CAR T-cell therapy for central nervous system lymphomas: blood and cerebrospinal fluid biology, and outcomes

Therapeutic options are limited for relapsed/refractory (R/R) central nervous system lymphomas (CNSL), including primary CNSL (PCNSL) and secondary CNSL (SCNSL).^1-3 Chimeric antigen receptor (CAR) T cells has been one of the most promising novel cancer therapies. Given that almost all CNSL express the CD19 antigen and that T cells are known to pass the blood brain barrier, there is a strong biological rationale for treating them with CAR T cells but they have been excluded from most clinical trials. Recent data have shown that CAR T cells could be efficient in R/R PCNSL and we recently reported an overall response rate (ORR) of 67% in nine patients.^4–7 However, the dynamics of CAR T cells and their trafficking and persistence in cerebrospinal fluid (CSF) have been rarely described. Identification of biomarkers associated with increased disease control could enhance understanding of the biological basis for efficiency and enables more effective treatment interventions. We investigated the clinical outcomes and the CAR T-cell expansion and phenotype in the peripheral blood and CSF within 21 patients with CD19⁺-R/R CNSL.

Patients with isolated R/R CD19⁺-PCNSL or SCNSL and treated with tisagenlecleucel (n=19) or axicabtagene ciloleucel (n=2) from January 2020 to January 2022 at PitiéSalpêtrière Hospital (France) were retrospectively selected. The disease response was assessed at day 28 (D28) by brain magnetic resonance imaging (MRI) and then every 2 months. Clinical results of eight PCNSL patients were previously reported with shorter follow-up and without any biological data.⁴ All patients gave written informed consent, the study was performed in accordance with the Declaration of Helsinki and was approved by national (CNIL 913611) and local (CPP Ile-De-France PP 13-022) Ethics Committees. Fresh blood samples were collected for realtime quantitative polymerase chain reaction (RT-qPCR) every 3 days the first 2 weeks, every week the following 2 weeks and then monthly. PMBC were collected and used for mass and flow cytometry at the peak of expansion. For detection of integrated CAR-expressing vectors, DNA was extracted from blood using a QIAamp DNA Blood Mini Kit. RT-qPCR was performed on 100 ng of extracted DNA using the 2X TaqMan Universal PCR Master Mix. The specific primers and TaqMan probes detected the CD28-CD3ζ (Yescarta) and the 41BB-CD3ζ (Kymriah) junctions of the CAR T transgene (Online Supplementary Figure S1). For CAR T immune profiling by mass cytometry, thawed PBMC were stained with the CD19 CAR Detection Reagent, Biotin (Miltenyi), washed, and stained with an anti-biotin-^106-116Cd antibody for 20 minutes as previously described.⁸ They were then incubated with the MDIPA panel (Fluidigm), plus five antibodies (anti-PD-1-¹⁷⁵Lu, anti-TIGIT-²⁰⁹Bi, anti-TIM-3-¹⁶⁹Tm, anti-CD69-¹⁶²Dy and anti-CXCR4-¹⁶⁵Ho). Samples were acquired on a Helios machine and analyzed with Maxpar Pathsetter, FlowJo, and OMIQ. For CAR T functionality assessment by flow cytometry, 1x10⁶ thawed PBMC were stimulated 5 hours with a CD19⁺ B-lymphocyte cell line immortalized by Epstein-Barr virus (Raji cell line), then incubated with biotin-labeled CD19 for 30 minutes and stained with Live/Dead, anti-biotin-PE, CD107a-APC-R700, CD3-APC780, CD8-BV605, CCR5-BV650, CCR6-BV700, CD69-PE-CF594, interleukin (IL) 17-BV421, interferon-γ (IFN-γ)-FITC, tumor necrosis factor-α (TNF-α)-PE-Cy7, and IL22-AF647. Fresh CSF samples were collected every 2 weeks during the first month and then monthly, and tested after incubation with biotin-labeled CD19 for 30 minutes, with anti-CD3-APC-H7, CD4-PercPCy5.5, CD8-APC-Alexa700 and anti-biotin-PE. IL-6 levels were measured by Cytometric Bead Array technique in freshly thawed CSF samples, on a FACSCanto II cytometer.

Twenty-one R/R patients (13 PCNSL, 8 SCNSL) were selected, all but one patients with brain parenchymal or meningeal involvement and nine with CSF involvement (Online Supplementary Table S1). The median age was 67 years and the median number of prior therapies three (range, 2–5), including ASCT in 16 patients. At the time of CAR T-cell infusion, 62% had progressive disease (PD), 29% partial response (PR), 5% stable disease and 5% complete response (CR). The median follow-up was 12 months (range, 1–29). At D28, ORR was observed in 67% cases, including 29% CR and 38% PR. At month 3 (M3), ORR and CR without new treatment were observed in nine (43%) (6/13 PCNSL, and 3/8 SCNSL) and six (29%) (5/13 PCNSL, and 1/8 SCNSL) patients, respectively. Among the nine patients with response at M3, the median response duration was 19 (range, 8–29) months. On final follow-up, eight (38%) patients had persistent response: six CR (4/13 PCNSL, and 2/8 SCNSL) and two PR (1 PCNSL and 1 SCNSL), and all the 13 remaining patients died because of R/R disease. The median overall survival was 15 months and tended to be higher in PCNSL than in SCNSL (20 vs. 12 months; P=0.63) (Figure 1), and the median progression-free survival was 3 months. For subsequent biological analysis, patients with response ≥6 months without new systemic treatment were defined as responders (R, n=8), and the others as non-responders (NR, n=13). Cytokine release syndrome and immune effector cell-associated neurotoxicity syndrome occurred in 16 (1 grade 3) and seven (2 grades≥ 3) patients respectively, all having been resolutive using tociluzumab and corticosteroids for 14 and five patients.

We first assessed the expansion of CAR T cells in the CSF for 16 patients. All tested patients demonstrated CSF positivity for CAR T cells regardless of clinical outcome with an initial phase of rapid expansion followed by a slow decrease with long-term persistence (Figure 2). On D30, the median number of CAR-T cells in the CSF was 0.17/mm³, reflecting 19% (range, 4–45) of CD3⁺ T cells, and the median CD4/CD8 CAR T-cells ratio was 4.2. The later CSF analysis assessed 17 months after the infusion in one R patient still demonstrated the presence of CAR T cells (41% of CD3). We demonstrated, focusing on the first month, that the expansion peak in the CSF was significantly higher in R than in NR patients (0.50/mm³ vs. 0.19/mm³ (P=0.01) (Figure 2). Finally, a transient increase in IL-6 dosage was detectable during the first month for 76% cases regardless of their outcomes.

We next addressed the phenotypic characterization of blood CAR T cells at the expansion peak with mass cytometry for 20 patients. We first demonstrated that R patients had a higher frequency of CD8 T cells with a terminal effector phenotype (CD8 TE, mean=24% vs. 11% of cells; P<0.05) and a lower frequency of regulatory T cells (Tregs, 0.5% vs. 1.8%; P< 0.05) compared to NR. We next performed a datadriven non-supervised analysis using the Phenograph clustering algorithm which identified 68 distinct cells subtypes (Figure 3A). Among them, PG-4 which presented a phenotype similar to T cells CD8 TE, confirming our previous observation, and PG-34 corresponding to CD4⁺ CAR T cells expressing high levels of IL7R and IL3 were particularly overrepresented in the R group (Figure 3B). Conversely, three clusters were less present among the responders: PG-31, phenotypically comparable to TIGIT-expressing Tregs, which refines the above findings; PG-8, CD4⁺ T cells positive for interleukin 7 receptor (IL7-R) and PG-13 which fits the description of natural killer (NK) cells. We further applied Phenograph specifically to CAR T cells and we confirmed the over-representation of CD4⁺ IL7R⁺ CAR T cells (PG-CAR13) among the responders (Figure 3C, D). In addition, PD1 showed a weaker expression among responders, and this was particularly striking regarding the CAR T cells (Figure 3E). This reduction of PD1 was paralleled by a CAR T-cell-specific reduction of CXCR4 expression, especially among CD4⁺ CAR.

Figure 1.Overall survival and progression-free survival. (A) Overall survival (OS) and progression-free survival (PFS) in (A) all the patients, (B) in primary central nervous system lymphomas (PCNSL) patients (N=13) and (C) in secondary CNSL (SCNSL) patients (N=8).

Finally, we analyzed the CAR T cells functionality and the expression of chemokine receptors at the peak of expansion by using thawed PBMC (n=16) after 5 hours of stimulation with tumor B cells in order to replicate the in vivo interaction with the CAR T-cell target. We showed that the IFN-γ expression by CD4⁺ CAR T cells was higher in R than in NR patients (P=0.03), and that the expression of CD107a and IFN-γ by CD8⁺ CAR T cells tended to be higher in R patients (P=0.08) (Online Supplementary Figure S2). In addition, the expression of CCR5 and CCR6 on CD8+ CAR T cells was higher (54% and 47%, respectively) than that on the CD8⁺ non-CAR T cells (6% and 4%, respectively; P<0.0001) after the stimulation with tumor B cells. This upregulation on CAR T cells seemed to be the result of the specific costimulation since their expression on the CD8⁺ CAR T cells cultured without target B cells (negative control) was significantly lower (22% for CCR5 and 14% for CCR6; P=0.0096). We first reported persistent response in 38% cases with highly R/R disease, and it is very encouraging in view of the very poor prognosis of CNSL. CNS CAR T-cell trafficking was reported in all of our patients, confirming their ability to migrate into the CNS and persist there despite the low level of target antigen in the blood. Of note, the R patients demonstrated higher CAR T-cell peaks in the CSF, reinforcing the recently published data⁶ but these results should be interpreted with caution due to far apart time points. We next suggested that a strong cytotoxic TE CD8⁺ T-cell response, and a diminished suppressive mechanism mediated by Tregs may be crucial for the efficacy of CAR T cells reinforcing that non-CAR T cells are also critical for the responses.⁹ We further demonstrated that a subset of CD4⁺IL3R⁺IL7R⁺ CAR T cells are associated with a better clinical response. Indeed, IL7 prolongs the survival time of tumor-specific T cells and the effector pool generation, and CAR T cells engineered for IL7-R constitutive activation had a higher anti-tumor activity and persistence in preclinical models.^10–14 On the other hand, the expression of PD-1 and CXCR4 paralleled with a poorer outcome, probably reflecting an impaired capacity to migrate to the target tissue and then to kill the tumor cells. Our results strongly support that engineered CAR T cells might increase the level of response. Finally, we showed for the first time that the production of IFN-γ was associated with clinical outcomes, suggesting that the peripheral immunological invigoration of CAR T reflects the activity at the tumor site. In conclusion, the use of CAR T-cell therapy in CNSL patients answers an unmet medical need, and we suggest that the CSF expansion as well as the functionality and phenotype of CAR T cells are implicated in the clinical outcome, paving the way for the development of novel CAR T cells with higher anti-tumor activity.

Figure 2.Cerebrospinal fluid CAR T-cell expansion. (A) Individual profiles of chimeric antigen receptor (CAR) T cells expansion (% of CD3) in cerebrospinal fluid (CSF) for each patient. Red color denotes responder (R) patients, circle plots denote the patients with primary central nervous system lymphomas (PCNSL) and triangle plots the patients with secondary CNSL (SCNSL). (B) Number of CAR-T cell/mm³ in CSF on the best expansion peak during the first month. (C) Interleukin 6 (IL-6) dosage (pg/mL) in CSF. The timing scale has not been respected for the first 30 days in order to better see all the dots. Red color denotes R patients, circle plots denote the patients with PNCSL and triangle plots the patients with SCNSL Mann-Whitney test.

Figure 3.Immune profiling of peripheral blood mononuclear cells in central nervous system lymphoma patients treated with CAR T cells. (A) t-distributed stochastic neighbor embedding (t-SNE) representation of single-cell cytometry by time-of-flight (CyTOF) data for all patients (N= 20), depicting 68 cell clusters after Phenograph unsupervised clustering based on the expression of each of the 35 markers. Numbers within colored boxes correspond to the respective cluster number. (B) Abundances of selected clusters identified in and comparison of their frequencies between responders (R, red) and non-responders (NR, black) patients. (C) t-SNE representation depicting specific 34 CAR T-cell clusters after Phenograph unsupervised clustering based on the expression of each of the 35 markers. Numbers within colored boxes correspond to the respective cluster number. (D) Abundances of selected clusters identified in (C) and comparison of their frequencies between R (red) and NR (black) patients. (E) Overlayed histograms comparing expression of PD-1 (left) or CXCR4 (right) of gated total CAR T cells, and specific CD8⁺ or CD4⁺ CAR between R (red) and NR (black) patients.

Footnotes

• Received February 23, 2023
• Accepted June 14, 2023

Correspondence

^*CL and JC contributed equally as first authors

M. BARON - marine.baron@aphp.fr

Disclosures

No conflicts of interest to disclose.

Contributions

AG, BA, VV, FN, DR-W and MB designed the study. CM collected and preserved CAR T-cell products. MC performed the cell banking. BF performed the PCR analysis. CL, NT, MB, EL, AV, G.G, MM and JC performed the immunological analysis. ES and MLG-T performed the cytokine dosage in the CSF. CP monitored the flow analysis. DR-W, SC, FN, VM, LS, MU, SN-Q, CL, MB and CS took care of patients. CL and MB collected clinical data. MB wrote the manuscript. All the authors reviewed the manuscript.

Data-sharing statement

All biological data were linked and shared at the Department of Immunology, and of Hematology, Pitié-Salpêtrière Hospital, Paris, France. External users with a formal analysis plan can request access to these data.

Funding

Acknowledgments

We acknowledge all the patients included in the study. We acknowledge the French national expert network for oculo-cerebral lymphomas (LOC) for the recruitment of the patients and the treatment decisions.