Hyperpolarized MRI Detects Increased [1‐¹³C]Pyruvate Metabolism in Early Liver Regeneration

ABSTRACT

Surgical resection remains the primary curative approach for malignant liver tumors and liver regeneration is essential for patient recovery following major hepatic resection. While its molecular and cellular mechanisms have been extensively studied, the in vivo metabolic dynamics underlying early regeneration remain incompletely characterized. Hyperpolarized [1‐¹³C]pyruvate MRI (HP‐MRI) offers a unique, noninvasive method to assess real‐time metabolic fluxes in regenerating liver tissue. Twelve male Wistar rats were randomized to either 70% partial hepatectomy (PH; n = 6) or nonsurgical controls (n = 6). On postoperative day 1, all animals underwent HP‐MRI and multiparametric proton MRI. Metabolic fluxes were quantified using area‐under‐the‐curve ratios for lactate‐to‐pyruvate (L/P), alanine‐to‐pyruvate (A/P), and lactate‐to‐alanine (L/A). The PH group showed significantly higher L/P (0.267 [95% CI: 0.225–0.310] vs. 0.168 [95% CI: 0.135–0.200]; p < 0.001) and A/P (0.236 [95% CI: 0.153–0.319] vs. 0.150 [95% CI: 0.128–0.172]; p = 0.028) ratios compared to controls, indicating increased exchange of pyruvate to lactate and alanine. L/A ratios remained unchanged. These findings were supported by elevated biochemical markers of hepatic injury and changes in quantitative MRI parameters, including reduced ADC and IVIM flow fraction. HP‐MRI revealed increased glycolytic and transaminase activity during early liver regeneration, consistent with the metabolic demands of hepatocyte proliferation. These results demonstrate the feasibility of HP‐MRI for noninvasive metabolic assessment of liver regeneration in vivo and additionally provide insights into early regenerative metabolism. This suggests HP‐MRI as a promising tool for assessing postoperative recovery in patients undergoing major hepatectomy.

Hyperpolarized [1‐¹³C]pyruvate MRI detected increased pyruvate‐to‐lactate and pyruvate‐to‐alanine exchange during early liver regeneration following 70% partial hepatectomy in rats. These findings support the use of hyperpolarized MRI for noninvasive assessment of regenerative metabolism.

1. Introduction

Surgical resection remains the primary curative treatment for malignant liver tumors [1, 2]. Liver regeneration is a critical determinant of postoperative recovery and long‐term outcome in patients undergoing major hepatic resection. Although the liver has a remarkable capacity to regenerate, the extent and efficiency of this response vary among individuals. This is influenced by factors such as underlying liver disease, age, comorbidities, and perioperative insult. Insufficient regenerative capacity may result in posthepatectomy liver failure, a serious complication associated with high morbidity and mortality [3, 4].

Regeneration is a highly orchestrated process that requires substantial energy and metabolic reprogramming to support hepatocyte proliferation and tissue remodeling [5]. Despite extensive research into the cellular and molecular mechanisms of liver regeneration, the in vivo metabolic dynamics during this process remain incompletely characterized, limiting early identification of patients at risk and hindering the development of targeted interventions. This limits our ability to predict regenerative success, identify patients at risk of liver failure, or develop metabolic interventions to support recovery. Recent advances in metabolic imaging, particularly hyperpolarized (HP) MRI, offer a unique opportunity to noninvasively monitor real‐time metabolic fluxes in living tissue by enhancing signal intensity more than 20,000‐fold [6]. Among the first applications of HP‐MRI were studies of altered cancer metabolism, including the Warburg effect, and early treatment response in preclinical models [7, 8]. Since then, HP‐MRI has been applied to various liver pathologies, including liver injury, fatty liver disease, and altered gluconeogenesis [9, 10, 11, 12, 13]. In oncology, HP‐MRI has shown promise for diagnosing prostate cancer [14] and evaluating treatment response in patients with prostate cancer, as well as for metabolic characterization of liver metastases [15, 16]. These developments highlight its potential for translational applications in both hepatic and oncologic diseases. HP‐MRI using [1‐¹³C]pyruvate enables dynamic assessment of key metabolic conversions, including the enzymatic exchange between pyruvate, lactate, and alanine—pathways that are tightly coupled to cellular energy state and anabolic demands. These metabolic routes are not only central to cancer metabolism but also play a critical role in rapidly proliferating cells, such as those involved in liver regeneration [17, 18, 19]. Recently, a porcine study demonstrated the feasibility of using HP‐MRI for regional assessment of liver regeneration following portal vein ligation, further supporting its utility for evaluating functional recovery after hepatic surgery [20].

The aim of this study was to quantify metabolic activity in the regenerating liver using hyperpolarized [1‐¹³C]pyruvate MRI in a rat model of 70% partial hepatectomy (PH). We hypothesized that the early regenerative process would be associated with increased metabolic flux through native pathways, reflected by elevated lactate‐to‐pyruvate and alanine‐to‐pyruvate ratios.

2. Methods

All procedures were approved by the Danish Animal Experiments Inspectorate (Permit No. 2021‐15‐0201‐00978) and conducted in accordance with the Guide for the Care and Use of Laboratory Animals, National Institutes of Health, and the ARRIVE guidelines [21, 22].

The animal model and experimental procedures have been described in detail previously [23]. The present study includes additional, previously unpublished data acquired from the same cohort of animals. A summary of the relevant methods is provided below. As previously reported for this animal cohort, liver regeneration following 70% PH was confirmed by immunohistochemical evaluations and further supported by stereological assessments and the calculated liver regeneration ratio [23].

Twelve 8‐week‐old male Wistar rats (Janvier Labs, Le Genest‐Saint‐Isle, France) were housed under standard conditions (23°C, 12‐h light–dark cycle) with ad libitum access to food and water. Animals were randomly assigned to either 70% PH (PH; n = 6, mean preoperative weight 296 g [range, 258–334 g]) or a nonsurgical control group (controls; n = 6, mean preoperative weight 305 g [range, 246–314 g]). One PH animal was replaced due to postoperative complications. Sample size estimation was performed a priori, and the number of animals was minimized in accordance with ethical guidelines.

Under inhalation anesthesia, a 70% PH was performed by resecting the median and left lateral liver lobes, leaving the right and caudate lobes as the liver remnant. On postoperative day (POD) 1, animals underwent hyperpolarized and multiparametric MRI. Prior to scanning, rats were anesthetized with 2%–3% sevoflurane, and a tail vein catheter was inserted for subsequent administration of hyperpolarized [1‐¹³C]pyruvate. Vital parameters, including respiratory rate, oxygen saturation, and temperature, were monitored using an MRI‐compatible system (SA Instruments, Stony Brook, NY). Following imaging, blood samples were collected, and animals were euthanized by cervical dislocation.

2.1. Hyperpolarized MRI

All MRI scans were performed on a 3T clinical scanner (GE, Healthcare, Brøndby, DK). Animals were placed on a fixed sledge that allowed changing coils without repositioning the animal. For HP‐MRI, a ¹³C/¹H volume coil (RAPID, Biomedical) with a diameter of 90 mm was used. Under free breathing, a T ₁‐weighted 2D gradient echo sequence with fat suppression, FOV = 80 × 80 mm, matrix 128×128, slice thickness = 16 mm, TR = 68 ms, TE = 2.68 ms, flip angle = 15°, number of averages = 2 was acquired for anatomical planning of ¹³C images. For ¹³C, a modified dynamic 2D spectral‐spatial imaging sequence with spiral readout was used to acquire [1‐¹³C]pyruvate, [1‐¹³C]lactate, [1‐¹³C]alanine, and [1‐¹³C‐bicarbonate], as described previously [24]. Scan parameters included a 16‐mm slice thickness, TR = 650 ms, TE = 10 ms, FOV = 80 × 80 mm, matrix 20 × 20, and flip angles of 30° for pyruvate and 70° for lactate, alanine, and bicarbonate. The 30° flip angle for pyruvate was intentionally applied to ensure sufficient pyruvate signal during the arterial phase in the liver, where the effective pyruvate delivery to the liver during the arterial phase is approximately 20%–25% of the injected dose, together with rapid downstream conversion to lactate and alanine, consistent with prior spectral‐spatial‐based liver studies [20, 25]. The acquisition alternated between pyruvate, lactate, bicarbonate, and alanine with a repetition time of 650 ms per metabolite, resulting in one image per metabolite every 2.6 s and a total of 30 time points acquired for each metabolite.

Multiparametric ¹H images were acquired with a 16‐channel flex coil (GE, Healthcare), wrapped around the animal on the sledge (coil diameter approximately 90 mm). T ₁, R ₂*, and diffusion maps were acquired. Scan parameters are presented in Table 1.

TABLE 1.

MRI acquisition parameters by sequence type.

Parameter	13C imaging	T 1 weighted	T 1 mapping	R 2* mapping	Diffusion‐weighted imaging (DWI)
Sequence type	2D SPSP GRE with spiral readout	2D GRE	2D spoiled GRE	Multiecho GRE (2D, 16 TE)	2D spin‐echo EPI
Repetition time (ms)	650	68	5.04	67.16	4210.5
Echo time (ms)/number of echos	10	2.68	2.51	[2.3–51.6]/16	68.4
Flip angle (°)	30° for pyruvate, 70° for lactate, alanine and bicarbonate	15	20	25	90
Field of view (mm2)	80 × 80	80 × 80	120 × 120	120 × 120	160 × 160
Acquisition matrix	20 × 20	128 × 128	128 × 192	160 × 160	128 × 96
Reconstruction matrix	256 × 256	256 × 256	256 × 256	512 × 512	256 × 256
Slice thickness (mm)	16.0	16.0	4.0	4.0	4.0
b values (s/mm2)	—	—	—	—	0, 30, 40, 50, 80, 150, 200, 300, 400, 600, 800, 1000
Inversion time (ms)	—	—	60–4180 (11 steps)	—	—
Number of signal averages	1	2	1	2	2
Respiratory control	Free‐breathing	Free‐breathing	Free‐breathing	Free‐breathing	Respiratory‐triggered

Abbreviations: EPI, echo‐planar imaging; GRE, gradient‐recalled echo; SPSP, spectral‐spatial.

A 3D multiecho IDEAL IQ sequence was used as previously described [23] and was used in the present study solely for anatomical reference and region of interest (ROI) placement. The liver fat content did not correlate with the metabolic readout. No quantitative data from this sequence are reported here.

2.2. Hyperpolarized [1‐¹³C]‐Pyruvate Sample Preparation

Hyperpolarization of 23 or 127 mg [1‐¹³C]pyruvate (Cambridge Isotope Laboratories) mixed with 30 mM or 15 mM of AH111501 (GE Healthcare) was performed in a SpinAligner (Polarize, Lyngby Denmark) for > 1 h (n = 11) or in a SPINlab (GE Healthcare, Brøndby, Denmark) for > 2 h (n = 1) yielding polarization levels over 40%. Subsequently, dissolution was carried out into deuterium‐enriched NaOH buffer, reaching an end concentration of 75–80 or 120–130 mM [1‐¹³C]pyruvate and a volume of ~2.5 mL or ~16 mL in the final sample– 1 mL used for bolus injection. ¹³C imaging was initiated at the start of pyruvate injection. Subsequent analyses were based on ratiometric metabolite measures, normalizing metabolite signals to the delivered pyruvate signal and thereby minimizing sensitivity to differences in injected pyruvate concentration; this variation did not affect the reported metabolic ratios.

2.3. Data and Image Analysis

HP‐MRI data were gridded, Fourier transformed, and denoised [26, 27] in MATLAB (R2023a, Mathworks Inc., Natick, MA, USA). Area‐under‐the‐curve (AUC) metabolic maps were calculated as voxel‐wise time‐integrated metabolite signals and, from these AUC maps, ratio maps for lactate‐to‐pyruvate (L/P), alanine‐to‐pyruvate (A/P), and lactate‐to‐alanine (L/A) were calculated. Subsequently, metabolic maps were linearly interpolated to match the T ₁‐weighted gradient echo images (matrix = 256 × 256).

ROIs for hyperpolarized images were drawn in MATLAB on T ₁‐weighted gradient echo images and subsequently moved onto HP‐MRI images. For ROI placement, the axial slice containing the right liver lobes was chosen. The caudate lobe was not included because of its small size and susceptibility to partial volume contamination from adjacent vessels. Additionally, time to peak (TTP) and mean transit time (MTT) were computed on ROI basis from Savitzky–Golay–smoothed curves interpolated by cubic spline. For pyruvate, both TTP and MTT were determined over the full dynamic curve; for lactate and alanine, calculations were restricted to time points ≥ 3 × temporal sampling (2.6 s) after injection to avoid spurious peaks from spiral acquisition artifacts caused by the initial pyruvate signal overload. MTT was calculated as the first temporal moment of the respective curve. All times are reported relative to the injection time.

For multiparametric MRI, ROIs were drawn on water T ₁‐weighted (IDEAL) images in in‐house software in the right liver lobes. To avoid motion artifacts, ROIs were placed at a distance to the liver edges. A ROI was drawn on three adjacent slices and copied to multiparametric MRI maps. Reported mean values are averages from these three ROIs. Due to differences in coil setup and slice geometry between hyperpolarized and multiparametric MRI acquisitions, ROIs were defined separately for each modality but consistently placed within the right liver lobes to ensure comparable anatomical sampling. Apparent diffusion coefficient (ADC) and intravoxel incoherent motion (IVIM) fittings were done in in‐house software on a ROI basis. ADC was fitted based on b values (0/30/40/50/80/150/200/300/400/600/800/1000) using single‐exponential fitting, and IVIM parameters were calculated using the same b values and fitting to a bi‐exponential model using the segmented approach as previously described [28], resulting in the biomarkers true diffusion (D), pseudo diffusion (D*), and flow fraction. DWI was respiratory‐triggered, and ROIs from water images (IDEAL) were in some cases displaced due to different respiration state, and, therefore, ROIs were manually drawn directly on diffusion images in these cases. Due to low signal intensity, [1‐¹³C]‐bicarbonate could not be reliably quantified, likely due to low signal‐to‐noise ratio, and was therefore excluded from analysis.

2.4. Statistics

All statistical analyses were conducted using GraphPad Prism (Version 10.5.0, GraphPad Software, San Diego, CA, USA). Normality of continuous variables was assessed separately for each experimental group using histograms and Q–Q plots. Log‐normality was evaluated by inspecting the distribution of log‐transformed values. If both groups displayed approximate normal or log‐normal distributions, group comparisons were performed using unpaired t‐tests. When data from either group did not meet these assumptions, the nonparametric Mann–Whitney U test was used. Data are presented as mean ±95% confidence interval unless otherwise stated.

3. Results

Representative A/P and L/P images from individual animals are shown in Figure 1, and group HP‐MRI measurements are summarized in Figure 2. Mean L/P was significantly higher in the PH group with 0.267 (95% CI: 0.225–0.310) compared to 0.168 (95% CI: 0.135–0.200) in controls (p < 0.001). PH also showed a significantly higher A/P mean value of 0.236 (95% CI: 0.153–0.319) versus 0.150 (95% CI: 0.128–0.172) in controls (p = 0.028). Mean L/A was similar between PH and controls, with 1.21 (95% CI: 1.00–1.42) versus 1.18 (95% CI: 0.835–1.53), respectively (p = 0.865). Additionally, representative pyruvate, lactate, alanine, and L/A maps are provided in Figure S1.

FIGURE 1.

Representative metabolic maps. Anatomical T ₁‐weighted 2D GRE images with red circles showing ROI placements (A, D) and corresponding alanine‐to‐pyruvate (B, E) and lactate‐to‐pyruvate (C, F) ratio maps in a control and a rat after partial hepatectomy. Ratio maps are overlaid on the anatomical images and displayed with identical color map and scaling. Only values within the animal are displayed.

FIGURE 2.

Hyperpolarized measurements in the liver. Horizontal line represents group means, and dots represent mean values for individual animals. p values from unpaired t‐test.

TTP was similar across all metabolites (Figure 3): pyruvate 4.67 s (95% CI: 3.81–5.52) in the control group versus 4.72 s (95% CI: 4.16–5.27) in the PH group (p = 0.902); lactate 10.93 s (95% CI: 10.12–11.75) versus 10.48 s (95% CI: 8.93–12.04; p = 0.524); alanine 8.60 s (95% CI: 7.29–9.91) versus 9.93 s (95% CI: 8.09–11.77; p = 0.160).

FIGURE 3.

Raw signal intensity dynamics and simple perfusion parameters in control and partial hepatectomy (PH) rats. Mean time–intensity curves for pyruvate (scaled/10), lactate, and alanine following intravenous hyperpolarized [1‐¹³C]pyruvate injection in control and PH rats, with shaded areas indicating ± SD For visualization, the first 50 s of the dynamics is shown. Simple perfusion parameters—time‐to‐peak (TTP) and mean transit time (MTT)—were derived from each curve. Group differences in TTP and MTT were assessed using unpaired t‐tests.

For MTT, pyruvate was higher in the PH group (10.06 s [95% CI: 8.58–11.54] vs. 8.27 s [95% CI: 7.44–9.10]; p = 0.022), while MTT for lactate (22.02 s [95% CI: 21.45–22.60] vs. 22.20 s [95% CI: 21.37–23.03]; p = 0.667) and alanine (20.48 s [95% CI: 18.90–22.05] vs. 20.50 s [95% CI: 19.00–22.01]; p = 0.977) was similar between groups.

Quantitative T ₁ and R ₂* values are shown in Table 2. Mean T ₁ was higher in the PH group with 591 ms (95% CI: 533–628) versus 469 ms (95% CI: 443–495) in the control group (p < 0.001). Mean R ₂* was lower in the PH group 64.1 s⁻¹(95% CI: 61.1–67.1) versus 70.5 s⁻¹ (95% CI: 68.7–72.3) in the control group (p < 0.001).

TABLE 2.

Quantitative MRI parameters in the PH and control groups at postoperative day 1.

Parameter	PH group (mean [95% CI])	Control group (mean [95% CI])	p
T 1 (ms)	591 (533–628)	469 (443–495)	< 0.001
R 2* (s−1)	64.1 (61.1–67.1)	70.5 (68.7–72.3)	< 0.001
ADC (×10−3 mm2/s)	1.25 (1.03–1.46)	1.90 (1.32–2.48)	0.021
D (×10−3 mm2/s)	1.06 (0.922–1.21)	1.12 (0.914–1.33)	0.554
D* (×10−3 mm2/s)	23.9 (8.94–38.8)	30.1 (10.8–49.5)	0.525
Flow fraction (%)	14.4 (8.12–20.6)	37.5 (22.1–52.9)	0.005

Note: Values are mean with 95% confidence intervals (CI).

Abbreviations: ADC, apparent diffusion coefficient; D, true diffusion coefficient; D*, pseudodiffusion coefficient; PH, partial hepatectomy.

ADC and IVIM parameters are shown in Figure 4. Mean ADC values were significantly lower in the PH group with 1.25 × 10⁻³ mm²/s (95% CI: 1.03–1.46) versus 1.90 × 10⁻³ mm²/s (95% CI: 1.32–2.48) in controls (p = 0.021). With IVIM fitting, mean D values were not statistically different between the PH group 1.06 × 10⁻³ mm²/s (95% CI: 0.922–1.21) and controls 1.12 × 10⁻³ mm²/s (95% CI: 0.914–1.33) (p = 0.554). Flow fraction was lowest in the PH group with mean values of 14.4% (95% CI: 8.12–20.6) versus 37.5% (95% CI: 22.1–52.9) in controls (p = 0.005), while D* was 23.9 × 10⁻³ mm²/s (95% CI: 8.94–38.8) for PH and 30.1 × 10⁻³ mm²/s (95% CI: 10.8–49.5) for controls (p = 0.525).

FIGURE 4.

Quantitative MRI parameters in the liver. Horizontal line represents group means, and dots represent mean values for individual animals. ADC, D, fractional perfusion, and D* in the liver. p values from unpaired t‐test.

Serum biochemical markers are summarized in Table 3. Compared with controls, animals undergoing partial hepatectomy demonstrated higher ALT, INR, bilirubin, and ammonium levels.

TABLE 3.

Serum markers in experimental animals.

Serum markers	PH group	Control group	p
ALT (U/L)	1877 (1297–3292)	67.5 (58.0–144)	0.002
INR	2.25 (1.48–3.98)	1.48 (1.44–1.57)	0.015
Bilirubin (μmol/L)	9.70 (3.60–21.9)	3.00 (3.00–3.00)	0.002
Ammonium (μmol/L)	163 (150–176)	58.0 (37.0–81.0)	0.002

Note: Data are presented as median (range). Statistical comparisons were performed using the Mann–Whitney U test.

Abbreviations: ALT, alanine aminotransferase; INR, international normalized ratio; PH, partial hepatectomy; POD, postoperative day.

Linear regression analysis showed associations between flow fraction and both L/P (R ² = 0.78, p = 0.020) and A/P (R ² = 0.76, p = 0.023) in the PH group, while no association was observed for L/A; no significant associations were found in controls. No consistent associations were observed for other diffusion parameters and metabolic ratios (Figure S2 and Table S1).

4. Discussion

This study investigated metabolic alterations during liver regeneration in a rat model of 70% PH using hyperpolarized [1‐¹³C]pyruvate MRI. The PH group demonstrated elevated L/P and A/P, reflecting increased exchange through LDH and ALT. The unchanged L/A ratio implies that glycolytic and transaminase pathways were similarly upregulated during early regeneration. This finding is likely explained by increased cellularity and perfusion in the regenerating liver, indicating overall metabolic upregulation without a shift in the relative balance between pathways.

4.1. Early Liver Regeneration Enhances Metabolic Exchange While Preserving Pathway Balance

In this study, both L/P and A/P were increased in the regenerating liver 1 day after 70% PH, whereas L/A remained unchanged. The elevated L/P is likely driven by a shift in redox state, specifically a relative reduction in the cytosolic nicotinamide adenine dinucleotide ratio (NAD⁺/NADH), as the reduced form (NADH) acts as a coenzyme in the enzymatic conversion of pyruvate to lactate and has been suggested to correlate with pyruvate to lactate conversion in HP‐MRI studies [8, 29]. Rather than reflecting a direct contribution of mitochondrial NADH, this may be explained by altered cellular redox handling in proliferating hepatocytes, where increased mitochondrial oxidative metabolism may limit transfer of cytosolic reducing equivalents into mitochondria via NADH shuttles, thereby favoring pyruvate‐to‐lactate exchange as a mechanism for cytosolic NAD⁺ regeneration [30]. The concurrent increase in A/P may result from elevated intracellular glutamate concentrations, due to increased amino acid catabolism and urea cycling—an interpretation supported by the elevated blood ammonia levels in the PH group. Additionally, enhanced glutaminolysis, which is commonly observed in proliferating cells, may contribute to the observed increase in alanine labeling [17, 18, 31]. Overall, these findings indicate an upregulation of both glycolytic and transaminase activity supporting the metabolic demands during early regeneration, without a preferential shift between pathways.

4.2. Hemodynamic and Structural Changes Shape the Regenerative Metabolic Response

The 70% PH model is a well characterized model of liver regeneration [5] with microstructural changes including hepatocyte hypertrophy and proliferation that were confirmed by stereology in this animal cohort [23]. Following two‐thirds PH, the portal vein flow per unit liver tissue triples, leading to significant hemodynamic alterations despite the absence of direct tissue injury and thereby increasing substrate delivery to the liver remnant. These changes, including reduced oxygen tension and increased delivery of gut‐derived factors such as insulin, EGF, and amino acids, are thought to initiate liver regeneration through mechanisms resembling a wound‐healing response [5]. Consistent with this, the prolonged MTT for pyruvate observed in the PH group likely reflects these early vascular changes. Supporting this interpretation, quantitative T ₁ and R ₂* mapping revealed distinct microenvironmental alterations consistent with vascular remodeling and changes in blood oxygenation. Specifically, the elevated T ₁ in the PH group likely reflects increased tissue water content or early extracellular matrix remodeling, both of which may result from hepatocyte hypertrophy and sinusoidal expansion. The concurrent reduction in R ₂* suggests that susceptibility effects from increased hepatic blood volume or perfusion may outweigh those associated with deoxyhemoglobin, indicating that vascular changes during regeneration exert a dominant influence on the R ₂* signal [32, 33]. Furthermore, diffusion‐weighted MRI further supports these findings. ADC and IVIM‐derived flow fraction were significantly lower in the PH group compared to controls, while D and D* remained unchanged. This pattern indicates that the decrease in ADC reflects altered intravoxel signal composition due to remodeling of the hepatic parenchyma characterized by hepatocyte proliferation and hypertrophy with accompanying sinusoidal expansion [34, 35], rather than impaired microvascular perfusion or altered cellular diffusivity. Given the well‐described macroscopic hyperperfusion of the liver remnant in the PH model [5, 34], hypoperfusion is an unlikely explanation; accordingly, the reduced flow fraction likely reflects regeneration‐associated changes in intravoxel tissue‐vascular balance rather than impaired microcirculatory perfusion. The stability of D and D⁎ indicates that the observed ADC changes are unlikely to be driven by alterations in intrinsic cellular water diffusivity or vascular flow velocity.

These imaging biomarkers provide noninvasive evaluation of the hemodynamic and microstructural shifts occurring early after resection. We speculate that these changes may contribute to the increased lactate and alanine labelling observed with hyperpolarized [1‐¹³C]pyruvate MRI. Consistent with this, linear regression analysis showed associations between IVIM flow fraction and both L/P and A/P in the PH group, while no association was observed for L/A; no significant associations were found in controls (Figure S2). This finding may indicate a link between perfusion‐related changes and increased metabolic labelling during early regeneration. If this is the case, perfusion‐driven changes may precede and amplify metabolic exchange, supporting the notion that altered hepatic blood flow is a critical initiator of regeneration‐associated metabolism.

4.3. Alanine and Lactate Labelling Reflect Model‐Specific Regenerative Demands

A previous HP‐MR spectroscopy study in a mouse model of 70% PH reported no change in L/P on POD 3, despite a decreased NAD⁺/NADH ratio, but did observe an increase in A/P [36]. The apparent discrepancy in L/P may be partly explained by the limited spatial resolution of spectroscopy, which cannot clearly isolate hepatic tissue from surrounding organs. In contrast, our imaging approach allowed for precise liver‐specific quantification. An additional explanation for the discrepancy between this study and ours may be the difference in timing. In rats, liver regeneration following 70% PH is a dynamic process that peaks around POD1–3 [37], whereas in the mouse study, imaging was performed on POD3, potentially after the peak of early proliferative activity. As a result, lactate production may have declined by the time of measurement, contributing to the absence of a significant change in L/P. This interpretation is supported by studies showing that the remnant liver reaches ~45% of its original volume by POD1 and ~80% by POD3, providing a larger functional cell mass to support metabolic homeostasis [38]. Furthermore, in a porcine model of partial portal vein ligation, a significant increase in pyruvate to lactate exchange was observed in the future liver remnant at early and late time points, while pyruvate to alanine exchange remained unchanged [20]. The difference in alanine labelling between the pig and the rat model in this study likely reflects both species‐specific variation and differences in regenerative context. While portal vein ligation induces minimal hypertrophy and regenerative response in the pig liver [20, 39], the 70% PH model presents a more severe regenerative stimulus, potentially requiring not only increased glycolytic flux but also enhanced amino acid metabolism to meet the biosynthetic demands of hepatocyte proliferation. The observed alanine and lactate labelling in our study may reflect the hepatic upregulation to meet the increasing demand for whole body homeostasis, which is supported by the unaltered change in L/A.

4.4. Lactate‐to‐Alanine Ratio as a Marker of Hepatic Metabolic Integrity?

In our study, the increases in both L/P and A/P ratios suggest an overall upregulation of pyruvate metabolism following 70% PH, likely reflecting the elevated metabolic demands during liver regeneration. Importantly, the L/A ratio remained stable despite the global increase in metabolite labelling. This preserved proportionality between lactate and alanine production indicates that pyruvate utilization across multiple pathways remains balanced during the regenerative phase. Notably, the PH model was performed in otherwise healthy animals without underlying liver disease, and these animals are known to recover fully [37].

This is consistent with other studies applying hyperpolarized [1‐¹³C]pyruvate in models including hepatic ischemia–reperfusion injury [9], liver fibrosis [40], chemically induced liver inflammation (e.g., CCl₄ and 1,3‐DCP) [11], and metabolic associated fatty liver disease [10], which report increased production of both alanine and lactate—likely reflecting preserved hepatocellular function and a capacity to upregulate exchange through lactate dehydrogenase and alanine aminotransferase in metabolically flexible hepatocytes.

In contrast, a recent study of a more chronic fatty liver model in rats showed reduced lactate and alanine production over several weeks, suggesting impaired hepatic function. This was associated with persistent liver injury characterized by fibrosis, hepatocellular ballooning, and reduced hepatocyte density due to severe steatosis [41]. This pattern was observed across multiple rodent models using both imaging and spectroscopy techniques. Other studies have reported discrepancies in lactate and alanine production, with ethanol infusion altering redox state in the rat liver [29]. Furthermore, different models of hepatocellular carcinoma have shown inconsistent findings, with some demonstrating increased levels of both alanine and lactate [42, 43], while others have reported elevation in only one of these metabolites or dynamic changes [44, 45] These findings likely reflect cellular reprogramming and a metabolic shift away from normal hepatocellular function, indicating altered enzymatic control and substrate handling in pathologically transformed hepatocytes.

Traditionally, hyperpolarized [1‐¹³C]pyruvate MRI relies on the bicarbonate signal to estimate pyruvate dehydrogenase (PDH) flux and infer TCA cycle activity. However, in the liver, this signal is complex and may reflect contributions from both PDH and pyruvate carboxylase, depending on the metabolic context [46, 47]. Moreover, the bicarbonate peak is particularly sensitive to signal‐to‐noise ratio limitations and is typically acquired via spectroscopy rather than imaging, making quantification challenging and variable across studies.

As evident in mechanistic studies using hyperpolarized [1‐¹³C]pyruvate, alanine production may serve as a more accessible and biologically meaningful surrogate of mitochondrial activity. For example, Yang et al. showed that inhibition of the mitochondrial pyruvate carrier suppresses both alanine and bicarbonate formation while lactate production remains preserved, directly linking alanine synthesis to mitochondrial pyruvate oxidation and TCA cycle flux [48]. In support of this, Klain et al. demonstrated that during early liver regeneration in rats, alanine‐derived carbons were increasingly routed through pyruvate carboxylase and into glutamate, indicating a shift in TCA cycle routing toward gluconeogenic and biosynthetic needs [49]. These findings highlight that alanine labelling is tightly coupled to mitochondrial glutamate and α‐ketoglutarate pools, both central to TCA metabolism.

Furthermore, in an MRS study of human fatty liver [12], Park et al. observed increased lactate production with concurrent decreases in both alanine and bicarbonate, suggesting impaired mitochondrial pyruvate handling even in noncirrhotic steatosis. Collectively, these observations support the use of alanine as a more stable and interpretable indicator of mitochondrial metabolic function in the liver, particularly when bicarbonate quantification is limited by physiological or technical constraints.

Taken together, the stable L/A ratio observed in our regenerating liver model contrasts with the imbalanced patterns seen in disease states, supporting the idea that proportional pyruvate utilization is a feature of functional recovery. We hypothesize that in settings where regeneration is impaired, such as in models or patients developing posthepatectomy liver failure, a mismatch in L/A, or a reduction in L/P and A/P with a maintained L/A, may serve as an early indicator of metabolic insufficiency. Future studies are warranted to test the predictive value of L/A as a metabolic biomarker of regenerative capacity.

This study has several limitations: While the 70% PH model in healthy male Wistar rats does not fully replicate the clinical complexity of human liver resections, it remains a widely accepted and well‐characterized model for studying regenerative responses in a controlled setting. Imaging was limited to a single postoperative time point, restricting our ability to characterize the temporal dynamics of metabolic reprogramming throughout the regenerative process; however, POD1 represents the peak of early regenerative activity in the rat liver, thereby providing optimal sensitivity for detecting metabolic alterations associated with the initiation of regeneration. Liver regeneration is initiated by an early inflammatory signaling response, and subtle cytokine‐mediated effects related to surgical stress have been demonstrated at POD1 [37, 50]. This cytokine surge is a well‐established driver of hepatocyte proliferation and is primarily mediated by resident nonparenchymal cells rather than by infiltration of circulating immune cells [5, 50]. In line with this, histological evaluation in the present cohort demonstrated liver tissue composed almost exclusively of hepatocytes, with no overt inflammatory cell infiltrates identified on routine histological examination. However, as no targeted assessment of inflammatory cell populations or cytokine expression was performed, the absence of detectable inflammatory infiltrates does not exclude the presence of active cytokine‐mediated inflammatory signaling. Consequently, while overt inflammation or infection is unlikely to represent a major confounder of the imaging findings, subtle cytokine‐driven effects associated with the early regeneration response cannot be excluded.

Although [1‐¹³C]‐bicarbonate was acquired, signal levels were low and excluded from quantitative analysis. Importantly, the absence of a quantifiable bicarbonate signal in this study precludes any definitive conclusions regarding hepatic tricarboxylic acid cycle activity, given the complex and limited pathway specificity of hepatic bicarbonate production following [1‐¹³C]pyruvate administration [46, 47]. The study did not include ex vivo biochemical or enzyme activity measurements (e.g., LDH, ALT, PDH, or PC), which could provide complementary information to the in vivo hyperpolarized MRI findings; however, hyperpolarized [1‐¹³C]pyruvate MRI primarily reflects in vivo metabolic exchange, which does not necessarily correspond to enzyme gene expression levels or ex vivo measurements of enzyme activity [20, 41]. Finally, hyperpolarized pyruvate was prepared using two different polarization systems. While this introduces theoretical variability in polarization levels, all samples exceeded 40% polarization, and the use of metabolite‐to‐pyruvate AUC ratios for analysis eliminates the impact of absolute signal differences between animals due to polarization levels.

In conclusion, this study shows that early liver regeneration after 70% PH is associated with an overall increase in metabolism, with balanced lactate and alanine production from pyruvate, reflecting coordinated upregulation of glycolytic and transaminase pathways. Hyperpolarized [1‐¹³C]pyruvate MRI enabled dynamic, spatially resolved assessment of these metabolic changes in vivo, offering new insights into the metabolic demands of hepatic regeneration.

Author Contributions

U.K., A.L., F.V.M., and C.L. contributed to the study conception and design. Data acquisition was carried out by U.K., A.L., L.B.B., E.S.S.H., and K.H.A. Data analysis and interpretation were undertaken by U.K., A.L., J.R.N., S.R., F.V.M., and C.L. The first draft of the manuscript was prepared by U.K. and A.L. All authors critically revised the work and approved the final version.

Funding

The study was funded by Aarhus University, the Danish Cancer Society, AP Møller Fonden, Lizzi and Mogens Staal Fonden, Dagmar Marshall Fond, and Fabrikant Einar Willumsens Mindelegat.

Ethics Statement

All experimental procedures complied with Danish legislation governing the use of animals in research. Approval was granted by the Danish National Committee for the Protection of Animals used for Scientific Purposes, Copenhagen, Denmark (License No. 2021‐15‐0201‐00978).

Supporting information

Figure S1: Representative maps from a control animal (A–D) and a PH rat (E–H). Raw intensity maps for pyruvate (A, E), lactate (B, F), and alanine (C, G), summed over the entire acquisition time. Lactate‐to‐alanine ratio maps are shown in (D, H). All maps are overlaid on anatomical T ₁‐weighted 2D GRE images with ROI placement indicated (red circle). Pyruvate, lactate, and alanine maps (A–C, E–G) are individually scaled in arbitrary units (A.U.). For the lactate‐to‐alanine ratio maps (D, H), only values within the animal are displayed, using identical scaling.

Figure S2: Linear regression analyses between IVIM‐derived parameters and metabolic ratios. Plots show relationships between ADC, D, flow fraction, and D* and L/P, A/P, and L/A in PH and control groups. Solid lines represent linear regression fits, with dashed lines indicating 95% confidence intervals. Associations were observed between flow fraction and both L/P and A/P in the PH group, while no consistent associations were observed for other parameters or in controls. ADC, apparent diffusion coefficient; D, diffusion coefficient; D*, pseudo‐diffusion coefficient; PH, partial hepatectomy.

Table S1: Linear regression analyses between IVIM‐derived parameters and metabolic ratios. β represents the regression slope with standard error (SE), and R ² denotes the coefficient of determination. ADC, apparent diffusion coefficient; D, diffusion coefficient; D*, pseudo‐diffusion coefficient; PH, partial hepatectomy. Statistically significant associations (p < 0.05) are indicated by *.

Data Availability Statement

Data and code are available from the corresponding author upon request.