A Novel Mutation in ELOVL4 Leading to Spinocerebellar Ataxia (SCA) With the Hot Cross Bun Sign but Lacking Erythrokeratodermia: A Broadened Spectrum of SCA34

Importance  Although mutations in 26 causative genes have been identified in the spinocerebellar ataxias (SCAs), the causative genes in a substantial number of families with SCA remain unidentified.

Objective  To identify the causative gene of SCA in 2 Japanese families with distinct neurological symptoms and radiological presentations.

Design, Setting, and Participants  Clinical genetic study at a referral center of 11 members from 2 Japanese families, which started in 1997.

Main Outcomes and Measures  Results of neurological examinations and radiological evaluations. The causative mutation was identified using genome-wide linkage analysis and next-generation sequencing.

Results  Affected members (9 of 11 members [81.8%]) showed slowly progressive cerebellar ataxia (all 9 members [100%]), ocular movement disturbance (all 9 members [100%]), and pyramidal tract signs (8 of 9 members [88.9%]) with an age at onset between the second and sixth decades of life. Besides cerebellar and pontine atrophy, magnetic resonance imaging of the brain revealed the hot cross bun sign (4 of 6 members [66.7%]), pontine midline linear hyperintensity (2 of 6 members [33.3%]), or high intensity in the middle cerebellar peduncle (1 of 6 members [16.7%]), which are all reminiscent of multiple system atrophy in tested patients. Using linkage analysis combined with exome and whole-genome sequencing, we identified a novel heterozygous mutation in the ELOVL fatty acid elongase 4 (ELOVL4) gene (c.736T>G, p.W246G) in both families. Haplotype analysis indicated that it was unlikely that these 2 Japanese families shared a common ancestor. Although a missense mutation in ELOVL4 (c.504G>C, p.L168F) was recently reported to be associated with SCA with erythrokeratodermia variabilis (SCA34) in a French-Canadian family, signs of erythrokeratodermia variabilis were absent in our families.

Conclusions and Relevance  Combined with the results of the family with SCA34 reported previously, this report confirms that mutations in ELOVL4 can cause dominantly inherited neurodegeneration severely affecting the cerebellum and brainstem. We should be aware that the presence of multiple system atrophy–like features on magnetic resonance imaging scans, together with cerebellar and brainstem atrophy, suggests SCA34, even when erythrokeratodermia variabilis is absent. The present study further broadened the spectrum of the clinical presentations of SCA34 associated with mutations in ELOVL4, which is involved in the biosynthesis of very long-chain fatty acids.

Spinocerebellar ataxias (SCAs) are autosomal dominant neurodegenerative disorders that show progressive cerebellar ataxia, often associated with various phenotypes of neurological dysfunction, such as ocular movement disturbances, pyramidal tract or extrapyramidal signs, and peripheral neuropathy. Twenty-six causative genes of SCAs have been described to date,¹ some of which have been identified by exome or whole-genome sequencing in combination with linkage analysis.^2-6 However, causative mutations in a significant number of families with SCA still remain to be identified.

Herein we report on 2 Japanese families with SCA whose members present with a slowly progressing gait ataxia, ocular movement disturbances (such as horizontal gaze nystagmus and supranuclear gaze palsy), dysarthria, and pyramidal tract signs. Magnetic resonance imaging (MRI) of affected family members showed not only cerebellar atrophy but also marked pontine atrophy with the hot cross bun sign or pontine midline linear hyperintensity, resembling multiple system atrophy. Using genome-wide linkage analysis and exome and whole- genome sequencing, we identified a novel mutation in the ELOVL fatty acid elongase 4 (ELOVL4) gene (c.736T>G, p.W246G) that segregated with the disease in these 2 families.

We first studied family A (Figure 1), with 7 of 9 members who have slowly progressive ataxia, pyramidal tract signs, and cerebellar and pontine atrophy detected on MRI scans of their brains. We conducted neurological and dermatological examinations for all 9 family members of family A (II-1, II-2, II-3, II-4, III-1, III-2, III-3, III-4, and III-5), extensive ophthalmologic examinations for only 2 affected family members (II-1 and III-1), and electrophysiological examinations for only 1 affected family member (II-1). A neurological examination was also conducted for 2 additional participants from another Japanese family (II-1 and III-1 in family B) with highly similar clinical and radiological features. These 2 families are unrelated and originated from distant regions of Japan.

Our study was approved by the local ethics committee of Tokyo Medical and Dental University, the University of Tokyo, and Yokohama City University. After obtaining written informed consent from the 11 members of the 2 families, blood samples were obtained, and genomic DNA extracted, using standard protocols. The initial screening by genetic testing excluded SCA types 1, 2, 3, 6, and 31 and denatatorubral-pallidoluysian atrophy in these 2 families. Plasma levels of very long-chain fatty acids (VLCFAs; represented by the ratios of C24:0 to C22:0, C25:0 to C22:0, and C26:0 to C22:0 [where C24:0 denotes a saturated fatty acid with C24 carbon chain length]) were measured in 2 affected family members in family A (II-1 and III-1) by use of gas chromatography (SRL Inc).

Genome-wide single-nucleotide polymorphism (SNP) genotyping, using the Genome-Wide Human SNP 6.0 Array (Affymetrix), was performed on the genomic DNA of members II-1, II-2, II-3, II-4, III-1, III-2, and III-3 in family A. Experimental procedures were performed according to the manufacturer’s instructions. Acquired data (.chp files) were further processed by use of the high-throughput linkage analysis system SNP HiTLink,⁷ and parametric linkage analysis was performed by assuming autosomal dominant inheritance using Merlin⁸ (eAppendix in the Supplement).

The Genome-Wide Human SNP 6.0 Array data were used to detect copy number variations with PennCNV (an integrated hidden Markov model designed for detecting high-resolution copy number variations in whole-genome SNP genotyping data) according to the manual’s default settings.⁹ The copy number variations detected were then filtered, with the candidate regions determined by the linkage analysis.

Genomic DNA from 3 affected members in family A (II-1, III-1, and III-2) were further subjected to exome sequencing using exome capture kits (SureSelect Human All Exon kit; Agilent Technologies) and next-generation sequencers (HiSeq2000; Illumina). Whole-genome sequencing was performed on the genomic DNA of 1 affected member in family A (III-1) using HiSeq2000. The acquired reads were mapped by use of the Burrows-Wheeler Alignment tool¹⁰ to human genome GRCh37, and the calling of single-nucleotide variations (SNVs) was performed using SAMtools.¹¹ The SNVs were filtered to exclude known variants in several databases: dbSNP (build 135), 1000 Genomes, National Heart, Lung, and Blood Institute Exome Sequencing Project, and HapMap. The SNVs that were either in segmental duplicated regions or registered in the in-house exome database of the Japanese population (N = 373) at the University of Tokyo were further excluded. Functional predictions (using Polymorphism Phenotyping version 2 [Polyphen-2],¹² the Sorting Intolerant From Tolerant [SIFT] program,¹³ a likelihood ratio test,¹⁴ and MutationTaster¹⁵) were performed to evaluate the effect of each SNV. For validation of the SNVs, genomic DNA were amplified using primers designed with Primer-BLAST.¹⁶ Amplified fragments were analyzed using Applied Biosystems 3130xl Genetic Analyzer (Life Technologies). For ELOVL4 (Refseq NM_022726), all the coding sequences were determined by direct nucleotide sequence analysis. Haplotypes were reconstructed using nearby SNVs identified by whole-genome sequencing of 1 affected member (III-1 in family A) (eFigure in the Supplement).

To filter signal peptide sequences that may cause erroneous annotations as a transmembrane domain, SignalP 4.0¹⁷ was first applied to ELOVL4 (Uniprot Knowledgebase: Q9GZR5) and ELOVL5 (Uniprot Knowledgebase: Q9NYP7) sequences; no signal sequences were identified.¹⁸ Then, the programs MEMSAT-SVM,¹⁹ MEMSAT3,²⁰ ENSEMBLE,²¹ PHOBIUS,²² and TMHMM2²³ were used to predict transmembrane α-helical segments and the orientation of amino and carboxy termini in either the cytoplasmic or noncytoplasmic compartment.

A summary of the clinical characteristics of the patients in the 2 Japanese families (as well as of patients in French-Canadian^3,24 and European² families) are shown in our Table, and detailed neurological and radiological findings of each examined patient are described in eTable 1 in the Supplement. All the 9 affected members in the 2 families (7 in family A and 2 in family B) showed slowly progressive gait ataxia as the cardinal manifestation. The mean age at onset was 33.9 years (range, 13-56 years). All the affected members subsequently developed dysarthria.

Disease progression was very slow, and it was not until the age of 60 years or older that affected members required a cane or walker. On neurological examination, truncal and limb ataxia and ataxic dysarthria were observed in all 9 affected family members. Horizontal gaze nystagmus was observed in 7 affected members (77.8%), and supranuclear ophthalmoplegia, more obvious in the vertical than horizontal direction, was observed in 3 affected members (33.3%) (II-1, II-3, and II-4 in family A). Pyramidal tract signs, such as elevated deep tendon reflexes in the limbs, or positive Babinski signs were observed in 8 affected members (88.9%). Autonomic symptoms, such as bladder disturbance (44.4%) and constipation (22.2%), were also observed, but none of the affected members had obvious orthostatic hypotension. Notably, we could not find any present evidence or history of skin lesions characteristic of erythrokeratodermia variabilis (EKV) in any of the 9 affected members. The MRI scans of the brains of the 6 affected members who could be investigated (II-1, II-4, III-1, III-4, and III-5 in family A and III-1 in family B) all showed cerebellar and marked pontine atrophy (Figure 2A-H). It is noteworthy that axial T2-weighted images showed either cruciform hyperintensity (which is referred to as the hot cross bun sign) in 4 of these 6 affected members (66.7%) (II-1, III-1, and III-5 in family A and III-1 in family B) or pontine midline linear hyperintensity in the other 2 affected members (33.3%) (II-4 and III-4 in family A) (Figure 2I-M), both of which often appear in patients with multiple system atrophy.^25-27

Furthermore, high intensity in the middle cerebellar peduncle, another radiological feature of multiple system atrophy, was observed in 1 of the 6 affected family members (16.7%) (II-4 in family A) on a fluid-attenuated inversion recovery MRI scan (Figure 2N). Extensive ophthalmologic evaluations, which included a visual acuity test, a color sensation test, and fundoscopy, revealed no abnormalities in 2 affected members (II-1 and III-1 in family A). The results of the Goldmann perimeter test, which was only performed for member III-1 in family A, were normal. A nerve conduction study performed for 1 affected member (II-1 in family A) showed normal conduction velocity and amplitude of compound motor action potentials in the median and tibial nerves and of sensory nerve action potentials in the median and sural nerves. The motor evoked potentials and somatosensory evoked potentials were normal for the upper and lower extremities.

A genome-wide linkage analysis of 7 members (II-1, II-2, II-3, II-4, III-1, III-2, and III-3 in family A) revealed multiple chromosomal regions of possible genetic linkage with logarithm of odds larger than 0. Among these regions, chromosomes 1, 2, 6, and 9 harbored regions showing logarithm of odds scores higher than 1.45, with total lengths of 33.4, 34.7, 58.7, and 18.0 megabases, respectively. We next performed exome sequencing of 3 affected members (II-1, III-1, and III-2 in family A) and whole-genome sequencing of 1 member (III-1 in family A) (eTable 2 in the Supplement), to detect candidate mutations within the above-mentioned candidate regions determined by the linkage analysis.

Data from whole-genome sequencing were used to detect variants in the coding sequences and rare variants outside the coding sequences for haplotyping. After selecting variants within the candidate regions, we excluded variants already recorded as SNPs or those within regions of segmental duplication. The selection of variants present in at least 2 affected members resulted in 4 novel nonsynonymous heterozygous SNVs in GCFC2 (NM_001201334.1; c.1298G>A; p.S433N), ELOVL4 (NM_022726.3; c.736T>G; p.W246G), ZBTB24 (NM_014797.2; c.1216C>T; p.P406S), and ENPP3 (NM_005021.3; c.1787C>T; p.T596I) remaining as candidates.

Analysis of copy number variations using SNP array data obtained from family A did not reveal any novel copy number variations that cosegregate with the disease. We further had the opportunity to analyze 2 more affected members of family A (III-4 and III-5), and we tested them for the cosegregation of each of the 4 candidate SNVs. This revealed that the novel variant identified in exon 6 of ELOVL4 (c.736T>G, p.W246G) was the only SNV cosegregating with the disease in family A (Figure 1 and Figure 3A). We next investigated whether the same or allelic variant of ELOVL4 was present in the 2 affected members of family B (II-1 and III-1) because these members showed clinical and MRI findings highly similar to those of the affected members of family A. Sanger sequencing confirmed the same heterozygous mutation in ELOVL4 (c.736T>G, p.W246G) in the affected members of family B. We screened 513 healthy controls in the extended Japanese in-house exome database and confirmed that none of them harbored this mutation.

To investigate whether the mutation identified in the 2 families (c.736T>G, p.W246G) shares a common ancestral origin, we analyzed the haplotypes around the ELOVL4 mutation locus in the 2 families. We used rare (mean allele frequency <0.10) SNVs identified by whole-genome sequencing in member III-1 of family A, and we reconstructed the haplotypes in the 2 families. Haplotyping around the ELOVL4 locus suggested that the 2 families are unlikely to share a common ancestor (eFigure in the Supplement) because the haplotype of a region of at least 211 kilobases in length around the ELOVL4 locus was not shared by the 2 families.

The bioinformatics functional prediction of the p.W246G mutation indicated that the amino acid residue W246 is highly conserved and that the mutation is damaging (Polyphen-2: 0.963 [damaging], SIFT: 0.000 [damaging], MutationTaster: 1.000 [damaging], and likelihood ratio test: 1.000 [damaging]; Figure 3B). We measured VLCFA contents in the blood samples obtained from 2 affected members (II-1 and III-1 in family A) and found that the ratios of C24:0 to C22:0, C25:0 to C22:0, and C26:0 to C22:0 were normal.

The present genetic study identified a novel ELOVL4 mutation (p.W246G) as the sole missense mutation cosegregating with the disease in 2 independent Japanese families with SCA. This mutation, which involved a highly conserved amino acid residue, was predicted to be functionally deleterious using bioinformatics tools. The allelic mutation p.L168F in ELOVL4 has recently been described in a French-Canadian SCA pedigree with EKV (SCA34; OMIM 133190) (Figure 3C).^3,24 Our findings further confirmed that heterozygous missense mutations in ELOVL4 cause SCA.

The previously reported French-Canadian family and our Japanese families are clinically similar in the cardinal clinical feature of slowly progressive cerebellar ataxia. However, several important differences are noted (Table). First, none of our mutation carriers had skin lesions of EKV, which were found with a high frequency (14 of the 19 mutation carriers [73.7%]) in the French-Canadian family. Second, neurological signs, such as pyramidal tract signs, ocular disturbances (gaze palsy and nystagmus), and autonomic symptoms (bladder disturbance and constipation), were more pronounced in our Japanese families compared with the French-Canadian family (Table). Thus, SCA34 caused by the p.W246G mutation did not result in skin lesions but showed broader neurological phenotypes than that caused by the p.L168F mutation. Intriguingly, missense mutations in ELOVL5, another member of the elongase family, have recently been reported to cause SCA38 (OMIM 615957).² In SCA38, a slowly progressive course of gait ataxia is also a common clinical feature,² whereas pyramidal tract signs, autonomic symptoms, and EKV skin lesions are absent (Table).

Of clinical importance, the MRI scans of the brains of the members in our Japanese families showed not only cerebellar atrophy but also marked pontine atrophy with the hot cross bun sign, pontine midline linear hyperintensity, and high intensity in the middle cerebellar peduncle. The hot cross bun sign was observed in 63.3% of patients with multiple system atrophy in a Japanese cohort,²⁷ and pontine midline linear hyperintensity precedes the hot cross bun sign in multiple system atrophy,²⁵ indicating that these 2 MRI-detected abnormalities constitute a continuum. Besides being observed in patients with multiple system atrophy, these MRI findings are also observed in patients with SCA; of 138 Taiwanese patients with SCA, 8.7% had the hot cross bun sign, and 39.1% had pontine midline linear hyperintensity.³¹ Because all 6 family members that we studied showed either the hot cross bun sign (66.7%) or pontine midline linear hyperintensity (33.3%), these MRI findings could be a marker for patients with SCA34 with the p.W246G mutation. In contrast, these MRI findings were not noted in the French-Canadian family with SCA34 with the p.L168F mutation or in SCA38 families (Table). Considering the pathological basis of the hot cross bun sign in multiple system atrophy,^26,32 the presence of these 2 MRI findings might indicate that pontocerebellar fibers and reticular formation are affected much intensely in SCA34 caused by the p.W246G mutation compared with SCA34 caused by the p.L168F mutation and SCA38.

ELOVL4 encodes a 314 amino acid protein (namely, elongation of very long chain fatty acids protein 4 [ELOVL4]) that is expressed in the retina, brain, thymus, testis, and skin.^28,33 ELOVL4 is mainly localized to the endoplasmic reticulum membrane, and in humans there are 7 elongases (ELOVL1-7) that are involved in the elongation of C26 fatty acids to C28 or longer, which are subsequently utilized for the biosynthesis of substances such as phosphatidylcholine, sphingomyelin, and ceramides.^34,35 Small amounts of phosphatidylcholine and sphingomyelin containing very long-chain polyunsaturated fatty acids exist in the retina, brain, and testis,³⁴ although their physiological roles in the brain are unclear. In the skin, ceramides containing VLCFAs of C28 or longer are important components of the epidermidis,³⁵ and the loss of ELOVL4 by homozygous nonsense mutations in humans (ichthyosis, spastic quadriplegia, and mental retardation [OMIM 614457]) and in Elovl4 knockout mice leads to severe neonatal skin symptoms.^29,30 These results suggest that a loss-of-function mechanism might partially explain the occurrence of EKV skin lesions, which are observed soon after birth in ELOVL4 p.L168F carriers. Because we have not directly measured these VLCFAs with carbon chains of C28 or longer owing to technical difficulties, further investigation is required to determine whether the production of such VLCFAs is affected or involved in the pathogenic mechanism of SCA34.

Heterozygous C-terminal truncating mutations in ELOVL4 are known to cause autosomal dominant Stargardt-like macular dystrophy (OMIM 600110) (Figure 3C).³⁰ Basic research on Stargardt-like macular dystrophy showed that the mutant protein lacking the C-terminal dilysine motif that confers endoplasmic reticulum localization by a retention mechanism has a dominant-negative effect on the wild-type protein and alters its trafficking, and it also disturbs the biosynthesis of VLCFAs.^30,36-40 These facts suggest that heterozygous mutations of p.W246G or p.L168F may also cause SCA through a loss-of-function mechanism by the depletion of VLCFAs in the brain and also through a gain-of-function mechanism by aberrant protein trafficking.

When considering the differences between the p.W246G and p.L168F mutations in ELOVL4 and the p.G230V mutation in ELOVL5 causing SCA38 (Table) from a molecular standpoint, we find that the sites of their mutated amino acids relative to the transmembrane helices are of particular importance because elongases are multipass transmembrane proteins. The ELOVL4 protein was previously predicted by Zhang et al,²⁸ using the classic method of Kyte and Doolittle,⁴¹ to have 5 transmembrane helices. However, considering the lack of a crystallographic study, there is uncertainty about its topology, and furthermore, Sur4p/Elo3p, the yeast homologue of ELOVL4, was recently predicted to be a 7-pass transmembrane protein.⁴² Therefore, we sought to predict the topology of ELOVL4 using recently developed and more accurate bioinformatics tools (see Methods).^19-23 All the tools unexpectedly predicted ELOVL4 and ELOVL5 to be 7-pass transmembrane helix proteins (Figure 4A and B; see details in eTables 3 and 4 in the Supplement) and suggested that both W246 in ELOVL4 and G230 in ELOVL5 are on the border between the small loop in the endoplasmic reticulum lumen and helix 7, whereas L168 in ELOVL4 is within helix 4.

Furthermore, alignment of ELOVL4 and ELOVL5 sequences showed that W246 in ELOVL4 and G230 in ELOVL5 are only 1 amino acid away from each other (Figure 4C). Considering that mutations in both of these elongases lead to SCA without skin or retinal lesions, the similar locations of these 2 mutated amino acids suggest a common molecular mechanism selectively resulting in SCA. In addition, the brain might be more vulnerable to the effects of these mutations than the skin or retina, possibly owing to the disturbed production of VLCFAs or the aberrant protein trafficking. On the other hand, because L168 is only 6 amino acids after the conserved histidine cluster that is essential for the catalytic activity of ELOVL4,⁴³ it is possible that a negative effect on the catalytic site leads to a partial loss of function, resulting in EKV skin lesions reminiscent of a recessive human disease (ichthyosis, spastic quadriplegia, and mental retardation) and Elovl4 knockout mice.

We described the clinical features of 2 Japanese families with SCA with multiple system atrophy–like features detected on MRI scans and discovered the novel causative mutation p.W246G in ELOVL4. Along with the study³ of the original French-Canadian family with SCA34, our study confirms that mutations in ELOVL4 can cause SCA34 and broadens its clinical spectrum. These mutations in ELOVL4 and the recently identified mutations in ELOVL5 comprise a spectrum of mutations in elongases that leads to SCAs.

Accepted for Publication: April 8, 2015.

Corresponding Author: Kinya Ishikawa, MD, PhD, Department of Neurology and Neurological Science, Graduate School of Medical and Dental Sciences, Tokyo Medical and Dental University, 1-5-45 Yushima, Bunkyo, Tokyo 113-8519, Japan (pico.nuro@tmd.ac.jp).

Published Online: May 26, 2015. doi:10.1001/jamaneurol.2015.0610.

Author Contributions: Dr Ishikawa had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: Ozaki, Sato, Yamane, Tanaka, Tsuji, Mizusawa, Ishikawa.

Acquisition, analysis, or interpretation of data: Ozaki, H. Doi, Mitsui, Sato, Iikuni, Majima, Yamane, Irioka, Ishiura, K. Doi, Morishita, Higashi, Sekiguchi, Koyama, Ueda, Miura, Miyatake, Matsumoto, Yokota, Tsuji, Mizusawa, Ishikawa.

Drafting of the manuscript: Ozaki, Yamane, Tsuji, Ishikawa.

Critical revision of the manuscript for important intellectual content: Ozaki, H. Doi, Mitsui, Sato, Iikuni, Majima, Yamane, Irioka, Ishiura, K. Doi, Morishita, Higashi, Sekiguchi, Koyama, Ueda, Miura, Miyatake, Matsumoto, Yokota, Tsuji, Mizusawa, Tanaka.

Statistical analysis: Sato, Morishita.

Obtained funding: Yamane, Mizusawa, Ishikawa.

Administrative, technical, or material support: H. Doi, Sato, Irioka, Sekiguchi, Ueda, Miura, Miyatake, Matsumoto, Tanaka, Tsuji, Mizusawa, Ishikawa.

Study supervision: Yamane, Yokota, Tanaka, Tsuji, Mizusawa.

Conflict of Interest Disclosures: None reported.

Funding/Support: This study was funded by the Health and Labour Sciences Research Grants on Ataxic Diseases from the Ministry of Health, Labour, and Welfare, Japan (Drs Mizusawa, Tsuji, and Ishikawa), Grants-in-Aid for Scientific Research from the Japan Society for the Promotion of Science (Drs Matsumoto, Mizusawa, and Ishikawa), the Strategic Research Program for Brain Sciences from the Ministry of Education, Culture, Sports, Science, and Technology of Japan (Drs Matsumoto and Mizusawa), a grant for Core Research for Evolutional Science and Technology from the Japan Science and Technology Agency, Saitama, Japan (Dr Mizusawa), and Grants-in-Aid for Scientific Research on Innovative Areas, “Exploring molecular basis for brain diseases based on personal genomics,” from the Ministry of Education, Culture, Sports, Science, and Technology of Japan (Drs Morishita, Tanaka, Tsuji, and Ishikawa).

Role of the Funder/Sponsor: The funding sources had no role in the design and conduct of the study; collection, management, analysis, or interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.