Results

Clinical Presentation
Individual 1:II-1 is a 28-month-old female of El Salvadorian descent (Figure 1; Video S1). Pregnancy was notable for reduced fetal movements. She was born at term by spontaneous vaginal delivery. Birth weight was 2.98 kg (10th–25th percentile), and length was 48 cm (25th percentile). There were no concerns regarding feeding or breathing in the neonatal period, and movements of the extremities were reported as normal. As an infant, she was noted to have poor head control and a soft cry. She was found to have congenital esotropia, which was managed with botulinum toxin A injection. By 8 months of age, she had increasing difficulty with lifting her arms and legs against gravity. Her strength declined further, which was exacerbated by periods of illness. By 9 months of age, she had lost the ability to vocalize. She rolled from back to side briefly but subsequently lost this skill. She had normal neonatal growth; however, by 21 months of age, her weight, height, and head circumference were all below the third percentile. She had a history of recurrent pulmonary infections requiring hospitalizations, and nighttime non-invasive ventilation in the form of bilevel positive airway pressure (BiPAP) was initiated at 20 months of age. Examination at the age of 28 months revealed a high-arched palate and tongue fasciculations. Extraocular movements were full but were notable for broken pursuits and gaze-evoked nystagmus in horizontal and vertical directions. Although there was a tendency for downward deviation of the eyes, upward gaze was also observed. Occasional oromotor dyskinesia was noted on examination. She had severe axial and appendicular hypotonia and profound weakness, causing her to require full head support. She had no antigravity movements proximally in the upper and lower extremities. Wrist extension was antigravity at 3/5 (Medical Research Council-grade), whereas elbow flexion, knee extension, and finger extension were 2/5. Facial strength was relatively spared. Sensation appeared to be normal. Reflexes were absent throughout. Plantar response was flexor bilaterally. There were multiple joint contractures (Figure 1A). Hands were fisted with thumbs adducted, and feet were maintained in a cavo-varus position (Video S1).
Figure 1 Clinical Presentation and Pedigrees
(A) Clinical photos of affected individual 1:II-1. She had multiple joint contractures. Hands were fisted with thumbs adducted, and feet were maintained in an extended cavo-varus position.
(B) Segregation of the EXOSC9 sequence variants in the four pedigrees.
(C) Electropherograms of the EXOSC9 sequence variants identified in this study.
Electromyography (EMG) and nerve conduction responses were consistent with an axonal motor neuronopathy. Muscle ultrasound revealed a pronounced streak-like and mixed increase in echogenicity with widespread evidence of active fasciculations, consistent with advanced neurogenic changes in the muscle in a non-length-dependent manner. Brain MRI performed at the age of 7 months revealed mild cerebellar atrophy with a normal-appearing pons and no significant abnormalities of the cerebral white matter or the basal ganglia (Figure 2A). A repeat brain MRI at 21 months of age revealed a mild progression of the cerebellar atrophy (Figure 2A). Muscle biopsy performed at 15 months of age demonstrated abundant, very small fibers, often in groups and intermixed with hypertrophic fibers, consistent with denervation and thus suggestive of a neurogenic process. Lysosomal enzymes, plasma amino acids, carbohydrate transferrin, coenzyme Q10, and plasma lactate levels were normal. No seizures have been observed, and a routine electroencephalogram (EEG) performed at 5 months of age did not reveal any epileptiform activity. Family history was non-contributory, and the family denied known consanguinity. Extensive genetic testing for congenital muscle diseases, neuronal ceroid-lipofuscinoses, and SMN1 were negative.
Figure 2 Radiographic Studies
(A) Axial and sagittal T1-weighted images of the brain from affected individual 1:II-1 at 7 months (top) and 21 months (bottom) of age show moderate cerebellar predominant volume loss, which appears mildly progressive between the two exams, as well as mild cerebral atrophy with resultant enlargement of the ventricles. There is no brainstem atrophy.
(B) Sagittal T1- and coronal T2-weighted images of the brain from affected individual 2:II-1 at 6 days of age show severe cerebellar and moderate cerebral and brainstem atrophy. A radiograph of the right leg also shows a mid-femoral fracture.
(C) Two axial CT images of the brain from affected individual 3:II-1 at 9 months of age show mild prominence of the sulci and ventricles reflecting mild atrophy both above and below the tentorium.
(D) Sagittal T1-weighted images of the brain from affected individual 4:II-1 at 6 months and 12 months of age show slightly progressive moderate cerebellar and cerebral atrophy. The brainstem is relatively spared.
Individual 2:II-1 was the first child of non-consanguineous, healthy parents of African-Canadian and Jamaican descent (Figure 1B). During pregnancy, reduced fetal movements, growth retardation, and oligohydramnios were noted. He was born at term with several congenital fractures, including fractures of the bilateral femurs and one fracture of the humerus (Figure 2B). Examination findings were consistent with arthrogryposis multiplex congenita with hip and hand involvement, severe hypotonia, and respiratory insufficiency. Dysmorphic features, including hypertelorism, prominent epicanthic folds, low-set ears, prominent lips, and a short neck, were noted. Sclerae were described as blueish, and skin appeared redundant. As a result of swallowing difficulties, he required gastric-tube feeding. Brain MRI performed at 1 week of age showed cerebellar atrophy, generalized cerebral atrophy, and possibly delayed myelination (Figure 2B). Muscle biopsy findings, including fiber-type grouping, grouped fiber atrophy, and type 1 fiber hypertrophy, were suggestive of a motor neuronopathy. EMG was not performed. Genetic testing for deletions in SMN1 and an SMA-related gene panel were negative. Chromosomal microarray analysis, methylation analysis for Prader-Willi syndrome, myotonic dystrophy testing, and metabolic screening were also negative. Individual 2 died of respiratory failure at the age of 15 months.
Individual 3:II-1 is a 4.5-year-old female born to consanguineous parents of Saudi-Arabian descent. She was noted at birth to have hypotonia, a poor cry, and difficulties with feeding. Subsequently, she was diagnosed with microcephaly and failure to thrive, and she was noted to have severe developmental delay. At the age of 5 months she was noted to have seizures, which were partially controlled by anti-epileptic drugs. Clinical examination at the age of 18 months showed minor dysmorphic facial features. She was unable to visually track objects. She was not vocalizing and appeared to have receptive language difficulties. There was axial hypotonia, generalized weakness of the upper and lower limbs, and an apparent increased tone of the extremities with evidence of exaggerated deep tendon reflexes. She underwent brain imaging (computed tomography [CT]), which revealed minimal cortical atrophy (Figure 2C). Family history is significant for an older sister who died at the age of 8 years and who reportedly manifested severe spasticity and epilepsy. DNA from the sister was unavailable for diagnostic testing. Routine biochemistry, renal profile, total creatine phosphokinase, ammonia, plasma amino acids, very-long-chain fatty acids, and urine for organic acids were normal. Chromosomal microarray was normal. EMG and muscle biopsy were not performed.
Individual 4:II-1 is a 19-month-old female born to non-consanguineous parents of African, European and Filipino ancestry. Pregnancy was uncomplicated, and delivery was induced at 40+ weeks. She was born via Caesarean section with vacuum assistance because of fetal distress. Birth weight was 2.81 kg. She was found to have mild jaundice and congenital nystagmus. At 2 weeks of age she was noted to have poor head control and a weak cry. She had excessive oral secretions and difficulty with clearing her airway. Over time, her head control slowly improved. She started to visually track and follow at the age of 6–7 months. Her weight gain had stagnated between 6 and 12 months, but by 19 months her weight was above the 50th percentile. Her length had been between the 25th and 75th percentiles and her head circumference between the 10th and 25th percentiles. On examination at the age of 12 months she was found to have a moderately high-arched palate and a weak cry. Extraocular movements were full, but she had a recurrent intermittent horizontal nystagmus. She had diffuse hypotonia and weakness with poor truncal control, requiring head support. She was unable to sit and reach for objects. Her muscles were atrophic, and she had minimal antigravity movements in all extremities. Sensation appeared to be normal. Reflexes were 2+ in upper extremities, 1+ at the patellae, and absent at ankles. Plantar response was flexor.
EMG was suggestive of a motor neuropathy or motor neuronopathy. Brain MRI at the age of 6 months showed cerebellar atrophy but a normal-appearing pons. Repeat brain MRI at 12 months revealed a slight progression of the cerebellar atrophy (Figure 2D). Muscle biopsy was suggestive of neurogenic atrophy with fiber-type grouping. Sural nerve biopsy was normal. EEG studies performed on multiple occasions were mostly read as normal, though one earlier study was interpreted as showing myoclonic discharges. Creatine kinase was borderline at 147 U/L (reference range < 143 U/L). Cerebrospinal fluid studies for neurotransmitters were normal. Extensive genetic testing, including a chromosomal microarray, Prader-Willi methylation testing, SMN1 deletion testing, IGHMBP2 sequencing, and a neuromuscular disorder gene panel, was negative.
The clinical presentation seen in these four unrelated affected individuals of early-onset, rapidly progressive weakness and respiratory impairment combined with the presence of cerebellar atrophy and a motor neuronopathy suggests a condition along the PCH1-related disease spectrum.

Identification of EXOSC9 Variants
SNP array testing in individual 1:II-1 revealed one isolated long contiguous stretch of homozygosity of approximately 14.0 Mb on chromosome 4 (117,649,360–131,644,865). Exome sequencing identified a rare homozygous variant, c.41T>C (p.Leu14Pro), in EXOSC9 (MIM: 606180; GenBank: NG_029848.1) within the region of homozygosity. This mutation is predicted to cause a disruption in the first alpha helix of EXOSC9.41 The parents of individual 1:II-1 were found to be heterozygous for the variant, consistent with autosomal-recessive inheritance (Figure 1B). The variant is a rare SNP (dbSNP: rs139632595) and has been reported six times in heterozygous state in the ExAC Browser in individuals from African descent with an allele frequency of 4.947 × 10−5. This variant was neither reported in individuals from Hispanic descent nor found in a homozygous state.
WES of individual 2:II-1 revealed the same c.41T>C variant as in individual 1:II-1 in compound heterozygosity with a EXOSC9 c.481C>T variant that leads to a premature stop of the protein (p.Arg161∗). The c.481C>T variant was listed three times in the ExAC Browser in the heterozygous state. To exclude variants in genes that are known to be associated with either SMA or with muscle diseases, we specifically screened the variant VCF files for variants therein but found none. Sanger sequencing confirmed compound heterozygosity (Figure 1).
Individuals 3:II-1 and 4:II-1 were subsequently identified to carry the same EXOSC9 c.41T>C in homozygosity.

Haplotype Analysis
The shared haplotype analysis revealed a common haplotype of 800 kb at approximately chr4: 122,400,000–123,200,000, encompassing ANXA5, TMEM155, EXOSC9, BBS7, TRPC3, and KIAA1109. The largest homozygous block was identified in individual 1:II-1 and extended approximately 11 Mb, whereas the other two homozygotes, individuals 3:II-1 and 4:II-1, showed ROHs of only approximately 1 Mb. Our common haplotype was estimated at 800 kb. Using the approximation given by Ying et al.,42 we can estimate a common ancestor approximately 125 generations ago. The results of ancestry PCA are consistent with self-reported ethnicities. Although these self-reported ethnicities are superficially divergent, PCA shows that these individuals do potentially share some common ancestry, most likely northern or eastern African. Additionally, compared with individuals in the primary super-population clusters, all four individuals are relative outliers in the PCA space (Figure S1).

The Exosome Complex Is Reduced in Skeletal Muscle and Fibroblasts
Immunoblotting for components of the exosome complex was performed on cultured fibroblasts of individual 1:II-1 and skeletal muscle of individual 2:II-1. EXOSC9 was less abundant in both affected fibroblasts and skeletal muscle than in controls. Additionally, in fibroblasts from individuals harboring mutations in EXOSC3, EXOSC8, and RBM7, EXOSC9 was reduced; the reduction was most pronounced in cells from the EXOSC3 and EXOSC8 cell lines (Figure 3A). These data suggest that a primary reduction in one component of the exosome complex, or in a related protein such as RBM7, leads to destabilization of the whole complex. BN-PAGE with protein lysates from affected individual’s fibroblasts probed with an antibody for EXOSC3 supported this hypothesis. Variants in EXOSC3, EXOSC8, EXOSC9, and RBM7 resulted in reduction of the exosome complex irrespective of which subunit carried the primary variant (Figure 3B).
Figure 3 The Exosome Complex Is Reduced in Affected Individuals’ Fibroblasts
(A) Immunoblotting of fibroblasts from affected individuals with variants in different components of the exosome complex (the homozygous c.92G>C in EXOSC3, the homozygous c.5C>T in EXOSC8, the homozygous c.41T>C in EXOSC9 [individual 1:II-1], and the homozygous c.236C>G in RBM7) shows reduced EXOSC9, but other components of the exosome complex were also reduced. Actin was used as a loading control.
(B) Blue native polyacrylamide gel electrophoresis (BN-PAGE) shows that there is a reduction of the assembly of the whole exosome complex in affected individuals’ fibroblasts. GAPDH was used as a loading control.
(C) Immunoblot on muscle extracts from affected individual 2:II-1 and four controls confirms that EXOSC9 was severely reduced in affected individual 2:II-1.

Variants in EXOSC9 Result in Abnormal RNA Metabolism
To study the effect of EXOSC9 variants on gene expression, we performed RNA-seq on RNA collected from cultured affected individuals’ fibroblasts (individual 1:II-1) and skeletal muscle (individuals 1:II-1 and 2:II-1). In the fibroblasts of individual 1:II-1, 69 genes (22 containing AREs) showed a significant increase and 138 genes (35 containing AREs) showed significantly less RNA expression than control individuals (Figure 4A; Table S4). Many of the enriched Gene Ontology (GO) terms from both the significantly increased and decreased genes described cellular and embryonic developmental processes of the neuronal system (Figure 4B). EXOSC9 mRNA and mRNAs encoding other subunits of the exosome complex did not show a significant difference in expression. Previously, we showed that expression of HOTAIR, HOXC6, HOXC8, and HOXC9 was significantly elevated in fibroblasts from individuals with variants in EXOSC8 and RBM7. However, in fibroblasts from individuals with variants in EXOSC9 and EXOSC3, only HOXC8 expression was higher than in control fibroblasts; interestingly, increased HOTAIR mRNA seems to be specific to the EXOSC8 and RBM7 mutant cells (Figure 4D).
Figure 4 RNA-Seq and qPCR in Fibroblasts of Affected Individual 1:II-1 Carrying the Homozygous c.41T>C Sequence Variant in EXOSC9
(A) Differential expression analysis (DESeq2) detected 69 genes that were significantly upregulated and 138 that were significantly downregulated in the affected individuals’ fibroblasts.
(B) Gene-set enrichment analysis of GO terms with differentially expressed genes in fibroblast RNA-seq.
(C) Comparison of the numbers of upregulated and downregulated genes in muscle biopsy specimens of RNA-seq between affected individual 1:II-1 and affected individual 2:II-1. 2,778 transcripts were upregulated in both affected individuals, whereas 2,864 transcripts were downregulated in both. 18 transcripts that were upregulated in affected individual 2:II-1 were downregulated in affected individual 1:II-1, and 45 transcripts were upregulated in affected individual 1:II-1 but downregulated in affected individual 2:II-1.
(D) Gene expression analysis of HOXC6, HOXC8, HOXC9, and HOTAIR through qRT-PCR in EXOSC3, EXOSC8, EXOSC9, and RBM7 mutant fibroblasts. Results were normalized to the average values obtained from two control fibroblast lines.
RNA-seq revealed that a high number of genes were significantly differentially expressed between muscle biopsies of individual 2:II-1 and control fibroblasts and fibroblasts of individual 1:II-1: 2,778 genes (497 containing AREs) were significantly upregulated and 2,864 (448 containing AREs) were downregulated in both individuals (Figure 4C; Table S4). Expression of several genes linked to motor neuronopathy and familiar amyotrophic lateral sclerosis (EPHA4, IGHMBP2, VAPB, BICD2, and DYNC1H1) or distal arthrogryposis (MYH8, ZC4H2, MUSK, RAPSN) were significantly upregulated or downregulated. In individual 2:II-1, who presented with congenital fractures, we detected a higher number of significant changes in gene expression than in individual 1:II-1. This included seven genes (LIFR, TMEM38B, PLS3, NANS, SLC26A2, ALX4, and PLS3) associated with skeletal dysplasia or bone disease, and four of them were ARE-containing genes with increased expression. A comparison of RNA expression between the fibroblasts and muscle of individual 2:II-1 showed that only 17 genes were significantly upregulated and that 16 genes were significantly downregulated in both samples, suggesting tissue-specific differences.

Knockdown or Variants in exosc9 Cause Developmental Defects in Zebrafish
Zebrafish have previously been used as model systems for investigating variants in exosome complex subunits16, 17 and associated proteins19 and are consistently used for modeling the cerebellar, hindbrain, and motor neuron dysfunction observed in human disease. We concluded that zebrafish would therefore make a suitable in vivo disease model for the effects of reduced exosc9 function for investigating whether a phenotype consistent with the other exosomal models would result.
Injection of morpholino oligonucleotides and the CRISPR/Cas9 system were used to knock down or induce variants, respectively, in exosc9 in zebrafish embryos (Figure 5A). The exosc9 morpholino oligonucleotides led to aberrant splicing of the exosc9 transcript, which was confirmed via RT-PCR (Figure 5B), where morphant zebrafish had a retained intron that was confirmed by sequencing. In addition to the appearance of mis-spliced transcripts, the amount of wild-type (WT) exosc9 transcript was reduced in injected embryos. The embryos injected with Cas9 and gRNA for exosc9 would be expected to be mosaic; genomic DNA of cells would be a mixture of WT and various mutated forms of exosc9 in varying proportions. To confirm mutagenesis in the crispants, PCR with primers flanking the sgRNA target area was performed on genomic DNA. The PCR product was then cloned into the pGEM-T easy vector and colony PCR, and sequencing was performed on individual clones. Sequencing showed that there was a variation in the amount of mutagenesis occurring and that there was a phenotype-genotype correlation (Figure 5D).
Figure 5 Strategies Targeting exosc9 in Zebrafish
(A) Schematic of exosc9 in zebrafish demonstrating the sites to where the morpholino, gRNA, and primers were targeted.
(B) RT-PCR of zebrafish morphants. The morpholino caused the retention of an intron and a reduction of the WT product in a dose-dependent fashion. The identity of the bands was confirmed by Sanger sequencing.
(C) The target sequence for exosc9 gRNA and an example of a mutation found in a crispant.
(D) The mutation rate found in crispants of differing phenotypes.
Zebrafish injected with the morpholino oligonucleotide (morphants) and Cas9 and exosc9 sgRNA (crispants) developed a similar range of morphological phenotypes (Figure 6A). Mildly affected embryos had smaller heads and eyes, whereas severely affected embryos had extremely small, sometimes absent, eyes, very small heads, and truncated bodies (Figure 6A). The relative distribution of phenotypes was also similar in morphants and crispants (Figure 6B).
Figure 6 Knockdown of exosc9 in Zebrafish Causes Abnormal Morphology
(A) Representative morphological scoring of morphant and crispant exosc9 zebrafish embryos at 48 hpf. Normal, identical to uninjected control clutchmates; mild, smaller head and smaller eyes; severe, very small head, smaller or absent eyes, and misshapen body.
(B) Relative distribution of morphological phenotypes in exosc9 morphants and crispants at 48 hpf. Scale bar represents 1 mm.

In Zebrafish, exosc9 Is Required for Brain and Neuromuscular Development
Next, we investigated whether exosc9 was required for brain and neuromuscular development. Whole-mount immunofluorescence performed on 48 hpf exosc9 morphants and crispants with an antibody against the neuronal marker, HuC, showed that the brain fails to properly develop (Figure 7A). In exosc9 morphants and crispants, it was common for the brain to be misshapen and for the cerebellum and hindbrain to be absent. Morpholino knockdown and CRISPR/Cas9 mutagenesis of exosc9 was also performed in the transgenic line of zebrafish, Islet1:GFP. This line of zebrafish produce green fluorescent protein in the motor neurons in the hindbrain.43 In WT zebrafish, the cranial nerves are very distinct and can be visualized in the Islet1:GFP zebrafish. Cranial nerve V is split into two distinct parts, anterior (Va) and posterior (Vp).43 In the exosc9 morphants and crispants, it was common for only Va to be present. This again suggests that functional exosc9 is required for the posterior part of the brain to develop in zebrafish (Figure 7B). Whole-mount immunofluorescence using an antibody against synaptic vesicle 2 (SV-2, presynaptic motor axons) and α-bungarotoxin (neuromuscular junctions) conjugated to Alexa Fluor 594 showed that the neuromuscular junctions developed abnormally in the 48 hpf morphants and crispants (Figure 7C). In both morphants and crispants, the neuromuscular junctions appeared closer together and the motor axons failed to migrate properly to the neuromuscular junctions, indicating a primary pathfinding defect of the motor axons. Pathfinding defects in motor axons have been illustrated in other zebrafish models of neuronopathies—exosc3, exosc8, RBM7,19 and SMA.44 Phalloidin staining in 48 hpf morphants and crispants showed a reduced amount of muscle and damaged and misaligned myofibres (Figure 7D). Together, these results show that exosc9 is also important in neuromuscular development.
Figure 7 Knockdown of exosc9 in Zebrafish Causes Abnormal Neuromuscular Development
(A) Whole-mount immunofluorescence of the pan-neuronal marker HuC shows that the midbrain (∗) appears abnormal and the cerebellum (#) and hindbrain (+) are absent in representative exosc9 morphants and crispants compared with uninjected controls.
(B) Islet1:GFP transgenic morphant and crispant zebrafish have absent cranial posterior nerve V (Vp).
(C) Whole-mount immunofluorescence of synaptic vesicle 2 (SV2, motorneurons, green) and α-bungarotoxin (neuromuscular junctions, red) shows that motoneurons and neuromuscular junctions failed to properly develop in exosc9 morphants and crispants compared with uninjected controls (white arrows).
(D) Phalloidin staining shows that muscle failed to develop properly. Fibers were sparser, more spread out, and irregular in the exosc9 morphants and crispants. All experiments were performed in 48 hpf zebrafish.