Complement-dependent synapse loss and microgliosis in a mouse model of multiple sclerosis

Abstract

Multiple sclerosis (MS) is an inflammatory, neurodegenerative disease of the CNS characterized by both grey and white matter injury. Microglial activation and a reduction in synaptic density are key features of grey matter pathology that can be modeled with MOG_35–55 experimental autoimmune encephalomyelitis (EAE). Complement deposition combined with microglial engulfment has been shown during normal development and in disease as a mechanism for pruning synapses. We tested whether there is excess complement production in the EAE hippocampus and whether complement-dependent synapse loss is a source of degeneration in EAE using C1qa and C3 knockout mice. We found that C1q and C3 protein and mRNA levels were elevated in EAE mice. Genetic loss of C3 protected mice from EAE-induced synapse loss, reduced microglial activation, decreased the severity of the EAE clinical score, and protected memory/freezing behavior after contextual fear conditioning. C1qa KO mice with EAE showed little to no change on these measurements compared to WT EAE mice. Thus, pathologic expression and activation of the early complement pathway, specifically at the level of C3, contributes to hippocampal grey matter pathology in the EAE.

Introduction:

Multiple sclerosis is an immune-mediated disease of the CNS involving damage to both the white and grey matter. Current immunomodulatory therapies for MS are effective at minimizing relapses and white matter lesion burden, but are less effective at preventing progressive neurodegeneration and cognitive deficits [1]. Grey matter degeneration is highly correlated with the progressive component of MS and its accompanying physical and cognitive disabilities [2–5]. Grey matter pathology involves focal demyelinated lesions and widespread atrophy of cortical and subcortical regions. MS demyelinated grey matter is characterized by activated microglia, axon damage, neuron and glial cell loss, and decreased synaptic density [6–8]. Interestingly, activated microglia and synapse loss also occur in myelinated areas of normal-appearing grey matter [9, 10] suggesting direct synaptic injury associated with activated microglia may be involved.

Microglia have been shown to prune synapses by phagocytosis during development by a process of complement-mediated opsonization and subsequent phagocytosis [11–14]. The complement system can be activated by deposition of C1q, which triggers the classical complement pathway, or via the alternative and lectin pathways that are independent of C1q. All three initiating pathways converge to cleave the alpha chain of C3 to C3a, a soluble chemokine that recruits phagocytic cells, and C3b, which is deposited on activating surfaces as an opsonin that facilitates phagocytosis via the microglia-specific complement receptor 3 (CR3) [11, 15]. C3b can also initiate the lytic pathway leading to formation of the membrane attack complex (MAC) that causes cell lysis. However, it is the early complement pathway (C1q - C3) that is implicated in synapse pruning by microglia [11, 16, 17].

While critical for development, complement is also recruited to mediate pathologic synapse loss by microglia during aging as well as in diseases including frontotemporal dementia, Alzheimer’s disease, viral encephalitis, traumatic brain injury, and schizophrenia [16–22]. Yet, microglial activation and synapse loss may occur independent of complement in other diseases such as the mutant SOD1 model of ALS and a model of HIV associated neurocognitive disorder (HAND) [23, 24]. Thus, the goal of this study was to evaluate whether complement-dependent synapse loss contributes to grey matter degeneration in MOG_35–55 EAE, which may also provide insight into its role in MS.

Complement involvement in the white matter pathology of MS is well-established. Complement proteins and activation products including the terminal MAC complex are abundant in active and inactive white matter lesions of MS patients [25, 26]. In the MOG_35–55 EAE mouse model, knockout of complement components C3, Factor B, or receptors to C3 cleavage products (C3aR, CR3, and CR4) all reduced white matter lesion pathology and motor impairment measured by the EAE clinical score [27–32], showing that the complement pathway plays a major role in driving spinal cord pathology in EAE. Evaluation of the role of the terminal lytic pathway in MOG_35–55 EAE severity has yielded mixed results [33–38]. In general, other EAE-inducing mouse models such as adoptive-transfer EAE, MP4 EAE, PLP EAE, spinal cord homogenate induced EAE, and autoantibody EAE, as well as toxin-induced demyelination models such as cuprizone all show significant dependence on complement for disease severity and spinal cord pathology, although differences for specific complement proteins do exist [38–43].

The role of complement in MS grey matter degeneration has only begun to be addressed. Recent studies on human postmortem samples indicate that early complement components C1q, C4d, Bb, C3b-iC3b, C3d, as well as the terminal MAC do accumulate in cortical, hippocampal and thalamic grey matter of MS patients [9, 25, 44]. Within the hippocampus, C1q and C3d were increased in both myelinated and demyelinated regions of CA1–3 that also displayed synapse loss and microglial activation. C1q and C3d also co-localized with synaptic markers within microglial processes, suggesting that complement-opsonized synapses were potentially phagocytized by microglia [9]. As the early complement pathways are correlated with the synapse reduction observed in MS patients, it is important to determine whether the relationship is causative and thus potentially preventable using animal models.

MOG_35–55-induced EAE in C57BL/6 mice provides a good animal model of several aspects of MS grey matter injury in the brain. The hippocampus of EAE mice have pronounced microglial activation, synapse loss, and some atrophy of pyramidal cell layers that occurs independently of extensive local myelin loss, axon injury or significant accumulation of peripheral immune cells [45–50]. Chronic EAE mice, 66+ days post immunization, exhibit significant reductions in brain volume including atrophy of the hippocampus [51]. EAE mice also show impaired synaptic transmission and deficits in hippocampal-dependent behavior [46, 52–54], indicating that EAE mice have cognitive impairments analogous to progressive MS patients.

We have used the MOG_35–55 EAE model to test the hypothesis that the EAE inflammatory environment results in excess complement production and activation in the brain that contributes to synapse elimination, microglial activation, and impaired learning and memory.

Methods:

Animals:

Male and female C57BL/6 mice were obtained from Jackson Laboratories (Bar Harbor, ME) at age 7 weeks and were housed for at least 1 week prior to EAE induction. A C3 knockout breeder pair (on a C57BL/6 background) was obtained from Jackson (Stock # 029661; [55]). A C1qa knockout (KO) breeder pair was obtained from Dr. Andrea Tenner (University of California, Irvine; [56]). C1q is an oligomeric protein complex originating from 3 genes: C1qa, C1qb, and C1qc. Loss of C1qa prevents correct folding and secretion of the C1q multimer with a complete loss of function [56]. Groupings of male and female mice were used for all reported experiments in relatively equal numbers.

EAE:

We immunized mice with a pre-mixed emulsion (Hooke Labs EK-2110) containing myelin oligodendrocyte glycoprotein peptide, amino acids 35–55 (MOG_35–55; 1 mg/mL of emulsion) in complete Freund’s adjuvant (CFA) containing heat-inactivated mycobacterium tuberculosis H37RA (2–5 mg/mL emulsion). 100μl of the emulsion was injected subcutaneously at each of two sites over the upper and lower back. Sham-immunized sex-matched littermate controls were injected with control emulsion without MOG_35–55 (Hooke Labs CK-21100), and otherwise received identical treatment. Mice of each genotype were randomly assigned to either the sham or EAE treatment groups. Mice were injected with pertussis toxin (Hooke Labs, males: 75–90ng, females 90–120ng) intraperitoneally (i.p.), with the dose adjusted according to potency of each lot per manufacturer recommendations) on 0 and 1 day post-immunization (dpi). In one cohort, WT mice received ~100ul i.p. twice daily of Vehicle solution: 5% DMSO, 40% polyethylene glycol 400, and 55% normal saline beginning on the first day of visible motor deficits (EAE score 0.5 or greater) with timing matched for sham controls. All mice were monitored daily for signs of clinical disease and scored for motor deficits as follows: 0, no deficit; 0.5, partial tail paralysis; 1, complete tail paralysis; 2 hind-limb weakness; 2.5, paralysis of one hind limb; 3, paralysis of both hind limbs; 3.5, hind limb paralysis and forelimb weakness; 4; quadriplegia; 5; dead.

Western blots:

At twenty-eight days post-immunization mice were anesthetized with ketamine and xylazine (100 and 10 mg/kg, respectively) and perfused briefly with PBS. Hippocampi and liver were isolated, immediately frozen on dry ice, and stored at −80°C. Hippocampal and liver tissues were homogenized in RIPA lysis buffer containing a protease inhibitor cocktail (MilliporeSigma: Calbiochem, SetV). The cell lysate was kept on ice with periodic vortexing for 30 minutes then centrifuged at 13,000 rpm for 10 minutes and repeated with the resulting supernatant. The protein concentration of the final supernatant lysate was assayed with a detergent-compatible Bradford assay (Pierce). Equivalent amounts of protein were then mixed with loading dye and run on an 10 or 15% SDS-PAGE gel and transferred to PVDF. Membranes were blocked with 5% milk in TBSt (tris buffered saline (20 mM Tris-Cl, pH 7.4; 150 mM NaCl) with 0.1% Tween 20) for 30 minutes and probed with primary antibodies: C1q (kind gift of Andrea Tenner, 1151, [57]; 1:1000), C3d (R&D systems, AF2655, 1:1000), actin (Santa Cruz, SC-47778; 1:2000) and GAPDH (MilliporeSigma, CB1001; 1:2000) in 5% milk in TBSt at 4 degrees overnight. Membranes were washed 3 times in TBSt and then incubated with HRP secondary antibodies (anti-Mouse and anti-Rabbit HRP, Biorad, 170–6516 and 170–6515 respectively, used 1:5,000; anti-Goat HRP Azure Biosystems, AC2149, 1:10,000) for 1 hour in 5% milk in TBSt. After washing we applied ECL substrate (Pierce) and developed membranes using a digital imager (Azure Biosystems). Membranes were stripped using buffer consisting of 200mM glycine; 0.1% SDS, 1% Tween 20, pH 2.2 and re-probed up to 3 times. Western blot band densities were quantitated using Image J/FIJI. Band densities for each animal were normalized to the control sham group mean from the same gel and graphed as mean ± SEM.

RNA isolation and qPCR:

At 28–30 days post-immunization mice were anesthetized with ketamine and xylazine, perfused briefly with PBS, and hippocampi were isolated and flash frozen using dry ice. RNA from a single hippocampus per mouse was obtained using the Nucleospin RNA kit (Macherey Nagel; Cat# 740955). cDNA was synthesized using random hexamer primers and the Superscript III first-strand synthesis system for RT-PCR (ThermoFisher, Cat# 18080–051). qPCR was performed using Taqman probes: C1qa Mm00432142_m1; C3 Mm_00437838_m1; IPO8 Mm_01255158_m1 (ThermoFisher). Delta Ct values for C1q and C3 were calculated relative to the housekeeping gene IP08 for each sample and fold changes were calculated relative to sham controls.

To obtain RNA specifically from CD11b+ microglia/myeloid cells, mice 28–30 days post-immunization were anesthetized with ketamine and xylazine, perfused with PBS and the cortex and hippocampi were dissected and combined. Single-cell suspensions free from myelin and RBCs were obtained using the Adult Brain Dissociation kit (Cat# 130–107-677) and gentleMACS Octo Dissociator from Miltenyi according to the kit instructions. Single cell suspensions were incubated with CD11b magnetic beads (Miltenyi, cat# 130–093-636) then CD11b+ cells were separated from other cell types using a magnetic MACs separator and LS columns (Miltenyi). CD11b+ cells were pelleted and RNA isolation, cDNA synthesis, and qPCR reactions were processed similar to the whole hippocampal tissue outlined above.

IHC:

At 28–30 days post-immunization, mice were anesthetized with ketamine/xylazine (100 and 10 mg/kg, respectively) and intracardially perfused for 1 minute with phosphate-buffered saline (PBS) containing EDTA (1.5 mg/ml) followed by 4% paraformaldehyde (PFA) in PBS. Brains were post-fixed for 18–24 hours in 4% PFA then stored in PBS at 4°C. Brains were cut into 40μm-thick coronal sections using a vibratome (Leica V1000) and stored in a cryoprotectant mixture of 30% PEG300, 30% glycerol, 20% 0.1 M phosphate buffer, and 20% ddH₂O at −20°C. We performed free-floating-section IHC. The sections were washed three times for 30 min in PBS to remove the cryoprotectant. Then sections were incubated in 100mM glycine in PBS for 30 minutes followed by citrate antigen unmasking at 37°C for 30 minutes (Vector, H3300 with 0.05% tween 20). Primary antibodies were diluted in blocking buffer consisting of 1.5% BSA (MilliporeSigma; A3294), 3% normal goat serum (Vector Laboratories; NGS; S1000) or 3% normal donkey serum (MilliporeSigma, 566460), 0.5% Triton X-100 (Promega, H5142), and 1.8% NaCl in 1× PBS. We used the following antibodies in these experiments: rabbit anti-Iba1 (Wako Biochemicals, 019–19741; 1:1000) rabbit anti-C1q (Abcam Ab182451 clone 4.8; 1:500), goat anti-C3 (MP Biomedicals 55730, 1:300), chicken anti-Homer1 (Synaptic Systems, 160006, 1:500), and mouse anti-PSD95 (NeuroMab, 75–028, 1:500). Sections were incubated in the primary antibody mixture for 1–3 days at room temperature with agitation. Sections were then washed three times for 30 min in 1× PBS with 1.8% NaCl and then incubated overnight at room temperature in Alexa Fluor-conjugated secondary antibodies (1:500 Invitrogen or 1:250 Jackson Immuno Research) in blocking buffer. Finally, sections were washed 3 times with 1×PBS with 1.8% NaCl, mounted on slides with Prolong Gold or Diamond mounting agents (Life Technologies; P36935 and P36961).

Imaging:

Most IHC sections were imaged with a Hamamatsu ORCA-ER camera on an Olympus BX-51 upright microscope with Quioptic Optigrid optical sectioning hardware and the following objectives: 4x, 0.13 NA; 10x, NA 0.4; 60x, NA 1.4. Differences in z-axis registration of various fluors were corrected by calibration with multicolor fluorescent beads. Identical light intensity and exposure settings were used for all animals within each image set. For 60x images, we imaged the CA1-stratum radiatum hippocampal area collecting 6 image stacks (5, 10, or 12μm thick depending on the antibody staining set; z-step = 0.3μm) usually from the left and right hemispheres of 3 separate brain sections. C1q and PSD95 high magnification images in Figure 2 D and E were alternatively imaged using an Olympus FV1000 laser scanning confocal microscope with 100x objective (NA 1.40) and with a 0.46 μm z-step and total z stack of 11.5 μm. Image analyses for all IHC were done using Volocity 3DM software (Quorum Technologies; version 6.3). All image stacks are displayed as maximum intensity protections. IBA1+ cell morphology measurements (volume, surface area and skeletal length) were obtained using an intensity-based automated segmentation algorithm in Volocity (“find objects”) followed by a fine filter. Homer1 and PSD95 puncta were identified using the “find spots” algorithm in Volocity software, which identifies puncta based on local intensity maxima that also exceed an intensity threshold value. For all measurements, values from image stacks in 6 tissue sections per mouse were averaged and expressed relative to sham-immunized controls from the same genotype and graphed as mean ± SEM. Experimenters were blinded to genotype or immunization group throughout each IHC experiment and subsequent data analysis.

Contextual/Cued Fear Conditioning:

Experiments were performed as previously reported [58]. Briefly, mice underwent fear conditioning at 25–28 days post immunization. Mice were placed in an isolation chamber with a metal grid floor and presented with a 15 second white noise cue co-terminating with an electric shock (2 seconds at 0.19 mA). The paired auditory cue and shock were delivered 3 times, 30 s apart. Mice were then returned to their home cage overnight. Twenty-four hours later, freezing behavior in response to the conditioned context, a novel context, and the auditory cue was measured. To test the contextual conditioning responses, mice were placed in the same isolation chamber, prepared identically to the conditioning setup, and video recorded for 5 minutes without shock or white noise cue for subsequent motion analysis (conditioned context test). After returning mice to their home cage for 3 hours, novel context and cued conditioning responses were tested by placing mice in a round test cage with smooth plastic walls wiped with ethanol, the floor covered with clean cage bedding, and illuminated with a red light to provide a novel context that differed from the conditioned context in shape, texture, odor and color. Video for motion analysis was recorded for 3 minutes without stimulus (novel context test) followed by 3 minutes of continuous presentation of the white noise auditory cue (cue test). FreezeFrame/FreezeView software (Actimetrics, Wilmette IL) or ANY-maze software (Stoelting Co.; Wood Dale IL) was used to analyze motion to detect episodes of freezing behavior which were defined as ≥ 0.5 s bouts of immobility except for the movements associated with breathing. Freezing responses were quantified as the percentage of total time spent freezing from 30–180 s after being placed in the conditioned context, 30–150 s after being placed in the novel context, or 0–60 s of cue presentation (beyond which responses tended to deteriorate at variable rates). Because EAE mice had increased rates of immobility at baseline likely reflecting motor deficits, we subtracted rates of freezing during exposure to the novel, unconditioned context from rates during the conditioned context and cue tests to generate a measure (ΔFreeze) that reflected increased freezing due to the mouse’s ability to discriminate between the conditioned and novel context, and in response to the cue. These values were normalized to those of the WT sham-immunized controls of the same sex for the same testing cohort. Mice with >75 % immobility over the duration of the novel context test (2 WT EAE mice of 36) or >25% more freezing/immobility during the novel context test than the conditioned context (1 C1qa KO EAE mouse of 20) were excluded from the Δ Freeze analysis, as high baseline immobility interfered with measurement of conditioning effects.

Statistics:

GraphPad Prism software version 8.3.0 for Windows (La Jolla California USA) was used to perform all statistics. N values (number of animals) for each experiment are reported in the figure legends and come from two or more independent cohorts. We defined significance as p<0.05. Statistical tests are listed in text and figure legends. In general, Student’s t-tests were used for statistical analysis comparing sham vs EAE in WT-only data sets. ANOVAs followed by Sidak’s multiple comparisons test were used when comparing sham vs EAE affects in multiple genotypes or in different sexes. Welch’s ANOVA with Dunnets T3 multiple comparisons test was used for analysis of the clinical scores. All data are expressed as the mean ± standard error of the mean (SEM).

Study approval:

Animal care and use were carried out in compliance with the US National Research Council’s Guide for the Care and Use of Laboratory Animals and the US Public Health Service’s Policy on Humane Care and Use of Laboratory Animals. Protocols were approved by the University Committee on Animal Resources at the University of Rochester.

RESULTS:

C1q and C3 expression in EAE hippocampus

Our lab has previously shown that EAE immunization results in significant synapse loss in the CA1-stratum radiatum of the hippocampus, which is largely spared from demyelination [45, 58]. To determine if the EAE hippocampus also exhibits increased complement production and deposition in EAE that could make synapses vulnerable to phagocytosis by glia, we first analyzed C1q and C3 protein and mRNA expression by Western blot and qPCR. By western blot, we found that C1q protein expression in the hippocampus of EAE mice was increased 2.5-fold compared to sham (Figure 1 A-B; p<0.001; t-test). Expression of full length C3 protein was also elevated in EAE mice 1.6-fold above sham controls (Figure 1 A-B; p=0.002; t-test). Hippocampal lysates from C1qa KO and C3 KO sham and EAE mice were used to validate the antibodies (Figure 1A). The small amount of C1q protein remaining in C1qa KO mice is likely due to C1qb and C1qc expression within cells that is non-functional without C1qa present to fold properly into the full C1q oligomer. The C3 antibody used for blotting has its epitope within the C3d region of C3 and thus it can detect full length C3 as well as cleavage products that signal ongoing or chronic complement pathway activation (see C3 cleavage schematic in Figure 1E). EAE hippocampal lysates have significant increases (6–7 fold; p<0.001, t-test) in the C3 cleavage isoforms of C3α’1 and C3d compared to sham control lysates where these isoforms are basically absent (Figure 1A-B). The EAE-induced increases in C3 expression and C3 cleavage products are specific to the CNS, as C3 levels in the blood and liver (the main producer of serum complement) are the same for sham and EAE mice (Supplementary Figure 1). C1q levels were likewise unaltered in blood, but slightly elevated (1.3 fold; p=0.05, t-test) in the liver by EAE (Supplementary Figure 1).

Figure 1: EAE mice have increased expression of complement components C1q and C3 in the hippocampus.

Hippocampi were isolated from sham or EAE mice at 25–30 days post immunization. A) Hippocampal lysates from sham or EAE WT, C1qa KO, or C3 KO mice were probed by western blot for expression of C1q, C3, and Actin. In the bottom set of blots (C3 & actin), all lanes came from the same gel and exposure. Extended images of all blots are shown in Supplementary Figure 3. B) Quantification of band densities from western blots. C1q/Actin and C3/Actin band densities are normalized to sham controls. Quantification included n=10–11 WT animals per group. C) RNA isolated from the hippocampus of sham or EAE mice was transcribed into cDNA and analyzed by qPCR for expression of C1qa and C3 with fold changes relative to sham controls. n=10 per group. D) qPCR results of C1qa and C3 gene expression from CD11b+ microglia/myeloid cells isolated from the hippocampus and cortex of sham or EAE mice. n=5 per group. E) Schematic of C3 cleavage products. The C3 antibody used for blotting recognizes the C3d region of the C3 alpha chain which is highlighted in dark grey. For all graphs: Statistical Analysis: Student’s t-test.

qPCR analysis of mRNA expression from hippocampal tissue shows there is a significant increase in the local expression of both C1qa and C3 in EAE mice compared to sham controls (Figure 1C; C1qa: 2.1 fold above sham controls p<0.001 t-test; C3: 8.4 fold above sham controls, p=0.001). Thus the increased protein expression of C1q and C3 in the EAE hippocampus is due at least in part to local gene expression and not simply due to blood brain barrier breakdown. Because microglia are thought to be the main producers of C1q in the healthy hippocampus [59] and can also upregulate C3 when activated, we used qPCR to analyze the local gene expression of the CD11b+ microglia/myeloid cells in sham and EAE mice. In order to obtain sufficient cells, we isolated CD11b+ cells from the hippocampus and cortex. We found that CD11b+ cells in EAE mice significantly over-express C3 at 28–30 days post immunization compared to sham controls (Figure 1 D; 54.5 fold increase; p=0.016, t-test). However, there was only a small trend for increased C1qa expression by CD11b+ cells in EAE, but no significant difference (Figure 1 D; 1.3 fold p=0.32 t-test). This shows that microglia (and other myeloid cells) contribute to the elevated C3 expression in EAE. However, we cannot rule out EAE-induced elevated expression of complement proteins by other cell types in the hippocampus, particularly as reactive astrocytes are also known to express high levels of C3 [60].

Next, we performed IHC using anti-C1q and anti-C3 antibodies to see where the elevated C1q and C3 were located in the EAE hippocampus (Figure 2). By IHC, we measured a 1.6-fold increase in mean C1q fluorescence across the hippocampus in EAE vs sham mice (Figure 2A-B; p=0.02; t-test) with significant increases ranging from 1.5–1.9-fold in all sub-regions of the hippocampus (Figure 2A,C). At 100x magnification within the CA1-stratum radiatum (SR), elevated levels of C1q were apparent in EAE. We observed that C1q was diffusely localized throughout the neuropil but C1q was also localized at higher density in small punctate regions, some of which co-localized with synapses (identified by the postsynaptic marker PSD95) or along the dendrites (identified by the linear PSD95+ synaptic pattern) in both sham and EAE brains (Figure 2D-E). However, we also observed abundant heterogeneous expression of C1q puncta that did not co-localize to synapses, but may contact other structures in the neuropil (Figure 2 E).

Figure 2: C1q protein levels are elevated in the hippocampus of EAE mice.

A-D) Brain sections from EAE and Sham control mice were immuno-labeled for C1q. A) Images show C1q in the hippocampus; Scale bar = 250μm. B) Quantification of mean C1q fluorescence intensity in hippocampus normalized to sham controls. n=4 Statistical Analysis: Student’s t-test. C) Quantification of C1q intensity within hippocampal sub-regions. SR=Striatum Radiatum; SO=Striatum Oriens; DG=Dentate Gyrus. n=17 animals per group. Statistical Analysis: Student’s t-test for each region. D) Higher-magnification (100x) images show considerable diffuse as well as punctate C1q expression in CA1-Stratum Radiatum (maximum intensity projection from 11.5μm z-stack). Scale bar = 20μm. E) C1q and PSD95 immuno-stain in brain sections within the CA1-Stratum Radiatum (100x magnification; 1 z-slice). Blue arrows highlight examples of co-localized C1q and PSD95. Scale bar = 5μm. F) C1q antibody (Abcam Ab182451) is specific for C1q as no staining is visible in C1qa KO mice. Scale bar = 250μm.

A similar global increase in diffusely localized C3 was observed across the hippocampus in EAE compared to sham controls using a polyclonal anti-C3 antibody that recognizes full length C3 as well as C3 cleaved fragments (mean C3 intensity, sham:140, EAE: 336; p=0.03; Dunnet’s T3; Figure 3A-B). All measured sub-regions around CA1 of the hippocampus, including the stratum radiatum, showed significant increases of C3 in EAE ranging from 1.5–2.5 fold (Figure 3A,C). Additionally, we measured increased C3 mean intensity co-localized to synapses (identified by antibodies to the postsynaptic protein Homer1) in EAE compared to sham controls (2.8 fold; p=0.047 t-test). It is important to note, however, that levels of C3 seemed to be elevated throughout hippocampal sub-regions and not just on synapses (Figure 3 D-F). As both C3 and C1q are secreted proteins and the antibodies used in this study do not specifically highlight sites of complement activation, the diffuse staining pattern throughout the neuropil is perhaps not surprising. Yet, the IHC results do support the conclusion that elevated levels of both complement proteins are abundant on and around synapses in the hippocampus in EAE.

Figure 3: C3 protein levels are elevated in the hippocampus of EAE mice.

A) C3 immuno-staining with Dapi in CA1 region of the hippocampus of Sham and EAE WT and C3 KO mice. Scale bars = 60μm. B) Quantification of mean C3 fluorescence intensity in 10x hippocampal images spanning CA1 (as shown in A). Statistical Analysis: two-way ANOVA with Sidak’s multiple comparisons test. WT sham n=8; WT EAE n=13; C3 KO Sham n=5; C3 KO EAE n=9. C) Quantification of C3 intensity within hippocampal sub-regions normalized to WT sham controls. D) Quantification of C3 mean intensity overlapping Homer1+ synaptic puncta. For both C and D, WT sham n=8; WT EAE n=13. Statistical Analysis: Student’s t-test. E and F) C3 (green) and Homer1 (red) immuno-staining in CA1-Stratum Radiatum at 60x magnification. E) 5μm thick image stacks; Scale bars = 12 μm. F) 1.5 μm thick image stacks with digital magnification; scale bar = 3 μm. Blue arrows show C3 co-localized with Homer1.

Comparison of EAE pathology in WT, C1qa KO, and C3 KO Mice

Loss of C3, but not C1q, results in less severe EAE motor impairment

Next, we immunized wild type (WT), C1qa knockout (KO) and C3 KO mice for EAE to determine whether the early complement pathway is a key driving component of synapse loss within the grey matter in EAE, particularly in the hippocampus. We also assessed whether the loss of C1q or C3 could protect against the EAE-induced motor impairment and paralysis caused by spinal cord damage (Figure 4A). As previously reported [27, 29, 32], C3 knockout resulted in less severe EAE motor deficits and paralysis with a significant decrease in the mean clinical score throughout the course of disease. On the other hand, C1qa KO did not alter the EAE disease course in any way, as the C1qa KO EAE mean clinical scores were virtually identical to WT EAE. Neither C1qa KO nor C3 KO prevented the EAE induced weight loss (Figure 4B). Thus C3 contributes to MOG_35–55 EAE white matter disease, whereas the C1q-initiating arm of the classical complement pathway does not contribute to classic EAE clinical symptoms.

Figure 4: Loss of C3, but not C1qa, resulted in a reduction in the severity of EAE motor deficits.

A) Graph of the mean clinical scores corresponding to days post symptom onset from EAE immunized WT, C1qa KO, and C3 KO mice. Scores from C3 KO EAE mice were significantly less (p<0.05) than WT EAE mice beginning on Day 2 through day 18. There were no significant differences between WT EAE vs C1qa KO EAE. B) Graph of the mean animal weights. All genotypes showed significant sham vs EAE differences in weight by day 3 through day 18 (p<0.05). There were no significant differences in weight loss between WT, C1qa KO or C3 KO EAE groups. Statistical Analysis: Welch’s ANOVA with Dunnet’s T3 multiple comparisons test for clinical scores and two-way ANOVA with Sidak’s multiple comparisons test for weights. For data set, group sizes are: WT Sham: n=36; WT EAE: n=35; C1qa KO Sham n=14; C1qa KO EAE: n=14; C3 KO Sham: n=21; C3 KO EAE n=24.

EAE-induced hippocampal synapse loss in WT, C1qa KO and C3 KO mice

As we have previously documented significant synapse elimination in the CA1-stratum radiatum layer of the hippocampus of EAE mice, we chose to focus on this same region for this study assessing the role of complement in grey matter synapse loss. Loss of complement gene C3, and perhaps C1qa, protected synaptic density during EAE (Figure 5). WT mice with EAE displayed a 15% decrease in the number of PSD95+ synaptic puncta in the CA1-striatum radiatum of the hippocampus compared to sham controls (p=0.002; Sidak; Figure 5A,C). C3 KO EAE mice were almost completely protected from EAE-induced synapse elimination, as they had only 2.6% fewer PSD95+ synaptic puncta than sham controls of the same genotype (p=0.95, Sidak). On the other hand, C1qa KO EAE mice exhibited just a small trend for protection, with 11% fewer synapses than sham controls, and the synapse loss was no longer statistically significant (p=0.28;). However, only C3 KO EAE mice had significantly reduced PSD95 puncta loss compared to WT EAE mice (p=0.02). EAE also induced a very small reduction in the intensity of PSD95 in PSD95+ synaptic puncta in WT mice compared to sham controls that was eliminated by C3 KO (WT EAE vs C3 KO EAE p=0.01; Figure 5C).

Figure 5: C3 KO mice are protected from EAE induced synapse loss in the CA1-SR region of the hippocampus.

A-B) Brain sections from WT, C1qa KO, and C3 KO sham and EAE immunized mice were immunostained for the postsynaptic markers PSD95 (A) and Homer1 (B). Scale bars = 10μm. C) Quantification of PSD95+ synaptic density and intensity normalized to sham controls of each genotype. D) Quantification of Homer1+ synaptic density and intensity normalized to sham controls of each genotype. For data set, group sizes are: WT sham: n=31–33-25; WT EAE: n=32; C1qa KO sham: n= 13–14; C1qa KO EAE: n=13–14; C3 KO sham: n=18; C3 KO EAE: n=21. Statistical Analysis: ANOVAs with Sidak’s multiple comparison tests.

When synapses were alternatively labeled with antibodies to the postsynaptic protein, Homer1, we also found that genetic loss of C3 prevented EAE induced-synapse elimination (Figure 5 B, D). WT EAE mice had a significant, 9% loss of Homer1+ postsynaptic puncta compared to sham controls (p=0.006, Sidak); whereas, C1qa KO EAE mice had a 7% loss and C3 KO’s were completely protected against EAE-induced Homer1+ synapse removal (0% change). Both C1qKO and C3 KO EAE groups were no longer significantly different from sham controls (C1qKO p=0.32; C3 KO p=0.99, Sidak), but only C3 KO EAE mice were significantly improved from WT EAE (p=0.006). We found no change in the intensity of Homer1 in Homer1+ synaptic puncta for any experimental group (Figure 5C). We also found no differences in absolute synapse counts between WT, C1qa KO, and C3 KO mice, as they all had ~4000 synapses per 10μm image stack (WT: 3933 +/− 113; C1qa KO: 3951 +/− 118; C3 KO: 3929 +/− 175) showing that the genetic knockout of complement genes did not change the development of the hippocampus in a way that affected synapse numbers. In summary, knockout of C3 prevents EAE induced-synapse elimination in the hippocampus, but knockout of C1qa results in only a small, insignificant trend toward reduced synapse loss.

C3 knockout mice, but not C1q knockout mice, have less EAE induced microglial activation in the hippocampus compared to WT

We next assessed whether C1qa or C3 KO could alter the morphometric parameters of microglial activation induced by EAE in the hippocampus. In WT mice, EAE immunization results in increased IBA1 expression by microglia, as well as, a change in microglial morphology that results in thicker and shorter processes that has been associated with a reactive microglial phenotype (Figure 6) with a disease-associated transcriptomic signature [61]. This change in microglial cell morphology can be measured by a decrease in the surface area/volume ratio and a decrease in the skeletal length/volume ratio of IBA1+ cells. WT EAE mice have a significant 1.2-fold increase in the average IBA1+ cell volume per field of view compared to WT Sham (p=0.005; Sidak, Figure 6B) and a significant 1.4-fold increase in the summated IBA1 intensity compared to WT Sham (p<0.001, Figure 6C). C1qa KO EAE mice show a similar 1.4-fold increase in IBA1+ cell volume and a 1.6-fold increase in IBA1 intensity compared to C1qa KO shams (p<0.001 for both). However, C3 KO EAE mice had only a small, insignificant increase in IBA1+ cell volume (1.1-fold; p=0.58) and a small increase in IBA intensity (1.27-fold; p=0.10) compared to C3 KO shams. Similar results were obtained for the surface area/volume ratio and skeletal length/volume ratio (Figure 6 D, E). C1qa KO EAE and WT EAE mice had similar significant decreases in these two microglial morphology measurements compared to genotype-matched sham controls (Surface area/volume: WT EAE 0.89, p<0.001; C1qa KO EAE 0.88, p<0.001 compared to 1.0 in genotyped matched sham controls; Skeletal Length/volume: WT EAE 0.77, p<0.001; C1qaKO EAE 0.76, p<0.001; compared to 1.0 in genotyped matched sham controls). C3 KO EAE mice, however, showed smaller decreases in the surface area/volume ratio (C3 EAE 0.94 vs C3 sham 1.0; p=0.008) and skeletal length/volume ratio (C3 KO EAE: 0.85 vs C3 sham 1.0; p=0.01). C3 KO EAE showed trends for maintaining homeostatic microglial morphology compared to sham controls across all measurements but only the surface area to volume ratio was significantly improved compared to WT EAE (p=0.03). Taken together, these results show that C3 KO limited hippocampal microglial activation induced by EAE, but C1qa KO had no effect on microglial morphology in EAE.

Figure 6: C3 KO mice with EAE have reduced microglial activation compared to WT EAE mice. Loss of C1qa had little to no effect on microglial activation induced by EAE.

A) Brain sections from WT, C1qa KO, and C3 KO sham and EAE immunized mice were immunostained for the microglia protein, IBA1, and imaged in the CA1-stratum radiatum of the hippocampus. Scale bars = 11μm. B-E) Quantification of microglia (IBA1+ cell) expression and morphology normalized to sham controls of each genotype. B) Quantification of the Sum Volume of IBA1+ cells per 60x field of view. C) Quantification of the Sum IBA1 Intensity per 60x field of view. D) Quantification of the IBA1+ cell surface area to volume ratio. E) Quantification of the IBA1+ cell skeletal length to volume ratio. For data set, group sizes are: WT sham: n=33; WT EAE: n=32; C1qa KO sham: n= 14; C1qa KO EAE: n=14; C3 KO sham: n=18; C3 KO EAE: n=21. Statistical Analysis: ANOVAs with Sidak’s multiple comparison tests.

Contextual/Cued Fear Conditioning in WT, C1qa KO and C3 KO sham and EAE mice

Next, we tested whether knockout of C1qa or C3 can protect hippocampal-dependent learning and memory specific to contextual fear conditioning (Figure 7).Contextual fear conditioning can be used to assess learning and memory in mice during the chronic phase of EAE because it is less confounded by EAE motor deficits than other learning paradigms that are dependent on extensive mobility and exploratory behavior. Freezing behavior in response to a specific environmental context or auditory cue stimulus that had been paired with a foot shock during a prior conditioning protocol is measured and distinguished from EAE-associated reductions in baseline mobility by comparison to freezing rates in an unconditioned novel context. We and other groups have reproducibly used fear conditioning to monitor learning and memory in EAE mice [52, 58, 62].

Figure 7: Contextual/Cued Fear Conditioning in WT, C1qa KO and C3 KO sham and EAE mice.

A) Mice were conditioned with a foot shock and paired auditory cue, and bouts of freezing behavior were measured 24 hours later when mice were returned to the conditioned context or in a separate test placed in a novel context and presented with the cue. B) Group averages of percentage of time spent freezing, shown for 30-s epochs in each csidaondition. C) Because EAE mice of all genotypes had reduced mobility and higher rates of freezing in a novel context which could be considered a baseline measurement, incremental increases in freezing in the conditioned vs. novel context (Δ context) and cue vs novel context (Δ cue) were used to measure changes in freezing behavior due to contextual and cued fear conditioning, respectively and is quantitated as Δ Freeze normalized to WT sham controls. D) Mean EAE Clinical Scores for each cohort on day of behavioral testing (27–28 days post immunization) For all graphs group sizes are: WT sham: n = 31; WT EAE: n = 34; C1qa KO sham: n = 24; C1qa KO EAE: n = 23; C3 KO Sham n = 15; C3 KO EAE n = 19. Statistical analysis: 2-Way ANOVAs with Sidak’s multiple comparison tests. For the Clinical Score Welch’s 2-Way ANOVAs with Dunnet’s T3 multiple comparison tests. Legend at bottom right applies to graphs in both C and D.

Our data show that after mice were conditioned with a foot shock and paired auditory cue within a defined contextual environment, WT EAE mice showed a significantly smaller Δ Freeze (% of time freezing in the conditioned context minus baseline % time freezing in a novel context) than WT sham mice (WT sham = 1; WT EAE = 0.62; p=0.011; Sidak; Figure 7B-C). C1q KO EAE mice had freezing responses that were basically identical to WT EAE mice (Context: Δ Freeze: C1qa KO sham = 0.96; C1qa KO EAE = 0.64; p=0.098; WT EAE vs C1qa KO EAE p=0.98). Interestingly, C3 KO EAE mice showed a trend for improved memory of the conditioned context as C3 KO EAE mice were no longer significantly different from C3 KO sham mice in the context Δ Freeze (C3 KO sham = 0.99; C3 KO EAE = 0.87; p=0.86). Yet, C3 KO EAE’s improved Δ Freeze was still shy of significance when compared to WT EAE mice (p=0.18). Freezing responses to the conditioned auditory cue (which does not involve hippocampal-dependent memory) relative to the novel context were more variable and showed no significant differences between the experimental groups (Figure 7 B,C). In summary, we found a relative deficit in contextual freezing above baseline mobility in WT EAE mice compared to sham controls that was potentially improved by C3 KO.

Sex-based differences

Combined cohorts of male and female mice in relatively equal numbers were used for all experiments reported in this study. As multiple groups have reported sex-dependent differences in complement protein levels and complement mediated damage in several pathological conditions in mice and humans [63, 64], we did analyze the key findings of this study for differences based on sex (specifically: C3 levels in the hippocampus (IHC), EAE clinical scores, synaptic density, IBA1+ cell measurements (surface area to volume ratio, skeletal length to volume ratio, IBA1 sum volume) and contextual fear conditioning Δ Freeze/control in the conditioned context). All two-way ANOVA tests showed no significant main effects of sex or significant interactions of sex with immunization status or genotype so the data was combined and reported from both sexes. Data separated by sex is available in Supplementary Figure 2. When split into male and female sub-groups, some experiments may be under-powered, yet most experiments showed the same trends in both sexes and maintained significant differences between WT sham and WT EAE and between WT EAE and C3 KO EAE. The one exception is the contextual fear conditioning experiment where much of the C3 KO EAE improvement compared to WT EAE was in female mice where it was statistically significant (Δ Freeze in Females: WT EAE 0.61 vs C3 KO EAE 1.05; p=0.04; Δ Freeze in Males: WT EAE 0.62 vs C3 KO EAE 0.66; p=0.98;) (Supplementary Figure 2E).

Discussion:

The overall goal of this study was to investigate complement expression and the contributions of complement proteins to synaptic damage, microglial activation, and learning and memory that contributes to contextual/cued fear conditioning in an EAE model with a focus primarily on hippocampal grey matter pathology. Our results show that mice immunized with MOG_35–55 to produce EAE have increased local expression of complement components C1q and C3 within the hippocampus deposited throughout the neuropil. C3 cleavage products were also present in the hippocampus indicating localized chronic complement activation. Knockout of C3, but not C1qa, significantly attenuated the severity of EAE motor impairments, protected against synapse loss, reduced microglial activation, and improved freezing and memory in contextual fear conditioning experiments. Thus, pathologic expression and activation of C3 contributes to hippocampal grey matter pathology in EAE.

Previous studies have shown that genetic ablation of C3, Factor B (an alternative complement pathway component), or receptors to C3 cleavage products (C3aR, CR3 (CD11b−/−), CR4 (CD11c−/−)) protects against MOG_35–55 EAE spinal cord pathology and associated motor deficits [27, 29–32, 65]. Additionally, loss of endogenous inhibitors like murine membrane-bound Crry exacerbate the EAE clinical score and microgliosis [66] and treatment with complement inhibitors that limit the C3 convertase, such as CR2-Crry, CR2-CFH, and sCR1, or antibodies that block the alternative pathway have been shown to attenuate the EAE clinical score [35, 39, 67–69]. On the other hand, components of the classical pathway (such as C4) seem largely dispensable for MOG_35–55-induced EAE motor impairments with genetic deletions resulting in little to no change in the EAE clinical score [70]. However, classical pathway components, C1q and C4, directly contribute to pathology in other EAE models such as an auto-antibody exacerbation model of EAE [70, 71]. Our data agree with these previous findings, as C1qa KO had no effect on the MOG_35–55 EAE clinical score, whereas, C3 KO mice were partially protected from EAE motor impairments.

The genetic background of mice may also be important to consider, as there are differences in complement activity [72, 73] and EAE disease penetration in various mouse strains [74]. Specifically, conflicting reports exist on whether C3 KO protects against MOG_35–55EAE disease severity assessed by the EAE clinical score. Our results align with three independent studies from the Barnum lab that used the C57BL/6 mouse strain and found C3 KO to be protective in EAE. One conflicting study used a 129sv x C57BL/6 F1 hybrid background and reported no effect of C3 KO compared to WT on MOG_35–55 EAE disease [27–29, 32].

This is the first study to investigate complement’s contribution to synaptic damage within EAE hippocampal grey matter, where genetic deletion of C3, but not C1q provided significant protection against EAE-induced synapse loss and reduced microglial activation. Complement opsonization of synapses followed by microglia phagocytosis is a mechanism for synapse pruning during development and in some disease models including Alzheimer’s, frontotemporal dementia, viral encephalitis, aging, traumatic brain injury, and schizophrenia [11, 15–22, 75–78]. Our results suggest that MOG_35–55 EAE (and perhaps multiple sclerosis) can be added to this list, although the precise mechanistic roles of C3 need additional studies to be fully elucidated. As C3-dependent aspects of the complement pathway contributed not only to hippocampal pathology, but also to motor deficits connected to white matter spinal cord pathology, it is possible that C3 KO may protect hippocampal grey matter secondary to systemic reduction in EAE severity. Yet, we did find evidence of chronic C3 activation directly in the hippocampus. Also, C3 KO only partially reduced the EAE clinical score, but almost completely protected synapses, suggesting that some complement-mediated synaptic pruning may be locally occurring in the EAE hippocampus by different mechanisms than in spinal cord.

Due to the many roles of complement in mediating immune cell functions, the role of C3 in EAE and MS is likely complex. C3 is known to contribute to co-stimulation strength of T cells and B cells, chemoattraction of infiltrating cells, cytokine production, opsonization of myelin and other debris, and likely opsonization of synapses [27, 29, 32, 79–81].Unfortunately, there is no currently available conditional C3 KO mouse to probe the local role of C3 within the hippocampus and it will be important in the future to tease apart the role C3 plays in complement-dependent synaptic pruning from those in white matter inflammation and systemic immune activation in EAE. Regardless though, in light of evidence of complement activation and synaptic phagocytosis reported in grey matter from MS patients [9, 44], the protective effects of C3 KO on EAE grey matter pathology are exciting given the paucity of treatments for grey matter neuroprotection in MS.

In summary, genetic loss of C3, but not C1q, provides protection from EAE induced synapse loss, microglial activation, and learning and memory impairment. This coupled with the protection from EAE motor impairments suggest that the early complement pathway may be an important therapeutic target for MS and may be especially relevant to the grey matter neurodegeneration in progressive MS.

Highlights.

• Expression of complement components, C1q and C3, is increased in the EAE hippocampus.

• Genetic loss of C3 prevents synapse elimination in the EAE hippocampus.

• Loss of C3, but not C1qa, reduces microglial activation and EAE motor impairments.

• C3 knockout protects memory and freezing behavior after contextual fear conditioning in EAE.