Figures and data
SIM imaging of dendritic spines, automated measurements of spine morphology, and generation of subtracted spine density plots for population-level analysis.
(A) Original SIM image of a dendritic shaft stained with lipophilic membrane dye DiI. (B) Binarized image of the same dendrite (top) and segmented spines numbered from 1 to 24 (middle). The bottom image shows segmented spines (white) and the dendritic shaft (blue). (C) Enlarged images of the individual spines shown in (B). The pseudocolor images indicating the relative positions of the spine segments from the base (blue) to the tip (yellow). (D) 3D views of three spines (No. 7, 15, and 24) viewed from different angles. (E) Principal component analysis (PCA)-based dimensional reduction of spine characteristics plotted in the plane of principal components (PC)1 and PC2. (F) Process of generating a subtracted density plot. The scatter plot of spine distribution in the PC1–PC2 plane based on morphological parameters was converted into a density plot for each culture source (genotype or treatment group), and the corresponding plots were subtracted to reveal differences in spine morphology at the population level. Bars, 4 μm for A, B, and E, 1 μm for C.
Spine population profiles for each model and corresponding control mouse line presented as subtracted density plots.
(A) The subtracted density plots for eight disease mouse models (Nlgn3R451C/(y or R451C), Syngap1+/−, POGZQ1038R/+, 15q11-13dup/+, 3q29del/+, 22q11.2del/+, Setd1a+/−, and CaMKIIαK42R/K42R) and three different culture conditions (immature culture at 13 DIV, AMPA glutamate receptor blocker CNQX treatment, and GABAA receptor blocker bicuculine treatment). The areas with a higher density of spines from mutant disease model mice are shown in yellow, and the areas with reduced density are shown in blue. The total number of spines (first number, n) analyzed from control and mutant mouse neurons and the corresponding number of dendrites (number in parentheses) are as follows: Nlgn3R451C/(y or R451C); n = 1,134 (58) and 1,204 (59), Syngap1+/−; n = 991 (65) and 1,371 (83), POGZQ1038R/+; n = 1,341 (72) and 1,271 (72), 15q11-13dup/+; n = 1, 208 (68) and 1,099 (63), 3q29del/+; n = 1,429 (66) and 1,408 (66), 22q11.2del/+; n = 1,143 (66) and 914 (63), Setd1a+/−; n = 1,381 (71) and 1,405 (71), CaMKIIαK42R/K42R; n = 880 (58) and 763 (54). All data are from three independent culture preparations. (B) Matrix of the 2D cross-correlations among subtracted density plots. In the lower right area, a spine group showing similar morphological changes can be identified. This group corresponded to the mouse models of schizophrenia. (C) Unbiased clustering of spine samples showing two distinct groups corresponding to schizophrenia (cyan) and ASD (red).
Distinct morphological properties of spines in cultured neurons derived from schizophrenia model mice compared with ASD model mice.
(A, C, and E) Relative enrichment of four different subgroups of spines in Nlgn3R451C/(y or R451C) neurons compared with wild-type neurons. In the projection plane of PC1 and PC2, four areas with distinct structural properties, namely (1) small and short, (2) small and long, (3) large and short, and (4) large and long, were defined (A). The map of areas with large differences (> 3 × SD) between control and mutant mice in spine number per area in the feature space (C) indicates that the area enriched with wild-type spines (area B, yellow region) overlaps with the area of large spines. The area enriched with the mutant spines overlaps with the area of small spines (area A, blue region). The relative enrichment of mutant spines was summarized in (E). (B, D, and F) Similar analysis of mutant spine enrichment in 22q11.2del/+ neurons. The map of spine distribution with four different properties (B), the map of areas enriched with wild-type or mutant spines (D), and the graph of relative enrichment of mutant spines (F) show that the mutant neurons were enriched with small and long spines. (G-J) Relative abundance of the four subgroups of spines in 8 mutant mouse models (1: Nlgn3R451C/(y or R451C), 2: Syngap1+/−, 3: POGZQ1038R/+, 4: 15q11-13dup/+, 5: 3q29del/+, 6: 22q11.2del/+, 7: Setd1a+/−, 8: CaMKIIαK42R/K42R). Three independent culture experiments with paired wild-type and mutant samples were performed. Group comparisons were performed using one-way ANOVA followed by Tukey’s post hoc test. (* p < 0.05, ** p < 0.01, *** p < 0.005).
Time-lapse imaging of neurons derived from 22q11.2del/+ mice and corresponding control mice.
(A and B) Images of dendritic segments from control neurons (A) and 22q11.2del/+ neurons (B) at two different time points. (C and D) Montages of time-lapse images from control neurons (C) and 22q11.2del/+ neurons (D). The curved dendrites were straightened, revealing newly formed spines (arrowheads) as fluorescent objects appearing at the edge of the dendritic shafts. Bar, 4 μm.
Turnover rate, lifetime, and growth/shrinkage profiles of dendritic spines in cultured neurons derived from 22q11.2del/+, Setd1a+/−, and Nlgn3R451C/(y or R451C) mice, as well as the corresponding controls.
(A-C) Spine turnover rates in the three mutant mouse models compared to the controls. Spine turnover rates were analyzed using a linear mixed-effects model with genotype as a fixed effect and plate, cell, and dendrite as nested random effects, showing a significant effect of genotype only in 22q11.2del/+ and Setd1a+/− mice (n = 16 dendrites/9 cells/5 plates in control and 11 dendrites/6 cells/4 plates in 22q11.2del/+, n = 10 dendrites/7 cells/4 plates in control and n = 14 dendrites/9 cells/6 plates in Setd1a+/−, n = 8 dendrites/8 cells/3 plates in control and n = 8 dendrites/8 cells/3 plates in Nlgn3R451C/(y or R451C)). (D-F) Lifetimes of transient spines for the three mutant mouse models compared with corresponding controls. Spine lifetime was analyzed using a linear mixed-effects model accounting for the hierarchical structure of the data (spines nested within dendrites, cells, and culture plates). The analysis revealed a significant effect of genotype only in 22q11.2del/+ and Setd1a+/−neurons (n = 186 spines/11 dendrites/7 cells/5 plates in control and n = 166 spines/7 dendrites/4 cells/4 plates in 22q11.2del/+, n = 82 spines/5 dendrites/5 cells/4 plates in control and n = 202 spines/8 dendrites/8 cells/5 plates in Setd1a+/−, n = 98 spines/8 dendrites/8 cells/3 plates in control and n = 125 spines/8 dendrites/8 cells/3 platess in Nlgn3R451C/(y or R451C)). (G-I) Temporal patterns of spine growth (n = 60 spines/11 neurons/7 cells/5 plates in control and n = 50 spines/7 dendrites/4 cells/4 plates in 22q11.2del/+, n = 25 spines/5 dendrites/5 cells/4 plates in control and n = 65 spines /8 dendrites/8 cells/5 plates in Setd1a+/−, n = 43 spines/8 dendrites/8 cells/3 plates in control and n = 41 spines/7 dendrites/7 cells/2 plates in Nlgn3R451C/(y or R451C)). Spine volume trajectories were analyzed using linear mixed-effects models incorporating nested random effects (spine within dendrite within cell within culture plate) to account for the hierarchical structure of the data. Newly formed spines in both 22q11.2del/+ and Setd1a+/− neurons were significantly smaller than those in wild-type neurons. In contrast, newly formed spines in Nlgn3R451C/(y or R451C) neurons were significantly larger than those in wild-type neurons. (J-L) Temporal patterns of spine shrinkage (n = 39 spines/10 dendrites/6 cells/5 plates in control and n = 37 spines/7 dendrites/4 cells/4 plates in 22q11.2del/+, n = 15 spines/5 dendrites/5 cells/4 plates in control and n = 55 spines/8 dendrites/8 cells/5 plates in Setd1a+/−, n = 28 spines/8 dendrites/8 cells/3 plates in control and n = 35 spines/8 dendrites/8 cells/3 plates Nlgn3R451C/(y or R451C) neurons). Spine volume trajectories were analyzed using linear mixed-effects models as in (G-I). In the 22q11.2del/+ neurons, spines undergoing elimination were significantly smaller than those in wild-type neurons. This effect was not detected in Setd1a+/− and Nlgn3R451C/(y or R451C) neurons.
Simulations of long-term spine turnover.
(A) The model of spine dynamics. Five successive phases of spine state transitions were defined. Phase 1: newly formed spines grow at speed V1. Phase 2: nascent spines are eliminated with probability P1. Phase 3: nascent spines are stabilized when volume reaches an upper threshold. Phase 4: stable spines are destabilized with probability P2. Phase 5: spines shrinking at a rate V2 are lost after reaching a lower threshold. (B) Pseudocolor maps of 625 different combinations of these parameters to identify those best fitting the experimental data. The color-coded values indicate the fitness of the simulation results to both the spine lifetime distribution profile and the turnover rate. (C) Frequency histograms of spine lifetimes for the three mouse models and controls (experimental data), and the results of simulations. The bin with a lifetime of 24 h corresponds to the spines that persisted throughout the imaging period. (D) Differences in V1, V2, P1, and P2 between mutant mice and controls, expressed as differences in the four parameters normalized by the control condition values. (E) Plots of individual spine turnover simulated using parameters that best fit the experimental data from control and schizophrenia-associated mouse models. The upper plots show the progression of spine formation along 200 dendritic segments (50 μm) over 10 days. The lower plots show the enlargement of the last 24 h of spine turnover. Color indicates spine size.
Manipulation of candidate gene expression in wild-type hippocampal neurons and its effects on spine turnover.
(A) Turnover rates of dendritic spines in neurons transfected with an overexpression plasmid encoding Cip4, Npas4, or Ecrg4, or a plasmid encoding an shRNA targeting Met or Arhgap15, together with a GFP expression plasmid. Spine turnover rates were calculated from the images of GFP-expressing dendrites taken at an interval of 24 h. Among the five DEGs, only the upregulation of Ecrg4 selectively increased the spine turnover rate. Linear mixed-effects modeling with experiment as a random factor revealed a significant effect of overexpression treatment (F(3,8)=4.59, p=0.038), driven by Ecrg4 (p=0.013), whereas shRNA manipulations showed no significant effect (F(2,6)=0.29, p=0.76). (results from n = 22 dendrites/11 neurons/3 culture plates derived from independent primary culture for all conditions except the Met shRNA groups, where n = 21 dendrites from 11 neurons were included in the analysis). (B) Fluorescence images of dendritic segments expressing GFP or GFP plus Ecrg4 on days 1 and 2. Newly formed spines are marked by asterisks. Bar, 2 μm. (C) Images of dendrites and axons expressing HA-tagged Ecrg4 together with GFP. Anti-HA immunocytochemistry revealed the presence of immunopositive puncta both in the dendrites (arrows) and axons (arrowheads). Some clusters could be detected in the extracellular space (asterisks). The upper image is the overlay of the anti-HA signal (magenta) and GFP (green). The lower image shows the distribution of the anti-HA signals. Bar, 5 μm.
Altered spine population profiles after suppression of Ecrg4 expression in neurons derived from two schizophrenia-associated mouse models: the 22q11.2del/+ mouse model (A-I) and Setd1a+/− mouse model (J-R).
(A and J) Density plot obtained by subtracting wild-type spines from mutant spines. Both wild-type and mutant neurons expressed control shRNA. (B and K) Area enriched with the mutant spines under the control shRNA condition. (C and L) Area depleted of the mutant spines under the control shRNA condition. (D and M) Density plot obtained by subtraction of the wild-type spines under control shRNA from the mutant spines under Ecrg4 shRNA. This subtraction illustrates the normalizing effect of Ecrg4 shRNA. (E and N) Reduction of mutant spines in area A by Ecrg4 shRNA. The density plot, obtained by subtracting the mutant spines under control shRNA from those under Ecrg4 shRNA, shows the extent of normalization by Ecrg4 shRNA. (F and O) The increase of mutant spines in area B by Ecrg4 shRNA using the same subtraction method. (G and P) The extent of rescue by Ecrg4 shRNA in Area A and B for three independent culture experiments using either 22q11.2del/+ or Setd1a+/− mouse models. The red lines show the values corresponding to statistical significance (p = 0.05), estimated from permutation analyses shown in (H, I, Q, and R). (H and Q) Permutation analysis for Area A to estimate the 95th percentile of the shuffled data. (I and R) Permutation analysis for Area B.
Cumulative frequency plots of spine length, surface area, and volume measured in four independent experiments performed > 2 months apart.
The cumulative frequency distributions of spine length (A), spine surface area (B), and spine volume (C) were obtained from wild-type data acquired by SIM imaging across four mouse models: Nlgn3R451C/(y or R451C) (S1), Syngap1+/− (S2), POGZQ1038R/+ (S3), and 15q11-13dup/+ (S4). The Kolmogorov–Smirnov test detected significant differences in only three of 18 possible pairwise comparisons: surface area of (S1) vs. (S3) (p = 0.017), volume of (S1) vs. (S3) (p = 0.032), and volume of (S3) vs. (S4) (p = 0.038).
Cumulative frequency plots of spine length, surface area, and volume for the eight mouse mutants: Nlgn3R451C/(y or R451C), Syngap1+/−, POGZQ1038R/+, 15q11-13dup/+, 3q29del/+, 22q11.2del/+, Setd1a+/−, and CaMKIIαK42R/K42R.
Areas where control (A: blue) or mutant (B: yellow) spines show a higher density within the feature space of the PC1-PC2 plane.
The plots for all 8 mouse models are presented. The differences in population-level spine properties between ASD- and schizophrenia-associated mouse models were preserved within two groups. The areas enriched with mutant spines in Nlgn3R451C/(y or R451C), Syngap1+/−, and POGZQ1038R/+ models were on the right side of the plot, suggesting the abundance of large spines in this group. In contrast, the areas enriched with mutant spines in 3q29del/+, 22q11.2del/+, Setd1a+/−, and CaMKIIαK42R/K42R models were positioned on the left side of the plot, indicating the abundance of small spines.
The relative numbers of spines within areas A and B in the feature space from Supplementary Figure 3.
The densities of mutant spines were higher in area B than in area A.
Profiles of different spine populations (spines in the control-dominant area A and the mutant-dominant area B for both wild-type and mutant neurons) were visualized by plotting the radius along the long axis.
The numbers of control and mutant spines included in the analysis are as follows: Nlgn3R451C/(y or R451C), n = 775 in area A and n = 513 in area B; Syngap1+/−, n = 668 in area A and n = 500 in area B; POGZQ1038R/+, n = 580 in area A and n = 815 in area B; 15q11-13dup/+, n = 378 in area A and n = 797 in area B; 3q29del/+, n = 627 in area A and n = 824 in area B; 22q11.2del/+, n = 308 in area A and n = 575 in area B; Setd1a+/−, n = 472 in area A and n = 808 in area B; CaMKIIαK42R/K42R, n = 237 in area A and n = 634 in area B.
Spine turnover (A), density (B), and size (C) from simulation data.
Simulation parameters were tuned to fit experimental results from control and mutant models. Means and standard deviations are shown.
Pseudocolor presentation of differentially expressed genes (DEGs) between mutant and corresponding control mice.
The number of shared DEGs was higher in the schizophrenia-related mouse models than in the ASD-related mouse models. Mouse gene identifiers (ENSMUSG) and gene names for DEGs shared by three or four schizophrenia-associated mouse models and not differentially expressed in ASD-associated mouse models are presented. DEGs analyzed for their effects on spines are shown in red. Color indicates the number of DEGs. The lower table shows all DEGs shared by three or four mouse models, irrespective of their grouping as ASD-related or schizophrenia-related.
(A-C) Validation of overexpressed tagged constructs by immunocytochemistry in hippocampal neurons. Bar = 20 μm. The data are presented as mean ± SEM (N = 5 neurons each). Red allows; transfected neurons, white arrows; non-transfected neurons. (A) Immunocytochemistry of hippocampal neurons expressing Ecrg4-HA with GFP using anti-Ecrg4 antibody. Anti-GM130 antibody staining detected the Golgi apparatus. (B) Immunocytochemistry of hippocampal neurons expressing Cip4-HA with GFP using anti-Cip4 antibody. Anti-MAP2 antibody detected neuronal dendrites. (C) Anti-NPAS4 and anti-HA immunocytochemistry of hippocampal neurons expressing NPAS4-Flag-HA with GFP. (D-E) Immunoblot verification of tagged construct expression in COS-7 cells. (D) Immunoblotting in total cell lysate from COS-7 cells expressing Cip4-HA or NPAS4-Flag-HA. The estimated molecular weights of Cip4-HA and NPAS4-Flag-HA are 64 kDa and 89 kDa, respectively. (E) Immunoblotting using both anti-HA and anti-Ecrg4 in total cell lysate from COS-7 cells expressing Ecrg4-HA or Ecrg4-HA-SEP. The estimated molecular weights of Ecrg4, Ecrg4-HA, and Ecrg4-HA-SEP are 17 kDa, 18 kDa, and 45 kDa, respectively. (F-H) Verification of knock-down effects of shRNAs by immunoblotting of COS-7 cells. Immunoblotting in total cell lysates from COS-7 cells transfected with the expression vectors of either Met-GFP (F), ArhGAP15-GFP (G), or Ecrg4-HA (H), together with dsRed2 and shRNAs targeting the corresponding transcripts. (I) Measurement of spine turnover rate in neurons transfected with a control shRNA construct together with GFP expression vectors. For the experimental control, we used neurons transfected only with GFP. The data are presented as mean ± SEM (n = 15 dendrites from 9 neurons for GFP control, n = 16 dendrites from 8 neurons for GFP plus control shRNA).
Expression and distribution of Ecrg4 in cultured hippocampal neurons.
(A) Preferential localization of Ecrg4 protein in the membrane fraction. Immunoblotting of Ecrg4, GluA1AMPA-type glutamate receptor, and α-tubulin in the remaining fraction after removal of nuclei (S1), supernatant after centrifugation (S2), and resulting pellet (P2). (B) Images of an immunostained hippocampal neuron expressing HA-tagged Ecrg4 and NPY-GFP. The fluorescence signals were partially colocalized, suggesting Ecrg4 protein accumulation in dense-core vesicles. Bar = 2 μm. (C) Quantification of the overlap between puncta immunopositive for HA-tagged Ecrg4 and NPY-GFP. N = 4 cells from 1 dish. (D) Surface labeling with the anti-GFP nanobody revealing clusters of SEP-tagged Ecrg4 in dsRed2-positive axons (arrows) and dendrites (arrowheads). A nanobody was applied at time t = 0 min. Bar = 2 μm. (E) Enlarged image of a single Ecrg4 puncta in the axon, marked by a yellow square in (D), with the fluorescence intensity profile along the dashed arrow. Bar = 0.5 μm. (F) The spine density of cultured hippocampal neurons 24 hours after administration of either Ecrg4 peptides or scrambled peptides (n = 26 dendrites and 13 neurons for each condition). (G) Spine turnover rate of neurons treated with Ecrg4 peptides or scrambled peptides (n = 26 dendrites and 13 neurons for each condition).
