Sublethal heat stress synergizes with the tumor microenvironment to drive recurrence of hepatocellular carcinoma after thermal ablation: mechanisms, molecular predictors, and targeted interventions

Abstract

Although thermal ablation has emerged as a minimally invasive and effective local treatment for hepatocellular carcinoma (HCC), its high postoperative recurrence rate remains a major clinical challenge. Sublethal heat stress can induce residual tumor cells to upregulate factors such as heat shock proteins (HSPs) and hypoxia-inducible factor-1α (HIF-1α), enhancing their survival tolerance. This process synergizes with components of the tumor microenvironment (TME), including myeloid-derived suppressor cells (MDSCs) and cancer-associated fibroblasts (CAFs), to collectively drive HCC recurrence. This article comprehensively reviews the research progress on the molecular mechanisms of tumor recurrence post-ablation, predictive biomarkers, and targeted therapeutic strategies. By deciphering multi-omics biomarkers, it provides new perspectives for predicting recurrence risk. Furthermore, this article also explores the potential of combination therapies, including targeting HSPs/HIF-1α, reversing immunosuppression, eliminating cancer stem cells (CSCs), and intervening in CAFs. This study provides a solid theoretical foundation for addressing the challenge of HCC recurrence, holding significant importance for improving patient prognosis and guiding clinical translation.

Graphical abstract

Mechanisms of HCC recurrence driven by sublethal heat stress and the tumor microenvironment after thermal ablation. Sublethal heat stress generated by thermal ablation simultaneously induces intrinsic adaptive changes in residual HCC cells and remodels the extrinsic tumor microenvironment. Within the cells, stress upregulates heat shock proteins (HSPs), stabilizes hypoxia-inducible factor-1α (HIF-1α), and activates post-translational modifications (e.g., SUMOylation), thereby initiating pro-survival and signaling reprogramming. Concomitantly, stressed cells release damage-associated molecular patterns (DAMPs) and extracellular vesicles (EVs). These mediators activate cancer-associated fibroblasts (CAFs) and recruit immunosuppressive myeloid-derived suppressor cells (MDSCs) and regulatory T cells (Tregs), thereby establishing an immunosuppressive niche. The remodeled microenvironment and the intrinsic adaptive processes interact continuously, collectively enriching a therapy-resistant cancer stem cell (CSC) population and activating the epithelial-mesenchymal transition (EMT) program, ultimately driving hepatocellular carcinoma (HCC) recurrence.

Introduction

Hepatocellular carcinoma (HCC) is the third leading cause of cancer-related mortality worldwide, with an annual incidence exceeding 900,000 cases (Bray et al. 2024). This high disease burden underscores the critical need for effective therapeutic strategies. For patients with early-stage HCC, thermal ablation is a recommended first-line curative option, particularly for solitary tumors ≤ 2 cm, with resection as an equivalent alternative. This aligns with the updated 2024 European Association for the Study of the Liver (EASL) Clinical Practice Guidelines, which refine patient selection based on tumor location and liver function(European Association for the Study of the Liver 2025). Primary modalities include radiofrequency ablation (RFA) and microwave ablation (MWA), which aim to eradicate tumor tissue by inducing coagulative necrosis through heating above 60 °C (Sasaki et al. 2005). RFA generates heat through high-frequency current-induced ionic agitation but is limited by uneven thermal distribution due to tissue charring and the heat sink effect, which increases the risk of local recurrence (Izzo et al. 2019; Sugimoto et al. 2025). In contrast, MWA heats tissue by exciting water molecules, creating more uniform and efficient ablation zones; however, its higher energy carries a risk of damaging adjacent critical structures such as bile ducts or intestines (Blair et al. 2025). Although other local ablation modalities such as cryoablation are also applicable for HCC treatment (Song 2016), this review focuses specifically on thermal ablation due to the more extensive and profound evidence base accumulated regarding its clinical application and the mechanisms underlying recurrence. Beyond its immediate tumorcidal effect, thermal ablation induces heterogeneous modes of hepatocellular carcinoma cell death, dictated by the spatial thermal gradient. At the ablation epicenter (> 60 °C), rapid protein denaturation and membrane disruption lead to coagulative necrosis, which predominates as the direct killing mechanism. In the adjacent sublethal zone (42–60 °C), however, cells face a fate determined by the balance of stress responses. While severe thermal stress can activate programmed death pathways such as apoptosis and pyroptosis, it concurrently triggers pro-survival and repair mechanisms. The ultimate fate of these cells—whether they undergo death or recover—depends on the dynamic equilibrium between these opposing signaling programs (Chen et al. 2025).

The core challenge of HCC thermal ablation therapy is the high postoperative recurrence rate (Kobayashi et al. 2011; Ni et al. 2013). A large multicenter study involving 1,055 HCC patients with nodules ≤ 3 cm reported a 5-year recurrence-free survival rate of only 42.7% after RFA (Kawaguchi et al. 2025). This high recurrence is closely associated with limitations in current assessment methods. Conventional imaging techniques often fail to accurately reflect true cellular viability within the ablation zone (Chiang et al. 2025). This underscores a key challenge highlighted in the 2024 EASL guidelines: even with standardized imaging criteria accurately identifying viable residual tumor at the ablation margin remains difficult(European Association for the Study of the Liver 2025). Supporting this limitation, Badar et al. reported that a significant proportion of lesions achieving complete radiological response still harbored viable tumor cells upon pathological examination (Badar et al. 2025). From a mechanistic perspective, this clinical phenomenon is linked to the formation of three characteristic biological zones post-ablation: a central zone of complete necrosis (> 60 °C), a peripheral sublethal transition zone (42–60 °C), and the surrounding normal tissue (Deng et al. 2017; Kostyrko et al. 2023). Critically, tumor cells within this 42–60 °C transition zone survive the corresponding sublethal heat stress, forming the cellular basis for recurrence (Kang et al. 2025). These residual cells enhance their survival and invasive capabilities through diverse mechanisms, including the upregulation of heat shock proteins (HSPs), activation of hypoxia-inducible factor-1α (HIF-1α) signaling, induction of epithelial-mesenchymal transition (EMT), acquisition of cancer stem cell-like properties, epigenetic reprogramming, and autophagy (Chen et al. 2024b; Xiao et al. 2025).

However, the molecular mechanisms driving recurrence, particularly under sublethal thermal stress, remain incompletely understood, and predictive models integrating molecular biomarkers are lacking. Furthermore, the distinct physical principles of RFA and MWA may result in technical variations in the sublethal zone extent, microvascular injury patterns, and DAMP release profiles. Such differences in initial damage signals could subsequently modulate the activation of core pathways (e.g., HSPs, HIF-1α) and tumor microenvironment remodeling, thereby influencing the drivers of recurrence. This review aims to elucidate these mechanisms, discuss predictive strategies based on multi-omics data, and summarize targeted therapeutic approaches against key pathways. By synthesizing current knowledge, we seek to provide a rational framework for improving clinical outcomes and guiding future research.

Key mechanisms of tumor recurrence mediated by ablation stress: cellular survival and microenvironment remodeling mechanisms

Survival and pro-recurrence mechanisms of residual tumor cells in the ablation zone

HSPs drive recurrence through a coordinated pro-survival program of proteostasis, death suppression, and cell cycle acceleration

HSPs are key orchestrators of the adaptive response in residual tumor cells surviving thermal ablation (Fig. 1). Their clinical relevance is underscored by transcriptomic analyses of The Cancer Genome Atlas and International Cancer Genome Consortium cohorts, which reveal significant upregulation of major HSPs such as HSP90, HSP70, and HSP27 in HCC tissues compared to normal hepatocytes (Shehata et al. 2020; Xiao et al. 2025). Within the sublethal temperature range of 42–60 °C, these HSPs are critically activated to maintain proteostasis via their chaperone functions (Iqbal et al. 2023; Mikhailova et al. 2025), facilitating the refolding or degradation of heat-damaged proteins and thereby shielding cells from heat-induced apoptosis (Wang et al. 2016).

Fig. 1.

HSP-driven mechanisms of HCC recurrence post-ablation. Following thermal ablation, residual hepatocellular carcinoma cells establish a protective mechanism through the high expression of heat shock proteins (HSPs). Among these, HSP90 plays a critical role by directly stabilizing oncoproteins such as c-Myc and sustaining the activation of the Akt signaling pathway, thereby driving aberrant cell proliferation and survival programs. HSP70 provides a survival advantage for residual cells by suppressing NOD-, LRR- and pyrin domain-containing protein 3 (NLRP3) inflammasome activity, effectively inhibiting Caspase-1 activation and subsequent pyroptosis. Furthermore, HSP27 efficiently scavenges reactive oxygen species (ROS) generated by thermal stress, which not only maintains genomic stability but also indirectly facilitates cell cycle progression by helping cells rapidly bypass the G1/S checkpoint

HSP90 emerges as a pivotal mediator of post-ablation stress and tumor recurrence (Sun et al. 2024). Supporting this, Yoshida et al. (2013) reported a significant increase in HSP90 expression in HCC cells exposed to 50 °C heat, a temperature within the sublethal range. HSP90 facilitates recurrence primarily by stabilizing a network of oncogenic client proteins, thereby promoting proliferation and suppressing apoptosis (Reynolds and Blagg 2024; Zhang et al. 2021). One key mechanism involves the stabilization of the cellular myelocytomatosis (c-Myc) oncoprotein. HSP90 binds to c-Myc, shielding it from ubiquitin-mediated degradation (Teng et al. 2004), which maintains high c-Myc levels that drive the expression of pro-proliferative genes (Zhou et al. 2019). This stabilization is further reinforced through HSP90's interaction with Bclaf1, creating a positive feedback loop that potently drives HCC progression and recurrence (Sun et al. 2024; Zhou et al. 2019). Concurrently, HSP90 directly binds to and conformationally stabilizes Akt kinase, maintaining its activity (Sumi and Ghosh 2022). Active Akt, in turn, stimulates downstream effectors like mTORC1, promoting biosynthetic processes such as protein and ribosome synthesis to fuel persistent proliferation (Cai et al. 2021). Through these coordinated networks, HSP90 enables heat-damaged HCC cells to survive, repair, and ultimately initiate clinical recurrence.

Beyond HSP90, HSP70 and HSP27 also function as critical drivers of recurrence via distinct mechanisms. For instance, in an orthotopic mouse xenograft model, Wang et al. (2021)confirmed that HSP70 expression increases after incomplete RFA (iRFA). The upregulated HSP70 inhibits the NLRP3 inflammasome, thereby reducing the cleavage of caspase-1 and the subsequent release of IL-1β/IL-18. This process suppresses pyroptosis, a form of inflammatory cell death, which undermines the efficacy of ablation and promotes tumor survival. Meanwhile, HSP27 promotes cell survival through a dual mechanism: it scavenges reactive oxygen species (ROS) to maintain genomic stability (Wang et al. 2022), and it facilitates the assembly and nuclear import of cyclin-dependent kinase (CDK)-cyclin complexes (Cushing et al. 2025). The latter action accelerates the G1/S phase transition, aiding cells in overcoming heat-induced cell cycle arrest (Parcellier et al. 2006). Collectively, the survival mechanisms mediated by HSPs constitute a critical foundation for tumor recurrence and metastasis post-ablation. Understanding these pathways unveils new avenues for developing rational combination therapies.

Activation of the HIF-1α signaling pathway

The sustained activation of the HIF-1α signaling pathway is a central driver of tumor recurrence following thermal ablation (Yu et al. 2024) (Fig. 2). Although thermal ablation effectively destroys tumor tissue, it concurrently damages the local microvasculature, causing vascular disruption and a precipitous drop in oxygen supply. This creates a profoundly hypoxic microenvironment that rapidly stabilizes HIF-1α protein levels (Hubbi 2024; Liu et al. 2025), enabling it to drive the progression of residual disease.

Fig. 2.

HIF-1α-mediated adaptive pathways in post-ablation HCC recurrence. The tissue damage and vascular disruption caused by thermal ablation establish a significantly hypoxic microenvironment within the tumor region. Hypoxia inhibits the activity of prolyl hydroxylase (PHD), thereby blocking the conventional von Hippel-Lindau protein (pVHL)-mediated ubiquitination degradation pathway for HIF-1α. This leads to the substantial accumulation and stabilization of HIF-1α protein within the cells. The stabilized HIF-1α subsequently initiates its downstream transcriptional program, upregulating various factors including VEGFA and Arginase-1, which effectively promotes angiogenesis and re-establishes microcirculation for the residual tumor. Concurrently, HIF-1α significantly enhances the expression of glucose transporters and hexokinase 2 to increase glucose uptake and utilization, while coordinately upregulating lactate dehydrogenase A (LDHA) and pyruvate dehydrogenase kinase 1 (PDK1). This shifts the dominant cellular energy metabolism pathway towards glycolysis, completing the metabolic reprogramming of the residual tumor cells. These adaptations collectively enhance the survival, proliferation, and stress resilience of the residual tumor cells

Mechanistically, the hypoxic microenvironment inhibits prolyl hydroxylase (PHD) activity, thereby blocking the von hippel-lindau protein (pVHL)-mediated ubiquitin–proteasome degradation pathway. This stabilization leads to the rapid accumulation of HIF-1α protein in residual tumor cells (Saber et al. 2024; Zhu et al. 2025b). HIF-1α then promotes HCC progression by upregulating key factors such as vascular endothelial growth factor A (VEGFA) and Arginase-1 (Alexander et al. 2020), which concurrently stimulate tumor angiogenesis and remodel the immunosuppressive microenvironment. Consistent with this, Halpern et al. (2021), using an HCC model of post-RFA recurrence, confirmed significant upregulation of HIF-1α, VEGFA, and Arginase-1 within the ablation zone compared to untreated HCC tissues. Stabilized HIF-1α then functions as a master transcriptional regulator, binding to hypoxia-response elements (HREs) to initiate the expression of a broad repertoire of downstream genes (Zhao et al. 2024a). This program enhances glucose uptake via upregulation of GLUT1 and HK2, while concurrently inducing key glycolytic enzymes like LDHA and PDK1, thereby shifting cellular energy metabolism toward glycolysis (the Warburg effect) to support survival under hypoxia (Liu et al. 2024a; Qiao et al. 2024). In summary, the hypoxic microenvironment induced by thermal ablation activates the HIF-1α signaling axis, which orchestrates a multi-faceted adaptive program—encompassing angiogenesis, immunosuppression, and metabolic reprogramming—to drive the survival and recurrence of residual HCC cells.

EMT drives post-ablation HCC recurrence

Local recurrence of HCC after thermal ablation frequently stems from occult residual cancer cells that evade current imaging detection (Weng et al. 2021). These residual cells drive invasive progression by activating EMT, a core cellular program that fuels aggressive recurrence and metastasis (Du et al. 2014; Fan et al. 2021). Supporting this, a comparative analysis of tumor samples from 10 patients with post-RFA recurrence and 78 patients undergoing initial RFA revealed significantly elevated mRNA expression of core EMT transcription factors—including TGF-β, Twist, and Snail-1—in the recurrence group (Iwahashi et al. 2016). This process is initiated by local tissue damage from ablation, which promotes the release of factors like TGF-β and IL-6. These cytokines, in turn, activate the Smad3/STAT3 signaling pathway, leading to the marked upregulation of key EMT regulators such as Twist1, Snail, and ZEB1 (Li et al. 2017; Mardiah et al. 2025; Zhou et al. 2025). These EMT regulators bind to E-box elements in the E-cadherin promoter, repressing the expression of this key epithelial adhesion molecule (Meng et al. 2015). Concurrently, they induce the expression of mesenchymal markers, including N-cadherin and Vimentin (Meng et al. 2015). This molecular switch results in the dissolution of cell–cell junctions and loss of polarity, thereby conferring enhanced migratory and invasive capabilities, as well as resistance to anoikis. Consequently, residual cells are empowered to form recurrent foci around the ablation zone or at distant sites (Ma et al. 2024).

Concomitant with EMT, tumor cells markedly increase the secretion of matrix metalloproteinases MMP-2 and MMP-9 (Dong et al. 2013), whose enzymatic activity can be enhanced 3 to fivefold compared to that in untreated cancer cells. These enzymes degrade key basement membrane components like type IV collagen and laminin, disrupting structural tissue barriers (Scheau et al. 2019). Simultaneously, the RhoA/ROCK pathway mediates actin cytoskeleton reorganization, driving the formation of invasive protrusions that confer cell motility. These mechanisms collectively promote local invasion and the dissemination of circulating tumor cells (Chen et al. 2018; Lawson and Ridley 2018). In summary, evidence indicates that thermal ablation can trigger EMT-related signaling pathways. This promotes a comprehensive invasive phenotype in residual cancer cells, characterized by enhanced motility and matrix degradation, which collectively drive the recurrence of HCC.

Acquisition and expansion of cancer stem cell properties

Cancer stem cells (CSCs) are a tumorigenic subpopulation within tumors capable of self-renewal and multi-lineage differentiation, which drives tumor recurrence, metastasis, and therapy resistance (Tsui et al. 2020; Zhang et al. 2023). Sublethal thermal stress (42–60 °C) from iRFA can induce a stem-like state in residual tumor cells (Balaji et al. 2024; Zhao et al. 2025b), leading to a significant expansion of the CSC pool within the residual lesion (Wang et al. 2017; Yamada et al. 2014). This population is considered a primary cellular reservoir for post-ablation HCC recurrence.

This CSC enrichment is an active process driven by iRFA-induced signaling and transcriptional reprogramming. IRFA promotes stemness through multiple axes. One axis involves the significant upregulation of the transcription factor SOX9, which drives CSC self-renewal and tumorigenicity (Yuan et al. 2018). Concurrently, iRFA activates the PKCα-ERK1/2 pathway, leading to Fra-1-mediated transcriptional upregulation of c-Myc. The elevated c-Myc expression further reinforces the stem cell state, establishing a positive feedback loop that robustly promotes tumor recurrence (Yuan et al. 2018; Zhang et al. 2019). In summary, iRFA orchestrates a stemness program in residual HCC cells by activating key transcription factors, including SOX9 and c-Myc. Therapeutic targeting of these core mechanisms represents a promising strategy for mitigating recurrence.

Rapid protein reprogramming mediated by post-translational modifications drives therapy resistance

Post-translational modifications (PTMs), the covalent addition of specific chemical groups to proteins after synthesis, enable rapid and precise control over protein activity, localization, and stability (Zhang et al. 2024b). Dysregulation of PTMs is an emerging area of research critically linked to HCC recurrence following thermal ablation. Utilizing in vivo RNA sequencing and in vitro models, Tang et al. demonstrated that RFA upregulates maternal embryonic leucine zipper kinase (MELK) in HCC. MELK, in turn, stabilizes fatty acid-binding protein 5 (FABP5) by inhibiting its ubiquitin-mediated degradation, a process that attenuates the anti-tumor immune response and facilitates immune evasion by residual cells (Tang et al. 2025). Furthermore, SUMO2-mediated SUMOylation is activated in post-ablation residual lesions. This PTM suppresses key immune signaling pathways and impairs immune cell function, thereby fostering an immunosuppressive microenvironment that supports the outgrowth of residual tumor cells (Liu et al. 2024b). Phosphorylation cascades rapidly activated by sublethal heat critically drive the aggressive transformation of residual cells. Evidence indicates that such stress induces epidermal growth factor receptor (EGFR) transactivation and downstream extracellular signal-regulated kinase (ERK) signaling, promoting proliferation—a process modulated by upstream Ca²⁺/calmodulin-dependent protein kinase II (CaMKII) (Dai et al. 2017). Additionally, p46-Shc phosphorylation potently activates the ERK1/2 pathway, inducing epithelial-mesenchymal transition and a stem-like phenotype, thereby heightening the invasive and tumorigenic potential of surviving HCC cells (Yoshida et al. 2013). Nevertheless, the exploration of PTMs in post-ablation HCC recurrence is still nascent, warranting further investigation to fully elucidate the underlying molecular mechanisms and therapeutic potential.

An integrated network of cellular adaptive mechanisms

The pro-survival mechanisms described above constitute an interconnected adaptive network. Rapid molecular responses, including HSP-mediated proteostasis and PTM-driven signaling rewiring (which directly modifies core signaling molecules to integrate recurrence signals at multiple levels), establish immediate cellular protection. This buffer permits the activation of sustained transcriptional programs, primarily through HIF-1α stabilization under hypoxia, a process regulated by oxygen-dependent changes in binding partners and post-translational modifications (Daly et al. 2021). Notably, these layers interact: HSP90 stabilizes clients such as c-Myc to promote CSC-like properties, while HIF-1α transcriptionally induces EMT regulators (e.g., Snail, Twist) (Li et al. 2025), whose stability and activity are, in turn, precisely controlled by phosphorylation and ubiquitination. The resulting EMT phenotype further reinforces the CSC state. This sequential reprogramming, often initiated by shared upstream signals such as HIF-1α, positions EMT as a critical driver for acquiring stemness. Concurrently, PTMs and HSPs (e.g., ROS-scavenging by HSP27) modulate the activity of these core pathways. Thus, molecular chaperone activity, master transcriptional regulation, rapid post-translational fine-tuning, and phenotypic plasticity converge into a unified, dynamic interactive network that drives recurrence.

Remodeling of the tumor microenvironment drives tumor survival and recurrence

Immune microenvironment reprogramming mediated by myeloid-derived suppressor cells

Thermal ablation therapy frequently induces a state of immunosuppression, a key driver of tumor recurrence. This process is mediated by the activation of immunosuppressive cells and signaling pathways that collectively dampen anti-tumor immunity (Ma et al. 2025; Yang et al. 2025) (Fig. 3). Following RFA, residual HCC cells produce chemokines like CCL2 and synergize with elevated immunosuppressive factors such as TGF-β to drive the recruitment of myeloid-derived suppressor cells (MDSCs) and regulatory T cells (Tregs) into the lesion, thereby establishing an immunosuppressive niche (Shi et al. 2019; Yang et al. 2025). Research by Tang et al. confirmed that a significant increase in MDSCs within HCC tissue post-ablation is a key factor mediating immunosuppression and recurrence (Tang et al. 2023). MDSCs suppress anti-tumor immunity through diverse mechanisms. First, they secrete inhibitory cytokines such as IL-10, which activates Tregs and upregulates programmed death-ligand 1(PD-L1) expression (Pal et al. 2019). Additionally, they directly impair T cell proliferation and cytotoxicity by depleting essential amino acids like arginine from the microenvironment (Yang et al. 2023). Concurrently, infiltrated Tregs inhibit T cell activation by engaging co-inhibitory receptors such as CTLA-4, which blocks essential co-stimulatory signals, thereby exacerbating T cell exhaustion and facilitating immune escape (Xue et al. 2025; Ying et al. 2025).

Fig. 3.

Ablation-induced immunosuppression and recurrence. Following thermal ablation, residual hepatocellular carcinoma cells release key factors such as CCL2 and TGF-β under thermal stress, effectively recruiting myeloid-derived suppressor cells (MDSCs) and regulatory T cells (Tregs) to the residual tumor site. Upon activation, MDSCs exert immunosuppressive effects through multiple pathways: they secrete IL-10 to promote the expansion and function of Tregs; they highly express programmed death-ligand 1 (PD-L1), which directly binds to programmed cell death protein 1 (PD-1) on T cells and inhibits their activity; and they release arginase 1, which depletes local arginine in the microenvironment, thereby disrupting T cell metabolism and impairing their proliferation and cytotoxic functions. Concurrently, Tregs competitively bind B7 molecules on antigen-presenting cells via CTLA-4, blocking the essential co-stimulatory signals required for T cell activation and further exacerbating T cell exhaustion. Additionally, the IL-6/JAK2/STAT3 signaling pathway, triggered by thermal ablation, remains persistently activated. This not only directly promotes the survival and proliferation of residual tumor cells but also sustains the immunosuppressive functions of MDSCs, establishing a positive feedback loop. These coordinated mechanisms collectively establish an immunosuppressive microenvironment that drives HCC recurrence and metastasis

Furthermore, sustained thermal stress activates the IL-6/STAT3 signaling axis, which directly promotes the expansion and functional maintenance of MDSCs (Meng et al. 2020; Wu et al. 2024). Studies have observed significant upregulation of IL-6 during RFA at sublethal temperatures compared to its levels following complete ablation (Markezana et al. 2021). IL-6 binding to its receptor triggers JAK2 kinase phosphorylation, thereby activating the transcription factor STAT3. The activated STAT3 translocates to the nucleus and regulates the expression of downstream target genes (Lokau et al. 2019), promoting the proliferation and survival of residual HCC cells. This multi-layered, interconnected immunosuppressive network results in persistent effector T cell dysfunction and excessive regulatory immune cell activity (Conway et al. 2022). Furthermore, ablation critically impairs anti-tumor immunity by driving pro-tumorigenic (M2-like) polarization of macrophages through damage-associated molecular patterns (DAMPs) and suppressing cytotoxic CD8⁺ T cell function via induction of exhaustion. These changes synergize with MDSC- and Treg-mediated suppression to establish an immunosuppressive microenvironment that fosters recurrence (Shi et al. 2019). Ultimately, this cripples immune surveillance and fosters local recurrence and distant metastasis (Fig. 4).

Fig. 4.

CAFs in post-ablation HCC recurrence. Thermal stress from ablation induces tumor cell necrosis, releasing damage-associated molecular patterns that provide critical activation signals for cancer-associated fibroblasts (CAFs). Among these, myofibroblast-like CAFs (myCAFs) secrete connective tissue growth factor to activate the Notch1/STAT3/Snail signaling axis, inducing epithelial-mesenchymal transition in tumor cells and enhancing their invasive and migratory capabilities. Periostin-positive CAFs promote EMT through the IL-6/STAT3-TG2 axis, while CD36⁺ and IL34⁺ CAF subpopulations mediate the recruitment of myeloid-derived suppressor cells and enrichment of regulatory T cells, respectively. These CAF subpopulations likely participate collectively in the post-ablation recurrence process, although their specific mechanistic roles require further elucidation

Activation of cancer-associated fibroblasts (CAFs)

In HCC, cancer-associated fibroblasts (CAFs), primarily derived from activated hepatic stellate cells (HSCs), promote tumor progression by remodeling the microenvironment (Cogliati et al. 2023) (Fig.4). The thermal stimulus from ablation induces necrotic cell death, leading to the substantial release of DAMPs that potently activate resident CAFs in the residual tumor (Ershaid et al. 2019; Rani et al. 2017). Recent single-cell studies have deepened the understanding of CAF heterogeneity, suggesting that ablation stress may differentially affect specific subsets (Cords et al. 2023; Zhang et al. 2024a). Consequently, activated CAFs drive post-ablation recurrence by directly inducing EMT in tumor cells and indirectly fostering an immunosuppressive microenvironment.

Among various CAF subtypes, myofibroblast-like CAFs (myCAFs) play a pivotal role. Studies using animal models confirmed a significant accumulation of activated myCAFs around the ablation zone post-RFA, identifying them as critical drivers of recurrence (Moussa et al. 2022). Emerging spatial transcriptomics and related technologies have begun to reveal that such pro-fibrotic subsets tend to be enriched at the tumor-stroma border, positioning them for direct interaction with residual cancer cells (Wang et al. 2023). Mechanistically, myCAF-derived extracellular vesicles transfer connective tissue growth factor (CTGF) to HCC cells. CTGF binds and activates the Notch1 receptor, triggering downstream signaling that upregulates the transcription factor Snail1 to induce EMT, thereby enhancing HCC cell proliferation, migration, and invasion (Zheng et al. 2025). Notably, Zheng et al. (2025) demonstrated that targeting the CTGF/Notch1 axis effectively blocks metastasis in preclinical models, highlighting its therapeutic potential. Other CAF subtypes also contribute to recurrence through distinct mechanisms. For instance, a periostin-positive (POSTN⁺) CAF subpopulation secretes IL-6 to activate the STAT3-TG2 axis, inducing EMT (Jia et al. 2020; Wang et al. 2024). Meanwhile, CD36⁺ and IL34⁺ CAF subpopulations help shape an immunosuppressive microenvironment by recruiting MDSCs and inducing cytotoxic T lymphocyte (CTL) exhaustion, respectively (Wang et al. 2023; Zhu et al. 2023). These CAF subtypes likely act in concert to promote post-ablation recurrence, although their precise mechanisms warrant further investigation. A comprehensive understanding of their precise regulatory networks in the post-ablation context awaits further elucidation.

Exosome-mediated intercellular communication

Residual HCC cells subjected to sublethal thermal damage during ablation exhibit significantly increased exosome secretion. These exosomes function as key intercellular messengers, delivering proteins and nucleic acids to shape the tumor microenvironment and drive multiple facets of recurrence (Zheng et al. 2024; Zhu et al. 2024). The team of Zhu et al. (2024) discovered that exosomes secreted by heat-treated HCC cells deliver circPTPRK, which upregulates PLA2G4E to activate angiogenesis in endothelial cells. This neovascularization provides a crucial material foundation and dissemination route for the proliferation, local recurrence, and distant metastasis of residual tumors. Furthermore, Ma et al. (2020) revealed that the long non-coding RNA ASMTL-AS1, transported by exosomes, functions as a competing endogenous RNA that sponges miR-342-3p. This sequestering action relieves the suppression of the target gene NLK, consequently activating the pro-oncogenic Yes-associated protein signaling pathway and ultimately driving the invasion and metastasis of residual tumors. Therefore, targeting exosomes and the pro-angiogenic and pro-invasive signals they mediate could represent a novel strategy for inhibiting recurrence after RFA.

Multi-omics biomarkers for predicting recurrence and prognosis after thermal ablation for HCC

Multi-omics analyses provide a powerful framework for predicting the risk of HCC recurrence after thermal ablation. By integrating genomic, transcriptomic, proteomic, and metabolomic data, these approaches can identify specific biomarkers that enable molecular-level risk stratification, thereby complementing the limitations of conventional imaging modalities (Table 1). It should be noted that the construction of a universally optimal, multi-omics biomarker panel for clinical use awaits validation through future large-scale, prospective multi-center studies.

Table 1.

Multi-omics biomarkers for predicting recurrence and prognosis after thermal ablation of hepatocellular carcinoma

Name	Type	Biospecimen Source	Expression	Mechanism/Function	References
TP53 gene mutation	Genomics	Plasma (ctDNA)	Detected in ctDNA	Poor prognosis indicator; high recurrence risk	(Chen et al. 2024a)
CTNNB1 gene mutation	Genomics	Plasma (ctDNA)	Detected in ctDNA	Indicates active tumor clones	(Chen et al. 2024a) (Sahu et al. 2025)
TERT promoter mutation	Genomics	Plasma (ctDNA)	Detected in ctDNA	Early driver event; predicts residual lesion risk	(Chen et al. 2024a) (Tang et al. 2024)
Methylated HIST1H3G	Genomics	Plasma (cfDNA)	Hypermethylated in plasma	Associated with tumor progression and malignancy	(Luo et al. 2022) (Zhu et al. 2025a)
ARF4 + EIF5B	Transcriptomics	HCC tumor tissue	Downregulated after insufficient RFA	Protective factors; knockdown promotes proliferation	(Zhang et al. 2025)
miR-122	Transcriptomics	Serum/Plasma	High expression in plasma	Associated with poor OS in HCC post-RFA	(Cho et al. 2015) (Colaianni et al. 2024)
Let-7c	Transcriptomics	Serum/Plasma	High expression predicts post-RFA relapse	Tumor suppressor; high expression indicates recurrence risk	(Canale et al. 2022)
miR-26a	Transcriptomics	Serum/Plasma	Low expression predicts poor prognosis	Tumor suppressor targeting oncogenic pathways	(Cho et al. 2017)
miR-29a	Transcriptomics	Serum/Plasma	Low expression predicts poor prognosis	Tumor suppressor inhibiting cell proliferation	(Cho et al. 2017)
circ-BANP	Transcriptomics	HCC samples	Higher in residual HCC	Promotes malignancy; inhibits progression on knockdown	(Li et al. 2022)
circPTPRK	Transcriptomics	Exosome	Significantly upreg in HCC exosomes	Promotes angiogenesis via PLA2G4E	(Zhu et al. 2024)
AFP	Proteomics	Serum	Post-RFA > 10 ng/mL denotes high risk	Predicts HCC recurrence post-RFA; low level = better prognosis	(Dohi et al. 2016)
AFP-L3	Proteomics	Serum	High expression in serum	Poor prognosis; high recurrence risk	(Zhao et al. 2025a)
Des-γ-carboxy Prothrombin	Proteomics	Serum	Serum marker; half-life < 48 h post-RFA	Correlates with RFA complete response	(Kobayashi et al. 2009) (Yao et al. 2023)
Albumin	Proteomics	Serum	Serum detection, cut-off < 3.5 g/dl	Independent risk factor for post-RFA OS	(Ma et al. 2014)
Glutamate + Aspartate	Metabolomics	Serum	Abnormal in HCV-related HCC pre-RFA	Predicts HCV-related HCC recurrence post-RFA	(Liu et al. 2018)
Glycerol	Metabolomics	Serum	Elevated in HCV-related HCC post-RFA	Predicts HCV-related HCC recurrence post-RFA	(Liu et al. 2018)
PC(30:2) + PC(30:1)	Metabolomics	Serum	Decreased in HCC (AUROC = 0.820)	Differentiates HCC-cirrhosis, evaluates post-RFA outcomes	(Nenu et al. 2022)
1,25-dihydroxy cholesterol	Metabolomics	Serum	Significantly elevated in HCC	Differentiates HCC-cirrhosis, evaluates post-RFA outcomes	(Nenu et al. 2022)
LysoPC(21:4) + LysoPE(22:2)	Metabolomics	Serum	Significantly elevated in HCC	Differentiates HCC-cirrhosis, evaluates post-RFA outcomes	(Nenu et al. 2022)

Genomics

Circulating tumor DNA (ctDNA) has emerged as a pivotal liquid biopsy biomarker for assessing recurrence risk after RFA in HCC. Mutations in key genes, including TP53, CTNNB1, and the TERT promoter, demonstrate significant predictive value (Chen et al. 2024a). Specifically, TP53 mutations are indicative of highly aggressive tumor behavior and warrant intensified postoperative surveillance. Monitoring the allelic frequency of TERT promoter mutations offers a dynamic means of evaluating ablation completeness, while CTNNB1 mutations may serve as potential markers for late, though not early, recurrence risk (Chen et al. 2024a; Sahu et al. 2025; Tang et al. 2024). The integration of these mutational markers enables a more accurate assessment of recurrence risk. Analysis of cell-free DNA (cfDNA) methylation provides another powerful approach. A methylation model developed using 120 HCC samples demonstrated high predictive accuracy (AUC = 0.98) (Luo et al. 2022), with hypermethylation of the HIST1H3G gene further identified as a key recurrent biomarker (Luo et al. 2022; Zhu et al. 2025a). An integrated liquid biopsy approach, which combines mutation and methylation markers, holds considerable promise for the early, non-invasive genomic assessment of recurrence risk following HCC ablation. To advance the clinical translation of liquid biopsy for post-ablation HCC recurrence risk assessment, future work must define optimal serial sampling timepoints and validate dynamic monitoring. Simultaneously, efforts should address cost control, assay standardization, and seamless integration with imaging surveillance.

Transcriptomics

Transcriptomic profiling of both coding and non-coding RNAs provides critical insights for predicting HCC recurrence after thermal ablation. Interrogation of multiple large-scale transcriptomic databases has led to the validation of a two-gene signature (ARF4 and EIF5B) that effectively identifies HCC patients at high risk of recurrence after RFA (Zhang et al. 2025). Circulating microRNAs (miRNAs) also show significant predictive value. For instance, low expression of miR-122 is an independent risk factor for poor prognosis (AUC = 0.86) (Cho et al. 2015; Colaianni et al. 2024). Furthermore, a profile characterized by high let-7c and low miR-26a/miR-29a expression is significantly associated with unfavorable outcomes, and combinations of these miRNAs demonstrate good sensitivity and specificity for identifying high-risk patients (Canale et al. 2022; Cho et al. 2017). Circular RNAs (circRNAs) represent another promising class of biomarkers. Sequencing of exosomes from ablated HCC cells revealed 229 differentially expressed circRNAs, with 211 upregulated compared to those from normal HCC cells (Zhu et al. 2024). Among these, circ-BANP (1.96-fold increase in residual lesions compared to normal liver tissue) promotes early recurrence by enhancing invasiveness (Li et al. 2022), while circ-PTPRK is a potential biomarker with a diagnostic sensitivity of 0.85 (Zhu et al. 2024). Future integration of these multi-class transcriptomic markers is expected to facilitate the construction of more accurate predictive models for recurrence.

Proteomics

Serum protein biomarkers provide a clinically accessible and non-invasive means to dynamically assess recurrence risk following HCC ablation (Jeon et al. 2023). Alpha-fetoprotein (AFP) remains the most widely used serum biomarker in this context. A retrospective study of 357 patients with early-stage HCC undergoing RFA demonstrated a significant positive correlation between post-ablation AFP levels and recurrence risk. Using a postoperative AFP level of < 5 ng/mL as a reference, patients with AFP levels exceeding 50 ng/mL faced a hazard ratio for recurrence of 3.13 (Dohi et al. 2016). However, the limited sensitivity and specificity of AFP constrain its predictive value when used alone (Canale et al. 2018). Among AFP variants, the lens culinaris-agglutinin-reactive fraction (AFP-L3) has been established as an independent predictor of poor postoperative prognosis (P < 0.001) and holds significant value for risk stratification (Zhao et al. 2025a). Beyond the AFP family, Des-gamma-carboxy prothrombin (PIVKA-II) demonstrates considerable predictive power. Elevated pre-ablation levels (≥ 100 AU/L) robustly predict poor survival and early recurrence, while a rapid decline following the procedure is strongly associated with a complete response and long-term disease-free survival (Kobayashi et al. 2009; Yao et al. 2023). Combining AFP with AFP-L3 and PIVKA-II effectively compensates for the limitations of any single marker and significantly improves the predictive accuracy for post-ablation recurrence (Canale et al. 2018).Furthermore, among traditional liver function parameters, a low serum albumin level (< 3.5 g/dL) has been independently associated with shorter overall and recurrence-free survival after RFA across multiple studies (Ma et al. 2014). In summary, these proteomic and clinical parameters collectively provide crucial evidence for guiding personalized post-ablation management strategies in HCC patients.

Metabolomics

Metabolomics provides a unique perspective for the non-invasive assessment of HCC recurrence risk after thermal ablation by tracking the dynamic changes of small-molecule metabolites. Its analytical scope encompasses metabolites such as amino acids, lipids, and carbohydrates (Esperança-Martins et al. 2021). Among pre-treatment serum biomarkers, elevated levels of glutamate and aspartate have been confirmed to correlate with poor prognosis. Utilizing random forest feature selection, researchers found that these two metabolites appeared in the predictive model with frequencies of 54.3% and 68.0%, respectively. When used in combination, they achieved a prediction accuracy of 83%, demonstrating considerable potential for clinical application (Liu et al. 2018).In post-ablation monitoring, glycerol—a key indicator of lipid metabolism—shows a significant correlation between its increased levels and elevated recurrence risk (P < 0.001). Concurrently, low expression levels of phosphatidylcholines (PC) PC (30:2) and PC (30:1) have also been identified as important prognostic indicators (Liu et al. 2018; Nenu et al. 2022). Furthermore, metabolites including 1,25-dihydroxycholesterol, lysophosphatidylethanolamine (LysoPE) LysoPE(22:2) and lysophosphatidylcholine (LysoPC) LysoPC(21:4) were identified as independent prognostic factors for survival in multivariate Cox regression analysis (Nenu et al. 2022). While these findings highlight promising metabolic markers, their precise role in predicting post-ablation recurrence necessitates further investigation. Predictive models constructed based on metabolomic biomarkers provide valuable complementary information for assessing the risk of recurrence in HCC patients following thermal ablation.

Precision therapeutic strategies targeting molecular mechanisms of recurrence after incomplete ablation

To address the challenge of HCC recurrence after thermal ablation, multifaceted therapeutic strategies are being developed. These span from established clinical regimens to novel agents targeting key recurrence mechanisms—including HSP upregulation, HIF pathway activation, CSCs, and the immunosuppressive microenvironment—currently under preclinical and clinical investigation.

In targeting HSPs, Pimitespib (TAS-116) is the only approved selective HSP90 inhibitor, while second-generation agents like Ganetespib are in clinical trials (Goel et al. 2025). Preclinical studies are exploring combinations of HSP90 inhibitors with autophagy inhibitors (e.g., 3-MA) or novel subtype-selective inhibitors (e.g., targeting GRP94), with the goal of integrating them into more potent combination regimens (Goel et al. 2025; Juthani et al. 2025; Xiao et al. 2023; Zhao et al. 2024b). Directly targeting the HIF pathway remains a promising but unrealized goal. Clinically, multi-kinase inhibitors (e.g., Sorafenib, Lenvatinib) or agents like Arsenic Trioxide are used to indirectly affect HIF signaling (Juthani et al. 2025). Meanwhile, potent direct inhibitors such as PT2385/PT2977 and the dual inhibitor 32-134D have shown efficacy in inhibiting tumor progression in preclinical studies (Juthani et al. 2025; Salman et al. 2022). Targeting CSCs and their associated immunosuppressive microenvironment represents another key frontier. Chimeric antigen receptor T cell (CAR-T) therapies and bispecific antibodies (e.g., Catumaxomab) directed against CSC surface markers have shown early promise (Dai et al. 2021). Within the microenvironment, targeting MDSCs is critical. Approved multi-kinase inhibitors like Cabozantinib and Sunitinib reduce MDSCs, and combinations of STAT3 inhibitors (e.g., Danvatirsen) with PD-L1 blockade have shown positive results in other cancers, providing a rationale for HCC (Lu et al. 2019). It is particularly worth emphasizing that this combination strategy, which leverages ablation-induced antigen release to potentiate systemic immunotherapy, represents one of the most promising clinical frontiers for improving long-term outcomes after thermal ablation. Leveraging the "in situ vaccine" effect of ablation, immune checkpoint inhibitors are being explored as consolidation therapy to prevent recurrence. Clinical trials evaluating combinations of ablation with Pembrolizumab (KEYNOTE-937) or Nivolumab (CheckMate 9DX) are underway (Chakraborty and Sarkar 2022; Leone et al. 2021). Additionally, early preclinical and clinical evidence supports the safety and potential synergy of combining HSP90 inhibitors (e.g., Pimitespib) with immune checkpoint inhibitors, though its specific efficacy and optimal dosing in the context of post-ablation HCC recurrence require further validation (Kawazoe et al. 2021). Thus, combining immunotherapy with novel agents targeting CSCs, HSP/HIF, and other core recurrence pathways presents a promising strategic direction for overcoming the challenge of post-ablation recurrence.

Conclusion

HCC recurrence following thermal ablation is orchestrated by a dynamic interplay between residual tumor cells and the remodeled tumor microenvironment. Sublethal heat stress triggers the concurrent activation of key pro-survival pathways—including HSP upregulation, HIF-1α signaling, and EMT—which collectively drive the acquisition of aggressive traits in residual cells. These processes are further enabled by an immunosuppressive microenvironment rich in MDSCs and Tregs. The integration of multi-omics technologies—spanning genomic, transcriptomic, proteomic, and metabolomic markers—facilitates the construction of predictive models that surpass the limitations of conventional imaging for early recurrence detection. However, persistent challenges hinder clinical outcomes. These include technical obstacles to achieving complete ablation (e.g., the heat-sink effect, tumor location), the limited sensitivity of conventional imaging for detecting minimal residual disease, the heterogeneity in study designs and patient populations (including diverse etiologies such as viral hepatitis and metabolic dysfunction) which complicates the identification of unified mechanisms and biomarkers, and an incomplete understanding of the interconnected molecular networks that facilitate tumor survival and immune evasion. Future efforts to fully delineate these molecular mechanisms will unveil novel therapeutic vulnerabilities. Integrating multi-targeted strategies holds the potential to transform thermal ablation from a localized procedure into a systemic therapy capable of eliciting durable, systemic antitumor immunity. Ultimately, such advances promise to catalyze a paradigm shift in HCC management—from reactive treatment of recurrence to its proactive and precise prevention. Guided by this paradigm, clinical translation may focus on three actionable strategies: targeting HSPs/HIF-1α to eliminate residual cells, applying multi-omics biomarkers for risk stratification, and combining ablation with immunotherapy to harness its in situ vaccine effect while reversing immunosuppression; optimizing the administration timing and regimen of such combinations represents a key translational challenge; and developing innovative regimens that co-target complementary microenvironmental components (e.g., CAFs and MDSCs) to achieve more comprehensive reversal of the pro-recurrence niche. Together, these can transform thermal ablation into a strategy for sustained systemic control and improved long-term outcomes in HCC.

Abbreviations

HCC

Hepatocellular carcinoma

EASL

European Association for the Study of the Liver

RFA

Radiofrequency ablation

MWA

Microwave ablation

HSPs

Heat shock proteins

HIF-1α

Hypoxia-inducible factor-1α

EMT

Epithelial-mesenchymal transition

iRFA

Incomplete radiofrequency ablation

ROS

Reactive oxygen species

CDK

Cyclin-dependent kinase

VEGFA

Vascular Endothelial Growth Factor A

HREs

Hypoxia-response elements

CSCs

Cancer stem cells

PTMs

Post-translational modifications

MDSCs

Myeloid-derived suppressor cells

Tregs

Regulatory T cells

CAFs

Cancer-associated fibroblasts

HSCs

Hepatic stellate cells

DAMPs

Damage-associated molecular patterns

myCAFs

Myofibroblast-like CAFs

CTGF

Connective Tissue Growth Factor

CTL

Cytotoxic T lymphocyte

ctDNA

Circulating tumor DNA

cfDNA

Cell-free DNA

miRNAs

MicroRNAs

circRNAs

Circular RNAs

AFP

Alpha-fetoprotein

AFP-L3

Lens culinaris-agglutinin-reactive fraction

PIVKA-II

Des-gamma-carboxy prothrombin

PC

Phosphatidylcholines

LysoPE

Lysophosphatidylethanolamine

LysoPC

Lysophosphatidylcholine

Authors contribution

Boran Li: Writing – original draft. Liou Zhang and Xiaoxi Bai: Conceptualization, Writing – review & editing.

Funding

No fund was required for this review.

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethics approval and consent to participate

Not applicable.

Clinical trial number

Not applicable.

Consent for publication

All authors agreed to the publication of the article in the journal.

Competing interests

The authors declare no competing interests.