Introduction

Vitamin E is a family of eight naturally occurring liposoluble molecules counting four tocopherols T and four tocotrienols T3 (Fig. 1). All these compounds have a hydroxylated chromane ring, which is responsible for their antioxidant activity, modulated by one to three methyl groups [1]. Tocotrienols differ from tocopherol by the presence of three unconjugated double bonds in their farnesyl isoprenoid tail. As well as its well-described antioxidant activity [2], vitamin E also has a function in inhibiting platelet aggregation related to thrombosis [3], could protect against neurodegenerative disease [4], and acts as an enzymatic activity regulator [5]. Among vitamin E congeners, α-tocopherol is the most active and is predominant in the European diet [6] because it is abundant in olive and sunflower oils [7]. In contrast, higher consumption of γ-tocopherol, which is more abundant in soybean and corn oils [7], is described for the American diet [8]. Because vitamin E supplements mainly contain α-tocopherol derivatives [9], i.e. α-tocopherol acetate, it is of high importance to control the level of other tocopherols and tocotrienols in food to diversify intake of vitamin E.

Fig. 1
figure 1

Chemical structures of the tocopherol T and tocotrienol T3 congeners

Full size image

Because of the low volatility of these compounds, liquid chromatography (LC) is the most suitable method for separation in biological samples. Although quantification of α-tocopherol among other liposoluble vitamins has been extensively described [10–13], baseline separation of all eight T and T3 molecules is mandatory when using UV or fluorescence detection, to avoid possible interferences when studying complex biological samples. This was previously achieved by normal-phase [14–16] and reversed-phase [17, 18] liquid chromatography (NPLC and RPLC, respectively). A way to improve specificity and sensitivity is to couple LC with mass spectrometry (MS) using either multiple-reaction-monitoring (MRM) [19] mode, available on triple-quadrupole instruments, or single-ion-monitoring (SIM) [20–22] mode or high-resolution-mass-spectrometry (HRMS) detection available with time-of-flight (TOF) or Fourier-transform mass analyzers [23]. Normal-phase analysis is not well suited to coupling with atmospheric-pressure ion sources, especially electrospray (ESI), which is widely used. This is because of the use of hazardous volatile organic solvents which are not compatible with efficient droplet fission and proton exchange with the analyte during ESI. In contrast, RPLC is well designed for MS coupling because a high percentage of water is commonly used for compound elution. Nonetheless, for T and T3 quantification by RPLC–MS a high percentage of methanol (85–95 %) was commonly used to achieve a good separation [17, 18]. Lanina et al. revealed that atmospheric-pressure chemical ionization (APCI) is more sensitive than ESI for T and T3 quantification [21]. However, sample preparation is more complex and must include removal of lipids through a saponification step to prevent contamination of the column. Finally, analysis time is identical for NPLC and RPLC at approximately 20 min, including equilibration time. It would thus be interesting to develop a fast, green, and sensitive quantification of T and T3, including a simple sample-preparation step by direct solvent extraction, using chromatography coupled to MS.

For several years, supercritical-fluid chromatography (SFC) has been the subject of renewed interest as a result of new instruments becoming commercially available [24]. Supercritical CO2 can be mixed with a modifier, for example alcohol or acetonitrile, for eluting compounds from a column. Because of its low viscosity and high diffusivity, supercritical CO2 enables high-flow-rate experiments and thus reduced analysis time. Moreover, SFC offers the unique advantage of being fully compatible with all types of stationary phase, enabling fine modulation of the interactions between the analytes and the stationary and mobile phases. Coupling with MS is also possible, even if some practical limitations are still present [25]. There have been several attempts at T and T3 detection by SFC. In 1993, Snyder et al. developed a capillary SFC–MS analysis of the four T in vegetable oils using pure CO2 [26]. At the same time, Yarita et al. revealed that addition of methanol as a modifier can improve resolution for quantification of the same species [27]. The first quantification by packed-column SFC–UV of T and tocopherol acetate was performed by Buskov et al. in 1999 [28]. In 2004, Han et al. introduced the quantification of all eight T and T3 by packed-column SFC–UV, including NRM and MS identification of the compounds but with a global analysis time of 40 min [29]. There have been few efforts in this field since, and quantification of T and T3 should benefit from recent advances in SFC instrumentation, column technology, and mass spectrometry.

This article will thus establish how SFC–MS technology can be of great interest for the quantification of tocopherol and tocotrienols in the food industry, especially for the analysis of vegetable oils. Special attention will be paid to the choice of the appropriate ionization source (ESI, APCI, or atmospheric-pressure photoionization, APPI) to achieve the best sensitivity and repeatability.

Materials and methods

Chemicals

Carbon dioxide (purity 99.999 %) was purchased from Alphagaz (Le Plessis Robinson, France). α, β, γ, and δ-Tocopherol and tocopherol acetate were purchased from Sigma–Aldrich (Saint Quentin-Fallavier, France) and α, γ, and δ-tocotrienol from Interchim (Montluçon, France). Because of its high cost, β-tocotrienol was not included in this study. Methanol (HPLC grade) was purchased from J.T.Baker, 2-propanol (HPLC grade) from Fisher Scientific (Illkirch, France), and ethanol (HPLC grade) and chloroform (HPLC grade) from Carlo Erba (Val de Reuil, France). Formic acid (purity 98 %) was obtained from Fluka (Buchs, Switzerland).

The stock solution was prepared at an equimolar concentration of 1 mmol L−1 in pure cyclohexane and was stored at −20 °C in amber glass vials. T and T3 are highly sensitive to light degradation. To avoid oxidative degradation, butylated hydroxytoluene (BHT) was added to the stock solution at a final concentration of 30 mg L−1.

Preparation of soybean-oil extract

Because Katsanidis et al. reported fast degradation of δ-T and δ-T3 when performing the saponification step [30], 10 μL soybean oil (La Vie Claire, Montagny, France, batch SOBB20803, 02/2014) was directly diluted in pure cyclohexane by a factor of 50. BHT was added to the sample solution at a final concentration of 30 mg L−1. The entire sample preparation was achieved under soft light and at room temperature (20 °C) and the samples were placed in amber glass vials in a temperature-controlled autosampler (5 °C) just before analysis.

SFC–UV

Chromatographic analyses were performed using a 1260 Infinity Analytical SFC system from Agilent Technologies (Waldbronn, Germany). This system is equipped with a modified SFC 1260 Infinity Binary LC and an additional module “1260 SFC”, generating supercritical CO2 from gaseous CO2. Outlet pressure was regulated to 130 bar by a back-pressure regulator (BPR). An autosampler with a 1 μL injection loop was used, and the compartment temperature was kept at 5 °C to avoid fast degradation of the analytes. Two thermostated column compartments, containing up to eight columns of 100 mm length each, coupled to a selection valve, were used for stationary-phase screening. The 0.17 mm internal diameter stainless-steel capillaries were replaced by 0.12 mm ones to decrease the total dead volume [31]. The SFC system was coupled to a diode-array detector (DAD-UV–visible) with a 6 mm path length high-pressure flow cell.

Six different analytical columns were compared to optimize the separation conditions. Five analytical columns were purchased from Agilent Technologies (Massy, France): Zorbax Eclipse RRHDXDB-C18 (150 mm × 2.1 mm × 1.8 μm), Pursuit XRs Diphenyl DPH (250 mm × 2 mm × 3 μm), Pursuit Pentafluorophenyl PFP (150 mm × 2 mm × 3 μm), Zorbax RRHD SB-CN (100 mm × 2.1 mm × 1.8 μm), and Pursuit RX Silica Si (100 mm × 2.1 mm × 1.8 μm). The Amine Luna NH2 column (150 mm × 2 mm × 3 μm) was purchased from Phenomenex (Le Pecq, France).

The Amine Luna NH2 column was selected as the optimum column and an elution gradient separation was selected using CO2 (solvent A) and ethanol with 0.1 % formic acid (solvent B). The gradient was: 0–4 min linear gradient from 3.5 to 8 % solvent B, 8 % solvent B isocratic plateau for 1 min, 5–7 min from 8 to 3.5 % solvent B, 7–10 min re-equilibration time with 3.5 % solvent B. The flow was fixed at 1.5 mL min−1. The separation temperature was maintained at 30 °C during the analysis. The back-pressure regulator was fixed at 130 bar.

SFC–MS

SFC–MS analyses were achieved by coupling the SFC–UV system to a quadrupole time-of-flight mass spectrometer 6540 Agilent (Agilent Technologies, Massy, France) equipped with an ESI dual, APCI, or APPI source, operating in positive and negative-ion modes. Experimental settings (capillary voltage, nebulizer pressure, vaporizer temperature, drying-gas temperature and flow rate, sheath-gas flow, sheath-gas temperature, nozzle voltage, and fragmentor voltage) were optimized for each ion source. Mass spectra were recorded from m/z 50 to m/z 1000 at 2 GHz leading to a mass resolution of 20,000 at m/z 922. Calibration solution, containing two internal reference masses (purine, C5H4N4, m/z 121.0509, and HP-921 [hexakis-(1H,1H,3H-tetrafluoro pentoxy)phosphazene], C18H18O6N3P3F24, m/z 922.0098), routinely led to mass accuracy below 2 ppm. The mass spectrometer was operated in full-scan mode. Recorded data was analyzed with Mass Hunter software. For each ion of interest, the extracted ion chromatogram (XIC) was calculated by selecting the theoretical m/z value with a selection window of 5 ppm.

To improve reproducibility and robustness [32], a make-up solvent was added between the UV detector and the BPR by a 1260 Infinity isocratic pump (Agilent Technologies, Massy, France). The chemical nature of the solvent, the presence of additives, and the flow rate were also optimized for each ion source to obtain the best signal-to-noise ratio (S/N) by MS detection. Finally a Caloratherm (Sandra Selerity Technologies, Kortrijk, Belgium), the temperature of which was fixed at 60 °C, was added just before the ion source to avoid freezing of the transfer line.

Data acquisition and processing were performed using Chemstation and MassHunter Workstation software (Agilent Technologies, Massy, France). Figures were generated using OriginPro 7.5 SR6 software (OriginLab Corporation, Northampton, MA 01060, USA).

For SFC–UV and SFC–MS experiments, limits of detection (LODs) and of quantification (LOQs) were estimated for signal-to-noise (S/N) ratios equal to 3 and 10, respectively.

Results and discussion

Optimization of SFC separation

To optimize the SFC separation of T and T3, a UV-absorbance detector was used at a fixed wavelength of 295 nm. Different conditions, including the column chemistry, the mobile-phase composition and flow rate, and the oven temperature, were optimized to obtain shortened analysis time and the best separation. For a primary fast screening, isocratic mode with a mixture of CO2 and methanol (95:5, v/v) was used at a flow of 1 mL min−1 and with a column-compartment temperature fixed at 30 °C. Results are shown in Fig. 2. PFP and CN stationary phases resulted in coelutions of the seven analytes, indicating that these types of stationary phase are not well suited for T and T3 analysis. DPH, C18, and Si columns resulted in several peak coelutions, with short analysis time (less than 1.5 min) for the last two. Finally, NH2 stationary phase separated the seven vitamin E congeners with an acquisition time of approximately 4 min, although a slight peak fronting was observed. This column was thus selected for further optimization steps.

Fig. 2
figure 2

Chromatograms recorded with Si (a), NH2 (b), CN (c), PFP (d), C18 (e), and DPH (f) columns, with 95:5 % CO2–EtOH (1 mL min−1) as mobile phase, at 35 °C, 1 μL injected at a concentration of 1 mmol L−1 for each T and T3, and UV detection at 295 nm

Full size image

Ethanol and 2-propanol were then tested as modifiers (Electronic Supplementary Material (ESM) Fig. S1). In this case, tocopherol acetate was added to the standard mixture to evaluate its retention time for quantitative purposes. Retention times were significantly longer when using ethanol compared with methanol, enabling baseline separation of all eight compounds. This was even more pronounced with 2-propanol, as expected. Because ethanol has the lowest environmental impact and toxicity, it was chosen for the following method optimization.

The effect of temperature was studied in the range 30–55 °C. An increase in temperature induced an increase in the retention time, as described elsewhere for the separation of vitamin A derivatives using a DPH column and the same mobile phase (ESM Fig. S2) [31]. A minor change of temperature can modify the retention as a result of variation in the thickness and composition of the adsorbed film of mobile phase on the stationary phase [33]. The change of selectivity with temperature might also be attributed to density variation. It must also be noted that the addition of 0.1 % formic acid to the ethanol slightly reduced peak fronting. Finally a temperature of 30 °C was chosen as being optimum and enabling the minimum analysis time.

Flow rate (in the range 1–1.75 mL min−1) was then tested (ESM Fig. S3) and a mobile-phase gradient was set to decrease analysis time while preserving a baseline separation of all eight analytes. From the data shown in Fig. S3, solvent consumption can be evaluated by determining the minimum analysis time, i.e. when the last peak ends. Solvent volumes were determined to be 8.5, 8.375, 8.1, and 8.05 mL for flows of 1, 1.25, 1.5, and 1.75 mL min−1, respectively. The flow was thus set at 1.5 mL min−1, to optimize solvent consumption and because the α-tocopherol acetate is eluted close to the injection peak for higher flow rates. A chromatogram using the fully optimized analytical conditions is shown in Fig. 3. The final analysis time was approximately 5 min; two and eight times shorter than published UPLC [18] and SFC [29] analyses, respectively.

Fig. 3
figure 3

Chromatogram recorded with NH2 column after optimization, with a gradient of CO2–EtOH + formic acid (0.1 %) (1.5 mL min−1), at 30 °C, 1 μL injected at 1 mmol L−1, at a concentration of 1 mmol L−1 for each T and T3, and UV detection at 295 nm

Full size image

To evaluate the linearity of the detection, calibration curves over 2.5 orders of magnitude were determined for each compound (ESM Fig. S4). Excellent linearity, with coefficients of determination (R 2) better than 0.996, was obtained for all the analytes. Limits of detection (LODs) were determined to be 25 pmol and limits of quantification (LOQs) were approximately 50 pmol, except for tocopherol acetate, for which these values were 10 pmol and 25 pmol, respectively. The coefficients of variation (CV) measured for triplicates were always below 10 %, except for the limits of detection (LOD), for which CVs were below 20 %. With an injected volume of 1 μL, the LOQs were in the range 9.9–11.8 ng. These values are very similar to those reported elsewhere for NPLC–UV [15, 16]. It must be noted that the particle diameter of the NH2 column used here is 3 μm. The sensitivity therefore cannot be directly compared with those reported for UPLC–UV with lower particle diameters (1.7 or 1.8 μm) [18].

SFC–MS

Because the final objective is the quantification of T and T3 in complex mixtures, mass spectrometry is necessary to significantly increase the specificity of the analysis. As already mentioned, several articles report the ionization of T and T3 by ESI and APCI in both positive and negative-ion modes [19–22]. However, very few data [34] are available describing the ionization of T and T3 by atmospheric-pressure photoionization (APPI), although this technique is particularly well suited for lipid analysis [35] and hyphenation with NPLC [36]. Thus, we decided to evaluate the ionization efficiency of ESI, APCI, and APPI coupled to the SFC system and to a high-mass-resolution mass spectrometer. For this purpose, several instrumental conditions had to be optimized, including the nature of the make-up solvent, its flow rate, the source temperature, and the flow rate of the drying gas. To evaluate the effect of each condition, the ratio between the peak area measured by MS and that measured by UV was systemically calculated. This enabled us to reduce the variation resulting from the SFC systems and to concentrate on the MS response. To exemplify the effect of each condition, the results for α-tocopherol will be described in detail. The behavior of the other analytes always followed the same trend. As a starting point, instrumental settings recommended by the manufacturer were used. Then optimization steps began with the choice of the best make-up solvent and its optimum flow rate. As shown in Fig. 4, increasing the flow rate led to a reduction in sensitivity for ESI in both ionization modes and for APPI in the positive-ion mode. Normalized peak areas were slightly higher for APPI compared with ESI for the three modifiers in the positive-ion mode. No ion related to T and T3 was detected for APPI in the negative-ion mode, whereas negative ESI was less sensitive than positive ESI. The opposite conclusion has been reported in the literature [21], but this difference could be explained by the different ion-source design leading to different ionization processes. For APCI, no clear tendency was revealed except that ionization in the positive mode is significantly higher than in negative mode. When comparing the best results for each ion source, the APPI source seemed the best choice for α-tocopherol. Interestingly, Bolanos et al. had already reported the great advantage of APPI over the other ion sources for the detection of non-polar compounds in the pharmaceutical field [37]. Hence, further optimization used the APPI source.

Fig. 4
figure 4

Ratio of the MS peak area over the UV peak area for α-tocopherol as a function of the chemical nature and flow rate of the make-up solvent

Full size image

When added at a low flow rate (0.1 or 0.2 mL min−1), methanol and ethanol obtained rather similar ionization efficiency of α-tocopherol (Fig. 4). Because ethanol is substantially less toxic than methanol, ethanol was chosen as make-up solvent at a flow of 0.2 mL min−1. To further increase the sensitivity, a dopant was added to the make-up solvent. The dopant is a small organic molecule with low ionization energy; below the energy of the 10.0 and 10.6 eV photons emitted by the krypton discharge lamp [38]. The dopant can have a major effect on the ionization efficiency by charge or proton transfers mediated by the solvent. We tested several standard dopants (acetone, toluene, and anisole) at 1 % and 5 % in the make-up solvent, but none of them led to a signal enhancement. At higher percentages, normalized peak areas of T and T3 decreased, probably as a result of ionization competition between the analyte and the dopant. Ethanol has an ionization energy of 10.48 eV [39], just below the energy of the second emission ray of the krypton discharge lamp at 10.6 eV. Therefore, ethanol itself could act as a dopant and addition of another was not necessary.

The next steps were the optimization of the vaporizer temperature and the capillary and fragmentor voltages (ESM Fig. S5). The first variable was most important for obtaining the highest sensitivity, whereas a peak intensity increase of less than 10 % was obtained after optimization of the others. The optimized method conditions are summarized in Table 1.

Table 1 Optimized settings for the APPI source
Full size table

Because quantification is performed using the extracted-ion-chromatogram (XIC) method, it is necessary to determine which types of T and T3 ion are produced by APPI. The MS spectrum of α-tocopherol is shown in Fig. 5. Major ion species were [M−H]+, M•+, and [M+H]+. The radical cation M•+, which is at a low relative intensity when the isotopic contribution of the [M−H]+ ion is removed, probably originates either from the direct photoionization of the neutral molecule or from charge transfer from ethanol as a dopant. In fact, the ionization energy of α-tocopherol was estimated to be ~6.7 eV by computational methods [40]. Protonated species originate from the proton transfer between a protonated solvent cluster and the neutral molecule. The [M−H]+ species was already described by ESI, APCI [21], and APPI [34] and probably originates from the dehydrogenation of [M+H]+ ions, as suggested elsewhere [10]. The in-source fragment ion at m/z 165 is the result of a well-described retro-Diels–Alder (RDA) reaction [41]. This fragment-ion peak is only a minor part (less than 5 %) of the total α-tocopherol signal and thus is not taken into account in the following. Consequently, extracted ion chromatograms of [M−H]+, M•+, and [M+H]+ species were reconstructed for each analyte and the sums of the peak areas were used for quantification.

Fig. 5
figure 5

APPI-MS spectra of α-tocopherol extracted from the SFC–MS chromatogram of T and T3 standard mixture

Full size image

Method validation

The method was validated for linearity, detection and quantification limits, selectivity, recovery, and repeatability.

Calibration curves were determined using the internal-standard procedure. Because tocopherol acetate is not naturally present in biological samples, this compound was chosen as an internal standard (IS). Least-squares linear regressions of the ratio between the peak area of the analyte and that of the IS as a function of the analyte concentration at 13 different levels (0.01–100 μmol L−1) were achieved in triplicate experiments. The linearity of the method was assessed for the entire concentration range over four orders of magnitude, with coefficient of determination R 2 better than 0.986 (Table 2). Limits of detection (LODs) and limits of quantification (LOQs) were determined for S/N = 3 and 10, respectively. Our values are of the same order of magnitude as those reported elsewhere, indicating that SFC–APPI-MS is as sensitive as standard LC–MS methods [19, 21, 22].

Table 2 Calibration curves and limits of quantification and detection for the vitamin E congeners using SFC–MS (LOD: limit of detection; LOQ: limit of quantification)
Full size table

The selectivity of the method was then evaluated by ensuring the absence of interfering peaks with different samples. Extracts from vegetable oils were spiked with standards and analyzed. When comparing the extracted MS spectra from the standard and from the spiked extract, no significant difference was observed, indicating that no major compound in the matrix produced strong interference.

Intra-day precision was evaluated by injection of standard solution at three different concentrations (low, medium, and high) in triplicate (Table 3). Relative standard deviation (RSD) was always better than 15 %, except for that of γ-tocopherol at low concentration (<20 %). These values can be regarded as acceptable for control analysis. Inter-day precision (Table 3) was determined for the medium concentration. RSD values were below 15 %, again indicating that the method is suitable for control analysis.

Table 3 Intraday recovery and intraday and interday precision for the vitamin E congeners using SFC–MS (RSD: relative standard deviation)
Full size table

Because vegetable oils free of T and T3 do not exist, final quantification was achieved by the standard-addition method. Known quantities of each standard were spiked into the sample before extraction. Intra-day recoveries (Table 3) were in the range 85–110 % for the high level and 75–105 % for the low level, indicating that the method is suitable for quantification.

Analysis of vegetable oil

The validated method was used for the detection and quantification of T and T3 in soybean oil. Samples were simply diluted by a factor of 50 in pure cyclohexane. Excellent matching between the retention time and the extracted MS spectra of standards and analytes enabled the identification of the different vitamin E congeners. Comparisons of the amount of T and T3 determined by the standard-addition method using SFC–APPI-MS and values reported in the literature are shown in Table 4. Although tocopherols were all quantified, only γ-tocotrienol was detected, but at a concentration below its LOQ. The γ-T was always determined to be the most abundant T in soybean oil; δ-T is the second most abundant, with a concentration approximatively two times lower. It must be noted that β-T was not detected by Wong et al. in the two different soybean oils [18], whereas a low amount was reported elsewhere [13, 27]. This could be explained either by photodegradation or by different extraction procedures; Wong et al. used 2-propanol whereas we used cyclohexane.

Table 4 Vitamin E composition of soybean oils (density: ~0.9)
Full size table

Conclusions and perspectives

An SFC–APPI-MS method for the simultaneous quantification of seven vitamin E congeners was described for the first time. The analytical approach was successfully validated and applied to the detection and quantification of T and T3 in soybean oil using a standard-addition method. Compared with chromatographic separation achieved by NPLC and RPLC, for which large volumes of organic solvents are required, SFC offers the unique advantage of consuming only CO2 and a low percentage of ethanol, which can be regarded as a “green” solvent. The analysis is fast (<5 min) and as sensitive as LC–MS when compared using similar column settings. To increase sensitivity, a column with a lower particle size (<2 μm) could be used. Moreover, a larger volume could be injected when using a column bearing 3 μm diameter particles. However, the commercially available SFC systems are still not fully compatible with such columns [42], and some efforts need to be made to decrease dead volumes. Another possibility, which should be investigated, is coupling of SFC and triple-quadrupole instruments enabling MRM analysis. Finally our method needs to be extended to the quantitative analysis of other biological materials, for example plasma, urine, or tissue extracts, to convince the analytical community that the SFC can have an important function in bioanalysis in the near future.