Background

Human epidermal growth factor receptor 2 (HER2) is overexpressed in 15–20% of invasive breast cancers (BCs).1 2 HER2 overexpression has traditionally been associated with worse prognosis, early metastasis, decreased disease-free survival and overall survival (OS) in BC.3 4 The addition of pertuzumab to neoadjuvant treatment with trastuzumab and paclitaxel, known as THP, has dramatically improved the pathological complete response (PCR) rates in HER2+ BC compared with paclitaxel alone.5–10 However, the impact on long-term survival outcomes remains modest and there is a subset of patients seem to be non-responsive.11

Although there are some researches that have revealed the mechanisms of resistance to trastuzumab alone, such as PI3K/NF-κb pathways,12 immune response13 and cell cycle control mechanisms,14 there is no research involving the resistance mechanisms of THP neoadjuvant therapy. What is more, a reliable prognostic and predictive biomarker to reflect response to THP has not been established. In our study, we found that delta-like ligand 4 (DLL4) accounted for the THP neoadjuvant therapy resistance and was reversely associated with the response of therapy, indicating DLL4 may serve as a potential biomarker for predicting response to THP therapy and a therapy target in HER2+ patients.

The Notch signaling pathway, a highly conserved system in evolution, plays a crucial role in determining cell fate.15 It remains a key regulator to control processes such as differentiation, proliferation, and apoptosis across various tissues.16–19 However, its dysregulation has been observed in various diseases. Aberrant activation of Notch signaling has been increasingly recognized as a driver of BC progression, particularly in enhancing tumor cell survival and resistance to chemotherapy.19–21 DLL4 is one of the main ligand that activates Notch signaling and its effects are profound, including tumor angiogenesis,22 micrometastasis,23 senescence and tumor growth.24 Studies have consistently shown a correlation between DLL4 expression and poor prognosis in pancreatic cancer, gastric cancer, and renal cell cancer.25–27 Additionally, DLL4 is overexpressed in the plasma and tumor tissues of patients with BC, which is associated with metastasis.28–32 However, the precise role of DLL4 in resistance to THP neoadjuvant therapy remains unclear. Our findings indicate that DLL4+ tumor cells exhibit characteristics of cancer stem cells (CSCs) and have the ability to induce neutrophil infiltration and neutrophil extracellular traps (NETs) formation, thereby impeding lymphocyte infiltration. Moreover, we used the DLL4 target chimeric antigen receptor (CAR)-T cells to eradicate DLL4+ tumor cells in the mouse model and sensitized the THP neoadjuvant therapy.

ResultsComprehensive bioinformatic analysis identified tumorous DLL4 as a potential target for neoadjuvant chemotherapy resistance in HER2+ BC

To delineate the factors contributing to neoadjuvant chemotherapy resistance and predict chemotherapeutic responsiveness, we designed a three-step comprehensive screening strategy (online supplemental figure S1A). First, the gene copy number variations containing amplifications and deletions were detected in the HER2+ BC from The Cancer Genome Atlas (TCGA) database (online supplemental figure S1B–D) and a candidate gene library comprizing 111 amplified genes (library 1) was established as pivotal biomarkers for tumor initiation (online supplemental figure S1E). Next, the RNA expression profiles from 44 patients with HER2+ BC who received neoadjuvant THP therapy and another 30 patients who received trastuzumab plus chemotherapy as neoadjuvant chemotherapy were obtained from the public database (GSE181574 and GSE76360) (online supplemental figure S1F,G). Gene enrichment analysis (GSEA) indicated enrichment of immune-related pathways in the non-pathologic complete response (NPCR) group, particularly the negative regulation of immune activity, alongside Notch signaling pathway enrichment (online supplemental figure S1H). Similar results were also obtained from the Kyoto Encyclopedia of Genes and Genomes (KEGG) and Gene Ontology (GO) analysis (online supplemental figure S1I). By analyzing the different expressed genes (DEGs) between the PCR and NPCR groups, the second candidate library (library 2) containing the upregulated genes in the NPCR group was determined (online supplemental figure S1J). Furthermore, considering that membrane proteins could act as valuable targets for therapeutic antibody development and other pharmaceuticals, we subsequently collected the candidate genes encoded membrane proteins from many databases related to membrane proteins (TMbase, Steve White, OPM database and MemProtMD database, etc) as our third candidate library (library 3) (online supplemental figure S1K).

Finally, the three candidate gene libraries were intersected and the common DEGs were acquired, including DLL4, SLC22A12, CD24 (online supplemental figure S1L). Among the DEGs, DLL4 was identified as the most differentially overexpressed membrane protein in resistant patients (online supplemental figure S2A–D) and could lead to the THP resistance by DLL4 overexpression (online supplemental figure S2E). Collectively, by using three candidate libraries, including the amplificated genes, the NPCR genes and membrane proteins, we finally identified DLL4 as the potential target that mediated the resistance of THP neoadjuvant chemotherapy.

The expression level of DLL4 was positively associated with the resistance to neoadjuvant chemotherapy in HER2+ BC

First, we evaluated the expression pattern of DLL4 in HER2+ BCs. As shown in figure 1A,B, the expression levels of DLL4 were abnormally overexpressed in THP-resistant tumor tissues compared with the matched non-tumor tissues. The overall survival (OS) and disease-specific survival (DSS) were significantly prolonged in HER2+ patients with low DLL4 levels in our center cohort and TCGA database (online supplemental figure S3A–D). Besides, DLL4 was overexpressed in the THP neoadjuvant chemotherapy unfavorable response population in the public database (figure 1C). Based on the DLL4 staining extent and staining intensity of tumor cells, DLL4 expression in HER2+ BC was divided into five levels. Consistently, retrospective analysis of the HER2+ BC cohort revealed that high baseline DLL4 expression was significantly associated with resistance to THP neoadjuvant chemotherapy and changes in tumor size (figure 1D–E, online supplemental figure S3E) determined by MRI (online supplemental table S1). Furthermore, only tumorous DLL4 expression level distinguished between PCR and NPCR patients (figure 1F, online supplemental figure S3F) and negatively correlated with the responsiveness (figure 1G).

Figure 1

The expression level of DLL4 was positively associated with the resistance to neoadjuvant THP chemotherapy in HER2+ breast cancer. (A–B) The protein expression levels (A) and IHC staining score (B) of DLL4 in paired tumor (T) and non-tumor tissue (P) from THP chemotherapy resistant patients. (C) DLL4 expression level between favorable response and unfavorable response patients in public database. (D) Representative MRI-scanning images of patients with HER2+ breast cancer. (E–F) DLL4 expression levels in favorable responsive and non-favorable responsive patients detected by multiplex IHC staining. n=130, bars, 50 µM. (G) Spearman’s correlation analysis of tumor DLL4 expression score and Miller-Payne score in patients with HER2+ breast cancer. n=130. (H) Schematic illustration for in vivo experiment of different tumor cell-bearing BALB/c mice with THP therapy. (I–J) Representative macroscopic images (I) and statistical analysis (J) of mammary gland orthotopic tumors of tumor-bearing BALB/c mice from different groups. Unpaired t-tests were used. (K) Six HER2+ breast cancer organoids were treated with THP and apoptotic levels were evaluated by Caspase 3/7 probe (green fluorescence). Bars, 100 µM. (L) The IC50 values of THP in each breast cancer PDO were determined by the CellTiter-Glo (CTG) assay. Spearman correlation were performed. (M–N) The apoptotic rates of 4T1-HER2 tumor cell detected by flow cytometry. Paired t-tests were used. (O) The THP IC50 of 4T1-vector and 4T1-DLL4 were detected by CCK-8 assay. ****p<0.0001, ***p<0.001, **p<0.01, *p<0.05. CCK-8, Cell Counting Kit-8; DAPI, 4',6-diamidino-2-phenylindole dihydrochloride; DLL4, delta-like ligand 4; HER2, human epidermal growth factor receptor 2; N-PCR, non-pathologic complete response; NR, non-responsive; ns, no significant; PCR, pathological complete response; R, responsive; THP, trastuzumab, pertuzumab and paclitaxel; PDO, patient-derived organoid.

To further elucidate the causal role of DLL4+ tumor cells in THP therapy resistance, murine 4T1-HER2-vector and 4T1-HER2-DLL4 cells were injected into the fat pad of BALB/c mice, followed by treatment with THP or control vehicle (figure 1H). Our results showed that THP treatment dramatically reduced the tumor burden in the control group, while it only had a minor antitumor effect in the DLL4 overexpression (4T1-HER2-DLL4) group (figure 1I,J, online supplemental figure S3G). We observed that DLL4-high organoids showed lower apoptotic rates and higher IC50 values than DLL4-low organoids (figure 1K,L, online supplemental figure S3H). DLL4 also decreased the apoptotic rate of 4T1 tumor cells treated with THP therapy (figure 1M-O). Consistent results were obtained in the in vivo and in vitro experiments of the SKBR-3-vector and SKBR3-OE cell line (online supplemental figure S3I–L). These findings supported that DLL4+ tumor cells contributed to the resistance of THP neoadjuvant chemotherapy.

DLL4 + cancer cells exhibited CSC characteristics and contribute to neoadjuvant chemotherapy resistance in HER2+ BC

To further elucidate the mechanisms through which DLL4+ tumor cells can induce THP therapy resistance in HER2+ BC, we used single-cell sequencing data (GSE166321) from HER2-overexpressing breast tumors from mice (figure 2A–C, online supplemental figure S4). GSEA showed significant enrichment in embryonic stem (ES)/stem cell gene signatures in DLL4 positive cancer cells (figure 2D). Therefore, we sought to investigate whether DLL4+ tumor cells displayed CSC properties in HER2+ BC and were associated with chemotherapeutic resistance. We found that DLL4+ tumor cells formed more tumor spheres, increased ALDH1 enzymatic activity and had a higher proportion of CD44+CD24− tumor cells (figure 2E–J, online supplemental figure S5A) compared with DLL4− tumor cells. Additionally, messenger RNA (mRNA) and protein levels of stemness-related genes (Sox9, Sox2, Oct4, and Nanog) were also elevated in DLL4+ tumor cells (figure 2K–N) in three patients. Furthermore, in vivo limited diluted assay confirmed that DLL4+ tumor cells increased tumor incidence (figure 2O). According to the DLL4 baseline expression, SKBR-3-OE, SKBR-3-KD, 4T1-HER2-OE and BT474-KD cell lines were established. The consistent results were also observed in BT474, 4T1-HER2 and SKBR-3 cell lines after DLL4 downregulation or upregulation (online supplemental figure S6A–S), underscoring DLL4’s critical role in maintaining the CSC phenotype in BC.

Figure 2

DLL4+ cancer cells exhibited cancer stem cell characteristics and account for THP therapy resistant. (A–C) The scRNA-seq data (GSE166321) of HER2+ breast cancer was analyzed by t-distributed stochastic neighbor embedding (t-SNE) projection and major cell types are marked according to the cluster marker genes. (D) The GSEA enrichment of DLL4_positive group and DLL4_negative group (E–F) Sphere formation assays were performed in indicated tumor cell lines, 50 µM. (G–H) The percentage of ALDH+ cells in DLL4− tumor cells and DLL4+ tumor cells detected by flow cytometry. (I–J) The percentage of CD44+ CD24− tumor cells were detected by flow cytometry. (K–N) The mRNA levels (K) and protein expression levels (L–N) of stemness-associated genes were detected in DLL4− tumor cells and DLL4+ tumor cells. (O) In vivo limited dilution assays in DLL4− tumor cells and DLL4+ tumor cells from different patients. Representative tumor incidence and CSC probabilities are shown in the table. (P–R) The protein expression levels of Notch signaling activation marker NICD detected by western blot in different patients. (S–U) The number of spheroids (S), percentage of ALDH+ cells (T), percentage of CD44+ CD24− tumor cells (U) of indicated tumor cells treated with Notch inhibitor MK-0752. (V) The protein expression levels of stemness-associated genes after treated with Notch inhibitor MK-0752. Paired t-tests were used. ****p<0.0001, ***p<0.001, **p<0.01, *p<0.05.BC, breast cancer; ns, no significant; CSC, cancer stem cell; DLL4, delta-like ligand 4; GSEA, gene enrichment analysis; HER2, human epidermal growth factor receptor 2; mRNA, messenger RNA; ns, no significant; THP, trastuzumab, pertuaumab and paclitaxel; SSC, side scatter; scRNA: single cell RNA sequencing.

Figure 3

DLL4 restrict T cell infiltration and further hampered the efficacy of THP therapy. (A) The immune cell distribution of each patient in PCR and NPCR group of public database analyzed with CIBERSORT method. The labels below the figure represent different immune cell populations. (B) The different immune cell infiltrations of PCR and NPCR group. (C–D) The percentage of CD3+ T cells detected by IHC staining in FNA specimen. Bars, 50 µM. (E–F) The percentage of CD8+T cells (E–F) and CD4+T cells (G–H) in THP resistant and THP sensitive patients were detected by multiplex IHC staining. Bars,50 μM. (I–J) The protein expression levels of CD4 and DLL4 detected by multiplex IHC staining in low DLL4 and high DLL4 patients. Bars,50 μM. (K, N) Quantification of average distance of CD4+T cells (K) and CD8+T cells (N) from different DLL4 expression level epithelial cells in low DLL4 and high DLL4 patients. n=130, bar, 50 µM. (L–M) The protein expression levels of CD8 and DLL4 detected by multiplex IHC staining in low DLL4 and high DLL4 patients. Bars, 50 µM. (O) The schematic illustration for in vivo experiment of CD3+T cells deletion using anti-CD3 antibody in tumor-bearing BALB/c mice. (P) The flow cytometry of CD3+T cells in PBMC after in vivo CD3 deletion using anti-CD3 antibody. (Q) The tumor weight of different groups treated with or without anti-CD3 antibody or THP therapy (six mice for each group). Unpaired t-test analysis was used. ****p<0.0001, ***p<0.001, **p<0.01, *p<0.05. DAPI, 4',6-Diamidino-2-phenylindole dihydrochloride; DLL4, delta-like ligand 4; FNA, fine needle aspiration; HPF, high power field; IHC, immunohistochemistry; NK, natural killer; ns, no significant; NPCR, non-pathological complete response; PBMC, peripheral blood mononuclear cell; PCR, pathological complete response; THP, trastuzumab, pertuzumab and paclitaxel.

We further asked whether DLL4 regulated cancer stemness in a Notch pathway-dependent manner. Notch signaling pathway was activated in DLL4+ tumor cells (figure 2P–R). Furthermore, the spheroid formation ability, ALDH1 enzymatic activity and CD44+CD24− cell proportion in DLL4+ tumor cells were all decreased after Notch pathway blockade by using γ-secretase inhibitor MK-0752 (figure 2S–U). MK-0752 also reversed the expression of stemness-related genes in DLL4+ tumor cells (figure 2V), highlighting the crucial role of Notch signaling in sustaining DLL4+ tumor cell stemness.

DLL4 rewired the tumor immune microenvironment and restrict T cell infiltration, which further hampered the efficacy of chemotherapy

Growing evidence suggests that CSCs may remodel the tumor immune microenvironment to enhance immune evasion.33–36 Considering that negative regulation of the immune activity pathway was enriched in the NPCR group in comparison to the PCR group (online supplemental figure S1H,I), we further explored whether DLL4 could reshape the immune landscape in HER2+ BC. So we performed deconvolution for the RNA bulk sequence gene expression data in the Gene Expression Omnibus database with CIBERSORT method37 to estimate the diversity of immune cell infiltrating and correlated cell fractions in the NPCR group and found that the proportion of T cells was dramatically decreased (figure 3A,B). As shown in figure 3C,D, the decreased total CD3+ T cells were found in the fine needle aspiration specimen of THP neoadjuvant chemoresistant patients, including both CD8+ T cells and CD4+T cells were decreased in the chemoresistant patients (figure 3E–H), indicating that T cell infiltration was restricted in the neoadjuvant chemoresistant patients. Furthermore, we found that the proportion of CD8+T cells and CD4+T cells were both decreased in the high-DLL4+ tumor cell patients (figure 3I,J,L,M). Moreover, we also found that CD8+ and CD4+ T cells were located more distant from the tumor cells in the high-DLL4+ tumor cell patients (figure 3K,N). Consistently, in vivo assays also confirmed that the proportion of CD3+ T cells, CD8+ T cells and CD4+ T cells were both decreased in the 4T1-HER2-DLL4 tumors (online supplemental figure S7A–C).

To further evaluate whether DLL4 could induce chemoresistance by reducing T-cell infiltration, in vivo T-cell deletion assay was carried out. As shown in figure 3O,Q, CD3+ T-cell deletion could weaken the effects of DLL4 in THP therapy resistance. What is more, among CD3+T cells, both CD4+T cells and CD8+T cells contributed to DLL4-induced therapy resistance, but CD8+T cells played a dominant role (online supplemental figure S7D–H). Collectively, these data highlighted that DLL4 could rewire the tumor immune microenvironment and was associated with the restriction of T-cell infiltration, thereby enhancing the tumor’s resistance to THP therapy.

DLL4 promote neutrophil recruitment and NET formation in HER2+ BC

To further investigate the impact of DLL4 on the immune microenvironment and elucidate the mechanisms underlying T-cell exclusion in HER2+ BC, we analyzed RNA bulk sequencing data and observed a significant increase in myeloid cell populations, particularly neutrophils and macrophages, in tumors with high DLL4 expression (figure 4A). However, we did not observe obvious alterations in macrophage (including M1 and M2) or dendritic cells (DCs) population in neoadjuvant chemotherapy-resistant patients or those with high levels of DLL4+ tumor cells (online supplemental figure S8A–H) in our cohort patients. Whereas, increased neutrophil infiltration was found in neoadjuvant THP therapy-resistant patients (online supplemental figure S8I). Consistent results were observed in patients with high levels of DLL4+ tumor cells (online supplemental figure S8J). Additionally, analysis of the intracellular communication network revealed enhanced interactions between DLL4+ cancer cells and neutrophils (figure 4C,D), particularly in the activation of Notch signaling within this context (figure 4E). In addition, we found a negative correlation between neutrophil infiltration and CD3+T cells (figure 4F). These findings suggested that a potential association between neutrophils and THP therapy resistance.

Figure 4

DLL4 can induce neutrophil infiltration and NET formation. (A) The diversity of immune cell infiltration in the GSE181574 data sets between DLL4+ high and DLL4+ low patients using CIBERSORT method. (B) KEGG pathway analysis of DLL4+ high and DLL4+ low patients in GSE181574 data sets. (C–D) The CellChat analysis of interactions between different cell types. The circle plot in number of interactions (C) and interaction weights/strength (D) were shown. (E) The heatmap of cell interactions in Notch signaling pathway. Red dashed line indicated the interactions of DLL4− cancer cell and DLL4+ cancer cell with other cell types. (F) Spearman’s correlation analysis of T cells and neutrophils in 130 patients with HER2+ breast cancer. (G–H) The percentage of neutrophils undergoing NETosis (MPO+ cells) detected by multiple IHC staining. Bars, 50 µM. (I) Spearman’s correlation analysis of DLL4+ cells and MPO+ cells in 130 patients with HER2+ breast cancer. (J) The schematic illustration for the NETs formation and transwell assay after isolating neutrophils from BALB/c mice PBMC. (K–L) The number of migrated primary mouse neutrophils after treated with conditional medium detected by transwell assay. Bars, 100 µM, paired t-tests were used. (M–N) Immunofluorescence staining for CitH3 (red) and DAPI (blue) on cultured mice primary neutrophils after treated with different conditional medium. ****p<0.0001, ***p<0.001, **p<0.01, *p<0.05. DAPI, 4',6-Diamidino-2-phenylindole dihydrochloride; FDR, false discovery rate; IHC, immunohistochemistry; KEGG, Kyoto Encyclopedia of Genes and Genomes; NES, normalized enrichment score; BC, breast cancer; DC, dendritic cell; DLL4, delta-like ligand 4; HER2, human epidermal growth factor receptor 2; NETs, neutrophil extracellular traps; ns, no significant; PBMC, peripheral blood mononuclear cell; Th1, T helper cells.

Neutrophils, as innate immune phagocytes, play a pivotal role in immune regulation, performing functions such as phagocytosis, degranulation, and the release of NETs.38 We found increased levels of total neutrophils and neutrophils undergoing NETosis (MPO+ neutrophils) in patients with high DLL4+ tumor cell levels (figure 4G–H) and a positive correlation between MPO+ cells and DLL4+ cells (figure 4I). Then the primary murine neutrophils were isolated and co-cultured with different conditional medium (figure 4J). Murine neutrophils exhibited a stronger migratory response (figure 4K–L) and an increased proportion of neutrophils undergoing NETosis (figure 4M–N) toward the supernatant of 4T1-HER2-DLL4 tumor cells. Similar results were obtained using the supernatant of primary BC tumor cells and SKBR-3 tumor cells in the human promyelocytic leukemia cell line (HL-60) (online supplemental figure S8K–R). These results suggested that the supernatant of DLL4-overexpression tumor cells could induce neutrophil infiltration and the formation of NETosis.

DLL4+ tumor cell restricted T cell infiltration and enhanced chemotherapy resistance via NET formation in HER2+ BC

To further ascertain whether DLL4+ tumor cells restricted T cell infiltration through inducing the NET formation in HER2+ BC, tumor-bearing mice were treated with protein arginine deiminase 4 inhibitor, GSK484 (online supplemental figure S9A), which could inhibit the neutrophil NETosis, and significantly inhibited NET formation without affecting neutrophil numbers (online supplemental figure S9B,C). Furthermore, GSK484 rescued the T-cell infiltration in the 4T1-HER2-DLL4 group (online supplemental figure S9D,E). What is more, we observed a notable reduction of tumor burden in the DLL4 overexpression group after treatment with THP plus GSK484 (online supplemental figure S9F–K). Taken together, DLL4+ tumor cells restricted T cell infiltration through inducing the NET formation and further promoted THP therapy resistance in HER2+ BC.

Soluble DLL4 activated Notch signaling in neutrophils and further induced NET formation by promoting the transcription of MPO, PDIA4 and ELANE

We further investigated the mechanism by which DLL4+ tumor cells induced NET formation and infiltration. DLL4 is a type I transmembrane protein composed of 685 amino acids and its extracellular signaling domain can be shed from the membrane to form the soluble DLL4 (sDLL4).39 But whether sDLL4 could directly interact with neutrophils and induce NETosis remains unknown. To explain this, primary murine neutrophils were treated with recombinant mouse DLL4 (rmDLL4). Results showed that the migration of neutrophils and NETosis levels were both increased after rmDLL4 treatment (figure 5A–D). The mRNA and protein levels of key regulator proteins of NETosis were also increased (figures 5E–6H). Consistent results were obtained in SKBR-3 cell lines (online supplemental figure S10A–C). Additionally, treatment with sDLL4 neutralizing antibody reversed neutrophil migration and NETosis formation in murine neutrophils (figure 5I–L), suggesting that sDLL4 shedding from the DLL4+ tumor cells directly led to neutrophil recruitment and NET formation.

Figure 5

Soluble DLL4 can activated Notch signaling in neutrophils and further induced NET formation. (A–B) The migrated mice primary neutrophils after treated with or without rmDLL4 were detected by transwell assay. Bars, 100 µM. (C–D) Immunofluorescence staining for CitH3 (red) and DAPI (blue). (E–F) The mRNA (E–F) and protein (G–H) levels of NETosis-associated genes, including PIDA4, MPO, ELANE in mice primary neutrophils (E,G) and HL-60 (F,H) after treated with different conditional medium. (I–L) The number of migrated mice primary neutrophils (I–K) and percentage of CitH3+cells (J,L) after treated with pLV-vector or pLV-DLL4 4T1 conditional medium with or without anti-DLL4 antibody. (M–N) The migrated mice primary neutrophils that treated with rmDLL4 were treated with or without γ-secretase inhibitor MK-0725. Bars, 100 µM. (O–P) Immunofluorescence staining for CitH3 (red) and DAPI (blue) on neutrophils with different treatment. Bars, 100 µM. (Q–T) The mRNA (Q–R) and protein levels (S–T) of NETosis-associated genes of mouse primary neutrophils (Q,S) and HL-60 (R,T) treated rmDLL4 and MK-0752. Paired t-tests were used. ****p<0.0001, ***p<0.001, **p<0.01, *p<0.05. DAPI, 4',6-Diamidino-2-phenylindole dihydrochloride; DLL4, delta-like ligand 4; HL-60, human promyelocytic leukemia cell line; mRNA, messenger RNA; NET, neutrophil extracellular trap; ns, no significant; PIDA4, protein arginine deiminase 4 inhibitor.

Figure 6

DLL4-targeted CAR-T therapy effectively sensitized HER2+ breast cancer to THP chemotherapy. (A) The schematic illustration of the structure and function of DLL4-targeted CAR-T. (B–C) DLL4-CAR-T secreted IFN-γ (B) and IL-2 (C) were detected by ELISA (D). Short-term cytotoxicity assay of 4T1-HER2-vector or 4T1-HER2-DLL4 with or without DLL4 CAR-T. (E–F) The apoptosis cells of DLL4-high and DLL4-low MMTV-Neu derived organoids treated with or without DLL4-CAR-T or THP therapy. (G) The statistical analysis of the apoptotic rate in THP sensitive and resistant MMTV-Neu derived organoids treated with or without DLL4 CAR-T or THP therapy. (H) The schematic illustration of CAR-T cell production and adoptive transfer. (I–L) The tumor weight (I), percentage of CD8+T cells (J), CD4+T cells (K) and MPO+CitH3+ cells (L) of different groups to test the effects of DLL4-CAR-T cells. Six mice for each group. Unpaired t-tests were used. (M) EGFP+ DLL4-CAR-T cells were monitored by flow cytometry in PBMC. Two-way ANOVA analysis was used. (N–O). The IL-2 (N) and IFN-γ (O) concentration of PBMC in these mice were detected by ELISA on day 28. Unpaired Student’s t-tests were used. (P) The percentage of EFGP+CAR-T cells in tumor were detected by flow cytometry. Unpaired Student’s t-tests were used. ****p<0.0001, ***p<0.001, **p<0.01, *p<0.05. ANOVA, analysis of variance; DLL4, delta-like ligand 4; E:T, effector cell: tumor cell; HER2, human epidermal growth factor receptor 2; IL-2, interleukin-2; NTD, untransduced T cells; ns, no significant; PBMC, peripheral blood mononuclear cell; THP, trastuzumab, pertuzumab and paclitaxel.

sDLL4 can bind to the Notch receptor and activate the Notch signaling pathway, which can regulate the transcription of downstream genes. To explore the function of a Notch signaling pathway in the process of sDLL4 inducing NET formation, γ-secretase inhibitor MK-0752 was used to block the Notch signaling in neutrophils and the increased migration ability and NET formation induced by the rmDLL4 were attenuated by MK-0752 (figure 5M–P). The mRNA and protein levels of key regulator proteins of NETosis were also decreased after MK-0752 treatment in primary murine neutrophils and HL-60 (figure 5Q–T), confirming the regulatory effect of Notch signaling on these enzymes. We further observed that RBPJ, an important transcriptional regulator of the Notch signaling pathway, had motifs and could bind to the promoter regions of PDIA4, MPO and ELANE genes (online supplemental figure S11A–G), which was further enhanced by rhDLL4 (online supplemental figure S11H–J) and promoted transcription of NETosis-related genes (online supplemental figure S11K–P).

DLL4-targeted CAR-T therapy effectively sensitized HER2+ BC to THP chemotherapy

Our results emphasized the significant role of DLL4+ tumor cells in THP neoadjuvant chemotherapy resistance, so we supposed that eliminating this cell subpopulation could potentially sensitize HER2+ BC to THP chemotherapy. To specifically eradicate the DLL4+ tumor cell subset, DLL4-targeted CAR-T cells were produced (figure 6A). Based on the single-chain Fragment Variable (scFV) sequence of the mouse DLL4 antibody, the mouse DLL4-targeted CAR with internal ribosome entry site (IRES) -driven EGFP was constructed. After a 24-hour co-culture, 4T1-HER2-DLL4 tumor cells (DLL4-high density) induced much higher IFN-γ and IL-2 secret