Hyperpolarized ¹³C Metabolic Magnetic Resonance Spectroscopy and Imaging

Dynamic nuclear polarization with subsequent sample dissolution has enabled real-time studies of metabolism in biological systems. Hyperpolarized [1-¹³C]pyruvate was used to study lactate dehydrogenase activity in a prostate carcinoma cell line in vitro.

In the past decades, new methods for tumor staging, restaging, treatment response monitoring, and recurrence detection of a variety of cancers have emerged in conjunction with the state-of-the-art positron emission tomography with ¹⁸F-fluorodeoxyglucose ([¹⁸F]-FDG PET). ¹³C magnetic resonance spectroscopic imaging (¹³CMRSI) is a minimally invasive imaging method that enables the monitoring of metabolism in vivo and in real time. As with any other method based on ¹³C nuclear magnetic resonance (NMR), it faces the challenge of low thermal polarization and a subsequent low signal-to-noise ratio due to the relatively low gyromagnetic ratio of ¹³C and its low natural abundance in biological samples. By overcoming these limitations, dynamic nuclear polarization (DNP) with subsequent sample dissolution has recently enabled commonly used NMR and magnetic resonance imaging (MRI) systems to measure, study, and image key metabolic pathways in various biological systems. A particularly interesting and promising molecule used in ¹³CMRSI is [1-¹³C]pyruvate, which, in the last ten years, has been widely used for in vitro, preclinical, and, more recently, clinical studies to investigate the cellular energy metabolism in cancer and other diseases. In this article, we outline the technique of dissolution DNP using a 3.35 T preclinical DNP hyperpolarizer and demonstrate its usage in in vitro studies. A similar protocol for hyperpolarization may be applied for the most part in in vivo studies as well. To do so, we used lactate dehydrogenase (LDH) and catalyzed the metabolic reaction of [1-¹³C]pyruvate to [1-¹³C]lactate in a prostate carcinoma cell line, PC3, in vitro using ¹³CMRSI.

Presently, the most widely used clinical method for tumor staging, restaging, treatment response monitoring, and recurrence detection of a wide variety of cancers is [¹⁸F]-FDG PET.¹ However, recently, several novel and alternative approaches have emerged. One of those methods is ¹³CMRSI. This technique involves the introduction of the ¹³C-molecule into a biological sample, followed by minimally invasive MRI to assess the metabolism in vitro or in vivo in real time. Nevertheless, the biggest challenge of ¹³CMRSI, compared to the other methods such as [¹⁸F]-FDG PET or computed tomography, is its low signal-to-noise ratio.

The NMR signal is directly proportional to the level of polarization, a ratio of the spin ½ nuclei population difference in two energy states to the total population (Figure 1A). The polarization is a product of the gyromagnetic ratio (γ) of the nuclei and the applied magnetic field strength over the temperature. A typical polarization of ¹H nuclei is in the order of 0.001% to 0.005% at 3 T, which gives a relatively poor signal-to-noise ratio. Today's state-of-the-art MRI has been a successful imaging method only due to the high abundance of ¹H in biological samples and the high gyromagnetic ratio of ¹H (γ_1H = 42.576 MHz/T). However, observing other nuclei, such as carbon, is more demanding. The only stable, magnetically active carbon isotope, ¹³C, makes up only 1.1% of all carbon atoms. In addition, the gyromagnetic ratio of ¹³C (γ_13C = 10.705 MHz/T) is four times lower than that of ¹H, leading to a lower detection efficiency. In summary, the low ¹³C abundance and low γ_13C cause thermal ¹³C measurements to achieve 0.0176% of the sensitivity of a ¹H-NMR measurement in vivo.

Dynamic Nuclear Polarization

A method to overcome the relatively poor sensitivity of ¹³C measurements is DNP. It was originally described for metals in 1953 by Albert W. Overhauser. In his article, he stated: "It is shown that if the electron spin resonance of the conduction electrons is saturated, the nuclei will be polarized to the same degree they would be if their gyromagnetic ratio were that of the electron spin."² Later that year, Carver and Slichter experimentally confirmed Overhauser's hypothesis³. In 1958, Abragam and Proctor described this effect for electrons in liquids and named it the "solid effect." At temperatures below 4 K, electron-spin polarization reaches nearly 100% and is more than three orders of magnitude higher than the nuclear-spin polarization (Figure 1B)⁴. This occurs because the gyromagnetic ratio of the electron (γ_e = 28024.944 MHz/T) is three orders of magnitude higher than the nuclear gyromagnetic ratios. The weak interactions between electrons and nuclei, such as the Overhauser effect, the solid effect, the cross effect, and the thermal mixing effect, allow the transfer of polarization from electron spins to nuclear spins using microwave irradiation with a frequency close to the corresponding electron paramagnetic resonance (EPR) frequency^5,6. DNP theory has been further developed to involve more electrons and thermal mixing. Nevertheless, to date, no unified quantitative theoretical description of DNP has been published^7,8.

Figure 1: Understanding Dynamic Nuclear Polarization and Hyperpolarization. A) A schematic comparison of the spin population in the thermal equilibrium polarization state and the hyperpolarized state. B) The polarization is dependent upon temperature. The polarization of an electron (e^-) reaches 100% below 1.4 K. The DNP allows the transfer of the polarization from the e- to the ¹³C nuclei, which increases their polarization up to 10⁵-fold. Please click here to view a larger version of this figure.

To introduce DNP in studies of biological systems using ¹³C NMR, subsequent rapid sample dissolution had to be developed. 50 years after Overhauser's hypothesis, Jan H. Ardenkjaer-Larsen et al. solved the technically challenging issue of bringing the hyperpolarized frozen sample into the liquid state with minimal hyperpolarization loss⁶. Dissolution DNP opened a new field of research called ¹³CMRSI, providing a new method to investigate and characterize various disease states^9,10. As stable carriers of an unpaired electron, a trityl radical tris (8-carboxy-2,2,6,6-tetra-(hydroxyethyl)-benzo-[1,2-4,5]-bis-(1,3)-dithiole-4-yl)-methyl sodium salt (OX063) or (2,2,6,6-Tetramethylpiperidin-1-yl)oxyl (TEMPO) is usually used. These are mixed with the desired ¹³C-labeled molecule and exposed to microwave irradiation with a frequency close to the corresponding EPR frequency. Using this technique, the polarization of ¹³C nuclei can be increased up to 37%¹¹. This results in a 10⁵-fold polarization enhancement compared to the thermal equilibrium polarization^11,12. However, as soon as the microwave irradiation is stopped and/or the ¹³C-molecule is transferred to the liquid state, the polarization decays with the longitudinal relaxation time (T₁) of the ¹³C nucleus that was polarized. Thus, the invention of fast dissolution techniques or any subsequent technique shortening the time before experimental measurement (i.e., injection) is crucial for biological applications¹³.

There are three major requirements that the candidate molecule needs to fulfill for successful ¹³CMRSI studies. First, the ¹³C nucleus of interest has to have a sufficiently long T₁ (> 10 s). The choice of the ¹³C-label is crucial. The best candidate nuclei are carbons with no direct contact with ¹H-nuclei via a bond. It also needs to be rapidly metabolized within 2 - 3 T₁ times, resulting in a downstream metabolic product with a significantly different chemical shift from the original substance. The sample mixture must also form an amorphous glass when in a solid state so that the spatial distribution decreases the distance between the electron and ¹³C, allowing the transfer of polarization. If the candidate molecule does not form amorphous glass naturally, it needs to be highly soluble in a glassing agent, such as glycerol or dimethyl sulfoxide¹⁴. These requirements result in a relatively small number of candidate molecules. However, even after the successful discovery of a suitable molecule, developing a working protocol for hyperpolarization can be technically challenging^9,14,15.

In recent years, several substrates have been successfully polarized, such as [1-¹³C]pyruvate^12,16-36, [2-¹³C]pyruvate³⁷, [1-¹³C]ethyl pyruvate³⁸, [1-¹³C]lactate³⁹, [1-¹³C]fumarate^40-43, ¹³C-bicarbonate^36,44,45, [1-¹³C]sodium acetate^43,46-49, ¹³C-urea^6,36,50,51, [5-¹³C]glutamine^15,52,53, [1-¹³C]glutamate^53,54, [1-¹³C]2-oxoglutarate⁵⁵, [1-¹³C]alanine, and others^14,56. A particularly interesting and commonly used substrate for hyperpolarization is [1-¹³C]pyruvate. It is widely used in preclinical studies to investigate the cellular energy-metabolism in various diseases^14,17,22. [1-¹³C]pyruvate meets all the requirements for successful hyperpolarization, including a relatively long T₁ and rapid transport across the cell membrane before subsequently being metabolized. Preclinical studies with [1-¹³C]pyruvate are currently being translated into the clinic⁵⁷.

Metabolism of Pyruvate

It is well known that there is a direct link between mutations in a cancer cells' DNA and changes in their metabolic pathways. Already in the 1920s, Otto Warburg discovered that there is an increased metabolism of glucose and production of lactate in tumors compared to healthy tissue^58-60. Subsequently, various alternations in other metabolic pathways, such as the pentose-phosphate pathway, the tricarboxylic acid cycle, oxidative phosphorylation, and the synthesis of nucleotides and lipids, have been described.

Pyruvate is the final product of glycolysis. In the tumor, it undergoes anaerobic glycolysis catalyzed by LDH⁶¹ and reacts with the reduced form of the coenzyme nicotinamide adenine dinucleotide (NADH), resulting in lactate and the oxidized form of the coenzyme (NAD⁺). Alternatively, pyruvate undergoes a transamination reaction with glutamate to form alanine, catalyzed by alanine transaminase (ALT). Both reactions are readily reversible. Pyruvate also undergoes decarboxylation catalyzed by pyruvate dehydrogenase (PDH) to carbon dioxide and acetyl-CoA, representing an irreversible reaction at this step. Alternations in these reaction rates can be linked to tumor metabolism^{17,21,22,25,62}. The metabolic pathways are summarized in Figure 2.

Figure 2: Diagram of the major metabolic reaction of pyruvate. Pyruvate/lactate conversion is catalyzed by LDH, and pyruvate/alanine conversion is catalyzed by ALT. Pyruvate is irreversibly converted to acetyl-CoA and CO₂ by PDH, and CO₂ is in a pH-dependent equilibrium with bicarbonate⁸⁰. Please click here to view a larger version of this figure.

The detection of hyperpolarized [1-¹³C]pyruvate and its metabolites has been previously demonstrated in the rat heart^37,63-65, liver⁶⁶, muscle, and kidney^62,67. One study demonstrated significant differences in the lactate-to-alanine ratio between the normal and fasted rat liver⁶⁶ and demonstrated a highly elevated and hyperpolarized [1-¹³C]lactate level in liver cancer^68,69. There is evidence that the tumor grade can be identified in a transgenic adenocarcinoma of mouse prostate (TRAMP) using hyperpolarized [1-¹³C]pyruvate²², with the hyperpolarized lactate levels showing a high correlation with the histological grade of the excised tumors. The alanine catalyzed from pyruvate by ALT has also been suggested as a useful marker in rat hepatocellular carcinoma²³.

Measuring the pyruvate-lactate metabolic flux has been used for monitoring ischemia^63,65,70 and as a response to treatment with cytotoxic chemotherapy^17,40, targeted drugs^24,25,41, or radiotherapy²⁶ in animal models. It has also been used for the detection of the phosphatidylinositol 3-kinase (PI3K) inhibitor LY294002 response in glioblastoma and breast cancer mouse models²⁵. Changes in pyruvate metabolism in brain tumors²⁶ and prostate cancer^24,71 have also been observed after treatment.

Prostate Carcinoma

Prostate carcinoma is the predominant cancer in elderly men and the second leading cancer related to death in men worldwide⁷². To date, no reliable, non-invasive methods are available for an early diagnosis and characterization of prostate cancer^73,74, emphasizing the urgent need for novel metabolic imaging techniques to enable stringent detection and staging of patients. Prostate carcinoma was used as a model to demonstrate the possibilities of dissolution DNP combined with ¹³CMRSI in patients⁵⁷. This work was continued in a first clinical trial employing [1-¹³C]pyruvate and ¹³CMRSI for the imaging of prostate cancer, and it has just recently has been completed (NCT01229618).

The motivation behind this work was to illustrate in more detail and for a wider audience the application of the ¹³CMRSI method in a preclinical setting with cells. Measuring the LDH-catalyzed metabolism of [1-¹³C]pyruvate to [1-¹³C]lactate in vitro in the PC3 prostate carcinoma cell line, we demonstrate the possible application of dissolution DNP in in vitro studies and address the crucial steps and challenges during experiments.

1. Sample Stock Solution Preparation

1. Add gadoterate meglumine (GadM, 0.5 mol/L) to concentrated [1-¹³C]pyruvic acid to give a final concentration of 1-mmol/L GadM. Add trityl radical tris (8-carboxy-2,2,6,6-tetra-(hydroxyethyl)-benzo-[1,2-4,5]-bis-(1,3)-dithiole-4-yl)-methyl sodium salt (OX063) to this mixture to give a final concentration of 15 mmol/L. Vortex until complete dissolution.
   NOTE: This stock solution preparation is designed for usage with a 3.35-T preclinical DNP hyperpolarizer. When a 7-T clinical hyperpolarizer is used, the gadoterate meglumine is not required because, at a higher magnetic field, its benefits are negligible. The addition of a gadolinium-based contrast agent increases the achievable solid-state polarization and also the polarization rate. However, in the liquid state, the contrast agent shortens the T₁ relaxation time.

2. Growing the Cell Culture

1. Grow PC3 cells in a culture flask with a 125-cm² growth area. Use F-12K medium containing 10% fetal calf serum (FCS) and maintain the cells at 37 °C in a humidified atmosphere at 5% CO₂. Before the dissolution step, remove the medium from the culture flask.
   NOTE: Each cell line requires a particular preparation protocol for cell propagation. Consult the requirements with the cell line provider.

3. Preparation of the Cells for the Experiment

1. Remove the cell medium and wash the cells with ~ 10 mL of phosphate-buffered saline (PBS).
2. Add 5 mL of trypsin to the flask and return the cell culture flasks to the incubator for 3 to 5 min.
3. Add ~ 5 mL of F-12K medium to deactivate the trypsin.
4. Count the cells using an automatic cell counter. Mix 10 µL of the cell solution with 10 µL of the stain solution. Mix well with the pipette and transfer 10 µL of the mixture into the chamber of a "counting glass".
5. Remove and count the cells in the flask(s). Transfer the appropriate volumes containing the desired number of cells (e.g., 5 x 10⁶ up to 10⁸) into plastic vials.
6. Centrifuge the cells at 1,200 x g for 5 min and discard the supernatant.
7. Re-suspend the cells in the F-12K medium containing 10% FCS to a total volume of 800 μL and transfer them into a reaction cup (2 mL). Place the reaction cup into a plastic vial filled with warm water.

4. Dissolution Agent Preparation

NOTE: The dissolution agent is a liquid that is used to dissolve the hyperpolarized sample. In biological applications, dissolution is usually performed with H₂O-based or deuterium oxide (D₂O)-based buffers, such as PBS or tris(hydroxymethyl)aminomethane (Tris), containing 1 g/L ethylenediaminetetraacetic acid (EDTA).

1. Preparation of 20 mmol/L PBS buffer
   1. To prepare 100 mL of the dissolution agent, dissolve 36 mg of monosodium phosphate (NaH₂PO₄), 247 mg of disodium phosphate (Na₂HPO₄), and 10 mg of EDTA in a solution of 20 mmol/L sodium hydroxide (NaOH) in D₂O. Mix properly until complete dissolution.
      NOTE: EDTA (1 g/L) is added to the buffer to eliminate possible ferromagnetic ions, which can spoil the hyperpolarization. The NaOH is used to neutralize the pyruvic acid in a 1:1 mol ratio to reach a pH of 7.4.

5. Variable Temperature Insert (VTI) Cooldown

1. In the DNP-NMR polarizer program main window, click on "Cooldown."
   NOTE: This switches on the vacuum pump and evacuates the VTI to approximately 5.0 mbar. Subsequently, the needle valve between the VTI and the liquid helium reservoir fully opens, allowing liquid helium to flow into the VTI. The flow rate is regulated by the needle valve to maintain the optimal amount of liquid helium in the VTI until it reaches the helium boiling temperature. Then, the VTI is evacuated to almost complete vacuum, and the temperature reaches approximately 1.4 K. The VTI is filled with liquid helium up to 65%. At this point, the instrument is ready for sample insertion.

6. Sample Preparation and Insertion

1. Using a micropipette, add ~ 8 µL of ¹³C-labeled sample stock solution into a plastic cup.
2. Attach the plastic cup to the insertion rod and initiate the sample insertion process by pressing "Insert sample" in the main program window. Select "Normal sample" and click "Continue."
   NOTE: During this process, the needle valve first closes to discontinue the flow of liquid helium into VTI, and the pressure in the VTI then increases. The sample holder inside the VTI is raised from the liquid helium, the inlet valve at the top of the VTI opens, and a gaseous helium flow is introduced from the inlet valve to prevent outside contamination by air moisture.
3. When prompted, push the insertion rod with the attached plastic cup down into the VTI. Make sure to reach the sample holder at the bottom of the VTI. Otherwise, the gaseous helium can push the sample out of the VTI.
4. Detach and remove the insertion rod.
5. Finish the procedure by clicking "Next" in the dialog window. The sample insertion procedure should not take longer than 10 s.
   NOTE: The inlet valve then closes, the gaseous helium flow is discontinued, the sample holder with the sample cup is submerged into liquid helium, and the needle valve is opened to allow the liquid helium to flow into the VTI. After 5 - 10 min, the VTI is cooled below 1.4 K, allowing all of the free electrons to be polarized.
6. Confirm that the plastic cup with the sample was introduced correctly into the VTI by checking that it is not attached to the insertion rod or pushed out from the VTI by helium gas. Then click "Finish."

7. Microwave Sweep (optional)

NOTE: A microwave sweep allows the determination of the optimal microwave frequency to maximize the hyperpolarization rate of the ¹³C nuclei in the target compound.

1. To measure the microwave sweep, start the RINMR program, type "HYPERSENSENMR," and click "Select Config" and "Do Microsweep."
2. To initiate the process, select the "calibrate" tab on the main program window.
3. Click "Generate" and choose the beginning and ending frequency (e.g., 94.100 GHz-94.200 GHz), the frequency step size (e.g., 20 MHz), the power (100 mW), and the time (60 s). Click "Continue," "Enable," and "Start."
   NOTE: With these settings, the hyperpolarizer first polarizes the sample for 60 s using a microwave frequency of 94.100 GHz and a power of 100 mW. Then, it applies a 90° radio-frequency (RF) pulse and acquires the hyperpolarized ¹³C signal using the built-in spectrometer. These steps are repeated for each step in specified frequency range. For subsequent hyperpolarization, choose the microwave frequency with the maximal signal amplitude measured.

8. Polarization

1. To measure the polarization build-up, start the RINMR program, type "HYPERSENSENMR," and click "Select Config" and "Solid Build-up."
2. In the DNP-NMR polarizer program main window, click "Polarization" to initiate the hyperpolarization process.
3. Choose the optimal microwave frequency (obtained during the microwave sweep) and the power (e.g., 100 mW) for the sample and click "Next."
4. Enable "Polarization build-up monitoring" and click "Finish."
5. Polarize the sample to > 95% (~ 60 min for [1-¹³C]pyruvate).
   NOTE: During the polarization, microwaves are guided into the VTI and to the sample, causing the ¹³C spins to align with the hyperpolarized unpaired electron spins. To measure the hyperpolarization buildup, RF pulses with a flip angle (FA) of 5° are applied periodically (e.g., every 300 s), and the resulting signal is plotted as a polarization build-up curve.

9. Dissolution

1. When the polarization reaches > 95%, initiate the dissolution process by clicking "Dissolution" in the DNP-NMR polarizer program main window.
2. Choose the dissolution process from the drop-down menu and click "Next."
   NOTE: The polarizer allows one to define the desired dissolution process by choosing the timing of the chasing gas.
3. Load ~ 5 mL of the dissolution agent through the top valve into a heated vessel in the dissolution part of the polarizer. Calculate the exact volume of the dissolution agent needed using following equation:
   
   where V_DA is the wanted volume of dissolution agent, m_PA, m_OX063, and m_Gad are the masses of the pyruvate, OX063 and gadoterate meglumine, respectively, added to the sample stock solution.
4. Place the dissolution stick in the active position above the inlet valve.
   NOTE: This allows the instrument to connect its dissolution instrumentation to the sample cup in the VTI.
5. Click "Finish" to start the dissolution process.
   NOTE: The dissolution agent is pressurized to 3 bar by helium gas and is subsequently heated up to 200 °C, causing an increase in pressure. When the pressure reaches 10 bar, the needle valve closes to discontinue the flow of liquid helium into the VTI. The sample holder raises the cup from the liquid helium. The dissolution stick is lowered into the VTI and connected to the sample cup. The conditioned dissolution agent is pushed by the pressure, which results from the heating vessel containing the dissolution buffer and helium gas, through the dissolution stick to the cup, causing a rapid dissolution of the sample. The solution then flows out into the collection flask via plastic tubing. The dissolution stick with the attached cup is then raised from the VTI.
6. Move the dissolution stick with the attached cup to the "cleaning" position and finish the process by clicking "Finish."

10. Detection of the ¹³C Hyperpolarized Signal

1. ¹³C metabolic magnetic resonance spectroscopy in vitro
   1. Mix 200 µL of the 20-mmol/L dissolved hyperpolarized sample from the collection flask with 800 µL of the cell solution.
      NOTE: The resulting final concentration of [1-¹³C]pyruvate is 4 mmol/L.
   2. Mix the suspension well using a micropipette and transfer ~ 600 µL into a 5-mm NMR tube.
   3. Insert the 5-mm NMR tube into the 1-T NMR spectrometer. In the main window of the software, click "Run" to start the measurement, applying series of one hundred 10° RF pulses every 3 s.
      NOTE: Measure the time between the initial mixing of the hyperpolarized sample with the cells and the start of the spectroscopic acquisition. Ensure that the mixing procedure does not exceed 30 s to minimize polarization loss.
2. ¹³C metabolic magnetic resonance imaging
   1. To build a container for the in vitro experiments using the MRI spectrometer, take a 5-mL syringe and connect it to a catheter (d = 1.2 mm) that is long enough to reach from the spectrometer's iso-center to the approachable area of the spectrometer.
   2. Fill the in vitro container with the cell solution of the desired concentration for the experiment (e.g., 10⁸) or with an enzymatic solution.
   3. Place an in vitro container at the isocenter of the MRI magnet. Place a ¹³C-tuned radio frequency receiver coil on the container. Place a concentrated ¹³C-labeled calibration phantom (e.g., 10-mol/L ¹³C-urea) nearby.
   4. Insert the "in vitro container" near the iso-center of the NMR scanner.
   5. Run the scanner's standard 3-plane localization sequence and adjust the in vitro container's position to the iso-center, as needed.
   6. Run a ¹H T2-weighted "anatomical" sequence covering the in vitro container localization. Use the following settings: 2D spin echo with axial orientation, repetition time (TR) = 2,000 ms, echo time (TE) = 20 ms, slice thickness = 1 mm, field of view covering the in vitro container, and 16 echoes per excitation. Ensure that field shimming is done on protons during this step.
   7. In the anatomical images, select 5 contiguous slices centered on the region of interest. Prescribe a ¹³C spectroscopic calibration acquisition covering the selected anatomical slices. Use the following settings: 2D Block-Siegert calibration sequence with axial orientation 12 x 12 centric encoded, TR = 1,000 ms, slice thickness = 5 mm, field of view matching anatomical images, number of scans (NS) = 64, bandwidth = 5,000 Hz, and FA = 90°.
   8. Select the ¹³C spectroscopic calibration sequence (for more information, see Schulte et al. 2011)⁷⁵ from the pulse sequence library. Download the pulse sequence to the scanner from the computer by clicking "Download." Click on "Spectra Prescan" to run the spectroscopic prescan. In the spectrum magnitude plot, adjust the peak from the ¹³C calibration phantom to the center of the scanner frequency. Set the receiver gains to the maximum. Click "Start" to run the ¹³C spectroscopic calibration sequence. Note the reported transmit gain and centric frequency.
   9. Set a ¹³C chemical shift imaging (CSI) acquisition covering the selected anatomical slices. Use the following settings: 2D echo-planar spectroscopic imaging (EPSI) with axial orientation 12 x 12 centric encoded, TR = 400 ms, slice thickness = 5 mm, field of view matching anatomical images, NS = 300, and bandwidth = 5,000 Hz.
      NOTE: EPSI samples a single line in k-space repeatedly after one RF excitation to acquire both spatial and spectral information simultaneously. For more information about the acquisition techniques, see the article by Durst et al. 2015⁷⁶.
   10. Download the ¹³C CSI sequence and run the spectroscopic prescan. Adjust the scanner frequency and transmit the gain as specified by the calibration sequence output.
   11. After the hyperpolarized solution is deposited in the collection flask, draw up ~ 3 mL into a syringe and then inject it into the catheter connected to the in vitro container. Start the acquisition. After the acquisition is complete, save the raw data file for subsequent reconstruction.

11. Data Reconstruction

1. Apply one of the two described kinetic models to analyze the acquired data.
   1. In the first method for describing the LDH kinetics, kinetic value (k), compare the sum of the lactate signal (M_LAC) to the signal of all hyperpolarized molecules (M_x)^21,77.
      
   2. In the other method, measure the lactate and pyruvate signals over time and fit these to a kinetic model^17,25,71. To solve the metabolic exchange rate, k_PA→LAC, and the effective signal decay rate of lactate, r_LAC, use the following linear differential equations using the two-site exchange differential model, yielding for lactate:
      
      

Note: The effective lactate signal decay rate r_LAC is dependent upon the lactate longitudinal relaxation time (T_1,LAC), the opposite metabolic exchange rate from lactate to pyruvate k_LAC→PA, the applied FA and TR, and the signal intensity of pyruvate (M_PA) and lactate (M_LAC), taking into account the irreversible signal reduction after each successive excitation:

Therefore, r_LAC results in a single, inseparable term of signal decay. Since it is possible to correct for the flip angle and the repetition time, and even though there is a flux LAC→PA, we assume that the exchange rate from lactate to pyruvate (k_LAC→PA) does not need to be included in the calculation, based on the results of Harrison et al. 2012⁷⁸. Their results show that the k_LAC→PA does not play as crucial a role as one would assume. This mode allows the T₁ relaxation time of lactate to be quantified. This model is independent of pyruvate administration to the measurement, which, in the case of in vitro experiments, is not crucial and can be neglected. It does, however, play an important role for in vivo measurements⁷⁹.

The results of the “microwave sweep” are illustrated in Figure 3. It shows that the optimal microwave frequency for the [1-¹³C]pyruvate sample is at 94.156 GHz for the local 3.35-T hyperpolarizer. All following hyperpolarization experiment (n = 14) were performed using this microwave frequency with a power of 100 mW. The microwave irradiation was applied for 60 to 80 min, leading to a solid-state hyperpolarization higher than 90%. The results are presented in Figure 4. The hyperpolarized [¹³C]pyruvate was mixed with 5 × 10⁶ (n = 2), 10⁷ (n = 2), 2 × 10⁷ (n = 1), 3 × 10⁷ (n = 2), 4 × 10⁷ (n = 1), 6 × 10⁷ (n = 2), 8 × 10⁷ (n = 2), and 10⁸ (n = 1) of the prostate cancer cell line PC3.

The resulting data are summarized in Figure 5 and Figure 6. Acquired data with spectral and temporal resolution are shown in Figure 5A-D and Figure 6A-D, with only a temporal resolution for each molecule observed (Figure 5E-H and Figure 6E-H), and with only a spectral resolution (Figure 5I-L and Figure 5I-L). We have observed three major hyperpolarized signals representing [1-¹³C]pyruvate, [1-¹³C]pyruvate hydrate, and [1-¹³C]lactate, with chemical shifts at 173 ppm, 181 ppm, and 185 ppm, approximately relative to the trimethylsilyl propanoic acid (TMSP) at pH 7.4 and temperature 20 °C. The signal ratios between the three metabolites are summarized in Table 1. The data show a clear correlation between the lactate signal and the number of cells present in the sample (Figure 7). However, the results from the experiments with less than 2 × 10⁷ cells exhibit significant deviation, likely due to a low signal-to-noise ratio. Therefore, we suggest using more cells than this for further experiments. When the relative lactate signal (kinetic value) is normalized by the number of cells (Figure 8), it clearly demonstrates similar uptake and metabolism throughout all of the cells. However, there is a trend of decreasing lactate production per cell with an increasing number of cells. We believe that one of the causes of reduced cell metabolic activity is a very high concentration of cells in a very small volume, resulting in the increased viscosity of the sample. The results of the two-site exchange differential model are summarized in Table 2 and shown in Figure 9. The data follow a trend similar to the previous model: increasing k_PA→LAC with an increasing number of cells. However, this model results in a steeper increase of the kinetics with the number of cells. When the metabolic exchange rate k_PA→LAC is normalized to the number of cells, we can again see a clear trend of decreasing k_PA→LAC with an increasing number of cells (Figure 10).

Figure 11 demonstrates the possibility of the addition of spatial localization to the experiment. It shows a phantom injected with 80 mmol/L hyperpolarized [1-¹³C]pyruvate next to a 10 mol/L ¹³C-urea phantom. The technique allows the attainment of a spectrum with temporal and special resolution (Figure 11A) or of the signal decay of the chosen metabolite signals in time (Figure 11B). The spectra in the time domain can also be summed to receive a better signal-to-noise ratio (Figure 11C). The special resolution allows the choice of the desired frequency region of the ¹³C spectrum belonging to certain metabolites, such as [1-¹³C]pyruvate (Figure 11D), [1-¹³C]pyruvate hydrate (Figure 11E), or reference ¹³C-urea (Figure 11F). It can be co-registered with a ¹H image. The pulse sequence used (EPSI) allows the acquisition of an image of the whole slice every 4.9 s. In summary, this technique can provide data with a spatial, temporal, and spectral resolution for any metabolite.

Figure 3: Results of a Microwave Sweep with [1-¹³C]pyruvate at the Local 3.35-T Hyperpolarizer. The result of the measurements determining the optimal microwave frequency to maximize the hyperpolarization rate of ¹³C nuclei in the target compound of [1-¹³C]pyruvate. The microwave sweep has a shape of an EPR absorption spectrum. The shape and separation of the peaks are based on the radical used (in this case, trityl radical), and the biggest influence have a solid effect and thermal mixing. Please click here to view a larger version of this figure.

Figure 4: Solid State Polarization Buildup of a [1-¹³C]pyruvate Sample. An average of n = 13 solid-state polarization buildups with the error represented by the standard deviation measured every 300 s for up to 4,500 s. Please click here to view a larger version of this figure.

Figure 5: Results of the ¹³C NMR Spectroscopy for the Number of Cells (5 x 10⁶ to 3 x 10⁷ cells). The acquired data plotted with spectral and temporal resolution (A-D), plotted with temporal resolution only for [1-¹³C]pyruvate, [1-¹³C]pyruvate hydrate, and [1-¹³C]lactate (E-H), and plotted with spectral resolution only, summing all time steps (I-L). Please click here to view a larger version of this figure.

Figure 6: Results of the ¹³C NMR Spectroscopy for the Number of Cells (4 x 10⁷ to 1 x 10⁸ cells). The acquired data plotted with spectral and temporal resolution (A-D), plotted with temporal resolution only for [1-¹³C]pyruvate, [1-¹³C]pyruvate hydrate, and [1-¹³C]lactate (E-H), and plotted with spectral resolution only, summing all time steps (I-L). Please click here to view a larger version of this figure.

Figure 7: Results of the Simple Metabolite Ratio Kinetic Modeling. Data represents the ratio of the [1-¹³C]lactate signal to the sum of [1-¹³C]pyruvate, [1-¹³C]pyruvate hydrate, and [1-¹³C]lactate versus the number of cells in the experiments. The error represents the standard deviation. Please click here to view a larger version of this figure.

Figure 8: Results of the Simple Metabolite Ratio Kinetic Modeling Normalized to the Number of Cells. The data represent the ratio of the [1-¹³C]lactate signal to the sum of [1-¹³C]pyruvate, [1-¹³C]pyruvate hydrate, and [1-¹³C]lactate normalized to the number of cells versus the number of cells in the experiments. The error represents the standard deviation. Please click here to view a larger version of this figure.

Figure 9: Results of the Two-site Exchange Differential Model. The data represent the metabolic exchange rate (k_PA→LAC) versus the number of cells in the experiments. The error represents the standard deviation. Please click here to view a larger version of this figure.

Figure 10: Results of the Two-site Exchange Differential Model Normalized to the Number of Cells. The data represent the metabolic exchange rate (k_PA→LAC) normalized to the number of cells versus the number of cells in the experiments. The error represents the standard deviation. Please click here to view a larger version of this figure.

Figure 11: Result of Magnetic Resonance Imaging of the Hyperpolarized [1-¹³C]pyruvate Probe. A) The spectrum acquired over the whole slice and all time steps. B) The decay of the [1-¹³C]pyruvate and [1-¹³C]pyruvate hydrate signal over time. The third signal is the 10 M ¹³C-urea localization reference. C) The spectrum acquired from the whole spatial and temporal resolution. D) The ¹H image overlaid with the ¹³C image of the summed [1-¹³C]pyruvate signal over all time steps. E) The ¹H image overlaid with the ¹³C image of the summed [1-¹³C]pyruvate hydrate signal over all time steps. F) The ¹H image overlaid with the ¹³C image of the summed ¹³C-urea signal over all time steps (reference). The ¹³C-signal in C-E is normalized to the maximum of the signal of the specific metabolite. Please click here to view a larger version of this figure.

Table 1: The Relative Ratio of Hyperpolarized Metabolites with Respect to the Different Number of Cells.

Table 2: Results of the Two-site Exchange Differential Model.

¹³CMRSI with hyperpolarized probes is a promising method to monitor metabolism in real time in vitro and in vivo. One very important aspect when employing this experimental process is the proper standardization, especially regarding in vitro experiments. First, the preparation of the sample needs to be done properly and consistently to achieve the same concentration of hyperpolarized material in each experiment. This requires a precise weighing of both the sample to be hyperpolarized and the buffer. If the concentration is not correct, the final pH of the solution is not precise, which can have an influence on T₁ and the cells' responses. It is also crucial to handle the cells as uniformly as possible. The cells should always be prepared in such a way that there is a minimal delay between cell harvest and the subsequent experiment in order to minimize the duration of time the cells are kept at a very high concentration and low volume. Variation in the cell preparation protocol, such as a different preparation times or temperatures, could result in substantial variations in the obtained data. The mixing of the sample with the cells should also be standardized. It is important to measure the time between the additions of the tracer to the cell suspension and the beginning of the measurement, because this can vary; during the data analysis, this should be considered.

The correct choice of the data analysis and kinetic modeling is crucial in the interpretation of the acquired data. The simple model is suitable for a linear one-way reaction with a constant exchange rate of two metabolites. As described in the introduction, pyruvate undergoes several enzymatic reactions and, more importantly, it also undergoes a non-enzymatic reversible-exchange reaction with pyruvate hydrate. This reaction played a crucial role in the experiments, and its effect is well demonstrated in the experiment with 8 × 10⁷ cells. Although Table 1 indicates that the pyruvate hydrate relative concentration is similar to other experiments, when closely investigated in Figure 6D, it shows a much higher pyruvate hydrate signal at the beginning of the experiment compared to the other experiments. However, when the temporal resolution is summed up, this important information is lost and causes an error in the reconstruction of the data. On the other hand, the two-site exchange differential model is a more robust and precise description of the kinetics because it includes the temporal resolution in the calculation. Thus, it includes the non-enzymatic exchange with pyruvate hydrate, even if it rapidly exchanges with pyruvate during the measurement.

There are various imaging strategies to choose between to observe the hyperpolarized signal or to track the metabolism of a hyperpolarized molecule in preclinical and clinical studies. Durst et al. demonstrated the advantages and disadvantages of different pulse sequnces⁷⁶. The free induction decay chemical shift imaging (FIDCSI) sequence is relatively robust but has limited use for multi-slice and temporally resolved imaging. Echo-planar spectroscopic imaging (EPSI) is robust for gradient issues and off-resonance effects but, it is prone to reconstruction artifacts. The iterative decomposition of water and fat with echo asymmetric and least-squares estimation (IDEAL)⁸¹, spiral chemical shift imaging (ISPCSI), pulse sequence³⁵, and spiral chemical shift imaging (SPCSI) have high encoding efficiencies but are sensitive to B₀ inhomogeneity. The choice of the sequence will depend on the scanner characteristics, the biological question, and the system being investigated.

There are many requirements that need to be fulfilled for successful hyperpolarization. However, there are also several limitations that the hyperpolarized ¹³CMRSI technique is nowadays facing. The primary and unchangeable limitation is the T₁ relaxation time of the ¹³C nucleus in the molecule, which defines the amount of detectable signal available at the specific time of measurement. The signal is lowered by each RF excitation that causes a loss of the hyperpolarization signal repeatedly during data acquisition. Another limitation is the relatively long time period that is required to hyperpolarize a molecule. This typically takes from 30 to 90 min.

In comparison to other techniques of molecule imaging, such as [¹⁸F]-FDG PET, hyperpolarized ¹³CMRSI does not require tumors with increased glycolytic metabolic pathways and therefore, increased glucose consumption. The technique shows a real metabolic flux in real time. On the other hand, [¹⁸F]-FDG PET does not give direct information about metabolism but only indirect information about accumulation in the metabolically active area. This could cause a false negative result, where the tumor seems to be metabolically inactive but is actually using different metabolic pathways, such as glutaminolysis, as the carbon source for proliferation.

In conclusion, dissolution DNP can be used in a variety of applications to study an unlimited list of diseases (such as diabetes)⁸², measure pH^15,36,45, or monitor metabolic changes in various types of cancer. These measurements can be accomplished on different levels, from in vitro cell experiments, through preclinical studies using animal models (such as mice, rats, rabbits, pigs, and dogs), to recent human clinical studies⁵⁷. The future clinical applications will feature a very powerful and noninvasive diagnostic tool that could not only detect and localize the disease but also allow the observation of the treatment response in real time⁸³.

Rolf F. Schulte and Marion I. Menzel are employed with GE Global Research.

E.K. gratefully acknowledges the support of the Graduate School of Bioengineering (GSB) at Technische Universität München. This work was supported by the German Research Foundation (DFG) within the SFB Collaborative Research Center 824, "Imaging for Selection, Monitoring, and Individualization of Cancer Therapies."