Anti-tumor effects of CD19-directed CAR T-cells against PCNSL have not only been demonstrated in vitro, but also in murine in vivo models [16,17]. Mulazzani et al. designed an orthotopic PCNSL model by combining a chronic cranial window with two-photon intravital microscopy, allowing the repetitive visualization of brain tumor growth [16]. A single dose of intracerebrally injected CD19-directed CAR T-cells was not only able to mediate regression, but also to completely eliminate established PCNSL in two out of three animals. These substantial anti-tumor effects lasted up to half a year until experiments were terminated, and CAR T-cells resided in the brain parenchyma as well as in draining and non-draining lymph nodes throughout the observation period. Importantly, intravenous CAR T-cell injection was associated with a low number of tumor-infiltrating CAR T-cells and therefore not able to sufficiently control PCNSL growth. Although the authors speculated that this might be due to poor trafficking of CAR T-cells across the blood–brain barrier or a rather low number of intravenously injected cells, the final mechanisms behind this observation were not elucidated. Importantly, the study was limited by the use of immunoincompetent mice lacking functional T-cells (but retain B-cells) as human lymphoma cells were utilized.

PCNSL regression after local but not intravenous administration of CAR T-cells was recently corroborated in another immunoincompetent mouse model (lacking function T- and B-cells) of PCNSL [18]. Wang et al. induced orthotopic PCNSL growth by intracranial injection of human lymphoma cell lines. CD19-directed CAR T-cells were either delivered via a single intraventricular or intravenous infusion. Bioluminescence was measured to quantify tumor growth in vivo over the course of weeks, and only intraventricularly injected CAR T-cells were able to control PCNSL growth. Single-cell RNA analysis of CAR T-cells sampled from the bone marrow of post-treatment mice, in vitro culture of CAR T-cells in either cerebrospinal fluid (CSF) or medium, and further mechanistic analyses showed that exposure to the CSF results in a distinct anti-tumor and memory effectivity of CAR T-cells.

These findings made from preclinical studies appear therefore promising in controlling PCNSL; however, they have not yet been validated in immunocompetent animal models. Given that only a limited number of preclinical studies on CAR T-cells and PCNSL is available, a high level of suspicion is therefore required when interpreting these results; however, some anti-tumor effects against PCNSL and CNS lymphoma in general might be assumed.

Patients with active involvement of the brain were excluded from almost all clinical trials on CAR T-cells, mainly due to dreaded more severe neurotoxic side effects. These trials have resulted in US Food and Drug Administration (FDA) approval of CD19-directed CAR T-cells for patients with systemic but not CNS disease [11,12]. So far, only three studies analyzing the clinical efficiency of CAR T-cells in patients with primary or secondary CNS lymphoma are available to date, all of them using CAR T-cells targeting CD19 [19,20,21] (Table 1).

In 2017, a first case report on CAR T-cell efficacy in secondary CNS lymphoma was published [19]. Abramson et al. enrolled a 68-year-old female with refractory diffuse large B-cell lymphoma in the TRANSCEND-NHL-001 trial on the CAR T-cell product lisocabtagene maraleucel (formerly known as JCAR017). After T-cell apheresis and prior to lymphodepletion and CAR T-cell infusion, re-staging studies were provided and a new right temporal mass consistent with disease involvement of the CNS was noted on imaging. The patient proceeded with lymphodepletion and intravenous CAR T-cell infusion (NCT02631044) as initially planned, and complete remission of the cerebral lymphoma site was seen one month after infusion. Of note, this remission was durable and ongoing for 12 months at the time the report was published. Neither cytokine release syndrome nor neurotoxicity was noted.

Another CD19-directed CAR T-cell product, tisagenlecleucel (formerly known as CTL019), has been approved in 2017 for large B-cell lymphoma patients with systemic but also secondary (not primary) CNS involvement. Based on the granted FDA approval, Frigault et al. treated and reported on a retrospective cohort of eight patients with secondary CNS involvement of the brain, spine, and leptomeninges [20]. All patients received lymphodepletion and a single intravenous CAR T-cell infusion of tisagenlecleucel (0.6 × 10⁸ to 6.0 × 10⁸ CAR T-cells). Only mild neurotoxic or systemic side effects were encountered, and none of these patients experienced CAR T-cell-mediated toxicities necessitating therapy with the anti-interleukin 6-receptor antagonist tocilizumab or steroids. Response assessment on day 28 after CAR T-cell infusion showed complete response in two patients, partial response in two more patients, and disease progression in four patients (including two fatalities due to progressive disease). Further follow-up on day 90 revealed ongoing disease control in three of the four patients who initially responded to CAR T-cells, and long-term follow up on day 180 was available in one of those patients showing complete response.

These results suggesting considerable anti-tumor effects in the treatment of CNS disease were recently corroborated by preliminary data from an ongoing prospective trial of CD19-directed CAR T-cells for B-cell non-Hodgkin lymphoma (NCT02153580) [21]. The studied CAR T-cell product is modified to express a truncated human epidermal growth factor receptor, which may serve as an antibody target to rapidly eliminate CAR T-cells in vivo in case of severe toxicities. Three patients with primary and four patients with secondary CNS lymphoma were treated by intravenous CAR T-cell infusion (2 × 10⁸ to 6 × 10⁸ CAR T-cells) following lymphodepletion, whereas when no life-threatening toxicities occurred, tocilizumab was provided for moderate cytokine release syndrome in two patients and steroids for neurotoxicity in three patients. Four patients had disease responses to CAR T-cells with one patient showing complete response and three patients showing partial response. On a cautionary note, follow-up time was only in the range of several weeks and it is unclear whether this response was durable.

Data from longer follow-up intervals after treatment of CNS disease were reported from Li et al. (ChiCTR-OPN-16008526) [22]. One patient with primary and four patients with secondary CNS lymphoma each received one intravenous infusion of CD19-directed (2.2 × 10⁶ to 7.1 × 10⁶/kg body weight) and one infusion of CD22-directed CAR T-cells (3.1 × 10⁶ to 7.0 × 10⁶/kg). CD22 is another pan B-cell marker which offers an additional target in the case of CD19 antigen loss [24]. In this cohort, one case of mild neurotoxic symptoms and one case of high-grade neurotoxicity was encountered which necessitated the use of steroids and plasmapheresis. All 5 patients responded within 60 days after CAR T-cell administration including two complete responses. However, four patients relapsed within three to eight months, and median progression-free survival was three months. Despite tumor relapse, tumor tissue analysis and CSF studies in one patient showed persistent target antigen expression and detectable CAR T-cells. The authors speculated that the immunosuppressive tumor microenvironment providing resistance against CAR T-cells might have contributed to tumor recurrence. However, the authors lacked sufficient evidence for this theory.

Based on the encouraging results of above-mentioned studies, different clinical phase I and phase II trials are currently testing safety and efficiency of CD19 CAR T-cells in primary and secondary CNS lymphoma patients (Table 2).

Primary brain tumors including PCNSL represent complex compositions of neoplastic and non-neoplastic cells which individually contribute to tumor formation [25]. Tumor-associated macrophages and microglia (TAM/M) constitute the majority of non-neoplastic cells in PCNSL, and these cells create a particular immunosuppressive pre-metastatic niche which facilitates tumor cell extravasation, survival, and expansion [26]. A spectrum of TAM/M activation phenotypes have been defined between the pro-inflammatory, anti-tumor M1 and the anti-inflammatory, pro-tumor M2 phenotype [27]. Accordingly, higher numbers TAM/M polarized towards M2 phenotype are associated with less favorable outcome in PCNSL [28]. In addition, immunosuppressive cytokines are strongly expressed in PCNSL, whereas cytokines promoting cell-based immune response are downregulated [29]. To mitigate the immunosuppressive milieu, different approaches including CAR T-cells expressing inducible proinflammatory cytokines [30] or combination therapies with checkpoint inhibitors [31] are currently being investigated in preclinical trials.

In addition to metabolic barriers for tumor infiltration by CAR T-cells, physical barriers including the blood–brain barrier may limit treatment success. Under physiologic conditions, the brain is virtually free of leucocytes, and their influx is tightly regulated. CAR T-cells have been shown to migrate across the blood–brain barrier and can be found in brain and CSF [16,32]. In turn, locally injected CAR T-cells have not only been found to travel to distant sites within the CSF but can also be detected in the systemic circulation [16,33]. After intravenous injection, the number of CAR T-cells within the CSF seems generally lower than in the systemic circulation [32]. In preclinical models, intravenous administration of CAR T-cells for CNS lymphoma but also other brain tumors such as glioblastoma has provided considerable anti-tumor effects [32,34]. However, direct comparison of different routes of application indicate that local delivery may be associated with improved treatment response in brain tumors [16,35]. Importantly, immunodeficient PCNSL mouse models may lack proper function of circulating B-cells and CD19-directed CAR T-cells may therefore not encounter their target immediately after intravenous injection (in contrast to local delivery). This may impair CAR T-cell expansion, and thus underestimate their anti-tumor effectivity. However, insufficient anti-tumor effects after intravenous injection of CAR T-cells have not only been observed in the murine model by Wang et al. [18] who used NOD scid gamma mice (lacking B- and T-cell function), but also in the murine model by Mulazzani et al. [16] who made use of Foxn1^nu/nu mice (lacking T- but not B-cell function). However, the murine CD19 on B-cells from Foxn1^nu/nu mice differs substantially from human CD19 which the CAR T-cells were directed against in the study by Mulazzani et al. It might therefore be indeed speculated that these models may underestimate CAR T-cell effectivity.

Clinical studies for glioblastoma show that local CAR T-cell delivery (delivery into a resection cavity or administration into the CSF) is feasible and effective [16,18,35]. Several clinical studies using local delivery of CAR T-cells to treat primary brain tumors other than PCNSL are currently recruiting (NCT03500991, NCT03638167, NCT04185038). Local injection of CAR T-cells for CNS lymphoma has so far not yet been conducted; however, it might represent an approach warranting evaluation especially for patients with primary CNS lymphoma given the exclusive CNS involvement.

Another factor contributing to recurrence after CAR T-cell therapy may be the loss or downregulation of the tumoral target antigens. Of patients with B-cell leukemia, 7–25% experience CD19-negative relapse [12,36]. The frequency of antigen loss in lymphoma is less clear given that biopsies are rarely obtained during relapse. However, several of such cases have been described after CAR T-cell therapy for systemic lymphoma [37], and antigen loss might therefore be also relevant for PCNSL [38]. In the future, (re-)biopsies of cerebral manifestations should be encouraged for antigenic profiling of the new lesion in order to substantiate the presence of druggable targets. Approaches to prevent or circumvent antigen loss as potential escape mechanism will also need to be evaluated in CNS lymphoma.

CAR T-cell therapy directed against CD19, but also other antigens might be accompanied by unique toxicities including cytokine release syndrome (CRS), immune effector cell-associated neurotoxicity syndrome (ICANS), on-target–off-tumor toxicities, and prolonged cytopenia [39,40].

CRS represents the most commonly encountered adverse effect and is characterized by a systemic increase of pro-inflammatory cytokines translating into sepsis-like symptoms [41]. A high number of up to 93% of patients with extra-axial lymphoma treated with CAR T-cells may experience some degree of CRS, and one out of ten patients may experience severe symptoms necessitating treatment at an intensive care unit [39]. So far, in the treatment of CNS lymphoma, only mild cases of CRS were seen with a frequency similar to what has been observed for systemic disease [11,12,21].

ICANS is the second most commonly observed toxicity following CAR T-cell therapy. Clinical presentation varies and includes moderate symptoms such as headaches, fatigue, or aphasia [42], but also more severe and potentially life-threatening symptoms such as seizures/status epilepticus, cerebral edema, and death [43]. Pathophysiology is likely multifactorial and involves IL-1- and IL-6-mediated systemic inflammation, blood–brain barrier disruption, endothelial activation, and cross-reactivity of CAR T-cells against brain tissue [44,45]. On-target–off-tumor toxicity refers to effects caused by CAR T-cells against non-pathogenic tissue due to shared expression of target antigens on neoplastic and healthy tissue. Parker et al. recently demonstrated by single-cell RNA sequencing and autopsy studies that brain mural cells, which are critical for blood–brain barrier integrity, express CD19 [45]. Thus, an on-target mechanism may contribute to the development of ICANS. The occurrence of severe ICANS has been associated with decreased survival after CAR T-cell therapy [46]. Treatment consists in the application of steroids. Tocilizumab, which can be used to treat CRS, does not seem to improve ICANS. As therapy escalation plasmapheresis, the application of immunoglobulins, and the IL1-antagonist anakinra may show beneficial effects in individual cases [47,48,49]. Although there has been major concern that treatment of CNS disease may be paralleled by strong neurotoxic symptoms, only one case of severe ICANS in a CNS lymphoma patient receiving CD19 and CD22 directed CAR T-cells has so far been described [22]. In clinical trials investigating CAR T-cell therapy for other primary brain malignancies like glioblastoma, only few cases with severe CRS and no clinical manifestation of ICANS were reported [50]. No data exist for locally injected CAR T-cells in CNS lymphoma therapy which would also circumvent direct exposition of brain mural cells to CAR T-cells, and future clinical trials will need to closely monitor for ICANS. In addition, long-term effects of CAR T-cell treatment on cognitive performance have been described [51], and the heavy pre-treatment burden including whole-brain radiotherapy of CNS lymphoma patients may aggravate such effects [7].

Hematological toxicities are among the most common, yet underreported adverse effects of CAR T-cell therapy [40]. As all commercially available CAR T-cell products are exclusively approved for patients with refractory or relapsed malignancies, all patients underwent extensive treatment prior to CAR T-cell infusion which frequently results in profound and long-lasting cytopenia. Moreover, lymphodepletion is usually used prior to CAR T-cell infusion as it may enhance anti-tumor efficacy by decreasing regulatory T-cells and myeloid-derived suppressor cells, increasing levels of proinflammatory cytokines, and enhancing innate immunity [52]. Such an approach is also often associated with profound cytopenia; however, the extent and frequency of observed cytopenia cannot fully be explained by lymphodepletion or prior chemotherapies alone and likely also includes CAR T-cell-mediated mechanisms [53]. A high level of suspicion is required as CAR T-cell patients are therefore highly susceptible for viral, bacterial, or fungal infections [54]. Problematically, therapeutic agents which alleviate cytopenia such as GM-CSF could potentially worsen other toxicities such as ICANS [55]. One could argue that local delivery of CAR T-cells (e.g., intratumoral injection of CAR T-cells or intraventricular application via indwelling ventricle catheter) holds the potential to decrease systemic adverse effects like cytopenia by decreasing circulating CAR T-cells in the blood stream; however, data on that subject remain scarce and future clinical trials will have to address different routes of administration for PCNSL immunotherapy.

Whereas patients with glioblastoma and primary brain tumors other than PCNSL often experience lymphopenia before, during, and after CAR T-cell therapy as stated above, patients with CNS lymphoma are at a particularly high risk due to aggressive myeloablative first-line therapies which also include the use of stem cell transplantation [56,57,58]. In selected patients with low lymphocyte counts, apheresis of an adequate quantity of autologous T-cells for CAR T-cell manufacturing might be challenging. Fortunately, with a continuous improvement in lymphapheresis protocols, sufficient yields of lymphocytes might be expected in most patients [59]. Of note, CAR T-cells from aged donors are of impaired quality including shorter persistence and less memory-like phenotypes and one may speculate that there might also be an association with more extensive pre-treatment burden [60].

Moreover, autoimmune disorders requiring immunosuppressive medication are overrepresented among patients with CNS lymphoma [61]. The pathomechanistic implications of this observations are not fully understood but likely involve Epstein–Barr virus-induced mutations and such patients might be at particular risk for less favorable outcome [62,63]. Importantly, individuals with autoimmune disease were excluded from the landmark clinical trials that resulted in the approval of commercial CAR T-cell products, and there is only little evidence whether CAR T-cell therapy may aggravate symptoms or not [64]. It remains to be shown whether CAR T-cell therapy is safe and beneficial also in patients with CNS lymphoma and a history of autoimmune disease.