Developing an Enzyme-Assisted Derivatization Method for Analysis of C27 Bile Alcohols and Acids by Electrospray Ionization-Mass Spectrometry

Enzyme-assisted derivatization for sterol analysis (EADSA) is a technology designed to enhance sensitivity and specificity for sterol analysis using electrospray ionization–mass spectrometry. To date it has only been exploited on sterols with a 3β-hydroxy-5-ene or 3β-hydroxy-5α-hydrogen structure, using bacterial cholesterol oxidase enzyme to convert the 3β-hydroxy group to a 3-oxo group for subsequent derivatization with the positively charged Girard hydrazine reagents, or on substrates with a native oxo group. Here we describe an extension of the technology by substituting 3α-hydroxysteroid dehydrogenase (3α-HSD) for cholesterol oxidase, making the method applicable to sterols with a 3α-hydroxy-5β-hydrogen structure. The 3α-HSD enzyme works efficiently on bile alcohols and bile acids with this stereochemistry. However, as found by others, derivatization of the resultant 3-oxo group with a hydrazine reagent does not go to completion in the absence of a conjugating double bond in the sterol structure. Nevertheless, Girard P derivatives of bile alcohols and C27 acids give an intense molecular ion ([M]+) upon electrospray ionization and informative fragmentation spectra. The method shows promise for analysis of bile alcohols and 3α-hydroxy-5β-C27-acids, enhancing the range of sterols that can be analyzed at high sensitivity in sterolomic studies.

One derivatization strategy that has lately become popular is enzyme-assisted derivatization for sterol analysis (EADSA, Scheme 1) [8,9,12,13,17,[40][41][42][43][44][45][46][47]. EADSA technology was designed to add specificity and sensitivity to sterol analysis [40,[44][45][46]. This is achieved by specifically targeting the 3β-hydroxy-5-ene or 3β-hydroxy-5α-hydrogen function in sterols and converting the 3β-hydroxy to a 3-oxo group with bacterial cholesterol oxidase from Brevibacterium or Streptomyces sp. [40,41]. Once introduced, the 3-oxo group is derivatized with the positively charged Girard hydrazine reagent, introducing a charge-tag to the target analyte and improving sensitivity in ESI-MS. A limitation of the existing protocol is that it is not applicable to sterols with a 3α-hydroxy group. Here we describe how this limitation is overcome for the analysis of bile alcohols and C27 bile acids with this stereochemistry. The methodology is potentially applicable for C24 bile acids but requires further optimization to achieve similar sensitivity as for C27 alcohols and acids.

Girard P (GP)-Derivatisation of 3-Oxo Groups
Previous studies have shown that hydrazone formation is very efficient towards 3-oxo-4-ene substrates and to other α,β-unsaturated ketones [40,41,49,50]. In earlier EADSA studies using cholesterol oxidase in phosphate buffer to convert 3β-hydroxy-5-ene sterols to their 3-oxo-4-ene equivalents, GP derivatization was achieved by simply adding methanol to the incubation solution to give a 70% methanol solution and then adding acetic acid and GP hydrazine reagent [40,41,45,46,[50][51][52]. However, the buffer required for 3α-HSD oxidation of the 3α-hydroxy group to the 3-one is 100 mM pyrophosphate buffer, pH 8.9, and upon methanol addition, necessary for subsequent hydrazone formation, a precipitate is formed. This can be avoided by limiting the methanol content to 5%, however, under these conditions hydrazone formation is reversed back to the hydrazine and free carbonyl. For this reason, following incubation with 3α-HSD, samples were desalted on an Oasis HLB reversed-phase column and eluted in methanol, a solvent suitable for subsequent GP hydrazone formation. The GP derivatization efficiency was assessed by LC-MS in the negative-ion mode by comparing the amount of underivatized oxidized acid present before and after the GP derivatization step and incorporating the data into Equation (2).

LC-MS Analysis of Oxidised/Derivatised 3α-Hydroxy-5β-Hydrogen Substrates
In this preliminary study we have not optimized the chromatographic or MS conditions for the GP-derivatized target compounds but rather used existing LC-MS conditions used previously for GP-derivatized sterols [44][45][46]52]. The logic behind this is that by using isotope-labelled GP reagent, the ultimate aim will be to analyze sterols oxidized with cholesterol oxidase or 3α-HSD in a single LC-MS run. Neither have we performed detailed investigations of limits of quantification or linearity of dynamic range in the current study. However, we find that the sensitivity obtained here for 3α-HSD oxidized/GP-derivatized C 27 sterols with a 3α-hydroxy-5β-stereochemistry is of the same order of magnitude to that obtained for GP derivatives generated after cholesterol oxidase treatment of 3β-hydroxy-5-ene substrates. For the C 27 substrates an on-column limit of detection (LOD, signal/noise, 5:1) of 250 fg was achieved. The on-column LOD for the C 24 acid, cholic acid, was 250 pg. More work is required to explain this discrepancy in sensitivity and the even poorer sensitivity with glycine-and taurine-conjugated acids. Optimization of the ion-source conditions for different groups of analytes, or at least compromise in the settings chosen, is likely to be necessary.

MS n Fragmentation
A major driver for the current study was the poor fragmentation properties of unconjugated C 24 and C 27 bile acids under conditions of ESI-tandem MS (MS/MS) at low collision energy (<100 eV) [5,[53][54][55][56] (see also MassBank of North America http://mona.fiehnlab.ucdavis.edu/). This has led to many studies in which the precursor ion at unit mass resolution is also used as the "product ion" for generation of LC-multiple reaction monitoring (MRM) chromatograms. Once derivatized with the GP reagent, both bile acids and bile alcohols fragment under low-energy conditions with the loss of the pyridine group, resulting in [M-Py] + ions (see Scheme 1). These ions can be fragmented further in ion-trap instruments to  3 is that it provides an extra dimension of separation compared to MS 2 , where spectra are a composite of fragment ions derived from desired and undesired coselected precursor ions.
The MS 3 spectra show considerable similarity, with many fragment ions in the spectrum of the C26 acid and tetrol being shifted by m/z 30 and m/z 16, respectively from the corresponding triol. This is explained by the introduction of a carboxylic acid group (+ O2 − H2) or hydroxy (+ O) group to the terminal carbon (C-26) of the sterol side-chain (Scheme 2). Postulated structures of fragment ions for C-3α,7α,12α-triol are shown in Scheme 3 and for C-3α,7α,12α,26-tetrol and CA-3α,7α,12α-triol in Supplemental Schemes S1 and S2, and are listed in Table 1. Reconstructed-ion chromatograms (RICs) and multistage fragmentation (MS 3 ) ([M] + →[M-Py] + →) spectra of oxidized/GP-derivatized 3α-hydroxy-5β-bile alcohols and acids: (a,b) C-3α,7α,12α-triol; (c,d) C-3α,7α,12α,26-tetrol; (e,f) CA-3α,7α,12α-triol; (g,h) cholic acid. The RICs were generated from mass spectra recorded in the Orbitrap mass analyzer at a resolution of 120,000 (FWHM definition at m/z 400), with an m/z window of ± 5 ppm. MS 3 spectra were generated in the linear ion-trap and recorded at the ion-trap detector of the Orbitrap-Elite mass spectrometer. Mass accuracy for fragment ion measurements made with the linear ion-trap is typically ± 0.3 Da. Postulated compositions of fragment ions are listed in Table 1. Note that the data for cholic acid was generated on an earlier version of instrument (i.e., Orbitrap-LTQ) at lower resolution and with reduced mass accuracy.
The MS 3 spectra show considerable similarity, with many fragment ions in the spectrum of the C 26 acid and tetrol being shifted by m/z 30 and m/z 16, respectively from the corresponding triol. This is explained by the introduction of a carboxylic acid group (+ O 2 − H 2 ) or hydroxy (+ O) group to the terminal carbon (C-26) of the sterol side-chain (Scheme 2). Postulated structures of fragment ions for C-3α,7α,12α-triol are shown in Scheme 3 and for C-3α,7α,12α,26-tetrol and CA-3α,7α,12α-triol in Supplemental Schemes S1 and S2, and are listed in Table 1.  Table 1. Note that the data for cholic acid was generated on an earlier version of instrument (i.e., Orbitrap-LTQ) at lower resolution and with reduced mass accuracy.
A key structurally distinct fragment ion for all the 3α,7α,12α-triols is the [A3-H-(H2O)2] + ion (or [A3-H-(H2O)3] + when an additional hydroxy group is at C-25, Table 1), a triply unsaturated carbonium ion consisting of B-, C-and D-rings plus the C17 side-chain, where charge is delocalized across the three double bonds in the ring system (Scheme 3, inset). An equivalent fragment ion is not observed in cholesterol oxidase-oxidized/GP-derivatized 3β,5α,6β-triols ( Figure 3).  Figure S1a-c respectively. Mass spectra recorded at the peak of the RIC for these and other sterols analyzed are shown in Supplemental Figure S2. Data was generated on the Orbitrap-Elite mass spectrometer as in Figure 1. See Table 1 to correlate m/z with fragment ion composition. Unsurprisingly, the MS 3 ([M + →[M-Py] + →) spectrum of the C 24 acid, cholic acid (Figure 1h), shows the same pattern of fragment ions as the C 27 acid CA-3α,7α,12α-triol (Figure 1f) but offset by m/z -42 (-C 3 H 6 ), corresponding to the mass difference between equivalent C 27 3 ] + when an additional hydroxy group is at C-25, Table 1), a triply unsaturated carbonium ion consisting of B-, C-and D-rings plus the C 17 side-chain, where charge is delocalized across the three double bonds in the ring system (Scheme 3, inset). An equivalent fragment ion is not observed in cholesterol oxidase-oxidized/GP-derivatized 3β,5α,6β-triols (Figure 3).

Discussion
In this communication we describe preliminary studies to develop an enzyme-assisted derivatization for C 27 bile alcohols and acids with a 3α-hydroxy-5β-hydrogen stereochemistry. The method still requires further optimization, particularly with respect to the GP-derivatization step which only gave a 45%-60% yield. Despite this, the considerable sensitivity of GP-derivatives makes the moderate yield tolerable. Although the LC-MS sensitivity for detection of C 24 acids was not as good as for C 27 acids, the rich MS 3 fragment ion spectra provide a significant advantage over conventional MS/MS spectra of unconjugated acids where few fragment ions are observed. The on-column detection limit of 250 fg for C 27 analytes translates to a limit of detection of about 0.2 ng/mL if 100 µL of biological fluid is worked up and 1% injected on-column, as in our usual procedure with EADSA [52]. For comparison, Johnson et al. could measure CA-3α,7α,12α-triol, after derivatization to the dimethylaminoethyl ester, at a concentration of about 60 ng/mL in as little as 5 µL of plasma, with 20% injected-on column [38], while DeBarber et al. determined the limit of quantification of 7α,12α-dihydroxy-5β-cholestan-3-one, the 3-oxo form of C-3α,7α,12α-triol, to be 20 ng/mL from 4 µL of plasma after derivatization to the oxime with (O-(3-trimethylammoniumpropyl) hydroxylamine) bromide [39]. We have not yet rigorously tested the repeatability of the EADSA methodology in biological samples. This will become relevant with the availability of isotope-labelled standards, which can be synthesized by methods described by Johnson et al. and by Shoda et al. [61]. Isotope-labelled internal standards will similarly facilitate the progression of the method to a quantitative format. We did not attempt to optimize LC-MS conditions for the GP-derivatives analyzed in this study; instead we used previously optimized conditions for derivatized oxysterols. The logic behind this was to allow the expansion of our sterol profiling method to include bile acids and alcohols derivatized with [ 2 H 0 ]GP after 3α-HSD treatment and oxysterols, and cholestenoic and cholenoic acids derivatized with [ 2 H 5 ]GP after cholesterol oxidase treatment, or vice versa, in a single LC-MS run. At present there are challenges with this strategy, as efficient ionization of glycine-and taurine-conjugated bile acids requires different ion-source conditions from the unconjugated GP-derivatives.
on an Oasis HLB column (60 mg) to remove excess derivatization reagent and was carried out as described in Abdel-Khalik et al. [52].

LC-MS(MS n ) Analysis
LC-MS(MS n ) was performed in the positive-ion mode as described in Abdel-Khalik et al. utilizing the Orbitrap-Elite hybrid MS preceded by a Dionex Ultimate 3000 LC system (Dionex, now Thermo Fisher Scientific) [52]. For analysis of underivatized acids in the negative-ion mode, other than for polarity reversal and a change of column from a Hypersil Gold C 18 to a Kinetex core-shell technology XB-C 18 column (2.6 µm, 2.1 mm × 50 mm, Phenomenex, Macclesfield, UK), the method was as for positive-ion mode LC-ESI-MS(MS n ) as described in Abdel-Khalik et al. [52].

Patents
The derivatization method described in this manuscript is patented by Swansea University (US9851368B2) and licensed by Swansea Innovations to Avanti Polar Lipids and to Cayman Chemical Company.
Supplementary Materials: The following are available online.
Supplemental Figure S1. Data were generated on the Orbitrap-Elite mass spectrometer as in Figure 1.
Supplemental Figure S2. Mass spectra of oxidized/GP-derivatized bile alcohols and acids recorded at the peak of their chromatographic elution.
Author Contributions: All authors contributed to conceptualization of the study, generation and analysis of data, writing reviewing and editing the manuscript.
Funding: This research was funded by the UK Biotechnology and Biological Sciences Research Council (BBSRC, grant numbers BB/I001735/1 to WJG, BB/L001942/1 to YW). JA-K was supported by a PhD studentship from Imperial College Healthcare Charities.