Sensitivity Enhancement for Detection of Hyperpolarized ¹³C MRI Probes With ¹H Spin Coupling Introduced by Enzymatic Transformation In Vivo

Abstract

Purpose

Although ¹H spin coupling is generally avoided in probes for hyperpolarized (HP) ¹³C MRI, enzymatic transformations of biological interest can introduce large ¹³C-¹H couplings in vivo. The purpose of this study was to develop and investigate the application of ¹H decoupling for enhancing the sensitivity for detection of affected HP ¹³C metabolic products.

Methods

A standalone ¹H decoupler system and custom concentric ¹³C/¹H paddle coil setup were integrated with a clinical 3T MRI scanner for in vivo ¹³C MR studies using HP [2-¹³C]dihydroxyacetone, a novel sensor of hepatic energy status. Major ¹³C-¹H coupling (1)J_CH = ~150 Hz) is introduced after adenosine triphosphate–dependent enzymatic transformation of HP [2-¹³C]dihydroxyacetone to [2-¹³C]glycerol-3-phosphate in vivo. Application of WALTZ-16 ¹H decoupling for elimination of large ¹³C-¹H couplings was first tested in thermally polarized glycerol phantoms and then for in vivo HP MR studies in three rats, scanned both with and without decoupling.

Results

As configured, ¹H-decoupled ¹³C MR of thermally polarized glycerol and the HP metabolic product [2-¹³C]glycerol-3-phosphate was achieved at forward power of approximately 15 W. High-quality 3-s dynamic in vivo HP ¹³C MR scans were acquired with decoupling duty cycle of 5%. Application of ¹H decoupling resulted in sensitivity enhancement of 1.7-fold for detection of metabolic conversion of [2-¹³C]dihydroxyacetone to HP [2-¹³C]glycerol-3-phosphate in vivo.

Conclusions

Application of ¹H decoupling provides significant sensitivity enhancement for detection of HP ¹³C metabolic products with large ¹H spin couplings, and is therefore expected to be useful for preclinical and potentially clinical HP ¹³C MR studies.

INTRODUCTION

Routine hyperpolarization of ¹³C-labeled small molecules using the process of dissolution dynamic nuclear polarization (1) has unlocked significant possibilities for clinical investigations of substrate-level metabolic activity in various disease states using ¹³C MRI. Hyperpolarized (HP) [1-¹³C]pyruvate (2) in particular has attracted considerable interest for applications to cancer and cardiovascular disease (3,4), with recent successful deployment into initial human studies (5,6).

¹H spin coupling is a major feature of ¹³C NMR, yet has thus far played little role in the new field of HP ¹³C MRI. In general, coupling to ¹H spins is conspicuously avoided in HP ¹³C MRI in the effort to minimize loss of sensitivity caused by J-coupling and dipolar relaxation. Two initial studies found that ¹H decoupling can be moderately beneficial for HP [1-¹³C]pyruvate MRI, by providing a modest degree of spectral narrowing of HP [1-¹³C]lactate signal as a result of the elimination of long-range coupling to the distant methyl protons (7,8). Although there has been little further work in this area, other more recently introduced probes could benefit much more substantially from the application of ¹H decoupling. In particular, certain HP probes of interest with long T₁ relaxation times and minimal ¹H couplings are enzymatically transformed in vivo to important metabolic products with major ¹H couplings. This can occur with dehydrogenase activity at the labeled carbon site, exemplified by the reduction via lactate dehydrogenase of HP [2-¹³C]pyruvate to [2-¹³C]lactate (9), which is observed as a well-resolved ¹³C signal doublet at 3 T (J = 146 Hz). Elimination of such large couplings, however, introduces high decoupling power requirements, which in turn require exceptionally good isolation between multinuclear RF channels, which is not typical for clinical MRI systems.

The new work presented here was inspired by the recent development of the new HP probe [2-¹³C]dihydroxyacetone (DHA) (10). Hyperpolarized [2-¹³C]DHA is rapidly taken up into cells and phosphorylated to the metabolic intermediate dihydroxyacetone phosphate in vivo, with subsequent reduction at the labeled site using glycerol-3-phosphate dehydrogenase to form [2-¹³C]glycerol-3-phosphate (G3P) (Fig. 1), which is observed as a ¹³C signal doublet (J = 148 Hz), among other metabolic fates. The adenosine triphosphate–dependent metabolic conversion of HP [2-¹³C]DHA to [2-¹³C]G3P is a promising new marker of liver energy status, and was recently shown to be markedly attenuated in rats challenged with an adenosine triphosphate–depleting fructose load (11). Derangement of normal hepatic energy metabolism underlies the progression of nonalcoholic fatty liver disease to nonalcoholic steatohepatitis (12), suggesting high potential for the use of HP [2-¹³C]DHA in the clinical assessment of nonalcoholic fatty liver disease. Development of improved clinical markers of nonalcoholic fatty liver disease status is in fact an urgent unmet clinical need because of the high prevalence of this disease and the serious limitations of liver biopsy, including invasiveness, sampling error, and limited applicability as a repeat measure (13).

FIG. 1.

Biochemical conversion of DHA to G3P in vivo. Position of the HP ¹³C label is indicated by stars.

In this work we show for the first time that application of ¹H decoupling for ¹³C MRI of HP [2-¹³C]DHA provides greatly enhanced sensitivity for in vivo detection of the metabolic conversion of [2-¹³C]DHA to [2-¹³C]G3P, in experiments conducted in a clinical MRI scanner environment, in which ¹H decoupling is not commonly used (14). To our knowledge, this is the first reported in vivo decoupling of HP ¹³C resonances with large C-H couplings (¹J_CH = ~150 Hz), and the first application showing a major sensitivity benefit of ¹H decoupling for HP ¹³C MRI studies in vivo.

METHODS

Hyperpolarization

For each experiment, a 40-μL sample consisting of 9M [2-¹³C]DHA (Sigma ISOTEC, Miamisburg, Ohio) in 2:1 water:DMSO (v/v), mixed with 22-mM OX63 and 1.0-mM Gd-DOTA, was polarized for a period of approximately 1 h in a commercial Hypersense polarizer (Oxford Instruments, Tubney Woods, United Kingdom) operating at 3.35 T and 1.3 K. Each sample was subsequently rapidly dissolved in 4.3-mL phosphate-buffered saline, to yield an 80-mM sample of HP [2-¹³C]DHA, which was then rapidly transferred to the MRI system.

Radiofrequency Coil Configuration

The MRI experiments were conducted in a clinical 3T MRI scanner (GE Healthcare, Waukesha, Wisconsin). A custom ¹³C/¹H curved paddle-shaped surface coil, depicted in Figure 2, was constructed for the purpose of these studies (Berkshire Magnetics, Berkeley, California). The coil consisted of a pair of concentric octagonal elements tuned to ¹H (outer) and ¹³C (inner) frequencies. As a single paddle, the B₁ fields of both coil elements, and consequently their transmit and receive profiles, fall off with distance from the coil. However, the larger outer loop falls off more slowly than the smaller inner loop. The inner loop was therefore tuned for ¹³C to provide high sensitivity in rat liver and constructed with sufficient size for sensitivity throughout the rat liver (with some weighting toward the proximal liver region), whereas the larger outer loop was tuned for ¹H to provide a relatively uniform decoupling field over this sensitive volume of the ¹³C loop. The ¹³C loop was equipped with passive ¹H blocking to prevent absorption of decoupling power. Although the ¹³C loop was strictly operated in transceive mode, the ¹H loop was alternately used for imaging/spectroscopy (transceive mode) and ¹H decoupling (transmit only). Alternate RF coil setups were also tested, but none performed well at the required decoupling power levels, as a result of insufficient isolation between ¹H and ¹³C channels.

FIG. 2.

Photographs of custom ¹³C/¹H paddle coil used for ¹H-decoupled ¹³C MRI experiments. a: Coil exterior and cabling. b: Coil internals consisting of concentric octagonal ¹H (inner) and ¹³C (outer) elements, with key components labeled by arrows.

¹H Decoupling

¹H decoupling power was provided by a commercial standalone rack-mounted assembly (GE Healthcare), with configuration similar to that described in a previous study (7), and depicted in Figure 3. A standalone device was used for this study, as neither of the clinical MRI systems sited near our polarizers support ¹H decoupling as a built-in feature. The assembly consists principally of a 2-kW, pulsed, 100-W, continuous-waveform radiofrequency (RF) amplifier (Herley, Lancaster, Pennsylvania), driven by a PC-controlled electronic signal generator (Agilent, Santa Clara, California). Output power was measured by feeding amplifier output through a 50-Ω dummy load with |S₂₁| = approximately −30 dB and into an oscilloscope (Tektronix, Mukilteo, Washington). Crucially, the decoupling signal path is equipped with separate low- and high-power filters to block introduction of RF power at 32 MHz, which can easily obscure ¹³C detection. Additionally, for this study, installation of a supplemental bandpass filter along the ¹³C transmit-receive line (i.e., between ¹³C loop and amplifiers) was required, to reduce receiver gain compression observed after activation of decoupling power approaching required levels. This filter was borrowed from a 3T small-animal MRI system (Bruker, Billerica, Massachusetts) sited nearby, and characterized by |S₂₁(128 MHz)| = −58 dB, |S₂₁(32 MHz)| = −0.8 dB.

FIG. 3.

Simplified schematic depiction of standalone accessory ¹H decoupler operation in conjunction with clinical 3T scanner used for HP ¹³C MRI. Major components include (in electronics room adjacent to scanner): PC controller, electronic signal generator (ESG), transistor-transistor logic (TTL) trigger signal from system exciter board, 2-kW peak RF amplifier, and low-and high-power filters; and in scanner room: concentric ¹H (outer) and ¹³C (inner, with passive ¹H blocking) coil elements, bandpass (BP) filter on ¹³C channel installed between ¹³C coil and ¹³C transmit-receive (T/R) switch.

Phantom Experiments

¹H decoupling for ¹³C MRI was first tested in a thermally polarized glycerol phantom, consisting of a standard nitrile glove filled with 200-mL glycerol (natural abundance). Couplings observed in glycerol are very similar to the relevant case of [2-¹³C]G3P. The glycerol phantom was placed on the described ¹³C-¹H coil, along with two standard 100-mL 0.9% (wt/wt) saline intravenous bags to approximate rat-like loading. Nonselective pulse and acquire spectroscopy (pulse width = 500 μs, center frequency = 75 ppm, spectral points = 4096, sweep width = 25 kHz) was conducted with and without ¹H decoupling. A WALTZ-16 sequence (individual pulse width = 0.5 ms) was applied for ¹H decoupling (15), which was=electronically triggered for activation during each readout interval using a scanner-derived transistor-transistor logic trigger signal. Outside of these intervals, decoupling power was attenuated by a factor of 49 dB using the decoupler’s bilevel irradiation feature. The WALTZ-16 scheme was chosen because its bandwidth and insensitivity to B₁ inhomogeneity compare favorably with other possible schemes (15,16), and for the practical reason that it was already implemented on the standalone electronic signal generator.

Hyperpolarized ¹³C MRI Experiments

Three rats each underwent back-to-back HP [2-¹³C]DHA scans, with and without ¹H decoupling. All animal studies were conducted in accordance with a protocol approved by the University of California, San Francisco, Institutional Animal Care and Use Committee. Each rat was anesthetized using 1.5% inhalational isofluorane before having a tail vein catheter implanted and being positioned prone onto the coil with liver region directly above the ¹³C loop. A heated water pad was placed under the rat to maintain body temperature. The first HP injection and scan applied ¹H decoupling, whereas the second injection and scan (approximately 90 min later, within the same scanning session) did not. A previous study applying similar methods showed that the extent of conversion of HP [2-¹³C]DHA to HP [2-¹³C]G3P in liver and kidney returns to a similar level within 80 min after the first injection (11). For in vivo experiments, effectiveness of decoupling was verified just before the HP experiment based on signal from a vial phantom containing 1-mL [2-¹³C]glycerol (Cambridge Isotopes, Tewksbury, Massachusetts) placed adjacent to the rat. This vial was removed for each HP experiment. Each injection consisted of 2.2 mL of 80-mM HP [2-¹³C]DHA delivered over 12 s, with dynamic nonlocalized spectroscopy as conducted in the phantom experiments (pulse width = 500 μs, center frequency = 75 ppm, nominal flip angle = 20°, spectral points = 4096, sweep width = 25 kHz). Excitation was repeated dynamically every 3 s over 30 s, starting 20 s after the start of injection. Decoupling power was again attenuated by 49 dB between readout intervals, so the duty cycle for application of ¹H decoupling power was 5%.

RESULTS

As configured, optimal ¹H decoupling of thermally polarized ¹³C spectra in the glycerol phantom was achieved at a measured decoupler output power of approximately 15 W (Fig. 4B, top row). The measured ¹³C signal enhancements for both doublet and triplet resonances were close to the two-fold increases predicted from theory—2.0x in the case of the C₂ doublet, and 1.8x in the case of the C₁/C₃ triplet. There was no significant nuclear Overhauser effect enhancement, as the bilevel irradiation attenuation factor was set to a very large value of 49 dB.

FIG. 4.

Multinuclear RF coil setup for in vivo ¹H-decoupled HP [2-¹³C]DHA scans in rats (a) and comparison of results with and without ¹H decoupling (b). a: Octagonal ¹H (outer) and ¹³C (inner) coil traces are shown overlaid on a standard coronal anatomic localizer image acquired using the paddle’s ¹H channel. b: ¹³C MR spectra (magnitude) acquired in the clinical 3T MRI scanner with (right column) and without (left column) ¹H decoupling. Top row: Comparison of thermally polarized, nonenriched glycerol phantom spectra. Bottom row: Summed dynamic spectra from a pair of back-to-back HP [2-¹³C]DHA experiments in one rat.

Application of ¹H decoupling in the HP [2-¹³C]DHA experiments collapsed the [2-¹³C]G3P signal doublet into a singlet and greatly enhanced the sensitivity for detection of G3P (Fig. 4b, bottom row). The placement of the RF coil is shown in Figure 4a. Across the three rats, the summed [2-¹³C]G3P peak signal was enhanced by an average factor of 71+/− 29% (P = 0.049, paired two-tailed t-test). Dynamic HP [2-¹³C]DHA spectra from one of the three rats scanned are shown in Figure 5, including data acquired with and without ¹H decoupling. In the absence of decoupling, the G3P doublet falls into the noise floor at an earlier time.

FIG. 5.

Dynamics of HP [2-¹³C]DHA spectra in one rat, for the cases of no ¹H decoupling (a) and ¹H-decoupled (b). The temporal dynamic spacing was 3 s, starting 20 s after the start of injection of HP [2-¹³C]DHA. Magnitude NMR data are shown, scaled to approximate signal-to-noise ratio level.

Occasional decoupling signal breakthrough was clearly detected in some ¹³C transients, but effects were mild until especially high power levels exceeding our requirements, and thus did not generally interfere with ¹H-decoupled ¹³C MR detection. Use of the supplemental bandpass filter along the transmit-receive line greatly reduced severe gain compression effects on the ¹³C receiver, but resulted in an approximate 1-dB increase in transmit gain required to achieved a given flip angle, corresponding well to the measured filter insertion loss |S₂₁(32 MHz)| = −0.8 dB, and establishing a relatively small upper bound on possible signal-to-noise ratio penalty associated with filtering.

DISCUSSION

Our results show the clear feasibility of ¹H decoupling of HP ¹³C resonances with large J-couplings (¹J_CH = ~150 Hz) such as found in [2-¹³C]G3P. This was accomplished using a standalone ¹H decoupler system integrated within a clinical MRI scanner environment, in which ¹³C MR (let alone ¹H decoupling) is not commonly used. Customization of scanner hardware, including specialized RF coil design and use of supplemental filtering, was necessary to facilitate ¹H-decoupled ¹³C MR at the required power levels.

Application of ¹H decoupling clearly resulted in a major sensitivity boost for detection of the HP metabolic product [2-¹³C]G3P. This result exemplifies a new approach for enhanced detection of HP ¹³C metabolic products, applicable for cases in which a long-lived HP precursor without significant ¹H coupling (e.g., [2-¹³C]DHA) is transformed enzymatically in vivo into a HP metabolic product with major ¹H coupling (e.g., [2-¹³C]G3P).

Excessive power deposition (i.e., specific absorption rate (SAR)) has practically limited the applicability of broadband ¹H decoupling schemes for ¹³C MRS studies of humans (14). The SAR is a complicated function of RF coil and subject size, shape, and electrical properties, as well as RF pulse frequency, amplitude, and duty cycle (17). The massive sensitivity enhancement afforded by hyperpolarization could greatly reduce the cumulative power deposition required for acquisition of high-quality broadband decoupled ¹³C MR spectra. With traditional thermally polarized ¹³C experiments, the nuclear Overhauser effect enhancement associated with constant broadband decoupling generates up to three-fold signal-to-noise ratio gain for protonated carbons. Because the HP sample has a polarization far in excess of that produced by the nuclear Overhauser effect, our experiments were run in a gated decoupling mode (decoupling only during signal acquisition, or 5% duty cycle), significantly decreasing the SAR. The lowered SAR values should enable the application of these decoupling schemes in full field-of-view ¹³C MR studies in humans.

Although precise SAR calculations or measurements are difficult, relatively simple calculations illustrate important considerations regarding the potential for clinical translation of these methods. Under simplifying quasi-static assumptions, the total instantaneous power deposition in a homogenous sphere irradiated by a uniform, linearly polarized RF magnetic field is given by P_i = (π/15)(D/2)⁵σω²B², where D is the sphere diameter, σ is its RF conductivity, ω is the frequency of the RF field, and B is the RF field amplitude (18). For this study, we can infer that B ≈ 23 μT from the 90° pulse duration of 0.5 ms. Then, assuming that σ = 0.4 s/m in a spherical sample with D = 8 cm (i.e., loosely approximating a 300-g rat, similar to those used in this study), we can estimate that P_i ≈ 3 W. In comparison to the measured amplifier output power of 15 W required to drive the surface coil used in this study, this is not an unreasonable estimate after accounting for cable losses and reflected power, as well as the relatively large coil size. Most importantly, the power equation shows that P_i scales sharply with sample diameter. For a 25-cm human abdomen, the calculated P_i required is predicted to rise to approximately 900 W. However, the average power is lowered by a factor of 20 by the low 5% duty cycle required in the described HP acquisition, leading to a more reasonable value of approximately 45 W. Moreover, the short total scan time of HP studies also leads to low cumulative RF exposure. If, however, we assume a worst-case scenario that all 15 W supplied by the amplifier is deposited into the subject, the expected power requirement for abdominal scanning scales to potentially excessive power levels, whereas head scanning (e.g., D = 17 cm) could still be reasonable, even though the amplifier used in this study is limited to 50-W output as configured. For optimal efficiency, use of a circularly polarized decoupling field would be highly desirable, as this would be twice as efficient per unit SAR (19). A further consideration is that dielectric effects are also more important in larger human-sized samples, requiring more sophisticated SAR analysis (20). In summary, fully establishing the feasibility of these ¹H decoupling methods in humans will depend on more detailed SAR analysis, and careful SAR monitoring will be required for translation into human studies.

Given that the metabolic conversion from HP [2-¹³C]DHA to [2-¹³C]G3P in vivo is a promising new marker of hepatic energy status, which is relevant to studies of nonalcoholic fatty liver disease, ¹H-decoupled HP ¹³C MRI is of great interest for preclinical and potentially clinical studies of liver disease. Furthermore, the described development of an integrated system for simultaneous dual stimulation of ¹³C and ¹H nuclei in a clinical scanner environment could also facilitate in vivo application of other advanced NMR methods potentially useful for HP ¹³C MRI, such as ¹³C-¹H polarization transfer (21).