Use of Trifluoro-Acetate Derivatives for GC-MS and GC-MS/MS Quantification of Trace Amounts of Stera-3β,5α,6β-Triols (Tracers of Δ5-Sterol Autoxidation) in Environmental Samples

Stera-3β,5α,6β-triols make useful tracers of the autoxidation of Δ5-sterols. These compounds are generally analyzed using gas chromatography–mass spectrometry (GC-MS) after silylation. Unfortunately, the 5α hydroxyl groups of these compounds, which are not derivatized by conventional silylation reagents, substantially alter the chromatographic properties of these derivatives, thus ruling out firm quantification of trace amounts. In this work, we developed a derivatization method (trifluoroacetylation) that enables derivatization of the three hydroxyl groups of 3β,5α,6β-steratriols. The derivatives thus formed present several advantages over silyl ethers: (i) better stability, (ii) shorter retention times, (iii) better chromatographic properties and (iv) mass spectra featuring specific ions or transitions that enable very low limits of detection in selected ion monitoring (SIM) and multiple reaction monitoring (MRM) modes. This method, validated with cholesta-3β,5α,6β-triol, was applied to several environmental samples (desert dusts, marine sediments and particulate matter) and was able to quantify trace amounts of 3β,5α,6β-steratriols corresponding to several sterols: not only classical monounsaturated sterols (e.g., cholesterol, campesterol and sitosterol) but also, and for the first time, di-unsaturated sterols (e.g., stigmasterol, dehydrocholesterol and brassicasterol).


Introduction
Autoxidation (free radical oxidation) of ∆ 5 -sterols mainly affords 7α-and 7βhydroperoxides and, to a lesser extent, 5α/β,6α/β-epoxysterols and 3β,5α,6βtrihydroxysterols [1]. The 7α-and 7β-hydroperoxides have been ruled out as possible markers of autoxidation processes in the environment due to their instability (fast degradation under environmental conditions) and lack of specificity (formation is also possible by allylic rearrangement of photochemically produced 5α-hydroperoxysterols) [2,3]. Unfortunately, the highly specific 5α/β,6α/β-epoxysterols have also been ruled out as they are too unstable under environmental conditions; they are quickly hydrolyzed to their corresponding triols by epoxide hydrolase [4] and under acidic conditions [5]. The 3β,5α,6βtrihydroxysterols, which are stable and only produced during autoxidation processes, have thus been proposed as specific tracers of sterol autoxidation in the environment [2,3].
Electron ionization (EI) provides more structural information than the soft ionization techniques such as electrospray ionization (ESI) or atmospheric pressure chemical ionization (APCI) employed in HPLC-MS analyses [6], and so quantification of ∆ 5 -sterols and their oxidation products in environmental samples is most often performed using gas chromatographyelectron ionization mass spectrometry (GC-EIMS). GC-EIMS analyses are generally carried out on a nonpolar silicone stationary phase after silylation [2,5,[7][8][9]. Silylation of sterol involves the replacement of the hydrogen of the hydroxyl group with an alkylsilyl (often trimethylsilyl) group. Trimethylsilyl (TMS) derivatives are highly volatile, thermally stable and present outstanding gas chromatographic characteristics. Moreover, the EI mass spectra of these derivatives often exhibit a significant [M-15] + ion formed by the loss of a siliconbonded methyl group, which is very useful for determining molecular mass, and are also very informative for structural elucidations [10,11]. However, since TMS derivatives can lose easily trimethylsilanol molecules under the effect of moisture, a short delay between derivatization and injection is needed. Despite this drawback, silylation is a very popular derivatization method and is often employed during sterol quantification using GC-MS [12][13][14][15][16]. Unfortunately, steric hindrance makes complete silylation of 3β,5α,6β-trihydroxysterols difficult, and common silylation reagents (such as bis(trimethylsilyl)trifluoroacetamide (BSTFA)/pyridine) afford derivatives that are only silylated at C-3 and C-6 [17]. Trisilylated derivatives can be obtained after treatment with BSTFA/dimethylsulfoxide (DMSO) [18], but conversion is still not complete (yields are close to 50%) and this treatment is still too complex to be applied for analysis of trace amounts in environmental samples. It is absolutely necessary to eliminate the DMSO before injection into the chromatographic column, and this operation (requiring: (i) addition of water, (ii) extraction with solvents and (iii) resilylation of 3β and 6β hydroxyl groups) cannot be carried out without significant losses of the lipidic material. The presence of a polar nonderivatized hydroxyl group at C-5 in the disilylated derivative strongly alters its chromatographic characteristics and leads to the formation of tailing peaks that substantially limit the sensitivity of the analyses [19].
Acetylation of steroids with acetic anhydride and trifluoroacetic anhydride is also very common [20][21][22][23]. However, it is generally considered that trimethylsilyl derivatives are more suitable for the GC-MS characterization and quantitation of sterols than acetate derivatives [16]. Acetylation involves replacement of the mobile hydrogen atoms of the hydroxyl groups of sterols with acyl or trifluoro acyl groups. Halogenated acyl groups enhance the electron affinity of the derivative and produce very specific fragmentation patterns in mass spectrometry [24]. It should be noted, however, that the use of fluorinated anhydrides requires the removal of any excess or byproducts prior to GC analysis to prevent deterioration of the column [25]. To our knowledge, trifluoroacetylation has not been used in the case of 3β,5α,6β-steratriols.
In this work, we set out to develop a trifluoroacetylation method able to derivatize the three hydroxyl groups of 3β,5α,6β-trihydroxysterols in order to reduce analyte adsorption in the GC system and improve detector response, peak separation and peak symmetry. We used trifluoroacetic anhydride, which is well known to be highly reactive in the case of steric hindrance [25]. This derivatization technique was then validated using environmental samples (desert dusts, marine sediments and particulate matter), where it allowed the detection of traces of several triols resulting from the oxidation of mono-and di-unsaturated sterols.

Formation and Characterization of Trifluoroacetate Derivative of Cholesta-3β,5α,6β-Triol
Reaction of cholesta-3β,5α,6β-triol with trifluoroacetic anhydride in tetrahydrofurane (THF) under the conditions described in Section 3.2 afforded a trifluoroacetate derivative at high yield (>95%). As expected, this derivative presented better chromatographic characteristics (shorter retention time and better peak shape) than the corresponding bis-trimethylsilyl ether (Figure 1). It is well known that the introduction of fluorine atoms strongly enhances analyte volatility and thus reduces analyte retention time [26]. Due to its high content of fluorine atoms (nine per molecule), the trifluoroacetate derivative of cholesta-3β,5α,6β-triol eluted 9 min faster than the corresponding disilylated derivative ( Figure 1) and 1.5 min faster than cholesterol trifluoroacetate on the 30 m capillary column employed.
( Figure 1) and 1.5 min faster than cholesterol trifluoroacetate on the 30 m capillary column employed.
Although negative inductive effects of the fluorine atoms in a derivatized product may drive hydrolysis in the presence of moisture [27], here, the trifluoroacetate derivative of cholesta-3β,5α,6β-triol was found to be highly stable. Indeed, in contrast to the corresponding TMS derivative, which was hydrolyzed in a few days, it could be stored at 4 °C for several months without significant alteration. The EI(TOF) mass spectrum of the cholesta-3β,5α,6β-triol trifluoroacetate derivative ( Figure 2A) exhibited ions at m/z 594.3134 (b +• ), 480.3209 (c +• ) and 366.3273 (d +• ) corresponding to the successive loss of one, two and three neutral molecules of trifluoroacetic acid by the molecular ion (a +• ), respectively ( Figure 3). Note that the abundance of the c +• ion resulted from the formation of a stable conjugated enol ester group. An ion at m/z 367.1876 (e + ) resulting from the loss of two molecules of trifluoroacetic acid and the sidechain was also formed. The shift of ions b +• , c +• and d +• by 7 m/z units and the lack of shift of the e + ion observed in the EI(TOF) mass spectrum of cholest-5-en- 25,26,26,26,27,27,27-d7-3β,5α,6β-triol trifluoroacetate derivative ( Figure 2B) further supports these attributions. Unfortunately, due to its instability under electron impact, the molecular peak of the cholesta-3β,5α,6β-triol trifluoroacetate derivative was not observable in its EI(TOF) mass spectrum. We therefore used electron-capture negative ionization (ECNI), which is generally considered a soft ionization technique that yields a mass spectral pattern with less fragmentation than EI ionization [28]. The ECNI mass spectrum of the cholesta-3β,5α,6β-triol trifluoroacetate derivative ( Figure 2C) appeared to be dominated by a peak Although negative inductive effects of the fluorine atoms in a derivatized product may drive hydrolysis in the presence of moisture [27], here, the trifluoroacetate derivative of cholesta-3β,5α,6β-triol was found to be highly stable. Indeed, in contrast to the corresponding TMS derivative, which was hydrolyzed in a few days, it could be stored at 4 • C for several months without significant alteration.
The EI(TOF) mass spectrum of the cholesta-3β,5α,6β-triol trifluoroacetate derivative ( Figure 2A) exhibited ions at m/z 594.3134 (b +• ), 480.3209 (c +• ) and 366.3273 (d +• ) corresponding to the successive loss of one, two and three neutral molecules of trifluoroacetic acid by the molecular ion (a +• ), respectively ( Figure 3). Note that the abundance of the c +• ion resulted from the formation of a stable conjugated enol ester group. An ion at m/z 367.1876 (e + ) resulting from the loss of two molecules of trifluoroacetic acid and the sidechain was also formed. The shift of ions b +• , c +• and d +• by 7 m/z units and the lack of shift of the e + ion observed in the EI(TOF) mass spectrum of cholest-5-en-25,26,26,26,27,27,27-d 7 -3β,5α,6β-triol trifluoroacetate derivative ( Figure 2B) further supports these attributions. Unfortunately, due to its instability under electron impact, the molecular peak of the cholesta-3β,5α,6β-triol trifluoroacetate derivative was not observable in its EI(TOF) mass spectrum. We therefore used electron-capture negative ionization (ECNI), which is generally considered a soft ionization technique that yields a mass spectral pattern with less fragmentation than EI ionization [28]. The ECNI mass spectrum of the cholesta-3β,5α,6βtriol trifluoroacetate derivative ( Figure 2C) appeared to be dominated by a peak at m/z 113 corresponding to the anion CF 3 -COO − and a smaller molecular peak at m/z 708, attesting that the observed derivative was well triacetylated. at m/z 113 corresponding to the anion CF3-COO − and a smaller molecular peak at m/z 708, attesting that the observed derivative was well triacetylated.     Figure 3. Proposed fragmentation of cholesta-3β,5α,6β-triol trifluoroacetate derivative. Note that another mechanism (not shown) involving initial loss of the 3β-acyl group is also possible (• = radical, +• = radical cation, + = cation, − = loss).
Based on its abundance ( Figure 2A) and specificity, the c +• ion corresponding to [M-2CF3COOH] +• was selected as the target ion for selected ion monitoring (SIM)-based quantification of the main 3β,5α,6β-steratriol trifluoroacetate derivatives present in environmental samples. Due to its high specificity, the less abundant b +• ion corresponding to

Validation of the Derivatization Method
Validation of the derivatization method was carried out using the cholesta-3β,5α,6βtriol trifluoroacetate derivative. Results of linearity tests in SIM and MRM modes are presented in Table 2. In the concentration range tested here (0.2325-46.5 ng/mL), the coefficients of determination of the linear regression curves were better than 0.995 and the intercepts did not differ significantly from 0.

Validation of the Derivatization Method
Validation of the derivatization method was carried out using the cholesta-3β,5α,6βtriol trifluoroacetate derivative. Results of linearity tests in SIM and MRM modes are presented in Table 2. In the concentration range tested here (0.2325-46.5 ng/mL), the coefficients of determination of the linear regression curves were better than 0.995 and the intercepts did not differ significantly from 0.  Table 3 reports the reproducibility of this derivatization technique. The precision (given by the standard deviation) and accuracy (defined as the difference between obtained concentration and expected concentration) were acceptable over the concentration range. The limit of detection (LOD) (defined by a signal-to-noise ratio of 5) was about 25.8 and 0.78 pg injected in SIM and MRM modes, respectively. For comparison, the LOD obtained for the corresponding disilylated derivative was 0.62 ng in SIM mode with the target ion m/z 456 corresponding to [M-TMSOH-H 2 O] +• . Due to the higher specificity of their MRM transitions, trifluoroacetate derivatives are much more suitable for the analysis of trace amounts of 3β,5α,6β-steratriols in complex samples than silylated derivatives.
The standard solutions of cholesta-3β,5α,6β-triol and internal standard were prepared by dissolving 10 mg measures of these compounds (weighted) in 10 mL of methanol. Dilutions were also carried out in methanol.

Environmental Samples
Detailed descriptions of the collection of samples of desert dusts, marine particulate matter and sediments used for validation of the proposed 3β,5α,6β-steratriol derivatization method can be found elsewhere [39][40][41][42]. Treatment of the whole material of the different samples involved reduction with excess NaBH 4 in MeOH (25 mL; 30 min) to convert labile hydroperoxides (resulting from oxidation) to their corresponding alcohols, which are more amenable to analysis using GC-EIMS, GC-EIMS/MS and GC-QTOF. Water (25 mL) and KOH (2.8 g) were then added and the resulting mixture saponified by refluxing (2 h). After cooling, the mixture was acidified (HCl, 2 N) to pH 1 and extracted with dichloromethane (DCM; 3 × 20 mL). The combined DCM extracts were dried over anhydrous Na 2 SO 4 , filtered and concentrated via rotary evaporation at 40 • C to give TLEs. All the solvents (pesticide/glass distilled grade) and reagents (Puriss grade) were obtained from Rathburn (Walkerburn, Scotland) and Sigma-Aldrich (Saint Quentin Fallavier, France), respectively. The different TLEs obtained were derivatized as described in the following section.

Trifluoroacetylation Method
In an effort to optimize the derivatization reaction, we tested several parameters, including the nature of the solvent (cyclohexane, tetrahydrofuran (THF), diethyl ether, dichloromethane, ethyl acetate and 1,4-dioxane), reaction temperature (50-100 • C), heating time (1-24 h) and volume of TFAA (25-200 µL). Although this reaction could be also carried out with pentafluoropropionic anhydride, we selected TFAA as the derivatizing reagent since it allowed the formation of fluorinated derivatives with a better yield (>95%). The best reaction efficiency was obtained with the following conditions. Samples to be derivatized (after evaporation to dryness under a stream of nitrogen at 50 • C) (2-100 ng), internal standard (66 ng), anhydrous THF (200 µL) and TFAA (100 µL) were put in glass vials (4 mL) with PTFE-lined screw caps, and the mixtures were maintained at 68-70 • C in a heating block for 24 h. After evaporation to dryness under a stream of nitrogen at 50 • C, the residues were dissolved in BSTFA to silylate the traces of trifluoroacetic acid formed during the reaction that could damage the GC column employed.

Silylation
3β,5α,6β-steratriols were silylated by dissolving them in 300 µL measures of a mixture of pyridine and BSTFA (2:1, v/v) and heating to 50 • C for 1 h. After evaporation to dryness under a stream of N 2 at 50 • C, the derivatized residues were dissolved in BSTFA.
ECNI analyses were carried out on the same apparatus with methane as the reagent gas at 50 mA emission current and 195 eV electron energy. During the experiment, the temperature of the source was held at 150 • C and reactant gas flow was 0.5-0.7 mL min −1 .