Genomic screening of candidate genes in Family OAK 587

Exonic mutations in OPA1, OPA3, PANK2, POLG, C10orf2 and in the complete mitochondrial genome (16.569 bp encoding 37 genes and the control region, NC_012920.1) as well as repeat expansions in FXN (Friedreich’s ataxia) and ATXN7 (spinocerebellar ataxia 7) were excluded. Gene rearrangements, including deletions and duplications at the OPA1 locus were excluded by multiplex ligation-dependent probe amplification in two affected subjects and the father (SALSA MLPA KIT P229-B1 OPA1, MRC-Holland). Larger chromosomal rearrangements were excluded by comparative genome hybridization applying a Nimblegen chromosome 3 specific 385K probe array (processed by ImaGenes) in one of the affected patients. The transcript NM_130837.1 is the most complete of the currently 12 protein coding OPA1 transcripts listed in Ensembl release 74, coding for a protein of 1015 amino acids and it is recommended by the OPA1 nomenclature working group of Mitodyn.org. The known OPA1 variant c.1146A>G, p.I382M (NM_015560.2) (Schaaf et al., 2011) corresponds to c.1311A>G, p.I437M based on transcript NM_130837.1. The nomenclature of all following variants refers to transcript NM_130837.1.

Molecular genetic investigation of OPA1 transcripts from patient-derived primary fibroblasts

For standard cell culture conditions of primary fibroblasts from pedigree members of Family OAK 587 as well as Puromycin treatment for detection of nonsense mediated messenger RNA decay, see the online Supplementary material. After complementary DNA synthesis of RNA from puromycin-treated fibroblasts, screening for mutations on the OPA1 transcript level was performed with primer pairs that amplify eight overlapping fragments covering the entire coding sequence of transcript NM_130837 (Supplementary material). Average fragment length was ~500 bp. Primer pairs that flank the alternative exons 4, 4b and 5b amplify multiple complementary DNA fragments of different lengths. These fragments were separated on an agarose gel, excised and cloned for mutation analysis as well as for verification of intact exon–exon boundaries.

Pyrosequencing assays were designed to discriminate between the OPA1 alleles and perform quantitative analysis of allelic expression for the common single nucleotide polymorphisms (SNP) rs7624750 in exon 4 and rs9851685 in exon 21. For primer sequences see the Supplementary material. The primers have been designed using the PyroMark Assay Design 2.0 software (Qiagen). Pyrosequencing was performed using the PyroMark Q96 ID instrument, the sample preparation and the data analysis according to the manufacturer’s instructions.

Western blot and assessment of mitochondrial morphology

Semi-quantitative analysis of OPA1 protein bands was performed from primary fibroblast cell lines of patients and a control (for details, see Supplementary material). Mitochondrial morphology was assessed in fibroblasts of all affected subjects of Family OAK 587 and their two parents as well as an age- and gender-matched control for each individual of this family. Their fibroblasts were cultivated and analysed in parallel (Supplementary material).

Screening for the deep intronic OPA1 mutation in large patient and control cohorts

Three hundred and sixty-nine patients with unexplained bilateral optic atrophy negative for exonic OPA1 mutations and 241 normal controls were screened for the c.610+364G>A by amplification of a 1166 bp fragment containing the entire intron 4b and a subsequent restriction fragment length polymorphism (RFLP) assay (Supplementary material).

Exon sequencing of candidate genes reveals a rare missense variant in OPA1

No exonic mutation was detected in OPA1, PANK2, POLG, C10orf2, OPA3, mitochondrial DNA and no repeat expansion in FXN and ATXN7, except for a known rare missense variant c.1311A>G/p.I437M in OPA1. The unaffected father in our family was homozygous for p.I437M, whereas the mother was negative for the variant. Correspondingly, the three affected subjects, as well as their unaffected siblings, were heterozygous for c.1311A>G/p.I437M (henceforth p.I437M), which indicates that this variant per se is not pathogenic, neither in heterozygous nor in homozygous state (Fig. 3). Analogous to a previously described family with the p.I437M variant in a compound heterozygous state and a complex early onset inherited optic neuropathy phenotype (Schaaf et al., 2011), we hypothesized that there is a second OPA1 mutation (in trans) in our family that—in concert with p.I437M—leads to the severe multisystemic neurodegenerative phenotype seen in the affected family members. Supporting this hypothesis we found by microsatellite marker-based haplotype reconstruction that all affected family members inherited the same maternal OPA1 allele, whereas the three unaffected siblings inherited the opposite maternal OPA1 allele (data not shown).

Allelic imbalance of OPA1 transcripts in patient fibroblasts

We therefore extended OPA1 mutation screening to the transcript level. RNA isolated from Patient II:2 fibroblasts was used for reverse transcriptase PCR of overlapping complementary DNA fragments. A first round of qualitative complementary DNA analysis did not reveal point mutations or aberrant splice products. We then performed relative quantification of allelic expression of OPA1 transcripts applying pyrosequencing technology. This assay revealed a strong allelic imbalance with a ratio of ~4:1 in favour of the paternally-derived transcripts, whereas equal allele ratios were found at the genomic DNA level (Table 2), indicating reduced steady-state levels of the maternally-derived OPA1 transcripts. These might be caused by reduced transcript stability due to nonsense-mediated messenger RNA decay of transcripts with premature termination codons. To suppress nonsense-mediated messenger RNA decay-dependent allelic decay we used puromycin treatment in patient derived fibroblasts. After this treatment, allelic balance on RNA level could be nearly restored back to normal, as proven by pyrosequencing. Control fibroblasts of healthy donors showed no allelic imbalance and no effect upon puromycin treatment (Table 2).

Identification of a deep intronic mutation that activates a cryptic exon

We therefore repeated qualitative transcript analysis with complementary DNA (cDNA) from puromycin-treated patient fibroblasts. Gel electrophoretic analysis of one of these complementary DNA amplicons yielded several additional products (Fig. 4). Gel extraction and analysis of these bands showed insertion of additional sequences between exons 4b and 5 (Fig. 5). These additional sequences derived from the intron between exon 4b and 5 and thus constitute a cryptic exon activated in these aberrant transcripts. Two forms of aberrant cDNA products were found: the larger product contained a 93 bp sequence (exon C+; corresponding to c.610+366 to c.610+458; NM_130837.2) that introduces a UAG stop codon after 87 bp into the open reading frame (Supplementary material), whereas the smaller aberrant cDNA contained a shorter variant of the cryptic exon (exon C) that comprises 61 bp (corresponding to c.610+366 to c.610+426; NM_130837.2). The latter starts at exactly the same nucleotide as exon C+ but lacks its last 32 nucleotides. Insertion of exon C into the transcript induces a frameshift and results in a premature termination codon another 84 bp downstream in exon 5. In both cases translation of the maternal allele is aborted before or within exon 5 supporting the hypothesis that nonsense-mediated messenger RNA decay is the cause for the observed allelic imbalance and that haplo-insufficiency contributes to the molecular pathogenesis in this family.

Open in a separate window

Figure 4

Unmasking of aberrant OPA1 transcripts in puromycin treated patient fibroblasts. Deep intronic mutations in OPA1 reverse transcriptase PCR amplification with primers located in exons 2 and 6 result in multiple cDNA products due to the alternation of constitutively (dark grey in scheme) and facultatively spliced exons (light grey) in this part of the OPA1 gene. The presence of a heterozygous coding SNP in exon 4 (rs7624750; depicted as A/G in the scheme) enabled to differentiate between maternal and paternal-allele derived transcripts in the analysed patient (see text). The pattern of reverse transcriptase PCR products obtained with RNA from untreated fibroblasts of Subject II:1 and the control is virtually identical. However treatment of the patient’s fibroblasts with puromycin before RNA isolation resulted in a drastically altered pattern with additional bands (arrow) and reduced amounts of regular fragments.

Open in a separate window

Figure 5

Novel splicing acceptor sites in intron 4b of OPA1 and resulting cryptic exon inclusion. Base transitions of G>A at positions 360 bp or 364 bp downstream of exon 4b create novel splice acceptor sites that exonize downstream sequences of 65 bp / 97 bp or 61 bp/93 bp. Top: Family OAK 587 (index) and Family OAK 302 share the c.610+364G>A mutation, whereas Families OAK 38 and OAK 344 share the mutation c.610+360G>A (centre). The cryptic exon is exclusively found in transcripts of the maternal allele (Family OAK 587) which can be determined e.g. by an SNP in exon 4 (rs7624750, bottom panel). Black bars = exons; dashed lines = introns; red bars = cryptic exons.

Using the heterozygous coding SNP rs7624750 in pyrosequencing as a marker to differentiate between paternal and maternal allele-derived transcripts we found that all maternal allele-derived transcripts were mis-spliced.

Sequencing of the intron between exons 4b and 5 in DNA samples of the affected patients unveiled a single nucleotide exchange G>A at a position 364 bp downstream of exon 4b and adjacent to cryptic exon C (chr3:193335990G>A). The mother and all of the affected siblings were heterozygous for this intronic mutation, whereas the father and the unaffected siblings did not carry the mutation (Fig. 3).

The bioinformatic tool Human Splicing Finder (Desmet et al., 2009) predicts that the deep intronic mutation, henceforth c.610+364G>A (reference: NM_130837.2), creates a novel splice acceptor site with a score of 92.2. The mutation c.610+364G>A was neither observed in 241 healthy Caucasian control samples nor in the 1092 genomes of the 1000 Genome Project (Genomes Project Consortium et al., 2012).

Additional identification of deep intronic mutations in patients with optic atrophy

To further confirm the pathogenicity of the c.610+364G>A OPA1 mutation as well as to determine its frequency in patients with unexplained optic atrophy, we investigated 369 index cases in which conventional OPA1 screening failed to detect the disease-causing mutation. Our RFLP analysis in intron 4b revealed three further index cases (OAK 302, OAK 38 and OAK 344), one carrying the same mutation c.610+364G>A as found in the original family and two other index patients with a G-to-A exchange 4 bp upstream (c.610+360G>A reference: NM_130837.2) (Fig. 5). Subsequent cDNA analysis with blood samples from two of these patients confirmed that both mutations result in aberrant splicing of OPA1 transcripts with pseudoexon activation between exons 4b and 5. Yet in the patients with the c.610+360G>A mutation, the 5’ acceptor site is shifted by 4 nucleotides further upstream whereas the alternative 3’ donor sites of the cryptic exon are conserved for both deep intronic mutations (as seen for exon C+; Supplementary material). Neither of the two newly identified deep intronic mutations were observed in the 1092 genomes of the 1000 Genomes Project Consortium (Genomes Project Consortium et al., 2012), demonstrating that they are not simply frequent SNPs. Each of the three additional deep intronic mutation carriers showed a classic, slowly progressing optic atrophy and no signs of extraocular neurodegenerative impairment.

Reduced OPA1 protein level in affected mutation carriers

We further investigated whether deep intronic mutations result in lower OPA1 protein levels, or whether compensatory mechanisms counteract the genetic defect. Western blot analysis with whole-cell lysates of primary fibroblasts from several members of the OAK 587 and OAK 302 families revealed a reduction of OPA1 protein normalized for ACTB (Fig. 6). This reduction caused by the deep intronic mutation was in the same range as in a patient with a well characterized stop mutation in exon 7 (c.868C>T/p.Arg290stop) (Puomila et al., 2005). We also observed a slight reduction of OPA1 protein in fibroblasts of the unaffected father in OAK 587 (homozygous carrier of p.I437M), suggesting some impairment of this modifier variant on OPA1 protein stability. Comparative immunoblotting with an antibody against the mitochondrial outer membrane protein TOMM40 revealed no significant difference in protein levels compared with the control indicating that reduced amounts of OPA1 do not by large affect mitochondrial mass in fibroblasts (Fig. 6).

Open in a separate window

Figure 6

Reduced OPA1 protein amount in patients with deep intronic mutation. Semi-quantitative western blot immunodetection with antibodies against OPA1 (top), ACTB (centre) and TOMM40 (bottom) in whole fibroblast cell lysates from a patient with autosomal dominant optic atrophy with heterozygous c.868C>T/p.Arg290stop in OPA1 (nonsense-mediated messenger RNA decay control), the affected index patient of Family OAK 302, two affected patients (Subjects II:1 and II:2) as well as the parents of Family OAK 587, and a healthy control (wild-type) (from left to right). A clear reduction in OPA1 protein amount can be observed in all samples except the control and the unaffected father (only slight reduction).

Increased mitochondrial fragmentation in mutation carriers

To assess the functional consequences of the two OPA1 variants reported here (c.610+364G>A; and p.I437M), we assessed the morphological characteristics of the mitochondrial network in fibroblasts of the three affected siblings and both parents in Family OAK 587. Compared to age-matched controls, surface (P < 0.05), aspect ratio (P < 0.001), and form factor (P < 0.001) of the mitochondria were significantly reduced in the three affected subjects (Fig. 7A and B). This finding is in line with the well-established effect of OPA1 mutations that lead to reduced mitochondrial fusion and increased mitochondrial fragmentation, respectively (Amati-Bonneau et al., 2005, 2008; Olichon et al., 2007; Zanna et al., 2008). Although clinically asymptomatic, the mother of the affected siblings (Subject I:2) also showed a significantly reduced surface (P < 0.01), aspect ratio (P < 0.05), and form factor (P < 0.01) of the mitochondria (Fig. 7), indicating that even the deep intronic mutation c.610+364G>A per se leads to OPA1-characteristic dysregulation of mitochondrial dynamics. Mitochondria of the father (Subject I:1) were reduced in aspect ratio (P < 0.001) and form factor (P < 0.001), but not in surface (P > 0.05) and volume (P > 0.05) (Fig. 7), compatible with a moderate effect of the homozygous p.I437M modifier variant on mitochondrial dynamics.

Open in a separate window

Figure 7

Mitochondrial morphometry. (A) Mitochondrial network stained with MitoTracker Green FM(R) and visualized with a fluorescence microscope. Compared to an exemplary control (i), where mitochondria build a network of lengthy and closely connected mitochondrial alignments, the mitochondrial network is fragmented in all three affected siblings (ii–iv). Similarly, also mitochondria of the mother (vi, Subject I:2) show signs of fragmentation and, to a lesser extent, mitochondria of the father as well (v, Subject I:1). (B) Statistical analysis of mitochondrial morphometry in fibroblasts of the three affected siblings (dark red), the father (dark blue) and the mother (dark grey) of Family OAK 587 compared to age-matched controls. Four aspects of morphology were analysed: (i) surface; (ii) aspect ratio (major axis/minor axis); (iii) form factor (surface²/4 π volume); and (iv) volume. *P < 0.05; **P < 0.01; ***P < 0.001; n.s. = not significant.

The p.I437M variant acts as a phenotypic modifier to reinforce the pathogenic effect of loss of function mutations

Mechanisms determining severe plus phenotypes rather than pure optic atrophy have been vague so far. Putative dominant negative effects of missense mutations in the GTPase domain, exerting a gain-of-function effect, are frequently associated with a severe phenotype (Amati-Bonneau et al., 2008; Yu-Wai-Man et al., 2010b). An additional mechanism might be a combined mutational effect from a biallelic OPA1 state of two OPA1 variants (which, for example, jointly lower the amount of functional OPA1 protein below a certain critical threshold). For example, it has been shown that, like in Family OAK 587, the I437M variant in compound heterozygous state with another OPA1 mutation (c.2873_2876del/p.V958Gfs*3) presents with a severe optic atrophy plus phenotype (Schaaf et al., 2011). Our findings corroborate and, at the same time, significantly extend this notion. We show that the later scenario does not follow a simple recessive mode of inheritance but rather that the I437M variant acts as an intralocus modifier. This hypomorphic allele causes a severe optic atrophy plus phenotype when combined with a compound heterozygous OPA1 null mutation. That second OPA1 mutation might be either exonic (Family DUK2976) or intronic (Family OAK 587).

The notion of I437M acting as an intralocus modifier is based on the finding that in all three families [OAK 587, DUK2976 and the family described by Schaaf et al. (2011)] the p.I437M variant is associated with a severe ‘optic atrophy plus’ phenotype if occurring with another OPA1 mutation. In contrast, these three truncating OPA1 mutations alone cause a pure optic atrophy phenotype. Compared to the compound heterozygous null mutations, the I437M variant alone does not cause disease per se, neither in heterozygous nor in homozygous state as observed in Family OAK 587. In fact, the homozygous father Subject I:1 is still healthy at the age of 76 years, without evidence for subclinical optic atrophy. In addition to our clinical observation in three independent families, several other lines of evidence also support an adverse effect of the p.I437M on OPA1 function.

First, amino acid sequence alignment of OPA1 and its homologues in diverse species shows that the isoleucine at position 437 (Ile437) localizes to the GTPase domain of OPA1 and is fully conserved in vertebrates, invertebrates (Drosophila melanogaster and Caenorhabditis elegans) as well as in yeast (Supplementary material).

Second, the p.I437M variant is very rare (5/~6500 exomes) in the Exome variant server (http://evs.gs.washington.edu/EVS/), thus demonstrating that it is not simply a frequent SNP.

Finally, Ile437 in OPA1 corresponds to the conserved Ile120 in the dynamin proteins DNM1, DNM2 and DNM3 and is part of a dynamin superfamily-specific β-strand (β2b). This β-strand directly follows the ‘cis-stabilizing loop’ that stabilizes nucleotide-dependent dimerization of the GTPase domains (Chappie et al., 2010). Furthermore, Ile120 points into the hydrophobic core of the GTPase domain and interacts, for example, with Ile101 of the preceding α-helix (Supplementary material). A substitution of Ile120(dynamin)/Ile437(OPA1) to methionine might locally destabilize the GTPase domain and thus partially interfere with its nucleotide-dependent function (Oliver Daumke, personal communication).

OPA1 plus: an autosomal-recessive ataxia mimic with cervical cord atrophy

The clinical presentation of the affected patients in Families OAK 587 and DUK2976—namely, early-onset, multisystemic neurodegenerative disease—mimics a picture well known from a variety of autosomal-recessive neurodegenerative diseases, such as autosomal-recessive ataxias, hereditary spastic paraplegias or lysosomal storage disorders, which might likewise include optic atrophy, spasticity and polyneuropathy [e.g. Friedreich’s ataxia, spastic paraplegia type 7, Gaucher’s disease or SPOAN (spastic paraplegia, optic atrophy and neuropathy) (Macedo-Souza et al., 2009)]. In addition, also the frequently observed reduced penetrance in OPA1 [as seen in Subject I:2 of Family OAK 587 and as described in the literature (Cohn et al., 2007)] might mislead towards an apparently recessive disease in these subjects. Thus, OPA1 mutations should be taken into the differential diagnosis of patients presenting with such seemingly recessive disease, and intronic mutations should be searched for if exonic OPA1 sequencing remains inconclusive.

Although most features of the affected siblings of our index family fit well with the reported phenotype of optic atrophy plus syndromes (Yu-Wai-Man et al., 2010b), our clinical findings modify and extend this phenotypic spectrum. Unlike most patients with optic atrophy plus (Amati-Bonneau et al., 2005; Yu-Wai-Man et al., 2010b; Schaaf et al., 2011), none of our patients was suffering from hearing impairment, demonstrating that this feature is not an obligatory part of the OPA1-plus cluster. Moreover, our MRI results demonstrate that atrophy of the cervical spinal cord can be part of the disease spectrum, whereas cerebellar atrophy might be only mild, despite severe ataxia [SARA (Scale for the Assessment and Rating of Ataxia) scores ≥29 points]. This resembles findings in patients with Friedreich’s ataxia, where atrophy of the spinal cord is also common, whereas cerebellar atrophy is mild even in late disease stages. This parallel might indicate that, as in Friedreich’s ataxia, also OPA1-ataxia might result largely from a spinal degeneration and only to a smaller extent from cerebellar degeneration.

Identification of more generalizable mutational mechanisms?

The mutational mechanisms identified in this study—deep intronic mutations and intragenic modifiers—might also be of relevance beyond the field of OPA1 genetics. This notion is corroborated by the previous identification of deep intronic mutations in another hereditary neurodegenerative disease (ataxia telangiectasia; deep intronic mutation in ATM; Cavalieri et al., 2013) as well as in a retinal degeneration (Leber’s congenital amaurosis, deep intronic mutation in CEP290; den Hollander et al., 2006). Taken together with our findings from OPA1, these observations make it highly likely that deep intronic mutations can be found also in a wide range of other neurodegenerative and optic neuropathy diseases where deep intronic mutations are not yet known and where the relative share of genetically unexplained patients is still high. This is important to acknowledge as these mutations will be missed even if using the current state-of-the-art whole-exome approaches.

The presence and action of genetic modifiers in neurodegenerative disease is frequently postulated, however, genetic mapping of such loci or even the identification of such factors has been rarely achieved. In Leber’s hereditary optic neuropathy, the other major form of inherited optic neuropathy, there is strong evidence that certain mitochondrial haplogroups govern clinical expression in subjects with principle mitochondrial DNA mutations at m.11778 (Hudson et al., 2007; Ji et al., 2008). Moreover, a modifier that may explain the pronounced gender difference in penetrance in this disease has been mapped on Xp21-q21 (Hudson et al., 2005) and Xq25-27.2 (Shankar et al., 2008), respectively, though their identity is still unknown.

Reports on intragenic modifiers that modulate phenotype severity and disease penetrance (as the p.I437M OPA1 variant in this paper) are still limited (Gouya et al., 2002). Nonetheless we suggest that such a mechanism represents an attractive explanation for expression variability of traits that rely on dose-sensitive or threshold dependent functions. As variability in disease severity and disease phenotype within families as well as between subjects carrying the same mutation have been described in a large number of genetic diseases [e.g. for SOD1 mutations in amyotrophic lateral sclerosis (Synofzik et al., 2010) or PRNP mutations in hereditary prion disease (Synofzik et al, 2009)], it is likely that such intragenic modifiers exist also for other neurological/ophthalmological diseases.

We thank the patients and their family members for participating in our study, Doron Rapaport (Tübingen) for the anti-TOMM40 antibody, Beate Leo-Kottler and Bernhard Jurklies for ophthalmological examinations and recruitment of patients with Autosomal Dominant Optic Atrophy, and in particular Oliver Daumke (Berlin) for the modelling of the 3D structures of the OPA1 GTPase domain and his support in evaluating the putative function of I437M.