Selection of Lactic Acid Bacteria from Alfalfa Silage and Its E ﬀ ects as Inoculant on Silage Fermentation

: The ﬁrst part of the study aimed to isolate, characterize, and identify wild lactic acid bacteria (LAB) strains from alfalfa silage produced in a tropical area. LAB strains were isolated from alfalfa silage ensiled for 1, 3, 7, 14, 28, and 56 days (d) and were identiﬁed by sequencing the 16S rRNA gene. The second part aimed to investigate the e ﬀ ects of wild LAB strains on the nutritive and fermentative characteristics of alfalfa silage. This trial was conducted according to a completely randomized design in a 4 × 2 factorial scheme [four inoculants (I) × two harvests (H)], (n = 4). The inoculants were: (1) no inoculant (CTRL), (2) Lactobacillus pentosus (AV 14.17); (3) L. pentosus + Lactobacillus brevis + Pediococcus acidilactici (Combo); and (4) commercial inoculant (CI). Alfalfa forage (7 kg) was ensiled in 10 L buckets and opened after 90 d. Seventy-seven strains were isolated. Pediococcus , Lactobacillus , and Weissella represented 52.0, 24.7, and 20.8% of the isolates, respectively. For the ﬁrst harvest, Combo, CI, and all inoculated silages showed lower acid detergent ﬁber ADF, neutral detergent ﬁber (NDF), and ammonia nitrogen (NH 3 -N), respectively. Silage fermented with AV14.17 presented greater residual water-soluble carbohydrate (WSC) in the second harvest and showed the lowest pH in both harvests. AV14.17 strain has potential as an inoculant for alfalfa silage production.


Introduction
In the tropical world, ruminant production systems use forage as a basic substrate for meat and milk yield. Thus, animal production follows the seasonality of forage production. In beef animals, for example, there is satisfactory weight gain during the rainy season and difficulty gaining or even maintaining weight during the dry season, increasing production cycles, with a decrease in the quality of the products. In this context, a viable strategy to overcome the effects of this productive gap would be the use of conserved forages.
Among conservation techniques, silage is the most used procedure in various parts of the world. Lactic acid bacteria (LAB) are the main group of microorganisms involved in silage conservation, being divided into homofermentative LAB and facultative and obligate heterofermentative LAB. The homofermentative LAB rapidly decrease pH and increase lactic acid relative to other fermentation products. Meanwhile, the obligate heterofermentative LAB silage additives slowly convert lactic acid to acetic acid and 1,2-propanediol during silo storage, improving aerobic stability [1]. Many species that were considered part of the homofermentative group are now in the facultative heterofermentative group, such as Lactobacillus plantarum and Lactobacillus pentosus [2].

Physiological Tests
The screening of LAB was performed based on the results of Gram staining, catalase activity, and acid production. Growth in MRS broth (Beton, Dickinson and Company, Sparks, NV, USA) was evaluated at different temperatures (15 • C and 45 • C) and pH levels (3.5, 4.0, 4.5 and 8.5) [22] after 24 h of incubation, and it was measured by the absorbance at 630 nm using a plate spectrophotometer (Thermo Fisher Scientific ® , Waltham, MA, USA). The growth at pH levels tested was done at 37 • C.
The metabolism classification of the isolates in heterofermentative or homofermentative was carried out by inoculating them in MRS broth with Durham tubes, which were incubated for 48 h at 37 • C. The presence of bubbles inside the tubes indicated heterofermentative metabolism while the absence of them indicated homofermentative metabolism.

Determination of Antimicrobial Activity
The antimicrobial activity of each isolate against the indicator bacteria Listeria monocytogenes ATCC 7644, Escherichia coli K12, Bacillus cereus ATCC 4904, and Staphylococcus aureus ATCC 25,923 was assessed by means of the deferred activity method [23]. Antimicrobial activity was defined by the formation of inhibition zones around the microdrops. The inhibition zone was measured with a ruler and calculated by subtracting the colony diameter from the external diameter of the inhibition area. This measurement was performed in duplicate, and the mean value was considered.

Extraction of LAB Genomic DNA
The DNA samples from the isolates obtained from fresh alfalfa and its silages at the various fermentation periods were extracted by using a commercial kit (Wizard ® Genomic DNA Purification kit, Promega, Madison, WI, USA), with some modifications in the protocol steps. The isolates were grown in 5 mL of MRS broth and incubated at 37 • C for 14 h. The cells were then centrifuged (Mikro 200 R, Hettich) at 1000× g for 5 min and washed once with 0.85% saline solution. The cell sediment was resuspended in 480 µL of EDTA (50 mM), and 50 µL of lysozyme 50 mg mL was immediately added. The subsequent DNA extraction steps were performed according to the kit manufacturer's instructions. The concentration of extracted DNA was assessed with a Nanodrop spectrophotometer (Thermo Scientific 2000, Thermo Fisher Scientific, Wilmington, NC, USA), and the DNA was stored at −80 • C.
The sequences were compared to references deposited in the GenBank database and aligned using the algorithm BLASTn (Basic Local Alignment Search Tool) (http://www.ncbi.nlm.nih.gov/BLAST) for nucleotides. The sequences of 16S rRNA genes that exhibited ≥97% of similarity were considered to be part of the same operational taxonomic unit (OTU) [25]. Moreover, the sequences of the isolates obtained in our study were deposited in the GenBank database. The isolates identification, access code, growth characteristics, and antimicrobial activity are available in Supplementary Materials Tables S1 and S2.

Selection of LAB Strains for Alfalfa Silage Production
Three strains of LAB were selected for evaluation as inoculant in alfalfa silage production. L. pentosus AV14.17 and Lactobacillus brevis AV14.2 were isolated from alfalfa silage on day 14 of fermentation. P. acidilactici AV56.13 was isolated from silage on day 56 of fermentation (Table S1). These strains were chosen based on their performance under the diverse conditions tested and the broad spectrum of inhibition against pathogenic microorganisms. It was not our goal to choose between homofermentative and heterofermentative strains. We were looking for strains that could control undesirable microorganisms and thus favor the growth of LAB during the fermentation process. In addition, it is necessary that these strains survive in various conditions, so they could act as soon as possible in the fermentation process. The commercial inoculant Sil All 4 × 4 W.S (Alltech, Sao Paulo, Brazil), which contained sucrose, L. plantarum, P. acidilactici, Enterococcus faecium, Lactobacillus salivarus ssp. salivarus, silicon dioxide, amylase, cellulose, hemicellulose, and xylanase in its composition, was rehydrated in 50 mL of distilled water and applied in an application rate of 1.0 × 10 6 CFU/g of fresh forage.
Alfalfa forage was harvested and wilted for six hours in the field reaching 331.4 and 435.2 g/kg of DM for harvest 1 and 2, respectively. The alfalfa forage characteristics before ensiling are shown in Table 1. Piles of 10 kg of forage were individually treated (each replicate) with the microbial inoculants and then approximately 7 kg of forage were packed in 10 L plastic buckets (mini-silo, 25 cm diameter and 25 cm height) sealed with tight lids. The buckets (16 buckets from each harvest) were stored in the laboratory at room temperature and opened after 90 days. Dry matter recovery was estimated by the weight and DM content of the fresh alfalfa and its silage [26].

Preparation of the Isolated Strains for Use as Inoculant
L. pentosus AV14.17, L. brevis AV14.2, and P. acidilactici AV56.13 were cultivated in MRS broth for 48 h at 37 • C for cell activation, since they were kept at −80 • C before their use. After that, each inoculum was standardized using a spectrophotometer (630 nm) at an optical density of 0.05, in 20 mL of MRS broth and kept at 37 • C for 16 h. This incubation time was used based on a pre-test, where the average time to obtain the microorganisms in their log phases was 16 h. The inoculum obtained in this last incubation was used to prepare the inoculants. The amount of this inoculum to achieve the final application rate of 1.0 × 10 6 CFU/g of fresh forage was centrifuged at 1000× g × 10 min. The supernatant was discarded, and cells were resuspended in 50 mL of distilled water.

Chemical Analysis
Forage and silages samples were analyzed for chemical composition. The samples were collected and dried at 55 • C for 72 h in a forced-air oven and ground to pass a 1 mm knife mill screen. Dried ground samples were used to quantify DM content (method 934.01), crude protein (CP) (method 984.13), acid detergent fiber (ADF) (method 973.18), and mineral matter (method 942.05) [27]. Neutral detergent fiber (NDF) was determined using heat-stable α-amylase without the use of sodium sulfite and was corrected for residual ash [28] and nitrogen [29].
An aqueous extract was prepared from 25 g of sample (forage or silage) and 225 mL of sterile Ring's solution (Oxoid, Hampshire, England) homogenized for 1 min. The aqueous extract was used to measure the pH using a potentiometer (Tecnal, SP, Brazil) and then it was acidified with 1:1 H 2 SO 4 diluted with distilled water for further analyzes of ammonia nitrogen (NH 3 -N), using the phenol-hypochlorite method [30], WSC [31], and organic acids. Samples for organic acids quantification were treated with calcium hydroxide and cupric sulfate and analyzed by high-performance liquid chromatography (HPLC; SPD-10 AVP, Shimadzu, OR, USA) [32]. The HPLC apparatus (SPD-10 AVP, Shimadzu) was equipped with a refractive index detector and an Aminex HPX-87H column (BIO-RAD, CA, USA) with the mobile phase containing 0.005 M H 2 SO 4 and a flow rate of 0.6 mL/min at 50 • C.

Quantification of Microbial Populations
A portion of the water extract was used to determine the microbial population. Serial dilutions were made and inoculated on MRS agar to determine the LAB count. Violet Red Bile agar (Difco) was used to quantify the enterobacteria population and Potato dextrose agar (Difco) [acidified with 1.5% of tartaric acid solution 10% wt/v] for yeasts and molds determination. Plates of LAB and enterobacteria were kept at 37 • C for 48 and 24 h, respectively. Yeast and mold plates were incubated at 25 • C for 120 h.

Statistical Analysis
Data from part 2 were analyzed as a completely randomized design, in a 4 × 2 factorial scheme. The model included the fixed effects of inoculants, harvests, and the interaction I × H, according to the following statistical model: where Yijk = dependent variable, µ = overall mean, αi = fixed effect of the ith inoculant, βj = fixed effect of the jth harvest, (αβ)ij = interaction between inoculant and harvest, and εijk = random error. Analyses of variance and multiple comparisons of means were performed using the GLM procedure of SAS (SAS Intitute Inc., Cary, NC, USA). The means were separated by Tukey's test, and differences and tendencies were declared at the levels of p ≤ 0.05 and p > 0.05 to p < 0.10, respectively.

Screening and Identification of LAB from Alfalfa Forage and Its Silage (Part 1)
One hundred and two strains were isolated from alfalfa forage and its silage. Among these, 77 strains (three from plant and 74 from silage) showed similarity above 97% with sequences of others LAB available in the GenBank database. The genera Pediococcus, Lactobacillus, and Weissella represented 52.0, 24.7, and 20.8% of the isolates, respectively ( Figure 1). P. acidilactici was the most frequent isolated species from alfalfa through the fermentation period, which corresponded to 28.6% of the isolates, followed by the L. plantarum (16.9%) (Figure 2). These genera and species represented the dominant microorganisms able to grow in MRS medium. In general, the species with the highest numbers of strains isolated from silage were Weissella cibaria, Pediococcus pentosaceus, P. acidilactici, and L. plantarum (Figure 2).
Enterococcus casseliflavus, Weissella confusa, W. cibaria, P. pentosaceus, L. pentosus, L. brevis, and Lactobacillus casei species were also identified (Table S2). All isolates were able to inhibit the growth of at least one indicator microorganism. Among the 77 isolates, 36 were able to grow under the tested conditions (Table S1).       The isolated strains, L. pentosus AV14.17 (GenBank access code-MK713801), L. brevis AV14.2 (GenBank access code-MK713790), and P. acidilactici AV56.13 (GenBank access code-KY613548) were selected from part 1 of the study to be used in part 2, as they showed antimicrobial activity against all indicator microorganisms, and good growth under tested conditions. The characteristics of the strains are presented in Table 2.

Chemical Composition (Part 2)
There was an effect of I × H interaction on NDF (p = 0.03) and ADF (p = 0.02) contents. From harvest 1, silage treated with CI and COMBO showed lower (p < 0.01) NDF and ADF, respectively, compared to CTRL and AV14.17. Dry matter content was affected by both inoculant (p < 0.01) and harvest (p < 0.01). Alfalfa silage treated with microbial inoculants had greater DM compared with CTRL silage. Similarly, silage from harvest 2 presented greater DM content than harvest 1. Crude protein was only affected by harvest, which was greater (p < 0.01) for silages from harvest 2 ( Table 3).

Fermentation Characteristics and Microbial Population (Part 2)
There was an effect of I × H interaction (p = 0.04) on residual WSC content and dry matter recovery (p = 0.02). A tendency (p = 0.06) of I × H interaction on NH 3 -N concentration was observed. There was no difference (p > 0.99) among inoculants on WSC content of silages from harvest 1. However, silage treated with AV14.17 strain showed greater (p < 0.02) WSC content compared to CTRL and CI, in harvest 2. Although interaction effect was found for dry matter recovery, there was no difference (p > 0.66) among inoculants in both harvests. Ammonia nitrogen concentration was lower (p < 0.05) for silages treated with inoculants compared with the CTRL silage in harvest 1. There was no difference among inoculants (p > 0.97). The pH was affected by inoculant (p = 0.04) and harvest (p < 0.01). The silage treated with AV14.17 strain showed lower pH. Moreover, pH was lower for silages from harvest 2 compared with harvest 1. Butyric acid was not detected in the silages. Lactic and acetic acids were affected only by harvest. Silage from harvest 2 showed greater lactic acid (p < 0.01) and lower (p < 0.01) acetic acid than harvest 1 (Table 4).   Enterobacteria were not detected in silages. There was a tendency of I × H interaction (p = 0.07) on yeast count. However, the interaction study showed no differences (p > 0.10) among treatments. There was a harvest effect on LAB and mold counts. Lactic acid bacteria were in greater number (p < 0.01) and molds in lower (p < 0.01) for silage from harvest 1 as compared to harvest 2 (Table 5).

Discussion
Isolation of LAB from fresh crop or silage to be used as microbial inoculant can improve silage quality [33] since epiphytic LAB are not always present in sufficient number to ensure efficient fermentation in the silo [34]. The LAB strains identified in the present study by 16S rRNA gene sequencing belonging to W. cibaria, L. plantarum, L. pentosus, L. brevis, L. casei, P. pentosaceus, and P. acidilactici were also found in other studies [8,[35][36][37].
Species of the genus Pediococcus are known to be able to grow quickly, and as a result, they represent the largest portion of LAB when they are found in association with bacteria from the genera Lactobacillus, Leuconostoc, and Weissella [38]. This statement is in agreement with our findings, where P. acidilactici was detected in larger numbers, followed by L. plantarum. Several studies have reported the presence of L. plantarum as the main species in silage [11,34], demonstrating the importance of this species for silage preservation.
Selecting strains able to grow at high temperature and low pH and with antimicrobial activity may be essential to the effectiveness of microbial inoculants in tropical conditions. A strain identified as L. casei isolated from tropical forage was able to grow at low pH and was able to improve silage quality [39]. Similar results were reported for P. pentosaceus isolated from alfalfa silage produced under tropical conditions, which showed broad-spectrum antimicrobial activity against pathogenic and harmful spoilage organisms and the ability to grow in different pH and temperature ranges [11]. In this study, alfalfa silage treated with this specie showed greater lactic acid content and lower NH 3 -N on day 28 of fermentation compared to the commercial inoculant and the lowest pH value from 14 d of fermentation. Silva et al. [40] also reported improvement in alfalfa silage fermentation when silage was prepared using P. pentosaceus isolated from Stylosanthes silage with antimicrobial activity of protein nature. Strains isolated in the present study were able to inhibit the indicator microorganisms. However, their antimicrobial activity has not been tested or characterized. Further investigation into the nature of the substance involved in inhibiting these spoilage microorganisms is required.
In the present study, the variation in characteristics of alfalfa forage between harvests resulted differences in the effect of the isolated strains associated with silage fermentation. Dry matter and WSC are the major factors affecting the fermentation process in the silo [41]. It is recognized that silage pH increases as forage DM content increases (above 40-45%) [41] because LAB metabolism is affected by the low water activity [42]. However, silages from harvest 1, which showed lower DM content compared to harvest 2, presented silages with greater pH and lower lactic acid concentration. Despite the difference on DM content and silage characteristics of the two harvests, the strain AV14.17 was effective in reducing the pH of the silage compared to that of the control silage. This result is in line with those found by Jones where inoculated silages showed higher rates of pH decline at all dry matter levels studied (330, 430, and 540 g/kg) [43]. Although the same wilting time was used for silage making in both harvests, DM was affected by weather variables, such as solar brightness (0.3 vs. 9 h; harvest 1 and 2, respectively), evaporative power (1.5 vs. 2.4 mL; harvest 1 and 2, respectively) and precipitation (2.6 vs. 0 mL; harvest 1 and 2, respectively) [44]. Furthermore, harvest alters carbohydrate metabolism and thus stimulates the mobilization of monosaccharides in relation to sugar polymers and the hydrolysis of starch reserves in soluble sugars [45], which may explain the greater concentration of WSC in harvest 2.
As already highlighted, DM and WSC are primary in guaranteeing adequate fermentation in the silo. Although silage fermented with AV14.17 presented similar NH 3 -N content compared to the COMBO and CI, these two did not differ from the control silage in some important aspects (pH, residual WSC). Addition of L. plantarum on silage production of several crops has shown improvement in silage fermentation [46][47][48]. However, few studies reported the use of L. pentosus on silage fermentation. It is important to emphasize, since L. plantarum is one of the most used species as silage inoculant, the similarity between L plantarum and L pentosus, which are genotypically related species and their sequences of the 16S rRNA gene differ only by 2 bp. Therefore, this species could also be a promising inoculant of silage.
The chemical composition of the silages was practically unaffected by inoculant. Differences in DM content were biologically insignificant. However, silage treated with COMBO showed lower ADF in harvest 1. Low ADF is desirable, as it is a fraction of the fiber that can compromise silage quality by affecting digestibility. Some meta-analysis studies showed the effect of inoculation with homofermentative and heterofermentative LAB strains on ADF and lignin silage decrease without affecting NDF and CP concentrations [5,49]. This behavior could be associated with the acidic hydrolysis of hemicellulose when organic acids are produced in higher concentration [50]. However, lactic and acetic acids were not altered by the inoculants in our study, and another unknown effect may be involved.

Conclusions
In conclusion, Weissela cibaria, Pediococcus pentosaceus, Pediococcus acidilactici, and Lactobacillus plantarum were the predominant species in alfalfa silage produced in a tropical region. Pediococcus acidilactici dominated the microbial community throughout the fermentation period in alfalfa silage. The strain of Lactobacillus pentosus AV14.17 has potential to be used as inoculant on alfalfa to improve silage fermentation. However, this strain must be tested on other forage legumes and other promising strains found in the present study should also be investigated.