Introduction

Hereditary spastic paraplegias (HSP) constitute a group of inherited disorders in which the chief clinical feature is progressive lower limb spasticity. HSP can be classified as “pure” or “complex” if additional features are present [1].

There is marked genetic heterogeneity in HSP, with at least 56 genetic loci (including at least 36 genes) identified so far. More than 11 genes are reported to cause autosomal dominant (AD) HSP. Mutations in the gene SPAST (SPG4) are the most frequent AD form, accounting for 45 % of AD kindreds [2], and often presenting with a pure HSP phenotype. So far, 22 genes have been implicated in autosomal recessive (AR) HSP. Mutations in SPG7 cause AR-HSP with a phenotype that is frequently complicated by cerebellar signs and cerebellar atrophy on neuroimaging [3]. SPG5 due to CYP7B1 mutations can also cause complex forms of HSP with cerebellar signs and white matter hyperintensities on brain magnetic resonance imaging (MRI) [4]. AR HSP with a thin corpus callosum (AR-HSP-TCC) is typically associated with mutations in the SPG11 or ZFYVE26 (SPG15) genes. HSP can also have an X-linked or a mitochondrial [5] mode of inheritance.

Molecular characterization is important for an accurate diagnosis of HSP [6]. Amplicon-based next generation sequencing (NGS) has recently been used to screen for CYP7B1 and SPG7 mutations in a large HSP cohort [6]. We used targeted next generation sequencing (NGS) to screen for multiple HSP-causing genes in an Australian sample in whom SPAST mutations had already been excluded.

Materials and methods

The study was approved by the Northern Sydney Central Coast Human Research Ethics Committee, is in accordance with the Declaration of Helsinki, and all participants gave written informed consent. Subjects were recruited from the adult neurogenetics clinic at the Royal North Shore Hospital, Sydney, Australia. Ethnicity was self-reported. All patients fulfilling Harding’s criteria for HSP [1] who presented to the clinic from January 2009 to April 2013 agreed to participate. Patients underwent a neurologist review, laboratory analysis, brain and spinal cord MRI, nerve conduction studies and motor evoked potentials (MEPs, central motor conduction time calculated according to the F-wave method [7]), and were categorized according to the likely mode of inheritance following detailed family history. Individuals were assigned to the AD panel when several generations of their pedigree were affected in a manner consistent with an AD pattern of inheritance. Patients were allocated to the AR panel when there were two or more affected family members from a single generation only (there was no history of consanguinity in this sample). Apparently sporadic cases were assigned an AR mode of inheritance, consistent with previous studies [6].

All patients underwent screening for SPAST mutations, regardless of the apparent mode of inheritance (e.g. AD, AR or sporadic). SPAST screening was performed using PCR and direct nucleotide sequencing of the entire coding region, as well as screening for deletions/duplications of the SPAST and atlastin genes using multiplex ligation dependent probe amplification. The initial SPAST screening was published in detail elsewhere [8], and SPAST-positive patients were excluded from further analysis in this study.

For those samples that did not have a SPAST mutation detected, sequencing of additional HSP genes was performed according to mode of inheritance, using NGS techniques. The AD panel included the atlastin, NIPA1, KIAA0196, KIF5A, HSPD1, BSCL2, REEP1, ZFYVE27, and SLC33A1 genes and repeat analysis of SPAST. The AR panel included CYP7B1, SPG7, SPG11, ZFYVE26, SPG20, SPG21/ACP33, FA2H, PNPLA6 and GJC2 genes. No patient had a family history suggestive of X-linked or mitochondrial inheritance. Genetic analysis was performed by PCR and NGS (Roche, 454 pyrosequencing) of both DNA strands including the entire coding region, the highly conserved exon–intron splice junctions, and most of the untranslated regions. A total of 173 and 217 fragments were amplified for the AD and AR panels, respectively. The length of the amplicons ranged from 200 to 450 bp. The PCR primers were extended at their 5′ end to include a universal M13 linker sequence and a 6 bp multiplex identifier (primer sequences can be provided upon request). The PCR primers, together with the linker sequences and multiplex identifiers, were synthesized by Eurofins MWG Operon and delivered in 2 ml tubes as 100 μM solution.

PCRs were carried out in 96-well plate format. The master mix for each singleplex reaction consisted of 15 μl of AmpliTaq Gold ready to use solution (Life Technologies), 12.4 μl of RNase-free water, 2 μl each of forward and reverse 10 μM primer solution, and 50 ng of genomic DNA. For GC-rich fragments, 2.8 μl of GC-enhancer (Life Technologies) was added to the reaction. The amplification reactions were performed on a C1000 Thermo Cycler (Bio-Rad) with initial denaturation at 95° for 15 min, followed by 30 cycles of 94° for 30 s, 62° for 30 s, and 72° at 30 s. The final extension step was performed for 5 min at 72°. Following the PCR, amplicons were purified using AMPure beads. The quantification procedure of the purified fragments was performed using Quant-iT PicoGreen dsDNA. All amplicons were diluted separately to 1 × 109 molecules/μl, pooled together, and the final DNA concentration was adjusted to 1 × 107 molecules/μl. Emulsion PCR and pyrosequencing on the 454 Junior Roche Sequencer was performed in accordance with the manufacturer’s protocols.

The data analysis was performed with custom Python scripts to evaluate the quality of the data and number of reads per amplicon. Fragments that had at least 30 reads were further analyzed with the “GS Amplicon Variant Analyzer” software from Roche. Any sequence variation that was present in at least 20 % of reads was further evaluated using the Alamut v.2.2 software (Interactive Biosoftware). Amplicons that failed to achieve 30 reads were resequenced with a conventional Sanger sequencing method. In addition, all pathogenic or potentially pathogenic variants, as well as fragments with homopolymeric regions within the exons, were also validated/resequenced on ABI Prism 3130XL Genetic Analyzer platform.

Results

Sequencing analysis

The mean coverage on target was 107 reads. A total of 93.5 % of NGS fragments had at least 30-fold coverage and passed quality control. The remaining 6.5 % (374 fragments) that did not pass the quality control were resequenced with the traditional Sanger method.

Clinical and genetic findings

A total of 27 unrelated SPAST-negative HSP patients were recruited with the following characteristics: eight pure versus 19 complex phenotypes, average age at onset 27.7 ± 20.5 years (range 1–60 years), average age at examination 49.1 ± 16.3 years (range 21–72 years) and 16 male versus 11 female. The majority of patients (92.6 %) were of European descent. Twelve subjects were categorized as AD inheritance and 15 as AR or sporadic inheritance. In addition, two patients who tested negative for the AD panel were subsequently screened using the AR panel, which was also negative.

Excluded from further analysis were 17 patients identified with SPAST mutations (17/44, 38.6 %). Of these, 16 patients had been published previously (including two with sporadic inheritance) [8]. One patient (L6681) who initially tested negative for SPG4 on Sanger sequencing was subsequently found to have a SPAST mutation on the NGS AD panel.

A genetic cause for HSP was identified in 25.9 % (7/27) of the SPAST-negative sample (Tables 1, 2, 3). Of the 12 SPAST-negative patients classified with AD inheritance, seven had pure phenotypes. A clear genetic cause was identified in one patient only, a 29-year-old man who was heterozygous for a known pathogenic REEP1 mutation (c.59C > A, p.A20E) [9].

Table 1 Mutations detected in HSP patients who initially tested negative for SPAST mutations
Full size table
Table 2 Variants of uncertain pathogenicity/benign polymorphisms identified in HSP patients who initially tested negative for SPAST mutations
Full size table
Table 3 Clinical features of SPAST-negative HSP subjects identified with probable disease-causing variants
Full size table

Fifteen SPAST-negative subjects were assigned to the AR-HSP panel, of whom 13 had sporadic disease and two had an AR pattern of inheritance. Nearly all (14/15) had complex forms of HSP, but only one had an AR HSP-TCC phenotype. A genetic cause was identified in 40 % (6/15). Four patients were identified with compound heterozygous SPG7 mutations, including one novel variant (c.1727C > G, p.S576W). Additional phenotypic manifestations in patients with SPG7 mutations included cerebellar ataxia, external ophthalmoplegia and ptosis. Two patients had compound heterozygous CYP7B1 mutations. The first patient with SPG5 (L-6871) carried the known c.1456C > T, p.R486C mutation [4], as well as a novel c.650dupT, p.L217Ffs*3 frameshift mutation. The second patient (L-6688) also carried the p.R486C mutation and another novel frameshift mutation (c.1460dupT, p.L487Ffs*10). The clinical features in patients with SPG5 included cerebellar ataxia and optic atrophy.

For patients with no genetic cause detected, there were seven pure versus 13 complex phenotypes. Additional clinical features in these patients included dementia (1/20), dysarthria (2/20), deafness (1/20), cataracts (1/20), ataxia (7/20) and Parkinsonism (2/20).

Discussion

A genetic cause of HSP was identified in approximately 25 % of patients in a non-consanguineous, SPAST-negative Australian sample. In the SPAST-negative AD-HSP group, the genetic basis of HSP (a REEP1 mutation) was found in one subject only. For SPAST-negative AR-HSP, 40 % had an identifiable cause (either SPG5 or SPG7). The other known genetic causes included in the screening panels were not found in this sample. No mutations in SPG11 or ZFYVE26 (SPG15) were detected, although only a single patient had evidence of a TCC on neuroimaging. Several phenotype–genotype correlations were observed. For example, subjects with compound heterozygous CYP7B1 mutations exhibited known complicating features of SPG5, such as optic atrophy and cerebellar ataxia [10]. Patients with SPG7 also had a complex phenotype that included ataxia. A chronic progressive external ophthalmoplegia-like phenotype was seen in two of our patients with SPG7 mutations, consistent with reports suggesting ptosis and supranuclear gaze palsy are manifestations of SPG7-AR-HSP [11]. The histochemical findings of COX-negative/SDH positive fibers on muscle biopsy in SPG7 have also been reported [12]. However, as previously observed [6], it appears difficult to reliably predict the genotype according to the phenotypic manifestations.

The results of this genetic study could have been affected by case ascertainment bias, since dominant families are more likely to be referred to an adult neurogenetic clinic, whereas infantile and childhood-onset HSP are usually seen elsewhere. In 75 % of patients, no genetic cause was identified, and there are several possible explanations for this finding. Patients may have been assigned an incorrect mode of inheritance of HSP due to variability of age at onset, the frequent occurrence of de novo mutations, or reduced penetrance [10]. Some patients may have an alternative aetiological explanation for the HSP phenotype. For example, the phenotype of HSP and primary lateral sclerosis can overlap, leading to difficulties distinguishing between these two conditions [13]. Furthermore, the cause of HSP in some patients may have been one of the recently described genes not covered in our screening panels, since we only covered 19 of at least 33 AD or recessive HSP-causing genes. However, we covered the most common and well-validated genes, and interrogation of all known genetic causes of HSP is difficult, given the rapid rate of new gene discovery. Another possibility is that these patients have a hitherto unknown gene causing HSP. Whole exome sequencing provides us with an opportunity to discover new causative genes in such patients, and has been used for this purpose by other investigators [14]. Indeed, an alternative approach would be to perform whole exome sequencing in lieu of testing specific genes in patients with AR-HSP, as has been proposed by Bettencourt and colleagues [15]. The disadvantage of performing whole exome sequencing only is that the coverage of specific genes of established importance in HSP may be inadequate. For example, in our study, at least 93.5 % of the targeted regions had a coverage of at least 30-fold, compared to 87 % of targeted regions using a whole exome sequencing approach [15]. It can also be argued that the high sensitivity of the PCR amplicon-based library capture method, as used in this study, is more universally accepted compared to the variety of library capture techniques utilized for whole exome sequencing. Given the relatively high frequency of SPG4 [8], SPG5, SPG7, perhaps a more cost-effective approach would be to initially restrict genetic analysis to screening for SPAST mutations in those with AD inheritance and CYP7B1 and SPG7 in those with sporadic or AR inheritance, in the absence of additional clinical features implicating a particular genetic diagnosis (e.g. TCC suggesting SPG11 or SPG15). It is also important to note that up to 7 % of patients with sporadic inheritance will have SPG4 [16, 17] (e.g. two patients with sporadic inheritance from our previous study had SPAST mutations) [8], and so patients with sporadic HSP should also be screened for SPAST mutations.

This study identified the genetic cause in approximately 25 % of patients in this sample; this is a high proportion of cases, given that one of the most common causes of HSP (SPG4) had already been excluded. Further studies are now required to determine the most cost-effective and practical method to utilise NGS technologies in order to achieve a molecular diagnosis in HSP.

Web resources

Align GVGD, http://agvgd.iarc.fr/

dbSNP database build 138, https://www.ncbi.nlm.nih.gov/SNP/

Exome variant server database, http://evs.gs.washington.edu/EVS/

Mutation Taster, http://www.mutationtaster.org/

NCBI dbSNP, http://www.ncbi.nlm.nih.gov/SNP

Polyphen 2, http://genetics.bwh.harvard.edu/pph2/

SIFT, http://sift.jcvi.org/