Mutations in the KIAA0196 Gene at the SPG8 Locus Cause Hereditary Spastic Paraplegia

Subjects

Protocols were approved by the Ethics Committee of the Centre Hospitalier de l’Université de Montréal. Patients gave informed consent, after which patient information and blood was collected. DNA was extracted from peripheral blood through use of standard protocols.

Genotyping and Locus Exclusion

PCR-amplified fragments incorporating α-³⁵S–2-deoxyadenosine 5-triphosphate were resolved on 6% denaturing polyacrylamide gels. Alleles were run alongside an M13mp18 sequence ladder and were scored on the basis of allele sizes and frequencies from the Fondation Jean Dausset-CEPH database. LOD-score calculations and multipoint analysis were performed using the MLINK program of the LINKMAP software package.¹⁸

Mutation Screening

The 28 exons of KIAA0196 were screened by automated sequencing, including at least 50 bp of each intronic region. Primers were designed using the PrimerSelect program (Lasergene) and were synthesized by Invitrogen Canada. Primer sequences and amplification conditions for each exon are listed in table 1.

Variants were first tested in 12 control individuals by sequencing, followed by allele-specific oligomerization (ASO).^19,20 In brief, 4 μl of PCR product was hybridized onto Hybond-N+ Nylon membranes (Amersham Biosciences) by use of a dot-blot apparatus. P³²-labeled probes specific to the mutation or normal sequence were hybridized and then visualized on autoradiographic film after overnight exposure. ASO primers for exon 11 are 5′-ACTAGAAAACCTTCAAGCT-3′ (normal) and 5′-ACTAGAAGACCTTCAAGCT-3′ (mutated). For exon 14, ASO primers of 5′-GGAGAGTTGGTATC-3′ (normal) and 5′-GGAGAGTTCGTATC-3′ (mutated) were used. Exon 15 ASO primers were 5′-CACTGAAGGTTTTG-3′ (normal) and 5′-CACTGAAGTTTTTG-3′ (mutated).

Protein-Sequence Alignment

Cluster analysis was performed using the Probcons (v. 1.09) program. Proteins from aligned species included Homo sapiens (Q12768), Canis familiaris (GenBank accession number XP_532327), Pan troglodytes (GenBank accession number XP_519952), Drosophila melanogaster (GenBank accession number CG12272), Caenorhabditis elegans (GenBank accession number CE13235), Xenopus tropicalis (GenBank accession number MGC89323), Rattus norvegicus (GenBank accession number XP_343250), Danio rerio (GenBank accession number BC045490), Gallus gallus (GenBank accession number XP_418441), Dictyostelium discoideum (GenBank accession number EAL63144), and Mus musculus (GenBank accession number NP_705776.2).

Homology Modeling

The size of the strumpellin protein (1,159 aa) made it prohibitive to obtain a template for the entire protein. Instead, 200 aa (amino acids 501–725) around the two mutations were selected and inputted in the Phyre program version 2.0 (Phyre Protein Fold Recognition Server). The template with the highest score was selected—namely, 1dn1b from the Neuronal-Sec1 syntaxin 1a complex. The SwissProt database viewer (v. 3.7)²¹ was used to visualize the model, with concentration on the α-helix in which the two mutations lie and on a second α-helix in nearby 3D space. Peptides incorporating one or the other identified point mutation were visualized in the same manner.

Expression Studies

Constructs

Each mutation was introduced into the KIAA0196 pBluescript clone by site-directed mutagenesis through use of the primers 5′-CTGGAGAGTTCGTATCCTATGTG-3′ for the exon 14 variant and 5′-CCTATGTGAGAAAATTTTTGCAGATC-3′ for the exon 15 variant, along with primers of their complementary sequence. Wild-type and mutant KIAA0196 cDNAs were cloned, upstream of Myc and His tags, into a pCS2 vector and were transcribed in vitro by use of the SP6 mMESSAGE mMachine kit (Ambion) for zebrafish studies. The protein expression from each of these constructs was validated after their transient expression in cell (HeLa) culture and subsequent western-blot analysis with an anti-Myc antibody. A band at the appropriate height (∼134 kDa for strumpellin) was observed.

Zebrafish Knockdown Studies

Clinical Information and Family Details

Family FSP24 with the SPG8 mutation is from British Columbia. It is composed of 13 members affected with a spastic gait and lower-limb stiffness (fig. 1A); genetic information is available for 10 of them. Symptoms were first observed in individuals between ages 35 and 53 years. Intrafamilial phenotypic heterogeneity exists, as shown by the symptoms presented and the range of disease severity in patients. Deep-tendon reflexes were brisk or increased, and decreased vibration sensation was also noted in three patients. Occasional bladder-control problems were also observed. Walking aids were required for some individuals, whereas one is confined to a wheelchair. Together, these features are consistent with a pure, uncomplicated HSP similar to that described for other families linked to the SPG8 locus. Family FSP29 is of European descent and resides in the United States. There are 31 affected individuals in the family, and genetic information is available for 10 of them (fig. 1B). Age at onset was quite variable, with symptom onset ranging in patients from their 20s to their 60s. The family was negative for mutations in the spastin gene.

Linkage Analysis

Two large families that map to the SPG8 locus were identified. In family FSP24, seven markers spanning the candidate region from markers D8S586 to D8S1128 were genotyped in the 18 individuals studied (fig. 1A). A disease haplotype segregated with the disease in all 10 affected individuals (table 2). A recombination event occurred in one individual (fig. 1A) between markers D8S586 and D8S1804, which defined the proximal border of the locus in this family. A lower recombinant was neither identified nor searched for, since the haplotype extended beyond the limits of the SPG8 locus. The maximum LOD score for this family was 3.43 at θ=0, by use of CEPH allele frequencies for the marker D8S1804, along with a maximum multipoint of 4.20 at marker D8S1799.

The same seven markers tested in family FSP24 were genotyped for family FSP29. A disease haplotype was established for all 10 studied affected individuals; it included many informative recombination events. The proximal recombinant occurred between markers D8S1799 and D8S1832 in three affected individuals (fig. 1B), and the distal recombinant was between markers D8S1774 and D8S1128 for another affected individual (fig. 1B). This yielded a candidate interval of 3.15 Mb. The maximum LOD score for this family was 5.62 (θ=0) for marker D8S1179 when CEPH allele frequencies were used. Multipoint analysis was also conducted for this family in this region, which yielded a maximum LOD score of 6.73, 0.5 cM centromeric to the D8S1128 marker.

Gene Screening

The previously published SPG8 locus spanned 3.4 cM (1.04 Mb) between markers D8S1804 and D8S1179 on chromosome 8q23-8q24. We screened nine known genes surrounding this candidate region, as annotated in the University of California–Santa Cruz Genome Browser (UCSC) May 2004 update, along with many clustered ESTs and mRNAs that aligned to the locus, without detecting a mutation. Therefore, we opted to redefine the candidate region on the basis of the critical interval determined by an upper recombinant in our FSP29 family at marker D8S1832, and a lower recombinant at D8S1774 was based on published data (fig. 2A).¹⁴ This increased the size of the region to 5.43 cM (3.15 Mb), a region that contains three additional known genes (fig. 2B). These additional genes were screened, and three mutations were identified in the KIAA0196 gene (fig. 2C).

Mutation Analysis

A valine→phenylalanine mutation was identified in amino acid 626 for families FSP24 and FSP29 (p.V626F) (fig. 3A). All studied affected individuals from each family were screened and were positive for this mutation. The same mutation was also found to segregate in a British family.¹⁵ This G→T nucleotide change is at position 1956 of the mRNA (GenBank accession number NM_014846.2). A total of 500 ethnically matched control individuals (400 from North America and 100 from CEPH) were negative for this mutation, on screening by a combination of ASO and sequencing. No unaffected members or spouse control individuals in any family were positive for the mutations.

A second mutation was identified in the Brazilian family¹⁶ in exon 14, a G→C transition at position 1937 of the mRNA (fig. 3B). This leucine→phenylalanine change (p.L619F) is only 7 aa away from the V626F mutation. It was not found, with use of ASO, in 500 controls.

The KIAA0196 gene was screened in probands from 24 additional dominant HSP–affected families that are negative for mutations in both spastin and atlastin, resulting in the identification of two more families with missense mutations in the KIAA0196 gene. Thus, the frequency of mutations in our SPG3A- and SPG4-negative autosomal dominant cohort is ∼8% (2 of 24). FSP34 has the same p.V626F change in its three affected studied family members. This family is originally from Great Britain and resides in Canada (fig. 1C). Haplotype analysis of this family with markers D8S1804, D8S1179, D8S1774, and D8S1128 indicated that there is allele sharing between this family and family FSP29, which suggests an ancestral haplotype (table 2). An additional mutation was found in three affected siblings of another North American family of European origin, family FSP91 (fig. 1D). This c.A1491G transition results in an asparagine→aspartate amino acid change (p.N471D) and is not present in the 500 controls tested (fig. 3C).

Mutated amino acids at positions 619 and 626 are strictly conserved across all 11 examined species, from human to the social amoeba, D. discoideum (fig. 3D). Indeed, the entire region surrounding these two mutations appears to be functionally relevant for the protein, since 73 consecutive aa (amino acids 576–649) are 100% identical in the human, dog, chicken, mouse, rat, and orangutan. Despite this high level of conservation, this region is an unknown domain, on the basis of searches of the National Center for Biotechnology Information (NCBI) Conserved Domain Database^, NCBI BLAST, and the Sanger Institute’s Pfam database. Position 471 is conserved across all species except D. melanogaster (with a glutamine residue) and X. tropicalis (with a histidine residue).

The exon 15 mutation is in the very first nucleotide of the exon, which leads to the speculation that the splicing of this exon might be compromised in our study families. Splice-site prediction programs, including NetGene2, suggested that the strength of the splice-site acceptor may be reduced by 33% in the mutant form. However, both normal and mutant alleles were observed in cDNA analysis, with use of several pairs of primers, of patient lymphoblasts. The KIAA0196 gene was expressed ubiquitously, including all regions of the brain that were examined by RT-PCR (fig. 3E). There were no alternative splice isoforms detected in control brain samples or patient whole-blood samples by RT-PCR and northern-blot analysis (fig. 3E and 3F). For the full KIAA0196 gene, all spliced ESTs and mRNAs from the ^UCSC browser, May 2004 draft, were analyzed for potential alternative splice products. One alternative first exon often appears; however, of the 356 entries, only 2 (AK223628 and DA202680) contain exons that are skipped. Thus, overall, the gene is not frequently spliced, and the two spliced entries may represent spurious transcripts.

KIAA0196 Profile

The KIAA0196 gene spans 59.7-kb pairs of genomic DNA, is 28 exons long, and codes for a protein of 1,159 aa (strumpellin). The European Bioinformatics Institute’s InterProScan program predicts a spectrin-repeat–containing domain in amino acids 434–518. Thus, the mutation at position 471 may abrogate the binding of the spectrin domain with other spectrin-repeat–containing proteins. In examination of the secondary structure by use of PSIPRED,²⁴ 74% of the protein is considered to be α-helical. The program further predicts an α-helix in the protein from amino acids 606–644, which encompasses the two other mutations that have been identified.

Homology Modeling

Given the high proportion of KIAA0196 considered to be α-helical, it is not surprising that the optimal homology-modeling candidates are similar in secondary-structure composition. This is true for 1dn1b, a stat-like t-SNARE protein neuronal-Sec1 syntaxin 1a complex. This is the most appropriate model for strumpellin, according to the Phyre program (Phyre Protein Fold Recognition Server). The two mutated residues lie within an α-helix from amino acids 619–628 that is in close 3D proximity to another α-helix from residues 665–670 (fig. 4A). A mutation in either Val-626 or Leu-619 to a phenylalanine residue would appear to have significant structural implications, given the change in bulkiness between the residues. In addition, Tyr-622 points in the same direction from the α-helix residue. To have two amino acids with aromatic rings in such a physical proximity could force apart the alignment of the two α-helices or induce alterations in the α-helix backbone. The N471D mutation was identified well after the two other mutations and so was not tested in homology modeling or in subsequent zebrafish-rescue experiments.

Zebrafish-Rescue Experiments

To validate the functional phenotype of the SPG8 mutations in vivo, we developed a zebrafish model. Morpholino oligonucleotide knockdown of the KIAA0196 protein ortholog in zebrafish (KIAA MO) resulted in an enlarged heart cavity, along with a curly-tail phenotype that severely impaired the ability of the fish to swim properly. The overall phenotype ranged in severity and was classified in three major groups: normal, slightly curly, and severely curly. This phenotype was clearly visible after dechorionating by 1 d post fertilization (dpf). At 3 dpf, wild-type zebrafish are ∼5 mm long, with a straight tail (fig. 5A). Fish injected with a mismatch-control morpholino (CTL MO) were initially used to titer a KIAA MO–specific nontoxic injection dose (fig. 5B). Injection of the KIAA MO resulted in 66 (37%) of 178 fish with a severely curly tail and 50 (28%) of 178 fish with a slightly curly tail (table 3 and fig. 5C and 5D). The KIAA MO fish had a significantly different distribution of phenotypic groups compared with those with CTL MO injections (P<.001). When wild-type human KIAA0196 mRNA was coinjected with KIAA MO, the curly-tail phenotype was rescued to levels comparable to CTL MO injections (P=.51) (fig. 5E and 5F). This suggests that, in zebrafish, human KIAA0196 mRNA can compensate for the loss of endogenous zebrafish mRNA. Conversely, coinjection of human KIAA0196 mRNA incorporating either the exon 14 or exon 15 mutation failed to significantly rescue the phenotype (fig. 5G and 5H). Injection of mutant exon 14 or exon 15 mRNA alone (without morpholinos) did not lead to a curly-tail phenotype or influence lethality in zebrafish, which suggests that the two mutations do not exert a dominant negative effect. Approximately 200 embryos were injected per experimental condition (table 3). The difference in distribution between KIAA MO injection alone and KIAA MO coinjection with wild-type mRNA was significant (P<.001). Similarly, coinjection of wild-type mRNA versus either exon 14 or exon 15 mutant mRNA was significantly different, with a P value <.001. There was no statistical difference between the coinjection of the exon 14 mutant and the exon 15 mutant (P=.10). On histochemical analysis of the embryos by use of an antiacetylated tubulin stain for growing axons, we found that the motor neurons in the spinal cord did not develop normally (fig. 6). Motor-neuron axons in fish injected with KIAA MO alone or with the mutant mRNAs were shorter and showed abnormal branching. The structure of interneurons in the spinal cord was also different. The absence of the KIAA0196 gene or mutations in this gene during early development thus seemed to hamper axonal outgrowth.