1H NMR spectroscopy for profiling complex carbohydrate mixtures in non-fractionated beer

Food Chemistry 150 (2014) 65–72

Contents lists available at ScienceDirect

Food Chemistry journal homepage: www.elsevier.com/locate/foodchem

Analytical Methods 1

H NMR spectroscopy for profiling complex carbohydrate mixtures in non-fractionated beer Bent O. Petersen a, Mathias Nilsson b,c, Marie Bøjstrup a, Ole Hindsgaul a, Sebastian Meier a,⇑ a

Carlsberg Laboratory, Gamle Carlsberg Vej 10, 1799 Copenhagen V, Denmark School of Chemistry, University of Manchester, Oxford Road, Manchester M13 9PL, UK c Faculty of Science, Dept. of Food Science, University of Copenhagen, Rolighedsvej 30, DK-1958 Frederiksberg C, Denmark b

a r t i c l e

i n f o

Article history: Received 26 February 2013 Received in revised form 24 October 2013 Accepted 26 October 2013 Available online 1 November 2013 Keywords: Dietary fibre Carbohydrates High-resolution spectroscopy Limit dextrins Mixture analysis NMR profiling

a b s t r a c t A plethora of biological and biotechnological processes involve the enzymatic remodelling of carbohydrates in complex mixtures whose compositions affect both the processes and products. In the current study, we employed high-resolution 1H NMR spectroscopy for the analysis of cereal-derived carbohydrate mixtures as exemplified on six beer samples of different styles. Structural assignments of more than 50 carbohydrate moieties were obtained using 1H1–1H2 groups as structural reporters. Spectroscopically resolved carbohydrates include more than ’’20 different’’ small carbohydrates with more than 38 isomeric forms in addition to cereal polysaccharide fragments with suspected organoleptic and prebiotic function. Structural motifs at the cleavage sites of starch, b-glucan and arabinoxylan fragments were identified, showing different extent and specificity of enzymatic polysaccharide cleavage during the production of different beer samples. Diffusion ordered spectroscopy supplied independent size information for the characterisation and identification of polysaccharide fragments, indicating the presence especially of high molecular weight arabinoxylan fragments in the final beer. Ó 2013 Elsevier Ltd. All rights reserved.

1. Introduction In modern agricultural societies, carbohydrates remain the major source of food energy intake (Cordain et al., 2005). Besides their caloric value, carbohydrates are well-suited food ingredients due to their physicochemical, organoleptic and prebiotic properties. The digestion of carbohydrates and the resulting contribution to blood glucose levels or prebiosis naturally depends on the structural complexity of the substrate (Kaput & Rodriguez, 2004). Diets rich in non-digestible oligo- and polysaccharides have been implicated in an increasing range of health benefits, including reduced risks of allergy, pathogen binding to epithelial cells, cancer, osteoporosis and obesity with its associated diseases type 2 diabetes and cardiovascular disease (Andersson, Porras, Hanson, Lagergard, & Svanborg-Eden, 1986; Kaput & Rodriguez, 2004; Kunz, Rudloff, Baier, Klein, & Strobel, 2000; Van Loo et al., 1999). Commercial food grade carbohydrates are mostly produced by enzyme action in batch processes that result in mixtures of feedstock carbohydrate, oligosaccharides and simple sugars, as cost-intensive purification steps are commonly avoided (Maitin & Rastall, 2005). Consumer awareness and scientific interest in the compositional analysis of foodstuffs and their possible impact on nutrition, health and well-being have been surging. Consequently, several

⇑ Corresponding author. E-mail address: [email protected] (S. Meier). 0308-8146/$ - see front matter Ó 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.foodchem.2013.10.136

NMR spectroscopic studies have analysed food quality, origin and authenticity mostly based upon amino acid, organic acid and aromatic metabolite profiles in the product (Wishart, 2008). Due to the severe signal overlap in the carbohydrate region and the overlap of the water signal with the carbohydrate spectral region, less attention is usually paid to the analysis of complex carbohydrate mixtures. At high magnetic field, however, spectral dispersion increases due to higher resonance frequencies. In addition, spectral appearance improves due to the reduction of strong-coupling induced distortions and favourable line narrowing as a consequence of longer 1H T2 values (Blundell, Reed, Overduin, & Almond, 2006). In addition, cryogenically cooled probes afford sensitivity gains that facilitate the identification of components by heteronuclear NMR spectroscopy at natural 13C isotope abundance (Kovacs, Moskau, & Spraul, 2005) and permit resolution enhancement using experimental approaches or processing schemes that are accompanied by sensitivity penalties (Aguilar, Faulkner, Nilsson, & Morris, 2010; Nilsson & Morris, 2007; Zangger & Sterk, 1997). At the same time, reference compound databases support the compositional analysis of foodstuffs, including some of their NMR detectable carbohydrate components (Cui et al., 2008; Jewison et al., 2012; Markley et al., 2008; Wishart et al., 2007). We assess the prospect of employing high-resolution NMR spectroscopy in the analysis of complex carbohydrate mixtures with emphasis on degradation products of cereal polysaccharides. Oligosaccharides were analysed in non-fractionated beer samples in order to draw conclusions about abundant and rate limiting

66

B.O. Petersen et al. / Food Chemistry 150 (2014) 65–72

enzyme activities in production processes. We employed H1–H2 cross peaks in phase sensitive double quantum filtered COSY experiments as reporter signals to thoroughly identify structural motifs of a-glucan, b-glucan and arabinoxylan fragments in beers. The completeness of starch amylolysis during beer production could be assessed by identifying and quantifying structural motifs in branched oligosaccharides deriving from starch degradation (limit dextrins). Such quantification became feasible with the use of homonuclear decoupling to substantially increase the resolution of one-dimensional spectra. Even in spectral regions with severe signal overlap, diffusion ordered spectroscopy (DOSY) (Johnson, 1999) near the highest available spectrometer fields permitted resolving beer carbohydrates according to their size. In total, more than 50 carbohydrate spin systems can be assigned to different carbohydrate types and oligosaccharide fragments. The ability to directly identify such structural motifs in polysaccharide-derived fragments, quantify fragment types and assess their molecular size in complex reaction mixtures encourages the increasingly detailed monitoring of glycan processing. 2. Materials and methods 2.1. Samples Beer was used as a cereal derived commercial product containing complex carbohydrate mixtures for a targeted approach to resolve cleavage site motifs in fragments of cereal polysaccharides alongside smaller carbohydrate constituents without sample fractionation. In order to demonstrate the extent of different enzyme usage in beer production, six beers representing different styles within the lager and ale beer categories were purchased from local stores, sampling popular commercial products within different beer categories. Lager and ale beer styles are produced by different yeast strains and at different temperatures. The beer styles were bottomfermented asian lager, Danish alcohol free lager and organic pilsner, Danish top-fermented ale and porter as well as a top fermented German wheat beer. Samples were prepared from fresh beverages following degassing of approximately 3 ml beer in a 5 ml vial by repeated sample agitation and release of CO2, until pressure buildup upon agitation stopped. 600 ll of these degassed samples were mixed with 5% (v/v) D2O (99.9%; Cambridge Isotope Laboratories, Andover, MA, USA) for NMR spectroscopic analysis. Due to the focus on non-volatile beer carbohydrates with NMR signals in the vicinity of the water signal, additional samples were prepared by lyophilisation of 600 ll fresh beer prior to resuspension in 600 ll D2O.

CryoProbe and an 18.7 T magnet (Oxford Magnet Technology, Oxford, UK), or on a 600 MHz Bruker DRX spectrometer equipped with a room temperature BBO SmartProbe. Homo- and heteronuclear spectra for component identification included DQF-COSY (4096  512 complex data points with acquisition times of 854 and 107 ms in the direct and indirect dimension, respectively), TOCSY with 10 kHz spin lock field during a mixing time of 60 ms (2048  256 complex data points sampling 307 and 38 ms, respectively), multiplicity edited 1H–13C HSQC (1024  256 complex data points with acquisition times of 153 and 11 ms), sensitivity-enhanced 1H–13C HSQC (1024  200 complex data points with acquisition times of 153 and 331 ms), and 1H–13C HSQC TOCSY (1024  256 complex data points sampling 153 and 10 ms). Diffusion ordered spectra of the lager beer were recorded with the recently proposed Oneshot 45 experiment suppressing multiplet phase distortion caused by J-modulation effects (Botana, Aguilar, Nilsson, & Morris, 2011). An airflow of 670 l/h and a temperature of 37 °C were used on an 800 MHz Bruker DRX spectrometer. The sample was not rotated, nor was an added reference used. For 32 gradient amplitudes that were incremented in steps of gradient squared between 3.6 and 27 G cm 1, 32 transients of 16384 complex data points (acquisition time 1.63 s) were acquired using a diffusion delay D = 200 ms and a gradient pulse duration d = 1.8 ms. Resultant decays were fit to the Stejskal–Tanner equation to yield translational self-diffusion coefficients (Stejskal & Tanner, 1965). Pure chemical shift 1H spectra were acquired with an improved version (Aguilar et al., 2010; Nilsson & Morris, 2007) of the homonuclear broadband decoupling method originally proposed by Zangger and Sterk (1997) and employed a rSNOB selective 180° pulse of 28 ms duration (bandwidth 83 Hz) with a slice selective gradient of 0.2 G cm–1. A total of 30 increments were acquired with sw1 = 100 Hz and a spectral width of 3000 Hz in order to warrant smooth chemical shift evolution between chunks. One-dimensional 1H spectra were recorded with inter-scan delays of 3 s. 2.4. Data processing and analysis

Wheat arabinoxylan, barley b-glucan, and barley b-limit dextrin produced by the treatment of lintnerized maize starch with b-amylase were purchased from Megazyme (Bray, Ireland). Wheat arabinoxylan (3 mg) or barley b-glucan (2 mg) were dissolved to 5 mg ml 1 in 600 ll phosphate buffer of pH 6.0 (50 mM in D2O) under gentle heating and whirlmixing. Both samples were partially digested by addition of 1 ll of the brewing enzyme mixture Ultraflo XL (Novozymes, Bagsværd, Denmark). The b-limit dextrin (20 mg in 600 ll phosphate buffer of pH 6.0, 50 mM in D2O) was partially digested by 150 units of porcine pancreas a-amylase (Megazyme, Bray, Ireland) at 37 °C. Upon partial digestion, a series of assignment spectra was recorded to identify structural motifs in the resultant fragments as described (Petersen, Meier, & Duus, 2012).

All spectra were acquired and processed using Topspin, except for diffusion ordered spectra, which were processed using biexponential fitting of anomeric signals using the DOSY toolbox (Nilsson, 2009). Spectra were processed with extensive zero filling and mild resolution enhancement in the direct dimension by applying a time domain Lorentzian-Gaussian transformation with a line broadening of 1 Hz and a Gaussian broadening of 0.3. Molecular sizes of carbohydrates were estimated from DOSY derived self-diffusion coefficients according to the parameterisation for oligosaccharides and polysaccharides of Mayo and coworkers (Miller, Klyosov, Platt, & Mayo, 2009). The molecular weight estimate for the well-resolved and strong a,a-trehalose signal was used to validate the size estimate, without extra adjustments to account for solution viscosity or temperature. Mono-, di- and trisaccharides were identified by comparison with reference compounds and published assignments (Vinogradov & Bock, 1998). Arabinoxylan cleavage site motifs were identified by published assignments (Gruppen et al., 1992; Hoffmann, Geijtenbeek, Kamerling, & Vliegenthart, 1992) that were validated with two-dimensional assignment spectra on Ultraflo XL degraded arabinoxylan in situ. Starch and beta glucan cleavage site motifs were identified by direct assignment in situ (Petersen et al., 2012; Petersen, Olsen, Beeren, Hindsgaul, & Meier, 2013). Chemical shifts were measured by reference to the glucose a-anomeric signal (d1H = 5.230 and d13C = 92.990 ppm at 37 °C).

2.3. NMR measurements

2.5. Chromatography

NMR spectra were recorded at 37 °C on an 800 MHz Bruker (Fällanden, Switzerland) DRX spectrometer equipped with a TCI

Hydrophilic interaction liquid chromatography (HILIC) was performed following fluorescence labelling of the lager beer sample

2.2. Chemicals

67

B.O. Petersen et al. / Food Chemistry 150 (2014) 65–72

with 2-aminobenzamide as previously described (Bojstrup, Petersen, Beeren, Hindsgaul, & Meier, 2013). Briefly, the labelled sample was injected into a Waters Acquity UPLC system equipped with a FLR detector (fluorescence detector, excitation wavelength of 320 nm and emission wavelength of 420 nm), binary solvent manager, sample manager, and column oven manager. Separation was performed using an Acquity UPLC BEH glycan 1.7 lm, 2.1  150 mm column with a VanGuard BEH glycan 1.7 lm, 2.1  5 mm precolumn at room temperature.

α(1-4)

“red. α”

α(1-6) lager

glc

asian lager pilsner porter

Mn

Mn ale wheat beer

3. Results 5.4

3.1. Starch-derived fragments Carbohydrate mixtures often comprise polydisperse sets of similar yet different molecules, where monomers that are structural isomers and stereoisomers are linked together in a variety of ways. Direct chromatographic or mass spectrometric analysis is complicated by the absence of intrinsic fluorophores, chromophores or readily ionisable groups. NMR spectroscopy has established itself as a common tool for mixture analysis due to its capacity to deliver quantifiable signals from one-dimensional spectra or – often more qualitative – high-resolution profiling from two-dimensional NMR spectra (Fan, 1996). NMR-based mixture analysis is further aided by the fact that NMR samples generally require minimal pretreatment and pose no demand on the structural features of the carbohydrates, such as free reducing ends, while resolving linkage types and isomers of identical mass. However, the similarity of proton environments in carbohydrates and the structural resemblance between different carbohydrates manifest themselves in a rather narrow NMR chemical shift range of pure compounds and mixtures thereof. The chemical similarity of carbohydrates thus results in rather comparable spin systems, so that two-dimensional NMR methods for spin system identification have limited benefits in carbohydrate mixture analysis. One-dimensional 800 MHz 1H NMR spectra of six different commercial beers representing different styles from the lager and ale categories are displayed in Fig. 1. The displayed spectral region primarily contained signals from maltooligosaccharides and limit dextrins containing a(1–4) and a(1–6) glycosidic linkages. Both maltooligosaccharides and limit dextrins derive from starch and their presence in beer reflects the incomplete conversion of starch to carbohydrates that are fermentable by yeast. Differences in the carbohydrate profile of a-glucans in different beer styles were evident with an increased maltooligosaccharide (5.39 and 5.24 ppm) level in the ale sample and increased glucose (5.23 ppm) levels in the asian lager sample (Fig. 1). The differences in starch-derived carbohydrates overall reflected different production conditions that include differences in starch sources and malting conditions, usage of adjuncts, brewing enzymes, mashing protocols and yeast strains or other processing resulting in different degrees of starch breakdown. Clear spectral differences in the beer samples were seen in the region of a(1–6) glycosidic linkages around a 1H NMR shift of 4.97 ppm. The presence of substantial amounts of branched limit dextrins in beer reflected incomplete starch debranching by free limit dextrinase in germinating barley. Different limit dextrin branch point structures result in sufficient chemical shift differences to resolve and assign the different branch point types (Fig. 2) in COSY spectra of branched limit dextrins. Lager type samples could be distinguished from ale type samples due to the presence of significant amounts of single terminal a(1–6) linked glucopyranosyl residues at the non-reducing end of limit dextrins resulting from extensive dextrin degradation by a-amylases. In general, the presence of limit dextrin signals and the relative

5.3

δ1H/ppm

5.1

5.0

Fig. 1. Anomeric region of 1H NMR spectra of six commercial beer samples representing different styles. The displayed spectral region mainly encompasses anomeric signals of a-glucopyranosyl residues. Signals corresponding to a(1–4), a(1–6), reducing end signals ("red. a"), glucose (glc) and maltooligosaccharides (Mn) are marked. Spectral differences are significant owing to different production sites and conditions.

prevalence of different limit dextrin branch type structures in complex foodstuffs afford direct conclusions about limiting amylolytic enzyme activities during production. 3.2. Homonuclear broadband decoupling in carbohydrate mixtures Limit dextrins deriving from a-amylase degradation of the amylopectin component of starch appear to be implicated in organoleptic properties of malt based beverages and thus may provide a relevant spectral probe for quality control (Bringhurst, Broadhead, Brosnan, Pearson, & Walker, 2001). Accordingly, direct spectral quantification of branch types was desirable. The limit dextrin signals were, however, poorly resolved in one-dimensional 800 MHz 1 H spectra for samples that do not contain a dominant limit dextrin type (Fig. 2). Lack of resolution is a general problem in spectroscopic mixture analysis and signal quantification. At the heart of this problem lies the homonuclear scalar coupling between protons which splits signals into multiplets and thus yields spectra with a number of 1H signals that is a multiple of the number of 1H sites (Aguilar et al., 2010). An attractive approach for the improvement of carbohydrate spectra is the application of homonuclear decoupling schemes, which allow the collapse of multiplet signals into single signals by methods that result in loss of sensitivity (Aguilar et al., 2010). As sensitivity usually is less of an issue than signal overlap in the 1H NMR-based analysis of complex mixtures, such homonuclear decoupling schemes may often prove beneficial. The benefit will be particularly noteworthy for mixtures of compounds with poor signal dispersion such as carbohydrates. Application of homonuclear decoupling schemes is challenging in the strong coupling regime, a problem that loses significance at increasingly higher magnetic fields. Accordingly, homonuclear broadband decoupling, which previously had proven beneficial in the resolution enhancement of NMR spectra of pure compounds and simple organic mixtures, was applied directly to the beer samples. The broadband homonuclear decoupled one-dimensional 1H NMR spectrum of the a(1–6) anomeric region in lager beer is displayed in Fig. 3. The spectrum indicated that the sensitivity penalty of broadband homodecoupled 1H NMR experiments is not prohibitive for application to complex samples and did not require pre-concentration of the samples. Overlapping multiplets could be resolved by homonuclear decoupling to yield quantifiable nonoverlapping signals (Fig. 3) of different limit dextrin branch types. Signal dispersion of the a(1–6) anomeric protons was greatly superior to that in a conventional 1H NMR spectrum recorded with identical shimming and sampling of the free induction decay and with

68

B.O. Petersen et al. / Food Chemistry 150 (2014) 65–72

B

A

lager

3.45

ale

pilsner

lager

3.50 3.55

asian lager 3.60 3.65

pilsner

porter

asian lager

3.45

wheat beer

3.50

porter

3.55

ale

3.60

wheat beer 3.65

5.01

δ1H/ppm

4.96

4.91

5.00

4.98

4.96

4.94

5.00

4.98

4.96

4.94

5.00

4.98

4.96

4.94

δ1H/ppm

Fig. 2. The a(1–6) spectral region containing signals from starch-derived limit dextrins of varying complexity in 1H NMR (A) and DQF-COSY (B) spectra. Ale, porter and wheat beers contain more complex branch points due to incomplete starch amylolysis. Lager type beers contain a higher fraction of limit dextrins that are cleaved adjacent to branch points due to more complete amylolysis.

processing using the identical window function (Fig. 3). In addition, the comparison of conventional and proton-decoupled 1H NMR spectra in complex samples directly clarifies the multiplet structure of 1H NMR signals and thus facilitates spectral assignments. Future development of increasingly sensitive two-dimensional 1H–1H NMR spectra employing broadband homodecoupling (Aguilar, Colbourne, Cassani, Nilsson, & Morris, 2012) clearly would be of great benefit in advancing the NMR-based resolution and assignment of mixture components. 3.3. Self diffusion of polysaccharide-fragments in beer Determining the molecular size of gycans in mixtures by their NMR relaxation properties is challenging, as the absence of a unique molecular reference frame obstructs the separation of local and global motion in NMR relaxation data analysis (Prompers & Bruschweiler, 2002). Instead, high-resolution NMR spectral information has been increasingly paired with a chromatographic dimension in hyphenated NMR methods such as LC-NMR (Sidelmann et al., 1996) or by adding a size dimension in form of the

5.00

δ1H/ppm

4.95

4.90

Fig. 3. Conventional (top) and pure shift (bottom) 1H spectra of lager beer. The limit dextrin signals are clearly resolved by homonuclear decoupling. Resolving signals in mixture analysis is paramount for improved quantification from one-dimensional spectra. The relative amount of indicated branch types can be integrated to 1:2.9:2.6 in this manner despite a highly overlapped conventional 1H NMR spectrum.

translational self-diffusion coefficient to NMR spectra (Groves et al., 2004). Using high-resolution diffusion ordered spectroscopy (DOSY) with biexponential fitting (Nilsson, Connell, Davis, & Morris, 2006) and the improved chemical shift dispersion at high magnetic fields allowed the characterisation of different carbohydrates in complex mixture based on their self-diffusion properties, even if the signals were not baseline separated. The diffusion ordered 1H spectrum of anomeric signals of lager beer carbohydrates is displayed in Fig. 4. Translational self-diffusion rates on the order of 4  10 10 m2/s for maltooligosaccharides signals (5.39 ppm) in lager beer indicated a weighted average degree of polymerisation for maltooligosaccharides on the order of 3–4, agreeing with predominating fractions of maltose, maltotriose and maltotetraose, and smaller amounts of longer oligosaccharides due to rather complete amylolysis of starch in the lager fermentations. The presence both of branched and linear glucans made the estimation of the degree of polymerisation (dp) based on signal areas of anomers and the reducing end alone problematic and highlighted the benefit of using a self-diffusion dimension in order to add an independent spectral dimension containing size information. Signals of a(1–4) anomeric protons in the vicinity of branches (5.35 ppm) indicated an average dp around 8. The signals of corresponding a(1–6) anomeric protons permitted a further resolution of the average sizes of oligosaccharides with different branch structure. The complex branch type carrying a(1–4) linked glucopyranosyl on both glucopyranosyl units of the branch occurs in limit dextrins with average dp  9, while limit dextrins carrying single internal a(1–6) linked glucopyranosyls showed an average dp  6 and limit dextrins carrying an a(1–6) linked glucopyranosyl at the non-reducing end showed dp  4, corresponding to a predominant fraction of 63-glucopyranosyl-maltotriose, underlining fairly complete amylolysis in the lager brew. By using one-dimensional diffusion filtering (pulsed-field gradient spin echo experiments) or CPMG spin-echo sequences, carbohydrate signals are separable into signals from smaller and larger oligosaccharides, the latter including signals from limit dextrins and other nondigestible oligosaccharides (Fig. 4 bottom). Particularly arabinoxylans have been implicated in filtration problems during the production of malt-based beverages. Consistent with their contribution to viscosity and filtration problems, self-diffusion

B.O. Petersen et al. / Food Chemistry 150 (2014) 65–72

A 60 Arabinoxylan

5

LD Mn Kes Koj

B 1

8

LD

3

approx DP

D/ 10−10 m2 s−1

1

Koj

Suc

3

α,α

Tre

5.4

δ1H/ppm

5.0

5.4

δ1H/ppm

5.0

H NMR

PFGSE CPMG

Fig. 4. 800 MHz 1H DOSY spectrum (top) of carbohydrates in the lager sample. Limit dextrins of Fig. 3 have apparent degrees of polymerisation (dp) of 6, 9 and 4 (signals at 4.95 ppm). Maltodextrins have an average dp of 3–4. Arabinose signals from arabinoxylan in the sample indicate average dp above 50 in agreement with previous notions that arabinoxylan often is limiting for beer filterability. Di- and trisaccharides of the cellobiose (Cell), kestose (Kes), sucrose (Suc), kojibiose (Koj) and trehalose (Tre) type are labelled. One-dimensional 1H NMR, diffusion-edited 1H NMR and CPMG spin-echo 1H NMR experiments are displayed for comparison.

data indicated high molecular weight specifically for arabinoxylans in final beer, with translational self-diffusion rates for arabinose moieties in arabinoxylans as low as 10 10 m2/s, corresponding to an apparent molecular weight on the order of 10 kDa. For validation, DOSY data were compared to a separation of the beer sample of Fig. 4 by hydrophilic interaction liquid chromatography (Fig. S1, supplemental data). This separation validates (i) the predominance of a tetrasaccharide different from maltotetraose, consistent with a significant fraction of 63-glucopyranosyl-maltotriose, (ii) the increasing predominance of oligosaccharides that are not maltooligosaccharides for carbohydrates larger than dp 4 and detectable at significant amounts up to approximately dp 9, consistent with DOSY data showing limit dextrin fragments with different branch type and average dp up to 9, (iii) considerable amounts of linear G2–G4, consistent with an approximate average dp of slightly below 4 for maltooligosaccharides and (iv) a considerable fraction of larger oligosaccharides, identified as arabinoxylans with an apparent molecular weight on the order of 10 kDa by DOSY. The strength of DOSY is that it combines chemical and size information in a single experiment, in contrast to chromatographic separations that are devoid of chemical information. DOSY thus provides unique information for the spectroscopic characterisation of main polymer fragments in beer, while individual components of smaller concentration require chemical characterisation with high-resolution data, specifically chemical shift measurement, in the indirect dimension. 3.4. COSY-based distinction of oligosaccharide fragments Usage of two-dimensional NMR spectra is a principal means of reducing peak overlap in complex samples (Fan, 1996). In particular, the 1H1–1H2 spectral region of double quantum filtered COSY spectra proved to provide sufficient signal dispersion to give a comprehensive fingerprint of carbohydrate composition in plant-

69

derived liquids (Fig. 5). Overall, more than 50 different spin systems can be assigned to different structural motifs in the 1 H1–1H2 spectral region of a DQF-COSY recorded on a 18.7 T magnet equipped with a TCI cryoprobe. Assignments were performed with homo- and heteronuclear two-dimensional experiments and exhaustive zero-filling in all dimension as well as mild resolution enhancement in the acquisition dimension. Besides starch fragments, isomers of carbohydrates encompassing kestose, allose, sucrose, maltulose, kojibiose, laminaribiose a,b-trehalose, a,b-trehalose, b,b-trehalose, cellobiose, gentiobiose, arabinose, ribose, galactose, xylose, xylobiose and nigerose type motifs were resolved, alongside fragments of the non-starch polysaccharides arabinoxylan and mixed linkage (1–3),(1–4)-b-glucans (henceforth referred to as b-glucans). Notably, all glucose disaccharides except sophorose were detectable in beer by the DQF-COSY of Fig. 5. Arabinoxylan and b-glucan are the main constituents of cell walls in the cereal endosperm and aleurone layer. Degradation of these cell-wall constituents allows the mobilisation of endosperm storage component that nourish the growing embryo in germinating cereals. In the processing of cereal extracts, arabinoxylan and b-glucan negatively affect rheological behaviour and extract filterability, so that extensive enzymatic degradation is desirable for improved processing. Hydrolysis of isolated cereal arabinoxylan and b-glucan by brewing enzymes yields signals that are in part detectable in beer (Fig. 5). Assignments of limit dextrin fragments, b-glucan and arabinoxylan spin systems were performed by in situ assignments of fragments derived by the enzymatic digestion of starch-derived a-glucans (Petersen et al., 2012) and barley b-glucan (Petersen et al., 2013). Published assignments of arabinoxylan fragments (Gruppen et al., 1992; Hoffmann et al., 1992) were verified by homo- and heteronuclear assignment spectra for the identification of cleavage sites in arabinoxylan. Mono-, di- and trisaccharides were assigned by comparison to reference compounds (Fig. S2, supplemental data). Unambiguous identifications are aided by the presence of several resolved 1H1–1H2 signals for most of these compounds (except symmetrical aa- and bb-Trehalose) due to resolved signals for a- and b-anomeric forms, together with 1 H1–1H2 signals from units other than the reducing end in di- and trisaccharides. In result, the likelihood of resolving non-overlapped 1 H1–1H2 reporter signals is increased. Assignments were validated by heteronuclear 1H–13C NMR for sufficiently concentrated carbohydrates (Fig. S3, supplemental data). As for the starch-derived oligosaccharides, fragment structures deriving from cell-wall polysaccharides should be indicative of enzymatic action during biotechnological processing. For b-glucan fragments, the anomeric 1H signals in b(1–3) glycosidic linkages (4.8–4.9 ppm) are structural reporter signals indicative of enzyme action on b-glucans. While b-glucan fragments are largely absent in the asian lager beer sample, varying degrees of b-glucan fragments were found in the other beer styles that were analysed. Varying degradation patterns were identifiable for b-glucans in different beer samples due to distinct endo- and exoglucanse activities in the different production processes. As an example, appreciable amounts of fragments with b(1–3) glucopyranosyl at the nonreducing end occured only for the lager-type sample (Fig. 6). Like b-glucan content and structure, arabinoxylan structure and content varied between the samples, with the lowest arabinoxylan content in the analysed bottom-fermented lager-style samples (Fig. 6). D-xylopyranosyl spin systems of arbinoxylan that are unsubstituted, 3-substituted or 2,3-substituted with a-L-arabinofuranosyl residues were resolved. Internal D-xylopyranosyl residues of these three substitution types were resolved from residues at the non-reducing end, thus permitting an assessment of arabinoxylan cleavage. As non-reducing D-xylopyranosyl termini are detected, the large molecular weight of arabinoxylan according to self-diffusion data seems to be a consequence of partially

70

B.O. Petersen et al. / Food Chemistry 150 (2014) 65–72 β

Gen GlcβXyl IMt2/3 R-Xylβ Gen

limit dextrin fragments arabinoxylan fragments beta glucan fragments

Nigβ

β,β

Tre Koj All Glcα-(1-1)-[Glcα-(1-2)-]-Frcβ Maltul α Glcα Xylose Suc Koj IMt3 Maltul Kes pyr Nigα,β Maltul Rib R-Xylα Tre Cellα Koj Malt Nigα α,α Tre

1H

3.4

Tre

Galβ Ara

3.6

Koj Lam Galα

3.8

Ribpyr

Araα

4.0 Ribfur

[ ]n

4.2 5.4

5.2

5.0

1H

4.8

5.4

5.2

5.0

ppm

4.8

Fig. 5. DQF-COSY spectrum of the lager sample and comparison to fragments obtained by enzymatic hydrolysis of starch (blue–green), b-glucan (red) and arabinoxylan (orange). Different structural motifs in starch and cell wall fragments are resolved. Filled symbols indicate the carbohydrate moieties from which the H1–H2 cross peaks derive. Vertical lines indicate a(1–6) bonds in a-glucan fragments, while diagonal lines indicate b(1–3) bonds in b-glucan fragments. In the case of arabinoxylan, circles indicate D-xylopyranosyl and diamonds indicate aL-arabinofuranosyl residues, with vertical and diagonal lines indicating a(1–2) and a(1–3) glycolytic bonds, respectively. Overall, more than 50 different carbohydrate moieties can be assigned. . Di- and trisaccharides of the cellobiose (Cell), Gentiobiose (Gen), isomaltotriose (IMt3), kestose (Kes), kojibiose (Koj), laminaribiose (Lam), maltose (Malt), maltulose (Maltul), nigerose (Nig), sucrose (Suc) and three different trehalose isomer (Tre) types are abbreviated for clarity. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

lager

pilsner

β

ale 3.2

β

Gen

3.4 3.6 porter

asian lager

wheat beer

3.2 3.4 3.6

4.9

4.8

4.7

4.6 4.5

4.9

4.8

4.7

4.6 4.5

4.9

4.8

4.7

4.6

δ1H/ppm

Fig. 6. Spectral region of DQF-COSY containing b-glucan and the D-xylopyranosyl signals as a fingerprint region of cell wall derived glycans. Only b-glucan signals are marked for clarity. As for starch-derived fragments, cell-wall fragments differ significantly in abundance between different beer samples. Specific cleavage products are distinguished, as glucans carrying terminal b(1–3) are only found in the lager sample, presumably due to lower b(1–3) glucanase activity in the production process of this sample.

retained crosslinking via arabinose moieties. Barley beers showed a major fraction of unsubstituted D-xylopyranosyl and slightly smaller fraction of 2,3-disubstituted D-xylopyranosyl moieties with only a small fraction of 3-substituted moieties. The fraction of 3substituted xylopyranosyl spin systems was higher in the analysed wheat beer sample. An increasing interest in the b-glucan and arabinoxylan content and fragment structures in food samples is expected due to the suggested health benefits of these fibres (Slavin, 2000). NMR spectroscopy can play a role in the detection and quantification of functional oligosaccharide fragments and their interactions with molecular targets. 4. Discussion 1 H NMR spectroscopy is a robust, quantitative and non-destructive method for food and nutrition analysis without derivatization and separation, but is rarely used for carbohydrate profiling due to the moderate chemical shift dispersion and spectral overlap with the water signal. In the current study, we showed that two-dimen-

sional 1H NMR employing 1H1–1H2 correlations or diffusion in the second dimension provides rich information for the targeted profiling of beer carbohydrates. We resolved and identified more than 50 carbohydrate spin systems in beer samples with minimal sample treatment by using 1H1–1H2 NMR cross peaks as reporters for carbohydrate profiling, identifying constituent components in six commercial samples of different beer styles. Carbohydrate composition with respect to small carbohydrates (predominantly monoand disaccharides), starch, b-glucan and arabinoxylan fragments was analysed. The composition varied significantly between the samples, including differences in soluble fibre content and fragment composition. Previous studies employing one-dimensional spectra and multivariate analysis had indicated that the carbohydrate spectral region contributes strongly to the distinction of beverage samples, thus calling for more detailed assignments of the carbohydrate fraction in plant derived foodstuffs (Duarte et al., 2002). It is expected that profiles of carbohydrate fragments reporting specifically on enzymatic degradation of starch and cell walls complement small molecule metabolites in the identification

B.O. Petersen et al. / Food Chemistry 150 (2014) 65–72

of production site and conditions, product style and product quality including organoleptic, nutritional and prebiotic value. Different degrees of amylolysis are evident from the varying abundance of limit dextrins that can be resolved in homonuclear decoupled one-dimensional spectra and two-dimensional DQFCOSY spectra. Even for spectra with severe signal overlap, the measurement of translational self-diffusion is feasible, thus adding an independent dimension with novel information content to carbohydrate spectra: resolving similar glycan fragments such as limit dextrins by their hydrodynamic properties. Sample derivatization and coelution problems in chromatography based analytical schemes are avoided altogether (Consonni, Cagliani, & Cogliati, 2012). Branched a-glucan fragment structures are of interest, as limit dextrins are increasingly recognised as functional compounds due to their prebiotic and organoleptic properties. Besides starch-derived fragments, various cell-wall fragments can be distinguished in situ that reflect the action of different cell wall degrading activities. In recent years, b-glucan and arabinoxylan fragments have received considerable attention as compounds with possible health benefits. Enzymatic action is a hallmark of biotechnological production processes. The prospect of deducing limiting enzyme activities from product composition, especially cleavage site motifs for polymer fragments, or by in situ monitoring of production steps under realistic conditions is appealing. The biotechnological use of exogenous enzymes for modulating production processes has grown to a multibillion-dollar industry. The methods and structural assignments detailed above permit conclusions to be drawn about enzyme action on polysaccharides in the production processes of cereal-derived products due to the noninvasive, label-free and high-resolution 1H NMR spectroscopic detection. Acknowledgments 800 MHz spectra were recorded at the instrument of the Danish National Instrument Center for NMR spectroscopy of Biological Macromolecules at the Carlsberg Laboratory. We gratefully acknowledge Sophie Beeren for numerous helpful discussions. This work was supported by the UK Engineering and Physical Sciences Research Council (Grant Number EP/I007989/1).

Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/ j.foodchem.2013.10.136. References Aguilar, J. A., Colbourne, A. A., Cassani, J., Nilsson, M., & Morris, G. A. (2012). Decoupling two-dimensional NMR spectroscopy in both dimensions: Pure shift NOESY and COSY. Angewandte Chemie (International ed. in English), 51(26), 6460–6463. Aguilar, J. A., Faulkner, S., Nilsson, M., & Morris, G. A. (2010). Pure shift 1H NMR: A resolution of the resolution problem? Angewandte Chemie (International ed. in English), 49(23), 3901–3903. Andersson, B., Porras, O., Hanson, L. A., Lagergard, T., & Svanborg-Eden, C. (1986). Inhibition of attachment of Streptococcus pneumoniae and Haemophilus influenzae by human milk and receptor oligosaccharides. Journal of Infectious Diseases, 153(2), 232–237. Blundell, C. D., Reed, M. A., Overduin, M., & Almond, A. (2006). NMR spectra of oligosaccharides at ultra-high field (900 MHz) have better resolution than expected due to favourable molecular tumbling. Carbohydrate Research, 341(12), 1985–1991. Bojstrup, M., Petersen, B. O., Beeren, S. R., Hindsgaul, O., & Meier, S. (2013). Fast and accurate quantitation of glucans in complex mixtures by optimized heteronuclear NMR spectroscopy. Analytical Chemistry, 85(18), 8802–8808. Botana, A., Aguilar, J. A., Nilsson, M., & Morris, G. A. (2011). J-modulation effects in DOSY experiments and their suppression: The Oneshot45 experiment. Journal of Magnetic Resonance, 208(2), 270–278.

71

Bringhurst, T. A., Broadhead, A. L., Brosnan, J. M., Pearson, S. Y., & Walker, J. W. (2001). The identification and behaviour of branched dextrins in the production of scotch whisky. Journal of the Institute of Brewing, 107(3), 137–149. Consonni, R., Cagliani, L. R., & Cogliati, C. (2012). NMR characterization of saccharides in Italian honeys of different floral sources. Journal of Agriculture and Food Chemistry, 60(18), 4526–4534. Cordain, L., Eaton, S. B., Sebastian, A., Mann, N., Lindeberg, S., Watkins, B. A., et al. (2005). Origins and evolution of the western diet: Health implications for the 21st century. American Journal of Clinical Nutrition, 81(2), 341–354. Cui, Q., Lewis, I. A., Hegeman, A. D., Anderson, M. E., Li, J., Schulte, C. F., et al. (2008). Metabolite identification via the Madison Metabolomics Consortium Database. Nat Biotech, 26(2), 162–164. Duarte, I., Barros, A., Belton, P. S., Righelato, R., Spraul, M., Humpfer, E., et al. (2002). High-resolution nuclear magnetic resonance spectroscopy and multivariate analysis for the characterization of beer. Journal of Agriculture and Food Chemistry, 50(9), 2475–2481. Fan, T. W. M. (1996). Metabolite profiling by one- and two-dimensional NMR analysis of complex mixtures. Progress in Nuclear Magnetic Resonance Spectroscopy, 28(2), 161–219. Groves, P., Rasmussen, M. O., Molero, M. D., Samain, E., Canada, F. J., Driguez, H., et al. (2004). Diffusion ordered spectroscopy as a complement to size exclusion chromatography in oligosaccharide analysis. Glycobiology, 14(5), 451–456. Gruppen, H., Hoffmann, R. A., Kormelink, F. J., Voragen, A. G., Kamerling, J. P., & Vliegenthart, J. F. (1992). Characterisation by 1H NMR spectroscopy of enzymically derived oligosaccharides from alkali-extractable wheat-flour arabinoxylan. Carbohydrate Research, 233, 45–64. Hoffmann, R. A., Geijtenbeek, T., Kamerling, J. P., & Vliegenthart, J. F. (1992). 1 H–N.m.r. study of enzymically generated wheat-endosperm arabinoxylan oligosaccharides: Structures of hepta- to tetradeca-saccharides containing two or three branched xylose residues. Carbohydrate Research, 223, 19–44. Jewison, T., Knox, C., Neveu, V., Djoumbou, Y., Guo, A. C., & Lee, J. (2012). YMDB: The yeast metabolome database. Nucleic Acids Research, 40(D1), D815–D820. Johnson, C. S. Jr, (1999). Diffusion ordered nuclear magnetic resonance spectroscopy: Principles and applications. Progress in Nuclear Magnetic Resonance Spectroscopy, 34(3–4), 203–256. Kaput, J., & Rodriguez, R. L. (2004). Nutritional genomics: The next frontier in the postgenomic era. Physiological Genomics, 16(2), 166–177. Kovacs, H., Moskau, D., & Spraul, M. (2005). Cryogenically cooled probes—A leap in NMR technology. Progress in Nuclear Magnetic Resonance Spectroscopy, 46(2–3), 131–155. Kunz, C., Rudloff, S., Baier, W., Klein, N., & Strobel, S. (2000). Oligosaccharides in human milk: Structural, functional and metabolic aspects. Annual Review of Nutrition, 20, 699–722. Maitin, V., & Rastall, R. (2005). Enzymatic synthesis of oligosaccharides. In K. Shettym, G. Paliyath, A. Pometto, & R. E. Levin (Eds.). Food Biotechnology (2nd ed.). CRC Press. Markley, J. L., Ulrich, E. L., Berman, H. M., Henrick, K., Nakamura, H., & Akutsu, H. (2008). BioMagResBank (BMRB) as a partner in the worldwide protein data bank (wwPDB): New policies affecting biomolecular NMR depositions. Journal of Biomolecular NMR, 40(3), 153–155. Miller, M. C., Klyosov, A., Platt, D., & Mayo, K. H. (2009). Using pulse field gradient NMR diffusion measurements to define molecular size distributions in glycan preparations. Carbohydrate Research, 344(10), 1205–1212. Nilsson, M. (2009). The DOSY toolbox: A new tool for processing PFG NMR diffusion data. Journal of Magnetic Resonance, 200(2), 296–302. Nilsson, M., & Morris, G. A. (2007). Pure shift proton DOSY: Diffusion-ordered 1H spectra without multiplet structure. Chemical Communications (Cambridge) (9), 933–935. Nilsson, M., Connell, M. A., Davis, A. L., & Morris, G. A. (2006). Biexponential fitting of diffusion-ordered NMR data: Practicalities and limitations. Analytical Chemistry, 78(9), 3040–3045. Petersen, B. O., Meier, S., & Duus, J. O. (2012). NMR assignment of structural motifs in intact beta-limit dextrin and its alpha-amylase degradation products in situ. Carbohydrate Research, 359, 76–80. Petersen, B. O., Olsen, O., Beeren, S. R., Hindsgaul, O., & Meier, S. (2013). Monitoring pathways of b-glucan degradation by enzyme mixtures in situ. Carbohydrate Research, 368, 47–51. Prompers, J. J., & Bruschweiler, R. (2002). General framework for studying the dynamics of folded and nonfolded proteins by NMR relaxation spectroscopy and MD simulation. Journal of the American Chemical Society, 124(16), 4522–4534. Sidelmann, U. G., Lenz, E. M., Spraul, M., Hofmann, M., Troke, J., Sanderson, P. N., et al. (1996). 750 MHz HPLC–NMR spectroscopic studies on the separation and characterization of the positional isomers of the glucuronides of 6,11dihydro-11-oxodibenz[b, e]oxepin-2-acetic acid. Analytical Chemistry, 68(1), 106–110. Slavin, J. L. (2000). Mechanisms for the impact of whole grain foods on cancer risk. Journal of the American College of Nutrition, 19(Suppl. 3), 300S–307S. Stejskal, E. O., & Tanner, J. E. (1965). Spin Diffusion Measurements: Spin Echoes in the presence of a time-dependent field gradient. The Journal of Chemical Physics, 42(1), 288–292. Van Loo, J., Cummings, J., Delzenne, N., Englyst, H., Franck, A., Hopkins, M., et al. (1999). Functional food properties of non-digestible oligosaccharides: A consensus report from the ENDO project (DGXII AIRII-CT94-1095). British Journal of Nutrition, 81(2), 121–132.

72

B.O. Petersen et al. / Food Chemistry 150 (2014) 65–72

Vinogradov, E., & Bock, K. (1998). Structural determination of some new oligosaccharides and analysis of the branching pattern of isomaltooligosaccharides from beer. Carbohydrate Research, 309(1), 57–64. Wishart, D. S. (2008). Metabolomics: Applications to food science and nutrition research. Trends in Food Science and Technology, 19(9), 482–493.

Wishart, D. S., Tzur, D., Knox, C., Eisner, R., Guo, A. C., Young, N., et al. (2007). HMDB: The human metabolome database. Nucleic Acids Research, 35(Suppl. 1), D521–D526. Zangger, K., & Sterk, H. (1997). Homonuclear broadband-decoupled NMR spectra. Journal of Magnetic Resonance, 124(2), 486–489.