Background
Neural injury associated with virus infection represents a multifaceted convergence of host–pathogen interactions that range from direct lytic killing of infected neurons to bystander pathology mediated by brain-infiltrating immune cells responding to chemotactic and inflammatory cues. Viral encephalitis is a trade-off between the need to clear pathogen from the brain and the need to preserve irreplaceable neurons and neural circuits: too little inflammation and the host dies of uncontrolled infection, too much inflammation and the host suffers permanent brain damage [
1]. And while much attention is rightly given to viral encephalitides associated with human mortality, there is likely a significant component of neural injury associated with low-level, sub-clinical viral infections of the central nervous system that are ultimately cleared by the host. Theiler’s murine encephalomyelitis virus (TMEV) is a model of such an infection [
2]. When C57Bl/6 mice are inoculated via intracranial delivery of the Daniel’s strain of TMEV, there is an acute viral encephalitis that culminates in the generation of an antiviral T cell-mediated response, development of virus neutralizing antibodies, clearance of the virus, and resolution of brain inflammation over the course of about 45 days [
3]. However, despite essentially complete resolution of the infection, permanent neurological sequelae such as impaired spatial learning [
2,
4], anxiety [
5], and epilepsy [
6] occur in most postinfectious animals. Uniquely, these neurologic problems largely stem from bystander loss of CA1 pyramidal neurons and subsequent disruption of hippocampal and hippocampal-cortical circuits [
4,
7,
8].
Our previous studies identified brain infiltration of inflammatory monocytes, a population defined as CD45
hiCD11b
+Gr1
+1A8
− cells, as the primary driver of hippocampal pathology during acute TMEV infection [
4,
8]. We showed that animals which mount a large inflammatory monocyte response exhibit extensive loss of CA1 neurons in the dorsal hippocampus and lose the ability to learn spatial navigation and novel object recognition tasks. In contrast, mice that mount a weak inflammatory monocyte response exhibit preservation of CA1 neurons and maintain cognitive performance, despite robust virus infection [
8]. In parallel, others have shown that monocyte-derived inflammatory factors such as interleukin-6 [
9,
10] and tumor necrosis factor-α [
11] drive ictogenesis in the TMEV model. These observations suggest that modulation of inflammatory monocyte responses during acute virus infection in the brain may confer neuroprotection. However, the specific mechanisms responsible for the recruitment of inflammatory monocytes to the brain in the TMEV model have not been previously characterized and open questions remain regarding the mechanisms of leukocyte infiltration into the brain in general.
The multistep process of leukocyte entry into the central nervous system is predominantly controlled by chemokines [
12]. In particular, inflammatory monocyte trafficking is thought to depend upon C-C motif chemokine receptor type 2 (CCR2) signaling in response to C-C motif chemokine ligand 2 (CCL2) [
13], though the specific details of this dependency vary with viral pathogen. For example, CCR2-deficient mice exhibited reduced monocyte recruitment to the brain but increased mortality during West Nile virus encephalitis [
14] and this effect was differentially regulated by CCL2 and CCL7, another CCR2 ligand [
15]. In a model of Japanese encephalitis virus infection, mice deficient in CCR2 had reduced monocyte infiltration and reduced mortality, whereas CCL2-deficient animals exhibited increased monocyte infiltration and increased mortality [
16]. Both CCL2- and CCR2-deficient mice with mouse hepatitis virus encephalitis had reduced monocyte infiltration, but only CCR2-deficient mice exhibited increased mortality [
17,
18]. Likewise, mice with CCR2-deficient hematopoietic cells mounted a reduced monocyte response to herpes simplex virus 1 infection and exhibited increased mortality [
19]. These studies clearly support a role for the CCR2:CCL2 axis in trafficking of inflammatory monocytes to the brain during viral encephalitis. However, the specific cellular source of the CCL2 is not identified in any of this work. In the current study, we demonstrate that inflammatory monocyte infiltration into the brain during acute TMEV infection requires CCR2 and that neurons are a key source of CCL2 driving this trafficking during the earliest stages of infection.
Methods
Mice
C57BL/6J (B6; #000664), Tg(Syn1-cre)671Jxm (Syn-Cre; #003966), Ccl2
tm1.1Pame (Ccl2-RFP
fl/fl; #016849), and Ccr2
tm1Ifc (CCR2
−/−; #004999) mice were acquired from The Jackson Laboratories (Bar Harbor, ME). Mice were acclimatized for at least 1 week following shipment and prior to use or were bred in-house. Female mice between 4 and 6 weeks of age were used for all experiments. LysM:eGFP mice were maintained in-house, as described [
4]. The CCR2-deficient mice were maintained as homozygotes; wildtype B6 mice were used as CCR2-sufficient controls for all experiments involving the CCR2
−/− mice. A Ccl2-RFP
flx/fl x Syn-Cre conditional knockout model was generated by interbreeding Ccl2-RFP
fl/fl and Syn-Cre mice. Tail DNA was screened by PCR using primers to detect the mutant CCL2 allele (forward: 5′-AGGACGGCGAGT TCATCTAC-3′; reverse: 5′-TGGTGTAGTCCTCGTTGTGG-3′; 288 bp product) and the wildtype CCL2 allele (forward: 5′- AACCACCTCAAGCACTTCTG-3′; reverse: 5’GCTTTGCAGTTTCCCTCAAG-3′; 363 bp product). PCR cycling conditions were 94 °C for 2 min, followed by 10 cycles of 94 °C for 20 s, 65 °C for 15 s (with − 0.5 °C decrease per cycle), and 68 °C for 10 s, followed by an additional 28 cycles of 94 °C for 15 s, 60 °C for 15 s, and 72 °C for 10 s, ending with 72 °C for 2 min, and 10 °C hold. Ccl2-RFP
fl/fl x Syn-Cre F2+ crosses were screened until all pups were homozygous for the mutant CCL2 allele. The Cre transgene was screened in all pups using generic Cre primers (forward: 5′-ACCAGCCAGCTATCAACTCG-3′; reverse: 5′-TTACATTGTCCAGCCACC-3′; 200 bp product). PCR cycling conditions were 94 °C for 5 min, followed by 35 cycles of 94 °C for 1 min, 60 °C for 1.5 min, and 72 °C for 2.5 min, ending with 72 °C for 5 min, and 4 °C hold. This PCR cannot distinguish heterozygotes from homozygotes; therefore, the line was maintained for Cre hemizygosity and Ccl2-RFP homozygosity and all litters were screened using a Transnetyx (Cordova, TN) panel that detects wildtype CCL2, RFP, and Cre. For experiments, mice were considered deficient in neuronal production of CCL2 (“CCL2−”) if they were positive for the Cre transgene, negative for the wildtype CCL2 allele, and positive for RFP. Mice were considered normal for neuronal production of CCL2 (“CCL2+”) if they were negative for the Cre transgene and positive either for the wildtype CCL2 allele or for RFP. In early experiments, the CCL2+ animals were selected as littermate controls for the CCL2− mice; in later experiments, the line was established such that all pups expressed the Cre transgene (presumptive homozygotes). All mice were group housed under controlled temperature and humidity with a 12-h light/dark cycle with ad libitum access to food and water. All animal experiments were performed according to the National Institutes of Health guidelines and were approved by the Mayo Clinic Institutional Animal Care and Use Committee (Animal Welfare Assurance number A3291-01).
Virus and infection
At 4–6 weeks of age, mice were infected by intracranial injection of 2 × 10
5 PFU of the Daniel’s strain of TMEV in 10 μL DMEM, prepared as previously described [
20]. Sham-infected mice received intracranial injection of 10 μL virus-free DMEM (Cellgro; Herndon, VA). In some experiments, virus was inactivated by exposure to short-wavelength UVc light (254 nm) for 15 min at room temperature immediately prior to inoculation.
RNA and microarray
RNA was isolated from the whole brain of mice following perfusion with ice cold PBS using the RNeasy Lipid Tissue Midi Kit (Qiagen, Valencia, CA). RNA integrity, purity, and concentration were assessed using an RNA Analysis kit (Agilent Technologies, Santa Clara, CA). Samples passing quality control were analyzed on Illumina mouse WG-6 v 2.0 expression BeadChips in the Mayo Clinic Medical Genome Facility Gene Expression Core. Expression data were analyzed using Excel and MATLAB, where fold change was calculated and converted to log2. Heatmaps and hierarchical clusters were derived using Gitools v2.2.2.
Cytokine measurements
Following TMEV infection, serum and whole brain or hippocampal homogenates were collected, clarified, and stored at − 80 °C until analysis. Mouse CCL2 was detected using the Quantikine ELISA kit (R&D Systems, Minneapolis, MN) following the manufacturer’s instructions. For each hippocampus and whole brain sample, 50 μL of neat homogenate was measured in duplicate. Serum samples were diluted twofold and measured in duplicate. Values were calculated from a standard curve included in every assay. Serum values are reported as per mL; tissue homogenate values were back-calculated to the total amount present in the entire animal.
Immunostaining and microscopy
Terminally anesthetized mice (isoflurane overdose) were perfused with 50 mL of 4% paraformaldehyde (PFA) via intracardiac puncture, and tissues were postfixed in 4% PFA for 24 h. Brain tissue was macrosectioned using a brain matrix, making cuts through the optic chiasm and infundibulum. The resulting tissue block containing the dorsal hippocampus was embedded in 4% agarose gel and sectioned at 70-μm thickness by vibratome. Free-floating sections were blocked in PBS containing 1% BSA, 10% normal donkey serum, 1% FBS, and 0.1% Triton-X 100 for 1 h, incubated overnight at 4 °C with primary antibody (anti-MCP-1/CCL2: Cell Sciences, CPM001, 1:400 in block), incubated with secondary antibody for 1 h (donkey anti-rabbit Cy3: Jackson ImmunoResearch, 711-166-152, 1:1000 in block), washed, and mounted in DAPI-containing mountant on charged slides. CCL2 immunoreactivity was imaged with the Zeiss AxioObserver.Z1 and ApoTome.2 structured illumination system (Carl Zeiss Microscopy GmbH, Jena, Germany) using a ×20 objective (LD Plan-Neofluar 20x/0.4 Korr Ph 2 M27, 0.55 NA), 538–562-nm bandpass excitation, 570–640-nm bandpass emission, 7-μm optical thickness, and 300-ms exposure time. For detection of the mCherry fluorophore in CCL2:RFP animals, free-floating 70-μm sections were mounted on charged slides in DAPI-containing mountant. Fluorescence images were captured with a laser scanning confocal microscope (LSM780, Carl Zeiss Microscopy GmbH, Jena, Germany) using a ×40 objective (C-Apochromat 40x/1.20 W Korr FCS M27, 1.2 NA) with water as the refractive medium. Validation of mCherry-specific emission and exclusion of autofluorescence was obtained with spectral imaging using a lambda scan at 488, 561, and 594 nm excitation and 8-nm-stepped emission spectra. Forty-micrometer z-stacks were acquired with a step thickness of 2 μm, pixel dwell time of 0.39 μs, and pinhole equivalent to 1 airy unit across all samples. Uncompressed TIFF-images were exported from Zen software (Zen Black 2012 64-bit, Carl Zeiss Microscopy GmbH, Jena, Germany) and post-processed in ImageJ (ImageJ v1.50b, Wayne Rasband, National Institutes of Health, USA) and Photoshop (Adobe Photoshop CC, 2014 Release, 64-bit). Levels were normalized, when appropriate, equally across images and groups; gamma values were not changed.
Isolation of brain-infiltrating leukocytes (BILs)
BILs were isolated as previously described [
20]. Briefly, following cardiac perfusion with 50 mL PBS, leukocytes were isolated from Dounce-homogenized whole brain tissue using a 30% Percoll gradient centrifuged at 7800
g
ave for 30 min at RT in a Beckman F0630 rotor. The floating myelin layer was removed, and the leukocytes were collected, strained at 40 μm via gravity, diluted in 50 mL PBS, and centrifuged at 600
g for 5 min at RT in a Beckman SX4250 rotor. The leukocyte pellet was resuspended in 1 mL PBS and underlaid with 1 mL of 1.100 g/mL Percoll to enrich for monocytes and neutrophils. With subsequent centrifugation at 800
g for 20 min at RT in a Beckman SX4250 rotor without brake, the mononuclear leukocytes were collected at the gradient interface, washed in PBS, and resuspended in flow cytometry buffer containing 1% bovine serum albumin and 0.02% sodium azide in PBS.
Flow cytometric phenotyping
BILs were incubated with 0.5 μg Fc block (anti-FcyRIII/II mAb) prepared from the supernatant of 2.4G2 hybridoma cells for 30 min on ice, followed by staining with CD45 (clone 30-F11, BD Biosciences), CD11b (clone M1/70, BD Biosciences), Ly6G/C (clone RB6-8C5, BD Biosciences), or Ly6G (clone 1A8, BD Biosciences). All antibodies were added to blocked wells at 1:200, incubated for 30 min, and washed three times prior to flow cytometric analysis. Brain-infiltrating cells were gated on CD45 expression. All CD45mid/hi cells were further assessed for expression of CD11b, Ly-6C/G (Gr1), and Ly-6G (1A8). We defined inflammatory monocytes as the CD45hiCD11b+Gr1++1A8− population, neutrophils as CD45hiCD11b++Gr1+1A8+ cells, and microglia as CD45midCD11bmidGr1− cells. For phenotyping experiments involving reporter LysM:eGFP mice, we defined inflammatory monocytes as GFPmid cells, neutrophils as GFPhi cells, and microglia as GFPneg cells within a specific forward- and side-scatter gate. Flow cytometric analysis was performed on an Accuri C6 flow cytometer with sampler arm (BD Biosciences, Mountain View, CA). Files were analyzed offline using FlowJo 10.08 (Windows version; FlowJo LLC, Ashland, OR).
Statistics and data analysis
α = 0.05 and β = 0.2 were established a priori. Post hoc power analysis was performed for all experiments, and significance was only considered when power ≥ 0.8. Statistical analyses were performed in JMP Pro 12 (SAS Institute Inc., Cary, NC). Normality was determined by the Shapiro–Wilk test and normally distributed data were checked for equal variance. Parametric tests were only applied to data that were both normally distributed and of equal variance. Dunnett’s method for pairwise comparison was used for all post hoc sequential comparisons following one-way ANOVA; the Tukey–Kramer test was used for two-way ANOVA pairwise comparisons. Error bars in all graphs are 95% confidence intervals.
Discussion
Identification of immune cells and effector molecules that contribute to brain injury is critical to the development of neuroprotective and neuroreparative strategies [
7]. We and others have used the Theiler’s murine encephalomyelitis virus model to investigate immune-mediated mechanisms of brain injury within the context of acute CNS viral infection [
3,
6]. Despite differences in nomenclature, host genetics, and details of the viral strain and inoculum, a general consensus has emerged that monocytic cells infiltrate the brain within hours of infection and create an inflammatory environment that kills neurons and alters neural circuitry. Understanding the mechanisms that drive monocyte infiltration may therefore reveal novel strategies to protect the brain.
Not surprisingly, intracranial inoculation with TMEV induced a robust inflammatory program in the hippocampus, involving upregulation of dozens of chemokine, cytokine, and adhesion factor genes (Fig.
1). Somewhat surprising, however, was the speed with which this induction occurred. We reproducibly detected large changes in transcription by 3 hpi, and in some experiments measured variable, but sizable, induction of chemokines and cytokines by 1 hpi (not shown). Moreover, these factors were detected at the protein level over the same acute period, with CCL2 measurable in the serum and brain as early as 1 hpi (not shown). Not only are these responses fast with regard to biosynthesis, but it is unlikely that a significant number of cells have even been infected in the brain by 3 hpi. Mathematical modeling constrains the infection rate to only 0.01 to 0.1 cells per minute [
26], suggesting that fewer than 20 cells would be productively infected by 3 hpi. Even if this rate is off by a factor of 100, fewer than 2000 cells are infected by 3 hpi. Obviously, some cells can respond to virions and virus constituents via mechanisms that do not require active infection, such as pathogen (or pattern) recognition receptors [
27], but our findings indicate that inoculation with UV-inactivated virus did not elicit CCL2 production in the brain at 3 hpi. This finding also rules out induction via non-specific trauma-induced effects associated with intracranial inoculation. Moreover, even assuming a direct effect of each active virion on target cells, we only inoculated the animals with 200,000 plaque-forming units and these were not introduced directly into the hippocampus [
20]. Yet, at 6 hpi the hippocampus produced essentially all of the CCL2 measured in the brain and presumably contributed the majority of serum CCL2. This issue is further confounded by the pattern of CCL2 expression revealed by immunostaining, which shows that effectively every neuron in dorsal hippocampal CA1, CA3, and dentate gyrus expressed this factor at 6 hpi (Fig.
2). As we have previously reported, even by 3 dpi, only a fraction of CA1 neurons are directly infected with TMEV and DG neurons are never positive for virus by immunostaining [
22]. These observations and discrepancies suggest that a currently unidentified amplification event occurs almost immediately after inoculation with live virus that results in widespread, albeit tissue-specific, upregulation of CCL2 production. Furthermore, this induction occurs almost exclusively in neurons, as synapsin-promoter driven deletion of CCL2 results in nearly complete suppression of both hippocampal and serum CCL2 at 6 hpi (Fig.
5).
Despite the gap in our knowledge about the pathway between virus inoculation and CCL2 induction, our data clearly support a model in which neurons are the primary source of this chemokine during the most acute phase of the host response. Notably, deletion of CCL2 from neurons recapitulates the effect of systemic deletion in mice treated with anti-CCL2 immunoglobulin (Fig.
6f compared to Fig.
4d)—the inflammatory monocyte infiltrate is reduced by about 70% in both conditions at 18–24 hpi. Presumably, the remaining infiltrate in both experiments is CCL7-dependent, although the cellular source of this other CCR2 ligand is currently unknown. These findings suggest an intriguing maladaptive response: neurons respond to brain inoculation with TMEV by producing copious amounts of a chemokine that serves to recruit the inflammatory monocytes into the brain that ultimately kill those same neurons [
4]. While the current study does not address the impact of a curtailed monocyte response on viral clearance or eventual lymphocyte recruitment, anecdotal evidence indicates that CCR2
−/− mice do not succumb to lethal infection and, indeed, do not show any apparent adverse effects at later timepoints (out to several months). Studies assessing the impact of CCR2 deletion on hippocampal neuropathology are ongoing.
Maintaining the focus on the most acute response to TMEV infection rather than downstream sequelae, several notable findings arise from the current study. First, genetic deletion of CCR2 in all cells almost entirely abrogates inflammatory monocyte infiltration at 18 hpi but actually increases neutrophil infiltration at this timepoint. This suggests that in the context of life-long absence of CCL2:CCR2 signaling, neutrophil responses are not compromised by the absence of an inflammatory monocyte response. Second, acute systemic immunodepletion of CCL2 reduces inflammatory monocyte infiltration by 70% but also reduces neutrophil infiltration by more than half. This suggests that acute inhibition of the CCL2:CCR2 axis either impacts neutrophil recruitment directly (for example, brain-infiltrating neutrophils express CCR2 and respond to CCL2) or via an indirect effect on monocyte-to-neutrophil communication (for example, infiltrating monocytes release neutrophilic chemokines) [
28]. Third, systemic immunodepletion of CCL7 blocks about 30% of monocyte infiltration but has no impact on neutrophils. This indicates that if brain-infiltrating neutrophils use a CCR2-dependent trafficking pathway, it is not responsive to CCL7 or that the CCL7-dependent brain-infiltrating monocytes are not responsible for creating the pro-neutrophil environment. While speculative, this may suggest that there are at least two functional subtypes of CCR2
+ brain-infiltrating monocytes that can be distinguished by CCL2 vs CCL7 responsiveness. Fourth, the Ccl2-RFP
fl/fl reporter line exhibits a trend toward increased neutrophil infiltration as compared to B6 mice. While not statistically significant due to the spread in values, this increase was visually evident in most of the mice, suggesting caution in studies analyzing leukocytic responses in these animals. Fifth, the robust inhibition of monocyte infiltration in the neuron-specific CCL2-deficient mice argues explicitly that this chemokine drives trafficking to the brain while side-stepping all of the issues regarding monocytopenia that arise in CCL2
−/− or CCR2
−/− mice. This is in contrast to the reported role of CCR2 in West Nile virus encephalitis [
14] but is consistent with the brain trafficking role described by Graham and colleagues in a Semliki Forest virus model [
29]. Sixth, manipulation of either CCL2 or CCR2 drove the inflammatory monocyte response in the same direction—down. This is in contrast to observations in a Japanese encephalitis model, in which CCR2
−/− mice had reduced monocyte infiltration while CCL2
−/− mice had a paradoxical increase in this population in the brain [
16]. Of note, however, our manipulation of CCL2 was via acute immunodepletion, not life-long genetic deletion.
The production of CCL2 by neurons during acute encephalitis may play a role in more than just leukocyte recruitment. For example, hippocampal neurons express CCR2 [
21] and CCL2 directly enhances both NMDA receptor- and AMPA receptor-mediated excitatory postsynaptic potentials [
30]. This suggests that the earliest stages of hippocampal circuit dysregulation associated with acute TMEV infection may be triggered locally by neuronal CCL2 acting in an autocrine fashion [
31,
32]. In this context, it is notable that CCL2 immunoreactivity is strongly upregulated in CA1 neuron apical dendrite tufts located in the stratum lacunosum moleculare at 6 h after infection (Fig.
2). This layer is an important site of integration between the entorhinal cortex and CA1 pyramidal neurons (the perforant pathway) and is of fundamental importance to spatial and episodic memory formation [
33]. It is also the site of GABAergic interneurons that are strongly activated by both entorhinal cortex and CA3-derived Schaffer collateral inputs [
34]. These interneurons exhibit both AMPA receptor and NMDA receptor excitatory postsynaptic potentials [
35] and are maximally activated at the theta oscillation peak and at the gamma oscillation trough [
36]. These cells appear to mediate a robust feedforward inhibition that controls the size and timing of excitatory inputs onto CA1 neurons [
37] and synchronizes network firing to theta frequency [
34,
36]. Given the important role for hippocampal theta oscillations in learning and memory [
38], it is likely that acute production of CCL2 in the stratum lacunosum moleculare would disrupt cognitive performance. It is also notable that neuronal CCL2 production has been observed following kainic acid- [
39] and pilocarpine-mediated seizure induction [
40] in rodents and both CCL2 and CCR2 are increased in tissue resected from humans with intractable epilepsy [
41,
42]. Inhibition of either CCL2 production or CCR2 signaling suppressed seizures in a mouse model of systemic inflammation and mesial temporal lobe epilepsy [
43]. Indeed, Caleo and colleagues have recently suggested that CCL2 “may act as a master regulator of inflammatory processes in the epileptic brain by both directly promoting hyperexcitability and regulating the activity of downstream inflammatory effectors” [
43]. Our findings echo this concept and add a third role in which hyperacute neuronal CCL2 production mediates CCR2-dependent inflammatory monocyte trafficking into the brain, resulting in the creation of an environment that further disrupts neural function and induces pyramidal neuron death [
4,
7].