Pathogenic UNC13A variants cause a neurodevelopmental syndrome by impairing synaptic function

Abstract

The UNC13A gene encodes a presynaptic protein that is crucial for setting the strength and dynamics of information transfer between neurons. Here we describe a neurodevelopmental syndrome caused by germline coding or splice-site variants in UNC13A. The syndrome presents with variable degrees of developmental delay and intellectual disability, seizures of different types, tremor and dyskinetic movements and, in some cases, death in early childhood. Using assays with expression of UNC13A variants in mouse hippocampal neurons and in Caenorhabditis elegans, we identify three mechanisms of pathogenicity, including reduction in synaptic strength caused by reduced UNC13A protein expression, increased neurotransmission caused by UNC13A gain-of-function and impaired regulation of neurotransmission by second messenger signalling. Based on a strong genotype-phenotype-functional correlation, we classify three UNC13A syndrome subtypes (types A-C). We conclude that the precise regulation of neurotransmitter release by UNC13A is critical for human nervous system function.

Fig. 1. Landscape of UNC13A patient variants identified in this study.

a, Location of validated pathogenic variants identified in 20 patients (Supplementary Data 1) mapped on a schematic representation of the UNC13A gene (NM_001080421.2). Variants with biallelic inheritance are in magenta (p.T117Rfs*18 and p.R202H, and c.767+1 G > T and c.4073+1 G > A are in a compound heterozygous state; others are in a homozygous state), heterozygous de novo variants are in black and a heterozygous, inherited variant is in brown. Variants p.G808D and p.P814L were detected in four and six patients, respectively. Exon colors correspond to the color of protein domains in b. b, Validated pathogenic missense variants overlaid on a schematic representation of the UNC13A protein. The sequence of the hotspot region, termed here as the ‘UNC13 hinge’, is magnified. c, UNC13A tolerance landscape generated based on the observed missense and synonymous variants in gnomAD and corrected for the sequence composition (MetaDome database). The degree of tolerance for missense variants is color-coded (lowest scores, red, highly intolerant; highest scores, blue, highly tolerant). d, AlphaMissense scores plotted for all missense variants (pathogenic, of uncertain significance (VUS) and (likely) benign) reported in this study (Supplementary Data 1; heterozygous de novo missense variants, bold typeface; heterozygous inherited or biallelic variants, non-bold typeface; validated pathogenic variants, red; VUS, gray; validated (likely) benign, blue). Notably, all validated pathogenic variants reported in this study are within the pathogenic AlphaMissense score range, and three out of four validated benign variants fall within the benign score range. We consider VUS variants in the pathogenic AlphaMissense score range as hot VUS (14 variants) and those within the benign/ambiguous score range as cold VUS (12 variants). e, Left: age of patients (with median and interquartile range; error bars, s.d.) at the latest physician visit, stratified by inheritance pattern. Middle: major clinical features of 20 patients with pathogenic variants (Supplementary Data 1), stratified by inheritance pattern. The presence or absence of features is shown by the indicated colors together with the respective percentage for different inheritance patterns. Right: standard deviation scores for growth parameters of patients with pathogenic variants at birth and at the latest visit are illustrated by inheritance pattern (HC, head circumference). Source data

Fig. 2. UNC13A disease-causing variants change UNC13A expression levels at synapses.

a, Location of studied variants within the UNC13A protein domains (not drawn to scale). b, Schematic representation of the autaptic cell culture system used in this study. c, Left: example images of cultured autaptic hippocampal neurons from Unc13a/b DKO mice that were transfected on DIV 2–3 with UNC13A–GFP cDNA encoding WT or disease-related variants and stained with antibodies against the presynaptic marker VGLUT1, the postsynaptic marker Shank2, the dendritic marker MAP2 and with an antibody against GFP (n = 24–37). Right: magnification of the regions indicated by the white boxes in the merged image. d, Quantification of VGLUT1–Shank2 juxtapositioned puncta per neuron. NT, not transfected. e, Fraction of VGLUT1–Shank2 juxtapositioned puncta where UNC13A–GFP co-staining was detected. Bars and circles in d and e represent means and values for individual neurons, respectively; error bars, s.e.m. The P values in e refer to pairwise comparisons of mean fractions with WT using Dunn’s multiple comparison test. f, Cumulative probability distributions of UNC13A–GFP intensities at VGLUT1–Shank2 juxtapositioned puncta, reporting relative UNC13A expression levels at individual synapses. For each cell analyzed, the mean Munc13 intensity over all synapses was obtained and, subsequently, the pooled mean was calculated for each genotype (Supplementary Data 2). a.u., arbitrary units. The P values in f refer to pairwise comparisons of mean intensities with WT using Tukey’s multiple comparisons test. See Supplementary Data 2 for further details. Source data

Fig. 3. Amino-terminal disease variants abate synaptic strength.

a, Amino acid sequence alignment of the C2A domain and adjacent region for UNC13A homologs in different species (disease variants are highlighted by a bar; E52K, orange; R202H, green; HS, Homo sapiens; RN, Rattus norvegicus; CE, C. elegans). b, Structure of the UNC13A C2A domain (orange) in complex with the RIM zinc-finger domain (blue) (PDB 2CJS). E52 is shown in a space-filled presentation. c, Pictures of three patients with pathogenic homozygous amino-terminal variants, showing narrow forehead, highly arched eyebrows, impression of hypertelorism, depressed nasal bridge, small nasal tip and underdeveloped, anteverted ala nasi, long philtrum, impression of retrognathia and large, mildly forward-facing earlobes. d, Representative pedigrees of patients with pathogenic biallelic amino-terminal variants, illustrating an autosomal recessive pattern of inheritance with apparently healthy carrier parents. e, Schematic illustration of electrophysiological recordings in autaptic neurons. f, Constructs used in electrophysiological experiments in g–m (NLS; nuclear localization signal). g–m, Characterization of synaptic transmission in autaptic hippocampal neurons expressing UNC13A^WT (black), UNC13A^E52K (orange) or GFP-NLS (gray). Example traces (g) and summary data showing mean frequencies (h) of mEPSCs. Example traces (i) and summary data showing mean charges (j) of eEPSCs. Example traces (k) and summary data showing mean charges (l) of sucrose-evoked EPSCs as a measure of RRP size. No eEPSCs were elicited in UNC13A^E52K-expressing neurons during high-frequency AP trains (40 Hz) (m). n, The constructs used in electrophysiological experiments in o–v. Example traces (o), mean amplitudes (p) and mean frequencies (q) of spontaneously occurring mEPSCs in neurons expressing UNC13A^WT (black) and UNC13A^R202H (green). r,s, Example eEPSC traces (r) and mean eEPSC charge (s). Smaller sucrose-evoked EPSCs (t) indicate a reduced RRP size in UNC13A^E52K-expressing neurons (u), while the mean vesicular release probability was unchanged (v). Data were obtained from at least three independent cultures per condition. Statistical analysis: Kruskal–Wallis test (g–m) and two-tailed Mann–Whitney test (o–v). In h, j, l, p, q, s, u and v, bars and circles represent means and values for individual neurons, respectively; error bars, s.e.m. See Supplementary Data 3 for further details.

Fig. 4. The UNC13A hinge region is a disease hotspot.

a, A sequence alignment of UNC13A homologs illustrating that the UNC13 hinge domain is conserved (residues found mutated in patients in bold). b, Facial portraits of patients with pathogenic de novo hinge variants, displaying high anterior hairline with an appearance of tall forehead, impression of hypertelorism, deeply set eyes, attached earlobes, smooth, rather short philtrum, thin upper lip vermilion, everted lower lip vermilion, bulbous nasal tip and retrognathia. c, Representative pedigrees of patients with pathogenic de novo variants, illustrating an autosomal dominant pattern of inheritance. d, Crystal structure of the UNC13A (PDB 5UE8) identifies the UNC13A hinge as non-structured (purple), connecting two structured domains. e–l, Characterization of synaptic transmission in autaptic hippocampal neurons expressing UNC13A^WT (gray) or UNC13A^G808D (blue). Example traces (e) and summary data showing mean amplitudes (f) and mean frequencies (g) of mEPSCs. Example traces (h) of sucrose-evoked EPSCs and summary plots of the measured RRP size (i). Example traces (j) of eEPSCs and summary plots of eEPSC charge (k) and vesicular release probability (l). m, Schematic illustration of the aldicarb assay in C. elegans. Aldicarb-induced accumulation of ACh in the extrasynaptic space leads to receptor desensitization and to worm paralysis. Synaptic phenotypes associated with ACh hypersecretion lead to faster ACh accumulation and to faster paralysis. Reduced secretion results in slower paralysis. n, CRISPR-generated heterozygous (green) and homozygous (blue) knock-in UNC-13^P814L (UNC-13^P956L) worms exhibit faster aldicarb-induced paralysis than WT worms. o, Similar results obtained in heterozygous (green) or homozygous (blue) UNC-13^G808C (UNC-13^G950C) knock-in worms. Statistical analysis was performed by using a two-tailed Mann–Whitney test (f, g, i, k, l) or Kruskal–Wallis test followed by Dunn’s test (n, o). Bars and circles in f, g, i, k and l represent means and values for individual neurons, respectively; error bars, s.e.m. Data were obtained from at least three independent cultures per condition. Circles in n and o represent averaged data from >10 worms. See Supplementary Data 3 for further details.

Fig. 5. UNC13 disease variations change the sensitivity of synaptic transmission to second messenger regulation by the DAG pathway.

a, Sequence alignment for UNC13 paralogs highlighting H554 (blue) and C587 (bold). Lines below indicate Zn²⁺-coordinating residues. b, Facial portraits of index patient 6, his father and aunt, carrying the pathogenic variant, displaying a high, broad forehead, impression of hypertelorism, deeply set eyes, fullness of the upper lateral eyelid, depressed nasal bridge, underdeveloped ala nasi and small nasal tip, smooth, broad philtrum and everted vermilion of the upper lip. c, A four-generation pedigree, with four individuals confirmed as harboring c.1760 G > T, p.(C587F). d, The C1 domain structure (top) and a zoomed-in view of one Zn²⁺-binding pocket. The corresponding amino acid numbers in the rat UNC13A are given in brackets. e, HEK293FT cells transiently transfected with UNC13A–GFP constructs encoding WT or patient variants, treated with 1 µM PDBu for 1 h (scale bar, 5 µm; n = 10–15 per condition; white arrowheads highlight UNC13A location (green)). f, Quantification of the cytosolic GFP signal in the presence or absence of PDBu. g–n, Characterization of synaptic transmission in autaptic hippocampal neurons expressing UNC13A^WT (gray) or UNC13A^C587F (red). Example traces (g) and summary plots showing mean amplitude (h) and mean mEPSC frequency (i). Example traces (j) and summary plots showing mean eEPSC charge (k). Example traces (l) and summary plots showing mean sucrose-evoked EPSC charge (m). n, Quantification of vesicular release probability. o, Time to paralysis in CRISPR-targeted C. elegans worms carrying a homozygous C587F (C729F) variation and exposed to aldicarb. p,q, Example eEPSCs before and after application of PDBu in neurons expressing UNC13A^WT (p) or UNC13A^C587F (q). r, The average time courses of eEPSC amplitude change. s,t, Similar data for UNC13A^G808D (s, blue) or UNC13A^R202H (t, green). In o and r–t, circles represent an average from >10 worms or an average of the indicated number of neurons. Bars and circles in f, h, i, k, m and n represent means and values for individual neurons, respectively. Data were obtained from at least three independent cultures per condition; error bars, s.e.m. Statistical analysis was performed by using a two-tailed Mann–Whitney test (f, h, i, k, m, n) or Wilcoxon test (o). See Supplementary Data 3 for further details.

Fig. 6. UNC13A disease variants modulate neuronal short-term synaptic plasticity.

a–d, Stimulation protocol (a) and average time courses of normalized eEPSC amplitudes recorded in autaptic hippocampal neurons expressing UNC13A^WT (gray) or UNC13A^R202H (b, green), UNC13A^C587F (c, red) or UNC13A^G808D (d, blue), before, during and after a 10 Hz train consisting of 30 APs. e–h, Average time courses of normalized eEPSC amplitudes recorded before, during and after a 40 Hz train consisting of 40 APs (e) in neurons expressing UNC13A^WT (gray) or UNC13A^R202H (f, green), UNC13A^C587F (g, red), or UNC13A^G808D (h, blue). Gray arrowheads mark the time point during which steady-state depression was quantified, and black arrowheads mark the first eEPSC following the train, used to estimate the magnitude of augmentation. i, Replenishment rate constants calculated based on a single-pool model of SV priming for neurons of the indicated genotypes. j,k, Magnitude of eEPSC augmentation after conditioning 10 Hz (j) or 40 Hz (k) stimulation. Data were obtained from at least three independent cultures per condition. Statistical analysis was performed by using a two-tailed Mann–Whitney test (i–k). Bars and circles in i–k represent means and values for individual neurons, respectively; error bars, s.e.m.

Extended Data Fig. 1. Facial photographs of six patients from this study harboring compound heterozygous (CH) or de novo (DN) UNC13A variants of uncertain significance (VUS).

See details in Supplementary Table 1.

Extended Data Fig. 2. Clinical and/or genetic information for patients 3 and 4.

a, Photographs of patient 3 harboring a pathogenic homozygous variant c.523+6 T > C in intron 7 of UNC13A. b, UNC13A c.523+6 T > C variant exhibits a strong, deleterious splicing impact leading to exon 7 skipping (a.u., arbitrary units). c, d, Neurophysiologic findings at Nasalis muscle in repetitive nerve stimulation (RNS) recordings. c, Decrementing response (19.7%) induced by low-frequency (3 Hz) repetitive nerve stimulation. d, Incrementing response (17.8%) induced by high-frequency (20 Hz) repetitive nerve stimulation. The findings are compatible with a compromised presynaptic neuromuscular junction function. e, A three-generation pedigree illustrating two affected cousins, patients 4a and 4b, harboring identical pathogenic homozygous variants p.(R202H) in UNC13A.

Extended Data Fig. 3. Minigene in vitro analysis of three UNC13A intronic variants.

a, b, UNC13A c.767+1 G > T (intron 9) and c.4073+1 G > A (intron 34) variants inherited in compound heterozygous state exert strong, deleterious impact on UNC13A exon 9 (a) and exon 34 splicing (b), respectively. c, an UNC13A c.2186+5 G > A variant (intron 18) inherited in a homozygous state revealed strong, deleterious impact on UNC13A exon 18 splicing. All events are predicted to cause mis-splicing and downstream nonsense-mediated mRNA decay. a.u., arbitrary units.

Extended Data Fig. 4. Electrophysiological characterization of UNC13A^R799Q and UNC13A^N1013S in Unc13a/b DKO neurons reveals no evidence for variant pathogenicity.

a-l, Data obtained from neurons expressing UNC13A^WT (gray) or UNC13A^R799Q (brown). a, Example traces of spontaneous miniature excitatory postsynaptic currents (mEPSCs), and b, quantification of mEPSC frequency and c, mEPSC amplitude. d, Example traces and e, quantification of the charge transfer during a single action-potential evoked excitatory postsynaptic currents (eEPSCs). f, Example traces and g, quantification of the readily-releasable pool of SVs by application of hypertonic sucrose. h, Average vesicular release probability ${\stackrel{\circledR }{p}}_{\mathit{vr}}$. i, Changes in eEPSC amplitude in response to PDBu application during a 0.1 Hz action potential stimulation recorded in UNC13A^WT (gray) or UNC13A^R799Q (brown) expressing neurons. j,k, Averaged time courses of normalized eEPSC amplitudes, before, during and after a high-frequency action potential train at 10 Hz (j) or 40 Hz (k), for the indicated genotypes. l, Replenishment rate constant calculated based on a single-pool model for neurons of the indicated genotypes. m-x, similar data for neurons expressing UNC13A^WT (gray) or UNC13A^N1013S (purple). Bars and circles in a-h, l, m-t and x represent means and values for individual neurons, respectively. Error bars, s.e.m. Statistical analysis: two-tailed Mann-Whitney test. See Supplemental table 2 and 3 for further details.

Extended Data Fig. 5. Landscape of the UNC13A disorder.

a, Illustrations of the synapse (left) and the core molecular machinery of the active zone (right). Variations in the genes encoding for these proteins are cause for SNAREopathy disorders. b, Spectrum of pathogenic variants, patterns of inheritance and associated phenotypes reported for genes encoding neuronal SNAREs and associated AZ proteins. Frequency of diverse disease-causing variants and their patterns of inheritance are shown based on the data in the Human Gene Mutation Database (HGMDv.4.1). Size of circles corresponds to the frequencies of reports in the literature with respective numbers labelled. Spectrum of associated phenotypes derived from clinical synopses in OMIM, GeneReviews and HGMD is shown with light blue boxes for each gene. AD, autosomal dominant; AR, autosomal recessive; ASD, autism spectrum disorder; CNV, copy number variant; DD, developmental delay; ID, intellectual disability; LGD, likely gene disrupting. c, Summary of the key data defining the UNC13A condition, stratified according to the three subtypes defined in the present study. d, Summary of the key experimental findings for each of the characterized disease variants. The data is normalized to the respective control (dotted vertical lines) to allow a comparison of magnitudes and directions of variant-associated changes. SEMs were calculated by Gaussian error propagation. Data for the P814L variant are obtained from. N.A; not analyzed, N.D; not determined.

Extended Data Fig. 6. UNC13A disease variants do not affect UNC13A stability in HEK293FT cells.

a, A GFP-P2A-UNC13A-FLAG (top) is cleaved in cells to produce two fragments (bottom), GFP and UNC13A-FLAG. b, An example blot including total protein stain (up), an anti-FLAG immunoblot (IB; middle), and an anti-GFP immunoblot (down) from one of four experiments used to determine the expression level of UNC13A-FLAG with or without disease variants in comparison to that of GFP. c, Changes in the ratio of UNC13A-FLAG to GFP intensity were statistically not significant, which agrees with the notion that disease variants do not interfere with UNC13A stability. Circles represent four independent biological replications. Error bars represent ± SEM. Statistical anaylsis: Kruskal-Wallis test.

Extended Data Fig. 7. Molecular Dynamics (MD) simulation predicts variant-specific reorganization of the UNC13A C1 domain.

The C1 domain structure is maintained by a zinc-cluster. Two Zn²⁺ ions are coordinated each by one histidine and three cysteine residues. C587, in which a disease variation was found, and H554 (H567 in the rodent UNC13A homolog; Fig. 5d) are Zn²⁺-coordinating residues. The finding that the C587F variant has a minor effect on basal synaptic transmission properties is surprising, because knock-in mouse neurons expressing a H567K mutation that abolishes PDBu/DAG binding^,, have impaired synaptic transmission properties. To understand the cause of this difference, we compared the effect of H554K and C587F on the structural integrity of the human UNC13A C1 domain using molecular dynamics (MD) simulations. The human UNC13A models comprising the C1, C2B, and MUN domains (residues 505 to 985) were predicted with Alphafold2 using the UniProt sequence Q9UPW8 as input. We ran five independent 1 µs-long MD simulations for the WT protein, H554K, C587F, and for other disease variants as controls (see materials and methods). a, Averaged dipole moments of the C1 domain during the last 250 ns of five simulations (black dots). We did not observe unbinding of the Zn²⁺ ion or unfolding of the domain within the simulation time period. b, Representative snapshots of the Zn²⁺ ion binding pocket in the WT UNC13A C1 domain and in the UNC13A C1 domain carrying the C587F and H554K substitutions, after 1 µs MD simulations (‘relaxed’) compared to the initial state. The Zn²⁺ ion is represented by as a blue sphere. The adjacent E586 replaces F587 in the coordination of the Zn²⁺ ion, but not K554. c, The averaged minimal distances between Zn²⁺ and E586 of the last 250 ns of five independent MD simulations (black dots). The structural and electrostatic changes observed in MD simulations support the notion that C587F exerts a different effect on the C1 domain structure than H554K, although both abolish PDBu sensitivity. In a and c the standard deviation is indicated by black bars. Averages are highlighted as gray dots.

Extended Data Fig. 8. The UNC13A R202H disease variant leads to reduced synaptic strength and enhances short-term facilitation and augmentation.

Recordings in autaptic hippocampal neurons were made in the presence of 2 mM Ca²⁺/2 mM Mg²⁺ in the extracellular solution. a, EPSC charge, b, RRP size and c, the average vesicular release probability ${\stackrel{\circledR }{p}}_{\mathit{vr}}$ are reduced in UNC13A^R202H-expressing neurons. Note that ${\stackrel{\circledR }{p}}_{\mathit{vr}}$ was unchanged under the experimental conditions in Fig. 3. d, Potentiation of the EPSC amplitude by the DAG analog PDBu is stronger in UNC13A^R202H-expressing neurons. e,f, Stronger facilitation during and larger augmentation after AP trains at 1 Hz or 10 Hz are observed in neurons expressing UNC13A^R202H. Note that before and after the high frequency trains, EPSCs were probed at frequency of 0.2 Hz, and not at 0.1 Hz as in the rest of the manuscript. Bars and circles in a-d represent means and values for individual neurons, respectively. Error bars represent ± SEM. Statistical anaylsis: two-tailed Mann-Whitney test.