Background
Sickness refers to a set of physiological and behavioral responses (e.g., fever, anorexia, immobility, reduced exploratory activity, social withdrawal, anhedonia) to systemic inflammation [
1‐
3]. Collectively, these changes serve as the body’s adaptive strategies to combat infections and injuries [
4]. However, increasing lines of evidence have indicated that the same responses can become deleterious if exacerbated [
1,
5‐
7]. Therefore, it will be beneficial to understand how sickness is regulated.
The regulation of sickness is a complex subject. Upon an inflammatory insult, innate immune cells sense pathogen-associated molecular patterns (PAMPs) and alarmins that are respectively released from pathogens and damaged tissues [
8,
9], and respond by upregulating cytokines [
10,
11], prostaglandins [
12,
13], and complement factors [
14,
15]. These systemic inflammatory mediators communicate to the brain via multiple humoral and neural routes, causing neuroimmune activation and sickness [
16‐
20]. Accordingly, inflammation is important for sickness development, and a reduction of inflammation should help to suppress sickness. Indeed, both peripheral [
10,
17,
21,
22] and central [
23‐
26] administration of exogenous inflammatory molecules acutely triggers sickness. On the contrary, pharmacological inhibition of the synthesis [
27‐
30] or the actions [
31‐
34] of endogenous inflammatory mediators can decrease sickness following immune challenge by lipopolysaccharide (LPS).
Protein kinase R (PKR) is a ubiquitously expressed serine-threonine kinase that was originally discovered as an antiviral defense mediator [
35,
36]. Upon viral infection, PKR binds to double-stranded RNA (dsRNA) from viruses through its N-terminal dsRNA-binding domain, resulting in PKR dimerization and autophosphorylation and activation of the C-terminal kinase domain [
37‐
39]. Activated PKR can inhibit viral infection by phosphorylating eukaryotic initiation factor 2 (eIF2α) to decrease protein translation [
40‐
42] and/or by inducing apoptosis of infected cells [
43,
44].
Apart from its traditional antiviral roles, our group is more interested in the role of PKR in regulating inflammation. For instance, PKR can be activated during inflammation, as illustrated by an increase of PKR phosphorylation in macrophages following CD40 ligation [
45] and when stimulated by toll-like receptor (TLR) ligands [
46]. Furthermore, PKR shows crosstalk with inflammatory pathways such as c-Jun-N-terminal kinase (JNK) [
46,
47], mitogen-activated protein kinase (MAPK) [
48], nuclear factor-kappa B (NF-κB) [
47,
49], signal transducers and activators of transcription 1 (STAT1) [
50], interferon-regulatory factor-1 (IRF-1) [
49], and inflammasome [
51]. It can affect cytokine production and release from cultured fibroblasts [
47,
48], macrophages [
46,
51], and mixed glia/neuron co-cultures [
52]. On the other hand, genetic deletion of PKR in mice attenuates plasma IL-6 and IL-12 increases triggered by LPS [
48]. Until recently, PKR has also been shown to control neuroinflammatory changes in animal models of viral encephalitis [
53] and excitotoxic injury [
54].
Since systemic inflammation triggers neuroimmune activation and sickness, and that PKR modulates inflammation, the purpose of this study is to investigate whether PKR can affect neuroimmune responses and sickness induced by systemic inflammatory challenge. Wild-type (WT) and PKR−/− mice were subcutaneously injected with live
Escherichia coli (
E. coli) or vehicle. As predicted, deficiency of PKR diminished peripheral inflammatory responses to
E. coli. However, to our surprise, the loss of PKR did not decrease sickness. Instead, PKR−/− mice displayed several behavioral components of sickness (reduced burrowing, exploratory deficits, and social withdrawal) that were not observed in WT mice. Moreover, these altered sickness behaviors were unlikely to be caused by exaggerated neuroimmune activation or increased bacterial load, because both strains of mice showed similar neuroimmune responses and bacterial titers throughout the course of infection. As systemic inflammation can activate the hypothalamic-pituitary-adrenal (HPA) axis which can potentially regulate sickness [
55‐
60], we further asked whether PKR modulates HPA axis activation following
E. coli challenge. Both strains of mice exhibited similar changes in plasma corticosterone levels post-infection. Nevertheless, PKR−/− mice displayed a delayed induction of corticotrophin-releasing hormone (CRH) in the hypothalamus as compared to WT mice, suggesting that the loss of PKR may postpone the CRH response to systemic inflammation. Taken together, our findings show that (1) deficiency of PKR could alter
E. coli-induced sickness behaviors and (2) this was not because of exacerbated neuroimmune activation, (3) increased bacterial load, or (4) dysregulation of the corticosterone response. However, knockout of PKR could delay the CRH response to systemic inflammation, and this may possibly affect sickness.
Discussion
Systemic inflammation triggers neuroimmune activation, leading to sickness. In this study, we investigated the role of PKR, a serine-threonine kinase that is activated by immune challenge [
45,
46] and regulates LPS-induced peripheral inflammatory responses [
48], on neuroimmune responses and sickness following subcutaneous
E. coli infection. While genetic deficiency of PKR in mice did not affect the core components of sickness (anorexia and motor impairments), it led to several behavioral components of sickness (decreased burrowing, exploratory deficits, and social withdrawal) that were not observed in WT mice (Fig.
1). Interestingly, such alteration in the behavioral components was not due to exacerbated inflammation, since the loss of PKR diminished peripheral inflammatory changes (Figs.
2 and
3), and it had minimal effect on the associated neuroimmune responses (Figs.
4 and
5). Likewise, bacterial titers (Fig.
6) and plasma corticosterone profiles (Fig.
7b) did not differ between WT and PKR−/− mice during the course of infection. Hence, the altered behavioral components in PKR−/− mice also did not result from an impaired host defense to the
E. coli infection or from a dysregulated corticosterone response. However, PKR−/− mice displayed a delayed induction of CRH after
E. coli challenge (Fig.
7), suggesting a postponed CRH response in these mice that may possibly modulate sickness.
As expected, genetic deletion of PKR suppressed peripheral inflammatory responses to
E. coli infection. Previous studies have demonstrated that deficiency of PKR attenuates plasma IL-6 and IL-12 increases after systemic LPS challenge [
48] and that PKR regulates inflammasome activation [
51] and TLR2/TLR4-dependent cytokine release [
46] from cultured macrophages. Moreover, PKR shows crosstalk with multiple inflammatory pathways including NF-κB [
47,
49], MAPK [
48], IRF-1 [
49], and JNK signaling cascades [
46,
47]. Our findings are therefore in agreement with these earlier reports. Since systemic inflammatory mediators can communicate with the brain by neural and humoral routes to trigger neuroimmune activation [
3,
19], we asked whether PKR deficiency could also suppress neuroimmune responses after
E. coli challenge.
To our surprise, WT and PKR−/− mice displayed mostly similar neuroimmune changes. This is in contrast to several earlier studies, which showed that pharmacological inhibitors of PKR can abolish inflammatory responses in glial cultures [
50,
52,
80]. Until recently, it has also been demonstrated that genetic deletion or pharmacological inhibition of PKR reduces neuroinflammation in animal models of viral encephalomyelitis [
53] and excitotoxic injury [
54]. It should be noted that in these studies, inflammatory changes were initiated directly in glial cultures or by direct brain insults, whereas in our scenario neuroimmune activation occurred subsequently to subcutaneous inflammation caused by
E. coli. This notable deviation suggests that PKR in the brain and PKR at peripheral tissues may act differentially to affect neuroimmune responses under different causes. It is known that systemic immune challenge can promote the recruitment of neutrophils [
81‐
84] and monocytes [
85,
86] into the brain and that infiltrating leukocytes can influence inflammatory changes in the brain [
84,
87,
88] and even LPS-induced depression-like behavior [
84]. For instance, a single injection of LPS dose dependently elevates the number of infiltrating neutrophils into the facial nucleus within 48 h, a phenomenon that can persist up to at least 96 h [
83]. Given that
E. coli challenge led to a sustained infection along with increased inflammatory factors up to 120 h, it is expected to have a similar effect as a high dose of LPS to cause considerable leukocyte infiltration into the brain. One possibility could be that PKR may modulate this leukocyte infiltration process. Should this be the case, even if deficiency of PKR can downregulate the neuroimmune changes mediated by glial cells, this effect may be masked by the inflammatory responses elicited by the infiltrating leukocytes. Indeed, PKR−/− mice display enhanced T cell recruitment into the brain during viral encephalomyelitis [
53], suggesting that PKR may also regulate the entry of other leukocyte cell types into the brain. Further investigation along this direction may provide clues as to how PKR controls neuroimmune activation during systemic immune insults.
Irrespective to the cause for the similar neuroimmune responses, PKR−/− mice did show decreased peripheral inflammatory changes. This finding led us to predict that sickness might also be diminished in PKR−/− mice, since sickness can be ameliorated by blockade of peripheral cytokine synthesis or their effects [
27,
29,
32,
89], or by genetic deletion of NF-κB [
90]. Therefore, we monitored both core (anorexia, motor impairments) and behavioral (burrowing and exploratory deficits, social withdrawal) components of sickness [
1,
5,
19,
71,
91] in WT and PKR−/− mice after
E. coli challenge. Previous literature has shown that these two types of sickness components can be temporally [
11,
92‐
94] and pharmacologically [
27,
29,
93,
94] dissociated, indicating that the two types of sickness components involve different regulatory mechanisms. For example, LPS-induced deficit in burrowing can persist up to 24 h, when the decrease in locomotor activity is no longer observed [
11]. On the other hand, while administration of cyclooxygenase (COX) inhibitors can ameliorate both core (hypothermia, impaired locomotor activity) and behavioral (decreased burrowing) components of sickness following LPS challenge, blockade of peripheral cytokine synthesis by dexamethasone or dexamethasone-21-phosphate only attenuates LPS-mediated hypothermia [
27,
29]. Here, our data demonstrate a rather unexpected finding. Knockout of PKR in mice did not elicit any reduction in sickness responses. Instead, the core components of sickness were similar between both strains of mice, and the behavioral components of sickness were only observable in PKR−/− mice. Of particular interest is that PKR−/− mice showed decreased activities in the open field test, object investigation test, and social interaction test even at 120 h, although these changes were absent in WT mice at the same time point. It should be noted that these exploratory and social deficits observed only in PKR−/− mice were not simply because of a general decrease in motor activity, because rotarod performance had been restored to the control level by 96 h. Thus, this is an indication that deficiency of PKR primarily affected the behavioral components of sickness but not the core components. We did not perform these tasks at an earlier time because we had specifically wanted to distinguish the roles of PKR on the core and behavioral components of sickness. Moreover, both strains of mice displayed decreased rotarod performance at 24 h, indicating that the use of these behavioral assays at or before this time would not allow us to separate the effects of the core and behavioral components of sickness. Given that LPS is well reported to abolish exploratory activity [
10,
95] and induce social withdrawal [
96‐
98] in WT animals within several hours after systemic administration, it is likely that
E. coli can also lead to these deficits in the two strains of mice under this time frame.
Interestingly, we did not detect any change in the burrowing activity of WT mice after
E. coli. This would seem contradictory to earlier reports which demonstrated that LPS acutely decreases burrowing in WT animals [
11,
27,
28]. However, it should be emphasized that these studies have used food pellets as the burrowing substrate, whereas here we used bedding material (wood chips) to avoid interference effects on the measurements of food consumption. It has been documented that different types of burrowing substrates can greatly influence the burrowing activity measured in rodents [
99,
100]. C57BL/6 mice typically burrow a greater proportion of the burrowing substrate when the burrowing tube is filled with bedding material than when it is filled with food pellets. For example, Kir6.2 knockout mice display severely impaired burrowing as compared to WT mice when food pellets are used, but this difference becomes less obvious when the burrowing substrate is replaced with bedding material [
99]. Based on these studies, it could be deduced that bedding material is more preferred than food pellets as a burrowing substrate by C57BL/6 mice, such that it will be harder to detect phenotypic differences with bedding material than with food pellets. Perhaps WT mice would have displayed burrowing deficits after
E. coli challenge if the burrowing tubes had been filled with food pellets.
Next, we asked whether the alterations in the behavioral components of sickness of PKR−/− mice would be correlated with exaggerated bacterial load. It has been well reported that loss of PKR can often enhance viral replication [
41,
44,
101,
102]. Another study has shown that PKR is involved in resistance to
Toxoplasma gondii [
45]. To our knowledge, however, the effect of PKR in bacterial infection has been little investigated [
51]. Here, we made use of two different methods, i.e., microbiological plate count and quantification of 16S rDNA by real-time PCR, and show that bacterial titers were similar between the two strains of mice throughout the course of infection. These results imply that altered sickness in PKR−/− mice was not due to an impaired host defense to
E. coli. In fact, even if PKR deficiency does affect
E. coli infection, it would likely suppress it. This point is supported by another study by Lu et al., in which PKR−/− mice had reduced bacterial titers following
E. coli-induced peritonitis [
51].
We then investigated if PKR could modulate the CRH response to
E. coli challenge. Previous studies have indicated that systemic inflammation leads to CRH induction and release from paraventricular nucleus (PVN) neurons of the hypothalamus [
56,
77]. Furthermore, administration of CRH into rodents acutely reduces exploratory activities to novel environments [
103] and to novel individuals [
104]. These behaviors are quite similar to what was observed in PKR−/− mice at 120 h after
E. coli challenge. Our data indicate that while
E. coli upregulated CRH expression at 4 h in the hypothalamus of WT mice, such increase was not found in PKR−/− mice. This finding is not surprising, given that systemic inflammation induces CRH, and PKR−/− mice showed reduced peripheral inflammation. However, at 48 h after
E. coli infection, there was a significant elevation of CRH in PKR−/− mice but not in WT mice. Hence, it appears that PKR deficiency can delay CRH induction in response to
E. coli. It is also possible that knockout of PKR may extend the CRH response period, although this requires validation in a more detailed temporal manner. The altered CRH response in PKR−/− mice suggest that these CRH-mediated effects may also be postponed or extended, which can possibly alter the behavioral components of sickness in PKR−/− mice. Indeed, chronic administration of CRH into the brain can delay behavioral inhibition induced by LPS [
105], thus providing indirect evidence to support this possibility. Future studies should address the role of PKR on the CRH response along with its implication on sickness.
We have known that CRH participates in the HPA axis to stimulate adrenocorticotrophic hormone (ACTH) synthesis and release from the anterior pituitary [
77]. ACTH in turn acts on the adrenal cortex, inducing corticosterone production and secretion. Upon systemic LPS challenge, corticosterone is increased in the circulation, and several reports have also demonstrated that endogenous corticosterone can suppress peripheral inflammation and sickness [
55,
59,
60]. As PKR−/− mice exhibited a delayed CRH induction, we questioned whether this would result in differential responses of corticosterone to
E. coli. Surprisingly, there was no significant change in the levels of plasma corticosterone between WT and PKR−/− mice. Plasma corticosterone levels were elevated by
E. coli to the same extent in WT and PKR−/− mice at 4 h, and by 48 h these increases had disappeared in the
E. coli-challenged groups. These results suggest that the altered sickness behaviors in PKR−/− mice were not due to a dysregulated corticosterone response. Furthermore, corticosterone profiles in the two strains of mice did not follow the same trend as that of hypothalamic CRH. While CRH induction was blunted at 4 h in PKR−/− mice relative to WT mice, corticosterone was similarly increased in both genotypes. At 48 h, CRH was induced in PKR−/− mice, but plasma corticosterone was not upregulated. Such a discrepancy between the CRH and corticosterone profiles might be due to the following possibilities. Firstly, CRH production is not limited to the hypothalamus. Extra-hypothalamic sources of CRH [
106‐
109] may likely serve as alternative source(s) of CRH in PKR−/− mice, such that hypothalamic CRH production would not be required at 4 h in these mice. Of particular relevance is that CRH is highly expressed at peripheral inflammatory tissues to regulate local immune responses [
109]. If deficiency of PKR can hyper-induce CRH at inflammatory sites during the
E. coli infection, such CRH may potentially spill over into the bloodstream and upregulate corticosterone. Secondly, while corticosterone production is regulated by CRH, it can also be triggered by vasopressin through the type 1b vasopressin receptor in the anterior pituitary [
110]. Perhaps PKR can act at the level of vasopressin system, thereby exerting its control over the level of corticosterone. Thirdly, elevated corticosterone has been shown to decrease CRH-R1 mRNA and CRH binding [
111]. Since corticosterone was increased before 48 h in PKR−/− mice, this could explain why these mice were unable to mount another wave of corticosterone response even though CRH was induced at 48 h.
It is noteworthy that despite the dissociation between hypothalamic CRH and circulating corticosterone profiles in PKR−/− mice, the possible involvement of the delayed CRH response in altering sickness behaviors should not be simply ruled out. CRH participates in the HPA axis, but CRH effects are not entirely mediated by ACTH or corticosterone. CRH is a neurotransmitter and it can bind to widely distributed receptors CRH-R1 and CRH-R2 in the brain [
112]. Indeed, many of the brain regions that express CRH receptors, including the hippocampus, amygdala, hypothalamus, midbrain, and cerebral cortex, are not responsible for ACTH production. Instead, direct manipulation of CRH signaling in these brain structures often results in behavioral changes that are also observed in sickness. For instance, specific deletion of CRH-R1 in midbrain dopaminergic neurons causes anxiety and inhibits dopamine release in the prefrontal cortex [
113]. In the same study, it was shown that deletion of CRH-R1 in forebrain glutamatergic neurons decreases anxiety and neurotransmission in the hippocampus and amygdala. Moreover, the direct infusion of CRH into the hippocampus enhances long-term potentiation and improves context-dependent fear conditioning [
114]. Interestingly, many of the brain structures that express CRH receptors are also involved in sickness [
20], thus supporting for a role of CRH in modulating sickness. A delay in the CRH response in PKR−/− mice suggests that endogenous CRH signaling at an early time after the
E. coli infection is required for normal sickness development.
In addition to the delayed CRH response, other parameters can be studied in future to better understand sickness in PKR−/− mice. For example, many inflammatory mediators such as NF-κB [
90], microsomal PGE synthase-1 (mPGES-1) [
115,
116], and transcription factor nuclear factor interleukin 6 (NF-IL-6) [
117] have been demonstrated to control sickness and are potential candidates to alter sickness behaviors in PKR−/− mice. In particular, PKR modulates NF-κB activation [
47,
49], and genetic deletion of NF-κB abolishes sickness induced by LPS and unmethylated cytosine-phosphate-guanosine motifs (CpG-DNA) [
90]. We have focused on quantifying inflammatory gene expression in the brain. This was because inflammatory gene expression serves as a good indicator of local neuroimmune activation and that it could enable us to assess multiple inflammatory markers with a limited amount of sample material. Subsequent investigation of neuroimmune markers at protein level would be important to further characterize neuroimmune responses and their relationships to sickness. For example, iNOS mRNA expression was increased to a slightly lesser degree in the PKR−/−
E. coli group than in the WT
E. coli group at 48 h in the hypothalamus. It is known that systemic LPS upregulates iNOS mRNA expression in the hypothalamus [
118] and that pharmacological inhibition of iNOS attenuates LPS-induced sickness responses [
34]. Our mRNA results raise questions as to whether
E. coli could indeed upregulate iNOS protein in the hypothalamus and if PKR could modulate iNOS protein production that can potentially affect sickness. Finally, systemic inflammation can activate the indoleamine-2,3-dioxygenase (IDO) pathway, leading to the catabolism of tryptophan (TRP) to kynurenine (KYN) in the brain and blood [
94]. In the same study, pharmacological inhibition of the IDO pathway was shown to alleviate LPS-induced depressive-like behavior. In another study, genetic deletion or pharmacological inhibition of IDO ameliorates depressive-like behaviors triggered by an intracerebroventricular injection of LPS [
119]. Since PKR−/− mice developed exploratory activity and social interaction deficits up to 120 h, a dysregulation of the IDO pathway is another reasonable possibility that may account for these altered sickness behaviors.
A major limitation of our study is that we have used a general knockout approach to model the effect of PKR on neuroimmune activation and sickness. Since PKR is ubiquitously expressed, it is difficult to pinpoint tissue-specific effects of PKR during systemic inflammation. For instance, we discussed the possibility that PKR deletion at peripheral tissues may enhance leukocyte recruitment into the brain, thereby masking the reduction of neuroimmune responses in glial cells. This issue can be more easily studied if PKR conditional knockout mice were used. Perhaps, it will be beneficial to generate different lines of mice having specific knockout of PKR in target tissues, so as to better characterize the role of PKR during systemic inflammation.
We believe that our findings have important implications. Firstly, we identify a novel role of PKR in regulating sickness, particularly the behavioral components of sickness. As an over-exaggeration of sickness may precipitate depression [
5,
19] and/or delirium [
3,
6], gaining a better understanding on sickness regulation can shed light on how to fine tune sickness responses. Ideally, one would want to preserve the physiological functions of sickness without causing severe side effects. Secondly, reports in the recent decade have demonstrated that PKR inhibition can be neuroprotective [
119‐
122] and improve cognition [
123]. Here, we provide additional information on PKR deficiency during systemic inflammation, and this can complement the earlier studies. While we did show that loss of PKR altered the behavioral components of sickness, we should not simply neglect the desirable effects of PKR inhibition in neurodegenerative processes and memory. Instead, the efficacy of PKR pharmacological inhibitors should be justified after considering both positive and negative effects and taking into account of the patient’s disease status.