Microenvironmental Cues Determine Tumor Cell Susceptibility to Neutrophil Cytotoxicity

Metastatic progression is a complex and multistep process where tumor cells acquire unique traits required for colonizing a distant organ (1). The fact that both primary and metastatic tumors display epithelial morphology prompted the idea that a transient phenotypic transformation, to a mesenchymal-like phenotype, is required for metastatic progression. Indeed, several groundbreaking studies have shown that epithelial-to-mesenchymal transition (EMT) is required for metastatic spread (2–5). Similarly, as ultimately cells undergoing EMT give rise to epithelial metastatic tumors, mesenchymal-to-epithelial transition (MET) was found to be critical for metastatic progression (6). Intriguingly, recent studies have demonstrated that EMT/MET are not required for establishing metastases (7–10). Taking into account heterogeneity between tumor types and the cellular heterogeneity within tumors, it is possible that some tumors will require EMT and MET for metastatic progression, whereas others will not (11).

Another component that was found to hold a very important role in metastatic progression is the tumor microenvironment (12). In this context, tumor-infiltrating neutrophils were shown to play a role in promoting tumor cell dissemination (13) and metastatic seeding (14–17). On the other hand, neutrophils were also shown to limit tumor progression and limit metastatic seeding of circulating tumor cells (CTC; refs. 18, 19, and others). In a recent study (20), we explored the mechanism utilized by antitumor neutrophils to eliminate tumor cells and found that neutrophil-secreted H₂O₂ induces a lethal influx of Ca²⁺ in tumor cells. The Ca²⁺ influx is mediated by the transient receptor potential cation channel, subfamily M, member 2 (TRPM2), a ubiquitously expressed H₂O₂-dependent Ca²⁺-permeable channel (21) that is frequently upregulated in cancer.

Although tumor cells at the primary site express TRPM2 they are not targeted by cytotoxic neutrophils. In contrast, disseminated tumor cells are efficiently targeted and eliminated by neutrophils, which limit metastatic seeding. This has led us to question whether disseminated tumor cells are more susceptible to neutrophil cytotoxicity than cells at the primary site and whether EMT plays a role in this process. We now show that TRPM2 expression is elevated in CTC compared with the primary tumor, rendering CTC more susceptible to neutrophil cytotoxicity. Furthermore, we demonstrate that TRPM2 expression is regulated by signals driving EMT/MET, which increase (EMT) or decrease (MET) tumor cell susceptibility to neutrophil cytotoxicity. Finally, we demonstrate that while EMT is often required for metastatic spread (22–24), the upregulation of TRPM2 during this process renders tumor cells more susceptible to neutrophil cytotoxicity. In contrast, cells undergoing MET are protected from neutrophil cytotoxicity. Taken together, our data demonstrate how signals in the different microenvironments determine tumor cell susceptibility to neutrophil cytotoxicity via modulation of TRPM2 expression. Furthermore, our data show that while EMT may be required for metastatic progression, MET at the distant site reduces the susceptibility to neutrophil cytotoxicity and may be regarded as a novel mode of immune evasion.

Five- to 6-week-old BALB/c mice were purchased from the Harlan. All experiments involving animals were approved by the Hebrew University's Institutional Animal Care and Use Committee.

4T1 mouse breast cancer cells were purchased from the ATCC and cultured in DMEM containing 10% heat-inactivated fetal calf serum (FCS). 4T1 cells were transduced with a lentiviral vector (pLVX-Luc, MigR1-Luc, or the triple-modality reporter gene TGL; ref. 25) to stably express firefly luciferase. The human mammary epithelial (HMLE) cell line kindly provided by R.A. Weinberg was cultured in 1:1 DMEM/HAMF12 medium (Biological Industries) complemented with 10% FCS, 10 ng/mL hEGF (PromoCell), 0.5 μg/mL hydrocortisone (Sigma) and 10 μg/mL insulin (Sigma). Cells were not authenticated, were never passaged for more than 5 times, and were tested routinely (monthly) for Mycoplasma.

Luciferase-labeled tumor cells (1 × 10⁴/well) were plated in 100 μL RPMI-1640 with 2% FCS in white 96-flat bottom wells (Corning). After 4 hours of incubation, 100 μL purified neutrophils (1 × 10⁵/well) or H₂O₂ (10–25 μmol/L) were added to the plated tumor cells and cocultured overnight in the presence or absence of TGFβ (10 ng/mL; R&D Systems). Following overnight incubation, luciferase activity was measured using Tecan F200 microplate luminescence reader. The extent of killing was determined by the ratio between tumor cells cultured alone and tumor cells cocultured with neutrophils. In vitro experiments were repeated at least 3 times.

EMT was induced by incubating 2 × 10⁵ tumor cells in the presence of 10 ng/mL TGFβ for 14 to 21 days. Alternatively, cells were incubated in serum free medium for 12 hours and then transferred to complete media containing TGFβ (10 ng/mL) for 4 days.

The 4T1 or HMLE cells infected with the tamoxifen-inducible lentiviral pWZL-Blast-Twist-ER vector were treated with 100 nmol/L 4-hydroxy-tamoxifen (4OHT, Sigma) for 21 days. Control cells were treated with vehicle.

The 4T1 cells were transduced with pBP-Snail retroviral vector and monitored for 7 to 10 days.

MET was induced by growing cells in low serum medium (0.5%) supplemented with 200 ng/mL of recombinant human BMP7 (cat. #120-03; PeproTech). Cells were treated for 5 to 7 days.

4T1 cells were infected with doxycycline inducible lentiviral TetO-FUW-Klf4 vector. Cells were treated with 2 μg/mL doxycycline for 4 to 5 days.

The open reading frame of firefly luciferase was prepared from pLVX-luciferase using forward primer 5′-CAGTCCGCTCGAGGCCGCCACCATGGAAGACGCCAAAAACATAAAG-3′ containing an XhoI restriction site, Kozac sequence followed by the ATG initiation code and reverse primer 5′-ACTTCCGGAATTCTTACACGGCGATCTTTCCGCCC-3′ containing an EcoRI restriction site and the stop codon and Phusion Flash High-Fidelity PCR Master Mix (Thermo Scientific). The amplified luciferase PCR product was digested by XhoI and EcoRI and inserted into the XhoI/EcoRI site of the MigR1 plasmid. The tamoxifen-inducible pWZL-Blast-Twist-ER retroviral vector and the pBP-Snail retroviral vector were kindly provided by Dr. Ittai Ben-Porath (The Hebrew University of Jerusalem, Jerusalem, Israel). Doxycycline-inducible lentiviral TetO-FUW-Klf4 vector was kindly provided by Dr. Yossi Buganim (The Hebrew University of Jerusalem, Jerusalem, Israel).

Neutrophils were purified from orthotopically injected (mammary fat pad) 3-week 4T1 tumor-bearing mice as described before (26). In brief, whole blood was collected by cardiac puncture using heparinized (Sigma) syringe. The blood was diluted with 5 volume of PBS containing 0.5% BSA and subjected to a discontinuous Histopaque (Sigma) gradient (1.077 and 1.119). NDN were collected from the 1.077–1.119 interface. Red blood cells (RBC) were eliminated by hypotonic lysis. Neutrophil purity and viability were determined visually and were consistently >98% for NDN.

Mouse α-E-cadherin (BD Biosciences, 610182), goat α-vimentin (Abcam, ab11256), rabbit α-vimentin (Cell Signaling, D21H3), donkey α-goat 488 (Jackson ImmunoResearch, 705545147), goat α-mouse 568 (Invitrogen, A11031). Rabbit α-Klf4 (Santa Cruz Biotechnology, sc-20691), mouse α-β-actin (Abcam, 8224), second antibody HRP-conjugated anti-mouse IgG (Jackson ImmunoResearch, 115-035-062), second antibody HRP-conjugated anti-rabbit IgG (Jackson ImmunoResearch, 111-035-144).

Female BALB/c mice were orthotopically (mammary fat pad) injected with parental 4T1 tumor cells. On day 6, the mice were randomized and injected daily with either control IgG or neutrophil-depleting anti-Ly6G (50 μg/mouse i.p.). On day 8, the mice were injected to the tail vein with 5 × 10⁵ GFP⁺ cells (control vs. BMP7, Klf4 overexpression, or TGFβ-treated cells). On day 10, the lung-associated GFP⁺ cells were analyzed by fluorescent microscopy.

Total RNA was isolated with TRI-Reagent (Sigma) according to the manufacturer's instructions. RNA was reverse transcribed into cDNA using the AB high-capacity cDNA kit (Applied Biosystems). Two nanograms of converted cDNA was used for each reaction, using Kapa Sybr-Green Master Mix (Kapa Biosystems) and 300 nmol/L of forward/reverse primer set. Analyses were done using the CFX384 BioRad Real-Time PCR. GAPDH was used as a reference gene for mouse analyses, and Actin was used as a reference gene for human analyses. The following primer sequences were used for mouse gene expression analyses: GAPDH-F 5′-GCCTTCCGTGTTCCTACC-3′, GAPDH-R 5′-CCTGCTTCACCACCTTCTT-3′, E-Cadherin-F 5′-TCCTGCTGCTCCTACTGTTTC-3′, E-Cadherin-R 5′-TCTCCACCTCCTTCTTCATCA-3′, Vimentin-F 5′-AACTGCACGATGAAGAGATCCA-3′, Vimentin-R 5′-GGACATGCTGTTCCTGAATCTG-3′, TRPM2-F 5′-GACCCAAGGAACACAGACAA-3′, TRPM2-R 5′-AGCCTCTTCAGCTCCATATCA-3′. Klf4-F 5′-GCAGTCACAAGTCCCCTCTC-3′, Klf4-R 5′-CACGACCTTCTTCCCCTCTT-3′.

The following primer sequences were used for human gene expression analyses: β-Actin-F 5′-ATGGATGATGATATCGCCGCGCTC-3′, β-Actin-R 5′-ATAGGAATCCTTCTGACCCATGCC-3′, Vimentin-F 5′-CCTTGAACGCAAAGTGGAATC-3′, Vimentin-R 5′-GACATGCTGTTCCTGAATCTGAG-3′, TRPM2-F 5′-TCGGACCCAACCACACGCTGTA-3′, TRPM2-R 5′-CGTCATTCTGGTCCTGGAAGTG-3′, E-Cadherin-F 5′-GTCACTGACACCAACGATAATCCT-3′, E-Cadherin-R 5′-TTTCAGTGTGGTGATTACGACGTTA-3′, Twist-F 5′-GGAGTCCGCAGTCTTACGAG-3′, Twist-R 5′-TCTGGAGGACCTGGTAGAGG-3′, Snail-F 5′-CCTCCCTGTCAGATGAGGAC-3′, Snail-R 5′-CCAGGCTGAGGTATTCCTTG-3′.

Tissues were harvested, minced, and digested in L15 medium (Sigma-Aldrich) containing 0.2 mg/mL collagenase type I, 0.05 mg/mL collagenase II, 0.2 mg/mL collagenase type IV (all from Sigma-Aldrich), 0.025 mg/mL DNase I (Roche Applied Science), and 0.025 mg/mL elastase at 37°C for 40 minutes. The digested tissue mixture was then filtered through a 70-μm filter and centrifuged at 1,250 rpm, 5 minutes at room temperature. The pellet was then resuspended in RBC lysis buffer for 1 minute, RPMI medium was added to stop the lysis, and centrifuged again 1,250 rpm, 5 minutes at room temperature. Pellet was resuspended in EasySep buffer, and Ly6G⁺ cells were isolated using the EasySep PE-positive Selection Kit (STEMCELL Technologies) with anti-mouse Ly6G PE antibody (BioLegend) according to the manufacturer's protocol.

The lungs or tumor tissues and 4T1 cells were homogenized in lysis buffer (25 mmol/L Tris [pH 7.5], 1% Triton X-100, 150 mmol/L NaCl, 5 mmol/L EGTA) and protease inhibitor cocktail (Sigma, P8340). The lysates were centrifuged at 14,000 × g for 20 minutes at 4°C, and the supernatants were transferred to new tubes. Samples (20 μg) of total protein were separated on 7.5% to 12% gel, transferred to a PVDF membrane, and blotted with primary for overnight then HRP secondary antibody for 1 hour. Proteins were detected using ECL.

For studies comparing differences between two groups, we used unpaired Student t tests. Differences were considered significant when P < 0.05. Data are presented as mean ± SEM.

Neutrophils have the capacity to kill disseminated tumor cells and limit metastatic seeding. They accumulate in large numbers in both primary tumor and premetastatic lung but are essentially excluded from metastatic tumors (Fig. 1A and B and Supplementary Fig. S1). Although neutrophils effectively limit metastatic seeding, they have no significant effect on 4T1 primary tumor growth or metastatic tumor growth (18). This suggests that the tumoricidal interaction between tumor cells and neutrophils is allowed only in the premetastatic tissue but not at the primary site. Indeed, while lung-associated neutrophils (LAN) show significant cytotoxicity toward tumors cells, primary tumor-associated neutrophils (TAN) are not cytotoxic (Fig. 1C). In addition, we observed that TAN but not LAN show increased phosphorylation of Smad2 and Smad3, indicating increased TGFβ signaling (Fig. 1D and E). Because TGFβ is a potent inhibitor of neutrophil cytotoxicity, these observations are consistent with the notion that neutrophil cytotoxicity at the tumor microenvironment is blocked by TGFβ (18, 26, 27). Importantly, TGFβ levels are very high at the primary tumor and are virtually absent in the naïve or premetastatic lung tissue (Fig. 1F). Taken together, the difference in neutrophil cytotoxicity in the primary tumor and the premetastatic lung is the consequence of high TGFβ levels and high TGFβ activity at the primary tumor but not in the premetastatic lung.

In a recent study, we demonstrated that neutrophil cytotoxicity is determined not only by neutrophil properties but also by tumor cell features. We found that tumor cell susceptibility to neutrophil cytotoxicity depends on the expression of TRPM2, an H₂O₂-dependent Ca²⁺ channel (20). We therefore hypothesized that the differences in the susceptibility to neutrophil cytotoxicity in the different microenvironments are also determined by changes in tumor cell expression of TRPM2. To address this question, we used a publicly available data set (28) to compare the expression levels of TRPM2 in primary 4T1 tumor cells and in disseminated circulating 4T1 tumor cells (CTC). As expected, CTC showed a more mesenchymal gene expression profile compared with primary tumor cells (Supplementary Fig. S2A). Importantly, 4T1 CTC express higher levels of TRPM2 compared with primary tumor cells, which accompanies their more mesenchymal phenotype (Supplementary Fig. S2A). Using another publicly available data set, generated by Sergeant and colleagues (29), we found higher expression of TRPM2 in CTC compared with primary tumor cells in patients with pancreatic cancer (Supplementary Fig. S2B). In addition, immunofluorescent staining of primary tumors and lung metastasis shows an inverse correlation between E-cadherin expression and TRPM2 expression (Supplementary Fig. S3A–S3B). Together, these observations suggest that CTC may be more susceptible to neutrophil cytotoxicity due to increased expression of TRPM2.

Although the absolute requirement of EMT for metastatic progression is currently a matter of debate (7, 10, 24), an obvious deduction from our experiments is that TRPM2 expression is associated with EMT. Moreover, our observations suggest that TRPM2 expression may be regulated by EMT-driving signals. To test this notion, we compared the morphologic differences and the expression of E-cadherin and vimentin in control (Fig. 2A–C) and TGFβ-treated cells (Fig. 2A, D, and E). TGFβ, a driver of EMT, not only changes the cellular morphology (compare Fig. 2B and D) and the expression of known epithelial and mesenchymal genes (Fig. 2A and compare Fig. 2C and E), but also upregulates TRPM2 expression (Fig. 2A). An increase in the TRPM2 is expected to enhance the tumor cell susceptibility to neutrophil cytotoxicity. However, TGFβ may also have a direct inhibitory effect on neutrophil function (18, 27) through inhibition of H₂O₂ secretion (30). Indeed, the presence of TGFβ in the coculture inhibits neutrophil cytotoxicity (Fig. 2F). Still, pretreatment of tumor cells with TGFβ, which is removed prior to the addition of neutrophils, results in an increased susceptibility of the tumor cells to neutrophil cytotoxicity (Fig. 2F). Because TGFβ has pleotropic effects and can both increase and reduce the apoptotic propensity of cells through various mechanisms (31, 32), we used additional approaches for inducing EMT, through overexpression of either Twist or Snail (33). As expected, Twist-overexpressing tumor cells (Fig. 2G) showed a more mesenchymal morphology, a reduction in E-cadherin expression and an increase in vimentin expression (Fig. 2H, compare Fig. 2B–I and compare 2C–J). Importantly, the mesenchymal-like cells overexpressing Twist showed an increase in TRPM2 expression (Fig. 2H) and enhanced susceptibility to neutrophil cytotoxicity (Fig. 2K). Similar observations were made using Snail overexpression to induce EMT (Supplementary Fig. S4A–S4D). We then questioned whether these observations may also be relevant to the disease in humans. We therefore used the same methodology to test the correlation between induction of EMT, TRPM2 expression, and the susceptibility to neutrophil cytotoxicity in HMLE human mammary tumor cells. Indeed, increased expression of TRPM2 and the ensuing enhanced susceptibility to neutrophil cytotoxicity was observed when EMT was induced by TGFβ (Supplementary Fig. S4E–S4G) or via Twist overexpression in the HMLE cells (Supplementary Fig. S4H–S4K), supporting the relevance to the human disease. Taken together, our observations suggest that TRPM2 expression is enhanced during EMT and that this renders tumor cells more susceptible to neutrophil cytotoxicity.

We then tested whether inducing a phenotypic transition toward a more epithelial state has the opposite effect on TRPM2 expression and the susceptibility to neutrophil cytotoxicity. To this end, we treated 4T1 cells with BMP7, a driver of MET (34). BMP7-treated cells maintained a highly epithelial morphology (Fig. 3A) concomitant with a dramatic reduction in vimentin expression (compare Figs. 2B with 3B, and Fig. 3C). E-Cadherin expression levels, which were already high in untreated cells, showed no significant change in BMP7-treated cells (compare Figs. 2B with 3B, and Fig. 3C). Most importantly, BMP7 decreased the expression of TRPM2 (Fig. 3C) and, as a consequence, decreased the susceptibility to neutrophil cytotoxicity (Fig. 3D). Similar results were obtained using Klf4—another potent inducer of MET (35). 4T1 cells overexpressing Klf4 (Klf4oe, Fig. 3E) exhibit a more epithelial morphology (Fig. 3F), lower expression of vimentin, and higher expression of E-Cadherin (Fig. 3G and H). In line with previous results, Klf4-induced MET was accompanied by downregulation of TRPM2 expression (Fig. 3H) and reduced susceptibility to H₂O₂ (Fig. 3I) and neutrophil cytotoxicity (Fig. 3J).

Together with the correlation made between EMT and increased TRPM2 expression, these observations provide a wider scope of relevance and show that TRPM2 expression and neutrophil cytotoxicity are inversely modulated by signals driving EMT and MET. However, in the metastatic cascade, disseminated tumor cells leave a TGFβ-rich microenvironment (the primary tumor, Fig. 1F) and arrive at a TGFβ poor microenvironment (the premetastatic site, Fig. 1F). Accordingly, we propose that the driving force for MET may be the withdrawal of TGFβ rather than the stimulation with MET-driving factors. This notion may be exemplified in vitro as withdrawal of TGFβ, from cells that have undergone TGFβ-driven EMT, reduces TRPM2 expression to basal levels (Fig. 3K and L). Notably, TRPM2 levels decrease within 48 hours of TGFβ withdrawal (Fig. 3L), even before vimentin/E-cadherin expression levels go back to basal levels (Fig. 3M and N), suggesting that TRPM2 expression might be regulated by EMT/MET-driving signals but is not necessarily associated with the epithelial or mesenchymal phenotype.

The experiments described above demonstrate that tumor cells undergoing EMT are more susceptible to neutrophil cytotoxicity and that tumor cells undergoing MET become less susceptible to neutrophil cytotoxicity. However, these experiments were done in vitro and may not faithfully recapitulate the consequences of EMT and MET on neutrophil-mediated elimination of tumor cells at the premetastatic site. To conclusively determine the role EMT/MET play in neutrophil-mediated inhibition of metastatic seeding in the lungs, we tested the consequences of neutrophil depletion on the seeding of control cells, cells that have undergone EMT, and cells that have undergone MET. To faithfully mimic the conditions CTC confront upon arriving at the premetastatic lung, we used mice bearing GFP⁻ primary tumors, which prime the premetastatic lung and mobilize neutrophils. We then tested metastatic seeding of GFP⁺ tumor cells injected via the tail vein to control or neutrophil-depleted tumor-bearing mice (Fig. 4A). Neutrophil depletion was confirmed by FACS analysis (Fig. 4B). Using this experimental platform, we were able to demonstrate that TRPM2kd cells are not susceptible to neutrophil cytotoxicity in the premetastatic lung and are therefore more efficient metastatic seeders than control cells (20). Imaging of GFP⁺ cells in the whole lungs showed that, as expected, neutrophil depletion dramatically enhanced the seeding of control tumor cells in the lungs (Fig. 4C and D), confirming the protective antimetastatic role neutrophils play in this context. We then tested how signals driving EMT or MET, which modulate TRPM2 expression (Figs. 1, 2 and Supplementary Fig. S4A–S4K), affect the antimetastatic protective function of neutrophils in the premetastatic lung. We found that TGFβ-treated cells, which have undergone EMT and have upregulated TRPM2 expression (Fig. 1C), seed as efficiently as control cells in control animals but seed significantly more than control cells in neutrophil-depleted animals (Fig. 4C–E). In contrast, BMP7-treated cells, which have undergone MET and have downregulated TRPM2 expression (Fig. 3A–C), do not benefit from neutrophil depletion, suggesting they are neutrophil resistant (Fig. 4C). However, BMP7-treated cells are poor metastatic seeders compared with control cells (Fig. 4C and D and Supplementary Fig. S5), making it difficult to draw a firm conclusion regarding the effect of MET on metastatic seeding from this experiment. We therefore repeated the seeding experiment using Klf4oe cells (Fig. 3F–H) that seed in the lungs with similar efficiency to control cells (Fig. 4F–H). Corroborating the observations made with BMP7-treated cells, Klf4oe cells, which have undergone MET, do not seed more efficiently when neutrophils are depleted (Fig. 4F–H). Taken together, these observations demonstrate that while EMT is required for efficient metastatic seeding, it concomitantly renders cells undergoing EMT more susceptible to neutrophil cytotoxicity. However, the observation that cells undergoing MET are protected from neutrophil cytotoxicity demonstrates that MET at the future site of metastasis is beneficial and provides a novel mode of immune evasion.

In a previous study (18), we have demonstrated that neutrophil cytotoxicity requires direct physical interaction with tumor cells and is mediated by hydrogen peroxide (H₂O₂). Further characterization of the molecular mechanisms by which neutrophils kill cancer cells revealed that neutrophil-secreted H₂O₂ results in a transient influx of Ca²⁺ via the H₂O₂-sensitive Ca²⁺-permeable TRPM2 channel (20). This influx of Ca²⁺ via TRPM2 ultimately results in tumor cell apoptosis (36–38). Furthermore, our findings revealed that reduced TRPM2 levels (via knockout or knockdown) in tumor cells, prevents both neutrophil and H₂O₂-induced cell death. More importantly, our data show that reduced susceptibility to neutrophil cytotoxicity in cells expressing reduced levels of TRPM2 is reflected by a higher metastatic potential. This might be explained by the important antimetastatic role neutrophils play in the premetastatic lung.

The role neutrophils play in the context of cancer is controversial. It has been established that neutrophils are not a homogeneous population and they feature both protumor and antitumor properties (30). Several studies, including our own, identified a critical role for cytotoxic neutrophils in the premetastatic site. However, neutrophil antitumor activity is rarely evident in the primary tumor. This could be explained by immunosuppressive factors abundant in the primary tumor microenvironment that can inhibit neutrophils activity and are absent in the premetastatic site (30, 39). One such factor is TGFβ, an immunosuppressive molecule that is highly abundant in the primary tumor microenvironment (40). In contrast, TGFβ levels and TGFβ activity are very low at the premetastatic site (Fig. 1D–F; ref. 18). This generates two very distinct microenvironments: the immunosuppressive primary site where neutrophil cytotoxicity is blocked and the immunopermissive premetastatic site where neutrophil antimetastatic properties are manifested.

We have previously demonstrated that TGFβ blocks neutrophil cytotoxicity by modifying neutrophil function (18, 27, 30). Here, we show that TGFβ affects the interaction between neutrophils and tumor cells by modifying tumor cells as well. Our data show that TGFβ upregulates TRPM2 expression and, consequently, increases the susceptibility to neutrophil cytotoxicity, in both murine and human breast cancer cells. This demonstrates that TGFβ plays a dual role—on the one hand, increasing tumor cell susceptibility to neutrophil cytotoxicity and, on the other, blocking neutrophil cytotoxicity (18, 27, 30). Importantly, even though tumor cells exposed to TGFβ are at greater risk of being eliminated by neutrophils, the inhibitory effect TGFβ exerts on neutrophil cytotoxicity predominates and tumor cells at the primary site remain protected. Notably, upregulation of TRPM2 in tumor cells is not limited to TGFβ but was also induced by other EMT-inducing transcription factors such as TWIST and Snail. As expected, upregulation of TRPM2 by these EMT transcription factors also enhances tumor cell susceptibility to neutrophil/H₂O₂ cytotoxicity. Indeed, we demonstrate that CTC, which have undergone EMT, express higher levels of TRPM2 compared with cells in the primary tumor, in both tumor-bearing mice (Supplementary Fig. S2A) and human patients with cancer (Supplementary Fig. S2B). EMT was previously shown to be affected by Ca²⁺ homeostasis in a mechanism that most likely involves more than a single channel (41, 42).

Although metastatic dissemination often requires a transition to a mesenchymal state (22–24), successful colonization in a distant organ also requires reversal of this process, i.e., a transition back to an epithelial state (5, 43, 44). It has been hypothesized that cells transition back to an epithelial state through MET to form macrometastases (45). MET can be driven by BMP7, Klf4, or by TGFβ withdrawal. We show that MET, driven by these strategies, is accompanied by a downregulation of TRPM2 with a concomitant reduction in the susceptibility to neutrophil/H₂O₂ cytotoxicity. Accordingly, our data suggest that while EMT is required for metastatic progression, MET at the distant site may be regarded as a novel mechanism for enhancing metastatic seeding by promoting immune evasion. Understanding how signals driving EMT and MET modulate tumor cell susceptibility to neutrophil cytotoxicity may have profound implications on our understanding of the metastatic process. The relevance of this notion is further supported by the understanding that CTC may display a spectrum of EMT phenotypes (46) and may undergo partial EMT (47, 48).

Together with our recent study, showing that neutrophil cytotoxicity is mediated via TRPM2 (20), we now propose a mechanism where EMT/MET-driving signals available at the different cancer-related microenvironments upregulate or downregulate TRPM2 expression and consequently the susceptibility to neutrophil cytotoxicity. We show that primary tumor cells undergoing EMT, a process shown to facilitate metastatic progression, upregulate TRPM2 expression (Fig. 5). The increased expression of TRPM2 makes cells undergoing EMT more susceptible to neutrophil cytotoxicity, which might contribute to the low efficiency of metastatic spread. The tumor microenvironment is rich in TGFβ, an established inducer of EMT, which has an inhibitory effect on neutrophil cytotoxicity (18, 30, 49). Therefore, although TGFβ levels at the primary tumor site are sufficient for EMT and for upregulation of TRPM2, tumor cells expressing higher levels of TRPM2 are protected from neutrophil cytotoxicity (Fig. 5). This is not the case at a distant site, where disseminated mesenchymal tumor cells expressing higher levels of TRPM2 are no longer protected by TGFβ and are targeted by neutrophils (Fig. 5). The low TGFβ activity at the distant site (18), as well as other factors such as versican (50), allows the reversal to an epithelial state (MET). MET is accompanied by a reduction in TRPM2 expression, which is translated to reduced susceptibility to neutrophil cytotoxicity. Together, these observations demonstrate how microenvironmental cues that are associated with cancer progression affect not only tumor cells per se but also their interaction with the immune system. Here, tumor cells undergoing EMT acquire features that make them more motile and enhance their metastatic potential. However, this transition comes with a “price”—the upregulation of TRPM2 and the ensuing increased susceptibility to neutrophil cytotoxicity.

Z.G. Fridlender is a consultant/advisory board member for Atrin Pharmaceuticals. No potential conflicts of interest were disclosed by the other authors.

Conception and design: M. Gershkovitz, R.V. Sionov, R.I. Aqeilan, Z.G. Fridlender, Z. Granot

Development of methodology: M. Gershkovitz, S. Khawaled, M.E. Shaul, Z. Granot

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M. Gershkovitz, T. Fainsod-Levi, S. Khawaled, M.E. Shaul, R.V. Sionov, L. Cohen-Daniel, R.I. Aqeilan

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M. Gershkovitz, R.V. Sionov, Y.D. Shaul, Z.G. Fridlender, Z. Granot

Writing, review, and/or revision of the manuscript: Z.G. Fridlender, Z. Granot

Study supervision: Z.G. Fridlender, Z. Granot

Z. Granot was supported by grants from the I-CORE Gene Regulation in Complex Human Disease, Center no. 41/11, the Israel Science Foundation, the Israel Cancer Research Foundation (RCDA), The Rosetrees Trust, the Lejwa Foundation and the Israel Cancer Association.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.