J. Agric. Food Chem.

1999, 47, 1549−1557 1549

Analysis of Food Oligosaccharides Using MALDI-MS: Quantification

of Fructooligosaccharides
Jian Wang,† Peter Sporns,*,† and Nicholas H. Low‡
Department of Agricultural, Food and Nutritional Science, University of Alberta,
Edmonton, Alberta, Canada T6G 2P5, and Department of Applied Microbiology and Food Science,
University of Saskatchewan, Saskatoon, Saskatchewan, Canada S7N 5A8

Matrix-assisted laser desorption/ionization mass spectrometry (MALDI-MS) is a powerful new

technique that will have a great impact on food analysis. This study demonstrates the applicability
of MALDI-MS performed directly on an aqueous food extract for qualitative and quantitative analysis
of food oligosaccharides. 2′,4′,6′-Trihydroxyacetophenone was found to be the best matrix for analysis
of oligosaccharides in the foods examined. The relationship between laser strength, resolution, and
the response factors of individual oligosaccharides using MALDI-MS was investigated. A MALDI-
MS method for quantitative analysis of fructooligosaccharides with standard addition of a pure
fructooligosaccharide was developed. High performance anion exchange chromatography with pulsed
amperometric detection was compared to MALDI-MS for the analysis of fructooligosaccharides. The
fructooligosaccharide analyses were performed on red onions, shallots, and elephant garlic.

Keywords: Cyclodextrin; maltohexaose; fructan; inulin; kestose; nystose

INTRODUCTION (HPLC). However, the procedure is also lengthy and

gives minimal information about which polymers are
Fructans are nonreducing water-soluble carbohy- present. Loo et al., 1995, describe a number of analytical
drates formed in higher plants composed of fructosyl procedures for determination of fructooligosaccharides
units but usually containing one terminal glucose including HPLC for low DP fructooligosaccharides and
moiety per molecule. They occur as linear, branched, a gas chromatography (GC) procedure. While the latter
or, less frequently, cyclic molecules (Darbyshire and can be used for sensitive determination of fructo-
Henry, 1978, 1981; Bancal and Gaudillère, 1991; Stahl oligosaccharides up to about DP of 12, extensive puri-
et al., 1997). Natural β-fructans have a degree of fication and derivatization with water sensitive reagents
polymerization (DP) ranging from 2 to 55 or more. are required. In addition GC conditions of very high
Lower mass (DP of 2-20) fructans are also called temperatures are required to volatilize the derivatized
fructooligosaccharides (Figure 1), while higher mass fructooligosaccharides. By far the most used procedure
polymers are known as inulin. It has been found that for analysis of fructooligosaccharides is high-perfor-
these nondigestible carbohydrates are effective in im- mance anion-exchange chromatography with pulsed
proving intestinal flora and increasing calcium and amperometric detection (HPAE-PAD) (Chatterton et al.,
magnesium absorption (Ohta et al., 1995, 1998). Fruc- 1989, 1993; Shiomi et al., 1991; Timmermans et al.,
tans have also been examined for their biological role 1994; Loo et al., 1995). The fructooligosaccharide re-
in plant osmoregulation, adaptation to low-temperature sponse with HPAE-PAD does vary (Timmermans et al.,
photosynthesis, protection from freezing stress (Darby- 1994) and the analyses often require significant sample
shire and Henry, 1978; Pollock, 1984; Nelson and Smith, purification.
1986; Chatterton et al., 1990; Livingston, 1990), and
Matrix-assisted laser desorption/ionization mass spec-
storage life of bulbs (Darbyshire and Henry, 1981;
trometry (MALDI-MS) was originally developed for
Suzuki and Cutcliffe, 1989).
measuring the mass of large molecules such as proteins.
Traditionally a variety of techniques have been used MALDI-MS has also been applied to carbohydrates since
to analyze for fructooligosaccharides. Gel permeation about 1991 (Mock et al., 1991; Stahl et al., 1991).
chromatography (Darbyshire and Henry, 1978) has been Fructooligosaccharides or inulin in some plants have
used, but detection and identification of the separated been qualitatively analyzed by using MALDI-MS (Metz-
fructooligosaccharides requires extensive additional ger et al., 1994; Stahl et al., 1997; Losso and Nakai,
methodology involving acid hydrolysis and various 1997) but there have been no reports using quantitative
enzymatic and colorimetric procedures to identify the analysis. Our purpose in this study was to develop
carbohydrates. The method by Manghi et al., 1995, methodology for both qualitative and quantitative analy-
detects fructooligosaccharides after various enzyme sis of fructooligosaccharides in selected food samples.
treatments and identification of the carbohydrates
produced by high performance liquid chromatography
MATERIALS AND METHODS

* Phone (780) 492-0375; fax (780) 492-4265; e-mail psporns@ Materials and Reagents. Red onion bulbs (Allium cepa
afns.ualberta.ca. L.), shallot bulbs (A. cepa L. var. ascalonicum), and elephant
† University of Alberta. garlic (Allium ampeloprasum) were purchased from local
‡ University of Saskatchewan. markets in Edmonton, Alberta, Canada. Inulin from Jerusalem

10.1021/jf9809380 CCC: $18.00 © 1999 American Chemical Society

Published on Web 03/10/1999
1550 J. Agric. Food Chem., Vol. 47, No. 4, 1999 Wang et al.

Figure 1. Chemical structures for fructooligosaccharides, maltohexaose, and γ-cyclodextrin.

artichokes, γ-cyclodextrin, maltohexaose, 4-hydroxy-R-cyano- cocrystallized with matrixes on the probe were ionized by using
cinnamic acid (HCCA), and sinapinic acid were purchased from a nitrogen laser pulse (337 nm) and accelerated under 20 kV
Sigma Chem. Co. (St. Louis, MO). A mixture of 1-kestose (GF2, by using pulsed ion extraction before entering the time-of-flight
DP ) 3), nystose (GF3, DP ) 4), and β-fructofuranosylnystose mass spectrometer. The preparation of matrixes and samples
(GF4, DP ) 5) (34%, 53%, and 10%) and these individual is shown in Table 1. Laser strength was selected to obtain
fructooligosaccharides in pure form were a gift from Dr. A. the best signal-to-noise ratios. The number of laser pulses
Ohta, Nutritional Science Center, Meiji Seika Kaisha, Japan. collected was determined as needed to obtain good responses
2′,4′,6′-Trihydroxyacetophenone monohydrate (THAP), [2-(4- of all oligosaccharides.
hydroxyphenylazo)benzoic acid] (HABA), 3-aminoquinoline (3- Quantification of fructooligosaccharides using MALDI-MS
AQ), 1-hydroxyisoquinoline, and 2,5-dihydroxybenzoic acid was achieved using standard addition. The frozen sample
(DHB) were obtained from Aldrich Chem. Co. (Milwaukee, WI). extracts were allowed to warm to room temperature. Samples
Extraction of Fructooligosaccharides from Onion, were prepared by taking 50 µL of extract and mixing this with
Shallot, and Garlic Samples. Fresh samples were peeled 50 µL of aqueous 0.01 M potassium chloride solution. The
to remove the dry outer layers and then chopped using a food standard addition samples had GF4 (7.9 × 10-4 M in 50 µL
processor (Braun, UK 100, Type 4259, Germany) for 5 min. aqueous 0.01 M potassium chloride) added to 50-µL sample
The mixed samples were freeze-dried. Each freeze-dried extracts. Samples and standard addition samples were each
sample (1 g) was extracted with two portions of 40-mL double spotted in five separate positions on the probe. A single
deionized water heated to reflux for 1 h. The cooled sample spectrum was then generated for each position on the probe
was centrifuged for 15 min at 10 000 rpm after each water (10 spectra in all, five for each sample and five for each
extraction; the supernatants were combined and made up to standard addition sample) by random selection of three
100 mL with water. The aqueous extract was then filtered with different spots for each probe position and collecting 60 laser
a Millipore HA 0.45-µm membrane (Chromatography Division/ pulses for each spot. That is, one spectrum represented the
Millipore Corp., Milford, MA) and the extract kept frozen at sum of 3 × 60 or 180 laser pulses. Peak heights (for potassium
-20 °C until needed. adducts) were determined for each fructooligosaccharide from
Moisture Content. Sample moisture content was deter- each spectrum. These peak heights were then scaled relative
mined according to AOAC Official Method of Analysis (1990) to the GF3 peak, which was arbitrarily set at a value of 1.0.
by using a vacuum oven (National Appliance Co.) overnight Each of the five spectra for the sample was compared to a
in vacuo at 70 °C. different spectrum from the five standard addition samples
MALDI-MS. MALDI-MS was performed with the Proflex and the average increase in scaled relative peak height for
III, Bruker Analytical Systems Inc. (Billerica, MA). Analytes GF4 (standard fructooligosaccharide added) determined. This
MALDI-MS of Fructooligosaccharides J. Agric. Food Chem., Vol. 47, No. 4, 1999 1551

Table 1. Performance of Matrixes for Desorption and Ionization of Maltohexaose and γ-Cyclodextrin
matrix preparation of spectra repeat- matrix analyte
matrix references concentration matrix and sample quality ability peaks molar ratio
2,5-dihydroxybenzoic Bruker, 1995; 1. 12.3 mg/mL in mix matrix and sample excellent good medium normal
acid (DHB) Mohr et al., 1995; ethanol:water ) 1:1 together in a ratio
Losso et al., 1997 of 1:1
2. 10.1 mg/mL in double as above good good medium normal
deionized water
2,4,6-trihydroxyaceto- Pieles et al., 1993 1. saturated in acetone put matrix on the probe excellent excellent few normal
phenone mono- first and then sample
hydrate (THAP) on top of matrix
2. 12.5 mg/mL in mix matrix and sample excellent excellent few normal
acetonitrile:water together in a ratio
) 1:1 of 1:1
3-aminoquinoline Metzger et al., 1994; 10.1 mg/mL in 10% as above good poor few normal
Stahl et al., 1997 ethanol
4-hydroxy-R-cyano- Bartsch et al., 1996 13.3 mg/mL in 50% as above good poor medium normal
cinnamic acid ethanol or matrix
(HCCA) saturated in ethanol
sinapinic acid Bruker, 1995 14.2 mg/mL in as above no no medium
acetonitrile:water
) 1:1
2-(4-hydroxyphenylazo)- Harvey et al., 1994 saturated in acetone put matrix on the probe poor good lots abnormal
benzoic acid (HABA) first and then sample
on top of matrix
2,5-dihydroxybenzoic Mohr et al., 1995 0.2 M DHB/0.6 M HIC mix matrix and sample good poor few normal
acid (DHB)/1-hydroxy- in acetonitrile:water together in a ratio
isoquinoline (HIC) ) 1:1 of 1:1

gave a value for the average increased response due to the RESULTS AND DISCUSSION
addition of GF4. This response factor was then used to
determine the average amount of each fructooligosaccharide For quantification of carbohydrates using MALDI-MS,
in the five sample spectra. The acquisition of the MALDI-MS several factors must be examined individually, including
data took about 20 min and the calculations necessary were the selection of matrixes, matrix and sample prepara-
rapidly carried out by using Microsoft Excel 97. All samples tion, and the selection of an appropriate internal
were analyzed in duplicate. That is, an entirely new 10 spots standard for quantification (Harvey, 1993; Jespersen et
(five samples and five standard addition samples) were al., 1995; Gusev et al., 1995; Abell and Sporns, 1996;
analyzed. Bartsch et al., 1996; Wilkinson et al., 1997). The best
HPAE-PAD. The frozen red onion, shallot, and garlic matrix offers spot-to-spot and sample-to-sample repeat-
sample extracts were allowed to come to room temperature ability and reproducibility, which makes quantitative
and diluted 1:5 (in 5-mL volumetric flask) with water. Fructo- analysis of the analytes of interest possible.
oligosaccharide standards (10.0 mg) were prepared in a 50- Selection of Matrixes. The main problems associ-
mL volumetric flask with HPLC grade water. Each sample
ated with matrix-assisted laser desorption/ionization
was passed through a 0.2-µm syringe filter (25 mm; Chro-
(MALDI) quantitative analysis are poor shot-to-shot
matographic Specialties, Brockville, ON). Filtered samples
were analyzed on a Waters 625 metal free gradient HPLC
repeatability or crystal inhomogeneity (Gusev et al.,
(Waters Chromatography, Milford, MA). All samples and 1995). Proper homogeneous crystallization over the
standards were injected (50 µL) with a Waters 712 Wisp entire probe area and a homogeneous embedding of
autosampler. Carbohydrates were separated on a Carbo Pac analyte molecules in the matrix are the prime criteria
PA1 column (250 × 4 mm; Dionex, Sunnyvale, CA) coupled for high repeatability and quantification. The selection
with a Carbo Pac PA1 guard column (50 × 4 mm). The solvents of matrixes is usually based on a comparison of spot-
used were 100 mM sodium hydroxide (solvent A), 100 mM to-spot or sample-to-sample repeatability and the ability
sodium hydroxide/400 mM sodium acetate (solvent B), and 300 to obtain a good quality spectrum with reasonable
mM sodium hydroxide (solvent C). The mobile phase flow rate signal-to-noise ratios with the best possible resolution.
was maintained at 1 mL/min, with a linear gradient profile 2,5-Dihydroxybenzoic acid (DHB) (Bruker, 1995; Mohr
consisting of solvent A with the following proportions (v/v) of et al., 1995; Losso and Nakai, 1997), 3-aminoquinoline
solvent B or C: 0-8 min, maintain 0% B and 0% C; 8-60 min, (3-AQ) (Metzger et al., 1994), 4-hydroxy-R-cyanocin-
100% B; 60-61 min, 100% C; 61-90 min, maintain 100% C; namic acid (HCCA) (Bartsch et al., 1996), and 2,5-
90-91 min, 0% B and 0% C. Sodium hydroxide (300 mM) was dihydroxybenzoic acid (DHB)/1-hydroxyisoquinoline
added postcolumn (Waters Chromatography) at a flow rate of (HIC) (Mohr et al., 1995) have all been recommended
0.70 mL/min to minimize baseline drift. Detection was achieved as matrixes for carbohydrate analysis using MALDI-
employing a Waters 464 pulsed amperometric detector (PAD)
MS. These matrixes were tested for their suitability to
with a dual gold electrode and triple pulsed amperometry at
ionize the oligosaccharides, maltohexaose and γ-cyclo-
a sensitivity of 50 µA. The electrode was maintained at the
following potentials and durations: E1 ) 0.05 V (T1 ) 0.2995);
dextrin (Figure 1). The preparation of matrixes and
E2 ) 0.60 V (T2 ) 0.2995); E3 ) -0.80 V (T3 ) 0.4995). their performance are listed in Table 1. DHB, 3-AQ,
Chromatograms were plotted employing Millenium 2010 chro- HCCA, and DHB/HIC matrixes all gave good quality
matography manager software (Waters Chromatography). All spectra (Figure 2). The repeatability with DHB was
samples were analyzed in duplicate. Quantification of fructo- acceptable, whereas the matrixes 3-AQ, HCCA, and
oligosaccharides was determined from peak areas using su- DHB/HIC could not meet the need for high repeat-
crose as the external standard (the peak area of sucrose was ability. DHB showed many matrix peaks in the low
used as the reference value to calculate all other response mass region, which could interfere with low molecular
factors as in Timmermans et al., 1994). weight analytes of interest, such as kestose with a mass
1552 J. Agric. Food Chem., Vol. 47, No. 4, 1999 Wang et al.

Figure 3. MALDI-MS positive ion spectrum of inulin from

Figure 2. MALDI-MS positive ion spectra of γ-cyclodextrin Jerusalem artichokes. Inulin was dissolved in double deionized
and maltohexaose in various matrixes. A: 4-Hydroxy-R- water to give a final concentration of 4 mg/mL. A 0.3 µL of
cyanocinnamic acid, 13.3 mg/mL in ethanol:water (1:1). A 20 saturated THAP in acetone was first placed on the probe and
µL of sample mixture containing γ-cyclodextrin (2.5 × 10-5 M, a 0.5-µL inulin sample was put on top of the crystallized matrix
marked M2) and maltohexaose (1.1 × 10-4 M, marked M1) in to dry. Sixty laser pulses at an attenuation of 22 were
double deionized water was mixed with 20 µL of matrix in accumulated for the final spectrum.
solution and vortexed for 30 s. Then 0.5-µL mixture of matrix
and sample was applied to the probe. Laser strength was set
at an attenuation of 44. Twenty shots were accumulated for (attenuation of 44 or 49) was used for HCCA or HABA
the final spectrum. B: 2,5-Dihydroxybenzoic acid, 12.3 mg/ to desorb and ionize maltohexaose and γ-cyclodextrin,
mL in ethanol:water (1:1). Other parameters were the same both of these matrixes showed numerous matrix peaks
as in A except the laser strength was set at an attenuation of making analysis of masses below 600 difficult. Further-
32. C: 2,4,6-Trihydroxyacetophenone monohydrate saturated more, the peak ratio (0.30) of maltohexaose to γ-cyclo-
in acetone. A 0.3-µL aliquot of matrix was applied to the probe
first and air-dried. Then, 0.5 µL of sample was put on top of
dextrin obtained from HABA was quite far from the
the matrix. Other parameters were the same as in B. D: actual molar ratio (0.63).
3-Aminoquinoline (10.1 mg/mL in 10% ethanol). Other param- The rate of the evaporation of the solvent affects the
eters were the same as in B. cocrystallization of matrix and sample. Fast evaporation
leads to fine crystals and more homogeneous incorpora-
of 504. Metzger et al. (1994) first introduced 3-AQ as a tion of sample. Improvement in sample homogeneity
matrix for inulin using MALDI-MS. Compared to DHB, using the fast-evaporation method enhanced both shot-
3-AQ showed sharper peaks (that is better resolution) to-shot repeatability and sample-to-sample reproduc-
and a lower background, but the high quality spectra ibility (Nicola et al., 1995). Fast evaporation could be
could not be repeatedly obtained in our experiments. enhanced with THAP using acetone as the solvent.
Mohr et al. (1995) noted that DHB crystals formed only THAP is very soluble in acetone which then evaporates
near the rim of the probe, complicating the location of rapidly giving small homogeneous crystals. Previously,
suitable laser ionization positions because in the center THAP was successfully used as a matrix for peptides
of the probe only a few crystals could be found. Stahl et and oligonucleotides (Kussmann et al., 1997, and Pieles
al. (1997) found 3-AQ to be more sensitive to contami- et al., 1993). However, Mohr et al. (1995) indicated that
nants, such as salts. Naven et al. (1997) found the THAP was not as good a matrix for carbohydrates such
spectra acquired from DHB exhibited the most abun- as DHB/HIC due to irregular crystallization and relative
dant fragmentation and that those using 3-AQ exhibited signal-to-noise ratios. However, they used water as
the least. Mohr et al. (1995) pointed out that DHB/HIC solvent for this matrix. With water both the lower
crystallized equally over the entire probe as a fine solubility of THAP and slower evaporation rate likely
powder using vacuum-drying for a few seconds. This lead to the noted problems. To get an appropriate excess
resulted in high quality spectra with few matrix peaks of matrix to analyte the THAP was first crystallized
and intense analyte peaks. However, this procedure was from acetone and then the aqueous sample applied on
not suitable for a multiple-position probe, because top of the formed crystals. The water redissolved some
samples cannot be spotted and vacuum-dried simulta- of the matrix and the remaining undissolved THAP
neously so irregular crystals are formed, leading to poor acted as seed crystals for rapid recrystallization of
spot-to-spot repeatability. analyte and matrix as the water evaporated. This
The laser strength used for the matrixes 3-AQ, DHB, technique resulted in high quality MALDI-MS spectra
and DHB/HIC was almost the same, with an attenua- (Figure 2) with high spot-to-spot repeatability. The
tion around 30-33, which was just above laser strength technique can be used to resolve the oligosaccharides
threshold values required to desorb and ionize analytes. in inulin up to a mass of 9000 (DP of about 55, Figure
(Note that for the MALDI-MS Proflex III, attenuation 3). Another advantage of this technique was its toler-
is opposite to laser strength; that is, the higher the ance to small amounts of protein or other impurities in
attenuation the lower the laser strength.) At the same samples, with few interfering matrix peaks from THAP
laser strength, DHB produced more matrix peaks than and reasonable signal-to-noise ratio at an attenuation
3-AQ or DHB/HIC. Whereas a much lower laser strength of between 30 and 32. Therefore THAP with acetone
MALDI-MS of Fructooligosaccharides J. Agric. Food Chem., Vol. 47, No. 4, 1999 1553

Figure 5. Relationship between laser strength and resolution.

The sample was prepared by first applying 0.3 µL of saturated
2,4,6-trihydroxyacetophonenone monohydrate saturated in
acetone, air-drying, and then applying 0.5 µL of maltohexaose
(1.1 × 10-4 M) as described in Figure 4. The laser strength
was changed from an attenuation of 34 down to 20 (x axis). At
each attenuation, two spots were randomly chosen to collect
two spectra (resolution shown for each spectrum) with 20 laser
pulses totaled for each spot. The resolution of the molecular
ion peaks was obtained from sodium adduct ion peak of
maltohexaose (y axis).
Figure 4. MALDI-MS positive ion spectra of fructooligosac-
charides from shallots. The sample was prepared by first
applying 0.3 µL of saturated THAP in acetone on the probe, Generally, laser strength has been chosen based on the
air-drying, and then applying a 0.5 µL aqueous sample solution signal-to-noise ratios (Bartsch et al., 1996; Naven et al.,
followed by further air-drying. Laser strength was set at 1997). However, the relationships between laser strength,
attenuation of 31 and 180 laser pulses were accumulated in resolution, and the analyte response have not been
three random positions for the final spectrum.
thoroughly investigated for carbohydrates. We found
prepared in a two step procedure was chosen as the that laser strength played a very important role in
matrix for later studies. obtaining quality spectra. As the laser strength in-
Alkali-Metal Adducts. In general, carbohydrates creased over a certain amount, the resolution deterio-
ionize in a MALDI-MS source only after cationization rated rapidly. Figure 5 indicates the trend of the
with alkali ions (Börnsen et al., 1995). For quantification resolution of maltohexaose sodium adduct peak with the
it was desirable that the investigated carbohydrate variation in laser strength. At the attenuation between
sample contained predominantly one kind of alkali 30 and 33, high quality spectra were obtained with well
metal, resulting in a single molecular ion peak. With resolved isotopic mass peaks having resolutions close
no modification the matrix and sample contain both to 3000 (full width at half-maximum, fwhm), and exact
sodium and potassium ions (Figure 2), resulting in isotopic mass could be determined with an accuracy 100
multiple carbohydrate peaks. The peak intensity of ppm or less by using internal calibration. However,
carbohydrate alkali-metal ion adducts in an unmodified when the attenuation was decreased to 27, the resolu-
sample is dependent on the concentration of the alkali- tion deteriorated rapidly to 500 (fwhm). This effect can
metal ions in final solution applied to the probe and the be easily seen in spectra since isotopic resolution is lost.
affinity between the metal and the carbohydrate. It has These isotopic peaks include the main peak plus one
been shown that the affinity of alkali metals to carbo- unit mass (M + 1) and M + 2 peaks due mainly to 13C
hydrates follows the order of H < Li < Na < K < Cs isotopes. At the same time, because of the loss of isotopic
(Mohr et al., 1995; Börnsen et al., 1995). Ion exchange resolution, molecular ion peaks became broad and the
and purification of carbohydrates on a Nafion mem- measured masses of peaks were shifted to high masses
brane has been successfully used as a sample pretreat- with increasing laser strength (data not shown). More
ment for MALDI-MS producing a single alkali ion important, the molar peak ratio between maltohexaose
adduct (Börnsen et al., 1995). However, there is another and γ-cyclodextrin changed, making quantitative analy-
simpler method to obtain a single alkali ion adduct peak. sis impossible using one as the internal standard for
By dissolving carbohydrates in a 0.01 M solution of the
the other, even though chemically these two molecules
alkali ion salt (e.g., potassium chloride) we were able
are very similar oligosaccharides. In Figure 6 the actual
obtain a single alkali ion adduct peak. The concentration
molar ratio of maltohexaose to γ-cyclodextrin was 4.4
of alkali ions was crucial, since too high a concentration
of salts would also suppress the molecular ions. Often and within the region of attenuation between 28 and
food samples, such as onions, shallots, and garlic, 34, the peak ratio of maltohexaose and γ-cyclodextrin
naturally contain a high concentration of potassium was very close to this molar ratio (4.1 ( 0.53) even
ions, and could be analyzed without further addition of though both sodium and potassium adduct peaks were
salts. The molecular ions seen in MALDI-MS for these used. However, with the increase in laser strength, the
food samples were almost entirely the potassium ad- peak ratio decreased. This indicated that relatively more
ducts (Figure 4). γ-cyclodextrin was desorbed and ionized during the
Laser Strength. Laser strength determines the ionization than maltohexaose. It can be concluded that
degree of the desorption and ionization of analytes in the behavior of molecules, even when they are only
MALDI-MS. Usually, with increases of laser strength, slightly different in their molecular structure, can be
more ions, including both matrix and molecular ions, quite different during the ionization process. While
are generated. Also higher laser strengths can lead to higher laser strength does result in a significant in-
more fragmentation. For a high quality spectrum and crease in ions formed, another advantage of lower laser
quantitation, the ideal laser strength is very important. strength is to limit any molecular fragmentation.
1554 J. Agric. Food Chem., Vol. 47, No. 4, 1999 Wang et al.

corresponding DP fructooligosaccharide because of its

cyclic structure).
Analysis of Fructooligosaccharides in Food
Samples. When γ-cyclodextrin was added as an inter-
nal standard to extracted food samples, the relative
responses noted above (about a 2:1 molar ratio) for pure
standards changed. It became obvious that in the
different food extract environments γ-cyclodextrin re-
sponded differently than fructooligosaccharides. Food
extracts from red onions, for example, completely sup-
pressed the production of ions from added γ-cyclodex-
Figure 6. Relationship between laser strength and analyte trin, even though the fructooligosaccharides could be
peak ratios. The sample, preparation of matrix and sample,
and other MALDI-MS parameters are the same as described seen. For this reason γ-cyclodextrin was abandoned as
in Figure 5. γ-Cyclodextrin was present at a concentration of an internal standard.
2.5 × 10-5 M. Total peak height of both sodium and potassium While working with food extracts, however, one
adduct peaks was used to plot this trend. feature seemed very consistent and that was the relative
ratios of the individual fructooligosaccharides. For this
reason it was decided to attempt the use of a single
purified fructoligosaccharide in a standard addition
method to quantitate all of the fructooligosaccharides.
Table 2 shows the results for standard addition using
nystose (GF3) as reference peak (that is all other peak
heights are compared to the peak height of this peak)
and GF4 as the standard added. Comparisons were
carried out for two samples of known concentration, a
supplied oligosaccharide mixture with known composi-
tion (standard A) and another mixture prepared from
Figure 7. Responses of individual fructooligosaccharides in pure fructooligosaccharide standards (standard B). Even
MALDI-MS. Diamond shapes: kestose (slope ) 2.1, R2 ) 0.93). though differing amounts of internal standard (GF4
Squares: nystose (slope ) 2.3, R2 ) 0.99). Triangles: GF4 added at both 0.72 and 0.33 mg/mL levels) were added,
(slope ) 2.1, R2 ) 0.98). Internal standard γ-cyclodextrin (4.0
× 10-5 M) was dissolved in 0.01 M potassium chloride solution. quantitation of all three oligosaccharide compounds
The concentration of individual fructooligosaccharides ranged compared nicely with known values.
from 1.0 × 10-4 to 1.0 × 10-3 M for kestose, 8.2 × 10-5 to 8.2 Of course the ultimate test was to examine food
× 10-4 M for nystose and 6.7 × 10-5 to 6.7 × 10-4 M for GF4 extracts. Table 3 shows the MALDI-MS results for
in 0.01 M aqueous potassium chloride. Other MALDI-MS determining the fructooligosaccharide content in red
parameters were the same as in Figure 4. Each data point
was the mean of three random positions for a total of 180 laser onions, shallots, and garlic analyzed by using the
pulses. Each spectrum from a single sample position was standard addition method.
collected from three random spots for a total of three of the Comparison of MALDI-MS and HPAE-PAD Re-
180 laser pulses. Peak heights were used as for quantification. sults. High performance anion exchange chromatogra-
Each error bar stands for the standard deviation from the
mean of three different spectra.
phy with pulsed amperometric detection (HPAE-PAD)
has been used to quantitatively analyze fructooligo-
saccharides or inulin (Chatterton et al., 1989, 1993;
Oligosaccharide Response in MALDI-MS. The Timmermans et al., 1994). Table 3 shows the result of
response of the analytes in MALDI-MS plays a very analysis for fructooligosaccharides contained in onion
important role in quantifying analytes of interest. In and shallot with both HPAE-PAD and MALDI-MS and
theory, the intensity or response of an analyte should garlic with MALDI-MS. In general, the sensitivity of
be linearly correlated to its molar ratios in the MALDI- PAD detector decreases rapidly from DP ) 2 to DP )
MS sample. Figure 7 indicates the linearity between the 6, while for longer oligomers (DP ) 7-17), the sensitiv-
concentration of individual fructooligosaccharides and ity of detector decreases only slightly (Timmermans et
their response in MALDI-MS using γ-cyclodextrin as an al., 1994). The calculation of fructooligosaccharides in
internal standard. The slopes or the relative response food using HPAE-PAD was based on the response
factors of the fructooligosaccharides using MALDI-MS factors and linear relationship reported by Timmermans
were 2.1 for kestose, 2.1 for nystose, and 2.3 for GF4. et al. (1994). Peak areas were integrated and compared
This indicated that the response of the individual to an external standard, sucrose (Figure 10) for quan-
fructooligosaccharides in MALDI-MS was very similar tification. Onion bulbs contain various isomeric fructans
and was more than twice that of γ-cyclodextrin. Al- (Darbyshire et al., 1981; Bancal et al., 1989, 1991; Stahl
though the responses of γ-cyclodextrin and the fructo- et al., 1997). These isomer ions yielded a more complex
oligosaccharides were different on a molar basis, the chromatogram pattern than for MALDI-MS (Figure 8
molar ratio of fructooligosaccharides to γ-cyclodextrin is the HPAE-PAD chromatogram of shallots, while
seemed to be consistent, making it possible to use Figure 4 is the MALDI-MS spectrum for this sample).
γ-cyclodextrin as an internal standard. γ-Cyclodextrin The retention times of individual peaks from both red
was a useful internal standard for fructooligosaccha- onions and shallots were very comparable. Kestose (DP
rides since it is readily available in pure form, was likely ) 3), nystose (DP ) 4), and GF4 (DP ) 5) in red onion
similar in chemical stability to fructooligosaccharides and shallot samples were determined in comparison to
since all compounds are nonreducing sugars, and had the retention times of fructooligosaccharide standards.
a unique mass that would not overlap with any other With higher DP fructooligosaccharides, because of the
fructooligosaccharide (18 mass units less than the lack of individual standards, we assumed that adjacent
MALDI-MS of Fructooligosaccharides J. Agric. Food Chem., Vol. 47, No. 4, 1999 1555

Table 2. Repeatability of MALDI-MS Analysis Data

fructooligosaccharides standard Aa fructooligosaccharides standard Bb
concn of GF4 (DP ) 5) degree of actual value, MALDI-MS actual value, MALDI-MS
added standard, mg/mL polymerization mg/mL datac mg/mL datac
0.72 DP ) 3 (kestose) 0.23 0.21 (0.01 0.37 0.41 (0.09)
DP ) 4d (nystose) 0.35 0.40 0.46 0.50
DP ) 5 (GF4) 0.070 0.072 (0.01) 0.37 0.36 (0.03)
total 0.67 0.68 (0.01) 1.20 1.26 (0.07)
0.33 DP ) 3 (kestose) 0.23 0.20 (0.03) 0.37 0.36 (0.04)
DP ) 4d (nystose) 0.35 0.37 0.46 0.50
DP ) 5 (GF4) 0.070 0.073 (0.01) 0.37 0.38 (0.04)
total 0.67 0.65 (0.03) 1.20 1.23 (0.02)
a A mixture of 1-kestose (34%), nystose (53%), and GF (10%) made in Nutritional Science Center, Meiji Seika Kaisha, Japan. b A
4
mixture of 1-kestose, nystose, and GF4 prepared in our experiment. c Mean of a duplicate with five replicates, each replicate from three
of 60 laser pulses and standard deviations (n ) 2). The peak of fructooligosaccharide with DP ) 4 (GF3) was taken as the reference
d

peak.

Table 3. Fructooligosaccharide Content Using HPAE-PAD and MALDI-MS

red onions, mg/g fresh; shallots, mg/g fresh; garlic, mg/g fresh;
water content ) water content ) water content )
fructooligosaccharide 88.3 (0.063) (fresh), 84.2 (0.15) (fresh), 63.1 (0.85) (fresh),
standard, % 5.2 (0.58)a (freeze-dried) 5.9 (0.22)a (freeze-dried) 2.1 (0.25)a (fresh)
deg of actual HPAE-PAD MALDI-MS HPAE-PAD MALDI-MS HPAE-PAD MALDI-MS MALDI-MS
polymn value meanb meanc meanb meanc meanb meanc meanc
3 34 37.06 (0.31) 30.83 (0.36) 3.92 (0.17) 6.81 (0.41) 8.43 (0.06) 14.88 (0.33) 14.69 (3.14)
4 53 49.98 (0.21) 58.01 (0.66) 3.99 (0.00) 3.88 (0.15) 8.27 (0.03) 11.37 (0.15) 10.44 (2.02)
5 10 12.96 (0.53) 11.24 (0.94) 2.67 (0.01) 2.19 (0.15) 8.15 (0.09) 7.69 (0.47) 9.34 (1.87)
6 3.42 (0.02) 1.29 (0.16) 10.86 (0.08) 4.84 (0.53) 8.67 (1.78)
7 1.96 (0.06) 0.82 (0.18) 7.86 (0.52) 3.24 (0.48) 7.92 (1.53)
8 1.25 (0.02) 0.52 (0.11) 5.73 (0.08) 2.07 (0.19) 6.81 (1.14)
9 0.96 (0.05) 0.43 (0.15) 4.37 (0.10) 1.43 (0.13) 5.55 (0.89)
10 0.40 (0.01) 2.73 (0.03) 1.01 (0.05) 4.64 (0.61)
11 0.30 (0.02) 2.02 (0.01) 0.87 (0.07) 4.07 (0.54)
12 0.18 (0.03) 1.33 (0.00) 0.90 (0.00) 3.54 (0.53)
13 0.78 (0.04) 3.11 (0.30)
14 2.80 (0.43)
15 2.37 (0.23)
16 2.14 (0.29)
17 1.87 (0.12)
18 1.64 (0.05)
19 1.53 (0.15)
total 19.06 (0.13) 15.93 (1.32) 60.55 (0.71) 48.3 (1.52) 91.74 (15.56)
a Numbers in parentheses indicate the standard deviation of triplicates (n ) 3). b Mean of a duplicate and standard deviations (n ) 2).
c Mean of a duplicate (same extract) with five replicates, each replicate from three of 60 laser pulses and standard deviations (n ) 2).

peaks (Figure 8) were isomers to calculate the amount fructooligosaccharides in onions, shallots, and garlic
of individual fructooligosaccharide content. The HPAE- during storage.
PAD technique was more sensitive in terms of detection The fructooligosaccharides in the garlic sample, how-
limit than MALDI-MS. The small amounts of higher DP ever, could not be analyzed by using HPAE-PAD be-
fructooligosaccharides in onion or shallot samples were cause of problems with baseline drift, but they could be
detected by HPAE-PAD (Table 3) and could only be seen analyzed by using MALDI-MS with a slightly higher
with MALDI-MS by using higher laser strength where laser strength at an attenuation of 28 or 29. We are
resolution and therefore quantitation suffered. uncertain if the lack of sample purity affected the
Both Loo et al. (1995) and Stahl et al. (1997) stated quantification of the other HPAE-PAD samples. It was
that for onion bulbs the major fructooligosaccharide had clear that the best correlation between the two analysis
a DP ) 5. However, using either analytical method, methods was for the pure fructooligosaccharide stand-
fructooligosaccharides other than DP ) 5 were the ard mixture.
major fructooligosaccharides in our onion sample. The While there are obvious differences in the quantita-
distribution of fructooligosaccharides in red onions, tion of fructooligosaccharides using HPAE-PAD and
shallots, and garlic seen by MALDI-MS followed a MALDI-MS, we feel that the MALDI-MS results more
definite pattern and the amounts of individual fructo- accurately reflect the true amounts of individual fructo-
oligosaccharides could be correlated to different expo- oligosaccharides in these food samples. As has already
nential distributions (correlation coefficients ranged been noted the response of a pulsed amperometric
from 0.97 to 99, Figure 9). The natural symmetry of detector (PAD) is different for different DP fructooligo-
these exponential distributions for fructooligosaccha- saccharides (Timmermans et al., 1994). However, while
rides in red onions, shallots, and garlic seems to support we tried to account for this changing response, this is
the relative amounts of fructooligosaccharides assigned only the detector response for linear fructooligosaccha-
by MALDI-MS. Also this distribution may be useful in rides. Onion bulbs contained various isomeric fructo-
predicating the amounts of fructooligosaccharides with oligosaccharides, including branched fructooligosaccha-
higher DP in a sample or elucidating the changes of rides (Darbyshire et al., 1981; Bancal et al., 1989, 1991;
1556 J. Agric. Food Chem., Vol. 47, No. 4, 1999 Wang et al.

Figure 10. HPAE-PAD chromatogram of carbohydrate stand-

ards. Glucose, 98 ppm/5.98 min; fructose, 104 ppm/6.68 min;
isomaltose, 98 ppm/10.28 min; sucrose, 98 ppm/11.50 min;
Figure 8. HPAE-PAD chromatogram of shallots. Peaks are isomaltotriose, 108 ppm/19.47 min; maltose, 102 ppm/22.55
identified by comparison of the retention time with standards. min; maltotriose, 100 ppm/34.58 min; maltotetraose, 98 ppm/
The higher DPs (DP ) 6 or more) of fructooligosaccharides 36.93 min; maltopentaose, 96 ppm/38.23 min; maltohexaose,
are labeled by retention time compared to DP ) 5 and two 98 ppm/39.28 min; maltoheptaose, 102 ppm/40.25.
adjacent peaks are treated as isomers.
replicates that were performed in our experiments,
standard deviations for MALDI-MS were still higher
than for the HPAE-PAD method. Finally, MALDI-MS
gives better assurance of correct molecular assignment
since the isotopic mass of each peak is available,
although because of similar masses branched and linear
isomers cannot be distinguished. In fact with MALDI-
MS the assignments can be further checked by substi-
tuting a different alkali metal in the sample preparation
procedure to see expected mass shifts for each oligosac-
charide adduct.
Figure 9. Distribution of individual fructooligosaccharides While we have concentrated on the analysis of fructo-
in red onions (Y ) 159.5e-0.48X, R2 ) 0.99), shallots (Y ) oligosaccharides in this study, our feeling is that
82.8e-0.35X, R2 ) 0.97), and garlic (Y ) 18.73e-0.14X, R2 ) 0.99). MALDI-MS can also be used to quantitate other oligo-
Each data point and the error bars were determined from saccharides found in food by developing similar analyti-
duplicate analyses. cal methodology. To our knowledge this paper repre-
sents the first use of standard addition to quantitate
Stahl et al., 1997), and nothing is known about the using MALDI-MS.
differing responses of the PAD detector to these differ-
ently linked (2f6 links) fructooligosaccharides. How-
ever, standards for similar mass branched and linear LITERATURE CITED
glucose oligosaccharides are available and their PAD Abell, D. C.; Sporns, P. Rapid quantitation of potato glyco-
responses vary a great deal (Figure 10). In fact the alkaloids by matrix-assisted laser desorption/ionization
branched glucose oligosaccharides (that is 1f6 linked time-of-flight mass spectrometry. J. Agric. Food Chem. 1996,
isomaltose and isomaltotriose; measured peak areas of 44, 2292-2296.
12.0 and 10.6 million, respectively) have significantly AOAC Official Methods of Analysis 1990.
greater PAD responses than their linear mass equiva- Bancal, P.; Gaudillère, J. P. Fructan chemical structure and
lents (maltose and maltotriose; measured peak areas sensitivity to an exohydrolase. Carbohydr. Res. 1991, 217,
of 8.1 million and 6.8 million, respectively). It would 137-151.
therefore be expected that using linear oligosaccharide Bancal, P.; Henson, C. A.; Gaudillère, J. P.; Carpita, N. P.
PAD response factors, one would overestimate fructo- Oiligofructan separation and quantification by high perfor-
oligosaccharides with branched forms present. As can mance liquid chromatography. Application to Asparagus
be seen from Table 3, HPAE-PAD values are generally officialis and Triticum aestivum. Plant Physiol. Biochem.
1989, 745-750.
higher than corresponding MALDI-MS values.
MALDI-MS is a somewhat faster analysis method Bartsch, H.; König, W. A.; Strassner, M.; Hintze, U. Quantita-
tive determination of native and methylated cyclodextrins
than HPAE-PAD, taking about 20 min (for determining by matrix-assisted laser desorption/ionization time-of-flight
peak heights for 10 probe positions) rather than an hour mass spectrometry. Carbohydr. Res. 1996, 286, 41-53.
for each analysis and MALDI-MS is more tolerant to Börnsen, K. O.; Mohr, M. D.; Widmer, H. M. Ion exchange and
impurities (as noted for our garlic sample). Further purification of carbohydrates on a Nafion membrane as a
reduction in MALDI-MS analysis time can be obtained new sample pretreatment for matrix-assisted laser desorp-
by a less rigorous quantitation procedure (i.e., less laser tion/ionization mass spectrometry. Rapid Commun. Mass
pulses, fewer spots), but even with the numerous Spectrom. 1995, 9, 1031-1034.
MALDI-MS of Fructooligosaccharides J. Agric. Food Chem., Vol. 47, No. 4, 1999 1557

Bruker. MALDI-TOF Mass Spectrometer. User’s Guide to the Naven, T. J. P.; Harvey, D. J.; Brown, J.; Critchley, G.
ProflexTM 1995. Fragmentation of complex carbohydrates following ioniza-
Chatterton, N. J.; Harrison, P. A.; Thornley, W. R.; Bennett, tion by matrix-assisted laser desorption with an instrument
J. H. Sucrooligosaccharides and cool temperature growth fitted with time-lag focusing. Rapid Commun. Mass Spec-
in 14 forb species. Plant Physiol. Biochem. 1990, 28, 167- trom. 1997, 11, 1681-1686.
172. Nelson, C. J.; Smith, D. Fructans: their nature and occurrence.
Chatterton, N. J.; Harrison, P. A.; Thornley, W. R.; Bennett, Curr. Top. Plant Biochem. Physiol. 1986, 5, 1-16.
J. H. Separation and quantification of Fructan (inulin) Nicola, A. J.; Gusev, A. I.; Proctor, A.; Jackson, E. K.; Hercules,
oligomers by anion exchange chromatography. Inulin and D. M. Application of the fast-evaporation sample preparation
Inulin-containing Crops 1993, 93-99. method for improving quantification of angiotensin II by
Chatterton, N. J.; Harrison, P. A.; Thornley, W. R.; Bennett, matrix-assisted laser desorption ionization. Rapid Commun.
J. H. Purification and quantification of kestoses (fructosyl- Mass Spectrom. 1995, 9, 1164-1171.
sucroses) by gel permeation and anion exchange chroma-
Ohta, A.; Ohtsuki, M.; Baba, S.; Hirayama, M.; Adachi, T.
tography. Plant Physiol. Biochem. 1989, 27, 289-295.
Comparison of the nutritional effects of fructooligosaccha-
Darbyshire, B.; Henry, R. J. Differences in fructan content and
rides of different sugar chain length in rats. Nutr. Res. 1998,
synthesis in some Allium species. New Phytol. 1981, 87,
18, 109-120.
249-256.
Darbyshire, B.; Henry, R. J. The distribution of fructans in Ohta, A.; Ohtsuki, M.; Baba, S.; Takizawa, T.; Adachi, T.;
onions. New Phytol. 1978, 81, 29-34. Kimura, S. Effects of fructooligosaccharides on the absorp-
Gusev, A. I.; Wilkinson, W. R.; Proctor, A.; Hercules, D. M. tion of iron, calcium and magnesium in iron-deficient anemic
Improvement of signal reproducibility and matrix/comatrix rats. J. Nutr. Sci. Vitaminol. 1995, 42, 281-291.
effects in MALDI analysis. Anal. Chem. 1995, 67, 1034- Pieles, U.; Zuercher, W.; Schör, M.; Moser, H. E. Matrix-
1041. assisted laser desorption ionization time-of-flight mass
Harvey, D. J. Quantitative aspects of the matrix-assisted laser spectrometry: A powerful tool for the mass and sequences
desorption mass spectrometry of complex oligosaccharides. analysis of natural and modified oligonucleotides. Nucleic
Rapid Commun. Mass Spectrom. 1993, 7, 614-619. Acids Res. 1993, 21, 3191-3196.
Jespersen, S.; Niessen, W. M. A.; Tjaden, U. R.; vanderGreef, Pollock, C. J. Sucrose accumulation and the initiation of
J. Quantitative bioanalysis using matrix-assisted laser fructan biosynthesis in Lolium temulentum L. New Phytol.
desorption ionization mass spectrometry. J. Mass Spectrom. 1984, 96, 527-534.
1995, 30, 357-364. Shiomi, N.; Onodera, S.; Chatterton, N. J.; Harrison, P. A.
Kussmann, M.; Nordhoff, E.; Rahbek-Nielsen, H.; Haebel, S.; Separation of fructooligosaccharide isomers by anion-
Rossel-Larsen, M.; Jakobsen, L.; Gobom, J.; Mirgorodskaya, exchange chromatography. Agric. Biol. Chem. 1991, 5,
E.; Kroll-Kristensen, A.; Palm, L.; Roepstorff, P. Matrix- 1427-1428.
assisted laser desorption/ionization mass spectrometry sample Stahl, B.; Lions, A.; Karas, M.; Hillenkamp, F.; Steup, M.
preparation techniques designed for various peptide and Analysis of fructans from higher plants by matrix-assisted
protein analytes. J. Mass Spectrom. 1997, 32, 593-601. laser desorption/ionization mass spectrometry. Anal. Chem.
Livingston, D. P., III. Fructan precipitation from water/ethanol 1997, 246, 1195-204.
extract of oats and barley. Plant Physiol. 1990, 92, 767-
769. Stahl, B.; Steup, M.; Karas, M.; Hillenkamp, F. Analysis of
Loo, J, V.; Coussement, P.; Leenheer, L. de; Hoebregs, H.; neutral oligosaccharides by matrix-assisted laser desorption/
Smits, G. On the presence of inulin and oligofructose as ionization mass spectrometry. Anal. Chem. 1991, 63, 1463-
natural ingredients in the western diet. Crit. Rev. Food Sci. 1466.
Nutr. 1995, 35, 525-552. Suzuki, M.; Cutcliffe, J. A. Fructans in onion bulbs in relation
Losso, J. N.; Nakai, S. Molecular size of garlic fructooligosac- to storage life. Can. J. Plant Sci. 1989, 69, 1327-1333.
charides and fructopolysaccharides by matrix-assisted laser Timmermans, J. W.; van Leeuwen, M. B.; Tournois, H.; de Wit,
desorption ionization mass spectrometry. J. Agric. Food D.; Vliegenthart, J. F. G. Quantification analysis of the
Chem. 1997, 45, 4342-4346. molecular weight distribution of inulin by means of anion
Manghi, P.; Vernazza, F.; Ferrarini, O.; Cagnasso, P.; Salva- exchange HPLC with pulsed amperometric detection. J.
dori, C.; Fockedey, J. Analytical method for the determina- Carbohydr. Chem. 1994, 13, 881-888.
tion of inulin in yogurt enriched with soluble fibre. Eur. J. Wilkinson, W. R.; Gusev, A. I.; Proctor, A.; Houalla, M.;
Clin. Nutr. Suppl. 1995, 3, S296-S297. Hercules, D. M. Selection of internal standards for quantita-
Metzger, J. O.; Woisch, R.; Tuszynski, W.; Angermann, R. New tive analysis by matrix-assisted laser desorption-ionization
type of matrix for matrix-assisted laser desorption mass (MALDI) time-of-flight mass spectrometry. Fresenius J.
spectrometry of polysaccharides and proteins. Fresenius J. Anal. Chem. 1997, 357, 241-248.
Anal. Chem. 1994, 349, 473-474.
Mock, K. K.; Daevy, M.; Cottrell, J. S. The analysis of
underivatized oligosaccharides by Matrix-Assisted Laser Received for review August 28, 1998. Revised manuscript
Desorption Mass Spectrometry. Biochem. Biophys. Res. received December 14, 1998. Accepted December 17, 1998. This
Commun. 1991, 177, 644-651. research and the MALDI-MS instrument used were funded
Mohr, M. D.; Börnsen, K. O.; Widmer, H. M. Matrix-assisted by the Natural Sciences and Engineering Research Council
laser desorption/ionization mass spectrometry: improved (NSERC) of Canada.
matrix for oligosaccharides. Rapid Commun. Mass Spec-
trom. 1995, 9, 808-814. JF9809380