3.2 Detailed analysis of CCC fractions in the elution range of tocochromanols.
Displacement of the stationary phase (on average 72 mL) was in the typical range of the BTF system (Müller et al., 2018). Sf values in all 11 CCC runs ranged from 67% to 72%, thus being sufficiently reproducible (see Electronic Supporting Material (ESM) Fig. S1). However, slight variations in run time and thus slight shifts in the collected fractions from run to run prompted us to normalize data of the individual CCC runs. This was carried out by subtracting the volume of displaced stationary phase (Vd, here: 72 mL, as an example) from the total volume of mobile phase (LP) that passed the CCC system at the end of the fraction (Vm). Division of this corrected volume by the total coil volume (Vt, here: 236 mL, section 2.4) resulted in the corrected elution volume (CEV) which is independent of individual run-to-run variations (Eq. 1).
CEV =\(\frac{V_{m}\ -\ \ V_{d}}{V_{t}}\ \ x\ \ 100\%=\ \frac{V_{m}\ -\ \ 72\ mL}{236\ mL}\ \ x\ \ 100\%\)(1)
with CEV - corrected elution volume in percent; Vm – volume of mobile phase; Vd – volume of displaced stationary phase; Vt: total coil volume
This normalization procedure of Hammann et al. (2013) allowed us to discuss data of all eleven CCC runs universally by means of CEV.
In head-to-tail mode (lower phase mobile), tocochromanol families were found to elute in the order T3 < T2 < T1 < T [14]. In addition, according to the recently presented equivalent chain length (ECL) rule of tocochromanols, the effect of one db in the side chain corresponded with one methyl group in the 6-chromanol moiety (Vetter et al., 2019). Hence, γ-T1 was expected to co-elute with δ-T and α-T2 (i.e. between α-T3 and α-T1; CEV: ~80%,Fig. 3 ). In order to elute all relevant tocochromanols, CCC fractions were collected between 50-105 min (57-150% CEV). As a benefit, the most prominent squalene-like compounds eluted into the post run fraction and thus could be removed.
Early tocochromanol-containing CCC fractions (57-75% CEV) also featured the more abundant isoprenoid compounds. Hence, purity of the first eluting tocochromanol γ-T3 (1 ) (58-74% CEV, Fig. 3 ) whose identity could be verified with a commercial reference standard, was only ~0.3%, although no other tocochromanols were detected in this CCC elution range. Similarly, the purity of the rare δ-T1 (2 ) (67-97% CEV) was also low but GC/MS analysis of the silylated main fraction (72-74% CEV) allowed its identification via (i) M+ at m/z 472 (base peak, Fig. 4a ), (ii) the diagnostic tropylium cation at m/z 209 and (iii)m/z 69, indicating the position of the db at C-11´ (Fiorentino et al, 2009). The elution range of δ-T1 also featured two γ-T2 isomers (γ-T2 isomer 1 (3 ), 67-81% CEV and γ-T2 isomer 2 (4 ), 67-84% CEV, Fig. 3 ) based on m/z 484 (M+, silylated, Fig. 4,3 ,4 ) and the tropylium cation at m/z 223. Both γ-T2 isomers featured the diagnostic allylic ion at m/z 69 which indicated that the remote db was located on C-11´ (Müller et al., 2020; Butinar et al., 2011). Also, the difference in GC retention times (ΔtR) of 0.7 min (24.2 min for γ-T2 isomer 1, 24.9 min for γ-T2 isomer 2) was similar to the gap between two α-T2 isomers recently discovered in palm oil, namely 3´,11´-α-T2 and 7´,11´-α-T2 (Müller et al., 2020). This analogy produced evidence that the two γ-T2 isomers in PSO were 3´,11´-γ-T2 (isomer 1) and 7´,11´-γ-T2 (isomer 2), respectively. As described by Müller et al. (2020), isomers with db closer to the 13´-end of the side chain usually elute later from GC columns.
However, the PSO sample featured further uncommon tocochromanols. Compounds 5 (tR 25.13 min, 70-74% CEV) and6 (tR 25.28 min, 72-74% CEV) shared the CCC elution range with the γ-T2 isomers. M+ at m/z486 and the diagnostic (silylated) tropylium cation at m/z222/223 indicated the presence of uncommon β- or γ-T1 isomers (β-/γ-T1u). This is remarkable because the classic γ-T1 (7 ) and β-T1 (9 ) isomers were additionally detected in the sample. Compared to them, the uncommon T1 isomers (5 ,6 ) did not elute according to the ECL rule of tocochromanols (Vetter et al., 2019) and also the GC retention time was higher than expected. Therefore, the substitution pattern in the aromatic ring of the T1 isomers (5 ,6 ) could not be established this time. Abundant low mass fragment ions at m/z69,73, 81 and 143 indicated co-elution with isoprenoid substances.
Next in elution was the target compound γ-T1 (7 ) as well as α-T3 (8 ; 72-82% CEV) which was verified by means of a reference standard. γ-T1 (tR: 23.77 min, 72-101% CEV) was identified by means of the silylated M+ atm/z 486 (Fig. 5 , 7 ) which is two atom mass units (2 u) less in comparison with γ-T. Additionally, presence ofm/z 69 indicated that the db was located on C-11´. Recently, it was found that the ratio between m/z  222 and m/z  223 in the GC/MS spectra can be used to distinguish silylated γ- from β-tocopherol (Hammann et al, 2019). Namely, a higher abundance ofm/z  223 compared to m/z  223 was characteristic for silylated γ-T (Fig. 6d , 14 ) while similarly high abundance of m/z 222 and m/z 223 was characteristic for silylated β-T (Fig. 6e , 15 ) (Hammann et al, 2019). Notably, this difference was recently found to be valid for silylated β- and γ-tocochromanols with unsaturated side chain (Müller et al., 2020), and it was also verified in all GC/MS spectra of silylated β-T, γ-T, and other β- and γ-tocochromanols detected in this study. Accordingly, the higher abundance of m/z 223 compared to m/z 2 22 in the GC/MS spectrum of silylated 7 and the fact that β-tocochromanols elute earlier from the GC column than γ-tocochromanols, further verified that the target compound was rather γ-T1 than β-T1 (11´-β-T1, 9 , tR 23.44 min, 75-92% CEV) which was also detected in the sample.
Presence of β-T1 (9 ) has been only mentioned twice before in the literature (Müller et al., 2020; Kruk et al., 2011). The slightly earlier GC/MS elution compared to γ-T1, M+ atm/z 486, similar abundance of m/z 222 and m/z 223, and presence of m/z 69 (Fig. 5 , 9 ) produced strong evidence for its presence in the PSO. Likewise, ΔtR between the potential 11´-β-T1 and 11´-γ-T1 was equivalent to ΔtR between β-T and γ-T, providing further evidence for the presence of 11’-β-T1.
In agreement with predictions, the known δ-T (10 ) co-eluted with γ-T1 contributing up to ~11% to total tocochromanol area in this range. Surprisingly, three more tocochromanols preceded the elution of γ-T (14 ) which was expected to elute next from the CCC system. The first one of these unexpected compounds (11 , tR 24.06 min, 77-92% CEV) featured M+ at m/z 488.4 which is isomeric with β- and γ-T, respectively. However, this is curious because only these two T isomers with two methyl groups in the aromatic ring can exist in theory. Nevertheless, compound 11 featured the diagnostic (silylated) tropylium cation at m/z 222/223 which supported the presence of a 6-chromanol moiety with two methyl groups in the aromatic part. This uncommon Tu (11 ) showedm/z 222 and m/z 223 (the higher abundance of m/z222 was more similar to β-T than γ-T) and it also featured the weakly prominent m/z 263 which is formed by β-/γ-tocochromanols by removal of the entire alkyl chain (Fig. 6a ). Accordingly, the GC/MS spectrum of silylated Tu (11 ) was similar to silylated β-T (15 , tR 22.53 min) and γ-T (14 , tR 22.79 min), but Tu(11 ) eluted much later tR (and even after γ-T1 ((7), tR: 23.77) (Fig. 5 ). Notably, the shift toward longer tR of Tu in GC/MS (1.53 min and 1.27 min relative to β-/γ-T) was similar to ΔtRbetween the uncommon T1 isomers 5 and 6 and β-/γ-T1. Tu (11 ) also eluted earlier from the CCC system than anticipated, namely together with δ-T (10 ) and β-T1 (9 ) (Fig. 3 ), which was also observed for the T1 isomers 5 and 6 . This pointed towards a structural relationship of compounds Tu (11 ), 5and 6 , which was, however, different to classic tocochromanols. One structural option might be formylated tocochromanols, which were isolated by Merza et al. from the stem bark of mangosteen treeGarcinia virgata (Merza et al., 2004) but another option will be discussed below.
In addition, compounds 12 and 13 could be verified to be 3´,11´-α-T2 (12 , tR: 27.73 min, 81-92% CEV) and 7´,11´-α-T2 (13 , tR 28.50 min, 84-92% CEV) due to identical features as described by Müller et al. who characterized these two 11´-α-T2 isomers in palm oil (Müller et al., 2020). Accordingly, the occurrence of two T2 isomers – if T2 is present – seems to be rather usual.
Presence of γ-T (14 , 86-150% CEV), β-T (15 , 89-114% CEV) and the common 11´-α-T1 (16 , 89-116% CEV) could be identified by means of authentic reference standards (Müller et al., 2018; Vetter et al., 2019). Similarly to observations of Müller et al. (2018) in palm oil, 11´-α-T1 was accompanied by much lower amounts of 12´-α-T1 (17 , tR 26.83 min, 99-101% CEV, M+ m/z 500, no m/z 69). Compared to the α-T1 isomers (small difference in tR), the two β-/γ-T1 isomers 3 and 4 discussed before showed a much higher ΔtR which produced strong evidence that both were more different in structure. Finally, α-T (18 , 109-150% CEV) was the last tocochromanol detected in the sample whose late elution is in agreement with the maximum number of methyl groups in the aromatic part and the lack of double bonds in the side chain (Fig. 1a ). Accordingly, all naturally occurring families of α-tocochromanols and γ-tocochromanols (T, T1, T2, T3) were present in PSO. Up to now, this variety, including several new detected tocochromanols is unique for plant oils.
Notably, PSO used for this study was a cold pressed oil from roasted pumpkin seeds. During roasting, pumpkin seed kernels are exposed to temperatures >100 °C and typically at 110 °C under mild conditions (Fruhwirth and Hermetter, 2008). This temperature was confirmed in correspondence with the manufacturer of the used PSO. In view of this thermal treatment, it cannot be excluded that the three uncommon compounds Tu (11 ), 5 and6 , were formed as artefacts during the roasting of the oil. Presence of a much higher number of tocochromanol isomers in dietary supplement capsules from rice bran oil had already indicated the lability of vitamin E compounds during the refining process (Hammann et al., 2016). However, it must be noted that conditions during refining are significantly harsher than the roasting conditions of the PSO used in this study (Van Hoed et al., 2006). Irrespective of the unsolved origins of compounds Tu (11 ), 5 and6 , 15 tocochromanols were found to be native constituents of PSO, and many of them were detected for the first time (Tab. 1 ).