Analytica Chimica Acta 711 (2012) 46–53

Contents lists available at SciVerse ScienceDirect

Analytica Chimica Acta

journal homepage: www.elsevier.com/locate/aca

Determination of aminoglycoside residues by liquid chromatography and

tandem mass spectrometry in a variety of matrices
A. Kaufmann ∗ , P. Butcher, K. Maden
Official Food Control Authority, Fehrenstrasse 15, 8032 Zürich, Switzerland

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

Article history: A quantitative LC–MS/MS method was developed for the determination of 13 commonly used amino-
Received 8 August 2011 glycoside antibiotics in meat (pork muscle, fish, and veal livers and kidneys). The proposed method is
Received in revised form 18 October 2011 sufficiently sensitive and highly selective. Unlike other previously reported methods, it uses a simple
Accepted 21 October 2011
clean-up procedure based on a strong cation-exchange solid-phase cartridge that permits high sample
Available online 29 October 2011
extract loading volumes. A unique elution regime based on a volatile buffer at intermediately high pH
value in combination with an organic solvent provides quantitative elution of the various aminoglyco-
Keywords:
sides. This methodology ensured that neither a breakthrough of weakly retained aminoglycosides (e.g.
Tandem mass spectrometry
Veterinary drugs
spectinomycin) nor the incomplete elution of strongly retained analytes (e.g. neo- and gentamycin) is
Aminoglycosides observed. The single-step clean-up is fast and produces clean extracts that minimize matrix-related signal
Solid phase extraction suppression in the electrospray interface.
© 2011 Elsevier B.V. All rights reserved.

1. Introduction proteins and their poor chromatographic retention. These were

Aminoglycosides (AG) are antibiotics with a chemical struc-
ture based on amino-modified oligosaccharides. Some of them
are produced by bacteria belonging to the Streptomyces and
Micromonospora genus, while others are chemically modified
molecules. Many of these drugs are extensively used in human
medicines, while others, such as dihydrostreptomycin, are typical
veterinary drugs. The use of these drugs as veterinary medicines
carries the risk of developing resistance among bacteria, which
might weaken the efficacy of these drugs as human medicines.
Therefore, it is important to have analytical methods capable of
monitoring the use or potential abuse of these drugs in the field of
animal husbandry.
Aminoglycosides have been less frequently included in meat
residue control programs than other drugs such as sulfonamides,
tetracyclines, or chinolones and others. This is less related to the
likelihood of their usage and respective abuse. As a matter of fact,
AGs represent the most frequent violations of the maximum resid-
ual level (MRL) among animals controlled in our local slaughtering
houses’ monitoring programs [1]. The reason that AGs are less fre-
quently analyzed than other veterinary drugs is probably related
to the analytical challenges. There are two properties of AG that
complicate their extraction, clean-up, separation, and detection:
high polarity with a tendency to undergo strong bindings to matrix
∗ Corresponding author at: Kantonales Labor Zürich, Fehrenstrasse 15, 8032
Zürich, Switzerland. Tel.: +41 43 244 71 97; fax: +41 43 244 71 01.
E-mail address: anton.kaufmann@klzh.ch (A. Kaufmann).
the reasons why the analysis of AG has been dominated for a
long time by enzyme-linked immunosorbent assays (ELISA). Early
liquid chromatography-based methods used a derivatization step
to enable chromatographic separation and photometrical detec-
tion. These early developments are comprehensively covered in
a review article [2]. The introduction of liquid chromatography
and tandem mass spectrometry (LC–MS/MS) permitted the analy-
sis of AG without having to rely on derivatization reactions. Still,
older methods were only able to monitor one or two different
AGs [3–8]. More sophisticated extraction, clean-up, and separa-
tion techniques were finally capable of covering all relevant AGs
with a single analytical method [9–16]. Few papers have reported
the analysis of AG together with other veterinary drugs [15,17,18].
The difficulties that have to be overcome are related to the extrac-
tion, clean-up, and separation and detection processes. Liberating
incurred AG from the matrix requires the use of strong acids [8,9].
The following clean-up step has been most often attempted with
solid-phase extraction (SPE). Ion pair-assisted C-18 sorbent-based
approaches were reported [4,5,11,12] besides the more commonly
used weak cation exchanger materials [2,3,6,7,10,15]. Strong cation
exchangers were only used for single residue methods like the
determination of neomycin in muscles [8] and streptomycin in
honey [20]. This is related to the very strong binding of many
AGs toward strong cation exchanger sites [7,9]. Large elution vol-
umes of 20 mL [20] and even 40 mL [7] were required to achieve
acceptable recoveries. These were the reasons for preferring weak
cation exchanger materials. The successful use of weak cation
exchange SPE requires the careful selection a pH range and low
ion strength of loading solutions [9]. This is a critical aspect, since

0003-2670/$ – see front matter © 2011 Elsevier B.V. All rights reserved.
doi:10.1016/j.aca.2011.10.042
 A. Kaufmann et al. / Analytica Chimica Acta 711 (2012) 46–53 47

high-ion strength solutions are obtained after neutralizing the Reference solution (10 g L−1 ). 2 mL of reference solution
strongly acidic extracts. This problem was mitigated by diluting (100 ␮g L−1 ) was transferred into a 20-mL volumetric flask and
extracts or reducing loading solution volumes. Significantly higher diluted to volume with a dilution solution.
loading volumes and analyte recoveries were reported [9] when We stored the stock solutions, spiking solutions, and ref-
treating the acid extract with an anion exchanger in the carbonate erence solutions in plastic vessels in a refrigerator. Stability
form. This treatment neutralizes the extract without introducing experiments indicated that the stock solutions are stable for six
cations that act as eluent for the AG. However, this additional step months, while the more diluted solutions should be prepared
relies on a laboratory-packed anion exchanger, which significantly weekly.
complicates the clean-up method. The difficulties associated with
the clean-up of the polar AG were the reason why even some recent
2.2. Reagents and solvents
papers [8,10] proposed the currently seldom utilized concept of a
previous derivatization step.
Trichloroacetic acid concentrated, formic acid 98–100%, ethy-
The high polarity of AG prevents the separation of these ana-
lene diaminitrilo tetraacetic acid disodium salt (EDTA), bromoth-
lytes on a classical C-18 column. As a consequence, underivatized
ymol blue, and lithium hydroxide were of analytical grade and
aminoglycosides were either separated with ion pair reagents
purchased from Merck (Darmstadt, Germany). Ammonium formate
[4,6,7,9,11–15] or more recently HILIC columns [2,3,16–18]. The
(HPLC grade), heptafluorobutyric acid, and ethanol (both of analyt-
currently preferred detection technique is tandem mass spectrom-
ical grade) were obtained from Sigma Aldrich (Buchs, Switzerland).
etry (LC–MS/MS) with electrospray in the positive ionization mode
Ammonium hydroxide 25% was from Scharlau (Barcelona, Spain)
[3–18].
and acetonitrile, LC–MS grade from Roth (Karlsruhe, Germany). The
Ion pair-based approaches suffer from signal suppression at
purified water was made in the lab by a lab water unit from Labtec
high ion pair concentrations. Furthermore, ion pair reagents have
(Wohlen, Switzerland).
the reputation of requiring long system rinsing times before
Extraction solution. 25 g of trichloroacetic acid was added into a
the original system performance is restored. HILIC has been
500-mL volumetric flask and diluted to volume with purified water.
extensively investigated for the described method. However, this
Trichloroacetic acid is a highly aggressive chemical. Use appropri-
approach was abandoned in favour of ion pair chromatography
ate safety measures.
(see Section 3).
EDTA solution. 7.5 g EDTA was dissolved with purified water in a
It was the aim of this paper to present a method that detects
100-mL volumetric flask and diluted to volume with purified water.
and quantifies all relevant AGs in a variety of food matrices. Unlike
Lithium hydroxide solution. 7.1 g lithium hydroxide was dis-
a previous approach [9], the method relies on a simple SPE step uti-
solved with 100 mL purified water in a beaker. The solution was
lizing a strong cation exchange material. The use of a novel elution
stirred for 15 min and the vessel was placed into an ultrasonic bath
buffer was crucial for obtaining good recoveries for all analytes.
for another 5 min. Finally, the solution was filtered through a fluted
filter.
Elution solution. 6.3 g of ammonium formate was dissolved in
2. Materials and methods
a beaker with 70 mL of purified water and 20 mL of acetonitrile.
Ammonium hydroxide was added until a pH value of 9.5 was
2.1. Standards and stock solutions
obtained. The solution was transferred to a 100-mL volumetric flask
and diluted to volume with purified water.
The reference substances (spectinomycin, streptomycin, dihy-
Heptafluorobutyric acid 25%. 2.5 mL of heptafluorobutyric was
drostreptomycin, amikacin, kanamycin (A, B, C), paromomycin,
diluted in a 10-mL volumetric flask to volume with purified water.
apramycin, tobramycin, sisomycin, neomycin B, and gentamycin)
Dilution solution. 100 mL acetonitrile, 10 mL acetic acid, and
were of the highest available purity and were bought from
0.65 mL heptafluorobutyric acid conc. were transferred into a
Sigma–Aldrich (Buchs, Switzerland).
500-mL volumetric flask and diluted to volume with purified
Individual stock solutions (1000 mg L−1 each) were prepared by
water.
accurately weighing 50 mg of reference substance (calculated as
Bromothymol blue solution. 0.25 g bromothymol blue was dis-
dry free base) into a 50-mL volumetric flask. The compound was
solved and diluted to volume in a 50-mL volumetric flask with
dissolved with a dilution solution (see Section 2.2) and diluted to
ethanol.
volume with the solution.
Mobile Phase A. 25 mL acetonitrile and 650 ␮L heptafluorobu-
Mixed the A spiking solution (5 mg L−1 ). The solution was pre-
tyric acid conc. were transferred into a 500-mL volumetric flask
pared by transferring 0.5 mL of each individual stock solution
and diluted to volume with purified water.
(1000 mg L−1 ) into a 100-mL volumetric flask and diluting it to
Mobile Phase B. 25 mL water and 650 ␮L heptafluorobutyric acid
volume with a dilution solution.
conc. were transferred into a 500-mL volumetric flask and diluted
Mixed the B spiking solution (1 mg L−1 ). 4 mL of the mixed A spik-
to volume with acetonitrile.
ing solution (5 mg L−1 ) was transferred into a 20-mL volumetric
Formic acid (10%). 10 mL of formic acid conc. was transferred into
flask and diluted to volume with a dilution solution.
a beaker and dissolved with 90 mL of purified water.
Reference solution (500 g L−1 ). 2 mL of mixed A spiking solu-
tion (5 mg L−1 ) was transferred into a 20-mL volumetric flask and
diluted to volume with a dilution solution. 2.3. Extraction and sample processing instrumentation
Reference solution (100 g L−1 ). 0.5 mL of mixed A spiking solu-
tion (5 mg L−1 ) was transferred into a 25-mL volumetric flask and Homogenization was done by a Polytron PT300 from Kinemat-
diluted to volume with a dilution solution. ica (Littau, Switzerland), while a Metrohm 691 pH meter (Herisau,
Reference solution (50 g L−1 ). 2 mL of reference solution Switzerland) was used for the measurement of pH adjustments.
(500 ␮g L−1 ) was transferred into a 20-mL volumetric flask and Extracts were centrifuged by a centrifuge from Sorval RC5C plus
diluted to volume with a dilution solution. Thermo (Zürich, Switzerland). Solid-phase extraction cartridges,
Reference solution (25 g L−1 ). 1 mL of reference solution OASIS MCX 150 mg; 6 mL from Waters (Millford, MA, USA) were
(500 ␮g L−1 ) was transferred into a 20-mL volumetric flask and processed by a Visiprep solid-phase extraction unit from Supelco
diluted to volume with a dilution solution. (Bellefonte, PA, USA).
48 A. Kaufmann et al. / Analytica Chimica Acta 711 (2012) 46–53

2.4. Sample preparation Table 1

MS parameters for monitored compounds. Columns: Analyte: Name of analyte;
Transition: first line corresponds to the quantification SRM, second line corresponds
Five grams of sliced tissue was weighted into a 250 mL cen- to the confirmation SRM; Tube lens: Tube lens voltage; Collision energy: Collision
trifugation tube. 25 mL of extraction solution and 0.5 mL of EDTA energy applied for fragmentation.
solution were added into the centrifugation tube. In the case of a
Analyte Transition Tube voltage [V] Collision energy [eV]
recovery test, 0.5 mL of mixed A spiking solution (5 mg L−1 ) was
added as well. The mixture was homogenized for 30 s and then Spectinomycin 351 > 333 100 19
351 > 207 100 21
centrifuged for 5 min (14,500 rpm = 30,000 × g). Proper safety pre-
Streptomycin 582 > 263 189 31
cautions were observed, because of the aggressive nature of the 582 > 246 189 36
extraction solution. Extracts that contained particulate matter were Dihydrostreptomycin 584 > 263 157 30
filtered through a fluted filter into a plastic beaker. Lithium hydrox- 584 > 246 157 34
Amikacin 586 > 163 108 32
ide solution was added until a pH value between 5.5 and 6.5 was
586 > 425 108 20
obtained. Kanamycin B 485 > 163 100 26
The SPE cartridge was conditioned with 4 mL of lithium hydrox- 485 > 324 100 17
ide solution and washed afterwards with 6 mL of purified water. Paromomycin 616 > 163 109 34
Then a reservoir was put on top of the cartridge and the whole 616 > 324 109 20
Apramycin 540 > 217 113 28
volume of neutralized extract was transferred into the reservoir.
540 > 378 113 17
The vacuum was adjusted to obtain a flow rate of approximately Tobramycin 468 > 163 100 24
two drops per second through the cartridge. This was followed by 468 > 324 100 16
a washing step of the cartridge, utilizing 5 mL of purified water and Sisomycin 448 > 254 100 21
448 > 271 100 18
5 mL of acetonitrile. The cartridge was afterwards eluted with 5 mL
Neomycin B 615 > 161 120 30
of elution solution into a 10-mL volumetric flask. Fifty microliters of 615 > 455 120 23
heptafluorobutyric acid 25% and a drop of bromothymol blue were Gentamycin C1 478 > 322 114 14
added. This was followed by a drop-wise neutralization of the mix- 478 > 157 114 21
ture with formic acid (10%). The adding of acid was stopped, as soon Gentamycin C1a 450 > 322 114 14
–
as the color of the solution changed from blue to yellow. Finally, the
Gentamycin C2/2a/2b 464 > 322 114 14
solution was made up to volume with purified water. Of the neu- –
tralized extract, 450 ␮L was transferred into an HPLC vial and 50 ␮L
of dilution solution was added to produce the injection ready solu-
tion. Another 450 ␮L of neutralized extract and 50 ␮L of mixed B regulated (having a defined maximum residue level), while others
spiking solution [1 mg L−1 ] were transferred into another HPLC vial are not regulated. The CD specifies different validation procedures
to produce a B spiking solution. for these two groups. Following the original protocol would require
twice the validation work. Furthermore, the fact that the maximum
2.5. UPLC separation residue levels (MRL) greatly varies from analyte to analyte but also
for a given analyte from matrix to matrix, suggested the use of a
The equipment consisted of an Acquity system (sample and relatively large spiking concentration range for the validation pro-
solvent manager) from Waters (Millford, MA) and a Kinetex cedure. There is a significant advantage in the proposed technique
C-18, 2.1 mm × 150 mm × 2.6 ␮m column with an installed pre- since it can accept future changes of the defined MRL levels. The
filter, both from Phenomenex (Torrance, CA, USA). The column unmodified CD requires repeated spiking experiments which are
was maintained at 25 ◦ C and the injector volume was 5 ␮L. symmetrically positioned around the defined MRL. In other words
The following linear gradient was used: 0–2 min 20%B, 2–5 min: the validation of the method has to be redone as soon as a sin-
20–30%B, 5–7 min: 30–60%B, 7–7.1 min: 60–100%B, 7.1–8 min: gle MRL value is being changed. Such changes are not infrequent
100%B, 8–8.1 min: 100–20%B, 8.1–9 min: 20%B. The flow was set and are introduced as soon as new veterinary science or toxicolog-
to 0.3 mL min−1 . A typical separation is shown in Fig. 1.. ical data suggest an adjustment of defined MRL. The modifications
of the validation protocol were extensively discussed [22] and are
2.6. MS/MS parameters in line with the Chapter 3.1.3 (Validation according to alternative
models) of the CD.
The utilized tandem quadrupole instrument was a TSQ Quantum The designed validation protocol included the processing of a
Access Max (triple quad) with heated electrospray interface and total of 21 blank liver matrix samples and seven blank kidney,
Xcalibur software (Thermo Fisher, San Jose, CA, USA). It was oper- muscle, and fish matrix samples, each to account for possible selec-
ated in the positive electrospray mode. The capillary voltage was tivity problems. The blind matrix samples were carefully selected
3.0 kV. The transitions, collision energies, and tube lines voltages for to account for the possible variations within a given matrix (e.g.,
the individual compounds are listed in Table 1. The discharge cur- race of animal, organic, conventional production, age). Liver was
rent was set to 4 ␮A. The temperature in the capillary was adjusted considered to be the most difficult matrix. Hence a larger number
to 250 ◦ C, while the vaporizer temperature was set to 500 ◦ C. Sheet of blank samples were investigated to test for the absence of any
gas pressure was set to 45 units, while the auxiliary valve flow false positive signal.
was set to 0. MS resolution values were defined to correspond to a The fortification range produced by spikes covered two orders
mass resolution of 0.7 Da. Other parameters were optimized to pro- of magnitude, described by four levels. Each spiking level (A-Spike)
duce a maximum analyte signal (Collision cell pressure: 1.5 mTorr was repeated four times (independent spiking, extraction, and
of Argon). injection). Furthermore, two blank samples extracts were spiked
prior to injection to account for possible signal suppression (B-
2.7. Validation concept Spike). The dynamic range was adapted to the analyte and matrix
dependent MRL values as stated in Tables 2–5.
The validation of the method was performed according to the The most relevant matrix (liver) was analyzed three times (dif-
EU Commission decision 2002/657/EEC (CD). Some modifications ferent days) according to this protocol to account for within and
[22] were required because most of the investigated compounds are between days variations. Robustness was tested by two means.
Table 2
Validation data for liver. Columns: Analyte: Name of compound; RSD: Relative standard deviation of repeated measurements (n = 4) at four different fortification levels. The first value refers to the RSD within a day, while the
value in brackets states the RSD between the days (n = 3); Recovery: Measured recoveries at various concentration levels. These values are corrected for signal suppression; CC␣: Decision limit; CC␤: Detection capability; Limit
of quantification: Corresponding to a s/n of 9:1; Coefficient of determination: r2 covering the dynamic range as described by the range defined by the stated levels; Relative signal suppression: Ratio of analyte response in matrix
as compared to the pure standard.

Liver

Analyte RSD [%] within days (RSD [%] between days) Recovery CC␣ (MRL) CC␤ (MRL) Limit of Coefficient of Relative
Analyte Analyte quantification determination signal
[␮g kg−1 ] [␮g kg−1 ] [␮g kg−1 ] [r2 ] suppression

Level A Level B Level C Level D Level A Level B Level C Level D

(100 ␮g kg−1 ) (500 ␮g kg−1 ) (1000 ␮g kg−1 ) (5000 ␮g kg−1 ) (100 ␮g kg−1 ) (500 ␮g kg−1 ) (1000 ␮g kg−1 ) (5000 ␮g kg−1 )
Spectinomycin 10.4 (11.2) 9.1 (19.1) 12.3 (27.4) 7.3 (11.2) 76.1 71.1 64.9 62.8 1257 (1000) 2060 (1000) 30 0.9932 0.46
Streptomycin 13.9 (19.6) 4.9 (6.4) 6 (21) 2.8 (3.7) 48.7 63.4 64.2 60.5 536 (500) 618 (500) 25 0.9986 0.66
Dihydro- 5.8 (10.6) 2.9 (6.5) 4.8 (10.5) 2.3 (3.8) 83.6 83.8 82.9 72.0 526 (500) 585 (500) 5 0.9983 0.61
streptomycin
Amikacin 4.6 (6.5) 3.2 (2.4) 3.4 (5) 1.9 (4.8) 83.2 87.2 85.9 83.8 (–) (–) 5 0.9995 0.61
Kanamycin B 5.2 (13.3) 3 (3.9) 2.7 (4.6) 1.7 (6.5) 89.0 87.3 85.2 82.9 630 (600) 698 (600) 5 0.9996 0.62
Paromomycin 7 (40.3) 5.2 (18.1) 5.4 (12.5) 2.5 (2) 78.8 84.4 85.1 77.7 1548 (1500) 1649 (1500) 10 0.9986 0.89

A. Kaufmann et al. / Analytica Chimica Acta 711 (2012) 46–53

Apramycin 7.8 (9.5) 5 (7.9) 6 (6.4) 1.3 (4.1) 79.5 84.7 84.0 77.1 10,116 (10,000) 10,350 (10,000) 5 0.9992 0.79
Tobramycin 7.5 (11.8) 2.6 (5.5) 7.6 (13.1) 2 (5) 79.9 81.2 80.2 75.5 (–) (–) 1 0.9989 0.91
Sisomycin 10.5 (22.4) 5 (10.5) 3.5 (7.1) 3.1 (0.8) 84.4 86.1 87.0 77.7 (–) (–) 10 0.9983 1.15
Neomycin B 7.1 (24.6) 5.7 (18.4) 14.3 (25.5) 3.6 (7.9) 76.3 69.8 66.5 61.9 548 (500) 664 (500) 5 0.9969 1.57
Gentamycin C1 26.2 (47.8) 7.3 (27.5) 10.7 (14.8) 5.8 (10.2) 80.8 82.1 86.2 76.1 228 (200) 297 (200) 30 0.9948 2.06
Gentamycin C1a 12.9 (23.3) 6.8 (15) 7.7 (14) 3.5 (1.5) 75.2 83.7 85.9 78.8 227 (200) 295 (200) 30 0.9979 1.29
Gentamycin 10.9 (40.6) 7.1 (12.3) 6.5 (14.3) 5.6 (3) 69.2 82.2 85.7 77.7 226 (200) 291 (200) 30 0.9958 1.59
C2/2a/2b

Table 3
Validation data for kidney. For legend, see Table 2.

Kidney

Analyte RSD [%] within days Recovery CC␣ (MRL) CC␤ (MRL) Limit of Coefficient of Relative
Analyte Analyte quantification determination signal
[␮g kg−1 ] [␮g kg−1 ] [␮g kg−1 ] [r2 ] suppression

Level A Level B Level C Level D Level A Level B Level C Level D

(100 ␮g kg−1 ) (500 ␮g kg−1 ) (1000 ␮g kg−1 ) (5000 ␮g kg−1 ) (100 ␮g kg−1 ) (500 ␮g kg−1 ) (1000 ␮g kg−1 ) (5000 ␮g kg−1 )
Spectinomycin 5.6 8.8 6.9 4.2 68.6 73.3 75.4 62.5 5346 (5000) 6116 (5000) 30 0.9960 0.69
Streptomycin 12 4.8 4.3 0.8 53.0 72.1 71.3 65.9 1076 (1000) 1245 (1000) 15 0.9994 0.78
Dihydro- 6.2 3.7 1.5 2.3 87.7 90.0 89.9 76.2 1029 (1000) 1090 (1000) 1 0.9980 0.69
streptomycin
Amikacin 2.4 2.8 1.1 1.2 86.8 87.3 89.1 84.6 (–) (–) 1 0.9997 0.64
Kanamycin B 2.3 4.2 0.6 0.9 89.3 91.2 89.9 87.1 2535 (2500) 2607 (2500) 1 0.9999 0.61
Paromomycin 5.3 4.4 1.6 2.2 80.0 84.1 85.7 79.7 1555 (1500) 1671 (1500) 5 0.9992 0.96
Apramycin 4.1 3.6 1.5 1.6 80.7 90.6 89.0 80.4 20,528 (20,000) 21,632 (20,000) 5 0.9992 0.81
Tobramycin 5.5 5.9 2.7 1.7 78.5 82.7 84.0 79.9 (–) (–) 1 0.9995 0.91
Sisomycin 3.5 9.1 4.2 2.4 92.8 91.6 89.8 80.4 (–) (–) 20 0.9986 1.18
Neomycin B 7 9.7 5 3 73.3 74.1 74.1 67.3 5243 (5000) 5767 (5000) 15 0.9984 1.4
Gentamycin C1 10.2 14.7 5.5 3.1 91.4 92.8 84.9 75.1 827 (750) 1009 (750) 30 0.9976 2.27
Gentamycin C1a 9 7.9 2.5 2 81.3 87.8 84.7 80.1 782 (750) 852 (750) 50 0.9993 1.29
Gentamycin 10.2 15.5 6.4 2.9 73.5 91.9 87.1 79.1 837 (750) 1045 (750) 60 0.9980 1.62
C2/2a/2b

49
 50
Table 4
Validation data for muscle. For legend, see Table 2.

Muscle

Analyte RSD [%] within days Recovery CC␣ (MRL) CC␤ (MRL) Limit of Coefficient of Relative
Analyte Analyte quantification determination signal
[␮g kg−1 ] [␮g kg−1 ] [␮g kg−1 ] [r2 ] suppression

Level A Level B Level C Level D Level A Level B Level C Level D

(100 ␮g kg−1 ) (500 ␮g kg−1 ) (1000 ␮g kg−1 ) (5000 ␮g kg−1 ) (100 ␮g kg−1 ) (500 ␮g kg−1 ) (1000 ␮g kg−1 ) (5000 ␮g kg−1 )
Spectinomycin 17.2 7.7 3.1 7.9 72.5 80.6 76.3 72.3 338 (300) 433 (300) 10 0.9932 0.69
Streptomycin 42.8 17.8 10.4 8.2 27.4 49.1 62.2 65.9 566 (500) 728 (500) 30 0.9918 1.27
Dihydro- 6.4 8.5 3.5 5.7 64.9 79.3 76.7 75.1 546 (500) 654 (500) 1 0.9964 1.07
streptomycin
Amikacin 7.8 4.8 5.6 1.5 72.3 80.7 81.3 84.5 (–) (–) 1 0.9995 0.6
Kanamycin B 6.8 2.8 2.1 2 83.9 81.4 84.4 83.8 103 (100) 113 (100) 15 0.9995 0.62
Paromomycin 15.4 18.8 8.3 3.9 69.4 73 80.9 85 532 (500) 603 (500) 15 0.9977 0.77
Apramycin 11.8 7.5 5.4 2.2 58.9 79.1 82.9 84.6 1036 (1000) 1113 (1000) 15 0.9991 0.77

A. Kaufmann et al. / Analytica Chimica Acta 711 (2012) 46–53

Tobramycin 17.3 5.3 3.2 2.1 69.4 70.8 79.5 79.6 (–) (–) 15 0.9993 0.85
Sisomycin 19 11.5 3.6 2.1 62.6 63.7 80.1 80.4 (–) (–) 10 0.9989 0.95
Neomycin B 5.9 5.4 9.8 3.6 55.6 74.2 72.5 68.4 529 (500) 593 (500) 30 0.9978 1.1
Gentamycin C1 12 24.2 12.5 13.7 50.2 56.2 64.8 87.6 55 (50) 69 (50) 30 0.9784 1.03
Gentamycin C1a 57.4 25.4 13.2 7.7 58.2 67.1 69.9 86.2 81 (50) 305 (50) 30 0.9914 0.87
Gentamycin 18.4 23.6 4.6 8.7 34.8 52.9 63.2 78.4 56 (50) 73 (50) 30 0.9903 1
C2/2a/2b

Table 5
Validation data for fish. For legend, see Table 2.

Fish

Analyte RSD [%] within days Recovery CC␣ (MRL) CC␤ (MRL) Limit of Coefficient of Relative
Analyte Analyte quantification determination signal
[␮g kg−1 ] [␮g kg−1 ] [␮g kg−1 ] [r2 ] suppression

Level A Level B Level C Level D Level A Level B Level C Level D

(100 ␮g kg−1 ) (500 ␮g kg−1 ) (1000 ␮g kg−1 ) (5000 ␮g kg−1 ) (100 ␮g kg−1 ) (500 ␮g kg−1 ) (1000 ␮g kg−1 ) (5000 ␮g kg−1 )
Spectinomycin 6.7 6.1 6.2 4.2 73.2 72.8 76.7 71 320 (300) 366 (300) 15 0.9977 0.69
Streptomycin 32.7 10 6.7 6.6 45.1 49.7 65 67.5 553 (500) 679 (500) 25 0.9947 0.74
Dihydro- 5.9 6.2 3.4 4.1 72.3 70.6 79.8 71.3 534 (500) 610 (500) 2 0.9975 0.76
streptomycin
Amikacin 1.6 3.2 1 4.3 79.2 79.4 80.2 77.5 (–) (–) 2 0.9981 0.6
Kanamycin B 4 2.9 3 5.7 78.5 81 80.5 79.7 105 (100) 115 (100) 5 0.9967 0.62
Paromomycin 16.7 9 10.9 7.7 66 79.2 82.2 78.3 563 (500) 715 (500) 10 0.9932 0.8
Apramycin 11.6 5.1 3.7 8.5 60 79 84.7 81.8 1139 (1000) 1483 (1000) 2 0.9924 0.75
Tobramycin 9.9 2.8 2.1 5 72 78.8 78.5 81.6 (–) (–) 3 0.9973 0.82
Sisomycin 12.2 12.3 5.9 4.4 70.6 69.5 74.8 79.9 (–) (–) 5 0.9975 0.89
Neomycin B 5.1 4.9 6.1 9.4 61.3 70 73.6 70 577 (500) 771 (500) 15 0.9908 1.08
Gentamycin C1 19.9 24.1 8 2.2 49.5 61.1 73.1 74.3 61 (50) 91 (50) 10 0.9984 1.07
Gentamycin C1a 28.1 15.7 14 5.8 47.4 71.4 78.8 73.8 65 (50) 112 (50) 10 0.9948 0.91
Gentamycin 24.8 17.7 7.3 4.3 37.6 61.8 69.3 77.1 60 (50) 86 (50) 10 0.9969 0.99
C2/2a/2b
 A. Kaufmann et al. / Analytica Chimica Acta 711 (2012) 46–53
The six series (three liver, one kidney, fish, and muscle each) were
processed by two different persons. Furthermore, the LC–MS/MS
instrument was used for other analytical work between the indi-
vidual aminoglycoside validation series. Hence, this required the
rinsing of the HPLC system (removal of ion pair agent). These two
measures were expected to introduce some variability and should
permit a realistic estimation of the method performance in daily
routine applications.
External calibration was employed. A quadratic calibration
curve was used without a forced intercept to account for possi-
ble non-linearity of the detection response. The signal suppression
effects were determined by spiking a final extract of blank matrix
samples with a known concentration of analyte. The calculated con-
centrations of this blank extract fortified right before injection, was
divided by the actual analyte concentration to estimate the relative
signal suppression, as given in Tables 2–5. The obtained values were
used to correct the measured peak areas of the analyzed samples.
The calculation of CC␣ and CC␤ for regulated compounds was
based on the confidence band (1.64 and 2.33 standard deviations
respectively) utilizing the defined MRL value and the regression
function covering the whole dynamic range covered by the spiking
experiments [22]. The detection sensitivity (limit of quantification)
of unregulated compound (without MRL) was determined by esti-
mating the concentration where a s/n ratio of 9 was observed.
3. Results and discussion
3.1. Extraction of AG from the matrix
Liberating incurred aminoglycosides from the matrix requires
harsh conditions [8,9]. Concentrations up to 10% [8] trichloroacetic
acid (TCA) have been reported. Authors trying to analyze amino-
glycosides together with other veterinary drugs were compelled to
use weaker extraction solvents in order to prevent the decompo-
sition of the less stable analytes. Formic acid has been suggested
[10] as extraction solvent, however, the authors did not provide
recovery data for aminoglycosides. Experiments indicated that the
exhaustive extraction from the matrix and the completeness of
the TCA-induced deproteination are important criteria. Concentra-
tions below 2% produced turbid extracts that frequently lead to
irreproducible blocking of the SPE cartridge. The use of higher TCA
concentrations significantly improved the recovery of some AG [9].
On the other hand, high concentrations of TCA result in high ion
strengths, negatively affecting the following SPE clean-up step [9].
The use of 5% TCA gave satisfactory recovery for fish and muscle
meat, however, very poor recoveries of neomycin in livers or kid-
neys were observed. There are indications the extraction efficacy
is inversely correlated to the number of primary amino moieties
within a given aminoglycoside. Adjacent amine groups are capa-
ble in forming chelate complexes with multiple charged ions. This
can be suppressed by the addition of complex binding reagents like
EDTA [8,12,14,17]. Even relatively small concentrations of EDTA are
sufficient in improving recoveries of affected compounds.
3.2. Clean-up by cation exchange SPE
Clean-up of extracts containing aminoglycosides has most often
been done by weak cation exchange SPE [3,4,7,8,12,17]. Such mate-
rials require a careful pH adjustment of the loading solution in order
to ensure, on the one hand, the deprotonation of the ion exchange
sites of the resine and, on the other hand, the protonation of the dis-
solved aminoglycosides to enable the retention of the AG [9]. Still,
a breakthrough of weakly retained analytes like spectinomycin can
be observed when loading matrix extracts instead of standard solu-
tions. This was traced to the high ionic strength of these extracts [9].
51
As mentioned above, the successful extraction of AG from matrix
requires high concentrations of acids. Hence neutralization of the
extract before SPE loading leads to a high ion load. The introduced
cations (sodium, ammonium, etc.) show a significant elution power
at the loading pH value. This was the reason why the removal of ions
by a HCO3 − conditioned anion exchanger step was proposed [9].
Such a step leads to the quantitative removal of TCA. The produced
carbonic acid decomposes to carbon dioxide that does not show
any more eluting power. This approach permitted higher loading
volumes and good recoveries for all relevant AGs. However, such
an additional step required the manual packing and precondition-
ing of a reservoir filled with ion anion exchange materials. It was
the aim of this paper to avoid such a time- and cost-intensive step.
Therefore the use of strong, instead of weak, cation exchanger was
thoroughly investigated. We are not aware that such materials have
been successfully used for multi-residue AG methods. There are
few reports using such a material for single residue method, e.g.,
streptomycin in honey [20] and gentamycin in swine tissues [21].
Furthermore, eluents capable of eluting strongly retained AG from
a strong cation exchanger show a high ion load and buffer capacity
(e.g., phosphate), which does not facilitate a following concentra-
tion step (e.g., evaporation in a rotavapor) or the direct injection
into an LC–MS system.
AGs are rather similar in their chemical structure, but
show significant differences regarding their interaction with ion
exchangers. The two secondary amino groups of spectinomycin
permit only weak interactions with a cation exchanger. On the
other hand, the six primary amine moieties of neomycin are respon-
sible for multiple interactions with negative ion exchanger sites.
Very strong interactions are also observed for the two low pKb
guanidine moieties of streptomycin. As a consequence, the weakly
retained spectinomycin is the limiting compound regarding the
loading capacity of a cation exchanger cartridge. Such losses could
be combated by adjusting the pH of the loading solution to a value
of six. The volume of the loading solution could be further increased
when employing a neutralization step that introduces a weakly
eluting cation. Lithium was found to show weaker elution power
than sodium or ammonium.
A more difficult problem was the quantitative elution of strongly
retained analytes. Various elution techniques were investigated.
Ammonium hydroxide shows an insufficient elution power. On the
other hand, sodium hydroxide was capable in eluting all investi-
gated AGs. However, solutions exhibiting sufficient ion strength
showed a pH value of 12–14. Unfortunately, such high pH val-
ues induced chemical decomposition of the AGs, as described by
another author as well [19]. Spectinomycin underwent a rapid
degradation, showing a t1/2 of only 10 min. Technically, it would be
possible to elute the AGs directly into a strongly acidified solution
to shorten the exposure time to the high pH environment. Still, this
approach was not considered to be robust enough. Hence, elution
under moderate alkaline conditions was investigated. This included
the used of multiple charged ions like cerium. Acceptable solubil-
ity (concentration) and sufficiently strong elution strength were
observed for strontium and barium hydroxide. Quantitative elution
of streptomycin at a non-critical pH value of 10.5 was also obtained
when utilizing a combination of ammonium acetate and barium
hydroxide. However, the injection of barium containing extracts
into the LC–MS/MS system produced an intensive background
due to the various barium adducts and isotopes. Furthermore,
this resulted in relevant signal suppression of some analytes. The
induced precipitation of barium by sulphate or oxalate did reduce,
but did not completely eliminate, this problem. Further investi-
gations showed that a high concentration of ammonium salts at
elevated pH values exhibited similar elution power as barium salts.
As a result, the barium-based elution approach was abandoned.
Some AGs like neomycin and sisomycin still showed interactions
52 A. Kaufmann et al. / Analytica Chimica Acta 711 (2012) 46–53
Fig. 1. Chromatograms of standard 250 ␮g L−1 (top) and spiked liver sample
500 ␮g kg−1 (bottom). The overlaid quantification trace for the following analytes
are given: (1) Spectinomycin; (2) Streptomycin; (3) Apramycin; (4) Amikamycin;
(5) Kanamycin; (6) Sisomycin; (7) Gentamycin; (8) Tobramycin.
with the ion exchanger that are probably not of an ion-exchange
nature. The elution of these compounds was facilitated by adding
a certain percentage of organic solvents like dimethylsulfoxide or
acetonitrile. No investigations were made to find out if this was
due to reversed phase interaction or due to the swelling of the
ion exchanger by the added organic solvent. Swelling of the bulk
material as induced by an organic solvent might make buried ion
exchange sites available to the elution buffer. The effectivity of the
optimized elution solution is shown in Figs. 2–4. Depicted is the
recovery of dihydrostreptomycin, sisomycin, and gentamycin C1
when using the described elution solution versus modified derived
elution solutions (containing no organic solvent, a reduced pH
value, and a reduced ammonium acetate concentration, respec-
tively). A quantitative elution of dihydrostreptomycin relies on the
combined effect of organic solvent, high pH value, and high salt
concentration (Fig. 2). Sisomycin can only be quantitatively eluted
Fig. 2. Recoveries of dihydrostreptomycin in liver by various SPE elution regimes.
Shown is the absolute recovery corrected for signal suppression effects. Eluate 1
refers to the normal elution procedure and Eluate 2 refers to a second elution to
recover possible remaining analyte from the cartridge. Legend: (A) Elution solution
as described in the experimental section. (B) Elution solution like A but without
organic solvent. (C) Elution solution like A but pH is reduced to pH 8. (D) Elution
solution like A but the ammonium formate is reduced by 50%.
Fig. 3. Recoveries of sisomycin in liver by various elution regimes. For legend, see
Fig. 2.
when an organic solvent is present in the elution solution (Fig. 3).
Quantitative elution of gentamycin requires the organic solvent and
the high pH value (Fig. 4). It was an interesting observation that
gentamycin produced only a recovery of some 40% when using an
organic solvent-free buffer (Eluent B). A second elution step, utiliz-
ing the same volume of organic solvent-free elution solution, did
not recover further gentamycin. Eluents containing organic solvent
did not only produce higher recoveries for gentamycin, but were
capable in eluting some additional analyte during a second elution
step. This is not a typical elution behavior for polymeric backbone
SPE materials. We speculate that the presence of organic solvent
causes a swelling of the polymeric backbone of the resin. This pro-
vides access to ion exchange sites that can otherwise be hidden and
therefore are not in direct contact with the cations of the elution
solution.
The alkaline extracts could not be injected directly into the
LC-system because the elevated pH and the relatively high buffer
capacity of the extract distorted the shape of eluting chromato-
graphic peaks. A neutralization of the extracts with formic acid and
the addition of heptafluorbutyric acid produced the injection-ready
samples. The utilized eluent ion concentration, the dilution of the
final extracts, and the injection volume were adjusted so that no
relevant peak broadening was observed.
Fig. 4. Recoveries of gentamycin C1 in liver by various elution regimes. For legend,
see Fig. 2.
 A. Kaufmann et al. / Analytica Chimica Acta 711 (2012) 46–53
3.3. Chromatography and detection
AGs are not retained on reversed phase columns. Therefore,
older separation methods were based on a previous derivatiza-
tion step and a following reversed phase separation. More recent
papers propose an ion pair based reversed phase or HILIC sepa-
ration. Neither technique requires a previous derivatization step.
However, ion pair agents were used by many authors as a means
of last resort. The reasons for this are the signal suppression effects
caused by these agents, the required equilibration time, and the
difficulty in cleaning an LC–MS system after having applied such
modifiers. As a consequence, alternative HILIC separation became
increasingly popular. HILIC was intensively investigated in our
laboratory for this redeveloped method. Two different kinds of
HILIC columns (BEH from Waters and a zwitterionic column from
Macherey Nagel) were tested. They all produced well-shaped peaks
for early eluting compounds, e.g., streptomycin. However, strongly
retained compounds like neomycin showed very poor peak sym-
metry and unacceptable low sensitivity. The peak shape could be
improved by adding high concentrations of volatile buffers. The
low analyte response, as caused by the high water content during
the final elution step and the presence of buffer additives further
reduced the detection sensitivity of the late-eluting analytes. These
negative aspects could be somehow improved by tuning each com-
pound dissolved with the mobile phase composition encountered
during peak elution. Still, late-eluting AGs showed up to 50 times
lower sensitivity than early eluting AGs. Unfortunately, some of the
late-eluting AGs are required to be detected at the lowest levels (the
MRL for gentamycin in muscle is 50 ␮g kg−1 ). Furthermore, HILIC
requires the injection of low water content injection solution (low
elution strength for HILIC separation) to maintain acceptable chro-
matographic peak shapes. Many AGs are not well soluble in such
solvents. Hence, after eluting from the SPE cartridge, the addition of
acetonitrile prior to injection is mandatory. Such high organic sol-
vent extracts showed significant analyte loss when stored in vials.
In addition, linearity of detection and the overall ruggedness of the
HILIC separation were not considered to be acceptably stable. As
a consequence, the use of ion pair modifiers and reversed-phase
LC was preferred over HILIC. The ion pair system was observed to
be stable after some two to three injections. It produced signifi-
cantly higher sensitivities for late-eluting AGs than HILIC and was
very stable regarding the injection of high-salt-containing extracts.
Cleaning the instrument was not found to be a relevant problem.
The LC–MS system can be used for a standard RP application after
a rinsing period of some 15 min. The high salt concentration in the
injection solution was observed to create signal enhancement for
the early eluting spectinomycin. This could be reduced by adjust-
ing the gradient to permit the elution of this analyte after most of
the salt plug was washed out of the system. There were issues of
relevant signal suppression and enhancement as caused by matrix
and mobile phase components. These effects were investigated by
post-column infusion experiments. These optimizations lead to the
evaluation of different analytical columns. Good separations of ana-
lytes from the detrimental suppression regions were obtained by
utilizing the described core-shell column and the stated gradient
profile.
The sensitivity requirements are rather different for the various
analytes (e.g., 50 ␮g kg−1 gentamycin in muscle but 10,000 ␮g kg−1
apramycin in liver). The underlying chemical structure of AGs
and the high ionic strength of the mobile phase are responsible
for the AGs producing a lower LC–MS response than many other
drugs. Hence, sensitivity and possibly the dynamic range can be
an issue for some LC–MS instruments. It is therefore advisable to
build acquisition methods containing retention time-based SRM
53
windows. Furthermore, this technology permits the definition of
relatively long dwell times for the low MRL compounds.
3.4. Validation
Good recoveries were obtained for all compounds in all matrices
(see Tables 2–5). The performance data (recovery) is significantly
better than the data reported for an older method [9]. All coef-
ficient of determinations were higher than 0.99. An exception is
gentamycin C1 in muscle (r2 = 0.978). Good results were obtained
for CC␣ and CC␤. The poorest values were observed for gentam-
cycin C1a in muscle CC␤ = 305 ␮g kg−1 as compared to the MRL of
50 ␮g kg−1 . The reason is the fact that commercially available gen-
tamycin reference material contains only some 20–40% gentamycin
C1 and 10–30% gentamycin C1a. Hence a 50-␮g kg−1 gentamycin
spike consists of some 5–15 ␮g kg−1 gentamycin C1a. The currently
defined MRL is based on the sum of gentamycins C1, C1a, and C2.
4. Conclusions
The proposed method has several advantages over previous
approaches. It covers all relevant AGs in a variety of matrices at
the required sensitivity. Furthermore, significant time and mate-
rial savings have been achieved compared with previous methods.
The extraction shows a good extraction yield of all compounds. The
introduced strong cation exchange SPE material ensures sufficient
retention of even weekly retained analytes (e.g., spectinomycin). It
was only the combination of a high ammonium formate solution at
intermediate high pH and the addition of organic solvent (acetoni-
trile) that permitted a quantitative elution form the strong cation
exchange SPE cartridge. The neutralized eluate shows a sufficiently
weak ionic strength that does not ruin chromatographic separation.
The proposed elution regime for strongly retained compounds
from strong cation exchangers might be relevant for other appli-
cations. This refers to quaternary amines like chlormequat or
mepiquat that are currently analyzed by methods based on weak
cation exchangers.
References
[1] Annual Report of the Official Food Authority of the Canton of Zurich, 2010,
http://www.klzh.ch/downloads/JB 2010 KLZH.pdf.
[2] T. McClinchey, P. Rafter, F. Regan, G. McMahon, Anal. Chim. Acta 624 (2008) 1.
[3] A. Gremilogianni, N. Megoulas, M. Koupparis, Chromatogr F J.A. 1217 (2010)
6646.
[4] A. Kaufmann, P. Butcher, P. Kölbener, Rapid Commun. Mass Spectrom. 17 (2003)
2575.
[5] D. Heller, J. Peggins, C. Nochetto, M. Smith, O. Chiesa, K. Moulton, J. Chromatogr.
A 821 (2005) 22.
[6] R. Granja, A. Montes Niño, R. Zucchetti, R. Montes Niño, A. Salerno, Anal. Chim.
Acta 639 (2009) 64.
[7] M. Cherlet, S. De Baere, P. De Backer, J. Mass Spectrom. 42 (2007) 647.
[8] Y. Chen, X. Hu, X. Xiao, Anal. Lett. 43 (2010) 2496.
[9] A. Kaufmann, M. Maden, J. AOAC Int. 88 (2005) 1118.
[10] P. Martos, F. Jayasundara, J. Dolber, W. Jin, L. Spilsbury, M. Mitchell, C. Varilla,
B. Shurmer, J. Agric. Food Chem. 58 (2010) 5932.
[11] S. Turnipseed, S. Clark, C. Karbiwnyk, W. Andersen, K. Miller, M. Madson, J.
Chromatogr. B 877 (2009) 1487.
[12] F. Holthoon, M. Essers, P. Mulder, S. Stead, M. Caldow, H. Ashwin, M. Sharman,
Anal. Chim. Acta 637 (2009) 135.
[13] W. Zhu, J. Yang, W. Wie, Y. Liu, S. Zhan, J. Chromatogr. A 1207 (2008) 29.
[14] Y. Babin, X. Fortier, J. AOAC Int. 90 (2007) 1418.
[15] S. Bogialli, R. Curini, A. Di Corcia, A. Laganã, M. Mele, M. Nazzari, J. Chromatogr.
1067 (2005) 93.
[16] H. Berrada, J. Moltó, J. Mañes, G. Font, J. Sep. Sci. 33 (2010) 522.
[17] R. Ishii, M. Horie, W. Chan, J. MacNeil, Food Addit. Contam. 25 (2008) 1509.
[18] C. Chiaochan, U. Koesukwiwat, S. Yudthaviorasit, N. Leepipatpiboon, Anal.
Chim. Acta 682 (2010) 117.
[19] J. Phillips, C. Simmonds, J. Chromatogr. A 657 (1994) 123.
[20] P. Edder, A. Cominoli, C. Corvi, J. Chromatogr. A 830 (1999) 345.
[21] X. Li, Y. Chen, X. Tang, J. He, Y. Shang, X. Xiao, J. Agric. Food Chem. 57 (2009)
7356.
[22] A. Kaufmann, Anal. Chim. Acta 637 (2009) 144.