Profiling of polar ionogenic metabolites in Polish wines by capillary electrophoresis‐mass spectrometry

Abstract The composition of wine is determined by a complex interaction between environmental factors, genetic factors (i.e., grape varieties), and winemaking practices (including technology and storage). Metabolomics using NMR spectroscopy, GC‐MS, and/or LC‐MS has shown to be a useful approach for assessing the origin, authenticity, and quality of various wines. Nonetheless, the use of additional analytical techniques with complementary separation mechanisms may aid in the deeper understanding of wine's metabolic processes. In this study, we demonstrate that CE‐MS is a very suitable approach for the efficient profiling of polar ionogenic metabolites in wines. Without using any sample preparation or derivatization, wine was analyzed using a 10‐min CE‐MS workflow with interday RSD values for 31 polar and charged metabolites below 3.8% and 23% for migration times and peak areas, respectively. The utility of this workflow for the global profiling of polar ionogenic metabolites in wine was evaluated by analyzing different cool‐climate Polish wine samples.

organic acids, polyphenols, vitamins, tannins, anthocyanins, amino acids, and biogenic amines. Metabolomics is a powerful tool providing a holistic view of the unique chemical composition of small molecules (≤1500 Da) in (biological) samples resulting from metabolic processes. This approach has gained momentum for the evaluation of food quality, toxicology, and processing over the past decades [2,3], and already has a wide application in viticulture in order to evaluate the quality, authenticity [4], factors affecting the sensory characteristic of wine, and terroir effect [5].
In Poland, the most commonly cultivated grapes are hybrid grapevine species that resulted from the crossing of European grapevines (Vitis vinifera) with North American grapevines (such as Vitis rupestis, Vitis riparia or Vitis abrusca) or Asian grapevines (such as Vitis amurensis). Crossing two or even more species of Vitis resulted in hybrid species of so-called cold-climate wines, which are resistant to temperatures below -30 • C and also less sensitive to fungal diseases. Specific characteristic features of Polish wines include a higher acidity and lower sugar content as compared to wines from warmer regions [6].
So far, various analytical tools have been developed for metabolomics analysis of grapes and wines [7], in particular GC-MS, LC-MS, and NMR have been considered for this purpose. CE-MS is an analytical approach that offers excellent selectivity for resolving a wide range of polar and charged metabolites as compared to reversed-phase LC or hydrophilic interaction chromatography (HILIC) [8,9]. The latter often suffers from a relatively poor retention time reproducibility, complex separation mechanisms, and long equilibration times of columns. The potential of CE-MS for the analysis of polar ionogenic metabolites in wine has only been reported in a limited number of studies so far [10][11][12][13]. Here, we propose a CE-MS-based analytical workflow for the untargeted profiling of polar and charged metabolites in wine and show the utility of this approach by comparative metabolic profiling of cool climate wines from Poland.
For the CE-MS methodology, we refer to our procedures described previously [14]. Briefly, CE-MS experiments were carried out on a 7100 CE system hyphenated with a 6230 TOF, both from Agilent Technologies. CE-MS coupling was realized via a co-axial sheath-liquid ESI interface equipped with a triple-tube sprayer, and sheath-liquid was of a mixture of water and isopropanol (50:50, v/v) with 0.03% acetic acid, which was delivered at 3 µl/min. As background electrolyte (BGE), 10% (v/v) acetic acid was used. Stable-isotope-labeled (SIL) histamine (5 µg/ml) was used as internal standard. CE-MS experimental data were acquired in positive ionization mode, between 50 and 1000 m/z with an acquisition rate of 1.5 spectra/s. The following MS settings were used; nebulizer gas: 0 psi, sheath gas (nitrogen) flow rate: 11 L/min, sheath gas temperature: 100 • C, ESI capillary voltage: 5500 V, fragmentor voltage: 100 V, skimmer voltage: 50 V. When in-source fragmentation was required for identification purposes, a fragmentor voltage of 200 V was used. Data treatment and analysis was performed as described previously [15]. Wine samples were purchased from Polish wine stores and included 10 red and 10 white wine samples (see Table S1). Wine samples were ultrasonicated and filtered by a 0.45 µM cellulose filter and stored under dark conditions at room temperature (21 • C).
Untargeted metabolomics of wine is a powerful tool for the assessment of wine authenticity and quality. However, before such an approach can be used for this purpose, it is important to assess the performance of CE-MS first for the analysis of target compounds. Only with acceptable performance metrics (i.e., for repeatability: area RSD% < 25%, migration time RSD% < 10%, linear response function and representable LOD values) obtained for targeted analysis the CE-MS method can be used for untargeted profiling of metabolites in wine. The analytical performance was evaluated using a metabolite mixture composed of 32 metabolites by considering aspects such as repeatability, response function, and LODs (Tables S2 and  S3).
A linear response (and with R 2 > 0.981) for the target metabolites in the concentration range from 0.05 to 10 µM was obtained with LODs ranging from 0.002 to 0.218 µg/ml (Table S4). In comparison with previous studies reporting the use of CE-MS for analysis of metabolites in wine, LODs were at least 2 to 12 times lower for biogenic amines (except for cadaverine) [10,13]. This improvement is probably due to the use of different CE-MS separation conditions, such as BGE composition and a lower sheath-liquid flow rate as no nebulizer gas was applied [9]. Repeatability of the CE-MS method for direct profiling of metabolites was assessed based on the consecutive analyses of wine samples spiked with metabolite standards (2.5 µg/ml). Intra-and interday RSD values for peak areas of all analytes were below 17% (n = 5) and 23% (n = 15) (except for cadaverine), respectively (Table S4). Migration time repeatability was assessed without internal standard correction, and was below 1.7% and 2.5% for intra-and interday analysis, respectively. Given that wine samples were directly analyzed by CE-MS, the obtained figures of merits for repeatability could be considered acceptable for comparative metabolic profiling studies. Figure 1 illustrates the applicability of CE-MS for the direct analysis of biogenic amines and amino acids in pooled red and white wines, respectively. These compound classes could be analyzed within 10 min by using an additional pressure of 40 mbar at the CE inlet during separation, thereby still maintaining a partial separation for the isobaric isomers isoleucine and leucine (R = 0.5). In case a better separation would be required, then the use of an additional pressure could be omitted and/or a longer separation capillary could be used.
Matrix effects were assessed using the standard addition method [16]. SIL standards were spiked into pooled wine at F I G U R E 1 Extracted-ion electropherograms obtained by CE-MS for the targeted analysis of (top) pooled white wine and (bottom) pooled red wine. Separation conditions: BGE, 10% acetic acid; sample injection volume 27.4 nl; separation voltage: 30 kV concentrations ranging from 0.05 to 25 µg/ml, and resulting response curves were compared to the accompanying response curves obtained when the SIL standards were dissolved in BGE ( Figure S5). The slopes were different for all compounds (up to 45% lower for wine samples), indicating that all metabolites experienced a matrix effect. Therefore, for quantitative studies, calibration curves constructed in wine or a SIL internal standard for each compound needs to be used to account for matrix effects.
Next, the CE-MS workflow was used for untargeted profiling of polar ionogenic metabolites in two groups of samples, that is, red wine (n = 10) and white wine (n = 10). Quality control (QC) samples (n = 5) were prepared by pooling all the wine samples and were analyzed periodically along the sequence to evaluate the performance of the method. A total of 94 features were detected (Table 1), after removing noise signals, duplicates, adducts and frag-ments. Data quality was assessed by clustering QC samples measurements in an unsupervised PCA-X model. The model showed a R 2 = 0.735, which indicated the high quality of the analysis (Figure 2) and the clustering trends of the groups were observed.
The compounds that play the most significant role in the discrimination between wines of different types and origin are amino acids. Key amino acids include proline, which is the most abundant amino acid in grapes. Its level is determined by the grape variety and aromatic amino acid phenylalanine, whose level depends on grape variety and alcoholic fermentation, where it is used by bacteria, fungi, and yeast to produce a highly polar aromatic alcohol, phenethyl alcohol [17]. A supervised Orthogonal partial least-squares-discriminant analysis model was built to evaluate the metabolic differences between white and red wine (Figure 2). The model showed a high grade of discrimination between the groups of samples and good quality parameters (R 2 = 0.947, Q 2 = 0.782). The model was validated by CV-ANOVA (p-value = 1.06 × 10 -03 ) and by the cross-validation leaving 1/3 out approach, showing a prediction accuracy of 100% ( Figure 2). In order to identify the features that were statistically significant by multivariate data analysis (MVDA) analysis, the confidence intervals of Jack-Knife, correlation p( corr ) > |0.5|, and variables importance in projection (VIP) > 1 were calculated. Additionally, the univariate data analysis (UVDA) statistical analysis was performed in order to obtain the statistical significance of each compound in the comparison of both groups. Metabolites with p < 0.05 were selected as significant metabolites using the Mann-Whitney U test; for correction of comparisons, the Benjamini-Hochberg method was applied to all p-values to control the false discovery rate (FDR) at the q = 0.05 level. The overall statistical analysis revealed 45 metabolic features as significantly different between the groups (Table S6). The annotation of these features was performed by m/z search in the FooDB database and CMM, a search tool that integrates different databases (Kegg, HMDB, LipidMaps, METLIN, NPAtlas, KNAp-SAcK, MINE) and an in-house library [18]. The annota-tion was carried out considering mass accuracy (20 ppm as maximum error as recommended [19]), electrophoretic mobility (<5% error to effective mobility database [9,20]), isotopic pattern, and adduct formation (confidence level 3). This was considered as tentative annotation. Additionally, in order to increase this confidence level, in-source fragmentation was performed (confidence level 2) [15], and when available, some metabolites were identified using commercial standards (confidence level 1). For unknown metabolites, only m/z was considered (confidence level 4). Interestingly, cyclic argininosuccinic acid was found to be one of the metabolites responsible for the classification of the two sets of wine samples, expressed more in white wines. This compound is found in a dynamic equilibrium with argininosuccinic acid in its open form; at an acidic pH as is the case in wines, cyclic argininosuccinic acid predominates [21].
Overall, a CE-MS workflow for the direct profiling of polar ionogenic metabolites in wine is proposed and a proof-of-principle study utilizing white and red coolclimate wines originating from Poland revealed the potential of this approach for assessing wine authenticity and quality in follow-up studies.