All chemicals were of analytical grade, purchased from Sigma-Aldrich (Milano, Italy) and used as received. HPLC-grade acetonitrile and methanol were purchased from Carlo Erba (Milano, Italy); HPLC-grade water was freshly prepared with the Milli-Q purification system (Millipore, Vimodrone, Italy).

Four commercial samples of freshly ground base malt, namely Pilser malt 2.8–3.8 EBC (Rhön Malz, Mellrichstadt, Germany), Monaco malt 18–20 European Brewing Convention (EBC) (Rhön Malz, Mellrichstadt, Germany), Vienna malt 6–10 EBC (Crisp Malting, Great Ryburgh Fakenham Norfolk, United Kingdom), Extra Pale Maris Otter malt 3–4 EBC (Crisp Malting, Great Ryburgh Fakenham Norfolk, United Kingdom), were purchased from Birramia.it — Enterprise SRL (Querceta, Lucca, Italy), each sample identified by batch number and expiration date. The samples, stored under vacuum, are available by our laboratory.

Before using the samples, water activity and absolute humidity were measured in duplicate by a Schaller Humimeter Rh2 (Schaller Messtechnik GmbH, St. Ruprecht an der Raab, Austria); the microbiological analysis for the total aerobic count, yeasts and molds, lactic acid bacteria and enterobacteriaceae were performed as previously described [36]. All measurements were performed in duplicate. Results are resumed in Table 1.

Aqueous extracts were obtained by adding 1.0 g of Vienna, Monaco, Pilsner and Extra Pale Maris Otter malt, in turn, to 10 mL of Milli-Q purified water, and keeping the mixture under magnetic stirring for 1 hour, at room temperature. Mixtures were then centrifuged at 3000 rpm for 10 min; supernatants were recovered, filtered at 0.22 $\mu$m, diluted 1:10 (v:v) with mobile phase (MilliQ/formic acid A:acetonitrile/formic acid B, A:B, 95:5, v:v) and injected for the analysis.

Malts were used for the wort production, with the aim to collect the spent grains (BSG or threshes) after mashing. Malts were mashed according to a typical multi-steps mashing procedure. In detail: 15 g of malt were put in a cylindrical glass apparatus equipped with a tap, complemented with 45 mL of mineral water, pre-heated at 52 °C, and immersed in a thermostated bath, for the following mashing program: 45 °C (20 min rest), 50 °C (10 min rest), 65 °C (60 min rest), 78 °C (15 min rest). A Crison GLP 21 (Crison Instruments) pHmeter was used to measure the pH along the mashing. The iodine test was performed to verify the absence of residual starch.

Wort was then collected, while threshes were washed by adding 45 mL of hot water (80 °C) acidified with HCl at pH = 5.5. After shaking, followed by 5 minutes of rest, the washing water was collected and centrifuged by an ALC Centrifugette 4206, at 3000 rpm for 10 min. Supernatant was recovered, filtered at 0.22 $\mu$m, joined to the wort previously collected, appropriately diluted with mobile phase (A:B, 95:5, v:v) and injected for the analysis.

Threshes, coming from mashing, were recovered and dried overnight at 60 °C, until they reached a constant weight. Water activity and absolute humidity were measured in duplicate by a Schaller Humimeter Rh2 (Schaller Messtechnik GmbH, Austria). Results are resumed in Table 2. The resulting product was stored under vacuum.

Aqueous extracts of the BSG samples were obtained by mixing the dried threshes and Milli-Q purified water in the weight/volume ratio 1:10 (w:v), and keeping the mixture under magnetic stirring for 1 hour, at room temperature. After centrifugation at 3000 rpm for 10 min, the supernatant was recovered, filtered at 0.22 $\mu$m, diluted 1:10 (v:v) with mobile phase (A:B, 95:5, v:v) and injected for the analysis.

All the samples, namely wort, aqueous extracts of malts and aqueous extracts of threshes, were analyzed for the targeted phenolic profile and the individual phenolic and tryptophan content, by a Waters apparatus composed of a 1525 $\mu$ HPLC (Waters, Milford, MA, USA), a 996 PDA detector (Waters, Milford, MA, USA) and a Quattro Micro Tandem MS/MS with an ElectroSpray Ionization (ESI) source (Micromass, Manchester, UK); the chromatographic separation was performed with an XBridge C18 (Waters Corporation) (150 $\times$ 2.1 mm i.d.) 5 $\mu$m analytical column, by using Milli-Q with formic acid 5 mM (A) and acetonitrile with formic acid 5 mM (B) as mobile phase, flow rate 0.20 mL/min, and applying a method previously developed [37]. Briefly: 0–1 min, 5% B; 1–20 min, 16.5% B; 20–30 min, 40% B; 30–35 min, 60% B; 35–36 min, 80% B; 36–40 min, 80% B; 40–41 min, 5% B; 41–61 min, 5% B. The PDA detector was set to record one UV–vis spectrum per second in the range of 200–800 nm, resolution 1.2 nm. Mass spectral data were acquired in negative ionization (ES-) by the Selected Ion Recording (SIR) technique, that uses separate acquisition channels, one for each monoisotopic value m/z relative to the deprotonated ion [M-H]${}^{-}$ of the searched compound. ES- parameters were previously optimized for the analysis of 26 phenolic compounds [38]. Briefly: capillary voltage 2.7 kV, cone voltage 27 V, source temperature 120 °C, desolvation temperature 350 °C, cone gas flow 40 L/h, desolvation gas flow 500 L/h, dwell cell value of 0.200 s. Mass spectral data were also recorded in full scan, in both negative (ES-) and positive (ES+) ionization, with the same source parameters. All the samples were injected (20 $\mu$L) in triplicate for quantitation. All validation parameters, including calibration curve, LOD and LOQ, accuracy, precision and matrix effect, used for the quantitation of the identified compounds, had been previously reported [36, 37, 38, 39]. Data acquisition, data handling, and instrument control were performed by MassLynx Software 4.1 v (Data Handling System for Windows, Micromass, UK).

All the samples, namely wort, aqueous extracts of malts and aqueous extracts of threshes, were analyzed for the untargeted metabolic profiling by direct infusion into the ESI source through an external syringe, flow rate 5 $\mu$L/min.

Mass spectral data were acquired in full scan for 2 minutes in both ES- and ES+ ionization, in the mass range 80–800 Da, cone voltage 27 (ES-) and 24 V (ES+), ionization source temperature 100 °C, desolvation gas temperature 150 °C, cone gas flow 30 L/h, and desolvation gas flow 400 L/h. The main signals evidenced in the full scan spectra (Figs. 1,2,3,4) were identified by the fragmentation pattern, obtained by selecting each m/z value as precursor ion and fragmenting with argon, using collision energy (CE) in the range of 10–25 eV (see Tables 3,4): fragmentation spectra were acquired for 2 minutes, in the appropriate mass range. All acquisitions were carried out in duplicate.

All samples were analyzed in triplicate for quantitation, and results were reported as mean values $\pm$ standard deviation ($\sigma$). Statistical analysis was performed using the one-way analysis of variance (ANOVA) and the Tukey test with a confidential level of 95% [40].

Phenolic chromatographic profiles of all samples were investigated through chemometric tools in order to study correlations among samples and to identify the variables affecting most the results. Multivariate statistical analysis is largely applied to study multiparametric data achieved by several techniques in order to simplify the interpretation of chemical problems and point out the relevant information as well as the variables that can be considered more informative for the analytical issue [41]. In this study, an explorative analysis based on Principal Component Analysis (PCA) was carried out on malts, wort and threshes, in order to study correlation among samples and between samples and variables.

Starting from chromatographic results referred to malts, wort and threshes, four matrices were imported into the chemometric package V-JDSU Unscrambler Lite (V-JDSU Unscrambler X 10.4, Camo software AS, Oslo, Norway) and different mathematical pre-treatments [42, 43] were tested in order to select the optimal one which separates samples according to the different chemical composition. The investigation of samples and variables correlation was performed by Principal Component Analysis using NIPALS algorithm.

Since all threshes were produced by the same mashing procedure, any differences can be ascribed to the starting raw material, that is represented by the four different base malts used herein, all barley malts. Similarities might be likely due to the same kind of raw cereal, namely barley, although differences may be expected mainly due to different environmental conditions, storage, or to the presence of other cereals in trace, as wheat, rye, oats [44, 45]. Conversely, differences might be likely due to the different malting procedures, leading from barley to the different malts [38, 39, 46, 47, 48, 49].

The untargeted mass spectral profile of the four samples of threshes obtained by infusion experiments resulted quite similar, as shown and discussed just below, suggesting that our results might be actually representative for this kind of brewery by-products, produced in very high amount all over the world.

Starting from the mass spectral analysis in positive ionization (ES+), carried out by direct infusion of the sample into the ESI source, a set of signals were well evidenced, corresponding to protonated ions [M+H]${}^{+}$ and/or to sodium [M+Na]${}^{+}$ and/or potassium [M+K]${}^{+}$ adducts, that are typically evidenced in positive ionization [36, 38, 39, 49, 50, 51, 52]. The ES+ mass spectral profile of BSG is shown in Fig. 1, in which threshes from Monaco, Extra Pale Maris Otter, Pilsner and Vienna malt are reported from bottom to top, respectively.

Among the signals, ten characteristic peaks, 1–10 in Fig. 1, were assigned on the base of the molecular mass, the behaviour in positive and negative ionization and fragmentation spectra, these last data also supported by literature. Peaks 1–9 at the monoisotopic values m/z 104, 116, 118, 120, 132, 147, 156, 166 and 175 were assigned to the protonated ion [M+H]${}^{+}$ of nine amino acids, namely $\gamma$-aminobutyric acid (GABA, 1), proline (Pro, 2), valine (Val, 3), threonine (Thr, 4), leucine/isoleucine (Leu/I, 5), lysine (Lys, 6), histidine (His, 7), phenylalanine (Phe, 8) and arginine (Arg, 9), respectively. The absence of signals for the corresponding deprotonated ion [M-H]${}^{-}$ in negative ionization (ES-) is in agreement with the amino acidic nature. Fragmentation data and reference literature [53, 54] are resumed in Table 3 (Ref. [53, 54]).

Although infusion experiments do not provide quantitative data, the relative abundance of compounds is given respect to the most abundant signal (100 %, height). In all BSG samples, GABA (1) is the main peak, followed by Pro (2), Val (3), with an inverted order in the case of Vienna, Leu/I (5) and Thr (4). Lys (6), His (7), Phe (8) and Arg (9) follow, with amounts around 10% compared to 1. The free amino acids evidenced by infusion experiments are in good agreement with literature reporting the content of free amino acids in craft BSG after pretreatment [25].

The presence of free amino acids in barley grains, malt, wort and finished beer were well reviewed recently [55]. Proline and glutamic acid, followed by leucine, are generally the main amino acidic component of barley; during malting, the activated proteolytic enzymes start to metabolize barley proteins, the cleavage of proteins continuing during mashing. Therefore, the proteogenic amino acidic profile of threshes is expected to reflect that one of the starting malt, with some changes occurred during mashing. The profile of BSG was compared with those ones obtained for the aqueous extract of malt (see Materials and Methods) and for wort (see Materials and Methods), obtaining four pools of samples BSG-malt-wort, namely Pilsner, Monaco, Vienna and Extra Pale Maris Otter. In Fig. 2, the ES+ mass spectral profiles of Monaco pool of samples, namely threshes, malt, wort from bottom to top, respectively, are shown as an example.

The amino acidic profiles of threshes, malt and wort (Fig. 2) are very similar from a qualitative point of view, although a different distribution can be observed among the samples. The main difference regards GABA (1), evidenced in all samples but becoming the main compound in threshes. GABA is not generally reported in barley [55], it is not a proteogenic amino acid and it generates by decarboxylation from glutamic acid, representative of barley [55]. Its presence in finished beers with different malt composition has been reported [38, 39]. Since glutamic acid was not evidenced in the profile of malt, as well in wort and threshes, GABA was likely generated from glutamic acid during malting, due to temperature or to the metabolic activity of microorganisms under different malting conditions. Moreover, the comparison among profiles in Fig. 2 suggests that among the identified amino acids, proline (2) is mainly transferred from malt to wort, while threshes remain richer in GABA (1). Noteworthy, besides the nutritional values, the free amino acids find applications as flavour ingredients in food, due to their organoleptic properties as the sweet taste (proline and threonine) and bitter taste (histidine, leucine/isoleucine, lysine, valine and phenylalanine) [25].

The other peaks evidenced in the ES+ profiles, 10 and 11 (Fig. 1 and Fig. 2), were identified as maltose and glucose, respectively: maltose 10 (M = 342 Da) is typically evidenced in ES+ by sodium and potassium adducts, [M+Na]${}^{+}$ = 365 m/z and [M+K]${}^{+}$ = 381 m/z, respectively, and glucose (M = 180 Da) is typically evidenced in ES+ by sodium and potassium adducts, [M+Na]${}^{+}$ = 203 m/z and [M+K]${}^{+}$ = 219 m/z, respectively [38, 39, 49].

Wort is obviously the sample richest in maltose and glucose, besides other oligosaccharides at higher m/z values (Fig. 2, wort Monaco), because starches are catalyzed into simple sugars, during mashing, by the activated carbohydrase enzymes of malt. In fact, the free simple sugars of wort are necessary for the subsequent fermentation by yeast to produce beer.

The ES- spectral analysis confirmed maltose 10 (Fig. 3), by the signals at m/z 341 and 377 corresponding to the deprotonated ion [M-H]${}^{-}$ and the chloride adduct [M+Cl]${}^{-}$, respectively, and glucose 11 (Fig. 3), by the signal at m/z 179 corresponding to the deprotonated ion [M-H]${}^{-}$. Peak at m/z 161, 12 in Fig. 3, was assigned to the ion [M-H]${}^{-}$ of an anydrohexose of glucose, and signal at m/z 503, 13 in Fig. 3, was assigned to the ion [M-H]${}^{-}$ of maltotriose, in good agreement with literature [38, 39, 49]. Signal at m/z 601, 14 in Fig. 3, was assigned to the ion [M-H]${}^{-}$ of an oligosaccharide, in agreement with literature [38, 39, 49]. The presence of glucose, maltose and maltotriose is in good agreement with literature [25].

Among the other signals, nine characteristic peaks at lower m/z values, 15–23 in Fig. 3, were assigned on the base of the molecular mass, the behaviour in negative and positive ionization, and fragmentation spectra, these last data also supported by literature. Signals 15–23 at the m/z monoisotopic values 87, 89, 97, 101, 103, 111, 133, 191 and 195 were assigned to the deprotonated ion [M-H]${}^{-}$ of nine small-size organic and inorganic acids, namely pyruvic acid (Pyr, 15), lactic acid (Lac, 16), phosphoric acid (Pho, 17), valerianic acid (Valer, 18), malonic acid (Malon, 19), 2-furoic acid (Fur, 20), malic acid (Mal, 21), citric acid (Cit, 22) and gluconic acid (Glu, 23), respectively. The absence of the corresponding ES+ signals is in agreement with their acidic nature. Fragmentation data and reference literature [56, 57, 58, 59, 60, 61] are resumed in Table 4 (Ref. [56, 57, 58, 59, 60, 61]).

By the comparison between ES+ and ES- profiles, higher peaks for glucose and its anhydrohexose, 11 and 12, respectively, are observed in ES-. Taking into account that their amount is obviously the same in ES+ and ES-, and admitting for them a similar behaviour in ES+ and ES-, the organic acids fraction content in threshes seems minor respect to the amino acids fraction. Similar profiles are evidenced for BSG from Monaco, Extra Pale Maris Otter and Pilsner, Lac (16) and Pho (17) being the most representative, followed by Fur (20), Valer (18) and Mal (21), although little differences in the distribution are observed. Conversely, some differences are evidenced for Vienna, from both a qualitative point of view and distribution: Pho (17) remains a dominant peak with Glu (23), found in very low percentage in the other samples. Noteworthy, gluconic acid occurs in nature from the oxidation of glucose, it is considered a valuable chemical for many applications [62], and its potential use in cancer therapy has been recently reported [63]. Gluconic acid, found abundant in Vienna threshes, was not detected in the corresponding malt and wort, suggesting that it is likely formed during mashing, but further research is needed to investigate on it. Furthermore, malonic acid (19) is found only in Vienna. Noteworthy, the content and proportion of organic acids in final malt is influenced by raw barley, differing for gene type and environmental conditions, as well by malting process and microorganisms [64].

Similarly to what reported above for profiling in ES+, the profile of BSG was compared with those ones obtained for the aqueous extract of malt (see Materials and Methods) and for wort (see Materials and Methods), obtaining four pools of samples BSG-malt-wort, namely Pilsner, Monaco, Vienna and Extra Pale Maris Otter. In Fig. 4, the ES- mass spectral profiles of Monaco pool of samples, namely malt, threshes and wort from bottom to top, respectively, in which the most significant compounds are evidenced, are shown as an example.

Simple sugars (10, 11, 12) are confirmed predominant in wort samples, 2-furoic acid (20), malic acid (21) and citric acid (22) remain in threshes, lactic acid (16) is the second abundant peak in wort as well in threshes. BSG has been evaluated as low cost raw material for the production of lactic acid by fermentation, for bioplastic applications [22]. The possible abundant presence of lactic acid in the by-product BSG encourages further studies, due to the increasing interest in lactic acid based bio-polymers [65].

Results obtained by the untargeted ESI-MS/MS infusion experiments on BSG samples laboratory-made obtained by base malts used in all the recipes for brewing, also supported by the comparative analysis of malts and wort, can be likely considered reasonably representative of the waste BSG, for which the mass spectral fingerprint has been provided in the present work. No similar comparative study had been reported before, at least to the best of our knowledge. Noteworthy, the BSG samples were extracted with water at room temperature under one hour stirring, conditions chosen because cheap and eco-friendly, and therefore potentially suitable for possible treatments of waste and by-products. However, the identified amino acids, organic acids and saccharides are molecules characterized by high polarity, and in fact they eluted within the first 3 minutes in the chromatographic run described below: by the fact, they are not retained under the used reverse phase chromatographic conditions. Therefore, further investigation is needed to evaluate the real possibility to selectively recover at least some of the identified compounds, for various applications.

However, the untargeted ESI-MS/MS profiling has been confirmed a powerful tool for fast and confident characterization of complex matrix.

The phenolic profile of the aqueous extracts of threshes was investigated by HPLC-PDA-ESI-MS in SIR mode with a method previously developed [37] for the simultaneous identification and quantitation of 26 phenolic compounds, in detail: p-hydroxybenzoic acid (pHBA), m-hydroxybenzoic acid (mHBA), o-hydroxybenzoic acid (salicyclic acid, SaA), 2,5-dihydroxybenzoic acid (gentisic acid, 2,5-DHBA), 2,6-dihydroxybenzoic acid (2,6-DHBA), 3,5-dihydroxybenzoic acid (3,5-DHBA), 3,4-dihydroxybenzoic acid (protocatechuic acid, PCA), 3,4,5-trihydroxybenzoic acid (gallic acid, GA), vanillic acid (VA), syringic acid (SyA), caffeic acid (CA), p-coumaric acid (CouA), sinapic acid (SA), ferulic acid (FA), 4-caffeoylquinic acid (4CQA), 5-caffeoylquinic acid (5CQA), catechin (Cat), rutin (Ru), myricitrin (My), quercetin-3-O-glucoside (Q3G), kampferol-3-O-rutinoside (K3R), hesperidin (He), quercetin (Q), kampferol (K), isoxanthohumol (IsoX), xanthohumol (X). Furthermore, comparison with literature [36, 66] was used for the identification of 3-feruloylquinic acid (3FQA) and 4-feruloylquinic acid (4FQA) that were quantitated as ferulic acid equivalents; ellagic acid was searched by comparison with standard; lastly, tryptophan was identified and quantitated in the same chromatographic run, as previously described [36]. The aqueous extracts of malts and the corresponding wort were also analyzed for a comparison.

The HPLC-PDA chromatograms of the aqueous extracts of threshes are shown in Fig. 5, in which samples Monaco, Extra Pale Maris Otter, Pilsner and Vienna are reported from bottom top, respectively.

Immediately a poor of peaks chromatogram catches the eye, suggesting that the matrix might contain a few polar molecules (water is the extraction solvent), or very low content of compounds (soft dilution 1:10), or molecules with no absorption in the UV-vis range (200–800 nm), or molecules not suitable for the ESI mass spectral analysis because of the absence of moieties able to be deprotonated or to be protonated (a very similar pattern is in fact observed in the ES- and ES+ full scan chromatograms, even less abundant). However, all the chromatograms are characterized by a peak at low retention time (t${}_{\text{R}}$ = 5.25 min) that was identified with standard and quantitated as tryptophan [36], and by another peak at high t${}_{\text{R}}$ (32.99 min), that was not identified and called unknown B in Fig. 5. Another abundant peak at 20.92 min was present only in Vienna threshes, it was not identified and called unknown A in Fig. 5. Therefore, Vienna sample provided a different chromatographic profile, in agreement with the other results discussed above. These three main peaks, well evidenced in threshes, differ enough in polarity (see elution time), suggesting that a potential selective recovery of tryptophan from the matrix might be reasonably possible.

Tryptophan and the phenolic profile were investigated by the Selected Ion Recording (SIR) mode, that is a highly selective and sensitive technique, and quantitation of each detected compound was carried by using the corresponding calibration curve, calculated with equal-weighted least-squares linear regression analysis of the SIR peak area against the standard nominal concentration, as previously reported [36, 37, 38, 39]. Results are resumed in Table 5 (Ref. [37, 38, 39]).

Tryptophan is found the most abundant metabolite in threshes, with a content in the range 62.33–75.35 $\mu$g per g of dried threshes. Tryptophan was not evidenced in the ESI-MS/MS profiling discussed above, neither in ES+ nor in ES-, although it was confirmed in the ESI-MS/MS analysis by selecting 203 m/z as precursor ion in ES- and by fragmenting it in collision cell, due to the sensitivity improved by the selectivity. Such a result may suggest that the amino acidic fraction as well the organic acids fraction should have higher content compared to tryptophan, up to hide its signal, in agreement with literature reporting for Pro, Val, Thr, Leu/I, Lys, His, Phe and Arg an average content in four craft BSG samples, after pretreatment, within 95.3–354.1 $\mu$g/g, dry base [25], actually from high to very high respect to our data of free tryptophan. No data on free tryptophan in BSG has been reported in the literature, at least up to our knowledge.

Last, tryptophan is chromatographically eluted under the used conditions, differently from the other amino acids and from the organic acids, and this might be an advantage to take in mind for future recovery strategies from the waste. Taking into account the main role of tryptophan in the metabolic pathways leading to many important biologically active molecules as serotonin, melatonin and vitamin B3 [31, 32, 33, 34, 35], these novel results encourage further studies for the selective recovery of free tryptophan from brewing by-products for a wide choice of applications, mainly in nutraceutical and pharmacological field.

The phenolic profile of threshes resulted quite poor, from both qualitative and quantitative point of view, as shown in Table 5: less than half among the searched compounds was detected.

Catechin resulted the second representative compound of threshes, after tryptophan, the content ranging between 1.13 and 4.24 $\mu$g per g of dry threshes. The low amount, obtained without any pretreatment of the sample, is in good agreement with literature, reporting even lower amounts (1.5 $\mu$g/g dry base [25], 1.08 $\mu$g/100 g dry matter [10]) in differently pretreated samples. Higher content of catechin were reported for hydro-alcoholic extracts, from 32 to 116 $\mu$g/g BSG by using 60% and 80% ethanol, respectively [27]. This comparison highlights the convenience of aqueous extraction of not pretreated BSG. Just a quick note, catechin is the most representative flavanol in nature, major building block of tannins, present also in cereals, among the others, and its antioxidant properties mainly prevent protein oxidation from oxidative stress [67]. Catechin is found the main phenolic antioxidant in threshes.

Ferulic acid FA was found as the third representative compound (range 0.68–1.73 $\mu$g per g of dry threshes) in all samples except Vienna. It was detected also as quinic acid ester, 4FQA and 3FQA, with content under LOQ, referring to the FA calibration curve [38]. Our results are in agreement [25] or lower (around 10 $\mu$g/g dry matter) than those reported in the literature for differently pre-treated samples or hydro-alcoholic extracts [26, 27, 68]. Higher contents were reported for samples after alkaline extraction [10].

Amounts $\le$1 $\mu$g per g of dry threshes were found in all samples for pHBA and VA in agreement with literature [25] or lower than those reported in the literature for differently pretreated samples or hydro-alcoholic extracts [26, 27, 68].

Conversely, most of phenolic compounds herein investigated, and generally not reported in the literature for BSG, were not detected in any threshes samples, namely the hydroxybenzoic acids 3,5DHBA, 2,5DHBA, 2,6DHBA, the hydroxycinnamic acids and quinic esters CA, 5CQA, 4CQA and SA, EA, most of flavonoids My, Ru, Q3G, K3R, He, K.

The absence of the prenylflavonoids X and IsoX, typical of beer, was expected because they come from hops. A very few other compounds were detected in low amount or under LOQ [37, 38, 39].

The analysis of malts evidenced a quite similar phenolic profile, in which in general Cat and FA are the most abundant, followed by CouA, VA and SA, besides tryptophan, whose higher content is confirmed also in the starting malt.

Noteworthy, the same results were obtained also after 2 hours extraction, suggesting that 1 hour is sufficient for a good extraction of metabolites.

If the phenolic profile of threshes appears quite definite from a qualitative point of view, differences among the samples, regarding the individual phenolic content, were well evidenced by the ANOVA test (Table 5, lowercase letters for the analysis among threshes). Differences among malt samples were also confirmed (Table 5, capital letters for the analysis among malts).

Based on the similarities occurring between threshes and the corresponding starting malt, although with an obvious decreased individual amount in threshes, residual metabolites in BSG available for a possible waste-recovery can be therefore expected, known the metabolic composition of the starting malt.

According to results reported herein, BSG can represents a good source of valuable compounds, first of all tryptophan and catechin, with satisfactory results obtained through a simple, fast, cheap and eco-friendly handling of the by-product.

In Table 6, for completeness, the phenolic distribution, beside tryptophan content, in the four wort samples is also reported. Due to the liquid nature of the matrix, the amounts are reported as $\mu$g/L of wort. The phenolic profile is confirmed, even enriched in some cases, likely due to mashing conditions, and differences are again evidenced by the ANOVA statistical analysis.

Interesting to note, such a comparative study well evidenced the phenolic contribution to beer antioxidants due to malt.

A more in-depth statistical analysis was carried on the four samples sets of threshes, malt and wort, with the aim of identifying specific parameters able to ensure a rapid and accurate quality control of raw materials and beer fermentation process. Therefore, a careful evaluation of the phenolic profiles of malts, wort and threshes, resulting from HPLC-ESI-MS/MS characterization, was performed using chemometric tools.

The exploratory PCA provided scores (Fig. 6a, 6c, 6e) and loadings plots (Fig. 6b, 6d, 6f), useful to highlight samples and variables correlations, respectively. The plot of scores points out similarities and differences between samples based on the distance among each other, showing the presence of possible samples grouping. Referring to scores plots reported in Fig. 6 (a, c and e), samples belonging to the same base malts group (Vienna, Monaco, Pilsner and Extra Pale Maris Otter) were well grouped and clearly separated from the others in the space described by the first two principal components (PCs), which explained the 75% of the total variance for malts scores plot (Fig. 6a), the 69% for wort (Fig. 6c) and the 78% for threshes (Fig. 6e). Factor loadings analysis permitted to highlight contribution of each variable on the concerns PCs: differences in phenolic characteristic compounds of the four studied samples groups and differences in the chemical composition of raw and waste materials as well as products of fermentation. Relevant information can be achieved considering the loadings position in the space: in fact, the longer the vector along a specific component, the greater the statistical weight of the variables over the PC. Based on this consideration, variables responsible for malts samples differentiation according to PC-1 and PC-2 may be identified: CouA, 3FQA, VA and Cat affected more PC-1 with respect to the other parameters, while SA and FA were considered more significant in PC-2 computing (Fig. 6b).

Observing both scores (Fig. 6a) and loadings plot (Fig. 6b), samples located in the right side of the plot (Pilsner and Vienna) were found characterized by higher content of 3FQA, CouA and VA respect to the others. On the contrary, Extra Pale Maris Otter malts were located in the left side of the plot and showed a higher value of Cat. Moreover, FA and SA were associated to PC-2 samples distribution resulting higher in Extra Pale Maris Otter and Vienna samples respect to Pilsner and Monaco. An important consideration could be made on the TRP variable. Position of the phenolic compound in the contribution plot allowed to detect its slight but significant weight on PC-1 and thus in the sample’s differentiation according to malt base origin.

Taking into account the same approach for wort and threshes loadings plots interpretation (Fig. 6d and 6f), it was possible to observe the contribution of Cat, CouA and 3FQA in samples grouping as a function of the different chemical composition of the four studied base malts. Moreover, wort loadings plot pointed out the possibility to examine other phenolic compounds for fermentation process monitoring such as 4FQA and SyA. As well highlighted in Fig. 6d, PC-1 was found to be affected by these molecules resulting higher in wort from Monaco malt with respect to the others.

A further significant outcome of the exploratory analysis consisted in the ability to monitor malts and threshers, simultaneously. Scores plot reported in Fig. 7a showed a clear distinction of malts and threshers samples according to PC-1 (63% of explained variance). As a result of the loadings plot interpretation (Fig. 7b), CouA, SA, 3FQA, FA and VA were found to be the variables affecting most PC-1 and thus molecules responsible for samples separation, confirming the importance of this parameters in the monitoring of raw and waste materials involved in beer process production.

Outcomes demonstrated the possibility to monitor a small number of molecules to ensure the quality of a final product that satisfies consumer’s request. Further, in this framework, chemometrics proved to be a valuable tool for time-saving and sensitive multiparametric chromatographic data evaluation that permitted to accurately detect characteristic parameters.