Convergence of autism proteins at the cilium

Abstract

Hundreds of high-confidence autism genes have been identified, yet the relevant etiological mechanisms remain unclear. Gene ontology analyses have repeatedly identified enrichment of proteins with annotated functions in gene expression regulation and neuronal communication. However, proteins are often pleiotropic and these annotations are inherently incomplete. Our recent autism functional genetics work has suggested that these genes may share a common mechanism at the cilium, a membrane-bound organelle critical for neurogenesis, brain patterning, and neuronal activity-all processes strongly implicated in autism. Moreover, autism commonly co-occurs with conditions that are known to involve ciliary-related pathologies, including congenital heart disease, hydrocephalus, and blindness. However, the role of autism genes at the cilium has not been systematically investigated. Here we demonstrate that autism proteins spanning disparate functional annotations converge in expression, localization, and function at cilia, and that patients with pathogenic variants in these genes have cilia-related co-occurring conditions and biomarkers of disrupted ciliary function. This degree of convergence among genes spanning diverse functional annotations strongly suggests that cilia are relevant to autism, as well as to commonly co-occurring conditions, and that this organelle should be explored further for therapeutic potential.

Figure 1:. Autism proteins are enriched at cilia.

(A) Enrichment of autism genes (Fu et al. 2022) and/or their protein-protein interactors (Wang et al. 2024) in three sets of cilia-related protein-protein interaction networks and one actin proteome (O’Neill et al. 2022; Gheiratmand et al. 2019; Thul et al. 2017). (B) Schematic of a ciliary basal body. (C) Enrichment of autism genes (Fu et al. 2022) within sub-ciliary structures of the post-mitotic neuronal centrosome protein-protein interaction network (O’Neill et al. 2022). DA: Distal Appendages (ODF2); SA: Subdistal Appendages (CEP170, ODF2); DL: Distal Lumen (POC5); CA: Cartwheel Assembly (CEP135); Proximal: proximal centriolar proteins (CEP62, CEP152); PCM: Pericentriolar Material (CEP192, CEP152, CDK5RAP2) from (O’Neill et al. 2022). Pericentriolar material (PCM) core protein (CENTROB) had 0/24 overlap and distal tip protein (CP110) had 0/85 overlap, so they were not included in the graph. (D) Schematic depicting localization screen in which 30 autism proteins with the strongest association with autism (FDR < 10⁻⁶, (Fu et al. 2022)) were tagged with GFP and overexpressed in motile cilia of Xenopus epidermis in vivo or primary cilia rat striatal neurons in vitro. Proteins listed were found to localize to cilia assessed by observation of GFP-tagged protein or endogenous antibody stain or both. (E) Human GFP-tagged SYNGAP1, STXBP1 localize proximally at the basal body (labeled by centrin-CFP, cyan), imaged live. See also Fig. S1–3 and Tables S1–3.

Figure 2:. Autism-associated chromatin regulators localize to and function at cilia.

(A) Human GFP-tagged constructs for autism-associated transcription factors SATB2 and RFX3 (green) localize to the nucleus (inset, deeper z-plane) and do not localize to motile cilia, while GFP-tagged constructs for autism-associated chromatin regulators ADNP, CHD8, CHD2, and POGZ do localize to motile cilia (labeled by acetylated alpha-Tubulin, magenta; basal bodies labeled by Centrin-CFP, cyan) when expressed in the X. laevis embryonic epidermis. Bottom panel shows basal bodies, actin, and autism-associated proteins in greater detail (actin network labeled by phalloidin, gray). (B) Human GFP-tagged constructs for POGZ, CHD8, and CHD2 (green) localize to primary cilia (cilia marked by Flag-tagged D1 receptor, magenta; basal bodies labeled by FGFR1OP antibody, red) when expressed in primary rat striatal neurons. (C) Endogenous staining for POGZ, CHD8, and CHD2 (green) in primary cilia (labeled by ARL13B, magenta; basal bodies labeled by FGFR1OP, red) in primary rat striatal neurons. (D) Morpholino-mediated knockdown (KD) of chd8, chd2, or pogz in X. tropicalis results in defects in cilia (labeled by acetylated alpha tubulin, magenta) on multiciliated cells of the embryonic epidermis. (E) Quantification of the data shown in B, with cilia.

Figure 3:. Autism-associated neuronal communication proteins localize to and function at cilia.

(A) Human GFP-tagged constructs for autism-associated ‘neuronal communication’ proteins (green) known to localize across neuronal subcompartments— presynaptic density, axon initial segment (AIS), and postsynaptic density—localize to cilia (axonemes labeled by acetylated alpha-Tubulin, magenta; basal bodies labeled by centrin-CFP, cyan) when expressed in X. laevis. NRXN1 localizes to the actin network surrounding cilia (labeled by phalloidin, gray). Bottom panel shows basal bodies and autism proteins in detail. H2B-GFP control localizes to the nucleus, in a deeper Z-plane as indicated in the inset “Nuclear Z-plane.” (B) Human GFP-tagged constructs for SLC6A1, SYNGAP1, PSD95, and SCN2A (green) localize to primary cilia (cilia labeled by Flag-tagged Dopamine D1 receptor, magenta; basal bodies labeled by FGFR1OP, red) when expressed in primary rat striatal neurons. (C) Endogenous staining for SLC6A1, SYNGAP1, PSD95, and SCN2A in primary cilia (labeled by ARL13B, magenta; basal bodies labeled by FGFR1OP, red) in primary rat striatal neurons. (D,E) Loss of syngap1, via CRISPR (syngap1^CR) or morpholino-mediated knockdown (syngap1^KD), results in cilia defects (labeled by acetylated alpha-Tubulin, magenta) and disrupts the apical actin network (labeled by phalloidin, gray) in the X. tropicalis multiciliated epidermis. (F) Quantification of syngap1^KD displayed in E as a percentage of embryos with no cilia phenotype vs. with a moderate or severe cilia phenotype. Fisher’s exact test (two-sided) followed by Pairwise Fisher’s Exact tests were used to calculate significance using raw counts of individual embryos. ** = p < 0.01, ns = no statistically significant difference. See also Fig. S8–11 and Tables S3 and S5.

Figure 4:. Patient-derived SYNGAP1 missense variants mislocalize away from cilia.

(A) Schematic of human SYNGAP1 protein (isoform alpha-2) with functional domains and locations of two autism-associated (Fu et al. 2022) patient-derived missense B variants annotated. (B-D) GFP-tagged human SYNGAP1 (green) localizes to the cilium (basal bodies labeled by centrin-CFP, cyan; axonemes labeled by acetylated alpha-Tubulin, magenta), while autism-associated variant SYNGAP1S140F often localizes away from the cilium at the membrane and SYNGAP1C233Y often localizes away from the cilium into aggregates. Actin network labeled by phalloidin (cyan). (E-G) GFP-tagged human non-mutant SYNGAP1 (green) localizes to the cilium when co-expressed with mCherry-tagged human non-mutant SYNGAP1 (E), but often localizes away from the cilium and to the membrane when co-expressed with mCherry-tagged autism-associated variant SYNGAP1S140F (F) or into aggregates when co-expressed with mCherry-tagged autism-associated variant SYNGAP1C233Y (G). mCherry proteins in white in the left panels. Right panels in 4G-I show GFP-tagged human non-mutant SYNGAP1 alone at higher magnification.

Figure 5:. Ciliary functional biomarker nasal nitric oxide is reduced in SYNGAP1 patients.

(A) SYNGAP1 is highly expressed in human airway ciliated cells and is significantly correlated with cilia markers FOXJ1 and ARL13B in the Tabula Sapiens epithelial dataset. (B) Nasal airway ciliated cells produce nitric oxide gas, but in the presence of cilia defects, this is greatly reduced. (C) Patients with de novo variants in SYNGAP1 have significantly lower nasal nitric oxide levels (p = 0.0015, one-sided Wilcoxon matched-pairs signed rank test, NIOX VERO nNO machine, tidal breathing method) compared to matched immediate family members (sibling or parent). (D) There is no relationship between family pair age difference and change in nitric oxide value (Simple Linear Regression, R² = 0.02078). ppb = parts per billion. See also Fig. S12.