Characterization of various isomeric photoproducts of ergosterol and vitamin D2 generated by UV irradiation

Vitamin D2 is produced from its precursor ergosterol under the impact of ultraviolet (UV) light which is also commercially carried out to increase vitamin D2 contents in mushrooms (‘Novel Food’). However, this process is accompanied by the formation of various isomers that partly co-elute with the target compound and are currently difficult to analyze. For this reason, vitamin D2 and ergosterol were irradiated with the goal to generate and characterize various isomeric photoproducts with three analytical methods. High-performance liquid chromatography with ultraviolet detection (HPLC–UV) was accompanied by using a chiral detector (CD) which was serially linked with the UV detector. Applied for the first time in this research area, HPLC-CD chromatograms provided complementary information which was crucial for the identification of several co-elutions that would have been overlooked without this approach. Additional information was derived from gas chromatography with mass spectrometry analysis. Diagnostic fragment ions in the GC/MS spectra allowed to distinguish four classes of tri- (n = 2), tetra-, and pentacyclic isomer groups. Despite several drawbacks of each of the applied methods, the shared evaluation allowed to characterize more than ten isomeric photoproducts of vitamin D2 including previtamin D2, lumisterol2, tachysterol2,trans-vitamin D2 isomers, and two pentacyclic isomers (suprasterols2 I and II), which were isolated and characterized by proton magnetic resonance spectroscopy (1H NMR).


Introduction
Vitamin D is the summarizing term for bioactive lipid compounds (vitamin D 2 -D 7 ) which are essential micronutrients in human nutrition [1]. These secosteroids vary in the number of carbons (C 27 -C 29 ) and double bonds (∆3-4), as well as the substitution pattern and stereochemistry in the side chain (SC) attached to C-17 ( Fig. 1, SC 2 -SC 7 ) [2,3]. Due to their comparably higher concentrations in foods, cholecalciferol (vitamin D 3 ) and ergocalciferol (vitamin D 2 ) play the predominant role in human nutrition. Both molecules are formed from distinct precursor sterols (i.e., cholesterol in animals and ergosterol in mushrooms) which are eventually transferred into the corresponding vitamins under ultraviolet (UV) radiation [4][5][6]. Hence, vitamin D is also known as the 'sunshine' vitamin, and levels in foods and humans naturally vary in dependence on UV exposure and lifestyle [5,7,8].
Notably, increasing numbers of vitamin D deficiencies in humans have been reported in the Western world [9]. Specifically, only a few dietary sources of animal origin (e.g., fatty fish, cod liver oil, and beef liver) contain noteworthy amounts of vitamin D 3 (1 − 250 µg/100 g fresh weight (fw)) [10,11], while mushrooms are the only relevant source of vitamin D 2 of non-animal origin. However, typical vitamin D 2 contents of edible mushrooms (< 10 µg/100 g fw) are comparably low but they are rich in provitamin ergosterol [12][13][14][15][16].
In the European Union, UV-treated button mushrooms and mushroom powders with increased vitamin D 2 levels * Walter Vetter walter.vetter@uni-hohenheim.de 1 Institute of Food Chemistry, Department of Food Chemistry (170B), University of Hohenheim, 70593 Stuttgart, Germany 1 3 (5-20 µg/100 g in fresh mushrooms and 100-130 mg/100 g in powders) were recently approved as Novel Food (EU 2020/1163 based on EU 2015/2283) [17]. By UV treatment of mushrooms, provitamin ergosterol is converted via C-C bond dissociation into previtamin D 2 and subsequently into vitamin D 2 [18]. Next to vitamin D 2 formation, various isomeric by-products [18][19][20] can be formed which are difficult to characterize due to their complexity and comparably low amounts. Several co-elutions are known to exist in highperformance liquid chromatography which cannot even be resolved by LC/MSMS which currently represents the 'gold standard method' in vitamin D analysis [21]. Contrary to vitamin D 3 which is also analyzed for metabolites in blood serum [13,16,[22][23][24], knowledge on liable and comprehensive analytical methods for isomeric photoproducts of vitamin D 2 is only scattered found and difficult to combine [25][26][27][28].
To overcome these difficulties, UV irradiation experiments were performed with ergosterol and vitamin D 2 . A rather long irradiation time was chosen in order to generate a high variety of isomeric photoproducts in sufficient amounts. Reaction products were investigated by high-performance liquid chromatography (HPLC) with diode array detector (DAD) and gas chromatography with mass spectrometry (GC/MS). In addition, HPLC with circular dichroism (CD) detection was used for the first time in this research area. Selected compounds were isolated by repeated HPLC fractionation and analyzed by 1 H NMR.

Irradiation of standard solutions ergosterol and vitamin D 2
UV lamps (n = 2) with a main emission line of λ max = 310 nm were from Waldmann (Villingen-Schwenningen, Germany). The distance between the two UV lamps and samples was 30 cm, respectively. Standard solutions of ergosterol and vitamin D 2 were prepared in ethanol (1 mg/mL, respectively) and placed in UV transmitting quartz glass vessels. Initial experiments with different exposure times showed that a long irradiation time of 60 min was best suited to generate relatively high amounts and a high variety of photoproducts. Therefore, standard solutions were irradiated for 60 min except in the case of preparative HPLC fractionation and subsequent 1 H NMR analysis (90 min irradiation time).
Due to the serial connection, HPLC retention times in the CD detector were ~ 30 s higher than in the UV detector. An analytical ET 250/8/4 Nucleosil 7 C 18 column (250 mm × 4 mm i.d. × 7 µm particle size, Macherey-Nagel, Düren, Germany) was used in combination with a binary gradient system (A: acetonitrile, B: methanol) according to Wittig et al. [25]. The solvent flow was set to 1 mL/min with 100% A at the beginning for 10 min. Then, over 10 min to 95% A and 5% B, which was held for a further 10 min. Finally, again to 100% A within 5 min and isocratic (100% A) for 8 min to equilibrate the system. Data were collected in the full spectral range of 190-800 nm (DAD) and at an excitation wavelength of 278 nm (CD). The HPLC fractionation of irradiated standard solutions (ergosterol or vitamin D 2 ) for subsequent GC/MS analysis was based on positive and negative signals in the HPLC-CD detector. At these points, the effluent was manually collected in different brown glass flasks (1 mL and 25 mL depending on fraction size).

Characterization strategy and quality control via 1 H NMR analysis
Photoproducts characterized either with available reference standards (i.e. previtamin D 2 , vitamin D 2 , tachysterol 2 , lumisterol 2, and 5,6-trans vitamin D 2 ) or 1 H NMR were assigned to level A (structure verified; il-A, not labeled in the following). All other isomers in UV irradiated solutions were specified by HPLC (λ max in UV spectra, retention times, and positive/negative signals in the CD detector) and GC/MS data (full scan mass spectra and relative retention times). In case of a good match with literature values, they were labeled identification level B (il-B) compared to il-C for tentative assignments.

HPLC-UV and HPLC-CD analysis of ergosterol and vitamin D 2 standards before and after UV irradiation
HPLC-UV chromatograms (reversed phase mode) of ergosterol and vitamin D 2 standards verified the much faster elution of the more polar tricyclic vitamin D 2 (3) (Δt R ~ 20 min, Fig. 3a,c) [25], and both isomers produced negative peaks in the HPLC-CD chromatogram (Fig. 3b,d). The split peak in the HPLC-CD chromatogram of the vitamin D 2 standard ( Fig. 3b) indicated the presence of two co-eluting compounds which could not be distinguished in the UV detector (Fig. 3a). Noteworthy, another vitamin D 2 standard from a different supplier did not show this split peak (Fig. S1, supporting information). Interestingly, in the case of vitamin D 3 (cholecalciferol), HPLC co-elution of 3α-and 3β-hydroxy epimers of 25-hydroxy-vitamin D 3 was recently described by Al-Zohily et al. [33]. The details of the C3-epimerization process are not fully understood but can reportedly occur at all major vitamin D metabolites [33,34]. The presence of the corresponding and hitherto unreported 3-hydroxy epimers of vitamin D 2 could thus explain our measurements, and this observation directly underlined the benefits the HPLC-CD detector added to the present study.
HPLC-UV chromatograms of UV irradiated ergosterol and vitamin D 2 together featured four additional signals, which were eluted before vitamin D 2 (peak #1 and peak #2), co-eluted with vitamin D 2 (peak cluster #3), or eluted between vitamin D 2 and ergosterol (peak #4) (Fig. 3e,g). HPLC analysis of additional standards (Sect. 2.1) verified that peaks #1 and #2 originated from 5,6-trans vitamin D 2 (7) and previtamin D 2 (2), respectively, due to identical HPLC retention times and UV spectra (λ max are given in Table 1 and UV spectra in Fig. S2, supporting information). The broad and fronting peak at the retention time of vitamin D 2 (3) in both HPLC-UV chromatograms indicated that several compounds including c(+)Zt trans-vitamin D 2 (peak #3b) as well as lumisterol 2 (peaks #3c) and tachysterol 2 (peak #3d)-both verified by standards-coeluted with vitamin D 2 (peak #3a) (Fig. 3e, g). This could be substantiated by five co-eluting compounds detected in the HPLC-CD chromatograms (Fig. 3f, h). While vitamin D 2 (peak #3a) and ergosterol (peak #5) standards generated negative signals in the CD (Fig. 3b, d), most newly generated peaks in the irradiated solutions including peaks #3b-d generated positive signals in the HPLC-CD detector (Fig. 3f, h). In agreement with that, the UV peak purity test and UV full spectrum analysis at HPLC-CD retention times of peak #3a,c (generated by UV irradiation of ergosterol, Fig. 3h) showed that vitamin D 2 (peak #3a, λ max at 265 nm) co-eluted with lumisterol 2 (peak #3c, λ max at 280 nm and 271 nm) (Figs. S1 and S2, supporting information). According to the CD signal, vitamin D 2 was only present at trace levels in both UV irradiated solutions, although the HPLC-UV chromatogram, which did not allow to distinguish both forms, showed an abundant peak at the corresponding retention time (Fig. 3e , the tetracyclic photoisomer lumisterol 2 (4) and the thermal isomerization products pyro-(10) and isopyro vitamin D 2 (11) and the pentacyclic photoproducts suprasterol 2 I (8) and II (9). Asterisk According to IUPAC nomenclature, the letter 'Z' specifies the orientation of the ∆5,6 double bond to the ring system which bears the 3-OH group. Letters before and after 'Z' refer to cis-(c) or trans-(t) orientation of the other double bonds (at ∆10,19 and ∆7,8, respectively) relative to the one at ∆5,6. Subscript letters 2 refer to the side chain as found in vitamin D 2 . Blue bullet Isomers marked with a blue dot (6, 7) are isomerized to pyro-and isopyro products of unknown structure at GC injection CD detection (black lines) of standard solutions of vitamin D 2 and ergosterol in ethanol (a-d) before and (e-h) after UV irradiation (60 min) with 5,6-trans vitamin D 2 (peak #1), previtamin D 2 (peak #2), and the non-isomeric ergosta-3,5,7,9,22-pentaene (peak #4). Peak #3 consisted of five isomers which were eventually assigned to vitamin D 2 (peak #3a; (c−)Zt form), (transvitamin D 2 (peak #3b; (c+) Zt-vitamin D 2 ), lumisterol 2 (peak #3c), tachysterol 2 (peak #3d), and suprasterol 2 (peak #3e). Subscript letters 2 refer to the side chain as found in vitamin D 2      n/a g). At this point, it became obvious that HPLC-UV data based on single wavelengths were not sufficient for peak characterization. Subsequent fractionation of peak #3 according to the HPLC-CD signal or recording of the UV full range spectrum (190-800 nm) followed by a peak purity test allowed to elaborate further details. Accordingly, peak #3b in the UV irradiated vitamin D 2 solution showed λ max at 265 nm which is in full agreement with vitamin D 2 (Figs. S1 and S2, supporting information). However, the positive signal in the CD detector (Fig. 3f) unequivocally showed that the bulk of peak #3b could not originate from vitamin D 2 [which produced a negative signal in the CD, (#3a, Fig. 3b)] but indicated the presence of a vitamin D 2 conformer without differences in the conjugated system. In line with that, previous investigations on the photochemistry of vitamin D (not specified in which form) reported the existence of two stable vitamin D conformers, namely c(−)Zt-vitamin D (3, Fig. 2) and the less twisted c(+)Zt trans-vitamin D (6) [18,32]. These two compounds were tentatively assigned to peaks #3a and #3b in Fig. 3f. The vitamin D conformers (3, 6, Fig. 2) differ in the orientation of the cisoid diene moiety [18,32] which could be the reason for the reversal of the CD signal while λ max (265 nm) remained the same (Fig. S3, supporting information). Accordingly, at this long irradiation time of 60 min, vitamin D 2 (3) was barely detected at all and predominantly transformed into c(+)Zt trans-vitamin D (6, Fig. 2). This indicated the strong dependency of vitamin D 2 formation and degradation on the reaction conditions and, again, the value added by implementing the CD detector.
In the following, we aimed to minimize these limitations of HPLC and GC by a shared evaluation of UV irradiated samples with both methods.

HPLC and GC/MS evaluation of photoproducts derived from the UV irradiation of vitamin D 2 (over-irradiation products)
GC/MS chromatograms of UV irradiated standards indicated isomeric photoproducts other than those mentioned above produced only one signal. Specifically, the GC/MS chromatogram of irradiated vitamin D 2 solution featured four additional peaks, three of which will be characterized in this section. Two prominent peaks of suprasterol 2 I (8) and suprasterol 2 II (9) in the GC/MS chromatogram (Fig. 4b) gave no signal in the HPLC-UV chromatograms which indicated the lack of conjugated double bonds. However, analysis of HPLC fractions by GC/MS allowed us to determine their HPLC retention times (i.e. within the peak #3 cluster) and subsequent isolation by preparative HPLC (Sect. 2.2) and 1 H NMR analysis confirmed their structures (Fig. S4, supporting information). Namely, the chemical shifts verified the pentacyclic backbones as reported by Kobayashi et al. [28]. Moreover, the peak abundance ratio of suprasterol 2 II (9) to suprasterol 2 I (8) in the present GC/ MS chromatograms of ~ 3: 1 (Fig. 4b) was in accordance with the one reported by Okamura et al. (2001) [27]. A small negative CD signal indicated the presence of the suprasterol 2 isomers (8,9) in the irradiated vitamin D 2 solution (peak #3e, Fig. 3f) but underrated their high relevance according to GC/MS. Last but not least, the identical GC/MS spectra of the suprasterol 2 isomers (M + , trimethyl silylated, m/z 468) featured the diagnostic base peak at m/z 118 [M-TMSiO-H-259] + with the more abundant suprasterol 2 II (9) eluting first [28,[35][36][37]. A further diagnostic GC/MS fragment ion of silylated suprasterols 2 was detected at m/z 209 which still included the TMSiO group [M-259] + (Table 1). In HPLC, suprasterol 2 II (peak #3e) eluted slightly after and partly coeluted with c( +)Zt trans-vitamin D (6) which produced the prominent positive peak #3b (Fig. 3f). The third newly formed peak formed during the UV irradiation of vitamin D 2 originated from 5,6-trans vitamin D 2 (7) (verified by a standard). In GC/MS, 5,6-trans vitamin D 2 (7) and c(+)Zt trans vitamin D 2 (6) were converted into pyro-and isopyro-vitamin D 2 (i.e. the same compounds, Fig. 4b). The pyro form of the trans-isomers left the column slightly earlier than isopyro vitamin D 2 (Fig. 4b). In HPLC, the isomers 5,6-trans vitamin D 2 (peak #1; 7) and c(+)Zt trans-vitamin D (peak #3b; 6) were resolved by ∆t R ≈ 3 min (Fig. 3f). Furthermore, 5,6-trans vitamin D 2 (7) and c(+)Zt trans-vitamin D 2 (6) with three conjugated double bonds each, differ in the number of alkyl substituents (Fig. 2). The higher number of five alkyl substituents on the conjugated triene system in the case of 5,6-trans-vitamin D 2 (7) versus four in c(+)Zt trans-vitamin D 2 (6) could be the reason for the shift of λ max from 265 nm (6) to 272 nm (7) (Fig. 2, Table 1) [38].
Again, λ max at 265 nm in combination with the positive signal detected at this retention time in the CD detector (Fig. 3f) unequivocally verified that the bulk of peak #3b originated from c(+)Zt trans-vitamin D and not from vitamin D 2 . In this context, it is important to note that vitamin D 2 was more abundant at shorter irradiation times (data Fig. 4 GC/MS full scan chromatograms of (a) trimethylsilylated standard solutions of vitamin D 2 (quantitatively isomerized to give in pyro-and isopyro vitamin D 2 ), tachysterol, and ergosterol and the resulting solutions after UV irradiation (1 h in ethanol) of (b) vitamin D 2 and (c) ergosterol. * and ** refer to coelutions as shown above the peaks. #A-#D are four minor isomers that could not be fully characterized 1 3 not shown). Given the fact that all compounds of the peak #3 cluster in HPLC (virtually the same t R ) were isomers, a differentiation could not be achieved by LC/MS (Fig. S5, supporting information). Additional HPLC measurements with a second connected HPLC column and UHPLC/HRMS measurements (Fig. S5, supporting information) showed that the separation of some (i.e. tachysterol 2 and lumisterol 2 ) but not all isomers (i.e. c(+)Zt and c(−)Zt isomers) can be improved using advanced chromatographic systems (e.g. UHPLC or 2D-LC). However, problems in the identification of the isobaric substances remained due to the identical mass spectra which were dominated by a prominent [M + H] + ion at m/z 397.3461 (Fig. S5, supporting information). This example again illustrates the high value added by implementing the CD detector.
Next to the compounds discussed so far, four additional isomeric photoproducts (isomers #A-D), trimethylsilylated M + at m/z 468, Table 1) were detected by GC/MS. The GC/ MS spectrum of isomer #D was similar to the one of pyro and isopyro vitamin D 2 (10, 11), ergosterol (1), and lumisterol (4) which was surprising because these four tetracyclic isomers cover all structural variants in 9 and 10 positions, respectively (Fig. 2). The GC/MS spectrum of isomer #C was similar to trans-vitamin D 2 , (6) and included the diagnostic base peak at m/z 119 (dimethyl-substituted tropylium cation) diagnostic for tricyclic, 5,7-dienic trans-vitamin D forms (Table 2) which could be a 3-hydroxy epimer of trans-vitamin D 2 due to the similar GC retention time as likewise observed for tentatively identified vitamin D 2 3-hydroxy epimer (Sect. 3.1). Isomer #A (present in both UV irradiated solutions, Fig. 4b,c) showed the base peak at m/z 253 [M-SC-TMSOH] + , which is also found in (silylated) tachysterol 2 (5) but otherwise, both isomers differed in their fragmentation patterns. Finally, isomer #B showed a unique GC/MS spectrum with the base peak at m/z 143.
In addition, a few smaller peaks originated from nonisomeric compounds or impurities already present in the standard. Exemplarily, two non-isomeric compounds (trimethylsilylated M + at m/z 466, base peak m/z 376, Table 1, and non-silylated M + at m/z 376, base peak m/z 251, peak #4 in Fig. 3g,h, Table 1) were detected which were most likely dehydrogenation products of ergosterol.

Conclusion
HPLC and GC/MS methods were used for the characterization of isomers of vitamin D 2 formed by the UV-irradiation of ergosterol and the title compound. Both separation methods, HPLC and GC, were found to have drawbacks. For instance, co-elution of vitamin D 2 with up to four isomers was observed in HPLC, which may have implications when only LC/MS is used for vitamin D 2 determination. These existing problems could be addressed by the serial linking of UV and CD detectors. UV full range spectra enabled the classification of several isomers by λ max . However, suprasterol 2 I and II could not be detected due to the lack of conjugated double bonds in these main irradiation products of vitamin D 2 . Particularly valuable proved to be the CD detector which was used for the first time in this research area. The serial linking of the CD after the UV detector allowed to indicate several co-elutions with vitamin D 2 , by means of the positive and negative signals. However, one disadvantage of the CD was the weak correlation between peak abundance and the amount of a compound. This problem could be addressed by the application of GC/MS. Despite the isomerization of vitamin D 2 and related isomers at mandatorily high temperatures in GC analysis, the good separation was paired with a comparably objective impression of the amounts of the isomers of vitamin D 2 and ergosterol. GC/MS allowed indicating the suprasterol 2 isomers which could be isolated by HPLC and characterized by 1 H NMR. After addressing the benefits and drawbacks of the individual methods, the shared application and evaluation of HPLC-UV, HPLC-CD, and GC/MS enabled the detection of > 10 additional isomers in irradiated ergosterol and vitamin D 2 solutions. The methods and data generated in this study are ready for being used in the evaluation of vitamin D 2 and isomeric photoproducts in food samples such as (UV-irradiated) mushrooms which are the most important non-animal source of vitamin D in human nutrition.