Animal experiments were performed according to the European Union Directive 2010/63/EU and the protocol was approved by the local authority (Regierungspräsidium Leipzig; reference numbers: TVV 51/14 for mice, TVV 34/11 for rats and TVV 56/15 for sheep). The study was carried out with n = 5 C57BL/6 mice of about 25 g body weight and n = 3 Wistar rats of about 300 g body weight (both breed by Charles River, Sulzfeld, Germany), which were randomly included, and with three male adult sheep from the nucleus breed of the educational manor (Faculty for Veterinary Medicine, University of Leipzig).

Focal cerebral ischemia in mice was induced by a right-sided pMCAO using a filament-based approach as originally described by Longa et al. (1989) with some minor modifications. Briefly, a standardized silicon-coated 6-0 monofilament (Doccol Corporation, Redlands, CA, USA) was inserted into the right external carotid artery and then moved forward into the internal carotid artery until bending was observed or resistance was felt. Mice were anesthetized with etomidate (33 mg/kg body weight i.p.; Hypnomidate, Janssen-Cilag, Neuss, Germany), added by local subcutaneous injections of lidocaine (Xylocitin 1%, mibe, Brehna, Germany) at the area of surgery. In parallel, the body temperature was monitored and adjusted to 37°C by a thermostatically controlled warming pad via a rectal probe (Fine Science Tools, Heidelberg, Germany). After surgery, mice were placed on a commercially available warming pad (37°C) until recovery from anesthesia.

In rats, focal cerebral ischemia was realized by applying a thromboembolic model as originally described by Zhang et al. (1997) with some minor modifications. Briefly, after surgical exposure of right-sided cervical arteries, a polyethylene (PE) tube with tapered end was introduced into the external carotid artery and moved forward through the internal carotid artery up to the origin of the middle cerebral artery (approximately 16 mm from carotid bifurcation). Blood was collected in a PE tube and allowed to clot on a warming pad (37°C) for 2 h followed by overnight storage at 4°C before surgery. Individual blood clots with a length of 45 mm were prepared and injected with a small volume of saline. Finally, the catheter was removed. During surgical procedure, rats were anesthetized using 2.0%–2.5% isoflurane (Isofluran Baxter; Baxter, Unterschleißheim, Germany; mixture: 70% N₂O/30% O₂; vaporisator VIP 3000, Matrix, New York, NY, USA). The body temperature was adjusted to 37°C by a thermostatically controlled warming pad with rectal probe (Fine Science Tools), also after surgery until recovery from anesthesia.

MCAO in sheep was induced as previously described by Nitzsche et al. (2016). Briefly, the subject was anesthetized by an intravenous injection of ketamine (4 mg/kg body weight; Ketamin, Medistar, Holzwicke, Germany), xylazine (0.1 mg/kg body weight; Xylazin, Ceva Sante Animal GmbH, Düsseldorf, Germany) and diazepam (0.2 mg/kg body weight; Temmler Pharma GmbH, Marburg, Germany). During surgery anesthesia was maintained by mechanical ventilation with 2% isoflurane and 40% oxygen (Primus, Dräger AG, Lübeck, Germany). After surgical preparation, the left temporal bone was exposed by incision of the covering skin. Subsequently, the temporal muscle was lifted and trepanation of the underlying bone was performed at 10,000 rpm with a 6 mm barrel burr (microspeed uni, scil animal care company, Viernheim, Germany). Next, the dura mater was incised and the proximal MCA was occluded by electrosurgical coagulation using neurosurgical bipolar forceps (ME 411, KLS Martin, Tuttlingen, Germany). Finally, the dura mater was repositioned, muscles and skin were sutured and the animals were treated with antibiotics (enrofloxacine, 5% Baytril, Bayer AG, Leverkusen, Germany) and analgesics (Butorphanol, Alvegesic 1%; CP-pharm, Burgdorf, Germany) and allowed to wake up after surgery. Two weeks after ischemia onset, sheep were anesthetized again by an intravenous injection of the above mentioned combination of ketamine/xylazine and midazolam (0.2 mg/kg bodyweight; Midazolam, Braun Melsungen, Melsungen, Germany) followed by euthanasia.

Successful induction of stroke in mice and rats was verified by the existence of a relevant neurobehavioral deficit. In detail, during the 24-h observational period, animals had to present at least 2 points on the Menzies score (Menzies et al., 1992), naturally ranging from 0 (no deficit) to 4 (spontaneous contralateral circling). In sheep, animals had to present a visible ischemic lesion in the cerebral magnetic resonance tomography (Biograph mMR, Siemens Healthcare, Erlangen, Germany) that was done at the end of the 2-week observational period.

Twenty-four hours after induction of ischemia, euthanized mice and rats were transcardially perfused with saline and 4% phosphate-buffered paraformaldehyde. The brains were post-fixed in the same fixative overnight and equilibrated in 30% phosphate-buffered sucrose. Forebrains of rodents were serially cut with a freezing microtome (Leica SM 2000R, Leica Biosystems, Wetzlar, Germany) producing 10 series of 30 μm-thick sections each.

The sheep were euthanized by intravenous administration of a mixture containing embutramide, mebozonium and tetracaine (20 mg/kg, 5 mg/kg and 0.5 mg/kg body weight, respectively; obtained as T61, Intervet Deutschland, Germany), followed by removal of the calvarium. Subsequently, about 10 mm-thick coronal slices were prepared and classified according to a sheep brain atlas (Nitzsche et al., 2015). Each slice was photographed and immersed into 4% buffered formaldehyde for the next 14 days. The slices were then equilibrated in 30% phosphate-buffered sucrose and the immersion-fixed tissue blocs consecutively cut at 40 μm-thickness using a freezing microtome (Microm HM 430, Thermo Fisher Scientific, Waltham, MA, USA). All specimens were collected in 0.1 M Tris-buffered saline (TBS), pH 7.4 containing 0.2% sodium azide and stored at 4°C prior to histochemical labeling.

Free-floating sections were rinsed with TBS prior to all histological procedures which commenced by blocking of non-specific binding sites for subsequently applied immunoreagents with 5% normal donkey serum in TBS containing 0.3% Triton X-100 for 1 h. Next, all sections were incubated overnight with mixtures of primary antibodies and biotinylated WFA—diluted in the blocking solution—as listed in Table 1. Subsequently, the sections were washed with TBS and transferred for 1 h into mixtures of appropriate fluorochromated secondary immunoreagents (all at 20 μg/ml TBS-BSA and obtained from Dianova, Hamburg, Germany) according to Table 1. In control experiments, the omission of primary antibodies or biotinylated WFA resulted in the expected absence of cellular labeling. Generally, sections were finally rinsed several times with TBS and briefly in distilled water, mounted onto fluorescence-free glass slides, air-dried and coverslipped with Entellan in toluene (Merck, Darmstadt, Germany).

An Axioplan fluorescence microscope (Zeiss, Germany) was used for the screening of fluorescently labeled brain sections. Images from entire sections and from selected regions at various magnifications were obtained with the microscope Biorevo BZ-9000 (Keyence, Neu-Isenburg, Germany) and the confocal laser-scanning microscope LSM 510 Meta from Zeiss. Panels of micrographs were created with Microsoft PowerPoint (version 2015; Microsoft Corp., Redmond, WA, USA). If necessary, brightness and contrast of micrographs were slightly adjusted without deletion or creation of signals.

Semi-quantifications were performed for collagen IV (Coll IV) and WFA-staining in the ischemia-affected neocortical layers II, III and IV of the somatosensory and parietal cortices, based on images that were captured with the Biorevo BZ-9000 microscope (Keyence). Thereby, three consecutive brain sections with a distance of 300 μm were selected, while the middle section had to demonstrate the most pronounced ischemic lesion involving neocortical areas. Next, analyses involved a forebrain region of about 600 μm in a rostral-occipital direction. In each of the three brain sections, the ischemic region and adjacent non-affected neocortex was subdivided into eight equally spaced regions of interest (ROIs) with a distance of 72 μm between their boundaries. From medial to lateral, the fourth ROI was positioned directly on the sharp border of changed Coll IV-immunoreactivity. For controls, the pattern of the eight ROIs in the ischemic hemisphere was mirrored to the contralateral, non-affected hemisphere, resulting in 16 ROIs in each of the three brain sections per animal. In detail, ROIs were captured using a 40× objective, a field dimension of 181 × 136 μm and a constant exposure time of 1/50 s for Coll IV and 1/55 s for WFA with a 12 bit CCD camera. For quantification of the fluorescence signals, the integrated density—reflecting mean gray values and analyzed area—was measured using ImageJ software (National Institutes of Health, Bethesda, MD, USA) in each ROI.

For all measurements, the threshold was adjusted to the contralateral hemisphere and kept constant at gray values of 1000 for WFA and 1250 for Coll IV to correct for any background intensities relating to non-labeled areas, such as the lumen of cross-sectioned vessels. Thresholds had been adjusted individually for each staining until the background intensity could be nearly excluded without affecting the visualization of targeted structures to prevent adulterations of the applied measurements. We thereby prevented an adulteration of our measurements, since background staining for the relevant markers was lowered in the ischemia-affected tissue. Data were processed by calculating mean values per animal to compare interhemispheric differences for Coll IV and WFA.

Calculations were performed with the IBM SPSS Statistics package version 24.0 (IBM Corp., New York, NY, USA). After descriptive statistics, the Wilcoxon test was applied to check for statistical significance between both hemispheres. Data are given as means ± standard deviation, unless otherwise indicated. In general, p < 0.05 was considered as statistically significant.

One day after induction of focal cerebral ischemia in mice, we observed opposite alterations for the vessel-associated Coll IV and binding sites for the PN marker WFA, while Coll IV-immunoreactivity became increasingly visible in ischemic areas including the striatum and neocortical structures. A considerable loss of WFA-stainable ECM and decomposed PNs were noted as exemplified by images from whole forebrain sections including affected somatosensory cortex and dorsal striatum (Figure 1A), as well as visual and auditory cortical areas, hippocampal formation and pontine gray matter (Figure 1B). Morphological changes of PN structures became visible at high magnification by confocal laser-scanning microscopy. While Figure 2A shows a representative PN with honeycomb-like meshes ensheathing the whole perikaryon and surrounding the proximal dendrites in a non-ischemic area, Figure 2B presents only WFA-stainable remnants devoid of net-like structures in close vicinity to apparently increased vascular Coll IV-immunoreactivity within the ischemia-affected tissue.

By using eight ROIs on the ischemia-affected hemisphere capturing the ischemic bordering zone, i.e., the transition from the non-affected to the ischemia-affected tissue in direction from medial to lateral, semi-quantitative analyses were added in order to verify the afore described patterns of Coll IV-immunolabeling and WFA-staining. To consider the natural patterns of PN components, the mirrored ROIs on the contralateral hemisphere served as controls (Figure 3A). Thereby, the Coll IV-immunoreactivity on the non-affected hemisphere was found to provide a consistent pattern with nearly equal values for the integrated density from medial to lateral (gray bars in Figure 3B). In contrast, ischemia led to a significant increase of Coll IV-immunoreactivity (white bars in Figure 3B) from the second (7.1 × 10¹¹ ± 4.3 × 10¹¹) to the third (13.1 × 10¹¹ ± 6.7 × 10¹¹) and continued to the fourth (24.2 × 10¹¹ ± 9.3 × 10¹¹) ROI (each p = 0.043; Wilcoxon test, n = 5). Consequently, direct inter-hemispheric comparison revealed a significant increase of Coll IV-immunoreactivity starting in the ischemic bordering zone and continued towards the most lateral ROI (each p = 0.043; Wilcoxon test, n = 5). Concerning WFA-binding sites, an ascending increase was observed in the non-affected hemisphere from medial to lateral (gray bars in Figure 3C), representing the natural pattern in neocortical areas. Thereby, a significant increase occurred from the second (19.2 × 10¹¹ ± 11.0 × 10¹¹) to the third (27.3 × 10¹¹ ± 15.8 × 10¹¹) as well as between the fourth (34.8 × 10¹¹ ± 18.2 × 10¹¹) to the fifth (45.7 × 10¹¹ ± 20.6 × 10¹¹) ROI (each p = 0.043; Wilcoxon test, n = 5). On the ischemia-affected hemisphere (white bars in Figure 3C), WFA-binding exhibited a shaped course in the bordering zone of ischemia by decreasing significantly from the first (17.7 × 10¹¹ ± 11.6 × 10¹¹) to the third (9.28 × 10¹¹ ± 10.7 × 10¹¹), followed by an increase towards the fifth (17.5 × 10¹¹ ± 11.4 × 10¹¹) ROIs (each p = 0.043; Wilcoxon test, n = 5). The pattern of WFA-binding sites provided—at least starting at the fourth ROI—a trend towards a diminished relative labeling. The direct inter-hemispheric comparison on the level of each ROI, however, barely missed statistical significance (regions 4–8, each p = 0.080; Wilcoxon test, n = 5).

By focusing on the relationship between PNs and vascular alterations as indicated by an altered Coll IV-immunoreactivity and glial NVU components, additional qualitative data were obtained from multiple-labeled forebrain sections of rodents 1 day after ischemia onset and from sheep 14 days after induction of ischemia. Additionally, net-associated parvalbumin-containing neurons were visualized to explore characteristics in the neocortical ischemic border zone of mice (Figure 4), rats (Figure 5) and sheep (Figure 6). Thereby, nearly complementary distribution patterns for WFA and massively enhanced Coll IV-immunoreactivity was found in all three species investigated (Figures 4A, 5A, 6A,A′). Remarkably, the infarcted zone of rodents was partly devoid of microglia immunoreactive for the ionized calcium-binding adaptor molecule 1 (Iba; asterisks in Figures 4A, 5A). However, in the representative image from the ischemic border zone of the sheep 14 days after infarction (Figure 6A), strongly Iba-labeled microglia/macrophages can be observed in the Coll IV-stained region, contrasted by ameboid and ramified microglial cells in vicinity to PNs.

Additional triple fluorescence labeling of Coll IV, WFA and glial fibrillary acidic protein (GFAP) demonstrated in the affected tissue from mice and rats also fragmented astrocytes at 1 day after ischemia onset (Figures 4B, 5B). In sheep, activated astrocytes became predominantly visible around the heavily Coll IV-stained region and formed a glia scar-like structure (Figure 6B). Analyses focusing on the spatial relationship to parvalbumin—as marker for most neocortical net-bearing neurons—revealed in all three models the expected up-regulated Coll IV-immunoreactivity (Figures 4C, 5C, 6C), delineating ischemia-affected tissue, whereas in the same area WFA-stained nets appeared decomposed (Figures 4C′, 6C′) or were even abolished (Figure 5C′). Parvalbumin-immunolabeling (Figures 4C″, 5C″, 6C″) detected shrunken cells and especially in the rat (Figure 5C″) combined with enhanced labeling of neuropil in the same area. In the ischemia-affected zone, merged staining patterns revealed only a few PNs around parvalbumin-containing neurons (Figure 4C‴) or even the absence of such net-enwrapped neurons (Figures 5C‴, 6C‴) while on the contrary, the healthy tissue in all models contained numerous WFA-stained PNs around parvalbumin-containing neurons. Strikingly, the overlay demonstrated nearly complementary staining patterns for Coll IV on the one hand and WFA and parvalbumin on the other hand. To address the CSPG aggrecan as a further PN constitute, triple fluorescence staining combined with the detection of WFA and parvalbumin was performed in neocortices of mice (Figures 7A–A″) and rats (Figures 7B–B″). Thereby, mice displayed a neocortical ischemia-affected area with lowered lectin-staining of the neuropil and devoid of PNs detected by WFA (Figure 7A) and aggrecan (Figure 7A′). After merging of staining patterns for PNs and parvalbumin, the calcium-binding protein was also observed in affected tissue, i.e., diminished in cells and in the neuropil with a coarser appearance than in healthy tissue (Figure 7A″). The overlay further visualized co-localized WFA-binding sites and aggrecan-immunoreactivity as well as parvalbumin-containing cells devoid of PNs. In parallel, neocortical ischemia-affected rat tissue showed a few remaining PNs recognized by WFA (Figure 7B, green) and anti-aggrecan (Figure 7B′, blue), each exemplified by an arrow. Concomitantly, parvalbumin-immunolabeling (Figure 7B″) was severely affected, while only a few apparently intact PNs surrounded cells with detectable calcium-binding protein remained and the neuropil appeared again coarse. As expected, the merge of staining patterns also visualized PNs double-labeled by WFA and anti-aggrecan (Figure 7B″).

To investigate further ischemia-affected net constituents in rodents, the immunodetection of neurocan, versican and the cartilage link protein (CRTL) was combined with double labeling of WFA and aggrecan.

Representative mouse sections displayed that the ECM stainable by WFA (Figures 8A–C; green) and aggrecan (Figures 8A′–C′; blue) were similarly abolished in the ischemia-affected area, which was also largely devoid of neurocan-immunolabeling (Figures 8A″,A‴). Notably, the apparently healthy tissue displayed a strong neuropil staining for neurocan. In parallel, immunolabeling of versican revealed several damaged PNs in close vicinity to the strongly affected tissue (Figure 8B″), whereas the detection of CRTL visualized in the presented area more PNs than WFA and anti-aggrecan (Figures 8C″,C‴). The overlay (Figures 8A‴,B‴,C‴) elucidated triple-stained nets (arrows), numerous WFA/aggrecan co-labeled nets (arrowheads) and nets mono-labeled by WFA (two-headed arrow).

The ischemic border in figures devoid of Coll IV-immunolabeling (Figures 8, 9) was localized by comparison of the presented sections with Coll IV-stained consecutive sections or by a lowered background staining typically seen in the ischemic tissue.

The distribution of WFA-binding sites in the PNs was compared with alterations of immunoreactivities for parvalbumin, Kv3.1b—which are both known as markers for fast-firing GABAergic neurons—and the “pan-neuronal” marker neuronal nuclei (NeuN) in neocortical neurons of mice (Figure 9) and rats (data not shown).

Thereby, lectin-positive PNs at the neocortical ischemic border were concomitantly revealed with Kv3.1b + parvalbumin (Figures 9A–A‴), Kv3.1b + NeuN (Figures 9B–B‴) and NeuN + parvalbumin (Figures 9C–C‴). In contrast to the rather evenly distributed PNs in the healthy tissue, the ischemia-affected areas were either devoid of WFA-labeled structures (Figure 9A) or exhibited only some remnants of nets (Figures 9B,C). In general, Kv3.1b (Figure 9A′) and parvalbumin (Figure 9A″) were co-localized, and in the overlay neurons co-expressing both markers appeared pink (Figure 9A‴). Notably, cells displaying only one of both proteins were not observed. On the contrary, the ischemia-affected region contained structures with diminished NeuN-immunoreactivity (Figures 9B″,C′) without an immunosignal for Kv3.1b (Figure 9B‴), and only parvalbumin-stained coarse neuropil as well as a few remnants of shrunken neurons (Figure 9C″). In contrast, triple fluorescence labeling in the same areas on the contralateral, non-affected hemisphere displayed homogeneous patterns for all three applied combinations of markers devoid of any cellular affection (Figures 9A^IV–C^IV).

Qualitative analyses in our study implicated the concomitant degradation of PNs and up-regulation of Coll IV in the ischemia-affected neocortices of mice, rats and sheep as the main finding of the present study. The added semi-quantification of Coll IV confirmed this observation in terms of a significantly up-regulated immunoreactivity in ischemic areas, confirming our previous findings using other mouse strains including a transgenic model with Alzheimer-like alterations and aged animals (Hawkes et al., 2013). Concerning the underlying causes, it remains to be clarified whether the increase in Coll IV-immunoreactivity corresponds to a reactive feature with the aim to stabilize vascular elements against the ischemic stimulus, or as kind of a degradation which leads to an increased level of antigens detectable by the applied immunohistochemical method. However, as matrix metalloproteinases have been considered to influence the vascular integrity and the ischemic stimulus is known to promote the production of matrix metalloproteinases (del Zoppo, 2009, 2010), the robustly observed up-regulation of Coll IV across the three animal models might be seen in the context of these mediators.

The present study revealed novel data on WFA-staining patterns at the border zone of the ischemia-affected neocortex 24 h after experimental stroke. Thereby, semi-quantification provided evidence for a shaped course with a significant drop directly at the border zone towards the ischemia-caused cellular alterations including clear cut up-regulation of Coll IV. Furthermore, the WFA-staining was relatively diminished in the area of ischemia. In detail, we were able to show an increase of WFA-staining in the medial to lateral direction at the non-ischemic hemisphere, which is assumed to represent the natural course in this neocortical region. In contrast, on the contralateral, ischemia-affected hemisphere such increased WFA-staining is lacking.

Additional novel qualitative data are the detection of reduced and structurally altered net components neurocan, versican and the CRTL. In line with previously reported data on the nucleus reticularis thalami (Härtig et al., 2016), the present study provides evidence for diminished neocortical PN markers in association with net-bearing, fast-firing GABAergic neurons such as parvalbumin and the potassium channel subunit Kv3.1b.

Immediately after stroke, matrix metalloproteinases are known to become dysregulated (Heo et al., 1999) and associated with acute neurovascular disruption. However, the same enzymes later contribute to beneficial mechanisms of neurovascular remodeling (Lo, 2008; Rosell and Lo, 2008; del Zoppo, 2010), which makes it difficult to interact pharmacologically in a consistent direction. Further candidates for the stroke-induced degradation of net components are aggrecanases, e.g., disintegrin and metalloproteinase with thrompospondin motifs (ADAMTS-1, 4, 8 and 15) and neprilysin (Cross et al., 2006; Lemarchant et al., 2013; Levy et al., 2015; Rossier et al., 2015). Interestingly, such enzymes were observed in net-ensheathed parvalbumin-immunoreactive neurons (Rossier et al., 2015). After focal cortical ischemia in rats, WFA- binding sites were found to be reduced in the demarcation zone at different time points between 24 h and 60 days after photothrombosis (Bidmon et al., 1998). Based on WFA staining, abolished and decomposed PNs were also reported for other animal models of experimentally induced stroke in neocortices of rats (Hobohm et al., 2005; Karetko-Sysa et al., 2011; Madinier et al., 2014; Quattromani et al., 2017). In line with these findings, the present study shows erased WFA staining in the neocortices from infarcted mice, rats and sheep extending own data on the affected ECM in the nucleus reticularis thalami of mice (Härtig et al., 2016).

Among the addressed net components, we detected strongly degraded aggrecan known as predominant CSPG in PNs occurring in several isoforms (Matthews et al., 2002; Morawski et al., 2012a) with differential distribution as shown for the human cerebral cortex (Virgintino et al., 2009). The fluorescence labeling also verified the occurrence of PNs double-positive for WFA and aggrecan or only for one of both markers. This finding is in line with the recently reported heterogeneity of PNs in the mouse hippocampus (Yamada and Jinno, 2017). The removal of their coatings might allow formerly net-enwrapped parvalbumin-containing neurons to re-open a more plastic state (Hensch, 2014). On the contrary, the loss of inhibitory neurons immunoreactive for parvalbumin and Kv3.1b might impair the GABAergic wiring in the cortical microcircuitry (Kubota, 2014).

Decomposed PNs are not restricted to stroke, as they were also demonstrated for other neurological disorders (Soleman et al., 2012). As for instance, in schizophrenia abnormalities CSPG markers such as WFA, aggrecan and chondroitin-6-sulfates in the ECM and associated glia were demonstrated in the amygdala and entorhinal cortex of cases with schizophrenia (Pantazopoulos et al., 2010). Based on WFA-staining of tissue from 86 post mortem human brains a strong decrease of PNs in the prefrontal cortical layers III and V contributes to the prefrontal cortex dysfunction in this disease (Mauney et al., 2013). Therefore, Berretta et al. (2015) discussed that the physiological remodeling of the ECM is disrupted and, in turn, contributes to the dysfunction of GABAergic neurons in schizophrenic patients (Berretta et al., 2015). Impaired PNs and remodeling of the ECM were also repeatedly reported for animal models of epilepsy (Dityatev, 2010; McRae et al., 2012; Rankin-Gee et al., 2015).

Notably, autoptic cases with bipolar disorder displayed in the amygdala and in the entorhinal cortex only negligible changes of CSPGs and only modest decrease of PN markers (Pantazopoulos et al., 2010, 2015). Unaltered PNs were also reported for an established animal model of β-amyloidosis in Alzheimer’s disease (AD), Tg2576 mice with cortical age-dependent vascular and parenchymal deposits of β-amyloid peptides (Morawski et al., 2010). Subsequently, Morawski et al. (2012b) extended these findings by comprehensive data on PNs in autoptic tissues from AD patients confirming the weak changes in PN structures during this neurodegenerative disease. By contrast, Suttkus et al. (2016a) found that the neuronal ECM restricts distribution and internalization of aggregated tau. Interestingly, the vast majority of cortical net-bearing neurons showed no tau hyperphosphorylation in numerous cortical autoptic samples (Brückner et al., 1999), and the neocortex of aged bisons displayed aberrantly phosphorylated tau exclusively in cells devoid of PNs (Härtig et al., 2001).

Apart from the preservation, an enhancement of net components was reported for addiction in experimental models. Repeated binge drinking caused an increase of PNs in the insular cortex of mice (Chen et al., 2015). Changes of brain ECM molecules resulting from addiction to opiates, psychostimulants and alcohol are contrasting to alterations in schizophrenia and mood disorders (Lubbers et al., 2014; Slaker et al., 2016). A single cocaine injection into the prefrontal cortex caused a decreased density of PNs, whereas five daily cocaine injections resulted in an increased density of PNs and lowered plasticity (Sorg et al., 2016).

An established tool for the remodeling the ECM and especially PNs in vivo and even under pathological conditions are bacterial glycosaminoglycan-degrading enzymes such as chondroitinase ABC (ChABC) as reviewed by Burnside and Bradbury (2014). Injections of ChABC into the rat frontal or occipital cortex resulted in the erasure of PNs within 1 day followed by the significant re-establishment of matrix components only after 4 weeks and the rebuilding of PNs only after 5 months (Brückner et al., 1998).

Following experimentally induced stroke in aged rats, the timely fine-tuned delayed treatment with ChABC promoted functional recovery and neuroplasticity, but did not protect the peri-infarct region (Soleman et al., 2012). Notably, the ChABC-induced removal of negatively charged ECM components such as glycosaminoglycans resulted in enhanced calcium diffusion within brain tissue which influences synaptic transmission, neuronal excitability and other physiological processes (Hrabětová et al., 2009). After thalamic stroke in hypertensive rats, the continuous intra-infarct infusion of ChABC rescued neuronal loss and enhanced levels of synaptophysin and growth-associated protein 43 indicating augmented axonal growth as well as synaptic plasticity to finally ensure overall neuronal function (Chen et al., 2014). Following cortical photothrombotic stroke in rats, enriched housing after ischemia induction reduced PNs, modulated activity and mRNA expression of ECM proteases and their inhibitors, accompanied by enhanced recovery of limb placement ability (Madinier et al., 2014; Quattromani et al., 2017).

Remarkably, ChABC injections into perirhinal cortices from transgenic mice with tau pathology restored even the object recognition memory of these animals (Yang et al., 2015).

Moreover, the ChABC-induced removal of PNs in the medial prefrontal cortex impaired acquisition and reconsolidation of conditioned place preference memory after its impairment by cocaine (Slaker et al., 2015).

The present study has some limitations. First, despite the use of a broad spectrum of antibodies, a comparable immunolabeling of diverse cellular and ECM structures across the three animal species was in some parts hampered because of some differences in the post-ischemic tissue processing including fixation. In detail, sheep brains were not perfusion-fixed, which was due to the experimental constraints of the underlying sheep studies. Notably, it was possible to carry out immersion fixation of about 1 cm-thick coronal forebrain slices from sheep. Consequently, frozen tissue sections from these blocs allowed for the detection of PNs and multiple vascular, glial and neuronal markers. Notably, several presented staining methods were well applicable in all three animal species analyzed. Thereby, the immunohistochemical toolbox for studies of sheep brain tissue was considerably extended. However, the amount of available sheep and rat tissue sections was restricted in this study, which allowed only qualitative investigations in these species. At least semi-quantifications were feasible in mice based on n = 5. Future studies are requested to re-produce the current quantitative findings in larger series of rats and sheep.

It is also noteworthy that experiments were limited to a single time point (i.e., 24 h after ischemia induction in rodents, and 14 days after ischemia induction in sheep). Given the fact that post-stroke reactions were described in a multi-phasic fashion (Dirnagl et al., 1999), additional work is necessary to explore the longitudinal course of addressed markers. This approach would also allow to gain more data on the underlying mechanisms of the observed histochemical alterations in terms of a primary consequence of the ischemic stimulus or a secondary effect driven by the ischemic affection of neighboring cellular structures.

This study confirmed that the ECM and in particular PNs are an important component of the NVU, which critically impacts on proper neuronal function. Therefore, novel insights into the spatial and functional relationship of the ECM and the cellular, especially vascular components of the NVU might help to elaborate additional strategies for an effective neuroprotection. In this context, future studies should not only address the acute phase of ischemic stroke to alleviate cellular damage due to the critical impairment of blood supply (Dirnagl et al., 1999), but further be aimed at later, chronic stages after stroke to improve reorganization and neuronal plasticity (Pekna et al., 2012).