Pediatric low-grade glioma (pLGG, WHO grade 1 or 2) is the most common brain tumor class in children (1–4). Although in general, prognosis, overall survival, and outcomes following treatment are better for pLGG than adult gliomas, the broad spectrum and heterogeneity of pLGG histopathology make these tumors challenging to treat (1, 3–6). For example, these gliomas may differ in their brain locations, histological spectrum, cell(s) of origin, genetic alterations, and the tumor microenvironment milieu (4). Additionally, since many pLGGs undergo senescence towards juvenile period or adulthood, a lot of pLGG patients end up not getting surgeries, making it harder to obtain tumor samples to be studied. For tumors located in brain areas accessible by surgery such as the cerebellum, total resection is a viable option and may be curative, but for less accessible tumors such as those in hypothalamus, midline areas, and optic pathway, additional treatments such as radiation and cytotoxic chemotherapies may be necessary (1, 7). Even with gross total resection, recurrence may still happen, and additional treatments may lead to allergies and long-term side effects (5, 7, 8). Furthermore, since pLGGs are chronic diseases, these children sustain significant life-long morbidity and general reduction in quality of life. Due to the complexity of the molecular landscape of pLGG, continuous efforts are directed towards characterizing the genomic and epigenomic alterations in these gliomas, and developing effective preclinical models are paramount in making informed decisions on therapeutic options.

Advances in sequencing technologies have aided in classifying the diverse spectrum of pLGG histopathology. The various types of pLGG that are more often encountered are summarized in Table 1 , including their locations and most common genetic alterations. pLGGs can be glial or mixed glial-neuronal tumors. Pilocytic astrocytomas (PAs) predominate in children younger than 15 years of age and is considered the most common type of pLGG (1, 7). However, it is still very challenging to accurately diagnose and stratify pLGG patients due to the broad spectrum of the histopathology and the complex molecular landscape (1, 4, 9, 14, 18). Moreover, in many cases, there are overlapping morphologies between the different groups of pLGGs, and even the more circumscribed tumors can contain infiltrative areas (4, 9, 14). The majority of pLGGs show convergence in MAPK (Mitogen-activated protein kinase)/ERK (Extracellular signal-regulated kinase) pathway alterations, as illustrated in Figure 1 and the next section will discuss in more details the genetic alterations found in various pLGGs based on advancement in molecular profiling studies (10, 11, 19).

As sequencing technologies advance, more mutations are continuously discovered in various types of pLGGs, complicating their molecular landscape. Moreover, different mutations may cause the different tumors to respond variably to targeted therapeutic agents, making risk stratification and development of therapies for pLGG difficult. Efforts to develop preclinical models of pLGGs are mainly focused on the more common molecular alterations including but not limited to KIAA1549-BRAF, BRAF V600E, and NF1 mutations, but many have not been successful. Table 2 lists the in vitro models that have been developed for preclinical testing and their varying degrees of success. One obstacle in generating viable pLGG patient-derived cell lines is the lack of availability of patient tissues. Tumors located in areas difficult to surgically excise are often not biopsied, making it difficult to obtain sufficient samples from these patients to generate cell lines (4). Another major obstacle in generating viable cell lines for non-NF1-pLGG (e.g. PA) is the oncogene-induced-senescence (OIS) phenomenon [ (48, 49), Figure 2 ]. Expression of BRAF V600E and KIAA1549-BRAF fusion led to MAPK pathway activation and subsequent OIS in human cortical neurospheres, human immortalized astrocytes, fetal astrocytes, and low-passage primary PA cultures [ (48, 49), Table 2 ]. The OIS phenomenon, however, may be the rationale for the lack of progression of PA to higher-grade tumors, in the absence of additional cooperating mutations (48, 49). Factors like senescence-associated secretory phenotype (SASP) and microRNAs have been implicated in senescence of pLGG cell lines (58, 59). On the other hand, inactivation of CDKN2A gene encoding the tumor suppressor p16INK4a might facilitate escape from senescence in these cells, explaining the association between p16INK4a loss and worse outcomes in PA patients (48, 49).

Attempts to generate pLGG patient-derived cell lines yielded varying success rates. Most of them could only be maintained as short-term cultures [ (52, 54, 60), Table 2 ] and/or they acquired additional alterations in culture that made them unrepresentative of the original tumors (47, 50). Bax et al. (2009) conducted molecular and phenotypic characterization of Res259 and Res186 cell lines derived from human pediatric astrocytomas to incorporate them in preclinical testing. Res259 and Res186 were derived from DA and PA patients, respectively. Although their immunophenotypes closely resembled low-grade lesions, both lines harbored genetic alterations reminiscent of higher-grade gliomas [ (47), Table 2 ].

As part of the Pediatric Preclinical Testing Program (PPTP), 2 astrocytoma cell lines were generated, BT-35 cells which had 2 to 5 copies of wild-type (WT) BRAF and BT-40 cells which had 5 copies of activated BRAF V600E [ (50, 51), Table 2 )]. BT-40 was further characterized as grade 2-3 astrocytoma as these cells lost CDKN2A and wild-type P53 (50, 54, 61).

An important milestone in the development of patient-derived PA cell lines was when Selt et al. (2017) transduced cells derived from a 2-year-old PA patient with Dox inducible system coding for simian vacuolating virus 40 large T antigen (SV40-TAg, Figure 2 ). Expression of SV40-TAg inhibited 2 pathways critical for OIS induction and maintenance and enabled the generation of long-term PA cell line which could be especially useful for preclinical drug testing. This KIAA1549-BRAF-expressing cell line was named DKFZ-BT66 [ (53), Table 2 )].

Chiacchiarini et al. (2021) reported additional pLGG cellular models which included primary cells derived from Grade I pLGG tumors. 3 of 9 tumors were infratentorial PAs, 1 of 9 tumors was infratentorial ganglioglioma, 3 of 9 tumors were supratentorial DNETs, and 1 of 9 was supratentorial angiocentric gliomas, and 1 of 9 was supratentorial ganglioglioma (54). Although these patient-derived cells also underwent senescence after about 30 days of culture, they were not genetically modified and thus were more representative of the tumors they were derived from [ (54), Table 2 )]. Hence, these cells could be useful for short-term in vitro experiments.

The overexpression of hTERT (human telomerase reverse transcriptase, the catalytic subunit of telomerase) was proposed to counteract OIS in pLGG cell lines so they regain their proliferative potential for long-term cultures [ (55), Figure 2 )]. Another method to extend lifespan of pLGG cell lines was proposed by Yuan et al. (2021). Instead of using feeder layer of 3T3 fibroblasts as first described (62), Yuan et al. (2021) co-cultured the pLGG cell lines in conditioned media from irradiated 3T3 cells and Rho kinase (ROCK) inhibitor Y-27632 [ (56), Figure 2 )]. They found that these culture conditions led to reversible blockage of senescence, increased proliferation, and allowed these cell lines to propagate longer in culture while still maintaining signature genetic changes of the original tumors (56). This method was tried on multiple pLGG cell lines including 4 PAs, 2 gangliogliomas, 2 anaplastic gliomas, 1 anaplastic PXA and others. One cell line derived from NF1-PA (JHH-NF1-PA1) and another from BRAF V600E anaplastic PXA (JHH-PXA1) patients exhibited growth sufficient for preclinical testing in vitro. More recently, Rota et al. (2022) developed an ex vivo culture system that utilized synthetic extracellular matrices (sECMs) to partially mimic properties of extracellular matrix in vivo. This culture method promoted proliferation and enabled propagation of several patient-derived pLGG cells for up to 1 month ex vivo (57). Continuous efforts are made to generate more pLGG patient-derived cell lines that can be cultured long enough for preclinical testing and still resemble the original tumors, but these studies have made major advancements in the field.

A recent report by Anastasaki et al. (2022) utilized human induced pluripotent stem cells (hIPSCs) to generate LGG xenografts harboring NF1 loss and KIAA1549-BRAF fusion. First, they engineered different hIPSC lines with patient-derived germline NF1 mutations and with KIAA1549-BRAF fusion, then they differentiated these hIPSCs to multipotent human neural stem cells capable of generating both neuronal and glial lineage cells (hINPCs). These hINPC lines exhibited increased proliferation, and KIAA1549-BRAF-expressing hINPCs had increased MAPK pathway activation (68). Since injections into the mouse optic nerve, the most common site for NF1-pLGGs, caused a lot of tissue damage, NF1-null hINPCs were injected into the brainstem of immunocompromised Rag1^-/- mice, which was the second most common site of NF1-pLGGs (68). Brainstem injections of NF1-null hINPCs and cerebellar injections of KIAA1549-BRAF-hINPCs formed LGGs at 1 month post injection. Moreover, these lesions exhibited many histopathologic features of human pLGGs. Mice with hIPSC-derived LGGs did not exhibit increased mortality, and the lesions had similar proliferative indices even though the sizes grew over time (68). These mirrored clinical observations in pLGG patients.

Orthotopic transplantation of NF1-null and KIAA1549-BRAF-expressing hIPSC-derived glial restricted progenitors (iGRPs) and oligodendrocyte progenitors (iOPCs), but not hIPSC-derived terminally differentiated astrocytes formed LGGs in Rag1^-/- mice (68). This was reminiscent of the genetically engineered NF1-OPG mice in which neuroglial progenitor cells could serve as initiating cells of optic glioma formation (82, 92). Interestingly, iGRPs gave rise to tumors resembling optic pathway and brainstem gliomas, while iOPCs gave rise to tumors similar to many cerebellar human PAs (68).

Furthermore, Anastasaki et al. (2022) demonstrated that formation of LGGs required CD4+ T cell depletion and reduced astrocytic Cxcl10 expression. So LGGs could form in NOD/SCID, CD4-deficient, CD4/CD8-deficient mice, but not in CD8-deficient mice or other strains lacking expression of microglia or T cell chemokine receptors (68). They also showed that primary human PA cell lines including that with an NF1-PA (JHH-NF1-PA) and that with sporadic PA (Res186) could form LGGs in Rag1^-/- and Cxcl10^-/- mice (68). Although additional work is still required to create preclinical models that more closely recapitulate patient tumors, this model provides an important milestone in the development of a humanized pLGG orthotopic xenograft model.

Another in-vivo model that has been utilized to study pLGGs is the zebrafish. It is cost-effective, has short experimental timeframe, and enables rapid investigation into tumor growth, invasion, metastasis, and drug screening (56, 102–105). Fewer cells are needed to establish tumor xenografts, and the availability of transparent transgenic zebrafish or their translucent larvae made it possible to track tumor cells in real-time (102–104, 106). Several transgenic zebrafish models have also been generated to interrogate glioma pathogenesis. For example, using the Gal4-UAS system, Mayrhofer et al. (2017) generated zebrafish brain tumor model that expresses oncogenes that activate MAPK and PI3K signaling in neural progenitor cells and discovered that activation of YAP signaling pathway promotes development of aggressive brain tumors (107). Luo et al. (2021) used CRISPR/Cas9 to establish transgenic zebrafish lines that express mutated NF1, Rb1 or TP53 under GFAP promoter and showed that various combinations of NF1, TP53, and/or Rb1 mutations can induce gliomas of different grades and phenotypes (108). Using transgenic zebrafish, Lee et al. (2010) demonstrated that NF1 knockdown increased ERK signaling and increased OPC proliferation in the developing spinal cord (109). Orthotopic xenografts and immunodeficient zebrafish models have also been developed for brain tumors (110–113). Although many of the zebrafish xenograft models used glioblastoma cells, Yuan et al. (2021) injected JHH-NF1-PA1 cells into the midline in the optic tectum of zebrafish at 2 days post fertilization, and they found that the pLGG cells survived over 6 days and migrated in the brains of larval zebrafish before their adaptive immune system matured (56). Sigaud et al. (2023) also used zebrafish embryos to evaluate therapeutic options for pLGG xenografts generated by injecting DKFZ-BT66 and BT40 cells (105). While zebrafish models certainly have advantages over mouse models in terms of time and cost efficiency and are useful for rapid drug screening, development of more models and/or a combination of different types of models will be necessary to recapitulate the complexity of human pLGGs.

Isakson et al. (2018) used Transcription activator-like effector nucleases (TALENs) flanking a known NF1 nonsense mutation NF1^R1947 to transfect fetal Ossabaw minipig fibroblasts. NF1 minipigs were generated through chromatin transfer and subsequent breeding (114). NF1 mutant allele exhibited germline transmission with Mendelian frequency and no evidence of reduced fitness (114). NF1 minipigs displayed features of NF1 patients such as skin abnormalities, neurofibromas, underwent biallelic inactivation of the NF1 gene, and they also developed OPG (114). Therefore, the generation of these NF1 minipigs paved a path for the field to interrogate NF1-related molecular pathogenesis, explore therapeutic options, and conduct preclinical testing in large animal models that hopefully present closer resemblance to human patients.

Although massive number of efforts have been directed at developing preclinical in vitro and in vivo models of pLGGs, there are many challenges to these endeavors. The generation of in vitro pLGG models has been hampered by many factors. First, the lack of availability of patient tissues as some pLGGs located in less accessible areas such as NF1-OPGs are rarely biopsied (4, 7), although a few NF1-PA cell lines are available (115). Second, pLGG samples obtained from young children as patients are usually small, hence only small number of cells can be obtained to culture. Third, the intrinsic slow growth and benign behavior of these tumors coupled with OIS made it difficult to grow these tumor cells in vitro (48, 49, 52, 54, 60). Several methods were developed to bypass OIS through genetic modifications of the pLGG cells so that they can be propagated long enough in culture (53, 55), but these methods might generate in vitro model systems that incompletely reflect the genetic/epigenetic background of the primary tumors. Nevertheless, as more research is being done into developing different ways to extend the lifespan of pLGG cells in vitro, certain methods such as conditional reprogramming culture conditions or using hIPSC-derived hINPCs might provide plausible solutions to generating viable in vitro models (56, 68).

Two main groups of pLGG in vivo models have been generated by several labs with varying degree of success: patient-derived xenograft models of mostly non-NF1-pLGGs (50, 51, 60, 65, 66, 116) and genetically engineered mouse models mostly for NF1-pLGGs especially NF1-OPGs (70, 72, 76–79, 82). However, some have also attempted to generate NF1-pLGG patient-derived xenograft models (68, 83) and transgenic non-NF1-pLGG model (67).

Genetically engineered NF1-OPG models have been extremely useful to investigate the pathogenesis of NF1-OPG, which could somewhat be extended to general NF1-pLGGs. In these models, tumors were readily detectable as they arose in predictable locations with near 100% penetrance, the histopathological features of these tumors resembled those in the patients, the initiating event (NF1 inactivation) was known, thus these models recapitulate many genetic and cellular abnormalities seen in NF1-OPG patients (94). Moreover, these tumors arose in immunocompetent mice allowing the interrogation into the tumor-immune axis that was necessary in NF1-OPG formation.

Nonetheless, there are many challenges presented to researchers in developing these pLGG in vivo models, and these reasons might vary depending on whether the pLGG is associated with NF1. For example, NF1-OPG formation depends on many factors including: temporal and spatial specificity (77, 78, 80, 91, 92), cell-of-origin (72, 76–78, 82, 83, 92), interaction with microenvironment (87–91, 93) and other signaling pathways (79, 84, 85, 117), nature of the NF1 mutation (97), among others. Meanwhile, the challenges in generating non-NF1 pLGG models include the low proliferation capacity and the specific permissive tumor microenvironment necessary for glioma formation (7). Interestingly, NF1-OPG requires NF1 heterozygous cells along with multiple different types of cells in the microenvironment to grow (77, 87–91, 93), whereas NF1 and non-NF1-pLGG xenografts needed loss of specific T-cell population to form (68). Consequently, the combination of patient-derived xenograft models and transgenic models are required to comprehensively interrogate the molecular pathogenesis of pLGGs and to assess potential immune response to therapeutic agents.

Although most in vitro and in vivo models that were generated often expressed BRAF and NF1 mutations, there are other genetic alterations that have been highlighted as hallmarks of certain types of pLGGs. Some reports proposed that FGFR1 mutation may be a relevant prognostic marker in PAs as in some cases FGFR1 mutations were associated with more adverse outcomes in patients (42). FGFR1 hotspot mutations were also relatively frequent in pLGGs especially in mixed neuronal-glial tumors without known genetic drivers, providing an additional way to classify these tumors (118). Moreover, Egbivwie et al. (2019) generated 5 grade 1 PA patient-derived cell lines and found that FGFR1 overexpression alone was able to increase tumor cell migration and drive tumor progression. They also found that there was higher expression of membranous phosphorylated FGFR1 in grade 2 tumors, so the presence of pFGFR1 could be associated with malignancy and tumor grade (119). Future research could add more focus into generating pLGG models that interrogate the pathogenesis of rarer alterations such as FGFR, NTRK, and MYB mutations. Development of pLGG models could also consider epigenetic events that might promote tumor survival, maintenance and/or progression as much less is known about the role of epigenetics in pathogenesis of pLGGs (4).

Due to the difficulties in generating long-term cultures of pLGG patient-derived cell lines, more alternative methods could be explored. For example, since there is such strong dependence of pLGG formation in vivo on microenvironment, certain factors could be added into the cultures such as cytokines or tumor cells could be co-cultured with other cell types to provide more supportive microenvironment. The relatively successful sECM method to culture pLGG cells demonstrates that elements in the microenvironment will be necessary to generate viable in vitro pLGG models (57). There is also a need to develop more sophisticated technologies for genetic manipulation of tissues and cells. Indeed, emerging somatic transgenic in vivo mouse (120–123) and 3D cerebral organoid-based human glioma models (124) genetically manipulated using electroporation also hold promise for more flexible and renewable modeling. These models can employ plasmid or mRNA-delivered transgenes, and/or CRISPR/Cas elements to enable more rapid recapitulation of mutations in diverse spatiotemporal contexts but have been mostly employed to generate HGG models (120–123).

A truly useful pLGG model would ideally recapitulate the molecular pathogenesis of human pLGGs along with the dependency of these tumors on the microenvironment milieu and the inevitable activation of the MAPK pathway. However, despite the constant challenges, many important milestones were achieved in the development of in vitro and in vivo pLGG models. The combination of advanced sequencing technologies and prognostic methods and in vitro and in vivo models should be utilized to comprehensively investigate the pathogenesis of this highly heterogeneous group of pediatric brain tumors and identify viable therapeutic options.

After the completion of this review, a related work was published (125). Table 5 lists the abbreviations commonly used in the manuscript.

GY: Writing – original draft, Writing – review & editing. JB: Funding acquisition, Resources, Supervision, Writing – review & editing.