Cell therapy for brain tumors: The first 60 years

Summary

Primary brain tumors remain among the most lethal cancers, but immunotherapy holds immense potential to overcome limitations of current standard treatment modalities. Since the late 1960s, early-phase clinical trials have iteratively tested cellular immunotherapies for the treatment of brain tumors. Six decades ago, in the earliest studies, brain tumor patients were treated with infusions of nonspecific leukocytes, peripheral blood mononuclear cells (PBMCs), and bone marrow cells. These earliest studies demonstrated safety and occasional durable antitumor responses, particularly when cell therapies were combined with conventional modalities or administered in the upfront setting. These early cell therapy approaches were chronologically followed by lymphokine-activated killer (LAK) cells, tumor-infiltrating lymphocytes (TILs), ex vivo nonspecifically expanded and antigen-specific T cells, natural killer (NK) cells, and chimeric antigen receptor (CAR) T cells. In this historical review, we summarize the clinical experience with adoptive cell therapies for brain tumors and review key findings from published clinical studies.

Graphical abstract

Cellular immunotherapies have been tested for the treatment of primary brain tumors since 1968, with clinical responses observed, albeit limited and inconsistent, from the earliest studies. Mehta et al. review clinical experience with adoptive cell therapies for brain tumors, highlighting key findings from published clinical studies.

Introduction

Primary brain tumors remain among the leading causes of morbidity and mortality across all age groups.^1,2 Rickman J. Godlee performed the first documented successful resection of a primary brain tumor in 1884.³ Since then, many advances have been made in neurosurgical procedures, radiation therapy, chemotherapy, and the molecular understanding of distinct brain tumor entities.⁴ Yet, outcomes for many brain cancers, such as glioblastoma (GBM) and diffuse midline gliomas (DMGs), remain extremely poor.^5,6 Furthermore, acute and long-term toxicities of current treatment modalities, which compromise quality of life, remain a major limitation.⁷

To address these limitations, immunotherapy has emerged as a promising tool to specifically activate immune cells against cancer antigens. Adoptive cell therapy is a type of immunotherapy that is based on the infusion of disease-targeting allogeneic or autologous immune cells into patients.⁸ The study of adoptive cell transfer dates back to 1913, when it was observed that inoculating rats with a mixture of sarcoma cells and splenocytes led to delayed tumor growth.⁹ Since then, as a result of extensive laboratory and clinical testing, cell therapy has revolutionized the treatment for relapsed/refractory B cell malignancies.^10,11 Meaningful results have also been achieved for patients with late-stage melanoma and refractory synovial sarcoma.^12,13 For the past 6 decades, cellular immunotherapies have similarly been tested in the clinic for the treatment of primary brain tumors (Figure 1).^{14,15,16,17,18,19,20,21,22,23,24,25,26,27,28} In this article, we review the approaches that have been tested and summarize the key takeaways from published clinical studies. While one could ask, why write a review on the history of cell therapy for brain tumors, we believe that it is as important to look back as it is to look forward. As the reader will appreciate, clinical response, albeit inconsistent, has been observed since the first cell therapy studies for brain tumors were conducted, and this insight alone puts the current results of chimeric antigen receptor (CAR) T cell therapy studies for brain tumors into perspective. As Mark Twain noted, “history doesn’t repeat itself, but it often rhymes.”

Figure 1.

The 60-year evolution of cell therapy trials for brain tumors

(A) A timeline of key clinical milestones in cellular immunotherapy for brain tumors; year of publication is shown unless the year of infusion was reported in the publication (denoted by ∗). PubMed Identifier, PMID; for the 1972 study with no PMID, see Takakura et al.²¹

(B) A timeline representing the progression of clinical implementation of major cell therapy products for brain tumors (Leukos, PBMCs, BM, leukocytes, peripheral blood mononuclear cells, bone marrow; LAKs, lymphokine-activated killer cells; TILs, tumor-infiltrating lymphocytes; NKs, natural killer cells; CARTs, chimeric antigen receptor T cells).

(C) Total number of patients infused with the different major cell therapy products in clinical trials for brain tumors.

Immunotherapy with leukocytes, PBMCs, and bone marrow cells

Between 1968 and 2001, more than 200 brain tumor patients were treated with leukocytes, bone marrow cells as adjuvant immunotherapy, or peripheral blood mononuclear cells (PBMCs), and the results from these studies are summarized in Table 1 and reviewed in the following text. In the 1960s, there was convincing evidence that the immune system can recognize tumor cells via the existence of tumor neoantigens.¹⁴ Lymphoid infiltration of brain tumors had been described and proposed as a predictor of favorable prognosis.¹⁴ While the natural antitumor immune response is weak and easily overcome, mouse experiments have shown that immunization via “pre-transplantation” of brain tumors to any part of the body could break immune tolerance, leading to the rejection of otherwise lethal intracerebral glioblastoma grafts.¹⁴ On the other hand, administration of immunosuppressive agents such as corticosteroids was known to impair immune function and even facilitate tumor growth.¹⁴ Finally, today it is well recognized that systemically administered immune checkpoint inhibitors can have therapeutic effects on brain metastases and biological effects against GBMs.^29,30 However, at the time, it was thought that antibodies cannot cross the blood-brain barrier, so an immunotherapy for the treatment of brain tumors should be cell based.¹⁴ This knowledge was translated into the first clinical cell therapy protocol for brain tumors in 1968.^14,31

Table 1.

Cell therapy with leukocytes, PBMCs, and bone marrow cells

Year(s)	Pt (#)	Diagnosis	Treatment	Cell dose	Best clinical outcomes
1969,14 197031	3	2 grade 4 GBM 1 grade 2/3 OA	auto cells from lymphatic fluid after antitumor immunization	6–22 × 109 (i.t. and SA)	OG pt: complete remission with no relapse for 17 mths 1 GBM pt: transient immune reaction at primary tumor site
1972,21 197532	21	recurrent tumors pediatric: MB, GBM, OPG, pinealoma, A, DIPG adult: GBM, A, EPN, OG	adjuvant pediatric: allo BM cells adult: allo or auto leukos	1–9 × 109 (BM: i.v.; other: i.t.)	significantly higher survival compared to historical controls
197733	17	recurrent grade 3–4 GBMs	auto PBMCs	0.01–1 × 109 (i.t.)	1 pt: sustained regression for 17 mths 7 pts: sustained neurological improvements and prolonged survival compared to historical controls
197834	4	primary grade 3–4 GBMs after surgery and radiation	auto PBMCs	0.001–5 × 109 (IT)	1 pt: transient improvement in speech and reduction in tumor burden 1 pt: transient reduction in tumor burden
197935	10	primary or recurrent GBM or grade 4 A after surgery, XRT, and/or chemotherapy	allo tumor RNA-sensitized leukos	0.7–1.5 × 109 (i.v.)	extended survival or delayed recurrence compared to conventional therapy
198136	9	primary GBM, EPM, DIPG, OG, A (after surgical resection)	auto PBMCs	0.01–1 × 109 (i.t.)	significant improved survival compared to historical controls (avg: 23 vs. 10.3 mths)
198437	4	recurrent GBM	auto PBMCs	0.08–4 × 109 (i.t.)	no benefit
198738	18	recurrent GBM	auto PBMCs labeled with indium-111	2.1–6.7 × 109 (IT)	no benefit
1987,39,40 199041	97	recurrent grades 2–4 glioma	auto PBMCs stimulated with PHA/IL-2	0.1–5 × 109 (i.t.)	18 pts: no recurrence for >6–33 mths incl grade 3 (n = 8) and 5 (n = 4) tumors 76 pts: tumor regression or stable disease for at least 2 mths
1989,42 199143	31	GBMs following initial resection	auto PBMCs mixed with 3 × 106 IU IFN-α	mean 8.6 × 108 (i.t.)	1 pt: transient regression for 6 mths 2 pts: stable disease for 5–6 mths
198944	4	MM, A, GBM	auto PBMCs stimulated with auto tumor cells	0.2–1.3 × 108 (i.v. or i.t.)	1 pt: neurological improvement
200145	12	refractory GBM or A	auto PBMCs stimulated with OK-432	1–11.2 × 107 (i.t.)	2 pts: transient stable disease 2 pts: transient minor regressions

Pt (#), number of patients; A, astrocytoma; DIPG, diffuse intrinsic pontine glioma; EPN, ependymoma; GBM, glioblastoma; MB, medulloblastoma; MM, metastatic meningioma; OA, oligoastrocytoma; OG, oligodendroglioma; OPG, optic glioma; XRT, radiation therapy; allo, allogeneic; auto, autologous; BM, bone marrow; IFN, interferon; IL, interleukin; leuko, leukocyte; OK-432, lyophilized Streptococcus pyogenes; PBMCs, peripheral blood mononuclear cells; PHA, phytohemagglutinin; IT, intrathecal; i.t., intratumoral; i.v., intravenous; SA, subarachnoid; avg, average; incl, including; mths, months; pt, patient.

In this inaugural attempt to induce “allergic encephalotumoritis” in humans, three patients’ primary tumors were resected, and bilateral tumor grafts were implanted subcutaneously into the mid-thighs. The tumor graft on the right thigh was boosted with the addition of Bacillus Calmette-Guérin (BCG) injections. Fourteen to 21 days after this procedure, immunity to the tumor grafts was confirmed by performing an intradermal reaction test using cryopreserved tumor homogenate. Once the patient was considered sensitized, lymphatic fluid was harvested via catheterization of the thoracic duct. Cells from the lymphatic fluid were concentrated, and 6–22 billion cells were injected near the primary tumor site, either intratumorally or into the subarachnoid space.^14,31

Following this extensive protocol, one of the patients with a grade 2/3 oligoastrocytoma achieved a complete response (CR) with no relapse or progression during 17 months of follow-up. The other two patients with grade IV GBMs had progressive disease. However, in one of the patients, histological analysis of a reoperation specimen post therapy revealed the presence of a “thick lymphocyte crown” evident of a cellular immunological reaction surrounding the primary tumor. Altogether, the authors concluded that while the therapy was safe, the overall efficacy of this treatment regimen was limited.^14,31

Ten other published clinical trials followed between 1972 and 2001 to evaluate the use of autologous and allogeneic leukocytes, bone marrow cell transfusions, and PBMCs.^{21,32,33,34,119,120,37,38,39,40,41,42,43,44,45} The majority of the patients had relapsed or refractory GBMs or anaplastic astrocytomas, although in 5 of the studies, several patients were treated in the upfront setting prior to relapse.^{21,32,34,119,120,42,43} The most promising clinical outcomes were produced by two studies.^33,41 In the first study, 17 recurrent grade III astrocytoma or GBM patients who had failed upfront radiotherapy and/or chemotherapy were treated with one or more infusions of autologous PBMCs into the tumor bed. One GBM patient, who was comatose at the time of infusion, had a sustained tumor regression and returned to full independence for 17 months post immunotherapy. Seven other patients had sustained neurological improvements and prolonged survival compared to historical controls.³³ In the second study, a total of 97 patients with recurrent malignant grade II–IV gliomas were treated with autologous lymphocytes. The lymphocytes were stimulated ex vivo with phytohemagglutinin (PHA) and interleukin (IL)-2 and then infused into the tumor bed following surgical debulking. Following immunotherapy, 76 patients had a positive initial response defined as either tumor regression or stable disease (SD) for at least 2 months after treatment. Eighteen of these responding patients, including 8 grade III and 5 grade IV tumors, remained free of recurrence during follow-up (6–33 months).⁴¹

Promising results were also observed in clinical studies of adjuvant bone marrow cell or peripheral leukocyte infusions, which were from an allogeneic source for the majority of patients.^21,32 Treated patients had a significantly improved survival compared to historical controls, and one GBM patient remained alive without recurrence during the 5-year follow-up period.³² However, many patients received their cell infusion in the upfront setting alongside conventional surgery, radiation, and chemotherapy regimens, making assessment of its benefit difficult. In addition, not all studies reported clinical benefits.^37,38 For example, the efficacy of an intrathecal injection of autologous leukocytes was compared with/without radiotherapy in 11 patients with recurrent GBM after resection. Five patients received only leukocytes and 6 leukocytes and radiation therapy, respectively. Patients who received only leukocytes had a shorter mean survival than patients who also received radiation therapy (4 vs. 11.4 months), highlighting that unspecific leukocytes do not have sufficient antitumor activity to prevent progression.³⁸

In conclusion, these early clinical trials began to demonstrate that adoptive cell therapy for brain tumors is safe and beneficial in a subset of patients who failed frontline conventional therapies. However, with less than 10%, CRs in the non-adjuvant setting were uncommon.^14,41

Lymphokine-activated killer cell therapy

In 1981, it was first reported that PBMCs from cancer patients could be activated and rapidly expanded to large numbers in vitro by culturing them with IL-2 for a minimum of 2 days.⁴⁶ The resulting cell product was coined lymphokine-activated killer (LAK) cells and consisted of a mixture of natural killer (NK) cells, T cells, and NK T cells. LAK cells can lyse otherwise NK cell-resistant autologous tumor cells (albeit at very high effector to target ratios) but do not kill healthy cells.⁴⁷ Following encouraging preclinical and clinical findings in the setting of pulmonary and hepatic metastases,^48,49 the first use of LAK cells and/or IL-2 in the treatment of human brain tumors was reported in 1986,^15,50 and until 2008, close to 300 patients were treated with LAK cells, which represents the largest group of patients treated with cellular immunotherapy. The results from these studies are summarized in Table S1 and reviewed in the following text.

In the first study, patients with progressive malignant glioma were treated with IL-2 with or without autologous LAK cells intratumorally at the time of surgical resection. While there were no objective antitumor responses, this trial demonstrated the feasibility and safety of locoregional LAK cell infusions. Since this original report, at least 20 early-phase clinical studies were conducted to evaluate the use of LAK cells for brain tumors.^{15,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73,74,75,76,77,78,79,80,81} The majority of enrolled patients were GBMs and grade III astrocytomas in the relapsed/refractory setting. However, in at least 5 studies, LAK cells were administered in the upfront setting alongside or immediately following conventional surgical resection, chemotherapy, and/or radiation therapy.^{53,57,65,66,74,77,78,80}

The best clinical responses after LAK cell therapy were observed in upfront studies where the patients were treated prior to relapse and in studies where relapsed/refractory patients were treated with LAK cells in the adjuvant setting alongside chemotherapy and/or other immunotherapies.^62,65,66,76 In one case report, a relapsed GBM patient’s tumor continued to rapidly progress despite re-irradiation, interferon (IFN)-β, and chemotherapy with ANCU (1-(4-amino-2-methyl-5-pyrimidinyl) methyl-3-(2- chloroethyl)-3-nitrosourea hydrochloride). Intratumoral infusions of autologous LAK cells were then added to the treatment regimen, and after 9 courses of LAK cells, IFN-β, and ANCU, the patient achieved a CR, which was sustained (last follow-up: >9 years).^62,76 Promising results were also observed in a study in which 10 patients were treated intratumorally with autologous LAK cells alone, and 10 patients were treated with autologous LAK cells that were pre-incubated with a bispecific antibody prior to infusion. The bispecific antibody consisted of a CD3 monoclonal antibody (mAb) chemically conjugated to a glioma-targeting mAb. At the time of the study, the target antigen of the glioma-targeting mAb was unknown but was later identified to be neural cell adhesion molecule (NCAM, CD56).⁸² All but one patient were treated after initial tumor resection and completion of radiation therapy and/or chemotherapy. One CR was observed in the group treated with LAK cells alone, and 4 CRs in the group that received bispecific antibody pre-incubated LAK cells, respectively.^65,66

However, LAK cells did not produce a consistent clinical benefit in any of the conducted clinical studies.^{15,50,67,68,69,72} While the reasons for inconsistent benefit remain largely unknown, one study highlighted that use of steroids during the month before leukapheresis was associated with a poor response to LAK cell therapy.⁸⁰ Likewise, high postsurgical tumor burden was associated with a poor response.⁵⁶ Finally, an autopsy study post LAK cell therapy demonstrated necrosis and infiltrating T cells close at the site of LAK cell administration post-surgical tumor resection but not at distant tumor sites within the brain, suggesting an inability of LAK cells to destroy macroscopic, non-resected tumor sites or their inability to migrate to tumor sites.⁶⁸

In conclusion, while some studies reported favorable outcomes in a subset of patients, the overall efficacy of LAKs with IL-2 in brain tumor patients was limited. Thus, a randomized phase 2 study for brain tumor patients was never conducted, and LAK cells were eventually eclipsed by more targeted and potent cell therapy approaches.

TILs and nonspecific and antigen-specific T cells

Between 1987 and 2018, more than 250 patients have been treated with ex-vivo-expanded tumor-infiltrating lymphocytes (TILs), non-specifically expanded T cells, or antigen-specific T cells, and the results from these studies are discussed in the following text and summarized in Table 2. In the quest to generate therapeutic cells with better antitumor activity than LAK cells, TILs were first identified in 1986.⁸³ TILs consist primarily of autologous cytotoxic T cells that are isolated from tumor specimens and expanded ex vivo in the presence of IL-2 prior to re-infusion. In vivo, TILs were 50–100 times more potent than LAK cells in pulmonary and hepatic metastatic tumor models.⁸³ An additional advantage of TILs is the presence of subpopulations of lymphocytes that can infiltrate growing tumors, being thus potentially more capable of homing to metastatic tumor sites. While TILs have been evaluated in the clinic for other cancers,¹² leading to their Food and Drug Administration (FDA) approval for melanoma in 2024, only one clinical study for brain tumors has been published.^16,84 TILs were successfully expanded from the tumor specimens of all 6 patients and reinfused intratumorally. At long-term follow-up, one recurrent anaplastic astrocytoma patient had a CR (45 months), and two patients (recurrent astrocytoma or GBM) had sustained partial responses (PRs) (48 and 47 months).¹⁶ Despite these encouraging results, no follow-up trials have been conducted in the US. Two phase 1 trials to evaluate TILs for brain tumors are currently ongoing (NCT06640582 and NCT04943913), but the results have yet to be published.

Table 2.

Cell therapy with tumor-infiltrating lymphocytes and nonspecific and antigen-specific T cells

Year(s)	Pt (#)	Diagnosis	Treatment	Cell dose	Best clinical outcomes
198722	5	primary or recurrent AO or GBM	auto T cells after in vitro antitumor immunization	5 × 107 (i.t.)	2 pts: PR for 20 weeks and >24 mths
199685	15	recurrent GBM or grade 3 A	auto T cells with IL-2 after antitumor + BCG immunization	0.1–9 × 1010 (i.v.)	10 pts: SD for 3 to >40 mths
1996,84 199916	6	recurrent malignant glioma	auto TILs and IL-2	1 × 109 (i.t.)	1 pt: CR for >45 mths 2 pts: PR for >47 mths 2 pts: PR for 6 mths
199786	5	recurrent grade 3–4 glioma	allo T cells after auto HLA sensitization and IL-2	0.1–5.15 × 109 (i.t.)	2 pts: no reoccurrence for >28 mths 1 pt: SD for >28 mths
199887	10	recurrent GBM or grade 3 A	auto T cells after antitumor + GM-CSF immunization	0.009–1.5 × 1011 (i.v.)	1 pt: no recurrence for >12 mths
199988	4	recurrent GBM or anaplastic A	auto T cells after in vitro antitumor immunization	0.7–7.3 × 107 (i.t.)	3 pts: tumor regression for 1 mth, 2 mths, and >1 year
200089	12	grade 2 A, anaplastic glioma, GBM after surgery and XRT	auto T cells after antitumor + GM-CSF immunization	0.6–5.5 × 1010 (i.v.)	4 pts: PR for 11 to >29 mths 2 pts: SD for >20 and >27 mths
200090	9	recurrent GBM or grade 3 A	auto T cells with IL-2 after antitumor + BCG immunization	1–9.6 × 1010 (i.v. or IA)	2 pts: CR for >4 and 5 years
200091	19	recurrent GBM or grade 3 A	auto T cells after antitumor + GM-CSF immunization	0.5–5 × 1011 (i.v.)	1 pt: CR for >17 mths 7 pts: PR for 3 to >28 mths 2 pts: SD for >7 mths
200392	10	recurrent GBM, anaplastic A, anaplastic OA	auto T cells after in vitro antitumor immunization	0.03–2.47 × 109 (i.v.)	1 pt: CR for 21 mths 4 pts: PR for 1–16 mths
200893	3	refractory anaplastic A or EPN	auto T cells after antitumor immunization	unknown (i.v.)	2 pts: tumor regression for 480 days and >7 years
201094	5	primary GBM	adjuvant auto T cells	1–1.7 × 1010 (i.v.)	3 pts: CR for 5, 12, and >14 years 2 pts: PR for 24 and 39 mths
201295	1	recurrent GBM	auto CMV-specific CD8+ T cells followed with TMZ after 10+ days	4 × 107 (i.v.)	tumor regression for >17 mths
201419	11	recurrent GBM	auto CMV-specific T cells	2.5–4 × 107 (i.v.)	4 pts: SD for >5 mths to >4 years
201796	23	primary or recurrent GBM, anaplastic OG, anaplastic OA, LGG, anaplastic EPN	auto lymphokine-activated T cells +/− TMZ	6.54–9.34 × 107 (i.v.)	10 pts (+TMZ): 3 PR, 7SD 4 pts (−TMZ): 1 PR, 3 SD
201797	91	primary GBM	auto T cells + TMZ	0.12–1.96 × 1010 (i.v.)	significant improved PFS compared to TMZ-only ctrl group (avg: 8.1 vs. 5.4 mths)
201898	23	recurrent GBM	auto T cells after cancer/testis antigen-expressing T helper cell immunization	0.0054–1.86 × 109 (i.v.)	3 pts: tumor regressions for >14, >22, and >27 mths

Pt (#), number of patients; A, astrocytoma; EPN, ependymoma; GBM, glioblastoma; LGG. low-grade glioma; OA, oligoastrocytoma; OG, oligodendroglioma; XRT, radiation therapy; allo, allogeneic; auto, autologous; BCG, Bacillus Calmette-Guerin; CMV, cytomegalovirus; GM-CSF, granulocyte-macrophage colony-stimulating factor; HLA, human leukocyte antigen; IL, interleukin; TILs, tumor-infiltrating lymphocytes; TMZ, temozolomide; IA, intraarterial; i.t., intratumoral; i.v., intravenous; avg, average; CR, complete response; ctrl, control; mths, months; PFS, progression-free survival; PR, partial response; SD, stable disease.

In addition to TILs, other strategies with ex-vivo-expanded T cells have been evaluated for brain tumors in at least 16 early-phase clinical trials conducted between 1987 and 2018.^{16,19,22,84,85,86,87,88,89,90,91,92,93,94,95,96,97,98} In these studies, about half of the patients have relapsed/refractory GBMs or grade III astrocytomas.^{16,19,84,85,86,87,88,90,91,92,93,95,98} The other half had newly diagnosed GBMs or other malignant gliomas, and were treated concomitantly or following primary conventional therapies.^{22,89,94,96,97} In two of the studies, autologous T cells were nonspecifically expanded from PBMCs using anti-CD3 and IL-2.^94,97 Both studies were randomized, treated newly diagnosed GBM patients, and gave multiple doses of autologous T cells intravenously. In one study, patients received T cells immediately following a standard regimen of maximum surgical resection, ANCU chemotherapy, and radiotherapy.⁹⁴ Three out of 5 patients had CRs lasting 5 to more than 14 years, and the median survival was 96.8 months longer than that of patients who only received the standard therapy regimen without T cells.⁹⁴ In the other study, patients received T cells simultaneously with standard temozolomide (TMZ) chemotherapy and radiotherapy following maximum surgical resection.⁹⁷ Median progression-free survival was significantly longer compared to patients who received TMZ and radiotherapy only (8.1 vs. 5.4 months), but there was no difference in overall survival.⁹⁷

In the remaining studies, T cells were sensitized to tumor antigens using various strategies prior to or during ex vivo expansion.^{16,19,22,84,85,86,87,88,89,90,91,92,93,95,96,98} In 6 of these studies, patients were immunized with an intradermal injection of their own irradiated tumor cells and an adjuvant, either granulocyte-macrophage colony-stimulating factor (GM-CSF) or BCG.^{85,87,89,90,91,93} Following vaccination, T cells isolated from draining lymph nodes near the vaccine site or from PBMCs were activated and expanded ex vivo before intravenous infusion. In total, 56 relapsed/refractory and 12 newly diagnosed brain tumor patients were treated with this strategy. For relapsed/refractory patients, clinical activity was observed in 33 out 56 patients, including CRs (3), PRs/transient tumor regressions (15), and SD (15).^{85,87,90,91,93} In the study with newly diagnosed patients, 6 out 12 patients had PRs (4) or SD (2).⁸⁹ While these findings are promising, especially for patients who have exhausted primary conventional treatments, a randomized phase 2 study has not been conducted.

Unlike the earlier clinical trials with leukocyte and LAK cell-based therapies, at least one patient in every T cell-based therapy trial experienced clinical improvement following immunotherapy.^{16,19,22,84,85,86,87,88,89,90,91,92,93,94,95,96,97,98} Yet, like for LAK cell-based studies, potent and lasting antitumor responses remained rare. Reasons for the limited efficacy of cell-based immunotherapy started to emerge from preclinical studies.⁹⁹ In the 1970s, investigators had discovered that the endogenous T cell response requires the presentation of antigens on major histocompatibility complex (MHC) molecules.¹⁰⁰ Yet, in 2005, one study demonstrated by immunohistochemical staining of 88 primary astrocytic tumors that MHC class I expression was lost in 50% of the GBMs analyzed and that loss correlated with tumor grade.¹⁰¹ Further, MHC class II expression was lost in 70% of GBMs, altogether identifying MHC downregulation as a major mechanism of immune evasion. The secretion of tumor-derived suppressive factors such as transforming growth factor (TGF)-β and prostaglandin E2 was also identified as a mechanism driving the reduced effectiveness of adoptively transferred cells.⁹⁹ T cell-based therapies have experienced a renaissance with the advent of genetic engineering, which is discussed in the CAR T cell therapy section of this review.

NK cells

Since 2004, about 50 patients have been treated with NK cell-based therapies, and the results from these studies are discussed in the following text and summarized in Table S2. Parallel to the rise of TILs and other T cell-based adoptive therapies, NK cell-based therapies emerged as a potential immunotherapeutic strategy for brain tumors. In the first NK cell clinical trial conducted for brain tumors, 9 patients received intravenous and intratumoral autologous NK cells together with IL-2 and IFN-β.²³ Three transient PRs (50% decrease) and 2 minor responses (25% decrease) were observed. In 3 of the responding patients, only the first NK cell dose induced tumor regression, suggesting outgrowth of an NK cell-resistant tumor cell population.²³ Five other NK cell therapy studies for brain tumors have been published.^{23,102,103,104,105,106} One of these studies was the first to explore the use of genetically modified NK cells for brain tumor patients.^102,103,106 In this study, 9 patients with recurrent human epidermal growth factor receptor 2 (HER2)+ GBM were treated with repeated intratumoral infusions of irradiated off-the-shelf HER2-CAR NK-92 cells. Following treatment, 5 patients experienced temporary SD, which lasted between 7 and 37 weeks.¹⁰⁵ In total, transient tumor regressions were observed in 4 out of the 6 brain tumor-specific NK cell studies, but no CRs were observed.^102,103,106 NK cell therapy for brain tumors remains under active preclinical and clinical investigation, particularly since it has become feasible to readily genetically modify NK cells to enhance their antitumor activity.¹⁰⁷

CAR T cells

While clinical responses were observed in the aforementioned early-phase cell therapy trials with antigen-specific T cells for brain tumors, the endogenous immune response was not sufficient to mount potent and lasting antitumor responses.^{16,19,22,84,85,86,87,88,89,90,91,92,93,94,95,96,97,98} The advent of T cell engineering offered a potential solution to this roadblock by not only reliably generating brain tumor-specific T cells but also generating T cells with enhanced effector function. Initial genetic engineering approaches focused on expressing CARs in T cells to generate brain tumor-specific T cells. The original CAR structure was published in 1993 and consists of an antigen recognition domain, a hinge/transmembrane domain, and a cytoplasmic signaling domain.^8,108,109 The antigen-binding domain consists of a single-chain variable fragment (scFv) derived from a mAb or a ligand specific for a tumor-associated antigen expressed on the surface of tumor cells. This enables CAR T cells to recognize tumor cells in a non-MHC-restricted manner, rendering them resistant to immune evasion tactics of tumor cells, including downregulation of MHC.⁸ The drawback of this design is that only the cell surface antigen can be targeted. However, the development of so-called “peptide-centric” CARs enables targeting of peptide derived from intracellular molecules presented by MHC molecules.¹¹⁰ The cytoplasmic domain of CARs consists of signaling domains for T cell activation and costimulation, and the most common domains used are derived from CD3ζ for T cell activation and CD28 and/or 41BB (CD137) for costimulation.

CAR T cells targeting CD19 or BCMA have revolutionized the care of patients with relapsed/refractory B cell malignancies or multiple myeloma over the last 15 years, and currently, there are 7 FDA-approved CAR T cell products.^10,11,111 In contrast, early-phase clinical studies evaluating CAR T cells for solid tumors and brain tumors have shown limited activity unless in the setting of low disease burden.^112,113,114 This limited efficacy is most likely multifactorial and includes (1) low level and heterogeneous antigen expression, (2) limited expansion and persistence of CAR T cells, (3) limited homing to and penetration of tumor sites, and (4) the immunosuppressive tumor microenvironment. These roadblocks have been reviewed in several recent excellent reviews.^{115,116,117,118,119,120,121,122,123,124,125,126} Since 2006, more than 200 patients with brain tumors have been infused with CAR T cells in 17 clinical studies, and we will focus here on reviewing clinical outcomes, which are summarized in Table 3.^{17,18,20,24,25,26,27,28,112,112,128,129,130,131,132,133,134,135,136,137,138,139,140,141,142}

Table 3.

Cell therapy with CAR T cells

Year(s)	Pt (#)	Diagnosis	Target	Treatment	Cell dose	Best clinical outcomes
CD8+ CAR T cell clones
2008,17 2015,24 2017128	3	recurrent GBM	IL13Rα2	Costim: no; LD: no other: HyTK transgene	0.1–1 × 108 (i.t.)	evidence of transient inflammation and targeting of IL13Rα2+ tu cells
2017,128 2022129	6	recurrent GBM		Costim: no; LD: no other: HyTK transgene; KO GR; IL2	0.2– 2 × 108 (i.t.)	2 pts: transient tumor necrosis
Polyclonal CAR T cells
2023,130 2025112	21	DMG	B7-H3	Costim: 41BB T cells: CD4:CD8 ratio, 1:1 LD: no	0.1–1 × 107 (i.c.v.)	1 pt: transient PR (60 days) 15 pts: SD (2 to >37.5 mths)
2021131	1	recurrent GBM		Costim: 41BB T cells: CD25/CD45RA depleted; LD: no	0.4–2 × 107 (i.t.)	tu reduction (50 days)
2025132	4	recurrent or refractory DMG, DHG, ATRT	EGFR806	Costim: 41BB T cells: CD4:CD8 ratio, 1:1 LD: no	1–2.5 × 107 (i.t. or i.c.v.)	1 DMG pt: SD followed by CR (2 years) to subsequent chemotherapy
2017,133 2021134	10	recurrent or progressive GBM	EGFRvIII	Costim: 41BB T cells: bulk; LD: no	1.75–5 × 108 (i.v.)	1 pt: sustained SD (15 mths) 4 pts: transient SD (1–2 mths)
2019135	18	recurrent GBM		Costim: CD28/41BB T cells: bulk; LD: yes other: IL2	1 × 107– 6 × 1010 (i.v.)	none
2024136	7	newly diagnosed GBM		Costim: 41BB T cells: bulk; LD: no other: anti-PD1	0.47–2 × 108 (i.v.)	none
2021137	3	recurrent GBM	EphA2	Costim: 41BB T cells: bulk; LD: yes	1 × 106/kg (i.v.)	1 pt: SD 3 mths
2022,25 2024127	12	H3K27M-mutant DMG, spinal DMG	GD2	Costim: 41BB T cells: bulk; LD: yes (pre i.v.) other: IC9 transgene	1–3 × 106/kg (i.v.) 1–3 × 107 (i.c.v.)	1 pt: CR (>30 mths) 1 pt: >30 mths (no change in tu vol) 3 pts: transient PR (>50%) 3 pts: transient tu vol reduction
2023138	8	recurrent or progressive GBM		Costim: CD28/41BB T cells: bulk; LD: yes other: iC9 transgene	2.5 × 106/kg (i.v.) 1 × 105/kg (i.t.)	4 pts: PR (2–24 mths) 1 pt: SD (4 mths)
201718	17	progressive recurrent GBM	HER2	Costim: CD28 T cells: virus specific; LD: no	0.01–1 × 108/m2 (i.v.)	1 pt: PR (9 mths) 7 pts: SD (8 weeks to 29 mths)
2021139	3	refractory anaplastic A, metastatic EPN		Costim: 41BB T cells: CD4:CD8 ratio, 1:1 LD: no	1–2.5 × 107 (i.t., i.c.v.)	1 pt: SD (<90 days)
2016,20 2024140	65	recurrent HGG	IL13Rα2	Costim: 41BB T cells: TCM, TSCM LD: no	0.02–2 × 108 (i.t., i.c.v.; i.t./i.c.v.)	2 pts: CR (7.5 mths, >43 mths) 2 pts: PR (1 mths, 9 mths) 10 pts: SD (3–15 mths) 16 pts: SD (51–90 days)
2025141	4	recurrent GBM	CLTX	Costim: CD28 T cells: TCM, TSCM LD: no	0.4–2 × 107 (i.t.)	3 pts: SD (1–2 mths)
Polyclonal CAR T cells with genetic modification to enhance antitumor activity
202426	3	recurrent GBM	EGFRvIII EGFR∗	Costim: 41BB T cells: bulk; LD: no other: T cell engager	1 × 107 (i.c.v.)	1 pt: PR lasting >5 mths 1 pt: transient PRs
2024,27 2025142	18	recurrent or progressive GBM	EGFR IL13Rα2	Costim: 41BB T cells: bulk; LD: no other: 2nd CAR	0.5–2.5 × 107 (i.c.v.)	1 pt: PR (3 mths), 2 pts: SD (>7.7 mths, >16.6 mths) 9 pts: SD (1–3 mths)
202428	11	H3K27M-mutant DMG, MB, ATRT	GD2	Costim: 41BB T cells: bulk; LD: yes other: C7R transgene (8)	(i.v.)	2 pts: PR (>1 year, <20 weeks) 5 pts: SD (<30 weeks) 3 pts: transient neurological improvement for <3 weeks

Pt (#), number of patients; A, astrocytoma; ATRT, atypical teratoid/rhabdoid tumor; DHG, diffuse hemispheric glioma; DMG, diffuse midline glioma; EPN, ependymoma; GBM, glioblastoma; HGG, high-grade glioma; MB, medulloblastoma; ∗, targeted with bispecific T cell engager; B7-H3, CD276; EGFR806, epidermal growth factor receptor epitope defined by mAb 806 (cancer specific); EGFRvIII, EGFR variant III (cancer specific); EphA2, EPH receptor A2; GD2, disialoganglioside; HER2, human epidermal growth factor receptor 2; IL13Rα2, IL-13 receptor subunit alpha 2; CLTX, chlorotoxin; Costim, costimulatory domain of CAR; GR, glucocorticoid receptor; HyTK, hygromycin phosphotransferase gene/herpes simplex virus type 1 thymidine kinase fusion gene; iC9, inducible caspase-9 gene; KO, knockout; pre i.v., before i.v. infusion; i.c.v., intracerebroventricular; i.t., intratumoral; i.v., intravenous; CR, complete response; mths, months; PR, partial response; SD, stable disease; tu, tumor; vol, volume.

Monospecific CAR T cells targeting IL13Rα2, HER2, GD2, EGFR variants (vIII; E806), and EphA2 have been explored in early-phase clinical studies in addition to bispecific CAR T cells expressing (1) two CARs specific for IL13Rα2 and EGFR or (2) one CAR specific for EGFRvIII and one bispecific T cell engager specific for EGFR. Second genetic modifications to enhance safety (inducible caspase-9; iC9) or improve antitumor activity (constitutive active IL-7 receptor, C7R) of CAR T cells have also been explored.^{25,28,127,138}

The first brain tumor patient was treated with CAR T cells targeting IL13Rα2 in 2006.²⁴ In this trial, 3 recurrent/refractory grade III or IV glioma patients were treated with repeated intratumoral injections of CD8⁺ CAR T cell clones that expressed an IL13Rα2-CAR with a CD3ζ signaling domain.^17,24,128 While no objective antitumor responses were observed, this study illustrated that repeat CAR T cell administration to brain tumor patients was both feasible and safe. CD8⁺ IL13Rα2-CAR T cell clones were explored in one follow-up study in which the glucocorticoid receptor was knocked out and clones were infused with IL-2 to boost their antitumor activity. However, only transient responses were observed in 2 out of 6 patients.¹²⁹

The first major clinical successes of CAR T cells for brain tumors were observed in a large follow-up phase 1 trial.^20,140 In this trial, polyclonal T cells were evaluated that expressed an IL13Rα2-CAR with a 41BB.CD3ζ signaling domain to not only provide T cell activation but also provide costimulation. A total of 92 patients with recurrent high-grade glioma received IL13Rα2-CAR T cells either into the tumor cavity post resection and/or intracerebroventricularly in this trial between 2015 and 2020. Two patients achieved a CR (7.5 months; 3 years, ongoing), 2 patients achieved transient PRs, 10 patients achieved sustained SD lasting 3–15 months, 16 patients achieved transient SD lasting 51–90 days, and 62 patients had no response.^20,140 This trial is also the first published report of delivering cell therapy to brain tumors using intracerebroventricular (i.c.v.) injection. The trial initially opened as a two-arm study evaluating only intratumoral delivery following biopsy or resection. However, three additional arms involving i.c.v. or dual intratumoral/i.c.v. delivery were later added. The rationale to add i.c.v. dosing stemmed from early clinical experience on the trial involving a patient with recurrent, highly aggressive multifocal leptomeningeal GBM.²⁰ This patient initially received six weekly CAR T cell infusions into the resected cavity of the largest tumor, resulting in stabilization of the local treated site. However, during this initial treatment, the patient’s nonresected tumor continued to progress, and multiple new brain and spine lesions appeared. The patient then received 10 additional i.c.v. CAR T cell injections, resulting in what appeared to be a complete elimination of all the patient’s tumors, which persisted for 7.5 months. This clinical experience, combined with supporting preclinical data suggesting that i.c.v. delivery is more effective for trafficking cells to sites of multifocal disease, provides a strong rationale for further exploration of i.c.v. administration of CAR T cells.^20,143

To date, findings from 18 ongoing and completed early-phase CAR T cell clinical trials for brain tumors have been published.^{17,18,20,24,25,26,27,28,112,127,128,129,130,131,132,133,134,135,136,137,138,139,140,142} Clinical responses have been consistently observed in clinical studies targeting HER2, GD2, and B7-H3 in subset of patients, particularly for pediatric DMG after i.c.v. or intravenous infusion of autologous GD2-CAR or B7-H3-CAR T cells.^{25,112,127,130} This led to a 2-fold increase in median survival in comparison to historical controls (∼20 vs. 10 months). In contrast, clinical responses were only observed in 1 out of 3 clinical studies targeting EGFRvIII.^{133,134,135,136} While these monospecific CAR T cell therapy studies overall demonstrated safety, toxicities were also observed due to local inflammatory responses at the tumor sites that included pseudoprogression. These side effects were coined tumor inflammation-associated neurotoxicity (TIAN) to contrast them from the immune effector cell-associated neurotoxicity syndrome (ICANS) observed in patients after CAR T cell therapy, who do not have brain tumors.¹⁴⁴

Three clinical studies have been conducted that explored a second genetic modification to improve CAR T cell function.^{26,27,28,142,142} In two of the studies, different bispecific CAR T cells were infused: one expressing a 2^nd CAR (EGFR and IL13Rα2) and the other expressing a CAR (EGFRvIII) and a bispecific T cell engager that recognizes EGFR and CD3ε.^26,27,142 In the third study, GD2-CAR T cells were genetically modified to express a constitutive active IL-7 cytokine receptor (C7R).²⁸ The safety and efficacy of i.c.v.-delivered EGFR/IL13Rα2-CAR T cells were evaluated in 18 adult patients with GBM.^27,142 Early-onset neurotoxicity was consistently observed, and several patients had clinical benefit, including one PR, one sustained SD (>7.7 months), and 9 transient SD (1–3 months).^27,142 Three patients with GBM have been reported after the i.c.v. administration of T cells expressing EGFRvIII-CARs and EGFR T cell engagers.²⁶ Early-onset neurotoxicity was observed in 2 out of 3 patients, and one patient had a sustained PR (>5 months) and two transient PRs, respectively.²⁶ Finally, the safety and efficacy of intravenously administered GD2-CAR T cells and GD2-CAR T cells expressing C7R (GD2-CAR.C7R) T cells were compared in one study for pediatric patients with DMG.²⁸ Three patients received GD2-CAR T cells, and eight received GD2-CAR.C7R T cells. Patients receiving GD2-CAR.C7R T cells had improvement from baseline neurologic deficits (2 to >12 months), and PRs were observed in two patients. In contrast, in the GD2-CAR T cell patient cohort, benefits were limited.²⁸

Correlative analyses from clinical trials of CAR T cell therapy for brain tumors have highlighted mechanisms of immune evasion, including the development of antigen-loss variants, adaptive changes in the tumor microenvironment such as an influx of myeloid cells and inhibitory regulatory T cells, and increased expression of inhibitory molecules such as PD-L1, PD-L2, PD1, TIM3, and indoleamine 2,3-dioxygenase 1 (IDO1).^{20,25,130,133,134,136,138}

Discussion and outlook

Cellular therapy has been explored for brain tumors for the last six decades. Early clinical studies aimed to augment antitumor immune response with infusions of nonspecific leukocytes, PBMCs, bone marrow cells, LAK cells, and ex-vivo-activated NK- and T cells. Durable responses were rare, but CRs in a subset of patients were observed, especially when cell therapy was combined with conventional therapy or administered in the upfront setting. However, it is important to acknowledge that many of these studies were conducted using standardized response criteria, such as the McDonald Criteria,¹⁴⁵ which were developed prior to the widespread recognition of pseudoprogression.¹⁴⁶ Pseudoprogression refers to transient treatment-related effects such as enhanced tumor lesions, increased contrast enhancement, and edema that occur in around 20%–30% of brain tumor patients following chemoradiotherapy and/or immunotherapy.¹⁴⁶ These effects can mimic tumor growth on MRI, but unlike true tumor progression, these effects eventually subside without any change in therapy.¹⁴⁶ The recognition of pseudoprogression and advancements in imaging technologies led to the widespread implementation of the more rigorous Response Assessment in Neuro-Oncology Criteria for High-Grade Gliomas (RANO-HGG) recommendations, published in 2010,¹⁴⁷ and the updated RANO 2.0 criteria, published in 2023.¹⁴⁸ Enrollment of patients and interpretations of outcomes from earlier clinical studies were likely impacted by the limited recognition of pseudoprogression at the time. Thus, these earlier historical clinical outcomes should be contextualized in relation to the imaging technologies and response guidelines available at the time, and caution should be exercised when directly comparing the findings to more recent clinical trials.

The advent of genetic engineering enabled the specific targeting of antigens expressed on brain tumor cells with CAR T cells. However, despite being able to directly target brain tumor cells, the clinical activity of CAR T cells has been limited. This lack of efficacy is most likely multifactorial and includes, but is not limited to, heterogeneous expression of the targeted antigens, limited ability of immune cells to traffic to and penetrate brain tumors, and the hostile tumor microenvironment. These roadblocks have been recently discussed in detail in several excellent review articles.^{115,116,117,118,119,120,121,122,123,124,125,126} Despite the caveats of pseudoprogression, many lessons from the earlier cell therapy trials can be applied toward overcoming the barriers to effective CAR T cell therapies. For example, the best clinical responses from the early clinical studies were observed when patients were treated in the upfront setting prior to relapse or when relapsed/refractory patients were treated with cell therapy in the adjuvant setting alongside chemotherapy and/or other immunotherapies.^{62,65,66,76,94,97} Currently, CAR T cells are being largely evaluated in the clinic as monotherapy for the treatment of relapsed/refractory CNS disease.^{17,18,20,24,25,26,27,28,112,127,128,129,130,131,132,133,134,135,136,137,138,139,140,141,142} Thus, moving CAR T cells into the upfront treatment setting and incorporating combinatorial approaches hold promise to improve their efficacy. In addition, steroid use during the month before leukapheresis was associated with poor responses to LAKs.⁸⁰ High doses of corticosteroids are similarly known to negatively impact CAR T cells in preclinical models,^143,149 yet they are currently widely used to manage CNS edema before and during CAR T cell treatment.^{17,18,20,24,25,26,27,28,112,127,128,129,130,131,132,133,134,135,136,137,138,139,140,141,142} Thus, investigation into alternative forms of edema management is warranted to preserve CAR T cell efficacy while managing baseline and treatment-related edema.

Altogether, counteracting these roadblocks with additional genetic modification of CAR T cells and combinatorial therapies holds the promise to improve their efficacy. Likewise, other immune cell subsets, including genetically engineered NK cells, warrant further exploration. We are hopeful that over the next two decades, these refined cell therapy approaches will induce durable remissions in pediatric and adult patients diagnosed with brain tumors.

Acknowledgments

This work was supported by NINDS grant nos. R01NS121249 and R01NS122859 to G.K., NCI grant no. U01CA281823 to G.K., Alliance for Cancer Gene Therapy, Alex's Lemonade Stand Foundation (ALSF), and the American Lebanese Syrian Associated Charities to G.K. and S.G. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

Author contributions

S.M. wrote the first draft of the manuscript and designed the figure. S.M., G.K., and S.G. reviewed and edited the manuscript.

Declaration of interests

G.K. and S.G. have patents or patent applications in the fields of cell or gene therapy for cancer. S.G. is a member of the Scientific Advisory Board of Be Biopharma and the Data and Safety Monitoring Board (DSMB) of Immatics.