lable at ScienceDirect

Carbohydrate Research 408 (2015) 134e141

Contents lists avai

Carbohydrate Research

journal homepage: www.elsevier .com/locate/carres

Quantification of sugars in breakfast cereals using capillaryelectrophoresis

Michelle R. Toutounji a, b, c, Matthew P. Van Leeuwen a, c, James D. Oliver a,Ashok K. Shrestha b, Patrice Castignolles a, *, Marianne Gaborieau a, c

a University of Western Sydney (UWS), Australian Centre for Research on Separation Science (ACROSS), School of Science and Health, Parramatta Campus,Locked Bag 1797, Penrith, NSW 2751, Australiab University of Western Sydney (UWS), School of Science and Health, Hawkesbury Campus, Locked Bag 1797, Penrith, NSW 2751, Australiac University of Western Sydney (UWS), Molecular Medicine Research Group, School of Science and Health, Parramatta Campus, Locked Bag 1797, Penrith,NSW 2751, Australia

a r t i c l e i n f o

Article history:Received 11 December 2014Received in revised form16 February 2015Accepted 9 March 2015Available online 17 March 2015

Keywords:Breakfast cerealsMonosaccharideDisaccharideCapillary electrophoresis (CE)Fehling method3,5-Dinitrosalicylic acid (DNS) assay

* Corresponding author. Tel.: þ61 2 9685 9970; faxE-mail address: p.castignolles@uws.edu.au (P. Cas

http://dx.doi.org/10.1016/j.carres.2015.03.0080008-6215/© 2015 Published by Elsevier Ltd.

a b s t r a c t

About 80% of the Australian population consumes breakfast cereal (BC) at least five days a week. Withhigh prevalence rates of obesity and other diet-related diseases, improved methods for monitoring sugarlevels in breakfast cereals would be useful in nutrition research. The heterogeneity of the complex matrixof BCs can make carbohydrate analysis challenging or necessitate tedious sample preparation leading topotential sugar loss or starch degradation into sugars. A recently established, simple and robust freesolution capillary electrophoresis (CE) method was used in a new application to 13 BCs (in Australia) andcompared with several established methods for quantification of carbohydrates. Carbohydrates identifiedin BCs by CE included sucrose, maltose, glucose and fructose. The CE method is simple requiring nosample preparation or derivatization and carbohydrates are detected by direct UV detection. CE wasshown to be a more robust and accurate method for measuring carbohydrates than Fehling method, DNS(3,5-dinitrosalicylic acid) assay and HPLC (high performance liquid chromatography).

© 2015 Published by Elsevier Ltd.

1. Introduction

About 80% of the Australian adult population consume breakfastcereal, either cooked or ready-to-eat, at least five days a week.1

Breakfast cereals, like most food products, contain a variety ofcarbohydrates as well as lipids, proteins and minerals. The het-erogeneity of this complex matrix can make sugar analysis inbreakfast cereals challenging.2,3

For sugars, total content is all that is required for the nutritioninformation panel (NIP), a mandatory labeling requirement of allprocessed foods in Australia.3 Methods for the measurement ofsugar in foodstuffs were often developed before carbohydratechemistry was established.4 Earlier quantitative chemical analyt-ical assays often relied upon the reducing properties of aldehydeor keto group found in monosaccharides and short-chain oligo-saccharides. In alkaline solutions, at elevated temperatures, these

: þ61 2 9685 9915.tignolles).

reducing sugars tautomerize to enediol forms, which are thenreadily oxidized by oxygen and oxidizing agents (such as metallicsalts). An estimate of glucose content was based on the colori-metric measurement of the oxide or the free metal formed. Theempirical nature of this reaction allowed analysts to developmethods such as the Fehling method and the 3,5-dinitrosalicylic(DNS) assay, which can produce reproducible and accurate re-sults for samples with simple matrices. Such methods are alsoinexpensive, technically easy to perform and highly applicable toroutine quantification. However, a strict control of experimentalconditions (rate of heating, alkalinity and strength of the reagent)in a non-automated setting is necessary to obtain repeatable andreproducible results.5

Due to their specificity and ease of operation, enzymatic assaysare the preferred reducing sugar method over Fehling method andDNS assay. Glucose and sucrose content has been determined in 79dry, North American BCs using the glucose-oxidase peroxidase(GOD-POD) method.6,7 The sugar content of these samples waspreviously assayed by the colorimetric condensation reaction withanthrone, which gave unsatisfactory reproducibility.

M.R. Toutounji et al. / Carbohydrate Research 408 (2015) 134e141 135

Most foods, including BCs, contain a mixture of sugars ratherthan a single type of sugar. Therefore, the methods previouslydiscussed are innately flawed by being either glucose-specific (e.g.,GOD-POD) or unable to distinguish between different reducingsugars as is the case for the Fehling method and the DNS assay.8 Inreducing-sugar assays the quantity of product formed andmeasured is not exactly equivalent to sugar content, and differentsugars yield different color intensities; this shows that the chem-istry involved in the assay is considerably more complicated than itappears.9 For certain foods in which the composition of the sugarsis known and the requirement of the analysis is routine, e.g. qualitycontrol, an estimate of total sugar values expressed as invert sugaror glucose may be sufficient. However, most BCs have sucroseadded during manufacture and thus total sugar determination re-quires a preliminary hydrolysis of non-starch polysaccharides (byacid or enzyme), which may cause sample loss or overestimation ofreducing sugars. In addition, mineral ions have been reported tointerfere with some reducing-sugar assays,10,11 a problem for mostAustralian BCs, which have been fortified.

In the area of nutrition research, the intrinsic accuracy of thequantities of the different carbohydrates present in a diet is oftenrequired for correlation with their metabolic behavior.4 Separationis used for this purpose. Separation methods have allowed forgreater accuracy of sugar analysis in foods. Individual sugarsmeasurements can be summed to calculate the ‘total sugar content’for the NIP. Gas chromatography (GC) is a popular method forcarbohydrate analysis and is very sensitive. It is the only chroma-tography method published in the peer-reviewed literature so farfor sugar quantification of BCs.12e15 The sample preparation re-quires multiple steps: grinding to pass through a 30-mesh(0.59 mm) screen, drying under vacuum, defatting with n-hex-ane, extraction with water for some of the sample and withaqueous methanol for the rest, centrifugation. In order to make thecarbohydrate volatile a multistep derivatization was then needed:concentration under nitrogen flow and then drying under vacuum,reaction with pyridine, hexamethylsilazane in presence of tri-fluoroacetic acid, followed by another centrifugation. The samplepreparation for GC is thus time consuming, laborious and has asignificant probability of sample loss.3 High performance liquidchromatography (HPLC) is the other established analytical methodfor measuring individual sugars in many foods. A number of col-umns have been tested for normal phase HPLC of carbohydrates, forexample ion-exchange columns for BCs,16,17 but they all have theirown disadvantages including co-elution, tedious sample prepara-tion and intolerance to salt or acid leading to short column life.18,19

For starchy-food sample matrices, such interfering substances notonly disrupt the analysis, they can damage the column leading to ahigh running cost. High performance anion exchange chromatog-raphy (HPAEC) is solving a number of these issues and is rapidlydeveloping for food analysis3 because of its high sensitivity withoutany required derivatization and its speed in carbohydrates sepa-ration.20,21 Released sugars from non-starch polysaccharides in arange of raw and processed foods (including BCs) were determinedby HPAEC.13

Capillary electrophoresis (CE) is becoming increasingly popularfor carbohydrate analysis.22 The most useful and simple separationmechanism of CE is free solution (devoid of a gel or polymernetwork medium). Free solution CE (FSCE), also known as capillaryzone electrophoresis (CZE), involves the flushing of an electro-phoresis buffer through a narrow-bore capillary prior to sampleinjection, application of voltage and separation.23 Like HPLC, FSCE isa fast and repeatable analytical tool for the qualitative and quan-titative analysis of carbohydrates in food and beveragesamples,24e27 including in foodomics.28 FSCE sensitivity can begreatly increased using on-line pre-concentration methods.29 One

FSCE method has been applied to a BC sample (flaked cereals) bycoating the capillary with cetrimonium bromide (CTAB).24 Thisdynamic coating is non-selective for sugars whereby interactionwith lipids and other components can become a problem forquantification in complex matrices. Until recently, indirect UVdetection of underivatized carbohydrates was considered superiorto direct UV detection; detection at low wavelengths after boratecomplexation generated poor sensitivity. However, direct UVdetection at 270 nm in high alkaline conditions has been discov-ered for the FSCE of sucrose, glucose and fructose in beveragesamples30 and the detection has been shown to be specific to car-bohydrates31,32 The method is considered robust by definition: ‘amethod that can be applied to analytes in a wide variety ofmatrices’33 and despite the complex matrix, no filtration is neces-sary during sample preparation.34 FSCE with direct UV detectionhas been applied to acid hydrolyzed plant fiber, fermentation,beverage, pharmaceutical and forensic samples.18,19,35e37 The CEmethod was shown to be cost effective, robust and repeatable.

The aim of this research was to analyze 13 Australian breakfastcereals using FSCE with direct UV detection18 and to comparequantitative results with that from traditional chemical analyticalmethods: Fehling method and DNS assay. Available data from theFood Standards Australia New Zealand (FSANZ) NUTTAB database38

of individual sugar quantities for relevant Australian breakfast ce-reals was also included for a comparison to our findings.

2. Results and discussion

2.1. Detection of individual sugars in breakfast cereals usingcapillary electrophoresis (CE)

The ground breakfast cereals (BC) were simply suspended inwater and injected in CE. The sugars in breakfast cereals wereseparated (Fig. 1) by CE. Identification of sugars was validated bycomparison of the electrophoretic mobility of observed peaks withthat of a standard sugar solution and previous literature (Table 1). Adouble correction was used to precisely determine the electro-phoretic mobility of each sugar peak. The first correction was usinga neutral species (DMSO) as an EOF marker (see Equation S-2). Thesecond correction involved a homothetic transformation with anelectrophoretic mobility marker (Equation S-3).30 Table 1 demon-strates how the mobility value of a sugar, much like elution time inHPLC, is useful in peak identification. Sucrose was detected in allBCs, while lactose, maltose, glucose and fructose were detected insome. Other components are not detected even though the matrixis complex: proteins, lipids are also present but the direct UVdetection has been shown to be specific to carbohydrates since it isdue to a photo-oxidation reaction taking place at the detectionwindow.32 To identify the carbohydrates, electrophoretic mobilityis used, and not migration time, since the former has a higherrepeatability than the latter (Table 1). Preliminary sugar identifi-cationwas confirmed by spiking BC samples (see Figs. S-2 and S-3).The samples, ‘Corn Flakes’, ‘Froot Loops’ and ‘Weet-Bix Multigrain’were selected as they contained the greatest number of peaksamong the seven BCs used in the first set of experiments.

Repeatability within the standards was sufficient with relativestandard deviation (RSD) values of no more than 1.3%, providing areliable set of values on which to base sugar identification. In theanalysis of BC samples, higher RSD values were observed inMRT (series1) ‘Nutri-Grain’ and MVL ‘Weet-Bix’, yielding mea-surement errors of 5.6% and 2.4%, respectively (see also Fig. S-1).Apart from these isolated cases, the repeatability of BC sampleanalysis was good with RSD �1.5%.

Reproducibility of the electrophoretic mobility values has beeninvestigated by comparing the results obtained by two different

Fig. 1. Migration time electropherogram (A) and mobility electropherogram (B) of‘Weet-Bix Multigrain’. From the left to the right, peaks correspond to sucrose, (un-known), maltose, glucose and fructose.

M.R. Toutounji et al. / Carbohydrate Research 408 (2015) 134e141136

operators, MVL and MRT, or by the same operator with differentsample preparation (differing only in dilution), MRT (series1) andMRT (series2), or in the literature. Comparison of analyzed samplesshows a reasonable level of reproducibility between values re-ported in Ref. 30, and those obtained by operator MVL and MRT(series1), showing a maximum variance between mobilities ofapprox. 5% (sucrose), with MRT reporting consistently higher mo-bilities. Identical operator with diluted sample (MRT (series2))showed reduced repeatability, yielding mobility value variance of5e10% compared to MRT (series1), while comparison with MVLshowed variance of 1e5%, with two values yielding a 10% variance(glucose and fructose in ‘Weet-Bix’). Results for diluted ‘Coco-Pops’lacked the identification of both glucose and fructose, present inprevious experiments, indicating dilution to reduce impacts ofoverloading is negatively affecting the sensitivity for low concen-tration sugar identification. Results reported by Oliver et al.18 pre-sent significantly higher electrophoretic mobilities for both glucoseand xylose. This illustrates the importance of a standard solution toestablish themobility values of sugars in each session to account forvariations between operators, equipment and injection series.

The results obtained in this work by both operators MVL andMRT showed a sufficient level of reproducibility between operatorsto yield identical sugar identifications in all cases. Same operator,

with dilution showed a reduced reproducibility, yielding slightlyhigher variance in mobility, though still indicates a sufficient levelfor the identification of sugars. CE is thus a viable method for theidentification of sugars in breakfast cereals.

2.2. Quantification of individual sugars in breakfast cereals usingcapillary electrophoresis (CE)

The calibration curve for each sugar was prepared with thesequential analyses of six sugar mixtures injected in triplicate. Thelinearity and repeatability were determined for 5 sugars, withxylose (0.5 g L�1) used as the internal standard. Sufficient linearitywas achieved for all tested sugars with correlation coefficient (R2)greater than 0.99 (Table 2) and reasonable standard error on the yestimates (see Figs. S-4 to S-6 in Supplementary data), as achievedin the literature applying this CE method to different matrices.18,39

The calibration for disaccharides, maltose, lactose and sucrose, hadslightly better linearity than that for monosaccharides glucose andfructose. The sugar concentrations determined by CE also showgood repeatability (see error bars on Fig. 2) of the peak area(normalized with respect to the peak area of the internal standard,each peak area being also divided with the corresponding migra-tion time) consistent with the literature (see Table S-3).18,30 The useof an electro-osmotic flow marker and the addition of an internalstandard are recommended for optimal repeatability of the peakarea. The high pH of the NP200 buffer made it prone to carbon-ation40 and it is thus recommended to use buffer within 13 h (or19 h) of its preparation to be within 10% (or 15%) of initial currentmeasurement (Equation S-5 to S-8, Table S-4 and Fig. S-7 inSupplementary data). Table 2 lists the relative sensitivity of thedetection of different sugars along through the limit of detection(LOD) and limit of quantification (LOQ). LOD and LOQ were calcu-lated from the signal-to-noise ratio (SNR) obtained with 32 Karatsoftware. The sensitivity of the direct detection in this study, withLOD values between 2.4 and 30 mg L�1, was comparable to studiesthat had used the same CE method on different types of analy-tes,31,41 (see Table S-1).

The method was applied to the quantification of sugars inseveral commercial breakfast cereals. Since fewer than six sugarshave been identified in BCs previously, a short capillary, total length61.8 cm, was employed for all experiments to decrease analysistime to 40 min (including xylose internal standard). Fig. 2 presentsthe sugar concentration results for ten BCs determined by CE. EachBC contained sucrose at a higher concentration than any othersugar detected (lactose, maltose, glucose and fructose). This is likelydue to the amount of sugar added during manufacture of theproduct. BCs with high sucrose concentrations (measured above15 g/100 g), including ‘Oats Apple & Blueberry Bake’, Nutrigrain’,‘Coco Pops’, listed sugar as the second highest ingredient after thecereal component on their packaging.

Nutritionally insignificant concentrations of maltose, glucoseand fructose were detected in 9 of the 11 BCs analyzed by CE. Barleymalt extract is listed as an ingredient on BCs ‘Sustain’, ‘Corn Flakes’and ‘Sultana Bran’ thus the low concentration values (1.8, 0.9 and0.6 g/100 g, respectively) for maltose could be expected. Similarly,trace levels of lactose detected in ‘Oats Apple & Blueberry Bake’(0.5 g/100 g) and ‘Oats Banana Bake’ (0.3 g/100 g) are in agreementwith the addition of the milk powder ingredient in these BCs.

Problems were reported with the sugar quantification in GC ofNorth American BCs with sampling and/or measuring aliquots ofindividual cereals.12 The variation of most samples, however, wasnot greater than the standards. They state that RSD was not greaterthan 2% for glucose and sucrose and not greater than 5% for lactoseand maltose. The average RSD from CE reported in this work isapproximately 7% for lactose, glucose and fructose and about 13%

Table 1Reproducibility of the separation of sugars in five breakfast cereal samples with CE. Operator MVL (n¼2) and Operator MRT (n¼3)

Breakfast cereal Operator or publication Average mep (10�8 m2 V1 s�1) (RSD in %)

Sucrose Lactose Maltose Glucose Fructose Xylose

Standard sugar solution Ref. 30 b �0.772 (ua) d d �1.176 (ua) d �1.365 (ua)Ref. 18 c d d d �1.518 (0.45) d �1.754 (0.40)MVLd �0.720 (1.08) d �1.116 (0.49) �1.181 (0.30) �1.265 (0.30) �1.395 (1.07)MRTe �0.801 (1.29) �1.003 (0.77) �1.172 (0.40) �1.224 (0.74) �1.285 (0.40) �1.419 (ua)

‘Coco Pops’ MVL �0.720 (ua) d d �1.152 (0.77) 1.237 (0.83) d

MRT (series 1) �0.810 (0.109) d d �1.223 (0.006) �1.289 (0.029) �1.419 (ua)MRT (series 2) �0.733 (0.470 d d d d �1.329 (ua)

‘Nutri-Grain’ MVL �0.720 (ua) d d �1.188 (0.77) �1.283 (0.79) d

MRT (series 1) �0.788 (5.59) d d �1.212 (1.18) �1.276 (0.792) �1.419 (ua)MRT (series 2) �0.730 (0.49) d d �1.118 (0.93) �1.201 (0.19) �1.333 (ua)

‘Sustain’ MVL �0.720 (ua) d �1.124 (1.06) �1.195 (1.15) �1.288 (1.46) d

MRT (series 1) �0.805 (0.070) d �1.172 (0.051) �1.224 (0.038) �1.285 (0.047) �1.419 (ua)‘Weet-Bix’ MVL �0.720 (ua) d d �1.240 (2.20) 1.348 (2.39)

MRT (series 1) �0.798 (0.119) d d �1.221 (0.062) �1.285 (0.026) �1.419 (ua)MRT (series 2) �6.834 (1.48) d d �1.090 (0.21) �1.161 (0.11) 1.272 (ua)

‘Oats: Apple and Blueberry Bake’ MRT (series 1) �0.808 (1.061) �1.025 (1.230) d d d �1.419 (ua)

a u stands for unavailable.b Mobility correction using methanol as an EOF marker.c Mobility double correction using DMSO as an EOF marker and lactose as an electrophoretic mobility marker.d Mobility double correction using DMSO as an EOF marker and sucrose as an electrophoretic mobility marker.e Mobility double correction using DMSO as an EOF marker and xylose as an electrophoretic mobility marker.

Table 2Calibration of response at 266 nm (y) as a function of sugar concentration (x) with its correlation coefficient (R2), for the sugars in standard (capillary of 61.8 cm total length).Xylose (0.5 g L�1) was used as the internal standard

Sugar LOD (mg L�1) LOQ (mg L�1) Linear equation R2 Concentration range (mg L�1)

Sucrose 5.88 21.6 y¼0.4517x�0.0395 0.999 50e1500Lactose 2.38 19.8 y¼0.3292xþ0.0208 0.998 20e500Maltose 20.7 41.7 y¼0.4104xþ0.0259 0.995 20e500Glucose 30.0 42.6 y¼0.2327xþ0.0338 0.992 20e500Fructose 15.9 44.1 y¼0.4263xþ0.0324 0.992 20e500

Fig. 2. Individual sugar quantification of 11 BCs by CE (n¼3 or 5 for all BCs).

M.R. Toutounji et al. / Carbohydrate Research 408 (2015) 134e141 137

Fig. 4. Total sugar content of 7 BCs by CE, Fehling method, available HPLC data38 andpackaging label information (NIP). No HPLC data was available for ‘All Bran FibreToppers’ and ‘Oats Apple and Blueberry Bake’.

M.R. Toutounji et al. / Carbohydrate Research 408 (2015) 134e141138

for sucrose andmaltose concentrations. The average overall error ofCE at ±9% is comparable to that of the reported GC method at ±7%.

2.3. Comparison of CE and HPLC for determination of individualsugars

The quantity of individual sugars in Australian BCs as measuredby HPLC is available on the Food Standards Australia New Zealand(FSANZ) NUTTAB 2010 database.38 This data was compared with CEresults from this work (see Figs. 3e5). Among the eight BCsincluded in this comparison, both methods determined ‘OatsTraditional’, ‘Weet-Bix’ and ‘Corn Flakes’ to have the lowest sucroseconcentrations. For these cereals, the HPLC method did not detectany sugar in Oats Traditional whereas CEmeasured 0.2% sucrose. CEalso detected approximately 36% more sucrose in ‘Weet-Bix’ and48% more in ‘Corn Flakes’ compared to HPLC (see Fig. 3A). The su-crose content of ‘Rice Bubbles’ as measured by CE was not signifi-cantly different from that from HPLC data. For the two BCs with thehighest sucrose concentration (‘Nutri-Grain’ and ‘Coco Pops’), lesssucrose was measured by CE compared to HPLC.38 Some of thesevariations may be due to changes in the recipes or even batch tobatch variations. The HPLC data from individual quantification ofsugars in BCs are available online to the public. However, the exactmethodology is unpublished and unreported. It is very likely that

Fig. 3. Quantification of sucrose (A) and glucose and fructose (B) in BCs by CE (thiswork) compared to available HPLC data.38

some filtration and/or centrifugation is required to prepare samplesfor carbohydrate analysis by HPLC. Thus sample loss could haveoccurred during sample preparation and may have caused an un-derestimation of sugar content for ‘Weet-Bix’ and ‘Corn Flakes’. CEanalysis also yielded a significantly higher sucrose concentration of12.5 g/100 g for ‘Sustain’ compared to HPLC data38 at 0.2 g/100 g(see Fig. 3A). In addition, CE data for glucose and fructose con-centrations in ‘Sustain’ were significantly lower (0.9 and 1.1 g/100 g) compared to HPLC data (7.4 and 8.7 g/100 g), see Fig. 3B. Thismay be due to differing sample preparation, in a sample, which isevenmore heterogeneous than the other BCs due to the presence ofpieces of fruit. During initial sample preparation of BCs for allexperimental methods in this study, the fruit pieces and some otherlarge particulate ingredients were resisting grinding and did notpass through laboratory sieves. Thus BCs with added pieces of fruitmeasured in this study, such as ‘Sustain’ and ‘Sultana Bran’, areacknowledged as not being representative of thewhole sample. The

Fig. 5. Reducing sugar content of 7 BCs by CE, DNS assay and available HPLC data.38

M.R. Toutounji et al. / Carbohydrate Research 408 (2015) 134e141 139

free form of glucose and fructose is found naturally in plants,including many fruits and vegetables. Typical sugar composition ofAustralian sultanas, for example, is 38% fructose and 35% glucose bydryweight.42 Themajority of sultanas and other pieces of fruit wereselectively removed from the BC samples in this study. This likelycaused an underestimation of glucose and fructose concentration,as seen in ‘Sustain’ and ‘Sultana Bran’ (Fig. 3B). More extensivesample preparation of BCs that contained fruit led however to up to55% error on themeasured sugars in the case of GC.12 The extractionprocedures of the BCs including fruits for GC requires aqueousmethanol. However, this still led to larger error, especially formaltose quantification.12 As the total sugar measured by CE andHPLC for ‘Sustain’ is similar, 16.3 and 18.2 g/100 g, respectively (seeFig. 5), the significantly lower sucrose levels by HPLC (see Fig. 3A)may be attributed again to sample loss during preparation for HPLCor to a change in the BC composition due to different time of pur-chase. Lower level of individual sugars quantified by HPLCcompared to CE with direct UV detection has been observed forother samples with complex matrices, namely plant fiber18 andethanol fermentation.19

2.4. Estimation of total sugar by CE and Fehling (LaneeEynon)method

Total sugar content is a legal measurement requirement for foodlabeling in Australia and many other countries. Total sugar contentof 11 BCs was determined by the traditional Fehling (LaneeEynon)method and the high performance CE separation method, andcompared with the NIP on the BC packaging label as well as withavailable HPLC data.38 Comparison of methods is important todetermine the accuracy of sugar content in BCs, both to ensure labelinformation is correct and to highlight differences betweenmethodin cost and efficiency.

The 11 BCs presented in Fig. 4 can be grouped into three cat-egories for total sugar content: (high (>12.5 g/100 g), medium(5e12.5 g/100 g) and low (<5 g/100 g) sugar content. The highsugar content cereals as measured by CE, in increasing order, were‘Sultana Bran’, ‘Sustain’, ‘Nutri-Grain’, ‘Coco Pops’ and ‘Oats Appleand Blueberry Bake’. ‘Sultana Bran’ and ‘Sustain’ had lower levelsof total sugar measured by CE and Fehling method in this workcompared to the packaging and available HPLC data38 likely due toremoval of fruit during sample preparation (as mentioned inSection, 2.3). ‘Nutrigrain’ was consistently ranked amongst the topthree BCs for high sugar content among all methods compared.The total sugar content for ‘Coco Pops’ measured by CE had arelatively high RSD of 20% and was therefore not significantlydifferent from that measured with other methods in the compar-ison. ‘Oats Apple and Blueberry Bake’ was determined to have thehighest total sugar content of all BCs by CE, 15% more than labeledon the NIP.

The medium sugar content BCs as determined by CE, inincreasing order, were ‘Rice Bubbles’, ‘Corn Flakes’, ‘Oats BananaBake’ and ‘All Bran Fibre Toppers’. All methods, except Fehling,which gave a large degree of variability between replicates, deter-mined ‘Rice Bubbles’ to have between 6 and 9 g/100 g total sugar.The CE results for ‘Corn Flakes’ were in agreement with the NIP,which was more than double the amount reported by HPLC; noconclusions could be drawn from Fehling data of this BC due to thelarge degree of error. Sugar content of ‘Oats Banana Bake’ and ‘AllBran Fibre Toppers’measured by CE was approximately half of thatreported on label or measured by Fehling.

The low sugar content BCs determined by CE were ‘Oats Tradi-tional’ and ‘Weet-Bix’. ‘Oats Traditional’ had the lowest sugarcontent measured by CE at 0.2 g/100 g. Interestingly, the NIPlabeled this BC to have 1 g/100 g total sugar, however HPLC data38

reported no sugar at all. Also, sugar contained in ‘Oats Traditional’was below the limit of detection for the Fehling method. The totalsugar content of ‘Weet-Bix’ was consistent between the NIP, theHPLC data and the CE method in this study. At this low level ofsugar, the Fehling method had very poor repeatability (RSD¼89%)as one of the repeats was below LOD.38

The overall error for the quantification of sugars in BCs by the CEmethod was much lower than that of the Fehling method, espe-cially for the BCs with low sugar content. The error for the Fehlingmethod is mainly caused by human technique (such as over-titration), whereas CE is automated and is more affected by oper-ation error. In the case of the low sugar content BCs, the Fehlingmethod was not sensitive enough to produce precise and accurateresults. The most criticized aspect of CE in terms of operation erroris related to volume variation between injections. Injections in CEare achieved by inserting a capillary into a sample solution vial andusing pressure to draw sample solution into the capillary (hydro-dynamic injection). Pressure variations lead to differences in in-jection volume and thus to relatively poor peak area precision. Aspreviously mentioned, an internal standard was used in this work(to correct the peak area) and this eliminated this type of error andgreatly improved precision as it had been observed with the CEmethod applied to plant fibers18 or fermentation monitoring.19

There was no available data on the precision of the NIP and HPLCmeasurements. However, CE was more comparable to HPLC thanNIP or the Fehling method.

2.5. Estimation of reducing sugar by CE and DNS assay comparedwith HPLC

A reducing sugar is classified as any sugar that contains or iscapable of forming an aldehyde functional group that can beoxidized to a carboxylic acid functional group. Though the largestproportion of sugar contained in BCs is sucrose, rather thanreducing sugars (lactose, maltose, glucose and fructose), thequantification of these sugars is importantdespecially glucose (keyproduct of digestion). The same ground samples (with fruitremoved) were used to measure reducing sugars with both CE andDNS methods (Fig. 5), which may explain the significantly lowercontent in reducing sugars in ‘Sultana Bran’ and ‘Sustain’measuredby CE and DNS compared to HPLC data. As previously discussed,individual CE results confirmed the reducing sugars quantified byHPLC38 were fructose and glucose (see Fig. 3B). The use of the DNSassay for the estimation of reducing sugars is a widely practicedassay and also recommended by the International Union of Pureand Applied Chemistry (IUPAC).43 There was no significant differ-ence between the sugar concentrations from CE and DNS assay. Thesugar concentrations were found to agree with that from CE andHPLC in some cases, but not for the most complex matrices (plantfiber).41 Overall, the repeatability of the CE results was better thanthat of the DNS data in this work. During the DNS assay, relativelyharsh reaction conditions (pH 13.0, 100 �C, 5e10 min) is likely tocause starch degradation. In addition, other side reactions (espe-cially involving minerals) may compete for the availability of theDNS reagent.10,11 The simple suspension in water used for samplepreparation in CE ensure no or much more limited degradation forthe use of CE on starchy food.

3. Conclusion

FSCE with direct UV detection was shown to be advantageousfor measuring sugar content in breakfast cereals compared totraditional reducing sugar and glucose-specific methods forseveral reasons: (1) BCs can contain as many as 5 sugars; CE is ableto separate and quantify all sugars in a sample compared to

Table 3Breakfast cereal samples according to total sugar content as listed on Nutrition In-formation Panel (NIP, or packaging label information). Sugar quantity is listed in gper 100 g of BC

Sample name (Brand name) Sugar (g/100 g)

All Bran Fibre Toppers™ (Kellogg's®) 19.6Coco Pops® (Kellogg's®) 36.5Corn Flakes® (Kellogg's®) 7.9Froot Loops® (Kellogg's®) 38.0Nutri-Grain® (Kellogg's®) 32.0Rice Bubbles® (Kellogg's®) 9.0Sultana Bran (Kellogg's®) 22.7Sustain® (Kellogg's®) 20.4Weet-Bix™ (Sanitarium™) 3.3Weet-Bix™ Multi-grain (Sanitarium™) 9.9Oats Traditional (Uncle Tobys®) 1.0Oats apple & blueberry bake (Uncle Tobys®) 25Oats banana bake (Uncle Tobys®) 22.7

M.R. Toutounji et al. / Carbohydrate Research 408 (2015) 134e141140

traditional methods, which give an estimate of total sugar afterinversion of sucrose (a great disadvantage when sample compo-sition is unknown), (2) BCs with the lowest measured sugar levels,‘Oats Traditional’ and ‘Weet-Bix Original’, had sugar contentsbelow the LOD of the Fehling method, (3) the precision of datacollected by CE is greater than that from traditional methods likelydue to automation of the CE. While GC has been used to measureindividual sugars in BCs, multi-step sample preparation involvingderivatization12 or enzymatic removal of starch and acid hydro-lysis of the non-starch polysaccharides13,14 is time consuming andpresents a high risk of sample loss. At present, high performanceliquid chromatography (HPLC) is the gold standard for measuringindividual sugars in food. This study showed that, for all BCs(excluding those containing fruit), CE was able to detect andquantify more sugars compared to data measured with HPLC.38 CEis also more flexible, has a much lower running cost and requiresmuch less sample preparation than HPLC. High performanceanion exchange chromatography (HPAEC) has some of theinherent disadvantages associated with normal phase HPLC, buthas a greater sensitivity range than CE. In summary, FSCE is, withHPAEC, the method to recommend for the analysis of carbohy-drates in breakfast cereals. Future applications of this simple androbust CE method in foodstuffs could be enzymatic hydrolysismonitoring (in vitro digestibility studies) as previously done byNMR.44

4. Materials and methods

4.1. Materials

Milli Q quality (Millipore, Bedford, MA, USA) water was usedthroughout the analysis. Sodium hydroxide pellets (NaOH),disodium monohydrogen phosphate powder (Na2HPO4 stored in adesiccator), 100% pure glacial acetic acid hydrochloric acid andmethylene blue (C.I. 52015) were sourced from Ajax Chemicals(Auburn, NSW, Australia). Copper (II) sulfate was purchased fromFisons (Homebush, NSW, Australia). Citric acid (anhydrous) wasobtained from Chem-Supply Pty Ltd (Gillman, SA, Australia). Zincacetate �99% was supplied by BDH AnalaR, Merck Pty Limited(Poole, Dorset, England). Sodium potassium tartrate 99%, sodiumbisulfite, phenol, 3,5-dinitrosalicyclic acid, Glucose �99.5% anddimethyl sulfoxide (DMSO) �99.5% were supplied by Sigma-eAldrich (Castle Hill, NSW, Australia). Xylose �99% was from AlfaAesar (Ward Hill, MA, USA). Fused-silica capillaries (50 mm in-ternal diameter, 360 mm outside diameter) were obtained fromPolymicro (Phoenix, AZ, USA). Infinity™ glucose oxidase liquidstable reagent, pH 7.5±0.1 at 20 �C, was obtained from ThermoScientific (TR-15221, Worthing, West Sussex, UK).

4.2. Initial sample preparation

Thirteen breakfast cereals (BC) products were purchased from alocal Woolworths supermarket (Marayong, NSW, Australia).Approximately 80e150 g of each BC wasmilled in a K-mart, m-miniglass jug blender for 20 s, speed level 1. The ground cereal waspassed through a laboratory sieve with pore size 1000 mm andretained in a 500 mm sized sieve, producing BC samples with par-ticle size between 500 and 1000 mm. Samples were stored in a coldroom at 4 �C. Basic information of the 13 BC samples, includingsugar content, is shown in Table 3.

4.3. Capillary electrophoresis

For CE separations, disodium hydrogen phosphate(NP200d130mMNaOH and 36mMNa2HPO4) buffer was prepared

according to Ref. 30. This buffer was prepared on the day of use,sonicated for 5 min and filtered with a Millipore membrane syringefilter (0.2 mm). A stock solution of sugars (standard) was preparedcontaining 1.5 g L�1 sucrose and 0.5 g L�1 of each maltose, glucoseand fructose in water. Standard curves were obtained using anundiluted standard and standards diluted by factors of 2, 4, 8, 16,and 32. Sample solutions of 10.0 g L�1 were prepared by adding15.0 mg of ground sample (500e1000 mm particle size) to 1.500 mLof water. To each of the standards and samples, 0.50 g L�1 xylosewas added as an internal standard as well as DMSO (5 mL per500 mL) to mark the electro-osmotic flow.

Separations were performed on a Beckman P/ACEMDQ capillaryelectrophoresis system (AB Sciex Separations, Mount Waverley,Australia) monitoring at 191 nm, 266 nm and 270 nmwith a 10 nmbandwidth. A capillary with a total length of 61.8 cm (51.8 cmeffective length) was used. The capillary was preconditioned beforeuse by flushing with 1 M NaOH, 0.1 M NaOH, water and NP200buffer for 20 min each. The cassette temperature was set to 15 �C.Samples were injected by applying 34 mbar for 4 s followed byinjection of NP200 buffer in the same manner. A voltage of 16 kVwas ramped up over 2 min. Between consecutive separations, thecapillary was flushed with NP200 buffer for 5 min. After the finalinjection, the capillary was flushed with 1 M NaOH for 1 min, fol-lowed by water and then air (10 min each). Carbohydrates weremonitored at 266 and 270 nm and the EOF was monitored at191 nm by 32 Karat software. The data was processed first usingeither Origin 8.5.1 (‘manual’ treatment) or 32 Karat (‘automated’treatment). Quantification of the carbohydrates was not done at270 nm as in earlier literature but at 266 nm since it gave theoptimal signal-to-noise ratio as also observed in the most recentliterature.19,32,39 Both data treatment yielding the same results(data not shown) the latter treatment was used for all the resultspresented in this work. Concentration of identifiable sugars in BCswas determined from normalized peak area (relative to the internalstandard) and the standard curve. Outliers were removed, whererelevant, after applying a Grubb test (see Equation S-4).

4.4. DNS assay in micro plate format

Reducing sugars in BCs were quantified by the DNS assay inmicroplate format. The DNS reagent was prepared exactly accord-ing to Ref. 9. Sample solutions of 10.0 g L�1 were prepared byadding 15 mg of ground sample (500e1000 mm particle size) to1.500 mL of water. Triplicates of 100 mL were made up as bothundiluted samples and diluted samples (with dilution factors of 1,2, 4, 8, 16, 32). Glucose standard solutions of 0.2, 0.4, 0.6, 0.8,1.0 g L�1 were prepared in triplicates for the purpose of procuring a

M.R. Toutounji et al. / Carbohydrate Research 408 (2015) 134e141 141

standard curve. A blank and a set of standards were included withthe samples tested on each microtitre plate. The contents of theplates were then mixed on a plate mixer for 5 s, sealed and incu-bated at 100 �C in a water bath for 10 min. The plates were thencooled to room temperature to stop the reaction by being placed onice. Absorbance values were measured at 640 nm on a microplatereader with the standard containing no glucose as the blank.

4.5. Fehling (LaneeEynon) method

Estimation of total sugar content in BCs was carried out usingFehling's solution reagents as described in Ref. 45. Fehling solutionA contained 69.3 g of copper sulfate in 1 L of water, Fehling solutionB contained 346 g of Rochelle salt (potassium sodium tartrate) in 1 Lof water. Carrez solution 1 contained 21.9 g of zinc acetate and 3mLof glacial acetic acid in 100mL of water, Carrez solution 2 contained10.6 g of potassium ferrocyanide in 100 mL of water. Each groundBC sample (5 g) was added to 100 mL water, mixed with 5 mL eachof Carrez solutions 1 and 2 andmade up to a total volume of 250mLwith water. The mixture was then decanted and filtered throughWhatman® 540 filter paper. The clarified solution (25 mL of it) wastransferred to a conical flask, mixed with 2.5 g citric acid and gentlyboiled on a hot plate for 5 min to ensure the inversion of any su-crose present. The sample solution was then neutralized to pH6.5e7.5 with 1 M NaOH and made up to a total volume of 250 mLwith water. This gave an approximate sample concentration insample preparation of 2 g L�1. Fehling's solution was titratedagainst the sample solution as reported previously for the standardglucose solution. The total sugars content was calculated followingEquation S-1 (see Supplementary data).

Standardization of Fehling's solutions was carried out using a10 g L�1 glucose solution. Fehling's solutions (2.5 mL each of A andB) were mixed with 30e50 mL water in a ceramic crucible andheated to 100 �C over a Bunsen burner for 15 s. Three drops ofmethylene blue indicator were added followed by dropwise addi-tion of the standard glucose solution until only a faint blue colorremained.

Acknowledgments

The authors acknowledge a UWS School of Science and HealthEquipment Grant for the CE instrument.

Supplementary data

The supplementary data includes the details of the calculation oftotal sugars in Fehling method, CE electropherograms with spikedBC samples, repeatability of CE electropherograms and tables withthe RSDs of the electrophoretic mobility, the comparison of LODsand LOQs with published values and the statistics applied to detectoutliers. Supplementary data related to this article can be found athttp://dx.doi.org/10.1016/j.carres.2015.03.008.

References and notes

1. Woods J, Walker K. Nutr Diet 2007;64:226e33.

2. Brummer Y, Cui S. Understanding carbohydrate analysis. In: Cui S, editor. Foodcarbohydrates. 1st ed. Florida, USA: CRC Press; 2005. p. 69e74.

3. Brummer Y, Cui S. Understanding carbohydrate analysis. In: Cui S, editor. Foodcarbohydrates. 1st ed. Florida, USA: CRC Press; 2005. p. 74e89.

4. Southgate DAT. Introduction. In: Southgate DAT, editor. Determination of foodcarbohydrates. 2nd ed. London, UK: Elsevier Applied Science; 1991. p. 1e7.

5. Southgate DAT. The carbohydrates in foods. In: Southgate DAT, editor. Deter-mination of food carbohydrates. 2nd ed. London, UK: Elsevier Applied Science;1991. p. 9e33.

6. Shannon IL. ASDC J Dent Child 1974;41:347e50.7. Shannon IL. Brand name guide to sugar: sucrose content of over 1,000 common

foods and beverages. Chicago, USA: Nelson-Hall; 1977.8. FitzGerald J, Vermerris W. Biotechnol Appl Biochem 2005;41:233e9.9. Miller GL. Anal Chem 1959;31:426e8.

10. Forouchi E, Gunn DJ. Biotechnol Bioeng 1983;25:1905e11.11. Leung DWM, Thorpe TA. Phytochemistry 1984;23:2949e50.12. Li BW, Schuhmann PJ. J Food Sci 1980;45:138e41.13. Englyst HN, Quigley ME, Hudson GJ. Analyst 1994;119:1497e509.14. Bravo L, Englyst HN, Hudson GJ. Food Res Int 1999;31:129e35.15. Baker D, Norris KH. Appl Spectrosc 1985;39:618e21.16. Jones AD, Burns IW, Sellings SG, Cox JA. J Chromatogr A 1977;144:169e80.17. Shanmugavelan P, Kim SY, Kim JB, Kim HW, Cho SM, Kim SN, et al. Carbohydr

Res 2013;380:112e7.18. Oliver JD, Gaborieau M, Hilder EF, Castignolles P. J Chromatogr A 2013;1291:

179e86.19. Oliver JD, Sutton AT, Karu N, Phillips M, Markham J, Peiris P, et al. Biotechnol

Appl Biochem 2015. http://dx.doi.org/10.1002/bab.1269 [in press].20. Hanko VP, Rohrer JS. Anal Biochem 2000;283:192e9.21. Cataldi TRI, Campa C, De Benedetto GE. Fresenius J Anal Chem 2000;368:

739e58.22. Klampfl CW, Himmelsbach M, Buchberger W. Analysis of simple carbohydrates

by capillary electrophoresis and capillary electrophoresis-mass spectrometry. NewJersey, USA: Humana Press Inc; 2011.

23. Camilleri P. Capillary electrophoresis: theory and practice. 2nd ed. Boca Raton,Fla: CRC Press; 1998.

24. Jager AV, Tonin FG, Tavares MFM. J Sep Sci 2007;30:586e94.25. Rizelio VM, Tenfen L, da Silveira R, Gonzaga LV, Costa ACO, Fett R. Talanta

2012;93:62e6.26. Soga T, Serwe M. Food Chem 2000;69:339e44.27. Noe CR, Lachmann B, M€ollenbeck S, Richter P. Z Lebensm Unters Forsch

1999;208:148e52.28. Garcia-Canas V, Simo C, Castro-Puyana M, Cifuentes A. Electrophoresis 2014;35:

147e69.29. Lee ISL, Boyce MC, Breadmore MC. Food Chem 2012;133:205e11.30. Rovio S, Yli-Kauhaluoma J, Sir�en H. Electrophoresis 2007;28:3129e35.31. Sarazin C, Delaunay N, Costanza C, Eudes V, Mallet J-M, Gareil P. Anal Chem

2011;83:7381e7.32. Oliver JD, Rosser AA, Fellows CM, Guillaneuf Y, Clement JL, Gaborieau M, et al.

Anal Chim Acta 2014;809:183e93.33. Anderson GH, Cho CE, Akhavan T, Mollard RC, Luhovyy BL, Finocchiaro ET. Am J

Clin Nutr 2010;91:932e9.34. Thevarajah JJ, Gaborieau M, Castignolles P. 2014;2014:11.35. Rovio S, Simolin H, Koljonen K, Sir�en H. J Chromatogr A 2008;1185:139e44.36. Sarazin C, Delaunay N, Costanza C, Eudes V, Gareil P. Talanta 2013;103:301e5.37. Sarazin C, Delaunay N, Costanza C, Eudes V, Gareil P. Talanta 2012;99:202e6.38. FSANZ Food Standards Australia New Zealand. Explanatory notes for use with

NUTTAB. 2010. http://www.foodstandards.gov.au/consumerinformation/nuttab2010/nuttab2010onlinesearchabledatabase/onlineversion.cfm [accessed15.04.13].

39. Oliver JD, Gaborieau M, Castignolles P. J Chromatogr A 2014;1348:150e7.40. Mala Z, Gebauer P, Bocek P. Electrophoresis 2011;32:1500e7.41. Vaher M, Helmja K, K€asper A, Kura�sin M, V€aljam€ae P, Kudrja�sova M, et al. Catal

Today 2012;196:34e41.42. Bennion EB, Bamford GST. The technology of cake making. 6th ed. London, UK:

Blackie Academic and Professional; 1997.43. Ghose T. Pure Appl Chem 1987;59:257e68.44. Dona A, Pages G, Gilbert R, Gaborieau M, Kuchel P. Biomacromolecules 2009;10:

638e44.45. Ranganna S. Handbook of analysis and quality control for fruit and vegetable

products. New Delhi, India: Tata McGraw-Hill; 1986.