The clinical trial in which patients with B cell malignancies participated was conducted at the First Affiliated Hospital of Zhengzhou University and has been reported previously (ClinicalTrials.gov number, NCT03156101).13 End points for this study included safety, overall survival, duration of response, and CD19 CAR-T cellular kinetics. These patients were subjected to lymphodepleting chemotherapy with fludarabine (25 mg/m² body surface area per day) for 3 days, and cyclophosphamide (500 mg/m² body surface area per day) for 2 days, followed by infusion of CD19 CAR-T cells after 3–5 days. Peripheral blood and bone marrow samples were drawn to evaluate the CAR-T and CD19⁺ B cells kinetics. Sequencing of RNA (RNA-seq) and DNA (DNA-seq) was performed to evaluate the CAR-T function and mutation of B cells. Positron emission tomography (PET)-CT or CT was performed to evaluate the overall disease burden. Production of CAR-expressing T cells were produced for the experiments as previously described.13 Patients and healthy donors provided written informed consents according to the Declaration of Helsinki presented at the International Conference on Harmonization Guidelines for Good Clinical Practice. All relevant ethical regulations were followed in this study.

CD19 CARs comprising a single chain fragment variable (scFv) derived from clone FMC63 or 21D4, 4-1BB (CD137) costimulatory, and CD3-zeta domains were generated as previously described.13 Peripheral blood samples drawn from patients or healthy donors were used to produce CD19 CAR-T cells. The transduced T-cells were cultured for about 10 days and the transduction efficiency was evaluated by flow cytometry before conducting infusion.

To generate CD19 or mutated CD19 (p.163. R-L) expressing vectors, the complete gene as well as the CD19 exon 1, 2, 3, or 4 deleted versions were synthesized commercially (Sheng gong, Shanghai) and cloned into pCDH-EF1-IRES-Puro vector. Lentiviral particles were generated by transfection of the above-mentioned constructs into 293 T cells using Lipofectamine-2000 (Invitrogen). Viral supernatants were harvested 48 hours post-transfection and used to infect luciferase expressing H322 or A549 cell lines in the presence of polybrene (4 µg/mL).

A549, H322, 293T, and Raji cell lines were obtained from the Cell Bank of the Chinese Academy of Sciences. Cell lines were cultured in DMEM (Gibco) supplemented with 10% fetal bovine serum (FBS; Hyclone) and antibiotics (Penicillin and Streptomycin; Gibco) at 37°C with a continuous supply of 5% CO₂. These cells were then infected with pLenti-CMV-luc2-IRES-Puro virus to express the firefly luciferase.

Peripheral blood mononuclear cells (PBMCs) were isolated from the peripheral blood samples drawn from patients. Briefly, 1×10⁶ cells were washed with phosphate-buffered saline (PBS) and resuspended in fluorescence activated cell sorting (FACS) buffer consisting of PBS and 2% FBS. Cells were then incubated with anti-CD3-PE-cy7 (clone UCHT1; BD Pharmingen), CD19 (20-291) protein-FITC (ACRO Biosystems), or anti-CD19-PE (clone HIB19; BioLegend) antibodies for 30 min. Flow cytometry analysis was performed on BD FACS CantoII (BD, USA).

The ability of CD19 CAR-T cells (FMC63 or 21D4) to kill the target cells expressing CD19 was evaluated by flow cytometry or bioluminescent imaging (BLI) using the IVIS Spectrum Imaging System (PerkinElemer). Approximately 3×10⁴ CD19⁺ cells from bone marrow pre-infusion and postinfusion of CAR-T cells, CD19-H322 cells, CD19-A549 cells, mutated CD19 (p.163. R-L)-H322 cells, mutated CD19 (p.163. R-L)-A549 cells, and exon 1-deleted, 2-deleted, 3-deleted, or 4-deleted H322-CD19 cells were incubated with CAR-T cells for 6 or 24 hours at various effector:target (E:T) ratios. The lysis of target cells was determined by Annexin-V-PI staining. The BLI was performed and the viabilities of target cells were determined based on their photon intensities. Tumor cells alone were used as the control.

Genomic DNA was isolated from PBMCs that were collected before the treatment and at multiple time-points after treatment by using the DNeasy blood & tissue kit (Qiagen). The kinetics of CAR-modified cells in vivo was assessed using validated PCR primers specific to the 4-1BB and CD3ζ fusion gene as previously described.14 15

CAR-T cells generated from PBMCs before infusion and at day 14 and 180 post-treatment were collected by using the cell sorter Moflo-XDP (Beckman-Coulter). CD19⁺ cells from bone marrow or tumor tissues before treatment and at days 7 and 180 post-treatment were collected to evaluate the molecular characteristics of B cells. Total RNA was isolated using TRIzol reagent (Invitrogen) and treated with RNase-free DNaseI at 37°C for 20 min. Integrity of RNA was assessed on the Agilent TapeStation (RIN), followed by preparation for sequencing using the TruSeq RNA v2 prep (Illumina) as previously described.12 RNA-sequencing and data analysis were performed at BGI (Wuhan, China).

Genomic DNA was extracted from the bone marrow or tumor tissue according to manufacturer’s instructions (Promega).16 17 The enriched DNA fragments were sheared to 150–200 bp fragments on an average and were subjected to standard Illumina Genome Analyzer library preparation according to manufacturer’s protocol. High-quality reads were aligned to the reference human genome (build hg19) using the Burrows-Wheeler Aligner (BWA-v0.7.15). All variants were subjected to visual inspection. Sequencing and data analysis were performed at BGI (Wuhan, China).

We established Raji/Luc lymphoma model via intravenous injection through tail, and CD19-H322/Luc and mutated CD19-H322/Luc solid tumor models via subcutaneous inoculation using 6–8 weeks old female SCID-beige mice. The volume of these cells (1×10⁶) per injection was 100 µL. Thereafter, 1×10⁷ CAR-T cells were resuspended in PBS and injected through the tail vein. Then, the tumor burdens were evaluated by BLI.

All the data were analyzed using Prism V.7.0 (GraphPad Software) and SPSS V.17.0 software. Paired or unpaired two-tailed t-tests were performed to analyze the cytotoxicity of CAR-T cells. Survival curves were analyzed using the log-rank test. Statistical significance was defined as p<0.05.

Patient 1 who was diagnosed with HGBCL showed progression after chemotherapy with R-CHOP, and Hyper CVAD A and B regimen. However, she was found to be refractory to other regimens, including MA (methotrexate and cytarabine), alternating IVAC (etoposide, ifosfamide, cytarabine), and combined treatment with intrathecal injection two times per cycle (methotrexate, cytarabine, and dexamethasone). Approximately 77.6% and 88.4% clonal mature B lymphocytes were found to persist in her bone marrow and cerebrospinal fluid, respectively. Thereafter, she was enrolled in a clinical trial of BinD19, which included treatment with a CD19-directed CAR (FMC63)-T cell product. Following lymphodepletion, she was infused (splitting with dose escalation) with 2.23×10⁸ total T cells, comprising 2.18×10⁷ CAR-T cells for over 3 days. The major therapeutic schedule followed is shown in figure 1A. She had fever on day 3 after infusion of CAR-T cells, as tocilizumab (160 mg), methylprednisolone (40 mg), and dexamethasone (10 mg) were administered repeatedly to control cytokine release syndrome. PET-CT scan images revealed multiple lymphadenopathy and elevated standard uptake values in the breast, uterus, bilateral ovary, liver, bones, and subcutaneous mass before treatment (figure 1B). After infusion of CAR-T cells, she achieved CR by day 60. However, the CD19⁺ relapse occurred in her left breast on day 180. She then received two times the amount of anti-programmed cell death protein 1 therapy and eventually the augmented tumor mass was removed by surgical resection. This patient recovered and exhibited CR state by April 2020 (figure 1A,B).

We assessed CD3⁺ CAR-T cells by flow cytometry in blood sample of patient obtained after infusion. The proliferative burst of CAR-T cells was apparent on day 14 (10.8%) and these cells still persisted until day 180 (3.1 %), which is the period of relapse (figure 1C). After PD-1 blocking therapy and surgery, this patient achieved CR again and the percentage of CAR-T cells was 1.4% and 1.1% in peripheral blood and bone marrow, respectively (day 579, figure 1C). Meanwhile, the expansion and long persistence of CAR-T cells were confirmed via quantitative real-time (qRT)-PCR using CAR-specific primers (figure 1D). We also observed a directed relationship between in vivo CAR-T and abnormal B cells in peripheral blood. Our data indicated that the CD19⁺ B cells were significantly decreased on day 14 (7.7%) and disappeared on day 28 (figure 1C). Additionally, the proportion of malignant CD19⁺ B cells in bone marrow was 61.7% at the baseline that rose up to 95.4% on day 7. However, it decreased to 0.3% on day 17 after CAR-T cell infusion (data not shown). At day 180, the patient experienced relapse, as noted by abundant infiltration (84.1%) of CD19⁺ B cells in the left breast tissues. However, the expression of CD19 was very low in blood (1.0%) and bone marrow (1.2%) (figure 1E). These results suggest that tumor cells could evade CAR-T cell-mediated recognition and clearance, despite the persistence exhibited by CAR-T cells.

To further characterize the CAR-T and B cells derived from the patient at different points, RNA-seq was performed (online supplemental table S1,2). Our results showed that the inhibitor markers, such as LAG3, TIGIT, and CTLA-4 exhibited a lower expression level, but the expression of PDCD1 was higher in CAR-T cells at the time of relapse than that of CAR-T cells in pre group (figure 2A). The transcription factors TBX21, ZEB2, and ID2 are critical for T-cell exhaustion in cancer,18 but their expression had no obvious difference in CAR-T cells on days 14 and 180. Meanwhile, the exhausted marker NR4A1 was highly expressed in CAR-T cells on day 180 compared with that in CAR-T cells before infusion and on day 14, while the expression level of memory differentiation related genes such as EOMES and TCF7 was high in CAR-T cells on day 180, as well as the memory stem cell associated genes such as CCR7, FAS (encoding CD95), PTPRC (encoding CD45RA), and SELL (encoding CD62L, figure 2A). Thus, these genes may cause a difference in persistence of CAR-T cells in vivo. Herein, CAR-T cells at the relapse also expressed high level of GZMA, IFN-γ, GZMB, and RPF1, suggesting that these cells might have possessed potential antitumor ability in vivo. To understand the reason of relapse in this patient, we further analyzed the gene profile and mutations in CD19⁺ B cells derived from bone marrow and tumor tissues. Among the genes whose expression was changed in B cells, some transcription factors captured our attention. For example, the IRF4, MEF2B, BCL6, and TCF3, which are associated with progression of lymphoma, were significantly upregulated on day 180, as well as the B cell receptor (CD79B), Toll like receptor (MYD88), and the NF-kappa B protein (RELA). PRDM1, a master regulator of plasma-cell differentiation, was downregulated in relapsed tumors and its loss may contributed to the poor prognosis of patients.19 Meanwhile, no difference was observed in the expression level of stem cell-related and drug resistance-related genes in bone marrow before and after CAR-T cell therapy (on day 7 and 180, figure 2B).

For further analysis of whether the malignant B cells had a mutation or not, we performed whole-exome sequencing (WES). Interestingly, the p.163 (exon3) R>L site mutation was identified in CD19⁺ cells (95 %) that were isolated from relapsed tumor tissues (figure 2C). However, the CD19⁺ cells derived from bone marrow before therapy had no mutation (data not shown). Further functional analysis of CAR-T cells revealed that specific lysis of CD19⁺ cells derived from bone marrow before treatment was higher than those derived from relapsed tissue, indicating that the relapsed CD19⁺ B cells could not be eradicated by autologous CAR-T cells (figure 2D). We then constructed CD19 CARs comprising scFV antibody fragments, FMC63 and 21D4, and obtained two types of CD19 CAR-T cells. When compared with FMC63, the 21D4 CAR-T cells exhibited better effector-specific function against mutated CD19⁺ B cells derived from tumor tissues (figure 2E). In summary, point mutation of CD19 confers resistance to autologous FMC63 CAR-T cells, and sensitivity to 21D4 CAR-T cells.

To validate the antigen-specific cytotoxic effect of 21D4 CAR-T cells, we conducted an in vitro experiment. Wild type or mutated (exon3, p.163. R-L) CD19⁺ H322 and A549 cells expressing luciferase were established and the expression of CD19 can be detected at the similar level in these tumor cells through labeling with anti-CD19 antibody (HIB19, online supplemental figure S1A, B). CAR-T cells were then incubated with the target cells for 24 hours, and H322 and A549 cells as a negative control. Our results showed that both the FMC63 and 21D4 CAR-T cells exhibited a low antitumor activity against H322 and A549 cells (online supplemental figure S2A–C). As shown in figure 3A, FMC63 CAR-T cells demonstrated potent cytotoxicity against CD19⁺ tumor cells, but not against CD19 mutated cells (10:1, H322-mCD19, 2.90E+07±9.68E+05; A549-mCD19, 2.26E+07±3.23E+06). However, 21D4 CAR-T cells exhibited similar cytotoxicity against both wild type and CD19 mutated cells (10:1, H322-mCD19, 6.25E+06±1.27E+06; A549, 1.11E+07±2.21E+06) at higher E:T ratios (online supplemental figure S2D, figure 3A). Furthermore, the results of CAR-specific lysis assay also supported the higher cytotoxic function of 21D4 CAR-T cells (10:1, 37.9%±5.0%) against CD19 mutated tumor cells than that of FMC63 CAR-T cells (10:1, 19.3%±5.8%). (figure 3B), suggesting that there is a difference between the cytotoxic function of FMC63 and 21D4 CAR-T cells. We next investigated which exon of CD19 is most important for the recognition of CAR-T cells. For this, CD19 exon 1, 2, 3, or 4 was depleted and cloned into pCDH-EF1-IRES-Puro vector. Finally, we obtained four different H322 cell lines with deleted CD19 exon (online supplemental figure S3A, B). Then these cells were cocultured with CAR-T cells at E:T ratio of 10:1. Our results showed that FMC63 CAR-T cells were unable to kill the target cells with 1, 2, or 3 exon deleted, but 21D4 CAR-T cells could still exert an effector specific function against these cells (figure 3C,D). However, neither FMC63 nor 21D4 CAR-T cells could eradicate CD19 positive H322 cells, in which the CD19 exon four was deleted (figure 3C,D). This analysis revealed that the deletion of CD19 exons 1–3 or mutation in exon 3 are sufficient for 21D4 CART-19 cells to exert their cytotoxic function.

Based on the observation that 21D4 CAR-T cells exhibit higher potency against mutated CD19⁺ cells in vitro, we speculated whether CAR-T cells with different scFVs fragments could exert similar outcomes in vivo. 5×10⁶ Raji/Luc cells were injected through tail vein and simultaneously 1×10⁶ wild type or mutated CD19⁺-H322/Luc cells were inoculated subcutaneously into the immune-deficient mice 1 week before CAR-T treatment, respectively. Tumor burdens across the different groups were balanced by BLI analyzes before CAR-T infusion. Thereafter, 1×10⁷ CAR-T cells cultured for 10 days were infused to mice via tail, and mock T cells were used as control. Both FMC63 and 21D4 CAR-T cells demonstrated anti-lymphoma activity in all the treated mice on BLI examination compared with that in uninfected T cell-treated mice (figure 4A). Moreover, FMC63 and 21D4 CAR-T cells effectively controlled CD19⁺ H322 tumor cell growth, while the tumor growth was observed in mice treated with control T cells (figure 4B). We also assessed the antitumor effect of FMC63 and 21D4 CAR-T cells in mutated CD19⁺ H322 model. As shown in figure 4B,C, 21D4 CAR-T cells significantly controlled mutated CD19⁺ tumor growth when compared with FMC63 CAR-T cells. Meanwhile, the 21D4 CAR-T cells treated H322-mCD19 mice had a longer overall survival than the mice treated with FMC63 CAR-T cells (figure 4D). These results suggest that there is a difference in effector-specific function of CD19 CAR comprising scFV antibody fragments FMC63 or 21D4.

Whether this type of mutation appears in patients with NR to CAR-T cell therapy is not clear. Therefore, we further analyzed the mutations in CD19⁺ cells from two different patients with NR and one patient with CR as a control. The clinical characteristics of these patients are presented in table 1. All of the patients exhibited progression after receiving chemotherapy regimen. Patient 2 was treated according to the anti-CD19 CAR protocol and achieved a CR state in 24 months. Two patients with lymphoma resisted the CAR-T cell therapy and died due to disease progression. We detected CAR-T cells by qRT-PCR analysis in the peripheral blood samples of these patients obtained after infusion and observed an increase in the number of CAR-T cells in these patients (figure 5A). Patient 3 with NR exhibited bone marrow involvement on day 37 and died on day 42 after infusion of CAR-T cells. Patient 4 with NR also died on day 87 after treatment with CAR-T cells. We further found that the percentage of malignant CD19⁺ B cells in bone marrow or tumor tissues from the two NR patients after CAR-T cell treatment were 85.1% and 99.8%, respectively (data not shown). Additionally, these two patients possessed a point mutation in exon 3 of CD19 (p.174 L>V, figure 5B). However, the patient with CR had no such mutation. Further analysis revealed that 21D4 CAR-T cells exhibit better effector-specific function compared with FMC63 CAR-T cells when co-cultured with CD19⁺ cells from patients with NR (figure 5C). Altogether, the patients possessing point mutation in CD19⁺ cells (exon 3, p.174 L>V) might exhibit resistance to CD19 CAR-T cell therapy, and these mutated cells can be effectively eradicated by 21D4 CAR-T cells.