that of the C 6 position of [ 13 C 6 ,D 8 ]2DG6P is likely due to the heptamer form of this compound, as previously reported for glucose 56,57 . These signals also appear in the experiment carried out without an enzyme (Fig. 2F, middle panel). As the lack of reaction of bGK on 2DG observed here was in agreement with prior publications 54,55 , this experiment was carried out only twice.
Interestingly, in the reaction with yHK, the signals of the β anomer appeared to decay faster compared to the α anomer, with an apparent T 1 that is about 31% shorter for the C 1 position and 29% for the C 2 position (Table 2). No significant differences were observed in the experiments carried out without an enzyme (Supplementary Information S1) or with bGK (  8 ]2DG and apparent also for its C 2 position (Supplementary Information S1). The C 6 signal in these experiments was useful for characterizing the reaction rate constants as the substrate and the product signals were well resolved in this site. Therefore, the time course of the C 6 signals of the substrate and product were fitted to a kinetic model which considers the metabolic reaction as a first-order reaction 39 . The metabolic rate constants that were obtained are provided in Supplementary Table S2.
The reaction of [ 13 c 6 ,D 8 ]2DG with yHK reveals a decrease in the β to α anomeric ratio. Figure 4 shows the change in the ratio of anomers over the course of the reactions of [ 13 C 6 ,D 8 ]2DG and [ 13 C 6 ,D 7 ]glucose with yHK and bGK. In addition to the fast decay of C 1 β of [ 13 C 6 ,D 8 ]2DG, this experiment also revealed different reaction kinetics for the two anomers present in the reaction medium. In Fig. 4A, it can be seen that the β/α ratio of the C 1 position is decreasing during the first 20 s and then increasing. We wished to determine if this behavior is unique to [ 13 C 6 ,D 8 ]2DG by inspecting the same parameter in the experiments with [ 13 C 6 ,D 7 ]glucose. As shown in Fig. 4B, the reaction of [ 13 C 6 ,D 7 ]glucose with yHK did not show this behavior, rather -a slightly increasing β/α C 1 ratio can be seen during the same reaction time.
To further investigate this behavior, we analyzed the reactions of both substrates with bGK in the same way. Figure 4C  anomers in terms of substrate and product during the course of the reaction. Due to the significant overlap between the signals of [ 13 C 6 ,D 8 ]2DG and [ 13 C 6 ,D 8 ]2DG6P in each of the anomer signals, a spectral deconvolution analysis was applied. The C 1 signals were the only signals where this information could be obtained because: 1) the C 3 -C 5 signals of [ 13 C 6 ,D 8 ]2DG and [ 13 C 6 ,D 8 ]2DG6P overlapped, 2) in the C 3 -C 6 signals of both compounds, anomer differentiation was not possible, and 3) as regards to the C 2 site signals, although this site showed resolved anomer signals, these signals were broader than the signals of the other sites as a result of more extensive splitting due to coupling with two attached deuterons. Consequently, deconvolution of the anomeric signals to their [ 13 C 6 ,D 8 ]2DG and [ 13 C 6 ,D 8 ]2DG6P components was not possible for this site (Supplementary Information S4). An example of the deconvolution operation on the C 1 signals for one spectrum is shown in Fig. 5.
The ability to deconvolve the complex signals to their respective [ 13 C 6 ,D 8 ]2DG6P and [ 13 C 6 ,D 8 ]2DG components allowed monitoring of the temporal evolution of each of these components in the course of the reaction. The deconvolution procedure relied on the ability to resolve (at least partially) a few unique signals for each of the components (assignment shown in Supplementary Information S5). A typical example of such unique signals is shown in Fig. 6. It can be observed that the [ 13 C 6 ,D 8 ]2DG6P component is increasing with time in both the C 1 α The overall proportion of the [ 13 C 6 ,D 8 ]2DG6P component and the [ 13 C 6 ,D 8 ]2DG component, as derived from this deconvolution procedure applied on the C 1 signals, was in agreement with an analysis of these components in the C 6 signal, in which the signals of the substrate and the product are resolved, but the anomers are not (Fig. 6B). This result served as a validation for the deconvolution results. The normalized root mean square deviation (NRMSD) for the ratio of product signal to total signal between the C 1 and C 6 based determinations was found to be 16 ± 9%, n = 5).
Anomeric selectivity of yHK for the β anomer of [ 13 c 6 ,D 8 ]2DG. During the reaction with yHK, the β/α ratio of [ 13 C 6 ,D 8 ]2DG (the substrate) appeared decreasing with time, and within no more than 20 s this ratio reached zero as no more β signal was observed (Fig. 6C, plus signs). This finding suggested a selective consumption of the β anomer. On the other hand, the β/α ratio of the product (Fig. 6C, diamonds) appeared increasing throughout the time that the β anomer of the substrate was present, then slightly decreased towards a steady level. Because the anomerization rate of G6P is known to be about 500-fold faster than that of glucose and assuming that the same is true for DG and DG6P, it appears reasonable that this result reflects the fast anomerization rate of the product, i.e. [ 13 C 6 ,D 8 ]2DG6P. Since the mutarotation rate of 2-deoxy-D-glucose is more than 10-fold higher than that of D-glucose 58 , these results and the difference from the reaction with [ 13 C 6 ,D 7 ]glucose, cannot be explained by differences in the mutarotation rate of the substrate. Therefore, the results suggest that [ 13 C 6 ,D 8 ]2DG6P is formed in the β anomeric form and then quickly reaches anomeric equilibrium. Thus, in the first part of the reaction, before the consumption of the β anomer of the substrate is completed, we observe two competing processes as regards to the β anomer of the product: on one hand -the production of the β anomer leads to an increase in its signal, and on the other, the continuous rapid establishment of anomeric equilibrium leads to a decrease in this signal. Once the consumption of the β anomer of the substrate is completed, we observe a decrease in the content of the β anomer of the product and arrival to an equilibrium represented by the steady level of this ratio (β/α = 0.92 ± 0.10, n = 5).

Discussion
2DG in its radioactive form -FDG, is the most abundant tracer in clinical PET examinations. Here, we were interested in researching the ability to monitor the activity of an analog of this tracer in a hyperpolarized state by magnetic resonance. To this end, we have used a 2DG analog uniformly labeled with 13 C and deuterium as previously done for observing glucose in a hyperpolarized state 38,46-50 . We show here the 13 C NMR signals of this 2DG analog ([ 13 C 6 D 8 ]2DG) in a hyperpolarized state with 13% polarization level for the C 1 site at about 10-11 mM concentration.
After validating that indeed this analog can be detected on single 13 C NMR acquisitions at this concentration, we utilized it for direct and real-time monitoring of the phosphorylation reactions. In the reaction with yHK, [ 13 C 6 ,D 8 ]2DG was phosphorylated rapidly, but the signal of C 1 β appeared to decay faster compared to the C 1 α signal. This effect was also manifested as a decrease in the β to α ratio of the overlapping C 1 signal of [ 13 C 6 ,D 8 ]2DG and [ 13 C 6 ,D 8 ]2DG6P during the reaction time. To investigate this effect further, we deconvolved the signal and found that the β/α ratio of the substrate, in particular, decreased until the β anomer was completely consumed. This result suggested that yHK has an anomeric selectivity for the β anomer of [ 13 C 6 ,D 8 ]2DG. The presence of the α anomer of the product is likely explained by a fast anomerization of the newly formed [ 13 C 6 ,D 8 ]2DG6P. This assumption is supported by prior investigations that found that the anomerization rate of glucose-6-phosphate is about 500-fold faster than that of glucose 59 . To the best of our knowledge, the anomerization rate of 2DG6P has not been reported. We note that the lack of anomeric selectivity and the slightly increasing C 1 β/α ratio seen in the reaction of yHK with hyperpolarized [ 13 C 6 ,D 7 ]glucose are in agreement with the slight anomeric preference of this reaction towards the β anomer observed by Okuda et al. in 1984 for hexokinase A in yeast 60 , while Miclet et al. 38 reported comparable kinetics for both anomers of [ 13 C 6 ,D 7 ]glucose.

Compound & Enzyme
Apparent T 1 of C 1 β (s) Apparent T 1 of C 1 α (s) T 1 of C 6 P (s) T 1 of C 6 (s) Apparent T 1 of C 2 β (s) Apparent T 1 of C 2 α (s)  Values are presented as mean ± standard deviation. # Differences in the apparent T 1 between the β and α anomers for the C 1 and C 2 positions were 31% (p = 0.00004) and 29% (p = 0.0002), respectively. p-values were calculated using a paired t-test. *Only sites that could be spectrally resolved were analyzed. The apparent T 1 calculation (Methods) for the C 1 and C 2 sites was done without taking into consideration the reaction kinetics as the substrate and product signals were not resolved for this site. For the C 6 position, a full kinetic analysis was performed and the reaction kinetic rate constant are provided in Supplementary  www.nature.com/scientificreports www.nature.com/scientificreports/ The yHK used in this study, from the yeast Saccharomyces cerevisiae, previously showed similarity to mammalian HKs in terms of nucleotide and amino acid sequence 51 , with 35% amino acid homology compared to the C-terminus end of mammalian hexokinases 53 . Specifically, it was shown that the substrate-binding regions for glucose are conserved between the yHK and the HK found in rapidly growing tumor cells 52 . When comparing the amino acid sequence by global pairwise alignment, we found 29.4% identity and 48.1% similarity between yHK and human HK-2 ( Supplementary Information S3). Therefore, this work may represent a preliminary step for studying HK properties of cancerous tissues.
If the anomeric selectivity revealed here for yHK is also present in human HK, the findings of the current study may shed a new light on this important reaction. Various glucose transporters have varying anomeric preferences. For example, intestinal cells 61 as well as Ehrlich ascites tumors cells 62 showed preference for the β anomer. In human erythrocytes, initial reports indicated that cross-membrane glucose transport has a preference for the β anomer 63 , but later studies showed no difference between the two anomers 64,65 . In the pancreas, β cells show www.nature.com/scientificreports www.nature.com/scientificreports/ preferential uptake of β-D-glucose as compared to α-D-glucose and increased insulin release in the presence of α-D-glucose as compared to β-D-glucose 66 . However, this anomeric specificity is perturbed in animal models of diabetes 67,68 . It was also previously shown that the anomeric preference of glucose metabolic enzymes that is observed in normal mammalian cells is absent in tumor cells 69 . If these perturbations are also manifested using [ 13 C 6 ,D 8 ]2DG, it is possible that by imaging the anomeric preference of tissues, cancerous lesions can be detected by such loss of anomeric preference and treatments for diabetes may be monitored as well.
So far, due to the relatively fast anomerization times in physiologic temperature, the investigation of anomeric preference in live tissues was limited. Owing to the high temporal resolution enabled by the dDNP technology, and its translational nature, it may now be possible to re-visit the anomeric preference of enzymatic processes in live tissues and eventually in the whole animal, with the long-term goal of utilizing these differences for differential diagnoses and treatment monitoring.
It is interesting to suggest implications of the anomeric selectivity found here for FDG-PET imaging, which is the most widely used molecular imaging technology for cancer detection and staging. Interestingly, FDG is commonly synthesized from a starting material that is in the β anomeric form 70 . However, it is likely that anomeric equilibrium is established during the synthesis (in the acidification stages 70 ) and later on during the time allowed for washout in PET imaging (from injection to imaging, ~1 h). Therefore, it is not clear that standard PET findings should be interpreted any different due to the current findings. However, it may provide useful information for quantitative first-pass PET imaging. In this type of imaging, which is mostly used for research purposes and where the relevant time scales are much faster -on the order of tens of seconds, it is very likely that HK activity is predominantly recorded and it could be useful to consider that the relevant substrate concentration may be only that of the β anomer. www.nature.com/scientificreports www.nature.com/scientificreports/ Figure 5. Signal deconvolution of the C 1 β and C 1 α signals to their respective [ 13 C 6 ,D 8 ]2DG and [ 13 C 6 ,D 8 ]2DG 6 P components in one spectrum from the reaction of [ 13 C 6 ,D 8 ]2DG with yHK. This specific spectrum was acquired 14 s after the beginning of the reaction with yHK. The spectrum is taken from the same experiment that is shown in Fig. 2B,A. (A) The fitted signal components of C 1 β and C 1 α of [ 13 C 6 ,D 8 ]2DG. The complex signal of each anomer consists of 6 Lorentzian lines due to D-13 C coupling (J 13C-D = 25.5 Hz) and 13 C- 13 C coupling (J 13C-13C = 39.7 Hz). The J-coupling constants reported here are in agreement with those previously reported for [ 13 C 6 ,D 7 ]glucose 50 and for [1-13 C]glucose 77 . The 13 C-13 C doublet component of each signal was allowed to be asymmetric as asymmetry was previously observed in 13 C-13 C split signals in a hyperpolarized state (and occasionally at thermal equilibrium) 38 www.nature.com/scientificreports www.nature.com/scientificreports/ Since its introduction, the only tracer used clinically with hyperpolarized MR was [1-13 C]pyruvate [32][33][34] . We believe that the current study may set the foundations for further research regarding imaging possibilities with [ 13 C 6 ,D 8 ]2DG. [ 13 C 6 ,D 8 ]2DG shows beneficial T 1 profile in the physiological temperature range compared to room temperature with a relative improvement of about 1.6-fold in T 1 (C 1 carbon, Supplementary Information S1). Because the short T 1 relaxation time limits the clinical utility of hyperpolarized 13 C labeled tracers, this is an important advantage. The current study may also imply that the T 1 of [ 13 C 6 ,D 8 ]2DG may be further prolonged at higher temperatures and this could enable better preservation of the polarization prior to injection of the hyperpolarized agent to the subject. This aspect will be further investigated in the future.
In the following we will review the pros and cons of a hypothetical MRI scanning with hyperpolarized [ 13 C 6 ,D 8 ]2DG compared to other imaging techniques using a 2DG or glucose tracers.
An important advantage for a dDNP-MRI method over FDG-PET imaging is the ability to detect enzymatic conversion of the imaging probes, i.e. that the substrate and product signals can be resolved, and that this detection occurs within seconds. 2DG does not go through further metabolic steps in the glycolysis pathway after its phosphorylation, and is mostly found in its phosphorylated form 71 , 2DG6P, which is intracellular. PET imaging relies on this principle and therefore employs a delay between contrast injection to imaging to allow extracellular FDG to wash out prior to image acquisition. However, with hyperpolarized 2DG there may be no need to wait for the washout of the injected substrate, as the detection of phosphorylated product is possible and occurs in real-time. This short time scale may allow better quantification of enzymatic rates which may provide a new window for tissue characterization and differentiation from normal background tissue activity. These factors may allow imaging of tissues for which the background uptake of FDG is high and limits diagnostic imaging such as the brain and the heart.
Another diagnosis that may benefit from shortening the time from tracer administration to imaging is the characterization of hepatocellular carcinoma (HCC) lesions. The intensity of HCC relative to the normal liver (as in other tissues) in FDG-PET depends on the ratio between the phosphorylation of FDG and the dephosphorylation of FDG-6P by glucose-6-phosphatase 72 . Because glucose-6-phosphatase expression is variable in HCC, FDG-PET imaging only has around 50% sensitivity in HCC detection 73 . It appears possible that shortening the time between the injection of tracer and image acquisition will open new imaging capabilities for HCC as it may enable monitoring of the phosphorylation reaction before the phosphatase reaction takes place.
The major advantage of hyperpolarized MRI as compared to PET imaging is the use of non-ionizing radiation. However, the con of hyperpolarized MRI in this regard is the use of non-trace amounts of the imaging probe. Hyperpolarized [ 13 C 6 ,D 7 ]glucose was visible in 0.12 g/kg in rats 47 , and it is likely that [ 13 C 6 ,D 8 ]2DG will be visible in a similar dose. It is important to note that 2DG was found to be non-toxic in doses up to 0.5 g/kg when administered intravenously to rats 74 . Therefore, it appears likely that [ 13 C 6 ,D 8 ]2DG can be observed in a hyperpolarized state in vivo without toxic effects.
Competing MR technologies are gluco-CEST 22,26,27 and deuterium metabolic imaging (DMI) 29 . The advantage of the gluco-CEST technology is the use of the 1 H channel that is available on all MRI scanners and the ability to use non-labeled imaging probes (2DG or glucose) 22,26,27 . However, the SNR on gluco-CEST imaging is low. The overall signal change from baseline on CEST imaging was not more than 2-fold 22 , which is far less than that of hyperpolarized MRI as in the latter the background signal is very low. In comparison, the SNR of hyperpolarized [ 13 C 6 ,D 7 ]glucose MRI was 100,000 47 . The advantages of the DMI technology include the ability to use oral administration of the imaging probe and resolve metabolism to lactate, and an improved SNR due to the low background signal 29 . However, the acquisition time of the images was long (ca. 30 min) 29 . The con of both gluco-CEST and DMI is the lack of ability to discern the phosphorylated product from the substrate. However, a recent CEST study did show that the signal change observed was higher on DG injection than on glucose injection, suggesting accumulation of DG6P in the brain 22 . Considering these pros and cons of each 2DG/glucose based imaging technology and the central role these imaging probes play in biology and pathology, it is possible that a synergistic combination between them can yield better results than each technique alone.
In summary, we show the potential of [ 13 C 6 ,D 8 ]2DG to serve as an imaging probe for hyperpolarized MRI and find that HK has a unique selectivity for the β anomer of this potential molecular imaging probe, which may provide a new contrast mechanism on MR. coupling (J 13C-D = 25.1 Hz) and due to 13 C-13 C coupling (J 13C-13C = 38.6 Hz). The 13 C-13 C doublet component of each signal was allowed to be asymmetric here as well. Both fit components are shown in solid cyan line, the sum of both components is shown in the solid blue line. (C) The combination of the 2DG and the 2DG6P fitted signals components. The sum of 2DG fit components is shown in the dashed dark green line, the sum of 2DG6P fit components is shown in the solid blue line, and the sum of both 2DG and 2DG6P fit components is shown in the black dotted line. (D) A comparison between the overall fitted signals (black dotted line) shown in (C) to the experimental spectrum (solid black line) demonstrates the utility of the deconvolution procedure. The normalized root mean square deviation (NRMSD) between the fit and the experimentally measured spectrum was found to be 9% and 17% for the C 1 β and C 1 α anomers, respectively. 2DG, [ 13  NMR spectroscopy. 13 C NMR spectroscopy was performed in a 5.8 T non-shielded high resolution NMR spectrometer (RS2D, Mundolshein, France) located about 2.2 meters away from the spin-polarization magnet Enhancement factor and polarization percent calculation. The enhancement factor was calculated by comparing the maximal integrated signal intensity obtained under hyperpolarized conditions to the integrated signal intensity of the same sample at thermal equilibrium, acquired with the same nutation angle under fully relaxed conditions. In order to compare the integrated intensities from the thermal and hyperpolarized acquisitions, the same acquisition (spectral width, number of points, receiver gain), and processing (exponential line broadening, zero-filling) parameters were used. The thermal equilibrium spectrum was corrected for number of scans. Percent polarization in solution was calculated by multiplying the enhancement factor by the theoretical percent polarization of 13 C at thermal equilibrium in the same magnetic field (5.8 T).

Methods
Experimental design and enzymatic assays. In experiments using yHK, ATP disodium was dissolved in 1 ml of TRIS HCl medium with MgCl 2 . The pH was then corrected with 10% NaOH solution to 7.6 75 and ~130 units of hexokinase were added and gently mixed until a clear solution was obtained. This solution was placed in an NMR tube which was placed within the spectrometer's probe. In experiments with bGK, ATP disodium was dissolved in 2 ml of TRIS HCl medium with MgCl 2 , the pH was corrected and the solution was added into 100 units of bGK. Half of this solution (50 units) was used per experiment. The final reaction concentrations, after combining these enzyme solutions with the dissolution medium, were 20.0 ± 0.5 mM ATP, 10 mM MgCl 2 , and 10.9 ± 0.8 mM of hyperpolarized substrate. The hyperpolarized solution containing [ 13 C 6 ,D 7 ]glucose or [ 13 C 6 ,D 8 ]2DG was injected via a Teflon line from the spin-polarizer directly into the NMR tube containing the enzymatic mixture (yHK or bGK) within 3 s of Helium (g) chase. For the experiments shown in Fig. 4D,F, mixing was confirmed by visual inspection for uniformly distributed greenish color indicating adequate mixing of the hyperpolarized substrate before starting data acquisition. For the other experiments, the dissolution of the hyperpolarized substrate was done while the NMR tube was already inside the probe of the NMR spectrometer. The yHK reactions were performed at room temperature (ca. 21 °C). The experiments with bGK were performed at 40 °C, achieved by heating both the NMR tube and the dissolution medium transfer line. The temperature of the reactions with bGK was confirmed in real-time by a temperature probe inside the NMR tube (Osensa, Burnaby, BC, Canada).
Spectral processing and data analysis. Spectral processing was performed using MNova (Mestrelab Research, Santiago de Compostela, Spain). Integrated intensities were calculated either with MNova or with DMFIT 76 . Signal deconvolution analysis was carried out with DMFIT 76 . t 1 calculation. T 1 determination in hyperpolarized decays measured without enzymatic reactions, and apparent T 1 determination of hyperpolarized decays measured during enzymatic reactions -not taking into account reaction kinetics, were performed by curve fitting of the signal decay to the following equation in which TR, the repetition time, and θ, the nutation angle of excitation, are known. Curve fitting analysis was performed using Matlab (Mathworks, Natick, MA, USA). In experiments in which enzymatic conversion occurred and for sites in which this reaction could be discerned (the C 6 position), the T 1 s were determined using a previously published model 39 , which enables the simultaneous determination of both the reaction rate constants and the T 1 s of the substrate and product.

Data availability
The data that support the findings of this study are available from the corresponding author on request.