MyD88 Is a Critical Regulator of Hematopoietic Cell-Mediated Neuroprotection Seen after Stroke

Neuroinflammation is critical in the neural cell death seen in stroke. It has been shown that CNS and peripheral responses drive this neuroinflammatory response in the brain. The Toll-like receptors (TLRs) are important regulators of inflammation in response to both exogenous and endogenous stressors. Taking advantage of a downstream adapter molecule that controls the majority of TLR signalling, this study investigated the role of the TLR adaptor protein myeloid differentiation factor 88 (MyD88) in the control of CNS and peripheral inflammation. Reversible middle-cerebral artery occlusion was used as the model of stroke in vivo; in vitro primary cultured neurons and glia were subject to four hours of oxygen and glucose deprivation (OGD). Both in vitro and in vivo Myd88−/− animals or cells were compared with wild type (WT). We found that after stroke Myd88−/− animals have a larger infarct volume compared to WT animals. Interestingly, in vitro there was no difference between the survival of Myd88−/− and WT cells following OGD, suggesting that peripheral responses were influencing stroke outcome. We therefore generated bone marrow chimeras and found that Myd88−/− animals have a smaller stroke infarct than their radiation naive counterparts if their hematopoietic cells are WT. Furthermore, WT animals have a larger stroke than their radiation naive counterparts if the hematopoietic cells are Myd88−/−. We have demonstrated that MyD88-dependent signalling in the hematopoietic cell lineage reduces infarct size following stroke and that infiltrating cells to the site of neuroinflammation are neuroprotective following stroke.


Introduction
Myeloid differentiation factor-88 (MyD88) acts as an adaptor protein in the signalling of the Toll-Like receptors (TLRs) and the interleukin-1 (IL-1) receptor [1,2]. These receptors detect pathogen associated molecular patterns (PAMPs), highly conserved regions among various pathogens. As well as PAMPs we now know that TLRs are capable of detecting endogenous damage associated molecular patterns (DAMPs). This study investigated whether MyD88-dependent signalling contributes to the damage associated with stroke.
Stroke induced hypoxia and ischemia causes cells directly affected to suffer energy failure and die. Upon doing so they release many intracellular components, several of which are TLR ligands, including heat shock proteins (HSPs), hyluronic acid, DNA complexes and heparin sulphate [3,4]. The binding of TLRs leads to the activation of kinases, and subsequent activation of transcription factors including AP1 and NFkB [5]. These transcription factors then go on to cause the release of pro-and anti-inflammatory cytokines including, IL-1, IL-10, IL-8, IL-12 and chemokines including chemokine (C-C motif) ligand 2 (CCL-2) [6,7,8]. The inflammatory environment created due to TLR activation is proposed to be both beneficial and detrimental, in the case of stroke this is particularly true as dead and dying cells need to be removed and it is difficult for the resident phagocytotic cells, microglia, to manage excessive inflammation. Therefore understanding the contribution of a crucial signalling pathway critical to the pathogenesis of sterile cerebral inflammation is important.
It has recently been shown that TLR2 contributes to the inflammation following cerebral ischemia [9], but did not alter the attraction of granulocytes to the infarct area. Attraction of invading cells is an important component of stroke pathophysiology as they can both propagate and control inflammation in the brain [10,11]. The release of IL-10 from T-regulatory cells has been shown to control the inflammatory response following stroke and can lead to a smaller infarct size [12]. TLRs are highly expressed on cells of a granulocytic origin and there is early evidence for TLR involvement in the recruitment of hematopoietic cells following CNS injury [13,14]. The response to stroke is a complex integration of both the CNS and invading cells from the periphery. This study employed MyD88 2/2 animals and bone marrow chimeras to elucidate the role of MyD88-dependent signalling in both components of this integrated response.

Animals
All animal experiments complied with the regulatory standards of, and were approved by, the Walter and Eliza Hall Institute Animal Ethics Committee and the University of Melbourne, Medicine Dentistry and Health Sciences Animal Ethics Committee (Ethics #0911133). MyD88 2/2 mice were on a C57Bl6 background, male and backcrossed to 15 generations and were generated by and kindly sourced from Professor Akira [6]. All surgery was performed under isoflurane anesthesia, and all efforts were made to minimize suffering.

Primary glial culture
Primary glial cultures were isolated from 1 day old pups as described previously [15]. Briefly, cortices were isolated and meningies removed the clean cortices were digested in 0.02% v/v trypsin and mechanically dissociated until a single cell suspension was achieved. Cells were initially plated at a density of 1 brain in 10 ml of DMEM with 10% v/v FBS and 100 U/ml penicillin G, and 100 mg/ml streptomycin sulfate, media was changed every three days, after 7 days in culture cells were split with trypsin and thereafter when 85% confluence was reached. Glial cultures were used between 14 and 28 days in vitro.

Primary neuron culture
Primary neurons were isolated as described previously [16]. Briefly, cortices of E14 pups were isolated and meningies removed, the clean cortices were digested in typsin A single cell suspension was ensured with mechanical disruption, cells were plated at a density of 6.0610 5 cells cm 2 in neurobasal media containing 2% v/v B-27 supplement, 0.25% v/v 200 mM l-Glutamine, 0.1% v/v gentamicin and 10% v/v fetal bovine serum (FBS). After 3 hours all media was changed to FBS free media and thereafter every 3 days a half media changed was conducted. All experiments took place after 10-12 days in culture and neuronal purity was deemed to be at least 90% with NeuN (Neuronal Nuclei).

Oxygen glucose deprivation
Both neurons and glia were plated on NUNC tissue culture plates and received a complete change of media to glucose free media (Gibco) at half the normal volume and were placed in the oxygen and glucose deprivation (OGD) chamber de-gassed for 5 minutes with 100% N 2 and subsequently placed in a normal CO 2 incubator for 4 hours. At the end of OGD glucose containing media (Gibco) was added to the cells to make up the normal volume. Control cultures received the same media changes with media containing glucose.

Chimera development
Endogenous haematopoiesis was ablated in adult C57BL/6 CD45.1 or Myd88 2/2 mice by irradiation (11 Gy in 2 equal doses 2-3 hours apart). The heads of the mice were shielded. These recipient animals were then injected intravenously with 1610 6 unfractionated bone marrow cells from un-manipulated C57BL/6 CD45.1 or Myd88 2/2 mice. 8 weeks post-transplantation, chime- A. Myd88 2/2 mice have a significantly larger volume of infarct than their WT counterparts (n = 9-10, * Significantly different from WT mice, P,0.05, error bars indicate 6SEM). B. Representative TTC staining demonstrates that the infarct in both the WT and Myd88 2/2 animals covers both the cortex and striatal regions of the brain. C. Neurological deficit wt (n = 9) and Myd88 2/2 (n = 10) mice. Scattergram shows the difference between the two groups. Animals were scored after 24 h of ischemia/reperfusion using the following criteria: 1, normal spontaneous movement; 2, unable to extend forepaw; 3, animal circling towards left; and 4, animal unresponsive to noxious stimuli. doi:10.1371/journal.pone.0057948.g001 rism was determined by flow cytometric analysis of peripheral blood leukocyte in conjunction with CD45-specific monoclonal antibodies. Only mice exhibiting donor engraftment of greater than 80% were used in this study.

MCAO
Middle cerebral artery occlusion (MCAO) surgery was conducted as described previously [17] and modified [18]. All MCAO experiments and analysis were carried out in a randomised operator blinded fashion. Briefly, mice were anesthetised with 5% isoflurane in O 2 and throughout surgery maintained on 2% isoflurane. The external carotid artery was isolated and whilst the common carotid artery was temporally clamped a silicone filament of 0.2160.02 mm (Doccol) diameter was inserted via the external carotid to occlude the middle cerebral artery for a period of one hour. Occlusion was monitored by laser doppler flowmetry and greater than 75% occlusion was deemed sufficient, which all mice achieved. To end occlusion the mice were re-anesthetized and the filament was withdrawn. Throughout and following surgery mice were maintained at 37uC. There was no surgical mortality WT mice and 7% in MyD882/2 mice. At 24 hours of reperfusion mice were killed and brains were immediately removed and sliced in a mouse brain matrix to 500 m m thickness. These were placed in a 2% 2,3,5-Triphenyltetrazolium chloride (TTC) in PBS solution at 35uC for 15 minutes. Photomicrographs were captured using a Zeiss Axioskop microscope and infarct area was determined using the Image J software (v1.47; NIH). Tissue swelling in the injured side was accounted for by dividing the infarct area from each section by the ratio of the areas of the injured relative to injured side. The Cavalieri formula was used to calculate total lesion volume [Volume = SA x t x ISF] where A = sum of the corrected infarct areas; t = section thickness (500 mm) and ISF = inverse of the sampling fraction.

Evaluation of neurological deficits
Neurological deficits of the mice that had undergone stroke surgery were measured on a scale of 1-4 [17], 24 h after surgery. The following grading system was used: 1, normal spontaneous movements; 2, animal unable to extend fore paw; 3, animal circling toward left; and 4, animal crouched and unresponsive to noxious stimuli.

Immunofluorescence and Infiltrate detection analysis
At 24 hours of reperfusion animals were anesthetised as above and tissue was fixed with a transcardial perfusion with 4% Paraformaldehyde in PBS. Brains were the cut into 10 mm sections and mounted on superfrost plus glass microscope slides. For immunofluorescence coronal sections were blocked in CAS-block (Invitrogen) for 45 min at room temperature and then exposed to primary antibody overnight at 4C. The primary antibody targeting NeuN (Millipore) was used at 1:1000 and Mac-1 [19] was used at 1:10 in a 1% w/v solution of BSA in PBS. Sections were then washed with TBS-t for 3615 min and incubated with secondary antibody for 45 min at room temperature. Secondary antibodies were used at a dilution of 1:1000in a 1% w/v solution of BSA in PBS (Alexa-fluor 488 and 595 anti rat and mouse, Invitrogen). Following secondary antibody incubation slides were was in TBS-T for 3615 min and cover slips mounted using a mounting media containing DAPI (Vectorshield-Vector) and images recorded on a Leica DMI 6000B fitted with wide field fluorescence. Leukocytes were detected with a naphthalene granulocyte assay (Sigma) as per manufacturers instructions. An optical fractionator stereological design [20] was employed to make unbiased estimates of Mac-1 positive cells in the infarcted cortex of WT, MyD882/2 and chimeric mice using Stereo Investigator software version 7. Cells that were positive to Mac-1 were counted on approximately 10 sections separated by 100 mm from each animal using a 620 objective and a 200 mm 6150 mm frame area.

Statistical analysis
All statistics were completed on Prism software, for multiple comparisons the means of the experimental groups were compared using the nonparametric Kruskal-Wallis test with a post-hoc Dunn's multiple comparison test. The nonparametric Mann-Whitney test was used for comparison between two groups. P,0.05 was considered statically significant all data is displayed as the mean 6 standard error of the mean (SEM).

MyD88 signalling affects infarct size following MCAO
The contribution of MyD88-dependent signalling to the inflammatory progression of infarct following stroke was assessed using the MCAO model of stroke. Following 2 hours of MCAO and 24 hours of reperfusion the Myd88 2/2 mice suffered a significantly larger infarct volume, WT 48.3866.10 mm 3 Vs. Myd88 2/2 84.2366.50 mm 3 P,0.05, P = 0.0076 (Figure 1A), despite a similar drop in blood flow to the area of infarct during occlusion ( Figure S1A). The infarct in both of the animals was spread across both the cortex and the striatum ( Figure 1B). The measurement of body weight, mean arterial blood pressure, blood pH, PCO 2 and O 2 saturation were found to be not significantly different between genotypes and consistent with previously published data [18,21] (supplementary data, table S1).

MyD88 signalling does not affect neuronal survival
To determine if the larger infarct size in Myd88 2/2 mice after stroke is due to the intrinsic response of the neurons and glia both WT and Myd88 2/2 primary cultured neurons and glia were subjected to 4 hours of oxygen and glucose deprivation (OGD). After 24 hours of reperfusion the survival of both cell types was assessed by MTT assay. The survival of neurons following 4 hours of OGD and 24 hours of reperfusion was the same regardless of the absence or presence of MyD88; this was also true for the WT and Myd88 2/2 glial cultures (Figure 2 A & B).

MyD88 is required for cellular infiltration following MCAO
To investigate the role of MyD88 in the attraction of haematopoietic cells to the injury site, WT and Myd88 2/2 brain sections were stained with a granulocyte specific esterase assay. WT mice brain sections demonstrate more leukocyte positive esterase cells than Myd88 2/2 animals ( Figure 3A). Sections were further stained for Mac-1, as Mac-1 is present on activated microglia, macrophages, and neutrophils, its detection within injured tissue is likely to be indicative of an inflammatory response. Mac-1 was detected alone and co-localised with the neuronal nuclear marker NeuN in the WT mice ( Figure 3B), primary antibody alone and secondary antibody showed no staining (data not shown). The few cells that were Mac-1 positive in the Myd88 2/ 2 mice were negative for NeuN. Both mice showed no Mac-1 positive cells on the contralteral side of the brain (data not shown). The number of Mac-1 positive cells per frame in the WT ipsilateral hemisphere was significantly higher than the number in the Myd88 2/2 brains ( Figure 3C, WT 8.3560.72 cells/frame Vs. Myd88 2/2 2.42 0.52 cells/frame P,0.05). The WT showed more Mac-1 positive cells than the Myd88 2/2 , and in the WT brains the Mac-1 co-localised with the neuronal marker NeuN. This is indicative of infiltrating phagocytic cells targeting neurons.

MyD88 signalling in hematopoietic cells affects infarct size
The results shown in figure 2 were suggestive that the detrimental response to the stroke in the Myd88 2/2 was not due to intrinsic defects to the brain cells. This suggested that peripheral cells, namely hematopeotic cells were contributing to the outcome. To address this issue we generated chimera mice, deficient in Myd88 except cells of hematopoietic origin (Myd88 2/2 + WT), only in cells of hematopoietic origin (WT + Myd88 2/2) or neither (WT + WT). All mice used in this study exhibited at least 80% donor engraftment at the time of the MCAO surgery, which was performed 8 weeks post-transplantation ( Figure 4C). There was no significant difference in infarct volume between the WT radiation naive and WT mice that were reconstituted with WT haemato-  Figure 4D). TTC staining demonstrated that infarcts in all groups were spread across both the cortex and striatum ( Figure 4A). The WT + Myd88 2/2 mice had a significantly larger infarct volume than the WT + WT mice. The role of Myd88 in the hematopoietic cells was further highlighted by the similar infarct volume between the WT + WT mice and the Myd88 2/2 + WT cells (Fig. 4A and B).

MyD88 signalling in the CNS contributes to haematopoietic cell infiltration
In the chimera studies the WT + WT and the Myd88 2/2 + WT display more cells positive for leukocyte esterase's ( Figure 5A). Mac-1 positive cells were identified in all chimera ipsilateral hemispheres, however there was significantly fewer Mac-1 positive cells in the WT + Myd88 2/2 compared to the WT + WT mice (WT + WT 9.9361.42 cells/frame Vs. WT + Myd88 2/2 4.861.31 cells/frame, P,0.05). There was no significant difference in the numbers of cells present between the WT + WT and the Myd88 2/2 + WT mice. However, the WT + WT Mac-1 positive cells were often co-localised with NeuN positive cells, whilst in the Myd88 2/2 this was rarely the case ( Figure 5B). The WT + Myd88 2/2 mice show very few Mac-1 positive cells, whilst the Myd88 2/2 + WT sections demonstrate Mac-1 positive cells that are rarely also NeuN positive. This data indicates that Myd88-dependent signalling is required in the CNS to attract infiltrating cells to the injury site. When the infarct volume is calculated from the TTC stained tissue the WT + Myd88 2/2 groups shows a larger infarct volume than the WT + WT groups, whilst there is no difference between the MyD88 2/2 + WT and the WT + WT infarct volume (n = 9). All chimera mice had above 80% donor cells present (C, n = 32-45) and there was no difference in percentage decrease in blood flow between the groups (D, n = 21-34 * Significantly different from WT + WT mice, P,0.05 error bars indicate 6SEM). doi:10.1371/journal.pone.0057948.g004

Discussion
As an adaptor protein for primarily pro-inflammatory pathways we hypothesized that Myd88-dependent signalling would contribute to brain damage following stroke, and that Myd88 2/2 mice would exhibit a smaller infarct following MCAO. In direct contrast, our study demonstrates that Myd88-dependent signalling in hematopoietic cells has a protective effect in stroke.
The major finding from this study is that hematopoietic cells exhibit a neuroprotective function after stroke that is mediated by MyD88. Other studies have demonstrated that invading cell numbers are protective [22,23,24] following stroke. These studies showed a decrease rather than a complete ablation of cellular infiltration. It is likely that infiltrating cells can have both beneficial and detrimental effects, and their relative number may contribute to this [25]. Interestingly, it has been reported that the removal of CCL-2 caused a decrease in the numbers of invading cells seen after stroke, which was mirrored by a corresponding increase in the activation of microglia, highlighting some form of coordinated response between endogenous and exogenous cell types in the brain [26]. One particular invading cell type that has been shown to be neuro-protective following stroke is the regulatory T cell (Treg). The release of IL-10 from T-regs has been shown to decrease inflammation and a subsequent decrease in infarct volume after stroke [27]. Reducing the infiltration of all cells in to the infarct area will also decrease the numbers of T-regs, and may over the longterm prolong inflammation. Modulating gross numbers of cells attracted to the infarct area may be less important than influencing the types of hematopoietic cells and at what time of reperfusion they arrive.
Our findings concerning infarct size differ from those of Yang et al., who found no difference in infarct size, oedema volume or neurological score between Myd88 2/2 and WT mice following MCAO [28]. However, in Yang's study the MCA was occluded for 6 hours, an unusually long occlusion for the intraluminal suture/filament method. This extended period reportedly caused the infarct to cover more than 55% of the prosencephalon, a very large infarct size. It is likely that in both WT and Myd88 -/mice the infarct had covered the entire ipsilteral hemisphere and therefore the damage was so severe that no difference could be detected. The model used in the current study utilized a 1-hour period of MCA occlusion. This is sufficient to produce a moderate infarct, with significant regions of undamaged and at risk tissue present in the ipsilateral cortex. We believe this approach allows the detection of more subtle changes in infarct size. Our data contrasts with that of Famakin et al., who found no difference in infarct size between Myd88 2/2 and WT mice in a permanently occlusive model [29]. This is perhaps unsurprising given that in the permanent occlusion model used by Famakin et al., there was no tissue reperfusion. Given that our results indicate a protective role for hematopoietic cells that requires Myd88 and is supported by our finding that WT mice with MyD88 2/2 haematopoietic cells suffered a larger infarct size than WT mice with WT haematopoietic cells, regardless of the genotype of the brain. This finding implicates the cells that invade from the periphery in the development of the infarct size following stroke, and shows that MyD88-dependent signalling in the invading cells in particular can control in the size of the infarct. Therefore in the model used by Famakin et al, a model that minimises invading cells access to the brain by excluding reperfusion, one would be expected to find little difference between the WT and MyD88 2/2 infarct volume.
Using a similar model of stroke others have demonstrated a primarily deleterious role for TLR signalling in stroke [9,30,31]. IL-1 signalling has also been shown through the use of release inhibition and receptor blockade to be similarly deleterious [32]. An explanation for the apparently disparate results between select TLR knockouts and the results reported here maybe found in the complexities of TLR and IL-1 signalling. TLR signalling is often described as either MyD88-dependent or MyD88 independent, a shift in the balance by removing MyD88, as opposed to removing or inhibiting just one TLR would affect all TLRs dependent on MyD88. In the inflammatory environment following cerebral infarction ligands for multiple TLRs are present. MyD88 dependent signalling following LPS stimulation has been shown to activate mitogen activated kinase phosphotases (MAPKP). These phosphotases are important negative regulators of many inflammatory pathways [33]. Therefore the absence of MyD88 dependent signalling may lead to less negative regulation of signal transduction pathways, and MyD88-independent pathways may lead to the activation of pro-inflammatory mediators unimpeded.
The role of both TLR and IL-1 signalling following stroke is complex as the response to CNS inflammation and not only involves resident cells but those from the periphery. These numerous cell types, belonging to disparate systems have variable expression of the receptors involved in Myd88-dependent signalling, adding to the complexity of the response. By isolating the expression of Myd88 to either the CNS or the hematopoietic cells this study has made it possible to understand the contributions of MyD88-dependent signalling pathways within a two separate systems. The activation and release of cytokines, attraction and infiltration of hematopoietic cells to the CNS and the subsequent infarct volume are all influenced by activation of MyD88dependent signalling. This study has demonstrated a role for MyD88-dpendent signalling in the response to MCAO further work to understand the location and particular pathways involved will significantly extend the field of neuro-immuno-biology. Figure S1 A. Lack of MyD88 causes no alteration in blood flow during MCAO. B. Generation of bone marrow chimeras has no effect on MCAO outcome. (DOCX) Table S1