Identification of flavonolignans from Silybum marianum seeds as allosteric protein tyrosine phosphatase 1B inhibitors

Abstract Protein tyrosine phosphatase 1B (PTP1B) is an attractive molecular target for anti-diabetes, anti-obesity, and anti-cancer drug development. From the seeds of Silybum marianum, nine flavonolignans, namely, silybins A, B (1, 2), isosilybins A, B (3, 4), silychristins A, B (5, 6), isosilychristin A (7), dehydrosilychristin A (8), and silydianin (11) were identified as a novel class of natural PTP1B inhibitors (IC50 1.3 7–23.87 µM). Analysis of structure–activity relationship suggested that the absolute configurations at C-7" and C-8" greatly affected the PTP1B inhibitory activity. Compounds 1–5 were demonstrated to be non-competitive inhibitors of PTP1B based on kinetic analyses. Molecular docking simulations resulted that 1–5 docked into the allosteric site, including α3, α6, and α7 helix of PTP1B. At a concentration inhibiting PTP1B completely, compounds 1–5 moderately inhibited VHR and SHP-2, and weakly inhibited TCPTP and SHP-1. These results suggested the potentiality of these PTP1B inhibitors as lead compounds for further drug developments.


Introduction
Protein tyrosine phosphatase 1B (PTP1B) is a non-transmembrane protein tyrosine phosphatase 1 , expressing ubiquitously in the classical insulin-targeted tissues, and negatively regulating insulin and leptin signalling pathway 2 . Inhibition of PTP1B has been demonstrated beneficial effects, such as increased insulin sensitivity, improved glucose tolerance, and resistance to high-fat-induced weight gain, but without side effects 3 . Thus, PTP1B inhibitors have attracted much attention for anti-diabetes and anti-obesity drug developments 4 . Recently, PTP1B also found a positive role in the tumorigenesis of breast cancer and colorectal cancer, extending the application of PTP1B inhibitors as anti-cancer agents 5 . Up to date, several PTP1B inhibitors, such as ertiprotafib 6 , and trodusquemine 7 , have been developed into clinical trials. Despite the fact that chemical synthetic PTP1B inhibitors have reached to quite potent inhibitory activities, some barriers are remaining, such as low PTP selectivity, low bioavailability, and insufficient in vivo efficacy 8 . In contrast, natural products are recognised as another important resource to discover novel PTP1B inhibitors 9 .
Silybum marianum (L.) Gaertn. (Asteraceae), also known as milk thistle, is a well-known medicinal plant both in Europe and Asia 10 . S. marianum has been reported to have various pharmacological effects, such as hepatoprotective, anti-inflammatory, antifibrotic, and antioxidant effects [11][12][13] . Recently, the effectiveness of flavanolignans from the seeds of S. marianum in ameliorating diabetes either in vitro or in vivo was also reported [14][15][16][17] .
During our ongoing investigations to discover novel PTP1B inhibitors from natural resources, we have reported a number of natural PTP1B inhibitors [18][19][20][21][22] , which included canthinone alkaloids from Simaroubaceae medicinal plants, isoprenylated flavonoids from Glycyrrhiza species and Sophora flavescens, neolignans from Magnolia officinalis, anthraquinones from Rheum officinale, and triterpenes and neolignans from Sambucus adnate. These natural PTP1B inhibitors shared a great structural diversity, which was different from synthetic inhibitors, providing more chance for development of novel PTP1B inhibitors. Herein, we report the identification of flavanolignans from the seeds of S. marianum as a novel class of natural allosteric PTP1B inhibitors.

Plant material
The seeds of S. marianum (L.) Gaertn. were purchased from Liaoning Shengbo Pharmaceutical Co., Ltd. (Shenyang, China), and identified by Professor Jincai Lu (Shenyang Pharmaceutical University). A voucher specimen (SM2014) has been deposited at the Department of Traditional Chinese Materia Medica, Shenyang Pharmaceutical University, China.

Extraction and bioassay-guided isolation
The seeds of S. marianum (30 kg) were powdered and defatted by extraction with petroleum ether. The residue was extracted with 95% ethanol (60 L) under reflux for three times. After filtration and evaporation, a crude extract (2.0 kg) (PTP1B inhibition rate: 73.1% at 1.0 mg/mL) was obtained. The crude extract was suspended in water and sequentially partitioned with CH 2 Cl 2 , EtOAc, and n-butanol to yield CH 2 Cl 2 (184 g), EtOAc (200 g), and n-butanol soluble fraction (390 g). The EtOAc fraction was chosen for further separation because it exhibited more potent PTP1B inhibitory activity (inhibition rate: 92.1% at 1.0 mg/mL) than CH 2 Cl 2 , n-butanol, and water fractions (inhibition rate: 62.0%, 49.3%, and 59.2% at 1.0 mg/mL, respectively). The EtOAc extract was fractionated by silica gel CC, eluting with gradient CH 2 Cl 2 -MeOH (from 1:0 to 0:1, v/ v), to afford seven fractions (A-G). Among them, fractions E and G exhibited the most potent inhibitory activity (inhibition rate: 97.8% and 91.2% at 1.0 mg/mL).   PTP1B inhibitory activity assay PTP1B inhibitory activity was measured using p-NPP as the substrate. A mixture consisting of p-NPP and PTP1B in a buffer containing 0.06 M citrate (pH 6.0), 0.1 M NaCl, 1 mM EDTA, and 1 mM DTT with or without a tested compound solution (prepared in the above buffer solution containing 30% dimethyl sulfoxide), was incubated at 37 C for 30 min. The substrate was used at a concentration of 4 mM. The reaction was terminated by adding 20 mL of 10 M NaOH. The reaction mixture was blended by a microplate mixer for 5 min and the amount of produced p-nitrophenol was tested by measuring the absorbance at 405 nm. The blank was measured in the same way except adding a buffer solution instead of the enzyme. The inhibitory activities were further measured at a number of stepwise gradient of concentrations ranging from 100 mM to 0.01 mM to obtain the IC 50 values by regression analysis. Ursolic acid was used as the positive control due to the commercial availability, similar inhibitory potency, and standardised bioassay process in our lab. All compounds used for bioassay were confirmed the purity >98% by HPLC-PDA and 1 H NMR spectroscopic analyses.

Molecular docking simulation
Docking experiment was carried out by using a software Biovia Discovery Studio 4.5 (Accelrys Inc., San Diego, CA). The stable structures of compounds were prepared by a standard dynamics cascade. The X-ray crystal structure of PTP1B (PDB code: 1T48, residues 1-283, and 290-298) was obtained from a protein data bank (http://www.rcsb.org). The residues 284-289 were built using the closed form PTP1B crystal structure (PDB code: 5KA9 (residues 1-294, including a7 helix)) to generate PTP1B 1-298 structure. Docking simulation was carried out using CHARMm-based DOCKER (CDOCKER). The docking results differing by <2.0 Å on the basis of a positional root mean square deviation (RMSD) were clustered together and were ranked on the basis of free binding energy. All other parameters were maintained as default.

PTP1B inhibition
The PTP1B inhibitory activities of compounds 1-11 were evaluated. A known PTP1B inhibitor, ursolic acid, was used as the positive control (IC 50 ¼ 3.52 mM). As the result, nine compounds (1-8 and 11) inhibited PTP1B in a concentration-dependent manner, and their IC 50 values were determined by regression analysis (Table 1). Among them, compounds 1 and 4 showed the most potent PTP1B inhibitory activity with IC 50 values of 1.54 and 1.37 mM, respectively. The stronger inhibitory activity of 1 than 2 (IC 50 ¼ 5.65 mM), and 4 than 3 (IC 50 ¼ 2.65 mM), suggesting that the absolute configurations at C-7" and C-8" are important for inhibitory activity. Compound 1 showed more potent activity than 5 (IC 50 ¼ 5.58 mM), suggesting that the B ring substituents pattern greatly affected the activity. The activities of compounds 5 (IC 50 ¼ 5.58 mM), 7 (IC 50 ¼ 6.55 mM), and 8 (IC 50 ¼ 6.55 mM) were almost same, indicating that the C ring double bond between C-2 and C-3 did not affect the activity. Furthermore, silydianin type flavanolignans 9 and 10 exhibited no activity (IC 50 > 50 mM).

Kinetic analysis
To elucidate the inhibition mode, the inhibition kinetics of compounds 1-5 were analyzed by the Lineweaver-Burk method with various substrate concentrations of p-NPP (4, 8, 16 mM). The initial reaction velocities were measured with and without the inhibitor. The V max values increased in a dose-dependent manner without changing the K m value, indicating that they inhibited PTP1B activity by a non-competitive mechanism ( Figure 2). The secondary plot of slope from the Lineweaver-Burk plot on the y-axis against the concentration of the compound on the x-axis, obtained the quadratic-like curves, exhibited a good linear plot. Dissociation constant (K i ) values of compounds 1-5 were calculated as 1.25, 4.05, 2.25, 1.03, and 3.95 mM, respectively.

Docking simulation
As compounds 1-5 were non-competitive type PTP1B inhibitors, their binding mode to the allosteric site of PTP1B was investigated by a docking simulation using Biovia Discovery Studio 4.5 19 . Preferred coordination mode of compounds 1-5 with the allosteric site of PTP1B is shown (Figure 3). The binding energy of compounds 1-5 for the docking experiment was calculated to be 32.5, 25.9, 31.7, 39.2, and 22.5 kcal/mol, respectively, suggesting that the docking energy was proportional to IC 50 values. Compounds 1-5 interacted with the amino acid residues belonging to a3, a6, and a7 helix of PTP1B (Figure 3). Four amino acid residues, namely, Ala189, Leu192, and Phe196 in a3 helix, and Glu276 in a6 helix of PTP1B seem to be important for the inhibitory activity of flavonolignans, as they were commonly binding to all five compounds. Compounds 1 and 2, which are different from their structures only by the configurations at C-7" and C-8", were binding to PTP1B similarly. The fact that compound 1 showing more potent inhibitory activity than 2, maybe interpretable by a hydrogen bond between 3''-OCH 3 and Ser286, which was only observed in 1. In contrast, compounds 3 and 4 are also a pair of diastereomers at C-7" and C-8", and they are binding to PTP1B similarly too. The structure of 3 was in favour of PTP1B inhibition due to the multiple interactions of Glu276 and Ala189 with 3. In a word, compounds 1-5 blocked the interaction between a7 and a3-a6, and prevented the closure of catalytic WPD loop, which is in agreement of the known allosteric PTP1B inhibitors 28 .

PTP selectivity
Because of the high structural similarity of the catalytic centre among the family of protein tyrosine phosphatases 29 , the inhibitory selectivity is an important evaluation index for development of PTP1B inhibitors. Four non-receptor-like and cytosolic PTPase homologous protein tyrosine phosphatases, namely, TCPTP, VHR, SHP-1, and SHP-2 were selected for comparison the PTPs inhibitory activities. Compounds 1-5 were evaluated their inhibitory activity against five PTPs at their PTP1B completely inhibitory concentrations, respectively. As the results, compounds 1-5 share the similar profile of PTP inhibitory selectivity (Table 2). Namely, when compounds 1-5 completely inhibited PTP1B, they moderately inhibited VHR (inhibition rate: 59.2-67.5%), and SHP-2 (32.7-50.1%), and weakly inhibited TCPTP (15.7-32.2%) and SHP-1 (22.9-33.4%). Among the five compounds, 1 showed the best inhibitory selectivity between PPTP1B and other PTPs but still, need further improvement.

Conclusions
In conclusion, nine flavonolignans from seeds of S. marianum were identified as a novel class of novel natural PTP1B inhibitors. Among them, compounds 1-5, which showed potent inhibitory activities (IC 50 1.37-5.65 mM), were demonstrated to be non-competitive PTP1B inhibitors, and with inhibitory selectivity between PTP1B and other PTPs. Docking analysis supported the results of PTP1B inhibitory activity assay and kinetics analysis. Flavonolignans from seeds of S. marianum have been demonstrated to have beneficial effects against diabetic mellitus. The antidiabetic effect of silybin A (1) was attributed to inhibition of gluconeogenesis in the liver and decrease of glucose-6 phosphatase activity 30 . In addition, isosilybin (3) has been identified as an agonist of peroxisome proliferator-activated receptor c 14 . Our investigation demonstrated that these PTP1B inhibitory compounds could have potential as lead compounds for further antidiabetes drug developments. 50.1 ± 1.5 48.6 ± 1.9 a Inhibition rate (%) is mean ± SD from three separate experiments at sample final concentrations: Compounds 1 and 4: 4.0 lM, 2: 12 lM, 3: 6.0 lM, 5: 20 lM.