Abstract This study employed a multi-omics approach to examine the impact of feed fermentation on the lipid profile and volatile flavor of duck eggs. Sensory evaluation and GC × GC-TOFMS analyses demonstrated that microbial fermentation in feed significantly reduced the off-odor of duck eggs. Among the thirty-nine differential volatile compounds identified, six—namely 3-methylbutanal, 1-octen-3-ol, hexanal, acetophenone, 2-heptanone, and 2-pentylfuran—were likely responsible for the alteration in yolk aroma. Lipidomics and metabolomics identified twenty-five key differential lipids (p < 0.05, VIP > 1.85) and a modified metabolic pathway associated with linoleic acid (LA), involving four metabolites. Correlation analysis revealed significant associations between LA-containing lipids (e.g., DG (18: 2/18:2)), LA metabolites, and differential volatiles (e.g., 3-methylbutanal) (p < 0.05). These results provide insights into the mechanisms underlying off-odor reduction and offer a potential strategy for enhancing the flavor profile of duck eggs. Keywords: Volatile flavor, Duck eggs, Lipidomics, Fermented feed, Off-odor Highlights * • Fermented feed altered the lipid composition and volatiles of duck egg yolks. * • 25 lipids and 6 volatile compounds are found responsible for duck egg off-odor. * • Linoleic acid metabolites were linked to enhanced volatile flavor compounds. * • Correlation analysis highlights relations among metabolites, odor, and lipids. * • Study offers strategies to improve odor in duck egg production. 1. Introduction Duck eggs enjoy significant popularity, particularly in Asia, owing to their high yield and rich nutritional content, which includes substantial amounts of proteins, polyunsaturated fatty acids (PUFAs), vitamins, trace elements, and other beneficial compounds ([41]Sinanoglou et al., 2011). Despite this, their acceptance lags behind that of chicken eggs, largely due to their distinctive and often off-putting flavor. To improve consumer acceptance, duck eggs are frequently processed with ingredients such as table salt or rice wine to mask or modify their odor prior to sale ([42]Li et al., 2023). Consequently, odor modification plays a vital role in enhancing their palatability and fostering the growth of the duck egg industry. The yolk, composed primarily of proteins and lipids, is the main source of the characteristic odor of eggs. Notably, alcohols, aldehydes, and ketones resulting from the degradation of oxidized fats significantly influence the overall flavor and aroma ([43]Gao et al., 2025; [44]Ren et al., 2022; [45]Zhao et al., 2025). Recent studies identify trimethylamine (TMA), which emits a distinctly fishy odor ([46]Zhou, Xu, et al., 2024), as a key contributor to the off-odor of duck eggs ([47]Tian et al., 2024). Typically, TMA forms during the digestion of feed and accumulates in the yolk. In liver tissue, flavin-containing monooxygenase 3 (FMO3) converts it into the odorless TMA oxide, which is then excreted through bodily fluids ([48]Bekhit et al., 2021; [49]Cho & Caudill, 2017). The TMA content in egg yolks may result from two primary factors: FMO3 dysfunction and the consumption of high-choline feeds (phosphatidylcholine and related phospholipids), both contributing to the undesirable odor of eggs ([50]Bekhit et al., 2021; [51]Li et al., 2019). Furthermore, lipid oxidation plays a significant role in the yolk's odor profile ([52]Ghorbani Gorji et al., 2016). Studies have shown that lipid oxidation produces low-volatility compounds, such as aldehydes (e.g., hexanal, 3-methylbutanal, and (E,E)-2,4-decadienal) and ketones (e.g., 1-octen-3-one) ([53]Ren et al., 2022; [54]Xue et al., 2022). Additionally, the oxidation of PUFAs contributes to the formation of distinct food flavors by promoting the breakdown of amino acids into Strecker aldehydes ([55]Hu, Suo, et al., 2024; [56]Wang, Wang, et al., 2023). This suggests that lipid oxidation is likely another major contributor to the off-odor of duck eggs. The increasing focus on modifying egg odor has gained significant attention in recent years ([57]Li et al., 2019; [58]Yang et al., 2025; [59]Zhou et al., 2025). Several strategies have been investigated, including the adjustment of choline levels in feed to lower trimethylamine concentrations in egg yolks ([60]Li et al., 2019), the use of β-cyclodextrin encapsulation technology to extract cholesterol from yolks, which reduces odor but also decreases PUFA content ([61]Tang et al., 2024), enzymatic treatments to enhance protein hydrophobic interactions, electrostatic interactions, and hydrogen bonding, thereby modifying the binding affinity of flavor molecules ([62]Tang et al., 2024), and heat-induced lipid migration, which leads to notable changes in the structure and flavor of yolk gels ([63]Xiang et al., 2020). Recently, microbial fermentation, a widely adopted food processing technique for flavor enhancement, has been explored to improve the quality and taste of eggs ([64]Jia et al., 2023; [65]Jia et al., 2024). Evidence indicates that microbial fermented feed can boost the levels of flavor-enhancing amino acids in duck eggs by modulating the cecal microbiota and reducing yolk TMA concentrations ([66]Tian et al., 2024). Moreover, fermented feed generates a range of metabolites that influence the flavor of duck meat ([67]Xu, He, et al., 2023), pork ([68]Liu et al., 2023), and crab meat ([69]Jiang et al., 2023). Despite these advancements, the specific compounds responsible for the undesirable odor in duck eggs remain unidentified, revealing a gap in understanding how fermented feed influences lipid composition and odor. It is proposed that microbial fermentation feed can modify the lipid profile of duck egg yolks, thus improving the odor by altering the production of volatile flavor compounds. In this study, the volatile flavor and lipidomic profiles of egg yolks were characterized and compared, with an analysis of their correlations to the metabolomic profile of the feed. The objectives include: (1) identifying differential metabolites and pathways of lactobacillus in feed, (2) pinpointing key volatile compounds and lipids in egg yolks, and (3) exploring mechanisms and strategies for odor flavor enhancement. The results offer novel insights into the components responsible for off-odors in egg yolks, establishing a foundation for improving their flavor profile. 2. Materials and methods 2.1. Sample preparation A total of 120 healthy 20-week-old female Shaoxing ducks were acquired from Jinhua Jinwu Agricultural Development Co., Ltd. (Jinhua, China). These ducks were randomly assigned to one of two dietary treatment groups: a commercial feed (CF) group and a fermented feed (FF) group, with five replicates per group and 12 ducks per replicate. FF was prepared by blending 70 % CF with 30 % CF that had undergone anaerobic fermentation using Lactobacillus. CF was sourced from Shanghai Nonghao Feed Co., Ltd. (Shanghai, China). The two experimental groups of laying ducks were raised under uniform conditions, with the only variation being the type of feed administered. One egg was randomly chosen from each replicate, and the egg yolks (EY) were isolated and subjected to a 100 °C water bath for 25 min, producing CF-EY and FF-EY samples corresponding to the respective dietary treatments. After the treatment, yolks were transferred into 50 mL centrifuge tubes and stored at −80 °C for further analysis. The heat treatment protocol for egg yolks was adapted from [70]Tang et al. (2024) and [71]Xiang et al. (2020), with minor modifications. The animals in this study were housed in compliance with the national Laboratory Animal Welfare Guidelines. All experimental procedures received approval from the Animal Use Committee of Zhejiang Academy of Agricultural Sciences (No. 2024ZAASLA027, March 5, 2024). 2.2. Chemicals and reagents The n-alkane standard (C7-C30) was sourced from Sigma-Aldrich (St. Louis, MO, USA). Ethanol (99.8 %) was supplied by Aladdin (Shanghai, China). n-Hexyl-d13 Alcohol (98.5 %) was procured from C/D/N Isotopes INC. (Quebec, Canada). LC-MS grade isopropyl alcohol (IPA) was acquired from Fisher Scientific (Loughborough, UK). LC-MS grade methanol (MeOH) was obtained from Dikma Technologies (51 Massier Lane, USA). Chloroform was provided by Chron Chemicals (Sichuan, China). Ultrapure water was supplied by Watsons (Guangdong, China). 2.3. Analysis of feed composition 2.3.1. Crude components analysis of feed The nutrient composition of CF and FF was compared to evaluate disparities in their dietary contents. Detailed methods and results are provided in supplementary Table S1. 2.3.2. Feed metabolomics analysis Each feed sample (100 mg) was precisely weighed and transferred to a 2 mL EP tube. Subsequently, 1000 μL of extract solution (methanol:water, 4:1) was added, followed by vortexing and a 5-min incubation on ice. The sample was then centrifuged at 15,000 rpm and 4 °C for 20 min. The supernatant was collected and transferred to a fresh glass vial for UHPLC-MS/MS analysis ([72]Want et al., 2006). UHPLC-MS/MS analysis was conducted using a Vanquish UHPLC system (Thermo Fisher, Germany) coupled to an Orbitrap Q Exactive™ HF mass spectrometer (Thermo Fisher, Germany). Samples were injected onto a Hypersil Gold column (100 × 2.1 mm, 1.9 μm) with a 17-min linear gradient at a flow rate of 0.2 mL/min. In positive ion mode, mobile phases consisted of solvent A (water, 0.1 % formic acid) and solvent B (methanol). In negative ion mode, solvent A contained 5 mM ammonium acetate (pH 9.0), and solvent B was methanol. The solvent gradient was as follows: 0–1.5 min, 2 % B; 1.5–12.0 min, 2 % to 100 % B; 12.0–14.0 min, 100 % B; 14.0–14.1 min, 100 % to 2 % B; and 14.1–17.0 min, 2 % B. The Q Exactive™ HF mass spectrometer operated in both positive and negative ion modes, with a spray voltage of 3.2 kV, a capillary temperature of 320 °C, a sheath gas flow of 40 arb, and an auxiliary gas flow of 10 arb. Raw UHPLC-MS/MS data were processed using Compound Discoverer 3.1 (Thermo Fisher) to screen metabolites based on retention time, mass-to-charge ratio (m/z), and other relevant parameters. Peak alignment across different samples was conducted with a retention time deviation of 0.2 min and a mass deviation of 5 ppm to ensure precise identification. Peaks were subsequently extracted using the following criteria: mass deviation of 5 ppm, signal intensity variation of 30 %, signal-to-noise ratio (S/N) of 3, and a minimum signal intensity threshold. Metabolite identification was carried out by comparing high-resolution MS/MS spectra against entries in the mzCloud, mzVault, and primary MassList databases. Identified metabolites were further annotated by comparing the sample's molecular weight (m/z) to data from the Kyoto Encyclopedia of Genes and Genomes (KEGG, [73]https://www.genome.jp/kegg/), Human Metabolome Database (HMDB, [74]https://hmdb.ca/metabolites), and LIPID MAPS ([75]https://www.lipidmaps.org/) databases. 2.4. Analysis of overall volatile flavor 2.4.1. Sensory evaluation Twelve healthy participants (7 females and 5 males, aged 20–26 years) were recruited from Ningbo University to serve as judges. Prior to the experiment, participants underwent training to familiarize themselves with the off-odor. The egg yolk samples were evaluated on a 0–5 scale for off-odor intensity (0, not detectable; 1, very weak; 2, weak; 3, moderate; 4, strong; 5, very strong) ([76]Ren et al., 2022). An independent sample t-test was conducted to analyze the sensory evaluation results, comparing the off-odor between the two egg yolks. All panelists provided informed consent to participate in the sensory evaluation, and ethical approval was not required. 2.4.2. Determination of electronic nose (E-nose) Egg yolks were analyzed using an E-nose (PEN3, Airsense, Schwerin, Germany) to assess the aroma profiles of each sample. For this analysis, 3.0 g of each egg yolk sample were placed in a headspace bottle, incubated at 50 °C for 20 min, and then detected. Five replicates were performed for both CF-EY and FF-EY groups. Experimental parameters included a 1-s sample interval, 60-s flush time, 10-s zero point trim time, 5-s presampling time, 100-s measurement duration, a chamber flow rate of 600 mL/min, and an initial injection flow rate of 400 mL/min ([77]Xu et al., 2024). 2.4.3. SPME/GC × GC-TOFMS analysis The chromatographic column selection and mass spectrometry scan range were informed by a review of the existing literature, with minor adjustments made to the measurement conditions to profile the volatile compounds in egg yolk ([78]Yang et al., 2024). A 5.0 g egg yolk sample was transferred into a 20 mL headspace vial, to which 10 μL of internal standard solution (n‑hexane‑d[13] alcohol at 1 mg/L) was added. The sample was quickly mixed, sealed, and incubated at 60 °C for 10 min. Solid-Phase Microextraction (SPME) using a DVB/CAR/PDMS fiber (50/30 μm × 1 cm, Supelco, Bellefonte, USA) was performed at 60 °C for 40 min to adsorb the sample. The adsorbed compounds were then desorbed in a GC injector at 250 °C for 5 min. Chromatographic conditions: Analyses were performed using a LECO Pegasus® 4D system (LECO, St. Joseph, MI, USA), which integrated an Agilent 8890 A GC (Agilent Technologies, Palo Alto, CA, USA) with a split/splitless injector, a dual-stage cryogenic modulator (LECO), and a TOFMS detector (LECO). The first-dimension column (1D) was an Rxi-5Sil MS (30 m × 250 μm I.D., 0.5 μm) (Restek, USA), while the second-dimension column (2D) was an Rxi-17Sil MS (2.0 m × 150 μm I.D., 0.15 μm) (Restek, USA). The oven temperature program began at 40 °C for 3 min, increased to 180 °C at 5 °C/min, held for 1 min, then ramped to 250 °C at 10 °C/min and held for 3 min. The secondary oven temperature was maintained 5 °C higher than the primary oven, and the modulator temperature was set 15 °C above that of the second column. The modulator operated with a 5 s modulation period. The GC injector temperature was set to 250 °C. Mass spectrometer conditions: Analysis of flavor substances was conducted on the LECO Pegasus BT 4D. The transfer line and TOF MS ion source temperatures were both set to 250 °C. Data acquisition occurred at 200 spectra/s. The mass spectrometer was operated in EI mode at 70 eV with a m/z range of 35–550 and a detector voltage of 2039 V. Raw data were processed using ChromaTOF software and identified with the NIST2020 mass-spectral library. 2.4.4. Analysis of odor activity value (OAV) The odor activity value (OAV) is calculated as the ratio of the odorant's absolute concentration (C) to its sensory threshold (T): OAV = C/T. In this study, the absolute thresholds of volatile compounds in egg yolks were semi-quantified using the internal standard method, while the sensory thresholds were derived from published values for each volatile compound. An OAV <1 within a specific range indicates that the compound likely does not significantly affect the overall odor, whereas an OAV >1 suggests that the compound may influence the odor profile and can be regarded as a key flavor compound ([79]Ren et al., 2021). 2.5. Lipidomics analysis of egg yolks 2.5.1. Lipid extraction Some modifications were implemented following previously established lipid extraction protocols ([80]Narváez-Rivas & Zhang, 2016; [81]Wang, Xiang, et al., 2023). A 50 mg sample was placed into a 2 mL centrifuge tube, to which 750 μL of mixed solvent (chloroform: methanol, 2:1, v/v) was added at −20 °C. The mixture was vortexed for 30 s and then processed in a tissue grinder for 60 s at 50 Hz, repeating the process twice. After grinding, the tube was kept on ice for 40 min. Subsequently, 190 μL of water was added, the mixture vortexed for 30 s, and incubated on ice for an additional 10 min. The sample was centrifuged at 12,000 rpm for 5 min at room temperature, and 300 μL of the organic layer was transferred to a new tube. Another 500 μL of the same mixed solvent (chloroform: methanol, 2:1, v/v) was added, and the centrifugation step was repeated, transferring 400 μL of the organic layer into the same tube. The sample was then concentrated under vacuum until dry. Finally, the residue was dissolved in 200 μL of isopropanol, and the supernatant was filtered through a 0.22 μm membrane to yield samples for LC-MS analysis. 2.5.2. UPLC/MS conditions Chromatographic and mass spectrometric conditions were adapted from prior studies with minor modifications ([82]Hu, Zhao, et al., 2024; [83]Jia et al., 2021; [84]Zou et al., 2022). Chromatographic separation employed an ACQUITY UPLC® BEH C18 (2.1 × 100 mm, 1.7 μm, Waters) column, maintained at 50 °C, with the autosampler set to 8 °C. Gradient elution was performed using acetonitrile: water = 60:40 (0.1 % formic acid +10 mM ammonium formate) (A2) and isopropanol: acetonitrile = 90:10 (0.1 % formic acid +10 mM ammonium formate) (B2) at a flow rate of 0.25 mL/min. After equilibration, 2 μL of each sample was injected. The separation followed this gradient: 0–5 min, 70–57 % A2; 5–5.1 min, 57–50 % A2; 5.1–14 min, 50–30 % A2; 14–14.1 min, 30 % A2; 14.1–21 min, 30–1 % A2; 21–24 min, 1 % A2; 24–24.1 min, 1–70 % A2; 24.1–28 min, 70 % A2. Mass spectrometer conditions: ESI-MSn experiments were conducted with spray voltages of 3.5 kV and 2.5 kV for positive and negative modes, respectively. Sheath gas and auxiliary gas flows were set to 30 and 10 arbitrary units, respectively, while the capillary temperature was maintained at 325 °C. The Orbitrap analyzer scanned a mass range of m/z 150–2000 for full scans at a resolution of 35,000. Data-dependent acquisition (DDA) MS/MS experiments utilized an HCD scan with a normalized collision energy of 30 eV. Dynamic exclusion was applied to eliminate irrelevant data from MS/MS spectra. Lipid annotation was performed using LipidSearch software (V4.2.28) on the raw mass spectrometer data, followed by peak alignment and filtering. To standardize data from varying magnitudes, sum peak normalization was applied for quantitative value correction. 2.6. Statistical analysis Data analysis of feed and egg yolk, along with result visualization, was performed using R (version 3.4.3), Python (version 2.7.6), Origin 2022 (Origin Lab, Massachusetts), and SPSS 26.0 (SPSS Institute, Chicago, USA). Independent sample t-tests were applied to assess inter-group differences, with p < 0.05 indicating statistical significance. Principal component analysis (PCA) and orthogonal partial least squares discriminant analysis (OPLS-DA) were conducted using metaX software. Metabolites exhibiting a variable importance in projection (VIP) score > 1, a p-value <0.05, and a fold change (FC) ≥ 2 or ≤ 0.5 (|log[2]FC| > 1) were classified as differential metabolites. Results were presented as mean and standard deviation (SD). 3. Results and discussion 3.1. Results of feed composition analysis 3.1.1. Results of crude components The crude components of the feed were presented in supplementary Table S1. Compared to the CF group, the FF group exhibited a significant increase in crude ash, crude protein, crude fat, crude fiber, and total amino acid content (p < 0.05). In contrast, total phosphorus and calcium content were significantly reduced (p < 0.05). These results suggest that fermentation induced substantial alterations in the feed's fundamental nutritional composition. 3.1.2. Analysis of metabolic differences in feed A total of 1186 metabolites were identified in the feed through combined positive-negative ion mode analysis, with 823 detected in positive-ion mode and 363 in negative-ion mode. Following data integration, the PCA score plot ([85]Fig. 1A) effectively differentiated between the two feed types. Metabolites were classified by chemical category, with lipids and lipid-like molecules comprising 281 (23.69 %) and organic acids and derivatives making up 178 (15.01 %), representing the primary components ([86]Fig. 1B). Using screening criteria of |log2FC| > 1, OPLS-DA/VIP > 1, and p-value <0.05, 519 differential metabolites were identified, including 288 upregulated and 231 downregulated metabolites ([87]Fig. 1C). The top 30 upregulated and downregulated differential metabolites, exhibiting the largest fold changes, were presented in [88]Fig. 1D. Fig. 1. [89]Fig. 1 [90]Open in a new tab Metabolic differences in feed analysis. (A) PCA score plot of CF and FF. (B) Pie chart depicting the proportion of metabolite classifications. (C) Volcano plot of differential metabolites: Red and blue represent significantly up-regulated and down-regulated metabolites, respectively, meeting the specified thresholds. Gray indicates metabolites that did not meet the screening criteria. Yellowish-brown represents metabolites that met the fold-change threshold but not the p-value threshold (|log[2]FC| ≥ 1, VIP > 1, p < 0.05). (D) Differential changes in the top 30 metabolites, combining up- and down-regulated metabolites. (E) KEGG enrichment map of differential metabolites: The horizontal axis shows the enrichment factor, defined as the ratio of the number of differential metabolites enriched in a specific pathway to the total metabolites in the background dataset; the vertical axis represents the KEGG-enriched functions. Circle size reflects the number of differential metabolites associated with each function. (For interpretation of the references to