Accepted Manuscript

Title: Determination of mineral oil aromatic hydrocarbons

(MOAH) in edible oils and fats by online liquid
chromatography–gas chromatography–flame ionization
detection (LC-GC-FID) − Evaluation of automated removal
strategies for biogenic olefins

Authors: Marco Nestola, Torsten C. Schmidt

PII: S0021-9673(17)30740-9
DOI: http://dx.doi.org/doi:10.1016/j.chroma.2017.05.035
Reference: CHROMA 358532

To appear in: Journal of Chromatography A

Received date: 3-2-2017

Revised date: 26-4-2017
Accepted date: 13-5-2017

Please cite this article as: Marco Nestola, Torsten C.Schmidt, Determination of
mineral oil aromatic hydrocarbons (MOAH) in edible oils and fats by online
liquid chromatography–gas chromatography–flame ionization detection (LC-GC-FID)
− Evaluation of automated removal strategies for biogenic olefins, Journal of
Chromatography Ahttp://dx.doi.org/10.1016/j.chroma.2017.05.035

This is a PDF file of an unedited manuscript that has been accepted for publication.
As a service to our customers we are providing this early version of the manuscript.
The manuscript will undergo copyediting, typesetting, and review of the resulting proof
before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that
apply to the journal pertain.
Determination of mineral oil aromatic hydrocarbons
(MOAH) in edible oils and fats by online liquid
chromatography–gas chromatography–flame ionization
detection (LC-GC-FID) – Evaluation of automated
removal strategies for biogenic olefins

Marco Nestolaa*, Torsten C. Schmidtb

a
Axel Semrau GmbH & Co. KG, Stefansbecke 42, D-45549 Sprockhövel

b
Instrumental Analytical Chemistry, University of Duisburg-Essen, Universitätsstr. 5, D-
45141 Essen

*Corresponding Author
Tel: +49 2339 12090, Fax: +49 2339 6030, [email protected]

Highlights
 MOAH quantitation after automated epoxidation of polyunsaturates was improved
 In combination with online LC-GC-FID accurate results for MOAH contents are
obtained
 The designed system is suitable for high-throughput routine environments
 Hydroboration and bromohydrin reaction failed for removal of polyunsaturates

Abstract
The determination of mineral oil aromatic hydrocarbons (MOAH) in foodstuffs gained in
importance over the last years as carcinogenicity cannot be excluded for certain MOAH. The
existence of olefins in foodstuffs, such as edible oils and fats, can be problematic for the
determination of MOAH by LC-GC-FID. Removal of these interfering substances by HPLC
1
based on polarity differences is not possible. During gas chromatographic separation heavily
overloaded peaks are observed rendering the detection of small mineral oil contaminations
almost impossible. Therefore, removal of these olefins is necessary before subjection of the
sample to LC-GC-FID. Epoxidation of olefins to increase their polarity proved to be a
valuable tool in the past. Precision and trueness of the results as shown in a collaborative trial,
however, are relying on exact reaction conditions. Additionally, it is known that certain
MOAH are oxidized during epoxidation and therefore get lost. In the scope of this work,
hydroboration, bromohydrin reaction, and epoxidation were examined for their potential for
derivatization of unsaturated hydrocarbons with increased robustness and higher recovery of
MOAH. Epoxidation by meta-chloroperoxybenzoic acid (mCPBA) delivered the best removal
of olefins. Factors influencing this reaction were enlightened. Adaption of the reaction
conditions and time-controlled automation increased the recovery of polycyclic MOAH. Good
precision (RSDr < 1.5 %) and recovery (95 – 102 %) for MOAH were also observed for
sunflower and olive oils spiked with a lubricating mineral oil (at 24.5 mg/kg of MOAH). The
trueness of the method was verified by analyzing collaborative trial samples.

Keywords
Misinterpretation of olefins as mineral oil, derivatization of olefins, epoxidation, bromohydrin
reaction, hydroboration, reaction conditions

1 Introduction
Hydrocarbons of mineral oil origin account for a large proportion of the known contamination
in foodstuffs. According to an EFSA (European Food Safety Agency) opinion from 2012,
MOH (mineral oil hydrocarbons) contribute also to a high degree to contamination found in
the human body [1]. They can be categorized into two main groups: Saturated (MOSH) and
aromatic hydrocarbons (MOAH) with variable alkyl chain lengths [2]. While the first group
consists of paraffinic and naphthenic saturated hydrocarbons, the second one is composed of
alkylated (possibly partially hydrogenated) aromatic hydrocarbons. The MOAH content of
MOH can roughly range between 0 – 35 % depending on the nature of the mineral oil [3].
While crude oils exhibit higher MOAH contents, refined hydrogenated oils show little to no
MOAH contribution.

2
The presence of MOH contamination in foodstuffs can be attributed to several sources.
Packaging material made from recycled paperboard (e.g. newspapers) that had been printed
on with mineral oil derived ink is one important origin. Additionally, lubricants used during
food processing, wax coatings directly applied to the food, environmental pollution, jute bags,
etc. can be sources for contamination [4]. Found contaminations ranged from below 1 mg/kg
up to several thousands of mg/kg [5]. In 2008, in Ukrainian sunflower oil more than 1000
mg/kg of mineral oil was found [6].

According to recent studies, acute toxicity upon oral intake of MOSH and MOAH is low [1].
Higher molecular MOSH are known to form microgranulomas in liver, spleen, lymph nodes,
and other organs [1, 7]. Because of the structural similarities to polycyclic aromatic
hydrocarbons (PAHs), some MOAH are suspected to have carcinogenic and mutagenic
potential. It is known that alkylated PAHs, e.g., 1-methylpyrene, show increased carcinogenic
potential compared to the parent compound (pyrene) [8]. In vitro assays gave indication that
MOAH from printing ink have genotoxic potential [9].

Although no legislation is established till now for upper limits of MOSH and MOAH,
minimization of both substance classes was advised by the EFSA and other national
authorities such as the German Federal Institute for risk assessment (BfR) [10]. Upper limits
of 0.6 and 0.15 mg/kg for MOSH and MOAH, respectively, were proposed in the past years
derived from a temporary ADI (acceptable daily intake) of 0.01 mg/kg body weight (for a 60
kg person) and a suspected MOAH contribution of 25 % [11]. In 2012, however, this ADI
was withdrawn by the JECFA (Joint FAO/WHO Expert Committee on Food Additives) due
to insufficient scientific data. In 2014, upper limits of 2.0 and 0.5 mg/kg for MOSH and
MOAH, respectively, found in foodstuffs packaged in recycled cardboard were proposed in
the latest draft for the 22nd amendment of the German consumer goods regulation [12]. In
2017, the upper limit for MOSH was removed from the draft and the use of a functional
barrier enjoined [13].

1.1 Analytics of MOSH and MOAH

The determination of MOSH and MOAH is routinely performed by online HPLC-GC-FID.
This method is based on a work by Biedermann et al. [14]. First publications regarding this
topic and hyphenated HPLC-GC can be found already in the early 1990es [15].

3
Shortly, HPLC on bare silica gel is used for the separation of MOH components from the food
matrix (lipids, sugars, etc.). Additionally, MOAH are separated from MOSH. The high
capacity for retention of triglycerides allows the direct injection of edible oils after dilution
[16]. Detection limits of approximately 5 mg/kg for MOSH and MOAH were reported for
selected edible oils. For low-fat containing matrices, such as rice or pasta, detection limits as
low as 0.5 mg/kg were feasible [14].

One remaining problem with certain matrices is the co-elution of biogenic olefins during the
HPLC separation. Some monoterpenes are eluted in the MOSH fraction, while
polyunsaturates, such as carotenes, squalene, and sterenes, can be found in the MOAH
fraction. Because of their natural abundance, these compounds may form large peaks
overloading the subsequent GC separation column hindering the detection of MOAH (see Fig.
1). Because of the low content found in edible oils and fats, the co-elution of monoterpenes
etc. in the MOSH fraction is mostly negligible and therefore out of the scope of this work.

Zoccali et al. described a method capable of removing the polyunsaturates from MOAH by
HPLC [17]. After a first cleanup on silica, the MOAH fraction was separated from the
polyunsaturates on Ag+-treated silica gel. To that end, a commercial silica HPLC column was
flushed with silver nitrate [17, 18]. Squalene from olive oil could be retained while MOAH
with up to three aromatic rings were eluted in a transfer volume exceeding 2 mL. MOAH with
larger ring systems were retained too strongly on the prepared column. The authors gave no
information regarding the elution of sterenes or carotenes present in vegetable oils. Moreover,
stability of silver-ion impregnated HPLC columns is known to be limited [19].

Independently, additional sample cleanup steps were developed [14, 20-21]. Treatment of the
sample with elemental bromine was used to derivatize the biogenic unsaturated hydrocarbons.
Because of the toxicity of bromine and insufficient selectivity, epoxidation by meta-
chloroperoxybenzoic acid (mCPBA) was proposed. Increased polarity of the polyunsaturates
was the aim in both cases. Consequently, a removal of these compounds during the HPLC
separation became feasible.

Typical reaction conditions for epoxide synthesis with mCPBA include the use of
dichloromethane as solvent and possible sub-ambient cooling for improved selectivity [22].
Quenching of the reaction is normally done by washing the sample with a reducing agent such
as sodium thiosulfate. Initial addition of sodium bicarbonate or a subsequent washing step

4
was reported to improve the recovery of acid-labile epoxides. Ring opening could otherwise
occur by catalytic amounts of meta-chlorobenzoic acid formed during the reaction [23].

The proposed reaction route by Biedermann et al. for the determination of MOAH in edible
oils and fats consisted of addition of mCPBA in dichloromethane at sub-ambient
temperatures, i.e., ice bath cooling [14]. The authors noticed that mCPBA also attacked
certain MOAH constituents, such as polycyclic aromatic compounds (PAHs, thiophenes), due
to its oxidation potential. Roughly 20 % of MOAH in non-refined mineral oils were reported
to be lost even at sub-ambient temperatures. Higher mCPBA amounts further increased the
loss of MOAH. The presence of a food matrix containing large amounts of unsaturated fatty
acids was found to be beneficial for the recovery of MOAH. Unsaturated fatty acids were
reported to be attacked prior to MOAH compounds. Therefore, addition of uncontaminated
edible oil as buffering agent was recommended for samples containing only small amounts of
unsaturated fatty acids [14]. This method is most widespread and established in routine
environments.

However, a collaborative trial showed high variances in the obtained results (z-scores > 2)
possibly originating from varying reaction conditions [24]. During routine application, it is
often seen that the recovery of the ISTD used for quantitation of MOAH is diminished by
epoxidation resulting in overestimation of the results. For use in routine environments,
method robustness is yet insufficient and needs to be improved.

For these reasons, aim of the current work was to explore possibilities for removal of
polyunsaturates from the MOAH fraction offering increased robustness. Therefore,
derivatization of the polyunsaturates was further explored. Hydroboration and bromohydrin
reaction were examined for their suitability for removal of polyunsaturates. Investigation of
the reaction conditions of mCPBA epoxidation enabled an optimization and automation of
this technique representing an important achievement to increase method robustness.

2 Materials and Methods

2.1 Samples

Extra virgin olive oil and refined sunflower oil were obtained at the local supermarket and
used for method development and validation. Additionally, edible oil samples from a
collaborative trial performed in 2015 within the CEN/TC275/WG13 work program (European

5
Committee for Standardization) organized by ITERG (Pessac, France) were available. They
consisted of refined olive pomace oil, extra virgin olive oil, and palm oil.

2.2 Chemicals and solutions

Acetone, acetonitrile, dichloromethane, ethanol, and n-hexane were from LGC Promochem
(Picograde quality, Wesel, Germany). The internal standard (ISTD) for MOH quantitation
(Cat. No. 31070 - n-undecane, n-tridecane, bicyclohexyl, α-cholestane, n-pentylbenzene, 1-
methylnaphthalene, 2-methylnaphthalene, 1,3,5-tri-tert-butylbenzene, perylene) and an EPA-
PAH standard (Cat. No. 31011) were supplied from Restek (Bellefonte, PA, USA). A
lubricating oil standard (K009) for spiking experiments was obtained from the Federal
Institute for Materials Research and Testing (BAM, Berlin, Germany). 9-
Borabicyclo(3.3.1)nonane (9-BBN, 0.4 M in hexanes), dibenzothiophene (DBT, 98 %) meta-
chloroperoxybenzoic acid (mCPBA, ≤77 %), N-bromosuccinimide (NBS, ReagentPlus®, 99
%), and sodium thiosulfate (purum p.a., ≥98 % (RT)) were from Sigma Aldrich (Steinheim,
Germany). Sodium formate and sodium sulfate were from Fluka (Buchs, Switzerland).
Quantofix® Peroxide 25 test stripes were obtained from Macherey-Nagel (Düren, Germany).
Water was supplied from a Milli-Q water purification system (Merck, Darmstadt, Germany).

2.3 Sample preparation

Three hundred milligrams of edible oil or fat were weighed into a 10-mL autosampler vial.
The vial was placed onto the autosampler, which performed all subsequent steps fully
automated. As first step, 50 µL of the ISTD solution (100 ng/µL in n-hexane) were added to
the sample.

Hydroboration

The autosampler added 550 µL of n-hexane and 100 µL of a 9-BBN solution in hexane. The
vial was placed into an agitator and was shaken at a speed of 500 rpm (revolutions per
minute) for 12 h at 60 °C. Afterward, the sample was directly injected into the LC-GC-FID
system.

6
Bromohydrin reaction

The autosampler added 650 µL of n-hexane, 3 mL of ethanol/water (90:10, v/v) and 700 µL
of an NBS solution (100 mg NBS in 700 µL acetonitrile/water (75:25, v/v)). The vial was
placed into an agitator and was shaken at a speed of 500 rpm for 30 min at 40 °C. Afterward,
2.5 mL of an aqueous sodium formate solution (100 g/L) were added to destroy excess NBS
and induce phase separation. The vial was shaken at 750 rpm for 30 s. Five hundred
microliters of the n-hexanic upper phase were transferred into a 2-mL autosampler vial
prefilled with a spatula tip of sodium sulfate. The dried organic phase was subjected to LC-
GC-FID.

Epoxidation

The autosampler added 650 µL of n-hexane and 500 µL of an ethanolic mCBPA solution (200
mg/mL) to the sample. The vial was placed into an agitator and was shaken at a speed of 500
rpm for 15 min at room temperature. Afterward, 500 µL of ethanol and 2 mL of an aqueous
sodium thiosulfate solution (100 g/L) were added to destroy excess mCPBA and induce phase
separation. The additional amount of ethanol was added to facilitate phase separation. The
vial was shaken at 750 rpm for 30 s. Five hundred microliters of the n-hexanic upper phase
were transferred into a 2-mL autosampler vial prefilled with a spatula tip of sodium sulfate.
The dried organic phase was subjected to LC-GC-FID.

2.4 LC-GC-FID method

LC-GC-FID experiments were performed on a system from Axel Semrau (Sprockhövel,

Germany). It consisted of a 1260 Infinity HPLC system (binary pump and variable
wavelength detector by Agilent Technologies, Waldbronn, Germany), Master GC with flame
ionization detector (DANI Instruments S.p.A., Cologno Monzese, Italy), and a DualPAL
autosampler (CTC Analytics AG, Zwingen, Switzerland).

Three rotatory switching valves (VICI AG International, Schenkon, Switzerland) were used to
guide the HPLC eluent from the HPLC into the GC. The gas chromatograph was equipped
with an on-column interface (Y-interface) and a solvent vapor exit [25]. The on-column
interface, the carrier gas, and solvent vapor exit were controlled by CHRONECT LC-GC
from Axel Semrau.

7
Typically, 50 µL of the prepared sample (corresponding to 15 mg of edible oil or fat) were
injected onto an Allure Si HPLC column (250 mm x 2.1 mm, 5 µm, 60 Å, Restek, Bellefonte,
PA, USA) without additional column temperature control. The mobile phase consisted of n-
hexane and dichloromethane. Starting at 100 % n-hexane with 300 µL/min, the mobile phase
was changed to 65 % n-hexane in 1.5 min after injection. It was held until 6.0 min. After
elution of the MOAH fraction (4.5 – 6.0 min, 450 µL), the column was backflushed with
dichloromethane at 500 µL/min for 9 min. Afterward, the column was reconditioned with n-
hexane at 500 µL/min for 15 min. The MOAH elution window was verified by UV detection
at 230 nm. 1,3,5-tri-tert-butylbenzene and perylene marked the starting- and end-point of this
window.

LC-GC transfer occurred by the retention gap technique and partially concurrent solvent
evaporation (PCSE) [25]. An uncoated, deactivated precolumn (MXT Siltek-treated stainless
steel, 10 m x 0.53 mm, Restek, Bellefonte) was followed by a steel T-piece union (modified
butt-to-butt connector, SGE Analytical Science by Trajan, Australia) connecting to the solvent
vapor exit and a separation column coated with a 100 % dimethyl polysiloxane film (MXT-1,
Siltek-treated stainless steel, 15 m x 0.25 mm x 0.10 µm, Restek, Bellefonte, PA, USA).

From HPLC, the MOAH fraction was transferred to GC at a carrier gas inlet pressure of 65
kPa (hydrogen) and an oven temperature of 60 °C. The solvent vapor exit was opened 0.5 min
prior to elution of the MOAH fraction and was closed 0.3 min after the fraction was
transferred. Recovery of n-pentylbenzene was quantitative under these conditions. At the
same time, the carrier gas inlet pressure was set to 150 kPa and maintained throughout the
whole analysis. The oven temperature was programmed at 30 °C/min from 60 °C (4 min) to
400 °C (4 min, total run time 18.00 min). The FID base temperature was set to 410 °C. The
gas flows for air, hydrogen, and nitrogen were set to 280, 40, and 25 mL/min, respectively.

Data processing was performed with Clarity 6.2 (DataApex, Prague, Czech Republic).
Quantitation was based on 2-methylnaphthalene (2MN) used as ISTD. The MOAH content
was calculated following the equation

(𝐴𝑀𝑂𝐴𝐻,𝑇𝑜𝑡𝑎𝑙 − 𝐴 𝑀𝑂𝐴𝐻,𝑃𝑒𝑎𝑘𝑠 𝑜𝑛 𝑇𝑜𝑝 ) ∗ 𝑚𝐼𝑆𝑇𝐷

𝐶=
𝐴𝐼𝑆𝑇𝐷 ∗ 𝑚𝑆𝑎𝑚𝑝𝑙𝑒

with C: Content [mg/kg], AMOAH, Total: Unresolved MOAH hump area, AMOAH, Peaks on Top: Area
of sharp peaks on top of the MOAH hump, AISTD: peak area of ISTD, mISTD: mass of ISTD
[mg], mSample: mass of test sample [kg].

8
For identification of brominated PAHs during bromohydrin reaction, the MOAH fraction was
collected in a vial and analyzed by GC-MS (EVOQ GC-TQ, Bruker Daltonik GmbH,
Bremen, Germany). The ion source and transfer line temperatures were set to 230 and 320 °C,
respectively. Data acquisition was performed in full-scan mode (50 – 750 amu) at a rate of 3
spectra/s with EI ionization at 70 eV. Data processing was performed with Bruker MS
Workstation 8.2.

3 Results and Discussion

3.1 Alternative derivatizations of polyunsaturates

As outlined in the introduction, recovery of certain MOAH can be troublesome after

epoxidation of polyunsaturates. Additionally, it is known that epoxidation of electron-
deficient species is much harder than epoxidation of π-electron-rich (conjugated) olefins [26,
27]. Therefore, other derivatization reactions can be of help.

Hydroboration of extra virgin olive oil with 9-BBN (as a longtime stabile alternative to BH3)
and subsequent injection into LC-GC-FID showed no removal of polyunsaturates even after
12 h at 60 °C. If the reaction occurred at all, polarity of the formed organoboranes did not
differ sufficiently from MOAH for a successful removal by silica HPLC. Oxidation of the
possibly formed organoboranes with H2O2/NaOH was avoided, since it was hardly
automatable. Besides corrosiveness, the application of NaOH on edible oil enabled
saponification, which hindered a clear identification of the n-hexanic upper phase. Therefore,
hydroboration was abandoned as removal step for polyunsaturates from edible oils and fats.

Alternatively, bromohydrin reaction, i.e., addition of bromine and a hydroxy group onto an
olefinic double bond, of extra virgin olive oil diluted by n-hexane with NBS was studied.
Opposed to elemental bromine, NBS supplies an electrophilic bromine atom in a safe manner.
Efficiency of the reaction was highly dependent on the specific solvent composition.
Compared to epoxidation, the observed hump after bromohydrin reaction originating from
residual polyunsaturates was significantly higher. A comprehensive interpretation of the
results can be found in the supporting information.

Furthermore, selectivity of the bromohydrin reaction proved to be insufficient. During

derivatization of an EPA-PAH standard, it was observed that several PAHs were either
brominated (as verified by shifted retention times and the bromine pattern in the MS spectra)

9
or absent in the MOAH fraction (due to oxidation). Monocyclic aromatic hydrocarbons were
not affected.

Thus, as long as the substances were only brominated, but not further oxidized, recovery was
quantitative due to the quasi-unity response of the FID. Indeed, MOAH recovery for
sunflower and olive oil spiked (at 24.5 mg/kg of MOAH) with a lubricating oil was found to
be quantitative indicating that alkylated aromatic compounds could be retrieved.

Nevertheless, loss of polycyclic MOAH and possible misinterpretation of residual

polyunsaturates as MOAH do not make the reaction the first choice as cleanup step. As add-
on procedure for persistent polyunsaturates hindering a successful MOAH quantitation,
bromohydrin reaction could still be a valuable tool. For instance, analysis of spices containing
large amounts of essential oils, e.g. cardamom, can be troublesome [28].

3.2 Optimization of reaction conditions for epoxidation

The use of dichloromethane as solvent for epoxidation showed two problems for the removal
of polyunsaturates in foodstuffs. On the one hand, sub-ambient cooling was necessary to
sufficiently slow down the reaction for acceptable recovery of certain MOAH. On the other
hand, initial dissolution of the foodstuff in dichloromethane and subsequent solvent exchange
to n-hexane were necessary to allow subjection of the sample to LC-GC-FID. Solvent
evaporation is a time-consuming step always involving the risk of losing volatile compounds.
Both inadequacies could be overcome by switching from dichloromethane to another solvent.

Epoxidation of an olefin by mCPBA passes a concerted butterfly-shaped transition state pre-

orientated by mCPBA due to intramolecular hydrogen bonding [29]. Typically, non- or
moderately-polar solvents are used for epoxidation reactions [30]. An important reason for
this is that these solvents do not disturb the formation of the transition state. Accordingly,
solvents, such as benzene, chloroform, or dichloromethane, show the highest reaction
velocities in comparison to other solvents [31]. Polar and especially protic solvents, e.g.
ethanol, however, retard the reaction considerably as they interact with the mCPBA by
intermolecular hydrogen bonding.

Epoxidation in pure n-hexane was inefficient as solubility of mCPBA was limited and
therefore higher solvent amounts would have been necessary. Due to this fact, ethanol was
added to the reaction as it offered several advantages.

10
Firstly, solubility (>100 mg/100 µL) and stability of mCPBA in ethanol proved to be well
suited for an automated approach. Secondly, ethanol and n-hexane were miscible in any
proportion. Thirdly, removal of ethanol from the n-hexanic sample was possible by addition
of water. Finally, the reaction kinetics could be adjusted by tweaking the amount of ethanol in
the reaction mixture. Consequently, sub-ambient cooling was no longer necessary as ethanol
retarded the reaction sufficiently and provided therefore a higher selectivity and robustness
even at room temperature. This represents an important step forward for routine
environments.

Several ratios of n-hexane/ethanol and mCPBA were tested for the epoxidation of extra virgin
olive oil. As seen in Fig. 2, the removal of polyunsaturates was in fact dependent on the
amount of ethanol. The best results were found with an n-hexane/ethanol ratio of 7/5 (v/v) and
100 mg of mCPBA. Five hundred microliters of ethanolic mCPBA (200 mg/mL) were added
to 300 mg of edible oil or fat dissolved in 700 µL of n-hexane. The reaction was finished after
15 min at room temperature.

Spiked extra virgin olive (squalene), palm (carotenes), and sunflower oils (sterenes) were
analyzed to check if the most common polyunsaturates could be removed. In Fig. 3, LC-GC-
FID chromatograms of these oils after epoxidation are shown.

In the case of olive oil, almost no distinct peaks could be detected after epoxidation. No traces
of residual squalene were found. For sunflower and palm oil, however, a few sharp peaks
could be observed. These signals originate from polyunsaturates resisting epoxidation for the
chosen reaction conditions [26]. They were observed to the same extent also with the
established epoxidation procedure (data not shown). As they do not interfere with the
quantitation of the underlying MOAH hump, no adaptions of the reaction conditions were
necessary. Nevertheless, by increasing of the mCPBA amount, reaction time or temperature,
epoxidation of persistent polyunsaturates can be enforced (in exchange for the recovery of
polycyclic MOAH).

Further information regarding epoxidation kinetics can be found in the supporting

information.

11
Besides varying reaction conditions, another reason for insufficient robustness of epoxidation
in routine environments could be the missing quenching step with a reducing agent. The
established method includes an improper sodium carbonate washing step. This step does not
remove excess mCPBA from the dichloromethanic phase as solubility in water is limited.
Even during evaporation of dichloromethane, mCPBA is able to react with the sample.
Although solubility of mCPBA in n-hexane is low (1.4 mg/100 µL), a continuation of the
reaction would still be possible during sample storage on the LC-GC instrument.

Application of a peroxide test stripe after finished reaction of mCPBA with sunflower and
olive oils did not show traces of left peracid in the solution supporting the hypothesis that
excess mCPBA is consumed by unsaturated fatty acids [14]. In absence of a matrix, however,
high residual amounts of peracid were detected. A carbonate wash did not reduce these
amounts significantly. While the established method uses blank edible oil as buffering agent,
substitution of sodium carbonate for sodium thiosulfate as quenching reagent solved this issue
in a simple manner.

3.3 Comparison of automated and established epoxidation

The suitability of automated epoxidation was examined by derivatization of an EPA-PAH
standard (3 mg/kg) with and without addition of edible oil. Although reaction of PAHs with
mCPBA is much slower compared to epoxidation of olefins, oxidation to the corresponding
quinones and phthalic acids were reported [32]. Contrary to the literature, in which PAH
losses were reported in absence of edible oil, the results provided here show that PAH
oxidation can be prevented by retarded reaction conditions, appropriate quenching and
automated sample handling with exact time control [14]. Recoveries ranged from 92 to 104 %
regardless whether edible oil was present or not.

DBT is known to be oxidized to the corresponding sulfoxide and sulfone by mCPBA and is
therefore well suited as indicator for worst-case MOAH losses [14, 32-33]. Derivatization of a
DBT standard caused losses of approximately 80 % in absence of a matrix. In literature,
complete loss of DBT was observed [14]. In presence of sunflower or olive oil, the losses
were reduced to approximately 45 % which was slightly worse compared to established
epoxidation (see Table 1). This reflects the higher amount of mCPBA used here opposed to
established epoxidation (100 mg vs. 30 mg for 300 mg of edible oil).

12
Higher amounts were anyhow chosen and routinely used, because the retarded reaction
conditions allowed these amounts without adverse effects which was not the case for
established epoxidation [14]. Individual foodstuffs, such as palm fatty acid distillates, needed
higher amounts of mCPBA for a sufficient cleanup (data not shown).

Finally, sunflower and olive oil spiked (at 24.5 mg/kg of MOAH) with a lubricating oil were
epoxidized automatically and compared to the established method. In Table 1, all obtained
results for automated and established epoxidation are listed.

3.4 Validation of the automated epoxidation approach

Repeatability and recovery were determined by six individual workups and injections of
spiked olive and sunflower oils. According to Horwitz, the predicted relative standard
deviation under repeatability conditions for a spiking of 24.5 mg/kg is fixed at 6.6 % [34].
The obtained repeatability was better than 1.5 % for both oils and complied therefore with the
Horwitz requirements. The recovery ranged from 95 to 102 % including the internal
standards.

Since no certified reference materials were available for the determination of MOAH in edible
oils and fats, collaborative trial material was used instead. Three oil samples were analyzed in
duplicate on three successive days to get an impression of the reproducibility and trueness of
the automated epoxidation with subsequent LC-GC-FID analysis (see Table 2 –
chromatograms found in Fig. 3, S-1, S-2).

In general, good agreement between quantitative collaborative trial mean values and
automated epoxidation results was observed. However, for the extra virgin olive oil a
reproducibility exceeding the predicted Horwitz limit was observed in own measurements.
This was related to the low quantified amount, which was hardly distinguishable from the
chromatogram baseline. As can be derived from the collaborative trial results, the participants
struggled with the same problem. Preceding sample enrichment is needed to cope with this
problem as was already suggested in literature [35]. Without it, contaminations higher than
approximately 3 mg/kg can be determined (in the chosen foodstuffs) with sufficient precision.

For the refined olive pomace oil, significantly higher amounts were quantified in own
experiments. This was clearly related to the high-boiling mass distribution of the MOAH
contamination in this case (as shown in Fig. 4). Contaminations exceeding elution
temperatures of 350 °C were detected. The use of metal GC columns allowed oven

13
temperatures of 400 °C beneficial for the elution of these high boiling compounds. It is
doubtful if all participants of the collaborative trial were aware of this problem as was
confirmed by the collaborative trial organizers in personal communication.

As can be seen in Fig. 4, the MOAH hump after epoxidation is clearly defined although some
peaks on top of the hump resisted epoxidation. No adaptions of the epoxidation conditions
were necessary, because the distinct signal could be precisely subtracted. However, in case of
smaller MOAH contaminations, harsher epoxidation conditions could become mandatory.

4 Conclusion
At the moment, epoxidation proved to be the method of choice for removal of polyunsaturates
from edible oils and fats. However, understanding and optimization of the reaction conditions
is of utmost importance for precise and valid results. Automation is therefore beneficial.

Substitution of the reaction solvent made the use of sub-ambient temperatures obsolete.
Solvent evaporation and sample reconstitution were no longer necessary either. Polycyclic
MOAH and PAHs could be quantitatively recovered in absence of a matrix, which is related
to retarded reaction kinetics and the use of an efficient quenching step. This represents an
important achievement in comparison to the established method. Quantitative results for
collaborative trial samples verified the trueness of the proposed method making the method
amenable to the analysis of edible oils and fats in routine environments.

While hydroboration turned out to be a dead end, bromohydrin reaction could be an

interesting additional tool for the cleanup of complex foodstuffs, such as spices containing
large amounts of essential oils, for which epoxidation is inefficient. As single cleanup step,
the bromohydrin reaction showed insufficient results and substantial losses of non-alkylated
PAHs. Unconventional reaction conditions proved to be the most successful, which was not
published in literature previously. Investigation of other factors influencing the reaction
kinetics could be of interest.

Exploration of new possibilities for removal of polyunsaturates need to be further pursued.

Evaluation of the suitability of the here developed method for the removal of olefins from the
MOSH fraction in other foodstuffs will be another important aspect.

14
Acknowledgements
Florence Lacoste and Loïc Leitner from ITERG are thanked for the supply of samples.

References
[1] European Food Safety Authority, Scientific opinion on mineral oil hydrocarbons in
food, EFSA J. 10(6) (2012) 2704.
[2] H.P. Neukom, K. Grob, M. Biedermann, A. Noti, Atmos. Environ. 36 (2002) 4839–
4847.
[3] M. Biedermann, K. Grob, J. Chromatogr. A 1255 (2012) 76–99.
[4] M. Biedermann, K. Grob, J. Chromatogr. A 1375 (2015) 136–153.
[5] S. Moret, L. Barp, G. Purcaro, L.S. Conte, J. Chromatogr. A 1243 (2012) 1–5.
[6] K. Grob, Eur. J. Lipid Sci. Technol. 110 (2008) 979–981.
[7] L. Barp, C. Kornauth, T. Wuerger, M. Rudas, M. Biedermann, A. Reiner, N. Concin,
K. Grob, Food Chem. Toxicol. 72 (2014) 312–321.
[8] W. Xue, D. Warshawsky, Toxicol. Appl. Pharm. 206(1) (2005) 73–93.
[9] P. Tarnow, C. Hutzler, S. Grabiger, K. Schön, T. Tralau, A. Luch, PloS One. 11(1)
(2016) 1–15.
[10] German Federal Institute for risk assessment (BfR), Stellungnahme 14 (2015) 1–19.
[11] Joint FAO/WHO Expert Committee on Food Additives (JECFA), Expert
Committee on Food Additives, 59th report, p. 11, WHO Technical Report Series
913 (2002) http://apps.who.int/iris/bitstream/10665/42601/1/WHO_TRS_913.pdf.
Accessed 26. April 2017.
[12] Federal Ministry of Food and Agriculture, Entwurf einer Zweiundzwanzigsten
Verordnung zur Änderung der Bedarfsgegenständeverordnung (BedGgstV).
24.7.2014.
http://www.bmel.de/SharedDocs/Downloads/Service/Rechtsgrundlagen/Entwuerfe/Ent
wurf22VerordnungBedarfsgegnstaende.pdf?__blob=publicationFile. Accessed 26.
April 2017.
[13] Federal Ministry of Food and Agriculture, Entwurf einer Zweiundzwanzigsten
Verordnung zur Änderung der Bedarfsgegenständeverordnung (BedGgstV).
24.2.2017.
http://www.bmel.de/SharedDocs/Downloads/Ernaehrung/Rueckstaende/MineraloelVO

15
 _Entwurf.pdf;jsessionid=F636D9DBDB7DF40D0E02A54932ECE219.2_cid358?__bl
ob=publicationFile. Accessed 26. April 2017.
[14] M. Biedermann, K. Fiselier, K. Grob, J. Agric. Food Chem. 57 (2009) 8711–8721.
[15] K. Grob, M. Biedermann, A. Artho, J. Egli, Zeitschrift für Lebensmitteluntersuchung
und Forschung. 193(3) (1991) 213–219.
[16] K. Grob, I. Kaelin, A. Artho, J. High Resolut. Chromatogr. 14 (1991) 373–376.
[17] M. Zoccali, L. Barp, M. Beccaria, D. Sciarrone, G. Purcaro, L. Mondello, J. Sep. Sci.
39(3) (2015) 623–631.
[18] M. Lommatzsch, M. Biedermann, T.J. Simat, K. Grob, J. Chromatogr. A 1402 (2015)
94-101.
[19] F. Ulberth, E. Achs, J. Chromatogr. A. 504 (1990) 202–206.
[20] C. Wagner, H.P. Neukom, V. Galetti, K. Grob, Mitteilungen aus
Lebensmitteluntersuchung und Hygiene 92(3) (2001) 231–249.
[21] C. Mariani, Stazione Sperimentale per le Industrie degli Oli e Grassi. (2009) (private
communication mentioned in [14]).
[22] R.D. Bach, C. Canepa, J.E. Winter, P.E. Blanchette, J. Org. Chem. 62(15) (1997)
5191–5197.
[23] F. Fringuelli, F. Pizzo, R. Germani, G. Savelli, Org. Prep. Proc. Int. 21(6) (1989)
757–761.
[24] F. Lacoste. Mineral oil in vegetable oils: Laboratory and Industrial Issues,
http://www.iterg.com/IMG/pdf/dgf_lacoste_mineral_oil__2016.03.10_vf.pdf.
Accessed 26. April 2017.
[25] M. Biedermann, K. Grob, J. Chromatogr. A. 1216 (2009) 8652–8658.
[26] C. Kim, T.G. Traylor, C.L. Perrin, J. Am. Chem. Soc. 120(37) (1998) 9513-9516.
[27] K.J. Shea, J.S. Kim, J. Am. Chem. Soc. 114(8) (1992) 3044-3051.
[28] M. Viuda-Martos, Y. Ruíz-Navajas, J. Fernández-López, J.A. Pérez-Álvarez, Acta
Chimica Slovenica 54(4) (2007) 921.
[29] P.D. Bartlett, Rec. Chem. Progr. 11 (1950) 47-51.
[30] N.N. Schwartz, J.H. Blumbergs, J. Org. Chem. 29(7) (1964) 1976-1979.
[31] R. Curci, R. DiPrete, J.O. Edwards, G. Modena, J. Org. Chem. 35(3) (1970) 740-745.
[32] J.T. Andersson, Int. J. Environ. Anal. Chem. 48(1) (1992) 1-15.
[33] J.M. McIntosh, R.A. Sieler, Can. J. Chem. 56(2) (1978) 226–231.
[34] W. Horwitz, R. Albert, J. AOAC Int. 89 (2006) 1095–1109.
[35] M. Zurfluh, M. Biedermann, K. Grob, J. Verbrauch. Lebensm. 9(1) (2014) 61-69.

16
Fig. 1. LC-GC-FID chromatograms of the MOAH fractions of extra virgin olive oil (top),
palm oil (middle), and sunflower oil (bottom). Signals originating from polyunsaturates can
either pretend the presence of MOAH or mask small MOAH amounts. (Due to the high
squalene content in olive oil, massive peak distortion of the ISTD peaks is observed.)

17
 100 0.5 mL EtOH
90 1 mL EtOH

80 3 mL EtOH
5 mL EtOH
70
Hump area [%]

60

50

40

30

20

10

0
0 20 40 60 80 100 120 140
mCPBA amount [mg]

Fig. 2. Residual hump area after epoxidation of extra virgin olive oil with various amounts of
ethanol and mCPBA (An area of 100 % corresponds to the total hump area prior to
epoxidation – Reaction conditions: 300 mg of olive oil, 0.7 mL of n-hexane, 15 min at room
temperature).

Fig. 3. LC-GC-FID chromatograms of palm (top), sunflower (middle), and extra virgin olive
oil (bottom) after epoxidation in n-hexane/ethanol. Samples were spiked at 24.5 mg/kg of
MOAH with a lubricating mineral oil. As can be seen for palm and sunflower oil, several
distinct peaks resist epoxidation under the chosen reaction conditions (see supporting
information for further information).

18
Fig. 4. LC-GC-FID chromatogram of the MOAH fraction of olive pomace oil from the
collaborative trial (70.6 mg/kg). Some peaks on top of the integrated hump resisted
epoxidation.

Table 1. Comparison of automated and established epoxidation for spiked samples

with and without edible oil as matrix

Recovery [%]

MOAHa DBTb EPA-PAHsc

Autom. Est. Autom. Est. Autom. Est.d

No matrix 95 107 20 <3 93–103 > 55
Sunflower oil 102 105 55 65 94–101 102
Extra virgin olive oil 101 99 56 65 92–104 95
All samples were measured in triplicate. Relative standard deviations were less than 3 %.
a
: BAM K009 lubricating oil MOAH spiking of 24.5 mg/kg
b
: Dibenzothiophene (DBT) spiking of 30 mg/kg
c
: EPA-PAH spiking of 3 mg/kg
d
: Results obtained from [14]

19
Table 2. Comparison of collaborative trial MOAH results and values obtained by
automated epoxidation

CTa MOAH Horwitz CT RSDR Automated RSDR

Mean value RSDR [%] [%]b epoxidation [%]c
[mg/kg] [mg/kg]
Extra virgin olive oil 1.7 14.7 49.5 1.5 18.2
Olive pomace oil 44.7 9.0 9 70.6 2.7
Palm oil 11.4 11.0 13.5 13.6 5.5
a
: Collaborative trial
b
: To eliminate the influence of the individual laboratories regarding differing
chromatogram integration etc., CT reproducibility was estimated from repeatability after
multiplying by an empirical factor of 1.5 [33].
c
: Based on the quantitative results in duplicate on three successive days (n = 6)

20