A review of organoid–immune co-culture platforms to model the immune microenvironment of hepatocellular carcinoma and guide immunotherapy

Abstract

Background

Hepatocellular carcinoma (HCC) is characterized by a highly immunosuppressive and heterogeneous tumor microenvironment that limits the effectiveness of current immunotherapies. Conventional two-dimensional cultures and animal models fail to fully capture patient-specific tumor–immune interactions, creating an urgent need for more physiologically relevant platforms.

Main body

This review summarizes recent advances in co-culture systems integrating patient-derived HCC organoids with defined immune cell populations to reconstruct essential features of the tumor microenvironment. We describe strategies for organoid establishment and validation, outline immune cell integration approaches, and compare static three-dimensional cultures, microfluidic organ-on-chip systems, and bioengineered multicellular platforms. We further highlight key tumor–immune interaction mechanisms that have been functionally interrogated in these systems, including immune checkpoint–mediated T-cell dysfunction, adenosine-driven metabolic suppression, and chemokine-regulated immune recruitment. Importantly, we critically evaluate current limitations, including immune cell exhaustion artifacts, lack of stromal and vascular complexity, and variability across protocols, which may affect the reproducibility and translational interpretation of these models. While emerging studies suggest potential for predicting immunotherapy responses, robust clinical validation in HCC remains limited.

Conclusions

Organoid–immune co-culture platforms represent an emerging translational framework that bridges mechanistic tumor immunology with functional precision oncology. With improved standardization and integration of multicellular bioengineering and multi-omics technologies, these systems have strong potential to guide personalized immunotherapy strategies, although further clinical validation is required.

Introduction

Hepatocellular carcinoma (HCC) accounts for nearly 90% of primary liver cancers and represents a major global health burden with increasing incidence and mortality [1]. Current therapeutic strategies—including surgical resection, locoregional therapies, systemic treatments, and immunotherapy—have improved outcomes in selected patients but remain insufficient for advanced disease [1, 2]. The tumor microenvironment (TME) plays a central role in HCC progression and treatment response [3]. Its complex cellular and molecular components interact to shape an immunosuppressive milieu that promotes immune evasion and treatment resistance [4]. Although immune checkpoint inhibitors and cellular immunotherapies have been introduced, their efficacy in HCC remains limited, partly due to intrinsic and acquired resistance mechanisms [5]. These include the expansion of immunosuppressive cell populations, such as regulatory T cells (Tregs) and myeloid-derived suppressor cells (MDSCs), which inhibit cytotoxic T-cell and natural killer (NK) cell function [6, 7].

Conventional two-dimensional (2D) cultures and animal models do not fully recapitulate the spatial architecture, cellular heterogeneity, and signaling complexity of the human TME [8]. Organoid technology has emerged as a promising alternative platform [9, 10]. Organoids are three-dimensional (3D), self-organizing structures derived from tissue-resident stem or progenitor cells, embryonic stem cells (ESCs), or induced pluripotent stem cells (iPSCs). They preserve key genetic, phenotypic, and functional characteristics of their tissue of origin, thereby enabling physiologically relevant and patient-specific tumor modeling [9, 11].

HCC organoids retain key genomic, transcriptomic, and histopathological features of the original tumor, thereby preserving tumor heterogeneity and enabling more representative disease modeling and drug screening [12, 13]. The integration of immune cells into these systems has led to the development of organoid–immune co-culture models, which allow dynamic and functionally controlled investigation of tumor–immune interactions in vitro [14]. These platforms provide a controllable and physiologically relevant setting to dissect mechanisms of immune evasion, assess immunotherapeutic responses, and explore potential predictive biomarkers, although their ability to recapitulate clinical responses remains under active investigation [14, 15]. While no in vitro system can fully reproduce the complexity of the in vivo tumor microenvironment, such co-culture models represent a significant advance in bridging experimental tumor immunology with translational oncology [14, 16]. This review summarizes current strategies for establishing HCC organoid–immune co-culture systems, highlights their applications in mechanistic and therapeutic studies, and critically discusses existing challenges and future directions for clinical translation.

A narrative literature search was conducted using PubMed, Web of Science, and Scopus databases up to January 2026. The search strategy included combinations of keywords such as “hepatocellular carcinoma”, “organoids”, “immune co-culture”, “tumor microenvironment”, and “immunotherapy”. Relevant original studies and review articles were selected based on their relevance to HCC organoid–immune co-culture systems, methodological rigor, and translational significance. The identified literature was then critically evaluated and synthesized to provide an up-to-date and comprehensive overview of the field.

Establishment of hepatocellular carcinoma organoids

Culture of HCC organoids

HCC organoids are most commonly established as patient-derived organoids (PDOs) from resected or biopsy specimens, although alternative sources such as cell lines or stem cell–derived systems have also been explored [12, 13]. Standard protocols typically involve enzymatic digestion of fresh tumor tissues, followed by embedding of dissociated cells within a three-dimensional extracellular matrix (ECM), such as Matrigel, and subsequent culture in growth factor–defined media [17]. While these approaches enable efficient organoid formation, variability in ECM composition and culture conditions may influence organoid morphology, differentiation status, and reproducibility across studies [16].

Media optimization is a critical determinant of organoid stability and expansion. The use of defined components, such as dexamethasone in place of conditioned media, has been shown to improve tumor purity and support long-term expansion [12]. However, such selective culture conditions may also introduce clonal selection bias, potentially limiting the representation of tumor heterogeneity [16].

Alternatively, HCC organoids can be generated from iPSCs through directed differentiation into hepatocyte-like cells, providing a renewable and scalable source for organoid production [18]. These two approaches offer complementary strengths in modeling HCC biology. PDOs preserve the genomic, phenotypic, and histopathological heterogeneity of the original tumor, making them particularly well-suited for personalized drug screening and investigation of native tumor–immune interactions, although their establishment efficiency remains variable and may introduce selection bias [12]. In contrast, iPSC-derived organoids provide long-term expansion capacity, genetic tractability (e.g., via CRISPR–Cas9), and more standardized culture conditions, which are advantageous for high-throughput mechanistic studies [19]. Therefore, PDOs are more directly applicable to translational precision oncology, whereas stem cell–derived models serve as powerful platforms for basic research and toxicological assessment, each with distinct advantages and limitations [12, 16].

Validation of organoid models

Comprehensive validation of organoid models requires a rigorous and integrative multi-tiered approach [20]. Histological validation evaluates architectural fidelity using H&E staining and immunohistochemistry (e.g., AFP, EpCAM) to assess cellular identity relative to the parent tumor tissue [21]. Molecular validation employs sequencing approaches (WES, RNA-seq) to determine whether key driver mutations (e.g., TP53, CTNNB1) and oncogenic transcriptional programs are preserved [21, 22]. Functional validation examines malignant phenotypes, including proliferation, invasion, and drug responsiveness through dose–response and viability assays [12]. In vivo validation, typically via xenotransplantation into immunodeficient mice, is widely used to assess tumorigenic potential [13]. Collectively, this integrative validation framework supports the biological fidelity of organoid models and provides a robust foundation for subsequent immune co-culture studies, although variability across models and assays remains an important consideration [17, 23]. To provide an integrated overview of organoid generation, model comparison, and multi-level validation strategies, these key processes are summarized in Fig. 1.

Fig. 1.

Generation, comparison, and validation of HCC organoid models for immune co-culture studies. Fresh tumor tissues obtained from surgical resection or biopsy are dissociated and embedded in extracellular matrix for three-dimensional culture in defined medium. Patient-derived organoids (PDOs) preserve tumor heterogeneity and are suitable for immune co-culture studies, whereas iPSC-derived organoids offer scalability and genetic manipulability but may lack native tumor context. Organoid fidelity is validated through histological, molecular, functional, and in vivo approaches, supporting their application in downstream immune co-culture studies

Selection and integration of immune cells in HCC organoid co-culture systems

Key immune populations in the HCC microenvironment

Within the HCC TME, immune cells play a central role in shaping tumor progression and therapeutic response [4]. These populations can be broadly categorized based on their functional roles, with distinct implications for organoid co-culture modeling. Innate immune cells, including tumor-associated macrophages (TAMs) and neutrophils (TANs), exhibit substantial functional plasticity [24]. TAMs can polarize into pro-inflammatory (M1) or immunosuppressive (M2) phenotypes, with M2-like TAMs promoting immune evasion through IL-10 and TGF-β secretion [24, 25]. Similarly, TANs may adopt tumoricidal (N1) or pro-tumorigenic (N2) states [26].

Immunosuppressive adaptive populations are primarily MDSCs and regulatory T cells (Tregs) [27]. MDSCs inhibit T-cell activity through metabolic disruption and reactive oxygen species (ROS) production [6], while Tregs suppress effector responses via CTLA-4 signaling and inhibitory cytokines [28]. In contrast, effector populations—including cytotoxic T lymphocytes (CTLs), natural killer (NK) cells, and dendritic cells (DCs)—mediate anti-tumor immunity through cytotoxic activity and antigen presentation [29].

From a co-culture modeling perspective, different immune subsets offer distinct advantages and limitations. TAMs are frequently incorporated due to their abundance in HCC and their central role in immunosuppression and cytokine-mediated crosstalk [30]. CTLs and NK cells are commonly used to model direct cytotoxic responses, although their function may be influenced by in vitro exhaustion and lack of sustained antigen stimulation [31]. DCs provide opportunities to study antigen presentation and T-cell priming but are less commonly included due to technical complexity [32]. Therefore, the selection of immune cell populations in organoid co-culture systems should be guided by the specific experimental objective, balancing biological relevance with technical feasibility [23].

Sources of immune cells for co-culture

The source of immune cells is a critical determinant of a co-culture model’s physiological relevance [14]. Murine-derived systems utilize immune cells from syngeneic mice, where DCs are differentiated from bone marrow, CTLs expanded from splenocytes, and macrophages generated from monocytes using cytokine cocktails [33]. These models are valuable for controlled preclinical studies of immune dynamics. For human translational studies, immune cells are typically sourced from donors [14]. Autologous peripheral blood mononuclear cells (PBMCs) or tumor-infiltrating lymphocytes (TILs) from matched patients offer the highest clinical fidelity [23, 34]. When autologous samples are unavailable, allogeneic PBMCs from healthy donors or immortalized cell lines (e.g., THP-1 monocytes) provide a reproducible, albeit less antigen-specific, alternative [35]. These systems improve experimental consistency but lack tumor-specific antigen recognition and may not fully recapitulate patient-specific immune interactions [36, 37].

Representative HCC studies demonstrate diverse co-culture strategies [37]. Peripheral blood–derived monocytes or PBMCs have been used to model macrophage polarization, while co-culture with autologous or donor-derived T cells enables assessment of tumor-reactive cytotoxicity. Immortalized monocytic cell lines such as THP-1 are commonly employed to investigate macrophage–tumor interactions under controlled conditions [17, 35].

Overall, the choice of immune cell source—autologous, allogeneic, or cell line–derived—should be guided by the specific experimental objective, balancing clinical relevance, antigen specificity, and experimental reproducibility [14].

Co-culture techniques: configurations and trade-offs

Co-culture systems can be broadly categorized along two principal dimensions: the mode of cellular interaction (direct vs. indirect) and culture dynamics (static vs. dynamic) [38]. Direct co-culture involves the physical integration of immune cells and organoids within the same microenvironment, enabling cell–cell contact, immune infiltration, and direct cytotoxic activity. This configuration is well suited for investigating immune-mediated killing mechanisms, but may also result in confounding effects such as immune cell overactivation, rapid exhaustion, or disproportionate cell expansion [23, 38].

In contrast, indirect co-culture systems typically employ permeable membrane inserts (e.g., Transwell) to physically separate cell populations while allowing the exchange of soluble factors. This setup facilitates the study of paracrine signaling, including cytokine- and chemokine-mediated communication, without direct cell–cell contact [17]. It is particularly useful for dissecting immunomodulatory interactions involving macrophages or DCs [38]. These major co-culture configurations are schematically illustrated in Fig. 2.

Fig. 2.

Organoid–immune co-culture platforms and their key features. (A) Direct co-culture enables physical contact between organoids and immune cells, allowing the study of cell–cell interactions and cytotoxic responses. (B) Indirect co-culture separates cell populations while permitting exchange of soluble factors, facilitating the investigation of paracrine signaling. (C) Microfluidic systems introduce controlled fluid flow and gradients, enabling analysis of immune cell migration and spatial interactions. (D) Bioreactor systems provide dynamic perfusion and controlled environmental conditions, supporting long-term co-culture. Panels A and B represent static systems, whereas panels C and D depict dynamic systems

Regarding culture dynamics, static systems (e.g., multiwell plates with Matrigel) are relatively simple and cost-efficient, and they preserve organoid architecture, making them well suited for initial screening applications [16]. However, these systems are limited by diffusion constraints and suboptimal long-term immune cell viability [39].

In contrast, dynamic systems—such as microfluidic organ-on-chip devices or bioreactors—represent more advanced platforms that introduce perfusable fluid flow, shear stress, and spatial nutrient gradients [40, 41]. These features allow such systems to more closely recapitulate aspects of physiological microenvironments, enable prolonged co-culture, and facilitate real-time observation of immune cell migration and interaction. However, their implementation requires specialized equipment, higher costs, and technical expertise [40, 42].

Therefore, the choice of co-culture strategy should be guided by the specific biological question. Dynamic systems may offer enhanced physiological relevance for studying complex tumor–immune interactions, whereas static platforms remain valuable for high-throughput and mechanistic investigations [39, 43].

Functional evaluation of co-culture systems

A robust, multi-parameter evaluation framework is required to assess and validate co-culture models [40]. Evaluation typically encompasses three core dimensions: immune cell function, tumor cell killing, and secretory profiling.

Immune cell functionality is commonly assessed by the expression of activation markers (e.g., CD69, CD25), effector molecule production (e.g., IFN-γ, granzyme B), and the upregulation of exhaustion markers (e.g., PD-1, Tim-3, Lag-3) under sustained stimulation [17, 44]. The characterization of memory subsets (T_CM, T_EM) and the identification of functionally potent populations, such as CD39⁺ CAR-T cells, represent additional readouts for evaluating persistence and functional capacity [45, 46]. Table 1 provides an overview of key experimental readouts and assays used to assess immune cell function and tumor cell killing in organoid–immune co-culture models.

Table 1.

Functional evaluation indicators of HCC organoid–immune co-culture models

Evaluate the indicators	Specific indicators	Detection method	Functional significance
Immune cell function and phenotype	Activation markers (e.g., CD69, CD25)	Flow cytometry	Indicates the early activation status of immune cells
	Secretion of effector molecules (e.g., IFN-γ, Granzyme B)	ELISA, intracellular factor staining Flow cytometry	Direct measure of immune cell killing ability
	Exhaustion markers (e.g., PD-1, Tim-3, Lag-3)	Flow cytometry	Assess the risk of failure under long-term stimulation
	Proportion of memory subsets (e.g., TCM, TEM)	Flow cytometry	Indicates long-term immune protection and durability
	Specific functional subsets (e.g., CD39 CAR-T)	Flow cytometry	Identify cell subsets with specific high-potency functions
Killing effect on tumor cells	Direct kill and vitality inhibition	LDH release assay, real-time cell analysis (xCELLigence), ATP assay (CellTiter-Glo)	Quantitative detection of tumor cell death and growth inhibition
	Induce apoptosis	Caspase-3/7 activity detection, Annexin V staining	Confirm the programmed death mode of tumor cells
	Long-term growth inhibition	In vivo tumor volume measurement (PDX model), organoid size measurement	Validation of antitumor effects in more complex physiological settings
Cytokine/chemokine secretion profile	Pro-inflammatory factors (e.g., IFN-γ, TNF-α, IL-12)	ELISA, cytometric bead array(CBA)	Reflects the strength and direction of the anti-tumor immune response
	Chemokines (e.g., MCP-1, IL-8, CXCL16)	ELISA, RNA sequencing	Reflects the ability to recruit immune cells and shape the immune microenvironment
	Engineering factors (e.g., IL-7, CCL19)	ELISA	Validation of functional secretion of engineered immune cells
	Global pathway activation	RNA sequencing, proteomic analysis	Fully validate activation of signaling pathways (e.g., TNF, inflammatory response pathways) at the genomic level

Note: Indicators and methods listed in this table are synthesized from standard immunological assay protocols reported in major organoid–immune co-culture studies [23, 47, 91, 133]

Tumor cell killing is commonly quantified using real-time impedance assays (e.g., xCELLigence), lactate dehydrogenase (LDH) release assays, or ATP-based viability assays [47, 48]. Apoptosis is typically assessed using Caspase-3/7 activity assays or TUNEL staining [49, 50]. Longitudinal changes in organoid growth can be monitored using imaging-based approaches, while in vivo validation using patient-derived xenograft (PDX) models provides supportive evidence of therapeutic relevance [21, 51].

Cytokine and chemokine profiling is widely used to characterize intercellular communication within co-culture systems. Pro-inflammatory cytokines (IFN-γ, TNF-α, IL-12) are commonly associated with immune activation, while chemokines (MCP-1, CXCL16) are associated with immune cell recruitment [52, 53]. For engineered immune cells, verification of synthetic cytokine production (e.g., IL-7, CCL19 from CAR-T cells) is important for evaluating functional activity [54, 55]. Transcriptomic and pathway analyses (e.g., TNF, IFN-γ response pathways) provide deeper mechanistic insights into immune activation and resistance [56].

Integrating these phenotypic, functional, and molecular readouts helps establish the biological relevance of co-culture models and supports their application in immunotherapy research, although their predictive value for clinical outcomes remains to be fully validated [14, 17].

Tumor-immune interaction mechanisms revealed by co-culture models

Co-culture models provide a valuable platform to investigate the bidirectional crosstalk between HCC cells and immune populations within the TME. These systems can help elucidate mechanisms by which tumors establish immunosuppressive microenvironments, as well as how immune effector functions may counteract these processes, although the extent to which specific pathways are faithfully recapitulated in vitro remains an important consideration [14].

Tumor-mediated immunosuppression

HCC organoids can suppress immune function through multiple interrelated pathways [14].

Immune checkpoint activation

One of the major mechanisms involves the upregulation of PD-L1 on tumor cells, which engages PD-1 on T cells and CAR-T cells, resulting in impaired proliferation, reduced cytokine (IFN-γ, TNF-α) secretion, and diminished cytotoxic activity—features consistent with T-cell exhaustion [57]. Co-expression of additional inhibitory receptors such as Tim-3 and Lag-3 may further exacerbate T-cell dysfunction [58].

Soluble factor secretion

HCC cells and associated stromal cells (e.g., CAFs, MSCs) can secrete immunosuppressive cytokines such as TGF-β and IL-10, which are known to inhibit CD8⁺ T-cell activity and promote Treg expansion [59]. Chemokines such as CXCL12 are implicated in the recruitment of suppressive MDSCs and Tregs into the TME [60].

Metabolic reprogramming (adenosine pathway)

HCC cells expressing ectonucleotidases CD39 and CD73 can convert extracellular ATP into adenosine [61]. Adenosine binding to A2A receptors on T and NK cells activates cAMP-dependent immunosuppressive signaling, thereby contributing to a metabolic checkpoint that limits anti-tumor immunity [62].

Stromal manipulation

Tumor cells can modulate stromal compartments (e.g., MSCs/CAFs) to promote immunosuppressive phenotypes. In HCC, MSC–PDO–PBMC co-culture platforms have been shown to support monocyte/macrophage survival and promote TAM (including M2-like) differentiation [63]. Chemokines such as CCL2 (and in some contexts CCL5) are implicated in monocyte recruitment and macrophage polarization programs in HCC-associated stroma [64, 65]. CAFs are associated with collagen-rich ECM remodeling and have been linked to T-cell exclusion in stroma-rich HCC, while angiogenic signaling (e.g., VEGF) may contribute to the establishment of immunosuppressive niches by impairing antigen presentation and favoring suppressive myeloid and Treg programs [66].

Innate immune evasion

HCC cells can evade NK cell surveillance through multiple mechanisms, including upregulation of inhibitory ligands (e.g., HLA-E), secretion of decoy ligands (truncated MICA/B), and production of cytokines such as IL-6 that can impair NK cell cytotoxicity [67, 68].

Collectively, these findings highlight multiple interconnected mechanisms through which HCC establishes an immunosuppressive TME, including checkpoint signaling, soluble factor secretion, metabolic reprogramming, stromal remodeling, and innate immune evasion. Co-culture systems provide valuable experimental platforms to interrogate these pathways in a controlled setting [14, 17, 63].

Immune effector functions against HCC

Conversely, co-culture models illustrate how immune cells can mediate anti-tumor activity and modulate the TME.

Direct cytotoxicity

Activated CD8⁺ T cells and NK cells mediate direct tumor cell killing. T cells can recognize tumor antigens and release perforin and granzymes to induce apoptosis [69]. Engineered GPC3-targeted CAR-T cells, particularly those engineered to secrete IL-7 and CCL19, have been reported to enhance immune cell infiltration and cytotoxic activity in co-culture systems [70].

Reversal of immunosuppression

Immune checkpoint blockade (e.g., anti-PD-1) in co-culture systems can partially reverse features of T-cell exhaustion and restore aspects of effector function [17]. Engineered ‶immune hotspot”-creating CAR-T cells (secreting IL-7/CCL19) have been shown to recruit dendritic cells and bystander T cells, and may promote a more immunologically active microenvironment in vitro [71].

Metabolic interference

Activated immune cells can compete with tumor cells for nutrients (e.g., glucose and amino acids), thereby contributing to metabolic stress within the tumor microenvironment [72]. NK cells can also generate reactive oxygen species (ROS), which can impair tumor mitochondrial function [73].

Inhibition of angiogenesis

CD8⁺ T cells and NK cells can produce IFN-γ, which has been reported to suppress endothelial cell proliferation and VEGF-driven signaling, thereby potentially limiting tumor vascularization and growth [74, 75].

Memory Formation

Sustained co-culture studies suggest that CAR-T cells can give rise to central memory and effector memory subsets, indicating potential relevance for persistence and long-term immune function [76].

Collectively, these findings demonstrate that co-culture systems can capture multiple dimensions of immune effector function against HCC, including direct cytotoxicity, reversal of immunosuppression, metabolic competition, modulation of angiogenesis, and the emergence of memory phenotypes. These platforms provide a controllable framework to evaluate functional immune responses and to explore mechanisms underlying therapeutic interventions [17, 69, 70, 75].

Integrative signaling pathways

Co-culture models provide a platform to investigate key signaling networks governing tumor–immune interactions.

Checkpoint and exhaustion pathways

The PD-1/PD-L1 axis is a key regulator of T-cell function, with Tim-3 and Lag-3 acting as important co-inhibitory receptors [77]. Genetic silencing of these receptors has been shown to enhance CAR-T cell activity, suggesting a cooperative role in T-cell exhaustion [78].

Metabolic Immunosuppression Pathway

The CD39/CD73–A2AR adenosine axis is recognized as a key metabolic checkpoint. Targeting this pathway has been shown to partially restore immune cell function and may synergize with immune checkpoint blockade [79, 80].

Cytokine networks

The balance between immunostimulatory (IFN-γ, IL-2, IL-15) and immunosuppressive (TGF-β, IL-10) cytokines plays a central role in shaping the immune milieu [81]. IFN-γ is known to enhance antigen presentation but may also exert context-dependent effects on cancer stem cell properties [82].

Chemokine axes

Opposing chemokine gradients contribute to shaping the TME. The CCL19/CCL21–CCR7 axis is associated with anti-tumor immune responses through the recruitment of dendritic cells and T cells, whereas the CXCL12–CXCR4 axis is implicated in the recruitment of suppressive cell populations and may facilitate metastatic progression [32, 83].

Effector and immunogenic death pathways

The perforin–granzyme pathway is a central effector mechanism for CTL/NK-mediated tumor cell killing [84]. Oncolytic viruses (e.g., V937) have been shown to induce immunogenic cell death (ICD), leading to the release of damage-associated molecular patterns (DAMPs), which can activate dendritic cells and promote T-cell priming [85].

Taken together, co-culture models suggest a dynamic equilibrium in the HCC tumor microenvironment, shaped by competing suppressive (e.g., checkpoints, adenosine, TGF-β) and activating (e.g., cytotoxic granules, IFN-γ, immunogenic cell death) pathways [86]. Dissecting these interwoven circuits provides a conceptual framework for understanding tumor–immune interactions and for informing the rational design of combination immunotherapies targeting multiple resistance mechanisms [87].

Overall, co-culture models provide an integrated framework to dissect tumor–immune interactions in hepatocellular carcinoma, capturing both immunosuppressive mechanisms and immune effector functions within a controllable experimental setting [16, 88]. By enabling the simultaneous interrogation of checkpoint signaling, metabolic regulation, cytokine networks, and cytotoxic pathways, these systems offer valuable insights into the dynamic balance that shapes the tumor immune microenvironment [89].

However, it is important to recognize that many of these observations are derived from simplified in vitro systems [16]. The extent to which these models fully recapitulate the spatial organization, cellular diversity, and temporal dynamics of the in vivo liver microenvironment remains incompletely defined [16, 89]. In particular, the predictive value of co-culture–based findings for clinical immunotherapy responses requires further validation through integrated in vivo studies and prospective clinical correlation [90]. Collectively, emerging evidence suggests that tumor-derived factors such as AFP may contribute to immunosuppressive remodeling through macrophage polarization, providing a conceptual framework for integrated tumor–immune interactions (Fig. 3).

Fig. 3.

Conceptual framework of tumor–immune interactions mediated by tumor-derived factors in HCC organoid–immune co-culture systems. Tumor-derived factors, such as alpha-fetoprotein (AFP), may contribute to immunosuppressive remodeling by influencing macrophage polarization toward an M2-like phenotype. M2-like macrophages can produce immunosuppressive cytokines, thereby modulating the tumor microenvironment and affecting tumor cell behavior. Potential downstream effects include the regulation of tumor cell stemness and therapeutic resistance through key signaling pathways. This schematic summarizes a putative interaction network and is intended to provide an integrative framework rather than a fully validated mechanistic pathway

Applications in immunotherapy research

Drug screening and mechanistic studies

HCC organoid–immune co-culture systems serve as a preclinical platform with high-content readouts for evaluating immunotherapeutic responses and dissecting resistance mechanisms [14]. These models can facilitate the optimization of monotherapies. For example, different CAR-T constructs (e.g., targeting GPC3) have been compared, and functional enhancers such as IL-7/CCL19 secretion have been identified to improve immune cell persistence and recruitment [70]. Co-culture studies have also helped to characterize markers of T-cell potency (CD39) and exhaustion (PD-1, Tim-3, Lag-3), suggesting that genetic or pharmacological blockade of these inhibitory receptors may restore CAR-T cell function [91]. Similarly, responses to immune checkpoint blockade (ICB) can be modeled, with studies suggesting context-dependent resistance associated with TGF-β signaling, adenosine metabolism, or MDSC recruitment [92].

A key application is the evaluation of rational combination strategies. Co-culture models have provided insight into potential synergistic interactions between oncolytic virotherapy and ICB [93]. Infection with the oncolytic virus V937 has been shown to induce immunogenic cell death (ICD), thereby promoting the priming of tumor-specific T cells [94]. These activated T cells can secrete IFN-γ, which may upregulate the viral receptor ICAM-1 on tumor cells, potentially enhancing viral oncolysis [95]. Concurrent PD-1 blockade may sustain T-cell effector activity and enhance antitumor responses to oncolytic virotherapy [96]. Likewise, combining CAR-T cells with PD-1 blockade may help to partially reverse CAR-T cell exhaustion and prolong cytotoxic activity, offering a potential strategy to overcome a suppressive TME [97].

Taken together, these findings highlight the utility of HCC organoid–immune co-culture systems for mechanistic dissection and preclinical evaluation of both monotherapy and combination immunotherapeutic strategies.

Personalized therapy and biomarker discovery

PDOs integrated with autologous immune cells have been proposed as “patient-in-a-dish” platforms for precision immunotherapy [98]. These systems may enable ex vivo assessment of treatment response [99]. Co-culturing PDOs with a patient’s own TILs, NK cells, or CAR-T cells can be used to evaluate cytotoxic efficacy in a patient-specific context, potentially informing patient stratification strategies [23]. In addition, these platforms support the functional screening of candidate personalized combination therapies, such as oncolytic virus plus ICB, based on the unique biological features of individual tumors [100].

Furthermore, co-culture systems provide a valuable framework for mechanistic biomarker discovery. On the immune side, CD39 expression has been associated with tumor-reactive CD8⁺ T cells, while co-expression of PD-1, Tim-3, and Lag-3 is commonly used to define exhausted phenotypes [37]. On the tumor side, molecules such as ICAM-1, which may facilitate viral susceptibility to V937, may link IFN-γ–mediated immune activation to increased viral susceptibility, suggesting their potential as predictive biomarkers for viro-immunotherapy [58]. In addition, transcriptomic profiling of PDOs can be used to identify gene signatures associated with immune resistance, which can then be functionally evaluated in co-culture systems [101].

Collectively, these findings support the potential of co-culture systems as patient-specific platforms for functional immunotherapy testing and biomarker discovery.

Dynamic studies of the tumor microenvironment

Advanced co-culture systems incorporating microfluidics and bioengineering enable the investigation of the spatiotemporal dynamics of the TME, which are increasingly recognized as important for immunotherapy outcomes [102]. These platforms can be used to model hypoxia, a key feature of solid tumors [102]. Hypoxic gradients activate HIF-1α, inducing PD-L1, VEGF, and CD73 expression, thereby contributing to the formation of an immunosuppressive, adenosine-rich niche that impairs T and NK cell function—a process that can be partially recapitulated in controlled co-culture systems [103].

The inclusion of stromal components (CAFs, MSCs, endothelial cells) enhances the physiological relevance of the model [104, 105]. CAFs have been reported to contribute to T-cell exclusion through physical ECM barriers and secretion of immunosuppressive factors such as TGF-β, while MSCs promote M2-like TAM polarization [63]. Endothelial cells regulate immune cell trafficking through adhesion molecules [106]. Using advanced imaging and spatial omics technologies, these systems can be used to map the formation of distinct immune niches, such as cytotoxic T cells at the tumor–stroma interface versus Tregs and MDSCs in deeper regions [107, 108].

Finally, long-term co-culture provides a framework to investigate aspects of the immunoediting process (elimination, equilibrium, escape) in vitro [109]. Sustained immune pressure may select for resistant tumor clones with features such as antigen loss, MHC-I downregulation, and upregulated immunosuppressive pathways [110]. Profiling these evolved organoids can reveal adaptive resistance-associated features (e.g., PI3K/AKT, adenosine signaling), thereby enabling preclinical exploration of strategies to counter immune escape, such as dual-target CAR-T cells or sequential therapy [111].

These observations collectively demonstrate the capacity of advanced co-culture systems to model spatiotemporal dynamics of the tumor microenvironment and to investigate mechanisms of immune adaptation and resistance.

Overall, organoid–immune co-culture systems provide versatile platforms for drug screening, mechanistic studies, personalized therapy evaluation, and dynamic modeling of the tumor microenvironment [16, 88]. By enabling functional interrogation of tumor–immune interactions in a controlled setting, these systems offer important opportunities for translational immunotherapy research [90].

However, it is important to recognize that many of these applications are currently based on simplified in vitro models [16]. Variability in culture conditions, immune cell composition, and experimental readouts may affect reproducibility across studies [112]. Moreover, the extent to which co-culture–based findings can reliably predict clinical immunotherapy outcomes remains incompletely established [90]. Further integration with in vivo models, standardized experimental frameworks, and prospective clinical validation will be essential to enhance their translational applicability [100].

Current challenges and future directions

Key limitations and controversies

Limited immune cell persistence and phenotype drift represent key challenges in organoid–immune co-culture systems. A major limitation is the restricted longevity of primary immune cells in vitro and the risk of phenotype drift under non-physiological cytokine conditions. As a result, short co-culture durations may lead to overestimation of acute cytotoxicity while failing to capture clinically relevant dynamics, including immune exhaustion, functional adaptation, and immunoediting processes.

In addition, heterogeneity in immune cell sources—including autologous versus allogeneic PBMCs, TILs, or immortalized cell lines—introduces variability in antigen specificity and baseline activation states, thereby complicating cross-study comparability [44, 113, 114].

Incomplete microenvironmental complexity and vascular deficiency. Most HCC organoid–immune co-culture systems remain highly simplified and often lack stromal compartments, functional endothelium, and perfusable vasculature. This limitation may distort immune infiltration, nutrient and oxygen gradients, and drug distribution—factors known to influence immunotherapy responses in vivo [38, 40, 43].

Reproducibility and lack of standardization. Protocol heterogeneity—including matrix composition, medium formulations, immune-to-organoid ratios, and readout selection—remains a major barrier to reproducibility and multi-center validation. Batch-to-batch variability of extracellular matrices and divergent definitions of “response” across assays further limit cross-study comparability. Establishing consensus SOPs and minimal reporting standards will be critical for benchmarking and for building clinically actionable datasets [115–117].

Uncertain clinical predictivity and validation gaps. While “patient-in-a-dish” paradigms are conceptually appealing, the predictive value of organoid–immune co-culture assays for clinical immunotherapy outcomes remains incompletely validated, particularly in HCC where underlying liver disease and etiologic heterogeneity influence immune responses. Many studies lack prospective correlation with patient outcomes or rely on small cohorts, thereby limiting generalizability. Robust clinical validation—ideally through prospective, blinded studies coupled with multi-omics readouts—will likely be required before these platforms can be routinely used for treatment stratification [99, 100, 118].

Technical and practical challenges

Despite significant progress, HCC organoid–immune co-culture models face technical hurdles that may limit their translational relevance [23].

Many current models remain simplified, often lacking critical stromal components (CAFs, MSCs) and vascular elements that shape the TME through biochemical signaling and physical architecture [119]. Immune diversity is also restricted, with most systems focusing on T cells while underrepresenting B cells, diverse DC subsets, and neutrophils, which collectively regulate immune responses [118]. Furthermore, static culture systems do not fully recapitulate the spatiotemporal heterogeneity of in vivo tumors, including oxygen and cytokine gradients, which are important for studying processes such as hypoxia-driven immunosuppression [120].

Maintaining functional immune cells in long-term co-culture represents a major bottleneck [14]. Primary T cells tend to undergo exhaustion or apoptosis due to insufficient antigen presentation and limited cytokine support (e.g., IL-2, IL-7, IL-15) [113, 121]. In addition, ex vivo–expanded immune cells, such as CAR-T cells, may diverge phenotypically from their in vivo counterparts [114].

Standardization and reproducibility are further challenged by inter-patient heterogeneity and the lack of uniform protocols for matrices, culture media, and cell ratios, resulting in inconsistent cross-laboratory data [115]. Finally, the high cost and infrastructure requirements associated with organoid biobanks and advanced co-culture platforms restrict access to well-resourced centers, thereby limiting broader adoption [116].

Future directions and strategic solutions

Future progress will require integrated solutions spanning bioengineering, immunology, and advanced analytics [14].

To address model complexity, next-generation systems should evolve toward multicellular, biomimetic ecosystems. Incorporating CAFs, endothelial cells, and MSCs may better recapitulate stromal crosstalk within the TME [122]. Bioengineering approaches such as microfluidic organ-on-chip platforms can be used to introduce physiological shear stress and nutrient gradients, while 3D bioprinting enables precise spatial organization of cells within perfusable, vascular-like architectures [120, 123].

To prolong immune cell function, potential strategies include engineered cytokine support (e.g., organoids modified to secrete IL-2 or IL-15), immune checkpoint modulation (e.g., low-dose PD-1 blockade in culture), and optimized immune-supportive media formulations [17, 98].

Improving standardization and throughput will be critical. This includes the establishment of annotated patient-derived organoid–immune cell biobanks and standardized operating procedures (SOPs) [117]. In parallel, automation, high-content imaging, and machine-learning–based analytics may enable high-throughput screening of drug combinations and resistance mechanisms [124].

Innovative analytical technologies will be important for decoding complex tumor–immune interactions. Single-cell and spatial multi-omics (e.g., spatial transcriptomics) can map immune subsets within defined tumor niches, while advanced 3D live imaging (e.g., light-sheet microscopy) enables longitudinal, high-resolution tracking of immune infiltration and cytotoxic activity [107, 125].

Ultimately, the goal is to facilitate clinical translation toward personalized immunotherapy. Organoid–immune co-culture platforms may serve as functional precision-oncology tools to evaluate patient-specific responses to immunotherapies (e.g., CAR-T cells and ICB combinations) ex vivo, potentially informing treatment selection [23, 98]. In addition, these systems can be used to model immunoediting processes, whereby sustained immune pressure selects for resistant tumor clones, thereby providing insight into the design of adaptive, multi-target therapeutic strategies to overcome immune escape [126, 127].

Core value and integration of co-culture models

Core value in HCC research

HCC organoid–immune co-culture models have emerged as a valuable framework with three key applications. First, they provide a preclinical platform for functional precision medicine. By reconstructing a patient-specific tumor–immune ecosystem, these systems enable functional evaluation of immunotherapies—such as CAR-T cells, immune checkpoint blockade, and oncolytic viruses—ex vivo, potentially informing personalized treatment strategies [17, 91].

Second, they serve as a platform for mechanistic discovery, facilitating the investigation of bidirectional crosstalk within the TME. These models can reveal critical feedback interactions, such as IFN-γ–induced ICAM-1 upregulation that may enhance immune- or virus-mediated killing, as well as the CD39/CD73–adenosine pathway contributing to metabolic immunosuppression [86]. They can also model aspects of the immunoediting process, illustrating how chronic immune pressure may select for resistant tumor clones, thereby providing a rationale for combination therapeutic strategies [128].

Third, they serve as platforms for therapeutic innovation. Co-culture systems can be used to optimize CAR-T cell designs, evaluate combination strategies, and investigate resistance mechanisms (e.g., antigen loss and T-cell exhaustion). These models also enable the study of key pathways (e.g., adenosine signaling) that influence therapeutic responses, integrating functional testing with biomarker discovery to support the development of next-generation immunotherapies [14, 91].

Collectively, these applications highlight the versatility of organoid–immune co-culture models in bridging mechanistic insights and translational research, although their predictive value for clinical outcomes requires further validation.

The necessity of interdisciplinary collaboration

Realizing the broader potential of organoid–immune co-culture models requires close interdisciplinary collaboration across bioengineering, immunology, and clinical oncology to advance these systems from experimental platforms toward clinically relevant tools.

Bioengineering provides the structural and technological foundation, including the development of microfluidic organ-on-chip platforms that can regulate fluid dynamics and biochemical gradients, as well as 3D bioprinting approaches that enable the construction of vascularized, biomimetic scaffolds for studying complex interactions under near-physiological conditions [129, 130].

Immunology and cell biology define the cellular components and molecular mechanisms underlying tumor–immune interactions. Immunologists optimize cytokine environments and apply spatial multi-omics approaches to map signaling networks [131], while cell biologists develop methods to incorporate stromal components such as CAFs and endothelial cells to improve biological relevance [132].

Clinical medicine and oncology play a critical role in validation and translational direction. Clinicians frame key translational questions (e.g., mechanisms of resistance) and provide patient-derived tissues that help ensure co-culture models better capture real-world heterogeneity. In turn, insights from co-culture studies can inform clinical trial design, facilitating a bidirectional bench-to-bedside feedback loop [17].

Through coordinated interdisciplinary efforts, these approaches may help advance the development of functional precision immunotherapy strategies for HCC, although continued validation and standardization will be required to support clinical translation.

Conclusions

HCC organoid–immune co-culture systems have emerged as a valuable experimental framework that integrates tumor biology, immunology, and bioengineering to model patient-specific tumor–immune ecosystems in vitro. These platforms enable mechanistic investigation of immune evasion pathways, functional assessment of immune-mediated cytotoxicity, and preclinical evaluation of immunotherapeutic strategies, including immune checkpoint blockade and adoptive cell therapies.

By preserving key features of tumor heterogeneity while allowing controlled immune modulation, co-culture models provide a potential translational bridge between reductionist cell culture systems and complex in vivo settings. Collectively, these systems represent a promising approach in functional precision oncology, with the potential to contribute to therapeutic stratification, biomarker discovery, and the development of rational combination immunotherapies in hepatocellular carcinoma.

However, the extent to which these in vitro platforms can reliably predict clinical outcomes remains to be fully established, and further validation through standardized protocols, in vivo studies, and prospective clinical investigations will be required.

Future perspectives

Despite substantial progress, several challenges need to be addressed to further advance the translational potential of organoid–immune co-culture platforms. Current models often lack stromal and vascular complexity, and the limited persistence and functional stability of immune cells may limit long-term modeling of tumor–immune dynamics. Variability in culture protocols and matrix composition may further hinder reproducibility across laboratories.

Future advancements are expected to involve multicellular bioengineered systems incorporating stromal and endothelial components, microfluidic organ-on-chip technologies that more closely mimic physiological gradients, and 3D bioprinting strategies enabling spatially organized tumor microenvironments. Integration with single-cell and spatial multi-omics approaches may further enhance the resolution of immune–tumor interactions and facilitate biomarker discovery.

Ultimately, standardized, high-throughput co-culture platforms combined with clinically annotated biobanks may enable more refined functional testing of patient-specific immunotherapies, potentially informing personalized treatment strategies and facilitating the clinical translation of precision immunotherapy for hepatocellular carcinoma.

However, the extent to which these advances can be consistently translated into reliable clinical applications remains to be fully established, and continued efforts in standardization, validation, and multi-center studies will be essential.

Author contributions

WWL conceived and designed the review, conducted the literature search, and drafted the manuscript. XSX assisted with literature screening, organization of references, and manuscript revision. XJY supervised the study, provided overall conceptual direction, and critically revised the manuscript for important intellectual content. All authors read and approved the final manuscript.

Funding

This work was supported by the Natural Science Foundation of Gansu Province (No. 25JRRA304), the Doctoral Fund of the Key Laboratory of Gastrointestinal Tumor Diagnosis and Treatment of the National Health Commission (No. NHCDP2022001), and the Doctoral Supervisor Cultivation Program of Gansu Provincial Hospital (No. 22GSSYA-3). The funding bodies had no role in the design of the study, literature analysis, or preparation of the manuscript.

Data availability

Data sharing is not applicable to this article as no datasets were generated or analyzed during the current study.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare that they have no competing interests.