Carbohydrate Polymers 179 (2018) 379–385

Contents lists available at ScienceDirect

Carbohydrate Polymers
journal homepage: www.elsevier.com/locate/carbpol

Quantification of food polysaccharide mixtures by 1H NMR MARK

a a a,⁎ a
Donny W.H. Merkx , Yvonne Westphal , Ewoud J.J. van Velzen , Kavish V. Thakoer ,
Niels de Rooa, John P.M. van Duynhovena,b
a
Unilever R & D, Olivier van Noortlaan 120, 3133 AT, Vlaardingen, The Netherlands
b
Wageningen University & Research, Laboratory of Biophysics, Stippeneng 4, 6708 WE, Wageningen, The Netherlands

A R T I C L E I N F O A B S T R A C T

Keywords: Polysaccharides are food ingredients that critically determine rheological properties and shelf life. A qualitative
Food polysaccharides and quantitative assessment on food-specific polysaccharide mixtures by 1H NMR is presented. The method is
Mixture analysis based on the identification of intact polysaccharides, combined with a quantitative analysis of their mono-
Non-instantaneous monosaccharide release saccharide constituents. Identification of the polysaccharides is achieved by 1H NMR line shape fitting with pure
qNMR
compound spectra. The monomeric composition was determined using the Saeman hydrolysis procedure, fol-
Saeman hydrolysis
lowed by direct monosaccharide quantification by 1H NMR. In the quantification, both the monosaccharide
degradation during hydrolysis, as well as a correction for the non-instantaneous polysaccharide dissolution were
taken into account. These factors were particularly important for the quantification of pectins. The method
showed overall good repeatability (RSDr = 4.1 ± 0.9%) and within-laboratory reproducibility
(RSDR = 6.1 ± 1.4%) for various food polysaccharides. Polysaccharide mixtures were quantitatively resolved
by a non-negative least squares estimation, using identified polysaccharides and their molar monosaccharide
stoichiometry as prior knowledge. The accuracy and precision of the presented method make it applicable to a
wide range of food polysaccharide mixtures with complex and overlapping 1H NMR spectra.

1. Introduction strong β-1,4-glycosidic bonds of for example cellulose, whilst being

Polysaccharides are widely used in food products and critically
determine rheological properties (Dickinson, 2003;
Saha & Bhattacharya, 2010) and shelf life (Wang, Li, Copeland,
Niu, & Wang, 2015). In order to understand and manipulate these
macroscopic properties, a detailed and quantitative knowledge on the
polysaccharide composition of complex food products is required. This
is a challenging task since food polysaccharides show a wide structural
diversity, are often formulated at low levels and engage in strong in-
teractions with the food matrix. Polysaccharide fractions can be readily
isolated from complex food matrices, with mg-scale yields (Grün et al.,
2015). The quantification of polysaccharide mixtures, however, re-
mains a significant challenge. Commonly used techniques for poly-
saccharide quantification are based on hydrolysis/methanolysis of the
polysaccharides, followed by quantification of the released mono-
saccharides.
Hydrolysis with hydrochloric acid (HCl) (Bertaud,
Sundberg, & Holmbom, 2002), trifluoroacetic acid (TFA) (De Ruiter,
Schols, Voragen, & Rombouts, 1992) and sulphuric acid (Saeman,
Moore, Mitchell, & Millet, 1954) are the most common approaches. The
Saeman hydrolysis procedure uses sulphuric acid (H2SO4) to break the
relatively mild towards most released monosaccharides (Saeman et al.,
1954). Typically, the monosaccharide content after degradation is de-
termined by combining GC and colorimetric assays
(Ahmed & Labavitch, 1977; Englyst & Cummings, 1984). These methods
are, however, not able to quantify neutral sugars and uronic acids in a
single experiment and are not able to distinguish between different
uronic acids. Furthermore, typical quantification of neutral sugars re-
quire laborious derivatisation prior to GC analysis, which limit the
overall sample throughput significantly.
Quantitative 1H NMR (qNMR) analysis has been introduced as a
high-throughput quantification method (van Duynhoven, van
Velzen, & Jacobs, 2013). It does not require monosaccharide specific
calibration and for quantification only one primary standard needs to
be used, either internal or external. qNMR can be applied directly on
the hydrolysate mixture in sulphuric acid, which significantly simplifies
sample preparation. Recently, two research groups (Carvalho de Souza,
Rietkerk, Selin, & Lankhorst, 2013; Mittal, Scott, Amidon,
Kiemle, & Stipanovic, 2009) have tailored the Saeman hydrolysis pro-
cedure towards 1H NMR quantification by exploring the use of deut-
erated sulphuric acid (D2SO4) and optimising the sample preparation
for 1H NMR. For quantification purposes, a degradation factor is

⁎
Corresponding author.
E-mail address: Ewoud-van.Velzen@Unilever.com (E.J.J. van Velzen).

http://dx.doi.org/10.1016/j.carbpol.2017.09.074
Received 26 June 2017; Received in revised form 20 September 2017; Accepted 22 September 2017
Available online 28 September 2017
0144-8617/ © 2017 Elsevier Ltd. All rights reserved.
D.W.H. Merkx et al.
commonly applied to correct for the degradation of the mono-
saccharides during the hydrolysis step (De Ruiter et al., 1992). This
factor does not take the rate of monomeric release into account and
inherently assumes that all monosaccharides are instantaneously re-
leased at timepoint to. This assumption, however, may not hold for all
monosaccharides. In cases where the monosaccharide release evolves
over a time interval Δt > 0, the quantification will result in over-
estimated results. Moreover, the currently available NMR methods have
been primarily optimised for analysing biomass and not to specifically
consider the compositional complexity of polysaccharides as they occur
in foods (Carvalho de Souza et al., 2013; Mittal et al., 2009).
In the currently reported study, we optimised and validated a rapid
qNMR method for the quantification of polysaccharide mixtures in
complex food systems. Firstly, we identified the individual components
of food polysaccharides mixtures from the 1H NMR spectra by means of
spectral line shape fitting (van Velzen et al., 2015; Wishart, 2008), also
referred to as targeted profiling (Jacobs, van Velzen, & Mihaleva, 2013;
Mercier, Lewis, Chang, Baker, & Wishart, 2011). In this approach the
NMR spectrum of the polysaccharide mixture is considered as the sum
of the individual spectra of the pure components. Secondly, we ex-
tended the qNMR procedure such that it utilises the monomeric com-
position of the polysaccharides, rather than its native forms
(Tojo & Prado, 2003) for quantification. We used the Saeman hydrolysis
procedure (Carvalho de Souza et al., 2013; Mittal et al., 2009; Saeman
et al., 1954) to obtain the monomeric composition from the poly-
saccharides. Thirdly, we quantified the monomeric composition by
means of Quantum Mechanical Total Line Shape (QMTLS) fitting
(Mihaleva et al., 2014; Tiainen, Soininen, & Laatikainen, 2014), and
furthermore adapted the qNMR method for polysaccharide-specific
differences in (polymeric) dissolution and (monomeric) degradation
properties as a result of the hydrolysis. Finally, we used the estimated
monosaccharide concentrations for quantification of polysaccharide
mixtures. This analysis requires prior knowledge on the polysaccharide
system and the associated monosaccharide stoichiometries. The per-
formance of the mixture analysis was tested with representative ternary
polysaccharide mixtures.
2. Materials and methods
2.1. Polysaccharide isolation and identification
The full procedure for isolation and subsequent identification of the
polysaccharide fraction from a food product matrix is described in
previous research (Grün et al., 2015).
2.2. Solvents and reagents
Deuterated water (D2O) was purchased from Euriso-Top (Saint
Aubin, France), deuterated sulphuric acid (D2SO4) was obtained from
Sigma-Aldrich (Zwijndrecht, the Netherlands). All monosaccharides
(glucose, galactose, fucose, xylose, arabinose, rhamnose, mannose,
glucuronic acid and galacturonic acid) were obtained from Sigma-
Aldrich. Xanthan gum (FNCS) was obtained from Jungbunzlauer Suisse
AG (Basel, Switzerland). GRINSTED® pectins and locust bean gum were
purchased from Danisco (Copenhagen, Denmark). The GRINSTED®
pectins had varying Degrees of Esterification (DE) and Degrees of
Amidation (DA): LC810 (DE 37%), LC710 (DE 48%), LA210 (DE 34%,
DA 17%). Rhamnogalacturonan (from soy bean pectin) from Megazyme
(Bray, Ireland), tara gum (HV grade) from Kalys S.A. (Bernin, France),
low acyl gellan (Kelcogel), high acyl gellan (Kelcogel LT100) and pectin
LM5 (DE 10%) from CPKelco (Atlanta, USA) and guar gum and cellu-
lose from Sigma-Aldrich. The 72% (w/w) D2SO4 in D2O solution was
freshly prepared. Maleic acid (standard for qNMR, TraceCERT®), pur-
chased from Sigma-Aldrich, was used as internal standard in D2O
(0.5 mg/mL), and accurately weighed (0.01 mg) on an analytical bal-
ance.
380
Carbohydrate Polymers 179 (2018) 379–385
2.3. Polysaccharide hydrolysis
Approximately 25 mg of a polysaccharide sample was accurately
weighed into a 20 mL glass tube. After adding 1 mL of 72% (w/w)
D2SO4 solution to the sample, the glass tube was sealed and stirred at
room temperature for 1 h. Upon dissolution, 6.2 mL of D2O was added
to obtain a final D2SO4 concentration of 13% (w/w) in D2O. The sample
was then incubated in a convection oven at 100 °C for 180 min. In the
specific case of the dissolution determination of polysaccharides, the
samples were incubated between 0 and 180 min, with increments of 15
or 30 min. After incubation, the sample was allowed to cool down to
room temperature, after which 1 mL of maleic acid solution (internal
standard) was added. The samples were, when required, centrifuged
and transferred to a 3-mm NMR glass tube.
2.4. NMR experiments
The NMR-experiments were recorded with a zg30 pulse sequence on
a Bruker Avance III 600 MHz NMR spectrometer equipped with a 5-mm
cryo-probe. The internal temperature of the probe was set at 290 K. The
size of the FID was 96k. In total, 32 scans were collected with a re-
laxation time of 10 s and an acquisition time of 4 s. The 90° pulse length
(∼12.5 μs) and receiver gain were determined automatically. The data
was processed with Bruker TopSpin 3.2 software. Fourier transforma-
tion with exponential window function and line broadening factor of
0.3 Hz were applied, followed by automatic phase correction and
baseline correction.
2.5. QMTLS fitting
QMTLS fitting was performed with PERCH NMR software (PERCH
solutions Ltd., Kuopui, Finland). A tailored spectral library was created
for the line shape fitting of the monosaccharides of interest: glucose,
galactose, mannose, rhamnose, xylose, fucose, arabinose, glucuronic
acid and galacturonic acid. Each of the compounds was described by a
set of spin particles and average, initial starting values for chemical
shifts, coupling constants, line widths and intensities (populations).
These parameters were obtained by fitting the pure model compounds
recorded in 13% (w/w) D2SO4 into the experimental spectra using the
PERCH NMR software. All parameters were iteratively optimised. The
internal standard (maleic acid) was used to calculate the absolute
compound concentrations from the fitted populations. Python scripts
were used for performing the QMTLS fitting in a non-supervised auto-
mated manner.
2.6. Monosaccharide quantification
Based on the NMR signal integral of the monosaccharide (Ix) and the
internal standard (IIS), the molar quantity (Nx) of the monosaccharide x
in the (reference) sample was estimated according to the general qNMR
equation (Malz & Jancke, 2005) (Eq. (1)):
Ix ⋅NPIS⋅WIS⋅FIS⋅PIS
Nx =
IIS⋅NPx⋅MWIS⋅VIS (1)
where, N is the molar quantity (mol), I is the signal integral, NP is the
number of protons, MW is the molecular weight (g/mol), W is the
weight (g), V is the weight of the solvent volume (g), F is the volume
fraction in the test solution (g) and P is the content mass fraction of the
internal standard. Index x stands for the respective monosaccharide and
IS for the internal standard.
During the hydrolysis phase in the experiment, the monosaccharides
underwent degradation. The degree of degradation is dependent on
monosaccharide x and the hydrolysis time th, and was expressed as the
degradation factor Dx (th) . Dx (th) was estimated from the mono-
saccharide quantity in a Sugar Recovery Standard (SRS) before hydro-
lysis (Nx (t 0) ) and after hydrolysis (Nx (th) ) following Eq. (2) (Carvalho de
D.W.H. Merkx et al.
Souza et al., 2013):
Nx (th)
Dx (th) = 1 − = k x th
Nx (t 0)
where, kx is the rate constant (min−1) of the degradation. Eq. (2) was
obtained by solving the differential equation of a zero-order reaction.
Here we assumed a constant degradation rate of x during the hydrolysis
phase.
When quantifying monosaccharides using Eq. (1), and estimating
the associated degradation factor using Eq. (2), the general convention
is that all dissolved monosaccharides are present at t0. This assumption
holds for the SRS solution, but does not hold for polysaccharide systems
where monosaccharides are gradually released during hydrolysis. To
prevent overestimation of the results, a modification of the degradation
factor Dx (th) was required.
The release of monosaccharide x at timepoint tr, which is dependent
on the polysaccharide y, is denoted as the derivative function f ′ y (t r) .
The associated non-derivative function f y (t r) represents the cumulative
dissolution curve of y. This dissolution function f y (t r) was directly es-
timated from a pure polysaccharide standard with a series of NMR
experiments collected over a time interval [to, th]. A selected char-
acteristic isolated polysaccharide peak was (semi)-quantified on each
measured timepoint by manual integration (Fig. 3). When combining
the release function ( f ′ y (t r) ) and the rate constant of degradation (kx),
the monosaccharide quantity Nxy (t r) was calculated at any timepoint tr
according to Eq. (3) using a solver (ODE45 in Matlab R2016a, The
MathWorks, Inc., Massachusetts, USA):
tr
Nxy (t r) = ∫ f ′ y (t r)(1 − k x (th − t r)) dt r
t0
in case a monosaccharide x does not degrade during hydrolysis, thus kx
is 0, then Nxy (t r) becomes solely dependent on the release function (Eq.
(4)). Here we introduce N 0y (t r) as the monosaccharide quantity (mol)
that could be obtained in the absence of any monosaccharide de-
gradation:
tr
N 0y (t r) = ∫ f ′ y (t r) dt r
t0
similar to Eq. (2), the modified degradation factor Dxy (t r) was calcu-
lated (Eq. (5)). Note that Dxy (t r) is dependent on x and y:
Nxy (t r)
Dxy (t r) = 1 −
N 0y (t r)
the monosaccharide concentration (Nx (t r) , mol) was finally calculated
with the modified degradation factor according to Eq. (6):
Ix ⋅NPIS⋅WIS⋅FIS⋅PIS 1
Nx (t r) = ⋅ saccharides were investigated in 5-fold at three different concentrations
IIS⋅NPx⋅MWIS⋅VIS 1 − Dxy (t r)
2.7. Polysaccharide quantification
Quantification of pure polysaccharides can be directly achieved
from the estimated concentration levels of the individual mono-
saccharides (Eq. (6)). When considering a collection of L mono-
saccharides for a polysaccharide y, we can define a row vector Nx
(1 × L) containing all molar monosaccharide quantities. Multiplication
of Nx with their associated molecular weights (MWx, size (L × 1)) re-
sults in the absolute total weight of that particular polysaccharide (Wy
in g) in the sample material as denoted in the matrix calculation in Eq.
(7). Note that the molecular weights of ‘anhydrous’ monomeric units in
polysaccharides differ from free ‘hydrous’ monosaccharides (Δ = 18 g/
mol).
Wy = Nx MWx
Carbohydrate Polymers 179 (2018) 379–385
2.8. Validation of polysaccharide quantification
Repeatability (r) and the within-laboratory reproducibility (R) were
(2)
determined on a series of twelve different food polysaccharides. Both r
and R were pooled over three concentrations levels (5, 10 and 15 mg/
mL) that were measured in duplicate on four different days. Moreover,
for a commercially available microcrystalline cellulose (Sigma Aldrich)
the trueness, linear working range and detection limit were determined.
The minimum specified purity of 95% (w/w), as given by the supplier,
was accepted as the reference value. Trueness was estimated with blank
samples that were spiked with increasing amounts of the cellulose
standard within the concentration range (0–60 mg/mL). The cellulose
standard was dried for 24 h at 70 °C in a drying oven prior to the
analysis. The linear working range was estimated for the same con-
centration range. Estimated cellulose concentrations were compared
with the actual cellulose concentrations using a Lack-of-Fit test. The
limit of quantification was based on an estimate of (10 times) the
standard deviation from 10 calibration samples containing a small
amount of the cellulose standard (1 mg).
2.9. Polysaccharide mixture analysis
Quantitative analysis of polysaccharide mixtures needs additional
information on the molar monomeric stoichiometry (per polysaccharide
y), in addition to Nx. The concatenated stoichiometries of all poly-
saccharides together form the stoichiometry matrix ST. In addition, if
the mixture contains J polysaccharides, Ny can be replaced by a row
vector Ny. The general quantitative equation for solving polysaccharide
mixtures (via least squares regression) is given in Eq. (8):
(3)
Ny = (STS)−1Nx S (8)
Where, Ny is a row vector (1 × J) of molar polysaccharide quantities
(mol), ST is a stoichiometry matrix (J × L), and Nx is a row vector
(1 × L) of molar monosaccharide quantities.
To solve this regression problem, we applied an alternating least
squares algorithm (PLS Toolbox 7.8.2, Eigenvector Research, Inc.) with
non-negativity constraints on Ny and equality constraints on ST in
Matlab R2016a.
(4)
2.10. Validation of polysaccharide quantification
A series of polysaccharide mixtures were analysed according to a
full 33 three-level factorial design with a centre point. Three poly-
(5) saccharides (factors) were included in the design, i.e. high acyl gellan,
xanthan and LBG. The estimated polysaccharide ratios were compared
with the actual (theoretical) quantities in order to investigate the ac-
curacy of the mixture analysis. All 9 combinations of the three poly-
(6) levels (5 mg, 2.5 mg, 0 mg). Agreement between estimated and theo-
retical values was evaluated in a trueness study and tested on off-set
bias and proportional bias.
3. Results and discussion
3.1. Identification of polysaccharides in food systems
The first step towards the quantification of polysaccharides in food
systems is their isolation and subsequent identification. Recently, an
extensive isolation procedure for unbiased and full recovery of poly-
saccharides from complex food matrices has been described (Grün
et al., 2015). This procedure yields isolates of polysaccharides that are
devoid of low molecular weight molecules. Fig. 1 shows an example of a
1
H NMR spectrum of a polysaccharide isolate that was obtained from a
food product, where the individual polysaccharides were present at 1%
(7) (w/w) or less. Even though the signals are broad and show a large
381
D.W.H. Merkx et al.
degree of overlap, the spectral contributions of different poly-
saccharides could clearly be recognised. Component identification is
aided by fitting the 1H NMR spectra of pure polysaccharides into the 1H
NMR spectrum of the isolate. This approach is referred to as line shape
fitting or curve fitting (Wishart, 2008) and showed that the mixture 1H
NMR spectrum was largely explained by the sum of few poly-
saccharides. Fig. 1 showed that the mixture analysis lead to the un-
ambiguous identification of four different polysaccharides which are
commonly found in food (guar gum, xanthan, LBG and low acyl gellan).
However, due to the unknown purity of most polysaccharide standards
we deem the line shape fitting approach unsuitable for direct quanti-
fication. Another downside of this method is that certain poly-
saccharides (like xanthan) yield low-intense, featureless 1H NMR
spectra which can hardly be distinguished from other polysaccharides
in complex mixtures. In addition, subtle differences in spectral features
between comparable polysaccharides within polysaccharide families
(e.g. galactomannans) may not be sufficient for attaining unambiguous
identifications in all cases. Therefore, we proceeded with a high-
throughput approach in which the polysaccharides (or mixtures of
polysaccharides) are converted into their constituent monosaccharides
prior to the quantification (Carvalho de Souza, 2017).
3.2. Rapid Saeman hydrolysis on food polysaccharides
The Saeman procedure was used to hydrolyse the polysaccharides
into their monomeric units. With 1H NMR it is possible to analyse and
quantify the monosaccharides in the Saeman hydrolysates without ex-
tensive additional sample preparation (Carvalho de Souza et al., 2013;
Mittal et al., 2009). We adapted the developed methods and optimised
382
Carbohydrate Polymers 179 (2018) 379–385
Fig. 1. 1H NMR spectrum (600 MHz) of a bouillon
cube isolate ( ). The polysaccharides in this
isolate were identified by line shape fitting using the
pure spectra of guar gum ( ), LBG ( ), low
acyl gellan ( ) and xanthan ( ).
them for food-specific polysaccharides. Firstly, for xanthan and pectin
the hydrolysis time was extended to 180 min (instead of 90 min) to
attain full conversion of the polysaccharides into their monosaccharide
constituents. To unify the hydrolysis protocol for all polysaccharides,
this prolonged hydrolysis time was standardised in the method. Sec-
ondly, we introduced an adapted pre-solubilisation step at room tem-
perature without temperature control. This makes the method faster
and simpler when compared with the original protocol that aimed for a
temperature-controlled pre-solubilisation in a thermostatic water bath
(30 °C). This adaptation does not affect the method performance and
makes it more convenient for high-throughput purposes. In the data
acquisition, a few optimisations were introduced as well. To avoid
suppression of the anomeric sugar-signals (δ 3.7–4.6 ppm), 1H NMR
spectra were recorded with a 30° pulse sequence without water sup-
pression. In addition, the relaxation delay was set to 10 s, thereby en-
hancing the analysis throughput (up to six samples per hour) while
maintaining the quantitative nature of the 1H NMR spectra. High field
NMR was required to obtain the desired spectral resolution on the α-
and β-anomeric monosaccharide signals, which were used for the
quantification. For quantification, we used both the α- and β-anomeric
monosaccharide signals. Up to nine different types of monosaccharides
were observed in the final hydrolysates. In mixtures, these mono-
saccharides often have overlapping anomeric signals, which hinder
manual quantification (Fig. 2). Quantification based on Quantum Me-
chanical Total Line Shape (QMTLS) fitting offers a suitable solution in
mixture analysis (Tiainen et al., 2014). QMTLS fitting can be used for
estimating the signal area of overlapping anomeric doublets in-
dependently and has a major advantage that the calculations can be
executed in automation.
Fig. 2. 1H NMR spectrum (600 MHz) of the anomeric
region of the nine monosaccharides considered in
this study: glucose (Glc), galactose (Gal), arabinose
(Ara), fucose (Fuc), xylose (Xyl) mannose (Man),
rhamnose (Rha), glucuronic acid (GlcA) and ga-
lacturonic acid (GalA).
D.W.H. Merkx et al.
3.3. Non-instantaneous polysaccharide hydrolysis
While optimising the method, we detected a systematic bias be-
tween the measured quantities and the true quantities for particular
polysaccharides. The measured quantities for pectin for example were
overall larger than their accepted values. It is however well known that
during the Saeman procedure (in strong acidic conditions at elevated
temperatures) not only hydrolysis of the polysaccharides takes place
but also degradation of the released monosaccharides (Shi, Yokoyama,
Akiyama, Yashiro, & Matsumoto, 2012). Hence current monosaccharide
quantification methods use a degradation factor to correct for the
monosaccharide degradation (Eq. (2)). A major advantage of 1H NMR
in combination with Saeman monosaccharide analysis is that poly-
saccharides and monosaccharides can be observed simultaneously, thus
allowing for direct assessment of the hydrolysis state. The time course
of hydrolysis is demonstrated in Fig. 3 over a period of 180 min for the
polysaccharide pectin. From top to bottom the 1H NMR spectra showed
increasing signals of released monosaccharides (e.g. δ 4.55 ppm, δ
3.97 ppm, δ 3.86 ppm, δ 3.65 ppm, δ 3.56 ppm, and δ 3.50 ppm) and at
the same time the disappearance of pectin (e.g. δ 4.42–4.18 ppm and δ
3.75–3.66 ppm). When integrating the characteristic signal area of
pectin within the spectral region δ 4.42–4.18 ppm (eCHOH), the
383
Carbohydrate Polymers 179 (2018) 379–385
Fig. 3. 1H NMR stacked spectra (600 MHz) of the
hydrolysis of a pectin, sampled at 15 min intervals
for the first 90 min, and sampled at 30 min intervals
for the last 90 min. The greyed area (δ
4.42–4.18 ppm), annotated to eCHOH-signals of the
pectin backbone, is used to estimate polysaccharide
degradation.
progressing degradation can be monitored. In Fig. 4A, the time courses
of both pectin degradation and monosaccharide release were plotted,
thereby indicating that the disappearance of the polysaccharide is in-
versely related to the release of monosaccharides. In Fig. 4A we ob-
served that monomeric release is non-instantaneous. Instead, we could
identify three consecutive stages from the experimental data. In the
initial lag phase of hydrolysis (0–1 h), pectin partially hydrolyses into
oligomers (Coenen, Bakx, Verhoef, Schols, & Voragen, 2007). As the
molecular mobility of pectin increased, also their associated 1H NMR
signals in Fig. 3 sharpen up. Degradation to monosaccharides particu-
larly occurred in the second hour. In this second phase, the majority of
all starting material (90%) degrades. After two hours, hydrolysis
reaches its decelerating phase. In this final stage, the remaining 10% of
pectin is converted at lower reaction rate. The characteristics of the
concentration-time curves in Fig. 4A are pectin-specific. Different
polysaccharides show different trends (Fig. 4B). For example, the de-
gradation of cellulose and xanthan is rapid and linear, whereas pectin,
as shown before, typically demonstrates a slow, non-continuous de-
gradation behaviour.
The concentration-time characteristics during hydrolysis were es-
tablished for a range of polysaccharides that are commonly used in food
products. We found that the conversion into monosaccharides could be
Fig. 4. A) Time course of the pectin degradation
( ), which is inversely related to the time course
of monosaccharide release ( ). The actual mea-
surement points include error bars, indicating the
inaccuracy in the individual estimations (standard
deviation). B) Monosaccharide release functions for
the polysaccharides pectin LC810 ( ), xanthan
( ), and cellulose ( ).
D.W.H. Merkx et al. Carbohydrate Polymers 179 (2018) 379–385

Table 1
Overview of the Dx and Dxy-values and calculation and validation results for a set of twelve polysaccharides.

Polysaccharide (y) Monosaccharide (x) Dx Dxy Concentration x based on Dx Concentration x based on Dxy Overestimation of Dx RSDra RSDRb
% (w/w) % (w/w) (%) (%) (%)

Cellulose Glc 0.07 0.07 92 92 0 1.8 3.3

Xanthan Glc 0.07 0.06 28 28 1 2.1 3.9
Man 0.10 0.09 20 20 1 3.7 4.3
GlcA 0.19 0.06 12 12 1 6.6 8.0
Locust Bean Gum Gal 0.10 0.07 16 16 1 4.5 8.7
Man 0.10 0.09 58 57 1 2.8 5.9
Guar Gum Gal 0.10 n.a.c 23 – – 4.0 6.7
Man 0.10 n.a.c 40 – – 2.9 3.4
Tara Gum Gal 0.10 0.07 18 18 1 3.7 10.8
Man 0.10 0.09 59 58 1 2.7 5.3
High Acyl Gellan Glc 0.07 0.06 25 24 1 3.9 5.0
Rha 0.09 0.08 10 10 1 4.6 5.3
GlcA 0.19 0.06 5 5 1 9.0 13.3
Low Acyl Gellan Glc 0.07 0.06 30 29 1 3.1 5.2
Rha 0.09 0.08 16 16 1 3.3 4.6
GlcA 0.19 0.06 7 7 1 12.6 18.8
Pectin LC 810 GalA 0.30 0.16 54 45 20 3.0 3.1
Pectin LC 710 GalA 0.30 0.16 53 44 20 5.0 7.8
Pectin LA 210 GalA 0.30 0.16 54 45 20 3.5 4.0
Pectin LM 5 GalA 0.30 0.16 40 34 20 5.1 5.6
Rhamnogalacturonan GalA 0.30 n.a.c 37 – – 5.6 7.7
Rha 0.09 n.a.c 7 – – 2.5 3.2

a
RSDr: Relative standard deviation of repeatability, based on calculations with Dx.
b
RSDR: Relative standard deviation of within-laboratory reproducibility, based on calculations with Dx.
c
n.a.: not analysed.

well-described via a polysaccharide-specific cumulative release func-
tion (fy). Moreover, it does not occur instantaneously at to. Among the
investigated polysaccharides, the release function for pectin was most
distinctive, due to its relatively long lag phase prior to monosaccharide
release. Because of this long lag phase, the effective time that remains
for monosaccharide degradation was relatively short. This should be
accounted for in the quantification when determining the degradation
factor Dx (Eq. (2)). Since Dx is based on the instantaneous release of
monosaccharides at to, and does not consider fy, the end results will be
prone to overestimation. We therefore aimed for an adapted version of
Dx, defined as Dxy (Eq. (5)) to compensate for delayed monosaccharide
release and resultant estimation bias. For a wide range of poly-
saccharides, the Dx and Dxy values and their corresponding mono-
saccharide concentrations were determined (Table 1). Especially for
pectin, the overestimation of Dx is significant (20%), whereas for i.e.
xanthan this overestimation is insignificant (1%). As expected, the
importance of Dxy increases with the instability of monosaccharides (i.e.
GalA and Xyl) and the stability of the polysaccharide (pectin) under
hydrolysis conditions.
3.4. Quantification and validation of pure polysaccharide standards
With NMR, we could identify and quantify the monomeric content
(%w/w) of a series of pure reference polysaccharides. The precision of
the estimations (in terms of r and R) were determined and presented in
Table 1. Overall, the method resulted in a good precision in terms of
relative standard deviation for repeatability (RSDr = 4.1 ± 0.9%) and
within-laboratory reproducibility (RSDR = 6.1 ± 1.4%).
In theory, the sum of the individual monosaccharide levels per
polysaccharide in Table 1 should provide an estimate of the total
polysaccharide recovery. However, since the exact purities of these
polysaccharides were unknown, or not well specified by the suppliers,
the estimated polysaccharide concentrations could not be compared
against their true values. Only for cellulose, we performed a trueness
study because the product purity was specified. The recovery for cel-
lulose was 92%, and neither a proportional nor an off-set bias were
detected when compared to the true value (95% w/w). This indicates
that the method was not only precise, but also accurate. Besides
trueness, we also determined the linear working range and the limit of
quantification. Based on a Lack-of-Fit test (F < Fcritical), the linearity of
the method was tested and proved to be significant over the in-
vestigated range (0–60 mg/mL). Based on the limit of detection, we
determined that a minimum of 0.24 mg cellulose was required for at-
taining recovery estimates with a relative standard deviation of 10%.
3.5. Quantification of polysaccharide content
An essential part for the quantification of polysaccharide mixtures
was the integration of prior knowledge on the individual poly-
saccharides. Hence, for each polysaccharide, a reliable estimate of the
molar monomeric stoichiometry was needed. The molar stoichiometry
for each polysaccharide could directly be derived from its theoretical
structure. In the current experiment, however, the molar stoichiometry
was empirically estimated from polysaccharide reference standards.
We demonstrate the applicability of the mixture analysis by com-
posing binary and ternary mixtures of polysaccharides that commonly
occur in food products, i.e. high acyl gellan, xanthan and locust bean
gum (LBG). This experiment was designed according to a 33 three-level
factorial design with a centre point. Signal integration in the 1H NMR
spectra of the resulting hydrolysates was done manually as well as by
QMTLS fitting. The two different approaches each resulted in a separate
matrix of monosaccharide quantities (Nx). Based on the stoichiometry
matrix ST the individual polysaccharide quantities (Ny) were estimated
via a non-negative least squares regression (Eq. (8)). The estimated
mixture ratios that were obtained from manual signal integration and
QMTLS fitting were plotted in the ternary diagrams in Fig. 5A and B,
respectively. The accuracy of the QMTLS-based procedure was slightly
better than the manual approach as demonstrated by the relative
tighter scatter patterns of the estimated values around the design
points.
For both quantification methods, the trueness was established as no
significant off-set and proportional bias were detected between the
estimated and the theoretical values. This observation indicated that
the non-negative least squares regression approach, with ST as prior
knowledge, is a valid quantitative approach to resolve polysaccharide
mixtures. Correct molar monosaccharide stoichiometries are therefore

384
D.W.H. Merkx et al.
critical for accurate polysaccharide quantifications.
4. Conclusions
A method has been presented for the quantitative assessment of
pure polysaccharides and polysaccharide mixtures. The Saeman NMR
method was tailored towards a convenient method for a wide range of
food polysaccharide systems. The method was repeatable
(RSDr = 4.1 ± 0.9%) and reproducible (RSDR = 6.1 ± 1.4%) with a
high throughput (40–60 samples a day). We employed automated
mixture analysis according to a non-negative least squares analysis in
order to increase the spectral processing speed and decrease operator
errors. This mixture analysis required a priori knowledge on the iden-
tity of the individual polysaccharides as well as on the molar stoi-
chiometries of the monosaccharide constituents. For the mono-
saccharide quantification, an adapted degradation factor was
implemented which accounts for both the monomeric degradation and
the non-instantaneous polysaccharide degradation.
The entire strategy for identifying and quantifying polysaccharides
and polysaccharide mixtures provides a fast and accurate analytical
tool. The combined method demonstrated its potential for a wide ap-
plication range within food analysis.
Acknowledgements
We gratefully acknowledge the expert advice from Adriana
Carvalho de Souza (DSM, Delft) and Christian Grün (Unilever R & D
Vlaardingen). We thank Seenakshi Dauwan (Unilever R & D
Vlaardingen) for carefully performing NMR measurements. We thank
Velitchka Mihaleva (Unilever R & D Vlaardingen) for writing the core of
the Python scripts.
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.carbpol.2017.09.074.
References
Ahmed, A. E. R., & Labavitch, J. M. (1977). A simplified method for accurate determi-
nation of cell wall uronide content. Journal of Food Biochemistry, 1, 361–365.
Bertaud, F., Sundberg, A., & Holmbom, B. (2002). Evaluation of acid methanolysis for
analysis of wood hemicelluloses and pectins. Carbohydrate Polymers, 48, 319–324.
Carvalho de Souza, A., Rietkerk, T., Selin, C. G., & Lankhorst, P. P. (2013). A robust and
universal NMR method for the compositional analysis of polysaccharides.
Carbohydrate Polymers, 95(2), 657–663.
385
Carbohydrate Polymers 179 (2018) 379–385
Fig. 5. Ternary diagrams representing the predicted
mixture ratios of xanthan, LBG and high acyl gellan.
Quantification of the three polysaccharides from the
1
H NMR spectra was based on (A) manual peak in-
tegration, and (B) QMTLS fitting.
Carvalho de Souza, A. (2017). Quantification of food polysaccharides by means of NMR.
In G. A. Webb (Ed.). Modern magnetic resonance (pp. 1–19). Springer International
Publishing.
Coenen, G. J., Bakx, E. J., Verhoef, R. P., Schols, H. A., & Voragen, A. G. J. (2007).
Identification of the connecting linkage between homo- or xylogalacturonan and
rhamnogalacturonan type I. Carbohydrate Polymers, 70(2), 224–235.
De Ruiter, G. A., Schols, H. A., Voragen, A. G. J., & Rombouts, F. M. (1992). Carbohydrate
analysis of water-soluble uranic acid-containing polysaccharides with high-perfor-
mance chromatography using methanolysis combined with TFA hydrolysis is superior
to four other methods. Analytical Biochemistry, 185, 176–185.
Dickinson, E. (2003). Hydrocolloids at interfaces and the influence on the properties of
dispersed systems. Food Hydrocolloids, 17(1), 25–39.
Englyst, H. N., & Cummings, J. H. (1984). Simplified method for the measurement of total
non-starch polysaccharides by gas – liquid chromatography of constituent sugars as
alditol acetates. Analyst, 109(7), 937–942.
Grün, C. H., Sanders, P., van der Burg, M., Schuurbiers, E., van Adrichem, L., van Velzen,
E. J., ... Schols, H. A. (2015). Strategy to identify and quantify polysaccharide gums in
gelled food concentrates. Food Chemistry, 166, 42–49.
Jacobs, D. M., van Velzen, E. J. J., & Mihaleva, V. (2013). Evaluation of approaches for
quantitative targeted profiling of complex compositions using 1D 1H NMR spectroscopy.
Magnetic resonance in food science: Food for thought3–13.
Malz, F., & Jancke, H. (2005). Validation of quantitative NMR. Journal of Pharmaceutical
and Biomedical Analysis, 38(5), 813–823.
Mercier, P., Lewis, M. J., Chang, D., Baker, D., & Wishart, D. S. (2011). Towards auto-
matic metabolomic profiling of high-resolution one-dimensional proton NMR spectra.
Journal of Biomolecular NMR, 49(3), 307–323.
Mihaleva, V. V., Korhonen, S.-P., van Duynhoven, J., Niemitz, M., Vervoort, J., & Jacobs,
D. M. (2014). Automated quantum mechanical total line shape fitting model for
quantitative NMR-based profiling of human serum metabolites. Analytical and
Bioanalytical Chemistry, 406(13), 3091–3102.
Mittal, A., Scott, G. M., Amidon, T. E., Kiemle, D. J., & Stipanovic, A. J. (2009).
Quantitative analysis of sugars in wood hydrolyzates with 1H NMR during the au-
tohydrolysis of hardwoods. Bioresource Technology, 100(24), 6398–6406.
Saeman, J. F., Moore, W. E., Mitchell, R. L., & Millet, M. (1954). Techniques for the
determination of pulp constituents by quantitative paper chromatography. Tappi, 37,
336–343.
Saha, D., & Bhattacharya, S. (2010). Hydrocolloids as thickening and gelling agents in
food: A critical review. Journal of Food Science and Technology, 47(6), 587–597.
Shi, Y., Yokoyama, T., Akiyama, T., Yashiro, M., & Matsumoto, Y. (2012). Degradation
kinetics of monosaccharides in hydrochloric, sulfuric, and sulfurous acid.
BioResources, 7(3), 4085–4097.
Tiainen, M., Soininen, P., & Laatikainen, R. (2014). Quantitative quantum mechanical
spectral analysis (qQMSA) of (1)H NMR spectra of complex mixtures and biofluids.
Journal of Magnetic Resonance, 242, 67–78.
Tojo, E., & Prado, J. (2003). A simple 1H NMR method for the quantification of carra-
geenans in blends. Carbohydrate Polymers, 53(3), 325–329.
van Duynhoven, J., van Velzen, E., & Jacobs, D. M. (2013). Quantification of complex
mixtures by NMR. Annual reports on NMR spectroscopy181–236.
van Velzen, E. J. J., Dauwan, S., de Roo, N., Grun, C. H., Westphal, Y., & van Duynhoven,
J. P. M. (2015). Quantitative NMR assessment of polysaccharides in complex food ma-
trices. Magnetic resonance in food science: Defining food by magnetic resonance. The
Royal Society of Chemistry39–48.
Wang, S., Li, C., Copeland, L., Niu, Q., & Wang, S. (2015). Starch retrogradation: A
comprehensive review. Comprehensive Reviews in Food Science and Food Safety, 14(5),
568–585.
Wishart, D. S. (2008). Quantitative metabolomics using NMR. TrAC Trends in Analytical
Chemistry, 228–237.