Journal of Chromatography A, 1207 (2008) 29–37

Contents lists available at ScienceDirect

Journal of Chromatography A
journal homepage: www.elsevier.com/locate/chroma

Simultaneous determination of 13 aminoglycoside residues in foods of animal

origin by liquid chromatography–electrospray ionization tandem mass
spectrometry with two consecutive solid-phase extraction steps
Wei-xia Zhu a,b , Ji-zhou Yang a,∗ , Wei Wei a , Ya-feng Liu a , Shu-sheng Zhang b
a
Henan Entry–Exit Inspection and Quarantine Bureau of China, Zhengzhou 450003, China
b
Department of Chemistry, Zhengzhou University, Zhengzhou 450001, China

a r t i c l e i n f o a b s t r a c t

Article history: A confirmatory and quantitative method based on liquid chromatography–electrospray ionization tandem
Received 29 April 2008 mass spectrometry (LC–ESI-MS/MS) has been developed for simultaneous determination of 13 aminogly-
Received in revised form 31 July 2008 coside antibiotics in various samples. The aminoglycoside analytes were released and extracted from
Accepted 11 August 2008
different matrices with 5% trichloroacetic acid. The influence of pH values on the solid-phase exaction
Available online 14 August 2008
(SPE) procedure has been studied. Due to different pKa values of the compounds, seven aminoglyco-
sides (AGs) were quantitatively retained on Oasis HLB cartridges at pH < 1 and then six aminoglycosides
Keywords:
were retained at pH 8.5. Thus, the combination of two HLB SPE cartridges with different pH values was
Aminoglycosides
Animal-origin foods
involved to simultaneously purify 13 aminoglycosides. The proposed two SPE steps produced high recov-
LC–MS/MS ery yields for every aminoglycoside in five different matrices. The LC–MS/MS method was validated
Solid-phase extraction according to the European Union Commission directive 2002/657/EC. Good performance characteris-
Residue analysis tics were obtained for recovery, precision, calibration curve, stability, specificity, decision limits (CC␣)
Method validation and detection capabilities (CC␤) in different matrices. The optimized procedure has been successfully
applied to real samples in our laboratories (n ≥ 200) for 1 year. It demonstrated that the new method was
robust and useful for identification and quantification of 13 aminoglycosides residues in foods of animal
origin.
© 2008 Elsevier B.V. All rights reserved.

1. Introduction able analytical methods, which monitor residues of trace-level AGs

Aminoglycosides (AGs) are a class of broad-spectrum antibiotics
that have bactericidal activity against some aerobic gram-positive
and gram-negative organisms [1,2]. The structures of these com-
pounds contain two or more aminosugars linked by glycosidic
bonds to an aminocyclitol component. Except for streptomycin and
dihydrostreptomycin having a streptidine component, the cyclitol
is usually 2-deoxystreptamine [3,4]. The AGs have been exten-
sively employed in animal husbandry for the treatment of bacterial
infections or growth promotion [5]. Due to their toxicity and pos-
sible antibiotic resistance, considerable attention has been paid to
the potential human health risk. The European Union (EU), China,
the USA, Japan, and other countries have issued strict maximum
residue levels (MRLs) for nine AGs in various animal-origin foods.
Consequently, there is a growing need to develop sensitive and reli-
∗ Corresponding author. Tel.: +86 371 65685749; fax: +86 371 65685841.
E-mail address: yangjz@haciq.gov.cn (J.-z. Yang).
in complex matrices.
Many analytical methods have been described in the literature
for determination of AGs in biological samples or pharmaceuti-
cals. Immuno-assays consisting of enzyme-linked immunoassay,
radioimmunoassay, and fluoroimmunoassay are only used as
screening methods [3,6]. Thin-layer chromatography has com-
paratively poor separation and cannot meet current needs of
multi-residue analysis [7]. Aminoglycosides are characterized by
thermal stability and non-volatility, requiring lengthy derivatiza-
tion time using GC or GC–MS [8]. Without the chromophores and
fluorophores, AGs are not amenable to direct ultraviolet (UV) or
fluorescence detection (FLD). LC analysis often entails pre- or post-
column derivatization, followed by the detection with UV or FLD
[1,9–11]. The methods result in poor reproducibility caused by
instability and low yield of derivatives. The pulsed amperomet-
ric detection [5], evaporative light scattering detection [12–14]
and chemiluminescence detection [15] without any derivatization
procedure can directly detect the AGs, but fail to satisfy the require-
ment of identification and sensitivity. LC–MS/MS is becoming more
common in the analysis of antibiotics residue because of its high

0021-9673/$ – see front matter © 2008 Elsevier B.V. All rights reserved.
doi:10.1016/j.chroma.2008.08.033
30 W.-x. Zhu et al. / J. Chromatogr. A 1207 (2008) 29–37
sensitivity and ability to provide confirmation. However, most of
the reported MS-based methods in the previous literatures only
allow the determination of one or two analytes [16–20] and there
are few reports for simultaneously determining multi-residues
AGs in products of animal origin [2,4,21–23]. In 2004, Kotretsou
reviewed the application of MS/MS for determination of AGs in
food and covered up to eight AGs [24]. Hammel et al. established
an approach of 42 antibiotic residues, including 3 AGs [25]. AGs
are extremely hydrophilic compounds due to having many amino
and hydroxyl groups in their chemical structures. Oertel et al. [4]
proposed hydrophilic interaction chromatography (HILIC)–MS/MS
to determine six aminoglycosides in serum. The HILIC–MS method
has great advantages in comparison with others, such as a more
volatile mobile phase and less ion suppression. Meanwhile, AGs
have unfavorable properties for developing a reversed-phase LC
method, such as binding silanol groups in the stationary phase and
high polarity. However, ion-pair chromatography (IPC) with MS/MS
served as a powerful tool widely used for multi-groups separa-
tion of aminoglycosides [21–23]. Volatile heptafluorobutyric acid
(HFBA) serving as an ion-pair reagent is compatible with mass spec-
trometry and can give rise to strong retention on the reversed-phase
column. The technology has been widely used in determination of
AGs.
The critical challenge for a valid determination of trace-level AGs
in complicated matrices is linked with the extraction and clean-
up procedure. Various purification methods including solid-phase
extraction (SPE) [21,22], on-line SPE [23], and matrix solid-phase
dispersion extraction [2] have been employed in simultaneously
extracting several aminoglycoside drugs. According to the char-
acter of polybasic cations and polarity, SPE adsorbents including
cation-exchange, C18, and hydrophilic–lipophilic balance have
been widely used to purify AGs. Kaufmann and Maden [22] reported
a method of determining eleven AGs, and cleanup was achieved by
three steps included the use of an anion-exchange resin, a HLB car-
tridge, and a weak cation SPE. The procedures are complicated and
time-consuming. Based on that method, Babin and Fortier [23] also
established an automated SPE cleanup to determine three AGs in
veal tissues. This high-throughput analytical method required an
extra pump and a multi-port valve system. The proposed methods
need to make use of two or more different sorbent materials, which
results in an overall time-consuming sample preparation, and many
losses can occur in any of these steps. From the environmental point
of view, it is important to minimize the use of undesirable organic
solvents in the determination of AGs. Additionally, the methods
included only one or two matrices. In fact, many matrices of animal
origin in the monitoring and surveillance program were controlled,
so it would be very preferable to be able to analyse several differ-
ent matrices with the same method. To our knowledge, there is
not any LC–MS/MS method for the determination and validation of
13 aminoglycosides residues in several animal tissues, honey, and
milk.
Therefore, the objective of the current study was to develop a
multi-residue method that would be simple and fast for routine
regulatory analysis of the 13 aminoglycosides residues in differ-
ent matrices. Fortunately, an effective SPE method was applied
to cleanup the high polarity of 13 AGs, which replaced the labor-
intensive and time-consuming steps. In addition, not only can the
developed IPC analysis be compatible with MS confirmation, but
resolved the separation of hygromycin B from other components,
which was previously regarded as impossible [24]. In this work,
we presented for the first time ultra-trace analysis of 13 AGs in
complex samples such as muscle, liver, kidney, honey, and milk.
Moreover, this is also the first report of validation according to the
EU Commission Decision 2002/657/EC guidelines for aminoglyco-
side antibiotics residues [26].
2. Experimental
2.1. Chemicals and reagents
Unless indicated otherwise, analytically pure substances and
HPLC-grade reagents were used. Tobramycin (TOBRA), neomycin
(NEO), paromomycin (PARO), gentamicin (consisting of GENTC1,
GENTC2, GENTC1a), apramycin (APRA), hygromycin B (HYGRO)
(sulfates, ≥92%) were obtained from Dr. Ehrenstorfer (Augsburg,
Germany). Kanamycin (KANA), amkacin (AMKI) (≥98%) were pur-
chased from Sigma–Aldrich (St. Louis, MO, USA). Streptomycin
solution (STREP), dihydrostrepmycin (DISTREP) solution, spectino-
mycin solution (SPEC) (1 mg/mL), and heptafluorobutyric acid were
supplied by Fluka (Buchs, Switzerland). Acetonitrile, methanol,
glacial acetic acid (99%), and formic acid (50%) were obtained from
Fisher (Bar-Bel, France). Analytical grade trichloroacetic acid (TCA)
and ammonium hydroxide (28%) were supplied by Beijing Chemical
Company (Beijing, China). Deionized water (18 M cm) was gen-
erated by a Milli-Q water-purification system (Millipore, Bedford,
MA, USA). The cartridges used for solid-phase extraction were Oasis
HLB cartridges (3 mL/60 mg) from Waters (Milford, MA, USA). Fil-
ter membranes (0.22 ␮m) provided by Agilent (Palo Alto, CA, USA)
were used to filter the extracts before the injection into chromato-
graphic system.
2.2. Standard solutions
Standard diluted solution was mixed with acetonitrile/
water/acetic acid (20:78:2, v/v/v). Individual stock solution of
aminoglycosides (100 ␮g mL−1 ) was prepared in diluted solution
and was stable for 6 months when stored in plastic tube at 2–4 ◦ C.
Tuning solution of each analyte (5 ␮g mL−1 ) was prepared by
diluting individual stock solution with diluted solution. A work-
ing solution (100 times MRL for MRL substances or 100 times
CC␣ for no MRL substances), a standard mixture used to fortify
the samples, was prepared by diluting individual stock solution
with diluted solution, and when lower fortification mixture was
needed, an extra dilution of the AGs was prepared. These solutions
were stored in plastic tubes at 2–4 ◦ C and remain stable for up to
1 month.
2.3. Samples
All samples including pork muscle, pork kidney, pork liver,
chicken muscle, chicken kidney, chicken liver, swine muscle, swine
kidney, swine liver, honey, milk were collected from supermarkets
and local markets in Zhengzhou (Henan, China). Typically, 500 g
tissue samples were first minced using a kitchen homogenizer and
stored at −18 ◦ C until experiment. Honey was mixed after thawing
at 50–60 ◦ C. Special control was taken in sample handling proce-
dures to prevent the possibility of accidental contamination or loss
of analytes.
2.4. Samples-preparation
2.4.1. Tissue samples
Five grams of aliquot of a sample was weighed into a polypropy-
lene centrifuge tube and fortified with the working solution at
appropriate concentrations. Ten milliliters of 5% TCA (w/v) was
added to the centrifuge tube. The mixture was homogenized thor-
oughly for 1 min by means of an Ultra-Turrax T-25 homogenizer
(Jankle & Kunkel, IKA-Labortechnik, Staufen, Germany) and then
centrifuged at 1000 × g for 5 min. The same extraction procedure
was repeated again with 10 mL of 5% TCA, and TCA supernatants
were combined into another centrifuge tube. A 5-mL volume of
 W.-x. Zhu et al. / J. Chromatogr. A 1207 (2008) 29–37
0.2 M HFBA was added to these extracts. After vortex-mixing for
1 min and centrifuging at 1000 × g for 2 min, the upper layer
extracts would be ready for the clean-up procedure.
2.4.2. Honey and milk samples
Five grams of aliquot of sample was weighed into a polypropy-
lene centrifuge tube and fortified with the working solution at
appropriate concentrations. Fifteen milliliters of 5% TCA and 5 mL
of 0.2 M HFBA were added to the centrifuge tube. The mixture was
vortexed for 3 min and then centrifuged at 1000 × g for 5 min, which
would then be ready for the clean-up procedure.
2.5. Solid-phase extraction
A HLB cartridge was preconditioned subsequently with 3 mL
methanol, 3 mL water, and 3 mL 20 mM HFBA. Five milliliters of
the previous extract was transferred onto the cartridge at the flow-
rate of 1 mL min−1 . After penetration, the cartridge was dried by
a vacuum pump (Visiprep vacuum manifold, Supelco, USA). All the
effluent was collected into another tube and adjusted to pH 8.5 ± 0.2
with 10% ammonium hydroxide (about 2 mL). Another HLB car-
tridge was preconditioned with the same step described above.
Then, the collected solution was loaded onto the column at the
flow-rate of 1 mL min−1 . After the sample loading, two HLB car-
tridges were hyphenated with vacuum joints. The columns were
rinsed with 5 mL water, then dried at less than 15 mmHg for
5 min. AGs residues were finally eluted with 6 mL acetonitrile/0.2 M
HFBA (8:2, v/v) and the elution was evaporated to 0.3 mL under a
steam of gentle nitrogen at 40 ◦ C. Finally, the residue was recon-
stituted to 1 mL with 20 mM HFBA. The resulting solution was
filtered through a 0.22 ␮m membrane filter and collected into an
LC autosampler vial, which was subjected to LC–ESI–MS/MS anal-
ysis.
2.6. LC–MS/MS analysis
LC was performed using an Agilent 1100 HPLC system
(Waldbronn, Germany) equipped with an automatic degasser, a
quaternary pump, and an autosampler. Chromatographic sepa-
ration was carried out using a Capcell Pak C18 UG120 column
(150 mm × 2.0 mm, particle size 5 ␮m, Shiseido, Chuo-ku, Tokyo,
Japan) at 30 ◦ C. The flow-rate of mobile phase was maintained at
0.3 mL min−1 and the injection volume was 10 ␮L. The mobile phase
A was acetonitrile/water (50:950, v/v) containing 20 mM HFBA and
mobile phase B was acetonitrile/water (500:500, v/v) containing
20 mM HFBA. A gradient elution program was started with 20% of
B, increased to 95% of B at 9 min, kept at 95% of B for 1 min, returned
to initial composition at 10.1 min, and held for 5 min to equilibrate
the column.
The HPLC system was connected to an API 4000 triple
quadrupole tandem mass spectrometer (Applied Biosystems/MDS
Sciex, Toronto, Canada) equipped with an ESI ion source. The opti-
mization of mass condition was achieved on infusing injection of
each compound separately at a flow rate of 10 ␮L min−1 . The source
block temperature was set at 550 ◦ C in the positive ion mode with
a capillary voltage 5000 V. Nitrogen gas was used as nebulizer gas
(40 psi), auxiliary gas (45 psi), collision gas (7 psi), and curtain gas
(15 psi). Detection was operated in the multiple reaction monitor-
ing (MRM) mode and the acquisition of two transitions made it
possible to obtain at least three identification points as required
by the guideline of 2002/657/EC. The higher intensity transition
was selected for the quantitative purpose and the resolution was
set at unit (0.7 U). The instrument control and data acquisition
were carried out by the analyst 1.4.1 software. Mass parameters
for each analyte including precursor ion (Q1), product ion (Q3),
31
dwell time, declustering potential (DP), entrance potential (EP),
cell exit potential (CXP), collision energy (CE) were summarized in
Table 1.
2.7. Method validation
Validation of the method strictly complied with the
2002/657/EC guideline for confirmation analysis procedure.
According to the criteria, the performance characteristics of a con-
ventional method include recovery, repeatability, reproducibility,
decision limit, detection capability, calibration curves, stability,
and specificity. We utilized muscle, kidney, liver, honey, milk for
full validation purposes and obtained validation data for these
matrices.
2.7.1. Calibration curves
Negative pork muscle, chicken liver, bovine kidney, honey, milk
were used as matrix samples. The matrix-match calibration stan-
dard curves were made by fortified with six levels of 0, 0.5, 1.5,
2, 5, and 10 times MRL. On the other hand, for compound (AMKI,
TOBRA, HYGRO) no MRL or special matrices no MRL values 0, 1,
1.5, 2, 5, and 10 times CC␣ should be spiked. The samples spiking
with aminoglycosides were operated with complete extraction and
purification procedure. The calibration curves were constructed
using a peak area from six concentration levels versus the con-
centration of analytes. Thus, there were all together 65 different
matrix-match calibration curves.
2.7.2. Recovery and precision
Pork muscle, chicken liver, bovine kidney, honey, milk known to
be noncompliant served as blank matrices. Recoveries and preci-
sions (intra-day, inter-day, and within-laboratory) were calculated
from the determination of six aliquots each sample fortified at three
levels (0.5, 1, and 1.5 times MRL or 1, 1.5, and 2 times CC␣ for no
MRL substances). The analyses were finished by the same opera-
tor in triplicate in a 2-week period. Within-laboratory was carried
out in the same laboratory, but performed by two different oper-
ators and environmental conditions on three separate occasions
in a 1-month period. The recovery was calculated by the measured
content/the fortified level × 100 and precision was expressed as the
RSD.
2.7.3. Decision limit (CC˛) and detection capability (CCˇ)
The decision limit (CC␣) is the lowest concentration at which a
method can discriminate with a statistical certainty of 1 − ˛ that
the analyte is present. In the case of AGs with established MRL,
CC␣ was established by the following: 20 blank matrix samples
of pork muscle, chicken liver, bovine kidney, honey, milk spiked
at the MRL level were analyzed. The concentration at the MRL
plus 1.64 times the corresponding SD is defined as CC␣ (˛ = 5%).
For no MRL substances, 20 blank matrix samples of pork muscle,
chicken liver, bovine kidney, honey, milk were analyzed and the
signal-noise ratio (S/N) is calculated at the time window in which
the analyte is expected. CC␣ values were defined as three times
of S/N.
The detection capability (CC␤) is the concentration at which the
method is able to detect truly contaminated samples with a statisti-
cal certainty of 1 − ˇ. In case of MRL substances, CC␤ was calculated
by analyzing 20 pork muscle, chicken liver, bovine kidney, honey,
milk spiked with the analytes at CC␣ and then the CC␣ value plus
1.64 times the corresponding standard derivation is equal to CC␤
(ˇ = 5%). For substances with no MRL, the analysis and calculation
were performed in the same way, but plus 1.64 times the SD of
within-laboratory reproducibility.
32 W.-x. Zhu et al. / J. Chromatogr. A 1207 (2008) 29–37

Table 1
Molecular weight and optimized MS parameters for the target analytes using ESI + mode

Compound MW Precursor ion (m/z) Product ion (m/z) Dwell time (s) CE (eV) DP (V) EP (V) CXP (V)
a
STREP 581.2 582.2 263.2 0.10 40 120 11 18
246.2 0.10 50 120 11 17
a
DISTREP 583.6 584.2 263.0 0.15 40 140 10 18
246.2 0.15 50 140 10 17
a
NEO 614.6 615.4 161.3 0.15 43 150 9 9.9
293.2 0.15 34 100 10 19
a
PARO 615.9 616.3 163.3 0.15 49 120 10 10
293.2 0.15 32 120 10 19
a
KANA 484.2 485.1 163.1 0.15 35 90 9 10
324.3 0.15 26 90 9 21
a
AMIK 585.3 586.2 425.3 0.10 25 100 9 26
163.4 0.15 40 120 8 11
a
TOBRA 467.2 468.3 163.3 0.15 33 100 9 8
324.4 0.15 20 97 8 8
a
SPEC 332.3 351.2 333.2 0.05 24 80 10 22
207.2 0.10 32 80 10 13
a
APRA 539.3 378.2 217.3 0.10 23 70 10 13.5
540.3 378.2 0.15 25 120 10 24
a
GENT C1 477.5 478.2 157.3 0.15 30 90 9 10
322.4 0.10 20 90 14 16
a
GENT C2 463.4 464.2 322.3 0.10 20 70 10 19
160.3 0.15 28 70 10 15
a
GENT C1a 449.4 450.2 160.2 0.10 15 70 8 9
322.2 0.10 30 70 8 9
a
HYGRO 527.2 528.2 177.4 0.15 40 110 9 10
352.2 0.15 32 110 9 21
a
Selected for quantification ion.

2.7.4. Specificity
The specificity of the method was demonstrated by testing all
matrices available in our research. We studied 30 muscle, 30 kid-
ney, 30 liver (from three species: chicken, pig, bovine), 10 honey,
and 10 milk (n = 110) samples. Then the fortification with identical
concentration (100 ␮g kg−1 ) in the matrices was investigated. The
results were evaluated by the presence of interfering substances
around the AGs retention time.
2.7.5. Stability
It is well known that instability of the analytes during storage or
analysis may result in significant deviations of the detected result.
Consequently, stability must be taken into account during the val-
idation of residue analysis. In order to simplify the operation, only
several important factors were investigated: (1) to evaluate the
stability in different solutions, all aminoglycosides (100 ␮g mL−1 )
were dissolved in acetonitrile, standard dilution solution, or 5% TCA.
(2) To assess the stability in three storage conditions, the stock
solutions (100 ␮g mL−1 ) and resulting extracts (100 ng mL−1 ) were
stored at room temperature (about 20 ◦ C), 2–4 ◦ C, or below −18 ◦ C.
(3) The stock solutions and purified extracts were placed in glass
or plastic tube. The measured values were compared to those of
freshly prepared standard solutions.
3. Results and discussion
3.1. Method development
Aminoglycosides antibiotics are extremely hydrophilic com-
pounds that do not bind strongly to protein in the matrix [1]. Thus,
it is not complicated to isolate the analytes from biological samples.
For deproteinization, TCA has been widely explored for extracting
AGs in animal tissues. Kaufmann and Maden [22] discussed recov-
ery efficiency using different concentration of TCA in extraction
process. To achieve a plateau for 13 AGs, we also tested 0.1%, 1%, 2%,
5%, and 10% TCA in various animal tissues, honey, and milk. Though
10% TCA reached the highest recovery (83–97%), it might give rise to
high ion suppression in ESI mode. Low concentration TCA (0.1%, 1%,
and 2% TCA) not only gave bad extract yield, but also led to turbid
extracts. Therefore, 5% TCA (74–94%) was applied to extract multi-
analytes and eliminate the turbidity with prominent effectiveness
of protein precipitation.
Animal tissues matrices are rich in protein and lipids com-
ponents. The analysis of AGs in a complex matrix like muscle,
liver, and kidney required an adequate purification. Liquid–liquid
extraction (LLE) is impossible for polarity drugs possessing many
amino-sugar groups in their chemical skeletons. Most reports
chose SPE cleanup, so several new SPE methods were designed .
The properties of poly-cation suggested us to choose the strong
cationic adsorbent (sulfonic groups) or weak cationic adsorbent
(carboxylic groups). Different commercially available cation SPE
cartridges (Varian bond elut certify, Waters Oasis MCX, Waters
Oasis WCX, Supelco SCX) were initially evaluated and then it was
observed that 13 AGs failed to be quantitatively retained or eluted
on the columns. Analysis was conducted by theory of IPC. There-
fore, copolymeric sorbent cartridge (Oasis HLB) was introduced in
SPE procedure with an addition of ion-pair reagent. HLB cartridges
have hydrophilic–hydrophobic interactions with the analyte. It was
found that when a mixture of TCA extracts and HFBA solution (about
pH 0.5) was directly loaded onto HLB cartridges, only seven AGs
were obtained with optimum extraction yields. While the aque-
ous extract was adjusted to pH 8.5 for loading HLB, high recoveries
for other six AGs were found (results were showed in Fig. 1). This
 W.-x. Zhu et al. / J. Chromatogr. A 1207 (2008) 29–37
Fig. 1. Sample pH optimization for the AGs in SPE procedure, a comparison between
no pH adjustment, adjusted to 8.5 and two combined pH.
could be attributed to the fact that AGs have different pKa val-
ues. After many trials, the optimized overall performance could be
obtained with two-coupled HLB columns. The precondition steps
were optimized properly by changing the nature and volume of
solvent. To effectively retain the analytes on HLB cartridges, it was
necessary for conditioning steps to acidify them with HFBA before
loading. In addition to the polarity dependence of eluents, ion-
pair reagent HFBA played a key role in the elution procedure. The
described process proved suitable for muscle, liver, kidney, honey,
milk and the absolute recovery followed a descending order of
milk > muscle > honey > kidney > liver.
Another interesting point to notice relates to the properties of
reconstituted solution before injection to LC–MS/MS, so we applied
several different solutions reconstituted the final extracts, such as
acetonitrile, 10 mM HFBA in acetonitrile, acetonitrile/10 mM HFBA,
acetonitrile/0.2% formic acid, acetonitrile/0.2% formic aicd/10 mM
HFBA, and 10 mM HFBA. The results proved that, by resolving in
acetonitril, acetonitril containing 10 mM HFBA, acetonitrile/10 mM
HFBA, or acetonitrile/0.2% formic acid, not only is the obtained
sensitivity low, but also peak shape is broad and tailing. Though
acetonitrile/0.2% formic aicd/10 mM HFBA was quite excellent,
the solution is relatively complicated. To reduce solvent con-
sumption, 10 mM HFBA was used for reconstituting the resulting
extracts, which rendered identical intensity. Moreover, aminogly-
coside antibiotics are more stable in strong acidic solutions than
other solutions.
To optimize chromatographic separation, a serial of preliminary
experiments were performed. An alternative for the determina-
tion of extremely polar compounds is the application of HILIC.
At the starting point of the trial, we successfully developed a
HILIC–MS/MS method which simultaneously separated 13 AGs at
a retention time of 8–13 min. To facilitate the analytes ionization
in the mass spectrometer, acetonitrile and 0.2% formic acid were
used as mobile phase, but splitting peaks were gained for several
target analytes. Finally, the addition of 20 mM ammonium acetate
gave a good peak shape with an elution program. However, the
separation was performed perfectly only in standard diluted solu-
tions but not in matrix. During the clean-up steps, the introduction
of ion-pair reagent produced significant effects on the separa-
tion mode and neutral HFBA-analytes failed to retain on the HILIC
stationary phase. The contradiction between preparation and chro-
matographic separation limited the usage of HILIC in our research.
Our chromatographic analysis step was improved and modified
from Kaufmann and Maden [22] with some modifications. Due
to there being many amino and hydroxyl groups in compound
structure, the AGs tend to have a strong ionic interaction with
underivitized residual silanols on the unend-capped C18 stationary
phase. Therefore, commercial end-capped C18 columns from differ-
ent companies were examined for separating 13 aminoglycisides.
Most of these C18 columns can separate analytes, but sensitivity
33
of neomycin, gentamicin, tobramycin, hygromycin, amikacin has
troubles meeting MRL. The Capcell Pak C18 column reached the
most excellent result and the S/N ≥ 3 in the standard solution were
2 ␮g kg−1 for dihydro-streptomycin and hygromycin, 5 ␮g kg−1 for
spectinomycin, amikacin, streptomycin, and kanamycin, 10 ␮g kg−1
for tobramycin, 15 ␮g kg−1 for apramycin, gentamicin, and paro-
momycin, 25 ␮g kg−1 for neomycin. Gentamicins consist of five
isomers, i.e. C1, C1a, C2, C2a, C2b, and the major components are
C1a > C2 > C1 while C2a and C2b are minor ones. The counts of gen-
tamicin varied from different manufactures, so the EU required the
contents of C1a (10–30%), C1 (25–50%), totals C2 (25–55%). Most
methods published for determination gentamicins only detect one
or two isomers [2,21,23]. The method could successfully detect
three gentamicins and three gentamicins reaching a baseline sep-
aration under the chromatographic method. As the result of the
same molecule and mass fragment for Gentamicin C2, C2a, C2b, so
the sums of C2 were determined. A good separation of 13 drugs
was achieved and no significant drift of their retention time was
observed after multiple tests (n > 500).
3.2. Mass spectrometry
Each aminoglycoside tuning solution was directly injected into
the electrospray source by syringe pump. Full scan and collision
activated dissociation (CAD) tests were operated to set up an
appropriate MRM method. Protonated molecular ions [M+H]+ were
dominated for the aminoglycoside drugs in positive ESI mode and
selected as precursor ions. Exceptions are SPEC, which produced an
intensive water adduct [M + H2 O + H]+ , and APRA, which resulted
in cone-induced fragments [22]. The confirmation of group B sub-
stances in foodstuffs, a minimum of three identification points (IPs)
must be required according to EU criteria and the method fulfilled
this requirement with the use of two MRM transitions (one pre-
cursor and two product ions) for each compound, which count for
four IPs, respectively. To achieve maximum sensitivity for each ana-
lyte, adequately long dwell times are required. The characteristic
fragment ions were presented in Fig. 2 by product ion scan mode
in good agreement with the literatures [4,6,22,20,24]. The promi-
nent fragmentation ions contributed to cleavage of glycosidic bonds
between the glycosidic oxygen and anomeric carbon accompanied
by hydrogen transfer to the glycosidic oxygen that result in the loss
of the amino-␣-d-glucopyranose or 2-deoxy-d-glucopyranose [24].
Therefore, the fragment pathway of such analytes mainly based
on a rearrangement of glycosidic bonds [4] and followed by the
openness of aminosugar ring. There existed two different opinions
about the fragment pathway of m/z 246 for STREP and DISTREP,
glycosidic cleavage or neutral loss of 263. We verified the fact by
using Q-TOF acquired extract mass of H2 O 18.0106 in neutral loss
mode. The patterns of AGs were consisted with the previous reports
[18,16,24,27].
It is well known that the addition of ion-pair reagent is prone to
signal suppression or enhancement effects in matrix. Matrix effects
on the ionization of analytes were evaluated by comparing the peak
area of standard solution with that in the matrix extract solution. In
general, matrix effects leading to a reduction of peak area are less
than 10%. To achieve a better quantitative result, the matrix effects
were compensated for using matrix-matched standard curves.
3.3. Method validation
Table 2 summarized all performance data of this procedure.
The aminoglycoside peak areas and concentrations were fitted to
a linear equation in the range of 0–10 times MRL or 0–10 times
CC␣. For each compound, five calibration curves were shown. There
are no significant differences between the curves. The correlation
34 W.-x. Zhu et al. / J. Chromatogr. A 1207 (2008) 29–37

Fig. 2. Product ion scan spectra of 13 aminoglycosides obtained from ESI + mode.
Table 2
The validation results and MRLs levels of 13 aminoglycosides in different animal-origin foods

STREP DI (STREP) NEO PARO KANA AMIK TOBRA SPEC APRA GENT (C1) GENT (C2) GENT (C1a) HY (GRO)

Muscle
r 0.9997 0.9992 0.9989 0.9995 0.9994 0.9962 0.9927 0.9998 0.9913 0.9991 0.9973 0.9982 0.9993
Intra-day RSD (%) n = 18 3–5 4–7 3–6 4–9 9–12 9–11 8–10 6–10 7–9 3–8 4–8 4–10 7–10
Inter-day RSD (%) n = 54 5–8 6–9 4–8 5–8 8–16 7–16 7–15 10–15 8–15 6–12 8–12 9–14 9–13
Within-laboratory RSD (%) n = 54 6–8 6–9 6–11 6–10 10–18 12–19 10–17 9–17 10–17 7–14 9–19 10–17 11–16
Recovery (%) 72–96 84–103 77–92 82–110 83–98 80–101 68–81 89–105 73–98 94–112 85–101 76–99 68–92
MRL (␮g kg−1 ) 500 500 500 500 40 / / 100 60 50 50 50 0
CC␣ (␮g kg−1 ) 575.3 594.2 558.2 574.6 49.5 18.2 19.3 119.6 73.3 56.9 61.8 60.3 23.5
CC␤ (␮g kg−1 ) 650.1 693.8 618.6 651.9 60.2 24.1 25.4 138.9 87.2 65.1 74.6 71.0 31.8

Liver
r 0.9983 0.9932 0.9906 0.9992 0.9921 0.9906 0.9934 0.9976 0.9925 0.9992 0.9986 0.9946 0.9915
Intra-day RSD (%) n = 18 4–6 4–7 4–8 3–8 6–10 7–11 8–16 6–9.4 7–11 5–5.6 3–7 4–9 4–11
Inter-day RSD (%) n = 54 6–8 6–9 7–8 5–8 9–14 8–14 10–19 8–12 10–15 9–15 5–9 6–10 10–17

W.-x. Zhu et al. / J. Chromatogr. A 1207 (2008) 29–37

Within-laboratory RSD (%) n = 54 6–10 7–10 8–10 5–8 12–18 12–19 10–22 9–15 8–18 11–18 8–11 9–12 13–19
Recovery (%) 82–103 75–98 74–93 85–99 72–104 75–99 69–104 73–99 72–102 80–112 92–109 81–103 65–91
MRL (␮g kg−1 ) 500 500 500 1500 40 / / 100 60 100 100 100 0
CC␣ (␮g kg−1 ) 578.7 566.4 596.1 1642.6 48.9 21.5 26.4 115.7 122.5 117.6 114.6 122.6 28.9
CC␤ (␮g kg−1 ) 650.9 631.4 680.3 1747.2 58.4 28.4 36.0 132.1 148.4 138.7 130.3 143.2 37.8

Kidney
r 0.9995 0.9987 0.9993 0.9997 0.9993 0.9992 0.9958 0.9989 0.9931 0.9997 0.9986 0.9924 0.9945
Intra-day RSD (%) n = 18 3–5 5–6 3–6 4–7 5–11 6–12 7–11 4–8 5–13 3–7 4–8 5–8 7–10
Inter-day RSD (%) n = 54 5–8 6–9 5–6 5–8 5–15 9–15 7–14 4–10 9–15 5–10 7–10 6–9 8–12
Within-laboratory RSD (%) n = 54 7–9 8–10 5–8 5–9 12–18 13–17 10–14 7–9 10–16 7–10 7–12 7–10 10–19
Recovery (%) 81–96 85–98 92–101 82–96 71–100 68–97 78–99 84–106 73–99 80–94 73–94 76–95 69–92
MRL (␮g kg−1 ) 1000 1000 5000 1500 40 / / 500 100 200 200 200 0
CC␣ (␮g kg−1 ) 1131.2 1080.4 5278.8 1595.4 49.1 19.7 22.5 577.9 121.5 223.3 235.8 216.1 27.6
CC␤ (␮g kg−1 ) 1313.0 1160.1 5538.5 1674.5 59.4 25.4 28.7 657.5 143.0 247.1 269.0 232.4 37.1

Honey
r 0.9997 0.9916 0.9925 0.9964 0.9964 0.9936 0.9992 0.9905 0.9993 0.9992 0.9915 0.9937 0.9996
Intra-day RSD (%) n = 18 5–9 6–10 9–12 9–11 7–9 8–12 10–15 7–10 9–17 6–10 5–11 5–9 8–11
Inter-day RSD (%) n = 54 10–15 10–16 9–16 12–18 9–13 10–14 12–17 11–16 11–19 9–16 11–17 9–12 13–16
Within-laboratory RSD (%) n = 54 12–17 11–18 13–16 16–20 11–14 14–19 15–20 16–19 13–19 10–17 10–19 11–16 14–18
Recovery (%) 79–116 72–107 64–85 71–101 81–99 61–89 73–104 71–97 78–106 68–96 72–102 82–108 61–89
MRL (␮g kg−1 ) / / / / / / / / / / / / /
CC␣ (␮g kg−1 ) 11.3 8.6 40.5 12.7 9.8 11.8 13.6 8.6 21.6 18.6 19.2 20.6 24.8
CC␤ (␮g kg−1 ) 14.7 11.4 52.1 17.1 12.4 15.6 18.1 11.5 28.4 23.8 25.2 26.1 32.5

Milk
r 0.9994 0.9990 0.9998 0.9982 0.9985 0.9916 0.9981 0.9953 0.9906 0.9995 0.9967 0.9991 0.9918
Intra-day RSD (%) n = 18 4–8 5–9 3–6 7–14 4–8 8–11 11–14 6–9 7–13 7–9 5–8 4–10 9–12
Inter-day RSD (%) n = 54 7–8 6–9 5––10 12–17 6–10 13–17 10–16 8–12 10–14 7–10 7–10 8–11 9–14
Within-laboratory RSD (%) n = 54 7–10 9–10 8–11 11–20 9–12 12–19 13–17 9–13 15–21 9–13 97–12 8–11 13–17
Recovery (%) 72–113 81–107 76–97 70–94 78–103 69–97 62–89 67–92 71–98 82–107 70–105 76–114 78–98
MRL (␮g kg−1 ) 200 200 500 / 100 / / 200 / 100 100 100 /
CC␣ (␮g kg−1 ) 225.9 237.1 224.3 19.4 121.5 9.4 11.5 245.9 18.8 115.1 123.4 121.5 23.2
CC␤ (␮g kg−1 ) 250.4 270.5 247.1 26.1 146.4 12.4 15.1 296.3 25.6 130.9 147.3 143.0 30.1

Note: For MRL substances, spiked levels are 0.5, 1, and 1.5 × MRL or 1, 1.5, and 2 × CC␣ for no MRL substances.

35
36 W.-x. Zhu et al. / J. Chromatogr. A 1207 (2008) 29–37
Fig. 3. MRM chromatograms of a spiked chicken liver (100 ␮g kg−1 ) obtained by LC–ESI–MS/MS.
coefficients (r) with each calibration curves are higher than 0.99.
The experimental data derived directly from matrix-match calibra-
tion curves. As can be observed in Table 2, the overall precisions
of less than 23% satisfactory and recoveries ranging from 61% to
114% were found for all aminoglycosides (chromatogram of spiked
sample showed in Fig. 3). After calculation, the CC␣ values lied
in the range of 8.6–5278.8 ␮g kg−1 and the CC␤ in the range of
11.5–5538.5 ␮g kg−1 (listed in Table 2) for all analytes.
After comparing with the background noise in various matrices,
the results demonstrated that, there were no interfering peaks that
could be detected on the expected retention time for these target
analytes (within 2.5%).
The optimum conditions for all analytes were as follow: the
compounds (100 ␮g mL−1 ) should be dissolved with standard
diluted solution; stock solutions and resulting extracts should be
stored in plastic tube at 2–4 ◦ C. Consequently stock solutions should
be stable for at least 6 months in plastic tube at 2–4 ◦ C and resulting
extracts for 1 week.
4. Conclusion
LC–MS/MS has indicated to be a reliable and powerful tech-
nique for the simultaneous quantification and confirmation of 13
aminoglycisides in products of animal origin. The performance
characteristics of the method fully comply with EU recommen-
dations and a good validation parameter was acquired. Moreover,
the clean-up step only required a kind of HLB sorbent cartridge
and the elimination of complicated procedures enables us to eas-
ily operate using routine analysis. This method has been applied
for daily analysis of aminoglycosides in our labs for 1 year. A large
number of incurred or contaminated aminoglycisides residues
in import–export samples from commercial company and some
authorities were diagnosed and quantified by this method. In terms
of overall data and ruggedness, this method has proved to be an
efficient approach for multi-residues detection of aminoglycosides
members.
Acknowledgements
The authors gratefully acknowledge the General Administra-
tion of Quality Supervision, Inspection and Quarantine of China for
financially supporting this research program (No. 2006IK056).
The authors would like to thank Professor X.W. Sun from
Nanyang Technological University (Singapore) for his valuable
assistance and discussions.
Appendix A. Supplementary data
Supplementary data associated with this article can be found,
in the online version, at doi:10.1016/j.chroma.2008.08.033.
References
[1] R. Tawa, H. Matsunaga, T. Fujimoto, J. Chromatogr. A 812 (1998) 141.
[2] S. Bogialli, R. Curini, A.D. Corcia, A. Laganà, M. Mele, M. Nazzari, J. Chromatogr.
A 1067 (2005) 93.
[3] D.A. Stead, J. Chromatogr. B 747 (2000) 69.
 W.-x. Zhu et al. / J. Chromatogr. A 1207 (2008) 29–37
[4] R. Oertel, V. Neumeister, W. Kirch, J. Chromatogr. A 1058 (2004) 197.
[5] Y. Cai, Y. Cai, J. Cheng, S. Mou, Y. Lu, J. Chromatogr. A 1085 (2005) 124.
[6] A.P. Provoost, W.V. Schalkwijk, O. Adejuyigbe, W.V. Leeuwen, J.T. Wagenvoort,
Antimicrob. Agents Chemother. 25 (1984) 497.
[8] M. Preu, D. Guyot, M. Petz, J. Chromatogr. A 818 (1998) 95.
[7] M.B. Medina, J.J. Unruh, J. Chromatogr. B 663 (1995) 127.
[9] A. Posyniak, J. Zmudzki, J. Niedzielska, J. Chromatogr. A 914 (2001) 59.
[10] P. Viñas, N. Balsalobre, M. Hernández-Córdoba, Talanta 72 (2007) 808.
[11] B. Kim, Y. Kim, J. Ok, J. Chromatogr. B 752 (2001) 173.
[12] I. Clarot, P. Chaimbault, F. Hasdenteufel, P. Netter, A. Nicolas, J. Chromatogr. A
1031 (2004) 281.
[13] A.K. Sarri, N.C. Megoulas, M.A. Koupparis, J. Chromatogr. A 1122 (2006) 275.
[14] A.E. Graham, E. Speicher, B. Williamson, J. Pharm. Biomed. Anal. 40 (2006) 1114.
[15] J.M. Serrano, M. Silva, J. Chromatogr. A 1117 (2006) 176.
37
[16] M.V. Bruijnsvoort, S.M. Ottink, K.M. Jonker, E.D. BoerM, J. Chromatogr. A 1058
(2004) 137.
[17] C. Lu, C. Feng, J. Chromatogr. A 1156 (2007) 249.
[18] R.E. Hornish, J.R. Wiest, J. Chromatogr. A 812 (1998) 123.
[19] M.X. Guo, L. Wrisley, E. Maygoo, Aanl. Chim. Acta 571 (2006) 12.
[20] R. Oertel, U. Renner, W. Kirch, J. Pharm. Biomed. Anal. 35 (2004) 633.
[21] D. Löffler, T.A. Ternes, J. Chromatogr. A 1000 (2003) 583.
[22] A. Kaufaann, K. Maden, J. AOAC Int. 88 (2005) 1118.
[23] Y. Babin, S. Fortier, J. AOAC Int. 90 (2007) 1418.
[24] S.I. Kotretsou, Crit. Rev. Food Sci. Nutr. 44 (2004) 173.
[25] Y.A. Hammel, R. Mohamed, E. Gremaud, M.H. LeBreton, P.A. Guy, J. Chromatogr.
A 1177 (2008) 58.
[26] EC Decision 2002/657, Off. J. Eur. Commun. L 221 (2002) 8.
[27] M. Zhou, D. Zhong, Y. Sun, C. Liu, Acta Pharm. Sinica 39 (2004) 826.