A total of 20 LAB strains were isolated from common vetch, tall fescue and perennial ryegrass silages (laboratory silos; 450 mL capacity/silo). These forage crops were harvested at a maturity of pod bearing period, heading period and heading period, respectively and ensiled on September 28th, 2013 at Shigatse Grassland Station (Tibet, China: N 29°16′, E 88°51′, elevation 3,836 meters, annual mean temperature 6.5°C and average annual precipitation 400 mm) and fermented for 45 days at ambient temperatures. All the forage crops were well preserved despite the low temperature (6.5°C). According to our previous experiment (not published), some epiphytic LAB contribute to the good fermentation quality of common vetch, tall fescue and perennial ryegrass silages at low environmental temperatures in Tibet, and these were a good donor source for low temperature-resistant LAB isolation. The 20 LAB strains isolated were tested for low temperature resistance using the methodology described by Cai et al [5] and only four strains (LCG3, LTG7, I5, and LI3) were selected. Gram stain, colony morphology, catalase activity and gas production from glucose were determined according to Kozaki et al [6]. Growth at different pH was observed in de Man, Rogosa, Sharpe (MRS) broth (adjusting pH with HCl or NaOH), after incubation at 35°C for 2 days. Growth at different temperatures was observed in MRS broth after incubation at 5°C, 10°C, 15°C, and 20°C for 5 days. The salt tolerance of LAB strains was tested using MRS broth containing NaCl at 3% and 6.5% after incubation at 35°C for 2 days. Carbohydrate fermentation of 49 different compounds was examined by API 50 CHL assay (BioMérieux, Marcy l’ Etoile, Lyons, France) according to the manufacturer’s instruction. Lactobacillus plantarum MTD-1, a commercial LAB inoculant (G; Ecosyl. Products. Inc. Madison, WI, USA) that is a high-performance silage inoculant and works well in both high- and low-dry matter (DM) silage over a wide pH (7.5 to 3.5) and temperature (7.8°C to 45°C) range was used as positive control for silage making.

Each strain was cultivated in 5 mL of MRS medium at 37°C overnight and centrifuged at 10,000×g for 5 min. The cells were washed twice with buffer (10 mmol/L Tris–HCl, 0.1 mmol/L ethylenediaminetetraacetic acid [EDTA], pH 8.0) in a clean 1.5 mL microcentrifuge tube and recentrifuged. Genomic DNA was extracted by a TIANamp Bacteria DNA Kit (Tiangen Biotech Co., Ltd, Beijing, China) following the manufacturer’s instructions [7]. The genomic DNA concentration of each strain was determined at 260 nm by spectrophotometer. The extracted genomic DNA of each strain was stored at −20°C until use.

The 16S rRNA coding region was amplified by polymerase chain reaction (PCR). One microlitre of diluted DNA was used as a template for the PCR reactions. The PCR primers were 27f (5′-AGAGTTTGATCCTGGCTCAG-35′) and 1492r (55′-TACGGCTACCTTGTTACGACT-35′) [5]. The PCR reaction was performed in 0.2 mL microcentrifuge tubes with 25 μL of reaction mixture according to the protocol described by Escalante et al [8] with slight modification: DNA template (approximately 80 ng), 10× PCR buffer (0.1 mol/L Tris-HCl, pH 8.0, 0.5 mol/L KCl), 1.5 mmol/L MgCl₂ pH 8.0, 100 μmol/L of each dNTP, AmpliTaq DNA recombinant polymerase (1 U) (all PCR reagents were obtained from Tiangen Biotech Co., Ltd., China). Primer 27f (0.4 μmol/L) and primer 1492r (0.4 μmol/L) were made up to 25 μL with sterile distilled water. The PCR protocol was as follows: 95°C for 5 min, followed by 30 cycles of denaturation at 94°C for 30 s, primer annealing at 55°C for 1 min, elongation at 72°C for 1.5 min and thermal retardation at 72°C for 10 min. A 1 μL sample of reaction mixture was analyzed by 1.5% agarose gel electrophoresis in Tris-borate EDTA (TBE) 1×buffer. The gel was stained with GoldView and the bands were visualized under UV illumination. The PCR products were purified using the DNA purification systems according to the manufacturer’s instructions (Promega, Madison, WI, USA) for 16S rRNA sequencing analysis. The 16S rDNA sequences were aligned with 16S rRNA sequences from GenBank to identify organisms using BLAST analysis [9]. Then, the sequence information from representative organisms was introduced into the CLUSTALW program (Hitachi Software Engineering Co., Tokyo, Japan) for assembly and alignment. The 16S rRNA gene sequences of isolates were compared with sequences from LAB-type strains held in GenBank. Nucleotide substitution rates were calculated and phylogenetic trees were constructed by the neighbor-joining method based on 1,000 random resamplings.

Italian ryegrass was harvested at the boot stage of maturity in the experiment field of the Shigatse District Grassland Station on 7 July 2014, and was chopped to a theoretical cut length of 2 cm using a paper-cutter. LAB inoculation (LCG3, LTG7, I5, LI3, and G) and the control at different storage temperatures (10°C, 15°C, and 25°C) were designed as follows: all LAB strains were inoculated at 1×10⁵ colony-forming units (cfu)/g of fresh material, and the same volume of distilled water was used for the control. About 650 g of fresh material was tightly packed into a plastic polyethylene bottle (1 L capacity; about 650 kg/m³ fresh weight) and sealed with a screw top and plastic tapes. Five replicates were made for each treatment, and ninety bottles in total (6 inoculant treatments×3 temperatures×5 replicates = 90) were prepared. Silos for each treatment were stored in the incubators according to the design temperatures, and for simulating the production practice, we chose 35 days to assess the practical feasibility of silage making in Tibet.

When the bottle silos were opened, the silage was mixed thoroughly, and a 35 g sample was taken from each bottle, mixed with 70 mL distilled water and stored in a refrigerator at 4°C for 24 h. Then the extracts were filtered through two layers of cheesecloth and a filter paper. The filtrate was used for chemical analyses [10] and its pH was measured immediately with a glass electrode pH meter (HANNA pH 211, Hanna Instruments Italia Srl, Villafranca Padovana, Italy). Buffering capacity (BC) was determined by Playne and McDonald [11]. The DM was determined by oven drying at 65°C for 48 h to a constant mass, and then ground to pass through a 1 mm screen for later analysis. The WSC content was analyzed by colorimetric after-reaction with anthrone reagent [12]. The ammonia nitrogen (AN) content was determined by the phenol-hypochlorite procedure [13]. Total nitrogen (TN) content was determined according to Kjeldahl procedure [14]. Crude protein (CP) content was calculated by multiplying TN by 6.25 (since plant protein contains 16% N). Organic acid and ethanol contents of the silage were analyzed using the Agilent HPLC 1260 (Agilent Technologies, Inc., Berlin, Germany; column: Carbomix H-NP5, Sepax Technologies, Inc., Santa Clara, CA, USA; detector: refractive index detector, Agilent Technologies, Inc., Germany; eluent: 2.5 mmol/L H₂SO₄, 0.5 mL/min; temperature: 55°C).

About 10 g of silage was blended with 90 mL of sterilized saline solution (0.85% NaCl), and the extract was serially diluted to quantify the microbial group in sterile saline solution. LAB were counted on de Man, Rogosa and Sharpe agar medium (Difco Laboratories, Detroit, MI, USA) after incubation in an anaerobic incubator (N₂:H₂:CO₂ = 85:5:10, YQX-II; CIMO Medical Instrument Manufacturing Co., Ltd, Shanghai, China) at 37°C for 2 to 3 days. Aerobic bacteria were counted on nutrient agar (Nissui-Seiyaku Ltd, Tokyo, Japan) and yeasts were counted on potato dextrose agar (Nissui-Seiyaku Ltd, Japan) acidified with sterilized tartaric acid solution to pH 3.5. The agar plates were incubated at 37°C for 3 days. All microbial data were transformed to log₁₀ and are presented on a fresh weight basis.

The 16S rRNA sequences of the LAB isolates were analyzed using MEGA 6.0 for Windows (The Biodesign Institute, Tempe, AZ, USA). The data of silage fermentation quality, chemical composition and microbial counts were analyzed using the general linear model procedure according to the model for a factorial treatment design as follows:

Where Y_ij is the dependent variable; μ is overall mean; I_i is the effect of LAB inoculation; T_j is the effect of temperature; (I+T)_ij is the effect of interaction between LAB inoculation and temperature; and e_ij is the residual error term. The effects were considered significant at p<0.05, Tukey’s tests were used for means separation. All statistical procedures were performed with Statistical package for the Social Sciences (SPSS 16.0 for windows; SPSS Inc., Chicago, IL, USA).

The morphological and physiological characteristics of the five LAB strains used in this experiment are presented in Table 1. All LAB strains were Gram-positive, homofermentative and catalase-negative. The strains LCG3, I5, LI3, and G were rod-shaped, while strain LTG7 was cocci-shaped bacteria. All isolated strains (LCG3, LTG7, I5, and LI3) could grow at 5°C to 20°C, pH 3.0 to 8.0 and NaCl (3.0%, 6.5%). All strains could ferment galactose, cellobiose and mannitol, whereas none of them could utilize starch. Strains LTG7, I5, and G could ferment rhamnose while strains LCG3 and LI3 could not.

Molecular homological analysis was conducted and the phylogenetic tree was constructed based on the 16S rDNA sequences for observing divergences. The phylogenetic tree of partial 16S rDNA sequences of LCG3, LTG7, I5, and LI3 is presented in Figure 1. Strain LCG3 was most closely related to Lactobacillus plantarum, supporting the 98% value from bootstrap analysis of the phylogenetic tree (Figure 1), and showing 100% similarity in its 16S rDNA gene sequences. With 100% similarity in their 16S rDNA gene sequences, strain LTG7 was most closely related to the Pediococcus acidilactici and strain I5 was most closely related to Lactobacillus paraplantarum and strain LI3 was most closely related to Lactobacillus casei, all supporting the 99% values from the bootstrap analysis of the phylogenetic tree (Figure 1). The accession numbers of LCG3, LTG7, I5, and LI3 were KJ806299, KJ806304, KJ806310, and KJ806307, respectively.

The chemical and microbiological characteristics of Italian ryegrass prior to ensiling are given in Table 2. The DM content was 240.6 g/kg fresh matter (FM) and WSC content was 103.5 g/kg DM. The BC and CP content were 82.3 mEq/kg DM and 64.9 g/kg DM, respectively. The numbers of LAB, aerobic bacteria and yeasts on the material were 4.12, 6.12, and 4.25 log cfu/g FM, respectively.

As shown in Table 3, storage temperatures, LAB inoculations and their interaction significantly (p<0.01) affected the fermentation characteristics of Italian ryegrass silage. Compared to the corresponding controls at three different temperatures, all these LAB strains significantly (p<0.05) reduced pH and ammonia nitrogen/total nitrogen (AN/TN), and significantly (p<0.05) increased lactic acid and residual WSC contents, and ratios of lactic acid to acetic acid (LA/AA) of the silage. In addition, no or only trace amounts of propionic and butyric acid were found in most silages (data not shown). Compared to the corresponding positive controls G at various temperatures, marked increases (p<0.05) in lactic acid content occurred, resulting in lower (p<0.05) pH and AN/TN in I5-inoculated silage. At 25°C, compared to G, LI3 showed the similar pH (p<0.05), while in LTG7-inoculated silage, the pH declined and lactic acid content increased (p<0.05). At 15°C, LI3 and LCG3 showed lower pH and AN/TN, and higher lactic acid contents and ratios of LA/AA than G (p<0.05). At 10°C, I5, LI3, and LTG7 had a higher ratio of LA/AA than G (p<0.05). In addition, G showed the lowest ratio of LA/AA among the five inoculants at 15°C and 25°C (p<0.05), respectively.

As seen in Table 4, storage temperatures, LAB inoculations and their interaction significantly (p<0.01) affected the microorganism populations of Italian ryegrass silage. Compared to the corresponding controls at three different temperatures, all the inoculants significantly (p<0.05) decreased aerobic bacteria counts, and significantly (p<0.05) increased LAB counts of the silage. Compared to the corresponding G at various temperatures, I5 had lower aerobic bacteria and yeast counts, and higher LAB count (p<0.05). At 25°C, LI3 and LTG7 had higher LAB counts than G (p<0.05). At 15°C, LI3, LCG3, and LTG7 showed lower aerobic bacteria counts and higher LAB counts than G (p<0.05). At 10°C, LI3 and LCG3 decreased aerobic bacteria and yeast counts as compared with G. Besides, the highest yeast count was observed in G inoculated silage among the all treatments at 10°C (p<0.05).

Comparison of the uninoculated silages at three different temperatures indicated that the silage at 10°C showed higher pH, AN/TN and aerobic bacteria count, and lower lactic acid, acetic acid contents, ratio of LA/AA and LAB count than that at 15°C and 25°C (p<0.05).