TBK1, a prioritized drug repurposing target for amyotrophic lateral sclerosis: evidence from druggable genome Mendelian randomization and pharmacological verification in vitro

Amyotrophic lateral sclerosis (ALS) is a fatal motor neuron disease characterized by progressive degeneration of nerve cells in the spinal cord and brain, leading to disability and eventual death of patients from respiratory failure. The prevalence of ALS is about 5 in 100,000, while patients usually die within 2 to 5 years of onset [1]. Due to a lack of understanding in its pathogenesis, formidable challenges have been historically posed on therapeutic approaches. Currently, riluzole, edaravone, AMX0035, and tofersen have been approved by FDA for treating ALS by reducing glutamate release, acting as an antioxidant, mitigating mitochondrial dysfunction, as well as down-regulating superoxide dismutase 1 (SOD1) protein level, respectively. However, most of them were demonstrated to slightly slow down ALS progression by a couple of months [2, 3]. Therefore, it is a priority to develop more efficient chemical compounds for urgent need to treat this devastating disease.

Accurately identifying drug candidates for particular disease subtypes and verifying their impacts on disease progression are prerequisites in drug development. However, due to the small sample size of the trial cohort, genetic complexity and heterogeneity, and insufficient drug efficacy or safety data, most clinical trials for newly developed drugs have to be terminated with unsatisfactory results [4]. Meanwhile, several studies have reported that the genetically supported drug-repurposing strategies could double the success rates in clinical development [5], and numerous large-scale genome-wide association studies (GWASs) have identified many single-nucleotide polymorphisms (SNPs) associated with ALS risk, offering possibilities for shortening the innovation period of ALS drug. According to Mendel’s law of independent segregation, genetic variants are randomly allocated to individuals during gamete formation and fertilization. Mendelian randomization (MR) applies these genetic variants (instrumental variables, IVs) to study causality with less confounding or unbias than that in traditional epidemiological methods. Therefore, MR has been considered an ingenious method to harness high power and fidelity of randomization in human biomedical research [6, 7]. The druggable genes, as described by Finan et al. [8], refers to the selection of genes that have the potential to serve as targets for pharmacological intervention. In drug-target MR analysis, cis-expression quantitative trait loci (cis-eQTL) located in the genomic region of the drug-target gene are often considered proxies, which have been intensively utilized to analyze the regulatory relationship between nucleotide variants and gene expression fluctuations [9, 10]. Such MR analyses have been applied to multiple diseases, such as Parkinson’s disease (PD) [11] and Alzheimer’s disease (AD) [12].

ALS GWAS have revealed numerous genes that exhibit statistical associations with the disease [13]. Among them, TNFSF12, also called TWEAK, has been researched in tumors, autoimmune disease, and inflammatory diseases [14‐18]. Studies have shown that TNFSF12 plays a crucial role as an apoptosis inducer in the pathogenesis of muscle atrophy and is associated with neuroinflammation in multiple sclerosis [19, 20]. GPX3 is abundantly present in neurons, the brain, and other tissues [21]. It is a well-known glutathione peroxidase, performing antioxidant functions associated with well-recognized ALS gene like SOD1 [22, 23]. TBK1 gene encoding TANK-binding kinase-1 (TBK1) has been shown to closely interplay with ALS development. Multiple loss-of-function mutations on TBK1 gene have been identified across sporadic and familial instances of ALS, as well as in cases of ALS/frontotemporal dementia (ALS-FTD/FTD) [24]. On the other hand, within ALS subtypes linked to gain-of-function mutations of SOD1 or TARDBP (encoding TAR DNA-binding protein, TDP-43), it has been observed that mitochondrial damage caused by mutated proteins could potentially stimulate the release of mitochondrial DNA, thereby activating the cyclic GMP-AMP synthase (cGAS)/stimulator of interferon genes (STING) pathway to potently initiate or aggravate neuroinflammation [25, 26]. Nowadays, extensive cellular and animal studies have demonstrated the protective effects of chemical inhibitors targeting upper stream factors like STING and cGAS in ALS models, while breakthroughs in clinical application remain to be achieved [25, 27].

In the current study, we obtained eQTL data for the druggable genome and conducted MR with the outcome of ALS GWAS to identify potential druggable genes associated with ALS. Through further colocalization analysis, we obtained the target genes with the stronger genetic association, including TBK1, TNFSF12, RESP18, and GPX3. Subsequently, we delved into ALS-repurposed drugs related to druggable genes through a series of criteria and identified R788 and AMX. Moreover, we conducted further research into druggable gene targets and the pharmacological effects of their targeted drugs in the context of ALS cell models. Lastly, AMX and R788 were verified to hold the therapeutic potential in attenuating ALS progression via dampened neuroinflammation.

All DNA positions are based on the human reference genome build hg19 (GRCh37). Data processing was completed using R software version 4.1.1. The main R package involved in this study included “TwoSampleMR”, “dplyr”, “ieugwasr”, “org.Hs.eg.db “, “Homo. sapiens”, and “coloc”. This MR study was in accordance with the guidelines of Strengthening the Reporting of Observational Studies in Epidemiology using Mendelian Randomization (STROBE-MR) [28, 29]. The checklist is elaborated in detail in Additional file 1: Table S1. Flowchart of our study is shown in Fig. 1A-D.

Currently, there is no established therapeutic approach for ALS that is proven to possess high efficacy across major patient populations. By integrating large-scale GWAS data with the druggable genome, we investigated more than 4000 potential gene targets to yield TNFSF13, CD68, TNFSF12, TBK1, RESP18, GDF9, PTPRN, and GPX3 target loci via systematic MR. TBK1, TNFSF12, RESP18, and GPX3 were then validated in the colocalization analysis. Further repurposing drug discovery and pharmacological verification analyses were applied to identify TBK1 as a promising repurposing drug target for ALS treatment. Although we regard TBK1 as one of the most promising targets for the treatment of ALS, it is essential to emphasize the significant value of TNFSF12, RESP18, and GPX3 in ALS drug research as well. Future studies should also focus on the interactions between these genes and others to gain a deeper understanding of their roles in the disease mechanism.

Among the abovementioned genes, TBK1 encodes a multimeric kinase regulating various biological processes [58] and serves as a known FDA-approved drug target in treating rheumatoid arthritis and multiple sclerosis [59]. Recently, several studies with large human genetic samples have suggested the multi-faceted roles of TBK1 mutations in the pathogenesis of ALS [60‐62]. In genetic results, TBK1 haploinsufficiency has been found to disrupt motor neuron autophagy, enhance mutated SOD1 protein accumulation, and promote ALS-like phenotypes in transgenic mouse models [63]. Meanwhile, researchers have also highlighted the bi-directional roles of TBK1 in the context of the SOD1^G93A ALS mouse model. Gerbino et al. indicated that diminished TBK1 function by point mutation, as well as in SOD1^G93A/TBK1 cKO (conditional TBK1 deletion in motor neurons) mice, promoted disease progression during the early stages but impeded it in later stages [26]. Similar research revealed that SOD1^G93A/TBK1^± mice carrying TBK1 heterozygous deletion displayed clinical symptoms earlier; however, decelerated disease progression and elevated survival rate [64]. TBK1 acts as an adaptor protein downstream of STING, leading to an augmentation of neuroinflammation via IRF3/IFNB axis [25]. The canonical cGAS/STING signaling is activated upon the presence of foreign DNA in the cytosol, such as that derived from invading bacteria or viruses or leaked self-DNA from the nucleus or mitochondria under pathological conditions [65]. In certain pathological conditions, the excessive TBK1 abundance might be associated with over-activated signaling. Clark et al. observed increased TBK1 protein level paralleled with elevated phosphorylation, which was partially due to intermolecular autophosphorylation [66]. The other study has also reported that overexpression of TBK1 in zebrafish could induce IRF3 phosphorylation [67]. Our MR analysis identified increased TBK1 expression in ALS patients, which may possess over-activated signal transductions. It has been demonstrated that in SOD1/TDP43-associated ALS mice models, the accumulation of toxic SOD1 or TDP43 proteins mislocated in mitochondria was found to trigger the release of mtDNA and RNA:DNA hybrids into the cytosol [52]. These two ALS-related toxic genes and their gain-of-function mutants were found to efficiently stimulate the cGAS/STING pathway through both the TBK1/IRF3 and the IKKε/NF-ҡB axes leading to the subsequent activated expression of inflammatory cytokines. Particularly, in a Prp-TDP-43^Tg/+ ALS mouse model, the disease was mitigated through genetic deletion or pharmacological inhibition of STING [25]. Therefore, targeting the pro-inflammatory effects of STING activation using specific inhibitors represents a promising therapeutic approach to impede the progression of ALS. A rising number of inhibitors targeting the cGAS/STING pathway have been reported [68].

Compared to innovating novel drugs, repurposing approved drugs for new indications has been demonstrated as a low time–cost and less investment strategy with an improved success rate [69]. For example, metformin has been proven to enhance prognosis in patients with non-diabetic heart failure [70, 71]. In addition, the kinase drug repurposing and novel target discovery have also been widely investigated. For example, imatinib, an FDA-approved ABL kinase inhibitor for chronic myeloid leukemia (CML), was later discovered to effectively treat advanced gastrointestinal stromal tumors by inhibiting KIT kinase [72, 73]. The aurora kinase inhibitor tozasertib has also been successfully used to treat CML-bearing ABL1 kinase mutant (T315I) [74]. Based on Mendelian randomization analysis and previous reports, our drug repurposing analysis on FDA-approved drugs identified the dual-specificity inhibitor AMX known to block TBK1 and IKKε enzymatic activities [75], which is utilized in several ongoing clinic trials [76, 77]. Our in vitro data suggested that AMX efficiently inhibited the IRF3/p65 phosphorylation induced by either cGAS/STING agonist MSA-2 or overexpression of ALS toxic proteins. These effects resulted in the attenuation of downstream pro-inflammatory gene expression (Fig. 4D, J–L).

Furthermore, polypharmacological effects are commonly observed in kinase inhibitor developments due to structural and functional conservation across ATP binding sites, which may have advantageous or detrimental consequences for their clinical applications [78]. Klaeger et al. showed that the FDA-approved MET/VEGFR inhibitor, cabozantinib, effectively inhibited the tyrosine kinase fusion product FLT3-ITD to potentially cure acute myeloid leukemia [78, 79]. As a known SYK inhibitor, R788 has been widely used in autoimmune diseases such as rheumatoid arthritis lymphoma [80] and immune thrombocytopenic purpura [81]. In the current study, we found a significant binding affinity of its active form R406 to TBK1 in serval independent online databases (Kd =  ~ 22 nM) (Lincs, DrugBank, and BindingDB) [47]. The molecular docking results suggested that both R788 and R406 were predicted to be located within the ATP-binding site or ATP-pocket of TBK1, with mostly overlapped interactions shown in the published R406/SYK co-crystal structure (Fig. 3B–D). Our in vitro results showed that R788 effectively inhibited the rise in TBK1/IRF3 phosphorylation levels induced by either cGAS/STING agonist MSA-2 or overexpression of ALS toxic proteins, resulting in the dampened pro-inflammatory gene expression (Fig. 4D, G–I). To this end, the repurposing potential of R788 on TBK1 has been demonstrated feasible by both bio-informative and pharmacological validations. Collectively, the results from in vitro findings and interaction analysis suggested that AMX and R788 might be utilized to suppress cGAS/STING-mediated neuroinflammation, thereby potentially treating ALS subtypes particularly related to overexpressed SOD1, TDP43, and their derivatives, respectively. Overall, by applying druggable genome-wide MR and pharmacological verification in vitro, we affirmed the promising application of AMX and R788 for repurposing uses in ALS. Because AMX and R788 have been approved for marketing by the FDA, their efficacy, safety, and adverse effects have been confirmed, while the expected benefit of conducting clinical trials is high.

In addition, we also identified that targets on TNFSF12 and GPX3 had potential protective roles in ALS. As we have discovered that TNFSF12 overexpression have a potential risk for ALS, MR and colocalization results strongly suggested that this gene could be a promising drug target for ALS. Studies have shown that TNFSF12 and its associated factor, fibroblast growth factor-inducible type 14 (Fn14), is an emerging apoptosis inducer to play an important role in the pathogenesis of muscle atrophy [19], which has also been validated in mouse models and clinical cohorts [82‐84]. We searched both ChEMBL and DrugBank Online for drugs that are currently in clinical trials to inhibit TNFSF12 expression and ultimately identified two drugs: BIIB-023 (ChEMBL ID: CHEMBL2109598) and RG7212 (ChEMBL ID: CHEMBL2109600). BIIB023, a humanized anti-TNFSF12 monoclonal IgG1 antibody, is currently in clinical trials for rheumatoid arthritis and has a favorable safety and tolerability profile [42]. RG7212, a fully humanized IgG1κ monoclonal antibody that attenuates the binding of Fn14 to TNFSF12, has been shown to inhibit tumor growth in multiple in vivo models, including renal, breast, and pancreatic cancers [85, 86]. We speculated that RG7212 and BIIB-023 might improve muscle atrophy in ALS patients through the TNFSF12 pathway, ultimately delaying disease progression. However, due to the uncertain safety and efficacy of RG7212 and BIIB-023, we have not prioritized them as top drug repurposing and have not conducted pharmacological research on them.

GPX3 gene was found as the unique gene altered in both ALS blood and brain samples. GPX3 is abundantly present in neurons, the brain, blood plasma, and other tissues. It has been reported to reduce oxidative stress by detoxifying hydrogen and soluble lipid peroxides catalyzed by reduced glutathione [21]. A recent study suggested that GPX3 genetic alteration may play an important role in the pathogenesis of ALS through reactive oxygen species (ROS) [23]. Interestingly, the brain eQTL results suggested that GPX3 expression reduced the risk of developing ALS, which was contrary to the results in the blood, suggesting that GPX3 transcriptional and translational regulations might not be coupled in different tissues. We searched both ChEMBL and DrugBank Online with GPX3 as a target and identified glutathione in documented clinical trial profiles. Glutathione binds to other drugs to make them more soluble for excretion, and it also serves as a cofactor for certain enzymes involved in protein disulfide bond rearrangement and reducing peroxide production [87]. Due to the molecular characteristics and mechanism of action of glutathione, it is better suited as a supplement rather than a unique ALS treatment.

The strengths and innovations of our study include the following. Firstly, our study employed druggable genome MR to investigate potential drug targets and identify repurposed drugs for ALS treatment. Secondly, we meticulously selected multiple tissue datasets with the largest sample size and rigorously conducted verifications using various instrumental variable selection criteria, thereby enhancing the credibility of our findings. Furthermore, our study also validated partial findings from previous research, such as TBK1 [13], a druggable target identified in our study; however, it was not reported as a prioritized drug target for ALS in other studies [23, 88‐90]. Additionally, we have made novel discoveries by identifying new druggable genes in blood samples, brain tissue, and spinal cord (e.g., TNFSF12, TNFSF13, and RESP18), which offer promising avenues for future investigation of ALS drugs.

Our study also has some limitations. Firstly, the results of the MR method may have weak instrument bias, i.e., the degree of phenotypic variation explained by a single genetic instrument is relatively limited, and the sample size needs to be expanded to obtain sufficient precision. Secondly, genetic associations for variants identified through genome-wide association studies tend to be overestimated in the original discovery data set, also known as “winner’s curse bias” [91]. Thirdly, we set a more stringent r² value to sufficiently remove the linkage disequilibrium to improve the accuracy of the study, which resulted in only 1 SNP for all our positive results. Thus, our results may suffer from a lack of statistical power but also possess a small intrinsic bias. Furthermore, our data are derived from European populations, which may lead to non-applicability between races. Additionally, our MR analysis was limited to selected known druggable genome eQTLs and ALS GWAS, potentially excluding other potential druggable genes from the analysis. Moreover, our analysis of IVs was based on the set selection criteria, which could introduce selection bias and result in false positive or false negative deviations in this study. It is worth noting that although our research relies on bulk tissue gene expression data, drug treatments may exert their effects specifically on certain cell types or tissues. Therefore, it is crucial to further investigate and address the heterogeneity of tissues and specific cell types. Lastly, it is crucial to underscore that despite the evidence provided by MR and pharmacological validation in cells, successful drug application cannot be guaranteed due to potential disparities in gene expression data between diseased and healthy states. Real-world applications may give rise to unexpected adverse effects or limited clinical benefits.

Using the druggable genome-wide MR approach, we obtained four potential drug targets for ALS. Repurposing drug discovery found five drugs with repurposing opportunities acting on TBK1, TNFSF12, and GPX3, respectively. After inclusion criteria filtration, we prioritized two FDA-approved TBK1 inhibiting chemicals that act as repurposing drugs for treating certain ALS subtypes and validated their efficacies by in vitro pharmacological analyses. Our research provides new insights into genetic-based drug development for ALS treatment, which will offer potential for successful clinical trials.