Introduction

Rho GTPases are small proteins belonging to the RAS superfamily functioning as signal transducers in pathways that control cell proliferation, differentiation and survival, being key regulators of actin polymerization [1]. The Rho GTPases family can be divided into eight subgroups comprising at least 22 members [2]. They are molecular switches cycling between active (GTP-bound) and inactive (GDP-bound) states through conformational changes in their Switch I and Switch II regions, and are characterized by a C-terminal CAAX motif, which undergoes post-translational modifications [2, 3]. Dysfunction of Rho GTPases and their regulators is a key event in various pathological processes, including oncogenesis and clinical evolution of adult-onset neurodegenerative diseases [4, 5].

Among Rho GTPases, the RAS-related C3 Botulinum toxin substrate (RAC) subfamily comprises 3 proteins (RAC1 to 3), which share about 90% of homology in their amino acid sequence [6]. RAC1 (MIM 602048) is one of the major regulators of actin cytoskeleton remodeling, controlling cell migration, axon outgrowth, dendritic arborization, and synaptogenesis [7, 8]. RAC1 is ubiquitously expressed [9]; differently, RAC2 (MIM 602049) is almost confined to hematopoietic cells, and RAC3 (MIM 602050) is specifically expressed in both developing and adult nervous system [10, 11].

Germline mutations affecting members of the RAC family have been causally linked to clinically heterogeneous developmental disorders [3, 7, 12, 13]. Heterozygous and biallelic pathogenic RAC2 variants cause primary immunodeficiency with variable presentation, ranging from granulocytes defects to combined immunodeficiency (MIM 618987, 608203 and 618986) [13]. RAC3 is mutated in an autosomal dominant neurodevelopmental disease (NDD) characterized by structural brain anomalies and facial dysmorphisms (NEDBAF; MIM 618577) [12]. Similarly, mutations in RAC1 underlie an autosomal dominant NDD characterized by a wide spectrum of clinical features, including facial dysmorphisms and intellectual disability (ID) (MRD48; MIM 617751) [7, 14, 15]. To date, missense variants affecting seven conserved residues located throughout the protein have been observed in eight unrelated individuals. In vitro functional characterization of a subset of variants demonstrated opposite effects on cell morphology, with two mutants (i.e., C18Y and N39S) having a dominant negative (DN) behavior, and one (i.e., Y64D) working as a constitutively active (CA) protein, with mirroring effects on filopodia/lamellipodia/ruffle protrusions and cellular roundness index (RI) [7]. The remaining mutants could not be clearly classified in the same experimental model although exhibiting some functional intermediate effect on both cellular morphology and RI. Among the studied variants, the C18Y substitution was also documented to prevent GTP-mediated RAC1 activation, with an increased dendritic spine density eventually precluding synaptic transmission [16].

Mirroring the molecular complexity of RAC1 mutations, affected individuals show a noteworthy phenotypic variability with different degrees of DD/ID, neurological abnormalities, and facial dysmorphisms [7, 15]. The most remarkable clinical finding is the stark variability in the occipital-frontal circumference (OFC), ranging from −5.0 to +4.5 SD, with a few subjects showing normal values (Table 1). This heterogeneity correlates with specific mutations, as the DN mutants (C18Y and N39S), which were reported in subjects with microcephaly, have also been demonstrated to induce a significant decrease in head size in zebrafish embryos [7].

Table 1 Clinical and molecular data of report individuals and present cases.
Full size table

Here, we describe two additional individuals with previously unreported de novo missense variants in RAC1 and a NDD characterized by normal OFC, facial dysmorphisms, mild/moderate DD/ID and a variable range of additional signs, confirming the clinical variability of MRD48. For both variants, we demonstrate their gain-of-function (GoF) impact. We clinically review the disorder and deeply delineate the common clinical and facial features. We also compare the main clinical presentations of MRD48 and NEDBAF, highlighting similarities and differences. Finally, we revise the mutation spectrum in RAC1-3 and CDC42, the latter being an additional Rho GTPase sharing considerable homology with RAC proteins previously associated with a highly heterogeneous phenotypic presentation [17, 18], evidencing the occurrence of a recurring pattern of mutational hotspots.

Materials and methods

Subjects and molecular analysis

Both subjects were enrolled in the context of a research program aimed at understanding the molecular causes of unclassified pediatric disorders at the Ospedale Pediatrico Bambino Gesù, Rome, Italy. The study has been approved by the local Institutional Ethical Committee (ref. 1702_OPBG_2018). Clinical data, pictures, DNA samples and other biological specimens were collected, used, and stored after signed informed consents from the participating subjects/families were secured. Permission to publish the clinical pictures was obtained for both individuals.

A trio-based whole-exome sequencing (WES) was performed on both subjects after the exclusion of any clinically relevant structural variant by high-resolution CMA analysis. Genomic DNA of the affected subjects and their parents was extracted from leukocytes and sequenced using Illumina paired-end technology coupled with the SureSelect AllExon V.7 (Aligent Technologies, Santa Clara, CA) enrichment kit. Sequencing protocol, WES statistics, reads alignment and variant calling pipelines, data processing and output are detailed in Supplementary Materials and Methods and reported in Table S1. Selected variants were validated via Sanger sequencing following standard protocols. The identified pathogenic RAC1 variants have been submitted to ClinVar (SCV002605544, SCV002605545).

Structural analysis

The crystal structure of RAC1 in complex with the KALRN DH1 domain (PDB accession: 5O33) was employed to display the sites affected by the identified amino acid substitutions. Molecular graphs were made with PyMOL (www.pymol.org).

Cell lines and cultures

HEK-293T and COS-1 cell lines were obtained from American Type Culture Collection (ATCC) and grown in high-glucose Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% fetal calf serum, 2 mM L-glutamine, 10 U/ml penicillin/streptomycin, and 1 mM sodium pyruvate (Merck, Darmstadt, Germany).

RAC1 levels and GTP binding assay

Wild-type human RAC1 (NM_006908.4) was cloned into a pcDNA3 backbone with an N-terminal HA tag. Mutant constructs expressing the E62K, A159T, A159V or G12V amino acid substitutions were generated by site-directed mutagenesis using the QuikChange XL kit (Agilent Technologies), and verified by direct sequencing. Each of the HA-tagged RAC1 constructs or the empty vector was transfected using Fugene 6 (Roche, Basel, Switzerland). Two days after transfection, cells were assayed for RAC1 levels, GTP-bound state and function, which were verified by western blot and pull-down assays (see Supplementary Materials and Methods) or processed for immunofluorescence analysis. For protein stability studies, cells were treated up to 6 h with MG132 (50 μM) and then lysed for RAC1 levels assays.

Cell shape assay by immunofluorescence and confocal microscopy analysis

Immunofluorescence analyses were performed as previously described [19]. Complete protocol is detailed in Supplementary Materials and Methods. Confocal laser scanning microscopy observations were performed with a LSM980 apparatus (Zeiss, Oberkochen, Germany), using a 63x/1.40 NA oil objective and excitation spectral laser lines at 405, 488, 594 nm. Cell roundness was calculated by using the formula roundness index = \(\frac{{perimeter}}{{2\sqrt {\pi \;area} }}\), as previously reported [20]. The mean perimeter and area of cells were calculated using the Image J software [21], drawing the region of interest based on f-actin fluorescence delimited borders. At least 20 cells were analyzed for each dataset. Data were pooled from three independent experiments and highly expressing cells were excluded.

Statistical analysis

Data were represented as means ± SEM. Statistical analyses were performed using GraphPad Prism version 5.0. (GraphPad Software, San Diego, CA). GTPase function and RI measurements were analyzed by multiple t-test, corrected for multiple comparisons using Holm-Sidak method and Chi-square’s test in 2 × 2 contingency tables (Fisher’s exact test), respectively. Statistical significance is indicated with asterisks: *p < 0.05, ***p < 0.001, ****p < 0.0001. Comparisons of clinical features were calculated by using the Fisher’s exact test.

Results

Genomic analyses

Molecular analyses identified two de novo missense variants, c.475G>A (p.Ala159Thr [A159T], subject 1) and c.184G>A (p.Glu62Lys [E62K], subject 2) within exons 7 and 3 of RAC1 (NM_006908.5, NP_008839.2). Sanger sequencing of probands’ DNA specimens collected from additional tissues proved both variants to be germline. The two variants had not previously been reported in neither public (e.g., gnomAD v.2.1.1) nor in-house (>2800 exomes) population databases, and were classified as pathogenic according to the ACMG/AMP criteria [22] (Supplementary Table S2). All tested in silico pathogenicity scores pointed out these changes as disease-causing. Consistently, a strong evolutionary constraint for the affected amino acids was evidenced, further supporting a deleterious effect (90% classification accuracy) of the reported changes [23].

In Subject 2, the segregation from an affected father of hearing impairment suggested a possible double diagnosis to be investigated. Filtering for private or low-frequency variants in genes associated with deafness allowed to identify a nonsense variant in GRHL2 (c.108G>A, p.Trp36*; NM_024915.4), which was classified as pathogenic according to the ACMG guidelines. Loss-of-function (LoF) variants in GRHL2 are associated with a form of late-onset autosomal dominant non-syndromic hearing loss (DFNA28; MIM 608641).

Functional characterization of the E62K and A159T substitutions

To investigate the effects of E62K and A159T amino acid changes on RAC1 function, we analyzed protein levels in lysates of COS-1 cells transiently transfected with HA-tagged RAC1 wild-type (WT), each of the two mutants, and RAC1 proteins carrying the A159V (oncogenic) or G12V (CA) variants [24, 25]. Western blot analysis documented similar levels of RAC1A159T with respect to the WT and RAC1A159V proteins, while showed a strong reduction of RAC1E62K and RAC1G12V (Fig. 1A). Treatment with the proteasome inhibitor, MG132, recovered levels of RAC1E62K, indicating an accelerated degradation of this mutant by the proteasome machinery (Fig. 1B). Next, we assayed the levels of active, GTP-bound WT and mutant RAC1 by performing pull-down assays on lysates from HEK-293T cells transfected with HA-tagged RAC1 mutants. Similar increased levels of normalized GTP-bound RAC1E62K and RAC1A159T proteins with respect to WT RAC1 were observed, which however were lower than those characterizing the RAC1A159V and RAC1G12V mutants (Fig. 1C). Considering the absolute GTP-bound RAC1 levels of proteins, we observed that RAC1E62K behaved as a weakly constitutively activated GTPase, while RAC1A159T showed a full CA behavior, similar to that characterizing RAC1A159V and RAC1G12V (Fig. 1D). To further confirm the GoF effect of the two RAC1 variants, we tested their impact on cell morphology. As shown in Fig. 2A, WT RAC1-expressing cells exhibited an irregular, variable shape, with only a few round cells, similarly to what was observed in cells expressing the DN RAC1T17N protein [26]. Conversely, the expression of all the other tested mutants variably promoted cell rounding, with a weaker effect exerted by RAC1E62K, supporting their activating impact. To quantitate these morphological changes, we evaluated the RI of cells (Fig. 2B), documenting a significantly lower RI in mutated cells, with RAC1A159T acting as the most activated between the two mutants (Fig. 2C). Overall, these data validate the GoF behavior of the two disease-causing variants and provide evidence for their differential impact on protein levels and function.

Fig. 1: RAC1 mutations variably affect protein stability and GTP binding in COS-1 transfected cells.
figure 1

A Western blot analysis of exogenous RAC1 levels in COS1 transfected cells. HA-tagged RAC1 fold-level changes are calculated with respect to cells overexpressing the WT protein. Data are presented as a mean fold change ± SEM (n = 3). Statistical significance of differences were assessed by t-test, corrected for multiple comparisons using Holm-Sidak method (***p < 0.001). B RAC1E62K undergoes accelerated degradation via the proteasome machinery. HA-tagged RAC1E62K changes are compared with cells overexpressing the WT protein. Data are presented as a mean fold change ± SEM (n = 3). Statistical significance of differences was assessed by t-test, corrected for multiple comparisons using Holm-Sidak method (*p < 0.05). C, D Representative western blot analysis of proteins after pull-down assay and bar charts of the mean fold change values of three independent experiments. Pull down of the GTP-bound HA-RAC1 from the total protein lysates is performed using a GST-human Pak1-(PBD). Immunoblot analysis of total cell lysates identified the levels of total protein (Total HA-RAC1). GTP-bound HA-tagged RAC1 level-fold changes are calculated with respect to cells overexpressing WT protein. In C, GTP-bound RAC1 levels were also normalized with respect to total HA-RAC1 levels. Differences are statistically significant with respect to WT RAC1-expressing cells. Data are presented as a mean fold change ± SEM (n = 3). Significance is determined by multiple t-test, corrected for multiple comparisons using Holm-Sidak method. GAPDH, glyceraldehyde-3-phosphate dehydrogenase.

Full size image
Fig. 2: RAC1 mutations affect cytoskeletal rearrangement and cell shape.
figure 2

A Representative confocal images of COS-1 cells expressing HA-tagged RAC1 WT or mutants cultured for the indicated period on fibronectin matrix and stained to detect HA-RAC1 (magenta), f-actin (green) and nuclei (blue). B Schematic representation of cells with respect to their shape and related roundness index (RI): lower values of RI correspond to higher cell roundness. RI was calculated as described in materials and methods and cells were grouped into four different morphological classes (round, 1.0 < RI < 1.25, partially round, 1.25 < RI < 1.50, partially complex, 1.50 < RI < 1.75 and complex, RI > 1.75). C Incidence of cell morphotypes. Bar graph representing the percentage of cells having different morphologies in HA-tagged RAC1 transfected cells 30 min after seeding on fibronectin matrix, based on the classification described above in B. Data are plotted as a percentage of the cells analyzed from the mean of three independent experiments. Two sided Fisher’s exact test in 2 × 2 contingency table is used (WT vs each mutant and each cellular type vs the other morphologies). Multiplicity adjusted p-values are reported. Round and complex morphotypes were extremely significant (****p < 0.0001) for all mutants except T17N. E62K, partially round and partially complex, not significant. A159T, A159V and G12V, partially complex, p < 0.001; A159T and G12V, partially round, p = 0.072; A159V, partially round not significant. T17N, complex, partially round and round, not significant; T17N partially complex p = 0.016. N ≥ 20. Bar = 10 μm.

Full size image

Structural analysis

To further explore the effect of the amino acid changes on protein function, we performed structural modeling analyses. A159 is a highly conserved residue mapping at the nucleotide-binding pocket (Fig. 3A). Specifically, it directly contributes to the stability of nucleotide binding by interacting with the purine ring of GTP/GDP. The alanine-to-threonine substitution at codon 159 is expected to locally alter the conformation of the nucleotide binding pocket and consequently RAC1 affinity for GTP/GDP. Of note, mutations affecting this region in RAS proteins have been proved to exert an activating effect on protein function both in cancer and in RASopathies by promoting the GEF-independent release of GDP [27,28,29].

Fig. 3: Structural impact of the pathogenic RAC1 amino acid substitutions and clinical pictures of affected individuals.
figure 3

A Sequence alignment and structure of RAC1. Multiple sequence alignment around residues, Glu62 and Ala159, in homologs and orthologs, and their positions in the crystal structure of human RAC1 complexed with the GEF kalirin DH1 domain (PDB 5O33). Glu62 and Ala159 are shown as sticks and balls with red surface, interacting residues as sticks, and guanosine-5’-diphosphate (GDP) in magenta. Critical interactions by the two residues are indicated by dotted lines. B Facial features of Subject 1 (upper line) and Subject 2 (lower line) at different ages. Note the high anterior hairline, arched eyebrows with tendency to lateral sparse (subject 2), mild synophrys, prominent nasal bridge, mild widely spaced eyes, wave-shaped palpebral fissures, overhanging columella, short philtrum, deep nasolabial folds thin upper lip (Subject 2), abnormally spaced teeth, everted lips, and long pointed chin.

Full size image

E62 is adjacent to Q61, a key residue mediating GTP hydrolysis representing one of the major mutation hotspots in cancer [28, 29], and is involved in multiple polar and electrostatic interactions with residues binding the phosphate group of GTP/GDP (Fig. 3A). It also contributes with interactions that are relevant for the proper conformation of the surface mediating binding to signaling partners, which are predicted to be perturbed upon replacement with the oppositely charged lysine (Fig. 3A). Thus, besides the activating effect, the E62K substitution is expected to have a more complex behavior in which opposite perturbing effects (i.e., disrupted GTP hydrolysis vs protein structural rearrangement affecting stability and protein-protein interactions) result in a mildly activated mutant, in line with the experimental findings.

Clinical findings

The natural history and clinical features of both subjects are detailed in Fig. 3B, Table 1, Supplementary Table S3, and Supplementary Clinical Reports.

Subject 1 is a 14-year-old boy with speech disorder and mild ID, multiple congenital heart defects (CHD), and facial dysmorphisms. He presented with hydronephrosis, and right inguinal hernia. Since 12 years old, he showed increasing difficulties in social interaction, worsening of intellectual competences, and behavior anomalies characterized by impulsivity, compulsivity, and anxious component.

Subject 2 is a 27-year-old woman with speech disorder, mild ID, facial dysmorphisms and peculiar behavior. Family history reveals late-onset deafness in her father (20 years), which was treated with hearing aids at 45 years. At 6 years, she was diagnosed with severe neurosensory deafness treated with hearing aids. A brain MRI scan revealed mild posterior corpus callosum hypoplasia and left fronto-insular pachygyria. At our observation (15 years), she presented with behavior anomalies characterized by impulsivity, compulsivity, stubbornness, anxious component, unexplained fears, and panic attacks.

We systematically assessed the clinical and facial features for the ten individuals with pathogenic RAC1 variants, which are summarized in Table 1 and Supplementary Table S3. A relatively homogenous presentation in the absence of any obvious genotype-phenotype correlation was apparent, except for macrocrania in two individuals carrying pathogenic variants involving the residue V51. Overall, all subjects shared ID with speech delay (10/10 and 7/7, respectively), a wide array of brain abnormalities (8/8) (Table 1 and Supplementary Table S4), behavioral abnormalities (5/6), hypotonia and neonatal/childhood feeding difficulties (5/7 and 5/8, respectively). CHDs were also common (4/8), while epilepsy was reported in 3/8 subjects. We also observed recurrent ectodermal anomalies, mainly represented by eczematous rashes, with or without ichthyosiform manifestations (4/10), and various skeletal anomalies (3/10), such as small hands and feet (3/10). Two individuals presented with congenital sensorineural deafness (subject 2 from the present report and subject E from [7]), and two showed hydronephrosis (subject 1 from the present report and subject K from [15]) (Table 1). We evaluated facial dysmorphisms in all the individuals whose available clinical pictures could be blind reviewed by three experienced clinical geneticists (MP, FCR and DM) (Supplementary Table S3). Common features included a high anterior hairline (9/9), arched eyebrows (8/9) with tendency to lateral sparse (7/9) and mild synophrys (4/9), a prominent nasal bridge (7/9), mild widely spaced eyes (8/9), wave-shaped palpebral fissures (7/9), overhanging columella (6/9) with a bulbous nasal tip (6/9), a short philtrum (7/9), deep nasolabial folds (8/9), thin upper lip (7/9), abnormally spaced teeth with or without diastema of superior incisors (6/6), everted lips (both superior and inferior) (5/9), and a long pointed chin (7/9).

Clinical profiling of RAC1 and RAC3-related disorders

The comparison among main clinical features associated with RAC1 and RAC3 variants revealed a common core of signs/symptoms characterized by DD/ID, brain abnormalities, hypotonia, failure to thrive/feeding difficulties, seizures, and behavior anomalies (Supplementary Table S4). Although reported in both cohorts, differences in MRI anomalies and ID were evidenced both in frequencies and severity. Facial dysmorphisms were quite distinctive between the two conditions. Although affected individuals shared high anterior hairline, arched eyebrows with tendency to lateral sparse and mild synophrys, a prominent nasal bridge, and a long pointed chin, RAC1 mutated subjects presented with wave-shaped palpebral fissures, short philtrum with overhanging columella, and a thin upper lip, while RAC3 mutated individuals showed upslanted palpebral fissures with short nose and upturned nares [3].

RAC1-3/CDC42 mutation spectrum

Rho GTPase superfamily is characterized by a highly conserved common structure of “G domain” characterized by five α helices, six β strands and five loops, also known as “G boxes”, which are the main mediators in GTP/GDP binding. RAC1-3 and CDC42 proteins show high similarity in G boxes consensus sequence, localization, and length: G1 box (P-loop/Walker A) (residues 10–17 in RAC3), G2 box (effector domain/Switch 1) (residues 25–40), G3 box (Walker B motif/Switch II) (residues 57–61), G4 box (residues 115–118), and G5 box (residues 158–160) [3, 30, 31]. However, slight differences in boundaries of these domains do exist among proteins. We revised all the pathogenic variants in these GTPases (Fig. 4), confirming a previously recognized hotspot in Switch II and highlighting a novel hotspot in Switch I involving highly conserved residues among orthologs and paralogs, located between residues 28–39. Notably, recurring variants at specific residues (E62 and Y64 in G3/Switch II) were reported in all proteins. We also evidenced specific recurring changes in G1 box (G12) and G2/Switch I box (P29 and P34) in RAC2, RAC3 and CDC42, and in G5 box (C157 and A159) in RAC1 and CDC42, respectively.

Fig. 4: Mutation spectrum in RAC1 and related GTPases.
figure 4

Affected residues in RAC1-3 proteins (green), CDC42 (gray) and RAS GTPases (representative recurrently mutated in RASopathies and cancer, various colors) are shown together with the major functional motif/domains. Only likely pathogenic/pathogenic variants are reported. Homologous residues are aligned. NS Noonan syndrome, CFCS cardio-facio-cutaneous syndrome, CS Costello syndrome.

Full size image

Discussion

MRD48 is a rare and poorly characterized syndromic NDD caused by dominantly acting variants in RAC1, with eight individuals reported so far. Here, we describe two additional subjects with previously unreported activating variants. We review the clinical and facial features of MRD48-affected individuals, identifying a core clinical presentation characterized by DD/ID with speech delay, brain abnormalities, behavior anomalies, hypotonia, neonatal/childhood feeding difficulties, and distinctive facies. Although previous reports failed in recognizing a facial gestalt, we identify recurrent features by systematical review of available pictures. Among them, a high anterior hairline, arched eyebrows with tendency to lateral sparse, wave-shaped palpebral fissures, overhanging columella with a bulbous nasal tip, a short philtrum, deep nasolabial folds, abnormally spaced teeth, with or without diastema of superior incisors, and a long pointed chin may be suggestive for this condition (Supplementary Table S3). Notwithstanding the heterogeneous functional impact of RAC1 mutations, the clinical phenotype does not appear to correlate with the functional effect of the identified variants with the exception of the association between OFC and variants involving key residues (V51, macrocephaly; C18 and N39, microcephaly). A wider series of affected subjects is needed to confirm these findings.

Rac1 knock-out (KO) mice are embryonic lethal [32]. To overcome this constraint, multiple conditional KO and Knock-in (KI) Rac1 mouse models have been reported in order to study specific effects of increased or abolished Rac1 activity in brain and other organs development and function (Supplementary Table S5 and Supplementary Text for a comprehensive revision). Many clinical features observed in affected subjects (i.e., microcephaly, ID, neurological signs, behavior anomalies, CHD, and oropharyngeal arches anomalies as those observed in Subject 2, ectodermal abnormalities, and neurosensory deafness) fit well with the malformations/anomalies observed in these models (Supplementary Table S5 and Supplementary Text). Notably, the perturbed development of neural crest cell (NCC)-derived tissues documented in NCC-targeted Rac1 conditional KO mice may explain both craniofacial and cardiac features recurring in affected individuals (see Supplementary Text) [33,34,35]. Cardiac NCCs contribute to the formation of the anterior second heart field (SHF) from which the right ventricle, the interventricular septum, the aortic-pulmonary septum, and the outflow tract (OFT) develop, generating the aortic and pulmonary valves and trunks [35]. Of note, 5/10 MRD48 individuals presented with a CHD, mainly valve anomalies and septal defects. SHF-specific KO Rac1 mice show right ventricular non-compaction and OFT defects. Three out of five subjects with CHD carried combined associations of septal defects (ASD and/or VSD) and OFT anomalies, such as bicuspid aortic valve and pulmonary valve stenosis as observed in subject 1. Finally, we noted a changing phenotype of the facial features with age. Specifically, widely spaced eyes, bulbous nasal tip and prominent nasolabial folds tend to smoothen in adolescents and young adults. A changing phenotype is observed in various developmental disorders and is likely due to the evolution of craniofacial morphology [36].

MRD48 and NEDBAF share many neurological signs (Supplementary Table S4). This is not surprising taking into account the overlapping expression of RAC1 and RAC3 in the nervous system [37]. However, specific MRI anomalies (i.e., hypoplasia/aplasia of corpus callosum, brainstem and cerebellar hypo/dysplasia) seem much more frequent in NEDBAF with respect to MRD48; on the other hand, a subset of RAC1 mutations correlate with microcrania, suggesting that RAC1 likely plays a major role in neurogenesis and/or apoptosis, while RAC3 is primarily involved in neuronal migration [3]. Pathogenic RAC1 variants exert variable activity ranging from DN to CA [7, 16]. This is not observed in RAC3 variants, which almost invariably show a CA behavior [3, 6]. This evidence suggests that downstream interactors may be differentially regulated by RAC1 with respect to RAC3 in a cell context-dependent manner. Eventually, Rac3 can partially compensate the lack of Rac1 function in late murine developmental stages, while double conditional KO mice are almost lethal [37]. This finding may in part explain the consistent and statistically significant differences in DD/ID in the two conditions; subjects with pathogenic RAC1 variants show a milder developmental involvement (5/10 with mild to moderate DD/ID) compared to individuals with RAC3 mutations, who were all diagnosed with severe to profound ID (Supplementary Table S4). CHD and involvement of other systems are quite rare in NEDBAF; this is expected when considering the confined brain expression of RAC3 with respect to RAC1.

The revision of all reported pathogenic variants in RAC1-3 and CDC42 proteins allowed us to further confirm a variation hotspot in Switch II (G3 Box, see RAC1-3/CDC42 mutation spectrum) and highlight a new hotspot in Switch I (G2 Box, see RAC1-3/CDC42 mutation spectrum) approximately spanning from residues 28–39. The Switch I and Switch II regions are crucial domains in RAS/RHO GTPase proteins, as they are directly involved in the conformational changes controlled by their GTP/GDP-bound state [38]. The most critical motifs that governs Mg2+ and GTP/GDP binding, reside within Switch I and Switch II regions [39]. Of note, Y32 is mutated in a RAC1 positive subject [15], P34H is a recurrent variant in both RAC2 and RAC3, and P34Q has been reported as pathogenic substitution in CDC42 [40]. All the critical residues within the DTAGQ motif (Switch II) are variably mutated in RAC proteins (Fig. 4). We evidenced the occurrence of variants at specific residues (i.e., E62 and Y64 in Switch II) in almost all these GTPases. Among them, the recurrent E62K has been observed in all RAC proteins and results to be in frame deleted in RAC3 [3, 7, 12, 13]. The Y64C variant has been described in RAC3 and CDC42, as well; the Y64D substitution has been reported in RAC1 [7]. The Y64C variant has recently been recorded in ClinVar as pathogenic (VCV001333690.1) in a single subject. While this manuscript was under review, a few other individuals with likely pathogenic RAC1 variants have been reported in ClinVar (W56C, VCV001027690.1; R66S, VCV000974895.2; S71F, VCV001685419.1; L165V, VCV001176531.7). While the clinical features and pictures were not available for these subjects, the spectrum of these variants well fits with the mutational diversity and hotspots (Fig. 4). Among the other recurrently mutated residues, G12 is a key residue in RAS superfamily. Amino acid substitutions at codon 12 results in a protein that has impaired GTPase activity and is blocked in its active conformation, and represent common somatic and germline events in cancer and RASopathies, respectively [28, 29, 41, 42]. A159 corresponds to A146 in HRAS, which is mutated in Costello syndrome (MIM 218040) [43, 44]. In general, the spectrum of RAC1-3 (including the variants reported in the present report) and CDC42 mutations widely overlaps with those observed in many RASopathies and in cancer.

Overall, our data delineate MRD48 as a clinically variable NDD caused by pathogenic variants in RAC1 with variegated effects on protein function. Both clinical review and functional evidence suggest a highly heterogeneous set of consequences due to RAC1 dysregulation. The pleiotropic and complex cell type-dependent impacts of the mutant RAC1 proteins possibly explain the differences observed in both MRD48 subjects and “in vitro” models with mirror contrasting results. As evidenced for the E62K RAC1 mutant, potentially different effects (i.e., disrupted GTP hydrolysis, protein structural rearrangement affecting stability and altered protein-protein interactions) may concur to the final functional behavior of each variant, eventually modulating their impact with counteracting consequences on different pathways. A similar behavior has been reported for pathogenic variants occurring in other members of this family of GTPases [3, 17,18,19]. Altered function of various RAC1 downstream interactors have been associated with NDDs with a wide range of clinical presentation, including PAK1 (MIM 602590) (IDDMSSD, MIM 618158) [45] and TRIO (MIM 601893) (MRD63, MIM 617061; MRD44, MIM 618825) [8]. The RAC1 brain network signaling has also been linked to some syndromic autism spectrum disorder-associated genes, such as AUTS2 (MIM 607270), SHANK3 (MIM 606230) and UBE3A (MIM 601623) suggesting that the neurobehavioral aspects in MRD48 might be the result of multiple and complex perturbed pathways regulating neuronal subtle cell-to-cell interaction, spine and dendritic formation and synaptogenesis [46,47,48]. Finally, the overlapping pattern of expression of RAC1 and RAC3 in both developing and adult brain, and the evidence of interaction with common downstream effectors further suggest a fine interplay among these two GTPases, which may eventually compensate each other in some functions and may contribute to the wide range of phenotypical presentation observed in both MRD48 and NEDBAF.