8u, a pro-apoptosis/cell cycle arrest compound, suppresses invasion and metastasis through HSP90α downregulating and PI3K/Akt inactivation in hepatocellular carcinoma cells

8u, an acridine derivative, has been proved effective anti-hepatocarcinoma effect, while the underlying mechanism remains unclear. Here, metabolomics and proteomics approaches were applied to study its anti-cancer mechanism and explore its effect on HepG2 cells’ invasion and metastasis abilities. The results showed that 8u significantly suppressed HepG2 cells migration and enhanced cell-to-cell junctions. The inhibition effect of 8u on invasion and metastasis disappeared after HSP90α gene silencing, and was reversed after HSP90α overexpression. The biological experimental results indicated that 8u also blocked PI3K/Akt pathway, thereby reducing fatty acid synthase (FASN) protein expression and disordering intracellular lipid metabolism to inhibit cell invasion and metastasis. In addition, HSP90α protein and PI3K/Akt pathway could co-adjust to each other. These findings demonstrated that 8u could efficiently suppress the invasion and metastasis of HepG2 cells by decreasing the expression of HSP90α protein and inhibiting the PI3K/Akt signaling pathway, which could be used as a potential candidate for the treatment of HCC.

biochemical systems, study pathophysiological processes, map mechanisms of action and discover targets of novel drugs [14][15][16][17][18] . The cancer metabolome, as the complete set of small-molecule chemicals found within a biological sample, could be an important source for the discovery of molecular targets and mechanisms of action 19,20 . As a major carrier and functional executor of cellular activities, proteins own more biological information as compared to metabolites 21 . Over the last two decades, "proteomics" has emerged as a fascinating tool to probe the biological perturbations occurring and contribute more important insights into the action mechanisms of drug by a global analysis of protein alterations upon drug treatments 22,23 . Combining of multiple omic techniques is an emerging approach, which aims to help identify latent biological relationships 24 . Recently, integrated metabolomic and proteomic technologies have already been applied in the antitumor mechanism researches 25 .
As part of the effort in the discovery of potent anti-hepatocarcinoma agents, our laboratory has developed several series of novel compounds with obvious antitumor activity. Among them, a multi-substituted benzyl acridone derivative 8u had good activity against human liver carcinoma HepG2 cells and showed low toxicity in vitro. Moreover, in vivo preliminary experiments showed that 8u might be a good lead compound in the treatment of HCC 26 . The results showed that 8u might have an anti-proliferative effect against human cancer cells through the induction of apoptosis. However, its possible molecular mechanism needs to be further improved, and its potential effect on cancer cell invasion and migration has never been observed before. In the current investigation, metabolomics and proteomics approaches were used to characterize alterations at the biochemical and molecular levels in control and 8u treated HepG2 cells. Furthermore, a series of bioassays were employed to in-depth discover the antitumor mechanism of 8u on HepG2 cells. This research revealed that acridone derivative 8u had the potential to develop into a new antitumor drug for HCC.

Results and Discussion
Differential metabolites identified and metabolism pathway analysis by LC/MS. Before the metabolomics experiment, the most optimal drug concentration was determined by MTT experiments. As shown in the Fig. 1A, the antiproliferative activity towards HepG2 cells increased significantly with increasing concentration of 8u. When the concentration increased to 2 μM, the inhibitor rate was approximately 20%. In this case, drug-induced changes could be observed, and sufficient cells could be ensured for subsequent detections. Thus, this concentration was appropriate for metabolomics study.
As Fig. 1B and C shown, 8u had a significant impact on metabolites in HepG2 cells. By differential analysis, 44 significantly changed metabolites were identified in positive and negative ion modes after 8u treated (Table S1). All of the altered metabolites were analyzed using KEGG database to identify the pathways involved. After searching against the database, the most abundant five pathways were presented in Fig. 1D. It could be easily observed that the differentially expressed metabolites were mainly involved in lipids metabolism (glycerophospholipid metabolism, linoleic acid metabolism, biosynthesis of fatty acids) and energy metabolism disorders (riboflavin metabolism and purine metabolism). 8u induced HepG2 cells apoptosis and cycle arrest. Lipids provide cancer cells with membrane structural units, signal molecules, protein post-translational modifications and energy supply to support rapid cell proliferation 27 . Hence, lipid metabolism disorders lead to the changes of membrane compositions, protein distribution and function, gene expression, and cellular functions, and then cause cell apoptosis 28 .
To determine the primary driver for the decreased cell viability, we analyzed the apoptosis of 8u treated HepG2 cells with flow cytometry. We observed a significant dose-dependent increase in the population of apoptotic cells in HepG2 cell lines when treated with 8u ( Fig. S1). Apoptosis cell populations in HepG2 cells treated with 10 μM 8u for 48 h, were increased to 34.6% and 55.9%, respectively, compared to DMSO-treated control cells (0.039% and 0.405%). These results suggested that 8u could induce apoptosis to inhibit the proliferation of HepG2 cells.
Furthermore, we observed cell cycle arrest by 8u in HepG2 cells. Propidium iodide (PI) staining for HepG2 cell cycle progression showed that the majority of the cells were arrested in the S phase (Fig. S2A,B) after 8u treatment. Then, we examined the levels of several cell cycle regulators during the S transition. As the western blotting results shown, 8u treatment elevated levels of the CDK inhibitors p18, p21 and p27, decreased levels of CDK2, 4, 6 and cyclin A, B1, D3 in HepG2 cells compared with those in control cells (Fig. S2C,D). Taken together, our data demonstrated that 8u induced cell cycle arrest at S phase through upregulation of CDK inhibitors and downregulation of cyclin-dependent kinases (Cdks). 8u reduced migration and invasive capabilities of HepG2 cells. Lipid metabolism also plays a crucial role in cancer metastasis. During the progress of metastasis, lipid metabolism will change to facilitate the transfer of tumor cells 29 . Boyden chamber assay was employed to assess whether 8u could regulate HepG2 cell invasion. As shown in Fig. 2A and B, 8u significantly reduced the number of invasion cells stained with trypan blue compared to the control group. The result indicated that 8u could inhibit migration of HepG2 cells.
Western blotting assay was further employed to explore the anti-metastasis mechanism of 8u. Enhanced cell-to-cell junctions can reduce the invasive and metastatic abilities of cancer cells 30 . Hence, the expressions of cell-to-cell junctions related significant factors β-catenin, CLDN-1, E-cadherin, N-cadherin, Vimentin, Slug and Snail were detected. E-cadherin and Vimentin are essential proteins for homotypic adhesion junctions, which are the major markers of epithelial and mesenchymal cells, respectively 31,32 . Epithelial properties promote the adhesion junctions, and inhibit the invasion and metastasis of the tumor 33 . Western blotting results showed that the expression of E-cadherin was significantly up-regulated by 8u. On the contrary, expression of Vimentin was remarkably down-regulated in the 8u treated cells compared to control groups (Fig. 2C,D). The results showed that 8u could enhance epithelial properties of HepG2 cells.
E-cadherin and β-catenin can form complexes to promote cell adhesion 34 . When adherens junctions are disrupted, β-catenin could release from the complex and transfer into the nucleus, while transactively activating the SCieNTifiC RepoRts | (2018) 8:309 | DOI:10.1038/s41598-017-18701-3 expression of β-catenin gene 35 . N-cadherin is another type of cell adhesion molecule opposite to E-cadherin. It can stimulate migration and invasion of cancer cells 36 . Transcription factor Snail and Slug (snail 2) could suppress the expression of E-cadherin gene, further reduce cell-cell connections and enhance tumor cell invasiveness 33,37 . 8u also enhanced cell adhesion junctions by blocking the expressions of β-catenin and N-cadherin proteins as well as Snail and Slug transcription factors (Fig. 2C,D).
The claudin-1 (CLDN-1) is a key constituent of tight junctions (TJs), which is another mainly component for cell-to-cell adhesiveness 30 . Overexpression of CLDN-1 could promote invasion and metastasis of HCC cells through regulating the expression of N-cadherin, E-cadherin, β-catenin, Vimentin proteins and transcription factor Slug 38 . As the Fig. 2C,D shown, 8u repressed the expression of CLDN-1 protein, enhanced tight junctions and cell adhesion junctions.
For validating the regulation effect of 8u on invasive and metastasis, immunofluorescence analysis was performed using E-cadherin, β-catenin and vimentin antibodies (Fig. S3). The results were consistent with the western blotting experiments. Taken together, these results indicated that 8u reduced migration and invasion capabilities of HepG2 cells via enhancing the cell-cell contacts and making their epithelial properties more pronounced. 8u induced protein expression changes and inhibited the expression of HSP90α in HepG2 cells. In order to comprehend the inhibition mechanism of 8u on invasion and metastasis, a proteomics approach was further performed to study the changes in proteins after 8u treated. A total of 137 differentially expressed proteins were discovered in the 8u treated group (Table S2). These proteins were functionally analyzed using the online DAVID, and were classified into three major categories (cellular component, molecular function, and biological process) according to GO analysis (Fig. 3).
These results validated our previous metabolic speculation, and were consistent with the results of transwell and western blotting experiments. To further identify the underlying protein complexes, we employed the STRING tool to generate a protein-protein interaction network and predict interactions between these proteins (Fig. 3D). The results showed that the HSP proteins family was the major linker protein. From proteomics data, 8u down-regulated the expression of HSP90α protein in HepG2 cells (Table S2). 8u could inhibit the expression of HSP90α protein and directly bind to it. Depending on the results of proteomics, the changes of HSP90α protein expression were examined after 8u treated on HepG2 cells. 8u could significantly reduce the expression of HSP90α protein in whole cells and cell membrane (Fig. 4A,B). To validated the results, immunofluorescence experiments were employed. Figure 4C visually showed that 8u inhibited HSP90α protein expression in HepG2 cells in a dose-dependent manner.
With the aim of observing whether 8u directly acted on HSP90α protein, the direct binding of 8u to HSP90α protein was assessed. First, molecular docking studies of the compound 8u with HSP90α model (PDB ID: 3TUH) were conducted using SYBYL-X v1.3 program. Ganetespib, an inhibitor of HSP90α, was as a control to analyze the binding sites of 8u and HSP90α protein. As seen from Fig. S4, four hydrogen bonds were formed between ganetespib and amino acid residues of HSP90α protein, two hydrogen bonds formed between the oxygen atom of carbonyl group of Ganetespib and the side chain of Asp93 and Lys58; another two formed between two nitrogen atom of triazole ring and the side chain of Gly97 and Thr184. It is notable that these amino acid residues (such as Gly97, Thr184, Asp93 and Lys58) constituted the key binding pocket for the HSP90α protein 39 . Similarly, 8u could also be well combined with HSP90α protein and formed two hydrogen bonds in the same binding pocket (Fig. 5A,B), one formed between the oxygen atom of nitro group and -COOH group of Gly97 (O···H-O); another formed between the oxygen atom of nitro group and -COOH group of Thr184 (O···H-O). Therefore, the docking results may suggest that the binding pocket of 8u and HSP90α protein is the same as Ganetespib.
To verify that 8u indeed binds to HSP90α, we performed a fluorogenic titration assay. Figure 5C showed that the fluorescence of HSP90α dramatically decreased in the presence of 8u. To confirm the binding affinity of 8u with HSP90α quantitatively, the classical Sterne-Volmer Equation (1) was utilized to calculate the binding constant 40 (Fig. 5D).   41 . This illustrated that the quenching effect of 8u on HSP90α was due to the static quenching caused by the formation of complexes. These analyses suggested that 8u might bind with HSP90α to contribute or partly contribute to its ability to inhibit tumor invasion and metastasis.

8u inhibited migration and invason by regulating the expression of HSP90α. Early research had
shown that HSP90α was closely related to tumor invasion and metastasis 42 . Secretory HSP90α could promote tumor cell invasion 43 . However, the regulation of intracellular HSP90α on invasion and metastasis is unclear. Transwell invasion assay were used to observe the migration ability of HepG2 cells after HSP90α protein silencing. The invasive ability of HepG2 cells gradually weakened, with the increase of 8u dose. Under action of 1 μM 8u, the numbers of HepG2 cells were significantly reduced. However, after the silencing of HSP90α, this phenomenon disappeared, even if the dose increased to 5 μM, 8u could not prevent the invasion and metastasis of HepG2 cells (Fig. 6A,B). Furthermore, the expressions of invasion and metastasis-related proteins in HepG2 cells were detected after silencing of HSP90α protein. In order to obtain a better effect of 8u, and a shorter acting time, we chose 3, 6, 9 μM 8u and 24 h for the subsequent Western blotting assays according to the IC 50 value (Table S3). The results showed that the expression of E-cadherin significantly up-regulated, on the contrary, the expressions of β-catenin, CLDN-1, N-cadherin, Vimentin, Slug and Snail remarkably down-regulated in the silencing HSP90α group compared to negative group (Fig. S5). However, the regulation of 8u on E-cadherin and Vimtenin was weakened or even disappeared after silencing HSP90α (Fig. 6C,D). Therefore, the invasion of HepG2 cells in the transwell experiment was not significantly affected.
For purpose of confirming that 8u inhibited HepG2 cells invasion and metastasis by regulating HSP90α protein, an HSP90α expression plasmid was transfected into HepG2 cells. After 12 h, cells were divided into two parts, which one was used for transwell experiments and the other for western blotting experiments. HSP90α overexpression resulted in higher numbers of migrated cells than controls and attenuated the migration inhibition of 8u on HepG2 cell (Fig. 6E,F). For western blotting assay, cell lysates were collected and analyzed using HSP90α, E-cadherin and Vimtenin antibodies to examine the effects of 8u on cell-to-cell junctions. As the Fig. 6G and H shown, HSP90α overexpression could decrease the expression level of E-cadherin and prmote Vimtenin. In additional, HSP90α overexpression could reverse the inhibitory effect of 8u on Vimtenin. And the change of Vimtenin protein is positive correlated with HSP90α (Fig. S6). These were consistent with transwell experimental results. But E-cadherin was further inhibited after HSP90α overexpression. This may be related to the apoptosis induced by 8u, while apoptosis could cleave and shed E-cadherin 44 . These results indicated that 8u inhibited the invasion and metastasis of HepG2 cells by governing the expression of HSP90α protein.
8u could inhibit the activation of PI3K/Akt pathway. To gain insight into how 8u might affect invasion and metastasis of HepG2 cells, integrated pathway analysis (IPA) was performed using Metaboanalyst. This analysis performed integrated metabolic pathway analysis on results obtained from the combined metabolomics and proteins expression. Our results showed that 8u mainly caused changes in biosynthesis of lipid, such as unsaturated fatty acids and fatty acid biosynthesis, and glycerophosphoipid metabolism (Fig. 7A), which were commonly found to enhance in HCC 45 .
FASN is a critical polypeptide enzyme for producing saturated fatty acids 46 . It can further influence the lipid metabolism 47 . When inhibiting metastasis, intracellular FASN protein expression is significantly decreased 48 . Thus, the effect of 8u on expression of FASN protein was examined. As the Fig. 7B and C shown, 8u significantly inhibited the expression of FASN protein in HepG2 cells. PI3K/Akt pathway is a key signal pathway to regulate the expression of FASN protein and cancer invasion and metastasis 49,50 . The phosphorylation of Akt represents the activation of the PI3K/Akt pathway. In this study, the inhibitory effect of 8u on Akt phosphorylation was analyzed using western blotting analysis. As shown in Fig. 7D and E, 6-9 μM 8u decreased the levels of phosphorylated Akt, while the total Akt protein levels remained constant. Next, LY294002, a PI3K inhibitor, was used to determine whether 8u inhibited invasion and metastasis were associated with the PI3K/Akt pathway. As seen from Fig. 7F and G, 6 μM 8u significantly decreased Akt phosphorylation. Changes of FASN protein after LY294002 treatment were also examined. The result showed that the expression of FASN protein also reduced after inhibiting the activity of the PI3K/Akt signaling pathway (Fig. 7H,I). These results suggest that 8u can inhibit the activation of the PI3K/Akt signal pathway and further inhibit the expression of FASN protein.

Activity of PI3K/Akt pathway interacts with the expression of HSP90α protein. In order to
in-depth understand anti-metastasis mechanisms of 8u, the link between HSP90α protein and PI3K/Akt signaling pathway were explored. First, the expressions of p-Akt and Akt in HepG2 cells were examined after silencing HSP90α protein. As the Fig. 8A and B shown, phosphorylation of Akt protein in the HSP90α silencing group significantly reduced compared to the control group, whereas Akt content was not affected. We also found that overexpression HSP90α could upregulate the p-Akt expression level (Fig. S7). Next, expression changes of HSP90α after inhibiting the activity of the PI3K/Akt signaling pathway by LY294002. As shown in Fig. 8C and D, the expression of HSP90α protein significantly decreased after inhibiting of PI3K/Akt signal pathway. These results indicated that the activity of the PI3K/Akt pathway and the expression of HSP90α protein could co-adjust the cancer invasion and metastasis after 8u treated.

Conclusion
In summary, the novel acridone derivative 8u could inhibit the invasion and metastasis of HepG2 cells by enhancing cell-cell adhesion junctions. Mechanistically, we confirm that 8u could directly bind to HSP90α protein, and inhibit the migration and invasion by suppressing HSP90α. Furthermore, 8u also blocked PI3K/Akt pathway, thereby reducing FASN protein expression and disordering intracellular lipid metabolism to inhibit cell invasion and metastasis. Notably, suppression of PI3K/Akt pathway could down-regulate the expression level of HSP90α, and down-regulate of HSP90α could in turn inhibit the activity of PI3K/Akt pathway. The anti-metastasis mechanism of 8u on HCC was showed in Fig. 9. Taken together, this study demonstrates that 8u could inhibit invasion The densitometry performed on the western blotting of C. All the western blotting data presented were means ± SD of three independent experiments, and the significant difference was set at p < 0.05. *p < 0.05, **p < 0.01 compared with the control group.

Methods
Cell culture and MTT assay. Detailed cell culture and MTT assay approaches were described in supporting information.
Metabolomics analysis. HepG2 cells were exposed to 2 μM 8u, and were harvested by cold methanol/water in a ratio of 4:1 (v/v). The samples were analyzed by UPLC/Q-TOF MS. The chromatographic separation was carried out on a Waters AcquityTM BEH C 18 column (100 mm × 2.1 mm, 1.7 μm). Raw UPLC-QTOF/MS ESI data were processed using the MarkerLynx software. Databases of HMDB (http://www.hmdb.ca/), Lipid Maps Database (http://www.lipidmaps.org) and METLIN (http://metlin.scripps.edu/) were used to identify the metabolite markers with tandem mass spectrometry. Some standards of metabolic interest were also used to confirm their structures.
Label-free quantitative proteomics analysis. Cell culture and dosing conditions were the same as metabolomics methods. The proteins of cells were extracted and quantified. Then equal proteins (approximately 30 µg) were desalted and digested. The peptides mixture were desalted using C 18 spin columns (Product # 89870, Thermo Scientific) and the flow-through was dried in a SpeedVac centrifugal evaporator. The dried peptides were dissolved in water for 2D LC-MS/MS analysis. A Waters system (Waters, Milford, MA) of 2D nano-Acquity UPLC coupled with Q-TOF premier tandem mass spectrometer was used for proteomics studies. Raw data from 2D nano-UPLC/Q-TOF were analyzed by the ProteinLynx Global Server (PLGS 2.5) with ExpressionE informatics. A Swiss-Prot database (release 51.0, October 2013) was used for database searches.
Cell invasion assay. Cell invasion assay was performed as described 51,52 . Invasion inserts with 8μm pore membranes from Corning (New York, USA) were coated with fibronectin from Sigma-Aldrich (Missouri, USA) as described 53 . After treatment, cells were seeded on the inserts to reach confluence, and then culture for 24 hours with drug treatment. After fixed, the invading cells were stained with trypan blue, and were counted by Image-Pro-Plus 6.0 of Media Cybernetics (MD, USA).
Western blotting analysis. Cell proteins were extracted by RIPA lysis as described 54 . Protein concentrations were measured by the Bradford protein assay. Equal amounts of protein were separated by 12% SDS polyacrylamide gel electrophoresis followed by transferring to PVDF membranes, and were subsequently analyzed with different antibodies. Specific-appropriate secondary antibodies were detected and measured by the Luminescence Image Analyzer Tanon 5200 (Shanghai, China). The density of the bands were measured by Image Quant software (Molecular Dynamics, Sunnyvale, CA, USA).
Immunofluorescence staining analysis. 8u treated cells were fixed with 4% paraformaldehyde for 25 min and permeabilized with 0.1% Triton X-100 for 10 min. Blocked with 2% bovine serum albumin (BSA) for 30 min at 37 °C and followed by the primary antibody against E-cadherin, β-catenin, Vimentin and HSP90α at 4 °C overnight. The cells were subsequently incubated with the corresponding Alexa 488-conjugated secondary antibody for 1 h at room temperature. The nuclei were stained with DAPI for 3 min. The images were captured using a DMI-4000B inverted fluorescence microscopy (Leica).

Molecular bocking.
Molecular modeling of small molecule compound 8u was performed with the molecular modeling package SYBYL-X 1.3 (Tripos associate Inc., St. Louis, MO, USA) according to the reported process 55 .
HSP90α protein binding assay. The emission spectra were carried out on Fluorolog spectrometer. 3.6 µg purified HSP90α protein (StressMarq Biosciences Inc., Canada) was incubated in 2 ml buffer solution (100 mM Tris/HCl pH 7.5, 20 mM KCl, 6 mM MgCl 2 , 0.0005% BSA). The final concentration of the HSP90α was 20 nM. And then 8u8u was also added with the final concentration from 0 to 200 nM. The excitation wavelength was set at 480 nm. The ambient temperature of this experiment was maintained at room temperature. The incubation time before testing was 3 min.
HSP90α overexpression. HSP90α plasmid was constructed by genecreate (Wu Han, China). After ligation, the amplicon was transfected into puncture bacteria competent cells followed by plasmid extraction and transient transfection of HSP90α into HepG2 cells using polyplus as manual. After 12 h transfection, cells were replated and cultured for subsequent experiments.