Udt

LSa

b

c

a

ARRAA

KTBBECA

1

htbfdttmptitprp

h0

Journal of Pharmaceutical and Biomedical Analysis 100 (2014) 50–57

Contents lists available at ScienceDirect

Journal of Pharmaceutical and Biomedical Analysis

j o ur nal ho me page: www.elsev ier .com/ lo cate / jpba

HPLC-MS/MS method for the determination of the cyclicepsipeptide mycotoxins beauvericin and enniatins in in vitroransdermal experiments

ien Taeverniera, Lieselotte Verysera, Kirsten Vandercruyssena, Matthias D’Hondta,tijn Vansteelandtb, Sarah De Saegerc, Bart De Spiegeleera,∗

Drug Quality and Registration (DruQuaR) Group, Faculty of Pharmaceutical Sciences, Ghent University, Harelbekestraat 72, B-9000 Ghent, BelgiumDepartment Applied Mathematics, Computer Science and Statistics, Ghent University, Krijgslaan 281 S9, B-9000 Ghent, BelgiumDepartment of Bioanalysis, Laboratory of Food Analysis, Faculty of Pharmaceutical Sciences, Ghent University, Harelbekestraat 72, B-9000 Ghent, Belgium

r t i c l e i n f o

rticle history:eceived 11 June 2014eceived in revised form 16 July 2014ccepted 19 July 2014vailable online 29 July 2014

eywords:

a b s t r a c t

Currently, dermal exposure data of cyclic depsipeptide mycotoxins beauvericin and enniatins are com-pletely absent with a lack of local skin and systemic kinetics, despite their widespread skin contact andintrinsic hazard. Therefore a sensitive and specific bioanalytical high-throughput UHPLC-MS/MS methodwas developed for the quantitative and simultaneous determination of cyclic depsipeptide mycotoxinsbeauvericin and enniatins (A, A1, B, B1, D, E, C/F) in human skin Franz diffusion cell samples. The limitsof detection ranged between 10 and 17 pg/ml, while the total run time was only 4.5 min. There was no

ransdermal Franz diffusion cellioanalytical UHPLC-MS/MS methodeauvericinnniatinsyclic depsipeptide mycotoxinsdsorption

significant effect of endogenous skin compounds on the mycotoxin MS signal observed, and the accuracy(0.68–24.68% bias) and precision (0.57–10.70% RSD) were considered acceptable for our purposes. More-over, it was demonstrated that these cyclic depsipeptides are stable for at least 7 days when formulatedin different organic or aqueous mixtures. Finally, adsorption to glass did occur: at least 50% ethanol oracetonitrile is required to prevent significant adsorption effects, which could be as high as 45%.

© 2014 Elsevier B.V. All rights reserved.

. Introduction

Studying the local pharmacokinetics of molecules throughuman skin is not only important within the pharmaceutical indus-ry but also in the field of environmental toxicology. The skin,eing the largest organ, is considered as a route of administrationor topically applied medicines [1,2], but is also important in theermal risk assessment of hazardous compounds, such as myco-oxins [3]. Both in vivo and in vitro methods can be used to measurehe skin absorption. Laboratory animals (such as guinea pigs, rats,

ice and pigs), readily available, provide indeed a reproducible,hysiologically and metabolically intact test system to investigatehe skin absorption of all kinds of compounds (e.g. pharmaceut-cals, cosmetics, hazardous chemicals). However, these have alsoheir limitations, i.a. inter-species variability, with often a higher

ermeability than for human skin. Therefore, human skin studiesemain the “gold standard” by which all methods for measuringercutaneous absorption should be judged. However, given the

∗ Corresponding author. Tel.: +32 9 264 81 00; fax: +32 9 264 81 93.E-mail address: Bart.DeSpiegeleer@UGent.be (B. De Spiegeleer).

ttp://dx.doi.org/10.1016/j.jpba.2014.07.021731-7085/© 2014 Elsevier B.V. All rights reserved.

extreme toxicity of some chemicals, such as mycotoxins, it is eth-ically unacceptable to use living human beings in the transdermalstudies [3,4]. In vitro Franz diffusion cell (FDC) methods are cur-rently the ideal alternative, since (i) it is possible to maintain thebarrier properties of the stratum corneum in excised skin, (ii) thereis good evidence that the obtained in vitro data are predictive forin vivo percutaneous absorption (in vitro in vivo correlation) and(iii) there are standardisation recommendations, guidelines andprotocols on how to execute these diffusion cell studies available,proposed by both regulatory entities and committees of interestedparties [4,5].

Cyclic depsipeptides are a large group of naturally occurringbioactive peptides. Some of these are secondary fungal metabo-lites, which are toxic to humans and animals, such as the emergingmycotoxins beauvericin (BEA) and enniatins (ENNs) [6–18]. Cur-rently, dermal exposure data of these compounds are completelyabsent with a lack of local skin and systemic kinetics, despitetheir widespread skin contact and intrinsic hazard. BEA and ENNs

are well-known Fusarium cyclic hexadepsipeptides, but they arealso produced by other fungi such as Beauveria and Paecilomyscesand Alternaria, Halosarpheia and Verticillium species, respec-tively [19–25]. These compounds, possessing cation-complexing

tical a

iavtt

bdcmQsienfilaaupaolddg

atascttw

2

2

sepwemEgfsw(10(fbh(tte

L. Taevernier et al. / Journal of Pharmaceu

onophoric and lipophilic properties, which act as inhibitors ofcyl-CoA: cholesterol acyltransferase, exert cytotoxic effects inarious cell lines [6–9,11–15,17,18] and have different effects onhe immune system [26–30]. Moreover, it was also demonstratedhat BEA is genotoxic [6,10].

UHPLC-MS/MS analysis of these cyclic depsipeptides has onlyeen reported in a few studies, all aiming for a multi-mycotoxinetermination in food. Considering the abundance (up to 191ompounds) and diversity of the investigated mycotoxins, theseethods have relatively long run times, e.g. up to 21 min [31–33].uantitative transdermal kinetics are characterised by multiple

kin donors with sufficient replicates and time points, which resultn a large amount of different samples, i.e. dose solutions, skinxtraction samples and receptor fluid samples. Hence, there is aeed for a sensitive, selective and rapid high-throughput method

or analysis of these frequently low concentrated samples, obtainedn each FDC experiment. During these FDC experiments, the ana-ytes are also exposed to elevated temperatures for significantmounts of time, i.e. ±32 ◦C (mimicking the human skin temper-ture) during 24 h, indicating the importance of a stability studynder these conditions. During analytical processes, adsorption ofeptides, which is believed to be mostly due to non-covalent inter-ctions and depending upon the experimental conditions, cannotnly lead to significant loss of the analyte, but also to increased ana-ytical variability [34,35]. The adsorption of these lipophilic cyclicepsipeptide analytes to the FDC glass wall, of which the qualityiffers from analytical volumetric glassware, was not yet investi-ated.

The goal of this study was to develop a sensitive, selectivend rapid high-throughput bioanalytical method to quantita-ively determine the cyclic depsipeptide mycotoxins beauvericinnd enniatins (A, A1, B, B1, D, E, F or C) in different FDCamples, using Ultra High Performance Liquid Chromatographyombined with electrospray ionisation (ESI) tandem Mass Spec-rometry (UHPLC-MS/MS). In addition, stability and adsorptiono glass under our in vitro test conditions were investigated asell.

. Materials and methods

.1. Chemicals and reagents

Mycotoxins beauvericin (BEA) and enniatin B (ENN B) wereupplied by BioAustralis (Smithfield NSW, Australia), while thenniatin mixture (ENNs) was obtained from Cfm Oscar Tro-itzsch (Marktredwitz, Germany). No formal ENN compositionas supplied by the manufacturer, therefore the composition was

xperimentally determined by our group (Appendix A, Supple-entary data): 43.8% ENN B, 34.4% ENN B1, 14.0% ENN A1, 3.6%

NN D, 1.8% ENN A, 1.8% ENN E and 0.4% ENN C or F. ULC-MSrade acetonitrile (ACN), formic acid (FA) and 2-propanol, usedor preparation of the mobile phase, were purchased from Bio-olve (Valkenswaard, The Netherlands). Ultrapure water (H2O)as produced by an Arium pro VF TOC water purification system

Sartorius, Göttingen, Germany), resulting in ultrapure water of8.2 M� × cm quality. Sigma–Aldrich (St. Louis, MO, USA) supplied.01 M phosphate buffered saline (PBS) and dimethyl sulfoxideDMSO). Ethanol (EtOH), used for the dose solutions, was purchasedrom Merck (Darmstadt, Germany) and UHPLC grade ACN wasought from Fisher Scientific (Waltham, MA, USA). Pharma gradeydroxypropyl-�-cyclodextrin (HPBCD) was supplied by Cerestar

Mechelen, Belgium). This was used as a solubilising modifier tohe receptor fluid (PBS), in order to guarantee sink conditions ofhe hydrophobic cyclic depsipeptide mycotoxins throughout thexperiment [36].

nd Biomedical Analysis 100 (2014) 50–57 51

2.2. Analytical method

2.2.1. Preparation of standard solutionsA separate stock solution of 100 �g/ml in ACN was prepared for

BEA and the ENN mixture. ENN B (pure) was used as an internalstandard (IS) for the determination of BEA, while BEA was usedfor the different enniatins present in the enniatin mixture. For eachinternal standard, a stock solution of 10 �g/ml in ACN was preparedand stored at −80 ◦C. For all experiments, except for the stabilityand adsorption tests, an internal standard was added to each sam-ple, with a final IS concentration of 20 ng/ml. From these four stocksolutions (BEA 100 �g/ml in ACN, ENNs 100 �g/ml in ACN, ENN B10 �g/ml in ACN and BEA 10 �g/ml in ACN), the standard solutionswere prepared by dilution in ACN–H2O (70:30, v/v).

2.2.2. In vitro human skin Franz diffusion cell protocolBriefly, the set-up consists of static Franz diffusion cells with

a receptor compartment of 5 ml and an available diffusion areaof 0.64 cm2 (Logan Instruments Corp., New Jersey, USA). Humanskin from the abdominal region is collected from patients whohad undergone cosmetic reduction surgery, with informed consentand confidentiality procedures in place (University Hospital, Ghent,Belgium). Immediately after surgical removal, the skin is cleanedwith 0.01 M PBS pH 7.4 and the subcutaneous fat is removed(OECD, 2004). The skin samples are wrapped in aluminium foiland stored at −20 ◦C for no longer than 6 months. Just before thestart of the experiments, the full-thickness skin is thawed, mountedon a template and dermatomed using an electrical powered der-matome (Integra Life Sciences, New Jersey, USA). The skin samplesare visually inspected for skin damage and are then sandwichedbetween the donor and acceptor chambers, with the epidermis fac-ing upwards, making sure all air under the skin is removed. Thewhole assembly is fixed on a magnetic stirrer and the receptor fluidwas continuously stirred using a Teflon coated magnetic stirringbar (600 rpm) to ensure sink conditions. Before starting the exper-iments, skin integrity is checked by measuring the skin impedanceusing an automatic micro-processor controlled LCR impedancebridge (Tinsley, Croydon, UK). Skin pieces with an impedance value<10 k�, a validated system-suitability cut-off, are discarded andreplaced by a new piece [37]. The dose solutions are topicallyapplied to the epidermal surface of the skin. Then, the donor cham-ber is covered with parafilm and the temperature of the receptorcompartment is kept at 32 ± 1 ◦C. Samples (200 �l) are drawn atregular time intervals from the sampling port and are immedi-ately replaced by 200 �l fresh receptor solution (the analyticallydetermined assay values in the FDC samples are correspondinglycorrected for these replenishments). At the end of the experiment(i.e. after 24 h), the skin surfaces are swabbed with cotton wool toremove the remaining donor solution and then epidermis and der-mis are separated. These samples are analysed as well and are usedto construct a mass balance.

2.2.3. Ultra high performance liquid chromatographyThe chromatography platform consisted of an Acquity UHPLC

equipped with a temperature controlled autosampler tray andcolumn oven, thermostated at 25 ◦C (±5 ◦C) and 45 ◦C (±5 ◦C),respectively (Waters, Milford, MA, USA). Chromatographic sepa-ration was achieved on an Acquity UHPLC charged surface hybrid(CSH) C18 column (1.7 �m, 100 mm × 2.1 mm, 130 A), attached toan Acquity UHPLC VanGuard pre-column (1.7 �m, 5 mm × 2.1 mm,130 A), both obtained from Waters. The injection volume was

10 �l and the needle wash consisted of DMSO-2-propanol-ACN(10:10:80, v/v/v). The isocratic flow rate was set to 0.6 ml/min,using ACN–H2O (70:30, v/v) containing 0.1% FA and 0.1% 2-propanol as mobile phase. The run time was 4.5 min, of which the

52 L. Taevernier et al. / Journal of Pharmaceutical and Biomedical Analysis 100 (2014) 50–57

ericin

fio

2

mTwwawvwTcM

2

2

itam1dHaaHwfw(1f

Fig. 1. Combined MRM overlay, normalised to the largest peak (ENN B). Beauv

rst 1.5 min was diverted to the waste. A typical combined MRMverlay chromatogram is shown in Fig. 1.

.2.4. Mass spectrometryFor the quantitative mycotoxin analysis, a multi-UHPLC-MS/MS

ethod was developed, with the mass spectrometer being a XevoQ-S detector (Waters, Milford, MA, USA). The mass spectrometeras operated in the positive electrospray ionisation mode (ESI+),ith an optimised capillary voltage of 3.50 kV, cone voltage of 50 V

nd source offset of 60 V. Source and desolvation temperaturesere set at 150 and 600 ◦C, respectively, while cone and desol-

ation gas flows were 150 and 1000 l/h, respectively. Acquisitionas performed in the multiple reaction monitoring (MRM) mode.

he selected precursor and product ions, together with the appliedollision energies are given in Table 1. Data were acquired usingasslynx software (V4.1 SCN 843, Waters, Milford, MA, USA).

.3. In-house method verification

.3.1. Adsorption to FDC glassA critical aspect in terms of bioanalytical method development

s the adsorption of analytes, which is often overlooked. Thereforehe adsorption of BEA and ENNs to FDC glass was determineds part of the analytical robustness. Therefore, BEA and ENNsixture were solubilised at a concentration of approximately

000 ng/ml in six different solvent mixtures (formulations), withifferent percentages of organic solvents: aqueous solutions, i.e.2O–EtOH (90:10, v/v) (1) and H2O–ACN (90:10, v/v) (4), as wells intermediate organic solutions, i.e. H2O–EtOH (50:50, v/v) (2)nd H2O–ACN (50:50, v/v) (5) and high organic solutions, i.e.2O–EtOH (5:95, v/v) (3) and H2O–ACN (5:95, v/v) (6). Theseere exposed to FDC glass in duplicate and left to equilibrate

or 24 h at 25 ◦C while continuously stirred (600 rpm), after

hich three independent aliquots were taken and diluted 1:10

v/v) with mobile phase, resulting in a final concentration of00 ng/ml and assayed (hence, n = 2 × 3). Responses were analysedor each compound separately, using ln-lin models which were

at a concentration of 20 ng/ml, whereas the enniatin mixture was 500 ng/ml.

fitted using generalised estimating equations with unstructuredcovariance to account for correlation within duplicates [38].QQ-plots confirmed the normality of the raw residuals in thesemodels. In general, the model can be described as: ln(meanresponse) = � + �1 × F2 + �2 × F3 + �3 × F4 + �4 × F5 + �5 × F6, withFk = 1 or 0, when the formulation is equal to or different from k,respectively. The complete models, including coefficients � and�1-5, can be found in Table B1 in Appendix B (Supplementary data).For each compound separately, the mean response ratios (in %)were evaluated: (i) ACN and EtOH were compared per concentra-tion level (10, 50 or 95% organic solvent) and (ii) the concentrationlevels (10 and 50%) were compared to the reference (95% = noadsorption assumed) per organic solvent (ACN or EtOH). Reported95% confidence intervals were Bonferroni-adjusted to accountfor multiplicity in the analysis of each compound separately. Anoverview of these results is given in Tables B2 and B3, respectively(Supplementary data).

2.3.2. Analytical stabilityFrom one duplicate of each of the six formulations from the

adsorption experiment (see Section 2.3.1), multiple aliquots weretaken and stored in HPLC glass vials protected from light at differ-ent conditions (−35, 5, 25 and 40 ◦C). These were analysed at T0,T2days and T7days, after 1:10 (v/v) dilution with mobile phase andeach compared with corresponding freshly prepared standard solu-tions (100 ng/ml). The percentage label claim (l.c.) was calculatedfor each compound separately, using the mean response factor ofthe standards, which were analysed on each experimental day,according to the following formulas: (1) response factor (RF) = areastandard/theoretical concentration standard and (2) label claimpercentage (%) = (area sample Tx/theoretical concentration sam-ple × RF) × 100%. Next, for the worst case stability scenario of 40 ◦C,

these percentages were plotted against time (days) for each com-pound and formulation. Stability was evaluated based on the 95%confidence interval around the slope in a linear regression analysisof the recovery (%) against time (days) for the worst case scenario

L. Taevernier et al. / Journal of Pharmaceutical and Biomedical Analysis 100 (2014) 50–57 53

Table 1MRM transitions and MS/MS parameters.

Compound Retention time (min) Precursor ion (m/z) Product ions (m/z) Collision energy (eV)

ENN B ±1.99 639.91 [M + H]+ 196.08 25527.26 22

ENN D ±2.29 653.99 [M + H]+ 196.09 23541.05 21

ENN B1 ±2.47 653.99 [M + H]+ 196.09 23541.05 21

BEA ±2.64 783.94 [M + H]+ 255.04 24623.23 23

ENN E ±2.86 668.07 [M + H]+ 209.9 24555.29 21

ENN A1 ±3.10 668.07 [M + H]+ 209.9 24555.29 21

ENN C or F ±3.61 682.47 [M + H]+ 209.93 26555.01 23

ENN A ±3.92 682.47 [M + H]+ 209.93 26

F ere ac

(o

2

u1(wm

2

no15ccc

2

topq

2

(mthtTawdt

2

(ls

or structure isomers (ENN D-B1, ENN E-A1 and ENN C/F-A) identical transitions w

40 ◦C). If this interval contains zero, no significant degradation isbserved.

.3.3. Calibration curveLinearity was evaluated by constructing a calibration curve

sing different standard solutions (1, 5, 10, 50, 100, 500 and000 ng/ml BEA or ENN mixture), containing the respective IS20 ng/ml ENN B or BEA). For each compound the linear rangeas evaluated and the coefficient of determination (R2) was deter-ined.

.3.4. Limit of detectionThe limit of detection (LoD) was determined using the signal-to-

oise approach, where LoD corresponds to a signal-to-noise ratiof 3:1. For BEA and ENNs B, B1, D, E, A and A1 a standard solution of

ng/ml beauvericin or enniatins mixture was used, while this was ng/ml enniatins mixture for ENNs C/F and A. For each ENN, theoncentration of the mixture was converted to the individual ENNoncentration, according to the previously determined chemicalomposition (see Section 2.1).

.3.5. Injection repeatabilityThe method injection repeatability was characterised by sex-

uplicate injections of standard solutions at 100 ng/ml beauvericinr enniatin mixture, containing their respective IS. The calculatedercentage RSD should be ≤10%, based on the EDQM guideline forualification of mass spectrometers [39].

.3.6. Accuracy and precision of receptor fluid samplesTo evaluate the precision and accuracy, the receptor fluid

0.01 M PBS +1% HPBCD) was spiked with beauvericin or enniatinixture at two concentration levels (10 and 100 ng/ml), each in

riplicate (n = 2 × 3), and left to equilibrate for 24 ± 0.25 h in pre-eated (32 ± 1 ◦C) FDCs. A 200 �l aliquot was taken at the end ofhe experiment, diluted with the respective IS in ACN and analysed.he percentage residual standard deviation (%RSD) and percent-ge bias were calculated. For the latter the following formulaas used: bias (%) = (concentration sample–concentration stan-ard)/concentration standard × 100%, where the concentration ofhe sample is calculated using the calibration curve.

.3.7. Effect of skin components on the mycotoxin MS signal

Receptor fluid (0.01 M PBS +1% HPBCD) and extraction solvent

ACN–H2O, 95:5, v/v), both as such and previously exposed to aarge piece of full-thickness skin for 3.5 ± 0.5 h in an incubationhaker set at 32 ◦C/150 rpm, were spiked with BEA or ENNs mixture

555.01 23

quired.

(100 ng/ml) and their respective internal standard and analysed.For the receptor fluid, this was done in triplicate (n = 3). This wasalso done in triplicate for the extraction solvent and combined withtwo dilution procedures, i.e. 1:2 (v/v), resulting in a final concen-tration of 50 ng/ml, and 1:4 (v/v), resulting in a final concentration25 ng/ml (hence n = 2 × 3). The matrix effect is expressed as per-centage recovery (responseskin soaked/responseas such × 100%).

3. Results and discussion

3.1. Method development

The method development was based upon the multi-mycotoxinmethod described by Van Pamel et al. [33], using an Acquity UPLCBEH (Ethylene Bridged Hybrid) C18 (1.7 �m, 100 mm × 2.1 mm,L1) column, thermostated at 45 ◦C. A gradient mobile phase con-sisting of H2O–ACN with 1 mM ammonium acetate, 0.1% FA and0.1% 2-propanol was used at a flow of 0.8 ml/min and an injec-tion volume of 10 �l was applied. Considering the objective todetermine only BEA and ENNs, an isocratic method was devel-oped, using the same Acquity UPLC BEH C18 column, by firstperforming a scouting gradient, from which the optimal mobilephase composition was found to be approximately 70:30 (v/v)ACN–H2O. However, this method resulted in significant and unac-ceptable tailing (see Fig. D1 in Appendix D). Beside this BEHUPLC C18 column, other research groups have mentioned theuse of other column chemistry types, i.e. Gemini LC C18 (5 �m,150 mm × 4.6 mm, L1), ZORBAX UPLC Eclipse Plus C18 (1.8 �m,150 mm × 2.1 mm, L1), Luna LC C18 (5 �m, 150 mm × 3 mm, L1),Shiseido Capcell LC C18 (5 �m, 250 mm × 4.6 mm, L1) and Gem-ini LC C6 phenyl (3 �m, 50 mm × 2 mm, L11) columns [40–44].Therefore, we switched to a different stationary phase, i.e. AcquityUPLC BEH phenyl (1.7 �m, 100 mm × 2.1 mm, L11), which unfortu-nately still led to bad peak shapes with unacceptable tailing (seeFig. D2 in Appendix D). Next, an Acquity UPLC BEH RP18 Shieldcolumn (1.7 �m, 100 mm × 2.1 mm, L1) was used, which incor-porates a hydrophilic group within the C18 chain. This change instationary phase resulted in significantly improved peak shapes.However, not all enniatin stereoisomers in the mixture were base-line separated, an issue which remained unresolved even afterlowering the organic amount in the mobile phase composition(see Fig. D3 in Appendix D). Therefore, a switch was made to the

Acquity UPLC CSH C18 column (1.7 �m, 100 mm × 2.1 mm), with alow level surface charge, designed to improve sample loadabilityand peak asymmetry in low-ionic-strength mobile phases. Indeed,higher resolution and overall better performance characteristics

5 tical and Biomedical Analysis 100 (2014) 50–57

wats(Cs[iaatws

3

ccc((wFb5l(gpfttmtcmfnlttagctfAtcsTTomi

3

wcanw

80

90

100

110

120

130

140

150

Resp

onse

ra�o

(%)

80

90

100

110

120

130

140

150

Resp

onse

ra�o

(%)

80

90

100

110

120

130

140

150

Resp

onse

ra�o

(%)

(A)

(B)

(C)

BEA ENN A ENN C/F ENN A1 ENN E ENN D ENN B1 ENN B

BEA ENN A ENN C/F ENN A1 ENN E ENN D ENN B1 ENN B

BEA ENN A ENN C/F ENN A1 ENN E ENN D ENN B1 ENN B

Fig. 2. Adsorption to FDC glass. Response ratios (in %) with 95% confidence intervalsof EtOH versus ACN at 10% (A), 50% (B) and 95% (C) organic solvent. The target

4 L. Taevernier et al. / Journal of Pharmaceu

ere achieved in comparison to the UPLC BEH RP18 Shield station-ry phase (see Table D1, Appendix D). It is also worth mentioninghat the retention order changed upon switching between thesetationary phases, from BEA eluting between ENN A1 and ENN C/FBEH RP18 Shield) to BEA eluting between ENN B1 and ENN E (CSH18). The MS instrument was operated in the positive ion electro-pray mode, previously confirmed to be the most sensitive mode33,40–42,44,45]. In our experiments, [M + H]+ ions gave the mostntense signal and were monitored, as did Van Pamel et al. [33]nd Sorensen et al. [40]. By infusing mixture solutions of both BEAnd ENNs at concentrations of 100 and 500 ng/ml, the MS parame-ers capillary voltage, cone voltage, source offset and probe positionere optimised towards maximal signal intensity, as well as the

election of product ions and optimisation of collision energies.

.2. Adsorption to FDC glass

The mean response ratios (in %) were evaluated for eachompound separately: (i) ACN and EtOH were compared per con-entration level (10, 50 or 95% organic solvent) (Fig. 2) and (ii) theoncentration levels (10 and 50%) were compared to the reference95% = no adsorption assumed) per organic solvent (ACN or EtOH)Fig. 3). The target response ratio of 100% is indicated by a green line,hile the pre-set specification limits are denoted by the red lines.

or the majority of compounds, a statistically significant differenceetween EtOH and ACN was observed (p values < 0.05) for 10 and0% organic solvent levels, and EtOH formulations tended to give

ower adsorption. If not, this might be related to a larger variabilitye.g. ENN B1). However, considering a 10% deviation from the tar-et response ratio of 100% as biopharmaceutically negligible (i.e. there-set specification limits), a biopharmaceutically significant dif-erence was only observed for BEA at 10% organic solvent (based onhe 95% confidence interval excluding the specification limits). BEA,he most lipophilic compound (highest log P), showed significantly

ore adsorption to FDC glass when formulated in ACN. Moreover,here was no biopharmaceutically significant adsorption effect at aoncentration of ≥50% EtOH for all investigated cyclic depsipeptideycotoxins. The latter also applied to the ACN formulations, except

or ENN B1, where a possible significant adsorption effect couldot be excluded. For BEA, significant adsorption was observed at

ow levels of organic solvent (10% EtOH or ACN), where adsorp-ion losses as high as 45% were observed (Fig. 3C). Most likely,he hydrophobicity of the compounds plays a role in the observeddsorption effects: compounds with higher log P values tend toive more adsorption to FDC glass. Log P values for all investigatedyclic depsipeptide mycotoxins are given in Table B4 (Supplemen-ary information). These results were taken into account during theurther method development: at least 50% organic solvent, eitherCN or EtOH, was used when preparing solutions or samples con-

aining BEA and/or ENNs. Only the receptor fluid (RF) itself did notontain any organic solvent, as the use of organic solvents is con-idered not well-suited as a biocompatible receptor medium [46].herefore, 1% HPBCD was added as a modifier (see Section 2.1).o investigate if this modifier can sufficiently inhibit adsorptionf the cyclic depsipeptides to FDC glass and to further verify thisethod, precision and accuracy of the obtained RF samples were

nvestigated as well (see Sections 2.3.7 and 3.7).

.3. Analytical stability

Linear regression analysis of the recovery (%) against time (days)as performed for each compound and formulation for the worst

ase scenario (40 ◦C), after which the 95% confidence intervalround the slope was investigated. If this interval contains zero,o significant degradation is observed. Fig. 4 presents the plots,hereas regression analyses data can be found in Table C1 in

recovery (100%) is indicated by the green line, while the pre-set specification limits(90–110%) are denoted by the red lines. (For interpretation of the color informationin this figure legend, the reader is referred to the web version of the article.)

Appendix C (Supplementary data). These data confirm that therewas no significant degradation of BEA and ENNs in these sam-ple solutions stored for a period of 7 days at 40 ◦C. Consideringthis is the worst case scenario, we conclude that there is like-wise no evidence of degradation for the other storage conditions(−35, 5 and 25 ◦C). In one case, namely for ENN A solubilised inACN–H2O (10:90, v/v), the 95% confidence interval around the slopedid not contain zero (−1.54, −0.49). However, following justifica-tions allow us to assume that there is also no biopharmaceuticallysignificant degradation for this compound in this formulation: (i) nosignificant difference between the different storage conditions at 7days was observed, (ii) for the other tested formulations there wasno significant degradation observed, (iii) given the many tests, this

may well be a false positive finding (e.g. the 99% confidence levelinterval does contain zero (−3.65, 1.62)), (iv) ENN A is a stereoiso-mer of ENN C/F and its chemical structure is also very similar to theother enniatins (A1, B, B1, C, D and E), which were demonstrated

L. Taevernier et al. / Journal of Pharmaceutical and Biomedical Analysis 100 (2014) 50–57 55

40

50

60

70

80

90

100

110

120

Resp

onse

ra�o

(%)

40

50

60

70

80

90

100

110

120

Resp

onse

ra�o

(%)

40

50

60

70

80

90

100

110

120

Resp

onse

ra�o

(%)

40

50

60

70

80

90

100

110

120

Resp

onse

ra�o

(%)

(A) (B)BEA ENN A ENN C/F ENN A1 ENN E ENN D ENN B1 ENN B BEA ENN A ENN C/F ENN A1 ENN E ENN D ENN B1 ENN B

BEA ENN A ENN C/F ENN A1 ENN E ENN D ENN B1 ENN B (C) BEA ENN A ENN C/F ENN A1 ENN E ENN D ENN B1 ENN B (D)

Fig. 3. Adsorption to FDC glass. Response ratios (in %) with 95% confidence intervals of EtOH10%/EtOH95% (A), EtOH50%/EtOH95% (B), ACN10%/ACN95% % (C) and ACN50%/‘95%. Thet limitsi

tfOnaitch

3

jciiftw51w

3

t11

arget recovery (100%) is indicated by the green line, while the pre-set specification

n this figure legend, the reader is referred to the web version of the article.)

o be stable under these experimental conditions, and (v) the dif-erence between T0 and T7days results were still <10% relative (7%).verall, it was concluded that there is neither biopharmaceuticallyor statistically significant degradation of enniatins or beauvericinfter 7 days, when formulated in an organic or aqueous mixture,ndicating no evidence of an unacceptable analytical stability ofhese compounds. Also no trend could be demonstrated, i.e. harsheronditions (ranging between −35 and 40 ◦C) do not translate intoigher degradation.

.4. Calibration curve

The linear calibration curve was forced through zero, which isustified by the fact that the difference in slope between the normalalibration curve and the one forced through zero is not significant,.e. the 95% confidence interval around the intercept (without zeronclusion) also contains zero. For BEA a linear range was obtainedrom 1 to 100 ng/ml with an acceptable coefficient of determina-ion of 0.998. For ENNs B and B1 the same linear range was used,hereas for ENN A1 this was 1–500 ng/ml and for ENN A this was

–1000 ng/ml. For all other ENNs a linear range was obtained from to 1000 ng/ml. For all enniatins the determination coefficientsere equal to 1.000.

.5. Limit of detection

All compounds have similar experimentally determined detec-ion limits, which was expected due to their structural similarity:7 pg/ml for BEA and ENN B, 14 pg/ml for ENN D and ENN B1,5 pg/ml for ENN E, ENN A1 and ENN A and 10 pg/ml for ENN C/F.

(90–110%) are denoted by the red lines. (For interpretation of the color information

3.6. Injection repeatability

The RSDs ranged between 0.15 and 1.84% for all cyclic dep-sipeptide mycotoxins investigated, which means that the injectionrepeatability of the method is below the pre-set limit of 10% andthus acceptable for our purposes.

3.7. Accuracy and precision of receptor fluid samples

At the 100 ng/ml concentration level, the precision rangedbetween 0.57 and 9.25%, which is within the normal expected vari-ability range of 10%. At the lower concentration level (i.e. 10 ng/ml),these values were slightly higher (3.62–10.70%), but still acceptablefor our purposes. The only exception was for ENN C/F at the lowestlevel with a precision of 24.37%: this can be ascribed to the concen-tration being below the limit of quantification (ENN C/F accountsfor only 0.4% of the total amount enniatins in the mixture). The biasat the highest concentration level (i.e. 100 ng/ml) ranged between0.91 and 24.47%. For the 10 ng/ml concentration level, this was 0.68and 24.86%. These values are slightly higher than the specificationlimits given by the FDA in their guidelines for formal bioanalyt-ical method validation (≤15.0% and 20.0% at LoQ, n = 5), but areconsidered sufficient for our purposes [47]. From these acceptableaccuracies, it is concluded that there are no significant adsorptionlosses upon using 1% HPBCD in PBS as a modifier in the receptorfluid.

3.8. Effect of skin components on the mycotoxin MS signal

All obtained recoveries were between 103.1 and 107.7% for theextraction solvent and between 95.0 and 113.4% for the receptor

56 L. Taevernier et al. / Journal of Pharmaceutical and Biomedical Analysis 100 (2014) 50–57

otoxin

flosfr

4

stmhmodml

A

oIn

Fig. 4. Stability profile for each cyclic depsipeptide myc

uid. These results indicate no significant effect of skin compoundsn the MS signal. Moreover, in an experimental FDC set-up thekin is only exposed at a 0.64 cm2 dermal surface area and notully soaked in the receptor fluid, as in this method verification,epresenting a worst-case situation.

. Conclusions

This manuscript described the development of a sensitive,pecific and high-throughput UHPLC-MS/MS method for the quan-itative and simultaneous determination of cyclic depsipeptide

ycotoxins beauvericin and enniatins (B, B1, A, A1, D, E, C/F) inuman skin Franz diffusion cell samples from in vitro transder-al experiments. The method has an acceptable range for limit

f detection (10–17 pg/ml), a total run time of only 4.5 min and noegradation in aqueous and organic solvents for at least 7 days. Ainimised adsorption of BEA and ENNs to glass was found when at

east 50% ethanol or acetonitrile is used.

cknowledgements

The authors would like to thank the Special Research Fundf Ghent University (BOF 01D23812 to Lien Taevernier) and thenstitute for the Promotion of Innovation through Science and Tech-ology in Flanders (IWT 110533 to Kirsten Vandercruyssen and

stored at 40 ◦C for 7 days in six different formulations.

IWT 101529 to Matthias D’Hondt) for their financial funding andscientific interest.

Appendix A. Supplementary data

Supplementary data associated with this article can be found, inthe online version, at http://dx.doi.org/10.1016/j.jpba.2014.07.021.

References

[1] H.A.E. Benson, A.C. Watkinson, Topical and Transdermal DrugDelivery—Principles and Practice, John Wiley and Sons, New Jersey, 2012, pp.446.

[2] A.C. Williams, Transdermal and Topical Drug Delivery—From Theory to ClinicalPractice, Pharmaceutical Press, London, 2003, pp. 242.

[3] J. Boonen, S.V. Malysheva, L. Taevernier, J.D. Di Mavungu, S. De Saeger, B. DeSpiegeleer, Human skin penetration of selected model mycotoxins, Toxicology301 (2012) 21–32.

[4] D. Howes, R. Guy, J. Hadgraft, J. Heylings, U. Hoeck, F. Kemper, H. Maibach,J.P. Marty, H. Merk, J. Parra, D. Rekkas, I. Rondelli, H. Schaefer, U. Täuber, N.Verbiese, Methods for assessing percutaneous absorption—the report and rec-ommendations of ECVAM workshop, ATLA 24 (1996) 81–106.

[5] Organisation for Economic Co-operation and Development (OECD), Guidancedocument for the conduct of skin absorption studies, in: OECD Series on Testingand Assessment, Paris, 2004, pp. 31.

[6] M. Celik, H. Aksoy, S. Yilmaz, Evaluation of beauvericin genotoxicity withthe chromosomal aberrations, sister-chromatid exchanges and micronucleusassays, Ecotox. Environ. Saf. 73 (2010) 1553–1557.

[7] R. Dornetshuber, P. Heffeter, M.R. Kamyar, T. Peterbauer, W. Berger,R. Lemmens-Gruber, Enniatin exerts p53-dependent cytostatic and

tical a

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

L. Taevernier et al. / Journal of Pharmaceu

p53-independent cytotoxic activities against human cancer cells, Chem.Res. Toxicol. 20 (2007) 465–473.

[8] L. Ivanova, E. Skjerve, G.S. Eriksen, S. Uhlig, Cytotoxicity of enniatins A, A1, B,B1, B2 and B3 from Fusarium avenaceum, Toxicon 47 (2006) 868–876.

[9] M. Jestoi, Emerging Fusarium-mycotoxins fusaproliferin, beauvericin, enni-atins, and moniliformin—a review, Crit. Rev. Food Sci. Nutr. 48 (2008) 21–49.

10] M.S. Klaric, D. Darabos, R. Rozgaj, V. Kasuba, S. Pepeljnjak, Beauvericin andochratoxin A genotoxicity evaluated using the alkaline comet assay: single andcombined genotoxic action, Arch. Toxicol. 84 (2010) 641–650.

11] H.S. Lee, H.H. Song, J.H. Jeong, C.G. Shin, S.U. Choi, C. Lee, Cytotoxicities ofenniatins H, I, and MK1688 from Fusarium oxysporum KFCC11363P, Toxicon51 (2008) 1178–1185.

12] L. Macchia, R. Di Paola, F. Fornelli, S. Nenna, A. Moretti, R. Napoletano, A.Logrieco, M.F. Caiaffa, M.F. TursiI, A. Bottalico, Cytotoxicity of beauvericin tomammalian cells, in: Abstracts of the International Seminar on Fusarium, Mar-tina Franca, Italy, 1995, pp. 72–73.

13] G. Meca, G. Font, M.J. Ruiz, Comparative cytotoxicity study of enniatins A, A(1),A(2), B, B-1, B-4 and J(3) on Caco-2 cells, Hep-G(2) and HT-29, Food Chem.Toxicol. 49 (2011) 2464–2469.

14] C. Nilanonta, M. Isaka, P. Kittakoop, S. Trakulnaleamsai, M. Tanticharoen, Y.Thebtaranonth, Precursor-directed biosynthesis of beauvericin analogs by theinsect pathogenic fungus Paecilomyces tenuipes BCC 1614, Tetrahedron 58(2002) 3355–3360.

15] A. Prosperini, G. Meca, G. Font, M.J. Ruiz, Study of the cytotoxic activity ofbeauvericin and fusaproliferin and bioavailability in vitro on Caco-2 cells, FoodChem. Toxicol. 50 (2012) 2356–2361.

16] A.A. Tonshin, V.V. Teplova, M.A. Andersson, M.S. Salkinoja-Salonen, The Fusa-rium mycotoxins enniatins and beauvericin cause mitochondrial dysfunctionby affecting the mitochondrial volume regulation, oxidative phosphorylationand ion homeostasis, Toxicology 276 (2010) 49–57.

17] W. Watjen, A. Debbab, A. Hohfeld, Y. Chovolou, A. Kampkotter, R.A. Edrada,R. Ebel, A. Hakiki, M. Mosaddak, F. Totzke, M.H.G. Kubbutat, P. Proksch, Enni-atins A1, B and B1 from an endophytic strain of Fusarium tricinctum induceapoptotic cell death in H4IIE hepatoma cells accompanied by inhibition of ERKphosphorylation, Mol. Nutr. Food Res. 53 (2009) 431440.

18] J.X. Zahn, A.M. Burns, M.P.X. Liu, S.H. Faeth, A.A.L. Gunatilaka, Search for cellmotility and angiogenesis inhibitors with potential anticancer activity: beau-vericin and other constituents of two endophytic strains of Fusarium oxysporum,J. Nat. Prod. 70 (2007) 227–232.

19] E. Gäumann, S. Naef-Roth, L. Etlinger, P.A. Plattner, U. Nager, Enniatin, einneues gegen mykobakterien wirksames antibiotikum, Experimentia 3 (1947)202–203.

20] R.L. Hamill, C.E. Higgens, H.E. Boaz, M. Gorman, The structure of beauvericin, anew depsipeptide antibiotic toxic to Artemia salina, Tetrahedron Lett. 49 (1969)4255–4258.

21] Y.C. Lin, J. Wang, X.Y. Wu, S.N. Zhou, L.L.P. Vrijmoed, E.B.G. Jones, A novel com-pound enniatin G from the mangrove fungus Halosarpheia sp (strain 732) fromthe south China sea, Aust. J. Chem. 55 (2002) 225–227.

22] T.C. McKee, H.R. Bokesch, J.L. McCormick, M.A. Rashid, D. Spielvogel, K.R.Gustafson, M.M. Alavanja, J.H. Cardellina, M.R. Boyd, Isolation and characteri-zation of newanti-HIV and cytotoxic leads from plants, marine and microbialorigins, J. Nat. Prod. 60 (1997) 431–438.

23] C. Nilanonta, M. Isaka, R. Chanphen, N. Thong-Orn, M. Tanticharoen, Y.Thebtaranonth, Unusual enniatins produced by the insect pathogenic fun-gus Verticillium hemipterigenum: isolation and studies on precursor-directedbiosynthesis, Tetrahedron 59 (2003) 1015–1020.

24] S. Supothina, M. Isaka, K. Kirtikara, M. Tanticharoen, Y. Thebtaranonth, Enniatinproduction by the entomopathogenic fungus Verticillium hemipterigenum BCC1449, J. Antibiot. 57 (2004) 732–738.

25] Q. Wang, L. Xu, Beauvericin, a bioactive compound produced by fungi: a shortreview, Molecules 17 (2012) 2367–2377.

26] A.S. Ficheux, Y. Sibiril, D. Parent-Massin, Effects of beauvericin, enniatin b andmoniliformin on human dendritic cells and macrophages: an in vitro study,

Toxicon 71 (2013) 1–10.

27] A. Gammelsrud, A. Solhaug, B. Dendele, W.J. Sandberg, L. Ivanova, A.K. Bolling,D. Lagadic-Gossmann, M. Refsnes, R. Becher, G. Eriksen, J.A. Holme, Enni-atin B-induced cell death and inflammatory responses in RAW 267.4 murinemacrophages, Toxicol. Appl. Pharmacol. 261 (2012) 74–87.

[

[

nd Biomedical Analysis 100 (2014) 50–57 57

28] K. Schaeuble, M.A. Hauser, A.V. Rippl, R. Bruderer, C. Otero, M. Groettrup, D.F.Legler, Ubiquitylation of the chemokine receptor CCR7 enables efficient recep-tor recycling and cell migration, J. Cell Sci. 125 (2012) 4463–4474.

29] H. Tomoda, H. Nishida, X.H. Huang, R. Masuma, Y.K. Kim, S. Omura, Newcyclodepsipeptides, enniatin-D, enniatin-E and enniatin-F produced by Fusa-rium sp FO-1305, J. Antibiot. 45 (1992) 1207–1215.

30] X.F. Wu, R. Xu, Z.J. Ouyang, C. Qian, Y. Shen, W.D. Wu, Y.H. Gu, Q. Xu, Y. Sun,Beauvericin ameliorates experimental colitis by inhibiting activated T cells viadownregulation of the PI3K/Akt signaling pathway, PLoS ONE 8 (2013) e83013.

31] A.G. Frenich, R. Romero-Gonzalez, M.L. Gomez-Perez, J.L.M. Vidal, Multi-mycotoxin analysis in eggs using a QuEChERS-based extraction procedure andultra-high-pressure liquid chromatography coupled to triple quadrupole massspectrometry, J. Chromatogr. A 1218 (2011) 4349–4356.

32] E. Varga, T. Glauner, F. Berthiller, R. Krska, R. Schuhmacher, M. Sulyok, Devel-opment and validation of a (semi-)quantitative HPLC-MS/MS method for thedetermination of 191 mycotoxins and other fungal metabolites in almonds,hazelnuts, peanuts and pistachios, Anal. Bioanal. Chem. 405 (2013) 5087–5104.

33] E. Van Pamel, A. Verbeken, G. Vlaemynck, J. De Boever, E. Daeseleire,Ultrahigh-performance liquid chromatographic-tandem mass spectrometricmultimycotoxin method for quantitating 26 mycotoxins in maize silage, J.Agric. Food Chem. 59 (2011) 9747–9755.

34] H.M.D.R. Herath, R.R. Kim, P.J. Cabot, P.N. Shaw, A.K. Hewavitharana, Inaccura-cies in the quantification of peptides—a case study using �-endorphin assay,LCGC North America 31 (2013) 58–61.

35] A. Pezeshki, V. Vergote, S. Van Dorpe, B. Baert, C. Burvenich, A. Popkov, B.De Spiegeleer, Adsorption of peptides at the sample drying step: influence ofsolvent evaporation technique, vial material and solution additive, J. Pharm.Biomed. Anal. 49 (2009) 607–612.

36] B. Baert, J. Boonen, C. Burvenich, N. Roche, F. Stillaert, P. Blondeel, J. VanBocxlaer, B. De Spiegeleer, A new discriminative criterion for the developmentof Franz diffusion tests for transdermal pharmaceuticals, J. Pharm. Pharm. Sci.13 (2010) 218–230.

37] B. De Spiegeleer, B. Baert, V. Vergote, S. Van Dorpe, Development of systemsuitability tests for in vitro skin integrity control: impedance and capacitance,in: The Eleventh International Perspectives in Percutaneous Penetration Con-ference, La Grande Motte, France, 2008, pp. 43–43.

38] K.Y. Liang, S.L. Zeger, Longitudinal data analysis using generalized linear mod-els, Biometrics 78 (1986) 13–22.

39] European Directorate for the Quality of Medicines and Health Care (EDQM),Qualification of equipment–Annex 7: qualification of mass spectrometers,2011, pp. 11.

40] J.L. Sorensen, K.F. Nielsen, P.H. Rasmussen, Development of a LC-MS/MS methodfor the analysis of enniatins and beauvericin in whole fresh and ensiled maize,J. Agric. Food Chem. 56 (2008) 10439–10443.

41] M. Jestoi, M. Rokka, A. Rizzo, K. Peltonen, Determination of Fusarium-mycotoxins beauvericin and enniatins with liquid chromatography-tandemmass spectrometry (LC-MS/MS), J. Liq. Chromatogr. Relat. Technol. 28 (2005)369–381.

42] M. Sulyok, R. Krska, R. Schuhmacher, A liquid chromatography/tandem massspectrometric multi-mycotoxin method for the quantification of 87 analytesand its application to semi-quantitative screening of moldy food samples, Anal.Bioanal. Chem. 389 (2007) 1505–1523.

43] S.M. Monti, V. Fogliano, A. Logrieco, R. Ferracane, A. Ritieni, Simultaneous deter-mination of beauvericin, enniatins and fusaproliferin by high performanceliquid chromatography, J. Agric. Food Chem. 48 (2000) 3317–3320.

44] M. Sulyok, F. Berthiller, R. Krska, R. Schuhmacher, Development and vali-dation of a liquid chromatography/tandem mass spectrometric method for thedetermination of 39 mycotoxins in wheat and maize, Rapid Commun. MassSpectrosc. 20 (2006) 2649–2659.

45] M. Devreese, S. De Baere, P. De Backer, S. Croubels, Quantitative determinationof the Fusarium mycotoxins beauvericin, enniatin A, A1, B and B1 in pig plasmausing high performance liquid chromatography-tandem mass spectrometry,Talanta 106 (2013) 212–219.

46] V.P. Shah, The role of dissolution testing in the regulation of pharmaceut-icals: the FDA perspective, in: J. Dressman, J. Kramer (Eds.), PharmaceuticalDissolution Testing, Taylor & Francis Group, Boca Raton, 2005, pp. 81–96.

47] Food and Drug Administration (FDA), Guidance for Industry—BioanalyticalMethod Validation, 2013, pp. 34.