High-Throughput Hyperpolarized ¹³C Metabolic Investigations using a Multi-Channel Acquisition System

Abstract

Magnetic resonance imaging and spectroscopy of hyperpolarized (HP) compounds such as [1-¹³C]-pyruvate has shown tremendous potential for new insight into disease and response to therapy. New applications of this technology in clinical research and care will require extensive validation in cells and animal models, a process that may be limited by the high cost and modest throughput associated with dynamic nuclear polarization. Relatively wide spectral separation between [1-¹³C]-pyruvate and its chemical endpoints in vivo are conducive to simultaneous multi-sample measurements, even in the presence of a suboptimal global shim. Multi-channel acquisitions could conserve costs and accelerate experiments by allowing acquisition from multiple independent samples following a single dissolution. Unfortunately, many existing preclinical MRI systems are equipped with only a single channel for broadband acquisitions. In this work, we examine the feasibility of this concept using a broadband multi-channel digital receiver extension and detector arrays that allow concurrent measurement of dynamic spectroscopic data from ex vivo enzyme phantoms, in vitro anaplastic thyroid carcinoma cells, and in vivo in tumor-bearing mice. Throughput and the cost of consumables were improved by up to a factor of four. These preliminary results demonstrate the potential for efficient multi-sample studies employing hyperpolarized agents.

Graphical Abstract

1. INTRODUCTION

The relatively high sensitivity and natural abundance of ¹H nuclei in the body permit noninvasive morphological and functional investigation of disease with magnetic resonance (MR) imaging (MRI) and spectroscopy (MRS). The excellent chemical specificity of MR further enables assessment of certain biochemical characteristics of tissue. It is well-known, for example, that many malignancies undergo higher levels of glycolysis and lactic acid fermentation despite normal tissue oxygenation, a condition that is often referred to as aerobic glycolysis or the Warburg effect [1, 2]. Therefore, dynamic changes in pyruvate and lactate levels in vivo could serve as useful biomarkers of cancer metabolism. The unique ¹³C spectroscopic signatures of pyruvate and lactate could permit their non-invasive observation through MRS, MRI, and MR spectroscopic imaging (MRSI), though such measurements have low sensitivity due to low natural abundance (~1%) and a gyromagnetic ratio that is one-quarter that of ¹H. Low sensitivity necessitates prohibitively long scan times for signal averaging, rendering the use of ¹³C signal from endogenous metabolites impractical for real-time measurement of metabolism in vivo [3].

To overcome sensitivity limitations, methods that create a hyperpolarized (HP) nuclear spin population may be used to dramatically enhance the signal of exogenous substrates. One such method, known as dissolution dynamic nuclear polarization (DNP), involves combining an enriched substrate with a polarizing radical, and transferring polarization from unpaired electrons to nuclear spins through microwave radiation at very low temperatures [4, 5]. The HP sample is rapidly dissolved with a heated buffer solution and quickly transferred to the MR scanner, where it may be used to perform experiments with greater than 10,000-fold signal enhancement [5]. Preliminary studies involving HP [1-¹³C]-pyruvate include very promising investigations of real-time cancer metabolism in cells [6, 7], cancer diagnosis and staging [8, 9], therapeutic response [10–19], and detection of cardiac metabolism [20, 21].

Although HP ¹³C MR shows great potential for in vitro and both preclinical and clinical studies in vivo, it is not without drawbacks. In addition to the technical challenges for HP data acquisition, which are confounded by T₁ relaxation and nonrenewable signal loss due to radiofrequency (RF) excitation, several practical limitations associated with cost, time, and efficiency may delay the widespread preclinical use of HP MRI. First, DNP requires significant build-up time to reach the desired polarization: ~45 minutes to reach 90% of the maximum polarization level for [1-¹³C] pyruvate, limiting the number of observations that can be made in a given interval of time. Second, costs associated with the radical and the ¹³C-enriched substrate can be significant. Third, the DNP process proceeds most efficiently at very low temperature ~1.4 K, requiring the use of cryogens that must often be replenished. System maintenance alone represents a large fraction of the overall cost. In addition, only on the order of 5% of the total dissolution volume may be used for in vivo studies in mice, and a smaller fraction may be used in vitro to maintain a physiologically relevant concentration in in cell studies, resulting in substantial waste [22]. More efficient use of HP ¹³C agents would reduce cost and facilitate integration of this technology into routine biomedical research.

Although the vast majority of preclinical MR systems are equipped with no more than one channel which may be used for detection of ¹³C signals, a handful of clinical scanners support multi-channel HP ¹³C acquisition. These receivers, in combination with RF coil arrays, have permitted improved sensitivity [23, 24], increased spatial coverage [25], and a means to perform partially parallel imaging [26].

To improve the efficiency of ex vivo, in vitro, and in vivo preclinical experiments involving HP ¹³C, we have implemented a low-cost, multi-channel, broadband RF receiver with various array coils for performing multiple measurements in parallel. Signal-to-noise ratio (SNR) calculations, based on data acquired from an enriched ¹³C urea phantom at thermal equilibrium, were used to evaluate consistency among channels and to compare performance against standard scanner hardware. The capability to simultaneously capture spectral dynamics from four distinct volumes was established using a 4-channel RF coil array and phantoms that contained varying concentrations of enzyme infused with HP [1-¹³C] pyruvate. High-throughput in vitro experiments, involving human thyroid cancer cells, were then performed with a dual Helmholtz coil to demonstrate the feasibility for improving measurement efficiency while preserving experimental and environmental consistency. Finally, two surface coils were placed over distinct subcutaneous tumors on one mouse, to demonstrate simultaneous in vivo measurement of metabolism from separate and distinct anatomic sites. This establishes a platform where paired examination of experimental therapies may be readily and rapidly performed, with reduced deviation due to biological or experimental variations.

2. Materials and Methods

2.1 Multi-channel Receiver

The home-built receiver system includes an RF module, a power supply module, an analog-to-digital conversion (ADC) acquisition board, and a Linux workstation. The RF module contains four RF channel cards, each supporting four independent receive channels, and a local oscillation (LO) signal distribution card. The LO signal is derived from the MRI console or alternatively provided by an external signal generator, and it provides a reference for down-converting the input RF signal to the baseband output. Baseband signals were acquired through a high speed ADC board (ICS-645D, Gloucester, Ontario, Canada) that was controlled by the Linux workstation. The ADC board is capable of processing streams from up to 32 channels simultaneously, although 16 channels are currently implemented. The digitized data was processed and visualized on the workstation.

The RF module was built on using a nuclear instrument module (NIM) system bin (4001C, ORTEC, Oak Ridge, TN), which includes a backplane with several dedicated pins for supplying power to each bin and the circuit boards they contain. An XLC01 (Excelsys, Rockwall TX) power source provides stable 6 V and 12 V signals. All electronics were assembled on an FR4 circuit board (thickness 0.059″, 1 oz copper, LPKF, Garbsen, Germany). Circuit layouts were drawn with Cadence (v16.3) software and manufactured using a ProtoMat C100/HF (LPKF, Garbsen, Germany) milling machine.

Summary diagrams for one 4-channel RF board are shown in Figure 1. Signal levels must be carefully managed for good conversion without distortion or addition of excessive noise. The local LO distribution circuit (Fig. 1A) requires a 13.5 dBm signal at its input; we used an ADL5601 (Analog Devices, Norwood, MA) amplifier to achieve the correct signal amplitude. After the LO signal is divided using a four-way splitter (PSC-4-1W+, Mini-Circuits, Brooklyn, NY), it is distributed to each of the four receive channels (Fig. 1B) on that board. Two gain stages (RAM-6+ and MAR-1SM, Mini-Circuits) on the receive boards provide ~35 dB of gain that for the MR signal path that can be adjusted using a voltage-controlled attenuator (G-1, Macom, Lowell, MA). Conversion from RF to baseband is performed with a TUF-3+ mixer (Mini-Circuits). Two series low-pass filters with ~830 kHz cut-off frequency (LPF-B0R6, Mini-Circuits) are used for anti-aliasing and noise rejection, and finally, a driver amplifier (OPA843, Texas Instruments, Dallas, TX) maximizes the dynamic range of the ADC and a cross-diode pair (HSMP-3892, Avago Technologies, San Jose, CA) protects the ADC board from any potentially damaging high-voltage signals.

Figure 1.

Schematics of the multichannel receiver boards. (A) Local distribution board – the local signal from Bruker spectrometer is amplified and distributed to each RF channel card with equal power. (B) RF channel card the – RF signals from various MRI coils are amplified and converted to baseband signals. Then the baseband signals are transferred to Analog-to-Digital converter in the Linux workstation for digital signal processing.

The SNR of the receiver, including the analog path and digital signal processing, was compared to the 7T Biospec MRI system by scanning a 2 mL phantom with 8 M ¹³C-urea at thermal equilibrium. The output signal from the preamplifier was split two ways and simultaneously fed to the console receiver and to each of four channels from the multichannel receiver in turn. A pulse-acquire sequence (TR = 2000ms, TE = 2.6 ms, 2048 readout points, BW = 4.96 kHz, 45° excitation angle, and 96 repetitions) was used for data collection. SNR was calculated by dividing the temporal mean of the phase-corrected real component of the urea peak (integrated over the full spectral width) by the standard deviation of a noise-only region.

2.2 Coil Design

RF array coils were designed, built, and tested in-house. Advantages of surface coils include high local sensitivity, spatial localization, and flexibility of positioning. For the ex vivo experiments with enzyme phantoms, a four-channel surface coil was implemented to measure four samples simultaneously. Each coil consisted of a single loop (outer diameter (OD) = 7 mm, inner diameter (ID) = 4.5 mm) with tuning and matching capacitors, and coil centers were positioned to correspond with four wells of a standard 96-well plate, in a 2 x 2 grid arrangement, with 20 mm separation between nearest coils. Inductive coupling between neighboring coils, as measured on a network analyzer (4395A; Agilent Technology, Santa Clara, CA), was between −34.8 dB and −35.7 dB at 75.53 MHz.

In vitro experiments to measure cell metabolism using HP-[1-¹³C]-pyruvate were performed using an array of two identical Helmholtz coils (10mm ID), illustrated in Figure 2. Coils were separated by 80 mm center-to-center and the inductive coupling between them was −31.4 dB at 75.53 MHz. The SNR of this Helmholtz coil array, including the home-built passive Tx/Rx switch and a preamplifier (MicroWave Technology Inc, Fremont, CA), was compared to the each coil located at isocenter by scanning a 250 μL phantom containing 8 M ¹³C-urea at thermal equilibrium.

Figure 2.

Two-channel array of Helmholtz coils. (A) Nested 5mm and 10mm NMR tubes fit just within the Helmholtz coil. The space between NMR tubes is filled with water to remove the susceptibility interface from the region of interest inside the interior tube. (B) Two independent coils are fixed to a structure that fits within the larger transmit coil (72mm ID).

For the in vivo experiment, two 15-mm ID dual-turn surface coils were similarly implemented for placement on distinct tumor sites. Interactions between transmit and receive coils was minimized by active decoupling, using a PIN diode switch under control of console electronics. Geometric decoupling between surface coils depends on the particular tumor locations, but efforts were made to place coils orthogonal to each other. Low input impedance preamplifiers provided up to an additional ~15 dB of isolation [27, 28].

2.3 Dynamic Nuclear Polarization

Hyperpolarized pyruvate was prepared using a mixture of 6.5mg to 26 mg of neat [1-¹³C] pyruvic acid (Isotech Sigma-Aldrich, St. Louis, MO) containing 15 mM OX063 trityl radical (GE Healthcare, Amersham, Denmark) and 1.5 mM aqueous solution of a gadolinium chelate (ProHance, Bracco Diagnostic, Singen, Germany) [5]. The ¹³C-labeled sample was polarized by DNP using a Hypersense system (Oxford Instruments, Abingdon, UK). The sample was inserted into a 3.35 T vertical bore magnet, frozen to 1.4 K, and irradiated at 94.12 GHz for approximately 45 minutes.

After a plateau in the solid-state polarization level was reached, the frozen sample was rapidly dissolved using a 4 mL buffer solution at 180°C containing 40 mM TRIS (7.6 pH preset), 20 mM to 80 mM NaOH, 0.1 g/L EDTA, and 50 mM NaCl. The final solution, containing 20 mM to 80 mM [1-¹³C] pyruvate, was flushed into a large syringe near the magnet and the appropriate volume was drawn and administered for scanning.

2.4 MR Signal Acquisition and Data Analysis

All experiments were performed on a 7 T Biospec small animal MRI scanner (Bruker Biospin MRI, Billerica, MA) equipped with a single channel for carbon excitation/reception and gradients with a 12-cm ID. A dual-tuned ¹H/¹³C birdcage coil with 72-mm ID (1P T10334, Bruker Biospin MRI, Inc., Ettlingen, Germany) was used for acquiring ¹H reference images. Dynamic HP ¹³C spectroscopic data were also acquired through the broadband multi-channel receiver extension, which was triggered by a TTL output signal from the Biospec. For ¹³C, the birdcage coil was used for transmission and the appropriate surface coil array was used for signal reception.

All data reconstruction and analyses were performed using Matlab (The MathWorks Inc., Natick, MA). The dynamic spectra were phase-corrected and line broadening was applied. Spectral information centered at the 75.53 MHz ¹³C Larmor frequency was mixed down to a sequence-dependent intermediate frequency (IF). The analog signal was then sampled and converted into a complex signal at baseband by digital I/Q demodulation. A 2 MHz sampling rate was selected to ensure Nyquist sampling of the full bandwidth of the low pass filters (f_c ~830 kHz). This avoided not only signal aliasing, but also aliasing of noise in the passband of the low pass filters. The signal was then filtered with a finite-impulse-response (FIR) filter, and data were decimated to the final signal length and bandwidth, then transformed by Fast-Fourier-Transform (FFT) for analysis and display in the spectral domain. The area under the real spectral peaks for pyruvate and lactate were integrated over full-width at half-maximum, and integrated again over time to calculate total signal from each metabolite. The normalized lactate (nLac) ratio was calculated as lactate divided by the sum of pyruvate and lactate signals.

2.5 Enzyme Phantom Experiment

Four 300 μL wells of a standard 96-well plate were filled with an enzyme mixture containing 4.6 mM β-NADH (Sigma Aldrich), a variable concentration (0.0, 1.0, 3.0, and 9.0 U/mL) of lactate dehydrogenase (LDH) (Worthington Biochemical Corp., Lakewood, NJ), and 4 mM final concentration of HP [1-¹³C] pyruvate in a buffer solution containing 203.3 mM NaCl and 81.3 mM Tris (7.6 pH preset). NADH and LDH were thawed from fresh aliquots that were stored at −80°C [29]. The HP pyruvate was mixed with other reagents outside the magnet bore and immediately moved to the magnet isocenter, after which a pulse-acquire scan (TR = 2000 ms, TE = 2.4 ms, 2048 readout points, 4.96 kHz BW, 15° flip angle, 96 repetitions) was initiated. Data were apodized by a 15 Hz exponential window.

2.6 Cell Experiments

Highly aggressive HTH-83 human anaplastic thyroid carcinoma (ATC) cells were cultured at 37°C in a humidified atmosphere consisting of 95% air and 5% CO₂ with 80–90% confluence of 10 x 15 cm dishes. Cells were treated with 5mM 2-deoxyglucose (2-DG) (Sigma) for 1 hour then harvested for experiments. Control cells and 2-DG treated cells were adjusted to the same concentration (4.5×10⁷ cells/mL) before imaging. A portion of the cell suspension was mixed with 0.4% Trypan blue solution (Hyclone) at a 1:1 ratio for 5 min at room temperature. Cell viability was measured using an Automatic Cellometer Reader (Nexcelom, MA). Cells demonstrated excellent viability in both control (96.1%) and 2-DG (97.5%) groups.

To assess the impact of 2-DG on cancer cell metabolism, control and 2-DG treated cells were simultaneously probed with HP-pyruvate. Concentric 5mm and 10mm NMR tubes (Wilmad-Labglass, Vineland, NJ) were placed vertically in the Helmholtz coils (see Figure 2A), and the space between NMR tubes was filled with water to facilitate placement, shimming, and removal of the air/water susceptibility interface from the test solutions. After dissolution, approximately 12.5 μL of 20 mM HP-pyruvate was pipetted into the tip of each 1 mL syringe containing 510 μL of cell slurry. Approximately 250 μL of the suspension (1.1x10⁶ cells) were delivered through 1 m PEEK tubing (Restek Co., Bellefonte, PA) that led from the edge of the bore to the coils at isocenter. The mixture contained final concentrations of ~1 mM HP [1-¹³C] pyruvate and ~0.5 mM unlabeled lactate. A slice-selective pulse sequence (TR = 2000ms, TR = 2.4 ms, 2048 readout points, 4.96 kHz BW, 15° flip angle, 96 repetitions) was initiated by the HyperSense system at the beginning of dissolution. 15 Hz line broadening was applied.

2.7 In vivo multi-tumor measurements

In some experiments, it is possible to achieve multiple experimental conditions within a single animal. Different doses of radiation therapy could be simultaneously administered, for example, to two or more tumors that were implanted within a single experimental animal. Thus, the ability to collect independent and localized metabolic information from different anatomic regions could reduce experimental variations, allow the use of paired statistical tests with higher power, and lower the number of animals that are required in order to detect statistically significant differences.

Athymic nude mice were used to demonstrate simultaneous multi-volume HP ¹³C MRS in vivo on a single animal. Luciferase-tagged 0805C ATC cells were implanted into opposing flanks two weeks before imaging. Prior to experimentation, an animal was anesthetized by isoflurane inhalation (2% in oxygen) and placed supine on a mouse sled with (1–2%) isoflurane in oxygen supplied through a nose cone. Body temperatures were maintained at ~37°C by circulating warm water through the sled. Respiratory rate and body temperature were closely monitored with a small-animal monitoring system (Small Animal Instrument Animal Instruments, Inc., Stony Brook, NY). Following anatomic ¹H imaging, a 10-mm axial slice that contained both tumors was prescribed. 200 μL of 80mM HP [1-¹³C] pyruvate solution was injected via tail vein. Dynamic spectra were acquired using a slice-selective pulse acquire ¹³C sequence (TR = 1500 ms, TE = 2.4 ms, 2048 readout points, 4.96 kHz BW, 15° flip angle, 120 repetitions). All procedures were approved by our Institutional Animal Care and Use Committee, which is accredited by the Association for the Assessment and Accreditation of Laboratory Animal Care International.

3 Results

3.1 SNR Performance

Measurements from the 2 mL 8 M ¹³C urea phantom at thermal equilibrium reveal that comparable SNR values are observed through the multi-channel receiver extension and the Biospec receiver. The paired SNR results are listed in Tables 1 and 2. The slightly higher average SNR in the multi-channel receiver (1.50 x 10⁴ vs. 1.13 x 10⁴) may be explained by imbalance in the splitter, differences in cable length and routing, and differences in gain and digital signal processing algorithms in the two systems. There was a larger variation among SNR measurements performed on the multi-channel receiver compared to the console receiver, likely reflecting relatively minor differences in components along the signal path in the home-built receiver.

Table 1.

SNR of Biospec receiver and each channel of one RF board from the broadband receiver extension

Receiver Channel	Multi-channel	Biospec
Channel #1	1.5112 × 104	1.1256 × 104
Channel #2	1.5228 × 104	1.1251 × 104
Channel #3	1.4932 × 104	1.1296 × 104
Channel #4	1.4604 × 104	1.1250 × 104
Mean	1.4969 × 104	1.1263 × 104
Std Dev	2.7211 × 102	2.1991 × 101

Table 2.

SNR of Helmholtz coils at isocenter and on dual-coil array setup

Coil	Array setup (offset)	Individual coils (isocenter)
	SNR	SNR
Helmholtz #1	1.96 × 103	2.05 × 103
Hemlholtz #2	2.18 × 103	2.27 × 103
Mean	2.07 × 103	2.16 × 103

Signal degradation due to suboptimal global shim is a significant concern for multi-volume MRS. SNR from was measured from 250 μL aliquots of 8 M ¹³C-urea to assess the impact of simultaneous offset measurements using the 2-channel Helmholtz array compared to single-channel measurements at with each coil at isocenter. Table 2 shows a slight degradation of less than 5% (2.07 x 10³ vs 2.16 x 10³) in SNR for the simultaneous/offset case. Linewidths were not significantly increased in the multi-coil configuration compared to coils placed at isocenter (0.5 ppm vs 0.8ppm) and were in all cases sufficient for analysis due to the wide chemical shift spacing for [1-¹³C]-pyruvate and its chemical endpoints in vivo.

3.2 Four Channel Test with Enzyme Phantoms

The feasibility of independent multi-channel dynamic acquisitions was initially demonstrated by scanning multiple enzyme phantoms simultaneously. Figure 3 shows that each channel presented different nLac values that were proportional to LDH concentration. The nLac values were 0.12 for channel 2, 0.48 for channel 3, and 0.76 for channel 4. No lactate production was expected or observed in channel 1 (nLac = 0) because this control case did not include the necessary enzyme. There is no evidence of coupling from the strong lactate signal in channels 3–4 into channel 1.

Figure 3.

Simultaneous dynamic measurement of the chemical conversion of HP [1-¹³C] pyruvate in four distinct samples using a four-channel array. Enzyme phantoms contained different doses of lactate dehydrogenase (LDH) to elicit varying rates of chemical conversion: (A) control: no LDH; (B) 1.0 U/mL; (C) 3.0 U/mL; and (D) 9.0 U/mL. The normalized area under the dynamic lactate curve (nLac) corresponds with LDH concentration.

3.3 Thyroid Cancer Cell Experiment

Multichannel cell experiments were performed to investigate the feasibility of high-throughput HP ¹³C experiments using a small volume (~250μL) of cells. In this experiment, differences in the conversion of HP pyruvate to lactate were compared in control cells and those exposed to 2-DG. Figure 4 shows the different normalized signal intensity of HP ¹³C pyruvate and lactate produced by the cells. In this experiment, the result indicates distinctive nLac values between two channels: 0.28 for the treated cells measured through channel 1 and 0.45 for the untreated cells measured with channel 2. These results are consistent with prior literature; 2-DG is a competitive inhibitor of glycolysis, and it has been shown to reduce the conversion of HP pyruvate to lactate by lowering intracellular reducing potential [30].

Figure 4.

Simultaneous dynamic conversion of HP pyruvate in anaplastic thyroid cancer cells. (A) cells exposed to 2-deoxyglucose, a competitive inhibitor of glycolysis; (B) untreated control cells. As expected, the relative fraction of HP pyruvate that is converted to lactate is higher in untreated cells.

3.4 Animal Model Experiment Using a Dual-coil

To explore the potential for paired tumor measurements while minimizing biological and environmental variations between groups, we investigated the utility and feasibility of simultaneous multi-volume HP ¹³C MRS in vivo. To confirm correspondence between acquisition through the multi-channel and standard receivers, the amplified signal from coil 1 (located over tumor 1) was split between the standard console receiver and the multi-channel receiver. The amplified signal from coil 2 (sensitive to tumor 2) was directly fed to a second channel of the multi-channel receiver. Figure 5 shows signal from the injected HP [1-¹³C] pyruvate and its conversion into HP lactate. The spectral data were also summed over time to generate cumulative spectra of the HP metabolites (Fig 6). These spectra demonstrate distinct tumor metabolism and the consistency of data quality acquired through the standard receiver and the broadband receiver extension developed in this work.

Figure 5.

Simultaneous dynamic measurement of HP pyruvate in two distinct 0805C ATC tumors including approximate locations and sizes of RF coils (A) The anatomic ¹H images show small tumors on the left and right flanks. (B) Dynamic metabolite spectra of tumor 1 (above) and tumor 2 (below) showing time courses of pyruvate (~173 ppm), lactate (~185 ppm), pyruvate-hydrate (~181 ppm), and alanine (~178 ppm).

Figure 6.

Temporal sum of dynamic spectra from dual-tumor animal model. The signal from coil 1 was split between (A) the receiver extension and (B) the standard X receive channel for comparison. (C) Coil 2 was placed over “Tumor 2” and signal was detected via receiver extension.

4 Discussion

In this work, we have demonstrated the versatility of simultaneous multi-sample HP ¹³C MR experiments made feasible by a broad-band multi-channel receiver extension and multinuclear array coils. Various responses to HP [1-¹³C] pyruvate were demonstrated in controlled enzyme phantoms ex vivo, cancer cells in vitro, and anaplastic thyroid tumors in vivo. The signal quality achieved with the multi-channel receiver closely matches that achieved with standard commercially available hardware. The receiver is relatively low-cost and is readily integrated into the experimental workflow. This strategy increased throughput, improved experimental efficiency, and reduced the overall cost of these pilot experiments. Investigators that are interested in multi-sample HP measurements such as these can implement a similar receiver system, or provision the configuration of a new or upgraded console with multiple broadband receiver channels.

Despite the dramatic increase in signal afforded by DNP, limited dosage of HP agents to achieve physiologically relevant concentrations results in measurements that remain limited by SNR. For in vivo measurement in mice, institutional guidelines for animal care restrict permissible injection volumes to a maximum of 200 μL per injection. To enhance sensitivity and extend the scale of HP experiments, the development of novel RF coils for HP ¹³C is essential. In the investigation for in vitro experiments with the optimized Helmholtz coils to improve SNR and reduce the number of cells for each sample, merely 12.5 μL of the HP agent was injected to maintain a physiologically relevant concentration of lactate and pyruvate in a solution containing a limited number of cells. Experiments with cells required careful coordination of cell preparation and delivery that was synchronized with preparation of HP measurements, and a well-rehearsed process for combining the HP substrate with the cell slurry. Despite the relatively low volume of HP agents, and noise predominantly determined by coil losses in all cases, metabolite time courses were readily measured with sufficient SNR for subsequent analysis. The multi-coil configurations provided improvements in sensitivity and spatial localization, and permitted simultaneous observation of multiple experimental conditions from a single dissolution. The apparent difference in SNR between the two measurements highlighted in Figure 4 could be the result of bubbles that displaced the cell slurry or variations in the final concentration following rapid manual transfer of such a small volume of HP pyruvate. However, close inspection of the data reveals that the overall SNR between the two measurements is remarkably similar, and that the apparently lower SNR in Figure 4B is caused by scaling both plots to the maximum value of HP pyruvate seen in each channel. Because cells are exposed to the HP agent for several seconds before the slurry is transferred to coils at isocenter, it is expected that some fraction of HP pyruvate would have been converted into HP lactate even before the first measurement. Thus the initial signal from pyruvate is higher in treated cells, where chemical conversion is seen to be lower; conversely, pyruvate signal is attenuated in the untreated cells, where conversion and initial signal from lactate is higher. In fact the sum total of pyruvate and lactate signals over the whole dynamic measurement, normalized to the standard deviation of noise in each channel, is almost identical.

Although the multi-channel receiver system can be readily scaled to 16 channels (and potentially to 32 channels), up to four-fold throughput improvements were demonstrated in this initial work. Other practical considerations, not related to the receiver, limited the parallelization of experiments. As an example, the location of two flank tumors on a single animal restricted placement of the coils and care was taken to geometrically reduce inductive coupling between coils. We can envision a case where multiple subcutaneous tumors, with different genotype and/or localized therapy, could be implanted in a single experimental subject. High sensitivity over a large FOV covering multiple regions of interest could be achieved using a more traditional imaging array structure. In imaging applications, such an array would support accelerated imaging strategies [26] and provide important reference information about [1-13C]-pyruvate metabolism in normal tissues and in areas between regions of interest that are known a priori. Such arrays also require broadband multichannel receiver capabilities. This work focuses on the use of coil configurations that achieve sufficient spatial localization (via coil sensitivity) to pre-defined experimental geometries for multi-channel dynamic spectroscopy without the need for additional spatial encoding beyond what might have been used in corresponding measurements from a single sample.

Another potential means to improve throughput of in vivo measurements is to scan multiple animals simultaneously [31, 32]. A preliminary experiment demonstrating this approach for orthotopic xenografts of ATC, and which leveraged the capability of the receiver developed in this work, was recently presented [33].

Scaling up enzyme or cell experiments is feasible, but will require optimization of larger coil arrays and systems for accurate distribution of substrates to multiple compartments. For the four-channel enzyme phantoms in this work, coils were separated to decrease inductive coupling. This approach required a tradeoff between isolation and shimming performance. This may not be critical for measuring pyruvate and lactate since their chemical shifts are widely separated, but may be necessary when using different ¹³C label positions or other HP substrates. In this preliminary work, the HP agents were manually delivered into phantoms, cells and animals. Future work will include a robust and repeatable substrate distribution system to improve experimental precision and accuracy. Such a solution should also distribute agents rapidly, since HP signal losses due to T₁ relaxation reduce signal strength. MR-compatible automated injection systems to improve the robustness of injections have been developed [34, 35] and can easily be extended to support multiple injections.

Despite the technical challenges that remain, the simultaneous HP detection schemes demonstrated in this work will be crucial to improve the time and cost efficiency for performing biological investigations with high statistical power. Without these strategies, measurements for one experimental time-point from a large cohort of animals may span multiple days, confounding logistical workflow and potentially leading to higher experimental and environmental variations.

In conclusion, a multi-channel acquisition system and purpose-built multinuclear arrays facilitate high-throughput HP ¹³C investigations. The methods described in this work have the potential to promote the integration of preclinical HP technology into biomedical research to accelerate drug development, improve our understanding of cancer biology, and ultimately lead to discoveries that may be translated into the clinic to improve patient care [36].

Highlights.

1. Most preclinical sites are limited to single-channel X-nucleus measurements. We developed a multi-channel acquisition system to demonstrate high-throughput measurements involving hyperpolarized ¹³C-labeled substrates.

2. Four different enzyme kinetics were measured simultanenously with enzyme phantoms ex vivo

3. The efficiency of measuring cell metabolism increased two-fold using multi-volume detection in vitro

4. Different tumor sites on single animal were measured simultaneously to investigate metabolic profiling in vivo

5. Experiment throughput was improved by up to four-fold, resulting in better workflow, conserved consumables, and lower overall cost of experiments