Profiling and activity screening of Dammarane-type triterpen saponins from Gynostemma pentaphyllum with glucose-dependent insulin secretory activity

The global prevalence of type 2 diabetes is increasing rapidly; consequently there is great need for new and novel therapeutic options. Gynostemma pentaphyllum (GP) is a traditional medicinal plant, mainly present in Southeast Asian countries, that has been reported to exert antidiabetic effects, by stimulating insulin secretion. The specific compound responsible for this effect is however as yet unidentified. Screening for discovery and identification of bioactive compounds of an herbal GP extract, was performed in isolated pancreatic islets from spontaneously diabetic Goto-Kakizaki (GK) rats, a model of type 2 diabetes, and from non-diabetic control Wistar rats. From this herbal extract 27 dammarane-type saponins, including two novel compounds, were isolated and their structure was elucidated by mass spectrometry and NMR spectroscopy. One of the dammarane-type triterpenoid showed a glucose-dependent insulin secretion activity. This compound, gylongiposide I, displays unique abilities to stimulate insulin release at high glucose levels (16.7 mM), but limited effects at a low glucose concentration (3.3 mM). Further studies on this compound, also in vivo, are warranted with the aim of developing a novel anti-diabetic therapeutic with glucose-dependent insulinogenic effect.


Results
In order to isolate and characterize the maximum number of saponin structures in the extract, an optimized chromatographic method was developed using a C-18 reverse-phase column (Fig. 1). The extract was fractionated by repeated preparative chromatography and the fractions were further purified by a final isocratic elution. Twenty-seven saponins were characterized and two of them, fractions 12 and 34 had new structures not previously reported. (Table 1 and Fig. 2). A compilation of the structures, NMR 1 H and 13 C chemical shifts and HSQC NMR spectra is available in the Supplementary Information. Biological activity. To locate the active saponins, ten sequential time fractions (Fig. 1) from the HPLC separation were collected and screened for biological activity. GP extract fractions or compounds isolated from the fractions were incubated with pancreatic islets of either W or GK rats to determine effects on insulin release at low (3.3 mM) and high (16.7 mM) glucose concentrations. No significant insulin stimulatory activity was found in TF1-TF2, TF4-TF8 and TF10, or the flavonoid fraction (Fig. 3), while such an activity was found in TF3 and TF9. TF9 enhanced insulin release to a similar extent regardless of in the presence of 3.3 or 16.7 mM glucose. TF3 dose-dependently increased insulin release at 16.7 mM glucose, while it did not enhance insulin release at 3.3 mM glucose (Fig. 4).
The saponins in fractions TF3 were further isolated and tested for their individual activity. Two compounds 20A and 20B in TF3 were found to stimulate insulin secretion. Compound 20B (Fig. 5) showed beneficial biological activity by only stimulating the insulin secretion at high glucose concentration (Fig. 6). This compound (20B) known as gylongiposide I has been isolated and its structure reported previously but has not been studied for its anti-diabetic activity 17-20 . Structural elucidation. Compound 12 (Fig. 2) was obtained as a white, amorphous powder, the molecular formula was established as C 47 H 76 O 18 and a molecular ion peak was observed by both positive and negative-mode ESI-MS (m/z 951.4863 [M + Na] + ; calc. 951.4924). The 1 H-13 C multiplicity edited HSQC NMR (CD 3         The configuration at C-20 and C-23 were determined by comparison with the most recent studies of configuration at C-20 and C-23 carried out by electronic circular dichroism (ECD) and X-ray diffraction analysis 21,22 . Therefore, compound 12 was assigned to be (3β,20S,23R)-3,20,23,26-tetrahydroxydammar-24-en-21-oic acid- Compound 34 (Fig. 2) was obtained as a white, amorphous powder, the molecular formula was established as C 49 H 78 O 18    Compound 29     Compound 41

Discussion
Biological activity. We have identified and characterized the dammarane-type triterpene saponin responsible for the insulin secretion induced by a Chinese extract of the Gynostemma Pentaphyllum (GP) herbal plant. Although several of the pooled time fractions, TFs, obtained from the HPLC chromatogram, had varying degree of effect on insulin secretion from isolated islets of control W and diabetic GK rats, significant effects on the insulin-producing beta-cells were seen with one of the fractions, TF3. In addition, we found that TF3 had the most favourable antidiabetic effect by enhancing insulin release significantly only at high glucose but not at low glucose concentrations. This is in contrast to the well-known antidiabetic sulfonylurea drugs, that stimulate insulin secretion also at normoglycemia or even at lower glucose levels 23 and thereby can cause serious hypoglycemia in patients. By further chromatographic separation of the saponins present in TF3 followed by bio-assay testing we were able to determine that this favourable activity, i.e. a glucose-dependent stimulation of insulin release was entirely related to the saponin compound 20B.
In this context, it is of great importance to note that this compound most likely retains its biological activity following oral administration, since we have shown prominent antidiabetic activity in diabetic GK rats, treated orally with GP extract (0.3 g/kg body weight) for 2 weeks 24 . In these animals, the improvement of glucose tolerance was demonstrated in parallel with augmented plasma insulin levels. A similar effect was previously reported by another compound isolated from GP extract, i.e. phanoside 12,25 . However, in contrast to the present saponin     20B, phanoside did not stimulate insulin secretion in a glucose-dependent manner. Recent mechanistic studies on the GP extract demonstrated that the stimulation of insulin secretion from rat islets was mediated by effects on several steps in the stimulus-secretion coupling to glucose in the pancreatic beta cells 24 . Among these steps are the K-ATP channels, the L-type Ca 2+ channels, the protein kinase A activity, and pertussis-toxin sensitive exocytotic G e -proteins.
Structure and activity. Gynostemma pentaphyllum extracts are known to possess numerous biological activities through their dammarane-type triterpene saponins. The possible structural-activity relationships (SAR) is related to the number and nature of the sugars, their acylation, the types of aglycones and the stereochemistry. A large number of studies have been conducted on anti-tumor, anti-inflammatory and anti-diabetic activities, where different structural components have been pointed out as important for the activities 2,24,26 . The presence of a hydroxyl group at position C-3 9 , a double bond between C20-C22 or C20-C21 27 , and a five membered unsaturated ketone in the side chain 28 in the dammarane-type structure has been mentioned as possible structural components responsible for anti-tumor activity. These results are from studies carried out on different cancer cell lines and are therefore difficult to compare. The differences in anti-inflammatory abilities have been explained by differences in sugar moieties 29 . From the present work, some structural features are worth noticing: The glycoside moiety at the C-3 position on the aglycone of the isolated saponins consists of a branched trisaccharide either α-L-Rha- 3)]-β-D-Glc with acetylation on C6 of glucose and in some cases also on C4 of rhamnose. In some cases, as in compounds 8, 10, 15 and 17 an additional glucopyranose unit is attached at C-21. The aglycone moiety shows structural variability at two positions. The substituent at C-19 can be either a methyl or an aldehyde group. The greatest variability between the structures is found in the side chain attached to C-17. Ten different side chains attached to C-17 were found, some composed of five membered rings, either as lactone or hemiacetal, other as open structures with different number of double bonds and/or additional glycoside moiety attached at C-21 (Fig. 7).
Interestingly, the potential to stimulate insulin secretion at high glucose concentration is limited to one single compound, 20B, while several saponins can stimulate insulin secretion independent of glucose concentration. Comparison of the structures of the different saponins shows that compound 10 has a structure very similar to that of 20B with the same trisaccharide unit at C-3, an aldehyde group at C-19 and the same open chain attached at C-17 with a double bond between C-24 and C-25. The only difference between the two compounds is the presence of an additional glucose residue at C-21 in compound 10. Thus, the presence of an additional sugar appears to suppress the insulin secretion at high glucose concentration.
Because of the increasing amount of data on the biological activities of saponins, rapid screening methods including activity and structural data such as LC-MS and NMR are valuable 30 . The 1 H-13 C HSQC NMR spectra of the 27 identified saponins together with the MS data and 1 H and 13 C NMR chemical shifts reported in the present work should help in structural analysis when investigating new extract, and may allow for more rapid screening of already known compounds.

Experimental Procedures
Material and reagents. Gynostemma Pentaphyllum (GP) herbal extract was purchased from Hanzhong TRG Biotech Co.,Ltd. Solid phase extraction column Hypersep C-18 was purchased from Thermo scientific. Deionized water was purified by Milli-Q system from Millipore. Ethanol absolute 100% Ph.Eur, acetonitrile (HPLC grade) and methanol (HPLC grade) were purchased from VWR chemicals. Methanol D4 99.80% D (NMR) were purchased from Euriso-top. Collagenase, HEPES, bovine albumin, D-glucose, L-glutamine, RPMI 1640 culture medium, and fetal calf serum were obtained from Sigma-Aldrich, Sweden.
Extraction and isolation. The dried and powdered GP extract was dissolved in 30% ethanol solution, and applied to a Hypersep C-18 cartridge. The column was washed with 30% ethanol solution (flavonoid fraction). The saponin fraction was collected by elution with 100% ethanol. The saponin fractions were dried and further separated through preparative HPLC (Gilson 305/306 system) on a 20 × 150 mm C-18 column (Kromasil 100-5C18 HiCHROM, UK) with a binary solvent system of an aqueous (Milli-Q) 0.1% formic acid solution (A) and acetonitrile (B), with the following gradient elution: initially 30% (B) -70% (A), changed to 50% (B) after 7 min, 50% (B) to 15 min, 60% (B) to 35 min, 80% (B) to 40 min and returned to 30% (B) after 41 min. The detection (Gilson 118 UV/VIS detector) was done at 210 nm, and a flow rate of 10 ml/min was used.
Ten sequential time fractions from the HPLC separation were collected (Fig. 1), lyophilized and stored at 4 °C until activity testing.
Furthermore, fractions 1-42 ( Fig. 1) were collected and lyophilized, through repeated injection to receive sufficient material for NMR analysis to determine if the fractions contain flavonoids, saponins or other components. Some of the 42 screened compounds were not saponins and are therefore not reported. Enough material were collected through repeated injections for exhaustive structural analysis of 27 saponins fractions. The fractions were then lyophilized and purified by isocratic elution according table S24 available in the Supporting Information. Fractions 20A and 20B were co-eluting and separated by an isocratic elution (50:50) of a binary solvent system of an aqueous 0.1% formic acid solution (A) and 80:20% acetonitrile:methanol (B). The purified fractions were stored at 4 °C until NMR and LC-MS analysis.

LC-MS.
The LCMS analysis were carried out on isolated fractions from the crude extract with an Agilent 1100 liquid chromatography system equipped with a variable UV wavelength detector and a Kromasil ® C-18 reverse-phase (4.6 mm × 150, 5 µm) column using suitable isocratic condition for each isolated fraction of aqueous:organic solvent with the inclusion of 0.1% formic acid into both aqueous (Milli-Q) and organic phase in which the later phase consists of acetonitrile and methanol 8:2 (v/v). The flow rate was set at 1 ml/min and the UV detection wavelength was set at 210 nm to verify and screen UV peak of the compounds with their respective total ion chromatogram (TIC) peak.
The accurate mass analysis on the saponins were determined using Maxis Impact Bruker ® MS-QTOF coupled to the LC system and using Compass ® Bruker data processing software.
Animals and pancreatic islet isolation. Male diabetic GK rats and control Wistar (W) rats were used.
The GK rats were bred within the department 31 , while the W rats were purchased from Charles River Ltd. The rats were maintained in a 12 h light and dark cycle and allowed food and water ad libitum. The GK rats (n = 25) had body weights ranging from 160 to 277 g, and their non-fasted plasma glucose levels were 6.4-10.9 mM, while W rats (n = 8) were 199-267 g with plasma glucose levels 4.4-5.9 mM. The animals were sacrificed by CO 2 to allow removal of pancreata for islet isolation by collagenase digestion as described 32 . Isolated islets were handpicked under a stereo microscope, maintained for 24 h at 37 °C and 5% CO 2 in culture dishes with RPMI 1640 medium, containing 10% heat-inactivated fetal calf serum, 2 mM L-glutamine, and 11 mM glucose prior to further experiments with incubations to determine insulin secretion 25,31,33 . All methods and experimental protocols were carried out in concert with relevant guidelines and regulations. In addition, the studies were performed after approval by the laboratory animal ethics committee of the Karolinska Institutet, Stockholm.

Islet batch incubations.
Isolated islets were pre-incubated for 45 min in Krebs-Ringer bicarbonate (KRB) buffer 25,33 , pH 7.4, and supplemented with 10 mM HEPES, 0.2% bovine albumin and 3.3 mM glucose. Thereafter, batches of three islets were incubated in KRB buffer with 10 mM HEPES, 0.2% albumin and either 3.3 or 16.7 mM glucose. In addition, GP fractions were filtered through a 0.20 μm syringe filter prior to the incubations to remove any debris from the sample. The GP fractions were diluted to final concentrations of 100 µg/ml, 10 µg/ml, 1 µg/ ml, and a control of 0 µg/ml in buffer with either glucose concentration. Each GP fraction was tested in triplicates, and each batch of islets was incubated for 60 min in 300 µl buffer in a shaking water bath at 37 °C, and then 200 µl of the incubation media was removed and stored at −18 °C until later determination of insulin. Experiments were repeated two times, and hence there were 3-6 observations from each fraction. Insulin was determined by radioimmunoassay as described 34 . Statistical analysis. All results from insulin secretion experiments were analysed and presented as mean and standard error of the mean (SEM): One-way Anova was performed to evaluate statistical significance of results. When this analysis indicated statistical significance, the results were further analysed by Fisher's least significant difference. Statistical analysis was conducted on each of the individual batch incubations and therefore n = 3-6 for each of the fractions tested at each concentration. The n value varied depending on the number of replicas of the experiment conducted with triplicates. A p value < 0.05 was considered statistically significant. To allow comparisons between the groups, the mean results for each concentration of the pooled fraction, or pure saponin, were divided by their respective control at low and high glucose levels. This allowed a direct comparison between the groups since the absolute values of insulin secretion from the islets varied between rats. The results were therefore presented as fold induction of insulin release compared against 3.3 mM or 16.7 mM glucose control values.