Lithocholic acid-activated VDR in macrophages promotes HCC recurrence post-ablation via SOCS3-mediated suppression of CXCL16

Abstract

Background

Incomplete thermal ablation (iTA) for hepatocellular carcinoma (HCC) fosters an immunosuppressive microenvironment driving local recurrence, yet the metabolic cues linking tissue injury to immune evasion remain undefined. Bile acids, particularly lithocholic acid (LCA), are elevated post-ablation, but their functional role in HCC recurrence is unknown. Here, we investigate whether LCA accumulation after iTA promotes tumor recurrence by suppressing antitumor immunity and identify the underlying molecular mechanism.

Methods

Orthotopic and subcutaneous HCC models were subjected to iTA. Bile acid metabolomics, multicolor flow cytometry, and immunofluorescence staining were performed. Human residual HCC specimens were analyzed for vitamin D receptor (VDR) expression, macrophage polarization and invariant natural killer T (iNKT) cell density. For mechanistic studies, tumor-associated macrophage (TAM)-specific conditional knockdown of Vdr or Socs3, along with iNKT cell co-culture systems, were employed. The therapeutic efficacy of tauroursodeoxycholic acid (TUDCA), a bile acid modulator that counteracts LCA-mediated effects, was evaluated in vivo.

Results

iTA markedly elevated intrahepatic LCA levels, which correlated with M2-like TAM polarization and reduced iNKT tumor infiltration. LCA functioned as a VDR agonist in TAMs, transcriptionally upregulating suppressor of cytokine signaling 3 (SOCS3) and suppressing C-X-C motif chemokine ligand 16 (CXCL16). This VDR-SOCS3-CXCL16 axis was necessary and sufficient for impaired iNKT chemotaxis, as genetic ablation of VDR or SOCS3 in TAMs restored CXCL16 secretion. Conversely, LCA treatment inhibited iNKT cell chemotaxis and interferon-γ secretion, which were restored by exogenous CXCL16 supplementation. In human residual HCC, post-ablation tissues exhibited M2-like TAM polarization, and VDR⁺ TAM density inversely correlated with intratumoral iNKT cells. Pharmacologically, TUDCA remodeled the bile acid pool, suppressed LCA-VDR signaling, reinstated CXCL16 expression, reprogrammed TAMs toward an M1-like phenotype, and significantly inhibited post-ablation tumor growth.

Conclusions

We identify a hitherto unknown metabolic-immune checkpoint: iTA-induced LCA activates macrophage VDR, triggering SOCS3-mediated epigenetic suppression of CXCL16 and evading iNKT surveillance. TUDCA resets this checkpoint by restoring CXCL16 expression and iNKT function. Thus, targeting this axis with TUDCA represents a readily translatable adjunctive strategy to abrogate post-ablation HCC recurrence.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• Ablation is effective for early-stage hepatocellular carcinoma (HCC), yet the post-ablation immunosuppressive microenvironment frequently drives recurrence and poor outcomes. Although bile acids are known to modulate liver immunity through specific receptors, how local bile acid signals contribute to immune evasion after incomplete HCC ablation remains largely unknown.

WHAT THIS STUDY ADDS

• Incomplete HCC ablation elevates local lithocholic acid (LCA) levels, which activates vitamin D receptor (VDR) in tumor-associated macrophages, upregulates suppressor of cytokine signaling 3, and suppresses C-X-C motif chemokine ligand 16 (CXCL16) secretion. Suppression of CXCL16 reduces invariant natural killer T (iNKT) cell infiltration into the tumor microenvironment, thereby impairing antitumor immunity and promoting residual HCC progression after ablation. Tauroursodeoxycholic acid (TUDCA), a bile acid modulator, restores iNKT cell infiltration, attenuates post-ablation tumor progression, and improves survival in preclinical models.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• This study opens a new direction investigating the role of bile acid-immune crosstalk in post-ablative HCC recurrence. Metabolic-immune biomarkers (eg, LCA, CXCL16) should be incorporated into future clinical trial designs for locoregional therapies, supporting regulatory consideration of TUDCA or VDR-targeted agents. Furthermore, TUDCA-already used clinically for cholestatic liver diseases-could be repurposed as an adjunctive therapy to improve post-ablation outcomes.

Introduction

Local tumor recurrence following thermal ablation remains a major obstacle to improving long-term survival in patients with hepatocellular carcinoma (HCC). Although thermal ablation is effective for early-stage tumors, incomplete thermal ablation (iTA) occurs in 15–40% of procedures due to technical limitations.^{1 2} iTA leaves behind a sublethal transitional zone where residual tumor cells, exposed to a profoundly altered microenvironment, can drive aggressive recurrence.³ Therefore, elucidating the mechanisms by which the post-ablation milieu enables tumor escape is of paramount clinical importance.

Accumulating evidence implicates an ablation-induced immunosuppressive microenvironment as a key facilitator of recurrence.³ However, the upstream metabolic signals initiating this immune dysregulation are poorly defined. Bile acids (BAs), which are central to liver physiology and frequently perturbed in chronic liver disease, have emerged as potent immunomodulators.^{4 5} Clinical studies report elevated systemic BAs in patients with HCC,^{6 7} and specific BAs can polarize tumor-associated macrophages (TAMs) towards an immunosuppressive M2-like phenotype or inhibit effector T-cell responses.^{4 5} Despite this, the impact of ablation on the BA landscape and its functional consequence for antitumor immunity are entirely unknown.

To address this gap, we analyzed BA profiles in a murine iTA model and identified a striking increase in the secondary BA lithocholic acid (LCA). This LCA elevation was accompanied by a marked loss of hepatic invariant natural killer T (iNKT) cells, a lymphocyte population critical for anti-HCC immunosurveillance.⁸ The clinical relevance of iNKT cells is well established: low intratumoral iNKT cell density independently predicts worse overall and recurrence-free survival in patients with HCC,⁹ and iNKT cell infusion has demonstrated therapeutic benefit in a phase II randomized trial.¹⁰ Our findings align with prior reports that LCA can suppress iNKT cell trafficking to the liver.¹¹ Given that LCA is a known ligand for the vitamin D receptor (VDR), a nuclear receptor highly expressed in macrophages,¹² we postulated that LCA might exert its immunomodulatory effect via VDR activation in TAMs, which are a major source of the iNKT cell-recruiting chemokine C-X-C motif chemokine ligand 16 (CXCL16) in the liver.¹³

To identify the potential downstream mediator of VDR in this context, we screened for VDR-regulated transcripts in residual tumor tissues from patients with HCC post-iTA. This analysis revealed a consistent and significant upregulation of suppressor of cytokine signaling 3 (SOCS3), a well-established inhibitory transcriptional target of VDR known to broadly repress cytokine and chemokine signaling.¹⁴ SOCS3 functions as an intracellular negative feedback regulator that limits inflammatory signaling by inhibiting the JAK/STAT pathway,¹⁵ thereby modulating immune responses and chemokine expression. Although a direct link between SOCS3 and CXCL16 has not been reported, the established role of SOCS3 as a broad negative regulator of cytokine/chemokine signaling led us to a novel mechanistic hypothesis: that iTA-elevated LCA activates VDR in TAMs, which upregulates SOCS3 and subsequently suppresses CXCL16, thereby impairing iNKT cell recruitment and promoting tumor recurrence.

In this study, we aimed to bridge this critical knowledge gap by integrating clinical observations, mechanistic investigations, and therapeutic exploration. We sought to define the role of LCA in driving post-iTA immunosuppression and HCC recurrence, and to decipher the detailed VDR-SOCS3-CXCL16 axis in TAMs that undermines iNKT cell recruitment. Tauroursodeoxycholic acid (TUDCA), a clinically available hydrophilic BA with cytoprotective and anti-inflammatory properties,¹⁶ is known to antagonize hydrophobic BA-induced signaling.¹⁷ Therefore, we evaluated its therapeutic potential to reverse the LCA-driven VDR-SOCS3-CXCL16 axis and prevent post-iTA relapse. Together, our study delineates a novel immunometabolic axis underlying ablation failure and positions the repurposing of TUDCA as a promising, immediately translatable strategy to enhance curative outcomes.

Materials and methods

Cell lines and culture

Mouse HCC Hepa1-6 cells and the iNKT cell hybridoma DN32.D3 (kindly provided by Professor Li Bai and Huimin Zhang, University of Science and Technology of China) were cultured in high-glucose Dulbecco's modified eagle medium (DMEM) and Roswell Park Memorial Institute 1640 medium (RPMI-1640) (BI, Israel), respectively. All media were supplemented with 10% fetal bovine serum (BI) and 1% penicillin/streptomycin. Cells were maintained at 37°C with 5% CO_₂ and routinely tested negative for Mycoplasma.

In vitro hyperthermia model

To simulate peri-ablational heat stress, Hepa1-6 cell suspensions (10⁶ cells/mL) were incubated at 45°C for 10 min in a precision water bath, then immediately returned to 37°C culture. Conditioned medium (CM) was collected from heat-treated cells after 24 hours of further culture and used for TAM maturation. For in vitro studies, cells were treated with 20 µM LCA (M2816, Abmole), 200 µM TUDCA (M5158, Abmole), or equivalent volumes of vehicles (dimethyl sulfoxide (DMSO) or phosphate-buffered saline (PBS)) for the indicated durations (24–72 hours).

Isolation and culture of bone marrow-derived macrophages

Bone marrow cells were flushed from femurs and tibiae of 9–12 weeks old male C57BL/6 mice, passed through a 70 µm strainer, and cultured for 7 days in RPMI-1640 containing 10 ng/mL macrophage colony-stimulating factor (M-CSF). For M1 polarization, cells were stimulated with 100 ng/mL LPS (#HY-D1056, MCE, New Jersey, USA) and 20 ng/mL IFN-γ (#315-05, PeproTech, New Jersey, USA) for 48 hours. For M2 polarization, cells were stimulated with 20 ng/mL IL-4 (#214-14, PeproTech, New Jersey, USA) for 48 hours. Polarization marker expression was validated by quantitative PCR (qPCR).

To generate TAMs, bone marrow-derived macrophages (BMDMs) were stimulated for 24 hours with 30% (v/v) CM from heat-stressed Hepa1-6 cells. TAMs were operationally classified as M1-like (CD86⁺) or M2-like (CD206⁺) based on established markers, recognizing that this represents a simplified framework for assessing polarization direction rather than capturing the full spectrum of TAM heterogeneity. Consistent with this framework, M1-like TAMs are considered pro-inflammatory and antitumorigenic, whereas M2-like TAMs are considered anti-inflammatory and protumorigenic.

Animal experimentation

Male C57BL/6 mice aged 5–6 weeks were purchased from the Laboratory Animal Center of Sun Yat-sen University. Following a 1-week acclimatization, mice were used for the subsequent experiment. All animal experiments were performed in compliance with the National Institutes of Health (NIH) Guide for the Care and Use of Laboratory Animals and were reviewed and approved by the Institutional Animal Care and Use Committee (IACUC) of Sun Yat-sen University (SYSU-IACUC-2023-000682). Each mouse was treated as an independent experimental unit. Sample sizes were estimated based on prior experience and literature, instead of formal calculation. Mice were excluded from analysis if they met any of the following a priori criteria: failure of orthotopic tumor implantation; death unrelated to tumor progression before the experimental endpoint; severe morbidity requiring early euthanasia according to institutional guidelines. All exclusions were predetermined and applied before data analysis. No additional data points were excluded. A detailed number for each experiment is provided in the figure legends.

Subcutaneous HCC-iTA model

To establish a self-control ablation model, 2×10⁶ Hepa1-6 cells were subcutaneously injected into bilateral axillary areas per mouse. When the tumor volume reached approximately 300 mm³, unilateral tumor was surgically resected and processed for flow cytometric analysis. The remaining tumor was treated using a bipolar radiofrequency ablation (RFA) device (Radionics, Massachusetts, USA). The ablation needle was inserted into the tumor core and treatment was applied for 30 s at a power setting of 3 W to simulate incomplete ablation. 7 days after incomplete radiofrequency ablation (iRFA), the residual tumor was collected and subjected to flow cytometric analysis.

Orthotopic HCC-iTA model

For the orthotopic HCC-iTA model, mice were anesthetized with tribromoethanol, and the left lobe of the liver was surgically exposed. A suspension of Hepa1-6 cells mixed with Matrigel at a ratio of 1:1 was injected directly into the left hepatic lobe, with a total of approximately 5×10⁶ cells per mouse. This relatively higher cell number was chosen based on pilot experiments to ensure consistent tumor formation and sufficient residual tumor tissue after iTA for subsequent mechanistic analyses. Tumor growth was monitored using a high-resolution small animal ultrasound imaging system (VisualSonics Vevo 2100, Canada). When tumors reached a volume of approximately 100 mm³, iRFA was performed using the same bipolar device. One day post-ablation, mice were randomly assigned to four treatment groups and administered 20 mg/kg LCA, 100 mg/kg TUDCA, or corresponding vehicles via oral gavage every other day for 7 days (LCA) and 14 days (TUDCA). To minimize potential confounders, all surgical procedures were performed by the same experienced operator, and all outcome assessments were conducted by investigators blinded to group allocation.

Patients and specimens

Formalin-fixed, paraffin-embedded tumor tissues were obtained from patients with HCC who underwent salvage resection for recurrent disease after initial RFA (n=8) or from ablation-naïve resection controls (n=8) at Sun Yat-sen Memorial Hospital (2003–2022). Patients receiving prior transarterial chemoembolization (TACE), chemotherapy, or immunotherapy were excluded.

Bile acid measurement

BAs in plasma and the liver tumor were analyzed by ultra-high performance liquid chromatography (UHPLC)-mass spectrometry, as described in.¹⁸ Approximately 20 mg of liver tumor tissue was homogenized. Following BA extraction and protein precipitation with methanol and acetonitrile, the supernatant and plasma were transferred to glass vials with microinserts and analysis was performed on an Agilent 1290 Infinity II UHPLC system coupled to a 6470B triple-quadrupole mass spectrometer equipped with a dual AJS electrospray ionization source (Agilent Technologies, Santa Clara, USA).

Flow cytometry analysis

Single-cell suspensions were stained with anti-mouse antibodies for surface proteins (detailed in online supplemental table S1). For iNKT cell staining, cells were incubated with anti-mouse α-GalCer-CD1d (#140506, BioLegend, UK) at 4°C for 30 min in the dark. After two washes with PBS buffer, cells were subsequently stained with surface antibodies (anti-CD45 and anti-T-cell receptor (TCR)-β; detailed in online supplemental table S1) for an additional 30 min at 4°C in the dark. For other surface protein stains (online supplemental table S1), cells were incubated with the corresponding antibodies for 30 min at 4°C protected from light. For intracellular targets, cells were fixed and permeabilized (a FIX and Perm Cell Permeabilization kit, GAS003, Invitrogen, USA) prior to staining according to the manufacturer’s instructions. Intracellular antibodies (detailed in online supplemental table S1) were diluted in 1X Permeabilization Buffer and incubated for 30 min at 4°C. Data were processed using Cytoflex LX (Beckman Coulter, Brea, USA) and analyzed with CytExpert software (V.2.5.0.77).

RNA interference in TAMs

TAMs were generated as previously described and transfected with 50 nM small interfering RNA (siRNA) using Lipofectamine 3000 reagent (Invitrogen, Thermo Fisher Scientific, Carlsbad, USA). Mouse-specific Vdr-targeting siRNA (GenePharma, Shanghai, China) was employed with the following sequence: sense: 5’-GCAGCCAA GACUACAAAUATT-3’; antisense:5’-UAUUUGUAGUCUUGGCUGCTT-3’. Mouse-specific Socs3-targeting siRNA (GenePharma, Shanghai, China) was used with the following sequence: sense: 5’-GGUCACCCACAGCAAGUUUTT-3’; antisense: 5’-AAACUUGCUGUGGGUGACCTT-3’. Non-targeting control siRNA (siNC) was used as a negative control. Transfected TAMs were allocated into four experimental groups for each knockdown experiment: (1) siNC+0.08% DMSO vehicle; (2) siVDR or siSOCS3+0.08% DMSO vehicle; (3) siNC+20 µM LCA (in 0.08% DMSO); (4) siVDR or siSOCS3+20 µM LCA (in 0.08% DMSO). Following a 48-hour or 72-hour incubation period post-transfection, cells were harvested for RNA extraction and protein detection, while supernatants were collected for quantification of secreted CXCL16 via ELISA.

Migration and functional analysis of DN32.D3 toward bile acid-treated TAMs

TAMs were pretreated with 20 µM LCA, 200 µM TUDCA, or vehicles for 24 hours, with or without 100 ng/mL recombinant CXCL16 (#250-28, PeproTech, New Jersey, USA) added to the LCA treatment group. For migration assays, 5 µm pore transwell inserts (Corning, New York, USA) were placed over TAM-containing wells. DN32.D3 cells (1×10⁶) were suspended in RPMI-1640 medium and placed in the upper chambers. After 24 hours of co-culture, the medium from the lower chamber was collected. Flow cytometry was used to analyze equal numbers of α-GalCer-CD1d⁺TCRβ⁺ cells, and the percentage of IFN-γ⁺ cells within this population was determined after Golgi blockade.

Quantitative real-time PCR

Total RNA was extracted using the EZB Rapid RNA Extraction Kit (EZBioscience, China) according to the manufacturer’s instructions. Complementary DNA was synthesized from total RNA using the EZB Reverse Transcription Kit (EZBioscience, China). Quantitative real-time PCR was performed using the EZB SYBR Green qPCR Master Mix (EZBioscience, China) on a real-time PCR system. Relative gene expression was calculated using the 2^−ΔΔCt method with β-actin as the internal control. The sequences of primers can be found in online supplemental table S2.

CXCL16 ELISA assay

Mouse serum, homogenized tumor tissues and concentrated supernatants from TAMs were prepared and stored at −80°C. CXCL16 was quantified by a commercial ELISA kit (#RX200282M, Ruixin Biotech, Guangzhou, China) with absorbance read at 450/570 nm following the manufacturer’s instructions. All samples were analyzed in triplicate with normalization to protein concentration or volume.

Western blot

Protein lysates were subjected to western blotting using primary antibodies (anti-VDR, 1:1000, #12550, CST; anti-SOCS3, 1:1000, #ab280884, Abcam; β-actin, 1:5000, #4967, CST) overnight at 4°C. After washing, membranes were incubated with HRP-conjugated secondary antibodies for 1 hour at room temperature. Protein bands were visualized using an enhanced chemiluminescence detection reagent.

H&E staining of liver tumor tissue

Liver tumor tissues were fixed in 4% paraformaldehyde, paraffin-embedded, and sectioned at 4 µm. Sections were stained with H&E following standard protocols, then dehydrated, cleared, and mounted. Images were acquired using a light microscope.

Immunofluorescence staining and multiplex immunofluorescence

TAMs were incubated overnight at 4°C with primary antibodies, including anti-VDR (#12550, CST, Danvers, USA), anti-SOCS3 (#FNab08100, FineTest, Wuhan, China), and anti-CXCL16 (#ER1906-85, Huabio, Hangzhou, China). SOCS3 and CXCL16 were detected with Cy5-conjugated secondary antibodies (#GB27303, Servicebio, Wuhan, China) while VDR used HRP-conjugated secondary antibodies (#GB23303, Servicebio, Wuhan, China), followed by signal amplification using tyramide signal amplification (TSA)-650 fluorescent dye (#RC008, Huilan Biotech, Shanghai, China).

Mouse and human paraffin-embedded tissue sections underwent sequential multiplex immunofluorescence (mIF) staining using a TSA-based multiplex immunostaining kit (RC0086plus-34RM, RecordBio, Shanghai, China). Mouse sections were incubated with primary antibodies against F4/80⁺ (#GB113373-100, Servicebio, Wuhan, China), CD86 (#19589, CST, Danvers, USA) and CD206 (#ET1702-04, HuaBio, Hangzhou, China). Human sections were incubated with primary antibodies against CD3 (#YM8134, Immunoway, San Jose, USA), CD161 (#ab302564, Abcam, Waltham, USA), CD68 (#HA601290, Huabio, Hangzhou, China), VDR (#12550, CST, Danvers, USA), CD86 (#26903-1-AP, Proteintech, Wuhan, China), and CD206 (#ET1702-04, HuaBio, Hangzhou, China). Each was detected by HRP-conjugated secondary antibodies (#GB23303, Servicebio, Wuhan, China) and Opal fluorophores. Images were acquired on a confocal microscope (Olympus, Japan) and analyzed with ImageJ (V.1.45, New York, USA).

Immunohistochemistry

Paraffin-embedded tissue sections (4 µm) were incubated overnight at 4°C with primary antibodies, including anti-Ki67 (#P46013, Servicebio, Wuhan, China), anti-VDR (#12550, CST, Danvers, USA), anti-SOCS3 (#FNab08100, FineTest, Wuhan, China), or anti-CXCL16 (#ER1906-85, Huabio, Hangzhou, China). Detection used HRP-conjugated secondary antibody (#GB23303, Servicebio, Wuhan, China) and DAB chromogen (#G1212-200T, Servicebio, Wuhan, China), with hematoxylin counterstaining. Images were captured using a bright-field microscope (Nikon NI-U, Tokyo, Japan) and protein expression quantified via positive cells per field.

Statistical analysis

Unless otherwise specified, data in figures are presented as mean±SD. Two-group comparisons were conducted using an unpaired two-tailed Student’s t-test. For bilateral subcutaneous tumor data, a paired t-test was employed to compare significant differences between the two groups. Multigroup analyses employed one-way analysis of variance followed by Tukey’s post hoc test. A p value<0.05 was considered statistically significant. All statistical analyses were performed using GraphPad Prism V.8.0 software (GraphPad Software, USA).

Results

Incomplete RFA reshapes the bile acid landscape in residual hepatocellular carcinoma

To investigate the metabolic consequences of sublethal thermal ablation, we profiled the BA composition in both residual tumor tissue and serum from an orthotopic HCC mouse model following iRFA. Analysis of residual cancer tissues revealed a significant decrease in the total bile acid (TBA) level compared with pre-ablation samples (figure 1A). This was accompanied by a concerted reduction in the concentrations of multiple specific BAs (figure 1C–L), including deoxycholic acid (DCA), hyodeoxycholic acid, ω-muricholic acid, ursodeoxycholic acid (UDCA), cholic acid (CA), α-muricholic acid (αMCA), and β-muricholic acid. In stark contrast, the concentration and the composition of LCA, a highly hydrophobic secondary BA, were markedly elevated in the residual tumors (figure 1B, online supplemental figure S1).

Figure 1. Alterations in bile acid profile following iRFA in HCC. An orthotopic HCC model was generated and liver tumor was collected before or 7 days after iRFA (n=8). (A) Total bile acid level in HCC. (B–L) Individual bile acid level in HCC. (M) Ratio of primary to secondary bile acids in HCC. (N) Ratio of conjugated to unconjugated bile acids in HCC. Data are presented as Tukey’s box. P values represent *p<0.05 by Wilcoxon test. BAs, bile acids; CA, cholic acid; CDCA, chenodeoxycholic acid; DCA, deoxycholic acid; HCC, hepatocellular carcinoma; HDCA, hyodeoxycholic acid; iRFA, incomplete radiofrequency ablation; LCA, lithocholic acid; MCA, muricholic acid; UDCA, ursodeoxycholic acid.

In serum, however, the TBA level remained unchanged after incomplete RFA (online supplemental figure S2A). Despite this stable TBA pool, profound shifts in BA composition were observed. The concentration of αMCA was significantly decreased, while the level of LCA was increased (online supplemental figure S2B,H). Other BAs showed no significant alterations compared with baseline (online supplemental figure S2C–G,I,J). Further analysis of BA ratios revealed a significant decrease in both the primary-to-secondary BA ratio and the conjugated-to-unconjugated BA ratio in the serum of post-ablation patients (online supplemental figure S2K,L). BA compositional analysis revealed a marked increase in the proportion of LCA post-iRFA (online supplemental figure S2M,N). These data indicate that incomplete RFA triggers a systemic imbalance in BA metabolism, favoring the accumulation of hydrophobic BAs.

Reduced iNKT cell infiltration and M2-like macrophage polarization in residual HCC tissues following iRFA

An immunosuppressive TME is frequently observed in HCC after iRFA,^{3 19} characterized by diminished infiltration of CD4⁺ and CD8⁺ T cells. Beyond conventional T cells, NKT cells play pivotal roles in antitumor immunity.²⁰ To minimize interindividual variability, we first used a bilateral subcutaneous tumor model enabling paired pre-ablation and post-ablation sampling from the same mouse. Our data revealed a significant reduction in iNKT cell infiltration in residual tumors post-iRFA compared with pre-ablation levels (figure 2A,B). Consistent with prior report,²¹ macrophage infiltration increased in residual lesions after iRFA (figure 2C), with a pronounced accumulation of CD206⁺ M2-like macrophages (figure 2D,E). Conversely, infiltration of CD86⁺ M1-like macrophages was markedly decreased relative to baseline (figure 2F,G). Consequently, the M1/M2 macrophage ratio declined by approximately 50% post-iRFA (figure 2H). These findings were further validated in an orthotopic HCC model (figure 2I), where iRFA also induced a significant decrease in M1-like macrophages and an increase in M2-like macrophages in residual tumor tissues (figure 2J–L). Collectively, these findings indicate that attenuated iNKT cell recruitment and enhanced M2-like polarization may cooperatively foster an immunosuppressive TME in residual HCC after iRFA.

Figure 2. Reduced infiltration of iNKT cells and M2 macrophage polarization in residual HCC tissues following iRFA. A bilateral subcutaneous HCC-iRFA model was established. One tumor was excised for flow cytometry prior to ablation, while residual neoplastic tissue from the contralateral tumor was harvested for flow cytometric analysis 7 days after incomplete ablation (n=9–12 mice/group). (A) Representative flow cytometry plots of TCRβ+ α-GalCer-CD1d+iNKT cells. (B) Calculation of iNKT cells infiltration. (C) Calculation of infiltration of macrophages. (D) Representative flow cytometry plots of F4/80+CD206+ macrophages. (E) Calculation of CD206 on the surface of macrophages. (F) Representative flow cytometry plots of F4/80+CD86+ macrophages. (G) Calculation of CD86 on the surface of macrophages. (H) The ratio of M1/M2 macrophages. An orthotopic HCC model received iRFA was generated. (I) Representative H&E staining of liver tumor tissues from pre-iRFA and post-iRFA mice (original magnification: 10×). (J) Representative images of immunofluorescence staining of CD86⁺ and CD206⁺ macrophages in pre-iRFA (n=4) and post-iRFA (n=4) groups. (K) Calculation of CD86+macrophages in the tumor tissue. (L) Calculation of CD206+macrophages in the tumor tissue. Bar represents 20 µm. Data are presented as mean±SD. P values represent *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001 by pair t-test. HCC, hepatocellular carcinoma; iNKT cells, invariant natural killer T cells; iRFA, incomplete radiofrequency ablation; TCR, T-cell receptor.

LCA promotes tumor recurrence and suppresses iNKT cell infiltration post-iRFA in HCC

To investigate whether elevated LCA contributes to immunosuppressive TME formation and residual tumor progression, we turned to the orthotopic model, which preserves the liver-specific BA milieu. HCC-bearing mice were subjected to iRFA followed by LCA administration (20 mg/kg, every other day) for 7 days (figure 3A). BA profiling revealed increased LCA proportion in the LCA-treated group versus controls (figure 3B,C), with concomitant elevation of plasma LCA levels (figure 3E). Other BAs remained comparable between groups (online supplemental figure S3), while the conjugated/unconjugated BA ratio was significantly increased with LCA treatment (figure 3G). No differences were observed in total plasma BA concentrations or primary/secondary BA ratios (figure 3D,F). Critically, LCA administration accelerated residual tumor progression and significantly reduced survival in iRFA-treated HCC mice (figure 3H–J). Flow cytometric analysis demonstrated attenuated infiltration of both iNKT cells and total macrophages in residual tumors of LCA-treated mice (figure 3K–M). Although CD86⁺ M1-like macrophages were significantly diminished, the M1/M2 macrophage ratio showed no intergroup difference (figure 3N–P).

Figure 3. LCA promotes tumor recurrence and inhibits iNKT cell infiltration after iRFA in HCC. (A) Scheme showing animal experimental set-up. An orthotopic HCC model received iRFA was generated and treated with 20 mg/kg LCA or vehicles by oral gavage for 7 days. (B) Bile acid composition in plasma in LCA-treated mice and control groups (n=9–11). (C) LCA% in plasma bile acids. (D) Total bile acid levels in plasma. (E) LCA levels in plasma. (F) Ratio of primary to secondary bile acids. (G) Ratio of conjugated to unconjugated bile acids. (H) Final photography of orthotopic tumors from each group. (I) Tumor volume. (J) Survival curve. Specifically, at the end of the CTR curve, 0 out of 11 mice died, and at the end of the LCA curve, 4 out of 10 mice died. (K) Representative flow cytometry plots of TCRβ+ α-GalCer-CD1d+iNKT cells. (L) Calculation of iNKT cells infiltration in residual tumor tissues. (M) Infiltration of macrophages. (N) Calculation of CD86 on the surface of macrophages. (O) Calculation of CD206 on the surface of macrophages. (P) The ratio of M1 to M2 macrophages in residual tumor tissues. Data are presented as mean±SD. P values represent *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001 by Student’s t-test. BAs, bile acids; CA, cholic acid; CDCA, chenodeoxycholic acid; CTR, control group; DCA, deoxycholic acid; HCC, hepatocellular carcinoma; HDCA, hyodeoxycholic acid; iNKT cells, invariant natural killer T cells; iRFA, incomplete radiofrequency ablation; LCA, lithocholic acid; MCA, muricholic acid; TCR, T-cell receptor; UDCA, ursodeoxycholic acid.

LCA suppresses iNKT cell chemotaxis via macrophage VDR activation and CXCL16 downregulation

To elucidate the mechanism underlying LCA-mediated inhibition of iNKT cell chemotaxis, we investigated whether LCA regulates CXCL16, a chemokine primarily derived from macrophages.¹³ Given that chromatin immunoprecipitation sequencing (ChIP-seq) data (from Cistrome Data Browser) suggest VDR as a putative CXCL16 regulator and LCA is a potent VDR agonist, we generated TAMs and treated TAM and heat-treated Hepa1-6 cells with 20 µM LCA for 48 hours (online supplemental figure S5A). The induction purity of TAMs reached 95% (online supplemental figure S4A). LCA did not affect the viability of heat-treated Hepa1-6 cells (online supplemental figure S4B) nor alter Vdr expression in these cells, but significantly upregulated Vdr in TAMs (online supplemental figure S5B,C), indicating that LCA primarily targets macrophages rather than tumor cells directly. Concordantly, LCA suppressed Cxcl16 transcription (online supplemental figure S5C) and protein secretion (online supplemental figure S5D) in TAMs. Immunofluorescence (IF) further confirmed that LCA induced VDR activation and concurrently downregulated CXCL16 (online supplemental figure S5E,F,H). We next assessed the functional consequence of LCA-mediated CXCL16 suppression on iNKT cells. Co-culture of TAMs with DN32.D3 iNKT cells showed that LCA treatment significantly inhibited iNKT cell recruitment and IFN-γ secretion (online supplemental figure S5I,J). Importantly, exogenous CXCL16 supplementation rescued both iNKT chemotaxis and function in this co-culture system (online supplemental figure S5K,L), confirming that CXCL16 is the key effector molecule downstream of LCA.

To identify the molecular mechanism by which LCA-VDR signaling suppresses CXCL16, we focused on SOCS3, a known VDR-interacting corepressor.¹⁴ RNA-sequencing analysis (GSE212604)²² of patients with HCC following iRFA revealed that SOCS3 expression markedly increased post-iRFA, while other VDR corepressors (NCoR1, NCoR2, Hairless)^{23 24} showed no consistent change (online supplemental figure S6). Consistent with these clinical observations, LCA treatment elevated Socs3 levels in TAMs (online supplemental figure S5C). IF confirmed that LCA induced VDR activation, upregulated SOCS3, and reduced CXCL16 in TAMs (online supplemental figure S5D–H), suggesting a VDR-SOCS3-CXCL16 regulatory axis.

To establish causality, we performed loss-of-function experiments in TAMs. siRNA-mediated Vdr knockdown (>95% efficiency, figure 4A,B,G) abolished LCA-induced SOCS3 elevation (figure 4B,F,H) and restored CXCL16 suppression (figure 4C,F,I). Conversely, Socs3 knockdown (~80% efficiency, figure 4D,L) significantly increased CXCL16 expression and secretion (figure 4E,J,M) without affecting VDR levels (figure 4J,K). These results confirm that SOCS3 acts downstream of VDR to suppress CXCL16.

Figure 4. Knockdown of Vdr or Socs3 in TAM restores CXCL16 secretion under LCA treatment. (A) Knockdown level of si-VDR in Vdr mRNA expression. (B) Protein levels of VDR and SOCS3 after treatment of si-VDR and vehicles. (C) CXCL16 secretion by TAM after treated with LCA and si-VDR. (D) Knockdown level of si-SOCS3 in Socs3 mRNA expression. (E) CXCL16 secretion by TAM after treated with LCA and si-SOCS3. (F) Representative images of VDR/SOCS3/CXCL16 immunostaining in si-VDR intervention (n=9). (G–I) Quantification of fluorescence intensity for VDR, SOCS3, and CXCL16, respectively, in the si-VDR group. (J) Representative images of VDR/SOCS3/CXCL16 immunostaining in si-SOCS3 intervention (n=9). (K–M) Quantification of fluorescence intensity for VDR, SOCS3, and CXCL16, respectively, in the si-SOCS3 group. Scale bar, 20 µm. Data are presented as mean±SD. P values represent *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001 by Student’s t-test or one-way ANOVA followed with Tukey’s multiple comparisons test. ANOVA, analysis of variance; CXCL16, C-X-C motif chemokine ligand 16; DAPI, 4',6-diamidino-2-phenylindole; DMSO, dimethyl sulfoxide; LCA, lithocholic acid; mRNA, messenger RNA; SOCS3, suppressor of cytokine signaling 3; TAM, tumor associated macrophages; VDR, vitamin D receptor.

Finally, we validated this axis in vivo. In orthotopic HCC mice, residual tumors post-iRFA exhibited increased Vdr and Socs3 but decreased Cxcl16 messenger RNA (mRNA) compared with pre-ablation tissues (figure 5A–C). LCA administration amplified this signature (figure 5D–F) and reduced tumorous CXCL16 protein (figure 5H), without affecting plasma CXCL16 levels (figure 5G). Consistently, LCA-treated mice showed elevated Ki67+proliferating cells (figure 5I,J), and immunohistochemistry (IHC) confirmed VDR/SOCS3 induction and CXCL16 suppression in tumor tissues (figure 5I–M). Collectively, these data demonstrate that LCA inhibits iNKT cell chemotaxis through a VDR-dependent, SOCS3-mediated suppression of CXCL16 in macrophages.

Figure 5. LCA and TUDCA regulate the VDR-SOCS3-CXCL16 axis in mice. An orthotopic HCC model received iRFA was generated and treated with 20 mg/kg LCA for 7 days or 100 mg/kg TUDCA for 14 days. (A–C) Gene expression of Vdr/Socs3/Cxcl16 in tumor tissues before and 3 days after iRFA (n=6). (D–F) Gene expression of Vdr/Socs3/Cxcl16 in residual tumor tissues after treated with LCA or TUDCA (n=3–5). CXCL16 levels in plasma (G) and in tumor homogenate (H) under the treatment of LCA or TUDCA in HCC mice receiving iRFA. (I) Representative images of Ki67/VDR/SOCS3/CXCL16 staining in each group (n=14–15). Scale bar, 10 µm. (J) Quantification of Ki67 expression. (K) Quantification of VDR expression. (L) Quantification of SOCS3 expression. (M) Quantification of CXCL16 expression. Data are presented as mean±SD. P values represent *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001 by one-way ANOVA followed with Tukey’s multiple comparisons test. ANOVA, analysis of variance; CTR, control group; CXCL16, C-X-C motif chemokine ligand 16; HCC, hepatocellular carcinoma; iRFA, incomplete radiofrequency ablation; LCA, lithocholic acid; PBS, phosphate-buffered saline; SOCS3, suppressor of cytokine signaling 3; TAM, tumor associated macrophages; TUDCA, tauro-conjugated ursodeoxycholic acid; VDR, vitamin D receptor.

Reduced iNKT cell infiltration, elevated macrophage VDR expression, and M1-to-M2 polarization in patients with HCC post-iRFA

We next assessed iNKT cell infiltration and macrophage polarization in residual tumor tissues from patients with HCC post-iRFA compared with HCC tumors without RFA. mIF analysis revealed significantly attenuated iNKT cell density in residual tumors post-iRFA (figure 6A,B). Consistent with our experimental models, VDR expression in TAMs was markedly elevated in post-ablation residual tumors versus non-ablation lesions (figure 6A,C). Furthermore, analysis of macrophage polarization showed a significant decrease in M1-like macrophages (CD86⁺) and a concurrent increase in M2-like macrophages (CD206⁺) in post-iRFA residual tumors compared with non-ablated lesions (figure 6A,D,E). These clinical findings corroborate that iRFA induces immunosuppressive reprogramming, characterized by suppressed iNKT cell recruitment, enhanced TAM VDR signaling, and a shift toward M2-dominant macrophage polarization in human HCC.

Figure 6. Decreased infiltration of NKT cells and elevated VDR expression in macrophages in patients with HCC following iRFA. (A) Representative images of immunofluorescence staining of CD3+ and CD161+ NKT cells and VDR in CD68+macrophages in tumor tissue of patients with HCC (pre-iRFA group, n=5) or residual tumor tissue of HCC after iRFA (post-iRFA group, n=5). Representative images of CD86⁺ and CD206⁺ macrophages in pre-iRFA (n=3) and post-iRFA (n=3) tissues are also shown. (B) Calculation of infiltration of NKT cells in the tumor tissue. (C) Calculation of VDR expression in macrophages. (D) Calculation of CD86+macrophages in the tumor tissue. (E) Calculation of CD206+macrophages in the tumor tissue. Bar represents 20 µm. Data are presented as mean±SD. P values represent *p<0.05, **p<0.01 by Student’s t-test. DAPI, 4',6-diamidino-2-phenylindole; HCC, hepatocellular carcinoma; iRFA, incomplete radiofrequency ablation; NKT cells, natural killer T cells; VDR, vitamin D receptor.

We further characterize the relationship between VDR and CXCL16 in distinct macrophage subsets and found that M1 macrophages exhibited low Vdr but relatively high Cxcl16 expression, whereas M2 macrophages showed high Vdr and significantly reduced CXCL16 expression (online supplemental figure S7A,B). These findings align with our observation that iRFA shifts the balance toward M2 enrichment, potentially enabling LCA to act on Vdr-high M2 macrophages to suppress CXCL16.

Heat treatment of BMDM-derived TAMs recapitulated key iRFA features, including reduced CD86 and Tnfα, elevated CD206 and Il-10, along with increased Vdr and Socs3 expression and decreased Cxcl16 expression (online supplemental figure S7C). These results confirm that the thermal stress of iRFA directly reprograms macrophages toward an immunosuppressive and VDR-high phenotype.

TUDCA confers survival benefits in iRFA-treated orthotopic HCC mice

Given that TUDCA competitively antagonizes LCA toxicity,^{5 16 25} we administered TUDCA (100 mg/kg, every other day) for 14 days post-iRFA in orthotopic HCC mice (figure 7A) to explore whether the TUDCA supplement can improve the outcomes of HCC after iRFA. BA profiling revealed a substantial increase in UDCA proportion (figure 7B,C) and plasma concentration (figure 7E) in TUDCA-treated mice versus controls. Besides, total plasma BA level was strongly decreased (figure 7D) with a significant reduction of CAs, chenodeoxycholic acids, MCAs and DCAs (online supplemental figure S8). The ratio of primary to secondary BAs and the ratio of conjugated to unconjugated BAs were strongly decreased in TUDCA-treated group compared with control group (figure 7F,G). Notably, TUDCA suppressed residual tumor progression and prolonged overall survival (figure 7H–J). Under the supplement of TUDCA, there was enhanced iNKT cell infiltration in residual tumors in HCC-iRFA mice (figure 7K,L). Although macrophages infiltration was comparable between TUDCA-treated group and control group (figure 7M), CD86⁺ M1-like macrophages were significantly increased and the ratio of M1/M2 macrophages increased between two groups (figure 7N–P).

Figure 7. TUDCA induces survival benefit in HCC mouse model after iRFA. Scheme showing animal experimental set-up. An orthotopic HCC model received iRFA was generated and treated with 100 mg/kg TUDCA or vehicles by oral gavage for 14 days. (B) Bile acid composition in plasma in TUDCA-treated mice and control groups (n=12). (C) UDCA% in plasma bile acids. (D) Total bile acid levels in plasma. (E) UDCA levels in plasma. (F) Ratio of primary to secondary bile acids. (G) Ratio of conjugated to unconjugated bile acids. (H) Final photography of orthotopic tumors from each group. (I) Tumor volume. (J) Survival curve. Specifically, at the end of the PBS curve, 8 out of 11 mice died, and at the end of the TUDCA curve, 3 out of 11 mice died. (K) Representative flow cytometry plots of TCRβ+ α-GalCer-CD1d+iNKT cells. (L) Calculation of iNKT cells infiltration in residual tumor tissues. (M) Infiltration of macrophages. (N) Calculation of CD86 on the surface of macrophages. (O) Calculation of CD206 on the surface of macrophages. (P) The ratio of M1 to M2 macrophages in residual tumor tissues. Data are presented as mean±SD. P values represent *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001 by Student’s t-test. BAs, bile acids; CA, cholic acid; CDCA, chenodeoxycholic acid; DCA, deoxycholic acid; HCC, hepatocellular carcinoma; HDCA, hyodeoxycholic acid; iNKT cells, invariant natural killer T cells; iRFA, incomplete radiofrequency ablation; LCA, lithocholic acid; MCA, muricholic acid; PBS, phosphate-buffered saline; TCR, T-cell ratio; TUDCA, tauro-conjugated ursodeoxycholic acid; UDCA, ursodeoxycholic acid.

TUDCA promotes iNKT cell recruitment via cxcl16 secretion enhancement

Our integrated in vitro and in vivo analyses demonstrate that TUDCA antagonizes LCA-driven effects on the VDR-SOCS3-CXCL16 axis. TUDCA enhanced CXCL16 transcription and protein secretion in TAM (online supplemental figure S5C,D), and co-culture of TAMs with DN32.D3 confirmed that TUDCA increased the recruitment and function of iNKT cells (online supplemental figure S5I,J). Expression of Vdr and Socs3 were downregulated by TUDCA (online supplemental figure S5B,C). In line with M1-like polarization of macrophages, expression of CD86 was increased and CD206 tended to decrease (online supplemental figure S5C). IF analysis confirmed coordinate VDR/SOCS3 suppression and CXCL16 induction in TUDCA-treated TAMs (online supplemental figure S5E–H). Consistently, in iRFA-treated HCC mice, TUDCA suppressed the mRNA expression of Vdr and Socs3 in tumor tissue, and upregulated the mRNA expression of Cxcl16 (figure 5D–F). Furthermore, TUDCA elevated CXCL16 protein in tumor homogenates but not plasma (figure 5G,H). A significant reduction in proliferative activity, evidenced by reduced Ki67⁺ cells, was observed in the TUDCA-treated group (figure 5I,J). IHC analysis confirmed coordinate VDR/SOCS3 suppression and CXCL16 induction in TUDCA-treated tumors (figure 5I,K–M). These findings establish that TUDCA inhibits post-iRFA tumor progression through CXCL16-mediated iNKT cell recruitment via modulation of the VDR-SOCS3 pathway.

Discussion

Our study identifies a critical and previously unrecognized immunometabolic cascade that drives HCC recurrence after iTA. We demonstrate that the hydrophobic BA LCA, which accumulates post-ablation, activates the VDR in TAMs. This triggers the upregulation of SOCS3, leading to suppressed production of the key chemokine CXCL16, impaired recruitment of antitumor iNKT cells, and consequent fostering of an immunosuppressive microenvironment conducive to relapse. This VDR–SOCS3–CXCL16 axis not only elucidates a fundamental mechanism linking metabolic perturbation to immune escape but also unveils a readily actionable therapeutic target.

While elevated total BAs have been reported in patients with HCC,^{6 26} our work reveals a distinct, procedure-specific shift towards LCA dominance following iTA. This likely stems from a dual perturbation: (1) ablation-induced systemic inflammation and gut barrier dysfunction may promote dysbiosis, enriching for bacteria (eg, Clostridium species) that convert primary BAs to LCA^{27 28}; and (2) transient post-ablation hepatic dysfunction may impair the sulfation/glucuronidation required for LCA detoxification and excretion.²⁸ The resultant LCA accumulation establishes a potent immunosuppressive niche, correlating with worse survival. This context-dependent, tumor-promoting role contrasts with its reported ferroptosis-inducing effect in gallbladder cancer,²⁹ underscoring that LCA’s function is dictated by organ-specific metabolism and disease state—a crucial consideration for therapeutic targeting.

The nexus between LCA and immunosuppression is mediated through its high-affinity receptor, VDR, prominently expressed on macrophages.¹² Within the cancer context, VDR activation is subverted from its canonical immunoregulatory role^{30 31} to induce SOCS3, a master inhibitor of cytokine signaling.³² We extend prior knowledge by defining a precise downstream mechanism: SOCS3 mediates the transcriptional suppression of CXCL16 in TAMs. Given that CXCL16 is the principal chemokine for iNKT cell recruitment via CXCR6,¹³ its inhibition directly cripples a pivotal arm of innate antitumor immunity, which is essential for dendritic cell activation and CD8⁺ T-cell cytotoxicity.³³ Clinically, low intratumoral iNKT cell density has been shown to independently predict worse overall survival in patients with HCC, underscoring the prognostic relevance of this immune axis.⁹ Although iRFA itself enriches M2-like macrophages, LCA treatment did not further alter M2 proportion. Instead, LCA preferentially suppressed M1-like macrophages, resulting in an overall reduction of total macrophage infiltration without shifting the M1/M2 balance. This suggests that LCA’s primary effect is on macrophage gene expression—specifically CXCL16 suppression—rather than on polarization per se. Thus, the dual impairment of iNKT cell chemotaxis and reduction of M1-like macrophages, combined with the iRFA-induced M2-rich milieu, creates a permissive niche for residual HCC outgrowth.

The translational power of this discovery lies in its therapeutic reversibility. The clinically available BA modulator TUDCA exerted a multimodal protective effect without requiring a reduction in intratumoral LCA levels. It systemically reshaped the BA pool, reducing total levels and limiting LCA bioavailability. At the molecular level, TUDCA antagonized the VDR-SOCS3 node, restoring CXCL16 expression and enhancing iNKT cell infiltration by over 60%. Furthermore, it reprogrammed TAMs towards an M1-like, tumor-phagocytic phenotype.³⁴ This coordinated metabolic-immune reprogramming by TUDCA underscores the therapeutic potential of targeting BA signaling as an adjunct to ablation. While our focus was on the VDR-CXCL16 axis, concurrent FXR activation by TUDCA may provide complementary anti-inflammatory benefits, including downregulation of programmed cell death ligand 1 (PD-L1) on tumor cells and enhanced CD8⁺ T-cell infiltration,³⁵ potentially synergizing with immunotherapies like anti-programmed cell death protein 1 (PD-1). The localized restoration of CXCL16 within the tumor, without a corresponding increase in plasma, suggests a favorable safety profile by minimizing risks of systemic autoimmunity.³⁶

Notably, while the core mechanism—LCA activating macrophage VDR to inhibit iNKT infiltration—is conserved between mice and humans, fundamental differences in BA composition must be contextualized. Murine BA pools are dominated by MCAs with minimal endogenous LCA, whereas humans exhibit higher proportions of DCA and LCA due to distinct gut microbiota 7α-dehydroxylation activities.³⁷ This divergence may amplify LCA’s immunomodulatory impact in human post-iRFA microenvironments, wherein LCA is anticipated to reach levels capable of potently activating VDR. Moreover, translating these findings into clinical applications requires careful consideration of interspecies differences in BA dynamics. Humans demonstrate more efficient hepatic uptake of conjugated BAs via the sodium taurocholate cotransporting polypeptide (NTCP),^{38 39} which may enhance targeted delivery of agents such as TUDCA to HCC lesions. While ongoing trials of TUDCA combined with PD-1/PD-L1 immunotherapy in advanced HCC (NCT07064668) may provide direct clinical evidence for the efficacy of this metabolic-immunologic approach.

Our study has limitations that point to future directions. First, while the axis is robustly defined in mice, validating LCA accumulation and its correlation with VDR⁺TAMs and iNKT cell deficiency in human post-ablation tissues is crucial. Second, the precise molecular mechanisms linking VDR, SOCS3, and CXCL16 remain incompletely defined—specifically, whether SOCS3 suppresses CXCL16 directly or indirectly. Future studies using ChIP-qPCR or promoter analyses are needed to resolve this question. Third, we acknowledge that the M1/M2 dichotomy does not fully capture the functional complexity of TAMs. Future studies employing single-cell transcriptomic profiling to resolve TAM subsets will provide deeper insight into the heterogeneity of post-ablation TAM responses. Fourth, further investigation is needed to identify the upstream triggers of ablation-induced BA dysregulation, and to determine whether precise profiling of individual BA species (eg, LCA) in patient blood could serve as superior prognostic biomarkers compared with total BA measurements.

In conclusion, our findings demonstrate that incomplete RFA profoundly and detrimentally reprograms the local tumor microenvironment and systemic BA homeostasis. We identified a mechanism whereby elevated LCA levels post-ablation inhibited CXCL16-dependent NKT cell accumulation in the liver, thereby promoting residual HCC progression following iTA. TUDCA effectively counteracted LCA-mediated effects and conferred survival benefits in HCC-iRFA mouse model, suggesting its potential as an adjunctive therapy to improve outcomes in patients with HCC receiving thermal ablation.

Data availability statement

Data are available upon reasonable request.