ATAD3 gene cluster deletions cause cerebellar dysfunction associated with altered mitochondrial DNA and cholesterol metabolism

Abstract

Although mitochondrial disorders are clinically heterogeneous, they frequently involve the central nervous system and are among the most common neurogenetic disorders. Identifying the causal genes has benefited enormously from advances in high-throughput sequencing technologies; however, once the defect is known, researchers face the challenge of deciphering the underlying disease mechanism. Here we characterize large biallelic deletions in the region encoding the ATAD3C, ATAD3B and ATAD3A genes. Although high homology complicates genomic analysis of the ATAD3 defects, they can be identified by targeted analysis of standard single nucleotide polymorphism array and whole exome sequencing data. We report deletions that generate chimeric ATAD3B/ATAD3A fusion genes in individuals from four unrelated families with fatal congenital pontocerebellar hypoplasia, whereas a case with genomic rearrangements affecting the ATAD3C/ATAD3B genes on one allele and ATAD3B/ATAD3A genes on the other displays later-onset encephalopathy with cerebellar atrophy, ataxia and dystonia. Fibroblasts from affected individuals display mitochondrial DNA abnormalities, associated with multiple indicators of altered cholesterol metabolism. Moreover, drug-induced perturbations of cholesterol homeostasis cause mitochondrial DNA disorganization in control cells, while mitochondrial DNA aggregation in the genetic cholesterol trafficking disorder Niemann-Pick type C disease further corroborates the interdependence of mitochondrial DNA organization and cholesterol. These data demonstrate the integration of mitochondria in cellular cholesterol homeostasis, in which ATAD3 plays a critical role. The dual problem of perturbed cholesterol metabolism and mitochondrial dysfunction could be widespread in neurological and neurodegenerative diseases.

Keywords: ATAD3; cerebellar hypoplasia; cholesterol; mitochondrial DNA; mitochondrial disease.

Figure 1

Pedigrees and brain MRI from five unrelated families with cerebellar disorders. (A) Pedigrees and ATAD3 genotypes for available members of Families 1–5. (B) Brain MRI of Subjects S1a, S3, S4 and S5. Top row: Sagittal images of Subjects S1a, S3 and S4 in the neonatal period and Subject S5 at 22 years of age. The neonates have severe brainstem and cerebellar hypoplasia with flat pons (short arrows) and tiny cerebellar vermis (long arrows). There is increase of the tegmento-vermian angle and ex vacuo enlargement of the posterior fossa CSF spaces. Arrowheads indicate the thin corpus callosum. Stars in Subject S3 show isointense blood products within and below the fourth ventricle. Subject S5 presents with severe hypoplasia/atrophy of the cerebellar vermis (thick arrow) with ex vacuo enlargement of the fourth ventricle; brainstem and normal corpus callosum are normal. Bottom row: Axial T₂-weighted images show simplified sulcation and gyration more marked frontally (short arrow) and diffuse white matter T₂ signal abnormality (long arrow) in Subjects S1a and S4. Similar but less severe changes are seen in Subject S3 with shallow simplified sulcation. Both subjects had a thin cortical ribbon, decreased white matter volumes with marked T₂ hyperintensity, ex vacuo ventriculomegaly (stars) and prominence of the extra-axial CSF spaces in keeping with brain atrophy. Hypointense material within the lateral ventricles of Subject S3 is haemorrhage. Subject S5 has normal ventricles and subtle ‘frosted glass’ aspect of the posterior periventricular white matter (thick arrow). del = ATAD3 deletion; WT = wild-type.

Figure 2

Identification of genomic ATAD3 deletions. (A) A custom CGH array was used to delineate homozygous deletions detected on chromosome 1 p in DNA from Subjects S1a and S2. Shaded boxes indicate the location of the ATAD3C, ATAD3B and ATAD3A genes. Details of deleted regions predicted by SNP and CGH arrays are summarized in Supplementary Table 2. (B) Long-range PCRs were performed on genomic DNA from subjects and controls. Primers OT472 and OT473 (middle panel) flank the ATAD3 deletion breakpoints predicted in Subject S1a by array CGH. Primers OT572 and OT575 (bottom panel) flank the ATAD3 deletion breakpoints predicted in S2 by array CGH. As a control, primers OT570 and OT575 (top panel) were used since primer OT570 is located within the predicted deleted ATAD3B/ATAD3A region. (C) Genomic DNA sequencing of the breakpoint-spanning PCR products determined the ATAD3B/ATAD3A deletion boundaries in each subject, with chromosome 1 coordinates indicated (hg19). Ambiguous regions flanking the deletion boundaries that have identical sequence in ATAD3B and ATAD3A are identified by dark grey boxes. The deletion predicted by high-density SNP array for Subject S5 is indicated, with maximum and minimum deletion boundaries labelled by hatched boxes.

Figure 3

ATAD3 gene cluster deletions result in decreased expression of ATAD3A, and the loss of ATAD3B at the mRNA and protein level. (A) ATAD3 deletions in relation to exon structure of full length ATAD3B and ATAD3A isoforms for Subjects S1a, S1b, S2 and S3. (B) Full length ATAD3A and ATAD3B isoforms were amplified from control and subject cDNA prepared from fibroblasts grown ± cycloheximide. Primers OT441 and OT443 were used to amplify ATAD3A, while primers OT441 and OT445 were used to amplify ATAD3B. The ATAD3A product amplified in S1a, S3 and S4 represents the fusion cDNA due to cross-hybridization of primer OT441. (C) The relative expression levels of ATAD3A versus ATAD3B were determined by qRT-PCR from four controls and Subjects S1a, S3, S4 and S5. Results were normalized to HPRT expression and presented as percent of average control ATAD3A expression; n = 3, error = SEM. Two-way ANOVA comparing all controls and subjects showed ATAD3B expression was significantly reduced in all subjects compared to individual controls (P < 0.0001), as was ATAD3A in S1a, S3 and S4 (P < 0.0001). For Subject S5, the reduction in ATAD3A was not significant in comparison to all controls. (D) Ranked gene expression list showing the positions of ATAD3A and ATAD3B in Subjects S1a and S5 versus control. The dotted line represents the position of genes that are expressed at the same level in subject and control samples; to the left, genes expressed more highly in the subjects; to the right, genes expressed less than the control. Total number of genes 26 689. S5 fibroblasts manifested reduced expression of ATAD3B, with log₂ fold change (FC) = −1.31; and to a lesser extent the intact ATAD3A log₂FC = −0.73. The transcripts of the ATAD3B/ATAD3A chimera of Subject S1a registered as decreased expression of both original genes; ATAD3B, log₂FC−3.0, ATAD3A, log2FC = −1.50. (E) ATAD3 proteins detected by immunoblotting using a pan-specific antibody in tissues and fibroblasts from subjects, relative to controls. Porin or GAPDH was used as a loading control.

Figure 4

ATAD3 deficiency is associated with mtDNA abnormalities. The DNA of human fibroblasts was stained with anti-DNA antibody and the number of cells with enlarged mtDNAs was scored for cell lines C1, S1a and S5. (A) A minimum of 150 cells was counted for each cell line (n = 4 independent experiments, error bars are 1 standard deviation from the mean). See Supplementary Fig. 6 for representative images for Subject S5. (B) The signals of foci in the cytoplasm (mtDNA) plotted as frequency distribution after particle point analysis of a minimum of 20 cells. To smooth the data, signals were sorted into 20 ‘bins’ based on intensity. Foci falling in bins 1–6 all correspond to one mtDNA molecule based on detailed study of the images, including ones subjected to deconvolution analysis (Akman et al., 2016), and the fact that most mtDNAs are organized as single copies (Kukat et al., 2011), with the variation across the range expected to be the result of differences in condensation (packing) and efficiency of antibody access and coating. Many of the foci of bins 7–9 could be resolved to two or three separate or overlapping smaller dots indicating they contained two or three mtDNA molecules, whereas larger foci in bins 10 and above contained more than three copies of mtDNA. Very large mtDNA foci (bins > 11) were significantly more numerous in Subjects S1a (P < 0.0001) and S5 (P < 0.0001) fibroblasts than controls. (C) PicoGreen® staining of mtDNA reveals larger foci in cells of S1a than those of controls. Representative images showing the PicoGreen® stained structures in the cytoplasm (mtDNA). Nuclear DNA staining with PicoGreen® is highly variable (often it is undetectable) (He et al., 2007) and so cannot serve as a reference. Microscope settings and image capture parameters were identical. PicoGreen® staining was as previously described (Ashley et al., 2005). (D) BrdU incorporation into mtDNA detected by immunofluorescence. Single confocal optical sections of C1 and S5 fibroblasts treated with 20 μM BrdU for 8 h and with anti-BrdU antibody (green) and anti-ATAD3 antibody (red).

Figure 5

U186666A, pravastatin and cholesterol increase the size of mitochondrial nucleoids in human fibroblasts. The DNA of control human fibroblasts was stained with anti-DNA antibody after treating cells for 7 days without or with 5 μM U18666A (A), or pravastatin (B); or for 5 days with 5 mM cholesterol (C). P-values for the difference between large mtDNA foci (bins > 8) of treated versus untreated cells were P = 0.0116 (U18666A); P = 0.0011 (pravastatin); P = 0.0002 (cholesterol). The data are plotted as frequency distributions after particle point analysis. Insets are representative images. (D) The proportion of fibroblasts with enlarged mtDNA foci, after treatment with U18666A, pravastatin, or cholesterol, compared to untreated cells, based on a minimum count of 50 cells from each of three independent experiments, error bars are 1 standard deviation from the mean. Fibroblasts of Subjects S1a and S5 carry deletions in the ATAD3 gene cluster. Data for the ‘untreated’ cells, derived from experiments carried out in parallel, are reproduced from Fig. 4A. Probability was determined using a one-way ANOVA test (uncorrected Fisher’s LSD). *P < 0.05; **P < 0.01, ***P < 0.001.

Figure 6

Niemann-Pick type C disease is associated with mtDNA disorganization. (A) Frequency distribution of mtDNA foci size in fibroblasts from a control and two individuals with NPC1 defects. (B) Proportion of cells with mtDNA clusters, error bars are 1 standard deviation from the mean. (C) Representative images. Anti-DNA (green) and anti-TOM20 labelling of fibroblasts. Yellow foci, and green foci bounded by red, are mtDNAs.

Figure 7

Abnormal cholesterol homeostasis and reduced expression of OXPHOS factors associated with ATAD3 gene cluster deletions. (A) Rank gene expression list showing the position of selected factors for Subjects S1a and S5 versus control (see also Supplementary Figs 8–10 and Supplementary material). (B and C) Proteins from fibroblasts of controls and S1a and S5 were fractionated by SDS-PAGE and probed with the indicated antibodies after blot transfer. There was no increase in the steady-state level of the larger isoforms of SREBF2 (located in the cytoplasm, SREBF2-C) among the subject samples but a SCAP processed SREBF2 isoform, which translocates to the nucleus (SREBF2-N) and resolves at ∼55 kDa, was increased in S1a and S5 samples (this particular species may be a phosphorylated form of SREBF2 (Krycer et al., 2012); CES1 was greatly diminished. (D and E) Free cholesterol detected by filipin labelling of fibroblasts exposed to (D) no treatment, or (E) 5 μM U18666A for 72 h. *P < 0.05, ns = not significant. (F) Gene set enrichment plots for OXPHOS in Subjects S1a and S5. Each vertical black line represents an mRNA, with the most positively correlated to the left (in the red zone). Clustering at the negatively correlated end of the spectrum (blue zone) indicates the pathway is repressed compared to the reference. Consistent changes across a gene set give sharp curves (green lines) and S1a gave the sharper curve of the two. Moreover, for S1a OXPHOS was the third most negatively differentially expressed pathway or process, whereas it was 17th for S5. These observations are consistent with greater OXPHOS impairment in S1a than S5. (G) Immunoblots of respiratory chain components of complex IV (COX2) and complex I (NDUFB8). (H) Proposed arrangement of cholesterol in the mitochondrial inner membrane, the mtDNA in particular is not to scale. (a) In normal conditions localized high concentrations of cholesterol impart rigidity to the membrane for optimal organization and segregation of the mtDNA, whereas other regions require greater flexibility to form the highly invaginated membranes characteristic of the inner mitochondrial membrane. (b) If cholesterol is scarce there is insufficient sterol to permit mtDNA segregation. (c) Alternatively, if cholesterol is present in normal amounts but dispersed, both rigidity and flexibility are suboptimal; hence, a key role of ATAD3 may be to concentrate cholesterol where the mtDNA is located. Such membrane abnormalities (b or c) could cause the increase in mitochondrial turnover (mitophagy) reported for other ATAD3 mutants (Harel et al., 2016).