Citric Acid Influences the Dynamics of the Fermentation Quality, Protease Activity and Microbial Community of Mulberry Leaf Silage

Mulberry (Morus alba) leaves has performed well as a high-quality protein supplement for livestock and enriches the edible resources of livestock. However, the harvest of mulberry leaves is seasonal and occurs mainly during the rainy season in southeast China; therefore, humid and sultry weather causes serious losses of mulberry leaf biomass, which pose a challenge for the preservation of mulberry leaves. In this study, we used the silage fermentation method to preserve mulberry leaves and investigated the effects of citric acid on the silage quality of mulberry leaves. Mulberry leaves were ensiled with or without 1% citric acid and 2% citric acid. The chemical composition, protein fraction and microbial community of mulberry leaf silages were analyzed. The results showed that the silage treated with citric acid had a higher dry matter recovery and lactic acid content and a lower acetic acid content, non-protein nitrogen content and ammonia-N content; citric acid also inhibited the activities of carboxypeptidase and aminopeptidase. Moreover, citric acid increased Lactobacillus abundance in silages and decreased the abundance of undesired microorganisms, such as Enterobacter. In summary, the addition of citric acid improved the fermentation quality of mulberry leaf silages, with 2% citric acid being more effective than 1% citric acid.


Introduction
Mulberry (Morus alba) is a traditional, economical woody plant and is cultivated across more than 780,000 ha in China [1]. The biomass yield of fresh mulberry leaves (ML) ranges from 25 to 30 mt/ha/yr [2]. The crude protein of ML ranges from 15 to 27.6% dry matter (DM) and ML is rich in minerals (2.42-4.71% Ca, 0.23-0.97% P, 1.66-3.25% K) [3]. As it is rich in nutrients, ML has performed well as a high-quality protein supplement for livestock [4]. In addition, ML has many biological benefits because it has antioxidant, anti-inflammatory, antidiabetic and antihyperglycemic properties, etc., and thus could potentially replace antibiotics to improve animal immunity [5,6]. However, the harvest of ML is seasonal and occurs mainly during the rainy season in southeast China, making it difficult to dry fresh ML naturally.
Ensiling is an efficient, low-cost and suitable method of preserving ML quality which can overcome the seasonal imbalance between forage resources and animal production. In a previous study, it was observed that proteolysis was a very serious problem in ML silage [7]. It is known that the activities of plant enzymes and microorganisms are mainly responsible for proteolysis, which can degrade proteins into peptides and further into ammonia-N [8]. However, information regarding the dynamics of proteolysis and the activities of protease and microorganisms in ML silage is lacking. Evaluation of the proteolytic mechanism would be beneficial in providing a theoretical basis for finding suitable proteolytic inhibitors of ML silage.
ML was cultivated and harvested from an experiment field in December 2020 at South China Agricultural University (Guangzhou, China, 113 • 35 E, 23 • 16 N). CA was purchased from Sinopharm Chemical Reagent Co. Ltd. (Shanghai, China). The ML was immediately chopped to a suitable length (1-2 cm) in a clean test bed when it was harvested. The ML was then ensiled with 1% or 2% CA. A control group (CK) involved no addition of CA. In detail, approximately 100 g ML was ensiled in an enclosed bag (20 cm × 30 cm) in triplicate for every treatment. In total, 36 sealed silage bags (3 treatments × 4 ensiling times × 3 repeats) were made and kept at ambient temperature (17-28 • C). After 3, 7, 14 and 30 days of ensiling, to evaluate the fermentation parameters, protease activity and the bacterial community, three silage bags from each treatment were opened at random.

Microorganism and Fermentation Parameter Analyses
We used the flat colony counting method to count the microbial population of silages according to Yan et al. [15]. A sample (10 g) from each silage bag and sterilized saline water (90 mL) were homogenized in a blender and then serially diluted. The population of lactic acid bacteria was counted in Man Rogosa Sharpe agar and that of coliform bacteria was counted in Violet Red Bile agar, and samples were subsequently incubated at 30 • C for 48 h. Molds and yeasts were cultured and counted in Rose Bengal agar at 28 • C for 48 h and then categorized by colony morphology. All microbial population counts were log10 transformed.
Another sample (10 g) from each silage bag was diluted in sterile distilled water (90 mL) and stored in a freezer at 4 • C for 18 h, then the homogenized sample was filtered. Part of the filtrate was immediately used to measure pH, and the remaining filtrate was used to measure the ammonia-N content and organic acid content. Following Broderick and Kang [16], the phenol-hypochlorite procedure was used to measure the ammonia-N content. The concentration of organic acids was measured using high-performance liquid chromatography (HPLC) (column, Shodex RSpak KC-811S-DVB gel C (8.0 mm × 30 cm; Shimadzu, Tokyo, Japan); oven temperature, 50 • C; mobile phase, 3 mmol/L HClO 4 ; flow rate, 1.0 mL/min; injection volume, 5 µL; detector, SPD-M10AVP) [17].
An oven was used to dry the remaining silage at 65 • C for 48 h for measurement of the DM content. For chemical analysis, these dry samples were ground with a grinder (FW100, Taisite Instrument Co., Ltd., Tianjin, China) and passed through a 1.00 mm screen. The powder samples were used to measure protein, fiber and water-soluble carbohydrates. The content of protein fractions, neutral detergent fiber and acid detergent fiber were determined by the method of Wang et al. [2]. Briefly, crude protein and true protein were analyzed using the Kjeldahl nitrogen analyzer (Kjeltec 2300 Auto-Analyzer, FOSS Analytical AB, Hoganas, Sweden). Non-protein nitrogen was calculated from the difference between crude and true protein. Neutral detergent fiber (NDF) and acid detergent fiber content were measured without use of heat-stable amylase and sodium sulfite in an A220 Fiber Analyzer (ANKOM Technology Corp., Macedon, NY, USA). The water-soluble carbohydrate content was analyzed by the anthrone method.

Protease Activity Analysis
The protease was extracted according to the method of He et al. [8], with some modifications. A sample (10 g) from the silage bag and a 0.1 M sodium phosphate buffer (50 mL, pH 6.0 with 5 mM hyposulfite) were homogenized in a blender, then the homogenate was filtered and the filtrate was centrifuged at 10,000× g at 4 • C for 15 min. The supernatant was retained at −80 • C for later use. According to the method of Guo et al. [18], acid protease activity, carboxypeptidase activity and aminopeptidase activity were measured.

Analysis of Microbial Communities
Following the manufacturer's instructions, microbial DNA was extracted by the HiPure Stool DNA Kits (Magen, Guangzhou, China). The primers 341F (ACTCCTACGGGAG-GCAGCAG) and 806R (GGACTACHVGGGTWTCTAAT) were used for amplifying 16S rDNA V3-V4 in a PCR thermal cycler (FlexCycler; Analytik Jena, Jena, Germany). DNA was amplified according to the procedure of Zhang et al. [19]. Illumina NovaSeq 6000 was used to sequence the purified PCR products. The raw reads were assembled according to the procedure set by Liu et al. [20].

Statistical Analysis
The analyses of the fermentation parameters were performed with SAS version 9.1 (SAS Institute, Cary, NC, USA) to evaluate the influence of treatment and ensiling time. Duncan's multiple range tests were used to measure mean values. p < 0.05 was considered significant and p < 0.01 was considered extremely significant. In addition, Pearson's correlation analysis between protein fractions and protease activity was conducted with SAS.

Characteristics of Pre-Ensiled ML
The characteristics of the pre-ensiled ML are shown in Table 1. The DM content of ML was 374 g/kg fresh matter (FM). The relatively high DM content of ML is comparable with that of tropical legumes, such as alfalfa and stylo [21]. The crude protein content of ML was 139 g/kg DM. It was lower than the 188 g/kg DM reported by Wang et al. [2]; this might be because it was at the mature stage [22]. Neutral detergent fiber is an inferior predictor of digestibility [7]. The low neutral detergent fiber content (440 g/kg DM) indicated that ML probably had better digestive efficiency as a fodder for animals. The water-soluble carbohydrates were a necessary fermentation substrate during the ensiling process and its content of greater than 60 g/kg DM was a guarantee of thorough fermentation [19]. Water-soluble carbohydrate content was 114 g/kg DM in this study, which was enough to support the acid production efficiency of microorganisms. Epiphytic lactic acid bacteria and coliform bacteria counts were 5.72 log10 cfu/g FM and 5.21 log10 cfu/g FM, respectively. Yeasts and molds were below detectable levels (2.00 log10 cfu/g FM). During the ensiling, lactic acid bacteria counts of more than 5.00 log10 cfu/g FM are considered to be a superior predictor of fermentation quality [23]. However, the relatively high coliform bacteria counts might result in poor fermentation, which suggests that fermentation inhibitors were necessary to inhibit undesirable microorganisms.

The Fermentation Parameters, Organic Acids and Microbial Population of ML Silages
DM recovery, the dynamics of the pH value, organic acid content and microbial population counts in ML silages are shown in Figure 1. The factor analysis revealed that CA had a significant effect (p < 0.01) on DM recovery, pH, lactic acid content, acetic acid content, lactic acid bacteria counts and coliform bacteria counts in ML silages. DM loss is a common issue during every stage of ensiling, which reduces the silage yield [24]. The silages treated with CA showed higher DM recovery than the CK, and the highest DM recovery occurred in the silages treated with 2% CA after 30 days of ensiling. The increase in DM recovery would significantly improve the economic efficiency of ML as an excellent protein supplement for livestock. The pH is the most intuitive fermentation parameter, which directly reflects the quality of silage. The pH value of silages treated with CA was lower than that of the CK during the ensiling process in this study. The pH value of ML silages reached stability in 3 days of ensiling, and the final pH was lower than 4.5 with the addition of 2% CA. It might be that enough CA directly reduced the pH value due to its strong acidity, with a pK1 3.13 [10]. However, the pH value of silage treated with 1% CA was maintained at around 5.0 and the final pH value of the CK remained approximately 6.0, which might be due to the high buffer capacity of ML [25]. The relatively low pH of ML silage ensiled with CA indicated that CA could effectively promote the fermentation of ML.
During the ensiling process, lactic acid and acetic acid were the dominant organic acids in ML silage, and no butyric acid or propionic acid was detected. The addition of CA significantly increased (p < 0.01) the lactic acid content and reduced the acetic acid content of ML silages. Similarly, Ke et al. [14] reported that 0.1% CA and 0.5% CA increased the lactic acid content. In the first 7 days of ensiling, the lactic acid accumulated rapidly in the ML silages. After 7 days of ensiling, the lactic acid content decreased in the silages treated with 2% CA, and after 14 days of ensiling in the CK and the silages treated with 1% CA. At the same time, the acetic acid content of the CK and the silages treated with 1% CA increased. The drop in lactic acid content and the rise in acetic acid content indicated that part of the lactic acid was converted into acetic acid, which might be caused by heterofermentation of the lactic acid bacteria, such as Lactobacillus buchneri [26]. Moreover, the ML silages treated with CA showed a lower (p < 0.01) ratio of acetic acid to lactic acid than the CK, which suggested that the addition of CA promoted the accumulation of lactic acid and inhibited the production of acetic acid during the ensiling process.
The dynamics of lactic acid bacteria, coliform bacteria and yeasts are shown in Figure 1. In the first 7 days of ensiling, lactic acid bacteria counts increased rapidly, which could explain the accumulation of lactic acid at the same time. The lactic acid bacteria counts then decreased from 8.0 log10 cfu/g FM to 6.0 log10 cfu/g FM when the ensiling time was prolonged, which was consistent with the report by Wang et al. [2]. The reduction in lactic acid bacteria counts might be explained by the lack of water-soluble carbohydrates as a fermentation substrate [27]. Compared with the CK, lower (p < 0.01) coliform bacteria counts were observed in the silages treated with CA during the ensiling process, which could explain the higher DM recovery and the lower acetic acid content in these silages. Moreover, the reduction in coliform bacteria counts of silages treated with CA was consistent with those reported by Mohan and Pohlman [28].
When the ensiling time was prolonged, yeast counts decreased (p < 0.01), which might be because the progressively anoxic situation inhibited its growth. Unexpectedly, no yeast was detected in the silages treated with 2% CA during the fermentation process. However, Lv et al. [13] reported that yeast counts were increased in alfalfa and Amomum villosum Lour silages with the addition of CA. This discrepancy was probably due to the water-soluble substances of the ML having a broad spectrum of antimicrobial activity [29]. Moreover, CA and the water-soluble substances of ML might have synergistic effects that inhibit the growth of yeasts. Therefore, these results indicated that more research is needed into the function of CA in controlling fungal growth in ML silage.
Fermentation 2021, 7, x FOR PEER REVIEW 5 of 16 counts were observed in the silages treated with CA during the ensiling process, which could explain the higher DM recovery and the lower acetic acid content in these silages. Moreover, the reduction in coliform bacteria counts of silages treated with CA was consistent with those reported by Mohan and Pohlman [28].

The Protein Fraction of Ml Silages
The dynamics of the protein fractions of ML silages are shown in Table 2. Proteolysis is a common phenomenon in silage and is the main factor affecting the protein quality of silage [7]. In the first stage, under the action of plant proteolytic enzymes, most of the proteins are degraded into non-protein nitrogen, such as free amino acids, oligopeptides and dipeptides, then these hydrolysis products are further converted to ammonia-N by microorganism activities [7]. When the ensiling time was prolonged, the content of nonprotein nitrogen increased (p < 0.01) from 122 g/kg TN to 380 g/kg TN, which indicated extensive proteolysis in the ML silage. The increase in non-protein nitrogen indicated a decline in the nutritional value of silage proteins, considering that the utilization efficiency of non-protein nitrogen in ruminants is lower than that of true protein [30]. Thus, inhibitors, which could inhibit protease and microbial activities, were necessary. After 30 days of ensiling, the highest non-protein nitrogen content was observed in the CK, whereas 2% CA decreased (p < 0.05) non-protein nitrogen content. Similarly, the addition of CA in alfalfa silages showed a quite different non-protein nitrogen content which decreased with the addition of 0.5% CA after 60 days of ensiling [31]. Moreover, compared with the CK, the non-protein nitrogen content varied less and had a lower content (310 vs. 380 g/kg DM) in the silages treated with 2% CA, implying that the proteins could be preserved better and more stably in silages treated with 2% CA. Ammonia-N, reflecting the degree of deamination of amino acids and peptide, is a more precise index compared with non-protein nitrogen. Consistent with the nonprotein nitrogen content, the content of ammonia-N increased (p < 0.01) when the ensiling time was prolonged, which was related to the deamination of amino acids and peptides caused by microbial activities [8]. During ensiling, the content of ammonia-N decreased (p < 0.01) in the silages treated with CA, indicating that CA inhibited the deamination of microorganisms. This inhibition could suggest that the amount of free amino acids, oligopeptides and dipeptides in non-protein nitrogen increased with the addition of CA, thereby improving the nutritive value of non-protein nitrogen. Similarly, Ke et al. [31] observed that ammonia-N decreased from 132 g/kg TN to 31.7 g/kg TN with the addition of 0.5% CA.
Compared with the CK, the relatively low content of ammonia-N in the silages treated with CA might be due to direct acidification and the antimicrobial properties of CA inhibiting the growth of undesirable microorganisms, such as Clostridium and Enterobacter, which is in accordance with the decrease in coliform bacteria counts [13]. Moreover, the silages treated with 2% CA showed the lowest ammonia-N content during the ensiling process. The above results indicate that CA improved the silage quality of ML.

The Protease Activities of ML Silage
The protease activities of ML silage are shown in Table 3. Carboxypeptidase, aminopeptidase and acid protease were the exopeptidases mainly responsible for the first stages of proteolysis [8]. Carboxypeptidase activity increased (p < 0.01) when the ensiling time was prolonged. The increased activity of carboxypeptidase may be related to the decrease in pH towards 5.4, at which point carboxypeptidase was optimally active [32]. However, carboxypeptidase activity remained stable in the silage treated with 2% CA during the ensiling process. However, carboxypeptidase activity was consistently higher than 60.0 units/h/DM at 30 days of ensiling in all treatments, which could partly account for the relatively high non-protein nitrogen content of ML silage ( Table 2). Aminopeptidase had low tolerance to pH and the optimal activity of aminopeptidase was observed at pH 7.0 [18]. Aminopeptidase activity gradually decreased (p < 0.01) in this study, which was consistent with the change in pH value ( Figure 1). Moreover, aminopeptidase activity sharply decreased (p < 0.01) with the addition of CA, which might be due to the direct acidification of CA. Similarly, Yuan et al. [32] and He et al. [8] observed that formic acid decreased aminopeptidase activity by rapid acidification of the silage. Furthermore, aminopeptidase can release free amino acid residues to provide an additional easily assimilated carbon source for lactic acid bacteria via deamination, accumulating extracellular ammonia-N [8]. Thus, the CA-inhibited aminopeptidase activity might partially account for the decrease in ammonia-N ( Table 2). In contrast to carboxypeptidase and aminopeptidase, acid protease activity was slightly increased (p < 0.05) in the silages treated with CA during the ensiling process. Although the pH in the silages treated with CA decreased, the pH still approached the optimal level for acid protease (4.50 in alfalfa). Thus, CA enhanced the acid protease activity. Although acid protease activity was lower than carboxypeptidase activity, it varied from 47.1 to 55.2 units/h/DM, which also promoted the hydrolysis of proteins. Overall, the relatively high activity of carboxypeptidase and the increased activity of acid protease in the silages treated with CA suggested that more CA should be added to ML silage to adjust the pH below 4.5, which might better reduce the degree of protein hydrolysis and improve the silage quality of ML.
To further study the relationship between protease and protein fractions, Pearson's correlations were conducted on the data for protein fractions and protease activity. As shown in Table 4, carboxypeptidase activity was negatively correlated (p < 0.01) with true protein content but positively correlated (p < 0.01) with non-protein nitrogen and ammonia-N content, which might partly explain the proteolysis in ML silage. The results indicated that CA inhibited (p < 0.01) carboxypeptidase activity could effectively inhibit extensive proteolysis in ML silage during the ensiling process, with 2% CA showing the best outcome.

The Dynamic Variation and Alpha Diversity of the Bacterial Community
The dynamic variation of the bacterial community was illustrated by principal component analysis (Figure 2). The results of the principal coordinate analysis was based on UniFrac (unweighted) distances. In the CK, the bacterial community showed the clearest distinctions when the ensiling time was prolonged. Moreover, after 7 days of ensiling, it was obvious that the silages treated with CA were separate from the CK, which suggested that the microbial community changed during the ensiling process. The distinctiveness of the bacterial communities among all treatments might account for the better fermentation that occurred in the ML silages treated with CA. the bacterial communities among all treatments might account for the better fermentation that occurred in the ML silages treated with CA. The alpha diversity of the bacterial community in ML silage is shown in Table 5. The coverage values of all samples were approximately 0.99, which indicated that most of the bacterial community was captured by the adequately large sequencing data. The Sobs, Ace and Chao1 indices decreased, but the Simpson index increased with a longer ensiling time, which indicated that the bacterial richness of ML silage decreased and dominant bacterial community richness increased. Compared with the CK, the silage treated with CA generally showed a lower Shannon index at the same stage of ensiling, which indicated that CA decreased the diversity of the bacterial community of ML silage.    The alpha diversity of the bacterial community in ML silage is shown in Table 5. The coverage values of all samples were approximately 0.99, which indicated that most of the bacterial community was captured by the adequately large sequencing data. The Sobs, Ace and Chao1 indices decreased, but the Simpson index increased with a longer ensiling time, which indicated that the bacterial richness of ML silage decreased and dominant bacterial community richness increased. Compared with the CK, the silage treated with CA generally showed a lower Shannon index at the same stage of ensiling, which indicated that CA decreased the diversity of the bacterial community of ML silage.

The Changes in the Dynamics of Relative Abundance among Bacterial Communities
The changes in the dynamics of relative abundance among bacterial communities by phylum are shown in Figure 3. The relative abundance of Proteobacteria, Firmicutes and Bacteroidetes (up to 69%, 11% and 10%, respectively) indicated that they were the dominant phylum in pre-ensiled ML; similar results were observed by Liu et al. [20]. With the prolonged ensiling time, the abundance of Proteobacteria and Bacteroidetes gradually decreased. After 30 days of ensiling, Firmicutes became the most abundant phylum in ML silages treated with CA, and Proteobacteria was still the most abundant phylum in the CK. Compared with the CK, the ML silages treated with CA showed a greater Firmicutes abundance but lower Proteobacteria abundance during all the stages of ensiling. Proteobacteria decreased in the silages treated with CA, which might be because the CA could enhance the permeability of the outer membrane of Gram-negative bacteria and inhibit the growth of Proteobacteria [33].
The dynamics of relative abundance among bacterial communities by genus are shown in Figure 4. The relative abundance of bacterial communities was very low in the pre-ensiled ML. Methylobacterium (4.01%) and Sphingomonas (1.03%) were the dominant genera in pre-ensiled ML, which was consistent with the results for pre-ensiled alfalfa reported by Ni et al. [34]. After 30 days of ensiling, Lactobacillus (24.9%), Kosakonia (9.67%), Enterobacter (8.96%), Methylobacterium (8.89%) and Enterococcus (6.26%) were dominant in the CK, whereas Lactobacillus was the only dominant genus in the silages treated with CA. Lactobacillus is usually the dominant genus in silage due to its more active and rapid growth when the environmental pH declines. Some Lactobacillus species are usually screened as inoculants due to their rapid growth, which enhances the fermentation quality, such as Lactobacillus plantarum and Lactobacillus casei [17,26]. The relative abundance of Lactobacillus increased from 0.4 to 24.9, 60.7 and 47.3% in the CK and the silages treated with 1% CA and 2% CA after 30 days of ensiling, respectively. In the same fermentation period, the silages treated with CA showed a higher Lactobacillus abundance, which was consistent with the higher accumulation of lactic acid content (Figure 1). Moreover, Lactobacillus abundance decreased in the silages treated with 1% CA after 14 days of ensiling, which might be due to the rapid depletion of fermentation substrates, which could not support the growth of Lactobacillus. The rapid depletion of fermentation substrate abundance might be responsible for the conversion of lactic acid to acetic acid in the silages treated with 1% CA because lactic acid becomes an alternative fermentation substrate for Lactobacillus. Similarly, Yan et al. [15] observed that relatively high Lactobacillus abundance corresponds to an elevation in acetic acid and a reduction in lactic acid. This conversion might be because different species of Lactobacillus have different metabolic patterns and functions. Similar to Lactobacillus, Enterococcus are desirable bacteria in the initial fermentation process, with the function of producing lactic acid to create a weak acid environment, which contributes to Lactobacillus fermentation. In the silages treated with CA, the relative abundance of Enterococcus decreased ( Figure 5), which might be due to its low tolerance to the acidic environment caused by CA [17,23].

The Changes in the Dynamics of Relative Abundance among Bacterial Communities
The changes in the dynamics of relative abundance among bacterial communities by phylum are shown in Figure 3. The relative abundance of Proteobacteria, Firmicutes and Bacteroidetes (up to 69%, 11% and 10%, respectively) indicated that they were the dominant phylum in pre-ensiled ML; similar results were observed by Liu et al. [20]. With the prolonged ensiling time, the abundance of Proteobacteria and Bacteroidetes gradually decreased. After 30 days of ensiling, Firmicutes became the most abundant phylum in ML silages treated with CA, and Proteobacteria was still the most abundant phylum in the CK. Compared with the CK, the ML silages treated with CA showed a greater Firmicutes abundance but lower Proteobacteria abundance during all the stages of ensiling. Proteobacteria decreased in the silages treated with CA, which might be because the CA could enhance the permeability of the outer membrane of Gram-negative bacteria and inhibit the growth of Proteobacteria [33]. The dynamics of relative abundance among bacterial communities by genus are shown in Figure 4. The relative abundance of bacterial communities was very low in the pre-ensiled ML. Methylobacterium (4.01%) and Sphingomonas (1.03%) were the dominant genera in pre-ensiled ML, which was consistent with the results for pre-ensiled alfalfa reported by Ni et al. [34]. After 30 days of ensiling, Lactobacillus (24.9%), Kosakonia (9.67%), Enterobacter (8.96%), Methylobacterium (8.89%) and Enterococcus (6.26%) were dominant in the CK, whereas Lactobacillus was the only dominant genus in the silages treated with CA. Lactobacillus is usually the dominant genus in silage due to its more active and rapid growth when the environmental pH declines. Some Lactobacillus species are usually screened as inoculants due to their rapid growth, which enhances the fermentation quality, such as Lactobacillus plantarum and Lactobacillus casei [17,26]. The relative abundance of Lactobacillus increased from 0.4 to 24.9, 60.7 and 47.3% in the CK and the silages treated with 1% CA and 2% CA after 30 days of ensiling, respectively. In the same fermentation period, the silages treated with CA showed a higher Lactobacillus abundance, which was consistent with the higher accumulation of lactic acid content ( Figure 1). Moreover, Lactobacillus abundance decreased in the silages treated with 1% CA after 14 days of ensiling,    It is known that Enterobacter is an undesirable facultative anaerobe in the fermentation process, being the principal competitor of the lactic acid bacteria for water-soluble carbohydrates under anaerobic conditions; its principal fermentation product is acetic acid [26]. Yang et al. [35] reported that Enterobacter was positively correlated with pH and negatively correlated with lactic acid. The high relative abundance of Enterobacter could explain the relatively high pH (5.99) and acetic acid content (11.0 g/kg DM) in CK after 30 days of ensiling. Moreover, Enterobacter can deaminate and decarboxylate amino acids, thereby enhancing ammonia-N production [36]. The relative abundance of Enterobacter decreased from 8.96 to 0.83% with the addition of CA after 30 days of ensiling in this study ( Figure 5), which might account for the lower acetic acid and ammonia-N content in the silages treated with CA. Methylobacterium are facultative methylotrophic bacteria and are commonly symbiotic with plants. The relative abundance of Methylobacterium decreased from 9.78 to 1.12% and 8.89 to 2.00% in the silages treated with CA after 14 days and 30 days of ensiling, respectively, because it is strictly aerobic and neutrophilic ( Figure 5) [37]. Methylobacterium are considered to be undesirable bacteria in silage because it positively correlates with levels of ammonia-N and negatively correlates with lactic acid [38,39]. The roles of Sphingomonas, Pantoea and Kosakonia in silage are still unclear. Fitzgerald et al. [40] observed that Sphingomonas is involved in nitrification under low dissolved oxygen conditions, which might resist the production of ammonia-N. The relative abundance of Sphingomonas decreased from 2.38 to 0.72% in the silages treated with CA after 30 days of ensiling. Ogunade et al. [39] speculated that Pantoea could preserve protein through their negative correlation with ammonia-N content. However, Li et al. [41] inferred that Pantoea might compete for the fermentation substrate with lactic acid bacteria and cause an accumulation of butyric acid in silage because Pantoea is separate from the Enterobacter genus. Furthermore, Guan et al. [42] reported that Pantoea was more active in high-moisture environments and positively correlated with butyric acid. However, no butyric acid was detected in the current study, which might be due to the low relative abundance of Pantoea (0.08%) in pre-ensiled ML. Similar to Pantoea, Kosakonia was separated from the Enterobacter genus. He et al. [43] reported that Kosakonia might reduce the ammonia-N content in silage due to its potential nitrogen fixing. The relative abundance of Pantoea and Kosakonia showed a similarly increasing trend in the CK during ensiling and was lower in the silages treated with CA in the current study. The relative abundance of Pantoea and Kosakonia decreased from 4.82 to 0.75% and 9.68 to 0.86% after 30 days of ensiling, respectively ( Figure 5). The potential capacity of Pantoea, Kosakonia and Sphingomonas to reduce ammonia-N might explain the relatively low ammonia-N in the CK after 30 days of ensiling (Table 2). However, the conventional functions of Pantoea and Kosakonia in silage are complicated; therefore, more research should be conducted to evaluate their function. Moreover, Roseomonas (Gram-negative bacteria) can split urea and assimilate arabinose and glucose, suggesting that it might compete for water-soluble carbohydrates with lactic acid bacteria and promote the production of ammonia-N [44]. The relative abundance of Roseomonas in the silages treated with CA decreased from 0.05 to 0.03%, from 0.10 to 0.03%, from 0.66 to 0.01% and from 0.20 to 0.03% after 3, 7, 14 and 30 days of ensiling in the current study, respectively.   The 16S rDNA gene-predicted functional profiles are shown in Figure 6. CA inhibited the metabolism of energy, carbohydrate, folding, sorting and degradation, and membrane

16S rDNA Gene-Predicted Functional Profiles
The 16S rDNA gene-predicted functional profiles are shown in Figure 6. CA inhibited the metabolism of energy, carbohydrate, folding, sorting and degradation, and membrane transport, which might be attributed to CA decreasing the pH value and inhibiting microbial activities directly. Moreover, CA inhibited the amino acid metabolism. Thus, the decrease in ammonia-N in the silages treated with CA might be because CA reduced the amino acid metabolism of Enterobacter, consistent with the decrease in Enterobacter abundance. All conversions of the abovementioned bacteria communities suggest that CA could enhance lactic acid bacteria fermentation by inhibiting the undesirable microorganisms in ML silage. The reduction in undesirable microorganisms and the increase in lactic acid bacteria might account for the lower pH, the increased DM recovery and the reduction in ammonia-N content in the silage treated with CA. Overall, CA had a positive effect on the fermentation quality and silage quality of ML. bial activities directly. Moreover, CA inhibited the amino acid metabolism. Thus, the decrease in ammonia-N in the silages treated with CA might be because CA reduced the amino acid metabolism of Enterobacter, consistent with the decrease in Enterobacter abundance. All conversions of the abovementioned bacteria communities suggest that CA could enhance lactic acid bacteria fermentation by inhibiting the undesirable microorganisms in ML silage. The reduction in undesirable microorganisms and the increase in lactic acid bacteria might account for the lower pH, the increased DM recovery and the reduction in ammonia-N content in the silage treated with CA. Overall, CA had a positive effect on the fermentation quality and silage quality of ML.

Conclusions
The study reveal that CA was useful for improving the fermentation quality and nutrition of ML silage. The content of acetic acid, non-protein nitrogen and ammonia-N, and coliform bacteria counts decreased in the silage treated with CA. The abundance of Enterobacter, Methylobacterium, Sphingomonas, Pantoea, Kosakonia and Roseomonas decreased, whereas Lactobacillus abundance increased when CA was added. Thus, the addition of CA could effectively preserve the quality of ML silage, and 2% CA was the most effective.

Conclusions
The study reveal that CA was useful for improving the fermentation quality and nutrition of ML silage. The content of acetic acid, non-protein nitrogen and ammonia-N, and coliform bacteria counts decreased in the silage treated with CA. The abundance of Enterobacter, Methylobacterium, Sphingomonas, Pantoea, Kosakonia and Roseomonas decreased, whereas Lactobacillus abundance increased when CA was added. Thus, the addition of CA could effectively preserve the quality of ML silage, and 2% CA was the most effective.