Rapamycin
One of the best known autophagy activators is rapamycin which inhibits the mTOR kinase activity. This compound is widely used as a drug inhibiting lymphocyte activation, thus, as an immunosuppressant it is employed in treatment of patients subjected to transplantations. Rapamycin binds to the cytosolic protein FKBP-12, and following formation of the tertiary complex with mTOR, the kinase activity of this protein is inhibited. As a result, the mTOR kinase substrate, 4EBP1, is not phosphorylated (Jacinto et al.
2004) which leads to destabilization of the mTOR-Raptor complex (Kim et al.
2002).
Rapamycin is used in studies on neurodegenerative diseases in both
in vitro and
in vivo models. Studies on cellular models gave promising results, indicating a possibility to enhance degradation of proteins which cause different disorders, including mutant huntingtin (mHTT) (Ravikumar et al.
2002; Sarkar et al.
2008) and alpha-synuclein (Webb et al.
2003). Nevertheless, more studies are being conducted with animal models. Effects of rapamycin on Huntington’s disease were tested in studies employing models of
Drosophila melanogaster (Ravikumar et al.
2004; Sarkar and Rubinsztein
2008; Berger et al.
2006), zebrafish (Williams et al.
2008; Sarkar et al.
2011), and mice (Ravikumar et al.
2004). Animal experiments indicated a decrease in levels of mHTT aggregates, amelioration of neurodegeneration and improved animal behavior. Despite encouraging results of these studies, clinical trials with the use of rapamycin for treatment of HD have not been started yet.
Another disease in which rapamycin was tested using animal models is AD. Levels of beta-amyloid and hyperphosphorylated tau protein were determined as basic parameters considered in AD pathology. Experiments with AD mice overproducing the mutated gene coding for the tau protein indicated a highly elevated level of hyperphosphorylated form of this protein, while treatment with rapamycin caused its significant reduction and a decrease in number of neurofibrylar tangles (Ozcelik et al.
2013). Similar tendency was observed in mice producing toxic beta-amyloid, where rapamycin decreased the amount of this compound in the brain and improved the cognitive deficits (Spilman et al.
2010; Majumder et al.
2011; Zhang et al.
2017a; Caccamo et al.
2010).
In rapamycin-treated cellular models of PD, levels of alfa-synuclein aggregates were decreased due to stimulation of both lysosomal (autophagy) and proteasomal degradation (Webb et al.
2003).
Another example of the disease in which rapamycin was tested as a potential drug is Gerstmann-Sträussler-Scheinker syndrome, belonging to the group of prion diseases. Studies on the mouse model of this disease revealed prolonged life span, delayed symptoms and milder phenotype in animals treated with rapamycin (Cortes et al.
2012). On the other hand, different results were observed in the mouse model of ALS. Despite induction of autophagy, in rapamycin-treated mice, degeneration of motor neurons was enhanced relative to untreated controls and the life span was shorter (Zhang et al.
2011b). No significant differences were observed in the levels of aggregated superoxide dysmutase (SOD) between both groups of animals. The cause of enhanced neurodegeneration in mice treated with rapamycin remains unknown, but it seems unlikely to be due to toxicity of mutated SOD.
Despite many encouraging results of studies on rapamycin in cellular and animal models of neurodegenerative diseases, clinical trials have not been performed yet. Some doubts appeared due to the presence of adverse effects occurring in patients treated with this compound as an immunosuppressant. They include severe infections, hemolytic-uremic syndrome, cancer, leukopenia, and bone atrophy. Such adverse effects might be perhaps acceptable in a short-term treatment, for example in the transplantation procedures, however, in a long-term use, which is necessary in neurodegenerative diseases, they would be dangerous for patients.
Trehalose
Trehalose activates both mTOR-dependent and mTOR-independent pathways of autophagy stimulation. It activates the AMPK/TSC/mTOR pathway, however, it also enhances expression of the gene coding for beclin, thus, it acts by stimulation of the JNK1/Beclin-1/PI3K pathway, which is an mTOR-independent mechanism of autophagy activation (Vidal et al.
2014). Translocation of the FoxO transcription factor, a substrate for the Akt kinase (in the mTOR-dependent pathway) has been demonstrated (DeBosch et al.
2016). This indicates that trehalose, similarly to L-NAME, is an inductor of at least two pathways leading to autophagy stimulation. This is supported by results indicating involvement of trehalose in the regulation of the AMPK-dependent pathway (DeBosch et al.
2016).
Promising results were obtained in studies on cellular models of PD in which induction of autophagy (Zhao et al.
2017) or proteasomal pathway of protein degradation (Lan et al.
2012) is accompanied with a decrease in the level of quickly aggregating form of A53T-mutated alpha-synuclein. In addition, the cells producing this toxic protein were protected against apoptosis.
In studies on the cellular AD model, it was found that the level of endogenous tau protein was reduced and the toxicity caused by formation of aggregates was alleviated after treatment with trehalose (Krüger et al.
2012). It was, therefore, suggested, that this compound might be effective not only in treatment of AD but also other tauopathies. Studies on animal models demonstrated a decreased number of neurons with tau protein aggregates and lower numbers of these aggregates in cells, as well as improved viability of neurons (Schaeffer et al.
2012) as a results of treatment with trehalose. Another study with similar approach indicated improvement of motor functions of animals and alleviation of fear (Rodríguez-Navarro et al.
2010).
Decreased quantity of mHTT aggregates and significant improvement of behavior were observed in analogous studies with HD mouse model (Perucho et al.
2016). When trehalose was tested in cellular models of ALS, autophagy-dependent degradation of SOD1 aggregates led to increased viability of neurons. These results were corroborated by studies on the animal model of ALS, in which prolongation of the life span was observed (Castillo et al.
2013).
Resveratrol
Resveratrol is another stimulator of the AMPK/TSC/mTOR pathway. It activates the AMPK kinase which leads to stimulation of the autophagy process (Burkewitz et al.
2014). Efficiency of this compound in removal of toxic protein has been tested in various diseases. Resveratrol has induced autophagy and enhanced degradation of mHTT in the cellular model of HD, in which neuroblastoma SH-SY5Y line was used. Moreover, level of the Atg4 (which is decreased in mHTT-accumulating cells) normalized (Vidoni et al.
2017). This compound has also been used in the 3-nitropropionic acid (3-NPA)-induced rat model of HD. However, it is worth to note that in such animals, neurodegeneration occurs due to changes in mitochondrial metabolism which leads to (i) production of reactive oxygen species, (ii) changes in cellular energetics, and (iii) induction of apoptosis. As a consequence, hypo- and hyper-motoric changes appear which resemble symptoms of HD. Nevertheless, chorea, dyskinesis and dystony never occur in 3-NPA-treated murine models, while they are the most characteristic symptoms in humans. Moreover, this compound does not cause appearance of mHTT, the primary cause of the disease, thus, only secondary effects can be tested (Túnez et al.
2010). Hence, mechanisms of autophagy-mediated degradation of mHTT could not be studied in this model, and the tests included only anti-oxidant properties of resveratrol, which caused an increase in the level of glutathione and a decrease in levels of nitrites, as well as less efficient peroxidation of lipids. Thus, results of memory and motoric tests were better in treated animals than in controls (Kumar et al.
2006). On the other hand, experiments with transgenic HD mice have also been performed. It was suggested that activation of SIRT1 (mammalian sirtuin) by resveratrol increases viability of neurons (Ho et al.
2010). However, this compound did not affect changes in the striatum, motor functions and life span of mice.
To test effects of resveratrol on PD, mouse neuroblastoma cell lines (N2a cells) were treated with this compound in combination with β-cyclodextrin. Number of alpha-synuclein aggregates decreased and viability of the cells increased (Gautam et al.
2017). In animals, PD can be induced by MPTP (1-methyl-4-phenyl-1,2,3,6-tetrahydroxypyridine), which is converted to MPDP
+ (1-methyl-4-phenyl-2,3-dihydropyridinium), and then to the active metabolite MPP
+ (1-methyl-4-phenyl-pyridinium) by astrocytes and acts as an inhibitor of complex I of the mitochondrial electron transport system. This active toxin is catched by dopaminergic neurons in striatum and leads to their degeneration (Porras et al.
2012). When resveratrol was administered to mice before MPTP, a protection against neurodegeneration was observed, dopamine was kept at normal levels, and animal behavior was significantly less changed relative to animals treated solely with the toxin. Molecular studies indicated that SIRT1 is activated by resveratrol which leads to LC3 deacetylation and induction of autophagy. As a result, number of alfa-synuclein decreased in dopaminergic cells (Guo et al.
2016). Other tests, performed with the use of the rat model, suggested an anti-oxidative mechanism of resveratrol action, since the red-ox balance has been re-established, endoplasmic reticulum stress was alleviated, and expression of genes coding for caspases was impaired, which might protected cells against apoptosis (Gaballah et al.
2016).
Calcium canal antagonists
Antagonists of calcium canals, which also activate the Ca
2+/calpain pathway, are relatively often tested in metabolic brain diseases. The list of such compounds include: latrepirdine, verapamil, loperamide, nitrendipine, nilvadipine, nimodipine, amiodarone, niguldipine, nicardipine, pimozide, penitrem A, fluspirilene, and trifluoperazine. When the calcium channel is blocked, the intracellular calcium level drops rapidly, thus, calpains are inactivated which stimulates autophagosome formation. Such compounds were tested in cellular models of HD, and it was found that they caused reduction of the mHTT level (Zhang et al.
2007; Williams et al.
2008).
Verapamil has been tested in experiments with a HD mouse model. This drug caused improvement in motoric activity and keeping balance by animals (Kalonia et al.
2011). Although verapamil has not been tested in clinical trials for HD, its efficacy was assessed in treatment of ALS patients. However, 5-month treatment did not result in improvement of the disease parameters (Miller et al.
1996b).
When nitrendipine and nilvadipine were studied using AD models, inhibition of beta-amyloid accumulation
in vitro and its enhanced degradation
in vivo, accompanied with memory improvement in mice, were observed (Paris et al.
2011). Interestingly, quite similar results were obtained during experimental therapy of AD patients when nilvadipine and nimodipine prevented the progression of congnitive problems (Nimmrich and Eckert
2013). On the other hand, in the clinical trial phase III with ALS patients, nimodipine appeared ineffective and many adverse effects were reported, including diarrhea, nausea, and lightheadedness (Miller et al.
1996a).
Amiodarone has been tested mainly as a potential anti-AD drug. Experiments were conducted
in vitro and
in vivo (with the Guinea pig model). Amiodarone was used as a compound which elevates pH, thus, secretases that cut the APP protein (the amyloid precursor) and require acidic environment, were inactivated. Thus, the level of amyloid decreased, however, a mechanism involving stimulation of autophagy was not considered, though the authors suggest that elevation of pH is perhaps not the only way of action of the tested compound (Mitterreiter et al.
2010).
Pimozide, is already used in medicine for treatment of schizophrenia and psychotic disorders (Mothi and Sampson
2013). However, its positive effects were observed also in HD. Experimental therapy with a low number of patients indicated an improvement in hyperkinesia (Girotti et al.
1984). Studies with the mouse model of AD which overproduces hyperphosphorylated tau protein, indicated that intraperitoneal administration of pimozide resulted in reduction of aggregates of this protein and to memory improvement in animals (Kim et al.
2017). Interestingly, when considering a possible molecular mechanism of pimozide action, the authors did not consider the calcium channel-dependent pathway. On the contrary, they have detected an increased level of phosphorylation of AMPK and ULK1 kinases, and inhibition of the mTOR kinase activation. They have suggested that pimozide may stimulate as yet unknown pathway of autophagy induction which is dependent on AMPK-ULK interactions, with no involvement of mTOR (Kim et al.
2017). One case of an AD patient treated with pimozide for 5 weeks has been described, and alleviation of dementia was noted; the effects remained unchanged for next 9 months (Renvoize et al.
1987). Despite these results, no clinical trial with this compound was reported to date.
Trifluoperazine has been tested as a potential anti-HD and anti-AD drug. However, it was suggested that this compound may inhibit apoptosis, and stimulation of autophagy has not been considered (Lauterbach
2013). Studies with patients included only a very limited number of individuals, though the results were quite encouraging (Stokes
1975). On the other hand, trifluoperazine was widely tested for treatment of AD. However, a clinical trial with this compound indicated that life span of treated patients was shortened by 12 months relative to untreated controls (Ballard et al.
2009). In another clinical trial, no improvement of the disease symptoms could be find (Ballard et al.
2008). Interesting studies were conducted on a PD mouse model, characterized by moderate expression of the gene coding for synucleine, thus the time of death of neurons could be determined precisely. It was found that non-induced autophagy is impaired by accumulated synuclein, while trifluoroperazine-induced autophagy causes a delay in neurons’ death (Höllerhage et al.
2014). When screening for anti-PD compounds was conducted using a zebrafish model, trifluoperazine was identified as a molecule preventing the loss of neurons. It was demonstrated that apart from blocking calcium channels, trifluoperazine stimulated translocation of the transcription factor EB (TFEB; a master regulator for lysosomal biogenesis that also activates autophagy, when present in the nucleus in a non-phosphorylated form, by enhancing transcription of relevant genes) which is a substrate for the mTOR kinase. Therefore, this compound is another factor which can stimulate autophagy through more than one pathway (Zhang et al.
2017b).
Studies on other blockers of calcium channels (loperamide, niguldipine, nicardipine, panitrem A, and fluspirilene) as anti-neurodegenerative agents were terminated after
in vitro studies (Zhang et al.
2007; Williams et al.
2008) due to adverse effects reported in the meantime by researchers using them for treatment of other diseases. These adverse effects included: constipation, dizziness, nausea, paralytic ileus, angioedema, anaphylaxis reactions, toxic epidermal necrolysis, Stevens-Johnson syndrome, erythema multiform, urinary retention, and heat stroke.
Calpastatin
Calpastatin inhibits activities of calpains, thus, inhibition of autophagosome formation is abolished. Overproduction of calpastatin induced autophagy decreased levels of mHTT, improved motor functions, and delayed appearance of other symptoms in the mouse model of HD (Menzies et al.
2015). Importantly, prolonged administration of calpastatin did not cause any severe adverse effects in animals. In studies on the cellular AD model, an inhibitor of histone deacetylase, trichostatin A, which also increases production of calpastatin, caused an increase in viability of cells (Seo et al.
2013). These results are in agreement with observations that silencing of expression of calpastatin-encoding gene causes changes in cytoskeleton and lowers cell viability (Rao et al.
2008). Moreover, long-term activation of calpains causes overstimulation of many proteases, which leads to degradation of a number of cellular substrates, including cytoskeleton elements and membrane receptors involved in homeostasis maintenance. When calpastatin is overproduced, such effects can be diminished (Schoch et al.
2013). Overexpression of the calpastatin gene in the mouse model of PD resulted in reduction of the number of alpha-synuclein aggregates and improved signal transduction through synapses (Diepenbroek et al.
2014).
Lithium
Lithium is tested as a potential drug for many diseases affecting central nervous system (CNS). This agent inhibits activity of inositol monophosphatase, decreasing the level of inositol and IP3 which allows formation of the autophagosome membrane (Sarkar et al.
2005). However, lithium negatively regulates also activity of another enzyme, GSK-3β, causing stimulation of the mTOR kinase and autophagy inhibition. Therefore, it was proposed to combine the use of lithium and rapamycin (an inhibitor of mTOR). This approach appeared significantly more effective than the use of each component separately (Sarkar et al.
2008). However, in studies with the 3-NPA-induced HD rat model, treatment with LiCl for 8 days caused an increase in pathological changes in the brain (Milutinović
2016). On the other hand, it is worth remaining that 3-NPA does not cause the appearance of mHTT aggregates, thus, this is not an adequate model for testing potential drugs which might activate autophagy. In such studies, genetic models of mHTT would be much more relevant. It was also reported that lithium causes a decrease in the level of histone deacetylase (HDAC1) which is correlated with effective degradation of mHTT (Wu et al.
2013). Lithium has also been used in experimental therapy in which 3 patients suffering from HD were involved. In one patient, some neurological parameters were improved, but no changes in chorea could be observed. The second patient responded with improvement in chorea with no neurological changes. In the third patient, stabilization of all symptoms, but no improvement, was noted. Nevertheless, all these patients received also other drugs, including carbamazepine, which makes interpretation of the results very difficult (Danivas et al.
2013). Other clinical trials with HD patients also did not give conclusive results regarding efficacy of lithium due to extremely different responses of various persons (Scheuing et al.
2014).
When mice overproducing hyperphosphorylated tau protein were treated with lithium, a significant improvement in behavior and cognitive functions was observed, levels and phosphorylation of tau decreased, as did efficiency of beta-amyloid formation, and levels of autophagy markers increased (Shimada et al.
2012; Zhang et al.
2011a). Efficacy of lithium in the mouse model of PD was tested in a combined therapy with valproic acid. Improvement in behavior and an increase in the number of dopaminergic neurons were evident. Deprivation of dopamine and its metabolite, dihydroxyphenyloacetic acid, was less pronounced than in untreated animals (Li et al.
2013). Analogous combination of drugs was tested in the mouse model of HD. Treated animals expressed improvement in motoric functions and memory (as tested in the Morris water maze). Reduction of the level of mHTT aggregates and less pronounced loss of neurons in striatum were observed. Interestingly, expression of genes coding for proteins involved in mitochondrial metabolism, antioxidative response, apoptosis and anti-inflammatory reactions were significantly modulated (Linares et al.
2016). These results indicate the broad spectrum of biological activities of lithium and valproic acid, as suggest that a complex network of processes is involved in the pathogenesis of HD.
Valproic acid
Valproic acid inhibits activity of myo-inositol-1-phosphate synthase, one of enzymes involved in the metabolism of inositol (Shaltiel et al.
2004), thus, causing a decrease of the level of the latter compound and activation of autophagy. Combination of valproic acid and lithium was tested in clinical trials with HD patients. However, in most cases either a lack of effects or only stabilization of symptoms (with no improvement) were observed (Scheuing et al.
2014).
AD model cells were treated with valproic acid, and no changes with the total amount of beta-amyloid were demonstrated while level of beta-amyloid oligomers (which are suggested to be more toxic) decreased and level of monomers increased, relative to untreated control cells (Williams and Bate
2018). This may suggest that beta-amyloid oligomers are converted to monomers in valproic acid-treated cells. Streptozotocin (STZ)-induced rat model of AD has been used in
in vivo studies. Intraventricular injection of STZ provokes neurodegeneration and accumulation of beta-amyloid and hyperphosphorylated tau protein, thus, mimicking the sporadic form of AD. Decreased levels of acetylcholine and neprylysine, and increased activity of acetylcholinesterase cause additionally enhanced neurodegeneration and cognitive defects. Treatment with valproic acid resulted in prevention of cognitive deficits and normalization of levels and activities of neurotransmitters (Sorial and El Sayed
2017). Using another animal model of AD, transgenic mice expressing a mutated
APP gene, effects of valproic acid in males and females were compared. Decreased levels of amyloid plaques were more pronounced in males than in females, while number of synaptic vesicles were similar in both genders. On the other hand, neurodegeneration was prevented more efficiently in males (Long et al.
2016).
Cellular models of PD were used to investigate the mechanism of action of valproic acid. This compound caused reduction of levels of proapoptotic proteins and ROS, while autophagy inhibitors diminished these effects, indicating a crucial role of this process in valproic acid-mediated improvement in PD cellular phenotypes (Zhang et al.
2017c). Other
in vitro studies were based on the use of murine neurons treated with human beta-amyloid. Defects in synaptic proteins and neurotransmitter transporting vesicles were observed. These effects were alleviated by addition of valproic acid into the cell culture. A mechanism has been proposed in which this compound negatively regulates cytoplasmic phospholipase A2 (cPLAS2), whose overactivity correlates with neurodegeneration. Autophagy has been suggested as an additional mechanism of the observed changes in cells (Williams and Bate
2016).
Controversial results were obtained in studies on prion disease. Early studies suggested that valproic acid causes an increased accumulation of PrP in neuroblastoma cells and model cells for the disease. However, administration of valproic acid to Chinese hamsters infected with prions did not cause any effects on the course of the disease (Shaked et al.
2002). Other studies performed with cellular models did not confirm effects of valproic acid on the levels of PrP (Legendre et al.
2007).
Carbamazepine
Mechanism of action of carbamazepine is similar to that by valproic acid and lithium (Williams et al.
2002). A decrease in the inositol level arises from deprivation of PIP2 and IP3 (Schiebler et al.
2015). Studies with the mouse model of AD indicated that carbamazepine improved learning abilities and memory, which was correlated with decreased number of amyloid plaques (Li et al.
2013; Zhang et al.
2017a). Apart from stimulation of mTOR-independent pathway of autophagy activation, carbamazepine inhibited the mTOR kinase activity. Therefore, it is another example of autophagy stimulation by more than one molecular mechanism (Li et al.
2013).
Since carbamazepine is known as an analgesic, anticonvulsant and antiepileptic drug, it has been used for treatment of HD patients (Danivas et al.
2013). It was proposed that its mechanisms of action is related to blocking calcium channels which cause inhibition of glutamate liberation (Kawata et al.
2001). Intriguingly, autophagy was not considered as a mechanism by which carbamazepine improves symptoms of HD.
An interesting case report has been published in which carbamazepine was administered to a patient suffering from hypertension, myocardial infraction, and atrial fibrillation. When high doses of the drug were used (as the patient became resistant to lower doses), many adverse effects were noted, including memory deficits, confusion, psychomotor slowness, hypersomnia, dysphasia, and postural instability with falls. The patient’s condition was continuously deteriorating. Psychological tests indicated attentional deficits, perseverations, severe non-fluent aphasia with paraphasias, and constructional apraxia. The EEG results were similar to those found in patients suffering from Creutzfeldt–Jakob disease. After cessation of treatment with carbamazepine both EEG results and patient’s conditions improved considerably. Cognitive deficit and motor dysfunctions normalized. Thus, it was concluded that carbamazepine caused Creutzfeldt–Jakob disease-like symptoms (Horvath et al.
2005).
Clonidine
Clonidine binds and activates the imidazoline receptor, which leads to a decrease in the level of cAMP in cells (Williams et al.
2008). However, it appears that there is an additional mechanism of action of this compound, namely activation of potassium channels which causes a decrease in concentration of calcium ions in the cytoplasm (Murphy and Freedman
2001). Clonidine was used as one of compounds activating autophagy in the screening for a potential drug for HD and PD. It was effective in reducing amounts of synuclein and mHTT in cells (Williams et al.
2008).
In vivo experiments were performed with reserpine-treated rat model of Parkinson's disease. Following injection of reserpine, severe akinesis was observed which could be prevented by previous treatment with clonidine (Hill and Brotchie
1999). However, stimulation of autophagy was not considered as a potential mechanism of action of this drug. In PD potential therapies concentrate on inhibition of movements while patients suffer also from cognitive deficits and mood swings. When clonidine, as an agonist of adrenergic receptor alpha-2, was tested as a potential drug at early phase of PD in a monkey model (
Macaca fasicularis), it was found that the treatment caused improvement in concentration and memory (Schneider et al.
2010).
Models of memory deficits were also used in studies on clonidine. In murine models, the symptoms were induced by administration of NMDA (N-methyl-D-aspartate) antagonist MK-801 or by excitotoxic hippocampal damage. Clonidine ameliorated symptoms caused by MK-801, but did not change behavior of rats in which hippocampus was damaged by excitotoxic agents (Bardgett et al.
2008).
The only studies on the use of clonidine in prion disease were performed with the yeast model. However, no significant effects on the level of PrPSC could be observed (Tribouillard-Tanvier et al.
2008).
Rilmenidine
Rilmenidine induces the autophagy proces through the cAMP/EPC/PLC pathway. Similarly to clonidine, it binds and activates the imidazoline receptor. It was tested in experiments with cellular HD and PD models, and caused a decrease in levels of mHTT and alpha-synuclein (Williams et al.
2008). In the mouse model of HD, reduction of mHTT levels was also observed but number of aggregates remained unchanged. Although rilmenidine could not prevent the body weight loss, it corrected the muscle parameters and general condition of the organism (Rose et al.
2010). Although a clinical study with 18 HD patients has been conducted, only 12 patients completed this trial. Some cognitive parameters and motor functions were improved (Underwood et al.
2017); however, a study with significantly higher number of patients is necessary to make solid conclusions.
Cellular and animal models of ALS were used to study effects of rilmenidine in this disease. A decrease in mutant SOD1 level was observed in cells in which macroautophagy and mitophagy were also evident. Similarly, administration of this drug to mice suffering from ALS resulted in autophagy induction in motor neurons. Unexpectedly, enhanced degeneration of these neurons was observed under these conditions. Moreover, accumulation of SOD aggregates and a decrease in number of mitochondria occurred in treated animals, and correlated with more severe symptoms relative to untreated mice. It was suggested that too extensive mitophagy could be responsible for these effects (Perera et al.
2017).