The hybrid elephant grass [(Pennisetum americanum × Pennisetum purpureum) × Pennisetum durpureum schum. cv. Guimu No. 1] was harvested from the Chongzhou experiment farm of Sichuan Agricultural University (N30°33′23.98″ E103°38′42.61″) on August 26, 2020, at which time the hybrid elephant grass was at nutritional growth period (1.8–2.0 m). These materials were chopped to 2–3 cm for the preparation of silage. The silage additives include heat-resistant Lactobacillus plantarum 149 (LP149), commercial inoculant Lactobacillus plantarum (LP, Sichuan Gaofuji Biotech Co., Ltd, Chengdu, China), and Acremonium cellulase (AC, Green Stone Swiss Co., Ltd., Shanghai, China). Strain LP149 was screened previously from silages in Southwest China as a humid and hot area, which has good growth characteristics that tolerate the low pH (3.5) and high temperature (45°C). The gene sequence of this strain has been uploaded to GenBank with the registration number MH 337263. LP and AC were supplied as lyophilized powders, where LP has a viable count of 50 billion colony-forming units (cfu)/g and AC has an enzymatic activity of over 1,000 U/g.

The chopped hybrid elephant grass was mixed thoroughly with three additives separately, packed into polyethylene plastic bags (200 × 300 mm, Shenzhen Sanfeng Plastic Packing Co., Ltd., Shenzhen, China, each bag weighed 300 ± 2 g) and sealed with vacuum sealer (DJVac, Wenzhou, China). AC was applied at a rate of 0.03% FW. LP and LP149 were inoculated at a rate of 5 × 10⁶ cfu/g FW. The appropriate amounts of AC and LP lyophilized powder were filled into the sterilized small spray and diluted with 10 mL of sterile water. An equal volume of sterile water was added to the hybrid elephant grass as a control group. Strain LP149 requires activation in MRS (De Man, Rogosa, and Sharpe) broth medium before inoculation and detection of viable bacteria by plate count method (Xie et al., 2022). In this study, a total of 120 bags of silage samples were prepared (4 treatments × 2 ensiling temperatures × 5 ensiling times × 3 replicates). The plastic bag silos were stored in incubators at 30 or 40°C for 1, 3, 7, 14, and 60 days of ensiling.

The plastic bag silos were opened at the designed ensiling time to analyze the fermentation and chemical characteristics. A 20 g sample of hybrid elephant grass silage was placed in a juicer, then 180 mL of distilled water was added and blended at high speed for 45 s, and the residue was removed through 4 layers of gauze. The obtained filtrate was poured into a clean conical flask, and the pH value was measured immediately. A volume of 50 mL of the filtrate, preserved in a freezer at −20°C, was used for subsequent analysis. The ammonia nitrogen (NH₃-N) content was analyzed by the phenol-hypochlorite method (Li et al., 2021). The organic acids were detected using a high-performance liquid chromatograph according to the procedure of Yan et al. (2019).

About 100 g of fresh hybrid elephant grass and silage samples were placed in a blast dryer at 65°C and dried for 3 days to determine the dry matter (DM) content. An appropriate amount of dried samples was crushed, sieved (1 mm), and stored in a desiccator for chemical composition analysis. The water-soluble carbohydrate (WSC) was detected using the anthrone-sulfuric acid method (AOAC, 1990). The nitrogen was analyzed by the Dumas combustion method (Rapid N exceed, Elementar, Germany), and a protein conversion coefficient (6.25) was used to calculate crude protein. The neutral detergent fiber (NDF), acid detergent fiber (ADF), and acid detergent lignin (ADL) were determined according to the method of Van Soest et al. (1991). Hemicellulose and cellulose contents were estimated from the differences between NDF and ADF, ADF and ADL contents, respectively (Li et al., 2019). The sucrose, glucose, and fructose contents were detected with a kit (Hexokinase method, G0545W, Suzhou Grace Biotechnology Co., Ltd, Suzhou, China).

One-way ANOVA and multi-way ANOVA were performed using SPSS (version 24.0) for the additives, ensiling time, and temperature. Duncan's new multiple-range test was used to compare the mean values of different chemical compositions and fermentation characteristics. Plate count results of microorganisms need to be log-transformed before statistical analysis. Differences were regarded as statistically significant when P < 0.05.

The initial DM content of fresh hybrid elephant grass was 172.62 g/kg DM. It is difficult to reduce the moisture content of fresh grass by wilting in humid and hot areas. High-moisture grass easily leads to undesirable microbial fermentation, nutrient loss, and aerobic spoilage during ensiling. In addition, the WSC, CP, NDF, ADF, and ADL contents of fresh hybrid elephant grass were 76.51, 103.82, 584.16, 336.28, and 33.73 g/kg DM, respectively. The number of LAB, coliform bacteria, yeast, and mold was 2.03, 4.00, 2.49, and 1.75 log₁₀ cfu/g FM, respectively. It is generally believed that the number of LAB ≥ 10⁵ cfu/g FM is more preferable for the preservation of silage (Ni et al., 2017). Therefore, exogenous addition is a key measure to improve the fermentation quality of high-moisture hybrid elephant grass in humid and hot areas.

The fermentation characteristics of hybrid elephant grass silages treated with additives at different ensiling times and temperatures are presented in Table 1. Compared to the control, three additive groups had a lower pH and were ≤ 4.2 after 3 days of ensiling, which is the threshold value indicating that the silages were well preserved (Mu et al., 2020). In addition, the pH decline rate after 1 day of ensiling at two temperatures was LP > LP149 > AC. The pH decline rate is regarded as a more important indicator for reflecting fermentation dynamics than the final pH value (Mu et al., 2020). However, it does not mean that lower pH is better, as the growth of LAB is suppressed when the pH ≤ 3.8 (Muck, 2010). The pH of AC and LP treatments decreased below 3.8 after 7 days of ensiling at 30°C. The accumulation of organic acids is the main reason for the pH decrease during anaerobic fermentation (Wang et al., 2021). Thus, AC and LP treatments may have facilitated the production of organic acids. Interestingly, the pH of all silages slightly increased after 60 days of ensiling at 45°C. With the soluble sugars depleted in the late stage of high-temperature fermentation, some acid-tolerant undesirable microorganisms such as yeasts take organic acids as a carbon source to proliferate, leading to a rise in pH.

Lactic acid (LA) is the most powerful organic acid capable of rapidly decreasing pH (Ali et al., 2020). Compared to the control, the silages treated with additives increased the LA content. The LP treatment had a higher (P < 0.05) LA content than the other treatments in the first 14 days of ensiling, but it decreased to different degrees at two temperatures after 60 days of ensiling. It is reported that some undesirable microorganisms can use LA as a substrate for fermentation when the soluble carbohydrate is low in late fermentation (Oliveira et al., 2017). The AC and LP149 treatments had higher LA content than the control during the whole fermentation process and were higher (P < 0.05) than the LP treatments after 60 days of ensiling at 30°C. These suggest that the addition of AC or strain LP149 may be more preferable than LP for the long-term preservation of silage.

The appropriate amount of acetic acid (AA) can enhance the aerobic stability of the silage to a certain extent (Kleinschmit and Kung, 2006). Throughout the fermentation process, AC and LP149 treatments had a higher (P < 0.05) AA content than the control and LP treatments. AC treatment can lead to the hydrolysis of lignocellulose to pentose, thus promoting the production of AA (Li et al., 2018b). The strain LP149 may boost the production of metabolites with bacteriostatic activity during ensiling. The LA/AA value is usually considered a qualitative indicator of ensiling, with a good fermentation ratio of about 2.5–3.0 (Guan et al., 2020). After 60 days of ensiling, the LA/AA of LP149 treatment at two temperatures (5.33 and 2.52) was closest to the ideal ratio, followed by AC treatment (5.55 and 2.45), while the LP treatment had a higher ratio (8.57 and 4.99). This may indicate that AC and LP149 are more preferable than LP for enhancing aerobic stability of silage in humid and hot areas. Throughout the ensiling process, a slight amount of propionic acid and butyric acid was detected (< 3.5 g/kg DM). It was worth noting that butyric acid content was lower (P < 0.05) in AC and LP149 treatments than that in the control and LP treatments.

A high concentration of NH₃-N is the result of excessive protein degradation. In this study, the NH₃-N concentration in all silages (< 75 g/kg TN) was consistent with the criteria for good quality silage (< 100 g/kg TN) (Mu et al., 2020). After 14 days of ensiling, the three additive groups had lower (P < 0.05) NH₃-N levels than the control. This might be because the additive treatments rapidly lowered the pH, which inhibits the undesirable bacteria from degrading the protein. Moreover, the silages stored at 45°C had higher NH₃-N levels than those stored at 30°C. It was reported that high temperature usually leads to butyric acid fermentation and more protein breakdown during ensiling (Chen et al., 2013).

The lignocellulose compositions (including NDF, ADF, ADL, cellulose, and hemicellulose) of hybrid elephant grass silages treated with additives at different ensiling times and temperatures are presented in Table 2. The AC treatment had lower (P < 0.05) contents of NDF, ADF, and cellulose than the other treatments after 7 days of ensiling. This reflected the powerful potential of AC to the degradation of lignocellulose during ensiling in humid and hot areas. Moreover, the LP149 treatment had lower (P < 0.05) contents of NDF, ADF, and cellulose than the LP treatment after 14 days of ensiling. This may prove that strain LP149 has better heat resistance and cellulolytic activity than LP.

After 3 days of ensiling, the LP and LP149 silages stored at 45°C had a higher content of NDF, ADF, and cellulose than those stored at 30°C. Generally, high temperatures can result in the deposition of lignocellulose in silage (Wilson et al., 1991). The Maillard reaction is promoted when the ensiling temperature is higher than 35–40°C, and the polymer produced by this reaction increases the lignocellulose in silage (Muck et al., 2003). However, the control and AC silages stored at 45°C had a lower content of NDF, ADF, and cellulose than those stored at 30°C. This is probably due to the high temperature promoting the growth of epiphytic bacteria with cellulolytic activity, and the optimum temperature of AC is 50–60°C (Li et al., 2018a).

The variations in the contents of fermentable carbohydrates (WSC, glucose, fructose, and sucrose) of hybrid elephant grass silages treated with additives at different ensiling times and temperatures are shown in Figure 1. Compared to the control, the contents of fermentable carbohydrates in three additive groups were higher (except for the LP treatment at 45°C). Throughout the fermentation process, apart from the glucose content of the AC treatment that appeared an increasing trend in the first 7 days of ensiling, other fermentable carbohydrates decreased gradually with the extension of ensiling time, which decreased sharply in the first 7 or 14 days of ensiling and then decreased slowly. This is closely related to the anaerobic fermentation process, where a large amount of fermentable sugar is consumed during the aerobic respiration period and the intensive fermentation phase, but the demand for substrates decreases when fermentation enters a stable phase (Ávila and Carvalho, 2020). In addition, the increased glucose content of the AC treatment in the first 7 days of ensiling may be attributed to the accumulation of glucose as a result of AC promoting the degradation of lignocellulose to produce more glucose than that consumed by microbial fermentation (Li et al., 2019). All fermentable carbohydrates of LP149 treatment were higher than LP treatment after 7 days of ensiling (except for sucrose at 30°C). This may indicate that strain LP149 was more beneficial than LP in promoting the degradation of lignocellulose during ensiling of hybrid elephant grass.

Correlation analysis of the compositions of lignocellulose and fermentable carbohydrates was performed to explore the linkage between them (Figure 2). Overall, lignocellulose was negatively correlated with fermentable carbohydrates where NDF, ADF, and cellulose were negatively (P < 0.01) correlated with WSC, glucose, and fructose. Taking into account the variation of their contents, this may demonstrate that lignocellulose is mainly degraded to WSC, glucose, and fructose during ensiling, whereas WSC mainly consists of glucose, fructose, and sucrose (Usman et al., 2022).

The Goods coverage of all silages was >99%, providing a possibility of microbial communities analysis (Table 3). In conventional silage, LAB inoculation could reduce alpha diversity (Bai et al., 2021; Liu et al., 2022). But in this study, three additive groups did not have the desired effect to reduce alpha diversity, and even alpha diversity was higher than the control. This is probably due to that the humid and hot environments promoted the growth of undesirable microorganisms, and the additive treatments indirectly provided them with fermentation substrates (Zhao et al., 2019).

PCoA reflected the similarity or dissimilarity of bacterial community composition during ensiling (Figure 3). Based on PCoA at different ensiling times (Figures 3A–E), bacterial communities were clearly separated among control and treated groups after 1, 3, 7, 14, and 60 days of ensiling. Apart from the 60th day of ensiling, the bacterial communities were clearly separated among the treated groups, and the bacterial communities of LP149 treatment were clearly separated from the other groups. Moreover, bacterial communities were clearly separated in silages stored at 30°C and those stored at 45°C. These suggest that temperature and additives are important factors affecting the bacterial communities of silage. Compared to additives, the temperature had a more profound effect on the bacterial community (Guan et al., 2020).

The number of LAB is a key factor to determine the quality of silage. LAB can inhibit undesirable fermentation by rapidly lowering the pH (Chen et al., 2021). In this study, the number of LAB in all silages increased sharply compared with fresh hybrid elephant grass after 1 day of ensiling (see Supplementary Table). The LP149 treatment had a higher (P < 0.05) number of LAB than the other treatments throughout the fermentation process at 30°C.

The bacterial community at the phylum and genus levels is presented in Figure 4. The main phyla in hybrid elephant grass silage were Firmicutes and Proteobacteria, and they are the most common phyla in silage. The main dominant bacteria in hybrid elephant grass silage were Lactobacillus, Weissella, and Klebsiella. In addition, the main bacteria involved in LA fermentation during ensiling were Lactobacillus, Weissella, and Pediococcus. The abundance of Lactobacillus in LP and LP149 treatments was higher than that in the control and AC treatments after 7 days of ensiling. This indicated that the addition of strains LP149 and LP could enhance the dominance of Lactobacillus. It is noteworthy that the AC treatment had a lower abundance of Lactobacillus than the control after 3 days of ensiling. This might be because some undesirable microorganisms have a growth advantage over epiphytic LAB under humid and hot environments (Pahlow et al., 2003), and AC treatment provides more fermentation substrate for them by degrading part of the lignocellulose into soluble sugars. Weissella might be an important bacterium for promoting LA fermentation in AC treatment. Generally, Weisseria will gradually decrease as prolonged fermentation time while acid-tolerant bacteria will increase (Ogunade et al., 2018). Bacillus has good acid tolerance and is found often in the late fermentation (Bai et al., 2020). After 60 days of ensiling at 45°C, Bacillus became one of the dominant bacteria. Particularly in control silage, Bacillus was the main dominant bacterium. It has been reported that Bacillus has the potential to improve the quality of silage (Bai et al., 2022), but the fermentation quality of the control silage was not good. Therefore, the role of Bacillus in silage in humid and hot areas needs further investigation.

Klebsiella, Acinetobacter, and Raoultella are the main undesirable bacteria in silage. These bacteria are usually considered undesirable microorganisms that compete with LAB for fermentation substrates (Du et al., 2022). Furthermore, Klebsiella and Acinetobacter can lead to reduced aerobic stability of silage (Lin et al., 2021; Muraro et al., 2021), whereas it was reported that Klebsiella, Enterobacter, and Raoultella are tolerant to low pH and can inhibit the pathogen (Gheibipour et al., 2022). Thus, it is necessary to carry out the in-depth study of bacteria normally considered undesirable during ensiling.

Bacterial systems have received much attention in the degradation of lignocellulose because of their outstanding functional diversity and adaptability (Georgiadou et al., 2021). The relationship among dominant bacteria with lignocellulose and fermentable carbohydrates of silage in humid and hot areas was investigated by Spearman's correlation analysis (Figure 5). The compositions of lignocellulose and fermentable carbohydrates were negatively (P < 0.05) correlated with Bacillus, Anoxybacillus, and Curvibacter. This demonstrated that microorganisms capable of promoting lignocellulose degradation may consume more fermentable carbohydrates during their action. Ahmed et al. (2018) reported that many members of Bacillus have ligninolytic and/or cellulolytic activities. In addition, Acinetobacter was negatively (P < 0.01) correlated with NDF, ADF, and cellulose. This is owing to that many species of Acinetobacter can promote the efficiency of carbohydrate catabolism and metabolism (Hošková et al., 2015). It is notable that the compositions of lignocellulose and fermentable carbohydrates were positively (P < 0.05) correlated with Lactobacillus and Weissella. This might be due to that these two bacteria rapidly lower the pH by involving in LA fermentation, which inhibits the consumption of soluble sugars by undesirable bacteria. The potential of LP149 to degrade lignocellulose may be due to that it directly or indirectly promoted the growth of bacteria with cellulolytic activity or as a result of its synergistic interactions with other bacteria.