henhaobio
Sign in

CellGeneticsPeer-reviewed journal article

Genome-scale perturb-seq in primary human CD4+ T cells maps context-specific regulators of T cell programs and human immune traits

Ronghui ZhuiD, Emma DanniD, Jun YaniD, Justine Reyes Retana, Ryunosuke GotoiD, Reese C Guitche, Lillian Brixi, Mineto OtaiD, Austin HartmaniD, Theodore L. RothiD, Ansuman T. SatpathyiD, Jonathan K. PritchardiD, Alexander MarsoniD

PDFDOIPreprintCodeData
In-depth full-text digest available · Jump to digest ↓

Abstract

Gene regulatory networks encode the fundamental logic of cellular functions, but systematic network mapping remains challenging, especially in cell states relevant to human biology and disease. Here, we perturbed all expressed genes across 22 million primary human CD4+ T cells from four donors and developed a probe-based perturb-seq platform to measure the transcriptome effects in cells at rest and after stimulation. These data allow us to map genes that regulate known and novel pathways, including novel regulators of cytokine production. Importantly, active regulators and the gene programs they control change dramatically across stimulation conditions. Perturbation signatures enabled us to model T cell states observed in population-scale transcriptomic atlases, nominating regulators of Th1 and Th2 polarization and of age-related T cell phenotypes. Finally, we leveraged perturb-seq to implicate context-specific gene regulatory pathways in autoimmune disease risk. Our data provide a foundational resource to decode human immune function and genetic variation and for new approaches to study gene regulatory networks.

In-depth summary

Claude Sonnet 4.6 · xhigh effort · 8 Nov 2026 · from the full text of version 1 (24 Dec 2025) · CC BY 4.0

Genome-scale CRISPRi perturb-seq in 22 million primary human CD4+ T cells maps stimulation-specific regulators and links perturbation signatures to T cell states and autoimmune genetics.

Genome-scale perturb-seq had been confined to immortalized cell lines, while primary-cell work used targeted libraries or arrayed formats that bias discovery toward known biology. This work supplies a multi-donor, multi-stimulation genome-scale map in primary CD4+ T cells.

Key results

  1. ShownA probe-based 10x Flex perturb-seq screen delivered genome-scale CRISPRi coverage in primary CD4+ T cells, with roughly double the per-perturbation cell coverage of recent cell-line datasets at comparable depth.

    33.4M cells passed QC; 21.99M (65.8%) carried a single guide; 23,573 of 26,504 guides seen in at least 50 cells; mean 575 cells per perturbed gene per condition; 73% of 24,174 detected guides significantly reduced target expression vs non-targeting guides. · A scalable platform for perturb-seq in primary…; Trans effects of genetic perturbations

  2. ShownPseudobulk DESeq2 yielded a large catalogue of trans-effects that replicates published arrayed knockout and FACS-based CRISPRi screens, and is only partly shared with K562 cells.

    11,527 genes testable; 7,807 (67%) affected at least 3 genes (FDR < 10%); 2,035,311 significant regulator-to-gene effects; 27 of 32 targets correlated with arrayed KO screens; about 8% of trans-effects across 3,081 shared perturbations also seen in K562; mean Pearson R = 0.32 among 1880 perturbations with measurable K562 effects. · Trans effects of genetic perturbations; Validation of trans-effects with complementary…; Comparison with K562 genome-wide perturb-seq

  3. ShownPerturb-seq nominated cytokine regulators, including new IL10 and IL21 regulators, with NFKB2 and KDM1A acting in opposite directions on the two cytokines; arrayed knockdowns largely confirmed them.

    1,556 perturbations affected at least one of 30 cytokine genes (FDR < 1%); arrayed knockdown of 9 regulators: 11 of 12 tested interactions matched direction, 8 significant; NFKB2 effect on IL10 protein did not mirror its transcript effect. · Identifying regulators of immune cytokines; Arrayed validation of IL10/IL21 regulation

  4. ShownUnsupervised clustering grouped regulators into condition-specific and shared-complex modules, and autoimmune-associated genes were enriched in clusters active after stimulation but not in Rest-only clusters.

    3341 strong perturbations from 1860 regulators gave 111 clusters; about one third had coherent effects in only one or two conditions; Open Targets autoimmune genes enriched in 33 of 77 tested clusters (regulators or downstream genes). · Functional clustering of T cell regulators; Interpreting natural genetic variant effects on…; Clustering workflow

  5. SuggestedRegressing natural cell-state signatures on perturbation z-scores reconstructed Th2/Th1 and CD4+ T cell aging signatures on held-out genes and nominated regulators such as FBXO32, mTORC1 components and SUMOylation enzymes.

    Mean cross-validation R = 0.39 (Th2/Th1 training cohort) and 0.27 (independent validation cohort), with CD4+ T cell models outperforming K562-trained ones in the validation cohort; Rest perturbations predicted the OneK1K aging signature (782 discovery donors) with cross-validation R = 0.366. · Predicting regulators of human T cell states…; Predicting regulators of CD4+ T cell states in…; Th1/Th2 polarization bulk RNA-seq data…; Aging scRNA-seq data processing and analysis

Methods, data and reproducibility

Pooled CRISPRi (2 gRNAs per gene) in primary CD4+ T cells from four donors, read out by probe-based 10x Flex scRNA-seq at Rest, Stim8hr and Stim48hr; pseudobulk DESeq2 per condition, consensus Leiden/HDBSCAN clustering, and elastic-net regression of cell-state signatures on perturbation z-scores.

  • Genome-scale CRISPRi perturb-seq of primary CD4+ T cells, 3 conditions (Rest, Stim8hr, Stim48hr), 12 pools · human · n = 22 million cells; 4 donors; 12,748 targeted genes · GSE314342 / SRP643211
  • Published bulk RNA-seq of sorted Th1 and Th2 cells (discovery and replication cohorts) · human · E-GEAD-397; GSE149090
  • OneK1K scRNA-seq, CD4+ T cell subsets, for age-associated signature · human · n = 782 discovery donors; 199 replication donors · CellxGene collection dde06e0f-ab3b-46be-96a2-a8082383c4a1
  • K562 genome-wide perturb-seq from scPerturb, compared with CD4+ T cell effects · human (K562 cell line) · n = 3081 perturbed genes compared
  • UK Biobank rare loss-of-function burden statistics for lymphocyte counts · human · n = 454,787 participants
  • Published arrayed knockout RNA-seq and FACS-based CRISPRi screens used for validation · human · n = 27 of 32 targets correlated (Methods: 28 genes); 4 FACS target genes
  • Arrayed CRISPRi validation of IL10/IL21 regulators (bulk RNA-seq and flow cytometry) · human · n = 9 regulators; 12 tested interactions
  • Open Targets genetic-association genes for autoimmune and control diseases · human · n = 14 autoimmune conditions; 3 non-immune control diseases
Methods & tools
10x Genomics Flex probe-based scRNA-seq, CellRanger, crispat (Poisson-Gaussian guide assignment), scanpy, scVI, DESeq2 (Python implementation), MASH (mashr), Leiden + HDBSCAN consensus clustering, Elastic net with truncated SVD (scikit-learn), gseapy, Open Targets Platform, GeneBayes
Software & environment
CellRanger with Chromium Human Transcriptome Probe Set v1.1.0 plus two custom puromycin-resistance probes; regulator-prediction model packaged as a stand-alone Python class; Human Protein Atlas v25.0; Open Targets API v4.
Reproducibility gaps
Raw sequencing data and CellRanger outputs are stated as forthcoming via SRA/GEO (SRP643211 / GSE314342); the genome-scale gRNA library is described as pending release on Addgene.
Limitations (5)
  • Only 2 gRNAs per target: off-target guide effects are not fully assessed, and donor-specific biology is not yet separated from batch effects, which would need more guides per gene and further work. · stated by the authors
  • Pseudobulk aggregation improves robustness but may hide cell-to-cell heterogeneity in perturbation responses. · stated by the authors
  • All screens used a non-polarized culture condition, so regulatory rewiring under polarizing cytokines is unmapped. · stated by the authors
  • Knockdown was significant for only 73% of guides, and proximal-TSS co-knockdown was seen for 60% of 1,782 genes with nearby TSSs, so some trans-effects may reflect off-target or undetected on-target effects. · noted in this summary
  • Regulators nominated by signature regression (e.g. FBXO32, SUMOylation enzymes, mTORC1 components) were not tested experimentally; mTORC1 effect directions ran opposite to expectation. · noted in this summary

Read this if you want a genome-scale primary T cell perturb-seq resource and a regression framework linking perturbation effects to cell-state signatures and genetics-derived gene sets.

Citations
24
FWCI
49.25
Field percentile
top 1%

Reproducibility ledger

Nobody has reported trying to reproduce this paper yet. Did you replicate an experiment or run the code? Your report saves the next lab a week.

Sign in to file a reproducibility report.

Related papers

Similarity Map
SelectedSame fieldCross-field
THISGate '19Freimer '22Moonen '26Mihai '25Wang '26Weinstock '24
74% semantic matchGenomics
Rachel E. Gate, Min Cheol Kim, Andrew C. Lu+5 more·bioRxiv·Jun 21, 2019· 8 citations