Characterization of brewer's spent grain extracts by tandem mass spectrometry and HPLC‐DAD: Ferulic acid dehydrodimers, phenolamides, and oxylipins

Abstract Brewer's spent grain (BSG) is a major by‐product of the brewing industry which is generated in high amounts. In recent years, sustainable food production has become more and more important. BSG mainly used as cattle feed has gained high interest due to not only its valuable ingredients such as fiber and proteins but also secondary metabolites remaining in BSG after the brewing process and known for many biological effects. In the present study, various methods were applied, such as acetone extraction (A), alkaline hydrolysis followed by ethyl acetate extraction (HE), and acetone extraction of alkaline hydrolysis residue (HA). Compounds present in the respective bioactive extracts were characterized by mass spectrometry to identify the active compounds. Various hydroxycinnamic acid derivatives as well as oxylipins and some dicarboxylic acids, such as azelaic acid, were present in HE and HA extracts. In contrast, some catechins and phenolamides, such as numerous hordatines, as well as oxylipins and phospholipids were detected in A extracts. Quantification using HPLC‐DAD revealed hordatine contents up to 172.2 ± 2.1 μg p‐coumaric acid equivalents/mg extract. Hydroxycinnamic acid derivatives content accounted for up to 48% of the total extract (HE extracts) but only around 3% of the total HA extracts. In summary, all extracts contained secondary plant metabolites belonging to different classes, ranging from hydroxycinnamic acids to phenolamides, such as not only hordatines but also oxylipins, which were identified for the first time in BSG.

Despite its high nutritional value due to its high protein (19-30% w/w) and fiber (30-50% w/w) content, BSG is often used as cattle feed. In recent years, other applications have been investigated, e.g., as an ingredient for human food to enhance its nutritional value (Merten et al., 2022;Sahin et al., 2021), used in biogas production (Szaja et al., 2020), or as an adsorbent to remove metals from waste water (Lu & Gibb, 2008). It is well-known that beer is rich in polyphenols such as hydroxycinnamic acids, flavonoids, and also phenolamides, as reviewed by Wannenmacher et al. (2018), but many of these bioactive compounds have been reported to be present in BSG as well (Bonifácio-Lopes et al., 2020;Verni et al., 2020).
BSG is a lignocellulosic material with cellulose, hemicellulose, and lignin being the most abundant components. Hemicelluloses may be present at up to 40% of the total dry weight (dw) of BSG (Mandalari et al., 2005), and due to cross-linkers built from ferulic acid dehydrodimers (DiFA), BSG is an interesting source of bound polyphenols (Stefanello et al., 2018). Lignin, a polyphenolic macromolecule, is found at up to 28% BSG dw and consists of not only monomers sinapyl/coniferyl and p-coumaroyl alcohol but also high amounts of polyphenols, such as ferulic acid (FA) or vanillic acid, as part of the lignin structure (Aura et al., 2013;Mussatto et al., 2010).
Besides the lignocellulosic compounds in BSG, high amounts of proteins (14%-31%) (Santos et al., 2003;Xiros et al., 2008) have been reported with essential amino acids, such as methionine, tryptophane, or lysine, representing around 30% of the total protein content (Lynch et al., 2016). Moreover, up to 13% of lipids have been found in BSG (Xiros et al., 2008). Niemi et al. reported that the lipids comprised mostly triglycerides (55%), and also free fatty acids (30%), phospholipids (9.1%), and diglycerides (5.7%) (Niemi et al., 2012). Del Río et al. reported similar distributions within lipid extracts of BSG, with palmitic, oleic, and linoleic acids being the predominant free fatty acids (Del Río et al., 2013). Previously, Anness and Reud reported that most lipids from malt remained in BSG but around 30% of lipids were oxidized during mashing (Anness & Reud, 1985), which might lead to modified lipid classes in BSG. Arts et al. reported the presence of hydroxylated fatty acids resulting from enzymatic oxidation during the mashing process (Arts et al., 2007). In line with these findings, trihydroxylated fatty acids have been detected in beer (Esterbauer & Schauenstein, 1977). They belonged to the class of oxylipins, comprising oxidation products of polyunsaturated fatty acids with oxygen as well as their primary oxidation products (Arts et al., 2007).
Regarding the polyphenols in BSG, a distinction between bound and free compounds can be made and has been investigated in different studies (Birsan et al., 2019;Verni et al., 2020). The first step in the extraction process is usually SLE with organic solvents (free polyphenols), followed by alkaline hydrolysis (bound polyphenols), which disrupts ester bonds in the cell walls, leading to the liberation of hydroxycinnamic acids. Free polyphenols mainly consisted of flavonoids like catechin, whereby bound polyphenols such as hydroxycinnamic acid derivatives, particularly FA and p-coumaric acid (pCA) were present in 50-to 100-fold higher amounts (Birsan et al., 2019;Verni et al., 2020). Ferulic acid oligomers are also part of the bound polyphenols and were reported to be contained in BSG (Birsan et al., 2019;Moreira et al., 2013;Verni et al., 2020).
We recently investigated the influence on glucose-metabolizing enzymes with BSG extracts prepared by solid-liquid extraction (SLE) with 60% acetone or alkaline hydrolysis followed by ethyl acetate as well as acetone extraction of alkaline hydrolysis residue (Becker et al., 2021). All of the extraction methods are well-known procedures and were already used for the extraction of polyphenols (Hernanz et al., 2001;Jay et al., 2008;Meneses et al., 2013).
Our approach was to generate different extract groups with varying compounds including solid-liquid extracts containing flavonoids and other polyphenols and extracts from alkaline hydrolysis mainly containing hydroxycinnamic acids. Extensive purification by ethyl acetate extraction and SPE was performed to ensure the elimination of interfering carbohydrates (Aarabi et al., 2016;Michalkiewicz et al., 2008). The respective extracts were characterized by the total phenolic content (TPC) and total flavonoid content (TFC), but individual compounds were not identified yet. In the present study, the same BSG extracts were analyzed by tandem mass spectrometry, and specific compounds were quantified by HPLC-DAD in a semiquantitative approach in order to identify active compounds.

| Chemicals
Chemicals were of analytical grade and obtained from Sigma-Aldrich (Taufkirchen, Germany) unless otherwise stated. HPLC LC-MS grade and supergradient HPLC grade acetonitrile and HPLC LC-MS grade methanol were obtained from VWR Chemicals (Darmstadt, Germany).

| BSG samples and extracts
Three different batches of BSG were used: one from the conventional Orval Brewery in Belgium (Florenville, Belgium, BSG 3) and two supplied by the brewing group of the Chair of Bioprocess Engineering at the Technische Universität Kaiserslautern (Kaiserslautern, Germany; BSG 1-2). The malts used for these batches are listed in Table 1.
Extracts were prepared using three different extraction processes: combinations of (a) ultrasound-assisted alkaline hydrolysis with 4 M sodium hydroxide (NaOH) followed by ethyl acetate extraction (HE extracts), (b) SLE with 60% acetone (acetone/water: 60/40; v/v; A extracts), and (c) acetone extraction of alkaline hydrolysis residue (HA extracts). All extractions were followed by a purification step using solid-phase extraction with C18ec material and a lyophilization process as previously described in more detail (Becker et al., 2021). As a result, three different extract groups, labeled A1-A7 (solid-liquid extraction with 60% acetone), HE1-HE6 (alkaline hydrolysis followed by ethyl acetate extraction), and HA1-HA3 (acetone extraction of hydrolysis residue), were obtained (Table 1).

| Structural elucidation of ferulic acid dehydrodimers and identification of oxylipins
Supplement Material A).

| Quantification of total hordatine content by HPLC-DAD
Total hordatine content was determined as p-coumaric acid equivalents (pCA-Eq) based as previously described (Becker et al., 2022) using a correction factor (CF) due to the different UV absorption of hordatines and pCA which was adapted from published data (Pihlava et al., 2016).
Two independent extract solutions were prepared and each was injected three times (except A2, six injections of the same extract solution). LOD was determined as 0.3 μg/ml and LOQ as 1.0 μg/ml. Significant differences between and within extraction groups were evaluated using a one-sample t-test (one sided) (Couallier, 2013;Fahrmeir et al., 2016). Differences were considered significant at p < .05; p < .01; and p < .001 levels.
Matrix effects within quantification methods were evaluated using a two-sided t-test and checked for (a) differences between internal standard response in extract matrix solution compared to solvent solution (for hordatine quantification) or (b) significant differences between experimental determined b-value, an indicator for multiplicative systemic errors achieved by indirect regression (see Section 2.9 for hydroxycinnamic acids quantification) and theoretical b-value of 0. Precision was evaluated by an inter-and intra-day repetition method.

| Method validation
Intra-day repeatability was assessed by five replicate analyses of one standard concentration; and inter-day reproducibility was obtained by analysis of one concentration level of the analyte for 5 days in a row.
Mean and SD of each experiment were determined, and the coefficient of variance was calculated. with regard to their deviation (%). Matrix effects within hydroxycinnamic acids' quantification was evaluated by indirect regression method using spiking (Wellmitz & Gluschke, 2004). Three different concentrations of FA, pCA, and CA (for HE extract) were each added to one HA extract (1000 μg/ml) and one HE extract (250 μg/ml) as follows: for HA extract pCA: 5, 7.5, and 20 μg/ml; FA: 10, 17.5, and 40 μg/ml; for HE extract: pCA: 20, 25, and 35 μg/ml; FA: 50, 75, and 100 μg/ml; and CA: 2.5, 5, and 7.5 μg/ml. Each extract was analyzed spiked as well as unspiked and each reference spiking solution was also analyzed in pure solvent. Criteria for matrix effects is the bvalue which was calculated as follows: where x 3 = AUC of spiked extract, x 1 = AUC of unspiked extract, and x + = AUC of reference spiking solution. Mean b-values were calculated (n = 2; duplicates; each replicate includes three spiking levels of each reference substance) and analyzed for significant differences to theoretical b-value of 0.

| RE SULTS AND D ISCUSS I ON
BSG extracts were prepared by three different extraction processes as previously described (see Section 2.2, Becker et al., 2021), resulting in three extract groups (HA, HE, and A;  (Hernanz et al., 2001;Jay et al., 2008). Ester-linked hydroxycinnamic acids and also sugars can be released by NaOH and are therefore found in the aqueous extracts from alkaline hydrolysis. Hence, ethyl acetate extraction was used to eliminate liberated carbohydrates from the raw extracts leading to HE1-HE6 extracts.
Furthermore, a second extraction step with 60% acetone was performed using the solid residue of the hydrolysis to ensure that all hydroxycinnamic acids were extracted and HA1-HA3 extracts were generated. Additionally, all three extract groups were purified and secondary metabolites were concentrated using SPE with C18ec material (Michalkiewicz et al., 2008). HPLC-MS/MS analysis showed that the extracts strongly differed regarding their main compounds: A1-A7 extracts contained some polyphenols but were mainly rich in hordatines. HE1-HE6 extracts were rich in not only hydroxycinnamic acids, which were released from cell wall arabinoxylans by alkaline treatment, as described in the literature (Stefanello et al., 2018), but also trihydroxy oxylipins were detected. HA1-HA3 extracts were generated to ensure almost complete extraction of hydroxycinnamic acids. In general, similar compounds as in HE extracts were identified in HA extracts, namely hydroxycinnamic acids and trihydroxy oxylipins, although the amounts were much lower.

| Compound identification in extracts by HPLC-MS/MS
Extracts were analyzed by HPLC-UV-ESI-MS/MS using two different systems but the same HPLC method; specific MS parameters are shown in the results ( Figure 1). For each extract group (A, HE and HA), a HPLC-DAD chromatogram recorded at 300 nm is presented in Table 2-4. In general, chromatograms within one extract group were very similar to each other and independent of the BSG batch recorded.
Tables 2-4 present the mass spectral data of compounds identified in the respective extracts by HPLC-ESI-MS/MS. Some signals were tentatively identified as hordatines (signal 2-4 in Figure 1a) or DiFAs and oxylipins (signals 6-11 in Figure 1b and signal 6-17 in Figure 1c) but will not be presented within this section since the methods had to be optimized in terms of chromatographic separation. All results regarding those compounds will be discussed in Sections 3.2 and 3.3. Identification was based on comparison of MS (parent ion) and MS/MS (fragment ion) data and the retention time (t R ) with those of commercially available standards or published data if no standard was available. Besides polyphenolic compounds, such as hydroxycinnamic acids, hordatines and catechin, tryptophan, phospholipids, and dicarboxylic acids were identified.

TA B L E 2 Identification of compounds in BSG
Some A extracts (A1, A2, A4, and A5) showed the presence of tryptophan (not visible in chromatogram recorded at 300 nm but at 280 nm, data not shown), an essential amino acid, which was verified using a reference substance. It has already been reported to be present in BSG (Essien & Udotong, 2010), but was not found in extracts from BSG 3 (A3, A6, and A7). This could be due to different storage times since microorganisms can degrade amino acids. Besides hordatines (see Section 3.2), coumaroyl hydroxyagmatine, a monomeric precursor of hordatines, was identified by mean of its fragmentation pattern (von Röpenack et al., 1998). It was not observed in the A1, A3, and A6 extracts, which was not surprising for extract A1 as it was produced from BSG composed of 50% wheat/50% barley grain malt, and hordatines are known to be largely barley-specific compounds. Extracts from BSG3 did not contain the hordatine precursor except in previously defatted samples (A7). This might be for two reasons. It is possible that the hordatine precursor was concentrated by the defatting process, or the defatting procedure may have caused modifications of the contained components. Catechin derivatives, such as (+)-catechin and procyanidin B, were detected in some extracts, mainly from BSG 2 (A2, A4, and A5), and were verified by using a reference substance.
Catechins have previously been detected in BSG. indicating that they might contribute to the A extracts inhibitory effects detected in our previous study (Becker et al., 2021). Besides the phenolic compounds, three lysophosphatidylethanolamines (LPEs) (m/z 478 and m/z 454) were detected by their characteristic fragments of m/z 337 and 313 resulting from a loss of m/z of 141 (Fang & Barcelona, 1998). Phospholipids accounted for around 9.1% in lipid fractions of BSG, which corresponded to around 11% of the total (Niemi et al., 2012). No single phospholipids have been reported to be present in spent grains so far. Regarding the different BSG batches, no clear correlation was observed, and phospholipids were detected in all batches under study. However, a recently published study demonstrated that phospholipids in milk thistle oil inhibit α-glucosidase (Harrabi et al., 2021). Therefore, they might also be important regarding the enzyme-inhibition by the A extracts found in our last study (Becker et al., 2021). Further compounds were detected but could not be identified; most of them eluted in a similar region as the LPEs, indicating relatively high lipophilicity. Additionally, most of them (m/z 348, 367, and 365) were observed in up to three isomeric forms. Compounds 6 and 7 only differed by 2 Da, which could indicate that compound 6 was the saturated form of compound 7. Also, both showed similar fragments corresponding to two times loss of 18 Da, which was probably due to the loss of water. A summary of all identified compounds in A extracts is shown in Table 2.  i.e., suberic and azelaic acid, were identified by their fragmentation patterns and showed good agreement with the reference substance fragmentation and t R . Cereals and whole grains are known to be a good source of carboxylic acids (Lohaus et al., 1983) and have been detected in BSG. However, as noted for the A extracts, some compounds remained unidentified.
HE extracts (Table 4) contained similar compounds to those in HA extracts, but the signals for hydroxycinnamic acids were more intense and isomeric forms of pCA and FA presenting the same fragmentation pattern were detected. Additionally, two isomers of CA already detected in BSG (Moreira et al., 2013;Verni et al., 2020) were observed in all HE extracts. The isomers for FA, pCA, and CA were tentatively assigned as cis-and trans-isomers which can be expected from literature (Callipo et al., 2010) with the trans-isomer being the main naturally occurring one. Moreover, a dimer of coumaric acid was identified by means of its fragmentation pattern (Spínola et al., 2018). As in HA extracts, DiFAs, dehydrotriferulic acids (TriFAs), and oxylipins were indicated in negative full scans (m/z: 100-1000 Da; data not shown), and the results are discussed in Sections 3.3 and 3.4.

| Characterization of A extracts: Identification of phenolamides and quantification of hordatines
The results of the initial LC-MS (MS) experiments (see Section 3.1.) showed that A extracts differed significantly from HA and HE extracts regarding the main compounds, which were tentatively identified as hordatines according to our previously reported study (Becker et al., 2022). In the frame of this work, 60% acetone was used, as it was described in literature as among the most appropriate solvents for the extraction of polyphenols (Meneses et al., 2013). However, a similar solvent (75% acetone) was also used by Kohyama and Ono to isolate hordatines from barley (Kohyama & Ono, 2013). Enhanced product ion scans were performed to specify each hordatine structure and led to further phenolamides being detected. Also, the total hordatine content as pCA-Eq was  (von Röpenack et al., 1998) in beer where they contribute to astringency (Kageyama et al., 2011;Pihlava et al., 2016) and also in BSG (Becker et al., 2022), which demonstrates their high heat and enzyme stability. Furthermore, some hydroxycinnamoyl putrescines and spermidines were already detected in barley beer (Pihlava, 2014). As also seen within our last study (Becker et al., 2022), many different isomers (cis and trans, regio-isomer, or epimers of the hexose unit) of each hordatine can occur and may be found in the extracts whereby a summary of all phenolamides is shown in Table 5.
Regarding the aglycons, the isomers found within this study varied slightly from our previous results. Especially, the amount of double hydroxylated hordatine isomers A2 to C2 was lower within the investigated extracts compared to the isolated fractions in our previously reported study (Becker et al., 2022). This might be due to the much more specified isolation process since the crude isolate which would be comparable to the extracts investigated here (A1-A7) was fractionated and thus hordatines were more concentrated in each fraction. Therefore, it might be possible that some isomers are only contained in very small amounts and have not been detected within our extracts. When comparing the different isomers for glycosides, similar differences can be found whereby for some hexosides (e.g., hordatine C and C1 hexoside) more isomers were detected within this study in each extract. Nevertheless, no hexosides of hordatines A2 to C2 were observed in none of the extracts, too. Furthermore, the two processes to extract the hordatines varied and it cannot be excluded that this might also lead to modifications of the original hordatine structure.
Comparing the amount of hordatines present in the extracts and taking into account the much higher measured concentrations for extracts A1-A3, it is likely that extracts A4-A7 contained considerably more different hordatine structures. This was an expected result for extract A1 since hordatines are known to be barley-specific compounds and A1 was produced from malt originating from 50% wheat.
Regarding the biosynthetic precursors of hordatines, similar results were already observed in our previously reported study.
Both precursors were also found in our hordatine-rich fractions prepared from BSG (Becker et al., 2022). Also, differences between the BSG batches were observed in the study presented here. Ncoumaroylagmatine was only detected in BSG 2 extracts (A2, A4, and A5). Also, N-feruloylagmatine isomers were more dominant in extracts from BSG 2 since the second isomer as well as the glycosylated structure were only both seen in extracts A2, A4, and A5. In addition to agmatine-containing phenolamides, two hydroxycinnamoyl spermidines were identified. Both were not observed or only in small amounts in extracts from BSG 3 (A3, A6, and A7). Spermidine conjugates have not been identified in BSG so far but were reported to be present in barley (Pihlava, 2014).

| Total hordatine content of extracts
Total hordatine content in A extracts was determined as pCA-Eq as previously described (Becker et al., 2022) and ranged from 14.2 ± 0.5 to 172.2 ± 2.1 μg pCA-Eq/mg extract. The amounts can be considered an estimation of total hordatines in BSG extract due to slight matrix effects and calculation as equivalents. However, significant differences between the extracts were observed (Figure 3).
Extract A1, originating from malt with around 50% wheat content, had a significantly lower total hordatine content than all the other F I G U R E 2 Structures of hordatines found in BSG extracts; all aglycons have been detected, monoglycosidic forms for hordatines A-C and A1-C1, and diglycosidic forms for hordatines A-C; glycosylation (1-3 hexose units) occurs at the hydroxyl group position indicated by the arrow; and sugar moiety: hexose (hex); numbering adapted from published data (Kageyama et al., 2013;Pihlava, 2014) TA B L E 5 Phenolamides (hordatines, hydroxycinnamic acid agmatines, and hydroxycinnamoyl spermidines) found in A1-A7 extracts tentatively identified by double-charged parent ion and characteristic fragments; x: contained and fragmentation sufficient, ( x (x) x (x) x (x) a Isomer at 17.7 min overlapped with hordatine B (m/z 291); therefore, there was no clear distinction between the fragments of hordatines B and A2.
Previous defatting of the raw material (A5 and A7) resulted in higher hordatine content compared to nondefatted samples (A4 and A6), which was significant for A6 and A7 (p < .001). To compare our findings with literature data, we calculated the yield-related hordatine content showing the hordatine content in BSG dw (Appendix S1: Supplement Material B), which should be treated with caution due to the last SPE purification step (Becker et al., 2021). Those BSGrelated values showed similar dependencies between the different BSG batches and extraction processes. Defatted BSG extracts (A5 and A7) had higher total hordatine contents than the corresponding nondefatted extracts (A4 and A6). Extract A1, from 50% wheat share, had the lowest content. The differences between extractions 1 and 2 were not so pronounced.
Quantitative data of hordatines content are relatively limited.
The total hordatine content determined as pCA-Eq was only reported for beer by the group of Pihlava et al. (Pihlava et al., 2016) and in our previous study for hordatine-rich fractions prepared from BSG (Becker et al., 2022). Total hordatines content in beer showed an average value of around 5.6 ± 3.1 mg pCA-Eq/L with maximum value of around 18.7 mg pCA-Eq/L (Pihlava et al., 2016). These values revealed that the relatively high amounts of hordatines present in malt were transferred to wort, and thereafter to the final product beer.
However, our previous results demonstrated that hordatines also remained in BSG in high amounts of around 40 μg pCA-Eq /g BS dw.
This was confirmed within our study presented here, where total hordatine contents of 242 to 1550 μg pCA-Eq/g BSG dw (Appendix S1: Supplement Material B) were determined. This was even higher than the amounts found in our previously reported study and also than the content observed in beer. However, the main target of the analysis was the identification of bioactive compounds within the BSG extracts. Extracts A2-A7 showed strong inhibitory activities towards α-glucosidase (Becker et al., 2021) and hordatines accounted for a relevant part of the A extracts with up to 17% (extract A7). Our previous reported study demonstrated that the hordatines contribute to the enzyme-inhibition but the specific active hordatine structure could not be identified. Nevertheless, our findings also indicate a correlation between hordatine content and bioactivity. However, a reference substance is needed for quantification in order to obtain unambiguous conclusions about the hordatine amounts. Also, a validation of the extraction process including repetition and determination of the recovery rates within the SPEs could be performed with a reference substance. Without this evaluation, the calculations of the amount contained in BSG remain merely an estimation.

| Characterization of HE and HA extracts: Quantification of hydroxycinnamic acids and structural elucidation of ferulic acid oligomers (FAO)
The results of the initial HPLC-MS(/MS) experiments (see Section 3.1) showed that HA and HE extracts differed significantly from A extracts regarding some of the main compounds tentatively identified as FAOs, which were not observed in A extracts. Since they are mainly part of the cell wall structure they have to be released chemically or enzymatically whereby in the frame of this work alkaline hydrolysis was used. Compared to studies focusing on isolation of FAO (Hernanz et al., 2001;Pedersen et al., 2015), the NaOH concentration was relatively high (4 M), probably leading to higher amounts of monomeric FA than DiFAs. MS/MS experiments were performed to clarify the different structures of FAOs, but only DiFAs could be identified. Furthermore, hydroxycinnamic acids were quantified, whereby the amount of DiFAs was determined as FA-Eq due to the lack of available reference substances.

| Structural elucidation of ferulic acid dehydrodimers (DiFAs) by mass spectrometry (HPLC-ESI neg -MS/MS)
DiFAs in HA and HE extracts were characterized using tandem mass spectrometry in the negative ionization mode. Characteristic fragment ions and the typical fragment distribution were used to determine the structures of each isomer by means of published data (Callipo et al., 2010). Altogether, a total of 10 DiFAs were identified whereby 2 were only present in low amounts or coeluted with more intense DiFA signals (Appendix S1: Fig. in Supplement Material C).
TriFAs, as indicated by the full scans from the first MS experiments (see Section 3.1.), could not be identified by tandem mass spectrometry due to low intensities.
A summary of all DiFAs, their tentative structures according to literature data, and characteristic fragment ions are summarized in F I G U R E 3 Total hordatine content in A extracts (acetone extraction) expressed as μg p-coumaric acid equivalents (pCA-Eq)/mg extract. Values are expressed as means ± SD of two independent extract solutions each injected three times. Significant differences between different BSG samples and extraction processes were analyzed: *p < .05; **p < .01; ***p < .001 showed specific fragmentation patterns, as already reported in detail before (Callipo et al., 2010) and described here in brief. The 8-8′-aryltetralin isomer (10.8 min; Appendix S1: Fig. Supplement C (1)) showed specific signals at m/z 217, 173, 158, 123, and 108, respectively. Linear 8-8′ isomers were found in two forms (eluting at 11.4 and 14.7 min, Appendix S1: Fig. Supplement C (2), (4)). The earlier eluting isomer was more abundant and therefore expected to be the trans-trans-isomer. Diagnostic signals of m/z 173, 159, and 145 were ascribed to cyclic fragments that can only be generated due to the conjugated structure between two aromatic rings.
The 8-5´ and 5-5′-isomers were more difficult to distinguish. One difference was the loss of water of the decarboxylated fragment ion m/z 341 for 8-5´-DiFA (Appendix S1: Fig. Supplement C (3)), leading to a phenolic lactone of m/z 323, which was found for the signal at 13.9 min and slightly also for the signal at 18.4 min (Appendix S1: Fig. Supplement C (5)), tentatively assigning them the trans-trans-and cis-trans-8-5´-DiFA according to literature data (Callipo et al., 2010). The 5-5´-DiFA (Appendix S1: Fig. supplement C, (6)) was characterized by a very stable parent ion as well as loss of 60 and 76 Da, giving rise to fragments of m/z 281 and 265, which was observed for the signal at 23.6 min. Additionally, it was confirmed using a reference substance (friendly provided by Prof.
Dr. Mirko Bunzel, KIT). Since the anti configuration is the most stable and the trans configuration the naturally occurring one, the isomer found was considered to be the anti-trans-isomer. All remaining signals (26.5, 28.5, 30.2, and 32.5 min; Appendix S1:

| Quantification of hydroxycinnamic acids
Hydroxycinnamic acid derivatives were quantified in HA and HE extracts by HPLC-DAD. DiFAs were quantified as FA-Eq and no CF could be calculated. Thus, the results, already expressed as equivalents, were considered to be an estimation. However, differences between the BSG batches and extraction groups could be evaluated.
All extracts contained relatively high amounts of hydroxycinnamic acid derivatives. HA extracts showed significantly lower contents (p < .001) than HE extracts for all four groups-CA, pCA, FA, and DiFAs-with the exception of CA in HE5, whose content was below the LOQ ( Figure 5). Altogether, the hydroxycinnamic acid content accounted to around 3% of the total HA extract and up to 48% of the total HE extract. Regarding our main goal of the study, the identification of active compounds especially the large percentage in the HE extracts is interesting. The HE extracts investigated showed inhibitory activity toward glycogen phosphorylase α (GPα) and αglucosidase (Becker et al., 2021). Hydroxycinnamic acids such as FA and DiFAs are reported to be potent inhibitors of different enzymes of the glucose metabolism (Adisakwattana, 2017;Narasimhan et al., 2015;Ye et al., 2022). Thus, their contribution to the enzymeinhibition detected in our study (Becker et al., 2021) is likely. The huge difference between the hydroxycinnamic acid contents in HA compared to HE extracts was expected since HA extracts were prepared from the residue of alkaline hydrolysis with the latter serving to produce the HE extracts; i.e., huge amounts of hydroxycinnamic acids were already extracted during the alkaline hydrolysis (HE extracts) whereby only the remaining hydroxycinnamic acids in the residue were extracted with acetone (HA extracts).
For all extracts, the contents could be ranked as follows: FA > pCA > DiFA > CA; data presented in Table 7 tent about twice as high as that of pCA in BSG dw. In our study, the FA content was around three-to fourfold higher than the pCA amount in the HE and HA extracts (around 10-fold higher for HE1).
However, the CA content in HE extracts was much higher than that reported by Birsan et al. All quantitative results reported in the literature used for comparison were provided in mg/g BSG. Therefore, to compare the amounts observed in our study, we calculated the hydroxycinnamic acid derivative amount in BSG dw in relation to extract yield that had previously been published (Becker et al., 2021).
It should be mentioned that these calculations only provided an estimation since many purification steps (SPE and LLE) were performed without determination of recovery rates. As the extraction was only performed once, no validation was available. Yield-related amounts were as follows: for HE extracts: 1251.5 ± 6.9-3283.  (Verni et al., 2020). They distinguished between free and bound polyphenols by using LLE with organic solvents as the first extraction step and alkaline hydrolysis as the second step.
Hydroxycinnamic acids were not detected in extracts from LLE, except for some dihydrohydroxycinnamic acids. High amounts of hydroxycinnamic acids were observed in extracts after alkaline hydrolysis, i.e., 53 mg/kg FA, 312 mg/kg isoferulic acid, 41 mg/kg pCA, 35 mg/kg CA, 568 mg/kg DiFAs, 450 mg/kg tetrameric FA, and 87 mg/kg TriFA. The amounts of monomeric hydroxycinnamic acids TA B L E 6 DiFAs found in HA1-HA3 and HE1-HE6 extracts tentatively identified; x: contained and fragmentation sufficient, (x): insufficient fragmentation due to low concentration, partially only unspecific fragment ions or signals in TIC detected, −: not detected; characteristic fragments are shown in bold; fragment ions listed in order of decreasing intensity; related HPLC-MS/MS-chromatogram can be found in the supplements (Appendix S1: Supplement Material C) Very high stability of parent ion.

TA B L E 7
Hydroxycinnamic acid contents in HA and HE extracts expressed as μg/mg extract and μg FA-Eq/ mg extract ± SD that reaction time, temperature, and concentration of NaOH influenced the hydroxycinnamic acids content and found the highest values for the maximum conditions of 2% NaOH (% w/v), 90 min, and 120°C (Mussatto et al., 2007).

| Identification of oxylipins in BSG extracts
Besides polyphenolic compounds, such as hydroxycinnamic acids and hordatines, hydroxylated fatty acids were detected in all extracts (HA, HE, and A) under study. Different isomers of trihydroxyoctadecenoic (TriHOME) and trihydroxyoctadecanoic acid (TriHODA) were identified by HPLC-MS/MS by means of published MS data (Bhunia et al., 2018;Martin-Arjol et al., 2010) and partially available reference substances. All of them were linoleic acid-derived oxylipins. Eight TriHOMEs and two TriHODAs were found in total, with the TriHOME isomer at 28.3 min and the TriHODA isomer at 28.9 min (Appendix S1: Supplement Material D) being the predominant structures in all extracts. It was reported previously that only 4%-5% of malt lipids are released into the wort and the majority of lipids remain in the BSG, where probably around 30% of the lipids are oxidized during mashing, which might have generated the oxylipins found in our studies (Anness & Reud, 1985).
An overview of all oxylipins detected in each extract is provided in Table 8. In general, most isomers of TriHOME were found in HE extracts, followed by HA extracts, with the lowest content in A extracts, where only the isomer at 27.5 min was present. Furthermore, TriHODA isomers were not identified in A extracts (except in A1 and A6) but were identified in HA and HE extracts. These findings indicate that most of the oxylipins were (a) released by alkaline hydrolysis, or (b) generated and modified during the extraction process with alkaline solvent. It has already been reported many years ago that TriHOMEs are present in beer (Esterbauer & Schauenstein, 1977) and are produced during the malting and mashing process by lipoxygenase (LOX) reaction, resulting in two isomers: (9, 10, 13)-and (9, 12, 13)-TriHOME, which further represent eight diastereomers and eight enantiomers.
However, the enzyme-catalyzed reaction by barley LOX is highly regio-and stereoselective, leading mainly to the formation of (9 S, 12 S, and 13 S)-TriHOME, which can be found in high amounts in wort and beer. This isomer elutes at 27.5 min, corresponding to the retention time of the isomer found mainly in A extracts. Even if the signal intensities in A extracts were relatively low, the fragment distribution was very similar. This indicates that residues of the characteristic (9S, 12S, 13S)-TriHOME in beer were extracted by 60% acetone, which was also used for the production of HA extracts. However, (9S, 12S, 13S)-TriHOME was rarely found in HE extracts, whose main isomer corresponded to the retention time of the standard (9S, 10S, 13S)-TriHOME at 28.3 min. In general, the TriHOME isomers differ in their fragment pattern, but it was difficult to compare in the frame of our study because most of the isomers coeluted. The large variety of isomers in HE and HA extracts may be due to the alkaline treatment. During the enzyme-catalyzed reaction to form TriHOMEs, intermediate allylic epoxy alcohols derived from 9-and 13-hydroperoxides are formed (Hamberg, 1991), which can then be hydrolyzed by alkali.
This hydrolysis, in contrast to enzymatic cleavage, is not regioor stereoselective, and therefore different isomers can be produced. To distinguish between the TriHOME isomers and to test the above hypothesis, optimization of the method is necessary.
Methods to separate 16 different isomers have already been reported (Fuchs et al., 2018 reported an inhibition of α-glucosidase by extracts containing trihydroxy fatty acids (Nadeem et al., 2020). Therefore, our assumption should be confirmed by quantification and isolation of oxylipins followed by the investigation of pure substances in the enzymeinhibition assays.

| CON CLUS ION
BSG extracts prepared by SLE with 60% acetone (A1-A7), SLE with 60% acetone and alkaline hydrolysis of the respective residue (HA1-HA3), or alkaline hydrolysis followed by LLE with ethyl acetate (HE1-HE6) from three BSG batches (Becker et al., 2021) were analyzed by HPLC-ESI-MS/MS. The main compounds were identified and afterward quantified by HPLC-DAD in order to draw conclusions on the active compounds in the extracts. In general, extracts A1-A7 differed strongly from HA1-HA3 and HE1-HE6 extracts. Catechin, phenolamides, some phospholipids, and tryptophan were detected in extracts A1-A7, whereas hydroxycinnamic acids, such as FA and pCA as well as DiFAs, were identified in HA and HE extracts. Furthermore, dicarboxylic acids such as azelaic and suberic acids were observed. Moreover, all extracts contained isomers of TriHOME, which have only been reported before in beer (Esterbauer & Schauenstein, 1977). The oxidized forms of linoleic acid were probably generated during mashing by enzymatic reactions, which are likely to be regio-and stereoselective, leading to at least one isomer (Garbe et al., 2005). A extracts only contained one TriHOME, whereas HA and HE extracts contained various isomers, TA B L E 8 Oxylipins found in BSG extracts tentatively identified; x: contained and fragmentation sufficient, (x): insufficient fragmentation due to low concentration, partially only unspecific fragment ions or signals in TIC detected, −: not detected  , 211, 329, 183, 193, 171, 129, 99, 293, 311 9S, 10S, 13S-TriHOME 28.3 329 171, 329,139,157,127,99,201,293,311,211 which might be due to the alkaline treatment. For the first time, oxylipins were identified in BSG. Also, the identification of phospholipids has not been reported so far. Both lipidic compounds are not well studied regarding the inhibiting potential toward glucose metabolism enzymes, but there are some recently published studies (Harrabi et al., 2021;Nadeem et al., 2020) indicating that the compounds might contribute to the bioactivity observed in our study (Becker et al., 2021).
Extraction of hydroxycinnamic acids from BSG by alkaline treatment is a well-established method (Hernanz et al., 2001;Verni et al., 2020). Our results are mostly comparable to published data.
However, the content of DiFAs was lower and the amounts of monomeric structures higher in our extracts. This indicates that the usage of more concentrated NaOH results in more extensive hydrolysis of the dimeric structures. Differences between the compositions after the use of different extraction methods were observed. HA extracts (prepared by acetone extraction of the residue after alkaline hydrolysis) had significantly lower amounts of hydroxycinnamic acids than HE extracts (prepared from alkaline hydrolysis followed by ethyl acetate extraction). This was expected since acetone was only used to extract the hydrolysis residue and most of the liberated polyphenols were already extracted by the NaOH solution. Hydroxycinnamic acids are already well-known inhibitors of α-glucosidase and GPα (Adisakwattana, 2017;Narasimhan et al., 2015) and recently also DiFAs were reported to be potent inhibitors of α-glucosidase (Ye et al., 2022). Especially for HE extracts, these compounds may be relevant for the inhibition since they account for up to 48% of the total extract. HA extracts did not inhibit the α-glucosidase, but GPα (Becker et al., 2021). However, HA extracts did not contain DiFAs and only 3% of the total extract accounted for hydroxycinnamic acids. Thus, our results gave strong evidence of these polyphenols to be relevant for the inhibiting potential of the BSG extracts.
Besides hydrolytic extraction methods, SLE has often been used for extraction of polyphenols from BSG with various extraction solvents (Birsan et al., 2019;Bonifácio-Lopes et al., 2020;Martín-García et al., 2019) and many phenolic compounds, such as catechin, syringic, and sinapic acid, vanillin, and proanthocyanidins, have been detected. Additionally, some phenolamides, such as hordatines and spermidine conjugates, were observed for the first time in BSG.
Altogether, 27 different phenolamide structures were identified, including 2 hydroxycinnamoyl spermidines, 4 hydroxycinnamoyl agmatines, and 21 hordatines, such as hordatines A, B, and C (Pihlava et al., 2016), hordatines A1, B1, and C1, as well as their glucosides and the aglycones hordatines A2, B2, and C2. Most hordatine structures were found in up to six isomeric forms. Furthermore, extraction process 3 (A4-A7) yielded more different hordatine isomers compared to A1-A3 (from process 1, Table 1). This was expected for A1 since hordatines are barley specific, and BSG 1, used as the A1 raw material, originated from 50% wheat. Relatively high amounts of up to 172 μg pCA-Eq/mg extract corresponding to 1550 μg pCA--Eq/g BSG dw were found, indicating BSG is a good source of phenolamides. Furthermore, hordatines might contribute to the inhibitory potential of A extracts on α-glucosidase which was also shown in our recently published study (Becker et al., 2022). However, the specific hordatine structure responsible for enzyme-inhibition is still not clear. Again, our results demonstrate the need for a reference substance for unambiguous quantitative data and investigation of pure substances in the enzyme-inhibition assay.
All extracts contained various compounds and notable amounts of different phenolic compounds. Alkaline hydrolysis resulted in hydroxycinnamic acid as the main compound (3% of the total HA extracts; up to 48% of the total HE extracts) and SLE with 60% acetone mainly extracted phenolamides such as hordatines (up to 17% of the total A extracts). Both compound classes were present in relatively high amounts. The quantification of the total hordatine content as well as the content of DiFAs has to be interpreted with caution as the results were only semiquantitative owing to the lack of available reference substances. However, differences in the amounts of hordatines and hydroxycinnamic acids between the different BSG batches and extraction methods could be seen as all extracts were analyzed in the same way. Thus, it was shown that extract A1 produced from BSG with 50% barley had the lowest hordatine content, and extracts from BSG 3 contained less hydroxycinnamic acids than BSG 1 and BSG 2 extracts. Nevertheless, for quantification, the precision can be improved by using reference substances. Methods to synthesize or isolate DiFAs have already been reported (Bunzel et al., 2004(Bunzel et al., , 2008. Nevertheless, our results provide good evidence of the bioactive components in the extracts investigated (Becker et al., 2021) and enable targeted follow-up studies to be carried out in the future. Thus, hordatines seem to be relevant for α-glucosidase inhibition and hydroxycinnamic acids including the DiFAs should be focused on both α-glucosidase-and GPα-inhibition. Finally, the quantification should be improved and the extraction process validated by the use of ref- erence substances to gain precise information on the amounts of bioactive compounds in BSG.

ACK N OWLED G M ENTS
The authors gratefully acknowledge financial support from the

FU N D I N G I N FO R M ATI O N
This research was funded by the EU-INTERREG project BIOVAL supported by the European Funds for Regional Development, project no. 018-4-09-021.

CO N FLI C T O F I NTE R E S T
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this article.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data presented in this study are available on request from the corresponding author.