CARs construction. The second generation of the anti-CD19 CAR were designed to contain DNA fragments encoding the following components in-frame from the 5’ to the 3’ end: a CD8α signal peptide sequence and the CD19 binding moiety, i.e., a single chain variable fragment (scFv) derived from the sequence of FMC63 mAb (GenBank ID: HM852952.1), followed by a CD8α-based hinge and trans-membrane domain, and the cytoplasmic domains of 4-1BB and CD3ζ. The CAR fragment was then linked with fragments encoding the human IL-7 and CCL19 via 2A peptide sequences (P2A and T2A) to facilitate their co-expression in CAR T cells. The codon-optimized DNA sequences were synthesized (Genscript Co., Ltd.) and assembled into a lentiviral vector backbone (pLenti7.3/V5-DEST Gateway Vector Kit, Thermo Fisher), of which the CMV promoter was replaced by EF1α promoter. To induce a nuclear factor of activated T-cell (NFAT)-inducible expression of IL-7 and CCL19 in CAR T cells, we engineered another expression cassette containing the DNA fragments encoding IL-7 and CCL19 separated by a P2A sequence, under the control of 5 NFAT response elements (REs) and the minimal TATA promoter. The NFAT-RE-IL-7/CCL19 cassette was assembled into the anti-CD19-CAR plasmid by seamless cloning. The final 7 × 19 constructs were propagated in Stbl3 competent cells (GeneCopoeia) and purified using plasmid maxiprep kit (Qiagen).
Lentivirus production and CAR T cell production for preclinical study. CAR T lentivirus was produced by transient transfection of HEK293T/17 cells (CRL-11268, ATCC) with lentiviral plasmid. Briefly, 70% confluent cells were co-transfected via polyethyleneimine (#24765-1, PolyScience) in 150 mm culture dishes with the lentiviral plasmid and the packaging plasmids (Addgene) pMDLg/pRRE (#12251), pRSV-Rev (#12253) and pMD2.G (#12259). Medium was replaced at 24- and 48- hours post transfection. Viral supernatant was harvested 48 and 72 hours post-transfection and concentrated by ultracentrifugation at 25,000 rpm at 4°C for 2 hours, then stored at -80℃ until use.
Peripheral blood mononuclear cells (PBMCs) were isolated from peripheral blood from healthy donors using Ficoll-Paque density gradient centrifugation and activated with anti-CD3/CD28 Dynabeads (Gibco) at 1:1 (bead: cell ratio) in complete AIM-V medium (Gibco) supplemented with 10% fetal bovine serum, 300 IU/mL of recombinant human IL-2 (PeproTech), 5 ng/mL IL-7 and IL-15 (Novoprotein) for 24 hours. After 2 days, the activated T cells were transduced with CAR lentiviral vectors, and then CAR-transduced T cells were proliferated at 0.5 × 106 cells/mL for 10 days.
Flow cytometry analysis. The expression of anti-CD19 CAR in human T cells were detected by immunofluorescence staining with Alexa Fluor 647-conjugated anti-mouse FMC63 scfv monoclonal antibody (BioSwan Laboratories) and flow cytometry analysis. The phenotype of CAR T cells was assessed by immunofluorescence staining and flow cytometry using the following monoclonal antibodies: APC/Cy7- anti-human CD3 (Cat#300426, UCHT1), FITC-anti-human CD4 (Cat#317408, OKT4), PE-anti-human CD8 (Cat#344706, SK1), PE/Cy7-anti-human CD45RA (Cat#304126, HI100) and APC-anti-human CCR7 (Cat#353214, G043H7), Percp/Cy5.5-anti-human CD45RO (Cat#304251, UCHL1), PE/Cy7-anti-human CD197 (CCR7) (Cat# 353226, G043H7). PE-anti-human CD95 (Cat#305608, DX2), APC-anti-human CD27 (Cat# 302810, O323). The expression of immune checkpoint molecules on CAR T cells was analyzed using PE/Cy7-anti-human PD-1 (Cat# 561272, EH12.1) and Brilliant Violet 421 anti-human LAG-3 (Cat# 369313, 11C3C65) antibodies. Dendritic cells were analyzed using PE-anti-human CD83 (Cat#: 305322, HB15e), APC-anti-human HLA-DR (Cat#307610, L243), PE-anti-human CD86 (Cat#305438, IT2.2) and APC-anti-human CD80 (Cat#305220, 2D10) antibodies. Flow cytometric data were acquired with NovoCyte flow cytometer (ACEA Biosciences, Inc, San Diego, California, USA). The fluorophore- antibodies were purchased from BioLegend, Inc. The PE/Cy7-anti-human PD-1 antibody was from BD Biosciences.
Cell lines and in vitro cytotoxicity assay. Raji (human Burkitt's lymphoma cell), Nalm6 (human B-cell lymphocyte leukemia cell), and K562 (human erythroleukemia cell) cell lines were purchased from the American Type Culture Collection (ATCC, Manassas, USA). CD19+ K562 cells were established by the transfection of vector expressing human CD19 antigen. The cells were cultured in RPMI 1640 medium containing 10% FBS and incubated at 37°C with 5% CO2. In vitro cytotoxicity assays were carried out using the luciferase-based cytotoxicity assay described previously20.
CAR T-cell proliferation and apoptosis assay. CAR T-cell proliferation was assessed by arboxyfluorescein succinimidyl ester (CFSE) dilution using CellTrace CFSE Cell Proliferation Kit (Invitrogen). Briefly, cells (1 × 105 cells) were stained with 1 μM CFSE dye at 37°C in dark for 30 min and then co-cultured with mitomycin C-treated Raji cells at the indicated E:T ratio for 24, 48, 72 and 120 hours. The dilution of CFSE was evaluated by flow cytometry analysis. The apoptosis of CAR T cells cocultured with mitomycin C-treated Raji cells for 3, 4, and 8 days was assessed by Annexin-V/7-AAD staining (BioGems, Westlake Village, CA) and analyzed by flow cytometry. The plots are gated on CD3+ lymphocytes.
Transwell migration assay. Monocyte-derived DCs were generated by interleukin 4 (IL-4) and granulocyte-macrophage colony-stimulating factor (GM-CSF) induced differentiation of monocytes isolated from healthy donors. Specifically, PMBCs isolated from blood by Ficoll-Paque density gradient centrifugation were cultured in AIM V medium containing 10% FBS, 100 ng/mL rhGM-CSF and 20 ng/mL rhIL-4 (PeproTech) for 5 days. 20 ng/mL TNF-α (PeproTech) was added to promote maturation of DCs for another 24 hour culture. Monocyte-derived DCs were harvested on day 7 for phenotypic and migration analyses. The migratory abilities of DCs and T cells in response to culture medium of CAR T cells were measured using 24-well transwell chambers (Corning) with a polycarbonate filter of 6.5 μm pore-size. Human T cells labeled with CFSE or DCs (5 × 106/mL) were added in the upper chamber, while the supernatant of CAR T cells stimulated with mitomycin C-treated Raji cells for 5 days was added to the lower chamber for the indicated times. The cells migrated from the upper chamber to the lower chamber were determined by flow cytometry analysis. For antibody blocking experiments, DCs or T cells were preincubated for 30 min with anti-CD127 (10 µg/mL, clone A019D5, Biolegend) or anti-CCR7 (5 µg/mL, clone 150503, R&D Systems) antibodies respectively.
Quantitative PCR (qPCR) analysis. Expression of Bcl-2, Bim and Survivin was quantified with qPCR. Briefly, total RNA was extracted from CAR T cells or control T cells with RNeasy Mini kit (Qiagen) and cDNA was then prepared with PrimeScript RT Regent Kit (TAKARA). PCR was performed on an Applied Biosystems 7500 real-time PCR system using SYBR Premix Ex Taq (TAKARA, Japan) with following PCR primers: Bcl-2: 5’-ACGACTTCTCCCGCCGCTAC-3’ and 5’- TTGACGCTCTCCACACACAT-3’; Bim: 5′- GTTCTGAGTGTGACCGAGAA-3′ and 5’-CTCCTGTCTTGTGGCTCTGT-3’; survivin: 5′-GGACCACCGCATCTCTACAT-3′ and 5′-GTTCCTCTATGGGGTCGTCA-3′; β-actin: 5′-TTGCCGACAGGATGCAGAA-3′ and 5′-GCCGATCCACACGGAGTACT-3.
CD19+ human lymphoma xenograft mouse model. All animal studies were performed with the approval of the Institutional Animal Care and Use Committee of Zhejiang University and in compliance with Chinese National Laboratory Animal Guideline for Ethical Review of Animal Welfare. The 6-8-week-old NSG (NOD.Cg-Prkdcscid IL2rgtm1Wjl/SzJ) mice (Biocytogen) were randomized into three groups and injected intravenously with 2 × 106 of luciferase-expressing Nalm-6 cells. The NSG mice were treated by intravenous injection at the E/T ratio of 2.5:1 with anti-CD19 CAR, 7 × 19 CAR T cells or control T cells on day 7 after Nalm6 inoculation respectively. Bioluminescent imaging was performed to assess tumor engraftment using an IVIS® Lumina LT instrument (PerkinElmer).
Quantification of CAR T cell expansion and trafficking in xenograft mice. To monitor CAR T cells in mice by real-time qPCR. 200 µL peripheral blood was collected by retro-orbital bleeding on day 4, day 12 and day 20. DNA was extracted using QIAamp DNA Blood Mini Kit (QIAGEN), and then amplified with a primer and probe set that is specific for the CD19-CAR. The sequence-specific primers and probes used in animal study were as follows: anti-CD19 CAR forward 5'-TATCGCCACCTATTTCTGCCAG-3' and reverse 5'-TTTCCTGCAGCTTCACTTC
G-3'; probe for anti-CD19 CAR reverse 5’-(FAM)- ACCTTTGGCGGCGGCACCAAGCTGGA-(BHQ1)-3’; β-actin forward 5’-CCACCATGTACCCTGGCATT-3’ and reverse 5’-CGGACTC
GTCATACTCCTGC-3’, probe for human β-actin reverse 5’-(HEX)-CCTGGCCTCGCTGTCC
ACCTTCCA-(BHQ1)-3’. The standard samples were made by a serial 1:2 dilutions of DNA from the infused CAR-T. Based on the percentage of CAR+ T cells determined by flow cytometry as mentioned above, the standard curve was obtained from the percentage of CAR+ T cells and the cycle threshold value. We then calculate the percentage of CAR+ T in the samples from the mice according to the standard curve. All samples were normalized to β-actin. After the CAR+ T cells' ratios were determined, the absolute number of CAR+ T cell was calculated by multiplying the percentage of CAR+ T by the sum of the total number of infused CAR T cells in the mice.
To monitor CAR T cell trafficking in vivo, anti-CD19 CAR T or 7 × 19 CAR T cells and control human T cells were transfected with a lentiviral vector expressing luciferase-GFP and GFP-luc+ cells were purified by FACS sorting to ≥ 98% purity. NSG mice were subcutaneously injected in the left and right flanks with 5 × 106 of Raji cells suspended in 200 µl PBS. When the average tumor volumes reached 100-150 mm3, mice were injected intravenously with 1 × 107 of GFP-luc+ CD19-CAR and GFP-luc+ 7 × 19 CAR T cells. Longitudinal bioluminescent imaging (BLI) was performed to monitor CAR T cell trafficking in vivo.
Study design of phase 1b trial and clinical protocol summary. A prospective, open-label, multicenter clinical trial evaluating anti-CD19 7 × 19 CAR T cells in adult patients with relapsed, refractory DLBCL, PMBCL, tFL, and MCL was conducted after approval by the Ethics Committee of the Second Affiliated Hospital, Zhejiang University and the Institute of Hematology, the First Affiliated Hospital, College of Medicine, Zhejiang University. All patient provided informed consent forms before enrollment in the study. The inclusion and exclusion criteria for this trial were generally the same as ZUMA-1 study42 except that patients had an ECOG performance status score of 0-3. The chemotherapy conditioning regimen consisted of 3 daily doses of 500 mg cyclophosphamide per m2 plus 30 mg fludarabine per m2. We assessed 4 doses of 7 × 19 CAR T cells, including 0.5 × 106, 1 × 106, 2 × 106 and 4 × 106 CAR+ T cells per kg of patient body weight (Supplementary Table 1).
Inclusion and exclusion criteria
Inclusion criteria:
- Patients should provide a written informed consent;
- Patients aged 18 to 75 years old, with ECOG Score of ≤ 3;
- Histologically confirmed CD19+ DLBCL not otherwise specified, MCL, tFL, or PMBCL;
• Confirmation obtained from central pathology review before enrollment;
• Sufficient formalin-fixed, paraffin-embedded tumor samples were required for histologically confirmed diagnosis and detection of CD19 expression;
• Relapsed DLBCL and tFL after ≥2 lines of chemotherapy that include rituximab and anthracycline, or refractory disease as defined in the SCHOLAR-1 study43: progressive disease after receiving ≥ 4 cycles of first-line therapy or stable disease (received 2 cycles of later-line therapy) as best response to chemotherapy or relapse ≤ 12 months after autologous stem cell transplantation (ASCT);
• Relapsed/refractory MCL after ≥ 2 lines of prior therapy, including immunochemotheapy and BTK inhibitor such as ibrutinib, or patient did not agree to receive ibrutinib treatment;
• At least one measurable tumor according to revised International Working Group (IWG) Response criteria43;
4. Life expectancy ≥ 4 months;
5. Adequate cardiac, pulmonary, liver, renal, and bone marrow functions, with the following laboratory values: an absolute neutrophil count > 1,000/mm3, platelets count ≥ 45,000/mm3, and hemoglobin > 8.0g/dl; alanine aminotransferase and aspartate aminotransferase ≤ 2.5 × the upper limit of the normal range (ULN), and total bilirubin ≤ 2.0 mg/dl; a serum creatinine of ≤ 1.5 × ULN; a left ventricular ejection fraction ≥ 50%;
Exclusion criteria:
- Prior treatment that included anti-CD19-targeted therapy, CAR T cell therapy, gene therapy, and allogenic hematopoietic stem cell transplantation (allo-HSCT);
- Chemotherapy other than lymphodepleting chemotherapy, therapeutic doses of steroids, immunosuppressive agent, any radiation therapy or anti-tumor targeted therapy including lenalidomide, bortezomib, ibrutinib, received within 2 weeks before cell collection;
- Clinical trial with investigational drug was performed within 4 weeks;
- History of other cancers;
- Active hepatitis B or hepatitis C. Hepatitis B: HBV-DNA ≥ 1,000 IU/ml; Hepatitis C: HCV RNA positive;
- HIV infection;
- Uncontrollable infection of active bacteria and fungi;
- Currently pregnant or refusal to practice birth control within 1 year;
- Active autoimmune or inflammatory diseases;
- No central nervous system lymphoma.
Manufacturing and immunophenotyping of CAR T cells for clinical trial. Autologous PBMCs were purified by Ficoll-Paque density-gradient centrifugation from leukapheresis products collected on a COBE Spectra apheresis instrument (Terumo BCT). Fresh or thawed PBMCs were suspended in AIM V medium (Gibco) with 10% human AB serum (Sigma) and 300 IU/mL IL-2 for incubation in a 37℃, 5% CO2 humidified incubator. After 6 hours, adherent cells were removed, and then suspension cells were cultured for 24 hours with AIM V medium containing IL-2 (300 IU/ml), 5 ng/mL IL-7 and IL-15 (Novoprotein) and anti-CD3 and CD28 dynabeads with the ratio of 1: 1. The 7 × 19 CAR lentiviral vector was used to transfect activated T-cells in the presence of 8 μg/ml polybrene at 32℃ with 1,200g for 1.5 hours, then stopped by resuspending the cells in fresh complete medium supplemented with IL-2, IL-7, and IL-15 for incubation at 37℃, 5% CO2. The expression of CAR on the surface of T cells was assessed on day 5 by flow cytometry analysis and the CAR T cells were harvested on day 11-15. Cell viability was determined by trypan blue exclusion and the frequencies of naïve (CD45RA+CD45RO-CCR7+CD95-), central memory (Tcm: CD45RO+CD27+), effector memory (Tem: CD45RO+CCR7-), and terminally differentiated effector (CD45RA+CCR7-), CD4+ and CD8+ T-cell subsets in 7 × 19 CAR T cells were evaluated by immunofluorescence staining with respective fluorochrome-conjugated antibodies and flow cytometric analysis. Release criteria for clinical CAR T cell products included the following:
- Cell viability: ≥ 90%;
- CD3+ cells: ≥ 90%;
- Endotoxin: ≤ 0.5EU/mL;
- Mycoplasma: negative;
- Bacterial culture: negative;
- Fungal culture: negative;
- CD3+ CAR+ T cells: ≥ 10%.
Clinical response assessment. Response was determined by whole-body PET-CT at month 3 according to the IWG Response Criteria for Malignant Lymphoma44, and then duration of response was evaluated by ultrasound and CT every 3-6 months until progression.
Toxicity evaluation. All observed AEs were monitored starting on the day of CAR T cell infusion to 1 month after infusion. CRS and ICANS were graded using the ASTCT Consensus grading system25, and all other AEs were graded according to the NCI Common Terminology Criteria for Adverse Events, v.4.03.
Cytokine-release assay. The serum levels of TNF-α, IFN-γ, IL-2, IL-4, IL-6 and IL-10 in patients were assessed using BD™ Cytometric Bead Array Human Th1/Th2/Th17 cytokine kit (BD Biosciences) by flow cytometric analysis according to the manufacturer’s instructions. The data were analyzed using FCAP Assay V3.0.1 software. The levels of IL-7 and CCL19 in culture supernatants of CAR T cells were assessed using Enzyme-Linked Immunosorbent Assay (ELISA) using IL-7 ELISA kit (ExCell Bio) and CCL19 ELISA kit (QuantiCyto).
Multiplex bead immunoassay. Blood samples from 26 patients were collected before and at several time points during the first 28 days after infusion and plasma samples were collected and stored at -80℃. The levels of selected cytokine/chemokine/growth factor in the plasma of patients were assessed by multiplex bead immunoassay using 45-ProcartaPlex Human Cytokine/ Chemokine/Growth Factor Panel (Invitrogen). The panel was used to analyze following cytokine and chemokines: MIP-1α, SDF-1α, IL-27, LIF, IL-1, IL-2, IL-4, IL-5, IP-10, IL-6, IL-7, IL-8, IL-10, PIGF-1, Eotaxin, IL-12p70, IL-13, IL-17A, IL-31, IL-1RA, SCF, RANTES, IFN-γ, GM-CSF, TNF-α, HGF, MIP-1β, IFN-α, MCP-1, IL-9, VEGF-D, TNF-β, NGF-β, EGF, BDNF, GRO-α, IL-1α, IL-23, IL-15, IL-18, IL-21, FGF-2, IL-22, PDGF-BB, and VEGF-α, and the 14-ProcartaPlex Human Immuno-Oncology Checkpoint Panel (Invitrogen) was used for immunological factors including TIM-3, CD28, CD137/4-1BB, CD27, CD152/CTLA4, HVEM, IDO, LAG-3, BTLA, GITR, CD80, PD-1, PD-L1, and PD-L2.
Quantitative PCR analysis of CAR T cells expansion and persistence in patients after infusion. To assess CAR T cells in patients' peripheral blood, genomic DNA was extracted from PBMCs of patients at multiple time-points after CAR T infusion using QIAamp DNA Blood Mini Kit (QIAGEN, Redwood City, California, USA). The percentage of PBMC CAR+ cells was determined by real-time qPCR using anti-CD19 CAR primers and probe as described above. The absolute numbers of CAR+ cells in the blood (μL) of patients were calculated by multiplying the percentage of CAR+ cells by the sum of the absolute numbers of lymphocytes and monocytes/μL as described previously20.
Peripheral blood B-cell quantification. The B-cell count was determined by immunofluorescence staining and flow cytometric analysis using following antibodies from BioLegend: Pacific Blue-anti-CD45 (Cat#304029), APC/Cy7-anti-human CD3 (Cat#300426), PE-anti-CD19 (Cat#302208), FITC- anti-CD20 (Cat#302304), PE/Cy7-anti-human Ig light chain κ (Cat#316520), and APC-anti-human Ig light chain λ (Cat#316610). Live cells were determined by 7-AAD staining, and normal B-lineage cells were defined as CD45+CD3-CD19+CD20+κ+λ+ lymphocytes. The normal range for blood B cells is 61~321/μL26.
Untargeted metabolomics analysis. Blood samples were drawn before and at 1 week, 2 weeks and 1 month after infusion and then centrifuged at 400g for 5 min to extract plasma samples that were stored at -80 ℃. Frozen plasma samples were subjected to untargeted metabolomics analysis. Metabolite Extraction: 100 µl samples were accurately weighed, and the metabolites were extracted using a 400 µL methanol:water (4:1, v/v) solution with 0.02 mg/mL L-2-chlorophenylalanin as internal standard. The mixture was allowed to settle at -10oC and treated by high throughput tissue crusher at 50 Hz for 6 min, followed by ultrasound at 40 kHz at 5oC for 30 min. The samples were placed at -20oC for 30 min to precipitate proteins. After centrifugation at 13,000 g at 4oC for 15 min, the supernatants were carefully transferred to sample vials for LC-MS/MS analysis. UHPLC-MS/MS analysis: Chromatographic separation of the metabolites was performed on a Thermo UHPLC system equipped with an ACQUITY UPLC HSS T3 (100 mm × 2.1 mm i.d., 1.8 µm; Waters, Milford, USA). The mass spectrometric data was collected using a Thermo UHPLC-Q Exactive HF-X Mass Spectrometer equipped with an electrospray ionization (ESI) source operating in either positive or negative ion mode. Data acquisition was performed with the Data Dependent Acquisition (DDA) mode. The detection was carried out over a mass range of 70-1,050 m/z.
Data pre-processing and annotation. After UPLC-MS analysis, the raw data were imported into the Progenesis QI 2.3 (Nonlinear Dynamics, Waters, USA) for peak detection and alignment. The pre-processing results generated a data matrix that consisted of the retention time (RT), mass-to-charge ratio (m/z) values, and peak intensity. Metabolic features detected at least 80% in any set of samples were retained. After filtering, minimum metabolite values were imputed for specific samples in which the metabolite levels fell below the lower limit of quantitation and each metabolic features were normalized by sum. The internal standard was used for data QC (reproducibility), metabolic features which the relative standard deviation (RSD) of QC >30% were discarded. Following normalization procedures and imputation, statistical analysis was performed on log transformed data to identify significant differences in metabolite levels between comparable groups. Mass spectra of these metabolic features were identified by using the accurate mass, MS/MS fragments spectra and isotope ratio difference with searching in reliable biochemical databases as Human metabolome database (HMDB) (http://www.hmdb.ca/) and Metlin database (https://metlin.scripps.edu/). Concretely, the mass tolerance between the measured m/z values and the exact mass of the components of interest was ±10 ppm. For metabolites having MS/MS confirmation, only the ones with MS/MS fragments score above 30 were considered as confidently identified. Otherwise, metabolites had only tentative assignments. A multivariate statistical analysis was performed using ropls (Version1.6.2) R package from Bioconductor on Majorbio Cloud Platform (https://cloud.majorbio.com). Principle component analysis (PCA) using an unsupervised method was applied to obtain an overview of the metabolic data, general clustering, trends, or outliers were visualized. All of the metabolite variables were scaled to unit-variances prior to conducting the PCA. Orthogonal partial least squares discriminate analysis (OPLS-DA) was used for statistical analysis to determine global metabolic changes between comparable groups. All of the metabolite variables were scaled to pareto scaling prior to conducting the OPLS-DA. The model validity was evaluated from model parameters R2 and Q2, which provide information for the interpretability and predictability, respectively, of the model and avoid the risk of over-fitting. Variable importance in the projection (VIP) was calculated in OPLS-DA model. P values were estimated with paired Student’s t-test on Single dimensional statistical analysis. Differential metabolites analysis: Statistically significant among groups were selected with VIP value more than 1 and P value less than 0.05. A total of 2,174 differential peaks were selected from 823 peaks in ESI+ and 1,350 peaks in ESI-. Differential metabolites among two groups were summarized, and mapped into their biochemical pathways through metabolic enrichment and pathway analysis based on database search (KEGG, http://www. genome.jp/kegg/). These metabolites can be classified according to the pathways involved or the functions they perform. Enrichment analysis was usually to analyze a group of metabolites in a function node whether appears or not. The principle was that the annotation analysis of a single metabolite develops into an annotation analysis of a group of metabolites. Scipy.stats (Python packages) (https://docs.scipy.org/doc/scipy/) was exploited to identify statistically significant enriched pathway using Fisher’s exact test.
Statistical analysis. Demographic and disease characteristics were analyzed using descriptive statistics and compared between two groups using a chi-square test. The probabilities of OS and PFS were estimated using Kaplan-Meier method, and survival curves were compared between groups with a log-rank test. PFS and OS were defined as the time from CAR T cells infusion to first relapse or death, with censoring at the last follow-up. DOR was defined only for subjects with a response at day 90 and is the time from the first responses to disease progression or death, with censoring at the last follow-up. ROC curves based on different populations were generated, and the predictive values were then evaluated through examination of the AUC. Specific statistical tests used for in vitro study and untargeted metabolomics analysis are described in the figure legends and Methods. Analyses were performed using GraphPad Prism version 9 software and R version 4.0.3 software. P values < 0.05 were considered significant.