Widespread detection of Polyethylene glycol reveals chronic human exposure through pharmaceutical drugs
Harsha GoudaiD, Patricia Kelly, Luke WhileyiD, Daniel Gómez‐PéreziD, Anna Mascellani Bergo, Wilhan Donizete Gonçalves NunesiD, Elena ChekmenevaiD, Ada H. Y. Yuen, Mark DavidiD, Shona MckirdyiD, Andrew NelsoniD, Darren L. Smith, Haoqi Nina ZhaoiD, Jeong In SeoiD, Edward J. MathiasiD, Gillian FarrelliD, Toon Anton Willem ScheurinkiD, Michael J. StrobeliD, Helena Mannochio-RussoiD, Konstantinos GkikasiD, Mingxun WangiD, Jaroslav HavlíkiD, Christopher QuinceiD, Zoltán TakátsiD, Matthew Lewis, , Pieter C. DorresteiniD
Abstract
Polyethylene glycol (PEG) is a synthetic polymer ubiquitous in pharmaceuticals, personal care products, food additives, and industrial manufacturing. Despite its widespread use and potential importance as an exposure chemical, the prevalence of PEG exposure and its excretion in human populations remain largely uncharacterized. Moreover, PEG detected in human biofluids is frequently assumed to arise from analytical contamination during sample preparation, potentially obscuring its contribution to the human exposome and confounding metabolic phenotyping studies. Here, we show that PEG is as component of the human xenobiotic exposome and is further metabolized into PEG hydoxy acid and diacid metabolites in humans. We further find that PEG exposure is associated with alterations in microbiome composition and short-chain fatty acid metabolism, suggesting its biological impact of its exposure. We identify PEG exposure in approximately 2.3% of publicly available metabolomics data files and provide a reusable 85,484 candidate PEG and PEGylated MS/MS spectral library for future use for the metabolomic community. Together, these findings establish PEG signal in human biofluids can reflect genuine exposure, and that PEG exposure is neither metabolically inert nor biologically silent.
02 /Specification & Classification
Click any tag to browse matching papers- Citations
- 1
- Organism
- Assay
- Research Focus
- Approach
In-depth summary
Claude Opus 5.5 · xhigh effort · 8 Nov 2026 · from the full text of version 1 (18 Sept 2026) · CC BY 4.0
Polyethylene glycol in human biofluids can mark drug exposure rather than contamination; it is oxidized to acid metabolites and appears in 2.3% of public metabolomics files.
Metabolomics labs usually treat polyethylene glycol (PEG) ions as labware or solvent contamination and discard them, which can hide a real chemical exposure and leave ion suppression uncorrected in LC-MS data. The work links PEG in stool, urine and plasma to laxative or medication intake, shows it is biotransformed in vivo and tied to gut microbiome shifts, and releases a public MS/MS library for annotating it.
Key results
ShownIn stool from young Crohn's disease patients (iPENS), PEG signal tracked documented laxative intake rather than sample preparation. Independent LC-MS/MS and 1H NMR calls agreed fully, and NMR pointed to a PEG 3350 grade.
144 stool samples (11 PEG+ from 5 participants): PEG+ total ion count median 2.10 × 10¹⁰ vs 1.27 × 10¹⁰ in PEG- (p < 0.001); PEG ions a median 31% of total ion current vs 0.07%; ~200-fold PEG abundance gap; oxonium fragments m/z 89.0597 to 265.1645 spaced 44.0262 Da; 100% NMR and LC-MS/MS agreement. · Laxative use is associated with a strong PEG…; Detection of PEG signals by 1H NMR spectroscopy…
ShownIngested PEG is absorbed and oxidized in humans. AIRWAVE urine carried PEG hydroxy acid (PEG-COOH) and diacid (PEG-2COOH) metabolites, and urine collected after a PEG-only gel capsule held high levels of both PEG and PEG-COOH.
Urine by RP-LC-MS (n=1040) and HILIC-LC-MS (n=990): PEG(n8) vs PEG(n8)-COOH intra-sample Pearson r =0.92; matched plasma PEG(n8) vs creatinine-adjusted urinary PEG-COOH(n8) r = 0.32, p<2.2e^-16 (n=990); the capsule (Nurofen Express) contained no PEG-COOH; PEG-COOH was absent from all plasma samples. · Human PEG metabolism following medication…; Metabolomics analysis of AIRWAVE cohort; Figure legends
ShownPEG MS/MS signatures occur in about 2.3% of public metabolomics files. The human data cover feces, blood and urine as well as saliva, skin, brain, milk and the nasal cavity. The hits were merged into an 85,484-spectrum candidate library on GNPS2.
MassQL oxonium query (zero false discovery against 2,003,180 GNPS reference spectra) retrieved 4,085,996 PEG ions from 66,644 of 2,477,862 MassIVE, 4,305 of 204,085 Metabolomics Workbench and 3,388 of 538,621 MetaboLights files; Falcon clustering gave 85,854 spectra, 85,484 after network cleanup; 13,352 precursors assigned chain lengths (<10 ppm). · Mass-spectrometry (MS2) library for PEG…; Design of PEG MassQL query; PEG library generation
ShownPEG exposure was associated with lower fecal short-chain fatty acid (SCFA) levels, including the three major SCFAs. The effect was small but remained after accounting for repeated sampling of the same patients.
1,014 iPENS stool samples, dry-weight normalised SCFA profiles: PERMANOVA F = 14.1, R2=0.015, p<0.0001; with repeated-sampling adjustment F = 4.2, R2=0.0047, p=0.0025; acetate, propionate and butyrate significantly reduced in PEG+, with further changes in isobutyrate, 2-methylbutyrate, caproate and others. · PEG exposure alters microbiome composition and…
SuggestedPEG+ stool showed a modest community shift. Clostridia were depleted, mainly SCFA-producing Lachnospiraceae (Blautia_A, Anaerostipes, Agathobacter, Fusicatenibacter), while Bacteroidota, Sutterella, Parasutterella and Desulfovibrio were enriched.
Bray-Curtis on species-level metagenome-assembled genomes (MAGs) in patients with both PEG+ and PEG- samples (n=16): PERMANOVA R2=0.0597, p<0.001; MetaPhlAn 4 R² = 0.0462, p < 0.001; centered log-ratio linear models (PEG + patient) on 824 samples from 112 patients found 69 differentially abundant taxa (21 PEG+-enriched, 48 depleted; FDR < 0.05). · PEG exposure alters microbiome composition and…; Metagenomics data analysis
Methods, data and reproducibility
Untargeted Orbitrap LC-MS/MS and 1H NMR of Crohn's disease stool classified PEG exposure, which was then related to SCFA levels and MAG profiles. AIRWAVE urine and plasma LC-MS tracked PEG oxidation products, and a MassQL oxonium-ion query mined public repositories to build a clustered PEG MS/MS library.
- iPENS stool from children and young adults (6-17 years) with Crohn's disease after exclusive enteral nutrition; LC-MS/MS PEG profiling · human · n = 144 samples (11 PEG+ from 5 participants)
- iPENS fecal 1H NMR spectra (500 MHz, 1D NOESY) used for PEG detection and grade estimation · human · n = 867 fecal samples
- iPENS stool short-chain fatty acid quantification, classified PEG+ or PEG- · human · n = 1,014 stool samples
- iPENS fecal metagenomes co-assembled per patient and binned into species-level MAGs · human · n = 824 samples from 112 patients
- AIRWAVE Health Monitoring Study (UK police personnel) urine and same-visit plasma profiled by RP- and HILIC-LC-MS · human · n = urine RP n=1040, HILIC n=990; plasma 990
- PEG400, PEG1000, PEG3350 and PEG8000 standards (reversed-phase LC-MS/MS) plus PEG1000 trapped ion mobility MS · n = Five replicates (PEG1000 TIMS) · MassIVE MSV000101530; MSV000102755
- LC-MS/MS of over-the-counter drugs, foods and personal care products, including foodMASST and personal care MASST searches · n = ~3,500 food and 500 personal care product files · MassIVE MSV000099832; MSV000097542; MSV000095003
- Public positive-mode MS/MS data in GNPS/MassIVE, Metabolomics Workbench and MetaboLights searched with MassQL · multiple (repository-wide; human subset via ReDU metadata) · n = 2,477,862 MassIVE; 204,085 Workbench; 538,621 MetaboLights files
- Methods & tools
- MassQL (Mass Spec Query Language), GNPS2 (molecular networking, massql_library_generation, foodMASST, personal care MASST), MZmine4, MSConvert (ProteoWizard), Progenesis QI 2.1, Falcon 1.3 spectral clustering, UHPLC with Orbitrap Exploris 240 LC-MS/MS, Bruker timsTOF PRO 2 trapped ion mobility MS, 1H NMR on Bruker AVANCE III HD 500 MHz (TopSpin 3.6.4), Cutadapt, MEGAHIT, CONCOCT, MetaBAT2, Galah (MAG pipeline), GTDB release 232 and MetaPhlAn 4, PERMANOVA, NMDS and CLR linear models with Benjamini-Hochberg FDR
- Data & accessions
- https://library.gnps2.org/https://library.gnps2.org/api/libraries/GNPS2-PEG-MASSQL-PROPOGATED/mgfhttps://gnps2.org/status?task=76eebe2be1a54088be028da1bf28a2afhttps://gnps2.org/status?task=798c50ed50984b578d550d5a729b3db0https://gnps2.org/status?task=02d9433450fb444caa16a7ef3393b273https://gnps2.org/status?task=f1e8b05483b744c09ee81db7c4f53af0
- Software & environment
- MZmine4 4.3.0 (Orbitrap data) and 4.10.6 (TIMS data); MSConvert (ProteoWizard); Progenesis QI 2.1; TopSpin 3.6.4; MATLAB R2020a; R studio 4.5.3; Falcon 1.3; GTDB release 232; MetaPhlAn 4
- Reproducibility gaps
- iPENS and AIRWAVE data not public (confidentiality; access on request); no accession for metagenomes or NMR spectra; in-house MATLAB script and analysis code not released; library ID spelled PROPAGATED vs PROPOGATED in URL.
Limitations (5)
- The authors call the idea that chronic PEG exposure erodes SCFA-producing commensals plausible but still speculative, and they place the long-term health effects of chronic exposure outside the article's scope. · stated by the authors
- Chain lengths were assigned to only 13,352 of 85,854 unique precursor m/z values; the authors attribute the remainder to PEG metabolites, PEGylated molecules, or ion forms such as in-source fragments, adducts and multimers that their calculation did not consider. · stated by the authors
- PEG spreads across the whole LC-MS/MS run and suppresses endogenous metabolite signals, so dropping PEG features alone can leave confounding from ion suppression. In 1H NMR, the PEG singlet overlaps resonances such as glucose, glycine, taurine and choline compounds. · stated by the authors
- SCFA and microbiome differences are observational, from one Crohn's disease cohort, and the reason for laxative use is not modeled, so PEG-specific causation is not established. Microbiome sample sizes also differ across analyses (n=16, n=109, and 824 samples from 112 patients). · noted in this summary
- The 2.3% repository figure counts any file with a PEG oxonium-ion spectrum, including non-human samples, product extracts and possible labware contamination, so it does not estimate human exposure. The gel-capsule metabolism evidence also relies on self-reported intake. · noted in this summary
Who should read:Read this if you run untargeted LC-MS or NMR metabolomics on human biofluids and filter out PEG features, or study how laxatives and drug excipients affect the gut microbiome.
Reproducibility ledger
Nobody has reported trying to reproduce this paper yet. Did you replicate an experiment or run the code? Your report saves the next lab a week.
Sign in to file a reproducibility report.
Related papers
- Assessment of exposure to Di (2-ethylhexyl) phthalate (DEHP) metabolites and Bisphenol A (BPA) and its importance for the prevention of cardiometabolic diseases
F. Carli, D. Ciociaro, A. Gastaldelli·bioRxiv·Dec 6, 2021
- An untargeted metabolomics analysis of exogenous chemicals in human milk and transfer to the infant
S. P. Thomas, J. M. Gauglitz, … S. M. Tsunoda·bioRxiv·Mar 31, 2022
- Precision environmental health monitoring by longitudinal exposome and multi-omics profiling
Peng Gao, Xiaotao Shen, … Michael P. Snyder·Genome ResearchPeer-reviewed·Jun 1, 2022
- In Vivo Tissue Distribution of Microplastics and Systemic Metabolomic Alterations After Gastrointestinal Exposure
M. M. Garcia, A. S. Romero, … E. F. Castillo·bioRxiv·Jun 2, 2023
- Empirically establishing drug exposure records directly from untargeted metabolomics data
Haoqi Nina Zhao, Kine Eide Kvitne, … Pieter C. Dorrestein·bioRxiv·Oct 7, 2024
- Lifetime exposure to known and emerging groundwater contaminants significantly alters poultry microbiome and metabolome
Chamia C. Chatman, Elena G. Olson, … Erica L.‐W. Majumder·Applied and Environmental MicrobiologyPeer-reviewed·Mar 20, 2026