Background
By mechanisms that remain to be fully established, systemic infection or inflammation can have a profound effect on the central nervous system (CNS), manifesting in a number of behavioural adaptations, as well as fever and increased neuroendocrine activation. Promoting energy conservation and minimizing heat loss, these sickness behaviours represent a sound strategy designed to help an organism overcome infection. Symptoms include fever, malaise, anorexia, lethargy and, in severe cases, neuropsychiatric disorders, such as depression and anxiety [
1]. Sickness behaviours occur during acute bacterial or viral infections, but also during chronic inflammatory diseases, such as rheumatoid arthritis, inflammatory bowel disease and psoriasis [
2‐
4]. In the case of the latter, what would be a beneficial, self-limiting, system can become dysregulated. The prolonged depression and anxiety that ensues represents a major burden to patients, not least because these detrimental comorbidities lead to a poorer clinical outcome.
It is becoming increasingly evident that sickness behaviours are triggered as a result of biological, inflammatory, pathways. In particular, inflammatory cytokines, such as tumour necrosis factor α (TNFα), interleukin-1 β (IL-1β) and interleukin-6 (IL-6) play a pivotal role in inducing symptoms of sickness behaviour. Although not a prerequisite for any of the psychiatric symptoms, IL-6 is required for the induction of a fever response [
5], whereas behavioural changes are thought to be attributable to IL-1β and TNFα. Using animal models, it has been demonstrated that most behavioural symptoms can be induced by peripheral, or central, administration of either IL-1β or TNFα [
6,
7]. Furthermore, several cytokines have been implicated in the manifestation of major depressive disorders in patients with chronic inflammatory diseases. For example, in a phase III clinical trial in which patients with moderate-to-severe psoriasis were treated with the soluble TNFα receptor etanercept, improvement in depression scores preceded the improvements seen in terms of psoriasis severity [
4]. Supporting the notion of cytokine-induced depression, patients receiving interferon (IFN) α or IFNβ therapy face a risk of experiencing depression as a side effect of treatment [
8‐
12]. Moreover, patients with major depressive disorders, and no clinical signs of inflammation, often present with elevated levels of circulating inflammatory cytokines [
13]. Therefore, significant quantities of literature back the immune system, in particular inflammatory cytokine production, as a key contributor to sickness-induced behavioural changes.
Once considered an immunologically privileged site, the CNS is well fortified against changes in the periphery. However, cross talk does occur and, as a result, much research has gone into elucidating putative routes of immune-to-brain communication. In spite of this, the precise effect of peripheral immune activation on neural circuitry remains unclear. With the aim of better unravelling neuroimmune communication pathways, and the downstream consequences of peripheral inflammation on the brain, we compared gene expression in the brains of mice following several different sterile (tumour necrosis factor (TNF) and 12-O-tetradecanoylphorbol-13-acetate (TPA) and toll-like-receptor-(TLR)-dependent (lipopolysaccharide (LPS), lipoteichoic acid (LTA) and Imiquimod) models of peripheral inflammation. We also compared gene expression in the brain with that of peripheral blood leucocytes (PBLs). Lastly, we explored a potential molecular mechanism by which inflammation originating in the periphery can induce transcriptional modulation in the brain.
Methods
Mice
Wild type C57BL/6 mice (7 to 8 weeks old, 20 to 25 g) were purchased from Harlan Laboratories. Mice were maintained in specific pathogen-free conditions in standard caging in the Central Research Facility at the University of Glasgow and treated with sterile (TNF and TPA) and TLR ligand-based (LPS, LTA and Imiquimod) inflammatory agents as described. For microarray experiments, a minimum of three biological replicates are required to allow statistical analysis of the data [
14]. Three mice were used per study arm for all our microarray experiments and four or five mice each for qPCR-based experiments, to ensure the statistical robustness of the data. All experiments received ethical approval and were performed under the auspices of UK Home Office Licences.
Acute inflammatory models
For acute LPS-induced inflammation, mice were injected intraperitoneally (i.p.) with 100 μl of 1 mg/ml LPS (≈4 mg/kg), derived from Escherichia coli serotype 055:B5 (Sigma, St. Louis, MO, USA, or an equivalent volume of vehicle (PBS). For TNFα- or LTA-induced inflammation, mice were injected intravenously (i.v.) with two doses of 1 μg recombinant TNFα (Peprotech, Rocky Hill, NJ, USA), two doses of 500 μg LTA (Sigma, St. Louis, MO, USA) or two doses of an equivalent volume (100 μl) of vehicle (sterile H2O) at 0 and 24 hours. Mice were euthanized by CO2 exposure 48 hours after initial injection and perfused for 5 minutes with 20 ml PBS.
Chronic inflammatory models
For chronic LPS-induced inflammation, and induction of endotoxin tolerance, mice received a daily i.p. injection of 100 μl of 0.5 mg/ml LPS (≈2 mg/kg) (Sigma, St. Louis, MO, USA) or an equivalent volume (100 μl) of vehicle (PBS) for 2, 5 or 7 consecutive days. For skin-inflammation models, mice were shaved on their dorsal skin 24 hours prior to receiving daily applications of ≈ 80 mg of 5% Imiquimod (Aldara™, MEDA Ab, Stockholm, Sweden) cream [
15], 150 μl of 100 μM TPA, or an equivalent volume of Vaseline (Unilever, Leatherhead, UK) or acetone control. Mice were treated for 5 consecutive days as described previously [
15]. All mice were euthanized by CO
2 exposure 24 hours after final treatment and perfused for 5 minutes with 20 ml PBS.
ELISA
Blood was collected from tail veins (approximately 300 μl) prior to termination of the mice. Plasma was isolated from whole blood by centrifugation. Throughout the study, plasma concentrations of soluble mediators, IL-1β, TNFα and IL-6, were determined using DuoSet ELISA kits (R&D Systems, Minneapolis, MN, USA) according to the manufacturer’s instructions.
RNA isolation from tissue and peripheral blood leucocytes
Whole brain tissue was snap frozen and stored at -80°C until use. Under RNase-free conditions, brains were homogenized using the TissueLyser LT (Qiagen, Hilden, Germany). RNA was extracted from homogenized tissue using Trizol® (Life Technologies, Invitrogen, Carlsbad, CA, USA) as described by the manufacturers. Isolated RNA was further purified and genomic DNA removed using an RNeasy Mini Kit (Qiagen, Hilden, Germany). Red blood cells were lysed from blood samples using red blood cell lysis buffer (Miltenyi, Cologne, Germany). Under RNase-free conditions, RNA was isolated and genomic DNA was removed from PBLs using an RNeasy Micro Kit (Qiagen, Hilden, Germany).
GeneChip microarray analysis
Microarray assays were performed in the Glasgow Polyomics Facility at the University of Glasgow [
16]. Briefly, 1 μg of purified total RNA was amplified by
in-vitro transcription and converted to sense-strand cDNA using a WT Expression kit (Life Technologies, Invitrogen, Carlsbad, CA, USA). cDNA was then fragmented and labelled using a GeneChip WT Terminal Labelling kit (Affymetrix, Santa Clara, CA, USA). Fragmented cDNA samples were then hybridized to GeneChip Mouse Gene 1.0 ST Arrays (Affymetrix, Santa Clara, CA, USA). Procedures were carried out as described by the manufacturers.
To maximize the identification of key differentially expressed genes we utilized two separate software analysis packages (Partek and GeneSpring) and focused on gene expression differences identified using both approaches. As shown in the results section, this reduced the number of genes requiring analysis and provided increased confidence in their validity.
Data generated using Partek Genomics Suite were normalized using the robust multichip average (RMA) method, adjusted for GC content. The normalized data were subsequently analyzed using one-way analysis of variance (ANOVA) to determine the significance of each gene in LPS-treated mice compared with vehicle-treated controls. Data generated using GeneSpring GX software were normalized using RMA 16. Normalized data were analyzed using unpaired t tests to determine the significance of gene expression differences in LPS-treated mice compared with vehicle-treated controls. In both analyses, P values were adjusted for multiple comparisons using the Benjamini-Hochberg multiple testing correction.
Gene ontology terms were assigned to differentially expressed genes using the Database for Annotation, Visualization and Integrated Discovery (DAVID) Bioinformatics Resources v6.7 [
17]. Analysis was performed in accordance with two protocols outlined by Huang
et al.[
18,
19]. Significance of enrichment was determined using a modified Fisher’s exact test and a Benjamini-Hochberg multiple testing correction was used to correct for the rate of type I errors. Co-expression of a gene cluster was considered significant for
P ≤ 0.05.
Genes were grouped into canonical pathways using Ingenuity Pathway Analysis software (Ingenuity® Systems [
20]). Significance of differentially altered pathways was determined using a Fisher’s exact test and a Benjamini-Hochberg multiple testing correction was used to correct for the rate of type I errors. Enrichment of a pathway was considered significant for
P ≤ 0.05.
qPCR
Total RNA was reverse transcribed using Quantitect® Reverse Transcription kit (Qiagen, Hilden, Germany) using random primers. Quantitative real-time PCR (QPCR) amplifications were performed in triplicate using PerfeCTa® SYBR® Green FastMix® (Quanta Biosystems, Gaithersburg, MD, USA). A 500 μM mix of forward and reverse primers was used per reaction. Primers were designed using Primer3 Input software (version 0.4.0) and generated by IDT technologies. Primer sequences are listed in Additional file
1: Table S1. qPCR reactions were performed using a Prism® 7500HT Sequence Detection System (Life Technologies, Invitrogen, CA, USA) for 40 cycles, in accordance with the manufacturer’s guidelines. The absolute copy number was calculated from a standard curve and normalized to a reference gene, TATA binding protein (TBP), as previously described [
21]. Fold change values were calculated by comparing the normalized copy number of individual samples with the mean of the control samples.
Histology
Skin samples were fixed in 10% buffered formalin prior to processing and paraffin embedding. Processing was performed using the Shandon Citadel 1000 automated tissue processor (Thermo scientific). Embedded tissue was then cut into 5 μm sections and stained with (H & E) to discern morphology.
Discussion
Here we demonstrate that systemic administration of LPS or topical administration of IMQ alters the gene expression profile in the brain, inducing the expression of a common panel of ISGs. Differing in kinetics and magnitude, this response is distinct from that of the PBLs. We also highlight ISG induction as a consequence of a TLR-induced type I IFN response. As considerable evidence links type I IFNs to psychiatric disorders [
8‐
12], IFN production in the brain may represent a significant mechanism, linking peripheral TLR-induced inflammation with neuropsychiatric symptoms.
The differential response of brain and PBLs provides evidence of a brain-specific inflammatory response resulting from both acute and chronic LPS challenge and IMQ-induced skin inflammation. This was not a downstream by-product of peripheral inflammatory cytokine production as similar sterile models of inflammation failed to induce the same response. Furthermore, ISG expression remained elevated in the brain following daily LPS or IMQ administration, long after the peripheral cytokine response was attenuated. Thus, ISG induction was brain-specific and not mediated by peripheral inflammatory cytokines in either the skin or the circulation.
As type I IFN production is a classic hallmark of both TLR4-induced IRF3 activation and TLR7-induced IRF7 activation, it would be attractive to propose that a TLR-induced IFN response is responsible for the induction of ISGs in the brain. Supporting this hypothesis, Skelly and colleagues documented a central induction of IFNβ within 2 hours of systemic LPS challenge [
33]. No central or peripheral type I IFN induction was observed following systemic TNFα or IL-1β injection. Furthermore, Leung and colleagues demonstrated that, following systemic administration of a TLR7 agonist, IFNα was induced in the brain in an IRF7-dependent manner [
34]. Both TLR4 and TLR7 are widely expressed in the brain. Although, to our knowledge, the capacity of IMQ to cross the blood–brain barrier (BBB) has not been investigated, a study with radiolabelled LPS suggested that negligible levels cross the intact BBB [
35]. Therefore, it is possible that the upregulation of target ISGs, following peripheral LPS or IMQ challenge, occurs via an indirect route downstream of IRF activation in the periphery. This may involve the direct action of peripherally produced IFNs on the CNS, the BBB or afferent nerves. Alternatively, introducing high or repeated doses of LPS or IMQ to the periphery may cause BBB breakdown, facilitating the direct action of these TLR ligands on the brain.
To further investigate the involvement of IRF activation in the neuroinflammation induced by systemic LPS injection, mice were challenged with the TLR2 ligand LTA. Consistent with previous reports [
36], no response was detectable in the brains of mice following peripheral LTA injection. This lack of response may be due to the inability of TLR2 ligands to stimulate IRF-dependent signalling. Conversely, peripheral stimulation with TLR3 ligands is known to trigger brain inflammation [
37,
38]. Like the MyD88-independent pathway downstream of TLR4, TLR3 signals through the adaptor molecule TRIF to activate IRF3, ultimately triggering IFNβ production [
31]. Subsequently, it would appear that IRF-dependent signalling, whether it occurs in the periphery or the brain, may be a requirement of ISG induction in the brain following systemic administration of TLR ligands.
As described, type I IFN therapy is intrinsically linked to severe neuropsychiatric disorders, mainly major depression [
8‐
12]. It is well known that injecting rodents with LPS initiates a number of behavioural adaptations, including a depression-like behaviour that perseveres after the other sickness behaviours have resolved [
1]. A recent report has also linked IMQ-treatment to the development of sickness behaviours in rats [
39]. The elevated transcription of type I ISGs in the brain following LPS or IMQ challenge is a strong indication that type I IFNs are produced during these models. As this family of cytokines are well known for their effects on behaviour, type I IFN production in either the periphery or in the brain, following peripheral LPS or IMQ challenge could contribute to the onset of depression-like behaviours in rodents. Formal demonstration of this will require the use of rodent behavioural models in conjunction with appropriate gene-deficient mice.
Conclusions
Toll-like receptor ligands have the capacity to modulate ISG expression distally in a manner that may be dependent on TLR-induced type I IFN production. Whether type I IFNs are produced in the brain or whether peripherally induced IFNs directly access the brain to modulate ISG expression remains open to further investigation; as does the downstream effects of central ISG induction. Owing to the well-established link between type I IFNs and depression, TLR-induced IFN production is worth investigating as a potential key mechanism, linking peripheral inflammation with sickness behaviour.
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
All data was collected, analyzed and interpreted by CT and AM. Experiments were conceived and designed by JC and GG. The manuscript was drafted by CT and all authors read and approved the final manuscript.