Identification of new PTP1B-inhibiting decipiene diterpenoid esters from Eremophila clarkei by high-resolution PTP1B inhibition profiling, enzyme kinetics analysis, and molecular docking

of new PTP1B-inhibiting decipiene diterpenoid esters from Eremophila clarkei by high-resolution PTP1B inhibition profiling, enzyme kinetics analysis


Introduction
Plants of the genus Eremophila (Scrophulariaceae) are endemic to Australia and mainly distributed in the arid and semi-arid regions of Australia.Eremophila consists of approximately 230 species, of which some have been used in traditional Australian Aboriginal medicine [1].Eremophila is rich in structurally diverse and unique diterpenoids, with serrulatane, decipiene, viscidane, cembrene, and eremane type diterpenoids being most abundant [2].A range of pharmacological properties have been reported for Eremophila diterpenoids, including antibacterial [3][4][5][6][7], anti-inflammatory [7], antidiabetic [8][9][10], and cytotoxic activity [6].Eremophila clarkei Oldfield & F.Muell., is an erect shrub widespread in West and South Australia, and also found in the extreme south western region of the Northern Territory [1,11].In a previous study, the identification of cembrene diterpenoids in E. clarkei was reported [12], but no pharmacological studies of this species have been reported.
Type 2 diabetes (T2D), which accounts for 90-95% of all diabetes cases [13], is characterized by chronic hyperglycaemia and is mainly caused by a lowered response to insulin (insulin resistance) and/or an inadequate insulin secretion by the β-pancreatic cells [14,15].Protein tyrosine phosphatases are a large family of enzymes playing an important role in cellular signaling by regulating the phosphorylation of target proteins [16].Protein tyrosine phosphatase 1B (PTP1B) is considered a promising T2D target due to its function as a negative regulator of the insulin response.In the insulin signaling pathway, PTP1B dephosphorylates the insulin receptor and insulin receptor substrate, which results in decreased insulin sensitivity [17].Even though a large number of PTP1B inhibitors have been reported, either from natural sources or through synthesis, no clinically approved drugs targeting PTP1B are available.The major challenges for development of PTP1B inhibitors as T2D drugs seem to be associated with selectivity and bioavailability [18,19].
Natural product-based drug discovery is strongly dependent on analytical techniques to pinpoint bioactive constituents in complex extracts, and high-resolution inhibition profiling has proven successful for this purpose.Thus, the eluate from analytical-scale HPLC is microfractionated into one or more 96-well microplates, followed by in vitro assaying of the dried contents in each well.The bioactivity of each well, calculated as percentage inhibition, is plotted against the corresponding retention time, and visually presented in the form of a high-resolution inhibition profile overlaid with UV and/or MS traces from the HPLC separation.This allows easy pinpointing of chromatographic peaks correlated with in vitro bioactivity.When combined with e.g., HPLC-PDA-HRMS or HPLC-PDA-HRMS-SPE-NMR analysis, the time used for structural identification of bioactive constituents in complex extracts is dramatically reduced.This bioanalytical platform has been successfully used for accelerated identification of different classes of α-glucosidase inhibitors [9,10,[20][21][22]24], α-amylase inhibitors [22,23], aldose reductase inhibitors [24], PTP1B inhibitors [8][9][10][20][21][22]25], and DPP-IV inhibitors [26] from complex plant extracts.
In the present study, high-resolution PTP1B inhibition profiling combined with HPLC-PDA-HRMS and NMR spectroscopy were applied for identification of PTP1B inhibitors in E. clarkei.Eight previously undescribed and one known decipiene diterpenoids, one biogenetically related new phenolic acid, as well as ten known O-methylated flavonoids were identified.Among all these compounds, the flavonoid hispidulin and four new decipiene diterpenoids exhibited promising PTP1B inhibitory activity.This is the first study reporting PTP1B inhibitory activity of decipiene diterpenoids.Structure-activity relationship (SAR) of all tested compounds are discussed.Plausible enzymatic as well as photochemically driven steps involved in decipiene metabolism are presented.

General experimental procedures
Analytical-scale HPLC experiments were conducted using an Agilent 1200 series instrument (Agilent Technologies, Santa Clara, CA, USA) consisting of a G1367C high-performance autosampler, a G1311A quaternary pump, a G1322A degasser, a G1316A thermostatted column compartment, a G1315C photodiode array detector, and a G1364C fraction collector, all controlled by Agilent ChemStation version B.03.02 software.A flow rate of 0.5 mL/min and a temperature of 40 • C were maintained for all experiments.HPLC-PDA-HRMS experiments were performed using an Agilent 1260 HPLC system (Agilent Technologies, Santa Clara, CA, USA) consisting of a G1311B quaternary pump with a built-in degasser, a G1329B autosampler, a G1316A thermostatted column compartment, and a G1315D photodiode array detector, coupled with a Bruker micrOTOF-Q II mass spectrometer (Bruker Daltonik, Bremen, Germany) equipped with an electrospray ionization source.NMR experiments were performed at 300 K using a Bruker Avance III system ( 1 H and 13 C operating frequency of 600.13 and 150.90 MHz, respectively) equipped with a Bruker SampleJet and a cryogenically cooled gradient inverse triple-resonance 1.7 mm TCI probe-head (Bruker Biospin, Karlsruhe, Germany).Experimental ECD spectra were recorded using a JASCO J-1500 CD spectrometer (JASCO, Tokyo, Japan).

Plant material
Leaves of Eremophila clarkei were collected on the Summit Walk, Kennedy Range National Park, Western Australia (24 40 19 S; 115 10 E) in July 2017.A voucher specimen was lodged at the University of Melbourne Herbarium, Melbourne, Victoria, Australia (accession number MELUD127857a).

Extraction
To extract leaf resins, leaf material (18.19 g) was submerged in mL of acetonitrile and shaken for 10 min (Ratek, Knox City, Victoria, Australia).The extract was filtered using a glass funnel and dried in vacuo at 40 • C using a rotary evaporator (IKA RV10).The dry extract (2.37 g) was transferred to amber-colored vials using a small volume of methanol and dried under nitrogen gas, then stored at − 20 • C.

PTP1B inhibition assay
The PTP1B inhibition assay was performed as described before [20].In short, the test substance in each well of a 96-well microplate was dissolved in 18 μL of DMSO, and diluted to a total volume of 180 μL by adding 52 μL of EDTA solution (3.4 mM in buffer), and 60 μL of substrate solution (1.5 mM p-NPP and 6 mM DTT in buffer).The buffer was prepared by mixing 50 mM Tris, 50 mM Bis-Tris and 100 mM NaCl in Milli-Q water, followed by adjustment of pH value to 7.0 with acetic acid.The microplate was incubated at 25 • C for 10 min before initiation of enzymatic reaction by adding 50 μL of 0.001 μg/μL PTP1B stock solution to each well.The absorbance of cleavage product p-nitrophenol produced in each well was measured at 405 nm every 30 s for 10 min to obtain the enzyme activity expressed as ΔAU/min.A Thermo Scientific Multiskan FC microplate reader (Thermo Scientific, Waltham, MA, USA) with a built-in incubator was used for incubation and absorbance measurement, controlled by SkanIt ver.2.5.1 software.DMSO was used as a blank sample and RK-682 as a reference compound.All measurements were performed in triplicate.The PTP1B inhibition was calculated using equation 1:

α-Glucosidase inhibition assay
The α-glucosidase inhibition assay was performed as described before [20].In short, the test substance was dissolved in 10 μL of DMSO in each well of a 96-well microplate, and 90 μL buffer and 80 μL C. Liang et al. α-glucosidase enzyme solution (2.0 U/mL in buffer) was added.The buffer was prepared by mixing 34 mM NaH 2 PO 4 ⋅2H 2 O, 66 mM Na 2 HPO 4 and 0.02% NaN 3 in Milli-Q water, followed by adjustment of pH value to 7.5 with NaOH.The microplate was incubated at 28 • C for 10 min before initiating the enzymatic reaction by adding 20 μL p-NPG solution (10 mM in buffer) to give a total volume of 200 μL in each well.The absorbance was measured at 405 nm every 30 sec for 35 min to obtain the enzyme activity expressed as ΔAU/s.The above-mentioned microplate reader was used for incubation and absorbance measurement, and all measurements were performed in triplicates.Percentage α-glucosidase inhibition was calculated using the same equation as for the PTP1B inhibition assay.

α-Amylase inhibition assay
The α-amylase inhibition assay was performed as described before [22].Similar to the procedure of the α-glucosidase inhibition assay, the test substance was dissolved in 20 μL of DMSO in each well of a 96-well microplate, and diluted in 80 μL buffer and 80 μL α-amylase enzyme solution (2.0 U/mL in buffer).The buffer (0.1 M) was prepared by mixing NaH 2 PO 4 ⋅2H 2 O, Na 2 HPO 4 , 60 mM NaCl and 0.02% NaN 3 in Milli-Q water, followed by adjustment of pH value to 6.0 with phosphoric acid and addition of 1 mM calcium acetate.The microplate was incubated at 37 • C for 10 min before initiation of enzymatic reaction by adding 20 μL CNP-G3 solution (10 mM in buffer) to give a total volume of 200 μL in each well.The absorbance was measured at 405 nm every 3 min for 30 min on the above-mentioned microplate reader.All measurements were performed in triplicate.Percentage α-amylase inhibition was calculated using the same equation described before.

Microfractionation and high-resolution PTP1B inhibition profiling
The acetonitrile extract of E. clarkei was microfractionated using the above-mentioned Agilent 1200 analytical-scale HPLC.The flow rate was 0.5 mL/min and the temperature was maintained at 40 • C. The mobile phase for the HPLC separation was a mixture of solvent A (acetonitrile-water-formic acid, 5:94.9:0.1, v/v/v) and solvent B (acetonitrile-water-formic acid, 94.9:5:0.1,v/v/v).For the high-resolution PTP1B inhibition profiling, 10 μL acetonitrile solution of crude leaf extract (50 mg/mL) was separated on a reversed-phase Phenomenex Luna C 18 (2) column (150 × 4.6 mm i.d., 3 μm particle size, 100 Å pore size) based on the following gradient: 0 min, 25% B; 30 min 100% B; 38 min, 100% B; 39 min, 25% B; 46 min, 25% B. The eluate from 5 to 35 min was fractionated into 88 wells of one 96-well microplate (resolution of 2.93 data points per minute), and the sampled eluate was evaporated to dryness using a Savant SPD121P speed vacuum concentrator equipped with an OFP400 oil-free pump and an RVT400 refrigerated vapor trap (Holbrook, NY, USA).The dried material in the microplate was subsequently assayed for PTP1B inhibitory activity, providing the percentage inhibition of each well, which was plotted against the corresponding retention time of the HPLC chromatogram to create the highresolution PTP1B inhibition profile.

Dose-dependent effect determination
Dose-dependent effects of PTP1B inhibition were determined according to the procedure described above.Test substances were dissolved in DMSO and two-fold serial dilutions for this stock solution were assayed for PTP1B inhibitory activity.The dilution series started from 2 mg mL − 1 and 2 mM for the crude extract and pure compounds, respectively, and all measurements were performed in triplicate.IC 50 values were calculated on the basis of equation 2 below using GraFit software, version 5.0.11(Erithacus Software Limited): ) slope (2) where min is the minimum and max the maximum concentrations, x is the concentration of the test sample and slope is the Hill slope.Results were expressed as IC 50 mean value ± standard error.

HPLC-PDA-HRMS experiments
High-resolution MS data were acquired in negative-ion mode with a capillary voltage of 3500 V, a drying temperature of 200 • C, a nebulizer pressure of 2.0 bar and a drying gas flow of 7 L/min.A solution of sodium formate clusters was injected at the beginning of the analysis to enable internal mass calibration.Chromatographic separations were conducted using the same column and conditions as used for the analytical-scale HPLC separation.

NMR experiments
NMR data of all isolated compounds were acquired at 300 K in methanol-d 4 , and the residual solvent signals (δ H 3.31 ppm and δ C 49.0 ppm) were used for chemical shift calibration. 1H NMR spectra were recorded with 30 • pulses, acquisition time of 2.72 s, relaxation delay of 1.0 s, spectral width of 20 ppm and 64 k data points. 13C NMR spectra were recorded with 30 • pulses, acquisition time of 0.90 s, relaxation delay of 2.0 s, spectral width of 240 ppm and 64 k data points.All 2D homo-and heteronuclear spectra were recorded with 12 ppm spectral width for 1 H, and 170 ppm (multiplicity edited HSQC) or 240 ppm (HMBC) for 13 C, and with 2048 (DQF-COSY, ROESY and HBMC) or 1730 (multiplicity edited HSQC) data points in the direct dimension and 512 (DQF-COSY and HMBC) or 256 (ROESY and multiplicity edited HSQC) data points in the indirect dimension.HSQC and HBMC spectra were both acquired with relaxation delay of 1.0 s.All 2D NMR data were zero filled to 1 k and 2 k data points in F1 and F2, respectively, with forward linear prediction employed in F1.Icon NMR (version 4.2, Bruker Biospin, Karlsruhe, Germany) was used for controlling automated NMR data acquisition, and Topspin (version 4.0.6,Bruker Biospin) for NMR data processing.

Calculation and acquisition of ECD spectra
The model structure of 11b was subjected to MacroModel interfaced to Schrödinger Maestro 11.9 (Schrödinger, LLC, USA) for conformational search with the Molecular Merck force field static (MMFFs) in gas phase [27].Geometry optimization for conformers showing relative energy within 5.02 kcal/mol were performed using the density functional theory (DFT) method with the B3LYP functional and the 6-31G(d, p) basis set of ground state [28,29].Polarizable continuum model in the integral equation formalism (IEF-PCM) was included for assessment of the acetonitrile solvent effect [30].The DFT and TDDFT calculations were conducted for lowest-energy conformers accounting for more than 1% Boltzmann distribution with the Gaussian 16 program package [31].Oscillator strengths and rotatory strengths of the first 60 excited states were calculated at TDDFT CAM-B3LYP/6-31G (d,p) level for each conformer, using IEF-PCM for acetonitrile solvent effect [32].The overall calculated ECD spectra were created after Boltzmann averaging the calculated spectra of each conformer, which was exported using SpecDis software ver.1.71 (Berlin, Germany) [33] with a halfbandwidth of 0.30 eV.The calculated ECD spectrum was redshifted by 9 nm.To evaluate the accuracy of absolute configuration assignment, we employed SpecDis to calculate the enantiomeric similarity index (Δ ESI ) [34], which facilitated a quantitative comparison of the calculated and experimental ECD spectra.
Experimental ECD spectra of all identified decipiene diterpenoids and ester derivatives were recorded in acetonitrile using a quartz cuvette with 1 cm path length.All measurements were conducted at 25 • C under constant nitrogen gas flow according to the following parameters: wavelength 190-400 nm, scan speed 50 nm/min, digital integration time 4 s, bandwidth 5.0 mm, and 3 accumulations.

Enzyme kinetics of PTP1B inhibition
To determine the inhibition mode of active ester derivatives of the decipiene diterpenoids, i.e., 13a, 13b, 13f and 14b against the PTP1B enzyme, various concentrations of each compound were evaluated for their PTP1B inhibition at three different concentrations of p-NPP (0.5, 1, and 2 mM), respectively.The kinetic data were analyzed by GraphPad Prism version 9.5.0 (GraphPad Software, Inc.) to be graphically interpreted as Lineweaver-Burk plot, from which the mode of inhibition was determined.Kinetic parameters were calculated by fitting the data to the Michaelis-Menten equation using non-linear regression analysis (GraphPad Prism version 9.5.0,GraphPad Software, Inc.).

Molecular docking simulation
Molecular docking was performed using Maestro version 13.0 software (Schrödinger, LLC).The 2D structures of the docked compounds were prepared using the 2D sketcher tool available in Maestro, while the 3D structures were generated using the LigPrep module.The X-ray structure of PTP1B in complex with the catalytic inhibitor 3-({5-[(N-acetyl-3-{4-[(carboxycarbonyl)(2-carboxyphenyl)amino]-1-naphthyl}-L-alanyl)amino]pentyl}oxy)-2-naphthoic acid (compound C, PDB ID: 1NNY [35]) at a resolution of 2.40 Å was used as reference to investigate the ligand-protein interaction at the catalytic binding site.The cocrystallized structure of PTP1B with an allosteric inhibitor 3-(3,5dibromo-4-hydroxy-benzoyl)-2-ethyl-benzofuran-6-sulfonic acid (4-sulfamoyl-phenyl)-amide (Compound A, PDB ID: 1T49 [36]) at a resolution of 1.90 Å was used as reference to explore the ligand-protein interaction at the allosteric binding site.Protein receptors were prepared using the Protein Preparation module with default settings and all water molecules removed.The docking grid was created in the Receptor Grid Generation module and defined as an enclosed box centered at the cocrystallized ligand's centroid and sized to match the ligand.Ligand docking and scoring were performed using Glide with default settings and with the SP (standard precision) scoring function.The figures were generated using PyMOL (2.5.2) and Discovery Studio Visualizer v21.1.0.20298.

Antidiabetic activity screening for acetonitrile extract of E. clarkei
As a part of our continued endeavor to investigate chemical constituents in and pharmacological properties of Eremophila spp., the crude acetonitrile extract of E. clarkei leaves was assessed for inhibitory activity against PTP1B, α-glucosidase, and α-amylase at a concentration of 50 μg/mL.The crude extract showed strong inhibitory activity against PTP1B with 82.2 ± 2.8 % inhibition, but no inhibitory activity against α-glucosidase and α-amylase with 3.2 ± 0.8 % and 1.0 ± 0.4 % inhibition, respectively.The dose-dependent inhibitory effect of the extract was therefore assessed for PTP1B, which resulted in an IC 50 value of 33.0 ± 0.8 μg/mL (Table 1 and Fig. S1).

High-resolution PTP1B inhibition profiling for acetonitrile extract of E. clarkei
A 46-min analytical-scale HPLC method provided satisfactory separation of all major constituents in the crude extract (Fig. 1), and the eluate from 5 to 35 min was microfractionated at a resolution of 2.93 data points per min.Following evaporation, the dried content in each well was tested for PTP1B inhibitory activity, which provided a highresolution PTP1B inhibition profile (Fig. 1).The biochromatogram shows that the material eluted with peak 12 is correlated with strong PTP1B inhibition (98.3%), the material eluted with peaks 13a-13f (indicated as fraction 13) and peaks 14b and 14c (indicated as fraction 14) are correlated with moderate to strong PTP1B inhibition (75% − 95%), and the material eluted with peak 11 is correlated with weak PTP1B inhibition (49.2%).

Plausible enzymatic as well as photochemically driven steps involved in decipiene metabolism
Although more than 200 structurally diverse diterpenoids have been identified from different Eremophila species [2], investigation of their biosynthesis is scarce.The single previous biosynthetic study available reported (Z,Z,Z)-nerylneryl pyrophosphate (NNPP) as the non-canonical terpene precursor of serrulatane, viscidane, and cembrane diterpenoids in E. lucida, E. drummondii and E. denticulata [45].Structural analyses of acyclic diterpenes present in E. glutinosa and E. exilifolia also supported their biosynthesis from NNPP [46,47].As first proposed by Ghisalberti, the biosynthesis of decipiene diterpenes may involve a cationic intermediate equivalent to the bisabolyl cation crucial in the construction of an array of diverse and complex architectures found in sesquiterpenoids [46,48].In contrast to the above-mentioned classes of diterpenoids, which are all made with (Z,Z,Z)-NNPP as precursor, the bisabolyl-type sesquiterpenoid cation is built with (E,E)-farnesyldiphosphate [48].In biosynthesis of the decipiene diterpenoids, the equivalent C 20 precursor Scheme 1. A. Plausible biosynthetic route for formation of the decipiene core skeleton.B. Plausible metabolic conversion of 1 into 2a.
C. Liang et al. would therefore be (E,E,E)-geranylgeranyl pyrophosphate (GGPP).The bisabolyl-type diterpenoid cation formed would give rise to a bicyclic intermediate and a [2+2] Diels-Alder type reaction would generate the decipiene core structure (Scheme 1A).None of the biosynthetic enzymes involved in core structure formation, including diterpene synthases and possibly a [2+2] Diels-Alderase, have been identified so far.Likewise, cytochrome P450s, flavine monooxygenases or α-ketoglutarate dependent dioxygenases involved in oxygenation of the decipiene core structure as well as the acyl-transferases further expanding the decoration pattern all remain unidentified.However, in this context, it should be noted, that many cyclobutane rings in natural products are thought to arise from photochemical reactions [49,50].Insights from quantum chemistry calculations may serve to more precisely characterize the specific nature of the cyclization reaction giving rise to the tricyclic decipiene core structure [49].
In 1981, Croft et al. demonstrated experimentally, that a decipiene diterpenoid isolated from E. decipiens could be converted into a bicyclic compound by UV irradiation [46,51].In a similar photochemical process, eremoclarkic acid (2a) isolated from E. clarkei in this study, may be considered derived from 1 in a retro [2+2] Diels-Alder type reaction accompanied by enolization and aromatization (Scheme 1B).Whether these latter reactions proceed non-enzymatically in the oxidative and light-exposed environment of the trichome cells at the surface of the E. clarkei leaves or whether the process is catalyzed by a retro [2+2] Diels-Alderase remain to be elucidated.
Enzyme catalyzed Diels-Alder reactions have been shown to be involved in the biosynthesis of several natural products [52,53].The characterized Diels-Alderases involved were catalyzing 4+2 cycloadditions.However, a cyclase that catalyzes competing [2+2] and [4+2] cycloadditions has recently been characterized [54].Directed engineering enabled a shift in the catalytic activity of the cyclase towards [2+2] activity.The reverse process, a retro-[2+2] cycloaddition, has also been shown to be enzyme catalyzed [55].It thus remains to be established whether the formation of the tricyclic decipiene ring system and its further metabolism proceeds as photochemical or enzyme catalyzed reactions.

Structure-activity relationship of isolated compounds
All metabolites isolated from E. clarkei were evaluated for PTP1B inhibitory activity, initially at a concentration of 100 μM, and the percentage inhibition can be seen in Table 5.For compounds displaying more than 50% inhibition at this concentration, i.e., 2b, 4, 12, 13a, 13b, 13f and 14b, dose-dependent effects were determined, and the results are given as IC 50 values in Table 5, with corresponding dose-dependent curves provided in Fig. S63.
Among all identified O-methylated flavonoids in this study, hispidulin (2b) showed the strongest inhibition against PTP1B with an IC 50 value of 24.8 ± 1.9 μM.Hispidulin has previously been reported as a PTP1B inhibitor with IC 50 > 33 μM [56].Other O-methylated flavonoids showed much weaker PTP1B inhibition than 2b.Comparison of 3 and 2b or 6 and 4, show, that the inhibitory activity decreases significantly when a methoxy group is introduced at C-3.This trend was also observed in some other O-methylated flavonoids, including 5, 7, 8, 9, and 11a.Interestingly, 12 which harbors a methoxy group at C-3 displayed similar inhibitory activity to 4, indicating that the introduction of methoxy groups to C-3′, C-4′ and C-5′ might compensate for the decline in activity due to the presence of methoxy group at C-3.Comparison of 4 and 2b shows, that the inhibitory activity decreases approximately by a factor of two when a methoxy group is added at C-3′.Comparison of 7 and 8 or 11a and 12 demonstrates that absence of the methoxy group at C-6 might lead to decreased inhibitory activity.
The decipiene diterpenoids 1, 10, and 11b were inactive against PTP1B.However, the ester derivatives of decipiene diterpenoids, i.e., 13a, 13b, 13f, and 14b showed PTP1B inhibitory activity with IC 50 values ranging from 22.8 ± 2.3 to 33.6 ± 8.8 μM as seen in Table 5.This indicates that the trans-p-coumaroyl unit in 13a and 13b, the trans-pisoferuloyl in 13f, and the trans-p-cinnamoyl in 14b might contribute to the increased inhibitory activity.Surprisingly, the cis-p-coumaroyl decipiene diterpenoid 13d showed five times weaker PTP1B inhibitory activity with 20.3 ± 2.5 % inhibition at 100 μM compared to the trans-pcoumaroyl analogue 13b with 103.9 ± 3.0 % inhibition at the same concentration.This implies that change of configuration of the Δ 2′,3′ double bond of the p-coumaroyl unit from trans to cis might result in decreased inhibitory activity.This is supported by a previous study which reported weaker PTP1B inhibitory activity of the triterpenoid 3β-O-cis-p-coumaroyl-2α-hydroxy-urs-12-en-28-oic acid containing a cis-pcoumaroyl unit (IC 50 = 26.67 ± 0.35 µM) than its trans-isomer jacoumaric acid (IC 50 = 11.93 ± 0.31 µM) [21].In a recent study, we showed that serrulatane ferulate esters reveal a saddle-shape-like binding in the active site of PTP1B, and that a cis-ferulate ester side chain forces the compound to adopt a less favourable conformation than a trans-ferulate ester side chain [57].Additionally, 13c exhibited stronger PTP1B inhibitory activity with 10.3 ± 1.5 % inhibition at 100 μM compared to its non-acetylated analogue 11b with − 1.3 ± 3.1 % inhibition at the same concentration, which supports that acetylation of the decipiene scaffold might positively impact the PTP1B inhibitory activity.

Enzyme kinetics analysis of PTP1B inhibition
The enzyme kinetics analysis of PTP1B inhibition was performed at different concentrations of the p-NPP substrate in the absence and presence of various concentrations of investigated inhibitors.The obtained enzyme kinetics data were analyzed by Lineweaver-Burk plot to determine the mode-of-action for PTP1B inhibition [21].The results reveal that 13a and 13b are competitive inhibitors, because all straight lines in Fig. 6A and 6B cross the same y-intercept.Thus, as inhibitor concentration increased, K m increased, and V max remained unchanged, and inhibition constant values (K i ) of 21.3 and 24.9 μM (Table 5) were obtained for 13a and 13b, respectively.The straight lines in both Lineweaver-Burk plot of 13f and 14b (Fig. 6C and 6D) intersect above xand to the left of y-axis, indicating mixed-mode inhibition (as inhibitor concentration increased, K m increased whereas V max decreased).These two inhibitors showed K i values of 22.3 and 36.6 μM (Table 5), respectively.

Molecular docking simulation of PTP1B inhibition
The active site of PTP1B is formed by a deep and solvent-accessible pocket that is surrounded by several loops within its catalytic domain, which are crucial for substrate recognition and enzyme activity.These mainly include WPD loop (Thr177-Pro185), P loop (His214-Arg221), Q loop (Gln262), R loop (Val113-Ser118), S loop (Ser201-Gly209) and pTyr loop (Tyr46-Val49) [58][59][60].Moreover, the enzymatic activity of PTP1B can be regulated by inducing conformational changes through the binding of ligands to its secondary and allosteric binding sites.The secondary site, also called the second aryl phosphate-binding site, which is composed of key residues Arg24 and Arg254, is an important nonconserved site that regulates the substrate specificity [59,60].The allosteric binding site includes the α3 helix (Glu186-Glu200), α6 helix (Ala264-Ile281), and α7 helix (Val287-Ser295) [36,59].
The enzyme kinetics studies, vide ultra, revealed 13a and 13b to be competitive inhibitors, and molecular docking simulations were therefore conducted to investigate the interaction of these two with the catalytic binding site of PTP1B.In general, 13a and 13b displayed a significant binding overlap with compound C, a potent and selective PTP1B inhibitor discovered by Szczepankiewicz et al. using linkedfragment strategy [35], in the catalytic site of PTP1B.The binding of compound C with PTP1B extends from the active site to the secondary site, while 13a and 13b only occupy the active site (Fig. 7A and 7B).The carboxylate group of the tricyclo[5.3.1.0 5,11]undecane ring of 13a formed hydrogen bonds with the backbone of residues S216, A217, G218, I219, and G220, and with the side chain of residue C215 at the active site.Moreover, a hydrogen bond interaction was observed between the carbonyl oxygen of the ester in 13a and the catalytic residue T263.The terminal phenyl ring in 13a showed a T-shaped π-π interaction with residue F182 in the WPD loop.In addition, the decipiene scaffold of 13a was involved in hydrophobic interactions with surrounding residues Y46, W179, A217, and I219 (Fig. 7D).Similarly, several hydrogen bond interactions were formed between the carboxylate group of the tricyclo[5.3.1.0 5,11]undecane ring of 13b and residues C215, S216, A217, G218, G220, and R221 of the active site.Several hydrophobic interactions were also observed between the tricyclo [5.3.1.0 5,11]undecane ring in 13b and the surrounding residues Y46, A217, and I219 (Fig. 7E).
Since 13f and 14b were found to be mixed-type inhibitors, vide ultra, docking simulations for these two compounds were performed for both the catalytic and the allosteric binding sites.At the catalytic site, 13f (Fig. 7C and 7F) and 14b (Fig. 7G and 7J) displayed binding conformations and interactions with PTP1B similar to those seen for 13a.Interestingly, the phenolic hydroxy group on the terminal phenyl ring of 13f formed a hydrogen bond with ASP265 (Fig. 7F).This may explain the differences in inhibitory activities among 13c-f, which share the same core skeleton but have structurally different side chains.Compound 13c lacks the terminal phenyl ring and corresponding hydrogen bond as seen for 13f, whereas 13d possesses a terminal cis-4-hydroxycinnamoyl unit.According to the docking pose of 13f, the cis-4hydroxycinnamoyl unit is likely to cause a steric clash with the binding site, and thereby impairing the binding of 13d to PTP1B.At the allosteric site, 13f and 14b exhibit substantial overlap with compound A, a novel and selective PTP1B allosteric inhibitor reported by Wiesmann et al [36], by forming similar interactions with important residues in the α3 and α6 helixes (Fig. 7H and 7I).The carboxylate group of the tricyclo [5.3.1.0 5,11]undecane ring of 13f formed a salt bridge with side chain residue K197 (α3 helix) and a hydrogen bond with residue N193 (α3 helix).The hydroxymethyl group of the tricyclo[5.3.1.0 5,11]undecane ring of 13f showed a hydrogen bond interaction with the backbone residues E200 (α3 helix) and K197 (Fig. 7K).As illustrated in Fig. 7L, 14b binds to the allosteric site of PTP1B in a similar manner as 13f.The ligand interactions of 14b at the allosteric site include a salt bridge to the carboxylate group of residue K197, and hydrogen bonds to residues N193 and K197, as well as hydrogen bonds from the hydroxymethyl  group to residues K197 and E200.Moreover, a hydrogen bond interaction was observed between the carbonyl oxygen of the ferulate ester and side chain residue N193.The tricyclo[5.3.1.0 5,11]undecane ring, the long alkyl chain, and the terminal phenyl ring of both 13f and 14b formed hydrophobic interactions with residues A189 and F196 in the α3 helix and with F280 in the α6 helix (Fig. 7K and 7L).This may also explain the low inhibitory activities of 13c and 13d compared to 13f, as the absence of a hydroxycinnamoyl unit in 13c and the cis-configuration of the 4-hydroxycinnamoyl unit in 13d may impair these hydrophobic interactions in allosteric sites.
In general, for the strongest PTP1B-inhibiting decipiene esters 13a, 13b, 13f, and 14b, the core decipiene skeleton was found to form multiple hydrogen bonds and hydrophobic interactions within the catalytic and allosteric binding sites.These interactions are considered fundamental in the protein-ligand interaction, and therefore might account for the inhibitory activity of these compounds.The docking results were in accordance with results from the enzyme kinetics analysis and demonstrated substantial binding capacity of 13a, 13b, 13f, and 14b to the catalytic and allosteric site of PTP1B enzyme, indicating the ester derivatives of decipiene diterpenoids to be a new class of naturalproducts-based compounds for development of PTP1B inhibitors.

Conclusion
In conclusion, high-resolution PTP1B inhibition profiling combined with repeated analytical-scale HPLC separation led to identification of eight previously undescribed decipiene diterpenoids and ester derivatives thereof as well as a biogenetically related new phenolic acid, from an acetonitrile extract of E. clarkei.In addition, one known decipiene diterpenoid and ten known O-methylated flavonoids were identified.Among all identified metabolites, the O-methylated flavonoid hispidulin (2b) and ester derivatives of decipiene diterpenoids harboring a trans-p-coumaroyl (13a and 13b), a trans-p-isoferuloyl (13f) or a trans-p-cinnamoyl (14b) moiety showed PTP1B inhibitory activity.The enzyme kinetics studies indicated that 13a and 13b were competitive PTP1B inhibitors, while 13f and 14b showed mixed-type PTP1B inhibition.The molecular docking simulation revealed comparable binding affinity of 13a, 13b, 13f, and 14b towards the PTP1B enzyme, supporting the results from in vitro PTP1B inhibition assay.These are the first decipiene ester derivatives identified and the first study to report this class of compounds as potential PTP1B inhibitors.This expands the diterpenoid chemical space from the genus Eremophila and may provide promising leads for development of future therapeutic agents for T2D.

Declaration of Competing Interest
The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: Dan Staerk reports financial support was provided by Novo Nordisk Foundation.

Fig. 4 .
Fig. 4. Comparison of the experimental ECD spectrum of 11b and the calculated spectrum of the 11b model compound without the long side chain.
a 13 C NMR data assigned by HSQC and HMBC experiments.b Multiplicities reported as apparent splittings: s = singlet, d = doublet, t = triplet, q = quartet, m = multiplet, br = broad.c Multiplicities undetermined due to overlapping signals.d Overlapped with solvent signal.

Table 5
Inhibitory activity of identified compounds against PTP1B.
a Results reported as mean ± standard error of three independent measurements.bPositivecontrol.C.Liang et al.