CAR T-Cell Therapy for CNS Malignancies

Clinical trials targeting IL-13R $\propto$ 2

In a phase 1 trial targeting recurrent/progressive GBM, Brown and colleagues used CAR T-cells against IL-13R $\propto$ 2 (see Section Interleukin-13 receptor alpha 2 ). In this trial, a first-generation CAR with the CD3 $\zeta$ signaling domain, but no costimulatory domain, was expressed in T-cells using electroporation.

Patients were enrolled after the initial diagnosis of GBM, and peripheral blood mononuclear cells (PBMCs) were collected by leukapheresis to manufacture the CAR T-cell product. After tumor recurrence, patients underwent surgical resection of the tumor with placement of a Rickham catheter. Tumor tissue was then tested for the expression of the target antigen, which had notable intra- and interpatient heterogeneity. The three patients whose tumor tested positive for IL-13R $\propto$ 2 expression subsequently received 11 (1/3 patients) or 12 (2/3 patients) intracavitary doses of CAR T-cells at 2-week intervals. An intrapatient dose escalation strategy was employed for the first three doses (1 × 10 ⁷ , 5 × 10 ⁷ , and then 1 × 10 ⁸ CAR+ T-cells), and the highest dose was then used for all subsequent doses. Patients were evaluated using brain MRI at week 3 and week 6 after the first infusion.

In this trial, there was only transient antitumor activity that was mostly localized to the site of CAR T-cell infusion. The disease progressed in all patients resulting in a mean overall survival of 11 months after relapse. The use of a less effective first-generation CAR lacking a costimulatory signal and the relatively long ex vivo manipulation process, which can affect the quality and potency of manufactured CAR T-cells, may partially explain the apparent lack of efficacy. Moreover, trafficking of the infused CAR T-cells following intracavitary delivery appeared to be limited as new lesions developed at distal sites. Of note, there was evidence of a reduction in expression of IL-13R $\propto$ 2 in a postinfusion tumor biopsy that was available for one patient who underwent resection for tumor recurrence. Although this might be considered as evidence of antitumor activity, it also raises concern over the potential for antigen escape.

In a follow-up clinical trial, Brown et al. used a self-inactivating lentiviral vector to transduce enriched central memory T-cells with a second-generation anti-IL-13R $\propto$ 2 CAR incorporating the 4-1BB costimulatory domain. A brief report detailing the treatment and response of a single patient treated as part of this trial was published recently. This patient had recurrent GBM with leptomeningeal spread documented around 6 months after standard-of-care therapy. Before the initiation of CAR T-cell infusions, three of five detectable tumor sites were resected, and a Rickham catheter was placed at one of these tumor sites. The patient then received a total of six scheduled intracavitary doses of CAR T-cells (dose 1: 2 × 10 ⁶ CAR+ T-cells, doses 2–6: 10 × 10 ⁶ CAR+ T-cells). Despite stabilization of the tumor at the infusion site, the other tumor sites continued to progress, and new lesions developed at distant sites, again suggesting these CAR T-cells were unable to traffic to distal sites following intracavitary infusion. A second intraventricular catheter was therefore placed in this patient, who went on to receive 10 additional intraventricular doses of CAR T-cells (doses 7–16: 10 × 10 ⁶ CAR+ T-cells).

The patient responded to CAR T-cell therapy with a gradual decrease in the size of all tumor lesions that had been detected prior to initiation of intraventricular CAR T-cell infusions. He was also successfully tapered off of systemic dexamethasone. These lesions eventually became undetectable by MRI/PET CT and did not recur. However, the tumor recurred at four new sites at around 7.5 months (228 days) after the first CAR T-cell infusion. Notably, there were no dose-limiting toxicities (DLTs) and no grade 3 or 4 adverse events recorded during treatment. However, the concomitant administration of supportive/symptomatic therapies, including dexamethasone (tapered dose $\le$ 4 mg per day), divalproex (7500 mg twice daily), and acetaminophen, may have reduced the risk for adverse events and DLTs.

Furthermore, in this patient, all detectable GBM lesions (including a spinal lesion) decreased in size after the start of intraventricular CAR T-cell infusions, suggesting that this route improved the trafficking of the infused cells compared with intracavitary delivery. Moreover, the elimination of tumor lesions that have a nonuniform expression of IL-13R $\propto$ 2 (as detected by immunohistochemistry [IHC]) is encouraging. However, whether this represents a “bystander” effect of activated CAR T-cells in the TME or the recruitment of the endogenous immune system is unclear.

Despite these encouraging observations, this case report also highlighted notable limitations of CAR T-cell therapies. Importantly, CAR T-cells failed to expand significantly following intraventricular administration, and their persistence was limited (detectable in the cerebrospinal fluid [CSF] for up to 7 days postinfusion). Multiple factors may have contributed to these observations including the concomitant administration of low-dose dexamethasone, the route of administration, and the highly immune-suppressive TME of GBM. Lastly, a decrease in IL-13R $\propto$ 2 expression was noted in the tumor that recurred following CAR T-cell infusion, raising concerns about the possibility of antigen escape after targeting a single TAA.

Clinical trial targeting human epidermal growth factor receptor-2

Another phase 1 clinical trial conducted by Ahmed et al. targeted the human epidermal growth factor receptor-2 (HER-2) (see Section Human epidermal growth factor receptor 2 ) to treat recurrent or progressive GBM. In this trial, autologous virus-specific T-cells (VSTs) were transduced with a gamma-retroviral vector encoding a second-generation anti-HER-2 CAR with the CD28 endodomain. VSTs reactive against viral antigens derived from cytomegalovirus (CMV), Epstein-Barr virus (EBV), or adenovirus were used in an effort to coopt the natural T-cell receptor (TCR)–mediated signaling and improve CAR T-cell functionality and longevity. The HER2 VSTs were delivered intravenously (IV), as prior clinical trials had demonstrated T-cell trafficking to the CNS.

A total of 17 patients (10 adults and 7 children) with recurrent GBM were treated, and 16 were evaluable as one patient was excluded from analysis because they received chemotherapy within 6 weeks from T-cell infusion. Patients received one or more IV infusions of autologous HER2-CAR VSTs at five dose levels (1–100 × 10 ⁶ HER2-CAR+ VSTs; with a cohort size of three patients per dose level). This dose-escalation strategy starting at a relatively low dose of 1 × 10 ⁶ cells/m ² was designed to address concerns for possible off-target adverse events. These concerns were fueled by a report of a patient who suffered lethal cytokine release syndrome (CRS) with multiorgan failure and possible on-target/off-tumor toxicity due to reactivity against lung epithelial cells after intravenous infusion of a high dose (1 × 10 ¹⁰ cells) of anti-HER2 CAR T-cells that incorporated the scFv derived from trastuzumab. Ahmed et al., however, utilized a CAR with an scFv derived from the FRP5 antibody that targets a more distal epitope compared with trastuzumab. FRP5-derived HER2 CAR T-cells were shown to be safe in a recent dose-escalation phase 1 trial targeting sarcoma. Importantly, no dose-limiting adverse events were noted in this trial, but one patient had grade 4 cerebral edema, two had grade 3 headaches, and one had grade 3 hydrocephalus ( Table 12.1 ), highlighting the need for close monitoring of patients receiving CAR T-cell therapies for local, CNS-related side effects.

All patients were evaluated with a baseline brain MRI prior to initiation of CAR T-cell infusions and a repeat MRI at 6 weeks postinfusion. The RECIST criteria were used to evaluate radiographic responses. Patients who had evidence of response were eligible to receive up to six additional doses of T-cells at 6- to 12-week intervals at the same dose level ( Table 12.1 ). In this study, the median overall survival was 11.1 months from the first infusion and 24.5 months from diagnosis.

Notable responses were documented in this trial, including a 17-year-old patient who received two doses at 1 × 10 ⁶ cells/m ² and had a partial response (PR) that lasted 9.2 months after the first infusion and survived for a total 26.9 months from that date. Seven patients had stable disease (SD) lasting 8 weeks to 29 months postinfusion, and three of these patients were alive at study conclusion (with SD lasting 24, 28.8, and 29 months). Although eight patients were considered to have progressive disease (PD) based on their follow-up MRIs, six survived for more than 6 months postinfusion. These patients were classified as having progressive disease based on the RECIST criteria, but the radiographic changes noted may have presented local inflammation owing to local T-cell expansion and activation, a phenomenon referred to as “pseudoprogression” in other immunotherapy trials. Since the end of this trial, criteria used for radiologic assessment of tumor responses in immunotherapy trials have been modified by multiple groups in an attempt to improve accuracy.

Additionally, HER2 VST levels in the peripheral blood were assessed serially by qPCR, with levels peaking at 3 hours in 15 of 17 patients and declining thereafter. This rapid peak and subsequent gradual decline suggests that the autologous HER2 VSTs had limited in vivo expansion. Although the persistence of HER2 VSTs was limited in the majority of patients, the cells were detectable in the peripheral blood up to 12 months after the initial infusion in two out of six patients evaluated after receiving multiple infusions (see Section Chimeric Antigen Receptor T-Cell Pharmacokinetics in the Setting of Central Nervous System Malignancies ). Of note, whether VSTs are more effective than nonselected T-cell populations remains an open question. Although CMV seropositivity was part of the inclusion criteria, only five patients had pp65-positive GBM (as assessed by IHC); of these patients, two had SD and three had PD.

Another interesting observation from this study is that patients who had received salvage therapies prior to trial enrollment had shorter median survival. Although these patients might have had more aggressive tumors prompting the use of salvage therapies, exposure to these potentially cytotoxic agents may have also led to T-cell defects that compromised the quality of the autologous CAR T-cell product. Such adverse effects following multiple cytotoxic chemotherapy cycles have been recently described in a cohort of pediatric patients with a range of different malignancies.

Clinical trial targeting epidermal growth factor receptor variant III

O'Rourke et al. conducted a clinical trial treating 10 adult patients with recurrent/progressive GBM with a single IV dose of autologous anti-EGFRvIII (epidermal growth factor receptor variant III) CAR T-cells (see Section Epidermal growth factor receptor variant III ). T-cells were transduced with a lentiviral vector to express a second-generation anti-EGFRvIII CAR incorporating a 4-1BB costimulatory domain. Patients with newly diagnosed or recurrent/progressive GBM were tested for EGFRvIII expression using an RNA-based next-generation sequencing assay. Those who tested positive and met the other eligibility criteria underwent leukapheresis for PBMC collection and storage. Any subsequent evidence of disease progression triggered CAR T-cell manufacturing.

Patients were evaluated with a baseline brain MRI followed by a single IV infusion of 1.75–5 × 10 ⁸ CAR+ T-cells. Of the 10 patients treated, six were evaluated with a repeat brain MRI at week 4 postinfusion. Five of these patients had SD and one had PD based on the RANO criteria. Given the complexity of the radiographic findings noted postinfusion, the authors focused subsequent analyses on the CAR T-cell engraftment and trafficking to the tumor, as well as their subsequent effects on the TME.

CAR T-cells were detectable (by flow cytometry and qPCR) in the peripheral blood of all patients, with peak levels seen at around 1–2 weeks postinfusion. However, the detection level declined thereafter, and CAR T-cells were no longer detectable in the peripheral blood by day 30 postinfusion in all patients. To test whether CAR T-cells were able to successfully traffic to the tumor site within the CNS, the authors determined the levels of CAR T-cells in postinfusion tumor biopsies whenever available. In all four patients who underwent surgical resection within 2 weeks postinfusion, CAR T-cells were detectable. In contrast, they were detectable in only one of three patients who underwent resection at later time points.

Further testing on the biopsies obtained from the four patients who underwent early resection provided insight into potentially important changes within the TME post-CAR T-cell infusion. Patchy lymphocyte infiltrates consisting of a mix of CD8+ and CD8− cells were noted, and a greater percentage of CD8+ Ki67+ cells were also detected postinfusion, indicating a proliferating cell population. However, there was a concomitant increase in the percentage of CD3+ T-cells expressing regulatory T-cell (T _reg ) markers postinfusion. Other markers of immune suppression in the TME (including IL-10, indoleamine 2,3-deoxygenase [IDO1], and programmed death ligand 1 [PD-L1]) also were upregulated postinfusion, and many of these factors are known to be upregulated by IFNγ, an effector cytokine secreted by activated T-cells. In contrast, the changes were not as dramatic in biopsies obtained at later time points (more than 2 weeks postinfusion).

The authors also analyzed the TCR repertoire of the CAR T-cell product and the tumor-infiltrating lymphocytes (TILs) in the tumor biopsy samples obtained pre- and postinfusion. A minority of TCRs detectable in the infusion product were also detected in the postinfusion biopsy, suggesting limited trafficking of peripherally infused cells to the tumor site. However, the detected T-cells represented a relatively large proportion of postinfusion TILs. Additionally, the postinfusion TCR repertoire of TILs was broader than their preinfusion TCR repertoire, indicating the possibility of epitope spreading; however, no further analyses were available to support this at the time of publication. Of note, the percentage of TILs with a regulatory phenotype also rose postinfusion, and this may represent a component of the immune evasion mechanisms employed by solid tumors in response to local CAR T-cell activation.

Importantly, the investigators also examined the effects of CAR T-cell infusion on the expression of the target antigen. In five of seven patients tested, the EGFRvIII mRNA expression levels (assessed by qPCR) decreased. The remaining two patients had stable expression levels. Of these, one patient had a spatially heterogeneous expression of EGFRvIII mRNA, whereas the other had poor CAR T-cell engraftment and early progression. The reduction in EGFRvIII level may reflect the activity of anti-EGFRvIII CAR T-cells, but it also raises concerns over the risk for antigen escape following CAR T-cell therapies targeting a single antigen, especially in malignancies that exhibit significant intratumoral antigen heterogeneity.

Epidermal growth factor receptor variant III

EGFRvIII is the first and only neoantigen to date used as a CAR T-cell target to treat brain tumors, primarily GBM. Amplification of the EGFR gene is the most frequent genetic change associated with GBM, common in the classic or receptor tyrosine kinase type 2 molecular subtype of isocitrate dehydrogenase (IDH) wild-type GBM. The EGFR-amplified GBM also expresses the tumor-specific deletion variant (EGFRvIII), present in 25%–33% of all patients with GBM. First identified by Sugawa and colleagues, variant III of EGFR is a truncated protein formed by identical splicing of exon 1 to exon 8 as a consequence of a deletion-rearrangement of the amplified gene. This rearrangement results in the loss of 801 coding bases (exons 2–7) and creation of a new codon for glycine at the novel splice site in their corresponding transcripts, leading to a neoantigenic epitope that is not present in any normal tissues. The truncated receptor protein lacks major parts of the extracellular domain, and while unable to bind its ligands, it remains constitutively active, leading to proliferation of tumor cells. The prognostic role of EGFRvIII positivity in EGFR-amplified primary and recurrent GBMs has been controversial, with earlier studies associating both prolonged and poorer overall survival to EGFRvIII positivity. A recent study by Felsberg et al. at German Glioma Network reported that EGFRvIII was not prognostic in EGFR-amplified GBM in their cohort. Additionally, this group reported changes in EGFRvIII expression levels at recurrence, thus recommending repeated biopsy for EGFRvIII status for recurrent GBM patients receiving EGFRvIII-targeted therapies. Apart from GBM, EGFRvIII has been previously reported in pediatric DIPGs.

The discovery of monoclonal antibody (mAb) 806 and development of other antibodies such as mAb139 that specifically inhibited the growth of tumor xenografts expressing either the EGFRvIII or amplified EGFR, but not wild-type EGFR, initiated immunotherapeutic targeting of EGFR-amplified GBM. The first CAR molecule, also referred to as chimeric immune receptor targeting EGFRvIII, was developed by Bullain and colleagues at Massachusetts General Hospital, Boston. Subsequently, investigators at the National Cancer Institute developed the first EGFRvIII targeted CAR T-cells based on mAb139. Their studies demonstrated specific recognition and antitumor activity against glioma stem cell lines and glioma cell lines expressing mutant EGFRvIII without reactivity to wild-type EGFR. Developing further on established preclinical data for EGFRvIII-targeted CAR T-cells, Johnson and colleagues rationally tested and characterized a battery of humanized anti-EGFRvIII CAR T-cells derived from murine mAb139 scFvs. The team's first-in-human study of intravenous delivery of a single dose of autologous EGFRvIII CAR T-cells against recurrent GBM (NCT02209376) demonstrated safety without any DLTs, and other findings from this study are summarized in Summary of Published Chimeric Antigen Receptor T-Cell Clinical Trials Targeting Central Nervous System Malignancies section. There are currently multiple active trials for EGFRvIII-targeted CAR T-cells recruiting patients at the National Cancer Institute and the Duke Cancer Institute (NCT03283631), the Abramson Cancer Center of the University of Pennsylvania (NCT03726515), and the Beijing Sanbo Brain Hospital in China (NCT02844062).

Human epidermal growth factor receptor 2

Human epidermal growth factor receptor 2 (HER-2, also known as receptor tyrosine-protein kinase erbB-2, CD340, or protooncogene Neu) is a 185-kDa protein receptor with tyrosine kinase activity and extensive homology to EGFR. When present in tumors, HER-2 acts as the preferred heterodimerization partner for other ErbB receptors and a potent signal amplifier. In tumors of the CNS, especially GBM, HER-2 expression is associated with poor prognosis. HER-2 is not expressed in the normal adult brain, and in brain tumors, its expression is associated with increased cell proliferation, metastasis, and the inhibition of apoptosis. Glial tumors expressing HER-2 are also less differentiated. Gilbertson and colleagues identified HER-2 expression in 83.6% of childhood medulloblastoma in their cohort and demonstrated that HER-2 signaling results in aggressive disease behavior in ependymoma by promoting tumor cell proliferation. Though not directly demonstrated, atypical teratoid/rhabdoid tumor (ATRT) responds to inhibition of ErbB2-EGFR pathway by the small molecule inhibitor lapatinib. Thus, HER-2 is an attractive candidate for immunotherapy-based approaches against CNS tumors.

Unlike many epithelial tumors including breast, ovarian, or gastric cancer, amplification of HER-2 is not observed in CNS tumors. Therefore, therapy with monoclonal antibody, trastuzumab (4D5), is ineffective in these tumors. Ahmed and colleagues at Baylor College of Medicine first demonstrated the antitumor efficacy of HER-2-targeted CAR T-cells based on the FRP5 mAb in preclinical models of medulloblastoma and GBM, and the latter study used patient-derived GBM cells and GBM stem cells. The team then completed a first-in-man clinical trial for patients with progressive or recurrent GBM using HER-2 CARs grafted on CMV-specific T-cells ( NCT01109095 ). The results of this trial are detailed in Summary of Published Chimeric Antigen Receptor T-Cell Clinical Trials Targeting Central Nervous System Malignancies section . Encouraged by the safety profile of this trial, in which T-cells were delivered systemically, the team is currently testing whether intracranial delivery of HER-2 CAR T-cells either to resection cavities or to the ventricular space will improve their trafficking and antitumor activity (NCT02442297). Another widely used HER-2 CAR for CNS tumors is based on the scFv from trastuzumab mAb (4D5). CAR T-cells using this scFv are currently being evaluated in clinical trials at City of Hope Medical Center in California (NCT03696030 and NCT03389230), Seattle Children's Hospital (NCT03500991), and Xuanwu Hospital, Beijing, China (NCT03423992).

Interleukin-13 receptor alpha 2

Interleukin-13 receptor alpha 2 (IL-13Rα2) is a GBM-associated protein that is overexpressed on up to 78% of tumors. Due to its negligible expression in the normal brain, IL13Rα2 is a promising candidate for targeted therapy. IL13Rα2 is an isolated receptor that binds IL-13 only and, unlike the widespread IL13Rα1 receptor, does not form a heterodimer with IL4R. Since the receptor lacks a signaling domain and has only a short cytoplasmic tail and a high affinity for the cytokine, it was initially thought to be a “decoy” receptor sequestering IL-13 and thus neutralizing its effect. In fact, IL-13Rα2 blocks IL-13-driven STAT6 signaling by binding IL-13 with high affinity. One recent study suggested that the IL-13/IL-13Rα2 axis is important in mediating signal transduction by increasing expression of AP-1 transcription factors in IL-13Rα2-positive, but not in IL-13Rα2-negative, glioma cell lines. Along with poor survival prognosis in GBM, IL-13Rα2 gene expression is reported to be associated with GBM resistance to temozolomide (TMZ) chemotherapy. The presence of IL-13Rα2 has also been described in pediatric cancers such as medulloblastoma and ependymoma. Recently, IL-13Rα2 has been shown to act in cooperation with EGFRvIII signaling to promote proliferation of GBM cells.

The initial iterations of CAR T-cells targeting IL-13Rα2 were based on membrane-tethered IL-13 as binding exodomains and termed as “IL-13-zetakine.” The substitution of an amino acid (E13Y) improved the specificity of the zetakine CAR to IL-13α2 and reduced its activity against targets expressing IL-13Rα1. The IL13 (E13Y)-zetakine CAR entered phase 1 trial at the City of Hope Medical Center (NCT02208362) and has demonstrated safety and objective responses after intracavitary or intratumoral administration. Later, an IL-13 mutein CAR was designed with two IL-13 mutations (E13K and K105R) and showed potent antitumor activity in xenograft models of GBM but recognized IL-13Rα1. Kong et al. also designed a zetakine CAR based on mutated IL13 extracellular domain (E13K and R109K) linked to intracellular signaling elements of the CD28 costimulatory molecule and CD3ζ. The IL-13 mutation enhanced the selectivity of CAR recognition of IL13Rα2 receptor compared with the IL13Rα1 receptor or the composite IL13Ra1/IL4Ra receptor. Recently, Krenciute and colleagues developed a series of IL13Rα2 CARs based on the scFv 47 and demonstrated that T-cells transduced with the CAR constructs bearing a short hinge showed the best antitumor activity in preclinical models of GBM. This study was followed by scFv-based IL13Rα2-specific CARs from groups at the University of Pennsylvania and City of Hope Medical Center. Another clinical trial testing the safety of IL13Rα2-directed CAR T-cells is ongoing at Xuanwu Hospital, Beijing, China (NCT03423992).

CD133

CD133, also known as AC133 and prominin-1, is a 97-kDa pentaspanning transmembrane glycoprotein that is expressed on cells with a stem-like phenotype. Indeed, CD133 has been used as a cell surface marker to isolate cancer stem cells from various solid tumors, including neoplasms of the brain. The marker was first used to describe cancer stem cells in pediatric tissue samples of medulloblastoma and glioma. Singh et al. found that CD133-expressing tissues could regenerate a heterogeneous tumor, attributable to the increased capacity for self-renewal observed in clinically aggressive medulloblastoma. Zhu and Niedermann at the University Hospital Freiburg, Germany, developed CARs targeting the AC133 epitope of CD133 and demonstrated antitumor activity against GBM stem cells in an orthotopic tumor model. Recently, a team from the Chinese PLA General Hospital reported on the phase 1 trial of CD133-directed CAR T-cells for advanced metastatic malignancies, but this trial did not accrue patients with CNS tumors (NCT02541370). Currently, one active clinical trial is reported at Xuanwu Hospital (NCT03423992) using CD133-directed CAR T-cells for malignant glioma.

Disialoganglioside II3(NeuAc)2GgOse3Cer

Disialoganglioside II3(NeuAc)2GgOse3Cer (GD2) is commonly overexpressed in pediatric and adult solid tumors, including neuroblastoma, glioma, retinoblastoma, most sarcomas, small-cell lung cancer, and melanoma. Traylor and Hogan first observed elevated proportions of gangliosides including GD2 in human gliomas compared with normal brain tissue. Later, another group demonstrated the reactivity of GD2-specific mAb DMAb-20 in 16 of 20 (80%) malignant glioma and 5 of 5 medulloblastoma cell lines. Mount et al. observed that GD2 was expressed at high levels on patient-derived DIPG cultures and was conserved as a surface marker across DIPG patients and other histone H3 K27M (H3K27M)–mutated diffuse midline gliomas (DMGs). Furthermore, Mount and colleagues showed that T-cells grafted with GD2-specific CARs derived from 14g2a mAb scFv and delivered systemically had potent antitumor efficacy against xenograft models H3K27M-mutant DMGs and DIPG. A randomized, open-label combined phase 1 and 2 trial for GD2 CAR T-cells was completed at Fuda Cancer Hospital, China (NCT03252171), but the study outcomes have not yet been reported. An active clinical trial with GD-2 CAR T-cells is ongoing at Xuanwu Hospital (NCT03423992).