Supporting information to

Polyfluorinated Surfactants (PFS) in Paper and Board Coatings
for Food Packaging

By : Xenia Trier1, 2, Kit Granby2, Jan H. Christensen1
1 The Faculty of Life Sciences, University of Copenhagen, Department of Basic Sciences and Environment, Thorvaldsensvej 40, DK-1871 Frederiksberg, Denmark.
2 Technical University of Denmark, The National Food Institute, Mørkhøj Bygade 19, DK-2860 Denmark.

2 Materials and Methods

2.1 Chemicals and materials

2.1.1 Chemical standards: Standards were from Sigma Aldrich Danmark A/S, Brøndby, Denmark (perfluorooctane sulphonate (PFOS) T-(technical, i.e. linear and branched) and perfluorooctane sulphonamide (PFOSA)), and diPAPS (4:2/4:2, 6:2/6:2, 8:2/8:2, 10:2/10:2 diPAPS) standards were synthesized and donated by Scott Mabury’s laboratory, University of Toronto, Canada. The industrial blends Zonyl BA-L, FSE, NF, UR, FSA, TM, FSN and FSO are products of Dupont (Dupont 2010) and were obtained from Sigma Aldrich, though Zonyl NF, together with FC 807 (3M) and Lodyne 2000 was donated by the Danish Veterinary and Food Administration, Søborg Denmark. FF-807, FX-8, FT-248, FF-09 and FC-10 were donated by Wuhan Fengfan Chemical Co. Ltd., China [Wuhan 2010]. Table 1 in the main article gives an overview of the structures, acronyms, chemical composition, their uses in relation to food contact materials and from where the chemicals were obtained.

Other chemicals: Methanol was from Fisher Scientific UK Ltd. (99.9%, LC-MS grade, Leischershire, UK) and deionised, distilled water (MilliQ water, MQ) was produced in-house to 18.2 MW, and glass distilled water as well. Ethanol (99.8%, pro analysis) was from Merckx, Germany and was diluted to 95/5 v/v with water. Ammonia (NH4OH, >25% purity in water) was from Fluka Analytical, Steinheim, Germany (bought through Sigma-Aldrich Denmark A/S). Ammonia formiate (NH4HCO2, 99%) from Alfa Aesar GmbH & Co (bought through VWR-Bie & Berntsen, Herlev, Denmark). Formic acid (98%, LC-MS grade) from Fluka Analytical. Sodium hydroxide (98%, NaOH) was from Riedel-de Häen (bought through Sigma-Aldrich Denmark A/S).

Preparation of standard solutions: 15 mL polypropylene tubes without septa were used to store stock solutions (Wheaton Industries Inc., Millville, New Jersey, USA). Separate stock solutions of 5 mg mL-1 were prepared in methanol. Working solutions of 0.025-5 µg mL-1 were prepared in mixtures of methanol and MQ (95/5, v/v). HPLC vials were of glass (12 mm × 32 mm) with PTFE/silicone septa turned around, so the silicone part was in contact with the sample (Waters, 10-425).

Mobile phases: The pHs of the mobile phases were adjusted to pH 9.7 with ammonia hydroxide into to A: 95% water/5% methanol (40 µL NH4OH) and B: 100% methanol (300µL NH4OH).

Mass calibration of mass spectrometer: The external Q-TOF calibrants were sodium formate (0.010 mg mL-1 in 80% aqueous methanol) and sodium iodide (2 mg mL-1 in 50% aqueous isopropanol).

2.2 Sampling and sample preparation

Both extracts, attempting full extraction, and migrates, resembling migration to food, were made. After removal of the food product, the paper and boards were rinsed for salts with distilled, deionised water. Unused food contact material that had not been in contact was chosen if possible. For the identification study, where extractions were made, 50 cm2 was taken from two samples (100 cm2 in total) and placed in 50 mL polypropylene centrifuge screw-cap tubes (Saarstedt, dimensions 115 mm × 28 mm). 40 mL of 95% ethanol, preheated to 60 °C, was poured over the subsamples (full immersion). The tubes were capped and ultrasonicated (Bransonic 1510E-DTH (70 W, 42 kHz), Branson Ultrasonic corporation, Danbury, CT) for 2 hours at 60 °C. For the migration experiments, composite sampling was performed to limit the variation from sample in-homogeneity: 10 sub samples of each 1 cm2 were taken from each of 10 samples, and the total area of 100 cm2. The same procedure was used as above, but the samples for migration testing were not ultrasonicated but instead placed in a thermostated oven (60 ± 1ºC , Termaks cabinet, series 8000, Denmark). Immediately after ended extraction or migration, the food simulant was decanted into new tubes that were centrifuged (9000 RPM for 10 min) and the supernatants were filtered through 0.2 µm nylon Minisart filters (polypropylene casing, from Sartorious, Goettingen, Germany) and transferred into HPLC vials. The clean-up was limited to centrifugation and filtering, in an attempt to make the analysis non-discriminative and not to loose material. The remaining liquid was transferred into new 15 mL centrifuge tubes, and kept refrigerated for future use. Some extracts and migrates were cloudy, with a gel-like structure, that could not be spun down. The final procedure was therefore to leave the extracts/migrates to settle overnight in a refrigerator (5 ºC). Before use they, as well as standard solutions were given ca. 2 min of ultrasonification to release adsorbed PFC from the walls, and then the samples were filtered.

2.3 Instrumental analysis for identification study of industrial blends and FCM extracts

UPLC-ESI- -QTOF MS was used for the identification study, while an HPLC-ESI- -QqQ MS was used for the migration study, as the linearity of this instrument was better.

2.3.1 ESI--QTOF mass spectrometry and software

A Micromass QTOF Ultima Global with a Maldi / Quattro Ultima cone (Micromass, Manchester, UK), was used in the negative electrospray ionisation (ESI-) mode. Temperatures: source 120 °C, desolvation 300 °C. Voltages: capillary -3 kV, cone voltage 90 V (Ultima Global cone) or 20 V (Quattro Ultima cone), Collision energy: 6 V (MS) and 10-70 V (MSMS). Desolvation gas flow 700 L hr-1, cone gas flow 50 L hr-1 of N2. Collision gas: Argon. External calibration immediately before the analysis: m/z 181 – 1472 Da, with 12-20 sodium formate clusters, polynomial order: 3-5 (RMS residual 1-3 ppm). In the MSMS mode, the quadrupole resolution was m/z ± 1.5 Da. MassLynx v. 4.1 (Micromass) was used for data processing and calculation of exact theoretical isotopic masses. M/zs from 75 to 1500 Da were recorded. Theoretical m/z’s are given with three decimal places, and experimental m/z values are given with one to two decimal places. Nominal masses are used when readability is of importance. Table S1 below gives a list of precursor and product ions for some industrial blends, in intermediates and in popcorn bag migrates, which can be used to chose ions from when setting up MSMS experiments.

Table S1 List of precursor and product ions for a number of industrial blends, intermediates and popcorn bag migrates measured from m/z 75-1500.

* m/z values of the most intense precursor ion of the homologue series is listed in bold (MS conditions).

** m/z values of product ions are listed in order of decreasing intensity under the applied MSMS conditions.

à Suggestion; as the molecular structure was not found (patent not found)

2.3.2 UPLC method

A Waters UPLC Acquity system (Waters Corp., Milford, MA, USA) was used. Fully resolved chromatographic peaks are essential to obtain pure MS spectra for the individual PFSs and of possible isomers. We used a C18 BEH-shield column (Waters, 2.1 mm i.d., 1.7 µm particle size, 150 mm). Injection volumes were 1-7.5 µL. To avoid cross contamination PEEK tubing was changed regularly, and no column pre-filter was used to prevent carry over (Voegt et al. 2006). Binary solvent system: Mobile phase A: water/methanol (95/5, v/v), B: methanol, with each mobile phases adjusted to pH 9.7 with ammoniahydoxide. Optimised UPLC gradient method: Initial composition 95% A, 0-3 min linear to 40% A, 3-24 min linear to 5% A, 24-31 min linear to 2% A, 31-33 min curved to initial composition, isocratic from 33-35 min. Flow: 0.28 mL min-1. The column temperature was 45 ºC, to decrease viscosity to keep the column pressure acceptable (<860 bar).

2.3.3. Quantification and quality assurance for UPLC-QTOF method

Identification study: Solvent blanks were run in between all samples, and the instrument was regularly cleaned. LODs were determined for the 15 min program, by using repeated injections of standards close to the detection limit (level 1) to determine the standard deviation from. Because the calibration curves were not linear near the LOD, only the concentration at the low level was used to determine the response factor, instead of the response factor for the calibration curve: LOD = (3 · std dev(peak area, level 1) + blank) · [average(peak area, level 1) · (concentration, level 1) -1]-1

The method performance data are given in Table S2.

Table S2: Method performance parameters of the UPLC-ESI--QTOF method for diPAPS, PFOA and PFOS

2.4 Instrumental analysis for the migration study

2.4.1ESI--QqQ mass spectrometry and software

A Micromass Quattro Ultima mass spectrometer (Micromass, Manchester, UK), was used in the negative electrospray ionisation (ESI-) mode. Temperatures: source 120 °C, desolvation 400 °C. Voltages: capillary -3 kV, cone voltage 60 V. N2 desolvation gas flow 630 L hr-1, N2 nebuliser gas flow: maximum, N2 cone gas flow 100 L hr-1. Collision gas: Argon (2.3 · 10-3 mbar). The precursor-to-product ion transitions and the collision energies for the MSMS experiments are shown in Table S3.

2.4.2 HPLC

A high pressure liquid chromatograph (HPLC, Waters 2695E) was used with an X-bridge TM C18 column (2.1 150mm 3.5 µm particle size) was used with mobile phases A: MQ water with methanol (95/5, v/v) and B: methanol, both adjusted to pH 9.7 with ammonia. The linear gradient elution programme was: Initial composition 50% A, 0-4 min to 30% A, 4-10 min to 25% A, 10-13 min to 20% A, 13-15 min to 10 % A, 15-18 min to 5 % A, 18-19 min to 2 % A, 19-20 min returned to the initial composition, and 20-22 min held constant. Flow: 0.3 mL/min. The retention times are given in Table S3.

Table S3: LC retention times and MS parameters for the HPLC-ESI-- QqQ MS method for diPAPS and S-diPAPS

2.4.3. Quantification and quality assurance of the migration study
The calculations were made by comparing the average peak areas of xx89 series or xx21 series for the samples vs. those of the xx89 series in an industrial standard (Zonyl NF) of a known concentration, with a “purity” of 19.5% of the sum of PFS. It was assumed that 1. only the diPAPS contributed to the 19.5% PFSs and that the response factor of all the peaks were equal, meaning that the areas could be taken as a measure of concentration. 2. that the intensities of the individual peaks in the homologue series was the same for the industrial blends and the migrates (which it is not necessarily true because some diPAPS might be better retained by the paper than others) 3. that the xx21 series had the same response factors as the xx89 series 4. that the interference from the matrix was equal in the industrial blends and the migrates. Conversion of units from mg L-1 food simulant to food was done using CEN guidelines (CEN 2002): The amount of PFS that had migrated per area was calculated by conversion from mg L-1 → mg dm-2 → mg kg-1 food, by relating the actual packaging area (eg. 10 dm2 for a popcorn bag) with the actual weight of food (100 g popcorn), so 1 dm2 ~ 10 g food. Blanks were run between the samples, and we used standard addition with Zonyl NF to check if there was matrix interference, e.g. by ion suppression. As this was not the case we used external calibration at the beginning and end of a series. We did not use internal standards, as isotopically labelled diPAPS standards were not available, and e.g. 13C PFOA is too different to respond to variations in extractions and instrumental conditions in the same way as diPAPS.

The repeatability of the sampling was checked by analysis of 2´5 paper subsamples, taken from 5 different popcorn bags, and an average of the RSD was 16-19% was found for the ions with m/z 789-889-989. The repeatability of the LC-MS analyses was checked on 2x5 times injection of the industrial standard Z-NF (at the 0.025 mg L-1 level), and the RSD was 8-11%. Procedural blanks were run in each sequence, and blank signals were subtracted before quantifying. The LOD of the MSMS analysis was 0.015-0.022 mg L-1 (calculated as 3*RSD+blank), and using the response factor of the calibration curves. Two examples of a-fragmentations of the PFS precursor ions (m/z 526 and 570), forming the product ions (m/z 419 and 526) are shown in Figure S1, together with the major carbon backbone fragments (m/z 419, 269, 219, 169).

Fig S1 a-fragmentations next to the hetero atom were common, together with backbone fragmentations giving the series of m/z 169, 219, 269 etc. separated by 50 Da corresponding to CF2 units. Here fragmentations next to the sulphur atoms is shown for two peaks present in the industrial blend FF10, whose EIC is shown in Fig. 1. When m/z peaks are topped by arrow heads, they are depicted at reduced intensity.

1