Lactobacillus plantarum inoculation delays decomposition of alfalfa silage under adverse ensiling conditions by regulating bacterial community composition

Background The objective of this study was to investigate the mechanism of Lactobacillus plantarum ( L. plantarum ) involved in improving fermentation quality of naturally ensiled alfalfa under poor conditions. Results High-moisture wilted alfalfa was ensiled without inoculants (CK) or with inoculation of two L. plantarum additives (LPI and LPII). The pH and fermentation products of silage were investigated, and the bacterial community compositions were analyzed. The L. plantarum inoculants significantly enhanced the lactic acid fermentation in terms of promotions in pH decline, lactic acid accumulation, and Lactobacillus abundance for both periods. At 90 d, silage in CK exhibited a high pH, a loss in dry matter, and a high concentration of ammoniacal nitrogen. The inoculations of L. plantarum significantly inhibited the growth of Clostridia, and reduced ammoniacal nitrogen concentration in silage ( P < 0.05). Inoculation with improved inhibited growth of microorganisms, and further delayed decomposition

harvested in spring, summer, and autumn in most areas in China other than the north. At the end of spring or early summer, mid-eastern and southern China have abundant rainwater, and the alfalfa ensiled for over two or three months will reach high temperatures during summer. Sudden rainfall and high air humidity may cause extended wilting periods, which affect characteristics of the ensiling material. Sustained high temperatures during the ensiling process may also affect the stability of fermentation.
The microbial community plays an important role in the fermentation process and is likely to be impacted by multiple factors, such as ensiling conditions, inoculants, and epiphytic microorganisms of the fresh forage (Guan et al., 2018;Kasmaei et al., 2017). The development of PCR-based techniques enables us to define the microbial communities more accurately, and a larger variation in the microbiota in alfalfa silages than in cereal silages has been documented (McAllister et al., 2018).
Inoculations of Lactobacillus plantarum could enhance the acidification of silage and adaption to a low pH environment, which contributes to the inhibition of competing microorganisms and effectively improves silage quality. Thus, L. plantarum is the most commonly used bacterial inoculant in forage ensiling studies (Oliveira et al., 2017). During ensiling, L. plantarum enhances LA fermentation and survives and adapts to the silage environment (Guo et al., 2018;Ogunade et al., 2018;Yang et al., 2019). However, the fermentation quality of the silage varies owing to different factors, such as BC, water activity, and presence of epiphytic microbes. The latter can differ owing to several factors, such as forage type, climatic conditions, and irrigation level. Adverse ensiling conditions, such as extremely high temperature, high humidity, and soil incorporation, can result in the poor fermentation of alfalfa silage. Under good conditions, silage without L. plantarum inoculation can produce a good fermentation quality, with a decline in pH, abundant LA accumulation, and sufficient LAB counts.
However, few studies have examined the effects of L. plantarum on alfalfa silage fermentation quality when poor fermentation conditions occur.
The aim of this study was to investigate the mechanism of L. plantarum in improving fermentation quality of naturally ensiled alfalfa under the poor conditions that commonly occur in spring and summer ensiling in China. Yellow River-irrigated alfalfa was harvested during the high precipitation season when the relative humidity exceeded 80%, and ensiled for 90 d after a 1-d wilting with or without L. plantarum inoculation. Particular attention was paid to bacterial communities, fermentation properties, and their interactions in silage.

Methods 2.1. Materials and silage preparation
Fresh alfalfa was harvested at the early bloom stage in Zhengzhou, Henan Province (temperate monsoon climate, 34.76°N, 113.65°E, altitude 110.4 m above sea level), and wilted for 24 h. The wilted material was chopped into 1-2-cm lengths. For inoculant preparation, L. plantarum YX was isolated from the Yaxin alfalfa ensiling additive (Yaxin Biotechnology Co., Ltd., Taiwan, China) and L. plantarum A345, an alfalfa silage epiphytic strain, was isolated from Shanxi, China. The L. plantarum strains were cultured in MRS medium at 30 °C for 12 h. Then, the culture was centrifuged at 12,000 g for 10 min at 4 °C. The precipitate was mixed with distilled water to an OD 600 of 0.8.

Analysis of fermentation products
The AOAC Method 934.01 (1990) was used for DM determination. Subsamples were dried in an oven at 65 °C for 48 h and pulverized to pass through a 1-mm screen using a laboratory knife mill (FW100, Taisite Instrument Co. Ltd., Tianjin, China). The WSCs were measured using anthrone colorimetry (Murphy, 1958).
Other subsamples of silage (10 g) were diluted with 90 ml of distilled water and filtered through a 0.45-µm membrane. The pH was measured using a pH meter. The organic acid contents were determined using high-performance liquid chromatography (Waters Alliance e2695, Waters, MA, USA).
Carbomix H-NP 10:8% (7.8 mm × 300 mm × 10 µm) was used as the stationary phase, and the column temperature was maintained at 55 °C. The injected sample volume was 10 µl. The mobile phase composition was 0.0254% sulfuric acid. The mobile phase was filtered through a 0.45-µm pore size, 47-mm diameter nylon membrane and degassed before use. The flow rate of the mobile phase was 0.6 ml/min. The detection wavelength for samples was 214 nm using a UV detector. The ammoniacal nitrogen (NH 3 -N) level was determined using Berthelot colorimetry (Broderick and Kang, 1980). The BC was determined using the hydrochloric acid-sodium hydroxide method (Playne et al., 2010). The fermentation coefficient of alfalfa silage was predicted using the formula of Addah et al. (2011). 2.3. Bacterial community analyses Total DNAs were extracted using a Bacterial DNA Kit D3350-02 (Omega Biotek, Norcross, GA, USA).
The purity levels and concentrations of DNAs were evaluated by 1% agarose gel electrophoresis. The PCR amplifications of the V3-V4 regions of the bacterial 16SrDNA gene were performed using Primer F (Illumina adapter sequence 1 + CCTACGGGNGGCWGCAG) and Primer R (Illumina adapter sequence 2 + GACTACHVGGGTATCTAATCC). The PCR products were extracted from a 2% agarose gel. The amplicon sequencing of 16SrDNA was conducted using the Miseq platform (Genesky Bio-Tech Co. Ltd., Shanghai, China) after the purification and quantification of the PCR products. All the raw reads were checked using FLASH2, and low quality sequences (quality scores below 20) were discarded according to the QIIME quality control process (version 1.7.0). Operational taxonomic units were clustered using Uparse (Uparse v7.0.1001, http://www.drive5.com/uparse/). The analysis of taxonomy assignment of representative sequences was performed using the Ribosome Database Project (Cole et al., 2009).

Statistical analyses
The statistical analysis of the fermentation products was performed using IBM SPSS version 21.0 (SPSS Inc., Chicago, IL, USA). The effects of different treatments were evaluated by one-way analysis of variance, with Duncan's multiple range tests. The alpha diversity of the bacterial communities was calculated using mothur (version 1.9.1). The beta diversity analyses and correlation analyses between bacterial compositions and environmental factors were performed using R software (version 2.15.3).

Characteristics of ensiling material before ensiling
The wilted alfalfa had pH of 6.06, DM of 266 g/kg FW, WSC concentration of 17.56 g/kg DM, and a high BC of 460 mEq/kg DM. Organic acid and NH 3 -N were not detected.
3.2. Effects of L. plantarum on fermentation properties of alfalfa silage Organic acid and NH 3 -N contents, as well as the pH level, of alfalfa silage inoculated with or without L. plantarum strains are shown in Table 1. The L. plantarum inoculants effectively accelerated the LA fermentation and acetic acid (AA) accumulation, resulting in a lower pH compared with the CK group (P < 0.001) at 30 d; and the enhanced LA accumulation was significantly greater in the LPII group (P < 0.001). For 90-d silage, increase in pH level and NH 3 -N concentration, reduction in LA and WSC concentration, as well as formation of propionic acid (PA), occurred compared with 30-d silage.
Significantly higher pH (P < 0.001) and NH 3 -N accumulation (P = 0.001), as well as a significant loss in DM (P = 0.017) were apparent in the CK compared with the inoculated groups. Silage in the LPII group had better fermentation quality compared with other groups, indicated by a relatively low pH, high LA and AA concentrations, and inhibition of NH 3 -N formation. The significantly lower WSC residual (P = 0.007) also indicated a more abundant fermentation in the LPII group.

Effects of L. plantarum on bacterial community composition
High-throughput analysis was used to determine the bacterial diversity in alfalfa silage, and the valid sequences were clustered into 552 operational taxonomic units based on a 97% sequence identity.
Richness and diversity indices of the bacterial communities, represented by observed species, the Chao1 and the Shannon indexes, respectively, are shown in Fig. 1A-C. The richness of the bacterial community decreased after ensiling for 30 d compared with the ensiling material, and then increased at 90 d (P = 0.005). The diversity of the bacterial community decreased after ensiling (P = 0.006). The diversity of the bacterial community in silage in LPII slightly and non-significantly increased at 90 d compared with 30-d silage (P = 0.1).
The principal coordinate analysis (PCoA) clearly reflected the variation within the bacterial community ( Fig. 2). The clear separation between bacterial communities of the ensiling material and alfalfa silage indicated a shift after ensiling. Divisions in the plots representing silage with and without inoculants for both periods indicated that the distribution of the bacterial community was shifted by L. plantarum inoculations. The distribution of the bacterial communities among the three replications within LPII was more similar compared to those within CK and LPI at 30 d. Although the Shannon indexes were similar between 30-d and 90-d silage (Fig. 1C), divisions in the plots representing silage for the two periods indicated a variation within the bacterial community.
LefSe was performed to further explore the variations in the bacterial communities among the treatments (Fig. 4). Significantly higher abundances of species of Enterobacter, Lactococcus, and Enterococcus were observed in CK at 30 d. Lactobacillus species grew well in LPII in 30-d silage. At 90 d, the relative abundance of some species belonging to spoilage genus Clostridium and pathogenic genus Listeria were significantly higher in the CK group. Although silage in LPII effectively inhibited the growth of Clostridium_XlVa (Fig. 3B) (13.42% in CK, 6.67% in LPI, and 0.67% in LPII, respectively), it showed a relatively poorer inhibition against some Clostridia genera with low abundance.

Correlation analysis of the bacterial community with fermentation products
Mantel tests were performed to reveal the relationships between bacterial community compositions at different taxon levels and silage properties ( Table 2). Results indicated significant correlations between fermentation properties and the bacterial community composition (P < 0.05) at different taxon levels, except for pH value and bacterial community composition at phylum level. Spearman's correlation analysis indicated that the pH value and LA concentration had significant correlations with richness and diversity indices of the bacterial communities (P < 0.05). A strong negative correlation was found between AA concentration and the Shannon index (P < 0.05). Meanwhile, the number of observed species and the Chao1 index of the bacterial community showed positive correlations with PA and NH 3 -N concentrations, and negative correlations with WSC concentration (P < 0.05).
Spearman's correlations further illustrated the relationships between bacterial genera and silage properties for different periods ( Fig. 5A and B). Lactobacillus was positively correlated with LA (r = 0.92 at 30 d and 0.87 at 90 d), and negatively correlated with pH (r = − 0.92 at 30 d and − 0.88 at 90 d ) for both periods. Positive correlations were found between NH 3 -N and two genera at 30 d (P < 0.05), Rosenbergiella and Bifidobacterium. The NH 3 -N was negatively correlated with Nocardioides, Methylobacterium, and Prevotella in 30-d silage (P < 0.05). The NH 3 -N was positively correlated with a series of nitrogen-fermenting genera at 90 d, including Anaerotruncus, Peptoniphilus, and genera belonging to Clostridia (Clostridium_XlVa, Clostridium_XI, and Finegoldia) (P < 0.05). Other genera also showed positive correlations with NH 3 -N at 90 d, including Weissella, Eremococcus, Brevibacillus, Vagococcus, Anaerococcus, Enterococcus, and Propionibacterium (P < 0.05). Nocardioides and Methylobacterium retained their negative correlations with NH 3 -N at 90 d. Another genus belonging to NH 3 -N were also apparent for the acid-producing genus Lactobacillus, nitrate-reducing genera Pantoea and Exiguobacterium, putrescine-fermenting genus Anaerovorax, and other genera, including Planomicrobium, Acinetobacter, Kocuria, and Janibacter, with unknown functions in the ensiling process.
4. Discussion 4.1. Characteristics of wilted alfalfa Adverse conditions during harvesting and ensiling could result in decomposition of silage. In this study, Yellow River-irrigated alfalfa was chosen as the ensiling material, and was harvested at the early bloom stage during the high precipitation season. Although the wilting period was extended compared with previous studies (Li et al., 2018a, b), the DM (266 g/kg FW) was still lower than the recommended content (300-400 g/kg FW). This might due to the high precipitation during the harvest season. Frequent rainfall resulting in high air humidity slowed the moisture loss during wilting. As reported by Agarussi et al. (2019) and Tao et al. (2017), wilting decreased the WSC concentration and crude protein content in fresh alfalfa. The 24-h wilting may have exacerbated the lack of WSC (17.56 g/kg DM) in this study. The WSC of the ensiling material was lower than 50 g/kg DM, which was insufficient for adequate fermentation during ensiling (Ni et al., 2018). The BC was still high after wilting in this study. Frequent rainfall might increase the incorporation of soil in fresh alfalfa during harvest, resulting in an increase in BC (Weinberg and Ashbell, 2003). This might stimulate the epiphytic aerobic microorganisms to remain active for an extended period and so reduce the quantity of WSCs for further LAB fermentation. The low fermentation coefficient (26.9 < 35) also suggested that the material was too hard to ensile naturally (Ke et al., 2015). This might be explained by better adaption of the alfalfa epiphytic inoculant in LPII. Good adaption to the alfalfa ensiling environment of the inoculant aids in improvement of silage quality (Zhao et al., 2020). The AA concentrations increased with L. plantarum inoculations at 30 d. The AA is a promoter of aerobic stability during the ensiling process (Schmidt and Kung, 2010) and an effective inhibitor of fungi (Le Lay et al., 2016). This enhancement in AA accumulation might help with pH decline and reduction in microbial richness at early stages of ensiling. Although inoculation enhanced LA fermentation in the 30-d silage, pH in the silage was still above ideal level (< 4.20). The increase in pH in 90-d silage also suggested that LA fermentation was insufficient to stabilize the ensiled mass.
The insufficient LA fermentation might be due to a lack of available substrate and high BC of the ensiling material.
The LA to AA ratio became much lower at 90 d in all treatments, indicating reduction in LA fermentation in the silage. The PA contents in silage were also out of the acceptable range of 1-10 g/kg DM (Agarussi et al., 2019). The NH 3 -N is considered a representative of amino acid deamination and decarboxylation (Scherer et al., 2015), which generally reduce the nutritional value of silage. The remarkable increase in pH value and NH 3 -N concentration, as well as the loss in DM in CK at 90 d indicate potential decomposition of silage. The delay in decomposition with L. plantarum inoculation was shown by the lower pH and NH 3 -N concentration, higher LA residual, and good DM maintenance. The NH 3 -N concentration in LPII was significantly lower compared with other treatments, indicating better protein maintenance with the inoculant.
Notably, the NH 3 -N concentration increased in LPI (P < 0.001), but the LA accumulation and pH decline were greater, compared with the CK group. Effects of Clostridia and plant proteolytic enzymes may be typical causes of NH 3 -N accumulation (Kung and Shaver, 2001). Tao et al. (2012) reported that most plant proteolytic enzymes in alfalfa silage showed greater activities at pH 5.0-6.0, which could explain the higher NH 3 -N concentration in LPI. Plant proteolytic enzymes tended to be more active at the pH level in LPI (5.60) than in CK (6.10). Although LPI and LPII had similar pH values (5.60 and 5.50, respectively), LPII showed low NH 3 -N formation. This might result from the greater acidogenic capability of the inoculant in LPII in terms of higher LA and AA concentrations, which contributed to an increased inhibition of proteolytic microorganisms.

Effects of L. plantarum on bacterial community of alfalfa silage
Although richness of bacterial community decreased compared with the ensiling material at the early stage of ensiling, the richness flourished at 90 d. Similar trends were also reported by Zheng et al. (2017) in direct-cut alfalfa silage with and without LAB inoculant and sugar and by Keshri et al. (2018) in untreated corn silage, in which richness of the bacterial community decreased at the early stage of ensiling compared with the alfalfa material, and later increased during the ensiling process.
Consistent with the study of Ni (2017a), diversity of the bacterial community decreased after ensiling.
Inoculation slightly reduced diversity of the bacterial community. Although the Shannon indexes were similar between the 30-d and 90-d silage, the PCoA showed variation within the bacterial community.
The clear separation between bacterial communities of the ensiling material and alfalfa silage indicated a shift after ensiling, consistent with previous studies (Ni et al., 2017b;Yang et al., 2019).
Divisions between the plots representing silage with and without inoculants indicated that the distribution of the bacterial community was shifted by L. plantarum inoculations, which could explain the differences in fermentation quality (Ni et al., 2018). The distribution of the bacterial communities among the three replications within LPII was more similar compared with those within LPI for both periods. The more stable bacterial community distribution in LPII might result from its stronger LA fermentation represented by the highest LA concentration (Table 1) (Ni et al., 2018;Pang et al., 2011). Low abundance of these genera was exhibited in the wilted alfalfa before ensiling, and aerobic microorganisms dominated the bacterial community. This was consistent with the results of Tao et al. (2017), in which a long period of wilting decreased the LAB count and promoted growth of aerobes in alfalfa before ensiling.
A series of Clostridia genera grew in the 90-d silage. Clostridia may lead to excessive protein degradation, DM loss, and butyric acid (BA) production in silage, and further promote the growth of other less acid-tolerant spoilage microorganisms (Wang et al., 2019). Thus, they are undesirable in silage. Clostridiales (64.47%) dominated the bacterial community in CK at 90 d at the order level.
Inoculation with L. plantarum effectively inhibited the dominance of Clostridiales (48.27% in LPI and 29.26% in LPII, respectively), which further reduced NH 3 -N formation, maintained the DM (Table 1), and inhibited less acid-tolerant Listeria pathogens in silage (Fig. 4). Although Clostridia is a major producer of BA in silage, presence of BA was not detected in our study. This might result from the activity of BA-utilizing microbes in silage. Flythe and Russell (2004) reported that some Clostridia could produce large amounts of AA apart from BA. This might explain the high AA concentration in CK at 90 d. A high abundance of Garciella, an anaerobic and thermophilic bacterium belonging to Clostridiales (Miranda-Tello et al., 2003), was apparent in the 90-d silage. A high abundance level of Garciella was also reported by Zhang et al. (2018) in alfalfa silage ensiled at a high temperature of 40 °C. The excessively high temperature during July (with the highest temperature over 40 °C during the day time) might be one cause of Garciella growth. Spearman's correlation analysis showed a weak relationship between Garciella and NH 3 -N (Fig. 5B). In all, the current evidence is insufficient to indicate a certain role of Garciella in the ensiling process.
Although L. plantarum inoculation promoted Lactobacillus abundance in silage, it failed to stabilize the abundance of Lactobacillus at 90 d. The overlong period of wilting before ensiling may have exacerbated the poor stability of the bacterial community. The period of wilting reduced the WSC concentration available for LA fermentation, and led to a lower LAB population before ensiling.
Excessively high temperature might be another factor influencing stability of the bacterial community.
LAB. An effect of high temperature on stability of the bacterial community was also reported in maize silage (Guan et al., 2018). Further research is needed to evaluate the influence of these factors on silage fermentation.

Relationships between bacterial community and fermentation properties
Recent studies illustrated an interaction of bacterial community and silage properties (McAllister et al., 2018). Mantel tests revealed close correlations of the pH, LA, AA, PA, NH 3 -N, and WSC with bacterial community composition. The pH and LA affected both the richness and diversity of bacterial community. The concentration of AA remained stable for both periods (Table 1) regardless of the shift in bacterial abundance. This is consistent with its poor relationship with bacterial richness illustrated by Spearman correlations. There was a negative correlation between AA and diversity of the bacterial community, possibly due to an inhibitory effect of AA toward some bacterial species. The PA and NH 3 -N had positive correlations with bacterial richness, but this did not affect diversity of the bacterial community. A probable explanation is that PA and NH 3 -N were mainly produced by some specific species in silage, so concentrations of these products mainly depended on the number of their producers rather than the composition of microbes. The WSC was negatively correlated with bacterial abundance. This is predictable, because WSC is consumed with the growth of microbes, as a main nutrient source of microbes in silage.
Spearman's correlation analysis illustrated some potential spoilage genera in alfalfa silage. Genera that correlated with NH 3 -N differed between the two periods. This indicates a potential shift in the main promoters of NH 3 -N formation during ensiling. At 30 d, Rosenbergiella and Bifidobacterium were potential promoters of NH 3 -N formation. Rosenbergiella is a genus belonging to Enterobacteriaceae.
An abundance of Bifidobacterium was also reported in direct-cut alfalfa silage with a high NH 3 -N concentration (Zheng et al., 2017). At 90 d, presence of Clostridia lead to increasing formation of NH 3 -N by fermenting amino acids (Flythe and Russell, 2004). Spearman's correlation analysis illustrates a series of NH 3 -N producers belonging to Clostridia, including Clostridium_XlVa, Clostridium_XI, and Finegoldia. Another two nitrogen-utilizing genera, Anaerotruncus and Peptoniphilus, were also considered to promote NH 3 -N formation. There is little evidence concerning the role of Eremococcus in silage fermentation. One recent study reported dominance of Eremococcus in the anodic biofilms in air-cathode microbial fuel cells (Zhang et al., 2016). In air-cathode microbial fuel cells, electroactive microorganisms, mainly in the form of anodic biofilms, accelerate the rate of electrochemical oxidation of complex organic substrates (Borole et al., 2011). This indicates the potential capability of Eremococcus in oxidizing organic materials; however, this study did not shown that oxidation by Eremococcus could occur under acidic conditions. Thus, further research is needed to explore its role in the ensiling system. Anaerococcus is a butyrate-producing saccharolytic genus, and so may exhibit synergy with Clostridium during ensiling -this might explain its positive correlation with NH 3 -N. Some species of Vagococcus were reported to exhibit capability for arginine hydrolysis (Teixeria et al., 1997), which might aggravate protein degradation in silage. In this study, Weissella was outcompeted by Lactobacillus following L. plantarum inoculation, and exhibited low abundance in the inoculated groups. This might explain the positive correlation between Weissella and NH 3 -N. A higher abundance of Weissella was exhibited in CK, for which the NH 3 -N concentration was significantly higher than the other treatments. Propionibacterium requires complex nutrition, including multiple vitamins and an abundant nitrogen source to produce PA and AA from glucose and other carbohydrates (Rogers et al., 2006). Accumulation of NH 3 -N might promote its growth and fermentation metabolism, resulting in the high PA concentrations in the 90-d silage. Brevibacillus was also positively correlated with NH 3 -N, but its role in ensiling is currently unclear.
The L. plantarum inoculation induced a rapid drop in pH at 30 d, which would inhibit proteolytic bacteria and contribute to true protein preservation (Weinberg and Muck, 1996). This was also illustrated by the inhibitory effect against Clostridiales in the inoculated silage and the negative correlation between Lactobacillus and NH 3 -N. Although L. plantarum inoculation aided in protein preservation in this study, it failed to stabilize the abundance of Lactobacillus at 90 d. This demonstrated that Inoculation of L. plantarum alone might be insufficient to overcome the adverse conditions during ensiling. And other strategies combining with L. plantarum inoculation might help to stablilize the fermentation quality. For instance, ensiling the freshly harvested alfalfa in combine with excessively wilted stover could help to reach an appropriate moisture for LA fermentation. Apart from Lactobacillus, correlation analysis suggests that some other genera may also contribute to protein preservation. However, the roles of these genera in the ensiling process remain little understood, and more evidence is needed to elucidate their relationships with protein preservation in silage.

Declarations
Ethics approval and consent to participate Not applicable.

Consent for publication
Not applicable.

Availability of data and material
The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.

Competing interests
Not applicable.

Funding
This work was supported by the National Natural Science Foundation of China (31772672). The sponsor had no involvement in study design; in the collection, analysis and interpretation of data; in the writing of the report; or in the decision to submit the article for publication.