Materials

The sources of the materials were as follows: Annexin V-Alexa Fluor 647 (BioLegend, San Diego, CA), DiOC-6, HEPES, bovine serum albumin, Hoechst-33342, poly-L-lysine, paraformaldehyde (PFA), and phosphate-buffered saline (PBS) (Sigma-Aldrich, St Louis, MO); fibrillar collagen type I (Chrono-Log Corporation; Havertown; USA); Aspirin (Bayer, Germany); Natalizumab (anti-α4β7) (Hospira Inc, USA); Cell Tracker Violet and secondary antibodies were conjugated with AlexaFluor 488 or AlexaFluor 568 (Invitrogen). Human von Willebrand factor (VWF) was a kind gift of Prof. Pierre Mangin (INSERM, Etablissement Français du Sang-Grand Est,UMR_S1255, Fédération de Médecine Translationnelle de Strasbourg, Université de Strasbourg, France), Monafram was a kind gift of Prof. Alexey V. Mazurov (NMRC of Cardiology, Moscow, Russia); human fibrinogen and fibronectin were isolated from human blood plasma and purified by Dr. Egor Osidak (IMTECK, Russia); mouse and rabbit anti-human monoclonal antibodies against myeloperoxidase and neutrophil elastase were a kind gift of Prof. Alexey V. Sokolov (Institute of Experimental Medicine, St. Petersburg, Russia).

Patients or Human subjects

SDS diagnosis was made based on the typical clinical picture and in all cases confirmed by detection of bi-allelic mutations in the SBDS gene.

Blood collection and handling

Blood was collected into Sarstedt-Monovette© hirudin (525 ATU/ml blood) vacuum tubes. In our previous paper [18], we have shown that heparin increases the percent of immobile neutrophils (probably due to chemotaxis inhibition [59]), and citrate introduces reproducibility issues due to the need of recalcification, without which no neutrophil adhesion was observed. Therefore, we choose hirudin as our primary anticoagulant. Experiments were performed within 3 h after blood collection. For shear stress experiments, blood was collected from adult healthy volunteers (3 female, 3 male donors, ntotal = 6), 22–27 years old. For experiments with different matrix proteins, blood was collected from adult healthy volunteers 18–45 years old (2 female, 2 male donors), ntotal = 4. For experiments with different inhibitors and with blood washout, blood was collected from adult healthy volunteers (5 female, 2 male donors, ntotal = 7), 19–39 years old.

For the assays involving Shwachman-Diamond syndrome patients, blood was collected from healthy pediatric donors (n = 5, 2 male, 3 female donors 0.25–33 months old), healthy adult donors (n = 5, 3 male, 2 female, 19–29 y.o.) or from patients with Shwachman-Diamond syndrome (Table 1) into Sarstedt-Monovette hirudin (525 ATU/ml blood) tubes. For NET-osis assay in flow chambers, blood was collected either from SDS patients (Table 1) or adult healthy volunteers 18–45 years old into Sarstedt-Monovette hirudin (525 ATU/ml blood). For neutrophil isolation and NET-osis smear preparation, blood from healthy donors and SDS patients was collected into Sarstedt-Monovette EDTA K3E (1.6 mg/ml blood) tubes.

Patient characteristics

Ten unrelated pediatric patients (6 girls and 4 boys) with Shwachman-Bodian-Diamond syndrome (SDS) were included in this study (Table 1). Median age at enrollment was 5.5 years. Nine patients harbored compound heterozygous mutations in the SBDS gene NM_016038.2:c.258 + 2 T > C and NM_016038.2:c.183_184delinsCT. All identified mutations were confirmed by Sanger sequencing. In one patient, isochromosome 7q, i(7)(q10) abnormality was found; all other patients did not have any confirmed cytogenetic abnormalities or signs of MDS/AML. At the time of enrollment, 6 patients received recombinant human granulocyte colony-stimulating factor (rhG-CSF) therapy.

Fluorescent microscopy

Parallel-plate flow chambers were described previously [18, 60]. Channel parameters were: 0.1 × 18 × 2 mm. Glass coverslips were coated with fibrillar collagen type I (0.2 mg/ml) for 1 h 30 min at 37 °C, washed with distilled water and then inserted into the flow chambers. Alternatively, untreated glass coverslips were used for oncological cell experiments. After addition of fluorescent reagents (DiOC6 (1 μM), Hoechst (2 μg/mL), and AnnexinV-Alexa647 (10 μg/mL)), blood was perfused through the parallel-plate chambers with wall shear rates 0–500 s−1 [61]. Thrombus growth and leukocyte crawling were observed in DIC/epifluorescence modes with an upright Nikon Eclipse Ni-U microscope (20x/0.50 Plan Fluor objective) for mathematical model preparation and quantitative assay of experimental neutrophil movement and upright Nikon Eclipse Ni-E (60x/1.49 Apo TIRF objective with oil immersion) for tumor cell experiments.

For blood washout, whole blood was perfused over collagen-coated glass for 10 min and then perfused with either Tyrode’s buffer (137 mM NaCl, 2.7 mM KCl, 12 mM NaHCO3, 0.36 mM NaH2PO4, 1 mM MgCl2, 2 mM CaCl2, 5 mM HEPES (pH 7.5), 0.36% BSA, 1 g/l D-glucose, pH 7.35) with 2 mM calcium or platelet poor plasma (PPP) from the same donor for 5 min at 200 s−1. PPP was obtained by centrifuging whole hirudinated blood for 10 min at 1600 g.

To study SDS neutrophil behavior on normal thrombi matrix, EDTA-anticoagulated LRP from either healthy controls or SDS patients was centrifuged at 100 g for 3 min. The pellet was resuspended in 200 μl Tyrode’s buffer and incubated with 1 μM CellTracker Violet fluorescent label for 30 min at 37 °C. The sample was then centrifuged at 100 g for 3 min, the supernatant discarded, and the pellet resuspended in healthy donor whole blood. For healthy controls, whole blood from another donor was used. DiOC6 was then added. Imaging was performed by fluorescent microscopy as described above.

For VWF, fibrinogen, and fibronectin matrix assays, cover glasses were silanized [62] before the experiments. Afterwards, the flow chamber was assembled as described above and incubated for 60 min at 37 °C with either 100 μg/ml vWF, 100 μg/ml fibronectin, or 100 μg/ml fibrinogen. Further assay was carried out as described above.

NET-osis level determination

For NET-osis observation in a parallel-plate flow chamber, whole blood was perfused over glass coated with 200 μg/ml collagen for 10 min at 100 s−1. Non-adherent cells were washed out with Tyrode’s buffer with the addition of 2 mM CaCl2. Cells were then incubated for 3 h at 37 °C. Before fixation, the chamber was rinsed with modified Tyrode’s buffer without BSA. Sample fixation was carried out using 1% PFA for 30 min. After rinsing and blocking in 3% BSA in PBS buffer, samples were incubated with primary mouse antibodies against myeloperoxidase (MPO) and rabbit antibodies against elastase for 1 h. Incubation with Hoechst 33,342 and Alexa488- and Alexa555-conjugated secondary antibodies against mouse and rabbit immunoglobulins, respectively, was performed for 1 h.

For smear analysis of NET-osis, leukocyte-rich plasma (LRP) was used [24]. LRP was obtained by sedimenting the blood sample at 37 °C for 45 min. LRP was smeared over a glass microslide and dried overnight at room temperature. The samples were then fixed with 1% formaldehyde for 30 min. After fixation, the smears were washed with 1% BSA and PBS. Subsequently, they were blocked with 10% normal goat serum for 30 min. Thereafter, samples were incubated with primary antibodies to neutrophil elastase and myeloperoxidase for 30 min. Afterwards, samples were incubated with secondary antibodies and Hoechst 33,342 for 20 min. The mounting medium was used to prevent burn-in of fluorescently labeled antibodies. Finally, samples were analyzed by confocal microscopy. The level of NET-osis was determined by the ratio between granulocytes released their DNA and the total number of granulocytes in the sample. Granulocytes were considered as nucleated cells containing MPO and NE that was confirmed with Nikon Ti2 fluorescent microscope with confocal AX attachment.

Cell culture

The cell lines MCF7 and SiHa (ATCC, USA) were used for our experiments. The cells were cultured as a 2D culture in monolayer in DMEM high glucose (HiMedia, USA) supplemented with 10% FBS (HiMedia, USA) and 100 U/ml penicillin, 100 μg/ml streptomycin solution, and 2 mM L-Glutamine (Sigma-Aldrich, USA). Cells were cultivated at 37 °C in a humidified 5% CO2-containing atmosphere. For cells localization in flow chamber, a suspension (106 cells/ml) of cells was incubated in flow chambers with untreated glass for 15 min at 37 °C and then washed with 20 w/v BSA in PBS and afterwards with Tyrode’s buffer (137 mM NaCl, 2.7 mM KCl, 12 mM NaHCO3, 0.36 mM NaH2PO4, 1 mM MgCl2, 2 mM CaCl2, 5 mM HEPES (pH 7.5), 0.36% BSA, 1 g/l D-glucose, pH 7.35) solution. For exclusive tumor cell staining, Hoechst fluorescent probe was added into the Tyrode’s buffer but not into the whole blood sample.

Data analysis

The Nikon NIS-Elements software was used for microscope image acquisition; ImageJ (http://imagej.net/ImageJ) was used for image processing. ImageJ manual tracking plugin was used for manual granulocyte tracking.

For automated cell tracking, particle tracking algorithm described in [63] was utilized. The algorithm was based on Python trackpy v.0.4.2 library. First, particle tracking was performed; then, the tracks belonging to leukocytes were selected manually. The platelet thrombi location was determined using ilastik (http://ilastik.org) pixel classifier. Platelet thrombi area was calculated as the percentage of the screen covered by platelet thrombi. Tracking Code listing and program operation examples are available at (https://github.com/juliajessika/Leukocytes2023).

The first step is the segmentation of thrombi and leucocytes from the background to produce binary (black and white) images. To do this, we use a pixel classifier trained within ilastik. Various pixel-level features including smoothed intensity and edge indicators are measured and used to train a random forest classifier with two outcomes: signal and background. Training images should be selected to ensure the full variability within the dataset is captured. Having trained the pixel classifier within ilastik, it is run on the full dataset.

Statistics

All experiments were performed at least in triplicate with whole blood from 3 different adult, 3 different pediatric donors, and 6 SDS patients. Statistical analysis was performed using Python 3.6; all statistical details are provided in the figure legends.

Computational modeling and algorithms

The mathematical model which describes the propagation of a chemoattractant in a flow is a system of differential equations integrated using the finite element method implemented in the COMSOL Multiphysics software (COMSOL Multiphysics® v. 5.4. www.comsol.com. COMSOL AB, Stockholm, Sweden). Briefly, we solved the Navier–Stokes equation under the assumption of laminar flow and then calculated the distribution of chemoattractants by solving reaction–diffusion-advection equation. Equations and modeling details are given in the Additional file 1: S1, Table S1. References: [30, 31, 64, 65].

For further analysis, we developed a stochastic algorithm simulating neutrophil movement within the previously computed chemokine field based on an existing model by Szatmary et al. [4]. In this model, neutrophil movement is segmented into discrete timesteps. At each timestep, every neutrophil assesses chemokine concentrations at its current position within the gradient and then orients itself in alignment with the gradient. Within the model, it is posited that neutrophils move at a uniform velocity in their newly determined directions during each timestep. The decision is driven by the differential receptor occupancy (DFRO). This DFRO represents the variance in the proportion of receptors bound by ligands over the length of the migrating cell. The DFRO is defined by the following equation:

$$\mathrm\;=\;\frac\frac\frac1^2},$$

where c is the chemoattractant concentration, Kd is the dissociation constant for the chemoattractant-receptor interaction, and ?C is the length of the cell [4]. The chemoattractant gradient sampling was performed using Sobel matrices [66]. The image was rotated, so neutrophil movement direction on the previous step would align with x-axis. Sobel kernel size for x-axis was taken equal to average neutrophil length \(}_\) in pixels. Sobel kernel size for y-axis was taken equal to \(}_/2\), to indicate leucocyte ellipsoid shape. A random bias \(_\) was introduced into neutrophil movement assuming that neutrophils’ orientations fall on a von Mises–Fisher distribution [67]. This is used to represent the observation that stronger gradient signals (i.e., higher DFRO) cause cell orientations to be more biased toward the gradient direction. The von Mises–Fisher distribution is given by \(f(_;\kappa )=\frac}(\kappa }(_))}_(\kappa )}\), where \(\kappa =k \star DFRO\), where k is the sensitivity constant, and l0() is the modified Bessel function of zero order. Based on these quantities, we calculate the components of the model neutrophil velocity on the step N by a recurrent formulae:

$$_=\frac1}v_^cos^N}_+\frac}}v_^cos\theta_^,$$

$$_=\frac1}v_^sin^N}_+\frac}}v_^sin\theta_^,$$

where θsum = θgrad + θrandom, θgrad is the direction of the gradient, \(_=L/(\triangle t)\) is neutrophil memory coefficient, \(_\) is the neutrophil average velocity, L is neutrophil persistence length [4], and △t is model time step. We chose \(_\) corresponding to the experimental average velocity of each neutrophil. Parameters such as \(}_\) and L were chosen for each leucocyte independently to fit the experimental trajectory as good as possible. Persistence length L was chosen to be in the same order of magnitude as in Vicker et al. [49] (\(L=22 s\)). Neutrophil size \(}_\) was chosen from the range of 10–20 μm [68]. The described algorithm was implemented in Python 3.6.