The anti-leprosy drug clofazimine reduces polyQ toxicity through activation of PPARγ

Summary Background PolyQ diseases are autosomal dominant neurodegenerative disorders caused by the expansion of CAG repeats. While of slow progression, these diseases are ultimately fatal and lack effective therapies. Methods A high-throughput chemical screen was conducted to identify drugs that lower the toxicity of a protein containing the first exon of Huntington's disease (HD) protein huntingtin (HTT) harbouring 94 glutamines (Htt-Q94). Candidate drugs were tested in a wide range of in vitro and in vivo models of polyQ toxicity. Findings The chemical screen identified the anti-leprosy drug clofazimine as a hit, which was subsequently validated in several in vitro models. Computational analyses of transcriptional signatures revealed that the effect of clofazimine was due to the stimulation of mitochondrial biogenesis by peroxisome proliferator-activated receptor gamma (PPARγ). In agreement with this, clofazimine rescued mitochondrial dysfunction triggered by Htt-Q94 expression. Importantly, clofazimine also limited polyQ toxicity in developing zebrafish and neuron-specific worm models of polyQ disease. Interpretation Our results support the potential of repurposing the antimicrobial drug clofazimine for the treatment of polyQ diseases. Funding A full list of funding sources can be found in the acknowledgments section.


INTRODUCTION
Polyglutamine (polyQ) diseases include 9 inherited hereditary neurodegenerative syndromes that are caused by the expansion of Q-coding repeats within the exons of several seemingly unrelated genes [1]. One of these pathologies is Huntington´s disease (HD), being one of the most frequent neurodegenerative diseases with an incidence of 3-5 cases per 100.000 worldwide [2]. In HD, the disease is linked to the expansion of a CAG repeat within the first exon of huntingtin (HTT), which becomes pathogenic above 35 repeats with the severity of the disease correlating with repeat length [3,4]. While HTT dysfunction has been proposed to contribute to HD [5,6], an alternative hypothesis is that the pathology is caused by gain-of-function toxicity of the polyQ-bearing mutant HTT (mHTT). Accordingly, early studies showed that transgenic mice expressing a fragment of the exon 1 from mHTT including the expanded polyQ track suffered from motor dysfunction and premature death [7,8]. Importantly, seminal work revealed that ectopic expression of polyQ expansions inserted in HPRT, a gene not mutated in patients, also led to neurodegeneration and premature death, highlighting the causal role of polyQ toxicity [9].
In what regards to the mechanisms of polyQ toxicity, this remains to be fully understood. An important feature of these expansions is their propensity to form insoluble aggregates that form intraneuronal inclusions, which were found in mouse models and also patients from several polyQ diseases including HD [10][11][12]. However, whether these inclusions are the real cause of the pathology has been the subject of intense debate, and it is clear that mHTT can also be toxic independently of the formation of large aggregates (reviewed in [1]). Regardless of whether it forms inclusions, mHTT has been shown to drive multiple cellular alterations in aspects such as mRNA transcription [13][14][15], protein degradation and post-translational modifications [16], synaptic function and plasticity [17][18][19][20][21] and mitochondrial activity [22][23][24][25][26].
Unfortunately, these mechanistic discoveries have not yet led to clinical improvements in the treatment of HD. The only approved treatments for HD, tetrabenazine and deutetrabenazine, are directed to alleviate the involuntary movements (chorea), but do not cure the disease [27,28]. In this context, it becomes urgent to try to find novel therapies for polyQ diseases, an area of intense research. Efforts are spread among strategies trying to prevent mHTT aggregates or promote their clearance, as well as to targeting their downstream pathological effects (reviewed in [29]). Noteworthy, several of the unbiased chemical screens have been focused on the identification of compounds that lower polyQ aggregates in biochemical assays, which often lead to compounds that show toxicity per se when evaluated in in vivo models [30,31]. Here, we present our results from a High-Throughput Imaging based drug-repurposing screening oriented to find compounds that reduce the toxicity of polyQ expansions.

A chemical screen for modulators of polyQ-toxicity
To conduct a chemical screen, we first generated an inducible system enabling the expression of an EGFP fusion protein containing the first exon of human HTT with an expanded polyQ tract of 94 glutamines (Htt-Q94 hereafter). The cDNA was cloned in a Tet-On gene expression system enabling the expression of Htt-Q94 upon the addition of doxycycline (dox). As polyQ expression is toxic for any cell type, the system was stably integrated in human osteosarcoma U2OS cells that are widely used in large chemical screens, and a clone with stringent regulated expression selected for further experiments (U2OS Q94 ). Before conducting the screen, we verified the Dox-inducible expression of Htt-Q94 as seen by a widespread accumulation of EGFP-expressing cells (Fig. 1A). Moreover, and as previously reported in similar setups, a 1-week treatment with dox led to the appearance of cells with Q94 aggregates (Fig. 1A, inset). At this time, Htt-Q94 expression led to a significant reduction in cell numbers, as quantified by detecting nuclei by High-Throughput Microscopy (Fig. 1B), confirming the toxicity driven by polyQ expression in this cell system.
The chemical library used combined 1,200 FDA-approved compounds and 94 additional drugs targeting components of the epigenetic machinery (Table S1).
We added epigenetic drugs given that several of them have been found to be of potential for the treatment of HD and other neurodegenerative diseases [32]. To conduct the screen, U2OS Q94 cells were seeded on 384 well plates at 100 cells/well, and treated with dox (50 ng/ml) and the library compounds (1µM) for 8 days. At this point, cells were fixed and DNA was stained with Hoechst enabling the quantification of nuclei (Fig. 1D). 35 compounds leading to an increase in nuclei numbers bigger that 3 SD from those found in the control wells (only treated with dox), were taken for a dose-response validation screen conducted at 0.5, 1, 5 and 10 µM. From this secondary screen, 4 compounds showed a significant rescue of toxicity in at least 2 of the doses tested: promethazine (PRM), amodiaquine (AMD), clofazimine (CFZ) and troglitazone (TZD) (Fig.   S1A).

Clofazimine and troglitazone rescue polyQ-toxicity in vitro
Next, and to evaluate whether the compounds were able to present a sustained effect in reducing the toxicity associated to Htt-Q94 expression, we conducted clonogenic survival assays. These experiments confirmed that all 4 compounds increased the number of colonies in dox-treated U2OS Q94 cells ( Fig.   2A and Fig. S2). Before entering into mechanistic analyses, we first wanted to discard hits that were acting by preventing the dox-dependent expression of Htt-Q94, an issue that we have previously faced when conducting similar screens using Tet-On systems [33]. To do so, we measured their effects on dox-induced Htt-Q94 levels both by immunofluorescence (IF) and western blotting (WB). While PRM and AMD significantly limited Htt-Q94 expression, CFZ and TZD rescued polyQ toxicity despite not affecting Htt-Q94 expression or the presence of its aggregates (Fig. 2B, C). We thus focused on these two compounds for subsequent analyses.
To further validate these results in vitro in an orthogonal model, we performed growth competition assays in the human leukemic KBM7 cell line. To do so, we co-cultured KBM7 cells expressing either mCherry or EGFP-Htt-Q94 (KBM7 Q94 ) for 10 days. In the absence of drugs, the percentage of KBM7 Q94 cells progressively declined, confirming that Htt-Q94 expression also impairs cellular fitness in this model. In contrast, treatment with CFZ or TZD rescued the relative decline of KBM7 Q94 cells, confirming the in vitro effects of both these drugs in rescuing polyQ toxicity (Fig. 2D,E). Interestingly, one of these two compounds, TZD, is a well-established agonist of the peroxisome proliferator activated receptor gamma (PPARg), an approach that has been previously studied as a potential therapy for various neurodegenerative diseases including HD, confirming the usefulness of our screen to identify potential therapies [34][35][36][37]26].
In contrast, CFZ, an antibiotic originally developed as a treatment for leprosy active against a wide range of mycobacteria [38], has not been previously investigated in the context neurodegeneration. We thus selected CFZ for further analyses.

Clofazimine rescues polyQ-induced mitochondrial damage
To understand how CFZ treatment was rescuing polyQ toxicity we conducted transcriptomic analyses by RNA sequencing (RNAseq) in dox-induced U2OS Q94 cells treated or not with CFZ for 8 days. Interestingly, these analyses revealed a general impact of CFZ in boosting the expression of multiple factors related to mitochondria such as voltage dependent ion channels, translocases, subunits of the ATP synthase and components of mitochondrial translation (Fig. 3A).
Consistently, Gene Set Enrichment Analyses revealed that CFZ treatment led to a significant enrichment of multiple pathways related to mitochondrial function To evaluate mitochondrial activity, we used mitotracker, a red dye that stains mitochondria in a membrane-potential-dependent manner [39]. In agreement with the mitochondrial dysfunction that has been repeatedly documented in cells from HD patients (reviewed in [40]), dox treatment led to a notable reduction of the mitotracker signal in U2OS Q94 cells, which was rescued by CFZ (Fig. 3C,D).
Similarly, transmission electron microscopy analyses revealed that Htt-Q94 expression had a profound impact on the mitochondria of U2OS Q94 cells, characterized by swelling and substantial abnormalities in external membranes and cristae, all of which were rescued by CFZ (Fig. 3E).

Clofazimine is a PPARg agonist
As mentioned, CFZ has been used as an antimycobacterial since the 1950s [41]. In addition, recent screens also identified that CFZ prevented infection by a wide range of viruses, including SARS-CoV-2 [42]. Surprisingly, despite its interesting medical properties, its target and mechanism of action remain unknown. To address this, we used the transcriptional signature of CFZ-treated U2OS cells to interrogate the Connectivity Map (CMap), a database from the Broad Institute at MIT that stores the transcriptional signatures of more than 5,000 drugs [43], aiming to identify drugs with a similar transcriptional impact.
Interestingly, these analyses revealed an enrichment of PPARg agonists among the compounds presenting a transcriptional signature that resembled that of CFZ ( Fig. 4A). In fact, PPARg itself was transcriptionally induced by CFZ in our transcriptomic analyses (Fig. 3A).

Consistent with bioinformatic analyses, molecular docking revealed that CFZ
is able to bind to the same pocket in PPARg as other agonists such as TZ, and with a similar binding affinity (Fig. 4B,C). Furthermore, cellular thermal shift assays (CETSA) [44] indicated that TZ and CFZ were able to have a similar impact in stabilizing PPARg at increasing temperatures, supporting a direct interaction (Fig 4D,E). Finally, concomitant treatment with the PPARg antagonist GW9662 reverted the effects of Cf and Tz in restoring the expression of PPARG or the mitochondrial factors TFAM or Citrate Synthase (CS) in dox-treated cells U2OS Q94 cells (Fig. 4F). Together, these results demonstrate that CFZ can bind to PPARg and stimulate its activity.

Clofazimine rescues polyQ toxicity in neurons and zebrafish
To further document the effect of CFZ in a neuronal model, we used SH-SY5Y neuroblastoma cells, which can be differentiated into a neuronal-like phenotype with retinoic acid (RA) [45]. Parental cells were infected with pLVX-UbC-rtTA-Htt-Q94-CFP lentiviruses, enabling dox-dependent expression of Htt-Q94-CFP (SH-SY5Y Q94 ). After 5 days of differentiation with 10 µm RA, SH-SY5Y Q94 cells were treated with dox for 3 additional days. As in all previous models, Htt-Q94 expression had a profound impact on SH-SY5Y Q94 cells, exemplified by decreased cell numbers and a reduction in the mitotracker signal; all of which were rescued by treatment with CFZ (Fig. 5A,B and Fig. S3). Of note, CFZ had a significant effect in increasing cell numbers and mitochondrial activity also in SH-SY5Y Q94 cells that were not previously exposed to Dox, highlighting its potentially beneficial effects in other pathologies associated to neuronal dysfunction [46].
Finally, we tested the impact of CFZ in alleviating polyQ toxicity in zebrafish.
To this end, we used a previously developed plasmid enabling the expression of Htt-Q94-CFP [47]. On day 0, fertilized zebrafish eggs were injected with the plasmid and exposed to CFZ at 12,5 µM (Fig. 5C). Consistent with previous studies [48,49], transgenic Htt-Q94 expression led to substantial embryonic lethality in developing fish. Importantly, CFZ was able to significantly increase embryonic survival in Htt-Q94-transgenic embryos, confirming its effects in alleviating polyQ toxicity in vivo (Fig. 5D,E).

DISCUSSION
As mentioned in the introduction, and despite the substantial advances made in understanding the molecular basis of polyQ-diseases, this has not yet led to effective treatments. Among others, substantial efforts are being dedicated to find therapeutic strategies to either reduce the expression of polyQ-containing proteins (e.g. antisense oligonucleotides (ASOs) or RNA interference), or that aim to either prevent the formation of polyQ aggregates or promote their clearance (reviewed in [29]). Our approach was rather to identify molecules capable of reducing the toxicity of polyQ-bearing proteins. In this regard, a similar approach was conducted by the Taylor laboratory where they searched for molecules that reduced apoptosis triggered by the expression of a truncated androgen receptor containing a 112-glutamine repeat in HEK 293T cells [50]. In our screen model, U2OS, Htt-Q94 did not trigger apoptosis but rather cell cycle arrest. Interestingly, we observed that the severity of this phenotype more acute when cells were sed at low densities, perhaps reflecting that the formation of polyQ aggregates is also enhanced at sub-confluence [51].
The usefulness of our approach was supported by the fact that we were able to identify compounds previously known to modulate the severity of polyQ pathology in preclinical models such as TZD [34,35,37,26]. Unfortunately, and despite being originally approved for the treatment of diabetes, TZD was later removed from the marked due to hepatic toxicity [52]. Nevertheless, cumulative data supporting that activation of the PPARg/PDC1a axis is a fruitful therapeutic approach for the treatment of neurodegenerative diseases [46], emphasizes the need of discovering new PPARg agonists that hopefully overcome the initial toxicities. In this regard, our work indicates that CFZ is a PPARg agonist, with a similar binding affinity as TZD, but which is seemingly safe as it is in clinical use for the treatment of infectious diseases. Of note, one limitation of CFZ is its poor efficacy in crossing the blood-brain barrier (BBB), which has limited its efficacy for the treatment of infections in the central nervous system. In this regard, there are already efforts dedicated to circumvent this problem such as nanoparticlebased formulations of CFZ [53]. In any case, our work suggests that CFZ could be a useful alternative to TZD for the treatment of pathologies outside the CNS.
In summary, our study further indicates the potential of PPARg stimulation to reduce the severity of pathologies of polyQ-diseases, and that these effects are primarily related to restoring mitochondrial function. In addition, our work adds a new example of the possibilities offered by drug repurposing to identify medically approved drugs that could be investigated in the context of neurodegenerative diseases. While acknowledging the current pharmacological limitations of the drug, we believe that exploring the efficacy of CFZ or its derivatives in polyQ diseases deserves further preclinical work.

High-throughput Screening (HTS)
Plate and liquid handling were performed using Echo550 (Labcyte, USA), The chemical collection used in the primary screening contained 1,122 medically approved compounds from the Prestwick library and 94 epigenetic-drugs available at CBCS collections (the list of compounds is available at Table S1).

Flow cytometry
For the analysis of the competition assay of KBM7 Q94 and KBM7-mCherry in, 4x 10 4 KBM7 Q94 and KBM7-mCherry cells were mixed at a 1:1 ratio and were seeded in T175 flasks. Cells were treated with CFZ and TZD at indicated concentrations. After 5 or 15 days, cells were analysed for green or red fluorescence by flow cytometry (Bio-Rad S3e cell sorter). Data were processed with the Flow Jo 10 software to calculate the percentage of each cell population.

Viability assays
For clonogenic survival assays, 500 cells per well were plated in 6-well tissue culture plates in the corresponding culture medium. Cells were treated with the indicated concentrations of drugs and maintained with the compounds for 12 days, changing the medium every 3-4 days, and then fixed and stained with 0.4% methylene blue in methanol for 30 min.

RNA-seq and data analysis
Total RNA was extracted from cell pellets using a Purelink RNA Mini Kit an Illumina platform. STAR [57] was used for sequence alignment based on the GRCh38 DNA primary assembly reference build [58], and quantification was done using featureCounts [59] with reference build GRCh38.101 [58].
Differential expression (DE) analyses between the groups were performed using DESeq2 [60]. Generalized linear model (GLM) was fitted to the expression data and shrunken log2fold-change (LFC) using adaptive Student's t prior shrinkage estimator [60,61]. Multiple testing correction was done using Benjamini-Hochberg (BH) method [62]. GSEA analysis was performed on the gene-level statistics from the DE analyses results against the molecular signatures from the Molecular Signatures Database (MSigDB) v7.5.1 [63,64] and the Reactome database [65]. Specifically, signatures from the ontology gene set C5 of MSigDB, containing Gene Ontology (GO)-derived gene sets [66,67], as well as the complete gene sets from the Reactome database, retrieved from fgsea 1.20.0 [68], were used. GSEA analysis was carried out using clusterProfiler 4.2.2 [69] to identify enriched terms. The transcriptional signatures, identified to be the sets of top up/down-regulated 150 genes (BH p-adjusted values < 0.05, ranked by LFC) from the DE analysis outcomes, were separately used as inputs to the Connectivity Map (CMap) Query clue.io tool (https://clue.io/) [70] to identify drugs with signatures similar to that clofazimine.

Molecular docking
The 3D crystal structure of PPAR-γ was downloaded from the Protein Data Bank (http://www.rcsb.com; #3ET0). Autodock vina was used to removed water molecules and add missing side or back chains and residues. Chemical

Zebrafish study
To test the impact of CFZ in alleviating polyQ toxicity in zebrafish. On day0, the injection mix (25ng/ul transposase, 50ng/ul vector, 0.3% phenol red) was injected into 150 eggs. The compounds were added in to the E3 medium at 12,5 µM, CFZ treatment naïve injected group and DMSO treated un-injected group are also inclouded. After 24h, dead embryos/well are counted and imaged. The experiment was triplicated.

DECLARATION OF INTERESTS
The authors declare no competing interests.

23.
Hayashida N, Fujimoto M, Tan K, Prakasam R, Shinkawa T, Li L, Ichikawa H, Takii R & Nakai A (2010) Heat shock factor 1 ameliorates proteotoxicity in cooperation with the transcription factor NFAT.  Scattered controls of cells not treated with dox, or only treated with dox but without additional compounds were used for normalization. (B) Hit distribution of the screen described in (C). Compounds that led to an increase in nuclei numbers higher than 3SD when compared to the numbers founds on wells only treated with dox were taken for secondary validation (Fig S1).    *p<0,05, **p<0.01, t-test.