Background
Neuropsychiatric disorders, including depression, anxiety and schizophrenia, are a significant and growing burden to society [
1] and are also commonly associated with chronic inflammatory disorders [
2]. However, we know very little about the mechanisms underpinning this relationship.
Increasingly compelling data have highlighted possible roles for a number of immune mediators, including inflammatory cytokines, in the pathogenesis of neuropsychiatric disorders. Examples include (a) treatment of hepatitis C with IFN-α results in the onset of major depressive disorder (MDD) in around 50 % of patients and is the most common reason for ceasing treatment [
3‐
5], (b) the treatment of psoriasis patients with the anti-TNF drug etanercept improved the clinical symptoms of depression independently of an improvement in psoriatic disease score in patients [
6,
7] and (c) patients suffering from long-term MDD or schizophrenia have an elevated inflammatory profile [
8,
9]. In essence these pathology-, or therapy-, associated depressive episodes represent exaggerated presentations of the so-called ‘sickness behaviour’, which is a recognised and normal response typically associated with microbial infection. Nonetheless, despite these correlations, it remains unclear how inflammation in the periphery relates to neuropsychiatric phenotypes.
One family of cytokines that might play a key role in this communication pathway is the chemotactic cytokine, or chemokine, family. The classical role of chemokines is in mediating cell trafficking and inflammatory chemokines are important for cell migration during immune and inflammatory responses [
10]. Many chemokines, such as CXCL12, are expressed in the brain in steady-state and play an important role in development and homeostasis [
11]. However, some inflammatory chemokines are also expressed in the brain and are upregulated during infection [
12,
13]. We hypothesised that, in systemic inflammatory disease contexts, brain-expressed chemokines would drive immune cell infiltration into the brain, leading to the activation of local cells resulting in co-morbid psychiatric disorders [
14‐
16].
To test this hypothesis, we used a well-characterised mouse model of skin inflammation. The topical application of Aldara™, a commercially available drug which contains the Toll-like receptor (TLR) 7/8 ligand imiquimod (IMQ), to the dorsal skin of mice, causes a psoriasis-like skin pathology. The ease and reproducibility of this model made it ideal for our study [
17,
18]. Here, we show that peripheral skin inflammation can induce a ‘remote’ chemokine response in the brain that appears to be maintained independently of a response in the periphery. In addition, we show that this is associated with leukocyte infiltration into the brain parenchyma, a reduction in neural precursor cells in the dentate gyrus of the hippocampus and behavioural alterations reflected in a suppression of burrowing behaviour. These findings highlight the impact of peripheral immune stimulation on homeostatic CNS function and implicate chemokines and their receptors as potential therapeutic targets for chronic inflammatory disease associated neuropsychiatric co-morbidities.
Methods
Mice
Wild-type C57BL/6 female mice (6- to 8-week old, 16–20 g) were purchased from Harlan Laboratories. Mice were maintained in specific pathogen-free conditions in the Central Research Facility at the University of Glasgow. All experiments received ethical approval and were performed under the auspices of UK Home Office Licence.
Models of inflammation
Several models of peripheral inflammation were used in this study. These included the topical application of a cream containing the TLR7/8 ligand IMQ (Aldara model), the topical application of a sterile inflammatory agent (TPA), the topical application of IMQ in the absence of active components in the Aldara vehicle (Topical IMQ) and the intraperitoneal administration of IMQ (soluble IMQ). The four models are described in detail below. Since three out of the four models were used to induce cutaneous inflammation, this study focused on female mice, as it has been shown that male mice respond differently to topically applied inflammatory agents [
19].
(i)
Aldara model of skin inflammation
Mice were shaved on their dorsal skin 24 h prior to Aldara treatment. Eighty milligrams of 5 % imiquimod (Aldara™, MEDA Ab, Stockholm, Sweden) cream, or aqueous control cream, was applied to the shaved dorsal every 24 h for 1, 3 or 5 days. Mice were euthanised 24 h following the final application.
(ii)
Topically applied IMQ model
In accordance with manufacturer’s instructions, soluble IMQ (Source BioScience, Nottingham, UK) was reconstituted in PBS prior to being dissolved in aqueous control cream. Mice were shaved on their dorsal skin 24 h prior to treatment. Eighty milligrams of 5 % IMQ cream, or aqueous control cream, was applied every 24 h for 5 days. Mice were euthanised 24 h following the final application.
(iii)
Soluble IMQ injection model
Soluble IMQ was reconstituted in PBS to a concentration of 1 mg/ml. Mice were injected, intraperitoneally, with 100 μl IMQ (100 μg) or PBS every 24 h for 5 days. Mice were euthanised 24 h following the final application.
(iv)
TPA model of inflammation
Mice were shaved on their dorsal skin 24 h prior to treatment. One hundred fifty microlitres of 100 μM 12-O-tetradecanoylphorbol-13-acetate (TPA) (Sigma Aldrich, Missouri, USA), or an equal volume of acetone control (Sigma-Aldrich, Missouri, USA), was applied to the mice every 24 h for 1, 3 or 5 days. Mice were euthanised 24 h after the final application.
Cardiac perfusion of animal tissues
Prior to tissue retrieval, all mice were extensively perfused as follows: the right atrium of the heart was cut to allow blood to drain into the chest cavity without compromising the circulation. Perfusions were performed by injecting with 20 ml of PBS, warmed to 37 °C, into the left ventricle of the beating heart using a 23G needle.
RNA isolation from brain tissue and peripheral blood leukocytes
Brain tissue was ‘snap-frozen’ and stored at −80 °C until use. Under RNase-free conditions, brains were homogenised using the TissueLyser LT (Qiagen, Hilden, Germany). RNA was extracted from homogenised tissue using QIAzol® Lysis Reagent (Qiagen, Hilden, Germany), isolated RNA was further purified and genomic DNA was 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). Cells were first passed through a QIAshredder (Qiagen, Hilden, Germany) before RNA was isolated from peripheral blood leukocytes (PBL) using an RNeasy Mini Kit.
GeneChip microarray analysis
Microarray analyses were performed in the Glasgow Polyomics Facility at the University of Glasgow (
www.polyomics.gla.ac.uk), as previously described [
20]. Briefly, 1 μg of purified total RNA was amplified by in vitro transcription and converted to sense-strand complementary DNA (cDNA) using a WT Expression kit (Life Technologies, California, USA). Fragmented and labelled cDNA samples were hybridised to GeneChip Mouse Gene 1.0 ST Arrays (Affymetrix, California, USA). Procedures were carried out as described by the manufacturers. Microarray profiling data were deposited in the National Center for Biotechnology Information Gene Expression Omnibus database with the series entry identifier GSE72214 and can be accessed using the following link:
http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?token=sbmpmemufzynhmj&acc=GSE72214.
Data generated using GeneSpring GX software were normalised using RMA 16. Normalised data were analysed using unpaired t tests to determine the significance of each gene in Aldara-treated mice compared to control mice. p values were adjusted for multiple comparisons using the Benjamini Hochberg multiple comparison test.
Gene ontology terms were assigned to differentially expressed genes using the Database for Annotation, Visualization and Integrated Discovery (DAVID) Bioinformatics Resources v6.7 (
http://david.abcc.ncifcrf.gov/). Analysis was performed in accordance with two protocols outlined by Huang et al. [
21,
22]. The significance of enrichment was determined using a modified Fisher’s exact test. A Benjamini-Hochberg multiple comparison test was used to correct for the rate of type I errors. Co-expression of a gene cluster was considered significant if it satisfied a
p value cutoff of 0.05.
QRT-PCR
Total RNA was reverse transcribed using Quantitect® Reverse Transcription kit (Qiagen, Hilden, Germany) with random primers. Quantitative real-time PCR (QRT-PCR) amplifications were performed in triplicate using PerfeCTa® SYBR® Green FastMix® (Quanta Biosystems, Maryland, USA). 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. A 750-nM mix of forward and reverse primers was used per reaction. QPCR reactions were performed using a Prism® 7900HT Sequence Detection System (Life Technologies, California, USA) for 40 cycles in accordance with manufacturer’s guidelines. The absolute copy number was calculated from a standard curve and normalised to the housekeeping gene, TATA-binding protein (TBP), as previously described [
23]. Fold-change values were calculated by comparing the normalised copy number of individual samples to the mean of the control samples.
Luminex
Blood was collected by cardiac puncture. Plasma was isolated from whole blood by centrifugation. Plasma concentrations of soluble inflammatory mediators were determined using mouse multiplex cytokine Luminex panel kits (Life Technologies, California, USA) in accordance with manufacturer’s instructions.
Generation of a single-cell suspension from brain tissue
Perfused brains were extracted from control and treated mice as described. Brains were digested for 45 min at 37°, 750 rpm in 10 ml digestion buffer (6 μg/ml Liberase TM (Roche), 5 U/ml DNaseI and 25 mM Hepes buffer diluted in HBSS (all Sigma Aldrich, Missouri, USA)). Following digestion, cell suspensions were passed through a 70-μm cell strainer before being washed twice with 2-mM EDTA in HBSS. Myelin removal was performed using myelin removal beads (Miltenyi Biotech, Cologne, Germany) as per the manufacturer’s instructions using an AutoMACS. Total cell number was determined using a haemocytometer.
Flow cytometry
Cells were first incubated with 1-μl FcR block (Miltenyi) per sample and then stained at 4 °C using the antibodies listed in Additional file
2: Table S2. Samples were analysed using an LSR II or FACSAria I/III cytometer (BD Biosciences) and FlowJo software (Tree Star).
Legendplex protein assay
Chemokine protein expression was measured using the Legendplex assay (BioLegend, California, USA) as per the manufacturer’s instructions. In brief, snap-frozen brain tissue was homogenised in N-PER™ Neuronal Protein Extraction Reagent (Thermo Scientific) at a ratio of 10 ml per 1 g of tissue for 20 min on ice. Samples were centrifuged at 10,000g for 10 min, and supernatant was collected. Legendplex beads were incubated with whole brain lysate for 2 h at 600 rpm on a plate shaker. Beads were conjugated with streptavidin-PE for 30 min and were washed twice prior to sample reading using BD LSRII flow cytometer. Samples were differentiated on the basis of bead size and APC fluorescence. Protein quantity was determined using SA-PE fluorescence calibrated to a standard curve. Limits of detection were 1.69 and 1.4 pg/ml for CCL2 and CXCL10, respectively.
Histology
Brain and skin samples were fixed in 10 % neutral buffered formalin prior to processing and paraffin embedding. Processing was performed using the Shandon Citadel 1000 automated tissue processor (Thermo scientific).
(i)
Haematoxylin and eosin staining
Skin samples from treated mice were cut into 5-μm sections and stained with haematoxylin and eosin (H&E) to discern morphology.
(ii)
CD3 T cell staining
Brains extracted from Aldara-treated mice were formalin fixed and embedded in paraffin as described. CD3 staining was performed, using a rabbit anti-mouse CD3 antibody (Vector), by the Veterinary Diagnostic Service Facility at the University of Glasgow. Slides were cut from three sequential regions through the mid-sagittal brain, from four mice per group. Slides were blinded and whole sections were counted for CD3.
(iii)
Doublecortin (DCX) staining for neurogenesis
Seven micrometre sections were cut from formalin-fixed brains from mice treated with Aldara or control cream for 5 days. Sections were stained with polyclonal goat anti-mouse DCX (Santa Cruz Biotechnology, Texas, USA). Slides were blinded, and the mean DCX counts from three areas of the dentate gyrus of the hippocampus were obtained per section.
Burrowing model of mouse behaviour
The assessment of burrowing behaviour was carried out as previously described [
24]. Briefly, groups were acclimatised overnight prior to baseline tests being performed. Forty-eight hours later, mice were put on procedure and were treated with Aldara or control cream as described above. Burrowing tests were carried out on individually caged mice for a 2-h time period, 4 h after Aldara treatment, for three consecutive days.
Statistical analysis
All data were analysed using the Prism 4 software (GraphPad, San Diego, CA, USA). Results are shown as mean ± standard deviation (SD) unless stated otherwise. Data were analysed using the unpaired two-tailed student’s t test if two groups were compared. If more than two groups were compared a one- or two-way ANOVA was performed using the Bonferroni post-test. The figure legends indicate which statistical test was used for each graph. A p value of ≤0.05 was considered as statistically significant.
Discussion
This study demonstrates that peripherally induced, tissue-specific inflammation can ‘remotely’ alter the transcriptional profile of the brain, independently of a similar response in the periphery. Topical Aldara treatment, which induces a psoriasis-like skin inflammation, causes the upregulation of a number of genes in the brain, including seven inflammatory chemokines and one chemokine receptor, a response that was not mimicked by PBL. This response peaks following the third application but persists beyond the fifth application suggesting a prolonged expression in the brain. That this response was also seen with topically applied reconstituted IMQ, but not with a sterile model of skin inflammation, implicates peripheral TLR ligation as an important mechanism in the induction of this response. In addition, the lack of a brain response following an I.P. injection of reconstituted IMQ suggests that the localised skin response is also an important factor in driving the brain response.
The use of reconstituted IMQ yielded an interesting result in that a significant brain response was observed in the absence of an overt skin inflammation, which is important when considering the mechanisms of communication between the periphery and the brain. Although the lack of cutaneous inflammation is reportedly due to the omission of isostearic acid [
25], it is difficult to hypothesise how the effects of peripheral inflammation are being transmitted to the brain without the generation of a local inflammatory response or circulating inflammatory cytokines. It is also notable that neither induction of a ‘sterile’ cutaneous inflammatory response nor intraperitoneal administration of IMQ resulted in the generation of a brain response in our study. One possibility that might unite these disparate observations is that the communication between the inflamed skin and the brain is not driven by inflammatory agents but by direct ligation of TLRs on sensory nerves within the skin which then transmit a response to the brain inducing transcriptional changes associated with ‘sickness behaviour’. A number of studies have shown TLR expression on peripheral nerves [
37‐
39], and a further series of experiments is planned by our lab to specifically address this issue. An alternative hypothesis is that tissue-resident myeloid cells, or other leukocytes exposed to the inflamed environment in the skin, could transmit the signal to the brain directly, by migrating through the circulation to the brain. Alternatively, they could transmit inflammatory signals indirectly, by causing damage to peripheral nerves that is then sensed by the brain. Whilst we do not feel this provides a mechanistic explanation for our observations, due to the fact that TPA treatment induces a potent inflammatory response in the skin and is not associated with chemokine upregulation in the brain, further studies would be required to investigate this potential route of immune-to-brain communication.
Strikingly, in addition to triggering the expression of chemokines in the brain, topical Aldara treatment was associated with a sizable recruitment of several leukocyte populations to the brain, including CD4+ T cells, CD8+ T cells, NK cells, NK T cells and monocytes. On day 3, the monocyte populations in the brain were Ly6C+CD64- and Ly6C+CD64+, whereas by day 5, they were predominantly Ly6C+CD64+ monocytes and Ly6C-CD64+ macrophages. This is likely indicative of monocyte to macrophage differentiation. Although we cannot directly link enhanced chemokine production to leukocyte recruitment, the chemokine profile in the brain was highly specific to the cell populations recruited.
Several groups have shown that chemokine induction in the brain can induce leukocyte infiltration, both under neuroinflammatory conditions [
16,
40‐
43], in response to peripheral inflammation [
44] and during CNS viral infection [
16,
45]. The brain specific chemokine profile that is triggered following topical Aldara treatment bears a notable resemblance to the array of chemokines that are induced in the brain following encephalitic virus infection [
16]. Indeed, the leukocyte populations that infiltrate the brain following Aldara treatment: monocytes, effector memory T cells, NK cells and NKT cells, are populations that play a prominent role during viral infection and are often crucial for pathogen clearance. Furthermore, as would be expected during viral encephalitis, IMQ, the active component of Aldara, stimulates TLR7 to mimic a viral response. Thus, in response to TLR7 ligation in the periphery, the brain may specifically produce monocyte and lymphocyte chemoattractants to recruit the appropriate populations required to protect the brain against a potential viral infection.
The infiltration of immune cells into the brain following a remote immune challenge is both surprising and interesting. The generation of a cellular brain response following peripheral inflammation could indicate the presence of a protective surveillance mechanism initiated when a potential CNS insult is anticipated to ‘prime’ the brain. However, further studies would need to be performed before we could begin to unpick the role of these infiltrating cells, and their fate beyond the time points examined.
A key question regarding this response relates to the temporal relationship between chemokine induction in the brain and leukocyte recruitment, as infiltrating leukocytes may either have responded to chemokines made in the brain prior to their recruitment, or may in fact have contributed to the expression of the detected chemokines. However, at least two key inflammatory chemokines, CXCL10 and CCL2, are upregulated in the brain at the protein level by 4 h respectively following topical Aldara treatment, whereas immune cells are not recruited into the brain until 72 h post-treatment. This leads us to propose that chemokine production within the brain parenchyma precedes, and is likely to be at least partially responsible for, the subsequent recruitment of immune cells.
In addition to immune cell infiltration, chemokines are also implicated in a number of neuromodulatory processes, including neurotransmission, development and neurogenesis [
11,
46‐
48], indicating that changes in chemokine expression could disrupt essential homeostatic events in the CNS [
49]. In this study, we have been able to link peripheral cutaneous inflammation with two relevant functional outputs: an impairment in adult neurogenesis at the dentate gyrus and a reduction in burrowing activity of mice, a behaviour that has been shown to be a sensitive representation of neuropsychiatric imbalance in a number of inflammatory CNS disease models [
24,
50]. It is worth noting that whilst food pellets were a convenient material to be used in large quantities, it is assumed that this model is not appetite-dependent and mice were not ‘food rationed’ at any time. Nevertheless, it remains possible that an additional contribution to the lack of burrowing seen in the inflamed mice may come from a generalised lack of interest in food as a consequence of an impaired sense of well-being. Of course, to fully characterise the behavioural deficits in response to Aldara treatment, and to associate these with neuropsychiatric conditions, more sophisticated, impairment-specific behavioural tests should be performed. Nonetheless, our data suggest that inflammation generated at distant sites has the potential to negatively influence homeostatic brain function.
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
All data was collected, analysed and interpreted by AM, CAT, LN, GJG and JC. Experiments were conceived and designed by JC and GJG. The manuscript was written by the authors and all authors read and approved the final manuscript.