Ablation of lysozyme M-positive cells prevents aircraft noise-induced vascular damage without improving cerebral side effects

Animals were treated in accordance with the Guide for the Care and Use of Laboratory Animals as adopted by the U.S National Institutes of Health. Approval was granted by the Ethics Committee of the University Medical Center Mainz and the Landesuntersuchungsamt Rheinland-Pfalz (Koblenz, Germany; permit number: 23 177-07/G 15-1-094 and extensions). Male LysMCre^+/+ iDTR^+/+ were bred with C57BL/6J females purchased from Janvier according to a published protocol [32, 68]. The resulting pups were heterozygous LysMCre^+/− iDTR^+/−, of which only the males were used in our experiments at an age of 6–8 weeks (female mice display inconsistent vascular responses due to hormonal changes during their menstrual cycle). For a minor part of the experiments (e.g., brain flow cytometry and vascular function measurement of mesenteric microvessels), homozygous LysMCre^+/+ iDTR^+/+ mice were used. A total number of 173 mice was used. All mice were housed in a standard 12-h light/dark cycle in an institutional animal facility with ad libitum access to standard rodent chow and water. After all treatments, the mice were sacrificed by cardiac puncture under isoflurane anesthesia and whole blood as well as organs were harvested for further investigations. Animal experiments were conducted at the same time of day with all four groups concurrently to account for circadian variation of several measured parameters.

Diphtheria toxin (Sigma; 322326) was dissolved in commercially available PBS to a concentration of 5 and 1 ng/µL. The toxin was then intraperitoneally injected once a day for 10 days, killing only the LysM⁺ cells expressing the inducible diphtheria toxin receptor, as previously reported [5]. For the first 3 days, mice were injected at a dose of 25 ng/g of mouse weight. For the remaining 7 days, the dose was reduced to 5 ng/g. This protocol was adapted from a published protocol [32, 68]. The time schedule of all treatments is shown in Fig. 1. The study protocol utilized four groups of age-matched male LysMCre^iDTR mice. Control (CTR, no noise exposure and no DTX treatment.), Noise (4 days of noise exposure), DTX (10 days of diphtheria toxin treatment), and DTX + Noise (10 days of DTX treatment and 4 days of noise exposure).

Non-invasive blood pressure (NIBP) measurements were performed using a CODA High-Throughput 2 Noninvasive Blood Pressure System (Kent Scientific, Torrington, USA) according to a previously published protocol [34, 39, 63]. This method has been proofed in accuracy in comparison with radiotelemetric measurement of implanted catheter devices by Feng et al. [15]. Measurements were preceded by three training sessions to acclimate the animal before the baseline measurement and avoid stress reactions that may affect vascular tone and resting blood pressure. Conscious animals were allowed to enter the restraining tube freely, placed on a preheated platform at 32 °C, and allowed to rest for 15 min. The occlusion (O) cuff was fitted at the base of the tail and the volume–pressure recording (VPR) cuff was placed distal to the O-cuff. 15 NIBP measurements were taken per animal at each time point, the first five of which were discarded as acclimation cycles. The mean values of the remaining ten NIBP measurements were used in calculations. Measurements took place in a quiet setting at a set time each day to account for diurnal variation of blood pressure.

The noise chamber maintains the same housing conditions as the institutional animal facility to minimize transfer stress. Aircraft noise events are 43-s-long noise events, irregularly distributed over a 2-h sequence, which is repeated constantly over a 4-day period. The noise events are separated by irregularly distributed periods of silence to avoid early adaptation processes. Noise is applied through downward-facing speakers positioned approximately 30 cm above open mouse cages as described [34, 39]. The sound system is a Grundig MS 540 with a total output of 65 W. The noise events had a maximum sound pressure level of 85 dB(A) and a mean of 72 dB(A), volumes which are below the threshold for cochlear damage [65]. Sound pressure levels were calibrated with a Class II Sound level meter (Casella CEL-246) within one of the cages at initial setup.

After dissection, aortas were maintained in ice cold Krebs–Henseleit (KH) buffer and cleaned of perivascular fat. 4-mm segments of thoracic aortas were cut and mounted on force transducers within the organ bath chamber (25 mL). The rings were adjusted to a stable basal vascular tone and maximally constricted using bolus of 1 mL of 2 M KCl followed by a wash with KH buffer, then further boluses of 62.5, 125, 250, and 500 µL of 2M KCl to create a control contractile curve. The rings were then washed again and pre-constricted with prostaglandin F2α (PGF2α) to yield approximately 80% of the maximal tone induced by KCl bolus. Concentration–relaxation curves in response to increasing concentrations of acetylcholine (ACh) and nitroglycerin (NTG) were performed as previously described [40, 48]. Relaxation responses are reported as a percentage of maximal PGF2α constriction.

Dihydroethidium (DHE)-dependent fluorescence microtopography was used to determine the vascular ROS formation in cryo-sections of the aorta, as described [47, 49, 70]. Freshly harvested tissues were dissected and directly embedded in Tissue Tek (Sakura; 4538) and either immediately frozen in liquid nitrogen or incubated with the NOS inhibitor l-NAME (0.5 mM) before snap freezing. Samples were sliced on a cryostat (Leica; CM3050S). To perform the staining, slides were thawed and warmed to 38 °C and incubated with 1 µM DHE solution for 30 min. ROS-derived red fluorescence was detected using a Zeiss Axiovert 40 CFL microscope, Zeiss lenses, and Axiocam MRm camera. To investigate the involvement of eNOS uncoupling in ROS production and endothelial dysfunction, red fluorescence of DHE oxidation products was only quantified in the endothelial cell layer in aortic rings that were incubated with l-NAME. l-NAME increases endothelial cell-dependent ROS signal (DHE oxidation-dependent fluorescence) in control samples when eNOS is producing nitric oxide, and decreases endothelial cell-dependent ROS signal, when eNOS is producing superoxide (uncoupled enzymatic state) [11]. Of note, DHE fluorescence was quantified for a defined area in each aortic cryo-section and was normalized to this specific area to ensure comparable read-out for all evaluated animals. Normalization of the DHE signal to specific area was explained in detail for endothelial DHE staining in [11] and for vascular wall DHE staining in [50].

Oxidative stress from superoxide was also measured by an HPLC-based method (modified from [77]) to quantify 2-hydroxyethidium levels as previously described [69, 70]. Briefly, tissue of aorta or heart was incubated with 50 µM DHE for 30 min at 37 °C in PBS. Tissues were homogenized in 50% acetonitrile/50% PBS using a glass homogenizer (heart) or pulverized in a mortar under liquid nitrogen and resuspended in homogenization buffer (aorta), centrifuged and 50 µL of the supernatant were subjected to HPLC analysis. The system consists of a control unit, two pumps, mixer, detectors, column oven, degasser and an autosampler (AS-2057 plus) from Jasco (Groß-Umstadt, Germany), and a C18-Nucleosil 100-3 (125 × 4) column from Macherey & Nagel (Düren, Germany). A high pressure gradient was employed with acetonitrile and 50 mM citrate buffer (pH 2.2) as mobile phases with the following percentages of the organic solvent: 0 min, 36%; 7 min, 40%; 8–12 min, 95%; 13 min, 36%. The flow was 1 mL/min and DHE was detected by its absorption at 355 nm, whereas 2-hydroxyethidium was detected by fluorescence (Ex. 480 nm/Em. 580 nm).

Protein was isolated from frozen tissue by homogenization under nitrogen, then incubated in lysis buffer for 30 min with intermittent vortexing. Following incubation, the samples were centrifuged at 12,000 × g at 4° for 20 min. The resulting protein-rich lysate was extracted from the pellet and measured in concentration by Bradford assay. Samples were then prepared in Laemmli buffer to a concentration of 2 µg/µL. Gel electrophoresis was conducted with 30 µg of protein on 6 and 12% gels, transferred onto a nitrocellulose membrane, and blocked with 5% milk protein in TBS-T. Primary antibodies NLRP3 (Cell Signaling #15101, 1:750) and TXNIP (Cell Signaling #14715, 1:1000) were incubated overnight at 4° with either α-actinin or β-actin as loading controls (Sigma, 1:2500). Membranes were then washed 3 × in TBS-T and incubated with respective anti-mouse or anti-rabbit peroxidase-coupled secondary antibodies (Vektor PI-2000 or PI-1000, 1:10,000) for 2 h at room temperature, at which point positive bands were detected using ECL development (Thermo Fisher, 32106) and Chemilux Imager (CsX-1400M, Intas). Densitometric quantification was performed using ImageJ software [42].

Following protein isolation and determination of concentration as described in the Western blot section, 20 µg of protein was transferred into each well to deposit on the nitrocellulose membrane (Sigma Aldrich, WHA10402506) via a Minifold I vacuum Dot-Blot system (Schleicher&Schuell, 10484138CP) [34, 39], washed twice with 200 µL of PBS then dried for 60 min at 60 °C to adhere the proteins. The membranes were then cut and incubated in Ponceau S solution (Sigma, P7170) for protein visualization. The stain was removed and the membrane blocked for 1 h at room temperature with 5% milk in PBS-T. The membranes were incubated with antibodies against malondialdehyde (MDA)-positive proteins (Abcam, ab27642, 1:1000) as well as CD68 (Abcam, ab955, 1:1000), and interleukin 6 (Abcam, ab6672, 1:1000) overnight at 4 °C. The membranes were then treated in the same manner as those of Western blots.

Procedure for immunohistochemistry has been previously described [34, 39, 63]. Briefly, aortic segments retaining adventitial and perivascular tissue were fixed in 4% formaldehyde, embedded in paraffin, and sliced into sections of 5 µm. Following deparaffinization, samples were blocked with normal horse blocking solution (Vector) and stained with a primary antibody against NOX-2 (1:200, #LS-B12365, LS Bio) or 3-nitrotyrosine (1:200, Merck-Millipore, Darmstadt, Germany), biotinylated with a secondary antibody (Thermo Fisher Scientific, Waltham, MA), and then imaged with Image ProPlus 7.0 (Media Cybernetics, Rockville, MD).

Total mRNA from aortic and brain tissue was isolated using the acid guanidinium thiocyanate–phenol–chloroform extraction method. 125 ng of total RNA was used for quantitative reverse transcription real-time PCR (qRT-PCR) analysis using QuantiTect Probe RT-PCR kit (Qiagen) as described previously [22, 51]. Primer–probe–mixes purchased from Applied Biosystems (Foster City, CA) were used to analyze the mRNA expression patterns of NADPH oxidase 2 (NOX-2 Mm00432775_m1), eNOS (NOS3, Mm00435204_m1) and vascular cell adhesion protein 1 (VCAM-1, Mm00449197_m1) in aorta and NFκB2 (NFκB2, Mm00479810_g1), CD40L (CD40L, Mm_00441911_m1) in brain. All samples and tissues were normalized on the TATA box binding protein (TBP, Mm_00446973_m1) as an internal control. For quantification of the relative mRNA expression, the comparative ΔΔCt method was used. Expression of target gene in each sample was expressed as the percentage of unexposed wild type.

Circulating catecholamines (adrenaline and noradrenaline) were determined in mouse plasma using a commercial enzyme-linked immunosorbent assay (ELISA) kit (TriCat, IBL, RE59395) following the instructions of the vendor [39]. Corticosterone plasma levels were determined by commercial ELISA (Enzo, ADI-900-097) [39].

Immunohistochemistry for microglia was performed on 4–5-μm-thick de-paraffinized sagittal brain sections as described previously [34]. Endogenous peroxidase was blocked using H₂O₂ (3%) in PBS-T for 30 min. Subsequently, sections were washed twice with 1 mL PBS-T (2 min) followed by incubation in 5% NGS for 20 min at RT, followed by 2 × 2 min washing with PBS-T. Blocking of endogenous biotin was performed with an Avidin/Biotin Blocking Kit (SP-2001; Vector laboratories). Sections were then incubated with the respective primary antibody overnight at 4 °C (polyclonal rabbit anti-Iba1, 1:400, WAKO). After washing with PBS-T (2 × 3 min), immunoreactivity was visualized by the avidin‐biotin complex method and sections were developed in diaminobenzidine (DAB, Sigma). For negative controls, the primary antibodies were omitted. The micrographs were processed using ImageJ software (https://​imagej.​net/​ImageJ [60]) as follows. First, a number of microglia were counted using the cell counter plugin and calculated as number of microglia/mm². Then, color deconvolution function (vector H DAB) was used to unmix RGB images. Micrographs from channel two (brown signal) were then processed using the threshold function. The Iba-1 immunoreactive area in % was then quantified using the analysis function.

Flow cytometry of aortas and whole blood was performed as described previously [33, 68]. Briefly, cleaned aortic vessels were digested with liberase™ (1 mg/mL, Roche, Basel, Switzerland) for 30 min at 37 °C and passed through a cell strainer (70 μm) to yield a single-cell suspension. Single-cell suspensions were treated with Fc-block (anti-CD16/CD32), washed, and surface stained with CD45 APC-efluor 780 (#47-0451-82, 2 µg/mL), NK1.1 PE-Cy7 (#25-5941-81, 2 µg/mL), F4/80 APC (#17-4801-82, 4 µg/mL) from eBioscience (San Diego, CA) and TCR-β V450 (#560706, 2 µg/mL), CD11b PE (#553311, 2 µg/mL), Ly6G FITC (#551460, 5 µg/mL), and Ly6C PerCP-Cy.5.5 (#560525, 2 µg/mL) from BD Biosciences (San Diego, CA). Dead cells were excluded by staining with Fixable Viability Dye eFluor506 (#65–0866-14, 1:1000, eBioscience, San Diego, CA). Based on a live gate, events were acquired and analyzed using a BD FACS CANTO II flow cytometer (Becton Dickinson, Franklin Lakes, NJ) and FACSDiva software (Becton Dickinson, Franklin Lakes, NJ), respectively. Gating was performed according to a previously published strategy [63].

The protocol was previously published [21, 57]. Brain tissue was dissected from LysMCre^+/+iDTR^+/+ mice which were transcardially perfused with 0.9% NaCl solution (B. Braun, Melsungen, Germany). The dissected brains were digested with papain (1 mg/mL, Sigma-Aldrich, St. Louis, MO) and DNase I (100 μg/mL, Roche, Basel, Switzerland) for 30 min at 37 °C and were mechanically homogenized using the gentleMACS™ Dissociator (Miltenyi, Bergisch Gladbach, Germany) during the incubation time. The cell suspension was filtered through a cell strainer (70 μm) to yield a single-cell suspension. Cells were separated using a 30%:70% Percoll (Sigma-Aldrich, St. Louis, MO) gradient centrifugation for 45 min, 500 ×g at 18 °C without brakes. Myelin was discarded, and the cells at the interface were collected and washed in PBS containing 2% FCS (Thermo Fisher, Waltham, MA). Single-cell suspensions were treated with Fc-block (anti-CD16/CD32, BioXCell, Lebanon, NH), washed and surface-stained with CD45 BV510 (#103138, BioLegend, San Diego, CA, 0.7 µg/mL), CD11b PE-Cy7 (#101216, BioLegend, San Diego, CA, 0.2 µg/mL), CD80 PerCP-Cy™5.5 (#560526, BD Biosciences, Franklin Lakes, NJ, 0.5 µg/mL), CD86 FITC (#553691, BD Bioscience, Franklin Lakes, NJ, 0.5 µg/mL), MHC Class II eFl450 (#48-5321-82, eBioscience, San Diego, CA, 0.05 µg/mL), Ly6G PE (#127608, BioLegend, San Diego, CA, 0.4 µg/mL), Ly6C V450 (#560594, BD Bioscience, Franklin Lakes, NJ, 0.7 µg/mL), and F4/80 APC (#123116, BioLegend, San Diego, CA, 1 µg/mL) for 30 min at 4 °C. Dead cells were excluded by staining with Fixable Viability Dye eFluor™ 780 (#65-0865-18, 1:1000, eBioscience, San Diego, CA). For intracellular staining, cells were fixed and permeabilized using the Foxp3/Transcription Factor Staining Buffer Set (#00-5523-00, eBioscience, San Diego, CA) and stained for 12 h at 4 °C with CD68 APC (#137008, BioLegend, San Diego, CA, 0.4 µg/mL). The samples were acquired and analyzed using a BD FACS CANTO II flow cytometer (Becton Dickinson, Franklin Lakes, NJ) and FACSDiva software (Becton Dickinson, Franklin Lakes, NJ). Microglia were identified by gating on live, single, CD45^intermediate CD11b⁺ cells. Brain infiltrates were separated according to their expression levels of CD11b into CD45^high CD11b⁺ and CD45^high CD11b^- populations.

Microvascular reactivity was measured in first-order arterioles of isolated retinas using video microscopy as previously described [16, 18]. After dissection, eyes were transferred into ice cold KH buffer for preparation of the ophthalmic artery, isolation of the retina, cannulation the ophthalmic artery and placing the retina onto a transparent plastic platform. Subsequently, retinal arterioles were pressurized to 50 mmHg, visualized under brightfield conditions, and equilibrated for 30 min. Mesenteric small arteries/arterioles of the 6th and 7th order were isolated and measured as previously described for other small blood vessels with minor modifications [19, 74]. Briefly, a mesenteric vascular tree was isolated and cleaned from surrounding tissue by fine-point tweezers and Vannas scissors. Next, a micropipette was inserted into the lumen of the proximal part of the vascular tree (typically into the 4th branch) and advanced into the 6th or 7th branch using the lumen as a guide channel. Once the micropipette tip was placed, the vessel was tied to the pipette with 10.0 nylon suture material. The other end of the arteriole was tied to another micropipette. The vessel was then pressurized to 50 mmHg and visualized under an inverted microscope. Next, concentration–response curves for the thromboxane mimetic, U46619 (10⁻¹¹ to 10⁻⁶ M; Cayman Chemical, Ann Arbor, MI, USA), were conducted for both retinal arterioles and mesenteric small arteries/arterioles. For measurement of vasodilation responses, vessels were then preconstricted to 50–70% of the initial luminal diameter by titration of U46619 and responses to the endothelium-dependent vasodilator, acetylcholine (10⁻⁹ to 10⁻⁴ M; Sigma-Aldrich, Taufkirchen, Germany) and to the endothelium-independent nitric oxide donor, sodium nitroprusside (SNP, 10⁻⁹ to 10⁻⁴ M, Sigma-Aldrich), were determined. ROS formation was measured in retinal blood vessels in 10 µm cryosections of the retina by dihydroethidium (DHE, 1 µM)-derived fluorescence (518/605 nm excitation/emission), as previously described [17, 75].

Systolic and diastolic blood pressure were significantly increased by noise exposure at days 8 and 10 of the treatment regimen (days 2 and 4 after beginning of noise exposure), respectively. Ablation of LysM⁺ cells had only minor effects on blood pressure in mice without noise exposure, but prevented a blood pressure increase in the noise-exposed animals (Fig. 2a). The endpoint blood pressure values (day 10 of treatment regimen) showed a significant difference in blood pressure between noise and all other groups (Fig. 2b). The acetylcholine (ACh) dose–response relationship was significantly impaired in response to noise, while endothelium-independent relaxation to nitroglycerin was not impaired (Fig. 2c). LysM⁺ cell ablation did not modify ACh-responses of mice aorta unexposed to noise, but prevented endothelial dysfunction in the noise-exposed group. Aortas of mice from the DTX + Noise group showed identical endothelium-dependent and -independent vasodilation as compared to the control group (Fig. 2c).

Noise exposure caused uncoupling of eNOS, exhibited by an increase in DHE signal in the endothelial cell layer and significant suppression by l-NAME (Fig. 3a). Endothelial eNOS uncoupling was mostly prevented by LysM⁺ cell ablation. Prevention of eNOS uncoupling was also evident in both LysM⁺ cell ablation groups by an increase in DHE signal upon treatment with l-NAME. DHE staining in aortic cryo-sections revealed that noise increases vascular superoxide production, which is prevented by ablation of LysM⁺ cells (Fig. 3b). Likewise, using quantitative HPLC analysis of the superoxide-specific DHE product, 2-hydroxyethidium, we established that noise increased superoxide production in aortic and cardiac tissue compared to control, which was prevented by DTX treatment (Fig. 3c, d).

Flow cytometry analysis showed an increase in the presence of all leukocytes (CD45⁺), myeloid cells (CD11b⁺), macrophages (F4/80⁺), and monocytes (Ly6G⁻Ly6C⁺) in the aortas of mice subjected to 4 days of noise exposure. Depletion of LysM⁺ cells prevented this increase (Fig. 4a–d). Complementary to the influx of immune cells into the aorta, we observed a drop in all subsets (CD45⁺, CD11b⁺, Ly6G⁻Ly6C⁺, F4/80⁺) in whole blood of noise-exposed mice (Fig. 4e) that we interpret as a shift of these cells from the blood stream to the aortic tissue and potentially other organs.

Noise induced a significant increase in expression of eNOS, VCAM-1, and NOX-2 mRNA as well as NOX-2 protein as previously reported [28]. While there was a reduced expression for eNOS and NOX-2 in DTX-treated LysMCre^iDTR mice, there were no increases in expression in LysM⁺ cell-depleted mice in response to noise exposure (Fig. 5a–c). A similar observation was made for NOX-2 protein expression in the aorta that was increased by noise, which was prevented at least in part by DTX treatment (Fig. 5d). In line with the upregulation of NOX-2 expression, the markers of oxidative stress, 3-nitrotyrosine, and malondialdehyde were increased in the aorta as well as plasma of noise-exposed control mice, whereas no such increase was observed upon noise exposure of LysM⁺ cell ablated mice (Fig. 5e, f).

The smooth-muscle-dependent vasoconstrictor, U46619, was used to generate constriction of retinal arterioles, demonstrating no difference between any of the groups (Fig. 6a). Similarly, there was no detectable difference from control in smooth-muscle dependent vasodilation via nitroprusside in any of the groups (Fig. 6b). In contrast, endothelium-dependent relaxation induced by acetylcholine illustrated a significant reduction in vasodilation in noise-exposed mice, which was partly alleviated by LysM⁺ cell ablation (Fig. 6c). DHE staining in retinal vessels revealed a similar pattern, with a significant increase in DHE signal upon noise exposure, where ablation generally defended against ROS formation (Fig. 6g). In mesenteric vessels, no significant differences were found in the U46619-induced constriction in any of the groups in homozygous males (Fig. 6d) nor in the endothelium-independent vasodilation (Fig. 6e). A stark contrast was found in endothelium-dependent vasodilation between the noise-exposed group and all others, demonstrating clear protection in the DTX + Noise group (Fig. 6f), whose function remained similar to control. These responses were mirrored in heightened levels of ROS formation upon noise exposure and normalization of ROS formation in DTX + Noise (Fig. 6h).

The moderate inflammation induced by noise was increased following LysM⁺ cell ablation. An increase in Iba-1⁺ staining in the DTX-treated groups and altered microglial phenotype indicates an increased presence and potentially activation of microglia (Fig. 7a, perimeter evaluation not shown). The immunohistochemical data were also supported by increased levels of cerebral NFκB subunit 2 mRNA as well as NLRP3 and TXNIP protein (Fig. 7b). Downstream of the NFκB-NLRP3-TXNIP signaling axis, IL-6 protein, CD40L mRNA as well as CD68 protein were upregulated by trend by noise and further increased in the DTX-treated groups supporting the neuroinflammatory phenotype (Fig. 7c). The stress response and release of stress hormones adrenaline, noradrenaline, and corticosterone was also more pronounced in the LysMCre^iDTR mice with DTX treatment (Fig. 7d). In line with the neuroinflammatory phenotype in the DTX groups, we also observed a trend of elevated cerebral malondialdehyde and p67^phox levels, a marker of lipid peroxidation and a regulatory subunit of the NOX-2, in all DTX-treated mice, and this tendency was most pronounced for malondialdehyde and even became significant for p67^phox in the noise-exposed DTX mice (not shown). Cerebral eNOS protein expression showed an almost identical pattern with most pronounced increase in the noise-exposed DTX-treated mice (not shown). Noise exposure increased all these parameters in the control mice at least by trend.

We sought to characterize the seemingly opposing central and peripheral immune responses following noise exposure and DTX treatment. Gating on live, single cells, we were able to discern immune cell types based on their expression of CD45 and CD11b. Namely, microglia are CD45^intermediate CD11b⁺, peripheral myeloid populations are CD45^high CD11b⁺ and peripheral lymphoid populations are CD45^high CD11b⁻ (Fig. 8a). We found significant increases in three markers of microglial activation (CD68, CD86, MHC-II) in Noise, DTX, and DTX + Noise groups, indicating an active central immune response to both noise and DTX treatment (Fig. 8b). Furthermore, we found infiltration of both lymphoid and myeloid peripheral cells in the brains of noise- and especially DTX-treated mice (Fig. 8c, d). Further analysis revealed these myeloid cells to be bone-marrow-derived monocytes and macrophages (Fig. 8e, f), implying either microglial signaling to the peripheral immune system, disruption of the blood–brain barrier, or failure of ablation of the few resident peripheral cells in brain tissue.