Peripheral blood samples were obtained from patients with diagnosed AML as well as from young (below 30) and aged-matched healthy donors. Sample collection was conducted in accordance with the principles of the Declaration of Helsinki and with the approval of the Research Ethical Committee of the University of Navarra. All subjects provided written informed consent. Jurkat-TPR cells (kindly provided by Dr. P. Steinberg; Medical University of Vienna) were cultured in RPMI 1640 (Lonza) supplemented with 10% FBS (Gibco). MOLM-13 was cultured in RPMI 1640 supplemented with 20% FBS. HEK293T cells were cultured in DMEM supplemented with 10% FBS. All media were supplemented with 1% penicillin/streptomycin (Gibco) and 1% L-Glutamine (Gibco). All cell lines were maintained at 37°C in 5% CO₂.

A third-generation self-inactivating lentiviral vector (pCCL) was used to express under the EF1a promoter second-generation CAR constructs targeting CD33, derived from M195 (23) or my96 (24) antibodies. CAR structure comprised the single-chain variable fragment (scFv), a panel of different hinge regions derived from CD8a or IgG4, a CD8 transmembrane domain, 4-1BB or CD28 co-stimulatory domain, and CD3ζ endodomain ( Figure S1A ). All constructs included an huEGFRt transduction marker separated from the CAR gene by a viral 2A sequence. Lentiviral vectors were produced in HEK293T cells following standard procedures. Briefly, 6×10⁶ cells were co-transfected with LV vector along with pMDLg/pRRE (Gag/Pol), pRSVRev, and pMD2.G (VSVG envelope) packaging plasmid using Lipofectamine 2000 (Invitrogen). Supernatants were collected 40h after transfection, filtered, concentrated using Lenti-X Concentrator (Takara) following manufacturer specifications, and stored at -80°C until use.

Jurkat-TPR cells, transduced at MOI of 1 with CAR constructs targeting CD33 were co-cultured in triplicate with MOLM-13 cells, at a 1:1 effector to tumor cell ratio. Non-transduced Jurkat-TPR cells or transduced with a previously described CAR construct targeting CD19 (25) were used as control. Activation of the NFAT, NF-κB, and AP-1 pathways was quantified before and 24h after co-culture with tumor cells measuring eGFP, eCFP, and mCherry emissions respectively, using a CytoFLEX LX Flow Cytometer (Beckman Coulter) ( Figure S1B ).

Peripheral Blood Mononuclear Cells (PBMCs) were isolated with Ficoll-Paque and CD4⁺ and CD8⁺ T cells were selected using CD4 and CD8 MicroBeads (Miltenyi Biotec) in the AutoMACS Pro Separator (Miltenyi Biotec). Isolated T cells were activated with 10 µl/ml T cell TransAct (Miltenyi Biotec) for 48h and infected with the CAR lentiviral vector at MOI 2 with 10 µl/ml of LentiBoost (Sirion Biotech). CAR-T cells were expanded during 10-12 days in RPMI 1640 culture medium supplemented with 3% human serum (Sigma), 1% penicillin/streptomycin, and 625 IU/ml of human IL-7 and 85 IU/ml of human IL-15 (Miltenyi Biotec). CAR-T cells were counted, and the concentration was adjusted to 1×10⁶ cells/ml every two days.

Phenotypic characterization of T cells and CAR-T cells was performed at day 0 and 14 of the production, respectively. All antibodies were purchased from Biolegend unless otherwise stated ( Table S1 ). Data was acquired on a BD FACSCanto II (BD Biosciences) and analyzed using the FlowJo Software version 10 (Tree Star).

Cytotoxicity was determined using MOLM-13-GFPLuc as target tumor cells. Briefly, MOLM-13-GFPLuc cells were cultured with CAR-T cells at different ratios in RPMI 1640 culture medium, supplemented with 3% human serum (Sigma) and 1% penicillin/streptomycin in Nunc™ 96-Well Round Bottom plates (ThermoFisher Scientific). After 24h, luminescence was measured using the Bright-Glo™ Luciferase Assay System (Promega) according to the manufacturer’s instructions. IFN-γ cytokine production was quantified using BD™ Immunoassay ELISA reagents (BD Biosciences) following manufacturer protocol.

CAR-T cells were co-cultured with irradiated MOLM-13-GFPLuc tumor cells for 21 days. Briefly, 1×10⁶ cells MOLM-13-GFPLuc were irradiated at 54 Gy to prevent tumor growth, and co-cultured with CAR-T cells, at a 1:1 effector: target ratio of cells, in RPMI culture medium. Every three days, CAR T cells were counted and irradiated tumoral cells were added at a 1:1 ratio. On day 21, the phenotype of CAR-T cells was studied.

All experimental procedures were approved by the Ethics Committee of the University of Navarra and the Institute of Public Health of Navarra according to European Council Guidelines. NOD-SCID-Il2rg^−/− (NSG) mice were purchased from The Jackson Laboratory (JAX) and bred and maintained in-house in a pathogen-free facility. Eight-to-twelve-week-old male or female mice were irradiated at 1.5 Gy at day -1 and 5×10⁴ MOLM-13-GFPLuc cells were intravenously injected the following day. Mice were randomized to ensure equal pre-treatment tumor burden before CAR-T cell treatment. At day 4 mice received i.v. injection of 3×10⁶ CAR-T cells. Mice were humanely euthanized when mice demonstrated signs of morbidity and/or hindlimb paralysis.

RNA-seq was performed following the MARS-seq protocol adapted for bulk RNA-seq (26, 27) with minor modifications. RNA-seq libraries quantification was done with Qubit 3.0 Fluorometer (Life Technologies) and size profiles were examined using Agilent’s 4200 TapeStation System. Libraries were sequenced in an Illumina NextSeq500 at a sequence depth of 10 million reads per sample. Samples were aligned to the Human genome (GRCh38) with STAR (v2.6.1). Gene expression was quantified with quant3p (github.com/ctlab/quant3p). Downstream analyses were performed in R (v3.6.2). Data transformation, normalization, and differential gene expression analysis were performed with DESeq2. Genes were considered adjusted p value (p-adj)<0.05 and absolute log transformed fold-change (|log2FC|)>1, unless otherwise indicated. vst expression values were used for data visualization and unsupervised analysis. Stem cell memory, T cell activation, and exhaustion gene signatures used in this work were obtained from previous publications (28–31) ( Table S2 ). The normalized gene expression matrix was used to discover the disrupted genes after stimulation of AML CAR-T cells in comparison to adult and senior CAR-T cells. These genes were found using the maSigPro package (version 1.72.0) (32) which applies a negative binomial model to the expression distribution and adjusts the false discovery rate using the Benjamini and Hochberg procedure. The degree of polynomial regression in this study was set to 2, and the two ways forward elimination algorithm was used to perform stepwise regression to select genes with alpha equal to 0.05. To extract the significant genes upon stimulation the following settings were used: min.obs=2, and rsq=0.7.

sgRNAs targeting exon 1 of the beta-2-microglobulin (B2M) gene and exon 1 of T-cell receptor α constant (TRAC) locus were designed and selected as described previously (33) using Benchling software (www.benchling.com). Sequences for B2M and TRAC sgRNAs can be found in Table S3 . In vitro cleavage efficiency was evaluated by TIDE (6) after transfection of Streptococcus pyogenes Cas9 (SpCas9) and sgRNA ribonucleoprotein complexes (RNP) in the Jurkat cell line. SpCas9 protein and sgRNAs were purchased from IDT. 2x10⁶ Jurkat cells were electroporated with 61 pmol of RNP (ratio 1:1 Cas9:sgRNA) using the SE Cell Line 4D-Nucleofector Kit and the CL120 program on a 4D-Nucleofector System (Lonza) according to manufacturer’s instructions. Genomic DNA was isolated with NucleoSpin Tissue for DNA extraction kit (Macherey-Nagel) 72h after electroporation and subjected to targeted PCR amplification using primer described in Table S4 . Indel percentage was calculated using the TIDE webtool (https://tide.nki.nl).

Minicircle (MC) encoding CD33 targeting CAR was generated from parental pT2 plasmids by PlasmidFactory, using site-specific recombination and purified by affinity chromatography. Poly(A)-tailed ARCA-capped SB100X mRNA was produced by in vitro transcription from the T7-SB100X plasmid (Addgene #34879) using the mMESSAGE mMACHINE kit and column purified using the MEGAclear kit (Ambion).

CD4⁺ and CD8⁺ T cells were isolated from PBMCs and activated as described above. 2x10⁶ T cells were electroporated 48h after activation with 1µg of MC, 1 µg of SB100X mRNA, 3µM of each sgRNA (targeting TRAC and B2M), and 1,5µM Cas9 previously mixed as an RNP, using the ExPERT GTx™ electroporation device from MaxCyte, according to manufacturer’s instructions. CAR-T cells were expanded in TexMACS™ culture medium (Miltenyi Biotec) supplemented with 3% human serum (Sigma), 1% penicillin/streptomycin, and 625 IU/ml of human IL-7 and 85 IU/ml of human IL-15 (Miltenyi Biotec). HLA-I^KO/TCR^KO CAR-T cells were negatively selected using the AutoMACS Pro Separator after incubation with anti-human HLA-I and TCRα/β antibodies and Anti-Biotin MicroBeads (Miltenyi Biotec) according to manufacturer’s instructions. For large-scale productions conditions were scaled proportionally to the number of cells.

SB copy number (CN) per cell was determined by qPCR. Genomic DNA was extracted using the DNeasy Blood and Tissue Kit (Qiagen). CN/cell was quantified by duplex detection of the WPRE sequence, normalized to ALBUMIN, using specific primers, and detected with the TaqMan probes ( Table S4 ). qPCR was performed using the Absolute qPCR Mix Low ROX Mix (Thermo Scientific) in a QuantStudio™ 3 Real-Time PCR System (Thermo Fisher Scientific). Results were analyzed in QuantStudio 3 Design and Analysis Software (Thermo Fisher Scientific).

Integration site analysis was performed in genomic DNA isolated with NucleoSpin Tissue for DNA extraction kit (Macherey-Nagel) from CAR-T and CAR-T^KO cells (3 independent productions) using INSPIIRED pipeline (34, 35) with minor modifications as described (20). Sequencing was carried out in an Illumina MiSeq at a depth of 3-10x10⁵ reads per sample using MiSeq Reagent Kit v2 300-cycles (Illumina) (Rd1: 179c; Index1: 12c; Rd2: 143c). Raw sequencing data was demultiplexed and trimmed of ITR-specific sequences. Then, sequences were filtered against the vector sequence. The remaining unique sequences were aligned to the Human genome (GRCh38) reference using BLAT. Alignment for R1 and R2 sequences were then joined together and filtered for quality alignments, yielding unique sites of integration or multihit locations. Data were stored within an R object. Unique sites were annotated using the clusterProfiler library. The virtual machine, software, and instructions are available at https://github.com/BushmanLab/INSPIIRED.

Libraries were prepared following the protocol described in iGUIDE (36). Genomic DNA from samples was purified with NucleoSpin Tissue for DNA extraction kit (Macherey-Nagel) and randomly fragmented by ultrasonication. Adapters were ligated to end-repaired DNA, and targeted DNA was amplified through a nested-PCR from the incorporated dsODN to the ligated adapter sequence. Amplicons were purified and sequenced on an Illumina MiSeq with 300 cycle v2 reagent kits. Output sequence data was analyzed using the iGUIDE pipeline. The software and instructions are available at https://github.com/cnobles/iGUIDE.

Statistical analyses were performed using GraphPad Prism for Mac version 10.0.0. The different tests used in this work are indicated in the figure legend.

A panel of 12 different 4-1BB second-generation CAR constructs targeting CD33, derived from 2 different monoclonal antibodies (scFv from my96 and M195) and presenting hinge regions with different lengths from CD8a or IgG4 molecules were generated ( Figure S1A ). CAR design also included a truncated version of the human EGFR (hEGFRt) as a reporter gene to facilitate tracking of the CAR-T cells (37). The specific activation of the main signaling pathways (NFAT, NFkB, and AP1) after tumor recognition was measured using a triple reporter system in Jurkat cells (Jurkat-TPR) (38, 39). After transduction with lentiviral particles coding for the different CAR constructs Jurkat-TPR was cocultured with MOLM-13 cells, an AML cell line expressing CD33 established from the peripheral blood of an acute monocytic leukemia relapsed patient (40). Non-transduced cells (UTD) or transduced with a CAR targeting CD19 were used as controls. The top three CAR constructs [my96(45aa), my96(228aa), and M195(45aa)] showing the highest specificity without tonic signal were selected for further in vitro and in vivo functional analysis ( Figure S1B ). CAR-T cells with the selected CAR constructs were generated from healthy donors and functionally characterized. Transduction efficiency (CAR⁺ cells) and expansion capacity of CAR-T cells were similar between the three selected constructs. Evaluation of their in vitro lytic activity by regular cytotoxic assay against MOLM-13 cells showed that all three CAR-T cells effectively killed tumor cells at low E:T ratios. The in vivo antitumoral efficacy was evaluated in a xenograft model in NSG mice by administration of 3x10⁶ CAR-T cells/animal 4 days after intravenous administration of 5x10⁴ cells/animal MOLM-13 cells expressing luciferase. CAR-T cells derived from my96 antibody presented a statistically significant increased antitumor efficacy in vivo compared to M195(45) CAR-T cells. Finally, the my96(45) CAR, containing the 45aa hinge from CD8a, was selected for the rest of the studies ( Figures S1C–F ).

Since AML is associated with the presence of abnormal T cells phenotypes (41), we performed a phenotypic characterization of T cells collected from a cohort of 21 AML patients ( Table S5 ). When compared to T cells from young adults (below 30) or aged-matched healthy donors (senior), T cells from AML patients presented a more differentiated phenotype, particularly in the CD8⁺ compartment, with significant enrichment of effector (T_E) cells. In contrast, T cells from healthy donors were enriched in naïve T cells (T_N) ( Figures 1A, B , S2 ). Moreover, T cells from AML patients also displayed significantly higher expression of activation (CD69, HLA-DR) and exhaustion (PD1, LAG3) markers compared to healthy donors ( Figures 1C , S2 ). To determine whether differences in phenotype and activation may play a role in the functionality of CAR-T cells, using our selected CAR construct, we generated CAR-T cells (CD33-CAR-T) from AML patients that were compared to CD33-CAR-T cells from young (adult) and elderly (senior) healthy donors ( Table S5 ). Non-transduced T cells (UTD) from each group were used as controls. The proliferation capacity of the CD33-CAR-T and UTD cells from the different groups was similar, reaching 4-5 population doublings during the expansion phase. Moreover, similar transduction efficiency was observed between the different groups, with around 50% of CAR⁺ cells, suggesting that transduction was not impaired in AML T cells ( Figures 2A, B ). Phenotypic analysis revealed that CD33-CAR-T cells from both senior donors and AML patients presented a statistically significant reduced percentage of memory stem (T_SCM) and central memory T cells (T_CM), with increased levels of activation markers like CD69 and HLA-DR. However, CD33-CAR-T cells from AML patients presented an increased proportion of terminal effector (T_E) T cells, with higher levels of PD1 and LAG3 exhaustion markers ( Figures 2C, D , S3 ). Altogether, these data suggest that the initial T cell phenotype has a clear impact on the final phenotypic features of the CAR-T cells, with significant differences in the differentiation, activation, and exhaustion states produced by both age-related and AML-specific factors.

Next, we decided to analyze the in vitro and in vivo functional capacity of the generated CAR-T cells. Regardless of their origin, all CD33-CAR-T cells were highly cytotoxic, producing similar levels of IFN-γ ( Figures 3A, B , S4 ). To further explore CAR-T cell functionality under more challenging conditions, a continuous repeated in vitro stimulation for 21 days with tumoral cells was performed. Under these conditions, we found that CD33-CAR-T cells from AML patients presented a more exhausted phenotype, with increased PD1 and LAG3 expression, and a decreased proliferation potential that was statistically significant ( Figures 3C, D , S4 ). These features were also associated with reduced cytotoxic capacity after repeated restimulations ( Figure 3E ). The antitumoral potential of the CAR-T cells was then evaluated in vivo in a xenograft model of NSG mice transplanted with MOLM-13 cells, as described above. Treatment with all CD33-CAR-T cells significantly increased animal survival compared to UTD controls with no differences between male and female animals ( Figure S4 ). Interestingly, CD33-CAR-T cells from AML patients showed a statistically significant reduction of the antitumor activity with reduced survival compared to CAR-T cells from healthy donors (adult and senior) ( Figure 3F ). To address whether the decreased capacity of CD33-CAR-T from AML patients could be reverted by modifying the design on the CAR construct, we replaced 4-1BB sequence by CD28 costimulatory domain and compared CD33-CAR-T cells generated from AML patients using both constructs. Similar transduction efficiency and proliferation potential were observed between 4-1BB and CD28 CAR-T cells, with no differences in the phenotypic subpopulations, activation/exhaustion markers, cytotoxic capacity, or IFN-γ production after coculture with tumoral cells ( Figure S5 ). Finally, both 4-1BB and CD28 CAR-T cells presented similar antitumoral efficacy in vivo in the xenograft model in NSG mice ( Figure S5 ). Altogether, these data strongly suggest that the CD28 costimulatory domain does not improve the functionality of CAR-T cells from AML patients.

Given the phenotypic and functional differences observed between CAR-T cells from AML patients and healthy donors, we further analyze CD4⁺ and CD8⁺ CD33-CAR-T cells at the transcriptomic level. RNAseq analysis before antigen recognition revealed that CAR-T cells from AML patients were transcriptionally similar to those from senior healthy donors, with only a few differentially expressed genes (DEGs) with an FDR<0.05 and a |log2FC|>1 (31 for CD4⁺ and 12 for CD8⁺ CAR-T cells) ( Table S6 ). In contrast, most of the differences were observed when CAR-T cells from senior healthy donors and AML patients were compared to CAR-T cells from adult healthy donors (752 and 853 DEGs in CD4⁺ and 807 and 705 in CD8⁺ CAR-T cells respectively) ( Figure S6A ). All these differences were reflected in a principal components analysis where CAR-T cells from AML patients clustered together and close to those from senior healthy donors, being separated from CAR-T cells from adult healthy donors in both CD4⁺ and CD8⁺ CAR-T cell subsets ( Figure 4A ). A closer analysis of the DEGs revealed that most of the differences observed with adult CAR-T cells were shared between senior and AML CAR-T cells, clearly indicating age-related differences. Thus, CAR-T cells from AML patients and senior healthy donors were enriched in genes related to activation, such as CIITA, with downregulation of genes associated with stem cell memory, like CD28 ( Figure 4B , S6D ). Moreover, gene ontology (GO) analysis of common DEGs showed enrichment in pathways related to the regulation of lymphocyte and T cell differentiation ( Figure S6C ), which would explain the more differentiated phenotype observed in the subpopulation analysis. Interestingly, we observed that about 30-40% of the DEGs between AML and adult CAR-T cells were unique to AML condition, indicating specific differences intrinsic to AML. We observed increased expression of genes related to lymphocyte exhaustion, such as LAG3 and NR4A1. Moreover, important genes related to T cell memory (BATF3, CCR7), T cell differentiation (GATA3), apoptosis (BAX, BCL family), or IFN response (OAS1) were also deregulated ( Figure 4C , S6E ). In accordance, GO analysis revealed enrichment in pathways related to lymphocyte activation, response to stress and DNA damage, regulation of the immune response, antigen receptor-mediated signaling, and regulation of the cell cycle ( Figure S6C ). We further characterized CAR-T cells from AML patients upon stimulation with tumoral cells. Two weeks after stimulation we observed a set of genes presenting disrupted expression patterns (different behavior after stimulation) compared to adult and senior CAR-T cells ( Figures 4D , S6F ). Those genes were related to CAR-T cell migration and cell adhesion (CD81, CCL5, MMP25), proliferation (IRF1, CCR7), or regulation of cytokine production (KLF2, IL32), which would explain the reduced proliferation potential and antitumoral efficacy of these AML CAR-T cells. Altogether, these data suggest that CAR-T cells from AML patients present dysfunctional features already identified in the initial T cells, exacerbated after tumor recognition that may compromise the long-term antitumoral efficacy.

Due to the reduced potential of the CD33-CAR-T cells from AML patients and, in no less measure, the fact that delaying treatment in patients with AML may be unacceptable, we decided to approach these issues by generating allogeneic CD33-CAR-T cells. To prevent the immune rejection and graft-versus-host disease (GvHD) associated with allogeneic cells, we used CRISPR-Cas9 to knockout HLA-I and TCR expression. First, we designed different single-guide RNAs (sgRNA) targeting exon 1 of the beta-2-microglobulin (B2M) gene and exon 1 of T-cell receptor α constant (TRAC) locus, and we selected three sgRNAs based on their location and the predicted on-target/off-target efficiency ( Table S3 ). Cleavage efficiency was evaluated in vitro by TIDE (42) after transfection of Streptococcus pyogenes Cas9 (SpCas9) and sgRNA ribonucleoprotein complexes (RNP) in the Jurkat cell line. β2M-sgRNA3 and TRAC-sgRNA2 were selected for further studies since they presented the highest cleavage efficacy, inducing frameshift mutations (mostly 1nt insertions) that resulted in the reduction of HLA-I and TCR levels ( Figure S7 ). Then, selected CRISPR RNPs were combined with the Sleeping Beauty transposon system to generate HLA-I^KO/TCR^KO CD33-CAR-T (CAR-T^KO) cells from healthy donors. The transposon vector containing the selected CD33 targeting CAR construct was used as minicircle (MC) and the SB100X transposase was provided as mRNA. Our optimized protocol allowed in a single electroporation an efficient depletion of HLA-I and TCR complexes, with more than 69% of double negative cells, and an efficient CAR delivery, that was slightly reduced when CAR and RNP were delivered together (average of 46.8% vs 32.4%) ( Figures 5A, B , S8 ). Genome-edited CD33-CAR-T cells were further purified using magnetic beads which yielded >98% double-negative cells while maintaining the percentage of transduction (CAR⁺ cells) ( Figures 5C, D , S8 ). The proliferation capacity of CD33-CAR-T^KO cells, although not significant, was slightly reduced compared to non-edited CAR-T cells, an effect that was also observed in UTD^KO cells, consistent with the effect of electroporation ( Figure 5E ). Moreover, no differences were observed in terms of CD4/CD8 ratio, population subsets, or the expression of activation and exhaustion markers ( Figures 5F, G , S9 ), indicating that HLA-I and TCR depletion did not affect CAR-T cell phenotype. Then in vitro cytotoxicity of CD33-CAR-T^KO cells against MOLM-13 cells was equivalent to non-edited CAR-T cells ( Figure 5H ). Moreover, although proliferation was slightly reduced after continuous restimulation with tumoral cells, CD33-CAR-T^KO cells also presented similar cytotoxic activity to non-edited CAR-T cells ( Figure S9 ). Finally, although the in vivo antitumoral potential in a xenograft model in NSG mice was slightly reduced compared to previous results, where some total remissions were observed, no differences were observed between CD33-CAR-T^KO and CD33-CAR-T cells ( Figures 5I , S9 ). These results indicate that CD33-CAR-T^KO cells generated from healthy donors could be an efficient approach to overcome dysfunctional features observed in CAR-T cells from AML patients.

The safety of edited cells remains a concern for human application. Analysis of genomic DNA found a similar number of transposon copies in CD33-CAR-T^KO cells compared to non-edited CAR-T cells generated with the Sleeping Beauty transposon system, with an average of 7.3 and 7.5 respectively ( Figure 6A ). Moreover, integration site analysis revealed a safe integration profile of the CAR, with no differences due to CRISPR modifications. We mapped and characterized a total of 94148 unique insertion sites of three independent CD33-CAR-T^KO and CAR-T cell productions. Most of the insertions were located at the expected AT-rich DNA regions of the Sleeping Beauty transposon system, detecting the palindromic ATATATAT motif, which contains the TA dinucleotide target sequence adjacent to all the insertions ( Figure S10 ). Moreover, CAR insertions presented a wide distribution within the genome, with no preferences for promoter or exonic regions, and most of the insertions were located at distal intergenic regions ( Figures 6B , S10 ). An additional potential issue associated with CRISPR-based genome editing would be the genotoxicity due to non-specific cleavage of the genome. Thus, we used iGUIDE (36), a modification of the GUIDE-seq method (43), to analyze the CRISPR-mediated cleavage specificity. We observed a highly specific editing of the β2M and TRAC locus, with no significant off-target sites ( Figure 6C ). Taken together, these data strongly suggest that CRISPR-mediated HLA-I and TCR knockout do not alter the safe integration profile of the Sleeping Beauty transposon system without inducing unspecific cleavage of the DNA.

Finally, to determine the feasibility and consistency of the process for further clinical application, we optimized the production of edited CD33-CAR-T^KO cells at a large scale. Thus, 50x10⁶ T cells were electroporated with equivalent scaled proportions of Sleeping Beauty transposon system (MC + mRNA) and RNP (Cas9 + sgRNA). Electroporated CAR-T cells were expanded using the G-Rex platform and CD33-CAR-T^KO cells were selected at the end of the expansion phase using AutoMACS. No differences were observed in editing efficiency before CD33-CAR-T^KO cell selection, with more than 70% of double negative cells, or in the purity of selected cells ( Figure 7A ). The analysis of the purified CD33-CAR-T^KO cells revealed no differences in the transduction efficiency and expansion capacity (population doublings), allowing the generation of at least >300x10⁶ of purified CD33-CAR-T^KO cells at the end of the procedure ( Figures 7B–D ). Moreover, CD33-CAR-T^KO cells produced at the large scale were equally functional in terms of cytotoxic activity and IFN-γ production than previously produced CD33-CAR-T^KO cells ( Figures 7E, F ). In summary, we have successfully developed a protocol for the efficient production and selection of fully functional gene-edited CAR-T cells, allowing the generation of large numbers of CD33-CAR-T^KO cells, that could be compatible with clinical applications.