Targeting tumor exosomal circular RNA cSERPINE2 suppresses breast cancer progression by modulating MALT1-NF-𝜅B-IL-6 axis of tumor-associated macrophages

Breast cancer tissues were obtained from surgical specimen archives of 136 patients from the First Affiliated Hospital of Gannan Medical University, between 2009 and 2012. The inclusion criteria of patients were as follow: i) Female; ii) Diagnosed as breast cancer by pathological confirmation; iii) No adjuvant treatment including chemotherapy, radiotherapy, hormone therapy or immunotherapy before surgery; iv) Complete clinical and pathological data. The patients had the exclusion criteria as follows: i) Male; ii) Combined with other tumors and serious diseases; iii) incomplete clinical and pathological data. Frozen tissues were used for quantitative reverse transcription PCR (qRT-PCR) analyses, while paraffin-embedded tissues were used for FISH and IHC analyses.

All samples were collected from patients with informed consent, and all related procedures were performed with the approval of the internal review and ethics boards of the First Affiliated Hospital of Gannan Medical University. The study was compliant with all relevant ethical regulations regarding research involving human participants.

When equal cell numbers of tumor cells (MCF-7 and EO771 cells) were seeded and reached 70% confluency in 150 mm culture dishes, cells were washed with PBS and cultured in exosome-depleted media to isolate exosomes after 48–72 h of further culture as previously described [14, 21]. Briefly, equal volume of conditioned media were collected and centrifuged at 3000 rpm for 10 min, followed by a centrifugation step of 10,000 × g for 30 min at 4℃ to remove cellular debris. Subsequently, the supernatant was filtered through a 0.22- μm filter. Exosomes were ultracentrifuged at 100,000 × g for 90 min and resolved in 100 μl PBS.

For exosomes protein quantification, we used 10 × RIPA buffer to lyse exsomes. Then, micro BCA Protein Assay Reagent Kit (Cat#23,235, Invitrogen, USA) were used to quantificate the protein concentration of lysed exosomes. Additionally, exosomes were examined by electron microscopy using negative staining, and quantified using a NanoSight NS300 instrument (Malvern Instruments) equipped with NTA 3.0 analytical software.

The animal study procedures were approved by the Animal Care and Use Committee of Sun Yat-sen University. 4–6 weeks-old female C57BL/6, Csf1^op/op and corresponding wide-type mice were randomized to each experimental group. For orthotopic transplantations, 10⁶ EO771 cells transduced with lentivirus carrying control or cSERPINE2 shRNA and with or without cSERPINE2 overexpression were resuspended in 100μl of PBS and injected into the fourth mammary fat pads on one flank of the mice.

In some experiments, intraperitoneal injection was performed with 200μg of IL-6 antibody or control IgG antibody three times per week as previously described [22]. For the exosome injection mouse breast cancer model, exosomes obtained from cells with a dose of approximately 10¹⁰ were resuspended in 100μl PBS per mouse, which was injected into mice every 3 days as previously described [23, 24].

The tumor volumes were calculated by the formula: volume = width² × length/2. The maximum permitted tumor diameter of 20 mm in any dimension was never exceeded. We recorded the tumor volumes every 3 days. Additionally, mouse tumor burden was monitored by weekly bioluminescence imaging using the IVIS-200 Lumina Imaging System (Xenogen). Lung metastatic lesions were confirmed by histological analysis.

Additional methods information can be found in Supplementary Materials.

A number of host genes of circRNAs play oncogenic roles in tumorigenesis, but it remains unclear whether oncogenes-derived circRNAs affect cancer progression. Given that SERPINE2 overexpression programmed breast cancer cells for vascular mimicry and contributed to distant metastasis [20], we wondered whether SERPINE2-derived circRNAs exerted influence on breast cancer progression. We first analyzed the expression of 8 SERPINE2-derived circRNAs in breast cancer and found that only hsa_circ_0001103 (cSERPINE2) was substantially upregulated (Fig. 1a, b). Consistently, the expression of cSERPINE2 was much higher in breast cancer tissues than in paired adjacent tissues in our cohort (Fig. 1c). Subsequently, we explored the relationship between cSERPINE2 expression and the clinicopathological characteristics of 136 breast cancer patients and found that cSERPINE2 might participate in the progression of breast cancer (Supplementary Table S1). Notably, breast cancer patients with larger tumor sizes and lymph node metastasis exhibited higher expression of cSERPINE2 (Fig. 1d, e). Moreover, Kaplan–Meier survival analysis revealed that breast cancer patients with higher cSERPINE2 expression exhibited shorter overall survival (OS) and recurrence-free survival (RFS) than those with lower cSERPINE2 expression (Fig. 1f). Additional Cox proportional hazards regression models suggested that upregulated cSERPINE2 expression was an independent prognostic indicator for OS and RFS in breast cancer patients (Supplementary Table S2 and S3).

cSERPINE2, derived from exon 2 and exon 3 of the SERPINE2 gene, formed a circular transcript of 509 nt (Fig. 1g). We investigated the expression of cSERPINE2 in breast cancer cell lines and normal breast cell line MCF-10A and found that cSERPINE2 was significantly upregulated in breast cancer cell lines (Supplementary Fig. S1a). For further characterization of cSERPINE2, RNase R treatment and half-life assay consistently showed that cSERPINE2 was much more stable than the linear transcript of SERPINE2 in MCF-7 and MDA-MB-468 cells (Supplementary Fig. S1b, c). Subsequent cell fractionation and fluorescent in situ hybridization (FISH) assays showed that cSERPINE2 was predominantly localized in the cytoplasm (Supplementary Fig. S1d, Fig. 1h). Collectively, these data suggest that cSERPINE2 is frequently upregulated in breast cancer, and it could serve as a potential indicator for unfavorable prognosis.

RNA-binding proteins (RBPs) participate in the biosynthesis of circRNA by binding to specific sequences to regulate the back splicing of circRNAs [25, 26]. To investigate which RBPs contributed to the formation of cSERPINE2, we chose 6 RBPs previously reported to participate in circRNA formation in cancers and detected the expression of cSERPINE2 after knocking down these RBPs (Supplementary Fig. S1e). Interestingly, we found that knocking down EIF4A3 significantly inhibited the expression of cSERPINE2 in MDA-MB-468 cells (Fig. 1i). Moreover, down-regulation of cSERPINE2 expression in EIF4A3 silenced cells could be restored by ectopic reintroduction of EIF4A3, indicating that EIF4A3 was involved in cSERPINE2 biogenesis (Supplementary Fig. S1f-h). EIF4A3 is a core component of the exon junction complex, which plays an essential role in pre-mRNA splicing [27]. To further explore the mechanisms by which EIF4A3 regulates cSERPINE2 biogenesis, we used CircInteractome to predict EIF4A3 binding sites and found two binding sites of EIF4A3 in the upstream flanking intron and across the self-exon 2 region of SERPINE2 pre-mRNA. Subsequently, we truncated the full length, upstream and downstream region of cSERPINE2 into four segments. RNA immunoprecipitation (RIP) assays indicated that EIF4A3 could bind the full-length sequence of cSERPINE2 (named “b”) and the upstream flanking sequence (named “d”) but not the other sites (“a” and “c”) (Fig. 1j). These results were further confirmed by RNA pull-down assays using in vitro transcript RNA fragments of the corresponding cSERPINE2 pre-mRNA (Fig. 1k).

In addition, the expression of cSERPINE2 in tumor tissues was positively related to the level of EIF4A3 in 27 clinical breast cancer specimens, suggesting that EIF4A3 was associated with cSERPINE2 expression (Fig. 1l). Furthermore, the expression of EIF4A3 was significantly increased in breast cancer tissues compared to normal tissues according to TCGA database analysis (Supplementary Fig. S1i). Kaplan–Meier survival analyses indicated that higher EIF4A3 expression was associated with poor OS and RFS in the breast cancer Kaplan–Meier plotter database (Supplementary Fig. S1j). Taken together, our findings suggest that EIF4A3 may facilitate the formation of cSERPINE2 by binding to flanking sequences.

To further investigate the biological functions of cSERPINE2 in breast cancer cells, we constructed stable cSERPINE2 knockdown cell lines using short hairpin RNAs that specifically targeted the back-splicing junction site of cSERPINE2. In addition, we also successfully constructed stable ectopic cSERPINE2 overexpression cell lines by transduction with the full-length cDNA of the cSERPINE2 lentiviral vector. The transduction efficiency was validated by qRT-PCR, but knocking down or overexpressing cSERPINE2 did not affect the level of SERPINE2 (Supplementary Fig. S2a). Unexpectedly, neither silencing nor overexpressing cSERPINE2 affected tumor cell proliferative and invasive capacity in vitro (Supplementary Fig. S2b, c). We observed high conservation between human and mouse cSERPINE2 through sequence alignment identification (Supplementary Fig. S2d). Therefore, we further explored the role of cSERPINE2 in vivo. Mouse cSERPINE2 was stably knocked down and overexpressed in EO771 cells which previously labeled with luciferase (Supplementary Fig. S2e), followed by orthotopic implantation into C57BL/6 mice. Contrary to in vitro results, tumor growth was significantly inhibited by cSERPINE2 silencing and accelerated by overexpression of cSERPINE2 (Fig. 2a). Furthermore, fewer formed lung metastases were detected following cSERPINE2 knockdown, while cSERPINE2 ectopic expression facilitated the occurrence of lung metastases (Fig. 2b). However, in vitro gain- and loss- of function assays showed that cSERPINE2 had no effect on the proliferative and invasive capacity of EO771 cells (Supplementary Fig. S2f, g). The substantial difference between cSERPINE2 function in vivo and in vitro indicated that cSERPINE2 might mainly influence the tumor immune microenvironment rather than the tumor cells themselves.

In an analysis of the association of CD45⁺ tumor-infiltrating leukocytes (CD8⁺ T cells, CD19⁺ B cells, CD56⁺ NK cells and CD68⁺ macrophages) and the cSERPINE2 level in breast cancer tissues (n = 27), immunohistochemistry (IHC) staining suggested that higher cSERPINE2 expression was notably associated with increasing CD68⁺ macrophages infiltration (Fig. 2c, d). Consistently, F4/80⁺ macrophages infiltration was elevated in the cSERPINE2-overexpressing xenograft tumors but decreased in the cSERPINE2-silenced xenograft tumors (Fig. 2e). To further investigate the role of macrophages in cSERPINE2-induced carcinogenesis, we constructed the Csf1^op/op mice that lack macrophages with mutated macrophage colony stimulating factor 1 (Csf1) and then constructed orthotopic tumor xenograft models [28]. As predicted, no significant differences in tumor size or lung metastasis formation were found in Csf1^op/op mice after transplantation of cSERINE2-silenced EO771 cells (Fig. 2f, g). Collectively, these results indicate that cSERPINE2 may promote breast cancer progression by increasing TAMs infiltration.

To further elucidate the effects of tumor cSERPINE2 on TAMs, we induced THP-1 monocytic cells to differentiate into macrophages by stimulation with PMA for 24 h, and these THP-1-derived macrophages (PMA-THP-1) were further stimulated with cell-free conditioned media (CM) from MCF-7 cells transduced with empty vector or cSERPINE2 overexpression vector to obtain TAMs (TAM^EV−CM or TAM^OE−CM) (Fig. 3a). Moreover, we isolated and differentiated murine bone marrow derived macrophages (BMDMs) from C57BL/6 mice, and then exposed them to cell-free CM collected from EO771 cells transduced with empty vector or cSERPINE2 overexpression vector to obtain BMDMs derived TAMs (mTAM^EV−CM or mTAM^OE−CM). Consistent with previous studies [9, 29], THP-1 and BMDM-derived TAMs both exhibited a mixed M1/M2 phenotype, which not only showed higher levels of the M2 markers CD163 and CD206 but also increased the expression of proinflammatory genes, including TNF-α, IL-1β and IL-6 (Supplementary Fig. S3a, b). Since cSERPINE2 promoted breast cancer progression through TAMs in vivo, we constructed a coculture system in which tumor cells (MCF-7 and EO771) were cocultured with THP-1 and BMDM derived TAMs for 24 h, and subsequently, tumor cells were collected for EdU staining and transwell invasion assays (Fig. 3a). We found that the proliferative and invasive capacities of tumor cells were obviously enhanced after coculturing with TAMs, especially coculturing with TAM^OE−CM, suggesting that TAMs promoted the proliferation and invasion of tumor cells and that this promotion effect was enhanced by cSERPINE2 overexpression in tumor cells (Supplementary Fig. S3c, d).

Previous studies have shown that tumor exosomal RNAs promoted tumor progression by regulating the interaction between tumor cells and TAMs [13, 30]. Hence, we speculated that cSERPINE2 could be shuttled from breast cancer cells to TAMs via exosomes and further educated TAMs, which impacted the proliferation and invasion of tumor cells. Therefore, we collected exosomes in CM from breast cancer cells using ultracentrifugation and the structure and size distribution of enriched exosomes were analyzed by transmission electron microscopy (TEM) and nanoparticle tracking analysis (Fig. 3b, c). Additionally, the exosomal markers CD63 and CD81 were detected on enriched exosomes (Fig. 3d). More importantly, the cSERPINE2 expression in enriched exosomes was positively related to the corresponding tumor cells, indicating that cSERPINE2 could be effectively packed into exosomes (Fig. 3e). Moreover, cSERPINE2 expression was significantly increased in TAMs, especially in TAM^OE−CM and mTAM^OE−CM, suggesting that tumor exosomal cSERPINE2 could be internalized by macrophages (Fig. 3f). Furthermore, we inhibited exosome secretion of tumor cells by applying GW4869 or depleted exosomes from CM by ultracentrifugation, and found that the expression of cSERPINE2 was largely decreased in TAM^OE−CM or mTAM^OE−CM (Fig. 3f). These results indicated that cSERPINE2 could be encapsulated into tumor exosomes and further internalized by TAMs.

To further investigate the function of tumor exosomal cSERPINE2 internalized by TAMs, we stimulated PMA-THP-1 cells and BMDMs with exosomes from tumor cells transduced with empty vector or cSERPINE2 overexpression vector to acquire corresponding TAMs (TAM^EV/OE−Exo or mTAM^EV/OE−Exo), respectively. Subsequently, tumor exosome-induced TAMs were cocultured with tumor cells for 24 h, and then, these tumor cells were collected for EdU staining and transwell invasion assays (Fig. 3a). Similar to the aforementioned results, the proliferation and invasion of MCF-7 and EO771 cells were notably increased after coculturing with exosome- pulsed TAMs, especially coculturing with TAM^OE−Exo and mTAM^OE−Exo (Fig. 3g-j). Taken together, our findings suggest that TAMs promote the proliferative and invasive behaviors of breast cancer cells by internalizing tumor exosomal cSERPINE2.

Since sponging miRNAs is a major mechanism of circRNAs [31], we analyzed the CircInteractome, circBank and StarBase online databases, and eventually identified 3 miRNAs as possible targets of cSERPINE2 (Fig. 4a). Moreover, we performed a circRNA pull-down assay to identify miRNAs interacted with cSERPINE2 in TAM^OE−Exo and found that miR-513a-5p could be pulled down by a biotin-labeled cSERPINE2 probe (Fig. 4b). Additionally, the miRNA pull-down assay further confirmed obvious enrichment of cSERPINE2 using biotin-miR-513a-5p in TAM^OE−Exo (Fig. 4c). Furthermore, we constructed the full-length cSERPINE2 sequence (cSERPINE2-WT) and cSERPINE2 sequences with mutant binding sites (cSERPINE2-MUT) followed by a luciferase reporter assay (Fig. 4d). Our results demonstrated that miR-513a-5p mimic and inhibitor significantly reduced and increased the luciferase activity of the luciferase reporter plasmid with cSERPINE2-WT respectively, while no change in the luciferase activity of the luciferase reporter plasmid with cSERPINE2-MUT was found (Fig. 4d and Supplementary Fig. S4a). Moreover, FISH assay suggested that cSERPINE2 colocalized with miR-513a-5p in the cytoplasm of TAM^EV−Exo and TAM^OE−Exo (Fig. 4e and Supplementary Fig. S4b). These results demonstrate that miR-513a-5p has a direct interaction with cSERPINE2 in TAMs.

To better understand the functions and mechanisms of cSERPINE2 in TAMs, we analyzed TAM transcriptomes by RNA-seq to compare differentially expressed genes (DEGs) between TAM^EV−Exo and TAM^OE−Exo. Heatmap showed the top 200 upregulated and downregulated genes (Supplementary Fig. S4c). KEGG analysis showed that the upregulated genes were prominently enriched in the NF-\(\upkappa\)B pathway (Fig. 4f). Subsequently, we used StarBase and TargetScan to identify the potential target genes of miR-513a-5p and overlapped the potential target genes with the 52 upregulated genes involved in the NF-𝜅B pathway. Finally, 7 DEGs were identified as potential targets of the cSERPINE2/miR-513a-5p axis (Fig. 4g). Among the 7 DEGs, the miR-513a-5p in vivo pull-down assay revealed that the relative MALT1 enrichment bound by miR-513a-5p was prominently decreased in TAM^OE−Exo compared to TAM^EV−Exo, suggesting that increased cSERPINE2 prevented miR-513a-5p binding to MALT1 (Fig. 4h). To further confirm that miR-513a-5p interacted with MALT1, we constructed a luciferase reporter gene containing the wild-type (WT) or mutant (MUT) MALT1 sequences. MiR-513a-5p mimics and inhibitor reduced and increased the luciferase activity, respectively, from 293 T cells transfected with MALT1-WT but not in MALT1-MUT group (Fig. 4i and Supplementary Fig. S4d). Furthermore, the expression of MALT1 was obviously upregulated in TAM^OE−Exo compared to TAM^EV−Exo, and the upregulation of MALT1 was retarded by the miR-513a-5p mimic in TAM^OE−Exo cells (Fig. 4j).

The paracaspase MALT1, a key regulator of immune responses, could induce a component of the signaling pathway mediating antigen receptor-dependent stimulation of the transcription factor NF-𝜅B [32, 33]. Although abnormal MALT1 expression is closely associated with lyphomagenesis and antoimmune diseases, the function of MALT1 in breast cancer is still unclear [34‐36]. MALT1 expression has been noted in several types of cancer, with low expression levels in breast cancer indexed in the Human Protein Atlas database (Supplementary Fig. S4e). In our cohorts of breast cancer patients, MALT1 expression was barely present in tumor cells but was enriched in CD68⁺ TAMs in breast cancer tissues with high expression cSERPINE2 (Fig. 4k and Supplementary Fig. S4f). Moreover, enforced cSERPINE2 expression had no effect on MALT1 expression in MCF-7 and EO771 cells, but obviously promoted MALT1 expression in PMA-THP-1 and BMDMs cells (Supplementary Fig. S4g-h). Collectively, these results suggest that the MALT1 levels are regulated by tumor exosomal cSERPINE2 by sponging miR-513a-5p in TAMs, while this regulatory mechanism of MALT1 is not present in breast cancer cells.

We observed that the IL-6 expression was substantially upregulated in TAM^OE−CM compared to TAM^EV−CM, indicating that the expression of IL-6 in TAMs may be regulated by tumor exosomal cSERPINE2 (Supplementary Fig. S3a, b). Indeed, ELISA results confirmed that the level of IL-6 was significantly increased in the supernatants of TAM^OE−Exo compared to TAM^EV−Exo (Supplementary Fig. S5a). Since MALT1 promoted IL-6 secretion by activating the NF-\(\upkappa\)B pathway in macrophages, we asked whether tumor exosomal cSERPINE2 promoted IL-6 secretion of TAMs through activation of the NF-𝜅B pathway [37, 38]. As expected, silencing MALT1 or the additional treatment with PDTC (NF-κB inhibitor) in TAM^OE−Exo inhibited the increasing IL-6 secretion (Supplementary Fig. S5a). Furthermore, western blotting indicated that the levels of phosphorylated I𝜅Bα and IKK-β were notably increased in TAM^OE−Exo, while these effects were retarded by MALT1 knockdown or the addition of PDTC treatment (Supplementary Fig. S5b-c). Similar findings were obtained in mTAMs using the above mentioned assays and western blotting quantification analysis (Fig. 5a-b and Supplementary Fig. S5d). Moreover, p65 nuclear translocation was notably increased in mTAM^OE−Exo, while silencing MALT1 or adding PDTC treatment significantly inhibited p65 nuclear translocation (Fig. 5c). These results suggested that tumor exosomal cSERPINE2 promoted the IL-6 expression by upregulating MALT1 and activating the NF-\(\upkappa\)B pathway in TAMs.

Cytokine secretion represents the major functional response of macrophages [8]. Next, we asked whether the increased IL-6 levels were responsible for the tumor-promoting function of tumor exosomal cSERPINE2. Breast cancer orthotopic xenograft model was constructed with EO771 cells transduced with empty vector or cSERPINE2 overexpression vector followed by intraperitoneal injection of IgG or anti-IL-6 antibody. Surprisingly, the promoting effect of cSERPINE2 overexpression on tumor growth and metastasis was blocked by anti-IL-6 antibody (Fig. 5d, e). Furthermore, IHC staining revealed that the increased expression of Ki67 and CD44 (invasive marker) and the elevated infiltration of F4/80⁺ macrophages in cSERPINE2 overexpressing tumor tissues were blocked by anti-IL-6 antibody (Fig. 5f). In particular, we injected exosomes enriched from EO771 cells transduced with empty vector or cSERPINE2 overexpression vector (Exo^EO771−EV or Exo^EO771−OE) followed by injection of IgG or anti-IL-6 antibody in orthotopic xenograft models. Notably, the tumor size and lung metastatic formation largely increased after injection of Exo^EO771−OE in the IgG group, while no significant difference was observed after injection of anti-IL-6 antibody (Fig. 5g, h). Additionally, the expression of Ki67 and CD44 and the infiltration of F4/80⁺ macrophages were not affected by treatment of Exo^EO771−OE in the anti-IL-6 antibody group (Supplementary Fig. S5e).

IL-6 usually binds to its receptor complex IL6R/gp130 and activates downstream Janus kinases (JAKs), which subsequently activate STAT3 through phosphorylation of tyrosine 705 [39]. Unsurprisingly, analysis of harvested xenograft tumors showed that the levels of p-JAK2 and p-STAT3 notably increased after injection of Exo^EO771−OE in the IgG group, while no significant changes were observed in the anti-IL-6 antibody group (Supplementary Fig. S5f-g). Consistently, the levels of p-JAK2 and p-STAT3 notably increased in EO771 cells after coculturing with mTAM^OE−Exo in contrast to mTAM^EV−Exo followed by IgG treatment, whereas anti-IL-6 antibody abrogated these effects (Supplementary Fig. S5h-i). Together, these results demonstrate that tumor exosomal cSERPINE2 promotes the secretion of IL-6 in TAMs by activating the NF-\(\upkappa\)B pathway to encourage breast cancer progression.

To better uncover the mechanism of IL-6/STAT3 signaling in TAM-educated tumor cells, we performed RNA-seq on MCF-7 cells after coculturing with TAM^EV−Exo or TAM^OE−Exo. Transcriptional profiling revealed that 590 genes were upregulated and 691 genes were downregulated (|log₂FC|≥ 1; FDR ≤ 0.05) (Fig. 6a). More interestingly, we found that EIF4A3 and CCL2 were prominently upregulated in MCF-7 cells educated with TAM^OE−Exo compared to those educated with TAM^EV−Exo. Gene set enrichment analysis (GSEA) analysis confirmed that the JAK2-STAT3 pathway was enriched as the top pathway in MCF-7 cells after coculturing with TAM^OE−Exo (Fig. 6b).

As described above, the RNA binding protein EIF4A3 is considered as an important regulator of post-transcriptional regulatory processes [40], and CCL2, as a powerful macrophage chemokine, was reported to be a target of STAT3 in breast cancer [41]. Here, we further confirmed that the EIF4A3 and CCL2 levels were notably increased in MCF-7 cells after coculturing with TAM^OE−Exo (Fig. 6c, d and Supplementary Fig. S6a). Nevertheless, ectopic expression of cSERPINE2 in MCF-7 cells did not regulate the expression of EIF4A3 or CCL2 or activation of the JAK2-STAT3 pathway, which indirectly indicated that cSERPINE2 overexpression did not affect tumor cell proliferative and invasive capacity in vitro (Supplementary Fig. S6b-d). Consistent with the above data, silencing STAT3 in the MCF-7 cells stimulated with IL-6 led to striking downregulation of EIF4A3 and CCL2, suggesting that STAT3 participated in the regulation of EIF4A3 and CCL2 (Fig. 6e and Supplementary Fig. S6e-f). By investigating the mechanisms by which STAT3 regulates EIF4A3 and CCL2 expression using the Jaspar database, we firstly found two binding sites (BS1/BS2) for STAT3 in the promoter of EIF4A3 (Fig. 6f). ChIP-qPCR revealed that MCF-7 cells cocultured with TAM^OE−Exo had significantly increased STAT3 occupancies on the BS1 and BS2 of the promoter of EIF4A3 (Fig. 6g). Furthermore, a luciferase plasmid with the top 2000 nucleotides of the promoter domain of the EIF4A3 gene (EIF4A3-WT) and a luciferase plasmid with mutant sequences in BS1 or/and BS2 of the promoter domain (EIF4A3-BS1/BS2-MUT) were generated and transfected into MCF-7 cells. Luciferase reporter assays revealed that the TAM^OE−Exo induced MCF-7 cells enhanced the luciferase activity of EIF4A3-WT, but mutation in either STAT3 binding site partially abrogated this promotion and mutations in both sites completely abrogated it (Fig. 6h). Similar to the results of EIF4A3, two binding sites (BS1/BS2) for STAT3 in the promoter of CCL2 were predicted in the Jaspar database (Supplementary Fig. S6g). ChIP-qPCR suggested that the STAT3 occupancies on the both BS1 and BS2 of the promoter of CCL2 were significantly increased in the TAM^OE−Exo induced MCF-7 cells (Supplementary Fig. S6h). Moreover, our results indicated that the supernatant from the MCF-7 cells cocultured with TAM^OE−Exo significantly promoted PMA-THP-1 migration compared to that of supernatant from the tumor cells cocultured with TAM^EV−Exo (Supplementary Fig. S6i-j).

We summarized our findings in a schematic (Fig. 6i). Briefly, breast cancer cells exosomal cSERPINE2 was shuttled to TAMs and notably elevated MALT1 levels, enhancing the secretion of IL-6 by activating the NF-\(\upkappa\)B pathway and leading to increased proliferation and invasion of breast cancer cells. More importantly, IL-6 in turn increased the EIF4A3 and CCL2 levels within tumor cells in a positive feedback mechanism, further enhancing tumor cSERPINE2 biogenesis and promoting the recruitment of TAMs.

To investigate the potential utility of cSERPINE2 as a therapeutic target in breast cancer, we applied a self-assembly strategy for formulating lipid-polymer hybrid NPs for systemic si-cSERPINE2 delivery, composed of the ionizable lipid-like compound G0-C14 for si-cSERPINE2 complexion, a biocompatible poly lactic-co-glycolic acid (PLGA) polymer to form a stable NP core to carry the G0-C14/si-cSERPINE2 and a lipid-poly (ethylene glycol) (lipid-PEG) layer for stability (Fig. 7a). The si-cSERPINE2 loaded NPs were ~ 114.63 nm in size, as measured by dynamic light scattering (DLS) (Fig. 7b). Their morphology was assessed by transmission electron microscopy and the zeta potential was ~ -9.09 mV (Fig. 7c, d). Moreover, there were no obvious changes in the size of si-cSERPINE2-NPs over a period of 96 h in the presence of 10% serum, indicating acceptable stability of these engineered NPs (Fig. 7e). Furthermore, the si-cSERPINE2-NPs efficiently silenced cSERPINE2 expression in tumor cells and more than 70% of cSERPINE2 expression could be inhibited at a siRNA dose of 30 nM (Fig. 7f). With a siRNA dose of 30 nM, si-cSERPINE2-NPs significantly silenced cSERPINE2 expression in tumor cells compared to si-ctrl-NPs (Fig. 7g and Supplementary Fig. S7a). We next examined whether si-cSERPINE2-NPs treatment of tumor cells could influence the cSERPINE2 and MALT1 levels in TAMs. As mentioned in the aforementioned assays, we found that cSERPINE2 and MALT1 expression was significantly decreased in the TAMs educated by exosomes of the tumor cells pretreated with si-cSERPINE2-NPs (TAM ^{si−cSERPINE2−NPs−Exo}) compared to TAMs educated by exosomes of the tumor cells pretreated with si-ctrl-NPs (TAM^{si−ctrl−NPs−Exo}) (Fig. 7h and Supplementary Fig. S7b-d). Moreover, the proliferative and invasive ability of tumor cells were obviously repressed after coculturing with TAM ^{si−cSERPINE2−NPs−Exo} compared to coculture with TAM^{si−ctrl−NPs−Exo} (Supplementary Fig. S7e-h).

To further evaluate the properties of si-cSERPINE2 NP delivery in vivo, we first conducted pharmacokinetics studies by injecting cy5-si-cSERPINE2-1-NPs or naked-cy5-si-cSERPINE2-1 into healthy C57BL/6 mice through the tail vein, and the results showed that the cy5-si-cSERPINE2-1-NPs had prolonged siRNA circulation (Supplementary Fig. S7i). Next, we investigated the biodistribution and tumor accumulation of these NPs in orthotopically incubated EO771 tumor-bearing mice. As shown in Supplementary Fig. S7j, cy5-si-cSERPINE2-1-NPs exhibited higher tumor accumulation. With the promising in vitro results, prolonged blood circulation and high tumor accumulation described above, the effects of si-cSERPINE2-NPs on breast cancer in vivo are promising.

To better illustrate the potential therapeutic effect of si-cSERPINE2-NPs on breast cancer in vivo, we randomly divided the tumor-bearing mice into four groups. The mice in the PBS, control NPs, si-ctrl NPs and si-cSERPINE2-1-NPs groups were injected via the tail vein with a siRNA dose of 700 μg/kg body weight every 3 days for 6 cycles after average tumor size had increased to ~ 120mm³ (about day 10) (Fig. 7i). At day 31, all mice were euthanized and we found that si-cSERPINE2-1-NPs obviously suppressed tumor growth (Fig. 7i). qRT-PCR results confirmed the consistent knockdown of cSERPINE2 in tumor tissues derived from the mice treated with si-cSERPINE2-1-NPs (Supplementary Fig. S8a). In addition, IHC staining showed that si-cSERPINE2-NPs treatment inhibited the expression of Ki67 and CD44 and the infiltration of F4/80⁺ macrophages (Fig. 7j).

Furthermore, we generated lung metastasis models to explore the effect of si-cSERPINE2-1-NPs in regulating the tumor metastasis in vivo. After 3 weeks of tail vein injection of EO771 cells, PBS, NPs, si-control NPs or si-cSERPINE2-1-NPs were injected into mice via tail vein every 3 days for 6 cycles (Fig. 7k). Surprisingly, we found that the lung metastatic tumor burden was significantly reduced in the si-cSERPINE2-1-NPs group compared to other groups (Fig. 7k). Regardless of the orthotopic tumor model or lung metastasis model mice, no remarkable changes in mouse body weight were found during the whole treatment period, indicating no apparent systemic toxicity for any treatment group (Supplementary Fig. S8b). Furthermore, the results of systemic toxicity revealed that intravenous administration of si-cSERPINE2-1-NPs exhibited no significant toxicity to major organs, and blood index analyses confirmed the absence of significant hepatotoxicity and renotoxicity (Fig. 7l and Supplementary Fig. S8c).