Abstract
Complex brain disorders like schizophrenia may have multifactorial origins related to mis-timed heritable and environmental factors interacting during neurodevelopment. Infections, inflammation, and autoimmune diseases are over-represented in schizophrenia leading to immune system-centered hypotheses. Complement component C4 is genetically and neurobiologically associated with schizophrenia, and its dual activity peripherally and in the brain makes it an exceptional target for biomarker development. Studies to evaluate the biomarker potential of plasma or serum C4 in schizophrenia do so to understand how peripheral C4 might reflect central nervous system-derived neuroinflammation, synapse pruning, and other mechanisms. This effort, however, has produced mostly conflicting results, with peripheral C4 sometimes elevated, reduced, or unchanged between comparison groups. We undertook a pilot biomarker development study to systematically identify sociodemographic, genetic, and immune-related variables (autoimmune, infection-related, gastrointestinal, inflammatory), which may be associated with plasma C4 levels in schizophrenia (SCH; n = 335) and/or in nonpsychiatric comparison subjects (NCs; n = 233). As with previously inconclusive studies, we detected no differences in plasma C4 levels between SCH and NCs. In contrast, levels of general inflammation, C-reactive protein (CRP), were significantly elevated in SCH compared to NCs (ANOVA, F = 20.74, p < 0.0001), suggestive that plasma C4 and CRP may reflect different sources or causes of inflammation. In multivariate regressions of C4 gene copy number variants, plasma C4 levels were correlated only for C4A (not C4B, C4L, C4S) and only in NCs (R Coeff = 0.39, CI = 0.01–0.77, R2 = 0.18, p < 0.01; not SCH). Other variables associated with plasma C4 levels only in NCs included sex, double-stranded DNA IgG, tissue-transglutaminase (TTG) IgG, and cytomegalovirus IgG. Toxoplasma gondii IgG was the only variable significantly correlated with plasma C4 in SCH but not in NCs. Many variables were associated with plasma C4 in both groups (body mass index, race, CRP, N-methyl-D-aspartate receptor (NMDAR) NR2 subunit IgG, TTG IgA, lipopolysaccharide-binding protein (LBP), and soluble CD14 (sCD14). While the direction of most C4 associations was positive, autoimmune markers tended to be inverse, and associated with reduced plasma C4 levels. When NMDAR-NR2 autoantibody-positive individuals were removed, plasma C4 was elevated in SCH versus NCs (ANOVA, F = 5.16, p < 0.02). Our study was exploratory and confirmation of the many variables associated with peripheral C4 requires replication. Our preliminary results point toward autoimmune factors and exposure to the pathogen, T. gondii, as possibly significant contributors to variability of total C4 protein levels in plasma of individuals with schizophrenia.
Introduction
Schizophrenia is a neurodevelopmental disorder with an etiology and pathophysiology increasingly linked to interacting environmental and genetic immune sources [1‒4]. Pathogen exposures, autoimmune comorbidities, gut dysbiosis, acute inflammatory episodes, and chronic low-grade inflammation are all implicated risk factors [5‒17]. The specific mechanism(s) underlying inflammation and driving the immune component of this complex brain disorder, however, has been difficult to identify. Distinguishing sub-categories of the operative inflammatory processes in schizophrenia is not only important from a mechanistic perspective but also as a guide if anti-inflammatory drugs are being considered for use in this disorder. For example, inflammatory pathologies or comorbidities are not managed using the same interventions, and each inflammatory type requires a different treatment (i.e., antibiotics, antivirals, antifungals, non-steroidal anti-inflammatories, and steroidal-derived immune inhibition compounds).
The complement system has been implicated in schizophrenia etiology and pathophysiology for a long time, and its components contribute a first line of defense against antigens, while also inducing inflammation as part of the innate immune response [18‒25]. Schizophrenia is also highly heritable, and the search for susceptibility genes has focused on immuno-genetically relevant loci, among other factors [26, 27]. Evidence for a genetic association of the complement component, C4, and its structural isoforms, C4A and C4B, with schizophrenia has been re-invigorated into an intriguing model connecting increasing C4 copy number variants (CNVs) with C4A overexpression in the brain, and excessive synapse pruning [28].
Much research effort has been directed at scrutinizing and perfecting this C4 model to better understand how peripheral functions of C4 are related to complement mechanisms operative in the central nervous system (CNS) in individuals with schizophrenia. While the aims may differ from study to study, a shared goal has been a desire to develop peripheral biomarkers that reflect pathologies relevant to schizophrenia. Such functional endpoints that could be linkable to a plasma C4 biomarker include abnormal synaptic pruning, cortical thinning, neuropil contraction, and dendritic spine deficiencies [28‒37]. Future potential biomarker goals might include detection of neuroinflammation, more rapid identification of early psychoses, identification of cognitive deficits and other symptomatology, choice of treatment, and monitoring of treatment efficacies [38‒43]. Interrogations of peripheral C4, however, have produced remarkably mixed results, with peripheral C4 sometimes elevated, sometimes reduced and sometimes unchanged in schizophrenia versus comparison groups [33, 36, 38, 40, 44‒50].
Analyses of cerebrospinal fluid (CSF) may offer a more broad perspective, given the functional role of this body fluid as a link between peripheral and CNSs. However, studies of C4 levels in CSF samples from patients with schizophrenia seem as similarly conflicted as the plasma and serum studies. For example, CSF C4, but not plasma C4, was increased in individuals with schizophrenia compared to controls, and this group difference was only detectable when comparisons were corrected for age and sex [51]. Investigators found C4A elevations in CSF from first episode psychosis patients who developed schizophrenia (FEP-SCZ) compared to healthy control group levels and levels from FEP patients who did not develop schizophrenia [52]. C4B levels in CSF were not different between groups, as expected if the C4A locus is driving the C4 genetic association in schizophrenia. Of note, C4B and related gene deficiencies have been reported associated with schizophrenia [19, 33]. In a study of individuals at clinical high risk for psychosis, antipsychotic-naïve FEP and healthy controls, an array of complement proteins including C4 was elevated in serum of both patient groups compared to controls, but no inter-group differences were present in CSF. Furthermore, body mass index (BMI) levels and storage times were associated with CSF complement measures [53].
Before we can effectively design a biomarker to study synapse pruning and related outcomes, we must understand why peripheral plasma or serum C4 levels are not straightforward reflections of peripheral inflammation. The inconsistencies repeatedly encountered with studies of C4 in plasma or serum are probably driven by many factors including heterogeneities of psychiatric diagnoses, problematic study designs and multiple types and sources of inflammation that are not uniformly measurable by a single inflammatory marker. Inconclusive findings may also result from intrinsic and extrinsic factors that interfere, confound, or otherwise co-associate with plasma C4 levels in schizophrenia and comparison groups. Published studies already are documenting significant associations of age, sex, and race with a variety of C4 and related measures [39, 51, 54‒56].
Our goal was to begin an exploratory C4 biomarker development process by systematically identifying factors that might contribute to peripheral C4 levels in schizophrenia and a nonpsychiatric comparison group. We examined plasma C4 associations with standard sociodemographic variables and clinical factors (age, BMI, socioeconomic status, race, sex, and cigarette smoking), C4 gene CNVs (C4A, C4B, C4L, C4S), and categories of immune variables related to autoimmunity, infections, gastrointestinal (GI) dysbiosis, and general inflammation. Details regarding the variables chosen are further described in the Methods section. Ultimately, we hope that this exploratory work will lead to well-matched prospective studies to determine if it is worthwhile pursuing peripheral C4 as a biomarker of neuroinflammation, synapse pruning, and other related neural mechanisms.
Materials and Methods
Study Population
Participants were recruited at Sheppard Pratt in Baltimore, MD, USA as part of a long-term, ongoing study of schizophrenia. Started in 1999, this cohort is a source of sociodemographic, clinical, and other types of data and biospecimens with an overall goal to understand the immune components of major psychiatric disorders. For our current study, we reviewed the database and identified individuals with schizophrenia (SCH; n = 335) and a non-psychiatric comparison group (NCs; n = 233) for whom we had plasma C4 biomarker data. Sample sizes varied for some of the follow-up multivariate comparisons, and these are so noted in the relevant sections. Basic sociodemographic and other standard data typically applied to studying psychoimmunology-derived cohorts are shown in Table 1.
Standard sociodemographic and other variables in the study populations
Sociodemographic/other test variable . | NCs (n = 233) . | SCH (n = 335) . | Test stat . | p value . |
---|---|---|---|---|
Age (mean years+SE) | 31.69+0.72 | 38.26+0.72 | T = −4.49 | 0.0001 |
BMI (mean score+SE) | 27.46+0.45 | 30.08+0.38 | T = −4.42 | 0.0001 |
Maternal education (mean years+SE) | 13.81+0.17 | 12.93+0.15 | T = 3.92 | 0.0001 |
Race, n (% Black) | 88 (37.77) | 167 (49.85) | χ2 = 8.11 | 0.004 |
RBANS (mean score+SE) | 84.96+0.76 | 64.26+0.64 | T = 20.73 | 0.0001 |
Sex, n (% female) | 134 (57.51) | 123 (36.72) | χ2 = 23.99 | 0.001 |
Smoker, n (%) | 37 (15.88) | 200 (59.70) | χ2 = 108.53 | 0.001 |
Sociodemographic/other test variable . | NCs (n = 233) . | SCH (n = 335) . | Test stat . | p value . |
---|---|---|---|---|
Age (mean years+SE) | 31.69+0.72 | 38.26+0.72 | T = −4.49 | 0.0001 |
BMI (mean score+SE) | 27.46+0.45 | 30.08+0.38 | T = −4.42 | 0.0001 |
Maternal education (mean years+SE) | 13.81+0.17 | 12.93+0.15 | T = 3.92 | 0.0001 |
Race, n (% Black) | 88 (37.77) | 167 (49.85) | χ2 = 8.11 | 0.004 |
RBANS (mean score+SE) | 84.96+0.76 | 64.26+0.64 | T = 20.73 | 0.0001 |
Sex, n (% female) | 134 (57.51) | 123 (36.72) | χ2 = 23.99 | 0.001 |
Smoker, n (%) | 37 (15.88) | 200 (59.70) | χ2 = 108.53 | 0.001 |
Maternal education was used as a surrogate variable for socioeconomic status.
NCs, nonpsychiatric comparison subjects; SCH, schizophrenia; Test stat, test statistic; BMI, body mass index; RBANS, repeatable battery for the assessment of neuropsychological status.
As described previously [57, 58], individuals received DSM-IV-TR diagnoses of schizophrenia, schizophreniform, or schizoaffective disorders [59] and were between the ages of 18 and 65. Cognitive functioning was examined with the Repeatable Battery for the Assessment of Neuropsychological Status (RBANS) Form A [60]. The comparison group was composed of individuals who did not have a history of psychiatric disorder based on the Structured Clinical Interview for DSM-IV Axis I Disorders Non-Patient Edition [61]. Exclusion criteria for both groups were the following: mental retardation; clinically significant medical disorder that would affect cognitive performance; any history of intravenous substance abuse or a primary diagnosis of substance abuse or substance dependence. For the comparison group, any active substance misuse was considered an exclusion criterion.
We received approval for these studies by the Institutional Review Boards (IRB) of Sheppard Pratt and the Johns Hopkins Medical Institution following established guidelines. Written informed consent was obtained from all participants after study procedures were explained. This research was performed in accordance with the Code of Ethics of the World Medical Association (Declaration of Helsinki) for experiments involving humans.
Laboratory Tests
Blood was drawn at the time of interview using Becton-Dickinson’s Cell Preparation Tubes containing sodium citrate. Plasma was separated and stored at −80°C for the blood biomarker studies. Peripheral blood mononuclear cells were isolated and stored at −80°C for the C4 gene marker studies.
C4 genotyping was performed on 108 individuals in the sychiatric comparison group and on 168 individuals with schizophrenia. For C4 genotyping, DNA was extracted from peripheral blood mononuclear cells using Qiagen’s DNAeasy kit and stored at −80°C. Complement C4A, C4B, C4S, and C4L CNVs were imputed from GWAS SNPs data using previously described methods including use of DNA microarrays on the extracted DNA (Sekar et al. [28]; https://github.com/freeseek/imputec4). Utilized SNP chips were PsychChip (PsychChip_15048346), Omni2.5 (HumanOmni25M-8v1-1), and GSA (GSAMD-24v1-0_20011747).
To examine plasma C4 associations with other plasma immune-related variables, we searched our database and selected the following targets: 5 autoimmune markers (double-stranded DNA [dsDNA IgG], an often-used biomarker for lupus [62], the Celiac disease-associated tissue-transglutaminase (TTG) IgA, IgG, and gliadin food antigen IgG [63], and NMDAR-NR2 subunit IgG, an autoantibody target often studied in psychoses and schizophrenia and related models [64‒66]); 4 markers of past infectious events with selections representing different taxa of pathogens (fungi (C. albicans IgG) [67], viruses (cytomegalovirus [CMV], EBV IgG) and a parasite (T. gondii IgG) [68], which is also cross-listed in the GI category, because infection produces GI inflammation [69]); 5 indices of GI inflammation or dysbiosis (gliadin IgG (cross-listed as an autoantibody target) [70], bacteria-associated lipopolysaccharide (LPS)-binding protein and its inflammation partner, sCD14 (a marker of monocyte response to LPS and also cross-listed as an inflammatory marker) [71], translocation of the dietary yeast, Saccharomyces cerevisiae, used to help diagnose Crohn’s Disease [72], and T. gondii IgG, cross-listed as a pathogen); and C-reactive protein (CRP) and sCD14 (cross-listed with GI) used as basic measures of inflammation [58, 73].
Plasma biomarker quantification was performed using commercially available enzyme-linked immunosorbent assays (ELISAs). Total C4 protein in plasma was measured using the ELISA kit manufactured by Abcam (catalog# ab108824, Waltham, MA, USA). The methods, ELISA kit sources, and statistical analyses to distinguish psychiatric case and comparison group levels of the other targeted biomarkers have been previously described for all except the dsDNA IgG [63, 67, 68, 71, 72, 74‒76]. The ELISA kit targeting the biomarker, dsDNA IgG, was purchased from IBL America, Minneapolis, MN, USA (catalog# ORG 604).
Data Analyses
ANOVAs with post hoc Sidak analyses and t tests were used to detect bivariate associations between continuous variables, such as the baseline C4 plasma levels between individuals with SCH and NCs. χ2 analyses were used to detect bivariate associations between categorical variables such as race and sex differences in SCH and NCs.
Multiple linear regression models were used to examine correlations among continuous variables such as C4 plasma levels with C4 CNVs, antibodies, and other biomarker levels. Biomarker positivity was defined based on quantitative levels of these markers in NCs. Biomarker values in the SCH population which exceeded the 90th percentile of NCs values were considered seropositive. Multivariate linear regressions included the covariates: age, BMI, maternal education (surrogate for socioeconomic status), race, sex, and tobacco smoking. An assay plate covariate was included in the ELISA data analyses regression models to correct for plate-to-plate variation. These regressions were performed both with and without robust standard error correction, a statistical test increasingly applied to maximize additional clustered data that we had for each individual in the form of repeated outcome measurements over time [77]. We did not formally account for multiple comparisons due to the exploratory nature of the study and to minimize the occurrence of type B (false negative) errors. Thus, a p value of less than 0.05 was considered significant.
Results
Plasma C4 and General/Systemic Inflammation
Plasma C4 levels in SCH were not significantly different from those in the non-psychiatric comparison group (NCs: n = 233; mean + SE = 1.23 + 0.07 μg/mL; SCH: n = 335; mean + SE = 1.32 + 0.05 μg/mL; T test t = −1.01, p > 0.15; ANOVA F = 1.02, p > 0.31). In contrast, plasma levels of the general marker for inflammation, CRP, were significantly elevated in SCH compared to NCs (NCs: 0.55 + 0.04 optical density units (od); SCH: 0.84 + 0.04 od; T test t = −4.55, p < 0.0001; ANOVA F = 20.74, p < 0.0001). In both NCs and SCH, plasma C4 levels were well-correlated with plasma CRP levels in multiple linear regressions that included age, BMI, maternal education, race, sex, and smoking (NCs: n = 233; R coefficient = 0.30, CI = 0.09–0.51, R2 = 0.12, p < 0.0003; SCH: n = 335; R coefficient = 0.24, CI = 0.11–0.37, R2 = 0.09, p < 0.0002).
Plasma C4 and C4 Gene Copy Number Variants
In multiple linear regressions, C4A CNVs were significantly associated with C4 plasma levels in NCs but not in SCH (Table 2; NCs: n = 108; R coefficient = 0.39, CI = 0.01–0.77, R2 = 0.18, p < 0.01; SCH: n = 168; R coefficient = −0.03, CI = −0.27 to 0.21, R2 = 0.06, p > 0.79). C4B, C4L, and C4S CNVs were not significantly associated with plasma C4 in either NCs or SCH (All p > 0.27).
Multivariate regressions of C4 gene copy number variants with plasma C4
. | NCs (n = 108) . | SCH (n = 168) . | ||||||
---|---|---|---|---|---|---|---|---|
Genetic test variable (C4 CNVs) . | R coeff . | 95th% CI . | R2 . | p value . | R coeff . | 95th% CI . | R2 . | p value . |
C4A | 0.39 | 0.01–0.77 | 0.18 | 0.01 | −0.03 | −0.27 to 0.21 | 0.06 | >0.79 |
C4B | −0.17 | −0.68 to 0.35 | 0.15 | >0.52 | −0.03 | −0.34 to 0.29 | 0.06 | >0.86 |
C4L | 0.04 | −0.22 to 0.30 | 0.15 | >0.75 | −0.09 | −0.26 to 0.07 | 0.07 | >0.27 |
C4S | 0.12 | −0.18 to 0.41 | 0.15 | >0.45 | 0.10 | −0.10 to 0.31 | 0.06 | >0.31 |
. | NCs (n = 108) . | SCH (n = 168) . | ||||||
---|---|---|---|---|---|---|---|---|
Genetic test variable (C4 CNVs) . | R coeff . | 95th% CI . | R2 . | p value . | R coeff . | 95th% CI . | R2 . | p value . |
C4A | 0.39 | 0.01–0.77 | 0.18 | 0.01 | −0.03 | −0.27 to 0.21 | 0.06 | >0.79 |
C4B | −0.17 | −0.68 to 0.35 | 0.15 | >0.52 | −0.03 | −0.34 to 0.29 | 0.06 | >0.86 |
C4L | 0.04 | −0.22 to 0.30 | 0.15 | >0.75 | −0.09 | −0.26 to 0.07 | 0.07 | >0.27 |
C4S | 0.12 | −0.18 to 0.41 | 0.15 | >0.45 | 0.10 | −0.10 to 0.31 | 0.06 | >0.31 |
CNV, copy number variant; NCs, nonpsychiatric comparison subjects; SCH, schizophrenia; R coeff, regression coefficient; CI, confidence interval.
Regressions included all variables found in Table 1.
Plasma C4 and Basic Sociodemographic and Clinical Variables
For these analyses, we first sought to identify if C4 plasma levels were associated with those sociodemographic and other variables that are commonly used to protect multivariate comparisons from confounding factors. Second, we have multiple samples per individual and wanted to maximize our ability to identify a C4 association if one were present. Therefore, we first compared analyses done using the standard dataset composed of one sample per individual (n = 233 NC; n = 335 SCH). We then applied robust standard errors to our regression models to accommodate additional data from those individuals for whom we had multiple samples per individual (n = 290 NCs samples; n = 630 SCH samples).
In multiple linear regressions, BMI and sex were significantly associated with C4 plasma levels in NCs using the standard sample size analysis (p < 0.003); the inclusion of robust standard error analyses confirmed these significant associations of C4 with BMI and sex and also identified an association with race (all p < 0.0001). In SCH, analyses of both sample sizes identified BMI and race as having significant associations with C4 plasma levels (p < 0.0009–0.02). These data are shown in Table 3.
Multivariate regressions of sociodemographic and related study variables with plasma C4
. | NCs . | SCH . | ||||||||
---|---|---|---|---|---|---|---|---|---|---|
Sociodem/other test variable . | n . | R . | 95th% CI . | R2 . | p value . | n . | R . | 95th% CI . | R2 . | p value . |
Individuals | ||||||||||
Age | 233 | −0.01 | −0.02 to 0.01 | 0.09 | >0.63 | 335 | 0.01 | −0.01 to 0.01 | 0.05 | >0.46 |
BMI | 233 | 0.03 | 0.01–0.05 | 0.09 | 0.003 | 335 | 0.02 | 0.01–0.04 | 0.05 | 0.02 |
Maternal education | 233 | 0.04 | −0.02 to 0.09 | 0.09 | >0.18 | 335 | 0.01 | −0.03 to 0.05 | 0.05 | >0.61 |
Race | 233 | 0.25 | −0.04 to 0.53 | 0.09 | >0.09 | 335 | 0.24 | 0.03–0.45 | 0.05 | 0.02 |
Sex | 233 | −0.37 | −0.63 to 0.12 | 0.09 | 0.003 | 335 | 0.21 | −0.02 to 0.42 | 0.05 | >0.07 |
Smoker | 233 | −0.01 | −0.36 to 0.35 | 0.09 | >0.96 | 335 | −0.01 | −0.23 to 0.20 | 0.05 | >0.92 |
Samples | ||||||||||
Age | 290 | −0.01 | −0.01 to 0.01 | 0.08 | >0.87 | 630 | −0.01 | −0.01 to 0.01 | 0.05 | >0.96 |
BMI | 290 | 0.03 | 0.01–0.04 | 0.08 | 0.0001 | 630 | 0.02 | 0.01–0.03 | 0.05 | 0.0009 |
Maternal education | 290 | 0.01 | −0.04 to 0.06 | 0.08 | >0.65 | 630 | 0.01 | −0.03 to 0.05 | 0.05 | >0.71 |
Race | 290 | 0.32 | 0.04–0.60 | 0.08 | 0.0001 | 630 | 0.27 | 0.08–0.46 | 0.05 | 0.0009 |
Sex | 290 | −0.33 | −0.58 to 0.07 | 0.08 | 0.0001 | 630 | 0.14 | −0.06 to 0.33 | 0.05 | >0.17 |
Smoker | 290 | 0.04 | −0.31 to 0.39 | 0.08 | >0.82 | 630 | −0.11 | −0.30 to 0.07 | 0.05 | >0.24 |
. | NCs . | SCH . | ||||||||
---|---|---|---|---|---|---|---|---|---|---|
Sociodem/other test variable . | n . | R . | 95th% CI . | R2 . | p value . | n . | R . | 95th% CI . | R2 . | p value . |
Individuals | ||||||||||
Age | 233 | −0.01 | −0.02 to 0.01 | 0.09 | >0.63 | 335 | 0.01 | −0.01 to 0.01 | 0.05 | >0.46 |
BMI | 233 | 0.03 | 0.01–0.05 | 0.09 | 0.003 | 335 | 0.02 | 0.01–0.04 | 0.05 | 0.02 |
Maternal education | 233 | 0.04 | −0.02 to 0.09 | 0.09 | >0.18 | 335 | 0.01 | −0.03 to 0.05 | 0.05 | >0.61 |
Race | 233 | 0.25 | −0.04 to 0.53 | 0.09 | >0.09 | 335 | 0.24 | 0.03–0.45 | 0.05 | 0.02 |
Sex | 233 | −0.37 | −0.63 to 0.12 | 0.09 | 0.003 | 335 | 0.21 | −0.02 to 0.42 | 0.05 | >0.07 |
Smoker | 233 | −0.01 | −0.36 to 0.35 | 0.09 | >0.96 | 335 | −0.01 | −0.23 to 0.20 | 0.05 | >0.92 |
Samples | ||||||||||
Age | 290 | −0.01 | −0.01 to 0.01 | 0.08 | >0.87 | 630 | −0.01 | −0.01 to 0.01 | 0.05 | >0.96 |
BMI | 290 | 0.03 | 0.01–0.04 | 0.08 | 0.0001 | 630 | 0.02 | 0.01–0.03 | 0.05 | 0.0009 |
Maternal education | 290 | 0.01 | −0.04 to 0.06 | 0.08 | >0.65 | 630 | 0.01 | −0.03 to 0.05 | 0.05 | >0.71 |
Race | 290 | 0.32 | 0.04–0.60 | 0.08 | 0.0001 | 630 | 0.27 | 0.08–0.46 | 0.05 | 0.0009 |
Sex | 290 | −0.33 | −0.58 to 0.07 | 0.08 | 0.0001 | 630 | 0.14 | −0.06 to 0.33 | 0.05 | >0.17 |
Smoker | 290 | 0.04 | −0.31 to 0.39 | 0.08 | >0.82 | 630 | −0.11 | −0.30 to 0.07 | 0.05 | >0.24 |
Sociodem, sociodemographic; n, sample size; NCs, nonpsychiatric comparison subjects; SCH, schizophrenia; R, regression coefficient; CI, confidence interval.
Regressions included all variables found in Table 1.
We further characterized these sociodemographic associations and found that C4 plasma levels were elevated in individuals who were Black in both the NCs and SCH groups (NCs: n = 88 Black, 1.43 + 0.11, n = 145 non-Black, 1.11 + 0.08; T = −2.33, 1-tail p < 0.01; SCH: n = 167 Black, 1.44 + 0.07, n = 168 non-Black, 1.19 + 0.07; T = −2.35, 1-tail p < 0.01), and in females of the NCs (n = 134 females, 1.38 + 0.08, n = 99 males, 1.03 + 0.11, T = 2.58, 1-tail p < 0.005). C4 plasma levels were not significantly correlated with RBANS measures of cognitive functioning in either SCH or NCs (all >0.05).
Plasma C4 and Immune-Related Variables
We detected numerous associations of plasma C4 with other immune-related plasma biomarkers, and this complete dataset is shown in Table 4. Further description of these findings is present in the next section.
Multivariate regressions of immune-related plasma biomarkers with plasma C4
. | NCs . | SCH . | ||||||||
---|---|---|---|---|---|---|---|---|---|---|
Immune-related test variable . | n . | R . | 95th% CI . | R2 . | p value . | n . | R . | 95th% CI . | R2 . | p value . |
Autoimmune | ||||||||||
dsDNA IgG | 82 | 3.07 | 0.60–5.53 | 0.32 | 0.0001 | 260 | −1.20 | −2.63 to 0.23 | 0.07 | 0.10 |
Gliadin IgGa | 290 | −0.31 | −1.11 to 0.49 | 0.09 | 0.45 | 630 | 0.08 | −0.06 to 0.23 | 0.05 | 0.26 |
NMDAR-NR2 IgG | 290 | −0.38 | −0.50 to 0.26 | 0.23 | 0.0001 | 630 | −0.39 | −0.48 to 0.3 | 0.18 | 0.0001 |
Tissue transglutaminase IgA | 202 | 0.15 | 0.06–0.25 | 0.16 | 0.0001 | 380 | −0.55 | −1.13 to 0.03 | 0.15 | 0.0001 |
Tissue transglutaminase IgG | 202 | 0.22 | 0.07–0.37 | 0.17 | 0.0001 | 380 | −0.04 | −0.64 to 0.56 | 0.14 | 0.90 |
Infection | ||||||||||
C. albicans IgG | 290 | 0.20 | −0.11 to 0.52 | 0.09 | 0.20 | 630 | 0.08 | −0.08 to 0.25 | 0.05 | 0.33 |
CMV IgG | 290 | 0.13 | 0.02–0.24 | 0.10 | 0.0001 | 630 | 0.08 | −0.01 to 0.17 | 0.06 | 0.07 |
EBV IgG | 290 | −0.06 | −0.29 to 0.18 | 0.09 | 0.64 | 630 | 0.10 | −0.07 to 0.27 | 0.05 | 0.26 |
T. gondii IgGa | 290 | 0.13 | −0.07 to 0.33 | 0.09 | 0.19 | 630 | 0.30 | 0.12–0.49 | 0.06 | 0.001 |
Gastrointestinal | ||||||||||
Gliadin IgGa | 290 | −0.31 | −1.11 to 0.49 | 0.09 | 0.45 | 630 | 0.08 | −0.06 to 0.23 | 0.05 | 0.26 |
LBP | 290 | 0.05 | 0.04–0.06 | 0.26 | 0.0001 | 630 | 0.03 | 0.03–0.04 | 0.17 | 0.0001 |
S. cerevisiae IgG | 290 | 0.13 | −0.13 to 0.39 | 0.09 | 0.33 | 630 | 0.02 | −0.13 to 0.18 | 0.05 | 0.77 |
sCD14a | 290 | 0.66 | 0.56–0.75 | 0.47 | 0.0001 | 630 | 0.46 | 0.39–0.52 | 0.31 | 0.0001 |
T. gondii IgGa | 290 | 0.13 | −0.07 to 0.33 | 0.09 | 0.19 | 630 | 0.30 | 0.12–0.49 | 0.06 | 0.001 |
Other inflammation | ||||||||||
CRP (general) | 290 | 0.35 | 0.17–0.53 | 0.13 | 0.0001 | 630 | 0.23 | 0.13–0.33 | 0.08 | 0.0001 |
sCD14a (monocyte activation) | 290 | 0.66 | 0.56–0.75 | 0.47 | 0.0001 | 630 | 0.46 | 0.39–0.52 | 0.31 | 0.0001 |
. | NCs . | SCH . | ||||||||
---|---|---|---|---|---|---|---|---|---|---|
Immune-related test variable . | n . | R . | 95th% CI . | R2 . | p value . | n . | R . | 95th% CI . | R2 . | p value . |
Autoimmune | ||||||||||
dsDNA IgG | 82 | 3.07 | 0.60–5.53 | 0.32 | 0.0001 | 260 | −1.20 | −2.63 to 0.23 | 0.07 | 0.10 |
Gliadin IgGa | 290 | −0.31 | −1.11 to 0.49 | 0.09 | 0.45 | 630 | 0.08 | −0.06 to 0.23 | 0.05 | 0.26 |
NMDAR-NR2 IgG | 290 | −0.38 | −0.50 to 0.26 | 0.23 | 0.0001 | 630 | −0.39 | −0.48 to 0.3 | 0.18 | 0.0001 |
Tissue transglutaminase IgA | 202 | 0.15 | 0.06–0.25 | 0.16 | 0.0001 | 380 | −0.55 | −1.13 to 0.03 | 0.15 | 0.0001 |
Tissue transglutaminase IgG | 202 | 0.22 | 0.07–0.37 | 0.17 | 0.0001 | 380 | −0.04 | −0.64 to 0.56 | 0.14 | 0.90 |
Infection | ||||||||||
C. albicans IgG | 290 | 0.20 | −0.11 to 0.52 | 0.09 | 0.20 | 630 | 0.08 | −0.08 to 0.25 | 0.05 | 0.33 |
CMV IgG | 290 | 0.13 | 0.02–0.24 | 0.10 | 0.0001 | 630 | 0.08 | −0.01 to 0.17 | 0.06 | 0.07 |
EBV IgG | 290 | −0.06 | −0.29 to 0.18 | 0.09 | 0.64 | 630 | 0.10 | −0.07 to 0.27 | 0.05 | 0.26 |
T. gondii IgGa | 290 | 0.13 | −0.07 to 0.33 | 0.09 | 0.19 | 630 | 0.30 | 0.12–0.49 | 0.06 | 0.001 |
Gastrointestinal | ||||||||||
Gliadin IgGa | 290 | −0.31 | −1.11 to 0.49 | 0.09 | 0.45 | 630 | 0.08 | −0.06 to 0.23 | 0.05 | 0.26 |
LBP | 290 | 0.05 | 0.04–0.06 | 0.26 | 0.0001 | 630 | 0.03 | 0.03–0.04 | 0.17 | 0.0001 |
S. cerevisiae IgG | 290 | 0.13 | −0.13 to 0.39 | 0.09 | 0.33 | 630 | 0.02 | −0.13 to 0.18 | 0.05 | 0.77 |
sCD14a | 290 | 0.66 | 0.56–0.75 | 0.47 | 0.0001 | 630 | 0.46 | 0.39–0.52 | 0.31 | 0.0001 |
T. gondii IgGa | 290 | 0.13 | −0.07 to 0.33 | 0.09 | 0.19 | 630 | 0.30 | 0.12–0.49 | 0.06 | 0.001 |
Other inflammation | ||||||||||
CRP (general) | 290 | 0.35 | 0.17–0.53 | 0.13 | 0.0001 | 630 | 0.23 | 0.13–0.33 | 0.08 | 0.0001 |
sCD14a (monocyte activation) | 290 | 0.66 | 0.56–0.75 | 0.47 | 0.0001 | 630 | 0.46 | 0.39–0.52 | 0.31 | 0.0001 |
n, sample size; NCs, nonpsychiatric comparison subjects; SCH, schizophrenia; R, regression coefficient; CI, confidence interval; dsDNA, double-stranded DNA; NMDAR-NR2, N-methyl-D-aspartate receptor-NR2 subunit; CMV, cytomegalovirus; EBV, Epstein-Barr virus; LPS, lipopolysaccharide; LBP, lipopolysaccharide-binding protein; sCD14, soluble CD14; CRP, C-reactive protein.
Regressions included all variables found in Table 1.
aCross-listed in multiple categories.
Summary of Plasma C4 with All Variable Types
As depicted in Table 5, we categorized the variables examined in this study according to if the C4 association was in (1) neither SCH nor NCs, (2) both SCH and NCs, (3) either SCH or NCs. We expected the last category to be the most informative since it would presumably identify differential C4 plasma levels between NCs and individuals with SCH. However, the C4 associations in both NCs and SCH are also informative as variables that might require matching in future studies.
Summary of plasma C4 associated variables in and out of schizophrenia
Test variable . | Presence of significant association . | ||
---|---|---|---|
NEITHER NCs nor SCH . | BOTH NCs and SCH . | EITHER NCs or SCHb . | |
p value . | p value . | p value . | |
Sociodemographic/other | |||
Age | NS | ||
BMI | 0.0001–0.0009 | ||
Maternal education | NS | ||
Race | 0.0001–0.0009 | ||
Sex | NCs: 0.0001 | ||
Smoker | NS | ||
C4 gene CNVs | |||
C4A | NCs: 0.01 | ||
C4B | NS | ||
C4L | NS | ||
C4S | NS | ||
Autoimmune | |||
dsDNA IgG | NCs: 0.0001 | ||
Gliadin IgGa | NS | ||
NMDAR-NR2 IgG | Both 0.0001 | ||
Tissue transglutaminase IgA | Both 0.0001 | ||
Tissue transglutaminase IgG | NCs: 0.0001 | ||
Infection | |||
C. albicans IgG | NS | ||
CMV IgG | NCs: 0.0001 | ||
EBV IgG | NS | ||
T. gondiia IgG | SCH: 0.0001 | ||
Gastrointestinal | |||
Gliadin IgGa | NS | ||
LBP | Both 0.0001 | ||
S. cerevisiae IgG | NS | ||
sCD14a | Both 0.0001 | ||
T. gondii IgGa | SCH: 0.001 | ||
Other inflammation | |||
CRP | Both 0.0001 | ||
sCD14a | Both 0.0001 |
Test variable . | Presence of significant association . | ||
---|---|---|---|
NEITHER NCs nor SCH . | BOTH NCs and SCH . | EITHER NCs or SCHb . | |
p value . | p value . | p value . | |
Sociodemographic/other | |||
Age | NS | ||
BMI | 0.0001–0.0009 | ||
Maternal education | NS | ||
Race | 0.0001–0.0009 | ||
Sex | NCs: 0.0001 | ||
Smoker | NS | ||
C4 gene CNVs | |||
C4A | NCs: 0.01 | ||
C4B | NS | ||
C4L | NS | ||
C4S | NS | ||
Autoimmune | |||
dsDNA IgG | NCs: 0.0001 | ||
Gliadin IgGa | NS | ||
NMDAR-NR2 IgG | Both 0.0001 | ||
Tissue transglutaminase IgA | Both 0.0001 | ||
Tissue transglutaminase IgG | NCs: 0.0001 | ||
Infection | |||
C. albicans IgG | NS | ||
CMV IgG | NCs: 0.0001 | ||
EBV IgG | NS | ||
T. gondiia IgG | SCH: 0.0001 | ||
Gastrointestinal | |||
Gliadin IgGa | NS | ||
LBP | Both 0.0001 | ||
S. cerevisiae IgG | NS | ||
sCD14a | Both 0.0001 | ||
T. gondii IgGa | SCH: 0.001 | ||
Other inflammation | |||
CRP | Both 0.0001 | ||
sCD14a | Both 0.0001 |
NCs, nonpsychiatric comparison subjects; SCH, schizophrenia; NS, not significant; dsDNA, double-stranded DNA; NMDAR-NR2, N-methyl-D-aspartate receptor-NR2 subunit; CMV, cytomegalovirus; EBV, Epstein-Barr virus; LPS, lipopolysaccharide; LBP, lipopolysaccharide-binding protein; sCD14, soluble CD14; CRP, C-reactive protein.
aCross-listed in multiple categories.
bPotentially diagnostic.
Variables associated with plasma C4 levels in both SCH and NCs included BMI, race, NMDAR-NR2 IgG, TTG IgA, LBP, SCD14, and CRP. Variables that were differentially associated with C4 levels in SCH versus NCs included sex, C4A CNV, dsDNA IgG, TTG IgG, CMV IgG, and T. gondii IgG. In this latter category, only the levels of T. gondii antibodies were positively associated with plasma C4 levels in schizophrenia; the remainder was associated with plasma C4 in the NCs. Most of the correlations detected in this study were positive except for the autoimmune variables, NMDAR-NR2, and tissue-transglutaminase autoantibodies which showed various patterns of inverse correlations with plasma C4. We further explored this autoimmune association in schizophrenia and found that when we removed individuals who were positive for the NMDAR-NR2 autoantibodies (n = 22), circulating C4 levels were significantly elevated compared to the NCs (NCs: n = 233, 1.23 + 0.07 μg/mL; SCH: n = 313, 1.42 + 0.05 μg/mL; ANOVA, F = 5.16, p < 0.02). Removal of individuals with schizophrenia who were seropositive for each of the other autoimmune markers did not similarly restore plasma C4 level differences between SCH and NCs.
Discussion
In this exploratory study, we report on a complex dynamic of sociodemographic, genetic, autoimmune, infectious, GI, and inflammatory factors that may influence plasma complement C4 levels relevant to schizophrenia. Complement activities in the circulatory system are primarily known to us in an immune context where complement molecules and pathways function to remove pathogens and cellular debris from our bodies. This innate immune response is characterized by a pro-inflammatory environment that subsides when the threat is gone [23]. An abnormally sustained chronic inflammatory period is a pathology that is increasingly reported in a subset of individuals with schizophrenia [14, 78, 79].
Our first step to evaluate plasma C4 as a potential biomarker was to assess baseline differences in a known C4 phenotype, peripheral inflammation [17], between individuals with schizophrenia and a nonpsychiatric comparison group. We found no significant differences between plasma C4 levels of individuals with schizophrenia and the comparison group, in direct contrast to a routinely used marker of inflammation, plasma CRP levels, which were significantly elevated in people with schizophrenia. These preliminary findings suggest that inflammation of some type was differentially present in schizophrenia and was detectable by the CRP. The failure of C4 to also detect this inflammation may suggest the presence of a different immunopathology and possibly other operative mechanism driving down plasma C4 levels.
In addition to its classic role removing pathogens and cellular debris, the complement system, when dysregulated, impaired, or mis-activated, has been shown to contribute to autoimmune disorders, and also increasingly to neurological disorders [80, 81]. Using multivariate regression models, we found that autoimmune markers and particularly autoantibodies directed against the NMDAR-NR2 subunit were associated with decreased plasma C4 levels. This finding is not surprising given that epidemiology studies have identified a significant autoimmune component comorbid to schizophrenia in some individuals [66, 82, 83]. Of further note, alterations of complement system factors, including plasma C4, can help diagnose suspected cases of autoimmunity [84]. It is possible that if peripheral C4A is upregulated as a function of high C4A gene CNVs, it may not always be observed peripherally due to masking by an undetectable saturation/consumption of C4 during an autoimmune event. This possibility should be examined in future longitudinal studies prospectively designed to trace plasma C4 changes over time in populations of individuals with SCH with and without known autoimmune comorbidities.
The associations of peripheral C4 with many diverse types of markers in both diagnostic groups motivated us to also look for other factors that may mediate apparent decreases of plasma C4. C4 gene deficiencies such as C4 null alleles may be prime candidates to interrogate in future studies, as these mutations are known to lead to autoimmune susceptibilities, as are C4B partial or complete deficiencies [19, 21, 33, 84]. It is of note that the ancestral human leukocyte antigen (HLA) 8.1 haplotype, one of the most associated haplotypes with autoimmune disorders, is thought to be protective against schizophrenia likely due to its natural lack of the C4A locus [28, 85, 86]. We examined CNVs of the C4 gene with the expectation that a high copy number would lead to elevated plasma C4 levels. Conversely, a low CNV (the closest we can come to in this study to a null allele) would in turn be hypothesized to associate with a low plasma C4 level. We found that C4A CNVs were positively correlated with levels of plasma C4 in NCs only, suggesting that a direct linear relationship between the number of C4A CNVs and plasma C4 levels in controls may be disrupted in schizophrenia. We re-analyzed these data with the NMDAR-NR2 autoantibody-seropositive individuals removed; however, the linear correlation of plasma C4 and C4A CNV variables was not restored (data not shown), suggesting that other factors may be operative.
Overall, these results further highlight that conditions leading to and expressed as inflammation may be highly heterogeneous. In addition to sociodemographic, genetic, and autoimmune markers, other indices including those derived from infectious disease sources and gut dysbiosis were found to be associated with plasma C4 patterns. Exposure to T. gondii, as measured by IgG antibodies, was the only marker that was significantly correlated with plasma C4 in schizophrenia but not the comparison group. Of note, T. gondii is a pathogen that is not only neurotropic, but it enters the host through ingestion and creates an inflammatory environment in the GI tract [69]. For this reason, we cross-listed T. gondii in the gut dysbiosis category. The viral marker, CMV IgG, was correlated with plasma C4 only in controls, suggesting that in schizophrenia, this linear relationship is disrupted, as similar to our observations of C4A copy number correlations with plasma C4 only in controls. More in-depth mechanistic studies are certainly required, but one explanation may be that these pathogens evade host immune systems in different ways, such as through complement regulation and inactivation [87‒89].
A long-term goal of these studies is to better understand if the complement system can be harnessed to serve as a biomarker of neuroinflammation, synaptic pruning or even if complement pathways can be modulated for therapeutic use. Recent evidence, for example, supports a complement and coagulation protein pathway mechanism by which omega-3 polyunsaturated fatty acids might contribute to cognitive and symptom improvements in individuals with early psychosis [41]. A role for C4 in neuroinflammation is based on positive, negative, and mixed associations of C4 levels with markers of white matter integrity, CSF biochemistry, cortical thinning, and patterns of expression with postmortem brain structures and functions [31, 32, 42, 51, 52, 54, 90‒97].
Our study has a number of limitations which further accentuate the very exploratory nature of our findings. Methodologically, we were limited by our measurement of total C4 protein levels in plasma, rather than a separate measurement of plasma C4A and C4B. It is generally difficult to distinguish C4A from C4B using ELISA-based methods, but this limitation was overcome in a recent study that used mass spectrometry to distinguish C4A and C4B components in CSF [52]. Our schizophrenia and comparison groups were significantly different for each sociodemographic measure, which reinforces the importance of having very-well-matched study populations and rigorous statistics to accommodate the many variables that may contribute to a study’s outcome. As a pilot study aimed to assess the prospects of a potential biomarker, we started with variables in our own database. This action might be construed as a biased selection based on convenience, rather than as a carefully constructed and systematic retrospective examination to prepare for future resource-intensive, prospective studies. Furthermore, individuals in our schizophrenia group were not medication-naïve. In a study of persons at high risk for psychosis and those with FEP, treatment with the antipsychotic aripiprazole was associated with decreased C4 plasma levels [98]. In our study, a small number of people (38 of 335 individuals with schizophrenia) were receiving aripiprazole; however, when we removed these individuals from our analyses, levels of plasma C4 between schizophrenia and the comparison group were still not significantly different (data not shown).
There are also more global limitations to consider about the emerging C4 model of schizophrenia. While this C4 model of schizophrenia has advanced the field technically and philosophically, substantial gaps in our knowledge of the pathophysiological steps that lead to schizophrenia are still present [33, 34, 42, 99]. For example, C4A genomic diversity and inferred C4 expression do not always equate with a risk of schizophrenia independently of the immune-gene rich major histocompatibility complex region or outside of European and African populations [55, 99, 100]. Furthermore, a role for synaptic pruning mechanisms attributable to C4 is predominantly based on experiments performed in rodent models, whose single C4 gene is not structurally homologous to the two-gene human C4 system [28, 34, 101] but see [52, 102]. C4A expression in the human brain was in part based on work that may have included diagnostically heterogeneous control groups. Big data investigations further indicate that schizophrenia risk may be preferentially conferred not by genes of the complement system but by synaptic pathways [103]. Strong associations of the complement system with measures of cognitive function may also operate independently of C4A allelic diversity [104].
At this research stage, it is currently unknown if plasma C4 is best considered a biomarker of a peripheral inflammation subtype or if it may have more far-reaching applicability to reflect CNS functions. The identification of autoantibodies that were associated with decreased plasma C4 measures serves as one example of how we may better balance population comparisons of plasma C4 by pre-identifying individuals with any evidence of autoimmune issues. Likewise, balancing population comparisons for pathogen exposures such as T. gondii, which were associated with plasma C4 levels in schizophrenia, may help further refine these analyses. Follow-up studies that directly examine C4 associations with T. gondii, with autoimmunity, and with the other significant variables found in Table 5 in larger cohorts of individuals with schizophrenia are certainly needed. Incorporating machine learning algorithms that model and predict outcome risks based on variable relatedness and interactions show promise for evaluating complex target such as plasma C4, but regression models are still a gold standard [105]. It is clear that the variation in plasma C4 derives from complex sources and that some variables contribute positively to plasma C4 levels and others contribute negatively. In this pilot study, we share a baseline collection of knowledge to assess if plasma C4 is transformable into an informative biomarker relevant to schizophrenia.
Statement of Ethics
As stated in the Materials and Methods section describing the study population, approval for these studies was granted by the Institutional Review Boards (IRB) of Sheppard Pratt and the Johns Hopkins Medical Institution following established guidelines. The Sheppard Pratt IRB Review Committee reviewed and approved two study protocols, with the approval numbers, 767141-23 and 332263-25, and approval dates, December 15, 2022, and May 11, 2023, respectively. The Johns Hopkins School of Medicine IRB-X Committee reviewed and approved one study protocol, approval number CR00042820/NA_00003294 and approval date, October 31, 2022. Written informed consent was obtained from all participants after study procedures were explained. This research was performed in accordance with The Code of Ethics of the World Medical Association (Declaration of Helsinki) for experiments involving humans.
Conflict of Interest Statement
Dr. Yolken is a member of the Stanley Medical Research Institute Board of Directors and Scientific Advisory Board. The terms of this arrangement are being managed by the Johns Hopkins University in accordance with its conflict of interest policies. None of the other authors report any potential conflicts of interest.
Funding Sources
This work was supported by the Stanley Medical Research Institute and ERA-NET NEURON grant ANR-18-0008-01. These funding sources had no role in the preparation of data or writing of the manuscript.
Author Contributions
E.G.S. conceived the idea for the paper. E.G.S., E.P., C.L.W., R.T., M.L., F.D., and R.H.Y. developed the conceptual and intellectual content. S.Y., F.L., and A.L. generated and managed the data. All authors helped interpret the data. E.G.S. and E.P. analyzed the data. E.G.S. wrote the first draft of the paper, and all authors participated in writing and reviewing subsequent drafts.
Data Availability Statement
The datasets analyzed for this study are available upon request in compliance with legal issues involving institutional data sharing regulations. Further inquiries can be directed to the corresponding author.