Skip to main content
NCBI home page
ACS AuthorChoice logoLink to ACS AuthorChoice
. 2023 Jun 1;95(23):8780–8788. doi: 10.1021/acs.analchem.2c05389

Combination of Structure Databases, In Silico Fragmentation, and MS/MS Libraries for Untargeted Screening of Non-Volatile Migrants from Recycled High-Density Polyethylene Milk Bottles

Qi-Zhi Su †,‡, Paula Vera †, Cristina Nerín †,*
PMCID: PMC10267890  PMID: 37262310

Abstract

graphic file with name ac2c05389_0004.jpg

Chemical contamination is one of the major obstacles for mechanical recycling of plastics. In this article, we built and open-sourced an in-house MS/MS library containing more than 500 plastic-related chemicals and developed mspcompiler, an R package, for the compilation of various libraries. We then proposed a workflow to process untargeted screening data acquired by liquid chromatography high-resolution mass spectrometry. These tools were subsequently employed to data originating from recycled high-density polyethylene (rHDPE) obtained from milk bottles. A total of 83 compounds were identified, with 66 easily annotated by making use of our in-house MS/MS libraries and the mspcompiler R package. In silico fragmentation combined with data obtained from gas chromatography–mass spectrometry and lists of chemicals related to plastics were used to identify those remaining unknown. A pseudo-multiple reaction monitoring method was also applied to sensitively target and screen the identified chemicals in the samples. Quantification results demonstrated that a good sorting of postconsumer materials and a better recycling technology may be necessary for food contact applications. Removal or reduction of non-volatile substances, such as octocrylene and 2-ethylhexyl-4-methoxycinnamate, is still challenging but vital for the safe use of rHDPE as food contact materials.

Recycling of food contact plastics attracts increasing interest recently under the context of circular economy. According to EU regulation 10/2011,1 recycled materials may not have a negative effect on consumer health. Several studies have been conducted over the last few decades to investigate migratable substances from rHDPE,2−7 including the examination of odorants8−11 and volatile substances.12 While these studies have primarily focused on volatile substances and utilized well-developed GC–MS and commercial libraries, to the best of our knowledge, non-targeted screening of non-volatile compounds in recycled polyolefins remains largely unaddressed. This type of screening is more sophisticated and requires the use of expensive high-resolution mass spectrometry (HRMS) as well as specialized expertise.

Identification of non-volatile substances in FCM employing HRMS together with vendor software has been well documented.13−15 In the last decade, various open-source tools like XCMS,16 MZmine 2,17 MS-DIAL,18 and in silico fragmentation tools, e.g., MetFrag,19 MS-FINDER,20 and SIRIUS 4,21 have been developed to facilitate and improve the handling of HRMS data in the metabolomics community. Nonetheless, they are, in principle, applicable to the identification of any small molecule.22 These tools are of major interest for novelty, easy accessibility, and reproducibility.23 In contrast to vendor software, they are able to leverage publicly available MS/MS libraries like MassBank, RIKEN, and GNPS libraries and can connect to advanced tools, for example, CAMERA,24 MS-CleanR,25 and CliqueMS26 for feature (mass–retention time pair) cleaning.

This study aims to comprehensively analyze non-volatile migrants coming from rHDPE milk bottles by ultrahigh-performance liquid chromatography coupled to quadrupole time-of-flight mass spectrometry (UPLC-QTOF-MS) and advanced data processing workflow. An R package, namely, mspcompiler, was developed to compile various publicly accessible and in-house MS/MS libraries. The library was then utilized for identification in MS-DIAL. MS-CleanR was employed to clean up redundant features including a number of adducts and in-source fragments. An in silico fragmentation tool (MS-FINDER) was finally applied to identify the remaining unknowns, taking advantage of the list of chemicals previously identified in the same set of samples by GC–MS and a list of chemicals associated with plastic packaging compiled by Groh et al27 After the chemicals were annotated, pseudo-multiple reaction monitoring (pseudo-MRM) using parent–product ion pairs exported from MS-DIAL was employed as a sensitive targeted analysis to determine the presence of each identified substance in the samples using the MRMPROBS program. Finally, the concentrations of the annotated compounds in the simulants were quantified.

Materials and Methods

Descriptions about the samples have been detailed in our previous article.12 In brief, rHDPE milk bottles in flake (F) and pellet (P) forms were supplied by three Spanish companies. Samples F1.1, F1.2, F1.3, F1.3′, P1.1, P1.2, P1.3, and P1.3′ were from company 1; samples F2.1, F2.2, F2.3, F2.4, F2.5, F2.6, P2.1, P2.2, and P2.3 were from company 2; samples F3.1, F3.2, P3.1, and P3.2 were from company 3.

Solid–Liquid Extraction

Samples were first ground into fine powders, weighed (1.00 g), and extracted 3 times with 5 mL of dichloromethane under an ultrasonic bath for 1 h. Subsequently, the extracts were merged and evaporated to dryness by a gentle flow of nitrogen at 40 °C. Then, the extract was re-dissolved with 0.4 mL of methanol for 5 min under ultrasonication and 30 s of vortex mixing and filtered (0.2 μm Acodisc GHP syringe filter) prior to UPLC-QTOF-MS analysis. Samples and procedural blanks were simultaneously prepared in triplicate. Quality control (QC) sample was pooled from filtered extracts (50 μL from each sample).

Migration Test

Migration test was carried out based on the regulation EU 10/2011, which has been well-documented in our previous study.12 Samples as well as procedural blanks were simultaneously prepared in duplicate.

UPLC-QTOF-MS Analysis

A Waters Acquity UPLC was used with an Atlantis premier BEH C18 AX column (2.1 mm × 100 mm, 1.7 μm particle size) at a column temperature of 40 °C and a flow rate of 0.3 mL/min. The mobile phase consisted of water (A) and methanol (B) with 0.1% formic acid in both positive and negative modes with a gradient elution over a 13 min run. The initial mobile phase was 95/5 A/B shifted to 100/0 A/B in 7 min, kept for 4 min, dropped to the initial phase in 0.1 min, and maintained for 1.9 min before the next injection. The injection volume was 10 μL. The UPLC and the QTOF-MS were interfaced by an electro spray ionization (ESI) probe. Low energy (6 V) and ramp high energy (10–30 V) were used for data acquisition (MSE mod) scanning from 50 to 1200 Da. Leucine enkephalin was employed for online mass correction. Test-mix from Waters was injected every 20 injections to ensure accuracy of the data.

The data was processed by MS-DIAL (version 4.38)18 with a mass tolerance of 0.01 and 0.025 Da for MS1 and MS2, respectively. Adducts in negative mode were [M-H]−, [M + FA-H]–, [M + Hac-H]−, [2 M-H]−, [2 M + FA-H]−, and [2 M + Hac-H]−, and those in positive mode were [M + H]+, [M + NH4]+, [M + Na]+, [M + K]+, [2 M + H]+, [2 M + NH4]+, [2 M + Na]+, and [2 M + K]+; the minimum peak height was 3000, and the identification score cut-off was 80%. Only peaks that had sample max/blank ratios higher than 5 were kept.

In-House MS/MS Library and mspcompiler R Package

An in-house MS/MS library containing 449 and 172 plastic-related chemicals in the positive and negative modes, respectively, was built following a strategy proposed by Tada et al.,28 which is available at https://github.com/QizhiSu/MS-libraries. In addition, we developed a R package, mspcompiler, to clean and compile various MS/MS libraries for identification in MS-DIAL. The details are available in Appendix A.

MS-CleanR Feature Cleaning and Structural Elucidation by MS-FINDER

Feature tables exported from MS-DIAL were cleaned by removing those with blank/QC and relative standard deviations higher than 0.5 and 30, respectively, using the MS-CleanR package.25 Subsequently, the features were grouped into clusters using pre-calculated links, Pearson correlation, and other criteria. Clusters had to have a minimum Pearson correlation of 0.8 at a maximum p-value of 0.05 and maximum mass and retention time differences of 0.005 and 0.025, respectively, for positive/negative merging. It is worth noting that identical mobile phases should be used in both positive and negative modes to ensure the correlation of peaks as the pH value of mobile phases is well known to have great impact on the retention behavior of certain compounds. For subsequent structural elucidation in MS-FINDER,20 both the most intense and most connected features were kept. In addition, features that did not have representative MS/MS spectra (see example in Figure A1) were eliminated as well since they are meaningless for structural elucidation.

In MS-FINDER, the MS1 and MS2 tolerances were 0.005 and 0.025 Da, respectively, considering elements C, H, O, N, P, S, F, and Cl. Three structure databases were applied for the identification, namely, volatile and semi-volatile substances identified by GC–MS in the same set of samples in our previous study (volDB),12 chemicals associated with plastic packaging (cppDB) compiled by Groh K. J. etc.,27 and a generic database (genDB) integrated in MS-FINDER. The genDB includes only FoodDB (Food), PlantCyc (Plant), T3DB (Toxin), STOFF (Environment), NPA (Natural Products Atlas), KNApSAcK (Natural Product), NANPDB (Natural Product), and UNPD (Natural Product) because they could be contaminants in recycled plastics. The weights given to the three databases in MS-CleanR were 2, 1.5, and 1, respectively. Manual inspection of the MS/MS spectra was conducted to excluded compounds that did not have representative spectra as demonstrated in Figure A1 since most of the fragments were simply noise and meaningless for subsequent elucidation. The identification flow chart is shown in Figure A2.

Pseudo-Multiple Reaction Monitoring by MRMPROBS

MRMPROBS is an open-source software designed for targeted metabolomics, providing automated posterior probabilistic data assessment to replace manual methods.29 In this study, MRMPROBS was used as a pseudo-MRM approach to detect identified compound in each sample. The top 5 product ions were exported as a MRMPROBS library which was then used for MRM analysis directly on a DIA data set. Unlike conventional MRM, no re-acquisition of MRM data is required. Parameters used included a smoothing level of 1, a minimum peak width of 5, and a minimum peak height of 100. Retention time tolerance, amplitude tolerance, and minimum posterior score were 0.1 min, 15%, and 60%, respectively. A peak area at least 3 times higher than the blanks was required for a compound to be considered present in a sample.

Results and Discussion

Identification by Matching against Libraries

As shown in Table 1, there were 66 compounds identified either in the extracts or migrates by library search. Although some compounds had low scores, they were still included because their scores were negatively affected by the large number of noisy signal present in the spectra. For instance, N-[3-(dimethylamino)propyl]dodecanamide and dimethyldiben-zylidene sorbitol had low scores (Figure A3), but they were confirmed by standards using retention time and exact mass of the precursor ion and the major fragments. The compiled library facilitated the identification of several pesticides (e.g., propanil and pyrimethanil) and plastic-related chemicals (e.g., caprolactam and Irgafos 168) by library matching.

Table 1. Migrants Identified in 95% Ethanol or 3% Acetic Acid Food Simulantsa.

no. Rt precursor adduct name CAS formula fill (%) matrix score lev remark
1 0.83 341.1081 [M-H]− sucrose 57-50-1 C12H22O11 14.3 95EtOH Lmatch (86) I  
2 0.95 232.1456 [M + H]+ aminophenazone 58-15-1 C13H17N3O 9.5 95EtOH genDB (6.2) IV drug
3 1.28 124.0759 [M + H]+ o-anisidine 90-04-0 C7H9NO 14.3 3HAC Lmatch (86) V CMR; SVHC
4 1.65 158.0972 [M + H]+ quinoline, 2,7-dimethyl- 93-37-8 C11H11N 28.6 3HAC volDB (5) IV  
5 1.77 122.0967 [M + H]+ 2,4-dimethylbenzenamine 95-68-1 C8H11N 42.9 3HAC Lmatch (88) II intermediate
6 2.61 202.0432 [M + H]+ thiabendazole 148-79-8 C10H7N3S 9.5 extract Lmatch (81) IV drug
7 3.11 114.0915 [M + H]+ caprolactam 105-60-2 C6H11NO 23.8 3HAC Lmatch (86) II nylon 6 monomer
8 3.46 195.0885 [M + H]+ caffeine 58-08-2 C8H10N4O2 47.6 extract Lmatch (87) IV  
9 3.92 142.0427 [M + H]+ 3-chloro-o-toluidine 87-60-5 C7H8ClN 14.3 3HAC cppDB (5.4) IV intermediate
10 4.78 274.2754 [M + H]+ N,N-bis(2-hydroxyethyl)dodecylamine 1541-67-9 C16H35NO2 61.9 3HAC Lmatch (97) II  
11 4.83 214.2535 [M + H]+ N,N-dimethyldodecylamine 112-18-5 C14H31N 100 3HAC cppDB (6.2) II antistatic
12 4.83 200.2371 [M + H]+ N-methyldodecylamine 7311-30-0 C13H29N 90.5 extract Lmatch (87) IV  
13 4.90 285.2917 [M + H]+ N-[3-(dimethylamino)propyl]dodecanamide 3179-80-4 C17H36N2O 33.3 3HAC Lmatch (72) IV antistatic
14 5.02 230.2482 [M + H]+ lauramine oxide 1643-20-5 C14H31NO 33.3 extract Lmatch (92) IV surfactants
15 5.02 150.0912 [M + H]+ N-(2,4-dimethylphenyl)formamide 60397-77-5 C9H11NO 14.3 extract Lmatch (84) II Insecticides
16 5.04 288.2896 [M + H]+ 2-aminoheptadecane-1,3-diol   C17H37NO2 14.3 extract Lmatch (87) III  
17 5.33 202.0854 [M + H]+ simazine 122-34-9 C7H12ClN5 19 extract Lmatch (90) V CMR; herbicide
18 5.34 242.2852 [M + H]+ N,N-dimethyltetradecylamine 112-75-4 C16H35N 100 95EtOH Lmatch (86) II antistatic
19 5.40 266.0975 [M + H]+ albendazole 54965-21-8 C12H15N3O2S 9.5 extract Lmatch (80) IV drug
20 5.70 200.1191 [M + H]+ pyrimethanil 53112-28-0 C12H13N3 28.6 3HAC Lmatch (92) IV fungicide
21 5.74 242.1439 [M + H]+ prometryn 7287-19-6 C10H19N5S 9.5 extract Lmatch (88) IV herbicide
22 5.79 242.1435 [M + H]+ terbutryn 886-50-0 C10H19N5S 9.5 extract Lmatch (78) IV herbicide
23 5.86 192.1384 [M + H]+ diethyltoluamide 134-62-3 C12H17NO 61.9 95EtOH Lmatch (93) II insect repellent
24 5.88 270.3173 [M + H]+ N,N-dimethylhexadecylamine 112-69-6 C18H39N 95.2 95EtOH   II antistatic
25 5.95 312.3632 [M + H]+ N-methyldidecylamine 7396-58-9 C21H45N 42.9 95EtOH Lmatch (92) II intermediate
26 6.14 404.1255 [M + H]+ azoxystrobin 131860-33-8 C22H17N3O5 9.5 extract Lmatch (89) IV fungicide
27 6.18 253.0307 [M + H]+ 3,3′-dichlorobenzidine 91-94-1 C12H10Cl2N2 14.3 95EtOH Lmatch (84) V intermediate
28 6.37 230.1168 [M + H]+ sebuthylazine 7286-69-3 C9H16ClN5 28.6 extract Lmatch (87) IV herbicides
29 6.43 215.9981 [M-H]− propanil 709-98-8 C9H9Cl2NO 9.5 95EtOH Lmatch (90) IV herbicides
30 6.43 182.0097 [M + H]+ 2-(methylsulfanyl)-1,3-benzothiazole 615-22-5 C8H7NS2 9.5 extract Lmatch (81) IV fungicides
31 6.48 194.1175 [M + H]+ ethyl 4-(dimethylamino)benzoate 10287-53-3 C11H15NO2 28.6 extract Lmatch (91) II paint additives
32 6.71 332.0665 [M + H]+ piroxicam 36322-90-4 C15H13N3O4S 9.5 3HAC genDB (5.9) IV drug
33 6.75 284.1428 [M + H]+ metolachlor 51218-45-2 C15H22ClNO2 9.5 extract Lmatch (86) IV herbicide
34 6.76 415.2133 [M + H]+ dimethyldibenzylidene sorbitol 135861-56-2 C24H30O6 23.8 extract Lmatch (64) I plastic additive
35 6.83 293.1735 [M-H]− 3-(3,5-di-tert-butyl-1-hydroxy-4-oxocyclohexa-2,5-dien-1-yl)propanoic acid   C17H26O4 100 3HAC cppDB (6.4) IV NIAS
36 6.86 229.0862 [M + H]+ oxybenzone 131-57-7 C14H12O3 33.3 extract Lmatch (85) V UV filter; EDC
37 6.91 198.1856 [M + H]+ 1-octylpyrrolidin-2-one 2687-94-7 C12H23NO 23.8 95EtOH cppDB (5.9) IV  
38 6.94 308.1541 [M + H]+ tebuconazole 80443-41-0 C16H22ClN3O 9.5 95EtOH Lmatch (83) V fungicides; CMR; EDC
39 6.97 305.1081 [M + H]+ diazinon 333-41-5 C12H21N2O3PS 9.5 extract Lmatch (86) IV insecticide
40 7.00 342.078 [M + H]+ propiconazole 60207-90-1 C15H17Cl2N3O2 9.5 95EtOH Lmatch (81) V fungicides; CMR
41 7.06 277.1817 [M + H]+ 7,9-di-tert-butyl-1-oxaspiro(4,5)deca-6,9-diene-2,8-dione 82304-66-3 C17H24O3 47.6 95EtOH Lmatch (83) IV NIAS
42 7.14 406.074 [M + H]+ difenoconazole 119446-68-3 C19H17Cl2N3O3 9.5 extract Lmatch (85) IV fungicide
43 7.19 200.202 [M + H]+ lauramide 1120-16-7 C12H25NO 9.5 95EtOH Lmatch (85) IV  
44 7.20 310.2367 [M + Na]+ lauric acid diethanolamide 120-40-1 C16H33NO3 14.3 95EtOH Lmatch (86) II antistatic
45 7.22 395.0808 [M + H]+ diflufenican 83164-33-4 C19H11F5N2O2 9.5 95EtOH genDB (6.3) IV herbicides
46 7.24 220.1125 [M + H]+ N-phenyl-2-naphthylamine 135-88-6 C16H13N 9.5 95EtOH volDB (5.4) V lubricant; CMR
47 7.30 383.2042 [M + Na]+ tributyl citrate 77-94-1 C18H32O7 85.7 95EtOH cppDB (6.3) IV plasticizer
48 7.32 421.2326 [M + Na]+ tris(2-butoxyethyl) phosphate 78-51-3 C18H39O7P 42.9 extract Lmatch (89) IV plasticizer
49 7.37 277.1812 [M-H]− 3-(3,5-di-tert-butyl-4-hydroxyphenyl)propionic acid 20170-32-5 C17H26O3 28.6 extract Lmatch (88) III NIAS
50 7.40 273.1853 [M + H]+ galaxolidone   C18H24O2 28.6 extract Lmatch (83) IV  
51 7.47 179.0701 [M + H]+ 3-methoxycinnamic acid 6099-04-3 C10H10O3 33.3 extract Lmatch (86) II  
52 7.48 599.1155 [M + Na]+ ethylene terephthalate cyclic trimer   C30H24O12 9.5 95EtOH cppDB (5.5) IV PET oligomer
53 7.54 322.1454 [M + H]+ pyriproxyfen 95737-68-1 C20H19NO3 14.3 extract Lmatch (84) IV insecticide
54 7.54 425.2149 [M + Na]+ tributyl acetylcitrate 77-90-7 C20H34O8 81 95EtOH Lmatch (81) I plasticizer
55 7.55 199.1328 [M + H-H2O]+ sebacic acid monomethyl ester 818-88-2 C11H20O4 19 extract Lmatch (80) II  
56 7.63 255.1746 [M + H]+ 4-methylbenzylidene camphor 36861-47-9 C18H22O 38.1 extract Lmatch (84) V UV filter; EDC
57 7.67 295.2272 [M-H]− 9-hydroxy-10,12-octadecadienoic acid 10075-11-3 C18H32O3 76.2 extract Lmatch (88) III  
58 7.77 507.2737 [M + Na]+ 1,2,3-trideoxy-4,6:5,7-bis-o-[(4-propylphenyl)methylene]-nonitol (NX8000) 882073-43-0 C29H40O6 33.3 extract Lmatch (82) II plastic additive
59 7.88 384.1934 [M + Na]+ octocrylene 6197-30-4 C24H27NO2 100 95EtOH Lmatch (90) IV UV filter
60 8.00 259.2065 [M + H]+ galaxolide 1 1222-05-5 C18H26O 28.6 95EtOH volDB (5.8) IV  
61 8.04 311.1645 [M + H]+ avobenzone 70356-09-1 C20H22O3 100 95EtOH Lmatch (91) IV UV filter
62 8.06 291.1998 [M + H]+ 2-ethylhexyl 4-methoxycinnamate 83834-59-7 C18H26O3 61.9 extract Lmatch (83) V UV filter; EDC
63 8.13 443.3351 [M + H]+ 1,2,3-propanetriol 1-stearate 2,3-bisacetate 33599-07-4 C25H46O6 100 95EtOH cppDB (6) II  
64 8.20 256.2647 [M + H]+ palmitamide 629-54-9 C16H33NO 42.9 extract Lmatch (85) IV  
65 8.47 381.298 [M + Na]+ 2-stearoylglycerol 621-61-4 C21H42O4 76.2 95EtOH Lmatch (92) II lubricant
66 8.50 327.1963 [M + H]+ Chimassorb 81 1843-05-6 C21H26O3 42.9 95EtOH Lmatch (88) II UV absorber
67 8.52 437.3062 [M-H]− 2,4-di-tert-butylphenyl 3,5-di-tert-butyl-4-hydroxybenzoate (UV 120) 4221-80-1 C29H42O3 9.5 95EtOH cppDB (5.7) I plastic additive
68 8.57 803.5445 [2 M + Na]+ dioctyl phthalate 117-84-0 C24H38O4 19 95EtOH Lmatch (88) V plasticizer; EDC
69 8.63 393.2969 [M + Na]+ bis(2-ethylhexyl) adipate 103-23-1 C22H42O4 66.7 95EtOH Lmatch (93) II plasticizer
70 8.81 255.2318 [M-H]− palmitic acid 57-10-3 C16H32O2 71.4 95EtOH Lmatch (91) I  
71 8.81 360.324 [M + Na]+ erucamide 112-84-5 C22H43NO 61.9 95EtOH Lmatch (85) I lubricant
72 8.86 647.4585 [M + H]+ Irgafos 168 31570-04-4 C42H63O3P 100 95EtOH Lmatch (85) I antioxidant
73 8.86 281.2475 [M-H]− oleic acid 112-80-1 C18H34O2 100 95EtOH Lmatch (85) I  
74 9.18 340.3574 [M + H]+ docosanamide 3061-75-4 C22H45NO 61.9 95EtOH cppDB (5.9) I processing aid
75 9.25 469.3289 [M + Na]+ diisodecyl phthalate 89-16-7 C28H46O4 33.3 95EtOH Lmatch (86) V plasticizer; EDC
76 9.36 1175.776 [M-H]− Irganox 1010 6683-19-8 C73H108O12 100 95EtOH Lmatch (89) I antioxidant
77 9.39 431.1784 [M + H]+ 2,5-bis(5-tert-butyl-benzoxazol-2-yl)thiophene 7128-64-5 C26H26N2O2S 71.4 95EtOH Lmatch (86) III plastic additive
78 9.61 199.0157 [M-H]− (4-chloro-2-methylphenoxy)acetic acid 94-74-6 C9H9ClO3 9.5 3HAC Lmatch (73) IV herbicides
79 10.24 663.454 [M + H]+ oxidized Irgafos 168 95906-11-9 C42H63O4P 100 95EtOH Lmatch (98) IV NIAS
80 10.27 385.3471 [M + H]+ (+)-4-cholesten-3-one 601-57-0 C27H44O 90.5 extract Lmatch (71) IV  
81 11.44 553.4598 [M + Na]+ Irganox 1076 2082-79-3 C35H62O3 100 95EtOH Lmatch (85) II antioxidant
82 11.60 591.4948 [M + Na]+ glycerol dihexanoate 502-52-3 C35H68O5 14.3 95EtOH Lmatch (69) II emollient
83 11.73 533.529 [M + H]+ (Z)-octadec-9-enyl oleate 3687-45-4 C36H68O2 9.5 95EtOH cppDB (5.1) II  
Open in a new tab
a

Chemicals in bold font were confirmed by a reference standard considering exact mass, retention time, and MS/MS spectra. The “fill (%)” column is the percentage of samples that detected the chemical in all samples (21 in total). The column “matrix” tells where the chemicals were identified. Some compounds were initially identified in the extracts by matching libraries, and their determination in the simulants were achieved by pseudo-MRM in MRMPROBS. Lmatch in the “score” column represents library match, and the number in the bracket were the scores (full mark 100) given by MS-DIAL, while others were the three structure databases that the compounds were finally identified, and the number in the bracket were the scores (full mark 10) given by MS-FINDER. The “lev” column gives the level of concern of the compounds based on the method previous proposed by our group.30 For detailed information, please refer to Appendix B. In the “remark” column, CMR represents carcinogenic, mutagenic, and reprotoxic chemicals, while EDC stands for endocrine-disrupting chemicals.

The retention time of some compounds differed between 95% ethanol and 3% acetic acid food simulants (Figure A4). This is expected as a reverse-phase column, and an initial mobile phase of 5% ethanol was used. For aqueous samples, such as 3% acetic acid, sample injection should not affect the mobile phase composition as the solvent and mobile phase are similar. However, for organic samples like 95% ethanol, the 10 μL injection volume takes time to mix with the mobile phase because the mobile phase contains more organic solvent, which elutes compounds faster and results in shorter retention time. In the field of food contact materials, various food simulants can be used, and compounds may have different retention time in these simulants, leading to misidentification. Additionally, caution should be exercised when using retention time information from MS/MS libraries or prediction models for identifying compounds in aqueous food simulants as organic solvents are commonly used to prepare standard solutions for building these databases.

MS-CleanR Feature Cleaning and Structural Elucidation by MS-FINDER

Some features with high matching scores were removed by MS-CleanR as they were fragments rather than individual compounds. For instance, feature 106.0866 m/z at 7.20 min was identified as diethanolamine (Figure A5A), but it was actually an in-source fragment of the feature 288.2544 m/z at 7.20 min, which was identified as lauric acid diethanolamide (Figure A5B). After this, the remaining features were considered individual compounds and subjected to MS-FINDER for structural elucidation, which we will explain using chemical structure databases.

Integrating Results from GC–MS

Using MS-FINDER to search the entire Pubchem can be slow and yield a large number of candidates, given that the database archives tens of millions of compounds. In certain types of samples, many compounds are suspect or already known to be present. This is why various structure databases are used for structural elucidation as they help in detecting suspect compounds in similar samples. While manually comparing fragmentation patterns to the ones shown in literature in similar samples is tedious and requires a lot of expertise, MS-FINDER can computationally fragment structures with the same molecular formula in selected databases and rank them based on factors like in silico fragmentation probability and frequency in the databases. Hence, using relevant structure databases can be beneficial in the investigation of the samples.

Over 200 compounds were previously identified in this set of samples using GC–MS by our group,12 but many of them can also be detected by liquid chromatography. If a compound has already been identified in GC–MS, it needs not be re-analyzed using UPLC-QTOF-MS. The challenge lies in identifying unknown peaks in UPLC-QTOF-MS that may correspond to a GC–MS-identified compound. To do this, one common approach is to calculate the exact mass and common adducts of the GC–MS identified compounds, like [M + H]+ and [M + Na]+ in positive mode and [M-H]− in negative mode, and compare them to the UPLC-QTOF-MS peaks. While identical mass (within tolerance) may suggest the same compound, relying solely on exact mass may not be sufficient.

In this study, we proposed a more convenient and reliable way for correlating UPLC-QTOF-MS peaks to the GC–MS-identified compounds. This method involved creating a structure database of the GC–MS-identified chemicals and using it in MS-FINDER for structural elucidation. MS-FINDER rates molecular formulas based on mass errors, isotopic ratios, product ions, and neutral losses.20 It then predicts the MS/MS spectra of all chemicals with the same formula in the database and compares them to the acquired MS/MS spectra, a process known as precursor-oriented spectral search. By combining the exact mass and in silico MS/MS spectra, we can confidently match peaks to the corresponding compounds. For example, using the volDB, we matched the peak 214.2535 m/z at 4.826 min to N,N-dimethyldodecylamine (6 A), which was previously identified in this set of samples by GC–MS, and its distributions in GC–MS and UPLC-QTOF-MS among the samples were consistent (Figure A6B), demonstrating good correspondence. Although medelamine A and tetradecyl-amine had higher scores than N,N-dimethyldodecylamine when using the genDB, the latter one was deemed more reliable. Moreover, the MS/MS spectrum of N,N-dimethyldodecylamine shared the same pattern with its homologue N,N-dimethyltetradecylamine, which was confirmed by a reference standard (Figure A7). Based on this similarity, the peak 270.3173 m/z at 5.884 min was identified and confirmed as N,N-dimethylhexadecylamine. Many other matches were found using this method, providing high-confidence identifications by combining library and retention index match in GC–MS and in silico MS/MS spectra.

Chemicals Associated with Plastic Packaging Database as a Useful Structure Database

To help characterize plastic-related chemicals, the chemicals associated with plastic packaging database (cppDB) compiled by Groh et al.27 can be useful. However, the initial version lacks vital structural information, such as Smiles and InChIKey, which is essential for in silico fragmentation. Hence, we added this information to cppDB and reorganized it to be compatible with MS-FINDER, which is available in https://zenodo.org/record/4454648.

To showcase the utility of the cppDB, the identification of compound 35 is shown below as an example using the database. The peak 293.1735 m/z at 6.832 min in negative mode had a formula of C17H26O4 with good isotope match (Figure 1A). While many candidates with scores of >6 were found using the genDB, using the cppDB led to the identification of compound 35 (Figure 1B), which has been reported as a hydrolysate of compound 41, a degradation product of a common polyolefin antioxidant, Irganox 1010.31,32 Compound 41 and Irganox 1010 were also identified in this set of samples. In MS-CleanR, identifications using volDB, cppDB, and genDB were merged, and the unknown was automatically identified as compound 35 with a weight of 1.5 for the cppDB, while the genDB candidates had a weight of 1. The identification was confirmed by comparing the exact mass, retention time, and MS/MS spectrum of the hydrolysis (70 °C for 1 h in water) product of compound 41 to this peak.

Figure 1.

Figure 1

Open in a new tab

Isotope match (A) and in silico fragmentation (B) of compound 35.

Similarly, many other compounds were identified, such as lauric acid diethanolamide. Although some commonly used polymer additives, e.g., Irgafos 168, Irganox 1010, and Chimassorb 81 were identified by library matching, they can also be automatically identified by combining MS-FINDER and the cppDB. It should be noted, however, that the confirmation of identification through a reference standard is the only foolproof method. Nevertheless, in situations where reference standards are not available, combining in silico MS/MS and the cppDB can increase the confidence level of identification in plastic materials.

Use of Generic Databases

Recycled plastics may contain environmental contaminants and food residues in addition to plastic-related chemicals. Therefore, only generic databases containing information related to food, environment, and natural product were used. Several compounds were identified by this way, and some of them were later confirmed by reference standards. For instance, C19H11F5N2O2 was found to be the best formula for the peak 395.0808 m/z at 7.215 min in the positive mode (Figure 2A). Within the genDB, diflufenican was the only candidate that had in silico MS/MS matching the experimental spectrum of the unknown (Figure 2B). Some pesticides/drugs were identified in these samples, especially in P2.2 and P2.3. Therefore, it was not surprising to detect other pesticides. Since this unknown was only found in P2.2 and P2.3 (Figure 2C), diflufenican, an herbicide, was a good candidate for this unknown, and its identification was later confirmed using a certified standard. Similarly, other pesticides/drugs such as pyrifenox and piroxicam were identified in samples P2.2 and P2.3. However, the identification of many other compounds was only based on in silico MS/MS spectra matches (score > 5), and their identification confidences were relatively low. Furthermore, there were 13 compounds that remained unknown, and their spectra are kept in Appendix C, which can be used by MS-FINDER for reproduction and further exploration.

Figure 2.

Figure 2

Open in a new tab

Identification of diflufenican: isotope match (A); in silico fragmentation match by MS-FINDER (B); distribution of this compound across samples (C).

Pseudo-MRM by MRMPROBS

To determine the presence of an identified substance, it is crucial to examine the samples where it may be present. MS-DIAL provides a means to verify the annotation for each sample (Appendix A, Figure A8). However, some samples may lack annotation due to insufficient intensity or the presence of noise. Assessing such samples based on EIC/TIC can be challenging. In addition, for substances that were not identified through library matches, the MS-DIAL approach cannot be used as there is no annotation data available for each sample. As such, we employed a pseudo-MRM (MRMPROBS) method, which utilized a posterior probabilistic approach to more sensitively, accurately, and automatically determine the presence or absence of a substance in each sample. This approach integrated peak intensity, retention time, precursor–product ion ratio, shape, and coelution similarity to enhance the accuracy of the results.29 For instance, the compound 2,2′-methylenebis(6-tert-butyl-4-methylphenol) was not identified in samples F2.1 and F2.2 by MS-DIAL (Figure A8), but it was automatically identified in these samples using pseudo-MRM (Figure A9). Moreover, the absence of this compound in sample F2.4 was also confirmed.

In the migration samples, compounds identified in the extracts (by matching libraries) were also subjected to pseudo-MRM analysis, in addition to those characterized in the migrates. While some of these compounds did not have representative spectra for structural elucidation, they were detected in the simulants by the pseudo-MRM. Examples include azoxystrobin and ethyl 4-(dimethylamino)benzoate (Table 1). Essentially, the pseudo-MRM functioned as a sensitive targeted analysis in this context.

Efficiency of Extra Decontamination on Chemical Removal

Previously, the cleaning efficiency of extra decontamination on volatile compounds was evaluated.12 Herein, the non-volatile profiles were assessed. All identified substances and their presence (by pseudo-MRM) in the extracts are listed in Appendix D, with some excluded from Table 1 as they were not detected in the migrates and may not pose a risk to human health. Similarly, N-phenyl-2-naphthylamine exhibited significantly higher intensity in both flake and pellet samples following extra decontamination (Figure A10). It was suspected to be a contaminant or reaction product resulting from the extra decontamination process. Another compound that caught our attention was pyriproxyfen, an insecticide, which showed higher intensity in F1.3′ but not in P1.3′ (Figure A11) and was suspected to be an accidental contaminant. Moreover, high-molecular weight substances such as octocrylene and diisodecyl phthalate were not significantly reduced as previously discussed,12 which is expected as non-volatile compounds are more difficult to remove. Many pesticides such as sebuthylazine, were only detected in the extracts (Figure A10) but not in the migration (Table 1) likely due to low responses in the extracts. For example, propiconazole showed a strong response in P2.2 and P2.3 (Figure A12), but its presence in other extracts was confirmed by sensitive pseudo-MRM. However, since it was not found in migration, it may pose a risk to humans.

Quantification of the Migrants

Quantification details including linear range, determined coefficient (R2), LOD, and LOQ are shown in Appendix E. In line with the previous study,12 samples from company 2 contained many high-concern substances (level V and IV) not found in milk bottle-origin rHDPE (Table S1). This can be attributed to the high proportion of non-milk bottle plastics from the agriculture field in these samples. To minimize contamination, better separation of recycled materials is necessary. In company 1 and company 3 samples, the main risks came from two endocrine disruptors, octocrylene and 2-ethylhexyl-4-methoxycinnamate, and they were not significantly reduced by the extra decontamination. For other migrants with SML in EU 10/2011, their migration was all below the SML. Moreover, only few compounds migrated to 3% acetic acid, and their migrations were quite low and might not be risky for human health.

Conclusions

Aided by our in house MS/MS library (publicly available now), the developed R package mspcompiler, and our data analysis workflow, non-volatile compounds in rHDPE were characterized for the first time using non-targeted screening. Manual examination of the spectra is necessary during feature cleaning when dealing with data independent acquisition data. The list of identified compounds in the same samples using GC–MS was employed as a structure database for the in silico fragmentation tool (MS-FINDER) and automatically found correspondences for some unknowns in LC-QTOF-MS, helping us avoid spending plenty of time on elucidating compounds that were already known in the samples. The cppDB was valuable for identifying plastic-related compounds in the recycled plastics and is expected to be helpful for the investigations of other plastic packaging. However, this strategy only works for known unknowns, not unknown unknowns, which is much more challenging as they might not be present in any existing structure databases. Once identified, pseudo-MRM using precursor–product ion pairs was more sensitive and precise for determining the presence of compounds in the samples. Many high-risk substances were detected in samples with high amounts of non-milk-bottle plastics, demonstrating the significance of good sorting. Even so, high-molecular weight substances such as octocrylene are still obstacles for the food contact applications of rHDPE as they are not easy to remove.

Acknowledgments

The authors acknowledge funds given by ECOEMBES, Gobierno de Aragón, and European Social Funds to GUIA group T53_20R (University of Zaragoza) and project RTI2018-097805-B-I00 financed by the Ministerio de Ciencia e Innovación, Spain and the National Nature Science Foundation of China (no. 42207550).

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acs.analchem.2c05389.

  • Experimental details, supplementary figures, and supplementary tables (PDF)

  • Extended version of Table 1, showing more details such as mass error between the acquired and theoretical mass of the precursor ions, CID (ID in Pubchem), SMILES, and InChIKey of the identified compounds; spectra of the unknowns; chemicals identified in the extracts; quantification details including linear range, determined coefficient (R2), LOD, and LOQ (ZIP)

The authors declare no competing financial interest.

Supplementary Material

ac2c05389_si_001.pdf (1.4MB, pdf)
ac2c05389_si_002.zip (2.4MB, zip)

References

  1. EC . Regulation No. 10/2011; European Union; 2011. [Google Scholar]
  2. Huber M.; Franz R. Identification of migratable substances in recycled high density polyethylene collected from household waste. J. High Resolut. Chromatogr. 1997, 20, 427–430. 10.1002/jhrc.1240200806. [DOI] [Google Scholar]
  3. Welle F. Post-consumer contamination in high-density polyethylene (HDPE) milk bottles and the design of a bottle-to-bottle recycling process. Food Addit. Contam.: Part A 2005, 22, 999–1011. 10.1080/02652030500157742. [DOI] [PubMed] [Google Scholar]
  4. Coulier L.; Orbons H. G. M.; Rijk R. Analytical protocol to study the food safety of (multiple-)recycled high-density polyethylene (HDPE) and polypropylene (PP) crates: Influence of recycling on the migration and formation of degradation products. Polym. Degrad. Stab. 2007, 92, 2016–2025. 10.1016/j.polymdegradstab.2007.07.022. [DOI] [Google Scholar]
  5. Dutra C.; Pezo D.; Freire M. T. d. A.; Nerín C.; Reyes F. G. R. Determination of volatile organic compounds in recycled polyethylene terephthalate and high-density polyethylene by headspace solid phase microextraction gas chromatography mass spectrometry to evaluate the efficiency of recycling processes. J. Chromatogr. A 2011, 1218, 1319–1330. 10.1016/j.chroma.2010.12.099. [DOI] [PubMed] [Google Scholar]
  6. Salafranca J.; Cacho J.; Nerín C. Compositional analysis and feasibility study of recycled high density polyethylene for direct food contact applications. Polym. Recycl. 1999, 4, 13–25. [Google Scholar]
  7. Salafranca J.; Cacho J.; Nerín C. Supercritical fluid extraction (SFE) optimization by full-factorial design for the determination of irganox 1076, irgafos 168, and chimassorb 81 in virgin and recycled polyolefins. J. High Resolut. Chromatogr. 1999, 22, 553–558. . [DOI] [Google Scholar]
  8. Strangl M.; Lok B.; Breunig P.; Ortner E.; Buettner A. Resour. Conserv. Recycl. 2021, 164, 105143 10.1016/j.resconrec.2020.105143. [DOI] [Google Scholar]
  9. Cabanes A.; Strangl M.; Ortner E.; Fullana A.; Buettner A. Odorant composition of post-consumer LDPE bags originating from different collection systems. Waste Manage. 2020, 104, 228–238. 10.1016/j.wasman.2020.01.021. [DOI] [PubMed] [Google Scholar]
  10. Strangl M.; Ortner E.; Buettner A. Resour. Conserv. Recycl. 2019, 146, 89–97. 10.1016/j.resconrec.2019.03.009. [DOI] [Google Scholar]
  11. Strangl M.; Schlummer M.; Maeurer A.; Buettner A. Comparison of the odorant composition of post-consumer high-density polyethylene waste with corresponding recycled and virgin pellets by combined instrumental and sensory analysis. J. Clean Prod. 2018, 181, 599–607. 10.1016/j.jclepro.2018.01.137. [DOI] [Google Scholar]
  12. Su Q. Z.; Vera P.; Salafranca J.; Nerín C. Resour. Conserv. Recycl. 2021, 171, 105640 10.1016/j.resconrec.2021.105640. [DOI] [Google Scholar]
  13. Nerín C.; Alfaro P.; Aznar M.; Domeño C. The challenge of identifying non-intentionally added substances from food packaging materials: A review. Anal. Chim. Acta 2013, 775, 14–24. 10.1016/j.aca.2013.02.028. [DOI] [PubMed] [Google Scholar]
  14. Martínez-Bueno M. J.; Gómez Ramos M. J.; Bauer A.; Fernández-Alba A. R. An overview of non-targeted screening strategies based on high resolution accurate mass spectrometry for the identification of migrants coming from plastic food packaging materials. TrAC, Trends Anal. Chem. 2019, 110, 191–203. 10.1016/j.trac.2018.10.035. [DOI] [Google Scholar]
  15. Peters R. J. B.; Groeneveld I.; Sanchez P. L.; Gebbink W.; Gersen A.; de Nijs M.; van Leeuwen S. P. J. Trends Food Sci. Technol. 2019, 85, 44–54. 10.1016/j.tifs.2018.12.010. [DOI] [Google Scholar]
  16. Benton H. P.; Wong D. M.; Trauger S. A.; Siuzdak G. XCMS2: processing tandem mass spectrometry data for metabolite identification and structural characterization. Anal.Chem. 2004, 5, 6382–6389. 10.1021/ac800795f. [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Pluskal T.; Castillo S.; Villar-Briones A.; Orešič M. MZmine 2: Modular framework for processing, visualizing, and analyzing mass spectrometry-based molecular profile dataV. BMC Bioinform. 2010, 11, 395. 10.1186/1471-2105-11-395. [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Tsugawa H.; Cajka T.; Kind T.; Ma Y.; Higgins B.; Ikeda K.; Kanazawa M.; Vandergheynst J.; Fiehn O.; Arita M. Nat. Methods 2015, 12, 523–526. 10.1038/nmeth.3393. [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Ruttkies C.; Schymanski E. L.; Wolf S.; Hollender J.; Neumann S. J. MetFrag relaunched: incorporating strategies beyond in silico fragmentation. J. Cheminformatics 2016, 8, 3–16. 10.1186/s13321-016-0115-9. [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Tsugawa H.; Kind T.; Nakabayashi R.; Yukihira D.; Tanaka W.; Cajka T.; Saito K.; Fiehn O.; Arita M. Anal. Chem. 2016, 88, 7946–7958. 10.1021/acs.analchem.6b00770. [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Dührkop K.; Fleischauer M.; Ludwig M.; Aksenov A. A.; Melnik A. V.; Meusel M.; Dorrestein P. C.; Rousu J.; Böcker S. Nat. Methods 2019, 16, 299–302. 10.1038/s41592-019-0344-8. [DOI] [PubMed] [Google Scholar]
  22. Ljoncheva M.; Stepišnik T.; Džeroski S.; Kosjek T. Cheminformatics in MS-based environmental exposomics: Current achievements and future directions. Trends Environ. Anal. Chem. 2020, 28. 10.1016/j.teac.2020.e00099. [DOI] [Google Scholar]
  23. Stanstrup J.; Broeckling C. D.; Helmus R.; Hoffmann N.; Mathé E.; Naake T.; Nicolotti L.; Peters K.; Rainer J.; Salek R. M.; Schulze T.; Schymanski E. L.; Stravs M. A.; Thévenot E. A.; Treutler H.; Weber R. J. M.; Willighagen E.; Witting M.; Neumann S. The metaRbolomics Toolbox in Bioconductor and beyond. Metabolites 2019, 9, 200. 10.3390/metabo9100200. [DOI] [PMC free article] [PubMed] [Google Scholar]
  24. Kuhl C.; Tautenhahn R.; Böttcher C.; Larson T. R.; Neumann S. Anal. Chem. 2012, 84, 283–289. 10.1021/ac202450g. [DOI] [PMC free article] [PubMed] [Google Scholar]
  25. Fraisier-Vannier O.; Chervin J.; Cabanac G.; Puech-Pages V.; Fournier S.; Durand V.; Amiel A.; André O.; Benamar O. A.; Dumas B.; Tsugawa H.; Marti G.. MS-CleanR: A feature-filtering approach to improve annotation rate in untargeted LC-MS based metabolomics bioRxiv Cold Spring Harbor Laboratory; 2020. [DOI] [PubMed] [Google Scholar]
  26. Senan O.; Aguilar-Mogas A.; Navarro M.; Capellades J.; Noon L.; Burks D.; Yanes O.; Guimera R.; Sales-Pardo M. CliqueMS: a computational tool for annotating in-source metabolite ions from LC-MS untargeted metabolomics data based on a coelution similarity network. Bioinformatics 2019, 35, 4089–4097. 10.1093/bioinformatics/btz207. [DOI] [PMC free article] [PubMed] [Google Scholar]
  27. Groh K. J.; Backhaus T.; Carney-Almroth B.; Geueke B.; Inostroza P. A.; Lennquist A.; Leslie H. A.; Maffini M.; Slunge D.; Trasande L.; Warhurst A. M.; Muncke J. Sci. Total Environ. 2019, 651, 3253–3268. 10.1016/j.scitotenv.2018.10.015. [DOI] [PubMed] [Google Scholar]
  28. Tada I.; Tsugawa H.; Meister I.; Zhang P.; Shu R.; Katsumi R.; Wheelock C. E.; Arita M.; Chaleckis R. Creating a Reliable Mass Spectral-Retention Time Library for All Ion Fragmentation-Based Metabolomics. Metabolites 2019, 9, 1–15. 10.3390/metabo9110251. [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Tsugawa H.; Arita M.; Kanazawa M.; Ogiwara A.; Bamba T.; Fukusaki E. Anal. Chem. 2013, 85, 5191–5199. 10.1021/ac400515s. [DOI] [PubMed] [Google Scholar]
  30. Su Q. Z.; Vera P.; Nerín C.; Lin Q. B.; Zhong H. N. Resour. Conserv. Recycl. 2021, 167, 105365 10.1016/j.resconrec.2020.105365. [DOI] [Google Scholar]
  31. Singh G.; Gollapalli R.; Blinder A.; Patel M. J. Pharm. Biomed. Anal. 2018, 152, 197–203. 10.1016/j.jpba.2018.01.053. [DOI] [PubMed] [Google Scholar]
  32. Beißmann S.; Stiftinger M.; Grabmayer K.; Wallner G.; Nitsche D.; Buchberger W. Polym. Degrad. Stab. 2013, 98, 1655–1661. 10.1016/j.polymdegradstab.2013.06.015. [DOI] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

ac2c05389_si_001.pdf (1.4MB, pdf)
ac2c05389_si_002.zip (2.4MB, zip)

Articles from Analytical Chemistry are provided here courtesy of American Chemical Society

RESOURCES