Introduction

Pancreatic cancer is a highly malignant disease that is primarily accounted for by pancreatic ductal adenocarcinoma (PDAC).1 2 Despite considerable advances in surgery, radiotherapy, and targeted therapies, the 5-year survival rate of PDAC is still only 9%.3 4 While immune checkpoint blockers (ICBs) have revolutionized cancer treatment, sustained clinical benefits for specific cancers are observed in only a minority of patients. Intrinsic resistance to ICBs is observed in immune-desert (‘‘cold’’) tumors, characterized by low mutation load and rare infiltrating immune effector cells, as opposed to immune-inflamed (‘‘hot’’) tumors.5 Our previous study revealed that pancreatic cancer tissue is richly infiltrated with macrophages, which significantly impacts tumor growth and metastasis.6 Therefore, strategies aimed at transforming the pancreatic cancer immune microenvironment from “cold” to “hot” through macrophage targeting hold promise for enhancing the antitumor effects of immunotherapy.

Within solid tumors, uncontrolled cell proliferation induces stress and heightened apoptosis.7 8 However, abundant macrophage infiltration in pancreatic cancer tissues efficiently clears apoptotic cells before membrane damage occurs, thus suppressing intratumoral inflammation and immune responses.9 10 The clearance of apoptotic cells by macrophages involves multiple cell receptors, including Annexin A1 (ANXA1), a member of the Annexin protein superfamily that enhances the recognition, binding, and internalization of apoptotic cells via a phagocytic process known as efferocytosis.11–13 Prior research has demonstrated impaired phagocytic function and reduced apoptotic cell clearance in ANXA1-deficient macrophages, which may be related to its anti-inflammatory effects.13–15 Elevated ANXA1 expression correlates with poor overall survival (OS) in pancreatic cancer.16 Notably, ANXA1 expression in tumor cells and extracellular vesicles has been implicated in promoting macrophage M2 polarization and tumor progression.17 Nevertheless, the precise mechanism of ANXA1 in tumor-associated macrophages (TAMs) within pancreatic cancer remains unresolved.

The cGAS-STING signaling pathway is a key factor in driving the antitumor type I interferon (IFN) response.18 cGAS, a cytoplasmic DNA sensor, binds to self-DNA that invades the cytoplasm, catalyzing the synthesis of cyclic GMP-AMP, which further activates the adapter protein STING and triggers a TBK1-IRF3-dependent signaling cascade that leads to the production of pro-inflammatory cytokines and type I IFN.19 DNA released from apoptotic tumor cells can trigger cGAS-STING signaling, which is involved in the progression of pancreatic cancer.20 21 Therefore, an increased understanding of the mechanism of cGAS-STING signaling activation in pancreatic cancer may lead to the development of new therapeutic approaches.

In this study, we systematically evaluated the expression of ANXA1 using public databases and pancreatic cancer tissue staining. Our results confirm that ANXA1 expression on TAMs is associated with lower apoptosis levels and worse OS in patients with pancreatic cancer. To investigate the role of macrophage ANXA1 in pancreatic cancer, we used mouse models with myeloid-specific ANXA1-knockout, which confirmed that myeloid ANXA1 deficiency slows tumor growth. We demonstrated that ANXA1 knockout in macrophages diminishes the clearance of apoptotic cells, consequently augmenting type I IFN-dependent inflammatory responses. Furthermore, we assessed the changes in tumor-infiltrating immune cell populations in ANXA1-deficient macrophages by single-cell RNA sequencing (scRNA-seq) and flow cytometry. Finally, we determined whether ANXA1 knockdown has synergetic effects when combined with gemcitabine and anti-programmed cell death protein-1 (PD-1) antibody for slowing the progression of pancreatic cancer in mice. Our results suggest that ANXA1 may be an effective target for the treatment of pancreatic cancer.

ResultsHigh ANXA1 expression in macrophages of patients with PDAC is associated with lower levels of cleaved Caspase-3 and poor survival

To elucidate the role of ANXA1 in TAMs in PDAC, we analyzed data from the Tumor Immune Estimation Resource (TIMER) database. The results indicate that ANXA1 was positively correlated with the infiltration of TAMs (R=0.235, p<0.001; online supplemental figure S1A). Moreover, ANXA1 expression was positively correlated with the messenger RNA (mRNA) expression of programmed death-ligand 1 (R=0.422, p<0.001; online supplemental figure S1B).

To visualize ANXA1 expression in the tumor microenvironment, we retrieved tumor specimens from 151 patients with PDAC (table 1). Representative immunohistochemistry (IHC) photomicrographs of ANXA1, the macrophage marker CD68 and the apoptosis marker cleaved Caspase-3 (c-Casp3) are shown in figure 1A. The results indicate that ANXA1 was highly expressed in 57.6% of the PDAC tissues, with corresponding higher levels of CD68 (53/87, 60.9%) and lower levels of c-Casp3 (18/70, 25.7%); and that despite the higher overall CD68 levels, the ANXA1+ TAMs were low (figure 1B–D). Patients with high compared with low expression of ANXA1 had worse OS (p=0.015; figure 1E). As observed in our previous study,6 CD68 expression was also significantly associated with worse OS (p<0.001; online supplemental figure S1C). Consistently, ANXA1 expression was positively correlated with CD68 expression (R=0.34, p<0.001; figure 1C). The percentage of c-Casp3-positive cells within the tumors was not significantly associated with OS (p=0.39; online supplemental figure S1D); however, both ANXA1 and CD68 were negatively correlated with the expression of c-Casp3 (R= −0.253, p=0.002; R= −0.302, p<0.001; figure 1D).

Table 1

ANXA1, CD68 and c-Casp3 expression and clinic-pathological characteristics

Figure 1

Immunohistochemistry and immunofluorescence of ANXA1, CD68, and c-Casp3 in human PDAC. (A) Staining with an anti-ANXA1, anti-CD68, and anti-c-Casp3 antibody in the human PDAC tissue samples at×100 magnification and×400 magnification. The images do not depict simultaneous high or low expression of these three indicators within the same samples. (B) Results of IHC and double IF staining. (C) Spearman rank correlation analysis between the expression of ANXA1 and CD68 in 151 patients with PDAC. (D) Spearman rank correlation analysis between the expression of ANXA1, CD68 and c-Casp3 in 151 patients with PDAC. (E) Kaplan-Meyer plot of OS in 151 patients with PDAC with high or low tumor ANXA1 expression. (F) The immunofluorescence staining of co-expression of ANXA1 and CD68. (G) Spearman rank correlation analysis between the expression of ANXA1+ TAMs and c-Casp3 in 151 patients with PDAC. (H) Kaplan-Meyer plot of OS in 151 patients with PDAC with high or low tumor-infiltrating ANXA1+ TAMs. ANXA1, Annexin A1; c-Casp3, cleaved Caspase-3; IF, immunofluorescence; IHC, immunohistochemistry; OS, overall survival; PDAC, pancreatic ductal adenocarcinoma; TAMs, tumor-associated macrophages.

To further evaluate the relationship between ANXA1+ TAMs and apoptosis in tumor tissues of patients with PDAC, we conducted double immunofluorescence (IF) staining with CD68 and ANXA1 antibodies in tumor specimens. Representative photomicrographs are shown in figure 1F. The results indicate a significant negative correlation between the level of ANXA1+ TAMs and c-Casp3 expression (R= −0.396, p<0.001; figure 1G), suggesting that ANXA1+ TAMs may have a role in reducing the apoptosis levels in PDAC tumors. We also performed univariate and multivariate analysis, which indicated that a high frequency of ANXA1+ TAMs was an independent predictor of OS in patients with PDAC (p<0.001; figure 1H, table 2). Taken together, these results are consistent with the possibility that ANXA1 expression in TAMs may exacerbate pancreatic cancer by negatively regulating apoptosis.

Table 2

Univariate and multivariate Cox proportional analysis for overall survival (N=151)

ANXA1 deficiency in mice diminishes macrophage-mediated clearance of apoptotic cells by efferocytosis

To address the potential role of ANXA1 expression in the clearance of apoptotic cells via efferocytosis, we extracted peritoneal macrophages and thymocytes from wild-type (WT) and ANXA1 knockout (Anxa1–/–) mice. The thymocytes were treated with dexamethasone to induce apoptosis and then labeled with pHrodo, which exhibits an increase in fluorescence intensity after endocytosis. Subsequently, the apoptotic thymocytes were co-cultured with peritoneal macrophages and the phagocytosis of the thymocytes was measured according to the fluorescence level (figure 2A). After 45 min, the phagocytosis by Anxa1–/– macrophages was significantly lower than the phagocytosis by WT macrophages (figure 2B). These results indicate that ablation of ANXA1 inhibits efferocytosis.

Figure 2

Myeloid-specific Anxa1 deficiency inhibited efferocytosis of macrophages. (A) Schematic diagram of in vitro efferocytosis experiments. (B) Anxa1 deficiency inhibited the uptake of AC (red) by macrophages (green). Data are representative of three independent experiments. (C) To evaluate the clearance of thymocyte apoptosis in vivo, mice were treated with dexamethasone. Annexin-V-APC and PI were used to detect apoptotic cells and dead cells. The data presented are the means±SD. ANXA1, Annexin A1; WT, wild-type.

To further corroborate the role of ANXA1 expression in macrophages in an in vivo model system, we injected mice with dexamethasone to induce apoptosis in the thymus. A large number of apoptotic thymocytes were identified at 8 hours, most of which were eliminated by resident macrophages in WT mice at 24 hours. However, the clearance process was largely impaired in Anxa1–/– mice, for which significantly more apoptotic cells remained at 24 hours (figure 2C). These results suggest that ANXA1 regulates the efferocytosis of apoptotic cells by macrophages in vivo and in vitro.

Myeloid-specific ANXA1 deficiency inhibits tumor growth in mice

To explore the role of macrophage ANXA1 expression in PDAC progression, we generated myeloid-specific ANXA1-knockout (Anxa1ΔMɸ) and control (Anxa1WT) mice by crossing Anxa1flox/flox mice with Lyz2-Cre mice. Subsequently, we subcutaneously injected Panc02 pancreatic cancer cells into the flanks of Anxa1ΔMɸ and Anxa1WT mice. Flow cytometry results indicated that the percentage of apoptotic cells in tumor tissues of Anxa1ΔMɸ mice as compared with compared with Anxa1WT mice was significantly increased (figure 3A). Furthermore, Western blot (WB) assays verified that the level of apoptosis (c-Casp3+) in tumor tissues of Anxa1ΔMɸ mice was significantly higher than that in Anxa1WT mice (figure 3B). In addition, the cell-free DNA (cfDNA) in blood circulation, which is released by damaged or dead cells, was obviously higher in Anxa1ΔMɸ mice compared with Anxa1WT mice (figure 3C). These results indicate that macrophage ANXA1 deficiency leads to an accumulation of apoptotic cells in tumors.

Figure 3

Myeloid-specific Anxa1 deficiency causes accumulation of apoptotic cells in tumors. (A) Flow cytometry analysis of apoptotic level of tumors from Anxa1ΔMɸ and Anxa1WT mice. Annexin-V-APC and PI were used to detect apoptotic cells and dead cells. (B) Immunoblot analysis of c-Casp3 expression in tumors from Anxa1ΔMɸ and Anxa1WT mice. (C) Quantification of host-derived cfDNA in the plasma collected from Anxa1ΔMɸ and Anxa1WT Panc02-tumor-bearing mice. Each group contained four mice. (D) Representative images of tumors from Anxa1ΔMɸ and Anxa1WT mice (left). Quantification of tumor volume and weight are shown (right). Each group contained five mice. (E) Immunoblot analysis of Bcl2 and Bax expression in Panc02 tumor cells that were untreated, irradiated with 25 mJ/cm2 UV-C (UV), or treated with 100 µM BIP-V5 (Bax inhibitor) and irradiated with 25 mJ/cm2 UV-C (BIP-V5+UV). (F) Flow cytometry analysis of apoptotic level of Panc02 tumor cells that from these groups. (G) Flow cytometry analysis of apoptotic level of tumors from these groups. (H) The growth of Panc02 tumor cells that were treated with control solvent or BIP-V5 (200 ug/day) in Anxa1ΔMɸ and Anxa1WT mice. Each group contained five mice. (I) Immunoblot analysis of ANXA1 in BMDM cells that transfected with lentiviruses control (BMDM-ctrl) and carrying plasmids overexpressing ANXA1 (BMDM-ANXA1). (J) Representative images of tumors from the PDAC model that macrophages were reconstituted by BMDM-ctrl and BMDM-ANXA1 (left). Quantification of tumor volume and weight are shown (right). Each group contained five mice. (K) Immunoblot analysis of c-Casp3 expression in tumors from BMDM-ctrl and BMDM-ANXA1 mice. The data presented are the means±SD. ANXA1, Annexin A1; BIP-V5, Bax inhibitory peptide V5; BMDM, bone marrow-derived macrophage; c-Casp3, cleaved Caspase-3; cfDNA, cell-free DNA; PDAC, pancreatic ductal adenocarcinoma.

Next, we evaluated the growth of subcutaneously implanted Panc02 cells in Anxa1ΔMɸ and Anxa1WT littermates. The volume and weight were significantly lower for tumors from Anxa1ΔMɸ compared with Anxa1WT mice (figure 3D). To confirm the tumor inhibitory effects of macrophage ANXA1 deficiency, we established mice bearing orthotopic pancreatic KPC tumors. Consistently, the tumor growth was lower and the amount of apoptotic cells was higher in tumors from Anxa1ΔMɸ compared with Anxa1WT mice (online supplemental figure S1E,F). These data suggest that ANXA1 deficiency in macrophages induces antitumor effects in PDAC.

To determine whether the antitumor effect of macrophage ANXA1 deficiency is caused by the accumulation of apoptotic cells, we applied the Bax inhibitory peptide V5 (BIP-V5) to suppress Bax-mediated apoptosis of cancer cells (figure 3E). BIP-V5 treatment did not affect the proliferation of Panc02 cells in vitro (online supplemental figure S1G), but instead conferred resistance to apoptosis triggered by apoptotic stimuli (figure 3F). In the Panc02 tumor model, BIP-V5 treatment reversed the effect of macrophage ANXA1 deficiency in increasing the percentage of apoptotic cells in tumor tissues (figure 3G) and blocked its antitumor effect (figure 3H). These results suggest that deficiency of ANXA1 in macrophages causes an accumulation of apoptotic cells that promotes antitumor efficiency.

To further verify the role of macrophage ANXA1 in the tumor microenvironment, bone marrow-derived macrophages (BMDMs) overexpressing murine ANXA1 were established. WB results confirmed the obvious ANXA1 overexpression in BMDMs (figure 3I, online supplemental figure S1H). In a PDAC model, we established a system in which macrophages were depleted and subsequently reconstituted by either control or ANXA1-overexpressed BMDMs. Overexpression of ANXA1 in BMDMs increased tumor growth in vivo (figure 3J). The WB assay indicated that overexpressing ANXA1 in BMDMs decreased the levels of c-Casp3 protein in the tumor microenvironment (figure 3K, online supplemental figure S1I).

Macrophage ANXA1 deficiency activates the cGAS-STING pathway to induce type I IFN in tumors

To explore the mechanism by which macrophage ANXA1 deficiency induces antitumor responses, we performed RNA sequencing of tumors from Anxa1ΔMɸ and Anxa1WT mice. Volcano plots revealed the expression patterns of the differentially expressed genes (DEGs), most of which were upregulated in tumors from Anxa1ΔMɸ as compared with Anxa1WT mice (online supplemental figure S1J). Specifically, among the 365 DEGs that were identified, 291 were upregulated and 74 were downregulated. Gene ontology (GO) enrichment analysis demonstrated that tumors of Anxa1ΔMɸ mice were differentially enriched for pathways associated with immune response and cellular response to IFN-β (figure 4A), which was further confirmed by gene set enrichment analysis (GSEA) (online supplemental figure S1K). We also performed quantitative proteomics analyses of tumors from Anxa1ΔMɸ and Anxa1WT mice. Consistent with RNA sequencing, 243 differentially expressed proteins (DEPs) were identified in tumors from Anxa1ΔMɸ as compared with Anxa1WT mice, including 175 that were upregulated and 68 that were downregulated (online supplemental figure S1L). GO enrichment analysis of the DEPs verified that tumors of Anxa1ΔMɸ mice were enriched for pathways associated with cellular response to IFN-β (figure 4B).

Figure 4

Macrophage Anxa1 deficiency activates cGAS-STING pathway to induce type I interferon in tumors. (A) GO enrichment analysis of upregulated genes in tumors from Anxa1ΔMɸ versus Anxa1WT mice. (B) GO enrichment analysis of upregulated proteins in tumors from Anxa1ΔMɸ versus Anxa1WT mice. (C) qPCR analysis of mRNA expression of Ifnb1, Usp18, Oas3 and Ifit1 in tumors from Anxa1ΔMɸ and Anxa1WT mice. Results represent the averages of three independent experiments. (D) Measurement of IFN-β protein in tumor homogenates. Tumors were collected 3 weeks after tumor cells implanted. (E, G) Immunoblot analysis of IFN-β, cGAS, STING, P65, p-P65TBK1, p-TBK1, IRF3 and p-IRF3 in tumor tissues from Anxa1ΔMɸ and Anxa1WT mice. (F) Shown are the image of the subcutaneous tumor of Anxa1ΔMɸ mice treated with IgG or anti-IFNAR1 (left). Quantification of tumor volume and weight are shown (right). Each group contained five mice. (H) qPCR analysis of mRNA expression of Ifnb1, Usp18, Oas3 and Ifit1 in tumors of Anxa1ΔMɸ mice treated with control solvent or H151. Results represent the averages of three independent experiments. (I) Measurement of IFN-β protein in tumor homogenates from Anxa1ΔMɸ mice treated with control solvent or H151. (J) Shown are the image of the subcutaneous tumor of Anxa1ΔMɸ mice treated with control solvent or H151 (left). Quantification of tumor volume and weight are shown (right). Each group contained five mice. (K) qPCR analysis of mRNA expression of Ifnb1, Usp18, Oas3 and Ifit1 in tumors of Anxa1ΔMɸ mice treated with control solvent or RU521. Results represent the averages of three independent experiments. (L) Measurement of IFN-β protein in tumor homogenates from Anxa1ΔMɸ mice treated with control solvent or RU521. (M) Shown are the image of the subcutaneous tumors of Anxa1ΔMɸ mice treated with control solvent or RU521 (left). Quantification of tumor volume and weight are shown (right). Each group contained five mice. (N) Shown are the image of the subcutaneous tumor size of Anxa1ΔMɸ mice treated with IgG or anti-CSF1R. Each group contained five mice. Quantification of tumor volume and weight that from Anxa1ΔMɸ mice treated with IgG or anti-CSF1R are shown. (O) qPCR analysis of mRNA expression of Ifnb1, Usp18, Oas3 and Ifit1 in tumors from Anxa1ΔMɸ mice treated with IgG or anti-CSF1R. Results represent the averages of three independent experiments. (P) Measurement of IFN-β protein in tumor homogenates from Anxa1ΔMɸ mice treated with IgG or anti-CSF1R. (Q) Immunoblot analysis of c-Casp3 and STING in tumor tissues from Anxa1ΔMɸ that treated IgG and anti-CSF1R mice. (R) Immunoblot analysis of ANXA1 in THP1 cells that transfected with lentiviruses control (THP1-ctrl) and carrying plasmids knockdown ANXA1 (THP1-shANXA1). (S) Representative images of tumors from the humanized mouse model of PDAC that ANXA1 was knocked down in macrophages (left). Quantification of tumor volume and weight are shown (right). Each group contained five mice. (T) Immunoblot analysis of c-Casp3, STING and IFN-β expression in tumors from THP1-ctrl and THP1-shANXA1 mice. (U) Immunoblot analysis of ANXA1 in THP1 cells that transfected with lentiviruses control (THP1-ctrl) and carrying plasmids overexpressing ANXA1 (THP1-ANXA1). (V) Representative images of tumors from the humanized mouse model of PDAC that ANXA1 was overexpressed in macrophages (left). Quantification of tumor volume and weight are shown (right). Each group contained five mice. (W) Immunoblot analysis of c-Casp3, STING and IFN-β expression in tumors from THP1-ctrl and THP1-ANXA1 mice. The Panc02 cells line was used in the mouse PDAC tumor model involved in (A–Q). The data presented are the means±SD. ANXA1, Annexin A1; c-Casp3, cleaved Caspase-3; CSF1R, colony-stimulating factor 1 receptor; GO, gene ontology; IFN, interferon; IFNAR1, IFN-alpha/beta receptor 1; mRNA, messenger RNA; PDAC, pancreatic ductal adenocarcinoma; qPCR, quantitative PCR; THP1, human monocytic cell line.

For additional confirmation of the role of IFN signaling, we performed reverse-transcriptase (RT)-quantitative real-time PCR (qPCR) assays. The results showed that the mRNA levels of multiple IFN-stimulated genes (ISGs), including Ifnb1, Usp18, Oas3, and Ifit1, were robustly increased in tumors of Anxa1ΔMɸ as compared with Anxa1WT mice (figure 4C). ELISA and WB assays further confirmed that IFN-β protein levels were significantly increased in tumors from Anxa1ΔMɸ mice (figure 4D,E, online supplemental figure S1M). In addition, we established a mouse pancreatic cancer model using the KPC cell line in Anxa1ΔMɸ and Anxa1WT mice. As expected, downregulated expression of ANXA1 in macrophages led to significantly suppressed tumor growth in vivo and increased IFN-β protein levels (online supplemental figure S2A,B). To determine whether the increased type I IFN signaling in Anxa1ΔMɸ tumors is functionally relevant in terms of the antitumor effect, we used a function-blocking antibody against IFN-alpha/beta receptor 1 (IFNAR1). Anti-IFNAR1 treatment completely abolished the antitumor activity of macrophage ANXA1 deficiency in both Panc02 and KPC tumor-bearing mice (figure 4F, online supplemental figure S2C). Anti-IFN treatment had no effect on the attenuation of efferocytosis caused by the lack of ANXA1 in macrophages, and WB results also confirmed that c-Casp3 protein levels were not changed significantly after anti-IFNAR1 treatment (online supplemental figure S2D). Therefore, the antitumor effect of ANXA1 deficiency in macrophages is mediated via enhanced type I IFN signaling.

Damage to the apoptotic cell membrane integrity is known to induce cGAMP release into the tumor microenvironment, which triggers the STING pathway.22–24 Furthermore, the cGAS-STING signaling pathway has been reported to represent a key mechanism driving antitumor type I IFN response.18 Therefore, we speculated that the cGAS-STING pathway may be involved in macrophage ANXA1 deficiency-induced type I IFN response in tumors. To assess this possibility, we performed a WB analysis of tumor tissues from both Panc02 and KPC tumor-bearing mice. The protein levels of cGAS, STING, p-TBK1, p-IRF3, and p-P65 were significantly increased in tumor tissues of Anxa1ΔMɸ compared with Anxa1WT mice (figure 4G, online supplemental figure S2E–I). To evaluate the role of cGAS-STING signaling in deficiency-induced type I IFN response, Anxa1ΔMɸ mice bearing subcutaneous Panc02 and KPC tumors were subjected to administration of H151, a selective and covalent antagonist of STING, or RU521, a potent and selective cGAS inhibitor. Consistently, the mRNA expression of ISGs and the levels of IFN-β protein in tumors were obviously decreased by H151 treatment (figure 4H,I, online supplemental figure S2J). Furthermore, H151 suppressed the antitumor effect induced by ANXA1 deficiency (figure 4J, online supplemental figure S2K). The type I IFN response in Anxa1ΔMɸ mouse tumors was obviously decreased after RU521 treatment (figure 4K,L, online supplemental figure S2L). Moreover, treatment with RU521 suppressed the antitumor effects caused by ANXA1 deficiency (figure 4M, online supplemental figure S2M). Therefore, these results support the role of the cGAS-STING pathway in mediating the type I IFN response induced by ANXA1 deficiency.

To further corroborate the predominant role of TAMs in producing IFN-β, we injected mice with an antibody against colony-stimulating factor 1 receptor (CSF1R), which is essential for macrophage survival. Indeed, CSF1R inhibitor-mediated macrophage depletion abrogated the antitumor activity induced by ANXA1 deficiency in both Panc02 and KPC tumor-bearing mice (figure 4N, online supplemental figure S2N). Furthermore, the mRNA expression of ISGs was obviously decreased in anti-CSF1R tumors (figure 4O). Consistently, treatment with anti-CSF1R antibody decreased the IFN-β protein levels in tissue homogenates prepared from freshly dissected tumors (figure 4P, online supplemental figure S2O). WB results demonstrated an elevation of c-Casp3 in tumors from Anxa1ΔMɸ mice following treatment with anti-CSF1R, while STING protein levels were reduced. (figure 4Q, online supplemental figure S2P,Q). These results confirm that TAMs are the major source of IFN-β in Anxa1ΔMɸ mouse tumors.

Additionally, a human monocytic cell line (THP1) was used to confirm the roles of ANXA1. THP1 cells were transfected with a lentiviral control and lentiviral shANXA1. The WB result confirmed that the expression of ANXA1 was effectively modulated in THP1 cells (figure 4R, online supplemental figure S3A). On induction to M0, the cells were co-inoculated with PANC-1 cells subcutaneously into immunodeficient mice, establishing a humanized mouse model of pancreatic cancer.25–27 Consistent with the results in both Panc02 and KPC tumor-bearing mice, downregulated expression of ANXA1 in macrophages led to significantly suppressed tumor growth in the humanized mouse model of PDAC (online supplemental figure 4S). WB assay results validated that the downregulation of ANXA1 by macrophages within tumors resulted in significant upregulation of c-Casp3, STING, and IFN-β protein levels (figure 4T, online supplemental figure S3B–D). To further verify the modulatory role of ANXA1 on macrophage function in the humanized mouse model, a THP1 cell line overexpressing hominine ANXA1 and a control were constructed. WB results confirmed the successful overexpression of ANXA1 (figure 4U, online supplemental figure S3E). As expected, overexpression of ANXA1 in macrophages led to significantly promoted tumor growth (figure 4V). The WB assay further validated that tumors with macrophages overexpressing ANXA1 exhibited significantly decreased levels of c-Casp3, STING, and IFN-β protein (figure 4W, online supplemental figure S3F–H).

Macrophage ANXA1 deficiency promotes remodeling of T cells in the tumor microenvironment

To identify changes in the tumor microenvironment caused by macrophage ANXA1 deficiency, we harvested tumors on 3 weeks of the Panc02 tumor model using syngeneic Anxa1ΔMɸ and Anxa1WT mice and performed scRNA-seq. Next, we applied Uniform Manifold Approximation and Projection (UMAP) technology for dimensionality reduction clustering and cell cluster annotation based on canonical markers, which revealed nine main groups: cancer cells, endothelial cells, fibroblasts, B cells, T and natural killer (NK) cells, neutrophils, mast cells, mononuclear phagocytes, and plasmacytoid dendritic cells (online supplemental figure S4A). The proportions of each of these cell subpopulations and the bubble map of marker genes are shown in online supplemental figure S4B,C.

For a more detailed understanding of the immune populations in the tumor microenvironment, we computationally separated the UMAP clusters and reanalyzed the data. The T and NK cells were categorized into eight distinct subclusters broadly defined by the distribution of classical marker genes (figure 5A). Of note, the proportions of NK cells, CD8 naive T cells and CD8 effector T cells (Teff) were higher in tumors from Anxa1ΔMɸ as compared with Anxa1WT mice, while the proportion of CD8 exhausted T cells (Tex) was lower (figure 5B). To gain further insight into the relationships among CD8+ T cells, we used Monocle 2 to infer the temporal dynamics of biological processes from the transcriptional similarities among cells. Pseudotime analysis for CD8+ T cells indicated that the starting point corresponded to CD8 naive T cells, followed by CD8 Teff, and ending with CD8 Tex (figure 5C). The group density stream verified the temporal order of these CD8 cell subpopulations and further confirmed that CD8 Teff cells were mainly distributed in Anxa1ΔMɸ mice, while CD8 Tex cells were mainly distributed in Anxa1WT mice (figure 5D). Furthermore, GSEA revealed that NK and CD8 Teff cells displayed a higher cytotoxicity score and a lower exhaustion score in Anxa1ΔMɸ mice (figure 5E,F). These results suggest that the T cells in Anxa1ΔMɸ mice tended toward higher toxicity than the T cells in Anxa1WT mice.

Figure 5

scRNA-seq analysis reveals the changes of T cells, NK cells and macrophages in tumor induced by macrophage Anxa1 deficiency. (A) UMAP plot showing T and NK cell clusters in tumor. (B) Fraction of eight clusters in each group. (C) tSNE plot with analysis of T and NK cells by Monocle 2 (the small picture in the middle). T and NK subpopulations overlaid on Monocle 2 pseudotime plot. (D) Pseudotime plot showed the proportion of each CD8 T cell type at different time points. The vertical coordinate is the pseudotime series, and the horizontal coordinate is the proportion of cell types at different time points. (E–F) Comparative cytotoxicity and exhaustion scores. The bar graph illustrates the cytotoxicity and exhaustion scores of two groups, across four cell types: CD8 Tex, CD8 naive T, CD8 Teff and NK cells. The y-axis represents the cytotoxicity score. (G) The number and proportion of clone types in different groups are displayed. The size of the circle represents the percentage of clone types (in the legend on the right). Single represents a single clone type; medium indicates that the clone frequency is greater than one and less than or equal to 10. Large indicates that the frequency of clone types is greater than 10. (H) Representative flow cytometry plot (up) and tabulated percentages (below) of NK cells in CD45+ TILs and Ki67, GZMB, IFN-γ in CD8+ TILs (n=4 per group). The data presented are the means±SD. Anxa1, Annexin A1; IFN, interferon; NK, natural killer; scRNA-seq, single-cell RNA sequencing; TILs, tumour-infiltrating lymphocytes; UMAP, Uniform Manifold Approximation and Projection.

For additional verification of the T-cell populations in Anxa1ΔMɸ mice, we performed an integrative analysis of the T cell receptor (TCR) repertoires represented in the scRNA-seq data. There were more clonotypes with a clonal frequency of greater than 10 in Anxa1ΔMɸ mice than in Anxa1WT mice, especially for the Teff cells, which is indicative of enhanced clonal expansion of effector T cells in Anxa1ΔMɸ mice (figure 5G). Using the cell–cell communication analytical tool in CellPhoneDB, we further found that there were more interactions between Teff cells and cancer cells in Anxa1ΔMɸ mice than in Anxa1WT mice, which is indicative of an enhanced immune response mechanism (online supplemental figure S4D). We also used flow cytometry to detect the status of NK and T cells in the tumors of two groups of mice (online supplemental figure S5A). The tumors of Anxa1ΔMɸ compared with