Multi‐Omics Reveals Phenethyl Acetate and Its Producer Lactiplantibacillus plantarum as Key Drivers of Enhanced Palatability in Alfalfa Silage
Zhihui Fu, Tianwei Wang, Jiaqi Zhang, Wenzhao Wang, Xiumin Zhang, Kaixuan Wei, Muhammad Tahir, Jin Zhong
TL;DR
This study shows that using Lactiplantibacillus plantarum and phenethyl acetate together improves the flavor and palatability of alfalfa silage, leading to higher feed intake in sheep.
Contribution
The study introduces a novel strategy for synergistic fermentation using Lactiplantibacillus plantarum and phenethyl acetate to enhance silage flavor and feed intake.
Findings
Lactiplantibacillus plantarum and phenethyl acetate together significantly improved silage fermentation quality and feed intake.
Key flavor compounds like phenylethyl alcohol and β-damascenone were increased with Lactiplantibacillus plantarum and phenethyl acetate.
Specific enzymes from Lactiplantibacillus plantarum played a crucial role in flavor compound formation during fermentation.
Abstract
High‐quality silage enhances palatability and feed intake; however, the effects of co‐fermentation with flavouring agents and lactic acid bacteria (LAB) on its flavour quality, core microbiota, and taste‐active amino acids remain unclear. This study investigated the effects of fermentation using Lactiplantibacillus plantarum (LP) alone or in combination with phenethyl acetate (LPP) on the flavour profile of alfalfa silage and its subsequent influence on feed intake. Both LP and LPP significantly improved fermentation quality versus control (CK), with markedly higher feed intake—LP > CK and LPP > LP. Key flavour compounds, including dimethyl trisulfide, 4‐ethylphenol and β‐damascenone, were significantly increased in the LP alone group. Contrarily, essential taste‐related amino acids including aspartic acid, alanine, proline, histidine, isoleucine, and valine were decreased, except for…
Genes, proteins, chemicals, diseases, species, mutations and cell lines named across the full text — each resolved to its canonical identifier and authoritative record.
Click any figure to enlarge with its caption.
FIGURE 1
FIGURE 2
FIGURE 3
FIGURE 4
FIGURE 5| Treatment (Test vs Reference) | Items | Test silage | Reference silage | SEM |
|
|---|---|---|---|---|---|
| LP VS CK | Feed preference (%) | 69.35 | 30.65 | 6.55 | < 0.001 |
| Average DM intake (g/day) | 213.55 | 97.59 | 19.94 | < 0.05 | |
| LPP VS LP | Feed preference (%) | 67.77 | 32.23 | 6.19 | < 0.001 |
| Average DM intake (g/day) | 182.85 | 90.98 | 18.55 | < 0.05 |
- —National Natural Science Foundation of China10.13039/501100001809
- —Strategic Priority Research Program of the Chinese Academy of Sciences
- —Science and Technology Program of the Inner Mongolia Autonomous Region
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Taxonomy
TopicsRuminant Nutrition and Digestive Physiology · Milk Quality and Mastitis in Dairy Cows · Plant and fungal interactions
Introduction
1
Alfalfa ( Medicago sativa L.) is one of the most widely cultivated forage legumes worldwide, due its high crude protein content and palatability (Liu, Chen, et al. 2024; Liu, Du, et al. 2024). Ensiling, a key method for preserving fresh alfalfa for ruminants (Zhou et al. 2024), relies on anaerobic microbial activity, primarily driven by the competition and synergy of lactic acid bacteria (LAB) (Guo et al. 2018). This microbial interactions not only help preserve the nutritional components of the raw material but also enhance its fermentation quality and palatability (Wang et al. 2024). Silage quality directly influences the production performance and health of ruminants. Flavour is a critical determinant of silage quality (Liu, Du, et al. 2024), encompassing both volatile aroma compounds and non‐volatile taste‐contributing amino acids. These components not only affect the feed intake of animals but are also associated with the quality and flavour of livestock products (Zhang et al. 2021). According to Scherer, Gerlach, and Südekum (2019), within the very first minutes of feeding, goats rely on the olfactory assessment of silage to guide their recognition and subsequent feeding decisions. Manzocchi et al. (2021) reported that the feeding of silage influences the flavour attributes and compositional makeup of milk and cheese. Silage flavour research forms a vital link between silage production and the livestock industry, as well as livestock products (including milk, cheese, etc.). Therefore, understanding the mechanisms of flavour formation in high‐quality silage is significant for promoting the development of animal husbandry and ensuring a stable supply of premium, flavorful livestock products to consumers.
Recent studies have reported on the flavour compounds in various silages (Deng et al. 2023; Liu, Du, et al. 2024; Zhang et al. 2021). Among these studies, particular attention has been paid to the role of microbial communities in shaping the flavour profiles of silage. Zhang et al. (2021) demonstrated that the microbial community significantly influences the volatile metabolites produced during silage fermentation of stylo and rice straw silage with or without L. plantarum . Liu, Du, et al. (2024) found that the metabolism of LAB influences the composition of volatile organic compounds in silage. In our previous multi‐omics analysis (Fu et al. 2025), phenethyl acetate was identified as the only aromatic ester significantly increased in high‐quality alfalfa silage and L. plantarum was a major contributor to its production. Notably, spraying phenethyl acetate onto alfalfa silage at 50 mg/kg was associated with increased feed intake. Phenethyl acetate is a well‐characterised aromatic compound widely present in various fermented foods (Fabricio et al. 2025; Paszkot et al. 2023; Wang et al. 2022) and is commonly used as a flavouring agent in the food industry (Sekar et al. 2022). However, its application as a fermentation additive in silage production, particularly in combination with microbial co‐fermentation strategies, remains unexplored. Scherer et al. (2021) investigated the feeding preferences of goats for red clover silage versus alfalfa silage and identified differential metabolites between preferred and avoided silages, but their study did not explore the relationship between these differential metabolites and the core microorganisms in silage. Furthermore, there is a lack of research focusing on non‐volatile taste amino acids in the flavour profile of silage. The synthesis and metabolism of numerous volatile and non‐volatile flavour compounds are intricately linked to microbial metabolism (Xia et al. 2023). Integrating metagenomics, flavoromics and metabolomics enables comprehensive elucidation of microbial community structure, metabolic dynamics and functional profiles during the fermentation process (Peng, Cheng, Lin, et al. 2025). Although multi‐omics technologies have been applied to alfalfa silage (Liu, Du, et al. 2024), studies specifically investigating the formation of key flavour compounds remain limited. Therefore, such research is essential to understanding how microorganisms contribute to flavour profiles and to identifying the critical enzymes involved in these metabolic pathways.
Based on the observed feeding preference of sheep for alfalfa silage in different treatment groups, this study aimed to investigate how the co‐fermentation of L. plantarum and phenethyl acetate influences the flavour profile of high‐quality alfalfa silage, compared to L. plantarum alone. To achieve this, we employed an integrative approach combining metagenomics, flavour substance analysis and microbial community profiling to characterise the metabolic pathways of key volatile aromatic compounds and taste‐free amino acids, and to identify their associations with core fermentative microorganisms. The results are expected to inform the development of synergistic fermentation strategies using aromatic compounds and microbial agents to improve silage palatability and feed intake, thereby promoting efficient utilisation of high‐quality forage resources.
Materials and Methods
2
Raw Materials and Alfalfa Silage Preparation
2.1
The third regrowth of alfalfa (at the booting stage) was harvested on August 11th, 2024, from the Zhuozhou Agricultural Science and Technology Park of China Agricultural University in Zhuozhou City, Hebei Province, China (39°28′ N; 115°510′ E). The harvested materials were wilted to approximately 40% DM and then cut into 2 cm lengths using a forage chopper (model: 5.2T, Yijie Machinery Factory, Zhengzhou, Henan, China). The chopped forage was then inoculated with L. plantarum B90 (10^6^ CFU/g FM) as the LP group and L. plantarum B90 (10^6^ CFU/g FM) combined with phenylethyl acetate (50 mg/kg FM; LD_50_ = 3670 mg/kg, rat, oral) as the LPP group. The equal amount of sterile water was used for the control group (CK). A total of 1500 kg, packaged in 20 kg per bag, was processed and fermented at room temperature (28°C ± 2°C) for 25 days for animal feeding experiments. A total of 18 samples (three treatments × six replicates each) were collected and analysed for nutritional quality, fermentation characteristics, volatile flavour compounds (VFCs), non‐volatile free amino acids (FAAs), and microbial community composition.
Animals, Feeding Management, and Experimental Design
2.2
All animal‐related experimental procedures were conducted in accordance with protocols approved by the Animal Care and Use Committee of the Institute of Subtropical Agriculture, Chinese Academy of Sciences (Approval No. ISA‐2024‐0036). The experiment was conducted at the Yellow River Estuary Tan Sheep Institute of Industrial Technology (Lijin County, Dongying City, Shandong Province, China). Twelve healthy 8‐month‐old sheep with similar body weights were randomly assigned to two groups (n = 6 per group) and subjected to a 21‐day feeding trial, consisting of a 7‐day pre‐adaptation period, a 6‐day adaptation period, and an 8‐day experimental period. Each sheep was fed in a single‐cage and dual‐trough choice test. The double tank design is to place two barrels on the feed inlet of the cage at the same time. Each barrel was filled with 500 g of samples. Alfalfa silage was fed twice a day, and the first 30 min of each feeding was fed. After 30 min of feeding, the remaining samples were collected, and the feed intake was weighed for each sheep. The total feed intake of the two feedings was used as the daily feed intake for each sheep, and the feed preference was calculated according to the daily feed intake. The feed preference was calculated according to Fu et al. (2025).
Assessment of Fermentation Characteristics and Nutrition Quality
2.3
To assess the fermentation characteristics of samples, approximately 10 g of each sample was homogenised with 90 mL of sterile water and oscillated for 30 min using an orbital shaker (Xia et al. 2024). The resulting filtrate was used to measure pH, ammonia nitrogen and organic acids contents. The pH value was measured immediately using a pH metre (pH 213; HANNA; Italy). High‐performance liquid chromatography (HPLC‐treated, 1200; Agilent, California, USA) was used to detect the content of organic acids, including lactic acid (LA), acetic acid (AA), propionic acid (PA) and butyric acid (BA) (Wang et al. 2020). To assess the nutritional quality, the samples were dried to a constant weight in a forced‐air oven at 65°C for dry matter (DM) content. After drying, all samples were ground into a fine powder for subsequent analysis. The crude protein (CP), neutral detergent fibre (NDF), and acid detergent fibre (ADF) were measured following the protocols established by the Association of Official Analytical Chemists (AOAC). Additionally, water‐soluble carbohydrate (WSC) content was quantified using a previously described method (Xia et al. 2024).
Analysis of VFCs and Odour Activity Value (OAV)
2.4
After thoroughly mixing all samples, approximately 10 g of each sample was weighed, quickly frozen in liquid nitrogen, and stored at −80°C. Before analysis, the samples were ground to a fine powder with liquid nitrogen, and approximately 1 g of the powdered sample was transferred to a 20 mL glass vial. The vial was immediately sealed with a magnetic screw cap equipped with a polytetrafluoroethylene (PTFE)‐silica septum, prepared for solid‐phase microextraction (SPME) analysis as described by Yang et al. (2022). 2‐Methyl‐3‐heptanone was used as the internal standard. GC–MS data were analysed using Xcalibur 4.1 and TraceFinder 4.0 software (Thermo Scientific). VFCs were identified based on their mass spectra and linear retention indices, with reference to the NIST17 (v2.3) and in‐house spectral libraries. Differential VFCs between alfalfa silage samples treated with various additives and the control (CK) group were screened using two criteria: a variable importance in projection (VIP) score > 1 from the orthogonal partial least squares discriminant analysis (OPLS‐DA) model, and a p value < 0.05 from Student's t‐test. Although an OAV > 1 is commonly used as an indicator of significant contribution of a volatile compound to the overall aroma, the OAV alone may not fully reflect its relative contribution to the aroma profile. To address this limitation, we introduced the Aroma Characteristics Impact (ACI) (Yao et al. 2023) to quantify the contribution of individual volatile compounds to the overall aroma profile of alfalfa silage. The calculation of ACI was according to Yao et al. (2023).
Analysis of Taste FAAs and Taste Activity Value (TAV)
2.5
Twenty milligrams of each sample were accurately weighed, followed by the addition of 141 μL of water and 100 μL of a 0.15% sodium deoxycholate (DOC) solution. The mixture was thoroughly mixed, and the free amino acid (FAA) content was quantified using LC–MS/MS according to the method described by Ma et al. (2022). LC–MS/MS analysis was performed using an ExionLC AD system interfaced with a QTRAP 6500+ mass spectrometer (Sciex, USA) at Majorbio Bio‐Pharm Technology Co. Ltd. (Shanghai, China). The raw LC–MS data were processed using Sciex OS software, with ion fragments automatically detected and peak integration performed under default parameters. Manual verification was performed on all integrations. Metabolite concentrations in samples were determined based on a linear regression analysis of standard curves. Non‐volatile FAAs with TAV ≥ 1 were considered to be taste actives, contributing significantly to taste characteristics, and it was calculated according to Shi et al. (2024).
Absolute Quantification of 16S rRNA Gene Sequencing Data
2.6
Microbial DNA was extracted from the samples using the FastPure Soil DNA Isolation Kit (MJYH, Shanghai, China), according to the manufacturer's protocol. Absolute quantification of the 16S rRNA gene was conducted by Majorbio Bio‐Pharm Technology Co. Ltd. (Shanghai, China) using the PacBio platform for sequencing, following the manufacturer's guidelines. High‐throughput sequencing data were analysed using methodologies established in prior studies (Chen et al. 2025). In short, quality control procedures for the raw paired‐end sequencing reads were conducted using Fastp software (version 0.19.6, available at https://github.com/OpenGene/fastp). Subsequently, sequence assembly was performed utilising FLASH software (version 1.2.11). The RDP Classifier (version 2.11) was utilised to align sequences against the 16S rRNA gene database (NT Taxon Core v2024).
Metagenomic Sequencing Analysis
2.7
Microbial DNA was extracted from samples using the FastPure Soil DNA Isolation Kit (Shanghai, China). Following the method described by Ruan et al. (2024), DNA extracts were fragmented using a Covaris M220 ultrasonicator (Gene Company Limited, China), and fragments of approximately 400 bp were selected. Metagenomic sequencing was conducted utilising the NovaSeq X Series 25B Reagent Kit (300 cycles) (Illumina, USA), following the manufacturer's protocols at Majorbio Bio‐Pharm Technology Co. Ltd. (Shanghai, China). High‐quality reads from each sample were then aligned against this non‐redundant gene set using SOAPaligner software (version 2.21) at a 95% identity threshold, and the gene abundance in each corresponding sample was quantified. Gene abundance was normalised using the RPKM (Reads Per Kilobase per Million mapped reads) method, which accounts for gene length and sequencing depth by expressing read counts per kilobase of transcript per million mapped reads (Lawson et al. 2017). Each functional gene was taxonomically linked to its most probable host species (e.g., L. plantarum ) according to BLAST alignment results, and the relative contribution of each species to enzymatic function was subsequently determined by weighting the normalised RPKM values at the gene level. DIAMOND (version 0.8.35) was used to align the amino acid sequences of the non‐redundant gene set against the KEGG database, thereby determining the functional annotations of the genes according to KEGG pathways. Gene‐specific primers for proline aminopeptidase, histidinol dehydrogenase, branched‐chain‐amino‐acid transaminase and aryl alcohol dehydrogenase were designed using NCBI Primer‐BLAST (https://www.ncbi.nlm.nih.gov/tools/primer‐blast/) based on homologous gene sequences retrieved from the KEGG database (Table S1). PCR amplification was performed using genomic DNA from L. plantarum B90 as the template, with 2 × Hieff Canace AdvanceFast PCR Master Mix (With Dye) (Yeasen Biotechnology Co. Ltd., Shanghai, China) as the reaction mixture. PCR products were validated by gel electrophoresis and further confirmed by sequencing. Total RNA was extracted from L. plantarum B90 using the TaKaRa MiniBEST Universal RNA Extraction Kit (9767), according to the manufacturer's instructions. The cDNA was synthesised using cDNA Synthesis Kit (TaKaRa, RR036A). Quantitative real‐time PCR (qRT‐PCR) was carried out in 96‐well plates with Bestar SYBR Green qPCR Master Mix (DBI Bioscience, Cat. No. DBI‐2043), following the manufacturer's protocol. Gene‐specific primers were designed using NCBI Primer‐BLAST (Table S2). The 16S rRNA gene was used as an internal reference for normalisation of target gene expression.
Statistical Analyses and Data Visualisation
2.8
Statistical analyses were performed using IBM SPSS Statistics 27 (version R27.0.1.0, 64‐bit). A one‐way analysis of variance (ANOVA) was applied to assess differences in nutritional quality, fermentation characteristics, and microbial alpha diversity indices—including Chao1 and Simpson—among the various treatment groups. Additionally, a t‐test was applied to compare differences in FAAs, feed intake, dry matter intake (DMI) and feed preference between groups. GraphPad Prism 8.0 (GraphPad Software, San Diego, CA, USA) was used for data visualisation. Pearson correlation analysis was utilised to assess significant correlations between high‐abundance microorganisms via ChiPlot (https://www.chiplot.online/). Bioinformatics analyses, including Orthogonal Partial Least Squares Discriminant Analysis (OPLS‐DA) and Principal Coordinate Analysis (PCoA), were carried out using OmicStudio tools (https://www.omicstudio.cn/tool).
Results
3
L. plantarum
Combined With Phenylethyl Acetate Improves the Alfalfa Silage Quality and Feed Intake
3.1
To gain deeper insights into the determinants of feed intake behaviour, we first analysed the nutritional profiles and fermentation parameters of alfalfa silage across treatment groups. As shown in Figure 1, no significant differences were observed among treatment groups in terms of nutritional composition, including CP, NDF, ADF, and WSC. However, notable differences were observed in fermentation quality among the groups. In particular, the pH values in both the LP and LPP groups were significantly (p < 0.05) lower than those in the CK group. The LP group exhibited the lowest pH of 4.48, followed by the LPP group of 4.55. The levels of total acids (TA), LA, and AA in both LP and LPP groups were significantly higher (p < 0.05) than those in the CK group. Additionally, the ammonia nitrogen content in the LP group was significantly lower (p < 0.05) than in the other two groups.
Nutrition and fermentation quality of fermented alfalfa in response to additives. (A) dry matter (DM) content. (B) crude protein (CP) content. (C) acid detergent fibre (ADF) content. (D) neutral detergent fibre (NDF) content. (E) water‐soluble carbohydrates (WSC) content. (F) pH value. (G) lactic acid (LA) content. (H) acetic acid (AA) content. (I) propionic acid (PA) content. (J) total acid (TA) content. (K) ammonia nitrogen (AN) content. CK: Fermented alfalfa inoculated with sterilised water; LP: Fermented alfalfa inoculated with Lactiplantibacillus plantarum B90; LPP: Fermented alfalfa inoculated with Lactiplantibacillus plantarum B90 + phenylethyl acetate.
In the comparison between the LP and CK groups (Table 1), the feeding preference rate of sheep for the LP group was 69.35%, indicating a significant preference (p < 0.001) compared to the CK group. Consistently, the dry matter intake (DMI) of sheep fed with the LP group was significantly higher (p < 0.05) compared to those fed with the CK group. In the comparison between the LPP and LP groups (Table 1), the feeding preference rate for the LPP group was 67.77%, indicating that sheep preferred the LPP group over the LP group. Similarly, the DMI of sheep fed with the LPP group was significantly higher (p < 0.05) than that of sheep fed with the LP group.
L. plantarum
Combined With Phenylethyl Acetate Regulates the Characteristics of VFCs
3.2
Among the characteristic volatile compounds across treatment groups, both the LP and LPP treatments exhibited a higher total abundance than the CK group (Figure 2A). The composition of these volatiles in alfalfa silage was dominated by aldehydes, alcohols and esters (Figure 2B), with aldehydes accounting for more than 50% of the total abundance (Table S3). The OPLS‐DA model analysis revealed that a total of 31 characteristic VFCs exhibited significant differences between the LP and CK groups (Figure 2C), as determined by variable VIP scores greater than 1 and statistical significance of p < 0.05. Among these, 23 main volatile compounds were further increased for their contribution to the overall aroma using the ACI value. Dimethyl trisulfide (47.32%), 4‐ethylphenol (22.56%), β‐damascenone (11.22%), eugenol (9.94%), phenylacetaldehyde (3.11%), isovaleric acid (1.96%) and methyl salicylate (1.53%) exhibited ACI values greater than 1%, suggesting their significant contribution as distinguishing volatile compounds between the LP and CK groups. A total of 25 characteristic volatile compounds exhibited significant differences between the LPP and LP groups (Figure 2D). Among these, phenylethyl alcohol (94.62%), benzyl alcohol (4.02%), and hexanal (1.36%) (Table S4) showed marked up‐regulation, with all having an ACI > 1%. Notably, phenylethyl alcohol possessed the highest ACI value.
Key VFCs characteristics of fermented alfalfa in response to additives. (A) Contents concentrations of different types of VFCs. (B) Relative concentrations of different types of VFCs. (C) The key flavour compounds were ranked by VIP of the LP group based on the partial least squares PLS‐DA model. (D) The key flavour compounds ranked by VIP of the LPP group based on the PLS‐DA model. CK, fermented alfalfa inoculated with sterilised water; LP, fermented alfalfa inoculated with Lactiplantibacillus plantarum B90; LPP, fermented alfalfa inoculated with Lactiplantibacillus plantarum B90 + phenylethyl acetate.
L. plantarum
Combined With Phenylethyl Acetate Optimises the Composition of Non‐Volatile Taste FAAs
3.3
Based on taste characteristics, amino acids can be divided into umami, sweetness, and bitterness categories. The types and contents of taste amino acids in different treatment groups are shown in Figure S1 and Figure 3A,B. Compared to the CK group, the LP and LPP treatments significantly (p < 0.05) reduced the concentrations of bitter and sweet amino acids (Table S5). In contrast, the LPP group showed a higher level of sweetness‐related amino acids than the LP group, suggesting that phenethyl acetate enhanced sweet‐taste compounds. Among the 16 detected FAAs (Table S5), seven were identified as characteristic taste‐active amino acids (TAV > 1): aspartic acid (Asp), alanine (Ala), proline (Pro), arginine (Arg), histidine (His), isoleucine (Ile), and valine (Val). These compounds are key contributors to the overall flavour profile. This suggests that these amino acids contribute to silage taste. Compared to the CK group, the LP group exhibited significant down‐regulation of six characteristic taste‐active amino acids (Asp, Ala, Pro, His, Ile and Val) and up‐regulation of Arg (p < 0.05). Compared to the LP group, the LPP group exhibited a significant down‐regulation of Arg (p < 0.05), a characteristic bitterness amino acid.
Key FAAs characteristics of fermented alfalfa in response to additives. (A) Radar images of the total amount of different types of key FAAs; (B) The total content of different types of key FAAs. CK: Fermented alfalfa inoculated with sterilised water; LP: Fermented alfalfa inoculated with Lactiplantibacillus plantarum B90; LPP: Fermented alfalfa inoculated with Lactiplantibacillus plantarum B90 + phenylethyl acetate.
L. plantarum
Combined With Phenylethyl Acetate Alters the Microbial Community
3.4
We used absolute quantitative 16S rRNA sequencing (AQ‐16S‐seq) to directly measure changes in the absolute abundance of microorganisms, thereby accurately determining the true increase or decrease of specific taxonomic units of community structure. To identify core microbial taxa in alfalfa silage, we quantified absolute microbial abundances across treatment groups (Figure 4A). The CK group was dominated by L. plantarum (32.73%), Companilactobacillus pabuli (9.92%), and Limosilactobacillus panis (9.82%), collectively accounting for 52.47% of total microbial abundance. In contrast, the LP group exhibited higher dominance of L. plantarum (51.92%) and Enterococcus mundtii (4.98%), totaling 56.90%, while the LP group was dominated by L. plantarum (41.57%) and Levilactobacillus brevis (7.17%), summing to 48.74%. Notably, the total microbial abundance in the LP and LPP groups was lower than in the CK group. To further explore the interaction between microbial flora in different treatment groups, we conducted a correlation analysis of the top 10 microorganisms (Figure 4B–D). The results showed that in the CK group, L. plantarum was positively correlated with C. pabuli and negatively correlated with L. panis . In the LP group, L. plantarum and E. mundtii were positively correlated. In the LPP group, L. plantarum and L. brevis were negatively correlated.
Microbial composition, contents, and correlation of fermented alfalfa in response to additives. (A) The relative abundance and absolute abundance of microbiota in different additives at the species level. (B) The correlation network analysis of key microorganisms in CK groups. (C) The correlation network analysis of key microorganisms in LP groups. (D) The correlation network analysis of key microorganisms in LPP groups. CK, fermented alfalfa inoculated with sterilised water; LP, fermented alfalfa inoculated with Lactiplantibacillus plantarum B90; LPP, fermented alfalfa inoculated with Lactiplantibacillus plantarum B90 + phenylethyl acetate.
Microbial Community Contributes to the Formation of Key Flavour Compounds
3.5
To further elucidate the regulatory role of fermentative microorganisms on key flavour compounds, metabolic pathways were reconstructed using the KEGG database. The biosynthesis of seven key flavour amino acids and 2 volatile compounds influencing sheep feed intake involved five metabolic pathways and 11 enzymes, respectively (Figure 5). Metagenomic data were integrated to predict the contributions of core microbial taxa to the key enzymes involved in the synthesis of nine important flavour compounds.
Metabolic pathways involved in the formation of key flavour compounds and the contributions of core microorganisms to these pathways. The heat map represents the content of key flavour compounds. CK, fermented alfalfa inoculated with sterilised water; LP, fermented alfalfa inoculated with Lactiplantibacillus plantarum B90; LPP, fermented alfalfa inoculated with Lactiplantibacillus plantarum B90 + phenylethyl acetate.
The key taste amino acids proline and arginine are produced through the arginine and proline metabolic pathways, in which the precursor substance peptide of proline is produced by proline aminopeptidase (EC 3.4.11.5). In both the LP and LPP groups, L. plantarum was the primary contributor to EC 3.4.11.5, with contribution rates of 93.30% and 93.13%, respectively. Proline was produced by 1‐pyrroline‐5‐carboxylate under the action of proline oxidase (EC 1.5.1.2). L. plantarum was the dominant contributor to EC 1.5.1.2 in both the LP and LPP groups, accounting for 73.41% and 80.41% of its contribution rate, respectively. The key taste amino acids alanine and aspartic acid are produced through the metabolic pathways of alanine, aspartic acid, and glutamic acid, in which alanine is produced by pyruvate under the action of alanine dehydrogenase (EC 1.4.1.1). Alanine can also be transformed with pyruvate under the action of alanine transaminase (EC 2.6.1.2). Sphingomonas sp. (LP: 29.37%; LPP: 28.69%) and E. mundtii (LP: 18.90%; LPP: 18.80%) were the main contributors of EC 1.4.1.1. Aspartic acid was converted to alanine under the action of aspartate 4‐decarboxylase (EC 4.1.1.12). In addition, aspartic acid was the most abundant of the seven key taste amino acids, followed by alanine. The key taste amino acid, histidine, is produced through the histidine metabolic pathway, and histidinal is produced under the action of histidinol dehydrogenase (EC 1.1.1.23). The main contributors of EC 1.1.1.23 in the LP group and the LPP group were L. plantarum (LP: 61.46%; LPP: 58.20%) and E. hormaechei (LP: 13.22%; LPP: 12.90%). Valine and isoleucine are produced by the valine, leucine, and isoleucine biosynthesis pathways. 2‐Oxoisovalerate was converted to valine under the action of valine‐pyruvate transaminase (EC 2.6.1.66). The main contributors of EC 2.6.1.66 in the LP group and LPP group were E. hormaechei (LP: 30.28%; LPP: 26.72%) and E. bacterium (LP: 19.95%; LPP: 21.65%). (S)‐3‐Methyl‐2‐oxopentanoate was converted to isoleucine by branched‐chain‐amino‐acid transaminase (EC 2.6.1.42), which also catalysed the conversion of 2‐oxoisovalerate to valine. The contribution of EC 2.6.1.42 in the LP and LPP groups was L. plantarum (LP: 65.60%; LPP: 64.15%).
The key VFCs phenylacetaldehyde and phenylethyl alcohol are produced through the phenylalanine metabolic pathway. Phenylethylamine was converted to phenylacetaldehyde under the action of monoamine oxidase (EC 1.4.3.4). The contributors of EC 1.4.3.4 in the LP group were Pseudomonas monteilii (29.48%) and Acinetobacter bereziniae (23.67%). Phenylacetaldehyde was converted to phenylethyl alcohol under the action of aryl alcohol dehydrogenase (EC 1.1.1.90). The main contributors of EC 1.1.1.90 in the LP and LPP groups were L. plantarum , with a contribution of 83.20% in the LP group and 85.75% in the LPP group. Notably, PCR amplification (Figure S2A), Sanger sequencing, and qRT‐PCR analysis (Figure S2B) collectively confirmed the presence and transcription of genes encoding aryl alcohol dehydrogenase, proline aminopeptidase, histidine dehydrogenase, and branched‐chain amino acid transaminase in L. plantarum B90—consistent with metagenomic predictions. These enzymes are implicated in the biosynthesis of key flavour‐related metabolites, including phenylethyl alcohol, proline, histidine and isoleucine, during fermentation.
Discussion
4
The quality and palatability of silage significantly influence feed intake of ruminants (Scherer et al. 2021). In this study, the application of LAB, either alone or in combination with flavour compound, enhanced silage fermentation quality, consistent with previously reported results (Su et al. 2024). This finding further demonstrates that fermentation characteristics, such as pH values, LA, AA and PA content, are among the factors influencing feeding preferences. The DMI and feeding preference of sheep fed with the LP group were higher compared to those fed with the CK group. These results are consistent with Scherer et al. (2021) findings that the inclusion of LAB during the ensiling of alfalfa and red clover markedly enhanced voluntary feed intake compared to non‐inoculated silage. The DMI and feeding preference of sheep fed with the LPP group were higher than those fed with the LP group. These findings align with the results reported by Gerlach et al. (2019), who observed that adding ethyl lactate to whole plant corn silage increased the feed intake of treated silage compared to the untreated control. This result indicates that the increase in sheep feed intake may be associated with the addition of volatile ester compounds to silage. However, studies on spraying ethyl acetate and ethyl lactate onto hay have shown that ethyl acetate or ethyl lactate, when used as a single volatile compound, is unlikely to affect the preference behaviour or feed intake of ruminants (Gerlach et al. 2019). This suggests that the higher feed intake observed in the LP group compared to the CK group is due to the synergistic effect of multiple flavour compounds. Furthermore, the higher feed intake of sheep fed with the LPP group compared to those fed with the LP group indicates that, in addition to the influence of phenethyl acetate, other key flavour substances also contribute to increased feed intake. While this study focused on short‐term intake responses, future work should investigate the long‐term effects of flavour‐modulated silage on animal health, metabolism, and the sensory quality of derived products such as lamb meat and sheep milk. Such insights could pave the way for developing functional forages that enhance both palatability and end product value.
Flavour, particularly volatile odour compounds, plays a crucial role in the palatability of feed and significantly influences the feeding behaviour of ruminants (Rapisarda et al. 2012; Scherer et al. 2021). The analysis of volatile compounds can explain ruminants' feeding preferences from the perspective of sensory perception (Rapisarda et al. 2012). Components with an OAV > 1 are considered characteristic volatile compounds (Yao et al. 2023). The increased total abundance of characteristic volatile compounds—predominantly aldehydes, alcohols, and esters, with aldehydes accounting for over 50%—in both LP and LPP silages compared to CK suggests that LAB inoculation, especially when combined with phenethyl acetate, enhances the aromatic profile of alfalfa silage, which may underlie the observed improvements in feed palatability and intake (Rapisarda et al. 2012). Aldehydes, largely generated through the degradation of available amino acids, impart a pleasant flavour to feed, characterised by green, orange, nut and herbal notes (Rapisarda et al. 2012). Rapisarda et al. (2012) have confirmed that ewes exhibit a higher intake of beet pulp, which was rich in aldehydes.
Compounds with an ACI > 1% were identified as key contributors to the distinctive aroma profile of the samples (Yao et al. 2023). Among the 31 differentially abundant volatile compounds distinguishing LP from CK silage, seven—dimethyl trisulfide, 4‐ethylphenol, β‐damascenone, eugenol, phenylacetaldehyde, isovaleric acid and methyl salicylate—exhibited ACI values exceeding 1%. These key volatile compounds contributed significantly to the enhanced feed intake observed in the LP group compared to the CK group, likely due to their distinct aroma profiles influencing olfactory perception and palatability. These compounds significantly contribute to the distinctive aromatic profile of traditional fermented foods. For example, beta‐damascenone was first isolated from rose oil, grapes, and wine, and has been identified as a major odorant in various fruits, teas, and beverages, with floral and fruity odours (An et al. 2023). Phenylacetaldehyde, renowned for its distinctive rose‐like aroma, is widely utilised as a flavouring agent in spice formulations (Sun et al. 2025). Additionally, it has also been identified as the most abundant aromatic compound in corn silage (Su et al. 2023). Eugenol finds application in animal husbandry, where its inclusion as part of a blend of essential oils in the diet has been shown to influence the feeding behaviour of high‐producing dairy cows (Diepersloot et al. 2025). The synergistic effect of these seven key volatile compounds contributed to the distinctive aroma profile of alfalfa silage in the LP group, which may be one of the reasons for the increased feed intake compared with the CK group.
Notably, phenethyl alcohol (94.62%), benzyl alcohol (4.02%) and hexanal (1.36%) were markedly increased in LPP silage compared to LP, with all three compounds exhibiting ACI values exceeding 1% and phenethyl alcohol having the highest ACI value. These three volatiles serve as the key volatile compounds distinguishing the LPP group from the LP group, and they are also pivotal in explaining the increased feed intake observed in the LPP group relative to the LP group. Phenethyl alcohol, which has a rose‐like aroma, is a precursor of phenethyl acetate (Zhang et al. 2020). This suggests that L. plantarum with phenethyl acetate co‐fermentation may promote the biosynthesis of phenethyl alcohol, thereby enhancing the overall floral/rose‐like fragrance profile. Similarly, a study has reported that phenylethyl alcohol is a key flavour component contributing to the floral and sweet notes characteristic of Huangjiu (Wei et al. 2023). Lin et al. (2024) identified benzyl alcohol as a significant aromatic component contributing notably to the aromatic profile of broken black tea. Therefore, the specific combination and concentrations of these key volatile compounds collectively shaped the distinctive flavour profile of the LPP group, which led to significantly higher feed intake compared with the LP group.
During the ensiling process, forages undergo varying degrees of protein and amino acid degradation. This degradation, facilitated by plant‐derived and bacterial proteases, results in a gradual accumulation of amino acids (Scherer, Gerlach, Taubert, et al. 2019). FAAs are generated from the degradation of soluble proteins by microbial proteases, contributing to flavour characteristics and serving as precursors for other flavour compounds (Sun et al. 2025). Amino acids with a TAV ≥ 1 are classified as characteristic amino acids, which are the key contributors to flavour perception and significantly influence the overall flavour profile (Shi et al. 2024). While both LP and LPP treatments reduced bitter‐ and sweet‐taste amino acids relative to CK, the LPP group retained higher levels of sweetness‐associated amino acids than the LP group, likely due to the modulatory effect of phenethyl acetate on amino acid metabolism. Although umami intensity also decreased, the overall reduction in bitter and sweet amino acids may contribute to improved flavour balance, potentially optimising the overall flavour profile and enhancing palatability through modulation of taste perception. Notably, while most taste‐active free amino acids—including Asp, Ala, Pro, His, Ile and Val—were decreased in the LP group relative to the CK group, arginine (Arg) was significantly increased, suggesting a shift in the umami–bitter balance that may influence palatability. Apart from that, the LPP group showed lower levels of arginine—a characteristic bitter‐tasting amino acid—compared to the LP group, potentially mitigating aversive bitter perception. The changes in these amino acids profiles significantly reflected the higher feed intake when LPP was fed to sheep compared to that of CK. These findings are consistent with those of Scherer et al. (2021), who identified bitterness amino acids Ile and methionine (Met) in avoided silage, suggesting that bitter taste perception may influence feed selection in livestock. Collectively, treatment with L. plantarum , whether applied singly or in conjunction with phenethyl acetate, resulted in modifications to the amino acid composition, particularly through a decrease in bitter‐tasting amino acids. These changes may contribute to an improved flavour profile and greater feed acceptance. Interestingly, compared to mono‐fermentation, the synergistic treatment exerted a less pronounced effect on the concentration of taste‐active amino acids, indicating its primary role in altering volatile flavour compounds rather than non‐volatile taste amino acids.
Silage fermentation is microbial‐driven, and microbial metabolic activities shape community dynamics, directly influencing metabolite profiles (Muck et al. 2018). Absolute quantification revealed that while the CK silage harboured a more diverse and abundant microbial community, both LP and LPP treatments led to a pronounced dominance of L. plantarum alongside an overall reduction in total microbial abundance. These results suggest that the additive reduced microbial community complexity and total abundance. This reduction may be attributed to the additive‐induced decrease in pH and concurrent increases in lactic and acetic acid concentrations, which likely created an unfavourable environment for undesirable or spoilage‐associated microorganisms. Consequently, the fermentation process exhibited enhanced efficiency and stability, ultimately contributing to improved fermentation quality. This observation aligns with findings in mixed Sesbania cannabina and sweet sorghum silage (Xia et al. 2024). The decreased metabolic activity of LAB strains was likely due to two key factors: alterations in the silage's chemical composition and competitive substrate utilisation among co‐occurring LAB species. According to Öztürk et al. (2023), E. mundtii contributes significantly to aroma development, maturation processes, and sensory characteristics in fermented foods, including meats and dairy products. L. brevis can increase the content of 1,8‐terpineol, 1‐hexanol, caproic acid, and other substances in radish paocai, thereby improving the flavour and sensory quality of its fermentation. The results suggest that L. plantarum , L. brevis and E. mundtii may play significant roles in the biosynthesis and modulation of volatile flavour compounds. Furthermore, phenethyl acetate, a common aromatic compound used to enhance flavour profiles, does not exert inhibitory effects on the growth of these microorganisms during fermentation. The flavour characteristics of fermented products are significantly affected by the interaction within the microbial community (Jiang et al. 2025). Notably, the interaction patterns of L. plantarum with other dominant taxa changed significantly across treatments—from positive correlations in CK and LP to a negative correlation in LPP. The results suggest that variations in microbial interspecific relationships across treatments may promote the formation and accumulation of specific flavour compounds. Previous studies have reported that L. plantarum (Munoz et al. 2024), E. mundtii (Peng, Cheng, Huang, et al. 2025), L. brevis (Zhang et al. 2023), and W. cibaria (Xiong et al. 2024) are common flavour contributors in fermented products and are essential for the formation of flavour compounds. However, the precise mechanisms by which these microorganisms drive flavour development remain unclear.
Integrating metagenomic and metabolic pathway analyses suggests that core fermentative microorganisms likely shape the flavour profile of silage by modulating key enzymes involved in central carbon metabolism that channel precursors toward the biosynthesis of taste‐ and aroma‐active compounds. Initially, the substrate, primarily consisting of soluble carbohydrates, is predicted to be enzymatically converted via glycolysis or gluconeogenesis into pyruvate, a critical precursor for flavour compound biosynthesis (Peng, Cheng, Huang, et al. 2025). Xiong et al. (2024) reported that isoleucine is produced by EC 2.6.1.42, primarily by species of Lactobacillaceae. Our results are consistent with this finding, as we identified high abundances of genes encoding EC 2.6.1.42 in both the LP and LPP groups, with L. plantarum being the dominant contributor (LP: 65.60%; LPP: 64.15%). In high‐quality alfalfa silage associated with increased feed intake, L. plantarum, which dominates fermentation, is predicted to play a critical role in enhancing the biosynthesis of proline, histidine and isoleucine. Specifically, the strain may promote the biosynthesis of proline, histidine and isoleucine via EC 1.5.1.2, EC 1.1.1.23 and EC 2.6.1.42, respectively. The conversion of phenylethylamine to phenylacetaldehyde in LP silage appears to be mediated by EC 1.4.3.4, primarily from Pseudomonas monteilii and Acinetobacter bereziniae . Zhang et al. (2022) reported that Rasamsonia emersonii and Byssochlamys spectabilis present in Daqu are capable of oxidising phenylethylamine to phenylacetaldehyde via EC 1.4.3.4. The reduction of phenylacetaldehyde to phenethyl alcohol is predicted to be catalysed by EC 1.1.1.90, an enzyme responsible for the biosynthesis of aromatic alcohols (Peng, Cheng, Huang, et al. 2025). In both LP and LPP silages, L. plantarum was the predominant microbial source of this enzyme, consistent with Peng, Cheng, Huang, et al. (2025), who reported that EC 1.1.1.90 is primarily encoded by L. plantarum and Pediococcus pentosaceus . In addition, phenylethyl alcohol, as a precursor of phenethyl acetate, may be produced from phenylacetaldehyde under the action of EC 1.1.1.90 with a high contribution of L. plantarum . It could also be due to the hydrolysis of phenethyl acetate under acidic conditions (Li et al. 2022). These findings suggest that the synergistic fermentation of phenethyl acetate and L. plantarum may further promote the production of phenylethyl alcohol, potentially contributing to a higher rose aroma in the overall flavour profile and promoting the feed intake of fermented alfalfa. However, it should be noted that these metabolic pathway interpretations are based primarily on metagenomic predictions and gene abundance data. While the strong correlations between microbial abundance and flavour compound accumulation support these hypotheses, we acknowledge that direct experimental evidence, such as enzyme activity assays or transcriptomics, was not provided in this study to definitively confirm the operation of these specific metabolic pathways. Future studies incorporating multi‐omics approaches and biochemical validations will be necessary to elucidate the exact enzymatic mechanisms underlying these flavour formation.
Conclusions
5
Fermentation by L. plantarum alone or combined with phenethyl acetate enhances the content of volatile flavour compounds in alfalfa silage and modulates the profile of taste amino acids. Key flavour compounds, including dimethyl trisulfide, 4‐ethylphenol and β‐damascenone, were significantly upregulated with L. plantarum alone, while crucial taste amino acids—such as aspartic acid, alanine, proline, histidine, isoleucine and valine—decreased, except for arginine, which increased. These metabolic shifts were associated with improved feed intake. The addition of phenethyl acetate further elevates phenethyl alcohol, benzyl alcohol, and hexanal, concurrently reduces arginine, and thereby enhances palatability. This synergistic effect of phenethyl acetate may represent a promising strategy to improve the palatability of low‐quality or less palatable forages—such as sesbania—particularly when their inherent microbial activity is limited and baseline flavour profiles are suboptimal. Mechanistically, the presence and transcription of genes in L. plantarum encoding aryl alcohol dehydrogenase, proline aminopeptidase, histidine dehydrogenase, and branched‐chain amino acid transaminase provide indirect genomic evidence for their potential involvement in the formation of key flavour metabolites during fermentation. Collectively, these results highlight the pivotal role of microorganisms in determining silage sensory attributes and suggest a strategy for enhancing feed utilisation through targeted modulation of LAB and exogenous flavour compounds.
Author Contributions
Zhihui Fu: writing – original draft, data curation and formal analysis. Tianwei Wang: writing – review and editing and funding acquisition. Jiaqi Zhang, Wenzhao Wang, Xiumin Zhang and Kaixuan Wei: investigation. Muhammad Tahir revised the manuscript. Jin Zhong: supervision and funding acquisition.
Funding
This work was supported by National Natural Science Foundation of China, 32201467. Strategic Priority Research Program of the Chinese Academy of Sciences, XDA26040201. Science and Technology Program of the Inner Mongolia Autonomous Region, 2025KJHZ0041.
Conflicts of Interest
The authors declare no conflicts of interest.
Supporting information
Data S1: mbt270332‐sup‐0001‐Tables.docx.
Data S2: mbt270332‐sup‐0002‐Figures.docx.
The reference list from the paper itself. Each links out to its DOI / PubMed record.
- 1An, T. T. , S. S. Shen , Z. Q. Zu , et al. 2023. “Changes in the Volatile Compounds and Characteristic Aroma During Liquid‐State Fermentation of Instant Dark Tea by Eurotium cristatum .” Food Chemistry 410: 135462.36669288 10.1016/j.foodchem.2023.135462 · doi ↗ · pubmed ↗
- 2Chen, B. , L. Guan , C. Wu , et al. 2025. “Gut Microbiota‐Butyrate‐PPARγ Axis Modulates Adipose Regulatory T Cell Population.” Advanced Science 12, no. 20: 2411086.39998325 10.1002/advs.202411086 PMC 12120792 · doi ↗ · pubmed ↗
- 3Deng, X. , Y. Jia , G. Ge , et al. 2023. “Microbiomics and Volatile Metabolomics‐Based Investigation of Changes in Quality and Flavor of Oat (Avena sativa L.) Silage at Different Stages.” Frontiers in Plant Science 14: 1278715.38023849 10.3389/fpls.2023.1278715 PMC 10657850 · doi ↗ · pubmed ↗
- 4Diepersloot, E. C. , M. R. Pupo , C. Heinzen Jr. , M. S. Souza , and L. F. Ferraretto . 2025. “Effects of Monensin and Essential Oil Blend Supplementation on Lactation Performance and Feeding Behavior in Dairy Cows.” Journal of Dairy Science 108, no. 3: 2517–2526.39710270 10.3168/jds.2024-25834 · doi ↗ · pubmed ↗
- 5Fabricio, M. F. , L. Schmidt , P. D. H. Rother , et al. 2025. “Targeted Metabolomics of Phenolic and Volatile Compounds During the Fermentation of a Potential Probiotic Tofu Whey Beverage.” Food Chemistry 478: 143689.40049137 10.1016/j.foodchem.2025.143689 · doi ↗ · pubmed ↗
- 6Fu, Z. H. , T. W. Wang , J. Q. Zhang , et al. 2025. “Multi‐Omics Profiling Reveals Microbial Regulation of a Key Aromatic Ester Phenethyl Acetate Formation in Fermented Alfalfa and Its Impact on Sheep Feed Preference.” Food Chemistry: X 32: 103249.41282313 10.1016/j.fochx.2025.103249 PMC 12639630 · doi ↗ · pubmed ↗
- 7Gerlach, K. , K. Weiss , and K. H. Südekum . 2019. “Effects of Ethyl Ester Supplementation to Forage on Short‐Term Dry Matter Intake and Preference by Goats.” Archives of Animal Nutrition 73, no. 2: 127–139.30784298 10.1080/1745039 X.2019.1575656 · doi ↗ · pubmed ↗
- 8Guo, X. S. , W. C. Ke , W. R. Ding , et al. 2018. “Profiling of Metabolome and Bacterial Community Dynamics in Ensiled Medicago sativa Inoculated Without or With Lactobacillus plantarum or Lactobacillus buchneri .” Scientific Reports 8: 357.29321642 10.1038/s 41598-017-18348-0PMC 5762819 · doi ↗ · pubmed ↗