Evaluation of C4 Gene Copy Number in Pediatric Acute Neuropsychiatric Syndrome

Pediatric acute-onset neuropsychiatric syndrome (PANS) is an abrupt-onset neuropsychiatric disorder that is characterized by obsessive-compulsive symptoms (OCS) and/or eating restriction with severe functional debilitation [1, 2]. Among children meeting criteria for PANS, a subset is diagnosed with pediatric autoimmune neuropsychiatric disorders associated with streptococcal infections (PANDAS) [3]. PANS patients have an increased prevalence of comorbid autoimmune illness, most commonly arthritis; a family history of autoimmune and psychiatric illness is also common [4‒6].

PANS/PANDAS are hypothesized to be inflammatory brain diseases that involve the basal ganglia [7]. Autoantibodies in sera from patients were found to preferentially bind cholinergic interneurons in human and mouse brain tissue in vitro [8, 9]. In a mouse model of PANDAS, there is neuroinflammation in the basal ganglia [10]. Similarly, brain imaging studies in PANS patients demonstrate differences in the basal ganglia (compared to controls), which likely reflects inflammation or injury [11, 12]. However, laboratory biomarkers specific for PANS have remained elusive [13]. Very large cohorts of PANS patients that would be needed to perform genome-wide association studies are not available. Whole exome sequencing studies and immunophenotyping studies are ongoing.

Decreased C4 protein concentrations have been reported in patients with PANS; a possible explanation is decreased production of C4 protein due to a low number of C4 gene copies (copy number [CN]) [1, 14]. Differences in peripheral C4 protein concentrations are often observed in autoimmune and psychiatric disorders, with disease risk correlated with differences in the CN of the A and B forms of the C4 gene [15‒18]. For example, reduced CN of the C4B gene is associated with an increased risk of rheumatoid arthritis, while higher C4A CN increases risk of schizophrenia [19, 20].

The C4 gene is found on the short arm of chromosome 6 in the class III region of the major histocompatibility complex (MHC) [21], a site of high variation and disease association [22]. There are two forms of the C4 gene: A-type and B-type, which differ by four amino acids due to five nucleotide polymorphisms [23, 24]. The C4A and C4B genes encode the C4A and C4B proteins, which have different binding properties and functions [24‒27]. Short- and long-forms of C4A or C4B are differentiated by absence or presence of an insertion of an endogenous retrovirus, HERV-K in intron 9 [28, 29]. Therefore, there are 4 total forms of the complement 4 genes: C4AS, C4AL, C4BS, and C4BL, where S stands for the short form of the gene that does not contain the HERV insertion and the L form of the gene is the long form, which contains the HERV insertion (Fig. 1a). On average, humans have four copies of complement 4 genes: 2 copies of the A-form and 2 copies of the B-form [30].

The high rates of autoimmune disease (AI), arthritis, psychiatric illness, and decreased complement protein concentration in PANS patients led us to hypothesize that either a high CN of C4A or low CN of C4B predicts PANS illness and/or time to diagnosis with juvenile idiopathic arthritis (JIA) or AI. To address this hypothesis, we characterized C4 CN in controls and in PANS/PANDAS patients in the largest collection of DNA samples from PANS/PANDAS patients to date. Group comparisons of mean total C4A and total C4B CN in ethnically matched PANS/PANDAS patients and controls were performed. Then, longitudinal clinical data from the Stanford PANS subset were used to determine the time to JIA onset and AI onset as a function of total C4A or C4B. Lastly, several hypothesis-generating analyses were performed to explore the correlation between individual C4 gene variants, sex, specific genotypes, and age of PANS onset (AO).

Only European ancestry (EA) individuals were included in this study.

DNA samples from participants meeting strict clinical criteria for PANS (or PANDAS in the case of the NIH cohort) were collected from several study sites from groups with longstanding expertise in PANS/PANDAS: Stanford University, the National Institutes of Health, the University of South Florida, and the University of Gothenburg. Details of patient recruitment, enrollment, and evaluation at each study site are provided in the online supplementary Materials (for all online suppl. material, see https://doi.org/10.1159/000531707). Exclusion of samples were based on ancestry, being related to another study participant, availability of DNA material, and successful C4 genotype characterization (online suppl. Materials).

All of the Stanford PANS patients in this study were followed longitudinally, with clinical notes and laboratory data available for study [5]. Clinical laboratory results of serum C3, C4, and plasma C4a were obtained through chart review. AO of JIA and AI were determined by a systematic chart review, performed between December 2, 2019 and August 1, 2020. AI diagnosis was based on clinical criteria. JIA diagnosis was based on diagnostic criteria from the International League of Associations for Rheumatology or assessment criteria of the Spondyloarthritis International Society [5].

We utilized C4 CN data from a random subset of psychiatrically healthy control participants from the Genomic Psychiatry Cohort (GPC) [31]. Additionally, healthy control samples obtained by the Stanford PANS program were recruited from the same community as the patient population (online suppl. Materials). Samples were excluded if there was evidence of possible OCD symptoms from the donor or a technical issue in the C4 genotype characterization (online suppl. Fig. 1).

DNA was extracted by standard methods from whole blood samples. A droplet digital PCR assay with specific probes for detection of the forms of the C4 genotype was performed, as previously described [20].

The participants selected for this analysis are detailed in the online supplementary Materials. The primary analysis in this study examined the difference in the mean total C4A and C4B CN between PANS/PANDAS patients and control groups CN. Population studies of both C4A and C4B CN in EA populations follow a normal distribution [18, 32]. Therefore, a two-tailed, non-paired Student’s t test was used to test the primary hypotheses. The power to detect a difference of 0.4 mean gene copies is 92% with the sample size of 182 PANS patients and 192 controls. The difference in total C4 (a sum of total C4A and total C4B for each participant), individual variants (e.g., C4AL, etc.), and group differences segregated by sex were analyzed in exploratory analyses (see below). Samples were also analyzed by segregation into their cohort of origin to check for differences that may be driven by individual cohorts (online suppl. Materials).

The time to JIA onset and time to AI onset were analyzed for the Stanford PANS cohort, using a Cox proportional hazard regression model separately for total C4A and total C4B as independent variables. Patients were classified into those with (a) C4A <2, C4A = 2, and C4A >2, and (b) C4B <2, C4B = 2, and C4B >2. Hazard ratios (HRs) were calculated comparing (a) C4A <2 to C4A = 2 and C4A >2 to C4A = 2 and, (b) C4B <2 to C4B = 2 and C4B >2 to C4B = 2. AO of PANS, JIA, and AI were compared to determine whether PANS diagnosis preceded JIA and/or AI diagnoses.

Bonferroni correction was used to correct for multiple tests. A total of four tests were performed in this study to test the primary hypotheses: (1) comparing C4A CN in PANS and controls, (2) C4B CN in PANS and controls, (3) time to JIA onset, and (4) time to AI Onset. An α value of 0.025 was required for significance.

To determine if there is a sex-dependent difference in C4 gene copies between groups, a two-tailed non-paired Student’s t test was applied separately to mean C4 gene CNs (total C4A and total C4B) between female and male EA patients and control groups. To examine group means of CNs of C4AL, C4AS, C4BL, and C4BS between patient and control groups as a whole and segregated by sex, a two-tailed non-paired Student’s t test was applied to mean C4 gene CNs (C4AS, C4AL, C4BL, C4BS) between patients and controls. Then, cases and controls were analyzed for males and females separately. Samples were also analyzed by segregation into their cohort of origin to check for differences that may be driven by individual cohorts (online suppl. Materials).

The genotype frequencies were then compared between PANS/PANDAS and controls, using a Fischer’s exact test.

To explore whether there is a relationship between the number of C4A or C4B gene copies and PANS/PANDAS AO, a Pearson correlation coefficient was computed between the PANS AO of each case and the number of gene copies in total C4A and total C4B.

Three hundred nine patient study participants were recruited across the four study sites. Participants were excluded from the analysis if: C4 droplet digital PCR assay results were not available, they were not of EA, or they were a sibling of another participant. One hundred ninety two patient participants were used for analysis in this study (Fig. 1a). The PANS/PANDAS group is male predominant: 124 (65%) of the patients were male and 68 were female (Fig. 1b).

All of the PANS patients in the Stanford cohort (n = 121 patients) were used in the time to JIA or AI analysis. Of the Stanford PANS participants included in this study: 29% were diagnosed with JIA, and 13% were diagnosed with AI disease. Ninety seven percent of PANS patients with JIA received a PANS diagnosis prior to JIA diagnosis. JIA diagnoses included enthesitis-related arthritis, spondyloarthritis, and psoriatic arthritis. Seventy six percent of PANS patients with AI received a PANS diagnosis prior to AI diagnosis. The most common AI diagnoses included were autoimmune thyroiditis, psoriasis, inflammatory bowel disease, and celiac disease. Thirty five percent of the Stanford PANS cases had a serum C4 protein below the normal range; 6% had a low serum C3 protein. Sixty nine percent of subjects with available clinical data (51 of 73 cases with available plasma C4a on clinical testing without excluding non-EA) had a plasma C4a level above the normal range on clinical testing (Fig. 1a).

A total of 233 control participants were recruited for inclusion in this study: (a) 45 from the Stanford study site and (b) 188 from the GPC. (a) In the Stanford group, participants with psychiatric, neurodevelopmental, or medical conditions were excluded upon screening (J.F. and M.T.). A total of 27 non-EA participants were excluded from this analysis. C4 CNV assay validation failed in 1 participant. (b) A subset of the GPC cohort was randomly selected for C4 CNV characterization. One hundred eighty-five were from EA and included in this study. Seventeen samples were excluded due to positive answers to questions regarding OCS on questionnaires. C4 CNV assay validation failed in 6 participants. A total of 182 control participants were analyzed from both cohorts (Fig. 1a). Forty-six percent of the patients were male (Fig. 1b). Additional details are provided in Online Supplementary Materials.

No differences in the mean total C4A or total C4B CN were found in PANS/PANDAS patients compared to controls as a group (Fig. 1c).

Stanford PANS cohort was classified into three C4A CN groups: <2 (n = 16), = 2 (n = 75), >2 (n = 30) and three C4B CN groups: <2 (n = 38), = 2 (n = 78), >2 (n = 5). Cox regression analysis demonstrates a statistically significant increase in risk for developing JIA in patients with PANS who had less than two copies of C4B compared to having two copies (HR = 2.7, 95% confidence interval: 1.4–5.2, p value = 0.004, survives Bonferroni correction for multiple comparisons, Fig. 2a, b). We also find a trend in risk for developing AI after PANS diagnosis with less than two copies of total C4B compared to having two copies (HR = 2.8, 95% confidence interval: 1.0–7.7, p value = 0.05, which does not survive Bonferroni correction for multiple comparisons, Fig. 2b). None of the other comparisons demonstrate differences in risk (Fig. 2b).

No differences were found in the number of gene copies of the C4 variants (C4AS, C4AL, etc.) in PANS/PANDAS compared to controls, either as a group or when separated by sex (Fig. 3). No differences were found in C4 variants, total C4, total C4A/B, either as a group or when separated by sex when samples were segregated by cohort of origin (online suppl. Fig. 2).

We found 6 genotypes that occurred >5% in the population (Fig. 4a). None of the comparisons were found to be different between PANS/PANDAS and controls, except for a possible nonsignificant trend in genotype: 2C4AL-2C4BL for females with PANS/PANDAS (Fig. 4b–d).

Average AO in our cohort is 8.9 +/− 3.3 years (Fig. 5a), similar for males and females (Fig. 5b). In PANS cases, C4A CN did not predict AO, but a weak relationship was observed between lower C4B CN and earlier AO of PANS (r = 0.12, nominal p = 0.09, Fig. 5c, e), primarily in males (r = 0.21, nominal p = 0.02, Fig. 5d–e). These correlations between C4B CN and AO of PANS were nonsignificant after Bonferroni correction for multiple comparisons.

We report the results of the first study of C4 CN in the largest collection of DNA from individuals with PANS/PANDAS. Our PANS/PANDAS patient cohort is predominantly male, consistent with prior reports [4, 33‒35]. We did not find a difference in the mean total C4A or C4B CN in PANS/PANDAS compared to controls. However, we found an increase in risk for subsequent diagnosis of JIA in PANS patients with low C4B CN. Furthermore, we observe a possible increase in risk of AI in PANS patients and, in the exploratory analysis, a possible correlation between lower C4B and PANS AO.

While we do not find an association between C4A or C4B CN and PANS, we do find an increase in risk of developing JIA in PANS patients with low C4B CN. Deficiencies of C4A and C4B have been reported in JIA cases [36]. An association between rheumatoid arthritis and low C4B CN has been reported previously [19]. However, patients with PANS develop arthritides that are thought to be biologically distinct from rheumatoid arthritis; specifically, enthesitis-related arthritis, spondyloarthritis, and psoriatic arthritis [5]. Complement proteins are found in synovium in patients with RA and play an important role in bone repair and homeostasis [37, 38]. However, whether complement plays a role in joint development or homeostasis is not known. It is possible that the increased risk for JIA in PANS patients with low C4B CN may be reflecting a more general mechanism by which low C4B confers risk for arthritis.

We also observe a possible increased risk for PANS patients to develop other AI with low C4B CN, and our exploratory analyses suggest possible correlation between lower C4B CN and a younger AO of PANS/PANDAS, especially in males. It remains to be seen whether this association reflects the association between PANS and arthritis rather than PANS illness itself.

We observe low C4 protein in an approximately one-third of the PANS cohort (Fig. 1a). The number of C4 gene copies, including the number of C4A CN and C4B CN, determines overall C4 protein expression [32]. Since the number of C4A or C4B CN is not different in PANS patients, low C4A or C4B CN cannot explain the low serum C4 protein observed in PANS patients. However, the almost two-thirds rate of elevated C4 protein activation product (high plasma C4a, Fig. 1a) in addition to the low C4 protein point to activation of the C4 protein [39]. A major caveat to the plasma C4a concentration reported here is that it was not available for all of the analyzed Stanford PANS cases. Therefore, plasma C4a concentration is being measured in controls and PANS patients in research collections of plasma by our group to test our hypothesis of C4 protein activation in PANS patients.

Our study of C4 CN variation in PANS/PANDAS has several limitations. We combine several cohorts in each group (patients and controls). Also, the psychiatric GWAS cohorts were adults, whereas the Stanford control cohort and all the PANS/PANDAS cohorts were children. However, we argue that the fact that DNA was collected from adults who met criteria as “controls” means that they are more likely to be without an OCS-related psychiatric condition. The school-age controls may develop a psychiatric illness in adolescence or adulthood when most psychiatric disorders arise. Furthermore, all the patients and the controls in the Stanford cohort completed extensive screening questionnaires and were interviewed by the research team. In the psychiatric GWAS cohorts, the clinical information was gathered using self-administered questionnaires, and no information regarding medical comorbidities is available. Therefore, the control cohort may include individuals with comorbid autoimmune conditions or individuals with psychiatric illness. The time to JIA and AI are only data from the Stanford cohort, which has longitudinal study data. However, the longitudinal data are limited to a pediatric population and do not capture the potential development of subsequent JIA or AI in adulthood.

We speculate that C4B deficiency in PANS could be contributing to autoimmunity through defective pathogen clearance and the production of autoantibodies [27]. Functional studies of C4B proteins found that C4B binds to hydroxyl group-containing substrates (e.g., bacterial cell walls) [40]. Group A streptococcal (GAS) infection often precedes PANS onset, and basal ganglia encephalitis is induced by repeated GAS infection in a mouse model. These observations support the hypothesis that lack of GAS clearance contributes to PANS pathogenesis [3, 10]. Children are commonly exposed to GAS. Furthermore, associations between bacterial infections and enthesitis-related arthritis, spondyloarthritis, and psoriatic arthritis, have been observed [41]. Thus, autoimmunity associated with persistence of GAS pathogenic material from C4B deficiency may prove to be a triggering mechanism for PANS and/or subsequent JIA.

We thank the investigators in that study (D.F.L., A.E.U., Michael Mindrinos, and Emmanuel Mignot) for making these genotypes available for the present analyses, and we thank the role of Julie Wilhelmy in carrying out the C4 genotyping for that study. We thank the GPC collaboration (Carlos Pato, P.I.) for allowing the HLA study to obtain GPC DNA samples in advance of public availability from the NIMH repository program, and for authorizing us to use C4 genotypes from GPC controls for the present study. We thank Professor Chiara Sabatti for helpful feedback on this manuscript. J.F. and M.T. would like to further thank current and former members of their research staff (listed on our website), the Immune Behavioral Health Clinicians and Staff, the Stanford PANS Basic Science Team, the many collaborating physicians, and the Stanford PANS Parent Action Committee. All authors want to thank the patients and families who are understanding of our limitations in clinical treatments and continue to lend us their time, energy, and cooperation with research participation and recruitment efforts. Figures are created with BioRender.com.

PANS genomic samples: use of the de-identified DNA specimens and their diagnostic status for this work was authorized (protocol number 26922) by the Stanford Institutional Review Board. Healthy controls from the GPC: use of specimens from the Molecular Genetics of Schizophrenia (MGS) cohort for this work was authorized (protocol IRB-3460) by the Stanford University IRB. Specimens were provided by the NIMH Repository and Genomics Resource (NRGR). De-identified DNA specimens and diagnostic status from the Genomic Psychiatry Cohort (GPC) were obtained for this work from NRGR, which was responsible for ensuring the appropriateness of the original human subject approvals for the GPC project. Written informed consent was obtained from participants (or their parent if under the age of 18) by the individual study sites for study participation.

The authors have no conflicts of interest to declare.

C4 genotypes were assayed for the GPC control samples by the NIMH-funded study HLA and schizophrenia: a high-throughput sequencing study (1R01MH096262 to D.F.L.). A.E.U. received funding from Mr. Bruce Blackie and from Dr. William McIvor. D.F.L. received support from the Walter E. Nichols MD Professorship in the School of Medicine, from Dr. William McIvor, and from the Stanford Schizophrenia Genetics Research Fund from an anonymous donor. A.K. was supported by fellowship funding from the Sierra Pacific MIRECC Fellowship. J.F. was supported by the National Institute of Mental Health (Pediatrics and Developmental Neuroscience Branch), the Dollinger Family PANS Biomarker Discovery Core, Caudwell Children, the PANDAS Physician Network (PPN), the Global Lyme Alliance, the Foundation for Children with Neuroimmune Disorders, the Stanford Maternal Child Health Institute, Stanford SPARK, and the Lucile Packard Foundation for Children’s Health. E.F., M.J., and C.G. are supported by the Swedish Brain Foundation (Hjärnfonden) for their studies in PANS.

A.K., J.F., E.M., and A.E.U. conceived the study design and methods. A.K. and L.T. performed the statistical analyses. R.P. and A.K. performed the assay to determine the C4 genotypes in the patient cohorts. H.O. and D.L. analyzed the C4 genotypes from the control cohorts. J.F., M.S., M.M., L.C., B.F., S.S., T.M., M.J., E.F., C.G., and M.T. recruited, enrolled, evaluated, and collected samples from participants. M.K., C.M., L.C., and J.F. assisted in data management and analysis. A.K., E.M., J.F., and D.L. wrote the manuscript. All authors provided input to the manuscript.

All data generated or analyzed during this study are included in this article and its online supplementary material files. Further inquiries can be directed to the corresponding authors. Some restrictions may apply because this study includes data collected from multiple studies.