Sensory and metabolite migration from tilapia skin to soup during the boiling process: fast and then slow

Animal skin soup is a longstanding traditional Chinese food believed to have health and therapeutic benefits. Tilapia skin, a by-product of processing, has been used for collagen/gelatin extraction and even biomedical applications but its soup is not commercially available, and the sensory and metabolite migration during boiling have not been characterized. Mass spectrometry-based untargeted metabolomics has proven powerful in food component, safety, and processing analyses. This study aimed to investigate, during boiling, the content, odor (E-nose), taste (E-tongue), and metabolite migration from tilapia skin to soup, to identify key time windows and differential metabolites underlying sensory and compositional changes.

The paper reviews applications of untargeted metabolomics in food science for component analysis, safety assessment, consumption monitoring, and processing effects, including prior work on Pu'er tea, Fu brick tea, Pinot noir wine, sea cucumbers, thermally processed tilapia muscle, and pufferfish soup. It cites studies linking specific metabolites to flavor/taste (e.g., oleic acid with juiciness/tenderness, linoleic acid with fried-food flavor, propylpyrazine with roasted notes), and discusses taste-active compounds (sweet amino acids, bitter valine/creatine/creatinine, N-acetylglycine). This positions metabolomics and sensory technologies as effective for profiling processing-induced changes and interpreting flavor formation.

Materials: Tilapia skins (Genetic Improvement of Farmed tilapia) were obtained post-fillet, stored at −18 °C. Reagents included HPLC-grade methanol, acetonitrile, chloroform, pyridine; methoxyamine hydrochloride; adonitol; ammonium acetate/hydroxide; BSTFA + 1% TMCS; fatty acid methyl ester standard mix. Boiling process: Thawed skins were de-scaled and defatted, cut into 5×6 cm pieces. For each replicate, 2 g skin + 10 mL ultrapure water were sealed in a glass vial and heated in a 100 °C water bath for 10, 30, or 60 min. Skin and soup were separated at each time point for analyses. Content analyses: Mass loss ratio was calculated from initial mass (skin 2 g; soup 10 g) and measured mass at each time. Moisture content of skin followed GB 5009.3-2016 (drying at 101–105 °C). Ash content followed GB 5009.4-2016 (ashing at 550 ± 25 °C). Soluble solids in soup were measured by drying at 101–105 °C. Protein content followed GB 5009.5-2016 using Kjeltec 8400; nitrogen × 6.25. E-nose: Fox 4000 (Alpha M.O.S.) with 18 MOS sensors measured volatiles of skin and soups (n=3). Data were analyzed by PCA and radar plots (Origin), with one-way ANOVA/Duncan’s test. E-tongue: For skin and soup taste analysis, samples (n=3) were extracted into water (homogenize, sonicate, centrifuge 10614×g at 4 °C, defat, filter to 100 mL). Measurements used ASTREE (Alpha M.O.S.), 120 s acquisition. PCA/radar analyses in Origin; ANOVA/Duncan’s test. Metabolite extraction: Skin and soup were snap-frozen in liquid N2, stored at −80 °C. Skin was cryo-ground (stainless beads); soup was prewarmed (50 °C, 10 min). 100±1 mg skin or 100 µL soup were extracted with 1 mL methanol:acetonitrile:water (2:2:1) with internal standards (clenbuterol for positive mode; chloramphenicol for negative). Bead-beating and ice sonication were repeated; centrifuge 13800×g, 4 °C, 15 min; quench at −40 °C for 1 h. Aliquots were used for UHPLC-MS/MS; pooled QCs prepared. For GC-TOF-MS, 200 µL supernatant was dried; derivatized with methoxyamine HCl (20 mg/mL, 80 °C, 30 min) then BSTFA+1%TMCS (70 °C, 1 h); FAMEs were added to pooled QC. UHPLC-MS/MS: Vanquish UHPLC with UPLC BEH Amide (2.1×100 mm, 1.7 µm). Mobile phase A: 25 mM ammonium acetate + 25 mM ammonium hydroxide in water (pH 9.75); B: acetonitrile. Gradient: 0–0.5 min 95% B; 0.5–7 min 95→65% B; 7–8 min 65→40% B; 8–9 min 40% B; 9–9.1 min 40→95% B; 9.1–12 min 95% B. Flow 0.5 mL/min; 4 °C autosampler; 3 µL injection. Q-Exactive HFX operated in IDA with ESI: sheath 30, aux 25, capillary 350 °C, full MS 60,000, MS/MS 7500, NCE 10/30/60, spray 3.6 kV (+)/−3.2 kV (−). GC-TOF-MS: Agilent 7890A with LECO Pegasus HT, DB-5MS (30 m×250 µm×0.25 µm); splitless 1 µL; He carrier 1 mL/min. Oven: 50 °C 1 min; ramp 10 °C/min to 310 °C; hold 8 min. Inlet/transfer/ion source: 280/280/250 °C; EI 70 eV; m/z 50–500, 12.5 spectra/s; solvent delay 6.27 min. Data processing: UHPLC-MS/MS raw data converted to mzXML (ProteoWizard); XCMS used for peak picking/alignment/integration; annotation via BiotreeDB v2.1 (score cutoff 0.3). GC-TOF-MS processed in ChromaTOF (peak deconvolution, alignment); identification against LECO-Fiehn RTX5; QC filtering (<50% in QC or RSD>30% discarded). Datasets merged in R and imported to SIMCA v16.0.2. Statistics: Data log-transformed and scaled (centered for PCA, unit variance for OPLS-DA). PCA and OPLS-DA performed; OPLS-DA validated by 200× permutation tests. Univariate analysis (Student’s t-test p<0.05) combined with VIP>1 from OPLS-DA defined differential metabolites. Volcano plots generated in SIMCA; hierarchical clustering heat maps generated in R using Euclidean distance and complete linkage. KEGG pathway mapping performed (species mapping to Oreochromis niloticus, pathway references to Danio rerio), with bubble plots and pathway annotations.

- Visual/content changes: Soup turned white by 10 min and yellow by 30–60 min, indicating rapid early migration (0–30 min). Skin mass increased at 0–10 min then decreased at 10–60 min; soup mass decreased at 0–10 min then increased at 10–60 min. Skin moisture increased at 0–10 min then stabilized. Soup soluble solids increased over time. Skin crude protein decreased while soup protein increased, consistent with collagen/gelatin extraction. Skin ash decreased then fluctuated; soup ash increased. Overall migration intensity: 0–10 min > 10–30 min > 30–60 min. - E-nose: PCA showed strong separation. Skin: PC1 75.1%, PC2 13.9% (total 89.0%); Soup: PC1 67.4%, PC2 18.8% (total 86.2%). Largest changes occurred by 10–30 min; changes after 30 min were smaller (overlaps at 30–60 min). Sensors with prominent responses (P40/1, PA/2, P10/1) indicated abundance of oxidizing gases, polar organics, and combustible gas-related volatiles. Migration pattern matched content trends. - E-tongue: Skin PCA: PC1 57.5%, PC2 36.6% (94.1% total); Soup PCA: PC1 54.3%, PC2 42.5% (96.8% total). Taste order across samples: bitterness > richness-A > richness-B > saltiness > umami > sweetness > sourness; minimal sourness. Sweetness changed slightly (skin: ↑ at 0–30 min then ↓ at 30–60 min; soup: ↓ at 10–30 min, little change at 30–60 min). Taste migration pattern: 0–10 > 10–30 > 30–60 min. - Metabolome coverage: 783 chemicals detected in skins (566 annotated metabolites), 761 in soups (550 metabolites). PCA/OPLS-DA indicated time-dependent separations; OPLS-DA models robust (R²Y near 1; permutation tests with declining Q² showed no overfitting; t(1)O ~20%). - Differential chemicals (UVA + VIP criteria) and volcano summaries: Skin 0–10 min: 44 increased, 252 decreased, 487 unchanged; Skin 10–30 min: 50 increased, 19 decreased, 714 unchanged; Soup 10–30 min: 40 increased, 10 decreased, 711 unchanged; Skin 30–60 min: 38 increased, 16 decreased, 729 unchanged; Soup 30–60 min: 20 increased, 10 decreased, 711 unchanged. Cluster heat maps confirmed the largest shifts in the first 10–30 min (especially 0–10 min). - Key differential metabolites: Skin (6) with VIP>1 and p<0.05: adenine (↓ sharply by 10 min), gingerol (↓), terephthalic acid (↓), vanillin (↑ by 30 min then ↓ by 60 min), pentanenitrile (low abundance), 2‑pyrrolidinone (↑; potential bitterness/astringency modulator). Soup (7): butyramide (↑ at 30 min; nutty note), lysoPE(0:0/20:4(5Z,8Z,11Z,14Z)) (↑), lysoPE(22:6(4Z,7Z,10Z,13Z,16Z,19Z)/0:0) (↑), linoleic acid (highest at 30 min), N‑acetylneuraminic acid (↑), L‑threose (highest at 30 min), benzoin (↑ at 10–30 min). - Flavor/taste-related observations: In skin, oleic acid increased; linoleic acid decreased 0–30 min; propylpyrazine increased; benzoin increased later; multiple bitter or sweet-related amino acids and derivatives changed (e.g., d‑alanine ↓, valine ↓, creatine ↓). - Pathways: Skin 0–10 min involved amino acid, lipid, carbohydrate metabolism, other amino acids, cofactors/vitamins, and translation; significantly enriched pathways included linoleic acid metabolism, taurine/hypotaurine metabolism, and glycine/serine/threonine metabolism. Skin 10–30 min: lipid metabolism with enrichment in biosynthesis of unsaturated fatty acids and glycerophospholipid metabolism. Skin 30–60 min: carbohydrate metabolism with purine metabolism enriched. Soup 10–30 min: purine metabolism enriched. Soup 30–60 min: carbohydrate metabolism with D‑arginine and D‑ornithine metabolism enriched. Genetic information processing was only implicated in the initial 0–10 min. Overall pathway impact followed 0–10 > 10–30 > 30–60 min. - Practical implication: Most migration and sensory changes occur within the first 30 min, especially the first 10 min; extended boiling beyond 30 min yields diminishing changes. A 10 min boil may optimally balance flavor changes and mitigate compounds like adenine.

The study addressed how boiling drives the migration of sensory-active compounds and metabolites from tilapia skin into soup. Convergent evidence from content measurements, E-nose/E-tongue, untargeted metabolomics, and pathway analysis showed that migration is rapid initially (0–10 min) and slows thereafter (10–60 min). The earliest phase involved broad metabolic processes (including translation-level signatures), large compositional shifts, and notable changes in volatile/taste-active compounds, explaining the strong initial sensory differentiation. Identified key compounds (e.g., vanillin formation trend, butyramide, lysophospholipids, N-acetylneuraminic acid) help rationalize the evolving flavor profile of the soup, while reductions in adenine and increases in 2‑pyrrolidinone suggest potential nutritional and taste-modulating impacts on the skin. These findings provide a mechanistic and time-resolved basis for recommending shorter boiling durations to achieve desirable sensory traits and to guide value-added utilization of fish skin by-products.

Boiling tilapia skin leads to a fast-then-slow migration of components into soup, with the majority of content, sensory, and metabolite changes occurring within the first 30 min and particularly during 0–10 min. Multi-platform metabolomics (UHPLC-MS/MS and GC-TOF-MS) combined with E-nose/E-tongue established that 783 (skin) and 761 (soup) chemicals are time-variant, and identified 6 skin and 7 soup key differential metabolites linked to flavor/taste and potential nutritional significance. Pathway analyses reinforced the primacy of early-time metabolic perturbations. Practically, a 10 min boil among the tested durations (10, 30, 60 min) may be optimal for desirable sensory outcomes and reduced adenine content. The study also offers a general analytical workflow for probing metabolite migration in foods and supports value-added uses of aquatic by-products. Future work should elucidate specific reaction mechanisms, quantify targeted flavor/taste compounds, and map detailed transformation routes underlying the observed changes.

The study employed E-nose, E-tongue, and untargeted metabolomics, which limit definitive identification/quantitation of specific compounds and mechanistic elucidation. Detailed reaction mechanisms and precise transformation pathways were not resolved; sensory results are instrumental rather than panel-based. Further targeted analyses, sensory validation, and mechanistic studies are needed.