High-Resolution Hyperpolarized In Vivo Metabolic ¹³C Spectroscopy at Low Magnetic Field (48.7 mT) Following Murine Tail-Vein Injection

Abstract

High-resolution ¹³C NMR spectroscopy of hyperpolarized succinate-1-¹³C is reported in vitro and in vivo using a clinical-scale, biplanar (80 cm-gap) 48.7 mT permanent magnet with a high homogeneity magnetic field. Non-localized ¹³C NMR spectra were recorded at 0.52 MHz resonance frequency over the torso of a tumor-bearing mouse every 2 seconds. Hyperpolarized ¹³C NMR signals with linewidths of ~3 Hz (corresponding to ~6 ppm) were recorded in vitro (2 mL in a syringe) and in vivo (over a mouse torso). Comparison of the full width at half maximum (FWHM) for ¹³C NMR spectra acquired at 48.7 mT and at 4.7 T in a small-animal MRI scanner demonstrates a factor of ~12 improvement for the ¹³C resonance linewidth attainable at 48.7 mT compared to that at 4.7 T in vitro. ¹³C hyperpolarized succinate-1-¹³C resonance linewidths in vivo are at least one order of magnitude narrower at 48.7 mT compared to those observed in high-field (≥ 3 T) studies employing HP contrast agents. The demonstrated high-resolution ¹³C in vivo spectroscopy could be useful for high-sensitivity spectroscopic studies involving monitoring HP agent uptake or detecting metabolism using HP contrast agents with sufficiently large ¹³C chemical shift differences.

Graphical Abstract

1. Introduction

The emergence of several hyperpolarization techniques enhancing the detected signal from nuclear spins by orders of magnitude [1–5] have led to a wealth of opportunities to pursue in the developing arena of high sensitivity magnetic resonance research. In non-hyperpolarized methods, the magnetic field strength of the detection magnet serves to polarize the nuclear spins (usually protons) with polarization levels of ≤3.5 ppm/Tesla constraining the detection limit [6] to large ensembles of nuclear spins (conventionally bulk water in clinical MR applications). Hyperpolarization techniques, however, produce spin polarization independent of the readout magnet, and therefore, they provide freedom to select a magnetic field strength, and correspondingly a detection frequency, optimized to the nucleus observed [7, 8] and the dielectric properties of the sample [9–11]. Of particular interest is the quickly developing frontier of low-field magnetic resonance (NMR) and imaging (MRI) [12–14], due to its ability to strongly complement the study of hyperpolarized (HP) nuclei, and its potential for application in areas of study including heteronuclear spin-labeled metabolites [15, 16] or other contrast agents [17–22] at low concentrations.

Key advantages have given rise to the desirability of pursuing low-field NMR and MRI. First, detection at low static magnetic field B₀ is known to yield far less vulnerability to magnetic susceptibility-induced field gradient artifacts in imaging, including those surrounding metallic implants [23] or interfaces between materials such as airways and tissues as occur in the lungs, inner ear, and nasal passages, thus significantly reducing the advanced shimming hardware requirements [24]. Second, the specific absorption rate (SAR) of energy deposited in the sample/subject during application of radiofrequency (RF) pulses scales $\propto B_0^2$. Consequently, lowering the field strength confers the benefit of reducing SAR to potentially negligible levels [11], which not only alleviates patient safety concerns, but can also lead to design and implementation of faster pulse sequences by permitting use of RF pulses for echoes instead of gradient echoes [11]. Third, through utilizing long-lived spin states (LLSS) [25–36] or taking advantage of reduced contributions from chemical shift anisotropy particularly below 0.1 T [37], hyperpolarized state lifetimes can significantly extend for protons and/or heteronuclei at low magnetic fields. Fourth, experimental validation of the theoretical equations for low-field NMR and MRI [12, 13] sensitivity for hyperpolarized states has favorably corroborated the weak frequency dependence of hyperpolarized state detection, where these theories indicate the potential for approaching or even exceeding the sensitivity of high-field NMR/MRI [12, 13]. Lastly, utilization of low-field magnets provides significantly less expensive purchasing and maintenance costs, which can pair favorably with lower-cost methods of producing hyperpolarized spin states such as Spin Exchange Optical Pumping (SEOP) primarily focused on ³He and ¹²⁹Xe [1, 38–42] or the high-throughput parahydrogen-based methods [5, 14, 43] including parahydrogen-induced polarization (PHIP) [44–46] and signal amplification by reversible exchange (SABRE) [47–51].

In this communication, we demonstrate proof-of-principle high-resolution ¹³C NMR spectroscopic detection of HP succinate-1-¹³C-2,3-d₂ (¹³C-SUX) at B₀ = 47.5 mT in vitro as well as in vivo. Here, we take advantage of the significantly more homogeneous B₀ field generated by a clinical-scale 80 cm-gap biplanar permanent magnet with ~13 ppm homogeneity over 40 cm diameter spherical volume (DSV) compared to a previously employed 8.9 cm-bore magnet with >20 ppm homogeneity over 4 cm [52]. Succinate was employed, because it finds potential application for cancer imaging [53–56]. With the improved DSV size and B₀ homogeneity, we report on in vivo high-resolution ¹³C NMR spectroscopy capable of resolving splittings ≥ 4 Hz arising due to ¹³C-¹H heteronuclear spin-spin couplings within the HP ¹³C-SUX molecule (note that although ¹³C-²H two- and three-bond couplings are present in HP ¹³C-SUX (Figure 1), they are significantly weaker than the corresponding ¹³C-¹H spin-spin couplings, and they are likely are manifested as line-broadening in the detected ¹³C NMR spectra). We compare 48.7 mT ¹³C in vitro and in vivo NMR spectra with those obtained at 4.7 T, and find at least one order of magnitude of ¹³C linewidth (in the units of Hz) improvement at low magnetic field compared to high magnetic field.

Figure 1.

Chemistry of ¹³C-SUX hyperpolarization and NMR spectroscopy. (a) the conversion scheme of fumarate-1-¹³C-2,3-d₂ to yield succinate-1-¹³C-2,3-d₂ via the PHIP process. (b) and (e) ¹³C NMR spectroscopy of HP ¹³C-SUX in hollow plastic spheres (2.8 mL volume) at 4.7 T (best shim) and an HCA-filled syringe at 48.7 mT, respectively. (c) and (f) thermal Boltzmann polarization reference of 1 g sodium 1-¹³C-acetate in D₂O at 4.7 T (single average) and 5 g sodium 1-¹³C-acetate in 15 mL D₂O at 48.7 mT (16 averages), respectively. (d) T₁ decay curve of HP ¹³C-SUX at 48.7 mT with simulated T₁ of 88 ± 5 s. See additional details in the text.

Material and Methods

1.1. PHIP Hyperpolarizer

A PHIP hyperpolarizer used to prepare the 1-¹³C-succinate-d₂ hyperpolarized contrast agent (HCA) was fully automated via open-source software and hardware and integrated with a dual-channel low-field NMR spectrometer (Kea2, Magritek, Wellington, New Zealand), with operations controlled by computer through a graphical user interface (GUI) [52]. The Helmholtz saddle configuration radiofrequency (RF) coils used for the RF polarization transfer sequence developed by Goldman and Johannesson [57] operate at the ¹H and ¹³C resonances for B₀ = 5.75 mT when capacitively tuned and matched to 50 Ω and enclose a high pressure plastic reactor with ~56 mL volume kept at temperatures between 50–60 ºC. Quality assurance of the produced agent before use is achieved via a small tipping angle RF pulse in order to establish the degree of polarization of the agent. The process of producing a dose of HCA takes less than a minute, and the routines for readying the system for the next dose, including cleaning, result in a total PHIP cycle time of less than 3 minutes. Full details of the PHIP polarizer instrumentation including schematics, software, and bill of materials are freely available elsewhere [52]. A home-built parahydrogen (p-H₂) generator using a low-power cryocooler is used to produce continuously 90+% p-H₂ from bulk hydrogen at flow rates up to 150 sscm and 25 bar pressure. A 1 L double-ended 316 SS sample cylinder with PTFE lining (316L-50DF4-500-T-PD, Swagelok) and brass Yor-Lok tube fittings (5272K291, McMaster-Carr) provides a gas reservoir for p-H₂ collection between experiments (~0.4 L is consumed per production of HCA, coincidentally matching conveniently the PHIP cycle time).

1.2. PHIP Hyperpolarization of 1-¹³C-succinate-d₂ (¹³C-SUX)

Preparation of aqueous Rh-based hydrogenation catalyst and fumaric acid-1-¹³C-d₂ precursor used to produce HP succinate-1-¹³C-d₂ (¹³C-SUX) followed previously the fully described previously procedure [52]. Briefly, precursor stock solution of fumaric acid-1-¹³C-d₂ (Cambridge Isotopes, CDLM-6062-PK, 1-¹³C 99%, 2,3-D₂ 98%, 3.00 mmol, 0.357 g) is prepared in D₂O, degassed, pH adjusted to 10.3, and mixed with the standard PHIP catalyst/ligand system of rhodium(I) catalyst, bis(norbornadiene) rhodium(I) tetrafluoroborate (0.200 g, 0.54 mmol, 45-0230, CAS 36620-11-8, Strem Chemicals, MA) and phosphorus ligand, disodium salt of 1,4-bis[(phenyl-3-propanesulfonate)phosphine]butane (717347, Sigma-Aldrich-Isotec, 0.360 g, 0.64 mmol). Following preparation, the catalyst is stored in a plastic bottle attached to the gas/liquid manifold of the PHIP hyperpolarizer and kept under slight overpressure (~2–5 psi gauge pressure) of ultra-high purity nitrogen or Argon (>99.999%, A-L Compressed Gases Inc., Nashville, TN) [52].

1.3. MRI magnet and RF coils

The upright biplanar permanent magnet clinical imaging system (Supporting Information (SI) and Figure 2) generates B₀ field of 48.7 mT statically shimmed to ~13 ppm homogeneity over a 40 cm DSV (SIGWA 48.7 mT, Boston NMR, Boston, MA). The open-bore configuration achieves an 80 cm-gap between the planar XYZ gradient coil inserts mounted to the pole pieces. Heteronuclear NMR spectroscopic detection involving ¹³C and ¹⁵N on this system has been observed to approach ~1 Hz linewidths (determined as full width at half maximum or FWHM), when measuring NMR samples with μL volumes (e.g., samples in 5 mm NMR tubes). We note that this system is particularly well suited for the direct NMR measurement of J-couplings of hyperpolarized compounds given its low field strength and excellent field homogeneity [30]. While not utilized for shimming in the reported in vivo study, when used for MRI the XYZ planar gradient coil set is capable of reaching 20 mT/m in all axes with linearity within approximately ±6% across the full FOV of 40 cm. Further details of the imaging system (including the Tecmag Redstone console) are provided in the SI.

Figure 2.

Experimental setup and in vivo ¹³C NMR spectroscopy. a) Parahydrogen at ~80% purity was prepared with a home-built parahydrogen generator and used to produce HP ¹³C-SUX from ¹³C-FUM in a home-built PHIP hyperpolarizer. b) Coinciding with the start of production of HP ¹³C-SUX in the PHIP hyperpolarizer, NMR acquisitions every 2 s were acquired with FA=18°, with (insets) showing a selected NMR spectrum from the series alongside the corresponding SVD de-noised spectrum (10 largest singular values). See additional details in the text and SI.

The RF coil built for use in this imaging system was designed to measure metabolic HP contrast agents in vivo in small animal studies involving mice. The solenoid coil was constructed to match the typical dimensions of the animal (17 mm diameter by up to 193 mm body length) to increase the coil filling factor and correspondingly the SNR. The coil had a quality factor of 55. The coil was tuned using 56H02 Johansen trimmer capacitors and ATC nonmagnetic ceramic capacitors to tune the resonant frequency to ¹³C resonance at 0.521 MHz or ¹H resonance at 2.074 MHz and match the coil impedance to the 50 Ω coaxial transmission line impedance. All component values are provided in Tables S1 and S2 in the SI. The resonant circuit for ¹H or ¹³C detection was comprised of a parallel LC resonator with a series capacitive matching element (see Figure S3 and S4, SI). The solenoid was constructed with 170 windings using 28 AWG gauge magnet wire with a measured inductance of 302 μH. The coil was enclosed in an RF shield constructed using square double-sided 1 oz. copper clad PCB materials (473–1007-ND, Digikey, Thief River Falls, MN) to form an enclosed cubic volume (1ft × 1ft × 1ft). This shielding provided an electric field suppression of 60+ dB. The coil was calibrated via a nutation sequence on the Redstone console using variable flip angle via pulse length calibration for fixed power, and the π/2 tip angle was determined to be 115 μs for ¹H and 475 μs for ¹³C for <1 W of applied RF power. Further description of the RF probe is provided in the SI.

1.4. ¹³C HP NMR spectroscopy and ¹³C NMR spectroscopy using thermal polarization

Sodium 1-¹³C-acetate was used for Boltzmann thermal polarization analysis at 48.7 mT and 4.7 T for HP ¹³C-SUX spectroscopic comparison given ~97-fold field strength ratio for B₀. HP ¹³C-SUX spectroscopy was performed by injecting HCA into a 2.8 mL hollow plastic sphere situated inside the corresponding RF coil at either 4.7 T (best ¹H shim settings on the acetate phantom) or 48.7 mT (static magnet shim) followed by application of a small tipping angle hard RF pulse to observe the shimmed NMR resolution obtainable. For in vivo detection, a time series of NMR acquisitions to allow observation of HCA dynamics was commenced on the Redstone console at the start of the PHIP hyperpolarization sequence that produces a batch of HP ¹³C-SUX.

1.5. ¹³C spectroscopy in small rodents using HP ¹³C-SUX

The feasibility of in vivo ¹³C high-resolution detection of spin-spin couplings was tested in mice for HP ¹³C-SUX at the low magnetic field of 48.7 mT. Procedurally, the 4T1 mammary carcinoma cells were maintained in RPMI-1640 medium supplemented with fetal bovine serum (10%, v/v) and the culture dish was incubated in a 37°C, 5% CO₂. Eight-to ten-week-old female nude mice were purchased from Jackson Laboratory. All animal experimental protocols were approved by the Vanderbilt University Institutional Animal Care and Use Committee. Nude mice 6–8weeks of age (from Jackson Laboratory, Bar Harbor, ME, USA) were implanted sub-cutaneously under anesthesia (isoflurane mixed with 2% oxygen) with 1.0×10⁶ cells in the mammary fat pad using a 27-G needle. The progress of tumor growth was monitored via every-other-day measurement of tumor size and animal weight. When the tumors reached approximately 4mm in diameter, in vivo ¹³C spectroscopy study was performed.

In preparation for the NMR experiment, the mice were anesthetized using ketamine:xylazine (75mg/kg: 5mg/kg) via intraperitoneal injection. The anesthetic dose was designed to immobilize animals for approximately 60 minutes. After anesthetization, with the animal lying on a thin film carrier, a 28-gauge needle affixed to a catheter line was inserted into the lateral tail vein. The animal with catheter in place was then transported and placed inside the RF probe with the catheter line fitting through the RF shield. Prior to acquiring NMR spectroscopy or MRI imaging, animals were slowly injected over ~5 seconds timespan with ~0.2 mL of HP ¹³C-SUX via the catheter in performance of a tail-vein injection. After NMR spectroscopy, animals were retrieved from the scanner and placed back into a warm cage under constant observation until recovered from anesthesia.

2. Results and Discussion

2.1. In vitro ¹³C NMR Spectroscopy at 4.7 T and 48.7 mT

¹³C NMR spectra of thermally polarized sodium acetate-1-¹³C (1 g dissolved in 2.8 mL D₂O DSV) were recorded using a 4.7 T Varian preclinical MRI scanner (Figure 1c, 1 scan) and using a Tecmag Redstone console (Figure 1c, 16 scans) at 50 MHz and 0.52 MHz ¹³C resonance frequency, respectively. The high-field acquisition was performed using active shimming, while the low-field acquisition was performed without any additional magnet shimming. Acetate-1-¹³C moiety results in a characteristic ¹³C quartet appearance due to ~6.0 Hz ¹H-¹³C two-bond spin-spin coupling. The quartet features are somewhat better resolved at 48.7 mT, Figure 1c vs. 1f.

¹³C spectroscopy of HP ¹³C-SUX was performed at 4.7 T using the same type of hollow plastic spherical container. The resulting spectrum (Figure 1b) shows one relatively broad line with FWHM ~ 36 Hz. ¹³C spectroscopy of HP ¹³C-SUX performed at 48.7 mT was conducted in a plastic syringe container. The spectrum (Figure 1e) shows a very well resolved multiplet pattern—a manifestation of complex network of two- and three-bond ¹³C-¹H spin-spin couplings. The half-height line width (HHLW) was estimated to be < 3 Hz, and the two outer peaks of the splitting pattern (~3.9 Hz away) are well differentiated. The ~12-fold narrower linewidth (and consequently better resolution in units of Hz) obtained at 48.7 mT is possible due to the fact that susceptibility-induced field gradients which scale linearly with B₀ are significantly reduced at 48.7 mT compared to 4.7 T (by the ratio of ~97 of the two field strengths). The chemical shift dispersion (in units of Hz) scales linearly with B₀, and as a result, it is reduced by a factor of ~97 at 48.7 mT compared to that at 4.7 T. The effective ~12-fold line narrowing (due to reduced susceptibility-induced B₀ field gradients) described above partially mitigates the above reduction in chemical shift dispersion, and the overall resolving power of ¹³C spectroscopy at 48.7 mT was approximately 8 times worse than that at 4.7 T (i.e., ~97:12).

Despite nominal reduction of spectral resolution, the significant in vivo ¹³C line narrowing at 48.7 mT vs. 4.7 T provides a certain advantage from the perspective of the detection sensitivity, or SNR. Because the sensitivity for HP detection can scale as B₀^1/4 [12] or as B₀^0.14 in the range of 10–500 kHz [58] depending on coil geometry and construction and the means of detection, which is a very weak field dependence, the significant reduction in FWHM is an important contributing factor to the overall detection sensitivity argument for low-field MR. In particular, this may translate into more sensitive LF MR detection schemes of HP contrast agents for some applications, which can take advantage of this significant narrowing of linewidth despite the commensurate reduction in the chemical shift dispersion, e.g. HP gas spectroscopy and imaging [5, 59].

2.2. In vitro ¹³C-SUX T₁ in deoxygenated D₂O

Previous measurement of ¹³C-SUX hyperpolarized lifetime in D₂O at 47.5 mT yielded 75 ± 3 s in vitro [15]. At 48.7 mT, the lifetime of HP ¹³C-SUX in D₂O was measured to be 88 ± 5 s in vitro according to the decay curve shown in Figure 1 and accounting for using a RF pulse with flip angle (FA) ~ 18°. This is in good agreement with previously reported value, and the minor discrepancy is likely caused by slight imprecision (< 1°) in calibrating the low-field RF pulse and small variations in the pH of the produced HCA, which is known to modulate ¹³C T₁ of HP SUX significantly [53, 54, 60]. ¹³C chemical shift anisotropy (CSA) decreases significantly in units of Hz at 48.7 mT compared to that at ≥3 T. As a result, ¹³C CSA likely has a negligible contribution to the T₁ relaxation of the ¹³C HP state at 48.7 mT, and dipole-dipole interactions likely represent the dominant T₁ relaxation mechanism at such low magnetic fields.

2.3. In vivo ¹³C NMR Spectroscopy at 4.7 T and 48.7 mT

HP ¹³C-SUX was injected as a bolus dose in the tail vein of anesthetized tumor-bearing mouse, which was positioned inside a RF coil placed in the center of 48.7 mT magnet (Figure 1). Non-localized ¹³C NMR spectra were recorded via the NMR console commencing at the start of HCA production on the PHIP hyperpolarizer every 2 seconds using a small angle pulse for every RF excitation. The time window for recording NMR spectra of HP ¹³C-SUX was ~80 seconds (Figure 2b). We note that the entire dynamic ¹³C in vivo scan was effectively completed ~3 minutes after the beginning of contrast agent production in the PHIP hyperpolarizer. The ¹³C NMR signal increases at the very beginning due to HCA entering the body of the animal, and it decreases after reaching a maximum post-injection due to (i) T₁ decay, (ii) RF pulse-associated losses, and (iii) possible metabolism. Despite the significant detection volume (whole mouse vs. a small syringe), the ¹³C NMR linewidth remained largely unchanged: ~3 Hz (in vivo, Figure 2b) vs. ~3 Hz (in vitro, see above). While this result is not surprising because the susceptibility-induced gradients are negligible, the demonstration of such high-spectral-resolution HP ¹³C low-field NMR spectroscopy in vivo has been performed for the first time to the best of our knowledge. Most in vivo preclinical experiments with ¹³C HP contrast agents are performed at 3 T and above, and ¹³C FWHM is typically ≥1 ppm corresponding to ≥30 Hz [55, 61–65]. The HP ¹³C FWHM (in units of Hz) demonstrated here in vivo is at least an order of magnitude improvement compared to high-field in vivo studies of HP HCAs. Moreover, the FWHM of ~3 Hz (corresponding to ~6 ppm) is only a factor of ≤ 6 worse than that observed in high-field studies (i.e. ≥1 ppm) [55, 61–65]. While the chemical shift resolution (in units of ppm) is certainly reduced at 48.7 mT, this reduction is offset through the gain in sensitivity via ¹³C line-narrowing, which represents an interesting trade-off between the detection sensitivity (i.e. SNR) and resolution.

2.4. Imaging and biomedical translation

Although HP NMR spectroscopy has been reported in the ZULF regime [37, 66, 67] including frequency-selective RF excitation [67], the use of conventional MRI frequency encoding in ZULF potentially is challenging due to the very low resonance frequency, and new MRI encoding techniques likely need to be developed. However, ¹³C (as well as ¹H, ³He and ¹²⁹Xe) MRI imaging can be performed at B₀ > 6 mT (including ~48.7 mT, the field studied here) as has been demonstrated by us [15, 52, 68–70] and others [8, 19, 20, 71–75] previously using conventional MRI approaches. In particular, ¹³C frequency encoding can still be accomplished at ~0.5 MHz resonance frequency (f₀) as employed here and using a RF coil with suitable quality factor Q and reasonable choice of spectral width (SW) to satisfy the imaging bandwidth condition of f₀/Q ≫ SW [15]. The direct use of conventional encoding approaches is a clear translational advantage of LF MRI compared to the current status of ZULF due to the ready availability of commercial hardware and the wide range of already established pulse sequences.

The ¹³C NMR spectroscopy demonstrated here at 48.7 mT with ~6 ppm spectral resolution (~3 Hz FWHM) could be useful for in vivo ¹³C spectroscopic studies aiming to: (i) differentiate HP pyruvate-¹³C and its main metabolic product in tumors of HP lactate-1-¹³C, because their HP ¹³C resonances have a sufficient chemical shift difference of ~12 ppm [61, 62]; (ii) pH imaging via spectroscopic detection of HP ¹³C-bicarbonate and ¹³CO₂, because their resonances are separated by ~35 ppm [76]; and (iii) potentially other injectable ¹³C HP contrast agents with sufficiently different chemical shifts of HP metabolites [65, 77].

3. Conclusion

High-resolution (FWHM ~ 3 Hz) ¹³C NMR spectroscopy of HP ¹³C-SUX has been successfully demonstrated at 0.521 MHz resonance frequency in vitro and in vivo using a homogeneous (13 ppm over 40 cm DSV) 80 cm-gap 48.7 mT magnet. Resolution of ¹³C line splittings of ~4 Hz were successfully observed in vivo. The in vivo ¹³C linewidth (FWHM) obtained at 48.7 mT was ~12 times better than that achieved with a 4.7 T small-animal MRI scanner using the same HCA. A significant gain in spectral linewidth for ¹³C low-field in vivo NMR spectroscopy is advantageous for improving the detection sensitivity of LF NMR. Potential biomedical applications can include monitoring HP ¹³C agent uptake as well as detecting metabolism of biomolecules with sufficiently different chemical shifts at such low fields including MRSI of HP pyruvate-1-¹³C/lactate-1-¹³C, HP bicarbonate-¹³C/¹³CO₂ and others [78–80].