2, 3-Dihydro-3β-methoxy Withaferin-A Lacks Anti-Metastasis Potency: Bioinformatics and Experimental Evidences

Withaferin-A is a withanolide, predominantly present in Ashwagandha (Withania somnifera). It has been shown to possess anticancer activity in a variety of human cancer cells in vitro and in vivo. Molecular mechanism of such cytotoxicity has not yet been completely understood. Withaferin-A and Withanone were earlier shown to activate p53 tumor suppressor and oxidative stress pathways in cancer cells. 2,3-dihydro-3β-methoxy analogue of Withaferin-A (3βmWi-A) was shown to lack cytotoxicity and well tolerated at higher concentrations. It, on the other hand, protected normal cells against oxidative, chemical and UV stresses through induction of anti-stress and pro-survival signaling. We, in the present study, investigated the effect of Wi-A and 3βmWi-A on cell migration and metastasis signaling. Whereas Wi-A binds to vimentin and heterogeneous nuclear ribonucleoprotein K (hnRNP-K) with high efficacy and downregulates its effector proteins, MMPs and VEGF, involved in cancer cell metastasis, 3βmWi-A was ineffective. Consistently, Wi-A, and not 3βmWi-A, caused reduction in cytoskeleton proteins (Vimentin, N-Cadherin) and active protease (u-PA) that are essential for three key steps of cancer cell metastasis (EMT, increase in cell migration and invasion).

vimentin phosphorylation 12 . More recently, some studies showed that the treatment with Wi-A containing root extract inhibited metastases and EMT process 13,14 . We had earlier demonstrated that Wi-A and Withanone inhibit cancer cell migration and invasion by downregulating heterogeneous nuclear ribonucleoprotein-K (hnRNP-K), VEGF and MMPs proteins 15 . 2,3-dihydro-3β-methoxy withaferin-A (3βmWi-A) is a natural and structurally close Wi-A analogue, having substitution of β-methoxy group at position 3 of Wi-A ergostane ring 16 . Structural analogue of natural compounds, having chemical modification i.e. methylation, acetylation or hydroxylation at key moieties has been shown to affect their bioactivity, underlining structure-function association 17 . More precisely, alkali analogues of organic steroid compounds have been shown to attain improved potency 18,19 . Markedly, 10-methoxy derivative of Camptothecin, was shown to induce anti-angiogenic response 20 . Moreover, dimethoxy substitution to Brartemicin, a trehalose-based metastasis inhibitor, was shown to improve potency in murine colon carcinoma model 21 . Similarly, cis-resveratrol methylated analogs were shown to augment anti-tumor potency in metastatic mouse melanoma cells 22 . 2-methoxyestradiol, a natural estrogen analogue was characterized to be a potent anti-angiogenic compound 23 .
In light of these findings, we had earlier characterized the activities of Wi-A and its methoxy analogue 3βmWi-A. Unlike Wi-A, 3βmWi-A was found to be safe and well-tolerated at even higher concentration in vitro 16 . On contrary to Wi-A, 3βmWi-A was found to be inert and lacked cytotoxicity, and shown to promote survival of normal human cells exposed to various stress conditions 24 . Consistently and the fact that some natural compounds at nontoxic doses have been shown to possess anti-metastasis activity 25,26 , we investigated the effect of 3βmWi-A on cancer cell migration using bioinformatics and molecular assays in cultured cells. We report that whereas Wi-A binds to vimentin and hnRNP-K and modulate VEGF and MMPs signaling, 3βmWi-A lacked these activities.

Results
Wi-A, but not 3βmWi-A, inhibited cancer cell migration and invasion. In order to investigate the effect of Wi-A and 3βmWi-A on cancer cell migration, we used their sub-toxic doses (0.3 and 0.6 μM) 16,27 . As shown in Fig. 1a, analyses of cell phenotype and viability showed no significant effect on U2OS (osteosarcoma) cells. In wound-healing assay, we found that whereas Wi-A inhibited U2OS cell migration at even the low dose (0.3 μM), 3βmWi-A did not cause any difference with respect to the control at both doses used (Fig. 1b). Quantitation of these data (Fig. 1b) showed dose-dependent and significant delay in migration of cells when treated with Wi-A, but not 3βmWi-A. To confirm that the anti-migration effect of Wi-A was not due to its cytotoxicity, time-lapse observations in control, Wi-A and 3βmWi-A (0.3 μM) treated GFP-labeled U2OS cells revealed that Wi-A, but not 3βmWi-A, delayed U2OS cell migration (data not shown); while no cell death was observed. We also confirmed the effect on three other metastatic cancer cell lines (MDA-MB-231, MCF-7 and HT-1080). As shown in Fig. 1c (photomicrographs), whereas sub-toxic doses of Wi-A caused significant delay in migration of all three cell types, 3βmWi-A remained inert. Quantitation of the data (Fig. 1c) showed Wi-A, not 3βmWi-A possesses significant anti-migratory activity. We also performed quantitative Matrigel invasion analyses that revealed a potent anti-invasion effect of Wi-A, but not of 3βmWi-A, both at 0.3 and 0.6 μM concentrations (Fig. 1d). Furthermore, a dose dependent effect (30 to 50% decrease) on invasiveness (both 0.3 and 0.6 μM) was observed in Wi-A treated U2OS cells. Conversely, 3βmWi-A did not exhibit anti-invasion activity. Therefore, these results suggested that Wi-A, and not 3βmWi-A, inhibited cancer cell migration and invasion (Fig. 1).

Molecular docking of Wi-A and 3βmWi-A with Vimentin.
In light of the information that Wi-A caused vimentin aggregation resulting in abrogation of invasion and metastases of the breast cancer cells 10-12 , we examined the interactions of these two molecules with vimentin by bioinformatics and molecular docking (MD) tools. The computationally modelled structure of vimentin dimer and tetramer was retrieved and vimentin monomer coordinates were extracted from dimer structure. MD simulation of vimentin modelled structure from residue 79-406 (monomer, dimer and tetramer) was carried out for 50 ns. The average representative structure of Vimentin was selected from stable simulation trajectories for each case. The RMSD of each structure was also analyzed to treat it as a control (Fig. S1a). The RMSD comparison of Vimentin monomer with its dimeric and tetrameric forms showed that Vimentin monomer is highly unstable whereas dimer and tetrameric forms have comparable and stable RMSD fluctuations. The interaction of Wi-A and 3βmWi-A was examined near Cys328 residue of Vimentin using molecular docking and simulation tools. Cys328 residue plays important role in reorganization of Vimentin network in response to oxidants as well as Vimentin polymerization by interacting with Zinc 28 . Interaction of any compounds at this critical residue may affect polymerization and assembly of Vimentin. Binding affinity and interactions formed by Wi-A and 3βmWi-A at the binding site of Vimentin monomer, dimer and tetramer are listed in Table 1. MM-GBSA was also used to find ligand binding affinity in docked complexes before and after MD simulations (Table 1).
Binding affinity of ligands reported by MM-GBSA were significant, hence, we proceeded with MD simulation of these complexes to investigate stability of interactions predicted at the binding site. Both molecules were found to be stable at the binding site of all forms of Vimentin, interacting with almost same residues but with slightly different orientations (Fig. 2a,b and Fig. S2a-d). Wi-A and 3βmWi-A did not show any interaction with Cys328, however, both interacted with its neighboring residues and could interfere with interaction of other entities with Cys328, thereby affecting polymerization and reorganization of Vimentin filaments. Next, we examined the effect of Wi-A and 3βmWi-A on stability of protein structure. RMSD computation of Vimentin monomer (Control) and in complex with Wi-A and 3βmWi-A showed that ligand binding around Cys328 increased the stability of Vimentin monomer structure, which was comparable to its dimeric and tetrameric forms (Fig. S1b). Of note, we found that Wi-A added more stability to Vimentin monomer (RMSD fluctuations in range of 20-30 Å) as compared to 3βmWi-A (that is 30-40Å). In line with this, the MM-GBSA calculated binding affinity of Wi-A (−49.61 Kcal/mol), was stronger than 3βmWi-A (−35.44 Kcal/mol). We also found that Wi-A formed both hydrophobic and hydrogen bond interactions, while 3βmWi-A interacted solely by hydrophobic interactions and did not form any hydrogen bonds ( Fig. S2a,b). For Vimentin dimer and tetramer, RMSD fluctuations of Vimentin (control), Vimentin-Wi-A complex and Vimentin-3βmWi-A complex were comparable (Fig. S1c). These data suggested that Wi-A and 3βmWi-A may not affect stability of Vimentin dimer and tetramer. In case of Vimentin dimer, 3βmWi-A showed stronger binding efficiency by laterally interacting with both chains of dimer (binding energy of −61 Kcal/mol). On the other hand, Wi-A showed longitudinal interactions with chain A only (binding energy of −37.59 Kcal/mol) (Fig. S2c,d). In addition, both Wi-A and 3βmWi-A formed hydrophobic interaction with Gln322, a conserved residue found in the highly conserved domain protein family of intermediate filament (Fig. S3) 29 . Similarly, in case of Vimentin tetramer both Wi-A and 3βmWi-A showed binding at the same site forming interactions with almost same residues with slightly different binding affinity of −93.51 and −78.48 Kcal/mol, respectively (Fig. 2a,b). Both Wi-A and 3βmWi-A interacted with conserved residue Gln322. However, we found that Wi-A caused local disruption of helical structure at the binding site while no such effect was observed in case of 3βmWi-A. Such configurations are predicted to affect interaction of Vimentin with metal ions (Zn) for formation of Vimentin octamer 28 . These data suggested that both Wi-A and 3βmWi-A can interact with Vimentin monomeric, dimeric and tetrameric forms. Wi-A, however, showed high binding affinity for and also caused disruption of secondary structure at the binding site of vimentin tetramer. Hence, it was predicted that Wi-A, but not 3βmWi-A, targets Vimentin causing anti-migration and anti-metastatic effects.
We next performed molecular analyses to confirm the above results. As shown in Fig. 3a, we found significant decrease in Vimentin, a key intermediate filament protein involved in cell migration, in Wi-A, but not 3βmWi-A, treated cells. We, next, analyzed the localization and assembly of Vimentin in control and treated cells. Whereas Wi-A (0.6 μM) treated cells showed distinct Vimentin aggregation (Fig. 3b), there was no difference in response to 3βmWi-A treatment. The results were confirmed with high resolution confocal microscopy ( Fig. 3b). Dose dependent effect of Wi-A and 3βmWi-A, and analysis of Vimentin expression by immunostaining further revealed its significant reduction in Wi-A treated cells, while its levels remain unchanged in response to 3βmWi-A treatment (Fig. 3b,c). Based on these data and Figs. 1 and 2, it was confirmed that whereas Wi-A targeted Vimentin protein, 3βmWi-A lacks such activity. The data was consistent to an earlier report that showed Vimentin as a target of Wi-A 10 . In order to get further insights to the differential bio-activities of Wi-A and 3βmWi-A, we examined their effect on the expression of N-cadherin and E-cadherin, two key mesenchymal (2019) 9:17344 | https://doi.org/10.1038/s41598-019-53568-6 www.nature.com/scientificreports www.nature.com/scientificreports/ and epithelial cell markers. As shown in Fig. 3d, Wi-A, but not 3βmWi-A, caused decrease in N-cadherin in dose-dependent manner (Fig. 3d). In contrast, there was no change in the level of E-cadherin as validated by immunoblotting (Fig. 3e) and immunostaining (Fig. 3f), respectively. These data suggested that Wi-A has potent anti-migratory activities, while 3βmWi-A lacks such effect.

Interactions of Wi-A and 3βmWi-A with the KH3 domain of hnRNP-K and functional consequences.
In light of the above findings and our earlier report that Wi-A targets hnRNP-K protein 15 , we next examined the effect of 3βmWi-A on hnRNP-K activity. Molecular docking and simulation analyses using KH3 domain of hnRNP-K that has been shown to bind to ssDNA revealed strong binding strength of Wi-A (docking score Wi-A −4.54 Kcal/mol) as compared to 3βmWi-A (docking score 3βmWi-A −3.17 kcal/mol) (Fig. 4a,b,e). Although the two withanolides shared similar configuration and surrounding residues, analysis of their superimposed interactions with hnRNP-K 3D-structure demonstrated that β-methoxy group of 3βmWi-A at C3 position caused steric hindrance and hence decreased its binding strength as compared to Wi-A (Fig. 4c,d). As shown in Fig. 4a,b,e, the superimposed images on the binding of Wi-A and 3βmWi-A with the ssDNA-binding interface revealed weaker docking capability of 3βmWi-A as compared to Wi-A. Following these results, we next examined the expression of hnRNP-K in control and treated cells. As shown in Fig. 5a, immunoblotting revealed dose-dependent decrease in its protein levels in Wi-A, but not 3βmWi-A, treated cells. Furthermore, immunostaining analysis also revealed a decrease in hnRNP-K expression levels in Wi-A treated cells (Fig. 5b). Of note, no change in hnRNP-K transcript levels were observed in Wi-A or 3βmWi-A treated cells (Fig. 5c, below).

Wi-A, but not 3βmWi-A, inhibits hnRNP-K-VEGF signaling.
In light of the fact that hnRNP-K is a key regulator of VEGF and promotes its mRNA stability and translation 30,31 , we next prompted to examine VEGF level in Wi-A and 3βmWi-A treated cells. As expected, Wi-A, but not 3βmWi-A, treated cells showed decrease in VEGF in immunostaining analyses (Fig. 6a). It was further confirmed by immunoblotting analyses that endorsed decrease in the VEGF protein in Wi-A treated cells only (Fig. 6b). Of note, 3βmWi-A treated cells, showed higher level of VEGF protein (Fig. 6a,b). VEGF transcript, on the other hand, showed no difference either in Wi-A or 3βmWi-A treated cells (Fig. 6c) suggesting that Wi-A may essentially affects VEGF at the protein level. Consistent to these results, we next examined the level of secreted VEGF in treated cells by multiplex immunoarray and found that whereas Wi-A caused decrease in secreted VEGF levels, it was increased by more than two-fold in 3βmWi-A treated cells (Fig. 7a). Interestingly, IL-10, IL-12, IL-13, key interleukin factors regulating VEGF, levels also showed decrease in response to Wi-A treatment and increase in 3βmWi-A treatment (Fig. 7a). In order to examine the protein targets of Wi-A relevant to cancer cell invasion and metastases effects, we next performed MMP Antibody array. Untreated U2OS cells were used as negative control. As shown in Fig. 7b, Wi-A treated cells showed remarkable decrease in MMP1, MMP2, MMP8, MMP9, MMP10 and MMP13 expression (Fig. 7b). On the other hand, 3βmWi-A treated cells exhibited a slight and insignificant change in their levels, when normalized with untreated control (Fig. 7b). Expression analyses of VEGF downstream effectors viz. u-PA 32 and MMP2 33 that play vital role in cancer cell migration, also exhibited their reduced levels in Wi-A, but not 3βmWi-A treated cells (Fig. 7c). The data suggested that Wi-A, consistent to the downregulation of hnRNPK-VEGF signaling, suppressed expression of several key MMPs/proteases involved in cell invasion and metastases, wherein 3βmWi-A remained largely ineffective. Also, we examined VEGF downstream targets phospho-p38MAPK and hsp27, two key VEGF markers promoting cancer cell survival during metastases/angiogenesis. As shown in Fig. 7d, a dose-dependent decrease in phospho-p38MAPK and hsp27 proteins was observed in Wi-A, but not in 3βmWi-A treated cells. www.nature.com/scientificreports www.nature.com/scientificreports/

Discussion
Wi-A and 3βmWi-A are natural co-occurring withanolides, isolated from Ashwagandha. 3βmWi-A, with the substitution of a β-methoxy group at Wi-A parental ring is suggested to be a natural Wi-A analogue. Wi-A is potent bioactive compound, and has been shown to inhibit metastases and EMT process in vitro and in vivo models essentially by promoting Vimentin aggregation [12][13][14][15][16] and regulating hnRNP-K function 17 . With evident anti-metastatic potency of Wi-A, and the concept that the alkali analogues retain augmented bioactivities 18,19 ; we investigated the activities of 3βmWi-A on cancer cell migration at cellular and molecular levels. Taking the sub-toxic concentrations that showed no effect on cell viability, we found Wi-A, but not 3βmWi-A, inhibited cancer cell migration and invasion. By molecular docking and dynamic simulation studies, Wi-A was predicted to possess greater binding affinity to Vimentin and tetramer than 3βmWi-A. It also caused local disruption of helical structure at Vimentin tetramer binding site. Consistently, Wi-A, but not 3βmWi-A, led to Vimentin aggregation. Wi-A treated cells showed decrease in Vimentin, N-cadherin and E-cadherin proteins, signifying compromised cell migration and invasion 34 . Wi-A was earlier shown to target hnRNP-K, a key regulator of cell migration and angiogenesis 15,30 . Inputs from in silico analysis confirmed a stronger binding of Wi-A, over 3βmWi-A, at the hnRNPK's KH3 domain i.e. ssDNA binding domain. It could lead to instability and decrease in hnRNP-K and its downstream effectors including VEGF and MMPs leading to compromised cell migration 15,32,35,36 . Reduced levels of MMPs, uPA proteases and VEGF targets viz. phospho-p38MAPK 37 , heat shock protein 27 (HSP27) that promote cancer cell migration [37][38][39] , were indeed found in Wi-A, but not 3βmWi-A, treated cells. Taken together, we demonstrate that 3βmWi-A lacked cytotoxicity to cancer cells and showed no effect on cell migration, invasion and angiogenic characteristics. Wi-A was found to affect the Vimentin assembly, yielding its decrease in treated cells; 3βmWi-A was ineffective. Wi-A, as compared to 3βmWi-A, also showed stronger binding to hnRNP-K. Additionally, it caused decrease in other key regulators of cell migration, invasion and metastasis, viz, MMPs and VEGF, u-PA, Hsp27 and p38MAPK. Taken together, we found that in contrast to Wi-A and the anticipation that alkali analogues may exhibit stronger potency, 3βmWi-A lacked anti-metastatic activity. Hence, structural-function predictions of natural compounds warrant careful experimental validation before their recruitment in disease preventive or therapeutic avenues.  www.nature.com/scientificreports www.nature.com/scientificreports/ Immunofluorescence staining. Cells were cultured as described above. 4 × 10 4 cells /well were seeded on glass coverslips placed in a 12-well plate for 24 h. Cells were treated with indicated doses of either Wi-A or 3βmWi-A for 24 h and then fixed with pre-chilled methanol on ice for 10 min. PBS-Triton-X-100 (0.2%) for 10 min were used to permeablized the fixed cells followed by blocking with 2% BSA (10 min). The coverslips were probed with antibodies against VEGF, hnRNP-K, Vimentin, N-Cadherin and E-Cadherin at room temperature (1 h) or 4 °C (overnight). Cells were then probed with Alexa Fluor-conjugated secondary antibodies (Molecular Probes, Invitrogen), and counterstained with Hoechst 33258 (Thermo Fisher Scientific). The stained cells were viewed under Zeiss Axioplan 2 microscope and images were captured using a Zeiss AxioCam HRc camera. Acquired images were quantified using ImageJ software, while fluorescence intensity values were normalized with respective controls and represented in percentage values. To examine the distribution and aggregation of Vimentin in Wi-A or 3βmWi-A treated cells, images were acquired under confocal laser scanning microscope (ZEISS LSM 700) and analyzed by IMARIS ® software (Bitplane, Switzerland).

Methods
Wound healing assay. Motility of cells was examined using the Wound-healing assay. 2.5 × 10 5 cells/ well were seeded in 6-well plate and allowed to attach to the substratum for next 24 h. Monolayers of cells were wounded on the following day by uniformly scratching the surface with a scrapper tip (20 gauge), followed by washing with PBS and then fresh medium was added. Cells were allowed to proliferate and migrate into the wound for, at least, 24 or 48 h. Movement of the control and treated cells in the scratched area were serially monitored under a phase contrast microscope with a 10 X phase objective (Nikon). Time-lapse live imaging was performed using a microscope (Carl Zeiss MicroImaging, ApoTome) equipped with a Plan-Apochromat 10x/0.45 DICII objective, motorized scanning table and a stage incubator at 37 °C with CO 2 . Images were captured with an AxioCam MR Rev 3 Monochromatic Digital Camera using the AxioVision rel 4.8 software for microscope control and data acquisition. Time-point captured images were processed with image J software adjusting contrast and color threshold options for representation. Percentage covered area was calculated using % area option and www.nature.com/scientificreports www.nature.com/scientificreports/ crosschecked with analyze particle options using normalized processed images (6-10 randomly captured). The assay was repeated in triplicate in parallel to the Cell Viability Assay.
Matrigel cell invasion assay. 4 × 10 3 cells were seeded into the upper chamber, coated on the surface with 1/10 dilution of Matrigel (BD BioSciences, Franklin Lakes, NJ), and allowed to migrate to the lower chamber as described in details earlier 41 . Migrated cells were fixed and stained with crystal violet and counted under phase contrast microscopy.

Reverse-transcription PCR (RT-PCR).
RNA was extracted from control, Wi-A or 3βmWi-A treated cells with Qiagen RNeasy kit. cDNA was synthesized from 2 μg of RNA using the ThermoScript Reverse Transcriptase (ThermoFisher Scientific) following the manufacturer's protocol. cDNA was subjected to PCR amplification along with gene specific sense and antisense primer set (as follows) using TaKaRa Ex Taq ® DNA polymerase, steps consisting of an initial 10 min denaturation step at 95 °C followed by 34 cycles at 95 °C for    DNA binding interface of KH3 domain was used to generate grid, so that Wi-A and 3βmWi-A could be screened for their docking potentials at this site 43 . Prepared protein and ligands were then docked using XP docking protocol of Glide module. During docking, hydrogens of aromatic rings were also allowed to form hydrogen bonds. ii. Molecular interaction study of Wi-A and 3βmWi-A with monomer, dimer and tetrameric forms of Vimentin. Since X-ray crystallography structure of full Vimentin is not available so far, we used computational model of Vimentin dimer and tetramer structure generated by Qin et al., 2009 44 . In order to investigate the effect of Wi-A and 3βmWi-A, first MD simulation of structure of Vimentin monomer, dimer and tetramer was carried out for 50 ns each. Simulated and stabilized Vimentin monomer, dimer and tetramer molecules were docked with Wi-A and 3βmWi-A around Cys328 residue using Glide extra precision (XP) algorithm.
The stability of inhibitors at the binding site was examined by performing molecular dynamic (MD) simulations of docked complexes. Vimentin monomer-ligand complexes and Vimentin dimer-ligand complexes were simulated for 50 ns each, while Vimentin tetramer-ligand complexes were simulated for 40 ns each. Prime MM-GBSA module of Schrodinger was used to calculate binding energy of protein-ligand complex 42 . Ligplus program was used to study protein-ligand interactions 45 . Root mean square deviation (RMSD), hydrogen bonds analysis and conformational changes over the simulation trajectories of protein-ligand complexes were monitored using VMD version 1.9.4 46 . Images for publication were generated using Pymol molecular graphics system 47 .
Human MMP Antibody Array. To examine expression levels of Matrix metalloproteinase family proteins, Human MMP Antibody Array -membrane (#ab134004) were used to determine the total secreted MMP protein quantity in condition media of U2OS cells treated with Wi-A and 3βmWi-A (0.6 μM) following the manufacturer's protocol.
Multiplex immunoassay. Bio-Plex Suspension Array System (Bio-Rad, Hercules, CA) was used for multiplex immunoassay profiling. Conditioned medium was collected from control and Wi-A or its methoxy analogue i.e. 3βmWi-A (0.6 μM) treated cells. Concentrations of the cytokines and growth factors in the medium were