C4A Copy-Number Variation, GRIN2A, and Synaptic Complement Pruning in Schizophrenia

Schizophrenia is a heritable neurodevelopmental disorder (h2 ≈ 0.65–0.80 in twin studies; SNP-based heritability h2_SNP ≈ 0.24) characterized by positive psychotic symptoms (hallucinations, delusions), negative motivational/affective deficits, and dorsolateral prefrontal cortex (dlPFC) cognitive impairment that typically manifests during late adolescence or early adulthood (ages 16–28). Convergent genome-wide association studies (GWAS across >76,000 cases in the Psychiatric Genomics Consortium PGC3), whole-exome sequencing (SCHEMA consortium across >24,000 cases), and human pangenome structural-variant assemblies demonstrate that polygenic and rare-variant risk architectures converge onto a unified neurobiological pathway: postnatal microglial complement-mediated synaptic pruning and excitatory glutamatergic postsynapse maturation.

At the population level, the single strongest common genetic association across the human genome maps to the Major Histocompatibility Complex (MHC) Class III locus on chromosome 6p21.32, driven by complex tandem copy-number variation (CNV) of the complement component 4 genes (C4A and C4B) and an intronic 6.36 kb human endogenous retrovirus (HERV-K(C4)) insertion that elevates brain C4A mRNA expression (OR ≈ 1.27–1.40 between extreme C4 structural haplotypes). At the rare coding variant level, heterozygous protein-truncating mutations in GRIN2A (Chr16p13.2, encoding the adult synaptic GluN2A subunit of the N-methyl-D-aspartate NMDA glutamate receptor) confer a ~24-fold increase in schizophrenia risk (OR ≈ 24.1), acting alongside common regulatory variants at voltage-gated calcium channels (CACNA1C) and the dopamine D₂ receptor (DRD2).

Tandem RCCX Copy-Number Modules and the HERV-K(C4) Retrotransposon

Within the MHC Class III region on chromosome 6p21.32, the complement C4 locus resides inside a polymorphic tandem segmental duplication unit termed the RCCX module (~30–36 kb per repeat), named for its four contiguous genes: RP1 (STK19), C4 (C4A or C4B), CYP21 (CYP21A2 or pseudogene CYP21A1P), and TNX (TNXB or TNXA). Because non-allelic homologous recombination (NAHR) frequently duplicates or deletes entire RCCX cassettes across human haplotypes, a single chromosome 6 typically carries one, two, three, or four tandem copies of the C4 gene, which collapse into unreadable reference alignments on linear GRCh38 without long-read human pangenome graph resolution.

Each copy of the C4 gene varies simultaneously along two independent structural axes:

1. C4A vs. C4B Isotype Specificity (Exon 26): Although C4A and C4B share >99% nucleotide sequence identity across 41 exons, four amino acid substitutions within a five-residue hexapeptide motif in Exon 26 (residues 1,101–1,106: PCPVLD in C4A vs. LSPVIH in C4B) alter the nucleophilic reactivity of the internal thioester bond (Cys991–Gln994) exposed upon C1s proteolytic cleavage into C4b. The PCPVLD motif of C4A reacts preferentially with free amino (-NH2) groups on protein antigens and neuronal synaptic membrane glycoproteins to form stable amide bonds, whereas the His1106 side chain of C4B (LSPVIH) catalyzes nucleophilic attack by hydroxyl (-OH) groups on bacterial carbohydrate surfaces to form ester bonds.
2. Long (L) vs. Short (S) Alleles (HERV-K(C4) Intron 9 Insertion): Approximately 76% of human C4 genes are Long (C4L, ~20.6 kb) because they harbor a 6.36 kb endogenous retrovirus (HERV-K(C4)) integrated in the antisense orientation within Intron 9, whereas Short (C4S, ~14.2 kb) alleles lack this retrotransposon.

Four common structural haplotypes—AL, AL-BL, AL-BS, and AL-AL—account for >90% of European and East Asian chromosomes. In post-mortem human brain tissue across five cortical regions, C4A RNA expression scales linearly with C4A copy number and is further increased by the presence of the HERV-K(C4) insertion (C4AL), such that each structural haplotype's predicted brain C4A expression level (and not C4B expression) correlates directly with its epidemiological odds ratio for schizophrenia.

The Classical Complement Cascade in Postnatal Circuit Refinement

During normal mammalian brain development, cortical and thalamic projection neurons initially overproduce excitatory glutamatergic dendritic spines during infancy and childhood, followed by an activity-dependent wave of synaptic pruning that eliminates weak, asynchronous, or underutilized synapses while stabilizing active ones. Whereas visual-system retinogeniculate pruning occurs during early postnatal life, layer III pyramidal neurons in the human dorsolateral prefrontal cortex (dlPFC) undergo prolonged synaptic pruning throughout late adolescence and early adulthood (ages 15–25)—coinciding with the peak age of onset for schizophrenia and the ~30% reduction in dlPFC layer III dendritic spine density consistently observed in post-mortem schizophrenia neuropathology.

This activity-dependent elimination is executed by the classical complement cascade and resident central nervous system phagocytes (microglia):

1. Synaptic Tagging (C1q and C4A): Inactive or weakly firing dendritic spines externalize phosphatidylserine (PS) or lose protective sialic acid/complement-regulatory (CD47/CSMD1) signals, recruiting C1q (C1q–C1r2–C1s2). Activated C1s cleaves astrocyte- and neuron-secreted C4A into C4b, exposing its reactive internal thioester (PCPVLD) to covalently crosslink (amide bond) directly onto synaptic membrane proteins.
2. C3 Convertase Assembly and Microglial CR3 Engulfment: Bound C4b recruits C2a to assemble the classical C3 convertase (C4b2a), which cleaves complement C3 into C3b/iC3b. Microglia expressing Complement Receptor 3 (CR3, the CD11b/CD18 or ITGAM/ITGB2 integrin heterodimer) recognize iC3b-opsonized dendritic spines and engulf them via phagocytic trogocytosis.

In humanized bacterial artificial chromosome (BAC) transgenic mice expressing human C4A versus C4B, C4A binds synaptic membranes far more efficiently than C4B and drives elevated C3 deposition, excessive microglial synaptic engulfment, reduced cortical spine density, and prefrontal working-memory deficits. Moreover, common schizophrenia GWAS hits independently identify CSMD1 (Chr8p23.2, encoding the brain-enriched CUB and Sushi Multiple Domains 1 protein, an endogenous inhibitor of C4/C3 convertase deposition at synapses), reinforcing the complement pruning pathway through balancing selection and polygenic epistasis.

Developmental GluN2B-to-GluN2A Switching and SCHEMA Exome Genetics

Why would excessive complement pruning or synaptic instability selectively affect the adolescent prefrontal cortex and trigger psychosis? The answer bridges the complement cascade to the glutamatergic NMDA receptor hypofunction hypothesis of schizophrenia (originally inferred from the observation that dissociative NMDA channel blockers such as phencyclidine PCP and ketamine reproduce positive, negative, and cognitive symptoms of schizophrenia in healthy adults).

Functional NMDA receptors are heterotetrameric ligand-gated ion channels composed of two obligate glycine/D-serine-binding GluN1 (GRIN1) subunits paired with two glutamate-binding GluN2 (GluN2A–D, encoded by GRIN2A–D) subunits. During embryonic and early juvenile brain development, cortical synapses predominantly express GluN2B (GRIN2B), which exhibits slow channel deactivation kinetics (τ_off ≈ 250–400 ms), permitting prolonged Ca²⁺ influx that favors structural plasticity and filopodial motility. During postnatal maturation and adolescence, cortical pyramidal neurons and fast-spiking parvalbumin-positive (PV+) GABAergic basket interneurons undergo a transcriptional switch from GluN2B to GluN2A (GRIN2A, Chr16p13.2, 1,464 aa), which accelerates NMDA receptor deactivation (τ_off ≈ 35–50 ms), anchors receptors to the postsynaptic density scaffold (PSD-95 / DLG4), and synchronizes cortical gamma-band (30–80 Hz) oscillations.

In the landmark Schizophrenia Exome Sequencing Meta-Analysis (SCHEMA, Singh et al., 2022) of 24,248 cases and 97,322 controls, GRIN2A emerged as one of the top 10 exome-wide significant genes (p = 7.3 × 10−10), with ultra-rare heterozygous protein-truncating variants (PTVs: nonsense, frameshift, and canonical splice-site mutations) conferring an OR ≈ 24.1 for schizophrenia, alongside common non-coding GWAS variants at the same locus (PGC3).

Circuit Convergence: PV+ Interneuron Disinhibition and Striatal Dopamine Hyperactivity

Heterozygous loss of GRIN2A—combined with polygenic variation in postsynaptic CACNA1C (Cav1.2 L-type calcium channels), GRIA3 (AMPA receptor subunit), and SHANK/DLG2 scaffolds—converges with C4A-mediated complement pruning via two circuit mechanisms:

1. Activity-Dependent Vulnerability to C4A/C3 Pruning: Synapses deficient in GluN2A-mediated Ca²⁺ influx and CACNA1C excitation–transcription coupling fail to stabilize their actin cytoskeleton (ARC/BDNF), marking them as "weak" synapses targeted by C1q–C4A–C3 for microglial elimination during adolescence.
2. Cortical PV+ Interneuron Disinhibition Driving Subcortical DRD2** Hyperdopaminergia:** Fast-spiking PV+ GABAergic interneurons rely heavily on GluN2A-containing NMDA receptors for excitatory drive. In Grin2a+/- mouse models, delayed maturation and hypofunction of cortical/hippocampal PV+ interneurons disinhibit glutamatergic pyramidal projection neurons, which in turn drive downstream hyperactivity of ventral tegmental area (VTA) and associative striatal dopamine (DA) neurons—explaining why cortical glutamatergic/complement pathology manifests clinically as striatal DRD2 hyperstimulation responsive to D₂ receptor antagonists.

Why do missense mutations in GRIN2A often cause childhood epilepsy-aphasia syndromes, whereas truncating LoF variants cause adolescent-onset schizophrenia?

Genotype–phenotype correlation across >400 clinical GRIN2A variants reveals a striking functional dichotomy. De novo gain-of-function (GoF) missense mutations clustered inside the transmembrane pore (M2–M4) and linker regions reduce Mg²⁺ voltage block or prolong channel opening, causing severe early-infantile developmental and epileptic encephalopathies. Conversely, protein-truncating loss-of-function (LoF) variants and ligand-binding domain (S1/S2) LoF missense variants cause milder childhood speech/rolandic epilepsy (such as Landau–Kleffner spectrum) that frequently resolves after puberty, followed by a high penetrance (OR ≈ 24) of schizophrenia in late adolescence when the prefrontal cortex switches to GluN2A dependence.

Why did the HERV-K(C4) insertion and multi-copy C4A alleles rise to high frequency in human populations despite increasing schizophrenia risk?

The MHC region on chromosome 6p21.3 evolves under intense pathogen-driven balancing selection and negative frequency-dependent selection. Because C4A covalently opsonizes protein antigens (such as viral capsids and immune complexes) via amide bonds whereas C4B targets carbohydrate cell walls via ester bonds, high-copy C4A alleles protect against systemic autoimmune diseases—specifically systemic lupus erythematosus (SLE) and Sjögren's syndrome, where C4A deficiency is a major genetic risk factor (OR > 2.0 for low C4A)—creating an evolutionary pleiotropic trade-off between peripheral immune clearance and central nervous system synaptic pruning.

How does CSMD1 (Chr8p23.2) act as an endogenous brake on C4A-mediated synaptic pruning?

CSMD1 (CUB and Sushi Multiple Domains 1) is a giant transmembrane protein (3,564 amino acids containing 14 CUB domains and 28 complement-control-protein Sushi repeats homologous to peripheral complement decay-accelerating factor CD55 and CR1) expressed selectively in developing cortical neurons and growth cones. CSMD1 binds activated C4b and C3b on neuronal membranes and serves as a cofactor for Factor I-mediated proteolytic cleavage of C4b/C3b, preventing excessive C3 convertase amplification on healthy dendritic spines; common intronic GWAS risk alleles in CSMD1 reduce its expression, removing this local brake on synaptic pruning.

1. Sekar, A., Bialas, A. R., de Rivera, H., et al. (2016). "Schizophrenia risk from complex variation of complement component 4." Nature, 530(7589), 177–183. https://doi.org/10.1038/nature16549
2. Singh, T., Poterba, T., Curtis, D., et al. (2022). "Rare coding variants in ten genes confer substantial risk for schizophrenia." Nature, 604(7906), 509–516. https://doi.org/10.1038/s41586-022-04556-w
3. Trubetskoy, V., Pardiñas, A. F., Qi, T., et al. (2022). "Mapping genomic loci implicates genes and synaptic biology in schizophrenia." Nature, 604(7906), 502–508. https://doi.org/10.1038/s41586-022-04434-5
4. Yilmaz, M., Yalcin, E., Presumey, J., et al. (2021). "Overexpression of schizophrenia susceptibility factor human complement C4A promotes excessive synaptic loss and behavioral changes in mice." Nature Neuroscience, 24(2), 214–224. https://doi.org/10.1038/s41593-020-00763-8