A Novel De Novo Splice Acceptor Variant in BICD2 Is Associated With Spinal Muscular Atrophy

Abstract

Spinal muscular atrophy lower extremity dominant 2 is an autosomal dominant neuromuscular disorder caused by disruption of the dynein activating adaptor protein Bicaudal-D2 (BICD2) [MIM: 609797] (Neveling et al. 2013). BICD2 is a ubiquitously expressed motor-adaptor Golgin protein involved in anterograde and retrograde transport of cargo from the Golgi to the endoplasmic reticulum (Martinez-Carrera and Wirth 2015). Studies support the role of BICD2 in the development and maintenance of lower motor neurons of the anterior horn of the spinal cord (Neveling et al. 2013). To date, reported pathogenic variants include missense and small in-frame deletions (Koboldt et al. 2020). Significant genotype–phenotype correlations have not been identified; however, variants in the coiled-coil domain 2 and 3 (CC2 and CC3) may be associated with more severe phenotypes (Koboldt et al. 2020). In addition, variants in the CC3 domain may be associated with greater upper extremity involvement (Koboldt et al. 2020). Functional studies showing a toxic accumulation of BICD2 and Golgi fragmentation, established mutational hot spots, and a lack of truncating variants in affected individuals and presence of such variants in healthy controls support a gain-of-function (GOF) mechanism of disease (Koboldt et al. 2020; Martinez-Carrera and Wirth 2015).

Over 30 families have been reported in the literature to date with variants in BICD2 (Koboldt et al. 2020). There are two recognized phenotypes, the severe congenital onset form (SMALED2B [MIM: 618291]) and the slowly progressive or non-progressive form (SMALED2A [MIM: 615290]) (Koboldt et al. 2020; Storbeck et al. 2017). SMALED2B is associated with de novo autosomal dominant inheritance, and clinical features of decreased fetal movements, severe hypotonia, respiratory insufficiency, arthrogryposis multiplex congenita, talipes equinovarus, fractures in utero, micrognathia, dysplastic ears, and central nervous system abnormalities including enlarged ventricles, cortical atrophy, thin corpus callosum, and cerebellar hypoplasia with or without intellectual disability. Children are non-ambulatory, and, in some cases, neonatal death occurs (Storbeck et al. 2017). For SMALED2A, both de novo and inherited BICD2 variants have been reported. There is significant phenotypic variability, with individuals presenting with congenital, childhood, or adult-onset disease of variable severity. The main characteristics of SMALED2A include lower extremity-predominant distal and proximal muscle weakness and atrophy with fatty replacement, hip and knee contractures, hip dysplasia, scoliosis, hyporeflexia or areflexia, and talipes equinovarus. Individuals present with varying degrees of motor impairment and ambulation. Some individuals have upper motor neuron signs including spasticity and hyperreflexia. Electromyography (EMG) and muscle biopsy typically reveal neurogenic abnormalities and atrophic and necrotic fibers, respectively (Frasquet et al. 2020; Neveling et al. 2013).

We report an additional affected proband with a de novo splice site variant in BICD2 (c.2107-2A>G) identified by exome sequencing (ES), in whom RNA sequencing (RNA-seq) studies were useful in clarifying the impact of the variant. The proband is an 8-year-old girl of French Canadian and First Nations ancestry who initially presented with multiplex arthrogryposis congenita of unknown etiology. Her mother had a history of congenital club feet that resolved without complications, and she exhibited no further skeletal or neuromuscular issues throughout her life. The rest of the family history was non-contributory. Prenatal ultrasound revealed talipes equinovarus. No other maternal or prenatal complications were reported. She was born via cesarean due to footling breech presentation at 37 + 5 weeks gestation. The birth was complicated by a femoral fracture. At birth, she was found to have multiple skeletal anomalies, including talipes equinovarus, bilateral contractures of the hand, slightly curved spine, hip dislocation, and hip acetabular dysplasia confirmed by ultrasound. APGARs were 8 and 8, and the birth weight was 2885 g (15th centile: CDC Girls (0–3 years-old)). She was diagnosed with multiplex arthrogryposis congenita after birth and was admitted for 1 month due to feeding difficulties requiring a nasogastric tube. She presented with hypotonia and weakness (lower extremity predominant) and areflexia. Height and head circumference at 12-days-old were within normal limits (CDC Girls (0–3 years-old) growth curves: Height: 48 cm (10th centile); head circumference: 33 cm (4th centile)). Magnetic resonance imaging (MRI) of the brain and spine showed soft tissue changes consistent with arthrogryposis with a normal brain shortly after birth. MRI of the hips showed decreased muscle of the psoas, iliacus, gluteal, and muscles of the appendicular skeleton with fatty displacement. An echocardiogram was normal.

In her first year of life, she had serial casting for her knee and ankle contractures as well as corrective bracing for scoliosis. At 23-months-old, she underwent a right acetabuloplasty with varus derotation osteomy. She has a history of recurrent shoulder dislocations with minimal trigger in early childhood. At 2-years-4-months old, x-rays revealed a 28° left proximal thoracic curve and a 65° long right thoracolumbar curve (previous measurement 46°) and recurrent ankle-foot valgum and left external tibial torsion. She underwent repeat below the knee casting and had a spine cast for progressive thoracolumbar scoliosis. Nerve conduction studies at 4-years-5-months-old showed features consistent with a severe motor neuropathy given length-dependent findings. Sensory responses of the left median, ulnar, peroneal and medial plantar nerves were normal. Motor response to the left median nerve was normal while the ulnar, peroneal and tibial nerve showed low compound motor actional potential (CMAP) amplitudes. Needle EMG, performed under sedation, revealed fibrillation potentials to the abductor hallucis and extensor hallucis longus. Creatine kinase was within normal limits. A repeat muscle MRI identified complete muscle atrophy of the lower limbs with fatty displacement (Figure 1a). At 5-years-4-months-old, she had telescopic growing rods inserted to correct her scoliosis (Figure 1b). She has persistent genu valgum and bilateral tibial external rotation. She has a history of recurrent respiratory tract infections and pneumonias, meibomitis, mild obstructive sleep apnea and borderline low bone density. She has no history of seizures, no hearing or vision loss, and no facial dysmorphisms. At 7-years-9-months-old, her height was 118 cm (10th percentile) and her weight was 41.5 kg (99th percentile).

From a developmental perspective, she has a history of fine and gross motor delay. She was unable to use her hands until 14-months-old due to contractures but this improved with physiotherapy, and she can now feed herself and draw. She crawled at approximately 16-months old, stood with a walking frame at 3-years-4-months-old, and sat unsupported at approximately 3-years-11-months-old. She is currently non-ambulatory and has generalized weakness, truncal instability, and hyperlaxity. Gross motor development has been complicated by weight-bearing difficulty. She had a normal speech-language assessment at 3-years-old and speaks in full sentences.

Initial clinical genetic testing was non-diagnostic. A microarray at 3-years-old was normal, followed by duo ES at 4 years old with her unaffected mother, which identified a canonical splice site variant in intron five of BICD2: c.2107-2A>G, p.IVS5-2A>G (NM_001003800.1) classified as a variant of uncertain significance (VUS). The variant was absent in the proband's mother, absent in gnomAD (Lek et al. 2016) and predicted by the laboratory to result in premature truncation and loss-of-function (LOF) due to exon 6 skipping. Follow-up targeted testing in the proband's father was negative suggesting the variant was de novo. Given than LOF is not an established mechanism of disease (Koboldt et al. 2020), the variant remained a VUS.

The proband was then enrolled in the Care4Rare Canada research program for RNA-seq to assess the impact of the splice site variant. RNA-seq was performed on primary fibroblast cells derived from a skin biopsy (Marshall et al. 2023). Rather than the initially predicted exon 6 skipping and subsequent frameshift and premature truncation, analysis of splice events in the RNA-seq data revealed that the c.2107-2A>G splice acceptor variant results in the activation of leaky cryptic splice acceptors in exon 6 in addition to intron 5 retention (Figure 2a). Of the 247 reads mapping at the canonical exon 5 donor, 167 (68%) splice correctly to the canonical exon 6 acceptor, whereas 38 (15%) continue into intron 5 (intron 5 retention), 38 (15%) splice to a novel cryptic acceptor 30 bp into exon 6, and 4 (2%) splice to a novel cryptic acceptor 9 bp into exon 6 (Figure 2a). Neither of the novel splicing events were detected in RNA-seq data from N = 288 fibroblast cell controls from the GTEx consortium (The GTEx Consortium 2020) or in N = 45 patient-derived fibroblasts in our in-house cohort. Intron 5 retention also was unique in the proband fibroblast cells (Figure 2a): the ratio of average coverage of intron 5 to the average coverage of exons 5 and 6 was 0.2 in the proband, whereas this same ratio was < 0.005 in all other fibroblast samples in our in-house cohort. The intron 5 sequence introduces 77 amino acids followed by a premature termination, therefore these transcripts are predicted to undergo nonsense-mediated decay (NMD). However, the use of the cryptic splice acceptors 30 or 9 bp into exon 6 results in transcripts that preserve the reading frame and effectively function as in-frame deletions of the first 10 or 3 amino acids, respectively, of exon 6.

Next, novel splicing was confirmed by RT-PCR using primers spanning exons 5–6, which showed a novel smaller product representing the 30-bp deletion in exon 6, in addition to the wild-type product and a faint larger product representing transcripts with intron 5 retention (Figure 2B,c). These data confirm that although some of the transcripts from the allele carrying the splice donor will result in premature truncation and NMD (those with intron 5 inclusion), a substantial proportion of the remaining transcripts are using an alternate exon 6 acceptor site that results in an in-frame deletion; therefore, the c.2107-2A>G splice acceptor variant is not simply a LOF variant.

To further assess the functional impact of the c.2107-2A>G splice acceptor variant, gene expression and protein levels of BICD2 were assessed. Expression of BICD2 by RNA-seq was not outlying compared to GTEx fibroblasts (The GTEx Consortium 2020), nor compared to our internal cohort (Figure 2d). This was confirmed by RT-qPCR, where no significant difference in BICD2 expression compared to 8 controls was found (Figure 2d). However, BICD2 protein levels were reduced in proband fibroblasts compared to controls (Figure 2e). This may be attributable to NMD of the subset of transcripts with intron 5 retention, as no smaller truncated protein was observed by Western blot.

RNA-seq and functional studies support the diagnosis of SMALED2B due to a de novo splice acceptor variant in BICD2 for our proband. Initial predictions from the reporting laboratory predicted haploinsufficiency, which is not an established mechanism of disease associated with this gene and is not expected based on the observation of LOF variants in BICD2 in several reportedly healthy individuals in population cohorts such as gnomAD for this autosomal dominant disorder (Karczewski et al. 2020). RNA-seq and functional studies clarified that while the abolishment of the canonical splice acceptor site at exon 6 results in some LOF transcripts that retain intron 5 and a reduction of BICD2 protein levels, the splice acceptor variant is not exclusively a LOF variant. There is also activation of leaky splice acceptors in exon 6 that preserve the reading frame, the most predominant of which is 30 bp into the exon which would result in a 10-amino acid deletion (p.Thr703_Lys712del). This is an interesting example in which the leakiness of novel splice acceptors may have had a protective effect, as only some transcripts from the variant allele use these novel in-frame splice junctions and are expected to generate an aberrant protein, whereas the remainder retain intron 5 and are likely subject to NMD. The finding of transcripts with novel splicing supports the GOF theory previously described in the literature (Koboldt et al. 2020). While most disease-associated variants in BICD2 have been missense variants, there are 3 reported likely pathogenic or pathogenic in-frame deletions in the literature and/or in the ClinVar database, all of which are single amino acid deletions (Koboldt et al. 2020). The largest in-frame deletion reported in ClinVar is a 21 bp deletion NM_001003800.1:c.613_633del(p.Phe205_Glu211del) interpreted as a VUS; therefore, the 30 bp deletion resulting from aberrant splicing in our current report represents the largest reported in-frame deletion in BICD2 associated with disease. The resulting p.Thr703_Lys712del is in the CC3 domain of BICD2, overlapping a hotspot with multiple disease-causing variants that is depleted for benign population variants (Koboldt et al. 2020). These residues are highly conserved (average GERP 3.22), and protein modeling using SWISS-Model (Waterhouse et al. 2018) predicts that the loss of these amino acids disrupts folding of the α-helix in the CC3 domain (Figure S1). This may alter some functions of this key domain, including binding cargo proteins such as RAB6 and RANBP2 (Splinter et al. 2010) and auto-inhibition of the N-terminal CC1 dynein-dynactin complex-binding domain (Splinter et al. 2012). While no significant genotype–phenotype data are available to date, other individuals with upper extremity involvement like our proband also have variants in the CC3 domain as opposed to the CC2 domain (Koboldt et al. 2020). In addition, our proband's presentation is on the severe end of the spectrum, supporting the phenotypic heterogeneity of SMALED2A and SMALED2B. This also supports the findings of Koboldt et al. (2020) in which variants in the CC2/3 domains typically lead to a more severe presentation as seen in our proband.

Taken together, the current evidence supports that this BICD2 variant c.2107-2A>G, p.IVS5-2A>G (NM_001003800.1) is likely pathogenic based on ACMG standards (Richards et al. 2015). The variant is de novo (without paternity confirmed, PM6), absent from population controls (PM2), in a suggested mutational hotspot (PM1, considered as supporting evidence), and our functional studies support that the variant impacts splicing and generates transcripts that could be in keeping with the disease mechanism (PS3, considered as moderate evidence).

This report adds to the growing literature supporting the utility of RNA-seq to clarify effects of splice-altering VUSs in rare disease diagnosis (Cummings et al. 2017; Lee et al. 2020; Wai et al. 2020). Beyond diagnoses, clarifying the impacts of splice-altering variants can also provide insight into the underlying mechanism of disease (e.g., LOF vs. GOF) which will be important to consider for directing any potential future targeted therapies. Although RT-PCR studies can be a highly sensitive tool evaluate the impact of potential splice-altering variants, mRNA isoforms detected are highly dependent on primer placement and PCR conditions. Additionally, they do not provide a quantitative measure of relative transcript abundance and it may be difficult to resolve small-scale changes. An example of this from our study is that we could not clearly identify the less abundant novel splice isoform with the small 9 bp deletion in our RT-PCR studies. RNA-seq on the other hand provides a hypothesis-free method to query and quantify all potential splice events related to a variant, provided these are adequately represented in the tissue studied. Because gene and mRNA isoform expression patterns are highly tissue and cell-type specific (The GTEx Consortium 2020), sample type is an important consideration for any RNA-seq or RT-PCR studies. Fibroblasts were used in the present study as expression of BICD2 in control cohorts is higher in fibroblasts than in other clinically accessible tissues such as blood (The GTEx Consortium 2020). This has been found for neuromuscular disease genes in general (Gonorazky et al. 2019). For similar studies, however, selection of sample to sequence may need to be determined on a gene-by-gene basis, and expression patterns weighed against accessibility of samples.

In conclusion, this case demonstrates the utility of RNA-seq to clarify the effect of this splice variant in BICD2 and ultimately support the diagnosis of SMALED2B. It also highlights the potential complex impacts of splice variants beyond simple LOF and emphasizes the need for implementation of RNA-seq as a secondary test to appropriately assess the consequences of intronic variants at or near splice sites and ultimately support rare disease diagnosis.

G.F.D.G. and S.K.M. drafted the manuscript with input from all authors. G.F.D.G., S.K.M., X.W., and G.L., contributed to data acquisition. G.F.D.G., X.W., and Y.L. analyzed data. G.L., H.J.M., and K.M.B. critically revised the manuscript, and all authors gave final approval.

The authors declare no conflicts of interest.