Adoptive cell therapy uses immune cells to target cancer epitopes, either by selecting tumor-targeting immune cell subsets, or by genetically modifying them. The current most successful modality, chimeric antigen receptor (CAR) T cells, uses the latter approach. CARs are synthetic molecules that combine the antibody recognition properties (non-MHC restricted) of a B cell and the effector functions of a T cell. They consist of two domains, an extracellular binder (typically a single chain variable fragment) against a specific cell surface tumor epitope, and an intracellular signaling domain that induces target killing and T cell proliferation (62–66). The CAR transgene is introduced into autologous or allogeneic T cells, enabling them to identify and eliminate any cells that express the target antigen. This includes tumor cells (on-target/on-tumor activity) and normal cells expressing the target epitope (on-target/off-tumor activity). In addition, CAR T cell activity and expansion is typically accompanied by significant systemic cytokine release (62–66). Other types of adoptive cell therapy that have been tested in children include genetically modified T cell receptors (TCRs) and Natural Killer (NK) cell therapy. Bispecific antibodies, which serve as a linker between immune cells and cancer targets, are covered in the antibodies section of this review.

CAR T cell therapy has revolutionized cancer treatment by inducing durable remissions in pediatric patients with relapsed/refractory B cell leukemia (B-ALL). Several studies have been published using CAR T cells for treatment of pediatric patients with relapsed/refractory B cell hematologic malignancies and have demonstrated impressive complete remission rates of 60-90% (67–72). The pivotal phase II global registration trial for CD19 targeted CAR T cells led to the historic, first pediatric FDA approved CAR product, tisagenlecleucel, for children and young adults up to age 25 with relapsed/refractory B-ALL (68, 73). Common toxicities (incidence > 20%) that occur post-CAR therapy in pediatric patients with B-ALL include CRS, neurotoxicity, infection, hypogammaglobulinemia, and fatigue (67–72, 74).

Neurotoxicity, now known as immune effector cell-associated neurotoxicity syndrome (ICANS) (15), is a common adverse event associated with CAR T cell therapy, with variable incidence of 7%-72% in published trials using CD19, CD22, or CD19/22 CAR T cells in children and young adults with hematologic malignancies (67–72, 74–77). In a recently published donor-derived CD7 CAR T cell trial for T cell ALL (T-ALL), incidence of ICANS was 15%, and all cases were grade 1 and associated with CNS disease (78). More severe cases of ICANS (≥ grade 3) have been described in up to 20% of pediatric patients treated with CAR T cell therapy. Clinical manifestations of ICANS vary, with most common presentations including depressed levels of consciousness, confusion, headache, tremors, and language disturbances (67–72, 74–77). More severe presentations of encephalopathy, seizures, and coma have occurred in up to 20% of patients, with rare cases of fatal cerebral edema (77, 79, 80). The incidence of seizures also varied significantly across studies, 0-20%, with most seizures being generalized tonic-clonic (67–72, 74–77). ICANS is often preceded by CRS onset, typically presenting within 7 days post-CAR infusion.

Acute neurocognitive functioning and patient reported outcomes measures have also been evaluated in pediatric patients receiving CAR therapy. Data collection in this patient population is feasible, with stable neurocognitive function and clinically meaningful health-related quality of life improvements demonstrated over time (17, 81).

As the application for CAR T cells continue to broaden, so too may the toxicity profile. Several hematologic CAR T cell trials in pediatrics are ongoing including those targeting CD5 and CD7 in T-ALL, and CD33 and CD123 in AML. Additionally, combinatorial CAR constructs with CD19, CD22, and/or CD20 are currently in clinical trials.

The pathophysiology of ICANS post-CAR therapy remains under active investigation. Endothelial cell activation, vascular and blood brain barrier disruption, vascular leak, systemic inflammatory response with CRS, and off-tumor/on-target toxicity are all possible mechanisms by which neurotoxicity can occur in pediatric patients (10, 76, 77, 82, 83). Indeed, the highest risk of ICANS is typically seen during the acute CAR T cell expansion, which suggests that the systemic inflammatory surge drives cytokine mediated neurotoxicity, rather than a local CNS cytokine production (77). Mouse models have suggested possible roles for IL-1 and GM-CSF signaling underlying the development of neurotoxicity (84, 85).Additionally, this systemic inflammation can also lead to endothelial activation, astrocyte injury, and increased permeability of the blood brain barrier (76, 77). However, an association of CAR T neurotoxicity with elevated Angiopoietin-2 levels, indicative of endothelial activation, was shown in adults (10) but not children (77, 86). Increased CSF glial fibrillary acidic protein (GFAP), and S100 calcium-binding protein (S100b), were found to be elevated in the acute setting post-CAR, indicating astrocyte injury and activation, which could lead to osmotic dysregulation in the brain potentially causing or contributing to cerebral edema (77, 83). In a mouse model, blockade of brain capillaries by circulating leukocytes was associated with the development of neurotoxicity (87).CAR T cell trafficking to the CNS has been demonstrated in several pivotal pediatric trials, but presence or quantity of CAR T cells in the CSF does not predict whether ICANS will occur (67, 68, 70, 88, 89). In fact, CAR T cells are detected in the CSF of patients with and without ICANS. Recently, a study demonstrated CD19+ mural cells surrounding the endothelium of the brain, suggesting that off-tumor/on-target toxicity of CD19 CAR T cells can contribute to neurotoxicity (90). However, this finding needs to be cautiously interpreted as ICANS is seen with CAR T cells targeting other antigens (B cell maturation antigen, or CD22) and not all patients who receive CD19 directed CAR T cells develop neurotoxicity.

The presence and severity of cytokine release syndrome is the strongest and most consistent risk factor for ICANS in pediatric CAR T cell patients (68, 76, 77, 91). There may also be additional risk conferred by pre-existing neurologic comorbidities (76, 91), including abnormal baseline brain MRI (77). Differences in demographics, CNS disease status, treatment history including prior brain radiation and bone marrow transplant, bone marrow disease burden at the time of CAR infusion, lymphodepletion regimen, or CAR T cell dose was not associated with development of ICANS in pediatric patients (70, 77, 91). It is important to note that disease burden, lymphodepletion regimen, and CAR T cell dose are associated with the development of ICANS in adult patients with hematologic malignancies, which likely affect the incidence and severity of CRS (10, 92, 93).

The American Society for Transplantation and Cellular Therapy (ASTCT) grading criteria are now used to grade ICANS (15). ICANS grading incorporates a 10-point encephalopathy score for children >12 years and adults (ICE) or the CAPD score for children <12 years, as well as additional neurologic domains including level of consciousness, seizures, motor symptoms, and signs of increased intracranial pressure (ICP) or cerebral edema. Grading is typically based on the most severe symptoms present in each category. Treating medical teams need to perform detailed past medical histories including evaluation of any prior treatment-related neurologic toxicities and neurologic risk factor comorbidities (e.g., seizure disorder) in each patient prior to receipt of CAR T cells. Baseline neurologic examination should also include ICE or CAPD scores. In patients with a history of neurologic adverse events, CNS disease, or focal findings on exam, baseline neuroimaging should be considered. Additionally, for those deemed high-risk (e.g., those who have a prior history of seizures), or patients receiving CAR products that have high neurotoxicity incidence, anti-seizure prophylaxis prior to receipt of CAR T cells should be initiated. Post-CAR T cell infusion, patients should be monitored for the development of neurotoxicity with routine neurologic exams inclusive of ICE or CAPD scores, especially once cytokine release syndrome has started. In patients who develop ICANS, work up should be performed based on the severity of symptoms present, with routine labs, neurology consult, neuroimaging, EEG, and evaluation for infectious etiologies, with consideration for a lumbar puncture if deemed safe (13).

Based on the clinical data that has emerged from CD19-directed therapy in pediatric and young adult patients receiving CAR T cells, management of ICANS remains largely supportive, with administration of anti-pyretics and/or antiseizure medications. Although tocilizumab has demonstrated remarkable results in abrogating CRS severity, it has not proven beneficial in treatment of ICANS (10, 67, 68, 77, 92, 94). Based on ASTCT grading criteria, use of corticosteroids for ICANS should be considered in patients with grade 2 ICANS, and is recommended for patients who have grade 3 or grade 4 ICANS (13), with consideration of intrathecal hydrocortisone administration in cases of severe ICANS (95). Additionally, in patients with seizures, levetiracetam is recommended (96). When there is suspicion for worsening ICANS or signs of increased ICP, patients should be transferred to the ICU for close monitoring and initiation of urgent management as clinically indicated. In a majority of pediatric ICANS cases, symptoms typically resolve by day 28, however, rare chronic neurologic sequelae and fatal cases have been reported (77).

There are many ongoing adoptive cell therapy trials for solid tumors utilizing a patient’s own T cells or NK cells to directly target tumor cells. There are currently published results from clinical trials of CAR modified T cells or NK cells targeting CD171/L1CAM, GD2, and HER2 in pediatric patients with solid tumors (97–99, 102). All trials thus far show feasibility and safety of infusing CAR modified T cells or NK cells, with no obvious neurotoxicity. GD2 and CD171/L1CAM are both proteins known to be expressed by cells in the central nervous system and peripheral nerves, but no overt toxicities have been observed in clinical trials targeting these proteins (96–101). There is one published study of T cells genetically engineered with an NY-ESO-1 reactive T cell receptor (TCR) that includes young adults with synovial sarcoma, where antitumor efficacy was demonstrated and no neurotoxicity was seen (103). Of note, Guillain-Barre syndrome has been observed in adults treated with NY-ESO-1 directed T cell receptor therapy (104).

Multiple pediatric CNS tumor CAR T cell trials are underway, as well as trials using tumor specific T cells, but toxicity data is just beginning to emerge (105). Common to all studies is that focal neurologic symptoms were frequently seen after CAR T cell treatment for brain tumors, but typically the authors were not able to definitively distinguish progressive disease from immune-mediated toxicity. Results from two HER2-targeting trials showed a good safety profile. In a mixed age trial of HER2-CAR transduced virus specific T cells for glioblastoma that included 7 children, one child had a neurologic adverse event. This was one of the two patients among the 17 total patients who developed seizures and/or headaches (106). In a study of intracranially delivered HER2 CAR T cells, 3 of 3 patients developed headache and/or back pain. One patient had transient worsening of baseline neurologic deficits after the first CAR T cell dose, which was accompanied by increased peritumoral edema and contrast enhancement on imaging. Although this constituted a possible immune response, the patient had confirmed progressive disease 2 months after initiating treatment and did not meet iRANO criteria for pseudoprogression (107). A trial of GD2-targeting CAR T cells for diffuse intrinsic pontine gliomas and spinal cord diffuse midline gliomas showed evidence of efficacy, and evidence of intratumoral inflammation which the authors termed Tumor Inflammation-Associated Neurotoxicity (TIAN) (105, 108). Additional trials targeting HER2, GD2, B7H3, EGFR, and IL13Rα2 are ongoing.

Immune checkpoint proteins are a part of the adaptive immune system and their role is to prevent the immune system from over-engaging and destroying healthy cells (109). Under normal conditions, inhibitory checkpoint proteins work by engaging and binding to partner proteins and often initiate signaling pathways that suppress T cell function. Sometimes, however, cancer cells can evade detection by simulating immune checkpoint proteins thereby inhibiting T cell mechanics. Drugs that block the inhibitory checkpoint proteins, known as immune checkpoint inhibitors (ICIs), target immune checkpoints and can overcome tumor-mediated inhibition of T cell function. This effectively restores immune system function and often leads to T cell activation and tumor killing. ICIs have revolutionized cancer treatment paradigms and are approved for the treatment of multiple adult and pediatric malignancies (110, 111). With disruption of the checkpoint pathway, effectively taking the brakes off the immune system, ICIs can lead to a spectrum of inflammatory side effects including gastrointestinal, dermatologic, and endocrine toxicities (112). In the adult population, the incidence of neurologic complications ranges from 2-4% (1). In large pharmacovigilance studies, the most commonly reported neurologic syndromes were inflammatory myopathies, myasthenia gravis, noninfectious encephalitis or meningitis, and peripheral neuropathy (113, 114). Relapses of preexisting autoimmune neurologic conditions such as multiple sclerosis can occur (115, 116).

Use of checkpoint inhibitors in pediatric clinical trials often includes patients with different oncologic diagnoses including hematologic (Hodgkin’s and Non-Hodgkin’s lymphoma) and solid tumor malignancies, with few cases reported in leukemia (117–121).

The overall reported incidence of neurotoxicity across trials ranges from 20-53%, with headaches being the most common neurologic symptom reported (118, 120, 121). Other neurologic side effects that occurred less commonly included blurred vision, seizures, and hemiparesis, with reports of severe toxicity such as stroke, blindness, encephalitis, and spinal cord compression occurring at approximately 1% incidence (120, 121). (See Table 3 for neurotoxicity related to individual agents in the larger pediatric clinical trials). Additionally, hypophysitis, which is a common irAE (122), can develop in patients treated with ICIs and often presents with concurrent neurologic symptomatology. Thus, assessment of endocrine function should always be part of the evaluation of neurologic complications of checkpoint blockade.

With ICIs, two types of neurologic toxicities can be expected: systemic immune-mediated neurologic disorders, which can occur with ICI use for any indication, and local immune activation at the tumor site, which is unique to brain tumors. Although these presentations are typically easy to distinguish, there may be overlap, and both will likely respond to steroids.

No randomized controlled trial results have yet been published on the use of ICI for treatment of pediatric brain tumors. Efficacy against malignant gliomas in pediatric patients with congenital mismatch repair deficiency (CMMRD) has been described in case reports, with clinical trials ongoing. Two siblings with CMMRD and glioblastoma responded to nivolumab, with their courses complicated by focal seizures and reversible radiographic pseudoprogression (123). In children without CMMRD, transient radiographic responses were seen in 3 of 10 patients with a variety of malignant CNS tumors, and no neurologic toxicities were reported (124). After treatment with 3 doses of nivolumab, a 10-year-old with glioblastoma developed successively worsening severe peritumoral edema; the first time it was reversible with steroids, the second time she required hemicraniectomy, and the third time she died (125). In this case, it was difficult to discern whether this death was from immune response or from tumor progression.

Antibody-based therapy in cancer therapeutics utilizes the recognition properties of an antibody to bind to a specific antigen on a malignant cell. Upon binding, the immune system triggers downstream processes that promote cell destruction and apoptosis either through direct cytotoxicity, non-restricted activation of cytotoxic T cells, and/or Fc-mediated immune effector engagement (135). The toxicity profile of the antibody-based therapy will be dependent on whether an antibody is conjugated (bound to another substrate such as an enzyme, toxin, or inorganic compound) or unconjugated, as conjugation itself can alter the side effect profile. This section will focus on three categories, bispecific T cell engagers, unconjugated antibodies, and antibody-drug conjugates.

Targeted therapy using monoclonal, conjugated, and unconjugated antibodies has been utilized in early phase clinical trials, with some drugs now included in upfront treatment in hematologic malignancies. These include antibodies targeting CD20 (e.g., rituximab), CD22 (e.g., inotuzumab, epratuzumab), and CD30 (e.g., brentuximab vedotin) with specific clinical trials discussed in Table 3 . Broadly, use of these antibodies in pediatric hematologic malignancies has shown an incidence of neurotoxicity of up to 30% in published trials, with the most common neurologic adverse events being pain, peripheral neuropathy, and headache (151–154). Other more rare and serious adverse events seen include chorioretinitis, cerebellar syndrome, cognitive disturbances, encephalopathy, intracranial hemorrhage, and meningitis (155, 156).

Clinical trials with BITEs in pediatric solid tumor are ongoing with no toxicity data available.

Many unmodified and modified monoclonal antibodies have been evaluated for high-risk, relapsed, and refractory solid tumors. Reported neurotoxicity from published clinical trials on specific agents is summarized in Table 3 . In clinical trials evaluating multiple agents, it is not possible to determine if neurotoxicity is directly caused by the monoclonal antibody. The disialoganglioside GD2 is a common monoclonal antibody target in pediatric oncology. Neuropathic pain is a well-described toxicity of anti-GD2 antibody therapy due to on-target binding of pain fibers and activation of the complement system (157). Co-administration of nonsteroidal anti-inflammatory drugs, gabapentinoids, and analgesic drugs, as well as temporarily stopping and slowing the infusion are primary ways to mitigate pain in patients. Common sites of pain are the abdomen, chest, back, and extremities. Pain is usually observed shortly after the start of the infusion, is most frequent during the first cycle of anti-GD2 therapy, and typically decreases in frequency during subsequent cycles (158).

Adie’s pupil is an ocular abnormality associated with anti-GD2 antibody therapy. It is characterized by mydriasis and accommodation deficits and is thought to be due to binding of antibody to neural structures in the ciliary and sphincter muscles of the eye. Most vision abnormalities can be corrected with prescription glasses and symptoms often improve or completely resolve after completion of therapy (159).

Posterior reversible encephalopathy syndrome (PRES) is also associated with GD2 targeting antibody therapy and can present with hypertension, seizures, headache, visual disturbance, and/or altered mentation. Characteristic radiographic findings of PRES on MRI of the brain are edematous changes most often seen in the parietal and occipital lobes. Prior brain irradiation is significantly associated with incidence of PRES in patients who received the anti-GD2 antibody, 3F8. Most patients have complete resolution of symptoms over weeks (160).

Bispecific T cell engagers are in clinical trials for pediatric CNS tumors however toxicity data is not yet available.

Bevacizumab is the most frequently used antibody-drug in brain tumors (161, 162). It acts by blocking angiogenesis via neutralizing free circulating VEGF-A. Since it reduces contrast enhancement of tumors, interpretation of tumor response by imaging criteria alone is problematic. In pediatric trials, bevacizumab has been used together with chemotherapy and/or radiation. While safety profiles are favorable, efficacy has not been well established due to small trial size. Serious nAEs are rare; the most common of these are intracranial hemorrhages (ICH), which affect 1-4% of patients and are typically intratumoral (163–166). Importantly however, control groups not receiving bevacizumab have similar rates of ICH, which has been seen in primary brain tumors in children and adults, as well as in those with brain metastases (167–170). Other rare but serious reported adverse events include CNS ischemia ( Table 3 ) (165).

EGFR is frequently expressed by pediatric brain tumors making EGFR a suitable epitope for tumor-targeting immunotherapies (171). The antibody drugs cetuximab, nimotuzumab, and depatuxizumab (172) have been tested in children, and generally have been well tolerated (173, 174), with rare cases of headaches and ICH seen. Depatuxizumab, which targets a tumor specific EGFR epitope, has been shown to be safe in adult studies (175). However, the drug-antibody conjugate depatuxizumab mafodotin has caused corneal keratopathy in 33-81% of adult patients with GBM (175, 176). Pediatric trials of depatuxizumab mafodotin for HGG have been completed but not yet been published.

Multiple radiolabeled antibody conjugates are in clinical trials for pediatric brain tumors but no toxicity data is available yet.