Retinoblastoma: unveiling molecular pathogenesis and pioneering organoid-driven therapeutic innovations.

Background
Retinoblastoma (RB), which is the most common intraocular malignancy in children, arises primarily from biallelic inactivation of the RB1 tumor suppressor gene and has an incidence of approximately 1 in 15,000–20,000 live births worldwide, representing approximately 8,000 newly diagnosed cases annually [1]. The predominant clinical manifestations of RB are leukocoria and strabismus, and leukocoria constitutes the presenting symptom in approximately 60% of cases [2]. It can metastasize or invade other tissues, poses a significant threat to vision and is life-threatening. Despite advancements in treatments, such as chemotherapy, radiotherapy and enucleation, challenges such as toxicity, relapse, and resistance persist, underscoring the need for deeper molecular insights and personalized strategies [3]. Many mechanisms have been shown to be associated with the pathogenesis of RB, including RB1 loss [4], MYCN amplification [5], epigenetic dysregulation [6, 7], and dysregulated pathways, resulting in the production of targeted medicines (e.g., CDK4/6 inhibitors and sunitinib).
Traditional models, including murine systems and two-dimensional cell lines, exhibit limited translational relevance owing to species-specific discrepancies and loss of tumor heterogeneity [8]. For example, RB1–/– mice fail to recapitulate human RB histopathology, highlighting the need for human-specific models [9]. Recent breakthroughs in organoid technology, particularly in retinal organoids (ROS) generated from human embryonic stem cells or induced pluripotent stem cells (iPSCs), offer transformative potential. These three-dimensional models faithfully mimic the tumor microenvironment (TME), retain patient-specific genetic profiles, and enable high-throughput drug screening [10, 11]. Notably, CRISPR/Cas9-edited RB organoids identified ARR3-positive cone precursors as the cellular origin of RB, resolving longstanding debates regarding tumorigenesis [8, 11].
This review systematically explores the molecular pathogenesis of RB, focusing on genetic alterations (RB1 loss and MYCN amplification), epigenetic dysregulation, and aberrant signaling pathways (e.g., PI3K/AKT/mTOR). It further evaluates the role of organoids in tumor biology, drug discovery, and precision medicine [12, 13].

Clinical feature
The progression of RB is marked by four clinical stages, each with distinct features. Leukocoria is one of the earliest visible signs and is characterized by yellow-white reflection in the pupils of children [14]. The associated symptoms included perceptual exotropia, strabismus, ocular redness, and pain. Advanced disease may cause iris discoloration, proptosis due to increased intraocular pressure, and grayish translucent retinal lesions that evolve into opaque white/yellow masses with dilated surface vessels and hemorrhages. Tumor extension beyond the globe into the optic nerve or orbit or metastasis to the central nervous system, blood, or lymph nodes is life-threatening [15, 16].

Mechanisms/pathophysiology

Genetic basis of the RB
RB is the first tumor to be clearly identified as having a genetic etiology, and its occurrence closely pertains to inactivation of the RB1 gene [4]. The RB1 gene is located in the q14 region of chromosome 13 and encodes pRB, a key protein involved in cell cycle regulation [4]. According to Knudson's two-hit theory [17], the occurrence of RB requires sequential inactivation of two RB1 alleles. Children with hereditary RB carry a germline mutation, so only a single postnatal somatic mutation in retinal cells is required for tumor formation. This somatic mutation can be caused by chemical, physical, or biological factors, including DNA viruses [18]. In nonhereditary RB, mutations of both alleles in the same cell are required [1].
The mechanisms underlying the inactivation of the RB1 gene are multifaceted and include point mutations, small insertions and deletions, loss of heterozygosity, and promoter hypermethylation [18]. Different inactivation mechanisms contribute to the loss of pRB function and differences in molecular effects through pathways, such as changes in the phosphorylation state, interference by viral oncoproteins, or abnormalities in upstream pathways [19].
pRB inhibits the progression of the cell cycle by combining with the E2F transcription factor. Unphosphorylated pRB in the G1 phase blocks the transcription of genes downstream of E2F. When CDK4/6 or CDK2 is activated, phosphorylated pRB dissociates from E2F, driving the cells into the S phase [20, 21]. Inactivation of the RB1 gene leads to inactivation of pRB function and cell cycle regulation, which is the core mechanism underlying the occurrence of RB. However, in approximately 2–5% of RB cases (especially the MYCN-amplified subtype), no mutations in the RB1 gene were detected. Moreover, the abnormal increase in the phosphorylation level of pRB resulted in its functional inactivation, confirming the existence of an alternative oncogenic pathway independent of gene mutations [22, 23]. This highlights the importance of studying the pathogenesis of nonmutated RB.

MYCN amplification
MYCN amplification is a key oncogenic driver in RB, and this alteration is observed in approximately 1–2% of RB cases [5]. The MYCN gene encodes the N-myc transcription factor, which regulates cell growth and differentiation. Its amplification often indicates poor prognosis and is an important prognostic indicator of RB [22, 24].
MYCN promotes RB occurrence through a dual mechanism. This leads to the overexpression of the N-myc protein, which activates the transcription of Myc target genes. In contrast, it abnormally activates kinases such as CDK4/6 by upregulating the expression of CDK genes, resulting in the overphosphorylation and inactivation of pRB. Even if the RB1 gene is not mutated, it can bypass the tumor suppressor function of RB1 [23, 25]. In addition, MYCN cooperates with DEAD box 1 and the opposite strand of MYCN to promote carcinogenesis. The MYCN opposite strand enhances the tumor-promoting effects of MYCN by stabilizing MYCN proteins [22, 26].
Studies using mouse models have demonstrated that MYCN overexpression accelerates RB occurrence. However, after MYCN expression was inhibited, the tumor initially regressed but eventually relapsed, and the relapsed tumor cells no longer relied on MYCN to drive tumor progression. This confirms the evolutionary adaptation of RB from MYCN dependent to MYCN independent; thus, achieving long-lasting therapeutic effects by targeting MYCN alone is challenging [25].
In addition to the inactivation of RB1 and amplification of MYCN, the development of RB involves various other molecular mechanisms, including the regulation of microRNAs (miRNAs) and long noncoding RNAs (lncRNAs), abnormal activation of signaling pathways, and epigenetic modifications.

miRNA regulation
miRNAs play critical regulatory roles in the pathogenesis and progression of RB. As a class of small noncoding RNAs, miRNAs adjust cellular homeostasis by triggering the degradation of target mRNAs [27]. For example, overexpression of miR-133a-3p suppresses the cell cycle and induces apoptosis in RB cells [28]. Similarly, miR-506 inhibits apoptosis by targeting SIRT1, whereas its downregulation promotes programmed cell death and inhibits RB proliferation [29]. Additionally, miR-361-3p is underexpressed in RB, and its overexpression inhibits cellular proliferation and stemness by targeting Gli1/Gli3, a key component of the Hedgehog signaling pathway, thereby exerting antitumor effects [30]. These findings highlight the potential of miRNAs such as miR-133a-3p and miR-361-3p as therapeutic targets. Moreover, mir-25-3p promotes the growth of tumor cells by regulating the PTEN/Akt signaling pathway [31]. Mir-144 serves as an indicator for the diagnosis and prognosis of RB [32]. Mir-21 functions in RB by targeting PDCD4 and downregulating the expression of RB1 [33]. Mir-378a-3p restricts the hyperplasia of RB cells but promotes apoptosis by inhibiting FOXG1 [34].

LncRNA regulation
LncRNAs are a class of noncoding RNA molecules that regulate gene expression, modify chromatin, and regulate cell function by sponging a specific miRNA that hinders binding to the mRNA target [35]. As a tumor suppressor, MEG3 reactivation inhibits tumor cell growth. Yan et al. [36]reported that the overexpression of MEG3 significantly downregulated the protein levels of p-PI3K, p-AKT, and p-mTOR, confirming that MEG3 suppresses the PI3K/AKT/mTOR signaling pathway. Additionally, Gao et al. [37] demonstrated that MEG3 may influence the invasion and metastasis of RB cells by modulating the Wnt pathway. These findings highlight the dual regulatory mechanisms of MEG3 in restraining tumor progression through the PI3K/AKT/mTOR and Wnt signaling pathways. Treatment with the demethylating agent 5-Aza-CdR restores MEG3 expression, suppresses tumor growth, and confirms its role as a tumor suppressor [38]. These results demonstrate that the epigenetic silencing of MEG3 contributes to RB pathogenesis, emphasizing its potential role as a therapeutic target.

Epigenetic modification
Epigenetic modifications, including DNA methylation, histone modification, and RNA modification, are heritable modifications that do not change the DNA sequence but affect gene expression [6, 7]. Paolo et al. [7] reviewed the role of the RB protein family in epigenetic regulation. The RB f protein family participates in the regulation of chromatin architecture and function, affects the establishment and maintenance of heterochromatin, and regulates gene expression. Inactivation of RB1 leads to changes in the epigenetic field of the whole genome and promotes tumorigenesis. Han et al. [6] found that m6A RNA methylation plays an important role in RB tumorigenesis. METTL3, a m6A methyltransferase, affects gene expression and signaling pathway activity by regulating the modification level of target gene mRNAs, facilitating the initiation and progression of RB.

Cancer stem cells
Multiomics data have been used to illustrate the existence of two RB subtypes [39]. Subtype 1 has an earlier onset and comprises most hereditary forms. In addition to inactivation of the initiating RB1 gene, few other genetic alterations are present in differentiated tumors that express mature cone markers. In contrast, subtype 2 tumors frequently exhibit recurrent genetic alterations, including MYCN amplification. They express fewer differentiated cone markers and neuronal/ganglion cell markers, with significant intra and intertumor heterogeneity. Cone dedifferentiation in subtype 2 is associated with stemness features including low immune and interferon responses, activation of E2F and MYC/MYCN, and high transferability. The increased stemness of subtype 2 tumors may also be associated with increased MDM4 expression. High SOX4 expression may be associated with local diffusion, whereas high OTX2 expression helps RB cells acquire proliferative and stem-like properties and induces the transdifferentiation of retinal cells through signaling pathways [40].
Cancer stem cells (CSCs) rely on the TME to maintain their stemness. CSCs can recruit monocytes and macrophages by secreting chemokines and cytokines and reprogramming them into tumor-associated macrophages (TAMs) [41]. Conventional cytokines produced by M2-like TAMs, including transforming growth factor-β, tumor necrosis factor α, and interleukin-6, play crucial roles in supporting CSCs. The CSF-1/CSF-1R signaling pathway facilitates macrophage migration, proliferation, and survival and is often abnormally activated upon interaction with CSCs. Studies [42] have shown that blocking CSF1/CSF1R expression inhibits TAM recruitment. TAMs promote angiogenesis by secreting vascular endothelial growth factor [16], and RB-derived exosomes promote angiogenesis through microRNA-92a-3p, providing nutritional support for RB CSCs [43]. In contrast, cancer-associated fibroblasts enhance tumor invasiveness by remodeling the extracellular matrix [44]. Additionally, the interaction between RB CSCs and retinal astrocytes activates the IGFBP-5/PI3K/AKT pathway, further promoting cell proliferation.
The immune cells in the TME include cytotoxic T lymphocytes (CTLs) and regulatory T-cells (Tregs). CTLs induce necrosis and apoptosis and prevent growth by secreting interferon-γ(IFN-γ) [41]. However, Tregs inhibit the proliferation of CD8+ cells, suppress macrophages and antigen-presenting cells, and reduce the cytotoxicity of natural killer cells. JAK1 and JAK2 are the two most important signaling pathways that increase the efficacy of immunotherapy [45]. Mutation of these two genes results in an insufficient response to IFN-γ, ultimately resulting in a reduced sensitivity of cancer cells to antiproliferative agents. Another mechanism by which IFN-γ induces drug resistance involves accelerating the sequestration and nuclear translocation of Yes-associated protein (YAP) and increasing the density of YAP in cancer cells instead of increasing the transcription of target genes to regulate anti-PD-1 in immunotherapy [46].
RB-derived exosomes (EXOs), which contain cytokines, miRNAs, and mRNAs, contribute to the growth and metastasis of RB cells [47]. The results of this study suggest that RB EXOs may be considered therapeutic targets and that their composition could be regulated to inhibit RB cells. Alternatively, inhibitors that disrupt the uptake of EXOs could be developed.
Studies have demonstrated that vitamin D regulates the inflammatory state of the TME by influencing CSCs [48]. Additionally, they can enhance the efficacy of anticancer drugs such as cisplatin, doxorubicin, and gemcitabine [49]. Notably, vitamin D reduces the expression of stem cell markers in stem-like cells, a process that inhibits tumor sphere formation [48] Fig. 1.
The main pathways of retinoblastoma include the PI3K/AKT/mTOR signaling, Hippo/YAP signaling, Hedgehog signaling, and pRB-p53 pathways. During DNA damage, p53 is phosphorylated and released from the MDM2 protein, allowing p53 to bind to DNA and stimulate the production of proteins. which inhibits CDK4/6. p16 also inhibits CDK4/6, preventing the phosphorylation of pRB. Unphosphorylated pRB sequesters E2F, regulating cell cycle progression. MDM2, influenced by MYCN and MDMX, negatively regulates p53. Collectively, these pathways integrate signals to control cell cycle progression and survival/decreased apoptosis, highlighting the intricate balance of regulatory mechanisms within the cell.

Abnormal activation of the signal pathways
RB tumorigenesis involves abnormal stimulation of various signaling pathways, including the PI3K/AKT/mTOR [6], Hippo/YAP signaling [30], and Hedgehog signaling [50], pathways, as well as the pRB, p53, HIF-1, and NF-κB signaling pathways. In the YAP/TAZ pathway, phosphorylated YAP/TAZ binds 14-3-3 and is degraded, whereas unphosphorylated YAP/TAZ interacts with TEAD. SOX2 influences this network and contributes to the regulation of the YAP/TAZ pathway. The SHH signaling pathway, which involves PTCH1 and SMO, affects GLI1, which is modulated by miR-133-2p. The effects of pRB have been previously described. Furthermore, p53 facilitates apoptosis and cell cycle arrest at the G1/S phase. In RB tumors, the regulators of p53 are inhibited, promoting cell cycle progression, and potentially contributing to oncogenesis [51]. HIF-1 reacts to hypoxic conditions and vitreous seeding triggers HIF-1 production in the RB, increasing VEGF and PDK1 expression [52]. This promoted cell proliferation and oncogenesis. NF-κB also affects RB and becomes constitutively activated, inhibiting apoptosis [51]. Han et al. [6] reported that the m6A RNA methyltransferase METTL3 is highly expressed in RB tissues and cell lines. Mechanistic studies have shown that METTL3 activates the PI3K/AKT/mTOR signaling pathway and promotes RB tumorigenesis. Zhao et al. [53] reported that the transcription factor SOX2 was abnormally highly expressed in RB tissues, promoting the stemness of RB cells and activating the Hippo/YAP signaling pathway. Dan et al. [30] reported that miR-361-3p inhibited the proliferation and stemness of RB cells by targeting the hedgehog signaling pathway. These aberrant signaling pathways do not operate in isolation; instead they form an intricate regulatory network that sustains tumor cell survival and proliferation by modulating multiple biological hallmarks.

Targeted drugs
Current treatment methods for RB include intravenous, arterial and intravitreal chemotherapy; cryotherapy; radiotherapy; and surgery [3]. However, these treatments may contribute to adverse reactions, such as toxic side effects, postoperative complications, blindness after surgery, and other complications attributed to radiotherapy. With technological advancements, treatment methods have gradually moved towards targeted drug therapy.
RB1 is a multifunctional protein that incorporates signals from multiple upstream events and plays an important role in cell cycle regulation [54]. The importance of the RB1–E2F signaling axis in cancer has been widely recognized. Currently, the only clinically approved drug targeting this pathway is a CDK inhibitor [55]. However, the use of CDK inhibitors for restoring RB activity at the protein level is challenging due to drug resistance and low protein expression levels. Currently, the use of small molecules that promote RB expression is considered an effective anticancer strategy.
Recognition of the expression of maturing cone precursors and the dependence of RB cells on MDM2, MYCN, SKP2, and SYK suggest the use of MDM2 synthetic lethal antagonists of MYCN-directed Aurora kinases [5] and inhibitors of SKP2 stability and SYK activity [3]. Therefore, MYCN is an attractive therapeutic target. MYCN overexpression sensitizes cone precursors to E2f inhibitors, whereas the RB1 −/− MYCN+ cell line is highly sensitive to drugs that target SKP2 [56], Aurora kinase [57], RAD51, and MDM2 [58]. Therefore, for recurrent MYCN motile RB, the most urgent challenge is the clinical validation of drugs targeting the identified vulnerabilities.
Studies have demonstrated that RB1 and E2F3a cooperatively regulate over one-third of spliceosome genes by binding to their promoter or enhancer regions. Pharmacological inhibition of spliceosome function results in extensive exon retention, reduced cell proliferation, and impaired tumorigenic capacity, highlighting the therapeutic potential of targeting the spliceosome in RB1-deficient retinoblastoma [59]. In addition, SKP2 deletion induces a synthetic lethal effect in RB1-null cells, making the SKP2 inhibitor MLN4924 a promising candidate for retinoblastoma treatment [60]. The histone H3K79 methyltransferase DOT1L has also been shown to enhance chemosensitivity in retinoblastoma cells [61]. Loss of RB1 alleviates transcriptional repression of ESRRG, promoting cancer cell survival under hypoxic stress. Targeting ESRRG using the specific inverse agonist GSK5182 or shRNA-mediated knockdown significantly induces cell death in RB1-null cells, with a paricularly pronounced effect under hypoxic conditions [62].
Han et al. [6] demonstrated that targeting the METTL3/PI3K/AKT/mTOR axis may be a potential strategy for RB therapy. The development of METTL3 inhibitors is expected to provide new drugs for the treatment of RB. Inhibitors of the PI3K/AKT/mTOR signaling pathway, such as rapamycin, may also be effective in patients with RB with high METTL3 expression. In the future, drug screening could be conducted by combining RB organoid models to identify METTL3 inhibitors or PI3K/AKT/mTOR signaling pathway inhibitors that are effective against METTL3 overexpressing RB cells, thereby providing new drug options for the precise treatment of RB.
(Table 1).

Advances in organoid technology
To address the biological limitations of traditional models that hinder clinical translation despite significant advances in understanding the molecular mechanisms of RB, it is imperative to develop novel in vitro models capable of preserving patient-specific tumor characteristics and supporting functional studies. This urgency arises from the limitations of current approaches; animal models struggle to simulate the genetic diversity of human RB, whereas two-dimensional cell lines frequently lose tumor heterogeneity during prolonged passaging.
In 2011, Sasai was the first to differentiate 3D reactive oxygen species in mouse embryonic stem cells (ESCs) in vitro. They generated completely layered ROS similar to in vivo retinal processes, including major retinal components such as photoreceptors, revealing a new method for the in vitro production of photoreceptors [10]. However, this method requires destruction of the embryo; thus, its feasibility in clinical practice is limited. ESCs are allogeneic in origin and trigger acute or persistent immune-mediated rejection in the host. Currently, the majority of iPSC lines are generated via the ectopic expression of Yamanaka factors, a set of transcription factors that reprogram somatic cells into iPSCs. While patient-specific iPSCs circumvent challenges related to immune rejection and ethical issues, gene manipulation may alter the genome and limit their clinical application [63]. Rouhani et al. [64] reported that strong screening pressure during Yamanaka factor-mediated somatic reprogramming of iPSCs resulted in the introduction of somatic mutations into the iPSC lines. Deng et.al. [65] successfully reprogrammed mouse embryonic fibroblasts into chemically induced pluripotent stem cells (CiPSCs) without introducing exogenous genes, which exhibit potential comparable to that of ESCs with respect to gene expression patterns, epigenetic status, differentiation, and germ cell transmission. Moreover, CiPSCs were established from autologous somatic cells to avoid immune rejection issues [65]. However, CiPSCs have only recently been successfully established, and their ability to differentiate in vitro, particularly 3D retinal differentiation, remains largely unexplored. 3D ROS derived from CiPSCs constitute an ideal cell source for cell transplantation therapy, suggesting a novel approach for retinal transplantation therapy aimed at treating retinal degenerative disorders [63]. CiPSCs have the capacity to differentiate into polarized and stratified 3D retinal organs containing all significant retinal cell types, such as photoreceptors, retinal ganglion cells (RGCs), horizontal cells, bipolar cells, and Müller glial cells [63]. The presence of cilia and outer segments confirmed the good maturation status of the photoreceptors. The expression patterns of retinal genes revealed that CiPSC-ROS production was similar to that associated with retinal development in vivo. Notably, CiPSC-derived photoreceptors can restore the light sensitivity of the degenerating retina. However, RGCs were not identified during the advanced phases of CiPSC-ROS differentiation. Because RGCs have axonal connections to the optic nerve in the retina in vivo and there is no generation of ROS in the optic nerve, RGCs are compromised because of the absence of axon development in the ROS. Moreover, if RGCs are distributed on the inner side of ROS, the oxygen and nutrient supply on the inner side of the organoid is significantly lower than that on the outer side, which also affects the survival of RGCs [63].
Commonly used organ culture methods cannot accurately reproduce the dynamic and complex microenvironments essential for organ development and maintenance in vivo [10]. The lack of a critical microenvironment for tumor growth inhibits the further growth and maturation of organoids, thereby distancing them from the physiological state of the original tumor-bearing organ [66]. Therefore, to simulate a more complex TME, tumor organoids are co-cultured with diverse cell types, such as cancer-associated fibroblasts [67], tumor-reactive T cells [68], chimeric antigen receptor T cells, and macrophages, as well as combinations of these cell types with other cell types [66]. Under physiological conditions, the levels of TAMs and astrocytes decrease during RB invasion, indicating that the RB environment is immunosuppressed and that fibroblasts might induce RB tumor proliferation [16]. Moreover, glial cells with astrocytic properties, such as the TME, promote the proliferation and survival of rb53 [69]. TAMs have also been linked to tumor vascularization and the invasion of RB tumor cells in cancer [16]. Macrophages from peripheral blood mononuclear cells (PBMCs) interact with soluble factors secreted by RB cells, thereby promoting the development of an immunosuppressive microenvironment within RBs [70]. Some research groups have identified differentially expressed genes that contribute to RB progression after the co-culture of RB tumor cells and PBMCs [71] or analyzed the effects of RB-derived exosomes on macrophages and bone marrow mesenchymal stem cells, the results of which demonstrated their tumor-promoting properties [47].. Although most of these results are based on sequencing analysis, they still provide valuable information concerning cell–cell interactions. To solve this problem, some researchers [72] have isolated heterogeneous RB derived stromal cell lines from magnetic cell separation to identify single stromal cell types. Finally, a 3D co-culture system of RB-derived stroma and tumor cells was established. Owing to the complexity of RB and its TME, tumor cells exhibit unique gene expression patterns and morphological characteristics according to different culture media, and a comprehensive examination of the interactions within the tumor stroma is expected to yield significant insights into the molecular mechanisms underlying tumor development and progression [72]..
Tumor organoid technology faces multiple challenges [13, 73]. First, the establishment, maintenance, and passage of organoids are costly. Second, the success rates vary significantly across cancer types owing to low reproducibility and instability, which are influenced by factors such as primary tissue cell density [73]. Third, standardized culture protocols are needed to increase the reproducibility of large-scale organoid production and enable high-throughput drug screening. Fourth, tumor heterogeneity raises concerns regarding the use of small biopsy samples to represent the entire tumor. Sampling multiple tumor regions may better capture heterogeneity for translational research [13]. Fifth, replicating patient-specific immune microenvironments remains challenging. While co-culturing tumor spheroids with immune cells aids in modeling tumor-immune interactions, variability in the immune cell composition (e.g., infiltrating vs. stromal immune cells) across tumor types limits the accuracy of immunotherapy response prediction. Sixth, vascularization challenges persist; current animal implantation or co-culture methods create structural vasculature without functional perfusion [13]. Microfluidic platforms for vascularization are limited by their crude adjustability and sensitivity to cytokine concentrations and flow rates, warranting the development of more precise systems to model antiangiogenic therapy responses. Addressing these challenges through standardized protocols, multiregional sampling, immune microenvironment reconstruction, and advanced vascularization platforms is crucial for advancing organoid applications in precision medicine and translational research. As described in Sect. “Cancer stem cells”, the TME of RB exhibits complex immunosuppressive features, comprising immune cell populations such as TAMs and Tregs, as well as intricate cytokine networks. Accurately recapitulating this environment in vitro is essential for investigating immune escape mechanisms and evaluating immunotherapeutic strategies. At present, the generation of immunocompetent RB organoids primarily relies on co-culture approaches, although their success rates vary.
The most commonly used strategy involves co-culturing RB organoids with immune cells, including peripheral PBMCs [70] and monocytes/macrophages [71]. Previous studies [70, 71] have demonstrated that RB cells can drive monocyte polarization toward immunosuppressive M2-like TAMs, suppress T-cell function, and identify associated differentially expressed genes. However, several limitations persist. Long-term co-cultures often suffer from low success rates and poor reproducibility, with limited immune cell viability and difficulty in maintaining the dynamic spatial interactions observed in vivo. Moreover, these systems are relatively simplistic and fail to capture the synergistic effects of multiple immune cell populations. In addition, the use of allogeneic or healthy donor-derived immune cells reduces physiological relevance and limits the ability to model patient-specific immune characteristics.
Although current co-culture models have provided valuable insights into RB-immune cell interactions, they remain insufficient to fully recapitulate the in vivo immunosuppressive network. Future efforts should focus on developing multi-component immunocompetent RB organoid platforms that incorporate autologous immune cells, stromal elements, and advanced bioengineering technologies. Such models will be critical for accurately assessing RB immunotherapies and for elucidating TME-driven mechanisms underlying retinoblastoma pathogenesis.

New generation of organoids
Common organoid models exhibit significant differences between batches, leading to poor reproducibility and reduced value for in vitro applications. Researchers have developed various methods to address these constraints, including the application of dynamic culture conditions using bioreactors [74], the vascularization of organs [75], and the ability to guide self-organization through multiple instructive signals. The integration of organoid systems with microfluidic chip technology is a promising solution for challenges such as reproducibility, environmental signals, maturity, and scalability. Achberger et al. [76] were the first to integrate ROS and retinal pigment epithelial (RPE) cells successfully into a chip. The open microfluidic device consists of a 3-channel microfluidic platform for moderate perfusion, which allows the perfusion of circulating immune cells. The choroid-to-chip system comprises three main cellular components. The epithelial layer was composed of a single layer of human iPSC-derived retinal RPE. The endothelial layer consisted of two consecutive single layers of human primary microvascular endothelial cells seeded in the central channel facing the RPE above and the melanocyte region below. This channel is separated from the RPEs by a semipermeable membrane. This design enhanced the development and maturation of ROS, as demonstrated by the invasion of retinaldehyde into the hydrogel between the two structures and the increased formation of inner and outer segment-like photoreceptor structures. The function of RPE cells is demonstrated by phagocytosis of the photoreceptor outer segments [76]. Xue et al. [77] reported another study demonstrating the feasibility of culturing ROS on a chip. In this study, a micrometer-scale microfluidic bioreactor was developed and optimized to improve the mass-transfer efficiency and uniformity of the medium concentration between the chambers in the chip. ROS at various differentiation stages were manually added, trapped in the chambers of a 30-channel chip, and cultured for more than one month. Compared with static culture, although there was no significant improvement in the differentiation and maturation of ROS, Xue et al. reported that the ROS produced by this chip were of comparable quality with reduced labor intensity [77]. As a new frontier technology, organoid chips can guide stem cell differentiation and organoid organization in a controlled cellular microenvironment. This technology can also integrate multiple analyses to monitor multiscale aspects of the culture environment and the behavior of organoids accurately.
In addition, 3D bioprinting technology plays an important role in technological integration and innovation in the field of organoids. Chen et al. [78] utilized 3D printing technology to develop a reverse mold (micropillars) and shaped a microporous platform for the cultivation of human brain organoids, which formed a robust structure with high-level characteristics. Sun et al. [79] achieved large-scale production of ROS in a high-throughput manner by combining modern 3D printing technology, providing powerful technical and platform support for exploring retinal pathogenesis and high-throughput drug screening [80].

Practical applications of organoids

Disease modeling
The RB organoid model is constructed primarily by knocking out the RB1 gene in hPSCs or patient-derived iPSCs via CRISPR-Cas9 gene editing technology. Liu et al. [8] generated organoids via biallelic knockout of RB1, recapitulating the key features of RB: tumor foci, Flexner-Wintersteiner rosettes, and proliferative preconal precursor cells (ARR3+Ki67+). Single-cell sequencing demonstrated that these cells re-entered the cell cycle and differentiated into RB-like or RB-like cell clusters, confirming that preconal precursors are the cells of origin for RB. Rozanska et al. [81] constructed a homozygous mutation model using patient-specific iPSCs carrying the heterozygous RB1 mutation c.2082delC and observed an accumulation of retinal progenitor cells, along with a reduction in horizontal and amacrine cells within the organoids, indicating that RB1 deficiency induces differentiation arrest. Additionally, patient-derived organoids can spontaneously form RB tumors in immunodeficient mice, accompanied by secondary mutations, such as MYCN/MDM4 amplification, which is consistent with aggressive clinical phenotypes [82]. This model not only validates the two genetic hit theories of RB occurrence but also reveals the dynamic changes in epigenetic and genomic instability during tumor progression, providing a reliable in vivo verification platform for screening targeted drugs against the MYCN/MDM4 pathway. Moreover, RB1 deficiency leads to RB development. By co-cluturing microglia with RB1 deficiency and ROS, the production of inflammatory factors and the increase in the level of phosphorylated extracellular regulated protein kinase (p-ERK) resulted in an enhanced innate immune response and the structural disruption of ROS [83]. These findings provide an important basis for understanding the role of microglia in the disease mechanism of RB. Furthermore, a recent study demonstrated that heterozygous RB1 mutations enhance ATP production in human iPSC-derived retinal organoids [84]. This finding not only advances our understanding of early tumor-initiating events and reveals a previously unrecognized role of tumor suppressor genes in metabolic regulation, but also enables iPSC-derived organoid models to more closely recapitulate the metabolic features of primary tumors, thereby providing a robust platform for subsequent mechanistic investigations and drug screening.

Drug screening and development
Conventional animal models frequently do not accurately predict drug toxicity and metabolism because of species-specific discrepancies, resulting in low clinical success rates in human trials [85]. The emergence of 3D organoid technology has enabled researchers to screen drug delivery for cancer and other diseases as well as in the development of highly biocompatible pharmaceuticals with minimal adverse effects [86]. The ROS derived from hPSCs provide a suitable platform for preclinical drug toxicity testing. Srimongkol et al. [71] screened 133 Food and Drug Administration-approved drugs using RB organoids and reported that sunitinib was highly cytotoxic against both RB1-deficient and novel MYCN-amplified RB organoids and was capable of significantly reducing the proliferation of tumor cones. Duangporn et al. [87] found that a retinal organoid model combining topotecan and melphalan was more effective than melphalan alone. Melphalan is extensively utilized in intravitreal chemotherapy; nevertheless, it is often ineffective in managing recurrent and refractory tumor seeds. Compared with melphalan alone, the combined drug regimen rapidly controlled the seeds and required fewer chemotherapy cycles. Zhao & Yan [12] reported that high-throughput screening of drugs could be performed in organoids combined with chips. Roche et al. [88] engineered a human retinal microvascular organ-on-a-chip model and used it for extended toxicity and pharmacokinetic evaluations. The model was developed via a high-throughput in vitro 3D functional vascular modeling platform capable of simulating a tight endothelial layer generated by retinal microvascular cells under continuous medium perfusion. A microdevice was used to simulate and quantify the transport and barrier disruption of small molecules during drug exposure. Intraocular drug delivery is one of the most challenging and popular topics in ophthalmic drug development and toxicological evaluation. Drug-loaded nanocarriers such as liposomes, nanoparticles, and extracellular vesicles may overcome many challenges and improve targeted retinal permeability, noninvasive delivery, enhanced drug bioavailability, and prolonged residence time [89]. These innovations, combined with high-throughput organoid screening [12], have reshaped ophthalmic drug development by bridging preclinical and clinical translatability.

Gene editing
CRISPR-Cas9 enables precise genome editing in organoids by utilizing guide RNA (gRNA) to direct Cas9 nuclease to induce targeted double-strand breaks (DSBs), facilitating gene knockout or donor sequence insertion via endogenous repair pathways [90, 91]. When integrated with organoid models, this technology has accelerated the development of genetically accurate disease systems for identifying cancer-associated oncogenes/tumor suppressors. Afanasyeva et al. [92] corrected LCA5 mutations in iPSCs to model Leber congenital amaurosis 5. Chirco et al. [93] employed allele-specific CRISPR-Cas9 to knockout mutant CRX, rescuing rod cell phenotypes in ROS and demonstrating the therapeutic potential of LCA7. Rozanska et al. [81] generated RB1-null iPSCs via CRISPR-Cas9, with derived organoids recapitulating RB tumor features (rosette and mitochondrial abnormalities) and anchorage-independent growth. CRISPR-Cas9-edited organoids provide a transformative platform for modeling genetic diseases, validating therapeutic targets, and advancing personalized oncology research through isogenic controls and patient-specific genomic recapitulation.

Precision medicine
Precision medicine leverages patient-specific organoids to tailor therapies on the basis of molecular and pharmacogenomic profiles [86]. ROS derived from patient iPSCs or engineered cell lines enable in vitro drug screening and intervention profiling, linking the genetic background to the therapeutic response. Preliminary drug screening via organoid biobanks marks the early stage of this technology; future developments include single-patient cell-based “chip biobanks”, enabling rapid functional differentiation for predictive diagnostics and tailored therapies [94]. By mapping mutation-specific phenotypes (e.g., RB1 loss and MYCN amplification) to improve intervention efficacy, organoid models can be used to optimize treatment strategies, bridge genomic insights, and facilitate clinical translation [86, 94, 95] Fig. 2.
Retinal organoids derived from stem cells are used in disease modeling and drug screening via chips-on-organoids for high-throughput analysis and toxicity assessment. Gene editing involves in vitro RNA editing and in vivo DNA synthesis. Organoids from autologous stem cells facilitate personalized medicine by performing drug sensitivity tests and optimizing chemotherapy regimens. Gene profiling in three-dimensional models investigates ocular diseases.

Conclusion
RB is a genetic tumor primarily associated with inactivation of the RB1 gene. Other contributing mechanisms include MYCN gene amplification, miRNAs, lncRNAs, epigenetic modifications, and certain signaling pathways. Numerous targeted drugs have been developed on the basis of these mechanisms of action. However, these drugs have certain limitations, such as drug resistance and severe side effects. RB organoids have revolutionized our understanding of tumor biology and therapeutic development. Their continued refinement and integration with advanced technologies will accelerate precision medicine approaches, ultimately improving the outcomes of patients with RB.