|Home | About | Journals | Submit | Contact Us | Français|
Lipopolysaccharide (LPS) preconditioning provides neuroprotection against subsequent cerebral ischemic injury through activation of its receptor, Toll-like receptor 4 (TLR4). Paradoxically, TLR activation by endogenous ligands following ischemia worsens stroke damage. Here, we define a novel, protective role for TLRs following ischemia in the context of LPS preconditioning. Microarray analysis of brains collected 24 hours following stroke revealed a unique set of up-regulated genes in LPS pretreated animals. Promoter analysis of the unique gene set identified an over representation of Type I IFN associated transcriptional regulatory elements. This finding suggested the presence of Type I IFNs or interferon regulatory factors (IRFs), which up-regulate interferon-stimulated genes. Up-regulation of IFNβ was confirmed by real-time RT-PCR. Direct administration of IFNβ icv at the time of stroke was sufficient for neuroprotection. TLR4 can induce both IFNβ and interferon-stimulated genes through its adapter molecule TRIF and the IRF3 transcription factor. We show in oxygen glucose deprivation of cortical neurons, an in vitro model of stroke, that activation of TRIF following stroke reduces neuronal death. Further, mice lacking IRF3 were not protected by LPS preconditioning in our in vivo model. Our studies constitute the first demonstration of the neuroprotective capacity of TRIF/IRF3 signaling and suggest that interferon stimulated genes, whether induced by IFNβ or by enhanced TLR signaling to IRF3, are a potent means of protecting the brain against ischemic damage.
It is increasingly clear that Toll-like receptor (TLR) signaling worsens stroke injury. Mice lacking TLR2 or TLR4 are less susceptible to damage in multiple models of cerebral ischemia (Cao et al., 2007; Lehnardt et al., 2007; Ziegler et al., 2007). TLRs are expressed by microglia, astrocytes and endothelial cells and are activated by the damage-associated molecules HSP70 (TLR4) and HMGB1 (TLRs 2 and 4), present in the brain following ischemia (Kinouchi et al., 1993a; Kinouchi et al., 1993b; Faraco et al., 2007). TLR activation induces production of the inflammatory molecules TNFα, IL1β, and iNOS, and other cytotoxic mediators that increase tissue damage.
Though TLR4 activation following stroke exacerbates injury, activation of TLR4 prior to stroke protects the brain from damage. Systemic administration of lipopolysaccharide (LPS), a potent TLR4 ligand of bacterial origin, renders animals tolerant to injury in several models of cerebral ischemia (Tasaki et al., 1997; Rosenzweig et al., 2004; Hickey et al., 2007). LPS-induced tolerance to ischemic injury mirrors the phenomenon of LPS-induced tolerance to LPS. Initial exposure of macrophages to LPS induces pro-inflammatory TNFα, but upon subsequent exposure to LPS, TNFα production is reduced markedly due to disrupted signaling through the TLR4 adaptor molecule MyD88 (West and Heagy, 2002; Fan and Cook, 2004; Liew et al., 2005). Conversely, macrophages produce little IFNβ upon initial exposure to LPS, but enhance IFNβ production upon secondary exposure (Broad et al., 2007), suggesting up-regulated TLR4 signaling through the TRIF adaptor molecule. Thus, pretreatment with LPS may cause cells to switch their dominant TLR4 signaling pathway.
TLR4 signaling through TRIF induces IFNβ via activation of the transcription factor IRF3. IFNβ, administered systemically, reduces ischemic brain damage (Liu et al., 2002; Veldhuis et al., 2003), likely through activation of interferon-stimulated genes (ISGs). IRF3 itself may have similar neuroprotective effects. IRF3 binds to interferon stimulated response elements (ISREs) within gene promoters, increasing the expression of many ISGs to the same extent of that elicited by Type I IFNs (Nakaya et al., 2001). Hence, activation of IRF3 may independently result in protection from ischemic stroke. Thus, enhanced TLR4 signaling to TRIF-IRF3-IFNβ would be expected to contribute to neuroprotection.
We propose that pretreatment or preconditioning with LPS changes the cellular environment such that subsequent activation of TLR4 increases signaling via TRIF to IRF3 and up-regulates the neuroprotective cytokine IFNβ. Thus, in this way, LPS preconditioning may reprogram subsequent activation of TLR4 during ischemia, which leads to an increase in neuroprotective Type I IFN signaling. Here we provide evidence for such reprogramming and its neuroprotective consequences.
C57BL/6 mice (male, 8–12 weeks, approximately 25 grams) were purchased from the Jackson Laboratories (Sacramento, CA). IFNβ knockout mice were kindly provided by Dr. Leanderson of Lund University. IRF3 knockout mice were procured from RIKEN BioResource Center, Tsukuba, Japan. Both strains were backcrossed onto the C57Bl/6 background for at least 8 generations. All mice were housed in an American Association for Laboratory Animal Care-approved facility. Procedures were conducted according to Oregon Health and Science University, Institutional Animal Care and Use Committee, and National Institutes of Health guidelines.
Mice were given a 200 ul intraperitoneal injection of saline or LPS (0.2 –1.0 mg/kg; Escherichia coli serotype 0111:B4; (Cat# L2630, purified by phenol extraction, protein content <3%) Sigma, St. Louis, Mo). Each new lot of LPS was titrated to determine the optimal dose that confers neuroprotection in the particular strain of mouse being tested.
Mice were anesthetized with 4% halothane and subjected to MCAO using the monofilament suture method described previously (Stevens et al., 2002). Briefly, a silicone-coated 8-0 monofilament nylon surgical suture was threaded through the external carotid artery to the internal carotid artery to block the middle cerebral artery, and maintained intraluminally for 40, 45, or 60 min. Duration of occlusion was based on pilot studies performed to determine the time necessary to obtain an infarct size that is between 35–45% in the control groups of mice. It is well known that genetic background can influence ischemic outcome and thereby affect infarct size. The suture was then removed to restore blood flow. Cerebral blood flow (CBF) was monitored throughout surgery by laser Doppler flowmetry. The mean CBF during occlusion was between 10 and 17% of baseline in each of the studies presented. Mice that did not maintain a CBF drop within the norm of the group during the occlusion were excluded (<4% of all animals in the combined studies). Body temperature was maintained at 37°C with a thermostat-controlled heating pad. The survival rate for the MCAO procedure was >80%.
To visualize the region of infarction, 6 × 1 mm coronal midsections were placed in 1.5% 2,3,5 triphenyltetrazolium chloride (TTC) in 0.9% phosphate buffered saline and stained at 37°C for 15 min. The infarct size was determined from computer-scanned images of the hemispheres using NIH images analyses. To account for edema within the infarct region, infarct area for each section was computed indirectly as: 100 × (contralateral hemisphere area - area of live tissue on ipsilateral hemisphere)/(contralateral hemisphere area) (Swanson et al., 1990).
C57/BL6 mice were divided into 10 groups with 4 animals per group: Groups 1–3 received a saline injection and were sacrificed at 3, 24 and 72 hr respectively. Groups 4–6 received an LPS injection and were sacrificed at 3, 24 and 72 hr respectively. Group 7 and 8 received a saline injection followed 72 hr later with a 45 min MCAO. Group 9 and 10 received an LPS injection followed 72 hr later with a 45 min MCAO. Groups 7 and 9 were sacrificed at 3 hr following start of occlusion with Groups 8 and 10 sacrificed 24 hr following start of occlusion. At the time of sacrifice mice were anesthetized, then perfused with heparinized saline. One group (n=6) was included as unhandled controls. Under RNase-free conditions, a 1 mm section was removed (4 mm from rostral end) to determine the area of infarct based on TTC staining. The ipsilateral cortex region from the frontal 4 mm was isolated and snap frozen in liquid nitrogen.
Total RNA was isolated using the Qiagen RNeasy Lipid Mini Kit (Qiagen Inc., Valencia, CA). RNA from individual animals was hybridized to single arrays as described below.
Microarray assays were performed in the Affymetrix Microarray Core of the Oregon Health & Sciences University Gene Microarray Shared Resource. RNA samples were labeled using the NuGEN Ovation Biotin RNA Amplification and Labeling System_V1. Hybridization was performed as described in the Affymetrix technical manual (Affymetrix, Santa Clara, CA). Labeled RNA was hybridized to test arrays containing control probe sets and samples that did not meet empirically defined cutoffs within the core facility were remade. Quality-tested samples were hybridized to the MOE430 2.0 array. The array image was processed with Affymetrix GeneChip Operating Software (GCOS). Data was normalized using the Robust Multichip Average method (Irizarry et al., 2003). The normalized data was then analyzed using a two-way ANOVA model for each gene, using conditions and time as groups. Post hoc comparisons were made using the unhandled mice as a control group. P-values were adjusted for multiple comparisons using the Hochberg and Benjamini method (Hochberg and Benjamini, 1990). Genes were considered significantly regulated if the adjusted p value was less than 0.05 and the fold change in regulation was greater than or equal to 2.
Using the web based program: Promoter Analysis and Interaction Network Toolset (PAINT) version 3.5 (Vadigepalli et al., 2003), we examined the predicted regulatory elements associated with the unique gene regulation identified by microarray. In brief, using PAINT we obtained the 5000 bp upstream sequence for the transcripts represented on the MOE430 Affymetrix gene chip (33,635 transcripts were identified with 5000 bp of upstream sequence). PAINT identified putative transcription factor binding sequences (TREs) in these upstream sequences using the TRANSFAC PRO database version 10.4. This pool of genes and identified TREs was used as our reference comparison group. The statistical component of PAINT (FDR adjusted p value set at ≤0.2) was used to determine the over represented TREs in individual gene clusters compared to the reference comparison group (i.e. uniquely expressed genes in LPS preconditioned mice compared to 33,635 member reference group).
rmIFNβ (Cell Sciences, Canton, MA) or vehicle (artificial cerebral spinal fluid (aCSF)) was injected into the left lateral ventricle as previously described (Meller et al., 2005). Injections (1 ul) of either rmIFNβ (200U) or aCSF were administered immediately before and after surgery (60 min MCAO). Infarct volume was measured 24 hr following stroke.
RNA was treated with DNase and transcribed into cDNA using the Omniscript RT Kit (Qiagen). Real-time PCR (RT-PCR) reactions were performed using TaqMan PCR Master Mix (Applied Biosystems, Foster City, CA). For IFNβ TaqMan Gene Expression Assay Mix for mouse IFNβ was used (ABI # Mm00439546_S1). Primers and probe for β-Actin were obtained from Integrated DNA Technologies: forward: 5'-AGAGGGAAATCGTGCGTGAC-3'; reverse: 5'-CAATAGTGATGACCTGGCCGT-3'; probe: CACTGCCGCATCCTCTTCCTCCC. Samples were run on an ABI-prism 7700 (Applied Biosystems). Results were analyzed using ABI sequence detection software. The relative quantitation of IFNβ was determined using the comparative CT method (2−ΔΔCT) described in ABI User Bulletin #2. Results were normalized to β-actin and presented relative to unhandled mice. All reactions were performed in triplicate.
Primary mouse mixed cortical cultures were prepared from E15-E17 mouse fetuses. Cortices were dissected and dissociated with Trypsin-EDTA (Gibco) and plated at a density of 4.5 × 105 cells/ml onto coverslips coated with poly-L-Ornithine (15 mg/L). Cells were cultured in Neurobasal media (containing 4.5 g/L glucose; supplemented with Glutamax and B27-AO; Gibco) for 5 days prior to each experiment. Cultures consisted of ~60% neurons (range 53–66%) as determined by staining for NeuN (Chemicon), with less than 5% astrocytes (GFAP+; Sigma) and less than 5% microglia (tomato lectin+; Vector Labs). Oxygen-glucose-deprivation (OGD) was performed by removal of the culture medium and replacement with D-PBS (Gibco) followed by incubation in an anaerobic atmosphere of 85% N2, 10% CO2, 5% H2 at 37°C for 3 h. The anaerobic conditions within the chamber were monitored using an electronic oxygen/hydrogen analyzer (Coy Laboratories). OGD was terminated by replacement of the exposure medium with Neurobasal medium (containing 4.5 g/L glucose; supplemented with Glutamax and B27-AO) and return of the cells to a normoxic incubator. Control plates were kept in the normoxic incubator during the OGD interval.
Cell death in vitro was examined 24 hr following OGD by means of fluorescent, cell-permeable, DNA-binding dyes: propidium iodide (PI), as an indicator of cell death, and 4',6-diamidino-2-phenylindole (DAPI), as an indicator of the total number of cells. Coverslips were incubated with PI (1.5ug/ml, Sigma) for 5 min, washed with PBS and fixed for 30 min in 10% formalin. Coverslips were mounted on slides with Fluoromount-G mounting medium containing DAPI (SouthernBiotech). Stained cells were visualized with a fluorescent microscope (Leica GMBH) and analyzed using Metmorph7 software (Molecular Devices Corp., Downington, PA). The number of PI and DAPI stained cells were counted in two random fields of view on each coverslip, and percent death was calculated as mean (PI)/(DAPI) × 100 per field of view. Each treatment was performed with triplicate coverslips within an experiment and the entire experiment was repeated three or more times.
As we have previously shown, systemic administration of LPS (0.2 mg/kg) given 3 days prior to MCAO substantially attenuates ischemic damage (Rosenzweig et al., 2004; Rosenzweig et al., 2007). To begin to elucidate possible mechanisms of neuroprotection we isolated RNA from the cortex of LPS treated and control mice at time points prior to MCAO. Using Affymetrix oligonucleotide microarrays we identified 263 genes (228 increased, 35 decreased) significantly regulated 3 hr following LPS injection and 176 genes (174 increased, 2 decreased) at 24 hr post LPS treatment. However, within 72 hr following LPS administration, most of the genomic changes had subsided to baseline with the exception of 5 differentially regulated genes that remained increased (Fig. 1A; Table 1). The saline treated controls showed no statistically significant gene regulation at any time point when compared to the unhandled baseline group.
We determined putative functions for the modulated genes using the Affymetrix Netaffx website, the Stanford-Online Universal Resource for Clones and ESTs website (http://genomewww5.stanford.edu/cgi-bin/source/sourceSearch) and the published literature. A large fraction (~50%) of the genes regulated at both 3 and 24 hr are involved in the defense/inflammation response, which includes genes associated with both the innate and adaptive immune response as well as genes involved in stress and wound responses (Fig. 1B). Thus a low dose of LPS given systemically induces genomic regulation of the inflammatory response in the brain as early as 3 hr post administration, which is resolved at the genomic level of RNA expression by 72 hr.
We compared the transcriptional response to MCAO in LPS preconditioned and control mice. The majority of genes regulated (~70%) at both 3 and 24 hr post MCAO were independent of the preconditioning stimulus. However a significant number of genes were uniquely regulated based on LPS preconditioning. At 3 hr post MCAO 66 genes (29%) were unique to the saline pretreated animals while 26 genes (14%) were only seen in mice preconditioned with LPS (Fig. 2; purple and blue regions respectively). Only one of the 26 genes unique to the LPS preconditioned mice was regulated at the time of ischemia. Saa3 an acute phase responder was increased 6 fold over unhandled controls at 72 hr post LPS injection. At 3 hr following MCAO Saa3 remained increased (5.5 fold) in mice preconditioned with LPS suggesting that this increased level was due to the preconditioning stimulus that had occurred 3 days earlier. The general absence of unique gene regulation just prior to ischemia is in contrast to the presence of unique gene regulation that occurs following ischemia and suggests that prior LPS preconditioning modifies the genomic response to ischemia.
The distinct responses to stroke are also evident at the 24 hr time point where 231 genes (29%) are uniquely regulated in the saline pretreated animals and 176 genes (23%) are unique to the LPS preconditioned mice (Fig 2; gold and green regions respectively). Table 2 shows the regulation of each of these 176 genes prior to and following MCAO in LPS preconditioned mice. These genes appear to be regulated in response to the MCAO, as there is little (less than 2-fold) or no regulation for each identified gene at the time of stroke (72 hr post injection). Hence, following stroke, LPS preconditioning induces the regulation of a unique set of genes as early as 3 hr following stroke that is not evident in saline pretreated mice. These findings suggest LPS preconditioning reprograms the genomic response to stroke in LPS preconditioned mice.
We identified transcriptional regulatory elements (TREs) associated with the unique gene regulation detected in the LPS and saline preconditioned animals using the web-based program Promoter Analysis and Interaction Network Toolset (PAINT) version 3.5. We compared the TREs identified in the cluster of genes uniquely increased in LPS preconditioned mice 24 hr following stroke (158 genes, Fig. 2) to a reference cluster consisting of ~33,000 transcripts from the MOE430 gene chip. This allowed the determination of over-represented TREs associated with the genes in the preconditioned cluster. We performed the same comparison using the cluster of genes uniquely increased in the saline pretreated mice 24 hr following stroke (128 genes, Fig. 2). Analysis of the LPS preconditioned group identified 14 TREs with an adjusted p value <0.2 while the saline pretreated cluster revealed only 5 over-represented TREs (Table 3). Five of the 14 identified TREs in the LPS preconditioned cluster are interferon-associated (IRF [V$IRF_Q6 and V$IRF_Q6_01], IRF8, ISRE, IRF7) and 2 are NFkB components (cRel, NF-kappaBp65). A network depiction of interactions between the identified TREs and the genes in the LPS preconditioned cluster is displayed in Figure 3. The interferon-associated TREs (in red) are linked to a substantial number of the genes shown (60%; 76 of 127). In fact, a large number of the genes with identified IFN TREs have been reported in the literature to be induced by type I interferons (Fig. 4; red asterisks). NFkB regulatory elements were also overrepresented, these sequences were found on 54 of the 127 genes (42%; Fig. 3, blue) with 30 of those also sharing IFN TREs (Fig. 4). It has recently been reported that cRel directly binds to the promoter and regulates several ISG genes upon IFN stimulation (Wei et al., 2008). Thus cRel/NFkB may play an integral role in the interferon fingerprint associated with LPS preconditioning. Collectively, 79% (100/127) of the genes induced 24 hr following stroke in LPS preconditioned mice contain a regulatory sequence for IFN or NFkB (Fig. 4). The predominance of the type I interferon signature prompted us to pursue the possible role of enhanced type I interferon signaling in LPS -induced neuroprotection.
The increase in interferon inducible genes and over-representation of interferon-associated TREs suggested that IFNβ may be present in the brain cortex following stroke in LPS preconditioned mice. Using real time PCR we examined the levels of IFNβ transcript in the brain following stroke in LPS preconditioned and saline treated mice. IFNβ levels were increased following stroke in the preconditioned and non-preconditioned mice compared to unhandled controls (Fig. 5). However, levels in LPS preconditioned mice were 9-fold higher at 3 hr (LPS treated 59.4 ± 22 vs saline treated 6.7 ± 3; p<0.001) and 3.5-fold higher at 24 hr (LPS treated 45.3 ± 23 vs saline treated 11.7 ± 6; p<0.001) post stroke. We examined levels of IFNβ just prior to MCAO (72 hrs post injection) to confirm that the increase in IFNβ following stroke was independent of any residual increase of IFN β resultant from the preconditioning LPS injection. Levels of IFNβ in LPS and saline treated mice were statistically equivalent to unhandled controls (1.49 ± 1.4 and 0.74 ± 0.6 respectively; data not shown). Thus, following stroke, mice preconditioned with LPS mount a more robust IFNβ response to ischemic injury.
IFNβ following stroke in non-preconditioned mice was increased 6.7-fold over unhandled mice (Fig. 5). Although not as robust as in LPS preconditioned mice, this does suggest that IFNβ may play a protective role in the endogenous response to stroke. To determine whether the increase following stroke alone could be protective, IFNβ knockout mice were subjected to 40 min MCAO followed by 72 hr of reperfusion. Wild type and IFNβ knockout mice displayed infarcts of similar size (38.5 ± 2% vs 39.5 ± 1%; p=0.7). Thus IFNβ itself does not appear to play a critical role in the brain's usual response to ischemia.
To determine whether the more robust increase in IFNβ (9 fold; Fig. 5) following stroke in LPS preconditioned mice would effect ischemia, we tested the effect of exogenous IFNβ administration in the brain following MCAO. We injected C57BL/6 mice i.c.v. with recombinant mouse IFNβ immediately prior to and following MCAO and measured infarct size 24 hr later. Animals treated with IFNβ showed a significant reduction in infarct volume versus vehicle treated mice (31.9 ± 4% vs. 49.4 ± 2%; p<0.001). This result supports the notion that increased expression of IFNβ within the brain would confer protection from ischemic injury.
The increase in IFNβ and type I interferon associated genes in response to stroke in the LPS preconditioned mice mirrors the secondary response to LPS in classic endotoxin tolerance, and supports a possible reprogramming of the TLR response to stroke, resulting in a TRIF mediated event. To determine whether signaling via TRIF following stroke would induce protection we used the synthetic TLR3 ligand, Poly I:C (InvivoGen), which signals exclusively via a TRIF dependent pathway. We reasoned that activation of this pathway at the time of ischemia would provide acute protection from ischemic damage. To test this, mixed cortical cultures from mice were exposed to oxygen glucose deprivation (OGD) for 3 hr followed by treatment with varying doses of Poly I:C and subsequently returned to normoxic coniditions. Twenty-four hours later cell death was determined. Acute Poly I:C treatment following OGD significantly reduced OGD-mediated cell death at all 3 doses tested (Fig. 6). Thus signaling via TRIF following ischemia provides protection against damage in vitro, which suggests that TRIF mediated signaling in the brain following stroke could reduce ischemic injury.
To further explore the TRIF/IRF3 pathway we tested whether IRF3 is a critical effector of LPS-induced ischemic protection. First we determined whether IRF3 is involved in the brain's endogenous response to stroke. IRF3 knockout mice were subjected to 40 min MCAO followed by 72 hr of reperfusion. IRF3 knockout mice displayed infarcts of similar size to wild type mice (41.2 ± 5% vs 43.4 ± 4%; p=0.8). Thus IRF3 does not play a critical role in the brain's usual response to ischemia. Next we determined whether IRF3 is a required effector of LPS-induced tolerance to ischemia. IRF3 knockout mice were pretreated with LPS (0.4 mg/kg) 72 hr prior to 40 min MCAO, and sacrificed 24 hr later. Figure 7 shows that IRF3 knockout mice fail to be preconditioned with LPS (17.3% vs 53.2% reduction). Hence, IRF3 is required for the protective effects of LPS pretreatment.
We propose a molecular model of LPS-induced neuroprotection from ischemic injury wherein systemic LPS preconditioning reprograms TLR4 signaling in response to stroke, directing it towards a neuroprotective pathway. Administration of systemic LPS induces an early inflammatory genomic response in the brain that has receded by 72 hr. However, the response to stroke in these LPS preconditioned mice is altered and a new pattern of gene regulation is induced as early as 3 hours following exposure to ischemia. The genomic changes in the brain in response to LPS suggest a possible activation of brain TLRs prior to the stroke which go on to respond with an altered signaling pathway following activation by the stroke event. The fact that IFN regulatory elements are over represented among the unique genes induced following stroke in LPS preconditioned animals suggests that TLR signaling in this setting may be altered. To pursue this we examined the possibility that increased signaling via the TRIF dependent pathway could serve as a neuroprotective mechanism associated with LPS preconditioning.
The induction of IFNβ is a hallmark of the TRIF dependent TLR4 cascade (Biswas et al., 2007). In accordance with this we found that LPS preconditioning increased IFNβ within the brain 3 and 24 hours following stroke. We then tested whether the increase in IFNβ was required to confer neuroprotection. Mice lacking IFNβ displayed infarcts of similar size to wild type mice, suggesting that endogenous IFNβ does not protect the brain from ischemic injury. However, exogenous administration of IFNβ i.c.v. at the time of stroke conferred significant protection against ischemic damage, indicating that local up-regulation of this cytokine may be neuroprotective. Our in vitro studies using modeled ischemia showed that activation of the TRIF-IRF3 pathway with the TLR3 ligand, Poly I:C, following OGD reduced neuronal death, thus supporting the notion that TRIF dependent signaling can provide protection to ischemic injury.
The role of the TRIF-IRF3 cascade in the brain's natural response to stroke appears relatively minor as mice lacking IRF3 displayed infarcts of similar size to wild type mice. However, there appears to be a critical role for IRF3 in LPS preconditioning because IRF3-deficient mice could not be protected from ischemic injury in the setting of LPS preconditioning. As these mice are deficient in IRF3 during the induction of tolerance as well as following the ischemic event we cannot rule out a requirement for IRF3 during the induction phase as well as the resolution phase. In preliminary studies, we have found that Poly I:C preconditioning is protective in stroke (unpublished observation) which suggests that TRIF signaling may be sufficient for the induction of tolerance. In models of endotoxin-endotoxin (LPS) tolerance where prior exposure to low dose endotoxin provides protection against subsequent high dose exposure, there is substantial support for the notion that TRIF/IRF3 signaling is required. Biswas and colleagues showed that endotoxin tolerance was induced in MyD88 deficient mice but not in TRIF or IRF3 deficient mice (Biswas et al., 2007). Thus, although TRIF dependent signaling is likely involved in the induction of tolerance, the enhancement of IRF3 dependent genes following stroke in LPS preconditioned mice and the protective response elucidated with Poly I:C following OGD supports a role for the TRIF-IRF3 pathway in the protective phenotype as well.
Our data support a model of redirected TLR4 signaling that resembles endotoxin tolerance. Cells made tolerant to endotoxin (LPS) are known to suppress the pro-inflammatory MyD88-TNFα pathway by up-regulating pathway inhibitors, namely IRAK-M, Tollip, Ship1 and Trim30α, among others (Lang and Mansell, 2007; Shi et al., 2008), which results in decreased inflammatory cytokine responses upon secondary exposure to TLR4 ligands. Inhibition of these pathways shunts subsequent TLR4 signaling down the TRIF-IRF3-IFNβ pathway and results in enhanced production of IFNβ (Bagchi et al., 2007). Similarly, LPS preconditioning may up-regulate inflammatory pathway inhibitors in the brain that shunt subsequent TLR signaling down the TRIF-IRF3-IFNβ pathway. Thus, in the setting of ischemia, release of endogenous TLR ligands would be expected to lead to TLR signaling that is shunted down the TRIF-IRF3-IFNβ pathway and result in up-regulation of IFNβ.
Our data support a model of redirected TLR4 signaling following stroke in preconditioned animals. Unlike control animals, LPS preconditioned mice demonstrate a significant up-regulation of IFNβ and IFN-associated genes following stroke, which, based on promoter region analyses are likely produced via the TLR4-TRIF-IRF3 pathway. Hirotani and colleagues have shown in endotoxin tolerance, using TRIF deficient mice, that induction of type I interferon associated genes following a secondary challenge is exclusively dependent on TRIF (Hirotani et al., 2005). Together, these data suggest that the Type I IFN “fingerprint” is generated downstream of TLR4-TRIF-IRF3 and supports the concept of TLR4 reprogramming.
The potential for the reprogrammed TLR response to be protective is evinced by the TRIF mediated neuronal protection in our in vitro model and via the protective effect of IFN β directly administered in the brain. Others have shown that systemic administration of IFN β reduces tissue damage in both a rat and rabbit model of ischemic stroke (Liu et al., 2002; Veldhuis et al., 2003). In these studies, the protective effects of IFN β may have been mediated through leakage across the BBB following stroke wherein IFN β may have then acted directly on the brain parenchyma. A recent study by Maier and colleagues in which BBB integrity is believed to be preserved failed to show attenuation of ischemic brain injury following systemic administration of IFN β (Maier et al., 2006). Thus as with our data, these results support the notion that the neuroprotective effects of IFN β occur centrally.
We show that IRF3 is required for LPS induced neuroprotection and that this is most likely through its induction of IFN β and other ISGs. Two potential ISG modulators of the protective effect, Trim30 and Ifit1--were identified in our microarray analysis. Trim30 has been shown to negatively regulate LPS-induced TNFα and IL6 expression via inhibition of TLR4 induced NFκB activation (Shi et al., 2008). In the brain we found increased mRNA levels of Trim30 at 3 and 24hr following LPS treatment and at 24hr following stroke in LPS preconditioned mice. This induction could play a role in the suppression of inflammatory cytokines in the brain following stroke. Although Shi and colleagues reported that Trim30 expression depends on NFκB (Shi et al., 2008), the promoter sequence also contains an ISRE site (Fig. 4). In addition, we find induction of Ifit1 following LPS administration and again 24hr following stroke in LPS preconditioned mice. The closely related gene Ifit2 has recently been shown to suppress the LPS mediated induction of TNFα and IL6 (Berchtold et al., 2008), although it has not been directly linked to suppression of TLR signaling. The induction of these two ISG genes and IFNβ in LPS preconditioned mice support the hypothesis of a reprogrammed TLR response to stroke resulting in a neuroprotective state.
Our data suggest that systemic administration of LPS reprograms TLR4-expressing cells within the brain. TLR4 is widely expressed in the brain (Lehnardt et al., 2002; Olson and Miller, 2004; Chakravarty and Herkenham, 2005) and many studies have shown that peripheral LPS induces a pro-inflammatory response within the brain (Chen et al., 2005; Qin et al., 2008). However, it is unclear whether LPS crosses the BBB and/or whether it induces peripheral cytokines, which in turn, cross into the brain. Recent evidence suggests that systemic LPS elicits TLR4 signaling in the brain independent of peripheral cytokine responses (Chakravarty and Herkenham, 2005; Gosselin and Rivest, 2008). However, other researchers have failed to find LPS within the brain parenchyma following systemic administration (Singh and Jiang, 2004). It is clear that LPS binds to cerebral endothelial cells (Singh and Jiang, 2004; Verma et al., 2006). As these cells are an interface between the systemic circulation and the brain parenchyma, they may help integrate information from both compartments. Hence, reprogramming of TLR4 may occur within the cerebral endothelium.
In summary, we have shown that LPS preconditioning reprograms the cellular response to stroke and causes a Type I IFN response, with a critical and protective role for IRF3. These reprogramming events may exemplify endogenous processes that protect the brain against further injury and suggest that LPS preconditioning fundamentally changes the cellular response to stroke. This is the first demonstration that a preconditioning stimulus results in an interferon “fingerprint” after the ischemic event and the first report of a neuroprotective role for TRIF-IRF3 signaling following injury.
The authors would like to thank Jo-Lynn Boule, Eric Tobar, Delfina Homen, Tao Yang and Dr. Nikola Lessov for excellent technical support, and Dr. Roger Simon for constructive discussion. Microarray assays were performed in the Affymetrix Microarray Core of the OHSU Gene Microarray Shared Resource. This work was supported by National Institutes of Health grant R01 NS050567 (MS-P).