[6,6′-²H₂] fructose as a deuterium metabolic imaging probe in liver cancer

Abstract

Hepatocellular carcinoma (HCC) is one of the leading causes of cancer-related deaths. Imaging plays a crucial role in the early detection of HCC, although current methods are limited in their ability to characterize liver lesions. Most recently, deuterium metabolic imaging (DMI) has been demonstrated as a powerful technique for the imaging of metabolism in vivo. Here, we assess the metabolic flux of [6,6′-²H₂] fructose in cell cultures and in subcutaneous mouse models at 9.4 T. We compare these rates with the most widely used DMI probe, [6,6′-²H₂] glucose, exploring the possibility of developing ²H fructose to overcome the limitations of glucose as a novel DMI probe for detecting liver tumors. Comparison of the in vitro metabolic rates implies their similar glycolytic metabolism in the TCA cycle due to comparable production rates of ²H glutamate/glutamine (glx) for the two precursors, but overall higher glycolytic metabolism from ²H glucose because of a higher production rate of ²H lactate. In vivo kinetic studies suggest that HDO can serve as a robust reporter for the consumption of the precursors in liver tumors. As fructose is predominantly metabolized in the liver, deuterated water (HDO) produced from ²H fructose is probably less contaminated from whole-body metabolism in comparison with glucose. Moreover, in studies of the normal liver, ²H fructose is readily converted to ²H glx, enabling the characterization of ²H fructose kinetics. This overcomes a major limitation of previous ²H glucose studies in the liver, which were unable to confidently discern metabolic flux due to overlapped signals of ²H glucose and its metabolic product, ²H glycogen. This suggests a unique role for ²H fructose metabolism in HCC and the normal liver, making it a useful approach for assessing liver-related diseases and the progression to oncogenesis.

1 |. INTRODUCTION

Early detection and treatment of hepatocellular carcinoma (HCC), one of the leading causes of cancer-related deaths due to its aggressive malignancy, remains a major challenge in healthcare.¹ Metabolic imaging is an important noninvasive method for the detection and localization of HCC, providing insights into therapeutic strategy and treatment response.² Current methods for metabolic imaging aim to differentiate healthy and cancerous tissues by characterizing changes in glycolysis, as enhanced aerobic glycolysis, termed the Warburg effect, is a metabolic phenotype of cancer.³ Fluorodeoxyglucose–positron emission tomography (FDG-PET) is one of the most widely used clinical modalities, by detecting enhanced glucose uptake.^4–8 However, information regarding downstream metabolic processes, which is of fundamental importance for tumor detection and evaluation of therapeutic treatment, cannot be obtained by FDG-PET. Hyperpolarized ¹³C MRI shows the capability of interrogating tumor metabolism, but is limited to molecules with long T₁ relaxation times on the ¹³C nuclei.^9,10 Biologically important molecules, such as fructose and glucose, which are the main sources of energy in HCC,¹¹ have short ¹³C T₁ lifetimes and thus remain challenging for applications in a clinical setting.

Recently, deuterium (²H or D) metabolic imaging (DMI), based on administration of a ²H-labeled substrate in animal models or human subjects followed by examination of its downstream metabolites in real time, is a simple and powerful alternative for clinical imaging of tumor metabolism.^12,13 This technique allows the observation of tumor metabolism in dynamic or steady-state measurements, providing insights into metabolic flux through disease-related metabolic pathways. Because of the low natural abundance (0.0115%) of ²H, ²H-labeled substrates and products can be detected using conventional MRI with minimum background (i.e., water or lipid) signal. Because of the short T_1,2H, relatively high ²H MRI sensitivity can be achieved with fast signal averaging, compensating for the low sensitivity of ²H (γ_2H ≈γ_1H/6.5). Additionally, a wide range of ²H-labeled substrates are applicable. Given these advantages, the use of DMI as a reporter of in vivo metabolism has gained considerable popularity.^14–26

In view of the interest in studying liver cancer, there has been experimental evidence that fructose, which is predominantly metabolized in the liver, promotes the Warburg effect for tumor growth and metastasis,^27,28 indicating the important role of fructose metabolism in tumorigenesis. More recent work has highlighted that fructose metabolism is specifically reprogrammed in liver cancer, leading to fructose carbons being shunted through glycolysis to lactate, as opposed to canonical flux through the enzyme ketohexokinase.²⁹ This specific switch in metabolic utilization makes probes of fructose-derived glycolytic flux potentially useful in liver imaging. Previous DMI studies have investigated metabolism in the normal liver and liver tumors with ²H-enriched substrates, including [6,6′-²H₂] glucose,^14,18 L-[methyl-²H₃] methionine,^30,31 and [²H₉]-choline,²⁵ providing the feasibility of such an approach. The current study describes quantitative characterization of [6,6′-²H₂] fructose metabolism in HepG2 cells, subcutaneous tumor models, and mouse normal liver followed by comparison with [6,6′-²H₂] glucose, the most widely used DMI precursor. By referencing to the natural abundance deuterated water (HDO) signal in vivo, the production of metabolites and thus the metabolic rates were obtained, allowing for the comparison of metabolic flux of the two precursors. The purpose of this study is to assess the feasibility of developing ²H fructose as a useful deuterium imaging probe for detecting and localizing liver tumors and other liver-related diseases.

2 |. MATERIALS AND METHODS

2.1 |. Cell culture and tumor induction

HepG2 human liver cancer cells were cultured in Gibco Dulbecco’s Modified Eagle Medium (Thermo Fisher Scientific, Waltham, MA, USA) supplemented with 10% fetal bovine serum, 100 mg/mL penicillin–streptomycin, and 2.5 mM L-glutamine. The cells were grown under standard conditions, incubated at 37°C with 5% CO₂. Suspensions of HepG2 cells were mixed with Matrigel in phosphate-buffered saline (PBS) (1:1 v/v) and injected (~1 × 10⁷ cells) subcutaneously into the right flanks of the female athymic nude mice (4- to 6-week-old, average weight 25 g, Jackson Laboratories, Bar Harbor, ME, USA). As sex-dependent differences in the ²H glucose or ²H fructose metabolism in vivo were not considered in the current study, mice with a single sex were used. Here, female mice were chosen because they are less aggressive compared with male mice, and were therefore easier to handle during the experiments. All mice were then housed under standard conditions in a 12-h light/dark cycle with access to food and water, and were further monitored by measuring tumor volume. The anatomical image of the tumor was acquired using a T₁-weighted sequence, where tumor volume can be determined by tumor length/2 × (tumor width)². Tumors were allowed to develop to a size of 0.25 to 0.35 cc for in vivo experiments.

2.2 |. Treatment conditions and sample preparation for in vitro NMR

1 × 10⁶ HepG2 cells were plated in each well of six-well plates and incubated overnight. Then the media were exchanged for glucose-deficient media containing 10 mM [6,6′-²H₂] glucose or 10 mM [6,6′-²H₂] fructose, followed by incubation for 1, 2, 4, 8, and 24 h at 37C with 5% CO₂. After each time point, cell media from each of the six wells (three wells with ²H glucose and three wells with ²H fructose) were collected and stored at −80°C overnight. The time points for each experiment were repeated three times. The media samples were filtered at 14,000 rpm and 4C for 30 min using Amicon Ultra 0.5-mL centrifugal filters (3kD; Merck Millipore Ltd, Burlington, MA, USA).³² For each NMR sample, 1 μL of 501.53 mM methylsulfonylmethane-d₆ (MSM-d₆) as an internal standard, and 1 μL of 1.95 M sodium azide as an antibacterial agent, were added to 648 μL of filtered media, resulting in a total 650-μL sample containing 0.77-mM MSM-d₆ and 3-mM sodium azide.

2.3 |. ²H and ¹H NMR data acquisition and processing

²H and ¹H NMR spectroscopy were performed on a 14.1-T NMR spectrometer (Bruker Biospin, Billerica, MA, USA) installed with a Bruker 5-mm DCH helium cryoprobe with ²H lock that has a broadband direct observe channel tunable to most nuclei except for ¹⁹F, and a 11.7-T NMR spectrometer installed with a Bruker 5-mm BBO probe with ²H lock, respectively. ²H detection was accomplished by pulsing on the ²H lock for both instruments. ²H and ¹H spectra were acquired using a one-pulse sequence. The flip angle α of the excitation pulse is 90 for the acquisition of both spectra. The acquisition parameters for ²H spectra are pulse strength (γB₁)/2π 1.67 kHz, complex points per scan 32,768, acquisition time 3.9 s, and no less than 1800 transients. The acquisition parameters for ¹H spectra are pulse strength (γB₁)/2π 16.7 kHz, complex points per scan 65,536, acquisition time 3 s, and no less than 64 transients. For ¹H experiments with ²H decoupling, the WALTZ-16 decoupling sequence was applied on the ²H channel during ¹H acquisition. Both ²H and ¹H resonances were identified and quantified by TopSpin 4.0 (Bruker Biospin).

2.4 |. Tail-vein cannulation for mice

All animal experiments were approved by the Institutional Animal Care and Use Committee at Memorial Sloan Kettering Cancer Center. All mice (subcutaneous HepG2 mouse models for imaging HepG2 tumors: a total of $n=7,n=4$ for fructose studies and $n=3$ for glucose studies; healthy mice for imaging normal liver: $n=3$) were in the fasting state. Mice were fasted overnight before the experiments in the morning. For each experiment, a female nude mouse was cannulated with a 23-gauge rodent tail-vein catheter (Braintree Scientific, Braintree, MA, USA). Also, 10 U heparin/mL in PBS was prefilled in the catheter to keep the mouse heparinized prior to injection with the precursor, and the mouse was anesthetized using a continuous flow of 1 L/min oxygen with 1.5% isoflurane.

2.5 |. ²H MRSI experiment setup

All imaging experiments were conducted using a Bruker 9.4-T 20-cm bore magnet imaging system. The mouse with tail-vein cannulation was loaded and restrained inside a transmit-receive ¹H volume coil on the MRI bed equipped with a heated circulating water bath (Thermal Fisher Scientific, Newington, NH, USA) for maintaining the body temperature of the mouse at a physiological level and nose cone for oxygen/isoflurane delivery. A 24-mm diameter ²H transmit-receive surface coil (Doty Scientific, Columbia, SC, USA; Figure S1a) tuned to 61.4 MHz (²H Larmor frequency) was centered about the subcutaneous tumor or mouse normal liver. During image acquisition, a physiological monitoring system from SA instruments (Stony Brook, USA) was used to monitor respiration, while the mouse was under 1.5% isoflurane (Baxter Healthcare Corp., Deerfield, IL, USA) in oxygen. The respiration rate was regulated by adjusting the isoflurane level in oxygen delivered to the mouse, typically between 1% and 2%. Axial anatomical ¹H T₂-weighted images were acquired using the RARE sequence with 156-μm resolution across 20 slices throughout the mouse abdomen. The magnetic field was then shimmed manually on the tumor using the ¹H volume coil prior to the acquisition of ²H magnetic resonance spectroscopic imaging (MRSI). Immediately after acquisition of the first time point of ²H MRSI, the mouse was injected with ~350 μL of [6,6′-²H₂] fructose or [6,6′-²H₂] glucose (1.95 g/kg body weight; dissolved in saline) over 10 s through the tail-vein catheter. Note that this injected volume excluded the 100 μL of 10 U heparin/mL solution in the dead volume of the catheter.

2.6 |. ²H MRSI experiments

²H MRSI data were acquired using an axial two-dimensional chemical shift imaging (2D CSI) sequence with a matrix size of 8 × 8, field of view (FOV) of 40 × 40 mm², and in-plane resolution of 5 × 5 mm². The sensitivity region of the ²H coil (24 × 24 × 10 mm³; Figure S1b) leads to a slice thickness of 10 mm. For each image, a total of 2048 complex points were acquired with an echo time of 1.218 ms, repetition time of 278 ms, a flip angle of 70, and 64 averages. A total of 12 images and eight images were acquired with a 5-min time interval for subcutaneous tumors and mouse normal liver, respectively. The data were zerofilled in the spectra dimension, apodized, and Fourier transformed using MATLAB (MathWorks, Natick, NA, USA). Metabolites’ peaks from the in vivo ²H MRSI data were fitted by MNova (Mestrelab Research) using the global spectral deconvolution method where the integral areas were obtained.

3 |. RESULTS

3.1 |. Identification of deuterated metabolites in vitro using ²H and ¹H NMR

Upon metabolism, ²H fructose can be converted to multiple glycolytic intermediates, potentially enabling ²H to facilitate detection of multiple metabolite pools (Figure 1). To assess the ability of ²H MRI to detect this flux, HepG2 cells were incubated with [6,6′-²H₂] fructose and [6,6′-²H₂] glucose for 24 h, respectively. The resulting cell media samples were measured by ²H NMR (see Materials and Methods). ²H signals of the metabolic products, deuterated lactate, glutamate/glutamine (glx), alanine, and HDO, were observed in the culture media (Figure 2A), indicating that HepG2 cells actively metabolize both precursors. The glx peak displays as a doublet with ²J_1H-2H ~2.3 Hz, indicating that mono-deuterated glx isotopomers dominate.

FIGURE 1.

Schematic representation of the metabolic pathway of [6,6′-²H₂] fructose during fructolysis and TCA cycle. Deuterium loss is shown in the form of ²HOH. Mono-deuterated lactate can be produced from mono-deuterated malate through the pyruvate-malate shuttle (dashed line). Unlabeled lactate can be produced from unlabeled DHAP via triose phosphate isomerase. Note that the metabolic pathway of [6,6′-²H₂] glucose follows the HK pathway of [6,6′-²H₂] fructose. Key products observed in in vitro NMR and in vivo DMI experiments are shown in the scheme. See Figure S9 for ²H labeling of all intermediates in the TCA cycle. DHAP, dihydroxyacetone phosphate; F1P, fructose 1-phosphate; F6P, fructose 6-phosphate; F1,6P, fructose 1,6-biphosphate; GA, glyceraldehyde; GA3P, glyceraldehyde 3-phosphate; HK, hexokinase; KHK, ketohexokinase; TCA, tricarboxylic acid.

FIGURE 2.

²H metabolites are identified by ²H and 1H NMR spectra and further quantified in kinetic studies, revealing differences in in vitro metabolic rates between ²H glucose and ²H fructose. (A) ²H NMR spectra of the cell media samples after incubating HepG2 cell suspensions with 10 mM [6,6′-²H₂] fructose (pink) and 10 mM [6,6′-²H₂] glucose (blue) for 24 h, respectively. The expanded views of the spectra show peaks of deuterated glx and deuterated alanine. (B) Overlaid ¹H NMR spectra with (pink) and without (black) ²H decoupling of the cell media samples incubated with [6,6′-²H₂] fructose in (A). Doublets of different lactate isotopomers are shown, displaying J-coupling constants of 7 Hz that rise from ¹H at the C2 position of lactate. Addition of ²H causes a shift to a higher frequency. The asterisks designate resonances from cell media. Note that ¹H NMR spectra with ²H decoupling display the same peak pattern but larger signals of lactate isotopomers for cells incubated with [6,6′-²H₂] glucose (Figure S2). Plots of concentrations of ²H metabolites produced (C–F) and precursors (G) consumed in the cell media with respect to the incubation time. HepG2 cell suspensions were incubated with 10 mM [6,6′-²H₂] fructose (pink) and 10 mM [6,6′-²H₂] glucose (blue) for 1, 2, 4, and 8 h, respectively. Solid lines show the best fit from the kinetic analysis (see Table 1), demonstrating the production rates of metabolites and consumption rates of precursors. For determination of the production rate $\left(r\right)$ of each ²H metabolite, the concentrations of ²H metabolites $c\left(t\right)$ produced against the incubation time $\left(t\right)$ were fitted to $c\left(t\right)=r\cdot t$ in (C) and (D) and $c\left(t\right)=c\left(t_{\mathrm{\infty }}\right)\cdot \left(1-e^{-r\cdot t}\right)$ in (E) and (F), where $c\left(t_{\mathrm{\infty }}\right)$ refers to the concentration of ²H glx or ²H Ala at the end of the metabolic process $(t=\mathrm{\infty })$ when all ²H glucose was consumed. The fit of the consumption of ²H precursor (G) used the same equation as (C) and (D), except that $c\left(t\right)$ is the concentration of ²H precursor that was consumed and $\left(r\right)$ is the consumption rate. Each of the experiments was repeated three times.

¹H NMR spectra with and without ²H decoupling of the cell media samples were further recorded. A comparison of the methyl signals in the two spectra (Figures 2B and S2) indicates the presence of three lactate isotopomers, which are lactate-CH₃, lactate-CH₂D, and lactate-CHD₂, as shown in the metabolic pathway (Figure 1). The methyl ¹H resonances are split by ²J_1H-2H and ³J_1H-1H for lactate-CHD₂ and lactate-CH₂D, resulting in a doublet of quintets and a doublet of triplets, respectively. Yet the small geminal couplings of ²J_1H-2H (²J_1H-2H ≈ ²J_1H-1H/6.5 due to γ_2H ≈γ_1H/6.5) make the J-splitting on the methyl ¹H appear as line-broadening and indistinguishable. Therefore, the methyl signal of ²H-labeled lactate appears broad (Figure 2A). This was resolved by implementation of ²H decoupling during the acquisition of ¹H signals, which shows well-resolved doublets of lactate-CHD₂ and lactate-CH₂D (Figures 2B and S2). This is consistent with the observation by Mahar et al. of lactate production from [²H₇] glucose metabolism in the media of Huh-7 cells.³³ Each of the isotopomers displays as a doublet in the spectra showing a coupling constant of ³J_1H-1H ~7 Hz, which further confirms their origination from the same species. A secondary isotope shift of ~8 Hz is observed for each replacement of a methyl ¹H with a ²H in the methyl group. Taken together, these data confirm that ²H-fructose is metabolized through glycolysis to lactate and that this can be detected with ²H NMR.

3.2 |. Kinetics of [6,6′-²H₂] fructose or [6,6′-²H₂] glucose metabolism in vitro using ²H NMR

The identification of deuterated metabolites enables monitoring of the kinetics of [6,6′-²H₂] fructose and [6,6′-²H₂] glucose metabolism in human-derived HepG2 liver cells (Figure 2C–F). ²H NMR spectra were recorded for the cell media samples by incubating the precursors with the HepG2 cell line as a function of time. As the incubation period increased, stacked plots of ²H spectra showed an increase in the signals of the metabolic products and a decrease in the signals of the precursors (Figure S3). Integral areas of each of the products and precursors at different time points were extracted and converted into concentrations by using MSM-d₆ as an internal standard. Concentrations were fit as a function of time to derive the metabolic rates (Table 1). The production rate of ²H lactate is one-half that of the consumption rate of the precursors, indicating that one-half of the precursor was converted into ²H lactate during glycolysis or fructolysis.

TABLE 1.

The fit results of metabolic rates in vitro and in vivo in subcutaneous HepG2 mouse models.

Metabolite	In vitro		In vivo
	Production rated (mM/h/106 cells)		Production rate (mM/h/cm3 tumor)c		Production rate (mM/h/106 cells)
	GLU	FRU	GLU	FRU	GLU	FRU
2H lac	0.147 ± 0.033	0.049 ± 0.012	50.9 ± 9.0	25.7 ± 3.8	0.051 ± 0.009	0.026 ± 0.004
HDO	0.449 ± 0.113	0.307 ± 0.111	140 ± 56	144 ± 48	0.140 ± 0.056	0.144 ± 0.048
2H glxa	(0.527 ± 0.227) × 10−2	(0.573 ± 0.592) × 10−2	b
2H alaa	(0.493 ± 0.507) × 10−2	(0.482 ± 0.696) × 10−2
Precursor	Consumption rated (mM/h/106 cells)
2H GLU	0.307 ± 0.062
2H FRU	0.090 ± 0.003

Abbreviations: ala, alanine; FRU, fructose; GLU, glucose; glx, glutamate/glutamine; HDO, deuterated water; lac, lactate; SNR, signal-to-noise ratio.

^aThe low SNR of deuterated glx and alanine in vitro leads to large variations in the fit results.

^bNo observation of deuterated glx and alanine in subcutaneous mouse models.

^c1 cm³ tumor ≈ 109 cells.

^dSee Figure 2 for details of the fitted rate.

3.3 |. Kinetics of [6,6′-²H₂] fructose or [6,6′-²H₂] glucose metabolism in vivo

For comparison of metabolic conversion of these molecules both in vitro and in vivo, we established subcutaneous HepG2 mouse models to determine the rates of conversion in vivo; 1.95 g/kg [6,6′-²H₂] fructose or [6,6′-²H₂] glucose was infused intravenously through the tail veins of subcutaneous HepG2 mouse models. Kinetics of the metabolism of the precursors was monitored in vivo by 2D CSI using a ²H surface coil (Figure S1). In the CSI spectra extracted from the tumor region, changes in the signals of [6,6′-²H₂] fructose and HDO were observed over time (Figure 3). Note that a similar signal change was observed for mice injected with [6,6′-²H₂] glucose (Figure S4). Each of the metabolite peaks was further fitted individually and converted into concentrations by referencing to the natural abundance HDO peak (~12.7 mM) acquired in the first time point and by correcting for the number of deuterons in each molecule (Figure 4A–C). Approximately 5 min after the start of the CSI acquisition, the precursors were injected, followed by washing into the tumor, and were almost completely metabolized in 60 min (Figure 4B,C). An increase in the HDO signal was observed initially for both precursors. Over the course of the experiment, the HDO signal continued to rise and reached a concentration ~2.98 and ~3.86 times of natural abundance HDO at the end of the experiment, corresponding to (37.3 ± 8.2) mM $(n=3)$ and (48.2 ± 5.2) mM $(n=4)$ from [6,6′-²H₂] glucose and [6,6′-²H₂] fructose, respectively.

FIGURE 3.

Kinetics of ²H fructose metabolism was monitored using HepG2 tumor models, showing the arrival of ²H fructose in the tumor followed by metabolic conversion. (A) The T₂-weighted image (axial view) acquired for anatomical reference. The orange circle shows the position of the ²H surface coil. The tumor is outlined in a yellow dashed line. (B) ²H spectra extracted from a CSI grid (purple square) with 5 × 5 × 10 mm³ voxels. The spectra shown were acquired between 20 and 25 min starting the infusion of [6,6′-²H₂] fructose. (C) Plots of time series of ²H spectra in the tumor voxel (red square) showing the consumption of [6,6′-²H₂] fructose and the production of HDO. [6,6′-²H₂] fructose (1.95 g/kg body weight and dissolved in saline) was infused in the mouse tail vein at ~5 min after the start of the CSI acquisition. Note that the acquisition of DMI using [6,6′-²H₂] glucose follows the same experimental procedure (Figure S4). The ²H fructose and ²H glucose experiments in vivo were repeated four and three times, respectively.

FIGURE 4.

Quantitative analysis of the in vivo kinetic data reveals the metabolic rates of ²H metabolites in the TCA cycle and glycolysis pathway. (A) Concentration ratio of HDO against the precursors plotted as a function of time showing the correlation between the consumption of the precursors and the production of HDO. Dashed lines show the linear fit of the concentration ratio against time, in the initial 25 min for ²H fructose (pink) and 35 min for ²H glucose (blue), respectively. The resulting fit equations are $\frac{c_{HDO}}{c_{\text{fructose}}}=0.051\cdot t+0.036,R^2=0.97$, and $\frac{c_{HDO}}{c_{\text{glucose}}}=0.0077\cdot t+0.78,R^2=0.65$. The time ranges displaying the linear correlation in (A) are indicative of the time ranges used for the fits in (B) and (C). (B) Concentrations of HDO and [6,6′-²H₂] fructose shown in Figure 3C (tumor voxel) were plotted as a function of time. The solid lines (pink) show the best fit of the initial increase of HDO, where in vivo production rates of HDO in the first round of TCA cycle were obtained. (C) The same as (B), except for [6,6′-²H₂] glucose (blue), as shown in Figure S4c. (D) Sum of the ²H spectra recorded between 30 and 60 min after the injection of [6,6′-²H₂] glucose and [6,6′-²H₂] fructose, respectively. The peaks were further fitted, and this was used to calculate the in vivo production rates of ²H lactate.

A plot of HDO signal intensities against precursors’ signal intensities over time (Figure 4A) shows a linear correlation between HDO production and precursor consumption in the initial ~25 min for [6,6′-²H₂] fructose and ~35 min for [6,6′-²H₂] glucose, implying HDO arising initially from the TCA cycle can serve as a robust reporter of the uptake of the precursors in the tumor. Thus, a fit of the initial HDO signal intensities versus time (Figure 4B,C) derives the in vivo HDO production rate in the first round of TCA cycle (as shown in Table 1), corresponding to a production of HDO (35 ± 14) mM/h from ²H glucose and (36 ± 12) mM/h from ²H fructose per tumor voxel (voxel size = 0.25 cm³). After the initial period, the concentration of HDO continues to increase rapidly, which could be attributed to the second round of TCA cycle producing additional HDO, or the influx of HDO generated from metabolism of the precursors in the surrounding tissues.

The lactate signal was not observable in the individual spectrum but was observed by summing the spectra recorded between 30 and 60 min (Figure 4D). Summing the spectra from earlier time points results in an overlap of the large ²H fructose or ²H glucose signal with the lactate signal. Fit of the metabolites signal in the summed spectra followed by signal integration derives the averaged in vivo lactate production rates within the 30-min time window (Table 1), corresponding to a production of ²H lactate 12.7 ± 2.2 mM/h from ²H glucose and 6.4 ± 0.9 mM/h from ²H fructose per tumor voxel. After injection, ²H glucose and ²H fructose reached the tumor with the highest concentrations being 31.6 ± 3.9 mM/voxel $(n=3)$ in 10–15 min and 46.1 ± 4.3 mM/voxel $(n=4)$ in 5 min, corresponding to 0.118 ± 0.015% injected dose per gram of mouse (%ID/g) and 0.172 ± 0.016%ID/g (total amount of dose, 1.07 M in 0.25 mL saline; mouse weight, 25 g). This suggests that the dose was readily distributed throughout the body following intravenous injection and thus a small percentage of the dose reached the tumor, followed by approximately 40% and 55% ²H glucose as well as 14% and 39% ²H fructose converted into ²H lactate and HDO per hour ([the total concentration of HDO or ²H lactate produced in 1 h]/[n × {maximum concentration of ²H fructose or ²H glucose in the tumor}], where $n=2$ for HDO and $n=1$ for ²H lactate). The in vivo production rates of HDO from ²H glucose and ²H fructose are comparable (Table 1); however, the percentage of ²H glucose (55%) converted to total amount of HDO per hour in the liver tumor is higher than that of ²H fructose (39%). Since glucose is metabolized in almost every organ; by contrast, fructose is predominantly consumed in liver, this indicates that HDO produced from ²H fructose is probably localized to the liver tumor region, while HDO produced from ²H glucose originated not only from liver tumor, but also from the inflows of HDO from surrounding organs. The production of HDO and ²H lactate from ²H fructose metabolism in the HepG2 tumors were further confirmed by analogous ²H tracing experiments followed by ²H NMR analysis (Figure S5). Additionally, the observed production rates of ²H lactate in vivo are slower than those in vitro $\left({\nu }_{lac-glu,invitro}/{\nu }_{lac-glu,invivo}\approx 3;{\nu }_{lac-fru,invitro}/{\nu }_{lac-fru,invivo}\approx 2\right)$, possibly caused by the differences between an in vivo and an in vitro environment (see details in Section 4).

To compare the metabolic rates observed in these tumor models, we sought to quantify flux in the mouse normal liver. ²H glucose suffers in the normal liver due to overlapping resonances of the substrate and potential glycogen products,¹⁴ thus limiting its use in many liver-related diseases (i.e., fatty liver disease and hepatitis). [6,6′-²H₂] fructose was injected intravenously ~10 min after the start of the CSI acquisition. In the time-resolved spectra from a 2D CSI voxel extracted from the liver region (Figures 5A–C, S6a–c, and S7a–c), ²H fructose was washed into the liver and reached its peak concentration of (31.4 ± 5.1) mM ([0.117 ± 0.019]%ID/g mouse) in 5–10 min after injection, followed by increased signals of HDO and ²H glx. HDO signal continued to increase and reached a concentration of (24.6 ± 1.7) mM in 40 min. The ²H glx signal appeared in ~5 min after injection, then peaked with a concentration of (5.8 ± 0.9) mM in 5–10 min, after which the signal decayed. Compared with HepG2 tumor models, no ²H lactate signal was observed in the normal liver (Figures 5B, S6b, and S7b), indicating an enhanced anaerobic metabolism in the tumor tissue. Fit of the ²H glx signal in the summed spectra followed by signal integration derives the in vivo ²H glx production rate, which is (115 ± 11) mM/h/cm³ liver ([0.115 ± 0.011] mM/h/10⁶ cells, assuming 1 cm³ liver = 10⁹ cells).

FIGURE 5.

In vivo kinetic studies of ²H fructose in mouse normal liver demonstrate the generation of ²H glx. (A) Plots of time series of ²H spectra in the mouse normal liver voxel showing the consumption of [6,6′-²H₂] fructose and the production of HDO and ²H glx. [6,6′-²H₂] fructose (1.95 g/kg body weight and dissolved in saline) was infused in the mouse tail vein at ~10 min after the start of the CSI acquisition. (B) Sum of ²H spectra recorded from 15 to 40 min in (A). (C) Concentrations of HDO (black), ²H glx (blue), and [6,6′-²H₂] fructose (red) in vivo plotted as a function of time. The solid lines (purple) show the best fit of the initial increase of HDO, where the production rate of HDO in the first round of TCA cycle in the normal liver was obtained. The integral of HDO peak cannot be obtained at 15 min because of the overlap with the large ²H fructose peak, resulting in a missing marker of HDO at 15 min in (C). Concentration ratio of HDO against the ²H fructose in (D) and ²H glx in (E) plotted as a function of time showing the correlation among HDO production, ²H fructose consumption, and ²H glx production in the TCA cycle. The experiment was repeated three times (see Figures S6 and S7 for the data of additional healthy mice).

The concentration of HDO was plotted against concentrations of ²H fructose or ²H glx to evaluate the correlation between HDO production, ²H fructose consumption, and TCA cycle activity. A linear correlation between HDO and ²H fructose was observed initially (Figures 5D, S6d, and S7d), indicating that HDO production originating from the first round of TCA cycle can serve as a simplified means to assess the uptake of ²H fructose in the normal liver, which was also observed in liver tumor models. A fit of the initial HDO concentration derives the HDO production rate, which is (140 ± 13) mM/h/cm³ liver. Furthermore, a plot of HDO concentration against ²H glx concentration reveals an initial c(HDO)/c(²H glx) value below 10, followed by a significant increase. This increase can be attributed to the continued production of HDO compared with ²H glx (Figures 5E, S6e, and S7e; see details in Section 4). The ²H glx peak in the liver was further confirmed by analogous ¹³C tracing experiments and ¹³C NMR analysis (Figure S8). Subsequently, HDO continues to increase due to the conversion of ²H citrate to isocitrate producing additional HDO (50% possibility) in the second round of TCA cycle (see details in Figure S9).

4 |. DISCUSSION

The metabolism of [6,6′-²H₂] fructose and [6,6′-²H₂] glucose in HepG2 cell cultures (Figures 1, 2A, and S9) produce ²H metabolites from multiple metabolic pathways. HDO can be produced from keto-enol tautomerization in pyruvate during glycolysis. Yet previous work by De Feyter et al.¹⁴ suggested that a maximum of 8% HDO was generated from tautomerization in pyruvate when using [6,6′-²H₂] glucose as a precursor, indicating the relatively low occurrence of the process. In addition to glycolysis, deuterium can also be lost as solvent water during the conversion of citrate to isocitrate in the TCA cycle and the conversion of pyruvate to alanine.³⁴ Deuterated glx containing signals from [4-²H] glutamate/glutamine and [4,4′−²H₂] glutamate/glutamine is produced from α-ketoglutarate, an intermediate originated from the TCA cycle. These metabolites were unable to separate in in vitro DMI and therefore a combined kinetics was recorded. Lactate-CHD₂ is produced from the glycolytic 3-carbon intermediate pyruvate-CHD₂ (Figure 1). Pyruvate-CHD₂ further enters the mitochondria, where it undergoes decarboxylation catalyzed by pyruvate dehydrogenase to acetyl-CoA followed by a turn of TCA cycle, or carboxylation catalyzed by pyruvate carboxylase to oxaloacetate-CD₂ and oxaloacetate-CHD followed by conversion to malate-CD₂ or malate-CHD. Malate is exported from the mitochondria to the cytosol where it is decarboxylated into pyruvate-CHD₂ and pyruvate-CH₂D catalyzed by malic enzyme with the reduction of NADP⁺ to NADPH (pyruvate-malate shuttle; Figure 1). The produced pyruvate can be converted to lactate-CHD₂ and lactate-CH₂D (Figure 2B) or re-enter mitochondria to participate in another turn of the shuttle. As pyruvate carboxylase is abundant in liver, the pyruvate-malate pathway is supported by experiments that showed malate export from isolated liver mitochondria.^35,36 Lactate-CH₂D can further be produced from pyruvate-CH₂D originated through keto-enol tautomerization in pyruvate-CHD₂. However, due to the limited occurrence of the process,¹⁴ its contribution to the formation of pyruvate-CH₂D followed by lactate-CH₂D production should be negligible.

The values for the ratio of lactate-CHD₂ to lactate-CH₂D, assessed by integrating the corresponding methyl ¹H signals and correcting for the number of protons in each isotopomer, are 7:5 and 9:5 using [6,6′-²H₂] fructose and [6,6′-²H₂] glucose as the precursor, respectively. This indicates the percentage contribution of each pathway (glycolysis and pyruvate-malate) to lactate production. The production of lactate-CHD₂ is due to the contribution from both pathways. Theoretically, the conversion of a methyl group in pyruvate-CHD₂ to a methylene group in malate is equal to a 33.3% or 66.7% statistical chance that two deuterons or a deuteron are passed from pyruvate-CHD₂ to malate in the pyruvate-malate pathway, which is then decarboxylated to pyruvate-CHD₂ or pyruvate-CH₂D followed by the production of lactate-CHD₂ or lactate-CH₂D. Therefore, possible values for the ratio of lactate-CHD₂ (glycolysis) to lactate-CHD₂ (pyruvate-malate) to lactate-CH₂D (pyruvate-malate) could be 4.5:2.5:5 and 6.5:2.5:5 using [6,6′-²H₂] fructose and [6,6′-²H₂] glucose as the precursor, respectively. The differences in the concentrations of lactate isotopomers could be caused by the different metabolic rates of the precursors of lactate isotopomers due to deuterium kinetic isotope effects, in which the replacement of a proton with a deuteron results in a reduced rate caused by the presence of a heavier isotope.³⁷ This suggests that the production rate of lactate-CHD₂ is slower than that of lactate-CH₂D. However, the higher concentration of lactate-CHD₂ implies that the production rate of lactate from both pathways is higher using ²H fructose and ²H glucose as precursors. Furthermore, the ratio of lactate-CH₂D (pyruvate-malate) to lactate-CHD₂ (glycolysis) is higher from ²H fructose than that from ²H glucose, indicating more ²H fructose goes through the pyruvate-malate pathway. Studies have found significantly increased hepatic malic enzyme in mice fed with fructose^38,39 and suggest that fructose is shunted more through this pathway. Interestingly, this differential flux revealed by ²H isotopomer analysis indicates the role of fructose in the malic enzyme-mediated pathway, which may be targeted for therapeutic approaches against liver cancer, because overexpression of malic enzyme is found in HCC associated with progression-free survival and lower overall survival of patients.⁴⁰

Because glucose or fructose is transported across the cell membrane by glucose transporters (GLUTs), trapped by hexokinase (predominantly HK2) and converted into lactate catalyzed by intracellular lactate dehydrogenase (LDH) with the reduction of nicotinamide adenine dinucleotide (NADH) to NAD+, followed by lactate export via the monocarboxylate transporters (MCTs), the production of lactate determined from the cell media samples is dependent on the interplay of the activities and concentrations of all enzymes in the glycolytic pathway. The three times faster ²H lactate production rate from glucose in vitro than that from fructose in vitro (Table 1 and Figure 2C) implies that [6,6′-²H₂] glucose is transported across the cell membrane and phosphorylated faster than [6,6-²H₂] fructose, followed by subsequent conversion and lactate export, resulting in a higher overall production of lactate. In addition, a comparable production rate of ²H glx for the two precursors in vitro (Table 1 and Figure 2E), in comparison with a higher production rate of ²H lactate in vitro for ²H glucose, indicates that the two precursors contribute to the TCA cycle similarly, suggesting that transport and upstream metabolism can perform in excess of the carbons needed to maintain the TCA cycle.

The kinetics of in vitro metabolites (Figure 2C–F) show a linear increase in the concentration of extracellular lactate, whereas concentrations of extracellular glx and alanine reached plateaus, indicating that influx and efflux of glx and alanine across the cell membrane reached equilibrium in ~8 h. This suggests that there is little exchange of glx in the cell with glx in the media. Given the concentrations observed, it is very likely that the detected ²H alanine and ²H glx is a result of exchange as a result of media change. On the contrary, lactate transport, mediated by MCTs,^41,42 was not saturated and continued to occur over time. This is likely due to high expression of MCT4 in HepG2 cells that is predominantly responsible for lactate efflux.^43,44 Because of the Warburg effect, cancer cells exhibit an increased production of lactate in the cytoplasm. The excessive accumulation of lactate acidifies the intracellular environment that threatens the survival of cancer cells; therefore, the overexpression of MCT4 is observed in multiple cancers,⁴⁵ including liver cancer.⁴⁶

A comparison of in vivo versus in vitro metabolic rates (Table 1) shows that the values for the ratio of ${\nu }_{lac-glu\text{,}invivo}/{\nu }_{lac-fru\text{,}invivo}$ and of ${\nu }_{HDO-glu,invivo}/{\nu }_{HDO-fru\text{,}invivo}$ (~2 for in vivo lactate; ~1 for in vivo HDO) are comparable with those observed in vitro (~3 for in vitro lactate; ~1.5 for in vitro HDO). The higher production rate of lactate from glucose indicates that the anaerobic metabolism of glucose in HepG2 cells is more active. The in vitro HDO production rate of glucose is comparable with that of fructose, indicating their contribution to the TCA cycle in vitro is similar. Each of the metabolic rates in vivo is slower than that in vitro, which can be caused by the differences between an in vivo and an in vitro environment. Compared with the in vitro experiment that occurs in a controlled environment, the in vivo experiment performs in a living organism, resulting in differences of several factors that affect the metabolic rates, such as the oxygen level, temperature, and the concentration of precursors (i.e., a constant concentration of precursors supplied in the cell culture, in contrast to varied concentrations of precursors present in the tumor over time due to perfusion). In addition, the slower in vivo rates derived per million cells may further be caused by the inaccurate estimation of cells per cm³ (cc) of tumor (1 cc tumor = 10⁹ cells for current estimation) due to tumor heterogeneity.

We were unable to monitor the kinetics of deuterated lactate due to insufficient signal-to-noise ratio (SNR) in each individual CSI spectrum, indicating the amount of deuterated lactate accumulated at each time point reached the detection limit with the current experimental setup and tumor model. Because of the relatively insensitive ²H detection (compared with ¹H and ¹³C) and broad ²H lines (because of short T_2,2H), a longer scan time or higher lactate production is needed to achieve detectable SNR. In comparison with hyperpolarized ¹³C MRI that offers high ¹³C signal enhancement of orders of magnitude beyond the Boltzmann polarization and thus allowing to image hyperpolarized ¹³C lactate dynamics in vivo,⁴⁷ DMI suffers from low signal magnitude, limiting its capability of probing dynamic imaging of ²H lactate with high spatial resolution. From this point of view, in vivo imaging by hyperpolarized ¹³C^48,49 may be superior to DMI for studying glycolysis quantitatively in liver tumors, providing insights into tumorigenesis and malignancy.⁵⁰ On the other hand, because of the short window of detection, hyperpolarized ¹³C MRI fails to capture downstream metabolites produced from oxidative pathways, that is, H₂O from the TCA cycle, which is critical in pathological circumstances; however, this is achievable by DMI. Taken together, future studies would likely benefit from utilizing these complimentary techniques together to more deeply characterize metabolic flux.

The current study conducted experiments using subcutaneous mouse models in which tumors were located around the right flank. This allowed for the characterization of ²H fructose metabolism in the tumor. However, if tumors were in the liver, this will pose a challenge for demarcating liver tumors from the surrounding normal liver at later time points in the DMI, and therefore the analysis could be complicated by the contamination of the background signal from adjacent liver tissue. Even with a slice-selective excitation pulse with a slice thickness that only excites signals in the tumor region, the inflows of HDO from the surrounding liver tissue could still contaminate the acquired signal in the tumor. We hope to develop methods to address this by better spatially and temporally resolving the production of HDO, ²H glx, and ²H lactate in time using recently developed bSSFP sequences⁵¹ and modeling local diffusion of HDO.

DMI experiments in the livers of healthy mice (Figure 5) revealed a similar HDO production rate from ²H fructose in the liver tumor and normal liver, indicating the comparable contribution of ²H fructose to the TCA cycle. Additionally, the in vivo production rates of ²H glx and HDO in the mouse normal livers are comparable; however, no ²H glx was observed in liver tumors. It is possible that detection of ²H glx in tumors is limited because in vivo many tumors have increased glutaminolysis,^52,53 indicating a much higher fraction of the TCA cycle is labeled from glutamine compared with that from glucose. Such increased glutaminolysis leads to a higher flux of unlabeled glutamate into the TCA cycle from unlabeled glutamine, as opposed to the flux of ²H labeled glutamate out of the TCA cycle from ²H glucose or ²H fructose. This would result in labeled α-ketoglutarate having reduced labeling of glutamate in comparison with the amount of HDO generated. Future experiments are necessary to delve deeper into the mechanisms that drive this labeling. Furthermore, during the second round of the TCA cycle, the conversion of ²H isocitrate to α-ketoglutarate or the conversion of ²H citrate to isocitrate results in the complete elimination of deuterium in the TCA intermediates (Figure S9),²³ indicating that ²H glx observed in the mouse normal livers is only produced from the first-round reactions. Therefore, the observed c(HDO)/c(²H glx) ratio (Figures 5E, S6e, and S7e) during the initial period implies the production of HDO against the production of ²H glx in the first round of TCA cycle.

De Feyter et al. used [6,6′-²H₂] glucose as a metabolic precursor to detect ²H labeled glycogen and glucose in the human liver using a surface RF coil at a clinical field strength of 3 T,¹⁴ demonstrating the potential of translating DMI into human studies at lower field. The concentration of ²H labeled glycogen and glucose in their work was 2.7 ± 0.3 mM. In the current work, we detected ²H labeled metabolic products at concentrations higher than 3 mM. Given that ²H detection is concentration-dependent and fructose is predominantly metabolized in the liver, this suggests that the concentrations of metabolic products in our work fall within the ²H detection range at 3 T, indicating the possibility of translating our findings to human studies at clinical field strength.

Compared with glucose, which plays a crucial role in every organ of the human body, fructose is predominantly metabolized in the liver, making it a useful imaging agent to study liver-related diseases. The metabolism of fructose follows ketohexokinase (KHK) and hexokinase (HK) pathways, where the metabolism via the HK pathway relates to the downstream glycolytic metabolism and pentose phosphate pathway (PPP) that are coordinated by changes in a wide range of oncogenic transformations, including HIF-1α and PI3K.⁵⁴ Therefore, changes in fructose metabolism could provide key insights into the therapeutic targeting of these pathways. The increased dietary consumption of fructose further leads to an increased risk of nonalcoholic fatty liver disease,⁵⁵ and thus makes fructose a valuable biomarker to study this disease. Previous studies of [6,6′-²H₂] glucose metabolism in rat and human livers showed overlapped ²H chemical shifts of ²H glucose and its metabolic product ²H glycogen in the image,^14,18 making the study of ²H glucose kinetics and the measurement of glycogen production rate challenging; however, characterization of kinetics and rates plays an important role in disease assessment and diagnosis, such as nonalcoholic fatty liver disease and type 2 diabetes. By contrast, [6,6′-²H₂] fructose metabolism and its metabolic product ²H glx can be distinguishable and monitored over time in mouse normal liver. This suggests ²H fructose can be applied quantitatively to study liver-associated disease. Therefore, considering the unique role of fructose in these metabolic processes, it is important to develop ²H fructose as a novel DMI agent for studying liver disease, providing the simpleness and powerfulness of the DMI method.

5 |. CONCLUSION

In summary, we demonstrate for the first time using [6,6′-²H₂] fructose as a novel DMI probe to study liver tumor metabolism. The idea follows the recent demonstrations of using [6,6′-²H₂] glucose to monitor metabolism in various tumor models, as well as the important role of fructose in liver metabolism. As fructose metabolism via the HK pathway relates to the metabolic switch of liver tumors to reduced flux through KHK and downstream glycolysis, the metabolic products from [6,6′-²H₂] fructose and [6,6′-²H₂] glucose metabolism in HepG2 tumor models observed by DMI are similar. Evaluation of [6,6′-²H₂] fructose and [6,6′-²H₂] glucose metabolism quantitatively both in vitro and in vivo identifies faster metabolic rates in vitro and further demonstrates different kinetics between the two precursors: the production of ²H lactate, predominantly via the glycolytic pathway, is faster from the latter than the former, resulting from the interplay of the activities of all enzymes involved in the process including glucose transport rate across the cell membrane, followed by subsequent conversion to lactate and export. Furthermore, quantitation of HDO production in vivo can serve as a simplified means to evaluate ²H fructose or ²H glucose consumption in HepG2 tumor models. A comparison of percentage conversion from the ²H precursors to HDO indicates that HDO produced from ²H fructose likely localized to the liver tumor, while HDO produced from ²H glucose is from liver tumor and the surrounding organs. Additionally, the study of [6,6′-²H₂] fructose metabolism in mouse normal livers indicates the capability of characterizing ²H fructose kinetics; by contrast, the metabolism of [6,6′-²H₂] glucose in the liver showed overlapped signals in the image, making kinetic quantification difficult and limiting its ability to discern tumor from surrounding normal tissue. This suggests the important role of fructose in liver metabolism and many liver-related diseases, making it a useful DMI probe for the study of these disease-associated metabolic processes.

Abbreviations used:

2D

two-dimensional

²H or D

deuterium

CSI

chemical shift imaging

DMI

deuterium metabolic imaging

FDG–PET

fluorodeoxyglucose–positron emission tomography

FOV

field of view

GLUT

glucose transporter

glx

glutamate/glutamine

HDO

deuterated water

HCC

hepatocellular carcinoma

HK

hexokinase

KHK

ketohexokinase

LDH

lactate dehydrogenase

MCT

monocarboxylate transporter

MRSI

magnetic resonance spectroscopic imaging

MSM-d₆

methylsulfonylmethane-d₆

NAD

nicotinamide adenine dinucleotide

PBS

phosphate-buffered saline

PPP

pentose phosphate pathway

RARE

rapid acquisitions with relaxation enhancement

TCA

tricarboxylic acid

DATA AVAILABILITY STATEMENT

The data generated in this study are available at https://drive.google.com/drive/folders/1CVQfIoXnJFg1nr9ZZp-NQsid7lCeiE6Q.