Peripheral blood mononuclear cells (PBMCs) were isolated from buffy coats obtained from healthy volunteers (Linköping University Hospital blood bank, Linköping, Sweden) as described elsewhere (Raffetseder et al., 2014). Healthy volunteers gave their written informed consent for the use of their blood for scientific purposes. PBMC isolation was performed using LymphoPrep (Axis-Shield), involving separation by a density gradient followed by multiple centrifugation and wash steps. The mononuclear cells were seeded in culture flasks in Dulbecco's Modified Eagle Medium (DMEM, Gibco), supplemented with 25 mM HEPES (Gibco), 100 U/ml penicillin (Gibco) and 100 μg/ml streptomycin (Gibco), and allowed to adhere for 2 h before the non-adherent lymphocytes were washed away. The adherent monocytes were allowed to differentiate into hMDMs at 37°C for 5–7 days in DMEM, containing 25 mM HEPES, 100 U/ml penicillin, 100 μg/ml streptomycin, 2 mM L-Glutamine (Gibco) and 10% active human serum pooled from 5 donors (Linköping University Hospital). The resulting macrophages have purity exceeding 99% (Raffetseder et al., 2014). One day prior to the experiment, cells were trypsinized and re-seeded in antibiotic-free medium in a 24-well plate with 0.13 mm glass coverslips (2.5 × 10⁵ cells/well).

The virulent M. tuberculosis strain H37Rv (American Type Culture Collection) harboring the pFPV2-plasmid encoding green fluorescent protein (GFP) or pCherry3 plasmid carrying m-Cherry gene or pSMT1 plasmid for luciferase expression, an ESAT-6 deletion mutant strain (H37Rv-ΔESAT-6-GFP) and Mycobacterium marinum also harboring the pFPV2-GFP plasmid were used in this study (Valdivia et al., 1996; Snewin et al., 1999; Guinn et al., 2004; Carroll et al., 2010). Five additional M. tuberculosis strains, all belonging to lineage 1 (Nebenzahl-Guimaraes et al., 2016), were derived from the reference database of clinical isolates of the National Institute for Public Health and the Environment (RIVM) in Bilthoven, the Netherlands. These strains were selected from a study by Nebenzahl-Guimaraes et al. (2016), in which the strains were epidemiologically proven to be at the extremes of the transmission spectrum. The clinical strains 201 030 and 400 731 were both not part of a cluster, while 500 395, 500 887, and 800 343 all showed a high level of clustering, meaning that these strains were able to cause multiple subsequent TB cases. These five clinical isolates were transformed with the pFPV2-GFP plasmid.

All strains were grown in Middlebrook 7H9 broth supplemented with albumin-dextrose-catalase (ADC, Becton-Dickinson), and 0.05% Tween-80 for 2–3 weeks at 37°C. For standing cultures, the bacteria were re-seeded in fresh broth in the presence of 0.05% Tween-80 (Tween_high) for an additional 7 days to reach log phase. The shaken cultures were passaged 3 days prior to the experiment in medium without Tween-80, leading to the final concentration of 0.01% (Tween_low), and were put on a shaker at 260 rpm. Kanamycin at 20 μg/ml and hygromycin at 100 μg/ml were used as selection antibiotics where required. Prior to this study, standardization experiments were performed to test culture conditions, and determine the optimal duration of culturing (3 and 7 days) and shaking speed (260 rpm; data not shown).

On the day of infection, the bacteria were collected and prepared as previously described (Eklund et al., 2010). In brief, both bacterial suspensions of shaken and standing cultures were centrifuged at 5,000 × g for 5 min in phosphate-buffered saline (PBS) supplemented with 0.05% Tween-80 and passaged through a sterile syringe equipped with a 27-gauge needle to remove aggregates. This procedure was repeated once and the bacteria were resuspended in DMEM supplemented with 25 mM HEPES. The numbers of bacteria/ml was determined both for cording and non-cording conditions (the OD was validated against CFU for both situations, Figure S1) by using optical density at 600 nm (OD600) as a function of CFU/ml. The bacteria were added to the hMDMs at a multiplicity of infection (MOI) of 10. The macrophages were incubated with mycobacteria for 1 h at 37°C before medium was changed to DMEM supplemented with 25 mM HEPES and 10% active human serum pooled from 5 donors (from blood bank at Linköping University hospital).

To characterize METs, 2.5 × 10⁵ hMDMs were seeded in wells on glass coverslips and infected as described above. After 24 h of incubation, the cells were washed once with PBS and incubated with 180 U of DNase I (Thermofisher) for 30 min. To characterize dsDNA, the cells were stained with Quant-iT Picogreen dsDNA kit (Invitrogen) for 5 min. To determine the dependency of MET-formation on ROS-production, hMDMs were incubated with 10 μM of the NADPH-oxidase inhibitor Diphenylene iodonium (DPI) or only dimethyl sulfoxide (DMSO) as a solvent control, for 30 min before infection. DPI and its solvent DMSO were purchased from Sigma Aldrich (Saint Louis, MO). After all the above procedures, cells were washed twice with PBS before fixation in 4% paraformaldehyde (PFA) at RT for 30 min. After fixation, samples were stained with the nuclear dye DAPI, mounted in mounting medium (Dako) and visualized with a Zeiss LSM 700 upright confocal system.

For immunofluorescence staining, infected hMDMs (2.5 × 10⁵), seeded on glass coverslips, were fixed with 4% PFA, blocked (2% BSA and 10% goat serum in PBS) and incubated with polyclonal rabbit anti-histone H4 Citrulline 3 (diluted 1:50; Millipore) at RT for 1 h. After washing three times with PBS, samples were incubated with secondary Alexa 594-lableled goat anti-rabbit antibody (diluted 1:800; Molecular Probes, Eugene, OR) for additional 30 min at RT, again washed three times with PBS and stained with DAPI before mounting onto the slides.

hMDMs were seeded onto glass coverslips for 24 h before infection as described above. After 24 h of infection the cells were washed twice with 0.15 M cacodylate buffer and fixed with 2% glutaraldehyde for 1 h at RT. Subsequently, the samples were dehydrated with a graded ethanol series (50, 70, 85, and 100%), critical point dried with liquid CO₂ and sputter-coated with 1.5 nm platinum. The samples were examined using a scanning electron microscope (JSM 6320F).

A previously set up luminometry-based method (Eklund et al., 2010) was used to compare the kinetic of mycobacterial growth harvested from either shaken or standing cultures during hMDMs infection. Aliquots of medium supernatant and cell lysate containing luciferase-expressing bacteria were transferred to pure water in white 96-well plates at selected time points. One percent of decanal (Sigma Aldrich) was injected through the instrument needle into each well and flash luminescence was measured using a Glomax Multiplus reader (Promega). Light emission, measured as arbitrary luminescence units (ALU), is linearly proportional to the number of viable bacilli. Both extracellular (supernatant) and intracellular (lysate) number of bacteria were obtained by subtracting the ALU values from uninfected cells. The ALU values from both supernatant and lysate were then standardized for dilutions and summed up to obtain total ALU values. To determine fold change of bacterial load, the median value of each triplicate of all time points was normalized to the initial time point of the same experiment.

For quantification of MET production, specimens of infected hMDMs were coded and analyzed blindly using a confocal microscope. Briefly, the total number of nuclei and the number of nuclei connected to METs were counted in eight individual images per sample, after which the mean MET production for each sample was calculated. Nuclei releasing longitudinal thread-like structures were considered as producing METs. For each sample, at least 700 nuclei were examined in total.

Reactive oxygen species (ROS) production was measured using luminol-enhanced chemiluminescence (5-amino-2,3, dihydro-1,4-phtalazinedione, Sigma-Aldrich) in a plate reader (GloMax Multiplus reader, Promega). ROS oxidizes luminol to an intermediate form, luminophore, which is in an excited state and emits light with an emission peak at 425 nm. The intensity of the chemiluminescence emitted is proportional to the amount of ROS in the sample. Before infection, the hMDMs were washed once with 200 μl 0.5% bovine serum albumin (BSA) HBSS in a 96-well round-bottom plate. The cells were either activated with M. tuberculosis (MOI 10), zymosan (833 μg/ml) as a positive control, or culture medium alone. Subsequently, chemiluminescence was measured at 37°C for 100 min and as a measure of ROS production, the area under the curve (AUC x time) was determined.

Statistical analysis was performed using the unpaired Student's t-test (for two groups) and one - way ANOVA (for more than two groups) with the GraphPad Prism 6 software. Data presented were expressed as median values ± SEM and P-values of 0.05 or less were considered statistically significant.

M. tuberculosis is routinely cultured in a liquid culture medium containing the detergent Tween-80 to avoid aggregation of the bacilli. As the cording phenotype represents a form of organized aggregate, we speculated that reducing the concentration of Tween-80 would promote cord formation. To test this, an H37Rv strain expressing GFP was cultured for 3 days on a shaker at 260 rpm in broth with either 0.01% Tween-80 (Tween_low) or 0.05% Tween-80 (Tween_high). These cultures were compared to standing cultures of M. tuberculosis with Tween_low or Tween_high broth for 7 days. Fluorescence microscopy analysis revealed that the standing cultures, especially in the Tween_high broth contained bacteria without any large aggregates (Figures 1A,C). In contrast, the shaken cultures formed larger, more organized aggregates, especially in the Tween_low broth (Figures 1B,D). Scanning electron microscopy performed on bacteria harvested from the two extreme culture conditions (standing cultures with Tween_high and shaken cultures with Tween_low) revealed the presence of cords only in shaken cultures with Tween_low (Figures 1E,F). Based on this observation, we hereafter define the bacterial phenotypes as “cording” (shaken with Tween_low) or “non-cording” (standing with Tween_high).

Next, we assessed whether cording M. tuberculosis can induce MET formation in hMDMs. We infected hMDMs with the GFP-expressing cording or non-cording phenotypes of H37Rv. Uninfected cells were included as a control. DAPI-staining of DNA revealed that cording bacteria induced the release of extracellular DNA in hMDMs (Figure 2B), while macrophages infected with non-cording bacteria produced relatively few METs (Figure 2A). In uninfected cells, no METs could be observed (Figure 2C). Kinetic analyses at 2, 4, 6, and 24 h showed that cording appeared from 6 h on (not shown). Also, with increasing MOIs of the standing culture no METs were observed (Figure S2). Quantification of the percentage of hMDMs producing METs confirmed the observation that infection with the cording phenotype induced significantly more METs (17.0%) compared to non-cording bacteria (1.0%; Figure 2D). Next, we used scanning electron microscopy to observe the ultrastructure of METs (Figures 3A–D). Human macrophages infected with the non-cording bacteria showed normal morphology even after 24 h of infection (Figure 3A). On the other hand, mycobacterial cords could be seen in the remnants of dead macrophages infected with the cording phenotype (Figure 3B) and the macrophages had released extracellular traps either in the form of threads (Figure 3C) or meshwork (Figure 3D).

In order to verify that the observed structures were indeed METs, we performed additional experiments. First, we stained the METs triggered by cording bacteria using another dye, Picogreen, which also binds DNA (Figure 4A). This approach revealed similar structures as with DAPI (Figures 4B–D). In addition, no METs could be observed in hMDMs treated with DNase I subsequently to infection with cording bacteria (Figure 4B). ETs of nuclear origin, as opposed to ETs of mitochondrial origin, have been described to be decorated with citrullinated histones (Liu et al., 2014) and therefore we used a histone antibody (anti-H4Cit3) targeting histones to determine the origin of the METs. Confocal images showed the presence of histones that co-localized with the DAPI stain (Figure 4D). The formation of NETs by human neutrophils is dependent on the generation of ROS by NADPH oxidase (Fuchs et al., 2007; Braian et al., 2013). To determine whether ROS production is also linked to MET formation, ROS production was measured in hMDMs infected with cording or non-cording M. tuberculosis using luminol-enhanced chemiluminescence. Zymosan, a known inducer of ROS-production, was used as a positive control. We found that both cording and non-cording mycobacteria were very weak inducers of ROS as compared to zymosan (Figure 5A). Also, inhibition of the NADPH oxidase using diphenyleneiodonium (DPI) did not prevent MET formation (Figure 5B), suggesting that MET formation is not dependent on ROS production in hMDMs.

In order to determine the replication of the bacteria during the 24 h infection of hMDMs at MOI10, we measured bacterial growth using a luminescence-based method previously established in our group (Eklund et al., 2010). The result showed a continuous growth of cording bacteria over the period as shown by an increase in bacterial load while the non-cording bacteria did not replicate in hMDMs during this time span (Figure 6). The results were not significantly different.

ESAT-6 is one of the most important virulence factors that M. tuberculosis utilizes during interaction with human macrophages. Its membrane-lysing activity allows M. tuberculosis to escape out of the phagosome, kill the host cell and spread from an infected cell to other cells (Gao et al., 2004; van der Wel et al., 2007). We assessed whether ESAT-6 plays a role in M. tuberculosis-induced MET formation. An ESAT-6 deletion strain (ΔESAT6) and the wild-type H37Rv were grown under the cord-promoting conditions as described above. Although the ΔESAT6 also formed cords (Figure S3), those clusters were not efficient in triggering MET formation in the human macrophages (Figure 7A). Quantification of the MET formation triggered by the WT and ΔESAT6 H37Rv is shown in Figure 7B. This observation was confirmed by using another mycobacterial species M. marinum that can also secrete a ESAT-6 homolog (Smith et al., 2008). Like wild-type M. tuberculosis, the cording phenotype of M. marinum produced METs in hMDMs after 24 h of infection (Figure 7C).

To test the effect of cording on MET formation in other M. tuberculosis strains, hMDMs were infected with five clinical strains in addition to H37Rv. The chosen clinical strains were epidemiologically proven to be at the extremes of the transmission spectrum (Nebenzahl-Guimaraes et al., 2016), and therefore we were interested in testing if they differed in their ability to induce METs. All six tested strains were able to form cords when cultured on a shaker in the presence of 0.01% Tween-80 (data not shown), and all mycobacteria displaying the cording phenotype were able to induce MET formation in hMDMs (Figures 8A–F). Quantification of the percentage of MET producing nuclei 24 h after infection confirmed that cording mycobacteria induce more METs compared to non-cording mycobacteria (Figure 8G). However, the cording phenotype of the different clinical isolates were equally efficient in inducing METs.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.