Electronic nose, flavoromics, and lipidomics reveal flavor changes in longissimus thoracis of fattening Saanen goats by dietary Allium mongolicum regel flavonoids

Abstract

Lipid type was closely related to meat flavor. The influence of Allium mongolicum Regel flavonoids (AMRF) on the flavor and lipid composition of goat meat is unclear. We integrated electronic nose, flavoromics, and lipidomics analyses to reveal the effects of AMRF supplementation on flavor changes in longissimus thoracis of fattening Saanen dairy goats. Sixteen characteristic volatile compounds were identified as contributors to lamb aroma in the AMRF group. A total of 2014 lipid compounds were quantified with high accuracy, and 26 were identified as key contributors to lamb flavor formation in the AMRF group. Correlation studies revealed that flavor regulation in goat meat was associated with direct AMRF-mediated modulation of lipid synthesis and degradation. It was also linked to indirect modulation of auto-oxidation of unsaturated fatty acid chains in phospholipid molecules, reduced by the intrinsic antioxidant action of AMRF. This dual action exerted a positive influence on goat meat flavor.

Highlights

• •The role of AMRF in the volatile compound and lipid profiles of goats was reported for the first time.
• •2831 volatile compounds were detected in goat meat using HS-SPME-GC × GC-TOFMS.
• •26 differential metabolites were identified, 16 were upregulated and 10 were downregulated.
• •Generation of CVCs is associated with AMRF-mediated regulation of goat lipid composition.

1. Introduction

Lamb is widely consumed worldwide due to its neutral cultural associations, which makes it distinct from beef and pork. The content of essential amino acids such as lysine, arginine, and threonine in lamb is higher than in other meats and meets the Food and Agriculture Organization of the United Nations/World Health Organization evaluation criteria. Its cholesterol content (30–60 mg/100 g) is much lower than that of pork (74.5–126 mg/100 g) and beef (75 mg/100 g) (Webb, 2014). The global production of sheep meat was 16.5 million tons in 2023. Lamb accounted for 10.272 million tons, while goat meat contributed 6.368 million tons. The global production of goat meat is further expected to rise in the future (data from the United States Department of Agriculture, USDA). Lamb offers high nutritional value, yet its global consumption remains low. According to USDA data, lamb accounted for only 1.71 % of total meat production in 2023. Lamb production is directly linked to consumer demand. The distinctive flavor and odor of meat are the key reasons limiting its consumption (Kaffarnik et al., 2015; Watkins et al., 2021). Historically, flavor has been the dominant factor influencing lamb purchases, followed by tenderness and juiciness. In contrast, beef consumers prioritize tenderness over juiciness and flavor (Lee et al., 2025).

People's perception of meat flavor is the combined effects of olfaction and taste, specifically the interaction of non-volatile and volatile compounds with the human sense of taste, olfactory chemoreceptors, and other sensory networks (Kaffarnik et al., 2015). The type (i.e. aldehydes, ketones, esters, alcohols, and organic heterocyclic compounds), concentration, and balance of volatile compounds are critical to meat flavor acceptance, while the composition and structure of meat influence the generation and release of these compounds (Bravo-Lamas et al., 2018). In particular, fat deposits and lipid categories in meat act as solvents for flavor compounds and undergo lipid oxidation and autoxidation, accompanied by Maillard and Strecker reactions (Xu et al., 2023). Lipids are small hydrophobic or amphiphilic molecules classified into six major groups: fatty acyls (FAs), glycerophospholipids (GPs), sphingolipids (SPs), glycerolipids (GLs), sterol lipids (STs), and prenol lipids (PLs). GPs include lysophosphatidylcholines (LPCs), phosphatidylcholines (PCs), phosphatidylethanolamines (PEs), phosphatidylserine (PSs), phosphatidylglycerol (PGs), phosphatidylinositol (PIs), and cardiolipin (CLs). SPs include sphingomyelins (GMs), phytosphingosines (phSMs), and ceramides (Cers) (Fahy et al., 2005). Raw meat odor and cooked meat flavor are closely related to lipid-derived compounds present in raw meat due to lipid oxidation and degradation (Li et al., 2020). Numerous studies have examined lamb flavor after processing and cooking, such as grilling (Cheng et al., 2024) and dry curing (Guo et al., 2019, Guo et al., 2022). Many flavor compounds, such as unsaturated fatty acids (UFAs) and flavor amino acids, either originate in raw meat fat or form through the Maillard reaction and persist after cooking (Yang et al., 2024). However, relatively few studies have explored the pathways behind raw lamb flavor development, particularly how changes in diet composition influence on flavor production.

Numerous studies indicate that flavonoids and polyphenolic compounds derived from natural plants have strong antioxidant properties capable of enhancing animal growth performance, particularly meat flavor. Moreover, the fatty acid and amino acid composition determines meat flavor and nutritional quality. Many natural substances can elevate UFA content, reduce their oxidation, and eliminate free radicals, thereby improving mutton flavor (Liu, Tang, & Ao, 2022; Zhao et al., 2023). Purple corn anthocyanin (a natural phenolic compound) enhances both the diversity and relative abundance of aroma compounds in lamb and enriches flavors related to plants, herbs, fats, and fruits (Tian et al., 2021). The basic chemical structure of anthocyanin is 3,5,7-trihydroxy-2-phenyl benzopyran, which efficiently traps peroxyl radicals during lipid oxidation, thereby preventing excessive lipid oxidation and increasing UFA content (Tian et al., 2021). Lipid peroxidation typically generates various off-flavor volatile compounds. Specifically, lipid hydroperoxides decompose into hydroxyl and alkoxyl groups. Subsequently, fatty acid chains attached to alkoxyl groups split and yield low molecular weight volatile compounds, including alcohols, ketones, and aldehydes (Qiu et al., 2018). Excessive lipid oxidation in lamb likely results in UFA loss, negatively affecting lamb nutritional value and flavor, shortening shelf life, and reducing consumer purchasing interest. Our previous study indicated that dietary supplementation with Allium mongolicum Regel (AMR) and its extracts significantly enhances the antioxidant activity of longissimus thoracis (LT) muscle in Small-tailed Han sheep (Liu, Tang, & Ao, 2022) and Angus calves (Liu et al., 2024). This supplementation effectively regulates lipid composition, increasing UFA content in muscle, aligning with flavonoid and polyphenol properties of strong radical scavenging and lipid metabolism regulation. Specifically, AMR flavonoids (AMRF, 2.8 g/d/animal) significantly reduced the content of characteristic flavor and odor (CFO) compounds in lamb, thereby enhancing the overall meat flavor in captive sheep (Liu, Hui, et al., 2022). AMR powder at various concentrations can induce muscle fiber conversion in LT muscle from Type II (fast-twitch glycolytic) to Type I (slow-twitch oxidative). This conversion enhances volatile compound generation, including esters (ethyl acetate), ketones (1-penten-3-one), and organic heterocyclic compounds (tetrahydrofuran), which are key flavor substances associated with UFAs in triglycerides (TGs) (Liu et al., 2024). However, no known studies have specifically reported AMRF effects on lipid composition and lamb flavor in fattening male dairy goats.

The dairy goat male lamb fattening industry is a major initiative aimed at meeting the consumer market demand for goat meat. Research on flavor regulation in goat lamb is an important current theme. We hypothesized that AMRF could modulate the lipid composition of goat meat, reduce the processes involved in lipid oxidation, and improve goat meat flavor. Therefore, this work aimed to (i) determine the effects of AMRF addition on the flavor profile of lamb meat using the electronic nose technique; (ii) identify the characteristic volatile compounds (CVCs) and key differential lipid molecules in goat meat; and (iii) determine the relationship between the identified CVCs and the key differential lipid molecules using Spearman correlation analysis. This work contributes to enhancing the flavor of goat meat and increasing consumer demand in the goat meat market.

2. Materials and methods

2.1. Preparation of AMRF

Fresh AMR was harvested in May 2024 in the Wuwei area (36° 29′ N, 104° 16′E), Gansu, China. The washed leaves of AMR were placed into a constant temperature drying oven (DZF-GW, Shanghai Binglin Electronic Technology Co., Ltd., Shanghai, China) at 60 °C. After reaching a constant weight, the leaves were ground with a herbal grinder (CWF-300S, Top Calendar Medical Equipment, Zhejiang, China), sieved through a 1 mm mesh, and converted into dry AMR powder (AMRP). The AMRP was stored at 4 °C in a refrigerator for later use. The extraction of AMRF included six processes: degreasing and decoloration, warming, filtration, concentration, drying, and pulverization, strictly performed as described by Ding et al. (2021). In this study, the extraction rate of AMRF was 28 %. Ultra-performance liquid chromatography-electrospray ionization-tandem mass spectrometry was used to detect the relative content of main active ingredients in AMRF and the results are shown in Supplemental Table 1.

2.2. Animal feeding and management

The experimental procedures and animal care followed the guidelines of the Gansu Agricultural University Animal Care Committee for the Protection of Animals Used for Scientific Purposes (GSAU-Eth-AST-2022-001). The study was conducted at the experimental base of Baicao Plateau, Huanxian County, Qingyang City, Gansu Province. The farm lies at an altitude of 1655 m above sea level, with latitude and longitude coordinates of 36° 56′N and 107° 31′E, respectively. The regional climate is predominantly cold and semi-arid, featuring a mean annual precipitation of 300 mm and a mean annual temperature of 9.2 °C. Twelve healthy Saanen milk goats (male, initial average body weight = 16.38 ± 0.74 kg, 3.0 ± 0.1 months) underwent stratification into two nutritional groups through a randomization protocol: the control (C) group was fed with the basal diet, and the AMRF (A) group was fed the basal diet supplemented with 2.8 g of AMRF per animal per day. The amount of AMRF was determined based on previous studies involving fattening Small-tailed Han sheep rams (Liu, Hui, et al., 2022). Each goat remained individually housed in a separate pen. Animals underwent feeding for 139 days, comprising a pre-experimental period of 15 days and a fattening period of 124 days. The fattening period included three phases: phase I (Day 1 to 30), phase II (Day 31 to 60), and phase III (Day 61 to 124). The ingredients and chemical composition of the basal diet are listed in Supplemental Table 2. Additionally, fatty acid composition data for the basal diets are listed in Supplemental Table 3.

2.3. Muscle sampling

All 12 Saanen dairy goats were transported to a commercial slaughterhouse (Huanxian Zhongsheng Sheep Industry Development Co., Ltd., Qingyang, China) and slaughtered following halal procedures at the end of the fattening period (final weight, 38.03 ± 3.57 kg) by a qualified imam. The LT muscle samples were collected from the 9th through 12th ribs of the left carcass and fat was trimmed. LT samples obtained from each goat were equally divided into 3 portions, packed in vacuum-sealed bags (12 cm × 20 cm), and immediately stored in an ultra-low-temperature refrigerator (DW-86L490J, Haier Group Company, Shenzhen, China) at −80 °C. The stored samples were used for electronic nose analysis, volatile compound analysis based on HS-SPME-GC × GC-TOFMS, and absolute quantitative lipidomics.

2.4. Electronic nose analysis

The bionic olfactory data from LT samples were collected using a PEN3 portable electronic nose (NOVA Enose3, Beijing Innovate Technology Development Co., Ltd., Beijing, China) as described by Zhou et al. (2024). Samples (3 g) were accurately weighed into a 15 mL headspace vial and equilibrated at 25 °C for 30 min. The machine parameters were set as follows: sampling interval, 1 s; sensor self-cleaning time, 100 s; injection flow rate, 400 mL/min; and detection time, 70 s.

2.5. Profiles of volatile compounds in LT by HS-SPME-GC × GC-TOFMS

2.5.1. Extraction and analysis of volatile compounds

For volatile compound analysis, a 2 g LT sample was placed in a 20 mL headspace vial with 10 μL of n-Hexyl-d13 (1 mg/L) added as an internal standard solution, followed by incubation at 80 °C for 10 min. Before extraction, the SPME fiber (DVB/C-WR/PDMS, Switzerland) was aged at 270 °C for 10 min, then transferred to the incubation chamber to adsorb compounds from the sample at 80 °C for 25 min. After adsorption, the SPME fiber was transferred to the GC injector, desorbed at 250 °C for 5 min, and then aged again at 270 °C for 10 min following injection. After sampling, the SPME fiber was aged for 10 min at 270 °C, and then 10 μL of n-alkanes was transferred into a 20 mL headspace injection bottle, incubated, extracted, and injected into the sample. Analysis was conducted using a LECO Pegasus BT 4D (LECO, St. Joseph, MI, USA) GC × GC-TOFMS system (Agilent Technologies, Palo Alto, CA, USA) equipped with a dual-stage injection modulator and a split/unsplit sampling module. Chromatographic column models and the relevant setup parameters are listed below: the one-dimensional column was DB-Heavy Wax (30 m × 250 μm × 0.5 μm) (Agilent, USA), and the two-dimensional column was Rxi-5Sil MS (2 m × 150 μm × 0.15 μm) (Restek, USA). High-purity helium was used as a carrier gas at a constant flow rate of 1.0 mL/min. For the 1D column DB-Heavy Wax (30 m × 250 μm × 0.5 μm), the initial temperature was 50 °C for 2 min, then it increased to 230 °C at 5 °C/min, and maintained for 5 min. The temperature program for the 2D column Rxi-5Sil MS (2 m × 150 μm × 0.15 μm) was higher than for the 1D column, and the modulator temperature was constantly 15 °C higher than the temperature of the 2D column. A modulation period of 6.0 s and an inlet temperature of 250 °C were used. The system included a high-resolution LECO Pegasus BT 4D mass detector (LECO, St. Joseph, MI, USA), operated with a mass spectrometry transmission line at 250 °C, an ion source at 250 °C, a spectral acquisition rate of 200 spectra/s, an electron bombardment source of 70 eV, a detector voltage of 1960 V, and a mass spectral scanning range of m/z 35–550.

2.5.2. Qualitative and quantitative analysis of volatile compounds and calculation of relative odor activity value (ROAV)

The raw data from the downcomer were qualitatively analyzed for volatile compounds using the Chroma TOF software based on the NIST2020 database. The sensory flavors of samples were analyzed and compared using Flavordb. Volatile compounds were structurally identified by combining: (a) mass spectral matching against the NIST reference library, (b) linear retention index (LRI) calibration with n-alkanes (C₇-C₃₀), and (c) comparison with authenticated flavor standards. Two-dimensional volatile characterization was performed through olfactory attribute tracking combined with retention index mapping. The detection frequency (DF) metric statistically validated aroma recurrence, where a critical DF value of ≥6 confirmed the presence of CVCs (Zhou et al., 2024). A minimum peak summation required a signal-to-noise (S/N) ratio of 10, retaining only volatile compounds with positive match values exceeding 700 (Stefanuto et al., 2017). Compounds detected in less than half of the samples were excluded, half-minimum values replaced zero values, and the data were normalized using an internal standard for magnitude comparison across samples. A previously published ROAV method (Fan et al., 2022) was used to evaluate each volatile compound for its flavor contribution in the LT samples. The volatile compound making the largest overall flavor contribution was assigned ROAV = 100. For other volatile compounds, ROAV was calculated as follows:

$$\text{ROAV}\left(\mathrm{A}\right)=100\times \left[\text{Relative Content}\left(\mathrm{A}\right)/\mathrm{T}\left(\mathrm{A}\right)\right]/\left[\text{Relative Content}\left(\text{stan}\right)/\mathrm{T}\left(\text{stan}\right)\right]$$

where Relative Content (A) represents the normalized quantitative value of the measured volatile compound; T(A) indicates the odor minimum value for the measured volatile compound (obtained from the odor database and literature data); Relative Content (stan) indicates the normalized quantitative value for volatile compounds assigned ROAV = 100; and T(stan) represents the odor minimum value (obtained from the odor database and literature data) for volatile compounds assigned ROAV = 100. Higher ROAV values indicate greater volatile compound contribution to the overall lamb flavor.

2.6. Lipidomic analysis

2.6.1. Sample pretreatment

Lipids were subjected to absolute quantification using a Q-Exactive Plus (Thermo Scientific) mass spectrometer connected to an ultra-high performance liquid chromatography (UHPL CNexera LC-30 A) system. We used the methyl tert-butyl ether method as described by Matyash et al. (2008), to extract lipids and cover the full chemical space of lipid hierarchies, by applying polarity-adjusted extraction protocols and charge-state modulated ionization efficiencies. We accurately weighed 40 mg of LT sample, added 200 μL water, 800 μL MTBE, and 240 μL pre-cooled methanol, vortexed the mixture, and sonicated it for 20 min in a low-temperature water bath. The sample was then left at room temperature for 30 min. Subsequently, samples were centrifuged at 14,000g for 15 min at 10 °C, after which we collected the upper layer of the organic phase and dried it under nitrogen gas. We then added 200 μL of 90 % isopropanol/acetonitrile solution for mass spectrometry analysis and vortexed thoroughly. Next, 90 μL of the compound solution was collected and centrifuged at 14,000g for 15 min at 10 °C, and 10 μL of the supernatant was analyzed.

2.6.2. Absolute quantification analysis of LT lipids

Analytical separation was conducted using ultra-high-performance liquid chromatography (Nexera LC-30A, Shimadzu) equipped with a C18 chromatographic column (1.7 μm, 2.1 mm × 100 mm). The analysis was performed at a flow rate of 300 μL/min and a temperature of 45 °C. The mobile phases consisted of phase A: an aqueous acetonitrile solution (acetonitrile to water ratio of 6:4, v/v) and phase B: an acetonitrile-isopropanol solution (acetonitrile to isopropanol ratio of 1:9, v/v). The gradient elution program was set as follows: from 0 to 3.5 min, phase B remained at 40 %; 3.5–13 min, phase B increased gradually from 40 % to 75 %; 13–19 min, phase B remained between 75 % and 99 %; 19–24 min, phase B returned to 40 %.

Electrospray ionization (ESI) was used in both positive and negative ion modes. The samples were separated by UHPLC and analyzed by mass spectrometry using a QExactive series mass spectrometer (Thermo Scientific). The ESI source conditions were set as follows: heater temperature at 300 °C, sheath gas flow rate of 45 arb, auxiliary gas flow rate of 15 arb, sweeping gas flow rate of 1 arb, spray voltage of 3.0 KV, capillary temperature at 350 °C, and MS1 scanning range from 200 to 1800. The mass-to-charge ratios (m/z) of lipid molecules and lipid fragments were recorded using 10 fragmentation profiles (HCD MS2 scans) collected after each full scan. The resolution was 70,000 for MS¹ scans at m/z 200, and 17,500 for MS² scans at m/z 200. Lipid annotation was performed on the raw data obtained by LipidSearch software. A data matrix containing peak response values (intensity), retention time (rt), and m/z was generated by pre-processing both lipid molecules and internal standard lipid molecules through steps including peak alignment, peak recognition, and peak extraction. Absolute quantification analysis of volatile compounds was performed using the isotopic internal standard method (Internal Standards: SPLASH® LIPIDOMIX MASS SPRC STANDARD, AVANTI, 330707-1EA). This involved calculating the ratio of the response abundance of the measured substance to the internal standard (peak area ratio) and the concentration of the internal standard. All annotation results from the samples were further identified (secondary identification) by LipidSearch software.

2.7. Statistical analysis

Statistical analyses were conducted using one-way ANOVA with SPSS (version 22.0; SPSS Inc., Chicago, IL, USA) to assess differences due to diet treatments at P < 0.05 (*P < 0.05, **P < 0.01, ***P < 0.001). The experimental data are presented as mean ± standard error. Raw LC-MS datasets were subjected to chemometric evaluation using MetaboAnalyst 5.0 (www.metaboanalyst.ca). Orthogonal partial least squares-discriminant analysis (OPLS-DA) was performed, with VIP >1.5 used as the significance threshold. The absolute content of CVCs and key lipid metabolisms in LT for groups C and A was visualized using GraphPad Prism 5.01 software. Differential volatile compounds and differential lipids are represented as heat maps using the Hiplot online version (https://hiplot.com.cn/). Relationships between 16 CACs and 26 key lipid compounds were analyzed using Pearson correlation, and P < 0.05 was regarded as statistically significant.

3. Results and discussion

3.1. Electronic nose analysis of LT sample

Subtle variations in flavor between samples can be identified through the detection of characteristic flavor compounds by each sensor of the electronic nose (Zhao et al., 2025). The electronic nose analyzer, equipped with 10 sensors (Fig. 1A), was used to distinguish LT samples from different sources to assess differences in the combined flavor profiles of groups C and A. As shown in Fig. 1B, changes in the radar plot area of LT samples indicated that dietary AMRF supplementation can influence the volatile compounds in lamb. The E-nose analyzer differentiated LT samples primarily using W1S, W1W, W2S, and W2W sensors.

Fig. 1.

Performance description of electronic nose (E-nose) sensors (A), radar map (B), principal component analysis loading polt (C), and score plot (D) of E-nose response data in the basal diet without additives and after dietary flavonoids from dried Allium mongolicum Regel leaf (AMRF) supplementation the left longissimus thoracis in fattening Saanen diary goats. ^⁎P < 0.05, ^⁎⁎P < 0.01; C, a basal diet without additives; A indicate 2.8 g per Saanen diary goat per day of AMRF supplementation in the basal diet, respectively. The number of observations for each mean value was three.

To further investigate and clarify the flavor changes caused by AMRF in LT samples, principal component analysis (PCA) was performed using the response values of E-nose sensors, and the results are shown in Fig. 1C–D. As shown in Fig. 1C, the first two principal components explained approximately 81.10 % of the total variance. This indicates that most odor information in LT samples can be explained by these two components. LT samples from group C were located in the right quadrant and showed positive associations with W1C, W5C, W3C, and W6S, which are sensitive to aromatic compounds, short-chain alkane aromatics, ammonia, and hydride compounds. In contrast, LT samples from group A appeared in the lower left quadrant and were positively associated with methyl (W1S), sulfides, pyrazines, many terpenes (W1W), alcohols, aldehydes, ketones (W2S), and organic sulfide compounds (W2W). These findings align with our past results, which showed that AMRF dietary supplementation (22 mg/kg diet) or 20 g/animal/d of AMR leaves powder (AMRP) significantly enhanced the flavor of lamb and beef from Small-tailed Han sheep (Liu et al., 2019) and Angus calves (Liu et al., 2024). This enhancement may relate to the pungent odor (dimethyl disulfide, methyl propyl disulfide M) and active ingredients present in AMRP (Zhang et al., 2022). The results also demonstrate that the electronic nose is a useful tool for distinguishing lamb flavor attributes. However, it remains difficult to determine the specific profiles of volatile compounds in these LT samples.

3.2. Volatile compound profile analysis in LT by HS-SPME-GC × GC-TOFMS

3.2.1. Visual topographic plots of flavor in LT

Notably, after AMRF was added, the volatile compounds in LT became more abundant, and the intensity of signal peaks significantly increased (Supplemental Fig. 1A, B, C, and D). Meanwhile, quantitative descriptive analysis (QDa) showed that the flavor of the lamb became richer (Shen et al., 2025). As shown in Supplemental Fig. 1E, ten key sensory attributes were recorded, including sweet, green, fruity, waxy, fatty, nutty, fresh, woody, herbal, and floral. The LT sample in group A had a more green, fruity, and floral aroma than that in group C. The onion and musty notes in LT from group A was more prominent. Previous findings also confirmed the important role of AMR and its extracts in modulating the production of volatile compounds in fresh livestock products (especially lamb and beef), as well as delaying the development of bad odors during storage and imparting a specific flavor recognized by consumers. Specifically, the addition of AMRF (2.8 g/lamb/d) to the diets of fattening small-tailed Han sheep significantly reduced CFO substances in lamb, which is the main reason consumers refuse lamb (Liu & Ao, 2021). The CFO substances in lamb are mainly medium-chained branched-chain fatty acids. The production of these substances is related to the biohydrogenation of UFAs by rumen microorganisms. Vasta et al. (2019) reported that plant-derived flavonoids and polyphenols generally inhibit Gram-positive fibrinolytic bacteria and ciliated protozoa, and can modulate biohydrogenation of UFAs through changes in rumen microbiota composition without affecting fiber digestion. The results confirmed that active ingredients such as flavonoids, flavonols, flavanones, isoflavones, and other bioactives in AMRF have strong antioxidant effects, which interfere with biohydrogenation by rumen microorganisms and reduce the generation of CFO substances (Liu, Hui, et al., 2022). Recent studies found that AMR contains substances with aromatic odors, such as isobutyl butyrate (fruity), isopentyl formate (aromatic; fruity), and acetophenone (aromatic; sulfurous), which may be deposited in beef, thereby enriching the flavor of Angus calves (Liu et al., 2024), consistent with E-nose results. Wan et al. (2023) reported that some special Allium plants, Illicium griffithii, and Pimpinella Tibet grow in the pastoral areas of Tibet, China. A specific flavor substance, anethole, was detected in grazing cattle meat after feeding, which rendered the beef aniseed flavor, further confirming that meat flavor can be enriched by the consumed plant feeds. Furthermore, flavonoids deposited in meat can manipulate changes in myofibril type. This is significantly correlated with meat quality and flavor (Xu et al., 2022). Myofibrillar proteins (MPs), although tasteless, can have their flavor-binding capacity altered by flavonoids, which in turn affects meat flavor (Sun et al., 2024). The atomic force microscopy microstructure analysis showed that eugenol formed rough, irregular shapes on the MPs surface, mainly due to eugenol–MP interaction through hydrogen bonding and hydrophobic forces. This led to an increase in MPs particle size and uneven distribution (Sun et al., 2024). Notably, eugenol binding to MPs buried the binding sites of volatile compounds such as hexanal, heptanal, and octanal, weakening interactions with aldehydes and enhancing flavor release (Wang & Arntfield, 2014).

3.2.2. Quantitative analysis of volatile compounds in LT

A total of 2831 volatile compounds were identified in LT from both group C and A. Specifically, 1375 volatile compounds were detected in LT from group C, comprising 222 hydrocarbons (8.38 %), 122 esters (13.66 %), 110 alcohols (21.86 %), 81 ketones (10.01 %), 56 heterocyclic compounds (4.79 %), 47 aldehydes (11.97 %), 43 carboxylic acids (2.23 %), and 694 other substances (27.10 %). In contrast, LT from group A had 1456 volatile compounds, including 262 hydrocarbons (8.80 %), 125 esters (12.56 %), 99 alcohols (26.20 %), 91 ketones (13.41 %), 57 heterocyclic compounds (5.43 %), 52 aldehydes (8.32 %), 46 carboxylic acids (1.78 %), and 724 other substances (23.50 %) (Fig. 2A and B). Li et al. (2022) reported that aldehydes, ketones, and alcohols are the main volatile compounds in lamb. These compounds are more easily perceived due to lower sensory thresholds. This observation aligned with our findings. Notably, the release of flavor substances from lamb was strongly influenced by lipid species and fatty acid composition. The supervised OPLS-DA results showed a clear separation of data points representing volatile compounds characterizing lamb flavor in Groups A and C (Fig. 2C), indicating significant differences. OPLS-DA reduced model complexity while enhancing explanatory power without reducing predictive ability. This allowed a better view of group differences. Therefore, we used OPLS-DA to conduct model replacement testing and screen differential volatile compounds. As shown in Fig. 2D, we validated model reliability by evaluating predictive accuracy and goodness-of-fit. The model showed R² = 1 and Q² = 0.3, confirming it was reliable and free from overfitting.

Fig. 2.

Volatile compound of the left longissimus thoracis in fattening Saanen dairy goats between group C and A identified with Headspace Solid-Phase Microextraction with Comprehensive Two-Dimensional Gas Chromatography Time-of-Flight Mass Spectrometric (HS-SPME-GC × GC-TOFMS). Types and quantities of volatile compounds (A). Proportion of contents of each kind of classifications in samples (B). OPLS-DA score polt (R2X = 0.441, R2Y = 0.999, Q2 = 0.710) (C), and corresponding OPLS-DA validation (D) of the test samples. Heat map analysis of 63 differential volatile compounds (E). The content histograms for the 9 classifications are based on 63 differentially volatile compounds (F). The screening conditions were VIP ≥ 1 and P < 0.05. C, a basal diet without additives; A indicate 2.8 g per Saanen dairy goat per day of AMRF supplementation in the basal diet, respectively. The number of observations for each mean value was six (n = 6).

Based on OPLS-DA screening criteria (variable importance projection; VIP > 1 and t-test; P < 0.05) 63 significantly different volatile compounds were identified (Supplemental Table 4). We visualized and analyzed qualitative and quantitative differences between groups C and A. Among the 63 compounds, 16 were down-regulated and 47 were up-regulated in group A (Fig. 2E), including 15 organoheterocyclic compounds, 14 esters, 7 alcohols, 7 hydrocarbons, 5 ketones, 3 benzenoids, 3 heterocyclic compounds, 2 organosulfur compounds, and 7 others. The classification and content of volatile compounds are shown in Fig. 2F. Ketones were the most predominant volatiles in LT samples from both groups, accounting for 51.14 % in group C and 65.76 % in group A. Further analysis revealed that AMRF treatment altered the volatile profile of lamb samples. Group A showed a significant (P < 0.05) increase in total esters (138.95 ± 29.13 μg/kg vs. 47.22 ± 8.89 μg/kg), hydrocarbons (37.37 ± 1.03 μg/kg vs. 22.62 ± 4.95 μg/kg), and ketones (434.87 ± 79.73 μg/kg vs. 159.13 ± 70.35 μg/kg). A highly significant (P < 0.01) decrease in alcohols (21.68 ± 4.89 μg/kg vs. 68.26 ± 13.26 μg/kg) was also observed in LT of group A (Fig. 2F).

Our previous study found that supplementing the diet of Angus calves with AMRP during fattening significantly increased the levels of esters, ketones, and organic heterocyclic compounds in beef, positively influencing beef flavor (Liu et al., 2024). Short-chain esters have a fruity flavor. Long-chain esters can impart a fatty flavor to lamb. Esters are formed by esterification of alcohols and carboxylic acids in muscle, and the extent of this biochemical reaction depends on esterase activity (Guo et al., 2019). The addition of AMRF significantly elevated ester content in lamb. We hypothesized that the active ingredient in AMRF may have accelerated the esterification rate and reduced esterase activity. One of our recent studies found that AMRP significantly reduced the accumulation of acetic acid, which provides pungent, cheesy, and vinegary odors to beef. It also facilitated the esterification of acetic acid and alcohol compounds under non-enzymatic conditions, increasing ethyl acetate, which imparts fruity and floral flavors in beef (Liu et al., 2024). This may explain the observed decrease in alcohol content. The synthesis of esters in muscle cell membranes helps maintain plasma membrane fluidity and regulates intracellular redox homeostasis (Mason & Dufour, 2000). This supports the observed improvement in total antioxidant capacity, total superoxide dismutase activity, and glutathione peroxidase activity in LT of lamb due to AMRF (Liu, Tang, & Ao, 2022). Li et al. (2024) reported that dietary supplementation with prickly ash seeds, which are rich in flavonoids and phenolic compounds, significantly improved muscle metabolism in Hu sheep. Metabolomics revealed enrichment in nucleotide and purine pathways. Metabolites such as inosinic acid, guanosine, and inosine participated in muscle energy metabolism and contributed to the development of muscle flavor. This intervention significantly improved the meat quality and flavor of lamb.

There is direct evidence that thermo-oxidative degradation of free UFAs in meat produces a wide range of aldehydes, alcohols, and ketones. Ketones often have creamy or fruity notes and are important for forming the typical lamb flavor (Yang et al., 2024). Previous studies showed that the addition of AMR extracts, enriched in flavonoids, essential oils, and polysaccharides, to lamb diets promoted n-3 PUFA deposition in LT of lamb. This may explain the increased ketone content observed in the current experiments. The supplementation also enriched free amino acids such as glutamate and glycine. These are known as precursors of flavor compounds and undergo Maillard reactions with soluble reducing sugars to produce such compounds (Zhao et al., 2023). Ketone production is also linked to amino acid degradation (Biller et al., 2016). This suggests that AMRF supplementation can increase the accumulation of flavor precursors, including UFAs and amino acids, in lamb, which impacts flavor. Liquid chromatography-electrospray ionization mass spectrometry (LC-ESI-MS) was used to monitor the oxidative degradation process. The results confirmed that the generation of ketones, aldehydes, and alcohols is associated with the lipid origin of UFAs. The oxidation degree of unsaturated fatty acyl groups is lower in phosphatidylcholines (PCs) than in TGs, due to the poorer thermo-oxidative stability of the sn-2 site (Zhou et al., 2014). Our previous study revealed that dietary supplementation with AMR extract, rich in flavonoids, essential oils, and polysaccharides, could promote flavor development (Liu et al., 2024). Hydrocarbons, which have a higher sensory threshold and minimal impact on lamb flavor, are mainly formed by the homolytic cleavage of alkoxylates from fatty acids (García-González et al., 2008).

3.2.3. Screening of CVCs in LT

In this study, 16 CVCs were screened and identified using HS-SPME-GC × GC-TOFMS and ROAV values, with mass spectrometry databases, linear retention index, and reference flavor standards (Supplemental Table 5). The dilution factor value and odor descriptions of the 16 CVCs are shown in Fig. 3A. These included 5 organoheterocyclic compounds, 3 esters, 2 benzenoids, 2 organosulfur compounds, 1 alcohol, 1 hydrocarbon, 1 aldehyde, and 1 ketone. In group A, ethyl dodecanoate (sweet, waxy, floral), 2-ethylhexanol (rose, green), ethyl hexadecanoate (wax), ethyl tetradecanoate (sweet, waxy), acetone (sweet, fruity, etherous), thiazole (nut, sulfur, stink), 2-methyl-5-(methylthio)-furan (mustard, onion), dimethyl sulfide (asparagus), and methyl-pyrazine (roasted almonds) each had ROAV values of 6. As shown in Fig. 3B, among these CVCs in LT from groups C and A, the concentrations of followings were higher: pyridine (O1), ethyl dodecanoate (O2), benzene (O3), ethyl hexadecanoate (O5), ethyl tetradecanoate (O6), propene (O7), acetone (O9), thiazole (O10), tetramethylpyrazine (O11), 2,3-dimethylpyrazine (O13), dimethyl sulfide (O14), methyl-pyrazine (O15), and aniline (O16). Conversely, these were lower: 2-ethylhexanol (O4), 2-propenal (O8), and 2-methyl-5-(methylthio)-furan (O12). The flavor profile of lamb depends not only on the type and concentration of volatile compounds but also on their odor threshold. The ROAV reflects the contribution of each compound to overall flavor intensity, with higher ROAV indicating a stronger effect (Zhu et al., 2020). 2-ethylhexanol (O4, C = 5.16 × 10⁻¹ vs. A = 1.43 × 10⁻¹), ethyl hexadecanoate (O5, C = 1.47 × 10⁻² vs. A = 4.16 × 10⁻²), and dimethyl sulfide (O14, C = 2.33 × 10⁻³ vs. A = 3.32 × 10⁻³) had high ROAV values in both groups. This study established a multidimensional analytical platform combining HS-SPME, comprehensive GC × GC-TOFMS, and ROAV modeling. This platform identified 16 CVCs as aroma-active markers that define LT aromatic characteristics. The levels of O1, O2, O3, O5, O6, O7, O9, O10, O11, O13, O14, O15, and O16 were significantly higher (P < 0.05) in group A than in group C.

Fig. 3.

Characteristic volatile compounds of the left longissimus thoracis in fattening Saanen dairy goats between group C and A identified with Headspace Solid-Phase Microextraction with Comprehensive Two-Dimensional Gas Chromatography Time-of-Flight Mass Spectrometric (HS-SPME-GC × GC-TOFMS) and ROAV. The detection frequency (DF) value and odor descriptions of 16 characteristic volatile compounds detected using HS-SPME-GC × GC-TOFMS (A). Concentration and ROAV of 16 key volatile compounds (B). ^⁎P < 0.05; ^⁎⁎P < 0.01; ^⁎⁎⁎P < 0.001.

Sixteen CVCs were identified, among which five were organoheterocyclic compounds: pyridine, thiazole, tetramethylpyrazine, 2,3-dimethylpyrazine, and methyl-pyrazine. These are major contributors to the aroma of lamb. This shows consistency with research on grilled lamb aroma by Chen et al. (2023). Pyrazines, such as 2,5-dimethylpyrazine and 2,3,5-trimethylpyrazine, significantly contribute to lamb aroma due to their low threshold and high sensory detectability. The production of similar compounds in raw lamb may result from changes in nonvolatile flavor precursors. We previously showed that AMRP and extracts of AMRP significantly increased PUFA and free amino acid levels in lamb (Liu et al., 2019; Zhao et al., 2023). These precursors were subjected to temperatures above 85 °C in the headspace injection unit of the gas chromatograph, triggering a Maillard reaction. However, a high-temperature environment is not essential for the Maillard reaction. Reducing sugars and amino acids in meat can react under non-enzymatic conditions even at low temperatures (Whitfield & Mottram, 1992). This explains why raw lamb can produce reactions similar to those of cooked meat. Ai-Dalali et al. (2022) identified oxidative lipid decomposition in cured raw beef, the Maillard reaction, thiamine degradation, and phenolics from curing spices as the main sources of alcohols, aldehydes, ethers, furans, and pyrazines in raw meat. In this study, benzene and aniline were significantly higher. Li et al. (2020) reported volatile compounds like benzene and toluene in Hu sheep psoas major muscle. These compounds do not cause meat odor but may contribute to raw lamb flavor.

Esters in meat are formed through esterification between free UFAs and lipid-degrading alcohols (García-González et al., 2013). Wang et al. (2021) detected esters in the LT of Jingyuan lambs at different ages, including ethyl acetate and amyl acetate in six-month-old lambs, and butyl butyrate in two-month-old lambs. These esters mainly contribute to the fruity flavor of lamb. The content of esters with long-chain fatty acids (ethyl dodecanoate, ethyl hexadecanoate, and ethyl tetradecanoate) was significantly higher in our study. This increase contributed to a distinctive fruity flavor, consistent with the electronic nose results. We also observed a significant decrease in 2-ethylhexanol content, a key substrate for ester synthesis. Simultaneously, ester compound levels rose significantly, which may explain the reduced alcohol content. Aldehydes are major degradation products of lipid oxidation, and acrolein levels were significantly reduced in lamb meat from group A. Acrolein has a pungent odor, indicating that AMRP can reduce undesirable odors and enhance overall flavor acceptability. We suggest that apigenin, quercetin, and isorhamnetin in AMRP exhibit strong antioxidant activity. These phenols scavenge free radicals during oxidation, show transition metal chelation, and release hydrogen atoms to stabilize radicals, forming non-radical products. They inhibit UFA oxidation and thereby reduce odor (Wang et al., 2019). Similar studies have confirmed that purple corn anthocyanins, a polyphenolic compound with natural antioxidant properties, inhibit plasma lipid metabolism in goats, thereby modulating the formation of meat flavor substances (Tian et al., 2021). Ren et al. (2020) reported that Tamarix ramosissima bark extract, rich in isorhamnetin, hispidulin, and cirsimaritin, significantly reduced phenylacetaldehyde formation in grilled lamb. Phenylacetaldehyde is a precursor of 2-amino-1-methyl-6-phenylimidazo[4,5-b]pyridine (PhIP), a highly carcinogenic compound in grilled meat. These flavonoids formed complexes—8-C-(E-phenylethenyl)isorhamnetin, 6-C-(E-phenylethenyl)isorhamnetin, and 8-C-(E-phenylethenyl)hispidulin—with phenylacetaldehyde, thereby preventing PhIP formation. Qi et al. (2022) reported that dimethyl trisulfide, with a low odor threshold, is a primary contributor to the sulfuric and meaty flavor of lamb. Parker (2015) explained that dimethyl trisulfide originates from the thermal degradation of S-methyl L-cysteine and its sulfoxide. Higher dimethyl sulfide content in group A versus group C (P < 0.05) aligns with electronic nose results. Allium plants are rich in characteristic sulfides with a distinct aroma, influencing lamb flavor (Zhang, Zhang, et al., 2024). Therefore, variations in these volatile compounds likely underlie aroma differences between groups C and A.

3.3. Lipid compound analysis

3.3.1. Absolute quantification of lipids in LT

To achieve absolute quantification of lipid compounds, we analyzed the lipid composition of LTs in groups C and A. Deuterium-labeled internal standards and untargeted lipidomics were used on a Q-Exactive Plus mass spectrometer equipped with a UHPLC Nexera LC-30A. A total of 2014 lipid compounds were identified in group C and A LT samples, including 1169 in positive ion mode and 845 in negative ion mode. As shown in Fig. 4A and B, these lipids were classified into 41 subclasses, including 410 (20.36 %) TGs, 307 (15.24 %) PCs, 210 (10.43 %) PEs, 210 (10.43 %) Cers, and 192 (9.53 %) diglycerides (DGs). Other subclasses included 121 (6.01 %) CLs, 76 (3.77 %) PSs, 70 (3.48 %) hexosyl ceramides (Hex1Cer), 67 (3.33 %) SMs, and 46 (2.28 %) PIs. Further subclasses consisted of 32 (1.59 %) PGs, 30 (1.49 %) hexosyl ceramides (Hex2Cers), 29 (1.44 %) acylcarnitines (AcCas), 25 (1.24 %) LPCs, 19 (0.94 %) gangliosides (GM3), 17 (0.84 %) lysophosphatidylethanolamines (LPEs), 17 (0.84 %) sphingosine bases (SPHs), and 17 (0.84 %) wax esters (WEs). In addition, 16 (0.79 %) monoglycerides (MGs), 11 (0.55 %) phosphatidic acids (PAs), 10 (0.50 %) (O-acyl)-1-hydroxy fatty acids (OAHFAs), 10 (0.50 %) zymosterols (ZyEs), 9 (0.45 %) hexosyl ceramides (Hex3Cers), and 8 (0.40 %) sulfatides (STs) were identified. Rare subclasses included 6 (0.30 %) N-acetylhexosyl ceramides (CerG2MNAc1s), 6 (0.30 %) phosphatidylinositol(4,5)bisphosphates (PIP2s), 5 (0.25 %) lysophosphatidylglycerols (LPGs), 5 (0.25 %) lysophosphatidylinositols (LPIs), 5 (0.25 %) phSMs, 5 (0.25 %) fatty acids, and 5 (0.25 %) coenzyme Qs (Cos). Additional minor lipids included 4 (0.20 %) sphingosine phosphates (SPHPs), 3 (0.15 %) ceramide phosphates (CerPs), 2 (0.10 %) cholesterol esters (ChEs), 2 (0.10 %) gangliosides (GD3s), 2 (0.10 %) phosphatidylinositol(3,4,5)triphosphates (PIP3s), and 5 (0.30 %) others. Notably, TGs, PCs, and PEs were key determinants of lipid composition in lamb. As shown in Fig. 4B, all lipids were further classified into six major classes: 865 (42.95 %) GPs, 618 (30.69 %) GLs, 452 (22.44 %) SPs, 61 (3.03 %) FAs, 13 (0.65 %) SLs, and 5 (0.25 %) PLs. GPs and GLs exhibited the highest contents, aligning with the findings of Li et al. (2022). Among all lipids, the six major classes showed no significant difference (P > 0.05) in content between the two groups (Fig. 4C). However, some subclasses displayed notable differences (Fig. 4D). Compared to group C, group A had significantly higher contents of HexCer (A = 6.58 ± 0.31 μg/g vs. C = 5.61 ± 0.86 μg/g) and Others (A = 7.65 ± 1.83 μg/g vs. C = 5.33 ± 0.81 μg/g) (P < 0.05). In contrast, phSMs (A = 58.09 ± 24.40 μg/g vs. C = 193.94 ± 37.61 μg/g) and SMs (A = 380.81 ± 18.54 μg/g vs. C = 425.58 ± 19.62 μg/g) were highly significantly lower in group A (P < 0.01). These results indicate that AMRF addition may alter lamb lipid composition and content, possibly leading to observed changes in lamb flavor. Similarly, Yu et al. (2024) found that addition of perilla (rich in flavonoids, terpenes, polyphenols, and amino acids) to Tan sheep (Ovis aries) diets was found to increase the content of ω-3 PUFAs and decrease the ω-6/ω-3 ratio, while decreasing the LPCs, increasing the Cos, and improving the flavor of the lamb.

Fig. 4.

Numbers of lipid species identified in 6 classes and 41 lipid subclasses (A). Percentages of each lipid subclass and class (B). The content of differential lipid molecules and their proportions in six major classes (GPs, GLs, SPs, FAs, SLs, and PLs) between group C and A (C). The content of differential lipid molecules and their proportions in differential lipid subclasses (Hex3Cers, phSMs, SMs, and others) between group C and A (D). ^⁎P < 0.05, ^⁎⁎P < 0.01 respectively. OPLS-DA score plots (E) and corresponding OPLS-DA validation (F) of the test samples in positive ionization mode. OPLS-DA score plots (G) and corresponding OPLS-DA validation (H) of the test samples in negative ionization mode. VIP score plots of OPLS-DA in the positive and negative ionization modes (I), the screening conditions were VIP > 1 and P < 0.05. The heatmap (J) of the 26 key lipid metabolites of longissimus thoracis samples in the C and A groups. C, a basal diet without additives; A indicate 2.8 g per Saanen milk goatf per day of AMRF supplementation in the basal diet, respectively. The number of observations for each mean value was six (n = 6).

3.3.2. Identification of key lipid compounds

To distinguish lipid species and LT distribution in group C and group A samples and screen for differentially enriched lipid compounds, we used multivariate statistics to analyze the lipidomics data (supervised OPLS-DA). The OPLS-DA model simplifies complexity and enhances explanatory strength without reducing predictive power. This maximizes group C and A separation, improves interpretability, and maintains model validity. Based on these outcomes, subsequent model testing and differential lipid screening were conducted using OPLS-DA results. As shown in Fig. 4 E and G, with AMRF intervention, the lipid composition of groups C and A separated under both positive and negative ion modes, indicating distinct lipid classes in LT samples. To confirm OPLS-DA model reliability, cross-validation and permutation tests were performed. Fig. 4F and H validation metrics: R²/Q² = 0.60/−0.681 (ESI⁺) and 0.511/0.192 (ESI⁻), demonstrating consistent reliability across ionization modes. Multivariate screening (Student's t-test, P < 0.05; OPLS-DA VIP > 1.0) identified 26 key lipid compounds differentiating groups C and A in lamb LT (Fig. 4I). Volcano plot analysis (Fig. 4J) revealed 16 significantly upregulated and 10 downregulated lipids in group A versus group C, highlighting quantitative divergence in LT lipidome (VIP > 1; P < 0.05). As shown in Supplemental Table 6, these 26 key lipids included 23.08 % PEs, 23.08 % PCs, 15.37 % PIs, 7.69 % Cers, 7.69 % MGs, 7.69 % DGs, 3.85 % phSM, 3.85 % LPE, 3.85 % AcCa, and 3.85 % GM1.

The 26 lipid molecules with the highest 39-fold change and P < 0.05 are summarized in Fig. 5A (positive ion mode) and B (negative ion mode). Phospholipids, as functional lipids, form a major component of all cell membranes and participate in several physiological and biochemical functions in muscle tissue. Lipid droplets, located in the endoplasmic reticulum, are covered by protein-rich phospholipid monolayers. These droplets play a key role in cellular energy metabolism and in vivo homeostasis. Endoplasmic reticulum phospholipids and surface tension determine lipid droplet formation (Ben M’barek et al., 2017). Regarding nutrient uptake, PCs are more effective PUFA carriers than triacylglycerols. When incorporated in membrane phospholipids, PUFAs exhibit greater cell permeability and bioavailability than when linked to triacylglycerols (Le Grandois et al., 2010).

Fig. 5.

Absolute quantification of 10 key lipid metabolites in positive ionization mode (A), absolute quantification of 16 key lipid metabolites in negative ionization mode. C, a basal diet without additives; A indicate 2.8 g per Saanen dairy goat per day of AMRF supplementation in the basal diet, respectively. ^⁎P < 0.05, ^⁎⁎P < 0.01, ^⁎⁎⁎P < 0.001 respectively. The number of observations for each mean value was six (n = 6).

In PEs, it was found that six types of PE molecules, all five PEs (including PE(37:2e) + Na; PE(16:0p_20:3) + H; PE(18:0_18:0) + Na; PE(14:1e_18:2)-H; PE(36:3e)-H) of group A, were significantly upregulated (P < 0.05), except for PE(34:2e)-H, which was significantly downregulated (P < 0.05). For PCs, except for PC(18:1e_22:5) + HCOO, which was significantly downregulated (P < 0.05), PC(35:5) + H, PC(12:1e_18:2)-CH3, PC(18:0e_20:4) + HCOO, and PC(16:0e_19:1) + HCOO were all significantly upregulated (P < 0.05). Additionally, PC(18:2e_18:2) + HCOO was highly significantly upregulated (P < 0.01) in our study. Lipoxygenase-catalyzed peroxidation of structural lipids (TG, PC, PE) and FFAs generates key volatile precursors. GC–MS quantification showed that 68 % of quantified aldehydes and ketones originated from ω-3/6 oxidative cleavage (Bravo-Lamas et al., 2018). sn-2 positional enrichment of ω-6 PUFAs (C18:2n6, C20:4n6) in glycerophospholipids (PC, PE) serves as a primary substrate reservoir for LOX-catalyzed oxidation. This reaction generates 72 % of characteristic aldehydes in beef (e.g., hexanal, 2-nonenal) via β-scission pathways (Zhou et al., 2024). In goat meat, although the specific β-scission pathway has not been fully characterized, it can be hypothesized that it bears some resemblance to similar processes in beef. However, due to the higher content of n-3 PUFAs in the fatty acid composition of lamb, the oxidation products are likely to be more diverse, which may have a unique impact on the formation of lamb flavor. Meanwhile, during thermal processing, PC and PE undergo hydrolysis to produce LPC and LPE, two lipid subclasses potentially associated with meat flavor retention. This suggests that the generation of CVCs in group A correlates with increased levels of PEs and PCs in the lipid subclasses. Phospholipids, especially PE and PC, are rich in polyunsaturated fatty acids (PUFAs) such as linoleic acid (C18:2n6) and arachidonic acid (C20:4n6), which are primary substrates for LOX-catalyzed oxidation. The lipoxygenase (LOX) pathway initiates the peroxidation of PUFAs at the sn-2 position of phospholipids, generating hydroperoxides. These hydroperoxides undergo further cleavage via β-scission reactions to form volatile aldehydes (e.g., hexanal, 2-nonenal) and ketones, which are critical contributors to meat flavor (Zhou et al., 2024). In our study, the upregulation of PE and PC species (e.g., PE(18:0_18:0), PC(18:2e_18:2)) in the AMRF group suggests enhanced availability of PUFA substrates for LOX-mediated oxidation, thereby promoting the generation of desirable flavor compounds. From our results, AMRF appears to reduce the conversion of PE to LPE during heat treatment, as LPE(18:2e)-H decreases significantly (P < 0.05), allowing for the preservation of lamb flavor. Another important reason is that PCs and PEs exhibit greater thermo-oxidative stability at the sn-2 site than other lipid classifications. Wu and Wang (2019) reported that the lipids PC, PE, and TG, with saturated fatty acids at the sn-1 site and unsaturated fatty acids at the sn-2 site, showed distinct patterns in lipid thermo-oxidation. The sn-2 site in PC and PE demonstrated greater thermo-oxidative stability compared to that in TG. These lipids may influence the formation of lipid-oxidized volatile compounds.

In the PI subclass, PI(18:0_22:3)-H and PI(16:0_20:4)-H from group A showed highly significant upregulation (P < 0.01); PI(18:0_20:3)-H and PI(18:0_22:4)-H were significantly upregulated (P < 0.05). Li et al. (2020) reported that intramuscular fat (IMF) deposition in Hu sheep correlated with increased levels of 14 lipids: TGs, DGs, PCs, PEs, FFAs, PIs, PSs, LPCs, PGs, SMs, Cers, cardiolipin, cos, and AcCas. These lipids were attributable to AMRF. Based on this, we hypothesized that AMRF supplementation could increase lamb IMF, and this was confirmed. Similarly, Zhou et al. (2021) revealed that saturated fatty acids, PIs, and PSs contributed to IMF deposition, using proteomics and lipidomics analysis in male Xidu black pigs. Our previous study found that 33 mg/kg AMRF supplementation (based on body weight) significantly increased IMF in LT of Small-tailed Han sheep, along with higher eicosapentaenoic acid, PUFAs, and MUFAs (Liu et al., 2019). Therefore, we further hypothesized that AMRP may influence flavor substance generation by increasing the content of PCs, PEs, and PIs in lamb meat.

Cers are a class of amide compounds formed by dehydration between long-chain fatty acids and amino groups of sphingolipids. They exist in all eukaryotic cells and play a major role in regulating cell differentiation, proliferation, apoptosis, aging, and other life processes (Zheng et al., 2019). In addition, excessive Cer accumulation promotes muscle cell apoptosis, and CerP (d18:1_18:0) content was significantly upregulated during cold storage of Mongolian sheep. This indicates its role in post-slaughter cell death and apoptosis (Zhang et al., 2023). In the Cer subclass, Cer(d32:0) + H and Cer(d12:0_24:2) + HCOO, and in the GM1 subclass, GM1(t28:0)-2H, were significantly downregulated in group A (P < 0.05). This suggests that AMRF may delay muscle cell apoptosis. Our earlier study found that adding ethanol-soluble AMR extract to diets effectively extended lamb shelf life, mainly due to the strong antioxidant activity of its active ingredients (Liu, Hui, et al., 2022). Lipid molecules play an important role in apoptosis mediation. Cers, CLs, and PSs may induce mitochondrial reactive oxygen species, which decrease membrane potential and impair the oxidative–antioxidant system of muscle (Zou et al., 2022). AMRF appears to mitigate this effect.

MGs are indirectly involved in phospholipid synthesis during muscle energy metabolism (Zhou et al., 2021). In our study, MG(23:1) + NH4, MG(19:1) + H, DG(21:1e) + NH4, and DG(19:1e) + Na of group A were all significantly downregulated (P < 0.05), a result consistent with the observed up-regulation of most PC, PE, and PI molecules. On the one hand, we found that the downregulated DG and MG species all contain unsaturated bonds susceptible to conversion into fatty acids catalyzed by TG lipase. These fatty acids then enter the oxidation pathway and form precursors of volatile compounds in lamb meat (Xu et al., 2023). On the other hand, after animal slaughter, lipases (e.g., monoacylglycerol lipase, hormone-sensitive lipase) may retain temporary activity in muscle cells. MGs are further hydrolyzed by these enzymes, producing free fatty acids (FFAs) and glycerol. FFAs enter mitochondria for energy generation via the carnitine transporter system, while glycerol is phosphorylated into glycerol 3-phosphate by glycerol kinase and enters glycolysis. A recent study found that high AMR levels induced a transition from type II (fast glycolytic fibers) to type I (slow oxidative fibers) and significantly increased hypothermic pH in slaughtered Angus heifers. Further studies confirmed that significantly increased malate dehydrogenase activity and decreased lactate dehydrogenase activity in muscle tissues, both linked to aerobic metabolism, were key factors responsible for elevated muscle pH (Liu et al., 2024). Therefore, we hypothesized that AMRF introduction could alter the post-slaughter muscle acid-base environment and enhance lipohydrolase activity. However, no change in enzyme activity was detected, and whether AMRF could influence muscle fiber transformation in goats remains to be investigated.

AcCas are acetylated derivatives of carnitine that act in muscle through multiple biochemical mechanisms, mainly involving energy metabolism, antioxidant defense, and signal regulation (Hettema & Tabak, 2000). In the current study, the content of AcCa(18:2) + H was significantly upregulated in group A (P < 0.05). This finding may be explained by the carnitine shuttle system, through which AcCas transport long-chain fatty acids into the mitochondrial matrix for β-oxidation. AcCas also act as a buffer pool for acetyl groups, regulating the acetyl coenzyme A/coenzyme A balance that supports ATP supply in the cell (Zhang et al., 2023). However, cellular respiration ceases, and the mitochondrial oxidative phosphorylation pathway becomes inhibited. This condition may be the main reason for AcCas enrichment during early postmortem. AcCas may function as precursors of lipid-derived volatile compounds and are preferentially oxidized in the early postmortem period. This is related to the degree of unsaturation in their structure. AcCas may also serve as potential biomarkers for predicting the storage process in cooled postmortem lambs. This role further illustrates how lipid molecules modulate flavor changes in postmortem lamb (Xu et al., 2023).

3.4. Correlation analysis of key lipid compounds and CVCs

To further understand AMRF-driven changes in lipid composition and the influence of lipid degradation and oxidation reactions on the formation of volatile compounds in lamb meat, 26 key lipid compounds and 16 CVCs were selected for correlation analysis. The results are shown in Fig. 6. All nine up-regulated CVCs (O1, O2, O3, O5, O6, O9, O13, O15, O16) showed significant positive correlations with most PEs, PCs, and all PIs, AcCa (AcCa(18:2) + H), and GM1 (GM1(t28:0)-2H), excluding PE(34:2e)-H and PC(18:1e_22:5) + HCOO. These CVCs also had significant negative correlations with DGs, MGs, LPE (LPE(18:2e)-H), and phSM (phSM(t38:3) + HCOO). In contrast, the other two identified volatile compounds (O4, O12) exhibited the opposite trend (Supplemental Table 7).

Fig. 6.

Spearman's correlation heatmap showing the correlation between key lipid compounds and characteristic volatile compounds. The colour represent correlation coefficients, with red indicating a positive correlation and blue indicating a negative correlation. ^⁎P < 0.05, ^⁎⁎P < 0.01 ^⁎⁎⁎P < 0.001, respectively. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.) (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

In organoheterocyclic compounds, O1 correlated positively with PC(18:2e_18:2) + HCOO and negatively with phSM(t38:3) + HCOO. O10 showed a negative correlation with MG(19:1) + H. O11 was positively correlated with PE(18:0_18:0) + Na. O13 exhibited a positive correlation with PI(16:0_20:4)-H (ρ = 0.741; P = 0.006) and a negative correlation with phSM(t38:3) + HCOO. O15 showed a positive correlation with PC(18:2e_18:2) + HCOO and a negative correlation with phSM(t38:3) + HCOO.

In esters, O2 had a positive correlation with PC(18:2e_18:2) + HCOO and a negative correlation with phSM(t38:3) + HCOO. O5 showed a positive correlation with PI(18:0_22:3)-H and a negative correlation with PI(16:0_20:4)-H. O6 was positively correlated with PI(18:0_22:3)-H and negatively correlated with phSM(t38:3) + HCOO. In organosulfur compounds, O12 was positively correlated with DG(21:1e) + NH4 and negatively correlated with GM1(t28:0)-2H. O14 showed a negative correlation with GM1(t28:0)-2H. In benzenoids, O3 exhibited a positive correlation with PC(18:2e_18:2) + HCOO (ρ = 0.818; P = 0.001) and a negative correlation with phSM(t38:3) + HCOO. O16 had a positive correlation with PI(18:0_22:3)-H and a negative correlation with phSM(t38:3) + HCOO. In alcohols, O4 was positively correlated with Cer(d12:0_24:2) + HCOO and PC(18:1e_22:5) + HCOO, and negatively correlated with PI(16:0_20:4)-H. In hydrocarbons, O7 showed a positive correlation with PC(35:5) + H. In ketones, O9 exhibited a positive correlation with PE(18:0_18:0) + Na and a negative correlation with DG(21:1e) + NH4.

There is direct evidence that lipid oxidative degradation significantly influences the production of aroma compounds. Specifically, organic heterocyclic compounds and carbonyls (i.e., aldehydes and ketones) are considered products of Maillard reactions, lipid oxidation, and/or a combination of the two. Sulfur-containing compounds result from reactions involving sulfur-containing amino acids. Hydrocarbons and alcohols are generated through lipid oxidation (Bravo-Lamas et al., 2018). Zhou et al. (2024) reported that up-regulated differential lipid compounds such as PCs and PEs [i.e., PC(16:1e_20:4), PC(16:0e_20:4), PC(18:2e_18:2), and PE(16:1e_20:4)] contribute to the formation of esters, ketones, and organohybrids. These are key aroma compounds in beef. Similarly, numerous studies have found that CVCs in roasted lamb are linked to the oxidative degradation of phospholipid molecules such as PCs, PEs, and PIs. In vitro modeling of the Maillard reaction confirmed that meat aroma improves with the addition of phospholipid molecules (Cheng et al., 2024; Liu, Hui, et al., 2022). Therefore, the lipid composition of goat meat and the saturation level of its fatty acids directly affect the type and concentration of volatiles produced by fat oxidation. This, in turn, strongly influences meat flavor. We hypothesized that supplementing goat diets with AMRF could regulate lipid composition in LT and affect meat flavor directly or indirectly. In recent years, several studies have described how flavonoids and polyphenols regulate muscle lipids in animals. Plant-derived polyphenols influence lipid metabolism by modifying intestinal microbial diversity, activating the energy transducer AMP-activated protein kinase (AMPK), promoting mitochondrial biosynthesis, and reducing oxidative stress. Zhang, Zhai, et al. (2024) and Zhang, Zhang, et al. (2024) showed that grape seed proanthocyanidins, a natural polyphenolic compound, altered lipid and fatty acid composition in LT through fat metabolism, cholesterol metabolism, and AMPK pathway activation in fattening pigs. Our previous studies also confirmed that AMRF regulates lipid metabolism in sheep, promotes MUFA and PUFA accumulation in LT, and reduces the content of methyl-branched fatty acids (4-methyloctanoic acid, 4-methylnonanoic acid, and 4-ethyl octanoic acid), which are associated with CFO in lamb, thereby improving lamb flavor (Liu & Ao, 2021; Liu, Hui, et al., 2022).

The above studies found that the formation of 16 CVCs may correlate strongly with UFA chains in phospholipid molecules. These chains are highly vulnerable to attack by free radicals and reactive oxygen species because of their instability. They oxidize to form hydrogen peroxide, which ultimately undergoes a complex reaction to form volatile compounds (Domínguez et al., 2019). This finding is consistent with Al-Dalali, who reported that phospholipids are significantly and positively correlated with most volatile compounds in raw beef. For instance, PC(16:1/(9Z/16:0)) was significantly associated with hexanal, 4-carvonmethanol, hexanoic acid, 1-hexanol, and benzaldehyde (Ai-Dalali et al., 2024). GC–MS analyses identified five key UFA-derived flavor classes: aldehydes (lipid oxidation markers), alcohols (mushroom aroma), ketones (fruity notes), esters (creamy profiles), and heterocyclics (roasted characteristics) (Xu et al., 2023; Zhou et al., 2014). Results from several correlation studies confirm that raw meat odor and cooked meat flavor are closely linked to lipid-derived compounds. Lipid oxidation plays a critical role in this relationship (Resconi et al., 2018). In a study on Xinjiang dry-cured mutton ham, GPs and SPs metabolism were highlighted as key metabolic pathways in lamb flavor formation (Guo et al., 2022). In our study, we found that phospholipid-like molecules (precursors of volatile compounds) were significantly more abundant than in the control group. However, the type and content of characteristic compounds such as alcohols and aldehydes decreased. We hypothesized this was related to AMRF, which contains several antioxidant components that inhibit lipid oxidation in raw meat, particularly that of phospholipids. Tian et al. (2021) reported that anthocyanins can transfer from feed to goat muscle and affect muscle lipid metabolism. Specifically, the phenolic hydroxyl group of anthocyanins strongly inhibits oxygen radicals and donates hydrogen atoms, thereby reducing peroxide value and protecting lipid integrity. Additionally, anthocyanins may preserve lipids through activation of peroxisome proliferator-activated receptors. They may also influence lipid composition in LT by activating these receptors, inhibiting stearoyl-CoA desaturase and lipoprotein lipase, and ultimately altering lamb flavor. Meanwhile, under frozen storage conditions (−18 °C), pyrolysis and lipolysis reactions in raw meat persist. These processes primarily generate organic nitrogenous compounds, lipids, and organic oxygenates, while the Maillard reaction halts (Ai-Dalali et al., 2024). However, Tsujihashi et al. (2022) presented a different view, stating that even at −24 °C, the Maillard reaction continues. These findings suggest that volatile compound generation in raw meat is influenced by lipid oxidation, lipid degradation, pyrolysis, and the Maillard reaction. In conclusion, dietary supplementation with AMRF in fattening dairy goat diets positively influenced goat meat flavor. This effect was associated with changes in lipid abundance. The relationship between this modulation, direct (alteration of fat synthesis and degradation) or indirect (antioxidant activity), and volatile compound formation needs further investigation. Future studies should also examine how flavonoids and phenolics interact with food matrices such as proteins, as this may affect meat flavor release and retention. AMRF, as a natural plant extract, has a remarkable potential for use in animal feeding. In contrast to other plant extracts (e.g. rosemarinic acid), AMRF not only possesses a strong antioxidant capacity, but also directly influences the production of flavor substances by modulating lipid metabolism. Its high extraction rate make it a promising feed additive for enhancing the flavor and nutritional value of meat products. Future studies could further explore the effects of AMRF in different animal species and its synergistic effects with other feed additives to promote its commercialization.

4. Conclusion

The effect of dietary AMRF on the production of CVCs was accurately investigated through alterations in the molecular composition and lipid content in LT of fattening dairy goat males. Techniques such as the electronic nose, HS-SPME-GC × GC-TOFMS, and the absolute quantitative lipidome were used. A total of 2831 volatile compounds were detected. Among these, 16 flavor compounds were defined as CVCs. The content of organoheterocyclic compounds, esters, and benzenoids was significantly higher. The electronic nose results also indicated improved lamb flavor. A total of 2014 lipid compounds were identified and quantified by lipidomics. Among these, 26 key differential lipid molecules were identified. AMRF may improve lamb flavor by influencing lipid composition, such as phospholipids, and metabolite production in LT of goats. By reducing the oxidation of unsaturated fatty acids in phospholipids, AMRF indirectly contributes to the formation of desirable flavor compounds. This dual action—direct modulation of lipid synthesis and degradation, and indirect modulation of lipid oxidation—exerts a positive influence on goat meat flavor. This finding provides a scientific and theoretical basis for developing AMRF-based strategies to improve flavor in goats. In future research, the regulation of body fat deposition in goats by AMRF will be investigated at cellular and molecular levels. The direct and indirect effects of AMRF on flavor production in goat meat will also be explored.

CRediT authorship contribution statement

Wangjing Liu: Writing – original draft, Visualization, Validation, Resources, Methodology, Investigation, Data curation. Aihuan Yu: Resources. Yaodi Xie: Resources. Xiao Zhang: Data curation. Beibei Guo: Data curation. Lei Xu: Resources. Wenliang Tao: Resources. Ruixin Yang: Resources. Chenxu Sun: Resources. Jiang Hu: Resources. Zhaomin Lei: Writing – review & editing, Validation.

Funding

This work was supported by the National Natural Science Foundation of China (grant no. 32260846; 32402789); Gansu Provincial Science and Technology Major Project (grant no. 25ZDNA008); Gansu Provincial Department of Education, Industrial Support Program Project (grant no. 2024CYZC-36); Agriculture Research System of China (grant no. CARS-38); Science and Technology Support Project for Modern Cold and Arid Agriculture Seed Industry Breakthrough (grant no. ZYGG-2025-15); Discipline Team Project of Gansu Agricultural University (grant no. GAU-XKTD-2022-22).

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

Supplementary material

Data availability

The data that has been used is confidential.