通过机构审查董事会批准的知情书面同意书（Stemcell Technologies），从新鲜的外周血白细胞（70500，Stemcell Technologies）中分离出CD4+调节和效应T细胞。使用离心机用1倍的EasySp Buffer（DPB，2％FBS和1 mM EDTA（pH 8.0））洗涤白细胞的含量两次。根据制造商的协议，将洗涤的细胞重悬于每毫升缓冲液中的200×106细胞中，并用EasySep人CD4+ CD127LOWCD25+调节性T细胞分离试剂盒（18063，Stemcell Technologies）分离。Following isolation with the kit, Treg cells were stained Alexa Fluor 647 anti-human IL-2Rα antibody (302618, BioLegend; diluted 1:25), phycoerythrin anti-human CD127 (557938, Beckon Dickinson; diluted 1:50) and Pacific Blue anti-human CD4 antibody (344620, BioLegend;稀释1:50），并用在BD FACS ARIA Fusion 1（656700）上进行的FACS分离，以确保纯种群而不会污染效应细胞。在对纯CD4+ CD127LOWCD25+ Treg细胞进行排序后，将细胞以1×106个细胞在XVIVO-15（02-053Q，LONZA）中以1×106的细胞接种，并补充了5％FCS，55 µM 2- erigAto乙醇，4 mm n- acetyl-cysteyl-cystey and-cystey and 2001111 IN-1 IIN-2-1 IN-AM AM AMIR-AM AMIR-AM AMIRER-2（卑尔根）。在补充有10％FCS的RPMI-1640中，将TEFF细胞以1×106个细胞播种，2 mM-谷氨酰胺（25030081，Fisher Scientific），10 mM HEPES（H08877-100ML，SIGMA，SIGMA），1X MEM非骨骼氨基氨基酸（1x Mem noce Enter Ectementy氨基氨基酸）（11360070，Fisher Scientific），100 U ML-1青霉素 - 链霉素（P4333-100ML，Sigma）和50 U ML-1 IL-2（10101641，Amerisource Bergen）。然后，用免疫型人类CD3/CD3/CD28/CD2 T细胞激活剂（10990，Stemcell Technologies）在Treg细胞中以25 µl ml -1和TEFF细胞的6.25 µL ML -1刺激两个细胞子集。用5％CO2在37°C培养细胞。激活和电穿孔后， 每48小时每48小时拆分细胞1：2，以维持每毫升1×106个细胞的密度，并用各自剂量的IL-2补充。

　　按照先前描述的协议进行合并屏幕19。简而言之，在刺激和镀T细胞后24小时，将反式调节室植物库19添加到每种培养物中（补充表6）。在转导之前对细胞进行计数，并在感染的多种感染中添加病毒，并使用温和的混合分散病毒培养基而不会破坏细胞捆绑。然后将细胞在37°C下再孵育24小时，通过离心结合，然后将病毒培养基替换为补充IL-2的新培养基。

　　洗涤后二十四小时，通过以150克离心10分钟将细胞固定，每17.8 µL添加为1.5×106个细胞，补充了p3原细胞核对象缓冲液（V4SP-3960，LONZA），并与7.2 µL核糖核纤维蛋白酶颗粒（Reprile riperile ryprile renp）组合在一起（Rep）106-1-1-1-1-1-1-1-1-1-1-1-1-1- rnps/1- rnp）。水库。混合细胞和RNP后，将25 µL混合物分布在96孔核维特板的孔中（V4SP-3960，Lonza的组件）。使用96孔班车的Lonza 4D-核对象系统上的Treg细胞的Code EO-115和Teff细胞的EH-115核定细胞。核反理后立即将90 µL预热的适合细胞的培养基添加到每个孔中，并将细胞在37°C下孵育15分钟。孵育后，在补充IL-2的培养基中，将细胞以每毫升的1×106细胞接种。

　　编辑之前，在编辑后至少将转导和电穿孔的细胞扩展至少6天。分离静止筛网后10天进行细胞分选。对于受刺激的TEFF屏幕，将细胞用免疫型人CD3/CD3/CD28/CD2 T细胞激活剂（10990，Stemcell Technologies）在初次分离后9天，然后在峰值IL-2Rα表达时进行重新求解后72小时进行排序。在分类之前，对细胞进行计数，用EaseSep缓冲液洗涤一次，并用Alexa Fluor 647抗人类IL-2Rα抗体染色（302618，Biolegend； Biolegend；稀释1:25）。然后将细胞洗涤并重悬于EaseSep缓冲液中。在分类过程中，将细胞在GFP+群体（慢病毒SGRNA库标记）上门控，并且将表达IL-2Rα的顶部和底部的20％的表达IL-2Rα的细胞分类为涂有FCS的15毫升锥形管。分离的细胞被整合，计数和裂解。使用苯酚 - 氯仿提取物进行基因组DNA提取，并将SGRNA文库放大并准备使用自定义引物进行测序。图书馆在UCSF猫的Illumina Hiseq 4000上进行了测序。

　　用Mageck42（v0.5.9.5）分析所有合并的屏幕。使用-norm方法对所有供体进行了Mageck计数，然后进行Mageck测试-Sort-Sort-Surteria POS，以识别导致IL-2Rα表达统计学上显着变化的基因。结果被计算为IL-2Rαlow箱/IL-2Rα高bin。屏幕可视化表示为IL-2RαHighbin/il-2rαlowbin，通过翻转折叠变化的符号。所有具有FDR调整后P的基因< 0.05 were considered significant.

　　Guide-loaded Cas9 RNPs were assembled with custom CRISPR RNAs (crRNAs) (Dharmacon), which were resuspended in IDT duplex buffer (11-01-03-01, IDT) at 160 µM. Sequences are provided in Supplementary Table 6. Dharmacon Edit-R CRISPR–Cas9 synthetic tracrRNA (U-002005-20, Dharmacon) also resuspended in nuclease-free duplex buffer at 160 µM was combined at a 1:1 molar ratio in a 96-well plate and incubated at 37 °C for 30 min. Single-stranded donor oligonucleotides (sequence: TTAGCTCTGTTTACGTCCCAGCGGGCATGAGAGTAACAAGAGGGTGTGGTAATATTACGGTACCGAGCACTATCGATACAATATGTGTCATACGGACACG; 100 µM stock) was added to the complex at a 1:1 molar ratio and incubated at 37 °C for 5 min. Finally, Cas9 protein (MacroLab; 40 µM stock) was added at a 1:2 molar ratio and incubated at 37 °C for 15 min. The resulting RNPs were frozen at −80 °C until the day of electroporation and were thawed to room temperature before use. Forty-eight hours following T cell activation, the cells were pelleted at 100g for 10 min and resuspended in room temperature P3 Primary Cell Nucleofector Buffer (V4XP-3032, Lonza) at 1.5 × 106 cells per 17.8 µl. Cells (1.5 × 106) were transferred to each RNP-containing well and mixed gently. Of the combined RNP cell solution, 25 µl was transferred to a 96-well electroporation cuvette plate (VVPA-1002, Lonza) and nucleofected with pulse code DS-137. Immediately following electroporation, the cells were gently resuspended in 90 µl warmed media and incubated at 37 °C for 15 min. After recovery, the cells were cultured in 96-well round-bottom plates at 1 × 106 cells per millilitre for the duration of the experiment. To prevent edge effects, the sgRNAs were randomly distributed across each plate, and the first and last columns and rows of each plate were filled with PBS to prevent evaporation. Unless otherwise specified, CRISPR–Cas9-edited cells were restimulated on day 8 following isolation for stimulation response arrayed assays with ImmunoCult human CD3/CD28/CD2 T cell activator (10990, STEMCELL Technologies).

　　On the final day of the respective assay, genomic DNA was isolated using DNA QuickExtract (QE09050, Lucigen) according to the manufacturer’s protocol. Primers were designed to flank each sgRNA target site. Amplicons of the region were generated by adding 1.25 µl each of forwards and reverse primer at 10 µM to 5 µl of sample in QuickExtract, 12.5 µl of NEBNext Ultra II Q5 master mix (M0544L, NEB) and H2O to a total 25 µl reaction volume. Touchdown PCR was used with the following cycling conditions: 98 °C for 3 min, 15 cycles of 94 °C for 20 s followed by 65 °C to 57.5 °C for 20 s (0.5 °C incremental decreases per cycle) and 72 °C for 1 min, and a subsequent 20 cycles at 94 °C for 20 s, 58 °C for 20 s and 72 °C for 1 min, and a final 10-min extension at 72 °C. Amplicons were diluted 1:200 and Illumina sequencing adapters were then added in a second PCR. Indexing reactions included 1 µl of the diluted PCR1 sample, 2.5 µl of each the forwards and the reverse Illumina TruSeq indexing primers at 10 µM each, 12.5 µl of NEB Q5 master mix and H2O to a total 25 µl reaction volume. The following PCR cycling conditions were used: 98 °C for 30 s, followed by 98 °C for 10 s, 60 °C for 30 s and 72 °C for 30 s for 12 cycles, and a final extension period at 72 °C for 2 min. Samples were pooled at an equivolume ratio and SPRI purified before sequencing on an Illumina MiSeq with PE 150 reads. Analysis was performed with CRISPResso2 (v2.2.7)43 CRISPRessoBatch --skip_failed --n_processes 4 --exclude_bp_from_left 5 --exclude_bp_from_right 5 --plot_window_size 10.

　　The BioLegend FoxP3 Fix/Perm kit (421403, BioLegend) was used for staining according to the manufacturer’s protocol. Cells were washed in EasySep buffer before extracellular staining. Cells were stained with Alexa Fluor 647 anti-human IL-2Rα (CD25) antibody diluted 1:25 (302618, BioLegend), Ghost Dye Red 780 diluted 1:1,000 (13-0865-T500, Tonbo) and BV711 anti-human CD4 diluted 1:50 (344648, BioLegend) for 20 min at 4 °C and then washed once with EasySep buffer. After fixing and permeabilizing according to the kit, intracellular staining was performed with phycoerythrin anti-mouse/human Helios antibody (137216, BioLegend), KIRAVIA Blue 520 anti-human CD152 (also known as CTLA-4) antibody (349938, BioLegend) and Pacific Blue anti-human FOXP3 antibody (320116, BioLegend) each diluted 1:50 in permeabilization buffer for 30 min at room temperature. Cells were subsequently washed in permeabilization buffer and resuspended in EasySep buffer before running on the Thermo Fisher Attune NxT flow cytometer (A29004). Analysis of flow data was performed in FlowJo (v10.8.1). Gating was performed to select for lymphocytes, singlets, live cells (Ghost Dye negative) and CD4+ cells in the specified order. This population was then used to calculate the median fluorescence intensity for IL-2Rα or CTLA4. Visualization was performed in R using ggplot2 (v3.4.1).

　　CRISPRi sgRNAs for Perturb-seq were selected from the Dolcetto library44 and cloned into the LGR2.1 plasmid backbone (Addgene #108098). A lenti EF1a-Zim-3-dCas9-P2A-BSD was generated using Gibson assembly as previously described45. Lentivirus was prepared according to the a previous protocol25.

　　Twenty-four hours after stimulation of isolated human Treg cells and Teff cells from two donors, the cells were transduced with Zim3–dCas9 lentivirus at 3% v/v. The following day, Perturb-seq sgRNA library lentivirus was added at 0.75% v/v (multiplicity of infection of 0.3). Forty-eight hours after transduction with Zim3–dCas9, 10 mg ml−1 blasticidin (A1113903, Gibco) was added to each sample to select for dCas9+ cells. Blasticidin was replenished every 48 h until the cells were processed for sequencing. Eight days after initial isolation and stimulation of cells, half of the Treg and Teff cell culture was restimulated with ImmunoCult human CD3/CD28/CD2 T cell activator (10990, STEMCELL Technologies). On the tenth day after initial isolation, the resting and 48-h restimulated samples were collected for 10X single-cell sequencing. First, cells from each donor within the same stimulation and cell-type condition were pooled at equal concentrations. Sorting was performed to isolate live GFP+ cells from each condition. Sorted cells were processed according to the Chromium Next GEM Single Cell 5′ HT Reagent Kits v2 (Dual Index) with Feature Barcode technology for CRISPR Screening and Cell Surface Protein guide User Guide, CG000513. In brief, sorted cells were pelleted and washed once with cell staining buffer (420201, BioLegend). Next, the samples were blocked with Human TruStain FcX Fc Blocking reagent (422302, BioLegend). Meanwhile, TotalSeq-C Human Universal Cocktail V1.0 (399905, BioLegend) was prepared using cell staining buffer (420201, BioLegend), and TotalSeq-C0251 anti-human hashtag antibodies 1–4 (394661, BioLegend) were added to aliquots of the cocktail. After blocking, cells were stained with TotalSeq-C cocktail including one hashtag per cell and stimulation condition. After staining, the cells were washed three times in cell staining buffer. The samples were then resuspended in PBS with 1% BSA (Gibco) for final counting. The resulting samples were pooled across conditions and approximately 65,000 cells per well were loaded into eight wells of a Chromium Next GEM Chip N Single Cell Kit (1000375, 10X Genomics) for GEM generation. The samples were prepared for sequencing using the Chromium Next GEM (Gel Bead-in-emulsion) Single Cell 5′ HT Kit v2 (1000374), 5′ Feature Barcode Kit (1000256) and 5′ CRISPR Kit (1000451) according to the manufacturer’s protocol. GEM generation and library preparation were performed by the Gladstone Genomics Core. The resulting libraries were sequenced using a NovaSeqX Series 10B flowcell (20085595, Illumina) at the UCSF CAT.

　　Fastqs for each 10X well were concatenated across lanes and flow cells. Alignment of Perturb-seq data and count aggregation for the gene expression, CRISPR sgRNA and antibody-derived tag (ADT) libraries was performed with cellranger46 count (v7.1.0) using the default settings and –expect-cells=45000 –chemistry=SC5P-R2. Gene expression fastqs were aligned to ‘refdata-gex-GRCh38- 2020-A’ human transcriptome reference acquired from 10X Genomics. SgRNA sequences were aligned to a custom reference file using the pattern TAGCTCTTAAAC(BC), whereas ADTs were aligned to the TotalSeq-C-Human-Universal-Cocktail-399905-Antibody-reference-UMI-counting.csv provided by BioLegend, also including the hashtag oligo (HTO) sequences, which were used to distinguish each cell-type and stimulation condition. Counts for each respective library were aggregated across wells with cellranger aggr using the default settings. Cells were assigned to a donor using genetic demultiplexing with Souporcell47 (https://github.com/wheaton5/souporcell). For each well, souporcell_pipeline.py was run using the bam file and cellranger count output barcodes.tsv as input in addition to the reference fasta. Donor calls shared across wells were identified using shared_samples.py using the vcf file outputs from Souporcell.

　　Perturb-seq analysis was performed in R (v4.3.1) using Seurat48 (v4.3.0.1) based on code previously published49. Count matrices were imported into R using the Seurat Read10X function. After creating a Seurat object with CreateSeuratObject, quality filtering was performed to retain cells with more than 1,000 RNA features identified and less than 7.5% mitochondrial RNA. Cells without a singular donor assignment were also excluded from the object as well as cells with more than one HTO assignment as determined after running HTODemux. Low abundance transcripts were filtered using the threshold of ten cells per feature and TCR genes were removed from the primary RNA assay as they were found to be a major source of variance in the dataset. No sgRNA targets were removed as the number of cells in each condition exceeded the threshold set of 150 cells. After filtering, gene expression counts were normalized and transformed using the Seurat SCTransform function with regression of both S phase score and G2/M phase score, as described on Satija (https://satijalab.org/seurat/articles/cell_cycle_vignette.html). ADT counts were normalized using the centred log-ratio (CLR) normalization method of NormalizeData. After generating principal component analysis of both normalized and transformed RNA and ADT data, Harmony50 (v0.1.1) was used to correct for donor-associated variability in the dataset. The resulting normalized and transformed counts were used for downstream analysis unless otherwise specified. Uniform manifold approximation and projections (UMAPs) were generated using the transformed and corrected RNA and ADT counts with Seurat function FindMultiModalNeighbors followed by RunUMAP using weighted.nn. Before cell-type-specific analysis, Treg cells were manually filtered to include only cells belonging to clusters with FOXP3 and IKZF2 expression to maximize cell purity (clusters 1, 7, 8, 15, 6, 4, 19, 20, 17 and 23).

　　Activation scoring was performed according to Schmidt et al.25,49. In brief, Seurat FindMarkers was used to identify differentially expressed genes between stimulated and resting non-targeting control cells within the Teff cells and Treg cells individually. Genes that had a log2-transformed fold change of more than 0.25 and were detected in 10% of restimulated or resting cells were used to generate gene weights for the score calculated as sum(GE × GW/GM), where GE is the normalized/transformed expression count of a gene, GW is the weight of the gene, and GM is the mean expression of the gene in non-target control cells of the respective cell type. Wilcoxon tests were performed to determine significance compared with non-targeting control cells with Bonferroni correction for multiple hypothesis testing (Supplementary Table 7). To observe the effect of each sgRNA within independent cell and stimulation conditions, the cells were subset by HTO. RNA and ADT normalization, transformation and donor variability correction were repeated for each subset as described above for the combined dataset. UMAPs were generated using the transformed and corrected RNA and ADT counts with Seurat function FindMultiModalNeighbors followed by RunUMAP using weighted.nn. Cell cycle quantification for each subset was performed using cycle assignments generated using the Satija cell cycle vignette referenced above.

　　Pseudobulking of resting and stimulated Treg and Teff cell samples was performed using Seurat Aggregateexpression grouped by HTO, target gene and donor pulling from the counts slot (sgRNAs targeting the same gene were collapsed within the same donor). Differential expression analysis was performed with the resulting pseudobulked raw counts for both RNA and ADTs. DESeq2 (v1.32.0)51 was used to identify differentially expressed genes and proteins between each sgRNA and non-targeting control sample within each cell-type and stimulation condition, using donor information as a covariate. Network plots of differentially expressed gene connections were visualized in R using influential52 (v2.2.7) and ggraph53 (v2.1.0), including only genes with an adjusted P < 0.05. Other visualization of differentially expressed genes and surface proteins was performed using ggplot2 (v3.4.1).

　　At their respective timepoints, resting and 48-h restimulated cells were pelleted and resuspended at 1 × 106 cells per 300 µl of RNA lysis buffer (R1060-1-100, Zymo). Cells were pipette mixed and vortexed to lyse and frozen at −80 °C until RNA isolation was performed. RNA was isolated using the Zymo-Quick RNA micro prep kit (R1051) according to the manufacturer’s protocol with the following modifications: after thawing the samples, each sample was vortexed vigorously to ensure total lysis before loading into the extraction columns. The optional kit provided DNAse step was skipped, and instead RNA was eluted from the isolation column after the recommended washes and digested with Turbo-DNAse (AM2238, Fisher Scientific) at 37 °C for 20 min. Following digestion, RNA was purified using the RNA Clean & Concentrator-5 kit (R1016, Zymo) according to the manufacturer’s protocol. The purified RNA was submitted to the UC Davis DNA Technologies and expression Analysis Core to generate 3′ Tag-seq libraries with unique molecular indices (UMIs). Barcoded sequencing libraries were prepared using the QuantSeq FWD kit (Lexogen) for multiplexed sequencing on a NextSeq 500 (Illumina).

　　RNA-seq data were processed using the pipeline previously described19. In brief, fastq adapter trimming was performed with cutadapt (v2.10). Low-quality bases were trimmed with seqtk (v0.5.0). Reads were then aligned with STAR54 (v2.7.10a) and mapped to GRCh38. UMI counting and deduplication was performed with umi_tools55 (v1.0.1) and gene counts were generated from the deduplicated reads using featureCounts (subread v2.0.1) using Gencode v41 basic transcriptome annotation. Quality control metrics were generated for each sample with Fastqc56 (v0.11.9), rseqc57 (v3.0.1) and Multiqc58 (v1.9). Differentially expressed genes between Mediator knockouts and AAVS1-knockout samples as well as stimulated and resting AAVS1-knockout samples (Supplementary Table 8) were identified from the deduplicated count matrix using DESeq2 (v1.32.0)51 in R (v4.1.0). Comparisons were made within each cell-type and stimulation condition across three donors, using donor ID as a covariate in the model. Normalized counts were generated using a DESeqDataSet containing all samples, followed by estimateSizeFactors and counts(normalized=TRUE). AAVS1-knockout normalized sample counts were then subset and averaged across donors for visualization.

　　Differentially expressed genes for MED12-knockout versus AAVS1-knockout samples were defined by a cut-off of adjusted P < 0.05 (Supplementary Table 3). Comparison of the effects of MED12-knockout differentially expressed genes across stimulation-responsive categories was performed by grouping MED12-knockout versus AAVS1-knockout differentially expressed genes according to their stimulation-responsive behaviour in control cells (stimulation response = adjusted P < 0.05 and abs(log2 fold change) > 1). The Bonferroni-adjusted P value resulting from a two-tailed t-test is displayed (Fig. 4a), comparing each stimulation-responsive group to the non-stimulation-responsive group. The boxplot centre line denotes the median; the box limits indicate the upper and lower quartiles; the whiskers denote the 1.5-times interquartile range (genes per group (downregulated, not stimulation responsive and upregulated) = resting Teff cells: 272, 954 and 218; stimulated Teff cells: 242, 1,432 and 467; resting Treg cells: 269, 1,491 and 241; and stimulated Treg cells: 245, 1,945 and 426).

　　A one-sided Fisher’s exact test for regulators of IL-2Rα within the differentially expressed genes downstream of MED12 was determined using screen results from the matched cell-type and stimulation conditions (Fig. 4b). Genes were subset to those targeted in the screen library and detected in CD4+ T cell bulk RNA-seq (genes per group: regulators, non-regulators = resting Teff cells: 62 and 807; stimulated Teff cells: 41 and 824; and resting Treg cells: 82 and 787). Pathway analysis was performed using PathfindR59 (v1.6.4) including KEGG, Reactome and GO-BP gene sets and the lowest P value is displayed. Visualization was performed after removing KEGG disease pathways. Apoptosis pathway visualization was performed using Cytoscape60 (v3.8.2). Gene set enrichment analysis was performed with clusterProfiler61 (v4.10.1) using msigdbr (v7.5.1) on all human gene sets.

　　SEL120-34A (S8840, Selleckchem) was reconstituted in ultrapure H2O according to the manufacturer’s recommendations. Cells were treated every 48 h with a 1 µM dose, and treatment was started 48 h following cell isolation to align with the time at which cells are edited in CRISPR-based experiments. Restimulation of cells for flow cytometry and CUT&RUN was performed 10 days after initial isolation.

　　Immunoprecipitation base buffer (0.05 M Tris-HCl pH 7.5, 0.15 M NaCl, 0.001 M EDTA and AP MS water) was prepared the day of the experiment. Of resting and 48-hour restimulated cells, 20 × 106 cells per sample and immunoprecipitation were washed twice with PBS. Samples were then lysed in 500 µl lysis buffer per 10 × 106 cells (base buffer, 1X PhosphoStop (04906837001, Roche), 1X Complete mini-EDTA protease inhibitor cocktail tablets (11836170001, Sigma-Aldrich), 0.50% NP-40 Surfact-Amps Detergent Solution (85124, Thermo Scientific) and incubated on nutator for 30 min at 4 °C. To digest chromatin, tip sonication was performed in round with incubation on ice between each step: 7 s 12%, 7 s 12%, 7 s 12% and 7 s 15% with four rounds of sonication total. Cell lysate was clarified by centrifugation at 3,500g for 10 min at 4 °C. A bicinchoninic acid (BCA) assay was performed for each sample, and protein concentrations were normalized across conditions. Of whole-cell lysate, 10% was reserved for input, and samples were split into MED12 (14360, Cell Signaling Technologies) immunoprecipitation and rabbit IgG isotype control (3900, Cell Signaling Technologies) immunoprecipitation conditions. In each case, 10 µg antibody was added to a 1.5 ml protein lo bind tube containing clarified protein and samples were incubated overnight at 4 °C, with rotation on a nutator. In the morning, Pierce protein A + G magnetic beads (88802, Thermo Fisher) were washed four times using 1 ml of lysis buffer per 1 ml of bead slurry, allowing the beads to bind to a magnet between each wash before removing the buffer. After the final wash, beads were resuspended in lysis buffer at the original bead slurry volume, and 50 µl was added to each sample. The lysate–antibody–bead mixture was then incubated at 4 °C for 2 h with rotation on a nutator. After incubation, beads were bound to a magnetic tube rack and washed one time with immunoprecipitation buffer + NP-40 (immunoprecipitation buffer + 0.05% NP-40) followed by three washes with a 900 µl immunoprecipitation buffer. The resulting purified proteins were processed for mass spectrometry or western blot.

　　After immunoprecipitation, bound proteins were lysed in 8 M urea + 25 mM ammonium bicarbonate followed by reduction (5 mM dithiothreitol for 1 h at 37 °C), alkylation (10 mM iodoacetamide for 45 min at room temperature in the dark) and digestion overnight with 1 µg of trypsin (Promega). Peptide samples were applied to activated columns, and the columns were washed three times with 200 µl of 0.1% trifluoroacetic acid. Peptides were eluted with 140 µl of 50% acetonitrile and 0.1% trifluoroacetic acid and dried down by speedvac.

　　Samples were resuspended in 0.1% formic acid and separated by reversed-phase chromatography using an EASY-nLC instrument (Thermo Fisher Scientific) with a 15-cm PepSep column (inner diameter of 150 µm; Bruker). Samples were acquired by data-dependent acquisition. Mobile phase A consisted of 0.1% formic acid in water, and mobile phase B consisted of 80% acetonitrile and 0.1% formic acid. Peptides were separated at a flow rate of 500 nl min−1 over the following 60 min gradient: 4–35% B in 44 min, 35–45% B in 5 min and 10 min at 88% B. Peptides were analysed by an Orbitrap Lumos MS instrument (Thermo Fisher Scientific). Data were collected in positive ion mode with MS1 resolution of 240,000, 350–1,350 m/z scan range, maximum injection time of 50 ms, radiofrequency lens of 30%. For data-dependent acquisition, MS2 fragmentation was performed on charge states 2–5 with a 20-s dynamic exclusion after a single selection and 10 ppm ± mass tolerance. All raw mass spectrometry data were searched using MaxQuant (v2.4.7) against the human proteome (UniProt canonical protein sequences, downloaded in September 2022) using default settings and with a match-between-runs enabled62.

　　Protein spectral counts as determined by MaxQuant search results were used for protein–protein interaction (PPI) confidence scoring by SAINTexpress63 (v3.6.1). Rabbit IgG pulldown samples were used as control. The total list of candidate PPIs was filtered to those that met the criteria of SAINTexpress Bayesian FDR ≤ 0.05. To quantify changes in interactions between resting and stimulated T cell states, we used a label-free quantification approach in which statistical analysis was performed using MSstats (v4.8.7)64 from the artMS (v1.18.0) R package. Visualization was performed in Cytoscape with additional connections included from the STRING database65.

　　After affinity purification of proteins, beads were resuspended in 100 µl 2X sample buffer (4× Laemmli Sample Buffer; 1610747, Bio-Rad) with 1:10 β-mercaptoethanol (63689-25ML-F, Sigma) diluted 1:1 with 500 µl lysis buffer. Samples were boiled for 5 min at 95 °C and stored at −20 °C until further processing. Western blots were performed as previously published66. In brief, cell lysates were subjected to SDS–PAGE on 4–15% acrylamide gels and electroblotted to polyvinylidene difluoride membranes. Blocking and primary (diluted 1:1,000) and secondary antibody incubations of immunoblots were performed in Tris-buffered saline + 0.1% Tween-20 supplemented with 5% (w/v) BSA (antibodies are provided in Supplementary Table 9). Horseradish peroxidase-conjugated goat anti-rabbit and IgG (Southern Biotech) were used at a dilution of 1:30,000, and immunoreactive bands were detected using Pierce ECL Western Blotting Substrate (32106) according to the manufacturer’s instructions.

　　CUT&RUN was performed on resting and 48-h restimulated cells according to the manufacturer’s protocol with the EpiCypher CUTANA ChIC/CUT&RUN Kit and provided reagents. Samples for H3K27ac CUT&RUN were lightly crosslinked before isolation using 0.1% formaldehyde (252549, Sigma) for 1 min and quenched with 125 mM glycine (50046, Sigma). In brief, 5 × 105 T cells per reaction were washed with PBS before nuclear isolation using the EpiCypher recommended lysis buffer consisting of 20 mM HEPES pH 7.9 (Sigma-Aldrich), 10 mM KCl (Sigma-Aldrich), 0.1% Triton X-100 (Sigma-Aldrich), 20% glycerol (Sigma-Aldrich), 1 mM MnCl2 (Sigma-Aldrich), 1X cOmplete Mini-Tablet (11873580001, Roche) and 0.5 mM spermidine (Sigma-Aldrich). The cells were resuspended in 100 µl per reaction cold nuclear extraction buffer and incubated on ice for 10 min. Following lysis, nuclei were pelleted and resuspended in 100 µl per reaction of nuclear extraction buffer. The isolated nuclei were then frozen at −80 °C in extraction buffer until DNA isolation. After thawing the samples at 37 °C, the nuclei were bound to activated conA beads. After adsorption of nuclei to beads, permeabilization was performed with 0.01% digitonin-containing buffer. Antibodies for H3K27ac (13-0045, EpiCypher), H3K4me1 (13-0057, EpiCypher), H3K4me2 (13-0027, EpiCypher), H3K4me3 (13-0041, EpiCypher) and IgG (13-0042, EpiCypher) were added at 500 ng per reaction. Following overnight antibody binding, pAG-MNase addition and chromatin cleavage, 0.5 ng of the provided Escherichia coli DNA was added to each sample following chromatin cleavage by MNase. Before DNA isolation, crosslinked samples were digested overnight with proteinase K (AM2546, Invitrogen) as recommended. The provided spin columns and buffers were used for DNA isolation and purification. The resulting DNA was prepared for sequencing using the CUTANA CUT&RUN Library Prep Kit (14-1002) according to the manufacturer’s protocol.

　　Pooled libraries were sequenced on a NextSeq 500 (H3K27ac) and NextSeq 2000 with 2 × 75 or 2 × 50 paired-end reads, respectively. Bcl2fastq (v2.19) with the settings --minimum-trimmed-read-length 8 was used to generate fastqs. CUT&RUN data analysis was performed according Zheng et al. with the recommended settings unless otherwise specified below67. In brief, the fastqs were trimmed with cutadapt (v1.18). Bowtie2 (v2.2.5)68 was used to align the trimmed fastqs to GRCh38 using settings --local --very-sensitive --no-mixed --no-discordant --phred33 --dovetail -I 10 -X 700 -p 8 -q and E. coli (EMBL accession U00096.2) with settings --local --very-sensitive --no-overlap --no-dovetail --no-mixed --no-discordant --phred33 -I 10 -X 700 -p 8 -q. Bam files were generated with SAMtools69,70 (v1.9) view -bS -F 0 × 04 and bam-to-bed conversion performed with bedtools (v2.30.0) bamtobed -bedpe. Bedfiles were filtered to include only paired reads of less than 1,000 bp with the command awk ‘$1==$4 & & $6-$2 < 1000 {print $0}’ samplename.bed before generating bedgraph files using bedtools (v2.30.0) genomecov -bg. Peak calling was performed using the bedgraph files as input with SEACR71 (v1.3). Each target bedgraph file was compared to the respective donor and knockout condition IgG file to identify peaks above the background using the norm and stringent options for H3K27ac samples. Spike-in scaling was performed before methylation peak calling with SEACR using the IgG file as background without normalization (non option) and with the stringent option.

　　Before generating a peak by sample matrix for each target, ChIP–seq blacklist regions were removed from the data. The sample matrix was reduced across all peaks within the dataset, and H3K27ac peaks were segmented into regions of 5,000 bp maximum length. Regions of differential acetylation or methylation between the regulator knockouts and AAVS1-knockout samples were identified for the peaks called across any of the samples from bam files using DESeq2 (v1.32.0)51 in R (v4.1.0; Supplementary Table 10). Comparisons were made within each cell-type and stimulation condition using AAVS1s prepared in the same batch of samples. Gene annotation was performed using the gene with the nearest TSS to each region with the GenomicRanges72 (v1.44.0) nearest function. Final bedgraph scaling was performed based on peak coverage across all samples and conditions using DESeq2 (v1.32.0) sizefactors. SEL120-34A and H2O treatment samples were compared as described for MED12-knockout and AAVS1-knockout samples, using the peak matrix from MED12-knockout and AAVS1-knockout samples to maximize detection of overlapping regions across datasets.

　　A portion of edited Teff cells were restimulated with ImmunoCult human CD3/CD28/CD2 T cell activator (10990, STEMCELL Technologies) 10 days following isolation and collected 48 h later. Up to 1–2 × 106 Teff cells were crosslinked in PBS with 1% methanol-free formaldehyde (28908, Thermo) for 10 min at 18–22 °C followed by quenching in glycine at 125 mM final concentration. Crosslinked cell pellets were snap-frozen in liquid nitrogen and stored at –80 °C. Nuclei were isolated from thawed, crosslinked cells via sequential lysis in LB1 (50 mM HEPES-KOH pH 7.5, 140 mM NaCl, 1 mM EDTA, 10% glycerol, 0.5% IGEPAL CA-360 and 0.25% Triton X-100), LB2 (10 mM Tris-HCl pH 8, 200 mM NaCl, 1 mM EDTA and 0.5 mM EGTA) and LB3 (10 mM Tris-HCl pH 8, 100 mM NaCl, 1 mM EDTA, 0.5 mM EGTA, 1% sodium deoxycholate (NaDOC) and 0.5% N-laurylsarcosine) supplemented with 0.5 mM phenylmethylsulfonyl fluoride (PMSF; P7626, Sigma) and 0.5X protease inhibitor cocktail (PIC; P8340, Sigma). Chromatin was sheared on a Covaris E220-focused ultrasonicator using 1-ml milliTubes (520128, Covaris) with 140 W peak incident power, 5% duty factor, 200 cycles per burst, 6 °C temperature setpoint (minimum of 3 °C and maximum of 9 °C), fill level 10, and time 12–14 min to obtain a target size of 200–700 bp. Formaldehyde crosslinked, sheared mouse CD8+ T cell chromatin was spiked in at 2.5% of human Teff chromatin based on fluorometric (Qubit, Q33238, Thermo) or OD260 (Nanodrop, 912A1099, Thermo) quantification. Triton X-100 was added to a final concentration of 1% before immunoprecipitation for 16 h at 4 °C with 2–8 µg of indicated antibodies (Supplementary Table 9) bound to a 1:1 mixture of protein A and protein G magnetic beads (10001D and 10003D, Thermo). Bead-bound antibody–chromatin complexes were sequentially washed three times with wash buffer 1 (20 mM Tris pH 8, 150 mM NaCl, 1 mM EDTA, 0.5 mM EGTA, 1% Triton X-100, 0.1% SDS and 0.1% NaDOC), twice with wash buffer 2 (20 mM Tris-HCl pH 8, 500 mM NaCl, 1 mM EDTA, 0.5 mM EGTA, 1% Triton X-100, 0.1% SDS and 0.1% NaDOC), twice with wash buffer 3 (20 mM Tris-HCl pH 8, 250 mM LiCl, 1 mM EDTA, 0.5% IGEPAL CA-360 and 0.5% NaDOC), twice with TET (10 mM Tris-HCl pH 8, 1 mM EDTA and 0.2% Tween-20) and once with TE0.1 (10 mM Tris-HCl pH 8, 0.1 mM EDTA, 0.5 mM PMSF and 0.5X PIC) supplemented with 0.5 mM PMSF and 0.5X PIC. Beads were resuspended in TT (10 mM Tris-HCl pH 8 and 0.05% Tween-20) before on-bead library preparation using the NEBNext Ultra II DNA Library Prep Kit (E7370L, NEB) as previously described73. ChIP–seq libraries were multiplexed for paired-end (2 × 50 bp) sequencing on an Illumina NextSeq 2000 instrument.

　　Reads were trimmed to remove adapters and low-quality sequences and aligned to the hg38 and mm10 reference genome assemblies with bwa74 (v0.7.17-r1188) before filtering to remove duplicates and low-quality alignments including problematic genomic regions75 using the nf-core/ChIP–seq pipeline76 (v2.0.0; https://doi.org/10.5281/zenodo.3240506) with default parameters. Normalization to mouse spike-in chromatin was performed by scaling counts to the quotient of the ratios of human:mouse ChIP reads and human:mouse input reads as previously described77. CXXC1 peaks for visualization were identified using bam files from all AAVS1-knockout donors for MACS2 (v2.2.6)78 callpeak -q 0.05 with input samples used to define the background. High-confidence MED12 peaks were identified using bam files from all AAVS1-knockout donors for MACS2 callpeak -q 0.05 with MED12-knockout samples used to define the background (Supplementary Table 11). Utilization of high-confidence peaks generated from knockout controls reduced potential false-positive signals from the ChIP samples, providing a more rigorous assessment of MED12 binding79,80. ChIP–seq blacklist regions were removed from CXXC1 and MED12 peaks before analysis.

　　The polymerase pausing index was calculated as previously described33 as (TSS coverage/TSS length)/(gene body coverage/gene body length). Gencode v43 gene structures were selected for APRIS genes and filtered to include only genes expressed in Teff bulk RNA-seq data (defined from AAVS1 Teff RNA-seq base mean > 10). The TSS region of each gene was defined as 200 bp upstream and downstream of the TSS. The gene body was defined as the region 400 bp downstream from the TSS plus 400 bp past the final exon of the gene. Rtracklayer81 (v1.62.0) was used to import spike-in scaled RNA Pol II CTD bigwigs, and GenomicAlignments (v1.38.2) summarizeOverlaps() was used to determine the coverage within the defined gene regions.

　　Visualization of scaled tracks was performed with rtracklayer (v1.62.0) and ggplot2 (v3.5.1) with smoothing. APRIS gene structure was used for gene annotation with gggenes (v0.5.0). CD4+ Treg STAT5A ChIP–seq data were accessed from ChIP Atlas82, SRX212432 and GSM1056923, and generated by Hoffmann et al.31. Deeptools (v3.5.5)83 was used to generate profile plots of ChIP–seq data using computeMatrix scale-regions -b 3000 --regionBodyLength 5000 -a 3000 –skipZeros with scaled bigwigs, and a bed file of all expressed genes (defined from AAVS1 Teff RNA-seq base mean >10）作为输入，然后是PlotProfile –pergroup。

　　MED12 CAR RNA-seq数据来自Freitas等人。从Gene表达式综合访问omnibus，使用下载器检索原始计数文件（gse174279_raw_counts_grch38.p13_ncbi.tsv.gz）。首先，使用DESEQ2（V1.32.0）来识别AAVS1敲除刺激和静止样品之间差异表达的基因。使用以下标准定义了顶部上调的基因：调整后的P< 0.01, log2 fold change >2和碱的平均值> 10。所得的797个基因用于生成激活的基因特征。用DESEQ2 VST生成Med12-Knockout和AAVS1敲除静止和刺激样品的归一化计数，并将其转换为使用SummarizedExperiment84（v1.22.0）的汇总实验（V1.22.0）。使用GSVA函数min.sz = 10，max.sz = 6000，kcdf ='poisson'，使用归一化计数矩阵和激活评分作为GSVA85（v1.40.1）的输入。使用GGPLOT2（v3.4.1）进行了所得基因得分的可视化，并使用RSTATIX（V0.7.2）生成了调整后的P值。

　　除IL-2的50 U ml-1外，使用滴定量的人CD3/CD28/CD2 T细胞激活剂（10990，Stemcell Technologies）进行激活诱导的细胞死亡测定。使用cellevent caspase-3/7绿色流式细胞仪测定试剂盒（C10427，Invitrogen）根据制造商的方案，进行了72小时的活动caspase-3/7染色。凋亡群体的门控在淋巴细胞门上进行，并定义为活性caspase-3/7阳性和Sytox核酸染色。使用Phycoerythrin抗人CD95（FAS）抗体（305608，Biolegend； Biolegend;稀释1:50）进行FAS染色。

　　在分离Teff细胞后的第12天，分离Treg细胞后的第8天，将细胞在无细胞因子培养基中的96孔板中以每个孔的2×105细胞的密度铺板。将细胞用免疫型人CD3/CD28/CD2 T细胞活化剂（10990，Stemcelll Technologies）重新刺激，并在24小时后收集上清液。将上清液存储在-80°C下，直到通过Luminex 200系统上的Luminex Xmap技术来处理Eve Technologies。在串行滴定以确定适当的稀释液后，以技术重复运行样品，Luminex 48 Plex Human Panel A用于TEFF细胞（稀释的1:20）和Treg细胞（稀释1：5）。TREG细胞（未稀释）也运行了多种物种TGF 3 PLEX面板。每个SGRNA和供体组合将夏娃平均进行技术重复，以确定蛋白质浓度。从分析中除去了一个以上样品以上样品的细胞因子，以排除低丰度蛋白（补充表12）。

　　分离供体匹配的TEFF细胞，并在-80°C下冷冻，直到测定前24小时，直到24小时。用10 u ml-1 IL-2用2×106个细胞解冻Teff细胞，并以2×106细胞培养过夜。在测定当天，使用1：2,000的染料稀释，根据制造商的方案对TEFF细胞进行计数并用CellTrace Violet（C34557，Invitrogen）染色。在96孔圆形底板中，将测定板与1×105的Teff细胞组装在一起，其滴定量的Treg细胞范围为1：1至8：1 TEFF细胞：Treg细胞。每条条件还包括1×105 Treg细胞和5×104 Teff细胞（1：2 Teff细胞：Treg细胞），以及静止和刺激的Treg细胞和Teff细胞作为对照组。根据制造商的建议，制备了Treg抑制检查员（130-092-909，Miltenyi Biotec）iMac颗粒并根据制造商的建议添加到适当的井中。对四个捐赠者进行了一式三份的技术一式三份测定，并在37°C下孵育96小时。读数时，用Alexa Fluor 647抗人IL-2Rα（302618，Biolegend），BV711抗人类CD4（344648，Biolegend）和Ghost Dye Red 780（13-0865-T500，Tonbo）染色细胞。

　　在FlowJo（v10.8.1）中，对流量数据进行分析，并选择门控以选择淋巴细胞，单细胞，活细胞（幽灵染料负），CD4+ T细胞和Teff细胞（CellTrace+ CD25LOW）。然后使用不刺激的仅TEFF的对照（CellTrace Violet High Peak）为每个供体设置一个栅极，以建立增殖性TEFF计数。还通过选择非淋巴细胞，使用正向散射区域（FSC-A）和Ghost Dye选择非淋巴细胞，珠子，为iMacs珠子设置了门。然后使用公式（增殖性TEFF细胞计数×输入珠计数）/（珠）建立了绝对增殖的TEFF细胞计数，该公式可以调整刺激和收集异常的变化。计算抑制百分比为（100 - （绝对增殖的TEFF细胞计数/仅刺激响应者条件的TEFF细胞计数））×100。技术复制收集板的中位数用于计算抑制百分比和每个捐赠者的绝对增殖TEFF细胞计数。

　　有关研究设计的更多信息可在与本文有关的自然投资组合报告摘要中获得。