Upregulation of BCL-2 by acridone derivative through gene promoter i-motif for alleviating liver damage of NAFLD/NASH

Abstract Nonalcoholic fatty liver disease (NAFLD)/nonalcoholic steatohepatitis (NASH) are global epidemic public health problems with pathogenesis incompletely understood. Hepatocyte excessive apoptosis is a significant symbol for NAFLD/NASH patients, and therefore anti-apoptosis therapy could be used for NAFLD/NASH treatment. Up-regulation of BCL-2 has been found to be closely related with anti-apoptosis. BCL-2 gene promoter region has a C-rich sequence, which can form i-motif structure and play important role in regulating gene transcription. In this study, after extensive screening and evaluation, we found that acridone derivative A22 could up-regulate BCL-2 transcription and translation in vitro and in cells through selective binding to and stabilizing BCL-2 gene promoter i-motif. Our further experiments showed that A22 could reduce hepatocyte apoptosis in NAFLD/NASH model possibly through up-regulating BCL-2 expression. A22 could reduce inflammation, endoplasmic reticulum stress and cirrhosis in high-fat diet-fed mice liver model. Our findings provide a potentially new approach of anti-apoptosis for NAFLD/NASH treatment, and A22 could be further developed as a lead compound for NAFLD/NASH therapy. Our present study first demonstrated that gene promoter i-motif could be targeted for gene up-regulation for extended treatment of other important diseases besides cancer.


INTRODUCTION
Nonalcoholic fatty liver disease (NAFLD) is a main course of liver disease including a wide spectrum of disorders ranged from steatosis to steatohepatitis. Around 30% of patients with NAFLD have nonalcoholic steatohepatitis (NASH) (1), which is featured by liver injury, inflammation and fibrosis (2,3). NAFLD is also an important component of metabolic syndrome, which has enormous risk of running into cirrhosis or hepatocellular carcinoma. However, specific pathological mechanism of NASH is incompletely understood (4), and there is no specific pharmacological targeted drugs for NASH treatment (5,6). Lifestyle interventions with diet and exercise to achieve weight loss goal remain the cornerstone of treatment for NAFLD and NASH (7). Thus, effective targeted therapy is urgently required.
Hepatocyte apoptosis has been proved to be a generally prominent feature of NAFLD patients (8). BCL-2 family modulates the checkpoint of apoptosis and has two major category members: apoptosis members like Bax, and anti-apoptosis members like BCL-2 (9). The anti-apoptotic BCL-2 expression has been proved to be diminished in both hepatocytes and serum in accordance with stage of NAFLD/NASH (10)(11)(12). Therefore, we hypothesized that up-regulation of BCL-2 expression could play important anti-hepatocyte apoptosis role in the course of NAFLD. It has been shown that stabilization of BCL-2 gene promoter imotif could up-regulate gene transcription and translation (13,14), and other gene promoter i-motifs have also been shown to play crucial roles in regulating gene transcription and translation (15). i-Motif structure is formed by C-C base pairs (C+:C) stacking force (16), and is stabilized in faintly acidic (17) or molecular crowding condition in vitro (18). It has been shown that i-motif can form in telomeres (19)(20)(21)(22), promoter regions of proto-oncogenes (23), centromeres (24,25) by using biophysical techniques. Recently, it has been further demonstrated that i-motif structures can form in promoters and telomeric regions by using i-motif specific antibodies (19). We have also revealed that an acridone derivative could selectively bind to oncogene C-MYC promoter i-motif and down-regulate gene transcription and translation (26).
In this study, after extensive screening of our small molecule library, an acridone derivative A22 was found to interact with the C-rich sequence in P1 promoter of BCL-2, which significantly up-regulated BCL-2 gene transcription and translation. A22 could bind to and stabilize the i-motif structure formed by the C-rich sequence, and then possibly recruit hnRNP LL protein to form a complex promoting transcription and translation. Besides, we evaluated therapeutic effect of A22 on the pathological apoptotic characteristics of NASH in mice model of NAFLD/NASH generated by prolonged feeding with a high-fat diet. The i-motif binding ligand A22 showed significant therapeutic effect for NAFLD model by reducing hepatocyte apoptosis, affecting expressions of caspase family members (cleaved-caspase 3&9), hepatic injury (hepatocyte ballooning, serous levels of ALT and AST), hepatic inflammatory response (inflammatory cytokines and lymphocyte mononuclear macrophage numbers) possibly through up-regulating BCL-2.
As we know, this is the first effort to target gene promoter i-motif with a small molecule for NAFLD/NASH treatment. It should be mentioned that gene promoter G-rich sequence could form G-quadruplex, which has been widely studied as potential target for cancer treatment, possibly because its stabilization by small molecule normally downregulate gene transcription. In comparison, stabilization of gene promoter i-motifs by small molecules have been found to either down-regulate or up-regulate gene transcriptions, possibly due to their different binding affinity to transcription factors. Therefore, up-regulation of certain gene transcription through its promoter i-motif could possibly be developed for treatment of other important diseases with reasonably good efficiency and selectivity.

General materials
Chemically synthesized DNA oligomers of HPLC purified grade were purchased from Sangon Biotec, as shown in Supplementary Table S1. All oligonucleotide concentrations were determined by measuring absorbance at 260 nm using a Nano Drop1000 spectrophotometer (Thermo Scientific). For obtaining i-motif structures, C-rich oligonucleotides were annealed in 1× BPES buffer (30 mM KH 2 PO 4 , 30 mM K 2 HPO 4 , 1 mM EDTA, 100 mM KCl) with different pH at 95 • C for 5 min, and then cooled to room temperature. For obtaining G-quadruplex, oligonucleotides were annealed in 20 mM Tris-HCl buffer containing 100 mM KCl (pH 7.4) by heating at 95 • C for 5 min followed with gradual cooling to room temperature. Formation of secondary non-B DNA structures were determined by using circular dichroism (CD) spectrophotometer. BCL-2 and other oncogene promoter i-motif structures had a positive peak at 287 nm and a negative peak around 260 nm. BCL-2 promoter G-quadruplex had a positive peak at 263 nm and a negative peak around 240 nm. Cells were purchased from American Type Culture Collection (ATCC, China). All cell culture reagents including DCFH-DA (D-6883), BSA-conjugated palmitic acid (PA, CID: 985), and BSA-conjugated oleic acid (OA, CID: 445639), were purchased from Sigma-Aldrich. Antibodies were purchased from Affinity Company. Annexin V-FITC/PI apoptosis kit (70-AP101-100) was purchased from Muti-sciences. Highfat diet (HF, 60% of calories from fat) was purchased from Research Diet (Catalog number D12492, New Brunswick, NJ, USA) with detailed information listed in supplementary information. Synthetic route for acridone derivative A22 and A22-HCl was shown in Supplementary Scheme I.

Surface plasmon resonance (SPR)
Surface plasmon resonance (SPR) measurements were conducted by using a GLH sensor chip on a ProteOn XPR-36 Protein Interaction Array system (Bio-Rad Laboratories, Hercules, CA, USA). 5 -Biotin labeled DNAs were annealed in different buffers, and then immobilized (∼12 000 RU) in flow cells. In screening experiment, all ligands were diluted to different concentrations with running buffers (imotif running buffer: 20 mM MES, 100 mM KCl, pH 5.5; G-quadruplex running buffer: 20 mM Tris-HCl, 100 mM KCl, pH 7.4). Ligands of different concentrations were injected at flow rate of 25 l/min for 240 s of association phase, and 300 s of disassociation phase at 25 • C. The GLH sensor chip was regenerated with short injection of 50 mM NaOH solution between consecutive measurements. The final graphs were obtained by subtracting blank sensorgram of corresponding i-motif or G-quadruplex. Kinetic and equilibrium measurements were analyzed with Pro-teOn manager software. Langmuir model was used to fit SPR kinetic sensorgrams. K D values were obtained through equation

Nuclear magnetic resonance
Nuclear magnetic resonance ( 1 H NMR) experiments were carried out at 298 K by using a Bruker Avance 600 NMR spectrometer equipped with a 5 mm TCI Micro CyroProbe. The 3-9-19 WATERGATE gradient spin echo was used to suppress water solvent signal. The number of scans in the experiments was 256 by double scan. All spectra were collected using a spectral width of 18 000 Hz (30 ppm), with an alternate acquisition time of 4 s and a relaxation delay of 6 s. The DNA samples were annealed to 500 M concentration with or without different equivalent of A22 in 1× BPES buffer of different pH before experiments.

Circular dichroic (CD) spectroscopy and CD-melting experiments
CD measurements were performed on a Chirascan circular dichroism spectrophotometer (Chirascan, England). A 10 mm path length quartz cuvette was used to record the spectra with a wavelength range of 230-350 nm with a 1 nm bandwidth. The oligonucleotides had wild type unmodified DNA sequences, with py39 for BCL2 promoter i-motif, and pu39 for BCL2 promoter G-quadruplex. Other oligonucleotides for oncogene promoter i-motifs were simplified with their corresponding names, such as C-MYC, C-KIT, KRAS and VEGF. CD and CD melting experiments were performed at a fixed i-motif or G-quadruplex concentration (1 M) either with or without a fixed concentration (20 M) of A22 in different buffers. Data collection was performed at intervals of 5 • C over a range of 25-95 • C with a heating rate of 1 • C/min.

Fluorescence titrations
Fluorescence titrations were performed on the HORIBA Jobin Yvon FluoroMAX-4 spectrofluorometer. Py39 dual labeled with 5 -FAM and 3 -TAMRA of 1 M concentration was successively titrated with different equivalent of A22 in buffer of different pH. Then dose-dependent spectra changes at 516 nm for various dual labeled oligomers were recorded when increasing concentration of A22.
Fluorescent intercalator displacement (FID) assays using thiazole orange (TO) was a simple way to screen ligands against DNA secondary structures. For thiazole orange displacement, 1 M annealed DNA was pre-incubated with 1 M thiazole orange for 1 h at 37 • C, and then different concentrations of A22 were added into the solution. After each addition, the reaction was stirred and allowed to equilibrate for at least 5 min, and the fluorescence scan was taken with excitation wavelength at 500-700 nm.

Cell culture and MTT
The following cell lines including cervix cancer cell line Siha, epithelial cancer cell line MCF7, lung adenocarcinoma cell line A549, osteosarcoma cell line U2OS, human hepatocellular carcinoma cell line HepG2, human pancreatic cancer cell PANC-1, human colon cancer cell line HCT116, human glioblastoma cell line U87MG, human bladder cancer cell line T24, human bladder transitional cancer cell line UM-UC-3, rat islet cell tumor cell line INS-1, mouse hippocampal neuron cell line HT22, human embryonic kidney cell line HEK293, human hepatic stellate cell line LX-2, were cultured in Dulbecco's modified Eagle's medium (DMEM) or RPMI 1640, with 10% fetal bovine serum at 37 • C in 5% CO 2 . Cells were seeded in 96-well plates (5000 per well), and incubated with various concentrations of A22 at 37 • C in a humidified atmosphere of 5% CO 2 for 24 h. Then 100 l culture medium containing 2.5 mg/ml methyl thiazolyl tetrazolium (MTT) solution was added to each well for further incubation for 4 h. The cells in each well were treated with dimethyl sulfoxide (DMSO) (200 l for each well) and the optical density (OD) was recorded at 570 nm. All drug doses were parallel tested in triplicate. The cytotoxicity was evaluated based on the percentage of cell survival in a dose-dependent manner comparing to the negative untreated control. Blank group wells just contained culture medium without cells. Cell viability was calculated with the following equation: HepG2 cells were cultured in DMEM with 10% fetal bovine serum and 10 U/ml penicillin/streptomycin (Thermo, Guangzhou, China) at 37 • C in a humidified incubator with 5% CO 2 . Cells were seeded and overnight cultured. After the cell density attained to almost 70%, the cells were incubated with 0.5 mM oleic acid (OA, Sigma, China) or 0.5 mM palmitic acid (PA, Sigma, China) for 24 h for cell models, and then harvested for subsequent analysis. Oil red O staining and TG detection method were described in the Supplementary Methods section.

Reverse transcription and real-time PCR
HepG2 cells were treated with or without A22, and corresponding total RNAs were isolated and purified. Reverse transcription was performed by using cDNA synthesis kit. The relative transcription levels for genes were measured with quantitative real-time PCR using SYBR Green-based method as described in the Supplementary Methods section. The primer sequences information and raw data were shown in Supplementary Tables S5 and S7.

Protein extracts and western blotting
Cells after 24 h treatment with or without A22 were harvested, and the proteins were extracted and quantified by using BCA Protein Quantitation Kit (Thermo Fish). Western blots were performed as described in the Supplementary Methods section.

Animal experiment
Eight weeks aged male C57/B6J mice, with weight of 18-22 g, were bred at the Laboratory Animal Center of Sun Yatsen university (Guangzhou, China) for experimental use. All procedures were approved by the Animal Care and Use Committee of Sun Yat-sen University and conformed to the legal mandates and national guidelines for the care and maintenance of laboratory animals. Animals were housed in specific pathogen free environment with stationary temperature at 22 ± 1 • C on a 12 h dark/light cycle. After 1 week of adaptive feeding, the mice were divided into two groups: chow diet group fed with chow diet (CH, Animal Center proved 70% calories from starch), and high-fat diet group fed with high-fat diet (HF, 60% calories from fat purchased from Research Diet) for 16 weeks. CH and HF groups were randomly divided into subgroups at the beginning of week 10. The HF group had two subgroups treated with different dosage of A22: 10 mg/kg and 40 mg/kg. The CH group was treated with 40 mg/kg A22 as a negative control. A22 was dissolved in normal saline and injected through intraperitoneal injection (i.p.) every other day. The body weight and food intake were daily monitored.

Serum biochemical test
The mice were fasted for 8 h, and then anesthetized with 10 mg/kg xylazine and 80 mg/kg ketamine through i.p. injection. After the mice were completely anesthetized, we withdrew blood, removed eyeball, and then had cervical dislocation. We weighted and took photos of target tissues and organs, and then freeze-clamped or fixed with 4% paraformaldehyde solution. Blood samples were centrifuged at the speed of 3000g for 15 min at 4 • C, and then serums were collected for analysis (AST, AST, ALP, CHO, TG, Glucose, insulin, etc.). to BCL-2 gene promoter i-motif was studied by using SPR with K D value determined to be 3.56 M. Bio-py39 was annealed in MES buffer at pH 5.5 for this experiment. (C) CD melting experiment was recorded at 288 nm for py39 annealed in 1× BPES buffer at pH 5.5 with or without A22, indicating that A22 could stabilize BCL-2 gene promoter i-motif. Py39 had its maximum absorption at 288 nm. (D) CD melting experiment was recorded at 266 nm for pu39 with or without A22, indicating that A22 had no significant effect on BCL-2 promoter G-quadruplex. Pu39 had its maximum absorption at 266 nm, which was annealed in buffer of 20 mM Tris-HCl, 100 mM KCl, pH 7.4. All the experiments were repeated for three times.

Histomorphological examination
Tissues fixed with 4% paraformaldehyde were dehydrated gradually with ethanol (70-100%), and then put on paraffin. The tissues were incised at 4 m thick by a rotary microtome (Thermo, USA) for hematoxylin and eosin (H&E), masson and Sirius red staining. Steatosis, inflammation, ballooning, and fibrosis in liver were assessed with NAFLD activity scoring system. Sirius red staining was carried out to detect collagen, and masson staining was for fibrosis of liver. Monocyte-derived macrophages and lymphocyte were assessed by immunohistochemistry labeled with CD11b antibody and CD45 antibody, respectively. The apoptosis factors BCL-2 and cleaved-caspase 3 were evaluated by using immunohistochemistry with corresponding antibodies.

Statistical analysis
Data were expressed as the mean ± SEM. Data between two groups were analyzed with unpaired Student's t-tests using Graphpad Prism (Graphpad Software Inc, California, USA). A *P value of ≤0.05 is considered statistically significant.

Screening of small molecules for binding to BCL-2 gene promoter i-motif
In order to find a small molecule interacting with BCL-2 gene promoter i-motif, we screened our small molecule library by using SPR. Molecules screened here include acridone derivatives, isaindigotone derivatives, benzo[c]acridine derivatives, bisacridine derivatives, and imidazole derivatives. The K D values for interaction of the i-motif with various small molecules were determined as shown in Supplementary Table S2. Among these compounds, acridone derivative A22 ( Figure 1A) exhibited strong binding affinity to BCL-2 gene promoter i-motif, with its K D value determined to be 3.56 M ( Figure 1B). We then selected some of the outstanding compounds for Thiazole Orange (TO) fluorescent intercalator displacement assay (27) to further evaluate their binding affinity to BCL-2 promoter i-motif formed with C-rich sequence py39. Our results (Supplementary Table S3, Supplementary Figure S2) showed that A22 could significantly replace Thiazole Orange in binding with BCL-2 promoter i-motif, further indicating its strong binding affinity to BCL-2 promoter i-motif compared with other compounds.

Compound A22 could selectively bind to and stabilize BCL-2 promoter i-motif
In order to know if A22 could specifically bind to BCL-2 promoter i-motif, we performed SPR, MST, and ITC experiments to study the binding affinity of A22 to BCL-2 promoter G-quadruplex and other DNA structures. The K D values for binding of A22 to BCL-2 G-quadruplex (estimated to be 48.9 M), duplex DNA (>50 M), VEGF promoter i-motif (36.4 M), C-KIT promoter i-motif (34.5 M) and C-MYC promoter i-motif (11.95 M) were found to be significantly higher than that for binding of A22 to BCL-2 promoter i-motif (3.56 M) in SPR assay, as shown in Supplementary Figure S1. The K D values for binding of A22 to BCL-2 promoter i-motif were determined to be 4.69 and 5.52 M by using MST and ITC (Supplementary Figure S1), respectively. To further understand the interaction of A22 to other gene i-motifs, we used fluorescence titration assay to determine their binding affinities. As shown in Supplementary Figure S1, A22 could induce BCL-2 promoter i-motif formation by decreasing the fluorescence intensity of BCL-2 promoter C-rich oligonucleotide, which was labeled with FAM at 5 -end and TAMRA at 3 -end. It has been known that the fluorescence of FAM and TAMRA double-labeled C-rich oligonucleotide is quenched when folding into i-motif structure. In comparison, the fluorescence intensities of other gene promoter i-motifs were not significantly affected upon addition of A22. These results are consistent with our TO replacement result. The relative replacement ratio of A22 for BCL-2 promoter imotif was 52%, comparing to 10% for BCL-2 promoter G-quadruplex, 12% for hairpin, and 6% for GC-rich duplex structures (Supplementary Figure S2). All these results showed that A22 could specifically bind to BCL-2 promoter i-motif.
In order to further confirm the binding between A22 and BCL-2 promoter i-motif, we carried out CD, 1 H NMR, and ESI-MS experiments to study their interactions. Our CD spectra showed an increase of the i-motif band upon addition of 20 equivalents of A22 at pH 5.5 and pH 6.2 (Supplementary Figure S2). Our 1 H NMR spectra for the imino proton peaks of BCL-2 promoter i-motif are shown in Supplementary Figure S3A. At pH 5.5, the imino proton peaks at 15−16 ppm for hemi-protonated C−C+ base pairs in i-motif structures were increased upon A22 titration at 30, 40, 50, 60 • C, indicating that A22 could induce the formation of BCL-2 promoter i-motif. When the temperature rose to 60 • C, the characteristic peak of BCL-2 promoter imotif disappeared. Upon addition of A22, a peak appeared at 15-16 ppm in imino proton region, suggesting that the compound could induce the formation of i-motif and stabilize the secondary structure at high temperature. Supplementary Figure S3B showed similar results at pH 6.0. A22 could induce the formation of i-motif at 30, 40 and 50 • C. At 50 • C, the free DNA showed no signal in the imino proton region, and the addition of A22 could induce i-motif formation with peak appeared at 15-16 ppm, and stabilize the i-motif at high temperature.
In ESI-MS experiments (Supplementary Figure S3), py39 was incubated with A22 at pH 7.0, and a peak of 12,186 Da for py39-A22 adduct appeared on mass spectrum, indicating that A22 could tightly bind to py39. The similar result was more obviously observed at pH 6.0, as shown in Figure 2A and B. These results strongly supported that A22 could bind to and induce the formation of BCL-2 promoter i-motif and stabilize the structure even under near physiological pH conditions. An ideal i-motif ligand should possess two essential features: high i-motif binding specificity and high i-motif stabilizing ability. In order to know whether A22 could stabilize BCL-2 promoter i-motif, we carried out circular dichroism (CD) melting and FRET melting studies. In the presence of A22, the CD melting temperature of py39 increased for 9.63 • C ( Figure 1C). In contrast, A22 had much weaker effect with ΔTm values ranged from 0 to 4.26 • C for BCL-2 promoter G-quadruplex ( Figure 1D) and other genes (C-KIT, KRAS, VEGF, C-MYC) promoter i-motifs (Supplementary Figure S2). In our FRET melting experiment, A22 could increase the melting temperature of BCL-2 promoter i-motif for 9.15 • C at 4 M concentration. Meanwhile, A22 had much weaker effect on melting temperatures of other gene promoter i-motifs (Supplementary Figure S2). These results showed that A22 could specifically bind to and stabilize BCL-2 promoter i-motif.

A22 could up-regulate BCL-2 gene transcription and translation in a dose-dependent manner in HepG2 cell line
Since A22 could efficiently bind to BCL-2 promoter i-motif, we then studied whether A22 could affect BCL-2 gene transcription and translation in HepG2 cells, which were analyzed in lip-induced and steatotic cell models. A22 was almost non-toxic to most cell lines, with its IC 50 value determined to be more than 100 M for HepG2 cells after 24 h incubation, as shown in Supplementary Table S4. HepG2 cells were treated with increasing concentrations of A22, as shown in Figure 3A. A22 was found to significantly up-regulate BCL-2 transcription in HepG2 cells in a dosedependent manner. In order to know its selectivity, we studied whether A22 could affect transcription and translation of some other well-known genes, since these genes also have i-motif structure on their gene promoters. A22 was found to have no significant effect on transcription of BAX, C-KIT, KRAS, C-MYC and VEGF ( Figure 3B). BCL-2 translation was also up-regulated in HepG2 cells upon treatment with A22, with its protein levels increased in a dose-dependent manner as shown in Figure 3C and D. We also studied the selectivity of A22 for its effect on translation of C-KIT, C-MYC and VEGF. As shown in Figure 3C and D, our results showed that A22 had no significant effect on expressions of these genes.
In order to understand the mechanism for BCL-2 gene up-regulation induced by A22, we carried out some experiments to study the binding interactions between BCL-2 promoter i-motif and transcriptional factor in the presence and absence of A22. It has been reported that protein hnRNP LL as a transcriptional factor can recognize and bind to BCL-2 promoter i-motif forming a transcriptional complex to activate gene transcription (13,14). We studied this interaction in the presence and absence of A22 at extracellular level by using MST and ELISA experiments. Our purified protein hnRNP LL was half-diluted 15 times in PBS buffer (pH 7.4, 0.05% Tween-20) for binding studies with BCL-2 promoter i-motif, and their binding constant K D value was determined to be 26.6 M by using MST. Upon addition of 20 equivalent of A22, their K D value was significantly decreased to 2.32 M, indicating that binding of A22 to BCL-2 promoter i-motif could greatly increase the binding of BCL-2 promoter i-motif with hnRNP LL (Supplementary Figure S4A and B). Our binding study result from ELISA experiment was consistent with our MST experimental result, with their binding affinity increased significantly upon addition of A22, as shown in Supplementary Figure S4C.
The binding interactions between BCL-2 promoter imotif and transcriptional factor hnRNP LL in the presence and absence of A22 was further studied in HepG2 cells. DNA pull-down experiment was performed to investigate their binding interactions. The biotin-labeled DNA was fastened on magnetic beads and incubated with cell ly-sis solution for 1 h in the presence or absence of A22. The amount of hnRNP LL bound with magnetic beads was analyzed by using western Blot. As shown in Supplementary Figure S4D, A22 could significantly increase the binding of cellular hnRNP LL to BCL-2 promoter i-motif with increased efficiency of 84.3% in the presence of 20 equivalent of A22, which showed that binding of A22 to BCL-2 promoter i-motif could promote its binding with hnRNP LL from HepG2 cells.
Chromatin Immunoprecipitation (CHIP) experiment was also carried out to study whether A22 could affect the binding of BCL-2 promoter i-motif with hnRNP LL in HepG2 cells. After 24 h incubation of HepG2 cells with or without A22 at 37 • C cell incubator, protein-DNA complexes of HepG2 cells were crosslinked with formaldehyde. Cells were lysed and analyzed with hnRNP LL specific antibody, and DNAs were purified and quantified with qPCR. Our result showed that BCL-2 promoter i-motif significantly recruited hnRNP LL in the presence of A22 in a dosedependent manner, as shown in Supplementary Figure S4E. These results showed that A22 could promote the binding of BCL-2 promoter i-motif to hnRNP LL. This indicated that A22 might up-regulate BCL-2 gene expression through its binding to BCL-2 P1 promoter i-motif which then recruited transcriptional factor hnRNP LL forming a complex to activate transcription.

A22 showed dose-dependent anti-apoptosis in lipid-induced apoptosis cell model and improved hepatocyte function of glucose and lipid metabolism
Excessive hepatocyte apoptosis has been found to be related with NASH (28), and therefore anti-apoptosis of hepatocyte could block pathological progression of NASH. Since A22 could up-regulate BCL-2 expression that is related with anti-apoptosis, we further studied its anti-apoptosis effect on lipid-induced hepatocyte apoptosis cell model. 0.5 mM palmitic acid oil (PA) induced hepatocyte apoptosis cell model was used. As shown in Figure 4A, after treatment with 0.5 mM palmitic acid oil (PA) for 24 h, cell viability was reduced to 43%, which was used as a lipid-induced apoptosis model. Interestingly, cell viability was improved by 11% (P = 0.029), 24% (P = 0.001), and 35% (P = 0.0004) upon addition of 10, 20 and 50 M A22. We measured mRNA and protein levels of BCL-2 and BAX in apoptotic cell models. As shown in Figure 4B, the relative mRNA level of anti-apoptosis factor BCL-2 decreased in PA model. With A22 addition, BCL-2 mRNA levels increased by 54% (P = 0.022), 128% (P = 0.003), and 190% (P = 0.001) upon addition of 5, 10 and 20 M A22. The pro-apoptotic factor BAX showed an opposite trend to BCL-2. Meanwhile, the protein level of BCL-2 was also significantly improved compared with the PA group ( Figure 4C). FITC Annexin V/PI apoptosis detection method was used for the study on 0.5 mM palmitic acid oil (PA) induced hepatocyte apoptosis cell model without or with increasing concentration of A22 treatment for 24h, and collected cells were analyzed by using flow cytometry. As shown in Supplementary Figure  S5A and C, in contrast to the control cells with an apoptotic ratio of 5%, ratio of apoptotic cells in PA induced group was significantly increased to 24.6%. Upon addition of 20 M A22, the apoptosis ratio was reduced to 16% (P = 0.019), and A22 displayed apparent anti-apoptotic activity in a dose-dependent manner. The mitochondrial membrane potential assay which is an indicator of early apoptosis was also performed by using JC-1 fluorescence probe (Supplementary Figure S5E). The membrane-permeant JC-1 dye is widely used in apoptosis studies to monitor mitochondrial health. Mitochondrial depolarization in early apoptosis is indicated by a decrease in the red/green fluorescence intensity ratio. As the concentration of A22 increased, the red/green fluorescence intensity ratio increased, indicating that A22 could decrease the occurrence of early apoptosis. In addition, we also found that A22 could reduce general oxidative stress in 0.5 mM PA-induced apoptosis model with H2DCFDA probe and protect mitochondria in living cells by the Mitortracker deep red probe ( Supplementary Figure S5G). Then we studied caspases of apoptosis signaling pathways in PA-induced models. It was found that A22 could effectively decrease PA-induced activation of apoptosis related proteins in a dose-dependent manner, with decreased cleaved caspase 3 (C-Caspase 3) and 9 (C-Caspase 9) ( Figure 4C). These results all showed that A22 could alleviate apoptosis and improve cell survival in PA-induced apoptosis model.
We also examined whether A22 could affect glucose and lipid metabolism of HepG2 cells after up-regulating BCL-2 gene expression against apoptosis. As shown in Supplementary Figure S5B and D, A22 could reduce intracellular lipid deposition of 0.5 mM oleic acid (OA) induced droplet formation. The ability of hepatocyte in absorbing glucose in insulin resistant model was also studied by using 2-NBDG (2-(N-(7-nitrobenz-2-oxa-1,3-diazol-4-yl)amino)-2-deoxyglucose). HepG2 cells were induced to IR (insulin resistance) with insulin (10 −6 M) treatment for 24 h, and effect of A22 on cellular glucose intake was measured by using confocal imaging. Our results (Supplementary Figure S5F) showed that A22 could significantly improve hepatocyte capacity of glucose intake, compared with positive drug Metformin (1 mM). In conclusion, our results were well consistent, which all showed that A22 could improve hepatocyte function through anti-apoptosis in lipid-induced cell models.

A22 could ameliorate apoptosis followed with attenuating hepatic injury and steatosis for high-fat diet-fed mice model
In order to evaluate whether A22 has anti-apoptosis activity in vivo, we used NAFLD/NASH mice model induced by using reported methods with high-fat (HF) diet to study potential therapeutic effect of A22 on apoptosis. Formulation of the high fat diet was list as shown in Supplementary Table S6. After 6 weeks intraperitoneal injection of increasing concentrations of A22, samples were then collected and analyzed. Compared with control group, serum levels of alanine aminotransferase (ALT), aspartate aminotransferase (AST), and alkaline phosphatase (ALP) in experimental group were all found to be significantly decreased in a dose-dependent manner, indicating its effect of attenuating hepatic injury ( Figure 5E-G). As expected, A22 relieved HF-induced phenotypes of metabolic syndrome indicated by reductions of body weight gain ( Figure 5A and B), fat mass ( Figure 5J), hyper glycemia ( Figure 5K), hyperinsulinemia ( Figure 5L), and high plasma level of total cholesterol (CHO) and triglyceride (TG) (Figure 5H and I). These effects were not affected by food intake ( Figure 5D).
Interestingly, A22 administration almost corrected all the phenotypes of NAFLD/NASH induced by HF diet, namely histological steatosis and hepatocyte ballooning ( Figure 6A), compared with normal levels of chow (CH) diet-fed mice. Ballooning (blue arrow indicated), hepatic steatosis (black arrow indicated), and fibrosis (blue arrow indicated) scores were obtained according to the NAFLD Activity Score (NAS) System as shown in Figure 6B-D respectively. Relative lipid droplet contents were determined as shown in Figure 6E. A22 also decreased liver weight of HF diet-fed mice ( Figure 5C). Then we examined transcription of BCL-2 gene and expression of apoptosis-related proteins in mice liver samples. For CH fed mice with 40 mg/kg A22 group, their BCL-2 and BAX transcriptions did not have significant difference from CH control group. However, significant changes of BCL-2 and BAX transcriptions occurred for HF diet-induced NAFLD/NASH model, as shown in Figure 7A and B. Their mRNA levels were analyzed by using qRT-PCR. The chronic HF diet-fed mice had increased hepatocellular apoptosis in the liver, as indicated by their increased expressions of cleaved-caspase 3, PARP, BAX, and decreased expression of BCL-2 ( Figure  7C). A22 treatment significantly affected the levels of apoptotic markers and increased BCL-2 expression. The similar results were also obtained for immunohistochemical stain of BCL-2 ( Figure 6F) and cleaved-caspase 3 (Supplemen-   tary Figure S7A and B). Our above experimental results were consistent with the dose-dependent reduction of hepatocyte apoptosis of A22 detected in Tunnel experiment (Figure 6G), which could directly reflect the apoptosis of hepatocytes in liver tissue.
It has been reported that increased hepatocyte apoptosis is a pivotal mechanism contributing to inflammation and fibrogenesis of NASH (29). Hepatocyte apoptosis can be activated by many causes including lipid toxicity (30), unresolved endoplasmic reticulum (ER) stress (31) and oxidative stress (32). ER stress has been reported to be increased during NAFLD/NASH, which can promote lipid synthesis and inflammation (30). Proteins PRKR-like endoplasmic reticulum kinase (PERK) and eukaryotic translation initiation factor 2 alpha (eIF-2␣) can initiate unfolded protein response (UPR) canonical pathway to increase the ER stress, and UPR accumulation can result in activation of apoptosis. Therefore, we examined potential therapeutic effect of A22 on hepatocyte ER stress. Our results ( Figure 7D) showed that A22 reduced expression levels of inositol-requiring enzyme 1␣ (IRE 1␣), phosphorylated PERK, phosphorylated eIF-2␣, and C/EBP homologous protein (CHOP) all in a dose-dependent manner. These indicated that A22 could ameliorate apoptosis partially through its effect on ER stress.
In order to know whether A22 has anti-inflammatory and anti-fibrosis effects, we studied its effects on HF dietfed mice liver. Our results showed that A22 treatment significantly inhibited HF diet-induced macrophage aggregation, which were analyzed with specific antigen CD11b (for monocyte-derived macrophages) ( Supplementary Figure S7A and C) and CD45 (for lymphocyte) (Supplementary Figure S7A and D). These were associated with decreased protein levels of inflammatory cytokines NF-B, TNF-␣, IL-6 of the liver ( Figure 7E). Besides, A22 also showed significant anti-fibrosis effect by reducing protein expressions of ␣-SMA and Collagen I ( Figure 7F), as well as reversing the fibrosis stained with masson and Sirius Red ( Figure 6A and D). The above results indicated that A22 could effectively alleviate not only hepatocyte apoptosis induced by high-fat diet, but also pathological changes including balloon-like degeneration, inflammatory response, fibrosis, and hepatocyte damage.
In order to evaluate whether A22 could affect the ability of liver in metabolizing sugar in vivo, intraperitoneal glucose tolerance test (IPGTT) and intraperitoneal insulin tolerance test (IPITT) were performed on HF diet-fed mice with 4 weeks of A22 treatment. Compared to CH diet-fed mice group, the Area Under Curve parameters (AUC glucose , AUC ITT ) and Home-IR of HF diet-fed mice were apparently increased (P < 0.0001, P < 0.0001), indicating increased glucose tolerance. With addition of A22, the AUC value and insulin resistance index (HOME-IR) decreased in a dose-dependent manner (Supplementary Figure S6), which indicated fast glucose disappearance and therefore improved glucose tolerance. These results showed that A22 could bind to and stabilize BCL-2 promoter i-motif resulting in up-regulation of BCL-2 expression, which could become a promising approach for anti-apoptosis based on our study in NAFLD/NASH mice model in vivo. A22 could become a multi-functional lead compound for further devel-opment for NAFLD/NASH treatment through improving hepatic functions including reducing lipid accumulation, glucose metabolism, anti-inflammation, and inhibition of fibrosis.

DISCUSSION
Thus far, lifestyle intervention and weight loss are main treatments for NAFLD/NASH approved by FDA. Although some agents, such as obeticholic acid, have been proved to alleviate NASH in phase 3 clinical trials (33), no pharmacotherapy is clinically available. It has been shown that activation of caspases (34), BCL-2 family proteins (35), and c-Jun N-terminal kinase (36) induced hepatocyte apoptosis, which play important roles in activation of NAFLD/NASH. Apoptotic hepatocytes can stimulate hepatic stellate cells and immune cells (37)(38)(39)(40) to produce inflammatory factors and cytokines in the progression of fibrosis in liver. Disorder of glucose (41) and lipid metabolism (42) associated with NAFLD/NASH patients can accelerate the progress. The activation of reactive oxygen species (43,44), oxidative stress, and endoplasmic reticulum stress (45) are also involved. The pathogenesis of steatohepatitis contains a complicated reprogrammed molecular network. Therefore, it is desirable to develop new strategy for NASH treatment by targeting a key factor involved in pathogenic pathways to produce multifunctional and synergistic effects. The anti-apoptotic factor BCL-2 seems to be an ideal target, and its inhibition or down-regulation can trigger mitochondrial pathway of apoptosis and initiate caspase cascade reaction. It has been shown that BCL-2 expression is significantly low for NASH patients (10,46), suggesting that the pathological progression of NASH was negatively correlated with BCL-2 expression level.
After extensive screening, we discovered acridone derivative A22 as a promising suppressor of steatohepatitis by stabilizing BCL-2 gene promoter i-motif followed with upregulation of anti-apoptosis factor BCL-2. A22 could significantly improve cellular and animal models in aspects of insulin resistance and lipid deposition after alleviating hepatocyte apoptosis. From a clinical perspective, our study highly implicated future development of small molecules with BCL-2 gene promoter i-motif as target for disrupting hepatocyte apoptosis as a promising avenue for NASH therapy. Our results showed that A22 could induce formation of BCL-2 promoter i-motif under near-neutral conditions. A22 could bind to BCL-2 promoter i-motif with good selectivity, since it had no significant binding to the i-motifs from other tested genes such as C-KIT and VEGF. The effect of A22 on gene expression in cells also showed that A22 could up-regulate BCL-2 expression without significant effect on other gene expressions. Compared with other complex methods for up-regulating gene expressions, our approach of using small molecules with gene promoter i-motif as target is relatively more practical.
Notably, BCL-2 not only controls classical caspase cascade, but also plays an important role in mitochondrial pathway in relation to apoptosis. The interference of BCL-2 associated apoptotic signaling pathway in hepatocyte can affect hepatocyte function, inflammatory response, and hepatic fibrosis. Our results showed that selective up-regulation of BCL-2 expression in hepatocyte not only increased mitochondrial membrane potential (red/green fluorescence intensity ratio) in early apoptosis, but also blocked caspase cascade reaction. By blocking hepatocyte apoptosis, A22 effectively gained function of hepatocytes, offset occurrence of pathological changes in NAFLD/NASH, and relieved steatohepatitis.
The expression levels of BCL-2 in liver and serum have been found to be down-regulated in NAFLD/NASH patients, suggesting that BCL-2 play an important role in hepatocyte apoptosis in NAFLD/NASH. By up-regulating BCL-2 expression, A22 showed apparent anti-apoptosis effect in many ways. The mitochondrial pathway of apoptosis is triggered by losing the integrity of mitochondrial outer membrane, which allows the release of proapoptotic factors (e.g., cytochrome c) from the mitochondria into the cytosol. This process is controlled by BCL-2. On the other hand, activation of caspase family members which are triggered by BCL-2 is also a critical mechanism of apoptosis. Our results showed that by up-regulating BCL-2, A22 could increase the red/green fluorescence intensity ratio (Supplementary Figure S4D) and inhibit the activation of caspase 3 and 9 ( Figure 4C), indicating its possible multi-functional antiapoptosis mechanisms.
NAFLD/NASH is characterized by liver injury, inflammation and fibrosis of pathological progression. Progression of hepatocyte apoptosis results in cell function impairment, liver inflammation and oxidative stress. Since A22 could act as an anti-apoptosis molecule for NAFLD/NASH, we studied whether it could offset occurrence of pathological changes. We found that A22 could effectively reduce liver damage as indicated by reductions of serum AST, ALT and ALP, and liver morphology tissue damage compared with HF diet-fed group. The accumulation of inflammatory factors occurred in HF diet-fed mice livers, characterized by up-regulation of inflammatory factors and accumulation of mononuclear macrophages and lymphocytes. A22 treatment effectively reduced expressions of inflammatory factors and attenuated inflammatory cell cluster. Liver fibrosis was also reversed in HF diet-fed mice after A22 addition through down-regulation of corresponding proteins such as collagen 1 and ␣-SMA involved in the development of fibrosis as well as collagen fibers aggression.
Liver is a key organ in pathogenesis of metabolic syndrome, which controls a vast majority of glucolipid metabolism. By blocking hepatocyte apoptosis, A22 could effectively alleviate the symptoms associated with metabolic syndrome including weight loss, glucose metabolism (IPGTT and IPITT), serum lipid levels, and insulin resistance index (HOME-IR) in mice. In addition, A22 reduced TG accumulation in the liver of HF diet-fed mice and cultured cells incubated with fatty acids. These strongly supported that increased expression of BCL-2 is significantly beneficial to hepatocyte function. Liver plays an important role in metabolism and utilization of glucose and lipid. We examined hepatocyte function of absorbing glucose both in lip-induced cells and animal models. The insulin resistance was significantly improved by A22 administration in a dose-dependent manner, as indicated by increase of 2-NBDG intake in HepG2 cells. The ability of glucose metabolism (IPGTT, IPITT, HOME-IR) after A22 administration in vivo was also improved. Interestingly, the accompanying weight loss was also observed. Overall, the present study demonstrated that A22 could up-regulate BCL-2 expression by targeting BCL-2 promoter i-motif, which showed significant anti-apoptosis effect to offset hepatocyte impairment, inflammation, and fibrogenesis in pathological progression of NAFLD/NASH.