Abstract
Ischaemic stroke induces endogenous repair processes that include proliferation and differentiation of neural stem cells and extensive rewiring of the remaining neural connections, yet about 50% of stroke survivors live with severe long-term disability. There is an unmet need for drug therapies to improve recovery by promoting brain plasticity in the subacute to chronic phase after ischaemic stroke. We previously showed that complement-derived peptide C3a regulates neural progenitor cell migration and differentiation in vitro and that C3a receptor signalling stimulates neurogenesis in unchallenged adult mice. To determine the role of C3a–C3a receptor signalling in ischaemia-induced neural plasticity, we subjected C3a receptor-deficient mice, GFAP-C3a transgenic mice expressing biologically active C3a in the central nervous system, and their respective wild-type controls to photothrombotic stroke. We found that C3a overexpression increased, whereas C3a receptor deficiency decreased post-stroke expression of GAP43 (P < 0.01), a marker of axonal sprouting and plasticity, in the peri-infarct cortex. To verify the translational potential of these findings, we used a pharmacological approach. Daily intranasal treatment of wild-type mice with C3a beginning 7 days after stroke induction robustly increased synaptic density (P < 0.01) and expression of GAP43 in peri-infarct cortex (P < 0.05). Importantly, the C3a treatment led to faster and more complete recovery of forepaw motor function (P < 0.05). We conclude that C3a-C3a receptor signalling stimulates post-ischaemic neural plasticity and intranasal treatment with C3a receptor agonists is an attractive approach to improve functional recovery after ischaemic brain injury.
Introduction
Stroke is the primary cause of disability in adults and the second most common cause of death (Feigin et al., 2014). Loss of function after stroke is due to cell death in the infarcted tissue and cell dysfunction in surrounding and remote brain areas that are connected to the damaged area (Wieloch and Nikolich, 2006). Ischaemic brain damage induces endogenous repair processes that include proliferation and differentiation of neural stem cells, resulting in partial replacement of lost neurons (Gu et al., 2000; Arvidsson et al., 2002) and extensive rewiring of the remaining neuronal connections (Carmichael et al., 2001). The latter process involves sprouting of axonal projections and establishment of new synaptic contacts that result in cortical map rearrangement (Winship and Murphy, 2009). Understanding the mechanisms controlling these ischaemia-induced neural plasticity processes and their modulation is paramount to identification of novel treatment strategies to promote functional recovery.
The complement system is a part of innate immunity that provides an effective first line of defence against invading microorganisms by contributing to opsonization and cytolysis, promoting phagocytosis of foreign particles and leucocyte recruitment. We have previously shown that C3a receptor (C3aR, encoded by C3ar1) signalling stimulates neurogenesis in unchallenged adult mice (Rahpeymai et al., 2006) and C3a regulates neural progenitor cell migration and differentiation in vitro (Shinjyo et al., 2009). Complement activation-mediated neutrophil infiltration is detrimental in several types of ischaemic injury. Consistent with this view, complement inhibition proved neuroprotective in cerebral ischaemia with reperfusion (Huang et al., 1999; De Simoni et al., 2003; Costa et al., 2006; Mocco et al., 2006; Arumugam et al., 2009; Yang et al., 2013; Gong et al., 2015). Treatment with a C3aR antagonist improved functional and morphological outcome following ischaemia-reperfusion in adult mice (Ducruet et al., 2012). However, the precise role of C3a in the ischaemic brain is unclear. In a permanent cerebral ischaemia model, deletion of the C3 gene was associated with the development of larger infarcts and reduced post-stroke neurogenesis (Rahpeymai et al., 2006). In an in vitro ischaemia model, C3a increased the survival of astrocytes (Shinjyo et al., 2015). Overexpression of C3a in reactive astrocytes in the immature brain was shown to be neuroprotective, and intraventricular treatment with C3a ameliorated memory impairment resulting from neonatal hypoxia-ischaemia in wild-type control (C3aR+/+) mice but not C3aR-deficient (C3aR−/−) mice (Järlestedt et al., 2013). In addition, C3 expression was upregulated in sprouting neurons isolated from rat cortex after ischaemic stroke (Li et al., 2010). As some growth factors have been shown to promote axonal regeneration and sprouting after spinal cord injury (Lu et al., 2004; Vavrek et al., 2006), the finding that C3a induces upregulation of neural growth factor (NGF) in microglia and astrocytes in vitro (Heese et al., 1998; Jauneau et al., 2006) implies that complement can also exert pro-regenerative functions indirectly. Taken together, these findings raise the possibility that the complement proteins, and C3aR signalling in particular, are involved in ischaemia-induced neural plasticity including cell replacement, reorganization of axonal circuitry, and consequently, regulation of synaptic input. However, the role of the complement system in ischaemic brain injury is complex and seems to depend not only on factors such as the type of ischaemic injury and the developmental stage of the brain but also on length of time after injury.
In the present study, we sought to determine the role of C3a in neural plasticity and functional recovery following permanent focal cerebral ischaemia. We used C3aR−/− mice, GFAP-C3a transgenic mice, in which the astrocyte-specific expression of C3a is driven by the promoter for GFAP leading to the production of biologically active C3a in the CNS, and their respective wild-type (WT) controls. Further, to avoid the potentially negative effects of C3a in the acute phase of stroke, we applied a pharmacological approach and treated wild-type mice with intranasal C3a starting 7 days after stroke induction. To assess neural plasticity, we quantified the expression of synaptic markers synapsin I and VGLUT1 (encoded by Slc17a7), and growth associated protein 43 (GAP43), a surrogate marker of axonal plasticity (Benowitz and Routtenberg, 1997), as well as behaviour related to the recovery of motor function. Our findings reveal a stimulatory effect of C3a-C3aR signalling on post-stroke brain plasticity and identify treatment with C3aR agonists as an attractive approach to improve functional recovery after ischaemic brain injury.
Materials and methods
Animals
C3aR−/− mice (Kildsgaard et al., 2000) were backcrossed onto the C57BL/6J genetic background (Jackson Laboratories) for 10 generations. Heterozygous mice were then intercrossed to generate homozygous C3aR−/− mice. Wild-type C57BL/6J mice (C3aR+/+) served as controls. GFAP-C3a mice on a C57BL6/CNr genetic background (Boos et al., 2004) and their wild-type littermates were used as controls. Male, 7–9-month-old mice weighing 35–45 g were used: C3aR−/− (n = 14), C3aR+/+ (n = 10), GFAP-C3a (n = 12) and wild-type (n = 13). For the treatment experiments, male, 5-month-old wild-type C57BL/6CNr mice (Charles River), weighing 30–35 g were used. For in vivo imaging of fluorescent peptide translocation after intranasal administration, 2.5-month-old male C57BL/6 Albino mice (Charles River) were used. Mice were housed under standard conditions on a 12-h light/12-h dark cycle with food and water ad libitum. All experiments were conducted according to protocols approved by the Ethics Committee of the University of Gothenburg (permit number: 146–2008, 170-2009, 308-2012, 41-2015).
Photothrombotic stroke induction
Cortical photothrombosis was induced using the Rose Bengal method (Watson et al., 1985; Lee et al., 2007) with some modifications. Anaesthesia was induced with isoflurane (Forene®, Abbott) in air and oxygen (1:1) initially at 5% and reduced to 2.5% during the surgical procedure. Body temperature was monitored by a rectal probe and maintained at 37°C using a homeothermic control unit (Harvard Apparatus). Anaesthetized mice were placed in a stereotaxic frame, the skull was exposed through a midline scalp incision and Rose Bengal (200 µl, 10 mg/ml solution in sterile saline, Sigma) was injected intraperitoneally. After 5 min, the skull and underlying brain tissue were illuminated for 12 min by a 2 mm diameter cold laser beam (50 mW, 561 nm; Cobolt AB) positioned at anterior-posterior (AP) +0.5 mm and medial-lateral (ML) −2.7 mm relative to Bregma, targeting the border between left primary somatosensory and motor cortex (Porritt et al., 2012). For intervention experiments targeting motor cortex, stroke was induced as above with the following modifications: transcranial illumination lasting 15 min was delivered using cold light source (LQ1600, Fiberoptic-Heim AG) equipped with 2 mm wide fibre optic probe and directed to AP +0 and ML −1.5 relative to Bregma. After illumination, the scalp was sutured and mice were placed in a warm cage for 45 min to recover from anaesthesia prior to being returned to the home cage. Mice were provided with moist mashed food placed on the floor of the home cage and their weight was monitored daily for 7 days after surgery. There were no significant differences in body weight, body temperature (36.5 ± 0.5°C), duration of surgery, or post-stroke mortality between cohorts. Each cohort contained an even distribution of mice from the matched strains or treatment groups.
Intranasal treatment
Purified human C3a peptide (Complement Technologies) was diluted in sterile phosphate-buffered saline (PBS) to a concentration of 200 nM and a total of 20 µl (10 µl/nostril; corresponding to ∼1.13 µg/kg body weight) of peptide solution or PBS was given intranasally to awake, hand-restrained mice held in a supine position. Solutions were administered through a pipette tip, drop-wise in 5-µl portions divided by 1-min intervals to allow for absorption. C3a or PBS was given every 24 h on Days 7 to 21 post-stroke for the short-term study or on Days 7 to 28 post-stroke for the long-term study. Mice were assigned to C3a or PBS treatment using randomization stratified by body weight to avoid potential confounding effects of body weight on behavioural performance. The investigators carrying out behavioural studies and analysing data were blinded to treatment group. For the assessment of potential systemic anaphylactic response due to intranasal C3a inoculation, body temperature was monitored using a rectal temperature probe (Harvard Apparatus) inserted ∼4 mm into the rectum of awake mice restrained by the scruff. Baseline temperature was taken before intranasal administration and 5, 15, 30, 45 and 75 min after C3a or PBS administration.
In vivo epifluorescent imaging
C3a (Complement Technologies) was labelled with VivoTag® XL 680 fluorescent tag (Perkin Elmer) and purified according to manufacturer’s instructions. A minimum of 10 µg of labelled C3a (0.4 mg/kg of body weight) was determined in a pilot experiment to be necessary for reliable detection of the fluorescent signal in live animals due to its significant attenuation by skull bones. Mice received 20 µl of PBS or 65 µM C3a-VivoTag® (0.48 mg/kg of body weight) intranasally as described above. One and 3 h later, mice were anaesthetized with 2% isoflurane and imaged in the IVIS® Lumina III Bio-imaging platform (Caliper Technologies). After imaging, mice were deeply anaesthetized with isoflurane and killed by cervical dislocation for ex vivo imaging. Brains were quickly dissected from the skull and imaged using the same fluorescent filter sets. All mice and brains were imaged simultaneously with the PBS-treated control (acting as the tissue autofluorescence reference) placed in the middle, to minimize the potential confound of weaker illumination toward the sides of the observation field. Acquired images were processed and analysed using Living Image software (Caliper Technologies). Epifluorescent signal intensities are presented as radiant efficiency [(photons/s/cm2/sr)/(μW/cm2)] after subtraction of the residual tissue autofluorescence signal defined by the PBS control.
Tissue preparation and infarct volume measurements
Twenty-one days after ischaemia induction, mice were deeply anaesthetized with thiopental (Hospira) and transcardially perfused with 0.9% saline, followed by 4% paraformaldehyde in 0.1 M PBS. Brains were removed and immersed in the same fixative overnight. Tissue was dehydrated, embedded in paraffin, and cut into 6-μm serial coronal sections. Every 20th section was stained with haematoxylin and eosin. Infarct size was evaluated morphometrically on digital images with ImageJ software (NIH, v. 1.47q) by manual delineation of the infarct and hemisphere areas on sections spanning the entire lesion along the anterior-posterior axis by an investigator blinded to experimental group. Volume of injury was derived by multiplying area of total tissue loss that includes shrinkage due to scarring [(contralesional hemisphere − ipsilesional hemisphere) + infarcted tissue] on each section by the total intersection distance.
Tissue immunostaining
For immunofluorescent evaluation, sections were deparaffinized, heated three times for 5 min in a microwave oven in 0.01M citric buffer (pH 6.0), and blocked in PBS containing 0.05% Tween-20 (Sigma) and 1% in bovine serum albumin (BSA, Sigma) for GAP43 and synapsin I staining, 3% normal goat serum for VGLUT1 staining or 4% normal donkey serum for GAP43 double stainings. Goat anti-synapsin Ia/b antibody (1:150; Santa Cruz, sc-7379) was followed by biotin-conjugated donkey anti-goat immunoglobulin (Ig) secondary antibody (1:200; Jackson Research Lab, 705-065-147) and Cy3-conjugated streptavidin (1:100; Sigma). Guinea pig anti-VGLUT1 (1:500; Millipore, AB5905) was followed by Alexa Fluor® 488 goat anti-guinea pig Ig (1:500; Molecular Probes, A11073). For single staining, mouse anti-GAP43 antibody (1:1000; Millipore, MAB347), was followed by biotinylated rabbit anti-mouse Ig secondary antibody (1:200; Dako, E0354) and Cy3-conjugated streptavidin (1:100; Sigma). For double stainings, mouse anti-GAP43 antibody (1:250; Millipore) together with either rabbit anti-synaptophysin antibody (1:200; Millipore, 04-1019), rabbit anti-β3-tubulin (1:200; Covance, Covance PRB-435P), rabbit anti-GFAP (1:200; Dako, Z0334) or rabbit-anti-S100β (1:200; Dako, Z0311), were followed by a mixture of donkey anti-mouse-Alexa555 (1:250, Molecular Probes, A31570), donkey anti-rabbit-Alexa488 (1:250 for synaptophysin and β3-tubulin or 1:2000 for S100β and GFAP; Molecular Probes, A11034) and DAPI (0.5 µg/ml; Molecular Probes, D1306). All antibodies and dye-conjugates were diluted in the respective blocking buffer. Sections representing all experimental groups were stained simultaneously, when more than one round of staining was necessary due to large number of slides. Sections stained with only the secondary antibody served as negative control, and no signal was observed for any secondary antibody including antibodies against mouse Ig (Supplementary Figs 1D’, H’ and 5F’).
Image acquisition and analysis
Highest signal intensity single plane images of immunostained sections were obtained by laser scanning confocal microscope (LSM TCS SP2, Leica Microsystems, ×63/NA 1.3 objective for synapsin I and GAP43; and LSM 700, Carl Zeiss for VGLUT1 ×40/NA 1.3) at 1024 × 1024 pixels resolution. Images from a 238 µm × 238 µm optical field (synapsin I and GAP43) or 160 µm × 160 µm (for VGLUT1) were taken from two adjacent but not overlapping optical fields (referred to as proximal and distal) in the medial (motor) and lateral (somatosensory) peri-infarct cortex, each at superficial (I–IV) and deep (V–VI) cortical layers as well as at two depths in medial and lateral dorsal striatum (total of four images per region). Corresponding images were taken in the contralesional hemisphere and corpus callosum, the latter serving as an internal background control. Images were acquired in a standardized way including controlled and standardized exposure time and number of exposures. As there were no significant differences in the parameters of fluorescence-positive GAP43+ puncta between cortical layers, these data were pooled and expressed as values per entire region. Similar values of all parameters for punctate staining in medial and lateral regions within the contralesional hemisphere were obtained for mice with sensorimotor stroke, so these values were pooled. Three standard sections per animal in 160-µm intervals were analysed. All sections were scanned with the same acquisition parameters.
For co-localization analysis, peri-infarct region of sections double-stained for GAP43 and neuronal or glial markers were imaged with ×63/NA 1.4 objective (LSM 700, Zeiss) using sequential scanning mode with a 20-nm wide exclusion window at emission spectra overlap to avoid any potential mixing of signal from the two channels. Images were collected as Z-stacks (voxel size: 0.09 µm × 0.09 µm × 0.34 µm − optical thickness) using 16-bit colour space.
Single-stained images were analysed using MetaMorph® software (Molecular Devices, v. 2.8.5) to obtain number, average size and intensity of positive punctuate structures per image. Average intensity per punctum was highly homogenous between the groups and regions; therefore this measure was not pursued further. Co-localization analysis was performed with ImageJ (ver. 2.0; Coloc2 plugin) using automatic thresholding and statistical verification of non-random findings (estimated probability of random co-localization P = 1.0) according to Costes’ method, following background subtraction. An experimenter blinded to experimental group performed all image acquisitions and quantifications.
Behavioural assessment
Functional impairment of mice treated intranasally with C3a or vehicle was assessed using a modified cylinder test and a grid walking task to closely evaluate forepaw function. Mice were tested once on each task 1 week before stroke induction to establish baseline performance. Next, they were tested on Days 7, 14 and 21 (short-term study) or Days 7, 14, 28, 42 and 56 (long-term study) post-stroke. Behavioural assessments were carried out at approximately the same time each day, during the first half of light cycle. The experimenter scoring behaviour was blinded to treatment group. Due to larger than expected behavioural variation within groups in terms of scores at baseline and impairment following stroke, and in order to increase power for detection of differences, plotted scores are presented and analysed as ratio between score on a particular day and baseline score for each individual mouse (i.e. as fold of baseline performance).
Grid walking task
Mice were allowed to walk on a 35 cm × 25 cm wire grid with 11-mm square mesh fixed 60 cm above the lab bench for 5 min as described previously (Baskin et al., 2003). A camera was placed beneath the grid to record video for later assessment of stepping errors (foot faults). Total foot faults for each forelimb, along with non-foot fault steps for that forelimb, were counted during frame-by-frame analysis of the videos. A ratio between the number of foot faults and total number of steps taken for the affected paw was calculated [n R foot faults / (n R foot faults + n R non-foot fault steps)]. A step was considered a foot fault if it was not providing support and the foot passed through the grid hole. If an animal was resting with the grid at the level of the wrist after a foot slip, this was also considered a fault. Foot fault scores are presented as fold of baseline performance.
Spontaneous forelimb asymmetry task (cylinder test)
The method of Schallert et al. (2000) was used with minor modifications. Mice were videotaped with an HD digital camera while rearing in a 15-cm wide Plexiglas® cylinder until they performed 10 rears (5–10 min). Two mirrors were arranged at 90° angle and placed behind the cylinder to assist with detailed analysis of all movements. All paw contacts with the cylinder wall during vertical exploration were scored on videos played back frame-by-frame. Due to marked muscle weakness resulting from injury to primary motor cortex, mice often place the other paw to support the body while rearing after initial single paw contact. Therefore, forelimb asymmetry index for mice was calculated as the percentage of individual right (affected) paw touches to total paw touches [n R contacts/(n R contacts + n L contacts + n both paws contacts)]. Asymmetry score is presented as fold of baseline performance.
Statistical analysis
Sample size required for detection of significant differences with 80% power and significance level at α = 0.05 was determined in a pilot study using wild-type untreated animals, and was estimated to be between 9 and 12, for infarct size compressions and behavioural experiments, respectively. Longitudinal behavioural data were analysed by two-way repeated measures ANOVA followed by Dunnett’s post hoc tests for within-group comparisons between specific time points or Sidak’s post hoc tests for between-group comparisons at particular time points. Other types of data were analysed by unpaired t-test for comparisons between two groups or one-way ANOVA followed by planned multiple comparisons using Sidak’s method for comparisons between more than two groups. For datasets with non-Gaussian distribution, as determined by omnibus K2 normality test, non-parametric equivalents of the above-mentioned tests were used. Specifically, behavioural data were analysed by Friedmann’s test followed by Dunn’s post hoc test or Wilcoxon signed ranked test for within-group comparisons, and Mann-Whitney U-test for between-group comparisons at individual time points. For other comparisons, a Kruskal-Wallis test followed by Dunn’s post hoc analysis was applied. Association between density and size of synapsin I+ puncta and behavioural performance was determined by simple linear regression. Pearson’s linear correlation was used to determine the association between density of synapsin I+ and VGLUT1+ puncta. Reported P-values are adjusted for multiple comparisons where applicable. Data are presented as mean ± standard error of the mean (SEM) or median ± interquartile range (IQR). All analyses were two-tailed, and P-values < 0.05 were considered statistically significant. Analyses were performed in Prism (GraphPad Inc.; ver. 6.05f).
Exclusion of data points
In the analysis of neural plasticity markers, occasional extreme values scored for single images that were confirmed to be due to tissue section artefacts were excluded from the analysis. Such outliers were defined as values differing by >2 SD from the mean value for the parameter or by >1.5 IQR from the median for non-normally distributed data. Animals that displayed marked left paw preference (left/right paw contacts >1.50 versus median ratio of 0.94) in the cylinder task at baseline and consequently did not show a significant impairment in the targeted (right) paw function after stroke, despite the presence of lesion of the expected size, were excluded from the analysis of this behavioural task. This criterion was established prior to the study and was based on our previous observations. As these mice did not display any marked difference in other parameters, they were not excluded from the remaining analyses so as not to unnecessarily reduce the group size. Distribution of outliers was comparable between experimental groups and the numbers of included and excluded animals are reported in the figures and figure legends.
Results
Signalling through C3aR positively regulates the number of synapses in the contralesional hemisphere
Signalling through C3aR stimulates an increase in synaptic density in the contralesional cortex. (A) Study design. (B) Haematoxylin-eosin stained coronal section through the approximate centre of the infarct showing infarct size and location. Scale bar = 1 mm. M1 and M2 = primary and secondary motor cortex, respectively; S1FL and S1HL = forelimb and hindlimb field of primary somatosensory cortex, respectively. (C and D) Infarct volume at 21 days post-stroke in C3aR−/− mice, GFAP-C3a mice, and their respective controls. (E) Schematic diagram indicating cortical regions chosen for analysis. CC = corpus callosum; Ctx = cortex; Str = striatum. (F) Representative confocal images of proximal peri-infarct and contralesional cortex stained with antibody against synapsin I at 21 days after stroke (images show standard segments of acquired and analysed images and correspond to layers II/III of somatosensory cortex). Scale bar = 10 µm. (G and H) Density of synapsin I+ puncta in the proximal peri-infarct and contralesional cortex (mean ± SEM; C3aR+/+ n = 10, C3aR−/− n = 14, wild-type n = 13, GFAP-C3a n = 12). One-way ANOVA with Sidak’s planned comparisons: **P < 0.01, ***P < 0.001, ****P < 0.0001 for ipsilateral versus contralateral comparisons; #P < 0.05, ##P < 0.01 for between-genotype comparisons. contra = contralesional cortex; ipsi M = ipsilesional motor cortex; ipsi S = ipsilesional somatosensory cortex; WT = wild-type.
Signalling through C3aR stimulates an increase in synaptic density in the contralesional cortex. (A) Study design. (B) Haematoxylin-eosin stained coronal section through the approximate centre of the infarct showing infarct size and location. Scale bar = 1 mm. M1 and M2 = primary and secondary motor cortex, respectively; S1FL and S1HL = forelimb and hindlimb field of primary somatosensory cortex, respectively. (C and D) Infarct volume at 21 days post-stroke in C3aR−/− mice, GFAP-C3a mice, and their respective controls. (E) Schematic diagram indicating cortical regions chosen for analysis. CC = corpus callosum; Ctx = cortex; Str = striatum. (F) Representative confocal images of proximal peri-infarct and contralesional cortex stained with antibody against synapsin I at 21 days after stroke (images show standard segments of acquired and analysed images and correspond to layers II/III of somatosensory cortex). Scale bar = 10 µm. (G and H) Density of synapsin I+ puncta in the proximal peri-infarct and contralesional cortex (mean ± SEM; C3aR+/+ n = 10, C3aR−/− n = 14, wild-type n = 13, GFAP-C3a n = 12). One-way ANOVA with Sidak’s planned comparisons: **P < 0.01, ***P < 0.001, ****P < 0.0001 for ipsilateral versus contralateral comparisons; #P < 0.05, ##P < 0.01 for between-genotype comparisons. contra = contralesional cortex; ipsi M = ipsilesional motor cortex; ipsi S = ipsilesional somatosensory cortex; WT = wild-type.
To assess post-stroke changes in synaptic and axonal plasticity, we visualized the presynaptic terminals by immunostaining with antibodies against a pan-synaptic marker synapsin I (Fig. 1F) and used high-content image analysis to quantify synapsin I immunoreactive puncta in the peri-infarct region and the corresponding regions of the contralesional hemisphere. In all four experimental groups, we found significantly higher density of synapsin I+ puncta in the injured cortex proximal to the infarct (Fig. 1G and H), and this difference was more pronounced in somatosensory cortex than in motor cortex. Although the density of synapsin I+ puncta in the ipsilesional hemisphere was comparable between groups, it was reduced by 50% in the contralesional cortex of C3aR−/− mice (P = 0.001, P = 0.032 in superficial and deep cortical layers, respectively, Fig. 1G). In all experimental groups, the average size of synapsin I+ puncta in the infarct-proximal region was increased in the ipsilesional compared with the contralesional cortex (P < 0.01, Supplementary Fig. 1A and B). There was a marked overall reduction in the density of synapsin I+ puncta in the contralesional cortex of C3aR−/− mice compared with C3aR+/+ mice in areas corresponding to distal peri-infarct regions (P < 0.001, Supplementary Fig. 2A). C3aR−/− mice had fewer synapsin I+ puncta in the deep layers of distal ipsilesional (i.e. secondary) motor cortex than C3aR+/+ mice (P = 0.036, Supplementary Fig. 2A). GFAP-C3a mice did not differ from their wild-type littermates in the density of synapsin I+ puncta in any of the cortical regions assessed (Fig. 1H and Supplementary Fig. 2B).
The average size of synapsin I+ puncta in the deeper layers of the ipsilesional secondary motor cortex was significantly smaller in C3aR−/− mice compared with C3aR+/+ mice (P = 0.014, Supplementary Fig. 2C). The differences between C3aR−/− and C3aR+/+ mice in synapsin I expression in the contralesional hemisphere appear to be induced by brain ischaemia, as the density and size of synapsin I+ puncta did not differ between genotypes in age-matched naïve mice (Supplementary Fig. 3). The markedly lower density and size of synapsin I+ puncta in naïve compared with injured mice point to active involvement of the contralesional hemisphere in stroke-induced synaptic remodelling.
Signalling through C3aR stimulates an increase in the density of glutamatergic synapses in the peri-infarct region. (A) Representative confocal images of proximal peri-infarct and contralesional cortex stained with antibody against VGLUT1 at 21 days after stroke (images show standard segments of acquired and analysed images and correspond to cortical layer V). Scale bar = 10 µm. (B and C) Density of VGLUT1 + puncta in the proximal peri-infarct and contralesional cortex (mean ± SEM; C3aR+/+ n = 6, C3aR−/− n = 6, wild-type n = 6, GFAP-C3a n = 7). One-way ANOVA with Sidak’s planned comparisons: *P < 0.05, ***P < 0.001, ****P < 0.0001 for ipsi versus contra comparisons; #P < 0.05, ##P < 0.01 for between-genotype comparisons. contra = contralesional cortex; ipsi M = ipsilesional motor cortex; ipsi S = ipsilesional somatosensory cortex; WT = wild-type.
Signalling through C3aR stimulates an increase in the density of glutamatergic synapses in the peri-infarct region. (A) Representative confocal images of proximal peri-infarct and contralesional cortex stained with antibody against VGLUT1 at 21 days after stroke (images show standard segments of acquired and analysed images and correspond to cortical layer V). Scale bar = 10 µm. (B and C) Density of VGLUT1 + puncta in the proximal peri-infarct and contralesional cortex (mean ± SEM; C3aR+/+ n = 6, C3aR−/− n = 6, wild-type n = 6, GFAP-C3a n = 7). One-way ANOVA with Sidak’s planned comparisons: *P < 0.05, ***P < 0.001, ****P < 0.0001 for ipsi versus contra comparisons; #P < 0.05, ##P < 0.01 for between-genotype comparisons. contra = contralesional cortex; ipsi M = ipsilesional motor cortex; ipsi S = ipsilesional somatosensory cortex; WT = wild-type.
Taken together, these data suggest that C3aR is important for the post-stroke increase in the number of presynaptic glutamatergic terminals, and possibly synapses, and this response is cortical region- and layer-specific.
Signalling through C3aR positively regulates the expression of GAP43, a marker of axonal, presynaptic and glial plasticity
Signalling through C3aR stimulates post-stroke GAP43 expression in the cortex. (A) Representative confocal images of proximal peri-infarct and contralesional cortex stained with antibody against GAP43 at 21 days after stroke. Each panel shows a standard segment of an acquired and analysed image and corresponds to layers II/III of somatosensory (C3aR−/− and C3aR+/+) or motor cortex (wild-type and GFAP-C3a). Scale bar = 10 µm. (B and C) Density and (D and E) average area of GAP43+ puncta in the proximal peri-infarct and contralesional cortex (mean ± SEM; C3aR+/+ n = 10, C3aR−/− n = 14, wild-type n = 13, GFAP-C3a n = 12). One-way ANOVA with Sidak’s planned comparisons: *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001 for ipsi versus contra comparisons; #P < 0.05, ##P < 0.01 for between-genotype comparisons. contra = contralesional cortex; ipsi M = ipsilesional motor cortex; ipsi S = ipsilesional somatosensory cortex; WT = wild-type.
Signalling through C3aR stimulates post-stroke GAP43 expression in the cortex. (A) Representative confocal images of proximal peri-infarct and contralesional cortex stained with antibody against GAP43 at 21 days after stroke. Each panel shows a standard segment of an acquired and analysed image and corresponds to layers II/III of somatosensory (C3aR−/− and C3aR+/+) or motor cortex (wild-type and GFAP-C3a). Scale bar = 10 µm. (B and C) Density and (D and E) average area of GAP43+ puncta in the proximal peri-infarct and contralesional cortex (mean ± SEM; C3aR+/+ n = 10, C3aR−/− n = 14, wild-type n = 13, GFAP-C3a n = 12). One-way ANOVA with Sidak’s planned comparisons: *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001 for ipsi versus contra comparisons; #P < 0.05, ##P < 0.01 for between-genotype comparisons. contra = contralesional cortex; ipsi M = ipsilesional motor cortex; ipsi S = ipsilesional somatosensory cortex; WT = wild-type.
Intranasal C3a treatment improves functional recovery in wild-type mice
To determine the therapeutic potential of C3a in a clinically relevant scenario, we next investigated whether delayed treatment with C3a affects functional recovery and neural plasticity processes in wild-type mice.
Intranasal administration allows for repeated, rapid, and non-invasive delivery of peptides to the brain. Since the transfer of molecules occurs mainly via peri-vascular bulk flow along olfactory and trigeminal nerves, this method does not rely on crossing the blood–brain barrier and allows peptides to reach CSF within minutes (Lochhead and Thorne, 2012). As C3aR activation is known to cause histamine release from basophiles and mast cells in a similar way as stimulation by IgE, we monitored body temperature change to verify that intranasal administration of C3a peptide does not cause systemic hypersensitivity or anaphylaxis (Kind, 1955). The transient (5–10 min) and very small drop in body temperature observed after administration of C3a or PBS (Supplementary Fig. 6) is consistent with a general response to intranasal administration of a non-sensitizing agent (Fang et al., 2013) and indicates the absence of an adverse systemic response to C3a.
Intranasal C3a stimulates recovery of forepaw function after stroke. (A) Translocation from the nasal cavity to the brain of fluorescently tagged C3a in anaesthetized mice (top) at 1 and 3 h after intranasal administration (visualized in IVIS® Lumina III Bio-imaging platform). Confirmatory ex vivo imaging (bottom) of dissected brains of the same mice. Colour scale = epifluorescent signal measured as radiant efficiency with a cut-off at the level of tissue autofluorescence in PBS control. (B) Study design. (C) Haematoxylin and eosin-stained coronal section through the approximate centre of the motor cortex infarct showing infarct size and location. Scale bar = 1 mm; M1 and M2 = primary and secondary motor cortex, respectively; S1FL and S1HL = forelimb and hindlimb field of primary somatosensory cortex, respectively. (D) Infarct volume at 21 days post-stroke is not affected by the treatment. Mean ± SEM, n = 14 mice/treatment group. (E) Left: A typical foot fault during grid walking task; right: fold change relative to baseline performance in right (R, affected) paw foot faults over time in the grid walking task. (F) Left: an example of behaviour scored in the cylinder test. Right: fold change relative to baseline performance in right forepaw usage while rearing in the cylinder test over time (four mice/group excluded). Grey bar indicates treatment period. Mean ± SEM. **P < 0.01, ***P < 0.001 post-stroke versus baseline performance of C3a-treated mice; §P < 0.05, §§P < 0.01 §§§P < 0.001 post-stroke versus baseline performance of PBS-treated mice; determined by two-way ANOVA repeated measures and Dunnett’s test for within-group comparisons.
Intranasal C3a stimulates recovery of forepaw function after stroke. (A) Translocation from the nasal cavity to the brain of fluorescently tagged C3a in anaesthetized mice (top) at 1 and 3 h after intranasal administration (visualized in IVIS® Lumina III Bio-imaging platform). Confirmatory ex vivo imaging (bottom) of dissected brains of the same mice. Colour scale = epifluorescent signal measured as radiant efficiency with a cut-off at the level of tissue autofluorescence in PBS control. (B) Study design. (C) Haematoxylin and eosin-stained coronal section through the approximate centre of the motor cortex infarct showing infarct size and location. Scale bar = 1 mm; M1 and M2 = primary and secondary motor cortex, respectively; S1FL and S1HL = forelimb and hindlimb field of primary somatosensory cortex, respectively. (D) Infarct volume at 21 days post-stroke is not affected by the treatment. Mean ± SEM, n = 14 mice/treatment group. (E) Left: A typical foot fault during grid walking task; right: fold change relative to baseline performance in right (R, affected) paw foot faults over time in the grid walking task. (F) Left: an example of behaviour scored in the cylinder test. Right: fold change relative to baseline performance in right forepaw usage while rearing in the cylinder test over time (four mice/group excluded). Grey bar indicates treatment period. Mean ± SEM. **P < 0.01, ***P < 0.001 post-stroke versus baseline performance of C3a-treated mice; §P < 0.05, §§P < 0.01 §§§P < 0.001 post-stroke versus baseline performance of PBS-treated mice; determined by two-way ANOVA repeated measures and Dunnett’s test for within-group comparisons.
Intranasal C3a stimulates neural plasticity in the peri-infarct cortex
Intranasal C3a increases synaptic density in the ipsilesional cortex 21 days after stroke. (A) Representative images of peri-infarct and contralesional somatosensory and motor cortex (standard segments corresponding to layers II/III) stained with antibody against synapsin I and VGLUT1 at 21 days after stroke. Scale bar = 10 µm. (B) Schematics indicating cortical regions chosen for analysis. CC = corpus callosum; Ctx = cortex; Str = striatum. (C) Quantification of synapsin I+ puncta. (D) Quantification of VGLUT1+ puncta. Mean ± SEM; n = 7 mice/treatment group. One-way ANOVA with Sidak’s planned comparisons: *P < 0.05, **P < 0.01, ****P < 0.0001 for ipsilesional versus contralesional hemisphere comparisons; #P < 0.05, ##P < 0.01, ###P < 0.001, ####P < 0.0001 for between-treatment comparisons. C = contralesional cortex; I = ipsilesional cortex.
Intranasal C3a increases synaptic density in the ipsilesional cortex 21 days after stroke. (A) Representative images of peri-infarct and contralesional somatosensory and motor cortex (standard segments corresponding to layers II/III) stained with antibody against synapsin I and VGLUT1 at 21 days after stroke. Scale bar = 10 µm. (B) Schematics indicating cortical regions chosen for analysis. CC = corpus callosum; Ctx = cortex; Str = striatum. (C) Quantification of synapsin I+ puncta. (D) Quantification of VGLUT1+ puncta. Mean ± SEM; n = 7 mice/treatment group. One-way ANOVA with Sidak’s planned comparisons: *P < 0.05, **P < 0.01, ****P < 0.0001 for ipsilesional versus contralesional hemisphere comparisons; #P < 0.05, ##P < 0.01, ###P < 0.001, ####P < 0.0001 for between-treatment comparisons. C = contralesional cortex; I = ipsilesional cortex.
Intranasal C3a increases GAP43 expression in the ipsilesional cortex 21 days after stroke. (A) Representative confocal images of peri-infarct and contralesional motor cortex stained with antibody against GAP43 at 21 days after stroke. Scale bar = 10 µm. (B) Quantification of GAP43+puncta. Mean ± SEM; n = 14 mice/treatment group. One-way ANOVA with Sidak’s planned comparisons: **P < 0.01, ****P < 0.0001 for ipsilesional versus contralesional hemisphere comparisons; #P < 0.05 for between-treatment comparisons. C = contralesional cortex; I = ipsilesional cortex.
Intranasal C3a increases GAP43 expression in the ipsilesional cortex 21 days after stroke. (A) Representative confocal images of peri-infarct and contralesional motor cortex stained with antibody against GAP43 at 21 days after stroke. Scale bar = 10 µm. (B) Quantification of GAP43+puncta. Mean ± SEM; n = 14 mice/treatment group. One-way ANOVA with Sidak’s planned comparisons: **P < 0.01, ****P < 0.0001 for ipsilesional versus contralesional hemisphere comparisons; #P < 0.05 for between-treatment comparisons. C = contralesional cortex; I = ipsilesional cortex.
Intranasal C3a leads to a faster and sustained functional recovery
Intranasal C3a leads to faster and sustained recovery of forepaw function. (A) Study design. (B) Right (R, affected) paw foot faults in the grid walking task, presented as fold change relative to baseline performance, over time. Data presented as median ± IQR; #P < 0.05, ##P < 0.01 for between-groups comparison, Mann-Whitney U-test. (C) Grid walking task, change in the performance of individual mice in between Days 7 and 14 after stroke. §P < 0.05, Wilcoxon signed rank test. (D) Right forepaw usage while rearing in the cylinder test over time (two and three mice excluded in C3a and PBS group, respectively), mean ± SEM. (E) Cylinder test, change in the performance of individual mice between Days 7 and 56 post-stroke. §§P < 0.01, paired t-test. Comparisons of individual time points against baseline and Day 7 post stroke are not denoted in B and D, refer to the ‘Results’ section for details. Grey bar indicates treatment period.
Intranasal C3a leads to faster and sustained recovery of forepaw function. (A) Study design. (B) Right (R, affected) paw foot faults in the grid walking task, presented as fold change relative to baseline performance, over time. Data presented as median ± IQR; #P < 0.05, ##P < 0.01 for between-groups comparison, Mann-Whitney U-test. (C) Grid walking task, change in the performance of individual mice in between Days 7 and 14 after stroke. §P < 0.05, Wilcoxon signed rank test. (D) Right forepaw usage while rearing in the cylinder test over time (two and three mice excluded in C3a and PBS group, respectively), mean ± SEM. (E) Cylinder test, change in the performance of individual mice between Days 7 and 56 post-stroke. §§P < 0.01, paired t-test. Comparisons of individual time points against baseline and Day 7 post stroke are not denoted in B and D, refer to the ‘Results’ section for details. Grey bar indicates treatment period.
A similar positive effect of C3a treatment on post-stroke functional recovery was observed in the cylinder task. The average scores on the last day of testing showed only a trend toward a difference between groups (Sidak’s test: P = 0.0687), although the C3a-treated mice displayed a sustained functional improvement compared with Day 7 at Days 28 and 56 post-stroke (Dunnett’s test: P = 0.032 and P = 0.0019, respectively) while changes in performance of PBS-treated mice were inconsistent and not statistically significant (Dunnett’s test: P = 0.982 for Day 56; Fig. 7D). Also, paired analysis of individual mice showed that C3a-treated animals readily increased their affected paw usage between the treatment initiation and 4 weeks after the completion of the treatment period (paired t-test: P = 0.009; 8/10 of mice improved), while in PBS-treated mice, overall right paw impairment did not change (paired t-test: P = 0.673; only 3/9 of mice improved, Fig. 7E). Taken together, these data indicate that intranasal treatment with C3a supports faster and more complete motor function recovery, which is sustained beyond the treatment period.
Discussion
In the present study, we evaluated the role of C3a and C3aR signalling in stroke-induced neural plasticity. We found that C3a overexpression in reactive astrocytes increased, whereas C3aR deficiency decreased expression of GAP43, a marker of post-stroke axonal, synaptic, and glial plasticity, without affecting the infarct size. Moreover, C3aR deficiency was associated with reduced expression of synapsin I, a structural element of presynaptic terminals and a marker of synaptic plasticity, as well as VGLUT1, a presynaptic marker of the majority of glutamatergic synapses (Micheva et al., 2010). Intranasal treatment with C3a starting 7 days post-stroke robustly upregulated the expression of neural plasticity markers and was associated with faster and sustained functional recovery in wild-type mice.
We have previously shown that C3-deficient mice had increased infarct volume at 7 and 21 days after permanent middle cerebral artery occlusion (MCAO) (Rahpeymai et al., 2006) and that GFAP-C3a mice were strongly protected from neonatal hypoxic ischaemic brain injury (Järlestedt et al., 2013). To study the role of C3a and C3aR in stroke-induced neural plasticity and functional recovery, we used the photothrombotic stroke model, which results in an irreversibly damaged ischaemic core in the targeted cortical region and a relatively narrow penumbra with limited possibilities of collateral blood flow. Consequently, this model offers high reproducibility of stroke location with a small infarct size, facilitating study of regeneration processes while evoking a similar early cellular response as the permanent MCAO model (Schroeter et al., 1994; Jander et al., 1995). Our findings that the infarct volume was not affected by the overexpression of C3a or the absence of C3aR are therefore not surprising and do not preclude a possible role of C3a and C3aR in neuroprotection or ischaemia-induced tissue injury.
Synaptic plasticity and functional remapping involving both the peri-infarct regions and the contralesional hemisphere are believed to play a critical role in the recovery of function after stroke (reviewed in Pekna et al., 2012). Axonal plasticity is a hallmark of regenerative plasticity and a mechanism that ultimately leads to the emergence of new synapses after an ischaemic insult. This phenomenon is associated with reactivation of the intrinsic neuronal growth program and robust upregulation in the peri-infarct cortex of the membrane phosphoprotein GAP43 (Carmichael et al., 2005), which associates with axonal growth cones and is used as marker of axonal sprouting (Benowitz et al., 1990; Benowitz and Routtenberg, 1997). GAP43 is also upregulated during reactive synaptogenesis (Benowitz et al., 1990; Lin et al., 1992) and involved in presynaptic plasticity through regulation of vesicle trafficking (Hou and Dahlström, 2000) and neurotransmitter release (Dekker et al., 1991). Recently, astrocytic GAP43 was shown to mediate glial plasticity during astrogliosis, attenuate microglial activation, and provide beneficial effects for neuronal survival and plasticity (Hung et al., 2016). Our findings that the expression of GAP43 is reduced in the absence of C3aR and increased when C3a is expressed in reactive astrocytes or administered intranasally, together with previous results showing that C3 is upregulated in sprouting neurons isolated from rat cortex after ischaemic stroke (Li et al., 2010), and that there is a stimulatory effect of C3a on neurite outgrowth in vitro (Shinjyo et al., 2009) support the conclusion that C3a signalling through C3aR plays a positive role in post-stroke neural plasticity, possibly including axonal sprouting.
While many growth-related genes, including GAP43, are induced shortly after ischaemia and expressed for at least 28 days in young adult (2-month-old) mice, GAP43 expression in aged (20-month-old) mice peaks only transiently at around 3 and 14 days post-stroke (Li and Carmichael, 2006). Our findings of a robust increase in C3a-associated GAP43 expression in the ipsilesional motor cortex in up to 9-month-old GFAP-C3a mice 21 days post-stroke indicate that C3a signalling extends the plasticity window and makes the post-stroke brain milieu more permissive for functional recovery.
We further observed a robust and to some extent cortical region- and layer-specific ischaemia-induced increase in the density of glutamatergic synapses. This is in contrast to previous reports focusing on the first month after stroke that found a cortical layer-specific and transient effect of stroke on the density of GABAergic synapses (Hiu et al., 2016) or an early reduction in the density of presynaptic terminals followed by gradual recovery of baseline levels 1 month after stroke (Liauw et al., 2008). These differences between studies are conceivably due to differences in stroke models and quantification methods used. Together with reports of the association between synaptic density and better recovery of function after ischaemic stroke (Chen et al., 2007; Liauw et al., 2008; Cui et al., 2010, 2013), our findings of improved recovery, increased expression of synapsin I and VGLUT1 in C3a-treated mice, and association between synapsin I expression and functional improvement point to increased synaptic density as an important contributor to functional recovery.
Another important finding of our study is the positive effect of C3a-C3aR signalling on synaptic density in the peri-infarct region. Neuronal C3aR is a part of a signalling pathway that results in increased synaptic strength, and treatment with a C3aR antagonist or C3aR deficiency in neurons co-cultured with wild-type astrocytes reduced dendritic complexity (Lian et al., 2015). However, excessive activation of neuronal C3aR alters dendritic morphology and synaptic function (Lian et al., 2015) and in the context of neurotropic viral infection, C3aR is required for the removal of presynaptic terminals by an unidentified mechanism involving microglia (Vasek et al., 2016). The net effect of C3aR activation in the CNS thus appears to depend on the context and on the extent of C3aR activation. The timing of interventions targeting C3aR may therefore need to be carefully optimized.
In light of the role of neuronal C3aR in modulation of synaptic strength and dendritic morphology (Lian et al., 2015), the C3a-C3aR-mediated upregulation of expression of GAP43 and increased number of presynaptic terminals, particularly glutamatergic terminals, observed in our study is conceivably due at least in part to a direct effect of C3a on neurons. However, given that C3aR is also expressed on glial, endothelial, stem, and immune cells, C3a can also exert its effects on post-stroke plasticity indirectly by modulating the functions of these cell types.
As the contralesional hemisphere becomes electrically activated after stroke (Dijkhuizen et al., 2001; Calautti and Baron, 2003), can be a source of transcallosal axonal sprouting (Carmichael and Chesselet, 2002), and shows evidence of synaptic plasticity, it cannot be regarded as a control region for neural plasticity studies. Increased turnover of mushroom-like dendritic spines and synapse number in contralesional somatosensory cortex was associated with establishing a new pattern of electrical circuit activity in the intact hemisphere and functional recovery (Luke et al., 2004; Takatsuru et al., 2009). Moreover, dendritic remodelling in the cortex contralesional to injury is characterized by the presence of enhanced-efficacy perforated and multiple synaptic bouton-containing synapses (Jones, 1999; Luke et al., 2004), both of which are morphologically larger than regular synapses (Toni et al., 1999; Ganeshina et al., 2004). Here, we found that C3aR−/− mice had a standard number of synapsin I+ as well as VGLUT1+ presynaptic terminals, and intranasal C3a treatment increased the density of synapsin I+ terminals in the contralesional cortex. Together with smaller average size of synapsin I+ and VGLUT1+ puncta in the peri-infarct motor cortex of C3aR−/− mice, our findings suggest that C3aR signalling may be important for long-distance synaptic plasticity after stroke. Importantly, we show that the differences in synaptogenic response observed in the contralesional hemisphere are not due to baseline differences between C3aR−/− and C3aR+/+ mice.
Increased synaptogenesis and axonal plasticity provide greater potential for new axono-dendritic connections for neuronal communication and post-stroke circuit rewiring. However, beneficial effects on outcome need to be verified at the functional level. As upregulation of GFAP expression in peri-infarct astrocytes starts within 24 h, peaks around 4 days, and persists for at least 2 months after photothrombotic stroke (Nowicka et al., 2008), a similar temporal pattern could be expected for C3a expression in GFAP-C3a mice. We reasoned that high acute C3a levels produced by reactive astrocytes might not provide an optimal milieu for regeneration. Therefore, to assess the role of C3a on functional recovery and focus on its post-acute effects, we used a pharmacological approach and treated wild-type mice with intranasal C3a starting 7 days after stroke. This C3a treatment, which avoids the potentially deleterious effects of C3a in the acute phase, was associated with increased synaptogenesis and GAP43 expression as well as better recovery of forepaw function. The positive effect of intranasal C3a on functional recovery was sustained even after treatment cessation. It is noteworthy that intranasal treatment in mice requires repeated restraint, which can be regarded as predictable chronic mild stress. This, however, would not be an issue in human patients. Given the profound negative effect of stress on functional recovery from stroke (Walker et al., 2014), the efficacy of intranasal C3a treatment could be underestimated in a mouse model. In light of its anaphylatoxic properties, it is important to note that we did not observe any adverse or systemic effects of intranasal C3a, even after repeated administration. These results show that delayed intranasal treatment with C3aR agonists is an attractive approach to improve functional recovery after ischaemic brain injury.
In conclusion, C3a-C3aR signalling stimulates post-stroke synaptogenesis and axonal plasticity, and intranasal C3a treatment in the post-acute phase after ischaemic stroke improves functional recovery. These findings open new avenues for translational research aiming to promote neural plasticity and recovery after brain injury.
Abbreviation
Acknowledgements
We acknowledge the Centre for Cellular Imaging at the Sahlgrenska Academy, University of Gothenburg for the use of imaging equipment and for the support from the staff. We would also like to thank Dr Noriko Shinjyo for fluorescent labelling of C3a peptide.
Funding
This work was supported by Swedish Research Council (20116), ALF Gothenburg (142821 and 431431), the EU FP 7 Program TargetBraIn (279017), STENA Foundation, W. and M. Lundgren’s Foundation, AFA Insurance, the Swedish Stroke Foundation, Torsten Söderberg’s Foundation, Edit Jacobson’s Foundation, Rune and Ulla Amlöv’s Foundation, the New Zealand Health Research Council.
Supplementary material
Supplementary material is available at Brain online.
