Non-viral, specifically targeted CAR-T cells achieve high safety and efficacy in B-NHL

Abstract

Recently, chimeric antigen receptor (CAR)-T cell therapy has shown great promise in treating haematological malignancies^1-7. However, CAR-T cell therapy currently has several limitations^8-12. Here we successfully developed a two-in-one approach to generate non-viral, gene-specific targeted CAR-T cells through CRISPR-Cas9. Using the optimized protocol, we demonstrated feasibility in a preclinical study by inserting an anti-CD19 CAR cassette into the AAVS1 safe-harbour locus. Furthermore, an innovative type of anti-CD19 CAR-T cell with PD1 integration was developed and showed superior ability to eradicate tumour cells in xenograft models. In adoptive therapy for relapsed/refractory aggressive B cell non-Hodgkin lymphoma (ClinicalTrials.gov, NCT04213469 ), we observed a high rate (87.5%) of complete remission and durable responses without serious adverse events in eight patients. Notably, these enhanced CAR-T cells were effective even at a low infusion dose and with a low percentage of CAR⁺ cells. Single-cell analysis showed that the electroporation method resulted in a high percentage of memory T cells in infusion products, and PD1 interference enhanced anti-tumour immune functions, further validating the advantages of non-viral, PD1-integrated CAR-T cells. Collectively, our results demonstrate the high safety and efficacy of non-viral, gene-specific integrated CAR-T cells, thus providing an innovative technology for CAR-T cell therapy.

Fig. 1. Non-viral, AAVS1-integrated CAR-T cells effectively eliminate tumour cells.

a, Specific integration of the CAR cassette into the target locus by homologous recombination through CRISPR–Cas9. b,c, Percentage of CAR⁺ cells (b) and number of viable CAR⁺ cells (c) detected 7 d after electroporation using equimolar amounts of DNA templates with different homology arm lengths (n = 2 independent healthy donors). d, CAR expression in cells from two representative healthy donors determined 7 d after electroporation. SSC, side scatter. K, ×1,000. e, Percentage of CAR⁺ cells detected 7 d after electroporation (n = 23 independent healthy donors). f, Expansion of CAR⁺ cells after repeated stimulation with Raji cells. Data are shown as the mean ± s.e.m. (n = 3 independent healthy donors). g, Median fluorescence intensity (MFI) of CD69, CD137, CD25, PD1 and LAG3 expression in T cells detected after 24 h of co-culture with Raji cells (n = 3 independent healthy donors). CD3⁺ (untreated T, control) or CD3⁺CAR⁺ (LV-19bbz, AAVS1-19bbz) gated cells were analysed. h, Cytokine secretion measured by bead-based immunoassay in the supernatant after co-culture with Raji cells for 24 h. Data are shown as the mean ± s.e.m. (n = 3 independent healthy donors). IL-2, interleukin-2; TNFα, tumour necrosis factor α; IFNγ, interferon-γ. i, In vitro cytotoxicity against Raji cells as determined by lactate dehydrogenase (LDH) assay. E/T ratio, effector/target ratio. Data are shown as the mean ± s.e.m. (n = 3 independent healthy donors). j, Bioluminescence imaging of tumour cell growth following different treatments on the indicated days after CAR-T cell infusion (n = 5). The radiance scale (p s^–1 cm^–2 sr^–1) is shown. Immunodeficient mice were injected intravenously with 2 × 10⁵ firefly luciferase (ffLuc)-transduced Raji cells, and 2 × 10⁶ CAR-T cells were administered intravenously after 5 d. Control samples were electroporated the same as AAVS1-19bbz cells except without single guide RNA (sgRNA) addition. The mean value is shown in b,c,e,g. P values were calculated by one-way ANOVA with Tukey’s multiple-comparisons test (g,h) or two-way ANOVA with Sidak’s multiple-comparisons test (f) or Tukey’s multiple-comparisons test (i). ***P   < 0.001, **P  <  0.01; NS, not significant. Source data

Fig. 2. Non-viral, PD1-integrated CAR-T cells outperform lentivirus-produced CAR-T cells.

a, CAR expression in cells from two representative healthy donors determined 7 d after electroporation. b, Percentage of CAR⁺ cells detected 7 d after electroporation (n = 20 independent healthy donors). c, Percentages of CAR integration and PD1 indels in total T cells detected 7 d after electroporation in five representative healthy donors. d, Percentage of cells with PD1 expression detected by flow cytometry in CD3⁺CAR⁺ gated cells after 24 h of co-culture with PD-L1-expressing Raji cells (n = 3 independent healthy donors). e, Expansion of CAR⁺ cells after repeated stimulation with PD-L1-expressing Raji cells. Data are shown as the mean ± s.e.m. (n = 3 independent healthy donors). f, Cytokine secretion measured by bead-based immunoassay in the supernatant after co-culture with PD-L1-expressing Raji cells for 24 h. Data are shown as the mean ± s.e.m. (n = 3 independent healthy donors). g, In vitro cytotoxicity against PD-L1-expressing Raji cells determined by LDH assay. Data are shown as the mean ± s.e.m. (n = 3 independent healthy donors). h–j, Immunodeficient mice were injected intravenously with 5 × 10⁵ ffLuc-transduced PD-L1-expressing Raji cells, and 1 × 10⁶ CAR-T cells were administered intravenously after 5 d. h,i, Bioluminescence kinetics (h) and imaging (i) of tumour cell growth following different treatments (n = 4 or 8). Data are shown as the mean ± s.e.m. in h. Imaging on the indicated days after CAR-T cell infusion and the radiance scale (p s^–1 cm^–2 sr^–1) are shown in i. j, Kaplan–Meier analysis of survival of the mice in i. Control samples were electroporated the same as PD1-19bbz cells except without sgRNA addition. The mean value is shown in b,d. P values were calculated by one-way ANOVA with Tukey’s multiple-comparisons test (f), two-way ANOVA with Sidak’s multiple-comparisons test (e) or Tukey’s multiple-comparisons test (g), or a log-rank Mantel–Cox test (j). ****P  <  0.0001, **P  <  0.01; NS, not significant. Source data

Fig. 3. Non-viral, PD1-integrated CAR-T cells potently eliminate tumour cells in patients with r/r B-NHL without serious toxicity.

a, Occurrence of CRS and ICANS after treatment. b, Percentage of CAR⁺ cells among the peripheral blood T cells of patients on the indicated days before and after infusion. c, CAR copy number in genomic DNA from the peripheral blood of patients on the indicated days before and after infusion. d, Treatment response and duration of response after infusion. PD, progressive disease. e, PET-CT scans of three representative patients before and after treatment. Red arrows indicate tumour lesions. Source data

Fig. 4. scRNA-seq analysis of non-viral, PD1-integrated CAR-T cells.

a, Percentages of cluster 1 (C1) and cluster 2 (C2) in two donor samples prepared by different methods. C1 and C2 were generated by clustering cells on the basis of expression of CD8⁺ memory and dysfunction/cytotoxicity marker genes, respectively. P1, patient 1; P2, patient 2. b, GSEA of CD8⁺ T cells comparing AAVS1-19bbz and PD1-19bbz cells. Enriched gene sets in PD1-19bbz cells and the normalized enrichment score (NES) are shown. c,d, Violin plots showing the expression of memory (c) and dysfunction/cytotoxicity (d) genes in CD8⁺CAR⁺ cells from three patients before and after infusion. Data for the sample from patient 3 taken after 28 d of treatment are not shown owing to an unreliable low CAR⁺ cell number. D, days after infusion; IP, infusion product. e, GSEA comparing CD8⁺CAR⁺PD1⁺ and CD8⁺CAR⁺PD1^– cells from three patients after 7 or 12 d of infusion. The top six most enriched gene sets in CD8⁺CAR⁺PD1^– cells and the NES are shown. Source data

Extended Data Fig. 1. Optimization of the conditions for producing non-viral gene-specific targeted T cells.

a-j, The sequence of the fluorescent protein mTurquoise2 was used as a target to optimize the conditions for generating non-viral gene-specific integrated T cells. a, Number of viable cells calculated 7 days after electroporation by using different protocols. Equal quantities of circular plasmid DNA and linear double-stranded DNA (dsDNA) were used. Due to acquisition of higher cell viability, templates in the form of linear dsDNA were chosen for all the following experiments. b-c, Homologous recombination efficiency of mTurquoise2 at two PD1 sites (b) and one AAVS1 site (c) by using different DNA templates. HDR, homology directed repair. HITI, homology-independent targeted integration. HITI (pb), HITI template with 50bp protection base pairs flanking the target sequence. MMEJ, microhomology-mediated end joining. d, Homologous recombination efficiency of mTurquoise2 using 33bp or 800bp homology arms. Equal mole or quantity of template harboring 33bp homology arms was used, compared with template with 800bp homology arms. e, Homologous recombination efficiency of mTurquoise2 by using unmodified or modified DNA templates with 200bp homology arms. PS, phosphorothioate. Biotin was modified at the first base pair from the 5’ side. PS was modified at the first three or five base pairs from the 5’ side. f, Homologous recombination efficiency of mTurquoise2 in unstimulated or stimulated T cells using different electroporation programs and DNA templates. g, Homologous recombination efficiency of mTurquoise2 in fresh or recovered T cells after stimulation for indicated days by using HDR templates with 800bp homology arms. h-j, Homologous recombination efficiency (h) and numbers of all viable cells (i) and viable mTurquoise2+ cells (j) were detected using mTurquoise2 templates with different homology arm lengths. k, Number of all viable cells was enumerated using CAR templates with different homology arm lengths. 800bp and 20bp homology arms were used in HDR and MMEJ templates, respectively (a–c, f). Equal moles of DNA template were used in b, c, e–k. The homologous recombination efficiency was determined 7 days after electroporation in b-k. All the experiments were performed in cells from two independent healthy donors. Mean value is shown in all the figures. P values were calculated by one-way ANOVA with Tukey’s multiple comparisons test (g) or two-way ANOVA with Tukey’s multiple comparisons test (e). Source data

Extended Data Fig. 2. Comparison of non-viral gene-specific integrated CAR-T cells and lentivirus-produced CAR-T cells.

a, Percentages of CAR integration and AAVS1 indels in total T cells were detected 7 days after electroporation in five representative healthy donors. b, Cell viability detected by trypan blue staining on indicated days post infection/electroporation. Data are mean ± SEM (n = 3 independent healthy donors). c-d, Absolute (c) and relative (d) rates of T cell growth in vitro (n = 3 independent healthy donors). Data are mean ± SEM in c. e, Representative histogram showing Cell Trace Violet staining of T cells after co-culture with mitomycin C-treated Raji cells for 5 days. This experiment was repeated at least three times using different donors. f-g, In vitro cytotoxicity against Raji cells determined by flow cytometry. f, Representative plots showing lysis of Raji cells following 18 h co-culture. g, The percentage of Raji tumor cell death detected by flow cytometry-based cytotoxicity assay (n = 3 independent healthy donors). h, Bioluminescence kinetics of tumor cell growth following different treatments (n = 5). Immunodeficient mice were injected intravenously with 2×10⁵ ffLuc-transduced Raji cells and 2×10⁶ CAR-T cells were administered intravenously after 5 days. i, Kaplan-Meier analysis of survival of mice in h. j, Percentages of CAR+ cells in CD4+ and CD8+ cells determined 3 or 6 days after infection/electroporation. Data are mean ± SEM (n = 4 independent healthy donors). k, Ratio of CD4+ and CD8+ cells on indicated days post infection/electroporation. Data are mean ± SEM (n = 3 independent healthy donors). l, MFI of CD69, CD137 and CD25 expression in T cells detected by flow cytometry after 24 h co-culture with PD-L1 expressing Raji cells (n = 3 independent healthy donors). m, For the samples of AAVS1-19bbz and PD1-19bbz, CAR+ cells were sorted by fluorescence-activated cell sorting (FACS). Genomic DNA was used as template to amplify PCR products across the homology arms. Sanger sequencing was performed from end to end, outside of homology arms. n-o, Sequences of 5’ and 3’ junction sites between the homology arm and CAR cassette at the AAVS1 (n) and PD1 (o) locus. p, Genotyping of PD1-19bbz cells in two independent healthy donors by calculating the intensity ratio of WT/indel and CAR specific bands in unsorted and sorted CAR+ cells. WT, wild type. q, Non-specific integration of CAR elements was tested 7 days after electroporation by using recombinant spCas9 and different combinations of DNA template and sgRNA (n = 2 independent healthy donors). For the groups of AAVS1, PD1 and TRAC templates, one B2M sgRNA with high cleavage efficiency was used as off-target sgRNA. For the B2M template group, one TRAC sgRNA with high cleavage efficiency was used as off-target sgRNA. The off-target groups were designed to detect non-targeted integration under a hypothesized condition that sgRNA had very high off-target cleavage efficiency. r, PD-L1 expression in Raji cell lines. Control samples were electroporated the same as AAVS1-19bbz or PD1-19bbz cells except without sgRNA addition. CD3+ (Untreated T, Control) or CD3+/CAR+ (LV-19bbz, AAVS1-19bbz, PD1-19bbz) gated cells were analysed in e, l. Mean value is shown in d, g, l, q. P values were calculated by one-way ANOVA with Tukey’s multiple comparisons test (g, l), two-way ANOVA with Tukey’s multiple comparisons test (b, d) and Sidak’s multiple comparisons test (j) or log-rank Mantel-Cox test (i). Source data

Extended Data Fig. 3. Non-viral gene-specific integrated CAR-T cells exhibit similar antigen-independent and -dependent tonic signalling as lentivirus-produced CAR-T cells.

a, MFI of CAR expression in T cells detected by flow cytometry without antigen stimulation (n = 6 independent healthy donors). b, Percentages of PD1 expression in CD4+/CAR+ and CD8+/CAR+ cells detected by flow cytometry without antigen stimulation. c-g, MFI of CD69, CD137, CD25, LAG3 and TIM3 expression in CD4+/CAR+ and CD8+/CAR+ cells without antigen stimulation. h, Percentages of PD1 expression in CD4+/CAR+ and CD8+/CAR+ cells after 24 h co-culture with Raji cells. i-m, MFI of CD69, CD137, CD25, LAG3 and TIM3 expression in CD4+/CAR+ and CD8+/CAR+ cells after 24 h co-culture with Raji cells. The experiments were performed in three independent healthy donors (b-m). Mean value is shown in all the figures. P values were calculated by one-way ANOVA with Tukey’s multiple comparisons test (a, c-g, i-m). Source data

Extended Data Fig. 4. Non-viral PD1-integrated CAR-T cells effectively eradicate tumor cells with either high or low PD-L1 expression at a low or high infusion dose.

a-e, Immunodeficient mice were injected intravenously with 5 × 10⁵ ffLuc-transduced PD-L1 expressing Raji cells and 1×10⁶ CAR-T cells were administered intravenously after 5 days. a, Bioluminescence kinetics of tumor cell growth in each treatment group (n = 4 or 8). b-d, Percentages of PD1, LAG3 and TIM3 expression in peripheral CAR-T cells detected 7 days after CAR-T cell infusion. Data are mean ± SEM (n = 3). e, After CAR-T cell infusion for 7 days, the T cell subset differentiation in peripheral CAR-T cells was analysed according to CD45RO/CD62L expression by flow cytometry. Data are mean ± SEM (n=3). f-i, Immunodeficient mice were injected intravenously with 5×10⁵ ffLuc-transduced PD-L1 expressing Raji cells and 2 × 10⁶ CAR-T cells were administered intravenously after 5 days. The T cells were harvested from the same healthy donor as that in Fig. 2h–j and a–e. f-h, Bioluminescence imaging (f) and kinetics (g, h) of tumor cell growth following different treatments (n = 4 or 5). Data are mean ± SEM in g. i, Kaplan-Meier analysis of survival of mice in f. j-m, Immunodeficient mice were injected intravenously with 5×10⁵ ffLuc-transduced PD-L1 expressing Raji cells and 2×10⁶ CAR-T cells were administered intravenously after 5 days. The T cells were harvested from one healthy donor who was different from that in Fig. 2h-j and a–i. j-l, Bioluminescence imaging (j) and kinetics (k, l) of tumor cell growth following different treatments (n = 4 or 6). Data are mean ± SEM in k. m, Kaplan-Meier analysis of survival of mice in j. n-q, Immunodeficient mice were injected intravenously with 2×10⁵ ffLuc-transduced Raji cells and 1×10⁶ CAR-T cells were administered intravenously after 5 days. n-p, Bioluminescence imaging (n) and kinetics (o, p) of tumor cell growth following different treatments (n = 4 or 8). Data are mean ± SEM in o. q, Kaplan-Meier analysis of survival of mice in n. The imaging on indicated days after CAR-T cell infusion and the radiance scale (p/s/cm²/sr) are shown (f, j, n). P values were calculated by one-way ANOVA with Tukey’s multiple comparisons test (b-d) or log-rank Mantel-Cox test (m, q). Source data

Extended Data Fig. 5. In vitro evaluation of non-viral PD1-targeted CAR-T cell products.

a, Percentage of CAR+ cells in the final products of r/r B-NHL patients (n = 8 patient donors). b, CAR expression determined in three representative patient donors. c, Gating strategy for the detection of CAR expression. d-e, Percentages of CAR integration (e) and PD1 indels (d, e) in the final products (n = 8 patient donors). f, Cell viability of the final products detected by trypan blue staining (n = 8 patient donors). g, IFN-γ secretion measured by ELISA in the supernatant after co-culture with Nalm-6 cells for 18–24 h. Data are mean ± SD (n = 3 technical replicates). h, In vitro cytotoxicity against Nalm-6 cells determined using LDH assay. E/T, effector/target. Data are mean ±SD (n = 3 technical replicates). Mean value is shown in a, d, f. Source data

Extended Data Fig. 6. Off-target detection in non-viral PD1-integrated CAR-T cell products.

a, Off-target detection by whole genome sequencing (WGS) and deep sequencing. The genomic DNA of untreated T cells and the infusion product of patient-2 was subjected to 100× WGS. A total of 2,219 potential off-target sites (not including those around the on-target site) were predicted by Cas-OFFinder and compared with exclusive indels in the edited sample by bioinformatics. No indel events were detected within 15bp upstream and downstream (±15bp) of the sites. Indels were found within 200bp upstream and downstream (±200bp) of eight sites. Deep sequencing (10,000× coverage) was then performed to validate these indel events. While no indels were detected at five sites, indels at the other three sites were variances of one unit length on nucleotide repeats and thus were not considered to be true off-target events. b-e, Identification of off-target sites by iGUIDE. b, Genomic distribution of oligonucleotide (dsODN) incorporation sites by bioinformatic characteristics. The ring indicates the human chromosomes aligned end-to-end, plus the mitochondrial chromosome (labeled M). The frequency of cleavage and subsequent dsODN incorporation is shown on a log scale on each ring (pooled over 10 Mb windows). The purple inner most ring plots all alignments identified. The green ring shows three or more unique alignments that overlap with each other (pileup alignment). The blue ring shows alignments that can be found on either side of a dsODN incorporation site (flanking pairs). The red ring shows reads with matches to the gRNA (allowing < 6 mismatches) within 100 bp (target matched). c, Distance distribution of identified incorporation sites from the PD1 on-target locus. The percentage of incorporations within 100 bp of the on-target site is shown. d, Potential off-target cleavage sites identified by iGUIDE. Abund., the total number of unique alignments associated with the site. MESL, maximum edit site likelihood. Gene_ID, an identifier indicating the nearest gene. Symbols after the gene name indicate: * the site is within the transcription unit of the gene, ~ the gene appears on the cancer-association list. e, Deep sequencing (50,000× coverage) to validate off-target events at the PHACTR1 site in different infusion products.

Extended Data Fig. 7. PD-L1 and CD19 expression in the tumor tissues, serum cytokine profiles and peripheral CD19+ cell percentage in the patients treated with non-viral PD1-targeted CAR-T cells.

a. Representative immunohistochemistry (IHC) staining of PD-L1 protein expression in the patients’ tumor tissues before treatment. b. Evaluation of PD-L1 expression level and percentage in IHC staining. c, Representative IHC staining of CD19 protein expression in the patients’ tumor tissues before treatment and after disease progression or relapse. d, Evaluation of CD19 expression level and percentage in IHC staining. e-l, Serum cytokines including IL-2, IL-4, IL-6, IL-10, IFN-γ, TNF-α and IL-17A were assessed in eight r/r B-NHL patients on indicated days after infusion. m, Percentage of CD19+ cells in the peripheral lymphocytes of patients on indicated months after infusion. Scale bars represent 50 μm. Source data

Extended Data Fig. 8. Single-cell RNA sequencing analysis of CAR-T cells prepared by different methods.

a, Overview of the 63,789 cells that passed quality control (QC) for single-cell analysis. Cells are color coded by sample, respectively, in the t-distributed stochastic neighbor embedding (tSNE) plot. b, Expression of T cell marker genes (CD3D, TRAC) in the tSNE plots. c, Cells are color coded by CD4+ and CD8+ cells in the tSNE plot. d, tSNE plots showing CD4+ and CD8+ cell clusters in each sample. e-f, CD4+ and CD8+ cell proportion in total (e), CAR+ and CAR− (f) samples. g-j, CD8+ T cells were analysed in the samples prepared by different methods. g, Distribution of CAR+ and CAR− cells in the tSNE plot. h, tSNE plot showing two clusters in the samples. Cluster 1 (C1) and cluster 2 (C2) were generated by clustering CD8 memory and dysfunction/cytotoxicity marker genes, respectively. i, Expression of representative CD8 memory genes (SELL, LEF1, IL7R) in the tSNE plots. j, Expression of representative CD8 dysfunction/cytotoxicity genes (LAG3, TIGIT, IFNG) in the tSNE plots. Source data

Extended Data Fig. 9. Proportion of CD8 memory and dysfunction clusters in CAR-T cells prepared by different methods.

a, Heat map showing scaled expression of memory, dysfunction and cytotoxicity genes in two CD8+ T cell clusters. The gene set variation analysis (GSVA) scores of CD8 memory, dysfunction and cytotoxicity signatures are shown at the top. b, tSNE plots showing C1 and C2 in each sample. c, Percentages of C1 and C2 in mixed samples. d-e, Comparison of C1 and C2 proportion between CAR+ and CAR− cells in mixed (d) and individual (e) samples. Source data

Extended Data Fig. 10. Single-cell RNA sequencing analysis of T cells collected shortly after preparation by different methods.

Single-cell RNA sequencing was applied to analyse the characteristics of T cells collected 4 h after preparation by different methods. a, Overview of the 46,558 cells that passed QC for single-cell analysis. Cells are color coded by sample, respectively, in the tSNE plot. b, Expression of T cell marker genes (CD3D, TRAC) in the tSNE plots. c-d, Metabolism (c) and other (d) pathway activities in T cells were scored using the quantitative set analysis for gene expression (QuSAGE) method.

Extended Data Fig. 11. Gene set enrichment analysis of non-viral PD1-integrated CAR-T cells.

a, Gene set enrichment analysis (GSEA) of CD8+ T cells comparing AAVS1-19bbz and PD1-19bbz cells. Enriched gene sets in PD1−19bbz cells and the normalized enrichment score (NES) are shown. b, GSEA of CD8+ T cells comparing the infusion products of patients with different prognosis (patient-1/patient-2, better prognosis; patient-3, worse prognosis). The top ten most enriched gene sets in patient-1/patient-2 group and the NES are shown. c, GSEA comparing different time point samples from patient-1 and patient-2 before and after infusion. CD8+ cells were analysed in the infusion products (IP). CD8+/CAR+/PD1− (left) and CD8+/CAR+/PD1+ (right) cells were analysed in D12 and D28/D29 samples after infusion, respectively. The serial numbers and names of enriched gene sets and the NES are shown. Positive values of the NES (red) represented the enrichment of gene sets in D12 (vs. IP) and D28/D29 (vs. D12) samples. Negative values of the NES (blue) represented the enrichment of gene sets in IP (vs. D12) and D12 (vs. D28/D29) samples. The patient-3 samples were not subjected to this analysis owing to an unreliable low CAR+ cell number in D28 sample.

Extended Data Fig. 12. Landscape of cell types in single-cell RNA sequencing analysis of patient samples.

a-b, Overview of the 54,774 cells that passed QC for single-cell analysis. Cells are color coded by cell type (a) and sample (b), respectively, in the tSNE plots. c, tSNE plots showing cell clusters in each sample. d, Proportion of cell types in each sample. e, Bubble heat map showing marker gene expression for different cell types. f, Overview of the 36,201 cells in the T/NK cell cluster. Cells are color coded by cell type in the tSNE plot. g, tSNE plots showing subtypes in the T/NK cell cluster in each sample. h, Proportion of subtypes in the T/NK cell cluster in each sample. i, Bubble heat map showing marker gene expression for different subtypes in the T/NK cell cluster. Source data

Extended Data Fig. 13. Proportion of CD8 memory and dysfunction clusters in infusion products.

CD8+ T cells were analysed in the infusion products of three patients. a, Distribution of CAR+ and CAR− cells in the tSNE plot. b, tSNE plot showing two clusters in the infusion products. C1 and C2 were generated by clustering CD8 memory and dysfunction/cytotoxicity marker genes, respectively. c, Expression of representative CD8 memory genes (SELL, LEF1, IL7R) in the tSNE plots. d, Expression of representative CD8 dysfunction/cytotoxicity genes (LAG3, TIGIT, IFNG) in the tSNE plots. e, Heat map showing scaled expression of memory, dysfunction and cytotoxicity genes in two CD8+ T cell clusters. The GSVA scores of CD8 memory, dysfunction and cytotoxicity signatures are shown at the top. f, Percentages of C1 and C2 in mixed and individual samples of infusion products. g, Comparison of C1 and C2 proportion between CAR+ and CAR− cells in mixed and individual samples of infusion products. Source data

Extended Data Fig. 14. Expression of CD8 memory and dysfunction/cytotoxicity genes in non-viral PD1-targeted CAR-T cells before and after infusion.

a, Heat map showing scaled expression of memory, dysfunction and cytotoxicity genes in CD8+/CAR+ cells from three patients before and after infusion. The GSVA scores of CD8 memory, dysfunction and cytotoxicity signatures are shown at the top. b,c, Violin plots showing the expression of memory and dysfunction/cytotoxicity genes in CD8+/CAR+ cells from individual (b) and mixed (c) samples of three patients before and after infusion. d, Violin plots showing the expression of memory and dysfunction/cytotoxicity genes in CD8+/CAR+ and CD8+/CAR− cells from two patients after 28 or 29 days infusion. The data of the patient-3 sample taken after 28 days treatment is excluded from mixed samples and not shown individually owing to an unreliable low CAR+ cell number.

Extended Data Fig. 15. Pathway activities in non-viral PD1-integrated CAR-T cells before and after infusion.

a-b, Metabolism (a) and other (b) pathway activities were scored using the QuSAGE method in CD8⁺/CAR⁺ cells from three patients before and after infusion. The data of patient-3 sample after 28 days treatment is not shown owing to an unreliable low CAR⁺ cell number.