Second-cut alfalfa (Medicago sativa L.) at early bud stage was harvested from an experimental field (117°49′E, 39°39’N) in Huanghua City, Hebei Province, China. The Crop was grown without herbicide or fertilizer application. Fresh alfalfa was chopped to 2–3 cm in length. Jujube powder was obtained from Cangzhou Defeng Jujube Industry Co., Ltd. (Qingtown of Cangzhou City, Hebei Province). Two treatments were prepared: (i) control (CK) composed of 300 g chopped alfalfa; and (ii) Jujube powder treatment composed of 300 g chopped alfalfa supplied with 4% (w/w) Jujube powder according to a previous study (Liu et al., 2016). Chemical characters and microbial compositions of the raw materials are shown in Table 1. The mixed materials were packed manually into 28 cm × 35 cm polyethylene bags and vacuumed (DZ-360 vacuum sealer. Jinqrui Co. Wuxue City, Hubei, China). A total of 42 bags (2 groups × 3 replicates × 7 sampling times) were prepared and stored at ambient (25–37 °C). Sampling was conducted at 0, 1, 5, 15, 30, 45 and 60 days of ensiling in triplicate. For each bag, 10 g of silage was collected, homogenized, and stored at −80 °C for microbial analysis. The remaining material was used for fermentation parameter analysis.

Chemical composition of the samples was analyzed according to standard methods. Dry matter content was determined by the weigh difference before and after drying the sample at 65 °C for 48 h. The dried sample was ground through 1.0 mm sieve for the subsequent nutrient analysis. Crude protein was analyzed according to standard procedure detailed by the Association of Official Analytical Chemists (Hasan, 2015). Neutral detergent fiber (NDF) and acid detergent fiber (ADF) were measured according to the method of Van Soest et al. (1991) by using fiber analyzer (Ankom 2000i full; Ankom Tech Co., Macedon, NY, United States).

To determine the fermentation parameters, fresh silage sample (25 g × 3) was mixed with 225 mL sterile water and incubated at 4 °C overnight, then homogenized for 1 min and filtrated with quantitative filter paper. The acquired filtrate was centrifuged at 4 °C, 4500 × g, for 15 min and the obtained supernatant was conducted for measuring pH with a pH meter (Mettler Toledo). Lactic, acetic, propionic, and butyric acids were quantitively estimated by HPLC (Shimadzu, Tokyo, Japan) equipped with a UV detector and set as follows: Shodex Rspak KC-811S-DVB gel column, eluent 3 mmol/L perchloric acid at a running rate of 1.0 mL/min, temperature of column oven 50 °C; wavelength of 210 nm, injection volume of 5 μL. NH₃-N content was measured by phenol-sodium hypochlorite method (Broderick and Kang, 1980). Water-soluble carbohydrate content was determined by the anthrone-sulfuric acid colorimetric method (McDonald and Henderson, 1964).

For counting the culturable lactic acid bacteria, yeasts and molds, each of the samples (20 g) was immediately blended with 180 mL sterilized saline solution (NaCl 8.5 g/L), and serially diluted. Aliquots (0.1 mL) of the dilutions were spread on plates of De Man–Rogosa–Sharpe agar (MRS) (1.10660, Millipore, according to ISO 15214, for Lactobacilli) and Rose Bengal agar (R1273, Millipore, for yeasts and fungi). The colony forming units were counted after incubation of 2–5 days at 28 °C (Wang et al., 2017).

The samples frozen at −80 °C were used for metagenomic DNA extraction according to Liu et al. (2019). Three repeat samples (10 g × 3) for each day and total 42 samples were respective mixed with 90 mL of sterile normal solution with vigorous shaking at 120 r/m for 2 h. Then the mixture was filtered and the filtrate was centrifuged at 10,000 rpm for 10 min at 4 °C. The deposit was suspended in 1 mL of sterile saline solution and the microbial pellets for DNA extracting were obtained by centrifugation at 12,000 rpm for 10 min at 4 °C. Then DNA was extracted using the MN NucleoSpin 96 Soi (Macherey Nagel, Düren, GA, United States) according to the manufacture’s protocols. All the metagenomic DNA extracts were sent to Beijing Baimaike Biotechnology Co., Ltd. (Beijing, China) for microbial community estimation through the paired-end high-throughput sequencing with Illumina HiSeq 2500 platform. For bacteria, the 16S rDNA V3–V4 variable region was amplified by PCR using the universal primers with barcode: 338F (5′-ACTCCTACGGGAGGCAGCA-3′) and 806R (5′-GGACTACHVGGTATCTAAT-3′) (Ding et al., 2020). For fungi (including yeasts), the ITS region was amplified with the primers ITS1F (5′-CTTGGTCATTTAGAGGAAGTAA-3′) and ITS2 (5′-GCTGCGTTCTTCATCGATGC-3′) (De Beeck et al., 2014). Three technical repetitions were sequenced for each sample. The reads in the range of 480–490 bp was retained after quality control filtering and the effective sequences were clustered into operational taxonomic units (OTUs) at the threshold of 97% similarity using Uparse pipeline (version 7.0, Edgar, 2013) based on the database of Silva 138 and Unite 7.0 (Quast et al., 2013; Pruesse et al., 2007).

ACE and Shannon index of alpha-diversity were evaluated with Mothur software (version 7.0, Schloss et al., 2009). The principal coordinate analysis (PCoA) was performed by R software (5.2) to assess community dissimilarity among samples, using Bray–Curtis distance. The co-occurrence patterns among bacterial and fungal community and redundancy analysis were analyzed in cloud platform of https://www.bioincloud.tech/. LefSe was conducted by Metastats to explore the dynamic change of microbial community during ensiling. Microbial function prediction was estimated at the Kyoto Encyclopedia of Genes and Genomes (KEGG) database using Phylogenetic Investigation of Communities by Reconstruction of Unobserved States (PICRUSt2) (Douglas et al., 2020) and FUNGuild (Nguyen et al., 2016).

Chemical composition and microbial community data were compared by one-way ANOVA with student test (p < 0.05) by SPSS 18.0 (IBM, United States). The α-diversity indices and β-diversity were estimated by permutation test, while PERMANOVA was used to analyze the PCoA and RDA results.

The chemical and microbial compositions of jujube powder and fresh alfalfa are shown in Table 1. Fresh alfalfa showed typical characteristics with pH 6.42, 25.31% dry matter, and 5.96% water-soluble carbohydrates, while containing no detectable organic acids or NH₃-N. In contrast, jujube powder exhibited significantly different properties with lower pH (5.20), higher dry matter (93.28%), and substantially greater water-soluble carbohydrate content (30.23%), but lower crude protein (6.55%) in comparison with alfalfa. Meanwhile, their neutral detergent fiber (NDF) and acid detergent fiber (ADF) proportions were comparable. Furthermore, JP contained a substantially more lactic acid bacteria (LAB) and less aerobic bacteria/molds than alfalfa did (Table 1).

The fermentation dynamics and chemical compositions of alfalfa silages with or without JP addition are presented in Table 2. In both the treatment groups, pH was dropped rapidly on the first day of fermentation. However, JP addition significantly (p < 0.005) reduced initial pH and maintained lower pH levels throughout the ensiling compared to the control (CK), though a slight pH increase occurred on day 5 in both groups, possibly due to reduced organic acid production and NH₃-N release from protein degradation on this time. During the whole fermentation procedure, pH value was decreased continuously alongside increasing organic acid accumulation, particularly lactic acid (LA). While pH in CK stabilized at 4.8 by day 30, it further decreased in JP group, reaching 4.26 by day 60 (p < 0.005).

The organic acid profiling revealed LA, acetic acid (AA), and propionic acid (PA) as dominant fermentation products, with no detectable butyric acid (BA) by the end of fermentation in both the silages with/without JP addition. In the final silage (60 days of fermentation), JP-treated silage vs. CK presented greater DM content (257.8 vs. 207.8%, p = 0.001) and LA content (35.6 vs. 32.7, p < 0.01); but lower pH (4.3 vs. 4.8, p < 0.005), acetic acid content (6.6 vs. 8, p < 0.005) and propionic acid content (2.1 vs. 5, p < 0.001).

Correspondingly, the LA:AA ratio in CK ranged between 4.09 (day 60) and 8.65 (day 15), whereas JP-treated silage exhibited higher ratios (6.52–13.09), indicating enhanced homolactic fermentation. As shown in Table 2, AA levels increased similarly in both groups during the first 15 days but were significantly lower (p < 0.01) in JP from day 30 onward, this decrease coinciding with a slight LA decline and continued accumulation of AA. PA followed a trend similar to AA, with JP significantly reducing PA concentrations in the later stages.

NH₃-N levels, which reflecting protein degradation, remained below the recommended threshold (<150 g/kg total N) in both treatments. However, JP addition significantly (p < 0.01) reduced NH₃-N compared to CK. Dry matter content was consistently higher in JP-treated silage (around 260 g/kg FW) than in CK (around 210 g/kg FW) throughout the ensiling process, matching pre-ensiling measurements. In contrast, ash content remained unaffected by JP treatment.

JP treatment significantly enhanced the population of culturable lactic acid bacteria (LAB) only at the initial stage of ensiling (Table 2), consistent with the 12.5-fold higher LAB content in the Jujube Powder compared to the raw alfalfa (Table 1). Both JP-treated and control (CK) silages showed similar dynamic patterns of LAB: a rapid increase on day 1, peaking at day 15, followed by gradual decline until fermentation completion, with no significant differences between treatments (Table 2). The LAB abundance dynamics displayed an inverse relationship with some key fermentation parameters: when the LAB abundance increased rapidly, the concentrations of lactic acid, acetic acid and propionic acid changed minimally or remained stable during the initial 15-days; and when the LAB population decline, the concentrations of lactic acid, acetic acid and propionic acid content increased during the subsequent 15–60 days (Table 2).

The microbial community analysis with high throughput sequencing revealed distinct patterns in alfalfa silage with and without JP addition. From the 42 DNA samples (in 3 replications), a total of 4,039,918 raw pair reads of fungal ITS were obtained and checked using FLASH (version 1), and 2,725,262 clean tags were obtained by two sections splicing. Fungi presented in relatively low abundances throughout the ensiling process in both the treatments (Figure 1; Supplementary Figure S2 and Supplementary Tables S1, S2). At genus level, 27 genera with more than 1% relatively abundances (Supplementary Table S2) were detected, with Humicola and Alternaria being the most dominant; in JP-treated silage, Alternaria declined by ensiling from 60.4% in raw material, and reached the lowest peak (8.6%) on day 45 (Figure 1 and Supplementary Table S2). Importantly, JP addition significantly reduced populations of several molds and yeasts including Humicola, Thermoascus, Thermomyces, Cladosporiuma, Aspergillus, Penicillium, Fusarium and Olpidium compared to the control.

High-throughput sequencing of 16S rRNA genes generated 6,775,441 quality-filtered 16S rRNA gene sequences, with mean coverage exceeding 4×, ensuring reliable results. From the 42 DNA samples (in 3 replications), a total of 7,522,867 raw pair reads were obtained and checked using FLASH (version 1), and 6,775,441 clean tags were obtained by two sections splicing. Shannon diversity indices increased progressively during ensiling (0.46 to 0.78 in CK; 0.63 to 0.91 in JP), except for day 1 in JP treatment (Supplementary Figure S1A and Supplementary Table S3). JP-treated ensiling showed significantly higher (p < 0.001) Shannon indices than CK at most the time points, likely due to initial community differences caused by JP addition. ACE, and Chao1 indices indicated greater initial richness in JP treatments and they were decreased during ensiling (Supplementary Figures S1B–D), while Simpson indices mirrored Shannon diversity patterns (Supplementary Figure S1C).

Analysis of the top 25 operational bacterial taxonomic units (OTUs) revealed 17 families containing 25 genera and 5 defined species (Figure 2 and Supplementary Tables S1, S4). JP addition initially increased abundances of Lactobacillus, Enterococcus and Lactococcus, while decreased those of Burkholderia, Sphingomonas and Pediococcus pentusaceus compared to CK (Figures 2, 3 and Supplementary Table S4).

Both treatments showed rapid disappearance of Burkholderia, Acinetobacter and Sphingomonas by day 1, coinciding with pH drop and LAB increase (Figures 2, 3; Supplementary Figure S2 and Table 2), suggesting oxygen consumption by the aerobic bacteria facilitated anaerobic LAB growth. Subsequent fermentation demonstrated similar successional patterns for dominant OTUs (Enterococcus, Lactococcus, Lactobacillus, Leuconostoc); timing differed Lactococcus peaked at day 1 then declined, while Enterococcus, Lactobacillus and Anaerosalibacter peaked on days 5, 30–45 and 45–60, respectively.

Notable differences included higher Clostridium sensu stricto 18 in CK (days 15–60) versus greater Clostridium sensu stricto 12 in JP. JP treatment also showed elevated minor LAB (Weissella, Lactobacillus paracasei, Pediococcus) and better-preserved LAB communities compared to CK (Supplementary Figures S2–S4). The stable Clostridia community (Garciella, Anaerotruncus, Anaerosalibacter, Lachnospiraceae) in JP versus declining CK populations suggested earlier silage maturation in JP-treated samples.

Network analysis of the top 25 OTUs at genus level in bacteria and 27 OTUS in fungi at genus level revealed distinct co-occurrence patterns between JP-treated and CK silages (Figures 3, 4 and Supplementary Tables S5–S8). In the bacterial neworks, the JP group displayed a higher number of correlation edges (143 in total, 90 positive, 53 negative) compared to (CK 109 edges, 58 positive, 41 negative) (Supplementary Tables S5, S6). The network structures differed markedly between two treatments, stronger connections observed among Clostridia (Garciella, Anaerotruncus, Anaerosalibacter) and Bacillus in CK, whereas in the JP group, edges were frequent among LAB species (Lactobacillus, Leuconostoc) and specific Clostridia (Clostridium sensu stricto 12, Garciella, Proteus mirabilis). While Minor populations like Weissella and Pediococcus showed minimal network impact despite their functional importance. By contrast, Enterococcus in both groups, and Lactobacillus together with Lactococcus in the JP group, exhibited high degree centrality and therefore exerted a stronger influence on overall network structure.

In fungi networks, the JP group displayed 47 correlation edges (27 positive, 20 negative) slightly fewer than the 50 edges observed in the control (CK) group (26 positive, 24 negative) (Supplementary Tables S7, S8). The two treatments exhibited obviously different co-occurrence structures. In CK, stronger correlations were observed among taxa such Humicola, Chaetomium, Byssochlamys, Rasamsonia, Filobasidium, Thermoascus and Thermomyces. In contrast, the JP group showed increased connectivity among Aspergillus, Olpidium, Alternaria, Malassezia, and Thermomyces. While, the overall dominant genus (Alternaria) remained similar in total abundance between CK and JP, the relative abundance of certain minor taxa shifted noticeably. For instance, Fusarium, which showed moderate abundance in CK, appeared as a minor population in JP. More importantly, functionally relevant taxa such as Humicola, which exhibited high network influence in CK, were substantially reduced in JP and other fungi like Humicola decreased in JP. Overall, taxa with high network degree—indicating greater structural influence—included Alternaria in both groups, and Humicola, Cladosporium, Thermomyces and Thermoascus, which were uniquely influential in the CK network.

The correlation between fermentation parameters and the top 23 OTUs in alfalfa silage were analyzed through multivariate approaches. Principal coordinates analysis revealed clear separation between the fresh materials (cluster 1) and the fermented samples along the X-axis (Figure 5A), reflecting the rapid microbial changes during ensiling. Within the fermented samples, JP-treated and CK groups were distinguished from the in all the sampling time points, demonstrating the influence of JP addition on microbial community in ensiling. However, samples of JP treatment and CK on the same timepoint always showed closer relationships along the Y-axis, while the samples from early-stage (days 1–15) and late-stage (days 30–60) formed two clusters, respectively, demonstrating predictable microbial succession during fermentation. Notably, JP-treated samples reached stable community composition by day 45, while CK samples continued community changes until day 60, suggesting JP accelerated silage maturation.

LEfSe analysis identified 15 OTUs presenting differential abundances in JP treatment and CK (LDA > 2.4, p < 0.05), which may explain JP’s effects on microbial community. JP addition enriched Lactobacillus (LDA = 4.58), Weissella, and Enterobacteriaceae members, while suppressed Clostridium sensu stricto 18 (LDA = −4.65) and Anaerotruncus (LDA = −4.14) (Supplementary Figure S4). These shifts correlated with improved fermentation metrics, particularly reduced pH and ammonia levels. Spearman correlation analysis revealed pH and acetic acid as the strongest influencers on microbial community structure (Supplementary Figure S5 and Supplementary Tables S9–S11). JP-treated silage showed negative correlations with pH throughout the fermentation (p = 0.001), indicating its superior acidification (Supplementary Figure S5).

Detailed correlation patterns emerged between specific OTUs and fermentation parameters are shown in Figures 5B, 6. Taxa of Non-LAB (Burkholderia, Sphingomonas) positively correlated with pH but negatively with production of acids during ensiling, while LAB showed inverse relationships. The JP-upregulated OTUs (Enterobacter, Serratia, Lactococcuss) were positively associated with dry matter but negatively with ammonia-N; while the downregulated taxa (Proteus mirabilis, Bacillus-Clostridium) showed opposite patterns. The OUT Pediococcus was uniquely positive correlated with crud ash content in JP-treated silage, suggesting that microbes may contribute to the mineral retention.

The KEGG functionalities of bacterial communities during alfalfa ensiling, with or without jujube powder (JP), revealed that pathways categorized under “Metabolism” were dominated at first level (Figure 7A and Supplementary Table S12). At level 2, “Amino acid metabolism” and “Carbohydrate metabolism” were the most prominent (Figure 7B and Supplementary Table S13). Compared to the control (CK), JP-treated silage exhibited consistently higher relative abundances (p < 0.05) of “Carbohydrate metabolism” and “Metabolism of other amino acids” across all fermentation time points. Additionally, pathways including “Metabolism of cofactors and vitamins,” “Amino acid metabolism,” and “Nucleotide metabolism” were significantly elevated on day 0, 1, 5, 15 and 60.

At level 3 (Figure 7C and Supplementary Table S14), bacterial communities in JP showed significantly higher abundances (p < 0.05) of “Valine, leucine and isoleucine biosynthesis” throughout the ensiling process. Multiple carbohydrate and amino acid metabolic pathways were also enriched across all sampling days, including “D-Alanine metabolism,” “D-glutamate metabolism Glutathione metabolism,” “Galactose metabolism,” “Starch and sucrose metabolism,” “Glycolysis/Gluconeogenesis,” “Pyruvate metabolism,” and “Pentose phosphate pathway.” Several other pathways—such as “Cysteine and methionine metabolism,” “Alanine, aspartate and glutamate metabolism,” “Glycine, serin and threonine metabolism,” “Lysine biosynthesis,” “Glyoxylate and dicarboxylate metabolism” and “Histidine metabolism,” were significantly enriched on day 0 and 5. Notably, key acidogenic pathways—including “Glycolysis / Gluconeogenesis,” “Pentose phosphate pathway,” and “D-Alanine metabolism”—were consistently higher in JP than in CK throughout the fermentation period.

Fungal functional composition, predicted by FUNGuild, is presented in Figure 8A (Supplementary Table S15). In CK, the initial community was dominated by “Undefined Saprotroph-Wood Saprotroph” (46.7%), followed by “Undefined Saprotroph” (17.5.6%) and “Plant Pathogen” (3.03%). During ensiling, “undefined Saprotroph” group increased from day 0 to day 1, then declined until day 30, before rising again toward day 60. In JP, this functional group accounted for 47.6% initially but declined steadily throughout fermentation, falling to 10% by day 60—a level lower than that in CK. “Plant Pathogen” remained at low abundances (<4%) in CK, but increased over time, stabilizing around 3% by the end, with only slight rises days 45 and 60. Meanwhile, “Undefined Saprotroph-Wood Saprotroph,” which dominated CK initially (46.7%), maintained low relative abundance in JP throughout the ensiling process.

Jujube powder (JP), a widely used feed additive, was employed by this study to address issues in alfalfa silage, such as inadequate fermentation and susceptibility to spoilage caused by insufficient water-soluble carbohydrates (WSC) and high buffering capacity. The pH value, content of dry matter, and water-soluble carbohydrate concentration in fresh alfalfa, and absence of organic acids or NH₃-N in both the raw materials (Table 1) were consistent with the previous reports (Yang et al., 2022).

By the end of ensiling, comparing with CK, the silage quality characters, pH values, LA content, NH₃₋N content (Table 2) and the microbial community composition (Figures1, 2) or diversity (Supplementary Figure S1) in JP treated ensiling reached to stable 15 days earlier, demonstrating that the JP addition accelerated the ensiling procedure. Although the silages of both the treatments fitted the thresholds for well-preserved silage (NH₃-N < 10% of total nitrogen; pH < 4.5) (Kung et al., 1998, 2018) by the end of ensiling, the fermentation procedure in JP-added ensiling was more effective for degreasing pH and production of LA, while inhibiting plant enzymatic activity (NH₃ release by proteolysis) (Ding et al., 2013; Tao et al., 2012). The temporary pH increase occurring on day 5 in both groups could be due to reduced organic acid production and NH₃-N release from protein degradation (Ding et al., 2013; Tao et al., 2012). In the final silage (60 days of fermentation), the greater contents of DM and lactic acid, and the lower pH value, NH₃-N levels, acetic acid content and propionic acid content in JP added silage than that in CK (Figure 1) were consistent with the previous results (Ding et al., 2013; Liu et al., 2016; Rajabi et al., 2017; Tian et al., 2017; Weiss and Underwood, 2009) and clear evidenced that the JP addition improved the quality of high moisture silage. In our study, all the important quality features (Figure 1) reached the values comparable to those achieved in pre-wilted alfalfa silage (Zhang et al., 2018), implying that 4% (w/w) JP addition can be used to replace the energy- and labor-intensive pre-wilting in alfalfa ensiling.

Although the compensative physiochemical/microbial traits of jujube powder (JP) in fresh alfalfa (Table 1) provided a basis for improving alfalfa silage quality, the underlying microbiological mechanisms and how these traits modify the fermentation process remained unclear. Therefore, we further investigated the fermentation and microbial community dynamics during ensiling.

First, the higher water-soluble carbohydrate (WSC) and dry matter (DM) contents, together with the greater initial lactic acid bacteria (LAB) counts in JP (Table 1), promoted rapid oxygen consumption by the aerobic-facultative anaerobic bacteria (Enterobacteria, Garciella, Acinetobacter, and Burkholderia etc.). This in turn stimulated the subsequent growth of anaerobic LAB (Figure 2). Resulting in substantial lactic acid production and a sharp decline in pH during the early stages (day 1) of ensiling (Table 2), consistent with typical fermentation patterns (Muck et al., 2018).

Second, JP treatment significantly modified bacterial community composition compared with CK. Throughout ensiling, JP silage contained fewer Enterococcus and Leuconostoc but more Lactobacillus, likely related to the changes in DM and LAB content induced by JP addition (Su et al., 2023). Notably, during days 15–60, JP-supplied silage showed lower abundance of Clostridium sensus stricto 18 and higher abundances of Clostridium sensus stricto 12, possibly due to fructose-enhanced Lactobacillus growth (Endo et al., 2018). Increased abundances of minor LAB genera such as Weissella, Sporolactobacillus and Pediococcus in JP silage (Supplementary Figures S2, S4) may have further contributed to quality improvement. The persistence of Enterococcus in CK implied incomplete fermentation (Liu et al., 2022), while its stabilization in JP indicated earlier silage maturation.

In both treatments, rapid microbial succession occurred during the early ensiling, with Burkholderia and Acinetobacter disappearing while Lactococcus, Enterococcus and Enterobacter increased markedly by day 1. This shift was likely driven by the large Lactococcus that rapidly initiated fermentation. Bacterial community analysis identified two ensiling phases: early (1–15 days) and middle-late (30–60 days). JP treatment promoted earlier maturation, as evidenced by tight clustering of samples from days 45–60, whereas CK samples remained dispersed throughout fermentation. These patterns align with previous reports on alfalfa silage microbial dynamics (Liu et al., 2019; Zhao et al., 2021).

JP addition also alters fungal community composition, potentially influencing silage quality through mechanisms described in previous studies (Cheli et al., 2013; Wang et al., 2023). JP addition suppressed several potentially harmful fungi, including Phaeosphaeria, Aspergillus, Penicillium and Fusarium, particularly, Alternaria population were decreased by the ensiling process, demonstrating its antifungal properties during ensiling, which might be related to the increased population and metabolic activities of LAB (Fugaban et al., 2023). Alternaria, Aspergillus, Penicillium and Fusarium are well known mycotoxin producers (Kushwaha et al., 2025), while Phaeosphaeria is a genus with many phytopathogens and bioactive compounds (El-Demerdash, 2018). This change in fungal composition further supports that JP addition could improve the quality of alfalfa silage by reducing the mycotoxin levels (Muck et al., 2018).

Third, JP supplementation regulated the microbial community structure. Unlike that in sorghum silage (Gallagher et al., 2018), JP-amended alfalfa silage showed an apparent decoupling between LAB community composition and total abundance, especially in later stages. This suggests that the composition of LAB may exert a stronger influence on fermentation outcomes than overall abundance. Specifically, homofermentative species such as Lactobacillus plantarum dominated early fermentation when sugars are abundant (Li et al., 2018; Wang et al., 2017), indicating significant microbial succession throughout the ensiling process. This pattern aligns with previous observations that only specialized species like Lactobacillus buchneri remain active at low densities during late fermentation stages (Driehuis et al., 1999). JP treatment significantly altered this succession by reducing the abundance of Enterococcus and Leuconostoc while increasing Lactobacillus populations-a shift associated with improved fermentation outcomes (Su et al., 2023). Both JP addition and ensiling duration enhanced microbial diversity and richness, contrary with early findings that LAB and molasses addition in soybean silage decreased bacterial diversity (Ni et al., 2017).

Network analysis revealed that JP restructured microbial interaction, strengthening association among beneficial lactic acid bacteria (LAB) while weakening undesirable connections involving Clostridia. Contrary to previous reports of higher bacterial diversity in untreated silage (Ni et al., 2017), our results showed that JP simplify microbial interactions while promoting beneficial microbial relationships—consistent with findings from LAB-inoculated silages (Liu et al., 2022) and other additive-treated systems (Bai et al., 2021). This reduced network complexity was associated with improved fermentation quality, in agreement with previous studies (Bai et al., 2021).

Certain bacterial taxa played central roles in shaping the network structure. In the control (CK), Enterococcus significantly contributed to network connectivity. Although taxonomically and physiologically classified as a lactic acid bacterium, Enterococcus can contribute to early acidification; however, in silage fermentation practice, it is generally not regarded as a beneficial or efficient probiotic lactic acid bacterium due to its slow acid production rate, limited acid yield, and the propensity of some of species to produce ammino N (Muck et al., 2018). The persistence of Enterococcus in CK may therefore be associated with elevated amino-N levels and accelerated silage deterioration in the mid to late stages of ensiling. These undesirable characteristics may further explain the stronger co-occurrence observed between Enterococcus and know spoilage-associated taxa as Clostridia (e.g., Garciella, Anaerosalibacter) and Bacillus—all of which are recognized contributors to ammonia production and undesirable fermentation (Borreani et al., 2018).

In contrast, JP treatment reshaped the bacterial network, with Enterococcus, Lactobacillus, and Lactococcus emerging as key structural taxa. This restructuring favored positive interactions among LAB species (e.g., Lactobacillus, Pediococcus) and specific Clostridia (Clostridium sensu stricto 12, Proteus mirabilis), which correlated with higher lactic acid accumulation and better pathogen control. Simultaneously, JP suppressed potentially harmful genera such a Garciella, Anaerotruncus and Anaerosalibacter (Supplementary Figure S3). The absence of butyric acid in both treatments corresponed to low clostridial abundance, contrasting with the report of strong clostridial activity in high-moisture alfalfa silage (Zheng et al., 2017). Furthermore, JP treatment reduced undesirable bacteria including Anaerosalibacter, Garciella, and Enterobacter—known to promote ammonia and butyric acid formation (Borreani et al., 2018; Lawson et al., 2004; Pahlow et al., 2003; Togo et al., 2019; Zhang et al., 2018). Compared with commercial LAB inoculants and other chemical additives (Ding et al., 2013; Ni et al., 2017), JP proved to be a more supported greater bacterial diversity, which may contribute to the improved silage quality.

Successional dynamics of three core LAB genera were distinct in JP-treated silage. Lactococcus upregulated by JP, rapidly initiated acid production, peaking on day1. This was early (day 1) subsequent fermentation stages showed characteristic successional patterns, with Lactococcus followed by followed by sequential increases in Enterococcus (day 5), Lactobacillus (days 30–45) and finally Anaerosalibacter (days 45–60). This pattern resembles succession reported corn silage, where early dominance of Lactococcus is succeeded by Lactobacillus and Pediococcus species (Wang et al., 2017). In contrast, down-regulated taxa such as Proteus mirabilis and members of the Bacillus-Clostridium group showed opposing trends, consistent with their roles in ammonification (Muck et al., 2018). JP treatment altered their co-occurrence patterns, generating negative correlations with Sphingomonas and positive links with Lactococcus and Pediococcus.

All these findings collectively demonstrate that jujube powder acts as a multifunctional silage additive that shift microbial community structure by establishing a cooperative consortium of Lactococcus, Lactobacillus and Pediococcus. These three genera likely play complementary roles: Lactococcus drives fast-start fermentation, Lactobacillus sustains strong homofermentative acid production, and Pediococcus maintains activity under prolonged acidic conditions. Together, they contribute to a more efficient and stable ensiling process.

The predominant bacterial functional pathways at level 2 in both CK and JP treatments were associated with carbohydrate and amino acid metabolism. This aligns with expectations, as anaerobic ensiling relies on the conversion of water-soluble carbohydrates (WSC) into organic acids—primarily lactic acid—by lactic acid bacteria (LAB) under oxygen-limited conditions (McDonald et al., 1991). Furthermore, the ensiling process in the JP treatment effectively suppressed amino acid metabolism in many undesirable bacteria, resulting in an overall reduction of this activity compared to CK. The enhanced activities of pathways such as “D-Alanine metabolism,” “Valine, leucine and isoleucine biosynthesis” and glycolysis/gluconeogenesis in the JP group were primarily linked to the rapid proliferation of LAB. This contributed to maintaining protein stability by limiting proteolysis.

D-alanine is not primarily a metabolic energy substrate; rather, it serves as an essential structural component for peptidoglycan biosynthesis in the cell walls of lactic acid bacteria (LAB) (Steen et al., 2005; Palumbo et al., 2004). This role support cell wall integrity, ensuring the survival and functional dominance of these key fermentative bacteria in the acidic, anaerobic silage environment (Palumbo et al., 2004). The presence of D-amino acids, produced by microorganisms during fermentation, has been associated with enhanced taste profiles (e.g., umami, sweet) in foods like cheese and vinegar (Kanauchi and Matsumoto, 2023). In silage, acetic acid derived from metabolism pathways involving alanine or pyruvate acts as an effective antifungal agent. An appropriate concentration of acetic acid significantly improves the aerobic stability of silage after opening, thereby helping to prevent spoilage.

In contrast to D-alanine, valine is a key indicator and participant in detrimental fermentation processes, notably the “Stickland fermentation” (Muck et al., 2018). This process represents the most harmful function of valine in silage, leading to the production of butyric acid and ammonia, along with substantial loss of true protein, which reduces the feed’s nutritional value. Moreover, during later ensiling stages or after exposure to air, certain microbes can maetabolize valine into compounds like isobutanol, imparting an undesirable alcoholic odor.

The higher NH₃-N content observed in CK compared to JP can thus be attributed to divergent roles of D-alanine metabolism and valine metabolism. Concurrently, the increased glycolytic activity detected in the JP treatment indicates an enhanced capacity of LAB to utilize carbohydrates, promoting efficient conversion of glucose to lactic acid by homofermentative strains (Giacon et al., 2022). This more efficient fermentation contributed to the improved preservation outcomes observed with jujube powder addition.

Fungal ecological functions were analyzed using the FUNGuild tool (Nguyen et al., 2016). The analysis revealed a consistently higher proportion of the functional guild “Undefined Saprotroph-Wood Saprotroph” in the control (CK) than in the JP treatment throughout fermentation. This indicates an elevated risk of saprotrophic (decay) activity in untreated alfalfa silage. Furthermore, an increase in the proportion of certain fungal species associated with animal pathogens was noted on days 45 and 60, suggesting prolonged fermentation may heighten this risk. The JP treatment effectively suppressed “Undefined Saprotroph” fungi. This suppression likely resulted from the rapid and dominant fermentation initiated by LAB in the JP-treated silage, which inhibited fungal populations through acid production and competitive exclusion.