The Novel Neuronal Ceroid Lipofuscinosis Gene MFSD8 Encodes a Putative Lysosomal Transporter

Patients, Controls, and Samples

Nine Turkish families with vLINCL (a, b, c, d, e, f, g, h, and l), one family of Turkish-Georgian origin (k), and one family of Indian origin (j), altogether comprising 13 patients, 21 parents, and five unaffected siblings, were analyzed. All families except one (j) were consanguineous. The clinical phenotypes of eight patients were described elsewhere (patients 17, 18, 22, 24, 25, 27, 28, and 29 in the study by Topcu et al.¹⁸). The corresponding patient codes in the present study are e3, a3, f3, b3, c3, d4, d3, and g3, respectively.

Patient e5 (the brother of e3) presented at age 3.5 years with a history of delayed speech development, ataxia since age 2.5 years, and stereotyped hand movements. Brain magnetic resonance imaging (MRI) showed cerebellar and cerebral atrophy and increased signal intensity in white matter. At age 5 years, he had prominent myoclonic seizures and rare nocturnal seizures that responded well to treatment with levetiracetam and intravenous immunoglobulin. Electroencephalogram (EEG) showed slow background activity and multifocal epileptic discharges. Eye ground examination showed retinopathy and optic atrophy. At age 6 years, he could not walk without support. For the past 6 mo, he has been wheelchair bound and unable to speak.

Patient h3 presented at age 5.5 years with a history of delayed speech development, as well as ataxia that had been worsening over the previous 6 mo, frequent falling, and sleep disorders since age 3.5 years. On examination at 5.5 years, she had ataxic gait with abnormal cerebellar tests. She was able to speak slowly. Brain MRI showed cerebellar atrophy. She is taking sodium valproate for myoclonic seizures. She has tested negative for MECP2 (MIM 300005) mutations and for cerebral folate deficiency. No vacuoles were found on analysis of peripheral blood smear.

Patient j3, who was of Indian origin, presented at age 3 years with a history of mild generalized developmental delay. Over the previous 3 mo, her gait had become increasingly ataxic, and, on examination, she walked with a wide-based gait. At age 4 years, she developed frequent clonic seizures that responded well to treatment with sodium valproate. A wide range of metabolic investigations, including enzymologic analysis of PPT1 and TPP1 and molecular analysis of CLN3 and CLN6, were normal. MRI of the brain showed mild cerebellar atrophy and evidence of delayed white matter maturation. A rectal biopsy showed neuronal curvilinear bodies. She had progressive neurological decline and, by age 6 years, was wheelchair bound and had limited communication, significant visual impairment, and frequent breakthrough seizures. At age 7 years, she developed treatment-resistant status epilepticus. She died just before her 8th birthday.

Patient k3, who was of Turkish-Georgian origin, presented at age 8 years with a history of delayed psychomotor development (she walked at age 2.5 years, spoke at age 3 years, and did not learn to read). She showed cognitive deterioration, occasional agitation (with a few hospitalizations in a psychiatric unit), loss of vision, and walking difficulties. At age 22 years, she was agitated, she had axial akinesia, hypertonia, palilalia, and blindness. She developed problems with swallowing. Peripheral blood lymphocytes showed vacuoles, and a skin biopsy revealed vacuoles with fingerprint profiles.

Patient l3 presented at age 2 years with a 6-wk history of ataxia and myoclonic and atonic seizures that responded well to treatment with sodium valproate. Psychomotor development, retinal examination, and MRI were normal, whereas EEG showed occipital spikes. An axillary skin biopsy showed fingerprint profiles.

Genomic DNA was extracted using standard techniques from EDTA blood samples collected after informed consent was obtained. The control panel for the identified MFSD8 mutations consisted of 212 Turkish and 92 Centre d’Etude du Polymorphisme Humain (CEPH) chromosomes. RNA for the analysis of a splice-site mutation was extracted from peripheral blood samples and for the analysis of the alternative splicing pattern of MFSD8, from a control fibroblast cell line. An institutional review board of the Helsinki University Central Hospital approved the study.

Candidate-Locus and Genomewide SNP-Scan Analyses

The previously known NCL loci (CLN1, CLN2, CLN3, CLN5, CLN6, CLN8, CLN10, CLCN3, and CLCN7) were excluded by microsatellite analysis, as described elsewhere.²⁰ Four microsatellite markers (D4S427, D4S2975, D4S2938, and D4S429) spanning ∼6.5 cM (according to the Marshfield Genetic Map in the Mammalian Genotyping Service) were genotyped around the novel CLN7 locus.

Ten families with vLINCL (a, b, c, d, e, f, g, j, l, and k) were included in the genomewide SNP scan. One affected individual was genotyped from each family, except from one pedigree (d) that had four individuals (two affected and two unaffected siblings) genotyped. The SNP scan was performed using Affymetrix’s GeneChip Human Mapping 50K Xba 240 arrays in accordance with the manufacturer’s instructions. SNPs were merged with the Rutgers map²⁴ by their physical locations (NCBI build 35). Interval-specific interpolation was performed. From a total of 58,960 SNPs, the ones without chromosome numbers and physical location (466) or with zero heterozygosity values or male heterozygotes on chromosome X (3,904) were removed. Because the sample size in this data set was too small to calculate SNP heterozygosity values, the values from the Affymetrix white sample were used. Since linkage disequilibrium between markers can cause false-positive results when both parents are not genotyped,²⁵ we selected markers for linkage analysis instead of using all of them. We did this by grouping the remaining 54,590 SNPs into clusters, with at least 0.1 cM between the clusters. One SNP with the largest heterozygosity value was chosen from each cluster, resulting in the use of 8,704 SNPs. The computer program Merlin (version 1.0.1) was used for multipoint homozygosity mapping.²⁶ The disease-allele frequency was set at 0.0001, with penetrances of 0, 0, and 1 for an autosomal recessive model.

Gene Identification and Mutation Analysis

The gene content of the ∼19.5-Mb genomic region between rs2051785 and rs984369 on chromosome 4q28.1-q28.2 was examined with the NCBI Map Viewer, build 35.1. Positional candidate genes were chosen on the basis of the known or putative functions of the encoded proteins. The exons and the exon-intron boundaries of the candidate genes were screened for mutations by genomic sequencing. PCR conditions and primer sequences are available from the authors on request. Sequencing of the purified PCR products was performed with an ABI 3730 DNA Analyzer (Applied Biosystems). Controls were screened for the identified mutations by genomic sequencing.

Northern Blot and RT-PCR

The tissue expression of MFSD8 was analyzed with Multiple Tissue Northern Blots (Human, Human Brain II, and Human Brain V [BD Biosciences Clontech]) with a 559-bp (c.36_594) probe that was PCR amplified from MFSD8 cDNA (I.M.A.G.E Consortium cDNA clone ID 5272664²⁷ and RZPD clone IRATp970E0532D). The probe was [³²P]-labeled with the Rediprime II Random Prime Labeling System (Amersham Biosciences) and was hybridized in accordance with the manufacturer’s protocol (BD Biosciences Clontech).

For RT-PCR, RNA isolated from blood or from cultured fibroblasts was reverse transcribed using M-MLV Reverse Transcriptase (Promega). The consequence of the c.754+2T→A mutation was analyzed using primers from exons 6 and 11, and the alternative splicing of MFSD8 in fibroblasts was analyzed using several exonic primer pairs that covered the entire coding region, as well as primers specific for various exon-lacking transcripts. The PCR products were extracted from agarose gels and were sequenced.

Construction of Expression Plasmids and Site-Directed Mutagenesis

The PCR-amplified ORF of the human MFSD8 cDNA was cloned in frame into the aminoterminal hemagglutinin (HA) tag containing pAHC expression vector (kindly provided by Prof. T. Mäkelä, University of Helsinki, Finland), a derivative of pCIneo (Promega) (^HAMFSD8 construct), and into the pcDNA3.1(+) expression vector (Invitrogen), into which a carboxylterminal HA tag was generated (MFSD8^HA construct) by site-directed mutagenesis by use of the QuikChange Site-Directed Mutagenesis Kit (Stratagene). Two nucleotide changes, c.929G→A (p.Gly310Asp) and c.1286G→A (p.Gly429Asp), corresponding to patient missense mutations, were introduced into the ^HAMFSD8 construct by site-directed mutagenesis. All constructs were verified by sequencing.

In Vitro Translation

In vitro translation was performed with the TnT Quick Coupled Transcription/Translation System (Promega) and the Canine Pancreatic Microsomal Membranes (Promega) by use of the ^HAMFSD8 and MFSD8^HA constructs as templates. The [³⁵S]-labeled translation products were analyzed on 12% SDS-PAGE gels and were visualized by autoradiography.

Cell Culture, Transfections, and Immunofluorescence Analysis

COS-1 and HeLa cells (American Type Culture Collection) were cultivated in Dulbecco’s modified Eagle’s medium (BioWhittaker) supplemented with 10% and 5% fetal calf serum (PromoCell), respectively, penicillin, streptomycin, and 1× GlutaMAX (Gibco). The cells (1.5–2×10⁵ cells) were plated onto 6-well plates on coverslips 1 d before transfection and were transfected with 2 μg of wild-type or missense mutation containing ^HAMFSD8 or wild-type MFSD8^HA constructs per well by use of FuGENE 6 Transfection Reagent (Roche). At 18 h after transfection, cells were incubated with cycloheximide for 2 h, were fixed with 4% paraformaldehyde, and were permeabilized either with 0.1% Triton X-100 in PBS or with 0.2% saponin in PBS supplemented with 0.5% BSA. Overexpressed proteins were detected using mouse monoclonal or rabbit polyclonal anti-HA antibodies (clones 16B12 [Covance Research Products] and Y-11 [Santa Cruz Biotechnology], respectively). Antibodies used as organelle markers were mouse monoclonal anti-early endosome antigen 1 (EEA1 [BD Biosciences]), rabbit polyclonal anti-cathepsin D (CTSD [DAKO]), rabbit polyclonal anti-giantin (BioSite), mouse monoclonal anti–lysosomal–associated membrane protein 1 (Lamp-1, H4A3 [developed by J. T. August and J. E. K. Hildreth and obtained from the Developmental Studies Hybridoma Bank, developed under the auspices of the National Institute of Child Health and Human Development and maintained by The University of Iowa, Department of Biological Sciences, Iowa City]), mouse monoclonal anti-lysobisphosphatidic acid (LBPA, 6C4 [kindly provided by Professor Jean Gruenberg, University of Geneva, Switzerland]), mouse monoclonal anti–mannose-6-phosphate receptor 46 (MPR46 [kindly provided by Professor Kurt von Figura, Göttingen, Germany]), and mouse monoclonal anti–protein-disulfide isomerase (PDI [Stressgen]). Cy2- or Cy3-conjugated secondary antibodies (Jackson ImmunoResearch) were used to visualize the primary antibodies. The cells were viewed using an Axioplan 2 microscope and were photographed with AxioVision 3.1 (Zeiss).

Bioinformatics

Alternatively spliced transcripts were identified using NCBI AceView. The topology of MFSD8 was predicted using HMMTOP^{, TMHMM}, and TMpred prediction programs. Protein domains were searched for in the Pfam 20.0 database. Proteins homologous and similar to MFSD8 were searched for by NCBI protein-protein BLAST. The peptide sequences were aligned using MAFFT version 5.8.

Genomewide SNP Scan and Homozygosity Mapping

Analysis of genomewide SNP-scan data in 10 families with vLINCL—for which we had excluded all known NCL loci either elsewhere²⁰ or in this study (data not shown) (see the “Material and Methods” section)—revealed three regions, on chromosomes 4, 8, and 15, with heterozygosity LOD (HLOD) scores >2 (see the tab-delimited ASCII file, which can be imported into a spreadsheet, of data set 1). Six families (a, b, c, e, f, and j) contributed to the highest HLOD score of 3.39 at SNP rs348085 on chromosome 4q28.1-q28.2, where an ∼19.5-Mb region (from rs7657655 to rs10518621) of overlapping homozygosity was detected in all but one of these families. In family a, the homozygous region was <1 Mb, which was considered to be too short to reflect a region homozygous by descent, considering the close consanguinity. The haplotypes across the region were different in each family (data not shown). Four families (d, e, g, and l) contributed to the HLOD score of 2.48 on chromosome 15 (at SNP rs2956147). Four families (b, c, e, and l) contributed to the HLOD score of 2.05 on chromosome 8 (at SNP rs2432960). We decided to first focus on the ∼19.5-Mb region on chromosome 4 because it had the strongest evidence for linkage. Genotyping four microsatellite markers (D4S427, D4S2975, D4S2938, and D4S429) in patients in families b, c, e, f, and j and in an additional family (h) did not narrow the region but revealed that the patient in family h also is homozygous for these markers.

Identification of Mutations in MFSD8 Underlying vLINCL

Within the critical region of ∼19.5 Mb, we identified at least 90 known or putative genes. After excluding two of the genes (TRAM1L1 and TRPC3 [MIM 602345]) by sequencing, we identified six homozygous mutations (table 1 and fig. 1) in MFSD8 (a HUGO Gene Nomenclature Committee–approved symbol; previously denoted “MGC33302”; GenBank accession number NM_152778), by screening the 12 protein-coding exons (exons 2–13) of the gene. All six mutations cosegregated with the disease phenotype in the respective families and were not found in 212 Turkish and 92 CEPH control chromosomes. Patient j3 was homozygous for nonsense mutation c.894T→G in exon 10, which created a premature stop codon (p.Tyr298X) and was predicted to truncate the protein by 221 aa. Patients c3 and b3 were homozygous for missense mutations c.929G→A (p.Gly310Asp) in exon 10 and c.1286G→A (p.Gly429Asp) in exon 12, respectively, which affect amino acids that are conserved across vertebrates (fig. 2). In two patients, we identified homozygous nucleotide changes at the exon-intron junctions that may either change an amino acid or affect the splicing of the transcript: an A→G transition at the second-to-last nucleotide of exon 7 (c.697A→G; p.Arg233Gly) in patient f3, and a transversion of G→C at the last nucleotide of exon 11 (c.1102G→C; p.Asp368His) in patient h3. Both Arg233 and Asp368 are conserved in vertebrates (fig. 2). The only intronic mutation detected (c.754+2T→A) was identified in intron 8 in two affected siblings, e3 and e5. An RT-PCR analysis performed from an RNA sample of patient e5 with primers from exons 6 and 11 revealed an altered pattern of PCR products (fig. 1B), implying aberrant splicing of the mutant transcript. No mutations in MFSD8 were identified in patients that lacked homozygosity over the CLN7 locus (families a, d, g, k, and l).

MFSD8 mRNA Expression and Alternatively Spliced Variants

In northern-blot analysis of MFSD8 expression, an ∼5-kb transcript, expressed at very low levels, was detected in all tissues analyzed (fig. 3). In all brain regions and in lung, it was the only transcript seen. In other tissues, transcripts that ranged in size from ∼1 to ∼3 kb and/or ∼6 kb were also present.

Both northern-blot and EST database analyses support the ubiquitous expression of the main MFSD8 transcript, including exons 1–13, whereas the alternatively spliced variants seem to have more-limited tissue distribution (NCBI AceView). In particular, four alternative, partial variants were present in EST databases: one lacking exon 2 (BI553701), one lacking exon 7 (BX341359), one lacking exons 7 and 8 (BG434527), and one lacking exon 11 (represented by three sequences: BX341358, BG542082, and BU618424). None of these variants produce full-length in-frame transcripts. Translation of the variant lacking exon 2 is predicted to start at methionine 46 at exon 3 (p.Met46). The variant lacking both exons 7 and 8 produces a transcript with an in-frame deletion of 67 aa, whereas the others lead to frameshifts and premature stop codons. We verified these and identified several other alternatively spliced transcripts by RT-PCR analyses (data not shown).

MFSD8 Protein

MFSD8 is predicted to encode a 518-aa protein of ∼58 kDa with 12 predicted transmembrane domains (fig. 4). Pfam analysis of the MFSD8 amino acid sequence showed that it contains an MFS domain (MFS_1) at amino acid positions p.42_477 and a sugar (and other) transporter domain (Sugar_tr) at amino acid positions p.72_147. In an in vitro translation assay, HA-tagged MFSD8 proteins were detected as ∼60-kDa bands on SDS-PAGE (data not shown), in agreement with the calculated molecular weight of the protein. In spite of several attempts under various conditions, we were not able to detect soluble overexpressed HA-MFSD8 in transfected cell lysates on western blot with the use of HA-antibody, which is likely to be a result of the predicted excessive hydrophobicity of the protein, poor transfer of the protein, or partial proteolysis of the HA tag.

MFSD8 appears to be evolutionarily conserved, since a BLAST search returned several homologues for MFSD8 in different species. In each vertebrate species, MFSD8 has a single orthologue. For instance, mouse and zebrafish orthologues are ∼82% and ∼61% identical to the human protein, respectively (fig. 2). In invertebrates, several weakly similar proteins (∼18%–35% of amino acids are identical to those in human MFSD8) were identified: 2 in Drosophila melanogaster, 18 in Caenorhabditis elegans, and 4 in Saccharomyces cerevisiae.

Intracellular Localization of MFSD8

To study the subcellular localization of MFSD8, we transiently overexpressed the ^HAMFSD8 construct in COS-1 cells and, by using HA antibodies in immunofluorescence analysis, detected the protein in punctate structures in the cytoplasm (fig. 5A, 5D, and 5G). This staining pattern was confirmed by overexpressing MFSD8^HA in COS-1 cells and also by transfecting HeLa cells with both constructs (data not shown). A variety of organelle markers were used to identify the MFSD8-positive punctate cytoplasmic structures in COS-1 cells. The HA-MFSD8 protein displayed the strongest overlap with lysosomal markers Lamp-1 (fig. 5A–5C), CTSD, and LBPA (data not shown). No overlap was observed with early endosomal protein EEA1 (fig. 5D–5F) or with markers used to visualize the compartments of the early secretory pathway, including endoplasmic reticulum (ER) resident PDI, Golgi protein giantin (data not shown), and trans-Golgi network protein MPR46 (fig. 5G–5I). Patient missense mutations p.Gly310Asp and p.Gly429Asp, which were introduced into the wild-type MFSD8, did not interfere with the lysosomal localization of the HA-MFSD8 protein in transiently transfected COS-1 cells (fig. 5J–5O).