The DWW was taken from the Jinsha Huisha Distillery in Guizhou Province, China. The two types of DWW were yellow serofluid and distillery spent wash. Yellow serofluid is a yellow-brown pouring slurry formed at the bottom of the cellar during the fermentation process of Sorghum bicolor raw materials, and the distillery spent wash is a mixed slurry left at the bottom of the pot during the distillation procedure. The pH values of the yellow serofluid and spent wash were 4.45 and 3.94, respectively, both of which contained a large amount of Lactobacillus and Weissella that were beneficial for silage fermentation. Sorghum propinquum was taken from Tongzi County, Guizhou Province, southwestern China (28° 7’ N, 106° 34’ E), with an altitude of 1,219.15 meters on July 21, 2022. The growth period was the heading period. Fresh crops were chopped to approximately 2–3 cm with a laboratory-type chopper, ensiled and vacuumized in duplicate polyethylene bags (30 × 40 cm). Fill each polyethylene bag with approximately 0.6 kilograms of fresh chopped forage and store it at ambient temperature (<25°C). The silages were used to test aerobic stability after 60 days of ensiling. Fresh forages and silages during the micro-permeation of air were sampled for assay of chemical composition, fermentation profile, and microbial community (three repetitions per treatment). The following treatments were applied: (1) no additives (CK); (2) commercial LAB inoculant (LAB, L. buchneri, 10⁶ cfu/g FM, supplied by Gaofuji Biotechnology Co., Ltd., Chengdu, China); (3) yellow serofluid additive (Y, treated at a rate of 4%); and (4) distillery spent wash additive (S, treated at a rate of 4%). All treatment groups of grass were completely mixed with additives.

Chemical analysis was determined in triplicate. The dry matter (DM) content of the fresh forages and silages after enzyme deactivation was measured by oven drying at 65°C for 48 hours. Crude protein (CP) was determined by the Dumas Nitrogen analyzer (model: N Pro) (Wiles et al., 1998). Determination of neutral detergent fiber (NDF) and acid detergent fiber (ADF) by a modified procedure using sulfite and amylase (Van Soest et al., 1991). The content of water-soluble carbohydrate (WSC) was measured by the anthrone reaction-rate method (Murphy, 1958). NH3-N was measured in the silages by the extraction of 20 g of frozen samples with 180 mL distilled water for 24 hours in cold storage (Weatherburn, 1967). The silage juice for the determination of pH value by desktop pH meter. Determination of lactic acid (LA), acetic acid (AA), propionic acid (PA) and butyric acid (BA) contents by Ultra-High Performance Liquid Chromatography (model: Vanquish Core) (Muck and Dickerson, 1988).

The ensiled forages with four treatments were mingled and put (unpackaged but compacted) in unpolluted buckets, made the air slightly infiltrate, and a thermocouple wire was slotted in the center of the silage stack. The thermocouple wires were bonded to a data logger, which logged the temperature every 10 min and then averaged the temperature every half hour. Corruption begins when the temperature in silage masses rises to 2°C higher than the ambient temperature.

Factorial analysis of variance was used to evaluate the effects of the micro-permeation time (D), treatment (T), and their interactions (D × T) on the silage parameters in the General Line Model of SPSS (SPSS 27.0 program SPSS Inc.). Differences were only considered significant when the probability level was below than 0.05 (P < 0.05).

Table 1 shows the chemical composition and dominant bacterial genera of fresh S. propinquum and DWW prior to ensilage. The DM of S. propinquum was 31.46%, and the contents of NDF and ADF were 62.35% DM and 29.57% DM, separately. The content of CP in S. propinquum was 9.75% DM, and the content of WSC was 9.56% DM. Mostly, the WSC content of the fresh forage serves as a considerable supplement for anaerobic fermentation, and that a forage with over 5% DM WSC content tends to excellent fermentation performance (Cai et al., 1998). Therefore, the content of WSC in fresh S. propinquum was adequate for silage fermentation.

The contents of Lactobacillus in S. propinquum, yellow serofluid and distillery spent wash were 8.33, 7.42 and 31.26%, and the contents of Weissella were 48.05, 11.96 and 63.64% respectively. The Chloroplast and Mitochondria contents in S. propinquum were 35.62% and 6.55% respectively. Moreover, yellow serofluid contained a large source of Steroidobacter (35.24%).

The LAB- and DWW-treated S. propinquum silage had a higher DM content than CK after 60 days of ensiling ( Table 2 ). The content of CP differed significantly among different treatments (P < 0.01). Y- and S-treated S. propinquum silage had increased CP content by 1.2 and 0.11%, respectively, but no additives and LAB inoculated S. propinquum silage had reduced it by 1.12 and 0.43%, respectively, after 60 days of anaerobic fermentation. This may be connected with the possession of considerable Lactobacillus and Weissella in the DWW. In this study, the S. propinquum silage with different treatments expressed no significant difference in WSC content. Moreover, fermented S. propinquum had further contents of NDF and ADF than fresh S. propinquum after 60 days of anaerobic storage, which perhaps due to the depletion of accessible nutrients generating a multiplication in NDF and ADF contents during the anaerobic fermentation (Li et al., 2022). Thus, we considered that the phenomenon discovered in this study probably owing to the treated LAB and DWW impacting on anaerobic fermentation, which reduces the loss of dry matter content in S. propinquum silage during the anaerobic fermentation.

During the micro-permeation of air, the time of all silages with different treatments at 2°C above the ambient temperature was equivalent (>192 h). LAB and DWW played important roles in improving the fermentation characteristics of S. propinquum silage after 60 days of ensiling and during the micro-permeation of air, as shown in Table 3 . All silages with different treatments reached a relatively low pH value (< 4.70) after 60 days of fermentation. In this study, the various S. propinquum silage with different treatments expressed no significant difference in the pH value. Studies have verified that inoculation with Lactobacillus buchneri or Bacillus subtilis either improved or didn’t impact silage pH (Kung et al., 2003; Lara et al., 2016). The pH values of the various S. propinquum silage with different inoculants gradually increased, but they were all significantly below 5.0 during the micro-permeation of air. This is because the silage is opened in the air, the invasion of oxygen will activate the inhibited aerobic microorganisms, increasing the pH value. These aerobic microorganisms will affect the aerobic stability of silage (Tennant et al, 2017). The DWW and LAB additives influenced the DM parameter (P > 0.05), and the time of air micro-permeation also affected the parameter (P < 0.01). LAB- and DWW-treated S. propinquum silage had higher DM content than CK after 60 days of fermentation. LAB- and Y-treated S. propinquum silage and untreated S. propinquum silage on the sixth day of micro-permeation of air had significantly higher DM contents than those at silage opening after 60 days of ensiling. However, the content of DM in S-treated S. propinquum silage was reduced during the micro-permeation of air. The micro-exposure of silage to air can begin the aerobic spoilage process, resulting in pH increases and DM losses (Woolford, 1990). However, in this experiment, the various S. propinquum silage did not lead to DM losses. Studies have indicated that the DM content of S. propinquum silage increased during the aerobic exposure stage (de Oliveira et al., 2009), which is consistent with our research. As a byproduct of protein decomposition, the content of ammonia nitrogen is influenced by the activity of Clostridium and plant protease. In this study, the CK treatment was significantly higher than the LAB- and DWW-treated S. propinquum silage after 60 days of anaerobic fermentation. This change is due to the inhibition of protein-degrading microorganisms by the addition of LAB and DWW (Bai et al., 2020). However, with increasing micro-permeation time, the LAB- and DWW-treated treatments showed an increasing trend. This may be because when silage is exposed to air, aerobic microorganisms begin to multiply, and amino acids are decomposed into ammonia, hydrogen sulfide, and amines, increasing the degradation rate and leading to an increase in ammonium nitrogen content (Hashemzadeh-Cigari et al., 2014). Surprisingly, the ammonium nitrogen treated with CK showed a decreasing trend. As is well known, ammonium nitrogen is easily volatilized when exposed to air, so we believe that the factor reducing the CK treatment comes from the volatilization of ammonium nitrogen. Although aerobic microorganisms produce ammonia nitrogen during the micro-permeation of air, this production is minimal.

The Y-treated S. propinquum silage had a higher lactic acid content than CK, which increased by 0.34% after 60 days of anaerobic fermentation, improving the fermentation profiles of silage. Y-, S- and LAB-treated Sorghum propinquum silage had higher lactic acid content than CK after 6 d of air micro-permeation, increasing by 0.73, 0.31 and 0.39%, respectively, which improves the air micro-permeation of silage. During the micro-permeation of air, all treatments showed a decrease in lactic acid, while acetic acid and propionic acid showed an increase. The reason may be that lactic acid is easily utilized by yeast during the micro-permeation of air, leading to secondary fermentation of silage. In addition, the heterotypic fermentation lactic acid bacteria attached to the silages themselves decompose lactic acid to produce acetic acid, which to some extent maintains a low pH level in S. propinquum silage (Romero et al., 2017). As acids with antibacterial properties, the increase in acetic acid and propionic acid content indicates an increase in aerobic stability of silage (Driehuis et al., 1999). In this experiment, after 6 days of air micro-permeation, the addition of DWW seemed to have the same effect as LAB, with similar levels of acetic acid and propionic acid (P > 0.05). This indicates that DWW seems to be a potential silage additive that is equivalent to L. buchneri in improving the aerobic stability of silage. The presence of butyric acid during the micro-permeation of air is common, due to the reproduction of butyric acid-producing bacteria caused by the micro-permeation of air. In summary, we believe that DWW has the same additive effect as L. buchneri as a silage additive. However, the underlying mechanism still needs further exploration.

To further explore the effect of DWW as an additive on microorganisms in anaerobic fermentation of S. propinquum silage, we analyzed the alpha diversity of S. propinquum silage during the micro-permeation of air in Table 4 . We found that there was no significant interaction between days and treatment in the different treatments of DWW additives (S and Y), LAB, and CK. A significant effect was found only in the Pielou E index (P < 0.05). The Pielou E index represents the difference in quantity between different species in S. propinquum silage. As micro-permeation of air progressed, the Pielou E index values of LAB- and DWW-treated S. propinquum silage were higher than those of the CK treatment. This may be due to the addition of DWW and LAB causing an increase in some species, leading to an increase in diversity between species. In previous studies, it was also found that the addition of LAB resulted in an increase in species during the aerobic exposure (Bai et al., 2020). In addition, DWW does contain some acid-resistant microorganisms that are inhibited during the anaerobic fermentation, but this state is relieved when exposed to oxygen (micro-permeation of air) (Tripathi et al., 2022). At the beginning of air micro-permeation, the Pielou-e index values of the S and Y treatments were lower than that of LAB, confirming the abovementioned mechanism.

The diversity component analysis we used was principal coordinate analysis (PCoA), which extracted the most significant elements and constructions from multidimensional data through a battery of feature values and feature vector sorting. We used the Bray−Curtis distance for PCoA, and the principal coordinate combination with the highest contribution rate was selected for graphical revealing, as shown in Figure 1 . If the sample distance was relatively close, it indicated that the species composition structure was similar. Hence, samples with extraordinary similarity in community structure tended to cluster together, while samples with significant community differences were farther apart. In this experiment, the distance between samples of the same treatment was very close, illustrating that the bacterial community composition structure was very similar and the excellent repeatability of this experiment. For different treated samples, the distance between them was relatively large, and the community composition varied greatly, which was due to the different additives used in our different treatments. From the figure, we can see that there are some distinctions in the bacterial community composition of different treatments, and there are also some distinctions in the bacterial community composition of each treatment when it is just micro-exposed to the air compared with the later period of air micro-permeation. This is because when silage is exposed to air, aerobic microorganisms begin to multiply (Hashemzadeh-Cigari et al., 2014).

Network cooccurrence at the genus level of S. propinquum silage treated with different additives is shown in Figure 2 . We calculated the Spearman correlation index of all samples to obtain the species correlation coefficient matrix and set the filtering conditions as follows: (1) connections with correlation coefficients <0.6 were removed, (2) node self-connections were filtered out, and (3) connections with node abundance less than 0.005% were removed to obtain the network graph. The dominant genera of bacteria in all treatment groups were Lactobacillus and Weissella, which had a negative correlation. However, the genus-level structure of S. propinquum silage varied among different additive treatments. We found that Y-, S- and LAB-treated S. propinquum silage increased the negative correlation coefficient between Lactobacillus and other adverse bacteria and enhanced the negative associations within bacteria but reduced the negative correlation coefficient between Lactobacillus and Weissella.

This study ascertained phylum-, genus-, and species-level alterations in the microbial composition of S. propinquum silage and distinctions in bacterial taxa ( Figure 3 ). The phylum-level bacterial composition of S. propinquum silage is displayed in Figure 3A . Compared with fresh forages, an increase in Firmicutes accompanied by a decrease in Proteobacteria became the only dominant phylum. During the fermentation period of silage, Proteobacteria gradually replaced Firmicutes. This is because Proteobacteria is the maximum phylum-level bacteria, commonly found in various raw silage materials (Du et al., 2021), and Firmicutes contains various lactic acid bacteria that are the principal participants in silage storage (Li et al., 2021). With increasing air micro-permeation time, Firmicutes decreased and Proteobacteria increased in the CK treatment group, but Firmicutes did not decrease in S. propinquum silage treated with LAB and DWW, which is owing to the inferior inhibition of aerobic micro-organisms in the CK treatments, resulting in a speedy growth in other phyla, which is consistent with the conclusion of Liu et al. (2022). The primary bacteria existed in the lactic acid fermentation of silage pertain to the genera Weissella, Lactobacillus, Leuconostoc, and Pedicoccus (Ni et al., 2018). The genus-level changes are shown in Figure 3B . There are high abundances of Weissella, Chloroplast and some Lactobacillus in fresh S. propinquum. After 60 days of fermentation of silage, Lactobacillus in S. propinquum silage becomes the dominant genus and contains many Weissella. There was no significant difference between the different treatments. During the micro-permeation of air, Lactobacillus and Weissella had almost no change and remained at a relatively high level, which also showed that the S. propinquum silage had good aerobic stability and did not start to decay after six days of air micro-permeation. In this experiment, we found that there are unique genera Roseburia and Faecalibacterium in the S. propinquum silage treated with DWW. The S. propinquum silage treated with Y contained a small amount of Roseburia compared with other treatments, and the S. propinquum silage treated with Y and S contained a small amount of Faecalibacterium compared with CK and LAB treatments. Roseburia and Faecalibacterium are bacteria that are beneficial for animal health. Dietary fibers were metabolized by Roseburia intestinalis and Faecalibacterium prausnitzii as primary short-chain fatty acid producers, supplying energy sources for enterocytes and attaining anti-inflammatory functions in the gut (Hiippala et al., 2018). Zhang et al. (2019) also confirmed stress-related microbial species and further tested the beneficial function of Roseburia, a butyrate-producing genus that was dramatically reduced in rats, on stress-induced visceral hypersensitivity. A study showed that colonization with Roseburia intestinalis reduces the levels of inflammation markers and atherosclerosis in a diet-dependent pattern and verified that the final product (butyrate) of bacterial metabolism guides these functions (Kasahara et al., 2018). There are findings that highlight that R. intestinalis is a main degrader of some dietary fiber and that this metabolic capacity could be utilized to selectively accelerate crucial members of the healthy microbiota using β-mannan-based therapeutic interventions (La Rosa et al., 2019). According to reports, in murine models, F. prausnitzii cells or their cell-free epipelagic can alleviate the severity of acute (Sokol et al., 2008), slow acting (Martín et al., 2014) and low-grade (Martín et al., 2015) chemical-induced inflammation. Therefore, when livestock are fed silage containing Roseburia and Faecalibacterium, it will have a good effect on the prevention of intestinal inflammation, which is no different from using DWW as a silage additive.

After 60 days of silage, the abundances of Lactobacillus coryniformis, Lactobacillus bifermentans, Lactobacillus acetotolerans and Lactobacillus camelliae in S. propinquum significantly increased, as shown in Figure 3C . Y-treated S. propinquum silage had higher Lactobacillus coryniformis content than the CK treatment, which increased by 9.51% ( Figure 3D ), and S-treated S. propinquum silage had higher Lactobacillus bifermentans and Lactobacillus acetotolerans content than the CK treatment, which increased by 8.27 and 9.75% ( Figures 3E, F ), respectively, after 60 days of anaerobic fermentation. During the micro-permeation of air, the content of Lactobacillus coryniformis in all treatments significantly decreased, but Y- and S-treated S. propinquum silage had higher Lactobacillus coryniformis content than the CK treatment from start to finish (increased by 1.57 and 2.93%). The contents of Lactobacillus bifermentans and Lactobacillus acetotolerans in the S treatment significantly decreased, but S-treated S. propinquum silage consistently had higher Lactobacillus bifermentans and Lactobacillus acetotolerans contents (increased by 8.26 and 9.75%) than the CK treatment silage. Therefore, we believe that DWW as a silage additive enhances the aerobic stability of S. propinquum.

A Spearman correlation between fermentation parameters and kinetics of the top 10 genera and species during the ensiling of S. propinquum was generated and illustrated in Figure 4 . Lactobacillus played an important role in improving the lactic acid content and diminishing the pH value in subsequent anaerobic storage (Cai et al., 1998). Hence, our study demonstrated that lactic acid was positively correlated with certain Lactobacillus genera (L. coryniformis, L. camelliae, and L. songhuajiangensis; P < 0.05). Y-treated S. propinquum silage had a higher Lactobacillus coryniformis content than the CK treatment, which led to a higher capacity for acid production and anaerobic storage. We also found that ammonia-N was positively connected with the Roseburia and Faecalibacterium genera, which explains why the increase in ammonia-N in DWW-treated S. propinquum silage was higher than that in the CK treatment during the micro-permeation of air.