An efficient urine peptidomics workflow identifies chemically defined dietary gluten peptides from patients with celiac disease

The study addresses the long-standing gap in directly identifying chemically defined, diet-derived gluten peptides that are absorbed, circulate, and are excreted in humans. While proteins are generally digested to amino acids or di-/tri-peptides, gluten proteins are resistant to proteolysis due to high glutamine and proline content, leading to immunotoxic peptides that can cross the intestinal barrier and trigger CeD. Existing approaches largely rely on in vitro digests and antibody-based detection, which cannot reveal exact peptide sequences or capture the full diversity of immunogenic peptides. The purpose is to develop an untargeted LC-MS/MS workflow capable of sequencing urinary peptides to characterize in vivo gluten digestion products, compare their occurrence in CeD patients versus controls, and provide insight into gluten peptide ADME and immunogenicity. This has importance for understanding disease mechanisms, improving diagnostics, and monitoring gluten-free diet adherence.

Prior work showed gluten-derived peptides in urine by chromatographic UV and antibody-based assays, but without sequence resolution. Most gluten detection relies on monoclonal antibodies (e.g., R5, G12, A1), which recognize specific motifs and thus miss peptides lacking those epitopes. Structure–function studies with synthetic peptides have demonstrated that minor sequence changes drastically alter HLA binding and T-cell responses. Reports also describe non–T-cell mediated effects of certain gluten peptides and their innate immunostimulatory capacity. Despite extensive in vitro work and epitope mapping, no chemically defined dietary peptide had been identified in human circulation or excretion, leaving gluten peptide ADME poorly understood.

The authors developed an optimized urinary peptidomics workflow compatible with standard LC-MS/MS. Key steps: (1) Creatinine normalization: urine volume adjusted to contain ~30 µmol creatinine (typically 1–10 mL). (2) Reduction/alkylation: cysteines reduced with DTT (2 mM) and alkylated with iodoacetamide (4 mM) to prevent disulfide bonds that could complicate analysis. (3) Protein depletion: centrifugal filtration with 10 kDa MWCO under acidic conditions (2% formic acid) to enrich low-MW peptidome. (4) Solid-phase extraction: mixed cation exchange cartridges (Oasis MCX) used to remove salts, metabolites, and pigments with sequential washes (2% formic acid with ammonium formate, methanol, re-equilibration, ammonium hydroxide-containing washes), followed by elution with 95% methanol/5% ammonium hydroxide; eluates dried and reconstituted in water. (5) LC-MS/MS: nanoLC on RP-C18 (24 cm, 100 µm i.d., ReproSil-Pur C18-AQ 3 µm) using a 70 min gradient (4–25% B for 50 min; 25–50% B for 20 min; A: 0.1% FA in water, B: 0.1% FA in acetonitrile). Data-dependent acquisition on an Orbitrap Fusion Lumos (top-speed 3 s cycle): MS1 at 120,000 resolution (m/z 400–1500), selecting charge 2–7 precursors (intensity >50,000) for CID (NCE 35%, 1.6 m/z isolation) and charge 3–7 for EThcD (ETD + 25% supplemental energy). MS2 in Orbitrap at 15,000 resolution (m/z 156–2000), dynamic exclusion 30 s, AGC 4e5 (MS1), 5e4 (MS2). (6) Database searching: PEAKS X/X+ against custom UniProt FASTA databases (human and wheat; or human, wheat, rye, barley). Fixed mod: carbamidomethyl Cys; variable mods: deamidation (N/Q), Met oxidation, N-terminal pyroglutamate (Gln/Glu), up to 3 variable mods/peptide. Tolerances: 10 ppm (precursor), 0.02 Da (fragment). Unspecific digestion mode. 1% FDR via decoy-fusion. (7) Assignment curation: peptides uniquely mapping to wheat counted as wheat; peptides mapping to both human and wheat counted as human. Sequences containing YVRPD were reassigned to human (immunoglobulin joining region artifact). Manual validation of deamidation assignments was performed by inspecting MS1 isotopic patterns versus theoretical native/deamidated distributions and, when necessary, MS2 fragment evidence; ~90% of auto-assigned deamidations were corrected to native sequences; ambiguous cases excluded from downstream analyses. Study designs: (a) Pilot healthy cohort: 4 participants—48 h gluten-free diet (GFD) with 8 h pooled urine collection on day 2, then gluten challenge (two wheat bagels ~18 g gluten) with 8 h urine collection on day 3. (b) Simplified healthy cohort: 8 participants—overnight fast, pre-challenge spot urine, consume two bagels, 8 h pooled urine. (c) Grain specificity test: 2 participants—grain-free control and separate wheat/barley/rye breakfast challenges (~1.5 cups flour), 8 h urine each day. (d) Banked CeD urine: 9 single-void samples previously tested by lateral flow (G12/A1 antibodies), analyzed by LC-MS/MS. (e) Prospective clinical study: adults under evaluation for CeD (n=11) and healthy controls (n=8); post hoc diagnosis split into CeD (n=6) and other GI disorders (n=5). All underwent standardized overnight fast, bagel challenge, and 8 h pooled urine collection; samples processed in duplicate on two days. Overall workflow time: ~6 h sample prep; LC-MS/MS per run ~70 min.

• The optimized workflow increased identifiable wheat- and human-derived peptides by 29-fold and 13-fold, respectively, while mitigating instrument contamination; sample prep reduced from 2–3 days to ~6 h. - Across studies, more than 600 distinct dietary peptides were detected; approximately 35% contained CeD-relevant T-cell epitopes and about 5% were known innate immune stimulators (from abstract). Pilot healthy cohorts: - After gluten challenge, an average of 24 unique wheat peptide sequences per participant were detected; none in GFD/fasting samples; human peptide counts were unchanged between conditions. - A common peptide, GQQQPFPPQQPYPQPQPFPS, was found in all 8 healthy participants; its truncated variant GQQQPFPPQQPYPQPQPFP appeared in 7/8. These peptides, linked to innate immune stimulation, lack common antibody recognition motifs, suggesting they evade traditional immunoassays. - Many peptides harbored known T-cell epitopes; most frequent were DQ2.5-glia-y4c/DQ8-glia-y1a (QQPQQPFPQ; 32 distinct peptides) and DQ2.5-glia-y5 (QQPFPQQPQ; 14 peptides). Other epitopes (e.g., PQQSFPQQQ; QQPQQPYPQ; PFSQQQQPV) were also detected. - Post-translationally modified peptides with N-terminal pyroglutamate (e.g., PyrQTFPHQPQQQVPQPQQPQQP) were observed, potentially conferring protease resistance. - Newly observed peptides not previously described as digestion-resistant, e.g., SCHVMQQQCC and its variants, were found in multiple participants (6/8). Method specificity: - Grain-free diet days yielded only a single likely false-positive grain peptide. After challenges, unique peptide counts: wheat 51, rye 43, barley 37; human peptide repertoires were similar across days. Peptides were specific to the challenged grain, and mapping supported grain origin, with rye mapping limited by proteome annotation. Banked CeD urine: - Samples negative by lateral flow (n=4) had no wheat peptides by LC-MS/MS. High apparent gluten peptide samples (n=5) yielded at least 17 wheat peptides; mean 74/sample versus 24/sample in healthy challenge studies; one sample had 206 wheat peptides. Prospective clinical study: - CeD patients (n=6) had ~5-fold more unique urinary wheat peptides than healthy controls (n=8) or other GI patients (n=5) (p=0.017 for both comparisons). - Peptides containing at least one CeD-relevant T-cell epitope were ~4-fold higher in CeD (p=0.019 vs healthy; p=0.046 vs GI). - A total of 289 wheat peptides were unique to CeD patients; only 37 peptides were shared among all three groups. - CeD patients displayed broader peptide diversity, including variants of control peptides and sequences mapping to distinct proteome regions (e.g., residues 194–199 in γ-gliadin X observed in three CeD patients, absent in controls). Additional observations: - The canonical α2-gliadin 33-mer was not detected in urine; however, a related γ-gliadin 26-mer (FLQPQQPFPQQPQQPYPQQPQQPFPQ), highly immunogenic and protease-resistant, was found in 2/7 CeD patients. - No definitive evidence of TG2-mediated deamidation sites was obtained due to ambiguity at MS1/MS2 level; stringent validation excluded ambiguous assignments. - A commonly observed innate-stimulating peptide (GQQQPFPPQQPYPQPQPFPS) appeared in all gluten-challenged individuals, implying utility as a monitoring target beyond antibody motifs like QPQLP(Y/F), which were rare in urine.

The workflow enables direct sequence-level detection of in vivo–derived dietary peptides, overcoming prior reliance on in vitro digests and antibody assays. Findings reveal that CeD patients excrete a more diverse and epitope-rich repertoire of gluten peptides than controls, indicating differences in absorption and/or metabolism. Potential mechanisms include increased intestinal permeability in active CeD, facilitating peptide translocation and systemic exposure, and/or altered digestion/metabolism of wheat proteins in CeD. The method distinguishes closely related grains (wheat, rye, barley) by urinary peptide profiles, providing a window into gluten ADME and immunogenicity. Although the α2-gliadin 33-mer was not observed, detection of a highly inflammatory γ-gliadin 26-mer underscores physiological relevance. The lack of definitive TG2-deamidated peptide detection suggests low abundance, enhanced metabolic lability, or analytical challenges; targeted assays may be needed. Clinically, relying solely on peptide counts is unlikely to suffice for diagnosis due to interindividual variability, but a subset of sequences is enriched in CeD and may serve as biomarker candidates. The commonly observed innate-activating peptide sequence offers an improved target for noninvasive monitoring of gluten exposure and GFD compliance. More broadly, the workflow’s specificity and throughput position it for studies of endogenous and exogenous urinary peptides as disease biomarkers and for exploring bioactive peptides from other dietary proteins.

This work introduces an efficient, 6-hour sample preparation and LC-MS/MS workflow for untargeted urinary peptidomics that successfully identifies chemically defined, diet-derived gluten peptides in humans. It demonstrates that CeD patients have qualitatively and quantitatively distinct urinary gluten peptidomes, enriched for known T-cell epitopes and broader sequence diversity, compared with healthy controls and patients with other GI disorders. The method differentiates peptides from wheat, rye, and barley and reveals commonly excreted innate-activating peptides that escape traditional immunoassays, informing improved monitoring of gluten exposure and GFD adherence. The study lays a foundation for elucidating gluten peptide ADME, advancing CeD diagnostics and management, and enabling broader characterization of urinary peptidomes. Future work should include larger cohorts, targeted LC-MS/MS for low-abundance modifications (e.g., TG2 deamidation), validation of candidate diagnostic peptides, and extension to other dietary proteins and disease contexts.

• Deamidation site assignment: Stringent validation led to exclusion of many putatively deamidated peptides due to ambiguity in MS1 isotopic patterns and lack of definitive MS2 site-determining ions; targeted methods with synthetic standards are needed. - Non-detection of the α2-gliadin 33-mer suggests potential absorption or detection biases, or rapid metabolism. - Small to moderate cohort sizes in pilot and clinical studies, with substantial interindividual variability in peptide repertoires, limit generalizability and diagnostic performance estimation. - Database/annotation limitations (e.g., rye proteome under-annotation) affect precise proteome mapping for some grain peptides. - Search parameters did not include certain variable modifications (e.g., hydroxyproline), potentially underrepresenting endogenous collagen peptides; however, raw data are publicly available for reanalysis. - Single time-window urine collections (8 h) may miss temporal dynamics of peptide excretion.