Study of the isomeric Maillard degradants, glycosylamine and Amadori rearrangement products, and their differentiation via MS2 fingerprinting from collision-induced decomposition of protonated ions

Article abstract of DOI:10.1002/rcm.9062

Rationale: The focus of this work was to study glycosylamine and Amadori rearrangement products (ARPs), the two major degradants in the Maillard reactions of pharmaceutical interest, and utilize their MS² fingerprints by liquid chromatography/high-resolution tandem mass spectrometry (LC/HRMS²) to quickly distinguish the two isomeric degradants. These two types of degradants are frequently encountered in the compatibility and stability studies of drug products containing primary or secondary amine active pharmaceutical ingredients (APIs), which are formulated with excipients consisting of reducing sugar functionalities. Methods: Vortioxetine was employed as the primary model compound to react with lactose to obtain the glycosylamine and ARP degradants of the Maillard reaction, and their MS² spectra (MS² fingerprints) were obtained by LC/MS². Subsequently, the two degradants were isolated via preparative HPLC and their structures were confirmed by one- and two-dimensional (1D and 2D) nuclear magnetic resonance (NMR) determination. Results: The MS² fingerprints of the two degradants display significantly different profiles, despite the fact that many common fragments are observed. Specifically, protonated glycosylamine shows a prominent characteristic fragment of [M_vort + C₂H₃O]⁺ at m/z 341 (M_vort is the vortioxetine core), while protonated ARP shows a prominent characteristic fragment of [M_vort + CH]⁺ at m/z 311. Further study of the Maillard reactions between several other structurally diverse primary/secondary amines and lactose produced similar patterns. Conclusions: The study suggests that the characteristic MS² fragment peaks and their ratios may be used to differentiate the glycosylamine and ARP degradants, the two isomeric degradants of the Maillard reaction, which are commonly encountered in finished dosage forms of pharmaceutical products containing primary and secondary amine APIs.

Full text of DOI:10.1002/rcm.9062

Received: 13 December 2020
DOI: 10.1002/rcm.9062
Revised: 23 January 2021
Accepted: 25 January 2021
R E S E A R C H A R T I C L E
Study of the isomeric Maillard degradants, glycosylamine and
Amadori rearrangement products, and their differentiation via
MS2 fingerprinting from collision-induced decomposition of
protonated ions
Shaolan Wang1,2 | Jinsheng Lin1 | Dan Li1 | Tianpei Huang1 | Wenquan Zhu1
|
Wenbin Chen1 | Min Li1,3
| Weiyang Shen2
1Center of Excellence for Modern Analytical
Technologies (CEMAT), Zhejiang Huahai
Pharmaceutical Co. Ltd., Xunqiao, Linhai,
Zhejiang, 317204, China
Rationale: The focus of this work was to study glycosylamine and Amadori
rearrangement products (ARPs), the two major degradants in the Maillard
reactions of pharmaceutical interest, and utilize their MS2 fingerprints by liquid
chromatography/high-resolution tandem mass spectrometry (LC/HRMS2) to
quickly distinguish the two isomeric degradants. These two types of degradants
are frequently encountered in the compatibility and stability studies of drug
products containing primary or secondary amine active pharmaceutical
ingredients (APIs), which are formulated with excipients consisting of reducing
sugar functionalities.
2Department of Pharmaceutical Analysis,
China Pharmaceutical University, Nanjing,
210009, China
3Huahai US, Inc., 700 Atrium Drive, Somerset,
NJ, 08873, USA
Correspondence
M. Li, Center of Excellence for Modern
Analytical Technologies (CEMAT), Zhejiang
Huahai Pharmaceutical Co. Ltd, Xunqiao,
Linhai, Zhejiang 317204, China.
Email: minli88@yahoo.com
Methods: Vortioxetine was employed as the primary model compound to react with
lactose to obtain the glycosylamine and ARP degradants of the Maillard reaction, and
their MS2 spectra (MS2 fingerprints) were obtained by LC/MS2. Subsequently, the
two degradants were isolated via preparative HPLC and their structures were
confirmed by one- and two-dimensional (1D and 2D) nuclear magnetic resonance
(NMR) determination.
W.-Y. Shen, Department of Pharmaceutical
Analysis, China Pharmaceutical University,
Nanjing 210009, China.
Email: shenweiyang@cpu.edu.cn
Results: The MS2 fingerprints of the two degradants display significantly different
profiles, despite the fact that many common fragments are observed. Specifically,
protonated glycosylamine shows
a
prominent characteristic fragment of
[Mvort + C2H3O]+ at m/z 341 (Mvort is the vortioxetine core), while protonated ARP
shows a prominent characteristic fragment of [Mvort + CH]+ at m/z 311. Further
study of the Maillard reactions between several other structurally diverse primary/
secondary amines and lactose produced similar patterns.
Conclusions: The study suggests that the characteristic MS2 fragment peaks and
their ratios may be used to differentiate the glycosylamine and ARP degradants, the
two isomeric degradants of the Maillard reaction, which are commonly encountered
in finished dosage forms of pharmaceutical products containing primary and
secondary amine APIs.
Rapid Commun Mass Spectrom. 2021;35:e9062.
wileyonlinelibrary.com/journal/rcm
© 2021 John Wiley & Sons Ltd
1 of 14
https://doi.org/10.1002/rcm.9062

2 of 14
WANG ET AL.
1
|
INTRODUCTION
Schiff base/glycosylamine may be reversible, that of the ARPs is
irreversible. In addition, an ARP is also a key intermediate for a
number of subsequent degradants containing aldehyde and α-diketo
functional groups. Thus, it is important to distinguish the
glycosylamine and ARP degradants.
Drug degradation products, or drug degradants, are inevitably formed
in drug products which are formally examined in various drug stability
studies, such as long-term1 and accelerated stability studies2 under
the ICH stability conditions. The rapid structural elucidation of drug
degradants is critical to the quality and safety of a drug product. For
formulated drug products, drug degradation can originate from the
interaction (or incompatibility) between the active pharmaceutical
ingredient (API) and excipients. The Maillard reaction is one of the
major reasons for the chemical incompatibility encountered during
pharmaceutical development.3–5 Over 100 years ago, Louis Maillard
reported that carbonyl compounds may react with amino acids and
proteins to produce complex mixtures of brown pigments; this
phenomenon has become known as the Maillard reaction.6
Pharmaceutical formulations containing primary and secondary amine
APIs as well as excipients of reducing sugars are prone to this
degradation chemistry, in which condensation between the amines
and aldehyde functional groups takes place as the initial step of the
Maillard reaction, resulting in the formation of Schiff bases, which
cyclize to become glycosylamines, under pharmaceutically relevant
conditions. The glycosylamines may not be stable enough and, quite
often, they can undergo the Amadori rearrangement to form
1-deoxy-1-amino-2-ketoses, also known as the Amadori
rearrangement products (ARPs).7 The Maillard reaction is a very
complex process; only the early-stage degradants, namely the initial
glycoylsamine and possibly the ARP degradants, are usually relevant
in drug degradation chemistry.8 While the formation of the initial
Nevertheless, the glycosylamine and ARP degradants are isomeric
toward each other; hence, they are not distinguishable based on their
molecular weights or formulae. In our laboratories, we have employed
a strategy that combines liquid chromatography/photo-diode array
UV/multi-stage mass spectrometry (LC/PDA-UV/MSn) with
mechanism-based stress studies to rapidly elucidate the structures of
unknown degradants with a very high confidence level.9–17 The most
critical and reliable component of this strategy is the MSn molecular
fingerprints, which can be used to differentiate structurally similar
diastereomers, through the often subtle but distinguishable
differences between their MSn fingerprints.18 Therefore, in the
current study, we envisaged that the MSn fingerprints of the two
isomeric Maillard degradants would be different and, as such, certain
fragments in the two isomers might be utilized as characteristic
markers to differentiate them. Such an approach would be much more
efficient than alternative analytical techniques such as nuclear
magnetic resonance (NMR) spectroscopy. Thus, we decided to use
vortioxetine, a drug for treating major depression,19 as the primary
model compound containing a secondary amine moiety to react with
lactose, one of the most utilized reducing sugars in pharmaceutical
formulations,20 to generate the corresponding glycosylamine and ARP
degradants (1 and 2; Scheme 1) for analysis by high-resolution
LC/MSn molecular fingerprinting.
SCHEME 1 A plausible mechanism for the reaction between vortioxetine and lactose; pathway b is the Amadori rearrangement

WANG ET AL.
3 of 14
In the subsequent study, the corresponding Maillard reaction
2.2.2
|
Method II
degradants obtained from reactions of vortioxetine with other
reducing sugars (degradants 3 through 6), as well as several other
structurally diverse primary/secondary amines with lactose, glucose,
and galactose (degradants 7 through 16), were subjected to the same
high-resolution LC/MSn analysis.
The mobile phase gradient was varied according to the following
program: 0 min (40% B), 1 min (40% B), 10 min (90% B), 12 min
(90% B), 12.1 min (40% B) and 15 min (40% B). The other
parameters were the same as for method I. This method was used
in the experiments involving the Maillard reaction with other
primary and secondary amines. Due to the fact that the elution
times of the relevant components in the Maillard reaction were
significantly shorter than those in the Maillard reaction with
vortioxetine, the run time of this method is approximately half of
that in method I.
Here we will demonstrate that two characteristic fragments
obtained from the MS2 fragmentation and their ratios may be used
to rapidly differentiate the glycosylamine and ARP degradants of
the Maillard reaction. During the final stage of preparing this
manuscript for publication, we have become aware of the work
just published by Xing et al,21,22 in which they reported that
MS/MS fragmentation patterns can be used to discriminate
between glucose-derived Schiff bases, Amadori, and Heyns
compounds with several amino acids. Their work as well as ours
indicates that this methodology can be generally applicable for
differentiating the early-stage isomeric Maillard products from
amino compounds of vastly different structures with reducing
sugars.
2.3
|
LC/PDA-UV/MSn (n = 1, 2) analysis
A 1260 series HPLC instrument (Agilent Technologies, Santa Clara,
CA, USA) interfaced to a quadrupole time-of-flight (Q-TOF) mass
spectrometer (6545 series, Agilent Technologies) was used for
comprehensive LC/PDA-UV/MSn (n = 1, 2) analyses of samples. UV
spectra were collected from 190 nm to 400 nm by the PDA
detector. The Q-TOF mass spectrometer was operated in positive
electrospray ionization (ESI) mode with the following source
parameters: fragmentor voltage 70 V, gas flow 6 L/min, nebulizer
pressure 60 psi, source temperature 320ꢀC, sheath gas temperature
350ꢀC, sheath gas flow 12 L/min, and capillary voltage 3.5 kV.
Nitrogen was used as the gas for all purposes including
collision-induced decomposition (CID). The mass acquisition range
was m/z 100–1700 and, for the MS2 analyses, the collision energy
was set at 10, 20, and 30 eV, respectively. The chromatographic
conditions of the LC/MS method were the same as those of
method I.
Some of the LC/PDA-UV/MSn experiments were also conducted
on an IT-TOF mass spectrometer (Shimadzu, Kyoto, Japan). The
IT-TOF mass spectrometer was operated in positive ESI mode with
the following parameters: nebulizing gas flow 1.5 L/min, drying gas
pressure 107 kPa, CDL temperature 200ꢀC, heating block
temperature 200ꢀC, detector voltage 1.6 kV, ion accumulation time
50 ms, and interface voltage 4.5 kV. Nitrogen was used as the gas for
all purposes except CID, for which argon was used. The mass
acquisition range was m/z 50–1000 and, for the MS2 analyses, the
collision energy was set at 50%. The chromatographic conditions of
the LC/MS method were the same as those for method I.
2
|
EXPERIMENTAL
2.1
|
Chemicals and reagents
Vortioxetine was manufactured by Zhejiang Huahai Pharmaceutical
Co., Ltd (Linhai, Zhejiang, China). Acetonitrile and ethanol were
purchased from Sigma-Aldrich Corp. (St Louis, MO, USA). Formic acid,
dimethyl sulfoxide (DMSO), dimethylformamide (DMF), lactose,
glucose, galactose, α-phenylethylamine, 1-phenylpiperazine and
benzylamine were procured from Shanghai Aladdin Biochemical
Technology Co., Ltd (Shanghai, China). Ammonium acetate and
sodium carbonate were purchased from Xilong Chemical Co., Ltd
(Guangdong, China).
2.2
|
HPLC conditions
| Method I
2.2.1
A Dionex Ultimate 3000 HPLC system (Thermo Scientific, Waltham,
MA, USA) equipped with YMC PACK ODS-AQ column
a
(250 mm × 4.6 mm, 5 μm) and a PDA-UV detector was used to
analyze the Maillard reaction results. The mobile phase system
consisted of A (water with 600 mg/L ammonium acetate) and B
(acetonitrile), with the gradient varied according to the following
program: 0 min (40% B), 10 min (40% B), 25 min (90% B), 30 min
(90% B), 30.1 min (40% B), and 33 min (40% B). The analyses were
performed at a flow rate of 1.0 mL/min and a column temperature of
40ꢀC. UV spectra were collected between 190 and 400 nm by the
PDA-UV detector and, for UV chromatograms, a wavelength of
250 nm was used. This method was used in the experiments involving
the Maillard reaction with vortioxetine.
2.4
|
Maillard reaction between vortioxetine and
lactose
A mixture of vortioxetine (5 mg), lactose monohydrate (47.55 mg),
and potassium carbonate (3.6 mg) in 1.2 mL water/acetonitrile (50:50,
v/v) was shaken at 60ꢀC for 24 h. Aliquots of the reaction solution
were analyzed by HPLC and LC/MSn (n = 1, 2) as described in
sections 2.2.1 and 2.3, respectively.

4 of 14
WANG ET AL.
2.5
|
Preparation and isolation of glycosylamine
acetate) and B (acetonitrile), with the gradient varied according to the
following program: 0 min (40% B), 10 min (40% B), 25 min (90% B),
30 min (90% B), 30.1 min (40% B) and 33 min (40% B). The analyses
were performed at a flow rate of 15 mL/min and the fractions at
ꢁ15–16 min were collected, combined, and then lyophilized to afford
220 mg of product (22% yield).
1 produced by the Maillard reaction between
vortioxetine and lactose
A relatively large quantity of the degradation product of vortioxetine
and lactose was needed for further investigation. Hence, a mixture of
vortioxetine (1 g), lactose monohydrate (9.51 g), and potassium
carbonate (0.72 g) in 120 mL water/acetonitrile (50:50, v/v) was
allowed to stir at 60ꢀC for 24 h. After reaction, the solvent was
removed by rotary evaporator. Then 100 mL ethanol was added and
the solution was allowed to stir at room temperature for 30 min. The
insoluble potassium carbonate was filtered and the filtrate was
concentrated to 10 mL. The mixture was loaded onto a LC-8A series
preparative HPLC instrument (Shimadzu) for isolation of the product.
The preparative HPLC instrument was equipped with a YMC PACK
ODS-AQ column (250 mm × 20 mm, 5 μm) and a UV detector. The
mobile phase system consisted of A (water with 0.6 g/L ammonium
2.6
|
Selection of the optimal conditions for
converting glycosylamine 1 into ARP 2
Six 1.5 mL HPLC vials were selected and labeled as vials 1, 2, 3, 4,
5 and 6. Then 1 mg of the isolated glycosylamine 1 solid sample was
added to each of the six vials. To vials 2, 3, 4, 5 and 6, 1 mL each of
the following solvents was added into each vial, respectively: water,
methanol, acetonitrile, DMF, and DMSO. The glycosylamine in vial
1 remained in the solid state. All the vials were capped and then
FIGURE 1 The key gHMBC correlations of glycosylamine 1 formed between vortioxetine and lactose

WANG ET AL.
5 of 14
heated at 80ꢀC for 24 h with vibration. Samples of the reaction solid
(vial 1) and solutions (vials 2 to 6) were analyzed by HPLC and
LC/MSn (n = 1, 2) as described in sections 2.2.1 and 2.3, respectively;
the solid sample in vial 1 was dissolved in methanol (1 mg/mL) prior
to injection, while all the solution samples were diluted with methanol
(1 mg/mL) and then injected (5 μL). The results show that the highest
yield of ARP 2 was achieved at ꢁ27% in vial 1.
preparative HPLC instrument was equipped with a YMC PACK ODS-
AQ column (250 mm × 25 mm, 5 μm) and a UV detector. The mobile
phase system consisted of A (water with 0.6 g/L ammonium acetate)
and
B (acetonitrile), with the gradient varied according to the
following program: 0 min (40% B), 10 min (40% B), 25 min (90% B),
30 min (90% B), 30.1 min (40% B) and 33 min (40% B). The
purification was performed at a flow rate of 15 mL/min and the
fractions at ꢁ17–18 min were collected. The fractions collected from
different runs were combined, and then lyophilized to afford ꢁ10 mg
of the product (2; 25% yield).
2.7
|
Isolation of ARP 2
A relatively large quantity of ARP 2 was needed for NMR structural
confirmation. Based on the results of section 2.6, a 20 mL headspace
GC vial was selected as the reaction vessel in which 40 mg of the
isolated glycosylamine 1 was placed, which was heated at 80ꢀC for
24 h with shaking. After that, 10 mL methanol was added to dissolve
the solid. The mixture was then loaded onto the preparative HPLC
instrument (LC-8A series, Shimadzu) for isolation of the product. The
2.8
|
1D and 2D NMR determination of
glycosylamine 1 and ARP 2
Approximately 10 mg of glycosylamine
1 and 10 mg of ARP
2, prepared and isolated as described in sections 2.5 and 2.7, were
dissolved in 1 mL of CDCl3, respectively. 1H-, 13C-NMR and 2D NMR
TABLE 1 1H-NMR and 13C-NMR
data of glycosylamine 1 formed between
vortioxetine and lactose
C δ (ppm)
20.6
C type
CH3
CH3
CH2
CH2
CH2
CH2
CH
CH
CH
CH
CH
CH
CH
CH
CH
CH
CH
CH
CH
CH
C
C number
H δ (ppm)
2.23
2.32
2.76/3.03
2.97
3.51
3.59/3.75
3.62
3.33
3.32
3.31
3.44
3.34
3.23
3.25
3.83
4.22
7.10
6.88
6.33
7.08
-
Peak
s
H number
Position
17
18
2, 3
1, 4
24
30
27
20
26
28
22
21
23
29
19
25
10
8
1
1
2
2
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
3
3
4
4
2
2
1
1
1
1
1
1
1
1
1
1
1
1
1
1
-
21.2
s
47.4
t
52.0
t
60.8
d
61.2
dd
m
m
m
m
m
o
68.5
69.4
71.0
73.7
76.0
76.4
76.9
d
81.4
o
94.4
d
104.3
120.3
124.6
125.6
126.1
127.7
128.5
132.1
134.0
136.5
140.0
142.3
150.0
d
o
t
d
7
o
9
-
11
15
13
6
CH
CH
C
7.10
7.24
-
o
1
1
-
br s
-
CH
C
7.35
-
d
1
-
16
14
12
5
-
C
-
-
-
C
-
-
-
The numbering of the carbon skeleton in glycosylamine 1 is shown in Figure 1.
o, overlapping signals; br s, broad singlet signal.

6 of 14
WANG ET AL.
spectra of the two compounds were acquired on
a
400 MHz
All the mixtures were heated at 80ꢀC for 24 h with shaking.
Aliquots of mixtures A and B were analyzed by HPLC and LC/MSn
(n = 1, 2) as described in sections 2.2.1 and 2.3, respectively. Aliquots
of mixtures C through G were analyzed by LC/MSn (n = 1, 2) as
described in section 2.3.
spectrometer (Agilent Technologies) at 25ꢀC. 1H and 13C resonances
were assigned and confirmed by the results from the following 2D
NMR experiments: gCOSY, gHSQC, and gHMBC.
2.9
|
Studies of the Maillard reactions of other
model amino compounds with reducing sugars
3
|
RESULTS AND DISCUSSION
To explore the general applicability of using the MS2 molecular
fingerprinting for differentiating glycosylamines and ARPs, the
Maillard reactions of vortioxetine with other reducing sugars were
also examined. Hence, vortioxetine (50 mg) was reacted with glucose
(240 mg) and galactose (240 mg), respectively, in solutions of 6 mL
water and 6 mL acetonitrile containing potassium carbonate (36 mg),
resulting in reaction mixtures A and B.
Glycosylamine 1 and ARP 2 degradants generated from the Maillard
reaction between vortioxetine and lactose are novel compounds that
have not been reported previously; they have been fully characterized
by LC/MS, 1H- and 13C-NMR in this study, which will be described in
detail below.
On the other hand, the Maillard reactions of other primary and
secondary amines with lactose or galactose were also explored.
Hence, 10 μL α-phenylethylamine and 280 mg lactose were dissolved
in 6 mL water and 6 mL acetonitrile, resulting in reaction mixture
C. Likewise, benzylamine and 1-phenylpiperazine were mixed with
lactose in solution in the same fashion, resulting in reaction mixtures
D and E, respectively. Finally, benzylamine (10 μL) was reacted with
glucose (165 mg) and galactose (165 mg), respectively, in solutions of
6 mL water and 6 mL acetonitrile, resulting in reaction mixtures
F and G.
3.1
|
LC/PDA-UV/MSn analysis of glycosylamine 1
Vortioxetine was reacted with lactose in order to obtain the
degradants of the Maillard reaction which was outlined in section 2.4.
In the UV chromatogram, a peak at 15.600 min was observed in
ꢁ31% yield, the accurate m/z value of which was 623.2598 (refer to
Figures S1 and S2, respectively, supporting information), consistent
with the formula of the adduct formed between vortioxetine and
lactose in protonated form, i.e., [C30H42N2O10S + H+], within an error
of −5.6 ppm. This adduct was suspected to be glycosylamine 1 as it
FIGURE 2 The key gHMBC correlations of ARP 2 formed between vortioxetine and lactose

WANG ET AL.
7 of 14
was the first product formed in the Maillard reaction and, thus, it
would usually be the main product observed in the early stage of the
Maillard reaction.
formed between the amino moiety of vortioxetine and the glucose
moiety of lactose (Table 1).
3.3
|
Conversion of glycosylamine 1 into ARP 2
3.2
|
Structural characterization of glycosylamine
In addition to ꢁ31% glycosylamine 1, the experiment described in
section 2.4 also produced a species that eluted at 17.863 min with a
yield of 1.48%. Its protonated ion was at m/z 623.2622, matching the
formula of [C30H42N2O10S + H+] within an error of −1.76 ppm, which
is the same as the molecular formula of glycosylamine 1. It was
suspected that the 17.863 min species is ARP 2. Nevertheless, its
content was too low to be isolated for NMR analysis. According to
the mechanism of the Maillard reaction, glycosylamine would
isomerize to ARP. Hence, conversion of glycosylamine 1 into ARP
2 was tried under different conditions including heating glycosylamine
1 in various solvents as well as in the solid state.
1 by 1D and 2D NMR
In order to confirm the structure of the degradant, approximately
10 mg of the isolated degradant was analyzed by 1D and 2D
NMR. The 13C-NMR spectrum showed that it has 30 carbon
atoms. The 1H-NMR and gHSQC spectra indicated that it contains
two methyl groups, six methylene groups, seventeen methine
groups, and five quaternary carbons, among which seven methine
groups are in the aromatic region. These results match the
structure of glycosylamine 1. More specifically, δH 2.76 (H-2,3) is
related to δC 94.4 (C-19) in the gHMBC spectrum (Figure 1),
further indicating that the degradant is glycosylamine 1 that was
The results showed that the yield of the suspected ARP 2 was
ꢁ3% in various solvents. This yield was still too low, which would
TABLE 2 1H-NMR and 13C-NMR
data of the Amadori rearrangement
product (ARP 2) formed between
vortioxetine and lactose
C δ (ppm)
20.2
C type
CH3
CH3
CH2
CH2
CH2
CH2
CH2
CH
CH
CH
CH
CH
CH
CH
C
C number
H δ (ppm)
2.23
2.32
2.98
2.75
3.51
3.48, 3.81
2.54, 2.62
3.88
3.66
3.62
3.39
3.33
3.38
3.78
-
Peak
s
H number
Position
17
18
1,4
2,3
23
30
24
29
26
20
22
28
21
27
19
25
10
8
1
1
2
2
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
1
3
3
4
4
2
2
2
1
1
1
1
1
1
1
-
20.8
s
51.5
m
m
o
54.5
60.4
62.4
d
62.4
d
66.3
m
m
m
o
67.5
68.2
70.6
73.0
o
75.5
o
77.8
dd
-
97.9
101.2
120.1
124.3
125.5
125.8
127.3
128.1
131.7
133.4
135.9
139.2
141.8
149.1
CH
CH
CH
CH
CH
C
4.26
7.13
6.89
6.35
7.09
-
d
1
1
1
1
1
-
d
t
d
7
t
9
-
12
15
13
6
CH
CH
C
7.08
7.23
-
d
1
1
-
br s
-
CH
C
7.33
-
d
1
-
16
14
11
5
-
C
-
-
-
C
-
-
-
The numbering of the carbon skeleton in ARP 2 is shown in Figure 1.
o, overlapping signals; br s, broad singlet signal.

8 of 14
WANG ET AL.
make the preparation and isolation of ARP 2 quite difficult. On the
other hand, heating glycosylamine 1 in the solid state at 80ꢀC for 24 h
generated the suspected ARP (2; the 17.867 min peak) in ꢁ25% of
isolated yield. The m/z value of the protonated ion was 623.2606
(refer to Figures S3 and S4, respectively, supporting information),
matching the formula of [C30H42N2O10S + H+] within an error of
−4.3 ppm, which is consistent with the protonated ARP 2 and
isomeric with the protonated glycosylamine 1.
spectrum (Figure 2), δH 2.54/2.62 (H-24), δC 54.5 (C-2, 3), and δC
97.9 (C-19) are related, indicating that the degradant at 17.867 min is
indeed ARP 2 (Table 2). It needs to be pointed out that ARP 2 is in the
cyclized form of pyranose.
3.5
|
N-Formyl vortioxetine: Degradation product
of ARP
In addition to ARP 2, the solid sample of glycosylamine also
decomposed to give back vortioxetine in ꢁ17.6% yield and a peak
at 24.920 min in ꢁ4% yield in the course of 80ꢀC heating for 24 h.
The protonated ion of the 24.920 min species showed an ion at m/z
327.1522, matching a formula of [C19H22N2OS + H+]. The latter
formula has an extra CO group than that of vortioxetine, suggesting
it is likely N-formyl vortioxetine, which was later confirmed via
NMR analysis of the isolated impurity.23 It has been reported that
this type of N-formyl degradant can originate from the Maillard
reaction.24
3.4
|
Structural characterization of ARP 2 by 1D
and 2D NMR
In order to confirm that the degradant is ARP 2, approximately 10 mg
of the isolated degradant was analyzed by 1D and 2D NMR. The 13C-
NMR spectrum showed 30 carbon atoms: two methyl groups, seven
methylene groups, fifteen methine groups, and six quaternary
carbons, among which seven methine groups are in the aromatic
region. Its overall structure is quite similar to that of glycosylamine,
except for the following differences: it has one more methylene group
and quaternary carbon, but one less methine group. In its gHMBC
All the Maillard reaction pathways described above are
summarized in Scheme 1.
FIGURE 3 Top: CID-MS2 of glycosylamine 1. The abundance of m/z 341 was much greater than the abundance of m/z 311 and the ratio of
m/z 341 to m/z 311 was ꢁ32. Bottom: CID-MS2 of ARP 2. The abundance of m/z 311 was greater than the abundance of m/z 341 and the ratio
of m/z 311 to m/z 341 was ꢁ11

WANG ET AL.
9 of 14
3.6
|
Differentiation of glycosylamine 1 and ARP
the protonated ion of vortioxetine. The most abundant fragment is
m/z 341, which appears to stem from β-fission of the open-ring form
(Schiff base) of glycosylamine (Pathway a, Scheme 2). Loss of
formaldehyde from m/z 341 would produce the weak m/z
311 fragment. A plausible mechanism for the fragmentation pathways
of glycosylamine 1 is proposed in Scheme 2.
2 via their MS2 fingerprints obtained from high-
resolution LC/MS2
3.6.1
|
CID-MS2 fragmentation/fingerprinting of
the precursor ions of glycosylamine 1 and ARP 2
The CID-MS2 spectra of the protonated ARP (2, Figure 3,
It should be noted that glycosylamine 1 and ARP 2 are isomers and it
would be very difficult to distinguish them by single-stage mass
spectrometry unless the analysis was performed by an MS instrument
equipped with ion mobility capability. The molecular structures of
glycosylamine 1 and ARP 2 were established in the previous sections
by 1D and 2D NMR. The main focus of this paper is to differentiate
between these two isomers via their low-energy CID MS/MS
analyses and the establishment of their unique fingerprinting.
Collision-induced decomposition MS2 (CID-MS2) of the
protonated glycosylamine (1, Figure 3, top), obtained at a collision
energy (CE) of 30 eV, exhibited several major product ions at m/z
605, 461, 341, and 299, among which the m/z 299 ion coincides with
bottom), obtained at a CE of 30 eV, showed the following
major product ions at m/z 605, 587, 498, 443, 311, and 296. The
abundance of the m/z 605 ion is particularly high, which is 18 less
than the m/z 623 ion, suggesting the presence of
a hydroxyl
group that is particularly easy to cleave from ARP 2. This observation
appears to be consistent with the presence of a hydroxyl group at
position-19 of ARP 2 (Pathway a, Scheme 3). The m/z 587 ion is
formed from the double dehydration of the protonated ARP 2. The
m/z 443 ion seems to result from the cleavage of m/z 461 by loss of
H2O. The m/z 299 fragment, which should correspond to protonated
ARP 2, is also present. A plausible mechanism for the fragmentation
pathways of ARP 2 is proposed in Scheme 3.
SCHEME 2 Proposed fragmentation pathways of the precursor protonated glycosylamine 1 (m/z 623)

10 of 14
WANG ET AL.
SCHEME 3 Proposed fragmentation pathways of the precursor protonated ARP 2 (m/z 623)
3.6.2
ARP 2
|
Characteristic ions in glycosylamine 1 and
the two MS2 spectra can be observed: (1) the abundance of m/z
341 is much greater than the abundance of m/z 311 in
glycosylamine 1 and the ratio of m/z 341 to m/z 311 is ꢁ32.
(2) The abundance of m/z 311 is greater than the abundance of
m/z 341 in ARP 2 and the ratio of m/z 311 to m/z 341 is ꢁ11.
Based on the above fragments observed and structures of the two
isomers, the probable structures of m/z 341 (Figure 4A) and
311 (Figure 4B) can be proposed. For the predominant formation
of the m/z 341 fragment from the Schiff base (the open-ring form
of glycosylamine 1), it apparently can be attributed to the facile
cleavage of the C-2–C-3 bond (of the glucose moiety) via a retro-
aldo mechanism of the Schiff base; this cleavage is analogous to
After carefully comparing the MS2 spectra, we found that the
product ions of the obtained glycosylamine 1 and ARP 2 were
very similar, except for some differences in abundance. As shown
in Figure 3, most of the product ions of glycosylamine 1 are the
same as those of ARP 2. The common product ions observed are
m/z 605 (loss of H2O), 587 (loss of two H2O), 461 (loss of
C6H11O5), 443 (loss of H2O and C6H11O5), and two smaller
fragments (m/z 299 and 256) containing the core structure of
vortioxetine. Nevertheless, two characteristic phenomena between

WANG ET AL.
11 of 14
FIGURE 4 Characteristic structures of the
prominent product ions from protonated
glycosylamine 1 (m/z 341) and protonated ARP
2 (m/z 311) generated from vortioxetine and
lactose
the one in the Schiff base formed between glycine and glucose as
described by Xing et al.21 On the other hand, the preferential
formation of the m/z 311 fragment (versus that of the m/z
341 fragment) from ARP 2 can be attributed to the α-fission of
the C-1–C-2 bond of the glucose moiety.21,24–25
Based on the MS2 fingerprints of the two degradants (1, 2)
obtained, two characteristic MS2 fragments were observed: one is
[Mvort + C2H3O]+ at m/z 341 (Mvort is the vortioxetine core), while the
other is [Mvort + CH]+ at m/z 311. The fragment of [Mvort + C2H3O]+
at m/z 341, which will be referred to as the C2 fragment, is much
more prominent from protonated glycosylamine 1, while the fragment
of [Mvort + CH]+ at m/z 311, which will be referred to as the C1
fragment, is much more prominent from protonated ARP 2.
abundant fragments can often be obtained. In fact, we have also
examined CEs other than 30 eV during the MS2 experiments, such
as 20 eV and 40 eV. At 20 eV, the ratio of C2 to C1 in
glycosylamine 1 is 47, while the ratio of C1 to C2 in ARP 2 is 12.
At a CE of 40 eV, the ratio of C2 to C1 in glycosylamine 1 is
12 and the ratio of C1 to C2 in ARP 2 is 13.
The results are summarized in Table 3, along with other model
systems (the results of the latter will be discussed in section 3.8). It
can be clearly seen that the ratio of C2/C1 in the glycosylamines
decreases as the CE increases for all the model systems examined. On
the other hand, no such clear trend can be observed for the ratio of
C1/C2 in the ARPs examined.
In addition, we also performed a sub-set of the vortioxetine
experiments (with three reducing sugars) on the Shimadzu LC/IT-TOF
mass spectrometer, an MS instrument with very different
configuration. The results show that the same rule prevails: i.e., the C2
fragment is significantly higher than C1 in the glycosylamines, while
the C1 fragment is significantly higher than C2 in the ARPs (Table 4).
The above phenomena indicate that glycosylamines and ARPs
can respectively produce C1 and C2 fragments with consistent
3.7
|
Results from using different CEs and different
MS instruments
The above results were obtained using a collision energy of 30 eV
on the Agilent LC/QTOF mass spectrometer, and, at this CE,
TABLE 3 The C2/C1 ratio in MS2 of protonated glycosylamines (1, 3, 5, 7, 9, 11) and C1/C2 ratio in MS2 of protonated ARPs (2, 4, 6, 8, 10,
12) at different collision energies
C2/C1 (glycosylamine)
C1/C2 (ARP)
Amine compound
Vortioxetine
Reducing sugar
Glucose
20 eV
42 (3)
41 (5)
47 (1)
29 (7)
61 (9)
6.0 (11)
30 eV
11 (3)
12 (5)
32 (1)
13 (7)
12 (9)
3.1 (11)
40 eV
5.3 (3)
4.2 (5)
12 (1)
20 eV
6.0 (4)
11 (6)
30 eV
8.5 (4)
10 (6)
40 eV
11 (4)
Vortioxetine
Galactose
Lactose
15 (6)
Vortioxetine
12 (2)
11 (2)
13 (2)
α-Phenylethylamine
1-Phenylpiperazine
Benzylamine
Lactose
6.5 (7)
7.0 (9)
1.8 (11)
20 (8)
11 (8)
5.2 (8)
4.5 (10)
8.0 (12)
Lactose
1.4 (10)
15 (12)
2.0 (10)
8.5 (12)
Lactose
All data were obtained on the Agilent LC/QTOF instrument (refer to section 2.3).
TABLE 4 The C2/C1 ratio in MS2 of
protonated glycosylamines (1, 3, 5) and
C1/C2 ratio in MS2 of protonated ARPs
C2/C1 (glycosylamine)
C1/C2 (ARP)
Amine compound
Vortioxetine
Reducing sugar
QTOF-MS
IT-TOF-MS
16 (1)
QTOF-MS
11 (2)
IT-TOF-MS
20 (2)
(2, 4, 6) on different MS instruments
Lactose
32 (1)
11 (3)
12 (5)
Vortioxetine
Glucose
Galactose
21 (3)
8.5 (4)
5.0 (4)
Vortioxetine
19 (5)
10 (6)
10 (6)
For the analytical methods, refer to section 2.3.

12 of 14
WANG ET AL.
TABLE 5 C2/C1 and C1/C2 ratios of fragments in glycosylamines (all compounds with odd numbers) and ARPs (all compounds with even
numbers) of the eight model systems
C2/C1
Reducing
sugar
(glycosyl
amine)
C1/C2
(ARP)
Model
Amine compound
C1 fragment
C2 fragment
Primary amine +
disaccharide
α-Phenylethylamine Lactose
13 (7)
11 (8)
Benzylamine
Benzylamine
Benzylamine
Vortioxetine
Lactose
Glucose
Galactose
Lactose
3.1 (11)
3.0 (13)
2.1 (15)
32 (1)
8.5 (12)
5.0 (14)
5.0 (16)
11 (2)
Primary amine +
monosaccharide
Secondary amine +
disaccharide
1-Phenylpiperazine Lactose
12 (9)
2.0 (10)

WANG ET AL.
13 of 14
TABLE 5 (Continued)
C2/C1
Reducing
sugar
(glycosyl
amine)
C1/C2
Model
Amine compound
Vortioxetine
C1 fragment
C2 fragment
(ARP)
Secondary amine +
monosaccharide
Glucose
11 (3)
8.5 (4)
Vortioxetine
Galactose
12 (5)
10 (6)
All data were obtained on the Agilent LC/QTOF mass spectrometer (refer to section 2.3). The data for benzylamine with glucose or galactose were
obtained at a CE of 20 eV, and the data for other model systems were obtained at a CE of 30 eV.
preference for one fragment over the other, under different CEs
and on different instruments. Therefore, it can be envisioned that
the preference for either the C1 or C2 fragment, i.e., the ratio of
C1/C2 or C2/C1, can be utilized to distinguish glycosylamines
from ARPs.
C1/C2 ranged from 2.0 to 11 in the ARPs produced by the seven
model systems.
4
|
CONCLUSIONS
In this study, we prepared and fully characterized the glycosylamine
and Amadori rearrangement products formed between vortioxetine
and lactose, two impurities that have not been reported previously.
The MS2 fingerprints of the two degradants display significantly
different profiles, despite the fact that many common fragments are
3.8
|
Analysis of the C1 and C2 fragments in
glycosylamines and ARPs from other model systems
In order to verify whether the preference for either the C1 or C2
fragment is limited to vortioxetine or may be universally applicable to
other primary and secondary amines, we replaced vortioxetine with
observed. Specifically, the protonated glycosylamine shows
a
prominent characteristic fragment of [Mvort + C2H4O]+ at m/z
341 (C2 fragment), while the protonated ARP shows a prominent
characteristic fragment of [Mvort + CH]+ at m/z 311 (C1 fragment).
Further study of the Maillard reactions between several other
structurally diverse primary/secondary amines and lactose, as well as
other reducing sugars, produced similar patterns.
α-phenylethylamine,
1-phenylpiperazine,
and
benzylamine,
respectively, in the Maillard reaction with lactose. Likewise, we also
replaced lactose with glucose and galactose, respectively, in the
Maillard reaction with vortioxetine (refer to section 2.9). The MS2
results are shown in Figures S5–S11 (supporting information) and the
C2/C1 ratios in the glycosylamines and C1/C2 ratios in the ARPs are
summarized in Table 5. Again, the same rule prevails: the abundance
of the C2 fragment is much greater than the abundance of C1 in the
glycosylamines and the ratio of C2/C1 ranged from 2.1 to 32 in the
seven model systems. On the other hand, the abundance of the C1
fragment is greater than the abundance of C2 in ARPs and the ratio of
The study suggests that the characteristic C1 and C2 fragments
and their ratios may be used to differentiate the glycosylamine and
ARP degradants, the two isomeric degradants of the Maillard reaction,
which are commonly encountered in finished dosage forms of
pharmaceutical products containing primary and secondary amine
APIs. Our results are consistent with those that have just been

14 of 14
WANG ET AL.
published by Xing et al,21,22 in which they reported that MS/MS
fragmentation patterns can be used to discriminate between glucose-
derived Schiff bases, Amadori, and Heyns compounds with several
amino acids. Their work as well as ours indicates that this
methodology can be generally applicable for differentiating the early-
stage isomeric Maillard products from amino compounds of vastly
different structures with reducing sugars.
13. Jin J, Wang Z, Lin J, Zhu W, Gu C, Li M. “Ghost peaks” of ezetimibe:
Solution degradant products of ezetimibe in acetonitrile induced by
alkaline impurities from glass HPLC vials. J Pharm Biomed Anal. 2017;
140:281-286.
14. Lin J, Lin Z, Li D, et al. “Ghost peak” of clofazimine: A solution
degradant product of clofazimine via nucleophilic substitution by
nitrite leaching from certain glass HPLC vials. J Pharm Biomed Anal.
2018;150:183-190.
15. Lin J, Huang T, Feng M, et al. Solution degradant of mirabegron
extended release tablets resulting from
a Strecker-like reaction
PEER REVIEW
between mirabegron, minute amounts of hydrogen cyanide in
acetonitrile, and formaldehyde in PEG during sample preparation.
J Pharm Biomed Anal. 2019;168:181-188.
The peer review history for this article is available at https://publons.
com/publon/10.1002/rcm.9062.
16. Lin J, Liu X, Wang J, et al. An artifactual solution degradant of
pregabalin due to adduct formation with acetonitrile catalyzed by
alkaline impurities during HPLC sample preparation. J Pharm Biomed
Anal. 2019;175:112788.
ORCID
Min Li
https://orcid.org/0000-0001-8690-3451
17. Tian Y, Lin J, Chen F, et al. Structure elucidation and formation
mechanistic study of
a methylene-bridged pregabalin dimeric
REFERENCES
degradant in pregabalin extended-release tablets. Int J Pharm. 2019;
575:118910.
1. Gandhi S, Ren D, Xiao G, et al. Elucidation of degradants in acidic
peak of cation exchange chromatography in an IgG1 monoclonal
antibody formed on long-term storage in a liquid formulation. Pharm
Res. 2012;29(1):209-224.
18. Li M, Wang X, Chen B, et al. Use of liquid chromatography/tandem
mass spectrometric molecular fingerprinting for the rapid structural
identification of pharmaceutical impurities. Rapid Commun Mass
Spectrom. 2010;23(22):3533-3542.
2. Bhattacharya A, Cui Y. A GPU-accelerated algorithm for biclustering
analysis and detection of condition-dependent coexpression network
modules. Sci Rep. 2017;7(1):4162.
19. Connolly KR, Thase ME. Vortioxetine: A new treatment for major
depressive disorder. Expert Opin Pharmacother. 2016;17(3):421-431.
20. Lovdahl MJ, Hurley TR, Tobias B, Priebe SR. Synthesis and
characterization of pregabalin lactose conjugate degradant products.
J Pharm Biomed Anal. 2002;28(5):917-924.
3. Abdoh A, Al-Omari MM, Badwan AA, Jaber AM. Amlodipine
besylate–excipients interaction in solid dosage form. Pharm Dev
Technol. 2004;9(1):15-24.
4. Murakami T, Fukutsu N, Kondo J, Kawasaki T, Kusu F. Application of
liquid chromatography-two-dimensional nuclear magnetic resonance
spectroscopy using pre-concentration column trapping and liquid
chromatography–mass spectrometry for the identification of
degradant products in stressed commercial amlodipine maleate
tablets. J Chromatogr A. 2008;1181(1–2):67-76.
5. Serajuddin ATM, Thakur AB, Ghoshal RN, et al. Selection of solid
dosage form composition through drug–excipient compatibility
testing. J Pharm Sci. 1999;88(7):696-704.
21. Xing H, Mossine VV, Yaylayan V. Diagnostic MS/MS fragmentation
patterns for the discrimination between Schiff bases and their
Amadori or Heyns rearrangement products. Carbohydr Res. 2020;491:
107985.
22. Xing H, Yaylayan V. Mechanochemical generation of Schiff bases and
Amadori products and utilization of diagnostic MS/MS fragmentation
patterns in negative ionization mode for their analysis. Carbohydr Res.
2020;491:108091.
23. Lin J, Huang T, Lan F, et al. N-Formyl vortioxetine and preparation
method therefore and solid preparation of vortioxetine: CN, WO
2019/056211 A1. 2019-03-28.
6. Hodge JE. Dehydrated foods – chemistry of browning reactions in
model systems. J Agric Food Chem. 1953;1(15):625-651.
7. Arena S, Renzone G, D'Ambrosio C, Salzano AM, Scaloni A. Dairy
products and the Maillard reaction: A promising future for extensive
food characterization by integrated proteomics studies. Food Chem.
2017;219:477-489.
24. Hoaglund HCS, Williamson ML, Jansen PJ, et al. Mechanistic studies
of the N-formylation of edivoxetine, a secondary amine-containing
drug, in
a solid oral dosage form. J Pharm Sci. 2017;106(5):
1218-1238.
8. Li M. Organic chemistry of drug degradation. In: Drug–Excipient
Interactions and Adduct Formation. Cambridge: Royal Society of
Chemistry; 2012.
SUPPORTING INFORMATION
9. Li M, Conrad B, Maus RG, et al. A novel oxidative degradation
pathway of indomethacin under the stressing by hydrogen peroxide.
Tetrahedron Lett. 2005;46(20):3533-3536.
Additional supporting information may be found online in the
Supporting Information section at the end of this article.
10. Li M, Wang X, Chen B, Chan TM, Rustum A. Forced degradation of
betamethasone sodium phosphate under solid state: Formation,
How to cite this article: Wang S, Lin J, Li D, et al. Study of the
isomeric Maillard degradants, glycosylamine and Amadori
rearrangement products, and their differentiation via MS2
fingerprinting from collision-induced decomposition of
protonated ions. Rapid Commun Mass Spectrom. 2021;35:
e9062. https://doi.org/10.1002/rcm.9062
characterization,
and
mechanistic
study
of
all
four
17,20-diastereomers of betamethasone 17-deoxy-20-hydroxy-21-oic
acid. J Pharm Sci. 2009;98(3):894-904.
11. Wang X, Peng Y, Li M, Rustum AM. Use of high resolution
LC–MSn analysis in conjunction with mechanism-based stress
studies: Identification of asarinin, an impurity from sesame oil in
an animal health product.
1015-1021.
J Pharm Biomed Anal. 2009;50(5):
12. Li M, Chen B, Monteiro S, Rustum AM. Mechanism of base-catalyzed
autooxidation of corticosteroids containing 20-keto-21-hydroxyl side
chain. Tetrahedron Lett. 2009;50(32):4575-4581.