Activation of immune responses to brain antigens after stroke

Abstract

Infection is common after stroke and is independently associated with a worse outcome. The predisposition to infection following stroke is in part related to a sympathetically mediated suppression of the peripheral immune response. The teleological explanation for this immune dysfunction is that it serves to prevent autoimmune responses to brain antigens. We believe that the systemic immune response in patients who develop infection, however, thwarts this seemingly protective response and predisposes to central nervous system autoimmunity. These autoimmune responses may mediate, at least in part, the worse outcome associated with post-stroke infection.

Infection is common after stroke and associated with worse stroke outcome. How infection confers a worse outcome is not clear. Our research suggests that the inflammatory response associated with systemic infection predisposes to the development of Th1-type autoimmune responses to CNS antigens exposed to circulating lymphocytes by virtue of stroke-induced breakdown of the blood–brain barrier (BBB) and ischemic injury to the brain. In addition, strategies to prevent development of these Th1-type responses are associated with improved outcome. We thus propose that the link between post-stroke infection and worse clinical outcome is the development of CNS autoimmune responses, which are provoked by infection.

Infection and stroke outcome

A link between infection and the risk of stroke has been appreciated for years (Grau et al. 2010). Recent data suggest that in addition to increasing the risk of stroke, infections occurring within a week of stroke onset increase early morbidity and mortality (Grabska et al. 2011). As suggested by clinical observations, animals with chronic systemic infections experience greater ischemic brain injury (Denes et al. 2010). Most patients who develop stroke, however, are infection free at the time of symptom onset. Given that infection is common after stroke, however, the effects of post-stroke infection on stroke outcome must also be considered. Depending on the patient population studied, published post-stroke infection rates can vary dramatically. In a systematic review of publications, post-stroke infection was estimated to occur in approximately 30% of patients (Westendorp et al. 2011). The most common infections are pneumonias (PNAs) and urinary tract infections (UTIs), each occurring in about 10% of patients with stroke; for patients admitted to the intensive care unit, the rates of infection are much higher (45% of patients with any infection, 28% with PNA, and 20% with UTI) (Westendorp et al. 2011).

In a prospective randomized controlled trial of a novel neuroprotective agent, 158/1455 (11%) of patients developed PNA in the first week, and 146/1455 (10%) developed UTI. For patients alive at day 7, those who had PNA were nearly two times more likely to die and three to four times more likely to have a poor outcome at 3 months, even after controlling for baseline prognostic variables like stroke severity, age, gender, and medical comorbidities. UTIs, on the other hand, were not associated with increased mortality, but did increase the likelihood (two to threefold) of poor outcome at 3 months (Aslanyan et al. 2004). Multiple other studies also suggest that infection is an independent risk factor for poor outcome after stroke (Tanzi et al. 2011; Grabska et al. 2011; Hong et al. 2008; Vermeij et al. 2009). Almost without exception, these studies suggest that PNA is more detrimental than UTI (Westendorp et al. 2011). Why PNA confers worse long-term outcome than UTI is unclear, but it seems likely that the systemic immune response associated with PNA is a more robust than that associated with UTI.

Few studies address infection in animal models of stroke. Given the incidence of post-stroke infection in the clinical setting, it would seem reasonable to track infection as an outcome in animal models of stroke; documentation of the incidence and severity of infection in animals receiving immunomodulatory therapies for the treatment of stroke is especially important. In at least some settings, post-stroke infection in animals is common and frequently fatal; prophylactic treatment with antibiotics prevents these infections and decreases mortality (Meisel et al. 2004). Data also show that the risk of infection is dependent on the strain of animal, suggesting that these strain differences could be capitalized upon to better understand the genesis of post-stroke infection (Schulte-Herbruggen et al. 2006).

Stroke-related immunodepression

Following experimental stroke, there is a profound dysfunction of the immune system that appears to be sympathetically mediated(Prass et al. 2003). This dysfunction is in part related to stroke-induced lymphopenia(Prass et al. 2003; Liesz et al. 2009a). In addition, lymphocytes from animals undergoing middle cerebral artery occlusion (MCAO) appear to have a defect in Th1-type immune responses, as demonstrated by impaired secretion of tumor necrosis factor (TNF)-α upon stimulation with lipopolysaccharide (LPS) and interferon (IFN)-γ upon stimulation with concanavalin A (Prass et al. 2003). A more recent study suggests that activation of the sympathetic nervous system following experimental stroke also impairs the response of hepatic invariant natural killer T cells predisposing to infection (Wong et al. 2011). Based on accumulated data, it seems likely that there are redundant pathways that lead to post-stroke immunodepression, primarily driven by the sympathetic response, but also with contributions from the endocrine system (Prass et al. 2003).

In patients, the risk of infection appears to be directly related to stroke severity/infarct volume (Chamorro et al. 2006; Hug et al. 2009; Haeusler et al. 2008; Tanzi et al. 2011). Several studies show an association between a variety of biomarkers, such as interleukin (IL)-10 and catecholamines, and post-stroke infection (Chamorro et al. 2006, 2007; Harms et al. 2011; Klehmet et al. 2009). As stroke induces a sympathetic response and a rise in a number of different cytokines, it is difficult to know whether these blood-born substances contribute to infection risk or merely serve as biomarkers for patients with severe stroke. Most of the studies reporting a link between biomarkers and infection risk do not systematically exclude patients who are already infected or control for stroke severity. In our study of 112 patients with ischemic stroke, however, we found that elevated plasma IL-1 receptor antagonist (IL-1ra) concentrations in infection-free patients were associated with an increased risk of later infection, even after controlling for stroke severity using the National Institutes of Health Stroke Scale (NIHSS) score (Tanzi et al. 2011).

Lymphopenia is common after stroke, and the degree of lymphopenia correlates with stroke severity (Czlonkowska et al. 1979; Vogelgesang et al. 2008; Urra et al. 2009b; Hug et al. 2009; Haeusler et al. 2012). Whether surviving lymphocytes function normally, however, is less clear. Some studies suggest that lymphocyte function is impaired, as detected by decreased proliferation of mitogens and decreased ex vivo secretion of pro-inflammatory cytokines (Rogers et al. 1998; Hug et al. 2009; Haeusler et al. 2012, 2008; Urra et al. 2009b). Other studies show that despite decreased lymphocyte numbers following stroke, their function is intact (Ferrarese et al. 1999; Vogelgesang et al. 2010; Tanzi et al. 2011). Table 1 shows data from our clinical study of patients with ischemic stroke and the correlation between stroke severity (based on NIHSS score or infarct volume), lymphocyte numbers, and lymphocyte function (as detected by the number of cells secreting IFN-γ in response to phytohemagglutinin). These data confirm the decrease in lymphocyte numbers associated with severe stroke, but do not show an alteration in their capacity to respond to the mitogen phytohemagglutinin. Whether these lymphocytes respond normally to typical infectious pathogens is unknown.

Table 1.

Correlation between stroke severity, lymphocyte numbers, and lymphocyte function

	Day 1 n = 25	Day 3 n = 90	Day 7 n = 89	Day 30 n = 86
Correlation with NIHSS score*
Lymphocytes	−0.656 p < 0.001	−0.331 p = 0.001	−0.227 p = 0.033	−0.046 NS
Lymphocyte response to PHA	−0.157 NS	0.075 NS	−0.125 NS	0.064 NS
Correlation with infarct volume†
Lymphocytes	−0.564 p = 0.003	−0.365 p < 0.001	−0.252 p = 0.024	−0.123 NS
Lymphocyte response to PHA	−0.211 NS	−0.058 NS	−0.072 NS	0.058 NS

Data are presented as *Spearman’s rho or ^†Pearson’s r.

Lymphocyte function was determined by the number of cells secreting interferon-γ in response to stimulation with PHA (as determined by enzyme-linked immunosorbent spot (ELISPOT) assay. NIHSS, National Institutes of Health Stroke Scale; PHA, phytohemagglutinin; NS, not significant (p > 0.200).

In addition to impaired lymphocyte responses following stroke, many studies demonstrate ‘deactivation’ of monocytes with loss of human leukocyte antigen-DR (Haeusler et al. 2008, Zhang et al. 2009; Urra et al. 2009a). At the same time, there are reports of increased expression of toll-like receptor (TLR) 4 on monocytes (Urra et al. 2009a). The number of monocytes in circulation following stroke, however, is elevated (Chamorro et al. 2006; Urra et al. 2009a; Hug et al. 2009). In our cohort of ischemic stroke patients, the number of monocytes was clearly related to the NIHSS score (ρ = 0.634, p < 0.001 at 72 h). Given this relationship to stroke severity, it might be expected that the number of monocytes was higher among patients who later developed an infection and that the strength of the association would be attenuated (but not lost) after controlling for stroke severity (Table 2). Other investigators have shown a similar relationship between monocyte numbers and the risk of post-stroke infection (Chamorro et al. 2006; Urra et al. 2009a). It is important to remember, however, that not all monocytes are the same, and the phenotype matters (Urra et al. 2009c).

Table 2.

The number of monocytes^* in patients with and without infection

Infection by day 15:	All patients				Patients with infection at 72 h eliminated (n = 7)
	Yes n = 26	No n = 77	Uncontrolled	Controlled for NIHSS	Yes n = 19	No n = 77	Uncontrolled	Controlled for NIHSS
			p	p			p	p
Monocytes* (thou/microL)	0.88 (0.58, 1.13)	0.65 (0.46, 0.83)	0.012	NS	1.04 (0.65, 1.20)	0.65 (0.46, 0.83)	<0.001	0.063

Predictive value	Uncontrolled		Controlled for NIHSS		Uncontrolled		Controlled for NIHSS
	OR (95% CI)	p	OR (95% CI)	p	OR (95% CI)	p	OR (95% CI)	p
Monocytes* (per thou/microL)	10.14 (2.09–49.26)	0.004	2.12 (0.30–15.16)	NS	34.35 (4.86–242.74)	<0.001	8.81 (0.68–98.12)	p = 0.097

The number is higher among patients destined for infection in the first 15 days after stroke onset, and the difference is even more marked after eliminating patients with infection in the first 3 days after stroke onset. The difference in monocytes numbers is attenuated by controlling for stroke severity. Similarly, the odds for infection rises dramatically with every 1000 monocytes per microliter of blood and is most notable in those patients who are infection free at the time the blood was drawn.

Data are presented as median (interquartile range) and statistics are by Mann–Whitney U test.

^*Highest value by 72 h after stroke onset. thou/microL, thousand per microliter; NIHSS, National Institutes of Health Stroke Scale; OR, odds ratio; CI, confidence interval; NS, not significant (p > 0.200).

Consistent with the idea of immunodepression following stroke, some recent studies suggest that serum TNF-α is decreased after stroke (Vogelgesang et al. 2010; Urra et al. 2009a). Most studies, however, show a dramatic increase in plasma TNF-α and other pro-inflammatory cytokines following stroke that correlates with stroke severity (Zaremba and Losy 2001; Montaner et al. 2003; Intiso et al. 2004; Mazzotta et al. 2004; Sotgiu et al. 2006; Tuttolomondo et al. 2009; Tanzi et al. 2011; Offner et al. 2006). If lymphocytes and monocytes are not the source of this TNF-α, it is not clear what the source is. Changes in plasma IFN-γ following stroke are rarely noted, but when they are reported, they are often noted to be elevated (Urra et al. 2009a). We were unable to detect plasma IFN-γ in most of our ischemic stroke patient cohort.

Relative immunologic privilege of the brain

It has long been appreciated that the immune response within the brain differs from that in other organs. Part of this difference is related to the presence of the BBB, which limits the transit of leukocytes, as well as large molecules like immunoglobulins, into the brain. Just as important as the BBB, however, is the fact that the normal brain does not readily support the development of an immune response (Bailey et al. 2006). For instance, microglia, the resident antigen-presenting cells (APCs) in the brain, express no or very low levels of the major histocompatibility complex (MHC) II molecule, which is necessary for T-cell recognition of the antigen. In addition, the costimulatory molecules needed for activation of T cells during antigen presentation are generally absent (Dangond et al. 1997; Becker et al. 2005). Finally, the transforming growth factor (TGF)-β1, a potent immunomodulatory cytokine, is present in the resting brain and rapidly up-regulated following ischemia (Ali et al. 2001; Krupinski et al. 1996). The presence of TGF-β1 thus serves to inhibit the development of inflammatory immune responses.

Following stroke, there is at least a transient compromise in the BBB that allows cells of the immune system entrance to the brain parenchyma (Chen et al. 2009). Indeed, multiple studies show that there is an early influx of neutrophils followed by a later influx of lymphocytes. Lymphocytes can be found within the ischemic territory as early as a day or so after stroke onset and become increasingly numerous over the ensuing days (Schroeter et al. 1994; Jander et al. 1995). Within the brain, microglia become activated following stroke and are able to express MHC II, although the degree of expression is somewhat limited in comparison with other cell types (Felger et al. 2010). For instance, dendritic cells from the peripheral circulation infiltrate the infarct core within days after MCAO, while brain-resident dendritic cells (distinct from microglia) can be found at the periphery of the infarct (Felger et al. 2010). These dendritic cells are seen in close apposition to infiltrating lymphocytes, respond to IFN-γ, and are efficient at driving antigen-specific T-cell responses (Felger et al. 2010; Gottfried-Blackmore et al. 2009).

Disruption of the BBB following stroke also allows for CNS antigens to ‘leak’ into the peripheral circulation early after the onset of ischemia. The concentration of these antigens parallels the severity of the stroke and reflects cell death (Jauch et al. 2006; Herrmann et al. 2000). It is thus possible that these antigens can be processed and presented to the immune system in peripheral organs such as the spleen and lymph nodes. In fact, animal studies demonstrate the presentation of CNS antigens to lymphocytes in the cervical lymph nodes within hours of stroke onset, and a recent clinical study found evidence of CNS antigen presentation in the tonsils of patients by within a few days after stroke onset (van Zwam et al. 2009; Planas et al. 2012).

Bystander activation, the nature of the immune response, and autoimmune diseases

T-lymphocyte activation to an antigen generally requires that the antigen be processed by an APC and presented to the lymphocyte in the context of MHC II. In general, this means that the T cell must traffic to the site of antigen presentation; following stroke, antigen presentation can occur in both the brain and in the periphery. Previous studies, including our own, have demonstrated that inhibition of very late antigen-4 can improve stroke outcome in the days following MCAO (Becker et al. 2001; Relton et al. 2001; Liesz et al. 2011b). Inhibition of lymphocyte trafficking into brain by other methods, including antagonizing sphingosine-1-phosphate (S1P) receptors, has yielded mixed results, but is largely found to be beneficial as well (Liesz et al. 2011a; Wei et al. 2011; Hasegawa et al. 2010; Czech et al. 2009).

All of these studies evaluate short-term outcomes following experimental stroke, however, and therefore do not address the adaptive immune response to CNS antigens, which takes time to develop. In addition, these studies address the effects of therapies which non-selectively block the trafficking of lymphocytes, including natural killer cells and other cells that act in an antigen-non-specific manner. For lymphocytes to become activated to a specific antigen, that antigen must be recognized by appropriate T-cell receptor, and the lymphocyte must receive adequate costimulation from the APC. The nature of the immune response that develops from the interaction between the APC and the lymphocyte depends upon the microenvironment at the site of antigen presentation. (Zygmunt & Veldhoen 2011) Continued or repeated antigen exposure then leads to expansion of the lymphocyte population, a process that takes days to weeks to occur, based on demonstrations by models of experimental allergic encephalomyelitis where disease does not appear until at least day 10 after active immunization (Mannie et al. 2009).

Following stroke, the sympathetically mediated immune dysfunction tends to be associated with a reduced efficacy of circulating costimulatory cells and loss of human leukocyte antigen-DR (Hug et al. 2011; Urra et al. 2009a; Hug et al. 2009). As mentioned above, the general milieu of the brain does not support the development of inflammatory immune responses (Romo-Gonzalez et al. 2012; Suter et al. 2003). Several studies show an increase in the number of Tregs (Fig. 1) following stroke (Urra et al. 2009b; Yan et al. 2012). In experimental studies, depletion of Tregs is associated with increased inflammation, increased infarct volume at 1 week (but not at 2 days) and worse functional outcome from experimental stroke (Liesz et al. 2009b). In another study, however, depletion of Tregs prior to MCAO did not appear to affect infarct volume at 4 days (Ren et al. 2011). The role of Tregs in these studies was evaluated through depletion of Tregs by administration of anti-CD25 antibodies, meaning that all Treg cells, regardless of their antigen specificity, were depleted. Because the development of endogenous antigen-specific Treg cells takes time, it seems very unlikely that the Treg cells depleted in these studies recognized CNS antigens. Following experimental stroke, we found that the predominant immune response to myelin basic protein (MBP) after MCAO was that of a Treg response, but this response took many days to develop (Becker et al. 2005). Table 3 shows the change in the number of cells secreting TGF-β1, a Treg cytokine, in an antigen-specific fashion in our clinical study. At day 4 after stroke, there were trends for more cells to secret TGF-β1 in response to stimulation with glial fibrillary acidic protein and tetanus toxin in comparison to ‘healthy’ controls. By day 180 after stroke, the TGF-β1 response to neuron-specific enolase (NSE) and proteolipid protein were much more robust among patients with stroke, with trends toward enhanced TGF-β1 responses to other antigens as well. Furthermore, these responses tend to become more robust over time, a finding which is especially true for NSE. Finally, among patients who experienced an early infection (first 15 days), the TGF-β1 response to NSE, S100, and proteolipid protein was significantly greater (p < 0.05) at 180 days than in patients without early infection.

Fig. 1.

The usual lymphocyte responses to brain antigens following stroke are either no response or a TREG response. In the setting of a systemic infection, however, the immune system is activated and can predispose to the development of TH1 cells that are potentially detrimental. PNA, pneumonia; TGF, transforming growth factor; IL, interleukin; IFN, interferon.

Table 3.

Relative increase in the number of cells secreting TGF-β1 in response to stimulation with different CNS antigens and TT in patients with stroke and in ‘healthy controls’

	Controls n = 39 Median (IQR)	Day 4		Day 180		Change over time p
		Median (IQR)	p	Median (IQR)	p
MBP	1.09 (0.90, 1.43)	1.17 (0.88, 1.69)	NS	1.53 (0.84, 2.66)	0.091	0.086
NSE	1.20 (0.90, 1.20)	1.41 (0.94, 2.19)	NS	2.19 (1.20, 5.20)	< 0.001	0.003
S100	1.22 (0.79, 1.78)	1.39 (1.00, 1.98)	NS	1.76 (0.85, 3.60)	0.068	0.198
PLP	1.28 (1.03, 1.61)	1.42 (0.92, 2.57)	NS	1.93 (0.98, 4.32)	0.033	0.185
GFAP	1.26 (0.92, 1.77)	1.64 (0.92, 2.52)	0.069	1.60 (0.99, 4.17)	0.073	NS
TT	1.24 (0.87, 1.78)	1.53 (1.02, 3.05)	0.059	1.66 (0.97, 2.95)	0.079	NS

Number of cells secreting transforming growth factor (TGF)-β1 determined by enzyme-linked immunosorbent spot (ELISPOT) assay. IQR, interquartile range; MBP, myelin basic protein; NSE, neuron-specific enolase; PLP, proteolipid protein; GFAP, glial fibrillary acidic protein; TT, tetanus toxin; significant (p < 0.05) are in bold; NS, not significant (p > 0.200).

The teleological explanation for induction of a Treg-type response is that it limits the possibility of developing detrimental autoimmune responses to brain in the setting of BBB compromise. Given the propensity for infection following stroke, however, we argue that this seemingly protective ‘immunodepressive’ response can be thwarted. Specifically, infections (PNA in particular) can induce systemic inflammation, activate APCs/dendritic cells through the TLR pathway, and alter the milieu of the brain and lymphoid organs, all of which serve to optimize the conditions for lymphocyte activation. For example, activation of TLRs through agents like LPS induces expression of MHC II and costimulatory molecules on dendritic cells and microglia (Bauer et al. 2009; Olson and Miller 2004; Raymond and Wilkie 2005). Systemic infection is also associated with an increase in circulating pro-inflammatory cytokines like TNF-α and IFN-γ (de Werra et al. 1997; Tamayo et al. 2011). The infection-induced changes in the immune system lead to a scenario whereby lymphocytes may become activated to self-antigens by virtue of a phenomenon referred to as ‘bystander activation’. Bystander activation is invoked as the trigger for a number of autoimmune diseases and explains how inflammation associated with an innate immune response (or an immune response to an unrelated antigen) can trigger a response to other antigens within the vicinity (Sfriso et al. 2010).

Autoimmunity: the link between infection and outcome?

Given the breakdown in the BBB following stroke and the potential for presentation of CNS antigens to the immune system, we assessed animals for the development of CNS autoimmune responses following stroke. In a model of severe stroke, we found that Th1-type immune responses to MBP were rare, but that the immune response could be pushed to a Th1-type response by systemic administration of LPS at the time of MCAO (Becker et al. 2005). Animals that develop a Th1-type response to MBP, in particular, have worse long-term outcome from stroke (Becker et al. 2005; Gee et al. 2008, 2009; Zierath et al. 2010). LPS is a component of the Gram-negative bacterial cell wall and a potent agonist of TLR4. Infections with Gram-negative pathogens thus have the ability to activate APCs/dendritic cells through TLR4 and create an environment more likely to result in lymphocyte activation. Increased expression of both TLR4 and TLR2 are seen after stroke and are linked to worse outcome and an increase in pro-inflammatory cytokines (Urra et al. 2009c; Yang et al. 2008; Brea et al. 2011).

On the basis of these observations, we considered the possibility that our experimental model might help explain the link between post-stroke infection and poor outcome in humans—specifically, infection may predispose to the development of deleterious autoimmune responses to brain antigens. In our clinical study, we found that PNAs, but not less serious infections, were associated with a higher proportion of patients developing a Th1 response to MBP (Becker et al. 2011). Furthermore, the immune response to MBP was an independent predictor of poor outcome. These findings appear to validate our assumptions, but need to be replicated in a larger study. The finding that PNA is more likely to be associated with the development of autoimmune responses to brain than other less severe infections (like UTIs) is consistent, however, with the observation that the long-term clinical consequences of PNA are more severe than that of UTI (Aslanyan et al. 2004; Westendorp et al. 2011).

Summary

On the basis of a number of different clinical observations regarding post-stroke infection, we considered the possibility that infections could be deleterious through activation of the immune response and predisposing toward CNS autoimmunity. Both animal and human studies seem to support this hypothesis. To date we have focused our research on the immune responses to a small number of potential CNS antigens. Years ago it was found that patients with stroke had evidence of a cellular immune response to MBP, but these studies were done in the context of studying multiple sclerosis (Youngchaiyud et al. 1974; Wang et al. 1992; Kallen et al. 1977). More recent studies have shown that autoantibodies to a variety of CNS antigens can be found after stroke (Dambinova et al. 2003; Bornstein et al. 2001). It is likely that immune responses to a multitude of different CNS antigens occur following brain injury, and that we are merely constrained in our ability to screen for these responses. The cellular and humoral responses to brain are clearly epiphenomena of injury; some of these responses may be of no consequence, but others could contribute to worse short-term as well as long-term disability. Whether autoimmune responses to brain following stroke can be prevented through prevention of infection or modulation of the post-ischemic immune response remains to be seen.

Abbreviations used

APCs

antigen-presenting cells

BBB

blood–brain barrier

IFN-γ

interferon-γ

MBP

myelin basic protein

MCAO

middle cerebral artery occlusion

MHC

major histocompatibility complex

NIHSS

National Institutes of Health Stroke Scale

NSE

neuron-specific enolase

PNAs

pneumonias

TGF-β1

transforming growth factor-β1

TLR

toll-like receptor

TNF

tumor necrosis factor

UTIs

urinary tract infections