Small Molecule Stabilization of PINK-1/PINK1 Improves Neurodegenerative Disease

Macroautophagic recycling of dysfunctional mitochondria, known as mitophagy, is essential for mitochondrial homeostasis and cell viability. Accumulation of defective mitochondria and impaired mitophagy have been widely implicated in many neurodegenerative diseases, and loss-of-function mutations of two regulators of mitophagy, PINK1 and Parkin, are amongst the most common causes of recessive Parkinson’s disease. Activation of mitophagy via pharmacological treatments may be a feasible approach for combating neurodegeneration. In this effort, we screened ∼45,000 small molecules for the ability to activate mitophagy. A high-throughput, whole-organism, phenotypic screen was conducted by monitoring stabilization of PINK-1/PINK1, a key event in mitophagy activation, in a Caenorhabditis elegans strain carrying a Ppink-1::PINK-1::GFP reporter. We obtained eight hits that induced mitophagy, as evidenced by increased mitochondrial fragmentation and autophagosome formation. Several of the compounds also reduced ATP production, oxygen consumption, mitochondrial mass, and/or mitochondrial membrane potential. Importantly, we found that treatment with two compounds, which we named PS83 and PS106 (more commonly known as sertraline) reduced neurodegenerative disease phenotypes (including delayed paralysis in a C. elegans Alzheimer’s model) in a PINK-1/PINK1-dependent manner. This report presents a promising step toward the identification of compounds that will stimulate mitochondrial turnover.

we named PS83 and PS106 (more commonly known as sertraline) reduced neurodegenerative disease Introduction 31 Although they are often simplistically characterized as the "powerhouse of the cell", mitochondria 32 have a wide range of cellular functions beyond that role, including amino acid metabolism, regulation of 33 iron and calcium homeostasis, production of reactive oxygen species (ROS), stress surveillance, and 34 control of apoptosis and other programmed cell death pathways (1-4). Considering the number and 35 variety of roles that they play, it is not surprising that mitochondrial maintenance is crucial for the health 36 of cells and organisms. 37 Unfortunately, cells have not many methods to repair mitochondrial damage. When damage is 38 limited, mitochondria fuse, allowing their content to be mixed and sorted, segregating the damaged 39 material (by an unknown mechanism) into low-quality mitochondria targeted for destruction via 40 macroautophagic degradation (hereafter referred to as mitophagy) (5). The best-known pathway for 41 triggering mitophagy is the PINK1/Parkin pathway, both of whose namesake members have been linked 42 to Parkinson's disease (5-7). Activation of this pathway begins with the PTEN-induced kinase 1 (PINK1), a 43 serine/threonine kinase that is constitutively expressed and targeted to mitochondria (5). Upon arrival, 44 healthy mitochondria will import the kinase and it will be destroyed by matrix-resident proteases. If 45 mitochondria are damaged, PINK1 accumulates on the outer membrane, where it cross-phosphorylates 46 and high-throughput compounds screening due to its small size, short generation and lifespan, and the 71 ability to simultaneously counterscreen for toxic compounds (24). Finally, as C. elegans is transparent, 72 intact worms can be used for high-content screens where more nuanced or complex phenotypes can be 73 used. 74 In this study, we used a high-throughput, high-content phenotypic screen using C. elegans carrying 75 GFP tagged full-length PINK-1/PINK1 (Ppink-1::PINK-1::GFP) to identify a panel of small molecules that 76 stabilize the kinase and verified their potentials in the activation of mitophagy. Four of the eight tested PS 77 molecules significantly reduced mitochondrial membrane potential, while two of the others reduced 78 mitochondrial mass loss. Neuroprotective properties of the selected PS compounds were tested in a 79 transgenic C. elegans model of Alzheimer's disease, where we discovered that two of the compounds 80 significantly delayed paralysis in a PINK-1/PINK1-dependent manner. One of the two compounds was also 81 able to reduce aggregate formation in a C. elegans polyglutamine-protein aggregation model. The 82 compounds generally showed relatively low toxicity to human astroglial and prostate epithelial cells, 83 increasing their promise. 84 85

Results 86
High-throughput screening of compound libraries identifies eight compounds that stabilize PINK-1/PINK1 87 As noted previously, preventing the degradation of PINK1 leads to activation of mitophagy (26-28) 88 ( Fig. 1a). To identify small molecules that promote PINK1 stability, we leveraged a C. elegans strain 89 carrying a GFP-tagged, full-length PINK-1/PINK1 driven by its native promoter (9,29). This reporter 90 provides a simple method to track PINK-1/PINK1 stabilization by visualizing GFP. To develop a high-91 throughput, high-content phenotypic screen in C. elegans, we optimized parameters by identifying 92 efficient conditions for the mitophagic activation by sodium selenite (Na2SeO3), which triggers the 93 production of mitochondrial superoxide (30, 31) (Fig. 1b). 94 DMSO mostly showed this state (>85%, Fig. 2a) 119 Qualitative analysis of mitochondrial network structure showed that treatment with the PS 120 compounds significantly increased the tendency towards intermediate (partially fragmented) or punctate 121 (strongly fragmented) mitochondrial network. Four of the compounds, PS30, PS34, PS127, and PS143, 122 induced substantial fragmentation, as shown by the dissolution of the tubular network into discrete 123 punctae. PS83, PS103, PS106, and PS135 induced a milder disruption of the mitochondrial network, 124 although each was statistically significant. Disruption of mitochondria fission-fusion homeostasis is 125 consistent with existing evidence that mitochondrial fragmentation is necessary for mitophagy to occur 126 (10, 37). 127 To confirm the formation of autophagosomes, we exposed a worm strain expressing mCherry::LGG-1 128 to the PS compounds. Under normal conditions, LGG-1, and its mammalian ortholog MAP1LC3, exhibit a 129 diffuse cytoplasmic localization (38). During the formation of the isolation membrane (which will become 130 the autophagosome), MAP1LC3 is crosslinked to the lipids that will comprise this membrane, converting 131 the localization from a diffuse pattern into bright punctae (38, 39). Punctae were qualitatively assessed 132 as few, medium, or many, with representative images shown (Fig. 2b). Compared to the DMSO control, 133 several of the PS compounds increased punctae localization of mCherry::LGG-1/LC3, including two strong 134 hits (PS30 and PS106) and three weak hits (PS83, PS103, and PS135), indicating increased autophagosomal 135 formation. Intriguingly, three compounds (PS34, PS127, and PS143) reduced the production of 136 autophagosomes, an unexpected outcome from stimulating the accumulation of PINK-1/PINK1. Two 137 mitochondrial disruptors, rotenone (which blocks Complex I of the electron transport chain (ETC)) and 138 carbonyl cyanide m-chlorophenyl hydrazine (CCCP, a proton uncoupler that dissipates the electrochemical 139 gradient on the mitochondrial membrane), were used to validate the assay. 140 After formation of the autophagosome is completed, the next step in mitophagy is for the 141 autophagosome to fuse with lysosomes to form autophagolysosomes (40). To observe this process, 142 worms were treated with PS compounds, CCCP, rotenone, or vehicle control for 15 h. Afterward, worms 143 were stained with LysoTracker Red DND-99, a dye that specifically labels acidic cellular compartments. 144 This dye is routinely used to label lysosomes and autophagolysosomes (41,42). Uptake was quantitatively 145 measured by using flow vermimetry (43). Increased LysoTracker Red fluorescence was seen after 146 treatment with PS34, PS127, PS143, and CCCP, but not with the rest of the PS compounds ( Fig. 2c) (44). 147 In contrast, treatment with PS30, PS83, PS103, PS106, or PS135 fragmented mitochondria but did not 148 result in substantial formation of acidified organelles. This outcome was also seen for rotenone, which 149 triggered the early stages of mitophagy (i.e., PINK-1/PINK1 stabilization, formation of the isolation 150 membrane and autophagosome) but precluded acidification of the autophagosomes, preventing 151 increased fluorescence from LysoTracker Red. This is consistent with reports elsewhere regarding the 152 accumulation of autophagosomes and decreased autophagic completion after rotenone exposure (45, 153 46). This failure has been attributed to ATP depletion, which prevents the lysosomal vacuolar ATPase from 154 consuming ATP to acidify the autophagolysosome (45). 155 Worms carrying a ubiquitously expressed firefly luciferase to provide real-time readout of ATP (47, 156 48) were treated with each of the eight PS compounds. Compounds that greatly induced mitochondrial 157 fragmentation (PS30, PS34, PS127, and PS143, Fig. 2a) also caused significant drop in ATP production ( Fig.  158 2d). Unexpectedly, there was no apparent correlation between ATP depletion and failure to acidify 159 autophagolysosomes. 160 To investigate on whether ATP production failure was due to the inhibition of the ETC, we monitored 161 the last stage of the chain by measuring oxygen consumption rate (OCR) (49). Only PS34, PS127, or 162 rotenone significantly lowered oxygen consumption rate (with PS127 completely abolishing respiration) 163 ( Fig. 3a). 164

PS compounds show different effects on multiple mitochondrial parameters 166
To identify the mechanisms giving rise to mitophagy after compound treatment, several 167 mitochondrial phenotypes were assayed. First, mitochondria were stained with MitoTracker Red,which 168 accumulates in mitochondria proportionally to their membrane potential. Nonyl-acridine orange, a dye 169 that binds to cardiolipin and is comparatively insensitive to mitochondrial membrane potential (50, 51), 170 was used as a proxy to measure mitochondrial mass. CCCP, which dissipates mitochondrial membrane 171 potential (52), was used as a control (Fig. 3b). CCCP reduced both mitochondrial mass and mitochondrial 172 membrane potential, which is consistent with uncoupler treatment. 173 Significant reduction of mitochondrial membrane potential was observed in worms treated with PS30, 174 PS103, PS106, and PS135 (Fig. 3b). This reduction did not appear to be accompanied by corresponding 175 decrease in mitochondrial mass, may be indicative of weak uncoupling activity. These data suggest that 176 for at least four compounds, the loss of mitochondrial membrane potential might be the trigger for 177 mitophagy activation. In contrast, treatment with PS34 or PS127 reduced apparent mitochondrial mass, 178 but not mitochondrial membrane potential, which may indicate that there are fewer mitochondria with 179 increased membrane potential. We previously observed similar changes when worms' diet was 180 supplemented with vitamin B12, which improved mitochondrial health (53). Neither PS83 nor PS143 181 significantly affected mitochondrial membrane potential or mass (Fig. 3b). Combined, these results 182 suggest that the loss of ATP content in PS30 and PS135 was not due to failure of the mitochondrial ETC, 183 as in the case for PS34 and PS127. Instead, PS30, PS103, and PS135, and PS106 to a lesser extent, may 184 have mild uncoupling activity. 185 Another common reason for the induction of mitophagy is the accumulation of ROS (54, 55). To test 186 whether the PS compounds induced ROS production, worms were treated with compounds for 15 h, and 187 then were stained with dihydroethidium, a non-fluorescent, redox-sensitive dye that is converted to 188 fluorescent 2-hydroxyethidium by reaction with superoxide (56). Surprisingly, none of the compounds 189 appeared to significantly increase ROS production (Fig. 3c). The positive control, CCCP, validated that the 190 assay was being performed correctly (Fig. 3c) (Fig. 4a). The same four compounds also activated the 199 conserved SKN-1/Nrf2 pathway, as shown by a transcriptional reporter (Pgst-4::GFP (58)) that is 200 commonly used to confirm SKN-1 activation, albeit to a lower level than DAF-16 (Fig. 4b). 201 Mitochondrial dysfunction that is sufficient to trigger mitophagy has also been shown to activate 202 innate immune pathways (59-61). To test whether this occurred after treatment with these compounds, 203 a worm strain carrying Pirg-5::GFP, a transcriptional reporter for the PMK-1 innate immune pathway (62), 204 was exposed to the PS compounds or to RPW-24, a positive control. Only three of the compounds, PS30, 205 PS135, and PS143, induced PMK-1 pathway activity (Fig. 4c). 206 207 Two compounds, PS83 and PS106, show neuroprotective effects 208 Defects in mitochondrial clearance have long been linked to declined physiological function and NDD 209 (as reviewed in (63)). On this basis, it is reasonable to hypothesize that the impacts of the compounds on 210 mitochondria may stimulate protective effects in C. elegans models of NDD. A C. elegans model of 211 Alzheimer's disease was used to test this prediction. This strain, GMC101, produces full-length human β-212 amyloid in body wall muscles (64). Once the peptide has been expressed, shifting the strain to a higher 213 temperature leads to β-amyloid aggregation and paralysis. Each compound was tested at 2-4 different 214 concentrations (Fig. S2). Two compounds, PS83 and PS106, substantially reduced paralysis at 5 µM and 215 25 µM, respectively, and were comparable to the positive control metformin (Fig. 5a-c and Fig. S2). 216 Another compound, PS103, provided a more modest, but still significant decrease in paralysis at 25 µM 217 ( Fig. S2). Alternatively, it is possible that PS83-and PS106-mediated rescue was an artifact and was 218 independent of PINK-1/PINK1 stabilization. To test this, RNAi was used to knock down pink-1 expression 219 in the β-amyloid-expressing strain prior to compound exposure. Consistent with our interpretation, pink-220 1(RNAi) completely removed the ability of PS83 or PS106 to delay paralysis, and made worms treated with 221 PS83 or PS106 indistinguishable from vehicle controls (Fig. 5d, e). 222 To test whether the compounds had a broad effect on NDD models, we obtained a worm strain that 223 expresses a YFP-tagged protein with an engineered polyglutamine (polyQ) repeat of 82 consecutive 224 glutamine residues (Q82::YFP) (65). PolyQ repeats are causative for at least ten different 225 neurodegenerative diseases, with the best-known being Huntington's chorea (66). Expression of the 226 chimeric Q82::YFP product under the control of a tissue-specific promoter (e.g., unc-54 or vha-6) causes 227 Q82::YFP aggregation in the target tissue (65). 228 Young adult worms expressing the Q82::YFP construct under the intestinal promoter vha-6 were 229 treated with PS83 (5 µM) or PS106 (25 µM) for 24 h, and then aggregates were manually counted under 230 low magnification (Fig. 6). We found that PS83 showed no clear difference from vehicle alone (Fig. 6a), 231 while PS106 treatment significantly reduced the number of aggregates (Fig. 6b). Previously, the Morimoto 232 lab implicated activated DAF-16/FOXO in limiting Q82 aggregation and paralysis when aggregates form in 233 body wall muscles (65). Since PS106 did not induce DAF-16 nuclear localization, the precise regulatory 234 pathway induced by PS106 and its role in providing neuroprotection will need to be further elucidated. 235

PS compound exposure shows limited toxicity 237
While activation of mitophagy can provide a means to overcome some aspects of NDD, overactivation 238 of mitophagy may lead to excess mitochondrial loss, bioenergetic deficits, and cellular death (67). To test 239 whether prolonged exposure to PS compounds causes death, survival of C. elegans and human cells was 240 measured. For C. elegans, worms were exposed to the eight PS compounds at four different 241 concentrations for 72 h and then were incubated with Sytox Orange, a cell-impermeant dye that stains 242 DNA in dead worms. Treatment with most compounds showed greater than 75% of survival (as 243 normalized to the DMSO control) at concentrations of 25 µM or less (Fig. 7a). In all cases, worms survived 244 the lowest tested dosage; in all but one (i.e., PS83), the highest dose caused at least partial death (Fig.  245 7a). This suggests that some optimization would be necessary to see protective effects. Importantly, the 246 two compounds that reduced paralysis rate in the C. elegans Alzheimer's model (PS83 and PS106) did not 247 impair survival at tested concentrations of up to 100 µM. 248 Cytotoxicity of chronic, 72 h exposure to PS compounds at similar concentrations was also measured 249 in two human cell lines, SVG-P12 (an astroglial cell line) and RWPE-1 (prostate epithelial cells) (Fig. 7b, c). 250 Differential Hoechst / propidium iodide staining was used to assess cell death. Exposure to either PS103 251 or PS106 at concentrations higher than 10 µM showed substantial toxicity in both cell lines. Interestingly, 252 PS34, PS127, PS135, and PS143 were less toxic to human cells than to C. elegans. 253 254 Discussion 255 Stimulation of mitophagy has proven to be a promising therapeutic target for neurodegenerative 256 diseases (68) and may be beneficial for healthy aging. Using a high-throughput, high-content phenotypic 257 screen, we obtained and characterized eight PINK-1/PINK1-stabilizing compounds. Interestingly, two of 258 the compounds had previously been associated with alterations in mitochondrial function. 259 PS103, commonly known as triclosan or irgasan, has been linked with a variety of mitochondrial 260 dysfunction, including uncoupling of the mitochondrial membrane potential by reversible protonation of 261 the phenoxy group (69-71) and inhibition of Complex II of the ETC (72). Triclosan has also been associated 262 with increased mitochondrial ROS, reduced mitochondrial mass, and disruptions in mitochondrial 263 morphology (73). The ability of triclosan to cause several types of mitochondrial damage, apparently with 264 different proximal factors (72), somewhat reduces its value as a therapeutic agent. 265 Despite its common appearance in a wide variety of consumer products, questions about the safety 266 of triclosan remain, even if used externally. For example, in addition to its role in mitochondrial disruption, 267 triclosan also has the potential to disrupt endocrine function, affect immunity, disrupt calcium and zinc 268 homeostasis, and alter lipid metabolism (74). Triclosan also has a strong potential for bioaccumulation 269 (75), which is undesirable in a maintenance medicine. Given these caveats, the potential for triclosan to 270 be developed into a treatment for NDD seems small. 271 The potential for PS106, more commonly known as sertraline, is substantially greater. Sertraline is 272 thought to bind to the serotonin transporter (SERT) in the presynaptic neuron, preventing reabsorption 273 of serotonin and prolonging synaptic signaling. Sertraline is a well-known compound with carefully studied 274 pharmacological effects and is one of the most commonly prescribed psychiatric medications in the US. 275 Our data contribute to an ongoing discussion about the potential for sertraline as a treatment for one or 276 more NDDs. 277 Probably the earliest hint that sertraline may have some unexpected effect on mitochondria came in 278 a report attempting to identify 'hidden' drug targets, where the authors determined that sertraline had 279 several characteristics similar to the well-known mitochondrial toxin rotenone (76). Not long after, Kumar 280 and colleagues demonstrated that sertraline treatment could ameliorate damage caused by the mitotoxic 281 agent 3-nitropropionic acid (77). Recently, it was demonstrated that sertraline prevents the function of 282 the mitochondrial VDAC1, reducing cellular ATP, increasing the ADP/ATP ratio, and activating autophagy 283 through mTOR (78). 284 Sertraline increases survival and neurogenesis at pharmacologically relevant concentrations in several 285 murine models of Huntington's (79, 80) and physiological outcomes (e.g., grip strength, coordination, 286 locomotor activity, etc.) in rat models of Huntington's and Parkinson's diseases (81,82). Given the 287 frequent co-occurrence of depression with NDD, it is not surprising that sertraline is often prescribed to 288 these patients suffering from these disorders. Promisingly, some Parkinson's patients receiving sertraline 289 have shown improvement in their symptoms (83,84). It is clear, given these findings, that a more 290 systematic study of the potential for sertraline for the treatment of NDD is warranted. 291 The remaining six compounds are considerably less well characterized. PS83, formally known as Two of the compounds, PS30 and PS135, appear likely to be mitochondrial uncouplers. They reduced 300 mitochondrial membrane potential and decreased ATP content, but oxygen respiration continued 301 unabated. Like PS83, relatively little is known about PS30 or PS135. However, an analog of PS30, known 302 as SMTC1100, has been shown to be helpful in Duchenne muscular dystrophy (85). This fatal, progressive 303 disorder is characterized by wasting muscle loss due to disruption of the dystrophin protein, which leads 304 to mitochondrial dysfunction (86,87). This suggests that a larger portion of the scaffold may have a 305 positive effect on mitochondrial recycling in chronic degenerative disorders. 306 The final group of three compounds, PS34, PS127, and PS143 share a number of characteristics that 307 indicate that they may be acting the same way. They show substantial mitochondrial fragmentation, 308 autophagolysosomal acidification, they reduced ATP and oxygen consumption, and considerably reduced 309 average mitochondrial mass. The compounds did not, however, reduce mitochondrial membrane 310 potential, suggesting that they disrupt degradation of PINK-1/PINK1 in a different fashion. Interestingly, 311 they were also amongst the strongest activators of GST-4/Nrf and DAF-16/FOXO, which may indicate that 312 they are causing other damage to the cells. 313 The accumulation of mitochondrial damage, and concomitant degradation of function, is associated 314 with both aging and neurodegenerative disease. Mitophagy also appears to be inherently limited in 315 mature neurons (reviewed in (68)), which may explain why this tissue is more sensitive to mitochondrial 316 damage in the first place. Increasingly, it has been hypothesized and demonstrated that increasing 317 mitophagy in these cells may promote better cellular health and aging (reviewed in (68)). 318 Unfortunately, a relative dearth of compounds appropriate for this purpose is currently available, and 319 identification of new compounds requires a relatively complex screening process, like the whole-organism 320 phenotypic approach demonstrated herein. Although the eight compounds we identified and studied 321 have considerable promise (especially sertraline), substantial additional study is needed to further 322 understand their effects. 323 324

C. elegans strains and maintenance 326
Worms were synchronized by hypochlorite isolation of eggs from gravid adults, followed by hatching 327 of eggs in S Basal. 6,000 synchronized L1 larvae were transferred onto 10 cm standard nematode growth 328 medium (NGM) plates seeded with Escherichia coli strain OP50 as a food source (88). After transfer, 329 worms were grown at 20˚C for 50 hours prior to experiments, or for three days for the next eggs isolation.  Student's t-test analysis was performed to calculate the p values when comparing two groups in an 360 experimental setting. All statistical test results were indicated in graphs as follows: NS not significant, *p 361 < 0.05, **p < 0.01, and ***p < 0.001. For each of the experiments described below, at least three biological 362 replicates were performed.

ATP production measurement 394
A worm strain carrying firefly luciferase gene followed by GFP (PE255) was used for ATP production 395 measurement. Worms were treated with compounds as described above. ATP measurement was carried 396 out according to the published protocol (94). Essentially, at 18 h of incubation, worms were washed three 397 times to remove any remaining compounds. Luminescence buffer was then added, incubated for 3 398 minutes, and fluorescence (485/20 excitation and 528/20 emission) and luminescence were measured 399 with Cytation5 (BioTek Instruments). 400 401 Oxygen consumption rate measurement 402 3,000 N2 worms were sorted into each well of a 6-well plate. PS compounds, vehicle control DMSO, or 403 positive control rotenone, and E. coli OP50 (final OD600: 0.05) were then added into each well to a final 404 concentration of 50 µM. Two wells for each compound were used, totaling to 6,000 worms per condition. 405 Upon 8 h of incubation, worms were collected and transferred into a 15 mL conical. Worms were washed 406 three times to remove residual compounds. Oxygen consumption was measured by using a biological 407 oxygen monitor (YSI 5300) and a Clark-type oxygen electrode (YSI 5301) (Yellow Springs Instrument) at 408 20°C as previously described (49). Oxygen consumption was recorded continuously for ten minutes. with subsequent automated cell counting was used as cytotoxicity assay as previously described (96). 445 Cytation5 Cell Imaging Multi-Mode Reader with DAPI and Texas Red filter sets (BioTek Instruments) and 446 Gen5 3.10 software were used for imaging and cell counting pipeline. 447 For each cytotoxicity experiment, cells were seeded at a density of 10 3 cells/well in 96-well plates and 448 cultured for 24 h prior to the drug treatment. Cells were then treated with PS compounds or DMSO 449 (solvent control) at specified concentrations (see Fig. 7) for 72 h in 100 µL of complete media. All viability 450 rates were normalized to the corresponding solvent-control wells. The DMSO concentrations in the 451 incubation mixtures or solvent-control wells never exceeded 0.5% (v/v). 452 453 Acknowledgements: C. elegans strains used were obtained from the CGC. We thank Daniel Kirienko   PS143. Concentrations of PS compounds were indicated on the graphs. At least three biological replicates 758 with ~180 worms/replicate were analyzed. p values were determined from Student's t-test. NS not 759 significant, *p < 0.05, ** p < 0.01, *** p < 0.001. 760 Table S1. Chemical information of the eight PS compounds. 761