Inhibition of XPO-1 Mediated Nuclear Export through the Michael-Acceptor Character of Chalcones

The nuclear export receptor exportin-1 (XPO1, CRM1) mediates the nuclear export of proteins that contain a leucine-rich nuclear export signal (NES) towards the cytoplasm. XPO1 is considered a relevant target in different human diseases, particularly in hematological malignancies, tumor resistance, inflammation, neurodegeneration and viral infections. Thus, its pharmacological inhibition is of significant therapeutic interest. The best inhibitors described so far (leptomycin B and SINE compounds) interact with XPO1 through a covalent interaction with Cys528 located in the NES-binding cleft of XPO1. Based on the well-established feature of chalcone derivatives to react with thiol groups via hetero-Michael addition reactions, we have synthesized two series of chalcones. Their capacity to react with thiol groups was tested by incubation with GSH to afford the hetero-Michael adducts that evolved backwards to the initial chalcone through a retro-Michael reaction, supporting that the covalent interaction with thiols could be reversible. The chalcone derivatives were evaluated in antiproliferative assays against a panel of cancer cell lines and as XPO1 inhibitors, and a good correlation was observed with the results obtained in both assays. Moreover, no inhibition of the cargo export was observed when the two prototype chalcones 9 and 10 were tested against a XPO1-mutated Jurkat cell line (XPO1C528S), highlighting the importance of the Cys at the NES-binding cleft for inhibition. Finally, their interaction at the molecular level at the NES-binding cleft was studied by applying the computational tool CovDock.


Introduction
Exportin-1 (XPO1, also known as chromosome region maintenance 1, CRM1) is the best-characterized nuclear transporter that mediates the traffic of high molecular weight molecules (i.e., proteins or RNA) from the nucleus to the cytoplasm [1,2]. Aberrant XPO1 function is implicated in different diseases, including different types of cancers, inflammation, neurodegeneration and viral infections [3][4][5][6][7][8]. In the nucleus, exportin-1 recognizes specific leucine-rich peptide stretches, known as nuclear export signals (NES) in cargo proteins, and forms a ternary complex with RanGTP. When this exportin-RanGTP-cargo complex reaches the cytoplasm, GTP is hydrolyzed to GDP, disrupting the ternary complex and releasing the cargo [2]. XPO1 has been found to be overexpressed in a variety of solid tumors and hematologic cancers, and in many cases, elevated XPO1 levels have been correlated to poor prognosis [7,9]. Pharmacological inhibition of XPO1 has been considered an appealing anticancer strategy [2,4,5,10]. Indeed, XPO1 cargo proteins include many tumor suppressors and cell growth regulators, such as p53, Topo2, FOXOs, etc., [7,8], and XPO1 inhibition has been shown to restore nuclear localization and function of these tumor suppressors, leading to apoptosis of the cancer cells [11].
The most potent XPO1 inhibitors described, either from natural or synthetic origin, are α,β-unsaturated carbonyl compounds that bind into the NES-binding cleft of XPO1 through covalent interactions with Cys528 [2,4,10,12,13]. These inhibitors prevent the interaction of cargo proteins with XPO1 and, hence, block cargo export to the cytoplasm [14]. Only recently, a non-covalent XPO1 inhibitor has been described [15], while a first report on allosteric inhibitors with moderate affinity has been published [16].
The first XPO1 inhibitor clinically tested was the natural product leptomycin B (1, Figure 1). However, clinical trials were discontinued due to severe cytotoxicity [17], probably associated with irreversible inhibition. Among synthetic compounds, the N-azolylacrylate compounds [12,13], also known as selective inhibitors of nuclear export (SINE) exemplified by KPT-330 (selinexor) (2),  or the second generation compound KPT-8602 (eltanexor) (4) (Figure 1), have been described as very slowly reversible inhibitors [12,[18][19][20]. Indeed, the recent FDA approval of selinexor (KPT-330, XPOVIO, 2) for the treatment of patients with heavily pretreated relapsed or refractory multiple myeloma and diffuse B cell lymphoma supports the significance of XPO1 as a target in hematological malignancies [21]. Other interesting XPO1 inhibitors are the 1-(pyridin-2-ylamino)-1H-pyrrole-2,5-dione derivatives S109 (5) and CBS9106 (6) [22,23], with the latter being evaluated in Phase 1 clinical trials [24].    The overall structure of XPO1 complexed with Ran and RanBP1 is shown in Figure 2A, where the NES-binding cleft is colored in grey. Structurally, this cleft contains five hy- drophobic pockets (named Φ0-Φ4, Figure 2B) that lodge the hydrophobic key residues of the cargo protein. Cys528 (which corresponds to Cys539 in Saccharomyces cerevisiae in the X-ray structure) is located between the Φ3 and Φ4 pockets. While leptomycin B occupies almost all Φ pockets [17], the KPT compounds exemplified by KPT-8602 ( Figure 2C) only occupy a small part of the NES-binding cleft [25], demonstrating that XPO1 inhibition can be accomplished by only occupying part of the hydrophobic cleft. By analysing the binding mode of KPT-8602, it can be concluded that an aromatic hydrophobic core is located below the Cys-reactive residue, while more polar substituents (a pyrimidinyl ring in KPT-8602) can be lodged over the Cys in the proximities of the Φ4 pocket. The overall structure of XPO1 complexed with Ran and RanBP1 is shown in Figure  2A, where the NES-binding cleft is colored in grey. Structurally, this cleft contains five hydrophobic pockets (named Φ0-Φ4, Figure 2B) that lodge the hydrophobic key residues of the cargo protein. Cys528 (which corresponds to Cys539 in Saccharomyces cerevisiae in the X-ray structure) is located between the Φ3 and Φ4 pockets. While leptomycin B occupies almost all Φ pockets [17], the KPT compounds exemplified by KPT-8602 ( Figure 2C) only occupy a small part of the NES-binding cleft [25], demonstrating that XPO1 inhibition can be accomplished by only occupying part of the hydrophobic cleft. By analysing the binding mode of KPT-8602, it can be concluded that an aromatic hydrophobic core is located below the Cys-reactive residue, while more polar substituents (a pyrimidinyl ring in KPT-8602) can be lodged over the Cys in the proximities of the Φ4 pocket. Based on these precedents, we envisioned the chalcone scaffold as a suitable α,βunsaturated carbonyl construct for XPO1 inhibition by interaction with Cys528 based on their ability to behave as Michael acceptors. Indeed, quite similar α,β-unsaturated carbonyl compounds such as caffeic acid phenylethyl ester (7, Figure 1) or curcumin (8) have been described as XPO1 inhibitors. Moreover, since chalcones can undergo a retro-Michael reaction, the expected inhibition should be reversible. Reversible covalent inhibitors may have some advantages versus their irreversible counterparts, such as the possibility to tune the residence time and/or to avoid the irreversible inhibition of off-targets [26], while as a disadvantage they may show lower potency. Compounds 9 and 10 ( Figure 3) were proposed as prototypes, so that ring A should be lodged below the Cys where the covalent interaction takes place, while the pyrimidine (ring B) should be located closer to the Φ4 pocket. The XPO1 inhibition obtained with these compounds triggered us to explore closely related structural analogues ( Figure 3). Thus, their synthesis, antiproliferative activity against a panel of cancer cell lines, XPO1 inhibition and docking studies at the NES-binding cleft of XPO1 are here described. Based on these precedents, we envisioned the chalcone scaffold as a suitable α,βunsaturated carbonyl construct for XPO1 inhibition by interaction with Cys528 based on their ability to behave as Michael acceptors. Indeed, quite similar α,β-unsaturated carbonyl compounds such as caffeic acid phenylethyl ester (7, Figure 1) or curcumin (8) have been described as XPO1 inhibitors. Moreover, since chalcones can undergo a retro-Michael reaction, the expected inhibition should be reversible. Reversible covalent inhibitors may have some advantages versus their irreversible counterparts, such as the possibility to tune the residence time and/or to avoid the irreversible inhibition of off-targets [26], while as a disadvantage they may show lower potency. Compounds 9 and 10 ( Figure 3) were proposed as prototypes, so that ring A should be lodged below the Cys where the covalent interaction takes place, while the pyrimidine (ring B) should be located closer to the Φ4 pocket. The XPO1 inhibition obtained with these compounds triggered us to explore closely related structural analogues ( Figure 3). Thus, their synthesis, antiproliferative activity against a panel of cancer cell lines, XPO1 inhibition and docking studies at the NES-binding cleft of XPO1 are here described.

Synthesis
It has been described by performing kinetic measurements in α-X-2'3,4,4´-tetra methylchalcones that the nature of the substituent at position α of the chalcone affect their reactivity as electrophiles versus thiol groups [27]. In such series, the α-COOEt sub stituent led to a similar reactivity as that of the α-H chalcone; the α-CH3 chalcone wa poorly electrophilic, while the α-CN chalcone was the most electrophilic. Thus, in th naphthyl series (Figure 3), the synthesis of the chalcones where R1 = H, COOCH3, CH3 o CN was accomplished. The reaction of 1-(naphthalen-2-yl)ethan-1-one (11) with pyrimi dine-5-carbaldehyde in the presence of Ba(OH)2 afforded α-H chalcone 9 in a 55% yield (Scheme 1). The ketone 11 reacted with dimethyl carbonate, as described, to provide th methyl 3-oxopropanoate 12 [28], whose reaction with pyrimidine-5-carbaldehyde resulted in the α-COOCH3 chalcone 13. Reaction of the ester 12 with NH3/dioxane at 110 °C [29 led to the amide 14 that was further transformed into the chalcone 15. On the other hand reaction of naphthalene 16 with propionyl chloride under Friedel-Crafts acylation condi tions, as described [30], provided the 2-acyl derivative 17, together with a small propor tion of its 1-isomer [31]. This mixture of isomers reacted with pyrimidine-5-carbaldehyd to yield the α-CH3 chalcone 18 (30%). Finally, the synthesis of the α-CN derivative wa addressed by the reaction of 2-methylnaphthoate (19) with acetonitrile to provide 3-ox opropanenitrile 20 [32], which was further transformed into the α-CN chalcone 21. Com pound 21 proved to be very unstable and readily decomposed in phosphate-buffered so lution (PBS), hampering its biological evaluation.
A second set of modifications involved the incorporation as ring B of different pyridines instead of the 5-pyrimidinyl of the prototype 9. To this end, the ketone 11 was reacted with different aldehydes (21)(22)(23) in the presence of Ba(OH) 2 , using a mixture of methanol and water at rt, to yield the chalcones 24-26 (Scheme 2). A second set of modifications involved the incorporation as ring B of different pyridines instead of the 5-pyrimidinyl of the prototype 9. To this end, the ketone 11 was reacted with different aldehydes (21)(22)(23) in the presence of Ba(OH)2, using a mixture of methanol and water at rt, to yield the chalcones 24-26 (Scheme 2). In parallel, we synthesized the 3,5-dimethoxyphenyl chalcone 10 (Scheme 3) by reaction of the 1-(3,5-dihydroxyphenyl)ethan-1-one 27 with methyl iodide to provide the ketone 28 [33], whose reaction with pyrimidine-5-carbaldehyde afforded the chalcone 10 (64% yield). Additionally, in this case, the α-COOCH3 chalcone was synthesized through the transformation of the ketone 28 into the methylpropionate 29 [34], and further reaction Scheme 2. Reagents and conditions: (a) Ba(OH) 2 , methanol/water, rt, 2-7 h, 32-73% yield.
In parallel, we synthesized the 3,5-dimethoxyphenyl chalcone 10 (Scheme 3) by reaction of the 1-(3,5-dihydroxyphenyl)ethan-1-one 27 with methyl iodide to provide the ketone 28 [33], whose reaction with pyrimidine-5-carbaldehyde afforded the chalcone 10 (64% yield). Additionally, in this case, the α-COOCH 3 chalcone was synthesized through the transformation of the ketone 28 into the methylpropionate 29 [34], and further reaction On the other hand, as will be later shown in the molecular modeling studies, one of the methoxy groups of the chalcone 10 seems to be pointing towards the lower part of the NES-binding cleft that lodges the aliphatic chain of leptomycin B. Thus, one of the methoxy groups in the chalcone 10 was replaced by either a 2-methoxyethoxy or a 3,3dimethylallyloxy group. To this end, the 3,5-dihydroxyacetophenone 27 was transformed in 2 steps into the corresponding ketones 31 and 32, whose reaction with pyrimidine-5carbaldehyde afforded the chalcones 33 and 34 in a 32% and 41% yield, respectively. On the other hand, as will be later shown in the molecular modeling studies, one of the methoxy groups of the chalcone 10 seems to be pointing towards the lower part of the NES-binding cleft that lodges the aliphatic chain of leptomycin B. Thus, one of the methoxy groups in the chalcone 10 was replaced by either a 2-methoxyethoxy or a 3,3-dimethylallyloxy group. To this end, the 3,5-dihydroxyacetophenone 27 was transformed in 2 steps into the corresponding ketones 31 and 32, whose reaction with pyrimidine-5carbaldehyde afforded the chalcones 33 and 34 in a 32 and 41% yield, respectively. For the α-H chalcones (9, 10, 24-26, 33, 34), the configuration of the double bond was easily assigned as E, as expected, based on the JH2-H3 value (15)(16) in their 1 H-NMR spectra [36]. The configuration of the α-substituted chalcones (13, 15, 18, 21, 30) was also assigned as E, based on previous reports [37,38].

Incubation with GSH
In order to determine if these chalcones could react with thiol groups through a hetero-Michael addition reaction, a few selected compounds (9, 10, 13 and 18) were incubated with glutathione (GSH). After different incubation times, the reactions were quenched by adding 5,5 -dithiobis(2-nitrobenzoic acid) (DTNB, Ellman s reagent) [39,40] (1:1 ratio with respect to the initial concentration of GSH). The course of the reaction was followed by HPLC, adapting reported procedures [41,42]. The prototype α-H chalcones 9 and 10 quickly reacted with GSH to provide the Michael addition products, as shown by HPLC-MS. These adducts underwent a retro-Michael addition reaction, generating the parent chalcone, demonstrating the reversibility of the reaction (see Figures S1 and S2). The α-COOCH 3 chalcone 13 also provided the Michael adducts with GSH after a short reaction time, while longer incubations revealed that the adducts reverted to the parent chalcone ( Figure S3). However, under the same incubation conditions with GSH, the α-CH 3 chalcone 18 remained almost unaltered even after 4 h of incubation, and only a very small proportion of the addition products was detected ( Figure S4). As already mentioned, the α-CN chalcone 21 could not even be tested due to instability in PBS, suggesting a very high electrophilic character. Thus, although this assay is only qualitative, the tendency observed among these chalcones nicely fits the reactivity described for α-X-2 3,4,4 -tetramethylchalcones [27].

Antiproliferative Activity
The synthesized compounds were tested for their antiproliferative activity against a panel of cancer cell lines using KPT-330 as a reference compound (Table 1). The two prototype compounds 9 and 10 inhibited proliferation against all cell lines tested at single digit µM IC 50 values, being between 5-to 30-fold less potent than the reference compound KPT-330. The α-substituted chalcones in the naphthyl series with an ester (compound 13) or methyl (compound 18) group were almost inactive in the proliferation assays, while the α-CONH 2 chalcone 15 was slightly less active than the parent chalcone 9. When the 5-pyrimidinyl ring in 9 was replaced by different pyridines (pyridin-3-yl in 24, pyridin-2-yl in 25 or pyridin-4-yl in 26), the IC 50 vales were maintained in the low µM range. Concerning the phenyl series, the α-COOCH 3 chalcone 30 was considerably less active than the α-H chalcone 10. Moreover, replacement of one of the methoxy groups in compound 10 by longer ethers was compatible with antiproliferative activity, although up to a 20-fold increase in IC 50 values was obtained for certain cell lines.

XPO1 Inhibition Studies
In order to determine if the observed antiproliferative activity was caused by XPO1 inhibition, the compounds were assayed in a reporter cell line based on the subcellular localization of a XPO1-dependent GFP reporter cargo protein [13]. Inhibition of XPO1mediated nuclear export of the reporter is evident by its nuclear accumulation, which can be visualized and quantified compared to the untreated control, where the GFP protein localizes in the cytoplasm. Using this reporter cell line, the capacity of the synthesized chalcones to inhibit XPO1-mediated nuclear export was determined (Table 2). The prototype chalcones 9 and 10 both in the naphthyl and phenyl series showed significant XPO1 inhibition, with IC 50 values of 2.5 and 0.55 µM, respectively. Interestingly, among the naphthyl series, the chalcones substituted with an ester or a methyl group at the α-position of the double bond (compounds 13 and 18, respectively) that were inactive in the antiproliferative activity assays, were also inactive against XPO1. On the other hand, the α-CONH 2 chalcone 15 or compounds with a pyridinyl ring (24)(25)(26) instead of the pyrimidinyl of the prototype 9, which had shown significant antiproliferative activity, inhibited XPO1 with IC 50 values around or below 10 µM. Among the phenyl series, the α-COOCH 3 derivative 30 was almost inactive against XPO1, while the extended ethers 33 and 34 were active, particularly the 3,3-dimethylallyl derivative 34, with an IC 50 value of 1.59 µM.
Once XPO1 inhibition was confirmed, an additional experiment was performed for compounds 9 and 10 employing a Jurkat leukemia cell line where the cysteine residue at position 528 of XPO1 had been replaced by a serine residue (Jurkat XPO1 C528S ) [19,43]. The effect of the mutation on the activity of the inhibitors was assessed by the visualization of the subcellular localization of the XPO1 cargo RanBP1. As shown in Figure 4A (control experiment), in both wild-type and mutant Jurkat cells, RanBP1 is localized in the cytoplasm as a consequence of the correct nuclear export mediated by XPO1. When both cell lines were treated with KPT-330 at 1 µM (used as a positive control, Figure 4B), in the wild-type cell line, RanBP1 was accumulated in the nucleus, since KPT-330 blocked its binding to XPO1, whereas in the mutant cell line, KPT-330 was unable to interact with XPO1; therefore, RanBP1 remained in the cytoplasm. Similarly, when both cell lines were treated with our prototype compounds 9 and 10 at 4 µM ( Figure 4C, D, respectively), RanBP1 accumulated in the nucleus in the case of the wild-type cells, whereas its nuclear export was not inhibited in the mutant cell line.  The IC50 values of compounds 9 and 10 regarding their XPO1 inhibitory activity in WT and C528S Jurkat cell lines are shown in Table 3. Both compounds 9 and 10 inhibited nuclear export of RanBP1 cargo protein with IC50 values of 2.2 and 0.3 µM, respectively, in the WT Jurkat cell line. These results were in agreement with the IC50 values obtained for XPO1 inhibition in the HeLa reporter cell line (Table 2), where the chalcone 10 was the most potent of the two. What is most relevant is that both compounds lost their inhibitory activity against the Jurkat XPO1 C528S cell line, similar to the positive control KPT-330. Thus, these experiments evidenced the importance of Cys528 for the XPO1 inhibitory activity of these chalcones. Table 3. IC50 values of 9 and 10 in means of XPO1 inhibition in wild-type and XPO1 C528S mutant Jurkat cell lines.

IC50 (µM) a Jurkat XPO1 WT Jurkat XPO1 C528S KPT-330
0.07 ± 0.01 >5 9 2.2 ± 0.03 >5 The IC 50 values of compounds 9 and 10 regarding their XPO1 inhibitory activity in WT and C528S Jurkat cell lines are shown in Table 3. Both compounds 9 and 10 inhibited nuclear export of RanBP1 cargo protein with IC 50 values of 2.2 and 0.3 µM, respectively, in the WT Jurkat cell line. These results were in agreement with the IC 50 values obtained for XPO1 inhibition in the HeLa reporter cell line (Table 2), where the chalcone 10 was the most potent of the two. What is most relevant is that both compounds lost their inhibitory activity against the Jurkat XPO1 C528S cell line, similar to the positive control KPT-330. Thus, these experiments evidenced the importance of Cys528 for the XPO1 inhibitory activity of these chalcones. Thus, compounds 9 and 10, able to react with GSH in incubation studies, were among the most potent XPO1 inhibitors and also showed good antiproliferative activity. On the other hand, the α-CH 3 chalcone 18, which poorly reacted with GSH in the incubation assays, was unable to inhibit XPO1 and also did not show antiproliferative activity. The most puzzling data comes from the esters 13 and 30, which were almost inactive against XPO1 and in antiproliferation assays. However, in the incubation studies with GSH, compound 13 seems to be as reactive as the α-H chalcone 9. Certainly, many factors can be involved in the lack of activity of this chalcone. A potential explanation may arise by taking into account that the XPO1 inhibition assays are performed in a cell culture, so that the presence of esterases might convert the ester into the corresponding carboxylic acid. If this is the case, the resulting α-COOH chalcone would be very poorly electrophilic based on the reactivity described for α-X-2 3,4,4 -tetramethylchalcones [27], and thus, its capacity to behave as a Michael acceptor against XPO1 would be seriously compromised.

Docking Studies
Docking molecular studies have been carried out for compounds 9 and 10 with CovDock [44,45], a computational tool developed by Schrödinger to perform covalent docking, using the coordinates of the S. cerevisiae XPO1 protein in its complex with KPT-8602 [20]. Using this tool, a covalent bond is created between the SH of Cys539 (Cys528 in human XPO1) in the NES-binding cleft and the β-position of the double bond of the chalcone.
The best-docked solution of compound 9 ( Figure 5A) shows that the compound nicely fits within the upper part of the NES-binding cleft, so that the 2-naphthyl (ring A) has favorable interactions with the hydrophobic residues Ile555, Leu536, Phe583 and Val559, while the pyrimidin-5-yl ring (ring B) is buried in an inner region delimited by Ala552. As for compound 10 ( Figure 5B), the best-docked solution indicates that the 3,5dimethoxyphenyl ring lays in a hydrophobic cavity surrounded by the residues Ile555, Leu536 and Phe583, whereas ring B is directed to the upper part of the cleft. In addition, this binding mode is compatible with a hydrogen bond interaction between the carbonyl group of the ligand and the side chain of Lys579. This stabilizing interaction might help explain the higher inhibitory activity of compound 10 compared to 9 against XPO1. Additionally, this binding mode also suggests that one of the methoxy groups in ring A can be replaced by other longer ethers to gain additional interactions in the lower part of the cleft. Indeed, such ethers (compounds 33 and 34) also showed XPO1 inhibition, but did not improve the inhibitory value of compound 10.
this binding mode is compatible with a hydrogen bond interaction between the carbonyl group of the ligand and the side chain of Lys579. This stabilizing interaction might help explain the higher inhibitory activity of compound 10 compared to 9 against XPO1. Additionally, this binding mode also suggests that one of the methoxy groups in ring A can be replaced by other longer ethers to gain additional interactions in the lower part of the cleft. Indeed, such ethers (compounds 33 and 34) also showed XPO1 inhibition, but did not improve the inhibitory value of compound 10.  Compounds were also analyzed by HPLC/MS with a e2695 LC(Waters, Milford, Massachusetts, USA), coupled to a Waters 2996 photodiode array detector and a Waters Micromass ZQ. The column used is a Waters SunFire C18 2.1 × 50 mm, 3.5 µm, and the mobile phases were A: acetonitrile and B: H 2 O, together with a constant 5% of C (H 2 O with 2% formic acid) to assure 0.1% of formic acid along the run.
Separations on silica gel were performed by preparative centrifugal circular thin-layer chromatography (CCTLC) on a Chromatotron R (Kiesegel 60 PF 254 gipshaltig (Merck)), with a layer thickness of 1 and 2 mm and a flow rate of 4 or 8 mL/min, respectively.
General procedure for the reaction of aromatic ethanones with aldehydes under basic conditions (General procedure A) [46].
To a solution of Ba(OH) 2 ·8H 2 O (1.0-1.2 mmol) in water (0.2 mL), the corresponding aldehyde (1.0-2.0 mmol) in methanol (1 mL) was added. To the resultant mixture, the appropriate aromatic ketone (1.0-1.2 mmol) in methanol (8 mL) was added dropwise over 10 min and the reaction was stirred at room temperature for 2-16 h. The workup and purification procedures are described individually.
General procedure for the reaction of aromatic ketones with pyrimidine-5-carboxaldehyde under acid conditions (General procedure B) [47].
General procedure for the alkylation of phenol groups (General procedure C). To a solution of the corresponding phenol (1.0 mmol) in anhydrous DMF (6 mL), Cs 2 CO 3 (1.2-1.5 mmol) was added. After stirring at rt for 10 min, the appropriate alkyl halide (1.2 mmol) in anhydrous DMF (2 mL) was added dropwise. The resulting mixture was heated at 80 • C for 0.5-5 h and then quenched with water (5 mL). Volatiles were removed and the residue was diluted with ethyl acetate (20 mL) and washed with a saturated solution of NH 4 Cl (10 mL). The organic layer was dried over Na 2 SO 4 , filtered and evaporated to dryness. The residue was purified by CCTLC in the Chromatotron.

GSH Reactivity Assay
A 500 µM sample of the tested compound was incubated with 5 mM reduced Lglutathione for 24 h at 37 • C with a final volume of 200 µL. As a solvent system, 20 mM PBS buffer pH 7.4 with 2 mM EDTA:DMSO (1:1) was used [50]. To perform the assay, stock solutions of 20 mM of the chalcones (9, 10, 13 and 18) and 15 mM of reduced L-glutathione were freshly prepared for every experiment and then diluted properly to give the final electrophile:nucleophile ratio (1:10). A 5 mM solution of 5,5 -dithiobis(2-nitrobenzoic acid) (DTNB) was prepared in the same solvent system to quench the reaction. After different time points (10 min, 1 h, 4 h and 24 h), an aliquot of 20 µL of the incubation mixture was quenched by adding 20 µL of the DTNB stock solution, mixed thoroughly, and then analyzed by HPLC. HPLC analysis was performed in Agilent 1120 compact LC, column ACE 5 C18-300 (15 cm × 4.6 mm); UV detection was performed at λ = 254 nm. Solvents used were acetonitrile for bottle A and H 2 O (containing 0.05% TFA) for bottle B, and the flow rate was 1mL/min. The gradients used were: (A) incubations with chalcone 9: from 10% to 80% of solvent A in 10 min; (B) incubations with chalcones 10 and 18: from 10% to 100% of solvent A in 10 min. The new peaks were analysed by LC-MS. As controls, a 500 µM solution of the α,β-unsaturated compound was also analysed, and a 5 mM reduced L-glutathione solution was quenched by adding DNTB at the same time points. In general, after 6 h of incubation, no GSH was detected.

Cell Culture and Reference Compounds
Cancer cell lines HCT-116, NCI-H460, HL-60, K-562 and Z-138 were acquired from the American Type Culture Collection (ATCC, Manassas, VA, USA). The DND-41 cell line was purchased from the Deutsche Sammlung von Mikroorganismen und Zellkulturen (DSMZ Leibniz-Institut, Brunswick, Germany), and the HAP-1 cell line was ordered from Horizon Discovery (Horizon Discovery Group, Water Beach, UK). All cell lines were cultured as recommended by the suppliers. Culture media were purchased from Gibco Life Technologies and supplemented with 10% fetal bovine serum (HyClone, GE Healthcare Life Sciences, Chicago, Illinois, USA).
Stably transfected HeLa NLS SV40 -AcGFP-NES PKI were cultured as described in Vercruysse et al. [19] CRISPR/Cas9 genome editing of the Jurkat cell line was performed as in Neggers et al. [43] to generate a XPO1 C528S mutant cell line.
Reference inhibitor KPT-330 was purchased from Selleckchem, and stock solutions were prepared in DMSO.

Cell Proliferation Assays
Adherent cell lines HCT-116, NCI-H460 and Hap-1 cells were seeded at a density between 500 and 1500 cells per well in 384-well tissue culture plates (Greiner). After overnight incubation, cells were treated with different concentrations of the test compounds. Suspension cell lines HL-60, K-562, Z-138 and DND-41 were seeded at densities ranging from 2500 to 5500 cells per well in 384-well culture plates containing the test compounds at the same concentration points. The plates were incubated and monitored at 37 • C for 72 h in an IncuCyte (Essen BioScience Inc., Sartorius; Göttingen, Germany for real-time imaging of cell proliferation. Brightfield images were taken every 3 h, with one field imaged per well under 10× magnification. Cell growth was then quantified based on the percent cellular confluence, as analysed by the IncuCyte image analysis software, and used to calculate IC 50 values by logarithmic interpolation. Compounds were tested in two independent experiments and represented as mean ± SEM.

XPO1 Phenotypic Reporter Assay
To study the XPO1-mediated nuclear export, stably transfected HeLa NLS SV40 -AcGFP-NES PKI reporter cells were seeded at 8000 cells per well in 96-well all clear tissue culture plates (TPP). After overnight incubation, cells were treated with different doses of compound or solvent (DMSO) for 2 h and then fixed and counterstained with DAPI. Fluorescence was read on an ArrayScan XTI High Content Reader (Thermo Fisher Scientific, Waltham, MA, USA). Nuclear and cytoplasmic compartments were segmented and their average pixel intensities in the green channel were quantitated employing the HCS Studio software. Genedata Screener software was used for dose-response curve fitting, and calculation of EC 50 values was based on the percentage of cells having a predominant nuclear localisation (ratio of nuclear to cytoplasmic signal equal or above 1.4) of the reporter construct. Compounds were tested in two independent experiments and represented as mean ± SEM.

Immunofluorescence Staining of RanBP1
RanBP1 immunofluorescence staining was performed on both wild type and mutant XPO1 C528S Jurkat cells treated with compound or solvent (DMSO) for 3 h. Cells were harvested at 400× g, washed in PBS and then transferred into an 8-well µ-Slide (Ibidi) pretreated with 0.1% (w/v) poly-L-lysine (Sigma). Cells were allowed to adhere to the slides and then subsequently fixed (4% PFA in PBS), washed and permeabilized (0.2% Triton X-100 in PBS). Further immunofluorescence staining was then performed according to standard procedures. Employed antibodies were rabbit anti-RanBP1 (ab97659, Abcam, Cambridge, UK) at a 1:500 dilution and secondary Alexa Fluor 488 goat anti-rabbit antibody (A11008, Invitrogen, ThermoFisher Scientific). Cell nuclei were counterstained with DAPI, and the samples were imaged by confocal microscopy on a Leica TCS SP5 confocal microscope (Leica Microsystems, Weitzlar, Germany), employing a HCX PL APO 63× (NA 1.2) water immersion objective. Subsequently, fluorescence was read on an ArrayScan XTI High Content Reader (Thermo Fisher Scientific, Waltham, MA, USA). Nuclear and cytoplasmic compartments were segmented, and their average pixel intensities in the green channel were quantitated similarly as for the XPO1 phenotypic reporter assay.

Computational Methods
The crystal structure of XPO1 complex with KPT-8602 was retrieved from the Protein Data Bank [51] (pdb id: 5JLJ [20]).
The Schrödinger Suite v2018-3 has been used for all the computational studies [52]. The 3D structure of all compounds used in the modelling studies were generated using the graphical interface Maestro, and these were then optimized using the tool Macromodel. For the docking studies, all ligands were prepared with LigPrep in Maestro, and the receptor protein was prepared with the Protein Preparation Wizard.
Covalent docking was performed with CovDock [44]. The KPT-8602-XPO1 X-Ray structure was used for all covalent docking studies. Cys539 was selected as the reactive residue and Michael addition was selected as the reaction type. A grid box of 10 Å was defined. CovDock was used with the default parameters published in reference [44], except for the number of final poses per ligand, which was set up to 10. MMGBSA analysis was also chosen. Results were visually inspected and analysed using the computer graphics program PyMOL [53].

Conclusions
Most of the described XPO1 inhibitors are α,β-unsaturated carbonyl compounds able to react with Cys528 at the NES-binding cleft of XPO1. Based on these examples, we synthesized two series of chalcones with a six-membered N-heterocycle as ring B and tested their XPO1 inhibitory activity. Most of the synthesized compounds inhibited XPO1 function in a reporter cell line, and this inhibition nicely correlated with their antiproliferative activity in cell culture assays, with compounds 9, 10, 24 and 34 as the most potent. Moreover, in a mutant Jurkat cell line where the Cys528 of XPO1 had been mutated to a Ser (Jurkat XPO1 C528S ), the capacity of the prototype compounds 9 and 10 to inhibit XPO1 mediated nuclear export of the cargo protein was abolished, indicating the importance of Cys at position 528 for the inhibitory activity of our compounds, as also demonstrated for KPT-330 [43]. Finally, the interaction of the chalcones 9 and 10 with the NES-binding cleft has been analyzed through covalent docking with CovDock. Thus, these chalcones may represent an alternative scaffold in the search for XPO1 inhibitors.