Download GLUCOCORTICOIDS INCREASE CNS INFLAMMATION

GLUCOCORTICOIDS INCREASE CNS INFLAMMATION,
WORSENING ACUTE NEUROLOGICAL INJURY
A DISSERTATION
SUBMITTED TO THE DEPARTMENT OF BIOLOGY
AND THE COMMITTEE ON GRADUATE STUDIES
OF STANFORD UNIVERSITY
IN PARTIAL FULFILLMENT OF THE REQUIREMENTS
FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY
Shawn Fletcher Sorrells
June 2011
© 2011 by Shawn Fletcher Sorrells. All Rights Reserved.
Re-distributed by Stanford University under license with the author.
This work is licensed under a Creative Commons AttributionNoncommercial 3.0 United States License.
http://creativecommons.org/licenses/by-nc/3.0/us/
This dissertation is online at: http://purl.stanford.edu/ww382cb9419
ii
I certify that I have read this dissertation and that, in my opinion, it is fully adequate
in scope and quality as a dissertation for the degree of Doctor of Philosophy.
Robert Sapolsky, Primary Adviser
I certify that I have read this dissertation and that, in my opinion, it is fully adequate
in scope and quality as a dissertation for the degree of Doctor of Philosophy.
Ben Barres
I certify that I have read this dissertation and that, in my opinion, it is fully adequate
in scope and quality as a dissertation for the degree of Doctor of Philosophy.
Firdaus Dhabhar
I certify that I have read this dissertation and that, in my opinion, it is fully adequate
in scope and quality as a dissertation for the degree of Doctor of Philosophy.
Anton Wyss-Coray
Approved for the Stanford University Committee on Graduate Studies.
Patricia J. Gumport, Vice Provost Graduate Education
This signature page was generated electronically upon submission of this dissertation in
electronic format. An original signed hard copy of the signature page is on file in
University Archives.
iii
Abstract
Glucocorticoids (GCs) are stress hormones that are well-known for their potent and
pleiotropic anti-inflammatory effects. In the injured CNS, their anti-inflammatory properties
could be particularly beneficial due to the often detrimental effects of excessive inflammation in
the brain. In more recent years, however, it has become clear that GCs do not always decrease
inflammation and can even augment aspects of the immune response. The research presented in
this doctoral dissertation sought to determine the impact of this increased inflammatory response
in the CNS using animal models of excitotoxicity and hypoxia/ischemia. Specifically, both
exogenous GC treatments and the endogenously released GCs post-injury were found to increase
immune cell activation (both in phenotype and in p65 nuclear translocation) in both rat and
mouse models of excitotoxic hippocampal neuron death and in a mouse MCAO stroke model.
These increased inflammatory responses are likely to be mediated by an unexpected GCsuppression of several anti-inflammatory cytokines including CX3CL1 and CD22 and failure of
GCs to activate some of their normal anti-inflammatory targets like IkBa, IL-1ra, and MKP-1.
Furthermore, these GC-augmented inflammatory responses are necessary for GCs to make more
neurons die from either of these injuries. Taken together, this work demonstrates that cellular
inflammation is not kept in check by GCs in the forebrain; instead, GCs worsen hippocampal and
cortical neuron death, at least in part, by increasing the neurotoxicity of CNS inflammation.
iv
Acknowledgements:
The work presented in this dissertation would never have been possible without the care and
devotion of the following people: Robert Sapolsky, Ben Barres, Firdaus Dhabhar, Tony WyssCoray, Theo Palmer, Javier Caso, Carolina Munhoz, Nathan Manley, Angela Lee, Bonnie Chien,
Kevin Tran, Caroline Hu, and all of the members of the Sapolsky lab.
The following individuals provided instrumental and lively discussions about the research along
the way: Norman Ruby, Dritan Agalliu, Tyler Cutforth, Brenda Kusler, Marta Perez, Andy Evans,
Eric Hoopfer, Takaki Komiyama, Dev Manoli, and Joe Lipsick.
In addition, many of the most interesting findings discovered in these investigations were a direct
result of the supreme generosity of Dr. Louis Muglia who freely provided floxed-GR mice. Drs.
Sam Okret and Ben Barres also kindly provided mouse lines for these investigations.
Finally, I would especially like to thank my brother Trevor and my parents Mark and Nancy
Sorrells for their love and support over these past years.
v
Table of Contents
Chapter 1: Background and Literature Review
1
1.1 Introduction
1
1.2 Glucocorticoid regulation of immunity outside the nervous system
3
1.2.1 Classic anti-Inflammatory effects of glucocorticoids
3
1.2.2 The pro-inflammatory effects of stress and GCs outside the nervous system
5
1.3 Mechanisms of neuron death and inflammation
10
1.3.1 Neuron death
10
1.3.2 The inflammatory response to neuron death
11
1.4 Glucocorticoid regulation of immunity in the nervous system
12
1.4.1 Normal glucocorticoid effects in the brain
12
1.4.2 Glucocorticoids endanger neurons
12
1.4.3 Glucocorticoid effects on inflammatory cells in the CNS
14
1.4.4 Glucocorticoid effects on inflammatory cytokines in the CNS
15
1.4.5 Glucocorticoid effects on inflammatory transcription factors
16
1.5 Overview of experimental hypotheses
16
1.5.1 Primary hypotheses tested
17
1.5.2 Secondary hypotheses tested
18
Chapter 2: GC Activation of CNS Immune Responses
20
2.1 Experimental model
20
2.1.1 MGRKO mice have GR deleted in myeloid cells
20
2.1.2 FBGRKO mice have GR deleted in forebrain neurons
21
2.1.3 MGROV and FBGROV mice have inducible overexpression of GR in myeloid
cells or neurons
22
2.1.4 GC levels in experimental system
22
2.1.5 Injury models
23
2.2 Results
24
vi
2.2.1 GR signaling in myeloid cells increases immune cell activation and recruitment
24
2.2.2 GCs increase immune cell activation in the injured rat hippocampus
26
2.3 Discussion
27
2.3.1 Endogenous and exogenous GCs affect different aspects of cellular inflammation
post-injury
27
2.3.2 GCs increase immune cell activation in the injured hippocampus
28
2.4 Conclusions
29
Chapter 3: Mechanisms of GC-Augmented Immune Responses
31
3.1 Results
31
3.1.1 Hypothesis (i): GCs increase pro-inflammatory cytokines and chemokines
31
3.1.2 Hypothesis (ii): GCs decrease anti-inflammatory signaling molecules
32
3.1.3 Hypothesis (iii) GCs increase NF-kB activity
34
3.2 Discussion
35
3.2.1 Hypothesis (i): GCs do not increase pro-inflammatory cytokines and chemokines
35
3.2.2 Hypothesis (ii): GCs decrease anti-inflammatory signaling molecules
37
3.2.3 Hypothesis (iii) GCs increase NF-kB activity
38
3.3 Conclusions
40
Chapter 4: GC-Endangerment of Neurons Requires GC-Augmented Inflammation
41
4.1 Results
41
4.1.1 GC-endangerment requires GC-augmented inflammation
41
4.1.2 GR signaling in myeloid cells endangers neurons in mice
41
4.1.3 GR signaling affects BDNF levels differently in the hippocampus and cortex
42
4.1.4 Endothelial cell GR is also detrimental to neuron survival post-injury
43
4.2 Discussion
44
4.2.1 GC effects on inflammation contribute to GC endangerment of neurons
44
4.2.2 GC endangerment requires myeloid cell GR, not neuron GR
45
4.2.3 Endogenous GCs are also detrimental to neuron survival, but less so
46
vii
4.2.4 GR signaling has cell-autonomous effects on BDNF levels
47
4.2.5 GR signaling in endothelial cells also endangers neurons during MCAO
48
4.3 Conclusions
50
Chapter 5: General Discussion and Conclusions
51
5.1 Primary hypotheses tested and interpretations
51
5.1.1 GCs augment inflammation
51
5.1.2 GCs suppress anti-inflammatory cues and increase NF-kB activation
55
5.1.3 GC-augmented inflammation is necessary for neuron death
57
5.2 Secondary hypotheses tested and interpretations
57
5.2.1 Timing and duration of GC exposure
57
5.2.2 Synthetic versus endogenous GCs
59
5.2.3 Basal/permissive levels of GCs versus stress levels
60
5.2.4 Regional differences in GC effects in the brain
61
5.2.5 Does GC-augmented inflammation and neuron endangerment occur in mice?
62
5.3 Summary and Conclusions
62
Materials and Methods
66
Figures
72
Tables
93
References
95
viii
List of illustrations
Fig. 1. MGRKO mice have GR deleted in the parenchymal microglia of
the cortex and hippocampus.
72
Fig. 2. LysM-CRE is efficiently expressed in the parenchymal microglia
of the cortex and hippocampus.
73
Fig. 3. FBGRKO mice have GR deleted in neurons of the cortex and hippocampus.
74
Fig. 4. Overexpression of GR in microglia and neurons using the same CRE-drivers.
75
Fig. 5. Blood GC levels in experimental model.
76
Fig. 6. GCs enhance cellular inflammation and can both enhance and suppress recruitment.
77
Fig. 7. Use of flow cytometry to identify CNS leukocytes. GR overexpression
does not alter cell-recruitment.
78
Fig. 8. Minocycline reverses GC-augmented immune cell activation in the hippocampus.
79
Fig. 9. GCs do not increase cellular inflammation by enhancing chemokine production.
80
Fig. 10. GCs suppress anti-inflammatory gene expression.
81
Fig. 11. GCs affect CX3CL1 levels oppositely in the hippocampus post-KA and
cortex post-MCAO.
82
Fig. 12. GCs suppress pro-inflammatory cytokines and are necessary for COX-2 expression. 83
Fig. 13. GCs have an impaired ability to inhibit NF-kB in the hippocampus.
84
Fig. 14. GCs increase p65 nuclear localization.
85
Fig. 15. GC-augmented inflammation is necessary for GC-endangerment.
86
Fig. 16. GC-endangerment requires GR signaling in myeloid cells but not neurons.
87
Fig. 17. GR overexpression has minimal effects on neuron death.
88
Fig. 18. GCs suppress BDNF levels in the hippocampus and cortex in a cell-specific manner. 89
Fig. 19. Endothelial cell KO and OV mice have higher GR levels in endothelial cells.
90
Fig. 20: Endothelial cell GR signaling also worsens infarct post-MCAO.
91
Fig. 21. Summary of endogenous GC and exogenous GC effects during different injuries.
92
ix
List of tables:
Table 1. Some experimental conditions where GCs have been found to increase inflammation. 93
Table 2. Summary of parameters that impact how GCs affect inflammatory responses.
x
94
Chapter 1: Background and Literature Review
1.1 Introduction
We are born with the ability to respond to stress, and this is fortuitous because the act of
being born is the first of many stressors that we will experience during life. In humans and other
vertebrates, stressful circumstances elicit a series of physiological changes that constitute the
stress response. This begins within seconds with the release of the catecholamines of the
sympathetic nervous system (epinephrine and norepinephrine). In the ensuing minutes, the
hypothalamic-pituitary-adrenal (HPA) axis is activated. CRF produced by the hypothalamus acts
on the pituitary to activate the release of ACTH into circulation, resulting in the release of the
adrenal steroid glucocorticoids (GCs). The role of GCs is largely to modulate and control the
stress response over a longer timeframe in the minutes to hours following a stressor. This
response is conducted largely on a genomic level and accordingly the actions of GCs are
numerous and varied (1).
Our understanding of the stress response aptly began during the Great Depression. In 1935,
Walter Cannon described the extraordinary flexibility of the body in its ability to respond to
stress, or “accidents of existence” (2). In 1936, Hans Selye described the “general adaptation
syndrome”, activated by an organism in order to overcome various challenges (3). Cannon called
the stress-induced increases in cardiac output the “fight or flight” response, and he had realized
the importance of adrenal hormones in this response as early as 1924 (4). “Fight or flight,” as
described by Cannon, reflects the actions of both catecholamines and GCs (although the latter
were unknown until the work of Selye) to immediately increase cardiovascular output and blood
flow to the brain and skeletal muscles. This is predominately mediated by catecholamines,
although GCs potentiate their effects (1). Both hormones mobilize energy stores from adipose and
hepatic cells, ensuring a supply of energy to exercising muscle. At the same time, GCs decrease
less immediately essential activities such as feeding, digestion, growth and reproduction (1).
Thus, two separate hormone systems conduct divergent effects over time on a variety of
physiological parameters as an organism responds to stress.
1
Stress also has potent effects on the immune system, and indeed one of Selye’s original
observations was that chronic GC exposure caused the thymus to atrophy (5). Over the next
decades, GCs began to be widely utilized in clinical practice for their particularly potent
immunosuppressive properties. As research began to reveal the mechanisms underlying GCmediated immunosuppression, it also became apparent that stress and stress hormones can also
have immune-activating properties. Several different research groups have identified parameters
that lead to stress increasing immune responses (6-10). Some of the most important of these
factors are (1) the type of stress hormone (2) timing of hormone exposure (3) duration of
hormone exposure, and (4) hormone concentration. Like GCs, catecholamines can increase or
decrease immune responses, but these effects are not considered further here. GC effects also
vary depending on whether they are synthetic or endogenous, but the effects of physiological
levels of endogenous GCs were selected as the primary focus for this work. The most immunestimulating effects are caused by acute exposure to GCs and these augmented immune responses
can be beneficial to injury recovery (11). In contrast, chronic exposure to GCs is more frequently
detrimental and immunosuppressive (6). This explanation fits well into an adaptive understanding
of the immune system during the stress response where it needs to be mobilized initially, but must
be kept under control in the long term (9).
In the injured CNS, the consequences of excessive or prolonged inflammation are
particularly dire, so GCs are widely used in clinical neurology. Additionally, CNS injury is a
stressor itself so endogenous GCs are also present. Given that GCs can sometimes increase
inflammation, it is essential to know how they affect CNS inflammatory responses to direct their
proper use. The next sections review the existing literature about how stress and GCs affect
immune responses, beginning with their well-studied anti-inflammatory effects and then
discussing the conditions that lead to increased immune responses in the periphery. Finally, the
small but growing body of literature studying GC-augmented inflammatory responses in the CNS
is discussed.
2
1.2 Glucocorticoid regulation of immunity outside the nervous system
1.2.1 Classic anti-Inflammatory effects of glucocorticoids
Hans Selye in the 1930’s obtained the first evidence for GCs being anti-inflammatory,
noting that sustained stress or GC exposure caused involution of the thymus. This was such a
dramatic and reliable effect that it was long used as a bioassay for circulating GC concentrations,
before the development of radio-immunoassays. The stress-induced thymic involution described
by Selye was subsequently shown to be caused by GR-mediated apoptosis in immature T cells
(12). In the wake of their establishment as profoundly anti-inflammatory, synthetic GCs came to
be used in many domains of clinical pharmacology.
Work since then has examined GC-induced immunosuppression on far more reductive
levels than the involution of an organ. On the cellular level, GCs can inhibit both the innate and
adaptive immune responses. GCs inhibit lymphocyte proliferation, induce apoptosis in basophils
and eosinophils, and cause a redistribution of T-cells (13). GCs also cause dendritic cells (DCs),
the antigen-presenting cells of the adaptive immune system, to maintain an immature state
(13). In addition, GCs can induce T-cell apoptosis and a shift in the population of T cells from
mediating cellular immunity with a Th1 phenotype to mediating humoral immunity with a Th2
phenotype (1).
On the molecular level, GCs limit the capacity of DCs to interact with immature T-cells by
inhibiting MHC II expression. Moreover, GCs inhibit or reduce the expression of a variety of
chemoattractants and pro-inflammatory cytokines (e.g. IL-1β, IL-6, TNF-α), while enhancing the
expression of anti-inflammatory cytokines (e.g. IL-10, TGF-beta) (1). The expression of other
pro-inflammatory molecules such as inducible COX-2 and prostaglandins is also inhibited by
GCs. The inhibition of cytokines and other pro- inflammatory molecules has been demonstrated
at the levels of both protein and mRNA synthesis. In addition, GCs enhance the production of
decoy receptors and antagonist molecules such as IL-1R2 and IL-1Ra, resulting in decreased IL-1
signaling (although the latter is disputed) (14).
GCs exert these varied effects through well-established mechanisms of steroid hormone
3
signaling. Upon cellular entry, GCs bind to either the Type I, mineralocorticoid receptor (MR), or
the Type II glucocorticoid receptor (GR). MR has a much higher affinity (0.5 nM) for GCs than
does GR (5.0 nM), and this difference is a large part of the reason for concentration-dependent
differences in GC activity (15). At basal GC levels, GCs predominantly bind to MR and only
slightly occupy GR. In the early stage of a stressor, MR becomes saturated and the occupancy of
GR is increased. It is not until a sustained, major stressor that GR occupancy is saturated. Thus,
MR is responsible for much of the effects of basal and low-stress levels of GCs (i.e. the
permissive effects), whereas GR largely mediates the effects of high stress GC levels. This,
combined with the fact that MR and GR signaling often have opposing outcomes, results in a Ushaped curve of GC action where intermediate (basal to low-stress) GC concentrations have the
opposite effect of no GCs or high-stress concentrations (1, 10).
Both MR and GR belong to a large class of nuclear receptors which homodimerize and
translocate to the nucleus after ligand binding. There, both MR and GR bind to the glucocorticoid
response element (GRE) consensus DNA sequence, and modulate the transcriptional activity of
downstream genes in a similar manner. Because both MR and GR can bind to the same GRE,
their often opposite effects arise from their different patterns of protein-protein interactions with
other components of the transcriptional machinery. Unlike MR, GR directly interferes with the
transcriptional activity of the pro-inflammatory transcription factors nuclear factor kappa B
(NFκB) and activating protein 1 (AP-1), thus decreasing expression of the inflammatory genes
they would normally activate. The result is an extensive decrease in inflammatory mediators
(reviewed in (14)). Interestingly, reports of rapid non-genomic signaling by these nuclear
hormone receptors suggest a mechanism by which GCs might be able to mediate temporally
distinct effects (16, 17). Such non-genomic effects likely occur via protein–protein interactions
between the GR and other cytosolic or nuclear factors; however, there are also more enigmatic
membrane-bound nuclear hormone receptors that are also thought to mediate some of the rapid
signaling effects of GCs.
As discussed, GCs do not always decrease inflammation. The next section describes how
4
stress was found to have immune-enhancing effects and the realization that GCs are actually
responsible for some of these increases.
1.2.2 The pro-inflammatory effects of stress and GCs outside the nervous system
Stress has been assumed to be anti-inflammatory due to the well-understood antiinflammatory actions of GCs; however, this is not necessarily the case. In particular, acute stress
just before an inflammatory challenge often has the adaptive effect of enhancing the immune
response, whereas chronic stress has the role of keeping the immune response in check by
suppressing it (6). For example, one type of the delayed type hypersensitivity (DTH) cellmediated immune response is exacerbated by acute stress but suppressed by chronic stress (18).
Moreover, peripheral and central nervous system (CNS) levels of pro-inflammatory cytokines
such as IL-1β are increased following an acute stressor in both rats (19) and humans
(20). Additionally, the activity of NFκB and expression of NFκB target genes increase following
acute stress in human peripheral blood mononuclear cells (21) and rat cortex (22). These studies
all concluded that acute stress enhanced the response to a subsequent immune challenge. The
timing of the relationship between stressor and immune challenge is critical, in that acute stress
after an inflammatory challenge decreases expression of IL-1β, TNF-α, IL-6, and other proinflammatory cytokines (23).
Given the widespread anti-inflammatory actions of GCs, it seems most probable that some
other aspect of the stress response is responsible for immune enhancement by acute stress. Some
evidence implicates catecholamine release by the sympathetic nervous system (SNS). For
example, acute stress enhances the recruitment of leukocytes to sites of immune activation (24).
Furthermore, stress-induced increases in NFκB signaling were modeled in human monocyte
THP-1 cells after stimulation with norepinephrine, suggesting that SNS activation may be
responsible for stress-induced cytokine increases (21). Injection of alpha and beta adrenoceptor
antagonists block the stress-mediated increase in IL-1β levels, and adrenoceptor agonists are
sufficient to substitute for stress (25). Because the actions of catecholamines are exerted on a
shorter timescale than those of GCs, these results fit with the picture of acute and early stages of
5
stress as pro-inflammatory. For further review of the pro-inflammatory actions of stress, see (6,
8).
While many pro-inflammatory effects of acute stress can be attributed to catecholamines,
GCs are also involved. Several recent microarray studies examined the effects of the high-affinity
GR agonist dexamethasone on human peripheral blood mononuclear cells and found that amidst
expected anti-inflammatory changes in gene expression, there were many pro-inflammatory
systems such as chemokines, complement proteins, and cytokines with increased expression (26,
27). Several more in depth investigations have identified other situations where GCs increase
peripheral inflammatory responses and those are discussed next. The instances discussed are by
no means comprehensive; for instance, the stimulatory effects of GCs on the acute phase response
to inflammation are not discussed. See (1, 6, 28, 29) for further review of GC effects on
peripheral inflammation.
As noted, GCs have profound negative effects on the numbers of circulating leukocytes and
can induce apoptosis (30). However, a closer look reveals that neutrophil proliferation and
survival are enhanced by GCs (31, 32) and that the large-scale depletion of other leukocytes may
actually reflect extravasation to a site of injury, rather than cell death (24, 33, 34). The depletion
of circulating leukocytes by GCs has been described metaphorically as a deployment of immune
soldiers to the battlefront rather than to the barracks (35). Acute stress-enhanced migration of
leukocytes from the bloodstream to wound sites can be mimicked by either a physiological dose
of corticosterone or a GR agonist (but not an MR agonist), and is blocked by adrenalectomy
(ADX) or inhibition of GC synthesis (34). Notably, chronic stress also fails to have the same
effects on leukocyte redistribution and instead decreases baseline leukocyte numbers as well as
the magnitude of acute stress effects on leukocyte redistribution (33). These findings implicate
GR occupancy at GC levels typical of an acute stressor as both necessary and sufficient to
mediate enhancement of this particular immune response.
Acute stress levels of GCs have also been shown to enhance a model of peripheral
inflammation. Acute stress-induced enhancement of the DTH immune response is blocked by
6
adrenalectomy, which removes both GCs and epinephrine and is restored by supplementation
with corticosterone (the GC specific to rodents) levels that replicate the early stages of the acute
stress-response (33). This restoration is achieved to a greater extent by the replacement of
corticosterone than epinephrine (18). This surprising result may be due in part to the GCenhanced leukocyte migration following acute stress. In contrast, chronic exposure to
corticosterone or acute exposure to dexamethasone does not have the same effect and is antiinflammatory (18, 36). It is an important caveat that in some models of DTH and for some types
of stressors, GCs and catecholamines have not been seen to play an important role in stressmediated augmentation (36, 37).
The differences between the effects of acute and chronic stress on immune function
described by Dhabhar and McEwen (33) cannot be explained by the concentration-dependent
effects of MR and GR signaling. As noted, the pro-inflammatory migration of leukocytes to
wound cites is GR- rather than MR-mediated. Thus, the most meaningful way of understanding
the dichotomy between the pro- and anti-inflammatory actions is not so much the contrast
between MR and GR occupancy, as the contrast between acute and major/prolonged stressresponses. Presumably, these differences are manifested on the molecular level through the
duration and extent of GR occupancy.
GCs can have permissive and suppressive actions through mechanisms other than
activation of the opposing effects of MR and GR signaling. While GCs reduce the expression of
most pro-inflammatory cytokines, they often simultaneously increase the expression of the
corresponding cytokine receptor (38). This means that at low GC levels, cytokines are in
abundance, but their receptors are not and signaling is low. Conversely, at high GC levels,
receptors are in abundance, but cytokines are suppressed so signaling is low. Thus, cytokine
signaling is minimal at extreme GC concentrations and peaks at an intermediate GC
concentration (39). This observation has been shown to have functional importance in the
proliferative effect of IL-2 signaling on T cells. GCs decrease IL-2 production, but they expedite
expression of the IL-2 receptor. The result is that acute stress-levels of GCs enhance T cell
7
proliferation by accelerating the ability of IL-2 to signal through its receptor (40, 41). Notably,
exposure to chronic, high levels of corticosterone prevented this effect (42). The applicability of
this model to other cytokines and their receptors remains to be investigated. Weigers and Reul
have suggested that these effects of GCs on cytokine signaling might serve to optimize the course
of inflammation (38).
Although GCs inhibit many pro-inflammatory cytokines, they increase expression of at
least one potent pro-inflammatory mediator. The macrophage migration inhibitory factor (MIF)
was one of the earliest discovered pro-inflammatory cytokines.
Its release from intracellular
pools in macrophages and T cells results in the production of nitric oxide and TNF-α in an
autocrine stimulatory manner (43). MIF also has been implicated in DTH, as MIF-blocking
antibodies attenuate the DTH immune response (44). In humans, circadian plasma levels of GCs
correlate with a circadian pattern of MIF concentrations (45). Moreover, basal concentrations of
GCs increase MIF production in T cells and macrophages. Strikingly, physiologically relevant
concentrations of MIF can completely overcome the anti-inflammatory effects of GCs on other
aspects of inflammation in LPS stimulated monocytes (46). Furthermore, in a high dose in vivo
LPS challenge where GC treatment reduces mortality, concomitant treatment with MIF negates
the protection conferred by GCs. Taken together, these findings suggest a physiological role for
MIF in countering the anti-inflammatory effects of GCs (47).
GC enhanced MIF production is dose-dependent and fits the model that low-levels of GCs
are pro-inflammatory and high doses are anti-inflammatory. Low physiological doses of cortisol,
the species typical GC of humans and most primates, stimulate MIF secretion in murine
monocytes, while high doses suppress its production (46). Surprisingly, dexamethasone was as
stimulatory as cortisol at low doses and became anti-inflammatory at approximately the same
dose. Given the striking difference in GR binding affinity between these two GCs, this result
suggests that these effects of GCs on MIF are not necessarily controlled by the degree to which
GR is bound. For the most part, the actions of GCs on MIF fit the model that low doses suppress,
and high doses enhance inflammation.
8
The timing of GC exposure relative to an immune challenge has additional importance in
determining their action. When GCs are administered prior to an inflammatory challenge, they
have actually been found to augment the subsequent cytokine response. Treatment with GCs
twelve hours prior to LPS and IFN gamma stimulation increases NF-κB signaling and IL-6, TNFα, and nitrite production in a murine macrophage cell line (48). In a similar manner, exposure to
GCs twelve hours before LPS stimulation increases plasma levels of TNF-α and IL-6 in humans
(49). This enhanced response is not necessarily due to preparative cytokine receptor upregulation
because similar increases are not observed when GCs are given six hours prior to inflammatory
challenge. Finally, acute stress twenty-four hours prior to LPS increases plasma levels of TNF-α,
IL-6, and IL-1β in the rat; however the role of GCs in this outcome was not tested (50). In each
instance of GC pre-treatment, GCs had been removed from the media at the time of the
inflammatory stimulation in vitro, and in the in vivo model, GC levels had returned to baseline at
the time of LPS exposure. Thus, these effects are not due to the concentration of GCs at the time
of the inflammatory response.
Instead, Maier et al described stress as having a preparative
“priming” effect on the subsequent inflammatory response to an immune challenge (50-52).
It is clear that much work remains to be done to fully understand what governs when GCs
are pro- or anti-inflammatory in the periphery. As proposed by (6-10), GC type and concentration,
time course of exposure, and the activation state of the immune system are all significant. Several
trends emerge among the diverse GC enhancing effects on inflammation: low-doses, acuteexposure, and GC presence prior to inflammation all tend to augment it, whereas high-doses,
chronic-exposure, and GC presence after inflammation tend to suppress it. The evidence suggests
that in the context of an acute stressor and immune challenge, GCs play a complex role in
orchestrating an optimal and timely response that maintains control over the potentially damaging
effects of an over-extended immune response.
Before considering how GCs affect the CNS inflammatory response, the unique nature of
CNS injury is discussed. Unlike in the periphery, the CNS has considerably lower tolerance for
inflammation and the consequences can be devastating to neurons when it goes unchecked.
9
Inflammatory responses in the injured CNS are also regulated in different in several ways than
those in the periphery due to the cellular architecture of the CNS. The next section discusses two
common types of acute neurological injury, namely excitotoxicity and hypoxia/ischemia. These
neurological injuries can cause extensive neuron death and inflammation, and more of one often
leads to more of the other.
1.3 Mechanisms of neuron death and inflammation
1.3.1 Neuron death
While there are many different types of acute traumatic injuries to the CNS (e.g. stroke,
TBI, seizure, toxin exposure), many of the fundamental things that lead to neuron death are the
same. At the heart of most of these traumatic injuries is an energy crisis. This can be an active
process as during seizure where neurons fire in excess ultimately depleting their energy reserve,
or it can be a more passive process, as during stroke where the lack of blood flow to the injury
site interrupts the supply of energy. In addition, seizure is frequently a secondary consequence of
TBI and stroke.
The depletion of energy has multiple consequences for neurons. First, it disrupts the active
process of glutamate re-uptake from the synapse, leading to heightened extracellular glutamate
levels. This stimulates post-synaptic neurons to continue to fire, amplifying the energy crisis.
Second, it disrupts the active process of Ca2+ sequestration in response to neuronal
depolarization. When neurons fire, intracellular calcium levels rise as Ca2+ flows in via voltagegated Ca2+ channels and NMDA receptors. The excessive Ca2+ levels lead to over-activation of
intracellular processes that ultimately result in the increased production of reactive oxygen
species (ROS), protein mis-folding, cytoskeletal breakdown and membrane oxidation. ROS
scavenging is also energy dependent, further worsening the problem.
One last-ditch option that cells can choose if they still have enough energy to begin the
process is a programmed cell-death sequence. This process, called apoptosis, is frequently
initiated in neurons that were far enough away from the first traumatic events that they survived,
but were further compromised during secondary injury processes. Apoptosis is a good option for
10
a cell that cannot sufficiently recover because it helps to contain the release of neurotoxic and
immune-activating factors and debris that occurs during necrotic cell death. Although costly, it
protects neighboring neurons by giving cells a non-toxic way to dismantle their components into
self-contained vesicles. Ultimately, phagocytes will clean up these remaining compartments,
neutralizing the threat. The effect is proportional in expense and benefit to getting rid of nuclear
warheads by dismantling them and dealing with the waste properly instead of detonating them.
1.3.2 The inflammatory response to neuron death
These processes rapidly activate the immune response and local microglia extend their
processes and activate pro-inflammatory gene transcriptional programs. Microglia are activated
by perturbations in the extracellular environment like the presence or concentration of metabolic
substrates that are normally intracellular (e.g. ATP (53)), thus signifying an imbalance in neuronal
homeostasis. Microglia release pro-inflammatory cytokines such as IL-1beta, IL-6, and TNFalpha and chemokines such as MCP-1 and CINC-1. These molecules increase microglial
proliferation in an autocrine fashion, and signaling to other immune cells to recruit them to the
injury site.
Whether this inflammatory response is beneficial or detrimental to the neurons that survive
the initial insult is important for understanding how to appropriately treat neurological insults. At
least some attributes of the inflammation appear to be beneficial because TNF-α receptor
knockout mice have increased lesion size following neuronal insult (54). In contrast, TNF-alpha
and IL-1beta have been seen to increase neuron death (55), and at least IL-1beta is implicated in
worsening the extracellular accumulation of glutamate and intracellular calcium influx during
seizure (56). Other cytokines such as IL-6 and IL-10 have anti-inflammatory or neuroprotective
properties (55).
The cellular inflammatory response is similarly divided. Astrocyte proliferation may have
protective effects by isolating and containing the inflammatory response to the immediate injury
site. Supporting this, inhibiting astrocyte proliferation worsened neuron loss in focal brain injury
models (57). In contrast, peripheral granulocytes that are recruited release additional pro-
11
inflammatory cytokines, proteases, and oxygen radicals and these factors are responsible for
much of the damage that occurs (58). The most neurotoxic peripheral cells are granulocytes and
lymphocytes, whereas monocyte recruitment has been attributed a protective role (59).
Broadly, the early stages of a moderate inflammatory response in the injured brain can be
beneficial, whereas more extensive inflammation can adversely affect surviving neurons (60).
The next section describes how GCs manipulate these different aspects of the immune
response to CNS injury and what is currently known about the consequences of this for neuron
death.
1.4 Glucocorticoid regulation of immunity in the nervous system
1.4.1 Normative glucocorticoid effects in the brain
Glucocorticoids have an enormous and varied array of effects upon the brain. In particular,
hippocampal cells express high levels of MR and GR and are a primary target of GC actions.
Such actions include various salutary outcomes mediated by MR occupancy and/or the small
increase in GR occupancy seen in the early phase of the stress-response. These basal and
permissive effects include prevention of neuron death in the dentate gyrus, enhancement of
synaptic plasticity, and facilitation of hippocampal-dependent cognition (1). However, GC actions
in the hippocampus show the classic inverse-U pattern, in that heightened and prolonged GC
exposure has an array of deleterious effects, mediated by high GR occupancy. These include
impaired cognition and synaptic plasticity, inhibition of neurogenesis, atrophy of dendritic arbors,
and a reduction in spine density (61).
1.4.2 Glucocorticoids endanger neurons
The adverse consequences of prolonged elevation of GC concentrations are particularly
dramatic when coincident with neuronal injury. An extensive literature now demonstrates that
chronic stress and GCs, signaling through GR, can compromise the ability of neurons in the
hippocampus, cortex and striatum to survive a variety of neurological insults. Endogenous GCs
are protective during the injury caused by intraparenchymal LPS-injection (62), but are
detrimental during excitotoxicity (63). Exogenous GC treatment is more clearly deleterious and
12
can exacerbate neuron death from a wide variety of injury types including hypoxia-ischemia (64),
seizure (63), hypoglycemia (65), oxygen radical generators (66), and β-amyloid (67). These
“endangering” actions are exacerbated by the fact that many of these neurological insults
stimulate additional GC secretion. A fair amount is now understood about the mechanisms
underlying this endangerment. Specifically, GCs inhibit glucose uptake throughout these brain
structures (68), an effect that is not sufficient to be energetically disruptive by itself, but it
exacerbates the declines in ATP concentrations and mitochondrial potential caused by these
insults. As a result, affected neurons and astrocytes have less energy available for the costly tasks
of high-affinity re-uptake of the potentially excitotoxic glutamate from the synapse (69, 70). In
addition, affected neurons are compromised in their ability to sequester and/or extrude free
cytosolic calcium from post-synaptic neurons (71), and to contain and quench oxygen radicals
(66), all of which imperil neurons (72, 73).
In light of their anti-inflammatory properties, this suggests that chronic GCs should exert
two opposing effects in the injured brain. On the one hand, GCs directly compromise the ability
of various neuron types to survive a number of neurological insults, augmenting damage. In
contrast, by decreasing the extensive inflammation caused by these insults, GCs should decrease
damage (and in support of this, non-steroidal anti-inflammatory agents that blunt inflammation
are neuroprotective (74, 75)). This raises the issue of whether the endangering GC effect upon
neurons, or their protective inhibition of extensive inflammation is more dominant. The fact that
exposure to high stress concentrations of GCs augments the extent of neurotoxicity and
inflammation in response to these insults in vivo suggests that the endangering effects prevail
(72).
These findings come as somewhat of a surprise given the anti-inflammatory potential of
chronic GCs. However, it turns out that at least part of the explanation is that the antiinflammatory effects of chronic GCs in the brain are actually less consistent and potent than in
the periphery.
13
1.4.3 Glucocorticoid effects on inflammatory cells in the CNS
GCs are effective at reducing certain types of inflammatory cytotoxicity in the brain. For
example, treatment with dexamethasone decreases iNOS-mediated toxicity of microglia (76).
Direct infusion of LPS into the prefrontal cortex causes necrotic cell death accompanied by an
extensive inflammatory response and is commonly used as a model for inflammatory
neurological disorders. Nadeau and Rivest found that acute-stress levels of GCs were both
necessary and sufficient to lessen damage due to direct LPS infusion into the brain (62, 77). Thus,
basal GC levels and the physiological response to acute stress are anti-inflammatory and reduce
inflammatory cytotoxicity (78).
In contrast, there are a number of circumstances where GCs fail to have these antiinflammatory properties. For example, dexamethasone fails to reduce inflammation in
tuberculosis meningitis, although it does correspond with an improved prognosis (79) and fails to
decrease cytotoxic edema following cerebral ischemia. For a recent synthesis of clinical
applications where GCs do and do not have theraputic benefit see (80).
Most strikingly, long-term or high stress levels of GCs appear to exacerbate similar
measures of cellular inflammation. The chronic unpredictable stress (CUS) paradigm exposes
animals to a different stressor each day and animals subjected to CUS after direct LPS infusion
into the prefrontal cortex have markedly increased neuron and astroglia death. This effect of CUS
is reversed by blocking GR, suggesting that GC signaling is the critical component of CUS
involved (81). These findings are in contrast to the aforementioned protective effects of basal to
low-stress levels of corticosterone on LPS toxicity. Use of the same neurological insult highlights
the concentration-dependent effects of GCs on neurological inflammation. Notably, GCs also
dose-dependently inhibit astrocyte proliferation (82).
The pro-inflammatory GC effects extend to the extravasation and migration of circulating
inflammatory cells to a site of neurological injury. This may not come as a surprise given the
previously mentioned effects of GCs on these cells in the periphery. GC concentrations positively
correlate with the degree of inflammatory infiltration of granulocytes and macrophages to the site
14
of hippocampal infusion of excitotoxin (63). Blocking GC synthesis reduces the migration and
abundance of these cells at the injury site. Furthermore, rats exposed to chronic high stress levels
of corticosterone show accelerated and augmented microglial activation and migration of
peripheral cells to the injury site (63). Similarly, GCs released during CUS enhance microglial
activation following direct LPS injection to the prefrontal cortex (81). These findings implicate
chronic stress levels of GCs in enhanced activation of the immune cell response to neurological
cytotoxicity.
1.4.4 Glucocorticoid effects on inflammatory cytokines in the CNS
Levels of pro-inflammatory cytokines are an index of the inflammatory signaling cascades
activated by injury. GCs potently decrease the production of these molecules in the periphery and
also promote secretion of anti-inflammatory cytokines. In many cases, this is also true in the
brain. Adrenalectomized or GR antagonist-treated animals have been found to have increased
production of many cytokines in many different models (62, 83, 84). The studies that have looked
at the effects of chronic GC exposure have also found similar results for basal GCs: locking GCs
at basal levels augments excitotoxin-induced upregulation of IL-1β and TNF-α expression and
protein levels in the hippocampus (63). Pharmacological inhibition of GR increases TNF-α, IL-1
beta, and iNOS protein levels in rats given LPS peripherally (85). Moreover, in hippocampal
cultures, low concentrations of corticosterone inhibit excitotoxin-induced expression of IL-1β and
TNF-α (86). These findings suggest that basal levels of GCs have anti-inflammatory properties on
these pro-inflammatory cytokines during necrotic cell death.
However, in the context of sustained stress, or stress levels of GCs, a very different picture
emerges. CUS fails to suppress the increased expression of IL-1β and TNF-α following infusion
of LPS into the prefrontal cortex (81). Even more striking, CUS augments TNF-α, IL-1β, and
iNOS protein levels in hippocampus and cortex when LPS is administered peripherally, an effect
that is GR-mediated (85). Similarly, high stress concentrations of corticosterone exacerbate
excitotoxin-induced increases in IL-1β and TNF-α mRNA and protein levels both in vivo and in
vitro (63, 86). A more detailed analysis of GC concentration reveals inverse-U properties where
15
intermediate concentrations enhance inflammation the most. Moderate stress levels of
corticosterone are even more effective than high stress levels in augmenting excitotoxin-induced
expression of TNF-α and iNOS in vitro and in vivo (86, 87). Dexamethasone causes a dose- and
time-dependent increase in expression and activity of prostaglandin D2 synthase in neuronal
cultures (88) and dexamethasone or corticosterone treatment increases the pro-inflammatory
leukotriene synthesis enzyme 5-lipoxygenase in both the hippocampus (89) and macrophages
(90). Finally, dexamethasone alone or more robustly in conjunction with retinoic acid, causes
increased COX-1 expression in neuroblastoma cell lines (91).
1.4.5 Glucocorticoid effects on inflammatory transcription factors
As discussed, a classic anti-inflammatory GC action in the periphery is to decrease the
activity of NFκB. Commensurate with that, in unstressed rats, GR antagonists increase LPSinduced NFκB DNA binding activity in the frontal cortex and hippocampus (85). Moreover,
dexamethasone treatment decreases basal NFκB DNA binding activity in cortex and hippocampus
(92, 93) and also following LPS treatment (94) (however there are some caveats to this
interpretation that will be discussed later.)
In contrast, CUS increases LPS-induced NFκB DNA binding and protein levels in these
brain structures, a GR-mediated effect (85, 87). Moreover, CUS following direct LPS infusion
into the prefrontal cortex increases components of the MAPK signaling pathway involved in
inflammatory pathogenesis including phosphorylated JNK, p38, and ERK (81). These effects of
CUS are also GR-mediated. Taken together, these findings suggest that as for other aspects of
neuroinflammation, basal GC levels decrease pro-inflammatory transcription factor signaling
while high GC levels enhance it.
1.5 Overview of experimental hypotheses
These findings support the hypothesis that some types of GC exposures might increase
inflammation in the CNS instead of decreasing it. This suggests that the effects of GCs on the
inflammatory response are actually detrimental to neurons, and might contribute to GCendangerment of neurons. The research presented in this dissertation sought to address the
16
following gaps in our knowledge about the effects of GCs during acute CNS injury:
1.5.1 Primary hypotheses tested
The primary goal of this work was to determine the cellular and molecular nature of GCaugmented inflammatory responses during acute CNS injury. It might seem obvious that GCs
would affect immune cells directly to increase inflammatory responses, but additional possible
explanations include GC effects on neuron vulnerability or astrocyte functions which could
indirectly lead to an increased immune response. Hypothesis: Exogenous GC treatment
augments inflammatory responses in immune cells directly, not by increasing neuron death.
In addition to the effects of exogenous GC-treatment, these studies also sought to determine the
normative function of endogenous GCs on immune cells during acute neurological injury.
Hypothesis: Endogenous GCs released during the stress-response also activate immune
cells. This work is described in chapter 2.
The existing literature reports that GCs increase a variety of different pro-inflammatory
mediators like NF-kB. In addition, GCs fail to induce some of their normal anti-inflammatory
targets like MKP-1, IL-1ra, and IkBa in the forebrain (87). GCs might augment inflammation via
several different possible mechanisms: Hypotheses: (i) GCs increase chemokine production
thereby increasing immune cell recruitment; (ii) GCs decrease anti-inflammatory signaling
molecules; and, (iii) GCs increase NF-kB activation leading to increased transcription of
pro-inflammatory genes. This work is described in chapter 3.
The lack of insult specificity of GC neuron endangerment predicts that GCs are affecting
high-level processes that are broadly deleterious to neuron survival. In agreement with this
prediction, the known mechanisms for GC endangerment (impaired glucose uptake, glutamate reuptake, calcium extrusion, and ROS scavenging) could affect neuron death from a variety of
different injury types. The emerging evidence that GCs can increase immune responses
depending on the dose and temporal context of their exposure (7, 9) predicts that GC effects on
the immune response might also contribute to GC endangerment. Hypothesis: GC-augmented
17
inflammation during acute CNS injury is detrimental to neuron survival. This work is
described in chapter 4.
1.5.2 Secondary hypotheses tested
Several factors contribute to the difficulty in determining GC effects. First, GCs are
subject to multiple levels of regulation from the time they are released into the blood until when
they act on the GC receptor (GR) so blood hormone levels do not necessarily predict GR
signaling within a single cell of a complex tissue (e.g. differential binding to CBG carrier proteins
in the blood, local inactivation by 11beta-HSD enzymes). Second, GR signaling is likely to carry
different consequences depending on the cell type in which it occurs and there is widespread GR
expression throughout many cell types in the CNS (14). Third, GC effects are sensitive to
multiple factors like the timing and duration of GC exposure, whether the GCs are synthetic or
the endogenous ones released post-injury, and the specific tissue in which GCs are acting. Finally,
GCs affect not only the genes that they directly activate, but also those regulated by other
transcription factors that GR interacts with in trans, potentially masking the original cause of the
effect.
In the course of investigating the primary hypotheses about GC actions, several
additional hypotheses were addressed as a side-effect of developing the treatment parameters
necessary to resolve the confounding issues described above. The timing and duration of GC
exposure is a good example of this. Acute GC exposure prior to injury promotes increased
inflammatory responses to a variety of CNS and peripheral injuries in rodents (24, 48, 95) and in
humans (26, 49). This GC priming effect on the subsequent inflammatory response can be
prevented by either giving GCs after injury (23, 95) or giving them chronically (for weeks) (33).
Thus, at some point there is a switch from acute GCs stimulating immune responses to chronic
GCs suppressing immune responses (96). To begin to assess when this switch might occur, a subacute duration of GC exposure was chosen for exogenous GC treatments in these studies. These
GC treatments began three days prior to injury, so they also implicitly test the ability of this
timing and duration of GC exposure to affect inflammation and injury.
18
Most of the previous observations of GC-augmented inflammation were made in rats (63,
81, 85-87) and many examples of GC endangerment are from rats as well (97). To be able to
utilize the molecular genetic tools available in mice, these initial observations and parameters had
to be extended to the C57Bl6/J mouse line. The application of experimental parameters from rats
to mice had to be determined empirically, so an additional side-hypothesis that was tested was the
ability of GCs to augment inflammation and neuron death in mice.
19
Chapter 2: GC Activation of CNS Immune Responses
Because of the wide variety of GC effects on many different cell types, the unusual
finding that sustained exposure to GCs can actually augment CNS inflammation could be easily
dismissed as a side-effect of one of their many other possible targets. To determine whether GCs
were activating the immune response directly via their actions on immune cells themselves, two
separate approaches were employed simultaneously. The first approach was to develop a mouse
genetic model where GR could be deleted or overexpressed in specific cell types, and the second
approach was to manipulate hormone levels and use pharmacological antagonists to study the
same effects in rats. These two approaches were largely synergistic and are considered together
throughout the rest of this dissertation.
2.1 Experimental model
To determine the role of stress-level GCs in neurons and myeloid cells during acute
neurological injury, homozygous floxed-GR mice (98) were crossed to either LysM-CRE
(myeloid GR-KO, MGRKO) or CamKIIa-CRE (forebrain GR-KO, FBGRKO) driver mice. These
conditional KO mice have GR deleted throughout myeloid-lineage cells (MGRKO) (99) or
forebrain neurons (FBGRKO) (100) and were an extremely generous gift from Dr. Louis Muglia
(Vanderbilt University).
2.1.1 MGRKO mice have GR deleted in myeloid cells
Although MGRKO mice are described to have efficient myeloid-cell lineage deletion of
GR (99), we wanted to determine the efficiency of this recombination in CNS myeloid cells. To
quantify the percentage of myeloid cells lacking GR, these cells were isolated from WT and
MGRKO brains on a percoll gradient (101). The extracted mononuclear cells were pooled from 5
animals and GR levels were quantified by western blot. This analysis found that MGRKO mice
have 33.6% (±11.1% s.e.m.) reduction in total GR protein levels compared with their WT
littermates (Fig. 1a).
The extraction of myeloid cells from the whole brain includes many different subtypes of
these cells including perivascular macrophages and parenchymal microglia. To determine whether
20
parenchymal microglia in the regions we were interested in have efficient GR recombination in
MGRKO mice, tissue sections were stained with an antibody for the GR and a myeloid cell
marker. We used CD68 because it is expressed at low levels in inactive monocytes in cytoplasmic
granules and inactive GR would also be expected to be present in the cytoplasm. Many but not all
CD68+ cells were missing the GR in the MGRKO cortex (Fig. 1b) and in the MGRKO
hippocampus (Fig. 1c). Because only exon 1C of the GR is floxed in these mice, it is possible that
C-terminal fragments of the GR are still produced. We found a similar lack of GR
immunoreactivity in MGRKO brains stained with antibodies raised against the N-terminal or Cterminal end of the GR, arguing against this possibility (Fig. 1c).
Other investigations using the LysM-CRE driver have reported that its expression is
largely specific to myeloid cells, (99, 102) and we verified this for the brain regions we intended
to study in two additional ways. First, LysM-CRE mice were crossed to a ROSA-lacZ reporter
mouse that expresses the beta-galactosidase gene in cells undergoing CRE recombination. These
mice express this reporter throughout the cortex (Fig. 2a) and hippocampus (Fig. 2c), and antibeta-galactosidase staining co-localized with CD68+ cells (Fig. 2d). Finally, staining for the CRE
enzyme itself revealed a similar co-localization with myeloid cells (Fig. 2b).
2.1.2 FBGRKO mice have GR deleted in forebrain neurons
FBGRKO mice have efficient deletion of the GR in forebrain neurons after 4 months of
age (100). We tested the expression pattern of the CamKIIa-CRE promoter using the ROSA-lacZ
reporter and found widespread expression throughout forebrain neurons as early as 3 months of
age (Fig. 3a). Expression in the cortex and CA3 region of the hippocampus particularly, was
further enhanced by allowing animals to be 5 to 6 months old. To determine whether this CRE
driver could detectably reduce GR protein levels in the whole brain, infarcted hemispheres were
collected 24 h post-MCAO and analyzed by western blot with the N- or C-terminal anti-GR
antibodies (Fig. 3b). Although neurons were not separated from other cells, a decrease in GR
immunoreactivity of 33.6% (± 32.9, 95% CI, N-terminal antibody) or 15% (± 8.6, 95% CI, Cterminal antibody) was still detectable. We further verified this deletion by staining for the GR in
21
WT and FBGRKO littermates and found areas of missing GR expression in the pyramidal CA3
neurons in the hippocampus, even though these animals were only 2 months of age (Fig. 3c).
2.1.3 MGROV and FBGROV mice have inducible overexpression of GR in myeloid cells or
neurons
In addition to testing the need for GR signaling for various effects during acute CNS
injury, we also wondered whether some of those same effects could be enhanced by increasing
GR protein levels. To accomplish this, the same CRE-drivers were crossed to mice harboring
ROSA::loxP-stop-loxP-rtTA (from Jackson Labs) and CMV-TRE-rGR that was generously
provided by Dr. Sam Okret. These mice overexpress GR in myeloid cells or neurons upon
administration of doxycycline (DOX) (Fig. 4a). To verify overexpression, MGROV and WT mice
were given DOX in their drinking water for 7 days and monocytes were isolated on a percoll
gradient. Anti-GR immunoblotting determined that GR overexpression was 1.76 times greater
(±0.39, 95%CI) in microglia isolated from MGROV brains than in microglia from WT brains
(Fig. 4b). In neurons, GR overexpression was clearly higher in forebrain neurons of mice that had
been given DOX for longer than 3 days (Fig. 4c,d). These observations suggested an
experimental design where GC manipulations begin on day 5 of DOX to take advantage of the
increased GR protein levels and fresh DOX is given regularly including 24 h prior to any
significant manipulation (Fig. 4e).
One caveat to consider with this method of overexpressing the GR is its reliance on DOX
to induce the overexpression. DOX has neuroprotective and anti-inflammatory properties, so all
experiments were done with WT and overexpression mice receiving the same DOX treatments.
Despite this, infarct volumes were smaller in DOX-treated mice and DOX could interfere in
unknown ways depending on the treatment (e.g. if GC-treatment affected inflammatory
parameters that were also more affected by DOX).
2.1.4 GC levels in experimental system
Having verified that MGRKO and FBGRKO mice have efficient and specific deletion of
the GR in the expected cell types, we next wanted to determine how the various GC
22
manipulations affected GC levels. It is possible that MGRKO mice have an abnormal stress
response so to test this we exposed them to 2 hours of immobilization stress and measured serum
corticosterone levels (Fig. 5a). Immediately post-immobilization, MGRKO mice had slightly
lower cort levels, but by 2 h post-stress, GC levels were back to normal in both groups. GC levels
at 6 h were around 10 µg/dL, consistent with a normal rise in circadian release, and there were no
lasting effects by 24 h. These data suggest that the stress response in MGRKO mice is largely
intact. In contrast, FBGRKO mice have elevated GC levels (100), as would be expected from
deleting the negative feedback from GR signaling in hippocampal neurons (1).
Comparison of WT and KO mice was used to assess the role of the endogenous GC
signaling in each cell type. To measure the role of exogenously administered GCs, animals were
implanted with a 10 mg s.c. corticosterone pellet. Corticosterone (CORT) was chosen instead of
pharmacological GCs because the latter are targets of the mdr-1 protein in the brain and could
actually lead to a decrease in corticosterone centrally, as less endogenous GCs are produced in the
presence of synthetic GCs (15). To determine the blood CORT levels achieved following pellet
implantation, weight-matched WT mice were given pellets and serum CORT levels were
measured 1, 3, or 6 days later (Fig. 5b). CORT levels exhibited a linear decrease over time, with a
pooled slope of -10.96 µg/dL/day, and a pooled Y-intercept of 102.0 µg/dL. A separate cohort of
WT mice was measured post-KA to see if blood CORT levels were still elevated 24 h after injury
(4 days post-pellet) or 72 h later (6 days post-pellet). Pellet-implanted animals still had serum
CORT levels equivalent to if they had not received KA and were still substantially elevated over
no pellet controls after 4 or 6 days (Fig. 5b). Finally, KA treatment on day 3 did not significantly
alter the slope of the line Together, these results defined the exposure to exogenous GCs that was
achieved by s.c. pellet implantation and helped to suggest the timing of the experimental design
in Fig 4e.
2.1.5 Injury models
After establishing the parameters of the GC treatments, mice were next subjected to two
different acute neurological injuries. The two injuries selected were the excitotoxin KA and the
23
distal model of MCAO. These injuries affect the hippocampus and the cortex, respectively, brain
regions with high GR abundance (15), and where GCs have been found to increase inflammation
and neuron death (63, 81, 85, 87). Furthermore, they share some but not all of the same
pathological characteristics so their comparison might indicate how generalizable the effects of
GCs are across injury models.
Many of the original studies that observed GC-augmented inflammation and neuron
death were performed in rats, and the ability to observe these same responses in mice was not a
guarantee. Thus, several experiments in rats were conducted in tandem with the mouse
experiments. This parallel set of investigations used the same KA model of excitotoxicity that
was used in the mice where KA was injected stereotactically into the hippocampus. This provided
a unique opportunity to investigate similar questions about how GCs affect CNS injury and
inflammation in separate model organisms using very different experimental approaches. In rats,
the manipulations were gross physiological changes or pharmacological interventions, whereas in
mice the manipulations were largely genetic. The fact that these very different investigation styles
yielded similar conclusions was nothing short of remarkable, and throughout the rest of this
dissertation the findings from rats and mice are presented together, organized by the inflammatory
parameters that were measured. The experimental design timeline for the rat studies is shown in
Fig. 15a, and a similar timeline for the mouse experiments is shown in Fig. 16a. Mouse
overexpression experiments utilized the slightly modified experimental design indicated in Fig.
4e.
2.2 Results
2.2.1 GR signaling in myeloid cells increases immune cell activation and recruitment
To measure immune cell activation in KA-treated mice, sections were stained for CD68, a
myeloid cell marker that is upregulated in activated cells (Fig. 6a). In WT mice, exogenous GCtreatment increased CD68+ area as well as the pixel intensity of that signal 72 h after KA (Fig.
6b,c). MGRKO mice still had the same GC-mediated increase in CD68+ area but had a reduced
pixel intensity of the CD68+ signal within that area (Fig. 6c).
24
To measure immune cell activation in mice post-MCAO, myeloid cells were isolated
from the infarcted hemisphere on a percoll gradient and labeled with CD11b+, CD45, and Ly6
antibodies for analysis by flow cytometry (Fig. 7a). Cells that were CD45+ and CD11b+ (Fig. 7b)
were gated based on their forward and side scatter profile and split into three populations (Fig.
6d–f). Granulocytes are Ly6G+CD45low (P1) whereas activated mononuclear cells are Ly6GCD45high (P2). The largest population, however, was the resting microglia Ly6G-CD45low (P3)
outside of the injury site that were extracted along with the activated cells. These populations
were quantified as a percentage of the parent gate (G2) and to control for variability between
extractions, the fold increase in each population % was expressed relative to that population’s %
in the contralateral hemisphere (representative dot-plots are shown in Fig. 7c). This yielded an
estimate of the change in each population due to the stroke infarct, or an approximation of the
recruitment of these cells into the injury site.
GCs affected the recruitment of granulocytes and monocytes similarly, but exogenous
and endogenous GCs had divergent effects on immune cell recruitment. Treatment with
exogenous GCs suppressed monocyte recruitment (Fig. 6h) and there was a trend towards the
same effect for granulocytes (Fig. 6g). This is the standard immunosuppressive effect that longer
exposure to GCs can have on immune cell recruitment, and it required GR signaling in monocytes
to occur because GC-treatment did not suppress granulocyte or monocyte recruitment in MGRKO
mice. In contrast, GR signaling in myeloid cells was actually required for normal levels of
granulocyte recruitment (Fig. 6g) and there was a trend towards the same effect for monocytes
(Fig. 6h). Thus, while GC-treatments suppressed peripheral inflammatory cell recruitment via GR
signaling, the recruitment of these cells also required normal GR signaling to occur. This analysis
did not detect any differences in the resting microglial cell populations (Fig. 6i).
To see if GR is also required in immune cells for their recruitment outside of the CNS,
mice were tested in a model of peripheral inflammation. Leukocytes were collected in Zymosan
A after a week of inflammation induced by s.c. injections of sterile air. More than a 50%
reduction in leukocyte recruitment was detected in MGRKO mice (Fig. 6j), suggesting that GR
25
signaling in immune cells is required for their migration to many injury sites not just those in the
CNS.
Finally, because GR signaling was necessary for both stimulation and suppression of
immune cell recruitment, the effects of GR overexpression on these populations were measured.
GR overexpression alone did not significantly alter any of the cell populations (Fig. 7d–f). Thus,
GR signaling in myeloid cells is necessary but probably not sufficient to mediate GC-augmented
inflammatory responses to acute CNS injury.
2.2.2 GCs increase immune cell activation in the injured rat hippocampus
Sub-acute (72 h) GC exposure prior to injury endangers neurons and increase
inflammation following KA-induced excitotoxicity in rats (63). This injury model was therefore
selected to study GC-augmented inflammation and its relationship with neuron death in rats.
Previous examples of GC-augmented inflammation compared intact rats with adrenalectomized
(ADX) rats, where the endogenous source of GCs is removed. We wanted to ensure that GCaugmented inflammation was not simply an artifact caused by ADX so in this study we compared
ADX rats given either low or high GC replacements to isolate GC-dose as a variable. These
treatments lock blood GCs at either basal (low-GC group, 2-4 µg/dL) or stress-levels (high-GC
group, 30 µg/dL (87)). Three days later during kainic acid-induced excitotoxic injury, rats were
also given injections of either indomethacin or minocycline (Fig. 8a). Three days post-injury,
brain sections were stained for activated infiltrating and resident immune cells (CD11b/c) and
astrocytes (GFAP) in the injury site (Fig. 8b-e). Intact rats were also evaluated as a reference;
however, the primary comparison was between low-GC and high-GC rats where GC exposure
was a constant, controlled variable.
In support of GCs augmenting cellular inflammation, GCs doubled the CD11b/c signal
(Fig. 8b,c). This increase could be blocked by concurrent treatment with minocycline, an inhibitor
of microglia activation. Minocycline also reduced GFAP signal in astrocytes by approximately
one-half (Fig. 8d,e). In contrast, the non-selective COX-inhibitor indomethacin had no effects on
CD11b/c or GFAP signal. These results confirm that GCs are sufficient to increase cellular
26
inflammatory responses and show that GC-augmented inflammation can be suppressed by
minocycline but not indomethacin treatment.
2.3 Discussion
2.3.1 Endogenous and exogenous GCs affect different aspects of cellular inflammation postinjury
In agreement with these observations in KA-treated rats and similar published findings
in rats (63), GC-treatment increased CD68 area by 1.5 times and CD68 intensity by 1.75 times in
WT mice. This observation supported the idea that GC-augmented inflammation is a general
enough phenomenon that it could be observed in multiple rodent models.
Removing GR in myeloid cells did not suppress the GC-mediated increase in CD68 area,
but it did reverse the pixel intensity of that signal. All treatment groups were counterbalanced
during the IHC staining protocol supporting pixel intensity as a rough estimate of CD68 levels
and CD68 area as a rough estimate of the size of the microglia population. This suggests that GR
signaling is required in these cells for GCs to activate immune cells, but not for cell proliferation
in response to injury.
Following MCAO, exogenous GC-treatment suppressed peripheral myeloid cell
recruitment and had a trend towards the same for granulocytes (Fig. 6g,h). This is in agreement
with literature demonstrating similar suppressive effects of exogenous GC treatments on
leukocyte recruitment to other injuries (18, 24). Furthermore, this GC-mediated suppression of
recruitment required GR in the immune cells to occur as the LysM-CRE driver is heavily
expressed in granulocytes and monocytes alike.
Although myeloid-cell GR signaling was necessary for GCs to suppress their recruitment,
it was also necessary for their recruitment in the first place. MGRKO mice had a significantly
lower number of granulocytes and a trend towards fewer monocytes (Fig. 6g,h), suggesting that
stress-activation of GR in myeloid cells normally is required for their recruitment to the injury
site. In support of this not being an effect limited to the CNS, intact GR signaling in myeloid cells
was also required for proper leukocyte recruitment to a peripheral site of inflammation (Fig. 6j).
27
These results suggest that stress-mediated GR signaling in myeloid cells is required at a proper
level or during a proper time for those cells to migrate to an inflammatory site.
Thus, endogenous GR signaling is required in myeloid cells for proper cell activation and
recruitment to the injury site. Exogenous GCs increased cellular activation in the KA model and
suppressed peripheral cell recruitment in the MCAO model.
2.3.2 GCs increase immune cell activation in the injured hippocampus
To see whether the amount of inflammatory cell activation post-injury could be
manipulated pharmacologically, rats were treated with two anti-inflammatory drugs with
alternative mechanisms of action, indomethacin and minocycline. Indomethacin is a non-selective
inhibitor of the COX enzymes that are the rate-limiting step in the production of prostaglandins.
Two months of treatment with indomethacin only slightly reduces microglial activation postirradiation (103) so we did not expect this drug to have a large effect in this injury model. In
contrast, minocycline is reported to be a potent inhibitor of microglial activation (104). This fits
with the finding that CD11b/c+ area was slightly (but not significantly) reduced by indomethacin
and was robustly suppressed by minocycline (Fig. 8b,c).
Comparison of low-GC and high-GC rats revealed that higher GC doses increase CD11b/
c+ area (Fig. 8b,c). This agrees with previous reports that GCs increase immune cell activation
both basally (105) and in this same excitotoxicity model (63). Previous work using this same
excitotoxicity model found that ADX reduced inflammatory cell activation and intact rats given
additional GCs had even greater inflammatory cell activation (63). These effects could potentially
be due to compensatory decreases in endogenous GC synthesis (for the comparison of intact with
intact given additional GCs) or they could be an artifact of ADX (for the comparison of intact
with ADX). The present finding that GC-augmented inflammation still occurs when GC dose is
an isolated variable in ADX rats argues against these alternative possibilities.
For these ADX studies, only certain comparisons could be used to reach these
conclusions. Intact rats release endogenous GCs in response to the stress of receiving KA-injury,
and also retain their normal circadian rhythm in GC release compared with ADX rats. Although
28
ADX rats were given physiologically relevant stress-GC levels (106), implanted GC pellets
cannot mimic the pulsatile release of endogenous GCs (107), and ADX also removes
catecholamines. Comparison between ADX rats was used to draw conclusions about the effects of
constant stress-level GCs relative to constant basal GCs. Comparison of either of these groups to
intact rats is included as a useful reference, but there are several uncontrolled variables between
these groups so care must be taken when interpreting the underlying cause of any differences.
Thus, the primary conclusions drawn from this study are based on comparison of low-GC and
high-GC rats where GC exposure was a constant, controlled variable.
2.4 Conclusions
The goal of these investigations was to measure how endogenously released GCs and
exogenous GC treatments affect the immune response to acute CNS injury. Despite being
exposure to the same GC, corticosterone, endogenous and exogenous GCs could differ in their
effects due to differences in the timing and duration of exposure. Endogenously-released GCs
were pro-inflammatory in that they increased CD11b/c+ signal in rats post-KA and they increased
granulocyte and, likely, monocyte recruitment in mice post-MCAO. Exogenous GC-treatment
also had pro-inflammatory effects including increasing CD11b/+c signal in rats post-KA and
CD68+ signal in mice post-KA. GCs augmented the area and the intensity of the CD68+ signal
but only the increase in intensity of the signal required GR signaling in myeloid cells to occur.
These effects strongly support both of the original hypotheses that exogenous GC treatments
augment inflammatory responses in immune cells directly, and that endogenous GCs released
during the stress-response also activate immune cells. Finally, these results suggest an explanation
for the immunosuppressive effects that are often reported for exogenous GCs: such treatment
strongly reduced the recruitment of peripheral monocytes and likely granulocytes to the infarct
post-MCAO and this effect of exogenous GCs was myeloid-cell GR-dependent.
Because the GC-augmented cellular activation required GR signaling in myeloid cells, it
suggested the possibility that GCs in fact were activating myeloid cells directly, continuing to
challenge the established dogma that sustained stress level GC exposure is immunosuppressive.
29
In an effort to further understand the basis for these effects, the effects of GCs on a number of
inflammatory mediators were measured. The next chapter explores some of the potential
molecular explanations for how GCs are increasing cellular inflammatory responses to acute CNS
injury.
30
Chapter 3: Mechanisms of GC-Augmented Immune Responses
3.1 Results
Next, several non-competing hypotheses about the molecular mechanisms behind this
GC-augmented cellular inflammation were tested: (i) GCs increase pro-inflammatory cytokine
and chemokine production thereby increasing immune cell activation; (ii) GCs decrease antiinflammatory signaling molecules; and, (iii) GCs increase NF-kB activation. Any or all of these
could be explanations for the increased cellular inflammatory response described in chapter 1, as
they all reflect GC actions that are counter to their normal anti-inflammatory effects. Even a
failure to suppress pro-inflammatory cytokines or a failure to induce anti-inflammatory cytokines
could potentially explain GC-augmented cellular responses since GCs normally play a role in the
recovery from the initial inflammatory response and GC-augmented inflammation could simply
be a failure of this recovery to occur.
3.1.1 Hypothesis (i): GCs increase pro-inflammatory cytokines and chemokines
One possible explanation for GC-enhanced immune activation is that GCs increase the
signals that activate or recruit immune cells. The injured CA3 region was micro-dissected at
several time points post-injury to measure the levels of two principal chemokines involved in
recruiting monocytes (CCL2) and granulocytes (CINC-1) (Fig. 9). GCs did not significantly alter
the release of either chemokine, suggesting that GC-augmented inflammation occurs via another
mechanism. ADX did affect the release of these chemokines, however, slightly hastening their
production regardless of GC treatment. This re-emphasizes the need to control for ADX as a
confounding factor when attempting to study the effects of GCs.
Previous reports of GC-augmented inflammation detected increased production of the
pro-inflammatory cytokine IL-1beta in response to an inflammatory challenge (63, 81, 85-87) and
this might explain an increased cellular response. Prior to injury, both low and high GC rats had
2-3 times more IL-1beta than intact rats, but high GCs did not increase IL-1beta levels relative to
low GCs (Fig. 10c). This suggested that while they may not stimulate more IL-1beta production
basally, GCs are at least partially deficient in their ability to suppress IL-1beta production. To see
31
if this deficiency involves GR signaling in myeloid cells, IL-1beta and IL-6 levels were measured
in the different KO mice.
IL-1beta and IL-6 levels were measured in MGRKO and FBGRKO mice and their WT
littermates at 12 h post-KA and 24 h post-MCAO to determine the cell-specific role of GR
signaling in their production (Fig. 12a–d). Following KA, exogenous GC treatments slightly, but
not significantly, increased IL-1beta and IL-6 levels and this effect was significantly enhanced in
MGRKO but not FBGRKO mice (Fig. 12a,c). Following MCAO, exogenous GC treatments
suppressed IL-6 and had a trend towards the same effect on IL-1beta. In MGRKO mice, this
exogenous GC suppression of IL-6 did not occur and the trend for IL-1beta was reversed
(although still not significant) (Fig. 12b,d). FBGRKO mice had only minor differences in their
cytokine response, irrespective of treatment or injury type.
Finally, the effects of GCs on iNOS and COX-2 were measured because these enzymes
are rate limiting in the synthesis of NO and prostaglandins in response to injury. T-cell GR is
required for a lethal activation of COX-2 during peripheral inflammation (98). During stroke, the
production of NO and prostaglandins is particularly detrimental and GCs can activate either of
these systems with potentially deleterious consequences for neurons (22, 108-110). GR signaling
was not required for normal iNOS induction post-stroke (Fig. 12e). In contrast, COX-2 induction
was reduced in FBGRKO but not MGRKO mice relative to WT controls (Fig. 12f). These results
demonstrated that GR signaling in neurons is necessary for COX-2 induction post-MCAO.
These findings indicated that GCs are unlikely to increase cellular inflammatory
responses by increasing the production of pro-inflammatory signaling molecules. In some cases,
GCs suppressed these molecules, but in more situations GCs had little effect.
3.1.2 Hypothesis (ii): GCs decrease anti-inflammatory signaling molecules
An alternative explanation for GC-enhanced immune activation is that GCs decrease antiinflammatory signals prior to injury. In the CNS, reduced expression of CX3CL1, CD200, and
CD22 is associated with increased immune cell responses (111-114). To determine whether the
GC treatments affected the expression of these genes, their levels were measured by qPCR
32
immediately prior to KA-injury (Fig. 10a). The expression of TGF-b1 was also measured because
GCs normally induce this gene as a component of their anti-inflammatory actions (1, 14).
In agreement with the cellular activation profile post-injury (Fig. 8b,c), GCs significantly
reduced the expression of several key anti-inflammatory signaling molecules in the uninjured rat
hippocampus. Relative to high-GC rats, low-GC rats had 2 to 4-fold increases in the expression of
CX3CL1, its receptor CX3CR1, and CD22 (Fig. 10b). Furthermore, GCs failed to induce TGFbeta or IL-1ra, targets that GCs activate as a component of their anti-inflammatory actions and the
recovery from the inflammatory response. Thus, GCs suppressed several anti-inflammatory
signaling molecules and failed to induce some of their known anti-inflammatory targets in the
hippocampus.
GC-mediated reductions in CX3CL1 were of particular interest because of the magnitude
of the effect and the possibility that such reductions could increase the toxicity of the
inflammatory response to neurons (115). To determine the cell-specificity of the effects of GCs on
CX3CL1, and also to measure GC effects on this chemokine post-injury, western blots for
CX3CL1 were quantified at 12 and 72 h post-KA or at 24 h post-MCAO in WT and MGRKO and
FBGRKO mice (Fig. 11). By 12 h post-KA, GC treatments in intact WT mice had the same effect
in reducing CX3CL1 as GC treatments in ADX rats. This effect could be blocked by knocking out
the GR in either neurons or myeloid cells. After 72 h, GR signaling in myeloid cells but not
neurons was required to maintain reduced CX3CL1 levels (Fig. 11b,e). These results demonstrate
that GR signaling in myeloid cells maintains reduced CX3CL1 levels in the hippocampus and that
GCs require intact GR signaling in both neurons and myeloid cells to suppress CX3CL1.
In contrast, at 24 h post-MCAO, CX3CL1 levels were highly increased by GC treatment
(Fig 11c,f). These increases occurred to nearly the same degree in WT and KO mice, regardless of
whether GR was deleted in neurons or myeloid cells. Furthermore, GR signaling in both neurons
and myeloid cells was necessary for increased CX3CL1 levels. Thus, GCs stimulate CX3CL1
levels in the cortex post-ischemia, but they suppress and maintain a reduced CX3CL1 level in the
hippocampus post-KA.
33
3.1.3 Hypothesis (iii) GCs increase NF-kB activity
If GCs decrease anti-inflammatory signals and increase cellular activation, then they
might be expected to do so by altering the transcriptional activity of one of their primary targets,
NF-kB. We measured the nuclear activity of the p65 subunit of NF-kB via EMSA immediately
prior to KA-injury (Fig. 10d) but there were no differences in baseline NF-kB activity following
GC-treatment. Thus, the expression changes in anti-inflammatory cytokines are not likely to be
mediated via changes in NF-kB DNA binding activity.
GCs mediate their classical anti-inflammatory effects on NF-kB activity either directly by
protein–protein interaction or indirectly by inducing the inhibitor of NF-kB, IkBa (14). The
inability of GCs to alter baseline NF-kB activity in the hippocampus prompted us to investigate
whether our GC treatments induced IkB in this brain region. Indeed, they did not (Fig. 10e), and
as a control we assayed IkB expression in the frontal cortex, another brain region with abundant
GR expression and found normal IkB induction by GCs there (Fig. 13c,d). This suggested that
GCs might have an impaired ability to inhibit NF-kB in the hippocampus.
To see if GCs could affect NF-kB activation post-injury we measured its activity at the
peak of the immune response (24 h post-KA in Fig. 9). High-GCs suppressed NF-kB activation
by 25.4% (±22.8%, 95%CI) relative to low-GCs (Fig. 13,b), demonstrating that GCs can at least
modestly reduce NF-kB activation in the hippocampus. In contrast, however, several different
intermediate GC levels (corresponding to moderate stress) increasingly activated NF-kB. This
suggested that NF-kB activation depends on GC dose, and given that NF-kB activation also has
different consequences depending on the cell it is activated in, we next measured the effects of the
cell-specific GR manipulations on NF-kB activation.
Endogenous and exogenous GCs increased p65 levels after both injuries in mice.
Endogenous GR signaling is necessary for normal p65 activation in WT animals regardless of
whether it happens in myeloid cells (Fig. 14b,c) or neurons (Fig. 14e,f) at any time point
following either injury. Exogenous GC treatment also potentiated p65 activity in both myeloid
cells and neurons individually, and removing GR in either cell type was insufficient to completely
34
reverse the effect. By 72 h post-KA, or 24 h post-MCAO, however, the MGRKO mice had
significantly suppressed p65 levels, suggesting that GR in these cells is necessary for more
prolonged p65 activation. These results are consistent with both endogenous and exogenous GCs
stimulating p65 activation in myeloid cells and neurons independently.
3.2 Discussion
3.2.1 Hypothesis (i): GCs do not increase pro-inflammatory cytokines and chemokines
Given that GCs doubled immune cell activation, a likely possibility is that they do so by
accelerating and/or increasing the magnitude of chemokine release in the injury site. The
predominant chemokine source is inflammatory cells themselves, so increased cellular activation
in the presence of GCs might correspond with higher chemokine levels. Instead, we found that
CCL2 and CINC-1 levels post-injury were unchanged by GC-treatments (Fig. 9) suggesting
either that GCs are suppressing chemokine production, or that greater cellular activation does not
equate with more chemokine production. From this we concluded that GCs must affect other
mechanisms to increase immune activation.
In other examples of GC-augmented inflammatory responses, GCs increase the proinflammatory cytokine Il-1beta in the uninjured hippocampus (63) and after generalized
inflammation from peripheral or centrally injected LPS (81, 85, 87), or excitotoxicity (63, 86). If
GCs increase the production of this cytokine prior to the injury, it might lead to augmented and
potentially more neurotoxic immune responses, particularly in the context of reduced antiinflammatory signaling by molecules like CX3CL1 (115). In this study both ADX groups had 2-3
times the IL-1beta protein levels in the uninjured hippocampus, irrespective of GC dose. This
finding is in agreement with other observations that ADX can increase pro-inflammatory
signaling molecules in the brain (1, 116); however, high GCs should suppress IL-1beta relative to
low if GCs were fulfilling their anti-inflammatory potential.
IL-1beta and IL-6 were increased by exogenous GC treatments in MGRKO but not
FBGRKO mice (Fig. 12a,c). This result suggests that GR signaling in myeloid cells normally
suppresses cytokine levels post-KA, a commonly observed anti-inflammatory GC effect.
35
Interestingly, GR signaling only was necessary for this reduced cytokine response in the presence
of additional GC treatment. This means that during normal GC-augmented inflammation, GR
signaling in myeloid cells is still suppressing some of the increased inflammatory response.
Following MCAO, exogenous GC treatments suppressed IL-6 and had a trend towards the same
effect on IL-1beta, another typically reported anti-inflammatory effect. This exogenous GC
suppression of IL-6 and possibly IL-1beta as well, required GR signaling in myeloid cells to
occur (Fig. 12b,d). Thus, GR signaling in myeloid cells is required for cytokine suppression in
both injury models, a classic anti-inflammatory GC effect. Finally, GCs were considerably less
able to suppress these pro-inflammatory cytokines during KA injury than during MCAO. This
difference in effect could be due to the considerably larger immune cell population involved in
the response to stroke compared with excitotoxicity.
Although GR signaling in neurons was necessary for increased COX-2 levels, it is
unlikely that this effect could explain GC-augmented inflammatory responses because GCs could
still increase cell activation in the presence of indomethacin, a COX-2 inhibitor (Fig. 8c). The
lack of effect in FBGRKO mice indicated that GR signaling in neurons does not affect IL-1beta
or IL-6 production in these injuries. This result is not surprising given that it would have
otherwise had to be due to an indirect effect of neuron GR signaling on cytokines that originate in
myeloid cells.
Thus, these investigations found little evidence to support the hypothesis that GCs
augment inflammation by increasing the production of pro-inflammatory signals. GCs did not
increase pro-inflammatory chemokine and cytokine production, nor did they augment
inflammatory responses via altering the levels of rate-limiting enzymes in the synthesis of these
molecules. During MCAO, GCs were able to suppress IL-6 and probably IL-1beta in a manner
that required GR signaling in myeloid cells to occur. During KA, GCs were significantly less
suppressive of the production of these pro-inflammatory cytokines, despite the fact that GR
signaling in myeloid cells was still required for that suppression, but this level of analysis did not
reveal any clear explanations for how GCs are increasing inflammatory responses described in
36
chapter 1. In general, these investigations found support for traditional anti-inflammatory GC
effects with a few notable exceptions where GCs simply failed to have this expected action.
3.2.2 Hypothesis (ii): GCs decrease anti-inflammatory signaling molecules
In contrast, GCs were found to suppress expression and protein levels of several antiinflammatory cytokines and chemokines (Fig. 10b, 11). The anti-inflammatory genes that GCs
suppress are at least partially redundant in their ability to restrain the immune response to injury.
CX3CR1 is a receptor present on immune cells and its ligand CX3CL1 is a soluble factor
expressed in neurons throughout the CNS (117). CX3CR1 knockout mice have a more neurotoxic
microglial response to a number of different CNS injuries (115). CD22 is less-extensively
studied, but is a soluble ligand for the pan-leukocyte marker CD45 and it inhibits microglial
release of cytokines in response to LPS (113). That GCs affect these molecules in the CNS is not
unprecedented as they have been seen to reduce CX3CL1 expression in respiratory epithelial cells
(118). Thus, exposure to GCs reduces the expression of several critical inhibitory inflammatory
signals in the hippocampus. Following an injury, GCs suppress CX3CL1 by displacing NF-kB
from the CX3CL1 promoter (118), but in the absence of a stimulus NF-KB is not present on this
promoter, suggesting that in the absence of injury, GCs must suppress CX3CL1 via an alternative
method and that GC suppression of CX3CL1 expression might be even greater post-injury.
Indeed, GCs did suppress CX3CL1 levels in the hippocampus at 12 h post-KA and GR
signaling in myeloid cells was necessary to maintain this reduced CX3CL1 expression at 72 h
post-KA. Furthermore, these results demonstrated that GC treatments can suppress CX3CL1 in
intact mice just as they do in ADX rats. These actions could lead to an increased cellular
inflammatory response to hippocampal injury. The GC-augmented CX3CL1 levels in the
ischemic cortex are also likely to contribute to increased immune cell activation since this
chemokine is thought to mediate leukocyte chemotaxis to the injury site (119). Given the minimal
effects of GCs on CCL1 and CINC-1 (Fig. 9), CX3CL1 is an attractive alternative chemokine
explanation for the ability of endogenous GCs to stimulate leukocyte recruitment post-MCAO.
Despite these opposing effects of GCs on CX3CL1 post-KA and post-MCAO, both alterations
37
could potentially harm neurons because CX3CL1 deficient mice had smaller infarct volumes
post-MCAO (120) but CX3CR1 deficient mice exhibited more neurotoxic inflammatory
responses (115).
The observation that GCs have a reduced ability to suppress IL-1beta following KA in
multiple studies including the present findings, suggested that in addition to suppressing antiinflammatory signals, GCs might also have a deficit in their ability to inhibit pro-inflammatory
signals. Supporting this, GCs failed to suppress IL-1beta in rats, and they also failed to induce
IL-1ra, the anti-inflammatory antagonist of IL-1beta (Fig. 10b) and a normal target of their antiinflammatory actions (87). To further investigate this possibility, the third hypothesis addressed
the transcriptional activation of NF-kB, one of the primary targets that GCs inhibit as a
component of their anti-inflammatory actions.
3.2.3 Hypothesis (iii) GCs increase NF-kB activity
Aside from their suppression of several anti-inflammatory genes, the evidence that GCs
also fail to induce some of their normal anti-inflammatory targets suggested that GCs might be
impaired in other aspects of their normal anti-inflammatory mechanisms like NF-kB inhibition.
GC treatments did not affect basal NF-kB activation (Fig. 10d) but did affect post-injury NF-kB
DNA-binding (Fig. 13a,b), and primarily activated NF-kB in the hippocampus, except for a
concentration-dependent inhibition that occurred in animals with blood GC levels higher than 30
µg/dL (high GCs), (Fig. 13b). Despite this dose-dependent inhibition, high GCs did not reduce
NF-kB activation below the level of intact rats, an anti-inflammatory effect that would be
expected from such constant high-dose exposure. This matches with other observations following
peripheral LPS where high GCs did not suppress NF-kB activity below intact levels in the
hippocampus, but were more than two-times more suppressive of NF-kB in the frontal cortex
(87). Correspondingly, GCs could induce IkBa better in the frontal cortex than in the
hippocampus (Fig. 13c,d), suggesting that IkBa induction might be necessary for effective
suppression of NF-kB by GCs in the hippocampus. GC induction of IkBa is specific to certain
38
cell types (14), so this difference could extend to brain regions populated by distinct neurons and
glia.
In the rat KA model, intermediate GCs were the most stimulatory of NF-kB activity (Fig.
13a,b), demonstrating an inverse-U response where moderate stress has the largest proinflammatory effects. This matches with similar augmented inflammatory responses observed
during chronic mild stress (81, 85) or daily GC injections, both of which would frequently expose
animals to intermediate concentrations (63). An identical dose-response analysis of GCtreatments following peripheral LPS identified the same enhancement of LPS-induced NF-kB
activity by intermediate GC concentrations (87).
Activation of NF-kB leads to an increased inflammatory response and inhibition of this is
a primary means by which GCs are anti-inflammatory so the finding that GC-treated mice have
increased p65 nuclear localization is in opposition to this. The CD68 cell activation corresponds
with p65 levels and both are reduced in MGRKO mice at 72 h post-KA suggesting that GR
signaling in myeloid cells is necessary for this increased cellular activation in terms of both
markers. Unexpectedly, MGRKO mice were not protected from p65 potentiation at 12 h post-KA,
but by 72 h post-KA or 24 h post-MCAO, GC-treatment could no longer increase p65 levels in
the absence of myeloid cell GR signaling (Fig. 14a–c). These time-dependent effects of myeloid
GR signaling could reflect separate p65-augmenting effects of GCs in neurons (perhaps induced
by the initial inflammatory cascade from acute injury) and p65-augmenting effects of GCs on
immune cells after many have been recruited.
Finally, at any time point in either injury model, removing GR in either myeloid cells or
neurons decreased nuclear p65 levels. This suggests that endogenous GCs are necessary for
increased p65 nuclear translocation and that they do this in both neurons and immune cells. Taken
together, these findings suggest that endogenous GCs are necessary for normal p65 activation in
neurons and in myeloid cells independently. In contrast, exogenous GCs activate p65 to a greater
extent and maintain its elevation via signaling predominantly in myeloid cells themselves. This
prolonged activation is very likely to contribute to damage in the injury site, and indeed elevated
39
NF-kB signaling in myeloid cells is required for neuron death in both of these same injury models
(102).
3.3 Conclusions
The pro-inflammatory effects of GCs began as an assortment of observations about
increased and decreased inflammatory mediators during a number of different immune challenges
and model systems. Chapter 2 characterized this augmented response during two different acute
neurological injuries and asked which effects could be attributed to GR signaling in myeloid
cells. To further understand the molecular mechanisms behind these increased inflammatory
responses, this chapter systematically measured the effects of various GC treatments on pro- and
anti-inflammatory cytokines as well as other well-known mediators of inflammation.
These results identify several anti-inflammatory signaling molecules that are suppressed
by GCs in a manner that could lead to an augmented cellular response. Furthermore, GCs have an
impaired ability to inhibit NF-kB in the hippocampus and fail to induce some of their normal antiinflammatory targets like IL-1ra and IkBa. In rats, this led to an impaired ability of GCs to inhibit
NF-kB activation (and striking increases in NF-kB caused by intermediate GC doses). In mice,
NF-kB was activated both by GC treatment and endogenous GR signaling in neurons and
myeloid cells alike. Thus, while GCs do not appear to increase pro-inflammatory signaling
molecules directly, and even retain some of their expected anti-inflammatory effects on these
molecules, they do suppress anti-inflammatory signaling molecules and have stimulatory effects
on NF-kB. Together, these effects of GCs on the key elements of the inflammatory response to
injury help to narrow the possible mechanisms underlying the GC-augmented inflammatory
responses described in chapter 2.
Having identified some of the GC-augmenting effects on inflammation amidst some of
their expected anti-inflammatory effects, it was next important to determine whether these
unusual effects had any functional significance for neuron survival. The next chapter addresses
the issue of whether these pro-inflammatory effects of GCs have consequences for neuron
survival during the two different acute CNS injuries.
40
Chapter 4: GC-Endangerment of Neurons Requires GC-Augmented Inflammation
4.1 Results
To see if GC-augmented inflammatory responses (and if GC effects on immune cells in
general) contribute to neuron death, the damage resulting from each of these injuries was
measured. The effects of knocking out the GR in myeloid cells and neurons were measured after
KA and MCAO in the MGRKO and FBGRKO mice. In the rat KA model, the same drugs
indomethacin and minocycline were given to see if they could protect neurons from GCendangerment.
4.1.1 GC-endangerment requires GC-augmented inflammation
The pro-inflammatory GC effects described in chapters 2 and 3 suggested that GCaugmented inflammation might actually be detrimental to neuron survival (58), so it was next
measured whether anti-inflammatory drug treatments were more protective in the presence of
GCs (Fig. 15a). To quantify neuron death, the damaged area in the CA3 and CA4 hippocampal
fields was measured in tissue sections subsequent to those stained in Fig. 8 (Fig. 15b–d). In
agreement with previous examples of GC endangerment, high GCs significantly worsened CA3
neuron death (Fig. 15b,c). Indomethacin was unable to protect CA3 neurons from GC
endangerment, but minocycline was robustly protective. This result is consistent with the
differing ability of these drugs to block the cellular response to injury and further supports the
hypothesis that GC-augmented inflammatory responses contribute to and are necessary for GC
endangerment.
4.1.2 GR signaling in myeloid cells endangers neurons in mice
Next, the effects of GCs on neuron death were measured in the mouse injury models (Fig.
16). Several representative images (Fig. 16b,f) show that most KA-damage was restricted to the
CA3 region, but GC-treatment frequently extended the damage to include CA1 neurons.
Representative TTC sections (Fig. 16d,h) revealed significant cortical damage with some GCtreated mice having damage extending into the striatum. Quantification of this damage found that
GC-treatment nearly doubled neuron death in both injury models. MGRKO mice were protected
41
from this GC endangerment, suggesting that GR signaling in immune cells is necessary for GCs
to make more neurons die (Fig. 16c,e). Thus, despite differences in the pathology and types of
immune cells responding to these two injuries, deleting GR in myeloid cells prevented GCendangerment.
To see if the protective effects of GR deletion were myeloid cell-autonomous, the effects
of GCs were next measured in FBGRKO mice where GR is deleted in the neurons throughout the
injured brain regions. In contrast to MGRKO mice, FBGRKO mice were still susceptible to GCendangerment in both injury models (Fig. 16g,i). Removing GR from forebrain neurons also had
a trend towards a protective effect post-MCAO. Together, these findings suggested that GR
signaling in myeloid cells but not neurons is required for GC-endangerment.
Given the protective effects of GR deletion in immune cells, we next tested whether
increased GR levels are sufficient to worsen neuron death. Despite the increased GR levels,
overexpressing GR alone was not sufficient to increase neuron death in either injury model (Fig.
17). These results support a pathological role for endogenous GR signaling in neurons postMCAO, and a pathological role for exogenous GR signaling in immune cells.
4.1.3 GR signaling affects BDNF levels differently in the hippocampus and cortex
We next measured levels of BDNF at 12 and 72 h post-KA and at 24 h post-MCAO to see
if GC treatments were affecting the production of this neurotrophin. GCs are reported to suppress
BDNF levels, so this served as a positive control for a non-inflammatory system that GCs are
supposed to affect to make sure that the molecular genetic manipulations could detect a similar
effect.
GCs suppressed BDNF levels in almost all groups as expected (Fig. 18). In the
hippocampus post-KA, where BDNF levels were an order of magnitude higher than in the cortex,
GCs suppressed BDNF levels in MGRKO mice but not in FBGRKO mice (Fig. 18a,b). Indeed,
FBGRKO mice had lower BDNF levels in general, indicating that GR signaling in neurons is
necessary for BDNF expression, but is also necessary for GCs to suppress BDNF. In contrast, in
the cortex post-MCAO, GCs suppressed BDNF levels in FBGRKO mice but not in MGRKO
42
mice (Fig. 18c,d). FBGRKO mice also had higher BDNF levels in general, indicating that GR
signaling in neurons reduces BDNF levels in the cortex, but it is GR signaling in myeloid cells
that is necessary for GCs to suppress BDNF levels. Together, these results demonstrate opposing
neuronal regulation of BDNF levels by endogenously released GCs signaling through neuronal
GR depending on the brain region and furthermore that GC treatment suppresses BDNF levels via
separate mechanisms in these different brain regions.
4.1.4 Endothelial cell GR is also detrimental to neuron survival post-injury
Finally, the pathology of stroke is considerably exacerbated by the edema that occurs in
the confined skull. The shifts in osmotic balance that lead to increased water uptake can be
cytogenic or vasogenic in nature depending on the type of ischemic episode that occurs and
whether it involves a disruption of BBB integrity. One potential reason why GCs might be
considered to be beneficial is their reported ability to tighten the BBB, thus reducing edema. The
role of endogenous GCs released post-stroke in affecting the BBB and neuron death is less clear,
however, and these investigations tested whether GR signaling in endothelial cells might be an
additional contributing factor to the endangering effects of GCs post-MCAO.
To determine the role of GR signaling in endothelial cells during stroke, the floxed-GR
and overexpression mouse lines were crossed to the Tie2-CRE driver (EGRKO and EGROV).
This promoter expresses in endothelial cells, myeloid cells, and the female germline. The female
germline expression resulted in the serendipitous generation of a deleted version of the floxed GR
allele (Fig 19a,b). This null allele was used to improve the recombination efficiency by placing
the floxed allele over the deficiency (Fig. 19a). “WT” mice were thus hemizygous for GR
expression in these tissues. To verify the endothelial cell expression of this promoter, it was
crossed to the ROSA-lacZ reporter (Fig. 19c). These mice exhibited a similar staining pattern to
mice where the Tie2-CRE was crossed to the overexpression mice with the GFP reporter of rtTA
expression (Fig. 19d). Finally, to determine whether GR protein levels were actually altered by
our manipulations, blood microvessels were isolated from the brains of WT, EGRKO, and
EGROV mice and GR levels were assayed by western blot (Fig. 19e). These experiments
43
demonstrated efficient overexpression of the GR in endothelial cells of EGROV mice and a
strong trend towards a decrease in EGRKO mice.
Having demonstrated the efficacy of the GR manipulations in EGRKO and EGROV
mice, the amount of neuron death post-MCAO was measured. Relative to their hemizygous WT
littermates, EGRKO mice had smaller infarcts post-stroke (Fig. 20a). In a similar manner,
EGROV mice had larger infarct. Due to low fecundity insufficient EGRKO mice could be
obtained to measure the effects of exogenous GC treatment; however, such treatment in EGROV
mice was particularly detrimental to neuron survival (Fig. 20b).
To see if GCs were having their expected effects on molecules that comprise the BBB,
their effects on occludin and eNOS levels were measured 24 h post-MCAO. EGROV mice had no
differences in either of these proteins (Fig. 20d,f), but EGRKO mice had increased occludin
levels and a trend towards increased eNOS levels (Fig. 20c,e). Together, these results suggested
that endogenous GR signaling in endothelial cells is detrimental to neuron survival and
suppresses occludin expression after MCAO.
4.2 Discussion
4.2.1 GC effects on inflammation contribute to GC endangerment of neurons
These experiments determined that the effects of GCs on inflammatory cells are not only
detrimental to neurons, they are required for GC endangerment to occur. Several of the molecular
explanations described in chapter 3 suggested the possibility of this outcome. The reduced ability
of GCs to inhibit pro-inflammatory signals in the hippocampus like IL-1beta could have
particularly deleterious consequences for neurons because the increased neurotoxicity of
microglia lacking CX3CR1 is mediated via IL-1b (115). Furthermore, NF-kB activation in
immune cells during excitotoxicity increases neuron death (102) and GCs had an impaired ability
to inhibit NF-kB and in many instances increased NF-kB activation. To see if GC-augmented
inflammation might actually worsen neuron death, we tested whether the anti-inflammatory
effects of minocycline could protect neurons where GCs could not.
44
Indeed, treatment with minocycline blocked GC endangerment, arguing that GCaugmented inflammation is necessary for GCs to endanger neurons. Fittingly, indomethacin did
not reduce immune activation and was also unable to protect neurons. GC-augmented
inflammation alone is not sufficient to endanger neurons, however, because minocycline also
suppressed cellular responses in intact rats but was not neuroprotective (compare Fig. 8b and Fig.
15c). The fact that decreasing the cellular response in the context of high-GCs was protective,
argues that GCs either make neurons more vulnerable to the potential toxicity of an augmented
inflammatory response or they increase the toxicity of that response itself. Minocycline can also
have anti-apoptotic effects, but GCs do not endanger neurons by increasing apoptosis (121),
making this an unlikely explanation for the effectiveness of minocycline. Thus, GC-effects on
inflammation are likely to contribute, at least in part, to GC endangerment.
This fits with the literature describing several different ways that GCs have been shown
to imperil injured neurons. During the energy crisis caused by excessive excitatory
neurotransmitter release, GCs impair multiple processes including glucose transport, glutamate
re-uptake, calcium sequestration, and the scavenging of reactive oxygen species (97). None of
these effects can individually account for GC endangerment, supporting the notion that GC
endangerment emerges from the varied effects of these hormones. These findings add a surprising
new angle to GC endangerment, since the anti-inflammatory properties of GCs might have
normally been expected to counteract some of their endangerment.
Taken together, these data support the conclusion that GC-augmented inflammation in the
hippocampus contributes to worsened neuron death and is not simply a consequence of GCs
making more neurons die. Furthermore, microglial activation is required for GC endangerment,
despite all of the alternative mechanisms by which GCs can imperil neurons. The present results
demonstrate a functional role for GC-augmented inflammation in worsening neuron death.
4.2.2 GC endangerment requires myeloid cell GR, not neuron GR
Deletion of the GR in myeloid cells and neurons much more clearly revealed differences
in how endogenous GCs and exogenous GCs affected neuron death. Exogenous GC-treatment
45
endangered neurons after either KA or MCAO but could only be reversed by deletion of the GR
in myeloid cells, and deleting GR in neurons did not confer the same protection. This suggests
that the neuron-endangering effects of exogenously applied GCs are largely mediated via their
effects on myeloid cells. This outcome is in contrast to the expectation that GCs endanger
neurons by impairing some of the energy-intensive processes required for neuron recovery.
Secondly, these results demonstrate that the endangering properties of GC-augmented
inflammation are general enough to worsen neuron death from either stroke or excitotoxicity.
4.2.3 Endogenous GCs are also detrimental to neuron survival, but less so
The endogenous GCs released post-injury were considerably less detrimental to neurons
than the exogenous GCs (and were less activating of immune responses (Fig. 6b,c)). The
recruitment of peripheral leukocytes to an injury site required endogenous GC signaling in
myeloid cells post-MCAO (Fig. 6g–j), and correspondingly, endogenous GC signaling in myeloid
cells had a trend towards increasing infarct volume (Fig. 16e.). These effects were mild, however,
with the slightly larger endangering effects of endogenous GCs emerging from their signaling in
neurons themselves. Supporting this, there were trends towards GR signaling in neurons
worsening infarct post-MCAO (Fig. 16i) and towards overexpressing GR in neurons making KA
injury worse (Fig. 17c). The neuron-specific endangering effects of GR signaling could involve
impairment of glucose transport into energetically compromised neurons or impairment of
calcium sequestration and reactive oxygen species quenching. These neuron-intrinsic effects of
GR signaling would be expected to be more visible in the larger injury caused by MCAO and
indeed they were (Fig. 16i). The increased KA damage from GR overexpression in neurons
implicates either the energetically demanding transgene-overexpression or the GR overexpression
in the increased damage due to KA. The fact that CRE-expression was not detrimental in
FBGRKO mice argues against the increased damage being an artifact of transgene expression.
In rats, the role of the endogenously-released GCs post-injury in mediating
neuroprotection or endangerment can be roughly determined by comparing the low GC rats with
the intact rats. Other studies have found a protective effect of removing GCs (63), and if intact
46
and low-GC rats are compared across all drug treatments, there is a nearly significant main effect
of low-GC treatment to reduce the damage relative to intact animals (F = 3.182, df = 1, P =
0.0823). This suggests that the power in this experiment may simply be insufficient to distinguish
between low and intact animals if indeed there is a difference. A post-hoc power analysis was
performed to determine the n that would be required to detect such a difference. The approximate
standard deviation in damage for either low or intact rats was ±15% making the difference
between the observed means approximately 0.5 standard deviations in size. To be able to detect
an effect this size within a 3x3 ANOVA, an n of approximately 14 mice per group would be
required. Instead, each of these groups had an n ranging from 5–13 animals due to mortality
(although there were no significant differences in mortality between any of the treatment groups).
Thus, it is likely that endogenous GCs also endanger neurons, but the effect was much bigger
when stress level GC treatments were given.
4.2.4 GR signaling has cell-autonomous effects on BDNF levels
As reported in the literature, GC-treatments suppressed BDNF levels in almost every
treatment group (Fig. 18). GR signaling mediated this suppression differently in the ischemic
cortex and injured hippocampus. In the injured hippocampus, endogenous GR signaling in
neurons was required for increased BDNF levels, and this effect was reversed by exogenous GCtreatments (Fig. 18b). In contrast, in the ischemic cortex, endogenous GR signaling in neurons
suppressed BDNF levels and exogenous GC-treatments did not require neuron GR signaling to
reduce BDNF (Fig. 18d). Instead, myeloid cell GR signaling was necessary for GCs to suppress
BDNF levels in the infarcted hemisphere (Fig. 18c).
Thus, GR signaling in neurons affects BDNF levels oppositely in the injured
hippocampus and cortex. The endogenous GCs released post-injury activate BDNF expression
post-KA but they suppress BDNF post-MCAO. In addition, the amount of BDNF produced by
endogenous GR signaling in neurons was equivalent to the amount suppressed by exogenous GC
treatments. This could mean that in hippocampal neurons, GC treatments block the induction of
BDNF by endogenous GR signaling. In cortical neurons, endogenous GCs do not induce BDNF,
47
they suppress it in a fashion that works in parallel with the ability of exogenous GCs to suppress
BDNF. This additional GC-mediated reduction in BDNF levels in the cortex might be mediated
by GR signaling in myeloid cells. In support of this, there is a much larger immune cell response
to injury in the ischemic cortex than in the injured hippocampus and BDNF levels are much lower
in the cortex than in the hippocampus, perhaps lowering the threshold to make it easier for nonneuronal signaling to influence BDNF levels.
Together, these results demonstrate opposing neuronal regulation of BDNF levels by
endogenously released GCs signaling through neuronal GR depending on the brain region and
furthermore that GC treatment suppresses BDNF levels via separate mechanisms in these
different brain regions.
4.2.5 GR signaling in endothelial cells also endangers neurons during MCAO
The Tie2 promoter is not ideal for studying brain endothelial cells because it also
expresses in myeloid lineage cells, making specific interpretations difficult. Fortunately, this
undesired non-specificity of the Tie2 promoter is likely to overlap considerably with the myeloid
lineage specificity of the LysM promoter. An imperfect solution might therefore be to compare
effects in MGRKO and EGRKO mice to identify differences that might be attributable to
endothelial cell GR actions.
Endogenous GR signaling in endothelial cells increased neuron death, and
overexpressing the GR increased neuron death even further, albeit not significantly. In addition,
GR overexpression was proportionally increased whether or not additional GC treatments were
given. These results suggest that in endothelial cells, GR expression is a rate-limiting factor in
determining neuron survival. GR overexpression was rarely able to achieve significant effects in
any of the mouse lines examined, but the near-significance measured here could result from the
combined effects of GR overexpression in endothelial and myeloid cells. Comparison of EGRKO
and MGRKO mice post-MCAO shows that relative to their WT controls, EGRKO mice had a
30% reduction in infarct volume whereas MGRKO mice had a 10% reduction in infarct volume.
This supports the conclusion that GR signaling is detrimental in both endothelial cells and
48
myeloid cells. Other plausible, but slightly less parsimonious explanations would include
differences in the population size or type of myeloid cells targeted by the two different promoters,
or synergistic effects of having GR removed in multiple cells exacerbating the injury beyond the
contribution of each individual cell type.
These results suggested that GR signaling in endothelial cells is detrimental to neuron
survival in a manner that is independent from previously described mechanisms by which GCs
endanger neurons. To measure the effects of GR signaling in endothelial cells, levels were
measured of two key proteins regulated by GCs and involved in the integrity of the BBB and the
response to ischemia, occludin and eNOS. GCs would be expected to increase occludin levels,
based on their reported ability to tighten the BBB by increasing tight junction protein expression
(122). Instead, at 24 h post-ischemia, occludin levels were significantly reduced by GR signaling
in endothelial cells. This effect may be more specific to the endogenous GCs released during
injury as opposed to the effects of exogenous GC treatments, which were not measured in this
case. Finally, in contrast to the detrimental effects of increased NO synthesis by iNOS in myeloid
cells, eNOS expression in endothelial cells has been reported to be rapidly increased by GR
agonists during ischemia in a fashion that confers neuroprotection (123). Increased NO synthesis
in the vasculature would be of particular therapeutic benefit during ischemia because it increases
vasodilation and facilitates blood flow. In contrast to this expectation, the present findings
indicated a trend towards GR signaling in endothelial cells suppressing eNOS, however this
remains to be investigated further because the time point analyzed here was 24 h post-stroke, and
eNOS is particularly important in the early time points post-stroke. Additionally, the caveats
associated with GR antagonists may apply, in that they potently suppress endogenous GC release
so if the drug-resistance transporters in the BBB are not overwhelmed, observed effects could be
the result of a decrease in GR signaling (see later discussion in section 5.2.2).
To summarize, endogenous GR signaling in endothelial cells endangered neurons via
several unique mechanisms including decreased production of tight-junction proteins and
49
potentially decreased activation of a neuroprotective vasodilatory response. The inflammatory
involvement in this endangerment cannot be completely ruled out.
4.3 Conclusions
GC treatments are particularly detrimental to neurons post-ischemia or excitotoxicity, and
even the endogenous GCs released in response to injury are detrimental to neurons. The neuron
endangerment caused by exogenous GC treatments requires microglial activation and specifically
GR in myeloid cells to occur. In contrast, GR effects in neurons mediated more of the detrimental
effects of endogenously-released GCs. GCs suppress BDNF via different mechanisms following
these injuries and the one case where endogenous GR signaling in hippocampal neurons
increased BDNF levels, this same activation was suppressed by GC treatments. Finally,
endothelial cell GR signaling was also detrimental via a separate set of mechanisms that likely
involve reduced protective responses and compromised BBB integrity.
Together, these findings refine the picture of GC endangerment, demonstrating that the
circumstances that lead to increased neuron death require GC-effects on immune cells to occur,
and further that endogenous and exogenous GCs differ in their capacity to endanger neurons.
50
Chapter 5: General Discussion and Conclusions
5.1 Primary hypotheses tested and interpretations
5.1.1 GCs augment inflammation
To summarize the situations where GCs can increase inflammation, several tables are
presented. Table 1 lists several representative examples of the literature on GC-augmented
inflammation prior to this dissertation. Table 2 lists the parameters that are the most important in
dictating whether GCs will increase or decrease inflammatory responses. Whether they are
generally pro- or anti-inflammatory is also indicated, although these are broad generalizations and
the priority of the parameters has yet to be worked out. Finally, Figure 21 summarizes the
findings of this dissertation by indicating which inflammatory factors are affected by GCs during
acute CNS injury.
One criticism of these findings is that more neurons dying should equate with more
inflammation, so it is difficult to disentangle the two when GCs are already known to endanger
neurons via mechanisms that are separate from their effects on the immune system (63). The
impact of this criticism is lessened by several observations about the nature of GC-augmented
inflammation. First, as described, acute GC treatments can increase peripheral inflammatory
responses (6), so it is not a far stretch that GCs might have similar effects in the CNS.
Furthermore, in this specific injury model the pro-inflammatory effects of GCs were observed
prior to the emergence of neuron death (63). This suggests that the enhanced inflammation is not
simply the result of GC-increased damage. Moreover, a correlative analysis of those data indicate
that the extent of enhanced inflammatory cell migration at a particular time point post-insult
predicts the extent of augmented neurotoxicity at a later time point. This finding is compatible
with the idea that GC-enhanced inflammation may, in part, contribute to the increased neuron
death. In contrast, the extent to which GCs augment neurotoxicity at a particular time point does
not predict the subsequent extent of inflammation (63). Finally, GC-enhanced neuroinflammation
can occur in the absence of neuron death (i.e. from peripheral LPS administration) (85, 87).
51
The present studies also made note of the relationship between inflammation and neuron
death. Whenever possible, the amount of neuron death in each individual animal was plotted
against each measurement of inflammation and a regression analysis was performed. Neither
CD11b, nor CD68 signal had any correlation with the amount of neuron death post-KA (measured
in the next subsequent tissue section). For protein measurements like p65 where neuron death was
impossible to quantify by histology in the same animal, the amount of neuron death was
estimated by quantifying the amount of spectrin cleavage in the same sample. Spectrin is a
cytoskeletal protein that is cleaved in stereotypical fashion by the caspases activated during
neuron death (124), and initial studies demonstrated that it is a reliable marker for the size of the
nissl-stained lesions post-KA (data not shown). The amount of p65 activity did not correlate with
spectrin cleavage either, and these observations together supported the conclusion that
inflammatory activation does not correlate with the amount of neuron death at the same time
point.
Because GCs are some of the most anti-inflammatory compounds known, the notion that
they can increase inflammation in the CNS has met with skepticism (125). Two central issues
comprised the critical response to our review article (126). The first issue concerned the
misunderstanding that the literature review was attempting to argue that GCs increase brain
inflammation in an uninjured animal and that this was support for the “GC cascade hypothesis.”
“This review’s attempt to make glucocorticoids inflammatory appears to represent the first
faltering steps to negotiate the value of the glucocorticoid cascade hypothesis, which is wholly
dependent on the primary role of excessive glucocorticoids in stress-induced pathology (127).
Although it has been established that acute stress-induced catecholamines can activate brain
inflammatory pathways, no data is presented in this review demonstrating that stress-induced
elevations of glucocorticoids can do the same, either acutely or chronically. Instead, the
hypothesis derives from more complex data regarding the capacity of chronic stress to enhance
brain inflammatory responses to lipopolysaccharide or kainic acid; effects that were reduced by
the glucocorticoid antagonist, RU486 (125). With increasing recognition that inflammation is a
52
central mechanism for a number of diseases that are exacerbated by stress, including
cardiovascular disease, diabetes, cancer and neurodegenerative disorders, the fact that
glucocorticoids are generally recognized as one of the most potent anti-inflammatory hormones in
the body (128) poses a major problem for the glucocorticoid cascade hypothesis; a dilemma that
the authors have chosen to resolve by renegotiating the impact of glucocorticoids on
inflammatory pathways. However, increasing data (largely from humans and non-human
primates) suggest that, if anything, chronic stress-related disorders are associated with decreased
(not increased) glucocorticoid signaling, whether it occurs at the level of the glucocorticoid
receptor and/or its ligand (129). Thus, a more empirical hypothesis is that reduced glucocorticoid
signaling leads to increased inflammation and thus exacerbation of disease, whether in the body
or brain. In sum, this review has contorted the immunologic effects of glucocortiocoids [sic] into
a pretzel in an attempt to reconcile the irreconcilable.”
The problem with this argument is several-fold. First and foremost, the GC cascade
hypothesis is irrelevant because it is about normative brain aging (127), and the focus of the
review was acute necrotic insults and how GCs affect inflammation during those injuries. The
finding that GCs can augment the inflammatory response during acute neurological injury is the
much more pertinent situation because this is precisely the setting where GCs normally decrease
inflammation. The criticism dismisses this scenario as “more complicated” despite agreeing that
GCs have been seen to unexpectedly increase inflammation in this setting. Interestingly, a study
that was not included in the review found that GR signaling can increase microglia proliferation
in the brain in the absence of any injury (105), perhaps contributing to a heightened state of
readiness for any subsequent immune challenge (130). Finally, the suggestion that GCs augment
inflammation because of lower GC signaling is in opposition to the finding that the enhanced
inflammation is prevented by blocking glucocorticoid receptors (81, 85) or GC synthesis (63).
That can only occur if there are elevated GC levels, not lower than normal ones.
The second issue raised was the “paucity of data supporting the authors’ hypothesis that
chronic stress-levels of glucocortiocoids [sic] “can actually increase inflammation” in the
53
brain...as acknowledged by the authors, definitive mechanisms for these effects of chronic stress,
including the potential impact of glucocorticoids on immune cell trafficking and neuronal toxicity
(which do not represent direct effects of glucocorticoids on inflammatory processes), cannot be
established without further experiments.” This argument could not be dismissed at the time, but
the fact that mechanisms are not yet delineated for these observations poses an interesting area of
future research, not an argument against their existence. Indeed, the rationale for repeating this
published dialogue here is to emphasize the relevance of the data described in this dissertation to
the key criticisms initially levied against this hypothesis.
The findings summarized in chapter 2 resulted from an effort to consolidate the proinflammatory effects of GCs into a set of repeatable observations that could be manipulated
experimentally to investigate their underlying mechanisms. It is also still possible that GCs are
having cell-autonomous effects on neurons that result in an increased inflammatory response
earlier and more neurons dying at later time points (in accord with the regression analysis
mentioned previously (63)). More simply stated, this is an issue of correlation and causation. To
resolve this issue more conclusively, the mouse genetic model was used where GR was deleted in
neurons or myeloid cells. If stress-level GCs are increasing inflammation via signaling through
the GR in myeloid cells, then it would be expected that they could not do this in MGRKO mice.
To summarize the observed pro-inflammatory effects of GCs, endogenously-released
GCs increased CD11b/c+ signal in rats post-KA and they increased leukocyte recruitment in a
GR-dependent fashion to CNS and peripheral injuries. These findings support the hypothesis that
GR signaling in myeloid cells due to endogenous GCs released post-injury actually activates
them and increases their recruitment to the injury site. Exogenous GC-treatment also had proinflammatory effects on CD11b/+c signal in rats post-KA and CD68+ signal in mice post-KA,
and the increased intensity of CD68 signal also required GR signaling in myeloid cells. These
results also showed that GC treatment could reduce the recruitment of peripheral monocytes and
likely granulocytes to the infarct post-MCAO, an effect that was myeloid GR dependent. Thus,
across multiple injury models and species, GR signaling in myeloid cells increases their
54
activation and can both activate and suppress their recruitment depending on the context of GC
exposure.
5.1.2 GCs suppress anti-inflammatory signals and increase NF-kB activation
To begin to assess the molecular mechanisms behind these divergent GC effects, several
of their most prominent targets were analyzed next. The existing literature reported mixed effects
of GCs, with some reports of increased NF-kB or cytokines, some reports of no suppression, and
some reports of GCs acting in an anti-inflammatory manner (Table 1). These disparate findings
suggested no obvious explanations for the cellular activation profiles established in chapter 2, so
a broad survey of the most likely candidates was conducted.
As a first-most likely possibility, GCs might increase pro-inflammatory signaling
molecules. This hypothesis turned out to be largely untrue, however, it did reveal that GCs had a
diminished capacity to suppress the production of some of these inflammatory mediators. Given
this diminished capacity, the ability of GCs to induce some of their common anti-inflammatory
targets was measured. Indeed, GCs did not induce some of their expected anti-inflammatory
targets like IL-1ra and IkBa. GCs also suppressed the production of several other antiinflammatory molecules in the rat hippocampus that restrain microglial activation and toxicity
like CX3CL1, CX3CR1, and CD22 (111).
In rats, the GC-mediated suppression of these anti-inflammatory genes (Fig. 10b)
corresponded with the amount of cellular activation post-injury (Fig. 8b). Low-GC rats have onehalf of the cellular activation of high-GC rats and have correspondingly elevated antiinflammatory gene expression. Because high-GCs did not suppress anti-inflammatory genes
beyond the level of intact rats, it could be that GC signaling is required in a permissive fashion to
facilitate an appropriately robust inflammatory response. It is possible that if intact rats were
given high-GC treatment that these genes might be even further suppressed since intact rats given
additional GCs have greater hippocampal immune cell responses to excitotoxicity than intact rats
alone (63) and indeed GC-treated intact mice had lower CX3CL1 levels and correspondingly
augmented inflammatory responses to injury (compare Fig. 11b,e and Fig. 14b,e).
55
Despite widespread potentiation of p65 nuclear levels by both exogenous and
endogenous GCs, one set of WT mice that were GC-treated did not have increased p65 nuclear
hormone levels. This could reflect the fact that FBGRKO mice were used at an older age to
ensure efficient recombination. Aged animals have elevated GC levels and could have a
decreased ability to respond to exogenous GCs (72, 131), although these animals were only 4-6
months of age so these effects should not be present yet. An alternative explanation is that subtle
differences in the GC dose that each group of animals was exposed to might have had an effect on
the amount of p65 activity post-injury. The dose-dependent effect of GCs on NF-kB activity postKA in rats supports this possibility (Fig. 13a,b). Finally, it could be the case that there is increased
NF-kB activity in these mice, but no observable difference in p65 levels in the nucleus, as the two
do not necessary equate. Instead, these findings support the conclusion that GCs, particularly
those released during moderate stress, can increase p65 nuclear localization and activity, and in
the hippocampus GCs are less able to reduce the activation of p65 than in other tissues.
Finally, one interpretive confound must be considered. Previous studies using the LysMCRE driver have found that its expression can increase in immune cells after they have been
activated (102). This may matter less for the floxed GR allele based on the 70% reduction in GR
protein levels in extracted, unactivated microglia, but it could still be the case that injury
increases GR deletion in MGRKO mice. The half-life of GR levels is thus an important factor to
consider when analyzing time-dependent effects. It is possible that CRE-expression is activated
by the injury such that there is insufficient time for GR protein levels to decrease by 12 h, but
there is sufficient time by 72 h leading to greater effects only at the later time point. Supporting
this, GR half-life is 4 h in cultured cells (132).
Together, these molecular investigations paint a more detailed picture of GC-augmented
inflammatory responses as GC-mediated suppression of anti-inflammatory signaling molecules,
particularly CX3CL1, accompanied by a failure of GCs to induce some of their normal antiinflammatory targets in the hippocampus post-KA. The GC-augmented inflammatory responses
56
post-MCAO are more likely to be due to effects of GCs more directly on immune cells
themselves.
5.1.3 GC-augmented inflammation is necessary for neuron death
All of the evidence describing the mechanisms of GC-augmented inflammation suggested
that it might also contribute negatively towards neuron survival post-injury. Indeed, following
either injury, GC effects on myeloid cells were required for GCs to endanger neurons.
Furthermore, GR signaling largely suppressed BDNF levels and GR signaling in endothelial cells
was also detrimental to neuron survival. These findings together offer strong support for GCendangerment and further characterize it as an effect that largely depends on these unusual effects
of GCs on inflammatory responses in the CNS.
Some of the most likely molecular mediators of this increased neurotoxic inflammatory
response are the decreased expression levels of anti-inflammatory cytokines like CX3CL1 that
normally keep the toxicity of microglial IL-1beta in check. Accompanied by a reduced ability of
GCs to stimulate some of their normal anti-inflammatory targets in the hippocampus, these
effects could contribute to neuron death post-injury. Although they regulated CX3CL1 oppositely
in the cortex, the increases in this chemokine during ischemic injury are also described to have
detrimental effects on neuron survival (120). Finally, the GC-augmented NF-kB activation is also
likely to carry pathological consequences, as activation of myeloid cell NF-kB is detrimental
during both MCAO and KA injury in mice (102). In support of this mechanism for GCendangerment via augmented inflammation, MGRKO mice were protected from GC-induced
prolonged p65 nuclear translocation following both of those injuries.
5.2 Secondary hypotheses tested and interpretations
In the course of studying their general effects on inflammation and neuron death, several
more specific side-hypotheses were tested about some of the parameters that determine how GCs
function.
5.2.1 Timing and duration of GC exposure
One of the central distinguishing features that determines how GCs will function is the
57
timing and duration of their exposure relative to the injury. An acute single exposure to stresslevel GCs has been reported to stimulate both CNS and peripheral immune responses (133, 134),
provided that it occurs prior to the inflammatory challenge. GC-treatments beginning 12 h prior
to LPS treatment in humans potentiate plasma IL-6 and TNF-alpha levels (48, 49). One report
began to see stimulatory effects on cytokine production in the CNS if GC exposure occurred 2 h
ahead of peripheral LPS challenge, contrasted with suppressive effects if GCs were given 1 h
after the challenge (133, 134). Indeed, several studies have found that GCs given at the same time
as or following an inflammatory challenge are more likely to have classic anti-inflammatory
effects (23, 62, 84). A number of these investigations looked at the effects of acute stress but not
GCs. Nonetheless, many of the effects of prior acute stress match with the effects of prior GC
treatment on similar measurements.
Prior acute stress also has measurable effects on subsequent inflammation in the CNS. If
animals are exposed to an acute stressor 24 hours prior to LPS treatment, the result is an enhanced
inflammatory response across several measures, including plasma and brain levels of IL-1β and
TNF-α (50). More specifically, prior acute stress was found to enhance LPS-induced production
of IL-1β in microglia (133). The enhanced response in the brain is relatively long-lived, being
observable for at least four days following the stressor (51). Injection of IL-1 receptor antagonist
completely blocked stress-enhancement of IL-1β production, illustrating that pro-inflammatory
signaling by IL-1β is necessary to observe its own potentiation by stress. In addition, centrally
injected human recombinant IL-1β was sufficient to reproduce the effects of an acute stressor
(135). These findings argued that acute stress and likely GCs can have stimulatory effects on
immune responses provided that they occur prior to the injury, an effect that was described as
GCs priming the inflammatory response to injury (130).
For the present studies, a sub-acute duration of exposure to stress-level GCs was selected
for several reasons. First, this duration is common in a clinical setting, lending the results
additional relevance. Second, previous work in rats used this duration of exposure to increase
CNS injury and inflammation (63, 87). This duration of GC exposure is short enough that GR
58
expression levels might not be decreased in compensatory fashion. This sort of receptor
downregulation is normally only seen after 3 weeks of chronic restraint stress (35). For the ADX
studies, GC synthesis can resume in ectopic tissues, so this timing represented a compromise to
ensure that the GC manipulations were not confounded by endogenous compensatory
mechanisms. In the mice, implanted GC pellets would no longer release stress-levels of the
hormones if the experiment continued for many more days. Finally, it is a duration of exposure
that is expected to suppress immune responses, but it begins to address the question of where the
switch between the stimulatory and suppressive actions of GCs occurs (96).
Thus, the sub-acute duration of GC exposure prior to the injury was more than sufficient
to increase both neuron death and CNS inflammation in both rats and mice. Indeed, GC
endangerment has been found to occur from GC exposure as proximally as 12 h on either side of
the injury (69, 70, 106). This duration of GC exposure also began to reveal the threshold for
situations where sustained exposure to GCs is supposed to suppress certain inflammatory
measurements like immune cell recruitment and cytokine production. Despite these
immunosuppressive effects of GC treatment, GCs also increased cellular responses and p65
nuclear activity, suggesting that their concentration dependent effects on different cell types might
result in their mixed pro- and anti-inflammatory actions.
5.2.2 Synthetic versus endogenous GCs
Because of their clinical relevance, the properties of synthetic GCs (most commonly
dexamethasone and methylprednisone) are pervasively explored in the literature. Care must be
taken when considering the actions of these compounds, as they are frequently quite different
than the effects of endogenous GCs. The synthetic GC dexamethasone differs from corticosterone
in that it exhibits greatly reduced binding to normal GC carrier molecules such as the
corticosteroid-binding globulin, and has an enhanced in vivo affinity for GR as opposed to MR
(15, 35). But an important additional difference is that dexamethasone crosses the blood-brain
barrier (BBB) poorly due to the activity of the multi-drug resistance transporter p-glycoprotein
(15). As a result, dexamethasone exerts a disproportionate percentage of its neuroendocrine
59
effects at the pituitary, rather than the brain. A consequence of a strong dexamethasone signal at
the pituitary is a negative feedback inhibition of the adrenocortical axis, resulting in decreased
secretion of ACTH and corticosterone. This means that little dexamethasone actually penetrates
the brain parenchyma and systemic corticosterone secretion drops. Given that the total amount of
GCs (either endogenous or synthetic) reaching the injured brain will decrease, are the proinflammatory effects of dexamethasone merely an artifact of a net decrease in GC antiinflammatory actions?
This possibility has been raised (136) as a possible explanation of observed proinflammatory effects of dexamethasone (137), but it cannot be universally applicable. First, GCs
can augment pro-inflammatory cytokine expression in primary brain tissue culture, a setting
where issues of BBB permeability and adrenocortical negative feedback regulation are irrelevant
(86). Second, a similar question has been raised concerning the effects of dexamethasone on
memory consolidation and for this topic there exist many lines of evidence that support the
conclusion that like corticosterone, dexamethasone is capable of exerting its effects centrally,
provided that it is given in excess (138). Finally, as reviewed, corticosterone is also capable of
increasing inflammation in vivo and antagonism of GR signaling blocks this effect (81, 85).
Thus, synthetic GCs like dexamethasone permeate tissues and activate nuclear receptors in
a way that is different than endogenous GCs. A concerted effort is needed to understand the
differences between how endogenous and synthetic GCs affect inflammation in the CNS.
5.2.3 Basal/permissive levels of GCs versus stress levels
The more than half-century old view of the effects of stress and GCs on immunity has been
modified in two important ways. The first is the recognition that the primary effect of stress on
inflammation is not suppression, but rather stimulation; this reflects the immunostimulatory
effects of sympathetic catecholamines in the first few minutes of the stress-response. These
findings lead to the revisionist view that GC anti-inflammatory actions are not mediating the
stress-response, but rather constraining it and facilitating recovery from it.
The second modification is the recognition that the early, immunostimulatory effects of
60
stress are not merely due to catecholamines, but also to basal GC levels. These findings led to the
revisionist view that at the beginning of the stress-response, basal levels (or even the earliest
phases of stress-induced increases in GC secretion) have permissive immunostimulatory actions;
this is the inverse-U of GC function. Importantly, this revisionist view of the stimulation of
immunity by GCs does not challenge the traditional view that stress levels of GCs, and the
pharmacological levels of synthetic GCs used clinically, are robustly immunosuppressive.
The finding that GCs can have pro-inflammatory effects in the brain might initially seem to
fit comfortably in this revisionist picture, in that these GC actions would “merely” be basal,
permissive effects, predominately mediated by MR occupancy at the beginning of a stressresponse. The studies reviewed and the present experiments, however, make clear that this is not
the case. It is basal, not stress levels of GCs that lead to a reduced inflammatory response in the
face of a neurological challenge, both in vivo and in cell culture (62, 84-86). In contrast, sustained
stress, stress levels of GCs, or heavy and selective occupancy of GR all have pro-inflammatory
effects on these endpoints (63, 81, 85-89).
5.2.4 Regional differences in GC effects in the brain
Virtually all the reports of pro-inflammatory GC effects have examined the hippocampus or
cortex (or cultures derived from them) (63, 81, 85-87). For example, GCs can have
simultaneously opposing effects on the generalized CNS inflammation induced by peripheral LPS
challenge; specifically GR signaling increased NFκB activation in the hippocampus and cortex,
while inhibiting its activation in the hypothalamus and the heart (139). The present findings
support these results because they identify many additional pro-inflammatory GC effects in these
forebrain structures. In addition, within these structures, simultaneous pro- and anti-inflammatory
GC actions were observed. GC potentiation of LPS-induced NFκB activation and cytokine
production was more pronounced in the cortex than the hippocampus across the full physiological
range of corticosterone concentrations (87). Any number of possible mechanisms could explain
these regional differences including, for example, differing concentrations of GR and MR in
neurons, microglia or the vascular endothelium. Understanding the mechanisms that drive these
61
regional differences is an important subject of future research.
5.2.5 Does GC-augmented inflammation and neuron endangerment occur in mice?
There was some indication that GCs might work the same in mice as they did in the rat
studies that comprise most of the evidence in the literature. A variety of different injury types and
GC exposures demonstrate that GCs can endanger neurons in mice (140-143); however, there was
a dearth in the knowledge about normal GC functions post-injury and the role of high-stress level
exogenous GC treatment in these particular acute neurological injuries.
This work robustly demonstrated that GC endangerment occurs in mice in vivo as it does
in rats. The reported effects of GCs on neuron death from the distal model of MCAO are mixed
with some reports of endangerment and some of no effect or protection (97). Thus, the
observation that GC treatment also increased the infarct volume post-MCAO further supported
the literature describing situations where GCs endanger neurons and demonstrated once again that
GC endangerment is a generalizable phenomenon that affects multiple injury types.
5.3 Summary and Conclusions
There are a wide variety of experimental approaches and parameters measured in the
literature that support the idea that GCs can increase CNS inflammation, but the number of
variables that differ between reports makes direct comparisons difficult. When summarized
(Tables 1, 2), however, several trends emerge. Based on the literature prior to this dissertation, the
classic picture of GCs as anti-inflammatory already needed to be modified in a number of ways:
1) The immunosuppressive effects of GCs during chronic stress are involved not so much in the
mediation of the immune stress-response as containment of and recovery from it.
2) While catecholamines are responsible for much of the initial immune activation by the stressresponse, GCs can also play both permissive and stimulatory roles during this phase.
3) These opposing GC actions depending on the duration of exposure fit well with other examples
of inverse-U GC actions; a common effect when two separate receptors (MR and GR) have
different thresholds for activation by the same hormone.
4) Prior exposure to stress or GCs can result in a “priming” of the immune response to a
62
subsequent inflammatory challenge.
5) In the CNS, sub-acute and chronic exposure to GCs and stress is not uniformly antiinflammatory and there are circumstances where GCs can actually increase inflammation;
this has been observed for immune cell activation, cytokine levels, and transcription factor
activation in response to necrotic and inflammatory challenges.
6) Whether pro-inflammatory GC effects occur in the brain can depend upon the timing and
duration of GC exposure relative to the inflammatory challenge, the use of synthetic or
endogenous GCs, and even the brain region in question.
Finally, integrating the investigations presented in this dissertation adds the following additional
conclusions about the role of GR signaling during acute CNS injury (summarized in Fig. 21):
1) Stress-level GC exposure prior to injury increases the cellular inflammatory response to CNS
injury in both rats and mice.
2) GC-augmented inflammatory responses may result from their suppression of antiinflammatory cytokine expression (CX3CL1, CX3CR1, CD22) and a failure of GCs to activate
some of their normal anti-inflammatory targets (IL-1ra, IkBa, MKP-1).
3) The effects of GCs on immune cells are necessary for GCs to endanger neurons.
4) Exogenous GC treatments are more detrimental to neuron survival than the endogenous GR
signaling that occurs following CNS injury.
5) Endogenous GCs endanger neurons during ischemia via GR-signaling in endothelial cells
leading to a compromised BBB.
These findings produce two challenges. The first is in the realm of basic science, namely to
further investigate the mechanisms underlying GC-augmented inflammation and its relationship
to GC-endangerment of neurons. In particular, the role of GR signaling in astrocytes during these
acute neurological injuries remains to be determined. Based on the expectation that GCs should
reduce glutamate re-uptake from the synapse by astrocytes, it is expected that astrocyte GR
signaling would also contribute to GC endangerment. GCs have many other targets in astrocytes,
however, so it is possible that this assumption is incorrect. To further delve into the molecular
63
mechanisms behind some of these diametrically opposite effects of GR, several approaches will
be necessary. The establishment of a reliable indicator of these effects in primary cultured
neurons would greatly facilitate these investigations. As one example, GCs can have diametrically
opposite effects simultaneously in the injured cortex and hypothalamus (85). Contrasts such as
these could well arise from differences in effects of GC/GR complexes upon GREs, from GC/GR
interactions with other components of gene transcriptional machinery, or from mechanisms
completely independent of the genome. These differences could be easily manipulated and
visualized in cultured cells from the different brain regions to identify potential mechanisms that
could be verified later in vivo. As another approach, chromatin immunoprecipitation for the GR
from WT and KO mice could be used to identify differences in GR-responsive genes or
interactions with other gene transcriptional machinery in different brain regions under the
different GC treatment conditions, potentially with cell-specific resolution as well. Finally, the
scope of this research should be broadened to study GR effects in different CNS cell types during
chronic degenerative CNS disorders where excessive, unchecked inflammation frequently
contributes to the pathology.
The second challenge is one in the realm of clinical neurology. There has often been a
dogma about the uniformly beneficial effects of GCs in the face of brain inflammation thanks to
their peripheral anti-inflammatory effects. Yet, as long emphasized by leaders in the field, GCs
are frequently ineffective in lessening post-stroke edema and, in many cases actually worsen the
outcome (144); furthermore, the same seemingly paradoxical worsening is seen when GCs are
used after traumatic brain injury (145). Finally, while there is increasing evidence that chronic
low-grade brain inflammation can increase the likelihood of late-onset Alzheimer’s Disease, and
that long-term NSAID treatment can buffer against this, long-term GC treatment has no such
beneficial effects (146). There is even some very preliminary evidence that GCs can augment
inflammation in humans in a pattern consistent with rodent studies. High-stress level GCtreatments prior to LPS challenge in humans potentiate plasma IL-6 and TNF-a levels (49, 147).
Even pharmacological GCs like dexamethasone have been reported to induce the expression of
64
factors associated with innate immune responses (amid decreasing components of adaptive
immunity) in human blood mononuclear cells (26). Additional studies of this phenomenon in
humans or human cell lines are warranted. The basic science findings considered in this thesis
may well have considerable clinical implications.
To conclude, these investigations demonstrate a functional role for GC-augmented
inflammation in CNS injury. Specifically, exposure to moderate to severe stress-levels of GCs for
as little as a few days prior to hippocampal injury is sufficient to increase cellular and molecular
inflammatory responses with detrimental consequences for neuron survival. These results support
the idea that prior GC exposure modifies the immune environment in the hippocampus making it
more toxic to injured neurons, and could reflect a unique response of the GR-rich forebrain to
GCs (85, 87). The biological significance of these pro-inflammatory effects is further supported
by a recent study showing that blockade of NF-kB signaling reverses the effects of chronic stress
on anxiety behavior (148). The pathological influence of GC-augmented inflammation thus
affects neurological processes as diverse as psychological disorders and CNS injury.
65
Methods:
Animals:
(Rats). Male Sprague-Dawley rats between 250–300 g (Charles River, Gilroy CA) were housed
in a 12 h light–dark cycle and fed ad libitum. All rats were treated in accordance with the
Stanford University Administrative Panel on Laboratory Care regulations.
(Mice). Mice were housed in a 12h light 12 dark cycle with ad libitum food and water. The floxed
glucocorticoid receptor and LysM-CRE drivers were a gift from Dr. Louis Muglia. CD11b-CRE
mice were a gift from Dr. Jean Vacher. Tie2-CRE and CX3CR1-GFP mice were a gift from Dr.
Ben Barres. CamKIIalpha-CRE mice were obtained from Jackson labs and bred into the floxedallele background. TRErGR mice were a gift from Dr. Sam Okret and were bred into the ROSArtTA line from Jackson labs until progeny were double-homozygous for both alleles. All animals
used for KA experiments were male littermates between 2-4 months of age for LysM-CRE mice
and 4-6 months of age for CamK-CRE mice at the time of the experiment. All mice used for
MCAO were > 3 months of age.
GC treatments:
(Rats). Rats were anesthetized with 2% halothane and bilaterally adrenalectomized (ADX) or left
intact as a control. ADX rats were given a 100 mg s.c. pellet with different corticosterone
percentages mixed with cholesterol. Low-GC pellets were 15%, intermediate were 30% or 60%,
and high-GC were 100% corticosterone. High-GC rats were also given daily 10 mg/kg s.c.
injections of corticosterone in peanut oil (15).
(Mice). Mice were given a subcutaneous corticosterone (Sigma) pellet weighing 10mg that was
implanted between the scapulae using a 12ga trochar (Innovative Research of America). Vehicle
animals were subjected to the same injection but no pellet was implanted.
Stereotactic surgery:
(Rats). Rats were anesthetized with ketamine/xylazine/acepromazine cocktail (77/7.7/1.54 mg/
kg, i.p.). Phosphate-buffered saline (PBS) or 0.06 µg of KA dissolved in PBS was injected in 1
µL at a rate of 0.2 µl/min into the dentate gyrus (coordinates: ±3.2 mm lateral, 4.5 mm posterior
66
to bregma, 3.3 mm ventral to dura). Buprenex (0.03 mg/kg, s.c.) was given as a postoperative
analgesic.
(Mice). Mice were given 5% isoflurane (air mixture) to induce anesthesia which was then
maintained with 1.5% isoflurane. KA was injected via stereotactic surgery into the dentate gyrus
of the hippocampus using the following coordinates: 2.0mm posterior and 2.1mm lateral to
bregma; 1.9mm ventral to dura. A total volume of 0.2µL was infused over the course of 4 minutes
for a final dose of 50 ng of KA. Buprenex was administered as a post-operative analgesic at a
dose of 0.1mL of the stock concentration/animal.
Histology:
(Rats). Rats were transcardially perfused 72 h post-KA with 0.9% saline followed by 4% PFA.
Brains were post-fixed in 4% PFA and 30% sucrose for 24 h prior to cutting 20µm frozen cryostat
sections. Sections were stained with cresyl violet and 40x images of the CA3 region were
acquired using an Olympus IX70 microscope and Hamamatsu digital camera. CA3 damage was
calculated by dividing the remaining CA3 cell area in the KA-injected hemisphere by that in the
contralateral PBS-injected hemisphere as previously described (14). Sections were air dried for
30 min, blocked in 5% normal goat serum in PBS + 0.3% Triton-X 100 (PBST) for 1 h at room
temperature, and stained overnight at 4°C with antibodies for CD11b/c (1:100, Serotec) and
GFAP (1:500, Dako). Sections were washed in PBST 3x for 10 min, labeled with 488 nm and 594
nm Alexa secondary antibodies (1:200, Invitrogen) in PBST for 1 h at room temperature, washed
again and stored in polyvinyl alcohol. Metamorph (Universal Imaging) was used to quantify the
amount of CD11b/c+ and GFAP+ area according to established guidelines for quantifying
fluorescent signal (39). Slides were background corrected and the hippocampus was traced to
exclude other regions and artifacts. Signal was only counted if it was 10 intensity units above
background and greater than 5 pixels but less than 25 pixels in diameter.
(Mice). Animals were transcardially perfused at 72h post-KA with 0.9% saline followed by 4%
PFA. Brains were post-fixed and cryoprotected in 4% PFA for an hour and 30% sucrose for 24h
prior to cutting several sets of representative 20µm frozen cryostat sections in intervals across
67
each hippocampus. Nissl stain and hippocampus lesion quantification were performed as
previously described (63). Briefly, tissue sections were dehydrated and stained with cresyl violet
and 40x images of the hippocampus were acquired on an Olympus IX70 microscope using a
Hamamatsu digital camera. Total hippocampal damage was calculated as the remaining cell area
in the CA3 region of the KA-injected hemisphere divided by the remaining cell area in the CA3
region of the contralateral PBS-injected hemisphere. The next serial set of representative
hippocampus tissue sections was used for immunohistochemistry. Sections were air dried for 30
min and were blocked in 5% NGS in PBS + 0.3% Triton-X 100 (PBST) for 1 hour at room
temperature then they were stained overnight at 4°C with antibodies for CD11b/c (1:100, Serotec)
and GFAP (1:500, Dako). Sections were washed with PBST 3 times for 10 minutes each and
incubated 1hr at room temperature with 488nm and 594nm Alexa secondary antibodies (1:200,
Invitrogen) in PBST. Slides were mounted in PVA and images were taken at both wavelengths.
Metamorph (Universal Imaging) was used to quantify the amount of CD11b/c+ and GFAP+ area
adhering to guidelines for quantifying fluorescent signal (149). Slides were background corrected,
the hippocampus was traced to exclude other regions and any artifacts. Signal was only counted if
it was 10 intensity units above background and greater than 5 pixels but less than 25 pixels in
diameter. Prior to quantifying any staining, no-primary antibody controls were imaged to calibrate
the detection limits for image acquisition. These negative controls were always dark, except in
injured tissue where halos around injured cells could often be seen due to increased background
in the injury site.
ELISAs:
(Rats). Protein was prepared from dissected CA3 tissue at the various time points in NP-40 lysis
buffer. Total protein was quantified using the BCA method (Pierce) and ELISAs for MCP-1,
CINC-1, and IL-1β (R&D) were run as per manufacturer instructions and quantified on a
Molecular Devices plate-reader.
(Mice). Hippocampus was microdissected 12 or 72 h post-KA and supernatant and nuclear
extracts were prepared as previously (12). Total protein was quantified using the BCA method
68
(Pierce) and ELISAs for IL-1β, IL-6, and BDNF (R&D) were run on the supernatants (cytosolic
fraction) as per manufacturer instructions and quantified on a Molecular Devices plate-reader.
Drug treatments (Rats). Indomethacin was injected i.p. at 10 mg/kg beginning 24 h prior to KA
and every 12 h until rats were killed. Minocycline was given s.c. at 22.5 mg/kg 30 min prior to
KA, and daily for the next 72 h. LPS (1 mg/kg) was given i.p. and rats were rapidly anesthetized
and decapitated. The hippocampus and frontal cortex from three rats were dissected on ice,
pooled, and flash frozen for later use. All drugs were purchased from Sigma (St. Louis, MO).
Quantitative PCR (Rats). Total RNA was extracted using the RNEasy Plus Mini Kit (Qiagen),
and was reverse-transcribed using the iScript cDNA synthesis kit (BioRad). Intron-spanning
p r i m e r s u s e d w e r e : C X 3 C L 1 : G C AT G A C G A A AT G C A A C AT C ,
T T G G A C C C AT T T C T C C T T T G ; C D 2 2 : C T G T G G C C G T G G A G AT A G AT ,
A A G A G G G T T T G G G G AT G T T C ; C D 2 0 0 : T G T T T G G AT C T G G G A A G G T C ,
G G A G G AT G C T G G T G A C A G AT; T G F - b 1 : G C G G A C TA C TA C G C C A A A G A ,
C G T G T T G C T C C A C A G T T G A C ; I κ B - a : T G A G TA C C T G G A C T T G C A G A A C G ,
TGTAGATGCCTCTCCAAGGATGG. Quantitative PCR was performed on a BioRad PCR
machine using SyBRGreen and PCR Miner software (40).
Electrophoretic mobility shift assay (Rats). CA3 was microdissected 24h post-KA and nuclear
extracts were prepared as previously (12). EMSA to NF-κB was performed using the gel shift
assay kit from Promega as previously (41) using the consensus oligonucleotide probe (5’AGTTGAGGGGACTTTCCCAGGC-3’). Autoradiographs of non-denaturing acrylamide
electrophoresis were quantified with the ChemImager detection system (Alpha-Innotech, San
Leandro, CA), using several exposure times to ensure linearity of band intensities.
Flow Cytometry (Mice): Microglia were extracted from adult brains as described (101). Briefly,
brains were homogenized and mononuclear cells were collected from the 70/30% percoll
interface, washed, and were either allowed to adhere to coverslips or stained with antibodies for
flow cytometry. Loosely adherent cells were post-fixed in either acetone (CD45) or 4% PFA
(CD11b) and were stained with antibodies for CD45 (Santa Cruz) or CD11b/c (Serotec) at
69
concentrations of 1:50 each. Staining for flow cytometry was performed in FACS buffer, which
included sodium azide. Cells were stained with directly conjugated antibodies for CD11b-APC,
CD45-FITC, and Ly6G-PE or their isotype controls (Abcam) and analyzed on an LSR 1 machine.
Flow cytometry analyses were performed using FlowJo.
Western blots (Mice): Antibodies for p65 (Santa Cruz), occludin, eNOS, CX3CL1, (Abcam)
were used at a concentration of 1:1000. The immunoreactive density of equally-loaded lanes was
quantified using ImageJ and all samples were normalized to internal beta-actin load controls.
Focal Cerebral Ischemia (Mice):
Mice were anesthetized as for stereotactic surgery. The MCA was exposed and permanently
occluded by electrocoagulation. An incision perpendicular to the line connecting the lateral
canthus of the left eye and the external auditory canal was made to expose and retract the
temporalis muscle. A burr hole was drilled to expose the MCA for elevation and cauterization by
cutting and retracting the dura. After surgery, the mice were returned to their cages and allowed
free access to water and food. The survival rate of the animals until the end of the experiment was
100%.
Infarct Size Quantification (Mice):
Brains were removed 24 hours after MCAO and cut into seven 1-mm coronal brain slices
(Braintree scientific), which were stained in 1% 2,3,5-triphenyltetrazolium chloride (Merck) in
0.1 mol/L phosphate buffer. Infarct volumes were measured by sampling stained sections with a
digital camera (Nikon Coolpix 990, Nikon Corp), and the image of each section was analyzed by
an image analyzer (Scion Image for Windows 2000, Scion Corp, Frederick, Md). The digitized
image was displayed on a video monitor. With the observer masked to the experimental
conditions, the contralateral hemisphere perimeter was overlapped onto the ipsilateral hemisphere
to exclude edema, and infarct borders were delineated with an operator-controlled cursor. The area
of infarct, which was unstained, was determined by counting the pixels contained within the
outlined regions of interest and expressed in square millimeters. Infarct volumes (in mm3) were
integrated from the infarct areas over the extent of the infarct calculated as an orthogonal
70
projection. All animals displayed infarcts after the occlusion procedure, which included the
cortex, subcortex, and striatum, depending on the intensity of the lesion.
Statistics:
Significance levels * P < 0.05, ** P < 0.01, *** P < 0.001 established by 2-way ANOVA followed
by Bonferroni post-hoc tests (87). Significance established by 1-way ANOVA was followed by
Tukey’s multiple comparison post hoc test. All data are expressed as means ± s.e.m. All statistical
tests were performed with Prism (GraphPad). Vector art was assembled in Adobe Illustrator
(Adobe Systems).
71
R
KO
KO
a
b
M
G
R
G
M
CD68
GR
OVERLAP
MGRKO
WT littermate
GR protein level
(relative to WT control)
W
T
Fig. 1. MGRKO mice
have GR deleted in
the parenchymal
microglia of the
cortex
and
*
125
hippocampus. (a)
Microglia from
100
MGRKO mice or WT
75
littermates isolated on
a percoll gradient and
50
analyzed by western
blot for GR. One WT
25
and two representative
MGRKO bands are
0
shown. All bands were
WT MGRKO
normalized to betaactin controls for
quantification.
c
(microglia from 5
N-terminal Ab
C-terminal Ab
Hippocampus
CD68 mice were pooled perGR band, n = 4 bands per
DAPI g e n o t y p e ) . ( b )
OVERLAP Confocal microscopy
of the cortex in
MGRKO mice
showing examples of
CD68+ cells that lack
GR co-localization
(top, arrowheads) and
examples of CD68+
cells that still retain
GR immunoreactivity
(bottom, arrows).
Scale bar = 10µm. (c)
Epi-fluorescence
microscopy of the
CA3 region of the
hippocampus in WT
(top) and MGRKO
(bottom) mice with
CA3
examples of CD68+
cells that lack GR col o c a l i z a t i o n
(arrowheads) and examples of CD68+ cells that retain GR immunoreactivity (arrows). Staining
patterns were similar regardless of whether the GR antibody used was raised against the N- or Cterminus of the receptor. Scale bar = 35 µm.
Cortex
MGRKO
72
a
b
LysM-CRE+
ROSA-lacZ
X-gal
Cortex
Cortex
LysM-CRE+
CD68
CRE
OVERLAP
WT
d
LysM-CRE+
Hippocampus
CA3
c
LysM-CRE+
ROSA-lacZ
X-gal
CA3
Hippocampus
WT
CA1
DG
CA3
CD68
B-gal
DAPI
OVERLAP
WT
Fig. 2. LysM-CRE is efficiently expressed in the parenchymal microglia of the cortex and
hippocampus. (a) X-gal staining of the cortex in LysM-CRE+ mice crossed to a ROSA-LacZ
reporter have CRE-activity (dark staining) consistent with the distribution of microglia in the
cortex. LysM-CRE negative littermates have no staining (inset). Both scale bars = 1 mm. (b)
Confocal microscopy of the cortex in LysM-CRE+ mice showing examples of CD68+ cells that
also have CRE immunoreactivity (arrowheads) and examples of CD68+ cells that do not
(arrows). Scale bar = 10µm. (c) X-gal staining of the hippocampus in LysM-CRE+ mice crossed
to a ROSA-LacZ reporter exhibit increased CRE-activity in the CA3 region of the hippocampus
(dark staining) compared with WT controls (inset). Both scale bars = 1 mm. (d) Magnification of
the CA3 region of the next tissue section from the animal in panel c. The beta-galactosidase
reporter activity co-localizes with CD68+ cells (arrowheads). WT controls do not exhibit betagalactosidase immunoreactivity (inset). Both scale bars = 100 µm.
73
a
Coordinates relative to bregma:
-2.9 (± 0.1) mm
-1.6 (± 0.1) mm
WT
FBGRKO
WT
GR protein level
(% WT control)
3mo
Fig. 3. FBGRKO mice have GR
deleted in neurons of the cortex
GR C-term
and hippocampus. (a) X-gal
dMCAO
staining of the cortex in
*
120
*
CamKIIa-CRE+ mice that were
100
crossed to a ROSA-LacZ reporter
80
have CRE-activity (blue staining)
60
throughout the cortex and
40
hippocampus at 3, 4, and 5
20
m o n t h s o f a g e . Tw o
0
representative regions of the
N-term C-term
hippocampus are shown with
coordinates indicated relative to
the bregma. (b) GR protein levels
in the infarcted cortex of WT and
GR
FBGRKO mice 24 h postDAPI
MCAO. Western blots were
performed using antibodies
raised against the N- or Cterminus of the GR. n = 4. All
bands were normalized to betaactin controls. (c) Epi-fluorescent
microscopy for GR protein in the
hippocampus of FBGRKO and
WT mice. GR is deleted in
pyramidal neurons in FBGRKO
mice by two-months of age
(arrowheads), but there are still
many examples of GR colocalization with these cells at
this age (arrows). A second
example shows more efficient
deletion of the GR in the same
region (bottom). WT littermates
have widespread GR expression
throughout the hippocampus
pyramidal neurons (inset). All
scale bars = 100 µm.
b
GR N-term
4mo
FB W
G T
R
KO
FB W
G T
R
KO
5mo
c
FBGRKO
WT
CA3
CA3
FBGRKO
WT
CA3
CA3
74
a Promoter-CRE
GR protein level
(relative to WT control)
Fig. 4. Overexpression
of GR in microglia
ROSA-loxP-STOP-loxP-rtTA +DOX
GR and neurons using the
same CRE-drivers. (a)
Schematic of the
TRE-CMVmin-rGR
interactions between
MGROV
WT
NeuN
the transgenes used to
b
c FBGROV rtTA-IRES-GFP
FBGROV
overexpress the GR.
DAPI
OVERLAP
The CRE recombinase
250
excises a STOP
**
200
sequence allowing the
ubiquitous ROSA
150
promoter to drive
100
expression of rtTA in a
50
tissue-specific manner.
CA3
In the presence of
0
WT MGROV
DOX, the rtTA binds to
d
the TRE upstream of a
10 days DOX
7 days DOX
3 days DOX
No DOX
CMVmin promoter to
NeuN
Cortex
GR
increase expression of
DAPI
OVERLAP
the rat GR gene. (b)
Western blot for GR in
10 days DOX
microglia extracted
7 days DOX
5 days DOX
3 days DOX
from the brains of WT
CA1
and MGROV mice that
were given DOX for 7
7 days DOX
5 days DOX
3 days DOX
CA1
Cortex
days
(top,
3
CA3
representative bands
CA3
shown). Quantification
of GR levels shows that
microglial GR protein
levels are nearly
doubled by the
overexpression system
e
(microglia from 5 mice
new KA or dMCAO new
new GC pellet
begin
KA
were pooled per-band,
DOX dMCAO quant DOX quant
DOX or sham
DOX
n = 5 bands per
genotype). (c) The rtTA
8
11
10
4
6
7
9
2
3
5
Days: 1
transgene has an IRESGFP reporter so the
tissue-specific expression of this transgene can be monitored. An example of this is shown for
FBGROV mice where GFP expression co-localizes with NeuN neuronal labeling in the CA3
region of the hippocampus demonstrating tissue-specific rtTA expression. Scale bar = 100 µm.
(d) DOX was given to FBGROV mice for different lengths of time and GR co-localization with
NeuN was assessed in the cortex and CA1 and CA3 regions of the hippocampus. Several
representative epifluorescent images are shown from these different brain regions after different
durations of DOX-treatment. Three days of DOX visibly elevates GR co-localization with NeuN
(co-localization in white) over animals not given DOX (arrows indicate examples of this). Giving
DOX for 5 days increases GR signal relative to 3 days (arrowheads indicate examples of this), but
further increases beyond 5 days are difficult to distinguish. Scale bars = 100 µm. (e) Based on
these observations, this experimental design was used for the overexpression experiments. Fresh
DOX was given every 72 h, with experimental manipulations beginning only after 5 days of DOX
treatment to allow sufficient time for overexpression to occur.
75
Serum corticosterone
levels ( g/dL)
a
60
*
WT
MGRKO
50
40
30
20
10
0
Basal 0
b
2
6
24
Serum corticosterone
levels ( g/dL)
Time post-stress (h)
120
100
80
60
40
20
0
GC pellet: +
Days post-pellet: 1
KA on day 3: –
r2=0.81
***
+
3
–
+
6
–
+
4
+
***
+
6
+
–
–
4
+
6
+
Fig. 5. Blood GC levels in experimental model. (a) Serum corticosterone levels in WT and
MGRKO mice under basal conditions or 0, 2, 6, or 24 h after 2 h of immobilization stress in the
morning (n = 6). Stress stimulates endogenous GC release to almost the same extent in WT and
MGRKO mice. Circadian rhythm increases in endogenous GC release can begin to be seen 6 h
post-stress and are not different in MGRKO mice. (b) Serum corticosterone levels in WT mice at
different times after GC pellet implantation (n = 3-5 per group). GC pellets increased serum
corticosterone concentrations which decay linearly over time (R-squared = 0.8176, first three
bars). KA-treatment on day 3 post-pellet does not alter this progression and even at 6 days postpellet, serum GC levels are still elevated over non-pellet animals.
76
b
250 KA
CD68+ Area
(% WT Control)
a
200
Fig. 6. GCs enhance
cellular inflammation and
can both enhance and
150
suppress recruitment. (a)
100
Tissue sections subsequent to
50
those used for KA-lesion
CA3
0
quantification were stained
WT MGRKO
c 250 KA
with CD68, a marker
expressed by activated
200
myeloid cells. Scale bar =
150
100 µm. Magnification
100
shows ameboid morphology
50
of activated cells (inset).
P = 0.03
0
Scale bar = 20 µm. (b)
WT MGRKO
CD68+ area was quantified
d
e
f
P1
in the hippocampus
G1
G2
following standards for
stereological quantification
P2
of fluorescent signal (1). n =
P3
6-11 per group. GCtreatment increases CD68+
area in both WT and
FSC
CD45
MGRKO mice. (c) The pixel
i
g
h
intensity of the staining in
Monocytes (P2)
Microglia (P3)
Granulocytes (P1)
4
10
1.8
the CD68+ area quantified in
b. GC treatment increases
**
GC
treatment
1.6
8
3
Vehicle
CD68 intensity in WT but
1.4
6
not MGRKO mice. The P2
1.2
4
value indicated is for the
1
1.0
significant interaction. (d–i)
2
CNS myeloid cells were
P = 0.056
P = 0.009
0
0
0
isolated 24 h post-MCAO on
WT
MGRKO
WT MGRKO
WT MGRKO
a percoll gradient and
j
**
22.5
analyzed via flow cytometry
20.0
for expression of CD45,
17.5
15.0
LY6G, and CD11b. (d)
12.5
10.0
Events were gated based on
7.5
forward and side scatter (G1)
5.0
2.5
to eliminate debris. (e)
0.0
WT MGRKO
CD11b+ cells from G1 were
also gated for CD45+ signal
(G2). (f) G2 cells were differentiated into three populations using the granulocyte marker Ly6G
and their relative expression of CD45 (high or low). Granulocytes (P1) are Ly6G+ and CD45 low.
Myeloid cells are Ly6- and were split into CD45 high (P2), which are recruited, activated
mononuclear cells and CD45 low (P3) which are resting microglia. Each population was
calculated as a percentage of the G2 gate. n = 7-14 per group. (g) Fold-increase in the granulocyte
population (P1) in the infarcted hemisphere relative to the contralateral hemisphere. MGRKO
mice have lower granulocyte recruitment than WT controls. (h) Fold-increase in the activated
monocyte population (P2) in the infarcted hemisphere relative to the contralateral hemisphere.
GC-treatment decreases this population in WT but not MGRKO mice. (i) Fold-decrease in the
resting microglia population (P3). The percentage of resting cells is lower in the infarcted
hemisphere (reflected by a number < 1). (j) Leukocyte recruitment to a peripheral air-pouch
injury in WT and MGRKO littermates. n = 8 per group.
WT
***
CD68 Pixel Intensity
(% WT Control)
*
GC treatment
GC treatment
Vehicle
**
Vehicle
MGRKO
CD68
4
10 4
10
10 3
10 3
10 2
10 2
SSC-H: SSC-Height
1
10
SSC-H: SSC-Height
10 1
400
600
800
10 0
1000
0
**
200
10 1
10 2
10 3
14.1
56.7
0
0
Total Leukocyte Count (x10^6)
Fold increase over
contralateral hemisphere
0
<FL2-H>: PE
1
10
Ly6G
0
10 2
Fold decrease relative to
contralateral hemisphere
SSC
10
Fold increase over
contralateral hemisphere
SSC
17.5
10 3
77
10 1
10 2
10 3
a
Fig. 7. Use of flow
cytometry to identify CNS
leukocytes.
GR
overexpression does not
alter cell-recruitment. (a)
Representative sample of
b
the mononuclear cells
unstained
unstained
unstained
CD45 iso
CD11b iso
Ly6 iso
isolated from WT mouse
CD45+
CD45lo
CD11b+
Ly6+
brains by percoll gradient
processes
that were used for flow
CD11b+
cytometry experiments.
CD45hi
Ly6+
Isolated cells were allowed
to adhere to coverslips and
were stained for CD45 or
CD11b (top) to verify their
Ly6
CD11b
CD45
myeloid lineage identity.
GC treatment
Vehicle
Cells exhibited a
c
MGRKO
MGRKO
WT
WT
morphology consistent with
P1
P2
monocytes (DAPI, middle;
Brightfield, bottom). Scale
bar = 30 µm. (b)
P3
Representative flow
cytometry histograms
demonstrating the
specificity of the three
fluorescent antibodies used
to stain for CD45, CD11b,
and Ly6G. Unstained cells
(red trace) exhibit little
Ly6
background fluorescence.
Isotype control antibodies
f 2.0 Microglia (P3)
d 40 Granulocytes (P1) e 20 Monocytes (P2)
(blue trace) for each
channel have slightly more
30
background signal than
1.5
15
unstained controls.
Antigen-specific antibodies
20
1.0
10
(green trace) have peaks
corresponding with the
10
0.5
5
expected cell populations.
Low and high CD45 peaks
0
0.0
0
WT MGROV
WT MGROV
WT MGROV
(left) are distinguishable
from the isotype control.
CD11b signal (middle) is split into a population of cells and an off-axis population of brightlystained processes that were excluded from the analysis. The Ly6+ cell population (right) is clearly
distinguishable from isotype controls. (c) Representative flow cytometry results for the
quantifications presented in Fig. 2 g–i. For all plots, the Y-axis is CD45+ signal and the X-axis is
Ly6G+ signal. Populations in the infarcted hemisphere (top row) were normalized to the
contralateral hemisphere (bottom row) as an internal control for the quantification. (d) Foldincrease in the granulocyte population (P1) in the infarcted hemisphere relative to the
contralateral hemisphere. Granulocyte recruitment is not different in MGROV mice. (e) Foldincrease in the activated monocyte population (P2) in the infarcted hemisphere relative to the
contralateral hemisphere. Monocyte recruitment/activation is not different in MGROV mice. (f)
Fold-decrease in the resting microglia population (P3). The percentage of resting cells is lower in
the infarcted hemisphere (reflected by a number < 1). n = 8-10 (d–f).
Light
DAPI
CD45
CD11b
100
100
100
80
80
80
% of Max
40
% of Max
40
20
20
20
0
0
3
10
2
Contralateral hemisphere
0
0
10
1
10
10
<FL1-H>: FITC
1
10
58.8
25.2
10 1
10 2
<FL1-H>: FITC
10 1
69.7
10 3
20.6
10
1
10
2
10
2.59
10
1
10
2
10
3
12.4
5
1
10
2
10
<FL1-H>:
FITC
10 1
3
0
10
1
10
2
10
3
10
3
21.8
69
10 1
10 2
10 3
10 3
1.04
10 2
<FL1-H>: FITC
10 1
90.2
5.58
0
0
2
0
10
10 2
<FL1-H>: FITC
10 1
90.8
10
2
0.91
0
0
1
2.52
10
10 3
10 2
<FL1-H>: FITC
10 1
88.9
10
10 3
0
1.57
10 2
0
<FL1-H>: FITC
10 1
82.1
3
10 3
2.17
4
0
0
10 3
10
10 2
0
0
3
0.34
3.7
10 2
0
2
10 3
10 3
7.61
0
Fold increase over
contralateral hemisphere
10 3
Fold increase over
contralateral hemisphere
Infarcted hemisphere
10
10 2
<FL1-H>: FITC
10 1
3.73
91
0
0
10 1
10 2
10 3
Fold decrease relative to
contralateral hemisphere
10 1
0
CD45
60
60
60
% of Max
40
Light
DAPI
78
0
10
1
10
2
10
3
a
Vehicle
KA
1 week
3 days
Indomethacin
Minocycline
Indomethacin
3 days
Minocycline
CD11b/c
Intact
b
Intact = sham surgery
Low GC = ADX+15%
High GC = ADX+100%
High GC
Low GC
CA3
CD11b+ area
(% intact control)
c
d
*
150
**
*
Intact
Low GC
High GC
n.s.
100
50
0
Vehicle
Vehicle
Indomethacin
Indomethacin
Minocycline
Minocycline
Intact
GFAP
High GC
Low GC
CA3
GFAP+ area
(% intact control)
e
***
150
Intact
Low GC
High GC
100
50
0
Vehicle
Indomethacin
Minocycline
79
Fig. 8. Minocycline reverses GCaugmented immune cell activation in
the hippocampus. (A) Experimental
design: GC manipulations were
performed three days prior to KA, and
brains were analyzed three days later.
Indomethacin was injected every 12 h
starting 24 h before KA. Minocycline
was injected every 24 h starting 30 min
before KA. n = 5–13 per group. (B)
Representative images of CD11b/c signal
in the hippocampal CA3 region 72 h
post-injury. (C) The total amount of
CD11b/c+ area was calculated according
to established guidelines for quantifying
fluorescent signal (see methods). High
GCs increase CD11b/c+ area relative to
low GCs (* P < 0.05) and this is reversed
by minocycline (** P < 0.01). (D)
Representative images of GFAP signal in
the hippocampal CA3 region 72 h postinjury. (E) Quantification of GFAP+ area
in the same tissue sections. Minocycline
treatment reduces GFAP signal across all
treatment groups (*** P < 0.001). Scale
bars = 500 µm, 100 µm inset.
a Intact = sham surgery
KA
b
CCL2 concentration
(pg/ g total protein)
Low GC = ADX+15%
High GC = ADX+100% 1 week 3 days
CINC-1 concentration
(pg/ g total protein)
c
100 CCL2
80
Dissection times (h)
2 4 8 12 18 24
48
Intact
Low GC
High GC
60
40
20
0
1.2
CINC-1
1.0
0.8
0.6
0.4
0.2
0.0
0
4
8
12 16 20 24 28 32 36 40 44 48
Time after kainic acid (h)
Fig. 9. GCs do not increase cellular inflammation by enhancing chemokine production. (A)
Experimental design: CA3 tissue was collected at 2, 4, 8, 12, 18, 24, and 48 h post-KA injury. n =
3–6 per group. (B) CCL2 production peaks between 18-24 h post-KA and is unaffected by GC
treatment. (C) CINC-1 production peaks at 8 h post-KA and is also unaffected by GC treatment.
80
*
CX3CL1 CX3CR1 CD22
ct
Lo
H wG
ig C
h
G
C
0
e
150 NF- B
n.s.
100
50
0
1.5 I B
1.0
n.s.
0.5
0.0
t
Lo
H wG
ig C
h
G
C
10
d
ac
**
*
CD200 TGF-beta IL-1ra
In
t
IL-1
20
In
**
mRNA fold increase
relative to intact control
*
3 days
Intact
Low GC
High GC
t
Lo
H wG
ig C
h
G
C
30
1 week
ac
40
ta
hippocampus IL-1
(pg/mg total protein)
c
qPCR
In
t
6
5
4
3
2
1
0
p65 nuclear activity
(% WT control)
b
Intact = sham surgery
Low GC = ADX+15%
High GC = ADX+100%
mRNA fold increase
(relative to intact control)
a
Fig. 10. GCs suppress anti-inflammatory gene expression. (A) Experimental design: Rats were
treated with GCs and the hippocampus was dissected at the time when the injury would normally
be given. n = 5–6 per group. (B) Anti-inflammatory gene expression in the uninjured
hippocampus following each GC treatment, expressed as fold-increase relative to intact. GCs
suppress CX3CL1, CX3CR1, and CD22 expression (* P < 0.05, ** P < 0.01). (C) IL-1beta
protein levels in the uninjured hippocampus following each GC treatment. Low and high GCs
have higher IL-1beta levels than intact. (* P < 0.05, ** P < 0.01). (D) Quantification of EMSA
autoradiographs for the p65 subunit of NF-kB in the uninjured hippocampus. Basal NF-kB
activation is not affected by GC-treatments. (E) Quantitative PCR of IkBa in the hippocampus of
uninjured rats. IkBa expression is unchanged by GC-treatments.
81
a
MGRKO
WT
Vehicle
CX3CL1
Vehicle
GC treatment
GC treatment
12 h post-KA
72 h post-KA
12 h
*
150
120
90 *
60
P = 0.015
P = 0.135
12 h
GC treatment
72 h
f
90
60
P = 0.332
WT MGRKO
GC treatment
150
120
90 *
60
P = 0.362
WT FBGRKO WT FBGRKO
12 h post-KA
24 h
210 dMCAO
180
150
***
120
***
CX3CL1 protein level
(% WT control)
180
P = 0.241
***
120
Vehicle
210 KA
0
150
GC treatment
Vehicle
FBGRKO
WT
Vehicle
180
0
WT MGRKO WT MGRKO
d
CX3CL1 protein level
(% WT control)
210 dMCAO
CX3CL1 protein level
(% WT control)
180
CX3CL1
24 h
***
210 KA
0
e
c
72 h
***
CX3CL1 protein level
(% WT control)
b
90
60
0
P = 0.561
WT FBGRKO
Fig. 11. GCs affect CX3CL1 levels oppositely in the hippocampus post-KA and cortex postMCAO. (a) Western blot for CX3CL1 in the injured hippocampus at 12 and 72 h post-KA in WT
and MGRKO mice. Two representative bands are shown, n = 5–9 per group. All bands were
normalized to beta-actin controls. (b) Quantification of CX3CL1 protein levels in the injured
hippocampus 12 and 72 h post-KA in WT and MGRKO mice. GC-treatment suppresses CX3CL1
levels and MGRKO mice have significantly elevated CX3CL1 levels 72 h post-KA. (c)
Quantification of CX3CL1 protein levels in the infarcted hemisphere 24 h post-MCAO in WT
and MGRKO mice. GC-treatment significantly elevates CX3CL1 levels and MGRKO mice have
significantly lower CX3CL1 levels. (d) Western blot for CX3CL1 in the ischemic hemisphere 12
h post-KA in WT and FBGRKO mice. Two representative bands are shown, n = 5–9 per group.
(e) Quantification of CX3CL1 protein levels in the injured hippocampus 12 and 72 h post-KA in
WT and FBGRKO mice. GC-treatment suppresses CX3CL1 levels and FBGRKO mice have no
difference in CX3CL1. (f) Quantification of CX3CL1 protein levels in the infarcted hemisphere
24 h post-MCAO in WT and FBGRKO mice. GC-treatment significantly elevates CX3CL1 levels
and FBGRKO mice have significantly lower CX3CL1 levels.
82
b
a
IL-1 in ischemic cortex
(pg/mg protein)
*
4
*
2
0
60
WT
MGRKO
*
GC treatment
Vehicle
40
30
**
20
10
0
WT
WT
MGRKO
150
iNOS protein levels
(% WT control)
MGRKO
WT
iNOS
125
100
75
50
25
M W
G T
R
KO
FB W
G T
R
KO
0
f
3
*
WT
WT FBGRKO
dMCAO
GC treatment
Vehicle
*
15
10
**
5
**
0
WT
MGRKO
20
WT FBGRKO
FBGRKO
6
25
KA
50
9
0
d
WT FBGRKO
GC treatment
Vehicle
12
GC treatment
Vehicle
6
dMCAO
WT
MGRKO
150
125
WT
MGRKO
WT FBGRKO
FBGRKO
COX-2
**
100
75
50
25
0
M W
G T
R
KO
FB W
G T
R
KO
e
8
IL-6 in ischemic cortex
(pg/mg protein)
IL-6 in injured hippocampus
(pg/mg total protein)
c
15
KA
COX-2 protein levels
(% WT control)
IL-1 in injured hippocampus
(pg/mg total protein)
10
Fig. 12. GCs suppress pro-inflammatory cytokines and are necessary for COX-2 expression.
IL-1beta and IL-6 levels were measured by ELISA at 12 h post-KA (a,c) and at 24 h post-MCAO
(b,d). n = 5–9 per group. (a) Hippocampal IL-1beta levels in KA-treated MGRKO or FBGRKO
mice and WT controls. MGRKO, but not FBGRKO mice, have higher IL-1beta levels than WT.
(b) Infarcted hemisphere IL-1beta levels post-MCAO in MGRKO or FBGRKO mice and WT
controls. MGRKO, but not FBGRKO mice, have higher IL-1beta levels than WT. (c)
Hippocampal IL-6 levels in KA-treated MGRKO or FBGRKO mice and WT controls. MGRKO,
and nearly FBGRKO mice, have higher IL-6 levels than WT. (d) Infarcted hemisphere IL-6 levels
post-MCAO in MGRKO or FBGRKO mice and WT controls. GC-treatment suppresses IL-6
levels in WT mice, but not in MGRKO mice (interaction, F = 10.16, P = 0.005). MGRKO mice
have lower and FBGRKO mice have higher IL-6 levels than WT. (e) Quantification of western
83
blot for iNOS in MGRKO, FBGRKO, and WT controls 24 h post-MCAO. Neither GR KO affects
the levels of iNOS significantly. All bands were normalized to beta-actin controls. (f)
Quantification of western blot for COX-2 in MGRKO, FBGRKO, and WT controls 24 h postMCAO. FBGRKO mice have lower COX-2 levels relative to WT.
b
In
ta
c
15 t
%
30
%
60
10 %
0
In %
ta
c
15 t
%
30
%
60
10 %
0%
NF-KB
(p50/65)
***
Hippocampus p65
225
***
***
***
200
175
150
125
100
75
50
25
24 h post-KA
In
ta
15 ct
30%
6%
100%
0%
0
In
ta
15 ct
30%
60%
10 %
0%
Uninjured
250
NF B p65 nuclear activity
(% intact, uninjured control)
a
Uninjured
4
3
Hippocampus I B mRNA
Intact
Low GC
High GC
2
1
0
SAL
d
I B mRNA fold increase
(relative to intact control)
I B mRNA fold increase
(relative to intact control)
c
LPS
4
3
24 h post-KA
Frontal Cortex I B mRNA
Intact
Low GC
High GC
**
2
1
0
SAL
LPS
Fig. 13. GCs have an impaired ability to inhibit NF-kB in the hippocampus. (a)
Representative EMSA autoradiograph of the uninjured hippocampus or the injured hippocampus
24 h post-KA. The NF-kB-specific band representing p65/p50 heterodimers is indicated with an
arrow. (b) Densitometric analysis of the NF-kB-specific band. GCs do not affect basal NF-kB
activity, but do affect the magnitude of NF-kB activation post-KA in a dose-dependent manner. n
= 6 per group. (c) Quantitative PCR of IkB mRNA in the hippocampus at baseline and 2 h after
LPS challenge. GCs do not significantly affect IkB expression in this brain region. (d)
Quantitative PCR of IKB mRNA in the frontal cortex at baseline and 2 h after LPS challenge.
GCs induce IkB in the frontal cortex. n = 6 per group (c,d).
84
a
GC treatment
Vehicle
WT
MGRKO
p65
MGRKO
WT
12 h
post-KA
72 h
post-KA
24 h
post-dMCAO
c
72 h
12 h
175 dMCAO
75
50
d
WT MGRKO WT MGRKO
Vehicle
WT
p65
FBGRKO
GC treatment
Vehicle
150
125
***
**
*
100 *
*
p65 nuclear activity
(% WT control)
125
***
p65 nuclear activity
(% WT control)
150 KA
0
24 h
***
b
100
75
*
50
0
WT MGRKO
GC treatment
WT
FBGRKO
12 h
post-KA
24 h
post-dMCAO
12 h
150 KA
125
***
100 *
75
50
0
WT FBGRKO
f
p65 nuclear activity
(% WT control)
p65 nuclear activity
(% WT control)
e
24 h
150 dMCAO
125
GC treatment
Vehicle
***
100
75
50
0
WT FBGRKO
Fig. 14. GCs increase p65 nuclear localization. (a) NF-kB nuclear localization was assayed by
measuring its p65 subunit in the nuclear fraction isolated from the injured hippocampus at either
12 or 72 h post-KA or the injured cortex 24 h post-MCAO. ATwo representative bands are shown
for each treatment. n = 5-6 per group. All bands were normalized to beta-actin controls. (b)
Quantification of nuclear p65 levels in MGRKO mice post-KA. GC-treatment potentiates p65
activity in WT mice at both times and MGRKO mice at 12 h but after 72 h p65 activation is
truncated in MGRKO mice. MGRKO mice alone have reduced p65 levels at both time points. (c)
Quantification of nuclear p65 levels in MGRKO mice at 24 h post-MCAO. GC-treatment
potentiates p65 activity in WT mice but not in MGRKO mice. MGRKO mice again have lower
p65 levels relative to WT mice. (d) NF-kB activity was assayed in FBGRKO mice in a similar
manner. (e) Quantification of nuclear p65 levels in FBGRKO mice 12 h post-KA. GC-treatment
potentiates p65 activity in WT and FBGRKO mice alike. FBGRKO mice have reduced p65 levels
relative to WT littermates. (f) Quantification of nuclear p65 levels in FBGRKO mice 24 h postMCAO. GC treatment does not increase p65 nuclear localization, but like MGRKO mice,
FBGRKO mice have lower p65 levels than WT mice.
85
a
Intact = sham surgery
Low GC = ADX+15%
High GC = ADX+100%
Vehicle
b
KA
1 week
Minocycline
Indomethacin
3 days
3 days
Indomethacin Minocycline
CA1
Intact
DG
CA3
High GC
Low GC
CA4
Damaged CA3 (%)
c
70
60
50
40
30
20
10
0
Damaged CA4 (%)
d
**
***
*
**
Vehicle
Indomethacin
Minocycline
Vehicle
Indomethacin
Minocycline
Intact
Low GC
High GC
100
80
60
40
20
0
Fig. 15. GC-augmented inflammation is necessary for GC-endangerment. (A) Experimental
design as in Fig. 1A. n = 5–13 per group. (B) Representative images of Nissl-stained sections
following each treatment combination at 72 h post-KA. The CA3 region has the most quantifiable
damage (magnified below each image). Hippocampal sub-regions are indicated with dashed lines.
Arrows point to areas of neuron loss. Scale bars = 500 µm, (200 µm, magnification). (C)
Quantification of CA3 lesioned area 72 h post-injury. High GCs significantly increase CA3
damage relative to low GCs (** P < 0.01). Indomethacin does not protect from this (* P < 0.05)
but minocycline does (*** P < 0.001, ** P < 0.01). (D) Quantification of CA4 lesioned area 72 h
post-injury. No treatments affect the widespread damage in this region.
86
KA or dMCAO
24 h
dMCAO infarct quantification
GC treatment
GC treatment
Vehicle
WT
MGRKO
CA1
WT
c
MGRKO
DG
CA3
WT
MGRKO
Vehicle
FBGRKO
WT
WT
MGRKO
GC treatment
treatment
GC
WT
e
g
FBGRKO
CA1
DG
CA3
60
50
40
30
20
10
0
35 dMCAO
30
25
20
15
10
5
0
60
50
40
30
FBGRKO
WT
FBGRKO
20
10
0
P = 0.743
WT FBGRKO
60 dMCAO
50
40
30
***
Infarct Volume (mm3)
WT
GCtreatment
treatment
GC
*
i
Vehicle
P = 0.085
WT MGRKO
80 KA
70
CA3
h
P = 0.061
WT MGRKO
**
f
GC treatment
Infarct Volume (mm3)
Vehicle
% hippocampus damaged
d
GC treatment
Vehicle
**
CA3
80 KA
70
*
b
KA lesion quantification
***
3 days
GC pellet or vehicle
72 h
% hippocampus damaged
a
20
10
0
P = 0.480
WT FBGRKO
Fig. 16. GC-endangerment requires GR signaling in myeloid cells but not neurons. (a)
Experimental setup: WT and KO mice were given an s.c. 10 mg GC pellet and three days later
subjected to either stereotactic infusion of KA or the distal model of MCAO. Mice were killed 72
or 24 h post-injury. n = 5-12 per treatment group. (b) Representative Nissl-stained hippocampal
images from WT and MGRKO mice with or without GC pellets. Scale bar = 200 µm. The CA3
region of the hippocampus is magnified on the bottom. Scale bar = 50 µm. Arrows indicate
damaged neurons. (c) Quantification of the damaged hippocampal area. GC-treatment doubles
KA-induced hippocampal damage in WT mice but not in MGRKO littermates. (d) Representative
TTC stained tissue sections post-MCAO from MGRKO mice with or without GC pellets. The
white infarcted region is outlined. (e) Quantification of the infarct volume. GC-treatment
increases infarct volume in WT but not MGRKO littermates. (f) Representative Nissl-stained
hippocampal images from WT and FBGRKO mice with or without GC pellets. Scale bar = 200
µm. The CA3 subfield of the hippocampus is magnified on the bottom. Scale = 50 µm. Arrows
indicate damaged neurons. (g) Quantification of damaged hippocampal area. GC-treatment
increases KA-induced hippocampal damage in both WT and FBGRKO mice. (h) Representative
TTC stained tissue sections post-MCAO from WT or FBGRKO mice with or without GC pellets.
(i) Quantification of infarct volume. GC treatment increases MCAO infarct in both WT and
FBGRKO mice.
87
40
20
0
WT
100 KA
80
d
GC treatment
Vehicle
*
20
0
30
20
10
WT
MGROV
70 dMCAO
60
50
GC treatment
Vehicle
40
30
20
10
0
WT FBGROV
GC treatment
Vehicle
40
0
MGROV
60
40
Infarct Volume (mm3)
60
50 dMCAO
Infarct Volume (mm3)
80
GC treatment
Vehicle
***
% hippocampus damaged
100 KA
**
c
b
***
% hippocampus damaged
a
WT FBGROV
Fig. 17. GR overexpression has minimal effects on neuron death. (a) Quantification of the
damaged hippocampal area as in Fig. 1 in WT and MGROV mice 72 h post-KA. GC-treatment
increases KA-induced hippocampal damage irrespective of genotype. (b) Quantification of the
infarct volume in WT and MGROV mice 24 h post-MCAO. GC-treatment increases infarct
volume irrespective of genotype. (c) Quantification of damaged hippocampal area in WT and
FBGROV mice post-KA. GC-treatment increases KA-induced hippocampal damage in WT but
not FBGROV mice. (d) Quantification of infarct volume in WT and FBGROV mice post-MCAO.
GC treatment increases MCAO infarct irrespective of genotype.
88
1200
400
WT
300
*
200
100
P = 0.077
WT MGRKO
*
4000
2000
MGRKO
400 dMCAO
0
6000
d
BDNF in ischemic cortex
(pg/mg protein)
0
GC treatment
Vehicle
P = 0.12
WT FBGRKO
0
400 dMCAO
300
**
200
100
**
BDNF in ischemic cortex
(pg/mg protein)
c
800
8000 KA
*
1600 KA
BDNF in injured hippocampus
(pg/mg total protein)
b
**
BDNF in injured hippocampus
(pg/mg total protein)
a
*
*
0
WT
FBGRKO
Fig. 18. GCs suppress BDNF levels in the hippocampus and cortex in a cell-specific manner.
BDNF levels were measured by ELISA at 12 h post-KA (a,b) and at 24 h post-MCAO (c,d). n =
5–9 per group. (a) Hippocampal BDNF levels in KA-treated WT and MGRKO mice. GC
treatment reduces BDNF and does not require myeloid GR signaling for this to occur. (b)
Hippocampal BDNF levels in KA-treated WT and FBGRKO mice. GC treatment requires neuron
GR signaling to suppress BDNF levels. FBGRKO mice have lower BDNF levels than WT. (c)
Infarcted hemisphere BDNF levels post-MCAO in WT and MGRKO mice. GC treatment reduces
BDNF levels in WT but not MGRKO mice. (d) Infarcted hemisphere BDNF levels post-MCAO
in WT and FBGRKO mice. GC treatment suppresses BDNF levels in both WT and FBGRKO
mice. FBGRKO mice have higher BDNF levels than WT.
89
a
b
ATG
GR genomic locus:
5’UTR
Exon1C
loxP
Floxed GR allele:
5’UTR
EcoRV
SpeI
ExonIC
SpeI
700bp
SpeI
1kb
Tie-2-CRE+ :: floxed GR /
Tie-2-CRE- :: floxed GR /
Tie2-CRE+
rtTA-GFP
Lectin
WT
Merge
EGRKO
200
WT
EGROV
*
150
100
50
0
EG WT
R
KO
DAPI
WT
EG WT
R
O
V
e
d
GR protein level
(relative to WT control)
EGRKO:
WT littermate:
5’UTR
X-gal
loxP
ATG
loxP
Null allele:
c
$
Fig. 19. Endothelial cell KO and OV mice have higher GR levels in endothelial cells. (a)
Schematic of the three different alleles at the GR locus that were used in the creation of EGRKO
mice. Tie2-CRE deletes in the female germline leading to the generation of a null allele at the GR
locus. To improve recombination efficiency, the floxed-GR allele was placed over this deficiency
so that only one recombination event had to occur for the GR to be deleted. (b) DNA agarose gel
electrophoresis of PCR on genomic DNA from WT, floxed, and null mice using primers that span
the deleted region. The expected size of the deletion-spanning PCR product is ~700bp (arrow).
(c) X-gal staining of the hippocampal fissure in Tie2-CRE+ mice crossed to a ROSA-LacZ
reporter have CRE-activity (dark staining) consistent with blood vessel morphology. Scale bar =
10 µm. (d) Epifluorescence microscopy of the hippocampal fissure in EGROV mice. The
overexpression GFP reporter (green, arrows) co-localizes with the blood vessel label lectin (red).
Scale bars = 30 µm. (e) Western blot for GR on extracted blood vessels from the brains of
EGRKO and EGROV mice. EGROV mice have higher GR levels than WT littermates after both
were treated with DOX for 7 days. n = 7–9. EGRKO mice have a trend towards lower GR levels.
n = 4–5.
90
Infarct Volume (mm3)
50
dMCAO
40
30
20
10
0
c
*
b
80
Infarct Volume (mm3)
a
EGRKO
P = 0.057
40
***
20
0
d
100
50
0
e
WT
250
eNOS Protein Levels
(relative to WT)
WT EGRKO
EGRKO
200
150
100
50
WT EGRKO
EGROV
WT
EGROV
200
150
100
50
0
f
WT EGROV
WT
EGROV
250 eNOS
eNOS
P = 0.1
0
Occludin Protein Levels
(relative to WT)
150
WT
250 Occludin
eNOS Protein Levels
(relative to WT)
Occludin Protein Levels
(relative to WT)
200 Occludin
*
GC treatment
Vehicle
60
WT EGRKO
WT
dMCAO
200
150
100
50
0
WT EGROV
91
Fig. 20: Endothelial cell GR
signaling also worsens infarct
post-MCAO. (a) Quantification
of infarct volume as in Fig. 1,
24 h post-MCAO in WT and
EGRKO mice. EGRKO mice
have a smaller infarct than WT
littermates. n = 7. (b)
Quantification of infarct volume
in WT and EGROV mice. GC
treatment increases infarct
volume and EGROV mice also
have more damage than WT
controls. n = 8–10 vehicle, 3–5
GC-treated. (c) Quantification
of western blot for occludin
levels in EGRKO mice 24 h
post-MCAO. EGRKO mice
have higher occludin levels than
WT mice. n = 3–5. (d)
Quantification of western blot
for occludin levels in EGROV
mice 24 h post-MCAO. GR
overexpression does not
increase occludin levels. n = 4.
(e) Quantification of western
blot for eNOS levels in EGRKO
mice 24 h post-MCAO. There is
a trend towards increased eNOS
levels in EGRKO mice. n = 3–
5. (f) Quantification of western
blot for eNOS levels in EGROV
mice 24 h post-MCAO. GR
overexpression does not
increase eNOS levels. n = 4.
Excitotoxicity (KA)
Hippocampus
BBB
Microglia/Macrophages
EXO
ENDO
ENDO
EXO
(ENDO)
CD11b/c
GR
EXO
ENDO
Neuron Death
GR
EXO
ENDO
NF-KB
IL-1B
IL-6
Blood
Brain
EXO
Peripheral
Leukocytes
CD68
GR
CX3CR1
ENDO EXO
CD45
EXO
ia l
CD22
EXO
ENDO
En
NF-KB
dot
hel
GR
BDNF
cel
ls
CX3CL1
EXO
ENDO
Exogenous GCs = EXO
Endogenous GCs = ENDO
Stroke (dMCAO)
Cortex
BBB
Microglia/Macrophages
Blood
Brain
EXO
ENDO
ENDO
GR
NF-KB
IL-1B
CX3CL1 (in neurons)
GR
Peripheral
Leukocytes
IL-6
EXO
ENDO
ENDO
Occludin
(eNOS)
ls
GR
CX3CL1
GR
GR
cel
Neuron Death
dot
hel
ial
EXO
ENDO
COX-2
ENDO
EXO
(ENDO)
NF-KB
BDNF
En
GR
ENDO
EXO
EXO
ENDO
ENDO
Astrocytes
GR
Fig. 21. Summary of endogenous GC and exogenous GC effects during different injuries.
Exogenous GCs and endogenous GCs have often-opposing effects on parameters post-injury that
could affect inflammation and neuronal survival. The effects of the endogenous GCs released
post-injury are indicated as “ENDO” whereas the effects of exogenous GC treatments are
indicated as “EXO.” Effects that were trends but had not yet established significance are indicated
in grey parentheses. For effects that are dependent on GR signaling in a particular cell type, an
open, dashed arrow indicates this relationship and the stimulatory or inhibitory effects of
endogenous and exogenous GCs is indicated. Stimulating effects are indicated by arrows and
inhibitory effects are indicated by blunt-headed arrows.
Microglia/Macrophages
EXO
(ENDO)
EXO
ENDO
ENDO
NF-KB
IL-1B
IL-6
EXO
ENDO
EXO
ENDO
CD11b/c
GR
EXO
92
Blood
GR
Peripheral
Leukocytes
CD68
GR
CX3CR1
CX3CL1
CD45
EXO
GR
ells
Neuron Death
BBB
Brain
CD11b/c
PERIPHERAL
GC dose
Acute stressor
GC treatment time
Concurrent
Basal cort levels
Concurrent
Dexamethasone
Peripheral Location
Human PBMCs ex vivo,
THP-1 cell line
Mouse macrophage cell line
Pro-inflammatory effect
NFkB activity
Inflammatory
stimulus
None
Reference
(Bierhaus et al. 2003)
MIF secretion
None
(Calandra et al. 1995)
Human PBMCs ex vivo
Cytokines, complement,
chemokines
5-lipoxygenase
None
(Galon et al. 2002)
None
(Riddick et al. 1997)
Dexamethasone
Concurrent
Concurrent
Dexamethasone
Concurrent
Human PBMCs ex vivo,
THP-1 cell line
THP-1 cell line
IL-1 beta
PMA
(Wang et al. 1997)
Dexamethasone
Concurrent
THP-1 cell line
NFkB activity
PMA
(Wang et al. 1997)
Dexamethasone or
prednisolone
>16 hours prior
Human lymphocyte cultures
Proliferation
Low stress cort levels
Acute 2hrs prior
Leukocyte migration
Low stress cort levels
Concurrent
Rat leukocytes in vivo
and Mouse leukocytes in vivo
Rat lymphocyte cultures
Proliferation
Mitogens, x-linking (Almawi et al. 1999)
Abs, or PMA
+ionomycin
DTH
(Dhabhar et al. 1996)
(Dhabhar et al. 2005)
IL-2
(Wiegers et al. 1995)
Low stress cort levels
>6 hours prior
Mouse macrophage cell line
NFkB activity
LPS and IFNγ
(Smyth et al. 2004)
Low stress cort levels
>6 hours prior
Mouse macrophage cell line
TNF-alpha, IL-6, nitrite
LPS and IFNγ
(Smyth et al. 2004)
Low stress cort levels
>12 hours prior
Human plasma
TNF-alpha, IL-6
LPS
(Barber et al. 1993)
Reference
(Madrigal et al. 2001)
CENTRAL
GC treatment
Acute stressor
GC treatment time
Concurrent
Central Location
Rat cortex
Pro-inflammatory effect
NFkB activity
Inflammatory
stimulus
None
Acute stressor
24 hours prior
Rat brain
IL-1 beta
Peripheral LPS
(Johnson et al. 2002)
Acute stressor
24 hours prior
Rat hippocampal microglia
IL-1 beta
Peripheral LPS
(Frank et al. 2006)
CUS
Chronic 9 days after
Rat prefrontal cortex
Microglia activation
Central LPS
(de Pablos et al. 2006)
CUS
Chronic 9 days after
Rat prefrontal cortex
TNF-alpha
Central LPS
(de Pablos et al. 2006)
CUS
Chronic 9 days after
Rat prefrontal cortex
MAPK signaling
Central LPS
(de Pablos et al. 2006)
CUS
Chronic 14 days prior
Rat hippocampus, frontal cortex
TNF-alpha, IL-1 beta
Peripheral LPS
(Munhoz et al. 2006)
CUS
Chronic 14 days prior
Rat hippocampus, frontal cortex
NFkB activity
Peripheral LPS
(Munhoz et al. 2006)
Dexamethasone
Concurrent
Mouse neuronal cell line
PGD2 synthase
None
Dexamethasone*
One hour prior
Mouse hippocampus
TMT
High stress cort levels
Chronic 3 days prior
Rat hippocampus
TNF-alpha, TNF-beta,
IL-1 beta
Leukocyte migration
(Garcia-Fernandez et
al. 2000)
(Bruccoleri et al. 1999)
Kainic acid
(Dinkel et al. 2003)
High stress cort levels
Chronic 3 days prior
Rat hippocampus
Microglia activation
Kainic acid
(Dinkel et al. 2003)
High stress cort levels
Chronic 24 hours prior
Rat hippocampal cultures
TNF-alpha, IL-1 beta
Kainic acid
Intermediate stress cort levels
Chronic 3 days prior
Rat hippocampus, frontal cortex
TNF-alpha, IL-1 beta
Peripheral LPS
(MacPherson et al.
2005)
(Munhoz in prep)
Intermediate stress cort levels
Chronic 3 days prior
Rat hippocampus, frontal cortex
NFkB activity
Peripheral LPS
(Munhoz in prep)
Low stress cort levels
Chronic 10 days
Rat hippocampus, cerebellum
5-lipoxygenase
None
(Uz et al. 1999)
Table 1. Some experimental conditions where GCs have been found to increase
inflammation.
The variety of experimental conditions in which GC enhanced inflammation has been observed
makes definitive conclusions difficult. The top table is a non-comprehensive list of proinflammatory effects of GCs on peripheral immunity. The bottom table lists conditions where
GCs have been found to enhance signs of central inflammation and is more extensive. The type of
GC treatment varies between experimental groups; however broad categorizations have been
assigned. Basal stress cort levels are between 1-10µg/dL, low stress cort levels are between
10-25µg/dL, and high stress cort levels are above 25µg/dL. Dexamethasone doses are roughly
equivalent to high stress cort doses at 10-7M. GC treatment time refers to the time of exposure to
GCs relative to a subsequent inflammatory stimulus or measurement. See chapter 1 for a more
detailed description of many of these investigations.
*study potentially affected by poor BBB permeability of dexamethasone (see text, section 5.2.2).
Ab = antibody; cort = corticosterone (in rodent studies) or cortisol (in primate studies); CUS =
chronic unpredictable stress; LPS = lipopolysaccharride; PMA is the NFκB activator phorbol 12myristate 13-acetate; THP-1 = human monocyte cell line; TMT = trimethyltin.
93
Type of GC
Endogenous (corticosterone)
Synthetic GC (dexamethasone)
PRO
ANTI
Timing of GC exposure
Before injury
During injury
After injury
Before, during and after injury
PRO
ANTI
ANTI
PRO
Duration of GC exposure
Acute (single exposure)
Sub-acute (3-7 days)
Chronic (>3 weeks)
PRO
PRO
ANTI
GC concentration
Basal
Moderate stress
High stress
Supraphysiological
PRO
PRO*
PRO
ANTI
Table 2: Summary of parameters that impact how GCs affect inflammatory responses.
Broad generalizations are given as to whether each variable tends to cause GCs to increase (PRO)
or decrease (ANTI) inflammatory measurements. Clearly this is an over-simplification and
counter-examples are prevalent, but a large percentage of these counter-examples may result from
unknown rules about which of these parameters has priority over the others. The * is an example
of this, as moderate GCs caused by chronic mild stress are often pro-inflammatory despite the
chronic nature of the exposure, which should be anti-inflammatory. This dissertation primarily
addressed the effects of endogenous GCs, both basally and during sub-acute exposure before,
during, and after the injury at moderate to high-stress levels.
94
References:
1.
Sapolsky RM, Romero LM, & Munck AU (2000) How do glucocorticoids influence
stress responses? Integrating permissive, suppressive, stimulatory, and preparative
actions. Endocr Rev 21:55-89.
2.
Cannon WB (1935) Stresses and strains of homeostasis. American Journal of the Medical
Sciences 189:1-14.
3.
Selye H (1936) A syndrome produced by diverse nocuous agents. Nature 138:32-32.
4.
Cannon WB & Pereira JR (1924) Increase of Adrenal Secretion in Fever. Proc Natl Acad
Sci U S A 10:247-248.
5.
Selye H (1936) Thymus and adrenals in the response of the organism to injuries and
intoxications. British Journal of Experimental Pathology 17:234-248.
6.
Dhabhar FS (2009) Enhancing versus suppressive effects of stress on immune function:
implications for immunoprotection and immunopathology. Neuroimmunomodulation
16:300-317.
7.
Sorrells SF, Caso JR, Munhoz CD, & Sapolsky RM (2009) The stressed CNS: when
glucocorticoids aggravate inflammation. Neuron 64:33-39.
8.
Black PH (2002) Stress and the inflammatory response: a review of neurogenic
inflammation. Brain Behav Immun 16:622-653.
9.
Yeager MP, Guyre PM, & Munck AU (2004) Glucocorticoid regulation of the
inflammatory response to injury. Acta Anaesthesiol Scand 48:799-813.
10.
Munck A, Guyre PM, & Holbrook NJ (1984) Physiological functions of glucocorticoids
in stress and their relation to pharmacological actions. Endocr Rev 5:25-44.
11.
Rosenberger PH, et al. (2009) Surgical stress-induced immune cell redistribution profiles
predict short-term and long-term postsurgical recovery. A prospective study. J Bone Joint
Surg Am 91:2783-2794.
95
12.
Tarcic N, Ovadia H, Weiss DW, & Weidenfeld J (1998) Restraint stress-induced thymic
involution and cell apoptosis are dependent on endogenous glucocorticoids. J
Neuroimmunol 82:40-46.
13.
Elenkov IJ & Chrousos GP (2002) Stress hormones, proinflammatory and
antiinflammatory cytokines, and autoimmunity. Ann N Y Acad Sci 966:290-303.
14.
De Bosscher K, Vanden Berghe W, & Haegeman G (2003) The interplay between the
glucocorticoid receptor and nuclear factor-kappaB or activator protein-1: molecular
mechanisms for gene repression. Endocr Rev 24:488-522.
15.
De Kloet ER, Vreugdenhil E, Oitzl MS, & Joels M (1998) Brain corticosteroid receptor
balance in health and disease. Endocr Rev 19:269-301.
16.
Karst H, et al. (2005) Mineralocorticoid receptors are indispensable for nongenomic
modulation of hippocampal glutamate transmission by corticosterone. Proc Natl Acad Sci
U S A 102:19204-19207.
17.
Mulholland PJ, et al. (2006) A 24 h corticosterone exposure exacerbates excitotoxic insult
in rat hippocampal slice cultures independently of glucocorticoid receptor activation or
protein synthesis. Brain Res 1082:165-172.
18.
Dhabhar FS & McEwen BS (1999) Enhancing versus suppressive effects of stress
hormones on skin immune function. Proc Natl Acad Sci U S A 96:1059-1064.
19.
O'Connor KA, et al. (2003) Peripheral and central proinflammatory cytokine response to
a severe acute stressor. Brain Res 991:123-132.
20.
Deinzer R, et al. (2004) Acute stress effects on local Il-1beta responses to pathogens in a
human in vivo model. Brain Behav Immun 18:458-467.
21.
Bierhaus A, et al. (2003) A mechanism converting psychosocial stress into mononuclear
cell activation. Proc Natl Acad Sci U S A 100:1920-1925.
22.
Madrigal JL, et al. (2001) Inducible nitric oxide synthase expression in brain cortex after
acute restraint stress is regulated by nuclear factor kappaB-mediated mechanisms. J
Neurochem 76:532-538.
96
23.
Goujon E, et al. (1995) Stress downregulates lipopolysaccharide-induced expression of
proinflammatory cytokines in the spleen, pituitary, and brain of mice. Brain Behav
Immun 9:292-303.
24.
Viswanathan K & Dhabhar FS (2005) Stress-induced enhancement of leukocyte
trafficking into sites of surgery or immune activation. Proc Natl Acad Sci U S A
102:5808-5813.
25.
Johnson JD, et al. (2005) Catecholamines mediate stress-induced increases in peripheral
and central inflammatory cytokines. Neuroscience 135:1295-1307.
26.
Galon J, et al. (2002) Gene profiling reveals unknown enhancing and suppressive actions
of glucocorticoids on immune cells. Faseb J 16:61-71.
27.
Reyes TM, Walker JR, DeCino C, Hogenesch JB, & Sawchenko PE (2003) Categorically
distinct acute stressors elicit dissimilar transcriptional profiles in the paraventricular
nucleus of the hypothalamus. J Neurosci 23:5607-5616.
28.
Dhabhar FS (2008) Enhancing versus Suppressive Effects of Stress on Immune Function:
Implications for Immunoprotection versus Immunopathology. Allergy Asthma Clin
Immunol 4:2-11.
29.
Dhabhar FS (2002) Stress-induced augmentation of immune function--the role of stress
hormones, leukocyte trafficking, and cytokines. Brain Behav Immun 16:785-798.
30.
Yoshimura C, et al. (2001) Glucocorticoids induce basophil apoptosis. J Allergy Clin
Immunol 108:215-220.
31.
Cox G (1995) Glucocorticoid treatment inhibits apoptosis in human neutrophils.
Separation of survival and activation outcomes. J Immunol 154:4719-4725.
32.
Liles WC, Dale DC, & Klebanoff SJ (1995) Glucocorticoids inhibit apoptosis of human
neutrophils. Blood 86:3181-3188.
33.
Dhabhar FS & McEwen BS (1997) Acute stress enhances while chronic stress suppresses
cell-mediated immunity in vivo: a potential role for leukocyte trafficking. Brain Behav
Immun 11:286-306.
97
34.
Dhabhar FS, Miller AH, McEwen BS, & Spencer RL (1996) Stress-induced changes in
blood leukocyte distribution. Role of adrenal steroid hormones. J Immunol
157:1638-1644.
35.
McEwen BS, et al. (1997) The role of adrenocorticoids as modulators of immune
function in health and disease: neural, endocrine and immune interactions. Brain Res
Brain Res Rev 23:79-133.
36.
Blecha F, Barry RA, & Kelley KW (1982) Stress-induced alterations in delayed-type
hypersensitivity to SRBC and contact sensitivity to DNFB in mice. Proc Soc Exp Biol
Med 169:239-246.
37.
Blecha F, Kelley KW, & Satterlee DG (1982) Adrenal involvement in the expression of
delayed-type hypersensitivity to SRBC and contact sensitivity to DNFB in stressed mice.
Proc Soc Exp Biol Med 169:247-252.
38.
Wiegers GJ & Reul JM (1998) Induction of cytokine receptors by glucocorticoids:
functional and pathological significance. Trends Pharmacol Sci 19:317-321.
39.
Munck A & Naray-Fejes-Toth A (1992) The ups and downs of glucocorticoid physiology.
Permissive and suppressive effects revisited. Mol Cell Endocrinol 90:C1-4.
40.
Wiegers GJ, et al. (1995) Glucocorticoids accelerate anti-T cell receptor-induced T cell
growth. J Immunol 155:1893-1902.
41.
Almawi WY, Hess DA, Assi JW, Chudzik DM, & Rieder MJ (1999) Pretreatment with
glucocorticoids enhances T-cell effector function: possible implication for immune
rebound accompanying glucocorticoid withdrawal. Cell Transplant 8:637-647.
42.
Sterzer P, Wiegers GJ, & Reul JM (2004) Long-term in vivo administration of
glucocorticoid hormones attenuates their capacity to accelerate in vitro proliferation of rat
splenic T cells. Endocrinology 145:3630-3638.
43.
Bernhagen J, et al. (1994) Purification, bioactivity, and secondary structure analysis of
mouse and human macrophage migration inhibitory factor (MIF). Biochemistry
33:14144-14155.
98
44.
Bernhagen J, et al. (1996) An essential role for macrophage migration inhibitory factor in
the tuberculin delayed-type hypersensitivity reaction. J Exp Med 183:277-282.
45.
Petrovsky N, et al. (2003) Macrophage migration inhibitory factor exhibits a pronounced
circadian rhythm relevant to its role as a glucocorticoid counter-regulator. Immunol Cell
Biol 81:137-143.
46.
Calandra T, et al. (1995) MIF as a glucocorticoid-induced modulator of cytokine
production. Nature 377:68-71.
47.
Donnelly SC & Bucala R (1997) Macrophage migration inhibitory factor: a regulator of
glucocorticoid activity with a critical role in inflammatory disease. Mol Med Today
3:502-507.
48.
Smyth GP, et al. (2004) Glucocorticoid pretreatment induces cytokine overexpression
and nuclear factor-kappaB activation in macrophages. J Surg Res 116:253-261.
49.
Barber AE, et al. (1993) Glucocorticoid therapy alters hormonal and cytokine responses
to endotoxin in man. J Immunol 150:1999-2006.
50.
Johnson JD, et al. (2002) Prior stressor exposure primes the HPA axis.
Psychoneuroendocrinology 27:353-365.
51.
Johnson JD, et al. (2002) Prior stressor exposure sensitizes LPS-induced cytokine
production. Brain Behav Immun 16:461-476.
52.
Johnson JD, O'Connor KA, Hansen MK, Watkins LR, & Maier SF (2003) Effects of prior
stress on LPS-induced cytokine and sickness responses. Am J Physiol Regul Integr Comp
Physiol 284:R422-432.
53.
Davalos D, et al. (2005) ATP mediates rapid microglial response to local brain injury in
vivo. Nat Neurosci 8:752-758.
54.
Bruce AJ, et al. (1996) Altered neuronal and microglial responses to excitotoxic and
ischemic brain injury in mice lacking TNF receptors. Nat Med 2:788-794.
55.
Allan SM & Rothwell NJ (2001) Cytokines and acute neurodegeneration. Nat Rev
Neurosci 2:734-744.
99
56.
Viviani B, et al. (2003) Interleukin-1beta enhances NMDA receptor-mediated
intracellular calcium increase through activation of the Src family of kinases. J Neurosci
23:8692-8700.
57.
Myer DJ, Gurkoff GG, Lee SM, Hovda DA, & Sofroniew MV (2006) Essential
protective roles of reactive astrocytes in traumatic brain injury. Brain 129:2761-2772.
58.
Dinkel K, Dhabhar FS, & Sapolsky RM (2004) Neurotoxic effects of polymorphonuclear
granulocytes on hippocampal primary cultures. Proc Natl Acad Sci U S A 101:331-336.
59.
Watanabe H, Abe H, Takeuchi S, & Tanaka R (2000) Protective effect of microglial
conditioning medium on neuronal damage induced by glutamate. Neurosci Lett
289:53-56.
60.
del Zoppo GJ, Becker KJ, & Hallenbeck JM (2001) Inflammation after stroke: is it
harmful? Arch Neurol 58:669-672.
61.
McEwen BS & Magarinos AM (2001) Stress and hippocampal plasticity: implications for
the pathophysiology of affective disorders. Hum Psychopharmacol 16:S7-S19.
62.
Nadeau S & Rivest S (2003) Glucocorticoids play a fundamental role in protecting the
brain during innate immune response. J Neurosci 23:5536-5544.
63.
Dinkel K, MacPherson A, & Sapolsky RM (2003) Novel glucocorticoid effects on acute
inflammation in the CNS. J Neurochem 84:705-716.
64.
Sapolsky RM & Pulsinelli WA (1985) Glucocorticoids potentiate ischemic injury to
neurons: therapeutic implications. Science 229:1397-1400.
65.
Tombaugh GC, Yang SH, Swanson RA, & Sapolsky RM (1992) Glucocorticoids
exacerbate hypoxic and hypoglycemic hippocampal injury in vitro: biochemical
correlates and a role for astrocytes. J Neurochem 59:137-146.
66.
McIntosh LJ & Sapolsky RM (1996) Glucocorticoids may enhance oxygen radicalmediated neurotoxicity. Neurotoxicology 17:873-882.
100
67.
Goodman Y, Bruce AJ, Cheng B, & Mattson MP (1996) Estrogens attenuate and
corticosterone exacerbates excitotoxicity, oxidative injury, and amyloid beta-peptide
toxicity in hippocampal neurons. J Neurochem 66:1836-1844.
68.
Olgemoller B, Schon J, & Wieland OH (1985) Endothelial plasma membrane is a
glucocorticoid-regulated barrier for the uptake of glucose into the cell. Mol Cell
Endocrinol 43:165-171.
69.
Stein-Behrens BA, et al. (1992) Glucocorticoids exacerbate kainic acid-induced
extracellular accumulation of excitatory amino acids in the rat hippocampus. J
Neurochem 58:1730-1735.
70.
Stein-Behrens BA, Lin WJ, & Sapolsky RM (1994) Physiological elevations of
glucocorticoids potentiate glutamate accumulation in the hippocampus. J Neurochem
63:596-602.
71.
Elliott EM & Sapolsky RM (1993) Corticosterone impairs hippocampal neuronal calcium
regulation--possible mediating mechanisms. Brain Res 602:84-90.
72.
Sapolsky RM (1999) Glucocorticoids, stress, and their adverse neurological effects:
relevance to aging. Exp Gerontol 34:721-732.
73.
Reagan LP (2002) Glucose, stress, and hippocampal neuronal vulnerability. Int Rev
Neurobiol 51:289-324.
74.
Patel PM, et al. (1993) Effect of ibuprofen on regional eicosanoid production and
neuronal injury after forebrain ischemia in rats. Brain Res 614:315-324.
75.
Hara K, Kong DL, Sharp FR, & Weinstein PR (1998) Effect of selective inhibition of
cyclooxygenase 2 on temporary focal cerebral ischemia in rats. Neurosci Lett 256:53-56.
76.
Golde S, Coles A, Lindquist JA, & Compston A (2003) Decreased iNOS synthesis
mediates dexamethasone-induced protection of neurons from inflammatory injury in
vitro. Eur J Neurosci 18:2527-2537.
101
77.
Nadeau S & Rivest S (2002) Endotoxemia prevents the cerebral inflammatory wave
induced by intraparenchymal lipopolysaccharide injection: role of glucocorticoids and
CD14. J Immunol 169:3370-3381.
78.
Rivest S (2003) Molecular insights on the cerebral innate immune system. Brain Behav
Immun 17:13-19.
79.
Simmons CP, et al. (2005) The clinical benefit of adjunctive dexamethasone in
tuberculous meningitis is not associated with measurable attenuation of peripheral or
local immune responses. J Immunol 175:579-590.
80.
Gomes JA, Stevens RD, Lewin JJ, 3rd, Mirski MA, & Bhardwaj A (2005) Glucocorticoid
therapy in neurologic critical care. Crit Care Med 33:1214-1224.
81.
de Pablos RM, et al. (2006) Stress increases vulnerability to inflammation in the rat
prefrontal cortex. J Neurosci 26:5709-5719.
82.
Crossin KL, Tai MH, Krushel LA, Mauro VP, & Edelman GM (1997) Glucocorticoid
receptor pathways are involved in the inhibition of astrocyte proliferation. Proc Natl
Acad Sci U S A 94:2687-2692.
83.
Goujon E, Parnet P, Cremona S, & Dantzer R (1995) Endogenous glucocorticoids down
regulate central effects of interleukin-1 beta on body temperature and behaviour in mice.
Brain Res 702:173-180.
84.
Goujon E, Parnet P, Laye S, Combe C, & Dantzer R (1996) Adrenalectomy enhances proinflammatory cytokines gene expression, in the spleen, pituitary and brain of mice in
response to lipopolysaccharide. Brain Res Mol Brain Res 36:53-62.
85.
Munhoz CD, et al. (2006) Chronic unpredictable stress exacerbates lipopolysaccharideinduced activation of nuclear factor-kappaB in the frontal cortex and hippocampus via
glucocorticoid secretion. J Neurosci 26:3813-3820.
86.
MacPherson A, Dinkel K, & Sapolsky R (2005) Glucocorticoids worsen excitotoxininduced expression of pro-inflammatory cytokines in hippocampal cultures. Exp Neurol
194:376-383.
102
87.
Munhoz CD, Sorrells SF, Caso JR, Scavone C, & Sapolsky RM (2010) Glucocorticoids
exacerbate lipopolysaccharide-induced signaling in the frontal cortex and hippocampus in
a dose-dependent manner. J Neurosci 30:13690-13698.
88.
Garcia-Fernandez LF, et al. (2000) Dexamethasone induces lipocalin-type prostaglandin
D synthase gene expression in mouse neuronal cells. J Neurochem 75:460-470.
89.
Uz T, et al. (1999) Glucocorticoids stimulate inflammatory 5-lipoxygenase gene
expression and protein translocation in the brain. J Neurochem 73:693-699.
90.
Riddick CA, Ring WL, Baker JR, Hodulik CR, & Bigby TD (1997) Dexamethasone
increases expression of 5-lipoxygenase and its activating protein in human monocytes
and THP-1 cells. Eur J Biochem 246:112-118.
91.
Schneider N, Lanz S, Ramer R, Schaefer D, & Goppelt-Struebe M (2001) Up-regulation
of cyclooxygenase-1 in neuroblastoma cell lines by retinoic acid and corticosteroids. J
Neurochem 77:416-424.
92.
Unlap T & Jope RS (1995) Inhibition of NFkB DNA binding activity by glucocorticoids
in rat brain. Neurosci Lett 198:41-44.
93.
Unlap MT & Jope RS (1997) Dexamethasone attenuates NF-kappa B DNA binding
activity without inducing I kappa B levels in rat brain in vivo. Brain Res Mol Brain Res
45:83-89.
94.
Wang Z, et al. (2006) The effects of dexamethasone on rat brain cortical nuclear factor
kappa B (NF-kappaB) in endotoxic shock. Toxicol Appl Pharmacol 214:263-269.
95.
Frank MG, Miguel ZD, Watkins LR, & Maier SF (2009) Prior exposure to
glucocorticoids sensitizes the neuroinflammatory and peripheral inflammatory responses
to E. coli lipopolysaccharide. Brain Behav Immun.
96.
Bowers SL, Bilbo SD, Dhabhar FS, & Nelson RJ (2008) Stressor-specific alterations in
corticosterone and immune responses in mice. Brain Behav Immun 22:105-113.
97.
Sapolsky RM (1996) Stress, glucocorticoids, and damage to the nervous system: the
current state of confusion. Stress 1:1-19.
103
98.
Brewer JA, et al. (2003) T-cell glucocorticoid receptor is required to suppress COX-2mediated lethal immune activation. Nat Med 9:1318-1322.
99.
Bhattacharyya S, Brown DE, Brewer JA, Vogt SK, & Muglia LJ (2007) Macrophage
glucocorticoid receptors regulate Toll-like receptor 4-mediated inflammatory responses
by selective inhibition of p38 MAP kinase. Blood 109:4313-4319.
100.
Boyle MP, et al. (2005) Acquired deficit of forebrain glucocorticoid receptor produces
depression-like changes in adrenal axis regulation and behavior. Proc Natl Acad Sci U S
A 102:473-478.
101.
Cardona AE, Huang D, Sasse ME, & Ransohoff RM (2006) Isolation of murine
microglial cells for RNA analysis or flow cytometry. Nat Protoc 1:1947-1951.
102.
Cho IH, et al. (2008) Role of microglial IKKbeta in kainic acid-induced hippocampal
neuronal cell death. Brain 131:3019-3033.
103.
Monje ML, Toda H, & Palmer TD (2003) Inflammatory blockade restores adult
hippocampal neurogenesis. Science 302:1760-1765.
104.
Yrjanheikki J, et al. (1999) A tetracycline derivative, minocycline, reduces inflammation
and protects against focal cerebral ischemia with a wide therapeutic window. Proc Natl
Acad Sci U S A 96:13496-13500.
105.
Nair A & Bonneau RH (2006) Stress-induced elevation of glucocorticoids increases
microglia proliferation through NMDA receptor activation. J Neuroimmunol 171:72-85.
106.
Sapolsky R, Brooke S, & Stein-Behrens B (1995) Methodologic issues in studying
glucocorticoid-induced damage to neurons. J Neurosci Methods 58:1-15.
107.
Stavreva DA, et al. (2009) Ultradian hormone stimulation induces glucocorticoid
receptor-mediated pulses of gene transcription. Nat Cell Biol 11:1093-1102.
108.
Madrigal JL, et al. (2003) Induction of cyclooxygenase-2 accounts for restraint stressinduced oxidative status in rat brain. Neuropsychopharmacology 28:1579-1588.
104
109.
Garcia-Bueno B, Madrigal JL, Perez-Nievas BG, & Leza JC (2008) Stress mediators
regulate brain prostaglandin synthesis and peroxisome proliferator-activated receptorgamma activation after stress in rats. Endocrinology 149:1969-1978.
110.
Madrigal JL, Garcia-Bueno B, Caso JR, Perez-Nievas BG, & Leza JC (2006) Stressinduced oxidative changes in brain. CNS Neurol Disord Drug Targets 5:561-568.
111.
Biber K, Neumann H, Inoue K, & Boddeke HW (2007) Neuronal 'On' and 'Off' signals
control microglia. Trends Neurosci 30:596-602.
112.
Ransohoff RM & Cardona AE (The myeloid cells of the central nervous system
parenchyma. Nature 468:253-262.
113.
Mott RT, et al. (2004) Neuronal expression of CD22: novel mechanism for inhibiting
microglial proinflammatory cytokine production. Glia 46:369-379.
114.
Hoek RM, et al. (2000) Down-regulation of the macrophage lineage through interaction
with OX2 (CD200). Science 290:1768-1771.
115.
Cardona AE, et al. (2006) Control of microglial neurotoxicity by the fractalkine receptor.
Nat Neurosci 9:917-924.
116.
Nguyen KT, et al. (1998) Exposure to acute stress induces brain interleukin-1beta protein
in the rat. J Neurosci 18:2239-2246.
117.
Harrison JK, et al. (1998) Role for neuronally derived fractalkine in mediating
interactions between neurons and CX3CR1-expressing microglia. Proc Natl Acad Sci U S
A 95:10896-10901.
118.
Bhavsar PK, Sukkar MB, Khorasani N, Lee KY, & Chung KF (2008) Glucocorticoid
suppression of CX3CL1 (fractalkine) by reduced gene promoter recruitment of NFkappaB. Faseb J 22:1807-1816.
119.
Fong AM, et al. (1998) Fractalkine and CX3CR1 mediate a novel mechanism of
leukocyte capture, firm adhesion, and activation under physiologic flow. J Exp Med
188:1413-1419.
105
120.
Soriano SG, et al. (2002) Mice deficient in fractalkine are less susceptible to cerebral
ischemia-reperfusion injury. J Neuroimmunol 125:59-65.
121.
Roy M & Sapolsky RM (2003) The exacerbation of hippocampal excitotoxicity by
glucocorticoids is not mediated by apoptosis. Neuroendocrinology 77:24-31.
122.
Kroll S, et al. (2009) Control of the blood-brain barrier by glucocorticoids and the cells
of the neurovascular unit. Ann N Y Acad Sci 1165:228-239.
123.
Limbourg FP, et al. (2002) Rapid nontranscriptional activation of endothelial nitric oxide
synthase mediates increased cerebral blood flow and stroke protection by corticosteroids.
J Clin Invest 110:1729-1738.
124.
Rau SW, Dubal DB, Bottner M, Gerhold LM, & Wise PM (2003) Estradiol attenuates
programmed cell death after stroke-like injury. J Neurosci 23:11420-11426.
125.
Miller AH (2007) Letter to the Editor Re: "an inflammatory review of glucocorticoids in
the CNS" by Sorrells et al. Brain, Behavior and Immunity 21, 259-272, 2007. Brain
Behav Immun 21:988-989; author reply 990.
126.
Sorrells SF & Sapolsky RM (2007) An inflammatory review of glucocorticoid actions in
the CNS. Brain Behav Immun 21:259-272.
127.
Sapolsky RM, Krey LC, & McEwen BS (1986) The neuroendocrinology of stress and
aging: the glucocorticoid cascade hypothesis. Endocr Rev 7:284-301.
128.
Rhen T & Cidlowski JA (2005) Antiinflammatory action of glucocorticoids--new
mechanisms for old drugs. N Engl J Med 353:1711-1723.
129.
Raison CL & Miller AH (2003) When not enough is too much: the role of insufficient
glucocorticoid signaling in the pathophysiology of stress-related disorders. Am J
Psychiatry 160:1554-1565.
130.
Perry VH (2007) Stress primes microglia to the presence of systemic inflammation:
implications for environmental influences on the brain. Brain Behav Immun 21:45-46.
131.
Stein-Behrens BA & Sapolsky RM (1992) Stress, glucocorticoids, and aging. Aging
(Milano) 4:197-210.
106
132.
Howard KJ & Distelhorst CW (1988) Evidence for intracellular association of the
glucocorticoid receptor with the 90-kDa heat shock protein. J Biol Chem 263:3474-3481.
133.
Frank MG, Baratta MV, Sprunger DB, Watkins LR, & Maier SF (2006) Microglia serve
as a neuroimmune substrate for stress-induced potentiation of CNS pro-inflammatory
cytokine responses. Brain Behav Immun.
134.
Frank MG, Miguel ZD, Watkins LR, & Maier SF (2009) Prior exposure to
glucocorticoids sensitizes the neuroinflammatory and peripheral inflammatory responses
to E. coli lipopolysaccharide. Brain Behav Immun 24:19-30.
135.
Johnson JD, O'Connor KA, Watkins LR, & Maier SF (2004) The role of IL-1beta in
stress-induced sensitization of proinflammatory cytokine and corticosterone responses.
Neuroscience 127:569-577.
136.
Glezer I & Rivest S (2004) Glucocorticoids: protectors of the brain during innate immune
responses. Neuroscientist 10:538-552.
137.
Bruccoleri A, Pennypacker KR, & Harry GJ (1999) Effect of dexamethasone on elevated
cytokine mRNA levels in chemical-induced hippocampal injury. J Neurosci Res
57:916-926.
138.
Roozendaal B (2000) 1999 Curt P. Richter award. Glucocorticoids and the regulation of
memory consolidation. Psychoneuroendocrinology 25:213-238.
139.
Nicholas A, Munhoz CD, Ferguson D, Campbell L, & Sapolsky R (2006) Enhancing
cognition after stress with gene therapy. J Neurosci 26:11637-11643.
140.
Braun S, Liebetrau W, Berning B, & Behl C (2000) Dexamethasone-enhanced sensitivity
of mouse hippocampal HT22 cells for oxidative stress is associated with the suppression
of nuclear factor-kappaB. Neurosci Lett 295:101-104.
141.
Johnson EA, O'Callaghan JP, & Miller DB (2002) Chronic treatment with
supraphysiological levels of corticosterone enhances D-MDMA-induced dopaminergic
neurotoxicity in the C57BL/6J female mouse. Brain Res 933:130-138.
107
142.
Limoges J, Persidsky Y, Bock P, & Gendelman HE (1997) Dexamethasone therapy
worsens the neuropathology of human immunodeficiency virus type 1 encephalitis in
SCID mice. J Infect Dis 175:1368-1381.
143.
Roberts AJ, Crabbe JC, & Keith LD (1993) Type I corticosteroid receptors modulate
PTZ-induced convulsions of withdrawal seizure prone mice. Brain Res 626:143-148.
144.
Fishman RA (1982) Steroids in the treatment of brain edema. N Engl J Med 306:359-360.
145.
Roberts I, et al. (2004) Effect of intravenous corticosteroids on death within 14 days in
10008 adults with clinically significant head injury (MRC CRASH trial): randomised
placebo-controlled trial. Lancet 364:1321-1328.
146.
Aisen PS, et al. (2000) A randomized controlled trial of prednisone in Alzheimer's
disease. Alzheimer's Disease Cooperative Study. Neurology 54:588-593.
147.
Yeager MP, et al. (2009) Pretreatment with stress cortisol enhances the human systemic
inflammatory response to bacterial endotoxin. Crit Care Med 37:2727-2732.
148.
Koo JW, Russo SJ, Ferguson D, Nestler EJ, & Duman RS (2010) Nuclear factor-kappaB
is a critical mediator of stress-impaired neurogenesis and depressive behavior. Proc Natl
Acad Sci U S A 107:2669-2674.
149.
Waters JC (2009) Accuracy and precision in quantitative fluorescence microscopy. J Cell
Biol 185:1135-1148.
108