The hereditary spastic paraplegias (HSPs) are a group of neurodegenerative diseases clinically characterized by progressive lower limb spasticity, pyramidal weakness and extensor plantar responses. The main neuropathological feature of these conditions is declining degeneration of the corticospinal tracts. Pure (uncomplicated) and complicated forms of HSP have been described according to the presence of additional neurological features such as cerebellar ataxia, peripheral neuropathy and/or cognitive impairment, in addition to spastic paraparesis. In addition to clinical manifestation, age at onset and genetic background are also highly heterogeneous.1, 2 To date, at least 46 different loci have been mapped, associated with autosomal dominant (AD), autosomal recessive and X-linked mode of inheritance.1, 2, 3 A total of 17 responsible genes have been identified according to the HUGO and OMIM databases. Consistent with this high variability, HSP gene products seem to be involved in a wide range of cellular functions and pathogenic mechanisms.4 AD inheritance is the most common trait in pure HSP. At present, 18 loci have been linked to ADHSP, and 9 genes have been identified. Mutations in genes SPASTIN (MIM# 604277; SPG4, MIM# 182601), ATL1 (MIM# 606439; SPG3, MIM# 182600), KIF5A (MIM# 602821; SPG10, MIM# 604187) and REEP1 (MIM# 609139; SPG31, MIM# 610250) account for 40, 10, 3 and 8% of all ADHSP patients, respectively.2, 5, 6, 7 The remaining HSP genes seem to be relatively rare.8, 9 Spastic paraplegia type 42 (SPG42, MIM# 612539) was recently mapped to the 3q24–26 chromosomal region in a single Chinese family presenting with pure ADHSP. On the basis of the detection of a serine-to-arginine substitution at codon113 in this family, the authors reported SLC33A1 (MIM# 603690) to be the responsible gene.10 The protein product of SLC33A1 is an acetyl-CoA transporter, which serves as a substrate of acetyltransferases that modify the sialyl residues of gangliosides and glycoproteins.11 It is hypothesized that the modification of gangliosides and glycoproteins by acetylation probably has a critical role in the outgrowth and maintenance of axons of motor neurons.10

The frequency of SPG42 and its associated phenotypes is not known, as no further families with SLC33A1 mutations have been described so far.10 The purpose of this study was to determine the frequency of SPG42 by screening 220 European (German, French, Norwegian) ADHSP patients for conventional and gene dosage mutations.



A total of 220 DNA samples of unrelated HSP patients were recruited through the European Spastic Paraplegia network (EUROSPA) and the Norwegian HSP outpatient clinic. All patients were assessed by experienced neurologists. For all patients, a family history consistent with dominant inheritance was reported. All index patients were of European descent (Germany (n=45), France (n=147) and Norwegian (n=28)). The criterion for inclusion in this study was a mutation-negative state for SPAST. Mutations in the ATL1 and REEP1 gene had previously been excluded by direct sequencing and multiplex ligation-dependent probe amplification (MLPA) at the following frequencies: SPG3A sequencing in 55% of cases, SPG3A MLPA in 85% and SPG31 (sequencing and MLPA) in 84% of cases.7, 9, 12, 13, 14 The HSP phenotype was pure in 125 (57%) and complicated in 95 (43%) patients. Informed consent was obtained in all cases.


Analysis of the SLC33A1 gene

DNA was extracted from peripheral blood samples following standard protocols. The six coding exons of the SLC33A1 gene (ENSG00000169359, ENST00000359479) were screened by high-resolution melting curve analysis (HRM) in all patients. The primers were designed to flank the coding regions and to amplify fragments of an average size of 250 bp (with the largest amplicon spanning 298 bp). PCR and HRM were performed in a single run on a LightCycler 480 instrument (Roche Diagnostics, Mannheim, Germany). A normal control DNA, which had been completely sequenced, served as the HRM baseline sample in each fragment. Samples with an aberrant melting profile were assigned for validation sequencing. These samples were reamplified from genomic DNA and directly sequenced to identify or exclude sequence variants. Forward and reverse sequence reactions were performed with the Big Dye Terminator Cycle Sequencing Ready Reaction Kit (Applied Biosystems Inc., Foster City, CA, USA) using the same primers. The sequence products were analyzed on an ABI3100 Genetic Analyzer (Applied Biosystems Inc.). See supplement for detailed experimental procedures and primer sequences (Supplementary Table 1).

To screen for copy number aberrations affecting SLC33A1, we designed an MLPA assay. It targets the coding sequence of each exon, as well as the promoter sequence of the SLC33A1 gene, with one probe each. Three probes localizing to different chromosomes are included as references (probe sequences are available on request). Pertinent synthetic oligonucleotides (MWG Biotech, Ebersberg, Germany) and reagents from the EK1 kit provided by MRC-Holland (Amsterdam, The Netherlands) were used. MLPA reactions were performed according to the instructions of the manufacturers. Analysis of MLPA data was carried out as described previously.15


HRM analysis detected an aberrant melting profile caused by an A>G substitution at position 512 of the coding sequence that causes a p. D171G substitution in 16 index patients (Figure 1). This sequence abnormality represents the known single-nucleotide polymorphism (SNP) rs3804769 according to the SNP database.16 No other sequence alterations were detected. As HRM analysis is a screening method with 99% sensitivity,17, 18 we handled the results with care. Along with this high sensitivity, HRM is a simple, rapid and low-cost method to screen for unknown sequence variants in a high-throughput modality. Compared with other methods, such as denaturing high-performance liquid chromatography, HRM offers at least two additional benefits: (1) the temperature gradient covers all potential melting domains of amplicons, which provides superior detection sensitivity; and (2) the closed-tube method reduces post-PCR manipulations and the risk of pipetting errors and contaminations.17, 18 Moreover, our identification of the rs3804769 SNP represents a positive control and argues for the general validity of our approach.

Figure 1
figure 1

Known SNP rs3804769 in SLC33A1. In total, 16 patients were found to be heterozygous for rs3804769 in SLC33A1 exon 1 (c.512A>G, p.D171G). A total of 16 variant alleles (7.0%) were present in 440 sample chromosomes, as assessed by high-resolution melting curve analysis (left part). Difference plot analysis revealed the heterozygous group (red), in addition to the homozygous samples (blue). Heterozygosity has been validated by conventional sequencing (right part).

Similar to all PCR-based techniques, large insertions or deletions are usually not detected during gene scanning by HRM.17 This mutational class is frequent (20%) in SPG4 and is also relevant for SPG31.7, 14, 19 Both forms are associated with pure ADHSP. To assess quantitative changes in SLC33A1 copy number, MLPA was used but no rearrangements were identified. As no duplications or deletions of SLC33A1 have been reported so far, no positive control was available. However, the technique can be regarded as rather robust, and homemade probe sets developed by other groups (ie, Ganesamoorthy et al20), as well as by our groups (unpublished), have been successfully used in unambiguously detecting copy number alterations. Moreover, high overall data quality (data not shown) argues for the applicability of this assay; therefore, we believe that our findings are not likely to be false negative.

Screening for gene dosage seemed especially important in SPG42, as haploinsufficiency is supposed to be the pathogenic disease mechanism.10 In addition, SLC33A1 has an unusually high content of Alu sequences (38.1% compared with the 10% genomic average21). In other genes, such high values are associated with an increased susceptibility to genomic deletions (ie, MSH2).22 Nevertheless, MLPA failed to identify this potential pathological mutation class in our study.

The 220 index patients screened in this study represent a cohort of more than 500 consecutive ADHSP index patients, as mutations in common HSP genes (SPASTIN, ATL1 and REEP1), which account for >50% of our ADHSP diagnoses, had already been excluded. On the basis of data published previously by Lin et al,10 the results of this study can be interpreted in two ways. First, the lack of SLC33A1 mutations in our sample may indicate that SPG42 is an extremely rare form of ADHSP, at least in our European cohort. An alternative explanation for the negative finding of our study may consider the SLC33A1 gene to not be the true disease gene for the SPG42 HSP subtype. As only a single SPG42 family has been described so far, it is difficult to draw any genotype/phenotype correlations. In this Chinese family, 20 affected subjects presented with a ‘pure’ form of HSP starting between 4 and 42 years of age; only some had wasting of lower limbs mentioned, apart from pyramidal symptoms.

Identification of additional SPG42 families would help to decide whether SPG42 represents a pure spastic paraplegia or whether a complicated course might be seen in other patients with different mutations. Moreover, a next-generation sequencing approach encompassing the whole locus in the only SPG42 family known so far should help to search for sequence alterations in the nearly 125 genes that have not been screened yet.