Introduction
Reperfusion of cerebral vessels after transient occlusion is associated with immune cell recruitment, which contributes to both damage of the vessel and the surrounding tissue. Polymorphonuclear granulocytes (PMNs) are considered to play a prominent role in microvascular responses to ischemia via protease-mediated tissue damage and neuronal cell death during reperfusion by direct contact with CNS tissue [
2,
7,
19,
39,
49,
60]. This necessitates PMN extravasation across the blood–brain barrier (BBB) into the brain parenchyma at early stages after the ischemic insult, prior to terminal neuronal damage [
25].
Yet, the effects of experimental neutropenia in rodent stroke models range from decreased infarct volume [
12,
62,
75] to little or no influence on lesion size [
11,
44,
45], and PMN mobilization from the bone marrow induced by granulocyte colony stimulating factor (G-CSF) does not worsen clinical outcome in murine stroke models [
83] or in a recent clinical phase II trial (AXIS) [
76]. Furthermore, although some animal experiments targeting adhesion molecules mediating the multi-step PMN migration across inflamed microvessels have shown reduced infarct size [
25], clinical trials targeting PMN adhesion to endothelial ICAM-1 [
9,
36,
47] or aiming at preventing PMN infiltration into the brain by neutralizing the αMβ2-integrin (CD11b/CD18) [
54] have failed to alleviate stroke severity, suggesting that the mode of PMN action requires reassessment.
These discrepancies may reflect differences between the pathogenic mechanisms involved in ischemia/reperfusion in animal experiments and human stroke cases or, alternatively, could be linked to the assumption that PMNs use ICAM-1 and β2-integrins to migrate into the brain parenchyma in reperfusion injury, as in other inflammatory scenarios. This, however, has not been formally shown and most PMN depletion studies or adhesion molecule blocking studies after cerebral ischemia either did not assess PMN/immune cell infiltration [
15,
16,
67] or employed methods that do not permit unequivocal identification of PMNs [
37,
52,
61,
78]. A common problem has been the use of broad specificity reagents such as the antibody clone, RB6-8C5, targeting the Gr-1 antigen, which recognizes both Ly6G and Ly6C and therefore stains PMNs and monocytes [
35]. Similarly, myeloperoxidase staining identifies PMNs but also monocytes and activated microglial cells [
13]. Hence, the use of such broad specificity reagents in flow cytometry and/or immunohistochemistry inevitably results in an over-estimation of PMN numbers. It is therefore important to reassess the role of PMNs after cerebral ischemia using more specific immune cell markers that are now available, in particular in relation to their association with adhesion molecules and their precise localization within the vasculature and/or brain parenchyma.
That CNS vessels and neurons are functionally coupled is well illustrated by the rapid response of neurons to focal ischemia. Mechanistically, this is achieved by the neurovascular unit (NVU), composed of a monolayer of specialized endothelial cells (EC) interconnected by complex tight junctions, the underlying endothelial basement membrane (BM) and a second BM, known as the parenchymal BM as it marks the border to the CNS parenchyma. Together with the associated ensheathing layer of astrocyte endfeet, the parenchymal BM defines the glia limitans [
3,
64,
89], which is structurally and functionally interconnected to the surrounding neurons via astrocytes and microglia [
14,
18]. With the exception of capillaries, where endothelial and parenchymal BMs fuse to form one composite BM, in all brain parenchymal vessels the endothelial and parenchymal BMs are structurally and biochemically distinct entities, which define the inner and outer limits of the perivascular space.
The NVU strictly controls immune cell emigration from the blood vessel that requires a cascade of adhesive interactions that are well described for neuroinflammation [
32]. Due to the specialized structure of the NVU, leukocyte entry into the brain parenchyma involves two differently regulated steps: migration of leukocytes across the endothelium into the perivascular space and progression across the glia limitans into the brain parenchyma. In a murine model of multiple sclerosis, experimental autoimmune encephalomyelitis (EAE), induction of disease symptoms occurs only upon immune cell penetration of the glia limitans into the CNS parenchyma, whereas accumulation of inflammatory cells within the perivascular space does not translate into clinical disease [
1,
8,
88], highlighting the significance of the glia limitans as the effective border to the CNS parenchyma.
While histological analyses of brain samples from murine models of transient ischemia and human stroke tissues have detected “inflammatory infiltrates” in the brain [
33,
68], there has been little attempt to either specifically identify PMNs using unique molecular markers or to localize their precise position within the brain at defined time points after ischemic stroke. This study presents a collaborative investigation involving stroke researchers, neuropathologists, and basic scientists to determine the temporo-spatial relationship between immune cells and blood vessel micro-architecture in the mouse and human brain at early (acute) stages after ischemia using a panel of markers for different subsets of myeloid cells, endothelial cell adhesion molecules, and the BMs of the NVU. We employ the specific PMN marker, Ly6G [
17], in immunohistochemistry and in double and triple immunofluorescence confocal microscopy to precisely localize PMNs in relation to the cellular and BM components of the NVU, and to investigate correlations between PMN localization and altered vessel permeability or expression of endothelial cell adhesion molecules known to be involved in PMN rolling (P-selectin), arrest and crawling (ICAM-1, ICAM-2) or diapedesis (PECAM-1, CD99, JAM-A). An in vitro model for the BBB [
82], where PMN migration across an endothelial monolayer is measured under physiological flow, is employed to investigate the effects of transient oxygen–glucose deprivation followed by reoxygenation. Our mouse and human data highlight the need for a critical reappraisal of the precise site of PMN action after stroke and molecular targets for therapies to reduce reperfusion injury after stroke.
Discussion
By bringing together stroke researchers, neuropathologists, cell biologists, and neuroimmunologists specialized on the cellular and extracellular matrix components of the NVU and immune cell penetration of the NVU, we have been able to comprehensively investigate the in vivo PMN localization after ischemic stroke in mouse and human samples. Our data support an early appearance of PMNs after ischemic stroke in both mouse tMCAO and in human samples, as shown by others [
21], but contrary to previous concepts our data show that PMNs are (1) limited in number, (2) associate with vessel lumina or the perivascular and leptomeningeal space, and (3) do not strictly correlate with either platelet aggregates, sites of increased vessel permeability, or sites of enhanced expression of endothelial adhesion molecules known to be required for PMN extravasation in inflammation. Studies of CNS autoimmune inflammation have shown that localization of immune cells within the leptomeningeal and perivascular space is not sufficient to induce disease symptoms [
1,
8]. Rather, penetration of the parenchymal BM is required before access to CNS parenchyma and induction of pathological processes is possible [
1,
84]. Hence, the localization of PMNs to the vasculature early after ischemic stroke necessitates reassessment of their role in stroke.
Although some adhesion molecules, such as VCAM-1, were upregulated on vessels in the ischemic hemisphere, the expression of adhesion molecules was heterogeneous with some vessels having low and others high expression levels, and there was no spatial correlation with PMN accumulation within vessels or in the perivascular space. While previous studies have investigated adhesion molecules in ischemic stroke showing results similar to those obtained here [
16,
51,
63], no previous study has correlated in vivo adhesion molecule expression with localization of PMNs. Indeed, most studies have involved flow cytometry or myeloperoxidase expression in excised brains to quantify PMNs while adhesion molecules were analyzed by immunofluorescence microscopy on tissue sections, which led to the false conclusion that the two are correlated. This has also been the justification for employing mice lacking ICAM-1, or the use of function blocking antibodies targeting adhesion molecules in MCAO experiments, which have produced variable results [
9,
36,
47]. Normally, extravasation of PMNs during inflammation occurs at the level of postcapillary venules [
69] and involves E- and P-selectin-mediated rolling on the endothelial cell surface, and subsequent ICAM-1 mediated arrest and diapedesis across the endothelial cell monolayer [
31,
55]. The absence of a spatial correlation between upregulated expression of endothelial P-selectin, VCAM-1 and ICAM-1 and vascular sites of PMN accumulation in the tMCAO samples suggests the absence of the complete cascade of these events and that the mode of endothelial activation that occurs after ischemic stroke is not sufficient to trigger PMN extravasation into the brain parenchyma. This is supported by the in vitro studies involving pMBMECs, which demonstrated that while OGD/reoxygenation can upregulate endothelial ICAM-1 this was not sufficient to support transmigration of PMNs across the pMBMEC monolayer. In addition, the failure of those few PMNs that enter the perivascular space to penetrate into the brain parenchyma proper also reflects the absence of the molecular signals required for their invasion into the CNS as observed in inflammation.
The concept that the brain parenchyma is a tissue that is unique in its resistance to leukocyte diapedesis has previously been suggested by others, who have shown that even direct intracerebral injection of chemotactic cytokines that are sufficient to induce PMN extravasation into other tissues fail to trigger PMN extravasation into the brain parenchyma [
4]. Thus, PMN migration from the blood stream across the BBB and the glia limitans into the brain parenchyma requires more than presence of chemotactic factors and induction of leukocyte adhesion to cerebral endothelium.
The precise molecular mechanism involved in PMN accumulation within vessels observed in the current study is not clear and we can only speculate on the molecules involved. The fact that PMNs accumulated in larger vessels, mainly arterioles, further supports the hypothesis that the molecular mechanism/s involved are distinct from those described above for postcapillary venules. A recent study involving transient ligation and reperfusion of the vena cava reported a similar PMN accumulation that was not associated with expression of the classical adhesion molecules or platelets, but rather with DNA NETs released by PMNs and resulting in their aggregation [
85]. This was not the case here, as NETS were not detected in any of the tMCAO samples. However, it is possible that hypoxia or changes in blood flow, resulting from vessel occlusion, alters the expression of as yet unidentified adhesion molecules, either upregulating molecules that promote PMN adherence but also downregulating anti-adhesive molecules. The endothelial glycocalyx, composed largely of glycosaminoglycans carrying highly negatively charged heparin sulfate chains, is known to repel cells in the circulation, and the depletion of heparan sulfate chains significantly enhances immune cell adhesion [
57,
58]. As the glycocalyx changes with blood flow [
42], it is conceivable that it may present part of a novel adhesion mechanism in larger vessels. This is supported by the recent identification of a role for myeloperoxidase, released from PMNs, in charged interactions between PMNs and the gylcocalyx of endothelium which promotes PMN adherence [
53]. While this is a promising field, the role of charge molecules such as cell surface glycans has rarely been considered in immune cell recruitment, mainly due to the complex chemistry associated with their analysis.
Importantly, our data provide evidence that previous studies have overestimated PMN contribution to stroke because of the methods used for their identification and shed new light on previous animal studies and clinical trials. The largest majority of previous studies have focused on permanent MCAO, rather than the temporary MCAO employed here, which may be one reason for the comparatively low PMN numbers detected in our study. In addition, the high numbers of PMNs detected in the leptomeningeal areas, especially surrounding the optical nerve as detected here but also reported previously [
38], suggest that previous studies solely based on flow cytometry or enzymatic assessment of PMN infiltration into the brain are likely to have equated these findings with number of PMNs present within the ischemic brain parenchyma. Myeloperoxidase, commonly used as a PMN marker, is expressed by PMNs, monocytes and activated microglia present in the ischemic brain [
13], while the Gr-1 antibody (clone RB6-8C5), frequently used to identify PMNs, recognizes both Ly6C and Ly6G expressed by monocytes and PMNs, respectively [
35]. The use of these tools would therefore inevitably lead to an over-estimation of PMNs. Only few studies have attempted to localize PMNs in the brain parenchyma at defined times after MCAO, including that of Garcia and Kamijyo [
38], who employed electron microscopy of permanent MCAO in rats. Consistent with our data, the majority of the PMNs were found in vessels, and while rare isolated PMNs were detected outside of the endothelial layer, the magnifications shown are too high to definitively state that the PMN is in the brain parenchyma (and not in the perivascular space) nor do they provide evidence for neuronal death in the close vicinity of PMNs, rather isolated images of damaged neurons are shown. The concept that PMNs contribute to reperfusion injury at early stages after ischemic stroke has led to clinical trials targeting PMNs to minimize infarct volume, most of which have not been successful [
25]. While this may in part be attributable to factors such as patient cohort size or demography, or adverse effects of the PMN targeting strategies [
36], the data presented here suggest that the role of PMNs in ischemic stroke remains unclear and needs further investigation before their consideration as a relevant therapeutic target.
The PMN specific anti-Ly6G antibody (clone 1A8) employed here [
17], in conjunction with immunofluorescence staining for defined markers of the vascular basement membranes and confocal analyses have permitted a reliable identification and localization of PMNs in ischemic brain sections, and highlight the importance of supplementing quantitative analyses of ischemic tissue by in vivo localization studies. However, the choice of such in vivo imaging technologies is also crucial; while intravital microscopy (IVM) via a cortical window has been used to assess in vivo leukocyte–endothelial cell interactions in the ischemic brain [
5,
48], the working depth achievable is limited and in addition only meningeal postcapillary venules run parallel to the surface of the skull, whereas cortical postcapillary venules are orientated perpendicularly to the surface of the brain and the focal plane [
46,
71]. Hence, it is impossible to discern leukocyte–endothelial interactions in cortical postcapillary venules as this requires tracing of leukocytes over a defined period along a certain vessel length. As meningeal and parenchymal blood vessels differ from one another in several aspects [
64], including the absence of a glial ensheathment in leptomeningeal vessels [
70] and the constitutive expression of P-selectin in meningeal but not parenchymal microvessels [
6], studying meningeal vessels in stroke is not a substitute for cortical microvessels. Given our detection of a high number of PMNs in the leptomeninges, intravital imaging of such areas might overestimate PMN recruitment into the brain after stroke. Taken together, this suggests that confocal microscopic analysis of thick brain sections using defined markers of immune cells and the borders of the NVU is, at present, the most efficient mode of assessing immune cell infiltration after ischemic stroke in experimental models.
The analysis of human autopsy samples corroborated the observation that PMNs rarely occur in the brain parenchyma early after stroke. Clearly, it is impossible to prove that the patient samples analyzed had undergone reperfusion after the vessel occlusion/brain ischemia and to define this histologically. However, spontaneous reperfusion is a frequent phenomenon in clinical stroke, which may also occur in parts of the ischemic area, and even in patients receiving thrombolysis often some brain tissue undergoes infarction despite reperfusion [
43,
74]. Hence, a spectrum ranging from no reperfusion to different times and extent of reperfusion are likely in such autopsy samples, validating the comparison with the mouse tMCAO data. The question therefore arises how the concept that PMN leukocytes strongly infiltrate the brain parenchyma within the first 24 h after onset of brain infarction has gained access into widely accepted neurological and neuropathological textbooks [
33]. Several studies have defined the time frame of PMN detection in the brain after ischemic stroke, mainly using SPECT and CT scan analyses of injected tracer immune cells [
2,
87]. Already these studies revealed conflicting results about when and over which time period PMNs are detectable in the brain; however, the discrepancies were attributed to the injection of mixed leukocytes [
87] versus purified PMNs prior to SPECT [
2]. While SPECT permits high temporal resolution of such tracer cells, it has poor spatial resolution and does not permit the distinction between PMNs within or outside the confines of the NVU. A second point of conflict might be the extent of foci of fresh bleedings in mainly ischemic infarction. Even selected pictures from classical textbooks [
33], employed as evidence for PMN entry into the brain parenchyma after ischemic stroke, depict fresh foci of hemorrhages and associated leakage of various cellular populations including erythrocytes and monocytes, indicative of more severe vascular disruption. Therefore, vascular disruption might be a very logical explanation for the detection of PMNs in the CNS parenchyma in stroke cases accompanied by hemorrhages. However, these findings differ considerably from classical forms of purely ischemic stroke lesions. Finally, pioneering studies addressing the role of PMNs in stroke did not have access to the complex panel of immunocytochemical markers, available now, to precisely define cell types and their localization in relation to the cellular and BMs layers of the NVU, but rather relied mainly on classical histomorphological examination [
38,
79]. These drawbacks might constitute problems if sections are assessed by histology only where PMNs may be confused with different stages of apoptotic figures, which also frequently occur in hypoxic-ischemic brain lesions, due to their similar morphology. While isolated studies exist, like that of Lindsberg et al. [
56], which included early human stroke samples without secondary hemorrhages showing CD15-positive cells in the lesioned CNS, even these state that granulocytes in early infarction are mostly found within the intravascular space and typically aggregate at the walls of CNS vessels and are, therefore, in accordance with our findings. However, Lindsberg et al. admit that a systematic evaluation of morphological details such as whether granulocytes were still surrounded by a capillary lumen was not possible in their cohort.
In conclusion, our data indicate that PMNs do not gain access to the brain parenchyma early after ischemic stroke and highlight the vascular compartment of the NVU rather than neurons, as previously suggested also by others [
20], as the site of potential PMN action.