Large-Scale CRISPR Transcription Factor Knockout Screens in CAR-T Cells: Protocols for Identifying Exhaustion Regulators
How SLICE-based CRISPR screens identify transcription factor regulators of T cell exhaustion in CAR-T cells using TFome libraries and multi-omic readouts.
Advanced Experimental Methods
Large-Scale CRISPR Transcription Factor Knockout Screens in CAR-T Cells: Protocols for Identifying Exhaustion Regulators
CAR-T cell therapy achieves remission rates exceeding 80% in B-cell acute lymphoblastic leukaemia but falters against solid tumours, largely because T cells become exhausted under persistent antigen pressure. Large-scale CRISPR screens using the SLICE platform and targeted TFome libraries now enable systematic identification of the transcription factors that govern this exhaustion programme, from TOX-driven epigenetic fixedness to FOXO1-mediated stemness maintenance.
What CRISPR transcription factor screens reveal about CAR-T exhaustion
Large-scale CRISPR transcription factor knockout screens in CAR-T cells systematically delete individual TF genes across thousands of primary human T cells in parallel, then subject the pooled population to chronic antigen stimulation to identify which knockouts resist or accelerate exhaustion. The SLICE platform (PMID: 30449619) solves the primary technical barrier: efficient gene editing in non-dividing primary T cells without compromising viability. By combining genome-wide or TFome-targeted sgRNA libraries with in vitro or in vivo exhaustion models, these screens map the transcriptional regulome of T cell dysfunction at a resolution previously impossible with candidate-gene approaches. The output is a ranked list of TFs whose loss-of-function either enhances or impairs T cell fitness under persistent antigen pressure, directly informing the engineering of next-generation CAR-T products.
Why unbiased CRISPR screens surpass candidate-gene approaches for T cell exhaustion
Before pooled CRISPR screens, the identification of exhaustion regulators relied on hypothesis-driven single-gene studies or transcriptomic correlation. This approach was inherently biased toward known biology: researchers tested TOX because it was differentially expressed, NFAT because it was a known activation factor. The problem was that many regulators, particularly metabolic enzymes (PRODH2), chromatin remodellers (BATF3), and bridging transcription factors (c-Jun), would never be prioritised by expression profiling alone (PMID: 35276062, PMID: 37945901, PMID: 31802004).
Unbiased screens change the logic from "which gene is differentially expressed?" to "which gene, when deleted, changes T cell fate?" This functional framing captures regulators that act post-translationally, that operate at low expression levels, or that function redundantly with other family members but become rate-limiting under chronic stimulation. The transition from loss-of-function (CRISPRko) to gain-of-function (CRISPRa) screens has further expanded the discovery space, identifying "functional boosters" that cannot be found by knockout alone (PMID: 35276062).
The practical impact is direct: FOXO1 overexpression, identified through these screens, increases the proportion of CD45RA+/CD62L+ stem-like CAR-T cells and correlates with peak clinical expansion and overall survival in CLL patients (PMID: 38600391). This is the kind of translational link that candidate-gene studies rarely achieve at scale.
Technical workflow: from sgRNA library to functional hit list
SLICE delivery in primary human T cells
T cells are activated at 1 million cells/mL using plate-bound anti-CD3 (10 microg/mL) and anti-CD28 (5 microg/mL) with 50 U/mL IL-2. At 24 hours, the sgRNA-encoding lentiviral library is added at 1:250 v/v. At 48 hours, recombinant Cas9 protein is electroporated (Lonza P3 buffer, 20 million cells per 100 microL, code EH115). Cells recover at 37 degrees C for 20 minutes, then expand in X-Vivo 15 supplemented with 5% FCS, 50 mM 2-mercaptoethanol, and 10 mM N-acetyl L-cysteine at 1 million cells/mL, split every 2 days with fresh IL-2. A 10-day rest follows before re-stimulation (PMID: 30449619).
Library options: genome-wide versus TFome
Genome-wide screens use the Brunello library (~77,441 sgRNAs targeting over 19,000 genes), requiring large cell numbers to maintain 300 to 500x coverage per sgRNA. Focused TFome libraries target approximately 1,612 human TF genes with 4 sgRNAs each, reducing cell requirements by an order of magnitude while retaining comprehensive coverage of the transcriptional regulome. Custom CRISPRi/a libraries can tile 1,000-bp windows around transcription start sites for epigenetic modifier screens (PMID: 37945901, PMID: 30449619).
In vitro exhaustion induction
Two primary protocols are used. The peptide/DC model co-cultures naive P14 CD8+ T cells with peptide-pulsed dendritic cells, re-administering cognate peptide (1 nM Db GP33-41) and IL-2 every 2 days for 7 to 10 days (PMID: 37595022). The continuous antigen exposure (CAE) model repeatedly stimulates CAR-T cells with tumour cells (e.g., AsPC-1 for mesothelin-directed CARs) at defined effector-to-target ratios over 20 to 35 days without tumour clearance (PMID: 34861191). In vitro models achieve a 50-fold reduction in mice compared to in vivo screens but underrepresent the TOX/NFAT axis and hypoxic signaling.
Multi-omic readout coupling
Pooled CRISPR delivery is coupled with single-cell RNA-seq or scATAC-seq using platforms like Perturb-seq to characterise the transcriptional and epigenetic programmes controlled by each hit gene at single-cell resolution. Surface phenotyping by FACS (CCR7, CD62L, IL-7Ralpha for memory; PD-1, TIM-3, LAG-3 for exhaustion) and CFSE dilution for proliferation provide complementary functional readouts. CD107a surface expression quantifies degranulation after antigen encounter (PMID: 37945901, PMID: 31375813, PMID: 30449619).
Where CRISPR exhaustion screens have identified actionable targets
B-cell haematological malignancies
CD19-directed CAR-T cells achieve remission rates exceeding 80% in B-ALL (PMID: 37569693), but relapse driven by T cell dysfunction remains a major limitation. CRISPR screens identified FOXO1 as a regulator of the stem-like memory state: overexpression of wild-type FOXO1 in human CAR-T cells increases the proportion of CD45RA+/CD62L+ stem-like cells, and the FOXO1 regulon is significantly associated with peak CAR-T expansion and overall survival in CLL patients (PMID: 38600391, PMID: 38600376).
PRODH2 (proline dehydrogenase 2), identified via a gain-of-function CRISPRa screen, enhances proliferation and anti-tumour activity of human T cells in B cell leukaemia models, establishing metabolic engineering as a distinct axis from transcriptional reprogramming (PMID: 35276062).
Solid tumours: pancreatic and breast cancer
Solid tumours present the most severe exhaustion challenge due to persistent antigen, hypoxia, and immunosuppressive stromal signaling. In the CAE model using AsPC-1 pancreatic cancer cells, mesothelin-directed CAR-T cells progressively lose TNF-alpha and IL-2 production while retaining PMA/ionomycin-responsive cytokine machinery, indicating proximal signaling failure rather than global dysfunction (PMID: 34861191).
BATF3 overexpression, identified through a TFome CRISPRi/a screen, enhances the potency of HER2-targeted CAR-T cells in orthotopic breast cancer models by countering the heterochromatinisation of the TCF7 locus, maintaining the progenitor exhausted population required for durable anti-tumour responses (PMID: 37945901).
Mesothelioma and checkpoint combination
Mesothelin-directed CAR-T cells combined with pembrolizumab achieved a 1-year overall survival of 83% in malignant pleural mesothelioma (PMID: 36216928). The rationale for this combination emerged directly from CRISPR screen biology: TCF7-positive progenitor exhausted cells, identified as the population responsive to PD-1 blockade, can be expanded by engineering approaches that prevent terminal differentiation. Understanding which TFs maintain this progenitor state (TCF7, FOXO1) versus those that drive terminal commitment (TOX) guides the selection of genetic modifications compatible with checkpoint co-therapy.
Chronic viral infection models
LCMV Clone 13 chronic infection remains the gold-standard in vivo exhaustion model and provides the biological framework against which all in vitro models are benchmarked. Screens in this system established the hierarchical loss of cytokine production (IL-2 first, then TNF, then IFN-gamma) and identified TOX as essential for initiating the exhaustion epigenetic programme (PMID: 19043418, PMID: 31207603). The calcineurin-independent, TOX-driven molecular programme enforces durable suppression that distinguishes true in vivo exhaustion from the milder "functional adaptation" seen in vitro (PMID: 37595022).
Validating CRISPR screen hits from pooled rank to clinical candidate
Arrayed single-gene validation
Top-ranked hits from pooled screens are individually validated using arrayed CRISPR editing or overexpression constructs in primary human CAR-T cells. This confirms that the phenotype is attributable to the specific gene rather than library-level artefacts or passenger mutations. Validation assays recapitulate the original readout (proliferation, surface phenotype, cytokine production) and extend to additional readouts not captured in the pooled screen, such as serial tumour killing assays and in vivo tumour challenge (PMID: 38600391, PMID: 37945901).
In vivo tumour model confirmation
Hits identified in vitro require confirmation in orthotopic or subcutaneous tumour models to account for the microenvironmental complexity absent from cell culture. BATF3-overexpressing HER2 CAR-T cells were validated in orthotopic breast cancer models (PMID: 37945901). FOXO1-engineered CAR-T cells were tested in multiple tumour types to confirm that the stem-like phenotype persists under in vivo antigen pressure (PMID: 38600376). In vivo models also reveal off-target effects: the risk of secondary malignancies from constitutively active transcription factor overexpression is only assessable in long-term animal studies.
Epigenetic and transcriptomic confirmation
Because exhaustion is epigenetically encoded (PMID: 38226974, PMID: 28648661), validation must confirm that hits alter the chromatin landscape, not merely the transcriptome at a single time point. scATAC-seq on validated clones confirms whether knockout or overexpression opens or closes specific loci (e.g., TCF7 promoter accessibility in BATF3-overexpressing cells). This distinguishes genuine regulators of the epigenetic programme from genes that transiently modify gene expression without altering the underlying chromatin state.
Clinical correlate analysis
The most compelling validation links screen hits to patient outcomes. The FOXO1 regulon signature, derived from CRISPR screen biology, was correlated with peak CAR-T expansion and overall survival in CLL patients, providing clinical-grade evidence that the screen-identified biology is relevant in humans (PMID: 38600391). Similarly, pre-infusion product phenotyping for exhaustion markers (PD-1, TIM-3) correlates with clinical failure, validating the screening endpoint itself (PMID: 29713085).
Evidence quality and methodological limitations
The evidence base for CRISPR exhaustion screens is strong and rapidly maturing. Core exhaustion hallmarks (upregulation of PD-1, TIM-3, LAG-3; hierarchical loss of polyfunctionality) replicate across all published in vitro and in vivo models. Key hits like FOXO1, TOX, and TCF7 have been independently validated by multiple groups using distinct library designs, delivery platforms, and tumour models. The clinical correlate analysis linking FOXO1 regulon activity to patient survival in CLL represents the highest tier of translational evidence available for a screen-derived target. The field draws from over 33 primary studies with a median publication year of 2023 to 2024, indicating active and recent investigation.
In vitro exhaustion models underrepresent the TOX/NFAT axis, meaning that targets identified exclusively in cell culture may address "functional adaptation" rather than the more durable "terminal exhaustion" observed in patients (PMID: 37595022, PMID: 40236693). In vivo screens are resource-intensive, requiring pooling of large mouse cohorts to achieve sufficient library representation, and recover small cell numbers that limit statistical power. Donor-to-donor variation introduces significant heterogeneity: exhaustion markers in pre-infusion products may reflect activation-driven phenotypes rather than baseline patient traits (PMID: 29713085, PMID: 38226974). There is a strong recency bias toward gain-of-function screens, and long-term safety data for constitutive TF overexpression (risk of secondary malignancies, malignant transformation of "super-active" T cells) are only beginning to emerge (PMID: 38195751). Finally, mouse thymic and metabolic biology differs from human biology, and screen hits validated only in murine models require human-specific confirmation before clinical translation.
Large-scale CRISPR transcription factor screens have transformed the study of CAR-T cell exhaustion from descriptive transcriptomics into functional regulome mapping. The SLICE platform and TFome libraries provide the technical infrastructure; in vitro chronic stimulation and in vivo LCMV/tumour models provide the biological context; and multi-omic readouts coupled with clinical correlate analysis close the loop to patient outcomes. The field is shifting from "blocking brakes" (PD-1 inhibitors) to "rewiring engines" (FOXO1, BATF3, c-Jun engineering), with the recognition that exhaustion is an epigenetically stable programme that requires chromatin-level intervention. As gain-of-function screens expand the targetable space and long-term safety data mature, CRISPR-informed CAR-T engineering is positioned to extend the successes of B-cell haematological malignancies into the more challenging landscape of solid tumours.
Frequently asked questions
What is the SLICE platform for CRISPR screens in T cells?
SLICE (sgRNA Lentiviral Infection with Cas9 protein Electroporation) is a two-step delivery platform for primary human T cells. T cells are first infected with an sgRNA-encoding lentiviral library, then electroporated with recombinant Cas9 protein 24 hours later. This staggered approach maintains viability and proliferative capacity during large-scale pooled screens.
What is a TFome CRISPR library?
A TFome library is a targeted CRISPR library designed to perturb all annotated transcription factors in the human genome. A typical TFome library targets approximately 1,612 human TF genes with 4 sgRNAs each, providing focused coverage of the transcriptional regulome at lower cell requirements than genome-wide libraries.
How does in vitro chronic stimulation model T cell exhaustion?
In vitro chronic stimulation recapitulates exhaustion by repeatedly exposing T cells to cognate antigen. Common approaches include peptide-pulsed dendritic cell co-culture with re-stimulation every 2 days for 7 to 10 days, or continuous antigen exposure where CAR-T cells are repeatedly stimulated with tumour cells at defined effector-to-target ratios over 20 to 35 days.
What transcription factors have been identified as CAR-T exhaustion regulators?
Key regulators include TOX (initiates the exhaustion epigenetic programme), TCF7/TCF1 (marks the progenitor exhausted population required for checkpoint inhibitor response), FOXO1 (promotes stem-like memory differentiation and correlates with clinical expansion), BATF3 (counters heterochromatinisation of TCF7), and c-Jun (displaces immunoregulatory AP-1/IRF complexes to restore effector function).
Why do in vitro models underrepresent TOX/NFAT signaling?
In vitro chronic stimulation models tend to enrich for terminal exhaustion gene signatures rather than the TCF1-positive progenitor populations found in vivo. The TOX/NFAT axis, which drives calcineurin-independent epigenetic remodelling and enforces durable cytokine suppression, requires the full complexity of in vivo tissue microenvironments including hypoxia and metabolic competition.
What functional readouts identify exhaustion regulators in CRISPR screens?
Readouts include CFSE dilution for proliferation, surface phenotyping for memory markers (CCR7, CD62L, IL-7Ralpha) and exhaustion markers (PD-1, TIM-3, LAG-3), CD107a for degranulation, CAR surface expression loss for antigen internalisation, and coupling with single-cell RNA-seq or scATAC-seq for high-dimensional transcriptional and epigenetic characterisation.
How does BioSkepsis generate these methods reviews?
BioSkepsis synthesises PubMed-indexed literature into structured reviews with citation-level verification. Every citation undergoes three independent checks to confirm it directly supports the associated claim. Unverified citations are flagged and excluded from the main synthesis.
Explore Experimental Methods with Citation-Verified PubMed Synthesis
BioSkepsis builds structured literature reviews with three-stage citation verification, turning PubMed evidence into actionable experimental protocols and methods summaries.
Start freeSources and further reading
- SLICE platform for genome-wide CRISPR screens in primary human T cells (PMID: 30449619)
- In vitro chronic stimulation model for T cell exhaustion benchmarking (PMID: 37595022)
- Continuous antigen exposure model for CAR-T cell dysfunction (PMID: 34861191)
- TFome CRISPRi/a screens identifying BATF3 in CAR-T cells (PMID: 37945901)
- FOXO1 overexpression and stem-like CAR-T cell phenotype (PMID: 38600391)
- FOXO1 regulon characterisation by RNA-seq and ATAC-seq (PMID: 38600376)
- Gain-of-function CRISPRa screens and PRODH2 identification (PMID: 35276062)
- Hierarchical cytokine loss in chronic LCMV infection (PMID: 19043418)
- TOX as initiator of the exhaustion epigenetic programme (PMID: 31207603)
- c-Jun overexpression displacing AP-1/IRF complexes (PMID: 31802004)
- BATF as bridge between activation and dysfunction (PMID: 34282330)
- TCF7 progenitor populations and checkpoint response (PMID: 36216928)
- CD19 CAR-T remission rates in B-ALL (PMID: 37569693)
- Epigenetic encoding of T cell exhaustion (PMID: 38226974, PMID: 28648661)
- CAR co-stimulatory domain effects on exhaustion tempo (PMID: 25939063, PMID: 33824268)
- NFAT in T cell activation and exhaustion (PMID: 25680272)
- Manufacturing heterogeneity and pre-infusion phenotyping (PMID: 29713085)
- Secondary malignancy risk in engineered T cells (PMID: 38195751)
- Terminal exhaustion versus functional adaptation (PMID: 40236693)
- Single-cell multi-omic profiling coupled with CRISPR perturbation (PMID: 31375813)
- In vivo exhaustion models: LCMV and orthotopic tumour transplantation (PMID: 34031411)