The alfalfa used for silage was harvested in June 2021 in Kailu County, Tongliao City, Inner Mongolia Autonomous Region (latitude 43°9′-44°10′ N, longitude 120°25′-121°52 E) and the alfalfa samples were cut to 1–2 cm with a forage crusher. The lyophilized powder of L. plantarum (10¹¹CFU/g) was purchased from Beijing Baiobowei Biotechnology Co. Ltd and molasses (42%-50% sugar, Brix ≥ 60) was purchased from Weifang Fengguan Biotechnology Co. Ltd.

A single-factor experimental design was adopted in the experiment, and four treatment groups were set up: control (CK), Lactobacillus plantarum (L), molasses (M), and Lactobacillus plantarum + molasses (LM). In groups L, M, and LM, 10⁶ CFU/g Lactobacillus plantarum lyophilized powder, 5% molasses, 10⁶ CFU/g Lactobacillus plantarum + 5% molasses were dissolved in 10 mL of sterile distilled water and evenly mixed in raw materials, respectively. In the CK group, 10 mL sterile distilled water was added and evenly mixed. The final mixes of each group (300 g) were packed into a special polyethylene plastic silage bag (27 cm × 39 cm, thickness 0.24 mm), sealed with a vacuum sealer, and fermented at a temperature of 23–25 °C away from light. Triplicate silages were prepared for each group, and samples were collected and analyzed after 14 days and 56 days, respectively.

Alfalfa silage was sampled and dried in a constant temperature drying oven (DHG-9240A, Shanghai, China) at 65°C for ~48 h to a constant weight to determine the dry matter (DM) (Zhang et al., 2016). The dried material samples were smashed and passed through a 1.0 mm sieve for nutrient analysis. The contents of crude protein (CP), water-soluble carbohydrate (WSC), and ether extract (EE) were determined using the Kjeldahl method (AOAC, 2000), anthrone colorimetry (Wu and Nishino, 2016), and the Soxhlet extraction method (AOAC, 2000), respectively. Neutral detergent fiber (NDF) and acid detergent fiber (ADF) were determined using a semi-automatic fiber analyzer (ANKOM 200i) according to the method of Van Soest (Van Soest et al., 1991).

The silage samples (20 g) were taken and homogenized with 180 mL of distilled water in a blender for 1 min and then filtered through 4 layers of gauze and qualitative filter paper. The pH value of the leachate was immediately determined with a portable pH meter (SX-620, Shanghai). The contents of lactic acid (LA), acetic acid (AA), propionic acid (PA), and butyric acid (BA) were determined by gas chromatography (model GC-6800, Beijing Beifen Tianpu Instrument Technology Co., Ltd.), and the chromatographic conditions were as follows: Φ6 mm × 2m quartz glass packed column (15% FFAP as stationary phase, 80–100 mesh Chromosorb as tensile strength), column temperature of 150°C, inlet temperature of 220°C. The column temperature was 150°C, the inlet temperature was 220°C, the injection volume was 1 μL, the FID detector temperature was 280°C, the carrier gas was high purity N₂ with a flow rate of 30 mL/min, and a pressure of 200 kPa, the gas was H₂ with a flow rate of 30 mL/min and the combustion gas was the air with a flow rate of 300 mL/min. The LA content was determined using high-performance liquid chromatography (Shimadzu LC-20A). Chromatographic conditions were as follows: the mobile phase was a solution of 3 mmol/L perchloric acid at a flow rate of 1 mL/min, the column temperature was 60°C, and the detector was an SPD-20A ultraviolet detector at a detection wavelength of 210 nm.

Another 20 g of silage was sampled and homogenized with 180 mL of sterilized normal saline and remained for 30 min to make a 10¹ bacterial diluent, then 10², 10³, 10⁴, 10⁵, and 10⁶ bacterial diluents were prepared sequentially in the aseptic operating table. LAB, yeast, and mold in silage were counted using the spread plate method with MRS agar medium (MRS) and potato dextrose agar (PDA), respectively. LAB was continuously cultured for 2 days in an anaerobic incubator at 37 °C, and yeast and mold were cultured continuously for 2–3 days at 30°C under aerobic conditions. Smooth, round white colonies on the medium were used for LAB counting; moist, smooth, opaque, large, and flat colonies on the medium were used for yeast counting; loose colonies with slender, fluffy, and flocculated mycelium on the medium were used for mold counting, and the unit of colony counting was expressed as log10 CFU/g.

After 14 days and 56 days of ensiling, samples (100 g) of each group were collected, respectively, and placed in a special polyethylene tank without sealing and compaction, but covered the mouth with 2 layers of gauze and a thermometer probe with a range of 50°C inserted into the center of the silage samples to determine the temperature. The temperature was recorded every 0.5 h and the recording was stopped when the silage temperature exceeded the ambient temperature by 2°C (that is, aerobic stability was reached).

The experimental data were summarized by Excel (2010) and statistically analyzed using SPSS version 19.0 for Windows (SPSS Inc., Chicago, IL, USA). One-way analysis of variance (ANOVA) and LSD methods were performed for statistical analysis and multiple comparisons.

The mean pH value of fresh alfalfa sap was 7.02, and the nutrient composition of alfalfa before silage is shown in Table 1, where the numbers of LAB, yeast and mold are shown in Table 2.

As shown in Table 3, after 14 days of ensiling, the DM content of the alfalfa silage in group M and LM was extremely significantly higher than in group CK (P < 0.01), but the difference between group L and CK was not significant (P > 0.05). There were no significant differences in the CP content among the groups (P > 0.05). The EE content was extremely significantly higher in groups L, M, and LM than in group CK (P < 0.01). The WSC content was extremely significantly lower in groups L, M, and LM than in group CK (P < 0.01). The NDF and ADF content was extremely significantly lower in groups M and LM than in group CK (P < 0.01).

After 56 days of ensiling, the alfalfa DM content was not significantly different among groups (P > 0.05) but was significantly higher in the LM group than in the M group (P < 0.05). There were no significant differences in the CP content among the groups (P > 0.05). However, the EE content was extremely significantly (P < 0.01) higher in groups M and LM than in group CK. The WSC content was extremely significantly higher in groups L, M, and LM than in group CK (P < 0.01). The content of NDF and ADF in all other groups were extremely significantly lower than in the CK group (P < 0.01).

As shown in Table 4, after 14 days of ensiling, the pH values in group L, M, and LM were extremely significantly lower than those in group CK (P < 0.01), while the difference in pH values between group M and LM was not significant (P > 0.05). The LA content of groups M and LM was extremely significantly higher than in groups CK and L (P < 0.01), while there was no significant difference in LA content between groups M and LM (P > 0.05). There was no significant difference in the content of AA and PA in all groups (P > 0.05), and BA was not detected in any of the groups.

After 56 days of ensiling, the pH values of group L, M, and LM were extremely significantly lower than that of group CK (P < 0.01), and the pH values were not significant between group M and LM (P > 0.05). While the LA content was extremely significantly higher in groups M and LM than in other groups (P < 0.01), and extremely significantly higher in group L than in group CK (P < 0.01), but there were no significant differences between groups M and LM (P > 0.05). The AA content was extremely significantly higher in group L than in other groups (P < 0.01), while there was no significant difference among groups CK, M, and LM (P > 0.05). The PA content was significantly higher in group CK than in groups M and LM. BA was not detected in any of the groups.

As shown in Table 5, after 14 days of ensiling, the bacterial count of LAB in groups from high to low was LM > M > L > CK (P < 0.01). And LAB in the LM group was extremely significantly higher than in other groups (P < 0.01), but the difference between the L and M groups was not significant (P > 0. 05). The yeast bacterial count in group L, M and LM were extremely significantly higher than in group CK (P < 0.01), but the difference between group L and LM was not significant (P > 0.05). Mold was detected in all groups. The aerobic stabilization time from long to short was L > LM > M > CK, respectively (P < 0.01).

After 56 days of ensiling, the bacterial count of LAB from high to low was LM > M > L > CK, respectively (P < 0.01). The yeast bacterial count in groups CK and M were extremely significantly higher than that in groups L and LM (P < 0.01). Mold was detected in LM group. The aerobic stabilization time from long to short was in group L > LM = M > CK (P < 0.01).

A person correlation analysis was performed between nutrients and fermentation parameters with bacterial communities. As in the heatmap shown in Figure 5A, after 14 days of ensiling, the LA content was significantly negatively correlated (P < 0.05) with Enterobacter, Hafnia-Obesumbacterium, unclassified_f_Enterobacteriaceae, Cosenzaea, Enterococcus, and Pantoea. The WSC content was extremely significantly negatively correlated (P < 0.001) with Lactobacillus and extremely significantly positively correlated (P < 0.001) with Hafnia-Obesumbacterium, Lactococcus, Enterococcus, unclassified_o_Lactobacillales, Enterobacter, and was extremely significantly positively correlated with unclassified_f_Enterobacteriaceae, Pantoea, Leuconostoc and Cosenzaea (P < 0.01). The pH value was extremely significantly positively correlated (P < 0.001) with Hafnia-Obesumbacterium, Enterobacter, and significantly positively correlated (P < 0.001) with Enterococcus, unclassified_f_Enterobacteriaceae, Cosenzaea, Lactococcus, Pantoea, and Leuconostoc were extremely significantly positively correlated (P < 0.01).

As shown in Figure 5B, after 56 days of ensiling, the LA content was significantly negatively correlated with Enterobacter, Anaerosporobacter, Caproiciproducens, Eubacterium_fissicatena_group, Clostridium_sensu_stricto_12 (P < 0.05). The WSC content showed a very significant negative correlation (P < 0.01) with Enterobacter, Anaerosporobacter, Leuconostoc, Clostridium_sensu_stricto_12, Eubacterium_fissicatena_group, and a significant negative correlation (P < 0.05) with Caproiciproducens, Enterococcus. The pH value was extremely significantly positively correlated (P < 0.001) with Enterobacter, Anaerosporobacter, Caproiciproducens, Clostridium_sensu_stricto_12, the Eubacterium_fissicatena_group, was significantly positively correlated with Leuconostoc, Enterococcus, and Clostridium_sensu_stricto_18 (P < 0.05).

The 16S rRNA gene predictive function profiles at level 1 (Figures 5A, B), level 2 (Figures 6C, D), and level 3 (Figures 6E, F) of the pathways are illustrated in Figure 6. After 14 days and 56 days of ensiling, the metabolic pathways of metabolism, environmental information processing, and genetic information processing were significantly higher than other metabolic pathways (Figures 6A, B). Among them, carbohydrate metabolism, membrane transport, and amino acid metabolism were much higher than other metabolic pathways (Figures 6C, D). A more detailed comparison at the level of the third pathway showed that the metabolism of ABC transporters, the two-component system, purine metabolism, and aminoacyl-tRNA biosynthesis was significantly higher than the other metabolisms (Figures 6E, F). However, there were almost no significant differences in the metabolism of all groups at three levels of pathways after 14 days and 56 days of ensiling. But the carbohydrate metabolism and membrane transport in group M were extremely significantly higher than in other groups after 14 days of ensiling (P < 0.01).

Studies have shown that the proliferation of undesirable microorganisms such as yeast, mold, and Brucella during silage is the main cause of DM loss (Wang et al., 2017). Compared to the CK group, the retention of DM was higher in the M and LM groups, suggesting that M and LM contributed to the reduction of the loss of DM from fermentation. On the contrary, the significant decrease in DM after the addition of L could be attributed to the addition of LAB to increase the intensity of fermentation, which promoted the decomposition of carbohydrates, and carbon dioxide that escaped within the conversion process to LA into AA and PA (Randby et al., 2012). This indicates that although the addition of L improved the fermentation speed and fermentation quality, it sacrificed a part of DM, and the final effect should be evaluated in combination with the content and digestibility of other nutrients in the silage. The CP content is an important indicator to determine the nutritional value of silage, which includes nitrogenous substances in the food as well as microbial proteins (MCP). In this experiment, the difference in CP content among the groups was not significant (P > 0.05), and even the addition of mycoprotein in group L did not cause a significant effect, indicating that L, M, and LM have less influence on the content of nitrogenous substances in the silage process, which is in line with the results of Zhu et al. (2022). The main reason for this is that the proteins in the raw material are preserved when the pH value is low enough to limit the activity of protein hydrolyzing bacteria (Luo et al., 2021). WSC is the main substrate for LAB fermentation, and during silage fermentation, it is consumed to produce LA, so WSC is the most important component for the production of high-quality silage (Silva et al., 2020; Ke et al., 2023). On the other hand, NDF and ADF are also important factors that influence animal feed intake and digestibility of silage (Li et al., 2019; Gül, 2023), and can be fermented by microorganisms to produce WSC or further fermented to produce organic acids. In this experiment, after 14 days of ensiling, L, M, and LM significantly reduced the content of WSC, NDF, and ADF, and increased the content of organic acids (LA) in alfalfa silage, indicating that the addition of Lactobacillus plantarum and molasses promoted the decomposition of carbohydrates and the production of organic acids. The microbial degradation of WSC, NDF, and ADF occurred almost simultaneously (Table 4). When compared in terms of WSC content, the degree of fermentation from fastest to slowest was LM > L > M > CK, and compared in terms of NDF and ADF content, the degree of fermentation from fastest to slowest was M > LM > L> CK. From this result, it is clear that the addition of Lactobacillus plantarum promotes the decomposition of nonstructural carbohydrates predominantly in the early stage of fermentation, whereas the addition of molasses promotes the decomposition of structural carbohydrates more effectively. According to the results of this experiment, the change in pH value (Table 4) was consistent with the WSC content (Table 3), and the lower WSC content was accompanied by a lower pH value, which, from this point of view, indicated that the addition of Lactobacillus plantarum (pH = 5.12) to promote the effect of early fermentation was better than molasses (pH = 5.50), while the best effect was achieved by adding the two at the same time (pH = 4.96). In terms of microflora, WSC was negatively correlated with Lactobacillus plantarum and positively correlated with Enterococcus, Enterobacter, Pantoea, and other undesirable fermentation-related microorganisms (Figure 5A), suggesting that the reduction of WSC and the resulting low pH values play an important role in the pre-silage period to inhibit the proliferation of undesirable microorganisms and improve the effect of silage.

After 56 days of ensiling, the content of NDF and ADF continued to decrease in groups L, M, and LM, but the content of WSC was extremely significantly higher than that of group CK (P < 0.01), indicating that at the later stages of fermentation, structural carbohydrates (NDF and ADF) were consumed mainly in group L, M and LM, but still maintained high organic acid production and low pH value, suggesting that carbohydrate fermentation at this time was still efficient and did not depend on fermenting WSC. From the changes in the bacterial genus shown in Figure 4, it can be seen that after 14 days of ensiling, the higher abundance of g_Carnobacterium in the CK group could promote the decomposition of NDF and ADF (Korsa et al., 2022), and after 56 days of ensiling, the increased abundance of f_Lactobacillaceae and g_Clostridium_sensu_stricto_12 led to a large amount of consumption of WSC (Tang et al., 2017).

In this experiment, the proliferation of LAB during fermentation was promoted in groups L, M, and LM, and LAB-produced plasma lysozyme can degrade NDF and ADF (Korsa et al., 2022). Meanwhile, LAB promoted the conversion of fiber degradation products (such as WSC) to organic acids and lowered the pH value of silage, while a lower pH value can allow acid hydrolysis of digestible plant cells and further lead to increased digestion of NDF and ADF (Luo et al., 2021). Although the difference in NDF and ADF between group L and CK was not significant (P > 0.05) after 14 days of ensiling, it was extremely significantly lower than the CK group after 56 days of ensiling (P < 0.01). From microflora data, it is known that this may be due to Weissella proliferation after 56 days, which has genes to break down hemicellulose and cellulose (Hernández-Oaxaca et al., 2021).

After 14 days and 56 days of ensiling, the EE content of the M and LM groups was extremely significantly higher than the CK group (P < 0.01), and group L was extremely significantly higher than the CK group after 14 days of ensiling (P < 0.01), which may be attributed to fiber degradation or WSC, increasing the relative percentage of EE.

The pH value in the silage should be low enough to inhibit the multiplication of spoilage microorganisms and protein hydrolysis activity, thus reducing the loss of DM in the silage (Hattori et al., 1996). Insufficient acidification promotes the vigorous growth of Clostridium butyricum in the later stages of silage, which is the main cause of clostridial fermentation in alfalfa silage (Li et al., 2020). Studies have shown that the quality of silage fermentation is good when the pH value is 4.5 or lower (Hashemzadeh-Cigari et al., 2014). In this experiment, the pH values of alfalfa silage in groups L, M, and LM were significantly lower after 14 days and 56 days of ensiling and reached the ideal pH value of 4.5, where the pH values from low to high were LM <M <L, respectively, and the pH value of group CK (5.60) was extremely significantly higher than other groups and did not reach the ideal value (P < 0.01). This trend was already established at the early stage of 14 days, when the pH value of the CK group was 6.46, and the LM, M, and L groups had already decreased below 5.5, respectively. The pH value was negatively correlated with Lactobacillus and positively correlated with undesirable microorganisms such as Enterobacter, Enterococcus, and Clostridium (Figure 5); a lower pH value inhibited the proliferation of undesirable microorganisms. Therefore, inoculation with Lactobacillus plantarum and the addition of molasses was very effective in reducing the pH value of the silage quickly and improving the quality of fermentation.

The production of organic acids is the main cause of the drop in the pH value in silage. Among them, LA is the ideal fermentation product in silage, produced mainly by WSC-consuming LAB, while AA, PA, and BA are less desirable (Besharati et al., 2022). According to the fermentation pattern of LAB, the fermentation products are mainly LA, while the products of mixed fermentation of multiple bacteria are mainly AA and other organic acids (Guo et al., 2018). In this experiment, the increase in LA content in groups L, M, and LM indicated that WSC was fully used, which was the main reason for the decrease in pH value, and the negative correlation between LA content and undesirable microorganisms, such as Enterobacter, Enterococcus and clostridium (Figure 5), indicated that LA could inhibit the proliferation of undesirable microorganisms. After 56 days of ensiling, the LA content of group M and LM was extremely significantly higher than that of group CK and L (P < 0.01), and the pH value of group M and LM was extremely significantly lower than that of group CK and L (P < 0.01), and the LA yield increased. Overall, although both Lactobacillus plantarum inoculation and molasses addition improved the fermentation quality of alfalfa silage, groups M and LM were more effective in controlling low pH values. Acetic acid (AA) is usually the second most concentrated organic acid in silage fermentation after lactic acid and has been proved to guarantee aerobic stability as an inhibitor of spoilage organisms (Danner et al., 2003). However, too much AA does not seem to be preferable, as too much acetic acid can reduce dry matter intake (DMI) of animals for palatability reasons. Studies have shown that an increase of 1 g acetic acid/kg DM led to a reduction of 1.2 g in DMI (per 100 kg body weight of animals) for acetic acid concentrations <17.3 g/kg DM (Gerlach et al., 2021). Therefore, the significant increase in AA content in group L caused by the addition of lactic acid bacteria does not seem to be the optimal result. PA is produced mainly by hetero-lactic bacteria, Propionibacterium, and Enterobacter, and is also degraded by citrate, malate, and amino acids (Tian et al., 2020). BA is produced primarily by fermentation of amino acids by Clostridium butyricum, leading to loss of nutrients and poor preservation, but BA is inhibited under well-fermented silage and lower pH value conditions (Krizsan and Randby, 2007). In this experiment, the PA content of groups L, M, and LM was kept at a lower level than that of group CK, indicating that a low pH value inhibits Propionibacterium. Meanwhile, the BA content was not detected in all treatment groups, indicating that alfalfa silage fermentation was good, with no increase in crude protein consumption (Table 3) and less acid production from poor fermentation.

The maturity and type of silage fermentation are influenced by changes in the composition and abundance of the microbial community attached to the feed; in terms of metabolism, LAB are anaerobic, yeast is parthenogenetic anaerobic, and both have a similar cellular structure, yeast can degrade starch and sugar, while LAB degrades organic matter (Thomsen et al., 2007). Aerobic microorganisms, such as yeast, in the raw material dominate the early stages of ensiling and consume oxygen and some nutrients. When oxygen is reduced and conditions are suitable, LAB proliferates rapidly and secretes LA to inhibit the growth of aerobic microorganisms such as yeast and eventually become dominant (Ferrero et al., 2021; Okoye et al., 2023). However, this may be a kind of dynamic inhibition because yeast can still be isolated from sealed silage. Therefore, the presence of LAB and a lower pH value are necessary to inhibit yeast growth.

LAB is effective in lowering the pH value of silage reducing the population of yeast, and improving aerobic stability (Besharati et al., 2020a; Ferrero et al., 2021). In this experiment, the LM group had the lowest pH value after both 14 days and 56 days of ensiling, indicating that the LAB in this group had the highest intensity of activity and high aerobic stability, which is consistent with the results of aerobic stability (Table 5). After 56 days of ensiling, the pH values of group L, M, and LM were all <4.5, while group CK had not yet been reduced to <6.5. From the microflora point of view, there is a risk of undesirable fermentation mainly since the percentage of Firmicutes abundance in group CK is still low, and especially Lactobacillus are far from being overwhelmingly predominant to allow a rapid reduction of the pH value to the desired value. The same can be seen from the results of the bacterial count, which were extremely significantly lower (P < 0.01) in the CK and L groups than in the M and LM groups after 56 days of ensiling. At the same time, the yeast count was extremely significantly higher in the CK and M groups than in the LM group. Yeast can consume LA, leading to an increase in the pH value of the silage, making the silage more susceptible to deterioration due to aerobic activity (Thomas, 1991). Furthermore, the content of AA during silage fermentation should also be emphasized, and AA is often used as a poor silage fermentation inhibitor. Studies have shown that silage containing high concentrations of AA has lower yeast and mold populations and higher aerobic stability, AA can effectively reduce the pH value of silage, and its inhibitory effect on yeast is even greater than LA (Danner et al., 2003; Liu et al., 2020; Selwet, 2020; Yi et al., 2023), which is obvious in its inhibitory effect. In this experiment, although the difference in AA content was not significant in all groups after 14 days of ensiling (P > 0.05), the AA content of group L was extremely significantly higher than other groups after 56 days of ensiling (P < 0.01), while the yeast count of group L and LM was also extremely significantly lower than in group CK and M (P < 0.01). Furthermore, the L group with the highest AA content also had the highest aerobic stability, which confirms the bacteriostatic effect of AA, which is consistent with the results of Kim et al. (2016).

The diversity and abundance of bacteria in the alfalfa silage was lower (P < 0.05) in group L and LM than in group CK and M after 14 days and 56 days of ensiling, which may be related to the addition of LAB so that they play a dominant role in the silage. This result is supported by several studies showing that the limited diversity of microbial in acidic environments is influenced by a decrease in the pH value (Kuang et al., 2013; Liu et al., 2020), and similar results were reported by Méndez-García et al. (2015). However, there was no decrease in the bacterial diversity of the silage in group M compared to group CK. This indicates that in the absence of exogenous LAB, the microflora did not change much in the early stage of ensiling when the WSC fermentation was carried out mainly. However, nonstructural carbohydrate fermentation was mainly carried out in the late stage, and significant changes in the microflora occurred. In summary, the richness of the microbial community in alfalfa silage was strongly influenced by the addition of exogenous LAB, whereas the diversity of silage microorganisms was strongly influenced by the addition of molasses or fermentation substrates.

Firmicutes and Proteobacteria are usually the dominant bacteria in alfalfa silage (McGarvey et al., 2013). Firmicutes are Gram-positive bacteria with a low G + C content in their genomes and degrade macromolecular compounds such as proteins, starch, and cellulose (Zeng et al., 2020). Proteobacteria are the largest bacterial phylum, including not only Erwiniaceae, Sphingomonas Paucimobilis, Methylobacterium, and Pseudomonas adaceae, but also pathogenic bacteria such as Escherichia Coli, Vibrio, and Spirillum (Santos et al., 2019). Proteobacteria play an important role in the degradation of organic matter and the carbon-nitrogen cycles during anaerobic digestion (Zhao et al., 2021). In this experiment, the structure of the bacterial community in alfalfa silage changed significantly with the fermentation time, the abundance of Firmicutes increased significantly and Proteobacteria decreased significantly in all groups from 14 days to 56 days of ensiling. After 56 days of ensiling, Firmicutes abundance was >99.72% in groups L, M, and LM, which was significantly higher than that in group CK and became the dominant phylum in groups L, M, and LM. This is related to the promotion of carbohydrate fermentation by the addition of Lactobacillus plantarum or molasses, and the dominant bacterial phyla of alfalfa silage with 3% molasses was reported to be Firmicutes, with an abundance of 95.58% (Luo et al., 2021). Molasses provides an additional fermentable substrate for LAB and promote the dominance of the silage bacterial community, thus shifting metabolism to LAB fermentation (Ni et al., 2017). Furthermore, the dominant bacteria in alfalfa silage supplemented with six different combinations of LAB were all Firmicutes with abundances ranging from 69.3% to 83.1%, followed by Cyanobacteria and Proteobacteria with abundances ranging from 3.9% to 26.4 and 0.4% to 14.6%, respectively (Wang et al., 2021). The results of the present study are consistent with the fact that a low pH value or anaerobic conditions during silage favor Firmicutes growth, which replaced Proteobacteria and were widespread after 60 days of silage (Keshri et al., 2018). The relative abundance of Firmicutes in LAB inoculated silage was higher at 15 and 30 days, but there was no difference after 60 days of ensiling (Zhao et al., 2021).

In this experiment, after 56 days of ensiling, the dominant bacteria in group M were Lactobacillus, Weissella, and Pediococcus, respectively, which all belong to Lactobacillaceae and predominantly use glucose, suggesting that the addition of molasses produced a promotional effect on them all. The relative abundance of Weissella was significantly higher than in other groups, indicating that Weissella was more sensitive to increasing glucose concentration, and, on the other hand, Weissella also had a great effect second only to LAB. This bacterium is visible in the digestive tract of animals and expresses β-glucosidase (Lee et al., 2012; Fusco et al., 2015), which facilitates the degradation of cellulose. In silage fermentation dominated by LAB, competing microorganisms will not be able to survive (Yang et al., 2019), and in this experiment, LAB had been promoted as the dominant genus in group L and LM, which inhibits the growth of other bacteria, and therefore explains a decrease in the abundance of the bacterial population. For example, the abundance of Lactobacillus in silage in groups L and LM was >97% on day 14 and decreased to 89% at 56 days. However, the abundance of Lactobacillus increased from 26.4% and 55.94% to 39.06% and 59.05% from 14 to 56 days in the group CK and M. LAB are the main producers of LA and play an important role in lowering the pH value and inhibiting mold growth in silage (Zhang et al., 2022). Conversion of LA to AA produces high concentrations of AA, which also lowers pH value and inhibits the growth of spoilage microorganisms (Tabacco et al., 2011). In this experiment, the highest LA concentration was found in group M, and the highest AA concentration was found in group L after 14 days and 56 days of ensiling, but among all groups of alfalfa silage, the best aerobic stability was found in group L. Although the combination of added Lactobacillus plantarum and molasses further optimized fermentation quality and nutritional value, Lactobacillus plantarum alone still provided excellent aerobic stability, which is consistent with the findings of Mu et al. (2021).

In general, metabolites produced by silage fermentation can affect the bacterial community and thus the quality of silage (Dong et al., 2020; Li P. et al., 2022). As shown in Figure 5, after 14 days of ensiling, the pH value and WSC content are significantly positively correlated with Lactococcus, Enterobacter, Enterococcus, and Leuconostoc (P < 0.01). Lactobacillus consumed a large amount of WSC to produce LA and decreased the pH value, which led to a very significant negative correlation between WSC and Lactobacillus (P < 0.001). And it is consistent with the results of Li et al. (2020). Low pH can inhibit the growth of Enterobacter and other harmful microorganisms during ensiling (Zhang et al., 2009; Yang et al., 2022). KEGG (Kyoto Encyclopedia of Genes and Genomes) has been used to explain the role of cells and organisms from a genomic perspective. During ensiling, various types of bacteria convert fermentable substrates to different metabolites through different metabolic pathways. Therefore, Tax4Fun was used to analyze the metabolic pathways of alfalfa silage in sandy grasslands based on KEGG. There were no significant differences in the metabolism of all groups at three levels of pathways after 14 days and 56 days of ensiling, indicating that the metabolism of the bacterial community stabilized after 14 days of ensiling. However, the carbohydrate metabolism and membrane transport in group M were numerically higher than in other groups after 14 days and 56 days of ensiling (Figures 6C, D), suggesting that adding molasses in silage may have the potential to enhances carbohydrate metabolism and membrane transport metabolism. This also explains the mechanism of LAB degrading WSC to produce LA and reduce the pH of silage.