Stability of stevioside in food processing conditions: unexpected

recombination of stevioside hydrolysis products in ESI source

ABSTRACT
Stevioside is the main and the sweetest glycoside of stevia plant. It is attractive as a
natural sweetener to diabetics and others on carbohydrate-controlled diets. This
paper discusses the stability of stevioside under food processing conditions. It was
found that stevioside was transformed not only to rubusoside, steviolbioside, steviol
monoside and steviol but also to previously unknown stevioside α-anomer and
rubusoside α-anomer.

Those two identified stevioside transformation products are formed not only during
the heating of acidic, neutral and alkaline stevioside standard solutions and stevia
leaves suspensions in water and ethanol/water solvents but also during the
processing of foods containing stevia.

Apart from presenting the new compounds, the paper additionally shows that the
recombination of sugar moiety with steviolbioside molecule in MS/ESI source can
occur. The effect of molecule recombination in the MS source is known from the
literature; however, it has not been reported previously in relation to stevioside
derivatives.

1. Introduction
Extreme sweetness of Stevia rebaudiana leaves, commonly used as a sweetener
under the trade name “stevia”, results from the presence of steviol glycosides in this
plant species (Brandle and Telmer, 2007). Those organic glycosides, isolated from
the plant growing in South America, are 30 to 320 times sweeter than saccharose
(Brandle et al., 1998), although this comparison is constantly discussed (Brandle,
Starratt, & Gijzen, 1998; Cardello, Da Silva, & Damasio, 1999). As steviol glycosides
do not cause blood glucose level fluctuations, they are an attractive sugar substitute
for diabetics. The safe daily dose (ADI) of steviol glycosides, calculated as pure
steviol, is estimated at 4 mg/kg of body weight per day (Aguilar et al., 2010). In the
European Union, steviol glycosides are legal food additives, labeled as E-960,
approved for sweetening all types of food products, especially soft drinks, beer, ice
cream, fruit and vegetable products, jams, chocolates, candies, chewing gum,
breakfast cereals, desserts, etc. 13-[(2-O-β-D-Glucopyranosyl-β-D-glucopyranosyl)
oxy]-ent-kaur-16-en-19-oic acid β-D-glucopyranosyl ester (popularly known as
stevioside) is the most abundant steviol glycoside. Its content in Stevia rebaudiana
leaves, depending on the cultivar and the growth conditions, is estimated at 4–20%
of the dry weigh (Gardana, Scaglianti, & Simonetti, 2010; Geuns, 2003).

Two aspects concerning the properties of individual steviol glycosides deserve closer
examination. They differ in their sweetness level and they hydrolyze to glycosides
with a lower number of sugar moieties and/or to steviol (Patent publication number
WO 2014/197898). This aglycon weakly dissolves in water and has a bitter taste. In
consequence, the sweetness level of food products containing stevia is difficult to
predict precisely as various conditions of food preparation may lead to different
hydrolysis degree of stevioside. The present paper discusses the stability of
stevioside in different food processing conditions. The influence of time and medium
pH on the formation of individual hydrolysis products of stevioside in water and
ethanol/water solutions was examined. Stevioside standard, stevia leaves and
selected foods containing stevia were applied in the experiments.

2. Materials and methods

2.1. Materials and reagents

Apple, plums, tomato paste and dried stevia leaves, were purchased at one of the
local grocery shops. Sodium phosphate, phosphoric acid, ethanol (all of analytical
grade) and acetonitrile (HPLC) were supplied by the Polish Chemical Plant POCh
(Gliwice, Poland). Stevioside, rubusoside, steviolbioside, steviol and formic acid
standards were purchased from Sigma Aldrich (Seelze, Germany).

Deuterated water was bought from Armar AG (Döttingen, Switzerland). The Milli-Q
system from Millipore (Millipore, Bedford, MA) was used for water purification.

2.2. Investigations of stevioside transformation during the heating of

stevioside solutions and stevia suspensions

The following investigations of the stevioside transformation process were performed:

– heating under reflux of stevioside standard solutions in water, phosphoric
buffer (pH 4 or pH 9), ethanol/water, and ethanol/ phosphoric buffer (pH 4 or
pH 9) mixture;
– heating under reflux stevia leaves suspensions in the same solvents;
– heating in a laboratory oven pulps from apple, plums and tomato with addition
of stevia leaves.

All mixed solvents contained 50% v/v of alcohol.

The glass equipment for the experiments with stevioside solutions and stevia
suspensions consisted of a boiling flask (100 mL) and a small condenser. The
stevioside solutions and stevia suspensions contained 10 mg of the standard or 200
mg of the plant material in 50 mL of a given solvent. Individual solutions and
suspensions were heated for 1, 3 or 5 h at the boiling point of the used extractant.
The experiments with fruits and vegetable pulps were performed using 22-mL
headspace vials and a laboratory oven. Each pulp sample contained 40 mg of the
stevia leaves in 10 g of plant material. Individual pulps were heated at 150 °C for 3 h.
Subsequently, each obtained solvent and supernatant isolated from plant suspension
and pulp sample was subjected to LC–MS–PDA analysis. To obtain clear
supernatants centrifugation (10 min at 13000 rpm) was applied.

2.3. HPLC measurements

LC–MS measurements were performed using the same equipment and conditions as
described in Dawidowicz, Bernacik, Typek, and Stankevič (2017). The composition of
mobile phase and ion monitoring were the only difference in this case. Mobile phase
components were: A − 25 mM formic acid in water and B − 25 mM formic acid in
acetonitrile. The gradient program started at 20% B increasing to 95% B over 50 min
and was followed by isocratic elution (95% B) for 25 min. The total run time was 75
min at a mobile phase flow rate of 0.5 mL/min.
To better visualize the chromatographic separation and to remove the signal
connected with the plant matrix and buffer components, the SIM function was used
during the analysis of all examined samples. The time periods and monitored ions
were as follows:

10–19 min, m/z 803 for compound 1; 19–21 min, m/z 803 for compound 2;
21–23 min, m/z 641 for compound 3; 23–24 min, m/z 803 for compound 4;
24–27 min, m/z 641 for compound 5; 27–32 min, m/z 479 for compound 6;
32–45 min, m/z 317 for compound.

To identify stevioside derivatives, MS2 and HRMS functions were applied. In MS2
the collision energy for each examined compound was chosen individually. Due to
the lack of stevioside anomer standards, their amounts were estimated by relating
their chromatographic responses to the calibration curves for stevioside, rubusoside,
steviolbioside and steviol.

2.4. Preparative chromatography

To prepare sufficient amounts of compounds 2 and 4 for NMR analysis, 200 mg of
the stevioside standard were dissolved in 20 mL phosphorus buffer (pH3) and heated
for 5 h under reflux. Next, 2-mL portions of the obtained solution were injected into an
ODS column LUNA C18 100 Å (5 μm, 150×10 mm; Shimadzu), which was installed
in an HPLC system from Shimadzu (Kyoto, Japan) equipped with a SPD- 20AV UV–
VIS detector and an automatic fraction collector FRC-10A. Gradient elution was
applied during chromatographic separation. The mobile phase was composed of
water (solvent A) and acetonitrile (solvent B). The following elution program was
applied: gradient from 20% to 95% of B over 75 min followed by isocratic elution
(95% B) for 5 min, all at a steady flow rate of 5 mL/min. The automatic fraction
collector was controlled by the UV–VIS detector working at 210 nm.

Protonated solvents from the collected fractions were removed by evaporation in a

nitrogen stream. The resulting dry residue was dissolved in deuterated water and
subjected to NMR analysis.

2.5. NMR measurements

NMR measurements were performed using an NMR system from Bruker (Ascend
600). The D2O solutions of the dry residues obtained from fractions collected by
preparative chromatography were examined using 1H, 1H–1H COSY, 13C,
DEPT135, HSQC and HMBC. The following results were obtained for:

2.5.1. Stevioside
H NMR (500 MHz, D2O) δ 0.93 (s, 3H), 0.98–1.19 (m, 3H), 1.27 (s, 3H), 1.38–1.52
(m, 3H), 1.52–1.70 (m, 3H), 1.79–2.09 (m, 8H), 2.14–2.27 (m, 3H), 3.29–3.38 (m,
3H), 3.38–3.47 (m, 3H), 3.47–3.59 (m, 5H), 3.64–3.76 (m, 4H), 3.82–3.92 (m, 3H),
4.68–4.77 (m, 2H), 4.94 (s, 1H), 5.17 (s, 1H), 5.43 (d, J8.0 Hz);

C NMR (126 MHz, D2O) δ 15.0, 18.7, 20.2, 21.4, 28.0, 36.0, 37.5, 39.2, 40.3, 40.9,
42.0, 43.8, 44.2, 46.9, 53.4, 56.8, 60.5, 60.7, 61.2, 69.2, 69.5, 69.8, 71.9, 74.3, 75.6,
76.1, 76.3, 76.7, 80.6, 86.9, 94.1, 95.8, 103.1, 104.8, 153.2, 178.8.
2.5.2. Steviolbioside
H NMR (500 MHz, D2O) δ 0.77–0.85 (m, 1H), 0.90 (s, 3H), 0.92–1.02 (m, 3H), 1.05
(s, 3H), 1.31–1.37 (m, 1H), 1.37–1.54 (m, 4H), 1.54–1.63 (m, 1H), 1.70–1.86 (m, 4H),
1.90–1.98 (m, 2H), 2.03–2.15 (m, 2H), 2.20–2.25 (m, 1H), 3.28–3.42 (m, 5H), 3.42–
3.48 (m, 1H), 3.50–3.55 (m, 1H), 3.62–3.69 (m, 2H), 3.82–3.89 (m, 2H), 4.70–4.74
(m, 1H), 4.91 (s, 2H), 5.08 (s, 2H); DEPT 135 (126 MHz, D2O) δ 15.7, 19.3, 19.9,
22.2, 28.9, 36.6, 38.8, 40.7, 41.1, 44.4, 46.8, 52.8, 56.7, 60.6, 61.1, 69.5, 69.7, 74.3,
75.5, 75.6, 76.2, 76.3, 79.9, 95.6, 102.6, 103.8.

2.5.3. Stevioside α–anomer

H NMR (500 MHz, D2O) (due to overlapping only selected peaks were described) δ
0.86 (s, 3H), 1.20 (s, 3H), 5.41 (d, J7.9 Hz).

2.6. Statistical analysis

All the results are presented as the mean of three independent measurements (n3).
Differences in the concentration of the formed stevioside derivatives were compared
using analysis of variance (p0.05). Differences in the studied groups were considered
significant for p values lower than 0.05 and F values higher than 4.066. Analysis of
variance revealed statistically significant differences for the tested groups.

3. Results and discussion

As reported in the literature (Dawidowicz, Bernacik, & Typek, 2016a, 2016b, Kroyer,
1999, 2010; Musa et al., 2014; Wianowska, Dawidowicz, Bernacik, & Typek, 2017;
Wölwer-Rieck, Tomberg, & Wawrzun, 2010), thermal lability is a characteristic
feature of many glycosides. These compounds not only can disconnect sugar
moieties during their hydrolysis but also can undergo anomerization. This is why the
anomeric transformation of stevioside was also assumed in our research, despite the
absence of reports about such conversion. Although many stevioside transformation
products can be imagined, the most probable ones from a thermodynamic point of
view are those illustrated in Fig. 1. The correctness of this identification was
confirmed by the
HPLC, MS and NMR data.

Fig. 2 presents SIM chromatograms of stevioside solutions in water (A) and in

phosphate buffers of pH 4 (B) and pH 9 (C); supernatants of stevia suspension in
water (D) and in phosphate buffers of pH 4 (E) and pH 9 (F), all heated for 3 h at the
boiling temperature of the used solvent.

The identification of individual peaks is given in the caption to Fig. 2. and results
from:

– MSn and HRMS (collected in Table 1);

– their retention data (following the general rule for RP-HPLC that
– more hydrophobic molecules exhibit longer retention);
– the retention data of the standards.
–
The analysis of the examined stevioside solutions and stevia suspensions revealed
the existence of compounds of the same m/z as that of the stevioside transformation
products shown in Fig. 1
To confirm the presence of stevioside anomers (compounds 2 and 4) suggested by
the MS and HPLC data, fractions from preparative chromatography containing these
new compounds were analyzed using 1H, 13C and 2D NMR techniques. Their NMR
spectra are presented in Figs. 1S, 2S and 3S (see Supplementary Material). The first
fraction contained only one compound (compound 4) which turned out to be not
stevioside anomer but steviolbioside. In order to confirm the correctness of the above
identification, NMR spectra for stevioside standard were considered. The H NMR
spectra (see bottom spectra in Fig. 1S) reveal a doublet at 5.43 ppm corresponding
to the anomeric proton in the sugar fragment bonded to carbonyl group (alcoholic
fragment in stevioside). As can be deduced from the HSQC experiment, the doublet
corresponds with the signal at 94.1 ppm in C NMR/DEPT135 spectra from tertiary
anomeric carbon (see Fig. 2S). Moreover, two additional signals from anomeric CH
carbon atoms are present in DEPT135 spectra at 95.8 ppm and 103.1 ppm. They are
connected with the presence of three sugar moieties in the structure of stevioside. In
1H NMR spectra of the first fraction (compound 4, see middle spectra in Fig. 1S), the
signal at 5.43 ppm is absent, which suggests the lack of one sugar fragment. The
same results are seen from the analysis of 13C NMR/DEPT135 spectra (see Fig.
3S), in which signals at 95.6 ppm and 102.6 ppm from only two anomeric CeH
carbon atoms are present. Moreover, the signals of CH3 groups in 1H NMR spectra
of compound 4 are remarkably shifted toward a higher field and are found at 0.91
ppm and 1.05 ppm, respectively. The same signals in stevioside were found at 0.93
ppm and 1.27 ppm, respectively. In stevioside, the signal at 1.27 ppm corresponds to
the methyl group at the

α-carbon atom relative to carboxyl group (as deduced by HMBC analysis), whereas
the singlet at 0.93 ppm corresponds to the distant methyl group in the steviol
skeleton. The upfield shift of the signal of the latter group can be associated with the
shielding effect of the carbonyl group which is in close proximity in the structure of
stevioside (see Fig. 4Sb). The downfield shift of the α-methyl group is associated
with the presence of

β-anomeric sugar moiety at the carboxylic group, which forces the C=O group to
deviate from the perpendicular position relative to methyl group (torsion angle
between CH3 and alkoxy oxygen is close to 109°), thus deshielding the latter. The
lack of this effect in compound 4 results in a shift of the α-methyl group at 1.05 ppm
(the torsion angle between CH3 and hydroxyl oxygen is close to 90°). All the above
arguments prove that compound 4 is steviolbioside.

NMR analysis of the second fraction (see upper spectrum in Fig. 1S) suggests that it
also contains steviolbioside. However, the small signals at 5.41 ppm, 1.20 ppm and
0.85 ppm indicate the presence of another product (compound 2), which is neither
steviolbioside nor starting stevioside. The presence of a doublet at 5.41 ppm
confirms the presence of a third sugar moiety placed at the carboxyl group.
Therefore, the minor product must be stevioside with α-anomeric sugar bonded to
the carboxyl group. The shifts of α-CH3 and distant CH3 groups in 1H NMR spectra
confirm the α-anomeric structure of stevioside. The simplified structure of the latter is
shown in Fig. 4S c. As in the case of stevioside, the presence of a sugar moiety at
the carboxy group forces the latter to bend out of perpendicular position relative to
the α-CH3 group (the torsion angle between CH3 and alkoxy oxygen is close to
103°), although the effect is slightly weaker due to the lack of interactions between
the α-CH3 group and the sugar fragment. The same effect places the C=O group
almost parallel to the distant CH3 group, which in turn shields the latter even more
strongly than in steviolbioside (0.93 ppm vs. 0.85 ppm).

The observed α-anomeric structure of stevioside (13-[(2-O-β-D-glucopyranosyl-β-D-

glucopyranosyl)oxy]-ent-kaur-16-en-19-oic acid α-Dglucopyranosyl ester) in the
heated solutions of stevioside and the heated suspensions of stevia is very
interesting because this stevioside stereoisomeric form, not yet reported in the
literature, confirms the initial supposition about anomerization of stevioside.

Compound 5 was identified as rubusoside α-anomer, i.e. (4α)-13- (β-D-

glucopyranosyloxy)kaur-16-en-18-oic acid α-D-glucopyranosyl ester, on the basis of:

– MSn and HRMS data (collected in Table 1) – they are the same as for
rubusoside β-anomer, compound 3;
– HPLC data – its retention in the RPHPLC system is longer than the retention
of rubusoside β-anomer and the same retention sequence as that for
stevioside α- and β-anomers is observed.

As the amount of compound 5 among the transformation products of stevioside is

very small, the confirmation of the proposed identification by NMR data would be too
expensive. Hence, the assignment should be treated tentatively.

Another new observation is the discrepancy in the identification of compound 4 by

NMR and MS techniques. According to NMR analysis, which is more reliable in
relation to MS, compound 4 is identified as steviolbioside, a compound with two
sugar moieties, whereas MS analysis indicates that this compound contains three
sugar moieties. The connection of a third sugar moiety to steviolbioside molecule
probably occurs in the ESI source. Such an effect is known from the literature
(Purwaha, Silva, Hawke, Weinstein, & Lorenzi, 2014); however, it was not yet
reported in relation to stevioside derivatives.

The influence of heating time on the amount of individual stevioside transformation

products formed in water (A), phosphate buffers pH 4 (B) and pH 9 (C), stevia
suspension in water (D) and stevia suspension in phosphate buffers pH 4 (E) and pH
9 (F) are presented in Fig. 3. The analysis of dependencies from Fig. 3 A–C leads to
two conclusions:

– the degree of stevioside transformation grows with the length of heating time;
– stevioside transformation process is pH-dependent and intensified by acidic
and alkaline environments.

A more detailed consideration of individual bars shows that steviolbioside is the main
transformation product of stevioside as the ester bond is weaker than the ether bond
and easily undergoes hydrolysis.

Analogous conclusions can be drawn from the dependencies in Fig. 3 D–F showing
the influence of heating time on stevioside transformation in suspensions of stevia
leaves. In this case, however, the stevioside transformation process is influenced by
the plant matrix components. The obtained results prove that stevioside degradation
is accelerated with hydronium cation concentration decrease in all stevia extracts.
Moreover, the matrix components significantly facilitate the formation of stevioside α-
anomer (compound 2).

The stevia sweetener is used in food products containing ethanol. Hence,

experiments analogous to those above were performed using ethanol/water mixture
as a solvent of stevioside and suspending liquid of stevia leaves. Fig. 4 presents the
SIM chromatogram of stevioside solutions in water/EtOH, 50/50 v/v (A), phosphate
buffer pH 4/EtOH, 50/50 v/v (B) and phosphate buffer pH 9/EtOH, 50/50 v/v (C),
stevia suspension in water/EtOH, 50/50 v/v (D), in phosphate buffer pH 4/ EtOH,
50/50 v/v (E) and in phosphate buffer pH 9/EtOH, 50/50 v/v (F), all heated for 3 h at
the boiling temperature of the solvent used. The comparison of chromatograms
corresponding to ethanol/water and water stevioside solutions (Fig. 4 A-C and Fig. 2
A–C, respectively) show that the presence of ethanol in stevioside solution leads to a
small difference in the qualitative transformation of stevioside. There is no rubusoside
α-anomer in alkaline ethanol/water stevioside solution, which may suggest (I), (III)
and (VI) pathways of stevioside degradation (see Fig. 1). More difference, however,
is seen in the case of stevioside transformation in ethanol/water stevia suspension
(compare Fig. 4 D–F and Fig. 2 D–F). There is no rubusoside α-anomer and steviol
in alkaline ethanol/water stevia suspension. The absence of these compounds
confirms the above hypothesis concerning the pathway of stevioside degradation in
an alkaline environment. The inhibiting action of the plant matrix constituents in
stevioside degradation cannot be excluded.

Fig. 5. presents the influence of heating time on the amount of individual stevioside
transformation products formed in water/EtOH, 50/ 50 v/v (A), phosphate buffer pH
4/EtOH, 50/50 v/v (B) and phosphate buffer pH 9/EtOH, 50/50 v/v (C), stevia
suspension in water/EtOH, 50/50 v/v (D), in phosphate buffer pH 4/EtOH, 50/50 v/v
(E) and in phosphate buffer pH 9/EtOH, 50/50 v/v (F). The general conclusion from
the figure is that ethanol suppresses stevioside degradation.

To confirm the possibility of stevioside degradation during food preparation,

additional experiments were performed processing selected plant materials
containing stevia in domestic conditions. The results of these experiments are
illustrated in Fig. S5 (see Supplementary Material) and prove the presence of some
stevioside hydrolysis products. As results from the relationships, compounds 1, 2, 3
and 5 appear in the prepared pulps; however, their concentration depends on the
plant material used.

To facilitate the comparison of previous results (Figs. 3 and 5) with the last results
(Fig. S5) they were presented in relation to 10 mg of the stevioside in each pulp.

4. Conclusions
At elevated temperature, stevioside is transformed to rubusoside, steviolbioside,
steviol monoside and steviol. The presented results indicate that three other
derivatives, 13-[(2-O-β-D-glucopyranosyl-β-Dglucopyranosyl) oxy]-ent-kaur-16-en-
19-oic acid α-D-glucopyranosyl ester, (4α)-13-(β-D-glucopyranosyloxy)kaur-16-en-
18-oic acid β-D-glucopyranosyl ester and (4α)-13-(β-D-glucopyranosyloxy)kaur-16-
en-18-oic acid α-D-glucopyranosyl ester, are also stevioside transformation products.
The first two were confirmed by NMR spectroscopy. The identification of the third one
should be treated tentatively as it is formed in very small amounts during stevioside
transformation and its preparative isolation for NMR analysis would be economically
unviable. Although alkaline and acidic environments strongly favor stevioside
degradation to the mentioned compounds, the concentration of most of them in
neutral environment is significant, too.

The obtained results not only show the formation of new anomeric forms but also
prove that the recombination of sugar moiety with steviolbioside molecule occurs in
the ESI source. The effect of molecule recombination in the MS ion source is already
known from the literature (Purwaha et al., 2014), but it has not been reported before
in relation to stevioside derivatives.

The presented results are very important in the context of the food industry, which
involves preparation of foods sweetened with stevia leaves.

Due to a great popularity of stevia as a sweetener in foods prepared at high

temperatures, often well above 100 °C, stevioside degradation to its stereoisomers is
worth studying at temperatures higher than those applied in the reported
experiments, the more so that the biological activity of both anomers and their impact
on human health have not been fully recognized yet.