Despite significant academic, industrial, and clinical efforts, successfully translating preclinical findings to clinical trials and practice remains challenging (1–4) and less than 10% of drugs entering oncology clinical trials are eventually approved for clinical use (1, 5). Furthermore, these efforts and failures come at high financial and ethical costs. Pediatric malignancies have special considerations, since clinical drug testing is even more restricted by the relatively small number of patients and the ethics related to long-term side-effects in children. Involvement of multiple stakeholders is important to address the lack of childhood-specific drug development in the pharmaceutical sector (6, 7). Thus, there is an urgent need for clinically relevant and biologically accurate preclinical models to minimize these current bottlenecks to drug development and implementation (8).

Neuroblastoma (NB) is the most common solid extracranial pediatric tumor, accounting for ~15% of pediatric oncology deaths (9, 10). NB can be regarded as an aberration of neural crest development, and although it can arise anywhere along the sympathetic nervous system, most primary tumors are found in the adrenal gland (11). NB is biologically and clinically heterogenous, and patients are stratified into different risk groups based on tumor characteristics and disease presentation (12, 13). Clinical responses vary from spontaneous regression to metastatic and drug-resistant disease despite intensive treatment (13). Furthermore, patients often suffer from severe therapy-related long-term adverse effects (14).

NB is a copy number-driven disease with few targetable somatic mutations found at diagnosis, especially when compared with adult malignancies (15). The MYCN oncogene is amplified in ~20% of cases and is strongly correlated with aggressive phenotypes and unfavorable clinical outcomes (16). Other common chromosomal copy number changes, including 11q loss and 17q gain, are also poor prognostic features (17). Recurrent mutations are rare in NB but include ALK (9%), ATRX (7%), and PTEN (3%) mutations (15). In relapsed tumors, genome-wide sequencing has revealed a higher prevalence of recurrent mutations in targetable pathways, such as RAS-MAPK (18, 19), but at relatively low frequencies. Recent transcriptional and epigenetic analyses suggest that NB cells can adopt at least two phenotypic cell states, known as adrenergic (ADR)/differentiated and mesenchymal (MES)/immature (20–24). These findings highlight the heterogeneous and dynamic nature of NB.

Reliable, predictive, and authentic NB models are important because: (i) NB is uncommon, so sufficiently powered clinical trials are challenging and patient material is scarce for molecular studies (25), placing extra weight on the translatability of preclinical results; (ii) preclinical models that accurately recapitulate known clinical, genetic, and transcriptional intra- and inter-tumor heterogeneity are important for the identification of effective therapeutic targets; and (iii) patient-derived (PD) models resemble the clinical scenario better than conventional models and can therefore be used to screen for and test the most promising and safe novel therapies.

Here we discuss recent progress in patient-derived xenografts (PDXs) and PD in vitro cultures as preclinical NB models. We summarize the development of different PD models, their utility in studying biological mechanisms and treatment responses, and how they can be utilized for preclinical drug testing to improve treatment strategies against NB ( Figure 1 ).

Conventional cell lines have been used as laboratory models for decades, and they have provided valuable knowledge about tumor biology and drug efficacy in many cancer types. However, cancer cell lines are usually passaged under serum-containing conditions for years and thus their molecular profiles often differ from the original patient tumor (26, 27). This matters in terms of model fidelity, especially when considering clinical applicability; for example, clinically important features such as drug resistance might be lost after long-term in vitro passaging (27). PD NB models are established directly from tumor material obtained from children after parental informed consent. PD models have been shown to better reflect the features (e.g., treatment response) of their original tumors, compared with conventional models (28–31). PDXs have now been established from many diverse tumor types of adult and pediatric cancers including NB (32–35). Over the last few decades, the cancer research community has gradually turned towards PD model systems (8), and the US National Cancer Institute recently decided to replace its panel of human cancer cell lines (NCI-60) with well-characterized PDX models (36) for drug screening.

Conventional cell lines, cell-line derived xenografts, and genetically engineered mouse models have been the main preclinical tools used to study resistance mechanisms and for drug testing. By using PD models that retain the main characteristics and heterogeneity of the original tumors, patient tumors and their treatment response can be better represented in the laboratory, thereby bridging the gap between preclinical models and the clinic.

Although the successful establishment of PD NB models is encouraging, certain aspects can still be improved. For example, the distribution and function of the extracellular matrix (ECM) has been shown to influence NB progression in patient samples (117, 118). Consistently, modulation of ECM components induces specific cell behaviors of PD NB cultures (63). Optimization of ECM conditions could thus contribute to improved NB modelling.

The lack of a complete immune system is a limitation of most PDX models, since PD tumors are generally implanted in immunocompromised mouse strains (e.g., NSG) to permit tumor engraftment. Reconstitution of a humanized immune system, for example by injection of human hematopoietic stem cells into sub-lethally irradiated mice, could improve the immune status of the models (119). A technically advanced humanized mouse strain (MISTRG) supports the intrinsic development of human natural killer (NK) cells after bone marrow transplantation (120). When combined with orthotopic NB PDXs, these mice have allowed the identification of immune modulating functions in common between PDXs and patient tumors and suggest that the model is useful for immuno-oncology studies in general and in NB in particular (121). The use of PDO and stromal/immune cell co-cultures can be applied in vitro, where it has been very challenging to optimize culture conditions for multiple cell types over longer periods (122, 123). Co-cultures of NB organoids and peripheral blood mononuclear cells (from a healthy donor) were recently used to test a novel immunotherapy (124).

Innovative technological advances have suggested that microfluidics (lab-on-a-chip) and bioprinting may provide future systems for studying tumor cell and stromal/immune cell interactions. Functional short-time cultures of both tumor cells and immune cells in a microfluidic system have been reported for adult cancers (125, 126) and might be applicable also to pediatric cancers. Very recently, the first bioprinted, vascularized NB microenvironment on a fluidic chip was reported (127). Implantation of cell line-derived NB spheroids led to NB cell survival for two weeks and successful micro-vessel infiltration of the spheroids (127). A different study managed to establish PD NB organotypic slice cultures that could potentially preserve an intact NB tumor microenvironment (88). This study used a perfusion-based bioreactor to force medium through the tissue, thereby providing continuous nutrient delivery to the whole tumor.

Orthotopic NB PDXs retain spontaneous metastatic capacity in vivo to the lungs, liver, and bone marrow, mimicking the entire process from primary tumor growth to invasion and metastasis (37, 39). However, the TME is generally murine (38) and there are uncertainties about cross reactivity between human NB cells and the mouse TME. The presence of human mesenchymal stem cells can increase growth and metastasis of NB cells in vivo (128), suggesting species preference. Recent advances in tissue engineering have produced in vivo models of humanized bone (so-called ossicles) in mice. Implanted PDX-derived NB cells form osteolytic tumor lesions in the ossicles and display higher and faster engraftment rates than in mouse bone (129). This model could thus be valuable for the investigation of human NB growth and treatment responses in a humanized metastatic niche.

Further optimization of PD models to account for more patient-like microenvironmental factors both in vivo and in vitro is ongoing and will likely contribute to improved translatability.

The use of clinically relevant preclinical models is of immense importance in childhood cancers, such as high-risk NB, where the access to patient material is limited. PD models reflect the characteristics of the tumor of origin better than conventional in vivo and in vitro models. Nevertheless, ongoing efforts may further optimize their translational relevance. Existing NB PDXs and PD cultures have been – and continue to be – used to decipher therapy resistance and for target identification and drug testing. Future studies will need to investigate how PD models can be used to exploit phenotypic plasticity and NB cell states in preclinical studies to better benefit NB patients.

KA, KR, AA, AS, AM and DB wrote and reviewed the manuscript. KA and DB supervised the project. All authors contributed to the article and approved the submitted version.