A Homozygous Mutation in KCTD7 Links Neuronal Ceroid Lipofuscinosis to the Ubiquitin-Proteasome System

Main Text

Neuronal ceroid lipofuscinosis (NCL) comprises a group of at least nine distinct and currently untreatable lysosomal storage diseases with overlapping clinical features, including progressive motor and cognitive decline, pigmentary retinal degeneration and vision loss in most cases, seizures, movement disorder, and eventual premature death.¹ NCL is estimated to have a worldwide prevalence of 1:100,000 individuals, and there is higher prevalence in some geographic regions.¹ Most forms of NCL are autosomal recessive and show autofluorescent lysosomal storage material with distinctive ultrastructural morphology in many cell types. At least nine genes have been implicated in the pathogenesis of NCL: PPT1 (CLN1 [MIM 600722]), TPP1 (CLN2 [MIM 607998]), CLN3 (MIM 607042), CLN5 (MIM 608102), CLN6 (MIM 606725), MFSD8 (CLN7 [MIM 611124]), CLN8 (MIM 607837), CTSD (CLN10 [MIM 116840]),² and DNAJC5 (CLN4; autosomal-dominant Kufs disease or late-onset NCL [MIM 611203]).^3,4 Despite the discovery of over 365 NCL-causing mutations in these genes, many clinically suspected cases of NCL remain molecularly undefined,⁵ and these include ∼8% of probands enrolled in the Massachusetts General Hospital (MGH) NCL Registry. With the aim of identifying novel causes of NCL, we focused our efforts on samples from a family with more than one affected member with strong clinical and histopathological evidence of NCL (Figure 1). We report the results of whole-exome sequencing and functional studies to identify the gene defect in this family, in which screening of the defined NCL-associated genes failed to identify known or predicted deleterious variants.

Medical history from several individuals in a Mexican family whose ancestors originated from the state of Michoacán was reviewed, and DNA samples were obtained after all subjects or legal guardians gave informed consent as required by the institutional review board at MGH. Both affected siblings (IV-1 and IV-2, Figure 2A) had a normal birth and infancy and suffered from seizures manifesting at 9 (IV-1) and 8 (IV-2) months of age. Seizure manifestations, confirmed by electroencephalography and video telemetry, included frequent myoclonic movements involving mainly the face and extremities, were often precipitated or worsened by fevers, and were refractory to multiple antiepileptic drugs. Normal developmental milestones were documented until the age of approximately 18 months, at which point motor and speech regression were noted. By the ages of 12 years (IV-1) and 10 years (IV-2), both siblings had microcephaly, were nonverbal, and were without spontaneous motoric function. Both IV-1 and IV-2 showed no response to visual threat and had markedly diminished pupillary light reflexes; in addition, IV-2 showed mild bilateral optic atrophy without apparent retinopathy. Brain imaging in both probands showed global cortical atrophy, particularly marked in the cerebellum, as well as thinning of the corpus callosum and some loss of subcortical white matter in IV-1. Comprehensive metabolic testing of plasma, urine, and cerebrospinal fluid was normal in both siblings. A skin biopsy performed on IV-2 for clinical diagnostic purposes at 11 years of age revealed prominent NCL-type storage material in fibroblasts, neurons, and eccrine secretory epithelial cells. Electron microscopy of a clinical buffy coat sample from IV-2 was also remarkable for lysosomal storage material containing fingerprint-like profiles and granular osmiophilic deposits in ∼15% of the analyzed lymphocytes (Figure 1A). Moreover, immunoblot analysis probing for subunit c of the mitochondrial ATP synthase, the main stored proteolipid in most forms of NCL with fingerprint, rectilinear, and curvilinear storage profiles, revealed that the affected sibling had higher levels of subunit c in a lymphoblastoid cell line than did the unaffected sibling. This was similar to what was seen in a lymphoblastoid cell line from a juvenile NCL (JNCL) subject with a homozygous nonsense change in CLN3 (Figure 1B). Full sequencing of exons and flanking regions for all known genes associated with autosomal-recessive NCL (PPT1, TPP1, CLN3, CLN5, CLN6, MFSD8, CLN8, and CTSD) identified no known or predicted pathogenic mutations. Molecular analysis of genes associated with seizure disorders (CDKL5, MECP2, SCN1A, SCN1B), progressive myoclonic epilepsy (EPM2A, EPM2B, EFHC1, CSTB, and dodecamer repeat expansion in CSTB), and mitochondrial disease (including POLG1 mutations and mutations associated with MELAS, MERRF, NARP, LHON, KSS, and CPEO) were unrevealing. Peripheral-blood lymphocytes from IV-2 were cytogenetically normal. Large genomic deletions, duplications, and insertions were ruled out on subjects IV-1 and IV-2 by array-based comparative genomic hybridization with an Agilent SurePrint G3 Human 4x180K microarray (MGH Diagnostic Molecular Pathology Laboratory, Boston, MA). Both siblings died from complications of progressive disease in their mid teens but showed no apparent extra-CNS disease. A third sibling in this family (IV-3) remains asymptomatic at age 7.

To identify a potential causative mutation in this family, we performed whole-exome capture and sequencing by using a NimbleGen SeqCap EZ Human Exome Library v.2.0 (Otogenetics Corporation; Norcross, GA) on samples from the two affected siblings (IV-1 and IV-2) and the unaffected sibling (IV-3, Figure 2A). The average coverage across each targeted exome was >50×, and >95% of all known SNPs were detected. The raw sequence data was processed through the variant-calling pipeline at the Broad Institute.^6,7

Because the pattern of inheritance in this family was most likely autosomal recessive, we screened the sequence data for all pairs of homozygous and compound-heterozygous nonsynonymous (i.e., nonsense, splice site, readthrough, missense, and indel) variants that were shared in the two affected siblings (IV-1 and IV-2), but not in the unaffected sibling (IV-3, Figure 2A). We required the variants to be present at ≤0.1% minor allele frequency in a set of 5,400 controls from the Exome Variant Server (National Heart, Lung, and Blood Institute [NHLBI] Exome Sequencing Project, Seattle, WA) given that the frequency of the most common known NCL-associated point and indel variants² in this dataset is no greater than 0.1% (data not shown). Only one such variant fulfilled these criteria, and it mapped to exon 4 of KCTD7, which is found in chromosomal region 7q11.21 and encodes potassium channel tetramerization domain-containing protein 7 (isoform 1 [RefSeq accession number NM_153033]; isoform 2 [RefSeq NM_001167961.2]). This KCTD7 variant—c.550C>T leading to the predicted missense change p.Arg184Cys—was confirmed by Sanger sequencing and was found to cosegregate in a recessive pattern with the clinical phenotype; both affected siblings were homozygous for the change, and the unaffected sibling (IV-3), both parents (III-5, III-6), and paternal grandmother (II-2) were heterozygous for the variant (Figures 2A and 2B). The c.550C>T variant in KCTD7 was also absent in another 822 control individuals from the 1000 Genomes Project;⁸ 58 of these controls were of Mexican ancestry. Moreover, the p.Arg184Cys missense change was predicted to be deleterious by HumVar-trained PolyPhen-2⁹ and SIFT¹⁰ and is found in a particularly highly conserved region of KCTD7 (Figure S1, available online).

Mutations in KCTD7 have previously been linked to progressive myoclonic epilepsy (PME) in two families (Table 1).^11,12 Similar to the subjects we describe here, affected members of these families presented before 2 years of age with myoclonic seizures primarily affecting the face and extremities. All affected individuals showed developmental regression, and their disease progressed to death or severe disability by the first or second decade of life. Given that PME is one of several seizure types observed in NCL and given the strong evidence that the p.Arg184Cys KCTD7 mutation is the causal variant in the NCL-affected family studied here, we performed Sanger sequencing of the KCTD7 coding region on an additional set of 32 NCL samples (comprising both recessive and dominant forms of NCL with heterogeneous seizure types) that remain without molecular diagnosis. However, no further KCTD7 mutations were identified in this collection of samples.

In the Van Bogaert et al. study, in which a homozygous KCTD7 nonsense change (p.Arg99^∗) was found in three PME subjects in a large consanguineous Moroccan family, the ultrastructural analysis of skin biopsies from two of the affected individuals was reported to be normal, implying that there was no lysosomal storage in this family.¹¹ Whether other PME subjects with KCTD7 mutations were evaluated for lysosomal storage was not evident (Table 1). Therefore, it is unclear whether KCTD7 mutations produce allele-specific forms of PME, with or without lysosomal storage, or whether these cases in fact represent forms of NCL with false-negative peripheral biopsy results; this latter scenario has been reported to be problematic in the diagnosis of other forms of NCL.¹³ Of note, allele-specific subphenotypes within the NCLs have also been described. For example, the homozygous c.70C>G mutation in CLN8 produces northern epilepsy (progressive epilepsy with mental retardation), which is an early-onset form of NCL without vision loss¹⁴ and is distinct from variant late-infantile forms of CLN8 associated with vision loss. Similarly, certain mutations in CLN6 cause adult-onset forms of NCL without vision loss and appear to be mutually exclusive with those that cause variant late-infantile NCL with vision loss.¹⁵

KCTD7 bears an N-terminal bric à brac, tramtrack, and broad complex/poxvirus zinc finger (BTB/POZ) domain that is homologous to the T1 tetramerization domain of voltage-gated potassium channels¹⁶ and is shared with several other functionally diverse members of the K+ channel tetramerization domain, or KCTD, protein family.^17–20 Little is known about the in vivo function of KCTD7, but Kctd7 transcripts are highly expressed throughout the murine brain, particularly in the mitral cells of the olfactory bulb, hippocampus, cerebral cortex, and Purkinje neurons of the cerebellum.²¹ Overexpression of KCTD7 was reported to decrease the excitability of cultured murine embryonic cortical neurons,²¹ but whether this effect was mediated through potassium or other ion channels was unclear.

To further investigate endogenous KCTD7 in the mouse, we harvested multiple organs from a wild-type 5-month-old C57BL/6J mouse, and we extracted proteins with Tris-buffered saline containing 0.2% Triton X-100. Immunoblotting for KCTD7 with a rabbit polyclonal antibody (ab83237 from Abcam [Cambridge, MA]) recognizing the N-terminus of the protein revealed in whole-brain tissue a high expression of a major ∼31 kDa form that comigrated with full-length KCTD7 that resulted from transient KCTD7 overexpression in HEK 293T cells (Figure 3A). Other major immunoreactive bands included a ∼28 kDa species, which was particularly prominent in the spleen, liver, and kidneys, a ∼37 kDa species most evident in the kidneys, and a ∼62 kDa form most likely corresponding to an SDS-stable dimer, which we also observed in the KCTD7-transiently-transfected HEK 293T cells, but not in untransfected HEK 293T cells (Figure 3A and Figure S2A). Evidence for KCTD7 homodimerization was also previously reported by Azizieh et al., who found that differently tagged forms of KCTD7 in COS7 cells coimmunoprecipitated.²¹ The specificity of the bands was assessed by secondary-only immunoblots, which yielded no bands (data not shown), and by immunoblotting with two additional commercially available KCTD7 antibodies, which also labeled the above-described bands and gave no signal when they were preincubated with the peptide antigen (Figure S2). Here, the presence of multiple immunoreactive bands is consistent with evidence of alternative splice variants²¹ and might indicate proteolytic cleavage products. Therefore, despite the apparent ubiquitous expression of KCTD7 as assessed by Northern blotting,²¹ these data strongly suggest tissue-specific differences in KCTD7 regulation.

Given the high expression of Kctd7 in brain tissue, including in cerebellar granule cells and Purkinje neurons (Allen Brain Atlas), and the prominent cerebellar pathology observed in patients, we assessed KCTD7 levels and localization in conditionally immortalized murine cerebellar granule cells, which have previously been extensively used for cell biological studies of NCL.²² Of the species detected on immunoblots of whole-tissue homogenates, the ∼28 kDa form was the most prominent. Interestingly, all species, i.e., ∼28 kDa, ∼31 kDa, and ∼62 kDa isoforms, were significantly elevated in cerebellar granule cells derived from a murine JNCL model (CbCln3^{Δex7/8/Δex7/8})²¹ (Figure 3A). This suggested an impact of CLN3 loss of function on KCTD7 turnover because compared with Kctd7 expression in cells derived from wild-type (CbCln3^+/+) littermates,²³ Kctd7 expression in CbCln3^{Δex7/8/Δex7/8} cerebellar cells was essentially unchanged. Because immunohistochemistry did not successfully detect KCTD7, we generated N-terminal green fluorescent protein (GFP)-tagged forms of wild-type and p.Arg184Cys KCTD7 by using Gateway Technology (Invitrogen, Carlsbad, CA) and transiently transfected cerebellar cells with the TransIt Express reagent (Mirus Technologies, Madison, WI) (Figure 3B and Figures S2 and S3) in order to assess whether the identified mutation affected expression and/or localization. Wild-type GFP-KCTD7 showed broad, somewhat punctate cytoplasmic localization and distinct signal at the plasma membrane. In contrast, GFP-KCTD7 p.Arg184Cys showed more diffuse cytoplasmic localization and markedly diminished plasma-membrane signal, as well as prominent cytoplasmic aggregates (Figure 3B and Figure S3). All together, these results suggest that the mutation affects the trafficking and/or solubility of KCTD7. Expression of GFP alone or of another fusion protein, GFP-LC3, a marker of autophagosomes,²⁴ did not produce similar aggregates in cerebellar cells (data not shown and Cao et al.²⁵). Additional localization studies in other neuronal subtypes, for example, in primary neurons, will help further clarify the impact of the p.Arg184Cys missense change on KCTD7 trafficking, turnover, and function.

Similar to other members of the KCTD family,^17,18 KCTD7 has been shown to interact with Cullin-3,²¹ a scaffolding component of several E3 ubiquitin-ligase complexes that selectively tag proteins for degradation by the proteasome.²⁶ These interactions are mediated, at least in part, through the BTB/POZ domain,^17,18,27 but evidence points to the role of critical residues outside this domain as well.^17,28,29 To determine whether the p.Arg184Cys variant, which falls outside the BTB/POZ domain (Figure S1), affects its interaction with Cullin-3, we transiently cotransfected HEK 293T cells with constructs encoding Myc-tagged Cullin-3 or Cullin-1 (another E3 ubiquitin-ligase protein component) and untagged wild-type or p.Arg184Cys KCTD7, and we immunoprecipitated lysates with a Myc antibody (clone 9E10, Origene, Rockland, MD). Cullin-3, but not Cullin-1, interacted robustly with the SDS-stable dimeric form of wild-type KCTD7, but the p.Arg184Cys missense change abrogated the interaction with Cullin-3 (Figure 3C). The abrogated interaction between KCTD7 and Cullin-3 might lead to an accumulation of substrate proteins, which might be toxic and might lead to altered homeostasis of the ubiquitin-proteasome system (UPS) and the other protein-degradation systems within the cell.

The identification of the KCTD7 mutation described in this family establishes a rare, infantile-onset subtype of NCL and indicates that KCTD7 should be considered in the diagnostic workup for molecularly undefined forms of NCL in addition to PME.^11,12 Moreover, further assessment for evidence of lysosomal storage in PME subjects with KCTD7 mutations is warranted. In light of our recent findings, we propose a designation of CLN14 for this form of NCL; the synonymous CLN14 gene symbol for KCTD7 was approved by the Human Genome Organization (HUGO) Gene Nomenclature Committee. The functional effects of the causal variant in the current study are consistent with the previously defined role of defective protein processing in NCL^25,30 and support a contribution from the UPS. In particular, increasing evidence shows significant crosstalk between autophagy and the UPS in a broad range of neurodegenerative diseases.^31,32 The activity of cullin family members has specifically been linked to endosomal³³ and autophagosomal^34,35 maturation, processes that are defective in a murine model of JNCL.²⁵ Therefore, determining the substrate(s) of the predicted KCTD7- and Cullin-3-containing ubiquitin-ligase complex and their potential role in NCL pathogenesis will be of great interest in future studies.

Supplemental Data

Web Resources

The URLs for data presented herein are as follows:

• Allen Brain Atlas, http://mouse.brain-map.org/

• Clustal Omega, http://www.clustal.org/omega

• HUGO Gene Nomenclature Committee, http://www.genenames.org/

• NCBI UniGene, http://www.ncbi.nlm.nih.gov/unigene/

• NCL Mutation Database, http://www.ucl.ac.uk/ncl/mutation.shtml

• NHLBI Exome Sequencing Project, http://evs.gs.washington.edu/EVS/

• Online Mendelian Inheritance in Man, http://www.omim.org/

• PolyPhen-2, http://genetics.bwh.harvard.edu/pph2

• SIFT, http://sift.jcvi.org/