The core treatment modalities for pediatric central nervous system (CNS) tumors are surgery, radiation, and cytotoxic chemotherapy. While effective for some patients, these modalities are inadequate for many CNS tumor types and can cause significant morbidity. Recently, with better understanding of the underlying molecular drivers of pediatric cancer, targeted therapy has emerged as a promising alternative, or adjunct, to traditional cancer treatment. Targeted therapy aims to disrupt specific molecular pathways that drive tumor growth and progression. The underlying principle is that by targeting specific molecules involved in the growth and spread of cancer cells, on-target effects will increase and damage to healthy tissues will be attenuated.

In this narrative review, we outline the key pathways implicated in pediatric CNS tumors and evaluate specific targets for therapeutic intervention. Using these molecular pathways as a framework, we present a primer on the clinical trial landscape for targeted therapies in pediatric CNS tumors by surveying key completed ( Table 1 ) and ongoing ( Table 2 ) trials. Finally, we discuss innovative approaches to employing targeted therapy, including combinatorial regimens and real-time drug screening for personalized treatment planning.

The mitogen-activated protein kinase (MAPK) signaling cascade is one of the first described and most extensively studied pathways in cell biology. MAPK signaling is complex with multiple upstream and downstream interactions with other major pathways involved in cell proliferation, differentiation and other metabolic signals supporting tumor sustenance and growth ( Figure 1 ). Direct downstream RAS signaling is composed of RAS–RAF–MAPK kinase (MEK) – ERK. In physiologic conditions, this cascade is activated by receptor tyrosine kinase (RTK), which triggers the GTPase-dependent RAS. Neurofibromin-1 (NF1) is a negative regulator of RAS, which in turn activates RAF kinase. RAF is a primary mediator of the MAPK pathway and is responsible for the sequential activation of downstream targets MEK1/2 and the transcription factor ERK 1/2. ERK1/2 transcriptionally regulates genes involved in proliferation and cell survival, including cAMP response element–binding protein, as well as the transcriptional regulator Myc-like (c-Myc) and nuclear factor kappa B (NFKB) (44).

The MAPK pathway is implicated in the vast majority of pediatric low grade gliomas (pLGGs), to the point that pLGG is thought to act as single pathway disease (45). Over 85% of pLGGs exhibit molecular aberrations of RAF, which ultimately upregulate the MAPK pathway (46, 47). Activating alterations of BRAF can occur as point mutations, in-frame deletions, or fusions with other kinases. The two most common alterations are the BRAFV600E mutation, caused by nucleotide transversion resulting in the substitution of valine (V) with glutamic acid (E) at position 600 (i.e., V600E point mutation) and tandem duplication resulting in BRAF-KIAA 1549 fusion (48–50). Amongst pLGGs, pleomorphic xanthoastrocytomas and gangliogliomas histologies are more commonly are associated with BRAFV600E, whereas pilocytic astrocytoma mostly harbor fusion of BRAF: KIAA1549 (51, 52). Although most data is in pLGG, 5-10% of pediatric high-grade gliomas also have MAPK pathway alterations (53). MAPK pathway activation has also been implicated in adamantinomatous craniopharyngioma (54), providing a target for multiple ongoing craniopharyngioma trials ( Table 2 ).

Receptor tyrosine kinases (RTKs) are a family of cell surface proteins which act as receptors for growth factors, hormones, cytokines, neurotrophic factors, and other extracellular signaling molecules. This family of receptors is divided into subfamilies including epidermal growth factor receptor (EGFR), platelet-derived growth factor receptor (PDGFR), fibroblast growth factor receptor (FGFR), insulin and insulin-like growth factor receptor (IGFR), vascular endothelial growth factor receptor (VEGFR), and hepatocyte growth factor receptor (HGFR/C-MET) (10, 73–77). Once activated, RTKs initiate a signal cascade primarily through two downstream pathways: RAS/MAPK/ERK and RAS/PI3K/AKT. Ultimately, these pathways result in cell proliferation, invasiveness, survival, and angiogenesis. Aberrations in RTKs are commonly found in both pediatric high- and low-grade gliomas and are therefore promising therapeutic targets for treatment (10, 73–77). The most common genetic RTK alterations in these tumors occur in the EGFR family, followed by altered PDGFR and MET tyrosine kinase pathways. Multiple RTK inhibitors have and are currently being developed and evaluated in clinical trials (10, 73–77).

High-grade gliomas have structurally and functionally abnormal vasculature. Beginning in the mid-1990s, studies showed that inhibition of vascular endothelial growth factor (VEGF) expression in GBM reduced vasculature formation and suppressed tumor growth (10, 74). The most widely used drug to target VEGF is the humanized monoclonal anti-angiogenic antibody bevacizumab, which is the first FDA-approved targeted treatment for recurrent GBM. Bevacizumab has showed improved progression-free survival (PFS) in GBM, but has not shown benefit to overall survival when used alone (14, 74). In pediatric neuro-oncology, bevacizumab has shown efficacy in combination with irinotecan in progressive or recurrent low-grade glioma (13, 93). Bevacizumab may also have specific application for optic pathway gliomas, where it has demonstrated a favorable association with visual outcomes (94, 95). The VEGFR inhibitor cediranib was investigated in a phase II clinical trial as monotherapy in recurrent GBM showing improved radiographic response at 6 months, but no overall survival benefit. The drug sunitinib, which targets PDGFRA/B in addition to VEGFR, showed preclinical promise, but it did not show improvement of PFS in patients with recurrent GBM in a phase II clinical trial (74, 77).

The PI3K/mTOR pathway is a signal transduction pathway involved in cell growth and proliferation and is another critical target for tumor-directed therapy in pediatric CNS tumors. Once activated at the cell membrane, phosphatidylinositol 3-kinase (PI3K) leads to the accumulation of PIP3, which in turn leads to phosphorylation and activation of Akt (96) ( Figure 2 ). Activated Akt inhibits the GTPase tuberin-hamartin (TSC) complex. Inactivation of the TSC complex disinhibits mTOR, initiating downstream promotion of cell proliferation and survival. Activation of PI3K, a critical entry point to the PI3K/mTOR pathway, can be achieved through activation of transmembrane proteins including RTKs (eg, FGFR) and insulin-like growth factor 1 receptor (IGF-1R). PI3K can also be activated directly by RAS, leading to signficant crosstalk between the PI3K/mTOR and MAPK pathways. Alterations in the PI3K/mTOR pathway are ubiquitous in human cancer and have been specifically implicated in pediatric high- and low-grade glioma (97).

Most efforts to target the pathway are focused on inhibiting PI3K or mTOR. For example, for subependymal giant cell astrocytoma (SEGA) associated with the tumor predisposition syndrome tuberous sclerosis, the mTOR inhibitor everolimus is now FDA-approved in the upfront setting based on a landmark phase 3 clinical trial showing its efficacy (15). Attempts to apply mTOR inhibitor monotherapy to other tumor types have been less successful. A phase 2 trial of everolimus monotherapy in recurrent or progressive pediatric LGG showed a partial response rate in only 2 of 23 (13%) patients (16), and a separate study demonstrated that PI3k/mTOR pathway activation did not correlate with response (98). Similarly low response rates were observed in a trial of recurrent NF1-associated LGG treated with everolimus (99).

The crosstalk between the MAPK and PI3K/mTOR pathways creates an appealing opportunity for combinatorial treatment approaches. The combination of everolimus with the MEK inhibitor trametinib is currently being evaluated in the clinical trial NCT04485559 for patients with pediatric gliomas (either low or high grade).

Alterations in the PI3K/Akt/mTOR pathway have also been identified as molecular drivers in H3K27-altered diffuse midline glioma (53). The PI3K inhibitor paxalisib has demonstrated efficacy in pre-clinical DMG models when used in combination with other agents (100, 101), prompting its incorporation into an ongoing clinical trial with using an adaptive platform design (PNOC022/NCT05009992). Preliminary data from this trial indicate that the combination of paxalisib and ONC201 is generally well-tolerated, with the most common treatment-related adverse events being myelosuppression, rash, mucositis, and colitis (17).

Cyclin-dependent kinases (CDKs) are protein complexes involved in cell cycle regulation. CDK4 and CDK6 are of specific relevance in cancer and promote the transition from G1 to S cell cycle phases through the phosphorylation of retinoblastoma protein (Rb) (102). Inhibition of CDK4/6 by the tumor suppressor protein P16 causes G1 cell arrest by inhibiting CDK4/6 activity (103). Homozygous deletions of CDKN2A, the gene that codes for P16, result in unregulated cell division through disinhibition of CDK4/6 and have been implicated in multiple cancer types, including in BRAF V600E gliomas. Indeed, CDKN2A deletion is an independent negative predictor of outcome in V600E-mutant pLGG (51). CDKN2A homozygous deletions are also common in IDH-WT HGGs, can occur independently of BRAF alterations, and are associated with a worse clinical outcome (104).

CDK4/6 inhibitors are a class of small-molecule drugs designed to recapitulate the physiologic function of P16 that is lost with CDKN2A deletion, thereby promoting cell cycle arrest. Ribociclib is a CDK4/6 inhibitor with good CNS penetration (105). In a phase I/II clinical trial for DIPG, ribociclib was well-tolerated and associated with increased necrotic tumor volume but did not provide significant clinical benefit (18). A putative explanation for failure of CDK4/6 inhibitor monotherapy is reversal of cell-cycle arrest when the drug is withdrawn. This has led to efforts for combinatorial therapies as such as the TarGeT-A trial, which combines ribociclib with the mTOR inhibitor everolimus ( Table 2 ).

Epigenetic changes involve modification to gene expression, rather than alteration to genes themselves. A complex interplay between DNA and histone modification results in a dynamic switching of genes “on” and “off”, as well as modulation of the level of gene expression (106). In the nucleus, DNA is packaged as chromatin. The basic structural unit of chromatin is the nucleosome, which is a coil of DNA wrapped around a histone core. The histone core is an octamer made up of 8 proteins, 2 each of H2A, H2B, H3, and H4. Each histone protein has an amino acid tail, which is relevant for gene expression, and are abundant in lysine and arginine. H3 is of specific relevance to pediatric brain tumors.

One of the ways that transcription is regulated is through modulating the wrapping and unwrapping of DNA around histone octamers. This is done through modification of the DNA itself, through methylation, and through modification of the histone amino acid tails - byacetylation and methylation. Histone acetylation results in a relative negative charge. DNA is also negatively charged, so when there is histone acetylation, there is loosening of the DNA coil around the histone, facilitating transcription. In turn, histone methylation creates a docking site for chromatin-associated proteins. Histone methylation can result in activated or repressed chromatin, depending on the site. Methylation of specific lysine residues on the amino acid tails of H3 and H4, including H3K9, H3K27, and H4K20 result in transcriptional suppression. Of particular interest in CNS tumors is trimethylation of H3K27, abbreviated H3K27me3, which results in transcriptional suppression by Polycomb Repressive Complex 2 (PRC2). By contrast, methylation of other lysine residues, including H3K4, H3K36, and H3K79 results in transcriptional activation.

In contrast to adult HGG, histone modifications are seen in about 80% of pHGG. The most common somatic alteration seen in pHGG are variants resulting in the presence of a methionine (M) instead of a lysine (K) at position 27 on the amino acid tail of histone H3, H3K27M. This results in chromatin remodeling and loss of trimethylation of H3K27 (H3K27me3) with subsequent transcriptional activation and tumorigenesis. The H3K27M mutation is a hallmark molecular finding in diffuse midline glioma (DMG) (53). DMG is associated with dismal clinical outcomes and multiple chemotherapeutic regimens have been evaluated without significant improvement on survival. In the past decade, there has been rising interest in the small molecule ONC201/dordaviprone, which showed signs of efficacy in adult patients with progressive H3K27M DMG (107). Recently, evidence has emerged indicating that the anti-tumor effect of ONC201 in DIPG is through disruption of the TCA cycle within mitochondria, leading to an inhibitory effect on histone lysine demethylases and increase in genomic H3K27me3 (20). A pooled analysis of two clinical trials (NCT03416530/ONC201-014 and NCT03134131/ONC201-018) evaluating ONC201 in non-recurrent H3K27M-mutant DMG identified a modest improvement in median overall survival compared to historical controls (OS 21.7 months vs 12.0 months, respectively) (20). ACTION (NCT05580562) is an ongoing randomized trial comparing ONC201 to placebo in newly-diagnosed H3K27M-mutant DMG (39). Another histone alteration involves replacement of glycine (G) with valine (V) or arginine (R) at position 34 on histone H3.3. G34R and G34V mutants result in transcriptional activation and are commonly seen in pediatric-type hemispheric high grade gliomas (53).

The Hedgehog/Glioma-associated oncogene homolog (HH/GLI) pathway controls various processes during embryonic development including cerebellar maturation and tissue regeneration. Sonic hedgehog (SHH), a component of the HH/GLI pathway, is critical for normal cerebellar development, but constitutive activation of SHH signaling results in tumorigenesis (8, 11, 114, 115). Hedgehog pathway signaling involves the 12 pass-transmembrane receptor, PATCH (PTCH1) and when bound releases its inhibition of smoothened (SMO), a protein that activates the downstream portion of the pathway by binding to the cell fusion inhibitor called suppressor of fused (SUFU) and induces nuclear translocation of activators Gli1 and Gli2 and a repressor Gli3 ( Figure 3 ). The Gli proteins regulate the expression of downstream targets including Cyclin D and MYC involved in cell survival, proliferation, and differentiation (8, 11, 114). Mutations in this pathway drive the initiation and progression of the SHH subtype of medulloblastoma as well as other solid tumors. Therefore, various agents have been developed targeting SHH, SMO and Gli1 and Gli 2 (8, 11, 114).

SHH inhibitors atezolizumab and avelumab prevent the binding of SHH to PTCH1 and are being studied in the preclinical setting. Vismodegib and sonidegib are SMO antagonists and were approved by the FDA in 2012 and 2015, respectively, for the treatment of advanced or metastatic basal cell carcinoma (8, 11, 114). In initial an early Phase I study with vismodegib only 1 of 3 patients with SHH activated medulloblastoma had antitumor response. This has been corroborated subsequent SMO inhibitor trials (22, 116), and may be explained by intra-group heterogeneity within the SHH subgroup. Vismodegib was also found to induce irreversible growth plate fusion with prolonged exposure (117). The drugs GANT61 and HPI-1 target Gli1 and Gli2 but have not been tested in medulloblastoma (8, 114). Resistance to SMO inhibition was first described in 2009 and the patient with medulloblastoma had mutations in the SMO gene. It has since been described that mutations in SMO lead to both de novo and acquired resistance and neither vismodegib or sonidegib are effective in certain mutations (114). Hyperactivation of GLI has been shown to lead to chemoresistance and radiation resistance in multiple cancers including medulloblastoma. This increase in the pathway ultimately suppresses the anti-tumor response from the immune system (8, 11, 114, 115, 118, 119).

The development of targeted therapies has been driven by a deepening understanding of the molecular underpinnings of pediatric CNS tumors. This expanding knowledge base has enabled the identification of specific molecular targets for therapeutic intervention. Nevertheless, success has been uneven across the pediatric neuro-oncology landscape. The greatest advances have been seen in low grade glioma, a largely single-pathway disease and currently the only pediatric CNS tumor with FDA-approved targeted therapies – dabrafenib/trametinib for upfront treatment of BRAF V600E mutant LGG and tovorafenib for BRAF-altered LGG in the recurrent setting. A common theme in this review is that monotherapy with targeted agents is rarely sufficient for durable treatment response. As illustrated in Table 2 , several current clinical trials are using combinatorial approaches to achieve the sweet spot of optimizing efficacy while mitigating toxicity. These include employing multiple targeted agents addressing discrete implicated pathways (eg, everolimus/trametinib for LGG and HGG, NCT04485559), combining conventional treatment modalities with targeted agents, and using molecular therapy along with immunotherapy to create a synergistic anti-tumor effect.

Other trials are using a personalized medicine approach via a “molecular tumor board” that incorporates a patient’s individual tumor molecular profile to determine a treatment plan. In addition to providing rationale treatment recommendations for a given tumor’s molecular profile, this centralized approach has appealing equity implications by improving access to specialized care (120). The PNOC003/NCT02274987 trial used a molecular tumor board to recommend personalized treatment regimens for children with DIPG (121). The multidisciplinary tumor board considered clinical and genomic data before providing a consensus recommendation of up to 4 FDA-approved drugs to be included in the treatment regimen within 21 days of surgery (122). Thirty-eight participants were enrolled, 28 of whom were evaluable by the tumor board (123). Nineteen (68%) patients followed tumor board treatment recommendations, supporting the feasibility of the approach. There was no difference in survival for those who followed tumor board recommendations compared to those who did not. Nevertheless, experience from this trial informed the development of PNOC022, on ongoing platform trial that includes an arm for molecularly-guided combinatorial molecular therapy for DMG.

Another innovative trial design uses real-time drug screening with live tumor tissue to provide the molecular tumor board with more robust information in developing an individualized treatment plan (124). For example, the ongoing trial PNOC027/NCT05057702 conducts high-throughput drug screening on freshly isolated tumor cells of children with relapsed medulloblastoma. The platform evaluates responses to 232 clinically-available compounds. A recent preliminary report of 9 patients enrolled on the study demonstrated the feasibility of the approach: 8 of the 9 patients successfully completed real-time drugs screening, with a median turnaround time of 7 days from sample receipt (125).

While outside the scope of this review, there is also considerable interest in optimizing delivery of targeted therapies beyond traditional oral, intravenous, or intrathecal routes. For example, convection enhanced delivery, low intensity focused ultrasound, and nanoparticle-based therapies have been employed as tools to circumvent the blood-brain barrier and modulate the immune microenvironment (126–129).

An underlying impetus for developing targeted therapy, in addition to increasing treatment efficacy, is avoiding the systemic toxicities seen in traditional cytotoxic chemotherapy. With notable exceptions (eg, bevacizumab), targeted therapies have the additional benefit of availability of oral formulations, sparing patients the need for durable central venous access or hospital admissions for drug administration. Many targeted therapies are also less immunosuppressive compared to cytotoxic chemotherapies, decreasing the risk for serious infections. Nevertheless, as presented in Table 1 , these drugs are not without adverse effects. Grade 1/2 “nuisance” toxicities including mucocutaneous and gastrointestinal effects are common, and serious effects on cardiac, liver, and bone marrow function have been observed. Additionally, due to their novelty, long-term effects are not well characterized.

Although promising, molecular targeted therapy remains in its infancy. Questions remain regarding the optimal sequence and duration of therapies. Strategies to address “rebound” phenomena, in which early growth is observed after therapy discontinuation, remain unsettled. Finally, while immediate toxicity profiles are generally favorable compared to traditional cytotoxic chemotherapy, long-term effects of targeted therapies on growth, fertility, and cognitive function are not yet known.