体内CRISPR双敲除筛选确定协同抗肿瘤基因配对
In vivo CAR-T double knockout screen maps synergistic antitumour gene pairs
研究团队使用体内CRISPR筛选技术,筛选出多个协同作用的抗肿瘤基因配对,这些配对可能增强CAR-T细胞的抗肿瘤效果。研究结果为CAR-T细胞治疗提供了新的靶点和策略。
研究团队通过体内CRISPR筛选技术,发现了协同抗肿瘤基因配对,为CAR-T细胞治疗提供了新的靶点。
Data availability
All data generated or analysed in this study are included in this article, its Supplementary Information, source data files and Supplementary Datasets, or are available through public repositories. Raw and processed high-throughput sequencing data generated in this study, including CRISPR screen sequencing, targeted amplicon sequencing, bulk RNA-seq, single-cell RNA-seq, CUT&RUN and WGS data, have been deposited in the Gene Expression Omnibus under the accession numbers GSE272887 and GSE335215, and the Sequence Read Archive under the accession number PRJNA1477126. Source data and statistical information for non-high-throughput experiments, including flow cytometry, protein analyses, imaging analyses and other molecular or cellular assays, are provided with this paper as source data files. Processed results and relevant quantifications from high-throughput experiments are provided in the Supplementary Datasets. The following publicly available reference datasets and resources were used in this study: the human reference genome GRCh38 Gencode v44 for bulk RNA-seq analyses; the human reference genome GRCh38 Gencode v37 for single-cell RNA-seq analyses; the human reference genome GRCh38 Gencode v47 and ENCODE blacklist regions for GRCh38 for WGS analyses; the Molecular Signatures Database (MSigDB v7.4); and the ProjecTILs single-cell T cell atlas dataset. Parental cell lines are available from the commercial sources listed in the Methods and Supplementary Information. Engineered cell lines, plasmids and other research materials generated in this study are available from the corresponding author on reasonable request and completion of any required material transfer agreements. Source data are provided with this paper.
Code availability
Custom code used for data analysis and figure generation in this study is available at GitHub (https://github.com/Prenauer/DKO-CAR-T).
References
Li, Y.-R., Zhu, Y., Halladay, T. & Yang, L. In vivo CAR engineering for immunotherapy. Nat. Rev. Immunol. 25, 725–744 (2025).
Zebley, C. C. & Youngblood, B. Mechanisms of T cell exhaustion guiding next-generation immunotherapy. Trends Cancer 8, 726–734 (2022).
Kirchhammer, N., Trefny, M. P., Auf der Maur, P., Läubli, H. & Zippelius, A. Combination cancer immunotherapies: emerging treatment strategies adapted to the tumor microenvironment. Sci. Transl. Med. 14, eabo3605 (2022).
Patel, U. et al. CAR T cell therapy in solid tumors: a review of current clinical trials. eJHaem 3, 24–31 (2022).
Wei, J. et al. Targeting REGNASE-1 programs long-lived effector T cells for cancer therapy. Nature 576, 471–476 (2019).
Carnevale, J. et al. RASA2 ablation in T cells boosts antigen sensitivity and long-term function. Nature 609, 174–182 (2022).
Freitas, K. A. et al. Enhanced T cell effector activity by targeting the Mediator kinase module. Science 378, eabn5647 (2022).
Arce, M. M. et al. Central control of dynamic gene circuits governs T cell rest and activation. Nature 637, 930–939 (2025).
Agarwal, S. et al. Deletion of the inhibitory co-receptor CTLA-4 enhances and invigorates chimeric antigen receptor T cells. Immunity 56, 2388–2407.e9 (2023).
Rowshanravan, B., Halliday, N. & Sansom, D. M. CTLA-4: a moving target in immunotherapy. Blood 131, 58–67 (2018).
Sharpe, A. H. & Pauken, K. E. The diverse functions of the PD1 inhibitory pathway. Nat. Rev. Immunol. 18, 153–167 (2018).
Sitaram, P., Uyemura, B., Malarkannan, S. & Riese, M. J. Beyond the cell surface: targeting intracellular negative regulators to enhance T cell anti-tumor activity. Int. J. Mol. Sci. 20, 5821 (2019).
Liu, X. et al. Genome-wide analysis identifies NR4A1 as a key mediator of T cell dysfunction. Nature 567, 525–529 (2019).
Chen, J. et al. NR4A transcription factors limit CAR T cell function in solid tumours. Nature 567, 530–534 (2019).
Seo, H. et al. TOX and TOX2 transcription factors cooperate with NR4A transcription factors to impose CD8+ T cell exhaustion. Proc. Natl Acad. Sci. USA 116, 12410–12415 (2019).
Henriksson, J. et al. Genome-wide CRISPR screens in T helper cells reveal pervasive crosstalk between activation and differentiation. Cell 176, 882–896.e18 (2019).
Shifrut, E. et al. Genome-wide CRISPR screens in primary human T cells reveal key regulators of immune function. Cell 175, 1958–1971.e15 (2018).
Shi, H., Doench, J. G. & Chi, H. CRISPR screens for functional interrogation of immunity. Nat. Rev. Immunol. 23, 363–380 (2023).
Zhou, L., Yang, L., Feng, Y. & Chen, S. Pooled screening with next-generation gene editing tools. Curr. Opin. Biomed. Eng. 28, 100479 (2023).
Zhou, X., Renauer, P. A., Zhou, L., Fang, S.-Y. & Chen, S. Applications of CRISPR technology in cellular immunotherapy. Immunol. Rev. 320, 199–216 (2023).
Dong, M. B. et al. Systematic immunotherapy target discovery using genome-scale in vivo CRISPR screens in CD8 T cells. Cell 178, 1189–1204.e23 (2019).
Ye, L. et al. In vivo CRISPR screening in CD8 T cells with AAV–Sleeping Beauty hybrid vectors identifies membrane targets for improving immunotherapy for glioblastoma. Nat. Biotechnol. 37, 1302–1313 (2019).
Dai, X. et al. Massively parallel knock-in engineering of human T cells. Nat. Biotechnol. 41, 1239–1255 (2023).
Chen, Z. et al. In vivo CD8+ T cell CRISPR screening reveals control by Fli1 in infection and cancer. Cell 184, 1262–1280.e22 (2021).
Huang, H. et al. In vivo CRISPR screening reveals nutrient signaling processes underpinning CD8+ T cell fate decisions. Cell 184, 1245–1261.e21 (2021).
Fu, G. et al. Metabolic control of TFH cells and humoral immunity by phosphatidylethanolamine. Nature 595, 724–729 (2021).
Guo, A. et al. cBAF complex components and MYC cooperate early in CD8+ T cell fate. Nature 607, 135–141 (2022).
Shen, J. P. et al. Combinatorial CRISPR-Cas9 screens for de novo mapping of genetic interactions. Nat. Methods 14, 573–576 (2017).
Chow, R. D. et al. In vivo profiling of metastatic double knockouts through CRISPR-Cpf1 screens. Nat. Methods 16, 405–408 (2019).
Campa, C. C., Weisbach, N. R., Santinha, A. J., Incarnato, D. & Platt, R. J. Multiplexed genome engineering by Cas12a and CRISPR arrays encoded on single transcripts. Nat. Methods 16, 887–893 (2019).
Li, R. et al. Comparative optimization of combinatorial CRISPR screens. Nat. Commun. 13, 2469 (2022).
Tang, S. et al. Generation of dual-gRNA library for combinatorial CRISPR screening of synthetic lethal gene pairs. STAR Protoc. 3, 101556 (2022).
Prinzing, B. et al. Deleting DNMT3A in CAR T cells prevents exhaustion and enhances antitumor activity. Sci. Transl. Med. 13, eabh0272 (2021).
LaFleur, M. W. et al. PTPN2 regulates the generation of exhausted CD8+ T cell subpopulations and restrains tumor immunity. Nat. Immunol. 20, 1335–1347 (2019).
Aki, D. et al. The E3 ligases Itch and WWP2 cooperate to limit TH2 differentiation by enhancing signaling through the TCR. Nat. Immunol. 19, 766–775 (2018).
Nurieva, R. I. et al. The E3 ubiquitin ligase GRAIL regulates T cell tolerance and regulatory T cell function by mediating T cell receptor-CD3 degradation. Immunity 32, 670–680 (2010).
Huang, C.-Y. et al. DUSP4 deficiency enhances CD25 expression and CD4+ T-cell proliferation without impeding T-cell development. Eur. J. Immunol. 42, 476–488 (2012).
Dashevsky, O. et al. Use of olfactory receptor genes as controls for genome-scale CRISPR functional genomic studies to define treatment resistance mechanisms. Blood 136, 36 (2020).
Zhao, D. et al. Combinatorial CRISPR-Cas9 metabolic screens reveal critical redox control points dependent on the KEAP1-NRF2 regulatory axis. Mol. Cell 69, 699–708.e7 (2018).
Yang, L. et al. OR7A10 GPCR engineering boosts CAR-NK therapy against solid tumours. Nature 652, 740–751 (2026).
Rosskopf, S. et al. A Jurkat 76 based triple parameter reporter system to evaluate TCR functions and adoptive T cell strategies. Oncotarget 9, 17608–17619 (2018).
Ye, L. et al. A genome-scale gain-of-function CRISPR screen in CD8 T cells identifies proline metabolism as a means to enhance CAR-T therapy. Cell Metab. 34, 595–614.e14 (2022).
Mise-Omata, S. et al. SOCS3 deletion in effector T cells confers an anti-tumorigenic role of IL-6 to the pro-tumorigenic cytokine. Cell Rep. 42, 112940 (2023).
Negishi, S. et al. Treatment strategies for relapse after CAR T-cell therapy in B cell lymphoma. Front. Pediatr. 11, 1305657 (2023).
Liberzon, A. et al. The Molecular Signatures Database (MSigDB) hallmark gene set collection. Cell Syst. 1, 417–425 (2015).
Shi, H., Chen, S. & Chi, H. Immunometabolism of CD8+ T cell differentiation in cancer. Trends Cancer 10, 610–626 (2024).
Dumauthioz, N. et al. Enforced PGC-1α expression promotes CD8 T cell fitness, memory formation and antitumor immunity. Cell. Mol. Immunol. 18, 1761–1771 (2021).
Andreatta, M. et al. Interpretation of T cell states from single-cell transcriptomics data using reference atlases. Nat. Commun. 12, 2965 (2021).
Liu, Y. et al. Chimeric STAR receptors using TCR machinery mediate robust responses against solid tumors. Sci. Transl. Med. 13, eabb5191 (2021).
Nelson, R. FDA investigating safety risks in CAR T-cell recipients. Lancet 402, 2181 (2023).
Ashburner, M. et al. Gene Ontology: tool for the unification of biology. Nat. Genet. 25, 25–29 (2000).
The Gene Ontology Consortium. The Gene Ontology knowledgebase in 2023. Genetics 224, iyad031 (2023).
Schmiedel, B. J. et al. Impact of genetic polymorphisms on human immune cell gene expression. Cell 175, 1701–1715.e16 (2018).
Joung, J. et al. Genome-scale CRISPR-Cas9 knockout and transcriptional activation screening. Nat. Protoc. 12, 828–863 (2017).
Martin, M. Cutadapt removes adapter sequences from high-throughput sequencing reads. EMBnet.journal https://doi.org/10.14806/ej.17.1.200 (2011).
Langmead, B., Trapnell, C., Pop, M. & Salzberg, S. L. Ultrafast and memory-efficient alignment of short DNA sequences to the human genome. Genome Biol. 10, R25 (2009).
Li, X., Cooper, N. G. F., O’Toole, T. E. & Rouchka, E. C. Choice of library size normalization and statistical methods for differential gene expression analysis in balanced two-group comparisons for RNA-seq studies. BMC Genomics 21, 75 (2020).
Lun, A. T. L., Chen, Y. & Smyth, G. K. It’s DE-licious: a recipe for differential expression analyses of RNA-seq experiments using quasi-likelihood methods in edgeR. Methods Mol. Biol. 1418, 391–416 (2016).
Ritchie, M. E. et al. limma powers differential expression analyses for RNA-sequencing and microarray studies. Nucleic Acids Res. 43, e47 (2015).
Clement, K. et al. CRISPResso2 provides accurate and rapid genome editing sequence analysis. Nat. Biotechnol. 37, 224–226 (2019).
Roederer, M., Nozzi, J. L. & Nason, M. C. SPICE: exploration and analysis of post-cytometric complex multivariate datasets. Cytometry A 79, 167–174 (2011).
Bray, N. L., Pimentel, H., Melsted, P. & Pachter, L. Near-optimal probabilistic RNA-seq quantification. Nat. Biotechnol. 34, 525–527 (2016).
Frankish, A. et al. Gencode 2021. Nucleic Acids Res. 49, D916–D923 (2021).
Hao, Y. et al. Dictionary learning for integrative, multimodal and scalable single-cell analysis. Nat. Biotechnol. https://doi.org/10.1038/s41587-023-01767-y (2023).
Welch, J. D. et al. Single-cell multi-omic integration compares and contrasts features of brain cell identity. Cell 177, 1873–1887.e17 (2019).
Trapnell, C. et al. The dynamics and regulators of cell fate decisions are revealed by pseudotemporal ordering of single cells. Nat. Biotechnol. 32, 381–386 (2014).
Aibar, S. et al. SCENIC: single-cell regulatory network inference and clustering. Nat. Methods 14, 1083–1086 (2017).
Andrews, S. FastQC: a quality control tool for high throughput sequence data. GitHub https://github.com/s-andrews/FastQC (2010).
Langmead, B. & Salzberg, S. L. Fast gapped-read alignment with Bowtie 2. Nat. Methods 9, 357–359 (2012).
Danecek, P. et al. Twelve years of SAMtools and BCFtools. Gigascience 10, giab008 (2021).
Quinlan, A. R. & Hall, I. M. BEDTools: a flexible suite of utilities for comparing genomic features. Bioinformatics 26, 841–842 (2010).
Amemiya, H. M., Kundaje, A. & Boyle, A. P. The ENCODE Blacklist: identification of problematic regions of the genome. Sci. Rep. 9, 9354 (2019).
Zhang, Y. et al. Model-based analysis of ChIP-seq (MACS). Genome Biol. 9, R137 (2008).
Heinz, S. et al. Simple combinations of lineage-determining transcription factors prime cis-regulatory elements required for macrophage and B cell identities. Mol. Cell 38, 576–589 (2010).
Ramirez, F. et al. deepTools2: a next generation web server for deep-sequencing data analysis. Nucleic Acids Res. 44, W160–W165 (2016).
Li, H. & Durbin, R. Fast and accurate short read alignment with Burrows-Wheeler transform. Bioinformatics 25, 1754–1760 (2009).
McKenna, A. et al. The Genome Analysis Toolkit: a MapReduce framework for analyzing next-generation DNA sequencing data. Genome Res. 20, 1297–1303 (2010).
Cameron, D. L. et al. GRIDSS: sensitive and specific genomic rearrangement detection using positional de Bruijn graph assembly. Genome Res. 27, 2050–2060 (2017).
Zhang, H., Meltzer, P. & Davis, S. RCircos: an R package for Circos 2D track plots. BMC Bioinformatics 14, 244 (2013).
Acknowledgements
We thank A. Abiri and other laboratory members for reagent sharing and/or assistance; R. A. Flavell, D. Schatz, C. Lucas and J. Craft for scientific comments; P. Steinberger and K. Anderson for sharing Jurkat reporter cell lines; the Shen laboratory for Seahorse experiment assistance; and all members of the Chen laboratory and various core facilities for technical support.
Funding
S.C. is supported by the Cancer Research Institute Lloyd J. Old STAR Award and KCA CLIP Award (CRI4964 and CRI13945), the Alliance for Cancer Gene Therapy (ACGT) and the Cinelli Family Foundation, the NIH/NCI (R33CA281702), the DoD (HT9425-23-1-0472 and HT9425-23-1-0860), the Pershing Square Sohn Cancer Research Alliance, the Prostate Cancer Foundation, Bill Sperry and the YCC Team Science Award. M.M. is supported by HHMI-55108547, NIH/NCI R35CA197628, R01CA282877 and R01CA157644, and NIH/NIAID R01AI164692 and R01AI192914. L.Z. is supported by the Gruber Science Fellowship and the Lo Fellowship of Excellence of Stem Cell Research. P.A.R. is supported by a Yale PhD training grant from the NIH (T32GM007499), the Lo Fellowship of Excellence of Stem Cell Research and the YCC T32 fellowship program. K.T. is supported by the Gruber Science Fellowship. C.D. is supported by the Boehringer Ingelheim Biomedical Data Science Fellowship. X.Z. is supported by NIH/NCI K99/R00 Pathway to Independence Awards (K99CA282989 and R00CA282989) and the Margaret Q. Landenberger Research Foundation.
Author information
Author notes
These authors contributed equally: Liqun Zhou, Paul A. Renauer, Kaiyuan Tang
Authors and Affiliations
Department of Genetics, Yale University School of Medicine, New Haven, CT, USA
Liqun Zhou (周立群), Paul A. Renauer, Kaiyuan Tang, Shanshan Li, Zhicheng Luo, Yanzhi Feng, Shawn H. Lin, Feifei Zhang, Daniel Park, Shao-Yu Fang, Grace Vest, Zhihan Zhao, William Dorst, Luojia Yang, Chuanpeng Dong, Lei Peng, Xiaoyu Zhou & Sidi Chen
Systems Biology Institute, Yale University, West Haven, CT, USA
Liqun Zhou (周立群), Paul A. Renauer, Kaiyuan Tang, Shanshan Li, Zhicheng Luo, Yanzhi Feng, Shawn H. Lin, Feifei Zhang, Daniel Park, Shao-Yu Fang, Grace Vest, Zhihan Zhao, William Dorst, Luojia Yang, Chuanpeng Dong, Lei Peng, Xiaoyu Zhou & Sidi Chen
Department of Immunobiology, Yale University, New Haven, CT, USA
Liqun Zhou (周立群), Yanzhi Feng, Shawn H. Lin, Zhihan Zhao & Markus Müschen
Immunobiology Program, Yale University, New Haven, CT, USA
Liqun Zhou (周立群), Yanzhi Feng, Shawn H. Lin, Zhihan Zhao, Markus Müschen & Sidi Chen
Molecular Cell Biology, Genetics, and Development Program, Yale University, New Haven, CT, USA
Kaiyuan Tang, Luojia Yang & Sidi Chen
Center of Molecular and Cellular Oncology, Yale University, New Haven, CT, USA
Yanzhi Feng, Lars Klemm & Markus Müschen
Department of Neurosurgery, Baylor College of Medicine, Temple, TX, USA
Xingxin Pan, S. Stephen Yi & Nidhi Sahni
Dan L Duncan Comprehensive Cancer Center, and Dan L Duncan Institute for Clinical and Translational Research, Baylor College of Medicine, Houston, TX, USA
S. Stephen Yi
Quantitative and Computational Biosciences Program, Baylor College of Medicine, Houston, TX, USA
Nidhi Sahni
Vaccine and Immunotherapy Center, The Wistar Institute, Philadelphia, PA, USA
Xiaoyu Zhou
Ellen and Ronald Caplan Cancer Center, The Wistar Institute, Philadelphia, PA, USA
Xiaoyu Zhou
Department of Pathology and Laboratory Medicine, University of Pennsylvania, Philadelphia, PA, USA
Xiaoyu Zhou
Yale Comprehensive Cancer Center, Yale University School of Medicine, New Haven, CT, USA
Sidi Chen
Yale Stem Cell Center, Yale University School of Medicine, New Haven, CT, USA
Sidi Chen
Department of Biomedical Informatics and Data Science, Yale University School of Medicine, New Haven, CT, USA
Sidi Chen
Department of Neurosurgery, Yale University School of Medicine, New Haven, CT, USA
Sidi Chen
Authors
- Liqun Zhou (周立群)
- Paul A. Renauer
- Kaiyuan Tang
- Shanshan Li
- Zhicheng Luo
- Yanzhi Feng
- Shawn H. Lin
- Feifei Zhang
- Daniel Park
- Shao-Yu Fang
- Grace Vest
- Zhihan Zhao
- William Dorst
- Luojia Yang
- Chuanpeng Dong
- Xingxin Pan
- Lars Klemm
- S. Stephen Yi
- Nidhi Sahni
- Lei Peng
- Markus Müschen
- Xiaoyu Zhou
- Sidi Chen
Contributions
L.Z. and S.C. conceived the study. L.Z., P.A.R. and X.Z. designed the screens. L.Z. performed the screens. L.Z. and K.T. performed most of the experiments with assistance from various co-authors. P.A.R. performed the screen, bulk RNA-seq and single-cell analyses. K.T. performed the targeted Amplicon-seq, CUT&RUN and WGS analyses. S.L., Z.L., Y.F., S.H.L., F.Z., D.P., S.-Y.F., G.V., Z.Z., W.D., L.Y., C.D., X.P., S.S.Y., N.S. and L.P. assisted with various experiments and analyses. Y.F., L.K. and M.M. provided reagents and samples. L.Z., P.A.R., K.T., X.Z. and S.C. wrote the manuscript with input from all authors. S.C. secured funding and provided overall supervision of the project.
Corresponding authors
Correspondence to Xiaoyu Zhou or Sidi Chen.
Ethics declarations
Competing interests
S.C. is a (co)founder of EvolveImmune, Cellinfinity, MagicTime and Chen Consulting. A patent application has been filed by Yale University. The other authors declare no competing interests.
Peer review
Peer review information
Nature thanks Joseph Fraietta, Dalia Haydar and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Peer reviewer reports are available.
Additional information
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Extended data figures and tables
Extended Data Fig. 1 Establishment and quality control of an in vivo DKO CRISPR screen for solid tumour infiltration by human primary CAR-T cells.
a, Quantification of CD3− cells 5 days after CRISPR-mediated gene editing in three independent donors. b, Quantification of live cells 5 days after editing with the indicated Cas9 mRNA concentrations, n = 3 biologically independent replicates. Data are shown as mean + s.e.m. Significance was calculated using one-way ANOVA with Dunnett’s multiple-comparisons test. ns, not significant. c, Left, representative flow cytometry plots showing CD3 and CD28 protein-level DKO efficiency in activated human primary T cells 5 days after Cas9 mRNA electroporation. Right, quantification of CD3+ CD28+ cells in DKO and vector control groups; data are shown as mean + s.e.m.; n = 4 infection replicates. Significance was calculated using a paired two-sided t-test, ****P < 0.0001. d, Quantification of CD3− or CD28− cells 5 days after editing; n = 3 donors per concentration. Data are shown as mean ± s.e.m. e, Classification and composition of targeted genes in the T cell DKO CRISPR library. f, Schematic of the two-step T cell DKO CRISPR library construction process. g, Quantification of plasmid library sgRNA-pair construct distribution. h, Flow cytometry histograms showing EGFRvIII expression in U87-GL and U87-GL-EGFRvIII cell lines. i, Flow cytometry histograms showing PD-L1 expression in U87-GL and U87-GL-EGFRvIII cell lines. j, Schematic of DKO CAR-T cell library generation pipeline, including human T cell isolation and activation, CAR-T cell generation and selection, lentiviral DKO CRISPR library infection and Cas9 mRNA electroporation. Schematic in panel j created in BioRender; Chen, S. https://biorender.com/4g9xapi (2026). k, Spider plots of tumour volumes for donor 1 in vivo screen experiment shown in Fig. 1d. n = 5 biologically independent mice. l, Representative H&E-stained tumour sections from two biologically independent tumours per group from the screen shown in Fig. 1c. Scale bar, 250 μm. m, sgRNA-pair detection rates from the total library across individual samples. D, donor; T, tumour. n, Heat map of Spearman correlation coefficients between the log2-normalized counts-per-million (log-norm.) sgRNA counts for each sample. Data were arranged by hierarchical clustering using the ward.D2 method. o, Multidimensional scaling analysis of log2-normalized screen data using the top 10,000 most variable sgRNA pairs across samples. Data points were clustered by k-means. p, Box-dot plot of library representation across individual samples, showing the distribution of all sgRNA-pair constructs counts. Boxes indicate the 25th–75th percentiles, center lines indicate medians, and whiskers extend to the most extreme values within 1.5 times the interquartile range. Individual points represent sgRNA-pair counts. q, Empirical cumulative distribution function of DKO library representation among detected sgRNA pairs in plasmid, pre-injection cell and tumour samples. Sample labels in m-p indicate sample type, plasmid, pre-injection cell or tumour, and donor ID.
Extended Data Fig. 2 Additional analysis and gene editing validation of DKO CAR-T cells.
a, Distribution of the top four performing sgRNA-pair constructs for top DKO hits across samples, ranked by SAMBA sgRNA z-score. b, Heat map of DKO screen scores for all pairwise gene combinations assessed in the CRISPR screen. Values were scaled and visualized within a range of −1 to 1. Hierarchical clustering was performed using the Ward.D2 method based on Euclidean distances. c, Scatter plot showing the relationship between screen enrichment and synergistic effect. SAMBA z-scores were plotted against synergy scores for each gene pair. Dashed lines indicate FDR-adjusted P value, q < 0.05, for epistasis and SAMBA results. d, Scatter plot of observed DKO SAMBA scores versus expected SAMBA scores based on the additive effects of the corresponding SKOs for each gene pair (SKO1 + SKO2). Points are colored by synergy score. Linear regression yielded R2 = 0.0152 (n = 2,485 gene pairs; 71 SKO genes), indicating minimal predictability of DKO outcomes from SKO additivity. e, Quantification of indel frequencies in DKO-edited CAR-T cells 5 days after electroporation across three independent batches. P values were calculated using unpaired two-sided t-tests. Data are mean ± s.e.m.; n = 3 human donors. Each dot represents one human donor. f, Quantification of indel frequencies in CAR-T cells from three donors generated across three independent batches 90 days after gene editing. P values were calculated using one-way ANOVA with Dunnett’s multiple-comparisons test versus GTC. Data are mean + s.e.m.; n = 3 independent donor-derived CAR-T cell preparations, generated from three donors in independent batches. NS, not significant, *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.
Extended Data Fig. 3 Additional in vitro validation of top DKO hits.
a, Quantification of live cells 5 days after editing in different DKO CAR-T cells compared with controls, n = 3 infection replicates per group. b, Schematic of generation of DKO CAR-T or CAR-Jurkat cell generation and downstream assays. Schematic in panel b created in BioRender; Chen, S. https://biorender.com/7vyyd1r (2026). c, Quantification of Ca2+ flux AUC and peak signal change in different DKO CAR-Jurkat samples. The resting region was defined as 1 min before antigen crosslinking, and the peak region was defined as 500 s after cross-linking. AUC is calculated as (mean peak signal – mean resting signal) × 500. Representative of three independent experiments. d, Quantification of pERK1/2 (pT202/pY204) signal in DKO CAR-Jurkat cells after U87-GL-EGFRvIII stimulation at the indicated time points; n = 3 per group per time point. e, Quantification of pERK1/2+ (pT202/pY204) DKO CAR-T cells after U87-GL-EGFRvIII cell stimulation for 30 min; n = 3 biological replicates per group. f-g, NF-κB CFP and NFAT/AP-1 GFP reporter activity after coculture of U87-GL-EGFRvIII cells with different DKO CD8+ CAR-Jurkat cells with GTC control at indicated time points. E:T = 1:2. n = 3 per group per time point. f, Schematic of the NFAT/AP-1 GFP reporter structure and quantification of GFP+ cells. g, Schematic of the NF-κB CFP reporter structure and the quantification of CFP+ cells. h, Quantifications of tonic signaling in CAR-Jurkat cells after 1 month cell culture without stimulation, n = 3. P value was determined by one-way ANOVA with Dunnett’s correction for multiple comparisons test versus GTC. i, Left, representative flow cytometry histograms showing CD107a expression in CD8+ DKO and GTC CAR-T cells after 2-h coculture with U87-GL-EGFRvIII cells. Right, quantification of CD107a+ CAR-T cells; n = 4 per group; E:T = 1:2. j, Quantification of IFN-γ, TNF, and granzyme B expression in CD8+ and CD4+ CAR-T cells after 4-h coculture with U87-GL-EGFRvIII cells at E:T = 1:8; n = 4 per group. k, Left, quantification of CellTrace Violetlow CAR-T cells after two rounds of repeated antigen stimulation, measured on day 5, n = 6 each. Right, quantification of absolute CAR-T cell numbers; n = 3 on day 0, n = 6 on day 2, 5, and 7 per group. l, Quantification of CellTrace Violetlow CAR-T cells in low IL-2 medium, measured on day 7; n = 3 per group. m, Quantification of live cells percentage in culture or after repeated challenge. n, Left, quantification of CD45RO+ CD62L+ DKO CAR-T cells after two rounds of tumour stimulation on day 5. Right, quantification of CD27+ DKO CAR-T cells after two rounds of tumour stimulation on day 5; n = 6 per group. For each round of coculture, E:T = 1:4. o, Quantifications of LAG-3, PD-1, and TIM-3 expression in CAR-T cells after repeated tumour challenge on day 5, n = 4 per group. For each round of coculture, E:T = 1:4. p, Quantification of LAG-3+, PD-1+, TIM-3+ CAR-T cells after two rounds of tumour stimulation on day 5; n = 4 per group. For each round of coculture, E:T = 1:4. q, SPICE analysis of LAG-3, PD-1, and TIM-3 expression in CAR-T cells on day 5. Data are group means. r, Quantification of absolute CAR-T cell number after IL-2 withdrawal; n = 3 per group at day 0, n = 4 per group at day 2, 5, and 7. Data in d-g,k right,m,r are mean ± s.e.m.; data in a,h-j,k left,l,n-p are mean + s.e.m. Significance levels were calculated using ordinary two-way ANOVA with Dunnett’s multiple-comparisons test comparing each group with GTC within each row or time point (d-g,j,k right,m,r), repeated-measures one-way ANOVA with Dunnett’s multiple-comparisons test versus GTC (a), or ordinary one-way ANOVA with Dunnett’s multiple-comparisons test versus GTC (h,i,k left,l,n-p). Dashed lines in c, i, j, n-p indicate mean values of the GTC group. n indicates biologically independent samples unless otherwise stated. Exact P values and detailed statistics are provided in the Source Data. Data are representative of three independent experiments using cells from human donors independent of those used in the screen. NS, not significant, *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.
Extended Data Fig. 4 Additional in vitro validation of NR4A1-SOCS3 DKO CAR-T cells.
a-b, NUR77 (NR4A1) and SOCS3 protein expression in human CAR-T cells 30 days after gene editing. a, Left, representative flow cytometry histograms showing NUR77 (NR4A1) protein expression. Right, quantification of NR4A1 MFI; n = 3 per group. b, Left, representative flow cytometry histograms showing SOCS3 protein expression. Right, quantification of SOCS3 expression MFI; n = 3 per group. c, Quantification of CD107a+ CAR-T cells after 2-h coculture with tumour cells at E:T = 1:4. n = 3 per group. d, Quantification of IFNγ+, TNF+, and granzyme B MFI in CAR-T cells after 4-h coculture with tumour cells at E:T = 1:4. n = 4 per group. e, Left, cytokine array image with 15-min exposure. Right, Quantification of expression fold change relative to GTC. n = 2 technical duplicate spots on the same membrane (uncropped image, Supplementary Fig. 1). f, Flow cytometry histograms showing HER2 expression in different HER2-expressing cell lines. g, Flow cytometry histograms showing EGFRvIII expression in different U87 cell lines. h-i, Quantification of 24-h tumour killing percentages of different U87 cells lines at E:T = 1:4 (h) or E:T = 1:8 (i). n = 4 per group. j, Quantification of 24-h killing of HT-29 or U87-GL-EGFRvIII cells by antigen mismatched CAR-T cells at E:T = 1:2. n = 4 per group. k, Schematic of long-term in vitro repeated tumour-challenge experiments with DKO, NR4A1 SKO, SOCS3 SKO, and GTC CAR-T cells, with re-challenge every 6 days. Schematic in panel k created in BioRender; Chen, S. https://biorender.com/wcree3v (2026). l-m, Quantification of tumour killing percentages in long-term repeated challenge at E:T = 1:4 (l) or E:T = 1:8 (m). n = 6 per group. n, Quantification of HIF-1α expression in CAR-T cells under normoxic or hypoxic conditions. n = 3 per group. o, Workflow for generating B-ALL and MCL PDX cells. p, Table showing PDX donor information. Schematic in panel o created in BioRender; Chen, S. https://biorender.com/6995oyk (2026). q, Flow cytometry staining of CD19 and CD20 expression PDX donor cells. r, Quantification of 24-h killing of different PDX donor cells by CAR-T cells. n = 4 per group. s, Quantification of IFNγ, TNF, or granzyme B positive CAR-T cells after 4-h coculture with PDX2 cells at E:T = 1:2; n = 4 per group. t, Quantification of LAG-3+PD-1+TIM-3+ CAR-T cells after two rounds of tumour stimulation at E:T = 1:10 per round. n = 4 infection replicates per group. u, SPICE analysis of LAG-3, PD-1 and TIM-3 expression in CAR-T cells after one round of tumour stimulation E:T = 1:10. Data are group means from three donors. v, Quantification of effector memory-like (CD45RO+CD62L−) and central memory-like (CD45RO+CD62L+) CAR-T cells after two rounds of repeated PDX2 challenge; E:T = 1:10 per challenge round; n = 4 per group. Data in h-j,l,m are mean ± s.e.m.; data in a-d,n,r-t,v are mean + s.e.m. Significance levels were calculated using one-way ANOVA with Dunnett’s multiple-comparisons test versus GTC (a-c,s,t), one-way ANOVA with Tukey’s multiple-comparisons test (d), ordinary two-way ANOVA with Šídák’s multiple-comparisons test comparing DKO with each indicated control within each antigen-expression condition (h,i), ordinary two-way ANOVA with Tukey’s multiple-comparisons test within each row or condition (j,n), two-way repeated-measures ANOVA with Šídák’s multiple-comparisons test comparing DKO with each indicated control at each time point (l,m), or ordinary two-way ANOVA with Dunnett’s multiple-comparisons test versus GTC within each row or condition (r,v). n indicates biologically independent samples unless otherwise stated. CAR-T cells were generated using human donor cells independent of those used in the screen. Exact P values and detailed statistics are provided in the Source Data. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.
Extended Data Fig. 5 Additional in vitro validation of other top DKO hits.
a, Quantification of 24-h killing of different HER2-expressing tumour cells; n = 4 per group. b, Left, representative flow cytometry plots showing CD107a expression in CAR-T cells after 2-h coculture with HT29-PL cells at E:T = 1:4. Right, quantification of CD107a MFI. n = 4 per group. c, Left, representative flow cytometry plots showing granzyme B expression in CAR-T cells after 4-h coculture with HT29-PL cells at E:T = 1:4. Right, quantification of Granzyme B+ CAR-T cells; n = 4 per group. d, Left, representative flow cytometry plots showing TNF expression in CAR-T cells after 4-h coculture with HT29-PL cells at E:T = 1:4. Right, quantification of TNF+ CAR-T cells; n = 4 per group. e, Quantification of effector memory-like (EM-like, CD45RO+CD62L−) and central memory-like (CM-like, CD45RO+CD62L+) CAR-T cells 5 days post two rounds of HT29 tumour stimulation. E:T = 1:2 for each round; n = 3 per group. f, Left, quantification of LAG-3+PD-1+TIM-3+ CAR-T cells after two rounds of tumour stimulation on day 5. E:T = 1:2 for each round; n = 4 per group. Right, SPICE analysis of LAG-3, PD-1, and TIM-3 expression of CAR-T cells on day 5. Data are group means. g, Quantification of 24-h killing of different HER2-expressing tumour cells; n = 4 per group. h, Left, representative flow cytometry plots showing CD107a expression in CAR-T cells after 2-h coculture with HT29-PL cells at E:T = 1:4. Right, quantification of CD107a MFI. n = 4 per group. i, Left, representative flow cytometry plots showing granzyme B expression in CAR-T cells after 4-h coculture with HT29-PL cells at E:T = 1:4. Right, quantification of Granzyme B+ CAR-T cells; n = 4 per group. j, Left, representative flow cytometry plots showing TNF expression in CAR-T cells after 4-h coculture with HT29-PL cells at E:T = 1:4. Right, quantification of TNF+ CAR-T cells; n = 4 per group. k, Quantification of effector memory (EM, CD45RO+CD62L−) and central memory (CM, CD45RO+CD62L+) CAR-T cells 5 days post two rounds of HT29 tumour stimulation. E:T = 1:2 for each round; n = 3 per group. l, Left, quantification of LAG-3+PD-1+TIM-3+ CAR-T cells after two rounds of tumour stimulation on day 5. E:T = 1:2 for each round; n = 4 per group. Right, SPICE analysis of LAG-3, PD-1, and TIM-3 expression of CAR-T cells on day 5. Data are group means. Data are shown as means + s.e.m. Significance levels were calculated by one-way ANOVA with Dunnett’s correction (b-d, f, h-j, l), or two-way ANOVA with Dunnett’s correction (a, e-g, k). n indicates biologically independent samples. CAR-T cells were generated using cells from human donors independent from those used in the screen. Exact P values and detailed statistics are provided in the Source Data. NS, not significant, *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.
Extended Data Fig. 6 Time-course RNA sequencing reveals pathway synergy in NR4A1-SOCS3 DKO CAR-T cells.
a, Schematic of the time-series RNA-seq experiment in DKO, NR4A1 SKO, SOCS3 SKO, and GTC CAR-T cells undergoing repeated tumour challenges. b, Principal components analysis (PCA) of all RNA-seq samples across all stimulation time-points (left) or separated by timepoint (right). The first two principal components are shown. Genotypes are indicated by dot color, and samples are grouped by time point (left) or genotype (right) using convex hulls. c, Volcano plots of differential expression (DE) analysis for epistasis at each time-point. Significant epistatic transcripts were defined by an FDR-adjusted P value (q value) <0.05 and an absolute log2 fold change (LFC) > 0.5 for the DKO interaction term in the statistical model (Methods). Top DE transcripts are labeled based on significance and LFC, with selected transcripts shown in bold. d, Heatmaps and pathway analysis of DE transcripts from pairwise comparisons between genotypes across stimulation times. For each time point, DE transcripts were hierarchically clustered and shown as heatmaps (left). Pathway analyses were performed independently for each transcript cluster using over-representation analysis (ORA) tests with MSIGDB v7.4 Hallmark gene-sets. Significant pathways (q < 0.05; up to four pathways per cluster shown) are shown on the right, color-coded by the transcript clusters. Dot-size indicates the percentage of pathway genes that were significantly differentially expressed. e, Heatmaps of DE transcripts significantly enriched as NF-κB and STAT5B targets by transcription factor (TF) over-representation analysis (ORA). TF analysis was performed using MSIGDB v7.4 TF gene-sets; ontology names are shown above the heatmaps. f-g, Representative flow cytometry histogram and quantification of p-STAT3 and p-STAT5 in CAR-T cell groups. n = 4 (f), and n = 3 (g) biological replicates per group. h, Representative flow cytometry histograms and quantification of BAX expression in CAR-T cells. n = 3 biological replicates per group. Data in f-h are mean + s.e.m. Significance levels in f-h were calculated using one-way ANOVA with Dunnett’s multiple-comparisons versus GTC (f-g) or DKO (h). Exact P values and detailed statistics are provided in the Source Data. NS, not significant, *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.
Extended Data Fig. 7 Single-cell transcriptomics annotation and pathway-state analysis of in vivo CAR-T subsets.
a, Schematic of the single-cell transcriptomics (SCT) data processing pipeline. b, UMAP projections showing the T cell filtering step in the SCT analysis. B cells were identified by CD20 expression, and T cells were identified by CD3E expression. c, Dot plot showing gene signatures used for immune subset classification. Cells were classified into major immune cell categories, CD8+ T cell subsets and phenotypic states. Dot color and size indicate average gene signature z-score and percentage of expressing cells. Gene-sets used for each gene signature of the dot plot are shown in the right panel tables. d, Dot plot showing transcriptional patterns of marker genes used to classify in vivo CAR-T cell subsets in the SCT data. Dot color and size indicate average gene expression z-score and percentage of expressing cells. e, Distribution of ungrouped T cell subsets across genotypes, shown as stacked bar plots (left) and quantified by subset (right). Cells from mice within each treatment group were pooled before single-cell RNA sequencing. Statistical comparisons of cell-subset proportions between genotypes were performed using two-sided pairwise Fisher’s exact tests with Holm adjustment for multiple comparisons. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. f, Dot plot showing transcriptional patterns for genes comprising signatures of key T cell signaling pathways and functions. Dot color and size indicate average gene expression z-score and percentage of expressing cells, respectively. Data are arranged by hierarchical clustering using the ward.D2 method, and dashed boxes mark DKO subsets.
Extended Data Fig. 8 Additional transcriptional and mitochondrial fitness analysis of NR4A1-SOCS3 DKO CAR-T cells.
a, PCA projections of in vivo CAR-T cells grouped by genotype and subset, based on gene expression patterns for each gene-set. PCA was performed using the gene expression data shown in Extended Data Fig. 7f. Points are color-coded by genotype, and polygons mark DKO subset points that clustered together. b, Volcano plots of differential transcription factor signature analysis between DKO and GTC genotypes. Upregulated and downregulated signatures, defined by adjusted P < 0.01 and absolute log2 fold change > 0.25, are shown as red and blue dots, respectively. Statistical comparisons were performed using cell-level Wilcoxon rank-sum tests, and P values were adjusted for multiple testing using Bonferroni correction. c, Dot plot showing transcriptional patterns for genes comprising transcription factor signatures. Dot color and size indicate average gene expression z-score and percentage of expressing cells. Data are arranged by hierarchical clustering using the ward.D2 method, and dashed boxes mark DKO subsets. d, PCA projection of in vivo CAR-T cells grouped by genotype and subset, based on gene expression patterns for each gene-set. PCA was performed using the gene expression data shown in c. Points are color-coded by genotype, and polygons mark DKO subset points that clustered together. e, Left, representative flow cytometry MitoTracker Green signal. Right, quantification of MitoTracker Green MFI; n = 4 per group. CAR-T cells were stimulated with U87-GL-EGFRvIII cells at E:T = 1:1 and cultured for 7 days. f, Left, OCR curves from Seahorse assays. Right, quantification of SRC and MRC; n = 3 per group. g, Illustration of epistasis involving PGC-1α downstream of NR4A1-SOCS3 DKO. Data in f left are mean ± s.e.m., in e,f right are mean +s.e.m. Significance levels were calculated using one-way ANOVA with Tukey’s multiple-comparisons test (e) or two-way ANOVA with Tukey’s multiple-comparisons test (f). n indicates biologically independent samples. CAR-T cells were generated using human donor cells independent of those used in the screen. Exact P values and detailed statistics are provided in the Source Data. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. OCR, oxygen consumption rate; SRC, spare respiratory capacity; MRC, maximal respiratory capacity.
Extended Data Fig. 9 Additional in vivo validation data.
a, Venn diagram of genes included in benchmark comparisons. b, Quantification of PD-1, TIM-3, or LAG-3 expression in round 5 repeated stimulated anti-HER2 CAR-T cells. n = 3 infection replicates per group. c, Quantification of CD45+ cells within live cells population, identified by Live/Dead NIR stain, from the B16F10-OVA syngeneic model in Fig. 5d–h. c-f, n = 5 mice for each group. d, Quantification of NK1.1+ cells within the CD45+ cell population. e, Quantification of CD11b+ F4/80+ cells within the CD45+ cells population. f, Quantification of CD11c+ MHCII+ cells within the CD45+ cells population. g, Quantification of tumour weights of day 21 in the U87-GL-EGFRvIII model. n is the same as in Fig. 5j. h, Spider plots of U87-GL-EGFRvIII tumour volumes in different CAR-T treatment groups. i, Quantification of splenic CAR-T cell number of day 21 in the U87-GL-EGFRvIII model. n is the same as in Fig. 5j. j, Quantification of CCR7+ CD27+ CD45RO+ CD62L+ CAR-T cells; GTC, n = 4; NR4A1 SKO, n = 5; SOCS3 SKO and DKO, n = 6. Samples with fewer than 100 FACS-isolated positive cells were excluded from this analysis. k, Quantification of CD27+ CAR-T cells; GTC, n = 6; NR4A1 SKO, n = 7; SOCS3 SKO, n = 6; DKO, n = 8. Samples with fewer than 100 FACS-positive cells were excluded from this analysis. l, Quantification of CD25 expression of CAR-T cells from U87-GL-EGFRvIII tumour samples. l-n, n is the same as in Fig. 5j. m, Quantification of CD69 expression of CAR-T cells from U87-GL-EGFRvIII tumour samples. n, Quantification of CD44 expression of CAR-T cells from U87-GL-EGFRvIII tumour samples. o, Spider plots of MCF7-GL tumour volumes in different CAR-T treatment groups. p, Bioluminescence imaging of the PDX B-ALL model after CAR-T treatment; n = 5 mice for each group. q, Schematics of mantle cell lymphoma experimental design. r, Quantification of bone marrow CD45RO+ CD62L+ CAR-T cells on day 12. n = 3 mice for each group. Data in b-g, i-n, r are shown as means + SEM. Significance levels were calculated using one-way ANOVA with Dunnett’s multiple-comparisons test versus the indicated control, GTC or PBS, as appropriate (b-f,i,r), or one-way ANOVA with Tukey’s multiple-comparisons (g,j-n). n indicates biological replicates unless otherwise stated. Exact P values and detailed statistics are provided in the Source Data. NS, not significant, *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. CAR-T cells were generated using human donor cells independent of those used in the screen.
Extended Data Fig. 10 NR4A1-SOCS3 DKO CAR-T cells maintain an in vivo safety profile.
a, Top two rows, H&E-stained lung and liver sections from representative mice in different treatment groups from the U87-GL-EGFRvIII validation experiment, from Fig. 5i. Bottom two rows, human CD3 IHC staining of lung and liver sections from representative mice in the same experiment. Similar results were observed in all examined mice. Scale bar, 200 μm. b, Quantification of serum human IL-6 and IFN-γ in NSG mice bearing U87-GL-EGFRvIII tumour after treatments with different CAR-T cell groups. Data are shown as means + SEM. n is the same as in Fig. 5j, and indicates each mouse. Significance levels were calculated using one-way ANOVA with Dunnett’s multiple-comparisons test versus GTC. c, Quantification of human inflammatory cytokines at different time points in the PDX2 model. Data are shown as means ± SEM. On days 5 and 16, n = 5 mice per group. On day 41, n = 2 for NR4A1 SKO, n = 1 for SOCS3 SKO and n = 5 for DKO; no PBS or GTC samples. d, Representative H&E, human CD3 IHC and human CD19 IHC staining of brain, liver, lung and spleen sections from four biologically independent DKO CAR-T-treated mice in the PDX2 validation experiment on day 87, shown in Fig. 5o. Arrows indicate representative positively stained cell signal. Similar results were observed in all examined mice. Scale bar, 300 μm. e, Variant distribution in four samples, DKO, NR4A1 SKO, SOCS3 SKO and GTC, visualized by Circos plots. Chromosomal translocations spanning genomic regions greater than 50 bp are shown as red links. f, Quantification of INDEL variants across all chromosomes in the four samples shown in e. g, Quantification of CAR-T cell numbers during G-REX production. Data are shown as means ± SEM. n = 3 biologically independent donors. Schematic in panel g created in BioRender; Chen, S. https://biorender.com/d6kqeuq (2026).
Rights and permissions
Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
About this article
Cite this article
Zhou, L., Renauer, P.A., Tang, K. et al. In vivo CAR-T double knockout screen maps synergistic antitumour gene pairs. Nature (2026). https://doi.org/10.1038/s41586-026-10956-z
Received: 21 December 2023
Accepted: 27 July 2026
Published: 04 November 2026
Version of record: 04 November 2026
DOI: https://doi.org/10.1038/s41586-026-10956-z
来源:Nature 综合主刊 · nature.com