Introduction

For many common neurological disorders, a genetic contribution or even a monogenic aetiology has long been established [1]. For less frequent disease entities, similar progress is beginning to emerge thanks to the recent advances in genomic technologies [2]. Identifying causative genes for ultra-rare neurological conditions, however, remains a great challenge. Many pertinent reports present only a single-small family or even only a single patient [3, 4], and a lack of subsequent confirmatory studies may cast some doubt on the validity of the proposed genotype–phenotype correlations.

As an example, a recent study suggested causality of biallelic GTPBP2 variants for a complex neurological phenotype [5]. The gene, which had not previously been involved in human inherited disease, was found to harbour a homozygous splice site variant in three affected siblings born to consanguineous parents from Iran. Clinical features observed in all patients included dystonia, ataxia, cognitive dysfunction, motor neuropathy, and retinal abnormalities, but also sparse, thin and brittle hair. Mild hypointensities in the Globus pallidus upon magnetic resonance imaging (MRI) were interpreted in favour of abnormal brain iron deposition. Despite partial overlap of some of the clinical symptoms to the phenotype of a Gtpbp2 knockout mouse (which, notably, carried a concomitant variant which affects function in a tRNA gene) [6], the authors recognised that confirming the causative nature of inactivating GTPBP2 variants requires identification of additional families [5].

We report three families in which distinct homozygous GTPBP2 non-sense variants are associated with a wide range of neurological and non-neurological manifestations. Our findings not only confirm that biallelic inactivation of GTPBP2 is pathogenic, but also broaden the phenotypic spectrum of rare GTPBP2-associated disorder.

Patients and methods

Patients

Parental consanguinity and/or positive family history suggested a genetic basis for the complex phenotypes observed in families 1, 2, and 3 (Fig. 1a, c); molecular genetic workup was therefore requested. In the single patient of family 1, the primary clinical findings were failure to thrive, psychomotor retardation, visual impairment, and midline hair defect, whereas the index patient of family 2 had initially been clinically diagnosed with intellectual disability and ectodermal dysplasia. The patient from family 3 presented with developmental delay, seizures, microcephaly, dysmorphic features, abnormal tooth colour, and sparse hair. Magnet resonance imaging in the index patient of family 1, as well as more detailed clinical re-evaluation of both index patients were initiated after identification of GTPBP2 variants. All genetic analyses were performed in concordance to the provisions of the German Gene Diagnostic Act (Gendiagnostikgesetz), and written informed consent was obtained from the patients’ parents and referring clinicians for publication.

Fig. 1
figure 1

Three families carrying homozygous GTPBP2 (NM_019096.4) non-sense variants. Stars denote individuals analysed by WES. Allelic status for the variants in question is provided below individual symbols in the pedigrees. Sequence traces confirming the primary WES data are shown to the right; the predicted consequences at protein are indicated below the traces. a Family 1 and variant c.1219C>T. b Family 2 and variant c.1408C>T. Several third and fourth-degree relatives of the index case are affected by a reportedly similar disease (grey symbols). Dotted relationship lines indicate that consanguinity is likely. c Family 3 and variant c.430C>T. d Magnet resonance imaging of the index patient III-3 from family 1 at age 6 years. On the left: sagittal T1-weighted spin-echo image discloses a thin corpus callosum (arrowhead) with absent rostrum (arrow), hypoplasia of cerebellar vermis (stippled arrow) and presence of mega cisterna magna (star). On the right: axial T2*-weighted fast field-echo image (section through upper tectum) does not reveal evidence for specific hypointensities in the globus pallidus bilaterally (arrows) or in the substantia nigra (arrow heads). e Index patient from family 3 at age 2 years. Note dysmorphic features and sparse hair

Full size image

DNA preparation and sequencing

Genomic DNA from several members of family 1 (Fig. 1a) was prepared from leucocytes using a standard salting-out procedure. For the index patients of families 2 and 3 (Fig. 1b, c), genomic DNA was isolated from dried blood spots in filter cards (CentoCard) using the QIAamp DNA Blood Mini QIAcube Kit (Qiagen, Valencia, CA) according to the manufacturer’s instructions. Whole-exome sequencing (WES) for all three index patients, as well as for the parents in family 1 was performed as described previously [7]. In short, the Nextera Rapid Capture Exome Kit (Illumina, San Diego, CA) or the SureSelect Human All Exon kit (Agilent, Santa Clara, CA) were used for enrichment, and a HiSeq4000 (Illumina) instrument for the actual sequencing; variants were annotated using Annovar (http://annovar.openbioinformatics.org) [8] and in-house ad hoc bioinformatics tools [7].

Considering parental consanguinity in all cases, we first filtered WES data for homozygous variants with minor allele frequencies <1% in each of three public variation databases (1000 Genomes Project; Exome Variant Server; Exome Aggregation Consortium database), as well as in CentoMD and an internal database. CentoMD is Centogene’s proprietary database, and includes data from nearly 140,000 individuals of diverse geographic origins [9]. Our internal database contains ~9000 and 750 index cases (plus relatives) which have been analysed by WES and whole genome sequencing, respectively. This resulted in candidate gene lists that contained between 21 and 40 genes per family (Supplementary Table 1). Truncating variants (non-sense, frameshift, core 2 bp splice motifs, affecting initiation codon) were prioritised, and reviewed by searching the Online Mendelian Inheritance of Man database [http://www.omim.org] and the clinical-genetic literature. GTPBP2 variants were confirmed by Sanger sequencing.

Results

Genetic findings

Families 1 and 2 were both analysed in 2016, whereas family 3 was analysed in 2017. The WES evaluation strategy as outlined above flagged the homozygous GTPBP2 non-sense variants c.1219[C>T];[C>T] p.(Gln407*), c.1408[C>T];[C>T] p.(Arg470*) and c.430[C>T];[C>T] p.(Arg144*) (NM_019096.4), respectively, and this was based on a single 2016 publication which suggested a homozygous GTPBP2 splice site variant to be causative for a complex neurological phenotype [5]. Sanger sequencing of the corresponding GTPBP2 exons confirmed the primary WES-based findings, and revealed complete segregation of c.1219[C>T];[C>T] with the phenotype (Fig. 1a). Segregation in families 2 and 3 could not formally been proven as only the index patients’ DNA samples were available. All three GTPBP2 non-sense variants were submitted to the locus-specific database at the Leiden Open Variation Database (https://databases.lovd.nl/shared/genes/GTPBP2); the corresponding individual identifiers are #00121857, #00121858, and #00121859.

Clinical findings

Family 1 originates from Saudi Arabia. The index is a 6-year-old female, the parents are first degree cousins, and family history is negative (Fig. 1a). The patient presented with a neurodevelopmental phenotype characterised by dysmorphic facial features, global developmental delay, postnatal microcephaly, failure to thrive, and optic atrophy. Brain MRI revealed hypogenesis of the corpus callosum, cerebellar hypoplasia, and Dandy-Walker malformation, but hypointensities suggestive of brain iron accumulation were not observed (Fig. 1c). A summary of the clinical and imaging findings is provided in Table 1. Parents and unaffected siblings were heterozygotes or non-carrier of the c.1219C>T p.(Gln407*) GTPBP2 variant (Fig. 1a).

Table 1 Genetic and clinical features of all presently and previously reported patients with GTPBP2-associated disorder
Full size table

The second family is from Kuwait. The index is currently 10 years of age. His parents acknowledged to be distantly related, and several third- and fourth-degree relatives are similarly affected (Fig. 1b). The initially reported symptoms, i.e., intellectual disability and ectodermal dysplasia, were accompanied by seizures and a range of muscular/skeletal abnormalities. The medical record also refers to an earlier MRI investigation, and mentions agenesis of the corpus callosum, but not hypointensities in the Globus pallidus or adjacent areas (Table 1).

The third family originates from Saudi Arabia. The index is a 2-year-old boy born to first degree cousins with negative family history (Fig. 1c). He presented from the 3rd month of life with global developmental delay. In addition, he had microcephaly, seizures, hypertonia, and choreoathetosis. He also has abnormal dentition and tooth colour, as well as sparse hair (Table 1, Fig. 1d). Brain MRI showed cerebellar hypoplasia and a thin corpus callosum (Supplementary Figure 1). The initial clinical suspicion had been Menkes disease, but this was ruled out by genetic testing of the ATP7A gene (Sanger sequencing and copy-number screening).

Discussion

Our study reports three novel families in which biallelic variants in GTPBP2 are associated with a rare Mendelian disease. It thereby confirms the previously suggested pathogenic nature of complete GTPBP2 inactivation [5]. Many of the characteristics observed in the corresponding patients have a definite or at least potential neurological nature, with global developmental delay and intellectual disability being prime features. Additional shared symptoms include visual impairment (of different causes) and skeletal anomalies (scoliosis, pectus deformity). Remarkably, hair, teeth and/or skin abnormalities are also observed in all three families, but whether these observations are related to a more generalised ectodermal involvement needs further clarification.

Despite substantial clinical overlap between all reported families, clear differences are evident as well (Table 1). While the clinical phenotype of our patients is reminiscent of a neurodevelopmental disorder with structural brain abnormalities and severe intellectual disability, patients from the initially reported family presented with moderate intellectual disability, movement disorders and cerebellar atrophy. These differences may be explained by a generally wide phenotypic spectrum associated with GTPBP2 inactivation and/or by family-specific homozygous variants in additional genes as predicted for highly inbred populations [10]. Of note, the apparent discrepancies comprise MRI-based evidence for neurodegeneration with brain iron accumulation (NBIA). The positive finding by Jaberi et al. [5]. prompted the authors to suggest that GTPBP2 may be a novel NBIA gene, while we did not observe suggestive MRI hypointensities in our patients. This negative finding could, however, be a consequence of earlier diagnosis and younger age of our patients. Additionally, differences in the magnetic resonance imaging technique applied (susceptibility-weighted by Jaberi et al. [5]. vs. T2*-weighted gradient-echo images in our case) may play a role. Still, the apparent lack of clinical progression and the extra-neurological presentations would be rather atypical for NBIA [11, 12]. By additionally considering the very early onset, one may instead argue that the primary defect is of neurodevelopmental rather than neurodegenerative nature.

Such neurodevelopmental pathogenicity is also implied by findings in the earlier published mouse model. Before the first clinical-genetic report on GTPBP2 [5], a random mutagenesis-derived mouse line had been found to carry a homozygous Gtpbp2 splice site variant [6]. The associated pathology involved several types of neurons, but was CNS-specific; the resulting motor deficits were rapidly progressive, with death occurring at ~2 months of age. Notably, development of this phenotype depended on concomitant presence of a homozygous variant in a brain-specific tRNA. We therefore specifically checked whether our primary candidate gene lists (Supplementary Table 1) contain tRNA genes, but this was not the case. Moreover, a digenic background [13] can firmly be excluded in human patients, as it would predict many healthy carriers of biallelic GTPBP2 inactivating variants. GTPBP2 data in the ExAC database (http://exac.broadinstitute.org), however, shows a high probability for loss-of-function intolerance (Table 2 and corresponding legend), supporting our hypothesis. The severely affected murine model of Ishimura et al. [6] in which CNS-specificity is apparently mediated by brain-specific expression of a certain tRNA, may therefore not directly be comparable with human patients. The proposed pathomechanism behind inactivation of GTPBP2/Gtpbp2, i.e., a detrimental effect on the translational machinery [14], is still likely to be shared. GTPBP2-associated disease may thereby be functionally related to certain peripheral neuropathies caused by variants in aminoacyl-tRNA synthatase genes [15].

Table 2 Known variants in GTPBP2 which are predicted to trigger non-sense-mediated decay of the major isoform NM_019096.4, and to therefore represent bona fide loss-of-function variants
Full size table

We propose that, with the data at hand, classification of the GTPBP2-associated disease as an NBIA-subtype may be a premature conclusion. Further and older patients will have to be characterised to solve the issue, and to define the core features of the condition. That such patients can be identified in greater numbers is suggested by the fact that three of our ~9000 clinically mixed exomes [7] turned out to be positive for homozygous truncating variants. GTPBP2-associated disease may, after all, be rare rather than ultra-rare. GTPBP2 should therefore be added to the growing list of genes that have to be considered during molecular diagnosis of hereditary neurodevelopmental disorders.