Genotype-phenotype correlations in RHOBTB2-associated neurodevelopmental disorders

Abstract

Purpose:

Missense variants clustering in the BTB domain region of RHOBTB2 cause a developmental and epileptic encephalopathy with early-onset seizures and severe intellectual disability.

Methods:

By international collaboration, we assembled individuals with pathogenic RHOBTB2 variants and a variable spectrum of neurodevelopmental disorders. By western blotting, we investigated the consequences of missense variants in vitro.

Results:

In accordance with previous observations, de novo heterozygous missense variants in the BTB domain region led to a severe developmental and epileptic encephalopathy in 16 individuals. Now, we also identified de novo missense variants in the GTPase domain in 6 individuals with apparently more variable neurodevelopmental phenotypes with or without epilepsy. In contrast to variants in the BTB domain region, variants in the GTPase domain do not impair proteasomal degradation of RHOBTB2 in vitro, indicating different functional consequences. Furthermore, we observed biallelic splice-site and truncating variants in 9 families with variable neurodevelopmental phenotypes, indicating that complete loss of RHOBTB2 is pathogenic as well.

Conclusion:

By identifying genotype-phenotype correlations regarding location and consequences of de novo missense variants in RHOBTB2 and by identifying biallelic truncating variants, we further delineate and expand the molecular and clinical spectrum of RHOBTB2-related phenotypes, including both autosomal dominant and recessive neurodevelopmental disorders.

Introduction

In 2018, de novo, heterozygous missense variants clustering in the Broad-complex, Tramtrack, Bric-à-brac (BTB) domain region of RHOBTB2 were found to cause a developmental and epileptic encephalopathy (DEE) (DEE64, MIM 618004).¹ This severe neurodevelopmental disorder (NDD) is characterized by early-onset seizures, severe to profound intellectual disability (ID), movement disorders, and postnatal microcephaly.¹ RHOBTB2 is an atypical Rho GTPase, containing a GTPase and 2 BTB domains. It interacts via the BTB domains with a Cullin3-dependent ubiquitin ligase complex, mediates its own ubiquitination, and recruits other substrates to the complex.² So far, only 1 substrate has been identified, the RNA-binding protein Musashi-2, encoded by MSI2. Overexpression of RHOBTB2 resulted in enhanced ubiquitination and thus decreased protein levels of Musashi-2, whereas knockdown of RHOBTB2 resulted in increased levels of Musashi-2.³

RHOBTB2 missense variants initially identified in individuals with DEE were shown to result in abundant levels of mutant RHOBTB2 in vitro, probably owing to impaired proteasomal degradation.¹ Consistent with these findings, flies with increased levels of the Drosophila ortholog RhoBTB showed seizure susceptibility in vivo.¹ Based on the observation of recurrent missense variants clustering in the BTB domain region and heterozygous, large deletions of RHOBTB2 apparently not being associated with a disease phenotype, a rather specific effect of the pathogenic missense variants was suggested.¹

Additional to the initial report of 10 cases in 2018,¹ 23 independent individuals with (likely) de novo missense variants in RHOBTB2 and a neurodevelopmental/neurological phenotype were reported.^4-13 Most of the identified pathogenic variants (28 of 33) are located in the BTB domain region, either in the first BTB domain or at the dimer interface of the second BTB domain,¹ and are associated with a relatively homogeneous DEE phenotype, including epilepsy, severe to profound ID, and further neurological abnormalities.^4-8,10-13 Two missense variants were identified in the GTPase domain (p.(Glu35Lys), p.(Arg116Cys)), one of them in an individual with DEE and the other in an individual with a dystonic movement disorder without ID and with only a single febrile seizure.⁹ Three reported missense variants were not located in any of the known domains. Two of them involved neighboring amino acid positions 239 and 241 (p.(Trp239Cys), p.(Ser241Tyr)) and were associated with a prominent dystonia phenotype with variable ID with or without epilepsy.⁴ The third variant (p.(Thr659Ala)) was identified downstream of the BTB domain region and was associated with developmental delay and infantile spasms.¹⁰

By assembling data on 23 cases with de novo missense variants in either the BTB domain region (16x) or the GTPase domain (6x) or in between (1x), we now further delineate the molecular and clinical spectrum of RHOBTB2-related autosomal-dominant NDDs. Variants in the GTPase domain are associated with a more variable phenotype compared with variants in the BTB domain region. They also behave differently in vitro, thus suggesting a genotype-phenotype correlation. Additionally, identification of 9 families with biallelic splice-site or truncating variants in RHOBTB2 and variable ID and neurological abnormalities indicate that complete loss or truncation of RHOBTB2 is causative also of an autosomal-recessive NDD.

Materials and Methods

Data collection

After the initial report in 2018,¹ we assembled mutational and clinical data of 36 additional individuals with RHOBTB2 variants by personal communication with clinicians or parents and by GeneMatcher.¹⁴ Individuals from family 2 were included in a previously published study¹⁵; the other cases have not been reported before. Variants in RHOBTB2 were identified by panel or (trio) exome sequencing in either diagnostic or research settings (Supplemental Table 1). Consent for publication of molecular and clinical data was obtained from the individuals, their parents, or legal guardians. Ethics approval for this study was obtained from the ethical review board of the University of Bern or the respective institutional review boards of the testing centers (Supplemental Table 1). The described variants are annotated based on the longest isoform of RHOBTB2 (GenBank: NM_001160036.2, NP_001153508.1, NC_000008.10). Because isoform NM_015178.3 has also been used in variant databases and some publications, we also indicate variants for this isoform in the Supplemental Tables 1 and 2.

In silico analyses and structural modeling

The model of RHOBTB2 was retrieved from the AlphaFold protein structure database (https://alphafold.ebi.ac.uk/entry/Q9BYZ6).^16,17 Residues at the sites of variants are modeled with high or very high confidence in the structure. A putative GEF binding site was mapped based on the crystal structure of the RhoA-PDZ-RhoGEF complex (PDB: 3KZ1).¹⁸ RasMol was used for structure analysis and visualization.¹⁹

Constructs

To investigate the variants p.(Ala474Gly), p.(Arg483His), and p.(Arg511Gln), we used the His-cMyc-tagged expression plasmids from the initial report.¹ Novel variants p.(Arg116Cys), p.(Arg154Gln), p.(Arg183Met), p.(Ala471Cys), and p.(Arg507Cys) from this study and p.(Trp239Cys), p.(Ser241Tyr), and p.(Tyr306Asp) from the literature were introduced using a modified version of the Quick-Change site-directed mutagenesis kit (Stratagene, Agilent) into the same plasmid as described previously.¹ Additionally, an HA-tagged Cullin3 vector¹ was used. An expression vector for Musashi-2, containing FLAG-tagged human MSI2, was obtained from Sino Biologicals (pCMV3_FLAG-MSI2:HG13069-NF).

Protein expression analyses

Protein levels of wild type and RHOBTB2 harboring the different variants and MSI2 were determined as described previously¹ and in more detail in Supplemental Methods. For statistical analysis, all values >5 were set to 5. Statistical analysis was performed using the one-sample t test with the hypothetical mean set to 1 followed by Bonferroni correction.

Co-immunoprecipitation

For co-immunoprecipitation, HEK293 cells were transiently co-transfected with 1 μg His-cMyc-tagged wild type or RHOBTB2 carrying a missense variant and 0.5 μg HA-tagged Cullin3 or FLAG-tagged MSI2 per 6-well. After 48 hours, cells were treated with 25 μM MG-132 proteasome inhibitor (Sigma-Aldrich) for 4 hours and harvested with lysis buffer (100 mM Tris-HCl pH 8, 150 mM NaCl, 1 mM EDTA, and 1 % Triton X-100) with protease inhibitor (Sigma-Aldrich). For immunoprecipitation, lysates were diluted with 1× TBS and incubated with 15 μL Protein A Mag Sepharose bead suspension (GE Healthcare) and 1.6 μg anti-Myc antibody (M4439, Sigma-Aldrich) at 4 °C overnight. Subsequently, the beads were washed once with lysis buffer and 3 times with 1× TBS, followed by elution with 1× Lämmli buffer. Protein separation and western blotting was performed as described in the Supplemental Methods. The amount of coprecipitated Cullin3 was quantified using the Image Lab software (Bio-Rad) and normalized to the amount of precipitated RHOBTB2 and compared with the wild type. Statistical analysis was performed using the one-sample t test with the hypothetical mean set to 1 followed by Bonferroni correction.

Results

Molecular and clinical spectrum of de novo heterozygous missense variants in RHOBTB2

We were able to assemble data on 23 individuals with de novo heterozygous missense variants in RHOBTB2 that can be categorized into different groups based on the location of the variants and the associated phenotypes. Detailed clinical information and genomic and complementary DNA position description of the variants are provided in Supplemental Table 1 and summarized in Table 1 and Figure 1A.

Table 1.

Comparison of main clinical features in individuals with variants in RHOBTB2 based on their location and zygosity

Variant Location	GTPase Domain New	GTPase Domain Published9	Total GTPase Domain	BTB Domain New	BTB Domain Published1,4-8,10-13	Total BTB Domain	Outside Domains New	Outside Domains Published4,10	Total Outside Domains	Recessive Variants in 9 Families15
n	6	2	8	16	28	44	1	3	4	13
ID/DD	5/6	1/2	6/8	15/15	27/27	41/41	1/1	3/3	4/4	7/9
Severe to profound	0/4	–	0/4	8/8	22/27	30/35	1/1	1/2	2/3	–
Moderate	2/4	–	2/4	0/4	5/27	5/31	0/1	0/2	–	–
Mild	2/4	–	2/4	0/4	0/27	0/31	–	1/2	1/2	2/2
Seizures/epilepsy/abnormal EEG	2/5	1/2	3/7	16/16	25/28	41/44	1/1	2/3	3/4	9/13
Febrile seizures only	0/5	1/2	1/2	0/15	2/27	2/41	0/1	–	–	2/13
Regression	3/5	–	3/5	4/214	10/20	14/34	0/1	1/2	1/3	2/10
Microcephaly	0/4, 1× macrocephaly	–	0/4	9/12	10/16	19/28	1/1		1/1	5/8
Movement disorder	1/3	2/2	3/5	12/15	24/27	36/42	1/1	2/2	3/3	8/10
MRI anomalies	2/5	1/2	3/7	8/15	15/24	23/39	1/1	1/2	2/3	2/7
Transitory neuronal deficits (postictal)	1/4	–	1/4	7/13	11/19	18/32	1/1	–	1/1	0/10

DD, developmental disorder; EEG, electroencephalography; ID, intellectual disability; MRI, magnetic resonance imaging.

Figure 1. Overview of published and novel variants in RHOBTB2.

A. Schematic drawing of RHOBTB2 with domains and identified missense variants clustering in the GTPase or BTB domain region (GenBank: NM_001160036.2) above the scheme and biallelic splice-site and truncating variants below the scheme. Domains were identified and recolored based on SMART prediction.^20,21 Variants in gray were described previously^1,4-13,15; variants in black have been identified in this cohort. Recurrent variants are underlined. p.(Arg154*) has been published before,¹⁵ and compound heterozygous variants are marked by a +. Possible genotype-phenotype correlations based on phenotype severity and variant location are indicated by blue, gray, and red boxes. # indicates variants included in experiments. The cancer variant p.(Tyr306Asn),² marked with a C, results in impaired binding to Cullin3. B. Conserved positions of the affected amino acids in the GTPase domain according to the UCSC Genome Browser.^22,23 C. Structural model of the GTPase domain. The domain is shown in ribbon presentations with α-helices in red and β-sheets in green. The variant sites (Asp114, Arg116, Arg154, and Arg183) and 1 interacting glutamate are shown in space-filled presentation (atom-type coloring) and are labeled. The potential GEF binding site is exemplarily illustrated for PDZ-RhoGEF (white-space-filled presentation). NDD, neurodevelopment disorder.

De novo, heterozygous missense variants in the BTB domain region

Sixteen individuals harbored 7 different heterozygous missense variants in the BTB domain encoding region of RHOBTB2 (Figure 1, Supplemental Table 1), located in the first BTB domain (3x) or at the interface of the second BTB domain (4x). In 14 individuals, the variant was shown to be de novo, for 3 individuals this information was not available. Six variants were recurrent and were either identified in another individual within this study and/or were described previously^1,5; the (p.(Ala471Cys) variant is novel. All missense variants in the BTB domain region were classified as pathogenic or likely pathogenic according to American College of Medical Genetics and Genomics (ACMG) guidelines (Supplemental Table 1).²⁴ Of note, several individuals additionally carried variants of unknown significance in other known disease genes (Supplemental Table 1), and a contributory effect of these variants to the phenotype cannot be excluded.

Individuals with missense variants in the BTB domain region presented with a severe DEE. In 13 individuals, onset of epilepsy was within the first 6 months of life, and in 1 individual, seizures started at the age of 6 years. Seizure types included generalized tonic-clonic or focal seizures, with status epilepticus reported in 2 individuals. Eight individuals were treated with levetiracetam. Seizures were reported to be treatment-responsive in most individuals, but they were refractory in 2.

Developmental delay and ID were noted in 15 individuals, for 1 individual, this information is missing. Developmental regression occurred in 4 individuals, correlating with onset of seizures. All but 1 affected individuals presented with severe language delay and lacking or severely impaired speech capacities. Motor development was also severely impaired with limited or lack of ambulation in most individuals. Microcephaly or a rather small head circumference was observed in all individuals for whom data were available.

Twelve individuals had movement disorders, including ataxia, dyskinesia, and choreoathetosis. Response to treatment with acetazolamide was reported for 1 individual in this study. In 7 individuals, (post-ictal) hemiparesis occurred, in 1 subject after head injury. Brain magnetic resonance imaging (MRI) anomalies included hypoplastic corpus callosum, atrophic changes in temporal lobes, or delayed occipital myelination.

Other common features included muscular hypotonia or hypertonia (12 of 14 individuals) and behavioral abnormalities, autism spectrum disorder, or stereotypic movements (11 of 14 individuals). Minor facial dysmorphisms were noted in the majority of individuals but were rather nonspecific.

De novo, heterozygous missense variants in the GTPase domain

Additionally, we assembled data on variants residing outside the BTB domain region. Five different de novo missense variants in 6 individuals were identified within the GTPase domain (Figure 1, Supplemental Table 1). One of them recurrently occurred in 2 individuals (p.(Arg116Cys)) in this study and was published previously in another individual.⁹ Another variant was located in close proximity (p.(Asp114His)). Two missense variants affected the same amino acid residue (p.(Arg154Gln) and p.(Arg154Leu)). The missense variant p.(Arg183Met) is located close to a splice site and is predicted to lead to loss of the splice donor (score 0.95).²⁵ Aberrant splicing by loss of exon 6 was confirmed in an in vitro splice assay (Supplemental Methods, Supplemental Figure 1A). However, as patient-derived material (blood or cells) was not available, the in vivo effects of this splicing variant situation remain unclear.

None of the missense variants in the GTPase domain was present in gnomAD (gnomAD v2.1.1).²⁶ They all affected highly conserved amino acid residues (Figure 1B) and were predicted to be deleterious by several in silico prediction tools (Supplemental Table 2). The recurrent p.(Arg116Cys) variant, the variants affecting the same amino acid position (p.(Arg154Gln), p.(Arg154Leu)) and the variant at position 114 in the GTPase domain were classified as likely pathogenic according to ACMG guidelines²⁴ (Supplemental Methods, Supplemental Table 1). The variant p.(Arg183Met) remained of unknown significance because of the possible effect on splicing that might result in nonsense-mediated messenger RNA (mRNA) decay of the variant carrying allele.

All 6 individuals with variants in the GTPase domain presented with a variable neurodevelopmental phenotype (Table 1, Supplemental Table 1). Seizures only occurred in 2 individuals with good response to antiseizure treatment. When data were available, cognitive impairment ranged from learning difficulties (1x) to mild (1x) and moderate (2x) ID. Although speech delay occurred in most, 2 individuals could speak in complete but rather short sentences. Developmental regression was reported in 2 cases, cooccurring with spasms in 1 of them. Dystonic movements or hemiparesis were reported in a single individual each, and brain MRI anomalies in 2 individuals. Behavioral abnormalities and minor, nonspecific facial dysmorphism were common. Head circumferences were normal in 3 individuals, and macrocephaly was noted in 1 individual.

De novo missense variants not located in any of the known domain regions

Another de novo missense variant (p.(Pro262Leu)) was located between the GTPase and the BTB domains. Pathogenicity remains unclear. The individual’s phenotype included neonatal onset epilepsy with 30 to 40 seizures per day and myoclonic jerks, which were not fully controlled. Profound ID and neurological impairment were present, with absent ambulation and speech at age of 4 years. Other features included microcephaly, severe spastic paraparesis with little voluntary movement, and brain MRI anomalies such as diffuse cerebral atrophy. This individual additionally harbored a hemizygous missense variant in UBE2A, a gene associated with severe neurodevelopmental phenotypes (ID disorder, X-linked, syndromic, Nascimento type, MIM 300860). However, the variant is located in a noncanonical transcript, and its clinical significance is uncertain.

Molecular and clinical spectrum of biallelic splice-site and truncating variants in RHOBTB2

We also collected data of 13 individuals from 9 independent families with homozygous (8 families) or compound heterozygous (1 family), potential loss-of-function variants in RHOBTB2 (Figure 1A, Supplemental Figure 2, Supplemental Table 1). Ten variants were nonsense, frame-shifting, or located in or close to splice sites. Aberrant splicing was confirmed by an in vitro splice assay (Supplemental Methods) for 3 splice-site variants (c.258+4A>C, c.1568–1G>A, c.2032+1G>C) (Supplemental Figure 1B and C). However, as patient-derived material (blood or cells) was not available, the in vivo effects of these splicing variants remain unclear.

Of note, 4 of the truncating variants are located in the pen-ultimate or ultimate exon, therefore possibly escaping nonsense mediated mRNA decay,²⁷ as predicted for the other, more N-terminal truncating variants. For these C-terminal variants, truncation and thus a gain-of-function mechanism cannot be excluded. Five of the recessive variants are listed in gnomAD in a heterozygous state with a very low frequency (p.(Arg179*) and p.(Arg670*) reported in 2 alleles, p.(Tyr700*) and p.(Trp105*) reported in 1 allele each, and c.258+4A>C reported in 8 alleles).²⁶ According to ACMG criteria, 7 of these variants were classified as likely pathogenic and 3 as variants of unknown significance (Supplemental Table 1).

All individuals with biallelic variants in RHOBTB2 presented with variable neurodevelopmental phenotypes. Cognitive impairment ranged from learning difficulties to moderate ID. Speech delay was common. Seizures or febrile seizures occurred in all but 2 individuals. Response to treatment ranged from poor/partial (3 of 6) to good (3 of 6). Most of the individuals started to walk within the first 2 years of life, often with unsteady gait or a movement disorder. Microcephaly was reported in 5 of 9 individuals for whom information was available.

Predicted structural consequences of heterozygous missense variants in RHOBTB2

In 2018, structural modeling of variants in the BTB domain region predicted impairment of intramolecular stability and formation, destabilization of the first BTB domain, and interference with dimer formation of the second BTB domain, the latter associated with variants at position 510 and 511, located at the dimer interface of the second BTB domain.¹

We now performed structural analysis of the GTPase domain, based on a model generated with AlphaFold (Figure 1C). Most affected residues (Asp114, Arg116, and Arg154) are located on the surface of the GTPase domain, making them candidates for protein-protein interactions with interaction partners and/or substrates. Because homologous, typical GTPases such as RhoA, interact with guanine nucleotide exchange factors (GEFs) to catalyze release of GDP, we exemplarily mapped the PDZ-RhoGEF binding site on the RHOBTB2 GTPase domain, revealing that Asp114 and Arg116 would be located in the RHOBTB2-GEF interface and that Arg154 is close to the interface (Figure 1C). We refrained from a more detailed analysis of the interactions because the exact physiological interaction partner recognizing this surface patch is not yet known and as a GTPase activity for RHOBTB2 has not been shown.^2,28-30 Regardless of the protein recognized by this surface region, we expect that variants have a significant impact because they all lead to a loss of positive charges ((p.Arg116Cys), p.(Arg154Gln), p.(Arg154Leu)), or negative charges (p.Asp114His), thereby disrupting the electrostatic complementarity of protein-protein interactions. In the model, Arg183 forms a salt-bridge with Glu189. Because this interaction cannot be formed by the uncharged Met183, the p.(Arg183Met) exchange rather destabilizes the GTPase domain itself.

Interaction with Cullin3 or Mushashi-2 is not impaired by RHOBTB2 missense variants

To further analyze the functional consequence of missense variants on the RHOBTB2 protein level, we selected variants from the different domains and from regions between the known domains (GTPase: p.(Arg116Cys), p.(Arg154Gln), p.(Arg183Met), between: p.(Trp239Cys), p.(Ser241Tyr), BTB cancer: p.(Tyr306Asp), BTB DEE: p.(Ala471Cys), p.(Ala474Gly), p.(Arg483His), p.(Arg507Cys), and p.(Arg511Gln)) identified in this and previous studies.^1,2,4,5

Because RHOBTB2 interacts with the scaffold protein Cullin3 to assemble into an ubiquitin ligase complex, we first tested if the variants resulted in impaired binding toward Cullin3. We confirmed reduced binding for the cancer variant p.(Tyr306Asp) as demonstrated before.² Consistent with previous findings,¹ we did not observe impaired binding to Cullin3 resulting from variants in the BTB domain region (p.(Ala471Cys), p.(Ala474Gly), p.(Arg483His), p.(Arg507Cys), and p.(Arg511Gln)). We now also tested binding to Cullin3 for variants located within the GTPase domain or between the domains (p.(Arg116Cys), p.(Arg154Gln), p.(Arg183Met), and p.(Trp239Cys)), and did not observe an alteration either (Figure 2A). Moreover, we also did not observe altered binding to the only known substrate of RHOBTB2, Musashi-2 for any of the constructs with variants (Supplemental Figure 3).

Figure 2. Consequences of missense variants regarding binding to Cullin3 and proteasomal degradation of RHOBTB2.

A. Co-immunoprecipitation of His-cMyc-tagged RHOBTB2 and HA-tagged Cullin3 shows equal coprecipitation of Cullin3 for both wild type and RHOBTB2 carrying the missense variants, except for the cancer variant p.(Tyr306Asn). Cells were treated with the proteasomal inhibitor MG132, and co-immunoprecipitation wa8s performed with an antibody against Myc. A representative image from 5 independent experiments is shown. For quantification, Cullin3 bands after co-immunoprecipitation were normalized to the corresponding RHOBTB2 bands and compared with RHOBTB2 wild type. Error bars represent the standard error. Statistical analysis was performed using the one-sample t test with the hypothetical mean set to 1 followed by Bonferroni correction (*P < .05, **P < .01, and ***P < .001). Uncropped blots are provided in Supplemental Figure 4. B. Representative western blot from 3 independent experiments after transfection of wild type and His-cMyc-tagged RHOBTB2 carrying a variant shows impaired proteasomal degradation for both cancer and BTB domain region variants but not for the others. Experiments were performed with (top) and without (bottom) proteasomal inhibitor MG132. For quantification, RHOBTB2 bands were normalized to the loading control GAPDH and compared with RHOBTB2 wild type. Error bars represent the standard error. For statistical analysis, all values >5 were set to 5. Statistical analysis was performed using the one-sample t test with the hypothetical mean set to 1 followed by Bonferroni correction (*P < .05, **P < .01, and ***P < .001). Uncropped blots are provided in Supplemental Figure 5. GAPDH, glyceraldehyde 3-phosphate dehydrogenase; WT, wild-type.

Proteasomal degradation is only impaired by missense variants in the BTB domain region

We next determined protein levels of RHOBTB2 without or with missense variants and observed increased levels for RHOBTB2 carrying any of BTB domain region variants, as described previously.¹ The differences in protein quantity decreased upon addition of proteasomal inhibitor, supporting the hypothesis of impaired proteasomal degradation as cause of protein abundancy. Protein levels of RHOBTB2 carrying a variant outside the BTB domain, however, were unaltered compared with wild type. Our findings therefore indicate different consequences on proteasomal degradation of RHOBTB2 depending on localization of missense variants (Figure 2B).

Discussion

By assembling data on individuals with either de novo missense or inherited, biallelic splice-site and truncating variants in RHOBTB2, we further delineate and expand the RHOBTB2-associated NDD spectrum. We are able to define several RHOBTB2-related genotype-phenotype correlations, based on (A) clinical manifestations; (B) location, nature, and functional consequences of variants; and (C) inheritance pattern.

Consistent with initial reports,^1,5 individuals with de novo missense variants clustering in the BTB domain encoding region of RHOBTB2 present with a rather homogeneous, severe DEE, including early-onset epilepsy, ID, microcephaly, paroxysmal movement disorders, and MRI anomalies. Furthermore, transient neurological deficits, such as hemiparesis, stroke-like episodes, and brain MRI anomalies, were reported in several individuals, often occurring postictally.^1,4,5,11,12 Head trauma also seems to trigger encephalopathic episodes in individuals with RHOBTB2-associated DEE.^31,32 In our study, hemiparesis was observed in a postictal setting in 6 individuals and once after head injury (individual 22).

Apart from the “typical” DEE phenotype associated with de novo missense variants in the BTB domain region, we found growing evidence for a broader spectrum of neurodevelopmental phenotypes associated with de novo missense variants located in other domains or regions of RHOBTB2. In addition to 2 previously reported missense variants,⁹ we now identified 6 further individuals with de novo missense variants in the GTPase domain. The phenotype associated with variants in the GTPase domain is more variable and/or milder compared with that of individuals harboring variants in the BTB domains. Affected individuals presented with mild to moderate ID, and seizures or movement disorders occurred less frequently. Microcephaly, which is commonly seen in individuals with BTB domain region variants, was not observed in any of the individuals with variants in the GTPase domain. Interestingly, even macrocephaly was reported in 1 subject in the latter group. Thus, our data show that the GTPase domain manifests as a second variant hotspot for a RHOBTB2-associated NDD. However, the small number of missense variants outside the BTB domain region still limits definitive conclusions.

Furthermore, we identified several families with biallelic truncating variants in RHOBTB2. Affected individuals presented with variable neurodevelopmental phenotypes, including ID and seizures. Therefore, RHOBTB2 is not only implicated in autosomal-dominant NDDs due to de novo missense variants but also in an autosomal-recessive NDD caused by biallelic truncating/loss-of-function variants. RHOBTB2 therefore adds to the growing list of genes associated with both autosomal-dominant and autosomal-recessive NDDs, such as PLXNA³³ (Dworschak-Punetha neurodevelopmental syndrome, MIM 619955) or ACTL6B (DEE76, MIM 618468, Intellectual developmental disorder with severe speech and ambulation defects, MIM 618470).³⁴ The phenotype of individuals with biallelic truncating variants in RHOBTB2 overlaps with those carrying de novo missense variants regarding ID and epilepsy. However, it is more variable and less specific than the DEE resulting from de novo missense variants in the BTB domain region.

Although a gain-of-function effect due to truncation and lack of nonsense-mediated mRNA decay for biallelic variants in the last or penultimate exon might be possible, a general loss-of-function effect is most likely at least for biallelic splice-site and truncating variants in the more N-terminal regions of RHOBTB2. In contrast, such a loss-of-function effect for the de novo missense variants is rather unlikely as heterozygotes of the familial truncating variants do not show a phenotype, and because RHOBTB2 is predicted to be tolerant toward loss-of-function variants (probability of loss-of-function intolerant = 0.01)²⁶ and because heterozygous deletions of RHOBTB2 were observed in unaffected individuals.^1,35 In accordance, not reduced but instead abundant RHOBTB2 protein levels were shown to result from pathogenic missense variants in the BTB domain region previously.¹ This might be due to impaired proteasomal degradation of RHOBTB2. Missense variants in the BTB domain region might, therefore, have a very specific, initial loss-of-function effect on the ubiquitin-proteasome pathway but result in a gain-of-function effect by increased RHOBTB2 levels.

Because we now observed phenotypic differences in individuals carrying de novo missense variants also in other domains or regions of RHOBTB2, we investigated if these missense variants might have different functional consequences from those in the BTB domain region. Although the BTB domains are known to interact with Cullin3 and with other RHOBTB2 molecules to form homo- and heterodimers with RHOBTB2 or RHOBTB3 molecules,³⁶ the function of the GTPase domain is still elusive. It is distinguished from typical Rho GTPases by various sequence alterations and lack of crucial amino acids for GTP binding and catalytic transformation.^30,37 Therefore, it is controversially discussed if this domain has GTP binding capacities and/or GTPase activity.^2,28-30 Similar to variants residing in the BTB domain region, we did not observe impaired binding to ubiquitin ligase complex scaffold protein Cullin3 for variants located in the GTPase domain. This would be in line with none of the DEE- or NDD-related variants being located in the specific Cullin3-binding region in the first BTB domain.^1,2 Although variants in the BTB domain region result in impaired proteasomal degradation of RHOBTB2 in vitro as shown before¹ and confirmed in this study, we did not observe such an effect when testing variants from the GTPase domain or outside the domains. This suggests a different functional consequence of missense variants in the GTPase domain compared with missense variants in the BTB domain region.

Missense variants in the BTB domain region were predicted to impair the intramolecular stability of RHOBTB2 and dimer formation of the second BTB domain.¹ Structural modeling for the GTPase domain variants now suggests that p.(Arg183Met) decreases stability of the GTPase domain itself, whereas the remaining variants at positions 114, 116, and 154 more likely affect protein-protein interactions (Figure 1C). Alternatively, these variants may also affect intramolecular domain interactions, which have been proposed to regulate the active and inactive state of RHOBTB2.³⁶ Here, the GTPase domain interacts with the first BTB domain, keeping the protein in an inactive state. Conformational changes could be induced by interaction with specific ligands or substrates, allowing RHOBTB2 to assemble into a Cullin3 ubiquitin ligase complex.^29,30,36

Because the physiological binding partner of the interface is yet unknown, we investigated the impact of the missense variants on the only known substrate of RHOBTB2 to date, Musashi-2, in vitro. Musashi-2 has been shown to interact with the C-terminal region of RHOBTB2.³ In accordance, our results did not indicate impaired interaction for any of the tested missense variants. However, it is still possible that variants in the GTPase domain lead to a destabilization of this protein-protein interaction or affect an alternative type of protein-protein interaction (eg, intramolecular interactions or interactions with GEF proteins).

Although the exact mechanism remains elusive, our observations and investigations indicate a different functional consequence resulting from variants in the GTPase domain compared with variants in the BTB domain region and may possibly contribute to the different severity and manifestation of associated phenotypes. Although the functional consequence of clustering missense variants in either the BTB domain region or the GTPase domain might be rather specific, no clear categorization into loss-of-function, gain-of-function, or dominant-negative effects is possible, so far.

Conclusion

By identifying a phenotype-genotype correlation regarding location and consequences of de novo missense variants in RHOBTB2 and the resulting neurodevelopmental phenotype and by newly identifying biallelic truncating variants, we further delineate and expand the molecular and clinical spectrum of RHOBTB2-related NDDs.

Supplementary Material

The online version of this article (https://doi.org/10.1016/j.gim.2023.100885) contains supplemental material, which is available to authorized users.

Data Availability

Novel variants in this paper have been submitted to LOVD (variant # 0000922394 – 0000922410).