A two-cell system for genetic analysis of inhibitory and activatory signaling in T cells

To study activatory T-cell signaling in the context of both co-activatory and co-inhibitory signals from CD28 and PD-1, we first established a range of signaling settings using Jurkat T cells in the two-cell assay system described by Leitner et al (2010). In this system, the Jurkat reporter cell lines, with or without PD-1, henceforth referred to as Jurkat/PD-1 and Jurkat/control, are stimulated with a previously described TCS, a murine thymoma cell line, BW5417, that co-expresses a membrane-bound single-chain Fv (scFv) variant of the anti-CD3ε antibody OKT3 and high levels of PD-L1 and CD86 (TCS/CD86/PD-L1) (Leitner et al, 2010). The Jurkat cells are engineered to express eGFP under the control of the NFκB transcription factor, allowing eGFP levels to be used as a measure of activation (see schematics in Fig 1A). We sought to study inhibitory signaling, in a single cell-line, in the context of low activatory signaling, in which only the TCR/CD3 complex is engaged (signal 1), and high activatory signaling, wherein both TCR and CD28 are engaged (signal 1 + 2). To achieve this, we co-cultured Jurkat cells with stimulator cell line TCS/CD86/PD-L1 that were pretreated with (i) anti-CD86, (ii) anti-PD-L1 or (iii) both anti-CD86 and anti-PD-L1 blocking antibodies, creating stimulator cells equivalent to TCS/PD-L1, TCS/CD86, and TCS only, respectively. Cultures of Jurkat/control cells to which no TCS were added were used as baseline controls. In the signal 1 condition, Jurkat/control cells stimulated with TCS produced eGFP expression levels equivalent to those stimulated with TCS/PD-L1 cells, whereas Jurkat/PD-1 cells stimulated with TCS had significantly higher levels of eGFP expression compared with those stimulated with TCS/PD-L1 cells, demonstrating the inhibitory effects of PD-1 on the activatory pathway upon PD-L1 engagement (Fig 1B). In the signal 1 + 2 context (TCS/CD86), all the cells produced more eGFP, but PD-1 engagement (TCS/CD86/PD-L1) resulted in significantly less signaling (Fig 1B).

To enable genetic manipulation of the Jurkat cells used in the two-cell assay, we then generated a high-efficiency Cas9 version of both the Jurkat/control and Jurkat/PD-1 cell lines, using lentiviral transduction (Fig S1). We then transduced cells with single gRNAs targeting NFKB1 and RELA and observed a decrease in the expression of eGFP when these cell lines were stimulated with TCS compared with control cells targeted with empty gRNA. This confirmed that eGFP production relied on the NFκB signaling pathway and the two-cell system was amenable to manipulation with CRISPR-based perturbations using single gRNAs (Fig 1C). Next, given the known roles of SHP1 and SHP2 phosphatases in regulating inhibitory signaling, we targeted PTPN6 (encoding SHP1) and PTPN11 (encoding SHP2) using two separate gRNAs and stimulated the PD-1 expressing reporter lines in Signal 1 conditions (TCS and TCS/PD-L1). Targeting PTPN11 resulted in an increase in signaling when PD-1 was engaged, whereas targeting PTPN6 had no effect, reconfirming the role of SHP2 in mediating the inhibitory effects of PD-1 (Fig 1D). We then tested the two PTPN11-targeting guides in activation under signal 1 + 2 conditions (TCS/CD86 and TCS/CD86/PD-L1) and observed that targeting cells with the better-performing guide (PTPN11-g2) rescued signaling (Fig 1E). Altogether, these observations are consistent with recent work showing that PD-1 recruits SHP2 but not SHP1, and indicate that PD-1 inhibition is effective even in the absence of co-stimulatory signaling by CD28, presumably by acting directly on the TCR(/CD3) signaling pathway (Xu et al, 2021).

Arrayed CRISPR-based screening of the role of SH2 domain-containing proteins in T-cell signaling

Having established the cell-based assay that allows signaling through the NFκB pathway and the effects of PD-1 on this pathway to be studied, we used the assay to study the role of SH2 domain-containing proteins in regulating protein tyrosine kinase-mediated signaling in T cells. We compiled a list of 64 SH2 domain-containing proteins that are expressed in Jurkat cells, which included SHP1 and SHP2, based on transcriptome analysis of Jurkat cells from Cell Model Passport (van der Meer et al, 2019, Table S1). We created a CRISPR-based KO arrayed library to target the remaining 62 SH2 domain-containing proteins with two gRNAs for each gene. We then used the two-cell assay to assess signaling outcomes in all five signaling contexts for all the “SH2-targeted” T cells (summarised in Fig 2A). For each stimulation condition, we then identified the targeted genes producing the most extreme changes in eGFP expression (Fig S2). To investigate how targeting of these genes affected signaling under different thresholds of activation and PD-1-mediated inhibitory conditions, we individually investigated each gene. We observed that CSK-, CBL-, and SLA-targeted cells exhibited increased background signaling in the absence of TCR/CD3 complex engagement and produced increased fold changes in eGFP fluorescence compared with the empty transduction control in all stimulation contexts (Fig 2B, upper panel). This confirmed the known role of CSK in phosphorylating, specifically, the LCK inhibitory tyrosine (Tyr505) to reduce the activity of the kinase. Loss of CSK expression allows LCK to remain active and phosphorylate the ITAMs of the TCR-associated CD3 chains, leading to activation without receptor ligation (Schoenborn et al, 2011). The ubiquitin ligase CBL is a well-known negative regulator of TCR signaling and its loss leads to reduced clearance of engaged TCR from the cell surface, thereby increasing TCR signals. SLA (also known as SLIP-1) is another known negative regulator of signaling by the TCR thought to act by linking ZAP-70 with CBL (Tang et al, 1999). Although stimulation of CBL- and CSK-targeted T-cells with TCS/PD-L1 cells under signal 1-type (TCS and TCS/PD-L1) conditions significantly increased activation levels compared with the non-targeted (“empty”) control, both mutants produced lower eGFP levels when PD-1 was engaged. In contrast, under signal 1 + 2 conditions (TCS/CD86 and TCS/CD86/PD-L1), the inhibitory effects of PD-1 were lost for all three targeted genes, indicating that activating signals can readily overwhelm even potent inhibitory signaling mediators (Fig 2B).

Next, we investigated the genes whose targeting led to reduced activation in all signaling contexts (Fig 2A), that is, NFKB1 (positive control), ZAP70, LCP2, HSH2D, VAV1, GRB2, GRAP and GRB10, SH2B2, ITK, PLCG1, LCK. In the heatmap distance clustering (Fig 2A), these genes clustered in two separate groups, prompting us to investigate whether this was because of the differences in gRNA efficiency, with one of the guides being more efficient, producing a smaller average effect for some of the targets. This was indeed the case. For all genes except LCK and PLCG1, a decrease in eGFP expression was observed under both signal 1 (TCS) and signal 1 + 2 (TCS/CD86) signaling contexts for at least one of the two gRNAs (Fig S3A and B). For the gRNAs with lower efficacy, we tested whether the signal 1 + 2 (TCS/CD86) conditions were masking the effects of the gRNAs by targeting representative genes (ZAP70, LCP2, GRB2, VAV1, LCK, PLCG1) using both gRNAs and reducing the degree of stimulation by decreasing the co-incubation time to 15 h and the Jurkat:TCS ratio from 3:1 to 10:1. Using this less-intense stimulating regimen, we observed significant decreases in eGFP expression in both signal 1 (TCS) and signal 1 + 2 (TCS/CD86) stimulation conditions for all gRNAs (Figs 2C and S3C). Nevertheless, LCK-targeted cells still had higher residual eGFP expression when stimulated under signal 1 + 2 (TCS/CD86) conditions, even though the gRNAs targeting LCK dramatically reduced the levels of LCK at the protein level (Figs 2C and S3C and D).

LCK is an essential SRC–kinase responsible for initiating TCR signaling (Smith-Garvin et al, 2009), therefore it was unexpected that Jurkat cells targeted with LCK-specific gRNAs would have residual eGFP expression when stimulated under low-intensity signal 1 + 2 (TCS/CD86) conditions, whereas the same targets produced very little residual eGFP expression under signal 1 (TCS) conditions. This implies that co-stimulatory signaling by CD28 is especially LCK-independent. In our original screen of the six SRC family kinase members present (BLK, FGR, FYN, HCK, SRC, and YES-1), only SRC targeting had a modest effect upon stimulation and only upon stimulation under signal 1 (TCS) conditions (Fig 2D). We therefore tested whether these kinases compensate for the absence of LCK in Jurkat T-cells. To test this, we targeted members of the SRC–kinase family in an LCK-targeted cell line and examined the responses of the doubly targeted cells to the lower intensity signal 1 + 2 (TCS/CD86) stimulation. Targeting each of these kinases in addition to LCK reduced eGFP expression, with the largest reduction seen with targeting HCK (Fig 2E). However, the effects were small, suggesting that all the kinases collectively compensate for the absence of LCK.

Overall, the arrayed screen offered a way to study the contributions of SH2 domain-containing proteins to signaling outcome under each of the five conditions tested. For many of the known positive and negative regulators, the assay produced the expected increases and decreases in eGFP expression. With respect to the identification of specific regulators of PD-1 inhibition, although we identified CBL, CSK, and SLA as targets whose KO influenced the potency of inhibitory signaling, we did not identify a new mediator of PD-1 signaling among the 64 SH2 domain-containing molecules.

Genome-wide screening of positive and negative regulators of T-cell signaling

The arrayed screen indicated that there are no undiscovered mediators of PD-1 inhibitory signaling among the 64 SH2 domain-containing proteins expressed in Jurkat T cells. A genome-scale screen could, in principle, identify new potential mediators of PD-1 signaling outside of this group, in an unbiased manner. Genome-scale loss of function screens to identify the regulators of proximal T-cell signaling have been performed previously both in Jurkat and human primary T cells and such screens have provided valuable insights into known and novel regulators of T-cell signaling (Shang et al, 2018; Shifrut et al, 2018). These screens have, however, been performed in non-cell–based stimulation settings. Given that the two-cell system allows stimulation screens incorporating the engagement of inhibitory ligands, we undertook a genome-wide screen to identify regulators of T-cell signaling while PD-1 is engaged by PD-L1.

Our initial targeting of SHP1/2 in the arrayed CRISPR screen revealed that overstimulating cells can mask the effects of perturbing inhibitory signaling mediators, implying that stimulation via the TCR/CD3 complex alone (signal 1) would likely allow more sensitive identification of mutants that rescue signaling from PD-1 inhibition. Hence, we undertook genome-wide screens in the context of signaling by the TCR, and inhibitory signaling triggered by PD-1. We first conducted a genome-wide KO screen by transducing Jurkat/PD-1 cells with a lentiviral library encoding ∼91,000 sgRNAs targeting ∼18,000 genes and stimulating the mutant cells using TCS/PD-L1 13 d post-transduction. We collected mutant cells expressing high or low eGFP using flow cytometry, amplified the gRNAs present in these sorted populations, and then compared the gRNA counts with those of the control unsorted population. This identified gRNAs that were enriched in the two sorted populations (the sorting strategy is set out in Fig S4; screen data is available in Table S2). Cells expressing low levels of eGFP were enriched in gRNAs targeting positive regulators of TCR signaling as expected (Fig 3A, left panel). Similar to the Jurkat T-cell screen undertaken by Shang et al (2018) the hits from this screen included gRNAs targeting genes encoding the components of the TCR–CD3 complex comprised of CD3-δ, CD3-ε, CD3-γ, and CD3-ζ (CD247) chains (the genome-wide library did not include gRNAs targeting TCRA and TCRB), the tyrosine kinases LCK, ZAP70, and ITK, and the LAT signaling complex, that is, LAT and LCP2 (Fig 3A, left panel). Enrichment analysis using the KEGG database revealed significant enrichment of immune-related signaling pathways including the TCR signaling pathway (Fig 3B). An interesting hit from this screen was BAP1, which is known to be essential for peripheral T-cell proliferation (Arenzana et al, 2018). There are reports that BAP1 regulates NFκB signaling in B lymphoblasts so, taken together, it is likely that BAP1 is involved in signaling leading to proliferation via the NFκB pathway in T cells (Takagi-Kimura et al, 2022).

We closely examined the genes that had false discovery rate (FDR) < 0.25 among the eGFP-high population of cells and categorised them according to cellular function (Table S3). We were interested especially in identifying signaling-related molecules that could mediate signaling by PD-1. Of the molecules involved in cellular signaling, the top hits included regulators of the NFκB-pathway (NFKBIA, NFKB2, CYLD, and TNFAIP3) and TNF receptor-associated factors including TRAF2, 3 (Fig 3A, right panel). Pathway enrichment analysis also revealed a significant enrichment of the genes controlling the NFκB pathway (Fig 3B). We also identified previously characterised negative regulators of T-cell signaling such as CBL and CSK. Notably, the signaling regulators PTPN11 (SHP2) and SH3KBP1, which has been shown to bind to CBL (Taipale et al, 2014), and PDCD1, the gene encoding PD-1 itself, were identified in the screen. The identification of PDCD1 and PTPN11 in this screen confirmed that this assay could identify mediators of PD-1-dependent inhibitory signaling. However, presumably because the consequent increases in activation after KO were smaller than those of general negative regulators, they were identified at a lower FDR. In general, the top negative regulator hits in this screen were mostly dominated by cellular factors whose targeting increased positive signaling to a very high level thereby reducing the power to detect mutants that more modestly enhance signaling, an effect we previously noted in the arrayed screen.

The genome-wide screen also revealed several hits in actin/cytoskeletal regulation pathways (Fig 3A, right panel). Among the top hits were ARPC4 and PFN1, which encode one of seven subunits of the human Arp2/3 protein complex and an actin binding protein, respectively. PFN1 is known to regulate NFκB pathway but also has been previously shown to be the only member of the profilin family to be expressed in primary human CD8⁺ T cells where it acts as a negative regulator of cytotoxicity (Schoppmeyer et al, 2017). To further investigate the role of ARPC4, we designed gRNAs targeting this gene and tested the levels of activation post stimulation with the different TCS in both Jurkat/control and Jurkat/PD-1 lines. Knocking out ARPC4 (Fig S5A) increased eGFP expression in both Jurkat/control and Jurkat/PD-1 in all stimulation contexts (Fig 3C). Together, these data suggest that ARPC4 enhances T-cell signaling and is not a specific regulator of the PD-1 pathway.

We then tested if targeting ARPC4 affected signaling responses in primary T cells. We targeted ARPC4 in CD8⁺ T cells and stimulated them using TCS/CD86 and measured the expression of two activation markers, CD69 and 4-1BB, after 24 h. We observed that targeting ARPC4 had no effect on the expression of CD69 but increased the expression of 4-1BB compared with parental CD8⁺ T cells reaffirming our previous findings in Jurkat cells (Fig S5B). We further tested the function of ARPC4 in primary T cells using an in vitro cancer cell-killing assay. Primary CD8 T cells transduced to express the 1G4 TCR were used in a killing assay with an mOrange-expressing A375 melanoma cell line, which expresses the tumour antigen NY-ESO-1 (Chen et al, 2005). The level of killing was quantified by tracking the “integrated red area intensity” (see Fig S5C). ARPC4-targeted cells, and CSK-targeted (negative regulator control), LCK-targeted (positive regulator control), and empty guide-targeted 1G4-expressing T cells were generated and used in the killing assays (see Fig S5D for ARPC4 KO efficiency). The ARPC4 KO cells maintained and even slightly increased the level of expression of CD25 post interaction with A375 cell lines compared with the empty-targeted 1G4 TCR-expressing T cells (Fig 3D) but surprisingly had significantly reduced cytotoxic ability (similar to the levels of LCK-targeted cells; Fig 3E and F). Cells with disrupted actin pathways, whereas presenting as hyperactive cells in terms of the expression of surface markers of activation, appear therefore to be profoundly impaired in terms of their effector functions.

Cell culture

Previously generated Jurkat NFκB::eGFP cell lines and BW cell lines expressing an anti-human CD3 single-chain fragment anchored by fusion to human CD14 (TCS cells) were used in this study (Leitner et al, 2010; Jutz et al, 2017). Jurkat cells and TCS cells were both cultured in an RPMI 1640 medium (Gibco) supplemented with 10% (vol/vol) FBS (lot #08F6480K; Gibco), 1% (vol/vol) penicillin/streptomycin/neomycin (final concentrations 50 U/ml penicillin, 50 μg/ml streptomycin, 100 μg/ml neomycin; Sigma-Aldrich) and 1% Hepes–NaOH (final concentration 10 mM; Sigma-Aldrich). HEK-293T cells for lentiviral transfection were cultured in DMEM (Gibco), supplemented with 10% (vol/vol) FBS (lot #08F6480K; Gibco), 1% (vol/vol) penicillin/streptomycin/neomycin (final concentrations 50 U/ml penicillin, 50 μg/ml streptomycin, 100 μg/ml neomycin; Sigma-Aldrich), 1% (vol/vol) sodium pyruvate (final concentration 1 mM; Sigma-Aldrich), 1% (vol/vol) L-glutamine (final concentration 2 mM; Sigma-Aldrich) and 1% (vol/vol) Hepes–NaOH (final concentration 10 mM; Sigma-Aldrich).

Two-cell assays

The two-cell assay was performed as described previously (Jutz et al, 2017). Briefly, Jurkat cells were cultured in a 3:1 ratio with TCS cells expressing high levels of CD86 and PD-L1. CD86 and PD-L1 binding was blocked using 5 μg/ml Ultra-LEAF anti-human CD86 and PD-L1 antibodies (Clone #IT2.2 and #29E.2A3; BioLegend, respectively) to create different stimulation conditions. After 24 h, cells were harvested and TCS cells stained with PE-conjugated anti-mouse CD45 antibody (Clone #I3/2.3; BioLegend). Expression of eGFP reporter gene in Jurkat cells was measured by flow cytometry.

sgRNA design and cloning

sgRNA sequences used were the top two candidates from CRISPick (Broad Institute) for each target gene. sgRNA sequences, as oligos, (Sigma-Aldrich) were annealed and cloned into the BbsI site of the CRISPR gRNA expression vector pKLV2-U6gRNA5(BbsI)-PGKpuro2ABFP-W (#67974; Addgene) as described before (Sharma & Wright, 2020). The full list of gRNAs for all the targets used in this study are listed in Table S4.

Lentiviral preparation

Lentiviruses were prepared by transfecting HEK-293T cells with the gRNA expression vector. 500 ng of gRNA expression plasmid DNA was incubated with 250 ng psPAX2 (#12260; Addgene) and 150 ng psMD2.G (#12259; Addgene) and 3 μl of PLUS reagent (Invitrogen) in 250 μl OptiMEM (Gibco). After 5 min, 3 μl Lipofectamine LTX (Invitrogen) was added to the mixture. The mixture was added to confluent HEK-293T cells in a 12-well plate. After 3 h of incubation, the OptiMEM mixture was removed and replaced with DMEM (Gibco), supplemented as described above. Lentiviral supernatant was harvested and filtered at 0.45 μm after 48 h. Viral supernatants were stored at −80°. This protocol has also been described in detail before (Sharma & Wright, 2020).

Arrayed screening

Jurkat cells (1 × 10⁵ cells/well in 96-well plate) were transduced with the arrayed library lentiviruses followed by centrifugation at 800g for 90 min. After 2 d, transduced cells were selected by addition of puromycin (final concentration 2 μg/ml; Gibco). After 7–10 d of growth, TCS cells (5 × 10⁴/well in a 96-well plate) were added to Jurkat cells, with different conditions created as described above. After 24 h, cells were harvested and eGFP expression was measured by flow cytometry. Jurkat cells were gated in flow cytometry using their BFP expression.

Lentiviral transduction for generation of Cas9 lines

Cas9-expressing Jurkat/control and Jurkat/PD-1 lines were generated by lentiviral transduction of pKLV2-EF1a-Cas9Bsd-W construct. Cas9 activity in the cells was tested using a “BFP-GFP” system described before (Sharma et al, 2018). Briefly, cells were transduced with lentivirus encoding GFP, BFP, and a gRNA targeting GFP (pKLV2-U6gRNA5(gGFP)-PGKBFP2AGFP-W) or the same construct with an “empty” gRNA (pKLV2-U6gRNA5(Empty)-PGKBFP2AGFP-W) as a negative control. The activity of Cas9 in the cell lines was assessed by examining the ratio of BFP only to GFP-BFP–double-positive cells transduced by the two lentiviruses.

Genome-wide screens

Human Improved Genome-wide Knockout CRISPR Library v1 was a gift from Kosuke Yusa (#67989; Addgene) (Tzelepis et al, 2016). Genome-scale screens were performed as described in a detailed protocol before (Sharma & Wright, 2020). Detailed protocol with all PCR steps, reaction times, and volumes are described in this protocol. Briefly, a genome-scale “knockout” library of Jurkat/PD-1 cells was produced by transducing 80 million cells such that ∼30% of the total cell population was transduced. The transduced (BFP-positive) cells were harvested 3 d after transduction using a cell sorter (MA900; Sony sorter). Libraries with at least 1.5 × 10⁸ cells were selected and maintained in media containing 2 μg/ml puromycin to remove the non-transduced cells and at every passage. 13 d post-library generation, 150 million cells were activated with TCS/PD-L1 for 24 h. Two-way sorting was performed to collect the top 15% and bottom 15% of eGFP expressing cells. Genomic DNA was extracted using a commercial kit (Blood and Tissue mini kit) and gRNA was amplified using the isolated DNA using L1/U1 primers and Q5 Hot Start High-Fidelity 2 × Master Mix. The PCR-amplified guides were diluted to 40 pg/μl in EB and tagged with Illumina index primers using second round of PCR with using 200 pg template, PE 1.0 as forward primer, appropriate index tags as reverse primers, and KAPA HiFi HotStart ReadyMix polymerase. The PCR products were cleaned using SPRI beads (Agencourt AMPure XP beads), quantified using qbit and pooled to 4 nM concentration and sequenced using NextSeq platform. Single-end sequencing (19 bps) was performed with the custom sequencing primer, 5′-TCTTCCGATCTCTTGTGGAAAGGACGAAACACCG-3′.

Data analysis

The read count for each gRNA and gene level enrichment analysis was carried out using the MAGeCK statistical package (version, v0.5.5) by comparing the read counts from the sorted population with those from the control population. Additional parameters in MAGeCK were used to only include genes that had more than 3/5 guides enriched in each test (“min-number-goodsgrna 3”). Pathway analysis was also carried out using the EnrichR package with KEGG annotated pathways. All further analyses were carried out using R.

Generation of 1G4-T cells

PBMCs from healthy blood donors were activated with anti-CD3/anti CD28 beads for 2 d. The cells were then transduced with a lentiviral construct encoding the 1G4 construct on a retronectin plate. The 1G4 construct consisted of a short LNGFR fragment. Transduced 1G4-expressing CD8 cells were isolated using FACS sorting of LNGFR+/CD8+ population. Cells were proliferated for 2 wk after which they were electroporated using the RNP complex to generate KO cells. For the generation of RNPs, commercially available synthetic gRNAs (from Synthego) were preincubated at room temperature with Cas9 (IDT; Alt-R S.p. Cas9 Nuclease) (75 pmol gRNA: 31 pmol Cas9) for 30 min. One million cells per reaction were resuspended in 20 μl P3 buffer (Lonza) as per the manufacturer’s instructions. The RNP complex was added to the cells and electroporated on a 16-well strip using Lonza electroporator with EO115 setting. 80 μl of cells were added immediately to the cells and allowed to recover for 10 min. Cells were then plated on 2 wells of 96 wells with 150 μl media in each well. The mutant cells were used 7–10 d post electroporation.

In vitro A375 cell-killing assay with 1G4-T cells

A375 cells were engineered to constitutively express mOrange via lentiviral transduction. For setting the killing assay, 30,000 A375 cells were plated per well on a 96-well plate for 24 h before adding 90,000 (1:3 ratio) of T cells. The T cells used for this purpose were cultured in media without IL-2 for 24 h before use on killing assays. The plate was imaged using IncuCyte for 20 h after which all the cells from the wells were extracted. Leftover A375 cells were extracted using trypsin. The T cells were stained with anti-CD25 antibody to assess the level of activation. Killing was determined by assessing the percentage of leftover A375 cells post 20 h of co-incubation with 1G4 T cells. Killing levels were also quantified by measuring the area and the intensity of “red” fluorescence.

Assessment of KO efficiency

1G4-TCR and Jurkat KO of ARPC4 were validated using an antibody against ARPC4 Proteintech rabbit anti-human ARPC4, polyclonal antibody (Cat no. 10930-1-AP). LCK KO Jurkat cells were validated using anti-LCK antibody (Clone 3A5).