Venous blood was collected from healthy volunteers (Ethical approval number: 10/H1208/48) and peripheral blood neutrophils were isolated using Percoll density gradients and subsequent red blood lysis as previously described (35). Neutrophil stimulation was induced by exposing cells to 50 nM Phorbol 12-myristate, 13-acetate (PMA; Sigma P8139) for up to 4 h and inhibited by treating cells for 30 min, prior to PMA stimulation, with 10 µM Celastrol (Cayman 70950), which is a well-characterized NET inhibitor (33).

A 20% paraformaldehyde solution (Sigma P6148) was added directly to the neutrophil culture medium providing a final concentration of 4% paraformaldehyde. Subsequently, samples were incubated for 10 min at room temperature. Plates were then centrifuged for 10 min at 1,800 rcf, medium was carefully removed and the resulting fixed samples were washed twice with PBS.

The 56 compounds were diluted in DMSO as per manufacturers’ instructions (detailed in Table S1 in Supplementary Material). Master plates with compounds at 1,000 times concentration were prepared in advance, sealed and stored at −20°C until use. Compounds were combined and diluted in PBS to obtain five concentrations ranging between 0.025 nM and 0.25 mM. 20 µl of each concentration were added to 30 µl of cell media containing 5 × 10³ neutrophils to obtain compound concentrations ranging between 0.01 nM and 100 µM.

Neutrophils were diluted in glucose-supplemented PBS and seeded at 1 × 10⁵ cells/well in white 96-well plates (Costar 3912), pre-treated overnight with 1% bovine serum albumin (BSA; Sigma A4503) in PBS. ROS release was quantified by using an enhanced chemiluminescence assay as previously described in Matthews et al. (35). Neutrophils were incubated with 450 nM Luminol (Sigma A8511) in a luminometer (Berthold Tristar² LB942) for 30 min before stimulation and further incubation for up to 3 h. Readings were obtained at 37°C as relative light units.

Neutrophils were diluted in RPMI (Sigma R7509) and seeded at 1 × 10⁵ cells/well, unless otherwise specified, in clear 96-well plates (Costar 3370), pre-treated overnight with PBS 1% BSA. Fluorometric quantification of NET-DNA was performed according to the protocol previously described by Palmer et al. (5). Neutrophils were pre-incubated at 37°C in 5% CO₂ for 30 min prior to stimulation and further incubation at 37°C in 5% CO₂ for up to 4 h. Post-incubation samples were treated for 10 min with 1 unit/ml micrococcal nuclease (MNase; Worthington 4789), centrifuged for 10 min at 1,800 rcf and 150 µl of supernatant was transferred to a black 96-well plate (Costar 3915). NET-DNA was stained with 1 µM Sytox green^® (Invitrogen S7020) and quantified with a fluorometer (Berthold Twinkle² LB970; ex:485 nm/em:525 nm). Readings were obtained at 37°C as arbitrary fluorescence units.

Neutrophils were diluted in RPMI and seeded at 1 × 10⁵ cells/well, unless otherwise specified, in black, clear-bottom 96-well plates (Greiner Bio-one 655096), and pre-treated overnight with PBS 1% BSA. Neutrophils were pre-incubated at 37°C in 5% CO₂ for 30 min prior to stimulation and further incubated at 37°C in 5% CO₂ for up to 4 h. Post-incubation, neutrophils were fixed with 4% PFA and stained with 1 µM Sytox green^® or 20 µg/ml Hoechst 33342 (Sigma 14533).

Neutrophils were diluted in RPMI and 5 × 10³ cells were seeded in black, clear-bottom 96-well plates (Greiner Bio-one 781096). The 56-compound library, at the range of concentrations specified in Table S1 in Supplementary Material, was applied to the seeded neutrophils followed by 30 min incubation at 37°C in 5% CO₂. Subsequently, where specified, NETs were induced with PMA as described above and following 3 h incubation, cells were fixed with 4% Formaldehyde and stained with 20 µg/ml Hoechst 33342.

Samples were loaded onto the Cellomics™ ArrayScan™ (Thermo-Fisher Scientific VTI) and high-resolution images of 30 fields/well were acquired and captured using an automated excitation filter wheel and automated Observer Z1 inverted microscope. Operations were driven by the HCS Studio 2.0 Client Software (ThermoFisher Scientific VTI) using a range of feature settings for autofocus, auto exposure, and cell sub-population.

Automated image analysis was performed using the ArrayScan HCS Reader software (Cellomics ThermoFisher) employing algorithms for Compartmental, Tube formation, NET detection or Nuclear decondensation analysis. Parameters for cell identification and exclusions were optimized for NETs or nuclei analysis (Imagen Therapeutics Ltd., UK).

Statistical analysis was performed using Prism versions 6 and 7 (GraphPad). Each experiment was performed in triplicate and on at least three occasions using cells from different healthy volunteers. Statistical tests were two-sided, with significance defined at α = 0.05.

Investigating biological samples using HCA requires cell imaging for accurate and rapid examination which, in turn, requires fixation and appropriate staining to preserve cellular features (26). Previously reported neutrophil fixation protocols involve numerous wash steps prior to image acquisition and analysis (36). Notably, the mechanical action applied during these wash steps can result in the loss of sample material and, in particular, the disruption of NET structures. Subsequently, we have established an appropriate protocol for NET fixation using PFA in microplate formats in order to prevent material loss and artifact generation. Images of NETs before and after modifications of the fixation protocol were observed to identify their effect on sample integrity. Assay variations included applying in different orders of the fixation protocol the addition of PFA, wash steps, and centrifugation procedures. Following a qualitative assessment of the outcomes of the various fixation protocols (data not shown), we identified the most appropriate protocol as consisting of careful addition of PFA directly into stimulated NET cultures followed by centrifugation and washing with PBS as described in Section “Materials and Methods.” Representative images of the outcome pre- and post-fixation are provided in Figure 1. Notably the fixation agent directly interacts with the stimulated NETs potentially protecting them from further downstream mechanical disruption that may result from subsequent steps in the procedure. Differences between pre- and post-fixation are not easily identifiable indicating the fixation protocol successfully preserves sample integrity.

High-content analysis parameters were explored and optimized in order to develop an innovative high-throughput automated approach for NET analysis that utilized quantitative imaging for monitoring neutrophils, their nuclei and NET-related responses. Image analysis was performed using a bioinformatics software with built-in algorithms suitable for a diverse range of biological applications. Initially, two pre-existing algorithms, namely “Compartmental Analysis” and “Tube Formation,” were assessed for neutrophil and NET identification and analysis. NETs were induced by PMA stimulation and fixed with PFA. Neutrophil nuclei and NET structures were stained with Sytox green^®. Images were acquired using the ArrayScan™ and analyzed using the HCS Reader software. NETs could potentially be recognized using the Compartmental Analysis algorithm, which quantifies the DNA in the circular space surrounding the nucleus (Figure 2A), or by the Tube Formation algorithm which quantifies DNA filaments extruded into the extracellular space (Figure 2B). Although NET quantification was effectively performed, we identified several limitations with these approaches including the lack of detection of early time-points of NET formation (data not shown). Therefore, two bespoke algorithms were developed for NET analysis and were termed “Nuclear Decondensation” and “NET Detection.” These algorithms were developed for the simultaneous monitoring of two fundamental characteristics of NET release, respectively, including nuclear decondensation (Figure 2C) and the subsequent release of DNA into the extracellular space (Figure 2D).

The NET Detection algorithm was designed to identify filaments of extracellular DNA and to provide information on area, size, and intensity of those selected sections. Subsequently, we applied the NET Detection algorithm for the quantification of PMA-induced NET release over a 3-h time-course. Results indicated the technique detected NET units as early as 60 min after exposure to the stimulus, this detection then increased and peaked at 3 h. Interestingly, a relatively low amount of NETs were also identified immediately after PMA treatment. These data suggest HCA might be able to identify very early stages of NET formation as nuclear changes have been observed as rapidly as a few seconds after stimulation [data not shown and in Ref. (3, 4, 24)]. The algorithm also identified relatively low amounts of NETs in unstimulated neutrophils and this marginally increased by 90 min (Figure 3A). The accuracy of the HCA approach and data were further validated by comparison with a standard quantification method. NET analysis was performed by HCA and compared with standard fluorometric quantification, which determined levels of extracellular NET-DNA post-MNase digestion. Data obtained using the fluorometric quantification identified a similar trend in PMA-stimulated NET production and confirmed the validity of HCA data (Figure 3B).

The sensitivity of the NET Detection algorithm was further evaluated using a range of neutrophil densities of between 1,000 and 100,000 cells/well which were subsequently stimulated or unstimulated with PMA for 3 h. The NET Detection algorithm outputs were compared, again, with data obtained by the standard fluorometric assay for validation purposes. The NET Detection algorithm registered significantly higher levels of NETs in PMA-stimulated compared with unstimulated neutrophils seeded at 50,000 cells/well and higher (Figure 4A). A similar trend of NET production was confirmed following fluorometric quantification, which also showed significant increases in PMA-induced NETs compared with controls in neutrophils seeded at 50,000 cells/well and above. This increase was also detected with as few as 5,000 neutrophils/well (Figure 4B). These validated results indicate that the HCA algorithm allowed sensitive and accurate NET quantification in samples with seeding densities above 50,000 neutrophils/well in a microplate format.

Alongside the algorithm for NET structure analysis, a further algorithm was developed to acquire information on cells in the early stages of NET formation, based on monitoring changes in nuclear morphology and chromatin decondensation. Monitoring nuclear area and staining intensity provides information on the state of chromatin decondensation and, therefore, can reportedly be used as a quantifiable indicator of the early stages of NET formation (24). The Nuclear Decondensation algorithm provided information on Hoechst-stained neutrophil nuclei and identified as “NET forming” those cells with relatively large areas and low signal intensity compared with unstimulated controls. Neutrophils were stimulated with PMA or were unstimulated (controls) for 3 h (Figures 5A,B). Results shown in Figure 5D indicated the Nuclear Decondensation algorithm in unstimulated neutrophil nuclei calculated an average nuclear area of between 500 and 700 pixels throughout the 4-h incubation period. A similar nuclear area was calculated in PMA-stimulated neutrophils for up to 3 h and increased to above 1,000 pixels being detected after 4 h. Opposing trends were seen for the analysis of nuclear staining intensity. PMA-stimulated neutrophil nuclear intensity was constantly found to be 200 units below the measured intensity in unstimulated neutrophils during the initial 3 h of exposure to PMA and decreased to 500 units at 4 h (Figure 5E). Nuclei which presented simultaneously enlarged areas and a diminished nuclear intensity were subsequently considered by the algorithm as NETotic events. Data presented in Figure 5F indicate the percentage of NET forming events calculated in unstimulated and stimulated neutrophils related to the previously described area and intensity data. The percentage of NET forming events remained at approximately 0% in unstimulated neutrophils, while in PMA-stimulated neutrophils approximately 60% of cells were determined to be in a NETosing state at 4 h exposure to PMA. To further validate and confirm the data obtained with the Nuclear Decondensation algorithm, we analyzed the set of images using the NET Detection algorithm. Interestingly, the trend for NET forming cells calculated with the Nuclear Decondensation algorithm was similar to the trend of NET release calculated using the NET Detection algorithm (Figure 5C): after 4 h exposure to PMA stimulated neutrophils presented higher levels of NET release compared with unstimulated neutrophils. Taken together, these data indicate that monitoring nuclear decondensation provides a valid indicator of NET forming activity and directly relates to NET formation.

The HCA method described above was applied for HCS of a pharmaceutically derived compound library to identify modulators of NETosis. Celastrol-treated cells were used as inhibitory controls while Benzethonium Chloride was used to induce cytotoxicity and to discern NETosis from other forms of neutrophil cell death. After 30 min of compound pre-treatment, neutrophils were stimulated with PMA or left unstimulated for 3 h. Following fixation, Hoechst was applied for nuclear staining and images were acquired using the ArrayScan™. HCA was performed using the Nuclear Detection algorithm and image analysis performed using two distinct methods for unstimulated and PMA-stimulated images. To identify whether compounds were able to induce NETosis in unstimulated conditions, rejection and threshold criteria for nuclear size and nuclear staining intensity were defined based on images of unstimulated and untreated neutrophils. The algorithm subsequently identified cells with increased nuclear size and with decreased nuclear intensity. Concomitantly increased size and decreased staining intensity was calculated by the algorithm and used as an indication of NETosis. Conversely, in order to identify whether compounds inhibited PMA-induced stimulation, rejection and threshold criteria were established based on images of PMA stimulated neutrophils with untreated neutrophils used as controls. In this case, the algorithm determined the proportion of cells with increased or decreased levels of both nuclear size and staining intensity compared with control cells. HCS identified 17 compounds potentially inducing NETosis in unstimulated neutrophils, while 13 compounds were potentially identified as inhibiting NETosis and 13 compounds potentially enhanced PMA-induced NETosis (Figure S1 in Supplementary Material). Rapamycin and Lapatinib treatments resulted in inhibition of NETosis at relatively low doses and increased NETosis at higher doses, while Bosutinib treatment resulted in an opposing effect. The compounds exhibiting modulatory activity in unstimulated and PMA-stimulated neutrophils were selected for further screening.

Images associated with the compounds identified by the initial HCS were further analyzed using the HCS Reader Software. The nuclei in each image were analyzed for false-negative or false-positive events and for apoptosis-related phenotypes as previously identified in neutrophils by Fuchs et al. (4). Although apoptotic bodies were visible as black spots, the nuclei still presented similar features to NETosing activity, which was misinterpreted by the algorithm and, therefore, these were manually excluded from the selection process (data not shown).

Following extensive literature searching, those compounds that were associated with neutropenia or other forms of neutrophil cell death were also excluded from further study. Subsequently, of the 56 compounds initially screened for modulation of NET formation, 8 were selected for further downstream characterization (listed in Table 1) as NET-inducers, inhibitors or with a biphasic modulatory activity. Images in Figure 6 show PMA stimulated and compound-treated neutrophils. PMA stimulated nuclei (Figure 6A), utilized for the determination of nuclear size and intensity thresholds, presented a characteristic NETosing morphology. In Figure 6B, Celastrol-treated neutrophils, as expected, exhibited nuclei with characteristically unstimulated features and represented controls for inhibitory activity (33).

The eight selected compounds were characterized for their neutrophil modulatory activity using more targeted approaches. Compounds were simultaneously tested in neutrophils from three donors at a range of concentrations from 1 nM to 10 µM. PMA treatment was used as a positive control for stimulation, while Celastrol and media alone were used as a negative control and to provide baseline readouts, respectively. Intracellular and extracellular ROS productions were monitored by chemiluminescence assay. Simultaneously, their ability to induce NETosis was evaluated by fluorometric quantification.

Neutrophil ROS production was evaluated for 3 h following compound treatment. PMA stimulation effectively induced ROS production, while data indicated none of the eight compound treatments induced the activation of the ROS cascade (Figure S2 in Supplementary Material). Celastrol treatment resulted in statistically significant lower levels of ROS compared with media control, indicating the compound-driven inhibition may also suppress the normal metabolic activity of unstimulated cells. Similar statistically significant differences were detected with Bosutinib and Ponatib treatments at relatively high doses (respectively, at 10 µM and at 1 and 10 µM), indicating the two compounds exert inhibitory effects similar to that of Celastrol.

Dose–response induction of NET formation in unstimulated neutrophils was evaluated by fluorometric measurement of extracellular NET-DNA release following 3 h treatment with the selected compounds. Among the eight compounds tested, only Crizotinib treatment at 10 µM induced a significantly higher amount of NET-DNA release compared with the unstimulated/untreated control (Figure 7A). Furthermore, Bosutinib and Ponatinib treatments at 10 µM resulted in increased NET-DNA release, although the difference compared with the media control was not statistically significant (Figure 7B). The remaining five compounds exhibited similar NET-DNA levels to that of the untreated control indicating their application to neutrophils did not trigger NET formation. Notably, the data for Crizotinib and Ponatinib treatments were consistent with the NET-inducing activity as identified by HCS.

The eight selected compounds were characterized for their ability to modulate responses in PMA activated neutrophils from three donors. Neutrophils were treated with the selected compounds at a range of concentrations from 1 nM to 10 µM and incubated for 30 min prior to PMA stimulation for up to 3 h. Treatment with PMA only was utilized as the positive control, while Celastrol treatment was used as a control for inhibitory activity.

The dose–response modulation of PMA-induced ROS release was monitored for 3 h and the maximum levels of ROS production following treatment were compared with the maximum levels registered in untreated cells throughout the stimulation period. Data presented in Figure 8 were normalized using the media-only control values in order to minimize the physiological differences between donors. Consistently, Celastrol treatment significantly reduced PMA-induced ROS activity. Results presented in Figures 8A,B show an overall dose–response inhibition of ROS production following compound treatment compared with PMA-stimulated neutrophils, although, in general, these results were not statistically significant. Interestingly, Crizotinib treatment at the highest dose (10 µM) was able to significantly reduce the amount of ROS produced in response to PMA stimulation (Figure 8A). Notably, Carmustine, Bosutinib, and Ponatinib treatments, all at the highest dose (10 µM), also reduced levels of PMA-induced ROS (Carmustine in Figure 8A, Bosutinib and Ponatinib in Figure 8B). These differences were, however, not statistically significant.

The dose–response modulation of PMA-induced NETosis by different compounds was evaluated by fluorometric measurement of extracellular NET-DNA release following 3 h incubation. Data presented in Figure 9 were normalized by media control values in order to minimize physiological differences between donors. Inhibition of NET release was identified following treatment with Lapatinib at 10 nM and 1 µM; Carmustine at 1 nM, 10 nM, and 10 µM; and Erlotinib at 10 nM and 1 µM (Figure 9A). Interestingly, Crizotininb treatment at the highest dose (10 µM) resulted in more than threefold higher DNA release, although this was not statistically significant, compared with the untreated control (Figure 9A). Notably, these NET-modulation findings were consistent with their activity as identified by HCS. Statistical significance was identified for the inhibitory effects presented in Figure 9A. Notably, Ponatinib appeared to increase the PMA-induced NET production at higher doses (10 µM), while Rapamycin inhibited NET release at the lowest dose (1 nM) applied (Figure 9B). These differences were, however, not statistically significant.