Serrulatane diterpenoids with unusual side chain modifications from root bark of Eremophila longifolia

The plant genus Eremophila is endemic to Australia and widespread in arid regions. Root bark extract of Ere-mophila longifolia (R.Br.) F.Muell. (Scrophulariaceae) was investigated by LC-PDA-HRMS, and dereplication suggested the presence of a series of diterpenoids. Using a combination of preparative-and analytical-scale HPLC separation as well as extensive 1D and 2D NMR analysis, the structures of 12 hitherto unreported serrulatane diterpenoids, eremolongine A-L, were established. These structures included serrulatanes with unusual side chain modifications to form hitherto unseen skeletons with, e.g., cyclopentane, oxepane, and bicyclic hexahydro-1 H - cyclopenta[ c ]furan moieties. Serrulatane diterpenoids in Eremophila have recently been shown to originate from a common biosynthetic precursor with conserved stereochemical configuration, and this was used for tentative assignment of the relative and absolute configuration of the isolated compounds. Triple high-resolution α -glucosidase/ α -amylase/PTP1B inhibition profiling demonstrated that several of the eremolongines had weak inhibitory activity towards targets important for management of type 2 diabetes.

Decoctions of leaves, twigs, and bark of Eremophila longifolia (R.Br.)F.Muell.(Scrophulariaceae) have been used by some Australian Aboriginal peoples to cure headaches and to heal sores, and an infusion of leaves has been applied to sore eyes and to treat insomnia (Richmond and Ghisalberti, 1994).No published studies have characterised diterpenoids from E. longifolia, with previous studies focussed on the volatile compounds found in the leaves.Several essential oil chemotypes including types producing predominantly isomenthone/menthone, karahanaenone or safrole/methyl eugenol have been shown (Sadgrove and Jones, 2014;Sadgrove et al., 2021).The iridoid glycoside geniposidic acid has been isolated as a cardioactive component (Pennacchio et al., 1996) and the antimicrobial compound (− )-genifuranal is produced in the heated or smoked leaves, reflecting their use as a smoking treatment in traditional medicine (Sadgrove et al., 2014).Neryl ferulate and neryl p-coumarate have also been isolated as antimicrobial components (Galappathie et al., 2017).The root chemistry of this species has not previously been investigated.
High-resolution inhibition profiling is a technique used for identification of bioactive constituents present in complex extracts with activity against therapeutic targets.The crude extract is microfractionated into one or more 96-well microplates using analytical-scale HPLC, and the dried content in all wells are assessed for inhibitory activity against the selected target.The results, expressed as percentage inhibition, are plotted against the average retention time of the constituents in each well to create a high-resolution inhibition profile.The microfractionation can be repeated and the constituents assayed with other target enzymes.In this way, our laboratory on a routine basis performs single- (Li et al., 2019), dual- (Wubshet et al., 2016), triple- (Tahtah et al., 2015), or quadruple (Zhao et al., 2018) high-resolution inhibition profiling with α-glucosidase, α-amylase, PTP1B, aldose reductase and/or radical scavenging -all related to T2D.Peaks in the high-resolution inhibition profile(s) allow identification of HPLC-separated constituents potentially correlated with bioactivity.These may afterward be isolated and structurally elucidated using fully hyphenated analytical techniques like high-performance liquid chromatography -photodiode array detection -high-resolution mass spectrometry -nuclear magnetic resonance spectroscopy, i.e.HPLC-PDA-HRMS and NMR (Lima et al., 2017;Zhao et al., 2019b).
Diabetes is a multifactorial metabolic disease with a global prevalence of 537 million incidences in 2021, which is estimated to increase to 783 million incidences by 2045 (Sun et al., 2022).Type 2 diabetes (T2D) accounts for over 90% of all diabetes cases worldwide, and physical inactivity and obesity are the primary risk factors for the development of T2D (International Diabetes Federation, 2021;World Health Organization, 2021).T2D is characterised by resistance to insulin action and/or reduced pancreatic insulin production, and current treatment constitute a combination of lifestyle changes and antihyperglycemic drugs to control blood glucose levels (Chatterjee et al., 2017;International Diabetes Federation, 2021).α-Glucosidase and α-amylase inhibitors are important antihyperglycemic drugs that decrease the cleavage-and absorption rate of carbohydrates in the small intestine.Thus, α-amylase is an enzyme that catalyses the cleavage of internal α-1,4-glycosidic bonds of e.g., starch into di-and trisaccharides (Hanhineva et al., 2010).These saccharides are substrates for α-glucosidase, which catalyses the cleavage of terminal non-reducing (1 → 4)-linked α-glucose residues into monosaccharides, which are absorbed into the bloodstream.Inhibition of α-amylase and α-glucosidase slows the elevation of blood glucose after a carbohydrate-rich meal (Hanhineva et al., 2010;Kumar et al., 2011).Another way to control blood glucose levels is through protein-tyrosine phosphatase 1B (PTP1B), a negative intracellular regulator of the insulin signalling pathway.Inhibition of PTP1B therefore prolongs the action of insulin in the targeted cells (Egawa et al., 2001;Kerru et al., 2018).No PTP1B inhibitors have yet been approved for clinical use due to low selectivity and undesirable side effects (Dowarah and Singh, 2020).
Previous studies have shown that extracts from aerial parts of Eremophila are rich sources of serrulatane diterpenoids and other compounds, some of which have α-glucosidase and/or PTP1B inhibitory activity (Tahtah et al., 2016;Wubshet et al., 2016;Zhao et al., 2019a), however the roots of Eremophila species have been less extensively investigated.Thus, as part of our ongoing search for new bioactive natural products with the potential to treat T2D, a crude root bark extract of E. longifolia was investigated using a combination of triple high-resolution α-glucosidase/α-amylase/PTP1B inhibition profiling, HPLC-PDA-HRMS, and NMR.

Triple high-resolution α-glucoside/α-amylase/PTP1B inhibition profiling of crude root bark extract
An analytical-scale reversed-phase HPLC method with a multi-step gradient profile of 68 min was developed, and the eluate obtained from 8 to 60 min after injection of 10 μL crude root extract (50 mg/mL) was microfractionated into 176 wells using 88 wells of two 96-well microplates (leaving 8 for control).This was repeated twice to generate three identical sets of two microplates.After evaporation of the HPLC eluate, the material in each well was assessed for α-glucosidase, α-amylase, and PTP1B inhibitory activity.The results, expressed as percentage inhibition for each of the enzymes tested, were plotted at the respective retention time for each fraction, yielding a triple highresolution α-glucosidase/α-amylase/PTP1B inhibition profile (Fig. 1).
All three inhibition profiles had a rather low signal-to-noise ratio.However, the constituents eluted with peaks 8, 11, 12, 16, and 17 were correlated with moderate to weak α-glucosidase inhibitory activity, and the constituents eluted with peaks 8, 11, 14, and 15 were correlated with α-amylase inhibitory activity, the latter very weak.The PTP1B inhibition profile suggested that all the constituents eluted with peaks 8-17 may be correlated with PTP1B inhibitory activity, but especially the constituents eluted with peaks 8, 11, 16, and 17 were correlated with moderate to strong inhibitory activity.
The constituent represented by peak 11 showed a [M+H] + ion at m/z 313.1795 (calcd for C 20 H 25 O 3 + , 313.1798, ΔM +1.0 ppm) corresponding to a molecular formula of C 20 H 24 O 3 and a hydrogen deficiency index of 9.The COSY spectrum revealed the same two spin systems as also seen for 5, but with the appearance of a methyl doublet for H 3 -18 instead of the oxygenated H 2 -18 seen in 5.In addition, a H-5 ↔ H 3 -19 spin system (Fig. 3) showed the absence of the tetrahydropyran ring closure seen in 5. Thus, compound 11 was identified as (S)-6-hydroxy-3,8-dimethyl-5-(6-methylhept-5-en-2-yl)naphthalene-1,2-dione (Fig. 2), for which the name eremolongine I is suggested. 1H and 13 C NMR data for 11 are provided in Table 3, HRMS in Supplementary data Figure S49, and 1 H NMR, COSY, HSQC, HMBC, and ROESY data and spectra are provided in Supplementary data Table S10 and Figures S50-S54 were observed in the COSY spectrum of 15, but with an oxygenated methine signal for H-13 at δ 5.02 instead of the methylene seen for H 2 -13 in 11.Both C-3 at δ C 161.4 and C-13 at δ C 69.9 are oxygenated, and based on the molecular formula and the hydrogen deficiency index, the two carbons must share one oxygen as an aryl alkyl ether, thus forming a dihydro-2H-pyran ring system.The chemical shift value of 35.2 ppm for C-12 is furthermore in close agreement with chemical shift values around 37 ppm for carbon atoms neighboring the ether bond in similar structures, whereas chemical shift values of approximately 46 ppm are reported for carbon atoms neighboring a hydroxyl (Georgantea et al., 2014;Green et al., 2011).The ROESY spectrum revealed a correlation between β-positioned H-13 and H 3 -18, and with the arguments for conserved configuration at C-11, vide supra, compound 15 was identified as (1S,3R)-1,6,9-trimethyl-3-(2-methylprop-1-en-1-yl)-2,3-dihydro-1H -benzo[f]chromene-7,8-dione (Fig. 2), for which the name eremolongine K is suggested. 1H and 13 C NMR data for 15 are provided in Table 3, HRMS in Supplementary data Figure S61, and 1 H NMR, COSY, HSQC, HMBC, and ROESY data and spectra are provided in Supplementary data Table S12 and Figures S62-S66.
The ROESY spectrum showed a weak correlation between H-1 and H 3 -16 and a strong correlation between H-14 and H-5, which is in agreement with, and only possible for, the S configuration of C-4 and R configuration of C-14, i.e., with the C-4-C-11 bond being β-oriented.This is the third occurrence of a non-conserved stereochemistry at C-4, and the inversion of configuration at C-4 must proceed, as for 6 and 9, as part of the unusual sidechain modifications observed with the serrulatane diterpenoids isolated from E. longifolia.Compound 17 was thus identified as (1S,3aR,7R,11bS)-1,4,4,7,10-pentamethyl-2,3,3a,4-tetrahydro-1H,7H-cyclopenta[c]naphtho[2,1-b]furan-8,9-diol (Fig. 2), for which the name eremolongine L is suggested. 1H and 13 C NMR data for 17 are provided in Table 3, HRMS in Supplementary data Figure S67, and 1 H NMR, COSY, HSQC, HMBC, and ROESY data and spectra are provided in Supplementary data Table S13 and Figures S68-S72.
modifications of the side chain.This includes ring closures between C-14 and C-18 to form a cyclopentane ring (2 and 14), probably via a nucleophilic attack from the C-14-C-15 double bond to C-18 of a precursor with an aldehyde in this position.The resulting C-15 carbocation can react with hydroxides to form the hydroxyl group at this position (2).The C-15 carbocation may also undergo spontaneous E1 elimination to form 14, which could also be formed via an enzyme-catalysed reaction by a dehydratase acting on 2. The formation of the oxepane ring as part of the hemiacetal in 9 could proceed via a 3α-hydroxy analogue with an oxo group at C-14 (as also seen on the side chain of 1).Compounds 7 and 15 most likely originate from structurally related precursors, with ring closure from C-13 to the hydroxyl group at C-3 to form a tetrahydro-2H-pyran moiety (15) or from C-13 to C-5 to form a cyclohexane moiety (7).The biosynthetic route for formation of the bicyclic hexahydro-1H-cyclopenta[c]furan moiety in 17 must be a multistep process, and whether the C-15-O-3 or the C-4-C-14 bonds form first remains unresolved.The unusual side chain modifications presented here are important, because the side chain not only plays an important role in the pharmacological properties of serrulatanes, but also because these side chain modifications and/or new skeletons represents an otherwise unexplored part of the serrulatane 3D chemical space.

Pharmacological evaluation of isolated compounds
Based on the results from the biochromatogram and the amount of material isolated for each compound, it was decided to determine the inhibitory activities of compounds 7, 8, 11, 16, and 17 against PTP1B.However, none of the compounds provided a full dose-response curve with the tested dilution series (100, 75, 50, 25, 12.5, 5, 1, 0.5, and 0.1 μg/mL), but 11 provided around 45% inhibition at 100 μg/mL (320 μM), 16 provided 58% inhibition at 100 μg/mL (318 μM), and 17 provided around 48% inhibition at 100 μg/mL (318 μM).These compounds are thus weak PTP1B inhibitors, which is in agreement with the moderate PTP1B inhibitory activities observed with peaks 11, 16, and 17 in Fig. 1.Compound 7 and 8 showed no noticeable inhibition at the highest tested concentrations and are considered inactive.

Conclusions
In conclusion, the combined use of triple high-resolution α-glucosidase/α-amylase/PTP1B inhibition profiling, HPLC-PDA-HRMS, analytical-and semipreparative-scale HPLC isolation, and NMR spectroscopy for investigation of crude root bark extract of E. longifolia, led to identification of 16 serrulatane diterpenoidswith several having unusual side chain modifications.In this study, 12 previously undescribed serrulatane diterpenoids were identified.This shows that roots of E. longifolia contain a wide range of serrulatane diterpenoids, a class of compounds not previously characterised in this species, and the current findings greatly expand our knowledge on the chemistry in the roots of Eremophila.
Eremophila spp.are culturally important plants for many of Australia's First Peoples, the Aboriginal peoples.This also applies to E. longifolia (R.Br.)F.Muell.(Scrophulariaceae) from which decoctions of leaves, twigs, and bark are used to cure headaches and to heal sores, and infusions of leaves have been applied to sore eyes and to treat insomnia (Richmond and Ghisalberti, 1994).As outlined in the introduction, diterpenoids isolated from other Eremophila species are reputed for additional medical properties not tested in our current study.Thus, if you use the information provided here to make commercial products, we urge you to strongly consider benefit sharing with the Aboriginal communities or groups in the areas where these species grow.We acknowledge that this work took place on the lands of Aboriginal peoples who are the custodians of this land and acknowledge and pay our respects to their Elders, past and present.Ethical dilemmas related to the above matters have recently been discussed in a perspective article (Semple et al., 2022).

Plant material and extraction
Roots of Eremophila longifolia (R.Br.)F.Muell.(Scrophulariaceae) were collected North-East of Kichner Australia (30 Melbourne, Department of Botany, Victoria, Australia (voucher number: MELUD122727a).The plant material was after collection shipped on dry ice and stored at − 20 • C before extraction.The root bark was peeled off using a knife and 45.4 g (wet weight) was milled and afterward extracted with 250 mL of dichloromethane and methanol (1:1), shaking for 20 min at 120 rpm (Ratek, Knox City, VIC, Australia), and macerated overnight.The next day the extract was vacuum filtered using a Büchner funnel with Whatman filter paper (No. 1 from Whatman International Ltd.Maidstone, England).The filtrates were subsequently concentrated in vacuo at 45 • C using an IKA RV10 rotary evaporator and finally freezedried on a Christ Alpha 2-4 LD freeze dryer (Martin Christ Freeze Dryers, Osterode am Harz, Germany).The crude extract (2.8 g) was stored at − 20 • C in amber-colored vials until it was shipped to Department of Drug Design and Pharmacology, University of Copenhagen, for further studies.

α-Glucosidase inhibition assay
The α-glucosidase inhibition assay was performed as previously reported (Schmidt et al., 2012).In short, 0.1 M phosphate buffer consisting of 0.034 M NaH 2 PO 4 ⋅2H 2 O and 0.066 M Na 2 HPO 4 was added 0.02% NaN 3 and adjusted to pH 7 with NaOH.First, 90 μL phosphate buffer and 10 μL DMSO were added to each well to dissolve the analytes, and the microplates were shaken 3 min.Subsequent 5 μL of α-glucosidase solution (0.002 U/μL) and 75 μL phosphate buffer were added, and the microplates were incubated for 10 min at 28 • C. Finally, 20 μL substrate solution (10 mM p-NPG) was added to initiate the reaction.The microplates were incubated for 35 min, with the absorbance of the cleavage product 4-nitrophenol measured at 405 nm every 30 s using a Thermo Scientific Multiscan FC microplate reader (Thermo Scientific, Waltham, MA, USA) coupled to SkanIt ver.2.5.1 software.All solutions were prepared by dissolving or diluting the individual compounds in the phosphate buffer to give the required concentrations.The total volume in each well was 200 μL with a final concentration of 5% DMSO, 0.01 U of α-glucosidase enzyme per well, and 1 mM p-NPG.The enzyme activity (cleavage rate) was recorded as ΔAU/s, and the percentage inhibition was calculated using the following equation: The high-resolution inhibition profiles were constructed by plotting the percentage inhibition against the respective chromatographic retention time obtained during microfractionation of the crude extract to produce a triple high-resolution inhibition profile.

α-Amylase inhibition assay
The α-amylase assay was performed as previously reported (Okutan et al., 2014).In short, 0.11 M phosphate buffer consisting of 0.038 M NaH 2 PO 4 ⋅2H 2 O and 0.066 M Na 2 HPO 4 was added 0.002% NaN 3 and 0.35% NaCl, adjusted to pH 6 with H 3 PO 4 and then added 0.3 mM Ca (OAc) 2 .80μL phosphate buffer and 20 μL DMSO were added to each well to dissolve the analytes, and the microplates were shaken 3 min.Subsequent 5 μL of α-amylase solution (0.002 U/μL) and 75 μL buffer were added.The microplates were incubated for 10 min at 37 • C, whereafter 20 μL substrate solution (10 mM CNP-G3) was added to initiate the reaction.The microplates were incubated for 30 min, with the absorbance of the cleavage product 2-chloro-4-nitrophenol measured at 405 nm every 30 s using the above-mentioned microplate reader.All solutions were prepared by dissolving or diluting the individual compounds in the phosphate buffer to give the required concentrations.The total volume in each well was 200 μL with a final concentration of 10% DMSO, 0.01 U of α-amylase enzyme per well, and 1 mM CNP-G3.

PTP1B inhibition assay
The PTP1B assay was performed as previously reported (Tahtah et al., 2016).In short, a buffer consisting of 50 mM Tris, 50 mM bis-Tris, and 100 mM NaCl was adjusted to pH 7 with AcOH.The following solutions were prepared by dissolving or diluting with the buffer to give the required concentrations.First, 18 μL DMSO and 52 μL EDTA solution (3.46 mM) were added to each well to dissolve the analyst, and the microplates were shaken 3 min.Subsequent 60 μL substrate solution (1.5 mM p-NPP and 6 mM DTT) was added, and the microplates were incubated for 10 min at 25 • C. Finally, 50 μL PTP1B solution (0.001 μg/μL) was added to initiate the reaction.The microplates were incubated for 10 min, with the absorbance of the cleavage product p-nitrophenolate measured at 405 nm every 30 s using the above-mentioned plate reader.All solutions were prepared by dissolving and diluting the individual compounds in the buffer described above to give the required concentrations.The total volume in each well was 180 μL with a final concentration of 10% DMSO, 1 mM EDTA, 0.5 mM p-NPP, 2 mM DTT, and 0.05 μg of PTP1B enzyme per well.

HPLC-PDA-HRMS
Samples for HPLC-PDA-HRMS analysis were prepared by adding 80 μL of MeOH to each of the 20 μL fractions in the LC-MS vials.The purity of the samples was analysed using the same chromatographic conditions, solvent gradient profile, and column as described in section 4.3 with an injection volume of 20 μL.Mass spectra were acquired in positive mode, and a solution of sodium formate clusters was automatically injected at the beginning of each run to enable internal mass calibration.

NMR experiments
Chemical shifts of 1 H and 13 C were referenced to the residual solvent signal of methanol-d 4 and chloroform-d at δ H 3.31 ppm, δ C 49.0 ppm, and at δ H 7.26 ppm, δ C 77.16 ppm, respectively. 1H NMR spectra were acquired with 30 • -pulses and 64k data points and zero-filled to 256k data points.2D homo-and heteronuclear experiments were acquired with 2048 (DQF-COSY, HMBC, ROESY) or 1730 (HSQC) data points in the direct dimension and 512 (DQF-COSY, HMBC) or 256 (multiplicity edited HSQC, ROESY) data points in the indirect dimension (Compounds 5, 15, and 17 had 256 data points in the indirect dimension in the HMBC experiments and compound 17 had 512 data points in the indirect dimension in the ROESY experiment).2D NMR data were zero-filled to 1k in F1 and double the number of data points in F2, employing forward linear prediction in F1 (LPBIN = 0).Processing of NMR data was done using Topspin ver.4.0.8(Bruker Biospin, Karlsruhe, Germany).

Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Fig. 1 .Fig. 2 .
Fig. 1.HPLC chromatogram of crude root bark extract of E. longifolia monitored at 210 (black) and 254 (yellow) nm with corresponding high-resolution α-glucosidase (green), α-amylase (blue), and PTP1B (red) inhibition profiles.(For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)
The constituent represented by peak 2 showed [M+H-H 2 O] + and [M+Na] + ions at m/z 315.1943 (calcd for C 20 H 27 O 3

Table 3 1
H NMR (600 MHz) and13C NMR (150 MHz) spectroscopic data of 11, 14, 15, and 17 (δ in ppm).and H 3 -19 to C-7 (δ C 143.3 for 3 and δ C 143.0 for 6) and from H-1 to C-8 (δ C 142.7 for 3 and δ C 143.5 for 6).Likewise, the three remaining quaternary aromatic carbon atoms were identified by HMBC correlations from H 3 -19 to C-6, from H-2β, H-5, and H 3 -20 to C-9, and from H-1 to C-10.The spectra of 3 and 6 display large differences in the coupling patterns for H-2α and H-2β, and the largest chemical shift differences between the two compounds are seen for H-1, H-2α, H-12A, H-12B, and H 3 -20.This indicates different conformations of the cyclohexenone ring systems between the two compounds, caused by different configurations at C-4, i.e., where the large side chains are positioned.

Table 2 ,
HRMS in Supplementary data FigureS25, and 1 H NMR, COSY, HSQC, HMBC, and ROESY data and spectra are provided in Supplementary data TableS6 and Figures S26-S30.The constituent represented by peak 7 showed a [M+H] + ion at m/z 311.1635 (calcd for C 20 H 23 O 3 + , 311.1642, ΔM +2.2 ppm), corresponding to a molecular formula of C 20 H 22 O 3 and a hydrogen deficiency of 10.