Skip to main content

Genetically enhancing the expression of chemokine domain of CX3CL1 fails to prevent tau pathology in mouse models of tauopathy

Abstract

Background

Fractalkine (CX3CL1) and its receptor (CX3CR1) play an important role in regulating microglial function. We have previously shown that Cx3cr1 deficiency exacerbated tau pathology and led to cognitive impairment. However, it is still unclear if the chemokine domain of the ligand CX3CL1 is essential in regulating neuronal tau pathology.

Methods

We used transgenic mice lacking endogenous Cx3cl1 (Cx3cl1−/−) and expressing only obligatory soluble form (with only chemokine domain) and lacking the mucin stalk of CX3CL1 (referred to as Cx3cl1105Δ mice) to assess tau pathology and behavioral function in both lipopolysaccharide (LPS) and genetic (hTau) mouse models of tauopathy.

Results

First, increased basal tau levels accompanied microglial activation in Cx3cl1105Δ mice compared to control groups. Second, increased CD45+ and F4/80+ neuroinflammation and tau phosphorylation were observed in LPS, hTau/Cx3cl1−/−, and hTau/Cx3cl1105Δ mouse models of tau pathology, which correlated with impaired spatial learning. Finally, microglial cell surface expression of CX3CR1 was reduced in Cx3cl1105Δ mice, suggesting enhanced fractalkine receptor internalization (mimicking Cx3cr1 deletion), which likely contributes to the elevated tau pathology.

Conclusions

Collectively, our data suggest that overexpression of only chemokine domain of CX3CL1 does not protect against tau pathology.

Background

Fractalkine signaling in the CNS represents a unique microglial-neuron receptor-ligand pair, where fractalkine (CX3CL1) is expressed by neurons and its cognate receptor CX3CR1 is exclusively expressed by the CNS resident microglia [1]. CX3CL1 is a 373-amino acid protein, which contains an extracellular chemokine domain linked to a mucin-like stalk [2, 3]. CX3CL1 is functional in its membrane-bound form but can also be cleaved through metalloprotease (ADAM10/ADAM17) activity to produce a ~ 95-kDa soluble moiety [4, 5]. It has been proposed that the heavily glycosylated mucin-like stalk of fractalkine provides rigidity to the chemokine domain for the adhesive potency of the chemokine domain during patrolling/crawling behavior [6]. Several mouse models have been used to elucidate the role of fractalkine in mediating neurodegenerative and neuroinflammatory processes [7,8,9,10,11].

CX3CL1-CX3CR1 signaling is regulated through direct neuron-microglia interaction, which acts to tether microglia until pathological activation, via an inflammatory influence, or through normal physiological activity, which disrupts this interaction through the cleavage of CX3CL1 [12, 13]. Disruption of CX3CL1-CX3CR1 signaling by chemical or genetic manipulation induces dramatic morphological activation and altered levels of scavenger/inflammatory receptors on the cell surface, alterations in pro-inflammatory chemokine production, and over-sensitization to pathological insults [14,15,16,17].

Previous studies from our group have explored the role of CX3CL1 signaling in the context of Alzheimer’s disease (AD) and related dementias. Notably, we found that disrupting the CX3CL1-CX3CR1 signaling axis reduces Aβ burden with concomitant increases in pro-inflammatory IL-1 and heightened microglial activation in both APPPS1/Cx3cr1−/− and APPPS1/Cx3cl1−/− transgenic mouse models of AD [18]. Interestingly, this phenomenon was unaffected by the presence of soluble CX3CL1 [18]. In a separate study, converse to the protective anti-amyloid phenotype observed in APPPS1/Cx3cr1−/− mice, deletion of Cx3cr1 in hTau mice resulted in hyperphosphorylation and aggregation of tau, worsened cognitive function, and increased microglial inflammation [17]. This effect was regulated via the same IL-1-p38 MAPK axis [17, 19]. The dichotomy between the two studies likely stems from the type of pathological insults present, namely Aβ is extracellular whereas hyperphosphorylated tau exists primarily intraneuronally [20]. The precise mechanism of how disrupting the CX3CL1-CX3CR1 signaling affects the microglia either to a beneficial (in the case of Aβ study) or to a detrimental degree (in the hTau study) is still unclear. However, it is possible that the IL-1β promotes phagocytic phenotype of microglia in clearing Aβ (in case of APPPS1/Cx3cr1−/− and APPPS1/Cx3cl1−/− mice), while causing collateral damage (for example, over-activation of p38 MAPK) in neurons and leading to tau hyperphosphorylation [17,18,19]. Seemingly, contrary work demonstrated that Cx3cl1 overexpression through viral transfection models reduces tau and α-synuclein pathology [10, 21]. The present study seeks to determine if genetically expressing only the soluble chemokine domain of CX3CL1 could prevent tau pathology in both chemical (LPS) and genetic (hTau) mouse models of tauopathy.

Methods

Experimental animals

A mouse line (Cx3cl1105Δ) exclusively expressing obligate soluble CX3CL1 featuring only the chemokine domain, without the mucin stalk, was generated by introducing bacterial artificial chromosome (BAC) transgene encoding truncated CX3CL1 (B6.Cg-Tg(Cx3cl1*)1Jung/J RRID:IMSR_JAX:027119) to Cx3cl1−/− mice (RRID: MGI_2388041) [22]. For the current study, hTau+/−;Mapt−/− [23] (acquired from the Jackson Laboratory) which expressed all six isoforms of the human MAPT under the control of the endogenous human MAPT promoter and backcrossed into Cx3cl1105Δ animals [22] was subsequently intercrossed to generate both hTau+/−;Mapt−/−/Cx3cl1−/− (referred as “hTau/Cx3cl1−/−”) and hTau+/−;Mapt−/−/Cx3cl1−/−/ Cx3cl1105Δ (referred as “hTau/Cx3cl1105Δ”). Mice were housed in both the Cleveland Clinic Biological Resources Unit and University of New Mexico Animal Research Facility. Both facilities are fully accredited by the Association and Accreditation of Laboratory Animal Care. The Institutional Animal Care and Use Committee at respective institutions approved all experimental procedures.

Lipopolysaccharide injections

Three milligrams per kilogram b.w. LPS (Sigma-Aldrich) was administered intraperitoneally (i.p) to 2-month-old mice and sacrificed 24 h post-injection. The hemi-brains were post-fixed in 4% paraformaldehyde (PFA) followed by cryopreservation in 30% sucrose for immunohistochemistry (IHC) experiments. The remaining half of the brains were microdissected into the hippocampal and cortical fractions and snap frozen in liquid nitrogen and stored at − 80 °C for biochemical analysis.

Western blotting

Microdissected cortical and hippocampal fractions were homogenized using T-PER reagent (Thermo #78510) containing phosphatase and protease inhibitor cocktails (Thermo #78429, #78443; Sigma-Aldrich #p5726) and briefly sonicated at 20% amplitude for 10 s. Homogenates were centrifuged at 15,000 rpm, and the protein in the supernatant measured via BCA assay (Thermo #23225). Total protein (30–60 μg) were resolved by SDS-PAGE, transferred to the PVDF membranes (#IPFL10100 Millipore), and probed with phosphorylated tau antibodies (AT8 for pS199/pS202/pT205, AT180 for pT231 at 1:5000; Thermo; and PHF-1; 1:10,000; a generous gift from Dr. Peter Davies), total tau (Tau5 1:10,000; Thermo), and GAPDH (1:20,000; Millipore) (loading control). The membranes were incubated with near-IR conjugated (Thermo #A11371, #A11367) or HRP-conjugated secondary antibodies (Jackson ImmunoResearch), visualized, and either quantitated using LICOR Odyssey imaging systems (for the data presented in Fig. 2) or developed with enhanced chemiluminescence reagent and quantified by Alpha Innotech® software (for the data presented in Fig. 1).

Fig. 1
figure 1

LPS-induced tau phosphorylation and microglial activation are exacerbated in Cx3cl1105Δ mice. ad Two-month-old fractalkine (Cx3cl1−/−)-deficient mice and the mice exclusively expressing the chemokine domain (lacking the mucin-like domain, red) (CX3CL1105Δ) with a Myc tag were injected with LPS (3 mg/kg b.w; i.p) or vehicle (VEH, Hank’s balanced salt solution or HBSS) and sacrificed 24 h post-injection. ef Western blotting of the hippocampi revealed significantly increased total tau (Tau5) (> 1.5-fold) in VEH-treated Cx3cl1105Δ vs. Cx3cl1−/− mice (mean + SEM; **p < 0.01; n = 3; two-way ANOVA followed by Tukey’s post hoc test). Both AT8/Tau5 and AT180/Tau5 ratios were significantly higher in LPS-treated Cx3cl1105Δ compared to LPS-treated Cx3cl1−/− or Non-Tg mice (mean + SEM; *p < 0.05; **p < 0.01; n = 3; two-way ANOVA with Tukey’s post hoc test). g Immunohistochemistry (IHC) analysis revealing a modest increase in AT8 (pS199/pS202 tau) among experimental genotypes or between VEH- or LPS-injected mice in the CA3 hippocampal areas. Scale bar, 20 μm. hk IHC showing elevated Iba1+/F4/80+ reactive microglia in VEH-treated Cx3cl1105Δ mice that is enhanced with LPS treatment. Quantification reveals statistically higher form factor units (higher number means more towards circular contour) for Iba1+ microglia in Non-Tg and Cx3cl1105Δ mice in LPS-treated groups (mean + SEM; ***p < 0.0001 vs. **p < 0.01 for Non-Tg with LPS; two-way ANOVA with Tukey’s post hoc test; n = 3–6 mice per group). Scale bars (h, j) 25 μm

Immunohistochemistry

Sagittal free-floating 30-μm sections were subjected to standard sodium citrate antigen retrieval for 10 min at 95 °C followed by blocking in normal goat serum containing 0.1% Triton-X. The sections were incubated overnight with AT8, AT180, or myeloid cell surface markers (CD45 #MCA1388 1:500; BioRad; F4/80 #MCA497G 1:500 BioRad; Iba1 #019-19741 1:500; Waco) and respective biotinylated secondary antibodies (Vector Laboratories Catalog # BA-2000, BA-9400, BA-1000) and developed using 3-3′-diaminobenzidine with or without the nickel enhancer. Data were quantified using percentage immunoreactive area or form factor analysis [24, 25]. First, percent immunoreactive area for CD45 and F4/80 was processed using ImageJ, where five random fields per section were manually defined as region of interest (ROI) in three random sections (focusing only on the ones containing dorsal hippocampal region) per mouse and in n = 6 mice per genotype and consistently scored to detect percentage of CD45 and F4/80 immunoreactive area using ImageJ software. Briefly, first, the RGB images were converted into 8-bit gray scale, and then the images were processed to adjust the threshold, which was kept constant for all images. Finally, using the analyze tool in ImageJ, the total immunoreactive area per field was scored by an automated routine. After completing the scoring for all sections, the mean percentage area along with standard error of mean was plotted. For quantifying the roundness of Iba1+ microglia, we utilized form factor (FF) algorithm in the ImagePro Plus® software. FF measures the roundness of an object; in this case, it was Iba1+ microglia. We [25] and others [24] have previously described FF-based quantification of microglial contour as an indirect measure of its phagocytic/morphological activation state. Briefly, three images at random were taken in three different frontal cortical sections from each animal with at least three animals per group. FF measures the contour irregularity of a cell, i.e., FF is higher (approaches 1.0) in bushy cells, characterized by larger cell bodies that are less ramified, while morphologically “resting” microglia appear smaller cell bodies with abundant branches of regularly ramified processes, which would have lower FF values (closer to 0.0). A total of 235 (Non-Tg-Veh), 184 (Non-Tg-LPS), 220 (Cx3cl1−/−-Veh), 192 (Cx3cl1−/−-LPS), 234 (Cx3cl1105Δ-Veh), and 200 (Cx3cl1105Δ-LPS) microglial cells were scored for FF analysis.

Behavioral analysis

Morris water maze

Mice underwent a 3-day training using a visible platform, which was relocated to different quadrants of an opaque, water-filled maze each of the four trials per day. The first 3 days of visible platform training was to allow animals to learn the procedures of the task (i.e., swim and get onto the platform to escape from the pool). Next, the animals received 5 days of memory testing in which the platform was submerged and remained in a constant location with static spatial cues around the room. Latency to reach a stationary hidden platform was recorded across four separate trials per day, for all 5 days. Mice were allotted 60 s to reach the platform during both training and experimental days. Latency to reach the platform, swim speed, within-day learning, and across-day learning was examined to determine cognitive differences between experimental genotypes. We analyzed three key parameters relevant to learning trends: (a) acquisition index is a measure comparing learning within each trial day across four trials; (b) savings index is a measure of memory consolidation from the final trial of one testing day to the first trial of the following day; and (c) slope plots the trajectory of learning curve. These indices were scored for all 5 days as previously described [26].

Y-maze

The Y-maze is used to assess spatial working memory during a 5-min trial where each mouse is allowed to freely explore each arm of the Y-maze [27]. Total arm entries, repeat ratio (defined as the number of times a mouse enters the same arm twice over a total number of arm entries), and the spontaneous alternation (defined as when a mouse consecutively entered three different arms) were recorded, as previously described [17].

Multiplex ELISA assay

Cytokine levels were all normalized to total protein concentration following BCA assay. Multiplex assays were performed according to the manufacturer’s instructions using reagents provided with the kit (Invitrogen Mouse 20-plex Cytokine Panel, Cat# LMC0006M). Following sample incubation, plates were washed, incubated with streptavidin-RPE for 30 min, washed three times, and followed by a final addition of 125 μl wash solution to all wells. The plates were read on a Luminex Magpix unit (Life Technologies), and initial analyses were performed by Xponent software and results exported into Microsoft Excel for further processing. The sample size was set to 50 μl, and the minimum count was set to 100 events/bead regions.

Flow cytometry

Mononuclear cells were isolated via a density centrifugation technique at the interface of a 30/70% Percoll gradient (Fisher Scientific #17-5445-01), as previously described [28]. Cells were blocked using Fc blocking reagent (BD Biosciences #553141) for 10 min before a 30-min incubation with fluorophore-conjugated flow cytometry antibodies against CD11b (FITC; BD Biosciences #553310), CD45 (APC; BD Biosciences; #559864), and CX3CR1 (PE; BD Biosciences #565798). Data were acquired using BD Biosciences Fortessa Flow Cytometer and analyzed using FlowJo single cell analysis software. Fifty thousand events were minimally collected before data processing. Mean fluorescent intensity was utilized in conjunction with total event counts in order to quantify the number of brain-resident microglia (CD11b+/CD45low cells) and relative expression of CX3CR1.

Statistical analysis

Data are presented as mean ± SEM unless otherwise noted. Comparisons between two groups were analyzed using Student’s t test (two-tailed; unpaired) at 95% confidence interval. Multiple group comparison or multiple comparisons were analyzed using ANOVA or MANOVA followed by Tukey’s or Dunnett’s post hoc tests. The analysis was performed using Prism GraphPad or SPSS software. Significance was determined at *p < 0.05, **p < 0.01, and ***p < 0.001. Individuals who were blinded to the genotype/treatment groups performed the data analysis.

Results

Enhanced microglial activation in Cx 3 cl1 105Δ mice during LPS-induced endotoxemia

To explore the effect of neuronal Cx3cl1 deficiency and the overexpression of the shed fractalkine moiety on LPS-induced tau pathology, we utilized fractalkine-deficient (endogenous Cx3cl1−/−) and cleaved soluble fractalkine (Cx3cl1105Δ)-expressing transgenic mice, which express only the soluble chemokine domain of CX3CL1 [22]. Previously, this model revealed a differential requirement for soluble and membrane-bound CX3CL1 in the context of dendritic macrophage processes within the gut epithelium [22]. First, we confirmed the expression of Cx3cl1105Δ in the whole brain lysate via detection of a c-Myc tag present in the C-terminal end of the Cx3cl1105Δ BAC construct (Fig. 1a-c) [22]. Interestingly, the mRNA levels of Cx3cl1 were significantly higher in Cx3cl1105Δ mice compared to non-transgenic controls (Additional file 1: Figure S1A). However, the protein levels of soluble CX3CL1 were comparable to that of non-transgenic mice (Additional file 1: Figure S1B).

Previous work from our lab has demonstrated that LPS induces tauopathy as early as 24 h following administration [17]. We administered LPS (3 mg/kg b.w., single dose; i.p) to 2-month non-transgenic C57BL/6J (Non-Tg), Cx3cl1−/−, and Cx3cl1105Δ mice (Fig. 1d). Based on our previous reports that LPS leads to tau phosphorylation within 24 h [17, 29], mice were sacrificed 24 h post-injection to determine the alterations in tau phosphorylation and microglial activation. Western blotting revealed a significant (> 1.5-fold) increase in the total tau (Tau5) levels in vehicle-treated Cx3cl1105Δ mice compared to Cx3cl1−/− mice (Fig. 1e, f). No differences were detected in the basal level of tau phosphorylation among LPS-treated Non-Tg, Cx3cl1−/−, and Cx3cl1105Δ groups. Notably, LPS administration elevated phosphorylated tau levels at AT8 (S199/S202/T205) and AT180 (T231) sites in Non-Tg and Cx3cl1105Δ mice that were nearly two- to fourfold elevated compared to vehicle-treated groups, whereas Cx3cl1−/− mice revealed only a modest increase in AT8+ tau in LPS-injected groups (Fig. 1e, f). However, the LPS-induced AT8 positivity was more robust in 6-month-old Cx3cl1−/− mice (data not shown).

Immunohistochemical analysis revealed a moderate increase in AT8 immunoreactivity in the hippocampus (CA3) of Cx3cl1105Δ mice compared with other genotypes (Fig. 1g). Furthermore, Iba1+ microglial immunostaining with subsequent form factor quantitative analysis revealed an increase in microglial activation with LPS in all three genotypes tested (Fig. 1h, i). Reactive microglia displayed thick, less ramified processes in Non-Tg, Cx3cl1−/−, and Cx3cl1105Δ mice treated with LPS compared with their respective vehicle-injected controls (Fig. 1h). While the expression of a major macrophage marker—F4/80—appeared elevated with LPS treatment in the IHC images, the differences were not statistically significant due to large variability. We also did not detect any differences among experimental genotypes in either vehicle or LPS-injected groups (Fig. 1j-k).

Overexpression of only chemokine domain of CX3CL1 fails to mitigate microglial activation and tau pathology induced by Cx 3 cl1 deficiency in hTau mice

To determine the effect of Cx3cl1 deficiency and specific effects of Cx3cl1105Δ expression in hTau mice, the brains of 6-month-old hTau, hTau/Cx3cl1−/−, and hTau/Cx3cl1105Δ mice were analyzed. Significant increases in AT8 site tau phosphorylation, but not AT180, PHF-1 sites, or total tau (Tau5) were detected in the hippocampus of hTau/Cx3cl1−/− and hTau/Cx3cl1105Δ mice compared to hTau mice (Fig. 2a, b). Immunohistochemical analysis revealed increased AT8 immunoreactivity in hTau/Cx3cl1−/− and hTau/Cx3cl1105Δ mice compared to hTau mice in the CA3 region of the hippocampus, where the AT8+ tau pathology was robust (Fig. 2c). Significant increases in Iba1, CD45, and F4/80 immunoreactivities were detected in hTau/Cx3cl1−/− and hTau/Cx3cl1105Δ mice compared to hTau mice (Fig. 2d, e). Multiplex ELISA analysis of hippocampal lysates revealed a significant increase in inflammatory IL-1α in hTau/Cx3cl1−/− mice and a modest increase (p = 0.07) for hTau/Cx3cl1105Δ mice compared to hTau mice (Fig. 2f). Notably, IL-1β levels were significantly elevated in both hTau/Cx3cl1−/− and hTau/Cx3cl1105Δ mice compared to hTau mice (Fig. 2f), which is consistent with our previous studies linking increased IL-1β production to the microglial p38 MAPK signaling pathway [17,18,19].

Fig. 2
figure 2

Increased tau pathology, IL-1α/IL-1β, and microglial activation in 6-month-old hTau/Cx3cl1−/− and hTau/Cx3cl1105Δ mice. a, b Western blotting revealed increases in AT8+ tau in the hippocampus of hTau/Cx3cl1−/− and hTau/Cx3cl1105Δ mice compared to hTau controls. c AT8 IHC revealed increased reactivity in the CA3 regions of hTau/Cx3cl1−/− and hTau/Cx3cl1105Δ groups compared to hTau controls. Scale bar, 30 μm. d Significant increases in both CD45 and F4/80 immunoreactivities were detected and quantified (e) in the cortex of hTau/Cx3cl1−/− and hTau/Cx3cl1105Δ mice compared to controls. f A significant increase in IL-1α and IL-1β was observed in both hTau/Cx3cl1−/− and hTau/Cx3cl1105Δ mice via ELISA. n = 6 mice per group except for ELISA (n = 10). Three independent experiments were performed for each analysis. Error bars represent SEM. One-way ANOVA followed by Tukey’s post hoc test: *p < 0.05, **p < 0.01, ***p < 0.001

Microglial cell surface level of CX3CR1 is significantly reduced in Cx 3 cl1 105Δ mice, mimicking a Cx3cr1 deficiency phenotype

Flow cytometric analysis was performed to further explore the possible role of CX3CL1105Δ in the regulation of microglial activation in the Cx3cl1105Δ mice. Isolated brain myeloid cells were stained with antibodies against CD45 and CD11b to differentiate brain-resident microglia (CD11b+CD45low) and peripherally derived myeloid cell (CD11b+CD45hi) population. There were no significant differences in total microglia or peripherally derived myeloid cells within the brains of Non-Tg, Cx3cl1−/−, or Cx3cl1105Δ mice (Fig. 3a, c). However, microglia from the Cx3cl1105Δ mice displayed significantly lower cell surface expression for CX3CR1 compared to both Non-Tg and Cx3cl1−/− mice (Fig. 3b, d). This is despite showing elevated levels of Cx3cr1 mRNA in the brain (Additional file 1: Figure S1C). Furthermore, the complete lack of fractalkine (in Cx3cl1−/− mice) was not sufficient to promote the downregulation of microglial CX3CR1 levels (unlike in Cx3cl1105Δ mice) (Fig. 3b, d), further supporting previously published reports of receptor downregulation in this model [18].

Fig. 3
figure 3

Expression of the microglial CX3CR1 is decreased in Cx3cl1105Δ mice. a, c Flow cytometry on isolated brain mononuclear cells revealed no alteration in the total number of resident microglia in Non-Tg, Cx3cl1−/−, or Cx3cl1105Δ mice (Cd11b+/CD45low). b, d Overall decreased CX3CR1 expression in the CD11b+CD45low (microglial population) in Cx3cl1105Δ mice compared to Non-Tg and Cx3cl1−/− mice (mean + SEM; one-way ANOVA followed by Tukey’s post hoc test: *p < 0.05, **p < 0.01; n = 3 mice per group). e Working model of microglial-neuronal fractalkine signaling axis. Note that the neuronal-derived CX3CL1 either as full length (in case of Non-Tg mice), as complete knockout (in Cx3cl1−/− mice), or as the only soluble form with chemokine domain (Cx3cl1105Δ mice) differentially regulates the expression of microglial CX3CR1 (which is a seven transmembrane G protein-coupled receptor) on the cell surface. This in turn may lead to over-activation of microglia in (Cx3cl1105Δ mice) and enhanced neuroinflammation and exacerbation of tau pathology in chemical (LPS) or genetic (hTau) model of tauopathy

Cx 3 cl1 deficiency leads to cognitive impairments in aged hTau mice

Behavioral and cognitive dysfunctions are key clinical deficits in AD and tauopathies. To explore the effect of Cx3cl1 deficiency in hTau mice, we generated and aged additional cohorts of hTau, hTau/Cx3cl1−/−, and hTau/Cx3cl1105Δ mice to 12 months of age and subjected each group to a Morris water maze behavioral analysis. No statistically significant differences were detected among experimental genotypes with regard to swimming speed or latency to reach the platform during visible or memory trials, respectively (Fig. 4a, b). hTau mice performed significantly better across all five testing days as measured by their respective learning slopes across days 1–5 compared to hTau/Cx3cl1−/− or hTau/Cx3cl1105Δ mice (Fig. 4c). Detailed analysis of the acquisition index, which is a measure comparing learning within each individual trial day across four trials, and the savings index, a measure of memory consolidation from the final trial of one testing day to the first trial of the following day, was performed across all 5 days to explore learning trends as previously described [26]. Of interest, hTau mice performed worse day-to-day as measured by savings index, which measures the average memory consolidation from trial 4 on one testing day to trial 1 on the following day, across all 5 days (Fig. 4d). Overall, hTau mice performed better within each respective testing day compared to hTau/Cx3cl1−/− or hTau/Cx3cl1105Δ mice with regard to new memory formation (acquisition index; Fig. 4e). These data suggest that hTau mice have deficiencies transferring information but attempt to compensate by learning well within each respective testing day, thereby overcoming these impairments and performing better overall than either hTau/Cx3cl1−/− or hTau/Cx3cl1105Δ mice. Further, the memory impairments induced by Cx3cl1 deficiency were unable to be overcome by the overexpression of only chemokine domain of CX3CL1. We also performed the Y-maze test to assess the working memory and did not find any significant differences in spontaneous alternation ratio between these groups (data not shown).

Fig. 4
figure 4

Impaired learning in hTau/Cx3cl1−/− and hTau/Cx3cl1105Δ mice. Morris water maze was performed on 12-month-old mice. Mice were subjected to a 3-day visible training paradigm (a), followed by a 5-day hidden trial period (memory testing). b Where latency to reach the platform was recorded (seconds, sec). c Analysis of the linear regression slope adjusted for each genotype revealed that hTau mice learned the task better than hTau/Cx3cl1−/− or hTau/Cx3cl1105Δ mice over the 5-day hidden trial period. d hTau mice had a much lower savings index than hTau/Cx3cl1−/− or hTau/Cx3cl1105Δ mice during the 5-day hidden trial period. e hTau mice show higher acquisition index than hTau/Cx3cl1−/− or hTau/Cx3cl1105Δ mice during the 5-day testing period. Mean + SEM; one-way ANOVA followed by Tukey’s post hoc test: *p < 0.05, **p < 0.01, ***p < 0.001; n = 10 mice per group

Discussion

CX3CL1-CX3CR1 represents a unique signaling axis between the microglia and neurons, which is profoundly involved in the suppression of innate inflammatory responses. Alterations in fractalkine signaling by chemical or genetic manipulations have dichotomous consequences within the context of canonical AD pathological outcomes. Notably, the absence of fractalkine signaling ameliorates Aβ plaque burden in APPPS1 transgenic mice [30], but exacerbates intraneuronal tau pathology in hTau mouse model of pure tauopathy [17], even though both events likely occur via dysregulation of IL-1β-p38 MAPK signaling pathway [19]. Here, we demonstrate that the expression of the chemokine domain of CX3CL1 does not suppress inflammation-induced tau pathology or mitigate microglial responses.

Earlier studies suggested that Cx3cr1 deficiency increased tau phosphorylation in both LPS and hTau models of tau pathology [17]. This suggests that the presence of CX3CR1 may downregulate microglial pro-inflammatory signaling and mitigate inflammation-induced tau hyperphosphorylation. For reasons currently unknown, unlike Cx3cr1−/− mice, the Cx3cl1−/− mice demonstrate only a modest increase in AT8+ tau following LPS administration. Tau phosphorylation in hTau/Cx3cl1−/− mice seems to mimic hTau/Cx3cr1−/− mice as previously reported [17]. Furthermore, the expression of chemokine domain of fractalkine has virtually no beneficial effects on either LPS-mediated microglial morphological alterations or AT8/AT180 site tau phosphorylation. This observation suggests that in the absence of membrane-bound form, chemokine domain of the fractalkine may, in fact, disrupt normal microglia-neuron signaling, leading to downregulation and/or internalization of fractalkine receptor on the microglial cell surface. Our flow cytometry analysis reveals decreased microglial CX3CR1 levels in the Cx3cl1105Δ mice compared with Non-Tg or Cx3cl1−/− and supports this hypothesis. Interestingly, a previous study observed prolonged downregulation of cell surface CX3CR1 on aged microglia in response to LPS [31]. This reduced CX3CR1 on the CD11b+ microglia corresponded with delayed recovery from sickness behavior, elevated IL-1β induction, and reduced TGFβ [31]. Reduced Cx3cr1 expression (both mRNA and protein) in monocytes was also reported following septic shock [32]. This loss of monocyte-specific CX3CR1, which causes sepsis-induced lethality in humans, compromised this cell’s ability to respond to a fractalkine challenge [32]. While these and our current results suggested pro-inflammatory and pathological effects of reduced CX3CR1 expression on the microglial cell surface, the exact intra-microglial alterations in Cx3cl1105Δ mice will need to be further explored using isolated microglia and high-throughput single-cell RNA sequencing. Surprisingly, the Cx3cl1105Δ mice also had an increased baseline expression of total tau. Finally, there may be the remote possibility of ligand-independent CX3CR1 negatively influencing TLR4 signaling in immune cells (in Cx3cl1−/− mice) and reducing pro-inflammatory cytokine secretion. Indeed, such un-liganded receptor function was recently reported for progesterone receptor B (without progesterone, acting alone) in the regulation of the function of estrogen receptor-α affecting the proliferation and survival of breast cancer cells following estradiol stimulation [33]. Alternatively, the chemokine domain of CX3CL1 has also been shown to induce intracellular signaling independent of CX3CR1 via binding to αvβ3 integrins [34].

Similar to our previously reported exacerbation of tau pathology in hTau/Cx3cr1−/− mice [17], neuronal fractalkine deletion seems to worsen tau pathology in hTau mice, although differences in tau phosphorylation were only detected at the AT8 (S202) site. Fractalkine deletion elevates the pro-inflammatory response from microglia in hTau/Cx3cl1−/− and hTau/Cx3cl1105Δ mice compared to hTau mice. Further, cognitive abnormalities are evident in aged hTau/Cx3cl1−/− or hTau/Cx3cl1105Δ mice regardless of increased production of the soluble chemokine domain of fractalkine in the latter group. Given that hTau mice display impaired performance in the Morris water maze at 12 months of age [35], it is plausible that Cx3cl1105Δ overexpression fails to prevent cognitive impairment in hTau mice.

Our results contrast with a previous report where AAV-transduced overexpression of soluble fractalkine rescued several pathological phenomena including the hyperphosphorylation of tau at multiple epitopes and microglial phenotypes in a mouse model of tauopathy, rTg4510 [10]. The discrepancies between our study and Nash et al. could be due to a number of factors including the following: (1) Inducible AAV approach vs. our germline genetic system—their animal model had intact membrane-bound CX3CL1, while the Cx3cl1105Δ mice did not. (2) Differences in the structure of the soluble fractalkine moiety—in the AAV study, the mucin stalk of fractalkine was included, whereas, in our germline Cx3cl1105Δ mice, only the soluble chemokine domain, without the mucin stalk, was present. A previous 3D structural analysis of different domains of CX3CL1 has suggested that mucin stalk of CX3CL1 is important for the presentation of the chemokine domain to the outer cell membrane and increases adhesive interaction between CX3CL1 and CX3CR1 [6]. Therefore, lack of the mucin stalk in CX3CL105Δ may not be sufficient to restrict LPS-induced or hTau-mediated microglial activation [36]. (3) Presence of endogenous CX3CL1 in rTg4510 mice vs. the lack of it in Cx3cl1105Δ mice—because of this, the levels of soluble CX3CL1105Δ levels in Cx3cl1105Δ mice (which is comparable to that of Non-Tg (see Additional file 1: Figure S1B)) may be insufficient compared to significantly higher levels of soluble CX3CL1 levels in the AAV study. (4) rTg4510 vs. hTau are two different types of tauopathy mouse models. In rTg4510 only, 4R-Tau with a P301L mutation is expressed and pathological tau is present at 13-fold higher than endogenous levels (AAV study), vs. only an approximate two- to threefold higher expression of all six isoforms, including both 3R and 4R tau, in hTau mice (current study). Based on the data from these two studies, we hypothesize that when there is a robust tau pathology (like in rTg4510 mice), the effect of soluble CX3CL1 (containing the mucin stalk) may be beneficial and the benefits are discernable. We also speculate that this beneficial effect could be due, in part, to the contributions from the membrane-bound form of endogenous CX3CL1, present in the rTg4510 mice, and the rigidity of the soluble form containing the mucin stalk facilitating the “anti”-inflammation. In contrast, hTau mice do not display as robust tau pathology as rTg4510 mice. Due to the complete lack of membrane-bound CX3CL1 in our hTau/Cx3cl1105Δ mice, CX3CL105Δ may not be as efficient and therefore leads to the downregulation of CX3CR1 and exacerbation of neuroinflammation/tau pathology. Together, these interpretations suggest that both membrane-bound CX3CL1 and the soluble form of fractalkine may make a concerted effort together to mediate both neuroinflammation and tau pathology.

Conclusions

Taken together, our data suggest that neuronal expression of only the chemokine domain of fractalkine fails to suppress tau-related pathological outcomes and microglial activation. Our data also suggest that fractalkine acts to tether microglia to neurons and, once this interaction is disrupted, microglia alter their functional phenotype. This signaling benefit is quickly negated under chronic pathological duress and offers little protection from cognitive deficits in advanced stages of the disease. In conclusion, the data presented here suggest that obligatory expression of the chemokine domain of CX3CL1 downregulates CX3CR1 levels on the microglial cell surface and consequently exacerbates tau pathology. These results could indicate the usefulness of potential therapeutics targeting ADAM10 or ADAM17, which cleave CX3CL1, to prevent the formation of excessive soluble CX3CL1 as a means to modify disease outcome for tauopathies.

Abbreviations

AAV:

Adeno-associated virus

AD:

Alzheimer’s disease

APPPS1:

Amyloid precursor protein presenilin-1

BAC:

Bacterial artificial chromosome

CX3CL1:

Fractalkine chemokine

Cx3cr1:

Fractalkine receptor

IHC:

Immunohistochemistry

IL-1:

Interleukin-1

LPS:

Lipopolysaccharide

MAPT:

Microtubule-associated protein tau

p38 MAPK:

P38 mitogen-activated protein kinase

TGFβ:

Transforming growth factor beta

TLR4:

Toll-like receptor-4

References

  1. Harrison JK, Jiang Y, Chen S, Xia Y, Maciejewski D, McNamara RK, Streit WJ, Salafranca MN, Adhikari S, Thompson DA, et al. Role for neuronally derived fractalkine in mediating interactions between neurons and CX3CR1-expressing microglia. Proc Natl Acad Sci U S A. 1998;95:10896–901.

    Article  CAS  Google Scholar 

  2. Imai T, Hieshima K, Haskell C, Baba M, Nagira M, Nishimura M, Kakizaki M, Takagi S, Nomiyama H, Schall TJ, Yoshie O. Identification and molecular characterization of fractalkine receptor CX3CR1, which mediates both leukocyte migration and adhesion. Cell. 1997;91:521–30.

    Article  CAS  Google Scholar 

  3. Bazan JF, Bacon KB, Hardiman G, Wang W, Soo K, Rossi D, Greaves DR, Zlotnik A, Schall TJ. A new class of membrane-bound chemokine with a CX3C motif. Nature. 1997;385:640–4.

    Article  CAS  Google Scholar 

  4. Garton KJ, Gough PJ, Blobel CP, Murphy G, Greaves DR, Dempsey PJ, Raines EW. Tumor necrosis factor-alpha-converting enzyme (ADAM17) mediates the cleavage and shedding of fractalkine (CX3CL1). J Biol Chem. 2001;276:37993–8001.

    CAS  PubMed  Google Scholar 

  5. Hundhausen C, Misztela D, Berkhout TA, Broadway N, Saftig P, Reiss K, Hartmann D, Fahrenholz F, Postina R, Matthews V, et al. The disintegrin-like metalloproteinase ADAM10 is involved in constitutive cleavage of CX3CL1 (fractalkine) and regulates CX3CL1-mediated cell-cell adhesion. Blood. 2003;102:1186–95.

    Article  CAS  Google Scholar 

  6. Ostuni MA, Guellec J, Hermand P, Durand P, Combadiere C, Pincet F, Deterre P. CX3CL1, a chemokine finely tuned to adhesion: critical roles of the stalk glycosylation and the membrane domain. Biol Open. 2014;3:1173–82.

    Article  Google Scholar 

  7. Cardona AE, Pioro EP, Sasse ME, Kostenko V, Cardona SM, Dijkstra IM, Huang D, Kidd G, Dombrowski S, Dutta R, et al. Control of microglial neurotoxicity by the fractalkine receptor. Nat Neurosci. 2006;9:917–24.

    Article  CAS  Google Scholar 

  8. Chung CW, Song YH, Kim IK, Yoon WJ, Ryu BR, Jo DG, Woo HN, Kwon YK, Kim HH, Gwag BJ, et al. Proapoptotic effects of tau cleavage product generated by caspase-3. Neurobiol Dis. 2001;8:162–72.

    Article  CAS  Google Scholar 

  9. Limatola C, Ransohoff RM. Modulating neurotoxicity through CX3CL1/CX3CR1 signaling. Front Cell Neurosci. 2014;8:229.

    Article  Google Scholar 

  10. Nash KR, Lee DC, Hunt JB Jr, Morganti JM, Selenica ML, Moran P, Reid P, Brownlow M, Guang-Yu Yang C, Savalia M, et al. Fractalkine overexpression suppresses tau pathology in a mouse model of tauopathy. Neurobiol Aging. 2013;34:1540–8.

    Article  CAS  Google Scholar 

  11. Poniatowski LA, Wojdasiewicz P, Krawczyk M, Szukiewicz D, Gasik R, Kubaszewski L, Kurkowska-Jastrzebska I. Analysis of the role of CX3CL1 (fractalkine) and its receptor CX3CR1 in traumatic brain and spinal cord injury: insight into recent advances in actions of neurochemokine agents. Mol Neurobiol. 2017;54:2167–88.

    Article  CAS  Google Scholar 

  12. Hashimoto D, Ohmuraya M, Hirota M, Yamamoto A, Suyama K, Ida S, Okumura Y, Takahashi E, Kido H, Araki K, et al. Involvement of autophagy in trypsinogen activation within the pancreatic acinar cells. J Cell Biol. 2008;181:1065–72.

    Article  CAS  Google Scholar 

  13. Barnett A, Brewer GJ. Autophagy in aging and Alzheimer’s disease: pathologic or protective? J Alzheimer’s Dis : JAD. 2011;25:385–94.

    Article  Google Scholar 

  14. Ghosh S, Wu MD, Shaftel SS, Kyrkanides S, Laferla FM, Olschowka JA, O’Banion MK. Sustained interleukin-1beta overexpression exacerbates tau pathology despite reduced amyloid burden in an Alzheimer’s mouse model. J Neurosci. 2013;33:5053–64.

    Article  CAS  Google Scholar 

  15. Corona AW, Huang Y, O’Connor JC, Dantzer R, Kelley KW, Popovich PG, Godbout JP. Fractalkine receptor (CX3CR1) deficiency sensitizes mice to the behavioral changes induced by lipopolysaccharide. J Neuroinflammation. 2010;7:93.

    Article  CAS  Google Scholar 

  16. Ashe KH, Zahs KR. Probing the biology of Alzheimer’s disease in mice. Neuron. 2010;66:631–45.

    Article  CAS  Google Scholar 

  17. Bhaskar K, Konerth M, Kokiko-Cochran O, Cardona A, Ransohoff R, Lamb B. Regulation of tau pathology by the microglial fractalkine receptor. Neuron. 2010;68:19–31.

    Article  CAS  Google Scholar 

  18. Lee S, Xu G, Jay TR, Bhatta S, Kim KW, Jung S, Landreth GE, Ransohoff RM, Lamb BT. Opposing effects of membrane-anchored CX3CL1 on amyloid and tau pathologies via the p38 MAPK pathway. J Neurosci. 2014;34:12538–46.

    Article  Google Scholar 

  19. Maphis N, Xu G, Kokiko-Cochran ON, Jiang S, Cardona A, Ransohoff RM, Lamb BT, Bhaskar K. Reactive microglia drive tau pathology and contribute to the spreading of pathological tau in the brain. Brain. 2015;138:1738–55.

    Article  Google Scholar 

  20. Cameron B, Landreth GE. Inflammation, microglia, and Alzheimer’s disease. Neurobiol Dis. 2010;37:503–9.

    Article  CAS  Google Scholar 

  21. Nash KR, Moran P, Finneran D, Hudson C, Robinson J, Morgan D, Bickford PC. Fractalkine over expression suppresses alpha-synuclein mediated neurodegeneration. Mol Ther. 2015;23(1):17-23.

  22. Kim KW, Vallon-Eberhard A, Zigmond E, Farache J, Shezen E, Shakhar G, Ludwig A, Lira SA, Jung S. In vivo structure/function and expression analysis of the CX3C chemokine fractalkine. Blood. 2011;118:e156–67.

    Article  CAS  Google Scholar 

  23. Andorfer C, Kress Y, Espinoza M, De Silva R, Tucker KL, Barde Y-A, Duff K, Davies P. Hyperphosphorylation and aggregation of tau in mice expressing normal human tau isoforms. J Neurochem. 2003;86:582–90.

    Article  CAS  Google Scholar 

  24. Soltys Z, Ziaja M, Pawlinski R, Setkowicz Z, Janeczko K. Morphology of reactive microglia in the injured cerebral cortex. Fractal analysis and complementary quantitative methods. J Neurosci Res. 2001;63:90–7.

    Article  CAS  Google Scholar 

  25. Varvel NH, Bhaskar K, Kounnas MZ, Wagner SL, Yang Y, Lamb BT, Herrup K. NSAIDs prevent, but do not reverse, neuronal cell cycle reentry in a mouse model of Alzheimer disease. J Clin Invest. 2009;119:3692–702.

    Article  CAS  Google Scholar 

  26. Kokiko-Cochran O, Ransohoff L, Veenstra M, Lee S, Saber M, Sikora M, Teknipp R, Xu G, Bemiller S, Wilson G, et al. Altered neuroinflammation and behavior after traumatic brain injury in a mouse model of Alzheimer’s disease. J Neurotrauma. 2016;33(7):625-40.

    Article  Google Scholar 

  27. Hughes RN. The value of spontaneous alternation behavior (SAB) as a test of retention in pharmacological investigations of memory. Neurosci Biobehav Rev. 2004;28:497–505.

    Article  CAS  Google Scholar 

  28. Bergmann CC, Altman JD, Hinton D, Stohlman SA. Inverted immunodominance and impaired cytolytic function of CD8+ T cells during viral persistence in the central nervous system. J Immunol. 1999;163:3379–87.

    CAS  PubMed  Google Scholar 

  29. Maphis N, Xu G, Kokiko-Cochran ON, Cardona AE, Ransohoff RM, Lamb BT, Bhaskar K. Loss of tau rescues inflammation-mediated neurodegeneration. Front Neurosci. 2015;9:196.

    Article  Google Scholar 

  30. Lee S, Varvel N, Konerth M, Xu G, Cardona A, Ransohoff R, Lamb B. CX3CR1 deficiency alters microglial activation and reduces beta-amyloid deposition in two Alzheimer’s disease mouse models. Am J Pathol. 2010;177:2549–62.

    Article  CAS  Google Scholar 

  31. Wynne AM, Henry CJ, Huang Y, Cleland A, Godbout JP. Protracted downregulation of CX3CR1 on microglia of aged mice after lipopolysaccharide challenge. Brain Behav Immun. 2010;24:1190–201.

    Article  CAS  Google Scholar 

  32. Pachot A, Cazalis MA, Venet F, Turrel F, Faudot C, Voirin N, Diasparra J, Bourgoin N, Poitevin F, Mougin B, et al. Decreased expression of the fractalkine receptor CX3CR1 on circulating monocytes as new feature of sepsis-induced immunosuppression. J Immunol. 2008;180:6421–9.

    Article  CAS  Google Scholar 

  33. Daniel AR, Gaviglio AL, Knutson TP, Ostrander JH, D’Assoro AB, Ravindranathan P, Peng Y, Raj GV, Yee D, Lange CA. Progesterone receptor-B enhances estrogen responsiveness of breast cancer cells via scaffolding PELP1- and estrogen receptor-containing transcription complexes. Oncogene. 2015;34:506–15.

    Article  CAS  Google Scholar 

  34. Fujita M, Takada YK, Takada Y. The chemokine fractalkine can activate integrins without CX3CR1 through direct binding to a ligand-binding site distinct from the classical RGD-binding site. PLoS One. 2014;9:e96372.

    Article  Google Scholar 

  35. Polydoro M, Acker CM, Duff K, Castillo PE, Davies P. Age-dependent impairment of cognitive and synaptic function in the htau mouse model of tau pathology. J Neurosci. 2009;29:10741–9.

    Article  CAS  Google Scholar 

  36. Fong AM, Erickson HP, Zachariah JP, Poon S, Schamberg NJ, Imai T, Patel DD. Ultrastructure and function of the fractalkine mucin domain in CX(3)C chemokine domain presentation. J Biol Chem. 2000;275:3781–6.

    Article  CAS  Google Scholar 

Download references

Acknowledgements

We would like to acknowledge the Cleveland Clinic and University New Mexico Flow Cytometry, Imaging, and Behavioral Cores for their continued valuable assistance in acquiring and analyzing the data. We would like to thank Dr. Richard Ransohoff for his invaluable input in preparing this manuscript. We also like to acknowledge Dr. Surojit Paul and Mr. Sathyanarayanan Rajagopal for their assistance in performing the CX3CL1 ELISAs.

Funding

This work was supported by a BrightFocus Foundation grant (to BTL), a National Research Service Award F31 # 1F31AG050409-01A1 and The Stark Neurosciences Research Institute, Eli Lilly and Company, and by the Indiana Clinical and Translational Sciences Institute, funded in part by grant # UL1TR001108 from the National Institutes of Health, National Center for Advancing Translational Sciences (to SMB), Multi-Center Program Grant (to BTL) and an Investigator Initiated Research Grant (to BTL) from the Alzheimer’s Association, NIH/NIA R01 AG023012 (to BTL and RF1 AG051495 (to BTL), NIH/NINDS R01 NS047804 (to BTL), NIH/NEI R01EY022358 (to SDC), Department of Defense grant W81XWH-13 MRPRA-CSRA (to BTL), the Jane and Lee Seidman Fund and a generous donation from Chet and Jane Scholtz. NIH SC1GM095426 and R01NS078501 to AEC, Additional support from Bright Focus Foundation Pilot Award AHAF0311, Alzheimer’s Association NIRG-11-204995, NIH/NINDS R21NS077089, R21 NS093442, and R01NS083704 (to KB). UNM Molecular Genetics Microbiology intradepartmental grant support, UNM Health Sciences Center bridge funding (KB), and Stephanie Ruby travel award (NM).

Availability of data and materials

The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Author information

Authors and Affiliations

Authors

Contributions

SMB, NM, AEC, SJ, BTL, and KB contributed to the conceptualization. SMB, NM, JLC, BTL, AEC, GX, SJ, and KB contributed to the methodology. SMB, NM, GNW, KB, JC, and OKC contributed to the formal analysis. SMB, SVF, NM, CMM, KWK, GNW, and OKC contributed to the investigation. SMB, NM, and KB contributed to the writing of the original draft. SMB, NM, SDC, JLC, AEC, SJ, BTL, and KB contributed to the writing, review, and editing. SMB, KB, BTL, AEC, and SDC contributed to the funding acquisition. SDC, SJ, JLC, and GX contributed to the resources. All authors read and approved the final manuscript.

Corresponding author

Correspondence to Kiran Bhaskar.

Ethics declarations

Ethics approval

All animal experiments performed in this study were in accordance with the ethical standards of the Institutional Animal Care and Use Committee at the Cleveland Clinic Foundation and the University of New Mexico.

Consent for publication

Not applicable

Competing interests

The authors declare that they have no competing interests.

Publisher’s Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Additional file

Additional file 1:

Figure S1. Altered expression of Cx3cl1 and Cx3cr1 and protein levels of CX3CL1 in non-transgenic, Cx3cl1105Δ and. Cx3cl1−/− mice. Quantitative real-time PCR analysis showing the expression of Cx3cl1 and Cx3cr1 mRNAs (A and C), as well as ELISA analysis for soluble CX3CL1 levels in non-transgenic (Non-Tg), Cx3cl1−/− and Cx3cl1105Δ mice (B). (PDF 371 kb)

Rights and permissions

Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Bemiller, S.M., Maphis, N.M., Formica, S.V. et al. Genetically enhancing the expression of chemokine domain of CX3CL1 fails to prevent tau pathology in mouse models of tauopathy. J Neuroinflammation 15, 278 (2018). https://0-doi-org.brum.beds.ac.uk/10.1186/s12974-018-1310-6

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: https://0-doi-org.brum.beds.ac.uk/10.1186/s12974-018-1310-6

Keywords