Increasing Infusion Times of Hyperpolarized [1‐ ¹³C]Pyruvate in a Mouse Brain Maintain Lactate Generation and Approach Pseudo‐Steady State Metabolism

ABSTRACT

Purpose

To investigate whether varying delivery times of hyperpolarized [1‐¹³C]pyruvate, enabled by the increased apparent T ₁ dissolved in deuterium oxide (D₂O), affects the observed kinetics of glycolytic brain metabolism in vivo.

Methods

Eighteen healthy mice were injected with 300 μL of hyperpolarized [1‐¹³C]pyruvate dissolved in D₂O at increasing injection times (15 s, 60s, 90s, 120 s). After collecting T₂‐weighted scans, slab dynamic ¹³C MRS data were acquired. Time‐course curves of [1‐¹³C]pyruvate, [1‐¹³C]lactate and [1‐¹³C]lactate/total carbon ratio were calculated. Mean full‐width‐half‐maximum (FWHM) and area‐under‐curve (AUC) values were compared across injection times. A simplified one‐compartment model of pyruvate metabolism was fit using the conversion rate constant (k _PL) and effective lactate decay rate (R _1eff ). Dynamic EPSI images, acquired using an injection time of 15 s and 60 s for comparison.

Results

The mean FWHM values of the time‐course curves of [1‐¹³C]pyruvate and [1‐¹³C]lactate showed a significant increase (p < 0.01) with increasing injection times, while no statistical significance was found across the AUC values. The time‐course curves of lactate/total carbon ratio showed elongated plateaus with increased injection times. Kinetic modeling showed good agreement between fitted and acquired lactate data, with AUC of normalized lactate profile remaining constant across infusion times. Dynamic EPSI images acquired with a longer infusion time (60 s) showed the ability to monitor brain metabolism as it approached pseudo‐steady state.

Conclusions

Increased delivery times of hyperpolarized [1‐¹³C]pyruvate dissolved in D₂O approaches pseudo‐steady state metabolism in vivo and allowing for the potential to cater new acquisition and reconstruction approaches for enhanced imaging.

1. Introduction

In vivo hyperpolarized (HP) ¹³C MRI [1, 2] is a powerful molecular imaging technique [3, 4], which allows real‐time investigation of specific‐metabolic pathways [5], giving unique information in vivo. It has been applied primarily in oncology [6, 7, 8, 9, 10, 11, 12, 13] and more recently in non‐oncologic neurologic disorders [14, 15, 16, 17, 18] owing to its ability to detect aberrant metabolism. The hyperpolarized ¹³C MRI methodology relies on the enhancement of the hyperpolarized ¹³C MR signal, by several order of magnitude [1] leading to a significant increase in the ¹³C MR sensitivity compared to conventional ¹³C MRI methods. The lifetime of the hyperpolarized MR signal in vivo is dependent on the longitudinal relaxation time (T ₁) of the hyperpolarized ¹³C‐enriched probe, as well as the sampling rate (due to radiofrequency (RF) excitation dependent signal loss) and the concurrent lifetime changes due to metabolic conversion. After dissolution, there is a limited window of time to detect the ¹³C hyperpolarized signal, typically within 2–3 T₁s of the substrate [19]. During this short period of time, the hyperpolarized probes must be transferred to the scanner, intravenously injected, perfused to the organ, and after cellular uptake, which leads to metabolic changes, detected by fast magnetic resonance spectroscopic imaging (MRSI) sequences. A sufficiently long T ₁ of the ¹³C‐enriched probes in biological fluids is a key factor to accomplish the hyperpolarized experiment [3]. The longer the ¹³C T ₁ of the substrate, the better its distribution to the target organ visualized before the polarization decays with capability to perform optimal metabolic imaging acquisitions.

Among the ¹³C‐enriched probes used in hyperpolarized MRI, [1‐¹³C]pyruvate, is currently the most commonly studied substrate in both preclinical and clinical settings due to its high achievable polarization, central position to numerous metabolic pathways [9], and relatively long ¹³C T ₁ in water at clinical magnetic field strength (67 s at 3 T) [19]. Due to the T ₁ limitation of ¹³C‐enriched probes dissolved in water, a bolus of the hyperpolarized substrate, such as [1‐¹³C]pyruvate (∼60 s), is typically injected intravenously at a rapid rate (on the order of 10–15 s in mice) [20, 21]. After injection, [1‐¹³C]pyruvate is rapidly distributed throughout the body and is metabolized within a short time after its injection [5]. The time window of investigation is therefore bounded by the T ₁ relaxation of the substrate and product and results in a dynamic non‐steady state. Fast ¹³C MRSI sequences [22], such as the echo planar spectroscopic imaging (EPSI) [23, 24, 25] and ¹³C imaging approaches such as echo planar imaging (EPI) with spectral‐spatial (spsp) excitation [26, 27, 28, 29], are required to image metabolic changes. These approaches still typically sample a non‐steady state curve thus in between each imaging acquisition the ratio of substrate to product changes significantly, making signal averaging difficult and also heavily weighting the imaging to the window used for acquisition relative to the bolus.

In the recent years, several strategies [19, 30, 31] have been employed to increase the lifetime of the apparent T ₁ relaxation time of a nucleus of interest in ¹³C‐enriched probes, with the aim of obtaining a wider time window for in vivo metabolic investigation by ¹³C MR imaging. To elongate the apparent T ₁ of ¹³C‐enriched substrates, a practical approach, consisting of dissolving HP probes in deuterium oxide (D₂O), has been used in preclinical studies, both in vivo [32, 33] and in vitro [34, 35], showing a dramatic increase in the apparent T ₁ of the substrate. The same approach was used in a clinical study [30] where dissolution of HP [1‐¹³C]pyruvate in D₂O led to an extension of ¹³C T ₁ both in solution (92 s at 3 T) and in vivo in humans, while the apparent T ₁ in D₂O was found to be increased 1.4‐fold compared to that obtained in water. This suggests that given an extended time window, more can be done to reimagine this experiment to visualize differential in vivo kinetics.

With increased ¹³C T ₁ of ¹³C‐enriched probes dissolved in D₂O, the time window of investigation can be increased. The aim of this study was therefore to determine whether the prolonged T ₁ of [1‐¹³C]pyruvate facilitates an extended time window in vivo to appreciate near steady‐state kinetics in the murine brain. A variable infusion pump approach was developed to deliver the hyperpolarized bolus intravenously with increasing infusion times, while dynamic ¹³C MR spectroscopic scans were acquired to monitor glycolytic metabolism in the murine brain in real time. Dynamic spectroscopic data were simulated with a simplified one‐compartment kinetic model of pyruvate‐to‐lactate metabolism. Additionally, dynamic echo‐planar spectroscopic images were acquired at the intermediate infusion time of 60 s to confirm the approximation of pseudo‐steady state metabolism in the mouse brain, by extending the time window of the investigation.

2. Methods

2.1. Hyperpolarization of [1‐ ¹³C]Pyruvate

Isotopically enriched [1‐¹³C]pyruvic acid (40 μL; Sigma‐Aldrich, St Louis, MO, USA) doped with 15 mM of AH111501 trityl radical (GE HealthCare, Chicago, IL USA) was polarized by dissolution dynamic nuclear polarization (dDNP) for 1–2 h in a SpinLab polarizer (GE, Healthcare) operating at 5 T and 0.8 K with microwave irradiation at 139.88 GHz. Following polarization, the frozen sample was rapidly dissolved in 8.5 mL of TRIS buffer (100 mM TRIS, 1 mM EDTA, pH = 7.3) in D₂O and collected in a flask containing a solution of 10 M sodium hydroxide in D₂O (51 μL; Sigma‐Aldrich) to neutralize pyruvate and reach a pH of ∼7.3.

Following dissolution, ∼700 μL of hyperpolarized [1‐¹³C]pyruvate was transferred to a 5‐mm NMR tube to perform ¹³C NMR spectroscopy using a benchtop 1 T ¹³C NMR spectrometer (Spinsolve, Magritek, Wellington, NZ). All spectra were acquired using a single‐pulse‐and‐acquire sequence with 67 μs block pulse, 2.5 kHz spectral width (SW) and 4096 points. The longitudinal relaxation time (T₁) of [1‐¹³C]pyruvate was measured by a dynamic acquisition (5° flip angle, TR = 5 s, 60 repetitions, 1 average, 5 min scan time) and subsequently estimated by a mono‐exponential fit with the resulting value corrected for the flip angle.

Thermal polarization was measured by acquiring an average spectrum (90° flip angle, TR = 10 s, 256 average, 43 min scan time) and the calculated values were corrected for the flip angle and apparent T₁. The concentration of [1‐¹³C]pyruvate was measured on a 14.1 T Bruker AVANCE III vertical bore NMR spectrometer equipped with a multi‐nuclear broadband observe probe (Bruker BioSpin, Billerica, MA, USA) in presence of 1 mM Gd‐DOTA (6 μL) and 100 mM [1‐¹³C]lactate standard (100 μL) added to the hyperpolarized [1‐¹³C]pyruvate solution (500 μL).

2.2. Animals

Female BALB/cJ mice (n = 18; Jackson Laboratories, Bar Harbor, ME, USA) aged 12–16 weeks and weighting 21.07–26.17 g were used in this study. Mice were housed (groups of five per cage) with free access to food and water, a 12‐h light/dark cycle, temperature of 18°C–23°C and humidity of 40%–60%. All animal experimental procedures were conducted under the approval of the Institutional Animal Care and Use Committee of Memorial Sloan Kettering Cancer Center in accordance with the standards for animal care and use, as set by the Federal law under the Animal Welfare Act. Before imaging, each mouse underwent a tail vein cannulation using a 23‐gauge rodent tail‐vein catheter (Braintree Scientific, Braintree, MA, USA) filled with 1 units/mL of heparin (Fresenius Kabi, Lake Zurich, IL, USA) in 0.9% of Sodium Chloride. Subsequently, it was anesthetized through inhalation of a mixture of isoflurane (induction: 3% for 2 min; regime: 1%–2%) and oxygen delivered at a rate of 0.5–1 L/min. Anesthetized mice were laid prone on a dedicated holder that ensured minimal motion during scans. Respiration rate and body temperature were monitored using an MR‐compatible system (Model 1025, SA Instruments Inc., Stony Brook, NY, USA) and maintained at 85 ± 20 breaths/min and 37°C ± 0.5°C by regulating isoflurane concentration and circulating warm water system, respectively. Approximately 10 s before dissolution, the oxygen flow was increased to 1.5–2 L/min, while isoflurane was reduced to 0.5%–1% to gradually awaken the mouse. Then, approximately 5 s before the start of the probe injection, isoflurane was completely turned off and kept in this status for the entire duration of the ¹³C hyperpolarized scan.

Immediately after the hyperpolarized scan (∼1–5 s), the anesthetic regime was restored until the end of the MRI session.

2.3. In Vivo Magnetic Resonance Imaging and Spectroscopy

All MR experiments were performed on a Bruker BioSpec 3.0 T/18 cm horizontal bore system (Bruker BioSpin, Billerica, MA, US) equipped with a high‐performance gradient set (max gradient strength: 900 mT/m; slew‐rate: 4200 mT/m/ms). A dual‐tuned ¹H/¹³C transmit/receive RF volume coil (30‐mm inner‐diameter; Bruker BioSpin MRI GmbH, Ettlingen, Germany) was used. After collecting scout images to assess the positioning of the mouse brain in the magnet isocenter, multi‐slice T₂‐weighted fast spin‐echo (TSE) images of mouse brain were acquired in the coronal orientation as an anatomical reference (repetition‐time (TR)/effective‐echo‐time (TE_eff) = 2500/33 ms, echo train length = 8, field‐of‐view = 32 × 32 mm², matrix size = 128 × 128, in‐plane resolution of 250 × 250 μm², 15 slices of 1 mm thickness covering the entire mouse brain, 2 averages, scan time of 100 s). Static magnetic field homogeneity was optimized using a localized field‐map shimming routine (ParaVision 360, v2.0 pL 1, Bruker BioSpin), resulting in water linewidths of 15–23 Hz over the selected volume of the mouse brain. Approximately 20–30 s before the start of the hyperpolarized ¹³C MR spectroscopic imaging (MRSI) scan the isoflurane was turned off to acquire the hyperpolarized scan on the awake mouse. For quantification purposes, a cylindrical phantom (radius = 0.5 cm, height = 2 cm) containing a 6 M ¹³C‐labeled urea solution was placed adjacent to the mouse brain with the cylindrical axis parallel to the z‐axis of the scanner to be included in the field‐of‐view of T₂‐weighted and ¹³C MRSI scan.

2.3.1. In Vivo ¹³C NMR Spectroscopy

Based on the increased T ₁ of [1‐¹³C]pyruvate dissolved in D₂O [30], four injection times (15, 60, 90, 120 s) were explored to evaluate whether increasing the infusion time of the hyperpolarized [1‐¹³C]pyruvate dissolved in D₂O, could lead to appreciating near‐steady‐state kinetics in vivo. Three independent experiments were performed for each infusion time. A syringe pump (Thermo Fisher Scientific, Waltham, MA, US) equipped with a custom‐built connection to the mouse catheter was used to inject 300 μL of hyperpolarized [1‐¹³C]pyruvate at each injection time. To ensure accurate substrate's delivery within the considered interval time (15, 60, 90, 120 s), syringe pump's flow rates were preliminarily determined for each injection time. Following the mouse setup and preliminary scans as described above, dynamic changes in mouse brain glycolytic metabolism were investigated using a ¹³C non‐localized spectroscopic sequence acquired in the coronal orientation (128 μs block pulse at 10° flip angle, 1.3 kHz spectral width (SW), 4096 points, 1 average, 80 repetitions, TR = 3072 ms, 15 mm slice‐thickness to cover the whole brain, scan time of 4 min). Immediately after dissolution, the start of the ¹³C dynamic scan was synchronized with the start of the substrate's injection to evaluate the dynamic time course of the hyperpolarized substrate and downstream metabolite production as a function of the injection time. Three independent experiments (n = 3) were performed for each injection time on BALB/cJ mice for a total of 12 mice.

2.3.2. In Vivo ¹³C Dynamic Echo Planar Spectroscopic Imaging

To assess whether increasing the substrate injection time could enhance ¹³C hyperpolarized MRSI acquisition in vivo, six independent experiments were performed by dynamically acquiring real‐time spectroscopic images of a mouse brain while monitoring the substrate injection and the downstream metabolites production. The same volume of [1‐¹³C]pyruvate (300 μL) was injected over a 60 s time (n = 3) and the resulting images compared to those obtained when injecting the substrate during the routinely used 15 s rapid injection (n = 3). Following the mouse positioning and shim optimization, the hyperpolarized ¹³C Echo Planar Spectroscopic Imaging (EPSI) scan was acquired in the coronal orientation of the mouse brain (TR/TE_eff = 250/3.536 ms, flip angle of 15°, 75 kHz bandwidth, field‐of‐view = 32 × 32 mm², image size 12 × 12, in‐plane resolution of 2.667 × 2.667 mm², 1 slice of 15 mm thickness covering the whole brain, 1 average, 80 repetition, scan time of 4 min). Immediately after [1‐¹³C]pyruvate dissolution and setting up the infusion pump, the start of the mouse intravenous injection was synchronized with the start of the ¹³C hyperpolarized MRSI scan, as performed for the dynamic spectroscopic experiments. To normalize the ¹³C signal of the substrate and downstream metabolites in the mouse brain, the ¹³C EPSI scan was repeated to image the 6 M ¹³C‐labeled urea phantom after optimizing the homogeneity of the magnetic field over the phantom volume.

2.4. Data Analysis

2.4.1. In Vivo ¹³C NMR Spectroscopic Data Analysis

All ¹³C NMR spectra were analyzed using MestReNova software (Mestrelab Research, S.L., Santiago de Compostela, Spain, v14.3.1‐31 739). Prior quantification of hyperpolarized [1‐¹³C]pyruvate and downstream metabolites in a mouse brain, each HP ¹³C spectrum was pre‐processed by manually correcting first‐ and second‐order phase, and baseline [36]. Quantification of NMR spectral peak areas was performed in time‐domain for each time point of the metabolite time course curve per experiment (n = 3 per injection time). The quantified value of each metabolite in the mouse brain was normalized to the 6 M ¹³C‐labeled urea phantom, which was used as a reference. Time course curves of [1‐¹³C]pyruvate and downstream metabolites were calculated for each mouse (n = 3) per infusion time (n = 4). Due to the low signal‐to‐noise ratio of [1‐¹³C]pyruvate hydrate, [1‐¹³C]alanine and [1‐¹³C]bicarbonate at longer injection times (60, 90, 120 s), only time course curves of the average integral signals (n = 3) per time point of [1‐¹³C]pyruvate and [1‐¹³C]lactate were computed for comparison between the different infusion times. To assess differences between the metabolite dynamic state when increasing the injection times, the full width at half maximum (FWHM) values of [1‐¹³C]pyruvate and [1‐¹³C]lactate time course curves were determined for each animal and injection time, determining the width of each time course curve at half the maximum value of the curve. For each injection time, mean FWHM values of [1‐¹³C]pyruvate and [1‐¹³C]lactate were then calculated. Additionally, area under the curve (AUC) values of [1‐¹³C]pyruvate and [1‐¹³C]lactate time course curves were calculated for all mice (n = 12) by summing the integral signal of each time point over time. For each injection time, mean AUC values of [1‐¹³C]pyruvate and [1‐¹³C]lactate were calculated for comparison across different injection times. For each injection time, time course curves of the ratio between the average integral signals of [1‐¹³C]lactate and total carbon, derived from HP pyruvate, were also calculated to further evaluate metabolite dynamic changes over time.

2.4.2. Kinetic Modeling and k _PL Determination

Kinetic rate modeling was performed jointly on all experimental pyruvate and lactate datasets (n = 12) using a custom MATLAB script (Matlab R2023b, The MathWorks). The pyruvate and lactate signals were normalized to “total carbon” as $M_{\mathit{Pyr},\mathit{Lac}}=M_{\mathit{Pyr},\mathit{Lac}}/\left(M_{\mathit{Pyr}}+M_{\mathit{Lac}}\right)$.

A simplified one‐compartment kinetic model of pyruvate to lactate metabolism was modeled as $d{M}_{\mathit{Lac}}(t)/\mathit{dt}=k_{\mathit{PL}}{M}_{\mathit{Pyr}}(t)-R_{1\mathit{eff}}{M}_{\mathit{Lac}}(t)$, where $k_{\mathrm{PL}}$ is the conversion rate constant from [1‐¹³C]pyruvate to [1‐¹³C]lactate and R _1eff is the effective lifetime decay rate ($R_{1\mathit{eff}}=T_{1e\mathrm{ff}}^{-1}$). For simplicity, the backward rate constant, $k_{\mathrm{LP}}$, was set to zero. The in vivo [1‐¹³C]lactate T ₁ was assumed to be 25 s [37] from the observed value of 45 s. The effective lifetime, accounting for flip angle losses, was initialized as $R_{1\mathit{eff}}=T_{1\mathit{eff}}^{-1}-(\ln \mathit{\text{cosθ}})T{R}^{-1}$ where $\theta$ is the effective flip angle and TR is the repetition time. The continuous pyruvate signal was interpolated and the differential equation solved. The kinetic model was fitted to the observed lactate signal using least‐squares regression method, with two parameters, the conversion rate constant k _PL and an R _1eff value, which additionally takes into account other metabolism and transport/export mechanisms. Confidence intervals were obtained by bootstrapping and sampling combinations of all datasets (n = 12).

2.4.3. In Vivo ¹³C 2D Dynamic Echo Planar Spectroscopic Imaging Analysis

The 2D EPSI data were processed using a custom MATLAB script. The acquired data were first re‐cantered in k‐space to smooth out spatial first‐order phase shifts and zero‐filled to twice the size. Exponential line broadening (2.5 Hz) was applied to the spectral dimension, prior to inverse Fourier transformation of the spectral and spatial dimensions. Brain masks were used to filter brain tissues from the T ₂‐weighted images, which were manually registered and resampled into the Waxholm mouse brain atlas [38] using 3D slicer [39] (https://www.slicer.org/). The matrix‐pencil method was used to identify the signal poles of the first five components [40] of a mean spectrum averaged over the spatial and temporal dimensions. Subsequently, the amplitude and phase of each component were fitted by least‐squares and Moore‐Penrose pseudo‐inverse algorithm in each spatial and temporal point. Pyruvate and lactate amplitudes were obtained by summing phase‐adjusted amplitudes within 2 ppm of 170.1 and 183 ppm, respectively. Time courses of normalized lactate/total carbon ratios were calculated up to 72 s and grouped in 18‐s intervals to compare the dynamic following the substrate injection at 15 s and 60 s. Kinetic rate modeling was performed to generate $k_{\mathit{PL}}$ maps using a simplified one‐compartment model as in the slab‐dynamic data. In this case the effective lactate lifetime constant was fixed to account for the expected R₁ decay and flip angle losses, $R_{1\mathit{eff}}=T_{1\mathit{eff}}^{-1}-(\ln \mathit{\text{cosθ}})\mathit{PE}·T{R}^{-1}$, where PE is the number of phase encodes. Resampled metabolic maps were overlaid on the corresponding T ₂‐weighted images.

2.5. Statistical Analysis

Statistical analysis was performed using GraphPad Prism version 10.0.3 (GraphPad Software, La Jolla, CA, United States). Values are reported as mean ± standard error (SE) for each time point of all time course curves, the FWHM and AUC data determined for [1‐¹³C]pyruvate and [1‐¹³C]lactate per infusion time. Statistically significant differences between the mean values of both FWHM and AUC across injection times, calculated for the averaged time course curves, were determined using the nonparametric Kruskal–Wallis test corrected for multiple comparisons using Dunn's hypothesis test. The same statistical analysis was used to compare the mean values of lactate/total carbon ratios across the 18‐s time intervals in which 2DEPSI data acquired for rapid (15 s) and longer (60 s) injections were grouped. Statistical differences in the mean total metabolite signal, both pyruvate and lactate, obtained from the corresponding 2D brain metabolic maps calculated for injection times of 15 and 60 s were considered using a two‐tailed unpaired, nonparametric Mann–Whitney test. The same statistical analysis was applied to the mean k _PL values obtained from the 2D brain k _PL maps calculated for injection times of 15 and 60 s. Means and 95% confidence intervals for kinetic parameters were obtained via bootstrapping statistical procedure.

3. Results

3.1. Hyperpolarization of [1‐ ¹³C]Pyruvate in Solution

A mean apparent T₁ of [1‐¹³C]pyruvate, measured at 1 T was determined as (130 ± 1) s. The polarization values of [1‐¹³C]pyruvate in D₂O solution, measured at 1 T, were found in the range of 30% ± 6% while the concentration values, measured at 14.1 T were found between 95 ± 15 mM.

3.2. Dynamic In Vivo ¹³C NMR Spectroscopic Data From the Murine Brain

A graphical experimental design of the study is illustrated in Figure 1, where the experimental timeline and setup applied to each mouse (Figure 1A) are displayed together with the sagittal view of a mouse brain showing the slice‐selective excitation slab acquired in all dynamic scans (Figure 1B). The evolution of the substrate and downstream metabolites in a mouse brain observed when injecting the hyperpolarized probe over a longer time interval (e.g., 120 s) is illustrated in Figure 1C, where a representative ¹³C hyperpolarized dynamic scan is shown. The dynamic response of brain metabolism in real‐time, was investigated by injecting the same volume of hyperpolarized [1‐¹³C]pyruvate (300 μL) at extended time intervals (60, 90, 120 s), compared to the rapid injection time of ∼15 s routinely used [20, 21], thus ensuring the same amount of hyperpolarized pyruvate delivered to each mouse per injection time.

FIGURE 1.

Dynamic MRS of hyperpolarized [1‐¹³C]pyruvate in a mouse brain when investigated with increasing injection times. Graphical experimental design (A) of the study showing timeline and setup of the experiment. Sagittal view of a mouse brain (B) with relative slice selective excitation (in red; 15 mm thick, covering the murine brain). Evolution of metabolic substrate and products (C) observed in a murine brain when 300 μL of hyperpolarized [1‐¹³C]pyruvate were injected over 120 s, at a constant rate, using a syringe pump. A representative ¹³C NMR spectrum is also shown (in blue) with all metabolites annotated.

In Figure 2, the average time course curves of [1‐¹³C]pyruvate (Figure 2, left column) and [1‐¹³C]lactate (Figure 2, right column), determined for each injection time (15, 60, 90, 120 s), are shown. With increasing injection time from 15 s to 120 s, the average time course curves of both [1‐¹³C]pyruvate and [1‐¹³C]lactate visibly show a reduction in their maximum integral signal along with an increased pseudo‐steady‐state level plateau over time of their normalized integral signals (Figure 2). Interestingly while the maximum normalized pyruvate signal significantly decreases (Figure 2C,E,G), the maximum lactate signal is maintained from 60s to 120 s infusion times (Figure 2D,F,H), suggesting that the metabolic system is already saturated at lower pyruvate concentrations delivered to the brain. The mean AUC values of the time course curves of [1‐¹³C]pyruvate and [1‐¹³C]lactate determined for each infusion time are shown in Figure 3A,B, respectively. No statistically significant differences were found between the mean AUC calculated at different injection times either for pyruvate (Figure 3A) or lactate (Figure 3B), indicating that the mean of the metabolite signal intensities, each weighted by its corresponding polarization and summed over time, is approximately the same in the mouse brain for each injection time. This further supports the notion that, although the maximum HP pyruvate signal is lower as the infusion time increases, it is longer maintained while the amount of HP is not reduced.

FIGURE 2.

Time course curves of hyperpolarized [1‐¹³C]pyruvate and [1‐¹³C]lactate in a mouse brain obtained by increasing the substrate injection times using a syringe pump. Time‐course curves of hyperpolarized [1‐¹³C]pyruvate (left panel) and [1‐¹³C]lactate (right panel) obtained by injecting the substrate in 15 s (A, red circle), 60 s (C, black square), 90 s (E, plum hexagon) and 120 s (G, cyan rhombus). The corresponding lactate generation curves are shown in B (15 s, red circle), D (60 s, black square), F (90 s, plum hexagon) and H (120 s, cyan rhombus), respectively. For each mouse, the [1‐¹³C]pyruvate and [1‐¹³C]lactate signals were normalized to the signal from the 6 M‐urea phantom, which was placed adjacent to the mouse brain during acquisition, and corrected for the calculated polarization of [1‐¹³C]pyruvate. Data are mean ± SE, n = 3 for each injection time.

FIGURE 3.

Histogram plots of AUC (areas under the curve) and FWHM (full width at half maximum) values of [1‐¹³C]pyruvate and [1‐¹³C]lactate time course curves. Mean AUC (top panel) and mean FWHM (bottom panel) values calculated for each infusion time (15 s, red bar; 60 s, black bar; 90 s, plum bar; 120 s, cyan bar) from the time‐course curves of hyperpolarized [1‐¹³C]pyruvate (A, C) and [1‐¹³C]lactate (B, D), respectively. The AUC units are expressed as time multiplied by an arbitrary unit (a.u.) given by the hyperpolarized ¹³C signal normalized to the ¹³C signal of the urea phantom (6 M) used as a reference. Data are mean ± SE, n = 3 for each injection time. Statistic: Nonparametric Kruskal–Wallis's test followed by Dunn's post hoc test.

The mean FWHM values calculated for the time course curves of [1‐¹³C]pyruvate and [1‐¹³C]lactate at each injection time are shown in Figure 3C,D, respectively. Compared to the rapid injection time of 15 s (Figure 3C,D; red bar), the mean FWHM values determined for each time course curve showed an increase with increasing injection time at 60, 90 and 120 s for both [1‐¹³C]pyruvate (Figure 3C) and [1‐¹³C]lactate (Figure 3D), which was significant for 90 and 120 s infusion times (Figure 4C,D). This indicates an increased steady level of lactate production in the mouse brain, balanced with delivery and T₁‐mediated decay, depending on the substrate infusion time, which intrinsically affects the time courses of the delivered pyruvate signals in the mouse brain. Additionally, no significant difference between FWHM values calculated for consecutive, longer infusion times (i.e., 60 to 90s, and 90 to 120 s) was found for lactate production, which showed a significant difference only between 60 and 120 s.

FIGURE 4.

Time course curves of hyperpolarized [1‐¹³C]lactate/total carbon ratios. Graphs showing mean time‐course curves of lactate/total carbon ratios determined for each injection time, 15 s (A; red circle), 60 s (B; black square), 90 s (C; plum hexagon) and 120 s (D; cyan rhombus). Data are mean ± SE, n = 3 per infusion time.

The time course curves of the [1‐¹³C]lactate/total carbon ratio, calculated for each infusion time, are shown in Figure 4. By increasing the infusion time, the observed lactate/total carbon ratio was found to reach an increasing plateau (Figure 4B–D.) over time, which approaches a pseudo‐steady state due to the temporary balance of multiple concurrent mechanisms, including, for example, pyruvate delivery, reductive metabolism to lactate, other metabolic and transport pathways of the substrate and its products, and signal decay due to both RF excitation and T ₁ relaxation.

3.3. Kinetic Modeling and k _PL Determination in the Murine Brain

A simplified one‐compartment model was used to describe the dynamics of hyperpolarized [1‐¹³C]pyruvate and [1‐¹³C]lactate in a mouse brain when the ¹³C‐labeled substrate was infused with variable infusion times (15, 60, 90, 120 s). The experimental time‐course curves of hyperpolarized [1‐¹³C]pyruvate and [1‐¹³C]lactate together with those obtained by fitting the experimental [1‐¹³C]lactate signal with the simplified one‐compartment kinetic model are shown for each dataset (n = 12) and infusion time (15 s, red; 60 s, black; 90 s, plum; 120 s, cyan) in Figure 5A–C, respectively. Experimental [1‐¹³C]lactate data were fit with the simplified kinetic model using the pyruvate‐to‐lactate kinetic rate constant, k _PL set to 0.047 s⁻¹ (95% CI = 0.038, 0.078) and the effective lifetime rate of lactate, R _1eff equal to 0.092 s⁻¹ (95% CI = 0.057, 0.11), which was higher than the initialized value set to 0.045 s⁻¹, while confidence intervals were estimated via a bootstrapping filter. Lactate fitted data (Figure 5C) showed a good agreement with the experimental lactate time‐course curves (Figure 5B). The AUC values of lactate time‐course curves, fitted to the simplified kinetic model, determined for each mouse and infusion time, and normalized to the corresponding total carbon values, are shown in Figure 5D. The normalized lactate AUC values (Figure 5D) were approximately constant across all infusion times and mice, thus confirming previous results (Figure 3B) and supporting the ability to leverage slower infusion times to reach a pseudo‐steady state without compromising total lactate produced.

FIGURE 5.

Fitting of one‐compartment model of pyruvate and lactate dynamics. Time‐course curves of [1‐¹³C]pyruvate (A) and [1‐¹³C]lactate (B), calculated for each mouse. Red, black, plum, and cyan color codes refer to mice (n = 3) injected for 15, 60, 90 and 120 s, respectively. The time‐course curves of [1‐¹³C]lactate (C) fitted with the forward kinetic rate constant, k _PL = 0.047 s⁻¹ (95% CI = 0.038, 0.078), and effective lifetime rate for lactate R_1,eff = 0.092 s⁻¹ (95% CI = 0.057, 0.11). (D) Area under the curve (AUC) of lactate time‐course curves for all experiments (n = 12), each of them normalized to total carbon.

3.4. In Vivo ¹³C 2D Dynamic Echo Planar Spectroscopic Imaging in the Murine Brain

To extend this approach to resolving brain metabolism specifically, we acquired dynamic imaging using a time resolved dynamic 2D EPSI scan. We compared infusion times of 15 s, which is routinely used [20, 21], and 60 s given the findings from our slab dynamic studies. Representative total brain dynamic curves of pyruvate (Figure 6A; dashed lines), lactate (Figure 6A; solid lines) and lactate fitted with one‐compartment model (blue dotted lines) are shown for infusion times of 15 s (Figure 6A, left panel; red lines) and 60 s (Figure 6A, right panel, black lines), respectively. Compared to experiments conducted with a rapid injection (15 s; Figure 6A, left panel), the metabolite dynamic curves show a plateau‐like condition for the longer infusion experiments (60 s; Figure 6A, right panel), confirming that this behavior occurs specifically in the murine brain.

FIGURE 6.

2D brain metabolic maps of pyruvate and lactate dynamics. Representative dynamic time course of [1‐¹³C]pyruvate (dashed line) and [1‐¹³C]lactate (solid line) with simplified one‐compartment k _PL fit (blue dotted line) for infusion times of 15 s (A, left panel; red lines) and 60 s (A, right panel; black lines). Representative color overlay spatial metabolic maps of the area under the curve (AUC) of [1‐¹³C]pyruvate (B, left panel) and [1‐¹³C]lactate (B, right panel) on the corresponding coronal T₂‐w anatomical image of a mouse brain, investigated by dynamic 2D EPSI scans when the substrate was infused in 15 s (B, left panel) and 60 s (B, right panel). Spatial metabolic maps are normalized to the highest pyruvate signal. Bar plots showing the mean AUC values of total pyruvate (C; n = 3) and lactate (D; n = 3) for both injection times, 15 s (red cycles and bar) and 60 s (black square and bar). Comparison of the time‐course of normalized lactate/total carbon ratios, grouped in 18‐s intervals, shown up to 72 s after the start of substrate injection for both injection times, 15 s (6E, left panel; red bars) and 60 s (6E, right panel, black bars). Data are (mean ± SD), n = 3 per infusion time. Averaged values of normalized lactate/total carbon ratios (F) over the 18‐s intervals up to 72 s (n = 4 intervals) and corresponding averaged standard deviations (G, n = 4) shown for 15 s (red circles and bars) and 60 s (black square and bars). Representative k _PL maps determined by a simple one‐compartment model quantify spatial pyruvate‐to‐lactate metabolism for infusion times of 15 s (H, top panel) and 60 s (H, bottom panel). Bar plot showing k _PL values (I) for infusion times of 15 s (red circles and bar) and 60 s (black squares and bar). To facilitate comparison between the two strategies (15 vs. 60 s), all pyruvate, lactate and k _PL maps are scaled to 1, 0.25 and 0.075, respectively. All data are mean ± SE (n = 3 per infusion time) except (F and G) where they are mean ± SD. Statistic: Two‐tailed, unpaired Mann–Whitney test; p < 0.05; Kruskal–Wallis's test with multiple comparison; p < 0.05.

Representative spatial metabolic maps calculated from the AUC of the hyperpolarized [1‐¹³C]pyruvate and [1‐¹³C]lactate time‐course curves are illustrated for substrate infusion times of 15 s (Figure 6B, left panel) and 60 s (Figure 6B, right panel). Pyruvate maps showed a relatively higher amplitude for rapid infusion times of 15 s (Figure 6B, left panel) than for longer infusion times of 60 s (Figure 6B, right panel), as confirmed by the comparison of mean AUC values of total pyruvate maps calculated for infusion times of 15 s (Figure 6C, red bar) and 60 s (Figure 6C, black bar). However, the amplitude of lactate spatial maps for both infusion times, 15 s (Figure 6B, right panel) and 60 s (Figure 6B, right panel) was approximately the same, as also shown by the corresponding bar plot (Figure 6D) showing the comparison of mean lactate AUC between the two infusion times 15 s (red bar) and 60 s (black bar). The corresponding k _PL maps (Figure 6H) deriving by the simplified one compartment model are illustrated for infusion times of 15 s (Figure 6H, top panel) and 60 s (Figure 6H, bottom panel). The k _PL maps (Figure 6H) underscore the increased relative conversion of pyruvate to lactate in the longer infusion experiments (60 s; Figure 6H, bottom panel), as also illustrated by the trend shown in the related bar plot (Figure 6I).

Time‐courses of normalized lactate/total carbon ratios, calculated for each injection time and grouped in 18‐s intervals, are shown up to 72 s after the start of the substrate injection (Figure 6E). While the time‐course of lactate/total carbon ratios for 60 s injection time (Figure 6E right panel; black bars) shows a plateau where no statistically significant differences were found between the mean values across the 18‐s intervals (n = 4), a significant increase was observed across the corresponding time intervals when the substrate was rapidly injected in 15 s (Figure 6E, left panel; red bars), thus suggesting a transient state metabolism for rapid injection time (15 s) and an approximation to a pseudo‐steady state of brain metabolism for the longer 60 s injection. The averaged lactate/total carbon ratios calculated over the time intervals (n = 4) (Figure 6F) for both injection times along with their averaged standard deviation (Sd) (Figure 6G), show increased values for rapid injections (15 s) compared with longer times (60 s: Figure 6F,G) further supporting a transient‐state of brain metabolism for rapid substrate injection compared with a more steady brain metabolism for longer substrate injections.

4. Discussion

Hyperpolarized ¹³C MRI acquisitions are usually performed by injecting the ¹³C‐labeled substrate as rapidly as possible after dissolution (∼10–15 s) [20, 21], due to the relatively short ¹³C T ₁ relaxation time of the substrate [3] and the need to achieve high signal‐to‐noise ratios. As a result, the observed metabolic kinetics are influenced by the bolus type infusion of the substrate at supra‐physiologic conditions. By increasing the T ₁ of hyperpolarized [1‐¹³C]pyruvate, through dissolution in D₂O, we were able to explore increased infusion times, which allowed instantaneously lower (Figure 2A,C,E,G) cumulatively similar ¹³C‐labeled substrate concentrations than those obtained by rapid injections (Figure 3A), to investigate brain metabolism in real time [22, 41, 42]. In contrast to other methods, namely variable flip angle techniques [43], which maintain constant signal amplitude throughout the experiment, our approach aims to maintain constant delivery of the substrate (300 μL injected for each infusion type) over a longer period of time.

Compared to the transient state of brain metabolism that occurs when the HP [1‐¹³C]pyruvate bolus is injected rapidly (15 s), we observed a plateau of the substrate delivery curve (Figures 2, 3, 4), approaching a pseudo‐steady state with increasing injection times up to 120 s (Figure 4). Dynamically acquired data showed robust ¹³C acquisitions throughout the duration of the substrate infusion and until complete depletion of the ¹³C signal. This was further confirmed by the 2D dynamic data (Figure 6E–J) showing an increased plateau (Figure 6E) approaching a pseudo‐steady state of brain metabolism for a longer infusion time (60 s). While our 1D dynamic data of the mouse head showed similar lactate AUC between the different infusion times (Figure 3B), the 2D dynamic data of the mouse brain showed a trend toward increased effective metabolic conversion of pyruvate to lactate at longer infusion times (60 s; Figure 6H,I), although lactate AUC values (Figure 6B–D) were comparable. This observed trend is most likely the result of the differential number of excitations with respect to the bolus dynamics. For the purposes of the comparison in this work, we kept the excitation flip angle and number of excitations (16 per EPSI image) constant resulting in a more significant excitation of the pyruvate signal in the faster bolus as opposed to the slab dynamic excitations where the effect of RF saturation is minimal with respect to the time evolution of the signal. Future studies will focus on the optimization of the image acquisition to take advantage of the time evolving magnetization.

One limitation of using longer infusion times is that the bicarbonate metabolite was not detectable when injection times were increased from 90 to 120 s (Supporting Information). This is probably a signal‐to‐noise limitation rather than a differential effect based on a lower effective substrate concentration but could be investigated in future studies. The pseudo‐steady state kinetics achievable using longer infusion times can potentially be utilized for several different types of experiments, which are admittedly outside the immediate scope of this work. For example, pseudo‐steady state schemes allow for classes of experiments where an additional experimental condition is varied across the substrate delivery plateau, such as diffusion gradient encoding to probe metabolite compartmentalization or additional spectroscopic dimensions. Furthermore, because it is possible to acquire the entire available signal within a given TR, such as with a 90° excitation pulse, it is possible to potentially simplify kinetic modeling. More complex metabolic kinetics, such as Michaelis–Menten kinetics, could be approached by varying the substrate infusion concentration.

We hypothesize that some of the favorable relaxation properties, resulting from the dissolution of the hyperpolarized substrate in D₂O, are retained during the infusion, despite mixing that occurs in the blood pool, either via some residual D₂O or via a possible slow exchange with a deuteration shell surrounding the molecule. By slowly infusing the substrate, a larger portion of it remains dissolved in D₂O in the fringe field of the magnet throughout the experiment, where it has favorable relaxation properties. This suggests exploring whether potentially more dilute or larger volume D₂O infusates can improve signal‐to‐noise ratio by further optimizing relaxation properties, within safety limits. Additionally, the slow infusion approach can act as a damper to smooth concentration fluctuations over discrete hyperpolarized dissolution/production events, and enable truly longer‐scale infusions (∼20 min) limited by subject tolerance. Lastly, the lengthened infusion approach has a simple but important implication for clinical imaging. Namely, the acquisition time can be increased and the design of hyperpolarized ¹³C spectroscopic sequences can be biased away from rapid acquisition, toward other parameters such as resilience against artifacts, higher signal to noise, or higher resolution.

5. Conclusion

Increased infusion times of hyperpolarized [1‐¹³C]pyruvate dissolved in D₂O are feasible and offer an alternate paradigm to bolus type infusions for probing metabolic kinetics approaching pseudo‐steady state. This can enable more robust in vivo metabolic imaging leading to new types of experiments with potential clinical application.

Conflicts of Interest

Unrelated to this work, K.R.K. is co‐founder of Atish Technologies and serves on the Scientific Advisory Boards of NVision Imaging Technologies, Imaginostics and Mi2. He holds patents related to imaging and leveraging cellular metabolism.

Supporting information

Figure S1: Dynamic MRS of hyperpolarized [1‐¹³C]pyruvate in a mouse brain when investigated with increasing infusion times.

Evolution of metabolic substrate and products observed in a murine brain when 300 μL of hyperpolarized [1‐¹³C]pyruvate were injected over 15 s (A–C), 60s (D–F), 90s (G–I) and 120 s (J–K), at a constant rate, using a syringe pump. For each acquisition (n = 12), a representative ¹³C NMR spectrum is also shown in the representative color code (red for 15 s, black for 60s, plum for 90s and blue for 120 s infusion times) with all metabolites annotated, [1‐¹³C]pyruvate (Pyr) at 171.1 ppm, [1‐¹³C]pyruvate hydrate (Pyr Hyd) at 179 ppm, [1‐¹³C]lactate (Lac) at 183 ppm, [1‐¹³C]alanine (Ala) at 176 ppm and [1‐¹³C]bicarbonate at 161.1 ppm. A magnification of the spectral region where the bicarbonate peak (161.1 ppm) is observed (158–165.5 ppm) is show for each spectrum.

Sixty repetitions of the dynamic scan are shown for each mouse.

Porcari P., Miloushev V., Patel S., Peyear T., Berishaj M., and Keshari K. R., “Increasing Infusion Times of Hyperpolarized [1‐ ¹³C]Pyruvate in a Mouse Brain Maintain Lactate Generation and Approach Pseudo‐Steady State Metabolism,” Magnetic Resonance in Medicine 95, no. 4 (2026): 2254–2265, 10.1002/mrm.70175.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.