Aurone derivatives as Vps34 inhibitors that modulate autophagy.

We report in this study the identification of a natural product-like antagonist (1a) of Vps34 as a potent autophagy modulator via structure-based virtual screening. Aurone derivative 1a strongly inhibited Vps34 activity in cell-free and cell-based assays. Significantly, 1a prevents autophagy in human cells induced either by starvation or by an mTOR inhibitor. In silico modeling and kinetic data revealed that 1a could function as an ATP-competitive inhibitor of Vps34. Moreover, it suppressed autophagy in vivo and without inducing heart or liver damage in mice. 1a could be utilized as a new motif for more selective and efficacious antagonists of Vps34 for the potential treatment of autophagy-related human diseases.

Introduction

Autophagy acts as a stress response pathway against pathologic stresses, and also maintains proper function in cells by increasing the turnover of proteins and organelles and degrading damaged cytoplasmic components1., 2.. Defective autophagy is implicated in the development of maladies, such as diabetes, myopathy, neurodegeneration, liver disease, cancer, infection and immune disease3., 4., 5., 6.. Vps34, a catalytic subunit of phosphatidylinositol 3-kinase (PI3K) class III, mediates endocytosis as well as autophagosome—autolysosome creation so as to regulate autophagy and maintain cellular homeostasis7., 8.. Among the components of the autophagy machinery, Vps34 is the only class III kinase responsible for generating phosphatidylinositol 3-phosphate (PI3P) that mediates the start of autophagosome biogenesis. Vps34 also plays an essential role in heart and liver function and its complete suppression in mammals can cause hepatomegaly, hepatosteatosis, and cardiomegaly10., 11.. Therefore, it is important to discover novel small molecule Vps34 modulators that can provide new opportunities for drug discovery and help understand the molecular mechanisms of autophagy, but without triggering the aforementioned heart and liver side effects.

As the C-terminus region of Vps34 binds to ATP, targeting the ATP-binding pocket of Vps34 is a potential approach for the discovery of novel Vps34 inhibitors. However, it is far more difficult to identify Vps34 ATP-competitive inhibitors compared to class I PI3K inhibitors due to the smaller size of the Vps34 ATP-binding pocket13., 14., 15.. Several ATP-competitive inhibitors of Vps34 have been reported in the literature, including SAR405, Vps34-IN1, and 3-methyladenine (3-MA). However, the potential hepatotoxicity and cardiotoxicity (or lack thereof) of those Vps34 inhibitors have so far not been demonstrated.

Natural products have long been regarded as a rich source of structural motifs for drug discovery19., 20., 21., 22.. Advances in virtual screening methodologies have allowed large numbers of natural products or natural products-derived compounds to be screened in silico with a dramatically reduction in costs when compared to traditional high-throughput screening23., 35., 36., 37., 38.. We report herein the structure-based discovery of a novel and potent natural products-like Vps34 inhibitor as an autophagy modulator that does not damage the heart or liver in mice.

Results and discussion

Screening and structure-based optimization of small molecules as Vps34 inhibitors

The X-ray structure of Vps34 complexed with SAR405 (PDB: 4OYS) was used to construct a molecular model for our investigations. A total of 90,000 natural products and natural products-derived structures were docked into the Vps34–ATP site of Vps34 in silico using the ICM-Pro (3.6-1d) docking algorithm. Eleven compounds 1a and 2–11 (Fig. 1) exhibited Gibbs free energy (ΔG) changes of lower than −30.0 kJ/mol, and were shortlisted for further biological testing.

Figure 1

Chemical structures of compounds 1a–1q, 2–11 and SAR405.

An in vitro enzyme-linked immunosorbent assay (ELISA) was employed to detect the inhibitory effects of compounds (1a, 2–11) on Vps34 kinase activity. Aurone derivative 1a displayed the highest inhibition of Vps34 activity, with 79.6% reduction in luminescence activity at 100 nmol/L (Fig. 2). Compounds 3, 4, 8, 10 and 11 showed moderate inhibitory activity in this assay, while little or no activity were exhibited by compounds 2, 5–7, and 9. Notably, 1a showed higher potency than SAR405, a known potent and selective Vps34 inhibitor. A dose analysis was subsequently carried out to quantitate the efficacy of the aurone derivative 1a at inhibiting Vps34 activity. The results showed that aurone derivative 1a inhibited Vps34 in a concentration-dependent fashion with an IC50 of 7.6 nmol/L (Supporting Information Fig. S1), while SAR405 exhibited an IC50 value of 38 nmol/L under similar conditions. Compound 1a also exhibit selectivity toward Vps34 over other PI3Ks isoforms, including p110α/p85α (IC50>1000 nmol/L), p110β/p85α (IC50>1000 nmol/L), p120γ (IC50 ca. 1000 nmol/L), and p120δ/p85α (IC50>1000 nmol/L) using ELISA (Supporting Information Fig. S2). Moreover, kinetic analysis showed that like SAR405, aurone derivative 1a acts as an ATP-competitive inhibitor of Vps34 in a manner similar to that of SAR405 (Supporting Information Fig. S3). The lowest-scoring binding mode of 1a in the ATP binding pocket of Vps34 is shown in Fig. 3. A high degree of shape complementarity is observed between the aurone derivative and the ATP binding pocket of Vps34, suggesting that this protein–ligand interaction could be stabilized by significant hydrophobic interactions. The side-chain carbonyl oxygen group of the aurone derivative 1a is calculated to hydrogen bond with the side-chain of Asp761 along with the furanone carbonyl group forming similar hydrogen bonding interactions with the backbone amide motif of Ile685.

Figure 2

Compounds 1a, 2–11 (100 nmol/L) inhibit the activity of Vps34 as determined by an ELISA assay. Error bars represent the standard deviations of results obtained from three independent experiments. Data are expressed as means ± SD (n=3). P values were calculated using a two-sided t-test. *P < 0.05, **P < 0.01 compounds vs control, NS (not significant, P > 0.05) 1avs SAR405.

Figure 3

Top view of aurone derivative 1a bound to the VPS34 generated by molecular docking. Vps34 (PDB: 4OYS) is depicted as a space-filling representation showing carbon (yellow), oxygen (red) atoms. The binding pocket of the Vps34 is represented as a translucent green surface (for interpretation of the references to color in this figure legend, the reader is referred to the web version of this article).

Structure—activity relationships

Inspired by the promising activity of aurone derivative 1a, a series of derivatives were synthesized to study the structure—activity relationships (SAR) surrounding this scaffold. Analogues 1b–1g retained the 3(2 H)-benzofuranone core of 1a, but varied in the nature of the substituents attached at the 2- and 6-positions (Fig. 1). In the ELISA assay, 1a was again the most potent inhibitor of Vps34 out of all the analogues (Supporting Information Fig. S4). Considering those results, preliminary SAR could be determined. After compound 1a, the second most potent molecule was 1g, suggesting that a pendant benzofuran group at the 6-position could be mildly tolerated in place of the styrene group. However, converting the ester group at C6 into a hydroxy group (as in 1d) led to the abolishment of activity. Adding fluorine atoms to the C2 pendant phenyl ring (as in 1b and 1c) was also undesirable for anti-Vps34 potency.

In the next round of screening, a series of further analogues (1h–1q) were synthesized. Interestingly, having a nitrile group at the para position (as in 1j) completely abolished activity, whereas a nitrile group at the meta position (as in 1k) produced significant activity. This suggests that steric effects around the C2 pendant ring may be important determinants of activity. Adding a fluorine atom to the C5 pendant group of 1o (giving 1p) led to a further weakening of activity. Finally, reducing the C6 ester carbonyl group of 1a to methylene (as in 1l) led to a significant decrease of activity.

Aurone derivative 1a affects vesicle trafficking in cellulo

Recent studies have shown that Vps34 can regulate the expression level of p62, which is negatively correlated with autophagy25., 26.. Hence, the level of p62 in human cervical cancer HeLa cells treated with aurone derivative 1a was evaluated. This was first confirmed by using an MTT assay that showed the aurone derivative to be only moderately cytotoxic towards HeLa, which gave an IC50 of 35 μmol/L (Supporting Information Fig. S6). The cytotoxicity of compound 1a was further evaluated in three other cell lines, including a human breast adenocarcinoma cell line (MCF-7), a human hepatic cell line (LO2), and a human embryonic kidney cell line (HEK293T) using the MTT assay. 1a also showed moderate cytotoxicity to MCF-7, LO2, and HEK293T cell lines with IC50 values of >100, 25.12, and 37.15 μmol/L (Supporting Information Fig. S7), respectively. Encouragingly, aurone derivative 1a (at 0.3–10 µmol/L) induced a dose-dependent increase in p62 in HeLa cells as revealed through Western blotting (Fig. 4A and Supporting Information Fig. S8). This result suggests that 1a could suppress autophagy by inhibiting Vps34 activity. In this assay, the aurone derivative 1a showed comparable activity to the positive inhibitor SAR405. The difference in in vitro and in cellulo activities between compound 1a and SAR405 could be due to multiple factors, such as differences in cell absorption or metabolism27., 28.. We also examined the effect of 1a on vesicle trafficking, which is an important biological function of Vps34. HeLa cells were stained with the lysosomal probe Lyso-Tracker Red and Hoechst 33258 in the presence of 1a before imaging. Many large red spots were observed in the 1a-treated HeLa cells (Fig. 4B), indicating the formation of late-endosomal compartments. By comparison, no such red spots were observed in untreated cells. These results reveal that aurone derivative 1a can suppress vesicle trafficking presumably by inhibiting the activity of Vps34 in cellulo.

Figure 4

Effect of aurone derivative 1a on Vps34 activity in cellulo. (A) The level of p62 in HeLa cells treated with 1a was detected by Western blotting. (B) Effect of aurone derivative 1a on vesicle trafficking in HeLa cells. HeLa cells were pretreated with 1a (10 μmol/L) for 24 h before cell imaging. Scale bar = 20 μm.

Aurone derivative 1a engages Vps34 in cellulo

A cellular thermal shift assay (CETSA) was conducted to confirm binding affinity of Vps34 by aurone derivative 1a. An obvious shift of ca. 5 °C in the Vps34 melting curve was observed when 10 µmol/L of 1a was present, which indicated that 1a could engage with and stabilize Vps34 in cell lysates (Fig. 5A and B). We also performed co-immunoprecipitation experiments to investigate whether 1a could affect the formation of Vps34 complexes. Aurone derivative 1a did not disrupt the interaction between Vps34 with its partner proteins Vps15 and Beclin 1 (Supporting Information Fig. S9). Furthermore, the levels of Beclin 1 were not significantly increased after aurone derivative 1a treatment in HeLa cells as revealed by Western blotting (Supporting Information Fig. S10), indicating 1a could regulate Vps34-Beclin 1 complexes without affecting the levels of Beclin 1.

Figure 5

Aurone derivative 1a selectively engages Vps34 as its prime cellular target. (A) Stabilization of Vps34 in HeLa cell lysates by 1a. (B) Densitometry analysis of the cell-based Western blot showing 1a-mediated stabilization of Vps34. (C) p62 levels in HeLa cells with or without knockdown Vps34 in the presence or absence of 1a. (D) Densitometry analysis of p62 levels on the Western blot. Data are expressed as mean ± SD (n=3). *P < 0.05, **P < 0.01 1avs DMSO.

Moreover, a knockdown assay was performed to further verify the targeting of Vps34 by 1a. HeLa cells were preincubated with siVps34 to silence Vps34 expression (Supporting Information Fig. S11). As expected, knockdown of Vps34 resulted in an accumulation of p62 (Fig. 5C). However, the aurone derivative was less able to further increase p62 accumulation in Vps34 knockdown cells compared with control cells (Fig. 5D). Collectively, our results indicate that 1a could target and engage Vps34 in cells, and that its inhibition is not mediated through disruption of the Vps34 complex. This is consistent with the molecular modeling and kinetic analysis performed above suggesting that aurone derivative 1a targeted the ATP-binding pocket of Vps34.

1a prevents starvation- and rapamycin-induced autophagy

Vps34 has a key function in both starvation and mTOR inhibitor-induced autophagy. Therefore, we examined the effect of aurone derivative 1a on autophagy induced by nutrient stress or mTOR inhibition. GFP-LC3 HeLa cells stably expressing GFP-LC3, a marker of autophagosomes, were used to monitor autophagy caused by starvation or mTOR inhibition31., 32.. Before treatment with 1a, GFP-LC3 HeLa cells were preincubated with an amino acid and growth factor-free medium (Earle׳s balanced salt solution (EBSS)) for 24 h to induce starvation leading to autophagy. EBSS-soaked HeLa showed the presence of GFP-LC3 spots, indicative of autophagosome formation, but these were drastically reduced by an increasing dose of aurone derivative 1a (Fig. 6A). Moreover, the expression level of LC3-II was reduced by aurone derivative 1a in starved HeLa cells in a dose-dependent fashion (Fig. 6B and Supporting Information Fig. S12). Rapamycin, a known mTOR inhibitor, was used to activate autophagy in HeLa cells. As expected, a large number of GFP-LC3 spots appeared in HeLa cells treated with rapamycin, as autophagosomes were formed (Supporting Information Figs. S13 and 14). In contrast, the formation of autophagosomes was significantly inhibited after treatment with 1a in a dose-dependent fashion (Supporting Information Figs. S13 and 14). To further evaluate the autophagy inhibition mechanism of 1a, HeLa cells were transfected with mCherry-GFP-LC3 plasmid, which allows monitoring of the change in fluorescence (yellow to red) as a measure of autophagic flux. As shown in Supporting Information Fig. S15, the formation of autophagosomes (green dots), and clearance of autophagosomes by lysosomes (red dots) were decreased dramatically in rapamycin-induced HeLa cells after incubation with 1a, indicating that 1a could inhibit rapamycin-induced autophagic flux. Furthermore, we have detected the level of mTOR to further confirm the autophagic inhibition of 1a. The results showed that 1a had no appreciable effect on the levels of mTOR in HeLa cells (Supporting Information Fig. S16), which further indicates that aurone derivative 1a could prevent rapamycin-induced autophagy through selectively targeting Vps34, rather than significantly regulating the levels of mTOR. Finally, transmission electron microscopy (TEM) was performed to verify the effects of aurone derivative 1a on autophagy in rapamycin-induced HeLa cells. Representative TEM images are shown in Supporting Information Fig. S17. Typical autolysosomes induced by rapamycin were decreased dramatically after 1a treatment in HeLa cells. Taken together, the results indicated that aurone derivative 1a can inhibit both starvation and mTOR inhibition-induced autophagy, which we attribute at least in part to its ability to inhibit Vps34 activity in cellulo.

Figure 6

Aurone derivative 1a inhibits starvation-induced autophagy in HeLa cells stably transfected with GFP-LC3 plasmid. HeLa cells were incubated with EBSS for 24 h before treated with 1a for 24 h. (A) Cell imaging was imaged by a Leica TCS SP8 confocal laser scanning microscope system in GFP channel. Scale bar = 10 μm. (B) Cell lysis was analyzed by Western blotting.

The effect of aurone derivative 1a on mice heart and liver

Considering the important effect of Vps34 in the function of mammalian liver and heart, we studied the changes in heart and liver function in mice exposed to aurone derivative 1a. Six-week-old mice were intraperitoneally injected either with vehicle (100% PEG 400) or with 60 mg/kg of 1a every three days for 30 days. The results showed that the aurone derivative had no observable effect on both total body weight and also the weights of the heart and liver of treated mice (Supporting Information Figs. S18 and S19). Similarly, the morphology of mice heart and liver showed no obvious changes (Fig. 7A). Moreover, 1a has no effect on the activity of alanine aminotransferase (ALT) and aspartate aminotransferase (AST), which are used to indicate liver damage or disease (Supporting Information Fig. S20). Consistent with the cellular experiments, 1a induced the accumulation of p62 in the heart and liver without having a discernible effect on Vps34 expression (Fig. 7B and Supporting Information Fig. S21). Taken together, these data suggest that 1a can suppress autophagy activity without damaging the heart and liver in mice. This result is significant because the genetic deletion of Vps34 induces heart and liver damage, and therefore aurone derivative 1a offers a potential therapeutic option for the treatment of autophagy-related diseases without inducing cardiotoxicity and hepatotoxicity.

Figure 7

Effect of aurone derivative 1a on the Vps34 in mice heart and liver. (A) Anatomical views of representative heart and liver treated with 1a. Scale bar = 2 mm. (B) The level of p62 in mice heart and liver was detected using a Western blotting assay.

Pharmacokinetic study of aurone derivative 1a

Prior to the pharmacokinetic study of aurone derivative 1a, the compound standard peak was determined by LC—MS/MS. The aurone derivative 1a displayed three peaks which were confirmed as their ion fragments under MRM mode (ion-pairs: m/z 395→131; m/z 395→265) (Supporting Information Fig. S23A).

The pharmacokinetic profile of aurone derivative 1a was studied at 2 h after compound injection by intraperitoneal injection in mice. The Cmax value occurred between 4 to 12 h, and the maximum mean peak area (807.4 mAu) of aurone derivative 1a in plasma was obtained at 8 h (Supporting Information Fig. S23B). However, due to the presence of isomers found in the standard aurone derivative 1a, the actual amount of aurone derivative 1a in the plasma sample could not be determined. Therefore, the pharmacokinetic profile of aurone derivative 1a is presented as the mean peak area (mAu) at different time points in order to allow evaluation of the approximate duration and maximum concentration of the drug in the plasma (Fig. 8).

Figure 8

Pharmacokinetic profile of aurone derivative 1a. Mean peak area of aurone derivative 1a in blood at different time points. C57BL/6J mice were administered with an intraperitoneal injection of aurone derivative 1a at 60 mg/kg. The concentrations of aurone derivative 1a in the blood samples were measured at 0, 15, 30, 60, 120, 240, 480, 720 and 1440 min after administration. All data represent mean ± SEM; n=3.

Conclusions

In summary, we have discovered aurone derivative 1a as a potent inhibitor of the Vps34 activity via computer-based modeling of a natural products and natural product-like library. The aurone derivative 1a upregulated p62 levels and inhibited vesicle trafficking in HeLa cells. Kinetic analysis suggested that it could function as a reversible ATP competitive inhibitor of Vps34, and is in accordance with the in silico modelling data showing that 1a binds to the ATP-binding pocket of Vps34. Moreover, 1a could prevent autophagy induced by starvation or by an mTOR inhibitor. Systemic suppression of Vps34 can cause severe cell and organ damage. Pleasingly, 1a was able to promote p62 accumulation in vivo without affecting the morphology of the mice heart and liver. Pharmacokinetic study of 1a showed that the time to reach Cmax in plasma was 8 h after single intraperitoneal administration (60 mg/kg) in mice. We envisage that aurone derivative 1a may be harnessed as an effective scaffold for the further discovery of efficacious antagonists of Vps34 for the potential treatment of autophagy-related human diseases without inducing cardiotoxicity and hepatotoxicity.

Experimental

Compounds 1a–11 (commercial available, purity >90%) used in cell experiments were purchased from J&K Scientific Ltd.(Hong Kong, China). We synthesized and purified the aurone derivative 1a, for the animal study. SAR405, XTT kit, Hoechst 33258, Lipo3000 transfection reagent and siVps34 were purchased from Sigma—Aldrich (Santa Clara, CA, USA). ADP-Glo™ Kinase Assay System was purchased from Promega (Madison, WI, USA). Lyso-Tracker Red was purchased from Beyotime Biotechnology (Nanjing, China). Vps34 antibody, Vps15 antibody, Beclin 1 antibody, p62 antibody, LC3 antibody and GAPDH antibody were purchased from CST (Cell Signaling Technologies, Danvers, MA, USA) and Proteintech Group (ProteinTec Group Inc., Chicago, IL, USA), respectively. All the complexes were immersed in dimethyl sulfoxide (DMSO). Further synthetic details and full descriptions of biological methods are contained in the Supporting Information Experimental Section.

(Z)-3-0xo-2-((E)-3-phenylallylidene)-2,3-dihydrobenzofuran-6-yl cinnamate (1a)

Cinnamoyl chloride (554 mg, 3.31 mmol) was added to a solution of alcohol 1d, (880 mg, 3.3 mmol), triethylamine (370 mg, 3.66 mmol) and DMAP (36 mg, 0.3 mmol) in dry tetrahydrofuran (33 mL) at 0 °C. The solution was allowed to warm to room temperature and left to stir for 16 h. The reaction mixture was quenched with saturated NH4Cl solution and the aqueous layer extracted with EtOAc (×3). The combined organic layers were washed with brine (×1) and dried over MgSO4. The solvent was removed under reduced pressure to give crude target material as a yellow solid. Purification by flash column chromatography (SiO2: EtOAc/hexanes=1:9–1:5) gave compound 1a as a yellow solid (581 mg, 1.47 mmol, 45%). 1H NMR (CDCl3, 600 MHz): δ 7.92 (d, J=16.0 Hz, 1 H), 7.79 (d, J=8.3 Hz, 1 H), 7.60 (dd, J=6.9, 2.3 Hz, 2 H), 7.58–7.53 (m, 2 H), 7.44 (q, J=5.0, 4.1 Hz, 3 H), 7.38 (t, J=7.5 Hz, 2 H), 7.36–7.24 (m, 2 H), 7.18 (d, J=1.9 Hz, 1 H), 7.05–6.99 (m, 2 H), 6.79 (d, J=11.5 Hz, 1 H), 6.63 (d, J=16.0 Hz, 1 H); 13C NMR (CDCl3, 150 MHz): 13C NMR (151 MHz, CDCl3) δ 182.3, 166.0, 164.3, 157.4, 147.8, 147.7, 141.5, 136.2, 133.8, 131.0, 129.3, 129.0, 128.8, 128.4, 127.4, 125.3, 120.4, 120.0, 117.3, 116.3, 114.4, 106.4.; IR (NaCl, neat) v: 3123, 2711, 1637, 1573, 1494, 1454, 1397, 1316, 1285, 1251, 1198, 1134, 1100, 966 cm-1; HR-MS (ESI) Calcd. for C26H19O4 [M++H]: 395.1283, Found: 395.1280.

ELISA assay

ELISA assay was performed according to the assay protocol. Each well of the pre-cooled 96-well plate containing 10 μL of diluted active PIK3C3 protein, 5 μL of 125 μmol/L solution of phosphatidylinositol (PI)/phosphatidylserine (PS), 5 μL of kinase dilution buffer with 0.1% triton-100 and 100 nmol/L of a compound or 1 μL DMSO. Initiate the reaction by the addition of 5 μL of 250 μmol/L ATP solution. Sonicate the reaction mixture in the 96-well opaque plate for 10 s and continue the incubation at 30 °C for 15 min. Add 25 μL of ADP-Glo reagent to stop the reaction and deplete the remaining ATP. Shake the 96-well plate and then incubate the mixture for another 40 min at ambient temperature. Then add 50 μL of the kinase detection reagent to the each well and incubate the reaction mixture for another 30 min at the ambient temperature. Read the 96 well reaction plate using the kinaseGlo luminescence protocol on an Envision Multilabel Plate Reader (Perkin-Elmer, Waltham, MA, USA). Determine the corrected activity (RLU) by removing the blank control value for each sample and calculating the kinase specific activity.

Kinetic assay

Each well of the pre-cooled 96-well plate contained different concentrations of PIK3C3 protein and aurone derivative 1a, 5 μL of 125 μmol/L solution of PI/PS, 5 μL of kinase dilution buffer with 0.1% Triton-100. The reaction was initiated by the addition of different levels of ATP solution. The ELISA assay was then performed as above.

Vesicle trafficking

HeLa cells were seeded at the density of 1 × 105 cells in 96 well plate for 24 h. 1a or DMSO was added to each well for the 24 further incubations. Cells were washed with PBS for 3 times and pretreated with Lysotracker Red (1:20,000) and Hochest 33258 (1:2000) for 15 min at 37 °C before cell imaging. The treated HeLa cells were imaged using the GE InCell Analyzer 2000 with the indicated channels.

mTOR inhibitor induced autophagy

GFP-LC3 HeLa cells were treated with 3 μmol/L rapamycin for 4 h in fed conditions before treated with DMSO or 1a for 24 h. Cell imaging of HeLa cells fixed with 4% PFA were then captured using a Leica TCS SP8 confocal laser scanning microscope system in GFP channel. Or cells were collected and lysed for Western blotting assay.

Starvation-induced autophagy

After cultured in fed conditions or starved in EBSS for 24 h, GFP-HeLa cells were then treated for with 10 μmol/L hydroxychloroquine to terminate the autophagy. DMSO- or 1a-treated HeLa cells were either fixed with 4% PFA and then captured in the GFP channel for cell imaging, or harvested for Western blotting assay.

Pharmacokinetic study

Six-weeks-old C57BL/6J female mice (body weight 25 ± 2 g) were purchased from The Chinese University of Hong Kong (Hong Kong). Aurone derivative 1a was dissolved into 12 mg/mL in diluent buffer (PEG400:EtOH:ddH2O=6:1:3). Each mouse took a single intraperitoneal administration of 60 mg/kg aurone derivative 1a. After drug administration, the blood samples were collected from the retinal vein into heparinized centrifuge tubes at appropriate intervals (0.25, 0.5, 1, 2, 4, 8, 12, 24 and 48 h). After centrifugation at 3000×g for 10 min, 100 μL of plasma was collected and stored at −80 °C until analysis. The amounts of aurone derivative 1a in the plasma were estimated by UPLC—MS/MS analysis as described in Supporting Information. A 50 μL aliquot of plasma sample was spiked with 3.2 μL resveratrol (internal standard, 1 μg/mL). After vortexing, 300 μL methanol was added, ultrasonic dissolving and centrifuged at 9391×g for 10 min. The supernatant was collected, and then dried using nitrogen blowing instrument. The residue was reconstituted with 200 μL of methanol, and then 5 μL of the aliquot was injected into the UPLC—MS/MS system.