EXOSC8 mutations alter mRNA metabolism and cause hypomyelination with spinal muscular atrophy and cerebellar hypoplasia

The exosome is a multi-protein complex, required for the degradation of AU-rich element (ARE) containing messenger RNAs (mRNAs). EXOSC8 is an essential protein of the exosome core, as its depletion causes a severe growth defect in yeast. Here we show that homozygous missense mutations in EXOSC8 cause progressive and lethal neurological disease in 22 infants from three independent pedigrees. Affected individuals have cerebellar and corpus callosum hypoplasia, abnormal myelination of the central nervous system or spinal motor neuron disease. Experimental downregulation of EXOSC8 in human oligodendroglia cells and in zebrafish induce a specific increase in ARE mRNAs encoding myelin proteins, showing that the imbalanced supply of myelin proteins causes the disruption of myelin, and explaining the clinical presentation. These findings show the central role of the exosomal pathway in neurodegenerative disease.

T he degradation of messenger RNAs (mRNAs) is an important regulatory step, which controls gene expression 1,2 . Unstable mammalian mRNAs contain AUrich elements (AREs) within their 3 0 -untranslated regions. Rapid degradation of ARE-containing RNAs is performed by a multiprotein complex, the exosome 3,4 . The versatility and specificity of the exosome regulate the activity and maintain the fidelity of gene expression 5 . Both in yeast and humans, nine proteins organize the 'exosome core' in a two-layered ring. The central hexamer channel is composed by six subunits (Rrp41p/EXOSC4, Rrp46p/ EXOSC5, Rrp45p/PM/Scl-75/EXOSC9, Rrp42p/EXOSC7, Mtr3/ EXOSC6 and Rrp43p/Oip2/EXOSC8), while the cap consists of three proteins (Rrp4/EXOSC2, Rrp40/EXOSC3 and Csl4/SOCS4) 4,6 . The exosome degrades RNA starting at the 3 0 -end by an exoribonucleolytic function and it also has an endoribonucleolytic function 7,8 . The catalytic activity of the exosome core is provided through the association with other proteins (RRP44/DIS3, RRP6/ PM/Scl-100/EXOSC10 ribonucleases) 1 . ARE recognition requires ARE-binding proteins that interact with the exosome for recruitment, thereby promoting the rapid degradation of target RNAs 9 .
The human exosome also regulates gene expression via diverse RNA processing reactions 10 . Many cellular RNAs that play key roles in important cellular processes such as translation (ribosomal RNAs, transfer RNAs and small nucleolar RNAs) and mRNA splicing (small nuclear RNAs) are produced as precursor molecules that are trimmed from their 3 0 -ends by the human exosome 11 . This complex organization of the exosome provides the versatility needed to cope with the huge variety of RNA substrates in the cell 12 . However, detailed in vivo analyses are technically challenging and many questions remain unresolved.
Here we report that deficiency of a core component of the human exosome leads to severe infantile overlap phenotype of psychomotor deficit, cerebellar and corpus callosum hypoplasia, hypomyelination and spinal muscular atrophy (SMA).

Results
Clinical presentation in 22 patients from three pedigrees. First we studied a large Hungarian family of Roma ethnic origin, where 18 children presented between 2-4 months of age with failure to thrive, severe muscle weakness, spasticity and psychomotor retardation. Vision and hearing were impaired in all patients and deterioration was usually triggered by inter-current infections. All affected children died of respiratory failure before 20 months of age (Fig. 1a, Supplementary Table 1). Detailed diagnostic workup excluded known metabolic, neurodegenerative and common genetic disorders. Electrophysiology was performed in one patient only (P1-V:10) and was not conclusive. Brain magnetic resonance imaging (MRI) showed variable abnormalities including vermis hypoplasia, immature myelination, cortical atrophy and thin corpus callosum (Supplementary Table 1).
EXOSC8 mutations were also identified in two additional patients from an independent Hungarian Roma family (Fig. 1b, and in two affected siblings from a consanguineous Arab-Palestinian family (Fig. 1e). The clinical presentation of the additional patients was compatible with a progressive, infantile onset neurological disease, presenting with severe muscle weakness, respiratory problems, developmental delay and early death (Supplementary Table 1). Vermis hypoplasia was more prominent in the third pedigree (Fig. 1f), while immature myelination was reported in pedigree 2 (Fig. 1d). Weakness in P3 was proximal more than distal with attendant tongue fasciculations. Motor neuronopathy was noted on electrophysiological examination in P3-II:1.
Muscle biopsy in patient P1-V:10 at 5 months of age detected variations in fibre size and increased subsarcolemmal nuclei, but no signs of SMA. Cytochrome c oxidase (COX) negative fibres were noted on histochemical staining and activities of respiratory chain complexes I and IV were moderately decreased (Supplementary Note 1). We excluded mitochondrial DNA (mtDNA)-mediated mechanisms for multiple respiratory chain enzyme deficiency (mtDNA deletion/depletion, point mutations). Muscle biopsy of another patient (P3-II:1) at 2 years of age showed groups of atrophic and hypertrophic fibres compatible with SMA.
Autopsy in eight patients detected profound lack of myelin in the cerebral and cerebellar white matter and in the spinal cord, predominantly affecting the descending lateral fibre tracts, while myelination was normal in the peripheral nerves ( Fig. 2a-n). The lack of myelin in the brain and spinal cord (Fig. 2g,n) was similarly severe as observed in a patient with Pelizaeus-Merzbacher disease (PMD) (Fig. 2m). However myelin basic protein (MBP) staining was stronger in our patients (Fig. 2k) as compared with the PMD patient (Fig. 2j), indicating the different composition of defective myelin. Autopsy was not performed in the deceased patient from P3-II:1, but MRI at 5 years of age did not indicate hypomyelination.
Homozygosity mapping and exome sequencing. Genome-Wide Human SNP Array 6.0 (Affymetrix) in six affected family members of the first pedigree revealed two broad regions of homozygosity on chromosome 13 (36212278-37767059 bp, size: B1.55 Mb; and 43243791-44640995 bp, size: 1.4 Mb) (Fig. 3a) 13 . Exome sequencing in genomic DNA of two patients (P1-V:10, P1-V:29) identified seven shared rare (major allele frequency o0.01) homozygous protein altering variants (Fig. 3b), but only one, the c.815G4C, p.Ser272Thr mutation in EXOSC8 (Gene ID:11340, NC_000013.10, mRNA:NM_181503.2) (Fig. 3d) was located in the larger B1.55 Mb homozygous region. This mutation affects a conserved amino acid and resulted in significantly decreased EXOSC8 protein ( Fig. 3e-g), and no other variants were found in complementary DNA (cDNA) of EXOSC8. The location of the mutation implies that it may affect the opening of the RNA channel within the exosome, where funneling of RNAs takes place in single-stranded conformation by an unwinding pore (Fig. 3f). This mutation segregated with the disease in 19 individuals within the family (8 affected: homozygous; 11 unaffected: heterozygous or wild type) (Fig. 1a). The same homozygous mutation was identified in two siblings with an identical clinical presentation from a second Hungarian Roma family (pedigree 2) from the same region. None of the other six homozygous mutations detected by exome sequencing of the index patients in pedigree 1 were homozygous in the patients in pedigree 2, thus excluding the possibility that these other mutations were causing the disease.
To estimate the frequency of this mutation in Roma population, mutation testing was carried out in 63 anonymized Roma controls from Bulgaria, of whom 33 were homozygous for an identical 229522 bp chromosome 13q13.1 haplotype (rs582091 to rs7327020) around c.815G4C, p.Ser272Thr (Supplementary Note 2). This screening identified two heterozygotes, that is, carrier rates of B3% in the general Roma population, which is in agreement with the frequency rates of other Roma disease causing mutations 14,15 .
A different missense mutation, c.5C4T, p.Ala2Val was identified on exome sequencing in two affected siblings of a consanguineous Palestinian family (pedigree 3) with SMA, cerebellar and corpus callosum hypoplasia (Fig. 3c). Homozygosity mapping in this pedigree resulted in the identification of three disease-linked genomic regions, one of them was an 8.4 Mb spanning chr13:31642481-40039652 (Hg19) (rs7996548 to rs12873765). Exome sequencing in genomic DNA of P3:II-3 identified four shared rare (minor allele frequencyo0.01) homozygous protein altering variants but only one of them affected an evolutionary conserved residue (Fig. 3e). This variant, c.5C4T, p.Ala2Val (chr13: 37574947 C4T) in the EXOSC8 gene segregated with the disease in the family, was not present in dbSNP138 and was also absent from the 6503 exome analyses available at the Exome Variant Server, NHLBI GO Exome Sequencing Project (ESP), Seattle, WA, USA (URL: http:// evs.gs.washington.edu/EVS/) (accessed on 26 January 2014) and in fibroblasts of the patient, immunoblotting detected severely decreased EXOSC8 protein (Fig. 3d). The c.5C4T mutation in the second codon of EXOSC8 might therefore either interfere with the Kozak consensus sequence and/or cause mRNA instability probably through interference with the 5 0 mRNA capping mechanism 16,17 . Nonsense-mediated messenger decay could be excluded because mutant copy numbers could not be increased after 300 mM puromycin treatment ( Supplementary  Fig. 1).
Downregulation of the yeast ortholog of human EXOSC8. RRP43, the yeast ortholog of the human EXOSC8 has a low degree of conservation with the human gene, lacking the C-terminal region, where the human mutation p.Ser272Thr is located. Based on the observed mitochondrial dysfunction in patient V:10 we investigated whether a defect in the yeast ortholog of EXOSC8 affects mitochondrial function. We took advantage of a promoter-shutoff tetRRP43 strain 18 . Wild-type and tetRRP43 cells were grown on non-fermentable carbon sources in the presence/ absence of doxycycline 0.125 mg ml À 1 . Treatment with doxycycline caused a strong growth reduction on glucose of tetRRP43 while only a slight reduction on glycerol suggesting that RRP43 downregulation does not primarily affect oxidative phosphorylation capacity (Fig. 4a). We speculate that in the presence of glycerol, that slows cell growth and metabolic rate, the exosome complex is able to maintain a basal RNA abundance, despite the downregulation of tetRRP43. Furthermore, the reduced RRP43 gene expression caused significantly decreased respiration that was paralleled by a decreased amount of mitochondrial cytochromes compared with wild type, although their structural integrity was not affected (Fig. 4c). No difference in mitochondrial translation rate was observed between wildtype and tetRRP43 mutant in the presence/absence of doxycycline ( Fig. 4d) suggesting that reduced expression of RRP43/EXOSC8 may have a secondary effect on mitochondrial function, possibly due to disturbed mRNA processing of mitochondrial genes containing ARE elements 19 .
Increased gene expression of selected ARE-containing mRNAs. After the exclusion of a primary mitochondrial phenotype in RRP43 depleted yeast we studied the possible mRNA targets of EXOSC8 in human cells. EXOSC8 encodes a 3 0 -5 0 exoribonuclease that specifically interacts with ARE-containing mRNAs    Table 2) and myelin-associated oligodendrocyte basic protein (MOBP 48.5fold, P ¼ 0.0158, Supplementary Table 2). However, no significant change was detected in mRNA levels of any of the other tested AU-rich and non-AU-rich genes (Supplementary Table 2). Based on the selective effect of EXOSC8 downregulation on oligodendroglia-related genes, quantitative reverse transcription-PCR (qRT-PCR) analysis has been performed in EXOSC8 downregulated human MO3.13 oligodendroglia cells (kind gift of Prof. Nalbantoglu, McGill). In support of our previous findings, we detected a highly significant increase in expression of MBP (4100-fold, P ¼ 0.013) (Fig. 5c), which led to increased MBP protein levels in differentiated oligodendroglia cells detected by immunostaining and immunoblotting ( Fig. 5h-j).
In patient fibroblasts (P3-II:1) we detected a significant Bfourfold increase (P ¼ 0.01984) in the expression of SMN1, which is the major causative gene leading to SMA (Fig. 5d). No significant change was detected in the other studied genes (16 ARE and 16 non-ARE mRNAs), suggesting that only a subset of ARE transcripts was significantly affected by EXOSC8 dysfunction in both cell types. EXOSC8 downregulation in fibroblasts resulted in significant increase of SMN1 (2.72-fold, P ¼ 0.00038), MBP (6.74-fold, P ¼ 0.01936) and MOBP (2.48-fold, P ¼ 0.04767), but no significant change was detected in mRNA levels of any of the other tested AU-rich and non-AU-rich genes (Fig. 5e Figure 4 | Characterization of the yeast RRP43 and tetRRP43 strains to access mitochondrial function. (a) Oxidative growth phenotype of RRP43 and tetRRP43 strains grown in the presence and absence of doxycycline (0.125 mg ml À 1 ). Equal amounts of serial dilutions (10 5 , 10 4 , 10 3 , 10 2 ) of cells from exponentially grown cultures were spotted onto YP plates supplemented with 2% glucose (left panel) or with 3% glycerol (right panel). The growth was scored after 3 days of incubation at 28°C. (b) Cytochrome spectra of RRP43 and tetRRP43 strains grown in the presence and absence of doxycycline (0.125 mg ml À 1 ). The peaks at 550, 560 and 602 nm (vertical bars) correspond to cytochromes c, b and aa 3 , respectively. The height of each peak relative to the baseline of each spectrum is an index of cytochrome content. (c) Respiratory activity of yeast RRP43 and tetRRP43 strains grown in the absence and in the presence of doxycycline (0.125 mg ml À 1 ). Wild-type RRP43 and tetRRP43 mutant strain were grown in YP medium supplemented with 0.6% glucose. Respiratory rates were normalized to the wild-type strain grown in the presence of doxycycline. (d) Mitochondrial protein synthesis was performed on RRP43 and tetRRP43 strains in the presence and absence of doxycycline (0,125 mg ml À 1 ), after 15 min of incubation with 35 S. Equivalent amounts of protein were separated by SDS-PAGE on a 17.5% polyacrylamide gel, transferred to a nitrocellulose membrane and analyzed by autoradiography.      23 . Confirming the efficiency of the splice-blocking exosc8 MO, RT-PCR indicated the loss of wildtype exosc8 transcript (Fig. 6b, Supplementary Fig. 2). In support of the phenotype caused by the splice-blocking exosc8 MO, we detected similar changes by using the second translation-blocking exosc8 MO in zebrafish. Knock down of exosc8 in zebrafish embryos resulted in a set of phenotypes ranging from mildly to severely abnormal external morphology (Fig. 6a,c). Swimming abilities and touch-evoked escape response at 48 h post fertilization (hpf) were impaired even in embryos with a mild phenotype (Supplementary videos). Importantly, downregulation of exosc8 led to abnormal brain development in islet-1-GFP transgenic zebrafish, which express GFP in the motor neurons of the hindbrain. The degree of brain abnormality correlated with the increased severity of morphant phenotype (Fig. 6d), suggesting that Exosc8 is essential for brain development in zebrafish. Although the recapitulation of an identical phenotype with translation-blocking MO is in support of a specific effect in our experiments, to exclude an off-target effect we introduced coinjection of p53 MO to block off-target cell death.
We performed several attempts to rescue the phenotype of MO-treated zebrafish with different doses of wild-type human EXOSC8 mRNA. We observed a dose-dependent toxic reaction after addition of the wild-type EXOSC8 mRNA to MO-treated zebrafish. The reason behind the toxicity may be that an imbalanced supply of the different exosomal proteins may have deleterious effect on the function of the exosome, as comparative assessment of protein abundance and localization in living cells showed, that EXOSC8 and EXOSC9 are present in a 1:1 stoichiometry within the complex 24 .
We also performed whole-mount immunohistochemistry in zebrafish larvae after exosc8 MO injection with an antibody against the zebrafish MBP (kind gift of Prof. Talbot, Stanford). In un-injected embryos, MBP expression at 4 days post fertilization (dpf) was observed in the ventral hindbrain, spinal cord, along motor axons exiting the spinal cord and the lateral line. In exosc8 morphants the MBP signal on motor axons in somites, on the posterior lateral line and in the head was missing or interrupted (Fig. 7b, Supplementary Fig. 3). We also performed combinatorial labelling (BrainStain Imaging Kit, Molecular Probes) including FluoroMyelin Green stain, which works via the high lipid content of myelin and provided further evidence, independent from immunostaining with MBP, that myelination is primarily impaired in exosc8 MO downregulated zebrafish (Fig. 7c) Severe, abnormal brain Moderate, abnormal brain Moderate, normal brain Mild, normal brain on exosc8 MO-treated zebrafish revealed a clear defect in myelination at 4 dpf (Fig. 7d). Simultaneous MO knock down of zebrafish mbpa (gene ID: 326281, NC_007130.5) and exosc8 was performed to explore whether the phenotype of exosc8 downregulated zebrafish embryos can be rescued by knocking down mbp, which is originally increased in exosc8 downregulated zebrafish. Survival rate of MO injected zebrafish embryos was better in the simultaneous mbpa, exosc8 knockdown group (73.6%) compared with embryos treated with exosc8 MO only (53.7%) (Fig. 8a). Furthermore, in embryos with severe phenotype the pattern of midbrain and hindbrain nuclei was slightly more preserved in simultaneous mbpa, exosc8 MO-treated embryos compared with zebrafish treated with exosc8 MO only (Fig. 8b). Taken together, these data suggest that the neuronal defect caused by exosc8 downregulation primarily impairs myelination by affecting the regulation of mbp genes. Furthermore, we show that additional knock down of mbp improved the survival rate and brain structure of exosc8 MO downregulated zebrafish, suggesting that the original increase of mbp expression may trigger downstream events resulting in loss of MBP and impaired myelination in zebrafish.

Discussion
The first human disease linked to an integral exosome component was EXOSC3 deficiency, which has been identified in pontocerebellar hypoplasia (PCH) and spinal motor neuron disease (PCH1, MIM 607596) [26][27][28] . A broader clinical spectrum was recently suggested in patients with EXOSC3 mutations including isolated cerebellar hypoplasia and spinal anterior horn involvement or intellectual disability, early-onset spasticity and cerebellar atrophy 29 . A severe form of PCH1 has been reported among Czech Roma due to a founder mutation in EXOSC3 (ref. 30).
We discovered pathogenic mutations in the exosomal protein gene EXOSC8 in 22 patients with profound infantile neurodegenerative disease combining features of cerebellar and corpus callosum hypoplasia, hypomyelination and SMA. The similar disease spectrum caused by the defect of a different exosome protein suggest that impaired mRNA metabolism due to exosome dysfunction is the major common pathomechanism. However, while EXOSC3 mutations affect mostly spinal motor neurons and Purkinje cells, oligodendroglia cells are also targets of mutations in EXOSC8. It is possible that different exosome components affecting different ARE mRNAs, potentially based on their length 9 , or mRNA metabolism may have neuronal cell-type specific differences, potentially influenced by the type or localization of the exosome defect. However, both studies clearly suggest the major role of ARE mRNA metabolism in this complex neurodegenerative disease.
In support of a close interaction between EXOSC8 and EXOSC3 within the exosome, reduced EXOSC3 protein levels were detected in our patients carrying EXOSC8 mutations, and also in control cells after siRNA downregulation of EXOSC8 (Supplementary Fig. 4).
Both peripheral and central myelination requires an exact ratio of several myelin proteins 31 . Central nervous system myelin is a multi-layered membrane sheath generated by oligodendrocytes for rapid impulse propagation. It has been recently shown that new myelin membranes are incorporated adjacent to the axon, and simultaneously, newly formed layers extend laterally, leading to the formation of a set of closely apposed paranodal loops 32 . This model can explain the assembly of myelin as a multi-layered structure, where increased amount of MBP, a peripheral membrane protein, may result in premature compaction and may block myelin growth 32 or compaction at aberrant locations, which could also be toxic for myelination 22 . Duplication of PLP1 is a frequent cause of PMD 33-35 and a similar gene dosage effect of PMP22 is responsible for the demyelinating peripheral neuropathy in Charcot-Marie-Tooth disease type 1A 36 . A link between ARE mRNA decay and demyelination is further supported by the progressive leukodystrophy caused by mutations in the AU-specific RNA binding protein (AUH) 37 .
We suggest that increased mRNA levels of ARE-containing myelin proteins (MBP, MOBP) resulting from EXOSC8 deficiency initiate a cascade of downstream events, and the disturbed balance between ARE and non-ARE myelin components results in demyelinating disease. Ultimately, the loss of oligodendroglia cells leads to a secondary decrease of myelin proteins, as shown in exosc8 MO injected zebrafish, where an initial increase, followed by a secondary decrease of myelin proteins resulted in defective myelination.
The genetic causes of isolated PCH are autosomal recessive mutations in genes involved in transfer RNA splicing and processing 38 . Splicing of the pre-mRNAs by the spliceosome depend on small nuclear ribonucleoproteins, which require Spinal Motor Neuron 1 (SMN1) protein for their assembly and defect of SMN1 results in SMA, a leading cause of infantile mortality 39 . The exact mechanisms behind the strict cell-type specific effect of these factors involved in RNA metabolism are still unclear; however, it is possible that some clinical presentations of EXOSC8 and EXOSC3 mutations, and potentially other human exosome related diseases, are due to abnormal RNA splicing.  In summary, patients with EXOSC8 mutations present with a characteristic spectrum of overlapping phenotypes of infantile onset hypomyelination, cerebellar hypoplasia and SMA. The clinical presentation may depend on the type and localization of mutations and provides a clue to unravel the complex molecular pathways caused by the defective exosome function.

Methods
Patients. We have received informed consent from the patients (parents) included in this study. Ethical approval was obtained from the County Durham and Tees Valley 2 Research Ethics Committee (08/H0905/106) and from the Ethics committee of the Hebrew University Hadassah Medical School (0485-09). Written consent to publish the photo of a patient was obtained.
Histological and biochemical analyses of skeletal muscle. Histological and biochemical analyses of skeletal muscle were performed by standard methods 40 .
Autopsy staining methods. The brain and spinal cord samples were fixed in 10% phosphate-buffered formal saline before embedding in paraffin wax. For Luxol Fast Blue staining 10-mm sections were dehydrated and stained with Luxol Fast Blue overnight at room temperature, and then rinsed in alcohol and water to remove excess blue colour. The sections were differentiated with a weak solution of lithium carbonate, followed by 70% alcohol solution. Immunohistochemistry was done following citrate (pH 6) and EDTA (pH 8) pre-treatment for 10 min. Immunohistochemistry was carried out on 5-mm sections with mouse monoclonal antibodies that recognize MBP (1:2000 SM194R, Covance, NJ, USA) and vimentin (1:5600 clone V9 Dako; Copenhagen, Denmark). Rabbit monoclonal antibody was used for p62 1:150 SQTM1 (Santa Cruz Biotechnology, TX, USA). Sections were also stained with haematoxylin. Stainings were performed on at least three different sections.
Cell culture and siRNA transfection. Cultured myoblasts of patient V:10 (pedigree 1) and controls were grown in skeletal muscle cell growth medium and supplement mix (PromoCell) supplemented with 10% (v/v) foetal bovine serum (FBS, Sigma Aldrich) and 4-mM L-glutamine (Invitrogen) and cultured as recommended by the supplier. Fibroblasts of patient P3-II:1 were grown in high glucose Dulbecco's modified Eagle's medium (DMEM, Sigma, Poole, UK) supplemented with 10% FBS. Myoblasts of patient V:10 and a control cell line were immortalized by transduction with a retroviral vector expressing the catalytic component of human telomerase 47 . Silencer EXOSC8 RNA (s22370 siRNA, Ambion-Life Technologies) was transiently transfected in control myoblasts, fibroblasts and in M03.13 human oligodendroglia cells at a final concentration of 5 nm using Lipofectamine RNAiMAX (Invitrogen), according to the manufacturer's specifications. Transfections were repeated on day 3 and cells were harvested on day 6. A non-targeting SilencerSelect Negative Control (#1) was used as a control.
Human oligodendrocytic cells (MO3.13) were grown in DMEM supplemented with 10% FBS. The human oligodendroglia cell line was received from the McGill University, where it was originally characterized. Medium was changed every 2-3 days. MO3.13 cells were differentiated in oligodendrocyte differentiation medium for 7 days. Each experiment was repeated at least three times.
Blue native PAGE (BN-PAGE). BN-PAGE was performed as described 48 .
Yeast strains and culture condition. Saccharomyces cerevisiae strains used in this study are R1158 (deletion consortium wild-type MATa his3D1 leu2D0 met15D0 ura3D0 BY4741 with TTA integrated at URA3) and RRP43 Tet-O 7 -promoter mutant (referred as tetRRP43), kindly provided by T. Hughes 18 . The promoterreplacement system approach allows modulating the downregulation of essential genes by replacing the native promoter with one tetracycline-regulatable promoter that can be repressed by addition of doxycycline to the growth medium.
Cells were cultured in yeast peptone (YP) medium (1% Bacto-yeast extract and 2% Bacto-peptone (ForMediumTM, UK)) 49 . Various carbon sources were added at 2% (w/v) (Carlo Erba Reagents, Italy). Media were solidified with 20 g l À 1 agar (ForMediumTM, UK). For respiration and mitochondria extraction, cells were grown to late-log phase in the YP medium supplemented with 0.6% glucose. Oxidative growth, respiration and cytochrome spectra in yeast were performed. Oxidative growth phenotype was performed by spotting decreasing concentrations of yeast cells on YP medium supplemented either with 2% glucose or with 2% glycerol. Differential spectra between reduced and oxidized cells of a suspension of cells at 500 mg ml À 1 (wet weight) were recorded at room temperature, using a Cary 300 Scan spectrophotometer (Varian, Palo Alto, CA, USA). Oxygen uptake rate was measured at 30°C using a Clark electrode in a reaction vessel of 3 ml of air-saturated respiration buffer (0.1 M phthalate-NaOH, pH 5.0), 10 mM glucose, starting the reaction with the addition of 10 mg of wet weight of cells 50 . Mitochondrial protein synthesis in yeast was performed by standard methods 51 .
Gene expression studies. cDNA was generated by reverse transcription of 500 ng of total RNA using the Superscript VILO cDNA synthesis kit (Invitrogen, 11754-050) according to manufacturers' instructions. qPCR (Applied Biosystems 7900HT) was performed in triplicate on cDNA using SYBR Green PCR Master Mix (Invitrogen, 4309155). For human cells samples were normalized using the average of three reference genes, GAPDH, b-tubulin and actin; and for zebrafish samples were normalized to actin, EF1a and Rpl13a. Primers are shown on Supplementary Table 4.
Measurement of EXOSC8 mRNA copy numbers. Patient and three control fibroblast lines were grown in DMEM supplemented with 15% FBS and 1% Penicillin/Streptomycin to semi-confluence. To half of the cultures 300 mM puromycin was added 6 h prior to harvesting the cells. mRNA was prepared with the Trizole method and reverse transcribed into cDNA with random hexamers using the Superscript II kit (Life Technologies). For qPCR detection we used SYBR Green (Life Technologies) and a primer pair that generated a 198-bp product that comprised all known splice isoforms of EXOSC8: forward, 5 0 -GCATCCGTGTCGA AAGTGC-3 0 ; reverse, 5 0 -CTGGGTTCAAAACCGTGGAACC-3 0 . The qPCR was run on an ABI 7700 machine using the -DDC t method with HPRT: forward, 5 0 -ACAATGCAGACTTTGCTTTCC-3 0 ; reverse, 5 0 -TCAAGGGCATATCCTACA ACAA-3 0 ) as the reference gene.
Zebrafish strains and maintenance. Zebrafish strains used in this study were the golden strain and the islet-1-GFP transgenic line (Tg(islet-1:GFP), expressing GFP driven by the islet-1 promoter) 23 . All zebrafish procedures were performed under Home Office UK licence regulations. Zebrafish embryos were collected and raised at 28.5°C according to standard procedures 52 and staged in hpf or dpf according to standard criteria 53 PTU (0.003%; Sigma) was used to suppress pigmentation when necessary. The p53 morpholino was purchased from Gene Tools (Pilomath, OR).
Antisense MO injections. The antisense MO to downregulate exosc8, mbpa and mbpb expression in zebrafish was purchased from Gene Tools. The MO was designed using the mRNA of the zebrafish exosc8 orthologue; (Gene ID: 323016, mRNA accession number: NM_001002865) and the corresponding genomic sequence on zebrafish chromosome 10. We designed a splice-blocking MO directed against the splice donor site of exon 2: 5 0 -AGATTAACTCTCACCAGAAAGCT CC-3 0 and a second translation-blocking MO 5 0 -TTTAAAACCAGCCGCCATG ATGTTT-3 0 . Embryos were injected with 10-12 ng exosc8 MO. We designed a translation-blocking MO directed against mbpa (5 0 -GGCCATTCTAGGTGTTGA TCTGTTC-3 0 ). Embryos were simultaneously injected with 1 ng MO, with the NATURE COMMUNICATIONS | DOI: 10.1038/ncomms5287 ARTICLE addition of 5-ng p53 MO. We also simultaneously injected mbpa (1 ng) and exosc8 (10-12 ng) MOs with 5 ng p53 in separate experiments with equal MO doses, on the same clutch of embryos.
Three independent MO injection experiments were performed and over 300 injected embryos were evaluated in total to establish morphant phenotypes. MO injection into embryos of the Tg(islet-1:GFP) strain were used to study brain abnormalities, whereas embryos of the golden strain were used for whole-mount antibody staining procedures. Zebrafish were anesthetized with Tricaine solution and phenotyped at 48 hpf to assess brain and tail morphology. Images were captured using a MZ16F fluorescent stereomicroscope (Leica).
RNA isolation and RT-PCR in zebrafish. RNA from zebrafish embryos and larvae was isolated with Trizol reagent (Invitrogen) following the manufacturer's instructions. For RT-PCR analysis in developing zebrafish embryos and following MO injection, RNA from B30 embryos was extracted and after cDNA synthesis from three different MO injections. Primers used for zebrafish RT-PCR are shown on Supplementary Table 5.
Whole-mount antibody immunofluorescence. For whole-mount immunofluorescence staining, zebrafish embryos and larvae were fixed in 4% paraformaldehyde in phosphate-buffered saline overnight at 4°C and then permeabilized in cold acetone at À 20°C. In addition, 4-dpf-old larvae were permeabilized with collagenase A (Roche Diagnostics, 1 mg ml À 1 ) for 90 min. Embryos/larvae were blocked in 5% horse serum in phosphate-buffered saline containing 0.1% Tween-20 (PBT). Embryos/larvae were incubated in blocking solution containing primary antibody overnight at 4°C followed by washing several times with PBT and incubation with secondary antibody (anti-rabbit Alexa Fluor 594 and anti-mouse Alexa Fluor 488, Invitrogen). The primary antibodies used in this study were a polyclonal rabbit anti-MBP antibody (1:50 dilution, a kind gift from Prof. William Talbot, Stanford University) and a mouse monoclonal antiacetylated tubulin antibody (1:500 dilution, Sigma) to label axon tracts. Combinational staining of myelin, neurons and nuclei in zebrafish embryos was performed using BrainStain Imaging Kit (Molecular Probes, B34650). Stained embryos/larvae were imaged with a Zeiss Axio Imager microscope. Immunofluorescence images were collected using a Zeiss Axio Imager Z1 fluorescence microscope equipped with Zeiss Apotome 2 (Zeiss, Germany) in AxioVision Rel 4.9 software.
Electron microscopy. Zebrafish at 4 dpf were fixed in 2% glutaraldehyde in sodium cacodylate buffer at 4°C overnight and suddenly washed three times (15 min each) in cacodylate buffer, and then stained with 1% osmium tetroxide (Agar Scientific) in dH2O for 1 h. Fish were dehydrated using graded acetone (25, 50 and 75% and twice in 100%). Fish were impregnated through increasing concentration of resin in acetone (25, 50, 75 and 100%) and then embedded in 100% resin at 60°C for 24 h (TAAB Lab. Equip). Ultra-thin transverse sections of B70 nm were cut using a diamond knife on a Reichert Ultracut E ultramicrotome. The sections were stretched with chloroform to eliminate compression and mounted on pioloform-filmed copper grids. The grids were then stained with 2% aqueous uranyl acetate lead citrate and subsequently examined using a Philips CM 100 Compustage (FEI) Transmission Electron Microscope and digital images were collected using an AMT CCD camera (Deben) at the Electron Microscopy Research Services, Newcastle University.
Statistical analysis. For the statistical analysis of qRT-PCR we used unpaired t-test.