The Relationship between Localized Subarachnoid Inflammation and Parenchymal Pathophysiology after Spinal Cord Injury

All animal protocols were approved by the animal ethics board of the University Health Network, Toronto, Ontario, Canada. Injuries were induced as previously described (Seki and Fehlings, 2008). Briefly, female Wistar rats approximately 300 g in weight were anesthetized with 2% isoflurane with oxygen and NO₂. The dorsal aspect of the T6, T7, and T8 vertebrae were removed, exposing the spinal cord. The experimental groups consisted of: (1) sham group (non-injured controls, laminectomy only), (2) arachnoiditis group (5 μL intrathecal injection of a 500-mg/mL kaolin mixture at T7 without injury), (3) SCI group (35-g clip compression injury at T7), and (4) PTS group (animals subjected to a 35-g clip compression injury at T7 followed by kaolin injection). Multilayer tissue closure was performed following the experimental manipulations.

Animals were perfused transcardially with ice-cold PBS and fixed by perfusion with 4% paraformaldehyde. Spinal cords were harvested and post-fixed in 4% paraformaldehyde containing 10% sucrose overnight, followed by PBS containing 20% sucrose on the following day. The cords were then frozen in OCT and sectioned longitudinally (sagittally). Sections were rinsed in PBS for 5 min and blocked in blocking solution (0.1% Triton-X 100, 1% BSA, 5% non-fat milk, and 2.5% normal goat serum in PBS) for 1 h. Primary antibodies were incubated overnight in blocking solution minus Triton-X 100 overnight at 4°C (antibody solution). Sections were rinsed 3×10 min in PBS, and fluorescent secondary antibodies were incubated for 2 h at room temperature in antibody solution. The sections were rinsed again in PBS (3×10 min) and cover-slipped in Mowiol mounting medium containing 4,6-diamino-2-phenylindole (DAPI; Vector Laboratories, Burlington, Ontario, Canada). Primary antibodies used for immunohistochemistry included: collagen IV (Abcam, Cambridge, MA), CS56 for chondroitin sulfate proteoglycans (CSPG; Sigma-Aldrich, Oakville, Ontario, Canada), glial fibrillary acidic protein (GFAP; Millipore, Billerica, MA), polymorphonuclear leukocytes (PMN; Cedarlane, Burlington, ON, Canada), and Iba-1 (Wako, Osaka, Japan). Secondary antibodies included Alexa Fluor 488 and Alexa Fluor 568 conjugates (Invitrogen, Carlsbad, CA). In all cases, the sections shown in the figures are representative of the sagittal midline of the spinal cord. All images are presented in the same orientation, with the top of the image representing the dorsal aspect of the cord, and the left side of the image representing the rostral side of the epicenter. Scale bars in all lower-magnification fluorescent images represent 1 mm. In higher-power confocal images, the scale bars represent 100 μm.

Animals were perfused transcardially with ice-cold PBS. Spinal cords were isolated and snap frozen in liquid nitrogen. For in situ zymography, samples were embedded in OCT and sectioned at 14 μm using a cryostat. For enzyme-linked immunosorbent assay (ELISA) and gel zymography, samples were embedded in OCT, and dorsal meningeal isolation was carried out by sectioning off approximately 300 μm of the dorsal surface of 0.5 cm of spinal cord tissue centered at the injury epicenter using a cryostat. This fraction contained the dorsal portion of the meninges exposed to the kaolin, and is referred to as the meningeal fraction. The remaining kaolin-free portion of the cord is referred to as the parenchymal fraction. Each isolated fraction was homogenized in RIPA buffer (Thermo Scientific, Waltham, MA).

Samples in RIPA buffer were processed using Rat Cytokine/Chemokine multiplex ELISA assays available from Millipore. Concentrations obtained from the assay were divided by protein concentrations determined by the Lowry method, and data are expressed as picograms of cytokine or chemokine per milligram of protein.

All gel zymography reagents were purchased from Bio-Rad (Hercules, CA) unless otherwise stated. Gel zymography was carried out using 10% polyacrylamide gels with 0.1% gelatin B (Sigma-Aldrich). Samples homogenized in RIPA buffer were prepared with equal protein concentrations following a Lowry protein assay. The sample buffer consisted of 65 mM Tris-HCl (pH 6.8), 20% glycerol, 4% SDS, 0.01% bromophenol blue. Following electrophoresis, 2.5% Triton-X 100 (Sigma-Aldrich) in water was used to renature the gels with gentle agitation for 30 min. Zymogram developing buffer (50 mM Tris, 0.2 M NaCl, 5 mM CaCl₂, and 0.02% Brij35) was used to equilibrate the gels for 30 min at room temperature with gentle agitation, followed by the addition of fresh developing buffer overnight at 37°C. Gels were then stained with 0.5% Coomassie Blue for 30 min and destained with methanol:acetic acid:water (50:10:40). Images were taken with a Fluor-S multi-imager (Bio-Rad).

In situ zymography was performed using DQ gelatin-fluorescein (Invitrogen) as previously published (George and Johnson, 2010). Briefly, fresh frozen tissue was sectioned in OCT and sectioned at 20 μm. Tissue was rinsed in PBS and 100 μM of DQ gelatin-fluorescein was applied to slides overnight at 37°C. The slides were rinsed in PBS (3×1 min), and cover-slipped with Mowiol as a mounting medium. Areas of gelatinase activity appear fluorescent, as the non-cleaved DQ-gelatin-fluorescein remains quenched (non-fluorescent).

The myeloperoxidase activity assay was carried out per the manufacturer's instructions (Enzo Life Sciences–Stressgen, Plymouth Meeting, PA). Briefly, the animals were perfused with ice-cold PBS, and 0.5 cm of spinal cord tissue centered at the epicenter was homogenized and centrifuged at 12,000g at 4°C. Solubilization buffer was added to the pellets, followed by sonication for 30 sec, and two rounds of freeze-thaw cycles before centrifugation as above. The supernatants from this step represent the granular fractions. In the presence of H₂O₂, myeloperoxidase (MPO) contained in the samples oxidized a non-fluorescent detection reagent into a fluorescent analog, which was detected using a fluorescent plate reader (Perkin Elmer, Waltham, MA). Data are reported as units of MPO activity per gram wet tissue weight.

Disruption in the BSCB was measured by extravasation of Evans blue (EB) as previously described (Yu et al., 2008), with few modifications. Briefly, 4% EB was injected intravenously and the animals were kept anesthetized for 30 min before sacrifice. The animals were then perfused with saline containing 10 U/mL heparin. Cords were removed (0.5-cm sections centered at the epicenter), and weighed prior to homogenization in N,N-dimethylformamide. Homogenized cords were left at 50°C for 3 days before centrifugation at 20,000g for 20 min. Absorbance at 620 nm was measured using a NanoDrop spectrophotometer (Thermo Scientific). Data are reported as the amount of EB per wet tissue weight (μg/g).

Data are presented as means and standard deviations (SD). Statistical analyses were done using StatPlus:mac Version 2009 Software (AnalystSoft Inc., Alexandria VA). Bonferroni post-hoc tests were carried out for analyses of variance (ANOVA).

When kaolin was injected into the subarachnoid space rostral to the site of injury, it remained present and recruited local and systemic inflammatory cells. Figure 1 shows cellular meningeal inflammation in response to SCI and induced arachnoiditis (PTS group). Figure 1A shows a schematic of the region where the images were taken. Global representations in Figure 1 are tiled fluorescence microscope images, whereas higher-power images are confocal micrographs. At 2 days post-injury there was evidence of both microglia/macrophages (Iba-1, labeled green in the left panel), and neutrophils (PMN, labeled green in the right panel) in the meningeal layers (dense layer of nuclei) surrounding the kaolin in PTS animals (Fig. 1B). By 7 days post-injury, the meningeal layers were more difficult to discern due to the inflammation and scarring in the area; however, some neutrophils (labeled red) remained around the injury epicenter at the dorsal aspect of the spinal cord (Fig. 1C), whereas there was little evidence of these cells surrounding the kaolin directly. In contrast, macrophages/microglia (labeled red) completely surrounded the kaolin, and were even found in the meninges at remote sites from the injury epicenter (Fig. 1D). GFAP staining for astrocytes (green) was included to delineate the parenchymal boundary.

To determine spatial information regarding the expression of inflammatory mediators, kaolin-exposed meningeal fractions were isolated and analyzed separately from the remaining cord tissue (see Fig. 2 for a schematic of sectioning). Multiplex ELISA was used to assay cytokines (IL-1α, IL-1β and IL-6) and chemokines (MCP-1, MIP-1α and GRO/KC) in these fractions. The results are shown in Figure 2 and are reported as pg/mg protein (n=3 per group; 2-way ANOVA, p<0.05 for each molecule). Non-injured sham animals are represented by hatched lines on the graphs. All cytokines and chemokines in meningeal fractions from the arachnoiditis group were significantly increased relative those from sham animals at 1 day post injury (*p<0.05, Bonferroni post-hoc test). SCI alone resulted in significant increases in meningeal IL-6, MCP-1, and GRO/KC at 1 day post injury relative to shams (p<0.05). Additionally, relative to sham animals, the combination of SCI and induced arachnoiditis (PTS group) resulted in significant increases in all of the meningeal cytokines and chemokines reported at 1 day post injury (p<0.05). Relative to the SCI group, there was a significant increase in all inflammatory mediators reported at 1 day post injury in the PTS group (p<0.05). Relative to the arachnoiditis group, the PTS group had a significant increase in only MCP-1 at 1 day post injury (p<0.05). At 3 days post injury, the arachnoiditis group had higher chemokine levels as compared to both the sham and SCI groups (p<0.05).

Spinal cord tissues free from areas of kaolin/arachnoiditis were also analyzed for cytokine and chemokine expression. Parenchymal fractions from the arachnoiditis group contained significantly more MIP-1α and GRO/KC at 1 day post-injury compared with sham animals (p<0.05 by Bonferroni post-hoc test; Fig. 2, bottom panel), with no significant differences seen at 3 days post-injury. SCI alone resulted in a significant increase in IL-6, MCP-1, MIP-1α, and GRO/KC at 1 day post-injury (p<0.05), and IL-1α, MCP-1, MIP-1α, and GRO/KC at 3 days post-injury, compared to sham controls (p<0.05). Relative to sham animals, PTS animals exhibited significant increases in all of the parenchymal cytokines and chemokines reported at 1 day and 3 days post-injury (p<0.05). Relative to the SCI group, there was a significant increase in IL-1α and IL-1β at 1 day post-injury (p<0.05), and a significant increase in each chemokine at 3 days post-injury (p<0.05) in PTS animals. Relative to arachnoiditis alone, PTS animals had a significant increase in all inflammatory mediators reported at 1 and 3 days post-injury (p<0.05).

The relative extent of neutrophil extravasation in animals was determined using an MPO activity assay and qualitative immunohistochemistry. Figure 3A shows results from an MPO assay that analyzed the enzyme activity in granular fractions, measured in units per gram of whole cord tissue (n=5 per group). Non-injured sham animals are represented by the black-hatched bars. Arachnoiditis animals did not contain significantly more MPO activity than sham animals at 1 and 3 days post-injury (p<0.05 by two-way ANOVA and Bonferroni post-hoc test). The SCI group and PTS group had significantly more MPO activity relative to sham animals at 1 and 3 days post-injury (p<0.05 for each). Relative to the SCI group, PTS animals did not have a significant increase in MPO activity at 1 day post-injury (p=1.0), but did exhibit an increase at 3 days (p<0.05). Relative to arachnoiditis alone, the PTS group had a significant increase in MPO activity at 1 and 3 days post-injury (p<0.05).

As neutrophils are the main source of post-SCI MPO, we qualitatively assessed neutrophil accumulation 1 day following injury using immunohistochemistry. Representative images were taken from the sagittal midline of spinal cords and are shown in Figure 3B. Note the significant increase in PMN (green) immunoreactivity in the parenchyma of the SCI and PTS groups compared to animals in the arachnoiditis group. Further, the number of parenchymal neutrophils in the PTS group was increased compared to the SCI group, suggesting that increases in MPO activity might correlate with an increase in neutrophil influx.

In order to determine BSCB permeability, EB was injected intravenously into injured animals and its extravasation into the spinal cord was imaged visually in fixed frozen tissue via fluorescence, and measured in homogenized tissue by absorbance spectrophotometry. Figure 4A shows representative EB fluorescence (red) images from arachnoiditis, SCI, and PTS animals counterstained with DAPI (blue), and taken 2 days following injury (with equal exposure and intensity settings). These images demonstrate that there is evidence of parenchymal EB extravasation in the arachnoiditis group, in addition to EB fluorescence in the meninges. Also note that the area of EB fluorescence in the parenchyma is larger in the PTS group than in the SCI group. Further, there is a strong fluorescence signal in the meninges in PTS animals, which includes areas remote from the injury epicenter and on the ventral side of the spinal cord. Figure 4B shows the level of EB detected in whole tissue (0.5 cm) homogenates at 1 and 3 days post-injury. Animals in the arachnoiditis group did not exhibit significantly more EB extravasation than sham animals at 1 or 3 days post-injury (p<0.05 by two-way ANOVA and Bonferroni post-hoc test). The SCI group had a significant increase in EB extravasation compared to sham animals at 1 day following injury (p<0.05), but not at 3 days (p=1.0). In contrast, PTS animals exhibited a significant increase in EB extravasation at 1 and 3 days post-injury compared to sham animals (p<0.05 for each day). Relative to the SCI group, PTS animals did not exhibit increased EB extravasation at 1 day post-injury (p=0.18), but did show significantly more at 3 days (p<0.05). Additionally, PTS animals exhibited significantly more EB extravasation compared to arachnoiditis animals at 1 and 3 days post-injury (p<0.05 for each).

As matrix metalloproteinase-9 (MMP-9) has been implicated in promoting acute blood vessel permeability associated with injury, we assessed its spatial activity at 2 days post injury. Figure 4C shows a representative gel zymogram of MMP-9 activity in meningeal and parenchymal fractions. MMP-9 activity was not detected in sham animals, and was only very slightly increased in arachnoiditis animals in both fractions. Similarly, there was slight MMP-9 activity in the meninges of animals in the SCI group, and the parenchymal activity was more pronounced. When induced arachnoiditis was associated with SCI (PTS group), there was an increase in MMP-9 activity in both meningeal and parenchymal fractions compared to the SCI group and arachnoiditis group. Additional spatial data are demonstrated in Figure 4D, which shows representative results from in situ zymography. Fresh tissue was incubated with a fluorescent gelatin conjugate 2 days post-injury. When cleaved, the previously quenched substrate becomes fluorescent, demonstrating areas of gelatinase (MMP) activity. Animals in the arachnoiditis group exhibited gelatinase activity in the meninges and slight activity in the parenchyma, though this activity was not extensive compared to sham animals. In contrast, animals from the SCI group exhibited more pronounced gelatinase activity in the parenchyma with little in the meninges. Also note the increase in gelatinase activity in the parenchyma and meninges of animals in the PTS group compared to the SCI group.

Next we set out to determine the impact of arachnoiditis on subacute fibrous scarring. Assessment of fibrous scarring was carried out immunohistologically on midline sagittal sections using collagen IV (red) and CSPG (green) at 7 days following injury. Figure 5 shows representative images revealing extensive scarring in the meninges caudal to and surrounding the kaolin in the arachnoiditis group and the PTS group. While there was little parenchymal collagen IV and CSPG immunoreactivity in arachnoiditis animals, the SCI and PTS groups contained extensive parenchymal fibrous scarring. Note the large increase in parenchymal fibrous scarring in animals from the PTS group compared to the SCI group.

At the basic science level, little has been studied regarding meningeal inflammation following SCI. Some studies of relevance demonstrate that the presence of blood in the CSF following injury can stimulate cells in the arachnoid layers to act as antigen-presenting cells and to initiate an inflammatory response (Xin et al., 2010). Additionally, when inflammatory mediators are introduced into the subarachnoid space, they cause a breakdown in the blood–cerebrospinal fluid barrier and potentiate inflammation (Ichikawa et al., 2011). Following human SCI, inflammatory mediators have been detected in the CSF (Kwon et al., 2010), though the source of these molecules is likely from a combination of parenchymal and meningeal cells. Importantly, we acknowledge the limitations of our spinal cord sectioning procedures, and acknowledge that it is unlikely that the “meningeal” fraction was completely devoid of parenchymal cells.

To our knowledge, no study has specifically isolated the meninges and looked at the expression of inflammatory mediators. As such, data from the literature are typically representative of a combination of products from parenchymal and meningeal cells. Inflammatory cytokine mRNAs (tumor necrosis factor-α [TNF-α], IL-1β, and IL-6) are elevated within minutes after SCI, peak after several hours, and return to basal levels after 3 days (Bartholdi and Schwab, 1997; Basu, 2004; Donnelly and Popovich, 2007; Pineau and Lacroix, 2007; Yang et al., 2004). In light of this fact, our assay is most likely assessing post-peak production of cytokines. Additionally, maximal expression of chemokines is achieved between 12 and 24 h post-SCI (Lee et al., 2000b; Pineau and Lacroix, 2009; Tonai, 2001), suggesting that we were seeing peak chemokine production at our 1-day post-injury time point.

Maximal BSCB permeability reported in the literature is reached 1 day following injury, with some permeability still evident at 1 week (Noble and Wrathall, 1989; Popovich et al., 1996). Our results are consistent with these results (Fig. 4). However, in PTS animals there was no decrease in EB extravasation from 1 to 3 days, as was the case in the SCI group, suggesting that there is prolonged BSCB permeability in these animals. This increase was likely due to a combination of meningeal and parenchymal blood vessel permeability. It should be noted that we did not examine other vasoactive substances that could have influenced BSCB, such as reactive oxygen species, kinins, histamines, nitric oxide, and elastases (Fleming et al., 2006; Noble et al., 2002). The increased permeability that was associated with kaolin+SCI can partly be explained by tight-junction breakdown due to increased MMP-9 activity. Following SCI, MMP-9 activity has been identified to contribute to BSCB permeability (Noble et al., 2002), and reduced MMP-9 activity has been associated with attenuation of tight junction degradation and BSCB permeability (Lee et al., 2012).

Upon the association of arachnoiditis with SCI (PTS animals), some parenchymal inflammatory mediators detected were merely additive (i.e., the inflammation seen was equal to that of arachnoiditis alone plus SCI alone). However, others were greater than that of each combined, suggesting a synergistic increase. These include IL-1α and IL-1β at 1 day post-injury (Fig. 2), the inflammatory chemokines reported at 3 days post-injury (Fig. 2), BSCB permeability at 3 days (Fig. 4), and the extent of parenchymal scarring (Fig. 5). It should be noted that the number of animals per group in most figures is small, and thus negative findings may reflect inadequate power. In the majority of cases, the variability is reasonable and we feel that additional numbers would not alter the interpretation of our data.

There was some modest dorsal meningeal inflammation induced by the moderate clip compression/contusion injury used in our study (SCI group). From this observation, we postulate that not all SCI patients develop PTS because (1) arachnoiditis is either cleared in a timely fashion following injury, or (2) it is not severe enough to lead to alterations in parenchymal pathophysiology or cause subarachnoid scarring. The clip contusion/compression injury causes extensive parenchymal hemorrhaging, as is seen in humans following SCI. It is likely that blood products released from damaged vessels in the parenchyma can diffuse into the CSF and stimulate meningeal inflammation.

Our data demonstrated that the induction of arachnoiditis following SCI can have a significant impact on parenchymal pathophysiology. This increase could make the spinal cord more susceptible to the development of syringes, as it has been demonstrated that increased severity of parenchymal lesions leads to an increased incidence of syrinx formation (Brodbelt et al., 2003a). Furthermore, studies suggest that complete—presumably more severe—injuries are more prone to syrinx development in humans (Vannemreddy et al., 2002). In parallel studies, we have demonstrated that a bioengineered hydrogel containing hyaluronan can reduce arachnoiditis caused by kaolin injection and improve histological and functional outcomes following SCI (Austin et al., 2012).

There is most likely a reciprocal inflammatory relationship when arachnoiditis is associated with SCI. Inflammatory cells from the parenchyma can migrate to sources of inflammation in the meninges, and vice versa. Similarly, local inflammatory mediators can flow via extracellular fluid/CSF to alter inflammation in areas remote to the site of the inflammatory stimulus. It is believed that approximately 30% of CSF is derived from ECF (Brodbelt and Stoodley, 2007), allowing for movement of such macromolecules from the parenchyma to the subarachnoid space and vice versa. We did not examine meningeal fibroblasts in this study, though they are expected to play a role in cytokine and chemokine production, as well as deposition of the fibrous scar in the parenchyma. Indeed, others have demonstrated that following SCI, proliferative meningeal cells can enter the spinal cord parenchyma (Brazda and Muller, 2009; Parr et al., 2007).

As we begin to recognize the heterogeneity in SCI, models like these are needed to properly mimic the permutations seen in certain subpopulations of patients. Previous studies from our lab demonstrated that our PTS model significantly decreased locomotor recovery and increased neuropathic pain (Seki and Fehlings, 2008). This suggests that arachnoiditis could have a substantial impact on patient quality of life.