Neutrophil extracellular traps drive epithelial–mesenchymal transition of human colon cancer

Neutrophil extracellular traps (NETs) are extracellular structures, composed of nuclear DNA and various proteins released from neutrophils. Evidence is growing that NETs exert manifold functions in infection, immunity and cancer. Recently, NETs have been detected in colorectal cancer (CRC) tissues, but their association with disease progression and putative functional impact on tumourigenesis remained elusive. Using high‐resolution stimulated emission depletion (STED) microscopy, we showed that citrullinated histone H3 (H3cit) is sufficient to specifically detect citrullinated NETs in colon cancer tissues. Among other evidence, this was supported by the close association of H3cit with de‐condensed extracellular DNA, the hallmark of NETs. Extracellular DNA was reliably differentiated from nuclear condensed DNA by staining with an anti‐DNA antibody, providing a novel and valuable tool to detect NETs in formalin‐fixed paraffin‐embedded tissues. Using these markers, the clinical association of NETs was investigated in a cohort of 85 patients with colon cancer. NETs were frequently detected (37/85, 44%) in colon cancer tissue sections and preferentially localised either only in the tumour centre or both in the tumour centre and the invasive front. Of note, citrullinated NETs were significantly associated with high histopathological tumour grades and lymph node metastasis. In vitro, purified NETs induced filopodia formation and cell motility in CRC cell lines. This was associated with increased expression of mesenchymal marker mRNAs (vimentin [VIM], fibronectin [FN1]) and epithelial–mesenchymal transition promoting transcription factors (ZEB1, Slug [SNAI2]), as well as decreased expression of the epithelial markers E‐cadherin (CDH1) and epithelial cell adhesion molecule (EPCAM). These findings indicated that NETs activate an epithelial–mesenchymal transition‐like process in CRC cells and may contribute to the metastatic progression of CRC. © 2021 The Authors. The Journal of Pathology published by John Wiley & Sons Ltd on behalf of The Pathological Society of Great Britain and Ireland.


Introduction
Colorectal cancer (CRC) is the third most common cancer type worldwide, in both men and women, and holds the fifth highest mortality rate among all malignancies [1]. Immune cell infiltration is highly associated with the outcome of the disease [2]. Furthermore, the identification of consensus molecular subtypes (CMS) showed that certain immune cell signatures are associated with mortality of patients with CRC [3]. Specifically, the unfavourable, 'mesenchymal' (CMS4) subtype of CRC was characterised by increased expression of immunosuppressive factors and chemoattractants, as well as the infiltration of innate immune cells [4]. In a novel metastatic CRC mouse model, tumours resembling the CMS4 subtype showed an increased neutrophil accumulation as compared with prognostically favourable CMS1-3 tumours [4].
Neutrophils are the most abundant type of innate immune cells in humans, acting as a first-line defence against pathogens [5]. Neutrophils exert their functions by phagocytosis, degranulation and the formation of neutrophil extracellular traps (NETs) [6]. NETs are extracellular structures, composed of nuclear DNA and various proteins released from neutrophils, most prominently represented by histones and granular enzymes, such as neutrophil elastase (NE) and myeloperoxidase (MPO) [6]. Neutrophils undergo several morphological changes during the formation of NETs. Initially, the nuclei of activated neutrophils lose their characteristic lobules, then chromatin begins to de-condense, while the nuclear membrane remains intact [7]. Part of this process is the citrullination of histones, catalysed by peptidylarginine-deiminase 4 (PAD4). Citrullination changes the charge and initiates the de-condensation of nuclear chromatin [8]. Eventually, both the nuclear and the granular membranes lose their integrity, allowing NET components to mix, which results in the disruption of the plasma membrane and the release of the NETs into the extracellular space [7].
NETs have first been described as a mechanism to trap and kill bacteria [6]. In recent years, however, evidence has accumulated of NETs negatively influencing various, also non-infectious, diseases. Among these are thrombotic events [9][10][11][12], pre-eclampsia [13], sepsis [14] and, most recently, the coagulopathy, organ damage and immunothrombosis that characterise severe cases of COVID-19 [15]. The first evidence for NETs playing a role in cancer was provided by a study on a small cohort of patients (n = 8) with paediatric Ewing sarcoma in 2013, suggesting that NETs might be associated with metastasis and early relapse [16]. Since then, various studies have connected NETs with cancer metastasis and progression. However, only a few focused on intratumour NETs [17][18][19][20].
The assessment of intratumour NET formation remains difficult, as there is no high-fidelity protocol for examining the presence of NETs in tissue samples. Most studies use a combination of different markers to show the presence of NETs. For example, in intact neutrophils, nuclear histones and cytoplasmic granular enzymes reside in separated compartments. NET formation leads to co-localisation of these proteins and the extracellular DNA in the NETs. This suggested that specific detection of NETs can be achieved by analysis of the co-localisation of extracellular DNA and histone H2B with granular proteins, such as NE and MPO [21]. In addition, several studies described a critical role of PAD4-dependent citrullination in NET formation, showing that inhibition of PAD4 reduced chromatin de-condensation and NET formation in vitro [22] and in vivo [23]. On this basis, citrullinated histone H3 (H3cit) is frequently used as a marker for NETs, both in single and combined staining protocols. Recently, multiplex immunofluorescence analyses have been reported, to detect NETs in CRC in small groups of 10 or 20 patients [24,25]. However, the association of NETs with specific clinical features in CRC remained elusive. Here, we used high-resolution stimulated emission depletion (STED) microscopy to identify the most appropriate and easily usable marker for the detection of NETs in human formalin-fixed paraffin-embedded (FFPE) tissues sections from colon cancer tissues and applied this approach to investigate the clinical association of NETs in a cohort of 85 patients with colon cancer. Finally, we determined whether NETs may directly act on the tumour cells inducing a phenotype, which could explain the appearance of clinical features associated with NETs in CRC.

Ethics approval
The study was performed in accordance with the Declaration of Helsinki after approval of all procedures by the local ethics committee of the Friedrich-Alexander University (FAU) Erlangen-Nuremberg, Germany (no. 159_15 B, 26

Patients
Human CRC tissues were retrieved as FFPE blocks after completion of routine diagnostics as a retrospective study cohort from the Institute of Pathology, Friedrich-Alexander University (FAU) Erlangen-Nürnberg. The inclusion criteria were patients with first manifestation of histologically verified colon cancer (stage UICC I-IV) [26]. Exclusion criteria were patients with hereditary tumours (hereditary non-polyposis colorectal cancer (HNPCC), familial adenomatous polyposis (FAP)) or tumours resulting from inflammatory bowel disease (ulcerative colitis, Crohn's disease). The detailed patient characteristics are given in Table 1.
Neutrophils were isolated from the blood of healthy volunteers.
Scoring for H3cit positivity was carried out blind (without knowledge of the clinical parameters) by two persons independently. In cases of disagreement, discussion to consent was made. The tissue sections of all patients were assessed using a fluorescence (DM6000B, Leica) microscope and classified into H3cit-positive and -negative tumours. Positive areas that were separated from the tumour section or only found at the outermost edge of the tissue were not included. Isotype control-stained sections were used to validate the signal specificity.

Neutrophil isolation and induction of NET formation
EDTA-blood from healthy human donors was drawn and neutrophils were isolated using a Ficoll-Diatrizoate density gradient (Lymphoflot, Biorad, Hercules, CA, USA, #824012) and incubated with 500 nM phorbol 12-myristate 13-acetate (PMA, Sigma-Aldrich, #P1585) for 4 h at 37 C to allow NET formation. NETs were collected, pelleted and resuspended in 1 Â PBS according to a protocol published by Najmeh et al [29]. The concentration of this NET stock was quantified using a NanoDrop spectrophotometer (ThermoFisher).

Calculation of PMA dilution in NETs stock
Initially, the neutrophils were stimulated with 500 nM PMA to induce NETs; after 4 h the PMA-containing medium was discarded and NETs were washed off the NETs in colon cancer 457 cell culture dish using 5-6 ml 1 Â PBS. Approximately, 200 μl PMA-containing medium remained on the dish, resulting in a 25-30-fold dilution of PMA. NETs were pelleted, the supernatant was discarded and the NETs were pooled together in 250-700 μl 1 Â PBS, depending on the pellet sizes. Less than 100 μl remained on the pellets, resulting in a further 3.5-9-fold dilution of PMA. For treatment of the cells, the NET stock was diluted to 500 ng/ml in cell culture medium. Depending on the initial NET concentration, the range of dilution at this step was 1:314-1:1,095. In total, this resulted in a 30,000-176,000-fold dilution of the initially added PMA when NETs were applied onto the cells. Accordingly, the maximum concentration of residual PMA was 0.016 nM. For the quantification experiments ( Figures 4C, 5A,B and supplementary material, Figure S1), a NET stock requiring an 82,000-fold dilution with a calculated residual PMA concentration of 0.006 nM was used. PMA even in 17-fold higher concentrations did not activate cell migration (see supplementary material, Figure S1).

Cell culture
DLD1 and SW480 CRC cell lines were purchased from ATCC (Manassas, VA, USA) and authenticated by analysing the DNA profile of 17 highly polymorphic sites of short tandem repeats using multiplex PCR (Leibniz Institute, Deutsche Sammlung von Mikroorganismen und Zellkulturen (DMSZ), Braunschweig, Germany) as described previously [30]. Both cell lines were cultured in RPMI medium (ThermoFisher) + 10% FCS (Merck) + 1% glutamine (ThermoFisher) at 5% CO 2 in the absence of antibiotics. Mycoplasma tests (MycoAlert, Lonza, Basel, Switzerland) were carried out on a regular basis and were negative. For all experiments, cells were seeded in full medium (10% FCS), after firm adhesion cultured in the absence of FCS for 10 h and subsequently stimulated for the indicated time periods.

Time-lapse microscopy
Time-lapse microscopy was conducted in 24-well plates (ThermoFisher) using phase-contrast microscopy (DMI4000 B, Leica) and an automatically controlled stage and incubation system (5% CO 2 , 37 C). Cells were seeded (DLD1: 15,000 cells/well; SW480: 20,000 cells/well) and stimulated with 500 ng/ml NET stock, corresponding volumes of PBS or different concentrations of PMA as controls. Pictures were taken every 10 min for a total duration of 24 h. Videos were processed using ImageJ software.

Wound healing assays
Cell migration was analysed using wound healing assays, following established methodology [31]. DLD1 and SW480 cells were seeded in triplicate into six-well plates (ThermoFisher, SW480: 1.2 Â 10 6 cells/well, DLD1: 1 Â 10 6 cells/well) and stimulated with (1) 500 ng/ml purified NETs, (2) an identical concentration of DNase-digested or DNase-digested and heat-denaturated NETs, (3) a corresponding volume of 1 Â PBS as a control or (4) different concentrations of PMA (Sigma-Aldrich). For DNase digestion, the NET stock was incubated with 2 U/μl DNase I (Roche, Rotkreuz, Switzerland) at 37 C for 30 min and for denaturation of proteins NETs were incubated at 95 C for 20 min. The scratch was formed using a 1 ml pipet tip and images of identical areas were taken at time points 0, 2, 4, 6, 8, 10, 12, 18 and 24 h for DLD1 cells and at time points 0, 2, 4, 6, 8, 10 and 12 h for SW480 cells using phase-contrast microscopy (Axiovert 25, Zeiss, Oberkochen, Germany). Quantification of open areas was carried out using ImageJ software. To confirm DNase I digestion of NETs, aliquots of the reaction were analysed using 0.8% agarose gel electrophoresis and subsequent ethidium bromide/ultraviolet light visualisation (see supplementary material, Figure S2A).

Immunofluorescence, western blotting and RT-PCR analyses
Immunofluorescence, western blotting and RT-PCR analyses were conducted following standard procedures and are detailed in Supplementary materials and methods.

Patient cohort
Statistical differences in clinical parameters were determined using Fisher's exact test.

Gene expression analysis
Differential gene expression was analysed using twosided, unpaired t-tests corrected using the Holm-Sidak method for multiple comparisons.

Wound healing assays
Statistical differences in migration between cells in different conditions were analysed by two-sided, paired t-test.

H3cit and extracellular DNA identify citrullinated NETs in colon cancer
In order to determine how NETs can reliably and most easily be detected in human colon cancer tissues, we stained for NE and H2B on the first and H3cit on the following tissue section, employing immunofluorescence microscopy. Tissue areas showing NE/H2B colocalisation ( Figure 1A, arrows) were not in all cases also positive for H3cit ( Figure 1A; compare white and red frames, arrowheads). This finding suggested that either H3cit may detect a subpopulation of NETs or NE/H2B co-localisation may also occur in intact neutrophils in the absence of NETs. In order to investigate this,

458
AM Stehr et al De-condensed DNA in NETs was strongly stained with the anti-DNA antibody selectively in areas where H3cit was also detected. These areas were also positive for NE and H2B. The coincidence of these four markers specifically identified citrullinated NETs. In contrast, co-staining with NE and H2B also occurred in the nucleocytoplasmic transition zone in intact neutrophils (see also Table 2).
NETs in colon cancer 459

NETs in colon cancer 461
did not exhibit the typical nuclear morphology as detected with DAPI ( Figure 1B, white arrows). In contrast, in specimens lacking H3cit, the H2B staining often co-localised with nuclear DAPI staining, whereas NE displayed a cytoplasmic localisation ( Figure 1B, arrowheads). This constellation indicates that H3cit colocalised with extracellular DNA, decorated with NE and H2B, as characteristically observed for NETs. In contrast, the co-localisation of NE/H2B signals in H3cit-negative areas is prone to be derived from overlapping signals obtained in intact neutrophils at the nucleocytoplasmic transition zone. This was further confirmed by staining with an anti-DNA antibody that led to strong signals in H3cit-positive ( Figure 1B, patients 1 and 2, yellow arrows) but not H3cit-negative areas ( Figure 1B, asterisks). This finding was in agreement Statistical significance was determined using two-sided, paired t-test. *p < 0.05; **p < 0.01. Scale bars correspond to 50 μm.

AM Stehr et al
with the fact that DNA in NETs is less condensed and accordingly better accessible to the large IgM anti-DNA antibody as compared with highly condensed nuclear DNA. These results further supported the notion that H3cit staining is specifically associated with the citrullinated NETs in colon cancer tissues, whereas

NETs in colon cancer 463
overlapping signals for NE and H2B may also be derived from intact neutrophils. Profiling of the NE/H2B signal intensities confirmed this conclusion ( Figure 1C). Optical sections of H3cit-positive tissues showed similar distribution of NE and H2B signals, indicating colocalisation of both molecules, a hallmark of NETs ( Figure 1C, patient 1). In contrast, in specimens lacking H3cit the intensities of NE and H2B followed an opposite course, reaching maxima at different sites with an intersection at the nucleocytoplasmic transition zone ( Figure 1C, patient 3). In total, we concluded that H3cit staining is sufficient to reliably detect citrullinated NETs in colon cancer tissues ( Figure 1D).

NETs are associated with high tumour grades and local metastasis in colon cancer
In order to translationally determine a potential association of NETs with specific clinical features in human colon cancer tissues, we analysed NE, H2B, H3cit and extranuclear DNA in a cohort of n = 85 patients (Table 1) using immunofluorescence microscopy. In agreement with the STED analysis, H3cit-positive sections exhibited high amounts of extranuclear DNA, as indicated by staining with an anti-DNA antibody (Figure 2A, left panels, arrows). In contrast, areas exhibiting merely NE/H2B staining in the absence of H3cit showed only low amounts of decondensed DNA that can be detected by immunofluorescence microscopy ( Figure 2A, right panels, asterisks). Quantitative evaluation of the immunofluorescence analyses showed that NETs were more frequently present in colon tumours with higher grade (grade 3/4), locoregional metastasis (pN1/2) and consequently in UICC stage III ( Figure 2B,C,E). In addition, NETs were more often observed in tumours with increased local invasion ( Figure 2D). Finally, we evaluated whether NETs may appear at specific sites within the lesions. In most of the samples, NETs were detected either in the tumour centre and the invasive front (35.5%) or in the tumour centre only (32.3%) ( Figure 3A, left panel, for tumour areas see Figure 3B). A minority of tumours showed NET formation in the tumour centre and the luminal side, in all three compartments or exclusively in the invasive front or the luminal site ( Figure 3A, left panel). The localisation of NETs was comparable in high-and low-grade tumours ( Figure 3A, right panel).

NETs induce motility and epithelial-mesenchymal transition (EMT) in human CRC cell lines
Higher tumour grade has been shown to be associated with EMT, which describes the formation of a mesenchymal tumour cell phenotype with increased motility [32]. In order to investigate this in CRC cells, NETs were isolated from PMA-stimulated human neutrophils in vitro. CRC cell lines (DLD1, SW480) incubated with NETs changed their morphology, which was most clearly reflected by the formation of long pseudopodia ( Figure 4A, arrows) and alterations of the actin cytoskeleton, detected by phalloidin staining (Figure 4A, right). Moreover, a wound healing assay showed that NETs significantly increased the migration of CRC cells ( Figure 4B,C and supplementary material, Figure S2B). The increased migration of DLD1 and SW480 cells in the presence of NETs was also confirmed by time-lapse video microscopy (see supplementary material, Movie S1). Of note, NET formation was induced by the treatment of neutrophils with 0.5 μM PMA, which may activate cell migration per se. However, during NET isolation, until application onto tumour cells the PMA concentrations were highly diluted down to less than 7 pM. In control experiments, PMA even at 17-fold higher concentrations (100 pM) did not activate CRC cell migration (see Materials and methods and supplementary material, Figure S1). This finding excluded residual PMA in NET preparations being responsible for the activation of migration.
In order to investigate whether the DNA or protein content of NETs increased CRC cell motility, NETs were either digested with DNase I alone or in addition subjected to thermal treatment (95 C, 20 min). Digestion with DNase I alone did not suffice to abrogate the effects of NETs on cancer cell motility. However, upon subsequent heat denaturation of NET proteins, migratory activity of DLD1 and SW480 cells was significantly lower as compared with stimulation with untreated NETs ( Figure 4C and supplementary material, Figure S2B).
Increased motility of epithelial cells is often associated with EMT, characterised by a downregulation of epithelial markers and simultaneous upregulation of mesenchymal markers and EMT promoting transcription factors. Interestingly, in SW480 cells, incubation with NETs increased the expression of genes encoding the mesenchymal markers vimentin (VIM) and fibronectin (FN1) and of the transcription factors ZEB1 (ZEB1) and Slug (SNAI2) and decreased the expression of the epithelial markers E-cadherin (CDH1) and epithelial cell adhesion molecule (EPCAM) ( Figure 5A). In DLD1 cells, a significant increase of FN1 as well as ZEB1 expression could be observed ( Figure 5B). The three most highly expressed genes altered upon treatment with NETs (FN1, ZEB1, VIM) were validated at the protein level by western blotting ( Figure 5C). Moreover, increased expression of vimentin in SW480 cells after stimulation with NETs was confirmed by immunofluorescence ( Figure 5D).

Discussion
We found that NETs are abundantly present in human colon cancer tissues and can specifically be detected by H3cit staining. The presence of NETs was associated with high histopathological tumour grades and lymph node metastases. Treatment of CRC cell lines with NETs induced a migratory phenotype with decreased expression of epithelial markers and increased expression of mesenchymal markers and EMT-promoting transcription factors. Thus, NETs induce an EMT-like phenotype and through this may actively contribute to tissue invasion and lymph node metastasis in colon cancer.
Recently published protocols suggested that NETs can be identified by co-localisation of NE and H2B, as in intact neutrophils both proteins reside in different cellular compartments; namely NE in the cytoplasmic granules and H2B in the nucleus [21]. Other studies used H3cit as a marker of NETs in tissues either solitary or in combination with other markers [18,19,25,33]. However, in solid tumours, arginine deimination (also referred to as citrullination) of histones has also been observed during epigenetic regulation of gene expression [34]. In agreement with this, PAD4 (the enzyme catalysing histone H3 citrullination) was reported to be frequently overexpressed in cancer cells [35]. Accordingly, a recent study used multiplex immunofluorescence combining the detection of H3cit with the granulocyte marker CD15 and the neutrophil marker MPO for the detection of NETs in various solid tumours [25]. However, multiplex analyses are difficult to integrate into routine diagnostic processes. In order to determine which marker may be the most appropriate to detect citrullinated NETs in colon cancer tissues, super resolution STED microscopy was applied here. We found that NE, H2B and H3cit co-localised in colon cancer tissues but NE/H2B co-localisation was clearly more common as compared with H3cit staining. The reason for this was that NE and H2B co-localisation was not specific for NETs and also appeared at the nucleocytoplasmic transition zone in intact neutrophils. This situation was demonstrated by signal intensity profiling. Interestingly, under the conditions used for immunofluorescence (FFPE tissues, target retrieval at pH 6.0, 20 min, 90 C), DNA was strongly stained with an anti-DNA antibody in areas where H3cit was also detected. In these areas, DNA staining presented with a fibrous randomly scattered pattern indicating NETs. In contrast, strongly reduced or no signals were obtained with the anti-DNA antibody in cells with condensed nuclear DNA, as the high molecular weight IgM antibody (900 kDa) has limited penetration of condensed nuclear DNA. In contrast, the reactivity of the low molecular weight compound DAPI (0.28 kDa) preferred condensed nuclear DNA over extracellular, decondensed DNA of NETs. In total, these observations suggested that the anti-DNA antibody reacts with extracellular de-condensed DNA of NETs, which reveals improved accessibility of respective epitopes as compared with highly condensed nuclear DNA. The coincidence of H3cit and extracellular DNA further confirmed that extracellular staining of H3cit is the most sensitive marker for citrullinated NETs in human colon cancer tissues ( Table 2).
As yet, the relationship of NETs in primary tumour tissues with clinical parameters in colon cancer has not been investigated in larger patient cohorts. A putative role of NETs in cancer development has been suggested by murine models showing that surgical stress and increased LPS levels after postoperative infections lead to NET formation and increased occurrence of metastases [33,36]. In addition, high MPO DNA levels in the serum have been associated with increased risk of tumour relapse and shorter overall and disease-free survival [20,33,36]. In murine models of breast and colon cancer it has been shown that NETs in the liver and lungs actively attract cancer cells to form metastases by interaction of the transmembrane protein coiled-coil domain containing protein 25 (CCDC25) with NETs. CCDC25 on cancer cells hereby acted as a receptor for NET DNA [18]. The presence of NETs in the primary tumour tissue of patients with CRC was only shown in small cohorts of 10 and 20 patients [24,25]. To our knowledge, we are the first to systemically investigate the presence of NETs in primary colon cancer lesions on a large cohort of patients. Intra-tumoural citrullinated NETs were associated with high tumour grade, lymph node metastases and a trend towards local tumour progression. Our results support the clinical relevance and association with an unfavourable clinical course of citrullinated NETs in colon cancer, indicating that they contribute to disease progression.
In epithelial-derived solid tumours such as CRC, high tumour grade, lymph node metastasis and advanced clinical stage have been connected to EMT [37]. Studies on the relationship between NETs and EMT indicated that NETs promote a mesenchymal, pro-metastatic phenotype in breast [38], gastric [39] and pancreatic cancer [40] cell lines. In agreement with these previous studies on various tumours, our findings provided clear evidence that NETs exert a direct effect on CRC cells by increasing their migratory capabilities and inducing features of EMT, such as the increased expression of VIM, FN1, ZEB1 and SNAI2, as well as the decreased expression of CDH1 and EPCAM. Previous studies proposed the proteins of NETs as the main contributors to EMT induction [38,41,42]. In accordance with this, we noted that the EMT-associated migration cannot be abrogated by degradation of the NETs' DNA. Instead, the migration-inducing activity of NETs was found to be thermo-labile, supporting the hypothesis that it is triggered by proteins.
In conclusion, the detection of NETs by immunofluorescence using the combination of H3cit and anti-DNA antibodies provides a possibility to determine the prognosis of CRC patients. In addition, our results suggest that NETs are a putative therapeutic target, to decrease the risk of metastasis in patients with CRC.