NOTCH1 signaling during CD4⁺ T-cell activation alters transcription factor networks and enhances antigen responsiveness

NOTCH plays context-dependent roles in cell differentiation and stem cell self-renewal.^1,2 In T cells, NOTCH coordinates thymic T-cell development from common lymphoid progenitors,^3-6 facilitates gene expression critical for effector function,^7-9 and promotes memory CD4⁺ T-cell (T_MEM) survival following infection.¹⁰ Naïve T cells (T_N) upregulate NOTCH receptors after T-cell receptor (TCR) ligation and receive signals from antigen-presenting cells expressing NOTCH ligands.^11-13 In activated CD4⁺ T cells, NOTCH signaling promotes T-helper (T_H) subset-defining transcription factor (TF) and cytokine gene expression irrespective of polarizing cytokine signals, broadly enabling effector responses.^7,14-19 Depending on experimental context, NOTCH signaling has also been reported to skew differentiation toward various CD4⁺ T_H subsets.^14-16,20-22 In contrast to NOTCH’s role in T-cell effector function, one study reported that coculturing previously activated T_MEM with OP9 stromal cells overexpressing the NOTCH ligand DLL1 (OP9/DLL1) induced a less differentiated T-cell surface phenotype and enhanced T-cell proliferation upon restimulation.²³ These diverse effects suggest that manipulation of NOTCH signaling could be used to improve T-cell cancer immunotherapy²⁴.

Treatment of advanced hematologic malignancies with T cells engineered to express chimeric antigen receptors (CAR-T) targeting tumor-associated molecules is effective, but responses are often incomplete.^25,26 Both CD4⁺ and CD8⁺ CAR-T contribute to antitumor activity through direct killing and host immune cell activation in the tumor microenvironment.^27,28 In preclinical models, CAR-T derived from less differentiated subsets mediate superior antitumor responses,²⁸ and limiting tumor-specific T-cell differentiation during ex vivo manufacturing enhances efficacy.^29-35 Ex vivo polarization of CD4⁺ T-cell differentiation toward particular T_H subsets also improves antitumor responses in vivo.^36,37 These findings suggest that NOTCH signaling, whether delivered in contexts that restrict CD4⁺ differentiation or promote acquisition of particular T_H subset behavior, could improve CAR-T therapy.

The effects of NOTCH signaling in CD4⁺ T cells can vary with induction method as soluble NOTCH ligands, NOTCH-specific monoclonal antibodies (mAbs), and overexpression of transcriptionally active NOTCH intracellular domains in T cells provide signals of different quality, strength, and duration.³⁸ Additional membrane-bound and soluble signals available during cell-based delivery of NOTCH ligands further complicate interpretation; indeed, culturing CD8⁺ T cells in OP9/DLL1-conditioned medium was sufficient to induce phenotypic changes reported to be NOTCH-specific.³⁹ We avoided using cell lines and instead developed a culture system using αCD3/CD28 mAb-coated beads and plate-coated αNOTCH mAb to simultaneously activate CD4⁺ T cells and engage NOTCH receptors. Using this method, we studied the effects of NOTCH1 signaling on CD4⁺ CAR-T gene expression, differentiation, and antitumor responses.

Lenti-X cells (Takara) were cultured as described.⁴⁰ Raji, K562, and JeKo-1 cells (ATCC) were cultured and stably transfected as described.^40,41 

CAR vectors encoding the CD19-specific FMC63 single-chain variable fragment and truncated EGFR (EGFRt) transduction marker or the ROR1-specific R12 single-chain variable fragment and truncated CD19 (tCD19) transduction marker with either 4-1BB/CD3ζ or CD28/CD3ζ costimulatory domains were previously generated.^42,43 “CAR” denotes FMC63:4-1BB/CD3ζ:EGFRt unless otherwise stated. GFP was removed from the SGEP plasmid (#111170, Addgene) and replaced with a selectable truncated CD34 (tCD34) marker. Maf-targeting short hairpin RNAs (shRNAs) were designed using SplashRNA and cloned into SGEP-tCD34 as described.⁴⁴ Lentivirus was generated as described.⁴⁰ 

Peripheral blood was obtained from healthy donors who provided informed consent to participate in a blood donation protocol approved by the Fred Hutchinson Cancer Research Center Institutional Review Board. T-cell subsets were isolated using EasySep CD4⁺ Naïve, CD8⁺ Naïve, and CD4⁺ Memory T-cell negative selection kits (Stemcell).

Non–tissue-culture-coated plates were coated overnight at 4°C with phosphate-buffered saline (PBS) plus 5 μg/mL RetroNectin (Takara) and either 2.5 μg/mL Ultra-LEAF-Purified αNOTCH1 mAb (MHN1-519, BioLegend) for NOTCH1 (N1) culture, 2.5 μg/mL mouse immunoglobulin G1κ (IgG1κ) isotype (BioLegend) for control culture, or various concentrations of DLL1-Fc. Coating solution was aspirated and plates washed twice with PBS before T-cell plating. On day 0, 2 × 10⁵ T cells and 6 × 10⁵ Human T-Activator CD3/CD28 Dynabeads (ThermoFisher) in 1 mL T-cell culture medium including interleukin-2 (IL-2)⁴⁰ were plated per N1/control-coated 24-well. On day 1, T cells were transduced with CAR or shRNA lentivirus.⁴⁰ T cells were expanded in N1/control-coated wells for 7 days before transfer to uncoated flasks. DynaBeads were removed at day 5. CAR-T were analyzed on day 11. In some experiments, EGFRt⁺ CAR-T were enriched as described.⁴⁰ tCD34⁺ T-cells were enriched by CD34 microbead positive selection (Miltenyi).

Three hundred CD4⁺ EGFRt⁺ CAR-T cells were fluorescence-activated cell sorted (FACS) into SMART-Seq HT lysis buffer (Takara). Complementary DNA was generated using 13 amplification cycles and purified using AMPure XP beads (Agencourt). Libraries were prepared using the Nextera XT kit (Illumina) and sequenced in an SP flow cell on a NovaSeq 6000 (Illumina) to read depths > 10 × 10⁶ per sample. Transcripts were aligned using STAR⁴⁵ and analyzed for differential expression using DESeq2⁴⁶ and gene set enrichment analysis (GSEA).⁴⁷ 

For cocultures, tumors were irradiated (10 000 rad) and plated at 1.25 × 10⁴ cells per well in a 96-well U-bottom plate. For assays with recombinant human CD19 (rhCD19), 100 μL PBS plus 10 μg/mL avidin (Fisher) was added per well of a 96-well flat-bottom non–tissue-culture-coated plate, incubated overnight at 4°C, washed with PBS, blocked 1 hour with PBS plus 2% bovine serum albumin, and washed again. One hundred microliters of PBS plus biotinylated rhCD19 at the indicated concentrations was added per well, incubated for 30 minutes at 4°C, and washed again. For assays with OKT3, 100 μL PBS plus OKT3 was added per well, incubated overnight at 4°C, and washed with PBS.

For proliferation analyses, CAR-T were labeled in 0.25 μM carboxyfluorescein succinimidyl ester (CFSE) or CellTrace Violet (CTV) (ThermoFisher). 2.5 × 10⁴ or 7.5 × 10⁴ CAR-T were added to wells containing tumor cells or plate-bound rhCD19/OKT3, respectively. Twenty-four hours postplating, culture medium was analyzed for CAR-T cytokine production by Luminex, and CAR-T were fixed/permeabilized and stained for IL-2R receptor (IL-2R) and phosphoproteins or FACS-sorted for RNAseq. Separately, CFSE-labeled CAR-T were analyzed by flow cytometry for CFSE dilution after 72 hours of coculture with tumor. CTV-labeled CAR-T were transferred from rhCD19/OKT3-coated to uncoated plates after 24-hour stimulation and analyzed for CTV dilution 48 hours later.

6- to 8-week-old NOD.Cg-Prkdc^scidIl2rg^tm1Wjl/SzJ (NSG) mice were engrafted IV with 5 × 10⁵ CD19⁺/GFP⁺/firefly luciferase–positive Raji or JeKo-1 lymphoma. One week later, mice were treated IV with specified doses of CD19-specific CAR-T. CAR-T quantification in peripheral blood and tumor imaging were performed as described.^40,43 

Student paired and unpaired 2-tailed t tests, 1-way analysis of variance (ANOVA) with Dunnett posttests, and Mann-Whitney U tests were performed using Prism 9 (GraphPad). Raw cytokine values were log transformed prior to ANOVA analysis to achieve normal distribution. P values <.05 were considered significant.

We measured NOTCH receptor surface expression in human CD4⁺ T_N at rest and after activation with αCD3/CD28 mAb-coated beads and IL-2. Initial experiments were performed using CD4⁺ T_N to avoid potential differences introduced by variable T_N and T_MEM frequencies between donors. Unstimulated CD4⁺ T_N expressed low levels of NOTCH1 and no detectable NOTCH2, 3, or 4. Following activation, NOTCH1 was rapidly and uniformly upregulated and durably expressed, whereas NOTCH2 and NOTCH3 were upregulated more slowly and expressed transiently on a smaller fraction of cells (Figure 1A). These data suggested that NOTCH1 ligation would most uniformly induce NOTCH signaling early during in vitro activation.

We evaluated the effects of using plate-coated αNOTCH1 mAb to engage NOTCH1 during αCD3/CD28 stimulation of CD4⁺ T_N. T cells were cultured at low density on RetroNectin-coated plates to promote adhesion and provide mechanical stress required to induce NOTCH signaling.^48,49 This method allowed efficient lentiviral transduction of a CD19-specific CAR (Figure 1B). The NOTCH target genes Hes1 and Dtx1 were induced after CD4⁺ T_N activation with plate-coated αNOTCH1, and their expression was further increased by RetroNectin (Figure 1C). NOTCH1-induced Hes1 expression was abrogated by γ-secretase inhibition, indicating the effect was NOTCH-dependent (Figure 1D). T cells cultured on RetroNectin and αNOTCH1 mAb (hereafter “N1”) exhibited equivalent fold expansion and viability and a higher transduction rate compared with cells cultured on RetroNectin and IgG1κ isotype (hereafter “control”) (Figure 1E).

Coculture of previously activated murine ovalbumin-specific and human Epstein-Barr virus–specific T cells with OP9/DLL1 to activate NOTCH resulted induced a less differentiated stem cell memory (T_SCM) phenotype.²³ We evaluated the phenotypes of CD4⁺ T_N and N1 and control CAR-T derived from CD4⁺ T_N at the end of culture (supplemental Figure 1A). More N1 CAR-T expressed CD45RA and CD62L and fewer expressed CD45RO than control cells, consistent with the T_SCM phenotype (Figure 1F; supplemental Figure 1B). N1 and control cells expressed similar levels of CCR7, CD27, CD28, and TCF1 and uniformly expressed CD95, indicating transition from T_N to T_MEM (supplemental Figure 1B). Fewer N1 cells expressed the chemokine receptors CXCR3 or CCR4, which are associated with T_H1 and T_H2 differentiation, respectively, and are expressed by CD4⁺ T_SCM less frequently than by central (T_CM) or effector memory cells (supplemental Figure 1C).^50,51 The phenotypic effects of N1 culture were independent of CAR construct (supplemental Figure 1D).

Endogenous CD4⁺ T_SCM express lower levels of transcripts associated with T_H1, T_H2, and T_H17 effector responses than T_CM and effector memory cells.⁵¹ To investigate whether NOTCH-induced phenotypic changes were reflected at the transcriptomic level, we performed RNAseq of N1 and control CD4⁺ CAR-T at the end of culture. GSEA showed that N1 CAR-T expressed lower levels of T_H1-, T_H2-, and T_H17-characteristic genes compared with control CAR-T⁵² (Figure 1G; supplemental Table 1). Together, the enriched T_SCM phenotype and diminished CD4⁺ T_H effector gene expression indicate that NOTCH1 agonism during CD4⁺ T_N activation attenuates differentiation.

To compare specific NOTCH1 agonism to other strategies delivering nonspecific NOTCH signaling, we measured NOTCH target gene expression and surface phenotype in CD4⁺ T_N after activation and culture with RetroNectin and either αNOTCH1 mAb, IgG1κ isotype, or recombinant DLL1-Fc.⁵³ T cells cultured on saturating amounts of DLL1-Fc expressed Hes1 and Dtx1 levels and demonstrated CD45RA⁺ CD62L⁺ cell frequencies similar to N1 CD4⁺ T cells (supplemental Figure 1E-F).

We asked whether early NOTCH1 agonism mediated similar changes in CD4⁺ T_MEM by isolating CD45RA⁻ CD4⁺ T_MEM and evaluating NOTCH receptor expression at rest and following activation. T_MEM activation induced uniform and persistent NOTCH1 expression similar to T_N (supplemental Figure 2A), but N1 culture induced only a minor T_SCM population and little change in chemokine receptor expression (supplemental Figure 2B-C). We therefore focused further experiments on the effects of NOTCH1 signaling on CD4⁺ T_N.

NOTCH can both direct production of specific cytokines^11,54 and facilitate expression of master regulator TFs that drive T_H effector cytokine production in murine T cells.⁷ In response to restimulation with tumor, N1 CD4⁺ CAR-T produced more granzyme B (GzmB) and interferon γ (IFNγ) than control CAR-T, similar tumor necrosis factor α (TNFα), variable amounts of IL-2 and granulocyte-macrophage colony-stimulating factor depending on tumor stimulus, and less IL-5 and IL-13 (Figure 2A). N1 cells produced markedly more IL-10 than control cells and, uniquely, produced IL-22 (Figure 2A). N1 and control cells produced similar amounts of IL-4, IL-9, IL-17, and IL-21 (Figure 2A). The effects of NOTCH1 agonism on cytokine production were independent of CAR costimulatory domains (supplemental Figure 3A). N1 primary culture supernatants contained elevated levels of IFNγ, GzmB, IL-10, and IL-22 after 3 days, indicating that NOTCH1 agonism imparted a functional program early after activation that endured upon restimulation (supplemental Figure 3B). To determine whether IL-22 and IFNγ were produced by the same cells, we performed intracellular staining of N1 and control CD4⁺ CAR-T after restimulation with phorbol 12-myristate 13-acetate (PMA)/ionomycin or Raji tumor. IL-22 was expressed only by N1 cells and was produced by a population distinct from those making IFNγ (Figure 2B).

Differences in T-cell cytokine production can be driven by altered expression of TFs, particularly T_H subset-defining master regulators.⁵⁵ To profile TF and target gene expression, we performed RNAseq on N1 and control CD4⁺ CAR-T 3 days after activation, when cytokine differences were evident. N1 CAR-T were enriched for 18 genes, including the transcription factor Maf (c-MAF) (Figure 3A; supplemental Table 2). Though the differences did not rise to statistical significance, N1 CAR-T also expressed more Ahr (aryl hydrocarbon receptor, AhR), less Gata3 (GATA-3), and equivalent Tbx21 (T-bet) transcripts (Figure 3A). Intracellular staining confirmed these differences at the protein level (Figure 3B). Consistent with these findings, N1 cells expressed higher levels of AhR- and c-MAF–driven genes than control cells (Figure 3C).^56-58 

We next evaluated whether AhR and c-MAF influenced N1 CD4⁺ CAR-T cytokine production in response to tumor. The role of AhR was assessed by growing N1 CD4⁺ CAR-T in the continuous presence an AhR inhibitor (CH-223191), then restimulating cells with CD19⁺ tumor or PMA/ionomycin and assaying cytokine production in supernatants by Luminex or by intracellular staining. AhR inhibition abolished IL-22 production by N1 CAR-T and reduced IL-10 and GzmB but enhanced production of IFNγ, TNFα, granulocyte-macrophage colony-stimulating factor, IL-2, IL-4, IL-13, IL-9, and IL-17 (Figure 3D-E; supplemental Figure 3C). These data indicate a dual role for AhR, both driving IL-22, IL-10, and GzmB secretion and restricting production of other cytokines induced in CD4⁺ T_N by NOTCH signaling during activation.

To evaluate c-MAF, N1 CD4⁺ T-cells were transduced with a lentiviral vector encoding tCD34 and an shRNA either containing a scrambled sequence or targeting Maf. Two Maf-specific shRNAs both reduced c-MAF expression in N1 cells by half (supplemental Figure 3D). Scrambled and Maf-targeted N1 cells were isolated by CD34 selection, rested overnight, then restimulated with PMA/ionomycin. c-MAF knockdown reduced N1 cell IL-4 and IL-17 secretion and inhibited IL-10 production to an even greater extent than AhR antagonism but had no effect on GzmB, IFNγ, or IL-22 production (Figure 3F-G; supplemental Figure 3E). These data indicate that c-MAF, together with AhR, promotes IL-10 production in the setting of NOTCH signaling but does not limit IFNγ or IL-2 production as observed in other contexts.^58,59 

Less differentiated tumor-specific T cells demonstrate superior proliferation capacity and antitumor efficacy in vivo.^28,60 To assess whether NOTCH1 agonism altered CAR-T proliferative capacity, CFSE-labeled N1 and control CD4⁺ CAR-T were cocultured with CD19⁺ tumors and evaluated by flow cytometry over time. More N1 than control CAR-T divided after 30 hours of stimulation, and this early proliferation advantage persisted over 3 days (Figure 4A). ROR1-specific N1 4-1BB/CD3ζ or CD28/CD3ζ CAR-T also proliferated more than control CAR-T after coculture with ROR1⁺ tumor (supplemental Figure 4A), indicating the effect of NOTCH1 agonism on CD4⁺ CAR-T proliferation was independent of target antigen or costimulatory domain.

Activated T cells produce IL-2 and upregulate IL-2R, enabling signaling through STAT3 and STAT5 that drives proliferation.⁶¹ We measured IL-2R chain expression and STAT3 and STAT5 activation in N1 and control CD4⁺ CAR-T by flow cytometry after 24 hours coculture with CD19⁺ or CD19⁻ K562 cells. N1 CAR-T upregulated CD25, CD122, and CD132 and phosphorylated STAT3 Y705 and STAT5 Y694 to greater extents than control CAR-T in response to CD19⁺ but not CD19⁻ tumor (Figure 4B). RNAseq of N1 and control CAR-T cocultured with CD19⁺ K562 showed that N1 cells were strongly enriched for expression of genes involved in c-Myc, E2F and mTORC1 activity, and G2-mitosis progression compared with control cells^47,62 (Figure 4C; supplemental Tables 3 and 4). Together, these data indicated NOTCH1 agonism during CD4⁺ T_N activation conferred a cell state capable of an enhanced proliferative response to subsequent antigen encounter.

The increased proliferative capacity of NOTCH1-agonized CD4⁺ CAR-T could result from enhanced antigen recognition. To test this, CTV-labeled CD19-specific N1 and control CD4⁺ CAR-T were restimulated on plates coated with titrated rhCD19. At low rhCD19 densities, more N1 than control CAR-T had divided after 72 hour, and N1 cells proliferated more on average than control cells (Figure 5A). N1 CAR-T also produced more IFNγ, TNFα, and IL-2 and upregulated IL-2R chains and phosphorylated STAT3 to a greater extent than control CAR-T at low antigen densities (Figure 5B-C). Thus, NOTCH1 agonism heightens CD4⁺ CAR-T antigen sensitivity independently of any potential differential costimulatory or coinhibitory receptor engagement by tumor.

Although N1 and control culture yielded similar transduction efficiencies, N1 CD4⁺ CAR-T expressed higher levels of CARs and transduction markers than control CAR-T (supplemental Figure 4B), providing a possible explanation for heightened N1 CAR-T antigen sensitivity. To control for CAR expression, we sorted EGFRt-intermediate (EGFRt^INT) N1 and control CD4⁺ CAR-T with similar CAR expression (Figure 5D) and restimulated them on rhCD19-coated plates. EGFRt^INT N1 CAR-T produced more cytokine than EGFRt^INT control CAR-T (Figure 5E), indicating NOTCH1 agonism enhanced T-cell responsiveness to CAR signaling irrespective of CAR expression. We further asked whether this enhanced responsiveness translated to signaling through CD3, which was expressed equivalently by N1 and control CD4⁺ T-cells (supplemental Figure 4C). Mirroring responses to low rhCD19 levels, untransduced N1 CD4⁺ T-cells stimulated over a range of plate-bound αCD3 OKT3 mAb concentrations exhibited greater proliferation, IFNγ and TNFα production, IL-2R chain upregulation, and STAT3 phosphorylation than control CD4⁺ T-cells (Figure 5F-H). These data show that NOTCH1-agonized CD4⁺ T- cells are intrinsically capable of heightened responses to low antigen levels via both CAR and TCR signaling.

The phenotypic and functional characteristics imparted by NOTCH1 agonism suggested N1 CD4⁺ CAR-T might exhibit superior antitumor activity in vivo. NSG mice were engrafted with CD19⁺ Raji lymphoma, treated 7 days later with CD19-targeting N1 or control CD4⁺ CAR-T, and monitored for CAR-T expansion and changes in tumor burden (Figure 6A). N1 CD4⁺ CAR-T rose to markedly higher peak frequencies in the blood than control CAR-T, and analysis of paired blood, splenocytes, and bone marrow demonstrated higher N1 CD4⁺ CAR-T numbers at all sites (Figure 6B-D; supplemental Figure 5A). Neither N1 nor control CD4⁺ CAR-T expanded in non–tumor-bearing mice, indicating tumor recognition drove T-cell expansion and that NOTCH1 agonism did not promote T-cell transformation (supplemental Figure 5B).

RNAseq of N1 and control CD4⁺ CAR-T isolated from tumor-bearing mouse blood 8 days posttransfer showed that N1 cells expressed higher levels of genes involved in T-cell cytotoxicity, including Gzmb, Gnly, Prf1, and Nkg7 (Figure 6E; supplemental Table 5). N1 cells were also strongly enriched for gene signatures of cell cycle progression and c-Myc, E2F, and mTORC1 activity compared with control cells, similar to transcriptional differences observed after tumor recognition in vitro (Figure 6F; supplemental Table 6). GSEA further revealed strong skewing of N1 CAR-T toward T_H1-like and away from T_H2-like gene expression relative to control CAR-T (supplemental Figure 5C). In line with these transcriptomic findings, a higher fraction of N1 CAR-T harvested from marrow and restimulated with PMA/ionomycin produced IFNγ (supplemental Figure 5D). The pronounced T_H1-like N1 CAR-T effector response was accompanied by transiently enhanced tumor regression compared with control cells, but this effect was modest and did not persist (Figure 6G; supplemental Figure 5E). These data demonstrate that NOTCH1 agonism strikingly improves CD4⁺ CAR-T proliferative potential and skews cells toward a T_H1-like fate in vivo but does not enhance tumor control by CD4⁺ CAR-T alone.

Although CD8⁺ T cells are often the primary mediators of antitumor responses, CD4⁺ T cells can enhance, and in some cases are required for, CD8⁺ T-cell antitumor efficacy.^{27,28,42,63-65} The striking difference in expansion between N1 and control CD4⁺ CAR-T in tumor-bearing mice suggested that NOTCH1 agonism might improve the ability of CD4⁺ CAR-T to support CD8⁺ CAR-T proliferation and function. To test this, NSG mice bearing Raji lymphoma were treated with 3 × 10⁵ control CD8⁺ CAR-T and either 3 × 10⁵ N1 or 3 × 10⁵ control CD4⁺ CAR-T (Figure 7A). N1 CD4⁺ CAR-T rose to higher peak frequencies in the peripheral blood than control CD4⁺ CAR-T and promoted much greater expansion of control CD8⁺ CAR-T (Figure 7B). The enhanced CAR-T expansion in mice treated with N1 CD4⁺ and control CD8⁺ CAR-T drove rapid and complete tumor regression in all mice, whereas half of mice treated with control CD4⁺ and CD8⁺ CAR-T failed to regress tumor (Figure 7C; supplemental Figure 6A).

We then tested whether N1 CD4⁺ CAR-T could mediate enhanced helper function against CD19⁺ JeKo-1 lymphoma, which expresses lower levels of costimulatory molecules supportive of T-cell proliferation (supplemental Figure 6B). Mice were engrafted with JeKo-1 lymphoma and treated 7 days later with 2.75 × 10⁵ control CD8⁺ CAR-T and either 2.75 × 10⁵ N1 or 2.75 × 10⁵ control CD4⁺ CAR-T targeting CD19 (Figure 7D). Both CD4⁺ and CD8⁺ CAR-T rose to higher peak frequencies in mice treated with N1 compared with control CD4⁺ CAR-T (Figure 7E). All mice treated with N1 CD4⁺ CAR-T experienced curative antitumor responses, whereas most CD4⁺ CAR-T–treated mice failed to control tumor long-term (Figure 7F; supplemental Figure 6C). Thus, NOTCH1 agonism during ex vivo culture enhanced CD4⁺ CAR-T capacity to support CD8⁺ CAR-T proliferation and promote durable antitumor function in vivo.

CAR-T therapy effectively treats refractory hematologic malignancies,²⁵ but responses vary depending on CAR-T differentiation state, expansion, and persistence in vivo.^26,66-71 NOTCH signaling induced by coculture of previously activated human CD8⁺ CAR-T with OP9/DLL1 and IL-7 promoted a T_SCM phenotype and improved antitumor activity, in part by inducing FOXM1.³⁹ How NOTCH signaling alters human CD4⁺ CAR-T phenotype and function has not been studied. Here, we developed a cell-free culture system to simultaneously activate T cells and induce NOTCH signaling during CAR-T production. Culture on immobilized αNOTCH1 mAb strongly induced NOTCH target genes, demonstrating that existing GMP antibody manufacturing processes could be leveraged to deliver cell-free NOTCH signaling during clinical tumor-specific T-cell production. Contrasting prior work,²³ early NOTCH1 agonism efficiently induced a T_SCM phenotype in CAR-T grown from CD4⁺ T_N but not T_MEM, indicating that the mechanisms by which NOTCH signaling alters T-cell differentiation may be context- and T-cell subset-dependent.

NOTCH signaling during CD4⁺ T_N priming facilitates cytokine and master regulator TF expression, amplifying other polarizing signals.⁷ Under stimulation conditions conducive to efficient CAR-T production, NOTCH1 agonism enhanced IFNγ, GzmB, IL-10, and IL-22 production. Different populations of N1 CD4⁺ CAR-T produced IFNγ and IL-22, indicating that NOTCH signaling permitted distinct cell fates and might be used to generate diverse T_H subsets for different applications in adoptive cell therapy. The effects of NOTCH signaling on CD4⁺ T-cell cytokine production have often been studied through the lens of master regulator TFs; here, we observed that NOTCH instead controls cytokine production primarily by inducing AhR and c-MAF. AhR drove N1 CD4⁺ CAR-T production of IL-22, IL-10, and GzmB, in agreement with studies documenting its role transactivating these genes.^72-74 Moreover, AhR antagonism during N1 culture disinhibited production of other CD4⁺ effector cytokines, suggesting that NOTCH-induced AhR restrains overexpression of inflammatory cytokines during CD4⁺ T_H responses. In conjunction with findings that attenuation of AhR activation can improve CD8⁺ T-cell antitumor activity,⁷⁵ our data invite further inquiry into whether manipulating AhR activity might improve T-cell function in cancer immunotherapy.

NOTCH-induced c-MAF also promoted IL-10 production, consistent with a recent report⁷⁶ and c-MAF’s described cooperation with AhR in programming T regulatory type 1 (T_R1) characteristics.^77-79 N1 CD4⁺ CAR-T did not exhibit regulatory activity in xenogeneic lymphoma models, instead providing robust help to CD8⁺ CAR-T and promoting curative antitumor responses. N1 cells lacked key T_R1 traits, including CD49b and LAG3 coexpression, IL-21 secretion, and reduced IL-2 production.^78-80 However, it is possible that NOTCH1-agonized CD4⁺ T cells could mediate T_R1-like regulatory effects in an immunocompetent syngeneic setting, in which antigen-presenting cells play critical roles in T-cell antitumor responses. Our findings add to the known pathways through which NOTCH can alter CD4⁺ T-cell function and indicate that unbiased profiling strategies might identify additional novel behaviors initiated by NOTCH in other activation settings.

Unexpectedly, NOTCH1 agonism during CD4⁺ T_N activation rendered T cells more responsive to activation through the CAR or TCR. Restimulated N1 CD4⁺ T cells strongly upregulated expression of IL-2R chains and activated STATs and genes associated with T-cell activation and proliferation, translating into improved proliferation and T_H1 function in vivo. Although N1 CD4⁺ CAR-T did not control tumor more effectively, they provided markedly better help to cotransferred CD8⁺ CAR-T, supporting CD8⁺ expansion to higher frequencies in vivo and driving curative antitumor responses. N1 CD4⁺ CAR-T enabled potent CD8⁺ CAR-T antitumor responses at low doses, suggesting NOTCH activation during CD4⁺ CAR-T production might be leveraged to improve both therapeutic efficacy and safety.

In summary, NOTCH1 signaling provided concurrently with CD3/CD28 stimulation restricted differentiation during CAR-T production from human CD4⁺ T_N, generating a cell product that responded to low levels of antigen, proliferated robustly, and provided superior help to cotransferred CD8⁺ CAR-T. We identified AhR and c-MAF as transcriptional mediators of NOTCH-induced changes to CD4⁺ T-cell cytokine production, expanding the mechanisms by which NOTCH can shape T_H behavior beyond modulation of master regulator TF expression. Our data demonstrate that short-term antibody-based NOTCH1 agonism during ex vivo culture is a viable strategy for enhancing tumor-specific T-cell performance.

The authors thank S. Lee (Fred Hutchinson Cancer Research Center) for providing unmodified SGEP plasmid, as well as members of Shared Resources at the Fred Hutchinson Cancer Research Center for their help: E. Gad, D. Parrilla, L. King, and M. Jess for performing mouse experiments; R. Lawler and E. Griffiths for performing Luminex assays; C. Sather and D. Covarrubias for assistance with Aseq; and R. Reeves, B. Janoschek, B. Raden, and A. Berger for assistance with flow cytometry and FACS.

This work was supported by National Institutes of Health National Cancer Institute CA257088-02 (A.B.W.) and CA114536 (S.R.R.), by the Department of Defense BC190327P1 (S.R.R.), and by research funding from Lyell ImmunoPharma (S.R.R.). S.R.R. is an American Cancer Society research professor.

Contribution: A.B.W., E.C.F., M.J.P., G.O.C., S.M.S., and M.R.H. performed experiments; I.L. developed the plate-bound rhCD19 assay and profiled tumor cell costimulatory molecule expression; A.B.W. and S.N.F. analyzed RNAseq data; A.B.W., I.D.B., and S.R.R. conceived and designed the experiments; and A.B.W. and S.R.R. wrote the paper.

Conflict-of-interest disclosure: S.R.R. is a cofounder and shareholder of Lyell Immunopharma and has received grant funding and has intellectual property licensed to Lyell Immunopharma. S.R.R. was a founder of Juno Therapeutics, a Bristol Myers Squibb company, and has served as a scientific advisor to Juno Therapeutics and Adaptive Biotechnologies. I.D.B. has received grant funding from and owns shares of Lyell Immunopharma. M.J.P. has equity interest in Lyell Immunopharma, has received consulting income from SpringWorks Therapeutics, is currently employed by CellPoint B.V., and has equity interest in CellPoint B.V. The remaining authors declare no competing financial interests.

The current affiliation for M.J.P. is CellPoint B.V.

Correspondence: Stanley R. Riddell, 1100 Fairview Ave N, S2-207, Seattle, WA 98109; e-mail: sriddell@fredhutch.org.