Complement C3a treatment accelerates recovery after stroke via modulation of astrocyte reactivity and cortical connectivity

Despite advances in acute care, ischemic stroke remains a major cause of long-term disability. Approaches targeting both neuronal and glial responses are needed to enhance recovery and improve long-term outcome. The complement C3a receptor (C3aR) is a regulator of inflammation with roles in neurodevelopment, neural plasticity, and neurodegeneration. Using mice lacking C3aR (C3aR–/–) and mice overexpressing C3a in the brain, we uncovered 2 opposing effects of C3aR signaling on functional recovery after ischemic stroke: inhibition in the acute phase and facilitation in the later phase. Peri-infarct astrocyte reactivity was increased and density of microglia reduced in C3aR–/– mice; C3a overexpression led to the opposite effects. Pharmacological treatment of wild-type mice with intranasal C3a starting 7 days after stroke accelerated recovery of motor function and attenuated astrocyte reactivity without enhancing microgliosis. C3a treatment stimulated global white matter reorganization, increased peri-infarct structural connectivity, and upregulated Igf1 and Thbs4 in the peri-infarct cortex. Thus, C3a treatment from day 7 after stroke exerts positive effects on astrocytes and neuronal connectivity while avoiding the deleterious consequences of C3aR signaling during the acute phase. Intranasal administration of C3aR agonists within a convenient time window holds translational promise to improve outcome after ischemic stroke.

intranasal treatment for 2 or 3 weeks (data presented in Figures 3, 4, 5, 7, S1 and S3), the protocol was modified to target the motor cortex as follows. Transcranial illumination was delivered for 15 min with cold light source (LQ1600, Fiberoptic-Heim AG) equipped with 2 mm-wide fiber optic probe and directed to AP 0 and ML -1.5 relative to the bregma. For the MRI study (data presented in Figure 6 and S2), photothrombosis was induced as described (58).
In brief, 150 μl (1.5 mg) of Rose Bengal was injected intraperitoneally followed by 50 mW laser radiation at 561 nm for 15 min at brain coordinates AP 0.5 mm and ML -2.5 to target the primary somatosensory forelimb area and primary motor cortex. The scalp was then sutured, wound was infiltrated with bupivacaine (50 μl, 0.25%, Marcain, Astra Zeneca) and mice were placed in a warm cage for 45 min to recover from anesthesia. Upon return to their home cage, mice were given moist mashed food. Body weight was monitored daily for 7 days after surgery.

Intranasal treatment
The intranasal treatment was done as described (27). Purified human C3a (Complement Technologies) was diluted in sterile phosphate-buffered saline (PBS) to a concentration of 200 nM, and 20 µl (10 µl/nostril; corresponding to approximately 1.13 µg/kg body weight) of C3a 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 at 1-min intervals to allow for absorption. C3a or PBS was given every 24 hours on post-stroke day (P) 7 to P21 (short-term study) or on P7 to P28 (long-term and MRI studies). Mice were assigned to C3a or PBS treatment by randomization stratified by body weight, and experimenters remained blinded for the treatment.

Behavioral assessment of motor function
Motor function of C3aR -/and GFAP-C3a mice and controls was assessed by round beam walk test, in which mice traversed a 60 cm-long round metal beam 1 cm in diameter suspended 60 cm over the lab bench. A foam cushion was placed beneath the beam to soften the landing after falls. A cage similar to the home cage was placed at the end of the beam to encourage directional crossing. Mice were trained to complete the task 72-24 hours before stroke induction and were tested on P2, P7, P14, and P21. Distance traveled until the first fall or within a maximum time of 2 min was measured, and changes in performance over time after stroke were assessed for each group. Due to changes in animal and stroke induction protocols, motor function recovery of the intranasally treated mice was assessed by the cylinder and grid walk tests as described (58). Briefly, mice were habituated to the test environment and procedure 3 times within 2 weeks before stroke induction. Baseline assessment was done 3-4 days before stroke induction, followed by tests on P3, P7, P14, P28, P42, and P56. All tests were video recorded and analyzed to quantify paw foot faults and paw drags (58, 80). Motor function improvement in the grid walk test was determined as the difference between the percentage of front paw foot faults on P7 and P56 ( Figure 3G) or difference between the percentage of front paw foot faults on P28 and P7 ( Figure 6K, inset). Grid walking improvement data in Figure 3G are based on results that were presented in full in a previous publication (27). Investigators who performed the behavioral studies and analyzed data were blinded to genotype and treatment group.

Magnetic resonance imaging
MRI was done at the Max Planck Institute for Metabolism Research, Cologne, using a 94/20USR BioSpec Bruker system with the 660 mT/m B-GA12SHP gradient system, RT-shim and related power supplies, the 1H receive-only mouse brain surface coil, and the 1H resonator 112/072 -operated with ParaVision v6.0.1 (Bruker, BioSpin). To reduce movement artifacts and provide reproducible mouse brain placement, mice were anesthetized with isoflurane (2-3% in 70/30 N2/O2) and head-fixed with tooth and ear bars in an animal carrier. Respiration and body temperature were measured with a custom-made system based on DASYLab (National Instruments) and electronically recorded (Small Animal Instruments). Body temperature was maintained at 37°C with a feedback-controlled water circulation system (medres Gmbh). After initial adjustments (RF power, shim, and B0 field), a high-resolution, whole-brain, T2-weighted RARE sequence (T2w-MRI) was acquired at coronal slice orientation with a RareFactor of 8, repetition time (TR) of 5,500 ms, an effective echo time (TE) of 32.5 ms, voxel size of 0.068 x 0.068 x 0.5 mm 3 , field-of-view of 17.5 x 17.5 mm 2 , a matrix size of 256 x 256 with 28 noncontiguous slices (0.3 mm gap), and a scan time of 5:52 min. With the same slice orientation, DTI was acquired with an 8-shot, spin-echo, echo-planar imaging (EPI) sequence with a b value of 670 s/mm 2 , gradient duration of 3.5 ms, gradient separation of 8 ms, TR of 3,000 ms, TE of 17.5 ms, voxel size of 0.141 x 0.141 x 0.5 mm 3 , field-of-view of 18 x 18 mm 2 , matrix size of 128 x 128 with 20 slices, and acquisition time of 14 min.

Image pre-and postprocessing
For atlas-based lesion mapping and quantification of infarct size and location on P7 after stroke, we used a workflow based on the combination of T2-weighted MRI and the 3D reference atlas, Allen Mouse Brain Common Coordinate Framework, CCFv3 (81). T2-weighted and DTI data were pre-processed with our in-house Python pipeline AIDAmri (82), which included brain extraction, bias-field correction, and two-step image registration with the atlas (81). Stroke masks were semi-automatically segmented with the 3D snake evolution tool ITK-SNAP (83).
Transformation from the registration atlas to T2-weighted MRI was applied to calculate the percentage of infarct per atlas region. Notably, the infarct volume data obtained on P7 reflect the primary lesion together with vasogenic edema.
DTI processing, deterministic fiber tracking, and calculation of diffusion measures fractional anisotropy, axial diffusivity, and radial diffusivity were done as described with the in-house tools AIDAmri and DSI Studio (http://dsi-studio.labsolver.org; (57). DTI global density was calculated as the ratio of detected connections divided by the maximum possible number of connections between all 96 brain regions (i.e., with a global density of 1, all brain regions would be structurally connected with each other).
Individual peri-infarct regions of interest for the diffusion calculations in AIDAmri were generated with a workflow written in Python (v1.1, https://github.com/aswendtlab/Project_C3a_peri-infarct). Briefly, the peri-infarct area was defined as a dilation of the stroke mask by about 15 pixels, yielding an individual irregular ringshaped border area around the stroke core at P7, using the SciPy function binary dilation. This area was aligned to the DTI data by applying the calculated co-registration parameters for individual time points and the template as well as the known relation between the scans (T2 vs. DTI). Template labels were transferred to the DTI space using the known relations and aligned in the same DTI image spaces. Peri-infarct regions of interest were created by including only pixels in the vicinity of the stroke core where template brain regions overlapped with the individual peri-infarct border area. As a result, cortical regions were replaced by the new periinfarct-delimited regions for further analysis. Astrogliosis was quantified by diffusion MRI as described (43).

Tissue preparation and infarct volume quantification
Mice were deeply anesthetized with thiopental (Hospira) and transcardially perfused with 0. All image acquisitions and quantifications were done by experimenter blinded to experimental group.

RNA isolation
Tissue for gene expression analysis was prepared as described (85) with modifications. Briefly, brains from male mice at P7, P14 (WT and C3aR -/-), and P28 (WT) or naive mice (WT) were retrieved and rapidly frozen in isopentane chilled with dry ice and stored at -80°C. Brains were cut into 1 mm-thick slices with a brain matrix and razor blades that were kept cold with dry ice.
Peri-infarct regions of primary motor and somatosensory cortex, homotopic regions in the contralesional cortex, or corresponding uninjured cortex in naive mice were dissected from the resulting slices with a 1 mm ID tissue micropuncher (Reusable Rapid Punch kit, WPI) in a setup that was kept cold with dry ice (85) and stored at -80° C. Tissue micropunches were lysed with Qiazol reagent (Qiagen) and homogenized by passing the lysate 10 times through a microsyringe fitted with a 23G needle and repeating this procedure with a 27G needle. Total RNA was extracted with an RNAeasy Mini Lipid kit (Qiagen), including the on-column DNAase I digestion step, according to the manufacturer's recommendations. The concentration and purity of isolated RNA were assessed with a Nanodrop 1000 spectrophotometer (Thermo Fisher Scientific).

Library preparation, sequencing, and processing
Samples from WT naive mice and from mice at P7 and P14 were selected for RNA sequencing.

Differential expression analysis
Differential expression was analyzed with the R package DEseq2 v.1.30.1 (94). Counts were normalized with the command "DESeq", and the count matrix was limited to data of compared samples. Genes with an average normalized number of reads <20 were removed. Differentially expressed genes (false-discovery rate-adjusted P < 0.1 and log2FC < -0.5 for downregulation and > 0.5 for upregulation) were plotted with EnhancedVolcano package v1.8.0 (95).
Functional enrichment analysis for biological processes within gene ontology terms was done with a web-based toolkit, WebGEstAlT 2019 (webgestalt.org; (96) using default parameters and visualized as a bar graph in "weighted set cover" mode. Gene score was calculated for every gene using DESeq2 output as -log10(pval) multiplied by log2FC and the gene list ranked according to gene score was submitted annotated with entrez IDs. Unadjusted p values were used for gene score calculation due to modest fold-change differences that resulted in a low number of genes with a false-discovery rate-adjusted P value <0.1, precluding meaningful enrichment analysis.
Deconvolution was done with the default settings described above. The reference matrix and the sample matrix were provided in UMI counts.
Overrepresentation analysis for deconvoluted DAAs and GFAP low expression profiles was done in WebGEstAlT (www.webgestal.org) (96), using the most strongly expressed genes [gene expression profile (GEP) score >90] specific for each subpopulation, i.e., genes that were negligibly expressed in other astrocyte subpopulation (GEP score <10). For GFAP low subpopulation, genes with high GEP scores in the Intermediate subpopulation were allowed due to the paucity of highly expressed GFAP low -unique genes ( Figure S1E, online resource) and the high positive correlation in GEPs between these two subpopulations ( Figure S1B, online resource). Gene lists were submitted as gene symbols/entrez IDs, and a built-in murine genome protein-coding database was chosen as reference. Gene ontologies were considered significantly enriched at a false-discovery rate-adjusted P value <0.05.

Quantitative RT-PCR
cDNA from 500 ng of RNA per sample was generated with a GrandScript cDNA synthesis kit (TATAA Biocenter, Gothenburg, Sweden) as recommended by the manufacturer. After cDNA synthesis, samples were diluted to 10 ng/μl of total RNA with nuclease-free water. This concentration was deemed optimal to achieve an assay-specific PCR efficiency of 99-100% for all genes, as judged from dilution series experiments for each of the primer pairs. Formation of a single PCR product for each of the genes was verified by melt curve analysis; the absence of primer dimers was confirmed by the lack of amplification of negative (no template) controls.  (98) as the gene with the most stable expression across all samples. Expression relative to that of Hprt1 was analyzed with Livak's method (99) and reported as fold change relative to naive WT mice.

Statistical analysis
Data are presented as individual values and mean ± SEM. Comparisons were made between time-points relative to stroke induction, the contralesional and ipsilesional cortex, C3aR +/+ and C3aR -/-, WT and GFAP-C3a, and PBS-and C3a-treated mice. Gaussian distribution of the data was verified with D'Agostino-Pearson's test. Imaging and quantitative gene expression (by RT-qPCR) data and estimated cell-type fractions from the deconvolution analysis were analyzed by two-way ANOVA followed by planned comparisons, using Sidak's method for three or more groups or paired t test for two groups. Round beam test data were analyzed by two-way ANOVA with repeated measures (Figure 1 and 2). Due to single time-point data exclusions, mixed-effects modeling approach (residual maximum likelihood) with time and treatment as fixed factors and subjects as random factor was used for grid walk and cylinder test ( Figure 5).
Sidak's correction was used to adjust for multiple comparisons. Association between relative , radial diffusivity (F), and fractional anisotropy (G) in peri-infarct regions in stroke-affected and corresponding contralesional cortical areas (SSp-ul/ll/un and MOp) before induction of stroke (BL) and on P56. Bar plots represent mean ± SEM. PBS, n = 7; C3a n = 10. Two-way mixed-effects analysis with false-discovery rate correction: ***P < 0.001 for comparisons between the time points; ### P < 0.001 for ipsilesional vs. contralesional comparisons.