Characterization of neutrophil-neuronal co-cultures to investigate mechanisms of post-ischemic immune-mediated neurotoxicity

Abstract

Background

Immune-mediated reperfusion injury is a critical component of post-ischemic central nervous system (CNS) damage. In this context, the activation and recruitment of polymorphonuclear neutrophils (PMNs) to the CNS induces neurotoxicity in part through the release of degradative enzymes, cytokines, and reactive oxygen species. However, the extent to which close-range interactions between PMNs and neurons contribute to injury in this context has not been directly investigated.

New Method

We devised a co-culture model to investigate mechanisms of PMN-dependent neurotoxicity. Specifically, we established the effect of PMN dose, co-incident neuronal ischemia, lipopolysaccharide (LPS)-induced PMN priming, and the requirement for cell-cell contact on cumulative neuron damage.

Results and comparison to existing method(s)

Pre-exposure of day in vitro 10 primary cortical neurons to oxygen-glucose deprivation (OGD) enhanced PMN-dependent neuronal death. Likewise, LPS-induced priming of the PMN donor further increased PMN-induced toxicity in vitro compared to saline-injected controls. Compartmentalization of LPS-primed PMNs using net wells confirmed the requirement for close-range cell-cell interactions in the process of PMN-induced neuronal injury. Moreover, time-lapse imaging and quantitative neurite analyses implicate PMN-neurite interactions in this pathological response. These experiments establish a platform to investigate immune and neural factors that contribute to post-ischemic neurodegeneration.

Conclusions

Ischemic and immune priming enhance neurotoxicity in PMN-neuronal co-cultures. Moreover, cell-cell contact and neurite destruction are prominent features in the observed mechanism of post-ischemic neuronal death.

Graphical Abstract

1.0. INTRODUCTION

Polymorphonuclear neutrophils (PMNs) are an essential element of the host defense against invading pathogens and play equally important roles in response to tissue injury and sterile inflammation (Mai et al. 2019; Park et al. 2018). In the setting of cerebral ischemia, PMN chemoattractants can be detected in cerebrospinal fluid within the first twenty-four hours after injury, and their levels predict infarct volume (Bao et al. 2010). Central nervous system (CNS) ischemia also triggers the release of mitochondrial DNA, heat shock proteins, high-mobility group box-1 (HMGB-1), and other damage-associated molecular patterns (DAMPs) into the circulation where they activate PMNs and other immune cells by binding to a broad range of pattern recognition receptors including the family of Toll-like receptors (TLRs) (Zindel and Kubes 2019; Pittman and Kubes 2013). Within hours, PMNs respond to these cues and accumulate within the cerebral microvasculature, where they interact with circulating platelets and the vascular endothelium before migrating from the vascular lumen into the perivascular space and tissue parenchyma (Denorme et al. 2019; Allen et al. 2012).

Relative to other leukocyte subtypes, PMNs are the first peripheral immune cells to respond to CNS ischemia (Jickling et al. 2015). Yet, despite decades of investigation, our understanding regarding the complex roles that PMNs serve in the context of cerebral ischemic injury remains incomplete. Like monocytes and microglia, PMNs are highly efficient at phagocytosing bacteria, platelets, and other targets (Maugeri et al. 2009). PMNs also have ample opportunity to induce damage to axonal and dendritic processes following post-ischemic transmigration, as described in models of ALS (Trias et al. 2018) and in the perilesional cortex following intracerebral hemorrhage (Moxon-Emre and Schlichter 2011). Additionally, HMGB-1 produced by ischemic neurons promotes the extrusion of neutrophil extracellular traps (NETs), which are webs of DNA and histones that aid in microbe clearance but are pathologic in sterile ischemia (Kim et al. 2019). These findings led to the belief that infiltration of the pro-inflammatory “N1” PMN is deleterious, though there is now evidence of an anti-inflammatory “N2” phenotype that promotes phagocytic clearance, decreases NETosis, and decreases overall stroke injury (Bao et al. 2010; Cai, Liu, et al. 2019; Cai, Wang, et al. 2019).

To date, a variety of PMN-targeted strategies have been tested in preclinical stroke models. To block the initial phase of innate immune activation, strategies employing an interleukin-1 receptor antagonist limit peripheral innate immune activation and recruitment to the post-ischemic CNS (Maysami et al. 2016). Alternatively, immune-based strategies targeting VLA-4 and ICAM-1, among other adhesion molecules involved in PMN CNS migration, have also shown benefit in preclinical studies (Neumann et al. 2015; Yang and Hagmann 2003). In the context of alternatively-activated PMNs, it is interesting to note that PMN depletion does not always improve outcomes. Cai and colleagues showed that N2 polarization was TLR4-mediated, and that PMN depletion in TLR4-deficient mice resulted in larger infarct volumes (Garcia-Culebras et al. 2019). Cuartero et al. used the PPAR-γ agonist rosiglitazone to drive N2 polarization and found that greater N2 PMN infiltration was associated with smaller infarct volumes (Cuartero et al. 2013). More focal strategies have focused on small molecule inhibitors targeting metalloprotease or elastase activity, which exhibit benefit in reducing post-ischemic blood-barrier injury and associated neurovascular damage (Ikegame et al. 2010; Yang and Rosenberg 2015). Co-administration of DNAses with thrombolytics, intended to reduce the effects of intravascular NETs on inflammation and blood flow, has also shown promise in preclinical studies (Pena-Martinez et al. 2019).

While PMN-dependent release of toxic factors is alone sufficient to produce neuron death in vitro (Nguyen, O’Barr, and Anderson 2007; Allen et al. 2012), an aspect of PMN-dependent neurotoxicity that remains largely overlooked relates to potential effects related to direct neuronal-PMN interactions. Jones et al. demonstrated that after focal stroke, PMNs could be found in remote sites of secondary neurodegeneration, particularly in the thalamus, thus playing a crucial role in the disruption of neuronal connections (Jones et al. 2018). Prior reports investigating PMN-neuronal interactions have used PMNs co-cultured with either neuroblastoma cell lines or neurons harvested from dorsal root ganglia (Shaw et al. 2008). These studies indicate that treatment with IL-1β, TNF-α, or INF-γ facilitates PMN-cell interactions mediated by ICAM-1, VCAM-1, and LFA-1 (Birdsall 1991; Birdsall et al. 1992). Purported mechanisms of toxicity stemming from these interactions involve PMN activation and oxidative burst activity, induced neuronal hyperexcitability, and production of neuronal ROS (Birdsall 1991; Shaw et al. 2008). To our knowledge, these interactions have not been visualized in real-time, and given PMNs’ many means of cell destruction (phagocytosis, NETosis, ROS production, etc.), receptors may represent only one mechanism of PMN-neuron interaction.

Given our interest in the potential role of PMN-neuronal contact in post-ischemic neuronal injury, we combined the use of a PMN-cortical neuron co-culture model with live-cell microscopy to visualize dynamic in vitro PMN-neuronal interactions in real-time. Analyses also included low-dose donor immune priming using lipopolysaccharide (LPS) and ischemic neuronal stress using OGD to address the additive effects of conditional activation. This model of PMN priming combined with OGD-reperfusion reflects human studies of cardiac arrest-resuscitation, in which endotoxin is detectable in the serum of up to 86% of patients within 24–48 hours of successful resuscitation (L’Her et al. 2005; Grimaldi et al. 2015; Grimaldi et al. 2013; Adrie et al. 2002). Our results reveal that neurotoxicity, which is dependent on PMN dose, is further enhanced by the priming effects of OGD and LPS. Also, time course live-cell imaging analyses support a role for direct cell-cell contact in PMN-dependent neuritic damage in dissociated neuronal cultures.

2.0. MATERIALS AND METHODS

2.1. Animals

All animal work was performed according to federal regulations and with approval by the University Committee on Animal Resources. Rats cortical neurons were dissociated from E15–17 rat embryos as described (Perry et al. 2004). Wild-type (WT) C57BL/6 mice were used for PMN isolation and co-culture experiments. LysM-EGFP mice were also used for PMN isolation and co-culture; these mice were obtained from the NAIAD Taconic farm registry and genotyped as described (Faust et al. 2000). All mice used in this study were male and matched for both age (4–6 months) and weight (30–40 g).

2.2. Neuron culture and oxygen-glucose deprivation

Cortical neurons were dissociated from E15–17 rat embryos and cultured in serum-free Neurobasal media (Thermo Fisher Scientific, Waltham, MA) supplemented with L-glutamine and B27 without antioxidants (Thermo Fisher Scientific) (Brewer et al. 1993; Li et al. 2005). Neurons were plated on poly-l-lysine-coated plastic at 16,000 cells/cm² and used at day in vitro (DIV) 7–8 for cytometer-based analyses. In parallel, neurons were plated at the same density on poly-l-lysine-coated glass slides in 24-well plates for immunocytochemistry. 5-Fluoro-2’-deoxyuridine (Sigma-Aldrich, St. Louis, MO) was added at DIV 3 for a final concentration of 20 μM in media to inhibit the growth of non-neuronal cells. At DIV 7, slides were fixed and stained with chicken polyclonal anti-MAP2 antibody, mouse monoclonal anti-GFAP antibody, and Hoechst for cell counting with 99.6 ± 1.1% neuronal purity (Figure S1). Staining duration and antibody dilutions are presented in Table S1.

Oxygen-glucose deprivation (OGD) performed on DIV 10 cultures was performed by replacing culture media with warmed HBSS and incubating cells at 0.5% O₂ in an oxygen-control glove box (Coy Laboratory Products, Grass Lake, MI) as described (Tasca, Dal-Cim, and Cimarosti 2015; Ryou and Mallet 2018). The media removed was saved for use during reperfusion. Cells were treated for 2, 4, or 6 hours and then reperfused by replacing HBSS with a mix of 50% fresh Neurobasal media and 50% conditioned media removed at the onset of OGD. Neuron cultures were reperfused for 24 hours at 21% O₂ before analysis. Injured neurites were visualized after staining with mouse monoclonal anti-MAP2 antibody, which exhibits decreased immunoreactivity in ischemia (Dawson and Hallenbeck 1996; Kitagawa et al. 1989); rabbit polyclonal anti-NeuN antibody, which is positive in all cortical neurons (Wolf et al. 1996); and Hoechst.

2.3. Cytometer-based viability assays

To determine viability after 4 hours of OGD, media was gently removed, and neurons were washed in PBS. Cells were then incubated for 30 minutes in 1 μM calcein AM and 3 μM ethidium homodimer (EthD-1, Thermo Fisher Scientific) in PBS, washed again, and maintained in PBS for immediate analysis. Cells were analyzed in PBS to prevent colorimetric interference from culture media. The non-fluorescent calcein AM dye is cell-permeant and is converted to its fluorescent calcein form by intracellular esterases in live cells. EthD-1 is impermeant in live cells but is taken up by dead cells and fluoresces when bound to DNA. Plates of neurons were analyzed on a Celigo cytometer (Nexcelom, Lawrence, MA) using the Cell Viability: Live + Dead application with illumination parameters 483/536 nm for calcein and 531/629 nm for EthD-1. A 90% well mask was used to eliminate edge effects.

The percentage of live cells was calculated as the percent of calcein(+) cells divided by the total number of calcein(+) and EthD-1(+) cells with doubly positive cells only counted once. The percentage of injured cells was calculated as the percent of doubly positive cells (with injured membranes allowing EthD-1 permeation but with active esterases converting calcein AM to the fluorescent calcein form) divided by the number of calcein(+) cells. Using brightfield, soma size for all calcein(+) cells was calculated on the Celigo with the following parameters for Direct Cell Counting: intensity threshold 6, saturation intensity 0, precision high, diameter 10 μm, background correction checked, minimum thickness 3 μm. A 90% well mask was used, the cell area range was set to >20 and <300 pixels, and the aspect ratio was set to >10 to exclude debris.

2.4. In vivo neutrophil priming and purification

For the isolation of primary PMNs, bone marrow was flushed from femurs and tibias of WT and LysM-EGFP mice using an 18-gauge needle and gently dissociated in HBSS using a 1000 ml plastic pipette. Cells were spun, washed, and incubated with red cell lysis buffer (BioLegend, San Diego, CA) for 2 minutes. White blood cells were incubated with a PMN biotin-antibody cocktail and anti-biotin microbeads prior to magnetic separation and negative selection per manufacturer instructions to yield a pure PMN population (Miltenyi Biotec, Bergisch Gladbach, Germany). For experiments involving primed and unprimed PMNs, mice were injected with either saline (SAL) or 50 μg/kg intraperitoneal lipopolysaccharide (LPS; Sigma), respectively, 6 hours prior to bone marrow harvest. We have previously characterized this dose of LPS as sufficient to induce transient PMN activation without inducing significant weight loss, fever, or behavioral change in the host. This dose also did not yield detectable endotoxin levels in serum (thus avoiding the direct effects on endotoxin on neurons) or change the concentration or ratio of circulating white blood cells (Mai et al. 2017). Plasma concentrations of IL-1β and TNF-α were measured using the LEGENDplex immunoassay (Biolegend, San Diego, CA).

2.5. Flow cytometry

Retro-orbital blood was collected 6 hours after saline or LPS injection. Whole blood was washed and lysed with red blood cell lysis buffer (Biolegend, San Diego, CA) for 5 minutes. The remaining white blood cells were stained for 30 minutes at 4°C with Brilliant Violet 421-labeled anti-Ly-6G clone 1A8 antibodies (1 μg/ml), a PMN-specific clone (Bruhn et al. 2016), and Alexa Fluor 647-labeled anti-CD11b antibodies (0.8 μg/ml) (both Biolegend). CD11b is a marker of immune cell activation that is upregulated in PMNs during both infectious and sterile inflammatory responses (Mai et al. 2017).Each sample was blocked with 4 μL Fc receptor (BD Biosciences, San Jose, CA), and cells were later fixed in 2% paraformaldehyde for 30 minutes. Fluorescence was measured using a BD LSR II Flow Cytometer (BD Biosciences). Data were collected for 10,000 events, and the following gating strategy was used: a forward vs. side scatter plot (FSC-area vs. SSC-area) was used to visually identify white blood cells and exclude cell debris. Gates were then manually drawn on plots of FSC-height vs. FSC-area plots to gate for single cells. Two-dimensional plots of Ly-6G vs. CD11b were used to identity Ly-6G^hi/CD11b^hi PMNs, and histograms were constructed for each PMN population based on CB11b fluorescence. Geometric mean fluorescence intensity (MFI) analyses were performed using FlowJo analysis software (TreeStar, Ashland, OR).

2.6. Co-cultures and live imaging

To determine the necessity of cell-cell contact in PMN-mediated neurotoxicity, PMNs were harvested from saline- or LPS-injected WT mice, suspended in HBSS, and added 4:1 to normoxic neurons for 24 hours and to 4-hour OGD neurons at the start of 24-hour reperfusion. This ratio was determined by measuring neuronal death with increasing PMN concentrations, with a ceiling noted between 4:1 and 8:1 in normoxic conditions (Table S2). The same volume of HBSS was added to all wells, including control wells without PMNs. The protocol was timed so that PMN harvest after 6 hours of in vivo LPS priming coincided with the end of 4-hour neuron OGD. This allowed for PMNs to be added right at the beginning of the 24-hour reperfusion period.

For analysis, PMNs were incubated in 5 μM CellTracker Green 5-chloromethylfluorescein diacetate (CMFDA, Thermo Fisher Scientific) for 10 minutes, washed, and resuspended in HBSS. CMFDA is a cell membrane-permeable fluorescent dye, and once taken up, will not be transferred to adjacent cells. PMNs were then added to neuron culture wells with media containing 0.1 μM EthD-1. PMNs were either directly added to neuron cultures or to suspended culture inserts with 0.4 μm pores permissive to macromolecules but not cells (Millipore). After 24 hours, total EthD-1(+) cells were counted on a Celigo cytometer with illumination parameters 483/536 nm for CellTracker Green and 531/629 nm for EthD-1. PMNs that were doubly CellTracker Green(+) / EthD-1(+) were subtracted so that only neurons were analyzed.

For observation of neurite fragmentation of single neurons in co-culture, rat cortical neurons were transfected with plasmids for farnesylated Td-tomato (3 μg per 5 million cells) per manufacturer instructions (Lonza, Cologne, Germany) before plating. Given the low efficiency of transfection using this method (~20%, data not shown), individual neurons could be imaged in isolation. EGFP(+) PMNs from LysM-EGFP mice, isolated via Miltenyi isolation columns as above, were added 4:1 to neurons. Direct PMN-neuron interactions were observed by live-cell microscopy using the Zeiss AxioObserver Z1 microscope equipped with temperature, humidity, and CO₂ control. Environmental conditions and automated image acquisition were controlled by ZEN 2 imaging software (Carl Zeiss Microscopy, Thornwood, NY). Images were acquired every 30 minutes until neuronal fragmentation was complete.

2.7. Immunohistochemisty and Image Analyses

To evaluate neurite injury during OGD, plated neurons were washed with PBS and blocked in 10% goat serum for 1 hour at 20°C prior to immunocytochemical staining with mouse monoclonal anti-MAP2 antibody, rabbit polyclonal anti-NeuN antibody, and Hoechst nuclear counterstain per Table S1. For skeleton analysis of PMN concentration on neurite injury, 20× images were acquired of MAP2-stained neurons and processed as previously described (Buczynski et al. 2018). Briefly, slab-pixel analyses measure the complexity of branching/intersecting structures, including neurites. Images were made binary in ImageJ using the same threshold values for each image (lower bound: 460, upper bound: 973) and skeletonized, resulting in a 1-pixel wide, monolayer map of neurites (Figure 4B). Characteristics if these skeletonized neurites were quantified using the Analyze Skeleton plugin (Arganda-Carreras et al. 2010), which classifies each pixel in the skeleton. The plugin recognizes “slab” pixels as pixels that have exactly 2 neighboring pixels. These pixels make up the majority of length of the neurite. “Junction” pixels have >2 neighbors and thus represent points where neurites branch or made contact (a surrogate for complexity) (Figure 4C-D).

Figure 4.

PMNs disrupt neurite branching in vitro. (A) Time lapsed image of EGFP(+) PMNs (green) interacting with neurites on a neuron labeled with farnesylated Td-tomato (red) over 16 hours. (B) Skeletonization of a MAP2(+) monolayer with orange denoting a slab pixel (neurite body) and blue denoting a junction pixel (branch or intersection). (C) Effects of OGD and PMNs at 4:1 or 8:1 on the total number of slab pixels representing total neurite length. (D) Effects of OGD and PMNs at 4:1 or 8:1 on the number of junction pixels representing branch points and points of neurite contact. Scale bar = 100 μm. Values represent means ± SD. N = 7. * p < 0.05, ** p < 0.01, **** p < 0.0001 between Ctrl and OGD; ^# p < 0.05 between different PMN:neuron ratios; 2-way ANOVA with Holm-Sidak post-hoc test.

2.8. Statistical analyses

Statistical analyses were performed using Prism (Graphpad, La Jolla, CA). Comparisons were made using 2-way ANOVA and Holm-Sidak post-hoc test for PMN titration experiments, co-culture vs. netwell insert experiments, and skeleton analyses. Initial normoxia vs. OGD experiments and serum cytokine levels were analyzed with 2-tailed t-tests. P values less than 0.05 were considered significant.

3.0. RESULTS

3.1. Optimization of oxygen-glucose deprivation

To optimize the duration of ischemia sufficient to induce sub-lethal injury, neurons were exposed to OGD conditions for 2, 4, and 6 hours, followed by 24 hours of reperfusion. We then performed immunocytochemistry using the neuronspecific nuclear marker NeuN (Wolf et al. 1996) and the dendritic marker MAP2 to assess cumulative damage in exposed cultures (Figure 1A). MAP2 is particularly useful in this regard as studies have associated early loss of MAP2 immunoreactivity with ischemic dendritic injury (Dawson and Hallenbeck 1996; Kitagawa et al. 1989). While cultures exposed to 2 hours of OGD exhibited minor reductions in MAP2 immunoreactivity, extending the period of OGD to 4 hours induced marked MAP2 loss, neuritic beading (Figure 1A, 4h arrowheads) and partial redistribution of NeuN within neuritic processes. Cultures exposed to 6 hours of OGD exhibited significant injury characterized by the global loss of distal MAP2(+) processes and the accumulation of NeuN(+) fragmented nuclei. Neurons were also stained with calcein AM and permeability marker EthD-1 to quantify fractional cell death. Results show that 4 hours of OGD caused a reduction in calcein AM signal (62.8 ± 12.8% vs. Ctrl, 99.0 ± 1.3 %; p < 0.0001) and proportionate increase in the fraction of EthD-1 positive neurons (23.5 ± 3.5% vs. Ctrl, 1.0 ± 0.7%; p < 0.0001) (Figure 1B-D). OGD exposure was also associated with a reduction in the somal area of the calcein (+) neuron population (150.4 ± 46.8 pixels vs. Ctrl, 162.8 ± 33.3 pixels; p < 0.0001) representing the fraction of neurons undergoing cell death (Figure 1E). Given the balance between low-level neuron loss and neuritic injury observed, we selected the 4-hour timepoint going forward. While we did not quantify cumulative injury at other time points, our findings were consistent with published protocols (Holloway and Gavins 2016).

Figure 1.

Titration of oxygen-glucose deprivation in vitro. (A) MAP2 (red) and NeuN (green) staining of neurons after increasing durations of OGD followed by 24-hour reperfusion. Dendritic beading was observed with 4 hours of OGD (arrowheads) with frank neurite injury at 6 hours. Scale bar = 100 μm. (B) Imaging cytometer-based analysis of neuron viability using calcein (green) and ethidium homodimer (EthD-1, red). Cell-permeant calcein AM dye is converted to its fluorescent calcein form by intracellular esterases in live cells. EthD-1 is taken up by cells with disrupted membranes and fluoresces when bound to DNA. Scale bar = 200 μm. (C) Measurements of OGD effects on the percentage of live cells and (D) apoptotic cells. (E) Effects of OGD on soma size in calcein(+) cells. Values represent means ± SD. N = 20. **** p < 0.0001 between control (Ctrl) and OGD; 2-tailed t-test.

3.2. Effects of PMN dose and OGD priming

To determine an appropriate PMN-to-neuron ratio, we delivered PMNs from naive mice to normoxic neuronal cultures at a P:N ratio of 0:1, 1:1, 2:1, 4:1, and 8:1. Studies of focal stroke or excitotoxic insults indicate PMN-to-neuronal ratios of 1:1 in the affected hemisphere (Dinkel, Dhabhar, and Sapolsky 2004). However, the in vivo distribution of PMNs and other immune subsets is not homogeneous as they tend to concentrate within and around the ischemic lesion. For this reason, we chose to study mechanisms of PMN-dependent toxicity across a range of in vitro PMN:neuron ratios to establish a threshold effect. Using EthD-1 staining as a marker of death, results indicate that after 24 hours PMNs were toxic to neurons at ratios >1:1 with a peak effect observed at a P:N ratio of 4:1 (3,814 ± 1,599 vs. 2:1, 2,866 ± 819 EthD-1 neurons; p = 0.013) (Figure 2, Ctrl). We also examined whether priming neuronal cultures with mild OGD (4 hours) would potentiate PMN-induced toxicity vs. Ctrl and found effects at 2:1 (p = 0.019), 4:1 (p = 0.0017), and 8:1 ratio (p < 0.0001) (Figure 2, Table S2).

Figure 2.

PMNs are neurotoxic in neuronal co-culture. PMNs were harvested from WT mice and added at different ratios to Ctrl (normoxic) neurons or neurons recovering from 4-hour OGD. Cell death was measured by the number of Eth-D1(+) neurons after 24 hours of co-culture. Values represent means ± SD. N = 12–36. * p < 0.05, ** p < 0.01, **** p < 0.0001 between Ctrl and OGD; n.s. = not significant, ^# p < 0.05, ^### p < 0.001 between incrementally higher PMN:neuron ratios; 2-way ANOVA with Holm-Sidak post-hoc test.

Low-level systemic inflammation heightens PMN priming and worsens cortical injury and neurological outcomes following transient global cerebral ischemia (Mai et al. 2017). Having established the priming effect of OGD on PMN-induced toxicity, we next asked whether inducing low-grade inflammation in the host before PMN harvesting would potentiate PMN-dependent damage to neurons recovering from 4-hour OGD. To test this, mice received either saline or 50 μg/kg LPS by intraperitoneal injection, and Ly-6G(+) PMNs were analyzed by flow cytometry for the activation marker CD11b (Figure 3A-B). Results show that 6 hours after LPS administration, the PMN population from LPS-injected mice express higher levels of CD11b, measured by mean fluorescence intensity (7,234 vs. 2,012). We next harvested PMNs from LPS-primed or saline-treated donor mice and labeled them with CellTracker Green before introducing them to OGD-exposed neuronal cultures to differentiate them from injured, EthD-1(+) neurons in vitro (PMN-induced neuronal damage in this context was characterized by EthD-1 uptake and somal swelling (Figure 3D). When added at a ratio of 4:1 (PMN:neuron), OGD-exposed neuronal cultures exposed to PMNs harvested from saline-primed donors exhibited increased levels of EthD-1(+) staining (32,896 ± 9,796 cells vs. Ctrl, 7,792 ± 3,867 cells; p = 0.0001) (Figure 3D-E). Cumulative levels of injury increased if PMNs were harvested from LPS-treated donors prior to delivery in culture (48,471 ± 25,253 cells; p = 0.009). To test the requirement for close-range signaling for PMN-induced toxicity, we also delivered PMNs to cultures containing netwell inserts that permit the exchange of soluble components but prevent cell-cell contact. Results show that compartmentalization prevented the death of neurons caused by either saline- (9,048 ± 2,043 cells; n.s.) or LPS-primed PMNs (9,156 ± 3,213 cells; n.s.) relative to untreated controls (7,791 ± 3,177 cells).

Figure 3.

Effects of PMN priming and direct cell contact on neuronal cultures. (A) Ly-6G^hi/CD11b^hi PMNs were collected from saline- (SAL) and LPS-injected (LPS) mice. (B) CD11b upregulation is observed in PMNs 6 hours after LPS injection. MFI = mean fluorescence intensity. (C) PMNs were harvested 6 hours after saline or LPS injection, labeled with Cell Tracker (green arrow), and added to neurons recovering from 4-hour OGD (4:1 ratio). Cells were either added directly to neuron cultures or into suspended netwell inserts, and EthD-1 (red arrow) was added to culture media. (D) Microscopy showing the effects of SAL and LPS PMNs co-cultured with neurons. Brightfield images were overlaid with channels showing Cell Tracker(+)/EthD-1(±) PMNs and Cell Tracker(−)/EthD-1(+) dead neurons. Black arrowheads indicate swelling soma of dying neurons. (E) Cytotoxic effects of PMNs from SAL- and LPS-treated mice in co-culture and in netwell inserts. Values represent means ± SD. N = 8. Scale bar = 100 μm. *** p < 0.001, **** p < 0.0001 between co-cultured PMNs and netwell inserts; ^### p < 0.001 between Ctrl and SAL; ^## p < 0.01 between SAL and LPS; 2-way ANOVA with Holm-Sidak post-hoc test.

3.3. Activated PMNs induce neuritic damage

We next employed live cell microscopy to investigate the spatial and temporal dynamics of PMN-induced neuron death. Neurons were electroporated with farnesylated Td-Tomato while EGFP-expressing PMNs were harvested from LysM-EGFP transgenic donors to facilitate tracking of cell-cell interactions (Figure 4A, Video S1). At the outset, neurons exhibited trigonal cell soma, smooth continuous processes, and active growth cones (Figure 4A, time 0:00). Over time, PMNs made frequent contact with both the neuronal soma and neuritic arbor with progressive fragmentation of these structures. To quantify the relationship between PMN-Neuron ratios and neuritic injury, we imaged co-cultures and performed skeleton analyses on neuronal monolayers transformed into 1 pixel-wide neurite maps. The data extrapolated from these image sets included slab pixels, reflecting aggregate neurite length, and junction pixels, which represent branch points or areas where ≥2 neurites overlap (Figure 4B). In the extreme case, OGD-treated neurons exposed at the 8:1 P:N ratio showed a marked and significant decrease neurite length (8,594 ± 3831 pixels) compared to 0:1 control (20,025 ± 6,905 pixels; p = 0.017). Likewise, we observed a non-significant trend toward decreased neurite length between 4:1 and 8:1 (p = 0.086). Similar results were obtained in analyzing junction pixels. OGD-primed cultures exposed to an 8:1 P:N ratio (1,392 ± 817 pixels) exhibited significant decline in cumulative junction pixels relative to 0:1 control (4,150 ± 1,646 pixels; p = 0.022) (Figure 4C-D, Table S3).

4.0. DISCUSSION

In this study, we investigated the extent to which cell-cell contact between PMNs and neurons contributes to post-ischemic neurodegeneration. Specifically, we tested the hypothesis that in vitro oxygen-glucose deprivation and in vivo delivery of the inflammogen LPS would exacerbate PMN-dependent neurotoxicity via effects on neuron and PMN priming, respectively. In addition to testing survival responses to a range of PMN:neuron ratios, we also tested the requirement for PMN-neuronal interactions in the toxicity model. Aside from demonstrating that antecedent LPS and neuronal OGD exposure enhanced PMN-dependent toxicity in vitro, we found that PMN-neuronal contact is required for neuronal death in this paradigm and that these interactions produce marked neuritic damage and degeneration.

It is well established that exposure of PMNs to pro-inflammatory mediators heightens their migration and cytotoxic potential (Leow-Dyke et al. 2012; Pflieger et al. 2018; Comen et al. 2016). However, we selected endotoxin as a PMN priming agent for these studies because it has been previously detected in the serum of up to 86% of cardiac arrest patients within 24–48 hours of resuscitation (L'Her et al. 2005; Grimaldi et al. 2015; Adrie et al. 2002). We had also shown that low-dose LPS priming produced sustained systemic neutrophil activation, CNS transmigration, and neurovascular injury in a mouse model of the post cardiac arrest syndrome (PCAS) (Mai et al. 2017). It is important to note that the dose of LPS used in the current and past work was undetectable in serum (< 0.01 ng endotoxin/ml) after IP injection (25 μg/kg) at 1/1000^th of the dose typically reported in mouse models of sepsis. Although not studied here, is would be interesting to test whether alternate TLR4 agonists (i.e., HMGB-1, heat shock proteins, hyaluronic acid) could be substituted for LPS to study in vitro PMN-induced neurotoxicity (Downes and Crack 2010; Gesuete, Kohama, and Stenzel-Poore 2014). Likewise, it would be informative to evaluate the in vitro behavior of PMNs exposed to endogenous DAMPs generated in vivo following CNS ischemia-reperfusion injury.

While there is precedent establishing PMN-dependent neurotoxicity in vitro, conclusions from these studies were based on binary assessments of cell survival (live/dead) and did not consider the separable effects of ischemia and immune priming in this context (Kim et al. 2019; Shaw et al. 2008). For example, Dinkel and colleagues evaluated PMN-dependent excitotoxicity in hippocampal cultures containing both neurons (40%) and glia exposed to 6 hours of OGD (Dinkel, Dhabhar, and Sapolsky 2004). Thus, it is likely that their findings were likely influenced by the protective effects conveyed by astrocytes (Rose et al. 2017). Allen et al. incorporated live imaging of neurons and PMNs activated by IL-1β-treated endothelium over 6 hours to model the toxicity-promoting effects of PMNs crossing an inflamed blood-brain barrier (Allen et al. 2012). In this context, activated PMNs proved to be highly toxic, causing rupture of the soma and neurite beading in 30–180 minutes. One of the considerations raised relates to arriving at a PMN-to-neuron ration that most closely replicates the in vivo situation. It has been suggested that an effector-to-target (E/T) cell ratio of 1:1 based on average hemispheric counts may be most appropriate (Dinkel, Dhabhar, and Sapolsky 2004). However, this metric does not accurately reflect the heterogeneous distribution of neutrophils, macrophages, among other immune cells within and around the ischemic core. For this reason, we investigated PMN-dependent toxicity up to an E/T of 1:8. Notably, pushing the PMN dose to this level uncovered profound neuritic degeneration in post-ischemic cultures.

Additional aspects of the experimental model merit further consideration. Aside from being more plentiful, neurons harvested from embryonic rats are typically longer-lived in culture and generate a dense neuritic network (Ray et al. 1993; Seibenhener and Wooten 2012). Together these aspects of the protocol allowed us to study PMN effects on neuron survival and neuritic injury. And while we focused on using male mice as PMN donors, future studies addressing sex-specific differences in PMN-dependent mechanisms of neuronal injury are certainly warranted (Pace et al. 2017). We cannot rule out the possibility that inter-species incompatibility may have contributed to basal levels of PMN activation and neurotoxicity. However, establishing mice as suitable PMN donors will enable a range mechanistic studies using available genetic models. Selecting an appropriate duration of OGD was also an important consideration in developing the model. In our hands, four hours of OGD/reperfusion produced an acceptable level of neuronal necrosis while at the same time, inducing neuritic beading considered a marker of sublethal injury (Li and Murphy 2008). Although not tested, extending the duration of OGD beyond the four-hour mark would likely produce neuron loss impeding the further study of PMN-neuronal interactions.

What then, are the possible mechanisms responsible for in vitro PMN-dependent neurotoxicity observed? Azurophilic (primary), specific (secondary), and gelatinase (tertiary) PMN granules contain various factors that possess antibacterial, immune, and toxic effects (Cassatella et al. 2019). The release of pro-inflammatory cytokines, proteolytic enzymes, and free radicals, including hypochlorous acid and hydrogen peroxide are known toxins, which may have damaging effects when released at close range (Nguyen, Green, and Mecsas 2017). Using live-cell microscopy, we observed neuritic tethering by PMNs suggesting that OGD/LPS priming induces the expression of DAMPs, adhesion molecules, among other factors on both neurons and PMNs that may support contact-dependent cell death signaling (Birdsall et al. 1992; Dinkel, Dhabhar, and Sapolsky 2004). Birdsall et al. showed that stimulation with TNF-α, IFN-γ, or IL-1 induces surface ICAM-1 expression on PMNs and that PMN-neuron adhesion in this context was mediated by ICAM-1—LFA-1 interactions (Birdsall 1991). In addition, the co-culture model can be readily applied to test the effects of alternate TLR4 agonists (i.e., HMGB-1, heat shock proteins, hyaluronic acid) on in vitro PMN-induced neurotoxicity (Downes and Crack 2010; Gesuete, Kohama, and Stenzel-Poore 2014). Likewise, it should be straightforward to study the functional attributes of PMNs harvested from donor mice following CNS ischemia-reperfusion injury in our co-culture paradigm.

Our findings may also have broader implications regarding the evolution of stroke pathology in vivo. Studies investigating the post-ischemic CNS immune response portray PMNs as playing a pivotal role in determining behavioral outcomes and mortality in experimental stroke (Neumann et al. 2015) and in patients following cardiac arrest (Patel et al. 2019). Our model demonstrates PMN-mediated neurite degeneration and disruption on neuronal networks, which may account for their role in neurologic dysfunction in vivo. Moreover, our work implies that direct cell contact may be required for PMN-mediated injury in the case of peripherally primed PMNs. Contrast these findings with studies by Allen et al., who demonstrated that conditioned media produced from both naive and activated PMNs were both sufficient to cause neuron death (Allen et al. 2012). Importantly, these studies differed in several ways, including our use of lower neuronal plating densities to facilitate live-cell imaging (16,000 per cm² vs. 180,000 per cm²), which could have reduced bystander-mediated inflammation and excitotoxicity. It is also possible that low-dose endotoxemia unmasked contact-dependent PMNs toxicity while having a limited effect on degranulation. Nevertheless, our study highlights the lethal consequences of introducing mild priming stimuli to neutrophils and neurons in tandem, with the intent of replicating salient cues encountered in the clinical setting.

In summary, we demonstrate that primed PMNs are harmful to both hypoxic and normoxic neurons and that direct PMN-neuron interactions are required for neuronal injury under conditions of low-dose LPS-mediated in vivo PMN priming. In addition to showing that cumulative PMN-induced neurotoxicity reflects the cumulative cell-autonomous effects of immune and ischemic priming, these studies provide more granular details regarding the damaging effects of PMNs on neuronal axonal and dendritic elements. We expect that this experimental platform will be helpful in identifying the adhesion molecules and other factors required for PMN-neuronal tethering. And given recent reports describing induced polarization of anti-inflammatory PMN subsets (Cuartero et al. 2013; Zhao et al. 2017), reductionist mixed culture systems as described will be particularly useful for screening the therapeutic potential of novel anti-PMN-based small molecules and biologics (Dong et al. 2019; Hou et al. 2019; Zhao et al. 2018).

HIGHLIGHTS.

• Oxygen-glucose deprivation induces neutrophil toxicity in cultured neurons.

• Endotoxin priming augments neutrophil-dependent toxicity in neuronal cultures.

• Close range cell-cell interactions are required for injury in the co-culture model.

• Injury in the co-culture model involves both cell loss and neuritic transection.