The underlying cause of phenotype variation from the same allele remains largely unknown in most cases when a particular genotype is inherited. Emerging evidence indicate that modifier genes may contribute to phenotypic variations [32]. For example, patients with thalassemia, a disorder caused by defective β-globin synthesis, have diverse clinical characteristics and variable expressivity. A number of factors underlie this phenotypic diversity, including the involvement of numerous modifier genes at other genetic loci that affect the production of β-globin [33]. Similarly, DYT1 dystonia patients have a wide spectrum of symptom severity, which reflects the incomplete penetrance of the pathogenic ΔE mutation and the variable expressivity of the disease. For most diseases, variable expressivity of the disease phenotype is the norm among individuals who carry the same disease-causing allele or alleles [34], despite the causes are not always being clear.
In this work, we describe the identification of 264 variants in 195 genes that are associated with DYT1 dystonia. Below, we will discuss the potential implications of our results on our understanding of the pathogenesis and pathophysiology of DYT1 dystonia. Specifically, we will explore the connections between the DYT1 variants identified here and the following established cellular functions of torsinA: cytoskeletal regulation, endoplasmic reticulum stress, and lipid metabolism. In addition, we will examine the relationship revealed between DYT1 dystonia and the neuromuscular and neuropschiatric disorders linked with the genes in which we identified DYT1 dystonia-associated genomic variants.
DYT1 variants and the cytoskeleton
Of the 195 genes that we identified as harboring 264 DYT1 variants, 34 genes encode proteins that constitute or associate with the cytoskeleton (Table 3). Specifically, we found a total of 23 DYT1 variants in 18 genes that encode proteins involved in the function of the microtubule cytoskeleton. We also found 12 DYT1 variants in ten genes encoding actin cytoskeleton-associated proteins as well as ten variants in six genes encoding intermediate filament cytoskeleton-associated proteins. Moreover, seven of the 34 genes described above were found to harbor at least two DYT1 variants, including CCDC74B, DYNC2H1, KRT6A, KRT6B, LIMCH1, NRAP, and TUBA3E.
The identification of DYT1 variants in genes encoding proteins related to cytoskeletal function is consistent with the emerging view of torsinA as a critical regulator of cellular mechanics. Since its discovery in 1997, torsinA function has been implicated in the regulation of cytoskeletal dynamics and organization [35]. The first evidence to suggest that torsinA might be involved in cytoskeletal regulation was the finding that the nematode torsinA protein OOC-5 was required for the rotation of the nuclear-centrosome complex during early embryogenesis [36, 37]. In addition, the fruit fly torsinA protein torp4a/dTorsin was implicated in the regulation of the actin cytoskeleton [38]. Furthermore, the over-expression of a torsinA construct containing the ΔE mutation was shown to inhibit neurite extension in human neuroblastoma cells and to increase the density of vimentin intermediate filaments around the nucleus [39]. The relationship between torsinA and the cytoskeleton is further strengthened by reports of the impaired migration of dorsal forebrain neurons and fibroblasts from torsinA-knockout mice as well as DYT1 dystonia patient-derived fibroblasts [29, 40, 41].
More recently, torsinA was identified as a key regulator of the mechanical integration of the nucleus and the cytoskeleton via the conserved nuclear envelope-spanning linker of nucleoskeleton and cytoskeleton (LINC) complex [28-30]. The core of LINC complexes is formed by the transluminal interaction between the outer and inner nuclear membrane Klarischt/ANC-1/SYNE homology (KASH) and Sad1/UNC-84 (SUN) proteins, respectively [42]. KASH proteins interact with the cytoskeleton and signaling proteins within the cytoplasm [43], whereas SUN proteins interact with chromatin, other inner nuclear membrane proteins, and the nuclear lamina within the nucleoplasm [44].
While the precise mechanism of torsinA-mediated LINC complex regulation remains unclear, torsinA interacts with the luminal domains of both KASH and SUN proteins [29, 45]. The ability of torsinA to interact with KASH and SUN proteins is thought to promote the disassembly of LINC complexes given the fact that most AAA+ proteins act as molecular chaperones that disassemble protein complexes [46, 47]. This hypothesis is supported by the finding that torsinA loss elevates LINC complex levels in the mouse brain, which impairs brain morphogenesis [48]. More recently, fibroblasts isolated from DYT1 dystonia patients were shown to have increased deformability similar to that of fibroblasts harvested from mice lacking the two major SUN proteins SUN1 and SUN2 [49].
DYT1 dystonia patient-derived fibroblasts were also shown to have increased susceptibility to damage by mechanical forces [49] strongly suggests that cellular mechanics may impact the pathogenesis and/or pathophysiology of DYT1 dystonia. All cells, including neurons, adapt their mechanical properties by converting extracellular mechanical stimuli into biochemical signals and altered gene expression through the process of mechanotransduction [50, 51]. Since mechanotransduction instructs neuronal differentiation, proliferation, and survival [52, 53], it is possible that defective mechanotransduction of neurons in the developing brain may contribute to the pathogenesis and/or pathophysiology of DYT1 dystonia. Based on the information provided above, it is intriguing that we identified DYT1 variants in the KASH protein nesprin-2-encoding SYNE2 gene and the NUP58 gene, which encodes the nuclear pore complex protein nup58 (Table 3 and Figure 2). In the future, it will be interesting to test if the DYT1 variants found in SYNE2 and NUP58 negatively impact LINC complex-dependent nuclear-cytoskeletal coupling and/or mechanotransduction.
It is tempting to speculate that the impairment of the microtubule cytoskeleton is particularly relevant to dystonia pathogenesis given the enrichment of DYT variants that we found in genes that encode microtubule-associated proteins. Microtubules are fundamentally important for the structure and function of neurons, which are some of the most highly polarized cells in the human body [54]. Microtubules establish the polarized architecture of neurons and serve as tracks for microtubule motor proteins as they carry proteins and lipids to where they are needed for proper neuronal function. Thus, defects in microtubule dynamics and organization underly a wide array of neurological and neuropsychiatric disorders [55-57].
Consistent with our identification of 4 DYT1 variants in the TUBA3E, which encodes the protein α-tubulin-3E, mutations in the β-tubulin-4A-encoding TUBB4A gene cause another hereditary dystonia, Whispering dysphonia or DYT4 dystonia [58, 59]. These mutations result in the formation of disorganized microtubule networks and the impaired growth of neuronal processes similar to the clinical phenotypes observed in DYT4 dystonia patients [60, 61]. Future experiments designed to test the impact of the DYT1 variants in TUBA3E on the organization and function of neuronal microtubules will help elucidate the role of the microtubule cytoskeleton to the manifestation of DYT1 dystonia.
DYT1 variants in association with protein synthesis and transport and ER homeostasis
Accumulating evidence indicate a role of TOR1A in the cellular protein quality control system in which TOR1A could be both substrate and effector [18]. In the 264 genome variants, we observed six variants in five genes, CHGB, DOP1B, MTMR6, P2RY13 and PPP1R15A, that are annotated with protein synthesis and transport functions (Table 4). Notably, CHGB and PPP1R15A has also been linked to endoplasmic reticulum stress [62-64]. These findings support the previously proposed hypothesis that elevated levels of endoplasmic reticulum stress contributes to DYT1 dystonia pathogenesis [65-73].
TorsinA functions to protect against insults from protein aggregates in the neural system [66]. Protein aggregates are products of protein misfolding commonly seen in neurodegenerative diseases such as Alzheimer's disease, Parkinson's disease, amyotrophic lateral sclerosis and prion disease, which triggers endoplasmic reticulum stress response [74, 75]. In the TOR1A ΔE mutation background, we identified six candidate modifier genome variants in five genes that have known functions in endoplasmic reticulum for protein post translational modification, protein translocation and endoplasmic reticulum stress response (Table 4). Among them, DOP1B has neurological roles in both human and mice [76, 77]. Whether DOP1B’s endoplasmic reticulum cellular function has a causal effect on its neurological role remains to be investigated. Collectively, our data provide clinical indications of candidate genes and genome variants for further investigation on the underlying mechanisms of TOR1A dependent ER dysfunction in DYT1 dystonia.
DYT1 variants and lipid metabolism
TOR1A also has a pivotal role in lipid metabolism as demonstrated by the hepatic steatosis of liver-specific torsinA-knockout mouse model [31] and the requirement for the Drosophila torsinA homologue for proper lipid metabolism in adipose tissue [78]. Because of its functional indication in lipid metabolism, TorsinA is thought to promote membrane biogenesis [19] and synaptic physiology [79]. There are 11 DYT1 dystonia associated genome variants identified in ten lipid metabolism genes ALOXE3, APOB, CYP1B1, CYP2A7, FAM135B, GAL3ST1, GPAM, MTMR6, PLA2G4F and PLCL1 (Table 4), which suggest potential genetic interactions between the ΔE mutation and genome variants that might change membrane homeostasis.
TOR1A regulates lipid metabolism in both fruit flies and mammals [31, 78]. TOR1A facilitates cell growth, raises lipid content of cellular membrane and is involved in membrane expansion [78]. The linkage between the TOR1A ΔE mutation and 10 lipid metabolic genes suggest the impact on lipid metabolism associated cellular functions could be amplified by clustered mutations and genome variants. Two genes in this category have known functions in the neural system. The GAL3ST1 gene encodes galactose-3-O-sulfotransferase 1 that involves in the synthesis of a major lipid component of the myelin sheath galactosylceramide sulfate [80]. Gal3st1 deficient mice develop tremor, progressive ataxia, hind limb weakness, aberrant limb posture and impaired limb coordination with morphological defects in the neural system [81]. PLCL1 Involves in an inositol phospholipid-based intracellular signaling cascade. PLCL1 is phospholipase C like protein lacking the catalytic activity. PLCL1 binds and sequesters inositol triphosphates to blunt the downstream calcium signaling [82]. PLCL1 has been linked to the trafficking and turnover of GABAA receptors in neurons [83, 84]. Physiologically, loss of PLCL1 increases the incidence of chemically induced seizure in mice [85]. These findings indicate an essential role of PLCL1 in controlling the neural signaling transduction. While the functional impact of the genome variants on GAL3ST1 and PLCL1 awaits further investigation, their association with the TOR1A ΔE mutation suggests potential functional interactions between these molecules in DYT1 dystonia.
Connections between the genes harboring DYT1 variants and their implicated neuromuscular and neuropsychiatric disorders
The loss of torsinA function in either the cerebral cortex or cerebellum result in motor dysfunction [86-88], indicating a neuronal component of TOR1A’s function in dystonia. Based on these observations, we examined the 195 genes that carry candidate ΔE mutation modifiers for their association with neuropsychiatric and neuromuscular disorders. Such link was identified in 32 genes with 40 genome variants (Table 5). These include the AHNAK2, ARHGEF3, CDRT1, GBE1 and NRG2 genes in associated with peripheral neuropathy (Charcot-Marie-Tooth disease and Polyglucosan body neuropathy, adult form). The AMPD2, ATXN7 and MICAL3 genes are linked to cerebellar diseases (Pontocerebellar Hypoplasia, type 9 and spastic paraplegia 63, autosomal recessive; Spinocerebellar ataxia 7; Joubert syndrome (cerebelloparenchymal disorder)). Lastly, the IRF3, TRAF3 and LIPT2 genes are associated with encephalopathy (acute, infection-induced; encephalopathy, neonatal severe, with lactic acidosis and brain abnormalities and lipoic acid biosynthesis defects. Overall, more than 16% of the identified 195 genes are in association with neuropsychiatric and neuromuscular diseases related disorders, demonstrating the significance of the linkage between DYT1 dystonia and these diseases.
Study Limitations
The present study examined five individuals who have the TOR1A ΔE mutation. Among them, two have disease presentation and three are asymptomatic carriers. Furthermore, one affected patient and the three asymptomatic carriers are in the same family, which is an advantage to have a relatively close genetic background for modifier screening. Data from this family identified 1725 of genome variants as candidate modifiers. With the addition of the second affected patient, the number of candidate modifier variants were further narrowed down to 264. This number could have been reduced if data from more affected patients or asymptomatic carriers are available. Unfortunately, family members of the second affected patient declined to participate in the study. Due to the rareness of DYT1 dystonia in Taiwan, it is difficult to increase sample size within the Taiwanese population in foreseeable future. Alternatively, meta-analysis of our dataset with WES results from other populations across the world, once publicly available, may help to identify the common modifiers in the general population [89, 90].
The WES data allows identification of candidate modifiers in the coding genome. However, majority of the GWAS signals are mapped to the noncoding regions of the genome and accumulating evidence point to disease associations with the noncoding genome [91]. Mutations in the noncoding genome may impact cis-acting element functions and chromatin conformations that direct gene expression. Future inclusion of the whole genome sequencing assay may help to identify additional modifiers for the DYT1 dystonia.