Complementary Sample Preparation Strategies for Analysis of Cereal β-Glucan Oxidation Products by UPLC-MS/MS

The oxidation of cereal (1→3,1→4)-β-D-glucan can influence the health promoting and technological properties of this linear, soluble homopolysaccharide by introduction of new functional groups or chain scission. Apart from deliberate oxidative modifications, oxidation of β-glucan can already occur during processing and storage, which is mediated by hydroxyl radicals (HO•) formed by the Fenton reaction. We present four complementary sample preparation strategies to investigate oat and barley β-glucan oxidation products by hydrophilic interaction ultra-performance liquid chromatography-tandem mass spectrometry (UPLC-MS/MS), employing selective enzymatic digestion, graphitized carbon solid phase extraction (SPE), and functional group labeling techniques. The combination of these methods allows for detection of both lytic (C1, C3/4, C5) and non-lytic (C2, C4/3, C6) oxidation products resulting from HO•-attack at different glucose-carbons. By treating oxidized β-glucan with lichenase and β-glucosidase, only oxidized parts of the polymer remained in oligomeric form, which could be separated by SPE from the vast majority of non-oxidized glucose units. This allowed for the detection of oligomers with mid-chain glucuronic acids (C6) and carbonyls, as well as carbonyls at the non-reducing end from lytic C3/C4 oxidation. Neutral reducing ends were detected by reductive amination with anthranilic acid/amide as labeled glucose and cross-ring cleaved units (arabinose, erythrose) after enzyme treatment and SPE. New acidic chain termini were observed by carbodiimide-mediated amidation of carboxylic acids as anilides of gluconic, arabinonic, and erythronic acids. Hence, a full characterization of all types of oxidation products was possible by combining complementary sample preparation strategies. Differences in fine structure depending on source (oat vs. barley) translates to the ratio of observed oxidized oligomers, with in-depth analysis corroborating a random HO•-attack on glucose units irrespective of glycosidic linkage and neighborhood. The method was demonstrated to be (1) sufficiently sensitive to allow for the analysis of oxidation products also from a mild ascorbate-driven Fenton reaction, and (2) to be specific for cereal β-glucan even in the presence of other co-oxidized polysaccharides. This opens doors to applications in food processing to assess potential oxidations and provides the detailed structural basis to understand the effect oxidized functional groups have on β-glucan's health promoting and technological properties.


Figure S4
Comparison of UPLC-MS base peak ion (BPI) and extracted ion chromatograms (XIC) from (A) harsh (100 mM H2O2) and (B) mild (250 µM AH2) oxidation of BBG after lichenase+β-glucosidase treatment/SPE (strategy II; 10x concentrated by evaporation under a stream of N2). Negative ion mode, a slower aqueous ACN gradient (0.17 mL/min) with 15 cm BEH amide column, and basic eluent (0.1% NH3 additive) were used. The peaks are labeled with their base peak m/z. Note that the mild oxidation conditions lead to a different product profile of oxo-Glcn species with oxo GlcGlc(n-1) not being the predominant product as is the case for n = 3,4 under the harsh conditions. This phenomenon is subject to further investigation. The BPI of the mild oxidation (B) shows native Glcn peaks (m/z 341, 503, 665, 827), all of which also occur in the non-oxidized control in the same proportions (whereas no oxo-Glcn signals in control), and might be residues of DP3-5 (from lichenase treatment) that were not fully hydrolyzed by β-glucosidase. The sharp m/z 341 peak (13 min) originates from the catalase material (disaccharide). *, oxo-Glcn isomers with the carbonyl not at the non-reducing end. **, in-source fragmentation peak of the respective n+1 species.

Figure S5
Comparison of (A) negative and (B) positive ion mode in the UPLC-MS analysis of BBG oxidation products after lichenase & β-glucosidase/SPE treatment (strategy II) with the respective base peak ion chromatograms (BPI) and extracted ion chromatograms (XIC) of oxo-Glc4 as example. Note that in the positive mode, isomeric oxo-Glc4 products have different preferences regarding ionization, with the main oxo GlcGlc3 product ionizing preferably as [M+H]or [M+H2O+NH4/Na] + , in agreement with the carbonyl-hydrate (geminal diol) equilibrium of the oxo-group (R2C=O + H2O = R2C(OH)2). Other oxo-Glc4 isomers with mid-chain oxo-groups (labeled with * in the BPIs) preferably ionize as ammonium or sodium adducts in the positive ion mode.

Figure S6 (A)
Collision induced dissociation (CID) of Glc-2AB from reductive amination strategy III: Proposed structures for the observed most prominent MS/MS fragments. MS/MS spectra of (B) 2-AB labeled standards, and (C) of 2-AB labeled reducing termini from harsh BBG oxidation after enzyme treatment/SPE (fraction 2; see Figure 7b for BPI). Fragments are labeled for (C), with labels in purple containing the 2-AB moiety, while labels in blue do not. Differences in relative fragment intensities of Ery-and Ara-2AB between standards and BBG oxidation products might originate from Ara and Ery being different isobaric pentoses or tetroses, respectively, e.g. formed through epimerization. Oxidation product GlcGlc-2AB could be identified to be the β-(1→3)-linked isomer due to the observed MS/MS fragment m/z 191 and matching retention time (Rt) of ~3.3 min (β-(1→4)-isomer elutes earlier). Due to lack of standards, the linkage type of GlcAra-2AB could not be unambiguously confirmed by MS/MS, but is assumed to be β-(1→2)-linked originating from a β-(1→3)-Glc unit, analogous to the β-glucosidase resistant, confirmed Glcβ(1→3)Glc-2AB.

Figure S7
Detection of lytic C5-oxidation products with C=O labeling strategy III (reductive amination): 1 Proposed mechanisms to explain the observed epimeric mixture of Glc 5oxo Glc-2ABo (2 peaks), but not for Glc 5oxo Glc-2ABo (predominantly 1 peak), on the basis of observations made by Baxter and Reitz (1994) in their aza-sugar synthesis from 5-oxo-hexoses (see Figure S8a for chromatogram, and Figure 8 for the full mechanism). Under the assumption that the hydride attack (NaBH3CN) on the intermediate iminium ion takes place axially from the side that avoids formation of a boat-conformation transition state, a (A) β-(1→3)-linked unit has disfavoring steric obstacles for both possible intermediates, namely 1,2-allylic strain (A1,2; left) pushing the equilibrium to the right side, and a blocked top side from R = Glcm (right). Consequently, products from both intermediates are formed. (B) A β-(1→4)-linked unit has two factors favoring the conformation on the right-hand side: higher 1,2-allylic strain due to R = Glcm (left), and an accessible top that additionally might have a directing effect of the free hydroxyl group at C3 by anchimeric assistance (right). This would explain why predominantly one product was formed for 5oxo Glc-2ABo (β-(1→4)-linked before β-glucosidase), while both epimers were detected in comparable amounts for β-(1→3)-linked Glc 5oxo Glc-2ABo. The high dependence on selectivity regarding substitution pattern and configuration was also observed by Baxter and Reitz (1994), as for instance unsubstituted 5-oxo-glucose gave high selectivity (>95%) for one epimer after reductive amination, while mannose (C2 epimer of glucose) and per-O-acylated 5-oxo-Glc had low selectivity (67:33 and ~50:50, respectively). 1 5oxo Glc is the only primary oxidation product with 6 carbons expected to result in such a cyclization: For instance, a γ-keto-aldehyde (C4-oxidation) could also lead to a cyclisation by reductive amination resulting in a 5-membered pyrrolidine derivative with the same m/z. However, the C4-oxidation would have to occur on a reducing end for C1 to be a free aldehyde (in equilibrium with its hemiacetal form), and reducing ends are in low amounts compared to the total sugar units, most of which (>99%) are mid-chain units. Lytic C5oxidation is the only process that directly leads to a suitable substrate for the observed cyclisation without the need for two oxidation processes happening on the same glucose unit. Misidentification of dehydration side reactions can also be excluded: A loss of 18 Da corresponds to -H2O, or a dehydration, but cannot be a side product of Glc reducing end labeling, as test reactions with glucose and oligomer standards under identical reductive amination conditions showed no such products. It also cannot be a result of lactone formation with the carboxyl of the label, as this would lead to the correct m/z for 2-AA (-H2O), but not for 2-AB (-NH3; would give the same m/z of 282.10 as 5oxo Glc-2AA, which was not observed).

Figure S8
UPLC-MS/MS of oxo-products from BBG oxidation (harsh conditions) detected as 2-AB labeled species in SPE fraction 1 after reductive amination, enzyme treatment & SPE (strategy III; negative ion mode, basic eluent). (A) Overlaid extracted ion chromatogram (XIC) and (B) MS/MS spectra of C=O labeled 5-oxo-reducing ends (stereocenter * of epimers set arbitrarily). The inset labeled with "T" is the structure of 2-AB labeled L-threo-tetrodialdose ( oxo Tet-2ABo), which is also a C5-oxidation product that was also observed by Schuchmann & von Sonntag in their Glc irradiation study (Schuchmann and Sonntag, 1977). (C) XICs of labeled oxo-Glcn products and (D) their respective MS/MS spectra. For each n, the average MS/MS is shown, as surprisingly little differences were found between the isobaric individual peaks of 2AB-(oxo-Glcn) resolved by UPLC-MS. 2 The fragments are labeled assuming the labeled oxo-group being at the non-reducing end as in the depicted structures (since they are the main isomers as detected in strategy II; see Figure 5a). 3 Fragments labels in red contain the oxidized unit (incl. 2-AB), while labels in blue do not.

Figure S10
MS/MS spectra of CO2H labeled standards and oxidation products from EDC-mediated amidation of carboxylic acids with PhNH2 (sample preparation strategy IV). Note that the observed cross-ring fragments (A-ions) of labeled GlcAGlc (m/z 430) indicate a β-(1→4)-linkage, which is unexpected and might be a peeling product of GlcAGlc2 (bottom spectrum).
b Oxidized BG solution treated with phosphate buffer and catalase first (except for strategy I: SPE). c Monosaccharide and other small products (<C6) are lost in all cases due to SPE purification/fractionation. d (Glc) and (oxo-) in parenthesis refer to products both with and without the indicated additional structural feature. e Labeling by reductive amination with 2-AB (as example, 2-AA also possible) and NaBH3CN. Enzymes: lichenase + β-glucosidase treatment. f Labeling by amidation with aniline (PhNH2) and EDC. Enzymes: lichenase + β-glucosidase treatment. g Information only partially lost, namely β-(1→4)-linked C1-oxidation products.