The primary strategy for identifying HCC CSCs relies on the detection of specific cell surface markers that distinguish HCC CSCs from bulk tumor cells. A study has revealed a core set of surface markers (CD24, ICAM1, CD13, and EpCAM) that are consistently expressed on HCC CSCs³. Another study involving resected HCC specimens from 60 patients elucidated the intertumoral heterogeneity of liver CSCs (LCSCs) based on flow cytometry and identified a population of CD24/CD13/EpCAM triple-positive LCSCs in > 50% of the specimens⁴. Transcriptomic analysis on LCSC marker-sorted HCC cells and functional validation experiments pinpointed epidermal growth factor receptor kinase substrate 8-like protein 3 (EPS8L3) as a common functional regulatory gene associated with all three LCSC markers. Notably, more effective inhibition of sphere formation was observed following combined, rather than individual, knockdown of CD24, CD13, and EpCAM, suggesting the need to develop multi-targeted therapeutic approaches to overcome intrinsic CSC heterogeneity⁴ (Figure 1).

The advent of single-cell RNA sequencing (scRNA-seq) has provided unprecedented resolution in delineating the unique transcriptional landscape of HCC CSCs. Studies have demonstrated that CSCs exhibit marked upregulation of proliferative pathways (e.g., G2M checkpoint, MYC signaling, and E2F targets) and a metabolic shift towards glycolysis coupled with suppression of lipid metabolic pathways, including fatty acid, bile acid, and steroid metabolism⁵. In addition, advanced computational profiling (Dorothea) has revealed that CSCs are defined by upregulated activity of the following specific transcription factors: proliferative regulators (MYC, E2F1, E2F4, and E2F7); hypoxia-inducible factor (HIF-1α); and the redox-determining factor (NRF1), which is essential for mitochondrial homeostasis⁵. This transcriptional reprogramming represents a critical adaptive response to the multifaceted stresses of the advanced HCC tumor niche, including hypoxia, high proliferative demand, and concomitant metabolic and oxidative stress (Figure 1).

CSCs maintain an immunosuppressive niche by inducing M2 polarization and T cell exhaustion through the ICAM signaling pathway³. CSCs release chemokines (CCL3, CCL16, and CCL15) that target CCR1, which is expressed on SPP1⁺ macrophages, thereby promoting recruitment of SPP1⁺ macrophages to create an immunosuppressive tumor microenvironment (TME)⁵.

In addition to soluble factors, extracellular vesicles (EVs), especially exosomes and small EVs (sEVs), are key mediators of intercellular communication in HCC. Exosomes and sEVs contribute to the disease by maintaining CSC properties and modulating the TME through the transfer of various exosomal cargos to recipient cells. For example, a search for exosomal circRNAs identified circ-AFAP1 expression to be significantly higher in tumor tissues and positively correlated with CD133. Subsequent functional validation demonstrated the contributing role of circ-AFAP1 in tumor progression by stemness enhancement and epithelial-mesenchymal transition (EMT) promotion². Another study showed that CD90⁺ Huh7 cells released sEVs containing lncRNA H19, which upon internalization by recipient endothelial cells, promoted H19 overexpression and stimulated VEGF production, thus increasing pro-metastatic angiogenesis to permit metastasis of the CSCs⁶. In addition, RAB27A-mediated sEV secretion from liver CSCs upregulates NANOG to expand immunosuppressive Treg populations¹. sEVs from CD133⁺ CSCs carry the RNA demethylase, ALKBH5, which activates the SOX4/SHH signaling axis and contributes to additional immunosuppressive mechanisms in HCC¹. Together these findings suggest active modulation of the TME by the HCC CSCs.

In contrast, reprogrammed stromal components enhance CSC aggressiveness through multiple approaches. For example, SPP1⁺ macrophages, which are often spatially co-localized with CSCs in the tumor area, foster a pro-tumorigenic niche that enhances extracellular matrix remodeling, angiogenesis, and cell adhesion through HIF-1α signaling⁵. Furthermore, TGF-β1 secretion from tumor-associated macrophages (TAMs) at the edge of the tumor potently enhances core CSC properties, such as epithelial-mesenchymal transition, invasion, and tumorigenicity⁷. Moreover, CSC apoptosis can be regulated by TAM-secreted factors. Specifically, CXCL1 and CXCL2 from M2-polarized TAMs upregulate anti-apoptotic BCL-2, while suppressing pro-apoptotic BAD and BAX to confer significant sorafenib resistance⁸. Similarly, TAM-derived IL-6 activates pro-survival STAT3 signaling in CSCs, further promoting expansion and enhancing chemoresistance¹. Collectively, this reciprocal relationship progressively amplifies tumor malignancy and creates an immunosuppressive TME (Figure 2).

Clinically, elevated CSC scores and higher SPP1⁺ macrophage infiltration are consistently displayed in patients who are non-responsive to cabozantinib-nivolumab therapy⁵. The co-occurrence of the above two events in non-responsive patients underscores the critical necessity to develop co-targeting strategies that simultaneously address both CSCs and CSC conducive niche components to overcome therapeutic resistance. In fact, therapeutic strategies are evolving to co-target CSCs and the immune system. For example, bispecific antibodies, such as those targeting EpCAM on CSCs and CD3 on T cells, have shown promise in preclinical models, demonstrating potent immune-mediated cytotoxicity and tumor elimination⁹. Second, chimeric antigen receptor- (CAR-) T therapies targeting CSC markers have currently been the focus of the following clinical trials: a phase II trial of CAR-CD133 T cells (NCT02541370) in advanced HCC showed a manageable safety profile and encouraging preliminary efficacy [median overall survival (OS) = 12 months with hyperbilirubinemia the most common grade 3 adverse event]; and a separate phase I/II trial in the patient recruitment phase targeting EpCAM (NCT03013712) with outcome data and status currently unavailable⁹. Third, natural killer (NK) cell-based immunotherapies represent another promising strategy for selectively targeting CSCs in HCC. Multiple phase I/II clinical trials are currently exploring the efficacy of autologous or allogeneic NK cell infusions, either as monotherapy or in combination with other treatments (NCT03319459, NCT04162158, and NCT03592706), anti-MUC1 CAR-NK cells (NCT02839954)⁹. Notably, studies indicate that conventional therapies, such as chemotherapy or radiation, can have a dual effect. While conventional therapies have been shown to enrich the CSC population in various tumors, conventional therapies also upregulate stress-induced ligands, like MICA and MICB on the CSC surface, thus activating the NKG2S receptor on NK cells to promote NK cell recognition and cytotoxicity against CSCs⁹. This finding suggests that combining NK cell-based immunotherapies with conventional treatments, particularly when strategically timed, can synergize to eliminate CSCs more effectively and improve therapeutic outcomes in HCC.