Alfalfa (Medicago sativa L.) was harvested from the Yellow River Delta Modern Agriculture Experimental Demonstration Base of the Shandong Academy of Agricultural Sciences, located in the Nonggao district of Dongying City (37°18′N, 118°37′E) at an altitude of 2 m. The region has a warm temperate monsoon climate. The alfalfa variety used was WL440, and the cutting period was the end of the fourth budding stage, with a stubble height of 5 cm. The moisture content of alfalfa was adjusted to three levels: (1) approximately 78%, without wilting; (2) approximately 68%, wilted for almost 2 h; and (3) approximately 58%, wilted for almost 4 h. Real-time moisture measurements were performed using a microwave oven. After wilting, the alfalfa was chopped into 1–2 cm pieces using a chaff cutter (9Z-0.4, Jinniu Machinery Factory, Rongyang, Henan Province). A total of three replicates of alfalfa at each moisture content were reserved for measuring dry matter (DM) and CP contents. In total, four additive treatment groups were established for each moisture content of alfalfa: (1) Blank control (CK, with an equal amount of distilled water), (2) commercial L. plantarum (1 × 10⁶ cfu/g FW, CL, Yiqing No. 2, purchased from Beijing Zhongqing Yicao Technology Co., Ltd.), (3) L. plantarum (1 × 10⁶ cfu/g FW, LP, screened from sheep skin soaking solution by the Grass Product Processing Laboratory of China Agricultural University), and (4) propionic acid (P, applied at 6 mL/kg FW, diluted to a concentration of 45%). The alfalfa was thoroughly mixed with the additives, and approximately 200 g of the mixture was placed into polyethylene vacuum bags. The bags were sealed using a vacuum packaging machine. In total, 3 moisture levels * 4 additive treatments * 3 replicates resulted in 36 silage bags. All bags were stored at 25 °C (room temperature) and opened after 45 days of ensiling.

On the day of opening, a 10 g sample was collected and mixed with 90 mL of distilled water, then homogenized using a liquidizer for 1 min. The mixture was subsequently filtered through four layers of sterile gauze followed by qualitative filter paper. The pH was measured using a LEI-CI PHS-3C pH meter (Shanghai Yidian Scientific Instrument Co., Ltd., Shanghai, China). As previously described (22), the contents of lactic acid (LA), acetic acid (AA), propionic acid (PA), and butyric acid (BA) were determined usingHPLC. The ammonia nitrogen (AN) content was measured following the method described by Broderick and Kang (23). Flieg’s score was used to evaluate the fermentation quality of alfalfa silage, calculated as follows (24): Flieg’s score = 220 + (2 × %DM − 15) − 40 × pH.

The DM content was measured as previously reported (9). Samples were dried at 65 °C for 48 h in a heating and drying oven (LC-202-00, Shanghai Lichen Instrument Technology Co., Ltd., Shanghai, China). The CP content was measured using the dried samples, which were ground with a grinder (BJ-800A, Deqing Baijie Electric Appliance Co., Ltd., Huzhou, Zhejiang, China) and passed through a 1 mm sieve. The CP content was calculated as total nitrogen (TN) multiplied by 6.25. TN was determined using an automatic Kjeldahl nitrogen apparatus (Kjeltec 2300 AutoAnalyzer, FOSS Analytical AB, Hoganas, Sweden) according to the AOAC (25) standard.

Microbial genomic DNA was extracted from ensiled alfalfa using the E.Z.N.A.® Soil DNA Kit (Omega Bio-tek, Norcross, GA, USA) according to the manufacturer’s instructions. Then, 1.0% agarose gel electrophoresis and a NanoDrop® ND-2000 spectrophotometer (Thermo Scientific Inc., Waltham, MA, USA) were used to assess the quality and concentration of the DNA, and the DNA was stored at −80 °C until further analysis. The primer pair 338F (5′-ACTCCTACGGGAGGCAGCAG-3′) and 806R (5′-GGACTACHVGGGTWTCTAAT-3′) was used to amplify the bacterial 16S rRNA gene hypervariable region V3-V4 on an ABI GeneAmp® 9700 PCR thermocycler (ABI, Foster City, CA, USA). PCR was performed following the previously described method (26). The PCR product was extracted using 2% agarose gel, purified using the AxyPrep DNA Gel Extraction Kit (Axygen Biosciences, Union City, CA, USA) according to the manufacturer’s instructions, and quantified using a Quantus™ Fluorometer (Promega, USA). Purified amplicons were pooled in equimolar amounts and sequenced using paired-end reads on an Illumina MiSeq PE300 platform (Illumina, San Diego, USA) according to the standard protocols by Majorbio Bio-Pharm Technology Co., Ltd. (Shanghai, China). Raw FASTQ files were de-multiplexed using an in-house Perl script, quality-filtered with fastp version 0.19.6, and merged using FLASH version 1.2.7. Then, the optimized sequences were clustered into operational taxonomic units (OTUs) at 97% sequence similarity using UPARSE 7.1.

The experiment was conducted as a completely randomized design with a 3 × 4 (three moisture levels and four additives) factorial arrangement. Analyses of fermentation characteristics and chemical composition were performed using the SAS software (version 9.4, SAS Institute Inc., Cary, NC, USA). A mixed model procedure (GLM) was used as described by Gao et al. (21) as follows:

Where Y_ijk is the dependent variable representing the response for the moisture content i observed under additive j, μ is the mean, M_i is the moisture content effect, A_j is the additive effect, (M × A)_ij is the interaction effect of moisture content and additive, and e_ijk is the residual error.

ANOVA was used to evaluate the effects of additives at each moisture level or the effects of moisture levels for each additive. The model used was as follows:

Where Y is the response variable, μ is the overall mean, M_i is the effect of the moisture content, A_i is the effect of the additive, and ε_ij is the residual error. Post hoc mean comparisons were performed using Tukey’s multiple range test. Statistical significance was set at a p-value of ≤0.05. Microbial data analysis was conducted on the Majorbio Bio-Pharm Technology Co., Ltd. (Shanghai, China) cloud platform.¹

The moisture content of alfalfa without wilting was 78.31%; after 2 h of wilting, it was 67.32%; and after 4 h of wilting, it was 59.73%. The crude protein (CP) content before ensiling was 23.32%.

Moisture had a significant impact on pH, AA, AN, Flieg’s score, DM, and CP (p < 0.05). The additive significantly affected pH, LA, PA, BA, AN, Flieg’s score, DM, and CP (p < 0.05). Moreover, the interaction between moisture and additive had a significant effect on pH, LA, PA, BA, AN, Flieg’s score, DM, and CP (p < 0.05) in alfalfa silage (Table 1).

The effects of the moisture contents and additives on the fermentation characteristics of alfalfa silage are shown in Table 2. The pH of the P group at the 78% moisture content was significantly lower than that of the other treatment groups (p < 0.05). The CL, LP, and P treatments at the 68% moisture content had significantly lower pH values than the CK group (p < 0.05). At the 58% moisture content, the pH of the P group was significantly higher than that of the CK, CL, and LP groups (p < 0.05). At the 78% moisture content, the LA contents of the CL and P groups were significantly higher than those of the CK and LP groups (p < 0.05). At the 68% moisture content, the LA contents of the LP and P groups were significantly higher than those of the CK and CL groups (p < 0.05). At the 58% moisture content, the LA contents of the CL and LP groups were significantly higher than those of the CK and P groups (p < 0.05). The LA content of the CL group at 78 and 58% moisture contents was significantly higher than that at the 68% moisture content (p < 0.05). In the P group, the LA content at 78 and 68% moisture contents was significantly higher than that at the 58% moisture content (p < 0.05). The PA content of the P group was significantly higher than that of the CK and LP groups at the 78% moisture content (p < 0.05) and higher than that of the other groups at the 68% moisture content (p < 0.05). The AN content of alfalfa silage showed a decreasing trend with decreasing moisture content. The AN content of the CK group was significantly higher than that of the other groups at 78 and 58% moisture contents (p < 0.05). The Flieg’s score of the P group was significantly higher than that of the CL, LP, and CK groups at the 78% moisture content (p < 0.05). The CK group had a significantly lower Flieg’s score than the other groups at the 68% moisture content (p < 0.05). The Flieg’s score of the CL and LP groups was significantly higher than that of the CK and P groups at the 58% moisture content (p < 0.05).

The effects of different moisture contents and additives on the DM and CP content of alfalfa silage are shown in Table 3. The DM of the P group was significantly higher than that of the CK and CL groups at the 78% moisture content (p < 0.05). The DM of the CL, LP, and P groups was significantly higher than that of the CK group at the 68% moisture content (p < 0.05). The DM of the LP and P groups was significantly higher than that of the CK group at the 58% moisture content (p < 0.05). The CP content of the CL, LP, and P groups was significantly higher than that of the CK group at the 68% moisture content (p < 0.05).

The alpha and beta diversity of alfalfa silage among the four additive groups showed no significant difference at ecah moisture contents (Figures 1A–E), except for Shannon index at the 58% moisture content (Figure 1F). Alpha diversity analysis revealed that the LP group of alfalfa silage had a significantly lower Shannon index than the CK and P groups at the 58% moisture content (p < 0.05). A marked divergence in the Shannon index among the four additive treatments at the 58% moisture content indicated that the LP treatment reduced the diversity of certain microbial communities. Principal coordinates analysis (PCoA) demonstrated clear separation among the additive groups at each moisture content (Figures 1G–I), indicating that different additives caused marked differences in bacterial community composition at each moisture content.

At the phylum level, the predominant phyla across the 36 samples were Firmicutes and Proteobacteria (Figures 2A,C,E; Supplementary Table S1). Cyanobacteria, Actinobacteriota, and Deinococcota showed significant differences among the four groups at the 78% moisture content. Deinococcota showed significant differences among the four groups at the 68% moisture content. Actinobacteriota, Cyanobacteria, and Verrucomicrobiota showed significant differences among the four groups at the 58% moisture content (Figures 3A–C; Supplementary Table S2).

At the genus level, the dominant genera across the 36 samples were Lactobacillus and Enterococcus (Figures 2B,D,F; Supplementary Table S2). Among the four treatment groups, 19 genera showed significant differences at the 78% moisture content, 13 at the 68% moisture content, and 37 at the 58% moisture content (Figure 3). The abundance of Lactobacillus was higher in the additive groups compared to the CK group, whereas the abundance of Enterococcus was lower in the additive groups compared to the CK group (Figure 2; Supplementary Table S2). The abundance of Lactobacillus was lower in the P group at the 58% moisture content, while Methylobacterium-Methylorubrum, norank_f__norank_o__Chloroplast, Sphingomonas, Aureimonas, and Allorhizobium-Neorhizobium-Pararhizobium-Rhizobium together accounted for more than 59% of the bacterial community (Supplementary Table S2). Enterococcus was the dominant genus in the CK group at all moisture contents.

At KEGG pathway level 1, six categories—metabolism, environmental information processing, genetic information processing, human diseases, cellular processes, and organismal systems—were identified across the 36 samples at each moisture content of alfalfa silage (Supplementary Figure S1). The metabolism pathway was the most active (Table 4). There was a significant difference between the CK group and the CL group at the 78% moisture content. A significant difference was observed between the CK group and the LP group at the 68% moisture content. At the 58% moisture content, the CK group showed a significant difference compared to the LP group, and the P group was significantly different from the CL and LP groups.

At KEGG pathway level 2, carbohydrate metabolism and amino acid metabolism were the most active pathways (Supplementary Figure S2). The LP group showed higher carbohydrate metabolism activity in alfalfa silage at all moisture contents (Table 5), which was significantly higher than that of the CK group at the 78% moisture content, the CL group at the 68% moisture content, and the P group at the 58% moisture content (Supplementary Figure S3). The additives promoted amino acid metabolism in alfalfa silage (Table 5). The amino acid metabolism activity of the LP group was significantly higher than that of the CK group at the 78% moisture content and the CL group at the 68% moisture content. The CL and LP groups showed significantly higher amino acid metabolism activity than the P group at the 58% moisture content (Supplementary Figure S3).

At KEGG pathway level 3, the top pathway for carbohydrate metabolism was starch and sucrose metabolism, while the top pathway for amino acid metabolism was cysteine and methionine metabolism (Supplementary Figure S5). No significant difference was observed in starch and sucrose metabolism at the 78% moisture content. A significant difference between the CL and LP groups was observed at the 68% moisture content. Starch and sucrose metabolism in the P group was significantly higher than that in the CL and LP groups at the 58% moisture content (Table 6). Cysteine and methionine metabolism in the LP group was significantly higher than that in the CK group at the 78% moisture content. The LP group showed higher cysteine and methionine metabolism than the CK and CL groups at the 68% moisture content. Furthermore, cysteine and methionine metabolism in the LP group was significantly higher than that in the CK and P groups at the 58% moisture content, and it was also significantly higher in the CL group compared to the P group (Table 6).

The Spearman correlation between bacterial diversity, fermentation characteristics, and chemical composition is shown in Figure 4. It was found that pH was negatively correlated with Lactobacillus and Sphingomonas and positively correlated with Enterococcus, Weissella, Leuconostoc, and Enterobacter. The LA content was negatively correlated with Enterococcus, Leuconostoc, and Enterobacter and positively correlated with Lactobacillus. The AN content was negatively correlated with Methylobacterium-Methylorubrum, Sphingomonas, and Aureimonas. The BA content was positively correlated with Weissella. Flieg’s score was negatively correlated with Enterococcus, Weissella, Leuconostoc, and Enterobacter and positively correlated with Sphingomonas, Aureimonas, and Lactobacillus. The DM content was negatively correlated with Weissella, Leuconostoc, and Enterococcus and positively correlated with Sphingomonas, Aureimonas, and Methylobacterium-Methylorubrum.