Aberrant splicing induced by the most common EPG5 mutation in an individual with Vici syndrome

Sir,

Recently, a study describing the clinical and genetic data of 50 patients (30 families) with Vici syndrome (OMIM 242840) was published in Brain (Byrne et al., 2016b). Vici syndrome is a severe autosomal recessive condition recently found to be caused by mutations in the EPG5 gene (NG_042838.1) (Cullup et al., 2013). Principal features of Vici syndrome include neurological sequelae (agenesis of the corpus callosum, profound developmental delay, and progressive microcephaly), neonatal cataracts, hypopigmentation involving both skin and hair, progressive dilated or hypertrophic cardiomyopathy, and both B and T cell-related immunodeficiency (Byrne et al., 2016a, b). The most common mutation identified in Vici syndrome patients is EPG5 NM_020964.2; c.1007A>G p.Gln336Arg. Based on the two large cohorts of patients recently studied (Cullup et al., 2013; Byrne et al., 2016b), this mutation accounts for >10% (9/86) of mutations. Ancestry data from previous studies show that this mutation may be associated with Ashkenazi descent (Byrne et al., 2016a, b). This variant was not detected in whole genome sequencing data from our cohort of 4456 unrelated adults seen at the Inova Translational Medicine Institute (Bodian et al., 2014, 2016). This variant is seen with a frequency of 3.129 ×10⁻⁵ in the European (non-Finnish) population of Exome Aggregation Consortium (ExAC) and has thus far never been reported in other ExAC populations (ExAC, 2016; http://exac.broadinstitute.org).

The EGP5 c.1007A>G mutation has been suggested to affect splicing through in silico prediction models, though multiple predictions are contradictory [Human Splicing Finder (Desmet et al., 2009; http://www.umd.be/HSF3/HSF.html: −6%) and NNSPLICE (Reese et al., 1997; http://www.fruitfly.org/seq_tools/splice.html: −100%)]. Likewise, in silico predictions of the missense mutation are in disagreement with regards to its pathogenicity [PolyPhen-2 (Adzhubei et al., 2010; http://genetics.bwh.harvard.edu/pph2/: probably damaging, 0.998/1.00); SIFT (Kumar et al., 2009; http://sift.jcvi.org/: tolerated, 0.36 > 0.05); and MutationTaster (Schwarz et al., 2014; http://www.mutationtaster.org/: disease causing)]. Unlike other mutations causing Vici syndrome, full functional characterization of this common mutation is lacking (Cullup et al., 2013; Byrne et al., 2016b). Therefore, it is imperative that definitive proof of whether this mutation induces aberrant splicing or not is assessed. Furthermore, understanding if the patient phenotype is due to a null allele (splice mutation) or a missense mutation (p.Gln336Arg) is important in order to comprehend the molecular mechanism of Vici syndrome.

To investigate the functional effect of this mutation, including whether splicing is affected, we studied samples from a patient with prototypical features of Vici syndrome including static encephalopathy, absent corpus callosum, and neonatal cataracts. This individual had been found via commercial exome sequencing with Sanger validation (GeneDx) to be homozygous for the EPG5 c.1007A>G mutation (Patient 29 in Byrne et al., 2016b). Samples were obtained with full informed consent through a National Human Genome Research Institute IRB-approved protocol (76-HG-0238). Cultured primary dermal fibroblasts from this individual were harvested and after Sanger validation of the variant, RNA was extracted; fibroblasts from an unaffected individual (PCS-201-010, ATCC) were cultured and processed simultaneously as a control. Extracted mRNA was reverse transcribed using the SuperScript® VILO™ cDNA synthesis kit (Invitrogen).

We first analyzed the EPG5 mRNA level in the affected patient’s cells. For this, we performed quantitative polymerase chain reaction (PCR) (TaqMan® assay Hs00826959_m1, Life Technologies), which showed a decrease of ∼50% in the affected fibroblasts compared to control (normalized to POLR2A expression, Hs00172187_m1), suggesting that the c.1007A>G, p.Gln336Arg variant induces a decrease of EPG5 mRNA (data not shown). This reduction of mRNA is most likely due to nonsense-mediated decay (NMD) of misspliced isoforms (Hentze and Kulozik, 1999). The next question we addressed was whether the observed, stable EPG5 mRNA is the product of the correctly spliced isoform or not.

To analyze the residual mRNA, we PCR amplified patient and control cDNA using primers in exon 2 (5’-atgctaaaagtcggctgtgg-3’) and exon 5 (5’-gccaactctttgtagcacgg-3’) with AmpliTaq Gold®360 (Applied Biosystems). In the control sample, one prominent PCR product was observed with an apparent length of 450 nucleotides. In the patient, however, bands at ∼470, 540 and >1000 nucleotides were observed by electrophoresis (data not shown). These different PCR products observed in the patient sample represent several different splice isoforms that are generated in the context of the missense mutation but not degraded by NMD.

To better characterize these different isoforms observed in the patient cDNA, we sequenced the amplified cDNA. For this, the PCR products were cloned into a pCR4-TOPO bacterial vector (Life Technologies) and over 100 individual clones were isolated and sequenced with both M13 forward and reverse primers. Surprisingly, 25% of clones sequenced from the patient sample showed normal mRNA splicing (isoform 1, Fig. 1). The other products observed in the patient sample represented several different splice isoforms (isoforms 2–11, Fig. 1; variant splice sites are indicated relative to the normal splice isoform cDNA positions). The most frequently occurring isoform, isoform 2, uses an alternate splice donor site 62 nucleotides into intron 2 (58% of clones, Fig. 1). Other splice isoforms include both an in-frame deletion of six nucleotides from the end of Exon 2, removing the c.1007A>G variant, as well as complex isoforms that include nucleotides from intron 2 with partial or complete skipping of exon 3. We occasionally (<6%) identified isoforms with missplicing in exon 4 (Fig. 1), incorporating 4 bp of intron 4. Although this missplicing was also seen in the control sample, of the 55 control clones sequenced only one harbored this 4 bp insertion at the end of Exon 4 (isoform 12, Fig. 1), with 98% of the clones mapping to the wild-type splice product, isoform 1 (Fig. 1).

Figure 1.

Aberrant splicing. Different splicing events identified after cDNA amplification of EPG5 (between exon 2 and 5) in control (Ct) and patient (Pt) fibroblasts compared to the normal isoform (NM_020964.2). The per cent of sequenced clones from each sample that map to the indicated isoforms are provided in the table along with the total number (n) of clones sequenced. Isoforms that are predicted to encode a premature termination codon are labelled in red. Coordinates for the aberrant splice sites are provided below each specific isoform and reference the normal isoform cDNA sequence.

Most of the isoforms (70%) identified would harbour a premature termination codon (PTC) (Fig. 1). For the splice products that encode part of intron 2, a PTC is inserted three residues downstream of the p.Gln336Arg missense mutation, such that the predicted protein sequence would be p.Gln336ArgfsX4. These isoforms encoding a PTC have, at least partially, evaded NMD. NMD is not a fully efficient process (Isken and Maquat, 2007) and different mechanisms can explain this escape, such as exon junction proximity (Nagy and Maquat, 1998), presence of upstream open reading frames (Stockklausner et al., 2006) or downstream translation reinitiation (Hamid et al., 2010). In our patient, the remaining 30% of transcripts that do not harbor a PTC are either normal splice products with the p.Gln336Arg missense mutation or carry an in-frame deletion (isoforms 3 and 11, Fig. 1).

In summary, we show that the most common pathogenic mutation in the Vici syndrome-associated gene EPG5, the missense NM_020964.2; c.1007A>G, p.Gln336Arg, induces a more complex effect than expected. It not only produces the normal isoform with the missense mutation but also multiple misspliced mRNA isoforms. Some of these isoforms are targeted for NMD, as shown by the decreased level of EPG5 mRNA. All of the variants in EPG5 associated with Vici syndrome are null alleles (splice site, stop gain or frameshift) with the exception of few missense alleles (Byrne et al., 2016a). With the proof that the missense p.Gln336Arg induces a splicing defect, it is possible that the other missense variants may affect the splicing, either directly by deleting or creating a splice site or affecting splicing regulatory elements (Ward and Cooper, 2010). Therefore, performing mRNA analysis in patients with missense variants might reveal that Vici syndrome is only associated with loss of function due to null alleles, where phenotype variability could be associated with the amount of residual protein produced (Byrne et al., 2016a). With the increased generation of sequencing data, laboratory-based evidence of pathogenicity is important, especially for mutations that appear relatively common within a specific population. We hope that this functional evidence will be useful for genetic counseling, as well as contribute to the understanding of the biological underpinnings of this disease.