Analytical and Bioanalytical Chemistry

https://doi.org/10.1007/s00216-020-02751-0

RESEARCH PAPER

Capillary electrophoresis-mass spectrometry for the direct analysis

of glyphosate: method development and application to beer
beverages and environmental studies
Benedikt Wimmer 1 & Martin Pattky 1 & Leyla Gulu Zada 2 & Martin Meixner 1 & Stefan B. Haderlein 2 &
Hans-Peter Zimmermann 3 & Carolin Huhn 1

Received: 4 December 2019 / Revised: 22 May 2020 / Accepted: 29 May 2020

# The Author(s) 2020

Abstract
In this study, we developed and validated a CE-TOF-MS method for the quantification of glyphosate
(N-(phosphonomethyl)glycine) and its major degradation product aminomethylphosphonic acid (AMPA) in different samples
including beer, media from toxicological analysis with Daphnia magna, and sorption experiments. Using a background electro-
lyte (BGE) of very low pH, where glyphosate is still negatively charged but many matrix components become neutral or
protonated, a very high separation selectivity was reached. The presence of inorganic salts in the sample was advantageous with
regard to preconcentration via transient isotachophoresis. The advantages of our new method are the following: no derivatization
is needed, high separation selectivity and thus matrix tolerance, speed of analysis, limits of detection suitable for many applica-
tions in food and environmental science, negligible disturbance by metal chelation. LODs for glyphosate were < 5 μg/L for both
aqueous and beer samples, the linear range in aqueous samples was 5–3000 μg/L, for beer samples 10–3000 μg/L. For AMPA,
LODs were 3.3 and 30.6 μg/L, and the linear range 10–3000 μg/L and 50–3000 μg/L, for aqueous and beer samples, respec-
tively. Recoveries in beer samples for glyphosate were 94.3–110.7% and for AMPA 80.2–100.4%. We analyzed 12 German and
2 Danish beer samples. Quantification of glyphosate and AMPA was possible using isotopically labeled standards without
enrichment, purification, or dilution, only degassing and filtration were required for sample preparation. Finally, we demonstrate
the applicability of the method for other strong acids, relevant in food and environmental sciences such as N-acetyl glyphosate, N-
acetyl AMPA (present in some glyphosate resistant crop), trifluoroacetic acid, 2-methyl-4-chlorophenoxyacetic acid, glufosinate
and its degradation product 3-(methylphosphinico)propionic acid, oxamic acid, and others.

Keywords Matrix tolerance . Transient isotachophoresis . Food . Preconcentration

Published in the topical collection Persistent and Mobile Organic

Compounds – An Environmental Challenge with guest editors Torsten C.
Introduction
Schmidt, Thomas P. Knepper, and Thorsten Reemtsma.
Glyphosate (GLP) is a widely used broadband herbicide with
Electronic supplementary material The online version of this article
(https://doi.org/10.1007/s00216-020-02751-0) contains supplementary an annual production of more than one million tons [1].
material, which is available to authorized users. Various applications exist including pre-emergence use up
to desiccation before harvest, as discussed, e.g., by Schmitz
* Carolin Huhn and Garvert [2] for Germany. GLP blocks the shikimate path-
carolin.huhn@uni-tuebingen.de way which is vital to plants for the synthesis of aromatic
1
amino acids. Despite the fact that GLP is the most widely
Institute for Physical and Theoretical Chemistry, Eberhard Karls
applied herbicide worldwide [3, 4], a thorough understanding
Universität Tübingen, Auf der Morgenstelle 18,
72076 Tübingen, Germany of its fate in the environment including mobility, bioavailabil-
2 ity, and microbial degradation is still lacking. In the environ-
Center for Applied Geosciences, Environmental Mineralogy and
Chemistry, Eberhard Karls Universität Tübingen, Hölderlinstr. 12, ment, GLP can sorb to soil minerals which may limit bioavail-
72074 Tübingen, Germany ability and thus prevent quantitative mineralization. GLP
3
Agilent Technologies, Hewlett-Packard-Str. 8, readily forms chelate complexes with di- and trivalent metal
76337 Waldbronn, Germany ions, resulting in a large variety of possible GLP species at
 Wimmer B. et al.

environmental pH, altering its environmental behavior, not yet However, also contradicting results were reported [22]. Even
well understood [5]. Maximum legal contaminant limits in after derivatization, metal cations are problematic as the
water largely differ between EU with 0.1 μg/L and the USA FMOC-GLP derivative was shown to form very stable com-
with 700 μg/L [6]. An overview on findings of GLP in water plexes with divalent metal cations, which are separated in
samples was summarized in the reviews by Vereecken [7] and RPLC [22]. The selectivity of the derivatization process is
Saunders and Pezeshki [6], and for AMPA by Grandcoin et al. limited as all primary and secondary amines present, e.g., in
[8]. Human dietary exposure to GLP was discussed by food samples may be co-derivatized. Side reactions are possi-
Stephenson and Harris [9], who also presented a large over- ble and may lead to isobaric interferences [23, 24].
view on GLP findings in foodstuff. Accordingly, sample preparation techniques for matrix re-
One major reason for the relatively limited knowledge on moval are employed, e.g., SPE with various materials includ-
the environmental fate and impact of GLP and its main me- ing ion exchange resins [21, 25] or liquid-liquid extraction
tabolite aminomethylphosphonic acid (AMPA) and difficul- [26]. Direct GLP analysis without derivatization using LC
ties of their quantification in foodstuff stems from their phys- was achieved using mixed mode columns, ion exchange chro-
icochemical properties, making their sensitive and precise matography, or hydrophilic liquid interaction chromatography
analysis difficult (e.g., [10, 11]). GLP contains two acidic [10]. Another possibility for direct GLP analysis taking ad-
(phosphonic acid and carboxylic acid) and one basic (second- vantage of its high charge is ion chromatography. IC-MS/MS
ary amine) moieties, which make the molecule highly polar was applied to quantify GLP in lettuce, oranges, and wheat
and doubly charged at the pH range of approximately 6–8 flour extracts [27] within 10 min separation time. The instru-
relevant for most surface water samples (pKa values given in mental setup is complex with a metal-free ion chromatograph,
the literature 2.0/-/0.88, 2.6/2.32/2.22, 5.6/5.86/5.87, and eluent generator (to produce KOH as eluent), and an electro-
10.6/10.86/10.89; from [12–14]). Similarly, AMPA has a chemically regenerated suppressor to replace cations from el-
phosphonic acid and a primary amine as functional groups. uent and sample by H+ prior to the MS. Strong matrix effects
Thus, its effective charge number at pH 6–8 values ranges were observed. LOD was 2.5 μg/L in the extract from 10 g
from − 0.5 to − 1 (pKa values [15] 1.85/2.35, 5.35/5.9, and sample. Also two-dimensional ion chromatography coupled
10.0/10.8). GLP’s high polarity and high charge at intermedi- to mass spectrometry was used requiring a further pump and
ate pH, complex speciation regarding pH and complexation/ suppressor [28]. The instrumentation required a generation
chelation with multivalent metal cations, ability to adsorb on chamber to produce KOH as eluent, two IC separation col-
surfaces (mineral oxides but also fused silica), and the lack of umns, a concentrator, and two suppressors prior to the MS
suitable chromophores pose significant challenges to most if instrument. With this setup, LODs down to 0.05 μg/L were
not all analytical techniques. Koskinen et al. [16] presented an reached, however, with analysis times of 30 min. Matrix ef-
overview of analytical techniques and their application to dif- fects were briefly discussed using a model sample made of salt
ferent sample types (water/aqueous fluids, plant material, and solutions.
soil). Sensor methods and electromigrative separation tech- In contrast to chromatographic techniques,
niques were summarized by Gauglitz et al. [11]. Currently, electromigrative separation techniques take advantage from
gas chromatography (GC) with MS, electron capture, nitrogen the fact that GLP is negatively charged over a broad pH range
phosphorous or flame ionization detection, and liquid chro- [11]. Various direct methods without derivatization were pub-
matography (LC) with fluorescence detection or coupled to lished using indirect UV [29–32], fluorescence [33], capaci-
mass spectrometry are commonly used for sensitive GLP tively coupled contactless conductivity detection [31, 34–37],
analysis [17, 18]. However, the polarity of GLP prevents its inductively coupled plasma-MS [38], or ESI-MS [39–43],
direct analysis by GC or reversed phase LC. These methods though applications with the latter coupling are still scarce.
require a laborious and expensive derivatization procedure to It is interesting to note that the full pH range has been used
provide the derivative with sufficient volatility, thermal stabil- for the background electrolyte (BGE) starting from 2.45 [40]
ity, and sufficient retention, and to allow sensitive detection to 10 [42]. However, as also shown in this study, adsorption
[19]. For RPLC, mostly 9-fluorenylmethoxycarbonyl chloride on the capillary wall might be problematic. Long rinsing pro-
(FMOC) is used, but various other label reagents may be used tocols with strong acids or bases (e.g., [30, 32, 44–49]) and
[19]. Derivatization may be impaired by metal ions present in permanent [39, 41, 43, 50] or dynamic coatings [29–32, 34,
the sample matrix. This effect can be reduced by adding eth- 35, 51–53] have been used to improve precision and resolu-
ylenediaminetetraacetic acid (EDTA) to evoke competitive tion. As for chromatographic techniques, derivatization of
metal cation chelation, as envisaged in the ISO method GLP and AMPA was conducted using various amine-based
16308 [20]. These complexes impair derivatization yield, labeling reagents, e.g., fluorescein isothiocyanate (FITC) [45,
most likely due to changing the basicity of the glycyl group 46, 52, 54, 55], FMOC [49, 56, 57], and others [44, 55, 58].
or having it involved in the ligand sphere rendering it less The limited resolution was overcome using cyclodextrins [46]
active for electrophilic attack by the FMOC reagent [21]. or detergents in micellar electrokinetic chromatography
Capillary electrophoresis-mass spectrometry for the direct analysis of glyphosate: method development and...

(MEKC) [44, 45, 53, 54, 57–59]. UV detection at low wave- Hydrochloric acid (32%, analytical grade) was obtained from
lengths [49, 53, 56, 57, 60] or LIF detection, mostly at 488 nm Fisher Scientific (Schwerte, Germany). Twenty-five percent
[44–46, 52, 54, 55, 58, 61], was often used for detection. aqueous ammonia solution (p.a. grade), aluminum chloride,
Literature reveals a high matrix tolerance of CE methods iron(III) chloride hexahydrate, manganese(II) chloride
for GLP analysis. When combined with a selective detection tetrahydrate, magnesium chloride hexahydrate (99.6%), zinc
method such as LIF or MS, impressive results were obtained acetate dihydrate (> 99%), aminomethylphosphonic acid (>
for complex samples including environmental samples such 99%), 2-methyl-4-chlorophenoxyacetic acid (99.2%), oxamic
as ground water and surface water [32, 33, 35, 37, 38, 44, 46, acid (> 98%), difluoroacetic acid (98%),
49, 51, 55, 56, 60] or soil [41, 55, 61], but also food such as 3-(methylphosphinico)propionic acid (98%), and
beverages [41, 42, 50], soy or wheat products [29, 39, 43, 57, iminodiacetic acid were obtained from Sigma (Steinheim,
62], and vegetables or fruits [43, 52, 59]. Furthermore, herbi- Germany). Labeled 13C2-15N-glyphosate, N-nitroso glypho-
cides and reaction mixtures from herbicide production [33, 40, sate, N-acetyl glyphosate, and N-acetyl AMPA were bought
42, 51], marijuana seizures [53], and human serum [63] were from TRC/BIOZOL (Echingen, Germany), 13C-15N-D2-
analyzed. Most studies used spiked samples to demonstrate AMPA from LGC Standards (Wesel, Germany), phosphonic
the matrix tolerance of the methods. Online preconcentration acid from HPC Standard (Cunnersdorf, Germany), and
methods for GLP analysis using CE presented so far comprise trifluoroacetic acid (99%) from VWR (Darmstadt,
acetonitrile stacking [52, 62], field-amplified sample stacking Germany). The low concentration standard tune mix, purine,
[29], large volume sample stacking [29, 47], and column- and HP-921 were from Agilent Technologies (Waldbronn,
coupled isotachophoresis (both in capillaries [36] and on- Germany). The OHNOON solution for the capillary coatings
chip [48]). Also, field-amplified electrokinetic injection was was prepared as described elsewhere [64]. Water from a puri-
used [29, 34, 57], however, with low matrix tolerance regard- fication system from ELGA LabWater (Celle, Germany) was
ing the salt matrix of many samples, including surface water used.
[29, 34].
In this study, we developed a direct (derivatization-free) Samples and sample preparation
highly selective and rapid analytical method for quantification
of GLP and AMPA using CE-QTOF-MS. The method proved Aqueous stock solutions of GLP, 13C2-15N-glyphosate ([M-
to be applicable for a large variety of matrices including beer H] − m/z 171.010, referred to as GLP171), AMPA, and
beverages and aqueous samples from environmental sorption 13
C-15N-D2-AMPA ([M-H]− m/z 114.001, referred to as
studies. Online sample preconcentration is discussed as well AMPA114) were prepared at a concentration of 1 g/L (5.92,
as matrix effects. Additionally, we demonstrate the method’s 5.81, 9.01, 8.69 mmol/L). GLP and AMPA stock solutions
potential to analyze further pollutants/contaminants relevant were diluted to obtain concentrations of 3 and 1 mg/L, and
for food and environmental samples such as N-nitroso glyph- 700, 500, 200, 100, 70, 50, 30, 10, 5 and 2 μg/L. For calibra-
osate (NNG), N-acetyl glyphosate (NAG), N-acetyl AMPA tion in beer matrix, a mixture was used of GLP171 and
(NAA), glufosinate (GLU), 3-(methylphosphinico)propionic AMPA114 with 4 and 10 mg/L, respectively. For GLP quan-
acid (3-MPPA), oxamic acid, 2-methyl-4- tification in different beer samples, aqueous stock solution of
chlorophenoxyacetic acid (MCPA), difluoroacetic acid GLP171 was diluted with water to 2 mg/L. Aliquots of the
(DFA), and trifluoroacetic acid (TFA), phosphonic acid and stock solutions were stored at + 4 °C for maximum 6 months.
iminodiacetic acid (IDA). Stock solutions of DFA, GLU, IDA, MCPA, 3-MPPA, NAA,
NAG, NNG, oxamic acid, phosphonic acid, and TFA were
prepared at a concentration of 1 g/L; a mixture of the former
Materials and methods analytes plus GLP and AMPA was prepared with a concen-
tration of 50 mg/L, and diluted to achieve injection solutions
Chemicals with final concentrations of 250 and 500 μg/L.

Isopropanol (LC-MS grade), methanol (LC-MS grade), Samples of sorption experiments and toxicological studies
glyphosate (> 99.7%), glufosinate ammonium salt (> 98%), Supernatants of sorption experiments with GLP on Al2O3 par-
and lead(II) nitrate (> 98%) were purchased from Fluka ticles were prepared with 0.5 mmol/L KCl and initial GLP
(Steinheim, Germany). Formic acid (FA) (98–100%, LC-MS concentrations between 2 and 8 mg/L, the solutions were
LiChropur), formic acid-d2, sodium hydroxide (30%, injected directly without further treatment, and quantification
Suprapur), cadmium(II) chloride monohydrate (> 98%), was achieved by external calibration (10 to 2000 mg/L) [65].
nickel(II) chloride hexahydrate (> 98%), calcium chloride Toxicological studies with Daphnia magna [66] were con-
dihydrate (> 99.5%), and copper(II) chloride dihydrate (> ducted in a medium made of 2 mmol/L CaCl2, 0.5 mmol/L
99%) were bought from Merck (Darmstadt, Germany). MgSO4, 0.75 mmol/L NaHCO3, and 0.08 mmol/L KCl
 Wimmer B. et al.

according to ISO 6341 [67]. Calibration was done in the range isopropanol (10 min) and BGE (20 min). Between runs, the
of 25 to 200 μg/L in one 25 and three 50 μg/L steps. The capillary was flushed for 5 min with BGE. For storage, the
medium was provided by R. Triebskorn and D. Werner from capillary was flushed with BGE, isopropanol, and air for
the Institute of Evolution and Ecology in Tübingen. 10 min each, and stored in dry conditions. The final BGE
chosen for analysis was 175 mmol/L FA titrated to pH 2.8
Metal complexation To investigate the influence of metal with ammonia (final concentration of approximately
complexation on CE separations, aqueous stock solutions 40 mmol/L ammonia). BGE was exchanged after 10 runs to
with a concentration of 10 mmol/L of divalent and trivalent keep a high migration time precision. Separations were con-
cations were prepared for each salt (MnCl2, CuCl2, CaCl2, ducted at 25 °C (inside the CE housing) using a voltage of −
MgCl2, Zn(OAc)2, NiCl2, CdCl2, Pb(NO3)2, AlCl3, and 30 kV. In order to reduce analyte migration times and ensure
FeCl3). The injection solution was prepared with a final con- stable electrospray conditions, an optimized inlet pressure of
centration of 1 mmol/L cation salt and 200 μg/L of GLP and 30 mbar was applied during CE-MS analysis.
AMPA (1.2 and 1.8 μmol/L, respectively, pH 2–4). The molar Standards were injected as aqueous solutions in 5 mmol/L
ratio between the cation and GLP was 850:1, and between ammonium formate buffer. Samples were injected at 75 mbar
cation and AMPA 560:1. Before analysis, injection solutions for 10 s (18 nL) if not stated otherwise. Large volume injection
were equilibrated for 12 h at room temperature. (LVI) with aqueous standards was done with 75 mbar for 40 s
(71 nL), for beer beverages with 100 mbar for 40 s (94 nL).
Beer samples Fourteen beer samples from 2016 (Pilsner and Analyte injection was followed by dipping the capillary into
naturally cloudy breed, see Application to beer samples two extra vials filled with BGE to avoid analyte carryover
section) were purchased from local stores. Beer samples were between consecutive runs. The transfer of the method devel-
degassed by sonication for 15 min; naturally cloudy beer bev- oped using aqueous standards to the analysis of samples with
erages were additionally filtered with Chromafil Xtra PTFE- complex matrices required several adaptations. (1) We had to
45/25 filters (Macherey-Nagel, Düren, Germany). Calibration use a voltage ramp, starting at − 15 kV, decreasing within
in beer matrix was done with an organic beer sample (Fidelio) 2 min to − 30 kV to account for the high conductivity of some
in a similar range as for aqueous calibration (2 μg/L to 3 mg/ of the samples: at the beginning of the separation, this high
L), with additionally 200 μg/L of GLP171 and 500 μg/L of conductivity zone evoked (a) high electric field strength in the
AMPA114 for quantification; to estimate the LOD for GLP BGE zone and thus run failures or (b) high current alarms and
using large injection volumes, calibration samples of 2 and voltage re-adjustment, resulting in migration time shifts. (2)
5 μg/L were used. For quantification with the internal standard After sample injection, a plug of running buffer was injected
method (ISM) for all other beer samples, 500 μL of beer was at 100 mbar for 5 s to avoid sample components diffusing into
mixed with 10 μL of GLP171 stock solution to achieve a final the running buffer vial. (3) To further avoid carry over effects,
isotope standard concentration in the injection solution of the electrode was washed with BGE after each run using the
39 μg/L. For quantification via standard addition, “wash inlet electrode” command of the CE software.
Hasseröder Premium Pils was spiked with 10, 20, and
30 μg/L GLP. The mixture solution of AMPA, DFA, GLP, Mass spectrometry An Agilent 6550 iFunnel Q-ToF-MS in-
GLU, IDA, MCPA, 3-MPPA, NAA, NAG, NNG, oxamic strument (Agilent Technologies, Santa Clara, CA) was
acid, phosphonic acid, and TFA containing each analyte at a coupled to the CE. A coaxial sheath liquid electrospray inter-
concentration of 50 mg/L was used to spike an organic beer face from Agilent Technologies (Waldbronn, Germany) and a
(Fidelio) at a final analyte concentration of 250 and 500 μg/L Dual-ESI ionization source were used with an electrospray
in the injection solution. All aqueous and real samples were needle consisting of 80% platinum and 20% iridium
stored at − 18 °C. (Agilent Technologies, Waldbronn, Germany). To improve
the resulting electrospray, the needle geometry was optimized
Instrumentation (compare section Platinum-iridium electrospray ionization
needle). The sheath liquid (final method with a 50:50 (v/v)
Capillary electrophoresis A 7100 Agilent CE System was used mixture of isopropanol:water containing 0.01% FA) was
for CE-MS analysis. Capillaries coated with polyvinyl alcohol degassed upon ultrasonication and delivered by an isocratic
(PVA) were obtained from Agilent (Waldbronn, Germany) 1260 infinity pump (Agilent Technologies, Waldbronn,
with an i.d. of 50 μm and cut to a length of 65 cm. Bare fused Germany) at a flow rate of 5 μL/min with a split ratio of
silica capillaries were obtained from Polymicro Technologies 1:100. For online recalibration during CE-MS analysis, the
(Phoenix, USA). For acidic and alkaline buffers, an aqueous sheath liquid contained 0.2 μmol/L purine and 0.1 μmol/L
FA solution was titrated with aqueous ammonia to achieve the HP-921 (both from Agilent Technologies; for HP-921, adduct
desired pH and degassed by sonication for 5 min prior to with FA (m/z 966.001) was the reference). MS instrument
analysis. Before first use, capillaries were flushed with parameters: Drying gas was delivered at 11 L/min at 150 °C,
Capillary electrophoresis-mass spectrometry for the direct analysis of glyphosate: method development and...

and nebulizer pressure was set to 5 psi during measurements. were used, most of them not compatible with MS detection. In
During preconditioning, injection and the first 6 s of the mea- our work using bare fused silica capillaries, peak shapes were
surement, the nebulizer pressure was lowered to 3 psi (reduced not acceptable in ammonium acetate–based BGEs at any pH.
suction effects during vial handling). The acquisition rate was For analysis by electromigration techniques, the low charge
2 Hz; m/z range 50–1700; fragmentor voltage 380 V; states of AMPA at low pH have to be taken into account,
electrospray voltage 4000 V. MS calibration was performed which led to a detection close to or with neutral substances.
with the low concentration Tune Mix from Agilent At higher pH, resolution was lowered (as also observed by
Technologies. Vidal [43]) and the overall selectivity for a direct method
without derivatization was low as many organic acids are
Data processing EICs were extracted and evaluated from mass charged at this pH regime and may thus impair the analysis
profiles with a mass accuracy of 10 ppm using Mass Hunter or necessitate further sample pretreatment. For our study, we
Qualitative Software. Calibration curves in aqueous solution decided to use permanent capillary coatings (both electrostat-
and beer matrix were evaluated from mass centroids with a ically adsorbed and covalently coupled) and a BGE of low pH
mass accuracy of 20 ppm using MassHunter Quantitative for analysis. A cationic OHNOON coating was tested without
Software, the linear range was determined by the signal areas, success; strong interaction between the coating and GLP and
matrix effects were expressed by (sensitivity matrix)/(sensitiv- AMPA was observed, even with effects from open tubular
ity aqueous solution) × 100% (with the sensitivity being the capillary chromatography (details can be found in the
slope of calibration curve (signal area vs. spiked concentra- Electronic Supplementary Material (ESM) and in Fig. S1).
tion)), and recovery in aqueous solution and beer matrix by A neutral PVA-coated capillary exclusively presents non-
(calculated concentration)/(spiked concentration) × 100%, acidic hydroxyl groups to the electrolyte solution; hence, in-
LODs (signal to noise ratio SNR = 3) were estimated based teraction with the phosphonate group was low at acidic pH. As
on the SNR (mass profiles, 10 ppm, Qualitative Software) at visible from Fig. 1 good peak shapes, resolution and separa-
the lowest calibration concentration, in case of AMPA in beer tion efficiency with plate numbers of up to 85,000 (130,000/
matrix at 50 μg/L. All figures were created with Origin 9.1.0G m) were achieved for the analytes under optimized conditions
(OriginLab Corporation, Northampton, USA). (175 mmol/L FA titrated to pH 2.8 using approximately
40 mmol/L ammonia). However, different capillary batches
significantly differed in performance regarding migration time
Results stability and signal shape, possibly due to an aged PVA coat-
ing (data not shown). We observed that the optimized BGE
Strategies to prevent glyphosate adsorption to the made of formic acid and ammonia was also suitable for sep-
capillary wall aration on bare fused silica capillaries (which was not possible
using acetic acid–based BGEs), when the samples contained
Strong tailing of GLP signals on bare fused silica capillaries significant amounts of phosphate, as present, e.g., in soil ex-
was observed previously [39] and in our study using BGEs at tracts, which competes with sorption sites [74]. However, the
pH 2 to 9. GLP was detected as a broad signal over several use of PVA-coated capillaries is required when samples con-
minutes. In later runs, we even failed to detect it, presumably tain proteins, since protein sorption onto the bare fused silica
due to irreversible binding and changes of the electroosmotic capillary inner surface can hardly be avoided [75]. Distinct
flow. On the first glimpse, adsorption may seem unlikely due
to ionic repulsion at high pH and the almost neutral silica
surface at low pH. However, from sorption studies, it is
known that GLP predominantly binds to oxidic soil minerals
via hydrogen bonding with its phosphonate group [7, 68–72].
In some cases, the carboxylic acid group is also involved in
the binding event. The adsorption thus evokes inner sphere
complexes, mostly with five- or six-membered rings with rel-
atively high binding constants.
Strategies against GLP adsorption in capillary electropho- Fig. 1 EICs of GLP and AMPA (1.69 and 1.11 mg/L (each 10 μmol/L),
resis were summarized by Gauglitz et al. [11] and may include respectively) separated on a PVA-coated capillary (length 65 cm, i.d.
derivatization, the use of phosphate-based BGEs (e.g., Chui 50 μm) using BGEs of ammonium formate at pH 2.8 (FA titrated with
et al. [62]), working at elevated pH [33, 37, 42], extremely low aqueous ammonia) with different ionic strength (see figure legend).
Injection was accomplished at 50 mbar for 5 s, separation voltage was
pH [40], reduction of hydrogen bonding by shielding the cap- − 30 kV, 50 mbar pressure were applied. Sheath liquid was
illary surface using dynamic [29, 34, 73] or permanent [39, 41, isopropanol:water 1:1 with additional 0.1% FA, flow rate 0.5 mL/min
43, 50] coatings. Also other separation modes, mostly MEKC, (1:100 split)
 Wimmer B. et al.

matrix effects and possible effects by sample-induced tran- GLP. For AMPA, however, peak tailing and low migration
sient ITP will have to be studied for such samples. time precision were observed (see ESM Fig. S3).
Using 100 mmol/L FA titrated to pH 2.8, peak area preci-
Platinum-iridium electrospray ionization needle sion and resolution were good for all analytes. Separation
efficiency was higher than for any other pH value tested, but
A standard stainless steel ESI needle used in negative MS for GLP, a slight tailing was observed (see Fig. 1a). It has to be
polarity and reverse CE polarity leads to strong corrosion of noted that AMPA is neutral at this pH (anionic only at pH >
the electrospray needle, accompanied by the migration of met- 4), so it is transported only by EOF and the additional pressure
al cations into the separation capillary, which may even impair applied during separation. For complex samples, AMPA
the separation [76]. For reliable CE-MS analysis of anionic quantification is impaired by the presence of neutral matrix
compounds, the use of a platinum-iridium needle is necessary components. We accepted this drawback as our major focus
[76]. However, the commercially available platinum-iridium was on GLP analysis and as the use of an isotopically labeled
ESI needle (Agilent Technologies, Waldbronn, Germany) has AMPA standard enabled its quantitative analysis (see below).
a cylindrical tip geometry, which lacks coaxial focusing of the
nitrogen gas flow at the sprayer tip. Increased migration times Ionic strength
compared with the iron needle were observed under identical
separation conditions due to lower suction effects. A tapering In preliminary experiments, we observed that the ionic
step as common for steel ESI needles by short electrochemical strength was important to achieve a high separation efficiency
polishing to decrease the tip wall thickness was not successful. and higher migration time precision (which was low when low
Finally, the performance of the platinum-iridium electrospray concentrations of FA were used without ammonia to reach the
needle was optimized with regard to ESI stability (TIC noise desired pH). In order to further optimize separation efficiency
RSD < 10%), signal intensity, and migration time precision by and signal shape, especially for GLP, BGEs with elevated
reducing the wall thickness via grinding and mechanical concentrations of 175 mmol/L and 250 mmol/L FA titrated
polishing of the needle tip (by goldsmith Ulrich Wehpke, to pH 2.8 using ammonia were investigated. The BGE with
Krefeld, Germany). Different geometries of the platinum- 175 mmol/L FA revealed improved peak shapes due to the
iridium needle tip are shown in ESM Fig. S2. higher ionic strength (see electropherograms in Fig. 1), and up
to 20% higher GLP signal areas compared with 250 mmol/L
Method optimization FA. Additionally, using 250 mmol/L FA, the separation volt-
age was automatically reduced within the first 3 min of sepa-
pH and ionic strength of the BGE were optimized using PVA- ration, to keep the separation current below 40 μA avoiding
coated capillaries. For method optimization, we chose com- Joule heating. With increasing ionic strength, the separation
mon MS-compatible BGEs made of FA and ammonia. current declined more extensively due to the counter ion of the
Method optimization focused on separation efficiency, signal BGE (ammonium ions) being gradually replaced by protons
shapes as well as migration time precision as indicators for from the sheath liquid upon electromigration to the inlet [77].
reduced adsorption phenomena, and the overall analysis time. Accordingly, the pH was reduced during the run depending on
For BGE optimization, an additional pressure of 50 mbar was the initial ammonium concentration in the BGE. This phe-
applied during separation; for the optimized method, the pres- nomenon can hardly be avoided. pH steps will be present in
sure was set to 30 mbar as a compromise between stable most CE-MS methods, even when BGE and sheath liquid
separation and ESI conditions and minimized signal broaden- contain the same ions and only concentration differences are
ing due to the parabolic flow profile. present [78]. We used 175 mmol/L FA with 40 mmol/L am-
monia as final buffer composition giving best signal shapes,
pH of the BGE responses, and reduced Joule heating. Just flushing the capil-
lary with BGE between runs was sufficient to maintain high
In principle, fast analysis is possible at basic pH due to the precision (migration times RSD < 0.4, peak area RSD < 3, n =
high analyte charge. However, we were not able to establish 10); for real samples the electrode was additionally washed to
stable separation conditions at pH 9.2. Increasing peak asym- avoid matrix carry over into the run buffer vial (see below).
metry and peak broadening (full width at half maximum in-
creased from 0.12 to 0.25 min) and migration time shifts (RSD Optimization of sheath liquid composition
> 10%, n = 10) unveiled pronounced interaction between GLP
and the PVA capillary surface at high pH. In contrast, peak The isopropanol:water ratio of the sheath liquid was varied in
widths and migration times for AMPA remained constant. the range of 35–65% (v/v), flow rates between 0.5 and
Using 100 mmol/L FA titrated with ammonia to pH 3.6, peak 0.65 mL/min, and a nebulizer pressure in the range of 3–6
area precision and migration time stability were very good for psig. In contrast to many other studies (e.g., ref. [79]), we
Capillary electrophoresis-mass spectrometry for the direct analysis of glyphosate: method development and...

did not find a pronounced impact of the solvent:water ratio on (Fig. 2a, 75 mbar for 10 s, 18 nL) and LVI (Fig. 2b, 100 mbar
GLP ionization efficiency, as long as the water content was for 40 s, 95 nL), large signals of inorganic salts were observed
kept at or below 50%. Similarly, little influence of the sheath as well as a moderate migration time shift for GLP from 6 to
liquid flow rate and nebulizer pressure was observed, pointing 6.5 min. Obviously, sITP was evoked by the sample compo-
to a high robustness of the ESI process for GLP ionization. nents chloride, nitrate, sulfate, and phosphate acting as tran-
With 35% water in the sheath liquid, the separation current sient leaders (in order of decreasing electrophoretic mobility),
fluctuated sinusoidally by 10% during the run, occasionally its relevance depends on its concentration. At LVI conditions,
accompanied by electrical contact loss and interruption of the c h l o r id e , n i t r a te , a n d s u l f a te w e r e st i l l s t a c k e d
measurement. For further measurements, an isotachophoretically upon detection. Destacking was proven
isopropanol:water ratio of 1:1 and a nebulizer pressure of to start at the front boundary (MS direction) of the sITP stack
1.36 bar (5 psig) were used. (data not shown). At too high injection volumes, the capillary
Principally, low concentrations of formic or acetic acid in length was too short to fully resolve the sITP stack and GLP
the sheath liquid of 0.1–1.0% guarantee adequate conductivity was stacked behind phosphate.
of the sheath liquid [80]. However, we observed pronounced Experiments with doubly deuterated FA indicated that FA
ion suppression at these concentrations, corroborating find- has an electrophoretic mobility suitable to be a transient ter-
ings for LC-MS analysis with eluents with a high FA concen- minator only for GLP. The extent of sITP depends on the
tration [81]. Decreasing the FA concentration from 0.1 to concentration of the transient leaders. In beer samples, the
0.01% increased signal intensity by 60%. Highest signal in- most critical macrocomponent was phosphate (phosphorous
tensities were observed at 0% FA in the sheath liquid, how- content in beer in the range of 0.3–15 mmol/L [86]). Using
ever, at the cost of a reduced electrospray stability, so 0.01% disodium phosphate at a concentration of 75–100 mM for
FA (2.6 mmol/L) was used as a compromise. Using alkaline GLP extraction from soil, the transient ITP did not fully re-
sheath liquid conditions with 0.02 or 0.5% ammonium hy- solve and GLP was detected in the isotachophoretic stack
droxide (2 and 50 mmol/L), ionization efficiency of GLP de- (data shown elsewhere [82]). At the highest injection volume,
creased by 20 and > 90%. In addition, migration of counter- poor resolution between GLP and phosphate led to quenching
ions, presumably ammonium ions from the SL into the capil- of the GLP signal; poor resolution between GLP and a near
lary, was observed (for discussion, see ESM Section D with isobaric matrix compound (m/z 167.996) impaired accurate
Fig. S4 and [82]), accompanied by reduced migration times by quantification (see ESM Fig. S5); in addition, co-migration
12 to 22% and lower migration time precision (see Ionic of a matrix compound near isobaric to GLP171 was also ob-
strength section). served. When using the Q-ToF in MS/MS mode, reduced
sensitivity due to ion losses during fragmentation was ob-
Online preconcentration served. Thus, injection conditions had to be limited to

Only few online preconcentration methods are suitable for

samples of elevated ionic strength [83]. For beer samples,
sample-induced transient isotachophoresis (sITP) was possi-
ble with (in)organic ions present as ionic macrocomponents
[84, 85] with suitable electrophoretic mobilities.

Aqueous standards In order to improve detection limits, LVI

was investigated with injection volumes of 18–89 nL (injec-
tion at 75 mbar for 10 to 50 s) using an aqueous GLP solution
of 10 μg/L (59 nM). At high injection conditions (75 mbar for
40 s, 71 nL and thus 4 times higher than without LVI), LOD
was lowered by a factor of 3 with acceptable peak broaden-
ing. AMPA is neutral at pH 2.8 and band broadening was
significant (20 to 70 s base width); the LOD was not Fig. 2 sITP in CE-MS electropherograms of beer samples with EICs of
anions in relative intensity (left axis), EIC of GLP in counts (right axis) in
improved. Wicküler Pilsner beer originally containing ca.14 μg/L GLP for a stan-
dard injection at 75 mbar for 10 s (18 nL) and b LVI at 100 mbar for 40 s
Beer samples LVI combined with sITP using natural ionic (95 nL). Separation was achieved on a PVA-coated capillary (i.d. 50 μm,
macrocomponents of the samples as transient leaders was test- length 65 cm) using a BGE with 175 mmol/L FA titrated to pH 2.8 with
ammonia, and a separation voltage of − 30 kV with additional 30 mbar
ed as preconcentration method for organic beer samples pressure. Within the first 2 min of separation, the voltage was ramped
spiked with 2 or 5 μg/L GLP and injection volumes of 70– from − 15to − 30 kV. Sheath liquid was isopropanol:water 1:1 with ad-
118 nL (100 mbar for 30–50 s). Comparing standard injection ditional 0.01% FA
 Wimmer B. et al.

100 mbar and 40 s to assure baseline separation of GLP and Method performance
these matrix compounds.
Separation selectivity and matrix tolerance

Influence of divalent cations on the separation of GLP The broad applicability of the developed method is exempli-
fied by the analysis of GLP in various samples differing in
Di- and trivalent metal cations were added to the sample to their matrix. GLP was quantified in aqueous solutions of sorp-
investigate influences on CE separations at molar ratios be- tion experiments to Al2O3 nanoparticles as a model of soil
tween cations and GLP of 850:1 (see Samples and sample minerals and in the exposure medium for Daphnia magna
preparation section). The electropherograms from injection during a toxicological study. Beer beverages represent a com-
solutions with divalent cations (Mn2+, Cu2+, Ca2+, Mg2+, plex matrix containing many organic acids. The conductivity
Zn2+, Ni2+, Cd2+, Pb2+) showed a slight increase in migration of the BGE (175 mmol/L FA titrated to pH 2.8 using ammo-
times of 5–10 s compared with sample solutions void of these nia) was 2.7 mS/cm, of the exposure medium 0.6 mS/cm, and
cations. This is caused either by increased conductivity and of beer beverages around 1.8 mS/cm. Hence, field-amplified
thereby reduced electric field strength within the sample plug sample stacking can be expected for the aqueous samples
[87], by transient ITP phenomena, or by a relatively fast (com- (sorption and Daphnia medium) but not for beer samples.
pared with the analysis time) on-capillary dissociation of Representative base peak currents (BPC) and extracted ion
GLP-metal complexes at low pH [88]. However, these in- currents (EIC) of GLP (m/z 168.007) for the three sample
creased migration times proved to be stable (RSD < 0.3%). types are shown in Fig. 3. For all samples, separation of
The signal area differences were within the quantitative GLP from most matrix components was achieved within 5–
precision and thus comparable for different cation solu- 8 min. Sorption experiment and Daphnia medium samples
tions (RSDs < 6%, n = 3 within one run series, over all (Fig. 3a and b) showed relatively clean electropherograms
run series RSD = 5%, n = 9). This low influence of met- with mainly chloride and sulfate visible. The signals at 8–
al cations in CE is presumably due to the high lability 9 min of the Daphnia medium were from organic
of these complexes in the BGE of pH 2.8: when GLP-
metal complexes (which would be neutral or positively
charged) are present but dissociate quickly at the begin-
ning of the run, GLP and the cations migrate in oppo-
site direction, hindering complex reformation. We did
not observe signal broadening so that we assume very
fast dissociation kinetics; high GLP signal areas of 96–
107% compared with sample void of cations were
obtained.
In contrast, the GLP complexes of trivalent Al3+ and Fe3+
proved to be relatively inert in the separation buffer and im-
paired separation; the recovery of GLP from Al-containing
solutions was only 49% at a molar ratio of 850:1. We ob-
served a strong broadening of the GLP signal with peak base
widths over 1 min. We were able to reduce but not fully
overcome this matrix effect by adding 50 mmol/L Na2HPO4
to the sample, which leads to precipitation of insoluble alumi-
num or iron phosphates. The addition of 20 mmol/L EDTA
(common for derivatization in LC [21]) to the injection solu-
tion spiked with Al3+ did not improve the separation; instead,
a dynamic equilibrium was induced between GLP-Al, EDTA-
Al, and GLP-EDTA-Al complex species, which were poorly Fig. 3 Base peak current (m/z-range 50–800, excluded masses are
reference masses, sulfate, phosphate, and clusters thereof) in
separated but could be discriminated with MS (data not logarithmic scale (left axis) and EIC of GLP (m/z 168.007) with shaded
shown). In contrast to our results for CE-MS, the presence GLP signal (right axis). The samples were from a aqueous sorption
of Al3+ and Fe3+ did not impair RPLC-ESI-MS, but strong studies on Al2O3 nanoparticles with 0.5 mmol/L KCl, b toxicological
retention time shifts were observed for divalent Cu, Zn, or Mn studies in culture medium (additional 50 mbar pressure during
separation), and c a German beer (Hasseröder Pils). GLP concentrations
[22]. These differences may in part be explained by the dif- were determined to be a 18.6 μg/L, b 1 mg/L, and c 14.4 μg/L. Injection
ferent pH values used for the separation (pH 9 for RPLC-MS was done hydrodynamically at 75 mbar for 10 s. Further separation
[22] and 2.8 for our CE-MS method). conditions as in Fig. 2
Capillary electrophoresis-mass spectrometry for the direct analysis of glyphosate: method development and...

components. In contrast, the BPC of beer samples (Fig. 3c) Quantitative aspects, linear range, and limits of detection
revealed many large signals over the whole electropherogram;
inorganic salts and organic acids are present. Various organic Three different methods for quantification were tested: exter-
matrix components (migration time range 3.5–6 min) and in- nal calibration with aqueous standards, standard addition, and
organic anions (compare Fig. 2) migrated faster than GLP and internal standard method (ISM). Quantification via external
were well separated from the analyte. A matrix component calibration can only be applied, if matrix effects are negligible.
with m/z 361.200 and possibly also organic phosphates mi- Due to the CE method’s high selectivity, this was the case for
grated directly in front of GLP; tailing or incomplete stacking samples from sorption experiments and toxicological studies
of these matrix components probably influenced ionization with Daphnia magna (both mainly containing inorganic
efficiency, but also GLP preconcentration by sITP. After salts). For beer samples, matrix effects (co-migrating
8 min, mostly small organic acids (not further identified) were compounds with high signal intensities) impaired ionization
observed, which were present in beer in relatively large quan- efficiency of GLP, so that standard addition or quantification
tities [89]. Overall, only strong acids can be expected as most via ISM had to be used. Matrix-matched calibration would not
organic carboxylic acids (fatty or amino acids) are neutral or be an option, since matrix composition of similar beer types
positively charged at the BGE pH of 2.8, giving rise to a high can still be different, and blank samples of the same beer are
matrix tolerance of the method. Due to the Q-TOF’s high most likely not available. Quantitative precision for GLP in
mass accuracy, only two potentially interfering matrix com- beer matrix was high even when the concentration was close
ponents both with m/z 167.996 were observed migrating close to the LOD. The RSD of concentrations determined via ISM
to GLP at 7 min (Fig. 3c). With standard volume injection, ranged between 4 and 11% (n = 3). Comparing ISM and stan-
resolution between GLP and the closer co-migrating near iso- dard addition (10, 20, and 30 μg/L GLP were spiked to the
baric compound in different beer beverages was 2.4–2.6, so sample), the calculated GLP contaminations of Hasseröder
that no interference is expected, while for LVI with 100 mbar Premium Pils were 14.4 and 13.2 μg/L, respectively. Hence,
for 40 s, they were just baseline separated. Theoretical plate both quantification methods showed good accuracy for GLP
numbers are 20,000–35,000 for standard injection (depending analysis. As ISM is more efficient for larger sample numbers
on the matrix load) and 17,000 for LVI. Since AMPA is neu- and accounts for temporal changes in ESI conditions, we
tral and migrates close to the EOF, distinct matrix effects were chose ISM using GLP171 for further analysis.
observed (see Quantitative aspects, linear range, and limits of The linear range for aqueous samples, Daphnia medium,
detection section). and beer matrix (Fidelio) was determined in triplicate for GLP
and AMPA via a calibration curve with standard injection. In
Long-term stability and precision general, GLP contamination of beer was below 40 μg/L (see
Application to beer samples section); AMPA was never de-
Using a single PVA capillary, we were able to perform more tected. Calibration curves for GLP and AMPA showed good
than 200 measurements of standards and matrix-loaded sam- linearity within the calibrated range; however, distinct matrix
ples without loss in separation performance and peak shape. effects were observed in beer samples, especially for AMPA:
Precision for standards was calculated for peak area to be 2– the higher LOD for AMPA in beer samples compared with
10% RSD and for migration time 0.3–0.8% RSD (n = 9). aqueous standards (30.6 vs. 3.3 μg/L, respectively) is due to
Intermediate precision (5 days) for migration time and peak AMPA being neutral at the pH chosen, so that co-migration
area was 6.4% and 17.5%, respectively (n = 36). This con- with neutral matrix components and thus ion suppression oc-
firms the suitability of neutrally coated PVA capillaries for curs, which reduces the sensitivity to below 9% compared
GLP analysis. with aqueous samples. Ionic matrix components proved to
With regard to precision analyzing real samples, absolute be advantageous for GLP preconcentration as they acted as
changes in migration times (see Fig. 3) were due to sITP transient leaders in sITP, which could still be well separated in
phenomena, differences in applied pressure during separation LVI. In contrast, the LOD for AMPA in beer matrix was not
(30 vs. 50 mbar), and the use of the voltage ramp in the case of significantly improved using LVI; however, the signal shape
beer samples. Migration times for a specific sample were improved compared with the aqueous standard measurements.
highly reproducible with < 0.7% RSD (for the most complex We assume that this is due to sITP preconcentration occurring
beer samples even < 0.3% RSD). Signal shape was fair, while in the sample plug having a pH high enough to have AMPA
at high concentrations of GLP, a slight tailing was observed. charged. All validation parameters are summarized in Table 1.
RSDs of signal areas of GLP and AMPA in aqueous samples
were in the range of 1–11%, but usually below 10% RSD, in Application to beer samples
beer matrix in the range 1–9%, while for AMPA, sometimes
higher values, but below 15% RSD, were observed (see The developed method was applied to the analysis of 14 beer
Table S1 in ESM). beverages using ISM. Prior to analysis, all samples were
 Wimmer B. et al.

Table 1 Linear range (n = 3) (correlation coefficient R2 with error weight 1/x2) and LODs for GLP and AMPA in aqueous samples, Daphnia medium
(see Samples and sample preparation section), and organic beer sample (Fidelio). If large volume injection (LVI) was tested, values are given in brackets

Sample GLP AMPA

Linear range R2 Matrix Accuracy LOD Linear range R2 Matrix Accuracy LOD
(μg/L) effect (μg/L) (μg/L) effect (μg/L)

Aqueous 5–3000 0.9996 95.5–105.0% 4.2 (1.4) 10–3000 0.9957 87.2–99.5% 3.3
standard
Daphnia 25–200 0.9967 – – 4.7 25–200 0.9938 – – 6.2
medium
Beer beverage 10–3000 0.9912 93% 94.3–110.7% 5.3 (2.1) 50–3000 0.9723 8.7% 80.2–100.4% 30.6

degassed by sonication for 15 min to avoid CO2 release inside applicability of this strategy to other environmentally relevant
the capillary. The pH of the degassed samples was around 4. strong acids using the method optimized for glyphosate: N-
All beer samples were spiked with 40 μg/L of 13C2-15N- nitroso glyphosate (NNG) is a byproduct of GLP synthesis
glyphosate (GLP171) solution. If the detected GLP concen- [92]. N-Acetyl glyphosate (NAG) and N-acetyl AMPA
tration was close to or below the LOD, LVI was used. (NAA) are relevant GLP metabolites in modified crops con-
Quantitative results are summarized in Table 2; a representa- taining gat gene [93]. Glufosinate (GLU) is a herbicide struc-
tive electropherogram is shown in Fig. 3c. turally related to GLP; a degradation product of GLU is
Six out of 14 beer samples were tested positive for GLP 3-(methylphosphinico)propionic acid (3-MPPA) [94].
with contaminations ranging from 3.4 to 35 μg/L. In Southern MCPA (2-methyl-4-chlorophenoxyacetic acid) is a herbicide
German beers, the concentration of GLP was below the LOD. with converse properties to GLP; it poorly adsorbs to soil
AMPA was not detected in any of the investigated samples. components and thus is highly mobile in the environment
This is possibly due to the high load of neutral matrix compo- [95]. Trifluoroacetic acid (TFA) is a persistent, highly water
nents, possibly quenching AMPA signals and deteriorating soluble industrial chemical, which is also derived by
the LOD to around 30 μg/L. (photo)degradation from refrigerants, (per)fluorinated
chemicals, pharmaceuticals, and pesticides [96].
Applicability of the method to other anionic contaminants Difluoroacetic acid (DFA) is probably mostly derived from
TFA and has not yet been considered in the environment
The high selectivity of a separation BGE of low pH for strong before [97]. Oxamic acid is an oxidation product formed upon
acids has been shown for other applications such as glucosin- ozonation during drinking water treatment, most likely de-
olates in A. thaliana seeds [90], and sulfates, sulfonates, and rived from N-containing organic matter [98]. Iminodiacetic
phosphates in urine [91]. We here want to show the principal acid (IDA) is a chelating agent (e.g., immobilized for ion

Table 2 GLP residues

determined with CE-MS via ISM Beer beverage [GLP] (μg/L) RSD (%) Latitude (North)
with RSD values of quantified
concentration (n = 3) from differ- Astra Rotlicht 35.1 ± 3.1 8.7 53.0°
ent beers brewed in Germany and Tuborg Pilsner 33.9 ± 2.9 8.5 53.0°
geological latitude of the brewery. Hasseröder Premium Pils 14.1 ± 1.4 10.2 51.8°
Beers were spiked with 40 μg/L
of GLP171. n.d., not detected Wicküler Pilsner 13.6 ± 0.8 6.1 51.5°
Jever Fun (alcohol-free) 12.7 ± 1.4 10.9 53.6°
Carlsberg Pilsner 3.4 ± 0.1 4.2 53.0°
Alpirsbacher Spezial n.d. – 48.4°
Fidelio (organic) n.d. – 47.8°
Kloster Landbier n.d. – 48.5°
Gold Ochsen Original n.d. – 48.4
Riegeler Landbier n.d. – 48.0°
Rothaus Tannenzäpfle n.d. – 47.8°
Stuttgarter Hofbräu Bügel Spezial n.d. – 48.8°
Wulle Vollbier hell n.d. – 48.8°
Capillary electrophoresis-mass spectrometry for the direct analysis of glyphosate: method development and...

exchange resin) used for medical diagnosis [99], and structur- and reverse polarity. Under these conditions, only strongly
ally related to GLP, but its phosphonic acid moiety is replaced acidic organic compounds and mineral acids have sufficiently
by a carboxylic acid moiety. Phosphonic acid acts as a fungi- high electrophoretic mobilities to reach the MS. However, this
cide, e.g., to control downy mildew in vineyards [100]. advantage was on the cost of limits of detection for AMPA,
Most of the analytes were baseline separated, as indicated which can only be quantified at concentration above
by resolution for two successively migrating compounds (see 30.6 μg/L in beer matrix using an isotopically labeled internal
Table 3). Electropherograms of the analytes are shown in standard. This is due to its co-migration with neutral matrix
Fig. 4. TFA and DFA are badly separated, probably due to components. Mostly, neutral to alkaline pH BGEs were used
similar hydrodynamic radii. The pairs of NNG/oxamic acid by other groups [33–35, 39, 42, 43, 62]; only Safarpour and
and NAG/NAA are co-migrating. For all compounds, the Asiaie [40], and Iwamuro [41] used acidic BGEs at pH 2.5 and
LOD in aqueous solution was below 10 μg/L (based on cal- 3.4 for GLP analysis, respectively. Safarpour and Asiaie [40]
culation of SNR at 250 μg/L), except for oxamic acid and analyzed GLP contaminations in dried granule formulations
TFA. The relatively high LOD for TFA (especially compared within only 5 min separation time by CE-MS. Matrix effects
with DFA) is due to high noise, caused by the used fluorinated due to anionic surfactants and salts possibly present in the
mass calibrant for online m/z calibration. In the case of beer formulations were not discussed. Iwamuro et al. [41] mixed
matrix, the LODs deteriorated with increasing migration time green tea with a GLP formulation to have a final concentration
due to co-migration of weak organic acids originating from of 820 mg/L GLP; no interferences by matrix components
the beer matrix (this is the case for IDA, MCPA, GLU, and were observed at this high concentration and due to the fact
AMPA), or due to co-migration of phosphate and phospho- that anions were separated by CE but detected in the positive
nate derivatives, resulting in analyte signal quenching in case ionization mode by ESI-MS. However, the LODs stated
of NAA. IDA, MCPA, GLU, and AMPA are co-migrating would not be sufficient for GLP quantification in beer and
with matrix constituents of low or no charge and suffer from food samples. Goodwin et al. [39] used a capillary coated with
quenching by having 6–56% of signal intensity compared linear polyacrylamide and a BGE of 1 mmol/L acetic acid in a
with aqueous standard. water:methanol (50:50, v/v) mixture. With an injection vol-
ume of 70 nL, LODs of 169 μg/L for aqueous standards but
422 μg/L in wheat extracts were achieved using a sheathless
Discussion CE-MS interface. In part, the relatively high detection limits
compared with our study were due to poor peak shapes from
The properties of GLP are unfavorable for direct LC analysis. the low ionic strength of the BGE and the high pressure of
In contrast, its high charge over nearly the whole pH range 2 psi (138 mbar) applied during separation, which was re-
(with an anionic charge down to pH 2.3) allowed us to devel- quired by the sheathless ESI interface and absence of EOF.
op a direct CE-MS method without derivatization. The high We achieved LODs for GLP in aqueous samples and beer
selectivity of our CE separation was due to a BGE of pH 2.8 beverages of 1.4 and 2.1 μg/L, respectively. Of all

Table 3 Different anionic

analytes listed in their order of Analyte Resolution LOD (μg/L) RSD signal area (%) Matrix effect (%)
migration in organic beer matrix
(Fidelio). Resolution of Water Beer Water Beer Water Beer
neighboring analytes, estimated
LODs based on SNR, RSD of 1 TFA 0.8 0.6 66.0 53.3 7.4 2.6 112.7
signal area (n = 3), and matrix 2 DFA 6.4 4.9 2.8 2.7 8.3 0.4 102.0
effects (expressed as ratio
3 Phosphonic acid 3.1 2.4 30.4 28.5 10.6 0.5 93.7
between (signal area in matrix)/
(signal area in aqueous solution) × 4 NNG 0.0 0.2 2.9 3.7 11.5 2.8 115.3
100%) were obtained from spiked 5 Oxamic acid 4.4 6.3 53.7 90.7 9.3 15.4 58.6
concentrations of 250 μg/L. 6 NAG 0.1 0.1 1.8 4.3 11.5 3.6 46.2
Experimental parameters as in
7 NAA 8.3 6.9 2.7 15.3 7.8 1.7 32.0
Fig. 4
8 GLP 6.1 6.8 2.7 6.2 6.1 3.5 84.5
9 3-MPPA 4.5 5.5 1.6 3.3 3.8 2.0 89.4
10 IDA 0.8 0.5 20.5 40.8 7.1 7.4 41.6
11 MCPA 6.5 8.9 9.2 14.3 7.1 3.6 55.8
12 GLU 2.2 2.9 6.8 39.9 3.1 1.7 23.8
13 AMPA 3.3 24.4 3.4 2.1 6.4
 Wimmer B. et al.

Fig. 4 Different pesticides and environmentally relevant pollutants (13) AMPA. Intensities of TFA and DFA (1, 2) are divided by factor 5;
spiked at a concentration of 250 μg/L to an organic beer sample and of phosphonic acid, oxamic acid, IDA, MCPA, GLU and AMPA (3,
(Fidelio). Total ion current and highlighted signals of sulfate and phos- 5, 10–13) are multiplied by factor 3 to fit the axis range. Separation was
phate (left axis). Electropherograms of 13 different analytes (left axis): (1) conducted on a PVA-coated capillary with a total length of 60 cm, all
TFA, (2) DFA, (3) phosphonic acid, (4) NNG, (5) oxamic acid, (6) NAG, other parameters as in Fig. 3c
(7) NAA, (8) GLP, (9) 3-MPPA, (10) IDA, (11) MCPA, (12) GLU, and

derivatization-free CE methods surveyed [11], only Safarpour derivatization-free LC-MS/MS method, they achieved LOQs
and Asiaie (CE-MS) [40], See et al. (CE-C4D) [34], and of 10 μg/kg together with a high accuracy and recovery; how-
Horčičiak et al. (chip-based column coupling of ITP/CE- ever, they obtained a base width of the GLP signal in standard
C4D) [31] achieved comparable or better LODs down to solutions of almost 1 min at a retention time of 6 min.
0.1 μg/L for aqueous samples. However, the matrix tolerance The influence of a high concentration (up to 1 mmol/L) of
and selectivity were low using conductivity detection [31, 34] divalent inorganic cations (such as Ca2+, Mg2+, Pb2+, see
and preconcentration methods such as large volume sample Samples and sample preparation and Influence of divalent
stacking or field-enhanced/amplified sample injection [34]. cations on the separation of GLP sections) on the CE separa-
Besides sample preparation strategies, further improvements tion and MS detection proved negligible. However, trivalent
of LODs may be possible using different CE-MS interfaces. Al3+ and Fe3+, present, e.g., in soil samples, evoked migration
However, when using conditions of very low EOF, sheathless time shifts. Adding Na2HPO4 to the sample reduced this ma-
ESI-MS approaches such as presented by Goodwin et al. [39] trix effect. Together with the use of an isotopically labeled
for GLP analysis or the commercialized sheathless interface internal standard, identification and quantification of GLP is
originally presented by Cao and Moini [101] are problematic possible, especially with regard to the low concentrations of
as the low volume flow from the capillary has to be increased these metal cations normally present in (food) samples (e.g.,
using pressure. In addition, the choice of the BGE is more iron in dried beans and peas < 1 mmol/kg [106]). In contrast,
limited as the requirements for MS compatibility are even for derivatization and chromatographic separation, divalent
higher than for sheath liquid interfaces. A compromise may cations, especially Ca2+ and Mg2+, were more problematic
be the sheath flow interface using electrokinetic pumping [20–22], as they occur in larger concentrations in many
[102] or nanoESI [103]; however, its applicability to GLP still samples.
has to be shown. Our new CE method clearly has a simple instrumental set-
Comparing our method to LC-MS-based approaches, ma- up compared with IC-MS, where an eluent generator and a
trix effects were low despite the negligible sample preparation suppressor become necessary. Similar LODs of 2.5 μg/L (IC-
with degassing only. Anastassiades et al. [18] tested five dif- MS) [27] and < 2 μg/L (CE-MS) in samples were obtained.
ferent LC columns (hypercarb, anion exchange, and HILIC) Matrix effects in CE-MS were low given the high separation
after matrix-adapted extraction with acidified methanol selectivity. For IC, a two-dimensional method was discussed
(“QuPPe”) from different commodities (among them grapes, to reduce matrix effects, however, at the cost of analysis times
barley, lentils, cucumber). LOQs of GLP in apple, barley, [28].
cucumber, and grape samples were 20 μg/kg. An application For our study, it is interesting to note that GLP was only
note by Sciex [104] demonstrated GLP analysis using a detected in German beer samples, where the brewery is locat-
mixed-mode LC column with QTrap-MS using multiple reac- ed at a geodetic latitude > 50° North. Assuming that the brew-
tion monitoring with retention times of underivatized AMPA ery predominantly uses local barley sources, the reasons may
and GLP of 1.2 and 2.2 min and LODs of 0.2 μg/L in beer in part be historic: succession rules favored large farms in the
beverages. However, the mass spectra clearly showed the lim- Northern regions of Germany. With the foundation of
ited selectivity, as GLP is poorly separated from several ma- Agricultural Production Cooperatives (Landwirtschaftliche
trix components. As in our study, AMPA was not detected in Produktionsgenossenschaft), the German Democratic
beer. Nagatomi et al. [105] analyzed GLP and AMPA in beer Republic further fostered large agricultural units with accord-
beverages and barley extracts after enrichment and purifica- ingly large fields [107], for which non-tillage agricultural
tion using anion exchange solid phase extraction. With this management (using GLP) may be advantageous [108]
Capillary electrophoresis-mass spectrometry for the direct analysis of glyphosate: method development and...

compared with the rather small agricultural units and fields in sample plug. We demonstrated that the method is also appli-
the South of Germany, where plowing dominates [2]. In ad- cable to other strong acids relevant in environmental samples
dition, climate conditions favor GLP application in the more such as TFA, DFA, MCPA, GLU, 3-MPPA, IDA, phosphonic
humid conditions present in the coastal areas, e.g., for desic- acid, oxamic acid, and derivatives of GLP and AMPA. Hence,
cation [2]. this method initially developed for GLP screening is also ap-
The CE-MS method was not only applied for samples con- plicable for strong organic acids, especially in samples with
taining high salt and matrix loads but also for a variety of other moderate phosphate loads. Overall, the detection limits
strong acids of environmental relevance or concern. The reached are well suited to match the maximum contaminant
PVA-coated capillary reduced sorption especially of phospho- limits of GLP in water in the USA and partly for foodstuff in
rous containing analytes, which are prone to interact with the the EU. Future investigations will address GLP and AMPA
capillary surface similar to glyphosate. For most analytes, the extraction and quantification in different food and environ-
LODs in aqueous media and beer matrix were in the lower mental samples.
μg/L range, except for phosphonic acid and oxamic acid,
which suffer signal intensities by adduct formation with sodi- Acknowledgements Additionally, we thank U. Wehpke from Krefeld for
delicate ESI tip modifications, D. Werner and R. Triebskorn from the
um formate. In case of TFA, tremendous background noise
Institution of Evolution and Ecology, University Tübingen, for samples
deteriorated the LOD, which was not the case for DFA which of Daphnia magna medium, and B. Bugsel and K. Röhler from the
has a roughly 20 times lower LOD. Severe matrix effects of Center of Applied Geoscience, University Tübingen, for conducting
analytes with low electrophoretic mobilities were observed; glyphosate sorption experiments.
diluting or extracting the matrix via SPE can be an option to
Funding information Open Access funding provided by Projekt DEAL.
enhance signal intensities and thereby LODs; however, good
This work was supported by the Collaborative Research Center 1253
recovery rates have to be ensured. The separation of the strong CAMPOS (Project 4: Floodplain Biogeochemistry), funded by the
acids DFA and TFA was not sufficient. Just recently, Höcker German Research Foundation (DFG, Grant Agreement SFB
et al. [103] demonstrated the separation of strong acids (e.g., 1253/12017). C.H. thanks for the support from the Excellence
Initiative, a jointly funded program of the German Federal and State
chlorinated and brominated acetic acids) using a new CE-
governments, organized by the German Research Foundation (DFG).
nanoESI-MS method with LOQs below 0.5 μg/L. For a broad
screening of strong organic acids, further BGE modifications
Compliance with ethical standards
may be necessary, adapting pH and possibly the addition of
organic solvents. Conflict of interest The authors declare that they have no conflicts of
interest.

Open Access This article is licensed under a Creative Commons

Conclusion Attribution 4.0 International License, which permits use, sharing,
adaptation, distribution and reproduction in any medium or format, as
A rapid, derivatization-free CE-MS method for GLP analysis long as you give appropriate credit to the original author(s) and the
in complex matrices was developed for different applications. source, provide a link to the Creative Commons licence, and indicate if
changes were made. The images or other third party material in this article
The use of a PVA-coated capillary enabled selective GLP
are included in the article's Creative Commons licence, unless indicated
analysis with reduced sorption to the capillary wall and high otherwise in a credit line to the material. If material is not included in the
separation efficiency. The use of volatile BGE components article's Creative Commons licence and your intended use is not
(formic acid and ammonia) enabled sensitive MS detection permitted by statutory regulation or exceeds the permitted use, you will
need to obtain permission directly from the copyright holder. To view a
with LODs in the low μg/L range for GLP and AMPA. For
copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
GLP, the combination of an acidic BGE and negative ESI
polarity provided high selectivity and high matrix tolerance.
Divalent metal cations did not disturb the analysis, which is in
contrast to LC methods with GLP derivatization. The combi-
nation of large volume injection with sample-induced ITP References
(natural phosphate in beer as transient leader) provided a sim-
ple and rapid online preconcentration method taking advan- 1. PR Newswire. Research Report on China's Glyphosate Industry
tage from the high salt loads of several sample types. LODs (2018-2022). 2018. Available from: https://www.prnewswire.
below 2 μg/L and an accuracy of 94.3–110.7% in beer were com/news-releases/research-report-on-chinas-glyphosate-
industry-2018-2022-300689149.html. Accessed 05 June 2020.
reached. Beer samples were analyzed directly after degassing.
2. Schmitz PMG, H. Die ökonomische Bedeutung des Wirkstoffes
No further sample preparation was necessary for precise GLP Glyphosat für den Ackerbau in Deutschland. J Kult. 2012;64(5):
quantification using isotopically labeled standards. AMPA 150–62.
analysis is possible with this method with some restrictions 3. Benbrook CM. Trends in glyphosate herbicide use in the United
as the compound was only transported by EOF within the States and globally. Environ Sci Eur. 2016;28(1):3.
 Wimmer B. et al.

4. McKnight US, Rasmussen JJ, Kronvang B, Binning PJ, Bjerg PL. and electrospray ionization tandem mass spectrometry detection
Sources, occurrence and predicted aquatic impact of legacy and of glyphosate, glufosinate, aminomethylphosphonic acid and 2-
contemporary pesticides in streams. Environ Pollut. 2015;200:64– aminoethylphosphonic acid in water. J Chromatogr A. 2016;1475:
76. 64–73.
5. Borggaard OK, Gimsing AL. Fate of glyphosate in soil and the 23. Steinborn A, Alder L, Michalski B, Zomer P, Bendig P, Martinez
possibility of leaching to ground and surface waters: a review. Pest SA, et al. Determination of glyphosate levels in breast milk sam-
Manag Sci. 2008;64(4):441–56. ples from Germany by LC-MS/MS and GC-MS/MS. J Agric Food
6. Saunders LE, Pezeshki R. Glyphosate in runoff waters and in the Chem. 2016;64(6):1414–21.
root-zone: a review. Toxics. 2015;3(4):462–80. 24. Botero-Coy AM, Ibanez M, Sancho JV, Hernandez F.
7. Vereecken H. Mobility and leaching of glyphosate: a review. Pest Improvements in the analytical methodology for the residue de-
Manag Sci. 2005;61(12):1139–51. termination of the herbicide glyphosate in soils by liquid chroma-
8. Grandcoin A, Piel S, Baures E. Aminomethylphosphonic acid tography coupled to mass spectrometry. J Chromatogr A.
(AMPA) in natural waters: its sources, behavior and environmen- 2013;1292:132–41.
tal fate. Water Res. 2017;117:187–97. 25. Hanke I, Singer H, Hollender J. Ultratrace-level determination of
9. Stephenson CL, Harris CA. An assessment of dietary exposure to glyphosate, aminomethylphosphonic acid and glufosinate in nat-
glyphosate using refined deterministic and probabilistic methods. ural waters by solid-phase extraction followed by liquid
Food Chem Toxicol. 2016;95:28–41. chromatography-tandem mass spectrometry: performance tuning
10. Hao C, Morse D, Morra F, Zhao X, Yang P, Nunn B. Direct of derivatization, enrichment and detection. Anal Bioanal Chem.
aqueous determination of glyphosate and related compounds by 2008;391(6):2265–76.
liquid chromatography/tandem mass spectrometry using reversed- 26. Goscinny S, Unterluggauer H, Aldrian J, Hanot V, Masselter S.
phase and weak anion-exchange mixed-mode column. J Determination of glyphosate and its metabolite AMPA
Chromatogr A. 2011;1218(33):5638–43. (aminomethylphosphonic acid) in cereals after derivatization by
11. Gauglitz G, Wimmer B, Melzer T, Huhn C. Glyphosate analysis isotope dilution and UPLC-MS/MS. Food Anal Methods.
using sensors and electromigration separation techniques as alter- 2012;5(5):1177–85.
natives to gas or liquid chromatography. Anal Bioanal Chem. 27. Boušová K, Bruggink C, Godula M. Fast routine analysis of polar
2018;410(3):725–46. pesticides in foods by suppressed ion chromatography and mass
12. Sprankle P, Meggitt WF, Penner D. Adsorption, mobility, and spectrometry. Thermo Fisher Scientific; 2016. Available from:
microbial degradation of glyphosate in the soil. Weed Sci. https://assets.thermofisher.com/TFS-Assets/CMD/Application-
2017;23(3):229–34. Notes/AN-661-IC-MS-Polar-Pesticides-Foods-AN64868-EN.
13. Wauchope D. Acid dissociation-constants of arsenic acid, pdf. Accessed 05 Jun 2020.
methylarsonic acid (MAA), dimethylarsinic acid (cacodylic acid), 28. Yang C, Henday S, Wang L, Schnute B. Analysis of glyphosate
and N-(phosphonomethyl)glycine (glyphosate). J Agric Food and AMPA in environmental water by ion chromatography
Chem. 1976;24(4):717–21. electrospray tandem mass spectrometry (IC-ESI-MS/MS).
14. Comer J, Chamberlain K, Evans A. Validation of pH-metric tech- Thermo Fisher Scientific; 2010. Available from: https://assets.
nique for measurement of pKa and log Pow of ionizable herbicides. thermofisher.com/TFS-Assets/CMD/Application-Notes/AN-491-
SAR QSAR Environ Res. 1995;3(4):307–13. Glyphosate-AMPA-IC-ESI-MS-AN-63068.pdf. Accessed 05
15. Kortüm G, Vogel W, Andrussow K. Dissociation constants of Jun 2020.
organic acids in aqueous solution. Pure Appl Chem. 1960;1(2– 29. Cikalo MG, Goodall DM, Matthews W. Analysis of glyphosate
3):187–536. using capillary electrophoresis with indirect detection. J
16. Koskinen WC, Marek LJ, Hall KE. Analysis of glyphosate and Chromatogr A. 1996;745(1–2):189–200.
aminomethylphosphonic acid in water, plant materials and soil. 30. Rojano-Delgado AM, Ruiz-Jimenez J, de Castro MD, De Prado
Pest Manag Sci. 2016;72(3):423–32. R. Determination of glyphosate and its metabolites in plant mate-
17. Ding J, Guo H, Liu W-W, Zhang W-W, Wang J-W. Current rial by reversed-polarity CE with indirect absorptiometric detec-
progress on the detection of glyphosate in environmental samples. tion. Electrophoresis. 2010;31(8):1423–30.
J Sci Appl Biomed. 2015;3:88–95. 31. Horciciak M, Masar M, Bodor R, Danc L, Bel P. Trace analysis of
18. Anastassiades M, Wachtler AK, Kolberg DI, Eichhorn E, glyphosate in water by capillary electrophoresis on a chip with
Benkenstein A, Zechmann S, et al. Quick method for the analysis high sample volume loadability. J Sep Sci. 2012;35(5–6):674–80.
of numerous highly polar pesticides in food involving extraction 32. Corbera M, Hidalgo A, Salvado V, Wieczorek PP. Determination
with acidified methanol and LC-MS/MS measurement. EURL- of glyphosate and aminomethylphosphonic acid in natural water
SRM; 2019. Available from: https://www.eurl-pesticides.eu/ using the capillary electrophoresis combined with enrichment
docs/public/tmplt_article.asp?CntID=887&LabID=200&Lang= step. Anal Chim Acta. 2005;540(1):3–7.
EN. Accessed 02 Dec 2019. 33. Chang SY, Liao CH. Analysis of glyphosate, glufosinate and
19. Arkan T, Molnar-Perl I. The role of derivatization techniques in aminomethylphosphonic acid by capillary electrophoresis with
the analysis of glyphosate and aminomethyl-phosphonic acid by indirect fluorescence detection. J Chromatogr A. 2002;959(1–2):
chromatography. Microchem J. 2015;121:99–106. 309–15.
20. DIN. Water quality - Determination of glyphosate and AMPA - 34. See HH, Hauser PC, Ibrahim WA, Sanagi MM. Rapid and direct
method using high performance liquid chromatography (HPLC) determination of glyphosate, glufosinate, and aminophosphonic
with tandem mass spectrometric detection (ISO 16308:2014). acid by online preconcentration CE with contactless conductivity
Berlin: Beuth, DIN, Committee WPS; 2017 2017-09-00. detection. Electrophoresis. 2010;31(3):575–82.
Available from: https://doi.org/10.31030/2679839. 35. See HH, Hauser PC, Sanagi MM, Ibrahim WA. Dynamic support-
21. Freuze I, Jadas-Hecart A, Royer A, Communal PY. Influence of ed liquid membrane tip extraction of glyphosate and
complexation phenomena with multivalent cations on the analysis aminomethylphosphonic acid followed by capillary electrophore-
of glyphosate and aminomethyl phosphonic acid in water. J sis with contactless conductivity detection. J Chromatogr A.
Chromatogr A. 2007;1175(2):197–206. 2010;1217(37):5832–8.
22. Skeff W, Recknagel C, Schulz-Bull DE. The influence of salt 36. Goodwin L, Hanna M, Startin JR, Keely BJ, Goodall DM.
matrices on the reversed-phase liquid chromatography behavior Isotachophoretic separation of glyphosate, glufosinate, AMPA
Capillary electrophoresis-mass spectrometry for the direct analysis of glyphosate: method development and...

and MPP with contactless conductivity detection. Analyst. 52. Wei X, Gao X, Zhao L, Peng X, Zhou L, Wang J, et al. Fast and
2002;127(2):204–6. interference-free determination of glyphosate and glufosinate res-
37. da Silva ER, Segato TP, Coltro WK, Lima RS, Carrilho E, Mazo idues through electrophoresis in disposable microfluidic chips. J
LH. Determination of glyphosate and AMPA on polyester-toner Chromatogr A. 2013;1281:148–54.
electrophoresis microchip with contactless conductivity detection. 53. Lanaro R, Costa JL, Cazenave SO, Zanolli-Filho LA, Tavares MF,
Electrophoresis. 2013;34(14):2107–11. Chasin AA. Determination of herbicides paraquat, glyphosate,
38. Wuilloud RG, Shah M, Kannamkumarath SS, Altamirano JC. The and aminomethylphosphonic acid in marijuana samples by capil-
potential of inductively coupled plasma-mass spectrometric detec- lary electrophoresis. J Forensic Sci. 2015;60(s1):S241–7.
tion for capillary electrophoretic analysis of pesticides. 54. Molina M, Silva M. Simultaneous determination of phosphorus-
Electrophoresis. 2005;26(7–8):1598–605. containing amino acid-herbicides by nonionic surfactant micellar
39. Goodwin L, Startin JR, Keely BJ, Goodall DM. Analysis of electrokinetic chromatography with laser-induced fluorescence
glyphosate and glufosinate by capillary electrophoresis-mass detection. Electrophoresis. 2001;22(6):1175–81.
spectrometry utilising a sheathless microelectrospray interface. J 55. Cao L, Deng T, Liang S, Tan X, Meng J. Determination of herbi-
Chromatogr A. 2003;1004(1–2):107–19. cides and its metabolite in soil and water samples by capillary
40. Safarpour H, Asiaie R. Determination of glyphosate as cross- electrophoresis-laser induced fluorescence detection using
contaminant in a commercial herbicide by capillary microwave-assisted derivatization. Anal Sci. 2014;30(7):759–66.
electrophoresis-electrospray ionization-mass spectrometry. 56. Chang SY, Wei MY. Simultaneous determination of glyphosate,
Electrophoresis. 2005;26(7–8):1562–6. glufosinate, and aminomethylphosphonic acid by capillary elec-
41. Iwamuro Y, Iio-Ishimaru R, Chinaka S, Takayama N, Kodama S, trophoresis after 9-fluorenylmethyl chloroformate derivatization. J
Hayakawa K. Analysis of phosphorus-containing amino acid-type Chin Chem Soc-Taip. 2005;52(4):785–92.
herbicides by capillary electrophoresis/mass spectrometry using a 57. Gotti R, Fiori J, Bosi S, Dinelli G. Field-amplified sample injec-
chemically modified capillary having amino groups. J Health Sci. tion and sweeping micellar electrokinetic chromatography in anal-
2010;56(5):606–12. ysis of glyphosate and aminomethylphosphonic acid in wheat. J
42. Kawai M, Iwamuro Y, Iio-Ishimaru R, Chinaka S, Takayama N, Chromatogr A. 1601;2019:357–64.
Hayakawa K. Analysis of phosphorus-containing amino acid-type 58. Molina M, Silva M. In-capillary derivatization and analysis of
herbicides by sheathless capillary electrophoresis/electrospray amino acids, amino phosphonic acid-herbicides and biogenic
ionization-mass spectrometry using a high sensitivity porous amines by capillary electrophoresis with laser-induced fluores-
sprayer. Anal Sci. 2011;27(8):857–60. cence detection. Electrophoresis. 2002;23(14):2333–40.
43. Vidal DTR, do Lago CL, Daniel D. Development of an analytical 59. Yang G, Xu X, Shen M, Wang W, Xu L, Chen G, et al.
method for the determination of glufosinate, glyphosate and Determination of organophosphorus pesticides by capillary
AMPA in soybean milk by CE-MS/MS. Agilent Technologies; electrophoresis-inductively coupled plasma mass spectrometry
2003. Available from: https://www.agilent.com/cs/library/ with collective sample-introduction technique. Electrophoresis.
applications/5991-2888EN.pdf. Accessed 05 Jun 2020. 2009;30(10):1718–23.
44. Sung IH, Lee YW, Chung DS. Liquid extraction surface analysis 60. Amelin VG, Bol'shakov DS, Tretiakov AV. Determination of
in-line coupled with capillary electrophoresis for direct analysis of glyphosate and aminomethylphosphonic acid in surface water
a solid surface sample. Anal Chim Acta. 2014;838:45–50. and vegetable oil by capillary zone electrophoresis. J Anal
45. Molina M, Silva M. Analytical potential of fluorescein analogues Chem. 2012;67(4):386–91.
for ultrasensitive determinations of phosphorus-containing amino 61. Orejuela E, Silva M. Rapid and sensitive determination of
acid herbicides by micellar electrokinetic chromatography with phosphorus-containing amino acid herbicides in soil samples by
laser-induced fluorescence detection. Electrophoresis. capillary zone electrophoresis with diode laser-induced fluores-
2002;23(7–8):1096–103. cence detection. Electrophoresis. 2005;26(23):4478–85.
46. Navarrete-Casas M, Segura-Carretero A, Cruces-Blanco C, 62. Chiu HY, Lin ZY, Tu HL, Whang CW. Analysis of glyphosate
Fernandez-Gutierrez A. Potential determination of aminated pes- and aminomethylphosphonic acid by capillary electrophoresis
ticides and metabolites by cyclodextrin capillary electrophoresis- with electrochemiluminescence detection. J Chromatogr A.
laser-induced fluorescence using FITC as labelling. Pest Manag 2008;1177(1):195–8.
Sci. 2005;61(2):197–203. 63. Tomita M, Okuyama T, Nigo Y, Uno B, Kawai S. Determination
47. Hooijschuur EWJ, Kientz CE, Dijksman J, Brinkman UAT. of glyphosate and its metabolite, (aminomethyl)phosphonic acid,
Potential of microcolumn liquid chromatography and capillary in serum using capillary electrophoresis. J Chromatogr B.
electrophoresis with flame photometric detection for determina- 1991;571(1–2):324–30.
tion of polar phosphorus-containing pesticides. Chromatographia. 64. Pattky M, Huhn C. Advantages and limitations of a new cationic
2001;54(5–6):295–301. coating inducing a slow electroosmotic flow for CE-MS peptide
48. Khrolenko. Determination of glyphosate in water samples with the analysis: a comparative study with commercial coatings. Anal
combination of cation-exchange chromatography and capillary Bioanal Chem. 2013;405(1):225–37.
electrophoresis. Ars Separatoria Acta. 2003;2:56–63. 65. Gulu Zada L. Sorption of selected pesticides on mineral surfaces:
49. Dong YL, Guo DQ, Cui H, Li XJ, He YJ. Magnetic solid phase factors and mechanisms [dissertation]. Tübingen: Eberhard Karls
extraction of glyphosate and aminomethylphosphonic acid in river Universität Tübingen; 2019. Available from: https://doi.org/10.
water using Ti4+-immobilized Fe3O4 nanoparticles by capillary 15496/publikation-28103.
electrophoresis. Anal Methods Uk. 2015;7(14):5862–8. 66. Werner D. Akute Toxizität von Kalium-Glyphosat und der
50. Kodama S, Ito Y, Taga A, Nomurad Y, Yamamoto A, Chinaka S, Formulierung Roundup PowerFlex im Daphnien-
et al. A fast and simple analysis of glyphosate in tea beverages by Immobilisationstest [Bachelorthesis]. Tübingen: Eberhard Karls
capillary electrophoresis with on-line copper(II)-glyphosate com- Universität Tübingen; 2017. https://uni-tuebingen.de/en/
plex formation. J Health Sci. 2008;54(5):602–6. faculties/faculty-of-science/departments/biology/institutes/
51. You J, Kaljurand M, Koropchak JA. Direct determination of evolutionecology/groups/animal-physiological-ecology/phd-and-
glyphosate in environmental waters using capillary electrophore- exam-theses.html#c28699. Accessed 05 Jun 2020.
sis with electrospray condensation nucleation light scattering de- 67. DIN. Water quality - Determination of the inhibition of the mo-
tection. Int J Environ Anal Chem. 2003;83(9):797–806. bility of Daphnia magna Straus (Cladocera, Crustacea) - acute
 Wimmer B. et al.

toxicity test (ISO 6341:2012). Berlin: Beuth, Committee WPS; 84. Gebauer P, Krivankova L, Pantuckova P, Bocek P, Thormann W.
2013 2013-01-00. Available from: https://doi.org/10.31030/ Sample self-stacking in capillary zone electrophoresis: behavior of
2679839. samples containing multiple major coionic components.
68. Sheals J, Granström M, Sjöberg S, Persson P. Coadsorption of Electrophoresis. 2000;21(14):2797–808.
Cu(II) and glyphosate at the water–goethite (α-FeOOH) interface: 85. Beckers JL, Bocek P. Sample stacking in capillary zone electro-
molecular structures from FTIR and EXAFS measurements. J phoresis: principles, advantages and limitations. Electrophoresis.
Colloid Interface Sci. 2003;262(1):38–47. 2000;21(14):2747–67.
69. Jonsson CM, Persson P, Sjoberg S, Loring JS. Adsorption of 86. Styburski D, Janda K, Baranowska-Bosiacka I, Łukomska A, Dec
glyphosate on goethite (α-FeOOH): surface complexation model- K, Goschorska M, et al. Beer as a potential source of
ing combining spectroscopic and adsorption data. Environ Sci macroelements in a diet: the analysis of calcium, chlorine, potas-
Technol. 2008;42(7):2464–9. sium, and phosphorus content in a popular low-alcoholic drink.
70. Pessagno RC, Dos Santos AM, Torres Sanchez RM. N- Eur Food Res Technol. 2018;244(10):1853–60.
(Phosphonomethyl)glycine interactions with soils. An Asoc 87. Huhn C, Pyell U. Diffusion as major source of band broadening in
Quím Argent. 2005;93:97–108. field-amplified sample stacking under negligible electroosmotic
71. Ramstedt M, Norgren C, Sheals J, Shchukarev A, Sjöberg S. flow velocity conditions. J Chromatogr A. 2010;1217(26):4476–
Chemical speciation of N-(phosphonomethyl)glycine in solution 86.
and at mineral interfaces. Surf Interface Anal. 2004;36(8):1074–7. 88. Ibanez M, Pozo OJ, Sancho JV, Lopez FJ, Hernandez F. Re-
72. Zheng TT, Sun ZX, Yang XF, Holmgren A. Sorption of phos- evaluation of glyphosate determination in water by liquid chroma-
phate onto mesoporous γ-alumina studied with in-situ ATR-FTIR tography coupled to electrospray tandem mass spectrometry. J
spectroscopy. Chem Cent J. 2012;6(1):26. Chromatogr A. 2006;1134(1–2):51–5.
73. Moraes MP, Gonçalves LM, Pereira EA. Determination of glyph- 89. Cortacero-Ramirez S, Segura-Carretero A, Hernainz-Bermudez
osate and aminomethylphosphonic acid by capillary electrophore- de Castro M, Fernandez-Gutierrez A. Determination of low-
sis with indirect detection using pyridine-2,6-dicarboxylic acid or molecular-mass organic acids in any type of beer samples by
3,5-dinitrobenzoic acid. Int J Environ Anal Chem. 2018;98(3): coelectroosmotic capillary electrophoresis. J Chromatogr A.
258–70. 2005;1064(1):115–9.
74. Gimsing AL, Borggaard OK. Competitive adsorption and desorp- 90. Bringmann G, Kajahn I, Neususs C, Pelzing M, Laug S, Unger M,
tion of glyphosate and phosphate on clay silicates and oxides. Clay et al. Analysis of the glucosinolate pattern of Arabidopsis thaliana
Miner. 2002;37(3):509–15. seeds by capillary zone electrophoresis coupled to electrospray
75. Hajba L, Guttman A. Recent advances in column coatings for ionization-mass spectrometry. Electrophoresis. 2005;26(7–8):
capillary electrophoresis of proteins. TrAC Trends Anal Chem. 1513–22.
2017;90:38–44. 91. Bunz SC, Weinmann W, Neususs C. The selective determination
76. Soga T, Igarashi K, Ito C, Mizobuchi K, Zimmermann HP, Tomita of sulfates, sulfonates and phosphates in urine by CE-MS.
M. Metabolomic profiling of anionic metabolites by capillary Electrophoresis. 2010;31(7):1274–81.
electrophoresis mass spectrometry. Anal Chem. 2009;81(15): 92. Kim M, Stripeikis J, Inon F, Tudino M. A simplified approach to
6165–74. the determination of N-nitroso glyphosate in technical glyphosate
77. Foret F, Thompson TJ, Vouros P, Karger BL, Gebauer P, Bocek using HPLC with post-derivatization and colorimetric detection.
P. Liquid sheath effects on the separation of proteins in capillary Talanta. 2007;72(3):1054–8.
electrophoresis/electrospray mass spectrometry. Anal Chem. 93. Norskov NP, Jensen SK, Sorensen MT. Robust and highly sensi-
2002;66(24):4450–8. tive micro liquid chromatography-tandem mass spectrometry
78. Mala Z, Pantuckova P, Gebauer P, Bocek P. Advanced electrolyte method for analyses of polar pesticides (glyphosate,
tuning and selectivity enhancement for highly sensitive analysis of aminomethylphosfonic acid, N-acetyl glyphosate and N-acetyl
cations by capillary ITP-ESI MS. Electrophoresis. 2013;34(5): aminomethylphosfonic acid) in multiple biological matrices. J
777–84. Chromatogr A. 2019;1605:360343.
79. Posch TN, Müller A, Schulz W, Putz M, Huhn C. Implementation 94. Raina-Fulton R. A review of methods for the analysis of orphan
of a design of experiments to study the influence of the back- and difficult pesticides: glyphosate, glufosinate, quaternary am-
ground electrolyte on separation and detection in non-aqueous monium and phenoxy acid herbicides, and dithiocarbamate and
capillary electrophoresis-mass spectrometry. Electrophoresis. phthalimide fungicides. J AOAC Int. 2014;97(4):965–77.
2012;33(4):583–98. 95. Morton PA, Fennell C, Cassidy R, Doody D, Fenton O, Mellander
80. Klampfl CW, Himmelsbach M. Sheath liquids in CE-MS: role, PE, et al. A review of the pesticide MCPA in the land-water
parameters, and optimization. In: de Jong G, editor. Capillary environment and emerging research needs. WIREs Water. 2019.
electrophoresis–mass spectrometry (CE-MS): principles and ap- https://doi.org/10.1002/wat2.1402.
plications. Weinheim: Wiley-VCH; 2016. p. 41–65. 96. Solomon KR, Velders GJ, Wilson SR, Madronich S, Longstreth J,
81. Mallet CR, Lu Z, Mazzeo JR. A study of ion suppression effects in Aucamp PJ, et al. Sources, fates, toxicity, and risks of
electrospray ionization from mobile phase additives and solid- trifluoroacetic acid and its salts: relevance to substances regulated
phase extracts. Rapid Commun Mass Spectrom. 2004;18(1):49– under the Montreal and Kyoto protocols. J Toxicol Environ Health
58. B Crit Rev. 2016;19(7):289–304.
82. Melzer T, Wimmer B, Bock S, Posch TN, Huhn C. Challenges 97. Dorgerloh U, Becker R, Kaiser M. Evidence for the formation of
and applications of isotachophoresis coupled to mass spectrometry difluoroacetic acid in chlorofluorocarbon-contaminated ground
- a review. Electrophoresis. 2020. https://doi.org/10.1002/elps. water. Molecules. 2019. https://doi.org/10.3390/
201900454. molecules24061039.
83. Osbourn DM, Weiss DJ, Lunte CE. On-line preconcentration 98. Fleischer S, Weiss SC, Lucke T, Seitz W, Schulz W, Weber WH.
methods for capillary electrophoresis. Electrophoresis. Formation of oxamic acid during drinking water treatment. Ozone
2000;21(14):2768–79. Sci Eng. 2015;37(5):441–9.
Capillary electrophoresis-mass spectrometry for the direct analysis of glyphosate: method development and...

99. Roger J, Heeley T, Graham W, Walsh A. Laparoscopic cholecys- 104. Schreiber A, Jin W, Winkler P. LC-MS/MS analysis of emerging
tectomy for ultrasound normal gallbladders: should we forego food contaminants. Trace level detection of glyphosate in water
hepatobiliary iminodiacetic acid scans? Can J Rural Med. and beer samples. Ontario: Sciex; 2016. Available from: https://
2019;24(2):61–4. sciex.com/Documents/technotes/food-contaminants6500_beer.
100. Margarey PA, Wicks TJ, Wachtel MF. Phosphonic (phosphorous) pdf. Accessed 05 Jun 2020.
acid controls plasmopara viticola the cause of downy mildew of 105. Nagatomi Y, Yoshioka T, Yanagisawa M, Uyama A, Mochizuki
grapevines. Australas Plant Pathol. 1990;19(4):126–7. N. Simultaneous LC-MS/MS analysis of glyphosate, glufosinate,
101. Cao P, Moini M. A novel sheathless interface for capillary and their metabolic products in beer, barley tea, and their ingredi-
electrophoresis/electrospray ionization mass spectrometry using ents. Biosci Biotechnol Biochem. 2013;77(11):2218–21.
an in-capillary electrode. J Am Soc Mass Spectrom. 1997;8(5): 106. National Research Council. Drinking water and health, vol. 3.
561–4. Washington, DC: The National Academies Press; 1980. p. 415.
102. Schiavone NM, Sarver SA, Sun L, Wojcik R, Dovichi NJ. High ISBN: 978-0-309-02932-2
speed capillary zone electrophoresis-mass spectrometry via an 107. Zikeli S, Gruber S. Reduced tillage and no-till in organic farming
electrokinetically pumped sheath flow interface for rapid analysis systems, Germany-Status quo, potentials and challenges.
of amino acids and a protein digest. J Chromatogr B. 2015;991: Agriculture-Basel. 2017. https://doi.org/10.3390/
53–8. agriculture7040035.
103. Höcker O, Bader T, Schmidt TC, Schulz W, Neusüß C. 108. Steinmann HH, Dickeduisberg M, Theuvsen L. Uses and benefits of
Enrichment-free analysis of anionic micropollutants in the sub- glyphosate in German arable farming. Crop Prot. 2012;42:164–9.
ppb range in drinking water by capillary electrophoresis-high res-
olution mass spectrometry. Anal Bioanal Chem. 2020. https://doi. Publisher’s note Springer Nature remains neutral with regard to jurisdic-
org/10.1007/s00216-020-02525-8. tional claims in published maps and institutional affiliations.