Complement receptor C3ar1 deficiency does not alter brain structure or functional connectivity across early life development

13 Previous studies suggest that genetic deletion of the complement C3a anaphylatoxin 14 chemotactic receptor (C3ar1), a key component of the innate immune response, 15 influences behaviours associated with psychiatric symptomatology in mice but when 16 and where C3ar1 is needed in the brain is not known. These questions are significant 17 because, as a G-protein-coupled receptor (GPCR), human C3AR1 serves as a 18 potential therapeutic target for disorders associated with complement dysregulation, 19 such as schizophrenia. To provide a brain-wide assessment of developmental C3ar1 20 activity, we used longitudinal tensor-based morphometry (TBM), fractional 21 anisotropy (FA) from diffusion-weighted magnetic resonance imaging (dMRI) and 22 blood oxygen-level dependent functional MRI (BOLD fMRI) in male and female 23 C3ar1-deficient mice and wild-type littermates, with behavioural assessment in 24 adulthood. Unexpectedly, we did not find robust C3ar1-dependent phenotype in 25 any of these measures. Therefore, our study does not support neurodevelopmental 26 hypotheses for C3ar1 which will likely have implications for targeting this receptor 27 in disease. 28

29 The complement system is a conserved immune pathway that participates in host 30 defence through pathogen clearance and regulating inflammation (Chaplin 2020) as 31 well as tissue homeostasis (Kunz and Kemper 2021; West and Kemper 2023), with 32 emerging roles in neurodevelopment, psychiatric disorders and neurodegeneration 33 (Stevens et al. 2007; Hong et al. 2016; Sekar et al. 2016). 34 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 26, 2025. ; https://doi.org/10.1101/2025.04.24.650541doi: bioRxiv preprint 2 The most convincing evidence for the involvement of the complement system in 35 neurodevelopment so far is its genetic association with schizophrenia. Schizophrenia 36 is a complex and highly heritable neurodevelopmental disorder characterised by 37 hallucinations, delusions, and impaired cognition, with symptoms typically emerging 38 in late adolescence or early adulthood (McCutcheon, Reis Marques, and Howes 39 2020; Howes, Bukala, and Beck 2024). Neurobiological hallmarks of schizophrenia, 40 among others, include grey matter loss (Vita et al. 2012) and a reduction in synaptic 41 density (Osimo et al. 2019). Genome-wide association studies (GWAS) of 42 schizophrenia have identified two complement-related risk loci, the complement 43 component 4 A (C4A) structural variant (Sekar et al. 2016), which encodes C4A 44 protein responsible for propagation of complement activation, and the CUB and 45 Sushi Multiple Domains 1 (CMSD1) mutation which encodes a putative 46 complement inhibitor protein (Schizophrenia Working Group of the Psychiatric 47 Genomics Consortium 2014; Baum et al. 2024). Preclinical studies link these 48 mutations to increased brain-specific complement activation and synapse loss (Sekar 49 et al. 2016; Yilmaz et al. 2021; Baum et al. 2024) potentially tying complement 50 activation to synaptic pathology in schizophrenia. Indeed, the C4A risk locus 51 associates with MRI markers of grey matter loss and reduced cognitive performance 52 in humans, even in the absence of neurological disorders (O’Connell et al. 2021). 53 Complement pathway proteins represent promising therapeutic targets for disorders 54 linked to abnormal complement activity. Complement modulation therefore has 55 potential in addressing unmet therapeutic need in schizophrenia, since elevated 56 complement proteins correlate with severity of negative symptoms in psychosis 57 (Byrne et al. 2024), which do not respond to anti-psychotic medication. 58 C3a anaphylatoxin chemotactic receptor (C3aR1), a G-protein coupled receptor 59 (GPCR) bound by complement activation product C3a and the granin family 60 neuropeptide TLQP-21 (Rodriguez et al. 2023), acts downstream of complement 61 activation and stands out as a pharmacologically tractable target for modifying 62 complement activity in the brain (Hauser et al. 2017). In mice, its transcript is 63 predominantly expressed by microglia, with minimal neuronal expression observed 64 in both healthy adult humans and mice according to single-cell transcriptomic 65 analyses (Hammond et al. 2019; Tasic et al. 2016). While its temporal expression 66 patterns are not yet well characterised, it appears to be active during early embryonic 67 development, potentially influencing progenitor cell proliferation (Coulthard et al. 68 2018; Bénard et al. 2008; Hammond et al. 2019). C3aR1 also appears to play a role in 69 facilitating developmental astrocyte phagocytosis in the retina by microglia 70 (Gnanaguru et al. 2023), as well as to regulate microglial reactivity and 71 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 26, 2025. ; https://doi.org/10.1101/2025.04.24.650541doi: bioRxiv preprint 3 neuroinflammation more broadly (Gedam et al. 2023; Ge, Guan, and Wang 2024; 72 Zheng et al. 2021; Lian et al. 2015; Vasek et al. 2016; Chew and Petretto 2019). Brain 73 morphological changes observed in C3ar1-deficient mice further support its 74 neurodevelopmental relevance (Westacott et al. 2021; Pozo-Rodrigálvarez et al. 75 2021), although there is no consensus on the precise neurodevelopmental actions of 76 C3aR1. Addressing this gap is important given its potential as a pharmacological 77 target. 78 C3aR1 may impact brain functions relevant to psychiatric symptomatology since a 79 range of behavioural phenotypes have been reported in C3ar1-deficienct mice. These 80 include abnormal anxiety-like behaviours (Westacott et al. 2022), hyperactivity 81 (Pozo-Rodrigálvarez et al. 2021), cognitive impairment (Coulthard et al. 2018) and 82 resilience to depressive-like behaviours induced by chronic stress or inflammation 83 (Crider et al. 2018; Zhang et al. 2022; Sun et al. 2024). The involvement of C3aR1 in 84 behaviour suggests that it is needed for healthy brain function, but it remains unclear 85 whether the observed phenotypes arise because C3aR1 is needed during 86 development or from ongoing tonic activity, necessitating longitudinal assessment. 87 Importantly, none of the aforementioned studies used control wildtype mice that 88 were littermates of genetically C3ar1-deficient mice, which presents a confound due 89 to the rapidly diverging genetic background between mutant and control in small 90 inbred colonies (Fitch and Atchley 1985), but also because litter environment affects 91 behaviour and brain development (Crews et al. 2009; Jiménez and Zylka 2021; 92 Valiquette et al. 2023). 93 To investigate the potential effects of C3ar1 deficiency, we adopted a global, 94 unbiased approach, conducting a longitudinal study of male and female C3ar1-95 deficient mice and their wild-type littermates during adolescence and adulthood. 96 Using structural and diffusion magnetic resonance imaging (MRI and dMRI), 97 alongside resting state functional connectivity (FC) analysis of the blood oxygen level 98 dependent (BOLD rsf) MRI signal, we aimed to assess whether the requirement for 99 C3aR1 is implicated in brain structural development during adolescence—a critical 100 period for psychiatric vulnerability (Westacott and Wilkinson 2022; Paus, Keshavan, 101 and Giedd 2008)—or only becomes evident in adulthood. 102 Our imaging measures included tensor-based morphometry (TBM) which is a 103 sensitive measure for mapping developmental impacts of genetic and immune 104 perturbations (Nasseef et al. 2018; Ellegood et al. 2018; Guma et al. 2022; Kielar et al. 105 2012), white matter fractional anisotropy (FA) from dMRI to evaluate white matter 106 organisation which is influenced by microglial activity in development (Chan et al. 107 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 26, 2025. ; https://doi.org/10.1101/2025.04.24.650541doi: bioRxiv preprint 4 2024; Falangola et al. 2023), and other connectivity metrics such as global efficiency 108 and clustering coefficient to analyse brain network topology (Zhu et al. 2017; Forlim 109 et al. 2024; Hadley et al. 2016). These techniques were complemented by behavioural 110 testing in adult mice, measuring cognition and emotional reactivity through 111 established paradigms like the open field (OF) test, elevated plus maze (EPM), novel 112 object recognition (NOR), and prepulse inhibition (PPI), the three former 113 paradigms being previously tested in C3ar1-deficient mice. Unexpectedly, we found 114 no robust brain or behavioural phenotype in our datasets, challenging the previous 115 interpretations for a neurodevelopmental role for C3aR1 under physiological 116 conditions. 117

118 Validation of the C3ar1tm1Cge mutation 119 Although we used an established C3ar1 knockout line, C3ar1tm1Cge (Humbles et al. 120 2000), we performed our own validation of the mutation, for which we designed a 121 79-base pair (bp) amplicon targeting the putatively deleted region (Figure 1). PCR 122 analysis of cDNA derived from bone marrow-derived macrophages–a cell type 123 consistently reported to express C3ar1 mRNA (Tao et al. 2021; Mommert et al. 124 2018; Mamane et al. 2009)–showed no detectable transcript in this region in the 125 mutant animals (Figure 1-i), confirming the absence of the canonical transcript. We 126 also investigated the possibility of an alternative transcript after an alternative start 127 codon but found no transcript there either (Figure 1-ii, Supplemental figure 1). 128 These results were further corroborated by quantitative PCR (qPCR, n = 8) in 129 homozygous knockout animals in either M0-like or interleukin 4 (IL4)-induced M2-130 like macrophages (Supplemental figures 2a-e). Together, these results confirm that 131 C3ar1tm1Cge is a true loss-of-function or a “knockout” allele resulting in no C3ar1 132 transcript. 133 C3ar1-deficiency does not influence total or regional brain volume 134 We conducted a longitudinal MRI study (referred to as cohort 1 hereafter or implied 135 when cohort is not specified) to investigate potential genotype-related differences in 136 brain structure and function using C3ar1tmCge homozygous knockout mice 137 (C3ar1-/-, C3ar1-deficient) and their littermate wild-type control animals 138 (C3ar1+/+) (Figure 2a) on C57BL6J (Charles River, UK) background. Both groups 139 underwent in vivo MRI in adolescence (postnatal day or PND27-31) and adulthood 140 (PND81-92). Structural MR images were additionally collected ex vivo from the 141 same mice sacrificed in adulthood after the in vivo scan to achieve higher isotropic 142 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 26, 2025. ; https://doi.org/10.1101/2025.04.24.650541doi: bioRxiv preprint 5 resolution (0.1 mm ex vivo vs 0.15 mm in vivo) and increased signal-to-noise ratio 143 (SNR). We also conducted MRI in adulthood in an independent study cohort, 144 which also represents a replication dataset, rereferred to as cohort 2. 145 We used tensor-based morphometry (TBM) analysis to estimate total brain volume 146 and to map regional brain volume differences. There were no genotype dependent 147 differences in total brain volume in vivo in adolescence (Figure 2b) nor in adulthood 148 (Figure 2c) although in both groups female mice exhibited significantly smaller total 149 brain volumes compared to males in adolescence. These sex differences were no 150 longer observed in adulthood which aligns with previous findings (Guma et al. 151 2024). 152 Using TBM, no significant genotype-dependent differences in regional brain volume 153 were detected in adolescence (Figure 2d-i). In adulthood (Figure 2d-ii), C3ar1-/- 154 mice showed a significant (p < 0.05) volume increase in the right pretectal area, and 155 subthreshold (0.05 < p < 0.5) increases in the left pretectal area and the right lateral 156 thalamus in vivo but these differences were no longer observed in the same study 157 cohort ex vivo despite improved spatial resolution (not shown), nor did we observe 158 any genotype-dependent differences using TBM analysis in cohort 2 (not shown). 159 We did not observe any genotype-by-sex interaction effects in adolescence or in 160 adulthood in vivo, or ex vivo in adulthood the same study cohort, nor cohort 2 (not 161 shown). 162 While no reproducibly significant genotype effects in regional volume were observed 163 in cohort 1, we detected sexually dimorphic effects (Figure 2d-iii and 2d-iv). In 164 adolescence, female mice had significantly larger volumes bilaterally in agranular 165 insular cortex (AI), superior colliculus (SC), medial septum (MS) and in CA1 region 166 of the hippocampus, whereas male mice had increased volumes in white matter areas 167 including the olfactory tract, the corpus callosum, the hippocampal commissure, and 168 notably also in the median preoptic nucleus (MEPO) which is well known to be 169 larger in male rodents (Gorski et al. 1978) (Figure 2d-iii, Supplemental figure 3 170 for absolute volume). Many of these sex differences were no longer observed in 171 adulthood (see also Supplemental figure 4 for cohort 2 data), but females showed 172 larger relative volumes in the dorsal anterior cingulate cortex (ACAd), secondary 173 motor cortex (MOs) and primary somatosensory cortex (SSp). Adult males had larger 174 volumes in the medial preoptic area (MPO), medial amygdala (MEA) and the bed 175 nucleus of stria terminalis (BST)–differences which are well documented sexual 176 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 26, 2025. ; https://doi.org/10.1101/2025.04.24.650541doi: bioRxiv preprint 6 dimorphisms (Hines, Allen, and Gorski 1992) and which were also observed ex vivo 177 in this study cohort (Supplemental figure 5). 178 There were also no genotype differences in regional brain volumes over time from 179 adolescence to adulthood, nor did we observe any sex-by-genotype interaction (not 180 shown). Overall, female somatosensory and motor cortices increased more in volume 181 between adolescence and adulthood than male (Supplemental figure 6), in line 182 with the observed smaller differences in these areas in adulthood compared to 183 adolescence. 184 To evaluate whether our study was sufficiently powered to detect genotype effects 185 on regional brain volume, we analysed intra-group variability using a region-of-186 interest (ROI)-based approach. The coefficient of variation for grey matter ROI 187 volume showed no differences between genotypes in vivo in adolescence (Figure 2e) 188 and adulthood (Figure 2f), nor ex vivo in adulthood (Figure 2g). Variability was 189 low, ranging from 4.2–5.5% in vivo and 4.2% ex vivo, while hippocampal variability 190 remained below the neuroimaging gold standard of 5%, at 3.7–3.8% in vivo and 4.2% 191 ex vivo (Lerch et al. 2012). These values suggest that our study was well-positioned to 192 detect a genotype effect if one had been present. 193 C3ar1-deficiency does not influence fractional anisotropy 194 To assess the potential impact of C3ar1 deficiency on white matter integrity, we 195 measured fractional anisotropy (FA) as an indirect marker of axonal microstructure 196 (Figure 3). Using voxel-wise analysis, we observed sub-threshold (0.05 < p < 0.5) 197 decreases in FA in C3ar1-deficient mice compared to wild-types (Figure 3a). In 198 adolescence, these sub-threshold reductions were noted in the corpus callosum (CC) 199 and optic tract (OPT, Figure 3a, upper panel), while in adulthood, they were 200 primarily localised to the corpus callosum (Figure 3a, bottom panel). These sub-201 threshold differences were not observed in ex vivo scans. Female mice had sub-202 threshold decreases in FA in vivo at PND90 in the internal capsule and in the third 203 ventricle (Supplemental figure 7) but this was again no longer observed ex vivo. We 204 did not observe any sex-by-genotype interactions in our FA datasets (not shown). 205 In our ROI-based FA analysis, we focused on predefined white matter regions, 206 hypothesising that FA alterations would primarily occur in areas containing axonal 207 tracts due to increased microglial developmental phagocytosis in C3ar1-deficient 208 mice (Gnanaguru et al. 2023; Falangola et al. 2023; Chan et al. 2024). In this analysis 209 also, no significant genotype effects were observed in adolescence (Figure 3b-i) or 210 adulthood (Figure 3b-ii). There were no genotype-dependent differences in the 211 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 26, 2025. ; https://doi.org/10.1101/2025.04.24.650541doi: bioRxiv preprint 7 extent of change but FA increased in all white matter areas except the olfactory tract 212 between adulthood and adolescence (Figure 3b-iii), which is in line with the early 213 maturation of the olfaction system in mice (Gretenkord et al. 2019). 214 C3ar1-deficiency does not influence global functional brain connectivity 215 To evaluate global brain FC, which has been reported to be altered in other microglia 216 gene knockouts (Filipello et al. 2018; Deivasigamani et al. 2023; Zhan et al. 2014), we 217 estimated FC through analysis of BOLD signal time-courses with the assumption 218 that the magnitude of correlation between these time-courses corresponds to the 219 strength of FC. We calculated pairwise correlation coefficients between 36 (18 per 220 hemisphere) grey matter regions. Non-zero correlation values were averaged across 221 proportional progressively decreasing sparsity thresholds to preserve biologically 222 meaningful weak correlations while minimising noise (Bassett et al. 2009). No 223 genotype-dependent differences in global FC were observed in adolescence (Figure 224 4a), adulthood (Figure 4d) or in the change over time (Figure 4g). 225 We applied graph theory to characterise global brain network connectivity across all 226 regions, focusing on two key metrics: clustering coefficient (Figure 4b, e and h) and 227 global efficiency (Figure 4c, f and i), and applying the same strategy for proportional 228 thresholding as for FC. The clustering coefficient reflects the tendency of nodes to 229 form connected local clusters, with higher values indicating the presence of more 230 highly interconnected subnetworks within the brain (Bullmore and Sporns 2009). 231 Global efficiency refers to the average of shortest paths linking nodes in a network 232 and can be a proxy of information integration abilities since it decreases with 233 cognitive deficit (Berlot et al. 2016; Hawkins et al. 2020) and increases with 234 development (Jiang et al. 2023). Clustering coefficient and global efficiency appeared 235 higher in C3ar1-deficient animals at both time-points, but this was not significant 236 (Supplemental table 1). Further, we found no effects of genotype across global 237 connectivity measures when we treated males and females as separate groups (Table 238 1, for p values, see Supplemental table 2). 239 FC, global efficiency and clustering coefficient appeared to increase with brain 240 maturation when groups were combined, but this was only significant in the case of 241 global efficiency when graph sparsity was lower, that is, when 30-50% of the 242 strongest connections were retained, with small effect sizes observed (Cohen’s d = 243 0.30-0.32). These findings indicate that while C3ar1 deficiency did not result in 244 alterations in global network properties under our experimental conditions, 245 developmental changes in global efficiency were detectable. 246 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 26, 2025. ; https://doi.org/10.1101/2025.04.24.650541doi: bioRxiv preprint 8 C3ar1-deficiency has no detectable effect on functional brain networks 247 Since C3ar1-deficient mice did not show statistically significant changes in global 248 connectivity metrics, we next examined specific networks after conducting t-tests for 249 each FC pair between genotypes. For this we used two hypothesis-free approaches; 250 false discovery rate (FDR) correction to identify strongly differing edges between 251 genotypes, and network-based statistics (NBS) (Zalesky, Fornito, and Bullmore 252 2010) to detect network-level differences while controlling for family-wise error rate. 253 Thresholding the resulting t-value matrices at |t| ≥ 2 for adulthood and adolescence, 254 and their changes over time (adulthood-adolescence; Figure 5a), no edges remained 255 significant after FDR correction, indicating an absence of strongly differing 256 functional connections between genotypes. Then, using NBS, which tests the 257 likelihood of detecting a connected component of a specific size, we determined that 258 the network sizes observed after thresholding at |t| ≥ 2 (n = 68 at PND30, n = 15 at 259 PND90, and n = 1 for change) could occur by chance in these datasets. 260 Given that neither the hypothesis-free approaches, the FDR correction and NBS, 261 detected genotype-related differences, we focused on anxiety-related regions. This 262 decision was motivated by prior evidence of anxiety-like behaviour in C3ar1-263 deficient animals (Westacott et al. 2022) and the inclusion of anxiety-specific tests in 264 our behavioural battery. We calculated the mean absolute connectivity strength of 20 265 a priori selected anxiety-related regions. We did not detect an effect of genotype 266 (Figure 5b), sex (Table 2) or a sex-by-genotype interaction (Supplemental table 267 3). 268 Next, we examined FC and global efficiency within resting state networks that have 269 been linked to anxiety and emotionality in humans (Coutinho et al. 2016; Geng et al. 270 2015; Schimmelpfennig et al. 2023) and that have been observed in mice (J. 271 Grandjean et al. 2020; Sforazzini et al. 2014; Hikishima et al. 2023), the default mode 272 network (DMN) and the salience network (SAL; Figure 5c). Additionally, we 273 analysed network connectivity within the predefined anxiety-related regions whose 274 overall connectivity was summarised in Figure 5b. Consistent with our earlier 275 findings, we did not observe genotype-dependent differences in these networks at 276 either time-point. 277 The only significant genotype-related differences were seen in voxel-wise seed-to-278 brain connectivity analysis. Here, we show two anxiety-related regions, the left 279 prefrontal cortex and left ventral hippocampus (Figure 5d; Supplemental figure 8 280 for the right hemisphere) where we observed weak but significant, widespread higher 281 connectivity in C3ar1-deficient animals compared to controls at both time-points. 282 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 26, 2025. ; https://doi.org/10.1101/2025.04.24.650541doi: bioRxiv preprint 9 Seed-based analysis further revealed higher FC across the brain when other anxiety-283 related regions were used as seeds, mostly in adolescence (Supplemental figure 9). 284 However, this effect was not confined to anxiety-related regions or specific networks 285 (Supplemental figure 10). Although the increase of FC in C3ar1-deficient mice 286 appears widespread, this is not robust, since our seed-based analysis controls for 287 voxel-wise comparisons within subjects but does not correct for testing multiple 288 seeds. 289 C3ar1-deficient mice do not have any discernible behavioural phenotypes 290 Another way of assessing functional consequences of genetic manipulations with 291 expected neurodevelopmental sequelae is behavioural testing (Crawley 2007). In this 292 study, we aimed to evaluate the impact of C3ar1 deficiency on anxiety-like 293 behaviour, locomotion and recognition memory, which were chosen based on prior 294 reports of being altered in C3ar1-deficient mice. We used a battery of well-295 established behavioural tests, including the OF test and EPM for anxiety-like 296 behaviour and locomotion, as well as NOR for recognition memory. We also tested 297 PPI which is a sensorimotor reflex consistently found to be attenuated in 298 schizophrenia (Ludewig, Geyer, and Vollenweider 2003; Mena et al. 2016) but has 299 not been tested in C3ar1-deficient mice. 300 We performed behavioural testing in both cohorts described in the structural MRI 301

sections. In cohort 1 (Figure 2a), behavioural testing (EPM, then OF) 302 occurred shortly before adulthood MRI scan. Behaviour of cohort 2 was also tested 303 in adulthood (OF, NOR, EPM, PPI, in that order) followed by MRI, but they did 304 not undergo an MRI scan in adolescence. Behavioural data were analysed separately 305 for each cohort to account for differences in study design (see also Methods). 306 To assess anxiety-like behaviour and locomotion (Figure 6a–b, Supplemental 307 table 4 for description of anxiety-like metric selection), we calculated z-scores for 308 C3ar1-deficient mice relative to wild-type controls (Table 3 for untransformed 309 means). In cohort 1, no significant genotype effects were detected on anxiety-like 310 behaviour or locomotion at either time-point (Supplemental table 5). To address 311 potential confounding from co-housing littermate mutants and wild-types (Kalbassi 312 et al. 2017), we also examined whether the number of C3ar1-deficient cage-mates 313 influenced wild-type behaviour (Supplemental figure 11) but found no consistent 314 pattern or evidence of systematic anxiety-like effects in wild-types. 315 In cohort 2, most anxiety-related and locomotion measures showed no genotype 316 differences (Supplemental table 6), except for reduced distance travelled in the 317 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 26, 2025. ; https://doi.org/10.1101/2025.04.24.650541doi: bioRxiv preprint 10 centre 70% of the OF arena by C3ar1-deficient mice (uncorrected two-sample t-test p 318 < 0.01, Cohen's d = 0.92, Supplemental table 6). However, no genotype effect on 319 distance was detected in the core 30% of the OF arena or in the centre and core of the 320 OF arena in cohort 1. 321 The sample size (n = 38) of cohort 2 was insufficient to test sex-by-genotype 322 interactions, limiting statistical power to detect only large effects (Cohen’s f = 0.5 at 323 alpha = 0.05 and 80% power), which were clearly not observed across behavioural 324 measures. No sex-by-genotype interactions were observed in cohort 1 325 (Supplemental table 5). 326 Additionally, PPI testing in cohort 2 exclusively showed no effect of C3ar1 deletion, 327 though PPI increased with prepulse intensity as expected (Figure 6c). NOR testing 328 (also exclusively in cohort 2) revealed no genotype differences, with all groups 329 demonstrating successful learning based on recognition indices significantly above 330 chance levels (Figure 6d). 331 Overall, these findings suggest that C3ar1 deficiency does not result in robust 332 anxiety-like or hyperactive phenotypes, nor deficits in recognition memory or PPI. 333

334 Here we used a longitudinal neuroimaging and behavioural testing to investigate the 335 impacts of C3ar1 deletion on brain structure and function across early life 336 development in mice. We found no robust evidence that C3ar1 deficiency affects 337 total or regional brain volume, white matter FA, global FC, global efficiency, 338 clustering, or specific functional networks although we were able to detect previously 339 reported sexual dimorphisms and brain maturation effects unrelated to C3ar1 status 340 in these datasets. 341 We also found no evidence of changed behavioural outcomes in male or female 342 C3ar1-deficient mice in adulthood. These results raise questions regarding the source 343 of discrepancies between these data, our previous behavioural work (Westacott et al. 344 2022; Westacott et al. 2021), and that of others (Crider et al. 2018; Coulthard et al. 345 2018; Pozo-Rodrigálvarez et al. 2021; Sun et al. 2024), which may be attributed to 346 our use of littermate controls, as will be elaborated upon in the following sections. 347 Overall, the lack of a behavioural phenotype combined with the paucity of genotype 348 effects in brain structure across time-points suggest the role of C3aR1 in 349 neurodevelopment and behaviour might not be as significant as originally thought, at 350 least in the absence of exacerbating stress or immune triggers. 351 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 26, 2025. ; https://doi.org/10.1101/2025.04.24.650541doi: bioRxiv preprint 11 Absence of C3ar1-dependent effects on brain structure 352 Our morphometric analysis showed that total and regional brain volume was not 353 changed between C3ar1-deficient and wild-type mice. We were, however, able to 354 detect sex-specific neurodevelopmental changes that have previously been reported 355 in rodents been such as brain growth occurring later in females and minimal total 356 and regional brain volume differences in adulthood (Qiu et al. 2018; Guma et al. 357 2024; Gorski et al. 1978; Hines, Allen, and Gorski 1992; de Courten-Myers 1999), 358 indicating that our method was sensitive enough to detect these. Given the low 359 variability in our sample and our ability to detect these known sex differences, we 360 believe our study was well-positioned to identify potential genotype effects, had they 361 been present. 362 Absence of C3ar1-dependent effects on white matter fractional anisotropy 363 C3ar1 is predominantly expressed by microglia in the brain (Quell et al. 2017; Tasic 364 et al. 2018), where it influences their reactivity and phagocytosis (Vasek et al. 2016; 365 Gedam et al. 2023; Zheng et al. 2021; Lian et al. 2015; Gnanaguru et al. 2023), and 366 developmental microglial characteristics are associated with FA changes in mice and 367 humans (Falangola et al. 2023; Chan et al. 2024). In this study, we found no C3ar1-368 dependent changes in white matter while we were able to detect brain maturation 369 effects on FA regardless of genotype, which increases during white matter 370 development between adolescence and adulthood in both humans and mice 371 (Brouwer et al. 2012; Hagmann et al. 2010; Reynolds et al. 2019; Piekarski et al. 372 2023; Chahboune et al. 2007). Lack of a phenotype after the deletion of a 373 predominantly microglial gene may not be surprising in the light of recent report 374 that suggest mice that have never had any microglia (achieved by deleting the super 375 enhancer for macrophage colony-stimulating factor receptor, Csf1r) do not have 376 overt neurodevelopmental phenotypes (Rojo et al. 2019), and only show severe 377 vulnerability on a neuroinflammatory 5X-FAD background which is used to model 378 Alzheimer’s disease (Kiani Shabestari et al. 2022). It could therefore also be that 379 C3ar1-deficient neurodevelopmental phenotypes can only be detected in the context 380 of neuroinflammation. 381 Widespread, weak seed-to-brain functional connectivity increase in C3ar1-382 deficient mice in adolescence 383 We did not detect genotype-dependent global FC changes nor changes to clustering 384 coefficient and global efficiency at either time-point, but we did observe a genotype-385 independent developmental increase in global efficiency with time which has been 386 reported before (Jiang et al. 2023; Hagmann et al. 2010; Koenis et al. 2018), 387 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 26, 2025. ; https://doi.org/10.1101/2025.04.24.650541doi: bioRxiv preprint 12 particularly at lower sparsities (Cai, Dong, and Niu 2018), which similarly to out 388 structural results indicates that our method was sensitive enough to pick up 389 developmentally relevant effects. 390 Similarly, we did not observe any genotype-dependent changes to specific brain 391 networks, including anxiety networks (Figure 5), which we previously hypothesised 392 to be affected based on our previous behavioural results (Westacott et al. 2022). 393 However, our anxiety network connectivity findings are internally consistent with 394 the absence of an anxiety-like phenotype in the cohorts tested in this study. 395 Combined with a recent null report for an anxiety-like effect in the EPM in C3ar1-396 deficient animals by another group (Sun et al. 2024), our data indicate that the effect 397 of C3ar1 on brain and behaviour correlates of anxiety-like behaviour must be 398 context-dependent and/or minimal. 399 Although we did not detect network-specific effects, C3ar1-deficient mice exhibited 400 widespread increases in resting-state FC in voxel-wise seed-based analyses during 401 adolescence across nearly all seeds examined, with a weaker effect persisting into 402 adulthood (Figure 5 and Supplemental figures 8-10). These subtle increases in 403 connectivity may reflect developmental alterations in circuit properties caused by the 404 absence of C3aR1. For instance, similar findings have been reported for other 405 microglial receptors—such as TREM2, CX3CR1, and CR3—where genetic 406 deficiency altered adult brain FC (Filipello et al. 2018; Zhan et al. 2014; 407 Deivasigamani et al. 2023). In these reports, TREM2 or CX3CR1 knockouts led to 408 impaired synapse elimination alongside decreased FC, accompanied by social 409 behaviour deficits and increased repetitive behaviour (Filipello et al. 2018; Zhan et al. 410 2014). In contrast, CR3 knockout mice exhibited no deficits in synapse or axon 411 refinement but showed reduced phagocytosis of perinatal cortical neurons and 412 higher cortical FC (Deivasigamani et al. 2023). It is indeed possible that similarly to 413 CR3, C3aR1 is also involved in perinatal neuron phagocytosis so that its deletion 414

in weak brain-wide increase in FC, particularly as C3ar1-deficiency leads to a 415 deficit of developmental astrocyte phagocytosis in the retina (Gnanaguru et al. 2023). 416 For now, we urge caution with this interpretation for several reasons. First, these 417 connectivity changes were not accompanied by any structural or microstructural 418 alteration across the brain. Second, the voxel-wise seed-based analysis is the least 419 robust presented here, as the observed changes were largely transient, weak, not 420 confined to a specific subnetwork, and not corrected for multiple comparisons across 421 seeds. Future studies investigating FC in C3ar1-deficient mice should incorporate 422 repeat measures from the same individual during adolescence to improve robustness 423 as well as measuring spontaneous and miniature excitatory synaptic current (sEPSC 424 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 26, 2025. ; https://doi.org/10.1101/2025.04.24.650541doi: bioRxiv preprint 13 and mEPSC respectively) in brain slices, although it is currently still unclear which 425 brain regions should be targeted currently with the latter approach. 426 Behavioural outcomes were unaffected by C3ar1 deficiency 427 Neither anxiety-like behaviours nor locomotor activity showed significant 428 differences between genotypes across multiple testing paradigms, apart from a 429 specific decrease in ambulation of C3ar1-deficient mice in the aversive central zone 430 covering 70% of the open field, which in the absence of a locomotor phenotype 431 could index anxiety-like processes. This latter change, however, was not reproducible 432 across study cohorts and was no longer observed in the same study in the core of the 433 open field, covering 30% of the arena. Tests for recognition memory and 434 sensorimotor gating, NOR and PPI tests, also yielded null results. 435 Differences between our current findings and previous reports of anxiety-like 436 behaviours (Westacott et al. 2022), hyperactivity (Pozo-Rodrigálvarez et al. 2021) 437 and cognitive deficits (Coulthard et al. 2018) may stem from variations in 438 experimental design, particularly environmental factors such as lighting and testing 439 time. Our behavioural experiments were conducted during the active phase (7 PM–440 11 PM, lights off) to minimise stress from sleep deprivation, which is known to 441 upregulate complement in the brain (Tillmon et al. 2024; Crider et al. 2018; Li et al. 442 2023; Tripathi et al. 2021). It is possible that previous studies, which tested animals 443 during their inactive phase, introduced a stress-dependent “second hit” to C3ar1-444 deficiency. However, this seems unlikely since inhibiting complement, including 445 C3ar1, has been shown to improve resilience to chronic stress (Li et al. 2023; Crider 446 et al. 2018; Tripathi et al. 2021; Madeshiya et al. 2022; Tillmon et al. 2024). 447 Unlike previous studies of C3ar1-deficient mice, our study employed littermate 448 control animals, meaning that wild-type mice were co-housed with their C3ar1-449 deficient siblings. This approach minimises potential environmental differences 450 between groups, which is especially important in neurobehavioral studies where 451 subtle environmental factors can significantly affect outcomes. While the use of 452 littermate controls is widely regarded as best practice in such contexts (Holmdahl and 453 Malissen 2012; Valiquette et al. 2023; Bailey, Rustay, and Crawley 2006), it is not 454 without its own potential confounds. For instance, abnormal behaviours exhibited 455 by genetically altered animals can influence the behaviour of co-housed wild-type 456 mice, particularly when the phenotype is pronounced, such as in cases of increased 457 aggression (Kalbassi et al. 2017). However, as aggression has not been reported in 458 C3ar1-deficient mice, and their previously described behavioural phenotype was not 459 notably severe, we considered such an influence unlikely in our study. Nevertheless, 460 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 26, 2025. ; https://doi.org/10.1101/2025.04.24.650541doi: bioRxiv preprint 14 we systematically examined whether the behaviour of wild-type mice correlated with 461 the number of C3ar1-deficient cage-mates and found no significant association 462 (Supplemental figure 11). 463 Prenatal environment and maternal care significantly influence phenotypic 464 development (McCarty 2017) and the complement system plays an essential role in 465 normal pregnancy and parturition (Girardi et al. 2020). A notable aspect of using a 466 littermate design is that litters of mixed genotype are born from exclusively 467 heterozygous parents, whereas in most prior studies of C3ar1-deficiency both the 468 mother and offspring were homozygous knockouts. This raises the possibility that 469 behavioural effects in C3ar1-deficient mice may result from altered intrauterine 470 environments or maternal care by C3ar1-deficient mothers. In line with this, our 471 inbred homozygous C3ar1 knockout colony showed increased pre-weaning deaths 472 and maternal cannibalism (unpublished data). If this is indeed the reason for 473 previously observed adult behavioural phenotypes, two scenarios should be 474 considered. In the first scenario, the adult phenotypes are specific and C3ar1-475 deficient offspring are uniquely sensitive to C3aR1-dependent in utero conditions or 476 maternal care deficits. In the second, more likely scenario, however, the adult 477 phenotypes are not specific to C3ar1-deficiency (i.e., wild-type mice would have 478 been similarly affected). If the latter is true, we were able to avoid measuring these 479 non-specific effects by using a littermate design. 480 Null results in our global knockout model could also be due to genetic 481 compensation, where genes with similar function compensate for the loss of a 482 mutated gene (El-Brolosy and Stainier 2017). While the mechanism for this is not yet 483 resolved, mutated RNA degradation can trigger this process (El-Brolosy et al. 2019). 484 We confirmed that the C3ar1tm1Cge mutant lacks detectable C3ar1 RNA, suggesting 485 the transcript is either not produced or undergoes nonsense-mediated decay, 486 potentially activating compensation. Confirming this would require longitudinal 487 single-cell RNA sequencing in relevant brain regions, which is beyond the scope of 488 this study. 489 Genetic background of transgenic models is another factor that varies between 490 laboratories and is known to affect phenotypic expression. For example, 491 heterozygous knockout of the autism-associated gene CHD8 has varying effects in 33 492 (Tabbaa, Knoll, and Levitt 2023) substrains, mirroring heterogeneity observed in 493 human CHD8-haploinsufficiency. While that study identified significant variability 494 across substrains, these profiles consistently differed from that of wild-type littermate 495 controls. In contrast, our study found no robust genotype-dependent differences in 496 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 26, 2025. ; https://doi.org/10.1101/2025.04.24.650541doi: bioRxiv preprint 15 the brain globally and across measures or in behaviour, apart from a single internally 497 non-reproducible behavioural outcome measure, meaning that the effect of C3aR1 498 on brain development would have to be entirely dependent on modifying genes or 499 environmental factors. 500 Finally, genetic drift could have precluded reproducibility of previously reported 501 phenotypes. Mice accumulate spontaneous mutations that rapidly reach 502 homozygosity within small colonies speeding up drift, a phenomenon recognised as 503 early as the 1980s (Fitch and Atchley 1985). So far, our current work is the first to 504 study C3ar1 deficiency using a littermate design, which minimises the number of 505 genetic loci that differ between the mutant and control mice, all of which could 506 influence measured phenotypes. Fully understanding the genetic causes of results 507 previously observed by our laboratory and others would require systematic 508 outcrossing, whole genome sequencing, and linkage analysis. Given that hundreds of 509 genes likely differ between mutant and control lines last crossed during mutation 510 creation or before, the value of such work is questionable. 511 Future directions 512 Aside from the role of C3aR1 in brain structure and function, many aspects of the 513 basic biology of C3aR1 remain unclear, including its expression pattern in the 514 healthy brain, with little information available on the cell types, contexts, or time-515 points where it is expressed. It is also unclear which G-proteins C3aR1 signals 516 through in different brain cell types. Future studies could integrate toolkits like 517 TRUPATH, a suite of Gαβγ biosensors for analysing G-protein coupling 518 preferences (Olsen et al. 2020), with transcriptomics to address these questions. This 519 approach would be particularly valuable, as it could enable the use of chemogenetics 520 to activate the same G-protein as C3aR1 in specific cell types and contexts to study 521 its function. 522 Most importantly, we propose that C3ar1-deficiency in development should be 523 studied after an immune challenge. Unlike humans, laboratory mice live in an 524 immune-privileged environment, hence investigating the consequences of a 525 combined immune insult with the genetic deficit would be relevant; or better yet, in 526 the context of C4A overexpression, which is a known genetic risk factor for 527 schizophrenia (Sekar et al. 2016) and is known to impact white matter integrity 528 (Caseras et al. 2024). 529 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 26, 2025. ; https://doi.org/10.1101/2025.04.24.650541doi: bioRxiv preprint 16

530 In this study, we provided high-level global data on C3ar1-deficient mice. While our 531 study had notable strengths, including the use of a littermate design and the 532 inclusion of female animals, it is not an exhaustive characterisation of the mutant. 533 While we measured brain region volumes, we cannot definitively address more 534 reductionist questions, such as cellular composition which would have required 535 antibody staining and microscopy. Similarly, for axonal integrity analysis for which 536 we used the proxy of dMRI-derived FA, the gold standard would have been electron 537 microscopy. However, a more granular approach would have necessitated a trade-off 538 with throughput–something that is hard to justify in the absence of strong 539 hypotheses regarding the developmental expression pattern of C3ar1. Additionally, 540 our neuroimaging functional analysis relied on a specific parcellation of the brain 541 consisting of 36 regions, and different parcellations could yield varying results 542 (Thirion et al. 2014; Zalesky et al. 2010). We encourage others exploring our datasets 543 to therefore experiment with alternative parcellations. 544

545 Contrary to expectations, we found no evidence for C3ar1-dependent effects across 546 imaging measures at either time-point nor could we replicate previously seen 547 behavioural phenotypes. The absence of detectable phenotypes suggests that C3aR1 548 plays a minimal role in brain development, at least in the absence of an immune 549 trigger. These findings challenge prior assumptions about its neurodevelopmental 550 significance and necessitate further investigation into C3aR1 function on sensitized 551 backgrounds, such as those involving immune activation. 552

and materials 553 Animals 554 All animal procedures complied with the UK Animals and Scientific Procedures Act 555 1986 and were approved by the local ethical committee at King’s College London 556 (KCL). Homozygous C3ar1-/- mice were generated by homologous recombination in 557 embryonic stem cells and kindly provided by Dr. Bao Lu and Prof. Craig Gerard 558 (Harvard Medical School, Boston, MA) (Humbles et al. 2000). These mice were 559 subsequently backcrossed onto the C57BL/6J strain for at least 12 generations and 560 maintained on a C57BL/6J background in Professor Wuding Zhou’s laboratory at 561 KCL. For this study, cryopreserved stocks were rederived at KCL and crossed to 562 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 26, 2025. ; https://doi.org/10.1101/2025.04.24.650541doi: bioRxiv preprint 17 C57BL/6J mice purchased from Charles River to refresh the genetic background 563 following Jackson’s Laboratories line refreshing protocol. 564 Experimental animals (C3ar1-/- and C3ar1+/+ littermates) were generated through 565 heterozygote incrosses and resulting genotypes followed Mendelian ratios. The 566 heterozygote breeders were generated by outcrossing heterozygous mice to bought 567 wild-type Charles River C57BL/6Js. The breeders used to produce the experimental 568 animals were derived either from the first, second or third of these outcrosses. Sibling 569 crosses were not conducted and parental age was between 2-4 months to minimise 570 genetic drift. 571 Experimental mice were housed in individually ventilated cages under controlled 572 temperature (20–25°C), humidity (50–60%), and a 12-hour light-dark cycle (lights 573 on at 7:00 AM, lights off at 7:00 PM). Environmental enrichment included nesting 574 materials, tunnels, and chew sticks. Mice had ad libitum access to irradiated rodent 575 chow and autoclaved water. Animals were group-housed (2–4 mice per cage), with 576 males and females housed separately after weaning (PND21±2). Genotyping was 577 conducted on ear biopsy DNA by Transnetyx using probes targeting the neomycin 578 cassette for the mutant allele and intron 1 for the wild-type allele. No mismatches 579 were identified through double-genotyping 20% of the study cohorts. 580 Validation of mutation 581 Bone marrow-derived macrophages were obtained from tibias and femurs of eight 582 three-month-old mice (n = 4 C3ar1+/+, n = 4 C3ar1-/-) following standard protocols. 583 Bone marrow was filtered through a 40 µm mesh, centrifuged at 450 x g for 5 584 minutes at 4°C, and treated with NH4CL haemolysis buffer (NH4CL 0.15M, 585 K2HCO3 0.01M, EDTA 0.0001M). After a second centrifugation under the same 586 conditions, cells were washed with PBS and resuspended in Gibco RPMI 1640 587 Medium (Thermo Fisher, #21875-034) supplemented with 50 ng/ml recombinant 588 mouse macrophage colony-stimulating factor (M-CSF, R&D Systems, #416-ML-589 010/CF), 1% penicillin, 1% streptomycin, and 10% heat-inactivated fetal bovine 590 serum (Sigma, #F9665-50ml). 591 Cells were seeded at 1x106 cells/ml in six-well plates (six wells per animal) and 592 incubated at 37°C with 5% CO2 for 72 hours. On day three, fresh medium was 593 replaced, and 50 ng/ml recombinant mouse IL4 (R&D Systems, #404-ML-594 010/CF) was added to half of the wells to skew them towards M2 phenotype. 595 Incubation continued for an additional 48-72 hours, depending on cell confluence. 596 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 26, 2025. ; https://doi.org/10.1101/2025.04.24.650541doi: bioRxiv preprint 18 RNA was extracted using the ReliaPrep™ miRNA Cell and Tissue Miniprep 597 System (Promega, #Z6211) according to the manufacturer’s instructions. RNA 598 concentration was determined with a NanoDrop spectrophotometer (Thermo 599 Scientific, NanoDrop 2000), yielding values between 25.5 ng/µl and 227.5 ng/µl. 600 Reverse transcription PCR (RT-PCR) was performed using the LunaScript® RT 601 SuperMix Kit (NEB, #E3010L), again following the manufacturer’s instructions. 602 Depending on RNA yield, either 100 or 500 ng of RNA was used per reaction. 603 For cDNA PCR and gel electrophoresis, we used GoTaq® G2 Master Mix (NEB, 604 #M7822). We amplified a 79 basepair (bp) fragment in the deleted region alongside a 605 372 bp region in the C3ar1 cDNA that was located outside the deleted region and 606 downstream of an alternative start codon identified through Benchling. Gapdh 607 primers were included in each reaction to confirm amplification efficiency. PCR 608 products were analysed on a 2% agarose gel stained with GelRed. 609 For qPCR, we amplified the previously mentioned 79 basepair (bp) fragment in the 610 deleted region. Samples were analysed in triplicate on 96-well plates (Applied 611 Biosystems, #4346906) with Luna® Universal qPCR Master Mix (NEB, #M3003L) 612 and readings were obtained using an Applied Biosystems StepOnePlus plate reader. 613 The amplification data were processed with the ΔΔCt method, normalising against 614 the housekeeping gene, Hypoxanthine phosphoribosyltransferase 1 (Hprt). 615 Genotypes were arranged alternately across the plate to minimise bias. To verify M2-616 like polarisation, the expression of M2-specific markers Arginase 1 (Arg1) and the 617 mannose receptor, Cluster of differentiation 206 (Cd206) were assessed. 618 Table 4 | Primer sequences 619 Primer name Forward Reverse Application C3ar1 AGGATTTGTTGGTG GCTCGCA CTCCATGGCTCAGTC AAGCACA PCR and qPCR deleted region C3ar1 GCTTCCTGGTGCCG TTTTTC AGTTGGTAGAGTGCG TGAGC PCR of putative alternative transcript (3’ end of exon 2 after alternative start codon) Hprt AGTCCCAGCGTCGT GATTAGCG TTGAGCACACAGAGG GCCACAA qPCR housekeeping gene Arg1 GGCTTGCGAGACGT AGACCC GTCCAGCCCGTCGAC ATCAAA qPCR verification of M2 polarisation Cd206 CCGGAGGGTGCAGA CAAAGG TCGTCCACAGTCCAC CGAAAC qPCR verification of M2 polarisation .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 26, 2025. ; https://doi.org/10.1101/2025.04.24.650541doi: bioRxiv preprint 19 Gapdh CCTAGACAAAATGG TGAA GACTCCACGACATAC TCAGC PCR amplification efficiency control 620 Table 5 | PCR reaction 621 Amount (µl) Reagent 12.5 GoTaq® G2 Master Mix (Green) 1 C3ar1 forward primer, 10 mM 1 C3ar1 reverse primer, 10 mM 1 Gapdh forward primer, 10 mM 1 Gapdh reverse primer, 10 mM 7.5 H2O 1 Template DNA Final volume: 25 µl 622 Table 6 | Cycling conditions 623 Step Time Temperature 1. Initial denaturation 5 min 94°C 2. Denaturation 30 sec 94°C 3. Annealing 30 sec 60°C 4. Extension 1 min 72°C 5. Repeat steps 2-4 Repeat 35 x 6. Final extension 5 min 72°C Study design 624 This study used two separate cohorts of male and female C3ar1-deficient and 625 littermate wild-type mice. The main, longitudinal MRI cohort termed cohort 1, had 626 in vivo MRI performed in adolescence (range 27-31 days) and adulthood (range 81-627 92 days), and the adulthood MRI scan was preceded by OF and EPM tests. The 628 adolescence time-point was chosen because mice reach puberty approximately 629 between PND24-34 (Semaan and Kauffman 2015; Brust, Schindler, and 630 Lewejohann 2015; Pintér et al. 2007). Ex vivo imaging was conducted in cohort 2’s 631 perfusion-fixed brains after the final adulthood scan. 632 For cohort 2, behavioural testing was conducted similarly to cohort 1 in adulthood 633 only (range 74-110), and consisted of OF, NOR, EPM and PPI followed by in vivo 634 structural and diffusion MRI (note that MRI was not conducted in adolescence in 635 this cohort). For both cohorts, behavioural testing was conducted 2-7 days before the 636 adulthood scan. 637 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 26, 2025. ; https://doi.org/10.1101/2025.04.24.650541doi: bioRxiv preprint 20 Sample size of cohort 1 was statistically powered to detect medium effect sizes in 638 regional volume using TBM across four groups (males and females analysed 639 separately), with a minimum sample size of n = 15 per group based on previously 640 observed variance with this method by our group (Serrano et al. 2023). For cohort 2 , 641 statistical power was calculated to detect medium effect sizes in regional TBM with 642 sexes combined, using a sample size of n = 16 per group. 643 In vivo MRI 644 Two to three days after behavioural testing, mice were imaged using a Bruker 645 BioSpec 9.4 T scanner with an 86-mm volume resonator for transmission and a 4-646 channel surface array coil. Anaesthesia was induced with 4% isoflurane in medical air 647 (1 L/min) and oxygen (0.4 L/min), maintained at 2% but adjusted based on 648 respiration rates. For functional BOLD MRI in cohort 1, we used a medetomidine 649 and isoflurane anaesthesia optimised for mouse fMRI (Joanes Grandjean et al. 2014). 650 This consisted of a subcutaneous medetomidine bolus (0.05 mg/kg) followed ten 651 minutes later by its continuous infusion (0.1 mg/kg/h), with isoflurane levels 652 gradually reduced to 0.45-0.65% over 15 minutes from the start of the infusion after 653 BOLD-weighted fMRI was conducted after the structural scans which took a further 654 45-60 minutes after reducing isoflurane level. The respiration rate was monitored 655 with a pressure sensor, and temperature was monitored with a rectal thermometer 656 and maintained at 36-37°C using a water circulation system. 657 Ex vivo MRI 658 Following the adulthood in vivo scan in cohort 1, mice were perfused transcardially 659 with 20 mL phosphate-buffered saline (PBS) followed by 4% paraformaldehyde 660 (PFA). Heads were stored in PFA for 48 hours, then transferred to PBS containing 661 0.05% sodium azide and 2 mM gadolinium-based contrast agent (Gd-DO3A-butrol). 662 Brains were scanned in cranio in groups of four using a custom-made holder 663 immersed in perfluoropolyether (Galden®, Solvay). 664 MRI acquisition parameters 665 For in vivo imaging, we first acquired an Actual Flip Angle Imaging (AFI) sequence 666 for B1 mapping. Then we acquired three types of 3D multi-gradient-echo images 667 that were used for creating study-specific templates: magnetization-transfer weighted 668 (MTw), proton-density weighted (PDw), and T1-weighted (T1w). Subsequently, 669 T2-weighted Rapid Acquisition with Relaxation Enhancement (RARE) images were 670 obtained. Diffusion-weighted images were then acquired using a single-shot spin-671 echo echo planar imaging (EPI) sequence. Finally, for the longitudinal MRI study, 672 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 26, 2025. ; https://doi.org/10.1101/2025.04.24.650541doi: bioRxiv preprint 21 BOLD rsfMRI data were acquired using a single-shot gradient-echo EPI sequence 673 with 720 repetitions. Additionally, spin-echo EPI image pairs with opposing phase-674 encoding polarity were recorded to enable correction of susceptibility-induced 675 distortions. The in vivo scanning session lasted 1-1.5 hours. 676 For ex vivo morphometric analysis, 3D T2-weighted images were acquired using 677 RARE sequences, with a total scan duration of 1 hour and 5 minutes. Ex vivo 678 diffusion-weighted images were obtained using Stejskal-Tanner pulsed gradient spin-679 echo sequences with a 3D segmented EPI readout. Three b0 images were collected at 680 the beginning of three blocks of 30 diffusion-weighted images. The total scan time 681 for this acquisition was 14 hours and 15 minutes. 682 Table 7 | MRI acquisition parameters 683 Image type TR (ms) TE (ms) Flip angle (°) Aver- ages Band- width (kHz) FOV Matrix Other AFI 20/ 100 2.85 55 1 25 16.2 x 16.2 x 9 42 x 42 x 24 MTw 24 2.5 6 2 100 16.2 x 16.2 x 9 108 × 108 × 60 6 echoes with 2.1 ms spacing; MT pulse: gaussian, 4 ms, amplitude 10 µT, offset -3 kHz, bandwidth 685 Hz PDw 20 2.5 4 2 100 16.2 x 16.2 x 9 108 × 108 × 60 7 echoes with 2.1 ms spacing T1w 20 2.5 20 2 100 16.2 x 16.2 x 9 108 × 108 × 60 7 echoes with 2.1 ms spacing T2w in vivo 5000 42 90/ 180 4 50 16 x 12 128 × 96 32 slices, slice thickness 0.5, RARE factor 8 T2w ex vivo 300 30 90/ 180 1 50 25 x 24 x 18 250 x 240 x 180 RARE factor 4, scan time = 1 h 5 min .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 26, 2025. ; https://doi.org/10.1101/2025.04.24.650541doi: bioRxiv preprint 22 DWI in vivo 3000 21 90/ 180 2 357 19.2 x12 96 x 60 Single shot spin-echo planar imaging, 30 slices, slice thickness 0.5 mm, three diffusion shells: b = 350 s/mm² with 9 directions, b = 1000 s/mm² with 34 directions, b = 2000 s/mm² with 78 directions (δ = 3 ms, Δ = 11 ms), 3 b0 images per shell DWI ex vivo 300 28.5 90/ 180 300 25 × 24 × 18 200 × 192 × 144 Stejskal-Tanner pulsed gradient spin echo sequences with a 3D segmented EPI readout. 12 segments and a total of 90 diffusion-weighted images acquired at a b-value of 4000 s/mm² (δ = 4 ms, Δ = 13 ms) BOLD fMRI 1000 15 55 1 200 20 x 20 64 x 64 Single-shot gradient- echo EPI sequence, 16 slices, slice thickness 0.5, 720 repetitions 684 Structural MR image processing 685 For preprocessing MTw, T1w, and PDw images, de-ringing was conducted using the 686 MRtrix3’s (Tournier et al. 2019) mrdegibbs command. MTw, T1w, and PDw 687 images were averaged across echo times, rigidly co-registered using Advanced 688 Normalization Tools (ANTS; Avants et al. 2011) antsRegistration, and used for 689 template construction (see below). 690 For DWI, MRtrix3 dwidenoise was used for de-noising, MRtrix3 mrdegibbs for de-691 ringing, and FSL’s (the FMRIB Software Library, Jenkinson et al. 2012; Smith et al. 692 2004) topup and eddy for susceptibility and eddy-current distortion and motion 693 correction. FSL’s dtifit was used for diffusion tensor imaging (DTI) model fitting, 694 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 26, 2025. ; https://doi.org/10.1101/2025.04.24.650541doi: bioRxiv preprint 23 enabling the calculation of fractional anisotropy (FA), mean diffusivity (MD) and 695 axial diffusivity (AD), the latter two which are not reported in this manuscript for 696 brevity. For a more detailed diffusion processing protocol see Kim et al. (2023). 697 Study templates 698 The antsMultivariateTemplateConstruction2.sh script from ANTs was used to 699 create study specific templates from processed images. For in vivo scans of cohort 1, 700 MTw, T1w, R2* map (generated from the multi-gradient-echo PDw, T1w and 701 MTw images using the qi mpm_r2s command in the QUIT package), S0, FA and 702 MD images were used. For the ex vivo scans of cohort 1, separate T2w and DTI 703 (comprising S0, FA, and MD images) templates were created. For cohort 2, PDw, 704 T1w, MTw, S0 (estimated non-diffusion-weighted image from dtifit), FA, and 705 MD images were used. 706 Jacobian determinant maps 707 To estimate volume, Jacobian determinant maps were generated from the 708 deformation fields corresponding to the transformation of each subject to the study 709 template using the CreateJacobianDeterminantImage command from ANTs. The 710 Jacobian determinant values of all voxels within the template brain mask were 711 summed to obtain the total brain volume of each subject. Jacobian determinants 712 were calculated from the combined rigid, affine, and Symmetric Normalization 713 (SyN) transforms as well as from only the SyN transforms to obtain maps of absolute 714 and relative volume (accounting for differences in global brain volume), respectively. 715 For TBM, the Jacobian determinants were subsequently log-transformed. 716 Voxel-wise analysis 717 For voxel-wise statistics, FSL randomise with permutation testing was used (10,000 718 for cohort 1, 5000 iterations for cohort 2) followed by a threshold-free cluster 719 enhancement (TFCE) and family-wise error (FWE) correction as described in Wood 720 et al. (2016) and Kim et al. (2023). Given the absence of genotype differences in total 721 brain volume, TBM regional volumes are reported relative to total brain size for 722 greater accuracy (Lerch et al. 2012), while absolute volume maps are reported in the 723 Supplementary files where sex differences in total brain volume were present. 724 Common coordinate space 725 The study template was registered to the Allen Mouse Brain Common Coordinate 726 Framework (CCFv3; Wang et al. 2020) using ANTs, and the Allen atlas was 727 subsequently transformed to the study template space with the inverse transform. 728 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 26, 2025. ; https://doi.org/10.1101/2025.04.24.650541doi: bioRxiv preprint 24 ROI-based analysis of volume 729 The images were segmented using an in house modified version of the Allen atlas of 730 72 regions (Serrano et al. 2023; Wang et al. 2020). These segmentations were 731 subsequently used to compute regional volumes by summing the Jacobian 732 determinants within each parcellation. 733 Regional volume variability was estimated by calculating a coefficient of variation 734 (CV) for each region with normalised root-mean-square method for each genotype 735 in each experiment, using the calculation: 𝐶𝑉 = ! " , where 𝜎 is the standard 736 deviation and 𝜇 the population mean for each region in the atlas (n = 72 regions). 737 Group differences in CV were calculated with a Kruskal-Wallis test (SciPy.stats, 738 kruskal). 739 Fractional anisotropy 740 For mass-univariate voxel-wise analysis of fractional anisotropy, values from dtifit 741 for cohort 1 were again analysed with FSL randomise with permutation testing 742 (10,000 permutations) followed by TFCE and FWE-correction. FA is not reported 743 for cohort 2 in this manuscript for brevity. For ROI-based analysis, voxel FA 744 medians within parcellation were used for all white matter regions. 745 To calculate the change over time in fractional anisotropy in the longitudinal study, 746 adolescence values for each voxel value or regional median were subtracted from 747 adulthood values. Differences between genotypes were calculated with a mixed 748 ANOVA with between-subjects factor of genotype and within-subjects factor of 749 region. Change from 0 was calculated with a two-sided one-sample t-test (SciPy.stats, 750 ttest_1samp), which was corrected with the Benjamini–Hochberg method. 751 BOLD fMRI pre-processing 752 Images were largely pre-processed using the Analysis of Functional NeuroImages 753 (AFNI) toolkit. Slice timing correction was performed using the 3dTshift package, 754 despiking with 3dDespike, and motion correction was applied with 3dvolreg. The 755 motion-corrected time average was then registered to each subject's own T2w image 756 using ANTs, followed by registration to a template T2w image derived from a 757 separate mouse study conducted at the BRAIN Centre (KCL). The use of this 758 external template was justified by the low resolution of fMRI, which does not 759 benefit from creating a study-specific template. 760 The images were also distortion corrected using FSL’s topup, with the distortion 761 estimated using auxiliary phase encoded spin-echo images with otherwise matching 762 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 26, 2025. ; https://doi.org/10.1101/2025.04.24.650541doi: bioRxiv preprint 25 acquisition parameters to the gradient echo EPI used for the BOLD signal. Prior to 763 analysis, corrected images were band-pass filtered at 0.01–0.2 Hz with AFNI’s 764 3dTproject to remove low-frequency scanner drift noise and high-frequency 765 physiological noise. Nuisance variables (movement and CSF signal) were also 766 simultaneously regressed out of the signal at this stage. Finally, spatial smoothing was 767 applied using AFNI’s 3dBlurInMask with a FWHM kernel. 768 Functional connectivity and graph theory analysis 769 For functional analysis, a high-level parcellation scheme was applied to segment 36 770 regions (18+18) excluding white matter from the acquired 3D volume. The BOLD 771 signal time-courses were averaged within each ROI, and the mean time-courses were 772 extracted using FSL’s fslmeants tool. Pearson correlation coefficients were 773 calculated for each time-course pair, producing a 36×36 correlation matrix for each 774 subject. These matrices were analysed as functional connectivity (FC) graphs, with 775 edge strength determined by the Fisher z transformed Pearson correlation coefficient. 776 To identify the strongest connections, graphs were thresholded at 5% intervals from 777 5% to 50%. At each threshold level, referred to as the graph sparsity interval, 778 connections below this threshold were set to zero, generating 10 sparsity graphs per 779 subject. FC was calculated as the average non-zero connectivity at each threshold 780 level. Global graph metrics, global efficiency and clustering coefficient, were 781 computed at each sparsity level using Brain Connectivity Toolbox algorithms 782 (Rubinov and Sporns 2010) implemented with Network X (3.4.2) 783 global_efficiency and clustering passing binary thresholded Pearson matrices to 784 prioritise topology in the presence of noise. For global efficiency and clustering 785 coefficient, random curves were calculated by shuffling the thresholded binary 786 matrix positions. Area under the curve (AUC) was computed per subject using 787 trapezoidal numerical integration (numpy.trapz), and the likelihood of observed 788 values was estimated with permutation testing (10,000 permutations). 789 To estimate changes over time, global graph metric values at adolescence were 790 subtracted from adulthood values at each sparsity interval. Group differences were 791 again tested by calculating the AUC and applying permutation testing. Changes 792 from baseline (zero) were evaluated using two-sided one-sample t-tests (difference 793 from 0), corrected for multiple comparisons with the Benjamini-Hochberg method. 794 FDR correction and network-based statistics 795 Matrices for C3ar1-deficient and wild-type mice were compared with Student’s t-796 tests for each pairwise connection, resulting in 36×36 t-statistic and p-value matrices. 797 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 26, 2025. ; https://doi.org/10.1101/2025.04.24.650541doi: bioRxiv preprint 26 To control for multiple comparisons, FDR correction was applied to the upper 798 triangle of the p-value matrix using the Benjamini-Hochberg procedure 799 (statsmodels.stats.multitest.fdrcorrection, α = 0.05). 800 For network based statistics (NBS), above described t matrices were thresholded at |t| 801 ≥ 2. To identify connected components within the FC graph, adjacency matrices 802 representing significant connections (|t| ≥ 2) were converted into graph objects using 803 NetworkX. Regions of interest (ROIs) were treated as nodes, and significant 804 connections as edges. Connected components were identified using a breadth-first 805 search (BFS) algorithm, which explores all neighbouring nodes before moving deeper 806 into the graph. Only components containing more than one ROI were retained for 807 further analysis. To generate a null distribution of maximal component sizes, group 808 labels were randomly shuffled across subjects for each permutation while preserving 809 matrix structure (10,000 permutation). The p-value for an observed component was 810 calculated as the proportion of permutations where the maximal component size 811 exceeded that of the observed component. 812 A priori node strength analysis 813 We selected 20 (10+10 left and right) anxiety and fear related regions and calculated 814 their average absolute connectivity to all other regions using Fisher z-transformed 815 Pearson correlation coefficients. We then used a mixed ANOVA with between-816 subjects factor of genotype and within-subjects factor of region followed by pairwise 817 testing with BH FDR correction of p values. 818 Within-network analysis 819 Two mouse resting state networks were subset from correlation matrices: the default 820 mode network (DMN) and the salience network (J. Grandjean et al. 2020; Sforazzini 821 et al. 2014). The DMN included bilateral prefrontal cortices, cingulate cortices, and 822 dorsal hippocampi, while the salience network comprised bilateral cingulate cortex, 823 amygdala, and striatum. Additionally, a third anxiety-related network was defined, 824 consisting of regions identified in the seed-based analysis described above. 825 For each network, mean FC was calculated as the average of all pairwise connections 826 between nodes within the network, without applying a threshold. For individual 827 network global efficiency analysis, thresholding was not used due to small amount of 828 nodes. Instead, weighted Pearson matrices were passed to bctpy (0.6.1) 829 efficiency_wei. 830 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 26, 2025. ; https://doi.org/10.1101/2025.04.24.650541doi: bioRxiv preprint 27 Student’s t tests were used to test significance, and the comparisons were corrected 831 within outcome measure with BH method. 832 Voxel-wise seed-to-brain analysis 833 We conducted voxel-wise seed-based FC analyses for anxiety-related regions, as well 834 as the colliculus and sensory cortex, which served as control regions not specific to 835 anxiety. For each seed region, the time-course of the BOLD signal was extracted and 836 regressed with the BOLD signal of every voxel in the brain, resulting in a 3D spatial 837 map of the connectivity with the seed. Group-level comparisons of these maps were 838 performed between genotypes using voxel-wise permutation tests with FSL’s 839 randomise (5000 permutation), converted with TFCE and statistical significance 840 corrected for multiple comparisons using FWE (seed-to-brain), but they were not 841 corrected for the presence of multiple seeds. 842 General behavioural procedures 843 For cohort 1, EPM was administered as the first test, followed by OF. For cohort 2, 844 OF was the first test, followed by NOR test after two days of low light habituation (4 845 lux), EPM, and PPI. 846 Handling of the mice began 2-3 days prior to the behavioural testing battery. By the 847 start of the experiments, the mice sat comfortably on the experimenter’s hand. Mice 848 were handled using cardboard tunnels to minimise stress, and tail handling was 849 avoided. 850 Behavioural tests were conducted during the dark phase (between 7:00 PM and 851 11:00 PM clock-time) of the light-dark cycle to align with the active period of mice. 852 Mice were randomised by genotype and counter-balanced by sex, with the 853 experimenter systematically blinded to genotype throughout testing and analysis 854 though the allocation of a study ID and test order ID respectively. 855 For cleaning of the test apparatus, we used 70% EtOH for all arenas except for the 856 EPM arena for which we used Virusolve (Amity International) to avoid damage to 857 the material. Behaviour was recorded using a Google Pixel 5a camera at 1080p/60 858 fps. Videos were then cropped and down sampled by shell scripting using FFmpeg. 859 Open field test 860 Mice were placed in a 40 x 40 x 40 cm white arena and allowed to explore freely. For 861 cohort 1, dim red light (4 lux) was used, while for cohort 2, the OF test was 862 conducted under bright overhead lighting (500 lux). 10-minute videos were analysed 863 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 26, 2025. ; https://doi.org/10.1101/2025.04.24.650541doi: bioRxiv preprint 28 using CleverSys (VA, USA). Outcome parameters included time spent in the centre, 864 total distance travelled, velocity, and thigmotaxis (edge exploration). 865 Novel object recognition 866 Cohort 2 mice were initially habituated to the arena under low light conditions (4 867 lux) over two days, five minutes per session. On the training day, two identical 868 objects were introduced, and mice were allowed to explore for five minutes. After a 869 one-hour delay, one object was replaced with a novel one. The videos were analysed 870 using CleverSys software stereotypic event “Sniffing” module. Objects were 871 manually outlined with the polygon tool. An interaction with an object was recorded 872 when the mouse's nose was within 5 mm of the object. The novelty preference was 873 determined by calculating the proportion of time spent exploring the novel object 874 relative to the total exploration time of both objects, with a recognition index chance 875 level of 50%. 876 Elevated plus maze 877 For cohort 1, EPM was administered to behaviourally naïve animals, while in the for 878 cohort 2, it was conducted after OF and NOR tests. In both cases, mice were placed 879 in the closed arm of the arena (65 x 65 x 55 cm, elevated 40 cm) under full overhead 880 lighting (500 lux) and allowed to explore for five minutes. The arena was divided into 881 closed, middle, and open areas for analysis with CleverSys software. The number of 882 head dips and stretch-attend postures was recorded using BORIS software, and 883 testing accuracy was compared with the results of an independent scorer. 884 Prepulse inhibition 885 Like NOR, PPI was only conducted in cohort 2. We used an acoustic startle 886 chamber (SR-LAB, San Diego Instruments, San Diego, CA, USA) with a cylindrical 887 Plexiglas enclosure horizontally mounted on a mobile platform within a sound-888 proofed isolation chamber. A high-frequency loudspeaker positioned above the 889 enclosure emitted continuous background noise at 65 dB, along with the 890 experimental acoustic stimuli. The startle response was recorded by converting 891 Plexiglas enclosure vibrations into millivolt signals using a piezoelectric unit. 892 Each session started with a five-minute acclimatisation period to the 65 dB 893

noise, followed by five startle-alone trials at 120 dB to ensure 894 habituation. Mice then received 10 prepulse stimuli at 3, 6, and 12 dB above 895 background, each preceding a 120 dB pulse in a pseudo-randomised order, 896 interspersed with 10 no-stimulus trials and 10 startle-alone trials. The session 897 concluded with five final startle-alone pulses. The inter-trial interval was randomised 898 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 26, 2025. ; https://doi.org/10.1101/2025.04.24.650541doi: bioRxiv preprint 29 between 9-15 seconds to prevent expectation-based modulation of the startle 899 response. %PPI was calculated for each prepulse using the formula: %𝑃𝑃𝐼 =900 #$%&' )%*+' (-.!"#) 0 #$%&' 1234 #5'#$%&' (-.!"#) #$%&' )%*+' (-.!"#) . 901 Statistical procedure for behavioural outcome measures 902 All statistical analyses were conducted using Python 3.11.7, using relevant libraries 903 such as SciPy and statsmodels. For both studies, the Shapiro-Wilk test was used to 904 assess the normality of data distributions, while Levene’s test was used to evaluate 905 homogeneity of variances. When either assumption of normality or equal variance 906 was violated, Kruskal-Wallis test was used, followed by Dunn’s test for post hoc 907 pairwise comparisons with Bonferroni correction for multiple testing where 908 applicable. For datasets meeting parametric assumptions, different approaches were 909 used based on the study design. For cohort 1, a two-way ANOVA was performed to 910 analyse the effect of sex, genotype and sex-by-genotype interaction, followed by 911 Tukey’s test for post hoc comparisons. Bonferroni correction was applied to adjust 912 for multiple comparisons. Due to the relatively small sample size (n = 38) in cohort 2, 913 sex-by-genotype interactions were not analysed. Instead, Student’s t-test was used for 914 comparisons between groups. 915 To balance statistical rigor with preserving power in exploratory contexts, 916 behavioural measures were not universally corrected for multiplicity. Instead, 917 robustness was inferred through replication across independent cohorts and broad 918 phenotypic consistency. The threshold for statistical significance was set at p < 0.05. 919 Individual mice served as the experimental units in all analyses. 920 Code availability 921 Data and code to reproduce figures will be added to 922 https://github.com/hannalemmik upon publication. The MRI processing code is 923 available upon request from Eugene Kim. 924 Data availability 925 Behaviour videos will be added to figshare, and MRI data will be added to 926 openneuro. The intermediary analysis files will be added to github at 927 https://github.com/hannalemmik. 928 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 26, 2025. ; https://doi.org/10.1101/2025.04.24.650541doi: bioRxiv preprint 30

929 We thank Bao Lu and Craig Gerard for providing the C3ar1 knockout mice, Dr 930 Marija M. Petrinovic for providing prepulse inhibition testing equipment and KCL 931 Biological Service Unit staff for animal care. This work was funded by the MRC 932 grant “Complement C3aR in adolescent synaptic pruning and risk for anxiety” 933 (MR/W004607/1). Hanna Lemmik was funded by the Wellcome Trust as part of 934 the “Neuro-Immune Interactions in Health & Disease” Wellcome Trust PhD 935 Programme (218452/Z/19/Z). 936 Author contributions 937 HL - writing - original draft, writing - review & editing, conceptualization, 938 methodology, software, formal analysis, data curation, investigation, visualization, 939 funding acquisition; EK - methodology, software, formal analysis, investigation, data 940 curation, writing - review & editing, visualization, funding acquisition; EM - 941 methodology, software, investigation, writing - review & editing; DM - validation, 942 methodology, investigation; MB - software, formal analysis, data curation, 943 visualization; ZL - investigation, validation; DA - investigation, Esther - investigation; 944 MES - funding acquisition; WZ - resources, supervision; AI - resources, supervision; 945 DC - writing - original draft, writing - review & editing, supervision, project 946 administration, funding acquisition, conceptualization, methodology; LW - writing - 947 original draft, writing - review & editing, supervision, funding acquisition, 948 conceptualization, methodology. 949

950 Avants, Brian B., Nicholas J. Tustison, Gang Song, Philip A. Cook, Arno Klein, and 951 James C. Gee. 2011. “A Reproducible Evaluation of ANTs Similarity Metric 952 Performance in Brain Image Registration.” NeuroImage 54 (3): 2033–44. 953 Bailey, Kathleen R., Nathan R. Rustay, and Jacqueline N. Crawley. 2006. 954 “Behavioral Phenotyping of Transgenic and Knockout Mice: Practical Concerns and 955 Potential Pitfalls.” ILAR Journal 47 (2): 124–31. 956 Bassett, Danielle S., Edward T. Bullmore, Andreas Meyer-Lindenberg, José A. Apud, 957 Daniel R. Weinberger, and Richard Coppola. 2009. “Cognitive Fitness of Cost-958 Efficient Brain Functional Networks.” Proceedings of the National Academy of 959 Sciences of the United States of America 106 (28): 11747–52. 960 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 26, 2025. ; https://doi.org/10.1101/2025.04.24.650541doi: bioRxiv preprint 31 Baum, Matthew L., Daniel K. Wilton, Rachel G. Fox, Alanna Carey, Yu-Han H. 961 Hsu, Ruilong Hu, Henna J. Jäntti, et al. 2024. “CSMD1 Regulates Brain 962 Complement Activity and Circuit Development.” Brain, Behavior, and Immunity, 963 March. 964 Bénard, Magalie, Emilie Raoult, David Vaudry, Jérome Leprince, Anthony Falluel-965 Morel, Bruno J. Gonzalez, Ludovic Galas, Hubert Vaudry, and Marc Fontaine. 2008. 966 “Role of Complement Anaphylatoxin Receptors (C3aR, C5aR) in the Development 967 of the Rat Cerebellum.” Molecular Immunology 45 (14): 3767–74. 968 Berlot, Rok, Claudia Metzler-Baddeley, M. Arfan Ikram, Derek K. Jones, and 969 Michael J. O’Sullivan. 2016. “Global Efficiency of Structural Networks Mediates 970 Cognitive Control in Mild Cognitive Impairment.” Frontiers in Aging Neuroscience 971 8 (December): 292. 972 Brouwer, Rachel M., René C. W. Mandl, Hugo G. Schnack, Inge L. C. van Soelen, 973 G. Caroline van Baal, Jiska S. Peper, René S. Kahn, Dorret I. Boomsma, and H. E. 974 Hulshoff Pol. 2012. “White Matter Development in Early Puberty: A Longitudinal 975 Volumetric and Diffusion Tensor Imaging Twin Study.” PloS One 7 (4): e32316. 976 Brust, Vera, Philipp M. Schindler, and Lars Lewejohann. 2015. “Lifetime 977 Development of Behavioural Phenotype in the House Mouse (Mus Musculus).” 978 Frontiers in Zoology 12 Suppl 1 (S1): S17. 979 Bullmore, Ed, and Olaf Sporns. 2009. “Complex Brain Networks: Graph Theoretical 980 Analysis of Structural and Functional Systems.” Nature Reviews. Neuroscience 10 (3): 981 186–98. 982 Byrne, Jonah F., Colm Healy, Melanie Föcking, Meike Heurich, Subash Raj Susai, 983 David Mongan, Kieran Wynne, et al. 2024. “Plasma Complement and Coagulation 984 Proteins as Prognostic Factors of Negative Symptoms: An Analysis of the NAPLS 2 985 and 3 Studies.” Brain, Behavior, and Immunity 119 (July): 188–96. 986 Cai, Lin, Qi Dong, and Haijing Niu. 2018. “The Development of Functional 987 Network Organization in Early Childhood and Early Adolescence: A Resting-State 988 FNIRS Study.” Developmental Cognitive Neuroscience 30 (April): 223–35. 989 Caseras, Xavier, Emily Simmonds, Antonio F. Pardiñas, Richard Anney, Sophie E. 990 Legge, James T. R. Walters, Neil A. Harrison, Michael C. O’Donovan, and 991 Valentina Escott-Price. 2024. “Common Risk Alleles for Schizophrenia within the 992 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 26, 2025. ; https://doi.org/10.1101/2025.04.24.650541doi: bioRxiv preprint 32 Major Histocompatibility Complex Predict White Matter Microstructure.” 993 Translational Psychiatry 14 (1): 194. 994 Chahboune, Halima, Laura R. Ment, William B. Stewart, Xiaoxian Ma, Douglas L. 995 Rothman, and Fahmeed Hyder. 2007. “Neurodevelopment of C57B/L6 Mouse 996 Brain Assessed by in Vivo Diffusion Tensor Imaging.” NMR in Biomedicine 20 (3): 997 375–82. 998 Chan, Shi Yu, Eamon Fitzgerald, Zhen Ming Ngoh, Janice Lee, Jasmine Chuah, 999 Joanne S. M. Chia, Marielle V. Fortier, et al. 2024. “Examining the Associations 1000 between Microglia Genetic Capacity, Early Life Exposures and White Matter 1001 Development at the Level of the Individual.” Brain, Behavior, and Immunity 119 1002 (July): 781–91. 1003 Chaplin, Hugh, Jr. 2020. “Review: The Burgeoning History of the Complement 1004 System 1888-2005.” Immunohematology / American Red Cross 21 (3): 85–93. 1005 Chew, Gabriel, and Enrico Petretto. 2019. “Transcriptional Networks of Microglia 1006 in Alzheimer’s Disease and Insights into Pathogenesis.” Genes 10 (10). 1007 https://doi.org/10.3390/genes10100798. 1008 Coulthard, L. G., O. A. Hawksworth, J. Conroy, J. D. Lee, and T. M. Woodruff. 1009 2018. “Complement C3a Receptor Modulates Embryonic Neural Progenitor Cell 1010 Proliferation and Cognitive Performance.” Molecular Immunology 101 (September): 1011 176–81. 1012 Courten-Myers, G. M. de. 1999. “The Human Cerebral Cortex: Gender Differences 1013 in Structure and Function.” Journal of Neuropathology and Experimental Neurology 1014 58 (3): 217–26. 1015 Coutinho, Joana Fernandes, Sara Veiga Fernandesl, José Miguel Soares, Liliana Maia, 1016 Óscar Filipe Gonçalves, and Adriana Sampaio. 2016. “Default Mode Network 1017 Dissociation in Depressive and Anxiety States.” Brain Imaging and Behavior 10 (1): 1018 147–57. 1019 Crawley, J. N. 2007. “What’s Wrong with My Mouse?: Behavioral Phenotyping of 1020 Transgenic and Knockout Mice.” (2nd ed.). John Wiley & Sons Inc. 1021 Crews, David, David Rushworth, Francisco Gonzalez-Lima, and Sonoko Ogawa. 1022 2009. “Litter Environment Affects Behavior and Brain Metabolic Activity of Adult 1023 Knockout Mice.” Frontiers in Behavioral Neuroscience 3 (August): 12. 1024 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 26, 2025. ; https://doi.org/10.1101/2025.04.24.650541doi: bioRxiv preprint 33 Crider, A., T. Feng, C. D. Pandya, T. Davis, A. Nair, A. O. Ahmed, B. Baban, D. 1025 Turecki, and A. Pillai. 2018. “Complement Component 3a Receptor Deficiency 1026 Attenuates Chronic Stress-Induced Monocyte Infiltration and Depressive-like 1027 Behavior.” Brain, Behavior, and Immunity 70 (May): 246–56. 1028 Deivasigamani, Senthilkumar, Mariya T. Miteva, Silvia Natale, Daniel Gutierrez-1029 Barragan, Bernadette Basilico, Silvia Di Angelantonio, Laetitia Weinhard, et al. 2023. 1030 “Microglia Complement Signaling Promotes Neuronal Elimination and Normal 1031 Brain Functional Connectivity.” Cerebral Cortex 33 (21): 10750–60. 1032 El-Brolosy, Mohamed A., Zacharias Kontarakis, Andrea Rossi, Carsten Kuenne, 1033 Stefan Günther, Nana Fukuda, Khrievono Kikhi, et al. 2019. “Genetic 1034 Compensation Triggered by Mutant MRNA Degradation.” Nature 568 (7751): 1035 193–97. 1036 El-Brolosy, Mohamed A., and Didier Y. R. Stainier. 2017. “Genetic Compensation: 1037 A Phenomenon in Search of Mechanisms.” PLoS Genetics 13 (7): e1006780. 1038 Ellegood, Jacob, Yohan Yee, Travis M. Kerr, Christopher L. Muller, Randy D. 1039 Blakely, R. Mark Henkelman, Jeremy Veenstra-VanderWeele, and Jason P. Lerch. 1040 2018. “Analysis of Neuroanatomical Differences in Mice with Genetically Modified 1041 Serotonin Transporters Assessed by Structural Magnetic Resonance Imaging.” 1042 Molecular Autism 9 (1): 24. 1043 Falangola, Maria Fatima, Siddhartha Dhiman, Joshua Voltin, and Jens H. Jensen. 1044 2023. “Quantitative Microglia Morphological Features Correlate with Diffusion 1045 MRI in 2-Month-Old 3xTg-AD Mice.” Magnetic Resonance Imaging 103 1046 (November): 8–17. 1047 Filipello, Fabia, Raffaella Morini, Irene Corradini, Valerio Zerbi, Alice Canzi, 1048 Bernadeta Michalski, Marco Erreni, et al. 2018. “The Microglial Innate Immune 1049 Receptor TREM2 Is Required for Synapse Elimination and Normal Brain 1050 Connectivity.” Immunity 48 (5): 979-991.e8. 1051 Fitch, W. M., and W. R. Atchley. 1985. “Evolution in Inbred Strains of Mice 1052 Appears Rapid.” Science (New York, N.Y.) 228 (4704): 1169–75. 1053 Forlim, Caroline Garcia, Leonie Klock, Jürgen Gallinat, and Simone Kühn. 2024. 1054 “Altered Resting-State Functional Connectivity in a Thalamo-Cortico-Cerebellar 1055 Network in Patients with Schizophrenia.” Scientific Reports 14 (1): 26284. 1056 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 26, 2025. ; https://doi.org/10.1101/2025.04.24.650541doi: bioRxiv preprint 34 Ge, Tong-Qi, Pei-Pei Guan, and Pu Wang. 2024. “Complement 3a Induces the 1057 Synapse Loss via C3aR in Mitochondria-Dependent NLRP3 Activating 1058 Mechanisms during the Development and Progression of Alzheimer’s Disease.” 1059 Neuroscience and Biobehavioral Reviews 165 (105868): 105868. 1060 Gedam, Manasee, Michele M. Comerota, Nicholas E. Propson, Tao Chen, Feng Jin, 1061 Meng C. Wang, and Hui Zheng. 2023. “Complement C3aR Depletion Reverses 1062 HIF-1α-Induced Metabolic Impairment and Enhances Microglial Response to Aβ 1063 Pathology.” The Journal of Clinical Investigation 133 (12). 1064 Geng, Haiyang, Xuebing Li, Jie Chen, Xinying Li, and Ruolei Gu. 2015. “Decreased 1065 Intra- and Inter-Salience Network Functional Connectivity Is Related to Trait 1066 Anxiety in Adolescents.” Frontiers in Behavioral Neuroscience 9: 350. 1067 Girardi, Guillermina, Joshua J. Lingo, Sherry D. Fleming, and Jean F. Regal. 2020. 1068 “Essential Role of Complement in Pregnancy: From Implantation to Parturition and 1069 Beyond.” Frontiers in Immunology 11 (July): 1681. 1070 Gnanaguru, Gopalan, Steven J. Tabor, Gracia M. Bonilla, Ruslan Sadreyev, Kentaro 1071 Yuda, Jörg Köhl, and Kip M. Connor. 2023. “Microglia Refine Developing Retinal 1072 Astrocytic and Vascular Networks through the Complement C3/C3aR Axis.” 1073 Development 150 (5). 1074 Gorski, R. A., J. H. Gordon, J. E. Shryne, and A. M. Southam. 1978. “Evidence for a 1075 Morphological Sex Difference within the Medial Preoptic Area of the Rat Brain.” 1076 Brain Research 148 (2): 333–46. 1077 Grandjean, J., C. Canella, C. Anckaerts, C. Ayrancı, S. Bougacha, T. Bienert, D. 1078 Buehlmann, et al. 2020. “Common Functional Networks in the Mouse Brain 1079 Revealed by Multi-Centre Resting-State FMRI Analysis.” NeuroImage 205 1080 (January). 1081 Grandjean, Joanes, Aileen Schroeter, Imene Batata, and Markus Rudin. 2014. 1082 “Optimization of Anesthesia Protocol for Resting-State FMRI in Mice Based on 1083 Differential Effects of Anesthetics on Functional Connectivity Patterns.” 1084 NeuroImage 102 Pt 2 (November): 838–47. 1085 Gretenkord, Sabine, Johanna K. Kostka, Henrike Hartung, Katja Watznauer, David 1086 Fleck, Angélica Minier-Toribio, Marc Spehr, and Ileana L. Hanganu-Opatz. 2019. 1087 “Coordinated Electrical Activity in the Olfactory Bulb Gates the Oscillatory 1088 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 26, 2025. ; https://doi.org/10.1101/2025.04.24.650541doi: bioRxiv preprint 35 Entrainment of Entorhinal Networks in Neonatal Mice.” PLoS Biology 17 (1): 1089 e2006994. 1090 Guma, Elisa, Antoine Beauchamp, Siyuan Liu, Elizabeth Levitis, Jacob Ellegood, 1091 Linh Pham, Rogier B. Mars, Armin Raznahan, and Jason P. Lerch. 2024. 1092 “Comparative Neuroimaging of Sex Differences in Human and Mouse Brain 1093 Anatomy.” ELife 13 (March). 1094 Guma, Elisa, Maude Bordeleau, Fernando González Ibáñez, Katherine Picard, Emily 1095 Snook, Gabriel Desrosiers-Grégoire, Shoshana Spring, et al. 2022. “Differential 1096 Effects of Early or Late Exposure to Prenatal Maternal Immune Activation on 1097 Mouse Embryonic Neurodevelopment.” Proceedings of the National Academy of 1098 Sciences of the United States of America 119 (12): e2114545119. 1099 Hadley, Jennifer Ann, Nina Vanessa Kraguljac, David Matthew White, Lawrence 1100 Ver Hoef, Janell Tabora, and Adrienne Carol Lahti. 2016. “Change in Brain 1101 Network Topology as a Function of Treatment Response in Schizophrenia: A 1102 Longitudinal Resting-State FMRI Study Using Graph Theory.” Npj Schizophrenia 2 1103 (1): 16014. 1104 Hagmann, P., O. Sporns, N. Madan, L. Cammoun, R. Pienaar, V. J. Wedeen, R. 1105 Meuli, J-P Thiran, and P. E. Grant. 2010. “White Matter Maturation Reshapes 1106 Structural Connectivity in the Late Developing Human Brain.” Proceedings of the 1107 National Academy of Sciences of the United States of America 107 (44): 19067–72. 1108 Hammond, Timothy R., Connor Dufort, Lasse Dissing-Olesen, Stefanie Giera, 1109 Adam Young, Alec Wysoker, Alec J. Walker, et al. 2019. “Single-Cell RNA 1110 Sequencing of Microglia throughout the Mouse Lifespan and in the Injured Brain 1111 Reveals Complex Cell-State Changes.” Immunity 50 (1): 253-271.e6. 1112 Hauser, Alexander S., Misty M. Attwood, Mathias Rask-Andersen, Helgi B. Schiöth, 1113 and David E. Gloriam. 2017. “Trends in GPCR Drug Discovery: New Agents, 1114 Targets and Indications.” Nature Reviews. Drug Discovery 16 (12): 829–42. 1115 Hawkins, R., A. S. Shatil, L. Lee, A. Sengupta, L. Zhang, S. Morrow, and R. I. Aviv. 1116 2020. “Reduced Global Efficiency and Random Network Features in Patients with 1117 Relapsing-Remitting Multiple Sclerosis with Cognitive Impairment.” AJNR. 1118 American Journal of Neuroradiology 41 (3): 449–55. 1119 Hikishima, Keigo, Tomokazu Tsurugizawa, Kazumi Kasahara, Ryo Takagi, Kiyoshi 1120 Yoshinaka, and Naotaka Nitta. 2023. “Brain-Wide Mapping of Resting-State 1121 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 26, 2025. ; https://doi.org/10.1101/2025.04.24.650541doi: bioRxiv preprint 36 Networks in Mice Using High-Frame Rate Functional Ultrasound.” NeuroImage 1122 279 (120297): 120297. 1123 Hines, M., L. S. Allen, and R. A. Gorski. 1992. “Sex Differences in Subregions of the 1124 Medial Nucleus of the Amygdala and the Bed Nucleus of the Stria Terminalis of the 1125 Rat.” Brain Research 579 (2): 321–26. 1126 Holmdahl, Rikard, and Bernard Malissen. 2012. “The Need for Littermate 1127 Controls.” European Journal of Immunology 42 (1): 45–47. 1128 Hong, Soyon, Victoria F. Beja-Glasser, Bianca M. Nfonoyim, Arnaud Frouin, 1129 Shaomin Li, Saranya Ramakrishnan, Katherine M. Merry, et al. 2016. “Complement 1130 and Microglia Mediate Early Synapse Loss in Alzheimer Mouse Models.” Science 352 1131 (6286): 712–16. 1132 Howes, Oliver D., Bernard R. Bukala, and Katherine Beck. 2024. “Schizophrenia: 1133 From Neurochemistry to Circuits, Symptoms and Treatments.” Nature Reviews. 1134 Neurology 20 (1): 22–35. 1135 Humbles, Alison A., Bao Lu, Christy A. Nilsson, Craig Lilly, Elliot Israel, Yuko 1136 Fujiwara, Norma P. Gerard, and Craig Gerard. 2000. “A Role for the C3a 1137 Anaphylatoxin Receptor in the Effector Phase of Asthma.” Nature 406 (6799): 998–1138 1001. 1139 Jenkinson, Mark, Christian F. Beckmann, Timothy E. J. Behrens, Mark W. 1140 Woolrich, and Stephen M. Smith. 2012. “FSL.” NeuroImage 62 (2): 782–90. 1141 Jiang, Weixiong, Zhen Zhou, Guoshi Li, Weiyan Yin, Zhengwang Wu, Li Wang, 1142 Maryam Ghanbari, et al. 2023. “Mapping the Evolution of Regional Brain Network 1143 Efficiency and Its Association with Cognitive Abilities during the First Twenty-1144 Eight Months of Life.” Developmental Cognitive Neuroscience 63 (101284): 101284. 1145 Jiménez, Jessica A., and Mark J. Zylka. 2021. “Controlling Litter Effects to Enhance 1146 Rigor and Reproducibility with Rodent Models of Neurodevelopmental Disorders.” 1147 Journal of Neurodevelopmental Disorders 13 (1): 2. 1148 Kalbassi, Shireene, Sven O. Bachmann, Ellen Cross, Victoria H. Roberton, and 1149 Stéphane J. Baudouin. 2017. “Male and Female Mice Lacking Neuroligin-3 Modify 1150 the Behavior of Their Wild-Type Littermates.” ENeuro 4 (4). 1151 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 26, 2025. ; https://doi.org/10.1101/2025.04.24.650541doi: bioRxiv preprint 37 Kiani Shabestari, Sepideh, Samuel Morabito, Emma Pascal Danhash, Amanda 1152 McQuade, Jessica Ramirez Sanchez, Emily Miyoshi, Jean Paul Chadarevian, et al. 1153 2022. “Absence of Microglia Promotes Diverse Pathologies and Early Lethality in 1154 Alzheimer’s Disease Mice.” Cell Reports 39 (11): 110961. 1155 Kielar, Catherine, Stephen J. Sawiak, Paloma Navarro Negredo, Desmond H. Y. Tse, 1156 and A. Jennifer Morton. 2012. “Tensor-Based Morphometry and Stereology Reveal 1157 Brain Pathology in the Complexin1 Knockout Mouse.” PloS One 7 (2): e32636. 1158 Kim, Eugene, Ines Carreira Figueiredo, Camilla Simmons, Karen Randall, Loreto 1159 Rojo Gonzalez, Tobias Wood, Brigida Ranieri, et al. 2023. “Mapping Acute 1160 Neuroinflammation in Vivo with Diffusion-MRI in Rats given a Systemic 1161 Lipopolysaccharide Challenge.” Brain, Behavior, and Immunity 113 (July): 289–1162 301. 1163 Koenis, Marinka M. G., Rachel M. Brouwer, Suzanne C. Swagerman, Inge L. C. van 1164 Soelen, Dorret I. Boomsma, and Hilleke E. Hulshoff Pol. 2018. “Association 1165 between Structural Brain Network Efficiency and Intelligence Increases during 1166 Adolescence.” Human Brain Mapping 39 (2): 822–36. 1167 Kunz, Natalia, and Claudia Kemper. 2021. “Complement Has Brains-Do 1168 Intracellular Complement and Immunometabolism Cooperate in Tissue 1169 Homeostasis and Behavior?” Frontiers in Immunology 12 (February): 629986. 1170 Lerch, Jason P., Lisa Gazdzinski, Jurgen Germann, John G. Sled, R. Mark 1171 Henkelman, and Brian J. Nieman. 2012. “Wanted Dead or Alive? The Tradeoff 1172 between in-Vivo versus Ex-Vivo MR Brain Imaging in the Mouse.” Frontiers in 1173 Neuroinformatics 0 (March): 6. 1174 Li, Chenchen, Bo Liu, Jingyi Xu, Bin Jing, Lijie Guo, Liyong Wang, Milin Wang, et 1175 al. 2023. “Phloretin Decreases Microglia-Mediated Synaptic Engulfment to Prevent 1176 Chronic Mild Stress-Induced Depression-like Behaviors in the MPFC.” Theranostics 1177 13 (3): 955–72. 1178 Lian, Hong, Li Yang, Allysa Cole, Lu Sun, Angie C-A Chiang, Stephanie W. Fowler, 1179 David J. Shim, et al. 2015. “NFκB-Activated Astroglial Release of Complement C3 1180 Compromises Neuronal Morphology and Function Associated with Alzheimer’s 1181 Disease.” Neuron 85 (1): 101–15. 1182 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 26, 2025. ; https://doi.org/10.1101/2025.04.24.650541doi: bioRxiv preprint 38 Ludewig, Katja, Mark A. Geyer, and Franz X. Vollenweider. 2003. “Deficits in 1183 Prepulse Inhibition and Habituation in Never-Medicated, First-Episode 1184 Schizophrenia.” Biological Psychiatry 54 (2): 121–28. 1185 Madeshiya, Amit Kumar, Carl Whitehead, Ashutosh Tripathi, and Anilkumar Pillai. 1186 2022. “C1q Deletion Exacerbates Stress-Induced Learned Helplessness Behavior and 1187 Induces Neuroinflammation in Mice.” Translational Psychiatry 12 (1): 50. 1188 Mamane, Yaël, Chi Chung Chan, Genevieve Lavallee, Nicolas Morin, Li-Jing Xu, 1189 Jingqi Huang, Robert Gordon, et al. 2009. “The C3a Anaphylatoxin Receptor Is a 1190 Key Mediator of Insulin Resistance and Functions by Modulating Adipose Tissue 1191 Macrophage Infiltration and Activation.” Diabetes 58 (9): 2006–17. 1192 McCarty, Richard. 2017. “Cross-Fostering: Elucidating the Effects of 1193 Gene×environment Interactions on Phenotypic Development.” Neuroscience and 1194 Biobehavioral Reviews 73 (February): 219–54. 1195 McCutcheon, Robert A., Tiago Reis Marques, and Oliver D. Howes. 2020. 1196 “Schizophrenia-an Overview.” JAMA Psychiatry (Chicago, Ill.) 77 (2): 201–10. 1197 Mena, Auxiliadora, Juan C. Ruiz-Salas, Andrea Puentes, Inmaculada Dorado, 1198 Miguel Ruiz-Veguilla, and Luis G. De la Casa. 2016. “Reduced Prepulse Inhibition 1199 as a Biomarker of Schizophrenia.” Frontiers in Behavioral Neuroscience 10 (October): 1200 202. 1201 Mommert, Susanne, Derya Aslan, Lisanne Ratz, Holger Stark, Ralf Gutzmer, and 1202 Thomas Werfel. 2018. “The Anaphylatoxin C3a Receptor Expression on Human 1203 M2 Macrophages Is Down-Regulated by Stimulating the Histamine H4 Receptor 1204 and the IL-4 Receptor.” Journal of Innate Immunity 10 (4): 349–62. 1205 Nasseef, Md Taufiq, Gabriel A. Devenyi, Anna E. Mechling, Laura-Adela Harsan, 1206 M. Mallar Chakravarty, Brigitte Lina Kieffer, and Emmanuel Darcq. 2018. 1207 “Deformation-Based Morphometry MRI Reveals Brain Structural Modifications in 1208 Living Mu Opioid Receptor Knockout Mice.” Frontiers in Psychiatry 9 (December): 1209 643. 1210 O’Connell, Kevin S., Ida E. Sønderby, Oleksandr Frei, Dennis van der Meer, Lavinia 1211 Athanasiu, Olav B. Smeland, Dag Alnæs, et al. 2021. “Association between 1212 Complement Component 4A Expression, Cognitive Performance and Brain 1213 Imaging Measures in UK Biobank.” Psychological Medicine 52 (15): 1–11. 1214 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 26, 2025. ; https://doi.org/10.1101/2025.04.24.650541doi: bioRxiv preprint 39 Olsen, Reid H. J., Jeffrey F. DiBerto, Justin G. English, Alexis M. Glaudin, Brian E. 1215 Krumm, Samuel T. Slocum, Tao Che, et al. 2020. “TRUPATH, an Open-Source 1216 Biosensor Platform for Interrogating the GPCR Transducerome.” Nature Chemical 1217 Biology 16 (8): 841–49. 1218 Osimo, Emanuele Felice, Katherine Beck, Tiago Reis Marques, and Oliver D. 1219 Howes. 2019. “Synaptic Loss in Schizophrenia: A Meta-Analysis and Systematic 1220 Review of Synaptic Protein and MRNA Measures.” Molecular Psychiatry 24 (4): 1221 549–61. 1222 Paus, T., M. Keshavan, and J. N. Giedd. 2008. “Why Do Many Psychiatric Disorders 1223 Emerge during Adolescence?” Nature Reviews. Neuroscience 9 (12): 947–57. 1224 Piekarski, David J., Natalie M. Zahr, Qingyu Zhao, Uran Ferizi, Kilian M. Pohl, 1225 Edith V. Sullivan, and Adolf Pfefferbaum. 2023. “White Matter Microstructural 1226 Integrity Continues to Develop from Adolescence to Young Adulthood in Mice and 1227 Humans: Same Phenotype, Different Mechanism.” Neuroimage. Reports 3 (3). 1228 Pintér, O., Zoltan Beda, Z. Csaba, and I. Gerendai. 2007. “Differences in the Onset 1229 of Puberty in Selected Inbred Mouse Strains.” Endocrine Abstracts 14 (April). 1230 Pozo-Rodrigálvarez, Andrea, Roosa Ollaranta, Jenny Skoog, Milos Pekny, and 1231 Marcela Pekna. 2021. “Hyperactive Behavior and Altered Brain Morphology in 1232 Adult Complement C3a Receptor Deficient Mice.” Frontiers in Immunology 12 1233 (February): 406. 1234 Qiu, Lily R., Darren J. Fernandes, Kamila U. Szulc-Lerch, Jun Dazai, Brian J. 1235 Nieman, Daniel H. Turnbull, Jane A. Foster, Mark R. Palmert, and Jason P. Lerch. 1236 2018. “Mouse MRI Shows Brain Areas Relatively Larger in Males Emerge before 1237 Those Larger in Females.” Nature Communications 9 (1): 1–15. 1238 Quell, Katharina M., Christian M. Karsten, Anna Kordowski, Larissa Nogueira 1239 Almeida, Daria Briukhovetska, Anna V. Wiese, Jing Sun, et al. 2017. “Monitoring 1240 C3aR Expression Using a Floxed TdTomato-C3aR Reporter Knock-in Mouse.” The 1241 Journal of Immunology 199 (2): 688–706. 1242 Reynolds, Jess E., Melody N. Grohs, Deborah Dewey, and Catherine Lebel. 2019. 1243 “Global and Regional White Matter Development in Early Childhood.” 1244 NeuroImage 196 (August): 49–58. 1245 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 26, 2025. ; https://doi.org/10.1101/2025.04.24.650541doi: bioRxiv preprint 40 Rodriguez, Pedro, Lauren J. Laskowski, Jean Pierre Pallais, Hailey A. Bock, Natalie 1246 G. Cavalco, Emilie I. Anderson, Maggie M. Calkins, et al. 2023. “Functional 1247 Profiling of the G Protein-Coupled Receptor C3aR1 Reveals Ligand-Mediated 1248 Biased Agonism.” The Journal of Biological Chemistry 300 (1): 105549. 1249 Rojo, Rocío, Anna Raper, Derya D. Ozdemir, Lucas Lefevre, Kathleen Grabert, Evi 1250 Wollscheid-Lengeling, Barry Bradford, et al. 2019. “Deletion of a Csf1r Enhancer 1251 Selectively Impacts CSF1R Expression and Development of Tissue Macrophage 1252 Populations.” Nature Communications 10 (1): 3215. 1253 Rubinov, Mikail, and Olaf Sporns. 2010. “Complex Network Measures of Brain 1254 Connectivity: Uses and Interpretations.” NeuroImage 52 (3): 1059–69. 1255 Schimmelpfennig, Jakub, Jan Topczewski, Wojciech Zajkowski, and Kamila 1256 Jankowiak-Siuda. 2023. “The Role of the Salience Network in Cognitive and 1257 Affective Deficits.” Frontiers in Human Neuroscience 17 (March): 1133367. 1258 Schizophrenia Working Group of the Psychiatric Genomics Consortium. 2014. 1259 “Biological Insights from 108 Schizophrenia-Associated Genetic Loci.” Nature 511 1260 (7510): 421–27. 1261 Sekar, A., A. R. Bialas, H. de Rivera, A. Davis, T. R. Hammond, Kamitaki N, Tooley 1262 K, et al. 2016. “Schizophrenia Risk from Complex Variation of Complement 1263 Component 4.” Nature 530 (7589): 177–83. 1264 Semaan, Sheila J., and Alexander S. Kauffman. 2015. “Daily Successive Changes in 1265 Reproductive Gene Expression and Neuronal Activation in the Brains of Pubertal 1266 Female Mice.” Molecular and Cellular Endocrinology 401 (February): 84–97. 1267 Serrano, Maria Elisa, Eugene Kim, Bernard Siow, Da Ma, Loreto Rojo, Camilla 1268 Simmons, Darryl Hayward, et al. 2023. “Investigating Brain Alterations in the 1269 Dp1Tyb Mouse Model of Down Syndrome - ScienceDirect.” Neurobiology of 1270 Disease 188 (November): 106336. 1271 Sforazzini, Francesco, Adam J. Schwarz, Alberto Galbusera, Angelo Bifone, and 1272 Alessandro Gozzi. 2014. “Distributed BOLD and CBV-Weighted Resting-State 1273 Networks in the Mouse Brain.” NeuroImage 87 (February): 403–15. 1274 Smith, Stephen M., Mark Jenkinson, Mark W. Woolrich, Christian F. Beckmann, 1275 Timothy E. J. Behrens, Heidi Johansen-Berg, Peter R. Bannister, et al. 2004. 1276 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 26, 2025. ; https://doi.org/10.1101/2025.04.24.650541doi: bioRxiv preprint 41 “Advances in Functional and Structural MR Image Analysis and Implementation as 1277 FSL.” NeuroImage 23 Suppl 1 (January): S208-19. 1278 Stevens, Beth, Nicola J. Allen, Luis E. Vazquez, Gareth R. Howell, Karen S. 1279 Christopherson, Navid Nouri, Kristina D. Micheva, et al. 2007. “The Classical 1280 Complement Cascade Mediates CNS Synapse Elimination.” Cell 131 (6): 1164–78. 1281 Sun, Rui, Meng-Yu Tang, Dan Yang, Yan-Yi Zhang, Yi-Heng Xu, Yong Qiao, Bin 1282 Yu, et al. 2024. “C3aR in the Medial Prefrontal Cortex Modulates the Susceptibility 1283 to LPS-Induced Depressive-like Behaviors through Glutamatergic Neuronal 1284 Excitability.” Progress in Neurobiology, April, 102614. 1285 Tabbaa, Manal, Allison Knoll, and Pat Levitt. 2023. “Mouse Population Genetics 1286 Phenocopies Heterogeneity of Human Chd8 Haploinsufficiency.” Neuron 111 (4): 1287 539-556.e5. 1288 Tao, Juan, Jing Zhao, Xiang-Ming Qi, and Yong-Gui Wu. 2021. “Complement-1289 Mediated M2/M1 Macrophage Polarization May Be Involved in Crescent 1290 Formation in Lupus Nephritis.” International Immunopharmacology 101 (Pt A): 1291 108278. 1292 Tasic, Bosiljka, Vilas Menon, Thuc Nghi Nguyen, Tae Kyung Kim, Tim Jarsky, 1293 Zizhen Yao, Boaz Levi, et al. 2016. “Adult Mouse Cortical Cell Taxonomy Revealed 1294 by Single Cell Transcriptomics.” Nature Neuroscience 19 (2): 335–46. 1295 Tasic, Bosiljka, Zizhen Yao, Lucas T. Graybuck, Kimberly A. Smith, Thuc Nghi 1296 Nguyen, Darren Bertagnolli, Jeff Goldy, et al. 2018. “Shared and Distinct 1297 Transcriptomic Cell Types across Neocortical Areas.” Nature 563 (7729): 72–78. 1298 Thirion, Bertrand, Gaël Varoquaux, Elvis Dohmatob, and Jean-Baptiste Poline. 1299 2014. “Which FMRI Clustering Gives Good Brain Parcellations?” Frontiers in 1300 Neuroscience 8 (July): 167. 1301 Tillmon, Haven, Breeanne M. Soteros, Liang Shen, Qifei Cong, Mackenna Wollet, 1302 Julianne General, Hanna Chin, et al. 2024. “Complement and Microglia Activation 1303 Mediate Stress-Induced Synapse Loss in Layer 2/3 of the Medial Prefrontal Cortex in 1304 Male Mice.” Nature Communications 15 (1): 9803. 1305 Tournier, J-Donald, Robert Smith, David Raffelt, Rami Tabbara, Thijs Dhollander, 1306 Maximilian Pietsch, Daan Christiaens, Ben Jeurissen, Chun-Hung Yeh, and Alan 1307 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 26, 2025. ; https://doi.org/10.1101/2025.04.24.650541doi: bioRxiv preprint 42 Connelly. 2019. “MRtrix3: A Fast, Flexible and Open Software Framework for 1308 Medical Image Processing and Visualisation.” NeuroImage 202 (116137): 116137. 1309 Tripathi, Ashutosh, Carl Whitehead, Katelyn Surrao, Ananya Pillai, Amit 1310 Madeshiya, Yong Li, Hesam Khodadadi, et al. 2021. “Type 1 Interferon Mediates 1311 Chronic Stress-Induced Neuroinflammation and Behavioral Deficits via 1312 Complement Component 3-Dependent Pathway.” Molecular Psychiatry 26 (7): 1313 3043–59. 1314 Vadnie, Chelsea A., Kaitlyn A. Petersen, Lauren A. Eberhardt, Mariah A. 1315 Hildebrand, Allison J. Cerwensky, Hui Zhang, Jennifer N. Burns, et al. 2021. “The 1316 Suprachiasmatic Nucleus Regulates Anxiety-like Behavior in Mice.” Frontiers in 1317 Neuroscience 15: 765850. 1318 Valiquette, Vanessa, Elisa Guma, Lani Cupo, Daniel Gallino, Chloe Anastassiadis, 1319 Emily Snook, Gabriel A. Devenyi, and M. Mallar Chakravarty. 2023. “Examining 1320 Litter Specific Variability in Mice and Its Impact on Neurodevelopmental Studies.” 1321 NeuroImage 269 (April): 119888. 1322 Vasek, Michael J., Charise Garber, Denise Dorsey, Douglas M. Durrant, Bryan 1323 Bollman, Allison Soung, Jinsheng Yu, et al. 2016. “A Complement-Microglial Axis 1324 Drives Synapse Loss during Virus-Induced Memory Impairment.” Nature 534 1325 (7608): 538–43. 1326 Vita, A., L. De Peri, G. Deste, and E. Sacchetti. 2012. “Progressive Loss of Cortical 1327 Gray Matter in Schizophrenia: A Meta-Analysis and Meta-Regression of 1328 Longitudinal MRI Studies.” Translational Psychiatry 2 (11): e190. 1329 Wang, Quanxin, Song-Lin Ding, Yang Li, Josh Royall, David Feng, Phil Lesnar, Nile 1330 Graddis, et al. 2020. “The Allen Mouse Brain Common Coordinate Framework: A 1331 3D Reference Atlas.” Cell 181 (4): 936-953.e20. 1332 West, Erin E., and Claudia Kemper. 2023. “Complosome - the Intracellular 1333 Complement System.” Nature Reviews. Nephrology 19 (7): 426–39. 1334 Westacott, Laura J., Niels Haan, Claudia Evison, Omar Marei, Jeremy Hall, Timothy 1335 R. Hughes, Malik Zaben, et al. 2021. “Dissociable Effects of Complement C3 and 1336 C3aR on Survival and Morphology of Adult Born Hippocampal Neurons, Pattern 1337 Separation, and Cognitive Flexibility in Male Mice.” Brain, Behavior, and Immunity 1338 98 (November): 136–50. 1339 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 26, 2025. ; https://doi.org/10.1101/2025.04.24.650541doi: bioRxiv preprint 43 Westacott, Laura J., Trevor Humby, Niels Haan, Sophie A. Brain, Emma-Louise 1340 Bush, Margarita Toneva, Andreea-Ingrid Baloc, et al. 2022. “Complement C3 and 1341 C3aR Mediate Different Aspects of Emotional Behaviours; Relevance to Risk for 1342 Psychiatric Disorder.” Brain, Behavior, and Immunity 99 (January): 70–82. 1343 Westacott, Laura J., and Lawrence S. Wilkinson. 2022. “Complement Dependent 1344 Synaptic Reorganisation During Critical Periods of Brain Development and Risk for 1345 Psychiatric Disorder.” Frontiers in Neuroscience 16 (May): 840266. 1346 Wood, Tobias C., Camilla Simmons, Samuel A. Hurley, Anthony C. Vernon, Joel 1347 Torres, Flavio Dell’Acqua, Steve C. R. Williams, and Diana Cash. 2016. “Whole-1348 Brain Ex-Vivo Quantitative MRI of the Cuprizone Mouse Model.” PeerJ 4 1349 (November): e2632. 1350 Yilmaz, Melis, Esra Yalcin, Jessy Presumey, Ernest Aw, Minghe Ma, Christopher W. 1351 Whelan, Beth Stevens, Steven A. McCarroll, and Michael C. Carroll. 2021. 1352 “Overexpression of Schizophrenia Susceptibility Factor Human Complement C4A 1353 Promotes Excessive Synaptic Loss and Behavioral Changes in Mice.” Nature 1354 Neuroscience 24 (2): 214–24. 1355 Zalesky, Andrew, Alex Fornito, and Edward T. Bullmore. 2010. “Network-Based 1356 Statistic: Identifying Differences in Brain Networks.” NeuroImage 53 (4): 1197–1357 1207. 1358 Zalesky, Andrew, Alex Fornito, Ian H. Harding, Luca Cocchi, Murat Yücel, 1359 Christos Pantelis, and Edward T. Bullmore. 2010. “Whole-Brain Anatomical 1360 Networks: Does the Choice of Nodes Matter?” NeuroImage 50 (3): 970–83. 1361 Zhan, Y., R. C. Paolicelli, F. Sforazzini, L. Weinhard, G. Bolasco, F. Pagani, A. I. 1362 Vyssotski, et al. 2014. “Deficient Neuron-Microglia Signaling Results in Impaired 1363 Functional Brain Connectivity and Social Behavior.” Nature Neuroscience 17 (3): 1364 400–406. 1365 Zhang, Man-Man, Min-Xia Guo, Qiu-Ping Zhang, Xue-Qin Chen, Na-Zhi Li, Qing 1366 Liu, Jie Cheng, et al. 2022. “IL-1R/C3aR Signaling Regulates Synaptic Pruning in 1367 the Prefrontal Cortex of Depression.” Cell & Bioscience 12 (1): 90. 1368 Zheng, Jingwei, Jia’nan Lu, Shuhao Mei, Haijian Wu, Zeyu Sun, Yuanjian Fang, 1369 Shenbin Xu, et al. 2021. “Ceria Nanoparticles Ameliorate White Matter Injury after 1370 Intracerebral Hemorrhage: Microglia-Astrocyte Involvement in Remyelination.” 1371 Journal of Neuroinflammation 18 (1): 43. 1372 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 26, 2025. ; https://doi.org/10.1101/2025.04.24.650541doi: bioRxiv preprint 44 Zhu, Hongru, Changjian Qiu, Yajing Meng, Minlan Yuan, Yan Zhang, Zhengjia 1373 Ren, Yuchen Li, et al. 2017. “Altered Topological Properties of Brain Networks in 1374 Social Anxiety Disorder: A Resting-State Functional MRI Study.” Scientific Reports 1375 7 (1): 43089. 1376 Figures 1377 1378 Figure 1 | C3ar1tm1Cge mutant does not make C3ar1 RNA. Schematic shows 1379 C3ar1 mRNA with its only protein coding exon and the design of the 79 bp PCR 1380 amplicon which targets an exon-1/5’UTR-exon-2 junction of the canonical 1381 C3ar1+/+ transcript. Also shown is another 372 bp amplicon which targets a region 1382 downstream of the deletion and after an alternative start codon. The start of exon 2 is 1383 deleted in C3ar1-/- mice, so PCR should result in no amplification. Similarly, if no 1384 alternative transcript is made, there should be no amplification. (i-ii) Gel images 1385 show PCR products of cDNA from bone-marrow derived interleukin 4 (IL4)-1386 induced M2-like macrophages. Each sample well has a 297 bp Glyceraldehyde 3-1387 phosphate dehydrogenase (Gapdh) control band. NEG: no reverse transcriptase 1388 negative control. i) canonical transcript, ii) hypothesised alternative transcript. -/- = 1389 C3ar1-/-, +/+ = C3ar1+/+. 1390 1391 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 26, 2025. ; https://doi.org/10.1101/2025.04.24.650541doi: bioRxiv preprint 45 1392 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 26, 2025. ; https://doi.org/10.1101/2025.04.24.650541doi: bioRxiv preprint 46 Figure 2 | Sex but not C3ar1 status influences regional brain volume. (a) 1393 Schematic of the MRI study shown in (d-g). 69 mice were scanned twice in vivo; in 1394 adolescence at ~ postnatal day (PND) 30 (range 27-31) and at adulthood ~ PND90 1395 (range 81-92), as well as once ex vivo after sacrifice after the adulthood scanning 1396 session. Behavioural tests were carried out before the adulthood scanning session. 1397 EPM = elevated plus maze, OF = open field. (b) Total brain volume at PND30. 1398 Two-way ANOVA, genotype, sex, genotype * sex, F[1,65] = 0.35, 16.72, 3.26, p = 0.56, 1399 <0.001, 0.76. (c) PND90, two-way ANOVA, genotype, sex, genotype * sex, F[1,65] = 1400 0.74, 0.26, 3.33, p = 0.39, 0.61, 0.07. (b-c) Data presented as mean ± 95% CI. (d) 1401 Panels showing relative regional volume changes (%) overlaid on study-specific 1402 coronal templates (grey). Red hues signify areas larger in C3ar1-/- (i-ii) or females (iii-1403 iv) and blue hues signify areas larger in C3ar1+/+ (i-ii) or males (iii-iv). Transparency 1404 of the colour overlay shows the statistical significance, ranging from family wise error 1405 (FWE)-corrected p value 0.5 to 0 (transparent to opaque, respectively). Areas where 1406 FWE-corrected p value 0.5 are grey (no overlay), meaning that in adolescence genotype comparison (i), no 1408 voxels had a p value < 0.5. The locations of the coronal slices in relation to bregma in 1409 the left most column from top: -7.6, -4.6, -1.6, 1.4 mm. C3ar1+/+ n = 35, 17 males 1410 and 18 females; C3ar1-/- n = 34, 18 males and 16 females. ACAd = dorsal anterior 1411 cingulate cortex, AI = agranular insular cortex, BST = bed nucleus of stria terminalis, 1412 CA = cornu ammonis, CP = caudoputamen, DG = dentate gyrus, MEA = medial 1413 amygdala, MEPO = medial preoptic nucleus, MOs = secondary motor cortex (MOs), 1414 MPO = medial preoptic area, MS = medial septum, OB = olfactory bulb, PAG = 1415 periaqueductal grey, PRT = pretectal area, SC = superior colliculus, SSp = primary 1416 somatosensory cortex. (e) Coefficient of variation (CV) in adolescence in vivo for 72 1417 regional volumes across all animals in the experiment. (f) CV in adulthood in vivo. 1418 (g) CV in adulthood ex vivo. (e-g) Data are shown with quartiles and whiskers show 1419 the extent of the distribution. Dotted line shows the intergroup mean excluding 1420 cerebrospinal fluid (CSF) areas. Kruskal Wallis p values, all ns. D. hipp = dorsal 1421 hippocampus, v. hipp = ventral hippocampus. 1422 1423 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 26, 2025. ; https://doi.org/10.1101/2025.04.24.650541doi: bioRxiv preprint 47 1424 Figure 3 | Fractional anisotropy does not depend on C3ar1 status but 1425 increases with age. (a) Panels showing voxel-wise fractional anisotropy analysis in 1426 adolescence (top) and adulthood (bottom) corrected for FWE. There are no 1427 significant voxels (no black contour) where genotype effect is significant (p < 0.05). 1428 CC = corpus callosum, OPT = optic tract. (b) In vivo ROI-based median fractional 1429 anisotropy (FA) values in white matter regions for (i) adolescence, (ii) adulthood 1430 and (iii) change over time (adulthood – adolescence). Two-sided one-sample t test 1431 (difference from 0) p values that were adjusted with Benjamini-Hochberg (BH) procedure 1432 (### p value < 0.001). (b-i-iii) Mixed ANOVA with genotype and genotype-by-region 1433 interaction effects, all ns. Data are presented as mean ± 95% CI. 1434 1435 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 26, 2025. ; https://doi.org/10.1101/2025.04.24.650541doi: bioRxiv preprint 48 1436 Figure 4 | Global functional connectivity is not changed in C3ar1-deficient 1437 mice. FC, clustering coefficient and global efficiency at decreasing graph sparsity 1438 levels in adolescence (a-c), and adulthood (d-f), and change over time (g-i). (a-i) 1439 Statistical significance was determined with sexes combined using AUC permutation 1440 testing (10,000 iterations) and the resulting p values were corrected for multiplicity 1441 with the Benjamini-Hochberg method (n = 3 tests per outcome measure). PND30: 1442 C3ar1+/+ n = 32 (15 males, 17 females); C3ar1-/- n = 32 (18 males, 14 females); 1443 PND90: C3ar1+/+ n = 33 (16 males, 17 females) and C3ar1-/- n = 32 (17 males, 15 1444 females); change: C3ar1+/+ n = 30 (14 males and 16 females); C3ar1-/- n = 31 (17 1445 males and 14 females). Data are shown as mean ± 95% CI. (g-i) Two-sided one-1446 sample t tests (difference from 0) for global connectivity changes at each sparsity level 1447 with genotypes combined, corrected using the Benjamini-Hochberg procedure (n = 1448 10 sparsity levels), # = adjusted p value < 0.05. 1449 1450 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 26, 2025. ; https://doi.org/10.1101/2025.04.24.650541doi: bioRxiv preprint 49 1451 Figure 5 | C3ar1-deficiency has no detectable effects on functional brain 1452 networks. (a) Thresholded adjacency matrices comparing C3ar1+/+ vs C3ar1-/- 1453 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 26, 2025. ; https://doi.org/10.1101/2025.04.24.650541doi: bioRxiv preprint 50 groups where |t| ≥ 2 in adolescence and adulthood, and the change between time 1454 points. No connections with |t| ≥ 2 remained significant after the false discovery rate 1455 (FDR) correction. Additionally, the number of connections in the |t| ≥ 2 adjacency 1456 matrix components were not significant when assessed using network-based statistics 1457 (NBS) correction. (b) Nodal strength (mean absolute connectivity) of a priori 1458 anxiety-related seeds (e.g. L cingulate cortex correlation coefficients with all other 1459 regions). P values were calculated using a mixed ANOVA with between-subjects 1460 factor of genotype and within-subjects factor of region. Only the region effect was 1461 significant at both time-points (both p values < 0.001, ηp2 = 0.69 in adolescence and 1462 ηp2 = 0.73 in adulthood). (c) Mean functional connectivity (FC) and global efficiency 1463 (GE). Student’s t tests within a metric (n = 6) corrected for multiple comparisons 1464 using the Benjamini-Hochberg method. DMN = default mode network, SAL = 1465 salience network, BH = Benjamini-Hochberg. (d) Examples of seed-based FC maps 1466 showing voxel-wise group differences between C3ar1-/- and C3ar1+/+ mice (using t 1467 tests) with seeds placed in the left ventral hippocampus and left prefrontal cortex in 1468 adolescence and adulthood datasets. The dual scale bar displays contrast value on the 1469 x-axis and threshold free cluster enhancement (TFCE) p-values (transformed 0.81-p) 1470 on the y-axis. The transformed p-values have been re-scaled to range from 0 to 1 for 1471 visualisation, with darker colours representing greater statistical significance. Black 1472 outlines demarcate regions where TFCE p-values are below 0.05. 1473 1474 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 26, 2025. ; https://doi.org/10.1101/2025.04.24.650541doi: bioRxiv preprint 51 1475 Figure 6 | C3ar1 deficiency does not cause behavioural abnormalities. (a) Z-1476 scored (normalised to C3ar1+/+ control mean = 0) locomotion metrics for C3ar1-/- 1477 animals from EPM and OF tests, showing results from two independent study 1478 cohorts. Two-way ANOVA for longitudinal study (4 groups, males and females 1479 separated) or Student’s t tests for the adulthood-only study cohort (2 groups, males 1480 and females combined due to smaller sample size) where parametric assumptions 1481 were met, otherwise the Kruskal Wallis test (with Dunn in the longitudinal study 1482 cohort); uncorrected for multiplicity, all ns. (b) Same as (a) but for anxiety-like 1483 metrics from EPM and OF tests; p values all ns. (c) Prepulse inhibition in the 1484 adulthood-only study cohort. PPI increased with increasing prepulse intensity in 1485 both, C3ar1+/+ vs C3ar1-/- mice (paired t tests p value ### < 0.001). Dotted line at 0 1486 indicates inhibition threshold. A score above 0 indicates inhibition (shaded area). 1487 Slope plot shows means and 95% confidence intervals. Individual mice are plotted 1488 three times at increasing prepulse intensity. Prepulse inhibition did not differ by 1489 genotype at any prepulse intensity (uncorrected Student’s t tests at 3 dB, 6 dB and 12 1490 dB, t[37] = 1.58, 1.99, 0.28, p = 0.12, 0.54, 0.17). (d) Novel object recognition (NOR) 1491 recall 1 hr after acquisition in the adulthood-only study cohort. Both groups showed 1492 novelty preference (one-sample t test, value > 50/chance, # = p < 0.05, ## = p < 0.01 1493 ### = p < 0.001). There were no differences between groups (Student’s t test t[37] = 1494 0.42, p = 0.67). The dotted line marks the chance threshold. (a-d) Adulthood-only 1495 study cohort C3ar1-/- n = 20, C3ar1+/+ n = 19 (males and females combined), 1496 longitudinal study cohort: C3ar1-/- n = 33 (17 male, 16 female), C3ar1+/+ n = 33 (16 1497 male, 17 female). Data are expressed as mean ± 95% CI. 1498 1499 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 26, 2025. ; https://doi.org/10.1101/2025.04.24.650541doi: bioRxiv preprint 52 Tables 1500 Table 1 | Mean network connectivity metrics ± 95% CI in males and females 1501 of both genotypes 1502 Age Adolescence Adulthood C3ar1 status +/+ -/- +/+ -/- +/+ -/- +/+ -/- Sex F F M M F F M M Connectivity metric (AUC ± 95% CI) FC 25.5 ± 2 27.5 ± 3.1 25.9 ± 5 29.5 ± 4.5 27.6 ± 3.3 28.9 ± 2.9 29.9 ± 5.3 31.9 ± 4.7 GE 11.5 ± 0.8 12.7 ± 1.2 11.4 ± 1.5 13.2 ± 1.7 12.17 ± 1.2 12.9 ± 1.3 13.3 ± 2 14.5 ± 2.1 CC 13.5 ± 1.1 14.9 ± 1.9 13.9 ± 2.8 15.9 ± 2.6 14.8 ± 2 15.8 ± 1.9 16.3 ± 3.2 17.4 ± 2.6 1503 Table 2 | Mean anxiety-associated node connectivity in males and females of 1504 both genotypes. 1505 Age Adolescence Adulthood C3ar1 status +/+ -/- +/+ -/- +/+ -/- +/+ -/- Sex F F M M F F M M Brain area (mean node connectivity strength) L cingulate cx 0.39 0.41 0.39 0.44 0.45 0.49 0.48 0.54 R cingulate cx 0.38 0.41 0.38 0.43 0.45 0.49 0.48 0.53 L prefrontal cx 0.29 0.33 0.32 0.37 0.41 0.44 0.37 0.49 R prefrontal cx 0.28 0.28 0.33 0.37 0.41 0.45 0.39 0.47 L amygdala 0.14 0.17 0.15 0.2 0.2 0.2 0.24 0.28 R amygdala 0.11 0.16 0.15 0.18 0.19 0.2 0.25 0.25 L pallidum & accumbens 0.23 0.27 0.23 0.31 0.3 0.29 0.35 0.4 R pallidum & accumbens 0.24 0.26 0.24 0.3 0.31 0.31 0.35 0.39 L striatum 0.28 0.32 0.3 0.36 0.32 0.36 0.4 0.43 R striatum 0.26 0.31 0.32 0.35 0.33 0.36 0.42 0.41 L hypothalamus 0.28 0.29 0.27 0.34 0.32 0.32 0.34 0.38 R hypothalamus 0.25 0.27 0.26 0.32 0.31 0.32 0.33 0.37 L dorsal hippocampus 0.4 0.43 0.42 0.49 0.44 0.46 0.48 0.52 R dorsal hippocampus 0.41 0.43 0.42 0.49 0.44 0.47 0.5 0.52 L ventral hippocampus 0.18 0.18 0.17 0.26 0.21 0.23 0.2 0.27 R ventral hippocampus 0.18 0.23 0.19 0.25 0.22 0.25 0.25 0.26 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 26, 2025. ; https://doi.org/10.1101/2025.04.24.650541doi: bioRxiv preprint 53 L PAG 0.28 0.3 0.3 0.38 0.27 0.3 0.32 0.36 R PAG 0.28 0.31 0.3 0.39 0.27 0.29 0.32 0.37 L brain stem 0.11 0.11 0.12 0.18 0.1 0.14 0.16 0.16 R brain stem 0.11 0.09 0.12 0.18 0.11 0.13 0.17 0.18 1506 Table 3 | Mean behavioural outcome measures ± 95% CI in males and females 1507 of both genotypes 1508 Cohort Cohort 1 Cohort 2 Sex F M F M C3ar1 status +/+ -/- +/+ -/- +/+ -/- +/+ -/- Locomotion-related OF locomotion speed (mm/s) 79.65± 3.47 79.48± 3.91 75.61± 4.31 74.72± 4.08 94.21± 6.64 84.33± 7.26 84.93± 7.55 84.93± 5.68 OF locomotion duration (s) 213.38 ±22.53 223.73 ±26.85 216.60 ±21.94 217.97 ±20.82 315.66 ±50.93 264.33 ±49.90 297.85 ±39.56 283.61 ±17.86 OF locomotion distance (cm) 1777.8 9±210. 25 1847.7 1±271. 41 1746.4 5±257. 47 1747.5 2±225. 53 2984.9 9±589. 73 2231.4 5±473. 78 2562.3 6±515. 37 2412.0 3±268. 13 EPM speed (mm/s) 67.07± 3.43 65.89± 2.40 67.71± 3.74 69.90± 2.60 75.27± 5.08 75.38± 6.40 73.61± 4.21 76.63± 3.76 EPM locomotion duration (s) 69.25± 15.36 61.22± 11.89 73.29± 13.14 67.01± 7.48 78.93± 13.23 82.53± 18.41 71.11± 17.69 62.51± 8.88 EPM locomotion distance (cm) 473.32 ±124.8 7 401.52 ±88.01 502.39 ±109.2 2 465.16 ±54.60 591.26 ±122.2 0 620.88 ±155.6 9 518.89 ±140.0 6 472.57 ±68.28 Anxiety-related OF duration core (s) 25.92± 12.12 18.74± 4.60 19.81± 5.35 22.99± 7.73 25.06± 10.65 29.96± 24.36 32.19± 18.93 16.71± 3.87 OF duration centre (s) 106.94 ±18.40 102.01 ±16.36 97.18± 13.68 109.40 ±17.99 105.42 ±40.78 120.94 ±39.65 133.80 ±31.20 95.62± 13.10 OF duration periphery (s) 493.13 ±18.41 498.07 ±16.36 502.90 ±13.68 490.66 ±17.98 493.16 ±42.58 478.89 ±39.60 466.11 ±31.25 504.15 ±13.10 EPM duration open (s) 29.52± 11.52 28.92± 12.57 28.06± 12.03 28.19± 14.72 18.87± 8.77 20.45± 16.56 15.13± 7.96 12.15± 7.29 OF latency core (s) 30.66± 17.23 44.47± 34.34 29.89± 16.18 40.95± 19.45 34.19± 36.91 24.15± 20.50 25.86± 25.45 29.44± 52.19 OF latency centre (s) 10.02± 3.52 13.26± 8.07 12.35± 7.40 11.97± 4.95 5.79±6 .00 7.45±7 .79 5.22±5 .01 5.28±4 .26 EPM duration middle (s) 109.53 ±17.89 102.17 ±19.15 112.33 ±24.24 122.79 ±17.27 85.38± 21.11 72.59± 29.02 71.13± 14.54 74.93± 17.85 EPM duration closed (s) 176.11 ±22.18 198.69 ±19.25 188.78 ±24.17 178.29 ±17.22 214.24 ±21.04 207.31 ±38.82 227.88 ±14.86 224.17 ±17.79 EPM bouts open (#) 4.65±1 .92 3.19±1 .66 3.44±1 .33 2.94±1 .18 2.62±1 .48 3.43±2 .77 2.30±1 .12 1.91±1 .49 EPM head dips (#) 21.53± 6.20 18.12± 5.17 19.62± 6.29 22.31± 5.36 16.75± 4.55 17.57± 13.16 11.90± 3.14 16.73± 8.52 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 26, 2025. ; https://doi.org/10.1101/2025.04.24.650541doi: bioRxiv preprint 54 EPM stretch- attend postures (#) 19.00± 4.15 21.81± 2.98 23.62± 3.14 23.38± 2.87 16.62± 6.16 9.43±6 .04 13.50± 2.68 14.73± 4.10 EPM latency open (s) 65.31± 27.63 106.76 ±53.38 92.95± 48.81 117.27 ±51.75 75.66± 81.38 83.26± 103.55 95.40± 71.29 137.52 ±70.66 Other PPI 3 dB (%) 21.11± 13.2 15.31± 6.40 22.87± 7.41 17.14± 7.70 PPI 6 dB (%) 40.63± 10.16 26.36± 8.74 39.49± 11.24 34.37± 10.65 PPI 12 dB (%) 56.81± 13.18 49.81± 13.28 57.07± 9.51 60.40± 8.38 Acoustic startle response at 120 dB (A.U.) 876.35 ±432.3 6 1068.3 2±284. 31 1472.9 9±360. 84 1908.7 9±655. 66 NOR recognition index (%) 62.49± 11.20 62.60± 9.77 65.02± 7.66 68.16± 8.87 NOR total exploration training (s) 39.45± 15.62 38.58± 10.66 38.50± 12.16 41.12± 11.07 NOR total exploration test (s) 31.68± 10.07 32.41± 6.83 30.81± 9.17 33.58± 5.39 1509 .CC-BY 4.0 International licensemade available under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is The copyright holder for this preprintthis version posted April 26, 2025. ; https://doi.org/10.1101/2025.04.24.650541doi: bioRxiv preprint