Locoregional Lactate Dehydrogenase Inhibition Potentiates Therapy and Overcomes Treatment Resistance in Hepatocellular Carcinoma

Abstract

Background & Aims:

Metabolic inhibitors have demonstrated limited efficacy for cancer therapy due to metabolic plasticity and systemic toxicity. Locoregional therapies (LRT), such as transarterial embolization (TAE) or transarterial chemoembolization (TACE), generate ischemic stress that reprograms the tumor microenvironment (TME) toward glycolytic dependency, creating an opportunity to sensitize hepatocellular carcinoma (HCC) to metabolic inhibition. This study investigated whether pharmacologic inhibition of lactate dehydrogenase (LDH) with NCATS-SM1441 could exploit TAE-induced metabolic vulnerabilities to improve therapeutic efficacy in HCC.

Approach & Results:

Human HCC cell lines were exposed to replete or ischemic (TAE-like) conditions and treated with the LDH inhibitor NCATS-SM1441. Glucose/lactate flux, adenosine triphosphate (ATP) levels, and viability were assessed. In vivo, a diethylnitrosamine (DEN)-induced rat HCC model was treated with intraarterial NCATS-SM1441, TAE, or their combination. Drug distribution, tumor metabolism, necrosis, and survival were analyzed using mass spectrometry imaging, histopathology, T2-weighted magnetic resonance imaging (MRI), and survival metrics. Ischemic conditions induced LDHA expression and glycolytic flux, enhancing susceptibility to LDH inhibition. The combination of intraarterial NCATS-SM1441 before embolization increased intratumoral drug accumulation, reduced systemic exposure, and synergized with TAE to suppress lactate production, promote tumor necrosis, and significantly extend local progression-free survival.

Conclusions:

TAE conditions the TME to create a therapeutically targetable glycolytic dependency. Combining TAE with LDH inhibition overcomes key limitations of metabolic inhibitors as monotherapies, enhancing local control and survival with minimal systemic toxicity, supporting integration of metabolism-targeted agents with LRT for unresectable HCC.

INTRODUCTION

Metabolism-targeting drugs have historically yielded underwhelming results in cancer treatment with significant toxicity (1). Recent studies have highlighted the importance of conditioning the tumor microenvironment (TME) to induce specific metabolic dependencies before applying metabolic inhibitors (1). By altering the tumor’s metabolic landscape, it is possible to enhance the therapeutic efficacy of metabolism-targeting drugs while minimizing off-target effects. In this context, locoregional therapies (LRT) such as transarterial embolization (TAE), transarterial chemoembolization (TACE) offer a unique opportunity to condition the tumor microenvironment and induce metabolic vulnerabilities (2). Indeed, LRTs are first-line therapies for patients with intermediate stage hepatocellular carcinoma (HCC), a devastating and notoriously chemo-resistant disease (3). Approximately 40% of all HCC patients receive LRTs, which have a proven survival benefit (4). However, overall patient response is heterogeneous and local recurrence is common, leading to poor long-term survival rates (<2 years) (5). Surprisingly, TACE may not be more effective than TAE, suggesting that embolization plays a more central role in tumor targeting as compared to chemotherapeutics (6). LRTs occlude the hepatic artery with embolic material, creating an ischemic TME which deprives HCC of oxygen and nutrients. HCC adapts to this ischemia via a switch toward glycolysis (7). Since ischemic HCC cells depend heavily on anaerobic glycolysis for energy production, combining LRTs with glycolytic inhibition could both create and target a metabolic vulnerability in HCC. Recent advances in the understanding of this metabolic reprogramming have resulted in the development of several promising therapeutic strategies including agents targeting glucose uptake as well as enzymes critical for bioenergetic homeostasis such as lactate dehydrogenase (LDH) (8).

LDH, a key enzyme involved in cellular metabolism, catalyzes the conversion of pyruvate to lactate (Fig. 1a), a process that is often upregulated in cancer cells to meet their high energy demands under low oxygen conditions (9). LDH is a tetrameric protein comprising two major subunits, Ldha (subunit M) and Ldhb (subunit H) genes, and is increasingly recognized for its role in cancer metabolism and progression (10). Recent studies have highlighted its role in the context of the ischemic tumor microenvironment that is characteristic of HCC (11). Indeed, elevated LDH levels are associated with aggressive tumor behavior and poor clinical outcomes in various cancers, including HCC (12). Additionally, the consequence of lactate accumulation is the acidification of the extracellular pH in the TME, which is believed to favor vascularization, metastatic and immunosuppressive processes and is associated with a worse clinical prognosis (13). Lactate is now considered a critical oncometabolite in the metabolic reprogramming of cancer cells, and LDH a potential therapeutic target for cancer treatment. Evidence to support the role of LDH in cancer progression stems from the finding that pharmacological inhibition of LDHA caused the regression and growth delay of tumors in experimental models such as HCC (14). Although LDHB knockdown has been reported not to significantly impact tumor cell survival, genetic alteration of LDHA induced cell death in several cancer cell lines and impaired HCC development in a c-Myc/h-Ras murine model (15). Despite this dependency, we have demonstrated in preclinical studies that systemic administration of an LDH inhibitor results in significant, but modest growth delays (14).

Figure 1. LDH expression and significance on survival in HCC tumors.

A, Simplified schematic of glycolysis. Glycolysis is a metabolic pathway that breaks down glucose into pyruvate, producing ATP and NADH, and is often upregulated in cancer cells like HCC to support rapid growth and survival. The final reaction, conversion of pyruvate to lactate, is performed by the lactate dehydrogenase (LDH) and sustains redox balance, becoming critical under hypoxia/ischemia. B, TCGA LIHC dataset analysis shows expression of lactate dehydrogenases (LDHA, LDHB, LDHDC, LDHD, LDHAL6A and LDHAL6B) in HCC tumors vs. normal liver tissue. 373 HCC tumors and 50 normal liver tissues were included. LDHA, Lactate Dehydrogenase A; LDHB, Lactate Dehydrogenase B; LDHDC, Lactate Dehydrogenase C; LDHD, Lactate Dehydrogenase D; LDHAL6A, Lactate Dehydrogenase A like 6A; LDHAL6B, Lactate Dehydrogenase A like 6B. Statistical analysis was applied with *, P < 0.05; ***, <0.001. C, Representative immunohistochemistry images of LDHA protein expression in HCC and normal liver tissues. Images obtained from the Human Atlas Protein (version 24.1, www.proteinatlas.org, https://www.proteinatlas.org/ENSG00000134333-LDHA, accessed on October 2025, Uhlén et al.(40)). Images used in accordance with the Human Protein Atlas terms of use. D, Relative abundance of LDHA phosphorylated at Y10 between HCC tumor and liver tissues. Phosphorylation at Y10 enhances LDHA activity by enhancing the formation of active tetrameric LDHA, and NADH binding. Median and upper and lower quartiles are represented by the line and box, respectively. Analysis performed from Gao et al. Dataset (21). Statistical analysis was applied with ***, <0.001. E, Kaplan-Meier analysis indicating that patients with tumors that expressed high levels of LDHA mRNA had a shorter overall survival (TCGA LIHC). 364 HCC tumors split into halves based on LDHA mRNA expression. F, Analysis of LDHA dependency scores across various cancer cell lines using DepMap dataset (Broad Institute), showing that among all cancer cell lines, HCC and liver cell lines are the most dependent on LDHA.

Using in vitro (HCC cell lines) and in vivo (diethylnitrosamine (DEN)-induced HCC rat) models, we demonstrate that HCC cells rely heavily on glycolysis and LDH under ischemic conditions produced via TAE/TACE. Our findings indicate that combining LDH inhibitor (NCATS-SM1441) with LRT may realize the potential of metabolic inhibitors to enhance therapeutic efficacy and outcomes for patients with unresectable HCC.

METHODS

Transarterial Embolization Procedure

Rats were induced and maintained under anesthesia with 2% isoflurane via nose cone, then pre-procedural MR imaging was used to identify all visible tumors. After depilation and aseptic preparation of the right hind limb and lower abdomen, rats were placed supine on a radiolucent surgical table equipped with heating pads for normothermia. Sterile drapes were applied, and analgesia was administered per IACUC-approved protocol, consisting of subcutaneous injections of bupivacaine and meloxicam. A small incision was made to access the right common femoral artery. Under a stereomicroscope (Leica M80), the femoral artery was isolated, and an arteriotomy was created using a 26-gauge needle. A 1.7-F microcatheter system (Terumo, Tokyo, Japan) was then advanced under fluoroscopic guidance into the hepatic artery.

Real-time fluoroscopy-guided TAE was conducted using Siemens Cios Alpha or Siemens ARTIS Pheno fluoroscopic imaging systems, featuring digital subtraction angiography (DSA) and digital cine modalities. Fluoroscopy images were acquired at 5–7 frames per second and iodinated contrast medium (Iodixanol) was injected to precisely visualize and select tumor-feeding arteries (Sup. Fig. 1). Tumors in the targeted vascular territory were designated as target tumors, and all other visible tumors outside were designated as non-target tumors. Embolization was performed using microspheres (Embosphere; Merit Medical Systems Inc., South Jordan, UT), diluted to volumes of 300–1000μL, adjusted based on vessel diameter and tumor size. Microsphere delivery was precisely controlled using a Standard PHD ULTRA^™ CP Syringe Pump (Harvard Apparatus, Holliston, MA) at a rate of 75–100μL/min. For LDHi treatment groups, NCATS-SM1441 (10mg/kg) was administered intra-arterially immediately prior to embolization. Post-embolization digital subtraction angiography confirmed successful occlusion of target tumors (Sup. Fig. 1). The catheter was subsequently withdrawn, the arteriotomy closed, and the incision site sutured using 4–0 absorbable sutures. Post-surgery, animals were monitored closely. We have previously described further procedural details in Gade et al., 2015 (16).

MRI Acquisition and Tumor Volume Analysis

Post-treatment tumor response was evaluated using weekly MRI scans for four weeks (or until endpoint), as described above. Tumor segmentations were performed in a semi-automated fashion, and volumes were subsequently measured using ITK-SNAP software (17). Gompertz-transformed tumor volumes and linear regressions from tumor volumes measurements on MR imaging provided specific growth rates for each tumor over time (18). In this transformed tumor volume data, negative slopes indicate continued growth, while positive slopes reflect regression. To assess treatment response on follow-up MR imaging, RECIST-1.1 criteria were applied to individual target tumors during the volume analysis (19).

Desorption Electrospray Mass Spectrometry Imaging (DESI-MSI)

Target and non-target tumors were suspended in 12.5% fish gelatin media (G7041-100G, Sigma Aldrich) and sectioned at a thickness of 10μm using a Leica CM1950 cryostat. The sections were thaw-mounted onto superfrost glass slides and stored at −80°C until analysis. Mass spectrometry imaging (MSI) data were acquired using a Xevo G2-XS QTOF mass spectrometer equipped with a DESI-XS source (Waters Corporation, Milford, MA, USA). Data were acquired in negative ion mode between m/z 50–1200. The spray solvent consisted of 98% methanol with 50pg/mL leucine enkephalin at a flow rate of 2μL/min. The spray voltage was set to −0.6kV, nebulizing N₂ gas flow maintained at 600L/H, and heated transfer line temperature set to 50°C. The sampling area was defined using High Definition Imaging (HDI) v1.7 with pixel size 50×50μm, and raster rate 200μm/s. Immediately following acquisition, sections were again stored at −80°C until H&E staining was performed. Processing, statistical analysis, and visualization of MSI data was performed with MSI.EAGLE v0.1 (20).

Statistics and data availability

Data are expressed as mean±SD. The results were analyzed by Student’s t tests, Wilcoxon signed-rank tests, Kaplan-Meier estimates, and linear regressions where appropriate. P-values less than 0.05 were considered statistically significant (*P<0.05, **P<0.01, and ***P<0.001). Graphs and statistical analyses were generated and performed using Prism10 (GraphPad Software, Boston, Massachusetts). Refer to Supplementary Data for complete methods. Further information and request for resources and reagents should be directed to Terence Gade (tgade@pennmedicine.upenn.edu). This study did not generate unique new reagents.

RESULTS

Although glycolysis is known to contribute to HCC adaptation under TAE/TACE-like ischemic conditions (7), analysis of RNA sequencing data from The Cancer Genome Atlas (TCGA) revealed that LDHA and LDHD expression are lower in HCCs relative to liver, while LDHC, LDHAL6A and LDHAL6B expression levels are comparable (Fig. 1b). LDHA protein expression appeared similar between HCC samples and liver in the Human Protein Atlas (Fig. 1c). Analysis of data from a large-scale phospho-proteogenomic study (21) showed that LDHA Y10 phosphorylation, which enhances LDHA activity, was significantly lower in HCC tumors as compared to liver (Fig. 1d). These findings are consistent with a prior gene set enrichment analysis demonstrating that anaerobic glycolysis genes are not significantly dysregulated in primary human HCC tumors compared to normal liver (22). Despite these findings, high LDHA expression correlated significantly with worse overall patient survival in TCGA datasets (Fig. 1e), while the other isoforms did not (Sup. Fig. 2a–e). Additionally, analyses of a genome-scale CRISPR-Cas9 essentiality screen from the Broad Institute revealed that LDHA is essential for proliferation of HCC cells. Gene dependency scores identified LDHA as a top vulnerability in both HCC and normal liver lineage contexts, with the strongest dependency observed in liver and pancreas-derived malignancies across all cancers (Fig. 1f).

These observations highlight a critical distinction between absolute expression levels and functional dependency. While HCC tumors exhibit lower basal LDHA expression and phosphorylation compared to normal liver (Fig. 1b–d), the significant correlation between high LDHA expression and poor survival, combined with the CRISPR essentiality data (Fig. 1e, 1f), suggests that HCC cells become selectively dependent on LDHA-mediated glycolytic metabolism. This selective dependency may reflect the metabolic reprogramming characteristic of HCC, where cancer cells rely more heavily on specific metabolic pathways despite reduced absolute enzyme levels (14). Importantly, the ischemic conditions induced by TAE/TACE would be expected to further amplify this dependency, as oxygen deprivation forces greater reliance on anaerobic glycolysis for ATP generation. In this context, even modest levels of LDHA become critical for HCC survival, while normal hepatocytes maintain greater metabolic flexibility and can utilize alternative oxidative pathways. Thus, LDHA represents a targetable vulnerability that is selectively essential in HCC, particularly in the ischemic post-embolization environment where glycolytic dependency is maximized.

We and others have reported that HCC cells rely on oxidative phosphorylation under replete oxygen conditions and rapidly switch their metabolism to anaerobic glycolysis when exposed to ischemic or TAE-like conditions (7,23), increasing LDHA dependence. We previously defined TAE-like conditions as DMEM supplemented with 1% FBS, 1mM glucose, 0.5mM glutamine and 1% oxygen (7). To characterize metabolic microenvironment-induced alterations in LDHA protein expression, gene expression, and activity we exposed HCC cell lines (HepG2, SNU398, Huh7 and SNU449) to replete (Standard - Std) or ischemic (Isch) conditions. Consistent with our previous findings (7), qPCR and Western blot analyses confirmed that LDHA mRNA and LDHA protein levels as well as other ischemic-responsive genes (SLC2A1, CA9, PDK1, Sestrin2, LDHD) increased under TAE-like conditions (Fig. 2a, 2b and Sup. Fig. 3). Glucose consumption and lactate production measurements at 24hrs also revealed enhanced glycolytic activity in HCC cell lines under ischemic conditions (Fig. 2c) demonstrating the metabolic reprogramming occurring in HCC cells upon exposure to TAE-like conditions.

Figure 2. LDHA expression and activity are enhanced under ischemic conditions and targetable.

A, Real-time qPCR analysis of LDHA mRNA levels in HepG2, SNU398, Huh7 and SNU449 cells grown in standard or ischemic conditions for 48 hours. 18S ribosomal RNA (18S) was used as the housekeeping gene. B, Phospho-LDHA and total LDHA protein expression in HepG2, SNU398, Huh7 and SNU449 cells grown in standard or ischemic conditions for 48 hours (left panel). Quantification of Phospho-LDHA and total LDHA expression normalized as Phospho-LDHA/HSP90 and LDHA/HSP90 ratios (right panels). HSP90 was used as the loading control. C, YSI analyzer measurements of glucose level in HepG2, SNU398, Huh7 and SNU449 cells grown in standard or ischemic conditions for 24 hours, and YSI analyzer measurements of lactate level in HepG2, SNU398, Huh7 and SNU449 cells grown in standard or ischemic conditions for 24 hours. D, YSI analyzer measurements of glucose and lactate levels in SNU387 cells grown in standard conditions in presence of various concentrations of LDH inhibitor (NCATS-SM1441) for 48 hours. E, YSI analyzer measurements of lactate level in SNU387 cells grown in standard and ischemic conditions in presence of various concentrations of LDH inhibitor (NCATS-SM1441) for 6 hours. F, YSI analyzer measurements of lactate level in SNU387 cells grown in modified standard conditions (low serum – 1% FBS, and low serum – 1% FBS + albumin add back) in presence of various concentrations of LDH inhibitor (NCATS-SM1441) for 6 hours. All experiments were performed in at least triplicates and statistical analysis was applied with *, P < 0.05; **, P < 0.01; ***, <0.001).

To assess metabolic microenvironment-induced alterations in LDH activity, we utilized NCATS-SM1441, a small-molecule compound developed through the National Center for Advancing Translational Sciences (NCATS-NIH) chemical library. The LDH inhibitor (LDHi, NCATS-SM1441) showed nanomolar efficacy for the inhibition of lactate production in SNU387 HCC cells (Fig. 2d). Intriguingly, LDHi efficacy was significantly increased under ischemic conditions (Fig. 2d) with a 7-fold reduction in the concentration required to achieve equivalent levels of lactate reduction. Among the primary differences between replete/Std media and ischemic media is the percentage of FBS. It is well-established that serum albumin can attenuate drug activity through binding and reducing its free concentration (24). To control for the possibility that the observed difference resulted from concentrations of serum albumin, we repeated lactate production measurements under standard, ischemic and low serum (1% FBS) conditions, as well as low serum supplemented with albumin (Fig. 2e, 2f). As expected, the effect of LDHi on lactate production observed under ischemic conditions (Fig. 2e) was recapitulated under low serum conditions (Fig. 2f); however, the addition of albumin to the low serum condition restored lactate production to a level similar observed under standard conditions (Fig. 2f), demonstrating that serum albumin significantly reduces the effect of NCATS-SM1441.

We next evaluated the potential to leverage the observed changes in LDH expression and activity to potentiate targeting of HCC cells. While LDHi impaired viability under standard conditions, 72-hour treatment under ischemia resulted in complete cell death (Fig. 3a), with a markedly lower IC50 (Fig. 3b). Flow cytometry confirmed that reduced viability was driven by increased apoptosis, as indicated by annexin-V/PI staining (Fig. 3c and Sup. Fig. 4a), without increased ROS production (Sup. Fig. 4b). Mechanistically, LDHi significantly reduced cellular ATP levels at 48hrs post-treatment (Fig. 3d), consistent with a bioenergetic crisis triggering apoptotic cell death. Of note, viability remained similar between standard and ischemic conditions at 48hrs (Sup. Fig. 4c), underscoring that the cell death phenotype requires sustained nutrient and oxidation deprivation together with metabolic inhibition. Collectively, these data demonstrate that LDH inhibition under ischemic conditions drives metabolic stress and cell death, providing a rationale for enhanced efficacy of LDHi in the TAE/TACE-like microenvironment.

Figure 3. LDHA inhibition impairs HCC cell viability and induces apoptosis through energetic crisis.

A, Viability, assessed by WST1 assay, of HepG2, SNU398, Huh7 and SNU449 cells grown in standard or ischemic conditions for 72 hours in presence of various concentrations of LDH inhibitor (NCATS-SM1441). B, LDH inhibitor IC50 values for HepG2, SNU398, Huh7 and SNU449 cells grown in standard or ischemic conditions for 72 hours. C, Annexin-V/PI staining and flow cytometry analysis performed HepG2, SNU398, Huh7 and SNU449 cells grown in standard or ischemic conditions and with or without LDH inhibitor for 48 hours. D, ATP measurement in HepG2, SNU398, Huh7 and SNU449 cells grown in standard or ischemic conditions for 48 hours in presence of various concentrations of LDH inhibitor. ATP signals were normalized to untreated control cells. All experiments were performed in at least triplicates and statistical analysis was applied with *, P < 0.05; **, P < 0.01; ***, <0.001).

To explore the potential of LDHi in combination with TAE to improve outcomes in vivo, we performed pharmacokinetic (PK) studies in a DEN-induced HCC rat model. Once tumors reached >100 mm³, animals were randomized into four groups: 1) Sham (locoregional intra-arterial administration of saline), 2) Embo (locoregional intra-arterial administration of embolics alone), 3) LDHi (locoregional intra-arterial administration of NCATS-SM1441 without embolics), and 4) Embo+LDHi (locoregional intra-arterial administration of NCATS-SM1441 followed by embolics). To assess systemic exposure and potential toxicity, we measured plasma LDHi concentrations over time following administration. When LDHi was administered without embolization (LDHi group), plasma concentrations were elevated at early time points and significantly increased by 30min post-administration (Fig. 4a). In contrast, embolization performed after LDHi administration (Embo+LDHi group) resulted in lower plasma LDHi levels, demonstrating that systemic exposure is limited when the tumor is embolized (Fig. 4a). Consistent with prior studies, these findings suggest that embolization after intraarterial administration of therapeutics minimizes systemic exposure and toxicity (25,26). We next assessed the systemic metabolic effects of LDHi by measuring plasma lactate levels. At 30min post-administration, lactate levels were highest in rats receiving LDHi alone, intermediate in rats that underwent embolization following LDHi and lowest in rats with embolization alone (Fig. 4b). At 24hrs, lactate levels were lowest in the LDHi alone group, potentially reflecting a systemic metabolic correction towards equilibrium. Plasma lactate levels in the Sham, Embo, and Embo+LDHi groups were similar at 24hrs, and by 48hrs, all groups had comparable lactate levels. These results suggest that rats exhibited a more rapid return to baseline blood lactate levels when embolized after LDHi treatment, supporting the conclusion that embolization limits systemic metabolic effects of NCATS-SM1441.

Figure 4. NCATS-SM1441 Pharmacokinetics in a DEN-induced HCC rat model.

A, LDHi concentrations in rat plasma over the course of 48 hours after administration of LDHi with or without post-treatment embolization. Non-embolized rats are labeled “LDHi” and rats which were treated with transarterial LDHi then embolized are labeled “Embo + LDHi”. Sham and Embo control samples yielded no LDHi signal and are not shown. B, Lactate abundance in rat plasma over the course of 48 hours after administration of LDHi with or without post-treatment embolization. Additional control rats which only received saline or were embolized but received no LDHi (“Sham” and “Embo”, respectively) are shown. C, Replicate fold-change in LDHi concentrations between paired non-target and target tumors resected from the same rat, and lactate signal relative to control sham injections in paired non-target and target tumors resected from the same rat. D, H&E-stained tumors annotated for necrotic and living regions, annotation masks applied to these necrotic and living regions, and mass spectrometry images depicting spatial lactate and LDHi drug abundance. E, A violin plot of lactate levels per-MSI pixel in only live regions of tumors treated LDHi administration then embolization, or embolization alone. A volcano plot of untargeted MSI data per-pixel in the live regions of embolized tumors with and without LDHi administration. Features were binned using 30 ppm tolerance and 20 S/N cutoff, resulting in 2,080 m/z values. Fold-change values for each feature’s peak height and FDR values for the comparison between regions were plotted. F, A violin plot of lactate levels and volcano plot of untargeted features for only necrotic regions, analyzed and presented in the same manner at panel F.

Measurements of intratumoral concentrations of LDHi revealed that the percent increase in LDHi concentration in target versus non-target tumors was significantly higher when embolization was performed after LDHi administration compared to LDHi administration alone (Fig. 4c). These results are consistent with lower systemic LDHi in the Embo+LDHi group (Fig. 4a) and indicate that embolization after LDHi administration increases drug concentration within tumors. Target tumors exhibited lower lactate levels compared to non-target tumors across all treatment groups (Embo, Embo+LDHi, and LDHi) (Fig. 4c), consistent with increased cell death, necrosis, and reduced metabolic activity characteristic of embolized tumors. H&E-stained tumor tissue sections confirmed this observation (Fig. 4d). From the same rats, livers, kidneys, spleens and hearts showed minimal changes in lactate levels and LDHi concentrations between groups (Sup. Fig. 5a and 5b). This is consistent with the observed return to baseline of circulating plasma lactate levels at 48 hours post-procedure (Fig. 4b).

Necrotic tumor regions annotated from H&E staining in both Embo and Embo+LDHi groups co-localized to lactate-depleted regions in mass spectrometry images (Fig. 4d and Sup. Fig. 5c). Additionally, LDHi levels in the Embo+LDHi tumor section co-localized with necrotic, lactate-depleted regions, indicating the contribution of the inhibitor to regional treatment response (Fig. 4d, Sup. Fig. 5d). In both metabolically active, “live” regions and necrotic regions of the tumors, lactate abundance per-pixel was significantly lower in the Embo+LDHi group relative to the Embo group (Fig. 4e), demonstrating on-target efficacy of LDHi. In necrotic, “dead”, tumor regions, an untargeted metabolomic screen comparing 2,080 features revealed a substantially greater number of significantly increased metabolites in the Embo group versus the Embo+LDHi group (Fig. 4f), indicating that combined embolization and LDH inhibition enhances tumor cell death and metabolic disruption. Unexpectedly, in “live” regions there are a greater number of significantly increased metabolites in the Embo+LDHi group versus the Embo group. The pathways enriched in the “live” Embo+LDHi include fatty acid, pyrimidine metabolism, and the pentose phosphate pathway – all indicative of compensation for a lack of glycolytic activity (Sup. Fig. 5e). Furthermore, PCA of untargeted MSI data showed distinct clustering: Sham tumors clustered together, Embo tumors clustered together, and LDHi-treated tumors (both target and non-target, with and without embo) formed a third cluster (Sup. Fig. 5c). This provides preliminary evidence of a broader metabolic phenotype which extends beyond lactate level perturbation, impacting both embolized and non-embolized tumors in the LDHi treatment groups.

These pharmacokinetic studies demonstrated that LDHi administered prior to embolization enhances tumor drug accumulation and metabolic disruption, resulting in bioenergetic crisis. We therefore asked whether these effects translate into improved outcomes in vivo. Rats bearing DEN-induced HCCs were randomized into three cohorts including sham embolization (Sham, n=7), TAE (Embo, n=11), and locoregional intraarterial LDHi followed by TAE (Embo+LDHi, n=12). The Embo+LDHi cohort demonstrated significantly improved local progression-free survival (LPFS) of target tumors as compared to both the Sham (P<0.01) and Embo cohort (P<0.05), (Fig. 5a and Sup. Fig. 6). Additionally, results of the survival analysis demonstrated significantly improved overall survival for the Embo (P<0.01) and Embo+LDHi (P<0.01) cohorts as compared to the Sham cohort. In addition, rats in the Embo+LDHi cohort showed significantly improved overall survival as compared to the Sham cohort (P<0.01), but not the Embo cohort (P=0.33) (Fig. 5a).

Figure 5. TAE combined with NCATS-SM1441 improves tumor response and survival in a DEN-induced HCC rat model.

A, Kaplan-Meier curves showing local progression-free survival (A) and overall survival (B) in three cohorts of rats treated with sham embolization (Sham), conventional TAE (Embo), and conventional TAE with LDHi (Embo + LDHi). B, Bar plots showing the mean best response in target tumor volume following treatment in the Embo + LDHi cohort (C) and Embo cohort (D). Markers on the graph indicate RECIST classification or complete response (CR), partial response (PR), stable disease (SD), and progressive disease (PD). C, Bar plot showing the mean best response in target tumor size as % change from time of treatment in the Sham, Embo, and Embo + LDHi cohorts. D, Box-and-whisker plots showing the slope of the Gompertz transform tumor growth versus time for non-target (NT) and target (T) tumors in the Embo and Embo + LDHi cohorts. The dotted line marks a slope of 0. Negative slopes indicate net tumor growth, whereas positive slopes indicate net tumor regression. Boxes denote IQR with the median as the central line; whiskers extend to 1.5×IQR; points denote outliers. E, and F, Linear regression plots of Gompertz-transformed tumor growth rates following treatment in the Embo + LDHi cohort (G) and Embo cohort (H). Lines of best fit were derived from linear regression of all tumors within each RECIST response group in the respective cohort. Markers indicate individual tumors, and shaded areas represent 95% confidence intervals.

At the tumor level, seven target tumors in the Embo+LDHi cohort achieved CR as compared to none in the Embo cohort (Fig 5b). Consistent with these findings, target tumors in the Embo+LDHi cohort showed a significantly greater target tumor size reduction following treatment (P<0.01) versus the Embo cohort (Fig. 5c). Both Embo+LDHi (P<0.01) and Embo (P<0.05) cohorts showed a significant target tumor size reduction versus the Sham cohort. We analyzed slopes of Gompertz-transformed tumor volume over time to evaluate treatment effects on growth kinetics. Target tumors in the Embo+LDHi cohort demonstrated significantly more positive median slopes (median β = +0.0241, IQR 0.00709–0.0746), consistent with slowed growth or regression, whereas target tumors in the Embo alone cohort exhibited significantly more negative slopes (median β = −0.00239, IQR −0.0731–0.0296), indicating ongoing progression (P<0.01) (Fig. 5d). Non-target tumors in both cohorts showed negative slopes, though the Embo+LDHi group had a wider range with a median significantly closer to 0 (P<0.05). Tumor stratification using RECIST-1.1 (19) and linear regression analyses on Gompertz-transformed volumes over time (Fig. 5e and 5f) showed that in the Embo+LDHi cohort (Fig. 5e), CR tumors exhibited significantly more positive slopes (β=0.0562, 95% CI 0.0337–0.0787; P<0.001) consistent with regression, PR tumors showed flattened slopes indicating growth arrest (β=0.00518, 95% CI −0.00237 to 0.0127; P=0.176), and the single PD tumor demonstrated a steeply negative slope (β=-0.0220, 95% CI −0.0508 to 0.00676; P=0.0930) indicating continued growth. In contrast, the Embo cohort (Fig. 5f) showed that PR tumors demonstrated shallow negative slopes (β=−0.00463, 95% CI −0.0198 to 0.0106; P = 0.55), SD tumors similarly exhibited minimally negative slopes (β=−0.00122, 95% CI −0.0207 to 0.0183; P = 0.89), and PD tumors consistently demonstrated steeply negative slopes (β=−0.103, 95% CI −0.164 to −0.0413; P = 0.002), consistent with growth arrest in the PR and SD groups and continued rapid progression in PD tumors.

Intriguingly, although only a limited number of tumors per animal were directly targeted, in most cases only one, with LDHi and embolization, the Embo+LDHi group showed significantly improved local progression-free survival versus Embo. This outsized survival benefit, together with the significant reduction in the growth rate for non-target tumors, suggest systemic effects beyond local control in this cohort. Given recent data indicating the ability of LDHi to potentiate anti-tumor immune responses (27), we performed IHC analyses on target and non-target tumors from Sham, Embo and Embo+LDHi groups (Fig. 6). While we did not detect significant differences in CD3 and CD8 staining between groups for the target tumors (Fig. 6a and 6b), we observed a significant increase in CD3+ T cell infiltration (P<0.01) and a trend toward increased CD8+ cytotoxic T cells in non-target tumors (Fig. 6c and 6d). These findings suggest that local metabolic stress and necrosis induced by LDHi and TAE may trigger systemic immune responses capable of suppressing growth in distant, untreated tumors, thereby contributing to improved survival outcomes in our rat model.

Figure 6. Embolization and LDHi treatment increase immune presence.

A, Representative images of CD3 staining in target tumors from rats in the Sham, Embo or Embo + LDHi groups, and graphs representing quantification of CD3 staining as number of CD3 + T cells per high-power field in target tumors from rats in the Sham, Embo or Embo + LDHi groups. B, Representative images of CD8 staining in target tumors from rats in the Sham, Embo or Embo + LDHi groups, and graphs representing quantification of CD8 staining as number of CD8 + T cells per high-power field in target tumors from rats in the Sham, Embo or Embo + LDHi groups. C, Representative images of CD3 staining in non-target tumors from rats in the Sham, Embo or Embo + LDHi groups, and graphs representing quantification of CD3 staining as number of CD3 + T cells per high-power field in non-target tumors from rats in the Sham, Embo or Embo + LDHi groups. D, Representative images of CD8 staining in non-target tumors from rats in the Sham, Embo or Embo + LDHi groups, and graphs representing quantification of CD8 staining as number of CD8 + T cells per high-power field in non-target tumors from rats in the Sham, Embo or Embo + LDHi groups.

DISCUSSION

This study demonstrates that TAE can condition the HCC tumor microenvironment to unlock the therapeutic potential of LDH inhibition with the selective inhibitor NCATS-SM1441. This approach addresses a fundamental limitation of LRT, as recurrence following LRT remains a significant clinical challenge (28). TAE/TACE effectively induce ischemic and cytotoxic stress in HCC and provide a significant advantage by confining drug delivery and ischemia to the tumor while limiting systemic exposure. However, limitations include incomplete tumor necrosis, hypoxia-driven pro-tumorigenic signaling, metabolic rewiring and the rapid development of adaptive resistance pathways (29). These limitations are underscored by clinical studies which demonstrate that, despite a proven survival benefit, durable responses remain limited with recurrence rates of 76% in patients with demonstrating complete responses on follow-up imaging (30). These limitations highlight the need for rational adjunct strategies capable of exploiting the metabolic stress induced by embolization.

The development of metabolic inhibitors as cancer therapy has faced many challenges due to systemic toxicity and poor tumor selectivity (31). These limitations can be overcome with LRTs, and recent studies underscore the potential of metabolic inhibitors to potentiate ischemia-induced cytotoxicity (29,32). In our study, pharmacologic LDH inhibition under these ischemic/TAE-like conditions induced profound bioenergetic collapse and apoptotic cell death in vitro. Importantly, we observed that serum albumin diminished the in vitro effect of NCATS-SM1441. This effect is a known pharmacokinetic barrier limiting the systemic efficacy of small-molecule inhibitors (33), reinforcing the rationale for LRT delivery to maintain high intratumoral drug levels. In vivo, intraarterial delivery of NCATS-SM1441 prior to TAE enhanced intratumoral drug accumulation, suppressed lactate production, increased necrosis, and significantly improved local progression-free survival while minimizing systemic exposure. While overall survival did not differ significantly between TAE and TAE/LDHi cohorts, this likely reflects persistent multifocal tumor burden from newly developed lesions inherent to the carcinogen-driven DEN model. Together, these findings highlight the potential of LRT-induced metabolic conditioning to overcome intrinsic resistance to metabolic inhibitors and provide a rational framework for combining LRT with metabolism-targeted therapies to enhance efficacy and safety for patients with HCC.

The drug delivery strategy employed in this study was designed to leverage intact tumor vascularity for optimal drug distribution, with subsequent trapping of the LDH inhibitor within the ischemic TME following embolization. Mass spectrometry data demonstrated enhanced intratumoral accumulation and retention of NCATS-SM1441 using this approach (Fig. 4c). Clinical TACE protocols frequently incorporate drug delivery vehicles such as Lipiodol emulsions or drug-eluting beads to maximize intratumoral drug retention and minimize systemic washout (25). However, we elected to use direct intraarterial infusion without carrier agents to reduce variability in the degree of embolization that could confound interpretation of the metabolic synergy between ischemia and LDH inhibition. Future translational studies should explore the integration of NCATS-SM1441 or related LDH inhibitors with established drug delivery platforms currently used in clinical TACE protocol. This optimization of drug delivery represents an important next step toward clinical implementation of this therapeutic strategy.

An important consideration for the clinical translation of this approach is the development of imaging biomarkers to prospectively identify patients whose tumors are most likely to benefit from LDH inhibition following TAE. Recent advances in hyperpolarized magnetic resonance spectroscopic imaging (HP-MRSI) using dynamic nuclear polarization (DNP) of ¹³C-labeled substrates offer a promising solution to this challenge. Using DNP-MRSI, we have previously shown that administration of NCATS-SM1441 in the same rat model used here led to decreased intratumoral lactate production and dose-dependent slowing of tumor growth, with post-LDHi intratumoral alanine levels robustly predicting the degree of growth inhibition (14). We also showed that HCC cells surviving TAE-induced ischemia demonstrate increased glycolytic flux through lactate dehydrogenase, with reduced alanine/lactate ratios that correlate with tumor viability in regions considered non-viable by conventional contrast-enhanced MRI (7). This metabolic fingerprint, characterized by preferential conversion of pyruvate to lactate over alanine, directly reflects the LDH-dependent glycolytic reprogramming that our current study shows is targetable with NCATS-SM1441. Tumors demonstrating high lactate/pyruvate ratios and low alanine/lactate ratios would be predicted to have maximal benefit from subsequent LDH inhibition. Moreover, DNP-MRSI could enable non-invasive monitoring of on-target LDH inhibition by measuring real-time changes in pyruvate-to-lactate conversion following inhibitor administration, providing a pharmacodynamic readout of therapeutic efficacy. The integration of functional metabolic imaging into clinical workflows would allow for precision medicine approaches, ensuring that LDH inhibitors are deployed in patients whose tumors exhibit the glycolytic dependency necessary for therapeutic response. Future clinical trials combining TAE with LDH inhibition could therefore incorporate DNP-MRSI to both stratify patients based on pre-treatment metabolic profiles and monitor therapeutic response, maximizing the likelihood of clinical success.

Interestingly, principal component analysis (PCA) of molecular profiles revealed that metabolically targeted (Embo+LDHi and LDHi) lesions and non-target lesions co-clustered (Sup. Fig. 5), suggesting that local LDH inhibition triggers a systemic metabolic reprogramming in the host. Taken together, the 1) metabolic reprogramming in LDHi-treated tumors, regardless of embolization or target status, 2) LDHi+TAE reshaping the non-target tumor immune microenvironment, and 3) extended LPFS compared to TAE alone, indicate a pattern of systemic effects on distant, untreated tumors. While this Abscopal effect is suggested by our preliminary findings, the experiments were not designed to provide detailed characterization of this phenomenon, and further studies are warranted.

Mechanistically, these results align with emerging evidence that excess lactate in the TME is a key driver of immune evasion. Lactate accumulation fosters an immunosuppressive milieu by inhibiting effector T and natural killer cells as well as promoting regulatory T cells, myeloid-derived suppressor cells and pro-tumor macrophage polarization via protein lactylation and altered signaling pathways (13,27,34). The reduction of lactate levels may diminish immunosuppressive signaling and promote effector T-cell activity, thereby converting an otherwise “cold” TME into an “immune-permissive” TME. This concept is underscored by a recent preclinical study combining an LDH-A inhibitor (e.g. oxamate) with TACE in a rabbit model of hepatic squamous cell carcinoma metastases which demonstrated significantly improved tumor control and increased CD8+ T-cell infiltration into tumors compared to TACE alone (35). This synergistic effect underscores the prospect that targeting tumor metabolism can both potentiate direct tumor cell death and render the TME more immunogenic. These findings are particularly salient in the context of recent clinical trials combining TACE with immunotherapy. The phase III EMERALD-1 and LEAP-012 trials, which combined TACE with immune checkpoint inhibition (± anti-VEGF/kinase therapy), achieved improvements in LPFS but failed to translate into significant overall survival benefits (36,37). These outcomes suggest that metabolic barriers in the post-embolization TME, such as hypoxia-driven lactate accumulation and resultant immunosuppression, may have limited the efficacy of immunotherapy in those studies. The data described above indicate that modulating the TME via LDH inhibition may help to overcome some of these barriers, as the incorporation of metabolic intervention (LDH blockade) into TAE/TACE regimens could prime the tumor for immunotherapy.

Our study includes several limitations. First, the DEN rat model is multifocal and immunocompetent, which, while clinically relevant, complicates survival interpretation due to emergence of new tumors. The DEN model we utilized shows greater transcriptomic similarity to metabolic dysfunction-associated (MASH) HCC than viral hepatitis-driven disease (38), and serum LDH levels have been reported to be higher in MASH patients versus non-MASH (39), suggesting our findings may vary based on the background in which HCC arises. Validation in viral models such as rat PDXs, and etiology-stratified clinical trials will be essential to address this point. Second, only one LDH inhibitor and dosing strategy were evaluated. The timing of LDH inhibitor administration relative to embolization is critical, as hypoxic stress intensifies progressively after arterial occlusion; Further optimization of dose, timing, and duration will be required. Third, heterogeneity of hypoxic stress post-embolization could result in spatially and temporally variable LDH expression within tumor regions, creating differing levels of dependency on LDH. For instance, mildly hypoxic regions may maintain oxidative metabolism and evade LDH dependence. Tumor cell adaptation to hypoxia and activation of alternative metabolic pathways could also lead to less LDH dependence. Finally, although the metabolomics and phenotypic results suggest immune involvement, in-depth immunologic characterization was not performed and represents an important future direction. Exploring resistance mechanisms to LDH inhibition and identifying biomarkers to predict patient response will also be critical for translating this approach into clinical settings.

In conclusion, this study illustrates a novel therapeutic paradigm in which LRT is used to metabolically prime the tumor, enabling a previously underperforming class of drugs (metabolic inhibitors) to have pronounced efficacy. Ultimately, our study lays the groundwork for connecting the role of LDH in tumor metabolism with the potential benefits of using LDH inhibitors in combination with TAE/TACE to improve treatment outcomes for HCC patients.

List of Abbreviations:

ARRIVE

Animals in Research: Reporting In Vivo Experiments

ATP

Adenosine triphosphate

CR

Complete response

DEN

Diethylnitrosamine

HCC

Hepatocellular carcinoma

LDH

Lactate dehydrogenase

LDHi

Lactate dehydrogenase inhibitor

LRT

Locoregional therapy

MRI

Magnetic resonance imaging

DESI-MSI

Desorption electrospray mass spectrometry imaging

PD

Progressive disease

LPFS

Local progression-free survival

PR

Partial response

SD

Stable disease

TAE

Transarterial embolization

TACE

Transarterial chemoembolization

TME

Tumor microenvironment