A copper chaperone-mimetic polytherapy for SOD1-

associated amyotrophic lateral sclerosis

McAlary L.

1,2

, Shephard, V.K.

1,2

*, Wright G.S.A.

, Yerbury J.J.

1,2§

1. Illawarra Health and Medical Research Institute, University of Wollongong, Wollongong, NSW, Australia

2. Molecular Horizons and School of Chemistry and Molecular Bioscience, Faculty of Science, Medicine and

Health, University of Wollongong, NSW, Australia

3. Department of Biochemistry & Systems Biology, Institute of Systems, Molecular and Integrative Biology,

University of Liverpool, Liverpool, United Kingdom.

Correspondence to lmcalary@uow.edu.au or jyerbury@uow.edu.au

*These authors contributed equally to the manuscript

Abstract

Amyotrophic lateral sclerosis (ALS)-associated mutations in Cu/Zn superoxide dismutase

(SOD1) reduce folding stability, resulting in misfolding, aggregation, and ultimately cellular

toxicity. A great deal of effort has focused on preventing the misfolding and aggregation of

SOD1 as a potential therapy for ALS, however, the results have been mixed. Here, we utilise a

small-molecule polytherapy of CuATSM and ebselen to mimic the metal delivery and disulfide

bond promoting activity of SOD1‟s cellular chaperone, the „copper chaperone for SOD1‟ (CCS).

We find that polytherapy using CuATSM and ebselen is highly effective at reducing inclusion

formation in a cell model of SOD1 aggregation, reduces mutant SOD1-associated cell death, and

promotes effective maturation of SOD1 beyond either compound alone. Our data suggest that a

polytherapy of CuATSM and ebselen may be an effective method of treating SOD1-associated

ALS.

Introduction

Over 160 mutations have been identified throughout the gene encoding Cu/Zn superoxide

dismutase (SOD1) that are known to cause the motor neuron disease amyotrophic lateral

sclerosis (ALS) [1,2]. These mutations result predominantly in amino acid substitutions found in

all SOD1 protein secondary structure elements, cofactor-binding sites and homodimer interface.

Each is thought to cause ALS by decreasing SOD1 folding stability, thereby creating a pool of

misfolded and aggregation-prone SOD1 [3,4]. While there is debate as to whether proteinaceous

aggregates or smaller, soluble non-native oligomers are toxic [5–7] the initial event considered to

spark cell death is SOD1 protein misfolding [3,4,8–11].

.CC-BY-NC-ND 4.0 International licenseavailable under a

(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made

The copyright holder for this preprintthis version posted February 23, 2021. ; https://doi.org/10.1101/2021.02.22.432389doi: bioRxiv preprint

SOD1 maturation comprises several sequential post-translational modifications (PTMs). Initially,

the spontaneous binding of zinc (Zn) to immature monomer provides some folding stability [12].

Zinc-bound SOD1 then associates with the copper chaperone for SOD1 (CCS) which facilitates

the input of copper (Cu) and subsequent oxidation of an intra-subunit disulfide bond between

Cys57 and Cys146. The stable monomer is then free to form enzymatically active homodimers

[13–15]. Metal-binding region (MBR) mutants affect copper or zinc coordination and activity

while wild-type-like (WTL) mutants retain high levels of enzymatic activity when mature

[16,17]. Immature SOD1, lacking PTMs, is prone to misfolding and is the central component of

intracellular aggregates found with ALS neuronal tissues [3,8,10,11,18–20], whereas mature

SOD1 is highly stable [21]. Maturation and misfolding are therefore antagonistic pathways that

dictate SOD1 toxicity. A cell has finite resources and a limited capacity to catalyze nascent

SOD1 maturation. The maturation pathway can be overwhelmed by high concentration nascent

SOD1 [22] or inhibited by mutations that prevent cofactor binding and disulfide PTMs

[3,4,9,23,24]. This results in increased misfolding pathway flux and the proteostasis pathways

becoming overwhelmed [25]. Decreasing SOD1 expression and thereby reducing traffic along

the misfolding pathway is the focus of knock-down strategies currently in clinical trials [26].

However, SOD1 has important metabolic functions and long-term ablation of its activity is

known to have detrimental effects. A second option is to increase maturation or cellular

proteostasis capacity. Heat shock protein molecular chaperone upregulation reduces SOD1

misfolding and clinical trials are again ongoing [27]. In addition, several small molecules have

been shown to act as direct pharmacological chaperones for SOD1 [28–31]. Pyrimidine

derivatives, 5-fluorouridine and telbivudine, reduce SOD1 in vitro aggregation and in vivo

toxicity respectively [28,29]. Treatment of SOD1-G93A mice with the 5-fluorouridine analogue,

5-fluorouracil, also delays symptom onset and increases survival[32].

The cognate chaperone of SOD1, CCS, is a uniquely placed chaperone that has evolved to

increase SOD1 maturation pathway throughput. It exerts molecular, copper, and oxidative

folding chaperone activity on nascent SOD1 mediated through a specific protein-protein

interaction [13,14,33,34]. While overexpression of hCCS reduces the accumulation of misfolded

SOD1 it can also recruit SOD1 mutants to the mitochondrial intermembrane space where they

accelerate vacuolization and toxicity [35–37]. Two small molecules have been shown to

recapitulate CCS activities. Copper(II)ATSM (CuATSM) promotes WTL mutant SOD1 Cu-

binding in several mouse models and increases lifespan [31,38–40] but it is ineffective against

MBR mutants [41]. The seleno-organic compound ebselen promotes the formation of the SOD1

intra-subunit disulfide bond in cultured cells [30] and increases SOD1 dimer affinity [42,43]. As

there are few SOD1 ALS-associated mutations that directly prevent disulfide formation (C146R

and truncation mutants) ebselen is likely to be effective for MBR mutants as well as WTL [2].

Indeed, recent evidence shows that ebselen and some of its derivatives can restore the viability of

cultured cells expressing G93A mutant SOD1, as well as delay disease onset in G93A mice

through dietary supplementation [44].

.CC-BY-NC-ND 4.0 International licenseavailable under a

(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made

The copyright holder for this preprintthis version posted February 23, 2021. ; https://doi.org/10.1101/2021.02.22.432389doi: bioRxiv preprint

Here, we report the effect of ebselen on intracellular mutant SOD1 inclusion formation in a

disease-relevant cell model. To achieve this, we developed a machine learning-based image

analysis pipeline for accurate measurement of protein inclusion formation in large microscopy

data sets. Application of this method to a subset of compounds showed ebselen was capable of

reducing inclusion formation for both WTL and MBR SOD1 mutants. We then utilized this

method to investigate CuATSM and ebselen co-therapy aiming to divert nascent SOD1 from the

misfolding pathway to the maturation pathway and thereby reduce mutant toxicity. We show

these compounds can act in a synergistic manner to reduce SOD1 aggregation through the

promotion of dimerization, disulfide formation and copper loading. All mutants analyzed display

positive outcomes for at least one marker of effective SOD1 maturation with resulting cell

viability increases for common or severely structurally destabilizing mutants. This work

highlights the unexplored possibilities of mutation-specific personalized therapy for SOD1-ALS

and the potential use of a CCS-mimetic polytherapy specifically targeting steps on the SOD1

PTM maturation pathway.

Results

Automated image analysis to identify cells containing inclusions

We, and others, have previously utilized genetically encoded fluorescent proteins as a tool to

investigate the inclusion formation of SOD1 in cultured cells finding that ALS-associated and de

novo mutations, as well as small molecules, can alter this process [3,29,41,45–48]. We have also

utilized several methods to detect fluorescent proteinaceous inclusions, including manual

counting [3], fluorescence intensity thresholding [29], and cell permeabilization to release

soluble GFP-tagged protein [29,41,49]. We now sought to enhance our detection of protein

100

inclusion formation in cells for use in larger mutational or drug screens. This was accomplished

101

by exploiting advances made in the area of microscopy image analysis with the application of

102

user-assisted machine learning to accurately classify cellular phenotypes [50].

103

104

To this end, we developed an image analysis pipeline using CellProfiler software [51] to identify

105

and segment transfected cells for measurement followed by classification in CellProfiler Analyst

106

software. Measurement parameters were chosen to append spatial data (shapes, texture,

107

granularity, radial intensity, and intensity) in order to generate cytoprofiles that were processed

108

by a random forest classifier (Figure 1A and Supp. Fig. 1 and 2A) [50]. Normalization of the

109

extracted cellular features demonstrated significant differences in texture, radial intensity, and

110

intensity measurements, but not shape or granularity (Supp. Fig. 2B). This indicated that these

111

measurements were appropriate for the profiling of cells. We found that our segmentation

112

parameters correctly identified transfected cells with high accuracy (97 ± 2%) and that the

113

chosen measurements effectively facilitated accurate classification for both cells with and

114

without inclusions (95 ± 0.5% and 98 ± 0.9% accuracy respectively) (Figure 1B).

115

.CC-BY-NC-ND 4.0 International licenseavailable under a

(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made

The copyright holder for this preprintthis version posted February 23, 2021. ; https://doi.org/10.1101/2021.02.22.432389doi: bioRxiv preprint

116

117

118

119

120

Figure 1. User-assisted machine learning to determine cells containing inclusions. (A) The image analysis

121

pipeline first performs illumination correction for both DAPI and GFP channels and then segments the transfected

122

cells for measurement. A user then identifies phenotypes in a small subset of the cell population to train the

123

classifier for identification of the entire cell population. (B) The classifier requires less than 100 to become accurate

124

at categorizing cells into inclusion containing (blue) and those that did not contain inclusions (orange). (C)

125

Correlations of the percentage of cells with inclusions in this work vs previously published examination of the same

126

SOD1-EGFP expression constructs in NSC-34s. (Left) vs manual counts from McAlary et al. 2016 [3], and (right)

127

vs saponin permeabilized cells from Farrawell et al. 2018 [25]. Error bars represent SD of the mean from at least 3

128

separate classification requests.

129

130

Ebselen, but not other compounds, reduces the formation of ALS-associated mutant SOD1

131

inclusions in cultured cells

132

133

Having established a rapid and accurate image-based method of classifying cells containing

134

inclusions, we next sought to examine the effect of ebselen and a small panel of other similar

135

.CC-BY-NC-ND 4.0 International licenseavailable under a

(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made

The copyright holder for this preprintthis version posted February 23, 2021. ; https://doi.org/10.1101/2021.02.22.432389doi: bioRxiv preprint

small molecules on SOD1 inclusion formation. The molecules, other than ebselen, were

136

omeprazole, clopidogrel, and lipoic acid. These compounds were chosen on the basis that they

137

contained sulfur moieties that may be redox-active in a similar manner to ebselen [30].

138

139

To this end, NSC-34 cells were transfected with SOD1 variants WT, A4V (WTL mutant), or

140

G85R (MBR mutant) and were treated with determined non-toxic concentrations of ebselen,

141

lipoic acid, omeprazole, or clopidogrel (Supp. Fig. 3) for 48 h prior to being fixed, imaged, and

142

analysed. We observed that WT SOD1 formed very few inclusions across treatments, with only

143

0.8 ± 0.3% of cells in the untreated group being classified as containing inclusions (Figure 2A),

144

in line with previous observations by us and others [3,29,41,45–48]. In contrast, both A4V and

145

G85R SOD1 readily formed inclusions in this system with 27.7 ± 5.8% and 24.1 ± 5.3% cells

146

containing inclusions respectively (Figure 2A). Treatment of A4V or G85R SOD1 expressing

147

cells with either lipoic acid, omeprazole, or clopidogrel at any of the tested concentrations had no

148

significant effect on inclusion formation (Figure 2A). We found that treatment with ebselen at

149

the highest concentration of 20 µM resulted in a significant reduction in both A4V and G85R

150

inclusion formation (Figure 2A), suggesting that ebselen was protective against both WTL and

151

MBR mutant inclusion formation in this model. The reduction in A4V mutant inclusion

152

formation was more substantial (3-fold decrease) when compared to G85R (2-fold decrease).

153

Previous examination of the capability of ebselen to facilitate SOD1 maturation in cells used a

154

ten-fold greater concentration of ebselen (200 µM) than we used here [30], where our results

155

suggest that ebselen may be more potent at facilitating SOD1 maturation than previously

156

suggested.

157

158

Ebselen was further tested on SOD1 variants including C6G, G37R, H46R, D90A, G93A,

159

E100G, G127X, and V148G in this model. Most of these ALS-associated mutants in this list are

160

WTL and induce inclusion formation to various degrees in NSC-34 cells [3,41]. H46R and G85R

161

are MBR mutants with minimal ability to bind copper [52,53], and G127X is a truncation mutant

162

that removes residues 127-153, including the disulfide forming Cys146 residue [54]. Treatment

163

of the transfected NSC-34 cells with increasing concentrations of ebselen showed that for cells

164

expressing C6G, G37R, D90A, G93A, E100G, and V148G there was a significant dose-

165

dependent response to ebselen (Figure 2B). The most effective dose observed in each case was

166

20 µM, although a significant difference between vehicle control and a concentration of 10 µM

167

was observed for G37R, D90A, G93A, and V148G (Figure 2B), indicating a greater effective

168

action of ebselen on these mutants. Cells expressing the truncated G127X SOD1 mutant, which

169

does not contain the disulfide-forming Cys146 residue, showed no significant effect of ebselen

170

on the percentage of cells containing inclusions at any concentration tested (Figure 2B). H46R

171

transfected cells showed no response to ebselen, however, inclusion formation is low for this

172

mutant. These data indicate that ebselen is likely acting through the previously proposed method

173

of promoting the formation of the SOD1 intrasubunit disulfide between residues Cys57 and

174

Cys146 [30].

175

.CC-BY-NC-ND 4.0 International licenseavailable under a

(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made

The copyright holder for this preprintthis version posted February 23, 2021. ; https://doi.org/10.1101/2021.02.22.432389doi: bioRxiv preprint

176

177

178

Figure 2. Ebselen reduces inclusion formation of SOD1 ALS-associated mutants in cultured cells. NSC-34

179

cells expressing (A) SOD1-EGFP variants WT, A4V, and G85R were treated with vehicle (black), clopidogrel (red),

180

ebselen (orange), lipoic acid (teal), or omeprazole (pink) for 48 h and the number of cells containing inclusions was

181

enumerated. (B) NSC-34 cells expressing SOD1-EGFP variants C6G, G37R, H46R, D90A, E100G, G93A, G127X,

182

and V148G were all treated with increasing concentrations of ebselen and inclusion formation was measured.

183

Ebselen treatment decreased inclusion formation for most variants except for G127X and H46R. Error bars represent

184

SD of the mean of at least 3 separate experiments. Statistical significance was determined using a student‟s t-test

185

against vehicle control (p < 0.001 = ***, p < 0.05 = *).

186

187

188

A Combination of Ebselen and CuATSM is Effective at Decreasing SOD1 Inclusion

189

Formation in Cells

190

191

Multi-drug polytherapies are used as a standard treatment against most types of cancer [55]. This

192

is a less common but growing strategy being adopted against neurodegenerative diseases [56,57].

193

.CC-BY-NC-ND 4.0 International licenseavailable under a

(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made

The copyright holder for this preprintthis version posted February 23, 2021. ; https://doi.org/10.1101/2021.02.22.432389doi: bioRxiv preprint

Previous research has shown that the copper carrying compound CuATSM is capable of

194

facilitating copper delivery to SOD1 in animals and cells [31,41]. Likewise, we have shown that

195

ebselen can promote SOD1 disulfide formation [30]. Therefore, we reason that CuATSM and

196

ebselen have the potential to act collaboratively to promote both copper binding and disulfide

197

formation respectively in SOD1 mutants [30,31,41].

198

199

Considering this, we set out to establish if a combination of both CuATSM and ebselen would

200

have a greater effect at reducing SOD1-associated fALS phenotypes in our cell model than either

201

drug alone. To this end, we performed a checkerboard treatment of NSC-34 cells expressing WT,

202

A4V, or G85R using different concentrations of CuATSM (0.5 - 0 µM) with ebselen (10 - 0

203

µM). Similar to our previous measurements with automated image analysis (Figures 3A), WT

204

formed very few inclusions in this assay at all CuATSM-ebselen combination treatments (Figure

205

3A). In comparison, both A4V and G85R formed more inclusions at lower drug concentrations

206

(Figure 3A). Heatmap visualisation of the checkerboard analysis shows that A4V responds to

207

both CuATSM and ebselen monotherapy, whereas G85R shows no response to CuATSM

208

monotherapy but responds to higher concentrations of ebselen monotherapy. Interestingly,

209

application of fractional inhibitory concentration index (FIC) measurement to the checkerboard

210

assays showed that ebselen and CuATSM acted synergistically for A4V but not G85R. Indeed,

211

ebselen at a concentration of 10 µM was capable of significantly reducing the number of cells

212

with inclusions to roughly 5% with as little as 0.03 µM CuATSM, which is over a 10-fold

213

decrease in the maximum CuATSM concentration. Likewise, a concentration of 0.25 µM

214

CuATSM was capable of reducing the necessary ebselen concentration for significant inclusion

215

formation reduction by roughly 8-fold, from 10 µM to 1.25 µM.

216

217

We next performed a live-cell time-lapse microscopy assay to count the relative numbers of

218

GFP-positive cells across time under the most potent treatment regimes used in the checkerboard

219

assay (CuATSM at 0.5 µM and ebselen at 10 µM). Data are reported as the number of cells that

220

are EGFP-positive relative to SOD-WT-EGFP transfected cells treated with CuATSM, or

221

ebselen, or a combination of both compounds. The expectation is that the compounds would

222

reduce the time-dependent decline in relative GFP-positive cell numbers. Similar to previous

223

reports [3,29,41,58], transfection of NSC-34 cells with mutant SOD1-GFP results in a decline in

224

relative EGFP positive cells over time (Figure 3B; left A4V, right G85R). In this assay, cells

225

transfected with A4V and treated with either ebselen or CuATSM alone, or with the combination

226

therapy saw a significant increase in the number of GFP-positive cells across time as determined

227

by measuring the area under the curve (Figure 3B; left and inset). Cells transfected with G85R

228

also showed a similar trend of decreasing numbers of relative GFP-positive cells across time,

229

however, CuATSM treatment alone had no effect on this decline whereas ebselen alone and

230

ebselen with CuATSM in combination did significantly reduce the decline in cell numbers

231

relative to WT (Figure 3B; right and inset).

232

.CC-BY-NC-ND 4.0 International licenseavailable under a

(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made

The copyright holder for this preprintthis version posted February 23, 2021. ; https://doi.org/10.1101/2021.02.22.432389doi: bioRxiv preprint

233

Figure 3. A combination treatment of CuATSM and ebselen is effective at rescuing SOD1-A4V folding. (A)

234

Heatmaps of CuATSM and ebselen checkerboard treatment of NSC-34 cells expressing either SOD1-WT (left),

235

SOD1-A4V (middle), and SOD1-G85R (right). Colours represent the mean percentage of transfected cells

236

containing inclusions from 3 separate experiments. (B) Transfected cell counts of NSC-34 cells expressing SOD1-

237

A4V (left) and SOD1-G85R (right) treated with vehicle DMSO (black), ebselen (10 µM; orange), CuATSM (0.5

238

µM; green), and ebselen/CuATSM combo (10 µM/0.5 µM; blue). Cell counts are relative to SOD1-WT transfected

239

cells treated with the same compounds. Inset in each panel is the area under the curve measurements for each drug

240

treatment. Error bars represent SEM of 3 separate experiments. Statistical significance was determined using a one-

241

way ANOVA with Dunnet‟s test against DMSO-treated cells (p < 0.01 = ***, p < 0.05 = **).

242

243

244

Polytherapy with Ebselen and CuATSM mimics CCS Activity and is Effective at Promoting

245

Mutant SOD1 Maturation

246

247

To investigate the mechanisms by which ebselen and CuATSM catalyse SOD1 maturation we

248

assessed the intra-subunit disulfide bond formation, dimerization, and activity of SOD1. In

249

comparison to the other redox compounds examined previously, only ebselen was able to

250

facilitate purified recombinant A4V disulfide formation (Figure 4A). Intra-subunit disulfide

251

formation is known to shift the SOD1 monomer-dimer equilibrium in favour of the dimer [59]

252

and we have previously shown ebselen binding to Cys111 can increase A4V homodimer affinity.

253

.CC-BY-NC-ND 4.0 International licenseavailable under a

(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made

The copyright holder for this preprintthis version posted February 23, 2021. ; https://doi.org/10.1101/2021.02.22.432389doi: bioRxiv preprint

However, this effect is negated by the presence of dithiothreitol or reduced glutathione at

254

physiological concentrations of 1 mM [30]. Ebselen, unlike oxidized glutathione, was able to

255

facilitate the formation of SOD1 homodimers even in the presence of 5 mM reduced glutathione

256

(Figure 4B). Under these conditions, ebselen cannot form stable conjugates at Cys111, therefore,

257

SOD1 homodimerization likely results from the catalyzed formation of the SOD1 intra-subunit

258

disulfide formation as we previously described in live cells [30].

259

260

Next, we compared the effectiveness of CuATSM and ebselen monotherapies and combination

261

therapy on their ability to promote the folding of intracellular SOD1 variants. Previous work

262

determined that non-reducing SDS-PAGE of SOD1 maintains the intramolecular disulfide bond

263

and that the disulfide bonded form of monomeric SOD1 migrates more rapidly during

264

electrophoresis [14]. SOD1 containing the intra-subunit disulfide was detected across treatment

265

groups for both WT and A4V, but not for G85R (Figure 4C). Semi-quantitative measurement of

266

the immunoblots showed a significant shift in the proportion of disulfide-containing SOD1

267

detected for both WT and A4V for each treatment in comparison to vehicle control (Figure 4D).

268

Out of the treatments, the combination treatment showed the greatest shift in the proportion of

269

SOD1 containing the intra-subunit disulfide for both WT and A4V (Figure 4C and D). The lack

270

of detection of disulfide bonded G85R may be a result of this mutant being highly destabilized

271

even in cells, and being highly susceptible to reduction even when free-thiols are chemically

272

blocked [60–62].

273

274

We next examined the effect of CuATSM and ebselen monotherapies and polytherapy on SOD1

275

dimerization and activity in cells by using Native-PAGE and in-gel zymography [63]. Here we

276

used SOD1 variants tagged with TdTomato due to EGFP-tagged SOD1 migrating too close to

277

endogenous cellular SOD1 for accurate densitometry measurements (Supp. Fig. 5). Assessment

278

of the SOD1-TdTomato signal in native-PAGE showed that dimer was only the most prominent

279

species for WT (60%) when cells were treated with vehicle control. Other variants were

280

predominantly monomeric, where G93A was the ALS-associated variant with the highest

281

proportion of dimer (Figure 4 E and F). Treatment with CuATSM, ebselen, or combination

282

therapy showed increases in the proportion of dimeric SOD1 for both WT and G93A, indicating

283

that compounds were promoting dimerization either through Cu input or disulfide formation

284

(Figure 4 E and F). Subsequent in-gel zymography showed that all variants (G85R data not

285

shown due to no activity observed under any treatment), exhibited a significant increase in the

286

activity of SOD1-TdTomato for CuATSM, and combination therapy treated cells compared to

287

vehicle-treated cells (Figure 4 E and G). Ebselen treatment did not appear to result in greater

288

levels of SOD1 activity, supporting a mechanism of stabilization that is related only to disulfide

289

formation. V148G was more enzymatically active when both CuATSM and ebselen were used in

290

combination, as compared to when CuATSM was administered alone, perhaps due to ebselen

291

further stabilizing the disulfide bond, which is an important PTM for SOD1 activity [64].

292

.CC-BY-NC-ND 4.0 International licenseavailable under a

(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made

The copyright holder for this preprintthis version posted February 23, 2021. ; https://doi.org/10.1101/2021.02.22.432389doi: bioRxiv preprint

293

Figure 4. A combination treatment of CuATSM and ebselen is effective at rescuing SOD1 ALS-associated

294

mutant folding. (A) Non-reducing SDS-PAGE AMS assay shows treatment with ebselen but not other redox

295

containing compounds resulted in SOD1-A4V disulfide formation (reduced = SH, oxidized intact =SS). (B) Size

296

exclusion chromatography shows ebselen promotes A4V homodimerization whereas oxidized glutathione (GSSG)

297

does not (D = dimers, M = monomers). (C) Differential SDS-PAGE migration of SOD1-EGFP from cell lysates

298

under reducing (+β-merc) and non-reducing (-β-merc) conditions shows that the proportion of disulfide bonded

299

SOD1 (SS) is increased with CuATSM (green; 0.5 µM), ebselen (orange; 20 µM), and a combination treatment

300

.CC-BY-NC-ND 4.0 International licenseavailable under a

(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made

The copyright holder for this preprintthis version posted February 23, 2021. ; https://doi.org/10.1101/2021.02.22.432389doi: bioRxiv preprint

(blue; 0.5 µM CuATSM/20 µM ebselen) compared to vehicle control (black) for both WT and A4V, but G85R

301

remains fully reduced. (D) Densitometry of disulfide formation immunoblots for WT and A4V showed that

302

CuATSM, ebselen, and the combination therapy were capable of promoting disulfide formation in living cells. (E)

303

Native-PAGE of SOD1-TdTomato lysates shows oligomers, dimers, and monomers of SOD1 variants (top) and in-

304

gel zymography of the same gels shows the relative activity of each species including dimer, monomer, and mouse

305

SOD1. (F) Quantification of the fluorescence signal from native-PAGE of the proportion of SOD1-TdTomato signal

306

present for the dimer, showing that CuATSM, ebselen, or combination therapy promoted the dimerization of both

307

WT and G93A SOD1, but not A4V, G85R, and V148G. (G) Quantification of the achromatic bands from in-gel

308

zymography showing that only treatment with CuATSM and the combination therapy increased the relative levels of

309

active SOD1 for WT, A4V, G93A, and V148G (G85R not shown due to lack of activity). Error bars represent SD of

310

the mean of at least 3 separate experiments. Significance was determined using one-way ANOVA with Dunnet‟s

311

multiple comparisons test with comparisons made against vehicle control data (p < 0.0001 = ****, p < 0.001 = ***,

312

p < 0.01 = **, p < 0.05 = *).

313

314

315

Discussion

316

317

Considering that ALS-associated mutations in SOD1 disrupt its maturation, some therapeutic

318

strategies have focused on catalysing proper SOD1 folding [30,41,65–67]. Initial methods to

319

stabilize SOD1 were focused on promoting the formation or maintenance of the SOD1

320

homodimer [65,68], which was a strategy based on the success of small molecules that

321

maintained familial amyloid polyneuropathy-associated mutants of the serum protein

322

transthyretin in its native tetrameric conformation [69]. A caveat to the approach of promoting

323

dimer stability for SOD1-associated ALS mutants is that SOD1 dimer formation primarily occurs

324

when monomers are already metal replete and disulfide oxidized: a species of SOD1, which is

325

still highly stable even when containing ALS-associated mutations [70]. Evidence points towards

326

immature metal depleted SOD1 being a precursor to the toxic forms of misfolded or aggregated

327

SOD1 [4,10,18,19]. Until recently, effective pharmacological chaperones targeting immature

328

SOD1 were elusive. Ebselen is considered to have a potential duel effect on SOD1 maturation,

329

facilitating disulfide formation, and increasing dimer affinity through binding at Cys111 [30],

330

although it should be noted that in vivo, ebselen binding at Cys111 is unlikely due to the

331

presence of reduced glutathione within cells. CuATSM is thought to facilitate the increased

332

delivery of Cu to SOD1, increasing the pool of Cu-bound SOD1 in several SOD1 ALS animal

333

models [31,38,40]. Here we considered that ebselen and CuATSM may be used in combination

334

to promote proper SOD1 folding at two different points, copper binding and disulfide formation

335

in effect acting as a CCS mimetic.

336

337

Ebselen was found to be more effective at attenuating the inclusion formation and toxicity of

338

A4V when compared to G85R. Previous investigations have shown that SOD1 with Cu bound is

339

less susceptible to disulfide reduction [59,71,72]. Since A4V is a WTL mutant that retains Cu-

340

binding capacity similar to that of WT, it would be expected that Cu-bound forms of A4V would

341

respond more favorably to ebselen-associated disulfide formation. Indeed, our data showed that

342

.CC-BY-NC-ND 4.0 International licenseavailable under a

(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made

The copyright holder for this preprintthis version posted February 23, 2021. ; https://doi.org/10.1101/2021.02.22.432389doi: bioRxiv preprint

combination treatment resulted in a greater level of enzymatically active V148G as compared to

343

either CuATSM or ebselen alone, suggesting a strong effect of polytherapy to stabilise even the

344

most destabilizing ALS-associated mutants. In contrast, the pool of G85R that binds Cu is

345

relatively low [52,73], meaning that the synergistic effect of Cu-binding and disulfide formation

346

that we observed for both V148G and A4V would not be expected to be seen for G85R. In line

347

with this, we saw no additive or synergistic effect of CuATSM in combination with ebselen

348

against G85R expression in our cell model.

349

350

Other pharmacological chaperones, such as those targeted against tryptophan-32 in SOD1

351

[28,29,74] or the molecular tweezer CLR01 [75], may result in more effective reduction of

352

inclusion formation and toxicity in the case of SOD1 MBR mutants such as G85R. Likewise,

353

derivatives of ebselen may also prove more effective at reducing inclusion formation and toxicity

354

in the case of SOD1 MBR mutants [42]. Both ebselen and CuATSM are currently in clinical

355

trials. CuATSM is currently in clinical trials against ALS [NCT04082832 and NCT02870634]

356

and ebselen is in clinical trials as a potential treatment for noise-induced hearing loss [76,77],

357

and is even being considered as a potential treatment for COVID-19 [78]. That these two

358

compounds have available safety profiles is promising for their application as a polytherapy for

359

ALS patients carrying SOD1 mutations. The effectiveness of CuATSM and ebselen for other

360

forms of ALS is currently not well understood. CuATSM has been found to inhibit the paraquat-

361

induced cytoplasmic localization of TAR DNA-binding protein 43 (TDP-43) into stress granules

362

[79] and to prevent TDP-43 phosphorylation and fragmentation in vivo [80]. Ebselen is yet to be

363

examined against TDP-43-associated forms of ALS.

364

365

Collectively, we have shown here that a biophysical understanding of the folding pathway of a

366

protein can be exploited to target it at key points to promote proper folding. However, the

367

strategy presented here is by no means the only one that may be pursued against SOD1-fALS.

368

Considering the concept of proteostasis (protein homeostasis) incorporates protein synthesis,

369

protein folding, protein trafficking, and protein degradation [81], there is potential to establish

370

combination therapies against SOD1-fALS at multiple points. These therapies may upregulate

371

the cellular chaperone networks [82], protein degradation pathways [83], or reduce SOD1

372

synthesis [84]. For example, the most clinically promising therapeutic for SOD1-associated ALS

373

are antisense oligonucleotides (ASOs) [84], which bind to and enhance the degradation of SOD1

374

mRNA, effectively reducing the concentration of SOD1 within a cell. Considering SOD1

375

deficiency may have negative effects in animals and humans [85], complete knockdown of

376

SOD1 via ASOs may potentially lead to complications. A different strategy may include partial

377

knockdown of SOD1 paired with supplemental treatment with pharmacological chaperones such

378

as CuATSM and ebselen to ensure the SOD1 that is synthesised folds properly. Indeed, cancer

379

researchers have made marked advances in establishing polytherapies as a highly effective

380

method by which to treat cancer [55]. Finally, future work against SOD1-associated ALS, and

381

even other neurodegenerative diseases, should incorporate our growing knowledge of the root

382

.CC-BY-NC-ND 4.0 International licenseavailable under a

(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made

The copyright holder for this preprintthis version posted February 23, 2021. ; https://doi.org/10.1101/2021.02.22.432389doi: bioRxiv preprint

mechanisms and downstream effects into therapy design. We hypothesize that further

383

improvement and expansion of the number of pharmacological chaperones that promote SOD1

384

folding will result in better outcomes in preclinical models and patients.

385

386

Conflict of Interest Statement

387

The authors declare no conflict of interest.

388

389

Author Contributions

390

Conceptualization: LM, GSAW, JJY; Methodology: LM, GSAW, JJY; Validation: LM, VS,

391

GSAW, JJY; Formal analysis: LM, VS, GSAW; Investigation: LM, VS, GSAW; Resources:

392

GSAW, JJY; Data curation: LM, VS, GSAW; Writing - original draft: LM, VS, GSAW;

393

Writing - review & editing: LM, VS, GSAW, JJY; Visualization: LM, VS, GSAW;

394

Supervision: LM, JJY; Project administration: LM, JJY; Funding acquisition: LM, GSAW,

395

JJY.

396

397

Acknowledgments

398

LM is the Bill Gole MND Fellow (Motor Neurone Disease Research Australia). GSAW is

399

funded by the Motor Neurone Disease Association, UK, Wright/Oct18/969-799. The authors

400

acknowledge the facilities and technical staff of the Illawarra Health and Medical Research

401

Institute. The authors acknowledge the facilities, the technical and scientific assistance of the

402

Fluorescence Analysis Facility in Molecular Horizons, Faculty of Science, Medicine and Health,

403

University of Wollongong.

404

405

Supporting Information

406

407

Methods and Materials

408

409

Plasmids for Mammalian Protein Expression

410

Vectors for the expression of C-terminally EGFP-tagged SOD1 variants WT, A4V, C6G, G37R,

411

H46R, D90A, G93A, E100G, G127X, and V148G on a pEGFP-N1 backbone have been

412

previously described [3,41,47]. Plasmids for the expression of human SOD1-A4V from

413

Escherichia coli have been described previously [30]. Plasmids were heat transformed into

414

subcloning efficiency chemically competent Escherichia coli DH5α cells (Thermofisher, USA)

415

and purified using miniprep kits (Thermofisher, USA) and maxiprep kits (Qiagen, Germany) as

416

per the manufacturer‟s instructions.

417

418

Mammalian Tissue Culture and Transfection

419

NSC-34 [86] cells were cultured in Dulbecco's modified Eagle‟s medium-F12 (DMEM-F12)

420

(Invitrogen, USA), supplemented with 10% (v/v) heat inactivated fetal bovine serum (FBS)

421

.CC-BY-NC-ND 4.0 International licenseavailable under a

(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made

The copyright holder for this preprintthis version posted February 23, 2021. ; https://doi.org/10.1101/2021.02.22.432389doi: bioRxiv preprint

(Bovogen, Australia). In order to passage and plate NSC-34 cells, they were washed once with

422

pre-warmed DMEM-F12 and treated with 0.25% trypsin, 0.02% EDTA dissociation reagent

423

(Invitrogen, USA) to lift off the adherent cells. The cells were pelleted via centrifugation (300 ×

424

g for 5 min) and resuspended in pre-warmed culture media. Following washing, plates were

425

seeded at a confluency of 40% and cultured at 37 °C in a humidified incubator with 5%

426

atmospheric CO

for 24 h prior to transfection (~70-80% confluent). Cells were transfected with

427

plasmid DNA (0.5 µg per well of a 24-well plate, 2.5 µg per well of a 6-well plate) 24 h post-

428

plating using TransIT-X2 reagent (Mirus Bioscience, USA) according to the manufacturer's

429

instructions.

430

431

Crystal Violet Assay for Cell Density

432

The half maximal inhibitory concentration (IC

) of various drugs on untransfected NSC-34 cells

433

was determined via a crystal violet assay (0.5 g/L crystal violet, 1 % methanol (v/v), 1× PBS) as

434

previously described [87]. NSC-34 cells were treated for 48 h with ebslen and CuATSM at

435

concentrations ranging from 0-500 μM with a final concentration of 1% (v/v) (dimethyl

436

sulfoxide (DMSO) (Sigma Aldrich, USA)). The drug treated NSC34 cells were then fixed via the

437

addition of pre-warmed 4% paraformaldehyde (PFA) in 1× PBS and incubated for 30 min at

438

room temperature prior to the addition of crystal violet solution. Crystal violet stained cells were

439

imaged using a Gel Doc XR+ gel imager (BioRad, USA). Glacial acetic acid (100 µL 33% (v/v))

440

was used to release the crystal violet stain back into solution for quantification of absorbance at

441

590 nm on a POLARstar plate reader (BMG Labtech, Germany). The resulting data were plotted

442

via Prism (GraphPad PRISM, Version 5.00 or Version 8.00) using a log (inhibitor) vs.

443

normalized response variable slope fit.

444

445

Preparation of Plates for Fluorescence Microscopy

446

NSC-34 cells were plated into 6-well culture plates at a confluency of 40% and incubated

447

overnight at 37 °C in a humidified incubator with 5% atmospheric CO

. To overcome the effect

448

of transfection efficiency differences within assays, cells were transfected as described above

449

and incubated for 5 h at 37 °C in a humidified incubator with 5% atmospheric CO

. Following

450

incubation, cells were aspirated or lifted off with trypsin/EDTA dissociation reagent and replated

451

into 96-well culture plates at a confluency of 30% in the presence or absence of various

452

compounds in a final volume of 100 µL and incubated for 48 h. Following incubation, cells in

453

culture medium were fixed via addition of 100 µL pre-warmed 4% paraformaldehyde (PFA) in

454

1× PBS with a 30 min incubation. Following fixation, cells were permeabilized using 0.1%

455

Triton X-100 in 1× PBS for 5 min, which was followed by a 5 min incubation in 1× PBS with a

456

1:5000 dilution of Hoescht 33342 (Life Technologies, USA). Finally, cells were washed twice in

457

1× PBS before being immediately imaged or stored in the dark at 4 °C. Stored cells were imaged

458

no later than 3 days after fixation.

459

460

Fluorescence Microscopy

461

.CC-BY-NC-ND 4.0 International licenseavailable under a

(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made

The copyright holder for this preprintthis version posted February 23, 2021. ; https://doi.org/10.1101/2021.02.22.432389doi: bioRxiv preprint

A LionHeartFX automated microscope (Biotek Agilent, USA) running Cytation software (Biotek

462

Agilent, versions 3.04 and 3.08) was used for plate-based image acquisition. NSC-34 cells

463

expressing SOD1 mutant EGFP-tagged constructs were excited via illumination with a 465 nm

464

LED and the emission was filtered through a 469/525 nm bandpass filter cube. All images were

465

taken using an UPLFLN PH 10× 0.3NA objective (Olympus, Japan). Each well was imaged in a

466

4×4 or 5×5 tile scan. No image overlapping was used in order to avoid duplicating cell counts in

467

later analysis stages. For imaging SOD1-WT-EGFP transfected cells, the LED power was set to

468

1 and integration time was 20 ms to prevent the acquisition of saturated fluorescent signal from

469

high expressing cells. Mutant SOD1 transfected cells were imaged using an LED power of 2 and

470

an integration time of 25 ms, set to account for variation in fluorescent intensity and lower

471

expression. A camera gain of 10 was consistent for both WT and mutant samples. Meta data for

472

individual wells was set to the following:

473

474

(WELL)_(IMAGE IN WELL)_(CHANNEL)_(IMAGE NUMBER)

475

476

to generate a unique image identity such as B4_2_GFP_2 allowing for easy metadata assignment

477

and data curation in image processing and analysis pipelines.

478

479

Image Analysis

480

All images generated via automated microscopy underwent pre-processing quality control to

481

omit out of focus images and to correct illumination variation in the data sets. Out of focus

482

images were manually assessed by users and excluded from the data set while illumination

483

variation within images was corrected using CellProfiler software modules Correct Illumination

484

Calculate and Correct Illumination Apply. Briefly, a 500 pixel gaussian smoothing filter was

485

used to generate illumination functions which display the illumination variation within a set of

486

images from one imaging session (multiple plates). The illumination function is then subtracted

487

from the data set to correct for variation across the image. After quality control processing,

488

images were processed in CellProfiler to segment cells within the range of 17-50 pixel units and

489

measure intensity, granularity, size/shape, intensity distribution and texture. Accuracy of

490

segmentation and thumbnail generation was assessed during the training of the machine learning

491

algorithm. To determine accuracy, the user requested 100 cells of a particular “bin” and noted

492

how many cells were incorrectly classified. This was repeated 3 times per number of cells the

493

machine was trained on and the average accuracy noted. Once a reasonable accuracy was

494

achieved (~97 %) all remaining cells were automatically scored via CellProfiler Analyst.

495

Pipelines are available from the authors upon reasonable request.

496

497

Purified SOD1 free-thiol and homodimerization assays

498

Human SOD1-A4V was expressed in Escherichia coli BL21 (DE3) and purified as described

499

previously [42]. The SOD1 intra-subunit disulfide bond was reduced with 40 mM dithiothreitol

500

(DTT) overnight at 4

C followed by desalting into N

purged 20 mM Tris-HCl, 150 mM NaCl

501

.CC-BY-NC-ND 4.0 International licenseavailable under a

(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made

The copyright holder for this preprintthis version posted February 23, 2021. ; https://doi.org/10.1101/2021.02.22.432389doi: bioRxiv preprint

with 5 mM reduced glutathione for SEC homodimerization assays and without reduced

502

glutathione for free-thiol assays. Compounds, including ebselen, were dissolved in DMSO and

503

added to 20 µM SOD1-A4V at 20 or 100 µM concentration for homodimerization and free-thiol

504

assays respectively. The reaction was incubated at 20

C for 1 h before addition of 400 µM 4-

505

acetamido-4'-maleimidylstilbene-2,2'-disulfonic acid and incubation at 37

C for 90 minutes.

506

Samples were then heated to 97

C in non-reducing SDS sample buffer then separated by SDS-

507

PAGE using a 15% polyacrylamide gel. Homodimerization assay samples were incubated at 20

508

C for 24 h then 10 µl was loaded on an Agilent BioSEC Advance 300 Å, 4.6 × 300 mm size

509

exclusion chromatography column along with controls for disulfide reduced and disulfide intact

510

SOD1-A4V without ligands.

511

512

Immunoblotting of cell lysates

513

NSC-34 cells were cultured in 6-well plates, transfected, and treated with compounds similar to

514

above methods. Importantly, following 6 h after addition of transfection complexes to cells, cells

515

transfected with specific constructs (SOD1-WT-EGFP, SOD1-A4V-EGFP, or SOD1-G85R-

516

EGFP) were lifted and mixed together and replated to ensure equal transfection per construct for

517

each drug treatment. Following 48 h incubation in drugs (vehicle DMSO, 0.5 µM CuATSM, 20

518

µM ebselen, or a 0.5 µM CuATSM / 20 µM ebselen combination), cells were washed with

519

prewarmed (37

C) serum-free DMEM/F12 once, and incubated for 5 min in prewarmed 0.25%

520

trypsin, 0.02% EDTA dissociation reagent. Once lifted, cells were harvested into microfuge

521

tubes and pelleted at 300 × g for 5 min. Pellets were gently resuspended in prewarmed 1× PBS

522

and spun again at 300 × g for 5 min. Supernatant was removed and cell pellets were resuspended

523

and lysed in 100 µL ice-cold 1× Tris-buffered saline (pH 7.4) with 1% TX-100 1 mg/mL N-

524

ethylmaleimide (NEM) supplemented with 1× Halt™ protease inhibitor (ThermoFisher, USA)

525

to release soluble SOD1 from cells. Resuspensions were centrifuged at 20,000 × g for 20 min at

526

C to pellet nuclei and insoluble material. Supernatants were carefully transferred to new

527

microfuge tubes and these samples were flash frozen with liquid N

and stored at -80

C prior to

528

use.

529

530

Cell lysates were defrosted on ice and mixed 1:3 with either 4× non-reducing thiol-blocking

531

SDS-PAGE sample buffer (200 mM Tris-HCl pH 6.8, 8% SDS (w/v), 40% glycerol (v/v), 50

532

mM EDTA, 0.08% (w/v) bromophenol blue, 40 mM NEM) or 4× reducing SDS-PAGE sample

533

buffer (200 mM Tris-HCl pH 6.8, 8% SDS (w/v), 40% glycerol (v/v), 50 mM EDTA, 0.08%

534

bromophenol blue (w/v), 4% β-mercaptoethanol (v/v)). Samples for SDS-PAGE were then

535

heated to 95

C for 5 min prior to being loaded onto 4–20% Criterion™ TGX Stain-Free™ gels

536

(BioRad, Australia). Gels were electrophoresed for 5 min at 100 V and then 1 h at 150 V.

537

Following electrophoresis, total protein on the gel was quantified using a Criterion Stain Free™

538

Imager (BioRad, Australia) prior to transferring for immunoblotting.

539

540

.CC-BY-NC-ND 4.0 International licenseavailable under a

(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made

The copyright holder for this preprintthis version posted February 23, 2021. ; https://doi.org/10.1101/2021.02.22.432389doi: bioRxiv preprint