Non-canonical Hedgehog signaling through L-type voltage gated Ca2+ channels controls CD8+ T cell killing

Cytotoxic CD8+ T lymphocytes (CTLs) are critical to the immune response against intracellular pathogens and cancer and act by eliminating infected and malignant cells through targeted secretion of cytotoxic granules. Hedgehog (Hh) signaling has been shown to be critical for CTL killing. Interestingly, Hh signaling in CD8+ T cells is not induced by extracellular Hh ligands but is initiated upon T cell receptor (TCR) engagement. How the TCR induces the Hh pathway independently of extracellular Hh ligands is unknown. Here we show that the Hh transcription factor Gli1 is essential for efficient CTL function and is induced downstream of the TCR by an extracellular Ca2+ influx selectively controlled by L-type voltage gated Ca2+ channels localized at the plasma membrane. We demonstrate that this novel mode of Hh signaling induction is independent of the canonical Hh pathway and represents the primary mechanism of Gli1 induction in naïve CD8+ T cells, while CTLs can also activate Gli1 via MAP Kinase signaling. Importantly, we show that this L-type voltage gated Ca2+ channel-controlled Gli1 induction is functionally required for CTL killing in mice and humans. Gli inhibitors are currently in clinical trials against various cancers and our observations indicate that they likely inhibit the anti-tumor response. Significance statement Cytotoxic CD8+ T cells (CTLs) kill infected and malignant cells by targeted secretion of cytotoxic granules. Hedgehog signaling is critical for effective CTL killing and is activated by the T cell receptor (TCR) independently of exogenous Hedgehog ligands. This study shows that Hedgehog transcription factor Gli1 is required for CTL killing and identifies L-type voltage gated Ca2+ channels (Cav1) as essential regulators of CTL killing in mouse and human, by virtue of their ability to activate Gli1 downstream of the TCR. This Cav1-Gli1 axis operates independently of canonical Hedgehog signaling. Our work suggests that caution is required when using Gli inhibitors, currently in trials as anti-cancer therapeutics, since they may dampen the anti-tumor response.

ligands is unknown. Here we show that the Hh transcription factor Gli1 is essential for efficient CTL function and is induced downstream of the TCR by an extracellular Ca 2+ influx selectively controlled by L-type voltage gated Ca 2+ channels localized at the plasma membrane. We demonstrate that this novel mode of Hh signaling induction is 25 independent of the canonical Hh pathway and represents the primary mechanism of Gli1 induction in naïve CD8 + T cells, while CTLs can also activate Gli1 via MAP Kinase signaling. Importantly, we show that this L-type voltage gated Ca 2+ channel-controlled Gli1 induction is functionally required for CTL killing in mice and humans. Gli inhibitors are currently in clinical trials against various cancers and our observations indicate that 30 they likely inhibit the anti-tumor response.

INTRODUCTION 45
Cytotoxic CD8 + T lymphocytes (CTLs) eradicate infected and cancerous cells by targeted release of cytotoxic granules. Two types of immune synapses are important for this to occur: the signaling and the cytotoxic synapse. The signaling synapse initiates CTL differentiation and is formed between a naïve CD8 + T cell and an antigen presenting cell (APC). The cytotoxic synapse is formed between a CTL and its target cell whereby 50 engagement of the TCR initiates the secretion of granules containing cytotoxic perforin and granzyme. Both synapses promote optimal TCR signaling 1 .
Immune synapses are structurally and functionally very similar to the primary cilium, a hairlike protrusion from the cell body present on most cells. For example, both structures dock the centrosome at the plasma membrane via distal appendage proteins and are key 55 signaling hubs and sites of focussed endo-and exocytosis 1, 2 . The similarities between immune synapses and primary cilia prompted us to study Hedgehog (Hh) signaling -a pathway functionally tied to the primary cilium in vertebrates -at the T cell synapse. Although various roles for Hh signaling during T cell development in the thymus have been proposed 9 , little is known about Hh signaling in mature T cells. We have previously shown that Hh signaling is necessary for CD8 + T cell killing and, interestingly, is initiated independently of extracellular Hh ligands 10 . We previously demonstrated that proximal 80 TCR signaling is the main inducer of the Hh pathway 10 but the mechanism by which Hh signaling is initiated downstream of the TCR remains unknown.
Gli1 is the only Gli transcription factor expressed in CD8 + T cells. Here we show that Gli1 is critical for CD8 + T cell killing. We demonstrate a key role for MAPK signaling inducing 85 Gli1 in CTLs, as has been previously described in other cell types and tumor cells 11 . Additionally, we find binding sites of the MAPK induced transcription factor activator protein 1 (AP-1) in the Gli1 promoter. Most importantly, we identify a previously unknown, non-canonical mode of Hh signaling that culminates in Gli activation via L-type voltage gated Ca 2+ channels (Cav1 family) in both CTLs and naïve CD8 + T cells. 90 Published work has shown that constitutive loss of Cav1 channels in all tissues leads to defects in T cell development and maturation, with subsequent functional impairment of peripheral T cells 12, 13 . However, it is not fully clear what cell-intrinsic role these channels play in fully mature T cells. Our work demonstrates that Cav1 channels are critical for mature CD8 + T cell killing by virtue of their ability to induce Gli1 via non-canonical Hh 95 signaling.

Gli1 is important for cytotoxic T cell (CTL) killing
Gli1 is the only Gli transcription factor reported to be present in CD8 + T cells 10 and functions as a reliable marker of Hh signaling activation 14 . Previous work has shown that CTLs treated with the small molecule Gli inhibitor GANT61 have diminished killing ability 10 . We wanted to confirm this observation genetically and generated Gli1 eGFP/eGFP (Gli1 105 KO) by breeding Gli1 eGFP/+ mice to homozygosity, disrupting the expression of the Gli1 gene. Gli1 KO mice have a phenotypically normal peripheral T cell compartment (Suppl. Fig. 1) and CD8 + T cells from the Gli1 KO mice lack Gli1 protein (Fig. 1A). CD8 + T cells from OTI TCR transgenic mice recognize ovalbumin (Ova) peptide residues 257-264 in the context of MHCI H2K b and we bred Gli1 KO mice expressing the OTI TCR. We 110 compared Gli1 WT and KO OTI cells in their ability to kill Ova-presenting EL-4 murine lymphoma target cells. Strikingly, specific T cell killing was reduced by 25-50% in the Gli1 KO cells compared with the Gli1 WT controls (Fig. 1B). Thus, Gli1 is functionally important for effective CTL killing.

MAP kinase signaling downstream of the TCR promotes Gli1 induction in CTLs
CD8 + T cell killing is initiated by engagement of the TCR and the TCR-associated proximal tyrosine kinase Lck has been shown to be required for the induction of Gli1 10 . Induction of Gli1 mRNA is a robust readout of Gli1 activation and thus active Hh signalling 14 . We 120 therefore wanted to investigate how signaling downstream of the TCR leads to Gli1 activation.
TCR engagement by cognate antigen presented on MHCI molecules leads to recruitment of Lck to the TCR complex ( Fig. 2A). Lck phosphorylates ZAP70 (ζ-chain-associated protein kinase of 70 kDa) which in turn phosphorylates the adaptor protein LAT (Linker 125 for activation of T cells) leading to the formation of the LAT signalosome. The LAT signalosome propagates signal branching into three major signaling pathways ultimately culminating in the activation of the transcription factors NFAT, NFkB, and AP-1 15 .
The three signaling branches of the TCR: NFAT, NF-κB, and AP-1, respectively, can be pharmacologically manipulated. Ionomycin is a Ca 2+ ionophore that increases cytosolic 130  Fig. 2B). When Gli1 induction was assessed at steady state and after 3 hours of restimulation, VIVIT-and DN IkBa -expressing CTLs had no defect in Gli1 induction compared to GFP-transfected controls (Fig. 2C). 150 The third branch of TCR signaling leading to AP-1 activation can be blocked by U0126, a selective MAP Kinase (MEK1 and MEK2) inhibitor (Suppl. Fig. 2C) 18 . U0126 treatment did not affect cell viability (Suppl. Fig. 3A), but led to a significant reduction of Gli1 induction after TCR stimulation in CTLs and to a much lesser degree in naïve CD8 + T cells (Fig. 2D). MAPK signaling has been shown to be able to activate Gli1 through one of two mechanisms. The first is through direct activation of Gli1 transcription by AP-1 heterodimer which is formed by one Fos and one Jun (c-Jun, JunB or JunD) family member. The second mechanism is via the kinase activity of MAPK which activates an unknown upstream regulator of Gli1 11 .

160
To determine which mode of MAPK signaling is responsible for the Gli1 induction, we

Ca 2+ is required for Gli1 induction 175
We could not clearly associate the induction of Gli1 with one specific branch of TCR signaling in naïve CD8 + T cells. While MAPK signaling accounted to some extent for the induction of Gli1 in CTLs, it was still not clear what other pathways drive the induction of Gli1 in CD8 + T cells. Ca 2+ is required for the initiation of canonical Hh signaling at the level of the Ihh:Ptch interaction 22 and is an important second messenger in T cells 23 . We 180 therefore investigated whether Ca 2+ signaling was important for Gli1 induction. We subjected naïve T cells and CTLs to treatment with a cell-permeable high-affinity Ca 2+ chelator, BAPTA-AM, prior to TCR stimulation. Ca 2+ chelation with BAPTA-AM led to complete abrogation of Gli1 induction in naïve T cells and a significant reduction of Gli1 mRNA and protein induction in CTLs upon TCR stimulation (Fig. 3A). BAPTA-AM can 185 chelate Ca 2+ in both intra-and extracellular compartments. To determine the relative contribution of each of these pools of Ca 2+ to Gli1 induction, we repeated the experiments using non-cell-permeable BAPTA or EGTA which both chelate extracellular Ca 2+ . Extracellular Ca 2+ chelation significantly reduced Gli1 induction in CTLs (Fig. 3B). Thus, Ca 2+ signaling is essential for Gli1 induction and relies on both extra-and intracellular 190 Ca 2+ pools.

L-type voltage gated Ca 2+ channels are critical for Gli1 induction and act independently of MAPK signaling
We next sought to determine whether Ca 2+ signaling through a specific Ca 2+ channel is 195 responsible for regulating Gli1 induction. We targeted channels previously shown to be expressed in T cells such as CRAC channels, L-type voltage-gated Ca 2+ channels (Cav1), TRPV, TRPA and P2X channels with small-molecule inhibitors (Fig. 3C) 23 . Treatment of CTLs with small molecule inhibitors at doses previously published in the literature had no effect on cell viability during the course of the experiment (Suppl. Fig. 3A). Interestingly, 200 inhibition of CRAC channels showed no effect on Gli1 induction and neither did inhibition of TRPV, TRPA and P2X channels.
Only inhibition of Cav1 channels with Nifedipine showed a dose-dependent reduction of Gli1 induction in CTLs (Fig. 3D). Cav1 channels have been shown to play a critical role in CD8 + T cell development 24 . Evidence in the literature has shown that constitutive 205 knockout of CACNA1F (encoding Cav1.4) during T cell development results in impaired MAPK signaling strength 12 . We hence sought to determine whether the effect of Nifedipine on Gli1 induction in CTLs was through inhibition of MAPK signaling or through an independent mechanism. Inhibition of both, MAPK and Cav1 channels, showed an additive decrease in Gli1 induction ( Fig. 3D) but importantly pErk levels were unaffected 210 by treatment with Nifedipine (Fig. 3E), indicating that Cav1 channels were regulating Gli1 induction through a mechanism distinct from MAPK signaling. Given that MAPK signaling is not a major driver of Gli1 induction in naïve CD8 + T cells (Fig. 2D) we sought to investigate whether Cav1 channels might be the main source of Gli1 induction. Nifedipine treatment of naïve CD8 + T cells resulted in complete abrogation of Gli1 induction post 215 TCR stimulation (Fig. 3F).
We wanted to confirm this finding by genetic ablation of Cav1 channels and performed CRISPR knockouts of each of the known Cav1 α1 subunit coding genes (CACNA1S/C/D/F) in primary murine CD8 + T cells. Knockout of individual genes did not show a significant decrease in Gli1 induction (Suppl. Thus, we have shown that Cav1 channels are the main inducers of Gli1 in naïve CD8 + T cells and in CTLs act together with MAPK signaling to induce Gli1.

CD8 + T cells and are important for killing
To interrogate whether Cav1 channels are also expressed in human CD8 + T cells we Next, we wanted to know whether the Cav1 channel inhibitor Nifedipine could block GLI1 induction in human CD8 + T cells as it did in murine cells (Fig. 3F). The concentration of 240 Nifedipine used had no effect on cell viability (Suppl. Fig. 3B) but induction of GLI1 was ablated in Nifedipine-treated human CD8 + T cells (Fig. 4B).
Cav1 channels have been hypothesized to localize to the plasma membrane in T cells 12 but have never been localized by immunofluorescence. Using an antibody against Cav1.4 25 , we stained human CD8 + T cells from different healthy donors and consistently 245 observed specific staining at the plasma membrane as well as on intracellular vesicles in all donors (Fig. 4C).
To investigate whether Cav1 channels are implicated in the ability of human CD8 + T cells to kill tumor targets, we treated human CTLs from various donors with Nifedipine or carrier control before subjecting them to a killing assay. While carrier-treated human CTLs were 250 able to kill tumor targets, Nifedipine-treated CTLs from all donors tested showed a marked decrease in killing ability (Fig. 4D).
Taken together, we have shown that, similarly to murine CD8 + T cells, human CD8 + T cells express Cav1 channels and depend on them for GLI1 upregulation and consequent target cell killing. 255

L-type voltage gated Ca 2+ channels regulate T cell killing in a Gli1-dependent manner
To definitively answer whether the effect of Cav1 channel blockade via Nifedipine on CTL killing is achieved through Gli1 or an independent mechanism, we treated Gli1 WT and 260 Gli1 KO CTLs with Nifedipine or carrier control (Fig. 5A). We find that Nifedipine treatment leads to a defect in killing in Gli1 WT CTLs, which we have shown to be independent of effects on cell viability or MAPK signaling (Fig. 3E, Suppl. Fig. 3A). This indicates that Cav1 channels are important regulators of CTL killing. Treatment of Gli1 KO CTLs with Nifedipine showed no additive defect in killing compared to carrier control (Fig. 5A), 265 indicating that the presence of Gli1 is required for Nifedipine to exert its effect on CTL killing.
Thus, we have shown that L-type voltage gated Ca 2+ channels are critical for CTL killing and that the mechanism for this killing phenotype is by regulating the induction of Gli1.

Gli1 induction upon TCR signaling is independent of canonical Hh signaling via Ihh and Smo
Given the role of Smo in maintaining Gli1 steady-state protein expression 10 we asked whether canonical Hh signaling was needed for the induction of Gli1 after TCR 275 stimulation. To investigate this, we generated two different conditional knockout lines (Suppl. Fig. 5A,B). For the first mouse line, we crossed GzmB ER T2 Cre mice 26 to Smo f/f mice, to generate mice in which Cre is only active in mature CTLs. The second mouse line was generated by crossing dLckCre mice to Ihh f/f mice, in which the Cre is active only after T cell development has been completed in the thymus. Ihh is the only Hh ligand 280 expressed by CD8 + T cells and Hh signaling has been suggested to be entirely intracellular in CD8 + T cells 10 . We therefore decided to knock out endogenous Ihh in  27 , resulting in 80000-and 300-fold overexpression, respectively (Suppl. Fig. 5C). Strikingly, we did not observe any further increase in the induction of Gli1 as compared to empty vector-transduced control CD8 + T cells. (Fig. 5D). 300

Thus, canonical Hh signaling via Ihh and Smo is not responsible for the induction of Gli1
downstream of the TCR, and is independent of the MAPK and L-type voltage-gated Ca 2+ channel-induced Gli1 activation.

DISCUSSION
In this manuscript we have uncovered a novel non-canonical Hh signaling pathway whereby a Cav1 channel-mediated Ca 2+ flux can induce Gli1. We show that this 310 mechanism of Gli1 induction is functionally critical for CTL killing, providing a direct mechanistic link between Cav1 channels and CTL killing (Fig. 5E).
There is currently great scientific interest to unmask the roles of Cav channels in T cell biology. Constitutive knockout of Cav1.4 or the Cav channel regulatory b3 subunit leads 315 to signaling defects in thymocytes and apoptosis of peripheral T cells, respectively 13 . This makes it challenging to draw conclusions about the role of Cav channels in mature peripheral T cells. Here we use CRISPR technology to functionally ablate Cav channels in primary mature T cells only.
All four family members of the L-type Ca 2+ channels (Cav1.1, Cav1.2, Cav1.3, and Cav1.4) 320 are expressed in murine T cells 24 and we find that only Cav1.4 and to a lesser extent Cav1.3 are expressed in human CD8 + T cells. We show that Cav1.3 and Cav1.4 (and to a lesser extent Cav1.1) are involved in the induction of Gli1 in CD8 + T cells. Apart from the localization of Cav1.4 to lipid rafts little is known about the cellular localization of Cav channels 12 . Here we find that Cav1.4 is expressed at the plasma membrane and on 325 intracellular vesicles in primary human CD8 + T cells.
There is evidence from the literature that in neuronal cells canonical Hh signaling can induce Ca 2+ spikes via Cav and TRPC1 channels 7 . Here we show that the "reverse" is true in T cells, where we show that Ca 2+ signaling via Cav channels induces Hh signaling. 330 Our work uncovers Ca 2+ signaling via Cav channels as a novel mechanism of noncanonical, Ptch/Smo-independent Hh signaling.
The literature supports the notion that our proposed Cav-Gli axis may be active in other cells. Humans with mutations in Cav channels phenotypically display heart arrhythmias and signs of autistic spectrum disorder (consistent with the critical roles of Cav channels 335 in the heart and brain) as well as signs of syndactyly 28 , a phenotype which has been robustly linked to inactivating mutations in the Hh pathway 29 . Furthermore, studies in mice showed that knockdown of Cav channels leads to abnormalities in skeletal development 30 , highly reminiscent of the defects observed in Ihh KO mice 31 . These studies suggest that loss-of-function phenotypes of Cav channels phenotypically mimic 340 known Hh loss-of-function phenotypes, consistent with the hypothesis that a Cav-induced mechanism of Gli activation is operational in other cell types.
Interestingly, we uncover that in CTLs MAPK activation downstream of the TCR critically contributes to Gli1 transcription which has been shown in other cell types and cancers 11 . 345 Of interest is the difference of Gli1 regulation between naïve CD8 + T cells, that seem to solely rely on Cav signaling, and CTLs, that also employ MAPK signaling. A possible explanation for this could be that naïve CD8 + T cells have much longer contact times with an APC and form more stable synapses in vivo when they are stimulated than CTLs, which have shorter contact times with their target cells and have been shown to often 350 require multiple rounds of contact in vivo to efficiently kill target cells 32 . Therefore, CTLs may have evolved two pathways to ensure appropriate levels of Gli1 activation.
We show that Gli1 activation by Cav channels is important for killing in murine and human CD8 + T cells (Fig. 5A, Fig. 4E) which is a rapidly-induced process with killing observed 355 within a few hours in vitro. This is consistent with findings in the literature that Gli1 activation can be rapidly induced 33 , but the question remains how the transcription factor Gli1 can exert its effect on killing in such a short timeframe. Given that Gli1 can regulate the cytoskeleton 10 , we hypothesize that Gli1-induced transcription in CTLs upon target recognition is required for cytoskeletal rearrangement and immune synapse formation 360 upon successive target cell contacts and serial killing.
Regarding the activation of Cav channels in T cells, it is tempting to speculate that this is PKC-driven because of two observations. First, Cav channels in T cells lack the ability to respond to depolarization and instead become activated upon TCR engagement 34 . And second, PKC is a well-known activator of Cav channels in excitable cell types and has 365 been hypothesized as the link between TCR and Cav channel activation in T cells 35,36 .
This would explain why we find that PMA (a PKC agonist) is able to drive Gli1 induction (Fig. 2B), particularly in naïve CD8 + T cells where this effect cannot simply be explained by PKC-related induction of MAPK signaling (which has no effect on Gli1 induction). The requirement for PKC for Cav activation and the different nature of the Ca 2+ flux induced 370 by ionomycin might also explain why ionomycin administration alone cannot induce Gli1.
The precise mechanism by which Cav1 channel activation results in Gli1 induction is not known at this point. One potential mechanism is via Ulk family proteins that can be activated by Ca 2+ and have been shown to activate Gli1 via proteolytic cleavage in a very 375 short timeframe 33,37 . And indeed we find Ulk3 expressed in CD8 + T cells (Suppl. Fig.   5D).
Canonical Hh signaling is active for long periods of time, is initiated by extracellular ligands, and computes a cell fate choice 3 . By contrast, CTL killing must happen much 380 more rapidly and should not rely on extracellular ligand gradients. Canonical Hh signaling is a more ancient signaling pathway than TCR signaling 2 . The pathway might have come under evolutionary pressure for the induction of Gli1 to be more rapid and isolated from extracellular signals resulting in direct activation by the TCR.

385
Taken together, we show that Gli1 is critical for CD8 + T cell killing of tumor cells and that Gli1 activation is intricately coupled to CD8 + T cell activation downstream of the TCR via

Mice
RAG2KO were a generous gift from Suzanne Turner (University of Cambridge) and OTI mice were purchased from the Jackson Laboratory (C57BL/6-Tg(TcraTcrb)1100Mjb/j, Stock no. 003831). OTI RAG2KO mice were generated from these. Gli1-eGFP mice were a 400 generous gift from Alexandra Joyner (Sloan Kettering Institute) 39

In vitro killing assay 485
CD8 + T cell cytotoxicity was assessed using a Cyto Tox 96® Non-radioactive Cytotoxicity Assay kit (Promega, Cat no. G1780) according to the manufacturer's instructions. Murine and human CTLs were generated as previously described and used on day 6/7 or day 14/15 post-stimulation, respectively. EL-4 cells were used as target cells for murine OTI CTLs and pulsed for 1h at 37°C with 1µM SIINFEKL. P815 cells were used as target cells for human 490 CTLs and pulsed for 1h at 37°C with 1µg/ml anti-CD3 antibody (clone UCHT1, Biolegend, Cat no. 300438). They were washed and resuspended at 1x10 5 cells/ml in killing assay media (RPMI without phenol red + 2% FCS) in a round-bottom 96 well plate.
T cells were washed, resuspended in killing assay media and plated at the indicated effector:target ratios. Plates were incubated at 37°C prior to collection of supernatant at the 495 designated timepoints. Absorbance was measured at 490nm using a CLARIOstar microplate reader (BMG Labtech).

Intracellular staining 505
Following surface staining, cells were fixed with BD Cytofix/Cytoperm Plus Fixation Buffer (BD Biosciences, Cat no. 554715) as per manufacturer's instructions and stained with fluorophore-conjugated antibodies at the appropriate concentrations (Table 3)

Small molecule treatment of CD8 + T cells
For small molecule treatment studies, naïve CD8 + T cells or CTLs were incubated for 1h (2h 535 in the case of BTP2 -40 ) prior to (re-)stimulation. Doses of Ca 2+ channel blockers were chosen based on doses established for use in T cells in the literature (see Table 4   Each sample was run in triplicate with Tbp and/or CD3e used as housekeeping genes. In 555 addition, each experiment included a non-template control and primer/probes were validated by no RT controls. Samples were run on a QuantStudio 6 Flex Real-Time PCR System (Thermo Fisher). Levels of Gli1 mRNA were determined using probe (Mm00494654_m1) and confirmed in some of the experiments with a second probe (Mm00494645_m1).
Expression of the gene transcript of interest was calculated with the ΔCt method 46 . The 560 cycle threshold (Ct) value from the gene of interest was subtracted from the housekeeping gene and transformed with a factor of 2^(-ΔCt) to give the fold expression relative to the housekeeping gene.

Immunofluorescence 565
Primary human CTLs at day 11 post-stimulation were plated onto glass slides at 4x10 6 /ml and allowed to adhere to the glass for 10min at 37°C in an incubator. Cells were fixed with 4% PFA (16% PFA solution, CN Technical Services, cat no. 15710-s, 1x PBS) for 10min at room temperature. Slides were washed 5 times with PBS and blocked with blocking buffer (PBS + 1% bovine serum albumin (BSA, Sigma, cat no. A3912; 50 g 570 lyophilized powder) + 0.1% TritonX-100 (Alfa Aesar)) for 30min at room temperature.
Confocal spinning disc microscopy was performed on an Andor Dragonfly 500 (Oxford Instruments) and images were processed using Imaris software (Bitplane/Oxford 590 Instruments).

Generation of retrovirus
One day prior to the generation of retrovirus, HEK 293T cells were seeded in a six well 610 plate resulting in 75% confluency the next day. Media was replaced with 2ml fresh DMEM

Retroviral transduction of CD8 + T cells 625
Naïve CD8 + T cells were isolated from C57BL/6 spleens and stimulated with plate bound anti-CD3e/CD28 antibodies as described previously. After 24h of stimulation, T cell media was partially withdrawn and replaced by retroviral supernatant in a 1:2 ratio supplemented with protamine sulphate (Sigma, Cat no. 1101230005) at a final concentration of 9µg/ml and IL-2 to maintain the standard CD8 + T cell polarization conditions as described above. 630 Cells were then centrifuged at 1800rpm for 15min at 32°C and subsequently placed in a humidified cell culture incubator. 48h cells were centrifuged to wash off retroviral supernatant and cultured after that in complete RPMI + IL-2. complete media prior to maintenance in culture with complete RPMI + IL-2. Purity of the sort was routinely over 95% as measured by FACS.

Nucleofection of CTLs
The GFP-VIVIT plasmid was a gift from Anjana Rao (Addgene Plasmid # 11106; A16046.AE) in PBS prior to fixation in 4% PFA for 10min light-protected at room temperature. Cells were washed twice in perm wash buffer before staining in 100µl perm wash buffer containing anti-p65 APC antibody (Table 4) for 30min light-protected at room 675 temperature. Cells were washed twice in perm wash buffer, subsequently resuspended in 50µl perm wash buffer and transferred to a 1.5ml Eppendorf tube. DAPI was added (0.5mg/ml) immediately prior to acquisition on an Amnis ImageStream (MilliporeSigma) imaging cytometer. Data was analysed with IDEAS Software (MilliporeSigma).

Analysis of Chip-Seq data
Coverage bigwig files were downloaded from GEO (GSE54191). The Integrative Genomics Viewer (IGV) was used for coverage analysis 47 .

DATA AVAILABILITY
This study includes no data deposited in external repositories.

ACKNOWLEDGEMENTS
We thank all members of the de la Roche lab for constructive comments on the manuscript. Special thanks and gratitude go to Ellie Pryor, Nicky Jacobs and Gemma 695 Cronshaw from the BRU for expert animal care; Fadwa Joud from the Microscopy core for assistance in confocal microscopy and the flow cytometry core for assistance with cell sorting. We thank Rose Zamoyska for advice on CRISPR experiments in primary T cells, Fanni Gergely for help with reagents, and Gillian Griffiths for advice on the manuscript.

CONFLICTS OF INTEREST
The authors declare that they have no conflict of interest. were co-cultured with ovalbumin-pulsed EL-4 target cells for 4 hours at the indicated effector to target ratios and subjected to an LDH cytotoxicity assay. Representative data of n=3 independent experiments. Error bars indicate SD.