FcεRγI promotes canine CD8 chimeric antigen receptor T cell cytotoxicity through a Syk-NFκB axis

Abstract

Comparative oncology has advanced cancer immunotherapy, although cellular mechanisms governing chimeric antigen receptor T cell (CART) therapy in canines are poorly understood. In a first-in-canine trial, anti-CD20 CART with canine 4–1BB-CD3ζ (cBBζ) domains induced CD20-negative lymphoma outgrowth but did not persist or deplete B-cells. Here we show that canine CARTs incorporating human (h)BBζ demonstrate superior therapeutic function, mediated by FcεRγI. HBBζ-CART showed greater cytolysis and CD8 T cell outgrowth than cBBζ-CART in repetitive killing assays and a canine B-cell leukemia xenograft model. Transcriptional profiling revealed upregulation of FCER1G and innate-like genes in CD8 hBBζ− versus cBBζ-CARTs. CRISPR-mediated FCER1G deletion and pharmacologic Syk/NFκB inhibition indicated that Syk-NFκB signaling regulates FcεRγI-mediated enhancement of hBBζ-CART cytotoxicity, associated with increased granzyme B and IFNγ/TNFα production. Syk-NFκB signaling promotes FcεRγI expression in hBBζ CARTs, and CAR-TCR interactions potentiate NFκB signaling to upregulate FcεRγI and enhance CART function. These studies identify a potent therapeutic subset of innate-like canine CARTs induced by hBBζ signaling, which holds potential to improve both canine and human CART therapy.

Graphical Abstract

Canine CARTs expressing human 41BB-CD3ζ (hBBζ) domains outperform canine CARTs with canine-derived domains. HBBζ signaling induces an innate-like transcriptional program, characterized by high expression of FCER1G, encoding FcεRγI. FcεRγI promotes CD8+ CART cytotoxicity, and strategies to enrich for FcεRγI⁺ CARTs may be therapeutically beneficial.

Introduction

Chimeric antigen receptor T cell (CART) technology has transformed the cancer treatment paradigm for refractory B-cell leukemias and lymphomas by genetically programming a patient’s own T cells to kill CD19 antigen-expressing B-cells. CARTs also proliferate and differentiate into memory CARTs that can provide potentially lifelong protection against cancer recurrence. The clinical approvals of autologous CD19- or BCMA-targeted CART products for B-cell malignancies and multiple myeloma have led to an exponential increase in studies to adapt CART technology for additional cancers, diseases beyond oncology, and allogeneic manufacturing, although clinical approvals in the latter applications have not yet been achieved.

The value of a comparative oncology approach to improve cancer therapy became apparent in the mid-twentieth century with studies of hematopoietic cell transplants in pet dogs with naturally occurring lymphoma, which established methods for stem cell engraftment, prevention of graft-versus-host disease, and pre-conditioning and post-transplant immunosuppression, many of which helped to improve the clinical success and safety of human stem cell transplantation.^1,2 Dogs spontaneously develop many of the same cancers as humans,^3,4 with shared pathophysiology and epidemiology demonstrated by similar patterns of disease onset and progression, as well as mutational landscapes.^5–8 Like humans, dogs have experienced immune systems that are serially challenged through infections, vaccinations, and environmental exposures. Furthermore, at a population level, dogs are outbred and genetically diverse, with lifespans allowing for accelerated longitudinal follow-up studies.

Recently, a comparative oncology approach was applied using CART technology. In a first-in-canine CART clinical trial, pet dogs with refractory diffuse large B-cell lymphoma (DLBCL) received anti-CD20 CARTs expressing canine 4–1BB-CD3ζ (cBBζ) signaling domains. CARTs exhibited biological activity in vivo, evidenced by outgrowth of CD20-negative lymphoma B-cells⁹ and development of fevers and elevated cytokines resembling human cytokine release syndrome (CRS),¹⁰ a common and potentially life-threatening side effect of human CART therapy. CRS was not observed in preclinical mouse models prior to human clinical trials, although subsequent studies showed that CRS hallmarks can be experimentally induced in mice by modifying methods to allow high levels of intraperitoneal tumor cell accumulation prior to CART intraperitoneal administration.¹¹ However, anti-CD20-cBBζ CARTs failed to induce B cell aplasia, which was accompanied by poor in vivo CART expansion and lack of persistence. Collectively, these preliminary studies highlight an essential need for a mechanistic understanding of CAR biology in canine T cells to achieve clinical remissions and advance the field of CART comparative oncology. Here we investigate the role of CAR co-stimulatory domains in CART biologic activity and identify a novel population of innate-like CARTs defined by expression of the gamma subunit of the Fc epsilon receptor (FcεRγI), which synergizes with the endogenous T cell receptor to enhance CART therapeutic function.

Results

Human BBζ signaling domains enhance canine CART efficacy in vitro and in vivo

Previous studies of human CARTs using xenogeneic mouse models have shown that CAR co-stimulatory domains modulate in vivo longevity and cytotoxic function.^12,13 In syngeneic mouse models, murine CARTs expressing human 4–1BB-CD3ζ (hBBζ) cytoplasmic domains exhibit greater in vitro cytolytic potency, in vivo tumor control, and CART persistence than murine CARTs expressing mouse CD28-CD3ζ or 4–1BB-CD3ζ.¹⁴ To determine whether hBBζ similarly enhances activity in canine CARTs, we designed CARs expressing hBBζ (Figure 1A). We also designed a canine CD28-CD3ζ (c28ζ) expressing CAR, as this mimics the anti-CD19 CD28-CD3ζ constructs that have been clinically approved for human cancer therapy. Both CARs were evaluated against the clinically-tested canine BBζ (cBBζ) as a reference control.⁹ All constructs expressed the same anti-canine CD20 single-chain variable fragment (scFv) extracellular domain with canine CD8α hinge and transmembrane domains, cloned into the MSGV1 γ-retroviral vector.¹⁵ CBBζ, hBBζ, and c28ζ CARs were comparably well-expressed across multiple canine donors (Figure 1B).

Figure 1. Human BBζ signaling domains enhance canine CART cytolytic potency and production of IFNγ and TNFα in vitro.

(A) Schematic of CAR constructs evaluated. HD, hinge domain; TMD, transmembrane domain; cBBζ, canine 4–1BB-CD3ζ; c28ζ, canine CD28-CD3ζ; hBBζ, human 4–1BB-CD3ζ. (B) Surface CAR expression levels from donor-matched cells were determined one week after viral transduction. Data are representative of 3 experiments (n = 3 donors). (C) Donor-matched CART cytotoxicity was assessed using luminescent readouts 20 hours after co-culture with GL1 or K562 WT versus canine CD20+ target cells. Data represent mean (SD) of technical quadruplicates and are representative of 3 experiments (n = 3 donors). (D) Luminex assessed canine IFNγ and TNFα production after co-culture with WT or CD20+ K562 cells. Data represent duplicates. CARTs underwent three rounds of restimulation and were assessed for (E) cytotoxicity against CD20+ GL1 or K562 tumor cells, (F) percentage outgrowth of CD8 versus CD4 CARTs, and (G) cumulative expansion (representative of 2–3 donors and experiments). Arrows denote days of CART restimulation. Where indicated, a 2-way ANOVA with Dunnett’s multiple comparisons test computed significance of hBBζ versus cBBζ or c28ζ CART; ns – not significant, ***p<0.001, ****p<0.0001. Cytotoxicity values reflect mean (SD) across technical quadruplicates.

To assess CART potency, we evaluated cytolytic activity following incubation with canine CD20+ versus WT GL1 or K562 tumor cells. As compared to primary canine B cells, GL1-CD20 target cells express similar CD20 surface expression and density, while human K562-CD20 cells demonstrate a comparable number of surface CD20 molecules but a 50% reduction in CD20 surface density (Figure S1A–C). As compared to c28ζ and cBBζ-expressing CARTs, hBBζ-expressing CARTs demonstrated significantly greater GL1-CD20 cell lysis at low effector-to-target (E:T) ratios and enhanced K562-CD20 lysis across all E:Ts tested (Figure 1C). We profiled cytokine production after co-culture with K562 cells to exaggerate differences between groups and limit unwanted detection of target-cell derived cytokines. HBBζ CARTs produced significantly more IFNγ and TNFα than either c28ζ or cBBζ-expressing CARTs across tested E:Ts (Figure 1D). Incubation of BBζ CARTs with WT targets resulted in some cytolysis at high E:Ts, with minimal accompanying cytokine production (Figure 1C–D).

We also performed an in vitro repetitive rechallenge assay to mimic in vivo target cell re-encounter and evaluate the durability of cytolytic capacity. HBBζ CARTs displayed greater serial target cell lysis against CD20+ GL1 and K562 target cells (Figure 1E). Sustained cytotoxicity by hBBζ CARTs was associated with higher CD8:CD4 T cell ratios compared to cBBζ CARTs, despite similar CD8:CD4 ratios after manufacturing (Figure 1F) and similar total T cell expansion after each target cell rechallenge (Figure 1G).

To assess whether hBBζ-expressing CARTs also elicit greater tumor control in vivo, we established an NSG mouse xenograft model of canine leukemia using luciferase-expressing GL1-CD20 tumor cells. Bioluminescence imaging revealed hBBζ-expressing CARTs demonstrated faster and more robust tumor control than cBBζ-expressing CARTs (Figure 2A–B). Improved tumor control by hBBζ CARTs correlated with greater CD8 T cell persistence compared to c28ζ or cBBζ CARTs (Figure 2C). Despite differences in the extent of tumor control, hBBζ and cBBζ CARTs exhibited comparable surface immunophenotypes, comprised of mostly T central memory (Tcm) and effector memory (Tem)-like populations (Figure 2D). C28ζ CARTs showed a significantly higher terminally differentiated Temra-like population than either cBBζ or hBBζ CARTs.

Figure 2. HBBζ CARTs induce superior tumor control and CD8 T cell persistence in vivo.

(A and B) NSG mice were injected with 7 × 10⁵ GL1-CD20+ tumor cells on day 0, followed on day 4 with 3.5 × 10⁶ CAR+ or donor-matched non-transduced (NTD) T cells (n = 3, NTD; n = 6, c28ζ; n = 7, cBBζ; n = 8, hBBζ). Tumor progression was measured by bioluminescence imaging every 2– 3 days. Plot reflects mean flux [photons per second (p/s)] (SD). (C and D) Mice were euthanized on day 21 and femurs were harvested to evaluate canine T cell (C) CD4 and CD8 persistence (2-way ANOVA with Tukey’s multiple comparisons test) and (D) surface immunophenotype (1-way ANOVA, Kruskal-Wallis test). Counts were back-calculated from percentages of GFP⁻/murine CD45⁻/canine CD3⁺CD5⁺ lymphocytes as determined by flow cytometry. ns – not significant, *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001.

Given that hBBζ CD8 CART outgrowth correlated with superior in vitro and in vivo cytotoxicity, we designed CAR constructs with hybrid human/canine 4–1BB and CD3ζ domains (Figure S2A) to define the relative importance of individual signaling domains supporting this phenotype. In an in vitro rechallenge experiment, CARTs produced similar expansion (Figure S2B). However, we observed significantly greater CD8 outgrowth by hBB.cCD3ζ and hBBζ CARTs compared to cBB.hCD3ζ CARTs, indicating that the human 4–1BB costimulatory domain promotes CD8 outgrowth in canine CARTs (Figure S2C–D). To determine the mechanism underlying CD8 outgrowth, we evaluated rates of CD8 CART apoptosis and proliferation. While the degree of apoptosis was similar between cBBζ and hBBζ CD8 CARTs (Figure S2E), CD8 proliferation was markedly higher in hBBζ CARTs (Figure S2F). Functionally, CART rechallenge produced similar percentages of CAR⁺ T cells between groups, although hBB.cCD3ζ-expressing CARTs showed higher CAR mean fluorescence intensity (MFI) (Figure S2G). Despite comparable CD8 CART outgrowth and higher CAR expression, hBB.cCD3ζ showed reduced target cell lysis versus hBBζ at low E:T ratios (Figure S2H). Therefore, CD8 outgrowth is insufficient to explain the observed enhancements of hBBζ CART activity.

HBBζ signaling promotes FcεRγI⁺ CD8 CARTs resembling innate-like T cells

To investigate the mechanism by which hBBζ-mediated signaling generates CD8 CARTs with enhanced cytotoxicity, we defined the transcriptional profiles of hBBζ and cBBζ CD8 CARTs after co-culture with CD20+ or WT GL1 tumor cells. Transcriptional profiling after incubation with WT cells served as a control to account for baseline differences in gene expression unrelated to CART-induced signaling. After co-culture with GL1-CD20 target cells, FCER1G, encoding the gamma 1 chain of the high affinity Fc epsilon receptor (FcεRγI), was most differentially upregulated in hBBζ CD8 CARTs (Figure 3A). HBBζ CD8 CARTs also displayed significant enrichment of genes encoding NK cell activating receptors (CD244, KLRB1, KLRC2, KLRD1), chemotactic chemokines (CXCL10, CXCL11), and innate signaling mediators (ITGAX, SPRY2). Of the genes upregulated in response to GL1-CD20 target cells, only CXCL10 was also upregulated after co-culture with WT GL1 cells (Figure 3B), demonstrating the observed signature is the result of hBBζ-mediated signaling. Incubation with WT GL1 cells also produced significant enrichment of NFκB-related transcripts (NFKB2, IRF8, NFKBIA) in hBBζ CD8 CARTs.

Figure 3. HBBζ signaling promotes innate-like FcεRγI+ CD8α+ canine CAR T cells.

(A and B) Transcriptional profiling of hBBζ versus cBBζ CD8+ CAR T cells after 36-hour co-culture with (A) CD20+ or (B) WT GL1 cells was performed using the Nanostring Canine IO Panel (n = 3 donors). The vertical dashed line reflects a 1.5-fold change and the horizontal dashed line defines a p-value of 0.05. (C) FcεRγI protein expression in cBBζ versus hBBζ CART was measured in unstimulated conditions and after 6 hours of co-culture with CD20+ or WT K562 cells. Significance was determined by 2-way ANOVA with Sidak’s multiple comparisons test (n = 3 donors). (D) Representative flow plots evaluating FcεRγI expression in CD4+ versus CD8+ hBBζ CARTs, pre-gated for live CD3+CAR+ lymphocytes. (E-H) Phenotypic associations of FcεRγI expression in hBBζ CART evaluated by flow cytometry. Populations were pre-gated for live CD3+CAR+CD8+ lymphocytes and evaluated for (E) forward and side scatter (n = 5 donors; 2-way ANOVA), (F) Granzyme B levels (n = 5 donors; Mann-Whitney test), (G) IFNγ production (n = 3 donors; 2-way ANOVA), and (H) NKp30 and NKp46 surface expression (n = 4 donors; Mann-Whitney test). ns – not significant, *p<0.05, ***p<0.001, ****p<0.0001.

A highly cytotoxic CD8 T cell subset has been described in humans, mice, and dogs, characterized by high expression of FCER1G and an innate-like gene signature.^16–19 However, roles for FCER1G or FcεRγI have not been previously validated in CART cells of any species. We first confirmed that hBBζ-directed signaling promotes FcεRγI protein expression. Both hBBζ and cBBζ CD8 CARTs expressed a low level of FcεRγI prior to target cell stimulation. After stimulation with K562-CD20 target cells, hBBζ CARTs, but not cBBζ CARTs, upregulated FcεRγI expression on CD8 cells (Figure 3C). CD4 T cells did not express FcεRγI (Figure 3D).

FcεRγI expression associated significantly with larger cell size and increased granularity (Figure 3E), as well as higher levels of Granzyme B and IFNγ production (Figure 3F–G), indicative of a cytotoxic phenotype. Similar to previously described human FcεRγI⁺ CD8 T cells, which co-express NKp30 but not NKp46,¹⁷ canine CD8⁺FcεRγI⁺ T cells demonstrate an NKp30⁺NKp46⁻ surface expression pattern (Figure 3H). However, unlike human CD8 T cells, in which FcεRγI serves as the signaling component for NKp30, canine NKp30⁺ CD8 T cells are mostly FcεRγI⁻ (Figure S3A–B).

FcεRγI enhances cytolytic potency and production of IFNγ and TNFα in hBBζ CARTs

To determine if FcεRγI is necessary for the enhancement of canine hBBζ CART cytotoxicity, we employed a CRISPR-Cas9 approach to disrupt FCER1G expression in canine hBBζ-CARTs. Since FcεRγI expression naturally varies between canine donors (Figure 4A), we selected a high-expressing and a low-expressing donor to assess whether differences in baseline FcεRγI expression influence the degree to which FcεRγI mediates cytotoxicity.

Figure 4. Loss of FcεRγI inhibits hBBζ CART cytotoxicity and production of IFNγ and TNFα.

(A) FcεRγI-expressing populations vary across NTD canine CD8+ T cells activated and expanded for 10–12 days (n = 6 donors). (B) Flow plots depicting CAR versus FcεRγI expression and viability in mock electroporated or FCER1G-knockout (KO) hBBζ CARTs. (C) FSC and SSC in FCER1G-KO versus mock hBBζ CARTs. (D and E) Determination of the effect of FCER1G-KO on (D) lysis of K562 WT and CD20+ tumor cells and (E) IFNγ and TNFα production. Significance determined by 2-way ANOVA. (F – H) Evaluation of whether FcεRγI overexpression rescues (G) cBBζ CART cytotoxicity (significance comparing cBBζ-FCER1G and cBBζ or hBBζ CART) and (H) IFNγ and TNFα production. Cytotoxicity data reflect mean (SD) across technical quadruplicats. Representative flow cytometry plots of CARTs after stimulation with K562-CD20+ target cells. Significance determined by 2-way ANOVA with Tukey’s multiple comparisons test. (F) Representative flow cytometry plots depicting FcεRγI expression in donor-matched CARTs. 1-way ANOVA, Krusal Wallis Test. Data are representative of 3 canine donors. Ns – not significant, *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001.

Disruption of FCER1G minimally affected CAR expression or viability but effectively eliminated FcεRγI expression (Figure 4B). Loss of FcεRγI expression reduced cell size and granularity to levels on par with those from naturally FcεRγI⁻ cells (Figure 4C). In FcεRγI^hi Donor 1, FCER1G-KO CARTs showed significant reductions in target cell cytolysis across a range of E:T ratios (Figure 4D). Similar trends were observed by FCER1G-KO CARTs from FcεRγI^low Donor 2, although the extent of cytolytic impairment was lessened. Loss of FcεRγI also significantly impaired IFNγ and TNFα production across the tested E:Ts from both donors’ CARTs (Figure 4E).

We next evaluated the sufficiency of FcεRγI to rescue cBBζ CART cytotoxicity. We generated a cBBζ-FCER1G-expressing CART and compared its functionality to donor-matched cBBζ and hBBζ CARTs (Figure 4F). FcεRγI partially rescued cBBζ CART cytotoxicity against K562-CD20 targets, with negligible lysis of WT targets (Figure 4G). FcεRγI also enhanced cBBζ CART production of IFNγ to levels comparable with hBBζ CARTs (Figure 4H). TNFα production also trended higher in cBBζ-FcεRγI CARTs.

Given the cytolytic advantages demonstrated by FcεRγI-expressing CARTs, we considered manufacturing strategies to enrich for FcεRγI⁺ hBBζ CARTs. In human and mouse CD8 T cells, exogenous IL15 induces FcεRγI expression.^16,17 We thus considered whether IL15 supplementation during ex vivo CART manufacturing induces greater FcεRγI expression in hBBζ CARTs than manufacturing with IL2. Surprisingly, FcεRγI expression trended lower in IL15 versus IL2-expanded CARTs (Figure S3C). Furthermore, T cells can produce IL15 and utilize trans-presentation to induce IL15 autocrine and paracrine signaling.^20,21 IL15 supplementation during hBBζ CART expansion caused complete ablation of surface IL15Rα expression (Figure S3D), the subunit of the IL15 receptor unique to IL15 trans-presentation. We thus evaluated overexpressing FCER1G in hBBζ CARTs to equalize starting FcεRγI expression across donors (Figure S3E). HBBζ and hBBζ-FcεRγI CARTs demonstrated similar cytolytic profiles and levels of IFNγ and TNFα production (Figure S3F–G), likely owing to the redundancy from hBBζ-signaling mediated induction of FcεRγI.

Syk-NFκB signaling mediates FcεRγI activity and promotes FcεRγI expression in hBBζ CARTs

FcεRγI signaling is propagated through a Syk-NFκB pathway. At baseline, Syk and FcεRγI do not associate, but after autophosphorylation or phosphorylation by Lyn, Syk activates and phosphorylates FcεRγI ITAMs.²² Phosphorylation of FcεRγI initiates a signaling cascade that activates the classical NFκB signaling pathway, resulting in nuclear translocation of NFκB subunit p65.^23,24 To determine the mechanism by which FcεRγI enhances canine CART cytotoxicity, we assessed Syk-NFκB signaling. We first compared phospho-Syk (p-Syk) levels in FcεRγI⁺ versus FcεRγI⁻ hBBζ CARTs. FcεRγI⁺ CARTs expressed more p-Syk than FcεRγI⁻ CARTs in baseline unstimulated conditions. P-Syk levels in FcεRγI⁺ CART rose in response to stimulation with CD20+, but not WT, K562 target cells (Figure 5A).

Figure 5. Syk-NFκB signaling promotes FcεRγI expression and activity in hBBζ CD8 CARTs.

(A) Representative histograms of p-Syk Y348 in FcεRγI+ versus FcεRγI− CD8+ hBBζ CART. (B) MFI of p-Syk Y348 in FcεRγI+ versus FcεRγI− CD8+ hBBζ CART after pre-treatment with DMSO or BAY61–3606. Statistical significance determined by 2-way ANOVA with with Tukey’s multiple comparisons test. (C) HBBζ CARTs were pre-treated with Syk inhibitor BAY61–3606 or DMSO vehicle control and evaluated for lysis of CD20+ K562 tumor cells at indicated time points. Values represent mean (SD) across technical quadruplicates. Significance of treatment groups versus DMSO control was determined by 2-way ANOVA with Dunnett’s multiple comparisons test. (D-F) Flow cytometric-based determination of the impact of Syk inhibition on (D) IFNγ production, (E) Granzyme B production (1-way ANOVA, Krusal Wallis Test), and (F) FcεRγI upregulation after stimulation. (G) FcεRγI expression of CD8+ hBBζ CARTs stimulated with K562-CD20 cells and pre-treated with BAY11–7082, which inhibits NFκB nuclear translocation, or DMSO control. Data are representative of 2–4 experiments and donors. Ns – not significant, *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001.

We determined the necessity of Syk signaling for FcεRγI-mediated enhancement of CART cytotoxicity by treating hBBζ CARTs with BAY61–3606, a highly selective Syk inhibitor,²⁵ or DMSO vehicle control. Pretreatment with BAY61–3606 significantly reduced p-Syk levels, the extent of which was higher in FcεRγI⁺ CARTs (Figure 5B). After 9 hours of co-culture with target cells, treatment with 2 μM BAY61–3606 inhibited CART cytolysis, whereas 4.5 μM completely abolished target cell lysis across a range of E:T ratios. After 20 hours of co-culture with 2–4.5 μM BAY61–3606, CART-mediated target cell lysis increased, although significant inhibition of cytolysis relative to DMSO-treated CARTs was observed at both concentrations (Figure 5C). IFNγ and Granzyme B production also decreased in response to treatment with BAY61–3606 (Figure 5D–E).

Notably, Syk inhibition caused FcεRγI expression levels after CAR stimulation to fall in a concentration-dependent manner (Figure 5F), even decreasing below pre-stimulation levels in response to 4.5μM Syk inhibition. We confirmed direct NFκB inhibition also reduced FcεRγI upregulation in response to CART-stimulation. Treatment with BAY11–7082, an inhibitor of IkBa phosphorylation, which is essential for p65 NFκB nuclear translocation, also reduced expression of FcεRγI in response to hBBζ CART stimulation in a concentration-dependent manner (Figure 5G). These data indicate that Syk-NFκB signaling not only mediates FcεRγI activity in hBBζ CARTs, but also promotes FcεRγI protein expression.

We next evaluated levels of NFκB p65 in canine CD8 CARTs to determine whether variable FcεRγI expression develops from differential NFκB signaling. Percentages of CAR⁺ T cells were similar in donor-matched hBBζ and cBBζ CARTs, although hBBζ CARTs displayed higher CAR MFI (Figure 6A). Unstimulated hBBζ CD8 CARTs expressed more p65 than cBBζ CD8 CARTs (Figure 6B). Following incubation with CD20+ K562 cells, both cBBζ and hBBζ CARTs experienced increases in expression of total and phosphorylated p65 (phospho-p65), although hBBζ CARTs displayed significantly higher levels (Figure 6B–C).

Figure 6. HBBζ CD8 CARTs demonstrate greater NFκB signaling than cBBζ CD8 CARTs.

(A) Representative histograms of donor-matched hBBζ and cBBζ CAR MFI versus NTD cells prior to NFκB analysis. (B and C) Flow cytometric determination of (B) total p65 and (C) phospho-p65 expression in resting versus stimulated CD8 CARTs. Representative of 2–4 experiments and 4 donors; ns – not significant, **p<0.01 by Mann-Whitney test. (D) Histogram of CAR expression in unstimulated cBBζ and hBBζ-expressing Jurkat NFκB reporter cells. (E) GFP expression in CAR+ NFκB-Jurkat reporter cells at rest and after co-culture with WT or CD20+ K562 cells. Significance determined by 2-way ANOVA with Sidak’s multiple comparisons test (n = 6); ns – not significant, ****p<0.0001.

To confirm the functional role of NFκB, cBBζ and hBBζ CARs were transduced into Jurkat NFκB reporter cell lines. Similar levels of CAR expression (Figure 6D) and comparable baseline NFκB activity (Figure 6E) were observed in Jurkat-NFκB reporter cell lines. Stimulation with CD20+, but not WT, K562 cells initiated NFκB signaling in both cBBζ and hBBζ-Jurkats, the extent to which was significantly greater in hBBζ-Jurkats (Figure 6E). Collectively, these data indicate that hBBζ-mediated signaling activates NFκB to induce FcεRγI expression.

HBBζ CARTs engage native CD3ζ to augment CAR-mediated NFκB signaling

In mouse CARTs, hBBζ expression enhances NFκB signaling through improved CAR-TRAF2 and TRAF3 interactions,¹⁴ and the magnitude of human CAR-TRAF2 associations dictates the strength of NFκB signaling.^26,27 To determine whether TRAF similarly mediates enhanced NFκB signaling in canine CARTs, we used protein L to immunoprecipitate the CAR through the extracellular scFv component, which was equivalent between constructs. However, in canine CARTs, expression of hBBζ did not enhance interactions with either TRAF2 or TRAF3 (Figure S4A).

Given the close structural homology between FcεRγI and CD3ζ,²⁸ we also considered whether FcεRγI differentially binds the CD3ζ component of hBBζ versus cBBζ CARs. However, FcεRγI did not co-immunoprecipitate with either CAR (Figure S4B). Additionally, FcεRγI did not heterodimerize with endogenous CD3ζ (Figure S4C).

Human 2^nd generation CARTs interact with endogenous CD3ζ to promote T cell receptor (TCR)-mediated NFκB signaling,²⁹ so we next evaluated whether canine CARs interact with native CD3ζ. At baseline, hBBζ CARTs exhibit significantly more interaction with total and phosphorylated CD3ζ than cBBζ CARTs, despite similar overall CD3ζ expression (Figure 7A–C). Inclusion of the human 4–1BB costimulatory domain mediated enhanced hBBζ-CD3ζ interactions (Figure S4D). CART-stimulation did not alter levels of hBBζ-CD3ζ interaction, but induced phosphorylation of CAR-bound CD3ζ (Figure 7B–C), indicative of TCR-mediated signaling. Both human signaling domains within the hBBζ CAR were necessary to confer maximal phosphorylation to native CD3ζ (Figure S4D).

Figure 7. HBBζ CARTs engage native CD3ζ to augment CAR-mediated NFκB signaling and stabilize CAR expression.

(A) Western blot evaluation of total CD3ζ and CAR protein levels in donor-matched cBBζ and hBBζ CARTs and NTD T cells. Representative of 3 blots. (B and C) Co-immunoprecipitation (Co-IP) Western Blot for determination of interactions between canine CARs and total or phosphorylated CD3ζ. (B) Blots are representative of 3 donors across 4–5 experiments with replicate values plotted in (C). *p<0.05, **p<0.01 by Mann-Whitney test. (D) Representative histogram and combined replicates quantifying MFI of surface CD3ε in donor-matched NTD T cells and CARTs. Representative of 5 experiments and donors. Significance computed by one-way ANOVA. (E and F) Surface expression of CD3ε, TCRαβ, and CAR at indicated time points after co-culture with K562-CD20 tumor cells. Expression was quantified as either MFI or percentage positivity and data are representative of 2 experiments and donors.

Notably, despite similar levels of CD3ζ expression (Figure 7A), unstimulated hBBζ CARTs express significantly greater surface CD3ε than either cBBζ CARTs or donor-matched non-transduced (NTD) T cells (Figure 7D). CBBζ CARTs exhibit comparable surface CD3ε expression as NTD T cells. We were thus interested in determining whether CART stimulation differentially affects surface stability of CD3ε and CAR in hBBζ versus cBBζ CARTs. We evaluated changes to surface protein expression at early (<1 hour) and late (≥1hour) time points after CART stimulation. Prior to stimulation, CD3ε and TCRαβ surface expression were greater in hBBζ than cBBζ CARTs (Figure 7E), despite comparable CAR expression (Figure 7F). Changes to CD3ε and TCRαβ surface expression followed similar trends in both groups of stimulated CARTs, although hBBζ CARTs maintained greater expression throughout (Figure 7E). Similarly, changes to hBBζ and cBBζ surface expression were mild and equivalent between groups at early stimulation time points (Figure 7F). However, by 1 hour after stimulation, surface cBBζ expression dropped in both percentage and MFI, while hBBζ expression was enriched for MFI and maintained a stable positivity percentage (Figure 7F). Superior maintenance of surface CAR and CD3ε expression after stimulation supports a role for hBBζ-CD3ζ interactions in both potentiating hBBζ-mediated NFκB signaling and stabilizing surface CAR expression.

Discussion

CAR costimulatory domains have previously been shown to differentially modulate human CART signaling and therapeutic efficacy,^12,13 although cellular mechanisms governing canine CART design are not well understood, and correspondingly, CART therapeutic efficacy in dogs has not yet been achieved. Here we show that human 4–1BB and CD3ζ CAR signaling domains confer enhanced cytotoxicity, persistence, and therapeutic efficacy to canine CARTs compared to CARTs engineered with either cBBζ or c28ζ signaling domains (Figures 1–2), dependent on the induction of an innate-like T cell gene signature characterized by FCER1G (Figures 3–4). The mechanism for FCER1G enhancement of canine CARTs depends on activation of Syk-NFκB signaling downstream of FcεRγI (Figures 5–6), as well as cooperative engagement of the native TCR through CAR interaction (Figure 7).

The cBBζ-CART design previously used in a first-in-canine CART clinical trial⁹ conferred inferior in vitro cytotoxic activity compared to hBBζ-CARTs during both initial and repetitive target cell encounter, associated with less CD8 CART expansion (Figure 1) and inability to eliminate GL1 canine B-cell leukemia in a murine xenograft model (Figure 2). In contrast, hBBζ-CARTs demonstrated superior in vivo cytotoxic activity, CD8 CART expansion, and persistence, resulting in the elimination of canine GL1 leukemia in a murine xenograft model (Figure 2). Collectively, these data establish the preclinical rationale for a future canine clinical trial of hBBζ-CARTs for canine B-cell lymphoma or B-cell mediated autoimmune disease. Given the modular design of CARs and development of canine scFvs for the re-direction of CARTs against many tumor antigens,^30–33 incorporation of hBBζ signaling domains may also enhance canine CART efficacy across a range of disease settings, underscoring the potentially broad therapeutic relevance of our findings to canine oncology and autoimmunity.

We acknowledge that preclinical CART studies in NSG xenograft models may not predict clinical outcomes in immunocompetent dogs with naturally occurring B-cell cancers or autoimmune diseases. NSG mice lack intact immune systems necessary for complete assessment of CART safety and efficacy. CRS, for example, arises from crosstalk between CARTs and myeloid cells such as macrophages, which are defective in NSG mice, and immunogenicity of human sequences in canine CARTs may mediate hBBζ-CART rejection in immunocompetent hosts. HBBζ demonstrates 83% sequence identity to cBBζ. However, murine CARTs bearing hBBζ domains exhibited greater in vivo persistence than m28ζ CARTs in an immunocompetent mouse model,¹⁴ despite only 76% sequence identity between hBBζ and mBBζ, suggesting that superior biological function outweighs risk of immunologic rejection. Additionally, clinically-approved anti-CD19 CARTs use a murine-derived FMC63 scFv ectodomain, which does not preclude therapeutic efficacy or long-term persistence. Nevertheless, the true test of hBBζ-CART function will occur in canine clinical trials, which will define their therapeutic properties in a naturally-occurring disease setting.

Transcriptional profiling revealed that hBBζ signaling induces an innate-like CD8 CART gene signature, characterized by high expression of FCER1G (Figure 3). An FCER1G^hi innate-like transcriptional profile has been reported to distinguish a potent cytotoxic subset of CD8 T cells in humans and mice.^16,17,19 In human CD8 T cells, the NKp30⁺NKp46⁻ subset is mostly FcεRγI⁺, and most FcεRγI⁺ cells are NKp30⁺, reflecting FcεRγI’s role as the signaling component for NKp30, although we show that canine NKp30⁺ CD8 T cells are mostly FcεRγI⁻ (Figure S3A–B). These data are consistent with single cell RNA sequencing data that define a subset of canine CD8 T cells that express NCR3 (which encodes NKp30) without FCER1G,¹⁸ suggesting that alternate signaling adaptors must couple with NKp30 in the dog. Our current study validates FCER1G as defining a highly cytotoxic subset of canine T cells, and additionally establishes a direct role for FcεRγI in CD8 CART anti-tumor function, which has previously not been validated in any species. We show that FcεRγI enhances canine hBBζ CART cytolysis and expression of granzyme B, IFNγ and TNFα (Figure 3F–G, 4D–E). We demonstrate that this phenotype is dependent on the induction of Syk-NFκB signaling (Figure 5C–E), which is also responsible for upregulating FcεRγI expression (Figure 5F–G). Notably, a recent report examining the expression profiles of human CART manufactured products for diffuse large B-cell lymphoma identified FCER1G as the most upregulated gene in CD8 CART manufactured products from patients who subsequently achieved complete remission after anti-CD19/CD20 tandem CART therapy.³⁴ Although the authors did not investigate the mechanism or therapeutic relevance of the finding, these data suggest that elevated FCER1G expression in pre-infusion CART products may predict clinical outcomes.

Prior to stimulation, hBBζ CD8 CARTs expressed significantly greater expression of p65 (Figure 6B) than cBBζ CD8 CARTs, indicating that expression of hBBζ primes CD8 T cells for enhanced NFκB activity. Pre-positioning of CARTs for greater NFκB signaling would allow faster conversion of CAR-stimuli into NFκB signals, which may explain the observed differences in kinetics and magnitude of hBBζ versus cBBζ-directed target cell lysis in vitro and in vivo (Figure 1C–E, Figure 2A–B). Heightened NFκB status may also explain improved CD8 persistence of hBBζ CARTs (Figure 2C), as 4–1BB-NFκB signaling has previously been shown to promote murine CD8 T cell survival through expression of Bcl-2 family anti-apoptotic proteins.^35,36 A similar mechanism has been established in human CARTs expressing 4–1BB co-stimulatory domains, where NFκB signaling promotes cell survival through enhanced Bim expression, a Bcl-2 family protein.³⁷

Surprisingly, increased TRAF engagement did not contribute to enhanced NFκB signaling in hBBζ CARTs (Figure S4A). This was unexpected since a prior study found that inclusion of human 4–1BB in murine CARTs augments NFκB signaling through improved CAR-TRAF binding,¹⁴ although this study was performed using HEK293 reporter cell lines, whereas our study was performed using primary canine CARTs. Additionally, human 4–1BB endodomains share 86% sequence identity with canine 4–1BB, but only 54% identity with mouse 4–1BB, with greater sequence variation within TRAF binding segments. Thus, inclusion of human 4–1BB in mouse CARs may substantially boost TRAF interactions, while in canine CARs, enhanced recruitment of TRAF is not the primary mechanism for downstream NFκB signaling. Instead, we find that hBBζ engages the native T cell receptor through physical interactions with CD3ζ to augment NFκB responses through TCR-mediated signaling (Figure 7B–C). Consistent with greater baseline NFκB expression, hBBζ-CD3ζ interactions are established prior to CART stimulation (Figure 7B–C). In addition to FcεRγI and CART-mediated NFκB signaling, the preexistence of CAR-CD3ζ interactions provides hBBζ CARTs a third pathway for amplifying the conversion of extracellular stimuli into NFκB signals, which enhances anti-tumor activity.

Prior studies associate the size of CAR constructs²⁹ or transmembrane domain composition³⁸ with CAR-CD3ζ interactions. In our studies, CAR sizes and transmembrane sequences are identical. In human CD8 T cells, expression of BBζ but not 28ζ CARs amplify endogenous TCR signaling.³⁹ Coupled with reports that 4–1BB recruits the linker for activation of T cells (LAT) to assemble into lipid rafts with endogenous TCR components,⁴⁰ the degree of which correlates with CART potency,⁴¹ future studies should investigate the possibility that human 4–1BB-expressing CARs may similarly recruit LAT to assemble with native CD3ζ in canine CARTs through lipid rafts. This possibility would also help explain the finding that despite comparable levels of surface CAR prior to stimulation, hBBζ CARTs exhibit higher surface CAR stability over time as compared with cBBζ CARTs (Figure 7F).

Future studies may also further investigate the biochemical interactions and subdomains allowing certain CAR designs to better integrate with native TCR machinery. Such studies will promote the design of CARs that exploit endogenous signaling mechanisms to potentiate CART responses. These findings may become particularly relevant to the field of allogeneic CARTs, which are commonly edited at the TRAC and B2M loci to eliminate TCRαβ expression. In this approach, loss of TCR expression prevents assembly with CD3 subunits, including CD3ζ, preventing the formation of CAR-CD3ζ interactions. This has been associated with impaired in vivo CART persistence,⁴² likely due to reduced NFκB signaling necessary for CART proliferation and/or survival. Thus, engineering strategies that either forcibly stabilize CAR-CD3ζ interactions or bypass the need for CAR interactions with native TCR for cytotoxic potency and long-term persistence hold promise to improve CART efficacy across a broad range of disease conditions.

To further investigate translational applications of our data, we considered manufacturing strategies to enrich for FcεRγI expression in canine CARTs. In human and mouse CD8 T cells, IL15 promotes FcεRγI expression.^16,17 However, IL15 supplementation does not enhance FcεRγI expression in canine hBBζ CARTs (Figure S3C–D), indicating that strategies to enrich for FcεRγI⁺ CARTs vary between species. We also show that overexpression of FcεRγI in hBBζ CARTs does not alter hBBζ cytotoxicity (Figure S3F–G), but significantly improves cBBζ cytotoxicity and IFNγ and TNFα production (Figure 4G–H). This finding is consistent with our data showing hBBζ but not cBBζ-mediated signaling induces FcεRγI expression and activity. Hence, forced FcεRγI expression in canine hBBζ CARTs may offer limited incremental benefit in an in vitro setting, although more refined temporal control of FCER1G expression, or FCER1G induction under conditions where potent cytotoxic efficacy is required, such as for solid tumors, may afford therapeutic benefit. Collectively, these data underscore the importance of future investigation into strategies that bolster FcεRγI-Syk-NFκB and/or CAR-TCR signaling in human and canine CARTs for improved therapeutic efficacy.

Materials and Methods

Co-Immunoprecipitation and Western Blotting

Lysate preparation

Donor-matched NTD or CAR T cells were co-cultured with K562-CD20 targets for 3 hours at a 4:1 T cell-to-K562 ratio. Cells were harvested and washed three times with cold PBS before resuspension in RIPA buffer supplemented with 1mM PMSF and protease and phosphatase inhibitors (CST 5872). 6e6 total T cells were prepared per Co-IP reaction. Samples were lysed for 20 minutes on ice, followed by sonication (5 pulses; 40% pulse, 30kHz) and centrifugation (13,000×g for 10min at 4°C). Supernatants were collected for co-immunoprecipitation and western blot analysis.

Protein L magnetic beads (MedChem Express HY-K0205) were used to immunoprecipitate CAR complexes from cell lysates. Beads were pre-washed four times in cold RIPA buffer and incubated with 90% of the whole cell lysate at 4°C overnight on a rotator. Immunocomplexes were captured by magnetic separation and washed five times with cold RIPA buffer supplemented with protease inhibitors. Complexes were eluted from beads with 2X laemmli buffer (BioRad 1610737) with 5% 2-Mercaptoethanol and boiled for 10 minutes at 95°C. Beads were removed by magnetic separation and supernatant was used for SDS-PAGE.

Non-Reducing Western Blot

Lysates were prepared as described above using Laemmli buffer prepared without 2-ME.

SDS-PAGE and Western Blot

Samples were run on a 4–20% Tris-HCl gel (BioRad 3450032) for 1.5 hours at 150V. Gel was transferred onto PVDF membrane at 0.35A for 80 minutes at 4°C. Membranes were blocked with 5% milk in PBS for 1 hour at room temperature prior to overnight incubation with primary antibodies (Table 1) at 4°C. The following morning, membranes were washed with PBS (3× 15 minutes at room temperature) and incubated with secondary antibody for 1.5 hours at room temperature. Membranes were washed again and developed using KwikQuant Ultra HRP Substrate Developer (Kindle Biosciences R1004) and KwikQuant Imager (Kindle Bioscience).

Table 1.

Antibodies used for Flow Cytometry and Western Blot experiments

Target	Clone	Fluorophore	Catalogue #	Company	Application
CD5	YKIX322.3	APC-eF780	47-5050-42	Invitrogen	FC
CD5	YKIX322.3	PE	12-5050-42	Invitrogen	FC
CD3e	CA17.2A12	AF700	MCA1774A700	BioRad	FC
CD3e	CA17.2A12	AF488	PFM UCDAVIS LABL		FC
CD3ζ	6B10.2	PE	12-2479-82	Invitrogen	FC
CD4	YKIX302.9	PB	MCA1038PB	BioRad	FC
CD8a	YCATE55.9	APC	17-5080-42	Invitrogen	FC
CD8a	YCATE55.9	SB600	63-5080-42	Invitrogen	FC
CAR (Rabbit anti-mIgG)	Polyclonal	AF594	315-585-003	Jackson	FC
CAR (Goat anti-mIgG)	Polyclonal	AF647	115-605-072	Jackson	FC
CAR (Goat anti mouse IgG)	Polyclonal	Biotin	115-065-146	Jackson	FC
CD62L	FMC46	PE	Sc-18918	Santa cruz	FC
CD45RA	CA21.4B3	Biotin	PFM UCDAVIS LABL		FC
Streptavidin		BUV395	564176	BD	FC
Mouse IgG2a	R19-15	BV711	744533	BD	FC
Mouse CD45	30-F11	AF594	563709	BD	FC
Human CD32	FUN-2	APC/Fire750	50-237-7612	Biolegend	FC
Fixable L/D	n/a	AF488	L23101	Invitrogen	FC
Fixable L/D	n/a	eF780	65-0865-14	eBioscience	FC
FcεRγI1	Polyclonal	FITC	FCABS400F	MilliMark	FC
TCRab	CA15.8G7	FITC	PFM UCDAVIS LABL		FC
p-Syk Y348	I120-722	PE	558529	BD Biosciences	FC
TNFα	MAb11	BV785	502948	Biolegend	FC
IFNγ	CC302	AF647	MCA1783A647	BioRad	FC
p65	532301	APC	IC5078A	R&D Systems	FC
p-p65 (Ser536)	93H1	PE	5733S	CST	FC
Granzyme B	GB11	RB780	568705	BD Biosciences	FC
NKp46 (CD335)	48A	Unconjugated	MABF2109	Millipore Sigma	FC
NKp30 (NCR3)	Polyclonal	Unconjugated	BS-2418R	Bioss	FC
Dk Fab2 anti-Rb IgG	Polyclonal	PE	12-4739-81	Invitrogen	FC
Annexin V	VAA-33	Biotin	BMS147BT	eBioscience	FC
CTY			C34567	Invitrogen	FC
CD215 (IL15Ra)	Polyclonal	Unconjugated	PA5-79467	Invitrogen	FC
P-CD3ζ Y142	EP265(2)Y	Unconjugated	Ab68235	Abcam	WB
CD3ζ	F-3	Unconjugated	Sc-166275	Santa-cruz	WB
TRAF3	G-6	Unconjugated	sc-6933	Santa-cruz	WB
TRAF2	12H7L9	Unconjugated	702256	Invitrogen	WB
FCER1G	Polyclonal	Unconjugated	PA5-28832	Invitrogen	WB
B actin	8H10D11	Unconjugated	3700	CST	WB
Anti-mouse IgG	Polyclonal	HRP	R1004	Kindle Biosciences	WB
Anti-rabbit IgG	Polyclonal	HRP	R1004	Kindle Biosciences	WB

FC: Flow Cytometry, WB: Western Blot.

Quantification of Co-IP

Western blot band intensities were quantified using ImageJ. Co-IP values reflects the amount of immunoprecipitated protein normalized to CAR.

Retroviral plasmids

The retroviral plasmid pMSGV1 (provided by Dr. Nicola Mason) was modified as follows: A gene fragment was synthesized (Integrated DNA Technologies) with flanking 5’ EcoRI and 3’ SalI restriction sites. The fragment encoded a kozak sequence upstream of a canine CD8α signal peptide, an anti-canine CD20 scFv with BspEI restriction site directly following, and a partially codon optimized sequence for canine CD8α hinge and transmembrane, CD137 costimulatory, and CD3ζ signaling domains. Gene fragments were digested with EcoRI and SalI, purified using a PCR purification kit (Wizard Promega), and ligated into the pMSGV1 vector to create pMSGV1.cCD20.cBBζ. To create additional constructs, gene fragments flanked by 5’ BspEI and 3’ SalI restriction sites were synthesized that encoded distinct combinations of human/canine costimulatory and CD3ζ signaling domains. These fragments were digested with BspEI and SalI and subcloned into the pMSGV1.cCD20.cBBζ vector.

To create a cBBζ FCER1G overexpression vector, a gene fragment flanked by 5’ NotI and 3’ SalI restriction sites was created encoding canine CD3ζ with a downstream P2A ribosomal skip site and full length FCER1G. This fragment was digested with NotI and SalI and subcloned into the pMSGV1.cCD20.cBBζ vector.

A gene fragment expressing full length FCER1G downstream of human CD3ζ and a P2A site was synthesized with flanking BmgBI and SalI restriction sites. This fragment was digested with BmgBI and SalI and subcloned into the pMSGV1.cCD20.hBBζ vector.

Retrovirus was produced by transfecting 293T cells with pMSGV1 CAR plasmid, packaging plasmid pGAG/POL, envelope plasmids pRD114 and pVSVg, and Lipofectamine 2000 (Life Technologies). Retroviral supernatants were collected 48 and 72 hours after transfection and passed through a 0.45 μm filter before use.

In vitro activation, expansion, and transduction of canine CAR T cells

Peripheral blood mononuclear cells (PBMCs) were obtained from leukapheresis products of canine donors housed at the University of Pennsylvania School of Veterinary Medicine. T cells were negatively selected from PBMCs as previously described¹⁵ and cultured in canine T cell culture media (IMDM; 10% FBS; 1% P/S; 1X NEAA; 1X sodium pyruvate; 1% Glutamax; 1x HEPES). To generate T cell activation beads, anti-canine CD3 antibody (Bio-Rad MCA1774) and recombinant human CD86-Fc protein (Biolegend 775606) were conjugated to M-450 Tosylactivated Dynabeads^™ (Invitrogen 14013) at a 1:1 molar ratio according to the manufacturer’s instructions. Canine T cells were activated with anti-CD3/rhCD86-Fc beads at a 3:1 bead:cell for 48 hours in media supplemented with 200 IU/ml rhIL-2 before retroviral transduction.

Transduction of canine T cells

To prepare viral-coated plates, polystyrene nontreated plates were coated with RetroNectin (Takara) overnight, then blocked with 2% BSA in PBS for 30 minutes at room temperature before centrifugation with viral supernatant at 2250×g for 2 hours at 37°C.

On day +2 after activation, canine T cells were added to the retroviral-coated plate and spinoculated at 700×g for 8 minutes. From days +3 to +12 post-activation, cells were expanded in canine media and maintained at 1e6 cells/ml. Fresh rhIL2 was added every 2–3 days with gradual tapering from 100 IU/ml and a final 25 IU/ml dose on day +7. Cells were counted using a Beckman Coulter Counter. Expression of CAR was detected ten days after activation by flow cytometry using anti-mouse IgG antibodies.

CART Restimulation, Expansion and Immunophenotyping

CAR T cells underwent multiple restimulation events by co-culturing CARTs 1:1 with irradiated GFP⁺CD20⁺ GL1 target cells. For each re-challenge event, cells were replated at a density of 1 × 10⁶/ml total cells in fresh media supplemented with 100 IU/ml IL2. Cells were also plated for cytotoxicity assays against non-irradiated target cells in parallel. Cells were expanded in canine growth media, and immunophenotypes were determined by flow cytometry at designated time points. Total cell counts were determined using a Beckman Coulter Counter. Live CD4+ and CD8+ CAR T cell counts were back-calculated using total cell numbers in combination with immunophenotypic data.

FCER1G Knockout

Canine CAR T cells were generated and expanded for 10 days to reach adequate cell numbers, at which point CRISPR-Cas9 was used to knockout FCER1G. Cells were washed three times with Opti-MEM before final resuspension in Opti-MEM with 10 mg Cas9 (TriLink CleanCap Cas9 5moU #L-7206–100) and 2 nmol sgRNA per 3× 10⁶ cells. Cells were electroporated using an ECM830 machine set to 500V and 700ms. 48 hours after electroporation, knockout efficiency was confirmed by flow cytometry and cells were plated in an in vitro killing assay. Mock conditions were electroporated without sgRNA. The sgRNA sequence was GGACGATACCGTACAGAAAC.

In vitro CART functional evaluation

Luciferase-based cytotoxicity assay

WT and caCD20⁺ K562 and GL1 cell lines expressing click-beetle green luciferase were obtained from Dr. Nicola Mason. Donor-matched CAR and NTD T cells were co-cultured in fresh T cell media with target cells at indicated effector-to-target (E:T) ratios. 4 hours after coculture, luciferase substrate (D-luciferin potassium salt, GoldBio) was added to each well and luminescence was measured using a plate reader (BioTek, Synergy HTX microplate reader) at indicated timepoints. The percentage of specific lysis was calculated using the luciferase activity of 5% SDS-treated cells as maximum cell death and media alone as spontaneous cell death using the formula: specific lysis (%) = 100 × ((experimental data − maximum death data)/(maximum death data − spontaneous death data)).

Cytokine profiling

Canine CARTs were co-cultured with WT or CD20+ K562 in fresh canine media at indicated E:Ts for 20 hours, at which point supernatants were collected and stored at −80C. Cytokines were assessed in duplicate using a custom formulation of the MILLIPLEX Canine Cytokine Magnetic Bead Kit to assess IFNγ, TNFα, and IL10. Assay was performed according to manufacturer’s instructions with assistance from the University of Pennsylvania Human Immunology Core.

Flow cytometric functional analysis

Donor-matched hBBζ and cBBζ canine CAR T cells were resuspended in fresh T cell media and co-cultured with K562-CD20 or K562-WT target cells at a 4:1 CART:K562 ratio. Co-cultures were performed for indicated time points. Cells were harvested and prepared for flow cytometric analysis of FcεRγI expression, IFNγ, granzyme B, and phospho-Syk levels. Cells were also evaluated for p65 and phospho-p65 expression.

Where appropriate, cells were stained for CAR and viability in FACS buffer (PBS^−/− with 1mM EDTA and 1% BSA) for 25 minutes at 4°C. Cells were washed twice and stained for additional surface markers for 25 minutes in FACS buffer at 4°C. For evaluation of intracellular receptors, cytokines, and cytoplasmic granules, cells were prepared using the eBioscience Fixation and Permeabilization Kit (88–8824-00) according to the manufacturer’s instructions. For evaluation of p65 and phospho-p65, cells were prepared using eBioscience Foxp3 staining buffer (00–5523-00). Cells were acquired using either a BD LSR II or BD FACSymphony A3 Lite Cytometer and analyzed with FlowJo software. Antibodies used for flow cytometry are described in Table 1.

Syk and NFκB inhibition studies

Syk inhibitor BAY61–3606 (MedChemExpress HY-14985) and NFκB inhibitor BAY11–7082 (Sigma-Adrich) were resuspended in DMSO to create 5mM stocks. Stocks were diluted with fresh media at the time of cell treatment. Cells were treated overnight. After treatment, cells were washed twice and resuspended in fresh media and used in downstream applications.

NFκB Jurkat Reporter

Jurkat NFκB-GFP reporter cells were retrovirally transduced to express cBBζ or hBBζ CAR. CAR-Jurkat cells were serum-starved overnight prior to assay. K562 tumor cells were stained with 5uM CellTrace Violet before incubation with serum-starved CAR-Jurkat cells at a 3:1 CART:K562 ratio for three hours. After co-culture, cells were immediately assessed for NFκB-GFP signal by flow cytometry.

In vivo canine CART evaluation using NSG xenograft models:

Target cell and T cell injection

NSG (NOD.Cg-Prkdc^scidIL2rg^tm1Wjl/SzJ) mice received 15mg intravenous immunoglobulin (Privigen, IVIg) via intraperitoneal (IP) injection on day −3 and day −1 before target cell injection to prevent Fc-mediated GL1–20 clearance. On day 0, NSG mice received 7 × 10⁵ GL1–20 target cells via tail-vein injection. Mice received additional IP IVIg on days +1 and +3.

Canine T cells were prepared as described, and on day +4 after target cell injection, donor-matched T cells (NTD, CD20.cBBζ, CD20.c28ζ, and CD20.hBBζ) were injected via the tail vein. NTD cells were used to standardize total T cell numbers between groups to account for differences in CAR positivity. 3.5 × 10⁶ CAR+ T cells or equivalent total NTD T cells were injected per mouse. Mice continued to be injected IP with IVIg every 2–3 days after treatment.

Bioluminescence imaging

Bioluminescence was measured with a Xenogen IVIS Lumina S3 (Caliper Life Sciences) from day +1 after target cell injection and every 2–3 days thereafter by injecting d-Luciferin potassium salt (Gold Bio) intraperitoneally at a dose of 150 mg/kg. Mice were anesthetized with 2% isoflurane and luminescence was measured until signals decreased in automatic exposure mode. Total flux was quantified using Living Image 4.4 (PerkinElmer) by drawing rectangles from head to the base of the tail. Radiance unit of p s⁻¹ cm² sr⁻¹ = number of photons per second per square centimeter that radiate into a solid angle of one steradian.

Flow cytometric analysis

At the time of sacrifice, mouse femurs were harvested and flushed for bone marrow cells. Red blood cells were removed with RBC Lysis Buffer (BioLegend). Remaining cells were washed with cold PBS and stained in FACS buffer on ice. Cell counts were back-calculated based on percentage muCD45⁻/GFP⁻/caCD3⁺/caCD5⁺ CD4⁺ or CD8⁺ cells.

CART gene expression profiling and analysis

Canine CAR T cells were generated and expanded for 10 days. On day 11 after activation, bulk CAR T cells (mixed CD4+/CD8+) were co-cultured 1:1 with irradiated WT or CD20+ GL1 target cells for 36 hours. Following co-culture, canine T cells were FACS sorted for live/CD5+/CAR+/CD8+ T cells. Sorted cells were immediately resuspended in Trizol, after which RNA was isolated, and quality was assessed. For each sample, 100ng RNA was hybridized for 19 hours using the nCounter Canine IO Panel and run on the nCounter SPRINT Profiler (NanoString Technologies). NSolver4.0 and ROSALIND were used for data QC, normalization, and differential expression analysis. Normalized counts are provided in Supplemental Table 1.

CD20 surface expression density

CD20 expression was determined by staining K562-CD20 and GL1-CD20 cell lines in parallel with canine PBMCs with anti-canine CD20 mAb (Invivogen, Clone 6C12) followed by PE-conjugated anti-mouse IgG2a (Southern Biotech, Clone SB84a). Canine PBMCs were pre-gated for live CD3⁻CD21⁺ lymphocytes to identify B cells. Samples were run alongside Quantibrite PE beads (BD) to determine the number of PE molecules per cell. Cell surface area was measured using a Beckman coulter counter, and PE density was defined as the number of PE molecules / cell surface area (μm²).

Data availability

All of the data are available in the main text or the supplementary materials.