Herein, we identified SLC38A6 variants in familial ET patients and sporadically affected individuals and revealed the relationships between cerebellar abnormalities and SNAT6 dysfunction. We generated Slc38a6 deletion mouse models to mimic the tremors and pathological changes of ET, thereby revealing the underlying molecular mechanisms involved. This study provides novel molecular insights into the mechanisms underlying ET and promotes the development of new therapeutic drugs and treatments.
We collected and characterized data from 773 Chinese families and 640 sporadic cases of ET from mainland China. Among these, there was a five-generation large Chinese family with 61 members (Family 1) from Tianjin, including 27 patients (22 males and 5 females), who exhibited an autosomal dominant inheritance pattern (Fig. 1a). The proband V:6 developed posture and kinetic tremors affecting both upper limbs at the age of 20 and was assessed as 11 and 15.5, respectively, on the basis of the Essential Tremor Rating Assessment Scale (TETRAS)-I and TETRAS-II (Supplementary Table 1). The tremors of proband V:6 could be exacerbated by emotional stress, fatigue, or performing fine movements and relieved by drinking alcohol. The affected individuals in Family 1 exhibited action tremors in both upper limbs, such as the mother of the proband (IV:6) (Supplementary Video 1), whereas III:5, III:13 and IV:4 developed head and voice tremors. Individuals with cognitive impairment, dystonia, cerebellar ataxia, pyramidal signs and parkinsonism were excluded from Family 1 on the basis of neurological examinations. In addition, the brain MRI/CT scans of these patients were normal (Supplementary Table 1).
To identify genetic variants associated with ET, WES ( >100-fold average coverage) was performed on three patients from Family 1, including the proband (V:6) and two other affected members (III:13, IV:17) (Fig. 1a). First, we screened 21 known ET candidate genes, including DRD3, HS1BP3, LINGO1, FUS, SLC1A2, TENM4, HTRA2, SCN4A, SORT1, NOS3, KCNS2, HAPLN4, USP46, CACNA1C, STK32B, PPARGC1A, CTNNA3, LRRK2, SNCA, TUB and TCP10L, but we did not find any mutations in these genes that segregated with the phenotype (excluding SNPs with a high population frequency >5% based on gnomAD), nor did we identify repeated expansions of NOTCH2NLC GGC via tandem repeat amplification analysis. We thus hypothesized the presence of novel pathogenic genes in Family 1. All variants were screened via the following criteria: (1) variants with a frequency less than 5% according to gnomAD; (2) missense variants that were predicted to be deleterious by at least 2 out of the 3 programs (Polyphen 2, SIFT, and CADD) or that are nonsense, splicing or frame-shift variants; and (3) considering the family history (suggestive of autosomal dominant inheritance), heterozygous variants need to be shared by three affected individuals. Six candidate rare variants in four genes (DPP6, KMT5A, SLC38A6 and GGT1) met these criteria (Supplementary Table 2). Sanger sequencing and familial segregation testing of the available family members revealed that only the heterozygous variant c.842 T > C/p.M281T (rs117560154) and c.952 G > A/p.G318S (rs61050538) of SLC38A6 (NM_001172702) (Fig. 1b) cosegregated with family 1, with frequencies of 0.0015 and 0.0021, respectively, in gnomAD (Genome Aggregation Database) for the total population and 0.0227 and 0.0278, respectively, in gnomAD for East Asians (Table 1). In addition, IV:10 is homozygous for the variant c.842 T > C/p.M281T and c.952 G > A/p.G318S of SLC38A6 and suffered more severe tremor, whereas her mother (III:4) might also carry these two variants. II:8, IV:2, and IV:14 carried both of the variants but were asymptomatic, suggesting incomplete penetrance or a variable age of onset. IV:12 did not carry any of the variants of SLC38A6 but resulted in mild tremor. He was diagnosed with a phenocopy because his symptoms did not worsen with age, and his offspring were normal.
To identify variants in SLC38A6, WGS was conducted on the remaining 1 412 ET patients in this study, including 772 familial ET probands (family history: similar manifestations in immediate family members or siblings) and 640 sporadic cases (family history: no similar manifestations in immediate family members or siblings). Variants were selected on the basis of the following criteria: (1) a frequency less than 5% on the basis of gnomAD and (2) missense variants predicted to be deleterious by at least one of three tools (Polyphen-2, SIFT, and CADD) or nonsense, splicing, or frameshift variants. In total, 118 (8.35%) patients were found to carry an SLC38A6 gene variant, including 71 (9.18%) familial ET probands and 47 (7.34%) sporadic cases (Supplementary Notes, Supplementary Tables 3 and 4). The clinical characteristics of these ET patients are detailed in Supplementary Table 5. As a result, 15 different SLC38A6 variants were identified (Supplementary Fig. 1a), all of which were confirmed through Sanger sequencing (Supplementary Fig. 1b; primers for SLC38A6 are shown in Supplementary Table 6). Subsequent follow-up and Sanger sequencing further revealed cosegregation of the SLC38A6 genotype with clinical phenotypes in multiple families (Fig. 1c and Supplementary Table 7).
To investigate the effects of SLC38A6 variants, we selected three variants (c.323 A > T/p.Y108F, 23.7%; c.842 T > C/p.M281T, 33.1%; c.952 G > A/p.G318S, 37.3%), which were the most common of the 15 different types (Table 1) and were highly conserved (Supplementary Fig. 2). Flag-tagged wild-type (WT) human SLC38A6 and its variants (at the putative extracellular C-terminus) were expressed in HeLa cells (Fig. 1d), and the unpermeabilized cells were immunostained with anti-Flag antibodies to examine the localization and expression of SLC38A6. We found no significant difference in either the localization or expression level between WT SNAT6 and SNAT6 variants (Fig. 1d, e).
SNAT6 is a member of the SLC38 family of sodium-coupled neutral amino acid transporters. To determine the substrate of SNAT6 and whether the identified SLC38A6 variants impair the transport function of SNAT6, we individually expressed WT SLC38A6 and three variants (encoding Y108F, M281T and G318S of SNAT6) in HeLa cells, which presented low constitutive SNAT6 expression and low baseline SNAT6 activity (see https://www.proteinatlas.org/). The transfected HeLa cells were incubated with seven types of H-labeled amino acids (L-Ser, L-Pro, L-Met, L-Gln, L-Glu, L-Asp or L-Arg). Compared with the blank control, WT-SNAT6 transfection resulted in specific uptake of L-Gln and L-Arg (Fig. 1f). Notably, transfection with SNAT6 variants resulted in significantly reduced uptake of L-Arg but similar uptake of L-Gln to that of WT-SNAT6 (mean uptake of L-Arg: Mock, 5573.3 ± 512.6; WT, 21213.7 ± 1543.3; Y108F, 10735.0 ± 2281.3, p = 0.015; M281T, 9899.0 ± 833.9, p = 0.007; G318S, 8937.3 ± 513.3, p = 0.003) (Fig. 1f). These results suggest that impaired transport of L-Arg by SNAT6 variants may be the cause of ET.
The above results suggest that loss-of-function mutations in SLC38A6 might be a cause of ET. The SNAT6 protein is highly conserved between humans (NP_722518.2) and mice (NP_001032806.2), with 84% (384/457) amino acid identity (Supplementary Fig. 3). We thus generated Slc38a6 mice via a "knockout-first" strategy (Supplementary Fig. 4a). The primary cerebellar neurons of Slc38a6 and Slc38a6 mice were cultured, and their uptake of L-Arg and L-Gln was measured. We found that the deletion of SNAT6 indeed affected the uptake of L-Arg but not L-Gln (Supplementary Fig. 4b), which is consistent with the in vitro results.
To assess the morphology of basket cell axons surrounding PCs in Slc38a6 and Slc38a6 mice, Bielschowsky silver staining was performed (Fig. 3a). We classified hairy baskets into the following categories: low (1, minimum number of axonal collaterals), intermediate (2, moderate number of axonal collaterals) and high (3, numerous axonal collaterals). By semiquantitative rating of the basket cell plexus, we found that the average rating of Slc38a6 and Slc38a6 mice was not significantly different from that of 2- to 9-month-old mice and was significantly different at 12.5 months of age (Slc38a6: 1.33 ± 0.21; Slc38a6: 2.17 ± 0.31, p = 0.029) (Supplementary Table 8). To further investigate alterations in the processes of the basket cell plexus, we used another quantitative analysis. The axonal collaterals surrounding visible PCs in Slc38a6 and Slc38a6 mice were classified into four degrees: 0 (no detectable processes), 1 (thin processes), 2 (moderate processes) and 3 (dense twining processes) (Supplementary Fig. 7a‒d). We detected an increased number of basket cell plexuses at degrees 1 to 3 in Slc38a6 mice compared with Slc38a6 mice, and this difference became more significant with age (Supplementary Fig. 7e‒g).
In the cerebellum, CFs are critical for the regulation of PC physiology and motor behavior. To label the CF synapses, we performed IHC staining for vesicular glutamate transporter type 2 (vGluT2), a specific marker for CF synapses, on the cerebellar sections. We found that there were no significant difference between Slc38a6 mice and Slc38a6 mice at younger ages (from 2‒6 months). Compared with Slc38a6, the CF‒PC synaptic density was significantly lower in Slc38a6 mice at 9 and 12.5 months (Fig. 3b, c) (at 9 months, Slc38a6: 24.1 ± 0.4, Slc38a6: 22.8 ± 0.2, p = 0.042; at 12.5 months, Slc38a6: 24.2 ± 0.3, Slc38a6: 19.5 ± 0.5, p < 0.0001). We then examined the distribution of CF‒PC synapses by quantifying the number of vGluT2 puncta in the outer 20% of the molecular layer, which is the parallel fiber (PF) region, and found more extended CF synapses at all ages, with a significantly greater number at 9 and 12.5 months of age in Slc38a6 mice than in Slc38a6 mice (at 9 months, Slc38a6: 69.2 ± 4.2, Slc38a6: 119.5 ± 9.2, p < 0.0001; at 12.5 months, Slc38a6: 62.0 ± 6.1, Slc38a6: 165.5 ± 9.4, p < 0.0001) (Fig. 3d, e).
All of the morphological changes described above in the mutant mice were similar to those reported in cerebellar samples from ET patients, suggesting that SNAT6 dysfunction causes pathological changes in ET.
To explore the role of SNAT6 in the cerebellum, we performed IHC against SNAT6 and found that it was widely expressed throughout the mouse brain but enriched in the PC layer of the cerebellum (Supplementary Fig. 8a), which was consistent with the expression pattern revealed by in situ hybridization. To examine the potential role of PCs in mediating tremor, we specifically deleted Slc38a6 in the PCs of mice (Slc38a6) (Supplementary Fig. 8b), which also exhibited tremors that worsened with age (Supplementary Video 7), similar to those in Slc38a6 mice.
Moreover, we found that, compared with L7-Cre control mice, Slc38a6 mice presented significant PC loss at 9 months of age, and the difference became more significant at 12.5 months of age (Fig. 3f, g). Abnormal PC axon morphology and reduced PC dendrites were observed in 12.5-month-old Slc38a6 mice (Supplementary Fig. 8c, d). Slc38a6 mice presented significantly increased "hairy" basket coverage of the PC soma at 12.5 months of age (Fig. 3h; Supplementary Table 9). Compared with control mice, Slc38a6 mice at 9 and 12.5 months of age presented significantly lower CF‒PC synaptic density and more CF synapses extending into the PF territory (Fig. 3i‒l). Therefore, dysfunction of SNAT6 on PCs might play a role in the pathogenesis of ET.
Slc38a6 and Slc38a6 mice developed tremors at an early age without any obvious morphological alterations in the tissues or cells of the cerebellum. We therefore speculated that dysfunction of SNAT6 might cause significant functional changes in PCs that precede morphological changes. We first analyzed the neuronal excitability of cerebellar PCs in 2-month-old Slc38a6 and Slc38a6 mice. Compared with that in Slc38a6 mice, the number of action potentials (APs) evoked by depolarizing currents in PCs from Slc38a6 mice was significantly lower, especially for larger currents (Fig. 4a, b). We also measured the threshold, amplitude and half-width of the first AP in response to depolarizing currents but did not find significant differences (Fig. 4c‒e), suggesting normal expression levels of voltage-gated sodium and potassium channels. To examine whether there was any alteration in the synaptic input from basket cells, spontaneous inhibitory postsynaptic currents (sIPSCs) were recorded (Fig. 4f). We found that both the frequency and amplitude of sIPSCs were significantly greater in Slc38a6 mice than in WT mice (Fig. 4g, h). This could be attributed to the increased "hairy basket" around the PC soma in Slc38a6 mice (Fig. 3a).
Similar electrophysiological changes, including impaired neuronal excitability (Fig. 4i‒m) and an increased mean frequency of sIPSCs (Fig. 4n, o), were also observed in Slc38a6 mice, confirming the important role of SNAT6 in postnatal cerebellar PCs. However, we found that the average amplitude of sIPSCs in Slc38a6 mice was not significantly different from that in WT mice (Fig. 4p), whereas a markedly increased amplitude was found in Slc38a6 mice.
To explore the mechanism of PC alterations caused by SNAT6 deficiency at the young stage, we performed proteomic analysis on cerebellums from Slc38a6 and Slc38a6 mice via 4D-data-independent acquisition (DIA) mass spectrometry. A total of 7,945 proteins were identified, and we identified 128 differentially expressed proteins (DEPs) using an FDR-adjusted P value < 0.05 cutoff and a fold-change threshold of 1.1 (Fig. 5a, Supplementary Table 10). KEGG and GO enrichment analyses revealed that DEPs in the Slc38a6 and Slc38a6 cerebellums were involved mainly in ferroptosis and the mitochondrial, neuronal and synaptic pathways (Fig. 5b, c). To assess ferroptosis in the Slc38a6 cerebellum, we measured the expression of the ferroptosis markers acyl-CoA synthetase long-chain family member 4 (ACSL4) and transferrin receptor (TFRC) and detected elevated levels of ACSL4 and TFRC in the Slc38a6 cerebellum compared with those in the Slc38a6 cerebellum (Fig. 5d‒f). The binding of extracellular Fe to transferrin is known to be transported into the cell by TFRC. Prussian blue staining for total iron content (Fig. 5g) and an iron assay kit for Fe, which is known to be involved in ferroptosis, revealed that the iron content of the cerebellum was significantly greater in Slc38a6 mice than in Slc38a6 mice (Fig. 5h). Further Western blotting revealed that the level of ATF3 increased significantly in Slc38a6 mice (Fig. 5i).
Increasing evidence suggests that mitochondria play a significant role in the execution of ferroptosis, and our proteomic analysis revealed that numerous DEPs were enriched in mitochondrial and metabolic pathways. We performed TEM analyses on cerebellar sections from 2-month-old Slc38a6 and Slc38a6 mice to examine whether there were ultrastructural alterations (Supplementary Fig. 9a). We detected several highly elongated mitochondria in Slc38a6 PCs, which exhibited end‒to‒end connections, resulting in a doughnut-like morphology (Supplementary Fig. 9b), and the number of mitochondria and volume density, as well as the number of mitochondrial cristae, were significantly reduced in Slc38a6 PCs (Supplementary Fig. 9c‒e). However, there was no change in the aspect ratio of the mitochondria or the length and width of the mitochondrial cristae (Supplementary Fig. 9f‒h). The changes in the morphology and number of mitochondria may be related to the impaired energy supply, leading to reduced neuronal firing in Slc38a6 mice (Fig. 4b).
In summary, SNAT6 deficiency in the cerebellum promoted ferroptosis via ATF3 activation, contributing to abnormal metabolic and synaptic pathways in Slc38a6 mice.