Patient enrolment and ascertainment

Patients were recruited through international collaboration, also using GeneMatcher,²¹ from several clinical and research centres in Canada, France, Germany, Italy, Spain and the USA (further details available in the Supplementary material). Patients were evaluated by paediatricians with expertise in neurological disorders, paediatric neurologists and geneticists with expertise in the field of neurogenetics. Vineland Adaptive Behavior Scale (VABS, third edition) tests were also administered for adaptive behaviour assessment to Patients 4, 5 and 8. Clinical and molecular data of Patient 6, who was partially described in a previous report,²⁰ were critically reviewed. Brain MRI scans, locally performed for patient care, were reviewed by an expert paediatric neuroradiologist (M.S.). Previously published neuroimaging studies were also thoroughly reviewed for comparison and delineation of the neuroradiological spectrum.

Previously reported cases assessment

All the articles indexed in PubMed (https://pubmed.ncbi.nlm.nih.gov/?term=RAC3&sort=date) between April 2019, when RAC3 variants were first associated with NEDBAF by Costain et al.,¹⁸ and May 2021²² were retrieved using the terms ‘RAC3’, ‘Rac3 GTPase’ and ‘neurodevelopmental syndrome’. All the articles concerning the molecular, clinical, and neuroradiological spectrum associated with NEDBAF were thoroughly reviewed. Inclusion criteria for previously published patients included clinical data availability, unambiguous identification of pathogenic/likely pathogenic RAC3 variants and lack of duplication from other previous reports. Ambiguous clinical presentation not consistent with NEDBAF and inconclusive genetic testing were exclusion criteria.

Next generation sequencing

We investigated 10 individuals presenting with syndromic NDD and neuroimaging abnormalities. Genomic DNA was extracted from peripheral blood leucocytes using standard protocols (Supplementary material). Different next generation sequencing (NGS) approaches were employed, including trio-exome sequencing (Trio-ES) (Patients 1–3, 8 and 10), singleton exome sequencing (Patients 4, 6 and 9), trio-genome sequencing and RNA sequencing (Patient 5) and a NGS panel including 480 intellectual disability genes (Patient 7). Sanger sequencing was performed for the validation of the candidate variants and the parental segregation analysis. The presence of copy number variants was investigated in all subjects, either through array comparative genomic hybridization (aCGH) (Patients 1, 2, 4, 5, 6, 7, 8, 9 and 10) or CNV detection on exome sequencing data (Patient 3) (Supplementary material).

Variant identification and analysis

The identified variants were filtered according to minor allele frequency ≤0.001 in Genome Aggregation Database (gnomAD, https://gnomad.broadinstitute.org),²³ conservation (Genomic Evolutionary Rate Profiling—GERP, http://mendel.stanford.edu/SidowLab/downloads/gerp/) and predicted impact on protein function. In silico prediction tools were used for the interpretation of candidate variants, including combined annotation dependent depletion (CADD, https://cadd.gs.washington.edu), Mutation Taster (http://www.mutationtaster.org), Sorting Intolerant From Tolerant (SIFT, https://sift.bii.a-star.edu.sg) and Polyphen-2 (http://genetics.bwh.harvard.edu/pph2/). Ultimately, the American College of Medical Genetics and Genomics and the Association for Molecular Pathology (ACMG-AMP) guidelines were used to classify the candidate variants based on their predicted pathogenicity.²⁴ All RAC3 variants are reported according to the NM_005052.3 transcript (NP_005043.1) (https://www.ncbi.nlm.nih.gov/nuccore/NM_005052.3). Further details are available in the Supplementary material.

Plasmids

RAC3 was amplified by PCR from a cDNA pool of human glioblastoma U251MG cells and cloned into the pCAG-Myc vector. The 11 variants, RAC3-G12R, -P29L, -P34R, -A59G, -G60D, -Q61L, -E62del, -E62K, -D63N, -Y64C and -K116N, were generated by site-directed mutagenesis using KOD-Plus Mutagenesis kit (Toyobo Inc) with pCAG-Myc-RAC3 as a template. RAC3 and the 11 variants were also constructed into pTriEx-4 vector (Merck). pRK5-Myc-PAK1 (p21-activated kinase 1)-KA, a kinase-negative variant with a single amino acid substitution p.K299A and pRK5-Myc-MLK2 (mixed lineage kinase 2/MAPKKK10)-KN, a kinase-negative variant lacking aa 139–183, were kindly provided by Prof. Alan Hall (University College London, UK), and subcloned into the pCAG-Flag vector. RAC-binding regions (RBRs) in human PAK1 (aa 67–150), human MLK2 (aa 401–550), human IRSp53 (aa 2–228), rat N-WASP (aa 191-270), mouse RTKN (Rhotekin) (aa 10–100) and human ROCK (Rho-associated coiled-coil containing protein kinase 1) (aa 67–150) were amplified by PCR from a cDNA pool of U251MG cells, rat brain or mouse brain, and constructed into the pGS21a vector (GenScript). For gene transcription analysis, pGL4.74[hRluc/TK] (control reporter plasmid), pGL4.44[luc2P/AP1-RE/Hygro] (AP1-luciferase reporter plasmid) and pGL4.32[luc2P/NF-κB-RE/Hygro] (NFkB-luciferase reporter plasmid) were purchased by Promega. The SRF reporter element, 5′-CAGACAGACGTGTTCTTAAGTCCATATTAGGACATCTACCATGTCCATATTAGGACATCTACTATGTCCATATTAGGACATCTTGTATGTCCATATTAGGACATCTAAAATGTCCATATTAGGAC-3′, was inserted into pGVB (Toyo, Inc) to yield SRF-luciferase reporter plasmid. All constructs were verified by DNA sequencing.

Antibodies and histochemical reagents

The following antibodies were used: anti-GFP (Medical and Biological Laboratories Cat #598, RRID: AB_591819 or Nacalai Tesque, Cat #04404-84, RRID: AB_10013361), anti-NeuN (Millipore, Cat #MAB377, RRID:AB_2298772) and anti-Myc (Medical and Biological Laboratories, Cat #M047-3, RRID: AB_591112). Alexa Fluor 488 and 568 (Invitrogen) were used as secondary antibodies. 4′,6-diamidino-2-phenylindole (DAPI; Nichirei Bioscience) was used for staining DNA. Rhodamine-phalloidin (Invitrogen) was used for staining filamentous actin.

GTP-hydrolysis and GTP/GDP-exchange assays

Preparation and purification of His-tag-fused RAC3 and the 11 variants were carried out according to the manufacturer’s instructions. To assess the basal GTP/GDP-exchange reactions, release of methylanthraniloyl (mant)-GDP (Sigma-Aldrich) was measured.²⁵ Briefly, loading of RAC3 proteins with ^mantGDP was performed by incubation for 90 min at 20°C in ^mantGDP loading buffer [20 mM HEPES-NaOH (pH 7.5), 50 mM NaCl, 0.5 mM MgCl₂, 5 mM EDTA, 1 mM dithiothreitol, 20-fold excess of ^mantGDP] and subsequently adding 10 mM MgCl₂ (final concentration) to stop the reaction. Remaining non-bound nucleotide was removed from the incubation mixture with a NAP-5 column (GE Healthcare Life Sciences). ^mantGDP loaded GTPase was incubated with GppNHp (Abcam), which has 100 times higher concentration than that of RAC3 proteins, and fluorescence (365/450 nm) change was measured using an MTP-800 microplate reader (Corona). Intrinsic GTP-hydrolysis activity was then undertaken by directly measuring changes in the GTP concentration with a luciferase-coupled GTPase assay kit (GTPase-Glo Assay Kit, Promega) according to the manufacturer’s instructions.²⁶ In short, 5 μM GTP and serially diluted RAC3 proteins were incubated in GTPase/GAP buffer for 60 min at room temperature. Once the GTP-hydrolysis reaction was completed, the remaining GTP was converted to ATP by adding GTPase-Glo Reagent. Subsequently, generated ATP was detected by the luciferase/luciferin-based reagent using a Turner Designs TD-20/20 luminometer (BMG Labtech).

Cell culture and transfection

COS7 (monkey kidney fibroblast-like cell) and primary hippocampal neurons derived from embryonic day (E) 16.5 mice were cultured as described.²⁷ Transient transfection was carried out using polyethyleneimine ‘MAX’ reagent (for COS7 cells) (Polysciences Inc) or Neon transfection system (for primary neurons) (Invitrogen).

Pull-down assay of GTP-bound RAC3 and the variants

Glutathione S-transferase (GST)-fused RBRs of PAK1, MLK2, IRSp53, N-WASP, RTKN and ROCK were expressed in Escherichia coli BL21 (DE3) strain. Recombinant GST-proteins were purified according to the manufacturer’s instructions. COS7 cells were transfected with pCAG-Myc-RAC3, RAC3-G12R, -P29L, -P34R, -A59G, -G60D, -Q61L, -E62del, -E62K, -D63N, -Y64C or -K116N (1.0 μg/60 mm-dish). After 24 h, cells were lysed with the pull-down buffer (50 mM Tris-HCl, pH7.5, 150 mM NaCl, 5 mM MgCl₂, 0.1% SDS, 1% Nonidet P-40 and 0.5% deoxycholate), and insoluble materials were removed by centrifugation. The supernatant was then incubated for 30 min at 4°C with Glutathione Sepharose 4B beads (GE Healthcare Life Sciences) with which GST-fused RBR of PAK1, MLK2, IRSp53, N-WASP, RTKN or ROCK was bound. The beads were washed twice with the pull-down buffer, and bound proteins were analysed by western blotting. Images were captured with an LAS-4000 luminescent image analyser (GE Healthcare Life Sciences).

Assay of SRF, AP1 and NFkB-mediated gene transcription

Assays were performed as described.²⁸ The control reporter vector was co-transfected into COS7 cells seeded on 24-well plates with the indicated RAC3 expression plasmid (0.1 μg/well) together with the SRF-, AP1- or NFkB-luciferase reporter plasmid. After transfection, cells were washed once with phosphate-buffered saline (PBS) and lysed with the passive lysis buffer according to the manufacturer’s instructions. Luciferase activities were determined with the dual-luciferase reporter assay system (Promega).

In utero electroporation

The surgery on pregnant mice and embryo manipulation in the uterus were performed as previously described.^29,30 At E14, pregnant ICR mice provided by Japan SLC were deeply anaesthetized with a mixture of three drugs: medetomidine (0.75 mg/kg), midazolam (4 mg/kg) and butorphanol (5 mg/kg).³¹In utero electroporation was then performed in the specific-pathogen-free animal facilities as described.³² Briefly, 1 µl of solution containing 0.1 μg of pCAG-Myc vector (control), pCAG-Myc-RAC3, -D63N, -E62del, -Y64C or -Q61L was injected with pCAG-EGFP (0.5 μg) into the lateral ventricle of mouse embryos with a glass micropipette made from a microcapillary tube (GD-1; Narishige). After the embryo in the uterus was placed between the tweezers-type disc electrode (5 mm in diameter) (CUY650-5; NEPA Gene), electronic pulses (50 ms of 35 V) were charged five times at 450 ms intervals with an electroporator (NEPA21; NEPA Gene). In this method, plasmids are introduced into the somatosensory area, which is included in the parietal lobe. Brains were fixed at indicated postnatal days, sectioned, and then analysed. As for the quantification of distribution of GFP-positive cells in brain slices, coronal sections of cerebral cortices containing the labelled cells were classified into three bins. The number of labelled cells in each region of at least three slices per brain was counted. A graphical timeline of the study design is shown (Supplementary Fig. 1). All experimental procedures were carried out in the daytime. Animals were neither excluded nor died during experimentation.

Time-lapse imaging

After in utero electroporation at E14, organotypic coronal slices (250 µm-thickness) were prepared at E16 from the interventricular foramen with a microtome, placed on an insert membrane (pore size, 0.4 µm; Millipore), mounted in agarose gel, and cultured. The dishes were then mounted in an incubator chamber (5% CO₂ and 40%O₂, at 37°C) fitted onto an FV1000 confocal laser microscope (Olympus) and the primary somatosensory cortex was examined as described.³³ Approximately 10–15 optical Z sections were acquired automatically every 15 min for 15 h, and about 10 focal planes (~50 µm-thickness) were merged to visualize the entire shape of the cells.

Quantitative analysis of axon growth

For estimation of axon growth, GFP signal intensity of the callosal axons was measured at P0 or P7 using ImageJ software in distinct regions (bin 1–3 or bin 1–4). The relative intensities of bins were normalized with bin 1 as 1.0 and compared using R software.

Immunofluorescence

Immunofluorescence analysis was conducted essentially as described.²⁷ Images of cultured cells were captured with a BZ-9000 microscope (Keyence) or an LSM-880 confocal laser microscope (Carl Zeiss). As for the cortical slice staining, brains were embedded in 3% agarose, cut into sections (100 μm-thickness) with a vibratome and photographed with an LSM-880 confocal laser microscope. Acquired images were analysed with ImageJ to determine cell morphological descriptor and fluorescence intensity.

Statistical analysis

For all cell imaging experiments, cell counting and traces were assessed in a blinded manner by a technical staff who were not aware of experimental conditions. Statistical significance for multiple comparisons was determined by Dunnett’s or Tukey’s test (https://cran.r-project.org/web/packages/multcomp/multcomp.pdf). Comparisons between two groups were performed with Welch’s t-test.³⁴P<0.05 was considered statistically significant. Statistical analyses were performed using R (https://intro2r.com/citing-r.html; https://cran.r-project.org/doc/FAQ/R-FAQ.htmlCiting-R).

Neuroimaging

Callosal anomalies were present in 10/10 participants, including hypoplasia with prevalent posterior involvement (eight cases), thick and short morphology (one case, Patient 3) and partial agenesis (one case, Patient 10) (Fig. 1C). White matter thinning and enlarged subarachnoid spaces with global reduction of the brain volume were noted in 8/10 and 9/10 cases, respectively. Malformations of cortical development were noted in 7/10 participants, including polymicrogyria involving the posterior regions (five cases), diffuse dysgyria (four cases), and small grey matter heterotopias dispersed in the cerebral white matter (one case, Patient 1). Cerebellar dysplasia with foliation abnormalities involving the inferior cerebellar hemispheres was detected in 6/10 participants, while brainstem anomalies characterized by a small midbrain and/or pons were seen in 5/10 cases. Chiari I anomaly was not detected in this series, but mild (<5 mm) caudal descent of the cerebellar tonsils was noted in three participants. Neuroimaging features of 16 cases reported in this and previous studies were summarized in Supplementary Table 2.

Identification and analyses of RAC3 variants

Based on allele frequency (<0.001 in gnomAD), involvement of conserved amino acid residues, and predicted impact on protein function, variants in RAC3 were the most plausible candidate variants in the study cohort. NGS led to the identification of seven distinct variants, including the previously reported c.184G>Ap.E62K in Patient 7 (ClinVar ID: 425149)¹⁸. Most of the novel variants were missense: c.187G>A p.D63N (Patients 1 and 9); c.191A>Gp.Y64C (Patient 3); c.179G>Ap.G60D (Patient 4); c.34G>Cp.G12R (Patient 5); c.348G>Cp.K116N (Patient 8) (Supplementary Fig. 3). All these variants are rare, affect conserved residues (GERP scores 2.5–3.91) (Supplementary Fig. 4), and are predicted to have a high pathogenic impact (CADD score 24.3–27.6) (Supplementary Table 3). The missense variant c.176C>Gp.A59G previously detected in Patient 6²⁰ is also predicted to be pathogenic by several in silico tools (Supplementary Table 3). In Patient 2, the novel in-frame deletion c.186_188delGGAp.E62del was detected. This variant results in a protein coding length change through the deletion of the conserved Glu62 (GERP score 3.79). No copy number variants with a pathogenic potential could be detected in the reported subjects (Supplementary Table 1 and Supplementary material).

A common structural and biochemical feature across the small GTPase superfamily is the G domain, which is defined by five G boxes with certain structurally conserved amino acid residues (Supplementary Fig. 3). The G1 box/P-loop (GxxxxGKS/T, where x is any amino acid) recognizes the β-phosphate and Mg²⁺ ion of guanine nucleotides. The G2 box (T) positioned in the Switch I region interacts with the γ-phosphate and Mg²⁺, while the G3 box (DxxGQ/H/T) localizing in the Switch II region is responsible for GTP-hydrolysis. Eventually, the G4 (T/NKxD) and G5 (C/SAK/L/T) boxes make specific contacts with the guanine base. The previously reported p.P34R and p.P29L variants^18,19 lead to substitutions of evolutionarily conserved amino acids in the Switch I region, with p.P34R being also included in the effector binding region. It is of note that all variations identified in this study occurred at residues highly conserved across species (Supplementary Figs 3 and 4). Among the RAC3 variants observed in our cohort, seven variants (p.A59G, p.G60D, p.Q61L, p.E62del, p.E62K, p.D63N and p.Y64C) are mapped to the Switch II region, which is recognized as a variation hot spot among RAC1, RAC3 and CDC42.^16,35,36 In addition, the p.G12R and p.K116N are the first variants affecting the conserved residues in the G1 and G4 boxes, respectively (Supplementary Figs 3 and 4).

In vitro biological and biochemical activities of the 11 disorder causative RAC3 variants

To get further insights into the pathophysiological significance of RAC3 variants associated with NDDs, activation states were first examined in vitro for the eight variants identified in the enrolled patients, together with the recently reported p.P34R¹⁹, p.P29L and p.Q61L¹⁸ variants. Of note, the p.Q61L variant is best characterized as a GTPase-defective constitutively active version of RAC1 involved in tumour cell metastasis and invasion.^37–39

When control pCAG-Myc vector was expressed in primary cultured hippocampal neurons, cells were normally differentiated in a time-dependent manner (Fig. 2A and N). Likewise, neurons expressing wild type RAC3 were normally differentiated with single axon and numerous dendrites, although its expression induced lamellipodia at the growth cone of axon and cell body (Fig. 2B). These results indicate that the basal activity of exogeneous wild type RAC3 had little effect on cell differentiation in vitro. In contrast, neurons transfected with the 11 studied variants displayed cell rounding with typical lamellipodia formation and less neurite extension, resulting in a prominent increase in the solidity (Fig. 2C–N). While the median of solidity was shifted upward when each variant was expressed, the degree was variant-dependent because cells expressing some variants such as RAC3-P29L, -A59G, -G60D and -K116N occasionally showed neurite extension (Supplementary Fig. 5). Taken together, all the 11 variants appeared to facilitate cytoskeletal reorganization to form lamellipodia to greater or lesser degrees. The respective variants should dysregulate intracellular signalling in variant-dependent manners and adversely affect neuronal morphology and function.

Open in a separate window

Figure 2

Effects of the disease-causative 11 RAC3 variants on neuronal morphology in vitro. (A–M) Dissociated hippocampal neurons from E16 mice were co-electroporated with pCAG-EGFP (0.1 μg) together with pCAG-Myc (−), pCAG-Myc-RAC3 (wild-type, WT), -G12R, -P29L, -P34R, -A59G, -G60D, -Q61L, -E62del, -E62K, -D63N, -Y64C and -K116N (0.3 μg each). Cells were fixed at 3 days in vitro (DIV), and co-stained with anti-GFP (green), rhodamine-phalloidin (red) and DAPI (blue). Scale bar = 20 μm. (N) Quantification of (A–M). The morphological descriptor (solidity) of GFP-positive neurons (≥200 cells each) was shown in violin plot with boxplot. ‘Solidity’ is the ratio of the area of a cell to the area of a convex hull of the cell. Values of the ratio were calculated by ‘Analyze Particles…’ method in ImageJ (https://imagej.nih.gov/ij/docs/menus/analyze.html). The significance of difference between (−) and each variant was determined using Dunnett’s test. WT versus (−), P=1; G12R versus (−), P<1 × 10⁻¹⁰; P29L versus (−), P<1 × 10⁻¹⁰; P34R versus (−), P<1 × 10⁻¹⁰; A59G versus (−), P<1 × 10⁻¹⁰; G60D versus (−), P<1 × 10⁻¹⁰; Q61L versus (−), P<1 × 10⁻¹⁰; E62del versus (−), P<1 × 10⁻¹⁰; E62K versus (−), P<1 × 10⁻¹⁰; D63N versus (−), P<1 × 10⁻¹⁰; Y64C versus (−), P<1 × 10⁻¹⁰; K116N versus (−), P<1 × 10⁻¹⁰. ***P<0.001.

We then performed biochemical analyses and measured GTP/GDP-exchange and GTP-hydrolysis activity. When the effects of these 11 variations on GTP/GDP-exchange activity were examined with recombinant RAC3 proteins, the assay documented a variably increased or decreased activity. When compared with wild type RAC3, log₁₀(k_obs: observed rate constant) was especially high for RAC3-K116N and moderately high for RAC3-Q61L, -P29L, -Y64C, -E62del, -P34R and -D63N in this order, indicating an accelerated exchange reaction (Fig. 3A and Supplementary Fig. 6A). In contrast, a statistically significant decrease of log₁₀(k_obs) was observed for RAC3-G60D, indicating suppression of the exchange reaction (Fig. 3A and Supplementary Fig. 6A). The exchange activity of RAC3-G12R, -A59G and -E62K did not show statistical difference in comparison to the wild type (Fig. 3A and Supplementary Fig. 6A). Then, GTP-hydrolysis activity was assayed for each variant. This was highly suppressed by p.Q61L, p.G60D and p.G12R, while moderate to weak suppression was observed for p.E62del and p.A59G (Fig. 3B and Supplementary Fig. 6B). We consider that p.G12R, p.G60D and p.Q61L primarily activated the protein by eliminating the GTP-hydrolysis activity like oncogenic RAS proteins. In contrast, the activity of RAC3-P34R was higher than the wild-type (Fig. 3B and Supplementary Fig. 6B). Of note, RAC3-P29L, E62K, -D63N, -Y64C and -K116N did not affect GTP-hydrolysis activity when compared to the wild-type (Fig. 3B and Supplementary Fig. 6B).

Open in a separate window

Figure 3

Characterization of activation states of the disease-causative 11 RAC3 variants in vitro. (A) Measurement of GDP/GDP-exchange activity. Recombinant His-tag-fused RAC3 (wild-type, WT) and the 11 variants (RAC3-G12R, -P29L, -P34R, -A59G, -G60D, -Q61L, -E62del, -E62K, -D63N, -Y64C and -K116N) were preloaded with fluorescent ^mantGDP and incubated with unlabelled GTP. Intrinsic nucleotide exchange was measured and ^mantGDP-dissociation rates of wild-type and respective variants were calculated as observed rate constants [k_obs (×10⁻⁵ s⁻¹)] from the results in Supplementary Fig. 5A. Number of replicates, n≥4. The significance of difference between wild-type and each variant was determined using Dunnett’s test. G12R versus WT, P=0.8305; P29L versus WT, P<0.001; P34R versus WT, P<0.001; A59G versus WT, P=0.7655; G60D versus WT, P=0.0135; Q61L versus WT, P<0.001; E62del versus WT, P<0.001; E62K versus WT, P=0.1018; D63N versus WT, P=0.0132; Y64C versus WT, P<0.001; K116N versus WT, P<0.001. *P<0.05, ***P<0.001. (B) Measurement of GTP-hydrolysis activity. Intrinsic GTP-hydrolysis activity was analysed by direct measuring of changes in the GTP concentration with GTPase-Glo assay kit. EC50 (half maximal effective concentration) was estimated from the sigmoidal fitting curve in Supplementary Fig. 5B. Number of replicates, n≥3. P-value was calculated as in (A). G12R versus WT, P<1 × 10⁻⁰⁴; P29L versus WT, P=0.174; P34R versus WT, P<1 × 10⁻⁰⁴; A59G versus WT, P=0.012; G60D versus WT, P<1 × 10⁻⁰⁴; E62del versus WT, P<1 × 10⁻⁰⁴; E62K versus WT, P=0.514; D63N versus WT, P=0.743; Y64C versus WT, P=0.952; K116N versus WT, P=0.833. *P < 0.05, ***P < 0.001. P.C. = positive control. (C) Summary of the results in A and B. → = no change; ↑ = activity increased; ↓ = activity decreased.

Based on the obtained results summarized in Fig. 3C, it is possible to categorize the biochemical properties of the 11 variants into three groups based on GTP/GDP-exchange activity. Group I encompasses RAC3-P29L, -P34R, -Q61L, -E62del, -D63N, -Y64C and -K116N. These variants demonstrated a higher GTP/GDP-exchange activity. While GTP-hydrolysis activity of RAC3-P29L, -D63N, -Y64C and -K116N was comparable to that of wild-type, RAC3-Q61L and -E62del showed lower hydrolysis activity. We assume that these seven variants are prone to be in GTP-bound active conformation, since they exhibited high GTP/GDP-exchange activity with GTP-hydrolysis activity equivalent to or lower than that of wild-type RAC3. On the other hand, although RAC3-P34R has a statistically high GTP-hydrolysis activity, we assume that the variant is activated due to the accelerated GTP/GDP exchange activity in the context of an intracellular environment in which the GTP concentration is substantially higher than that of GDP. It is notable that the substitution of the corresponding Pro34 in RAC2 with His was suggested to activate the protein based on crystal structure modeling.⁴⁰ Group II includes RAC3-G12R, -A59G and -E62K, all showing a GTP/GDP-exchange activity similar to the wild-type. Since RAC3-G12R and -A59G displayed suppressed GTP-hydrolysis activity, they were likely to prefer GTP-binding. As for RAC3-E62K with normal GTP-hydrolysis and GTP/GDP-exchange activities in vitro, it appears that abnormal interaction with unidentified GEFs and/or GAPs holds RAC3-E62K in an active state in primary neurons. In this context, RAC2-E62K has been reported to be hyper-activated through altered GEF specificity and impaired GAP function.⁴¹ Eventually, Group III includes RAC3-G60D, which may prefer GTP-binding because of its very low GTP-hydrolysis activity even if the GTP/GDP-exchange rate is also low. Based on the results in Figs 2 and ​and3,3, we concluded that all the 11 variations confer a gain-of-function phenotype with respect to the RAC3 function. Biological and biochemical diversity among variations may variously dysregulate RAC3 function in vivo, leading to variation-specific clinical phenotypes.

Possible interaction of the RAC3 variants with various downstream effectors

While the patients described in this study share core NDD phenotypes associated with callosal and cortical malformations, clinical features peculiar to each individual may be attributable to the position and type of the involved amino acids changes. Since the 11 RAC3 variants analysed in this study are considered to be differently activated based on cell biological and biochemical analyses (Figs 2 and ​and3),3), we looked into the downstream signalling pathways interacting with these variants. To this end, a pull-down assay was performed to examine the interaction of respective RAC3 variants with recombinant RBRs of various effectors including PAK1, MLK2, IRSp53, N-WASP, ROCK and RTKN. Consequently, the RAC3 variants were observed to interact with these RBRs in variant-dependent manners (Fig. 4A). The RBR of PAK1, which is crucial for neuronal migration and variated in patients with NDDs,^42–46 was associated with all the variants except RAC3-P34R. High affinity was displayed for RAC3-G12R, -Q61L, -E62del and -K116N in this order, and low affinity was observed for RAC3-D63N (Fig. 4A). The RBR of MLK2, which activates the JNK-MAP kinase pathway downstream of RAC,⁴⁷ significantly interacted with RAC3-Q61L and then -Y64C, and moderate to weak affinity was observed for all other variants (Fig. 4A). The RBR of IRSp53, an adaptor protein acting at the membrane-actin interface and related to the formation of filopodia and lamellipodia,^48,49 showed relatively high affinity towards RAC3-Q61L, -Y64C and -G12R (Fig. 4A). Although RAC3-E62del and -K116N had moderate affinity, a very weak interaction was observed with other variants (Fig. 4A). The RBR of N-WASP, a regulator of actin polymer reorganization in the cytoplasm and nucleus,^50,51 displayed relatively strong affinity to RAC3-Q61L, -Y64C and -G12R in this order, while weak to moderate affinity was detected for all other variants but -P34R (Fig. 4A). The RBR of ROCK, a key regulator of actin cytoskeleton and cell polarity,^52,53 exerted strong binding to RAC3-G12R, -Q61L, -E62del and -Y64C (Fig. 4A). While moderate interaction was shown for RAC3-A59G, -G60D and -E62K, other variants except RAC3-P34R showed a weak affinity (Fig. 4A). RTKN has been shown to regulate the septin cytoskeleton.⁵⁴ Although RTKN was first reported as a Rho-specific effector,⁵⁵ we have found that RTKN interacts with activated RAC3 but not RAC1 through the Rho-binding domain. We thus asked whether this domain, which is referred to as RBR here, interacts with the 11 variants. We determined that RTKN-RBR interacted with the 11 variants, with particularly strong affinity to RAC3-Q61L and -Y64C (Fig. 4A). Taken together, given that the respective RAC3 variants are activated versions, they are strongly suggested to hyper-activate different sets of effectors in variant-dependent manners and, thereby, dysregulate specific downstream signalling pathways in context-dependent manners. Full cropped data of western blotting were depicted in Supplementary Fig. 7.

Open in a separate window

Figure 4

Possible interaction of the disease-causative 11 RAC3 variants with various downstream effectors. (A) Interaction with the RBRs of PAK1, MLK2, IRSp53, N-WASP, ROCK and RTKN. COS7 cells were transfected with pCAG-Myc-RAC3 (wild-type, WT), -G12R, -P29L, -P34R, -A59G, -G60D, -Q61L, -E62del, -E62K, -D63N, -Y64C and -K116N (0.3 μg each). Cell lysates were prepared, and the pull-down assay was conducted as described in the ‘Materials and Methods’ section with respective GST-fused RBRs (5 μg each). Bound RAC3 proteins were detected by western blotting with anti-Myc. RAC3 activity was indicated by the amounts of RAC3 bound to respective GST-RBRs, and total cell lysates were also immunoblotted with anti-Myc for normalization (input). Relative band intensity was shown when the value of RAC3-Q61L was taken as 1.0. (B–D) Involvement of the RAC3 variations in SRF- (B), NFkB- (C) or AP1-dependent (D) gene transcription. COS7 cells were co-transfected with each luciferase expression vector together with pCAG-Myc-RAC3 (WT) or the expression vector for the 11 variants. Luciferase activity obtained with wild-type was taken as 1.0, and relative activities are shown as box plot. Number of replicates, n=4. The significance of difference between wild-type and each variant was determined using Dunnet’s test. (B) G12R versus WT, P<0.001; P29L versus WT, P=0.81326; P34R versus WT, P=0.94861; A59G versus WT, P<0.001; G60D versus WT, P=0.99999; Q61L versus WT, P<0.001; E62del versus WT, P=0.00997; E62K versus WT, P=0.27663; D63N versus WT, P<0.001; Y64C versus WT, P<0.001; K116N versus WT, P=0.99999. (C) G12R versus WT, P<1 × 10⁻⁰⁴; P29L versus WT, P=0.00159; P34R versus WT, P<1 × 10⁻⁰⁴; A59G versus WT, P<1 × 10⁻⁰⁴; G60D versus WT, P=1.00000; Q61L versus WT, P<1 × 10⁻⁰⁴; E62del versus WT, P<1 × 10⁻⁰⁴; E62K versus WT, P<1 × 10⁻⁰⁴; D63N versus WT, P<1 × 10⁻⁰⁴; Y64C versus WT, P<1 × 10⁻⁰⁴; K116N versus WT, P=1.00000. (D) G12R versus WT, P<1 × 10⁻⁰⁴; P29L versus WT, P=0.00275; P34R versus WT, P<1 × 10⁻⁰⁴; A59G versus WT, P<1 × 10⁻⁰⁴; G60D versus WT, P=0.00017; Q61L versus WT, P<1 × 10⁻⁰⁴; E62del versus WT, P<1 × 10⁻⁰⁴; E62K versus WT, P<1 × 10⁻⁰⁴; D63N versus WT, P<1 × 10⁻⁰⁴; Y64C versus WT, P<1 × 10⁻⁰⁴; K116N versus WT, P=0.99999. **P<0.01, ***P<0.001.

We next examined signal transduction pathways underlying the aberrant activation states caused by these variations. We focused on SRF-, NFkB- and AP1-mediated gene expression, since their related signalling pathways include Rho-family proteins and MAP kinases.^56–58 When effects of respective variants on SRF-dependent gene transcription were assessed, prominent transcriptional activation was observed in cells expressing RAC3-Q61L, while moderate activation was induced by RAC3-G12R, -A59G, -E62del, -D63N and -Y64C (Fig. 4B). In contrast, RAC3-P29L, -P34L, -G60D and -K116N had marginal effects. Meanwhile, NFkB-dependent gene transcription was highly increased by RAC3-E62del, -G12R and -Q61L (Fig. 4C). Noteworthy, these three variants were observed to show high affinity to PAK1-RBR (Fig. 4A). Other variants except for RAC3-G60D and -K116N, which showed no effects, demonstrated moderate NFkB-activation (Fig. 4C). Then, when we looked into the effects on AP1-mediated gene expression, each RAC3 variant induced the expression similar to that of NFkB; the RAC3 variants other than RAC3-A59G, -G60D, -Q61L and -E62K might share common signalling pathways in terms of NFkB and AP-1 (Fig. 4C and D). These results further support the hypothesis that the disease-causing RAC3 variants may drive different and/or common downstream signalling pathways, leading to variant-dependent dysregulation of cellular processes and, eventually, common as well as patient-specific clinical phenotypes.

In vivo effects of the four RAC3 variants in the Switch II region on neuronal migration and morphology during corticogenesis

Based on the data obtained from in vitro analyses, the disease-causative RAC3 variants are most likely to be activated and induce abnormal neuronal morphology. Since cell morphology is closely associated with migration, we examined the effects of the RAC3 variants on the migration of newly generated excitatory cortical neurons in vivo. We especially focused on the Switch II region and selected the p.Q61L, p.E62del, p.D63N and p.Y64C, since the region is a variation hotspot among RAC1, RAC3 and CDC42. Using an in utero electroporation-mediated acute gene transfer method, pCAG-Myc-RAC3 or pCAG-Myc vector harbouring each variant was co-electroporated with pCAG-EGFP into the progenitor cells in the ventricular zone of E14.5 embryonic brains, and localization of transfected cells and their progeny was observed at P0. Neurons expressing the control vector or pCAG-Myc-RAC3 migrated normally to the superficial layer (bin 3; layers II/III) of the cortical plate (Fig. 5A). In contrast, most cells transfected with pCAG-Myc-RAC3-Q61L, -E62del, -D63N or -Y64C remained in the ventricular and subventricular zones and the intermediate zone (bin 1) (Fig. 5A). Quantitative analyses confirmed that each variant exhibited statistically similar effects (Fig. 5B). The result that the expression of the wild type did not statistically affect neuronal cell positioning indicates that the basal activity of RAC3 had no effects on neuronal cell migration and that physiological balance between GTP- and GDP-bound states of RAC3 should be essential for the establishment of cortical architecture during corticogenesis. Based on western blotting analyses with cortical neurons where Rac3 proteins were electroporated, the expression level of each protein was found to be comparable and lower than endogenous Rac3 (Supplementary Fig. 8).

Open in a separate window

Figure 5

Effects of the four RAC3 variants in the Switch II region on excitatory neuron migration during corticogenesis. (A and C) Migration defects of neurons expressing each variant. pCAG-EGFP (0.5 μg) was co-electroporated in utero with pCAG-Myc (−), pCAG-Myc-RAC3 (wild-type, WT), -RAC3-Q61L, -E62del, -D63N or -Y64C (0.1 μg each) into the ventricular zone progenitor cells at E14.5. Coronal sections were prepared at P0 (A) or P7 (C). Coronal slices were double-stained with anti-GFP (white) and DAPI (blue) for A or triple-stained with anti-NeuN (red), anti-GFP (green) and DAPI (blue) for C. Boxed areas in the left panels were magnified in right panels (C). Scale bars = 50 μm (A), 200 μm (C, left) and 30 μm (C, right). (B) Quantification of the distribution of GFP-positive neurons in distinct regions of the cerebral cortex (bin 1–3) for each condition in A. Number of replicates, n≥5. The significance of difference between control (−) and each variant was determined using Dunnett’s test and shown in box plot. (Bin 1) WT versus (−), P=0.9498; Q61L versus (−), P=0.0045; E62del versus (−), P=0.0172; D63N versus (−), P=0.0119; Y64C versus (−), P=0.0356. (Bin 2) WT versus (−), P=0.413; Q61L versus (−), P=0.535; E62del versus (−), P=0.185; D63N versus (−), P=0.439; Y64C versus (−), P=0.133. (Bin 3) WT versus (−), P=0.212; Q61L versus (−), P<0.001; E62del versus (−), P<0.001; D63N versus (−), P<0.001; Y64C versus (−), P<0.001. ***P<0.001, **P<0.01, *P<0.05.

Although cells expressing respective variants were dominantly distributed at bin 1, a small portion of such neurons still reached the superficial layer of cortical plate (Fig. 5A and B), perhaps due to a relatively low amount of the expression vector incorporated. Given that transfection efficiency into each cell depends on the size of the cell surface area which is physically exposed to the ventricular lumen (CSF) where plasmids are injected, neurons incorporating low amount of the expression vector are supposed to undergo partial effects of the variant.

When long-term effects of expression of the four RAC3 variants were examined, we noticed the formation of neuronal cell clusters in the ventricular and subventricular zones in developing cerebral cortex at P7 (Fig. 5C). The cells incorporated in the cluster were positive for NeuN, indicating they were differentiated at abnormal positions. These cells also extended neurites in the cluster (Fig. 5C). The results obtained indicate that the four variants prevent, rather than delay, cortical neuron migration, and that RAC3 plays a pivotal role in neuronal migration.

Time-lapse imaging of migration of cortical neurons expressing the four RAC3 variants in the Switch II region

Newborn cortical neurons generated at the ventricular zone primarily exhibit multipolar shapes in the lower intermediate zone, where cells show a slow and irregular movement termed multipolar movement for ~24 h.⁵⁹ Neurons then transform into a bipolar shape with a leading process and an axon in the upper intermediate zone, move into the cortical plate, and exhibit a saltatory movement termed radial migration toward pial surface. Given the tight correlation between cell movement and morphology, the abnormal accumulation and cluster formation of the RAC3 variant-expressing neurons in the subventricular zone and intermediate zone should be associated with impaired shape change into the bipolar status (multipolar-bipolar transition). We thus carried out time-lapse imaging to further investigate the morphology of cells stuck in the subventricular zone and intermediate zone during corticogenesis. To this end, ventricular zone progenitor cells were co-electroporated with pCAG-EGFP together with wild-type RAC3- or respective variant-expression vectors at E14.5, and cell migration was monitored from E16 for 15 h in the subventricular zone, intermediate zone and lower cortical plate. Consequently, time-lapse imaging revealed clear differences in migration profiles between control cells and those expressing the respective variants. In the control experiments with wild-type RAC3, while a considerable number of GFP-positive cells were positioned in the intermediate zone, many cells advanced slowly toward the pial surface (Fig. 6A and B and Supplementary Video 3). In stark contrast, cells expressing RAC3-Q61L, -E62del, -D63N or -Y64C were stuck in the intermediate zone. When looking closer, cells expressing RAC3-E62del or -D63N remained round and appeared not to obtain the multipolar status (Fig. 6A and B and Supplementary Videos 4 and 5). On the other hand, cells expressing RAC3-Q61L or -Y64C were observed to transform into the multipolar shape but then fail to undergo the multipolar-bipolar transition (Fig. 6A and B and Supplementary Videos 6 and 7). Quantification analyses revealed that migration distance of the variant-expressing neurons at the intermediate zone–cortical plate boundary was much shorter than that of the control cells (Fig. 6C). The average migration speed of neurons expressing each RAC3 variant was also reduced when compared to that of control cells (Fig. 6D).

Open in a separate window

Figure 6

Time-lapse imaging analyses of migration of cortical neurons expressing the four disease-causative RAC3 variants in the Switch II region. Electroporation was performed as in Fig. 5. Analyses were repeated three times for each case and representative results were shown in (A and B). (A) Tracing of neurons expressing Myc-RAC3 (wild-type, WT), -RAC3-Q61L, -E62del, -D63N or -Y64C in upper intermediate zone (IZ)–lower cortical plate (CP). Migratory tracks of over 100 cells were demonstrated as colour lines. (B) Time-lapse imaging of neurons, which express Myc-RAC3 (WT) or the variants, migrating in the IZ–CP boundary. (C) Migration distance of neurons expressing Myc-RAC3 or the four variants. Over 100 cells selected in (A) were analysed. The distance was shown in violin plot with box plot. The significance of difference between wild-type and each variant was determined using Dunnett’s test. Q61L versus WT, P<2 × 10⁻¹⁶; E62del versus WT, P<2 × 10⁻¹⁶; D63N versus WT, P<2 × 10⁻¹⁶; Y64C versus WT, P<2 × 10⁻¹⁶. ***P<0.001. (D) Migration velocity of cells expressing Myc-RAC3 or the four variants. Over 100 cells selected in (A) were analysed. The velocity was shown in violin plot with box plot. The significance of difference between wild-type and each variant was determined using Dunnett’s test. Q61L versus WT, P<1 × 10⁻⁰⁷; E62del versus WT, P<1 × 10⁻⁰⁷; D63N versus WT, P<1 × 10⁻⁰⁷; Y64C versus WT, P=1.16 × 10⁻⁰⁷. ***P<0.001.

Role of PAK1 as a downstream effector of RAC3 in neuronal migration in vivo

Since PAK1 is a downstream effector for RAC3 and activating PAK1 variants cause NDD,⁴⁵ dysregulation of this kinase is most likely to be involved in the pathogenesis of RAC3 variants. We thus investigated the possible involvement of PAK1 in the migration defects caused by the four RAC3 variations localized in the Switch II region. When pCAG-EGFP was co-electroporated with pCAG-Myc-RAC3-D63N, -E62del or -Y64C, together with pCAG-Flag-PAK1KA encoding a kinase-negative version of PAK1, the positional defects of GFP-positive cells were partially rescued at P0 (Fig. 7A–C). The observed rescue effects by PAK1KA confirmed that these three variants cause a hyper-activation of PAK1, which is supposed to be a crucial pathogenic mechanism of RAC3-related disorder caused by these three variations. To confirm the variant-mediated activation of PAK1, these three variants were expressed with PAK1 in COS7 cells, or without PAK1 in primary cultured cortical neurons. Consequently, activation of exogenous and endogenous PAK1 was observed in COS7 cells and cortical neurons, respectively (Supplementary Fig. 9A, B, D and E). In addition, when these variants were electroporated into embryonic mice brains, endogenous PAK1 activation was again detected in cortical neurons (Supplementary Fig. 9C and F). On the other hand, PAK1KA had no effects on the migration defects produced by RAC3-Q61L under the same experimental conditions (Fig. 7D). This result implicates the possibility that other downstream effector(s) is also involved in the pathogenicity of RAC3-related disorder. To answer this question, we selected another kinase MLK2, an activator for broad MAP kinase cascades including JNK (c-Jun N-terminal kinase), ERK (extracellular signal-regulated kinase) and p38,⁴⁷ as the next candidate effector molecule involved in RAC3-Q61L-mediated pathophysiological mechanism. However, MLK2 failed to rescue the aberrant migration phenotype by RAC3-Q61L (Fig. 7D), indicating that yet unidentified downstream effector(s) plays a crucial role in the variation-mediated signalling dysregulation.

Open in a separate window

Figure 7

Rescue effects by a dominantly negative PAK1 on migration defects caused by the four variants in the Switch II region. pCAG-Myc-RAC3-E62del (A), -D63N (B) or -Y64C (C) (0.1 μg each) was co-electroporated with pCAG-EGFP (0.5 μg) together with pCAG-Flag vector (1.0 μg, control) (left) or pCAG-Flag-PAK1KA (1.0 μg) (right). (D) pCAG-Myc-RAC3-Q61L (0.1 μg) was co-electroporated with pCAG-EGFP (0.5 μg) together with 1.0 μg each of pCAG-Flag vector (control) (left), pCAG-Flag-MLK2KN (middle) or pCAG-Flag-PAK1KA (right). Analysis was done as in Fig. 5. Quantification results of the distribution of GFP-positive neurons in distinct regions (bin 1–3) of the cerebral cortex were shown in boxplot for each condition. Number of replicates, n≥5. The significance of difference between control and each rescue condition was determined using Welch’s t-test (A–C) or Dunnett’s test (D), and shown in boxplot. (A, bin 1) E62del versus +PAK1KA, P=0.005591. (A, bin 2) E62del versus +PAK1KA, P=0.0006076. (A, bin 3) E62del versus +PAK1KA, P=0.01177. (B, bin 1) D63N versus +PAK1KA, P=0.0007075. (B, bin 2) D63N versus +PAK1KA, P=0.0009054. (B, bin 3) D63N versus +PAK1KA, P=0.001021. (C, bin 1) Y64C versus +PAK1KA, P=0.005963. (C, bin 2) Y64C versus +PAK1KA, P=0.04256. (C, bin 3) Y64C versus +PAK1KA, P=0.003158. (D, bin 1) Q61L versus +PAK1KA, P=0.338; Q61L versus +MLK2KN, P=0.807. (D, bin 2) Q61L versus +PAK1KA, P=0.311; Q61L versus +MLK2KN, P=0.993. (D, bin 3) Q61L versus +PAK1KA, P=0.393; Q61L versus +MLK2KN, p=0.673. ***P<0.001, **P<0.01, *P<0.05; Scale bar = 50 μm.

We thank Mses Noriko Kawamura and Nobuko Hane for technical assistance. We also acknowledge the Telethon Undiagnosed Disease Program (TUDP) for technical and scientific support.