Deuterium Metabolic Imaging Enables the Tracing of Substrate Fluxes Through the Tricarboxylic Acid Cycle in the Liver

ABSTRACT

Alterations in tricarboxylic acid (TCA) cycle metabolism are associated with hepatic metabolic disorders. Elevated hepatic acetate concentrations, often attributed to high caloric intake, are recognized as a pivotal factor in the etiology of obesity and metabolic syndrome. Therefore, the assessment of acetate breakdown and TCA cycle activity plays a central role in understanding the impact of diet‐induced alterations on liver metabolism. Magnetic resonance‐based deuterium metabolic imaging (DMI) could help to unravel the underlying mechanisms involved in disease development and progression, however, the application of conventional deuterated glucose does not lead to substantial enrichment in hepatic glutamine and glutamate. This study aimed to demonstrate the feasibility of DMI for tracking deuterated acetate breakdown via the TCA cycle in lean and diet‐induced fatty liver (FL) rats using 3D DMI after an intraperitoneal infusion of sodium acetate‐d3 at 9.4T. Localized and nonlocalized liver spectra acquired at 10 time points post‐injection over a 130‐min study revealed similar intrahepatic acetate uptake in both animal groups (AUC_FL = 717.9 ± 131.1 mM▯min⁻¹, AUC_lean = 605.1 ± 119.9 mM▯min⁻¹, p = 0.62). Metabolic breakdown could be observed in both groups with an emerging glutamine/glutamate (Glx) peak as a downstream metabolic product (AUC_FL = 113.6 ± 23.8 mM▯min⁻¹, AUC_lean = 136.7 ± 41.7 mM▯min⁻¹, p = 0.68).

This study showed the viability of DMI for tracking substrate flux through the TCA cycle, underscoring its methodological potential for imaging metabolic processes in the body.

Deuterium metabolic imaging (DMI) combined with an acetate infusion demonstrated the viability of DMI for tracking substrate flux through the TCA cycle. Using DMI, we documented the acetate breakdown by identifying an increasing Glx peak as a resulting metabolic product in healthy and fatty livers.

Abbreviations

AUC

area under the curve

CSI

chemical shift imaging

DMI

deuterium metabolic imaging

FL

fatty liver

Glx

glutamine/glutamate

GTT

glucose tolerance test

HCL

hepatic lipid content

HDO

deuterated water

HFD

high‐fat diet

MASLD

metabolic dysfunction‐associated steatotic liver disease

MRS(I)

magnetic resonance spectroscopy (imaging)

PET

positron emission tomography

SD

standard diet

SEM

standard error of the mean

SNR

signal‐to‐noise ratio

TCA

tricarboxylic acid

(F)FA

(free) fatty acids

1. Introduction

Obesity and weight‐related metabolic disorders have become increasingly prevalent in recent years, elevating metabolic dysfunction‐associated steatotic liver disease (MASLD) to the leading cause of chronic liver disease worldwide [1]. Impaired insulin sensitivity in white adipose tissue in the obese state causes unrestrained triglyceride lipolysis, which leads to uncontrolled free fatty acids (FFA) influx into the liver ultimately resulting in MASLD [2]. Numerous other mechanisms including increased lipogenesis, decreased lipid export, impaired mitochondrial function, and beta‐oxidation have been implicated in developing MASLD. In addition, alterations in tricarboxylic acid (TCA) cycle activity are observed in individuals with MASLD and related conditions [3]. Thus, a better comprehension of nutrient flux and TCA cycle in different metabolic states (i.e. lean vs. obese) is important to understand this complex metabolic liver disease.

Lately, magnetic resonance‐based deuterium metabolic imaging (DMI) has been implemented to evaluate the metabolism of various pathologies, primarily in the brain [4, 5, 6]. Applications in the abdominal region have been introduced in animal models and high‐field human investigations [7, 8] and DMI has recently also been demonstrated for monitoring inflammation‐related metabolic alterations [9]. The proven reliability and robustness of DMI, even at clinical magnetic field strengths [10], suggest its potential translation into clinical practice in the near future.

The availability and metabolism of the shortest‐chain fatty acid acetate are associated with metabolic syndrome, type 2 diabetes, and obesity [11]. Understanding the role of acetate and TCA metabolism in MASLD may shed light on the metabolic changes associated with this disease. Since investigating acetate metabolism with conventional techniques such as hyperpolarized ¹³C MRS or PET comes with specific technologic or methodologic limitations, and thermal equilibrium ¹³C MRS data of fatty liver are prone to lipid signal contamination, in vivo DMI offers a promising complement or alternative to gain deeper insights into these processes.

Despite acetate's central role in various metabolic pathways, including the TCA cycle (see Figure 1), its assessment via DMI remains limited. While De Feyter et al. [13] demonstrated the feasibility of acetate DMI in the brain, a subsequent study applied it to evaluate myocardial energy metabolism [12]. However, the application of acetate DMI in imaging hepatic metabolism has yet to be investigated.

FIGURE 1.

Metabolic breakdown of [²H₃]acetate: After reaching the hepatocyte, the acetate transforms into acetyl‐CoA, occurring either in the cytosol where it serves in fatty acid synthesis or within the mitochondrion. There, the acetyl‐CoA subsequently enters the tricarboxylic acid (TCA) cycle. The first deuteron gets lost during the conversion to citrate, resulting in only two labels in the detectable downstream Glx. In the TCA cycle steps from succinate to fumarate and from malate to oxaloacetate, the remaining deuterons are potentially transferred to water, contributing to an HDO increase throughout the experiment [12].

Studying the TCA cycle, we mainly anticipate the emergence of glutamine/glutamate (Glx) as a metabolic downstream product. Glx signals have been detected in ²H and uniformly labeled glucose DMI experiments in the brain [13, 14]. However, in the liver under postprandial conditions, glucose is primarily incorporated into glycogen, limiting label detection in Glx [15, 16]. Thus, acetate is a more suitable marker for investigating TCA cycle‐related metabolic processes and detecting downstream Glx.

Leveraging the capability of DMI to trace the fate of deuterated substrates during metabolic breakdown, our study aimed to assess potential alterations in hepatic TCA cycle metabolism in rodents with MASLD using acetate as a metabolic tracer.

2. Experimental

All animal procedures adhered to the European Commission's Directive 2010/63/EU and FELASA guidelines for animal research, receiving approval from the Austrian Federal Ministry of Science, Research, and Economy (license number: BMBWF 2020‐0.078.441).

The animal preparation and DMI experiments followed established protocols, as described previously [15]. Male Sprague–Dawley rats, aged 4 weeks (n = 14, Janvier Laboratories, France), were evenly divided into two groups of standard (SD, n = 7) and high‐fat diet (HFD, n = 7). While the SD group was fed the regular carbohydrate‐rich chow (LASQCdiet Rod16, altromin, Lage, Germany), the HFD group obtained a diet with 60% of calories as fat (Research Diets D12492, New Brunswick, NJ, USA). After 6 weeks, this induced insulin resistance and fatty liver (FL) disease in the HFD group, as confirmed by a glucose tolerance test (GTT) and ¹H magnetic resonance spectroscopy of the liver, respectively. The GTT was performed as described previously [15], one week before the DMI measurements.

For the DMI, sodium acetate‐d₃ ([²H₃]Na‐acetate) (Sigma‐Aldrich, St. Louis, Missouri, USA) was dissolved in water (aqua ad iniectabilia) to a concentration of 2 M and intraperitoneally infused over 120 min [13] at a weight‐dependent infusion rate of 30.8–46.8 μL/min, resulting in a total administration of 1.5 g/kg body weight of acetate. This acetate concentration and infusion rate proved to be physiologically tolerable in both groups of animals.

2.1. Imaging Experiments

MR experiments were performed when the HFD group had developed FL disease, as verified by ¹H magnetic resonance spectroscopy (MRS). Following an overnight fast (12–16 h), rats were anesthetized with isoflurane (2%–3% in air). During anesthesia, a heating pad kept their body core temperature stable and breathing and heart rate were monitored (SA Instruments, Stony Brook, NY, USA).

2.2. Deuterium Metabolic Imaging

For the acetate DMI experiments, we used a 9.4T BioSpec 94/30 (Bruker BioSpin, Ettlingen, Germany) MR system with ParaVision 360.3.3. The animals were positioned prone with the liver atop a ²H/¹H surface coil (Rapid, Rimpar, Germany). DMI measurements were performed according to the procedure set up in a prior study [15]. To acquire the anatomical ¹H MR images, an axial T₁‐weighted FLASH sequence (TR = 30 ms, FA = 70°, NA = 20, matrix = 120 × 120, FOV = 50 × 36 mm²) was applied. Nonlocalized ²H MRS control data were obtained using a pulse‐acquire sequence (TR = 400 ms, NA = 384, 2048 spectral points, bandwidth = 5.21 kHz, acquisition time = 2.5 min). The unlocalized measurements served only as a fast control during the measurements and were not quantified.

Localized ²H MRS data were acquired with a 3‐dimensional chemical shift imaging (CSI) sequence (TR = 100 ms, FA = 52.7°, NA = 36, 512 spectral points, matrix = 12 × 12 × 8 mm³, FOV = 50 × 36 × 20 mm³, bandwidth 6.06 kHz, acquisition time = 10.3 min, no motion correction applied). The infusion started after measuring the natural abundance water signal, with nonlocalized and localized measurements performed 10 times over a 130‐min study duration.

The intrahepatic fat content (HCL) was calculated from a single‐voxel MRS, employing short TE stimulated echo acquisition mode (STEAM) (TE = 5.5 ms, TR = 3000 ms, FA = 69.5°, NA = 64, 2048 spectral points, VOI = 6 × 6 × 6 mm³, bandwidth = 7.9 kHz, respiratory gated, acquisition time = 3.2 min), following established protocols from prior research [17].

2.3. MRS Processing

For pre‐processing and analyzing the DMI data we used the AMARES [18] time domain algorithm within jMRUI (v7.0) [19]. Spectra of interest were manually chosen from the CSI grid nonblinded, based on the respective voxel's position within the liver parenchyma, ensuring adequate SNR (visual check) and sufficient distance from the subcutaneous fat layer. The accepted spectra were located in the ventral part of the liver close to the surface coil to ensure sufficient signal. The selected spectra were frequency‐aligned and averaged to generate one spectrum per time point, then quantified using linear least‐squares fitting with a predefined set of prior knowledge. In this set, the starting value for the signal position for acetate was set to 1.9 ppm (constrained to the range of 1.7–2.0 ppm), for Glx to 2.35 ppm (range 2.1–2.6 ppm), and for water to 4.7 ppm (range 4.6–4.8 ppm). The linewidth was soft constrained to 10–40 Hz for acetate, 10–30 Hz for Glx, and 30–75 Hz for water. All peaks were fitted with a Lorentzian line shape. Figure 2 shows an exemplary fitted spectrum with the three identified peaks. The liver's naturally abundant deuterated water (HDO) signal before infusion served as an internal reference for quantification (8.94 mM) [13]. To convert the signal intensities to absolute concentrations (mM), the number of deuterons per molecule (three for acetate and two for Glx, Figure 1) was used as a normalization factor.

FIGURE 2.

Example of a DMI spectrum showing the three signals (water, Glx, acetate) that could be identified following the acetate infusion. (a) The CSI spectrum averaged over all time points, (b) the individual fitted peaks with HDO (dark blue), Glx (lilac), and acetate (bright blue), and (c) the entire fitted spectrum. For visual reasons, the averaged spectrum in panel (a) was zero‐filled.

The HCL was calculated from ¹H STEAM data via the signal intensities of water (W) and lipids (L, signals at 1.3 and 0.9 ppm) to HCL = [L/(L + W)], expressed as a percentage. The average HCL of all SD and HFD animals was determined, respectively, and the standard deviation was calculated. The results are presented as mean ± standard error of the mean (SEM).

2.4. Statistics

The results of the phenotyping characterization and the metabolite concentrations over time are presented as mean ± standard error of the mean (SEM). GTT was additionally evaluated by calculating the area under the curve (AUC). The AUC was also determined to assess the time course of the DMI‐based hepatic acetate concentrations. The AUC for the GTT and intrahepatic glucose and fatty acid concentration were determined using the trapezoid rule (AUC = (C₁ + C₂)/2(t₂ − t₁)). Group distinctions were analyzed via a two‐way repeated measures ANOVA, with statistical significance at p < 0.05. To test if individual time points show significant differences in the acetate or Glx signal between the two groups, we performed a post‐hoc analysis. Effect size was calculated as an additional parameter to evaluate the different phenotypes of the two animal groups. GraphPad Prism 8.2.1 (GraphPad, San Diego, CA, USA) was used for statistical calculations.

3. Results

The phenotypes of the healthy and metabolically compromised animals were confirmed by GTT (AUC_HFD = 20,595 ± 1370 g/dL▯min⁻¹, AUC_SD = 15,590 ± 1128 g/dL▯min⁻¹, p = 0.01) and ¹H MRS (mean HCL HFD = 5.8 ± 2.0% vs. SD = 0.9 ± 0.4%, p < 0.01). The effect size indicated notable differences between the two groups across all assessed parameters (Table 1). Acetate DMI acquisition using 3D CSI provided adequate SNR and spectral resolution in both metabolically healthy and impaired rodents, enabling the identification of three metabolite peaks across multiple ²H MRS imaging (MRSI) voxels (SNR_SD = 36.8 ± 4.6, SNR_HFD = 22.9 ± 1.7). Figure 3 displays localized spectra for one representative animal per group during the acetate infusion experiment, demonstrating the presence of the acetate and the HDO peaks and identifying a Glx peak.

TABLE 1.

Phenotypic characteristics (mean ± SEM) of the study animals.

	SD a (n = 7)	HFD b (n = 7)	p	Effect size
Mass [g]	462.1 ± 8.2	539.2 ± 17.6	< 0.01	2.97
Intrahepatic fat content [%]	0.9 ± 0.1	5.8 ± 0.6	< 0.01	12.20
Fasted blood glucose [mg/dL]	105.2 ± 2.9	114.8 ± 3.0	0.04	1.05
GTT c [mg/dL] 60 min	141.6 ± 6.6	187.2 ± 11.1	< 0.01	2.17
90 min	128.0 ± 1.2	151.0 ± 7.3	< 0.01	6.01

^a Standard diet.

^b High‐fat diet.

^c Glucose tolerance test.

FIGURE 3.

²H MR spectra after glucose injection in one exemplary animal of the healthy (SD, (a)) and the FL group (HFD, (b)). The spectrum in the front shows the baseline signal before infusion. A Glx peak emerged as a downstream metabolic product during the acetate breakdown. The SD animal clearly indicates the rise in HDO due to label loss in the TCA cycle (Figure 1). The panel on the right displays color‐coded amplitudes of acetate and Glx signals in ²H MRSI voxels over time co‐registered with liver parenchyma.

After 1 h of infusion, the intrahepatic acetate concentration in obese animals was similar to that in healthy animals (c_HFD = 5.6 ± 1.1 mM, c_SD = 6.9 ± 1.5 mM, p = 0.42) (Figure 4a). The area under the curve also revealed no significant group effect (AUC_HFD = 717.9 ± 131.1 mM▯min⁻¹, AUC_SD = 605.1 ± 119.9 mM▯min⁻¹, p = 0.62) and no significant time‐group interaction (p = 0.96). However, the time effect was significant (p = 0.0002).

FIGURE 4.

Acetate (a), Glx (b), and HDO (c) concentrations during the 130‐min study in SD (blue) and HFD (purple) animals. Data are mean ± SEM.

We could observe an increase in Glx as a metabolic product of the acetate breakdown in both animal groups throughout the measurement (Figure 3 and Figure 4b), with no significant difference between the groups (AUC_HFD = 113.6 ± 23.8 mM▯min⁻¹, AUC_SD = 136.7 ± 41.7 mM▯min⁻¹, p = 0.68). Here, both the time‐group interaction (p = 0.02) and the time effect (p = 0.007) of the ANOVA showed a significant difference. Despite the difference in the time‐group effect, the post‐hoc analysis revealed no significant differences for individual time points in Glx or acetate signal with all p‐values above 0.81. The continuous rise in HDO over time was similar for both groups (p = 0.80, Figure 4c), showing no difference in water labeling between control and diseased animals.

4. Discussion

This study investigated intrahepatic acetate metabolism in healthy and FL rats using DMI. Based on findings from previous investigations where we could not observe hepatic Glx signal after ²H glucose infusion [15], we tested acetate as an alternative tracer to gain insights into hepatic TCA cycle activity. We did not observe a significant difference in hepatic acetate uptake between healthy and impaired animals. Nevertheless, we documented the acetate breakdown by identifying an increasing Glx peak as a resulting metabolic product in both groups.

4.1. Label Transition in the TCA Cycle

Both animal groups consistently exhibited a clear increase in hepatic HDO concentration over time, suggesting a label transfer from acetate to water during the metabolic breakdown. While kinetic isotope effects arising from proton substitution with deuterons offer one explanation, in vivo studies have shown their relatively minor impact [20]. Labels are predominantly lost in the TCA cycle due to keto‐enol tautomerism (Figure 1). Within the mitochondrion, acetyl‐CoA synthetase 1 (ACSS1) transforms the labeled acetate into ²H acetyl‐CoA, initiating its entry into the TCA cycle. During the conversion from acetyl‐CoA to citrate, the first deuteron is passed over to the HDO. The breakdown continues with ²H citrate turning into isocitrate and α‐ketoglutarate. Parts of the latter undergo exchange with glutamate dehydrogenase, resulting in ²H glutamine and/or ²H glutamate formation [21, 22]. This contributes, at least partially, to the observed increase in intrahepatic Glx concentration in both rodent groups. Following this, α‐ketoglutarate leads to the generation of succinate via succinyl‐CoA and the conversion to fumarate. Here, the potential loss of the second ²H atom may occur. Subsequently, fumarate converts to malate, which finally transforms into oxaloacetate, possibly losing the last deuteron [12].

It is not clear yet if an increase in HDO can be attributed mostly to metabolic processes in the liver, as HDO rapidly interacts with the body's water pool and therefore will not solely indicate liver metabolism [20]. However, the concurrent increase in HDO and Glx concentrations over time reveals metabolic activity in the liver, suggesting that such a strong rise in hepatic HDO very likely indicates hepatic metabolism. Nonetheless, the identifiable Glx signal emerging in both animal groups demonstrates that we can trace the metabolic breakdown of acetate via the TCA cycle using DMI.

4.2. Acetate for Imaging Metabolic Processes in Vivo

Administering acetate as a substrate for metabolic imaging proved challenging, requiring careful consideration of injection quantity and methodology to ensure both physiological tolerability and adequate signal intensity for peak identification. To achieve a detectable acetate signal, the injection solution had to be produced in a way that the osmolarity turned out to be very high. This led to an instant death of the animals when injected intravenously as a bolus in pilot experiments for this study. Infusing the acetate intraperitoneally over a long period ensures tolerability while providing a satisfactory signal. The substantial intragroup variability underscores the complexity of acetate administration for imaging purposes, highlighting its intricate suitability as a substrate. The resulting injection duration of 120 min used in this experiment imposed certain constraints on observing downstream metabolic processes (see limitations section). This limitation may partly explain the rare utilization of acetate DMI to date. Nevertheless, combining DMI with acetate infusion presents a viable methodological approach for metabolic imaging, despite the absence of discernible group differences in our study. We present a physiologically bearable experimental procedure that is comparable to glucose DMI measurements under similar experimental conditions, demonstrating the potential of acetate as a metabolic tracer.

4.3. Prospects for DMI and Future Perspectives

In this study, we demonstrated acetate breakdown by detecting the downstream metabolic product Glx in the DMI spectra. Consequently, acetate emerges as a more suitable substrate for imaging short‐term hepatic lipid metabolism, contrasting with prior DMI research that failed to detect Glx peaks following palmitic acid administration consistently [15]. Back then, we attributed this in part to the relatively small Glx pool size in the liver compared to the brain, where Glx has been measured with DMI before [13]. Nonetheless, our current study shows that the hepatic lipid metabolism yields sufficient amounts of Glx to be detected with DMI.

To enhance the understanding of the role of mitochondrial lipid oxidation in the pathogenesis of MASLD, Befroy et al. [23] introduced a ¹³C‐MRS‐based method for noninvasive measurement of mitochondrial oxidative rates in vivo. In this context, they demonstrated the significance of acetate in elucidating the potential impact of altered oxidative metabolism in the etiology of fatty liver. Since acetate metabolism can serve as a surrogate for fatty acid oxidation [13] and given the importance of this topic, DMI could offer a robust, more sensitive, and technically simpler alternative to ¹³C liver MRS, additionally providing increased morphologic or time‐resolution. These characteristics make DMI an attractive tool for potential integration into clinical investigations. However, the primary drawbacks, namely prolonged scan durations and limited spatial resolution, pose challenges that must be tackled. Recent research has aimed to address these limitations, particularly emphasizing on expediting data acquisition methods [8, 24, 25].

4.4. Limitations

To ensure physiological tolerance to acetate in animals, a bolus injection is impractical and necessitates infusion over an extended period. With our study design, this limitation precludes the comprehensive monitoring of acetate breakdown and degradation. Nonetheless, demonstrating the viability of imaging acetate metabolism using DMI and tracing the acetate metabolism at least in the first steps of the TCA cycle is achievable with acetate administration via infusion.

We investigated acetate DMI at 9.4T with only small differences in the chemical shifts of acetate and Glx being resolved, despite the high field strength. Thus, with the current knowledge and technical possibilities, it may be challenging to translate acetate DMI into clinics. Further research could deal with finding a way to produce better tolerable acetate solutions that still yield sufficient SNR to detect the acetate signal in DMI spectra.

Furthermore, no significant differences in acetate uptake were observed between animals with healthy and fatty livers, thus limiting the strength of our conclusions. This can be partly related to the small sample size in the presented experiment.

Another potential limitation of exact quantitative results of any DMI study using natural abundance HDO signal as a concentration reference may be the variation of natural abundance ²H across the planet. The calculation of De Feyter and colleagues [13] in the first paper introducing DMI for metabolic imaging in vivo is based on a publication by Harris et al. [26], however, we do not know if these values are specific to Connecticut, which would be different to our site. Our calculations are based on De Feyter's work, which may lead to an inaccuracy in the numbers for deuterated metabolite concentrations.

Future studies should explore acetate DMI combined with PET or ¹³C MRS to validate the differentiation ability of DMI for tracing hepatic acetate metabolism and enhance the understanding and utility of DMI in assessing liver metabolism and disease progression. Acetate DMI could be complemented with [¹¹C]acetate PET to address this limitation, as previous research has shown elevated hepatic uptake of [¹¹C]acetate in subjects with hepatic liver disease (steatosis) [27]. Additionally, this study identified a correlation between [¹¹C]acetate uptake and the severity of steatosis. Therefore, integrating multimodal acetate PET/DMI could elucidate differences in acetate metabolism between healthy and diseased livers. In this context, however, the different metabolic conditions between [¹¹C]acetate for PET tracing natural levels of plasma acetate and the ²H acetate used for DMI acting as a substrate at much higher concentrations have to be considered.

5. Conclusion

We could demonstrate the feasibility of DMI for assessing acetate metabolism in rats with healthy and fatty livers, revealing the metabolic breakdown by detecting intrahepatic Glx production in both groups. Combined with complementary methods such as PET, mass spectrometry, or electron microscopy, DMI could contribute to unraveling the complexity of the nutrient flux in different metabolic conditions and its contribution to FL and gain a deeper insight into metabolic alterations related to FL disease.

Supporting information

Figure S1. Original spectrum, fitted spectrum, and individual fitted peaks of the two exemplary animals shown in Figure 3 (SD in the left panel, HFD in the right) 91 min after the start of the acetate infusion.

Data Availability Statement

The data that support the findings of this study are openly available in Open Science Framework at https://osf.io/y8pt5/, reference number DOI: 10.17605/OSF.IO/Y8PT5.