Capillary electrophoresis-mass spectrometry characterisation of secondary metabolites from the antihyperglycaemic plant Genista tenera

is endemic to the Portuguese island of


Introduction
Genista tenera (Jacq. Ex Murr) O Kuntze, which belongs to the Leguminosae family, is endemic to the Portuguese island of Madeira where an infusion of the aerial parts of the plants is used as a traditional medicine for the treatment of diabetes. A study of the activity of an ethyl acetate extract of G. tenera in normal and streptozotocininduced diabetic rats has shown antihyperglycaemic activity ([1], and authors' unpublished data). These preliminary results, together with the evidence that flavonoids can act as aldose reductase inhibitors, blocking the sorbitol pathway commonly linked with diabetes [2], provide a strong motivation for the study of these plant compounds.
Previous investigations have revealed the presence of alkaloids and flavonoids as the major secondary metabolites present in extracts of the aerial parts of G. tenera.
A recent study using GC-MS has identified the presence of ten different alkaloids [3] while two previous studies of organic extracts of G. tenera have identified a total of 11 flavonoids. In the first, FAB-MS/MS and NMR allowed identification of five flavonoids and one flavonoid-O-glucoside [4]. In the second, more recent, study using LC-MS and NMR analyses, a further two flavonoid-O-glucosides, two flavonoid-O-diglycosides and one flavonoid-C-glucoside were identified [5].
The combination of RP-LC with MS is routinely used for the analysis of phenolic compounds in plant extracts, because it is robust and fully automatable and methods are well precedented in the literature. Combination of LC with mass spectrometric detection enables structural information to be obtained. However, typical LC-MS analyses take in excess of 40 min (e.g., [5][6][7]). In contrast, separation of phenolic compounds using CE instead of LC offers the advantage of rapid analysis times. Typical analysis times are between 10 and 20 min (e.g., [8][9][10]). [11], only 47 papers describing CE-MS applications to all classes of phenolic compound (i.e. including analysis of standards, drugs and plant metabolites) had been cited since 1994.
The work reported in this paper is part of an ongoing project to study the bioactivity and isolate the active principles of G. tenera, and is the first application of a CE-MS method for this plant species. This approach has allowed us to distinguish at least 26 different phenolic components in only 10 min and demonstrates the benefits of CE-MS/MS as a rapid screening technique.

Plant material
Aerial parts of the plant were collected on the island of Madeira at the onset of anthesis (flowering). Affiliates of the Jardim Botânico da Madeira, Funchal, Madeira collected the samples and retained a voucher specimen (MADJ 2508).

Preparation of the infusion
Air-dried and powdered plant (162.71 g) was extracted in a Soxhlet apparatus with 2.5 L EtOH. The crude extract was filtered in a Büchner funnel to dryness under vacuum. The residue (40.49 g) was redissolved in 150 mL of warm water, filtered and successively partitioned against 150 mL diethyl ether, 500 mL ethyl acetate and 1 L BuOH. The fractions were then taken to dryness. One hundred milligrams of the total BuOH extract (7.41 g) was rehydrated in 5 mL of water containing 3% v/v formic acid. SPE was performed using Strata C-18e cartridges (Phenomenex, Torrance, CA, USA); the 5 mL of extract was loaded onto the column, conditioned following the manufacturer's recommendations, and then washed with 10 mL of 3% v/v formic acid; the analytes were then recovered in 2 mL of methanol and the solution taken to dryness in a centrifugal concentrator. The analytes were reconstituted in 1 mL of 3% v/v formic acid and the solu-tion was then filtered through a PVDF membrane filter unit (Millipore, Bedford, MA, USA) with a pore size of 0.45 mm and aliquots were directly analysed by CE-MS.

CE/ES-ITMS analysis
Separations were carried out using a PrinCE instrument (Prince Technologies, Emmen, The Netherlands) interfaced to an ITMS (Deca LCQ XP Plus, Thermo Electron, San Jose, CA, USA) via a Thermo Finnigan coaxial sheath flow electrospray interface. The separation was performed in an untreated fused-silica capillary, 65 cm in length with an internal diameter of 50 mm. The outlet end of the capillary was tapered by holding a small length of the capillary in a butane flame and pulling laterally to a point using a pair of tweezers. A ceramic tile capillary cutter was used to trim the tapered capillary to a tip with an approximate internal diameter of 20 mm. The untapered end of the capillary was threaded through the sheath flow interface and through a hole in the side of the CE instrument and then threaded through the manufacturer's cartridge to the electrode nearest the ITMS, thus minimising the length of the separation capillary. New capillaries were conditioned by flushing at 1000 mbar with MeOH (2 min), H 2 O (2 min), 0.1 M HCl (10 min), H 2 O (2 min), 0.1 M NaOH (10 min) and BGE (10 min). An initial start-of-day wash procedure was used to condition and equilibrate the capillary. This consisted of a 10 min wash with 0.1 M NaOH for 10 min and BGE (10 min) followed by a 10 min wash with the BGE using a pressure of 1000 mbar.
The sample was hydrodynamically injected using a pressure of 10 mbar for 0.20 min (sample volume ,9 nL) followed by an injection of BGE (75 mbar for 0.19 min). The BGE used was 95:5% v/v water:propan-2-ol containing 10 mM ammonium carbonate adjusted to pH 9.25 with ammonium hydroxide (35% v/v). A separation voltage of 30 kV (normal polarity) was used. The sheath flow liquid used was 50:50% v/v IPA:H 2 O, with flow rate 0.5 mL/min.
To tune the ITMS, the voltages on the lenses were optimised using the TunePlus optimisation function of the Xcalibur Software (version 1.3), whilst infusing the plant extract through the separation capillary with a pressure of 100 mbar and tuning on m/z 285 (a dihydroxy-methoxy flavone). Positive ion mode was selected for MS and MS/ MS experiments, using the following conditions: spray voltage was 4 kV, tube lens offset 20 V, transfer capillary temperature 2507C; no nebulising or auxiliary gas was used. The scan range was m/z 150-1000, the maximum injection time was set to 100 ms and two microscans were used. Whilst performing the MS/MS analyses two scan functions were performed; the first scan event was a full-scan mass spectrum and the second an MS/MS experiment performed using the data-dependent acquisition mode. MS/MS data were acquired on the three most intense ions, using a dynamic exclusion list.

Optimisation of CE separation
Previous CE separations of flavonoids, e.g. [9,10] have typically used borate buffer solutions, but these are not MS compatible. Dissociation constants of some flavonoids have recently been determined by CE, using mobility measurements in a set of buffer solutions including ammonium buffers in the pH range 9.6-10.8 [12]. For pH values greater than ,8-9, flavonoids migrate as anions, and thus migrate in the opposite direction to the EOF in this experimental set-up. In the present study ammonium carbonate adjusted to pH 9.25 with NH 4 OH was used as an MS-friendly BGE. Three different concentrations of BGE were evaluated; 1, 5 and 10 mM. It was observed that as the concentration of the BGE was increased the resolution was improved but analysis time increased. The increase of analysis time arises due to the decrease in EOF velocity with increasing ionic strength. The 10 mM concentration BGE was found to give the best compromise between analysis time and separation quality. The measured current was ,12 mA. In our experience using the PrinCE instrument, where the capillary is not thermostatted, currents greater than ,40 mA should be avoided to prevent excessive Joule heating and reduced robustness of the CE-MS system. The BGE also contained 5% v/v organic modifier, chosen to match the organic component in the sheath liquid. A sheath liquid comprising 50:50% v/v MeOH:H 2 O was initially used; however, this gave rise to an unstable sheath flow, even at flow rates in excess of 3 mL/min. Observation of the CE-MS interface using a video camera showed that the sheath liquid failed to flow steadily to the end of the capillary and gave periodic discharges, which were observed as arcing from the capillary to the sheath flow capillary. Use of a more viscous sheath liquid, 50:50% v/v IPA:H 2 O, yielded a steady spray and resolved the problem of arcing; thus, this sheath liquid and propan-2-ol as organic modifier in the BGE were used for all subsequent experiments.

CE-MS and CE-MS/MS results
The total ion electropherogram (TIE) obtained from the CE-MS analysis of the plant extract is shown in Fig. 1. Thorough examination of the mass spectral and product ion data allowed recognition of 26 components, which are listed in Table 1 Eight ions with m/z values corresponding to aglycones were detected. In addition, 13 variously glycosylated and five acetylated forms of these basic structures were also detected. Table 1 presents a detailed list of all 26 compounds in migration order. Some m/z values were observed more than once in the electropherograms. For example, there are at least four separate peaks with m/z 433 (Fig. 2). A similar number of peaks with this m/z value were also observed using LC-MS (unpublished data), suggesting that the peaks arise from differential migration of different isomers.
The acid dissociation constants pK a1 and pK a2 of flavonols lie in the range 5-10, with exact values depending on the positions of the different -OH groups in the aglycone rings, and the extent of conjugation and charge delocalisation associated with deprotonation [12,13]. The average net charge at pH 9.25 will therefore be different for the different structural isomers, and this may explain their varying migration times.   Previous studies have shown that MS/MS can be used to distinguish between O-and C-glycosylated flavonoids [14]; fragmentation induced by CID gives rise to Y 0 and B 0 ions for O-glycosides but results in cross ring cleavages (X ions) for C-glycosides (Fig. 3). Although there is literature precedent for product ion spectra allowing isomeric  aglycones to be differentiated [15], the MS 2 data obtained using the IT did not contain diagnostic aglycone cleavage ions. From the tandem data some of the flavonoid glycosides could be assigned as either O-or C-hexosides. For example, an analyte with m/z 435 was detected at 6.03 min. From the tandem data it was deduced that this was a trihydroxy-flavonone C-hexoside from the many cross ring fragment ions that were detected (Fig. 4).
C-glycosides are only commonly found with the glycan attached to either C6 or C8. It has been previously reported [16] that ion intensity ratios from high-energy CAD tandem data can be used to distinguish the site of glycosylation. Previous low-energy collision-induced experiments performed on ITMS used MS 3 to distinguish between C6 and C8 isomers [17]. As the peaks obtained in the CE-MS/MS separations were narrow (,0.2 min), frequently only single scan product ion data were recorded and so it was not possible to obtain the necessary MS 3 data. The SIE of m/z 447 gave rise to two major peaks at 3.68 and 4.66 min (Fig. 5). From the tandem data it is proposed that the peak at 3.68 min corresponds to a dihydroxy-methoxyflavone hexoside, as only the Y 0 ion (m/z 285) was produced (Fig. 5B). However, the tandem data arising from the m/z 447 ion at 4.66 min (Fig. 5C) showed an intense Y 0 ion (at m/z 285), together with cross ring fragment ions, e.g. 0,2 X 1 , 0,4 X 1 -2*H 2 O and 2,3 X 1 -2*H 2 O 1 , which are indicative of a C-hexoside. This suggests that both the O-and the C-glycoside are comigrating.   C-glycosides were formed. The majority of the secondary metabolites determined are flavonoids and glycosylated flavonoids; acetylated flavonoids were also detected.

Concluding remarks
The development and application of a CE-MS/MS method for the study of an extract of the antihyperglycaemic plant G. tenera has enabled us to distinguish the following in only 10 min: five flavonoid aglycones, five flavonoidmonoglycosides, two flavonoid-diglycosides, one flavonoid-triglycoside, three monoacetyl-flavonoids, one diacetyl-flavonoid and one acetyl-flavonoid-glycoside. Importantly, our MS/MS data made it possible to identify the presence of comigrating O-and C-glycosides. A previous publication [5] describing analysis of an ethyl acetate extract of G. tenera identified five different components, using LC-UV, LC-MS and NMR. Using LC retention times of authentic standards, the isomeric structures in four of the LC-UV peaks were assigned (ES tandem mass spectrometric data had been obtained for two of these), with the final component being identified using NMR spectroscopy and MS. In contrast, data consistent with these five components, as well as a further 21, were provided using our CE-MS/MS approach, which offers the additional advantage of taking one tenth of the time and requiring significantly less sample. Our data demonstrate that CE-MS/MS is particularly suited to the analysis of plant metabolites in complex mixtures where concentration is not a limitation.