Recurrent de-novo gain-of-function mutation in SPTLC2 confirms dysregulated sphingolipid production to cause juvenile amyotrophic lateral sclerosis

Abstract

Background:

Amyotrophic lateral sclerosis (ALS) leads to paralysis and death by progressive degeneration of motor neurons. Recently, ALS patients were identified with specific gain-of-function mutations in SPTLC1. SPTLC2 encodes the second catalytic subunit of the serine-palmitoyl transferase complex.

Methods:

We used the GENESIS platform to screen 700 ALS whole-genome and whole-exome data sets for variants in SPTLC2. The de-novo status was confirmed by Sanger sequencing. Sphingolipidomics were performed using high resolution mass spectrometry.

Results:

In two unrelated patients presenting with early-onset progressive proximal and distal muscle weakness, oral fasciculations, and pyramidal signs, we found the novel de-novo SPTLC2 mutation p.Met68Arg. This variant lies within a single short transmembrane domain of SPTLC2, suggesting that it renders the SPT complex irresponsive to regulation through ORMDL3. In patient plasma, both ceramides and complex sphingolipids were significantly increased, confirming this hypothesis. Accordingly, excessive sphingolipid overproduction was shown in mutant-expressing HEK cells.

Conclusions:

Specific gain-of-function mutations in both core subunits affect the homeostatic control of SPT. SPTLC2 represents a new Mendelian ALS gene, highlighting a key role of dysregulated sphingolipid synthesis in the pathogenesis of juvenile ALS. Given the direct interaction of SPTLC1 and SPTLC2, this knowledge might open new therapeutic avenues for motor neuron diseases.

Introduction

Amyotrophic lateral sclerosis (ALS) is a progressive degenerative disease of upper and lower motor neurons. Recently, pathogenic variants in SPTLC1 were associated with juvenile-onset ALS^1,2,3. SPTLC1 is an essential subunit of serine-palmitoyltransferase (SPT), conjugating L-serine and palmitoyl-CoA to form sphingoid bases - the first and rate-limiting step of sphingolipid (SL) de-novo synthesis. Forming a heteromeric complex, SPT is typically composed of the two core subunits SPTLC1 and SPTLC2 and the regulatory subunits ORMDL1–3 and ssSPTa/b⁴. Various missense mutations in SPTLC1 and SPTLC2 induce a shift in substrate specificity from L-serine to L-alanine, which forms a class of neurotoxic 1-deoxysphingolipids (1-deoxySL) that cause Hereditary Sensory and Autonomic Neuropathy type 1 (HSAN1)⁵. In contrast, ALS-associated SPTLC1 mutations alter the enzyme’s homeost tic control resulting in neurotoxic overproduction of canonical ceramides and complex SLs^1,3. All previously reported SPTLC1 ALS variants cluster in the N-terminal transmembrane domain (TMD) that is essential for interaction with the inhibitory ORMDL3 subunit. ORMDL3 likewise interacts with SPTLC2 residues, suggesting that mutations in SPTLC2 may be similarly associated with juvenile-onset ALS.

Materials and methods

Using the GENESIS database⁶, we screened 700 clinically well-documented and genetically unsolved motor neuron disease cases for variants in SPTLC2, using genome or exome sequencing. All identified coding (n = 15) and splice region (n = 10) variants in SPTLC2 were further stratified using allele frequency (MAF^gnomAD < e⁻⁷), conservation (GERP > 4), and in-silico predictions (CADD > 20). For this study, clinical HSAN cases and cases with other neurological and non-neurological phenotypes were excluded. Co-segregation studies were performed either by Sanger sequencing (patient 1) or by trio exome sequencing. Patients were diagnosed at specialized neuromuscular centers. Clinical examinations have been performed by experienced (pediatric) neurologists. Juvenile ALS was diagnosed using the revised El Escorial and re-confirmed with the new Gold Coast criteria. A detailed family history included aspects of motor neuron or other neuromuscular diseases, dementia, and metabolic disease. We obtained written informed consent from all examined individuals. The study was approved by the local ethical committees/institutional review boards (Bogazici University, Istanbul; National Institutes of Health protocol 12-N-0095) and conducted following the local and legal guidelines and in concordance with the Declaration of Helsinki. We obtained Li-heparin plasma collection tubes from probands and controls, and frozen supernatants were shipped to the Institute of Clinical Chemistry, University Hospital Zürich (Zürich, Switzerland). SPTLC1 and SPTLC2 deficient Flp-In T-Rex 293 cells were transfected with plasmids carrying different ALS- and HSAN-associated mutations and WT. Sphingolipid profiles were measured by liquid chromatography mass-spectrometry³. For an in-depth description of methods, see the supplementary methods.

Results

In a male patient in his twenties of Turkish-Bulgarian origin, whole-genome sequencing (WGS) revealed a heterozygous missense variant in SPTLC2: ENST000000216484.2/ NM_004863.4; c.203T>G, p.Met68Arg. Sanger sequencing and co-segregtion studies confirmed the varient’s de-novo occurrence (Fig. 1A). The patient’s motor milestones were delayed, and he started walking at 26 months of age. In early childhood, he displayed toe walking and had difficulty rising from the floor. Hand tremor was reported at the age of 3 years. At 15 years, he developed swallowing difficulties and hoarseness of voice. At the age of 18 years, he was not able to walk without assistance and presented with proximal and distal muscle weakness, fasciculations, abolished distal tendon reflexes and positive Babinski’s signs, pes cavus, high arched palate, tongue atrophy, hypophonia, gynecomastia, and hypogonadism. Superficial and deep sensation were normal. The patient underwent tracheostomy at the age of 22 years. At his most recent examination, he was bed-ridden and only able to use two fingers on both hands (Fig. 1B; suppl. Fig. 1). He was diagnosed with juvenile ALS at the age of 14 years. There is no family history of motor neuron or metabolic disease, and genetic screening for spinobulbar muscle atrophy was negative. Furthermore, tandem repeat expansions in C9ORF72 and ATXN2 have been excluded.

Figure 1: Genotype, phenotype, and structural effects of SPTLC2 p.Met68Arg.

A) Pedigree illustrating de-novo occurrence of the SPTLC2 p.Met68Arg variant in family 1.

B and C) Clinical pictures showing tongue atrophy in patients 1(B) and 2(C).

D) Tertiary model of the SPT complex consisting of SPTLC1 (purple), SPTLC2 (green), ORMDL3 (yellow), and SSPTa (orange), localized to the endoplasmic reticulum (ER) membrane. Magnification shows ORMDL3 in relation to SPTLC1 and SPTLC2 transmembrane domains (TMDs) in blue and pink, with ALS-associated mutations highlighted.

Through our collaborative network, we identified an African American female with slurred speech at age five years. Following normal motor milestones and independent walking by 13 months of age, she had developed scoliosis and was found to have muscle weakness, reduced muscle bulk, atrophy, and fasciculations of the tongue by ten years of age. An electromyography performed at age 12 years revealed diffuse neurogenic changes. At the age of 18 years, the patient was unable to rise from the floor and showed progressive bulbar and respiratory weakness requiring intermittent non-invasive ventilation. Neurological examination showed perioral fasciculations (figure 1C, suppl. video 2) and proximal muscle weakness (hip flexors: 3/5 on MRC scale) with a positive Gower’s’ sign. Deep tendon reflexes were brisk at knees and ankles, nd pyramidal signs were positive without sensory deficits. Muscle ultrasound showed increased muscle echogenicity (Heckmatt 1–2) and spontaneous fasciculations in multiple proximal and distal muscles. Trio whole-exome-sequencing, confirmed by Sanger sequencing, revealed the de-novo SPTLC2 variant p.Met68Arg.

This variant was not present in large control datasets (gnomAD: 125,748 WES and 15,708 WGS; GENESIS: 16,950 WES and WGS), nor in any other out of 1,800 ALS and motor neuron disease families in GENESIS and 2,800 cases in ALSdb. It affects a highly conserved amino acid position (GERP score 4.88, suppl. fig. 2) within the single short transmembrane domain (amino acid positions 67–87) of SPTLC2. The recently published CryoEM model of the SPT-ORMDL3 complex⁴ showed that Met68 is a key residue for the interaction of SPTLC2 with ORMDL3 (Fig. 1D). While methionine is hydrophobic, arginine has an additional guanidine group that is protonated at neutral pH.

We hypothesized that impaired ORMDL3 interaction might consequently result in an altered homeostatic control and increased SL formation. Indeed, a principal component analysis based on patient plasma SL profiles showed a separation of the two SPTLC2p.Met68Arg patients from unaffected family members, unrelated healthy controls, and a patient with HSAN1 (SPTLC1-p.Cys133Tyr) (Fig. 2A). Despite level differences in individual sphingolipid species arising from variable enzyme activities downstream of SPT, both p.Met68Arg carriers showed a significant increase in dihydro-sphingolipids (dhSL) such as dihydro-ceramides (dhCer) and dihydro-sphingomyelins (dhSM) (Supplementary figures 4, 5).

Figure 2: Lipid signature of SPTLC2 p.Met68Arg in patient plasma and mutant-expressing HEK cells.

A) Principal component analysis (PCA) of plasma sphingolipid profiles revealed separate clustering for the two ALS patients carrying the SPTLC2 Met68Arg mutation, compared to samples from an HSAN1 patient (SPTLC1 Cys133Tyr), patient 1’s unaffected parents, and unrelated controls.

B) De-novo formed SLs in SPTLC1- or SPTLC2-deficient cells transfected with either SPTLC2-WT or SPTLC2-Met68Arg. Cells were cultured in presence of isotope-labelled, D₃¹⁵N -L-serine and D₄-L-alanine for 16h. SLs were analyzed by LC-MS. Compared to WT, SL formation was increased in Met68Arg-expressing cells. Data are represented as mean ± SD, n=3, two-way ANOVA with Dunnett’s adjustment for multiple comparisons, **** p <0.0001)

C) Clustering heat map comparing the SL profile in SPTLC1 or SPTLC2 deficient cells, expressing either the empty vector (vec), respective wild type (wt) proteins, the HSAN1 variant SPTLC1-Cys133Trp, previously reported SPTLC1-ALS variants (SPTLC1-Ala20Ser, Leu39del, Ser40Phe41del, ex2del) and the newly identified SPTLC2 p.Met68Arg variant. Ceramides and sphingomyelins (SM), especially those with C₂₀-C₂₂ N-acyl chains, were significantly increased in SPTLC2 p.Met68Arg expressing cells.

To confirm the stimulating effect of the SPTLC2-Met68Arg mutation, we expressed it together with the SPTLC2 WT in an SPTLC2-deficient HEK293 cell line cultured in serine-deficient medium supplemented with isotope labelled (¹⁵N-D₃)-L-serine and (D₄)-L-alanine for 16h³. De-novo SL formation was significantly increased in p.Met68Arg-expressing cells compared to WT controls (Fig. 2B) with most pronounced differences seen for dhCer and dhSM (d18:0) as well as Cer and SM (d18:1) (Fig. 2B). Although being slightly increased overall, 1-deoxySL formation was significantly less compared to HSAN1 mutant (SPTLC1 p.Cys133Tyr) expressing cells (Fig. 2B). Compared to previously reported SPTLC1-ALS variants^1,3, a heat map showed an even higher SL de-novo formation for the SPTLC2 p.Met68Arg variant than the one observed for the SPTLC1-ALS variants (Fig. 2C). Similar to the plasma profile, we saw a significant increase in SL species with a long chain N-acyl (C_20:0, C_22:0) in the ALS-mutant expressing cells, with the most significant and consistent increase seen for dihydroSL species (dhCer, dhSM). While total Cer levels were unaltered and SM even reduced in patient plasma, the mutant-expressing cells showed a significant increase in both, dhSL and SL species. The conversion of dhCer to Cer is mediated by DEGS1, desaturase that introduces the characteristic Δ4E double bond into the sphingoid base backbone of Cer. The observed increase in dhSL species indicates that DEGS1 activity becomes limiting under conditions of the overactive SPT enzyme. A similar change in the SL pattern was described previously in conditions of DEGS1 deficiency, which is characterized by significantly increased dhSLs causing leukodystrophy⁷.

Discussion

It is striking that different gain-of-function mutations in both SPT subunits cause two opposing phenotypes: late-onset sensory neuropathy (HSAN1) or early-onset motor neuron disease (juvenile ALS). The former primarily involves peripheral nerves with emphasis on sensory and autonomic fibers, although when progressing, lower motor neurons become involved as well. The latter affects upper and lower motor neurons, rarely involving sensory systems. While SPT-associated HSAN commonly occurs later in life (3^rd to 5^th decade), mutations in SPTLC1 and now in SPTLC2 as well have exclusively been identified in juvenile-onset ALS patients. Our observations showed that both phenotypes correlated strictly with specific changes in the sphingolipid profile. Interestingly, two previously reported SPTLC1 variants, p.Ser331Phe and p.Ser331Tyr, are associated with a mixed sensory and motor phenotype, mimicking an atypical form of HSAN1 with additional ALS-like features⁸. This is mirrored by a mixed lipid signature with both, increased canonical and 1-deoxySL^3,8. Encoding another regulator of the SPT complex, two SPTSSA variants have been associated with increased overall sphingolipid synthesis and a congenital-onset syndrome comprising spastic paraplegia, developmental delay, and electroencephalographic abnormalities⁹. Interestingly, these children do not display any signs of polyneuropathy or lower motor neuron involvement by NCS and EMG. Despite originating from the same pathway of sphingolipid de-novo synthesis, differences in onset and tissue selectivity between our juvenile ALS patients and this more complex neurodevelopmental syndrome will have to be investigated further.

In conclusion, SPTLC2 p.Met68Arg occurred de-novo (Fig. 1A) in two independent patients with a similar phenotype (Fig. 1B, C). The exchange results in an increased ceramide and SL formation likely due to an impaired homeostatic control by ORMDL3. The two patients’ phenotype was highly consistent with the one reported previously in association with ALS-causing SPTLC1 mutations, and biochemical data closely match previous findings from pathogenic SPTLC1-ALS variants (Fig. 2A)). In vitro studies reproduced the specific change in sphingolipids clearly differentiating ALS- from HSAN-associated mutations (Fig. 2B). The direct association of juvenile ALS with Mendelian mutations in SPTLC1 and SPTLC2 opens new avenues for a gene-based therapy in these patients.

Supplementary Material

Supp2

Supplementary figure 1: Patient 1 phenotype

The patient is male and in his twenties, diagnosed with early-onset amyotrophic lateral sclerosis (ALS). Panel A shows him with a tracheostoma, leaning on a pillow to compensate for proximal muscle weakness. The picture further shows atrophies of the upper arm muscles (especially deltoid muscle) as well as pseudo-contracted fingers due to profound muscle weakness. Panel B shows talipes equinovarus and lower leg atrophies. Panel C illustrates that the patient is mostly bed-ridden, but can partially breathe on a T-piece.

Supplementary video 2: Patient 2 phenotype

Patient 2 is in her late teens with childhood-onset ALS. The video shows pronounced tongue atrophy and fibrillations.

Supplementary figure 3: Evolutionary conservation of Met68 among chordates and patient mutation

Position Met68 (SPTLC2 c.203T>G, p.Met68Arg, Chr14:78063653 (hg38), transcript: ENST000000216484.2/NM_004863.4.) is highly conserved throughout species, GERP score: 4.88.

Supplementary figure 4: Plasma levels of sphingolipid subclasses in SPTLC2-ALS patients

Total plasma levels of sphingolipids in individual patients and controls. SPTLC2 p.M68R patients show significant increase in the saturated dihydroceramides and dihydrosphingomyelins. In the two SPTLC2 p.M68R patients, 1-deoxysphingolipids that are elevated in the hereditary sensory neuropathy type 1 (HSAN1) patient with a SPTLC1 p.C133Y mutation remain comparable to controls. Controls (ctrl 1 and 2) represent related controls for patient 1 and controls 1 to 3 are unrelated healthy controls. Data are represented as mean ± SD, n=4, one-way ANOVA with Tukey’s’s adjustment for multiple comparisons, ns, not significant, * p<0.05, ** p<0.01, *** p<0.001, **** p<0.0001.

Although carrying the same variant, the increase in dhSL was more pronounced in patient 2 than in patient 1 (Supplementary Fig. 4). In contrast, total plasma ceramides (Cer) were not altered, while total sphingomyelins (SM) were reduced in both patients (Supplementary Fig.4). In contrast, 1-deoxySL were elevated in HSAN1 patient plasma (SPTLC1 p.Cys133Trp), but not in the ALS plasma (Supplementary Fig. 4). On the species level, we observed primarily an increase in SLs with long chain N-acyls (C₁₈–C₂₂), which are normally minor or even absent in human plasma (Supplementary Fig. 5).

Supplementary figure 5: Sphingolipid species profiles in SPTLC2-ALS patients

Heat map clustering panel illustrates the plasma sphingolipid species distribution in SPTLC2 p.M68R patients compared to related (Pat. 1, control 1 and 2) and unrelated (Control 1 to 3) healthy controls as well as a Hereditary sensory neuropathy type 1 (HSAN1) patient with a SPTLC1 p.C133Y mutation. Abbreviations: Cer, ceramide; doxCer, 1-deoxyceramides, HexCer, Hexosylceramide; SM, sphingomyelin. Shown are results for individual sphingolipid species from samples that were extracted and quantified independently. Log transformed (base 10) data with Euclidean distance measure were plotted using MetaboAnalyst Suite 5.0.

Supplementary methods