Effects of Different Cutting Stages and Additives on the Fermentation Quality and Microbial Community of Sudangrass ( Sorghum sudanense Stapf.) Silages

: (1) Background: Previous studies have indicated that ferulic acid esterase (FAE), cellulase and xylanase have synergistic effects in lignocellulose degradation, and the cutting stage has a major impact on silages. Whether these additives affect the silages at different cutting stages is unclear. (2) Methods: Sudangrass height at the tested cutting stages was 1.8 m (S1) and 2.0 m (S2). The silage from the two cutting stages was treated with FAE-producing Lactobacillus plantarum (LP), cellulase and xylanase (CX) and a combination of LP and CX (LP+CX) for 30 and 60 days. (3) Results: Compared with CK, adding LP+CX signiﬁcantly decreased the pH and the content of neutral detergent ﬁber (NDF) and acidic detergent ﬁber (ADF) ( p < 0.05) and increased the lactic acid (LA) concentration ( p < 0.05), dry matter (DM) content and crude protein content. Adding LP+CX effectively degraded lignocellulose in sudangrass, and the NDF and ADF degradation rates at the two stages were all more than 30%. In comparison, cutting at the S2 stage led to a lower pH and higher LA and DM contents ( p < 0.05). Additives and the cutting stage exerted a strong effect on the silage microbial community, and Firmicutes and Lactiplantibacillus became the most dominant bacterial phyla and genera, especially at the S2 stage. (4) Conclusions: The results suggest that FAE-producing L. plantarum , cellulase and xylanase had synergistic effects on sudangrass silages, especially at the S2 stage, and their use can thus serve as an efﬁcient method for ensiling.


Introduction
Sudangrass (Sorghum sudanense Stapf.) is an important annual grass and is a C4 plant, with excellent characteristics such as a high yield, barren and drought tolerance, and strong resistance to diseases and insects [1,2]. It can also be used as forage grass for livestock due to its high nutritional value, good palatability and strong regeneration capability, which is better than most annual grasses, and it can be mowed three to four times a year [3][4][5]. Therefore, sudangrass is used as animal feed, for example, as silage, in many countries, including China [6,7]. Sudangrass silage can help overcome the challenges posed by insufficient feed supply in winter, especially in southwest China (e.g., Guizhou), where the climatic conditions make it difficult to produce high-quality hay [8]. However, few studies have focused on improving the nutritional value and quality of sudangrass silage, especially in different cutting stages and via treatment with different additives.
Ensiling is a microbe-driven process in which epiphytic lactic acid bacteria (LAB) convert water-soluble carbohydrates (WSCs) into organic acids (mainly lactic acid (LA)) under anaerobic conditions, which rapidly reduces the pH in the fermentation mixture and allows the long-term preservation of the nutritional value of forage grass [9,10]. Many factors can affect silage quality, such as the growth stage of the materials, epiphytic microbes, storage conditions and additives [11,12]. Moreover, researchers believe that the cutting stage has a major impact on the digestibility of silage [11]. Over the course of maturation, nitrogen compounds in forage grass leaves are replaced by fiber and lignin [13], thus reducing the digestibility of silage. Studies have reported that the growth stage affects the chemical composition (such as the WSC content) and epiphytic microbiota of the raw materials [14,15].
Tall grasses, such as sudangrass, have a relatively high fiber content, which could affect silage quality and digestibility, especially with delayed mowing [5,16]. Therefore, in many studies, cellulase and LAB have been added to silage, improving silage quality to a certain extent, but the degradation of lignocellulose has not been satisfactory [17][18][19]. The reason for this is that hemicellulose and lignin in plant cell walls are mainly bonded via ferulic acid ester bonds formed by phenolic acids (especially ferulic acid and ferulic acid dimers), forming a lignin-carbohydrate complex to wrap cellulose [20,21]. The complex physical barrier structure limits the accessibility of exogenous enzymes to structural polysaccharides. To improve the degradability and nutritional value of feed, researchers have utilized ferulic acid esterase (FAE) to destroy the ester bonds between lignin and structural polysaccharides [22,23]. For example, a previous study found that adding FAEproducing LAB and cellulase to corn stalk silages resulted in higher quality and a higher degradation rate [24]. However, the ester bonds in the plant cell wall link lignin and hemicellulose [20,21]. Whether the use of hemicellulase can achieve a better effect on silage is unclear. Previous research found that hemicellulases (xylanases and acetyl xylan esterases) and ferulic acid esterases have functional synergies [25]. A previous study also found that xylanase interacts with FAE and could achieve a better degradation rate [26], and in another study, the combination of FAE and xylanase improved the production of ferulic acid and the degradation of cereal and wheat bran [27,28]. Moreover, our other study showed similar results and obtained higher silage quality.
In summary, adding FAE-producing LAB, cellulase and xylanase to sudangrass silage at different cutting stages should have different effects on the silages. To verify this hypothesis, the silage quality, fiber degradation and microbial community were explored to evaluate the optimal cutting stages and additives. This study lays a research foundation for sudangrass silage production.

Experimental Materials and Additives
Raw sudangrass was harvested in Tongren, Guizhou, China, on 14 July 2022. The raw materials were collected from two sudangrass fields with a cultivation time difference of 20 days, but both samples were harvested at the vegetative stage and 52 and 72 days after planting. The cutting heights of the sudangrass samples were 1.8 m and 2.0 m according to a previous study [5]. The materials were immediately returned to the laboratory on the same day. The grass was wilted for approximately 5 h until the moisture content was approximately 70% by natural air-drying and then cut into fragments with a length of approximately 2 cm with a chopper. The added FAE-producing L. plantarum was screened from rumen fluid previously from a total of 110 LAB strains using Man Rogosa Sharpe (MRS) agar. Ethyl ferulate (1% w/v in dimethyl formamide) was added to the agar without glucose (pH 6.5) and incubated at 37 • C for 48 h. The LAB strains producing FAE were confirmed by visualization of a ring of clearance around the colonies on the plates. Cellulase and xylanase were purchased from Beijing Solebo Technology Co., Ltd., and their enzyme activities were 50 U/mg and 20 U/mg, respectively.

Experimental Methods and Silage Preparation
This experiment was performed with a completely randomized design: 2 growth stages (S1: 1.8 m; S2: 2 m) × 4 additive treatments (control (CK); mixture of cellulase and xylanase (CX); FAE-producing L. plantarum (LP); cellulase, xylanase and FAE-producing L. plantarum (CX + LP)) × 2 fermentation times (30 and 60 days). The amount of LP added was 1 × 10 6 cfu/g fresh matter (FM), and the amount of both cellulase and xylanase added was 25 U/g FM, according to previous research [18,19]. The prepared LP solution and enzyme solution were placed into sterile and enzyme-free mini sprayers, respectively. Then, the solution was sprayed evenly for sample treatment and the same amount of distilled water was sprayed evenly for the control treatment to ensure thorough mixing. Three bags (300 mm × 230 mm) were prepared for each treatment, and each bag was loaded with 300 g of sample and then vacuum sealed. A total of 48 bags were stored in a carton and placed in the storage room (27 ± 1 • C) without being turned over. After 30 and 60 days of ensiling, the bags were opened for analysis of the chemical composition and fermentation characteristics of the silage, respectively. The microbial community was analyzed in the 60-day samples.

Chemical Analyses
To evaluate the fermentation characteristics of the silage, 10 g of fresh silage samples was taken from each bag, mixed with 90 mL of sterile water and then placed in a refrigerator at 4 • C for 24 h. Then, the mixture was mixed with a juicer and filtered with four layers of medical gauze. The pH of the supernatant was immediately measured with a pH meter (PHS-3E, Shanghai INESA Scientific Instruments Co., Ltd., Shanghai, China). Next, the organic acid (lactic acid (LA), acetic acid (AA), propionic acid (PA) and butyric acid (BA)) concentrations were determined by HPLC (KC-811 column, Shodex; Shimadzu Co., Ltd., Tokyo, Japan) via the methods described above [29]. Ammonia nitrogen was measured by phenol colorimetry [30]. The WSC content was quantified by the anthrone method according to the methods described above [31].
The remaining silage samples from each bag were dried in an oven at 65 • C for 48 h, and the dry matter (DM) weight of the samples was determined immediately. After weighing, the dried samples were ground using a grinder separately and passed through a 1 mm sieve for chemical composition analysis. The crude protein (CP) content was determined by a Kjeldahl nitrogen analyzer (Kjeltec 8400, FOSS, Sweden) using the Kjeldahl method [32]. The WSC content was determined by the anthrone method [33]. The neutral detergent fiber (NDF, assayed without alpha-amylase), acid detergent fiber (ADF) and acid detergent lignin (ADL) contents were analyzed by the method reported by Van Soest et al. [34]. Acid-insoluble ash was determined by incineration for 3 h at 550 • C according to the method described by Bergero et al. [35].

Microbial Population Analysis
The number of microbial populations was determined by the plate counting method according to Li et al. [29]. The filtered liquid was serially diluted with a gradient of 10 −1 to 10 −7 and shaken evenly, and three dilutions (10 −3 , 10 −5 and 10 −7 ; 100 µL each) were applied to the medium. MRS agar medium (GCM188, Beijing Land Bridge Technology Co., Beijing, China), Rose Bengal medium (YM01435, Shyuanmu Biomart Biotech. Co., Shanghai, China) and eosin methylene blue agar medium Beijing (02-002, Aoboxing Biotechnology Co., Beijing, China) were used to count the number of LAB, yeast and Enterobacteriaceae, respectively. Then, the Petri dish used for counting LAB was sealed with a sealing membrane. Subsequently, the LAB were incubated at 37 • C for 48 h, and yeast and Enterobacteriaceae were incubated at 30 • C for 48 h. Finally, the number of microorganisms on each plate was counted, and the values were transformed to log10 cfu/g FM values.

Bacterial Community Analysis
The DNA of the silage was extracted by the standardized cetyltrimethylammonium bromide (CTAB) method: An appropriate amount of sample was added to an EP tube containing CTAB lysate and lysozyme. Then, the sample was mixed with phenol, chloroform and isoamyl alcohol and centrifuged, and the upper layer was collected; this process was repeated with chloroform and isoamyl alcohol. Next, isoamyl alcohol was added to the upper layer, blended and precipitated at −20 • C. After centrifugation, the liquid was poured out, ddH2O was added to dissolve the DNA sample, and RNase A was added to digest the RNA. The V3~V4 variable region of bacterial 16S rRNA was amplified, and the PCR primer amplification sequences were CCTAYGGGRBGCASCAG (F) and GGACTACNNGGTTATCTAAT (R). Then, the amplicons were sequenced on an Illumina NovaSeq6000 platform in paired-end(PE) 250 bp mode, and the original sequencing was completed by Beijing Nuohe Biotechnology Co., Ltd., China. Finally, alpha diversity (Chao 1, Shannon, and Ace) and coverage value analyses were performed on the Magic platform (https://magic.novogene.com/, accessed on 13 March 2023).

Calculations and Statistical Analysis
The degradation rates of lignocellulose (NDF and ADF) during the ensiling process were calculated according to the following equation: IBM SPSS 27.0 software (SPSS Inc., Chicago, IL, USA) was used to statistically analyze the data, and one-way, two-way and three-way analyses of variance (ANOVAs) were performed on the fixed effects of growth stages, additives and ensiling days. The mean separation under the growth stages, additives and fermentation time was tested by Duncan's multiple range test for significance. When p < 0.05, the difference was considered statistically significant. The statistical significance of differences between datasets was assessed by PerMANOVA using the weighted PCoA scores on the Magic platform (https://magic.novogene.com/, accessed on 15 March 2023). A heatmap based on Spearman correlation coefficients between the bacterial population and fermentation quality was calculated with SPSS software. All graphics and thermal maps were prepared with GraphPad Prism 9.4.

Characteristics of Fresh Sudangrass before Ensiling
The chemical composition and microbial population of the fresh sudangrass are presented in Table 1. With increasing cutting stage, the sudangrass DM, WSC, CP, NDF, ADF and ADL contents significantly increased (p < 0.05). Sudangrass harvested at S2 showed relatively high contents of DM (p < 0.01) and WSCs (p < 0.05), indicating that it could provide more fermentation substrate [14] and higher NDF and ADF contents, which could be attributed to the decrease in the leaf-stem ratio [36]. The higher contents of NDF (higher than 650 g.kg −1 DM) and ADF (higher than 317 g.kg −1 DM) in the two stages of sudangrass were not conducive to ensiling fermentation and animal digestion [37]. Compared with the S1 stage, a higher abundance of LAB was found at the S2 stage of sudangrass (p < 0.01), which was consistent with a previous study [15]. The abundance of epiphytic LAB increased with the delay in the cutting stage, which was consistent with previous findings [14,15]. This is probably due to the increased mature tissues, which release some nutrients that could promote microbial growth and microbial community diversity [38]. Enterobacteriaceae were not detected in either growth stage.

Chemical Characteristics of Sudangrass after Ensiling
The effects of different growth stages and additives on the DM, CP, WSC and fiber contents at 30 and 60 days are presented in Table 2. In this study, the D × A × S interaction affected the CP (p < 0.01) and ADF contents (p < 0.05); the D × A interaction affected the WSC content (p < 0.001) and CP content (p < 0.05); the D × S interaction affected the CP content (p < 0.001), NDF content (p < 0.05) and ADF content (p < 0.05); the A × S interaction affected the WSC content (p < 0.05), NDF content (p < 0.05) and ADF content (p < 0.001); D significantly influenced the DM content (p < 0.01), CP content (p < 0.001), NDF content (p < 0.001) and ADF content (p < 0.001); A significantly influenced the DM content (p < 0.001), WSC content (p < 0.001), CP content (p < 0.001), NDF content (p < 0.001), ADF content (p < 0.001) and ADL content (p < 0.05); and S significantly influenced the DM content (p < 0.05), WSC content (p < 0.05), CP content (p < 0.001), NDF content (p < 0.001) and ADF content (p < 0.001). The DM content at S2 was higher than that at S1, which could be attributed to the higher DM content of the raw material and lower pH of S2 as the decrease in pH could suppress undesirable fermentation to conserve more nutrient substances [24]. Adding LP+CX resulted in a higher DM content at 30 and 60 days; moreover, at S1, the DM content of LP+CX was significantly higher (p < 0.05) than that of the CK sample at 60 days. This result was consistent with the results of previous studies [16,19]. The reason might be the synergistic effects of cellulase, xylanase and FAE-producing L. plantarum (homofermentative LAB), which degraded the sudangrass and provided more WSCs for LAB and then rapidly reduced the pH of the silages, resulting in more DM being preserved [19,39]. However, the DM content was lowest in the CX treatment at the two ensiling days, and the amount was significantly lower (p < 0.05) than that in CK-S1 at 30 days, which was similar to a previous study [39]. The above result occurred possibly because adding cellulase and xylanase led to increased lignocellulosic degradation and fermentation substrate production, and more substrate was consumed by the heterolactic fermentation of epiphytic LAB, which convert the substrates into liquid and gases, mainly ethanol and CO 2 , as LP was not added [40]. Adding CX resulted in a higher WSC content compared with the CK and LP treatments (p < 0.05) because cellulase and xylanase could degrade lignocellulose and produce more WSCs. Moreover, the WSC content of LP+CX was significantly higher than that of CX (except at the S1 stage at 30 days), which may contribute to the interactions of ligninolytic enzymes with FAE [26].
At the S1 stage, the additive treatments had higher CP contents than CK at 60 days (p < 0.05), and at the S2 stage, LP and LP+CX had higher CP contents than CK at 60 days (p < 0.05). The above results might be attributed to the higher content of organic acids and a quick decrease in the pH of the samples that added LP and CX (Table 3), which could inhibit the activity of undesirable microorganisms and proteolytic enzymes [16,41]. For the same treatment and 60-day silages, the CP contents in CK and CX at the S1 stage were higher than those in the samples at the S2 stage, but LP and LP+CX at the S1 stage had lower CP contents than those at the S2 stage. The reason might be the higher concentration of NH 3 -N in CK and CX and lower pH in LP and LP+CX at the S2 stage. For 60-day silages, the CK, CX, and LP treatments in S1, as well as the LP treatment in S2, showed higher CP contents compared to 30-day silages, which might be due to the lower NH 3 -N concentration (Table 3) at 60-day silages and the accumulation of mycoprotein with the extension of ensiling days [42]. Table 2. Chemical compositions of sudangrass silages at 30 and 60 days.

Items
Ensiling Days S1 S2 SEM . Different lowercase letters show that the same ensiling time treatment was significantly different in different additives and stages (p < 0.05). S1, Sudangrass cut at 1.8 m; S2, sudangrass cut at 2.0 m; D, ensiling days; A, additive; S, growth stage; D × A, the interaction between the ensiling days and additive; D × S, the interaction between the ensiling days and growth stage; A × S; the interaction between the additive and growth stage; D × A × S, the interaction among the ensiling days, additive and growth stage. "***", p < 0.001; "**" p < 0.01;"*", p < 0.05. . Different lowercase letters show that the same ensiling time treatment was significantly different in different additives and stages (p < 0.05). S1, Sudangrass cut at 1.8 m; S2, sudangrass cut at 2.0 m; ND, not detected; D, ensiling days; A, additive; S, growth stage; D × A, the interaction between the ensiling days and additive; D × S, the interaction between the ensiling days and growth stage; A × S; the interaction between the additive and growth stage; D × A × S, the interaction among the ensiling days, additive and growth stage. "***" p < 0.001; "**" p < 0.01; "*" p < 0.05.

p-Value
The changes in the NDF and ADF contents were similar to that of WSCs, and CX and LP+CX exhibited lower levels than CK and LP (p < 0.05). The lowest contents of NDF and ADF were found in LP+CX, especially at the S1 stage, and the contents in LP+CX were significantly lower than those in CX at 60 days of ensiling (p < 0.05). The results indicated that the combination of cellulase, xylanase and FAE-producing LAB had a better ability to degrade lignocellulose in sudangrass. This may be attributed to the structure of the plant cell wall; hemicellulose and lignin are crosslinked mainly via ferulate ester bonds [20]. Therefore, the synergistic action of FAE and lignocellulase (especially xylanase) can improve the degradation rate. Previous studies have reported that FAE combined with xylanase or cellulase showed improved degradation ability [26,39]. At 60 days, the NDF degradation rates in LP+CX were 34.69% (S1 stage) and 38.80% (S2 stage), while the ADF degradation rates were 41.06% (S1 stage) and 38.92% (S2 stage). Moreover, the NDF and ADF contents of S1 were lower than those of S2. The reason for this may be that, at the early stage, sudangrass has a low lignocellulose content and is more easily degraded. For the 60-day silages, the contents of NDF and ADF (p < 0.05) were lower than those for the 30-day silages. This phenomenon was also found in a previous study [24], which might be attributed to the enzymatic or acid hydrolysis of lignocellulose [16,24]. The cutting stages and additives had almost no effect on ADL and AIA, possibly because their components are difficult to degrade [19,43].
Previous studies have indicated that pH is an important parameter for silage fermentation, and only low pH values can inhibit the fermentation of undesirable microorganisms [19,44]. Compared with that at 30 days, the pH of the silage at 60 days decreased at the two stages (except LP-S2), especially for LP+CX, and the decrease was significant (p < 0.05). After 60 days of ensiling, the additives all decreased the silage pH, and almost all decreased it below 4.2. The lower pH suggested that adding LAB and enzymes could improve silage fermentation and produce much LA (Table 2), which was consistent with previous studies [16,24]. However, compared with CK, only CX-S1 and LP+CX (two stages) exhibited significant changes in pH (p < 0.05), which may be due to the large number of epiphytic LAB and adequate WSC content of gramineous plants. Moreover, the synergistic action of FAE-producing LAB and lignocellulase might have resulted in the significant difference between LP+CX and other treatments (p < 0.05), as observed in previous studies [16,39]. In general, the pH in the S2 stage was lower than that in the S1 stage, which might be due to the higher epiphytic LAB abundance [15] and fermentation substrate level (Table 1).
LA is the most important substance responsible for decreasing the silage pH, and increased LA production can accelerate pH decline and improve silage quality [41]. Therefore, adding LP and CX increased the concentration of LA in sudangrass silage, and the pH of these silages decreased. However, after 60 days of fermentation, only CX-S1 and LP+CX (two stages) exhibited significantly higher LA concentrations than CK (p < 0.05), which was consistent with the pH values and could be attributed to the characteristics of fresh sudangrass (the high content of NDF, ADF and ADL and low content of WSC) ( Table 1) and the synergistic action of LP and CX. Similarly to pH, the S2 silage had a higher LA content than the S1 silage. At 30 and 60 days, the AA content in the S2 stage was significantly lower than that in the S1 stage (p < 0.05), and later harvesting led to lower AA concentrations, which might be due to the greater fermentation substrate levels (Table 1), higher abundance of Lactobacillus ( Figure 1) and stronger homofermentative pathway present in the S2 stage. All the treatments (except LP-S1, CX-S2 and LP-S2 on day 60) exhibited higher LA/AA ratios than CK. Especially for LP+CX-S1, the ratio was significantly larger than that in CK (p < 0.05), which could be attributed to the addition of homofermentative LAB and the synergistic action of LP and CX, leading to increased homolactic fermentation and LA production [16]. Similarly, the LA/AA ratios in S2 were higher than those in S1 (p < 0.05).
The PA and BA concentrations in S2 were lower than those in S1 (p < 0.05), which was similar to the results for the AA concentration. The increased production of PA and BA often results in a large amount of nutrient loss [19]. The PA concentrations in silages increased with the extension of the ensiling process, especially for CK-S1, CX-S1 and LP-S1 (p < 0.05), which was similar to the results of a previous study [24]. In the present study, adding CX and LP significantly decreased the concentration of NH 3 -N, especially in LP+CX, which exhibited a lower NH 3 -N concentration than the other treatments (p < 0.05). The above result may be attributed to the addition of LP and CX, resulting in a lower pH and then suppressing the proteolytic action of plant enzymes and the growth and/or activity of undesirable bacteria ( Figure 1) [16,45]. However, S2 had higher NH 3 -N concentrations than S1, contrary to the pH, AA, PA and BA results, which might be due to the higher relative abundance of Enterobacter ( Figure 1). Moreover, for 60-day silages, LP+CX-S1, CK-S2, CX-S2 and LP+CX-S2 had significantly higher NH 3 -N concentrations (p < 0.05) than 30-day silages, which might be due to protein degradation by undesirable microorganisms during long-term storage [9]. The PA and BA concentrations in S2 were lower than those in S1 (p < 0.05), which was similar to the results for the AA concentration. The increased production of PA and BA often results in a large amount of nutrient loss [19]. The PA concentrations in silages increased with the extension of the ensiling process, especially for CK-S1, CX-S1 and LP-S1 (p < 0.05), which was similar to the results of a previous study [24]. In the present study, adding CX and LP significantly decreased the concentration of NH3-N, especially in LP+CX, which exhibited a lower NH3-N concentration than the other treatments (p < 0.05). The above result may be attributed to the addition of LP and CX, resulting in a lower pH and then suppressing the proteolytic action of plant enzymes and the growth and/or activity of undesirable bacteria (Figure 1) [16,45]. However, S2 had higher NH3-N concen-  Table 4. The D × A (p < 0.001), D × S (p < 0.05) and A × S (p < 0.001) interactions affected the LAB abundance. Compared to other treatments, LP+CX exhibited lower LAB abundance at the two stages. This result could be attributed to the lower pH in these silage samples, as many LAB are less tolerant to lower pH [46]. However, for the CX treatments, the silage samples had a lower pH than those in CK and LP, and the number of LAB in CX-S1 was lower than that in CK and LP-S1, but CX-S2 had different results, which might be caused by other reasons that need further study. Moreover, the LAB population was decreased at 60 days of ensiling (LP and LP+CX treatments in S1 and CX and LP+CX in S2) compared to 30 days of ensiling, which might be attributed to the lower pH at 60 days [46]. Except for the LP-S2 and LP+CX-S2 treatments, no Enterobacteriaceae were detected in any of the treatments, which might be due to the lower pH. However, a certain amount of Enterobacteriaceae was detected in LP-S2 and LP+CX-S2, as shown in Figure 1, and a previous study showed the presence of bacteria in Enterobacter at silages of a lower pH [9], which was likely due to the presence of acid-tolerant species [47,48]. Yeast was not detected in either stage.

Bacterial Diversity and Community of Fresh Material and Sudangrass Silage on Day 60
The diversity and richness of the bacteria according to Pielou's e, the Chao1 index, the Shannon index, the Simpson index and Good's coverage index are presented in Table 5. Pielou's e reflects the uniformity of the samples. The value for S2 was lower than that for S1 (p < 0.05), and adding LP decreased Pielou's e, indicating that adding LP at S2 increased the uniformity of the bacteria. Similarly to Pielou's result, LP and LP+CX had lower Chao 1, Shannon and Simpson values, which could reflect the richness (Chao 1) and diversity (Shannon index and Simpson index) of bacteria [8]. The above result could be attributed to the lower pH, which could suppress the growth of some bacteria [16,49]. The Good's coverage of all samples was above 0.999, which suggested that most bacteria could be detected by high-throughput sequencing, making it feasible to analyze microbial communities [16].
The relative abundances of bacterial communities at the phylum and genus levels in the fresh sudangrass sample and after ensiling for 60 days are presented in Figure 1. As shown in Figure 1A, Proteobacteria and Firmicutes were the most abundant phyla in the fresh sudangrass, which was consistent with the results of a previous study [50]. After ensiling, the relative abundance of Proteobacteria decreased greatly, but the relative abundance of Firmicutes increased greatly to 71.65~95.22%, especially for CK-S1 (95.22%), LP-S1 (95.43%) and LP-S2 (91.42%), compared with that in the fresh material (31.05% and 46.90%). The dominant phyla were Firmicutes and Proteobacteria, which were also detected in another study [16,19]. At the S1 stage, adding LP and CX increased the relative abundance of Firmicutes, which is the most common bacteria in silage, compared to the CK treatment, possibly because most bacteria cannot live in anaerobic and acidic conditions and are replaced by bacteria belonging to Firmicutes [51]. A higher relative abundance of Proteobacteria was found in CX-S1, CK-S2 and CX-S2; this phylum includes various pathogenic bacteria that can digest organic matter [16,49].
CK, Control; LP, ferulic acid esterase−producing Lactobacillus plantarum; CX, mixture of cellulase and xylanase; LP+CX, combination of ferulic acid esterase−producing Lactobacillus plantarum, cellulase and xylanase. Different capital letters show that the same additive and stage treatment have significant differences in different ensiling times (p < 0.05). Different lowercase letters show that the same ensiling time treatment was significantly different in different additives and stages (p < 0.05). LAB, Lactic acid bacteria; cfu, colony-forming units; FM, fresh matter; SEM, standard error of means; ND, not detected; S1, sudangrass cut at 1.8 m; S2, sudangrass cut at 2.0 m; D, ensiling days, A, additive; S, growth stage; D × A, the interaction between the ensiling days and additive; D × S, the interaction between the ensiling days and growth stage; A × S, the interaction between the additive and growth stage; D × A × S, the interaction between ensiling days, the additive and growth stage. "***", p < 0.001; "*", p < 0.05. In Figure 1B, the most dominant genera were Lactiplantibacillus, Aureimonas and Sphingomonas in the fresh sudangrass sample, but Lactiplantibacillus, Ralstonia and Weissella became the dominant genera after 60 days of ensiling. After ensiling in an anaerobic environment and adding additives, the relative abundance of Lactiplantibacillus increased to 65.52~94.08%, indicating a vital role in ensiling, as these bacteria could rapidly decrease the silage pH and have high tolerance to a low pH [16,52]. S1-CK had the highest abundance of Lactiplantibacillus (94.08%), but the pH of these silage samples was higher than that of the other treatments, possibly because the amount of fermentation substrate in the S1 stage of sudangrass was not sufficient for LAB and was fermented by epiphytic and heterofermentative LAB, which might be the main component of Lactiplantibacillus in S1-CK [40]. However, another reason may need further study. A low pH could inhibit the survival of many LAB [49], and the pH above 4.13 in S1-CK might have resulted in the presence of more bacteria in the genus Lactiplantibacillus compared with the pH below 4.0. Moreover, the heterolactic fermentation of epiphytic LAB belonging to Lactiplantibacillus (such as Lactiplantibacillus brevis) may reproduce in large quantities. The above reasons might result in the increase in the relative abundance of Lactiplantibacillus in S1-CK. At the S2 stage, adding LP and CX enhanced the Lactiplantibacillus abundance, especially for LP+CX, which led to an increased production of LA and could rapidly decrease the pH of these silages, which is consistent with a previous study [19,24]. The silage samples in the S1 stage had a higher relative abundance of Ralstonia than those in the S2 stage. A previous study also detected this genus in their study, and some of the bacteria in this genus were plant pathogens that might have negative effects on ensiling fermentation [53]. S1-LP exhibited a 6.75% abundance of Weissella, but the S2 silage exhibited a lower abundance. This genus contains heterofermentative LAB that can produce both LA and AA and are generally considered early colonizers during ensiling [54,55]. Therefore, S1-LP produced more AA (Table 3). CK, CX and LP had a higher abundance of Enterobacter (1.56~4.26%) at the S2 stage compared to the S1 stage and S2-LP+CX possibly because some bacteria in the genus Enterobacter could protect themselves under low pH environments [47,48]. Previous studies have reported that Enterobacter could compete with other LAB to consume WSCs and degrade proteins to increase the NH 3 -N content [16].
Principal coordinate analysis (PCoA) of the bacterial community, as shown in Figure 2, was used to assess the variance of the bacterial communities in different silage samples [50]. Principal component 1 (PC1) and principal component 2 (PC2) accounted for 89.26% and 6.37%, respectively, with a total variance of 95.63%. The fresh samples of the two stages were in the first and fourth quadrants, while the silage samples were all in the second and third quadrants, indicating differences in the bacterial community between fresh and treated silage samples (p > 0.05, Table S1), which was similar to the results of the previous study [19]. However, at the two stages, the CK and additive treatments showed no significant difference (p > 0.05, Table S1). For the S2 stage, CK2 was distributed in the second quadrant, while most additive treatments were in the third quadrant, suggesting that the additives had an effect on the S2 stage silages.
Fermentation 2023, 9, x FOR PEER REVIEW 13 previous study [19]. However, at the two stages, the CK and additive treatments sho no significant difference (p > 0.05, Table S1). For the S2 stage, CK2 was distributed i second quadrant, while most additive treatments were in the third quadrant, sugge that the additives had an effect on the S2 stage silages.

Correlation Analysis of the Microbial Community and Some Silage Parameters
The Spearman correlation between bacterial abundance and the silage param was established and visualized via a heatmap, as shown in Figure 3. Previous studies shown that the metabolites produced during the process of ensiling affect the bact community, and microbial diversity can also improve silage metabolite levels and qu [45]. Therefore, the correlation between bacterial abundance and silage characteristic been explored previously [19,24]. In this study, the abundance of the most dominan nus, Lactiplantibacillus, was negatively correlated with the pH and the concentratio NH3-N (p < 0.05) and positively correlated with the WSC content (p < 0.05), CP conten the concentrations of organic acids (LA concentration (p < 0.05), AA and PA), indic that this genus could increase the concentrations of LA and AA and decrease the d dation of CP. The negative correlation of Lactiplantibacillus abundance with the NDF tent (p < 0.001) and ADF content (p < 0.001) is shown in Figure 3, suggesting that increa the abundance of Lactiplantibacillus improved the degradation of NDF and ADF. above result could be attributed to the addition of FAE-producing L. plantarum. O

Correlation Analysis of the Microbial Community and Some Silage Parameters
The Spearman correlation between bacterial abundance and the silage parameters was established and visualized via a heatmap, as shown in Figure 3. Previous studies have shown that the metabolites produced during the process of ensiling affect the bacterial community, and microbial diversity can also improve silage metabolite levels and quality [45]. Therefore, the correlation between bacterial abundance and silage characteristics has been explored previously [19,24]. In this study, the abundance of the most dominant genus, Lactiplantibacillus, was negatively correlated with the pH and the concentration of NH 3 -N (p < 0.05) and positively correlated with the WSC content (p < 0.05), CP content and the concentrations of organic acids (LA concentration (p < 0.05), AA and PA), indicating that this genus could increase the concentrations of LA and AA and decrease the degradation of CP. The negative correlation of Lactiplantibacillus abundance with the NDF content (p < 0.001) and ADF content (p < 0.001) is shown in Figure 3, suggesting that increasing the abundance of Lactiplantibacillus improved the degradation of NDF and ADF. The above result could be attributed to the addition of FAE-producing L. plantarum. Other genera of bacteria, such as Weissella, Enterobacter and Pseudocitrobacter, were negatively correlated with the content of CP and positively correlated with NH 3 -N concentration, which was also observed in previous studies [16,19]. The reason may be that Weissella are heterofermentative LAB and probably slow the pH decline [9,16], and Enterobacter and Pseudocitrobacter both belong to the family Enterobacteriaceae, which could probably result in proteolysis and produce ammonia-N [9,18]. However, the reason for the relationship between the Lactococcus abundance and CP content and NH 3 -N concentration needs further study.
genera of bacteria, such as Weissella, Enterobacter and Pseudocitrobacter, were negatively correlated with the content of CP and positively correlated with NH3-N concentration, which was also observed in previous studies [16,19]. The reason may be that Weissella are heterofermentative LAB and probably slow the pH decline [9,16], and Enterobacter and Pseudocitrobacter both belong to the family Enterobacteriaceae, which could probably result in proteolysis and produce ammonia-N [9,18]. However, the reason for the relationship between the Lactococcus abundance and CP content and NH3-N concentration needs further study. Figure 3. Heatmap of Spearman correlations between bacterial abundances and silage fermentation parameters on day 60. WSC, Water−soluble carbohydrate; AN, ammonia nitrogen; LA, lactic acid; AA, acetic acid; PA, propionic acid. The colors in the heatmaps indicate the Spearman correlation coefficient r, which ranges from −0.7 to 0.7. r < 0 indicates a negative correlation, and r > 0 indicates a positive correlation. '*', and '**' represent p < 0.05, p < 0.01 and p < 0.001, respectively.

Conclusions
Overall, adding FAE-producing LP, cellulase and xylanase could improve the quality of sudangrass silage by effectively decreasing the pH, NDF content and ADF content; conserving more of the DM and CP content; and producing more LA, especially for the LP+CX additive. Taken together, a higher DM content, similar CP content and lower pH were found at the S2 stage and after ensiling for 60 days. Moreover, the additives and cutting stage had a strong effect on the microbial community, and LP+CX had a higher abundance of Lactiplantibacillus at the S2 stage. Based on the results of this research, cutting at the S2 stage, adding FAE-producing L. plantarum, cellulase and xylanase and ensiling for 60 days can be an appropriate method for preserving sudangrass silage and degrading lignocellulose.
Supplementary Materials: The following supporting information can be downloaded at www.mdpi.com/xxx/s1. Table S1: p values of community difference between different treatments (PerMANOVA). Heatmap of Spearman correlations between bacterial abundances and silage fermentation parameters on day 60. WSC, Water−soluble carbohydrate; AN, ammonia nitrogen; LA, lactic acid; AA, acetic acid; PA, propionic acid. The colors in the heatmaps indicate the Spearman correlation coefficient r, which ranges from −0.7 to 0.7. r < 0 indicates a negative correlation, and r > 0 indicates a positive correlation. '*', and '**' represent p < 0.05, p < 0.01 and p < 0.001, respectively.

Conclusions
Overall, adding FAE-producing LP, cellulase and xylanase could improve the quality of sudangrass silage by effectively decreasing the pH, NDF content and ADF content; conserving more of the DM and CP content; and producing more LA, especially for the LP+CX additive. Taken together, a higher DM content, similar CP content and lower pH were found at the S2 stage and after ensiling for 60 days. Moreover, the additives and cutting stage had a strong effect on the microbial community, and LP+CX had a higher abundance of Lactiplantibacillus at the S2 stage. Based on the results of this research, cutting at the S2 stage, adding FAE-producing L. plantarum, cellulase and xylanase and ensiling for 60 days can be an appropriate method for preserving sudangrass silage and degrading lignocellulose.