Non-invasive In-cell Determination of Free Cytosolic [NAD+]/[NADH] Ratios Using Hyperpolarized Glucose Show Large Variations in Metabolic Phenotypes*

Background: Free cytosolic [NAD+]/[NADH] ratio maintains cellular redox homeostasis and is a cellular metabolic readout. Results: Pyruvate/lactate ratios show distinct metabolic phenotypes and are used to derive free cytosolic [NAD+]/[NADH] ratios. Conclusion: Determination of free cytosolic [NAD+]/[NADH] ratios using hyperpolarized glucose is applicable to a wide selection of cell types. Significance: This metabolic phenotyping may be a crucial tool to understand pathologies, and to diagnose and measure effects of therapies. Accumulating evidence suggest that the pyridine nucleotide NAD has far wider biological functions than its classical role in energy metabolism. NAD is used by hundreds of enzymes that catalyze substrate oxidation and, as such, it plays a key role in various biological processes such as aging, cell death, and oxidative stress. It has been suggested that changes in the ratio of free cytosolic [NAD+]/[NADH] reflects metabolic alterations leading to, or correlating with, pathological states. We have designed an isotopically labeled metabolic bioprobe of free cytosolic [NAD+]/[NADH] by combining a magnetic enhancement technique (hyperpolarization) with cellular glycolytic activity. The bioprobe reports free cytosolic [NAD+]/[NADH] ratios based on dynamically measured in-cell [pyruvate]/[lactate] ratios. We demonstrate its utility in breast and prostate cancer cells. The free cytosolic [NAD+]/[NADH] ratio determined in prostate cancer cells was 4 times higher than in breast cancer cells. This higher ratio reflects a distinct metabolic phenotype of prostate cancer cells consistent with previously reported alterations in the energy metabolism of these cells. As a reporter on free cytosolic [NAD+]/[NADH] ratio, the bioprobe will enable better understanding of the origin of diverse pathological states of the cell as well as monitor cellular consequences of diseases and/or treatments.

Dysregulated glycolytic metabolic networks are found in the pathophysiology of diseases such as cancer, inflammation, neurodegenerative diseases, atherosclerosis, obesity, and diabetes (1)(2)(3). The fundamental changes in cell energy metabolism offer large potential for intervention and thereby control of entire biochemical networks by identifying and targeting key control points (4). One such control point is NAD (5-7), which plays a pivotal role in cellular metabolism, acting as a coenzyme in numerous central housekeeping redox reactions. The concentration ratio between the oxidized and the reduced forms ([NAD ϩ ]/ [NADH]) is critical to the cell and significant shifts from normal signify profound alterations in metabolism that are often linked to pathological states (8).
NAD is either protein bound or free and separate pools are present in the cytosol and in the mitochondria of mammalian cells. Several direct detection methods are commercially available for measurement of the total cellular NAD ϩ and NADH concentrations. Measurements of whole cell total concentrations of NAD ϩ and NADH do not, however, distinguish between pools in different compartments or between free and protein bound forms and thus can only provide information about the ratio of the NAD ϩ -NADH couple overall (9). Most NADH present in the mitochondria is protein-bound (10) and will not contribute to the cellular redox potential. Whereas the total cellular [NAD ϩ ]/[NADH] ratio is typically on the order of 2-4 (11,12), the free cytosolic [NAD ϩ ]/[NADH] ratio is usually several orders of magnitude higher (13,14). Thus, the importance of the ratio of free cytosolic [NAD ϩ ]/[NADH] to the cellular redox potential has lead to a focus on methods capable of estimating this ratio.
To assess the free cytosolic NAD redox potential, two main strategies have been reported in the literature. One strategy takes advantage of the proportionality between measured product/substrate ratios of suitable near-equilibrium redox reactions and the concentrations of the involved coenzymes (14). Many studies have successfully estimated the [NAD ϩ ]/ [NADH] ratio using this indirect metabolic method (15)(16)(17); however, the use of chemical cellular extraction makes this approach incompatible with dynamic studies in intact cells. Another less invasive strategy directly estimates the [NAD ϩ ]/ [NADH] ratio with optical imaging (18). In particular, two recent reports use circularly permuted fluorescent protein to create targeted biosensors for binding of NAD ϩ and NADH (19,20). The biosensor for binding of cytosolic NAD ϩ and NADH is made as a genetically encoded construct in intact cells, which allow optical estimation of [NAD ϩ ]/[NADH] ratios in single intact cells. Inherent to this method, however, it cannot be applied in animal studies.
Ten years ago, a technique was reported which dramatically enhances the signal that can be achieved in NMR experiments. This technique allows metabolites that have been hyperpolarized with dynamic nuclear polarization (21) to retain their hyperpolarization in a short time-window when dissolved (22). This technique opens the way for numerous in vivo and in-cell applications using non-invasive NMR measurements (23). Hyperpolarized 13 C NMR provides detailed chemical information on molecular transformations with sufficient time resolution for the detection of transient metabolic intermediates. The NMR sensitivity enhancement provided by hyperpolarization fades with the spin lattice relaxation rate (1/T 1 ), and the dynamic experiments are best conducted on a time scale of Ͻ3 ϫ T 1 . Recently, we have developed hyperpolarized markers for visualization of metabolic intermediates in the entire glycolysis (24). Here, we report a further development of hyperpolarized D-[1,2,3,4,5,6,6-13 C 6 ]glucose-d 7 (fully 13 C-enriched and deuterated glucose, 2 renamed 13 C 6 -d 7 ) into a cellular bio-probe for estimation of the free cytosolic ratio of [NAD ϩ ]/[NADH] using the dynamic, non-invasive tool, hyperpolarized nuclear magnetic resonance. Hyperpolarized glucose is perifused to living cancer cells in suspension where it is rapidly metabolized to the glycolytic products pyruvate and lactate. NMR signals from the hyperpolarized [1-13 C]pyruvate and [1-13 C]lactate products are followed over 1 min with a time resolution of 2 s and are used as input data for a kinetic model of the glycolysis. The model applied allows calculation of rate constants and ratios between concentrations of metabolites. It has previously been demonstrated that conversion between pyruvate and lactate according to the reaction, pyruvate ϩ NADH ϩ H ϩ 7 lactate ϩ NAD ϩ , is very fast in vivo and is very close to chemical equilibrium (14). Thus, it is in principle possible to use the ratio between pyruvate and lactate, catalyzed in vivo by lactate dehydrogenase, to calculate the free cytosolic [NAD ϩ ]/[NADH] ratio (25)(26)(27)(28) [NADH] ratio of the prostate cells was significantly increased relative to that of the breast cancer cells. This observation is likely to be correlated with a preference for fatty acids as cellular energy source as well as an increased de novo fatty acid synthesis of these cells.
Cell Culture and Harvest-Both cell types were cultured in RPMI 1640, 10% FBS, 100 units/ml penicillin, 100 g/ml streptomycin at 37°C in a 5% CO 2 atmosphere. Glucose consumption and lactate formation was measured with standard biochemical methods on both cell types as a function of time in the chosen culture medium. This analysis prevented nutrient depletion under the chosen growth conditions. MCF7 cells were plated in 75-cm 2 flasks at 6ϫ 10 6 /flask and PC3 cells in 175-cm 2 flasks at 5 ϫ 10 6 /flask and were cultured for 48 h. Cells were harvested by trypsination, resuspended in PBS buffer, and assessed for viability by trypan blue exclusion (viability in all experiments Ͼ98%). Cells were pelleted and resuspended in an appropriate volume of dissolution buffer (40 mM phosphate buffer, pH 7.3) to obtain a cell concentration of 40 million cells/ml. Preparation, Polarization, and Dissolution of Hyperpolarized 13 C NMR-Glucose (22.8 mg, 0.118 mmol) was dissolved in polarization medium (25.0 mg) prepared from Ox063 radical (19.1 mg, 13.3 mol) and gadoteridol (40 mg of 50 mol/g solution in water) in water (465 mg). The concentration of radical and gadolinium complex in the preparation was 17 and 2 mM, respectively. A sample of the preparation containing 20 mol of glucose was hyperpolarized to equilibrium at 3.3 Tesla and ϳ1.2 K with a build-up time constant of 1300 s in a prototype polarizer (22). The sample was dissolved in 5 ml of 40 mM phosphate buffer, pH 7.3. 20 million MCF7 or PC3 cells suspended in 500 l of 40 mM phosphate buffer, pH 7.3 were placed in a flat-bottomed 10-mm NMR tube adjusted in the NMR spinner to cover the active volume (1.5 ml). The NMR tube, with a connected inlet tubing, was placed in a 14.1 Tesla magnet at 310 K. 1 ml of the dissolved hyperpolarized glucose was perifused to the cells via the tubing resulting in a glucose concentration of 2 mM in the cell suspension. A series of 20 degree pulses every 2 s (30 scans in total) was acquired. The acquisition was started simultaneously with injection of the hyperpolarized glucose. Generally, liquid state 13 C spin polarizations of 19 Ϯ 2% were obtained for an average of the 13 C-labeled positions at the time of substrate infusion. The liquid state T 1 at ϳ37°C and 14.1 Tesla was 14 Ϯ 1 s for all of the signals originating from the uniformly isotope-labeled glucose molecule. The relatively short T 1 of glucose caused variations in the polarization level of the samples at the time of the experiment due to variations in time from dissolution to injection of the sample. To compensate for this variation, the signals in the different experiments were normalized to initial glucose signal.

Determination of Lactate and Pyruvate Concentrations Using
Conventional NMR Methods-Cells were grown, harvested, and resuspended following the procedure described for the hyperpolarized experiments. The conditions of a hyperpolarized experiment were mimicked in the bench-top NMR experiments: 167 l of a 40 million/ml cell suspension (either PC3 cells of MCF7 cells) was placed shaking at 37°C to which 333 l of a 2 mM glucose solution was added. The cells were incubated with glucose for 70 s after which perchloric acid extractions were made of the entire samples. In short, 200 l of 2.2 M perchloric acid was added to the glucose incubated cells on ice. The samples were neutralized with ϳ50 l 10 M KOH as verified by a pH indicator. To a 500-l extraction, 100 l D 2 O was added, 10 l of a 34 mM 13 C-labeled urea solution and 15 l of a 0.5 M Omniscan solution. 13 C-labeled urea was added as a standard for quantification, and Omniscan was added to reduce the T 1 for faster acquisition. Direct 13 C NMR acquisition was acquired for 48 h per experiment to obtain a high enough signal on the pyruvate resonances for quantification.
Measurement of Lactate and Pyruvate Using Biochemical Methods-Lactate export was measured via an NADH-coupled lactate dehydrogenase spectrophotometric assay. Briefly, supernatant samples were mixed 1:20 with 16 units/ml L-lactate dehydrogenase (Sigma), 2.5 mM NAD ϩ , 0.5 M glycine, 0.6 M hydrazine, pH 9.2, in a UV-compatible 96-well microtiter plate and for incubated 20 min at 37°C. Absorbance at 340 nm was determined and compared with a six-step L-lactate reference series included in each assay. Pyruvate export was determined in a 2,4-dinitrophenolhydrazine-capture assay to stabilize the pyruvate formed. 2 million freshly harvested PC3 or MCF7 cells were added 0 or 5 mM glucose and were incubated for 15 min at 37°C. After incubation, the cells were sedimented by centrifugation at 2000 ϫ g for 2 min.
A 300-l aliquot of supernatant was diluted 3:7 with 2,4dinitrophenolhydrazine to 0.125 mg/ml in 50 mM HCl and derivatized for 10 min at 37°C. The violet color of the hydrazone was developed with NaOH to 0.5 M and determined spectrophotometrically at 550 nm and compared with a series of pyruvate included in the assay.
Measurement of Intracellular pH in Human Breast and Prostate Cells-The intracellular pH was estimated from the hyperpolarized glucose 13 C NMR experiments applying previously reported principles (29). With the assumption that the intracellular enzyme carbonic anhydrase instantaneously equilibrates the produced CO 2 with bicarbonate, these two hyperpolarized NMR signals were quantified from a sum of spectra Ͼ30 s of the individual hyperpolarized glucose 13 C NMR experiments. The Henderson-Hasselbalch equation using a pK a value of 6.35 was hereafter used for calculation of the intracellular pH.

T 1 Determination of [ 13 C 3 ]
Lactate-Because the correctly labeled [ 13 C 3 ]lactate is not commercially available, this compound was produced from glucose in MCF7 cells. The metabolite was then extracted with perchloric acid following standard protocols and freeze-dried. A dynamic nuclear polarization sample of the extract was prepared from dissolving it in a glycerol/water mixture and adding Ox063 to 17 mM, and the sample was hyperpolarized as described above for 1 h. The T 1 of the extracted [ 13 C 3 ]lactate was measured to 34 s Ϯ 2 s (14.1 Tesla, 37°C, n ϭ 2) in a buffer solution and with a set-up as described for the in cell experiments above.
Data Analysis-Experiments for each cell type was repeated at least three times, and mean Ϯ S.D. for each value was calculated. Statistical analysis of the results was performed using the Student's t test. NMR data analysis was performed with the software MNova.
Kinetic Modeling-Kinetic modeling was performed using the ordinary differential equation solver implemented in the program Scilab (version 5.4.1, SciLab Enterprises) with least squares fitting to the data. The input data for the model were the integral values of the hyperpolarized signals from glucose, [1-13 C]pyruvate, and [1-13 C]lactate. The model is described by four variable parameters: the rate of glycolysis, r gluc ; the T 1 of intracellular lactate and pyruvate, T 1 in ; the rate constant of lactate export, k exp ; and the lactate/pyruvate ratio at equilibrium, K lac/pyr . Assuming that the [NAD ϩ ]/[NADH] ratio and the pH is not affected over the course of the 1-min experiment, the conversion of pyruvate to lactate was initially modeled as a bimolecular reversible Michaelis-Menten reaction.
V f is the maximum forward reaction rate and K pyr and K lac are the pyruvate and lactate concentrations at half-maximum velocity. Because the kinetic term always resulted in values in instant equilibrium, this term was substituted with the fixed number of 10 5 to simplify the model. The following set of equations describes the model, Prior to fitting, the data were corrected for pulsing with the radio frequency pulse of 20°used to collect the data. The points in the data set were divided with cos(20) n ϭ cos(20) (t/TR) , where TR is the repetition time of the pulse in the experiment. Due to a fast equilibrium between pyruvate and lactate metabolite pools, the intracellular T 1 of [1-13 C]lactate and [1-13 C]pyruvate are expected to be a mean of the two individual T 1 weighted by the concentrations of the two pools. Therefore, these parameters were fitted as one parameter, T 1 in . In the experiments with PC3 cells, poor initial mixing rendered the initial data points invalid. Hence, data points for the first 2 s were excluded. The following experimental parameters were either measured in different assays or determined from the in-cell hyperpolarized experiments: Gluc (ϳ75,000) ϭ glucose concentration expressed as the hyperpolarized signal in the first spectrum of the time series. This value was obtained by dividing integrals originating from glucose with 6 to get a signal/mol value. T 1 gluc (14 s) ϭ T 1 of glucose. Less than 10% of the hyperpolarized glucose takes part in the glycolysis (the remaining stays extracellular). T 1 ex (34 s) ϭ T 1 of extracellular lactate is measured in a separate experiment (see above). The model was tested for robustness and overfitting. The robustness of the model was addressed by selecting starting parameters obtained from fits to data sets from the other cell line. The model resulted in consistent and robust results for data sets from each cell line independent of the starting values. It was assured that overfitting was not being done by fixing parameters at values obtained from the other cell line while allowing all other parameters to float and compensate for the fixed parameter. All cases showed that it was impossible to produce a good fit to the data with any parameter fixed to a value obtained from the other cell line.

Hyperpolarized Glucose as Substrate for Human Breast and
Prostate Cancer Cells-We have translated the method previously developed for visualization of glucose metabolism in yeast (24) into a tool to study dynamic glucose metabolism in human cells with the aim to be able to identify and quantify biologically relevant control points for changes in human glucose utilization. To this end, hyperpolarized glucose was infused to human breast cancer cells (MCF7) or human prostate cancer cells (PC3) in concentrations where glucose uptake was saturated. The choice of a fully 13 C-labeled and deuterated substrate allowed detection of product signals originating from different carbon positions in the glucose molecule over a total experimental time window of 1 min. Within this experimental time frame, the in-cell conversion of hyperpolarized glucose could be followed through 10 enzyme-catalyzed transformations ending in the products pyruvate and lactate (Fig. 1). The main glycolytic metabolites observed are dihydroxyacetone phosphate (212.6 ppm), pyruvate (C2, 206.4 ppm, C1, 171.6 ppm), and lactate (C1, 183.5 ppm). In the pentose phosphate pathway, 6-phophogluconate (179.8 ppm) and 6-phosphogluconolactone (177.0 ppm) can be detected. The singlet at 161.4 ppm is assigned to bicarbonate, which arises from carbonic anhydraseassisted equilibrium with CO 2 . The CO 2 originates either from decarboxylation of pyruvate or of 6-phosphogluconate. The generally higher signals from metabolites in MCF7 cells compared with PC3 cells indicate a higher rate of glucose metabolism per cell in the former. MCF7 cells are smaller than PC3 cells and contain less soluble protein (ϳ25% less), suggesting that MCF7 cells are more glycolytic than PC3 cells also per soluble protein.

Dynamics of the Hyperpolarized Glycolytic Products, [1-13 C]pyruvate and [1-13 C]lactate, in Human Cancer Cells-
Dynamic nuclear polarization enhancement of NMR signals permits measurement of in-cell dynamic data. Hyperpolarized glycolytic intermediates and end products were followed with a temporal resolution of 2 s for 1 min in MCF7 and PC3 cells (Fig.  2, A and B). A close inspection of the dynamics (Fig. 2C) shows the order in which metabolites appear as a function of their transformation in the glycolysis or in the pentose phosphate pathway. During the first seconds, signals from metabolites in the pentose phosphate pathway are formed faster than pyruvate and lactate. However, lactate and pyruvate eventually dominate. Dihydroxyacetone phosphate upstream in the glycolysis also clearly appears before pyruvate and lactate. In Fig. 3A, the signal development in the two cell types is shown for hyperpolarized pyruvate and lactate. Maximum signal of hyperpolarized [1-13 C]lactate and [1-13 C]pyruvate is obtained ϳ11 s into the experiment. The ratio between hyperpolarized signals from lactate and pyruvate is significantly different (p Ͻ 0.0001)  between the two cell types with an average ratio of ϳ3.2 for PC3 and of ϳ12.5 for MCF7 cells. This ratio was approximately constant in MCF7 cells over the full time course of the experiment, whereas in the PC3 cells, the ratio increased presumably due to a more extensive export of lactate to the extracellular media (Fig. 3, B and C).
Kinetic Model for Production of Hyperpolarized [1-13 C]pyruvate and [1-13 C]lactate from Hyperpolarized Glucose-To be able to quantify the conversion of hyperpolarized glucose and calculate the respective intracellular 13 C-labeled proportions of glucose ending up in lactate and pyruvate, a kinetic model was established. Given the available data and the known pathways, the simplest kinetic model that describes the biochemical system is shown in Fig. 4. The rate and thermodynamic equilibrium ratio of the reversible conversion of pyruvate to lactate depends on the pH and of the NAD ϩ , NADH, pyruvate, and lactate concentrations. In the model, it is assumed that the [NAD ϩ ]/[NADH] ratio and pH is unchanged over the course of the experiment, leaving the thermodynamic ratio at equilibrium between lactate and pyruvate, K lac/pyr , constant. In a hyperpolarized NMR experiment the parameters intrinsic to the method (longitudinal relaxation time (T 1 ) and radio frequency pulsing) lead to signal decays that are independent of metabolic action. Thus, these parameters must be included in the kinetic modeling of metabolite signals (see "Experimental Procedures"). Applying the kinetic model to the hyperpolarized data obtained for MCF7 and PC3 cells led to the fits of hyperpolarized [1-13 C]lactate and [1-13 C]pyruvate shown in Fig. 5. The data fit the kinetic model with high accuracy; R 2 values of 0.89 and 0.99 for pyruvate and lactate signals, respectively, in MCF7 and R 2 values of 0.94 and 0.99 for pyruvate and lactate signals, respectively, in PC3. The lower intensity of the pyruvate signal results in a lower R 2 for the fit to the signal from pyruvate. The lactate export in PC3 cells is significant already 20 s into the experiment where ϳ40% of the lactate is exported (Fig. 5B, top  panel). This supports the interpretation that lactate export is the reason for the increasing ratio between lactate and pyruvate noted in the raw data collected in PC3 cells (Fig. 3C). MCF7 cells are clearly more glycolytic than PC3 cells. The calculated rate of glycolysis in MCF7 cells is a factor of two higher than in PC3 ( Table 1). The lactate concentration in MCF7 cells is more than double as high as in PC3 cells, whereas the pyruvate concentration is similar in the two cell types. The export rate of lactate is ϳ4 times higher in PC3 than in MCF7 cells. The fitted T 1 in of lactate and pyruvate was approximately half of the measured extracellular T 1 of lactate in both cell types. The ratio between lactate and pyruvate differs significantly between the two cell types. It is ϳ4 times higher in the MCF7 cells, suggesting that the free cytosolic [NAD ϩ ]/[NADH] redox conditions differ in the two cell lines. The ratios are well defined in the model with S.D. ϳ5% of the values.

Calculation of the Cytosolic Free Ratio of [NAD ϩ ]/[NADH] in Human Breast and Prostate Cancer
Cells-Based on the ratio between hyperpolarized [1-13 C]lactate and [1-13 C]pyruvate extracted from the kinetic model the cytosolic redox status can be calculated. To this end, the fitted ratio between [1-13 C]lactate and [1-13 C]pyruvate was used as input in the lactate dehydrogenase equilibrium equation, where pH is experimentally estimated and K Eq ϭ 1.11 ϫ 10 Ϫ11 M (14).
pH was measured from the ratio of quantified hyperpolarized CO 2 and hyperpolarized bicarbonate. The pH differed only slightly in the two cell types and was ϳ7.0 Ϯ 0.05 in MCF7 cells and 7.1 Ϯ 0.05 in PC3 cells. The resulting calculated free cytosolic [NAD ϩ ]/[NADH] ratios are shown in Fig. 6. The calculated ratio concentration of [NAD ϩ ]/[NADH] is 2.55 ϫ 10 3 Ϯ 346 in PC3 cells and 7.38 ϫ 10 2 Ϯ 92 in MCF7 cells. The obtained glycolytic rates and the measured signal of lactate were evaluated using conventional NMR techniques. The 13 C NMR data on perchloric acid extractions of PC3 and MCF7 cells incubated with glucose showed a glycolytic rate of 0.16 Ϯ 0.014 fmol/s/cell for MCF7 cells and 0.08 Ϯ 0.007 fmol/s/cell for PC3 cells, showing the glycolytic rate of MCF7 cells to be twice as high as that of PC3 cells.

DISCUSSION
As still more studies relate the pathologically perturbed homeostasis of the cell to changes in the free cytosolic [NAD ϩ ]/ [NADH] ratio (6,13,30), methods that can determine this ratio become increasingly important. Classically, the cytosolic [NAD ϩ ]/[NADH] ratio has been determined indirectly based on steady-state metabolite concentrations measured following cell disruption (31). Pool sizes at steady-state conditions are extremely labile, and during extraction, mixing of compartments and transient shifts in pools are likely to occur. Thus, this method is highly dependent on efficient quenching and effective extraction. To circumvent these challenges, genetically encoded sensors, based on green fluorescent protein, have been developed that allow determination of [NAD ϩ ]/[NADH] ratios directly in living cells. Although this technology is exceedingly elegant, it does require cells to be genetically transformed and will only be applicable to cells in culture.
In the present study, we describe a very different approach based on glucose, which is magnetically hyperpolarized on 13 Clabeled nuclei. The hyperpolarized substance, which is chemically indistinguishable from natural glucose, was perifused into living cells, and glycolytic metabolite formation was measured directly inside cells. This non-invasive method allowed determination of pyruvate-lactate dynamics and their ratios. As no genetic modification is needed this tool is potentially also applicable to ex vivo studies in tissue and in vivo animal studies. ratio, or in pH, over the extent of the experiment would be expected to alter the equilibrium ratio between pyruvate and lactate. However, non-steady-state models did not describe the data better than an instant and fixed equilibrium model. Hence, in the model, it is assumed that the [NAD ϩ ]/[NADH] ratio is not affected over the short duration of the experiment (1 min) at the given substrate concentrations (a few mM). In summary, the model supports the notion that the lactate dehydrogenasecatalyzed reaction is faster than the glycolytic steps leading up to it. This is in accordance with literature, where e.g. in neoplastic cells, the potential lactate dehydrogenase activity is 20 -200 times  higher than hexokinase and phosphofructokinase and transport into mitochondria (32). Thus, it is likely that lactate and pyruvate will also be in equilibrium with the free cytosolic [NAD ϩ ]/ [NADH] pool in non-cancerous cells with an active TCA cycle, allowing the model to be employed in these systems as well. The low NMR signals from metabolites in the pentose phosphate pathway, 6-phophogluconolactone, 6-phophogluconate, and bicarbonate ( Fig. 1) confirm that the flux of glucose through the pentose phosphate pathway is low compared with the flux through glycolysis. The kinetic model assumes the glycolysis within a specific cell type to take place at the same rate during the entire experiment. This assumption was supported by the observation that the rate of lactate generation was constant for Ͼ10 min after exposure of cells to the external glucose bolus, as measured with standard biochemical techniques. Similarly, standard biochemistry was performed to assure that extracellular pyruvate concentrations were very minor. No pyruvate could be detected within the limit of detection of the applied assay. On this basis, pyruvate export was excluded in the kinetic model. On the contrary, the export of lactate was significant in both cell types. The extracellular lactate concentrations were measured to 550 and 600 M after a 15-min incubation with glucose underlining the importance of including lactate export in the model. The glycolytic rates obtained in the hyperpolarized experiments compare well with those obtained by the conventional biochemical methods (0.09 fmol/s/cell and 0.08 fmol/s/cell, respectively, in PC3 cells and 0.18 fmol/s/cell and 0.16 fmol/s/ cell, respectively, in MCF7 cells). The glycolytic rates calculated from the hyperpolarized experiments are ϳ10% higher than those estimated from the perchloric acid extracts. This is seen as a negligible discrepancy in light of the number of involved experimental steps in both methods. The free cytosolic [NAD ϩ ]/[NADH] ratio in the two cancer cells types was calculated using a glycolysis model. The ratio was 7.38 ϫ 10 2 Ϯ 92 in the breast cancer cells (MCF7; cytosolic pH 7.0) and 2.55 ϫ 10 3 Ϯ 346 in the prostate cancer cells (PC3; cytosolic pH 7.1). The MCF7 ratio compares well with the ratios between 550 and 1164 reported in the literature for healthy liver or fibroblast cells and tissue (9, 14, 16 -17). The ratio of [NAD ϩ ]/[NADH] in PC3 cells, on the other hand, is more than four times higher than in the MCF7 cells, and higher than previously reported values even for 2-deoxyglucose-treated cells (16).
Interestingly, healthy prostate cells, and even more so prostate cancer cells have a distinctive metabolic phenotype that suggests an explanation for the extreme [NAD ϩ ]/[NADH] redox balance observed (33). A central role of prostate epithelia is to produce and secrete citrate to the prostatic fluid as energy resource for the spermatozoa. Citrate is diverted away from the TCA cycle into the cytosol for excretion due to zinc inhibition of mitochondrial aconitase. In prostate cancer cells, the inhibition of aconitase is relieved whereby the TCA cycle and citrate oxidation are reactivated (34,35). Even though the cells no longer excrete citrate, some of it is exported to the cytosol for fatty acid biosynthesis (34,35,36). So, unlike most cancer cells, prostate cancer cells are characterized by slow glycolysis (37), slow glucose uptake (38,39), and active mitochondria capable of oxidizing fatty acids. That the latter is important specifically in PC3 prostate cancer cells is supported by the finding that citrate synthase is present in PC3 cells but not in MCF7 cells (40). A slow glycolysis is in accord with the glycolysis rate determination for PC3 cells, which was only half of that of the breast cancer cells, MCF7 (Table 1). A slow glycolytic rate is by itself expected to limit the reduction of NAD ϩ to NADH in glycolysis and increase [NAD ϩ ]/[NADH] as shown in practice by treating primary human fibroblasts with 2-deoxyglucose resulting in a 2.6-fold increase in free cytosolic [NAD ϩ ]/[NADH] ratio (16). Fatty acid biosynthesis is shown to be one of the most important cytosolic NAD(P)H sinks, and it has been shown to be active in PC3 cells (36,41,42). Taken together, this suggests that a significantly increased free cytosolic [NAD ϩ ]/[NADH] redox potential could be expected in prostate cancer cells in accordance with the hyperpolarized measurements, due in part to increased mitochondrial activity, reduced glycolytic flux, and fatty acid biosynthesis. If the bias in metabolism toward fatty acids is ultimately the reason for the shift in free cytosolic [NAD ϩ ]/[NADH] ratio seen in the prostate cancer cells, this ratio could function as a read-out for the redox status of the cells and as such be used for monitoring of therapies targeting the enhanced dependence on fatty acid catabolism of the prostate cancer cells.
Despite the very different glycolytic fluxes in the two studied cell types, the sensitivity of the measurements was excellent under both experimental conditions. The study shows that the method enables measurement of [NAD ϩ ]/[NADH] ratios even in cell types that do not particularly favor glucose as a nutrient. Thus, the use of hyperpolarized glucose for metabolic phenotyping is expected to be applicable to a wide selection of healthy and diseased cell and tissue types.