Characterizing Metabolic Heterogeneity of Hepatocellular Carcinoma with Hyperpolarized ¹³C Pyruvate MRI and Mass Spectrometry

Abstract

Purpose

To characterize the metabolomic profiles of two hepatocellular carcinoma (HCC) rat models, track evolution of these profiles to a stimulated tumor state, and assess their effect on lactate flux with hyperpolarized (HP) carbon 13 (¹³C) MRI.

Materials and Methods

Forty-three female adult Fischer rats were implanted with N1S1 or McA-RH7777 HCC tumors. In vivo lactate-to-pyruvate ratio (LPR) was measured with HP ¹³C MRI at 9.4 T. Ex vivo mass spectrometry was used to measure intratumoral metabolites, and Ki67 labeling was used to quantify proliferation. Tumors were first compared with three normal liver controls. The tumors were then compared with stimulated variants via off-target hepatic thermal ablation treatment. All comparisons were made using the Mann–Whitney test.

Results

HP ¹³C pyruvate MRI showed greater LPR in N1S1 tumors compared with normal liver (mean [SD], 0.564 ± 0.194 vs 0.311 ± 0.057; P < .001 [n = 9]), but not for McA-RH7777 (P = .44 [n = 8]). Mass spectrometry confirmed that the glycolysis pathway was increased in N1S1 tumors and decreased in McA-RH7777 tumors. The pentose phosphate pathway was also decreased only in McA-RH7777 tumors. Increased proliferation in stimulated N1S1 tumors corresponded to a net increase in LPR (six stimulated vs six nonstimulated, 0.269 ± 0.148 vs 0.027 ± 0.08; P = .009), but not in McA-RH7777 (eight stimulated vs six nonstimulated, P = .13), despite increased proliferation and metastases. Mass spectrometry demonstrated relatively increased lactate production with stimulation in N1S1 tumors only.

Conclusion

Two HCC subtypes showed divergent glycolytic dependency at baseline and during transformation to a high proliferation state. This metabolic heterogeneity in HCC should be considered with use of HP ¹³C MRI for diagnosis and tracking.

Keywords: Molecular Imaging-Probe Development, Liver, Abdomen/GI, Oncology, Hepatocellular Carcinoma

© RSNA, 2024

See also commentary by Ohliger in this issue.

Summary

Metabolic heterogeneity of hepatocellular carcinoma can result in variable lactate flux, as measured with hyperpolarized ¹³C pyruvate MRI, which may be taken into account when evaluating therapeutic options.

Key Points

• ■ Hyperpolarized (HP) carbon 13 (¹³C) pyruvate MRI showed greater lactate-to-pyruvate ratio in N1S1 rat hepatomas compared with normal liver (0.564 ± 0.194 vs 0.311 ± 0.057; P < .001) but not in McA-RH7777 rat hepatomas (0.453 ± 0.137 vs 0.520 ± 0.120; P = .44).

• ■ Mass spectrometry confirmed that the glycolysis pathway was increased in N1S1 tumors, whereas McA-RH7777 tumors exhibited decreased glycolysis with preference for alternative metabolic pathways.

• ■ HP ¹³C MRI may be used to select hepatocellular carcinomas that may be amenable to glycolytic-targeted therapies and tracking.

Introduction

Hepatocellular carcinoma (HCC) is one of the most prevalent cancers worldwide and ranks third in all cancer-related mortality. Hepatocarcinogenesis involves complex genetic and cellular dysregulations, which cause high inter- and intratumor heterogeneity (1,2). This heterogeneity poses a challenge for targeted therapies and diagnostics.

Metabolic alteration is widely considered a hallmark feature of cancer (3). One common feature is the increased glucose uptake and fermentation of glucose to lactate via glycolysis, known as the Warburg effect (4). This reflects an adaptation mechanism to provide energy and macronutrients for tumor growth, even in the presence of oxygen. Hyperpolarized (HP) carbon 13 (¹³C) MRI can noninvasively depict conversion of the HP ¹³C-labeled pyruvate into lactate and other downstream metabolites (5). This exploits the Warburg effect, offering the ability for detecting tumors earlier and for tumor grading. Previous clinical studies have demonstrated a grade-dependent increase in ¹³C lactate flux in prostate (5,6), breast (7), and renal (8) tumors, which is associated with poor clinical outcomes.

Although HCC typically exhibits the Warburg effect, its metabolic profile can be highly variable, especially in more advanced disease, which presents a therapeutic challenge (9–11). Recent studies showed that tumor-derived lactate can inhibit the efficacy of lenvatinib (12) and cisplatin (13) and promote resistance to glucose starvation (14). Metabolic processes range from glucose metabolism to amino and fatty acid metabolism (15–17). We previously demonstrated high dependence of N1S1 HCC tumors on PFKFB3, a key glycolytic modulator, offering a potential therapeutic target and a way to monitor these changes with HP ¹³C pyruvate MRI (18).

In this study, we aimed to obtain a more comprehensive overview of glycolysis dependency in HCC and its effect on diagnostic evaluation with HP ¹³C pyruvate MRI. To do this, we compared two different rat hepatoma models at both nonstimulated and stimulated states: N1S1 is a single-lesion HCC tumor with limited metastasis potential, and McA-RH7777 regularly produces both intra- and extrahepatic metastases. We chose these models because they mimic the wide range of behavior in human HCCs. We then used both hydrogen 1 (¹H) and HP ¹³C pyruvate MRI to track these changes in vivo and correlated them to quantitative metabolomics via ex vivo mass spectrometry.

Materials and Methods

The study protocol was approved by the Institutional Animal Care and Use Committee, in accordance with the National Research Council Institute of Laboratory Animal Resources.

Cell Lines

Rat HCC cell lines N1S1 (CRL-1604, male) and McA-RH7777 (CRL-1601, male) were obtained from the American Type Culture Collection (ATCC). Cells were maintained at 37 °C in a humidified atmosphere with 5% CO₂. N1S1 cells were grown in Iscove’s modified Dulbecco’s medium (GIBCO), and McA-RH7777 cells were grown in Dulbecco’s modified Eagle medium, with 1% antibiotic-antimycotic (100×) (GIBCO).

HCC Model and Study Design

According to the information provided by ATCC, the N1S1 cell line was isolated from a male rat, whereas the McA-RH7777 cell line was isolated from a female rat. Based on our trial experiments, the two sexes did not differ in terms of tumor development in either N1S1 or McA-RH7777. We chose female rats for both HCC models primarily to standardize sex for metabolic comparison between N1S1 and McA-RH7777 tumors. In addition, their relatively smaller size was better suited for the MRI equipment used. The study design is summarized in Figure 1. Forty-six adult (120–150 g) female Fischer (CDF) rats (Charles River) were used. Using aseptic techniques, 10⁶ cells and 5 × 10⁶ cells were implanted in the left hepatic lobe for N1S1 and McA-RH7777, respectively. For McA-RH7777, 20 mg/kg cyclosporin A was injected intraperitoneally daily for 5 consecutive days to provide transient immunosuppression only during the implantation phase, with the implantation performed on the second day (19). We treated rats with cyclosporin A for only 4 days after implantation. The half-life of cyclosporin A is up to 27 hours (20), so tumors were grown for 2–3 weeks before the initiation of experiments to ensure that cyclosporin A had no active effect on tumor metabolism during imaging or at the time of tumor extraction.

Figure 1:

Animal experimental design. (A) Measurement of the baseline metabolism in N1S1 tumors, McA-RH7777 tumors, and normal liver tissues. (B) Measurement of the metabolism level in nonstimulated tumors (NS_T) and stimulated tumors (S_T) in both N1S1 and McA-RH7777 models. Red arrows demarcate the time of tumor and liver tissue harvesting. ¹H MRI = proton MRI, HP ¹³C MRI = hyperpolarized ¹³C MRI, MS = mass spectrometry, RFA = radiofrequency ablation.

First, we compared the metabolic profile of N1S1 and McA-RH7777 tumors against normal liver. To achieve this aim, healthy liver from nontumor rats were considered the baseline level for both HCC models. Three of the rats did not undergo any implantation and served as normal liver controls (Fig 1A). We then compared the metabolic profiles of nonstimulated tumors (sham radiofrequency ablation [RFA] treatment) and stimulated tumors (RFA treatment) for both N1S1 and McA-RH7777 (Fig 1B). Twelve of the 46 rats were used in a previous study (18). That study focused on PFKFB3 as a potential therapeutic target for HCC, whereas the current study compared metabolism reprogramming across different HCC subtypes.

Generation of Stimulated N1S1 and McA-RH7777 Variants

To generate stimulated variants of N1S1 or McA-RH7777, we exploited the use of off-target hepatic RFA, previously shown to stimulate tumor proliferation through activation of the c-MET pathway (21–23). When tumors reached 8–10 mm, rats were randomized to a sham-treated (nonstimulated) group or an RFA-treated (stimulated) group. With use of aseptic technique, the right hepatic lobe was exposed. In the RFA-treated group, off-target RFA was applied to adjacent normal hepatic tissue (7-mm active tip, RF tip temperature of 70 °C for 5 minutes) (21). For the sham-treated group, all the surgical steps for RFA were replicated without application of energy.

HP ¹³C pyruvate MRI was performed before RFA or sham treatment, and again 3 days after RFA or sham treatment (24). N1S1 rats were euthanized immediately after the last HP ¹³C MRI study for tumor harvesting. McA-RH7777 rats were euthanized 14 days after RFA or sham treatment to allow for the development and tracking of metastases. A portion of the harvested tumors were fixed in 10% formalin for histology, and the rest were immediately preserved at -80°C for metabolomic analysis.

¹H and HP ¹³C MRI

Imaging was performed with a 9.4-T MRI scanner (Biospec 94/20; Bruker) with an 84-mm quadrature proton coil combined with a transmit-receive ¹³C surface coil. HP ¹³C pyruvate solution was prepared by dynamic nuclear polarization using a commercial polarizer (Hypersense; Oxford Instruments Molecular Biotools). [1-¹³C] pyruvic acid was combined with 15 mM of OX063 radical (GE HealthCare) and 1 mM of gadoteridol (ProHance; Bracco) and polarized for 40 minutes or more at 1.4 K and 100 mW microwave power. The hyperpolarized material was then dissolved in saline containing 50 mM of Tris and 125 mg/L EDTA and adjusted to physiologic pH with sodium hydroxide to obtain a 96-mM hyperpolarized pyruvate solution.

Rats were anesthetized with isoflurane (4% induction, 2.5%–3% maintenance) with oxygen (700 mL/min), and the breathing rate was maintained at a mean (SD) of 80 breaths per minute ± 15. Core body temperature was maintained at 36 °C ± 1 with a feedback-controlled warm air circulator equipped with a fiberoptic temperature probe. The lateral tail vein was cannulated with a 24-gauge flexible catheter connected to an 84-cm extension tube with 0.5-mm inside diameter (SAI Infusions). Animals were positioned with tumors within 5 mm of the center of the coil during imaging to minimize the effects of the surface coil sensitivity profile.

A total of 2.5 mL of HP ¹³C pyruvate was administered via tail vein. The ¹³C data were acquired in the axial plane across the middle of the tumor using an echo-planar spectroscopic imaging nongated sequence (tip angle, 4 degrees; matrix, 16 × 16; field of view, 4.5 × 4.5 cm; section thickness, 6–9 mm; number of spectral points, 512; spectral width, 4 kHz; 64 frames at 2.2 seconds per frame). A [1-¹³C]-acetate phantom was placed at the opposite side of the surface coil to help identify the pyruvate and lactate peaks, given a known chemical shift of 0.5 ppm from lactate. Lactate flux was measured as the ratio of integrated signal relative to pyruvate, or lactate-to-pyruvate ratio (LPR). Lactate flux was also measured by kinetic modeling of lactate formation and expressed as a rate constant k_PL, detailed below.

¹H MRI was performed using T2-weighted axial, coronal, and sagittal images with a rapid imaging with refocused echoes sequence (field of view, 6 cm; matrix, 128 × 128; echo train length, 8; repetition time msec/echo time msec, 1245/11; section thickness, 2 mm).

Measurement of Lactate Flux

To quantify ¹³C-pyruvate to ¹³C-lactate, we integrated across each spectral peak and across all time points within a manually drawn region of interest (ROI) using the ¹H anatomic images as guidance. The tumor ROI covered the whole tumor area, and the liver ROI was placed at the largest region of hepatic parenchyma separate from the tumor. Lactate flux was measured as the ratio of the time-integrated signal of each species relative to pyruvate, summed over the ROI, expressed as the LPR. Kinetic modeling of lactate formation was performed by fitting the measured lactate and pyruvate time courses to a simplified two-site exchange model:

where L(t) is lactate signal, P(t) is the pyruvate signal, k_PL is the forward rate constant, and the effective T1 decay rate R1_eff includes the effects of tracer outflow, T1 decay, and saturation by RF pulses. The parameters k_PL and R1_eff were determined by performing a least-squares fit of the analytical solution of this equation to the pyruvate and lactate signal, summed over the ROI.

Quantitative Polar Metabolomic Profiling

Mass spectrometry was performed at our proteomics-metabolomics core facility. Three 15-mg random samples were acquired from each tumor or normal liver. Polar metabolomics profiling was performed using the 5500 QTRAP system (SCIEX) with selected reaction monitoring and MultiQuant version 3.0 software (SCIEX) with polarity switching for approximately 300 metabolites (25). The relative levels of metabolites in normal and tumor tissue were normalized and analyzed with analysis of variance–simultaneous component analysis using MetaboAnalyst version 5.0 (26). Data were analyzed using R software (RStudio 2023.03.0+386).

Histologic Analysis

Samples were fixed in 10% neutral formalin, and paraffin-embedded sections were stained with Ki67 (1:1000, CST12202S; Cell Signaling). Staining was labeled with goat anti-rabbit IgG (1:10000;ab214880, Abcam) conjugated with horseradish peroxidase polymer and 3,3-diaminobenzidine (Vector Laboratory). Slides were scanned with an Axio Scan.Z1 (Zeiss) with the whole mount digitalized at 10× magnification. Three regions in each sample were randomly selected, and Ki67% was determined using ImageJ software (National Institutes of Health).

Quantitative Reverse Transcription–Polymerase Chain Reaction

Total RNA was extracted from frozen samples using the RNeasy Mini Kit (74134; Qiagen) as described elsewhere (18), then reverse transcribed into complementary DNA using iScript gDNA Clear cDNA Synthesis Kit (1725035; Bio-Rad). The level of messenger RNA (mRNA) expression was normalized to the housekeeping gene cyclophilin A (CYPA). The mRNA expressions of GLUT1, HK2, PFKFB3, PFK1, LDHA, PKM, G6PD, TALDO1, PGLS, PGD, RPIA, and TKT were measured. The primer sequences are listed in Table 1.

Table 1:

Primer Sequences for Reverse Transcription–Polymerase Chain Reaction

Statistical Analysis

Statistical analysis of MRI and histopathology data was performed using Prism 9 (GraphPad). Comparisons between two groups (eg, LPR between tumor and background liver) were made using the Mann–Whitney test. For comparing N1S1, McA-RH7777, and normal liver mRNA expression levels, a Kruskal–Wallis test was used to determine whether there were significant differences across the three. If a significant difference was detected, paired comparisons were made with a Mann–Whitney test. Statistical significance was defined as a P value less than .05. Errors are reported as SD unless otherwise indicated.

Results

N1S1 and McA-RH7777 Differ in Glycolytic Activity with ¹³C Pyruvate MRI

Nine N1S1 and eight McA-RH7777 tumors were successfully implanted. We measured the baseline glycolytic activity in the N1S1 and McA-RH7777 tumors using HP ¹³C pyruvate MRI when the tumors reached 10 mm in diameter. Peak ¹³C-pyruvate signal was typically visualized at approximately 6 seconds after injection, followed by peak ¹³C-lactate signal approximately 6 seconds later (Fig 2A). The LPR was greater in N1S1 tumors than in normal liver (0.564 ± 0.194 vs 0.311 ± 0.057; P < .001 [n = 9]), but not for McA-RH7777 (0.453 ± 0.137 vs 0.520 ± 0.120; P = .44 [n = 8]) (Fig 2B). The lactate flux rate constant k_PL was greater in N1S1 tumors than in McA-RH7777 tumors (0.035 sec⁻¹ ± 0.021 vs 0.013 sec⁻¹ ± 0.004; P = .002) (Fig 2C). In summary, McA-RH7777 tumors demonstrated lower lactate flux compared with N1S1 tumors and could not be differentiated from background liver lactate flux.

Figure 2:

N1S1 and McA-RH7777 hepatocellular carcinoma (HCC) subtypes exhibit different degrees of lactate flux. (A) Hyperpolarized (HP) carbon 13 (¹³C) pyruvate and HP ¹³C lactate spectroscopic signals over time and representative images of ¹³C-lactate MRI (green) superimposed on T2-weighted ¹H images (grayscale) in both N1S1 and McA-RH7777 tumors. Tumor outlined in dotted red; adjacent normal liver outlined in dotted white. Blue arrow: kidney; yellow arrow: spine; green arrow: stomach. ¹³C lactate signal intensity scaling was adjusted between the N1S1 and McA-RH7777 images to allow visualization of the lactate signals in the latter. (B) Lactate flux, measured as the lactate-to-pyruvate ratio (LPR), in tumor (T) and adjacent normal liver (NL) in both HCC subtypes. (C) Baseline lactate flux as expressed in k_PL for N1S1 (N) and McA-RH7777 (M) tumors, also showing greater flux in N1S1. Error bars represent SD.

Quantitative Mass Spectrometry Reveals Different HCC Metabolomic Profiles Consistent with ¹³C Pyruvate MRI

Four N1S1 tumors, four McA-RH7777 tumors, and three normal livers from control rats were compared. Partial least-squares discriminant analysis revealed that the baseline metabolomics in N1S1, McA-RH7777, and normal liver exhibited different distributions, with a Q² of 0.97 and a R² of 0.99 (Fig 3A). To better understand the metabolic differences across these two HCC subtypes, we compared N1S1 and McA-RH7777 tumor samples against normal liver tissue. In N1S1 tumors, there were 51 metabolites downregulated and 81 metabolites upregulated. In McA-RH7777 tumors, there were 63 metabolites downregulated and 79 metabolites upregulated. Unsupervised clustering analysis among the significantly changed metabolites identified them according to the following six combinations: increased in both N1S1 and McA-RH7777, decreased in both, or increased or decreased in each tumor (Fig 3B, 3C). We then used KEGG (Kyoto Encyclopedia of Genes and Genomes) pathway enrichment analysis to identify the related metabolic pathways between N1S1 and McA-RH7777. Several amino acid metabolism pathways were upregulated in both N1S1 and McA-RH7777 tumors compared with normal livers (Table 2). Glycolysis and the tricarboxylic acid cycle were upregulated in N1S1 tumors compared with normal liver, whereas both glycolysis and the pentose phosphate pathway (PPP) were downregulated in McA-RH7777 (Table 2). Compared with normal liver, lactate levels were greater in N1S1 tumors but unchanged in McA-RH7777 (Fig 3D). These findings are consistent with the HP ¹³C MRI results.

Figure 3:

Metabolic profiles of N1S1 and McA-RH7777 hepatocellular carcinoma (HCC). (A) Multivariable analysis of metabolomic data using the partial least-squares discriminant model. (B, C) Altered metabolites (upregulated, B; downregulated, C) in N1S1 (N) and McA-RH7777 (M) relative to normal liver. (D) Normalized lactate levels in normal liver from control rats (C_NL), McA-RH7777 tumor (M_T), and N1S1 tumor (N_T). Error bars represent SD. F1,6BP = fructose 1,6-bisphosphate; F6P = fructose 6-phosphate; G3P = glyceraldehyde 3-phosphate; G6P = glucose 6-phosphate; NADPH = nicotinamide adenine dinucleotide phosphate; R5P = ribose 5-phosphate.

Table 2:

Enriched Metabolic Pathways in Tumors from N1S1 and McA-RH7777 Rat Hepatocellular Carcinoma Cell Lines Compared with Normal Liver from Control Rats

Figure 4 provides a summary map of the crosstalk between glycolysis and PPP with normalized concentrations of these related metabolites. In N1S1 tumors, fructose 1,6-bisphosphate, glyceraldehyde 3-phosphate, and lactate were upregulated in glycolysis, whereas other intermediates were unchanged. These glycolytic intermediates were downregulated in McA-RH7777. Metabolites along the early portion of the PPP were also decreased in McA-RH7777.

Figure 4:

Expression of metabolites relating to glycolysis and the pentose phosphate pathway (PPP) in normal liver from control rats and tumors from N1S1 and McA-RH7777. Each graph shows the metabolite expression in normal liver (C_NL, left bar), McA-RH7777 (M_T, middle bar), and N1S1 (N_T, right bar). Error bars represent SD. 1,3-BPG = 1,3-bisphosphoglyceric acid; 3PG = 3-phosphoglyceric acid; 6-PG = 6-phosphogluconic acid; 6-PGL = 6-phosphogluconolactone; DHAP = dihydroxyacetone phosphate; E4P = erythrose 4-phosphate; F1,6BP = fructose 1,6-bisphosphate; F2,6BP = fructose 2,6-bisphosphate; F6P = fructose 6-phosphate; G3P = glyceraldehyde 3-phosphate; G6P = glucose 6-phosphate; PEP = phosphoenolpyruvate; R5P = ribose 5-phosphate; Ru5P = ribulose-5-phosphate; S7P = sedoheptulose 7-phosphate; X5P = xylulose-5-phosphate.

To further elucidate the regulation of glycolysis and PPP in both tumors, we also assessed the mRNA expression of several key genes involved in both pathways (Fig 5). The key glycolytic enzymes GLUT1, HK2, PFKFB3, PFK1, and PKM were increasingly expressed in N1S1 tumors but unchanged in McA-RH7777. These results are consistent with both mass spectrometry and ¹³C MRI results. The PPP regulators G6PD and TALDO1 had increased expression in N1S1 tumors but were mostly unchanged in McA-RH7777.

Figure 5:

Messenger RNA expression of glycolysis and pentose phosphate pathway (PPP)–related genes in N1S1 tumors, McA-RH7777 tumors, and normal liver. Glycolysis-related key enzymes include GLUT1, HK2, PFKFB3, PFK1, LDHA, and PKM. PPP-related key enzymes include G6PD, TALDO1, PGLS, PGD, RPIA, and TKT. Densitometry qualification of band intensity is presented as a percentage of relative densitometry normalized to the CYPA gene. Blue arrow, target protein. P values represent results from Mann–Whitney test, performed only after a significant difference across the three groups was found with a Kruskal–Wallis test. Error bars represent SD. C_NL = normal liver from control rats, M_T = tumor from McA-RH7777 rats, N_T = tumor from N1S1 rats.

HP ¹³C Pyruvate MRI Tracks Metabolic Alteration in Stimulated N1S1 Tumors but Not McA-RH7777 Tumors

A total of 12 N1S1 and 14 McA-RH7777 tumors were successfully implanted. Of these, six N1S1 and eight McA-RH7777 were treated by off-target RFA. Stimulated McA-RH7777 and N1S1 tumors demonstrated greater Ki67 expression compared with sham RFA-treated tumors (stimulated vs nonstimulated McA-RH7777: 58.08% ± 8.19 vs 47.68% ± 7.69 [P = .03]; stimulated vs nonstimulated N1S1: 57.43% ± 8.64 vs 45% ± 12.97 [P = .03]) (Fig 6A). The number of distant metastases in the stimulated McA-RH7777 group was also greater (13.13 ± 9.31 for stimulated vs 1.33 ± 1.51 for nonstimulated; P = .007) (Fig 6B). Compared with sham-treated tumors, LPR was significantly increased in stimulated N1S1 tumors (0.269 ± 0.148 for stimulated_after-before vs 0.027 ± 0.08 for nonstimulated_after-before; P = .009) but not in stimulated McA-RH7777 (-0.083 ± 0.131 for stimulated_after-before vs 0.003 ± 0.08 for nonstimulated_after-before; P = .13) (Fig 6C). The LPR levels in background normal liver in N1S1 and McA-RH7777 were unchanged following the stimulation (data not shown).

Figure 6:

Proliferation and metabolic alteration in radiofrequency ablation–stimulated N1S1 and McA-RH7777 tumors. (A) Increased Ki67 expression is observed in stimulated tumors relative to nonstimulated tumors. Scale bar, 50 µm. (B) The number of metastases in stimulated McA-RH7777 (S) is greater compared with nonstimulated tumors (NS). (C) ¹³C lactate-to-pyruvate ratio (LPR) is increased in N1S1 tumors from prestimulation to 3 days after treatment, but not in McA-RH7777 tumors. Error bars represent SD.

Mass spectrometry revealed elevated glycolysis pathway components for stimulated N1S1 tumors compared with nonstimulated tumors (false discovery rate, 0.048), but not for McA-RH7777 tumors (false discovery rate, 0.282), consistent with MRI results (Table 3). The alternative PPP metabolic pathway, in contrast, was upregulated in McA-RH7777 while downregulated in N1S1 following stimulation.

Table 3:

Altered Metabolic Pathways between Stimulated Tumor and Nonstimulated Tumor in N1S1 and McA-RH7777 Models

Discussion

Metabolic alterations are necessary for biosynthesis and homeostasis in cancer, including HCC. Glycolysis-based lactate production is known to contribute to HCC resistance to current first-line treatments, such as sorafenib (11) and lenvatinib (27). HP ¹³C pyruvate MR spectroscopy has previously demonstrated detection of latent HCC cells based on lactate flux following transarterial embolization, predicting local recurrence (28). However, in this study, we demonstrated that HCC does not necessarily rely on glycolysis as a primary metabolic driver for survival and progression. This has been noted previously, for example in mouse HCC models showing that MYC-driven tumors show a high degree of aerobic glycolysis, but not in MET-driven tumors (29). Our study explores this further by providing, to our knowledge, the first comprehensive analysis of how such variations affect in vivo HP ¹³C pyruvate MRI and the underlying contributing molecular signaling pathways that lead to these changes. We also show how these metabolic profiles are shifted when these tumors were stimulated by off-target RFA compared with nonstimulated tumors, in particular if that shift requires upregulation of glycolysis. In vivo HP ¹³C pyruvate MRI demonstrated high lactate flux in N1S1 tumors but no such elevation in McA-RH7777, even with metastatic progression. This was concordant with mass spectrometry, which revealed markedly different metabolic profiles of the two HCC subtypes, with glycolysis and the tricarboxylic acid cycle pathway elevated in N1S1 tumors relative to normal liver, and with glycolysis and PPP relatively decreased in McA-RH7777 tumors.

The mRNA analysis revealed upregulation of the key glycolysis mediators GLUT1, HK2, PFK1, PFKFB3, and PKM in N1S1. GLUT1 aids in glucose transport across the plasma membrane, a key first step in glucose metabolism and a factor commonly expressed in tumors (30). Previous studies report that the expression of GLUT1 is associated with HCC differentiation, with significantly greater expression in poorly differentiated tumors (31–33). PFKFB3 encodes iPFK2, an enzyme that catalyzes fructose-6-phosphate to fructose-2,6-bisphosphate (F-2,6-BP) to activate PFK1, which commits the cell toward glycolysis (34). N1S1 is therefore highly dependent on glycolysis for maintenance and progression, and these mediators represent potential therapeutic targets. In a previous study, PFKFB3 expression in baseline and high-grade N1S1 was closely coupled to lactate flux, and this metabolic pathway and subsequent tumor growth could be suppressed with direct inhibition and tracked with HP ¹³C pyruvate MRI (15).

In contrast to N1S1 tumors, both glycolytic and PPP-related metabolites were decreased in McA-RH7777 tumors, revealing that glucose metabolism does not serve as its main energy source, consistent with previous studies using the same tumor line (35). We confirmed that in this glycolysis-independent variant of HCC, HP ¹³C pyruvate MRI could not depict the McA-RH7777 lesion or track its progression when stimulated to higher proliferation and metastasis. Rather, mRNA and mass spectrometry revealed that amino acid metabolism, including purine, arginine, proline, polyamines, and branched-chain amino acids, is the primary metabolic driver of McA-RH7777 at baseline. Purine metabolism, which is critical for nucleotide synthesis, and branched-chain amino acids, which are associated with tumor proliferation (36), were further upregulated in stimulated McA-RH7777 tumors with higher proliferation and metastasis. Although this demonstrates a limitation of HP ¹³C pyruvate MRI, this technique may instead be used not for tumor detection but as a technique to assess the level of glycolytic dependence, which could inform therapeutic options.

Both HCC models demonstrated upregulated amino acid metabolism, consistent with other studies (37). This is known to play a vital role in tumor maintenance and in the tumor microenvironment by serving as an alternative energy source, helping to maintain redox balance, and contributing to epigenetic regulation and immune response for tumorigenesis and metastasis (9). One example metabolite is glutamate, a major bioenergy substrate for cancer cell growth and activator of mitogen-activated protein kinase and phosphoinositide 3-kinase (PI3K), which promote cell survival, proliferation, and migration (38), and PI3K/Akt signal has been linked to RFA-induced tumorigenesis (39). These offer alternative diagnostic targets to assess HCC growth with potentially greater accuracy than techniques that target glycolysis. Hyperpolarization of nitrogen 15, which would potentially allow amino acid labeling, may offer this capability in the future (40).

Our study was limited by comparison between only two HCC variants, although this was sufficient to show different metabolic profiles, with implications for glycolysis-based imaging and therapy. In clinical practice, most human HCCs express high glucose uptake and downstream glycolysis similar to the N1S1 model, whereas variants show low glycolysis and GLUT1 expression similar to the McA-RH7777 model (41). Although our study focused on preclinical hepatoma models, the results provide a foundation and justification for further studies with human-derived tumor models and clinical trials.

In this study, we showed that HCC subtypes can exhibit glycolysis independence, which can lead to false-negative results with HP ¹³C pyruvate MRI. HP ¹³C pyruvate MRI may instead be used to select HCCs that may be amenable to glycolytic-targeted therapies and tracking.