Small Molecule Improvement of Trafficking Defects in Models of Neurodegeneration

Abstract

Disrupted cellular trafficking and transport processes are hallmarks of many neurodegenerative disorders (NDs). Recently, efforts have been made toward developing and implementing experimental platforms to identify small molecules that may help restore normative trafficking functions. There have been a number of successes in targeting endomembrane trafficking with the identification of compounds that restore cell viability through rescue of protein transport and trafficking. Here, we describe some of the experimental platforms implemented for small molecule screening efforts for rescue of trafficking defects in neurodegeneration. A survey of phenotypically active small molecules identified to date is provided, including a summary of medicinal chemistry efforts and insights into putative targets and mechanisms of action. In particular, emphasis is put on ligands that demonstrate activity in more than one model of neurodegeneration as retention of phenotypic activity across ND models suggests conservation of biological targets across NDs.

Graphical Abstract

INTRODUCTION

Neurodegenerative disorders (NDs) are a family of diseases characterized by progressive loss of neurons in the brain and peripheral nervous system, manifesting in physical symptoms such as deterioration of motor control and nonmotor symptoms like cognitive decline, memory loss, and depression. NDs occur with prevalence,^1,2 from more common disorders like Alzheimer’s disease (AD) and Parkinson’s disease (PD) to less frequently occurring disorders like amyotrophic lateral sclerosis (ALS) or Huntington’s disease (HD).^3–7 The majority of these diseases affect older adults, cause years of lost life or years living with disability, and often are ultimately fatal. While these NDs can result from numerous triggers or factors, a common pathological hallmark among them is disruption of trafficking and transport processes within the cell.^8–13 While it is not always clear whether defective trafficking is a primary cause of neurodegeneration or a result of other primary defects, the disruption of transport mechanisms contributes to cellular toxicity and, subsequently, to neuronal death. Despite each of the NDs having distinct causes, symptoms, and treatments, the occurrence of a common mechanism (in this case, disruption of cellular trafficking) provides the potential to identify conserved pathways of dysfunction that could reveal common cellular targets for therapeutic rescue of trafficking across the ND spectrum. Identifying targets and therapeutic leads to restore defective trafficking could provide a new avenue for treatment of NDs, many of which have limited therapies and lack neuroprotective or neurorestorative treatment options.

In this review, we provide a brief overview of endomembrane trafficking pathways affected in NDs and experimental models used to study these systems and to review progress toward identifying and advancing small molecules that may restore trafficking defects associated with neurodegenerative disorders. Further, targets and mechanisms of lead compounds identified to date are discussed, with particular focus on ligands that demonstrate activity in more than one model of neurodegeneration.

ENDOMEMBRANE SYSTEM, TRAFFICKING PATHWAYS, AND DEFECTS IN NEURODEGENERATION

The viability and function of eukaryotic cells depends on the endomembrane system, an intracellular trafficking network for transport between membrane-bound organelles with compartmental specificity. While trafficking is an essential mechanism in all cells that exhibit compartmentalization, the maintenance of intracellular trafficking pathways is especially important in neurons. Neurons are particularly long-lived cells, reaching a terminally differentiated, postmitotic state early in development but going on to function for decades.¹⁴ Further, neurons exhibit distinct and complex morphologies, requiring membrane trafficking processes to shuttle proteins and biomolecules between the distant extensions via axonal transport.^9,15–17 The function of membrane trafficking in neurons varies widely, including maintenance of cellular proteostasis, synaptic vesicle signaling, and regulation of surface receptors involved in synaptic signaling.^9,15–17 As trafficking is involved in nearly all aspects of neuronal maintenance and function, it is not surprising that dysregulation of cellular trafficking is a hallmark of most neurodegenerative disorders.

There has been significant research into the effects of trafficking dysregulation in neurodegeneration broadly, revealing that trafficking defects can be both a cause and an effect of disrupted proteostasis. In some cases, NDs are associated with aggregation of an aberrant protein that alters neuronal proteostasis and induces trafficking defects downstream. This is the case in α-synuclein-associated Parkinson’s disease, where accumulation of the protein aggregate is associated with disruption of ER-to-Golgi trafficking, among other neurotoxic effects.^18–20 Similarly, in Huntington’s disease, the presence of polyglutamine (polyQ) expansions in key proteins results in formation of protein aggregates and subsequent trafficking disruption.^21–23 Alternatively, some NDs are associated with a defect in a specific trafficking pathway such as in lysosomal storage disorders,^24–26 where a defect in lysosomal processing itself is associated with neurotoxic pathology. Through extensive research, reviewed elsewhere,^{8–10,13,17,27,28} it has been shown that several main endomembrane trafficking pathways are disrupted in one or more ND, affecting both secretory and endocytic trafficking. Affected pathways include the autophagic, lysosomal, ER-to-Golgi, synaptic vesicle, membrane recycling, and axonal transport pathways (Figure 1).^9,16,17

Figure 1.

Defects occur in various stages of endomembrane trafficking in the presence of ND toxicity.^9,16,17 Stages of endomembrane trafficking are disrupted by neurotoxicity including ER-Golgi trafficking (PD, ALS, HD, AD), autophagy and lysosomal trafficking (AD, HD, PD, ALS), axonal transport (AD), and synaptic vesicle transport (PD, HD). Disrupted trafficking steps are indicated by adjacent ND abbreviation in red. PD = Parkinson’s disease; AD = Alzheimer’s disease; HD = Huntington’s disease; ALS = amyotrophic lateral sclerosis. Figure adapted from Kiral et al. with permission.¹⁷

EXPERIMENTAL MODELS OF TRAFFICKING DEFECTS IN VITRO AND PHENOTYPIC SCREENING PLATFORMS

In efforts to study neurotoxicity and to identify neuroprotective compounds, a series of cellular models of neurodegeneration have been established, varying in ease of use and therapeutic translatability.^29,30 Experimental models in simpler organisms like Saccharomyces cerevisiae (yeast) have proven to be hugely useful for preliminary screening efforts.^31–35 The use of yeast is favorable due to its ease of handling, rapid growth rate, and robust genetic tractability. Extensive analysis of neurotoxicity models in yeast has revealed conservation of toxicity mechanisms observed in neurodegeneration, providing a robust primary experimental model of ND.

Various mammalian cell models have also been established for studying neurotoxicity and trafficking defects. In general, the use of immortalized mammalian cell lines provides more relevant model systems that are amenable to transfection and genetic manipulation, thus enabling genetic validation of targets or pathways identified through screening efforts.³⁰ For more advanced mammalian cell models, neurons derived from induced pluripotent stem cells (iPSC) or from patients are used.^30,36–39 These systems are particularly advantageous as they can be noninvasively derived from individual patients, enabling generation of patient-specific cell lines. These systems are also amenable to gene editing for generation of isogenic controls and can be used to study protein overexpression and associated toxicity. These models, however, are typically more difficult to culture in the laboratory and require experiments to be performed on a slower time scale due to the slower rate of cell growth.

While cellular models of neurotoxicity enable high-throughput screening of small molecule libraries to identify phenotypically active compounds, the models are limited in therapeutic relevance and are thus coupled with secondary validation in more advanced models. As with cellular models, there is a spectrum of more advanced organismal models that are used to study neurodegeneration. While a thorough discussion of these models is outside of the scope of this review, we want to highlight the burgeoning use of nematode Caenorhabditis elegans (C. elegans) for the study of neurogenerative mechanisms.^40–42 C. elegans has emerged as a useful model system as the organism contains many orthologs to human ND-related proteins, and models of NDs in C. elegans recapitulate many cellular features of human ND pathology. Further, the model is genetically tractable and amenable for use of common biochemical tools yet provides a whole animal model through which behavioral responses can be quantifiably monitored. This system, as demonstrated in some of the screening efforts described below, provides a useful secondary screen that is more advanced than cellular models but precedes rodent or primate models of neurodegeneration.

CELLULAR MARKERS OF TRAFFICKING FOR PHENOTYPIC SCREENING

There are various experimental approaches to monitoring protein aggregation and trafficking defects (Table 1). Commonly, in efforts to discover small molecules that can restore trafficking defects, the trafficking process itself is monitored in parallel with associated cellular markers of toxicity. These markers include generation of reactive oxygen species^43–45 or protein aggregation/localization^45,45–47 as well as global cell viability^45,46 and neurodegeneration (monitored in C. elegans^45,48 and Drosophila^47,49 models).

Table 1.

Experimental Methods To Monitor Trafficking and Cell Viability in Experimental Models of ND Toxicity

trafficking step (or associated toxicity marker)	organism	cellular marker monitored	experimental detection method	ref
ER-Golgi trafficking	Yeast	Carboxypeptidase Y maturation	Radiolabeling and immunoprecipitation	44–46
Autophagic clearance	Mammalian cell lines	Abundance of HA- or EGFP-tagged protein of interest	Immunoblotting	47, 55
		Number of EGFP-LC3 vesicles	Microscopy	55
		LC3-II levels as a measure of autophagosome levels	Immunoblotting	55
		GFP-LC3 puncta	Microscopy coupled with high-throughput image analysis	57
Protein aggregation/localization	Yeast	Localization of GFP-fused protein	Live cell microscopy	45
		Vesicular foci formation	Microcopy with fixed cells and automated detection	46
	Mammalian cell lines	EGFP-fusion aggregation	Fluorescence microscopy	47
Generation of reactive oxygen species	Yeast	CM-H2-DCFDA	Flow cytometry	45
Cell viability	Yeast	Growth rate	OD600	44–46
		Propidium iodide staining	Flow cytometry	45
	Primary neurons (rat)	Dopaminergic viability as indicated by MAP2 and TH levels	Immunocytochemistry	45
	iPSC-derived neurons	Neuronal survival via RFP morphology signal where loss of fluorescence, neurite loss or cell blebbing indicate cell death	Microscopy	46
Neurodegeneration	C. elegans	Neuronal counting and morphology analysis	Microscopy	45, 48
	Drosophila	Photoreceptor degeneration	Pseudopupil technique	47, 49

To monitor trafficking defects, common experimental approaches rely on trafficking-dependent maturation of protein zymogens as an indicator of proper cellular trafficking. In yeast, ER-Golgi trafficking is measured through maturation of vacuolar protein carboxypeptidase Y. Carboxypeptidase Y is synthesized as an inactive zymogen in the ER and is activated once transported to the Golgi. The activation of procarboxypeptidase Y involves a mass change of ~6000 Da, a difference that is typically monitored by SDS–PAGE-coupled immunoblotting.^{44–46,50–52}

By use of a similar experimental approach, mammalian autophagy is often measured as a function of microtubule-associated protein 1A/1B-light chain 3 (LC3) processing.^53,54 LC3 is a soluble, cytoplasmic protein that is ubiquitously expressed in mammalian cells. During autophagy, LC3 is converted from its basal form (LC3-I) to derivative LC3-II via conjugation to phosphatidylethanolamine. LC3-II is recruited to autophagic vesicles and is subsequently degraded when autophagic vesicles fuse with lysosomes to generate autolysosomes. Autophagy and autolysosome processing can therefore be monitored by maturation of LC3-I to LC3-II and by LC3-II degradation, respectively. As with carboxypeptidase Y maturation in yeast, LC3 processing is often monitored by immunoblotting, immunoprecipitation, or immunofluorescence.^47,53–57 LC3 localization can also be monitored as a measure of trafficking. In this case, the formation and abundance of LC3-GFP puncta are indicative of autophagosome formation, and this can be monitored by microscopy using a fluorescent reporter LC3 fusion to track autophagy. It should be noted that some of the detection methods for these processes like immunoblotting, in particular, are relatively low throughput. Therefore, it is common to track global viability in a primary screen and then assess trafficking defects in secondary validation experiments of initial hit compounds. Further, trafficking monitoring techniques have recently been made more amenable to screening through use of automated, high-throughput image detection and analysis tools.⁵⁷

PHENOTYPIC SCREENING IDENTIFIES STRUCTURALLY DIVERSE SMALL MOLECULES THAT RESTORE TRAFFICKING DEFECTS IN NDS

Through phenotypic screening efforts, a number of small molecules have been identified that are able to restore ND-associated trafficking defects. To date, these molecules are primarily active through stimulation of ER-to-Golgi trafficking or of autophagy.

Stimulation of ER-to-Golgi Trafficking.

It has been demonstrated that ER-Golgi trafficking is disrupted in multiple NDs, including PD, AD, HD, and ALS.^58–64 Specifically, defects in anterograde (PD, ALS)^18,65 and ER-to-Golgi trafficking as well as in trans-Golgi (AD)^66–68 and post-Golgi (AD, PD, HD)^69–71 secretory trafficking have been observed (Figure 1). Each of these steps is essential for maintenance of proteostasis, and disruption in multiple NDs poses the potential for discovery of small molecules that can restore these pathways in a pan-ND manner.

In efforts spearheaded by the lab of Susan Lindquist and associated therapeutic discovery venture Yumanity Therapeutics, a screening platform was developed that couples a phenotypic primary screen in yeast with genetic target identification and a secondary screen in primary or patient-derived neuron models. With this platform, a number of small molecules have been found that can restore ER-Golgi trafficking defects associated with both PD and ALS.^{31,44–46,72,73}

In a first demonstration of small molecule restoration of ER-Golgi trafficking, Su and co-workers used a yeast-based model to identify a series of 1,2,3,4-tetrahydroquinoline compounds that restored ER-Golgi trafficking induced by PD-associated α-synuclein toxicity (Figure 2A; Table 3).⁴⁵ The compounds had various potencies, with compound 1a demonstrating activity at the lowest concentration followed by 1b and 1c, while 1d showed low activity at all concentrations tested. In primary screening in yeast, the compounds restored cell viability in the presence of α-synuclein toxicity but did not alter α-synuclein levels nor prevent oligomerization of the neurotoxic protein. This activity was not conserved in yeast models of polyQ toxicity (HD), indicating that the effect is specific to PD-associated toxicity and that ligand mechanism is independent of generalized cell stress alleviation. Further testing of commercially available compounds 1e and 1f revealed no activity in restoration of viability or decrease of toxicity. The activity of compounds 1a–d was retained in C. elegans models of PD toxicity, with slight variances of potency, even when administered significantly after the induction of toxicity, indicating that the compounds arrest cell death after the onset of cellular stress. The activity of 1a, 1b, and 1d was further retained in primary neurons transduced with PD-associated mutant α-synuclein-A53T, but the activity of 1c was not. As the structural similarity of active versus inactive derivatives is very high, it was hypothesized that the compounds bound to the same target but with varying affinities. This was proven in a competition assay where the efficacy of compound 1a was decreased in yeast models with coadministration of increasing concentrations of 1e or 1f. This indicates that the compounds bind to the same molecular target but do not all induce the desired phenotypic effect. While a target of the compound was not identified through chemical genetic screening, retention of activity in three organismal models indicates that the target is conserved with minimal structural variation even across distant organisms.

Figure 2.

Phenotypically active small molecules and derivatives that restore ER-Golgi trafficking defects in the presence of ND toxicity. (A) 1,2,3,4-Tetrahydroquinoline compounds (1a–d) restore trafficking in models of PD but show sensitivity to small modifications of the scaffold.⁴⁵ (B) N-Arylbenzimidazole (NAB) lead 2a and optimized derivative 2b restored trafficking defects induced by TDP-43 and α-synuclein toxicity in ALS and PD, respectively.^44,72 (C) Analysis of the NAB scaffold by SAR reveals that amide linker and benzimidazole core are essential for activity while there is some tolerance for small substitutions on the benzylamine and N-aryl moieties.⁴⁴ (D) 1,2,4-Oxadiazole scaffold 3a and more potent compound 3b restore ER-Golgi trafficking in PD by inhibiting fatty acid desaturase Ole1/SCD.⁴⁶ (E) Variation of substituents on the oxadiazole core enabled confirmation of target Ole1/SCD.⁴⁶

Table 3.

Stimulation of Autophagy Was Induced by Numerous Small Molecules and Previously FDA-Approved Drugs in Various Models of HD and PD

compd	ND model	model system	FDA-approved	other purpose	putative target/mechanism	ref
5	HD	Mammalian cell models	Y	antipsychotic	Inhibits dopaminergic D2 receptor	57
6	HD		Y		Blocking adrenergic and dopaminergic neural transmission (23); inhibit excitotoxicity of glutamate (25)	57
7	HD		Y		Aminergic receptors	57
8	HD		Y	Cardiac indications	Inhibitors of intracellular Ca2+ current	57
9	HD		Y			57
10	HD		Y
11	HD, PD		Y	Diarrhea	μ-Opioid receptor agonist; blocks Ca2+ channels in vitro cellular experiments (27, 28)	55, 57
12	HD		N	Neurotoxin	Irreversible blocker of Ca2+-activated K+ channel	57
13	HD, PD		Y	Hypertension	L-type Ca2+-channel antagonist	55
14	HD, PD		Y	Hypertension	L-type Ca2+-channel antagonist	55
15	HD, PD		Y	Sedative, hypertension	α2-adrenergic and ILR receptor agonist	55
16	HD, PD		Y	Vasodilator	ATP-sensitive K+ channel agonist	55
17	HD, PD	Yeast, mammalian cells, Drosophila	na			47
18	HD, PD		na			47
19	HD, PD		na			47
20	PD	Mammalian cells	na			104

Using a similar screening platform, Tardiff and co-workers identified another scaffold, characterized by an N-arylbenzimidazole (NAB) core, that demonstrated phenotypic rescue of α-synuclein toxicity and restoration of ER-Golgi trafficking (Figure 2B; Table 2).⁴⁴ Compounds were initially screened in yeast models, where the NAB scaffold was first identified to restore viability in yeast models of TDP-43-associated ALS. Further analysis of the lead NAB compound 2a revealed that activity was retained in yeast models of α-synuclein toxicity (PD), and subsequent studies in secondary screens revealed activity was conserved in C. elegans and in iPSC-derived neuron models of α-synuclein toxicity.^44,72 Structure–activity analyses of the scaffold revealed that the benzimidazole core and amide linker are essential for activity but the scaffold is moderately tolerant of substitutions on the N-aryl and benzylamine moieties (Figure 2C). SAR provided an improved ligand 2b, with the only modification from lead compound 2a being addition of ortho-methyl substituent on the N-aryl moiety. The improved compound 2b rescued viability in the presence of α-synuclein toxicity with an EC₄₀ of 4.5 μM, demonstrating moderate activity and indicating that affinity and potency may need further optimization if the compound were to be carried forward.

Table 2.

Phenotypically Active Small Molecules That Restore ER-Golgi Trafficking Defects in Models of NDs

Unlike studies of the 1,2,3,4-tetrahydroquinoline scaffold, chemical genetic screening was performed to identify putative targets of the NAB scaffold. These analyses, initially performed in yeast and then in iPSC-derived neurons, revealed the target to be E3 ubiquitin ligase Rsp5 or its mammalian homolog Nedd4, respectively. The identification of Rsp5/Nedd4 as a target was promising for a number of reasons, particularly because Nedd4 has been previously implicated in the ubiquitin-dependent rescue of α-synuclein toxicity^74–78 and E3 ligases have become of interest as potential druggable targets. Despite the genetic identification of Nedd4 as a putative target, target engagement of the NAB scaffold with target Nedd4 was not demonstrated until recent efforts by our own group to expand our understanding of the NAB mechanism in rescue of α-synuclein toxicity.⁷⁹ Through the first analyses and our lab’s own contributions, it was established that NAB2 binds to Nedd4 but does not alter its enzymatic activity or conformation in vitro. Subsequent proteomic analyses indicate that NAB2 may alter Nedd4 specificity but that its mechanism of action is nuanced and other proteins may be involved in its phenotypic activity.⁷⁹

In a final report of small molecule rescue of ER-Golgi trafficking, Vincent and co-workers of Yumanity Therapeutics recently reported a series of 1,2,4-oxadiazoles that demonstrate cytoprotective activity in yeast models of α-synuclein toxicity (PD) (Figure 2D, Table 2).⁴⁶ In these primary yeast-based screens, 1,2,4-oxadiazoles 3a and 3b, along with other analogues tested, did not retain activity in models of β-amyloid (AD) or TDP-43 (ALS) toxicity indicating a PD-specific effect. Through yeast-based chemical genetic screening and analysis of drug-resistant mutants, it was determined that these compounds (3a,b) restored viability in yeast-based models and restored ER-Golgi trafficking by targeting Ole1, the sole fatty acid desaturase expressed in yeast. SAR of the oxadiazole core (Figure 2E) in complement with mutational analyses of Ole1 further established Ole1 as the target of phenotypically active 1,2,4-oxadiazole compounds. Chemical genetic screening revealed that deletion of genes encoding Mga2 and Ubx2, which serve as a transcription factor and bridging element, respectively, and regulate Ole1 expression, induced sensitivity to 3b. Finally, the activity of 3b in rescuing α-synuclein toxicity was mitigated when cells were supplemented with oleic and palmitoleic acid, the enzymatic products of Ole1. Thus, it was established that 3b serves to rescue α-synuclein toxicity by inhibiting Ole1 activity. It was further confirmed that 1,2,4-oxadiazole activity mitigated ER-Golgi trafficking defects induced by α-synuclein toxicity. As in previous screening efforts, activity was then tested in mammalian models of α-synuclein toxicity where it was confirmed that 1,2,4-oxadiazole activity was conserved and dependent upon stearoyl-coA desaturase (SCD), the mammalian homolog of Ole1. SCD has been separately identified as a target for neurotoxicity treatment, providing additional validation to the mechanism and target of the 1,2,4-oxadiazoles.^80,81

The compounds identified to date demonstrate the power of robust phenotypic screening platforms for identifying cytoprotective small molecules and associated targets in the restoration of ER-Golgi trafficking defects. In all cases, restoration of trafficking occurred in parallel with improved viability and alleviation of other markers of toxicity (protein aggregation, generation of reactive oxygen species, etc.). This furthers the idea that trafficking defects are intricately linked to other neurotoxic defects as it is not inherently clear if trafficking restoration is a direct effect of the small molecule or a secondary effect resulting from the small molecule alleviating other markers of toxicity. Further, the identified targets of the NAB scaffold (2a,b) and 1,2,4-oxadiazole scaffold (3a,b) are not directly implicated in trafficking regulation itself but are instead associated with the ubiquitination system^82–84 (Rsp5/Nedd4) and fatty acid metabolism^85–88 (Ole1/SCD), pathways that respectively are dependent upon or directly affect endomembrane trafficking. The identification of targets by subsequent chemical genetic screening demonstrates the utility of phenotypic screens for discovery of active small molecules as enzymes Nedd4 and SCD may not have inherently be the focus of a targeted drug discovery platform for alleviation of neurotoxicity. The idea of phenotypic screens as a powerful tool for neurodegeneration drug discovery was recently discussed further in a perspective by Brown and Wobst.⁸⁹

When considering the generalizability of cytoprotective molecules, it is particularly exciting that the NAB scaffold demonstrated activity in models of both PD and ALS. While putative target Nedd4 is strongly linked to α-synuclein in PD,^44,74–78 there is little evidence to date of a link between Nedd4 and ALS-associated protein TDP-43 outside of a recent report that Nedd4 expression is upregulated in neurons with ALS toxicity relative to control neurons.⁹⁰ If other proteins are implicated in the NAB mechanism of action, we anticipate that the targets are involved in trafficking pathways directly affected by both α-synuclein and TDP-43 toxicity, underlying the generalizable activity of the NAB scaffold in rescue of ER-Golgi trafficking defects induced by both types of cellular stress. The 1,2,3,4-tetrahydroquinoline scaffold (1a–d) and 1,2,4-oxadiazole scaffold, on the other hand, demonstrated activity in models of α-synuclein toxicity (PD) but not in polyQ toxicity (HD) or in β-amyloid (PD) or TDP-43 (ALS), respectively. This result, in contrast with the activity of NABs, indicates that targets and compounds may not be consistently generalizable despite similarities in the trafficking pathway disrupted by toxicity and restored by small molecule treatment thereof. We anticipate, however, that there is utility in screening phenotypically active compounds in more than one model of NDs, especially if there is evidence of putative targets being affected in more than one type of neurotoxicity.

Stimulation of Autophagy.

The process of autophagy is largely responsible for the degradation of cytosolic proteins or organelles and requires formation of autophagic vacuoles (also called autophagosomes) that subsequently fuse with lysosomes to generate autolysosomes.^91–94 Following autolysosome formation, the contents of the membrane-bound compartment are degraded by acidic hydrolases. This pathway, in addition to the ubiquitin proteasome system, is essential for maintenance of cellular proteostasis and is tightly regulated. This process is known to be disrupted across NDs, with demonstrated defects in models of PD, ALS, HD, AD, and tau-associated dementias.^82,94–98 In these disorders, functional autophagic clearance of neurotoxic proteins is particularly important as the oligomerized state of the proteins makes them inaccessible to the proteasomal machinery.^99,100 Thus, there have been significant efforts toward identifying methods to stimulate autophagy as a means to clear neurotoxic proteins from the cell and to restore cell viability.

The first known activator of autophagy was natural product rapamycin (4),^101,102 which stimulates autophagy through protein mammalian target of rapamycin (mTOR) (Figure 3A). While the stimulation of autophagy is favorable, the TOR proteins also regulate other essential cellular processes like ribosome biogenesis and protein translation,¹⁰³ likely contributing to the unfavorable therapeutic profiles associated with prolonged rapamycin use. Therefore, there have been significant efforts to identify autophagy activators that act independent of the mTOR or rapamycin-mediated pathway.

Figure 3.

Small molecules stimulate autophagy via various targets and mechanisms in models of PD and/or HD. (A) Natural product rapamycin, the first compound identified as an enhancer of autophagy, was used as a positive control in screening for autophagy regulators. (B) Screening efforts by Zhang and co-workers identified compounds 5–11, all previously FDA-approved drugs, as well as 12 as stimulators of autophagy in cell models of HD.⁵⁷ (C) Williams et al. expanded the class of autophagy activators by demonstrating activity of 13–16, as well as compound 6, in models of both HD and PD.⁵⁵ (D) Sarkar and co-workers identified structurally diverse compounds as novel activators of autophagy and demonstrated synergistic activity with rapamycin.⁴⁷ (E) Compound 20 stimulates autophagy by altering expression of autophagy related genes.

To this end, Zhang and co-workers developed an image-based high-throughput screen in mammalian cells to identify small molecules that increased the number of autophagosomes (as indicated by LC3-GFP puncta, Table 1).⁵⁷ This screen was coupled with a protein degradation assay to ensure that increased LC3-GFP puncta levels coincided with increased protein degradation, providing a two-pronged approach to study autophagy rates. By use of this platform, eight compounds were identified from a library of 480 small molecules which stimulated autophagy in wild-type mammalian H4 cells. While these compounds were not initially screened in models of neurodegeneration, all eight retained activity when tested in cellular models of HD by improving autophagic clearance of polyQ protein accumulation. Interestingly, seven of the eight compounds have been previously FDA-approved for other purposes, providing promising leads with established therapeutic tolerability (Figure 3B, Table 3). Compounds 5–11 have been approved for a variety of pharmacological uses, ranging from antipsychotic drugs (5–7) to cardiovascular treatment (8–10) to diarrhea (11) (see Table 3). The only compound which was not FDA approved was penitrem A (12), a fungal neurotoxin which is known to induce toxicity in vivo despite not showing toxicity in the H4 model used in the screen. In these cases, the specific mechanism of compounds 5–12 in stimulating autophagy is not known. Further, the utility of these compounds in stimulating autophagy was not tested outside of the model of HD described. In a separate screening effort, Tsvetokov et al. independently identified trifluoperazine (6) as a stimulator of autophagy and demonstrated that structurally related analogues (10-NCP, promethazine, promazine, chlorpromazine, triflupromazine), all of which contain a phenoxazine or phenothiazine core, also induce autophagy in models of HD (structures not shown).⁵⁶

In another effort, Williams and co-workers sought autophagy enhancers in a library of compounds that have previously been administered to humans without major toxic side effects.⁵⁵ In this case, a library of 253 compounds were screened first in two mammalian cell models of PD, where toxicity was induced by overexpression of PD-associated mutants α-synuclein-A30P and α-synuclein-A53T. From this screen, seven hits were identified that fell into three categories: antagonists of L-type Ca²⁺ channels, an agonist of an ATP-sensitive K⁺ channel, and a compound that activates G_I-protein coupled pathways via α₂-adrenergic and type I imidazole receptors (I1R). Of these hits, five were prioritized and screened against mammalian cell models of HD toxicity (Figure 3C; Table 3). All five compounds screened were effective in enhancing clearance of soluble mutant huntingtin exon 1 with a polyQ expansion in PC12 cells and reduced aggregation and toxicity in a neuroblastoma-derived cell line. These compounds include loperamide (11), which was also identified by Zhang and co-workers as an enhancer of autophagy,⁵⁷ verapamil (13), nimodipine (14), clonidine (15), and minoxidil (16). The conservation of loperamide (11) across two independent screening efforts provides strong evidence for its utility in the autophagic clearance of neurotoxic proteins. Further, the identification of verapamil (13) and nimodipine (14), which, like loperamide, target L-type Ca⁺ channels, reinforces the utility of this target in stimulating the pathway. Additionally, the core scaffold of nimodipine (14) is also present in active compounds nicardipine (8) and niguldipine (9) identified by Zhang et al.⁵⁷

In an alternative approach to identify small molecule modulators of autophagy, Sarkar and colleagues identified small molecules that could both inhibit and enhance rapamycin-induced autophagy in yeast.⁴⁷ These compounds, called small molecule inhibitors of rapamycin (SMIRs) or small molecule enhancers of rapamycin (SMERs), were identified by screening in yeast cultures pretreated with rapamycin under the hypothesis that compounds that were identified as SMERs would also demonstrate the ability to enhance autophagy as single agents. From the primary screen, both SMIRs and SMERs were identified and carried forward to a secondary screen in mammalian models of PD toxicity induced by α-synuclein-A53T. In this screen, three SMERs demonstrated significant clearance of α-synuclein-A53T (compounds 17–19; Figure 3D, Table 3) in the absence of rapamycin treatment. These compounds demonstrated further activity in models of HD (polyQ expansion of huntingtin exon 1) without indication of cytotoxic side effects. It was further confirmed that these compounds directly affect autophagy as activity was diminished in cells lacking an essential autophagy gene (Atg5^−/−) and treatment of wild-type cells increased the formation of EGFP-LC3 vesicles and increased LC3-I processing to LC3-II. As in yeast, this activity was enhanced in the presence of rapamycin treatment, and activity was conserved in polyQ toxicity (HD) in Drosophila. Importantly, the lack of structural redundancy in the identified SMERs indicates that the small molecules likely act via different targets or mechanisms, and the demonstration of activity in models of both PD and HD suggests that the targets are conserved across ND pathologies.

As a final demonstration of ligand-induced stimulation of autophagy, Kilpatrick and co-workers demonstrated that methyl-β-cyclodextrin (compound 20; Figure 3E, Table 3) enhanced autophagic clearance of α-synuclein by increasing the nuclear localization of transcription factor EB (TFEB).¹⁰⁴ TFEB regulates expression of genes that control autophagosome and lysosome biogenesis.^105–107 Therefore, chemical stimulation of autophagy-related gene expression through modulation of transcription factors provides an orthogonal method for enhancing autophagic clearance of cytotoxic proteins.

Through the screening efforts described, it is evident that there are several mechanisms and targets by which autophagy can be stimulated to alleviate the cytotoxic burden of protein aggregation in PD and HD. The identification of compounds that are active in both PD and HD models provides further evidence that some targets may be generalizable across NDs. While further efforts to identify the specific targets and mechanisms of each of these compounds is necessary to fully understand their activity and generalizability, the preliminary characterization is promising. Further, a large proportion of the compounds identified are already FDA-approved for other purposes or have been administered to humans without toxicity, providing promise that these compounds may be tolerable as therapeutics. It is not clear, however, whether the compounds will demonstrate sufficient blood–brain barrier permeability to elicit neuroprotective activity in humans.

Other Approaches To Restore Trafficking Defects.

Other efforts have been made to target systems upstream of endomembrane trafficking that can affect protein transport and proteostasis. For instance, there have been numerous efforts to regulate the unfolded protein response (UPR), a stress response system that is triggered in the ER upon accumulation of unfolded proteins.^59,108–113 While this is independent of endomembrane trafficking between organelles, the ER is a key component of the secretory pathway, and thus, disruption of ER function can also alter the efficiency of downstream trafficking steps. While the UPR in the context of neurodegeneration^108,110,111 and methods to target the UPR with small molecules^{112,114–121} have been widely studied, it is outside the scope of this review and has been discussed extensively elsewhere.

Another area of research in the context of trafficking defects in NDs has focused on directly targeting proteins that control trafficking processes. Endomembrane trafficking is tightly regulated by a number of mechanisms, including numerous small GTPase enzymes.^{12,17,122–124} As the GTPases are widely understood to regulate trafficking steps affected in neurodegeneration, efforts have been made to identify small molecules that can alter GTPase activity or regulation (by targeting protein–protein interactions)^125,126 to alter associated trafficking steps. There has been some success in this area, with identification of pan-GTPase activators^127,128 or inhibitors of some Rab GTPases,^129–132 but the development of small molecules specific to a single GTPase is made difficult by the high structural similarity among members of the GTPase classes. Further, there is not precedence for the efficacy of these compounds in the context of ND models to date.

Utility for Future Screening and SAR Studies.

As demonstrated by the efforts described above, the experimental approaches employed show promise for use in future discovery and optimization efforts. In particular, the yeast-based primary screening approach implemented by the Lindquist lab and Yumanity therapeutics shows particular promise.^{44–46,72,80} In the cases described, the yeast-based primary screen enabled high-throughput screening efforts while retaining sufficient sensitivity to detect subsequent SAR-driven lead optimization, with ~10-fold differences in activity detected. The method also allows parallel monitoring of trafficking defects and associated toxicity markers, enabling identification of molecules that address multiple facets of ND toxicity. Finally, it afforded hit compounds that retained activity once advanced to higher order model systems.

Other screening efforts, such as those by Williams et al.⁵⁵ and Zhang et al.⁵⁷ that focused on mammalian models of autophagy defects, proved effective in identifying bioactive compounds. The efforts reported, however, primarily focused on repurposing known bioactive molecules and did not include SAR optimization. Therefore, it is not directly clear if the experimental assays employed provide sufficient sensitivity to detect SAR-driven differences in activity. Despite this, the assays employed are similar to those used in secondary screens by Lindquist and co-workers. Thus, it is possible (though not directly evidenced) that these methods will be useful for hit identification and SAR optimization.

SUMMARY AND OUTLOOK

For the ligands discussed, there is still limited evidence as to whether the compounds will retain activity in higher models of NDs or as therapeutic, neuroprotective agents. A primary limitation in these analyses is that demonstrated activity, even in various cell and organismal models, does not ensure activity in higher order organisms as these experiments do not provide insight into the tolerability or bioavailability of the compounds. Even though some of compounds identified (5–11, 13–16) have been FDA-approved and do not show toxicity in humans, the compounds may not be bioavailable in the brain due to lack of permeability across the blood–brain barrier (BBB). For instance, loperamide (11), which was identified in two independent screens demonstrated activity in the stimulation of autophagy in cellular models of HD and PD, but does not cross the BBB.⁵⁷ Therefore, its utility as a neuroprotective therapeutic is limited.

Despite this, there is precedence for these screening platforms to provide new targets and hit compounds that can be moved forward for development of novel therapeutics for neurodegeneration. Yumanity Therapeutics, for example, is conducting clinical trials for a number of ND treatments, including testing of SCD inhibitors for treatment of PD and of dementia with Lewy bodies as well as two undisclosed programs for treatments of ALS/frontotemporal lobar degeneration (FTLD) and one for PD, AD, ALS/FLTD.^133,134 Progress in these clinical trials and research programs is particularly exciting and timely with the recent FDA-approval of aducanumab for treatment of AD.¹³⁵ Though controversial,^136–138 the approval of aducanumab will hopefully stimulate additional attention and energy toward development and advancement of other therapies for AD and for NDs more broadly.

From a basic science perspective, the discovery of compounds that can activate trafficking pathways also provides a suite of chemical probes that can be used to explore the role of these pathways in different disease or cellular models. While there are many well-established chemical tools to selectively inhibit steps of trafficking pathways,¹³⁹ there are fewer probes available to activate pathways of interest. This is a particularly compelling use for compounds like 17 and 18, which were first identified for their ability to stimulate autophagy in the absence of cytotoxic pressure in models of PD or HD. The discovery and characterization of these compounds, therefore, expand our chemical toolbox, may improve our ability to experimentally interrogate the endomembrane trafficking system, and may help us to better identify and validate targets.

Finally, for the compounds discovered for which a cellular target is not known, elucidation of targets could enhance our understanding of the compound mechanism and of endomembrane regulation. This may enable target-driven optimization of scaffolds for improved activity, tolerability, or bioavailability in relevant animal models of neurodegenerative diseases.