Association of Plasma Complement System With Brain Structure Deficits in Bipolar and Major Depressive Disorders


 Objective Inflammation plays a crucial role in the pathogenesis of major depressive disorder (MDD) and bipolar disorder (BD). However, the underlying neurobiological mechanisms are poorly understood. This study aimed to examine whether the dysregulation of complement components contributes to brain structure deficits in BD and MDD patients. Methods A total of 52 BD patients, 35 MDD patients, and 53 mentally healthy controls were recruited from the inpatient and outpatient departments of West China Hospital of Sichuan University. The human complement panel 2-immunology multiplex assay was used to measure the levels of complement C1q, C3, C3b, C4, factor B, factor H, and properdin. Whole brain-based comparison was performed to investigate differences in gray matter volume and cortical thickness among the BD, MDD, and control groups, and relationships were explored between neuroanatomical differences and levels of complement components.Results The gray matter volume in the medial orbital frontal cortex (mOFC) and middle cingulum decreased in both patient groups, while the cortical thickness of the left precentral and left superior frontal gyrus was affected differently. Log10-transformed concentrations of C1q, C4, factor B, factor H, and properdin were higher in both patient groups than in controls, while levels of C1q, factor H, and properdin showed a significant negative correlation with gray matter volume in the mOFC at the voxel-wise level.Conclusion Greater inflammation in mOFC was observed in BD and MDD patients than in controls. Structural deficits in both patient groups were associated with elevated levels of certain complement factors, providing insight into the neuro-inflammatory pathogenesis of mood disorders.

dysregulation of interleukin 6 (IL-6), tumor necrosis factor-α, and C-reactive protein than MDD patients [30,31], but whether the two types of patients also differ in plasma levels of complement factors is unclear.
Rodent and human brain imaging studies have shown that several cortical and subcortical structures in BD and MDD patients, including the amygdala, striatum, insula, ACC, orbitofrontal cortex, and hippocampus/parahippocampus, are particularly sensitive to changes in peripheral in ammation, especially to elevated levels of C-reactive protein and cytokines [32][33][34]. However, only a few studies have examined the association between complement components and brain structure in mood disorders. Signi cantly increased C3 expression was found in the post-mortem prefrontal cortex of depressed suicide subjects as well as in mice with chronic stress-induced depressive-like behavior [26]. In another study, genetic variants of C4, which encodes a protein of the classical pathway, have been associated with schizophrenia development, providing a solid basis for establishing a causal relationship between complement-mediated synaptic pruning and cortical thinning, which is frequently associated with psychiatric illness [35]. Animal studies have also consistently shown that activation of C1q, C4, and C3 contribute to synapse loss in Alzheimer's disease, leading to gray matter loss and brain atrophy [36]. In addition, complement components have been found to enhance pro-in ammatory signaling, leading to increased IL-1β production [26].
Despite previous attempts to determine the role of the complement system in the etiology of mental disorders, its effect on brain structure in BD and MDD remains unexplored. Since brain volume and CT can de ne endophenotypes for BD and MDD [37,38], the relationship between in ammation and disturbed brain structure might provide insights into disease pathophysiology [39]. Therefore, in the present study, we evaluated the association of altered levels of complement components with brain structure de cits in patients with BD and MDD. We hypothesized that the two groups of patients would show common and distinct patterns of GMV and CT alterations, and that both groups would show increased levels of complement factors, especially the BD group. We also explored whether elevated complement factor levels in plasma, an indicator of in ammation, correlated with GMV or CT differences among BD patients, MDD patients and controls. Finally, we examined whether the observed differences in brain structure correlated with cognitive function.

Participants
Patients with BD and MDD were recruited from the inpatient and outpatient departments of West China Hospital of Sichuan University. BD or MDD was diagnosed using the structured clinical interview for DSM-IV Axis-I disorders (patient version) [40]. All subjects, with ages between 16-55 years, were right-handed Han Chinese who were physically and neurologically healthy. Their demographic characteristics were recorded, including disease duration, treatment history, illness stage, age, sex, education level, and body mass index (BMI). Manic and depressive symptoms were assessed using the Young Mania Rating Scale [41] and the Hamilton Rating Scale for Depression [42]. Patients with concurrent neurological illness, mental retardation, cardiovascular disease, schizophrenia, anxiety disorder, alcohol or drug abuse, and history of loss of consciousness were excluded from the study. None of the patients had received electroconvulsive therapy before participation.
For comparison, demographically matched healthy controls (HCs) were recruited from the local community through advertising. HCs were included if they (1) were 16-55 years old; (2) did not meet the diagnostic criteria of the structured clinical interview for DSM-I (non-patient edition) [43]; (3) had no current or past signi cant medical or physical disease, such as diabetes, thyroid diseases, hypercholesteremia, liver diseases, epilepsy, stroke, or systemic lupus erythematosus; and (4) had no history of psychiatric illness in rst-degree relatives. All participants provided written informed consent after the study procedures were explained. The study was approved by the Institutional Review Board of Sichuan University.

Cognitive function measurements
Intelligence quotient (IQ), verbal IQ, and performance IQ scores of all participants were assessed using the seven-subtest short form (information, arithmetic, digital symbol, digital span test, block design, picture completion, and similarities) of the revised Wechsler Adult Intelligence Scale in Chinese [44]. The Logical Memory Test is a standard test used for assessing immediate and delayed recall memory function [45]. The Trail Making Test A/B [46] was used to measure attention, visual screening, and processing speed. The digital symbol substitution test was used to measure working memory, visuospatial processing, attention, and processing speed [44] (see the Supplementary Materials and the Methods section for more details).

Measurement of complement component levels
Blood samples were collected by venipuncture between 4.00 p.m. and 4.30 p.m. using ethylenediaminetetraacetic acid as an anti-coagulant. Peripheral blood mononuclear cells were removed by refrigerated centrifugation at 1,000 g for 10 min, and the separated plasma was immediately divided into 0.5-mL aliquots and stored at − 80°C. Plasma factors were evaluated using MILLIPLEX ® MAP kits (Merck KGaA, Darmstadt, Germany). The human complement panel 2-immunology multiplex assay (HCMP2MAG-19K, Merck Millipore, Billerica, MA, USA) was used to measure the levels of C1q, C3, C3b/iC3b, C4, factor B, properdin, and factor H, following the manufacturer's instructions. Data were analyzed using a FLEXMAP 3D ® instrument (Luminex, Merck Millipore, Billerica, MA, US) operated with xPONENT ® software (version 4.0, Luminex). Standard curves were generated using standard samples in the multiplex assay kit. Median uorescent intensity data were analyzed using a weighted 5-parameter logistic method to calculate the factor concentrations.

Image acquisition
Magnetic resonance imaging (MRI) was performed on a 3.0-T MR scanner (Achieva; Philips, Amsterdam, The Netherlands) using an eightchannel phased-array head coil. Foam padding and earplugs were used to minimize head movement and scanner noise. During scanning, participants were reminded to remain still with their eyes closed, without falling asleep or thinking of anything particular, which subjects con rmed immediately after MRI. High-resolution T1 images were acquired by 3D magnetization-prepared rapid gradient-echo sequence (repetition time, 8.37 ms; echo time, 3.88 ms; ip angle, 7°; in-plane matrix resolution, 256 × 256; eld of view, 24 × 24 cm 2 ; number of slices 188). During MRI, none of the participants showed more than 2 mm displacement in the x, y, or z direction or more than 2° angular motion.

VBM and SBM
VBM and SBM analyses were conducted using the Computational Anatomy Toolbox 12 (CAT12, version 12.5, Jena University Hospital, Jena, Germany, http://dbm.neuro.uni-jena.de/cat/) [47], an extension of Statistical Parametric Mapping (University College London, London, UK, http://www. l.ion.ucl.ac.uk/spm/software/spm12/), following the manufacturer's instructions (http://dbm.neuro.uni-jena.de/cat12/CAT12-Manual.pdf). The T1 MRI scans of all subjects were registered to the Montreal Neurological Institute space. The whole-brain structural data were then segmented into white matter, gray matter, and cerebrospinal uid, and bias was corrected to remove intensity non-uniformities. The images of segmented gray matter were used to assess the number of volume changes based on spatial registration, while the modulated images of gray matter could re ect the tissue volumes for VBM analysis. The total intracranial volume (TIV) of each subject was calculated and used as a covariate in further statistical analyses. The normalized gray matter images were smoothed using a Gaussian lter with a fullwidth half-maximum (FWHM) of 8 mm. The CT of the left and right hemispheres was automatically estimated by CAT12 based on the projection-based thickness method, and the corresponding images were smoothed with a 15-mm FWHM Gaussian kernel [48].

Statistical analysis
Statistical analyses were carried out using SPSS 24.0 (IBM, Chicago, IL, USA). Differences in demographic and clinical data were assessed using one-way analysis of variance for continuous variables and Fisher's chi-squared test for categorical variables. Data normality was evaluated by visual inspection and the Kolmogorov-Smirnov test. The concentrations of all complement components were log 10 -transformed prior to analysis in order to normalize their distributions. Complement factors and cognitive function were compared using analysis of covariance (ANCOVA), while controlling for age, sex, education level, and BMI. Clinical parameters between the two patient groups were compared using the two-samples t test. To explore correlations of complement components with cognitive function or other clinical parameters, partial correlation analysis was conducted after controlling for demographic parameters. Differences associated with P < 0.05 were considered statistically signi cant.

Image analysis
Voxel-wise GMV and vertex-wise CT differences among the three subject groups were investigated using ANCOVA, while co-varying for age, sex, education level, and BMI. For GMV analysis, all voxels with GMV probability value < 0.1 (absolute threshold; range, 0-1) were excluded to avoid possible edge effects around the margin between different tissue types. TIV was also controlled as a covariate in GMV analysis. While controlling for age, sex, education level, and BMI, statistical maps were created to identify potential linear correlations between plasma levels of complement components and whole-brain GMV or CT in BD patients, MDD patients and HCs considered separately as three groups or together as one large group [49]. The threshold was set at P < 0.05 after correcting for family-wise error rate at the cluster level with conservative voxelwise P < 0.0005. To further analyze the relationship between GMV or CT and other variables, GMV and CT values were extracted, respectively, from the anatomical automatic labeling atlas or the Desikan-Killiany (DK40) cortical atlas, then subjected to partial correlation analysis using SPSS 24.0. The extracted values were used to identify regions that differed signi cantly between patients and controls or between different types of patients, as well as to identify neuroanatomical alterations that correlated signi cantly with levels of complement components.

Demographic and clinical characteristics of patients and HCs
A total of 140 subjects, including 52 BD patients, 35 MDD patients, and 53 age-and sex-matched controls were enrolled in the study (Table 1). Across the three groups, there were no signi cant differences in sex, or smoking status. Moreover, there were no signi cant differences between BD and MDD patients in age at onset, total illness duration, number of depressive episodes, general assessment functioning, or anti-depressant usage. Both patient groups had a lower education level than HCs, while BD patients had a higher BMI than MDD patients and HCs. Although BD patients had lower Hamilton depression total scores than the MDD group, they were treated more often with mood stabilizers and/or

Complement component levels
The log 10 -transformed concentrations of all complement components were higher in the BD group than in HCs. The same trend was also observed in the MDD group, except for C3. Levels of C3, C4, and factor H were higher in patients with BD than in those with MDD, after covarying for age, sex, education level, and BMI. (Table 1). No other signi cant correlations were found between complement component levels and clinical parameters in either patient group (Table S1). C3b was excluded from the analysis, as it was not detected in more than half of the samples.

GMV and CT changes in BD and MDD patients
Voxel-based comparison of GMV in the mOFC, middle cingulum, and right precentral gyrus using age, sex, education years, and intracranial volume as covariates indicated signi cant differences among the three groups (Table 2 and Fig. 1a). Post-hoc analysis indicated common brain volume de cits in the mOFC and middle cingulum in the two patient groups (Table 2 and Figs. 1b-c), and MDD patients additionally showed lower GMV in the left precentral gyrus than HCs (Table 2 and Fig. 1c). 2. Age, sex and education years were co-variated out in gray matter volume and cortical thickness analysis. The intracranial volume was additionally controlled in gray matter volume analysis. Voxel-wise threshold was set at p < 0.0005. Signi cance level was set at p < 0.05, cluster wise family wise error corrected. *p < 0.05, **p < 0.01, ***p < 0.001. Bold represents statistically signi cant results.
Vertex-based comparisons of CT in the left precentral gyrus and left SFG revealed signi cant differences among the three groups after covarying for age, sex, and education level ( Table 2 and Fig. 1d). CT of the left precentral gyrus was signi cantly smaller in the MDD group than in HCs, while CT in the left SFG was signi cantly smaller in the BD group than in HCs (Table 2 and Fig. 1d). However, no signi cant differences were observed in GMV or CT between BD and MDD patients after controlling for individual variations in intracranial volume (for GMV), age, sex, and education level.
Correlation of complement component levels in plasma with whole-brain GMV or CT Whole-brain correlation analysis across all subjects after controlling for age, sex, education level, BMI, and TIV revealed a negative association between GMV in the mOFC and log 10 -transformed C1q, factor H, and properdin (Table 3 and Fig. 2). However, whole-brain correlation analysis in each group did not reveal any signi cant correlation between GMV and plasma levels of complement components. We did not detect any signi cant results for CT among all subjects and in the individual groups after multiple correction analysis (Table S2). Correlation of cognitive function and altered brain structure Across all subjects, the GMV in the left mOFC positively correlated with immediate logical memory recall (r = 0.20, P = 0.023), while the CT of the right precentral gyrus positively correlated with immediate logical memory recall (r = 0.18, P = 0.047) and delayed logical memory recall (r = 0.23, P = 0.010). However, these relationships were not signi cant after correction for multiple comparisons (Table S2). Therefore, we could not identify any speci c correlation between disrupted brain structure and cognitive function.

Discussion
In the present study, we examined the association between complement components and brain structure de cits in BD and MDD patients, and found that the two disorders have common and distinct patterns of structural alterations. Compared to HCs, both patient groups showed reduced GMV in the middle cingulum and mOFC. In addition, MDD patients showed lower GMV in the right precentral gyrus and smaller CT in the left precentral gyrus than HCs, while the BD group showed smaller CT in the left superior frontal gyrus than HCs. Complement component levels were signi cantly higher in both patient groups than in HCs, consistent with previous ndings suggesting that both disorders are associated with hyperin ammatory responses [50,51]. In fact, the log 10 -transformed levels of C3, C4, and factor H were signi cantly higher in BD than in MDD patients, even after controlling for age, sex, years of education, and BMI. These results con rm our hypothesis that BD involves a more severe in ammatory response than MDD. Furthermore, we found that C1q, factor H, and properdin were negatively associated with GMV in the bilateral mOFC across all subjects. To the best of our knowledge, this is the rst HC-controlled study examining relationships between brain structure and complement components in mood disorders.
Our ndings of shared structural alterations in the brains of BD and MDD patients are consistent with the results of a recent meta-analysis, where the volume of the ventromedial prefrontal cortex was found to be lower in both patient groups, suggesting a consistent volume de cit pattern in the two disorders [13]. A previous study also reported signi cantly smaller middle cingulum volume in MDD patients who did not respond to treatment than in HCs and in patients who did respond to treatment [52]. Although no similar reduction has been reported for BD patients, our results suggest that altered middle cingulate gyrus volume may underlie abnormal brain activation in both disorders. Our study also showed that MDD patients have signi cantly smaller GMV in the precentral and postcentral gyri as well as smaller CT in the left precentral gyrus than HCs. As both the precentral and postcentral gyri are key parts of the sensorimotor network [53], our results con rm the potential role of the precentral region in MDD pathogenesis [54].
MRI studies have implicated the SFG, the gyral-based representative of the dorsolateral prefrontal cortex [1], in self-awareness, complex cognitive behavior, executive function and emotional regulation [55]. In the present study, BD patients showed signi cantly smaller CT in the left SFG than HCs did. Our results are consistent with a previous whole-brain meta-analysis showing widespread cortical thinning in the SFG of BD patients, suggesting that structural de cits in the SFG may re ect the main emotional regulatory and processing symptoms of BD [56].
However, we did not nd signi cant differences in GMV or CT between MDD and BD patients, similar to a recent study showing that brain structural data could not differentiate the two types of patients [13].
The BD group of the present study showed higher levels of all complement components than HCs. In addition, BD patients showed higher levels of C3, C4, and factor H than MDD patients. Our results are consistent with a previous report that BD patients have higher levels of proin ammatory cytokines than MDD patients [31], but our results should be veri ed in further studies of complement components in the two disorders. Interestingly, signi cantly increased levels of C3, C4, and factor B have also been detected in the serum plasma of patients with mania [29,57]. All these observations suggest that complement components may serve as effective biomarkers of mood disorders.
Factor H and properdin, which are key regulators of the alternative complement pathway [58,59], were also signi cantly increased in the BD group in our study. Although the roles of factor H and properdin in mood disorders remain unclear, dysregulation of the alternative pathway in neurological and neuropsychiatric disorders, including schizophrenia and epilepsy, has been reported [60,61]. Consistent with these results, we found here that the levels of complement factors involved in the classical and alternative complement pathways were signi cantly increased in BD patients, suggesting that the complement system might be involved in disease neuropathology.
Levels of all complement components except C3 were signi cantly higher in MDD patients than in HCs in our study. Although our results are inconsistent with previous studies reporting high C3 levels in MDD patients [22], other studies have found high peripheral protein levels for C4, but not for C3, in patients with MDD [62]. Even allowing for differences in speci c components, our results and the literature concur in supporting the idea that elevated levels of complement components may lead to hyperin ammatory responses in MDD.
Our results indicated a signi cant association of increased C1q, factor H, and properdin with GMV reduction in the mOFC of HCs and patients.

Limitations
Our study has some limitations, including its cross-sectional design. The single assessment of peripheral blood samples could not provide evidence of causality, leaving open the question of whether the observed alterations in complement levels and brain structure are a cause or consequence of BD and MDD [67]. Since C1q is associated with illness duration, longitudinal studies should be performed to explore the effect of illness duration on the expression of in ammatory markers. Moreover, the association between elevated complement components and mOFC structural changes needs to be carefully interpreted. In the present study, patients with severe physical illnesses, such as systemic lupus erythematosus, were excluded, and blood samples were collected from all subjects at the same time to avoid potential confounding due to variation in in ammation state. However, other factors affecting complement levels were not taken into account, such as concurrent immunological conditions (infections, atopy, autoimmune disorders) or sleep level [68], so future studies should address potential confounding from pre-existing in ammatory conditions. Furthermore, our BD patients did not show severe symptoms, although they showed increased levels of in ammatory markers, which may limit the generalizability of our ndings to different BD subpopulations. All our BD patients had been receiving medications, so we were unable to assess the effects of treatment history on our results.

Conclusion And Outlook
To the best of our knowledge, our study is the rst to compare structural neuroimaging analysis with complement factor levels in BD, MDD, and HCs. Consistent with previous brain structural ndings, we found that BD and MDD patients have common and distinct patterns of GMV and CT de cits, and that BD involves more severe in ammatory responses than MDD. We expect that this study will help improve our understanding of the association between peripheral in ammation and altered brain structure in bipolar disorders.

Declarations
Availability of data and materials All data used in the current study are available from the corresponding author on reasonable request.

Figure 1
Gray matter volume and cortical thickness differences between participants with bipolar disorder and major depressive disorder, compared to healthy controls. (a). Three group comparisons: The gray matter volume of the middle cingulum, medial orbital frontal cortex, right superior temporal gyrus, right pre/post central gyrus and right occipital gyrus differed signi cantly among the three groups (color in yellow); (b).
Compared with controls, bipolar disorder patients showed decreased gray matter volume in middle cingulum, right superior temporal gyrus, medial frontal orbital cortex and right superior occipital cortex (color in blue); (c). Compared with controls, the major depressive disorder patients showed decreased gray matter volume in right pre/postcentral gyrus, middle cingulum and medial orbital frontal cortex (color in blue); (d). The gure in the left showed three group comparison results of the cortical thickness, including bilateral precentral gyrus and left superior frontal gyrus (color in red); the gure in the middle showed that cortical thickness in the left superior frontal gyrus was decreased in bipolar patients group compared with controls (cold color); the gure in the right showed that compared with controls, major depressive disorder patients showed decreased cortical thickness in bilateral precentral gyrus (cold color). (All the results were co-variated out for age, gender and education years, and the total intracranial volume was additionally co-variated out for gray matter volume analysis, voxel threshold signi cance was set at p < 0.0005, cluster level was set at p < 0.05, family wise error corrected). BD: bipolar disorder, MDD: major depressive disorder, HC: healthy control.