Complement component 4A protein levels are negatively related to frontal volumes in patients with schizophrenia spectrum disorders

.


Introduction
Schizophrenia-spectrum disorders (SSD) are considered developmental disorders, with diagnosis typically at the debut of a first psychotic episode during (early) adulthood.Due to the considerable heterogeneity in both clinical and etiological factors in SSD patients, identifying the exact underlying mechanisms remains a challenge.The >100 associated risk genes suggest multiple underlying mechanisms that vary considerably between affected individuals.Yet, in accordance with the GWAS data of the psychiatric genomics consortium, the complement system is being increasingly studied as a promising pathophysiological factor in SSD.
The major histocompatibility complex (MHC) takes a prominent place in large genetic studies as the genetic locus most strongly associated with schizophrenia risk (Schizophrenia Working Group of the Psychiatric Genomics Consortium, 2014).The MHC locus contains several complement related genes, such as the complement component 4 (C4) gene isotypes C4A and C4B.The complement system is the first line of defence upon infections and is involved in opsonisation (i.e.tagging pathogens, cells and synapses for elimination), cell lysis and the promotion of inflammation (Hogenaar and van Bokhoven, 2021;Woo et al., 2020).C4A protein is associated with binding to immune complexes (i.e.antigen-bound antibody), and reduced levels skew towards development of immune complex mediated disease (Traustadottir et al., 2002).C4B protein, in contrast, is associated with binding to pathogens and deficiencies may increase vulnerability to bacterial infections (Jaatinen et al., 2003).The level of C4A gene expression can be predicted based on gene copy number and structural genetic variant (i.e.short or long variants) (Sekar et al., 2016).Structural variations in alleles of the C4 gene were shown to result in increased C4A gene-expression in the brain which may drive the genetic association between the MHC locus and increased schizophrenia risk (Sekar et al., 2016).Predicted C4A expression also showed a positive relationship with positive symptom severity (Melbourne et al., 2018) and cognitive symptoms (Donohoe et al., 2018;Holland et al., 2020) in SSD.Post-mortem studies further corroborate the pathological role of complement, as brain tissue of individuals with SSD (n = 196) displayed an over-expression of total C4 in the stiatum, frontal and parietal cortex relative to healthy controls (HC) (Rey et al., 2020).
It is hypothesized that the role of complement in reduced synaptic spine density is mediated via microglia (i.e.resident immune-cells of the brain).Indeed, in an in vitro cell model with microglia-like cells and neurons of individuals with SSD, the C4A risk gene was associated with excessive synaptic pruning (Sellgren et al., 2019).Furthermore, recent preclinical studies reported that C4 over-expression in mouse brains results in increased microglial engulfement, decreased synaptic spine density, as well as the neurobiological (e.g.NMDA receptor hypofunction) and behavioural phenotypes (e.g.reduced working memory performance and social interaction) similar to SSD (Druart et al., 2021;Yilmaz et al., 2021).The most prominent neurobiological phenotypes of SSD include reduced brain volumes and cortical thinning (van Erp et al., 2018).These alterations precede the onset of psychosis and treatment with antipsychotics (Cannon et al., 2015;Ding et al., 2019).The reduction of volume and cortical thickness is associated with symptom severity, cognitive functioning, and disease outcome in SSD (Oomen et al., 2022;Walton et al., 2018).
The reduced brain volumes, cortical thinning, as well as cognitive symptoms in SSD may be linked to abnormal complement-mediated microglial synaptic pruning occurring during critical developmental periods (Germann et al., 2021;Schalbetter et al., 2022).The most recent genome wide association study also confirmed that schizophrenia was most strongly associated with genes that are involved in synaptic organization, differentiation and transmission (Trubetskoy et al., 2022).Other recent studies demonstrated significantly decreased levels of synaptic vesicle glycoprotein 2a (SV2a), which is indicative of lower synaptic density, in the frontal and anterior cingulate cortex of individuals with SSD compared to HC (Onwordi et al., 2020;Radhakrishnan et al., 2021).Reduced SV2a in frontal regions was also associated with positive symptoms and cognition in SSD (Radhakrishnan et al., 2021).Furthermore, genetic predisposition for high C4A expression has been shown to be related to high levels of a marker for microglial activity across the schizophrenia spectrum (n = 111) (Da Silva et al., 2021).These findings could be interpreted as excessive synaptic pruning in SSD as a result of abnormal microglia activity, which specifically affects the (frontal) cortex, but not the subcortical regions (Berdenis van Berlekom et al., 2019).Excessive synaptic pruning results in low synaptic density, which is considered to be reflected by reduced brain volumes and cortical thickness (Howes et al., 2023).
Only a handful of studies related C4A expression to brain morphology and cognition in HC (O'Connell et al., 2021) and SSD (Holland et al., 2020;Ji et al., 2021).In general, these studies conclude that high C4A expression and C4 protein levels are related to reduced cortical thickness and cognitive performance in HC and SSD (Ji et al., 2021;O'Connell et al., 2021).These prior studies were mainly performed in combined groups with HC and affected individuals (Da Silva et al., 2021;Holland et al., 2020;Ji et al., 2021), which obscures differentiation of the effects of C4 in each group.In addition, most studies were based on genetic predisposition for high C4A expression.However, environmental factors (e.g.infection, microbiome dysbiosis, trauma, obesity, smoking) influence the expression of complement genes and subsequent serum protein levels (Germann et al., 2021;Kim et al., 2021;Nimgaonkar et al., 2017;Severance et al., 2021).Studies that include total C4 protein levels (i.e.combined C4A and C4B) are unable to assess the specific effects of the C4A-gene product (i.e.C4A protein).
Using isotype-specific enzyme-linked immunosorbent assays (ELI-SAs) (Wouters et al., 2009), we were able to specifically assess C4A protein levels in serum in SSD and HC.The current study aims to investigate the association of C4A serum protein levels with brain morphology (i.e.cortical thickness and brain volumes) and cognition, and specifically whether this association differs between SSD (n = 69) and HC (n = 40).Results from previous studies suggest that elevated C4A levels are reflective of SSD pathology and therefore we hypothesize that C4A levels will be negatively related to the main outcomes (overall cortical thickness, brain volumes and cognition), while this association will be less strong or even absent in HC.Follow-up analyses on individual brain regions and cognitive domains may show effects in the same direction.We hypothesized the strongest effects in the frontal cortex, as that region is associated with both C4 over-expression and lower synaptic density.

Participants
This study sample and data are part of the baseline measurements of a completed clinical trial (ClinicalTrails.gov:NCT01999309).Patients aged between 18 and 50 and diagnosed with schizophrenia, schizoaffective, schizophreniform disorder (295.x) or psychotic disorder not otherwise specified (298.9)within 3 years prior to screening were included.HCs had no prior or family history with psychiatric disorders.Participants without C4A protein levels or brain volumetric measurements were excluded from the analyses, which resulted in a dataset of 69 SSD patients and 40 HCs.
Table 1 shows demographic and clinical characteristics separately for SSD (n = 69) and HC (n = 40).Antipsychotic medication use at time of assessment was attained from patient report.Dose of antipsychotic medication intake (mg/day) was converted into an antipsychotic dose equivalent (i.e.chlorpromazine equivalent) for each patient (Gardner  , 2010).Educational attainment was provided in years of education, both from participants and the mean of both parents.
Cognitive performance was assessed with the Brief Assessment of Cognition in Schizophrenia (BACS).The BACS is comprised of six subtasks, which measure several cognitive domains most affected in schizophrenia: list learning (verbal memory), digit sequencing task (working memory), token motor task (motor speed), category instances and controlled oral word association test (verbal fluency), symbol coding (attention & information processing speed) and the tower of London (executive function).The scores of these subtasks were standardized and corrected for age and gender according to procedures described by Keefe et al. (Keefe et al., 2004).Standardized scores from all subtasks are averaged to calculate the composite BACS score, reflecting overall cognitive performance.

C4A protein measurements
Blood was drawn in the morning of the baseline measurements of the clinical trial (Begemann et al., 2015).Serum was prepared and stored at -80 • C in aliquots within 4 h after blood draw by the Central Biobank of the University Medical Center Utrecht.C4A levels were determined by Sanquin Diagnostic Services (Amsterdam, the Netherlands) using previously described isotype-specific enzyme-linked immunosorbent assays (Wouters et al., 2009).In short, the micotiterplates are coated with an antibody specific for C4A (clone C4-4; (Wouters et al., 2005)).Samples and controls are analyzed in three dilutions in duplo.Bound C4A is detected with a biotinylated sheep-anti-human C4c polyclonal antibody.Biotin is detected by streptavidin-HRP and the assay is developed with tetramethylbenzidine. Concentrations are determined using a standard line with known concentration and log-log transformation (Wouters et al., 2009).The quantified C4A concentrations (in mg/L) were logtransformed prior to further analyses.

MRI acquisition and processing
MRI data processing was described in a previous publication (Oomen et al., 2022).T1-weighted scans were acquired in the UMC Utrecht, using a 3 T Philips Ingenia CX and a 32-channel SENSE head-coil (3D T1weighted Turbo Field Echo sequence; repetition time = 10 ms, echotime = 4.6 ms, flip angle = 8 • , reconstructed voxel size = 0.75 × 0.75 × 0.8 mm 3 , field of view = 240 mm × 240 mm × 160 mm).Brain volume and cortical thickness were calculated with the standard processing pipeline of FreeSurfer (version 6.0.1;http://surfer.nmr.mgh.harvard.edu/).T1 images were visually checked for errors in segmentation.We regarded the Euler number as a quantitative proxy of structural image quality.The Euler number reflects the topological complexity of the reconstructed cortical surface and is highly related to manual ratings of image quality (Rosen et al., 2018).FreeSurfer calculates the Euler number per hemisphere as part of the standard processing pipeline.The average Euler number of both hemispheres was used as a covariate for image quality in further analyses.Regions of interest (ROIs) included all regions from the Desikan-Killiany cortical atlas (Desikan et al., 2006) and FreeSurfer's subcortical atlas (Fischl et al., 2002).Brain volumes and cortical thickness of these areas were taken as a reflection of low synaptic density, since synaptic density is considered to be reflected more by brain volumes and cortical thickness, rather than cortical surface area or gyrification (Howes et al., 2023).

Statistical analyses
The difference in C4A levels between groups was analyzed using ANCOVA.This group comparison as well as all subsequent analyses included the following covariates: age, sex, BMI, and antipsychotic dose equivalent, as these factors have been previously related to C4A (Jaros et al., 2015;Kamitaki et al., 2020;Mohd Asyraf et al., 2022;Purves-Tyson et al., 2020;Ritchie et al., 2004).
Regression analyses included total gray matter volume, mean cortical thickness and BACS composite scores as main outcomes.Regression analyses on these three main outcome variables were performed including the covariates, C4A level, group (SSD vs. HC), and C4A*Group interaction terms.The interaction term was included to assess whether the relation between C4A and the outcomes differs between the groups (i.e.SSD vs. HC).Without a significant interaction, the main effect of C4A for both groups combined was reported.The regression analyses with the three main outcomes were evaluated at a significance threshold of α = 0.05.
For follow-up analyses, every ROI volume (n = 103), ROI thickness (n = 70), and score per cognitive task (6 BACS sub-scores), a separate regression analysis was performed with the same independent variables and covariates as the main analyses.Covariates for regression analysis on brain morphometry included handedness and mean Euler number.In addition, total intracranial volume was included as covariate in all volumetric analyses and mean cortical thickness was included as covariate in all thickness analyses.Analyses with cognitive outcomes included years of education as covariate.All follow-up regression analyses on subscores or ROIs were evaluated at a Bonferroni corrected significance threshold (α = 4.9 × 10 -4 , 7.1 × 10 -4 and 0.007, for brain volume, cortical thickness, and cognition, respectively).
The follow-up analyses resulted in effect sizes of the association between C4A and brain volumes (Fig. 3) and cortical thickness (Fig. 4), for HC (upper row), SSD (middle row) and both groups combined (lower row).It is interesting to note that C4A is mainly positively associated with brain volumes in HC (blue), but negatively associated with brain volumes in SSD (red).However, these effects did not reach statistical significance.Detailed statistical results for each ROI are presented in Supplementary Table 1 and 2.

Cognition
For the composite BACS score, there was no significant C4A *Group interaction (adj.R 2 = 0.33, β = -0.02,p = 0.93) or C4A main effect (adj.R 2 = 0.30, β = 0.03, p = 0.73).No significant C4A *Group interaction effect on executive functioning was found (adj.R 2 = 0.12, β = 0.04, p = 0.85).However, a positive C4A main effect on executive functioning in both groups combined was found (adj.R 2 = 0.14, β = 0.26, p = 0.006), yet the effect was not statistically significant considering bonferronicorrected thresholds.Remaining individual cognitive subdomains assessed with the BACS did also not demonstrate significant interaction (p > 0.15) or C4A main effects (p > 0.28) (Supplementary Table 3).C4A does not moderate the relationship between relevant brain structures (i.e. total brain volume, mean cortical thickness, l-mOFC and the left frontal pole volume) and cognition, as three-way interactions between C4A, group and the brain structures were not statistically significant (p > 0.43).When taking both groups together, C4A also did not significantly moderate the relationship between the brain structures and composite BACS score (C4A*brain structure interaction; p > 0.24).

Discussion
The present study investigated the association between C4A serum protein levels, brain morphology (i.e.cortical thickness and brain volumes) and cognition as well as whether this association differs between SSD (n = 69) and HC (n = 40).First, there was no significant difference in C4A serum protein levels between the groups when controlling for age, sex and BMI.Next, C4A was not significantly related to cognition, total gray matter volume or cortical thickness.In ROI analyses, C4A levels showed a negative relation with mOFC and frontal pole volumes in SSD, but a positive relation in HC.These results suggest that peripheral levels of C4A may have a differential association with brain morphology in SSD compared to HC.

C4A group differences
The current study did not demonstrate a significant difference in C4A protein levels between SSD and HC participants.This finding is in accordance with a recent meta-analysis which reported no elevated total C4 protein levels in SSD compared to HC (Mongan et al., 2020).The authors argue that this may potentially result from a lack of adequate control for confounding variables and a wide range of methods with insufficient sensitivity for the assessment of complement protein levels.The current investigation, however, accounted for these potential shortcomings by using methods specifically sensitive to C4A rather than total C4 protein levels, and by including relevant covariates in all analyses.Antipsychotics were not added as a covariate, since only the SSD group uses them and they did not correlate with C4A levels.Nevertheless, the potential effect that antipsychotics may have on the complement system are hard to ascertain, as studies are scarce and their results are conflicting (Woo et al., 2020).While there is some evidence that antipsychotics influence the complement system (Mongan et al., 2020), future (longitudinal) studies are required to determine the exact effects.Another important consideration is limited statistical power due to the naturally occurring wide range of C4 serum protein levels, as a result of genetic variation of the C4 gene (Yang et al., 2003).
The absence of an overall group difference in C4 proteins may also be due to the heterogeneity of SSD.That is, C4 protein levels may not be elevated among first-episode psychosis FEP, but were significantly increased in chronic patients relative to HC (Laskaris et al., 2019).This may explain the findings of the current study, as we included recentonset participants (mean time since diagnosis= 1.2 ± 0.99y).Furthermore, C4A protein levels were also elevated in a subgroup of SSD with elevated levels of peripheral cytokines (Ji et al., 2021), which further demonstrates increased complement levels in specific subgroups of SSD.Complement levels also showed to be predictive of treatment response (Mondelli et al., 2020;Susai et al., 2023) and transition from clinical high risk to first-episode psychosis (Mongan et al., 2021).Future studies should therefore further explore the clinical utility of complement proteins as biomarkers for transition and SSD disease course.

C4A and frontal volumes
The main finding of this study is the negative association between C4A serum protein levels and frontal volumes (i.e.mOFC and frontal pole) in SSD.According to recent preclinical and clinical studies, aberrant complement activity results in increased microglia-mediated synaptic pruning during critical developmental periods, which may underlie reduced brain volumes in SSD (Da Silva et al., 2021;Druart et al., 2021;Radhakrishnan et al., 2021;Sellgren et al., 2019;Yilmaz et al., 2021).In addition, a composite score of upregulated complement mRNA and complement protein was associated with reduced cortical thickness across SSD and HC (combined n = 86) (Ji et al., 2021).Based on these findings, a negative relation between C4A levels and brain volumes in SSD has been hypothesized.Although Fig. 3 of the current article suggests a negative association between C4A and volumes of regions across the brain in SSD, it only reached statistical significance for the mOFC and frontal pole and not for the total brain volume.
The frontal cortex is particularly affected by both SSD and complement pathology.For instance, post-mortem research in the dorsolateral prefrontal cortex shows C4 protein increases in the earliest stages of life (0-5y), which drops in teenagers (14-18y) and young adults (19-26y) (Sager et al., 2021).This drop of C4 levels also coincides with an increase in complement inhibitors (Ji et al., 2021;Sager et al., 2021).This C4 reduction may be present and possibly necessary for healthy neurodevelopment (Sager et al., 2021), whereas persistent C4 elevation could occur during a later developmental ages in SSD.Indeed, especially frontal regions display a higher rate of gray matter loss in clinical high risk who later convert to psychosis (Cannon et al., 2015).
C4A serum protein levels showed a statistically significant positive association with mOFC and frontal pole volumes in HC.However, O'Connell et al. (2021) reported a negative association between predicted C4A expression and mOFC thickness in undiagnosed UK biobank participants (n > 10.000).Yet, they also reported positive associations between predicted C4A expression and cortical surface area and cortical thickness in one third of their investigated brain regions (O'Connell et al., 2021).The authors suggest that increased gray matter volume in combination with high C4A may be explained by altered cell migration and deficient synaptic pruning (O'Connell et al., 2021).As synaptic pruning in the frontal cortex persist throughout the third decade of life (Petanjek et al., 2011), altered or incomplete synaptic pruning in the young healthy adults in this study would indeed explain high levels of C4A that coincide with relatively large frontal cortical volumes.Alternatively, dependent on the isotype and structural form, C4 has both protective and deleterious effects (Sekar et al., 2016;Woo et al., 2020).In that case, the possible protective effects in HC (from the C4B gene and C4 short(S)-form) may outweigh the deleterious effects (from the C4A gene and the C4 long(L)-form).Support for this comes from findings demonstrating C4B/C4S to be protective of age-related white matter microstructural changes in undiagnosed UK biobank participants (Traylor et al., 2021).Ultimately, as C4B/C4S and C4A/C4L are inversely correlated (Woo et al., 2020), future efforts should be made to investigate the role of either one or the balance of both genotypes in SSD pathology.

C4A and cognition
While previous studies report a negative association between C4 expression and cognitive functioning, the current study did not find a significant relationship.In earlier work, elevated C4 expression was significantly related to poor fluid intelligence, visual memory, and processing speed in >80.000HC (O'Connell et al., 2021) and to poor episodic memory in 141 HC and 503 SSD combined (Donohoe et al., 2018).However, the relation between episodic memory and C4 gene expression (Donohoe et al., 2018) was no longer significant when polygenic risk scores of other complement genes were included (Holland et al., 2020).Of these cognitive domains, only processing speed was part of the current study and demonstrated no associations with C4A levels in HC, SSD or in the groups combined.
Cognitive functioning and brain volumes of the present study sample were previously studied by Oomen et al. (2022).This study reported that 29 % of SSD patients in that sample (n = 86) showed a cognitive profile similar to HCs (Oomen et al., 2022).Especially executive functioning displayed the smallest differences between cognitive subgroups (Oomen et al., 2022).The present study also reported that C4A and executive functioning were positively related (C4A main effect), which did not remain after bonferroni-correction.This positive direction of this association may be a result of the relatively unaffected executive function in the group as a whole, since HC showed a positive relation between C4A and frontal volumes as well.As cognitive performance in SSD shows this heterogeneity, the absence of current significant effects in contrast to previous reports with large study samples may be due to a lack of sufficient statistical power.

Strengths and limitations
The use of C4A protein levels to investigate the relation between the complement system and the brain and cognition is novel.At the same time, this method could have formed a potential limitation since it is unknown to what extend peripheral proteins are a reflection of protein levels in the brain (Mongan et al., 2020) and no CSF was sampled for the current study to investigate this.The majority of studies used genetically predicted C4A levels instead of mRNA expression or protein levels of C4 (Da Silva et al., 2021;Donohoe et al., 2018;O'Connell et al., 2021).It is assumed that these predicted levels are reflective of C4A in the brain.Recently, however, genetically predicted C4A gene expression and C4A protein levels were found to be significantly related as well (Ji et al., 2021), which supports the use of either method.On the other hand, C4 protein in CSF and serum did not show a correlation (Gallego et al., 2021).Future studies on the most optimal, yet feasible, method to probe complement system activation in the brain would therefore be a great addition to the field.
Important to note is that the current as well as previous imaging studies assume reduced brain volume and thickness to be an indirect result of excessive synaptic pruning during the critical developmental period in SSD.Future studies may consider more specific proxies of synaptic pruning, such as molecular PET imaging of SV2a (Onwordi et al., 2020).Unsurprisingly, the contrasting effects of C4A on brain volume in SSD and HC in the present study resulted in no significant findings in the main C4A analyses with both groups combined.A previous study that combined SSD and HC (n = 88) also reported no significant associations between predicted C4A levels and brain morphometry (i.e. total brain volume, total cortical thickness or total surface area) (Holland et al., 2020).Furthermore, lack of statistical power is a limitation that prevented previous studies from stratifying their analyses based on clinical group.Results of the current study, however, demonstrated that the addition of the C4A *group interaction effect could yield interesting results.
As the morphological brain changes occur in the prodrome of psychosis (Cannon et al., 2015) and complement proteins could aid in the prediction of transition to psychosis (Mongan et al., 2021), associating the complement system in relation to brain morphology in SSD patients may not reflect the initial pathophysiological processes.However, as C4A protein levels are related to genotype (Ji et al., 2021), it is expected that SSD patients with relatively high C4A levels during adulthood, also showed this elevation in prodromal stages (Ji et al., 2021;Sager et al., 2021).The negative association between C4A and frontal volumes seen in SSD patients in this study, might thus be a reflection of earlier pathological processes.A strength of this study is that we included a sample with recent onset SSD, therefore the effects found are less obscured by the pathological processes that occur in chronic stages of SSD.Nevertheless, longitudinal studies with clinical high risk and SSD patients are warranted.

Conclusions
This study showed that C4A concentration in serum was negatively related to frontal brain volumes (i.e.mOFC and frontal pole volume) in SSD, but this relation was inverse for HC.These results suggest that peripheral levels of C4A may have a differential association with brain morphology in SSD compared to HC.The negative relation between C4A and frontal volumes support the hypothesis of complement-mediated microglial synaptic pruning as a pathological process in SSD.

Fig. 2 .Fig. 3 .
Fig. 2. Interaction effects of C4A*Group for left medial orbitofrontal cortex (A) and left frontal pole (B).Estimated regression lines are corrected for covariates.The displayed datapoints are uncorrected.HC: healthy controls (blue); SSD: schizophrenia spectrum disorders (orange).(For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Table 1
Demographic and clinical characteristics of the participants.BACS: Brief Assessment of Cognition in Schizophrenia; IQR: interquartile range; PANSS: Positive and Negative Syndrome Scale; SD: standard deviation; W: Wilcoxon test statistic. 1Parental years of education was available for 55 SSD and all HC participants.