Cul3 regulates cytoskeleton protein homeostasis and cell migration during a critical window of brain development

Abstract

De novo loss of function mutations in the ubiquitin ligase-encoding gene Cullin3 (CUL3) lead to autism spectrum disorder (ASD). In mouse, constitutive Cul3 haploinsufficiency leads to motor coordination deficits as well as ASD-relevant social and cognitive impairments. However, induction of Cul3 haploinsufficiency later in life does not lead to ASD-relevant behaviors, pointing to an important role of Cul3 during a critical developmental window. Here we show that Cul3 is essential to regulate neuronal migration and, therefore, constitutive Cul3 heterozygous mutant mice display cortical lamination abnormalities. At the molecular level, we found that Cul3 controls neuronal migration by tightly regulating the amount of Plastin3 (Pls3), a previously unrecognized player of neural migration. Furthermore, we found that Pls3 cell-autonomously regulates cell migration by regulating actin cytoskeleton organization, and its levels are inversely proportional to neural migration speed. Finally, we provide evidence that cellular phenotypes associated with autism-linked gene haploinsufficiency can be rescued by transcriptional activation of the intact allele in vitro, offering a proof of concept for a potential therapeutic approach for ASDs.

Fig. 1. Behavioral defects in Cul3 haploinsufficient mice.

a Hind limb clasping in adult Cul3^+/− mice, not observed in wild-type littermate controls (a top); scoring 0–1 (normal) to 3 (most severe) (a bottom, n = 25 animals per genotype; ***P < 0.0001; two-tailed Mann–Whitney U-test). b Cul3^+/+ and Cul3^+/− strides, forepaws (blue), and hind paws (red). c Altered gait of Cul3^+/− mice evidenced by comparison of sway (c top) and stance length (c bottom) (n = 19 mice per genotype; *P = 0.03, ***P = 0.0004; two-tailed t-test). d, d′ Accelerating RotaRod test revealing defects in motor performance and coordination in Cul3^+/− mice. Shown: sum of daily latencies of three trials per day on three consecutive days and final rpm on day one—trial 1, i.e,: initial coordination (d′) (n = 30 mice per genotype; *P = 0.021, **P = 0.001, ***P = 0.0005; 2-way ANOVA and Sidak’s multiple comparison test and unpaired two-tailed t-test). e, f Heat maps of three-chamber social interaction test (left), quantification of interaction times (middle), and social preference/novelty index (right). Sociability: Cul3^+/− and control mice spend more time with a stranger mouse (M1) than an object (Ob.) (e); social novelty: Cul3^+/− mice do not prefer a novel stranger (M2) over the familiar mouse (M1) (f) (n = 24 mice per genotype; *P = 0.01, ***P < 0.0001, n.s. not significant; 1-way ANOVA and Sidak’s multiple comparison test). g Both genotypes distinguish and familiarize to non-social and social odors in the olfaction habituation and dishabituation test, yet Cul3^+/− mutant mice are hyper-reactive to the presentation of social odors (n = 24 mice per genotype; **P = 0.002, ***P = 0.0009, n.s. not significant; 2-way ANOVA and Sidak’s multiple comparison test; details in Supplementary Data 1, 2). h, h′ Contextual memory retention and extinction scored as percent freezing during a 3 min exposure to the context (h), and fear-acquisition training (h′) (n = 26 mice per genotype; *P = 0.027, n.s., not significant; 2-way ANOVA interaction: (F1,100)= 6.18; P = 0.015; Sidak’s multiple comparisons test: Extinction P = 0.027). Sex-matched littermate animals were analyzed. Data presented either as mean ± SEM, as well as scatter plot (a, c, d′, e, f, h) or as boxplot showing median value and 25–75th percentile, whiskers show minimum and maximum (d, g, h′). Detailed statistics are provided in Supplementary Data 1.

Fig. 2. Cul3 loss after completion of main developmental milestones does not lead to major behavioral abnormalities in mice.

a, b Juvenile Cul3^+/fl Cag-CreER mice were injected for 5 days with either 100 mg/kg tamoxifen (TM) or vehicle (V), Cul3^+/fl animals were injected with TM. Behavioral tests were performed after ≥21 days post-last injection, followed by western blot analysis of brain Cul3 levels (a scheme; b representative western blot). c Quantification of Cul3 levels of tamoxifen-treated mice, normalized to vehicle-injected controls (n = 14 mice per condition; ***P < 0.0001; unpaired two-tailed t-test). d Hind limb clasping scoring from 0–1 (normal) to 3 (most severe) in all conditions (n(Cul3^+/fl + TM) = 13, n(Cul3^+/fl Cag-CreER + V) = 18, n(Cul3^+/fl Cag-CreER + TM) = 20; 1-way ANOVA and Sidak’s multiple comparison test). e, e′ Accelerating RotaRod revealed reduced motor learning abilities in Cul3^+/fl Cag-CreER + TM mice (e), initial coordination was unaffected (n(Cul3^+/fl + TM) = 13, n(Cul3^+/fl Cag-CreER + V) = 8, n(Cul3^+/fl Cag-CreER + TM) = 14; *P = 0.03; 2-way ANOVA (e), 1-way ANOVA (e′) and Sidak’s multiple comparison tests). f, g Normal behavior in three-chamber sociability test upon juvenile Cul3 loss, interaction times (left), social preference/novelty index (right). Test and control mice significantly prefer a stranger mouse (M1) over the object (Ob.) (f), and the novel stranger (M2) over the familiar mouse (M1) (g) (n(Cul3^+/fl + TM) = 13, n(Cul3^+/fl Cag-CreER + V) = 18, n(Cul3^+/fl Cag-CreER + TM) = 20; *P < 0.05, ***P < 0.0001; 1-way ANOVA and Sidak’s multiple comparison test). h All test-groups distinguish and familiarize similarly to social odors in the adapted olfaction habituation and dishabituation test (n(Cul3^+/fl + TM) = 13, n(Cul3^+/fl Cag-CreER + V) = 8, n(Cul3^+/fl Cag-CreER + TM) = 10; 2-way ANOVA and Sidak’s multiple comparison test; details: Supplementary Data 1, 2). i, i′ Contextual memory retention and extinction scored as percent freezing during exposure to context (i), and fear-acquisition training (i′) (n(Cul3^+/fl + TM) = 13, n(Cul3^+/fl Cag-CreER + V) = 8, n(Cul3^+/fl Cag-CreER + TM) = 10; 2-way ANOVA and Sidak’s multiple comparison test). Sex-matched littermates were analyzed. Data presented either as mean ± SEM, as scatter plot (c, d, e′, f, g, i) or as boxplot showing median value and 25–75th percentile, whiskers show minimum and maximum (e, h, i′). Significance levels: *P < 0.05, **P < 0.01, ***P < 0.001, n.s. not significant. Detailed statistics are provided in Supplementary Data 1.

Fig. 3. Abnormal lamination of the cortex and migration defects in Cul3 mutant mice.

a–d Immunofluorescent stainings for Ctip2 and Cux1 on coronal brain sections revealed laminar thinning in adults (a, b) and newborn (P0) Cul3^+/− animals (c, d) (n(adults) = 3 littermates per genotype; n(P0) = 6 littermates per genotype; *P = 0.04; **P < 0.01; 2-way ANOVA and Sidak’s multiple comparison test). e, i Bin-wise comparison of relative cell numbers (in %) revealed a shifted Cux1/Ctip2 layer profile, indicating laminar defects at P0 (n = 3 littermates per genotype; *P < 0.02, **P < 0.01, ***P < 0.001; 2-way ANOVA, Sidak’s multiple comparison test). f Nissl-staining of P0 coronal, Cul3^+/fl, Cul3^+/fl Emx1-Cre, and Cul3^fl/fl Emx1-Cre brain sections show severe brain malformations in Cul3^fl/fl Emx1-Cre pups (n = 3 littermates per genotype). g, h Immunofluorescent staining with antibodies against Ctip2 and Cux1 reveal cortical laminar thinning in both Cul3^+/fl Emx1-Cre and Cul3^fl/fl Emx1-Cre pups at P0 (n = 3 littermates per genotype; *P < 0.05, **P < 0.01, ***P < 0.001; 2-way ANOVA, Sidak’s multiple comparison test). j Scheme of the BrdU birthdate labeling experiments. k, m Injection of BrdU at E16.5 and anti-BrdU immunofluorescent (IF−) staining and analysis of total BrdU positive (BrdU+) cells in cortical columns at P0 shows severely decreased number of BrdU+ cells in Cul3^fl/fl Emx1-Cre brains, but not in the Cul3^+/fl Emx1-Cre and Cul3^+/− cortex (n = 3 littermate pairs pups per genotype; ***P = 0.0001, n.s. not significant; 1-way ANOVA and Sidak’s multiple comparison test and unpaired two-tailed t-test). l, n Bin-wise analysis of relative numbers of BrdU+ cells showed decreased numbers of BrdU+ cells in upper bins and increased numbers of BrdU+ cells in lower bins in Cul3^+/fl Emx1-Cre, Cul3^fl/fl Emx1-Cre (c) and Cul3^+/− (e) mice (n = 3 littermate pairs per genotype; *P < 0.05, ***P < 0.001; 2-way ANOVA and Sidak’s multiple comparison test). Data presented as stacked bar-plots of mean ± SEM in b, d, h, and connected mean ± SEM in e and i. Data presented as mean ± SEM and scatter plot in k, m and connected mean ± SEM in l and n. Scale bars: 50 µm in a, 25 µm in c, g and 200 µm in f, 25 µm in k, m; numbers in c, g, k, m indicate depth in the cortex. Detailed statistics are provided in Supplementary Data 1.

Fig. 4. Cul3 haploinsufficiency leads to migrations deficits in vivo and in vitro.

a Scheme of lentiviral labeling experiments followed by cortical acute slice preparation and time-lapse live-imaging. b, d Lentiviral injections of hSyn-eGFP at E13.5 and time-lapse imaging of migrating neurons over 16 h at E17.5 reveal migration deficits in the Cul3^+/− cortex (arrowheads: eGFP-labeled cells); (n(animals) = 5 per genotype, n(Cul3^+/+) = 52 cells, n(Cul3^+/−) =  56 cells). c,e Representative cell trajectories of Cul3^+/+ and Cul3^+/− cells indicate reduced path length upon Cul3 haploinsufficiency (e). f Quantification shows a significant decrease in cumulative distances and migration velocity in Cul3^+/− cells (n(animals) = 5 per genotype, n(Cul3^+/+) = 52 cells, n(Cul3^+/−) = 56 cells; **P = 0.001; two-tailed Mann–Whitney U-test). g, h In vitro migration assay of matrigel embedded neurospheres generated from Cul3^+/+ and Cul3^+/− NPCs reveals decreased migratory abilities (r = radius of the furthest migrated cell) 22 h (o representative images) and 46 h after plating. Radius was normalized to initial sphere size; (n(spheres) = 7/6 Cul3^+/+ and Cul3^+/− respectively; *P = 0.011, ***P = 0.0008; 2-way ANOVA and Sidak’s multiple comparison test). i, j Cell tracks of Cul3^+/+ and Cul3^+/− NPCs detaching from the neurosphere into embedding bovine collagen matrix (n(spheres) = 3 per genotype, n(cells) = 30 per replicate) imaged in a single plane. Cell trajectories of each cell fixed at origin plotted in Euclidean plane (i). Mean instantaneous speed (j top) and total cell path length (j bottom) quantification (**P = 0.002; ***P = 0.0004; Wilcoxon rank-sum test). Data are shown as mean ± SEM and scatter plots in f, h. Data presented as mean ± SEM and violin plots with median and first and third quartiles (j). Scale bars: 50 µm in d, 200 µm in g, overview and 40 µm in g, close-up. Detailed statistics are provided in Supplementary Data 1.

Fig. 5. Deregulation of cytoskeletal proteins in Cul3 mutant embryonic forebrain tissue.

a Sample preparation for proteomic analysis (n(Cul3^+/−) = 5 per genotype, n(Cul3^+/fl Emx1-Cre) = 3 per genotype, n(Cul3^fl/fl Emx1-Cre) = 3 per genotype; male mutant and control littermate pairs). b–d, Volcano plot of deregulated proteins at 10% FDR cut-off in the Cul3^+/− developing cortex (b), Cul3^+/fl Emx1-Cre (c), and Cul3^fl/fl Emx1-Cre (d) developing forebrain; cytoskeleton related proteins (purple), proteins differently regulated in more than one data set are underlined, Cul3 (red) (b–d) (details: Supplementary Data 3–5). e DAVID functional annotation identified down- (left) and up- (right) regulated proteins at 20% FDR of the Cul3^fl/fl Emx1-Cre forebrain involved in regulating the activity of RNA polymerase II, the proteasome core complex, neurogenesis and actin and microtubule cytoskeletal organization (selected GO-terms, significant GO-terms in red: RNA polymerase II core complex, GO: 0005665, adj. P-value = 0.002; proteasome core complex GO: 0005839, adj. P-value= 4.9e−07; details: Supplementary Data 6). f, g Comparison of ratios of up- and downregulated proteins in the Cul3^+/−, Cul3^+/fl Emx1-Cre and Cul3^fl/fl Emx1-Cre cortex at log 0.1 change (details: Supplementary Data 7). h, Fitting the mean raw expression levels of Pls3 and INA to a linear regression model indicates that these proteins follow a genotype-dependent dose-response (Pls3: R² = 0.999; INA: R² = 0.932; Cul3^+/fl < Cul3^+/fl Emx1-Cre < Cul3^fl/fl Emx1-Cre; boxplot shows the median value and 25–75th percentile, whiskers show minimum and maximum). i Western blot (left) and analysis of Gapdh normalized intensities (right) confirm increased levels of the cytoskeletal proteins Pls3 and INA in Cul3^+/− lysates (n = 5 per genotype; *P = 0.04, ***P < 0.001; unpaired two-tailed t-tests). j Western blot and analysis of Gapdh normalized intensities confirm increased levels of cytoskeletal proteins Nisch, Pls3, and INA in Cul3^fl/fl Emx1-Cre cortical lysates (samples on the same membrane cut for better visualization, n(Cul3^+/fl) = 5, n(Cul3^fl/fl Emx1-Cre) = 3; **P = 0.003, ***P < 0.001; unpaired two-tailed t-tests). k Quantitative real-time PCR analysis of mRNA levels of Pls1, Pls3, and INA normalized to wild-type levels (ΔCq expression values; n = 5 per genotype; *P = 0.02; unpaired two-tailed t-tests). Data presented as or as box and whiskers, min to max (h) and mean ± SEM in i–k. Detailed statistics are provided in Supplementary Data 1.

Fig. 6. Pls3 regulates cell migration dynamics in Cul3 haploinsufficient NPCs.

a 2D random migration assay was performed over a period of 15 h in Pls3 wt, Pls3 ko, and Pls3 over-expressing NPCs. Representative cell paths/trajectories of Pls3 wild-type, Pls3 ko, and Pls3 over-expressing cells indicate a larger migration area being covered by cells upon Pls3 ko (a middle), while over-expression of Pls3 leads to a smaller migration area (a right). b, c Quantification revealed a significant increase in migration velocity and cumulative distance in Pls3 ko cells, and a significant decrease in Pls3 over-expressing NPCs relative to Pls3 wild-type cells (n(Pls3 wt) = 49 cells, n(Pls3 ko) = 31 cells and n(Pls3 over-expr.) = 37 cells; *P = 0.02, ***P < 0.0001; 1-way ANOVA and Sidak’s multiple comparison test). d 2D random migration assay in Cul3^+/+, Cul3^+/− and Cul3^+/− Pls3 ko NPCs. Representative cell paths/trajectories show a smaller migration area covered by Cul3^+/− NPCs (d middle) as compared to Cu3^+/+ cells (d left). Reducing Pls3 levels in Cul3^+/− mNPCs by Pls3 ko is sufficient to rescue the Cul3 haploinsufficient phenotype (d right). e, f Quantification revealed a significant decrease in migration velocity and traveled distance in Cul3^+/− NPCs, a phenotype rescued when downregulating Pls3 levels through Pls3 ko (n(Cul3^+/+) = 34 cells, n(Cul3^+/−) = 40 cells and n(Cul3^+/− Pls3 ko) = 35 cells; ***P < 0.0001; one-sided Kruskal–Wallis test and Dunn’s multiple comparison test). Data are shown as mean ± SEM and scatter plots in (b, c, e, f). Detailed statistics are provided in Supplementary Data 1.

Fig. 7. Actin cytoskeleton disorganization in Cul3 haploinsufficient and Pls3 over-expressing NPCs.

a Scheme of the actin orientation analysis at the cell front of NPCs. b,e, NPCs were stained using SiR-actin (b) and an anti-tubulin antibody (e). Cell protrusions were imaged employing STED-microscopy (close-up images in insets b′, e′). c, f, i, l, Processed, rotated and analyzed images, color code: hue as the orientation angle, saturation as coherency, and brightness represent photon counts in the STED image (close-ups in insets c′, f′, i′, l′). d, g Orientation distributions for actin (d) and microtubules (g) aligned to the dominant orientation angle of microtubules are shown for Cul3^+/+ and Cul3^+/− cells (n(cells) = 43 per genotype from three independent NPC preparations; ***P = 0.0003, n.s. not significant; two-tailed Welch’s t-test). h, k Cul3^+/− and Pls3 over-expressing NPCs stained using SiR-actin (h) and anti-tubulin antibody (k). Cell protrusions were imaged employing STED-microscopy (close-up images in insets h′, k′). j, m Orientation distributions for actin (j) and microtubules (m) aligned to the dominant orientation angle of microtubules are shown for Cul3^+/− and Cul3^+/+ Pls3 over-expressing cells (n(Cul3^+/−) = 43 cells), (n(Cul3^+/+ Pls3 over-expr.) = 30 cells from three independent NPC preparations; n.s. not significant; two-tailed Welch’s t-test). The dominant orientation is computed as the average orientation angle inside the cell. Plotted is the average angle distribution per group ± 95% confidence intervals. Scale bars: 5 µm in b, e, h, k, 2.5 µm in (insets b′, e′, h′, k′). Detailed statistics are provided in Supplementary Data 1.

Fig. 8. Transcriptional activation of Cul3 rescues the cell migration defects displayed by Cul3 haploinsufficient cells.

a Scheme of CRISPR-mediated activation (CRISPRa). While Cul3^+/+ + Ctrl sgRNA NPCs express wild-type mRNA levels and migrate physiologically (a top, gray), Cul3^+/− + Ctrl sgRNA NPCs show reduced Cul3 mRNA levels and are characterized by cell migration defects (a middle, blue). Upregulation of Cul3 transcription, through a nuclease-deficient dCas9 fused to a transcriptional activator targeting the genomic Cul3 promoter (a bottom, green), increases Cul3 mRNA levels and thereby restores migration dynamics. b Quantitative real-time PCR analysis of Cul3 mRNA levels normalized to the housekeepers Gapdh and Pgk1 as well as the Cul3 levels of wild-type NPCs transfected with their respective guides. Transfection of Cul3^+/− cells employing two different sgRNAs sequences (sgRNA 1 and 2) targeting the Cul3 promoter increased Cul3 expression levels (ΔCq expression values; n = 3 per condition; *P < 0.05, n.s. not significant; 1-way ANOVA and Sidak’s multiple comparison test). c–e 2D random migration assay was performed over a period of 15 h in Cul3^+/+ and Cul3^+/− NPCs transfected with Ctrl sgRNA, as well as Cul3^+/− cells transfected with sgRNA1 and 2. Representative cell trajectories indicate reduced migration by Cul3^+/− + Ctrl sgRNA NPCs (c top right) as compared to Cul3^+/− + Ctrl sgRNA controls (c top left). This Cul3^+/− haploinsufficient phenotype is rescued through transcriptional activation of the Cul3 promoter using two different sgRNAs (Cul3^+/− + sgRNA 1 and 2; c bottom left and right). Quantification revealed a significant increase in migration velocity (d) and traveled distance (e) when restoring Cul3 expression levels in transfected Cul3^+/− NPCs (n(Cul3^+/+ + Ctrl sgRNA)= 53; n(Cul3^+/− + Ctrl sgRNA)= 62 cells, n(Cul3^+/− + sgRNA1)= 48 cells, and n(Cul3^+/− + sgRNA2)= 28 cells; **P < 0.01, n.s. not significant; one-sided Kruskal–Wallis test and Dunn’s multiple correction test). Data are shown as mean ± SEM and scatter plots in b, d, e. Detailed statistics are provided in Supplementary Data 1.