The Notch signaling pathway: a potential target for cancer immunotherapy

The precursors of Notch receptors (pre-Notch receptors) are originally synthesized, their S1 portion is cleaved in the ER and Golgi apparatus, and then the cleaved Notch receptors are transported into the cell surface to further integrate with their ligands [6]. Previous studies showed that the inhibition of sarcoendoplasmic reticulum Ca²⁺-ATPase (SERCA) or zinc transporter impaired pre-Notch receptors synthesis, rendering them to be potential therapeutic targets [24, 25].

Curcumin, a natural phenolic compound that binds and inhibits SERCA, has been tested in pancreatic cancer [26], colorectal cancer (CRC) [27, 28], and prostate cancer (PCa) clinical trials. These results showed that oral administration of curcumin is generally safe and tolerated in CRC patients [27, 28], and in pancreatic cancer, two  out of 21 patients showed clinical biological activity and one of the two experienced transient but significant 73% tumor regression [26]. NVS-ZP7-4, an inhibitor that inhibits zinc transporter in ER, has been tested in T-ALL at the preclinical stage [16]. FLI-06, an inhibitor that inhibits the secretion of pre-Notch receptors before leaving ER, has been tested in esophageal squamous cell carcinoma (ESCC) at the preclinical stage [23]. In both in vitro and in vivo studies, these two inhibitors have shown inhibitory effects on the synthesis of pre-Notch receptors, which is worth future testing in the clinic. The results of each clinical trials and preclinical studies are shown in Tables 1 and 2.

After Notch ligands binds to NECD of Notch receptors, the extracellular subunits of Notch receptors are dissociated from their transmembrane subunits, resulting in NICD release and activation [6]. Some blocking antibodies that block the function of Notch receptors or ligands have been developed.

Blocking antibodies of Notch receptors: (i) OMP-52M51 (also called brontictuzumab), an anti-Notch1 monoclonal antibody (mAb), was tested in the clinic to treat lymphoid malignancies [29], solid tumor [30], adenoid cystic carcinoma (ACC) [31], and metastatic CRC. Overall, OMP-52M51 was well tolerated, exhibiting moderate anti-tumor activity with one partial response (PR) and two stable disease (SD) in twenty-four lymphoid malignancies [29], two PR and four SD in thirty-six (17%) assessable patients with a solid tumor [30], and one PR of one patient with Notch1-mutant ACC [31]. Diarrhea is the main toxicity after MP-52M51 treatment. (ii) OMP-59R5, an anti-Notch2/3 mAb, has been tested in clinical trials in solid tumor [32], stage IV pancreatic cancer [33], and stage IV small cell lung cancer (SCLC). Overall, the therapeutic effect of OMP-59R5 is not impressive. Either as a single agent or in combination with the first-line chemotherapy drugs (e.g., gemcitabine), OMP-59R5 did not improve overall survival (OS), progression-free survival (PFS), or objective response rate (ORR) of patients.

Blocking antibodies of Notch ligands: (i)–(ii) Rovalpituzumab tesirine (also called Rova-T) and SC-002, two anti-Dll3 mAbs, were each tested in multiple clinical trials, especially in SCLC (e.g., NCT01901653, phase I/II) [34‐42]. Rova-T has controllable associated toxicities. In the treatment of SCLC, Rova-T exhibited moderate clinical activity. The results of clinical trial NCT01901653 showed that eleven of sixty (18%) evaluable patients received confirmed objective responses, including ten of twenty-six (38%) patients with high Dll3 expression [36]. Thus, it seems that Rova-T exhibits encouraging single dose anti-tumor activity with controllable safety, especially in patients with high Dll3 expression. (iii) MEDI0639, an anti-Dll4 mAb, was tested on clinical trials for solid tumors [43], and (iv) demcizumab, an anti-Dll4 mAb, was clinically tested in non-small cell lung cancer (NSCLC) [44], pancreatic cancer, primary peritoneal carcinoma [45], and other solid tumors [46]. For MEDI0639, in twenty solid tumor patients, only one melanoma patient with PRs; seven patients had stable disease lasting more than 12 weeks [43]; for demcizumab, in forty-six treatment-naive patients with NSCLC, twenty of forty (50%) evaluable patients had objective tumor responses [44]. CTX014, an anti-Jagged 1/2 mAb, was tested in solid tumor at the preclinical stage, and results showed that CTX014 treatment overcame tumor-induced T cell tolerance, increased the infiltration of reactivated CD8⁺ T cells into tumors, and enhanced the efficacy of T cell–based immunotherapy [17]. The results of each of clinical trials and pre-clinical studies are summarized in Tables 1 and 2.

Two important cleaving enzymes, a disintegrin and metalloprotease (ADAM) and γ-secretase, catalyze the cleavage of Notch receptors [6]. Specifically, ADAM promotes the cleavage of NECD from the transmembrane (TM) NICD domain (S2 cleavage), while γ-secretase promotes the release of NICD from the TM domain (S3 cleavage), thereby achieving nuclear translocation [6, 12]. Therefore, ADAM and γ-secretase are important targets in blocking Notch signaling.

ADAM inhibitors: INCB7839 (also called Aderbasib), a small molecule drug targeting ADAM, has been proposed for a phase I clinical trial of high-grade gliomas (NCT04295759). This clinical trial is recruiting, and no results have been reported yet. The curative effect of other ADAM inhibitors (e.g., ZLDI-8) was tested in hepatocellular carcinoma (HCC)-bearing mice. Results showed that ZLDI-8 significantly inhibited tumor growth [18].

γ-secretase inhibitors (GSIs): since 2004, at least six types of γ-secretase inhibitors have been clinically tested in various cancer patients. The details are as follows: (i) MK0752 was tested in clinical trials designed for leukemia, advanced breast cancer (BC) [47], metastatic BC [48], pancreatic cancer [49], or other solid tumors [47]. Overall, the main toxic side effect of this drug was diarrhea. But some patients also showed positive treatment reactions. In high-grade gliomas patients, one patient completely response and additional ten patients remained stable for more than four months to MK0752 treatment in one hundred and three patients in total [47]. Among forty-four eligible pancreatic cancer patients, thirteen patients achieved stable disease after MK0752 was combined with first-line chemotherapy drug gemcitabine, and one patient achieved a confirmed PR, indicating that MK0752 has the potential to be used in combination with first-line chemotherapy drugs [49]. (ii) RO4929097 has conducted clinical trials in BC, sarcoma [50], melanoma [51], adult solid neoplasm, and other solid tumors [52]. Overall, only one of thirty-two metastatic melanoma patients treated with RO4929097 achieved PR. Although RO4929097 is well tolerated, but it has significant toxicity. (iii)–(v) LY900009 in advanced cancer [53], PF-03084014 in triple-negative breast cancer (TNBC), and LY303947 in solid tumors [54‐56] have also been tested. Overall, the clinical treatment effects of these three drugs were not impressive, and participants showed limited clinical responses. (vi) AL101 has been studied in clinical trials of TNBC and in adenoid cystic cancer. These two clinical trials are recruiting, and no results have been reported yet. DAPT has been tested in head and neck squamous cell carcinoma (HNSCC) at the pre-clinical stage. Results showed that DAPT decreased tumor burden in a mouse model after prophylactic treatment [19]. The results of each clinical trials and preclinical studies are shown in Tables 1 and 2.

When activated NICD enters the nucleus, NICD binds with RBP-J and MAML to form a transcriptional complex, recruiting co-activators and triggering the transcription of Notch target genes [11, 13]. Therefore, targeted inhibition of the Notch transcription complex can also be an effective approach to block Notch signaling.

Notch transcription complex inhibitors: CB-103, the first drug to effectively control the Notch transcription complex, has been studied in advanced tumors and hematological malignancies (NCT03422679) in a phase I/II clinical trial. Results showed that CB-103 was well tolerated in cancer patients. The curative effect of the other two inhibitors, SAHM1 and IMR-1, has been tested in leukemic cells and an esophageal adenocarcinoma (EAC) patient-derived xenograft tumor model, respectively. Results showed that SAHM1 suppressed genome-wide suppression of Notch-activated genes in leukemic cells [20], and IMR-1 inhibited the growth of Notch-dependent EAC patient-derived xenograft tumors [21]. The results of each of clinical trials and pre-clinical studies are shown in Tables 1 and 2.

Collectively, among drugs that targeting Notch signaling, blocking antibodies of Notch ligands (e.g., anti-Dll3 mAb) and γ-secretase inhibitors (e.g., MK0752) have demonstrated encouraging therapeutic effects in clinical trials. Unfortunately, the therapeutic efficacy of other drugs does not seem to meet expectations, and further research is needed.

Natural killer (NK) cells are crucial anti-viral and anti-tumor cells in the innate immune system [61, 62]. Early studies have found that Notch signaling plays an indispensable role in regulating their development and effector functions. For example, human umbilical cord blood (UCB) CD34⁺ precursors become committed to differentiate into NK cells in vitro after they are stimulated with Notch ligands (mainly Dll1, Dll4, and Jagged2) in the presence of cytokines [e.g., interleukin 7 (IL-7), Fms-like tyrosine kinase 3 (Flt3) ligand, and interleukin 15 (IL-15)]. These NK cells were able to lyse tumor cells because they upregulate their transcription and release of granzyme B (GZMB) and interferon-gamma (IFN-γ) [63‐65]. As human NK cells mature, Dll1-mediated Notch signaling is activated to promote the expression of CD16 and killer Ig-like receptors (KIRs), resulting in cytotoxicity against tumor cells [66]. In human peripheral blood and decidual NK cells, activation of Notch1 and/or Notch2 by Dll1 and/or Dll4 promotes IFN-γ secretion [65]. Compared with normal human NK cells, Zakiryanova et al. found that the expression of Notch1 was significantly decreased in NK cells of patients with lung cancer or gastric cancer, while Notch2 was significantly reduced in patients with gastric cancer but not in those with lung cancer [67]. In murine NK cells, Kijima et al. found that DC-mediated NK cell activation was controlled by the interaction of Notch with Jagged2. Enforced expression of Jagged2 in DCs significantly enhanced the cytolytic effects of murine NK cells against YAC-1 cells by activating the NK cells’ Notch signaling [68]. Enforced expression of Jagged2 in A20 cells (a BALB/c-derived B cell lymphoma cell line with low expression of Jagged2) also enhanced the cytolytic efficacy of murine NK cells against A20 cells in vivo and in vitro [68]. Together, these observations suggest that Notch activation can significantly enhance the anti-tumor properties of NK cells. Therefore, targeted activation of Notch signaling in NK cells might be a promising strategy for enhancing NK cell therapy (Fig. 2A).

Innate lymphoid cells (ILCs) are newly discovered and defined lymphocytes involved in regulating innate and adaptive immune responses. They govern immune responses against viruses, intracellular pathogens, helminths, and tumors [69‐71]. ILCs are widely distributed in various tissues and organs (e.g., liver, lymph nodes, small intestine lamina propria, and other mucosal tissues), and the various murine ILC lineages are distinguished by differences in transcription factor profiles and cytokine production. For example, T-bet [encoded by T-box transcription factor 21 (Tbx21)]-expressing group 1 ILCs (ILC1s) secrete mainly IFN-γ; GATA3-expressing group 2 ILCs (ILC2s) secrete mainly IL-5 and IL-13; and RORγt-expressing ILCs (ILC3s) secrete mainly IL-22 and IL-17. Murine ILC3s can be further divided into NKp46⁺ ILC3s, NKp46⁻ ILC3s, and lymphoid tissue-inducer cells (LTi cells) [72, 73].

During the past decade, researchers have uncovered multifaceted roles for Notch signaling in ILC subsets. In 2011, Possot et al. cultured murine bone marrow (BM) common lymphoid progenitors (CLPs) on OP9-Dll4 stroma (to activate Notch signaling) or in the presence of DAPT (to inhibit γ-secretase and therefore Notch signaling) and found that the maturation of adult BM-derived RORγt⁺ ILCs (also defined as ILC3s) was Notch2-dependent manner [74]. A subsequent study of murine gut ILC22 cells (now known as ILC3s), which include NKp46⁺ILCs [CD3⁻NKp46⁺NK1.1^lo–negRORγt⁺ cells (also defined as NKp46⁺ ILC3s)] and LTi cells, showed that Notch signaling was crucial for the downstream signaling of aryl hydrocarbon receptor (AhR) during the generation of murine NKp46⁺ ILCs (likely ILC3). In contrast, LTi-like cells were partly dependent on Notch signaling [75]. Compared with WT mice, RBP-Jκ-CD mice (conditional deletion of RBP-Jκ expression in the hematopoietic compartment) had considerably fewer NKp46⁺ ILCs in the lamina propria [75]. Mechanistically, the binding of AhR ligands on AhR promotes the translocation of AhR into the nucleus, where it binds to regulatory sites and promotes the expression of Notch receptors (mainly Notch1, Notch2), enhances IL-22 secretion, and ultimately sustains the NKp46⁺ ILC population and partly sustains LTi-like cells in small intestine (SI) lamina propria [75]. A subsequent study by Rankin et al. demonstrated that T-bet-mediated NKp46⁺ ILC development could also be achieved via Notch (mainly Notch1, Notch2) signaling [76]. Specifically, after being exposed to Dll1 for 9 days, murine SI lamina propria Rorc(γt)^+/GFP Tbx21^+/+ LTi cells generated NKp46⁺ ILCs, whereas Rorc(γt)^+/GFPTbx21^−/− LTi cells did not generate that subset in vitro. This suggests that Notch signaling plays an integral role in the T-bet-mediated transition of LTi cells into NKp46⁺ ILCs [76]. In addition, in murine SI lamina propria, RBP-J-mediated Notch2 signaling contributed to the transition of NCR⁻ ILC3 precursors (NKp46⁻ ILC3s) into NCR⁺ ILC3s (NKp46⁺ ILC3s) in a cell-autonomous manner. Mechanistically, activation of RBP-J-mediated Notch2 signaling mainly stimulates the expression of genes encoding transcription factors, such as T-bet, AhR, and RORγt [77]. These murine studies support the notion that Notch signaling regulates ILC3 plasticity by controlling the fate of NKp46⁺ cells. In human ILCs, researchers have also found that, in combination with IL-7, Notch signaling induces the differentiation of hematopoietic progenitor cell subpopulation one [HPC-1, CD45RA (RA)⁻Flt-3⁺c-Kit^hi cells] into NKp44⁺ ILC3s [78]. Thus, Notch signaling plays an essential regulatory role in the development and phenotypic transition of ILC subsets, mainly ILC3s. However, our understanding of the role of Notch signaling in ILC-mediated immune responses in cancer is still preliminary, as even ILCs’ role in cancer is not yet very clear. We recently started to elucidate the role of ILCs in cancer [79, 80]. Therefore, further studies that explore the role and regulatory mechanism of Notch signaling in immune responses mediated by ILCs are warranted (Fig. 2B).

Macrophages are specialized phagocytes in innate immunity. As one of the first responders to infection, they recognize and degrade tumor cells [81]. In the TME, macrophages are extremely plastic, and their interaction with tumor cells and/or the stromal microenvironment usually polarizes M1-like tumor-associated macrophages (M1-TAMs) into M2-like tumor-associated macrophages (M2-TAMs) [82]. Generally, M1-TAMs promote anti-tumor inflammatory responses and exert anti-tumor effects [82, 83]. In contrast, M2-TAMs are involved in neovascularization [84] and matrix deposition and remodeling [85], and they participate in immunosuppression, promoting tumor growth [84]. Therefore, promoting the polarization of macrophages into M1-TAMs or reversing M2-TAMs into M1-TAMs are key strategies for targeting macrophages in cancer immunotherapy [82, 86, 87].

In recent years, our research has suggested that Notch signaling plays a crucial regulatory role in switching macrophage phenotypes and thus in remodeling the TME [88‐91]. Specifically, some Notch signaling molecules (e.g., Notch1, Notch2, Hes1) were expressed higher in M1-TAMs than in M2-TAMs in a B16F10 melanoma in vivo model. Forced activation of Notch signaling by co-culture with OP9-Dll4 cells promoted anti-tumor activity by polarizing macrophages into IL-12-producing M1-macrophages but not into M2-macrophages. Mechanistically, knockout of RBP-J-mediated Notch signaling inhibited M1 polarization by inhibiting LPS-induced suppressor of cytokine 3 (Socs3) expression [88]. Using NIC transgenic mice controlled by Lyz2-Cre (NIC^CA), we subsequently showed that forced activation of Notch signaling in macrophages in vivo repressed tumor growth while diminishing TAM phenotypes. Mechanistically, miR-125a has been identified as a key downstream miRNA of RBP-J-mediated Notch signaling activation. Overexpression of miR-125a promoted M1 polarization and suppressed M2 polarization, boosting anti-tumor activity [89]. In addition, signal regulatory protein α (SIRPα), a key inhibitor of macrophages, was identified as the key downstream molecule of RBP-J-mediated Notch signaling. Notch activation repressed SIRPα expression through the Hes family co-repressors and then enhanced tumor cell lysis partly by promoting polarization into the M1 phenotype. Soluble mSIRPα^ext polypeptides, which possess the extracellular domains of mouse SIRPα, promoted M1 polarization and increased phagocytosis of tumor cells by macrophages [90]. This study indicates that specifically activating Notch signaling to inhibit the SIRPα-CD47 axis might be a promising strategy for releasing macrophages from phagocytic inhibition. We recently generated a type 1 herpes simplex virus-based oncolytic virus (oHSV) that expresses a full-length anti-CD47 antibody (αCD47) to block the CD47 ‘don’t eat me’ signal. This engineered virus suppressed tumor growth in both glioblastoma and metastatic ovarian cancer models, partly by promoting the M1 polarization of macrophages [92, 93]. In the TME of murine orthotopic HCC, myeloid-specific RBP-J knockout significantly promoted the growth of orthotopic tumors [91]. Compared with control mice, the infiltration of CCR2-independent TAMs—mainly Kupffer cell-like TAMs (kclTAMs) but not monocyte-derived TAMs (moTAMs)—in the liver was significantly higher in RBP-J knockout mice [91]. Mechanistically, RBP-J deficiency in myeloid cells impeded the differentiation of moTAMs, but promoted the proliferation and pro-tumor cytokine [e.g., interleukin 10 (IL-10)] production of kclTAMs by upregulating WNT-β-CATENIN signaling, and then accelerating the progression of murine orthotopic HCC [91]. Together, these findings suggest that intrinsic activation of Notch signaling promotes the M1 polarization and suppressed M2 polarization of macrophage to boost anti-tumor activity, while intrinsic inhibition of Notch signaling promotes the proliferation of kclTAMs to boost pro-tumor activity (Fig. 2C).

Myeloid-derived suppressor cells (MDSCs) are another major immune response modifier in cancer, as they interfere with immune responses against tumors and facilitate tumor metastasis and angiogenesis [94]. According to the differences in cell surface markers, MDSCs can be divided into different subtypes. Granulocytic-MDSCs [G-MDSCs or polymorphonuclear (PMN)-MDSCs] and mononuclear MDSCs (M-MDSCs) are two important immunosuppressive subsets [95, 96].

Recently, Wang et al. showed that Notch signaling was significantly inhibited in PMN-MDSCs of tumor-bearing mice [97]. Compared with MDSCs of control mice, the MDSCs (mainly PMN-MDSCs) of mice with specific knockout of RBP-J in myeloid cells were significantly less immunosuppressive. Mechanistically, knockout of RBP-J inhibits the signal transducer and activator of transcription 3 (STAT3) signaling and reduces the inhibition capability of PMN-MDSCs on the proliferation and activation of allogenic T cells, while the deficiency of the Notch signaling has not much effect on M-MDSC [97]. Therefore, blocking RBP-J-mediated canonical Notch signaling, specifically in PMN-MDSCs, might be an ideal strategy for inhibiting tumor progression [97]. Using Cybersort and Gene Set Enrichment Analysis (GSEA), Otani et al. analyzed TCGA database and revealed that a higher Notch score positively correlated with M-MDSC recruitment in glioma patients [98]. Mechanistically, treating mice bearing intracranial glioma with oHSV induced Jagged1 expression on macrophages. These Jagged1-presenting macrophages spread Notch activation in the TME, especially in TAMs. TAMs with Notch activation induce the secretion of CCL2, further amplifying M-MDSCs recruitment and attenuating anti-tumor immune response of T cells [98]. Blockading Notch signaling with GSI (γ-secretase inhibitor) significantly reduced the M-MDSC-mediated immunosuppressive TME and activated CD8⁺ T cell-dependent anti-tumor memory response [98]. A study from our group showed that activating Notch signaling in murine myeloid cells significantly inhibited tumor progression. Activated NICD inhibited lactate import 2 (MCT2) expression via Hes1, thus reducing lactate intake in myeloid cells. Activated NICD also promoted the differentiation of M-MDSCs into M1-type TAM but not into PMN-MDSCs in the TME [99].

Together, these studies highlight the complex roles of Notch signaling in the differentiation of different MDSC subtypes, suggesting that targeting activation or inhibition of Notch signaling in MDSCs for cancer treatment must be context-dependent. However, oncolytic virotherapy combined with Notch blockade may be a promising strategy for synergistically inhibiting the immunosuppressive function of M-MDSCs and thereby enhancing therapeutic benefits in glioma or other tumors that respond positively to inhibition of Notch signaling (Fig. 2D).

Dendritic cells are professional antigen-presenting cells (APCs) that can efficiently intake and process antigens, and then present them to T cells, leading to activation of adaptive immune responses against pathogens and tumors [100]. Meng et al. identified a new human DC subset that highly expresses the Notch ligand Dll4 (Dll4⁺ DCs) [101]. Compared with Dll4⁻ DCs, these Dll4⁺ DCs can better promote the differentiation and expansion of T helper (Th) cells (e.g., Th1 and Th17 cells) and effector CD8⁺ T cells because they upregulate the transcription of differentiation-related transcription factors (e.g., GATA3, T-bet, RORC) and the production of anti-tumor effector cytokines (e.g., IL-4, IFN-γ, and IL-17). This suggests that high levels of Dll4 in DCs indicate the high anti-tumor potential because of upregulated antigen presentation and adaptive immune responses [101, 102].

In addition to Notch ligands, we found that Notch receptors and their downstream effectors are essential for DCs’ effector function. Thus, activation of RBP-J-mediated Notch signaling was critical in DC-dependent anti-tumor immune responses [9]. Compared with murine RBP-J^+/− DCs, RBP-J^−/− DCs (specific knockout of RBP-J in DCs) lost inhibition of tumors (e.g., B16F10 melanoma, H22 hepatoma, and Lewis lung carcinoma) in vivo because DC migration and antigen presentation to T cells were inhibited [9]. During the progression of colitis-associated CRC, mice whose DCs were deficient in Notch signaling were more susceptible to the disease than mice with normal DCs [103]. In contrast, adoptive transfer of Notch-primed DCs in mice restrained the progression of inflammation-associated CRC [103]. Mechanistically, chemokine receptors [mainly CC-chemokine receptor 7 (CCR7)] of DCs were identified as a critical downstream component of RBP-J-mediated Notch2 signaling, and upregulation of CCR7 mediated by activated Notch2 signaling facilitated DC migration and cross-presentation of antigens to CD8⁺ T cells [103]. Kirkling et al. found that Notch signaling facilitated the differentiation and CCR7-dependent migration of conventional DCs (cDCs) and then promoted their cross-presentation of antigens to T cells [104]. In addition, Notch signaling can be activated in DCs by a polysaccharide [Lycium barbarum polysaccharide (LBP)] and can then induce the phenotypic and functional maturation of DCs to promote DC-mediated cytotoxicity of T lymphocytes (CTLs) [105]. In general, Notch signaling is a positive regulator of DC maturation, antigen presentation, and adaptive immune responses, but the specific regulation mechanisms remain to be explored (Fig. 2E).

T cells, especially CD8⁺ T cells, are well known for their cytolytic effects that require prior sensitization during adaptive immune responses [106]. Previous reports indicated that activation of Dll1-mediated Notch signaling (mainly Notch2 signaling) promoted the differentiation and cytolytic function of murine T cells both in vitro and in vivo [107]. By using a Notch2^f/fE8I-Cre⁺ mouse model (lacking Notch2 expression in peripheral CD8⁺ T cells but not in CD4⁺ T cells), the authors found that the knocking out of Notch2 inhibited (compared to the control) the differentiation of naive CD8⁺ T cells into CTLs and could not control the growth of OVA-expressing EG7 thymoma cells and EG7 cells in vivo. These observations indicate that Notch2 is crucial for the anti-tumor response of CTL cells [107, 108]. Using a ChIP assay to explore the mechanism, the authors found that a complex of activated NICD, phosphorylated-CREB1, and transcriptional coactivator p300 bound to the promoter of the Gzmb gene, enhancing its transcription [107, 108]. Of note, in vitro and in vivo tumor models (e.g., breast adenocarcinoma, lung cancer, thymoma) and three follow-up studies also demonstrated that activation of Notch signaling significantly enhanced the anti-tumor and/or anti-tumor memory capacity of CD8⁺ T cells by promoting the expression of IFN-γ and GZMB. In these studies, Notch signaling was activated by (i) using mice whose CD8⁺ T cells contained a specifically activated NIC (Notch1 intracellular domain) [109], (ii) treating CD8⁺ T cells with the proteasome inhibitor bortezomib [110], or (iii) treating CD8⁺ T cells with the Notch ligand Dll1 [111]. However, in a different TME (e.g., in murine HCC or ovarian cancer), Notch signaling was regulated by a series of genes [e.g., transcriptional and immune response regulator (1810011O10 Rik, also known as TCIM) and enhancer of zeste homolog 2 (EZH2)] to modulate the anti-tumor immune response of T cells. Specifically, high expression of TCIM in T cells inhibited the nuclear translocation of activated NICD of the Notch2 receptor and thus suppressed the activation of downstream effector molecules, thereby reducing the cytotoxicity of CD8⁺ T cells [112]. Restricting glucose uptake of T cells from TME inhibited EZH2 expression, which indirectly inhibited the Notch signaling through suppressing the two Notch repressors, Numb and Fbxw7, leading to dampening anti-tumor activity of T cells [113]. Together, these studies indicate that Notch signaling plays a positive regulatory role in the anti-tumor properties of T cells. Therefore, activating Notch signaling in T cells, especially CD8⁺ T cells, might be a good strategy for enhancing anti-tumor responses.

As well as boosting T cells’ anti-tumor effects, Notch signaling might accelerate T cell exhaustion. The transcriptional activation complex of canonical Notch signaling directly binds to the promoter of Pdcd1 (encoding PD-1, a marker gene that promotes T cell exhaustion) to promote Pdcd1 transcription in CD8⁺ T cells [114]. Compared with colorectal T cells from healthy individuals, the expression of PD-1 and Notch signaling molecules (NOTCH1, NOTCH2, HES1, and HES5) was elevated in tumor-infiltrating CD8⁺ T cells from CRC patients [115]. Inhibition of Notch signaling not only promoted the cytotoxicity of tumor-infiltrating CD8⁺ T cells, but also enhanced CD8⁺ T cells’ production of proinflammatory cytokines [including IFN-γ, tumor necrosis factor alpha-like (TNF-α), interleukin-1 beta (IL-1β), IL-6, and IL-8] in those patients. This process was accompanied by decreased PD-1 expression in CD8⁺ T cells but did not affect cell proliferation [115]. This result suggests that Notch signaling has potential immunosuppressive properties that might inhibit the cytolytic and non-cytolytic functions of CD8⁺ T cells by inducing PD-1 in colorectal cancer patients [115]. In addition, the single-cell RNA sequencing of T cells in the human TME (e.g., lung cancer, pan-cancer) demonstrated that RBP-J expression was also related to the cytotoxicity or exhaustion of T cells [116, 117]. Together, the above evidence suggests that Notch activation can significantly enhance anti-tumor properties but may also potentially promote T cell exhaustion. However, targeted activation of Notch signaling in T cells combined with immune checkpoint blockers (ICBs), such as αPD-1, might be a promising strategy for enhancing T cell therapy (Fig. 2F).

In summary, Notch signaling plays a “double-edged sword” role in regulating immune responses, as Notch signaling can modulate the functions of anti- or pro-tumor immune cells. Specifically, for innate immune cells (e.g., NK cells, DCs, and macrophages), activation of Notch signaling mainly: (1) enhances the anti-tumor property of NK cells directly; (2) promotes the maturation and antigen presentation of DCs; and (3) facilitates the transition of macrophages into an M1 type. All of these can inhibit tumor progression. However, in the different TME, Notch signaling plays different roles in MDSCs-mediated tumor immunity. In the TME of murine lung carcinoma model, activating Notch signaling in myeloid cells promotes the differentiation of M-MDSCs into M1-TAMs and thus inhibits tumor progression. However, in the glioma TME, activating Notch signaling promotes M-MDSC-mediated immunosuppression and thus facilitates tumor progression. For adaptive T immune cells (e.g., T cells), activation of Notch signaling enhances their anti-tumor property, but Notch signaling also potentially enhances the exhaustion of T cells by upregulating PD-1 expression.

In BC patients, Jagged1 expression correlated with tumor progression. High Jagged1 expression correlated positively with infiltration of stromal M2-TAMs, which predicts poor patient survival and resistance to aromatase inhibitor therapy. However, BC cells pretreated with GSI and co-cultured with macrophages significantly inhibited the polarization of macrophages into M2-TAMs [132]. In BC, high expression of a long noncoding RNA, Linc00514, also promoted the expression of Jagged1, which in turn activated Notch signaling to promote the secretion of IL-4 and IL-6 from BC cells; these events then induced M2 polarization of macrophages. This suggests that activation of Notch signaling mediated by Jagged1 positively promotes M2-TAM polarization [133]. In multiple spontaneous BC models (e.g., 4T1 BC, PyMT-A BC), overexpression of tumor-derived Jagged1 promoted tumorigenesis [134]. By utilizing genetically engineered murine models of mammary-gland-specific Jagged1 overexpression or knockout mice, the researcher found that Notch activation by tumor-derived Jagged1 promoted the secretion of multiple cytokines (e.g., IL-6, WISP1) and TAM recruitment; the proliferation and tumoricidal activity of T cells were then inhibited, partially through upregulation of the T cells’ PD-1 [134]. Also, the combination of Notch inhibitor (GSI) with ICBs (αPD-1) significantly inhibited tumor growth in TNBC [134]. In pancreatic ductal adenocarcinoma (PDAC) patients, high Jagged1 expression in PDAC cells was associated positively with CD68⁺ macrophage infiltration and decreased patient survival [135]. Zhang et al. have shown that the upregulated expression of Dll1 in BC cells induces long-term normalization of tumor vascular and promotes the accumulation of CD8⁺ T cells and the polarization of M1-TAMs [136]. By recruiting 130 patients with invasive BC for bioinformatics and statistical analysis, the researcher found that high expression of Dll3 was associated with poor survival and with high levels of Treg cell infiltration [137]. High infiltration of tumor-associated neutrophils (TANs) was associated with immune tolerance and dismal prognosis in epithelial ovarian cancer (EOC) [138]. High expression of Jagged2 in tumor cells enhanced TAN infiltration, in turn inhibiting CD8⁺ T cell cytotoxicity. Blockade of Notch signaling [anti-Jagged2 antibody or LY3039478 (γ-secretase inhibitor)] reactivated CD8⁺ T cell-mediated anti-tumor properties, inhibiting tumor progression [138]. The above studies show that, in BC, PDAC, and EOC, high expression of Notch ligands (mainly Jagged1, Jagged2, and Dll3) of tumor cells promotes an immunosuppressive microenvironment in the TME, eventually allowing tumors to tolerate immunotherapy.

In addition to Notch ligands, abnormal expression of tumor-derived Notch receptors and downstream signaling genes affects the infiltration of immune cells and therefore tumor progression. By analyzing tumor samples from 152 patients with hormone receptor-positive and -negative phenotypes (luminal and triple-negative/basal-like) of BC, the author found that low mRNA levels of Notch receptors (mainly Notch1, Notch2, and Notch4) mainly in tumor cells were associated with higher infiltration of Treg cells into the tumors, predicting poor prognosis and poor survival [139]. However, in murine B16F10 melanoma models with subcutaneous and lung metastases, ectopic over-expression of Notch1 in B16F10 cells accelerated tumor progression and promoted tumor immunosuppression by upregulating TGF-β1. Specifically, forced high expression of Notch1 in B16F10 cells reduced the release of IFN-γ into the TME and inhibited the infiltration of CD8⁺ T cells and NK cells, while enhancing Treg cell and MDSC infiltration in vivo. PD-1 of CD4⁺ cells and CD8⁺ T cells were upregulated, accelerating T cell exhaustion [140]. In a mouse TNBC model, loss of function of ubiquitin-specific peptidase 9x-linked (USP9x) in tumor cells abolished NICD activation reduced the production of proinflammatory cytokines (e.g., CCL2, IL-1β), which further reduced the tumor inflammation through inhibiting the infiltration of CD206⁺ TAMs and Treg cells, augmenting the anti-tumor immune response through increase the infiltration of CD8⁺ T cells, suppressing BC tumor cell growth in vivo [141]. In a Tgfbr1/Pten knockout mouse model of HNSCC, Notch1 − Hes1 signaling was activated [19]. A γ-secretase inhibitor-DAPT, which inhibited Notch signaling, significantly decreased the burden of HNSCC tumors in that model [19]. Flow cytometry analysis demonstrated that the γ-secretase inhibitor also reduced the infiltration of MDSCs, TAMs, and Tregs into the spleen, draining lymph nodes and the TME as well as decreasing the expression of ICs (e.g., PD-1, CTLA-4, TIM-3, and LAG-3) in T cells in the circulation and tumor TME [19]. This study suggests that blocking Notch1 − Hes1 signaling in HNSCC might be an effective way to reduce immunosuppression and enhance therapeutic efficacy [19] (Table 2).

In the glioma TME, tumor cells escape immune surveillance and increase invasiveness by reducing Notch signaling. Specifically, loss of Notch signaling (mainly Notch1, Notch2, RBP-J, and Hey1) in glioma cells suppressed the expression of MHC-I and cytokines [e.g., C-X-C motif chemokine ligand 9 (CXCL9) and IL-15], reduced the recruitment of anti-tumor immune cells (e.g., CD8⁺ T cells), but favored the infiltration of microglia and pro-tumor TAMs [142]. In gastric cancer (GC) patients, both tumor tissue and peripheral blood showed significantly higher expression of Notch receptor (NOTCH1, NOTCH2) mRNA than normal human gastric tissue, and they also had higher proportions of Treg cells and Th17 cells [143]. Inhibiting Notch signaling with DAPT significantly suppressed Treg cell function in GC patients [143]. Another group also found that high Notch receptor (NOTCH3) expression was a poor prognostic factor when compared with 395 other genes in GC patients [144]. Specifically, high expression of Notch3 was associated with lower infiltration of anti-tumor immune cells (e.g., activated CD8⁺ T cells) and higher infiltration of immunosuppressive cells (e.g., Treg cells, M2-TAMs). In addition, high expression of Notch3 was accompanied by increased expression of a series of ICs [e.g., CD276, adenosine Aa2a receptor (ADORA2A)], resulting in a dampened anti-tumor immune response [144]. In addition to solid tumors, abnormal expression of Notch signaling in tumor cells of hematologic malignancies can also affect the infiltration of immune cells. For example, in diffuse large B cell lymphoma (DLBCL), mutation or knockdown of lysine methyltransferase 2D (KMT2D) in tumor cells indirectly activated Notch signaling (increased NICD protein), boosted the expression of downstream molecules (e.g., MYC and TGF-β1), and accelerated tumor progression by recruiting Treg cells [145]. Also in DLBCL, another group found that mutations in histone acetylation-related molecules [CREB binding protein (CREBBP) or E1A binding protein p300 (EP300)] in tumor cells contributed to tumor progression through indirectly upregulate Notch signaling [146]. Mechanistically, CREBBP or EP300 mutations indirectly activate Notch signaling (increased NICD protein, HEY1,and HEY2 mRNA) and downstream CCL2 − colony-stimulating factor 1(CSF1) in tumor cells, altering macrophage polarization into M2-TAMs and accelerating tumor progression[146] (Table 2). Based on these findings, we conclude that the abnormally expression of Notch receptors and their downstream signaling molecules in tumor cells affects tumor progression partially by regulating immune cells infiltration. Meanwhile, the specific regulatory mechanism of Notch signaling in tumor cell is complex and context-dependent (Table 3).

As well as tumor cells, stromal cells in the TME can also regulate immune cell infiltration and function through Notch signaling. In the TME of KRAS^G12D-driven CRC, Jackstadt et al. found that epithelial Notch1 signaling was critical in disease subtypes with the poorest prognoses and liver metastasis of CRC [147]. Mechanistically, activation of Notch1 in epithelial cells promoted the secretion of TGF-β into the TME, increased the recruitment of TGF-β-dependent neutrophils, and inhibited the anti-tumor function of CD8⁺ T cells [147]. In contrast, recruitment of neutrophils was significantly inhibited by 1D11 (a ligand-trapping antibody targeting TGF-β1/2/3) and then suppressed CRC tumor liver metastasis [147] (Table 2). As the author demonstrated that epithelial Notch1 signaling was critical for the secretion of TGF-β [147], we speculate that blocking Notch1 signaling of intestinal epithelial cells might be a potential strategy for inhibiting CRC metastasis. It could therefore suggest a clinical treatment for liver metastasis in CRC.

Oncolytic virotherapy, an emerging cancer immunotherapy, has received extensive attention in recent years [170]. One of the most widely investigated oncolytic viruses is oHSV, as it efficiently lyses tumor cells while leaving normal cells unscathed. In 2015, the US Food and Drug Administration (FDA) approved the first oncolytic HSV—oHSV-talimogene laherparepvec (T-VEC)—for treating melanoma patients (Clinical Trial.gov identifier: NCT02173171) [171]. Now, oHSV is used to treat glioblastoma (GBM) [93], melanoma [171], breast cancer [172], and ovarian cancer [92]. One study from our group showed that customized oHSV had significant efficacy against GBM in pre-clinical mouse models [173]. Our customized OV-CDH1 oncolytic virus was able to spread into tumors and lyse tumor cells more effectively than control oHSV. It also selectively prevented KLRG1⁺ NK cells from lysing OV-CDH1-infected tumor cells, improving the efficacy of cancer virotherapy [173]. In two recent studies, we customized an oHSV to express a full-length anti-CD47-IgG1 antibody [92, 93]. After that OV-αCD47 infected murine GBM or ovarian tumor in vivo, it lysed tumor cells, released αCD47 into the TME, and induced antibody-dependent cellular cytotoxicity (ADCC) of NK cells and antibody-dependent cellular phagocytosis (ADCP) of macrophages, thus cooperatively enhancing the therapeutic efficacy of cancer virotherapy [93].

In the TME of GBM, oHSV infection abnormally activates Notch signaling, causing TAMs to secrete large amounts of cytokines (e.g., CCL2, IL-10). It then recruits MDSCs to inhibit the therapeutic effect of oncolytic virotherapy [98]. Adding GSI, a pharmacological blocker of Notch signaling, rescued the oHSV-induced immunosuppressive TME and activated CD8⁺ T cell-dependent anti-tumor memory responses, resulting in therapeutic benefits [98]. Therefore, by combining our previously developed oncolytic viruses (e.g., OV-αCD47) with gene sequences encoding antibodies (e.g., αNotch1, αDll1) that block Notch signaling or by combining DAPT, GSI, or other inhibitors of Notch signaling with OVs, we can inhibit cytokine (e.g., CCL2) secretion of TAMs and the recruitment of MDSCs, reactivating CD8⁺ T cells for cancer immunotherapy. These combination therapies have important implications for the clinical treatment of solid tumors (e.g., glioma) that respond positively to inhibition of Notch signaling (Fig. 4A).

Targeted delivery of drugs into specific immune cells or the TME to transform “cold tumor” into “hot tumor” is an emerging and promising strategy for cancer immunotherapy [174]. In recent years, nanoparticles (NPs) have shown great clinical potential in drug delivery systems, as they can accurately and effectively deliver many types of drugs (e.g., oligonucleotides, siRNAs, or protein-based drugs) into TAMs of the TME. For example, siRNAs that modulate NF-κB signaling [175] and VEGF signaling [85] can be payloaded into polymeric NPs; anti-CSF-1R siRNA can be incorporated into lipid-based NPs [176]; and cytosine-phosphate-guanine (CpG) [Toll-like receptor 9 (TLR9) agonist] can be payloaded into carbon NPs [177]. These NPs have been successfully delivered into TAMs of the TME, affecting the cells’ functionality.

Our group and others have found that activating Notch signaling in TAMs of the TME (e.g., in murine lung cancer) promotes TAM polarization into proinflammatory M1-TAMs, thereby increasing the infiltration of CD8⁺ T cells, further inhibiting tumor progression [88, 91]. By integrating Notch-activating ligands (e.g., Dll1) and/or Notch1 overexpression plasmid into mannose-NPs, Notch signaling can be activated in TAMs but not in other cells. Thus, TAMs can be polarized into M1-TAMs, fulfilling the goal of remodeling the TME to improve tumor immunotherapy (Fig. 4B).

ICs are key regulators of immune system suppression [178]. ICBs (e.g., αCTLA-4, αPD-1/αPDL-1) can block inhibitory checkpoints, thereby unleashing suppressed anti-tumor immune responses [179]. In recent years, ICB-based immunotherapy, including αPD-1/αPD-L1 and αCTLA-4, has significantly improved the survival rates of patients with metastatic solid tumors, especially melanoma and lung cancer [180, 181]. Also, ICBs have correlated significantly with Notch signaling changes (activation or inhibition) in various tumors. Activation of Notch signaling in human neuroendocrine (NE) SCLC cell lines induced low NE differentiation and increased intrinsic tumor immunity [182]. Activation of Notch signaling was found to be an important predictor of the clinical benefit of ICB used in two relapsed SCLC cohorts [182]. In CRC, Notch signaling mutations in tumor cells were associated with the enrichment of cytotoxicity-related molecules (e.g., GZMB and PRF1) but also exhaustion-related molecules (e.g., PD-1) [183]. We found that, in some tumors (e.g., TNBC), ICBs combined with GSI inhibited tumor progression [134]. In other tumors (e.g., HNSCC, GC), high Notch expression promoted the expression of ICs, indicating that ICBs may have better therapeutic effects in these cancer patients with high Notch expression compared to those with low expression [19, 140, 146].

In summary, activation or inhibition of Notch signaling in tumor cells can affect the expression of ICs, thus likely modulating the therapeutic effect of ICBs. However, a previous study found that, in anti-tumor T cells, activation of Notch signaling enhanced the cells’ cytotoxicity but could also promote the expression of PD-1, potentially promoting T cell exhaustion. Therefore, we should adopt different synergistic therapeutic strategies in different contexts: (i) For a TME in which inhibition of Notch signaling enhances IC expression (e.g., PD-1), we could combine clinically used inhibitors that target Notch signaling (e.g., γ-secretase inhibitors, ADAM inhibitors) with ICBs (e.g., αPD-1/αPDL-1) to obtain synergistic anti-tumor effects; (ii) For a TME in which activation of Notch signaling enhances the expression of ICs (e.g., PD-1), we could develop NPs loaded with a Notch signaling activator [e.g., NICD, Dll1, Dll3, Dll4, or Notch homolog 1-translocation-associated (Notch1 TFA)] specifically into target cells (e.g., T cells). Combining these NPs with ICBs (e.g., αPD-1/αPDL-1) would produce additive or synergistic anti-tumor activity (Fig. 4C).

CAR (chimeric antigen receptor) protein is a synthetic cell surface receptor that confers immune cells (e.g., T cells, NK cells, and macrophages) with specific anti-tumor properties that can target corresponding antigenic proteins [184]. CAR-T cells have achieved unprecedented success with some hematological malignancies, and a couple of products have already been approved by the US FDA [185, 186]. Other CAR immune cells [including CAR-NK cells, CAR-NKT, CAR-macrophage (CAR-M), and CAR-γδT] have been approved for or are in clinical trials, as documented in our recent review of CAR immune cells’ great potential for improving cancer immunotherapy [184]. Because allogenic CAR-NK cells are efficacious against tumor cells but do not produce cytokine storms or graft-versus-host disease (GVHD) [187], they are being developed into ‘off the-shelf’ drugs for immunotherapy. Using animal models, we obtained a significant anti-tumor effect when we recently used ‘off-the-shelf’ human EGFR-CAR-NK cells and human PSCA-CAR-s15NK cells to treat solid tumors (e.g., glioma and pancreatic cancer) [188, 189]. Also, CAR-M cells have attracted great interest as potential immunotherapies in recent years. Researchers have found that modifying human macrophages with specific CARs can improve the presentation of tumor antigens (especially those on solid tumors) and increase macrophages’ phagocytic activity. These CAR immune cell-mediated tumor therapies have produced good results or shown great potential in the majority of hematological tumors and some non-homogeneous solid tumors. However, very few antigens are truly tumor-specific, and thus, conventional CAR-T cell therapies often cause lethal toxicities such as on-target, off-tumor cross-reaction of CAR cells with normal tissues; they also have poor specificity [126, 190‐192]. In fact, the majority of tumor antigens are often expressed heterogeneously, and treatment with conventional CAR cells allows antigen-negative tumor cells to escape immune surveillance [193]. Therefore, there is an urgent need to develop a new tumor recognition system of CAR cells—one that can recognize tumor cells carrying multiple antigens—to deal with tumor heterogeneity and thereby increase the therapeutic effectiveness of CAR immune cells against solid cancers.

Recently, the synNotch system, developed by the Lim group at the new frontier of cancer research, was launched. This system can accurately teach T cells (especially CAR-T cells) to recognize two or three antigens of solid tumors (e.g., mesothelioma, ovarian tumor, or glioblastoma) [122]. In the section—“SynNotch can be used as a tool to increase T cell cytotoxicity and specificity”, we explained how the synNotch system enhances anti-tumor specificity mediated by T cells (mainly CAR-T cells), especially in solid tumors. SynNotch circuits allow CAR-T cells to integrate extracellular and intracellular antigen recognition signals and accurately identify and kill tumor cells, as they use positive or negative logic to combine two or multiple different antigens [122, 126‐128]. Important future developments will likely include: (i) synNotch circuits that can simultaneously recognize multiple different antigens on tumor cells by CAR-T cells to precisely kill highly heterogeneous solid tumors; (ii) synNotch circuits that can simultaneously recognize multiple antigens on tumor cells by CAR-NK or CAR-M cells; (iii) CAR immune cells can persist in vivo and kill solid tumors by more accurately dissolving or swallowing tumor cells (Fig. 4D).