|Home | About | Journals | Submit | Contact Us | Français|
After stroke, the blood–brain barrier is transiently disrupted, allowing leukocytes to enter the brain and brain antigens to enter the peripheral circulation. This encounter of normally sequestered brain antigens by the systemic immune system could therefore present an opportunity for an autoimmune response to brain to occur after stroke. In this study, we assessed the immune response to myelin basic protein (MBP) in animals subjected to middle cerebral artery occlusion (MCAO). Some animals received an intraperitoneal injection of lipopolysaccharide (LPS; 1 mg/kg) at reperfusion to stimulate a systemic inflammatory response. At 1 month after MCAO, animals exposed to LPS were more likely to be sensitized to MBP (66.7% versus 22.2%; P=0.007) and had more profound and persistent neurologic deficits than non-LPS-treated animals. Exposure to LPS was associated with increased expression of the costimulatory molecule B7.1 early after stroke onset (P=0.009) and increased brain atrophy 1 month after MCAO (P=0.03). These data suggest that animals subjected to a systemic inflammatory insult at the time of stroke are predisposed to develop an autoimmune response to brain, and that this response is associated with worse outcome. These data may partially explain why patients who become infected after stroke experience increased morbidity.
After stroke, the integrity of the blood–brain barrier (BBB) is compromised and cells of the immune system encounter novel central nervous system (CNS) antigens in both the brain and in the systemic circulation. The concentration of circulating CNS antigens after stroke reflects the severity of cell injury and predicts the outcome (Herrmann et al, 2000; Wunderlich et al, 1999). If these antigens were presented to lymphocytes in the proper context, an autoimmune response to brain could develop. Indeed, CNS autoreactive cells and CNS-specific immunoglobulins are seen in patients with a history of stroke (Bornstein et al, 2001; Dambinova et al, 2003; Kallen et al, 1977; Wang et al, 1992; Youngchaiyud et al, 1974). The clinical consequences of this CNS autoimmune response are unknown.
Infection in the poststroke period is associated with poor outcome (Grau et al, 1999; Johnston et al, 1998). There are several plausible mechanisms by which infection could worsen ischemic brain injury, but definitive mechanistic data are lacking. Encounter of antigen by the immune system can lead to a Th1 immune response, a Th2/Th3 immune response, or no immune response. A Th1 immune response (or sensitization) is characterized by secretion of proinflammatory cytokines (interleukin (IL)-2, IL-12, tumor necrosis factor (TNF)-α, interferon (IFN)-γ) upon reencounter with the sensitizing antigen; a Th2/Th3 immune response (or tolerance) is characterized by secretion of immunomodulatory cytokines (IL-4, IL-10, transforming growth factor (TGF)-β1) on re-encounter with the tolerizing antigen (Romagnani, 2000; Weiner, 2001). The microenvironment of the brain on antigen encounter influences the type of response generated, because lymphocyte activation requires that T cells encounter antigen in the context of the major histocompatibility class II (MHC II) molecule and receive an additional costimulatory signal (Croft and Dubey, 1997). In brain, microglia can express MHC II and act as antigen-presenting cells (APCs) (Hickey and Kimura, 1988; Kato et al, 1996). The B7 molecules, among the most important costimulatory ligands, are also expressed on microglia and deliver a costimulatory signal to lymphocytes through CD28 (Croft and Dubey, 1997). Stimulation of CD28 by B7.1 results in a Th1 immune response; stimulation of CD28 by B7.2, which is preferentially expressed in both the normal and infarcted brain, generally results in a Th2/Th3 immune response (Bechmann et al, 2001; Dangond et al, 1997; Kuchroo et al, 1995; Racke et al, 1995).
Activation of the immune system during a systemic inflammatory response alters the microenvironment of the brain in a way that could promote sensitization to brain antigens during episodes of cerebral ischemia (Kissler et al, 2001). Since Gram-negative bacteria are the predominant organisms causing infection after stroke, we modeled the systemic inflammatory response to infection with lipopolysaccharide (LPS), a component of the Gram-negative bacterial cell wall (Hilker et al, 2003; Puri et al, 2002). The aim of these experiments was to determine whether LPS-induced inflammation during stroke would increase the likelihood of developing an immune response to brain antigens.
Experiments were approved by the Institution's Animal Care and Use Committee. Male Lewis rats (250 to 300 g) were used for all studies.
Anesthesia was induced with 5% and maintained with 1.5% halothane. After midline neck incision, the right common carotid, internal carotid, and pterygopalantine arteries were ligated. A monofilament suture (4.0) was inserted into the common carotid artery and advanced into the internal carotid artery (Zea-Longa et al, 1989). Animals were maintained at normothermia during surgery. Reperfusion was performed 3h after MCAO and animals killed at various time points thereafter. Rectal temperature and body weight were assessed at set time intervals. In sham-operated animals, the suture was inserted into the carotid but not advanced.
In a subset of animals, LPS (1 mg/kg) (Sigma, St Louis, MO, USA; serotype 026:B6) was injected intraperitoneally at reperfusion (3h after MCAO or sham surgery). Lipopolysaccharide-treated animals are henceforth referred to as LPS(+), non-LPS-treated animals as LPS(–).
Mononuclear cells (MNCs) were isolated from the brain and spleen using previously described methods (Becker et al, 2003).
MNCs were cultured (1 × 105 cells/well) for 48 h in 96-well plates (MultiScreen®-IP; Millipore) in media alone or in media supplemented with bovine myelin basic protein (MBP) (25 μg/mL). Experiments were performed in triplicate. Interferon-γ or TGF-β1 capture antibodies were used and the reaction product developed with alkaline phosphatase. Spots were independently counted under a dissecting microscope by two individuals masked to treatment status. The difference between the number of MBP-stimulated and unstimulated spots was considered indicative of an antigen-specific response. Secretion of IFN-γ was used to indicate sensitization (Th1 response); secretion of TGF-β1 was used to indicate tolerance (Th2/Th3 response). Results are expressed as the number of MBP-specific cells per 1 × 105 total MNCs. If the ratio of MBP-specific IFN-γ-secreting cells to MBP-specific TGF-β1-secreting cells was greater than two, the animal was considered sensitized to MBP, or to be Th1(+).
On killing, brains were frozen and stored at –80°C. To determine the infarct size, brains were sectioned (10 μm) at 6 predetermined levels (bregma +2.40, +1.00, –0.40, –1.80, –3.20, and –4.40) and stained with cresyl violet; infarct volume was calculated and corrected for edema (Swanson et al, 1990). Quantitative immunocytochemistry (ICC) was performed on sections at the level of the anterior commissure. Sections were fixed in acetone and methanol and stained with antibodies to CD4 (clone W3/25), CD8 (clone MRC OX-8), B7.1 (CD80; clone 3H5), B7.2 (CD86; clone 24F), MHC I (OX-18), and MHC II (OX-6). OX-18 and OX-6 antibodies were obtained from Pharmingen, San Diego, CA, USA; all other antibodies were obtained from Serotec, Raleigh, NC, USA. Sections were developed using peroxidase and 3,3′-diaminobenzidine (Vector) and counterstained with cresyl violet. The NeuroTACS™ kit (R&D Systems) was used to detect apoptosis. Sections were examined at × 100; the numbers of labeled cells within six different high-power fields in six different brain regions (Figure 1) were counted by two independent investigators masked to treatment status. The phenotype of cells expressing B7.1 was determined using double-label ICC for astrocytes (glial fibrillary acidic protein (GFAP); Sigma), microglia (OX-42; Serotec), and neurons (Neurotrace Nissl stain; Molecular Probes, Carlsbad, CA, USA).
Responses over time were compared using one-way analysis of variance. Data from different treatment groups were compared using the t-test or the Mann–Whitney U-test, and expressed as mean±standard deviation (s.d.) or median±s.d. as appropriate. Categorical data were evaluated using the χ2-test statistic. Correlations were performed using Spearman's rho. Significance was set at P≤0.05.
Lymphocytes specific for MBP were enriched in brain as very few MBP-reactive cells were seen in the spleen (data not shown). Based on the cytokine profile of MBP-reactive MNCs from the ischemic hemisphere of brain 1 month after MCAO, LPS(+) animals were more likely to develop a Th1 response to MBP (12/18 (66.7%)) than LPS(–) animals (4/18 (22.2%); P=0.007). When viewed as a continuous variable, the ratio of IFN-γ:TGF-β1-secreting MBP-reactive cells in brain 1 month after MCAO was higher in LPS(+) animals (5.0±5.0) than in LPS(–) animals (1.4±1.8; P=0.008), sham-operated LPS(–) controls (1.5±1.1; P=0.006), and sham-operated LPS(+) controls (1.7±1.1; P=0.02) (Figure 2A). In fact, LPS(–) animals were more likely to develop a Th2/Th3-type response to MBP (Figure 2B). The number of MNCs from the brain of LPS(+) animals secreting IFN-γ in an MBP-specific fashion 1 month after MCAO was also greater than in sham-operated LPS(–) controls (5.63±4.63 per 1 × 105 cells versus 2.41±2.92 per 1 × 105 cells; P=0.02), sham-operated LPS(+) controls (2.08±1.54 per 1 × 105 cells; P=0.01), and LPS(–) ischemic animals at the same time point (2.36±2.81 per 1 × 105 cells; P=0.02) (Figure 2C). Conversely, 1 month after MCAO there was a trend towards increased numbers of TGF-β1-secreting MBP-specific cells in the brains of LPS(–) compared with LPS(+) animals (3.35±3.96 per 1 × 105 cells versus 1.46±2.00 per 1 × 105 cells; P=0.08) (Figure 2D). Lipopolysaccharide administration did not predispose sham-operated animals to become either sensitized or tolerized to MBP (data not shown).
There was no mortality in LPS(+) or LPS(–) shamoperated animals and no difference in mortality among LPS(+) and LPS(–) animals undergoing MCAO: 7/38 (18.4%) versus 8/81 (9.9%), respectively. Death occurred almost exclusively in the first 24h after MCAO. All animals undergoing MCAO developed hyperthermia; the highest recorded temperatures occurred 6h after stroke onset: 38.4°C in LPS(+) and 38.6°C in LPS(–) animals. The temperatures of LPS(+) and LPS(–) animals differed only at 72h after MCAO (36.9°C±0.8°C versus 37.6°C±0.6°C, respectively; P=0.04). Similarly, the temperatures of Th1(+) animals, irrespective of LPS treatment status, were less than that of Th1(–) animals 72h after MCAO (36.2°C±0.6°C versus 37.8°C±0.4°C; P=0.006).
Neurologic scores were higher (i.e. neurologic dysfunction greater) in both LPS(+) and Th1(+) animals early after MCAO (Figures 3A and 3B). The degree of sensitization to MBP at 1 month was correlated to the neurologic deficit 72h and 7 days after MCAO (Figures 3C and 3D). Animals subjected to MCAO performed worse on the ‘sticky tape test’, a measure of sensation and dexterity. At 1 month after MCAO, LPS(+) animals took longer to recognize the presence of the tape (34.0±44.2 versus 6.5±62.8 secs; P<0.05) and tended to take longer to remove the tape (600±196 versus 400±251 secs; P=0.08).
All animals experienced a significant decrease in body weight after MCAO; the degree of weight loss did not differ between LPS(+) and LPS(–) animals. Animals that developed a Th1 response to MBP, however, were less likely to regain weight 1 month after MCAO; Th1(+) animals lost 5.9%±5.4% of their body weight in the month after MCAO, while Th1(–) animals gained 2.7%±10.4% of their initial body weight (P=0.01) (Figures 3E and 3F). The difference in body weight between Th1(+) and Th1(–) animals was apparent by 96h after MCAO (P=0.04).
More CD4+ cells were present in the ischemic hemisphere of LPS(+) animals 24h after MCAO (12.4±11.1 versus 3.4±2.6; P=0.002) (Figures 4A and 4B). CD8+ cells, however, were present in similar numbers in LPS(+) and LPS(–) animals until 1 month after MCAO, when the number of CD8+ cells was greater in both the ischemic (16.4±7.7 versus 4.2±4.3; P=0.02) and nonischemic (9.9±4.6 versus 4.2±4.3; P=0.005) hemispheres of LPS(+) animals (Figures 4C and 4D). More CD8+ cells were seen in LPS(+) animals undergoing MCAO also 1 month after MCAO than in sham-operated animals 1 month after LPS administration (P=0.02).
By 24h after MCAO, more cells expressed B7.1 in both the ischemic and nonischemic hemispheres of LPS(+) animals (21.4±19.0 versus 8.0±7.4; P=0.01 and 11.0±9.5 versus 4.8±4.1; P=0.02, respectively). This difference persisted to 72h (Figures 5A and 5B). Lipopolysaccharide(+) animals undergoing MCAO also expressed more B7.1 than sham-operated animals 24h after LPS administration (P=0.003). Double-label studies confirmed that cells expressing B7.1 were microglia (Figure 5C). B7.2 expression, on the other hand, was increased in the nonischemic hemispheres of LPS(–) animals 720h after MCAO (11.6±12.9 versus 3.0±4.5; P=0.02) (Figures 5D and 5E). Middle cerebral artery occlusion did not alter the expression of B7.2 among LPS-treated animals, although MCAO was associated with an increase in B7.2 expression 720h after MCAO in LPS(–) animals (P=0.01). More cells expressed MHC I and MHC II in LPS(+) animals early after MCAO (Figures 5E and 5F). The morphology of cells expressing MHC II was largely that of infiltrating leukocytes, although some cells with the morphology of microglia also appeared to express MHC II (Figure 5G).
Infarct volume did not differ between LPS(+) and LPS(–) animals at any time point (Figure 6A), and the amount of cerebral edema (volume increase of ischemic compared with nonischemic hemisphere) was similar in LPS(+) and LPS(–) animals from 6 to 72h after MCAO. At 1 month after MCAO, however, LPS(+) animals evidenced increased atrophy of the ischemic hemisphere (total volume of ischemic hemisphere/total volume of nonischemic hemisphere; 80.9%±6.8% versus 89.8%±6.6%; P<0.05) (Figure 6B). The hemispheric atrophy might be a reflection of the fact that there were more apoptotic neurons in both the ischemic (65.8±28.6 versus 23.3±28.7; P=0.03) and nonischemic (26.0±17.8 versus 8.4±10.7; P<0.05) hemispheres of LPS(+) animals 1 month after stroke (Figures 6C to 6E). The percentage of apoptotic to total neurons was also significantly greater in LPS(+) animals in the ischemic (16.4%±6.8% versus 5.1%±6.7%, P=0.02) and nonischemic (14.7%±9.2% versus 3.8%±5.0%; P=0.02) hemispheres. Sham-operated animals treated with LPS (N=4) had no evidence of neuronal apoptosis 1 month after surgery/LPS administration.
In these experiments, we show that an autoimmune response to brain does not generally occur after stroke; in fact, animals that experience cerebral ischemia seem predisposed to develop tolerance (a Th2/Th3 immune response) to CNS antigens. Animals that are exposed to a systemic inflammatory insult (i.e. injection of LPS) during stroke, however, tend to develop a Th1 or sensitizing response to brain antigens (in this case, MBP). The systemic inflammatory insult early after stroke onset leads to the expression of the necessary costimulatory and MHC molecules in brain essential for initiation of an adaptive immune response. Lipopolysaccharide administration was also associated with increased infiltration of lymphocytes into the brain after stroke. The histologic consequences of this autoimmune response include persistent inflammation (increased numbers of CD8+ cells) and enhanced brain atrophy/augmented neuronal apoptosis 1 month after stroke onset.
The fact that animals become sensitized to brain antigens after stroke would merely be an interesting observation unless there was a clinical correlate to the immune response. It could be argued that the differences in the neurobehavioral outcome of LPS(+) and LPS(–) animals were related directly to LPS, but it seems unlikely that LPS could account for the higher neurologic scores seen 1 week after administration or worse performance on the ‘sticky tape test’ 1 month after administration (Dantzer et al, 1998; Larson et al, 1996). In fact, LPS(–) animals that were Th1(+) also had higher neurologic scores 72h after stroke than LPS(–) animals that were Th1(–) (P<0.05). Furthermore, changes in weight are often used to indicate the activity of immune-mediated disease, and Th1(+), but not LPS(+), animals lost weight after MCAO (Pollak et al, 2000).
Infection in the poststroke period is common and is associated with worse outcome (Enlimomab Acute Stroke Trial Investigators, 2001; Davenport et al, 1996; Grau et al, 1999; Johnston et al, 1998). The intent of LPS administration in these experiments was to mimic the inflammatory response associated with systemic infection. The dose of LPS used in these experiments is relatively high and would be expected to compromise the BBB (Singh and Jiang 2004). Our data, however, argue that it is the combination of a cerebral insult and LPS that leads to the CNS autoimmune response. Given that there was no mortality in LPS(+) sham-operated animals and no significant changes in body temperature or weight, the plasma levels of LPS achieved in these animals seem clinically relevant. As such, our data may offer an explanation as to why infection during stroke is associated with worse outcome from stroke. Because Gram-negative bacteria are the predominant organisms causing infection after stroke, the use of LPS in these experiments is clinically relevant (Hilker et al, 2003; Puri et al, 2002). Future studies will need to address whether infection with other pathogens after stroke would similarly predispose to CNS autoimmunity.
Observational studies do not address the issue of whether infection worsens outcome from stroke or is just a marker for persons with more severe strokes; there are, however, several lines of evidence to suggest that infection is deleterious. Firstly, induction of an aseptic systemic inflammatory response after stroke is associated with increased morbidity and mortality (Enlimomab Acute Stroke Trial Investigators, 2001). Secondly, fever, a cardinal feature of the immune response in infection, is associated with worse neurologic outcome after stroke (Azzimondi et al, 1995; Reith et al, 1996). Thirdly, the cytokines secreted by leukocytes during the effector phase of an immune response might be toxic to neurons and glia (Barone et al, 1997; Hanisch et al, 1996; Touzani et al, 1999). Finally, because the bacterial products associated with infection (i.e. LPS) and the biologic mediators of the immune response (cytokines) can alter the microenvironment of the brain and peripheral lymphoid organs, lymphocytes could become sensitized to brain antigens in the context of an infection and contribute to cerebral ischemic injury.
Fever does not appear to be a primary mechanism by which LPS worsens outcome from MCAO, because body temperature was not elevated in LPS(+) or Th1(+) animals relative to LPS(–) or Th1(–) animals. Also, given that infarct sizes were similar in LPS(+) and LPS(–) animals, it does not appear that the initial injury was worsened by LPS administration. That LPS facilitates the induction of an autoimmune response to brain therefore seems tenable. In these experiments, LPS was used to simulate an infectious insult. Stimulation of the innate immune response can promote an adaptive (antigen-specific) immune response by creating an environment in which lymphocytes can become activated to antigens. During stroke, lymphocytes may encounter many novel CNS antigens to which an autoimmune response could occur. We evaluated the response to MBP, but it is plausible to assume that autoimmune response to other CNS antigens may also occur.
The long-term clinical consequences of the CNS autoimmune response after stroke are unknown. When the BBB is compromised by recurrent stroke or systemic inflammation, CNS autoreactive T cells could transit into brain and cause injury. This potential is illustrated by the fact that transgenic animals in which >95% of CD4+ T cells express receptors specific for MBP experience less recovery after spinal cord injury than wild-type controls and that animals immunized to MBP are more likely to die after MCAO (Becker et al, 1997; Jones et al, 2002). Moreover, T lymphocytes obtained from animals shortly after spinal cord injury precipitate histopathologic changes similar to experimental allergic encephalomyelitis (EAE) when injected into naïve animals (Popovich et al, 1996). While the absolute number of MBP-sensitized MNCs isolated from the brains of LPS(+) animals after MCAO in our study is only a fraction of the total number of inflammatory cells in the brain, it is similar to that seen in animals with EAE (where the numbers of MBP-reactive IFN-γ-secreting cells range from 10 to 20 per 1 × 105 MNCs) (Di Rosa et al, 2000; Li et al, 1998; Targoni et al, 2001). It is therefore conceivable that, in the appropriate clinical setting, MBP-specific MNCs generated after stroke could be of clinical consequence.
In summary, we showed that animals treated with LPS during MCAO are more likely to become sensitized to MBP. Furthermore, animals that develop a Th1 response to MBP experience worse outcome. Previously, we showed that immunologic tolerance to MBP is neuroprotective, showing that the balance of the Th1 and Th2/Th3 immune response towards CNS antigens is important in determining the consequences of the immune response (Becker et al, 2003, 1997). These data suggest that manipulating the immune response to brain antigens could be of therapeutic value if clinical studies show that patients develop a similar detrimental CNS autoimmune response after stroke.
This study was supported by a grant from the NINDS (KO2 NSO2160 NST).