Childhood amyotrophic lateral sclerosis caused by excess sphingolipid synthesis

Abstract

Amyotrophic lateral sclerosis (ALS) is a progressive, neurodegenerative disease of the lower and upper motor neurons with sporadic or hereditary occurrence. Age of onset, pattern of motor neuron degeneration and disease progression vary widely among individuals with ALS. Various cellular processes may drive ALS pathomechanisms, but a monogenic direct metabolic disturbance has not been causally linked to ALS. Here we show SPTLC1 variants that result in unrestrained sphingoid base synthesis cause a monogenic form of ALS. We identified four specific, dominantly acting SPTLC1 variants in seven families manifesting as childhood-onset ALS. These variants disrupt the normal homeostatic regulation of serine palmitoyltransferase (SPT) by ORMDL proteins, resulting in unregulated SPT activity and elevated levels of canonical SPT products. Notably, this is in contrast with SPTLC1 variants that shift SPT amino acid usage from serine to alanine, result in elevated levels of deoxysphingolipids and manifest with the alternate phenotype of hereditary sensory and autonomic neuropathy. We custom designed small interfering RNAs that selectively target the SPTLC1 ALS allele for degradation, leave the normal allele intact and normalize sphingolipid levels in vitro. The role of primary metabolic disturbances in ALS has been elusive; this study defines excess sphingolipid biosynthesis as a fundamental metabolic mechanism for motor neuron disease.

Extended Data Fig. 1 |. Sphingolipid biosynthesis pathway.

Serine palmitoyltransferase (SPT) catalyses the first and rate-limiting step in sphingolipid biosynthesis. The most abundant acyl-CoA used by SPT is palmitoyl-CoA, which after condensation with L-serine results in an 18-carbon long-chain base, 3-dehydrosphinganine. 3-dehydrosphinganine is rapidly reduced by 3-dehydrosphinganine reductase (KDSR) to form dihydrosphingosine (sphinganine). Sphinganine is acylated by a variety of ceramide synthases (CerS) to form dihydroceramides, which can be converted to ceramides by sphingolipid desaturase (DEGS). Many complex sphingolipids can be synthesized from a ceramide backbone. However, the only true exit from the pathway depends on activity of sphingosine kinase (SK) and sphingosine-1-phosphate aldolase (SGPL) to form phosphoethanolamine. SGPP = sphingosine-1-phosphate phosphatase.

Extended Data Fig. 2 |. SPTLC1 variant associated with amyotrophic lateral sclerosis and their effects on splicing.

a, Gel electrophoresis of PCR amplification of cDNA obtained from patient fibroblast cultures or whole blood. Fibroblasts were not available for c.58G>T patient. PCR primers were in exon 1 and 6 of SPTLC1 (NM_006415.4). SPTLC1 c.58G>T variant resulted in two bands with an apparent difference of ~100 bp. All other variants and controls resulted in a single band. b, Sequencing of the gel purified PCR product showed that the c.58G>T variant resulted in complete skipping of exon 2 (lower band). Neither the full-length product (upper band) nor the internally deleted transcript (lower band) contained the c.58G>T variant. However, it is possible that the missense variant exists in very small amounts and escaped amplification by PCR and sequencing. All other variants were confirmed and do not affect splicing.

Extended Data Fig. 3 |. Amyotrophic lateral sclerosis associated variant SPTLC1 cellular localization.

a, Immunostaining for SPTLC1 in patient derived fibroblast cultures shows normal subcellular localization in these cells, with an apparent distribution in the endoplasmic reticulum. Immunostaining for TDP43 also appears normal. Scale bar=40 μm b, Immunostaining for SPTLC1 in iPSC-derived human motor neurons with p.F40_S41 variant do not show any changes in its subcellular localization. Scale bar=20 μm c, Western blot of SPTLC1 in iPSC-derived human motor neurons confirms comparable expression levels in p.F40_S41 line and its isogenic control. Two independent differentiation experiments followed by western blotting were performed with similar results.

Extended Data Fig. 4 |. Induced pluripotent stem cell-derived human motor neurons (iPSC-hMN) with SPTLC1 ALS variants.

a, After differentiation by addition of doxycycline, SPTLC1 mutant iPSC-hMNs, one with heterozygous SPTLC1 p.F40_S41del variant and the other with compound heterozygous variants SPTLC1 p.F40_S41del; E2del (complete deletion of exon 2) and their isogenic control assume neuronal morphology. They express choline acetyltransferase (ChAT), Hb9 (a motor neuron specific transcription factor) as well as Tuj1 (neuronal microtubule marker). Scale bar = 20 μm b, Electron microscopy of iPSC-hMNs shows normal neuronal cell body morphology, nuclear heterochromatin, and organelles. In the processes (right), microtubules and occasional mitochondria and vesicles are noted. Scale bar left = 1 μm, middle and right = 200 nm. c, Sphingolipidomic analysis of de novo sphingolipid synthesis using isotope labelling (with 3,3-D₂ L-serine) depicted in the form a heatmap with each row representing a sphingolipid species. Row Z-scores were calculated and depicted in colour. Canonical sphingolipids ceramides and glucosyl ceramides (glucCer) are increased in p.F40_S41del iPSC-hMNs compared to their isogenic control and increase further in the double mutant iPSC-hMNs. The bar graphs show the increase in total 2+ long-chain bases (LCB), ceramides (Cer), and glucosyl ceramides (GlucCer).

Extended Data Fig. 5 |. Allele specific knockdown of variant SPTLC1 mRNA.

a, The two allele-specific siRNAs targeting the L39del and F40_S41del alleles do not reduce SPTLC1 mRNA levels in control cells at concentrations as high as 100nM. Error bars are SEM. Four replicates were performed. b, Sphingolipidomic analysis of patient derived fibroblasts with SPTLC1 ALS variants (F40_S41del and two patient lines with L39del) after treatment with 10 nM allele-specific siRNA targeting each corresponding variant were analysed and compared to control fibroblasts. Following the knockdown, the cells were labeled with D₃,¹⁵N-L-serine and D₄-L-alanine. The de novo synthesized +3 ceramide and sphingomyelin levels show a reduction in patient cells when treated with variant-specific siRNAs. Error bars are SEM. Three replicates were performed.

Fig. 1 |. SPTLC1 variants in patients with childhood-onset ALS.

a, SPTLC1 gene with ALS variants (red) in exon 2, compared with classic (blue) and atypical (green) HSAN1 variants. The numbers for ALS variants indicate the number of individuals in this report with that variant. b, Pedigrees of families with ALS and SPTLC1 variants. Solid symbols indicate ALS/motor neuron disease; a half-solid symbol indicates sensorimotor neuropathy. Empty symbols indicate unaffected family members. F, family c, Evolutionary conservation of altered SPTLC1 amino acids associated with ALS. Due to the repeat sequence, distinction of the deletion of L38 versus L39 residues in humans is not possible. However, the presence of both residues in series is highly evolutionarily conserved. AA number, amino acid number of human SPTLC1 protein. d, Muscle biopsies showing angular atrophic fibers (arrowhead) and fiber type grouping (ATPase stain) highlighting acute and chronic neurogenic atrophy, respectively. Pyknotic nuclear clumps and grouped, fascicular atrophy (arrow) both also suggest chronic neurogenic atrophy. Scale bars, 50 μm; H&E, hematoxylin and eosin; P, patient. e, Biopsy of sural nerve (a sensory nerve). Semithin toluidine blue (TB) stain (left) and electron microscopy (EM) photograph (right) showing normal density and morphology of myelinated and unmyelinated axons. There are no degenerating axonal profiles or empty Remak Schwann cells. Scale bars, 5 μm. f, T1 axial MRI images of the lower extremities in patient 1 showing global skeletal muscle atrophy and patchy, heterogeneous T1 hyperintensity consistent with denervation-related changes in the muscle.

Fig. 2 |. Elevated levels of the canonical sphingolipid products of SPT in SPTLC1-associated amyotrophic lateral sclerosis.

a, HSAN1-causing variants increase alanine usage by SPT leading to the formation of deoxysphingolipids (blue), while ALS-causing variants result in increased production of sphinganine and ceramides (red). Glycine (not depicted) usage by SPT similarly increases deoxymethylsphingolipids. CerS, ceramide synthases. b, Sphingolipidomic analysis of patient serum distinguishes HSAN1 (high 1-deoxydihydrosphingosine (dDHS) and 1-deoxymethyl DHS (dmDHS)) from patients with SPTLC1 ALS (high DHS, high ceramide). Thirteen control samples are represented in box-and-whisker plots as the median, first and third quartiles, and minimum and maximum. c, SPTLC1 knockout HEK293 cells were transfected with WT SPTLC1 and SPTLC1 with ALS- and HSAN1-associated variants. Sphingolipidomic analysis of de novo sphingolipid synthesis using isotope labeling (with D₃, ¹⁵N-l-serine and D₄-l-alanine resulting in +3 sphingolipids) shows a biochemical profile that distinguishes HSAN1 (high 1-deoxyceramides) and ALS variants (high ceramides). Three replicates for each variant were performed. d, Addition of exogenous serine (3 mM) increases canonical sphingolipid synthesis in SPTLC1 mutant iPSC-hMNs but not in controls. Error bars denote s.e.m. Four mass spectrometry measurements for each variant and condition were performed. Ordinary one-way ANOVA with Tukey’s adjustment for multiple comparisons was used for statistical comparisons.

Fig. 3 |. Homeostatic regulation of SPT mediated by ORMDL proteins.

a, The SPT multisubunit complex includes SPTLC1, SPTLC2 and a small subunit (ssSPT). ORMDL proteins bind the first transmembrane domain of SPTLC1 (TM1) and inhibit SPT activity in the presence of high ceramide (Cer) levels. SPTLC1 with ALS variants in TM1 does not bind ORMDLs efficiently, and high ceramide levels are less effective in inhibition of mutant SPT activity. Lcb, long-chain base; ER, endoplasmic reticulum. b, Cryoelectron microscopy structure of the SPT/ORMDL3/ssSPTa complex. Hereditary sensory neuropathy SPTLC1 residues (for example, C133 and V144) are in the cytosolic portion of SPT, near its active site and the pyridoxal phosphate-binding residue of SPTLC2 (K379). In contrast, ALS-associated variants are near the ORMDL3-interacting transmembrane domain (inset). c, HEK293 SPTLC1 knockout cells transfected with WT or ALS-associated SPTLC1 variants. De novo synthesized ceramides (labeled with D₃-¹⁵N-l-serine and D₄-l-alanine) were measured in the presence of increasing, exogenously added C₆-ceramide. Bar graphs show normalized +3 ceramides for each cell line. d, WT HEK293 cells were cotransfected with SPTLC2, ssSPTa, either WT or ALS variant SPTLC1 and increasing amounts of ORMDL3 plasmid. SPT activity was evaluated by measurement of sphinganine levels normalized to baseline for each cell line. Increasing ORMDL3 has a more robust inhibitory effect on sphinganine synthesis (that is, SPT activity) in WT versus ALS-associated SPTLC1 variants. Error bars are s.e.m. from five independent replicates.

Fig. 4 |. Allele-specific knockdown of transcripts containing SPTLC1 variants associated with ALS.

a, Allele-specific knockdown using siRNA. Precision-designed, mutation-specific siRNAs target the variant allele mRNA while not interfering with the WT transcript. Bar graphs show real-time reverse transcriptase PCR results from patient-derived fibroblasts treated with varying concentrations of allele-specific siRNAs. All values are reported as linear fold change compared to untreated cells in the same experiment. The levels of mutant allele transcripts L39del (yellow) and F40_S41del (green) show a dose-dependent decrease with increasing siRNA concentration. The levels of WT allele transcript in L39del cells (white) show no notable change, indicating allele specificity. In F40_S41del cells (gray), the level of total transcript including both mutant and WT alleles was measured and did not decrease by more than the expected 50% of total, akin to haploinsufficiency. Error bars are s.e.m. from four replicates. b–d, Corrective effect of allele-specific siRNA treatment on de novo synthesis of sphinganine (b), sphingosine (c) and ceramide (d) levels in patient-derived fibroblasts. Cells were treated with siRNA and labeled with 3,3-D₂ l-serine (3 mM) 24 h before analysis. Data represented by box-and-whisker plots are the median, first and third quartiles, and minimum and maximum from four replicates. Ordinary one-way ANOVA with Tukey’s adjustment for multiple comparisons was used for statistical comparisons. Unt, untreated; scr, scrambled.