Food Chemistry 266 (2018) 9–16

Contents lists available at ScienceDirect

Food Chemistry
journal homepage: www.elsevier.com/locate/foodchem

Simultaneous determination of aminoglycosides and colistins in food T

a a a b a a b
G. Saluti , I. Diamanti , D. Giusepponi , L. Pucciarini , R. Rossi , S. Moretti , R. Sardella ,
⁎
R. Galarinia,
a
Istituto Zooprofilattico Sperimentale dell’Umbria e delle Marche “Togo Rosati”, Via G. Salvemini, 1, 06126 Perugia, Italy
b
University of Perugia, Department of Pharmaceutical Sciences, Section of Chemistry and Technology of Drugs, Via del Liceo 1, 06123 Perugia, Italy

A R T I C LE I N FO A B S T R A C T

Keywords: A novel method for the simultaneous identification and quantification of twelve aminoglycosides (AGs) and two
Aminoglycosides colistins in meat and bovine milk has been developed. The analysis was carried out using liquid chromatography
Colistins coupled to quadrupole-Orbitrap mass spectrometry (LC-Q-Orbitrap). Among the HILIC (Hydrophilic Interaction
Muscle Liquid Chromatography) stationary phases tested, the bare silica Poroshell 120 provided the best results. The
Milk
samples were extracted with an aqueous solution followed by an SPE clean up based on the weak cation ex-
Liquid chromatography high-resolution tandem
mass spectrometry
change mechanism. The validation study was performed carrying out 72 experiments per matrix at six different
Method validation concentrations in a range encompassing the Maximum Residue Limits. The recoveries were from 72 to 87% in
meat (except colistins) and from 82 to 96% in milk. Repeatabilities and intra-lab reproducibilities were lower
than 10 and 15%, respectively. Limits of detection were lower than or equal to 33 µg kg−1. Finally, test materials
containing AGs prepared for interlaboratory studies were successfully analysed.

1. Introduction
Aminoglycosides (AGs) are multifunctional hydrophilic sugars that
possess several amino and hydroxyl functionalities (Fig. 1). They are
widely used in veterinary practices as antibiotics against bacteria and
parasites in pork, chicken and beef production. Maximum Residue
Limits (MRLs) have been set by Regulation 37/2010 in various animal-
origin food (European Regulation, 2010). Due to high polarity of their
structure with multiple ionization sites, it is well known that AGs are
difficult to extract from complex matrices and are poorly retained on
reversed-phase columns. As consequence, their determination at trace
levels has always been an analytical challenge also considering that, in
the past, the lack of chromophores and fluorophores forced to deriva-
tize AGs to allow sensitive and specific detection (Farouk, Azzazy, &
Niessen, 2015, McGlinchey, Rafter, Regan, & McMahon, 2008). For-
tunately, today the widespread application of mass spectrometric ana-
lysers enable their detection without any derivatization step. Reviewing
multiresidue procedures to determine AGs in food with liquid chro-
matography coupled to tandem mass spectrometry (LC-MS/MS), the
oldest published papers are those of Kaufmann and Maden (2005) and
Zhu and coworkers (Zhu, Yang, Wei, Liu, & Zhang, 2008). Kaufmann
and Maden (2005) described a protocol able to determine eleven AGs in
meat with three solid-phase extraction (SPE) steps prior to LC injection.
Zhu et al. (2008) carried out two consecutive SPE purification steps to
analyze thirteen AGs in various food types. Although these procedures
were effective, they were not efficient being complicated, time-con-
suming and hard to be operate. In the last ten years, simpler sample
treatments have been proposed. They were substantially based on the
extraction of AGs with an aqueous solution containing trichloroacetic
acid and Na2EDTA with or without other salts such as sodium chloride
or ammonium acetate. The extract was then purified applying the SPE
technique. Three kinds of SPE sorbents are used: i) weak cation ex-
change (WCX); ii) reversed-phase (RP); iii) MIP (molecularly imprinted
polymer). Ishii, Horie, Chan and MacNeil (2008), Kumar, Rubies,
Companyó and Centrich (2012), Tao et al. (2012), Lehotay et al.
(2013), Diez et al. (2015) and Asakawa, Uemura, Sakiyama, and
Yamano (2017) used SPE clean-up based on weak cation-exchange
mechanism, whereas Bohm, Stachel and Gowik (2013), Alechaga,
Moyeno and Galceran (2014) and Bazzan Arsand et al. (2016) applied
reversed-phase SPE. Finally, a very recent approach introduced MIP
SPE to purify AGs from complex matrices (Moreno-Gonzales, Hamed,
Gracia-Campaña & Gámiz-Gracia, 2017; Yang, Wang, Chunying, Xixi, &
Chengjun, 2017) (Table S1 - Supplementary Material). With regards to
the chromatographic approach, the classical way to separate very polar
compounds on reversed stationary phases with the addition of ion-
pairing reagents has been widely applied (Kaufmann & Maden, 2005;
Zhu et al., 2008; Tao et al., 2012; Kaufmann, Butcher, & Maden, 2012;
Lehotay et al., 2013; Bazzan Arsand et al., 2016). On the other hand, in

⁎
Corresponding author.
E-mail address: r.galarini@izsum.it (R. Galarini).

https://doi.org/10.1016/j.foodchem.2018.05.113
Received 13 December 2017; Received in revised form 3 May 2018; Accepted 24 May 2018
Available online 25 May 2018
0308-8146/ © 2018 Elsevier Ltd. All rights reserved.
G. Saluti et al.
Fig. 1. Chemical structures of aminoglycosides and colistins.
the last ten years the hydrophilic interaction chromatography (HILIC)
has been a valuable alternative for the analysis of AGs (Ishii et al. 2008;
Kumar et al., 2012; Bohm et al. 2013; Alechaga et al. 2014; Diez et al.,
2015; Asakawa et al. 2017; Moreno-Gonzales et al. 2017; Yang et al.,
2017). Unlike normal-phase chromatography, HILIC works well with
typical RP mobile phases avoiding the use of solvents immiscible with
water. The HILIC retention mechanism primarily involves the parti-
tioning of polar analytes between a water-enriched layer of solvent near
the sorbent surface and the slightly more hydrophobic bulk eluent
(typically acetonitrile). In addition to hydrophilic partitioning, other
separation mechanisms are involved, such as hydrogen-bonding, ion
exchange or dipole-dipole interactions. The great advantage is that
HILIC columns are fully compatible with the MS detectors and, with
respect to the traditional reversed-phases, better sensitivity can be ob-
tained since the use of fluorinated ion-pairing reagents can be com-
pletely avoided overcoming the well-known problems of ion suppres-
sion and frequent maintenance of the LC system. In 2010,
Gremilogianni, Megoulaus and Koupparis demonstrated that a method
based on HILIC for the determination of streptomycin and dehydros-
treptomycin in milk reached lower detection limits when compared to
an analogous procedure, which instead used reversed-phase ion pairing
chromatography (Gremilogianni, Megoulaus and Koupparis, 2010).
Colistin A (or polymyxin E1) and colistin B (or polymyxin E2) be-
long to the antibiotic class of polymyxins, a group of cationic poly-
peptides (Fig. 1). They have been mainly used against the infections
caused by Enterobacteriaceae in pigs, poultry, cattle, sheep, goats and
rabbits and MRLs have been fixed in several matrices. Recently, their
administration in farm has been restricted because of the evidence of
antibiotic-resistance cases (European Medicines Agency, 2016), further
increasing the interest about the development of efficient analytical
methods for their determination at trace levels. However, similarly to
AGs, colistins are polar drugs and therefore they are not easily ex-
tractable from tissues forming bonds with proteins or phospholipids
(Kunin, 1970). For this reason, few methods have been developed so far
for the determination of polymyxins in food. Before 2000s, these drugs
have been mainly analyzed after derivation to introduce chromophore
or fluorophore groups enabling the analysis with the traditional LC
detectors. The first paper using LC-MS/MS was published by Sin, Ho,
Wong, Ho, and Ip (2005). These authors described a method for the
determination of colistin A and bacitracin in milk, liver and kidney.
Later, the same research group (Wan, Ho, Sin & Wong, 2006) proposed
an improved procedure in milk, liver and muscle including colistin B,
too. More recently, Xu et al. (2012) developed an analytical protocol for
colistin A and B specifically suited for fishery products. Kaufmann and
Widmer (2013) published a multiresidue method for five polymyxins in
various food, whereas Boison, Lee and Matus (2015) included in the
method scope seven polymyxins in chicken muscle. All the
10
Food Chemistry 266 (2018) 9–16
aforementioned works shared the same sample treatment strategy, that
is an acidic extraction with a mixture of water, methanol or acetonitrile
in various proportions followed by a reversed-phase SPE (mainly with
polymeric sorbents) to reduce the amount of interfering substances. The
reported chromatographic approach to separate polymyxins has been
the ion-pair reversed-phase LC (Table S2 - Supplementary Material).
Some structural similarities between AGs and colistins such as high
polarity and the presence of several amino-groups suggested that there
could be a chance to optimize a common analytical strategy for the both
drug classes. Therefore the aim of this work has been the development
and validation of a quantitative multiclass method for the simultaneous
determination of twelve aminoglycosides (amikacin, apramycin, dihy-
drostreptomycin, gentamycin C1, gentamycin C1a, gentamycin
C2 + C2a, kanamycin A, neomycin B, paromomycin, spectinomycin
and streptomycin) and two colistins (A and B) in meat and bovine milk.
The determination was performed on an LC system coupled to a hybrid
quadrupole-orbitrap analyser (LC-Q-Orbitrap) using a HILIC analytical
column to separate the fourteen drugs.
2. Experimental
2.1. Chemical and reagents
Acetonitrile (ACN) (LC-MS CHROMASOLV®) was supplied by Fluka
(St. Louis, MO, USA). Formic acid 99% for LC-MS and ammonium
formate were obtained from VWR International Ltd (Lutterworth, UK).
EDTA sodium salt dihydrate and ammonium acetate were purchased
from Sigma-Aldrich (St. Louis, MO, USA). Polymeric Oasis WCX
(150 mg/6 mL), Sep-Pak® Accell™ Plus CM SPE cartridges (500 mg,
6 mL; 46 µm) with silica support were provided by Waters (Milford,
MA, USA). Deionized water was HPLC grade generated by a Milli-Q
purification system (Millipore, Molsheim, France). Amikacin (100%),
apramycin (> 95%), colistin A (and colistin B (> 75%), dihydros-
treptomycin (≥98%), kanamycin A (> 90%), neomycin B (≥70%),
paromomycin (≥98%), ribostamycin (≥70%), spectinomycin (≥98%)
and streptomycin (≥95%) standards were purchased from Sigma-
Aldrich (St. Louis, MO, USA). Gentamycin C1 (≥95%), gentamycin C1a
(≥90%) and gentamycin C2 + C2a (≥95%) were obtained from TRC
Inc. (Toronto, Canada).
Extraction solution for muscle: 10 mM ammonium acetate, 0.4 mM
EDTA, 0.5% NaCl and 2% TCA.
2.2. Apparatus
A refrigerated centrifuge from Eppendorf (Hamburg, Germany, DE),
a 12-position SPE manifold from Macherey-Nagel (Düren, Germany,
DE), a knife mill from Retsch (Haan, Germany, DE), an ultrasonic bath
G. Saluti et al. Food Chemistry 266 (2018) 9–16

Table 1
Retention times, MS acquisition experiments and selected ions: confirmatory run.
Peak Analyte RT Molecular formula Adduct Monoisotopic exact Fragment 1a,b Fragment 2a,b
(min) mass
(m/z) Monoisotopic exact Ion formula Monoisotopic Ion formula
mass (m/z) exact
(m/z)

1 Spectinomycin 5.07 C14H24N2O7 [M+H3O]+ 351.1762 207.1339 C8H19N2O4+ 140.0706 C7H10NO2+

2 Streptomycin 6.45 C21H39N7O12 [M+H2O 300.6454 263.1462 C8H19N6O4+ 176.0917 C7H14NO4+
+2H]++
3 Dihydrostreptomycin 6.46 C21H41N7O12 [M+2H]++ 292.6479 263.1462 C8H19N6O4+ 409.2041 C14H29N6O8+
3 Colistin A 6.80 C53H100N16O13 [M+3H]+++ 390.5958 101.0709 C4H9N2O+ 202.1186 C8H16N3O3+
4 Colistin B 6.83 C52H98N16O13 [M+3H]+++ 385.9239 101.0709 C4H9N2O+ 227.1754 C12H23N2O2+
5 Amikacin 6.86 C22H43N5O13 [M+2H]++ 293.6501 264.1554 C10H22N3O5+ 425.2242 C16H33N4O9+
IS Ribostamycin 6.86 C17H34N4O10 [M+H]+ 455.2348 – – – –
6 Kanamycin A 6.89 C18H36N4O11 [M+H]+ 485.2453 163.1077 C6H15N2O3+ 324.1765 C12H26N3O7+
7 Paromomycin 6.95 C23H45N5O14 [M+2H]++ 308.6554 163.1077 C6H15N2O3+ 144.0655 C6H10NO3+
8 Apramycin 6.95 C21H41N5O11 [M+2H]++ 270.6474 217.1183 C9H17N2O4+ 163.1077 C6H15N2O3+
9 Gentamycin C1 6.98 C21H43N5O7 [M+2H]++ 239.6654 160.0968 C7H14NO3+ 322.1973 C13H28N3O6+
10 Gentamycin C1a 6.98 C19H39N5O7 [M+2H]++ 225.6498 129.1022 C16H13N2O+ 322.1973 C13H28N3O6+
11 Gentamycin C2-C2a 6.98 C20H41N5O7 [M+2H]++ 232.6576 160.0968 C7H14NO3+ 163.1077 C6H15N2O3+
12 Neomycin B 7.00 C23H46N6O13 [M+2H]++ 308.1634 161.0921 C6H13N2O3+ 455.2348 C17H35N4O10+

a
Exact masses of fragment ions are determined by Mass Frontier™ 7.0.4.17 SR2 (HighChem Ltd, Slovakia).
b
Fragment ions are used only for qualitative purposes.

from Falc Instruments (Bergamo, Italy, IT), a rotary shaker from Ika
(Staufen, Germany, DE), a pH meter from Radiometer (København,
Denmark, DK) were used. The pH meter was calibrated before use with
certified calibration standards (pH 4 and pH 7).
2.3. Standard solutions
Individual stock standard solutions were prepared at 500 µg mL−1
(except gentamycins and colistins at 100 µg mL−1) and stored in plastic
vessels such as the intermediate solutions prepared at 10 and
1 µg mL−1. The details and stabilities are reported in Table S3.
2.4. Chromatographic conditions
Chromatographic separation was performed on a Thermo Ultimate
3000 High Performance Liquid Chromatography system (Thermo
Scientific, San Jose, CA, USA). Analytes were separated on a InfinityLab
Poroshell 120 HILIC column (100 × 2.1 mm; 2.7 µm, Agilent
Technologies, Santa Clara, CA, USA) connected with the InfinityLab
Poroshell 120 HILIC guard column (5 × 2.1 mm, 2.7 µm). HPLC eluent
A was an aqueous solution containing 1% (v/v) formic acid (FA) and
1 mM ammonium formate (AF), eluent B was ACN. The gradient was
initiated with 20% eluent A for 2 min, continued with linear increase to
35% A in 5 min. In 1 min eluent A increase to 95% and this condition
was maintained for 7 min. The system returned to 20% B in 0.1 min and
was re-equilibrated for 4 min (run time: 17 min). The column tem-
perature was 40 °C and the sample temperature was kept at 25 °C. The
flow rate was 0.25 mL min−1 and the injection volume 5 µL.
analytical batch with a direct infusion of a LTQ Velos ESI Positive Ion
Calibration Solution (Pierce Biotechnology Inc., Rockford, IL, USA).
The individual compounds were infused with a syringe through a T
union connected to the LC system with a mobile phase flow rate of
0.25 mL min−1. The product ions were found by increasing the collision
energy (CE) using the Q-Exactive Tune 2.3 software (Thermo Fisher
Scientific). After choosing the more intense product ions, fragmentation
energy scans were carried out to obtain the optimal CE for complete
fragmentation of precursor ions. Q-Exactive parameters (RP, AGC and
IT) were optimized to improve sensitivity and selectivity.
The screening run was performed with full scan/dd-MS2 using an
inclusion list with the ion species, retention time interval and optimized
CEs. Four different mass ranges were acquired: m/z 290–490, m/z
220–275, m/z 260–320 and m/z 382–392. The resolution was 70,000
FWHM (m/z 200), except for the last range (m/z 382–392 at 35,000
FWHM). The AGC representing the maximum capacity in C-trap was set
at 3 × 106 ions for a maximum injection time of 300 ms, except for the
last range (150 ms). The precursor ions were filtered by the quadrupole,
which operates at an isolation window of m/z 1.0. During the frag-
mentation experiments, a resolution of 17,500 FWHM (m/z 200) was
adopted. The AGC target was set at 5 × 105 ions for a maximum in-
jection time of 80 ms. The confirmatory run was achieved with SIM/
PRM experiments: resolution of 70,000 FWHM (m/z 200) and injection
time 300 ms (SIM); resolution 35,000 FWHM (m/z 200) and injection
time 150 ms (PRM). The AGC and the isolation window were always set
at 1 × 106 and m/z 1.0, respectively. The retention times, the adducts
and the selected fragment ions are listed in Table 1. All extracted mass
traces were based on a window of 5 ppm (mass accuracy).

2.5. MS conditions 2.6. Sample preparation

The mass spectrometer Q-Orbitrap (Q-Exactive, Thermo Scientific,
San Jose, CA, USA) was equipped with heated electrospray ionization
(HESI-II) source. The HESI-II temperature was set at 300 °C, the capil-
lary temperature at 250 °C, the electrospray voltage at 3.00 kV (positive
mode) and S-lens value was adjusted at 50 V. Sheath and auxiliary gas
were 35 and 25 arbitrary units, respectively. The mass spectrometer
was controlled by the Xcalibur 3.0 software (Thermo Fisher Scientific,
San Jose, CA, USA). The exact mass of the compounds was calculated by
Qualbrowser in Xcalibur 3.0 and the exact mass of the fragment ions
was deduced by Mass Frontier™ 7.0.4.17 SR2 (HighChem Ltd,
Slovakia). Instrument calibration in positive mode was done every
The sample preparation steps are summarized in Fig. S1. In detail,
for muscle, 3 g of minced tissue were weighed in a 50 mL polypropylene
centrifuged tube. The sample was spiked with 15 µL of a ribostamycin
(IS) solution at 10 µg mL−1. The muscle was extracted twice with 15 mL
(10 mL + 5 mL) of the extraction solution and mechanically shaken for
20 min. After sonication and centrifugation (10 min), the reunited su-
pernatants were once again centrifuged at 0 °C, their pH adjusted to 6.5
with 3 M NaOH and they were diluted with water to 50 mL. The diluted
extract was loaded under vacuum control onto a WCX SPE cartridge
(Sep-Pak® Accell™ Plus CM) previously conditioned with 3 mL of ACN
and 6 mL of water. Later, the cartridge was rinsed with 6 mL of water,

11
G. Saluti et al.
dried for 10 min with vacuum and, finally, the analytes were eluted
with 3 mL of elution solution (175 mM ammonium formate at pH 3).
After centrifugation, the sample was injected. For milk analysis, 2 g of
raw, thawed and homogenized sample were weighed in a 50 mL poly-
propylene centrifuged tube. The sample was spiked with 10 µL of a ri-
bostamycin (IS) solution (10 µg/mL−1) and extracted twice with 15 mL
(10 mL + 5 mL) of a 0.25% TCA aqueous solution, mechanically shaken
for 20 min, sonicated and centrifuged for 10 min. The pH of reunited
extracts was adjusted to 6.5 with 0.5 M NaOH, diluted to 50 mL with
water and purified according to the same procedure applied to tissue.
The eluate was centrifuged at 0 °C and injected. Final extracts of both
matrices were stable at room temperature for at least 96 h (Table S3).
2.7. Method validation
The validation was carried out in accordance with the requirements
of Commission Decision 2002/657/EC (European Commission, 2002).
The linearity has been evaluated both in solvent and in matrix. Preci-
sion (repeatability and within-laboratory reproducibility), recovery
(trueness), decision limit (CCα), detection capability (CCβ), LOD and
LOQ were assessed through the experimental plans detailed in Table S4.
An analyte-free bovine muscle or milk was spiked at the beginning of
the extraction procedure with the appropriate standard solutions. Six
spiking levels were performed: 10, 33, 100, 333, 1000 and
1500 µg kg−1 (3300 µg kg−1 for milk). For the spiking levels higher
than 100 µg kg−1, the final extract was diluted (175 mM ammonium
formate at pH 3) prior to LC injection. Four replicates for each spiking
level were analysed during the same day along with matrix-matched
calibration curves prepared adding the analytes immediately prior to
LC injection and a blank muscle sample. Each series was repeated on
three different days (72 spiked samples) varying time, operator and
calibration status of LC-HRMS/MS equipment. The ribostamycin (IS)
was used as internal quality control to daily check the recovery factors
within each analytical batch.
In addition, bovine, swine, poultry and fish muscles were analysed
in parallel at 33 µg kg−1 (four replicates in two different days for a total
of eight observations) to check the method applicability to the main
food-producing animal species. For muscle, two ruggedness experi-
ments (minor changes) were carried out investigating deliberate
changes of four variables, i.e. the concentrations of ammonium acetate,
EDTA, NaCl and TCA of the extraction mixture. The details are listed in
Table S5. Finally, test materials (muscle) from Proficiency Tests orga-
nized for antibiotics were analyzed.
3. Results and discussion
3.1. Optimization of LC-MS/MS conditions
During the development of the chromatographic conditions, three
LC columns have been tested: i) Thermo Hypercarb (100 × 2.1 mm,
5 µm); ii) Thermo Syncronis Hilic (100 × 2.1 mm, 5 µm); iii) Agilent
Poroshell 120 HILIC (100 × 2.1 mm, 2.7 µm). Thermo Hypercarb is a
stationary phase composed by 100% of porous graphitic carbon and
(even though it is not a “pure” HILIC phase) it is highly suggested for
the separation of very polar compounds. Unfortunately, it did not retain
spectinomycin and, on the other hand, neomycin B did not elute within
acceptable retention time. With Thermo Syncronis column (zwitterionic
phase), neomycin B was highly retained with bad peak shape. Finally,
with the Agilent Poroshell 120 HILIC column (bare silica) encouraging
results were obtaining using acetonitrile and an aqueous solution con-
taining 1% of formic acid (FA) and 1 mM of ammonium formate (AF) as
mobile phases.
Traditionally, electrospray (ESI) in positive mode has been the io-
nization technique of choice for the LC-MS analysis of AGs and colistins
because protonation is favored by the presence of amine groups (Fig. 1).
At pH < 4 the most abundant species are the polycationic ones as
12
Food Chemistry 266 (2018) 9–16
indicated by pKas listed in Table S4. In this work, we have obtained the
mass spectra using a H-ESI source. For this purpose, preliminarily, a
mobile phase composed of ACN:water (1% of formic acid and 1 mM of
ammonium formate) (1:1) at 0.25 mL min−1 was used injecting each
compound individually by flow injection analysis and acquiring their
full-scan mass spectra in positive mode. These data were then refined in
spiked matrices evaluating also the effect of the interfering ion species.
Contrary to what proposed by other authors, bi-charged ions were
chosen as quantifier species (Table 1) in order to improve the limits of
detection. The only exceptions were spectinomycin and kanamycin A
(mono-charged) and colistins (tri-charged). The fragmentation patterns
and ions were widely described in literature both for AGs and colistins
(Sin et al., 2005; Lehotay et al., 2013).
A typical phenomenon connected to Orbitrap detectors is the so-
called post-interface ion suppression (Moretti, Cruciani et al., 2016). It
is explainable with the saturation of the C-trap device by the interfering
substances co-eluting with the compounds of interest, leading to a re-
markable signal suppression. To overcome this problem, the reduction
of the acquisition scan range (using the quadrupole) was needed to
achieve suitable detection limits in matrix. Two examples demon-
strating this phenomenon for streptomycin and apramycin in meat are
shown in Figs. S1 and S2, respectively (Supplementary Material). Ac-
quiring a wide m/z range (m/z 205–610) the signals of analytes were
highly suppressed (Figs. S1a and S2a), whereas a narrower m/z range
(m/z 260–320) increased remarkably the peak signals (Figs. S1a and
S2b). The relevant MS spectra are also shown (a′ and b′). For this reason
the analysis of unknown samples (screening run) have to be performed
using four narrow m/z ranges (m/z 290–490, m/z 220–275, m/z
260–320 and m/z 382–392). Suspect samples were then confirmed with
an “ad hoc” SIM/PRM (Parallel Reaction Monitoring) acquisition,
which allows the reaching of the required Identification Points (IPs).
3.2. Optimization of sample preparation
Sample preparation was at first developed for muscle reviewing the
main published methods for AGs determination in various food (tissue,
milk and honey). Among the various purification approaches, these
based on weak cation exchange (WCX) SPE seemed the more promising
and practicable (Ishii et al., 2008; Kumar et al., 2012; Tao et al., 2012;
Diez et al., 2015). On the other hand, strong cation exchange sorbents
were unsuitable (Kaufmann, Butcherm, & Kölbener (2003) ; Ishii et al.,
2008; Kumar et al., 2012) or too laborious (Kaufmann et al., 2012).
Other approaches for AGs analysis, such as C18 dispersive clean-up
(Bazzan Arsand et al., 2016) when reproduced in our laboratory did not
allow the recovery of colistins. Similarly, the intrinsic nature of MIP
sorbents with tailor-made selectivity was not suitable for the simulta-
neous determination of AGs and colistins.
Among the various manufacturers and types, at first we tested the
Oasis WCX cartridges (150 mg, 6 mL, Waters) which contain a mixed-
mode polymeric sorbent based on weak cation exchange and reversed-
phase mechanism. Unsatisfactory recoveries (< 30%) of the last eluting
analytes (gentamycins and neomycin B) were observed whereas colis-
tins were not detectable at all. Later, the silica based Sep-Pak® Accell™
Plus (Waters) cartridges demonstrated good recovery and precision, as
also reported by Kumar and coworkers (Kumar et al., 2012) who de-
veloped a method for the determination of ten AGs in honey and
kidney. For AGs determination silica based WCX sorbents are preferred
instead of the polymeric WCX ones. Among the papers applying WCX
SPE, only Diez et al. (2015) opted for polymeric WCX cartridges but the
obtained recovery factors were undoubtedly less satisfactory than those
observed with silica based WCX (this study; Ishii et al., 2008; Kumar
et al., 2012; Tao et al., 2012).
Sep-Pak® Accell™ Plus sorbent contains a weak cation exchanger
(carboxypropyl group) with a pKa of about 4.8. At pH 6.5, this sorbent
is essentially negatively charged and its retention is due to ion inter-
actions with analytes. Accordingly, the muscle extract was adjusted at
G. Saluti et al.
Fig. 2. Recovery factors as function of NaCl percentage in the extraction mixture of muscle.
pH 6.5 prior to the loading onto SPE to make the majority of carboxyl
groups carry negative charges to retain AGs and colistins. At pH 3.0, the
charge on the sorbent is almost completely neutralised and the elution
can take place. However, pKa concept cannot by itself explain the be-
haviour of these analytes during the purification steps. A crucial factor
is also the ionic strength since the NaCl percentage in the aqueous ex-
traction mixture and the dilution of the extract before the loading onto
SPE cartridge were fundamental, mainly for spectinomycin and colis-
tins recovery. At first, the composition of the extraction mixture was
exactly that used by Kumar et al. (2012) containing 10 mM ammonium
acetate, 0.4 mM EDTA, 1% NaCl and 2% TCA. In Fig 2, the influence of
the NaCl concentration in the extraction mixture is shown. These data
perfectly agreed with those of Kumar et al. (2012) who were forced to
accept a recovery factor of 14% for spectinomycin in kidney. Although
the use of 1% NaCl generally improved the recoveries of all other
analytes, in our opinion spectinomycin was too penalized by this per-
centage and the NaCl concentration was reduced to 0.5%. This behavior
can be explained considering that only two ionisable amine sites are
present on the spectinomycin structure and, in addition, one of these
probably is not completely protonated at the loading pH (6.5) as sug-
gested by the pKas listed in Table S6. Two different packages were used
to predict these values: Moka® (version 2.6.6. Molecular Discovery ltd,
London, UK) (Cruciani, Milletti, Storchi, Sforna & Goracci, 2009;
Milletti et al. 2010) and software available at the website www.
chemicalize.org. In this context, the cations (Na+, NH4+) added in
the extraction mixture can compete with spectinomycin for the –COO−
sites on the sorbent (Kaufmann & Maden, 2005). Since all the other
compounds have from three (streptomycin and dihydrostreptomycin) to
six (neomycin B) ionizable amine groups, their retention on SPE sorbent
was less critical. Similar considerations were carried out by Ishii and
coworkers (Ishii et al., 2008). In addition the dilution of the tissue
extract prior to the SPE loading was also needed to have satisfactory
recovery of spectinomycin decreasing the ionic strength of the loaded
solution and then avoiding it SPE breakthrough.
Later, the importance of ionic strength was confirmed by the be-
havior of milk. Surprisingly, applying to bovine milk the same extrac-
tion mixture developed for muscle, unsatisfactory recovery factors were
observed for spectinomycin (< 20%), whereas for all other analytes
very satisfactory results were anyhow obtained. This probably hap-
pened because milk contains a natural high concentration of salts
(about 8–9 g L−1), such as calcium, magnesium, sodium and potassium
(cations) and phosphate, citrate and chloride (anions). These salts
produced an “intrinsic” high ionic strength, which, together with that
“added” by the ion species dissolved in the extraction mixture, ham-
pered the spectinomycin retention on SPE stationary phase. Therefore,
further experiments varying the composition of extraction mixture were
performed. Since the aim was to decrease the ionic strength of
13
Food Chemistry 266 (2018) 9–16
extractant mixture maintaining acidic conditions, solutions with only
TCA at different percentages were tested: 2, 1, 0.5, 0.25 and 0.1%. Good
results were finally observed adding 0.25% of TCA, but also decreasing
he amount of processed milk, i.e. two grams instead of three. Lower
concentrations of TCA were also suitable (e.g. 0.1%), but the pH of
extract was judged too high (about 5) to assure the successful cleavage
of bonds between analytes and proteins. However, it must be under-
lined that, although all the authors agree with the need of an acidic
environment to quantitatively extract AGs, there is not precise in-
formation about the exact pH regulation.
3.3. Method validation
All the selectivity requirements were verified when a suspect sample
was reanalysed to confirm both the identity and the concentration of
the analyte detected during the first run. Commission Decision 2002/
657/EC (European Commission, 2002) has not been updated to include
high-resolution mass analysers such as Orbitrap or Time of Flight and
the here adopted criteria for the identification were those suggested by
SANTE/11945/2015 (SANTE, 2015). The selected MS2 fragments are
listed in Table 1. It must be underlined that the second run to confirm a
suspect sample was not carried out using the screening run experiments
(first analysis), but with SIM/PRM (or SIM/dds) acquisition specifically
targeted on the suspect molecule. That was in order to obtain a suffi-
cient number of points for the chromatographic peak (quantitative
determination) and more than one product ion spectrum. This latter
aspect is important to minimize deviation in the ion ratio and then false
negative results due to the high irreproducibility in the measurement of
ion ratio (Berendsen, Meijer, Mol, van Ginkel & Nielen, 2017).
Linearity in solvent has been tested in the concentration range
3.3–300 ng mL−1 injecting nine concentration points. The coefficients
of determination were generally satisfactory (r2 > 0.99), however for
colistins the first calibration point was discarded and for kanamycin
even the first two (3.3 and 10 ng mL−1). The constancy of the response
(ratio y/x) was verified and the observed maximum deviation was
lower than 15%. The linearity in matrix (muscle and milk) was tested in
the range 3.3–200 ng mL−1 applying the same evaluation criteria.
When levels higher than 100 µg kg−1 have to be confirmed, the final
extract was appropriately diluted as detailed in Table S4.
The validation plan allowed the investigation of method accuracy,
decision limits and detection capabilities of each analyte without se-
parate experiments as described elsewhere (Galarini, Saluti, Moretti,
Giusepponi & Dusi, 2013). Six validation levels were performed in the
range 10–1500 µg kg−1 for muscle and 10–3330 µg kg−1 for milk.
Method repeatabilities (sr) and within-laboratory reproducibilities (swR)
were estimated, applying the Analysis of Variance (ANOVA) on the
available 12 experiments at each spiking level for each matrix. The
G. Saluti et al. Food Chemistry 266 (2018) 9–16

Table 2
Summarized validation data in bovine muscle and milk.
Analyte Bovine muscle Milk

a
Recovery (%) CVr,pooled CVRw,pooled ME Recovery (%) CVr,pooled CVRw,pooled MEa
(%) (%) (%) (%)

Spectinomycin 79 6.7 9.2 −5 95 5.5 7.4 14

Streptomycin 81 4.1 8.4 −1 85 4.9 6.6 4
Dihydrostreptomycin 87 4.0 6.7 −1 96 4.9 7.5 7
Colistin A 43 8.4 12 62 82 7.5 9.9 25
Colistin B 59 5.0 8.3 25 85 5.0 6.7 19
Amikacin 86 3.7 8.2 −52 94 5.3 7.1 −5
Kanamycin A 87 10 12 −28 92 7.3 8.0 33
Paromomycin 83 4.3 11 −38 93 5.2 6.6 −15
Apramycin 82 4.2 10 −41 92 5.3 6.6 −5
Gentamycin C1 83 5.8 11 −33 91 5.4 6.0 11
Gentamycin C1a 81 7.2 12 −34 91 6.7 7.5 1
Gentamycin C2 + C2a 82 5.2 10 −35 91 5.9 6.6 8
Neomycin B 72 3.6 7.9 −20 88 5.8 6.9 −6

a
Matrix effect. Values in bold are considered significant.

recoveries were established comparing the peak area of each compound
in the spiked samples against its peak area in the matrix matched
standards in which the 14 analytes were added immediately prior to LC
injection. In Table 2 the average recoveries obtained at all validation
levels together with the precision indexes are listed, whereas the de-
tailed results are reported in the Supplementary Material (Tables S7 and
S8). For AGs the recoveries in muscle ranged from 72 (neomycin B) to
87% (dihydrostreptomycin). Worst values were obtained for colistins
(43% and 59% for A and B, respectively). This happened for the above
mentioned reasons which forced us to choice between spectinomycin
and colistin recoveries. The precision was satisfactory with coefficients
of variation (CVs) generally lower than 11% (repeatability conditions,
CVr) and lower than 13% (within lab reproducibility conditions, CVwR).
The accuracy measured in milk was better than in muscle due to the less
critical influence of the extraction step. Recovery factors were higher
than 80% for all the analytes and CVs lower than 10% (Table 2). Since
the quantification was performed with external standardization, the
results were always corrected for the relevant recovery factor evaluated
during the validation study and checked within each analytical batch
by means of matrix-matched calibration curves. In Figs. S3 and S4, the
chromatograms of a blank and spiked bovine muscle at 33 µg kg−1 were
shown. Similarly, the chromatograms of a blank and a spiked milk
(10 µg kg−1) are reported in Figs. S5 and S6. LOD and LOQ were esti-
mated evaluating the method performances at the lowest spiking levels
(Tables S7 and S8). For all the analytes LODs and LOQs were equal to
10 µg kg−1 both in muscle and milk, except for kanamycin and colistins
(LOD and LOQ = 33 µg kg−1) and for neomycin B (LOD = 10 µg kg−1;
LOQ = 33 µg kg−1) in muscle. These values were comparable to those
reported in meat by the other researchers and, for milk, they are even
better (Tables S1 and S2).
Matrix effects (MEs) were assessed comparing the slopes of matrix-
matched curves (analytes added prior to LC injection) versus the same
curves prepared in solvent (175 mM ammonium formate at pH 3). The
more considerable matrix effects were measured in muscle as shown in
Table 2. However, the use of matrix-matched calibration standard
curves allowed to reach satisfactory accuracy. On the other hand, the
more severe MEs were generally “paid” with lower precision (Table 2)
as demonstrated by regression analysis (data not shown).
In Table 4 the results of the experiments carried out analysing bo-
vine, swine and poultry muscles spiked at 33 µg kg−1 in within-lab
reproducibility conditions demonstrated the method applicability to
other food-producing animal species. The concentrations were calcu-
lated applying the relevant matrix-matched curves (bovine for bovine,
swine for swine and poultry for poultry). Since recoveries and CVs
among the species were in good agreement (ANOVA) common decision
limits (CCα) and detection capabilities (CCβ) were fixed (Table 3). To
conclude the study in tissues of species other than beef, the same series
of experiments was also performed in trout muscle, given that for
spectinomycin, gentamycins, neomycin and colistins there are fixed
MRLs in fin fish. The results in Table 4 demonstrated the complete
applicability of the here reported method also in trout, except for
spectinomycin (recovery factor lower than 20%) which demonstrated
once again to be the most critical compound. Probably this was because
a higher content of salts was present in trout compared to the pre-
viously investigated terrestrial species. As matter of fact, further
method optimization was necessary to include spectinomycin, at first
changing the composition of extraction mixture. On the other hand, it is
hard to exactly known in advance the detrimental influence of meat
composition on spectinomycin performances taking into account all
edible fish species or all the terrestrial ones (turkey, sheep, rabbit?).
Therefore, during routine activity, the method management should
involve an “ad hoc” internal quality control program recognising
spectinomycin as the critical compound to judge the success of the daily
method application. On the other hand, low recovery factors are fre-
quently reported for spectinomycin. Kumar et al. (2012) and Diez et al.
(2015), after long method optimization, obtained values lower than
20% in kidney and milk, respectively, for this drug. However, in the
relevant tables showing the validation data these authors reported a not
well explained “trueness” percentages of about 100%. It is worth re-
membering the difference between the “recovery factor” (recovery) and
“apparent recovery” concepts (Burns, Danzer & Townshend, 2002).
Considering the various critical factors involved in the meat ex-
traction process (mainly pH and ionic strength), two ruggedness ex-
periments were carried out to investigate the influence of the con-
centrations of acids and salts dissolved in the extraction mixture: TCA,
ammonium acetate, EDTA disodium salt and NaCl. The purpose was to
assure the successful future application of the procedure in routine
activity as well as its transfer to a different laboratory. The adopted
design of ruggedness study was a typical Youden scheme, which re-
quires eight experiments for a maximum of seven factors/variables
(Commission Decision 2002/657/EC, 2002). The changes (levels) were
chosen taking care to does not introduce excessive (unrealistic) oscil-
lations (Table S5). At this scope, the uncertainties associated to the
preparation of extraction solution (mass weighting and volume mea-
surement) were assessed and the changes fixed five-time higher, as
suggested by Vander Heiden and coworkers (Vander Heyden, Nijhuis,
Smeyers-Verbeke, Vandeginste, & Massart, 2001). In the first experi-
ment, the altered concentrations were at a lower level than the nominal
one and, viceversa, in the second one (Table S5). The results were in-
terpreted applying two different approaches (intermediate precision
and Dong algorithm) which perfectly agreed indicating that the chosen
factors and changes did not affect the results.

14
G. Saluti et al. Food Chemistry 266 (2018) 9–16

Table 3
Maximum Residue Limits (MRL), Decision limits (CCα) and Detection capability (CCβ) in bovine, swine and poultry muscle and bovine milk.
Analyte Muscle Milk

MRL CCα CCβ MRL CCα CCβ

(µg kg−1) (µg kg−1) (µg kg−1) (µg kg−1) (µg kg−1) (µg kg−1)

Spectinomycin 300 345 397 200 224 252

Streptomycin 500 569 647 200 222 245
Dihydrostreptomycin 500 555 616 200 225 252
Colistin A 150 180 214 50 58 68
Colistin B 150 170 194 50 56 62
Amikacin – 10 10 – 10 10
Kanamycin A 100 120 143 150 170 192
Paromomycin 500 590 597 – 10 10
Apramycin 1000 1164 1355 – 10 10
Gentamycin C1 50 59 70 100 110 121
Gentamycin C1a 50 60 72 100 112 126
Gentamycin C2 + C2a 50 58 68 100 111 123
Neomycin B 500 565 638 1500 1670 1860

Table 4
Recoveries and precision obtained in meat of four food producing species (33 µg kg−1).
Analyte Bovine Poultry Porcine Fish (trout)

Recovery % SD CV % Recovery % SD CV % Recovery % SD CV % Recovery % SD CV %

Spectinomycin 82 11 13 74 9 12 73 3 3 12 2 17
Streptomycin 75 6 8 75 9 12 75 5 7 87 6 7
Dihydrostreptomycin 79 7 9 85 9 10 84 8 9 96 5 6
Colistin A 45 9 20 44 7 17 55 6 11 49 2 3
Colistin B 62 6 9 67 2 3 71 5 8 69 4 5
Amikacin 87 7 8 87 6 7 84 9 11 92 6 7
Kanamycin A 93 15 16 82 6 8 88 5 6 91 6 7
Paromomycin 80 6 8 82 8 9 82 7 8 87 5 5
Apramycin 82 6 7 87 5 6 82 6 8 85 5 6
Gentamycin C1 76 7 9 84 5 6 82 6 7 82 9 11
Gentamycin C1a 76 7 10 78 7 9 84 6 7 81 6 8
Gentamycin C2 + C2a 77 6 8 82 8 9 78 6 8 83 9 11
Neomycin B 76 5 6 66 8 12 71 8 12 65 5 7

Participation in proficiency tests (PT) is generally carried out to
conclude an intra-lab validation study and, later, to check the method
performances over the time. However, there are very few PTs for AGs in
food and none for colistins. Probably this happens because few la-
boratories have implemented this particular (but important) analytical
determination. Since 2014, our laboratory accredited a multiclass
method for ten classes of antibiotics in muscle and in milk (Moretti,
Dusi et al. 2016; Moretti, Cruciani et al., 2016). That method included
amphenicols, beta-lactams, diamino-pyrimidine, lincosamides, macro-
lides, pleuromutilins, quinolones, rifamycins, sulfonamides and tetra-
cyclines. Accordingly, we began the regular participation in the PTs
organized by Rikilt (Wageningen, NL) for antibiotics in meat. This PT
was the first organized to assess the performances of multiclass anti-
biotic procedures, whereas traditionally the PT rounds include only one
class of compounds. In the Reports of years 2014, 2016 and 2017 the
presence of AGs was declared by the organizer, but at that time, our
laboratory was not able to analyze AGs. Therefore, in June 2017 the
stored frozen muscle samples were reanalyzed with the new developed
procedure. The obtained data are listed in Table S9 and Z-scores cal-
culated using the information in the Reports (assigned value and
standard deviation for proficiency). The results were very satisfactory
proving the good performances of developed method and, secondarily,
the stability in matrix of the two found compounds in the muscle pre-
pared in 2014 and 2016, i.e. dihydrostreptomycin and kanamycin.
These data were coherent with the observations of Gaugain, Chotard
and Verdon (2013) who demonstrated the stability of these two AGs in
pork muscle stored for one year at −18 °C. Unfortunately, the test
materials provided by Rikilt contained spiked AGs and not incurred
residues, not giving complete information about the efficiency of the
extraction step.
4. Conclusions
The here reported method was rapid and accurate allowing the
screening/confirmation analysis of twelve AGs and two colistins in
meat and milk. To the best of our knowledge, this is the first procedure
able to detect simultaneously these two classes of antibiotics in food at
trace levels with performances comparable to those of the relevant
single-class methods. During the screening phase, different mass spec-
trometric experiments with limited m/z ranges of acquisition (full scan/
dds and SIM/dds) have to be performed in order to avoid the post in-
terface ion suppression, a typical phenomenon of Orbitrap analyser,
which can dramatically increase the detection limits. By doing that, this
type of platform can successfully compete with triple quadrupole ana-
lysers also from a quantitative point of view. Using WCX SPE pur-
ification, the composition of the extraction mixture must be carefully
regulated due to the influence of pH and ionic strength on the analyte
retention. Spectinomycin was the most critical compound and its ac-
curacy should be systematically monitored, especially when novel
matrices are analysed. Finally, acidic solutions are universally applied
to extract AGs and colistins from complex matrices. However, there is a
lack of information about this aspect due to the absence of reference
materials and proficiency tests involving the analysis of incurred sam-
ples. Therefore, the efficiency of the hydrolysis during the extraction
step should be better studied to assess the trueness of the current
analytical procedures.

15
G. Saluti et al.
Acknowledgements
This work was supported by the Italian Health Ministry
(“Development of methods for the determination of aminoglycosides in
meat and milk” - Ricerca Corrente IZSUM RC022014).
Appendix A. Supplementary data
Supplementary data associated with this article can be found, in the
online version, at https://doi.org/10.1016/j.foodchem.2018.05.113.
References
Alechaga, E., Moyeno, E., & Galceran, M. T. (2014). Mixed-mode liquid chromatography
coupled to tandem mass spectrometry for the analysis of aminoglycosides in meat.
Analytical Bioanalytical Chemistry, 406, 4941–4953.
Asakawa, D., Uemura, M., Sakiyama, T., & Yamano, T. (2017). Sensitivity enhancement of
aminoglycosides in hydrophilic interaction liquid chromatography with tandem mass
spectrometry by post-column addition of trace sodium acetate in methanol. Food
Additives & Contaminants: Part A, 1–11.
Bazzan Arsand, J., Jank, L., Martins, M. T., Barcellos Hoff, R., Barreto, F., Pizzolato, T. M.,
& Sirtori, C. (2016). Determination of aminoglycoside residues in milk and muscle
based on a simple and fast extraction procedure followed by liquid chromatography
coupled to tandem mass spectrometry and time of flight mass spectrometry. Talanta,
154, 38–45.
Berendsen, B. J. A., Meijer, T., Mol, H. G. J., van Ginkel, L., & Nielen, M. W. F. (2017). A
global inter-laboratory study to assess acquisition modes for multi-compound con-
firmatory analysis of veterinary drugs using liquid chromatography coupled to triple
quadrupole, time of flight and orbitrap mass spectrometry. Analytica Chimica Acta,
962, 60–72.
Bohm, D. A., Stachel, C. S., & Gowik, P. (2013). Validation of a method for the de-
termination of aminoglycosides in different matrices and species based on an in-
house concept. Food Additives & Contaminants: Part A, 30, 1037–1043.
Boison, J. O., Lee, S., & Matus, J. (2015). A multi-residue method for the determination of
seven polypeptide drug residues in chicken muscle tissues by LC-MS/MS. Analytical
Bioanalytical Chemistry, 406, 4065–4078.
Burns, D. T., Danzer, K., & Townshend, A. (2002). Use of the terms “recovery” and
“apparent recovery” in analytical procedures. Pure and Applied Chemistry, 74,
2201–2205.
Cruciani, G., Milletti, F., Storchi, L., Sforna, G., & Goracci, L. (2009). In silico pK(a)
Prediction and ADME Profiling. Chemistry & Biodiversity, 6, 1812–1821.
Diez, C., Giullarme, D., Staub Spörri, A., Cognard, E., Ortelli, D., Edder, P., & Rudaz, S.
(2015). Aminoglycoside analysis in food of animal origin with a zwitterionic sta-
tionary phase and liquid chromatography-tandem mass spectrometry. Analytica
Chimica Acta, 882, 127–139.
European Commission (2002). 2002/657/EC: Commission decision of 12 August 2002
implementing council directive 96/23/EC concerning the performance of analytical
methods and the interpretation of results. European Union Official Journal, L221, 8.
European Commission (2010). 2010/37/EU: Commission Regulation of 22 December
2009 pharmacologically active substances and their classification regarding max-
imum residue limits in foodstuffs of animal origin. Official Journal of the European
Union, L51, 1.
European Commission Health & Consumer Protection Directorate General, (2015).
SANTE/11495/2015 Guidance document on analytical quality control and validation
procedures for pesticide residues analysis in food and feed. < https://ec.europa.eu/
food/sites/food/files/plant/docs/pesticides_mrl_guidelines_wrkdoc_11945 > .
Accessed 04.12.17.
European Medicines Agency (EMA), (2016). Updated advice on the use of colistin pro-
ducts in animals 3 within the European Union: development of resistance 4 and
possible impact on human and animal health. EMA/CVMP/CHMP/231573/2016.
URL < http://www.ema.europa.eu/docs/en_GB/document_library/Scientific_
guideline/2016/07/WC500211080.pdf > . Accessed 04.12.17.
Farouk, F., Azzazy, H. M. E., & Niessen, W. M. A. (2015). Challenges in the determination
of aminoglycoside antibiotics, a review. Analytica Chimica Acta, 890, 21–43.
Galarini, R., Saluti, G., Moretti, S., Giusepponi, D., & Dusi, G. (2013). Determination of
macrocyclic lactones in food and feed. Food Additives & Contaminants: Part A. 30(6),
1068–1079.
Gaugain, M., Chotard, M. P., & Verdon, E. (2013). Stability study for 53 antibiotics in
16
Food Chemistry 266 (2018) 9–16
solution and in fortified biological matrixes by LC/MS/MS. Journal of AOAC
International, 96, 471–480.
Gremilogianni, A. M., Megoulaus, N. C., & Koupparis, M. A. (2010). Hydrophilic inter-
action vs ion pair liquid chromatography for the determination of streptomycin and
dihydrostreptomycin residues in milk based on mass spectrometric detection. Journal
of Chromatography A, 1217, 6646–6651.
Ishii, R., Horie, M., Chan, W., & MacNeil, J. (2008). Multi-residue quantitation of ami-
noglycoside antibiotics in kidney and meat by liquid chromatography with tandem
mass spectrometry. Food Additives and Contaminants, 25, 1509–1519.
Kaufmann, A., Butcher, P., & Kölbener, P. (2003). Trace level quantification of strepto-
mycin in honey with liquid chromatography/tandem mass spectrometry. Rapid
Communications in Mass Spectrometry, 17, 2575–2577.
Kaufmann, A., Butcher, P., & Maden, K. (2012). Determination of aminoglycoside re-
sidues by liquid chromatography and tandem mass spectrometry in a variety of
matrices. Analytica Chimica Acta, 711, 46–53.
Kaufmann, A., & Maden, K. (2005). Determination of 11 aminoglycosides in meat and
liver by liquid chromatography with tandem mass spectrometry. Journal of AOAC
International, 88, 1118–1125.
Kaufmann, A., & Widmer, M. (2013). Quantitative analysis of polypeptide antibiotic re-
sidues in a variety of food matrices by liquid chromatography coupled to tandem
mass spectrometry. Analytica Chimica Acta, 797, 81–88.
Kumar, P., Rubies, A., Companyó, R., & Centrich, F. (2012). Hydrophilic interaction
chromatography for the analysis of aminoglycosides. Journal of Separation Sciences,
35, 498–504.
Kunin, C. M. (1970). Binding of antibiotics to tissue homogenates. The Journal of Infectious
Diseases, 121, 55–64.
Lehotay, S. J., Mastovska, K., Lightfield, A. R., Nuñez, A., Dutko, T., Ng, C., & Bluhm, L.
(2013). Rapid analysis of aminoglycoside antibiotics in bovine tissues using dis-
posable pipette extraction and ultrahigh performance liquid chromatography-tandem
mass spectrometry. Journal of Chromatography A, 1313, 103–112.
McGlinchey, T. A., Rafter, P. A., Regan, F., & McMahon, G. P. (2008). A review of ana-
lytical methods for the determination of aminoglycoside and macrolide residues in
food matrices. Analytica Chimica Acta, 624, 1–15.
Milletti, F., Storchi, L., Goracci, L., Bendels, S., Wagner, B., Kansy, M., & Cruciani, G.
(2010). Extending pK(a) prediction accuracy: High-throughput pK(a) measurements
to understand pK(a) modulation of new chemical series. European Journal of Medicinal
Chemistry, 45, 4270–4279.
Moreno-Gonzales, D., Hamed, A. H., Gracia-Campaña, A. M., & Gámiz-Gracia, L. (2017).
Evaluation of hydrophilic interaction liquid chromatography-tandem mass spectro-
metry and extraction with molecularly imprinted polymers for determination of
aminoglycosides in milk and milk-based functional foods. Talanta, 171, 74–80.
Moretti, S., Cruciani, G., Romanelli, S., Rossi, R., Saluti, G., & Galarini, R. (2016).
Multiclass method for the determination of 62 antibiotics in milk. Journal of Mass
Spectrometry. 51, 792–804.
Moretti, S., Dusi, G., Giusepponi, D., Pellicciotti, S., Rossi, R., Saluti, G., ... Galarini, R.
(2016). Screening and confirmatory method for multiclass determination of 62 an-
tibiotics in meat. Journal of Chromatography A, 1429, 175–188.
Sin, D. W., Ho, C., Wong, Y., Ho, S., & Ip, A. (2005). Analysis of major components of
residual bacitracin and colistin in food samples by liquid chromatography tandem
mass spectrometry. Analytica Chimica Acta, 535, 23–31.
Tao, Y., Chen, D., Yu, H., Huang, L., Liu, Z., Cao, X., ... Yuan, A. (2012). Simultaneous
determination of 15 aminoglycoside(s) residues in animal derived foods by auto-
mated solid-phase extraction and liquid chromatography–tandem mass spectrometry.
Food Chemistry, 15, 676–683.
Vander Heyden, Y., Nijhuis, A., Smeyers-Verbeke, J., Vandeginste, B. G., & Massart, D. L.
(2001). Guidance for robustness/ruggedness tests in method validation. Journal of
Pharmaceutical and Biomedical Analysis. 24, 723–753.
Wan, E. C., Ho, C., Sin, D. W., & Wong, Y. C. (2006). Detection of residual bacitracin A,
colistin A, and colistin B in milk and animal tissues by liquid chromatography tandem
mass spectrometry. Analytical Bioanalytical Chemistry, 385, 181–188.
Xu, I., Tian, X., Ren, C., Huang, H., Gong, X., Liu, H., ... Zhang, L. (2012). Analysis of
colistin A and B in fishery products by ultra performance liquid chromatography with
positive electrospray ionization tandem mass spectrometry. Journal of
Chromatography B, 899, 14–20.
Yang, B., Wang, L., Chunying, L., Xixi, W., & Chengjun, S. (2017). Simultaneous de-
termination of 11 aminoglycoside residues in honey, milk, and pork by liquid chro-
matography with tandem mass spectrometry and molecularly imprinted polymer
solid Phase extraction. Journal of AOAC International, 100, 1869–1878.
Zhu, W., Yang, J., Wei, W., Liu, Y., & Zhang, S. (2008). Simultaneous determination of 13
aminoglycoside residues in foods of animal origin by liquid chromatography–elec-
trospray ionization tandem mass spectrometry with two consecutive solid-phase ex-
traction steps. Journal of Chromatography A, 1207, 29–37.