Neutrophil extracellular traps formed during chemotherapy confer treatment resistance via TGFβ activation

Summary

Metastasis is the major cause of cancer death and the development of therapy resistance is common. The tumor microenvironment can confer chemotherapy resistance (chemoresistance), but little is known about how specific host cells influence therapy outcome. We show that chemotherapy induces neutrophil recruitment and Neutrophil Extracellular Trap (NET) formation, which reduces therapy response in mouse models of breast cancer lung metastasis. We reveal that chemotherapy-treated cancer cells secrete IL-1β, which in turn triggers NET formation. Two NET-associated proteins are required to induce chemoresistance: integrin-αvβ1, which traps latent TGFβ, and matrix metalloproteinase 9, which cleaves and activates the trapped latent TGFβ. TGFβ activation causes cancer cells to undergo epithelial-to-mesenchymal transition and correlates with chemoresistance. Our work demonstrates that NETs regulate activities of neighboring cells by trapping and activating cytokines and suggests that chemoresistance in the metastatic setting can be reduced or prevented by targeting the IL-1β-NET-TGFβ axis.

Graphical Abstract

Mousset et al. show that chemotherapy promotes neutrophil extracellular trap (NET) formation which reduces therapy efficacy against breast cancer lung metastasis. NETs induce a TGFβ-dependent epithelial-mesenchymal transition (EMT) in cancer cells and pharmacological targeting of the NET/TGFβ-axis ameliorates chemotherapy efficacy.

Introduction

Breast cancer accounts for the highest incidence and death rates in women worldwide, and metastasis is responsible for most breast cancer-related deaths ¹. Most patients with metastases are treated with chemotherapy, but in the majority of cases, resistance to chemotherapy develops and it is a major challenge in treating metastatic cancers ^2,3. While cancer cell intrinsic factors, such as its genetic and epigenetic traits, play a role in therapy resistance, it is now well established that the tumor microenvironment (TME) modulates therapy response ⁴. Cancer extrinsic resistance is mediated by growth factors, cytokines and extracellular cellular matrix (ECM) proteins secreted by host cells. Among host cells, inflammatory cells, such as monocytes and macrophages, provide cues for the behavior of cancer cells during cancer progression ⁵ and have emerged as potent regulators of the therapeutic response in primary cancer ^6–11. Interestingly, chemotherapy has itself been linked to inflammation, which is considered a common, persistent, and potentially debilitating complication of chemotherapy ^12–14. Altogether, while inflammation has been linked to cancer progression, it is still unclear whether and how chemotherapy-induced inflammation is responsible for chemoresistance, and if so, by which mechanisms.

Despite the fact that neutrophils constitute the largest percentage of leukocytes in human blood and that they have well recognized roles in cancer progression, little is known about their roles in chemoresistance ^15,16. Outside of cancer, neutrophils can kill harmful microorganisms via (i) phagocytosis; (ii) degranulation of cytotoxic enzymes and proteases into the extracellular space; or (iii) formation of neutrophil extracellular traps (NETs). NETs are scaffolds of DNA associated with cytotoxic enzymes and proteases — such as Neutrophil Elastase (NE) and Matrix MetalloProteinase 9 (MMP9) — that are released into the extracellular space. NETs are generated through a signaling process that involves citrullination of histones by the Protein Arginine Deiminase 4 (PAD4) enzyme, chromatin decondensation, and disintegration of the nuclear membrane ¹⁷. Contents from the neutrophil’s secretory granules — including NE and MMP9 — associate with the decondensed chromatin. The plasma membrane ruptures, and the protease-associated chromatin fibers are finally released into the extracellular space. Recently, we and others determined that NETs have pro-metastatic activities ^18–22. In addition, NETs play a role in diseases involving tissue damage, including infectious and noninfectious inflammatory diseases, such as COVID-19, rheumatoid arthritis, and atherosclerosis ^23,24. Neutrophils are rapidly recruited to damaged tissue and, through the release of NETs, neutrophils crosstalk with both innate and adaptive immune cells during acute and chronic inflammation ²⁵. Chemotherapy, when effective, causes cell damage and death, and accordingly, high neutrophil infiltration is associated with poor response to chemotherapy in several cancer types ^25–28. In this context, chemokine (C-X-C motif) ligand 1 (CXCL1), a neutrophil chemoattractant, has specifically been linked to chemoresistance in the metastatic setting ²⁹. Identifying and understanding the potential roles of neutrophils and NETs, and more generally inflammation, in chemoresistance, could lead to the identification of novel ways to improve both chemotherapy efficacy and survival for patients with cancer.

Results

Neutrophil recruitment triggered by chemotherapy reduces treatment response of lung metastases

To determine how chemotherapy modifies the metastatic lung environment, syngeneic Balb/c mice were injected intravenously with 410.4 murine breast cancer cells and metastases were allowed to form for 7 days. Then, mice were treated with cisplatin or Adriamycin/Cyclophosphamide (AC) chemotherapies (Figure 1A). While chemotherapy caused no significant changes in the lungs of non-tumor bearing mice, we observed increased neutrophil recruitment in metastatic lungs as determined by flow cytometry (Figure 1B) and confirmed by immunofluorescence (Figures 1C and S1A). Interestingly, these changes in the lungs were not accompanied by differences in the number and the percentage of neutrophils, detected by flow cytometry, in the blood and in the bone marrow (Table S1). The marked increase in neutrophil infiltration in the lungs, led us to hypothesize that neutrophil recruitment might modulate therapy response. To test this hypothesis, we depleted neutrophils using Ly6G antibodies (Figures 1D, S1B and S1C) ³⁰ which dramatically improved both cisplatin and AC treatment responses (Figures 1E, 1F and 1G). Neutrophil chemotaxis is mediated by the expression of C-X-C motif Chemokine Receptor 2 (CXCR2) ¹⁵. We found that in vitro cisplatin treatment upregulated two CXCR2 ligands, CXCL1 and CXCL5, in cancer cells (Figure S1D). Furthermore, conditioned media (CM) from cisplatin-treated 410.4 cancer cells (410.4 Cis CM) promoted neutrophil chemotaxis in vitro and this was abrogated by a CXCR2 inhibitor, and by CXCL1 and CXCL5 blocking antibodies (Figures S1E–G). In mice, the pharmacological inhibition of CXCR2 reduced chemotherapy-induced neutrophil recruitment in the lungs and improved chemotherapy treatment response (Figures 1E, 1H, 1I, S1H and S1I). Of note, neither neutrophil depletion nor CXCR2 inhibition given alone were able to significantly reduce metastases (Figures 1D–I).

Figure 1. Neutrophils recruited during chemotherapy counteract treatment efficacy.

(A) Schematic showing experimental design. (B) Flow cytometry analysis showing the immune composition of lungs from mice injected or not with 410.4 cancer cells and treated as indicated in (A) (n=3 mice per group; mean±SD). (C) Representative immunostaining of neutrophils (red, myeloperoxidase [MPO]) and DNA (blue, DAPI) in the lungs of mice treated as indicated in (A). The tumor-stroma interface is outlined with a white dashed line. Scale bar: 50 μm. (D) Schematic showing experimental design. (E) Representative images of Hematoxylin and Eosin (H&E) staining of lungs from animals injected intravenously with 410.4 cells and treated as indicated in 1D. Scale bar: 1 mm. (F-I) Mice injected intravenously with 410.4 cells were treated as indicated, and the size (F, H) and the number of metastatic foci (G, I) in the lungs were quantified from H&E staining (in (F, G) n=4 mice per group; in (H, I) n=12 mice for the Veh. group, n=8 mice for the Cis and AC groups, n=5 mice for the CXCR2 inh. group and n=4 mice for the other groups; mean±SD).

Two-way ANOVA followed by Dunnett’s procedure where the Veh. treated non-tumor bearing mice group is the reference was used for (B), and two-way ANOVA followed by Tukey’s procedure was used for (F, G, H, I). See also Figure S1 and Table S1.

NETs promote chemoresistance of breast cancer lung metastases

To determine how neutrophils modulated chemotherapy response, we designed a co-culture system where 410.4 cancer cells were cultured with freshly isolated bone-marrow derived neutrophils (Figure 2A). In vitro co-culturing the breast cancer cells with primary neutrophils reduced the response to chemotherapy (Figures 2A and 2B). Interestingly, we noticed that neutrophils formed NETs when co-cultured with the 410.4 cells, but only in the co-cultures treated with chemotherapy (Figure 2A). However, neutrophils cultured alone did not form NETs when treated with chemotherapy (Figures S1J and S1K). Due to prior research showing that NETs promote cancer progression ³¹, we hypothesized that NETs modulated the therapy response. Indeed, treatment with a PAD4 inhibitor to block NET formation or DNase I to digest the NET-DNA scaffold, overcame neutrophil-mediated chemoresistance but had no effect on cancer cell numbers in the untreated conditions (Figure 2B). We have previously reported that some cancer cells (e.g., D2.A1 and 4T1) induce neutrophils to form NETs while others do not (e.g., D2.0R and 4T07) ^18,32, and decided to test whether these differences could modulate the effect of NETs on chemotherapy efficacy. Interestingly, PAD4 inhibition ameliorated chemotherapy efficacy of all these cell lines in vitro (Table S2). Of note, chemotherapy had no effect on D2.0R cells when they were cultured on Matrigel to mimic cellular dormancy ¹⁸ (Table S2). Further supporting a direct role of the NETs on chemoresistance, the addition of NET-containing conditioned media (CM, induced with Phorbol Myristate Acetate [PMA], NET CM) also caused breast cancer cell resistance to chemotherapy (Figure 2C). In contrast, CM from unstimulated neutrophils (PMN CM) and CM from degranulated neutrophils (obtained using complement 5a; PMN C5a CM) had no effect on response to chemotherapy (Figure 2C). As observed for the co-culture experiments, targeting NETs during PMA-induced NET formation also overcame the chemoresistance induced by the NET CM in vitro (Figure 2C).

Figure 2. NETs formed during chemotherapy counteract treatment efficacy.

(A) mCherry-expressing 410.4 cells were cultured either alone or with neutrophils and treated as indicated for 48 hours. Images show immunostaining of mCherry (red) and DAPI (green). Cancer cells are displayed in red with a green nuclei and neutrophils/NETs in green. Scale bar: 50 μm. (B) Bioluminescence imaging (BLI) signal from luciferase-expressing 410.4 cells cultured as indicated at day 2 (n=3; mean±SD). (C) BLI signal from luciferase-expressing 410.4 cells at day 2 after indicated treatments. PAD4 inh. and DNase I were used during the neutrophil culture to block or digest NET formation, respectively (n=3; mean±SD). (D) Representative immunostaining of neutrophils (red, MPO), NETs (green, citrullinated histone H3 [cit-H3]), and DNA (blue, DAPI) in the lungs of mice treated as indicated. Scale bar: 50 μm. (E) ELISA quantification of NETs in the plasma of mice treated as indicated (n=3 mice per group; mean±SD). (F, I, J) Representative images of H&E staining of lungs from animals injected intravenously with 410.4 (F), PyMT (I) or TS/A (J) cells and treated as indicated. Scale bar: 1 mm. (G, H, K, L) Mice injected intravenously with 410.4 (G, H), PyMT (K) or TS/A (L) cells were treated for 21 days as indicated, and the lung metastatic burden was quantified from H&E staining (in (G) n=13 mice for the Veh. group, n=11 mice for the Cis group, n=5 for the AC group, n=4 mice for the PAD4 inh. group and n=6 mice for the other groups; in (H) n= 10 mice for the Veh. group, n=5 mice for the Cis, AC and DNase I + Cis groups, n=4 mice for the DNase I group and n=6 mice for the other groups; mean±SD; in (K) n=4 mice for the Veh. group, n=5 mice for the AC group and n=6 mice for the other groups; in (L) n=3 mice for the Veh., PAD4 inh. and AC + PAD4 inh. groups and n=4 mice for the other groups). (M-O) ELISA quantification of NETs in the plasma of patients treated as indicated. See also Table S5.

Two-way ANOVA followed by Tukey’s procedure was used for (B, C, G, H, K, L), one-way ANOVA followed by Tukey’s procedure was used for (E) and a two-sided t-test was used for (M-O). See also Figures S1, S2 and S3, Tables S2, S3, S4, S5 and S6.

In vivo, chemotherapy treatment also led to NET formation in the lungs (as detected by co-localized staining for citrullinated histone H3 and myeloperoxidase [MPO]) and the presence of NETs in the blood (as detected by ELISA from plasma samples) at day 21, four days after the last treatment (Figure 2D and 2E). We next collected plasma and lungs at different time points after chemotherapy treatment: in the lungs, increased neutrophil infiltration was first evident two days after chemotherapy, while NETs were detected in both lungs and plasma after three days, and present until five days after treatment (Figures S1L–Q). These results support that chemotherapy did not directly act on neutrophils to cause NETs. To determine whether NETs contributed to chemoresistance in vivo, we targeted NETs in our mouse model of lung metastasis (Figures 1D). Targeting NETs with either a PAD4 inhibitor or DNase I enhanced chemotherapy efficacy (Figures 2F–H, S1R and S1S). Long term PAD4 inhibition and DNase I treatment not only eliminated NETs in the metastatic lungs and in the plasma, but also reduced neutrophil recruitment to the lungs (Figures 2D, S1T–X). However, while short term PAD4 inhibition only led to a non-significant decrease of neutrophil recruitment, long term PAD4 inhibition decreased CXCL1 and CXCL5 levels in the metastatic lungs (Figures S2A–F). These results are consistent with previous reports suggesting that the presence of NETs promotes further neutrophil recruitment ^18,21,33. Inhibiting NETs in the absence of chemotherapy had no effect on the metastatic burden of mice bearing 410.4 cancer cells (Figures 2F–H, S1R and S1S). This is consistent with the inability of 410.4 cells to directly induce pro-metastatic NETs, while inhibiting NETs reduces metastases of cancer cells that induce NETs (e.g., the murine breast cancer cell line 4T1) ^32,34. We confirmed the specificity of NETs in chemoresistance by using Nexinhib20, a degranulation inhibitor. Nexinhib20 did not block chemotherapy-induced NET formation and accordingly did not ameliorate chemotherapy efficacy in vivo (Figures S2G–K). We next confirmed our results using two additional metastatic breast cancer cell lines, the TS/A and PyMT cell lines. In vitro, PAD4 inhibition counteracted neutrophil-mediated chemoresistance of TS/A and PyMT cancer cells (Figures S2L and S2M). In vivo, chemotherapy treatment of mice injected intravenously with TS/A and PyMT cells also led to NET formation in the lungs and targeting of NETs with either a PAD4 inhibitor or DNase I enhanced chemotherapy efficacy (Figures 2I–L and S2N–U).

The inhibitory effects of neutrophils and NETs on chemotherapy response at the lung metastatic site, prompted us to assess the effects of NETs on chemotherapy response of primary tumors (Figure S3A). Only a few neutrophils were present in the primary tumors formed by 410.4 cells (Figure S3B). We detected no NETs regardless of whether the mice were treated with chemotherapy (Figures S3B and S3C) and targeting NETs with a PAD4 inhibitor did not affect chemotherapy response at the primary site (Figures S3D). However, PAD4 inhibition ameliorated chemotherapy efficacy against spontaneous lung metastasis of mice bearing a primary tumor (Figures S3E–J). Interestingly, chemotherapy was more effective against primary tumors when compared to its efficacy against lung metastasis (Figures S3K). To confirm the effects of NETs on lung metastasis specifically, we next injected 410.4 cells both in the mammary fat pad and in the tail vein of the same mice. Using this model, we confirm our results that chemotherapy treatment induced neutrophil recruitment and NET formation specifically in the lungs and that targeting NETs improved greatly chemotherapy response of lung metastasis (Table S3). In this model, chemotherapy was also more efficient at the primary site than in metastatic lungs (Table S3). Thus, the effects of NETs on chemotherapy response are specific to lung metastatic disease in mice. Interestingly, CXCL1 and 5 were increased in the metastatic lungs but not in the primary tumor after chemotherapy treatment, suggesting that the absence of these factors might be responsible for the absence of neutrophil recruitment within the primary tumor following chemotherapy, which could explain chemotherapy selectivity between metastasis and primary tumors. (Table S4).

To evaluate the translational relevance of our findings, we analyzed plasma NETs levels from 16 patients with metastatic breast cancer before and two weeks after the first cycle of their second line chemotherapy. Our results showed a significative increase of plasma NETs level after chemotherapy in patients with chemoresistant disease (i.e. having progressive disease as best response), while the level of plasma NETs remained low in patients experiencing a partial tumor response (Figure 2M–O, Table S5). However, we observed two subgroups within the group of patients with stable disease (Figure 2M–O, Table S5). Additionally, we analyzed 18 primary tumors samples from patients with breast cancer before and after chemotherapy and NETs were not detected and only a few neutrophils were detected before and after chemotherapy (Figures S3L, S3M and Table S6). Altogether these results support the data obtained in mice indicating that, in patients, the presence of NETs correlates with chemoresistance of metastasis.

A known, severe complication of cisplatin administration is Acute Kidney Injury (AKI) ³⁵, which limits the use of this drug in the clinics and can require patients to stop treatment. Evidence of AKI was observed in our mouse model with an increased concentration of plasma-creatinine and Blood Urea Nitrogen (BUN) following cisplatin treatment but not AC treatment (Figures S3N and S3O). NETs were detected in the kidney after cisplatin treatment and inhibiting NETs reduced kidney damage (Figures S3P–V and Table 1). Suggesting that cisplatin toxicity via NET formation is independent of tumor burden, we noticed that whereas chemotherapy did not trigger neutrophil recruitment and NET formation in the lungs of non-tumor bearing mice (Figures S3W and S3X), cisplatin treatment increased plasma NETs even in non-tumor bearing mice (Figures S3Y and Table 1). Accordingly, cisplatin also induced NET formation and kidney injury in non-tumor bearing mice and mice bearing a primary tumor and PAD4 inhibition reduced cisplatin-induced kidney damage in non-tumor bearing mice (Table 1). We did not detect NETs in the liver and in the spleen of mice following chemotherapy treatment, suggesting a specificity toward kidney injury (Table 1). Altogether, we found that inhibition of NETs dramatically increased the response of metastases to chemotherapy, and overcame kidney injury complications of the therapy, complications that can require cessation of treatment.

Table 1:

NETs participate in cisplatin-induced kidney injury

		Kidney neutrophils		Kidney damage		Plasma NETs	Liver neutrophils		Spleen neutrophils
	Conditions	Neutrophils	NETs	BUN	Creatinine		Neutrophils	NETs	Neutrophils	NETs
Non-tumor bearing mice	Veh.	-	-	++	++	-	-	-	-	-
	Cis	++	++	++	++	+	-	-	-	-
	AC	N/A	N/A	N/A	N/A	N/A	-	-	-	-
	Cis + PAD4 inh.	+	-	-	-	-	N/A	N/A	N/A	N/A
	Cis + NLRP3 inh.	+	-	-	-	-	N/A	N/A	N/A	N/A
	Cis + IL1β Ab	+	-	-	-	-	N/A	N/A	N/A	N/A
	Cis + TGFβR1 inh.	++	++	++	++	+	N/A	N/A	N/A	N/A
	Cis + MMP9 inh.	++	++	++	++	+	N/A	N/A	N/A	N/A
Mice with a primary tumor	Veh.	-	-	-	-	-	N/A	N/A	N/A	N/A
	Cis	++	++	++	++	+	N/A	N/A	N/A	N/A
	AC	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A
Mice with lung metastases	Veh.	-	-	-	-	-	-	-	-	-
	AC	-	-	-	-	+	-	-	+	-
	Cis	++	++	++	++	++	-	-	+	-
	Cis + PAD4 inh.	+	-	-	-	-	N/A	N/A	N/A	N/A
	Cis + NLRP3 inh.	++	-	-	-	-	N/A	N/A	N/A	N/A
	Cis + IL1β ab.	+	-	-	-	-	N/A	N/A	N/A	N/A
	Cis + TGFβR1 inh.	++	++	++	++	+	N/A	N/A	N/A	N/A
	Cis + TGFβ ab.	N/A	N/A	++	++	+	N/A	N/A	N/A	N/A
	Cis + MMP9 inh.	++	++	++	++	+	N/A	N/A	N/A	N/A
	Cis + ITGαvβ1	N/A	N/A	++	++	+	N/A	N/A	N/A	N/A

Non-tumor bearing mice, mice injected intravenously with 410.4 cancer cells and mice injected with 410.4 cancer cells in the mammary fat pad were treated for two weeks as indicated. Kidney, Liver and Spleen Neutrophils and NETs were quantified based on immunostaining for myeloperoxidase and Citrullinated-H3 respectively. For neutrophils: “-” indicates the basal level of neutrophils; “+” indicates a 2 to 3-fold increase; “++” indicates more than a 3-fold increase. For NETs: “-” indicates the absence of NETs, and “++” indicates the presence of NETs. For plasma NETs: “-” indicates the basal level of neutrophils; “+” indicates a 2 to 4-fold increase; “++” indicates more than a 4-fold increase. For BUN: “-” indicates a concentration lower than 25mg/dL and “++” indicates a concentration above 65 mg/dL. For creatinine: “-” indicates a concentration lower than 0.25 mg/dL and “++” indicates a concentration above 0.75 mg/dL. (n=3 mice per group; mean±SD). See also Figure S3, S4, S5, S7 and S8.

Chemotherapy triggers NLRP3-mediated IL-1β secretion in cancer cells, leading to NET formation and chemoresistance

We next sought to determine the mechanism driving NET formation after chemotherapy. Neither cisplatin nor AC chemotherapy directly induced primary neutrophils to form NETs (Figures S1J and S1K). However, CM from chemotherapy-treated cancer cells (410.4 Cis CM and 410.4 AC CM) promoted NET formation (Figures 3A and S4A), suggesting the presence of a NET-inducing, secreted factor released by cancer cells after chemotherapy. To identify candidate factors, we used a proteome cytokine array on the CM from chemotherapy-treated cancer cells (Figure 3B). As before (Figure S1D), our analysis showed upregulation of the neutrophil chemoattractants CXCL1 and CXCL5 but also revealed upregulation of interleukin-1β (IL-1β), specifically in the CM of chemotherapy-treated cancer cells (Figure 3B). Importantly, neutralization of CXCL1 and CXCL5 and inhibition of CXCR2 did not block NET formation induced by 410.4 Cis CM and 410.4 AC CM (Table S7). Consistent with prior reports ^36–38, recombinant IL-1β triggered NET formation in vitro (Figures S4B and S4C). Furthermore, secretion of IL-1β by chemotherapy-treated cancer cells (410.4, TS/A, PyMT, D2.0R and 4T07 cancer cells) was confirmed by enzyme-linked immunoabsorbent assay (ELISA) and neutralization of IL-1β with an IL-1β blocking antibody inhibited NET formation induced by chemotherapy-treated cancer cells CM (Figures 3C, 3D, S4D–J and Table S2). Additionally, the IL-1β blocking antibody improved chemotherapy response of all these cell lines in the co-culture assay in vitro (Figure 3E and Table S2). As reported before, the CM from untreated D2.A1 and 4T1 cells was sufficient to induce NET formation, but the ability of these cells to induce NET formation was not dependent on IL-1β (Table S2). While both 4T1 and D2.A1 cells secreted IL-1β in response to chemotherapy, neutralization of IL-1β did not counteract NETs formation and therefore did not counteract chemoresistance in vitro (Table S2). This is in accordance with the fact that 4T1 cells secrete granulocyte colony stimulating factor (G-CSF) which induces NETs at the steady state ³². Interestingly, the IL-1β blocking antibody did not increase the chemotherapy response of 410.4 cancer cells when using NET CM, where the NETs had already been formed (Figure 3F). These data support a role for IL-1β in NET formation but not in the downstream effects causing chemotherapy resistance. Also, in the context of lung metastasis, the IL-1β blocking antibody inhibited NET formation, reduced neutrophil recruitment (Figures 3G–J) and improved chemotherapy response (Figures 3K–M). As observed with the PAD4 inhibitor (Figures S2A–D), short term inhibition of IL-1β only led to a non-significant decrease of neutrophil recruitment in the lungs (Figures S4K–M). However, the effect of IL-1β blockade on neutrophil recruitment was not direct as IL-1β blockade did not block neutrophil chemotaxis toward chemotherapy-treated cancer cells CM in vitro (Figure S4N). Interestingly, long term inhibition of PAD4 decreased the levels of IL-1β which were induced by chemotherapy within the metastatic lungs (Figure S4O). In addition, IL-1β blockade counteracted cisplatin-induced NET formation in the kidney of non-tumor bearing mice and mice with lung metastasis and improved kidney function following cisplatin treatment (Table 1, Figures S4P and S4Q).

Figure 3. Chemotherapy triggers the secretion of IL-1β by cancer cells, which drives NET formation.

(A) Immunostaining of mouse neutrophils cultured as indicated. DAPI (blue), anti-MPO (red), and anti-histone H2B (green) staining were used to assess NET formation. Scale bar: 50 μm. (B) Protein content was investigated in the CM from vehicle- and chemotherapy-treated 410.4 cancer cells using protein arrays. (C) Immunostaining of mouse neutrophils cultured as indicated. DAPI (blue), anti-MPO (red), and anti-histone H2B (green) staining were used to assess NET formation. Scale bar: 50 μm. (D) ELISA quantification of IL-1β in the CM from 410.4 cells treated as indicated (n=3; mean±SD). (E) BLI from luciferase-expressing 410.4 cells cultured as indicated at day 2 (n=3; mean±SD). (F) BLI signal from luciferase-expressing 410.4 cells at day 2 after indicated treatments (n=3; mean±SD). (G) Representative immunostaining of neutrophils (red, MPO), NETs (green, cit-H3), and DNA (blue, DAPI) in the lungs of mice treated as indicated. Scale bar: 50 μm. (H, I) Quantification of images represented in (G) from the lungs of mice treated as indicated and stained for MPO (neutrophils, H) and cit-H3 (NETs, I) (n=3 mice per group; mean±SD). (J) ELISA quantification of NETs in the plasma of mice treated as indicated (n=3 mice per group; mean±SD). (K) Representative images of H&E staining of lungs from animals injected intravenously with 410.4 cells and treated as indicated. Scale bar: 1 mm. (L, M) Mice injected intravenously with 410.4 cells were treated for 21 days as indicated, and the size (L) and the number of metastatic foci (M) in the lungs were quantified from H&E staining (n=6 mice for the Cis and AC groups and n=4 mice for the other groups; mean±SD).

One-way ANOVA followed by Dunnett’s procedure where the Veh. group is the reference was used for (D), two-way ANOVA followed by Tukey’s procedure was used for (E, F, H, I, J, L, M). See also Figure S4 and Table S7.

It was unclear why chemotherapy-treated cancer cells would secrete IL-1β. IL-1β is generally secreted after activation of inflammasomes, large protein complexes assembled when cells sense danger, e.g., components such as those from dead cells ³⁹. The assembly of the inflammasome complex activates caspase-1, which then cleaves pro-IL-1β to generate mature IL-1β ⁴⁰. We determined that NOD-like receptor family pyrin domain containing 3 (NLRP3) inflammasome and caspase-1 inhibition inhibited IL-1β secretion by chemotherapy-treated cancer cells (Figure 4A). NLRP3 inflammasome activation requires two signals: (i) a priming signal, which leads to Nuclear Factor-κ-light-chain-enhancer of activated B cells (NF-κB)-mediated inflammasome and pro-IL-1β transcription; and (ii) a danger signal, which leads to inflammasome assembly and caspase-1 activation. Constitutive activation of NF-κB signaling pathway in cancer cells has been well described in the past ⁴¹ and was identified in 410.4 cancer cells with the constitutive phosphorylation of IκB in the absence of any external signal (Figure S4R). Adenosine triphosphate (ATP) binding to its purigenic receptor P2X 7 (P2RX7) is a well-known NLRP3 danger signal and cancer cell death induced by chemotherapy causes the release of ATP into the extracellular space ⁴². We confirmed that chemotherapy treated 410.4 cancer cells released ATP (Figure 4B). In addition, we determined that both NF-κB and P2RX7 inhibition blocked chemotherapy-induced IL-1β secretion but had no effect on ATP release (Figures 4C and S4S). We therefore hypothesized that ATP released by dying chemotherapy-treated tumor cells triggers NLRP3 inflammasome activation in neighboring non-dying cancer cells. To test this hypothesis, we first generated the CM from cisplatin-treated cancer cells in which we identified the presence of ATP (Figure 4B). We next used this CM to treat naive cancer cells and used immunofluorescence for Apoptosis-associated Speck-like protein containing a CARD (ASC) assembly as a readout of inflammasome activation (Figures 4D and S4T) ⁴³. CM from cisplatin-treated cancer cells triggered ASC assembly, and it was dependent on NF-κB, P2RX7, and NLRP3 activity. As expected, caspase 1 inhibition did not block ASC assembly as caspase-1 only mediates the cleavage of pro-IL-1β into IL-1β following ASC assembly (Figures 4D and S4T). Similarly to the blocking of IL-1β secretion, pharmacological inhibition of NF-κB, NRLP3, P2RX7 and caspase-1 counteracted the ability of cisplatin-treated cancer cells to promote NET formation (Figures 4E and S4U). In vitro, NLRP3 inhibition improved chemotherapy efficacy in the co-culture system (Figure 4F) but had no effect when NETs were already formed (Figure S4V). Additionally, NLRP3 inhibition blocked chemotherapy-induced NET formation in vivo and significantly improved both treatment efficacy and kidney function (Figures 4G–L, S4W–Y, and Table 1). Altogether, our results suggest that as chemotherapy induces cancer cell death, the dying cancer cells respond by releasing CXCL1 and CXCL5, which promote neutrophil recruitment, and ATP, which triggers NLRP3 inflammasome activation and IL-1β secretion from neighboring cancer cells. IL-1β then promotes NET formation and these NETs inhibit chemotherapy efficacy.

Figure 4. Chemotherapy triggers NLRP3 activation in cancer cells which drives IL-1β secretion.

(A) ELISA quantification of IL-1β in the CM from 410.4 cells treated as indicated (n=3; mean±SD). (B) ATP quantification in the CM from 410.4 cells treated as indicated (n=3; mean±SD). (C) ELISA quantification of IL-1β in the CM from 410.4 cells treated as indicated (n=3; mean±SD). (D) Immunostaining of 410.4 cancer cells cultured as indicated on the left scheme and the different panels. DAPI (blue) and anti-ASC (green) staining were used to assess inflammasome oligomerization. Scale bar: 50 μm. Schematic created with BioRender.com. (E) Immunostaining of mouse neutrophils cultured as indicated on the left scheme and different panels. DAPI (blue), anti-MPO (red), and anti-histone H2B (green) staining were used to assess NET formation. Scale bar: 50 μm. Schematic created with BioRender.com. (F) BLI from luciferase-expressing 410.4 cells as indicated at day 2 (n=3; mean±SD). (G) Representative immunostaining of neutrophils (red, MPO), NETs (green, cit-H3), and DNA (blue, DAPI) in the lungs of mice treated as indicated. Scale bar: 50 μm. (H, I) Quantification of images represented in (G) from the lungs of mice treated as indicated and stained for MPO (neutrophils, H) and cit-H3 (NETs, I) (n=3 mice per group; mean±SD). (J) ELISA quantification of NETs in the plasma of mice treated as indicated. (n=3 mice per group; mean±SD). (K) Representative images of H&E staining of lungs from animals injected intravenously with 410.4 cells treated as indicated. Scale bar: 1 mm. (L) Mice injected intravenously with 410.4 cells were treated as indicated, and the lung metastatic burden was quantified from H&E staining (n=13 mice for the veh. group, n=11 mice for the Cis group, n=5 mice for the AC and NLRP3 inh. group, n=3 mice for the NLRP3 inh. + Cis group and n=6 mice for the NRLP3 inh. + AC group; mean±SD). The same Veh., Cis and AC groups were used for mice with 410.4 cells and treated with the NLPR3 inh. and the PAD4 inh.

Two-way ANOVA followed by Tukey’s procedure was used for (A, C, F, H, I, J, L), and one-way ANOVA followed by Dunnett’s procedure where the Veh. group is the reference was used for (B). See also Figure S4.

NETs activate TGFβ signaling pathway in cancer cells, causing EMT and chemoresistance

To identify the signaling pathways mediating chemoresistance following NET formation, we performed RNA-sequencing of 410.4 cells stimulated with NET CM. This analysis identified Transforming Growth Factor β (TGFβ) signaling pathway and EMT related signaling pathways as two of the most significant pathways to be increased (Figure 5A). EMT causes chemoresistance in several cancer types, including breast cancer, and Transforming Growth Factor β (TGFβ) is a major inducer of EMT ⁴⁴. qRT-PCR and Western blot analysis revealed that NETs and TGFβ induced a similar upregulation of most mesenchymal markers analyzed in the cancer cells, but NETs and TGFβ induced the downregulation of only Claudin 1 among the epithelial markers tested (Figures 5B, S5A and S5B). Accordingly, NETs induced SMAD family member 2 (SMAD2) phosphorylation in cancer cells, indicating TGFβ signaling activation, and TGFβ Receptor 1 (TGFβR1) inhibition inhibited NET-mediated SMAD2 phosphorylation and EMT genes expression (Figure 5B). In addition, targeting TGFβ signaling improved chemotherapy response of cancer cells in vitro (Figures 5C and 5D) and in vivo (Figures 5E–K, S5C–F). Within the lung tissue, chemotherapy increased the expression of the mesenchymal marker N-cadherin and decreased the expression of the epithelial marker Claudin 1 in cancer cells, only in the presence of NETs (Figures 5L–O). Indeed, this switch was abrogated when targeting NETs with a PAD4 inhibitor and DNase I (Figures 5L–O). The same effect was observed after TGFβR1 inhibition, consistent with the role of TGFβ in NET-induced EMT (Figures 5L–O). Interestingly, while recombinant TGFβ had no effect on NET formation in vitro, targeting TGFβ led to decreased neutrophil recruitment and NET formation in vivo (Figures S5G–P). Interestingly, TGFβR1 inhibition did not block chemotaxis of neutrophils toward 410.4 AC CM in vitro (Figure S5Q) and short-term inhibition of TGFβR1 in vivo did not block the effect of chemotherapy on neutrophil recruitment and NET formation (Figures S5R–T). This suggests that targeting TGFβ signaling indirectly limits lung inflammation which further improves chemotherapy treatment. This is in accordance with the pleiotropic role of TGFβ signaling in immunity and cancer ⁴⁵. Accordingly, TGFβR1 inhibition decreased the levels of CXCL1, CXCL5 and IL-1β in the metastatic lungs following two weeks of chemotherapy (Figures S5U–W). Interestingly, TGFβ-targeting strategies did not block cisplatin-induced NET formation in the kidney and did not reduce cisplatin-induced kidney damage, suggesting that the effect of NETs on kidney injury are not TGFβ dependent (Figures S5X–AA and Table 1). Altogether, these results indicate that NETs induce a TGFβ-dependent EMT in cancer cells which correlates with chemoresistance.

Figure 5. NETs promote chemoresistance through a TGFβ-dependent EMT in cancer cells.

(A) Gene ontology analysis of the genes that were uniquely up-regulated in 410.4 cells treated with NET CM for 48 hours. (B) Western blot for protein listed in 410.4 cells after 48 hours of indicated treatments. Results are representative of 2 independent experiments. (C) BLI signal from luciferase-expressing 410.4 cells cultured as indicated at day 2 (n=3; mean±SD). (D) BLI signal from luciferase-expressing 410.4 cells at day 2 after indicated treatments (n=3; mean±SD). (E, H, J) Representative images of H&E staining of lungs from animals injected intravenously with 410.4 (E), PyMT (H) or TS/A (J) cells treated as indicated. Scale bars: 1 mm. (F, G, I, K) Mice injected intravenously with 410.4 (F, G), PyMT (I) or TS/A (K) cells were treated for 21 days as indicated, and the lung metastatic burden was quantified from H&E staining (in (F) n=12 mice for the Veh. group, n=8 mice for the Cis and AC groups, n=5 mice for the TGFβR1 inh. group, n=4 mice for the other groups (the same Veh., Cis and AC groups were used for mice with 410.4 cells and treated with the CXCR2 inh. and the TGFβR1 inh.); in (G) n=6 mice for the Cis and AC groups and n=4 mice for the other groups; in (I) n= 4 mice for the Veh. and TGFβR1 inh. groups, n=5 mice for the AC group and n=6 mice for the AC + TGFβR1 inh. group (the same Veh. and AC groups were used for mice with PyMT cells and treated with the PAD4 inh., DNase I and the TGFβR1 inh.); in (K) n=4 mice for the AC group and n=3 mice for the other groups (the same Veh. and AC groups were used for mice with TS/A cells and treated with the PAD4 inh., DNase I and the TGFβR1 inh.); mean±SD). (L, M) Representative immunostaining of Claudin 1 (red, L), N-cadherin (green, M), and DNA (blue, DAPI) in the lung metastasis of mice injected intravenously with 410.4 cells and treated as indicated. Scale bar: 50 μm. (N, O) Quantification of images represented in (L) and (M) (n=3 mice per group; mean±SD).

Two-way ANOVA followed by Tukey’s procedure was used for (C, D, F, G, I, K, N, O). See also Figure S5.

NET-associated MMP9 activates latent TGFβ which counteracts chemotherapy efficacy

Our results suggested that TGFβ plays a role in NET-mediated chemoresistance. TGFβ is synthesized and secreted as a latent complex which is then processed to its active form ⁴⁶. Using an ELISA test to detect both total- and active- TGFβ, we found a large amount of latent-TGFβ in the CM from 410.4 cancer cells, but not in the CM from neutrophils including after NET formation (Figure 6A). Active-TGFβ was barely detectable in all CM (Figure 6A). Chemotherapy did not change the levels of active and total TGFβ present in the CM from cancer cells (Figure 6A). However, when we incubated the CM from 410.4 cells with the CM from neutrophils, NETs activated latent TGFβ released by 410.4 cells (Figure 6B). NETs did not modulate transcription of TGFβ family members or receptors in 410.4 cells (Figure S6A) but targeting of NETs abrogated the activation of latent TGFβ (Figures 6B).

Figure 6. NET-associated MMP9 activates latent TGFβ which counteracts chemotherapy efficacy.

(A) ELISA quantification of total and active TGFβ secreted in indicated CM (n=3; mean±SD). (B) ELISA quantification of total and active TGFβ in the CM from 410.4 cells incubated as indicated for 24 hours. PAD4 inh. and DNase I were used during the neutrophil culture to block or digest NET formation, respectively (n=3; mean±SD). (C) Western blot for TGFβ and LAP in the CM of 410.4 cells after 24 hours of indicated treatments. Results are representative of 2 independent experiments. (D) Immunostaining of mouse neutrophils cultured with IL-1β. DAPI (blue) and MMP9 (red) localized on the DNA scaffold of NETs. Scale bar: 50 μm. (E) ELISA quantification of total and active TGFβ in the CM from 410.4 cells incubated as indicated for 24 hours (n=3; mean±SD). (F) Western blot for TGFβ and LAP in the CM of 410.4 cells after 24 hours of indicated treatments. Results are representative of 2 independent experiments. (G) BLI signal from luciferase-expressing 410.4 cells cultured as indicated at day 2 (n=3; mean±SD). (H) Representative images of H&E staining of lungs from animals treated as indicated. Scale bar: 1 mm. (I) Mice injected intravenously with 410.4 cells were treated for 21 days as indicated, and the lung metastatic burden was quantified from H&E staining (n=12 mice for the Veh. group, n=8 mice for the Cis and AC groups, n=5 mice for the MMP9 inh. group and n=4 mice for the other groups; mean±SD). The same Veh., Cis and AC groups were used for mice with 410.4 cells and treated with the CXCR2 inh., the TGFβR1 inh. and the MMP9 inh. (J, L) Representative immunostaining of Claudin 1 (red, J), N-cadherin (green, L), and DNA (blue, DAPI) in the lungs of mice injected intravenously with 410.4 cells and treated as indicated. Scale bar: 50 μm. (K, M) Quantification of images represented in (J) and (L) from the lungs of mice injected intravenously with 410.4 cells, treated as indicated and stained for Claudin 1 (K) and N-cadherin (M) (n=3; mean±SD).

Two-way ANOVA followed by Tukey’s procedure was used for (A, B, E, G, I, K, M). See also Figures S6 and S7.

Latent TGFβ is secreted either as a Small Latent Complex (SLC), associated with Latency Associated Protein (LAP) or as a Large Latent Complex (LLC), associated with LAP and Latent TGFβ Binding Protein (LBTP) ⁴⁷. LBTP was not detected in cell lysate, ECM, or CM from 410.4 cancer cells, but the LAP-TGFβ complex was detectable in all three isolates of the cancer cells (Figures S6B and S6C). In agreement with the ELISA results (Figure 6B), Western Blot revealed that incubation of cancer cell CM with NETs led to the degradation of LAP and the release of active TGFβ and targeting NETs abrogated LAP degradation (Figure 6C). In contrast, CM from unstimulated and degranulated neutrophils had no effect (Figures 6B and 6C). Activation of latent TGFβ into active TGFβ was also observed when culturing cancer cells with NET CM or after chemotherapy treatment when using a co-culture system and targeting NETs abrogated the activation of TGFβ also under these conditions (Figures S6D and S6E).

NETs are characterized by the association of neutrophil proteases with the DNA scaffold ^48,49. MMP9 can proteolytically activate TGFβ and is present within the NET-DNA scaffold (Figure 6D) ^18,50. Using inhibitors against some of the major NET-associated proteases - Neutrophil Elastase (NE), MMP9 and Cathepsin G - and MPO (which indirectly can activate TGFβ activation ⁵¹), we found that only MMP9 inhibition inhibited NET-mediated TGFβ activation (Figures 6E, 6F, S6F and S6G) and chemoresistance in vitro (Figures 6G and S6H). MMP9 inhibition also sensitized cancer cells to cisplatin and AC chemotherapy in vivo (Figures 6H, 6I and S6I). Accordingly, MMP9 inhibition counteracted NET-mediated EMT in vitro (Figure S7A) and in vivo (Figures 6J–M). MMP9 inhibition additionally decreased the number of neutrophils and NETs in the lungs (Figures S7B–E) but did not block the ability of chemotherapy-treated cancer cells to promote NET formation in vitro (Figures S7F and S7G). In addition, as observed with TGFβR1 inhibition, MMP9 inhibition did not block the chemotaxis of neutrophils toward 410.4 AC CM in vitro (Figure S7H), and short-term inhibition of TGFβR1 in vivo did not block the effect of chemotherapy on neutrophil recruitment and NET formation (Figures S7I–K). These results suggest that TGFβ signaling indirectly limits inflammation within the metastatic lungs. As observed with TGFβ-inhibiting strategies, MMP9 inhibition did not reduce cisplatin-induced NET formation within the kidney and did not reduce kidney damage (Figures S7L and S7M and Table 1). Altogether, our results show that NET-associated MMP9 can activate latent TGFβ, secreted by cancer cells, and that this activation induces EMT in cancer cells and correlates with chemoresistance.

NET-associated ITGαvβ1 traps latent TGFβ

NET-mediated proteolytic activation of TGFβ was required for cancer cell resistance to treatment and digesting NETs with DNase I improved treatment efficacy in vivo and in vitro (Figures 2B, 2C, 2H, 2K, 2L and S1S). Yet, we previously showed that DNase I digestion of NETs does not reduce NET-MMP9 activity ¹⁸. Therefore, we tested whether the NET-DNA scaffold would contribute to TGFβ activation through different means. While 410.4 cancer cells expressed the NET-DNA receptor ccdc25, siRNA against ccdc25 ²⁰ did not counteract neutrophil-mediated chemoresistance in vitro (Figures S8A and S8B). We therefore decided to focus on integrins (ITG) which are large transmembrane proteins, but parts of them are also found in NETs ⁵². Moreover, integrins, including ITGαvβ1, have been shown to bind latent TGFβ through the LAP-RGD domain ^46,47. We hypothesized that latent TGFβ was trapped and then processed within the NET-DNA scaffold. We found that ITGαvβ1 and MMP9 were both present within the NET-DNA scaffold by immunofluorescence (Figures 7A and 7B). A potent and highly specific inhibitor of ITGαvβ1 has been developed ⁵³, and this inhibitor, as well as blocking antibodies against ITGαvβ1 and a RGD peptide inhibited TGFβ activation and LAP degradation (Figures 7C, 7D and S8C), indicating that NET-associated ITGαvβ1 is necessary for TGFβ activation by NET-associated MMP9. Next, using NETs incubated with CM from chemotherapy treated cancer cells, we showed that TGFβ, secreted by cancer cells, was trapped within the NET-DNA scaffold and this trapping was prevented when targeting ITGαvβ1 (Figures 7E and 7F). In addition, and as observed with TGFβR1 and MMP9 inhibition, ITGαvβ1 inhibition counteracted NET-induced EMT in vitro (Figure S8D). Accordingly, targeting ITGαvβ1 improved chemotherapy response both in vitro (Figures 7G and S8E) and in vivo (Figures 7H–J) and reduced neutrophil recruitment and NET formation (Figures S8F–I). However, ITGαvβ1 targeting did not block NET formation induced by chemotherapy-treated cancer cell CM in vitro (Figures S8J and S8K). Targeting ITGαvβ1 in vivo did not limit cisplatin-induced kidney injury (Figures S8L, S8M and Table 1) suggesting that the activation of TGFβ by NETs is not responsible for NET-mediated kidney injury. Altogether, our results suggest that NETs act as a scaffold that traps cancer-cell derived latent TGFβ, which can then be cleaved efficiently by NET-associated MMP9 and released as active TGFβ to promote chemoresistance.

Figure 7. NET-associated ITGαvβ1 traps latent TGFβ.

(A) Immunostaining of mouse neutrophils cultured with IL-1β. DAPI (blue), ITGβ1(green) and MMP9 (red). Scale bar: 50 μm. (B) Immunostaining of mouse neutrophils cultured with IL-1β. DAPI (blue), ITGβ1 (green) and ITGαv (red). Scale bar: 50 μm. (C) ELISA quantification of total and active TGFβ in the CM from 410.4 cells incubated as indicated for 24 hours (n=3; mean±SD). (D) Western blot for TGFβ and LAP in the CM of 410.4 cells after 24 hours of indicated treatments. Results are representative of 2 independent experiments. (E, F) NETs were formed from IL-1β-treated murine neutrophils and 410.4 CM were incubated on top of the NETs for 30 min with the indicated treatment before fixation and immunostaining. (E) DAPI (blue), TGFβ (green) and ITGβ1 (red). (F) DAPI (blue) and TGFβ (red). Scale bar: 50 μm. (G) BLI signal from luciferase-expressing 410.4 cells cultured as indicated at day 2 (n=3; mean±SD). (H) Representative images of H&E staining of lungs from animals treated as indicated. Scale bar: 1 mm. (I, J) Mice injected intravenously with 410.4 cells were treated for 21 days as indicated, and the size (I) and the number of metastatic foci (J) in the lungs were quantified from H&E staining. (n=12 mice for the Veh. group, n=8 mice for the Cis and AC groups, n=5 mice for the ITGαvβ1 inh. group, n=3 for the ITGαvβ1 + Cis group and n=4 mice for the ITGαvβ1 + AC group; mean±SD). The same Veh., Cis and AC groups were used for mice with 410.4 cells and treated with the CXCR2 inh., the TGFβR1 inh., the MMP9 inh. and the ITGαvβ1 inh.

Two-way ANOVA followed by Tukey’s procedure was used for (C, G, I, J). See also Figure S8.

Discussion

The major obstacles to the cure of metastatic breast cancer are metastasis and the emergence of resistance to chemotherapy, two phenomena that are closely linked clinically ⁵⁴. In this context, we found that chemotherapy induces NET formation, which promotes chemoresistance of breast cancer lung metastasis in mice. Interestingly, it was recently shown that NETs promote radiation resistance to bladder cancer through unknown mechanisms ⁵⁵ and that radiation injury to the healthy lung tissue enhances neutrophil-dependent metastatic colonization in mice ⁵⁶. While recent studies identified that NETs can promote EMT ^57–61, we here unveil the molecular mechanism of this promotion, via the binding and the activation of latent TGFβ by two NET-associated proteins: 1) ITGαvβ1, which traps latent TGFβ; and 2) MMP9, which cleaves and activates latent TGFβ. While we focused on cancer cells as the main source of latent TGFβ in our study, we cannot exclude that stromal cells within the metastatic TME also release TGFβ that can be trapped and activated by NETs. TGFβ has pleiotropic effects on many cell types in vivo, with both physiological and pathological consequences ⁶²; therefore this function of NETs could have broad implications in development and disease progression.

Recently, NETs have emerged as targets in the treatment of non-infectious diseases such as cancer ¹⁷ highlighting the importance of identifying the mechanisms responsible for NET formation. The chemotherapy did not act directly on the neutrophils, rather our study reveals that the release of ATP from chemotherapy-treated, dying cancer cells drives NLRP3-mediated IL-1β secretion by other cancer cells, and that IL-1β in turn triggers NET formation. This is consistent with prior reports highlighting the importance of IL-1β in NET formation ^36–38. In addition, our study re-affirms the role of NETs as a proteolysis scaffold, as we and others described previously ^18,63. Our data suggest that the NET scaffold, through associated integrins, acts as a specialized local site that allows NET-proteases to cleave their substrate efficiently.

Interestingly, NET inhibition had no effect on the efficacy of chemotherapy against murine primary breast tumor, where no NETs were detected, either before or after chemotherapy. Accordingly, we did not detect any NETs but only a few neutrophils in the primary tumors from breast cancer patients before and after chemotherapy. These results are consistent with previous studies that have shown that neutrophils and NETs are scarce within the primary breast tumors both in mice and humans, and that neutrophils do not influence primary tumor growth in mice ^20,32. It is possible that neutrophils do not mediate pro-tumoral activities at the breast primary tumor site because their recruitment is limited by the primary tumor microenvironment, in which CXCL1/CXCL5 were not upregulated after chemotherapy in mice. While the TME plays a crucial role in cancer progression, its role changes in primary tumors when compared to metastatic sites, especially regarding immune cell population ⁶⁴. Altogether, our results suggest that resistance mechanisms differ between the primary tumor and metastatic sites, indicating that different treatment protocols could be needed to treat the patients. It is increasingly clear that neutrophils are key players in breast cancer metastasis, especially in the lungs ^{18,20,21,32,34} but further studies will be needed to investigate the importance of the NET-ITGαvβ1/MMP9-TGFβ axis in other metastatic organs such as the bones, the liver and the brain, but also in other cancer that metastasize to the lungs such as colorectal, bladder and prostate cancer.

Acute kidney injury (AKI) is an important limitation for cisplatin use, as there is approach treatment to reduce the onset of AKI. Interestingly, we identified NET-forming neutrophils in the kidneys of mice treated with cisplatin and showed that targeting NETs improved kidney function. This is consistent with the recently identified role of NETs in ischemic AKI ⁶⁵. Additionally, our study demonstrated that NLRP3 and IL-1β targeting-strategies improved kidney function, and interestingly, NRLP3 and IL-1β cytokines were reported to play a role in AKI ^66,67. Targeting TGFβ did not improve kidney function, indicating that the mechanisms by which NETs promote chemoresistance and cause kidney damage are different.

Our work identified a significative increase of plasma NETs level 15 days after chemotherapy treatment specifically in patients with progressive disease. This suggest that the detection of NETs in the blood of patients could be used as a biomarker to identify individuals where therapy responses against metastasis are impaired and could be improved by targeting the NETs/αvβ1/MMP9/TGFβ axis identified here.

In conclusion, our study identifies upstream and downstream effectors of the chemoprotective role of NETs in lung metastasis as therapeutic targets to improve chemotherapy efficacy, which could have immediate relevance to thousands of cancer patients.

STAR Methods

Resource Availability

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Dr. Jean Albrengues (jalbrengues@unice.fr).

Materials availability

This study did not generate new unique reagents.

Data and Code availability.

• RNA-seq data generated for this study are available through the Gene Expression Omnibus (GEO: GSE218119).

• This paper does not report original code.

• Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

Experimental model and subject details

Animals

Six to eight-week-old immunocompetent female BALB/c and C57BL/6 mice were purchased from Charles River Laboratories. All mice had a Virus Antibody Free (AVF) immune status (pathogen free). Animals were housed 4 mice per cages in a room with controlled temperature and humidity under a 12h light/12h dark cycle. Mice were fed with food and autoclaved water provided by the animal facility of the Institute for Research on Cancer and the Aging, Nice. All procedures were conducted in accordance with French laws and European recommendations following a protocol approved by the French Ministry of Education and Research and by the local ethical committee for animal experimentation (CIEPAL Azur; Projects APAFIS#23972-2020011415206448 v2, APAFIS#29349-2021012013049209 v4).

Patient samples

Human breast cancer tissue

Formalin fixed and paraffin embedded tumor tissue from cancer patients before and after treatment with neoadjuvant chemotherapy were selected retrospectively by database extraction from the archives of the Department of Pathology at Odense University Hospital. Clinical and pathological characteristics of the patient tumors are presented in Table S6. All tissue samples were collected in compliance with informed consent policy. All clinical samples were coded to maintain patient confidentiality and studies were approved by the Ethics Committee of the Region of Southern Denmark (approval #S-20210115). Paraffin-embedded tissues were used for detection of neutrophils and NETs.

Patients and blood sampling

Patients in this study were included in the prospective single arm phase II clinical trial “MONDRIAN” (clinicaltrials.gov: NCT04720729), treated and followed at Institut Curie Hospitals. The trial was approved by an ethical committee (“CPP Ouest I”) on December 4th, 2020. In addition, the consenting to study protocol itself, patients consented to the re-use of plasma samples for other researches. Main eligibility criteria were as follows: patients aged ≥ 18 years old with HER2-negative metastatic breast cancer (MBC) (triple negative breast cancer (TNBC) or hormone receptors (HR) positive, HER2-negative breast cancer), performance status (PS) of 0–2, eligible to a second line of chemotherapy for MBC and written informed consent. All patients were enrolled at the beginning of the second line of chemotherapy. The second line of treatment was managed by clinical and radiological evaluations (RECIST v1.1). For each patient, blood samples were collected at day 1 (D1) prior to treatment start and day 15 (D15) after the first cycle of chemotherapy. All information regarding the patients is presented in Table S5.

Cell lines and cell culture

410.4 murine cancer cells were obtained from Tobias Pukrop (University Hospital Regensburg, Regensburg, Germany)⁶⁸. D2.OR and D2.A1 murine cancer cells were obtained from Jeffrey Green (National Institutes of Health)⁶⁹. PyMT murine cancer cells were obtained from Bodu Liu (Cold Spring Harbor Laboratory, Cold Spring Harbor, USA). 4T07 murine cells were obtained from Erik Sahai (The Francis Crick Institute, London, UK)⁷⁰. TS/A murine cancer cells were purchased from Merckmillipore (#SCC177). 4T1 murine cancer cells were purchased from ATT (#CRL-2539). 293T cells were purchased from ATCC (#CRL-3216). TS/A, PyMT, 4T07, D2.OR, D2.A1 4T1 and 293T cells were cultured in Dulbecco’s Modified Eagle Medium (DMEM) supplemented with 10% fetal calf serum (FCS) (FCII, Hyclone), 2 mM L-Glutamine, 100 units/mL penicillin, and 100 μg/mL streptomycin. The PyMT cell line was isolated from the primary tumor cells obtained from one MMTV-PyMT mouse on the C57BL/6J background. The isolated cells were cultured in DMEM supplemented with 10% FCS until spontaneous immortalization. Immortalized cells were then sorted based on CD326 expression using Flow cytometry, and the sorted cells were then maintained in complete medium. To generate mCherry- and luciferase-expressing cells, a pGIPz vector containing cDNA for mCherry and luciferase was used. Lentiviral supernatants were collected from a 293T packaging cell line transfected with the vector. Cancer cells were infected with the viral supernatants overnight in the presence of 10 μg/mL polybrene and selected with 1 μg/mL puromycin. To generate 4T07 GFP-expressing cells, a pLJM1 Puro vector containing cDNA for GFP was used. Lentiviral supernatants were collected from a 293T packaging cell line transfected with the vector. Cancer cells were infected with the viral supernatants overnight in the presence of 8 μg/mL polybrene and selected with 5 μg/ml puromycin. Cells were tested repeatedly for mycoplasma over the course of the study and were never positive. Neutrophils and lung cells were isolated freshly as described below.

Method details

Lung metastasis model and treatments

To generate lung metastasis, we intravenously injected 410.4 (500,000), TS/A (80,000) or PyMT (500,000) mouse cancer cells in 100 μL of PBS into eight-week-old female BALB/c mice. After 7 days (for 410.4 and TS/A cells), or 28 days (for PyMT cells) chemotherapy treatments were started twice a week with cisplatin (2.5 mg/kg) or Adriamycin/Cyclophosphamide (AC, 1mg/kg and 30 mg/kg, respectively). Inhibitors were injected intraperitoneally daily using compounds targeting PAD4 (20mg/kg), CXCR2 (1mg/kg), NLRP3 (40 mg/kg), MMP9 (50 mg/kg), TGFβR1 (10mg/kg), and ITGαVβ1 (70 mg/kg). DNase (300 U/mouse) treatment was used to digest the DNA strands of NETs. Antibodies were injected intraperitoneally 3 times a week using anti-Ly6G (200 μg/mouse) and its IgG control (200 μg/mouse); anti-IL-1β (200μg/mouse) and its IgG control (200μg/mouse); anti-TGFβ (200μg/mouse) and its IgG control (200μg/mouse). Two weeks after treatments were started, mice were sacrificed for analysis otherwise stated in figure legends.

Primary tumor model and treatments

To generate mammary tumors, we injected 410.4 mouse cancer cells (1.2 million cells per mouse) in 100 μL of Matrigel/PBS (1:1 ratio) into the mammary fat pad of eight-week-old female BALB/c mice. After 5 weeks, chemotherapy treatment was started twice a week with Adriamycin/cyclophosphamide (AC, 1mg/kg and 30 mg/kg, respectively). PAD4 inhibitor was injected intraperitoneally daily (20mg/kg) and two weeks after chemotherapy treatment, mice were sacrificed, the tumor and lung removed for analysis. Tumors were pictured on millimeter paper to quantify tumor size using ImageJ.

Simultaneous lung metastasis and primary tumor model and treatment (For Table S3).

410.4 mouse cancer cells were injected the same day in the tail vein (300 000 cells per mouse in 100μL PBS) and in the mammary fad pad (2.5 million cells per mouse in 100μL Matrigel/PBS 1:1) of eight-week-old female BALB/c mice. After 2 weeks, chemotherapy treatment was started twice a week with Adriamycin/cyclophosphamide injected intraperitoneally (AC, 1mg/kg and 30 mg/kg, respectively) and inhibitors were injected intraperitoneally daily using PAD4 inhibitor (20mg/kg). Two weeks after chemotherapy treatment, mice were sacrificed, the tumor and lung removed for analysis.

Isolation of mouse neutrophils

Neutrophils were harvested from six to eight-week-old female BALB/c mice. The bone marrow of the femurs and tibias was isolated in sterile Hank’s buffered salt solution (HBSS 1X) without Ca²⁺/Mg2⁺. Bone marrow cells were flushed in HBSS and after centrifugation (400g, 5 min, 4°C), the cells were resuspended for 3 min on ice in an ammonium-chloride-potassium (ACK) buffer, and centrifuged (400g, 5 min, 4°C) before being washed twice with HBSS 1X (400g centrifugation was used for the washing steps, 5 min, 4°C). Neutrophils were then separated from mononuclear cells by plating 2 mL of the cell suspension onto a Percoll gradient consisting of 3 mL of 81% Percoll under 3 mL of 62% Percoll, followed by centrifugation at 700g for 20 min at 4°C. The middle layer containing the neutrophils was washed twice in HBSS 1X and cells were resuspended in serum-free DMEM.

Activation of neutrophils and preparation of neutrophil-conditioned media

Primary isolated neutrophils were cultured in 24-well plates (250,000 per well) containing 500 μL of serum-free DMEM and were activated overnight with PMA to induce NETs or with recombinant complement 5a (C5a) to induce degranulation. The study used conditioned media (CM) from unstimulated neutrophils (PMN CM), NET-containing conditioned media (CM, induced with PMA and referred as NET CM) and CM from degranulated neutrophils (obtained using complement 5a, PMN_C5a CM). GSK484 was added to inhibit PAD4 and therefore NET production and DNase I to digest NET-DNA. The next day, the neutrophil CM were collected, centrifuged at 450g for 10 min to remove neutrophils, and used to treat the cancer cell culture. To assess NET formation, an immunofluorescence was performed using histone H2B and myeloperoxidase on neutrophils cultured overnight on poly-L-lysine-coated coverslips in serum-free DMEM or CM (with vehicle or cytokines and/or inhibitors as indicated).

Effect of chemotherapy on cancer cells in vitro

Luciferase-expressing cancer cells (2,000) were plated in 96-well culture plates and incubated in complete media at 37°C overnight. For experiments with neutrophil CM, media were replaced the next day with 100 μL of neutrophil CM with the indicated treatment. For co-culture experiment, the next day, media were replaced, and neutrophils (10 000 per well) were plated with cancer cells in a serum-free medium with the indicated treatment. After 2 days, 100 μL of medium containing 5 μg/mL of luciferin were added and bioluminescence was measured (BLI) using a plate reader. To induce dormancy, D2.0R cells were cultured on Matrigel as described before ¹⁸. To quantify the number of GFP-4T07 cancer cells, GFP intensity was measured at excitation/emission wavelengths of 488/520 nm using a plate reader.

NETs quantification using sytox green

Neutrophils (100,000) were seeded onto 96-well plates and incubated in serum free DMEM or CM (with vehicle or cytokines and/or inhibitors as indicated) for 4 hours. Sytox green (impermeable DNA dye, 50 nM) was then added to the plate and after 5 minutes, fluorescence intensity was measured at excitation/emission wavelengths of 488/520 nm using a plate reader.

Conditioned media preparation

Cancer cells were grown in complete medium, washed twice with PBS, and subsequently incubated at 37°C in a serum-free medium with cisplatin, AC or as indicated in Figure 3E. After 48 hours, CM were collected, centrifuged at 500g for 5 min to remove cell debris and the supernatant were stored at −80°C. The CM generated from 410.4 cancer cells treated with cisplatin was referred as 410.4 Cis CM while the CM generated from cells treated with AC was referred as 410.4 AC CM. The same nomenclature was used for TS/A and PyMT cells.

Cytokine array

A Proteome Profiler Mouse XL Cytokine Array test was performed following the manufacturer’s instructions.

Chemotaxis assay

Chemotaxis was assayed using a boyden chamber with a 0.3 μm pore size cell culture insert. The upper and the lower wells were separated by a light-tight polyethylene terephthalate (PET) membrane. Serum-free DMEM or CM containing the indicated neutralizing antibodies and inhibitors were applied to the lower wells, and the cells (50,000) were seeded in each of the upper wells and incubated overnight at 37°C with 5% CO2. The next day, membranes were fixed in 4% PFA for 15 min, washed twice with PBS and stained for 10 min with Sytox green (5μM). Membranes were then mounted between glass slides and cover slips using mounting media. The number of migrating cells was then manually counted using confocal microscopy and a 20x objective.

Flow cytometry analysis

After treatment, the mice were sacrificed, and the lungs were removed. Lungs were then mechanically dissociated and digested at 37°C for 30 min with 2% FCS RPMI containing Dispase (2.5 U/mL), Collagenase D (0.1 mg/mL), DNase I (25 U/mL) and Liberase DL (0.2 mg/mL). Suspensions were then passed through a 70μm nylon cell strainer with RPMI and centrifuged at 600g for 1 min at 4°C. The supernatant was removed, and the cells were incubated at 4°C for 15 min in ice-cold Fluorescence Activated Cell Sorter (FACS) buffer (2% FCS and 0.5 mM EDTA) with diluted Fc Block (1:25). After one wash in FACS buffer, the cells were incubated with conjugated antibodies for 30 min at 4°C in the dark which allowed the detection of different immune cell populations. Cells were then washed twice in FACS buffer and resuspended in 450 μL of FACS buffer before being analysed using a CytoFLEX operated by Cytexpert software. The different immune cell populations were detected among CD45 positive cells. Among TCRβ cells, CD8 Lymphocytes were detected as CD8⁺ cells, and CD4 Lymphocytes were detected as CD4⁺ cells. B cells were detected as CD19⁺ cells. NK cells were detected as NKp46⁺ cells. Dendritic cells were detected as CD11c⁺ and CD11b⁻ cells. Among CD11b⁺ cells, monocytes were detected as Ly6C⁺ cells, neutrophils as Ly6G⁺ cells, and macrophages as F4/80⁺ cells. For blood analysis, blood was obtained from cardiac blood collection using a syringe with a 25G needle. Blood was then directly transferred in a 15ml EDTA-coated tube (0.5 M EDTA, pH=8), incubated with 10 ml of ACK lysis buffer for 5 min at room temperature and centrifuged for 5 min at 400g. Cells were then resuspended in 3ml of ACK lysis buffer for 2 minutes at room temperature, centrifuged (400g, 5 min) and resuspended in FACs buffer (using the volume of blood collected originally). For bone marrow cells analysis, bone marrow cells were flushed in HBSS and after centrifugation (400g, 5 min, 4°C), the cells were resuspended for 3 min on ice in an ammonium-chloride-potassium (ACK) buffer, and centrifuged (400g, 5 min, 4°C) before being washed twice with HBSS 1X (400g centrifugation was used for the washing steps, 5 min, 4°C) and resuspended in FACs buffer (10 million/ml). Blood and bone marrow cells suspensions were then incubated with CD45, Ly6G, Ly6C and CD11b antibodies for FACS analyses, following the same protocol described as above for lung cells.

Antibodies

For Western Blot analysis, antibodies against Latent Associated Protein, TGFβ, LBTP, pIκB, IκB, pSMAD2, SMAD2, Twist1, N-cadherin, E-cadherin, Twist1, Zeb1, Snail, Claudin 1 and tubulin were used.

For immunofluorescence staining of in vitro cultures, antibodies against myeloperoxidase and Histone H2B were used to assess NET formation, and an antibody against ASC was used to assess inflammasome oligomerization. Antibodies against ITGβ1, ITGαv, MMP9 and TGFβ were used to assess the presence of these proteins on NETs.

For immunofluorescence on tissue samples, antibodies against myeloperoxidase and citrullinated histone H3 were used to detect NETs and antibodies against N-cadherin and Claudin1 were used to detect the epithelial/mesenchymal status of cancer cells.

The neutralizing/blocking antibodies used in vitro were: anti-integrin β1 antibody clone Ha2/5, anti-integrin αV, anti-mouse CXCL5, CXCL1 and IL-1β and were used at 10 μg/mL.

For NET-ELISA, an anti-neutrophil elastase antibody and an anti-DNA peroxidase conjugated antibody were used.

For flow cytometry analysis, antibodies against CD4 (1:50), CD8a (1:100), TCR β (1:50), CD19 (1:50), CD45 (1:50), NKp46 (1:20), F4/80 (1:20), CD11b (1:50), CD11c (1:50), Ly6C (1:50), Ly6G (1:50), and CD107a (1:20) were used.

Immunostaining of tissue samples

For paraffin-embedded tissue, the lungs were fixed overnight at 4°C in 4% PFA, rinsed with PBS and transferred into 70% ethanol, processed using conventional methods, embedded in paraffin, and sectioned at 8 μm. Paraffin-embedded tissue sections were deparaffinized and rehydrated, and antigen retrieval was performed in EDTA buffer (10 mM Tris Base, 1 mM EDTA solution, 0.05% Tween 20, pH=9.0). Sections were blocked with Fc Receptor blocker and incubated with 1X blocking buffer (5% donkey serum, 2.5% BSA, 0.1% Triton X-100 in PBS). Then, sections were incubated overnight at 4°C with anti-myeloperoxidase (1:100) and anti-citrullinated histone H3 antibodies (1:250) (to detect neutrophils and NETs), and with anti-claudin1 (1:150) or anti-N-cadherin (1:150) (to detect EMT phenotype), in 0.5X blocking buffer. After three washes with PBS, the sections were incubated with the suitable fluorochrome-conjugated secondary antibodies (1:150) in 0.5X blocking buffer for 45 min in the dark at room temperature. After two washes with PBS and one with water, sections were counterstained with DAPI and rinsed in water, and the slides were mounted onto coverslips using mounting media.

Hematoxylin and Eosin staining

Paraffin-embedded tissue sections were first deparaffinized and rehydrated. The slides were then incubated with hematoxylin (15 min), an ammonia solution (0.08 % in water for 10 sec), and eosin (30 sec) and washed with tap water between each step. After dehydration, the slides were mounted onto cover slips using mounting media. Metastasis areas and foci were then quantified with ImageJ software.

Plasma sample collection from mice

Plasma samples were collected from cardiac blood using a syringe with a 25 G needle and placed into EDTA tubes. Whole blood was centrifuged at 4°C at 1300g for 10 min, and the top plasma layer was collected.

Plasma sample collection from human and storage

Blood samples were collected in Cell-Free DNA BCT tubes. Plasma was isolated by centrifugation at 1500g for 10 min, RT and then stored at −80°C before use.

Enzyme-linked immunosorbent assay (ELISA)

For NETs, 96-well Enzyme ImmunoAssay/Radio ImmunoAssay (EIA/RIA) plates were coated overnight at 4°C with an anti-elastase antibody (1:250) in 15 mM of Na2CO3, 35 mM of NaHCO3, at pH 9.6. The next day, the wells were washed three times with PBS, blocked in 5% BSA for two hours at room temperature, and washed three times with PBS. Then, 50 uL of plasma samples were added to the wells, incubated for two hours at room temperature on a shaker, and plates were washed three times with wash buffer (1% BSA, 0.05% Tween 20 in PBS). Next, an anti-DNA-peroxidase conjugated antibody (1:50) in 1% BSA in PBS was added to the wells for 2 hours at room temperature, and the wells were washed five times with wash buffer before the addition of 2,2′-azino-bis (3-ethylbenzothiazoline-6-sulphonic acid). Optical density was read 40 min later at 405 nm using a plate reader.

An IL-1β kit and a TGF-β1 kit were used on CM as indicated and following the manufacturer’s instructions.

For creatinine and urea analysis, ELISA kits were used to evaluate the plasma concentration of creatinine and urea, following the manufacturer’s instructions.

An IL-1β kit, a CXCL1 and a CXCL5 kits were used on normal lung, metastatic lungs, mammary fat pad and breast primary tumor cell lysate as indicated and following the manufacturer’s instructions. Tissue lysate was isolated by washing the tissues once with cold PBS, cutting them into smaller pieces, and homogenizing thoroughly in radioimmunoprecipitation assay (RIPA) buffer (50 mM Tris HCl [Ph=7.4], 1% NP-40, 0.5% sodium deoxycholate, 0.1% SDS, 150 mM NaCl, 50 mM NaF, and 2 mM EDTA) with protease inhibitors. For 10 mg tissue, 500 μL of RIPA buffer was used. After 30 min on ice, the samples were centrifuged at 10,000 rpm for 20 min at 4 °C, and protein concentration was determined on the supernatant using a bicinchoninic acid (BCA) protein assay. 200 μg of proteins were used for each ELISA assay.

ATP quantification

ATP was quantified in CM using an ATP determination kit following the manufacturer’s instructions.

Cell culture reagents

The following reagents were used as stated, unless otherwise mentioned in figure legends. Phorbol 12-myristate 13-acetate (PMA) was used at 20 nM; recombinant mouse complement 5a was used at 100 ng/mL. Recombinant IL-1β and TGFβ-1 were used at 1 ng/mL. Cisplatin was used at 10 μM and Adriamycin/Cyclophosphamide (referred as AC) were used at 0.1 μM and 1 μM respectively. The PAD4 inhibitor GSK484 was used at 10 μM to inhibit NET formation, and 1.5 units/mL of DNase I was used to digest NET scaffolds. The NRPL3 inhibitor MCC950 was used at 5 μM; AC-YVAD-CMK was used at 100 μM to inhibit caspase 1 and Bay11-7082 was used at 10 μM to inhibit NF-κB activity; Sivelestat, MMP9 inhibitor 1, Cathepsin G inhibitor 1, 4-Aminoebnzoic Acid hydrazide were used at 10 μM to inhibit NE, MMP9, Cathepsin G and MPO activity respectively, RGD peptide was used at 1 mM to counteract integrin β1-dependant adhesion; αVβ1 integrin-IN-1 TFA (HY-100445A, Medchem Express) was used at 100 μM to inhibit ITGαvβ1 activity; anti-Integrin β1 antibodies clone Ha2/5 and anti-integrin αv were used at 10 μg/mL to inhibit integrin β1 and αv activity respectively; anti-mouse CXCL5, CXCL1 and IL-1β were used at 10 μg/mL.

Western blot

For immunoblotting analysis, cells grown on plastic plates were lysed on ice in lysis buffer (25 mM Tris [pH=6.8], 2% sodium dodecyl sulfate (SDS), 5% glycerol, 1% β-mercaptoethanol, 0.01% bromophenol blue) and the samples were sonicated. Equal amounts of protein from each sample were loaded on SDS-polyacrylamide gel electrophoresis, separated, and transferred onto nitrocellulose. The immunoblots were incubated in blocking buffer (5% BSA, 10 mM Tris-HCl [pH=7.5], 500 mM NaCl) for 30 min at room temperature and probed with specific antibodies overnight at 4°C. Then, the immunoblots were washed three times for 10 min in Tris-buffered saline Tween 20 (TBST, 10 mM Tris-HCl [Ph=7.5], 500 mM NaCl, 0.1% Tween 20), incubated with secondary antibodies for one hour at room temperature in blocking buffer, and washed three times in TBST again. Immunodetection was performed using chemiluminescent horseradish peroxidase (HRP) substrate.

Immunofluorescence of cell cultures

Cells were fixed with 4% paraformaldehyde (PFA) for 20 min. After fixation, they were rinsed twice in PBS, incubated in 50 mM of NH₄Cl for 10 min and permeabilized with 0.5% Triton X-100 for 5 min. Cells were next blocked in PBS containing 1% bovine serum albumin (BSA) for 60 min and incubated with anti-H2B (1:200), anti-myeloperoxidase (1:400) antibodies in blocking buffer overnight at 4°C. After two washes in PBS, cells and matrices were incubated in the presence of fluorochrome-conjugated secondary antibodies (1:250) for 40 min, rinsed twice in PBS, stained with 4’,6-diamidino-2-phenylindole (DAPI) for 5 min, rinsed in water, and the coverslips were mounted onto glass slides using mounting media.

mRNA analysis

RNA was isolated from total cell lysates using a RNeasy Mini kit according to the manufacturer’s instructions. Reverse transcription was performed using 500ng cytoplasmic RNA using Superscript II reverse transcriptase. Real time PCR was performed using Fast SYBR Green Master Mix in duplicates according to the recommendations of the manufacturer on an ABI Prism 7900HT sequence detection system (Applied Biosystems, Foster City, CA).

Primer sequences are listed in the Key resources table and Table 8.

Key resources table.

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Latent Associated Protein (LAP) antibody	Novus	Cat #NBP2-22114SS; RRID:AB_2909813
TGFβ	Cell Signaling Technology	Cat #3711; RRID:AB_2063354
Latent Binding TGFβ Protein (LTBP)	Abcam	Cat #ab78294; RRID:AB_1952060
pIκB	Cell Signaling Technology	Cat #2859T; RRID:AB_561111
IκB	Cell Signaling Technology	Cat #4814T; RRID:AB_390781
tubulin	Sigma	Cat #T4026; RRID:AB_477577
Twist1	Thermo Fisher Scientific	Cat #PA5116093; RRID:AB_2900727
Zeb1	Thermo Fisher Scientific	Cat #PA595862; RRID:AB_2807664
Snail	Thermo Fisher Scientific	Cat #PA5115940; RRID:AB_2900574
E-cadherin	Cell Signaling Technology	Cat #3195S; ID UniProt : P12830
Phospho Smad2	Cell Signaling Technology	Cat #3108; ID UniProt : Q15796
Smad2	Cell Signaling Technology	Cat #3103; ID Uniprot : Q15796
myeloperoxidase	Dako	Cat #A0398; RRID:AB_2335676
Histone H2B	Abcam	Cat #ab52484; RRID:AB_1139809
RFP (mcherry)	Rockland	Cat #600-401-379; RRID:AB_2209751
ASC	EMD Millipore	Cat #04-147; RRID:AB_1977033
ITGβ1	BD Biosciences	Cat #555003; RRID:AB_395637
ITGαv (clone RMV-7)	Invitrogen	Cat #14-0512-82; RRID:AB_467296
MMP9	Abcam	Cat #ab38898; RRID:AB_776512
TGFβ	Cell Signaling Technology	Cat #3711; RRID:AB_2063354
anti-myeloperoxidase antibody (used for mouse tissue)	R & D Systems	Cat #AF3667; RRID:AB_2250866
anti-myeloperoxidase antibody (used for human tissue)	R & D Systems	Cat #MAB3174; RRID:AB_2250873
anti-citrullinated histone H3 antibody (used for mouse and human tissue)	Abcam	Cat #ab5103; RRID:AB_304752
anti-N-cadherin	Abcam	Cat #ab18203; ID Uniprot:P19022
anti-Claudin1	Thermo Fisher Scientific	Cat #71-7800; RRID:AB_2533997
Donkey anti-Goat IgG secondary antibody, Alexa Fluor 647	Invitrogen	Cat #A21445; RRID:AB_2535862
Donkey anti-Rabit IgG secondary antibody, Alexa Fluor 568	Invitrogen	Cat #A10042; RRID:AB_2534017
ITGβ1 (clone Ha2/5)	BD Biosciences	Cat #555003; RRID:AB_395637
ITGαv (clone RMV-7)	Invitrogen	Cat #14-0512-82; RRID:AB_467296
CXCL5	R & D Systems	Cat #Mab433; RRID:AB_2086587
CXCL1	R & D Systems	Cat #Mab453; RRID:AB_2087696
IL-1β	R & D Systems	Cat #AF-401-NA; RRID:AB_416684
anti-mouse Neutrophil Elastase	Santa Cruz Biotechnology	Cat #sc-9521; RRID:AB_2096537
anti-human Neutrophil Elastase	Santa Cruz Biotechnology	Cat #sc-9520; RRID:AB_650173
anti-DNA peroxidase conjugated	Roche	Cat#11774425001; RRID:AB_2909412
CD4 (clone GK1.5)	BD Biosciences	Cat #612974; RRID:AB_2870246
CD8a (clone REA601)	Miltenyi Biotec	Cat #130-120-806; RRID:AB_2752203
TCR β (clone REA318)	Miltenyi Biotec	Cat #130-104-856; RRID:AB_2654019
CD19 (clone T45-2342)	Miltenyi Biotec	Cat #130-112-042; RRID:AB_2655828
CD45 (clone REA737)	Miltenyi Biotec	Cat #130-110-803; RRID:AB_2658224
NKp46 (clone 29A1.4)	BD Biosciences	Cat #741029; RRID:AB_2740647
F4/80 (clone T45-2342)	BD Biosciences	Cat #743281; RRID:AB_2741399
CD11b (clone REA754)	Miltenyi Biotec	Cat #130-113-808; RRID:AB_2751173
CD11c (clone REA754)	Miltenyi Biotec	Cat #130-110-837; RRID:AB_2654705
Ly6C (clone REA796)	Miltenyi Biotec	Cat #130-111-920; RRID:AB_2652814
Ly6G (clone 1A8)	BD Biosciences	Cat #561236; RRID:AB_10611860
CD107a (clone 1D4B)	BD Biosciences	Cat #565533; RRID:AB_2739285
CD326	Biolegend	Cat #118214; RRID:AB_1134102
anti-Ly6G antibody (clone 1A8)	BioXCell	Cat #BP0075-1; RRID:AB_1107721
IgG for antiLy6G antibody (clone 2A3)	BioXCell	Cat #BP0089; RRID:AB_1107769
anti-IL-1β antibody (clone B122)	BioXCell	Cat #BE0246; RRID:AB_2687727
IgG for anti-IL-1β antibody	BioXCell	Cat #BE0091; RRID:AB_1107773
anti-TGFβ antibody (clone 1D11.16.8)	BioXCell	Cat #BE0057; RRID:AB_1107757
IgG for anti-TGFβ antibody (clone MOPC-21)	BioXCell	Cat #BE0083; RRID:AB_1107784
Biological samples
Human plasma samples from metastatic breast cancer patients	Institut Curie, Paris, France	Table S5
Paraffin-embeded breast cancer tissue	University of Southern Denmark	Table S6
Chemicals, peptides, and recombinant proteins
Cisplatin	Sigma	Cat #C2210000
Doxorubicin	Sigma	Cat #D1515
Cyclophosphamide	Sigma	Cat #C0768
PAD4 inh.	Cayman Chemical	Cat #17488
CXCR2 inh.	Selleckchem	Cat #SB225002
NLRP3 inh. (MCC950)	EMD Millipore	Cat #5.38120.0001
MMP9 inh. (in vivo)	Medchem Express	Cat #HY-12354
TGFβR1 inh.	Selleckchem	Cat #SB431542
ITG αVβ1 inh.	Medchem Express	Cat #HY-100445A
DNAse I	Sigma	Cat #04536282001
HBSS 1X without Ca2+/Mg2+	Gibco	Cat #14175-053
ACK	Gibco	Cat #A10492-01
Percoll	GE Healthcare	Cat #17-0891-02
serum-free DMEM (Dulbecco’s Modified Eagle Medium)	Gibco	Cat #21969-035
Dharmafect 3	Dharmacon	Cat #T-2003-02
Opti-MEM	Gibco	Cat #31985-047
Cytiva HyClone™ Sérum FetalClone™ II (FCII)	Cytiva	Cat #SH3006603
Matrigel	Corning	Cat #356234
Luciferin	Goldbio	Cat #Luck-1G
Phorbol 12-myristate 13-acetate (PMA)	Sigma	Cat #P1585
Recombinant mouse complement 5a (C5a)	R & D Systems	Cat #2150-C5/CF
Recombinant IL-1β	Peprotech	Cat #211-11B
Recombiant TGFβ-1	Biolegend	Cat #763102
Caspase I inh. AC-YVAD-CMK	abcam	Cat #SML0429
NF-kB inh. Bay 11-7082	Cayman Chemical	Cat#10010266
NE inh. Sivelestat	Tocris	Cat #3535
MMP9 inh. 1 (in vitro)	Calbiochem	Cat #44278
Cathepsin G inhibitor 1	EMD Millipore	Cat #219372
Myeloperoxidase inh. (4-Aminoebnzoic Acid hydrazide)	Cayman Chemical	Cat #14845
RGD peptide	abcam	Cat #A8052
Sytox green	Thermo Fisher Scientific	Cat #S7020
Chemiluminescent horseradish peroxidase (HRP) substrate	GE Healthcare	Cat #RPN 2109
RPMI	Gibco	Cat #31870-025
Dispase	Stem Cell	Cat #07913
Collagenase D	Sigma	Cat#11088866001
Liberase DL	Sigma	Cat #05466202001
Fc block	BD Biosciences	Cat#553142
PFA	Electron Microscopy Science	Cat #15714
Triton X-100	Thermo Fisher Scientific	Cat #BB151-500
Fc receptor blocker	Innovex	Cat #NB309
Donkey serum	Sigma	Cat #D9663
BSA	Sigma	Cat #A3294
DAPI	Thermo Fisher Scientific	Cat #D1306
Hematoxylin	Thermo Fisher Scientific	Cat #6765008
Eosin	Sigma	Cat #HT110180
Polybrene	Sigma	Cat #TR-1003-G
Puromycine	Sigma	Cat #P8833
L-Glutamine	Gibco	Cat #25030
Penicillin-streptomycin	Gibco	Cat #15140-122
Critical commercial assays
96-well Enzyme ImmunoAssay/Radio ImmunoAssay (EIA/RIA)	Costar	Cat #3590
ELISA kit for TGF-β1	R & D Systems	Cat #DY1679
ELISA kit for Urea	MyBiosource	Cat #MBS2611085
ELISA kit for creatinine	MyBiosource	Cat #MBS763433
ELISA kit for IL-1β (for tissue analysis)	R & D Systems	Cat #MLB00C
ELISA kit for IL-1β (for CM analysis)	Thermo Fisher Scientific	Cat #88-7013-22
ELISA kit for CXCL1/KC	R & D Systems	Cat #DY453
ELISA kit for CXCL5 (LIX)	R & D Systems	Cat #P313546
ATP determination kit	Thermo Fisher Scientific	Cat #A22066
RNeasy Mini kit	Qiagen, Turnberry Ln Valencia, CA	Cat #217004
Proteome Profiler Mouse XL Cytokine Array test	R & D Systems	Cat #ARY028
PierceTM BCA Protein Assay Kit	Thermofisher	Cat #23225
Illumina Stranded mRNA Prep	Illumina	Cat #20040532
Deposited data
RNA-seq data	This paper	GEO: GSE218119
Experimental models: Cell lines
410.4 murine cancer cells	Blazquez et al. 68	Obtained from Dr. Tobias Pukrop (University Hospital Regensburg, Regensburg, Germany)
PyMT murine cancer cells	This paper	Obtained from Dr. Bodu Liu (Cold Spring Harbor Laboratory, Cold Spring Harbor, NY 11724, USA)
TS/A murine cancer cells	Merckmillipore	Cat. # SCC177
D2.OR murine cancer cells	Barkan et al. 69	Gift from Dr. Jeffrey Green (National Institutes of Health)
D2.A1 murine cancer cells	Barkan et al. 69	Gift from Dr. Jeffrey Green (National Institutes of Health)
4T07 murine cancer cells	Avgustinova et al. 70	Gift from Dr. Erik Sahai (The Francis Crick Institute, London, UK). RRID:CVCL_B383
4T1 murine cancer cells	ATCC	Cat. #CRL-2539
293T cells	ATCC	Cat. #CRL-3216
Experimental models: Organisms/strains
Mouse Balb/c female	Charles River	Cat #028
Mouse C57BL/6 female	Charles River	Cat #027
Oligonucleotides
siRNA ON TARGET Plus Non-targeting pool target sequences: UGGUUUACAUGUCGACUAA; UGGUUUACAUGUUGUGUGA; UGGUUUACAUGUUUUCUGA; UGGUUUACAUGUUUUCCUA	Dharmacon	Cat #D-001810-10-05
siRNA ON TARGET Plus mouse ccdc25 #1: -J-053396-09: GCAAGAUGAACAACGUUAA	Dharmacon	Cat #L-053396-01-0005
siRNA ON TARGET Plus mouse ccdc25 #2: -J-053396-10: CCAUGUGUACCUACGAUUA	Dharmacon	Cat #L-053396-01-0005
Additional primer sequences	This paper	Table S8
Software and algorithms
ImageJ software	ImageJ	N/A
GraphPad Prism software version 7	GraphPad	N/A
Cytexpert	Beckman Coulter	N/A
STAR	Dobin et al. 71	N/A
FeatureCounts	Liao et al. 72	N/A
DEseq2	Love et al. 73	N/A
ReactomePA R package	Yu et al. 74	N/A
Other
70μm nylon cell strainer	BD Falcon	Cat #352350
96-well plate	Falcon	Cat #353072
Plate reader FLUOstar Omega	BMG Labtech	N/A
CytoFLEX	Beckman Coulter	N/A
Mounting media immunostaining	Electron Microscopy Sciences	Cat #17985-16
Cytoseal 60 mounting media for histological staining	Thermo Fisher Scientific	Cat #8310-4
Syringe 25 G	Terumo	Cat #AN*2516R1
2,2′-azino-bis(3-ethylbenzothiazoline-6-sulphonic acid) (antibodyTS)	Thermo Fisher Scientific	Cat #37615
Superscript II reverse transcriptase	Invitrogen, Carlsbad, CA	Cat #18064-014
Fast SYBR Green Master Mix	Applied Biosystems, Foster City, CA	Cat #18064-014
Boyden chamber	Corning	Cat #351151
96-well Enzyme ImmunoAssay/Radio ImmunoAssay (EIA/RIA) plates	Costar	Cat #3590
Cell-Free DNA BCT tubes	Streck	N/A

Relative expression of the respective gene was determined after normalization to GAPDH and calculated using the following formula: relative expression level = 2ddCT.

Sample and library preparation for RNA-seq

410.4 cells (200,000 per well) were plated in 6 well plate in 5% FCII DMEM media. The day after, the media was removed and replaced by NET CM or fresh control media for 48h. RNA was isolated from total cell lysates using a RNeasy Mini kit according to the manufacturer’s instructions. 500ng of total RNA was used to prepare RNA libraries. There were generated with Illumina Stranded mRNA Prep and performed following the manufacturer’s instructions. Each library was normalized to 8pM and sequences on an Illumina NextStep 2000.

Analysis of RNA-seq data

FASTQ sequencing files were processed and analyzed using DRAGEN included in the Nextseq2000 Illumina sequencer. Reads were aligned to the GRCm39/mm10 reference mouse genome using STAR ⁷¹. Gene read counts were determined using FeatureCounts ⁷². Differential gene expression analysis was performed using DESeq2 ⁷³, with a 0.05 false discovery rate (FDR)-adjusted p-value cutoff for significance. Signaling pathways were revealed using ReactomePA R package ⁷⁴. The option “pAdjustMethod” was set to “fdr”. Raw data are available on GEO number GSE218119

siRNA Transfections

410.4 cancer cells were transfected using Dharmafect. Briefly, cells were plated at 60% confluence and subjected to transfection the following day in Opti-MEM using 100nM final concentration of siRNA. The next day, cells were detached using trypsin and used in a co-culture assay with freshly isolated neutrophils. SiRNA was purchased from Dharmacon and sequences are listed in the Key resources table.

Quantification and statistical analysis

For the quantification of the lung metastatic burden, the number of lung metastatic foci, NET-ELISA, lung neutrophils and NETs (based on immunofluorescence imaging), BUN and plasma creatinine: the vehicle, cisplatin and AC mice groups used for the following experiments were the same as the experiments were conducted at the same time to reduce the number of animals used in our study. i) mice with 410.4 cells and treated with the PAD4 inh. and the NLRP3 inh.; ii) mice with 410.4 cells and treated with the CXCR2 inh., the TGFβR1 inh., the MMP9 inh. and the ITGαvβ1 inh.; iii) Mice with PyMT cells and treated with the PAD4 inh., DNase I, and the TGFβR1 inh.; iv) Mice with TS/A cells and treated with the PAD4 inh., DNase I, and the TGFβR1 inh; v) Non tumor-bearing mice treated with the PAD4 inh., the NLRP3 inh., the TGFβR1 inh and the MMP9 inh. For the quantification of lung neutrophils using Flow Cytometry (Figures S1I and S1X), the same Veh., cisplatin and AC mice groups were used for the CXCR2 inh. and the PAD4 inh. as the experiments were conducted at the same time to reduce the number of animals used in our study.

Statistical tests are presented within figure legends. A p-value less than 0.05 was considered significant, and p-values are indicated in figures as ***P<0.001, **P<0.01, and *P<0.05. All statistical analyses were performed using GraphPad Prism software version 7, unless otherwise stated.

• Chemotherapy promotes NET formation in breast cancer lung metastasis

• NETs promote breast cancer lung metastasis chemotherapy resistance

• NETs activate latent TGFβ, promoting EMT in cancer cells

• Targeting NETs ameliorates chemotherapy efficacy against lung metastasis

Inclusion and diversity

We support inclusive, diverse, and equitable conduct of research.