Chemicals and materials
All chemicals were purchased from either Sigma-Aldrich or VWR and where of at least analytical grade. For the mobile-phase, LC-MS CHROMASOLV (Fluka, Sigma-Aldrich) grade water and acetonitrile (ACN) were used. Deionized water was generated by a Milli-Q system (Millipore, Belgium). Formic acid ∼98 % (FA), and ammonium formate (NH4FA), as LC-MS grade mobile-phase additives, were purchased from Fluka. An equimolar mix of FA and NH4FA of pH 3.7 was prepared as follows: 4.6 g FA and 6.3 g NH4FA were mixed and diluted with water to 34 mL. This NH4FA pH 3.7 solution was used as additive for the buffered mobile phase and was equivalent to 10 % FA (v/v).
The certified reference material ERM-AC057 (AFB1 in acetonitrile) with a certified mass fraction w = 3.79 μg/kg and an expanded measurement uncertainty (k = 2) of 0.11 μg/kg (the combined uncertainty contained contributors from purity assessment, stability testing, and certification) was obtained from IRMM. The spike, isotopologue 13C17-AFB1 in ACN (c = 0.502 μg/mL), was purchased from Romer Labs-Biopure (Tulln, Austria). All subsequent dilutions of AFB1 and the spike were prepared gravimetrically in neat ACN.
PT materials investigated were a maize-based feed material, a neat maize material, and a cereal-based baby food material, all naturally contaminated with aflatoxins and used in an EU-RL mycotoxin PT in 2011. All the above PT materials were packaged as ground powders and of each material three test units were selected at random for the investigation. Analyte-free materials matching the PT materials were from the material pool of the EU-RL for mycotoxins. Absence of analyte signal was verified with the method described here.
Instrumentation
The 2D LC-LC system consisted of an Accela low-pressure gradient solvent delivery unit and an Accela auto liquid sampler (ALS) as LC1 (Thermo Scientific, Belgium). LC2 was a high-pressure gradient system made up of two LC-20AD pumps with a microvolume mixer and a DGU-20A degasser (Shimadzu Benelux, The Netherlands). The MS was a TSQ Quantum Ultra triple-quadrupole mass spectrometer with a HESI 2 ion source (Thermo Scientific, Belgium).
First-dimension separation was afforded by a Supelco Ascentis C18 column (50 × 2.1 mm, 3-μm particle size) with an Ascentis Express C18 guard column (5 × 2.1 mm, 2.7 μm; Sigma-Aldrich, Germany) at isocratic conditions of 38 % B at 200 μL/min and 40 °C. To prevent the build-up of late eluting substances, a 1.5-min step-up to 90 % B after elution of the analyte was included. Mobil phase A was water/FA (999/1, v/v) and B was ACN/FA (999/1, v/v).
A Supelco Ascentis phenyl column (50 × 2.1 mm, 3 μm) at isocratic conditions of 53 % B at 200 μL/min and room temperature was used for the second-dimension separation. Here, a step-up to 100 % B for 1.5 min was also included in the gradient. Second-dimension mobile phase A was water/NH4FA pH 3.7 (999/1, v/v) and B ACN/water/NH4FA pH 3.7 (900/99/1, v/v/v). The addition of NH4FA led to the suppression of [AFB1 + Na]+ and increase in [AFB1 + H]+.
The integrated six-port, two-position divert valve of the TSQ Quantum Ultra was used for the transfer of the heart-cut of the first-dimension separation to the second-dimension column. To achieve this, a 100-μL loop was used to trap the analyte eluting from the first-dimension column. During preliminary tests, the switching time was determined by connecting the outlet of the loop directly to the MS. The retention time of the front of the analyte peak minus the delay caused by the internal volume of the ESI probe and the tubing is the run time at which the analyte peak fills the loop. After switching the content of the loop was loaded in reverse onto the phenyl column which was installed between the valve and the ion source.
For 1D separations, the Shimadzu solvent delivery system was connected to the Accela ALS and a Supelco Ascentis Express C18 column (75 × 2.1 mm, 2.7 μm) with an Ascentis Express C18 guard column (5 × 2.1 mm, 2.7 μm). Mobile phase A and B were identical with the second-dimension conditions above. Separation was performed isocratically at 35 % B, 300 μL/min, and 40 °C.
The MS ion source settings are listed in Table 1. The MS analyzer was used in selected reaction monitoring (SRM) mode with argon as collision gas at 0.2 Pa (1.5 mTorr) and the monitored ions are listed in Table 2. Scan cycle time was set to 0.7 s for the seven transitions measured to record >30 scans per peak.
Table 1 Ion source settings of the mass spectrometer; the gas flows are in arbitrary flow units (afu.)
Table 2 Ions monitored by MS during SRM: the protonated species [M+H]+ was selected as precursor at unit resolution
Blend preparation
To minimize potential within-unit inhomogeneities, the entirety of each test unit (ca. 30 g) of each material was additionally comminuted/homogenized for 15 min with a Mortar Grinder with a hard porcelain grinding set (Retsch, Haan, Germany).
The sample blend (SB) consisted of 2 g test material weighed into a 50-mL conical screw-cap polypropylene centrifuge tube (VWR, Belgium) to which 4 mL of water was added. After the material was fully suspended by vortex mixing, the spike was weighed in. The amount of spike was chosen such that the observed isotope ratio in the SB (R
′
B
) of the total ion current (TIC) of analyte ion over spike ion would be near unity.
Calibration blends (CB) consisted of 2 g of a matched analyte-free material. After suspending in 4 mL of water, the same amount of spike as in the SB was added. Then, AFB1 was weighed in such that the observed isotope ratio in the CB (R
′
Bc
) would also be near unity. All weighing was performed with an analytical balance of readability d = 0.01 mg (Sartorius ME235S, Belgium) and weights were recorded with full precision. The balance is recertified annually by the manufacturer and checked daily with a 1-g weight of Class E2 with full traceability to the SI unit.
From each of the three units of the baby food and maize test materials, two SBs were prepared for a total of six SBs per material. Of the feed test material, one unit was used up for preliminary tests and of the remaining two units, three SBs each were prepared for a total of six SBs. One matching CB was prepared per test unit, .i.e., three CBs for baby food, three CBs for maize, and two CBs for feed.
Preparation of injection solutions and measurements
For extraction, 16 mL ACN were added to the blends. For the feed and maize material, this was done in a single addition; while for the baby food, it was done in four portions with intermediate vortex mixing to prevent the sudden precipitation of the milk protein and the resulting loss of analyte in the precipitate. The blends were then agitated on an orbital shaker (KS 260 control, IKA-Werke, Germany) for 30 min and centrifuged (Centrifuge 5810R, Eppendorf, Germany) at 3,200×g for 10 min.
Of the clear supernatant, 4 mL for the baby food and 2 mL for the maize were transferred into silanized glass vials (Supelco 45 × 15 mm, Sigma-Aldrich). After evaporation to dryness under a stream of N2 at 70 °C, the dry residues were reconstituted with 120 μL of ACN, vortex mixed, and then diluted with additional 280 μL of water. For the feed, because of the higher contamination, 300 μL of clear supernatant were diluted by addition of 500 μL of water.
Of the reconstituted and diluted solutions, 20 μL were injected (“no waste mode”) without any further treatment. For the determination of matrix effects, the same procedure as above was performed with the feed material but the spike was added after the extraction into an aliquot of clear supernatant.
The measurement batches began with several blank runs until the instrument was fully equilibrated, especially the ion source temperatures. The next injection was a SB followed by a corresponding CB. This pair was repeated ten times and followed by a blank run again. This sequence of ten SB/CB pairs and a blank run was repeated for every SB prepared. Always, SBs of the same test unit shared the respective CB for that test unit. Isotope ratios in the SBs (R
′
B
) and CBs (R
′
Bc
) were calculated from the TIC of analyte ion over TIC of spike ion.
Calculations
Since the following assumptions were met, the simplified version [26] of the model equation (Eq. 2) for double IDMS could be used to calculate the mass fraction w
X,i
of analyte in the ith SB: occurrence of the spike ion signal in the native test materials and in the reference material of the native analyte was negligible; occurrence of the analyte ion signal in the spike material was negligible; “exact matching” was achieved.
$$ {w}_{X, i}={w}_Z\frac{m_{Y, i}{m}_{Z, i}}{m_{X, i}{m}_{Y c, i}}{\overline{R}}_i^{\prime } $$
(2)
where w
Z
= mass fraction of analyte in reference material, m
X,i
= mass of test material in ith SB, m
Y,i
= mass of spike added to ith SB, m
Z,i
= mass of reference material in ith CB, m
Yc,i
= mass of spike added to ith CB, and \( {\overline{R}}_i \) = mean of all measurements of R
′
B,ij
/R
′
Bc,ij
for the ith SB/CB pair with R
′
B,ij
= observed isotope ratio of the jth measurement of the ith SB and R
′
Bc,ij
= observed isotope ratio of the jth measurement of the ith CB.
The combined uncertainty of w
X,i
can then be expressed by Eq. 3 [25] as follows:
$$ {u}_{c, i}\left({w}_{X, i}\right)={w}_{X, i}\sqrt{{\left(\frac{u\left({w}_Z\right)}{w_Z}\right)}^2+{\left(\frac{u\left({m}_{Y, i}\right)}{m_{Y, i}}\right)}^2+{\left(\frac{u\left({m}_{X, i}\right)}{m_{X, i}}\right)}^2+{\left(\frac{u\left({m}_{Z, i}\right)}{m_{Z, i}}\right)}^2+\left(\frac{u\left({m}_{Y c, i}\right)}{m_{Y c, i}}\right)+{\left(\frac{u\left({\overline{R}}_i^{\prime}\right)}{{\overline{R}}_i^{\prime }}\right)}^2} $$
(3)
where u denotes the standard uncertainty of the respective term of Eq. 2, e.g., u(\( {\overline{R}}_i \)) is the standard error of the mean of the ten measured ratios R
′
B,ij
/R
′
Bc,ij
in the ith SB/CB pair.
The mass fraction w
T
of a test material is then calculated by Eq. 4 as follows:
$$ {w}_T={\overline{w}}_X{F}_X $$
(4)
where \( {\overline{w}}_X \) = mean of all six w
X,i
of one test material and F
X
= a factor of unity representing the mean of the relative combined uncertainties of w
X,i
of one test material. The combined uncertainty of w
T
is then expressed by Eq. 5 as follows:
$$ {u}_c\left({w}_T\right)={w}_T\sqrt{{\left(\frac{u\left({\overline{w}}_X\right)}{{\overline{w}}_X}\right)}^2+{\left(\frac{u\left({F}_X\right)}{F_X}\right)}^2} $$
(5)
where u(\( {\overline{w}}_X \)) = the standard error of the mean of \( {\overline{w}}_X \) and u(F
X
) = the mean of all u
c,i
(w
X,i
) / w
X,i
per test material.
All calculations were performed with R, a language and environment for statistical computing [27].