Cysteine Modification by Ebselen Reduces the Stability and Cellular Levels of 14-3-3 Proteins

The 14-3-3 proteins constitute a family of adaptor proteins with many binding partners and biological functions, and they are considered promising drug targets in cancer and neuropsychiatry. By screening 1280 small-molecule drugs using differential scanning fluorimetry (DSF), we found 15 compounds that decreased the thermal stability of 14-3-3f. Among these compounds, ebselen was identified as a covalent, destabilizing ligand of 14-3-3 isoforms f, e, c, and g. Ebselen bonding decreased 14-3-3f binding to its partner Ser19-phosphorylated tyrosine hydroxylase. Characterization of site-directed mutants at cysteine residues in 14-3-3f (C25, C94, and C189) by DSF and mass spectroscopy revealed covalent modification by ebselen of all cysteines through a selenylsulfide bond. C25 appeared to be the preferential site of ebselen interaction in vitro, whereas modification of C94 was the main determinant for protein destabilization. At therapeutically relevant concentrations, ebselen and ebselen oxide caused decreased 14-3-3 levels in SH-SY5Y cells, accompanied with an increased degradation, most probably by the ubiquitin-dependent proteasome pathway. Moreover, ebselen-treated zebrafish displayed decreased brain 14-3-3 content, a freezing phenotype, and reduced mobility, resembling the effects of lithium, consistent with its proposed action as a safer lithium-mimetic drug. Ebselen has recently emerged as a promising drug candidate in several medical areas, such as cancer, neuropsychiatric disorders, and infectious diseases, including coronavirus disease 2019. Its pleiotropic actions are attributed to antioxidant effects and formation of selenosulfides with critical cysteine residues in proteins. Our work indicates that a destabilization of 14-3-3 may affect the protein interaction networks of this protein family, contributing to the therapeutic potential of ebselen. SIGNIFICANCE STATEMENT There is currently great interest in the repurposing of established drugs for new indications and therapeutic targets. This study shows that ebselen, which is a promising drug candidate against cancer, bipolar disorder, and the viral infection coronavirus disease 2019, covalently bonds to cysteine residues in 143-3 adaptor proteins, triggering destabilization and increased degradation in cells and intact brain tissue when used in therapeutic concentrations, potentially explaining the behavioral, anti-inflammatory, and antineoplastic effects of this drug.


SIGNIFICANCE STATEMENT
There is currently great interest in the repurposing of established drugs for new indications and therapeutic targets. This study shows that ebselen, which is a promising drug candidate against cancer, bipolar disorder and the virus infection Covid-19, covalently bonds to cysteine residues in 14-3-3 adaptor proteins, triggering destabilization and increased degradation in cells and intact brain tissue when used in therapeutic concentrations, potentially explaining the behavioral, antiinflammatory and anti-neoplastic effects of this drug.
Dimers of 14-3-3 protein present an overall W-like shaped structure with the ligand binding grooves facing each other. Each subunit has a characteristic "C" shape and is made up of nine antiparallel αhelices ( Figure 1A). The canonical binding site for protein-protein interaction (PPI) in 14-3-3 is situated on the concave surface of each subunit, and binding of a partner protein to 14-3-3 proteins is typically dependent on a phosphorylated Ser or Thr flanked by Arg and Pro residues (Aitken 2006; Obsil and Obsilova 2011).
Here we report the characterization of a previously unobserved effect of ebselen on 14-3-3 proteins at the molecular, cellular and animal levels. These effects are associated with a covalent interaction with 14-3-3 cysteine residues. Ebselen can interact covalently with cysteine residues in several protein targets (Mukherjee et al. 2014;Favrot et al. 2013;Joice et al. 2013;Lieberman et al. 2014). It 6 seems to bind to all 14 Cys residues of the nonstructural protein 3 helicase (Mukherjee et al. 2014) but has also been reported to only interact with specific Cys residues in e.g. superoxide dismutase 1 (Capper et al. 2018). Modification of Cys residues by ebselen has been reported to inhibit enzymatic function by forcing conformational changes in target proteins (Favrot et al. 2013;Mukherjee et al. 2014). Ebselen has been proposed as a safe lithium mimetic Singh et al. 2013;Singh et al. 2016) and has been subject to phase II clinical trials for bipolar disorder. Ebselen is also in clinical trials for hearing disorders, diabetes and Covid-19 (https://clinicaltrials.gov/). Out of 10 000 approved drugs and compounds in clinical trials, ebselen showed the strongest inhibition of the main protease (M pro ) of SARS-CoV-2 (Jin et al. 2020). Ebselen has also shown activity against other virus protein targets and has been proposed to protect against organ injury caused by Covid-19 (Sies and Parnham 2020;Menendez et al. 2020). Here we show that although all Cys residues in 14-3-3 proteins have the potential to be modified by ebselen, they have different reactivities, and a single residue (C94) is responsible for the protein destabilization observed at low concentrations of ebselen. Ebselen is capable of crossing the blood brain barrier and is pharmacologically active in the brain, making it a compound of interest in drug discovery campaigns related to neurological disorders (Singh et al. 2013). We propose that our findings may provide a foundation for further studies on the repurposing of ebselen and its use as a covalent modulator of 14-3-3 PPI with several therapeutic implications. 7.4. The GST tagged 14-3-3ζ was immobilized on Glutathione Sepharose 4B (GE Healthcare) and cleaved overnight at 4 o C using 10 U/mL of thrombin (Sigma-Aldrich). The cleaved 14-3-3ζ was collected as flow through. The flow through was filtered using Ultrafree-MC GV centrifugation filters (Merck) and purified further on Superdex 200 increase 10/300 GL (GE Healthcare) gel filtration column using 20 mM HEPES, 200 mM NaCl, pH 7.4. The 14-3-3ζ peak was collected and concentrated using Amicon Ultra 15 mL Centrifugal Filters with a cut-off of 30K (Merck). 14-3-3γ, η and ε were purified as previously reported (Ghorbani et al. 2016;Kleppe et al. 2014). 14-3-3 cysteine mutants were prepared as previously reported (Jandova et al. 2018). TH expression, purification and phosphorylation was performed as reported (Kleppe et al. 2014). Dephosphorylation assay was performed in 15 mM Hepes, 150 mM NaCl, pH 7.4, 10 % glycerol, using TH labelled on Ser19 by PRAK and shrimp alkaline phosphatase essentially as described (Kleppe et al. 2014;Ghorbani et al. 2016), except that no reducing agents were used. Thus, TH (2 mg/ml) was phosphorylated using active p38regulated/activated protein kinase (PRAK, 7 U/ml) essentially as described (Kleppe et al. 2014, Ghorbani et al. 2016), but in 15 mM Hepes (pH 7.4), 150 mM NaCl, 10 % glycerol without reducing agents, and the PRAK inhibitor epigallocatechin gallate was added (100 μM) after 30 min (25 C) and pSer19TH kept on ice until used for dephosphorylation. TH (2 μM) with or without 14-3-3 (10 μM) was preincubated on ice for 5 min before starting the dephosphorylation reaction (25 C) by adding shrimp alkaline phosphatase (0.12 U/μl) and sampling remaining [ 32 P]-labelled TH on phosphocellulose filters washed in phosphoric acid. Dephosphorylation sample without added phosphatase showed stable level of [ 32 P]-pSer19TH throughout the experiment and was used as 100 %. Except when otherwise indicated, ebselen and ebselen oxide -treated 14-3-3 proteins were prepared by incubating purified 14-3-3 proteins (without any reducing agents) with a 10 fold molar ratio ebselen : 14-3-3 subunits on ice for 30 min, followed by gel filtration to remove excess of compound. Once modified, the Tm-values of the proteins were stable at 4 o C for at least 6 h, and after several cycles of freezing and thawing.

Differential scanning fluorimetry (DSF) and DSF-monitored screening
For DSF-based screening the Agilent Bravo Automated Liquid Handling Platform was used for liquid handling. The final screening volume was 10 μL, with 0.1 mg/mL 14-3-3 in 20 mM HEPES, 200 mM NaCl, pH 7.4, a compound concentration of 400 μM and 4% DMSO, in 384 well plates, with 24 DMSO (4%) controls in each plate. 5X SYPRO Orange (Sigma) was used to monitor protein unfolding in Roche Lightcycler 480. Unfolding curves were recorded with 0.2 °C-intervals with a scan rate of 2 o C/min from 20 to 95 o C, with 4 acquisitions per o C including a 10 s hold at 20 o C before and after the experiment, and monitored at λex = 465 nm and λem = 610 nm. The experimental unfolding curves were normalized to fraction of unfolded protein and analyzed to extract the Tm-values for the protein without (DMSO controls) and with compounds (Urbaneja et al. 2017). Tm-values (average ± standard deviation (SD)) were calculated for 24 DMSO controls. ∆Tm-values were calculated as the Tm for the protein with compound minus the Tm-value for DMSO control. Any compounds above or below the +/-10xSD (for DMSO control) cutoff value were selected as hits for further validations. The first validation of the initial hits was performed by concentration dependent DSF by measuring 20 points between 0-400 μM and 0-16 μM (n=4). DSF experiments with other 14-3-3 isoforms (γ, η and ε) and 14-3-3ζ mutants were carried out at the same conditions as above. The effect of ebselen on the time dependent unfolding of 14-3-3ζ was measured by monitoring the increase in SYPRO Orange fluorescence, at 37 o C for 17 h, with measurement each 2:17 (min:sec), using the same conditions as for DSF, with 0.1 mg/mL of ebselen-treated and untreated 14-3-3ζ (n=3). The final 14-3-3 protein concentration in these experiments was 0.1 mg/ml (3.7 μM subunit concentration with 2% DMSO in the controls) and the final ebselen concentration was 200 μM (with 2% DMSO).
Quartz-cuvettes with 0.1 cm path length were used. The samples were measured at 25 o C covering a range of 180-260 nm with a scan rate of 50 nm/min, data pitch of 0.5 nm, response of 0.25 s, sensitivity of 100 mdeg, and 3 accumulations with a band width of 1 nm. The data pitch was 0.2 o C, sensitivity was 100 mdeg with a response of 1 s with a band width of 1 nm. The mean residual 9 ellipticity was determined using the formula [θ]mrw = θ/(n·c·l), where θ is the ellipticity (mdeg), c is the protein concentration (M), n is the number of amino acids and l is the path length of the cuvette (cm). Three parallels were averaged after subtraction of baseline spectra and calculation of mean residual ellipticity. Circular dichroism by neural networks (CDNN) (Böhm, Muhr, and Jaenicke 1992) was used to estimate secondary structure content.

Surface plasmon resonance (SPR)
Unless otherwise stated, the method below is the same for all SPR experiments. The Biacore T200 instrument (GE Healthcare Life Sciences) was used. Immobilization of full length 14-3-3ζ (1 mg/ml) was carried out in 10 mM sodium acetate pH 5.1 onto the CM5 sensor chip (GE Healthcare Life Sciences) using standard amine coupling procedure by using PBS as running buffer. The

Mass spectrometric LC-MS/MS analysis
20 g of protein pellet from ebselen-treated (see above) and untreated 14-3-3ζ were solubilized and trypsinized at 37 °C overnight. Tryptic peptides were desalted and purified using a reverse-phase Oasis HLB μElution Plate 30 μm (2-mg HLB sorbent, Waters, Milford, MA). After purification of tryptic peptides on Oasis c18 columns, the samples were applied on LC column coupled to Orbitrap Elite in the collision-induced dissociation (CID) and higher energy collisional dissociation (HCD) modes, or on LC column coupled to Orbitrap QExactive. The raw files were analyzed by either Proteome Discoverer software version 2.1.0.81 or Peptide Shaker version 1.16.27 (Vaudel et al. 2015). Three search engines were used MS-Amanda, Sequest and XTandem with a fragment ion mass tolerance of 0.02 Da in HCD mode and 0.6 Da in CID mode and a parent ion tolerance of 10 ppm.
Oxidation of methionine and modification of cysteine by ebselen were specified as variable modifications, and a human database was used. Scaffold (version Scaffold_4.8.7, Proteome Software Inc., Portland, OR) was used to validate MS/MS based peptide and protein identifications.

MALDI-MS
The protein samples were incubated with porcine trypsin (Promega) for 2 h at 40 o C in 50 mM ammonium bicarbonate. Matrix-assisted laser desorption/ionization time of flight (MALDI-TOF) mass spectra were obtained using an Ultraflextreme instrument (Bruker Daltonics, Bremen, Germany) operated in linear and reflectron positive ion detection modes for analysis of intact mass of the protein and after proteolysis by trypsin, respectively. The identity of the tryptic peptides was verified by means of MALDI-MS/MS. was used as a nuclear stain. Images were obtained by using Leica microscope TCS SP5 (Leica Microsystem GmbH).

Lysate preparation
Lysates from SHSY5Y cells and zebrafish brain tissues (see below) were prepared in lysis buffer (20 mM HEPES pH 7.0, 10 mM KCl, 1% NP40) containing protease and phosphatase inhibitor. Triton X-100 (1%) was applied to the brain lysates prior to 30 min incubation at 4 °C. Lysates were centrifuged at 10000 rpm for 10 min at 4 o C, supernatant was collected, and protein concentration was measured by using direct detect spectrometer (Merck).

Western blotting (WB)
This article has not been copyedited and formatted. The final version may differ from this version.

Zebrafish strains, care and maintenance
AB wild-type adult zebrafish (originally from the Zebrafish International Resource Centre; RRID: ZIRC_ZL1) were maintained at the University of Leicester using standard zebrafishkeeping protocols and in accordance with institute guidelines for animal welfare.
They were fed twice per day with ZEBRAFEED 400-600 dry food (Sparos). Zebrafish used for behavioural analysis were randomly netted from a large home tank containing approximately 40 fish, and included a mixture of 3 month old male and female AB wild-types. All zebrafish experiments have been approved by the local Animal Welfare and Ethical Review Board (AWERB) and are covered by a UK project licence to Dr Norton, PPL P56FB749.
For experiments with zebrafish brain tissue ebselen was diluted in system water and applied to AB wild-type zebrafish by immersion for 30 min before behavioral testing and for 18 h before dissection for Western blot analysis. Treatment duration and concentrations were chosen according to pilot experiments in our lab. In initial screening experiments, we tested ebselen concentrations between 1-5 μM to make conditions comparable with cell culture experiments. At 5 μM ebselen, we observed approximately 50% lethality of the zebrafish, while at 1 and 1.5 μM all fish survived during the observation period, without obvious symptoms of distress. In accordance with animal welfare regulations (RRR), we decided to limit the subsequent experiments to a fixed concentration of 1 μM and group sizes of three individuals. At this concentration we observed consistent behavioral effects This article has not been copyedited and formatted. The final version may differ from this version. without any lethality. Three adult zebrafish were used for Western blotting. They were anesthetized with MS222 and sacrificed by decapitation. The brains were quickly dissected in PBS and frozen in liquid nitrogen.

Behavior experiments
All behavioral experiments were carried out between 11:00 and 17:00. Experiments were performed in a dedicated room with light and temperature kept constant. These behavioral experiments are classified as mild on our project license. We minimized any potential suffering by maintaining zebrafish in oxygenated water at the correct temperature of 28 o C during the experimental period. The time-line of the study was as follows: Visually-mediated social preference for novelty. The social preference for novelty experiment was performed in a transparent rectangular tank composed of five chambers. One central chamber (19 × 13 cm) was surrounded by four identical chambers (9.5 × 6.5 cm). The transparent walls dividing the chambers contained perforated holes to permit water to move between compartments. Social preference test. 3 female adult zebrafish were used as a social stimulus. Their behavior was not included in the data presented here. They were placed in the top left chamber and left to habituate to the novel setting for 5 min. A mixture of ten male and female adult zebrafish were recorded as focal animals. These test fish were placed individually in the central chamber and their behavior was recorded for five minutes from above. The videos were then analyzed to measure time spent swimming and freezing, the primary and secondary endpoints of this study. The zebrafish treated with ebselen were exposed to 1 μM for 30 min. Control zebrafish were immersed in 0.01% DMSO. No exclusion criteria were used in this test, meaning that all animals tested were included in the statistical analysis.

Novel tank test and open field test.
Tests were performed in a standard 1.5 L trapezoid tank (Egan et al. 2009) (novel tank test) or a large rectangular arena (43 x 22 x 8 cm). A mixture of ten male and female adult zebrafish were placed individually into the tank and recorded for 5 min. Ethovision software (EthoVision XT, RRID:SCR_000441; Noldus) was used to measure locomotion and the amount of time spent freezing. No exclusion criteria were used in this test, meaning that all animals tested were included in the statistical analysis.

Statistical analysis and study design
Triplicate measurements with different sample preparations were performed and either representative spectra or resulting values as mean ± SD are provided in text or in figure legends. The two-tailed Student's t-test and Welch's two-tailed t-test were used to evaluate significant differences between the samples. For Western blot-based analyses with lysates from neuroblastoma cells and zebrafish, treated samples were referenced to untreated samples, which were given the arbitrary value of 1. All data are presented as mean ± SD of measurements with at least three independent cell culture preparations. Assessment of normality of the data was done using Shapiro-Wilk test. Statistical signification was analyzed by Student's t-test (two-tailed) or one-way ANOVA with the Holm-Sidak method. For the behavioral studies with zebrafish we used 10-12 zebrafish, as indicated, for both the social preference for novelty and novel tank diving assay. This sample size was selected based upon a power analysis using recently reported data (Dalla Vecchia et al. 2019). No blinding was performed.
After recording, behaviour was analysed using Ethovision XT tracking software, permitting unbiased comparison of results between treatment groups. All animals were included in the analysis. Zebrafish were then killed using a schedule 1 procedure: overdose of MS222 followed by decapitation as specified by ASPA 1986. Data was first assessed for normality using the Shapiro-Wilk normality test.
We did not test for or remove outliers. Statistical analyses were performed using a Mann-Whitney U test or an unpaired Student's t-test with Welch's correction. Statistical analyses were performed using the software GraphPad Prism 8.1.1 (GraphPad Prism, RRID:SCR_002798) and SigmaPlot 13.1 (SigmaPlot, RRID:SCR_003210), and statistical significance was set at p < 0.05 for all experiments.
This article has not been copyedited and formatted. The final version may differ from this version.

RESULTS
Differential scanning fluorimetry (DSF)-based screening for drugs targeting 14-3-3ζ. Identification of ebselen as a destabilizer that covalently bonds to 14-3-3ζ To identify compounds capable of binding to 14-3-3ζ, a DSF-based high throughput screen was carried out using an established protocol for identification of binders to protein targets (Niesen, Berglund, and Vedadi 2007;Aubi et al. 2015;Urbaneja et al. 2017). A similar technique has previously been used to identify peptide stabilizers of 14-3-3s (Valenti et al. 2019). The Prestwick Chemical Library (PCL) containing 1280 compounds, 95% of which are approved drugs, was screened. The midpoint melting temperature (Tm) of control samples of 14-3-3ζ with 4% DMSO was 61.1 ± 0.5°C, in agreement with Tm-values previously measured for this protein by thermal dependent CD (Ghosh et al. 2015). Hits were identified based on the change in Tm for the protein (ΔTm). A I∆TmI ≤ (10 x SD (for the DMSO controls)) was selected as the cutoff-value for hit identification. Only destabilizing drugs were identified for 14-3-3ζ and 15 drugs that decreased the Tm of 14-3-3ζ below 56.01 ± 0.5 o C (I∆TmI ≥ 5.1 ± 0.5 o C) were selected as primary hits ( Figure 1B and Supplemental Table S1). These 15 hits were subjected to detailed concentration dependent DSF analyses at concentrations up to 125 μM.
Thimerosal and ebselen showed the most consistent concentration dependent binding curves at these conditions, with large maximal ΔTm-values -11.6 ± 0.2 C and -8.1 ± 0.2 C, respectively; Figure   1B), without inducing 14-3-3ζ denaturation. The organomercury antiseptic compound thimerosal was, however, excluded from further studies on effects on 14-3-3 proteins due to its associated cytotoxicity (Parran, Barker, and Ehrich 2005;James et al. 2005) and ebselen was pursued as the primary hit compound to understand its destabilizing effect and possible interference on partner binding to 14-3-3ζ. The concentration of ebselen that provided a half-maximal decrease in ∆Tm was defined as the EC50,, estimated to be 1.50 ± 0.30 (SD) μM for the interaction of ebselen with 14-3-3ζ ( Figure 1C). The effects of ebselen on 14-3-3 conformation were studied by CD spectroscopy. The CD signal was slightly lower after ebselen treatment ( Figure 1D). 14-3-3ζ is a largely α-helical protein ( Figure 1A) and, accordingly, the percentage of α-helix calculated using CDNN (Böhm, Muhr, and Jaenicke 1992) for untreated and ebselen treated 14-3-3ζ was approx. 86.4 ± 0.8% and 78.6 ± 1.2%, respectively. These results indicate that the destabilization of the protein caused by ebselen triggers a discrete conformational change of 14-3-3ζ, and is not associated with a large denaturation.

< Fig. 1 >
This article has not been copyedited and formatted. The final version may differ from this version. Ebselen covalently attaches to all cysteine residues in 14-3-3ζ and modification of C94 induces protein destabilization The interaction between 14-3-3ζ and ebselen was further validated by surface plasmon resonance (SPR). The sensorgram shows that ebselen covalently attaches to the protein and does not dissociate at concentrations up to 25 μM, when a plateau is reached for the response signal quantified in resonance units (RU), suggesting covalent bonding (Figure 2A). Ebselen has previously been shown to modify protein targets mainly through covalent interaction with cysteine residues (Mukherjee et al. 2014;Favrot et al. 2013;Joice et al. 2013;Lieberman et al. 2014). To investigate the possible involvement of cysteines in ebselen-induced destabilization of 14-3-3ζ, we prepared Cys mutants that were analyzed by mass spectroscopy (MS) to identify cysteine residues modified by ebselen treatment of 14-3-3ζ. Using MS/MS with CID detection, we detected ebselen-derivatized by selenylsulphide bonds to tryptic peptides containing either C25, C94 or C189. The identification of a peptide with C94 is illustrated in Figure 2B. Detailed liquid chromatography-tandem MS (LC-MS/MS) and MALDI-TOF MS experiments using the mutants C25A-C189A-14-3-3ζ ( Figure 2C and S2A) and C94A-14-3-3ζ ( Figure 2D and S2B) contributed to identify C25 as the preferential site of ebselen conjugation with high confidence, followed by C94 and C189. A preferential S-glutathionylation (Kim et al. 2014) and S-nitrosylation (Greco et al. 2006) labelling of C25 in 14-3-3ζ has also been previously reported.

< Fig. 2 >
We also investigated the thermal stability of the Cys mutants by DSF (Table 1). It has been shown that the buried C94 is a determinant for the stability of 14-3-3ζ (Jandova et al. 2018). Accordingly, we noted a protein destabilization caused by mutation of C94, observed in the single mutant C94A and in the double (C25A-C94A and C94A-C189A) and triple mutants, which presented a decrease in Tm of 3.2-5.9 o C compared with wild-type (WT)-14-3-3ζ (Table 1). On the other hand, mutations at C25 and C189 did not affect the protein stability, as seen with the C25A and C25A-C189A mutants (Table 1).
With respect to the ebselen effect, both triple mutants (C25A-C94I-C189A and C25A-C94V-C189A) were unaffected by ebselen treatment, which excluded the possibility that the destabilization originated from modification of 14-3-3ζ at other residues than cysteine. The results also indicate that ebselen induced only minor changes in Tm in C94-mutants, while the two mutants at other residues This article has not been copyedited and formatted. The final version may differ from this version. but with C94 intact (C25A and C25A-C189A) showed a similar destabilization (8.6-9.0 o C decrease in Tm) as WT-14-3-3ζ (Table 1), pointing to C94 as the site for ebselenation-mediated destabilization.
Altogether, the MS analyses of 14-3-3ζ and Cys mutants revealed a complex pattern of Cys modification by ebselen, which appeared to occur at all Cys residues, to a larger degree for C25 and lesser extent for C94 and C189, whereas the DSF assays of mutated proteins (Table 1) indicate that the destabilization is dependent on the ebselen-modification of C94.

Ebselen modification of 14-3-3ζ affects its kinetic stability and its interaction with TH phosphorylated at Ser19 (pSer19-TH)
Based on the discrete destabilization of 14-3-3ζ observed by ebselen treatment, leading to an approx.
7 o C reduction in Tm (see above and Table 1), it is not straightforward to predict subsequent physiological consequences of ebselen modification. By measuring the fluorescence-monitored isothermal loss of conformational stability of the protein at 37 o C we observed a several foldincreased rate of unfolding for the ebselen-treated 14-3-3ζ compared with the untreated control (Fig.   3A). This indicates that the protein may show altered stability also at physiological conditions. The functional consequences of the destabilizing conformational change of 14-3-3ζ was explored by examining its interaction with a physiologically relevant protein partner. To this end, we measured the binding of the archetypical 14-3-3 partner Ser19-phosphorylated TH (pSer19TH) to ebselentreated and untreated 14-3-3ζ by SPR. Complex formation of pSer19TH with 14-3-3 proteins is of high affinity, and activates and stabilizes TH (Ghorbani et al. 2016). As seen in Figure 3B, the binding of ebselen treated 14-3-3ζ to pSer19TH was reduced compared to the untreated protein.

Ebselen destabilizes several 14-3-3 isoforms
We further investigated if the destabilizing effect of ebselen on 14-3-3ζ could also be observed for other 14-3-3 isoforms. All these proteins have conserved Cys residues at positions equivalent to C94 in the ζ isoform; C189 is conserved in all isoforms except in σ and C25 is not conserved ( Figure S1).
14-3-3η showed the lowest thermal stability (Tm = 47.9 ± 0.2 o C), followed by 14-3-3γ (56.9 ± 0.2 o C), while 14-3-3ε (59.7 ± 0.1 o C) and 14-3-3ζ (59.9 ± 0.3 o C) had similar Tm-values (Table 1). All tested This article has not been copyedited and formatted. The final version may differ from this version. isoforms were destabilized by ebselen, however, in contrast to 14-3-3ζ, ebselen treatment did not elicit a discrete destabilizing conformational change in these isoforms, but rather a major destabilization, as observed for the ε and γ isoforms, or denaturation for η ( Figure 3C, Table 1 and data not shown). Curiously, a concentration effect threshold was seen at low concentrations of ebselen in the concentration dependent destabilization (∆Tm-values) for 14-3-3 ζ ( Figure 3C), which was not observed in the other tested isoforms, and 14-3-3ζ was also the most resilient isoform against the destabilizing effects of ebselen. The apparent concentration threshold of 14-3-3ζ modification may indicate that an intact C25 exerts a stabilizing effect on this protein. Only when C25 has been modified, the consecutive bonding of ebselen to C94 would lead to destabilization of the ζ isoform.
We have reported that the interaction between pSer19TH and 14-3-3 is conserved between different isoforms, but that 14-3-3 had the strongest effect on activation and inhibition of pSer19TH dephosphorylation (Ghorbani et al. 2016). 14-3-3 proteins also modulated access to other regulatory sites for multi-site phosphorylation of TH (Ghorbani et al. 2020). As inhibition of TH dephosphorylation is a sensitive functional readout of 14-3-3, we conducted dephosphorylation experiments with [ 32 P]-labelled pSer19TH in the presence or absence of 14-3-3 or 14-3-3 pretreated with ebselen or ebselen oxide (Fig. 3D). PRAK phosphorylated TH (pSer19TH) was stably labeled throughout the dephosphorylation experiment as shown in absence of added phosphatase ( Fig. 3D) and was rapidly dephosphorylated in the absence of unmodified 14-3-3 , which inhibited dephosphorylation substantially. This was however not the case for ebselen treated 14-3-3, which showed no significant inhibition compared to no 14-3-3, whereas ebselen oxide treated 14-3-3 showed some residual inhibition (Fig. 3D).
In conclusion, it appears that 14-3-3 isoforms behave differently in the presence of ebselen, and these differences can be hypothesized to be due to differences in cysteine residues within the protein.

< Fig. 3 > Ebselen treatment of SHSY5Y cells triggers proteasomal degradation of 14-3-3
To test the effects of ebselen on 14-3-3 protein stability and turnover in intact cells, we performed immunofluorescence imaging of SHSY5Y cells treated with ebselen. After evaluating the viability of This article has not been copyedited and formatted. The final version may differ from this version. the cells at different ebselen and DMSO concentrations, 5 μM ebselen with 0.05% DMSO was selected, using untreated as well as a DMSO-control (0.05%) cells for comparison ( Figure 4A). The levels of DAPI, 14-3-3, and β-actin were virtually unchanged for the untreated and DMSO control cells. However, the 14-3-3 signal was significantly reduced in cells treated with ebselen. This was confirmed by Western blotting experiments and densitometric analysis, showing that ebselentreated cells had 28 ± 13% decreased 14-3-3 levels compared to untreated cells ( Figure 4B). To investigate the mechanism for 14-3-3 depletion, cells were treated with ebselen, a proteasome inhibitor (Bortezomib), and a caspase inhibitor (Z-VAD-FMK). Cell samples pretreated with ebselen, and then treated with either the proteasomal inhibitor or a mixture of proteasomal and caspase inhibitors, showed a significant increase in 14-3-3 level compared to cell samples only treated with ebselen. However, samples treated with ebselen and then with the caspase inhibitor showed similar level of 14-3-3 as the samples treated only with ebselen. These results indicate that ebselen treatment triggers a proteasome-mediated degradation of 14-3-3 in SHSY5Y cells ( Figure 4C).

Comparison of ebselen and ebselen oxide
Previous studies have shown that ebselen can act as an antioxidant and reduce hydroperoxides.
During this process, ebselen is converted into the antioxidant inactive derivative ebselen oxide (Lass et al. 1996). To explore whether the cellular effects of ebselen might be mediated by its oxidized derivative, we compared the effects of ebselen and ebselen oxide. As shown in Fig. 5, ebselen and ebselen oxide had slightly different effects on the viability of SHSY5Y cells (Supplementary Figure S3).
At 20 μM ebselen oxide, the cellular viability was reduced by 71 %, compared to 41 % reduced viability by ebselen at this concentration (P= 0.007, for comparison, t-test). However, the effects of ebselen and ebselen oxide on cellular levels of 14-3-3 were rather similar, as reported for other protein targets (Wang et al. 2017). For both compounds, the levels of 14-3-3s gradually decreased at concentrations exceeding 2 μM and the effects were abolished by treatment with bortezomib. A similar decrease of 14-3-3 levels was observed using either14-3-3ζ specific antibodies or pan-14-3-3 antibodies, suggesting that the decrease was not limited to one particular isoform. Based on these studies, we conclude that the observed effects of ebselen on intact cells could (partially) be mediated by its oxidized derivative(s).

< Fig. 5 >
This article has not been copyedited and formatted. The final version may differ from this version.

In vivo action of ebselen in zebrafish brain
The effect of ebselen on 14-3-3 proteins was also studied in adult zebrafish. Zebrafish were treated with 1 μM ebselen for 18 h before the brains were collected. Western blot was performed using the brain lysates of ebselen treated zebrafish vs untreated (DMSO control). Densitometric analysis showed a significant reduction of 14-3-3 in the brain of zebrafish treated with ebselen, compared to untreated control ( Figure 6A). This reduction in 14-3-3 abundance was in accordance with our observation in SHSY5Y cells (Figure 4).
A visually mediated social preference test was carried out following treatment with 1 μM ebselen for 30 min (n = 30) ( Figure 6B) (Carreño Gutiérrez et al. 2019). Zebrafish immersed in 0.01% DMSO control water froze for approximately 25 s. In comparison, zebrafish treated with 1 μM ebselen froze for more than 100 s ( Figure 6C). Moreover, zebrafish treated with DMSO swam for approximately 1250 cm, whereas the ebselen treated zebrafish only swam about 500 cm ( Figure 6D). The novel tank test ( Figure 6E) and the open field test ( Figure 6H) were also carried out. These tests revealed no significant differences between DMSO and ebselen treated zebrafish in terms of time spent frozen and distance travelled ( Figure 6F,G,I,J). This is similar to previous research reporting that zebrafish treated with lithium display approximately half the mobility of untreated zebrafish (Nery et al. 2014).
The lithium effect has been associated to direct binding and inhibition of targets such as glycogen synthase kinase 3ß (GSK3B) and inositol monophosphatase (IMPase). Interestingly, IMPase is also efficiently inhibited by ebselen (Singh et al. 2013;Masaki et al. 2016;Singh et al. 2016). To explore this connection, we used a DSF-monitored binding assay to investigate whether lithium would directly bind to 14-3-3ζ. However, no apparent binding was observed when lithium tested in the concentration range 0.01 μM to 10 mM (data not shown).

< Fig. 6 >
This article has not been copyedited and formatted. The final version may differ from this version.

DISCUSSION
In this work, we screened a chemical library to find compounds that interact with 14-3-3ζ. This isoform was selected for initial screening due to its high abundance in the CNS and strong involvement in human physiological functions and disorders (Torrico et al. 2020). Strikingly, only destabilizing compounds were found. This is unusual for DSF-based screens, which customarily result in identification of both stabilizing and destabilizing ligands (Niesen, Berglund, and Vedadi 2007;Aubi et al. 2015). The 14-3-3 destabilizing drugs belong to different therapeutic classes, with an overrepresentation of proton pump inhibitors and calcium channel blockers, several of them including S or Se in their structure (Table S1). Ebselen was identified as a destabilizer through covalent bonding to 14-3-3ζ, bonds, and similarly shown to modify and destabilize 14-3-3ε, γ and η, but with markedly large differences between these highly conserved isoforms.
A concentration effect-threshold was observed at the lowest concentrations in the concentration dependent DSF screen of 14-3-3ζ, most probably explained by the stabilizing effect of C25, which is not present in isoforms ε, γ or η (Supplemental Fig. S1), whereas destabilization appears associated to ebselen modification of C94. Protein destabilization brought about by conformational changes has been previously shown as an effective mechanism of drug action (Ren et al. 2018). In the case of the ebselenated Mycobacterium tuberculosis antigen 85 (Mtb Ag85) complex, it has been shown that ebselen bonding triggers structural changes accompanied by a decrease in protein thermal stability This article has not been copyedited and formatted. The final version may differ from this version.  (Goins et al. 2017). Thus, the CD results corroborate the destabilizing effect of ebselen due to a discrete conformational effect on 14-3-3ζ without denaturing the protein, even after treatment with a high and saturating concentration of ebselen (200 μM).
The functional consequences of the destabilizing conformational change of 14-3-3ζ were demonstrated by showing a decreased interaction with the physiologically relevant protein partner TH. Together with the tryptophan hydroxylases (TPHs) 1 and 2, TH was amongst the first proteins identified to bind to 14-3-3's (Ichimura et al. 1987). The interaction between 14-3-3 proteins and TH primarily takes place at phosphorylated Ser19 on the N terminal tail of TH. This phosphorylated amino acid interacts with two conserved Arg and one Tyr residue of 14-3-3 within the main binding cleft (PDB 4J6S) (Skjevik et al. 2014). TH catalyzes the rate-limiting step in the biosynthesis of dopamine. Thus, compounds that modulate the TH:14-3-3 interaction could represent potential therapeutics for disorders related to altered dopamine function, such as attention deficit hyperactive disorder, parkinsonism, and Parkinson's disease (Del Campo et al. 2011;Waloen et al. 2017).
We also explored how ebselen treatment influenced 14-3-3 levels in intact cells and tissues.
Increased protein instability caused by mutations or chemical modifications usually results in dysfunction and reduction of cellular protein levels due to increased processing of the proteins by quality control systems, notably through the ubiquitin-dependent proteasome system (Vilchez, Saez, and Dillin 2014). Caspase-dependent degradation has been reported to be implicated in the reduction of 14-3-3ζ levels following intracellular oxidation (Kim et al. 2014). On the other hand proteasome-mediated 14-3-3 degradation of the γ isoform has also been reported (Chen et al. 2015).
The function and regulation of 14-3-3 proteins have been studied in several cell types derived from the nervous system, notably the SHSY5Y neuroblastoma cell line. These cells express multiple 14-3-3 isoforms, including 14-3-3 ε and ζ that were of particular interest in our study. It was recently shown that these 14-3-3 isoforms are subject to ubiquitin regulated proteasomal degradation in SHSY5Y cells (Jiang et al. 2019). Our observation that ebselen treatment triggered ubiquitin-mediated proteasomal degradation of 14-3-3s in SHSY5Y cells (Fig. 4C) is consistent with these findings.
Proteasome-mediated degradation of ebselen, and ebselen oxide modified proteins has previously been reported (Li et al). It nevertheless possible, that the effect of ebselen on 14-3-3 levels may be indirect, by i.e. increasing the phosphorylation, ubiquitination, or other covalent modification of 14-This article has not been copyedited and formatted. The final version may differ from this version. 3-3s, as such modifications are reported to trigger 14-3-3 degradation (Jiang et al. 2019;Chen et al. 2015).
14-3-3 proteins constitute approx. 1% of total soluble brain proteins (Aitken 2006). They are involved in multiple functions, including metabolic regulation and cortical development (Cornell and Toyo-Oka 2017). Common and rare genetic variants in 14-3-3s are implicated in multiple neuropsychiatric disorders, making the 14-3-3s and their protein-partner networks attractive as therapeutic targets in neuropsychiatry (Pagan et al. 2017;Jacobsen et al. 2015;Torrico et al. 2020). The in vivo experiments showed that zebrafish treated with ebselen exhibited a freezing behavior and reduced locomotion ( Fig. 6). Similarly, ebselen reduces impulsivity in rodent models and has been suggested as an alternative to lithium in the treatment of bipolar disorder and other mood disorders (Singh et al. 2013;Singh et al. 2016). Lithium has severe side effects, whereas ebselen has limited toxicity at brain-active therapeutic concentrations, which makes ebselen a potential safer treatment alternative (Singh et al. 2013;Singh et al. 2016). 14-3-3s regulates the established lithium target GSK3B (Liao et al. 2005;Sugden et al. 2008), pointing to crosstalk of protein networks targeted by lithium and by alteration of 14-3-3 functionality. This indicates that alteration of protein interaction networks involving 14-3-3s may contribute to the lithium-mimicking effect of ebselen. Nevertheless, elucidation of the molecular basis for the phenotype caused by ebselen treatment requires further investigation. Moreover, due to the broad and rather non-specific effects of ebselen it is very probable that other molecular targets also are involved in its lithium-mimicking activity. Interestingly, we did not observe any direct binding of lithium to 14-3-3ζ at therapeutically relevant concentrations.
Overexpression of 14-3-3 has been found in several cancer forms and is associated with poor prognosis and aggressive tumor growth (Xiao et al. 2014;Liu et al. 2014). Reduction of 14-3-3 abundance makes cancer cells more susceptible to chemotherapy treatment, and 14-3-3s are considered attractive targets for cancer therapy (Cao et al. 2015;Woodcock et al. 2015). Thus, the ebselen effect on 14-3-3 turnover and consequent alteration of the PPI network is also expected to be relevant for cancer treatment. Actually, cysteine residues in 14-3-3 are targets of photodynamic light cancer therapy mediated by reactive oxygen species, where oxidization of these residues is expected to mediate the apoptosis triggered by such treatment (Helander et al. 2016). Our work, showing the ability of ebselen to reduce the abundance and function of 14-3-3 in vivo through cysteine modification, also reinforces the promise of ebselen as an exciting candidate to use in cancer treatment in addition to mood disorders. Furthermore, this study also identified 15 other small-molecule drugs with destabilizing effects on 14-3-3ζ (Table S1). These compounds share redox activity properties with ebselen, and include S atoms, indicating that their destabilizing effects on 14-3-3s may also be mediated by interaction with cysteine residues. As ebselen is a redox active compound, its effects may be modified in cells and in vivo by glutathione, free cysteine or other cellular sulfhydryl compounds, by reducing the selenylsulphide bonds.
In summary, the destabilizing effect of ebselen binding to 14-3-3 proteins suggest that disruption of 14-3-3 protein interactions may contribute to the biological effects of ebselen in cancer, neuropsychiatric and infectious disorders. Moreover, ebselen has recently also been proposed as a treatment of Covid-19 and other respiratory viral infections (Sies and Parnham 2020). Here we show that ebselen bonding has some specificity and has differential effects also within the highly conserved group of 14-3-3 proteins. However, even more selective binding to specific Cys residues in protein targets may be necessary to achieve the specificity required for further clinical applications of this drug. Figure 1. Screening of drugs interacting with 14-3-3ζ, and the destabilization caused by ebselen. (A)
Densitometric analysis was used to determine the reduction of 14-3-3 abundance in ebselen treated zebrafish (n 3, independent lysates of 1 zebrafish each, *p<0.05). (B) Tanks used to measure visually-