Effects of Malic Acid and Sucrose on the Fermentation Parameters, CNCPS Nitrogen Fractions, and Bacterial Community of Moringa oleifera Leaves Silage

The present study investigated the effects of malic acid, sucrose, and their mixture on the fermentation parameters, Cornell Net Carbohydrate and Protein System (CNCPS) nitrogen fractions, and bacterial community of Moringa oleifera leaves (MOL) silages. The trial was divided into four treatments and labeled as CON (control group) and MLA, SUC, and MIX (respectively denoting the addition of 1% malic acid, 1% sucrose, and 1% malic acid + 1% sucrose to the fresh weight basis). The silage packages were opened on the 2nd, 5th, 10th, 20th, and 40th days of ensiling for subsequent determination. Malic acid and sucrose increased the lactic acid content (p < 0.05) and pH value, and the acetic acid contents of MLA and MIX were lower than those in CON (p < 0.05). Compared with sucrose, malic acid had a better capacity to preserve nutrients and inhibit proteolysis, and thus exerted better effects on the CNCPS nitrogen fractions. The results of 16S rRNA showed that the dominant phyla were Firmicutes and Proteobacteria and that the dominant genera were Lactobacillus and Weissella. With the application of silage additives and the processing of fermentation, there was a remarkable change in the composition and function of the bacterial community. The variation of the fermentation parameters and CNCPS nitrogen fractions in the MOL silages caused by malic acid and sucrose might be attributed to the dynamic and dramatic changes of the bacterial community.


Analyses of the Fermentation Parameter and Chemical Composition
Approximately 2 g of fresh sample was maintained at −80 • C from the analyses of the bacterial community. Then, 10 g of fresh sample was mixed with 90 mL of distilled water and stored in a 4 • C refrigerator for 24 h. After filtration using a sterilized fourlayer gauze, the filtrate was collected for the determination of the fermentation parameters. Thereinto, pH was measured via a glass electrode pH meter (FE28-Standard, Mettler Toledo, Switzerland). Lactic acid (LA) and volatile fatty acids (VFAs) were tested according to the methods described by Rumsey et al. [19]. Ammonia nitrogen (AN) was determined using phenol-hypochlorite colorimetry [20].
The rest of the fresh sample was dried at 65 • C for 48 h and grounded into powder for the determination of the chemical composition. Briefly, CP and WSC were tested according to the Association of Official Analytical Chemists (AOAC) [21]; NDF and ADF were tested according to Van Soest et al. [22]. True protein (TP), non-protein nitrogen (NPN), neutral detergent insoluble crude protein (NDICP), and acid detergent insoluble crude protein (ADICP) were measured according to the procedure of Licitra et al. [23]. Soluble crude protein (SCP) was measured on the basis of the method described by Yang et al. [24].
The calculation of the CNCPS nitrogen fractions was performed according to previous researches [14,16,17].

Analyses of the Bacterial Community
The total genomic DNA was extracted via the DNeasy Power Soil Kit (QIAGEN, Inc., Venlo, The Netherlands), and the purity, concentration, and integrity of the DNA samples were determined. Thereafter, the 16S rRNA V3-V4 regions of genomic DNA was amplified via Pyrobest DNA Polymerase (TaKaRa, DR500A) with the primer pairs of 338F (5 -ACTCCTACGGGAGGCAGCA-3 ) and 806R (5 -GACTACHVGGGTATCTAATCC-3 ). After amplification, Agencourt AMPure Beads (Beckman Coulter, Indianapolis, IN) was adopted for the purification of the PCR products, and the PicoGreen dsDNA Assay Kit (Invitrogen, Carlsbad, CA, USA) was used for quantification. The equimolar and pairedend sequencing (PE250) was performed on the Illumina Novaseq 6000 platform (Personal Biotechnology Co., Ltd., Shanghai, China).
After high-throughput sequencing, QIIME (V 1.8.0) was adopted for the processing of the sequenced data. After the filtration of chimera and low-quality sequences described by previous reports [25,26], the effective sequences were obtained and clustered into operational taxonomic units (OTUs) via UCLUST with a 97% similarity threshold [27]. A representative sequence from each OTU was selected for further taxonomic classification conducted via the Basic Local Alignment Search Tool (BLAST). Subsequently, an OTU table was generated [28].
The bioinformatics analyses of the bacterial community were mainly performed using the QIIME and R softwares (V 4.0.0). In detail, α-diversity was analyzed via the alpha_diversity.py script in QIIME; β-diversity was analyzed by a vegan package in the R software [29]. The functions of microflora were predicted on the basis of the PICRUSt database.

Statistical Analysis
The test data of the fermentation parameters, chemical compositions, CNCPS nitrogen fractions, and α-diversity indices were analyzed using a two-way ANOVA method in the SPSS 25.0 software. The model for data processing was: Y ij = µ + D i + A j + (D * A) ij + ε ij , where, Y ij is the dependent variable, µ is the overall mean, D i is the effect of ensiling days, A j is the effect of different silage additives; (D * A) ij is the interaction effect of ensiling days and silage additives, ε ij is the random residual error. The LSD method was adopted for multiple comparisons, and p < 0.05 indicated a statistical significance. A Pearson correlation coefficient was adopted for the correlation analysis, with p < 0.05 indicating relevance [30].

Fermentation Parameters of MOL Silages
The fermentation parameters affected by malic acid, sucrose, and their combination are shown in Table 1. The pH values of the MOLs in the four treatments decreased dramatically at the beginning of fermentation (p < 0.05); however, the pH values in CON, MLA, and SUC recovered from the 10th day (p < 0.05). Compared with that in CON, the pH values in MLA and MIX were significantly lower during the ensiling period (p < 0.05). With fermentation, the contents of LA and AN increased obviously (p < 0.05). Specifically, the LA contents of MLA, SUC, and MIX were significantly higher than that in CON (p < 0.05), whilst the AN content of MLA and MIX were significantly lower than that in CON (p < 0.05). The contents of AA increased in the first 10 days of ensiling (p < 0.05). Relative to the content in CON, the AA in MLA and MIX decreased (p < 0.05). Meanwhile, the WSC contents of CON and SUC decreased during the whole fermentation stage (p < 0.05); in MLA and MIX, the reduction of WSC contents occurred in the first 10 days (p < 0.05). Toward the end of fermentation, the WSC content of MLA was significantly higher than that of CON (p < 0.05).

Chemical Compositions and Nitrogen Fractions of MOL Silage
The results of the chemical compositions are displayed in Table 2. The DM contents in the four treatments changed significantly with fermentation (p < 0.05), but no obvious tendency was obtained. However, the DM content in CON was always the minimum compared to those in the other three treatments (p < 0.05). The contents of NDF and ADF showed a downtrend as the fermentation proceeded (p < 0.05). For the first 20 days, NDF in SUC was always lower than that in CON (p < 0.05), but no significant difference was found amongst the four treatments on the last day of ensiling. For ADF, the content in SUC was consistently the minimum (p < 0.05). The contents of CP increased with the processing of fermentation (p < 0.05), and no significant difference was observed on the last day (p > 0.05).
The nitrogen fractions of MOL showed dramatic changes during fermentation (Tables 3 and 4). The proportions of TP, NDICP, ADICP (PC), PB2, and PB3 showed a descending tendency (p < 0.05). Moreover, the ratios of these fractions in MLA and MIX were significantly higher than that in CON (p < 0.05). However, NPN (PA), SCP, and PB1 showed opposite tendencies. With fermentation, the ratio of the three fractions increased gradually (p < 0.05) and was obviously lower in MLA and MIX than in CON (p < 0.05).

Diversity of Bacterial Community in MOL Silage
In the present study, α-diversity, including the observed species. The Chao1 index, Shannon index, Simpson index, and Pielou evenness were studied. As shown in Table 5, with fermentation, no regular tendency was obtained. However, on the 5th, 10th, 20th, and 40th day of ensiling, all the five indices in MLA were the minimum amongst the four treatments (p < 0.05). The observed species and Chao1 index in MIX were also significantly lower than those in CON during the whole ensiling stage (p < 0.05). In addition, no noticeable difference was noted between CON and SUC as the ensiling proceeded (p > 0.05).     For β-diversity, a principal coordinates analysis (PCoA) was performed, and the results are shown in Figure 1. On the 2nd day of ensiling, no obvious separation was obtained among the four treatments; however, at the 5th, 10th, 20th, and 40th day of fermentation, a remarkable separation was moted between CON and the other three treatments. In addition, significant differences were observed between MLA and SUC, whilst the differences between MLA and MIX on the 5th, 10th, and 20th day were not particularly obvious.

Composition of the Bacterial Community in MOL Silage
During the 40-day ensiling period, Firmicutes was always the most dominan lum, accounting for more than 90% of the bacterial community on the 5th, 10th, an day ( Figure 2a). As the fermentation proceeded, other phyla increased and include teobacteria and Bacteroidetes, especially in CON and SUC. On the 40th day, the r abundance of Proteobacteria and Bacteroidetes in CON and SUC was higher than MLA. At the genera level (Figure 2b), the most dominant phylum on the 2nd da Weissella, replaced by Lactobacillus on the 5th, 10th, 20th, and 40th day. On the 2n 10th, and 20th day of ensiling, Lactobacillus in CON and SUC was lower than that in and MIX. On the last day, other genera, such as Inhella, unclassified Burkholde Env.OPS_17, Caulobacter, and Ideonella, were obtained; however, compared with MLA, the relative abundance of Lactobacillus showed a decline. At the species lev predominant bacteria were L. paralimentarius, L. brevis, L. spicheri, L. namurensis, and sphingobium capsulatum (Figure 2c). The amount of L. brevis in MA was particular The relative abundance of L. paralimentarius and L. spicheri was high in MIX, whilst L. namurensis was high in CON.

Composition of the Bacterial Community in MOL Silage
During the 40-day ensiling period, Firmicutes was always the most dominant phylum, accounting for more than 90% of the bacterial community on the 5th, 10th, and 20th day ( Figure 2a). As the fermentation proceeded, other phyla increased and included Proteobacteria and Bacteroidetes, especially in CON and SUC. On the 40th day, the relative abundance of Proteobacteria and Bacteroidetes in CON and SUC was higher than that in MLA. At the genera level (Figure 2b), the most dominant phylum on the 2nd day was Weissella, replaced by Lactobacillus on the 5th, 10th, 20th, and 40th day. On the 2nd, 5th, 10th, and 20th day of ensiling, Lactobacillus in CON and SUC was lower than that in MLA and MIX. On the last day, other genera, such as Inhella, unclassified Burkholderiaceae, Env.OPS_17, Caulobacter, and Ideonella, were obtained; however, compared with that in MLA, the relative abundance of Lactobacillus showed a decline. At the species level, the predominant bacteria were L. paralimentarius, L. brevis, L. spicheri, L. namurensis, and Novosphingobium capsulatum (Figure 2c). The amount of L. brevis in MA was particularly low. The relative abundance of L. paralimentarius and L. spicheri was high in MIX, whilst that of L. namurensis was high in CON.

Predicted Functions and Pathways of Bacterial Community in MOL Silages
As shown in Figures 3a and S1, the top five predicted functions were DNA helicase, DNA-directed DNA polymerase, Histidine kinase, Non-specific serine/threonine protein kinase, and NADH: ubiquinone reductase (H(+)-translocating). For the predicted pathways (Figures 3b and S2), the top five were Aerobic respiration I (cytochrome c), Acetylene degradation, Superpathway of pyrimidine nucleobases salvage, Peptidoglycan maturation (meso-diaminopimelate containing), and Superpathway of adenosine nucleotides de novo biosynthesis I. The heatmaps of the predicted functions and pathways ( Figure  3a,b) implied that the addition of malic acid might have affected the dominant functions and pathways.

Predicted Functions and Pathways of Bacterial Community in MOL Silages
As shown in Figures 3a and S1, the top five predicted functions were DNA helicase, DNA-directed DNA polymerase, Histidine kinase, Non-specific serine/threonine protein kinase, and NADH: ubiquinone reductase (H(+)-translocating). For the predicted pathways (Figures 3b and S2), the top five were Aerobic respiration I (cytochrome c), Acetylene degradation, Superpathway of pyrimidine nucleobases salvage, Peptidoglycan maturation (meso-diaminopimelate containing), and Superpathway of adenosine nucleotides de novo biosynthesis I. The heatmaps of the predicted functions and pathways (Figure 3a,b) implied that the addition of malic acid might have affected the dominant functions and pathways. The PCoA analyses of the predicted functions and pathways were performed. The results showed that the individuals in MLA and MIX always flocked together and that these plots were isolated from the plots in CON and SUC (Figure 4a, b).  The PCoA analyses of the predicted functions and pathways were performed. The results showed that the individuals in MLA and MIX always flocked together and that these plots were isolated from the plots in CON and SUC (Figure 4a,b). The PCoA analyses of the predicted functions and pathways were performed. The results showed that the individuals in MLA and MIX always flocked together and that these plots were isolated from the plots in CON and SUC (Figure 4a, b).

Correlation Analysis
In the present study, the correlation analysis of various indicators and dominant bacteria was performed ( Figure 5 and Table S1)

Correlation Analysis
In the present study, the correlation analysis of various indicators and dominant bacteria was performed ( Figure 5 and Table S1)  In addition, we analyzed the correlation between dominant bacteria and dominant functions ( Figure 5 and Table S2  In addition, we analyzed the correlation between dominant bacteria and dominant functions ( Figure 5 and Table S2

Discussion
In this study, the DM content of MOL was 243.62 g/kg FM, which was lower than the ideal content of 30-35% DM reported by Guyader et al. [31]. High moisture increases the risk of spoilage and nutrient loss for silages. Therefore, the use of silage additives is necessary [20]. The CP content of MOL was 279.32 g/kg DM, which was higher than those in the previous studies by 21.5%, 26.0%, and 25.3% [20,32,33]. The differences might be attributed to variations in cultivation, harvest time, planting environment, and climate conditions. WSC is the main fermentation substrate, and its content in this study was 44.77 g/kg DM, which was lower than the minimum requirement of 60-70 g/kg DM for high-quality silages [33]. In sum, MOL is an excellent forage, with high protein and low NDF and ADF. However, the characteristics of high moisture and insufficient WSC hinder the process of ensiling without silage additives.
PH value is an important indicator to evaluate the quality of silages. Generally, when the pH value is less than 4.2, the silage could be regarded as well fermented [34]. In this study, the pH values in MLA and MIX were significantly lower than those in CON and SUC, thereby indicating that malic acid possibly played vital roles. Malic acid could reduce pH rapidly because of its acidity. Moreover, the application of malic acid could promote the growth of LAB, thereby accelerating the production of LA and reducing pH [35]. In addition, sucrose solved the problem of the insufficient fermentation substrate in MOL and promoted LA production. On the 20th day of ensiling, the pH values in CON and SUC presented a dramatic upswing possibly because of the accumulation of AN and other alkaline substances caused by proteolysis [9]. AN reflects the degree of peptide hydrolysis, as well as the deamination of amino acids or peptides [36]. The results of the study indicated that the addition of malic acid reduced the pH value of the silage, thus inhibiting the activity of protease. According to another research, AN production is caused by clostridial fermentation, which might be inhibited by the low pH value created by malic acid [37]. The differences in AA contents in the current work might be associated with the fermentation type. The application of malic acid was speculated to be capable of reducing the abundance of heterofermentative LAB. In addition, propionic acid and butyric acid were undetected during the fermentation period. Butyric acid is the product of secondary fermentation caused by clostridium. It is undesirable and reflects nutritional damage in the forage [32]. This indicates that the preservation of silage is relatively good. Like sucrose, some organic acids (such as citric acid, malic acid, succinic acid, and fumaric acid) and their salts can provide fermentation substrates for silage. This feature might explain why the WSC content in MLA was significantly higher than that in CON during the whole ensiling period [35]. In sum, malic acid and sucrose exerted a positive effect on the fermentation quality of MOL silages, but the addition of the former might be more effective.
The DM of silage was mostly consumed by the metabolism of aerobic microorganisms such as Clostridia and yeasts. In general, the lost fractions were mainly digestible organic acids and carbohydrates [38,39]. According to previous studies, malic acid and sucrose could increase the DM content after ensiling [14,15]; however, their combination has not been widely explored. On the last day of ensiling in the current work, the DM contents in MLA and SUC and that in CON showed no significant difference, whilst those in CON and MIX showed obvious differences. This result indicated that relative to malic acid or sucrose alone, their mixture had positive effects on the DM content and reduced DM loss. During the fermentation period, the NDF and ADF contents in SUC were always at a minimum; on the last day, the ADF contents in SUC and MIX were significantly lower than that in CON. This result indicated that sucrose played a crucial role in fiber degradation, which could be attributed to the promotion of acid hydrolysis in plant cells [40]. As in our research, Li et al. reported that the addition of glucose, molasses, sucrose, and their mixtures reduces the NDF and ADF contents of king grass silages [41]; Zhao et al. found that molasses decrease the NDF and ADF contents of rice straw silages [42].
On the last day of ensiling in the current work, no significant difference was found amongst the four treatments, whilst tremendous changes occurred in the protein fractions. The purpose of silages is to extend the storage life of the forage and reduce nutritive losses. Ensiling is a dynamic process accompanied by enzyme and microbial activities [34]. Here, one of the most important biochemical reactions is proteolysis, which transforms proteins into NPN (such as small peptides, free amino acid, and AN) under the activities of microorganisms and proteases [43]. The conversion of TP to NPN (PA fraction) reduces the N utilization in ruminants, further increasing urinary and fecal N losses and causing environmental pollution [44]. Thus, the application of malic acid may have a good effect in terms of nutrient preservation. We can speculate that the addition of malic acid reduces the pH value of silages, thus leading to the decreased activity or even inactivation in protease [35]. SOP consists of PA fraction and PB1 fraction, with PB1 defined as a rapidly degradable true protein, which is lost quickly in the rumen and is difficult to be further utilized. NDICP is composed of a PB3 fraction and a PC fraction (ADICP). NDICP and ADICP belong to proteins bonded with plant cell walls. Relative to NDICP, ADICP cannot be digested and utilized by ruminants [23]. In this study, the proportions of NDICP, PB3, and PC in MLA and MIX were higher than those in CON and SUC. According to our speculation, the enzymes which break chemical bonds that are linked to proteins and structural carbohydrates were inhibited under the acidic conditions caused by malic acid, resulting in the high content of bonding protein [15]. However, because the related research is limited, this topic requires further research. The PB2 fraction is a part of TP for degradation. The increase of PB2 might indicate the promotion of protein digestion in the rumen [45]. Moreover, PB2 + PB3 can be regarded as a ruminal bypass protein or rumen undegradable protein, and its increase means that a large amount of high-quality amino acids can be provided for absorption [46]. In conclusion, malic acid has good properties for avoiding nutrition loss in the forage and could increase the proportion of available proteins. However, it also results in a significant increase in the proportion of undesirable proteins.
Microorganisms play important roles in the ensiling process. Therefore, to make highquality silages, the dynamic changes of bacterial communities during fermentation should be clearly understood. For the past few years, 16S rRNA high-throughput sequencing technology has been widely used in silage. Changes in community structure are always quantified through a range of nonparametric ecological indices [47]. On the basis of such indices, α-diversity is adopted to measure the richness, diversity, and evenness of species in bacterial communities. In the present study, we analyzed the observed species, the Shannon index, the Simpson index, the Chao1 index, and the Pielou evenness in MOL silages. The observed species, Shannon index, and Simpson index reflect the diversity of the bacterial community. The Chao1 index and Pielou evenness represent the richness and evenness, respectively. The α-diversity indices changed significantly at different ensiling stages, thus indicating that silage is a dynamic microbial reaction. The composition and function of microflora showed great differences as well [48]. The results of this study showed that the addition of malic acid remarkably reduced the five α-diversity indices. Similarly, Zi et al. found that silage king grass silage treated with citric acid had a lower OTU, Chao1 index, ACE index, and Shannon index, possibly because the addition of organic acids inhibits the growth of harmful bacteria [49]. The result of β-diversity reflected the distinction of the bacterial community in each individual or treatment. On the 5th, 10th, 20th, and 40th day of ensiling, CON presented great distinctions from the other three treatments, especially MLA and MIX. In sum, the results of the bacterial diversity corresponded to the fermentation parameters and nitrogen fractions, thus indicating that malic acid might play important roles in regulating bacterial community and thereby affecting silage quality.
The composition of microflora was associated with silage quality. At the phylum level, the dominant bacteria were Firmicutes and Proteobacteria, similar to those reported by Wang et al. [50]. Firmicutes can secrete various cellulases, lipases, and proteases and is able to survive in an anaerobic and low pH environment [51]. In addition, almost all LABs belong to Firmicutes. In this study, Proteobacteria were positively correlated with AN, thereby indicating that the accumulation of AN increased the pH value and provided favorable conditions for the growth of Proteobacteria. A positive correlation was also noted between Proteobacteria and PA (NPN), and it suggested the involvement of Proteobacteria in proteolysis. On the last day of fermentation, MLA had the lowest abundance of Proteobacteria; hence, malic acid might have prevented proteolysis by inhibiting the growth of Proteobacteria. The relative abundance of other phyla was inferior, but they also played important roles in the ensiling process. Bacteroidetes was positively correlated with PA, LA, and AN and negatively correlated with NDF, ADF, and WSC, probably because Bacteroidetes is mainly involved in the hydrolysis of complex macromolecules [52]. In addition, Sa et al. reported that Chloroflexi is associated with acid production [53], but our results showed no correlation between Chloroflexi and organic acid (LA and AA). We could speculate that Chloroflexi only accounted for 0.01% of the bacterial community and that the effects of Chloroflexi on fermentation were negligible. For the other phyla, Acidobacteria, Actinobacteria, Gemmatimonadetes, and Verrucomicrobia were reported by He et al. [20], whilst Deinococcus-Thermus was reported by Dong et al. [54]; however, the roles of these phyla in silages remain unreported. Although we found that these phyla had a correlation with some fermentation parameters and chemical compositions via correlation analyses, the results require further verification and exploration with regard to the mechanism.
At the genera level, the dominant bacteria were Lactobacillus and Weissella. On the 2nd day of ensiling, Weissella accounted for 80% and thus dominated the bacterial community. As fermentation proceeded, the dominance of Weissella was gradually replaced by that of Lactobacillus. According to a previous study, Weissella colonizes at the early stage of ensiling. With the accumulation of organic acid and the decline of the pH value, the growth of Weissella is inhibited [55]. On the contrary, a low pH condition favors the growth of Lactobacillus. At the initial stage of fermentation, the growth of aerobic bacteria and the respiration of plant cells lead to a high consumption of oxygen. Then, Lactobacillus grows rapidly when the environment is completely anaerobic [56]. Notably, the pH value in this work had a positive correlation with Weissella and a negative correlation with Lactobacillus. The results further indicated that the regulation of malic acid on the two genera might be related to the pH value. Leuconostoc is a type of lactate-producing bacteria, and its relative abundance decreased with the processing of ensiling and showed a negative correlation with LA. As a microbial additive generally recognized as safe, Leuconostoc is popular in the food fermentation industry and is able to utilize a variety carbon sources, including arabinose, fructose, galactose, glucose, lactose, and sucrose [57]. However, Leuconostoc was positively correlated with the pH value and showed a downtrend with the decrease of the pH value. Actually, Leuconostoc is regarded as an early colonizer and has poor tolerance to low pH conditions. As fermentation progresses, Leuconostoc is gradually replaced by acid-tolerant LABs, such as Lactobacillus [58]. As for other genera, Pannonibacter and Inhella have been reported to exert the effect of biological denitrification and reduce nitrate to nitrite [59,60]; Caulobacter plays functional roles in enhancing plant growth by synthesizing siderophores, solubilizing phosphate and producing indole-3-acetic acid [61]. However, the roles of these genera in silage are still undefined and thus require further research.
At the species level, we found that the relative abundance of L. brevis was lower in MLA and MIX and that L. brevis was positively correlated with AA. As a type of heterofermentative microorganism, L. brevis could convert 1 mol glucose to 1 mol LA, 1mol CO 2 , and 1 mol AA [62]. L. parabrevis was also found to have a positive relation with AA. Mu et al. reported that the increase of LA and AA in silages was closely related to L. parabrevis [63]. In addition, Le. fallax was found to proliferate at the heterofermentative stage [64]. Compared with homofermentative fermentation, heterofermentative fermentation is generally characterized by a higher pH, a high AA, and a lower LA content [65]. In the present study, both L. parabrevis and Le. fallax were inhibited by malic acid, and malic acid regulated the fermentation type by altering the bacterial composition. L. brevis showed a positive correla-tion with AN and PA, but no evidence that L. brevis was involved in proteolysis. On the contrary, L. brevis has been reported to be a potential silage inoculant capable of mitigating nutrient losses [66]. The proteolysis in CON and MIX might be attributed to the relative high pH caused by heterofermentative fermentation. Some undesirable microorganisms, such as Clostridium, were rapidly inhibited when the pH was lower than 4. However, when the pH rose again to 4.5, the activities of the harmful microbes recovered [39]. The rise of pH in CON and SUC at the late stage of ensiling might have accelerated proteolysis to a certain extent. For the other species, L. spicheri is considered as a probiotic strain and could be adopted in the production of health care food [67]. L. paralimentarius was the most abundant species, but its role in ensiling remains unknown. In the present study, we observed that L. paralimentarius was positively correlated with CP and negatively correlated with pH and PB1, thereby suggesting that this microorganism has the potential to be a silage additive.
PICRUSt is a computational approach for predicting the function and pathway compositions of metagenomes according to the databases of reference genomes and marker gene data [68]. In this work, the application of malic acid increased the abundance of Histidine kinase, NADH: ubiquinone reductase (H(+)-translocating), N-acetylmuramoyl-Lalanine amidase, and Serine-type D-Ala-D-Ala carboxypeptidase, etc.; while it decreased the abundance of DNA-directed DNA polymerase, Glutaminyl-tRNA synthase (glutaminehydrolyzing), DNA helicase, and Alcohol dehydrogenase, etc. In addition, the PCoA analyses of the predicted functions and pathways corroborated the viewpoint that malic acid exerts great effects on the function of bacterial communities. These results could be ascribed to the tremendous variation of some functional bacteria caused by silage additives [39]; however, the mechanisms remain unclear. Moreover, we performed correlation analyses of bacteria and predicted functions, and the results may provide potential biomarkers for the regulation of silage fermentation. In sum, we hypothesized that the regulation of silage quality by malic acid might be related to the composition and function of the microflora.

Conclusions
In this work, malic and sucrose were able to enhance the fermentation quality of MOL silages. Malic acid regulated the fermentation type by altering the bacterial community. Malic acid also showed a protective effect on the nutrients by inhibiting proteolysis, thereby improving the CNCPS nitrogen composition of MOL silages. Relative to sucrose, malic acid had a remarkable effect on the composition and function of the bacterial community, thereby improving the silage quality. In addition, L. paralimentarius, which has not been reported in silages, was found to have the potential to be developed as a silage inoculant. As far as silage preparation is concerned, different additives have different advantages. Therefore, maximizing their respective advantages and forming synergistic effects are the key to the efficient utilization of silages. These steps are worth studying further.
Supplementary Materials: The following are available online at https://www.mdpi.com/article/ 10.3390/microorganisms9102102/s1, Figure S1: The top 20 predicted functions of the bacterial communities analyzed via PICRUSt. CON, control group; MLA, 1% malic acid addition on the fresh weight (FW) basis; SUC, 1% sucrose addition on the FW basis; MIX, 1% malic acid and 1% sucrose addition on the FW basis; Figure S2: The top 20 predicted pathways of the bacterial communities analyzed via PICRUSt. CON, control group; MLA, 1% malic acid addition on the fresh weight (FW) basis; SUC, 1% sucrose addition on the FW basis; MIX, 1% malic acid and 1% sucrose addition on the FW basis; Table S1: Correlation analysis between various indicators and dominant bacteria (phyla, genera, and species level); Table S2: Correlation analysis between dominant predicted functions and dominant bacteria (phyla, genera, and species); Table S3: Correlation analysis between dominant predicted pathways and dominant bacteria (phyla, genera, and species). Data Availability Statement: The sequences in this study were submitted to the Sequence Read Archive (SRA), and a BioProject number PRJNA748826 was obtained.