Ll Dong-xia, Nl Kui-kui, ZHANG Ying-chao, LlN Yan-li, YANG Fu-yu

1 College of Animal Science and Technology, China Agricultural University, Beijing 100193, P.R.China

2 Beijing Sure Academy of Biosciences, Beijing 100085, P.R.China

Abstract This study assessed the effects of lactic acid bacteria (LAB), cellulase, cellulase-producing Bacillus pumilus and their combinations on the fermentation characteristics, chemical composition, bacterial community and in vitro digestibility of alfalfa silage. A completely randomized design involving a 8 (silage additives)×3 or 2 (silage days) factorial arrangement of treatments was adopted in the present study. The 8 silage additive treatments were: untreated alfalfa (control); two commercial additives (GFJ and Chikuso-1); an originally selected LAB (Lactobacillus plantarum a214) isolated from alfalfa silage; a cellulase-producing Bacillus (CB) isolated from fresh alfalfa; cellulase (C); and the combined additives (a214+C and a214+CB). Silage fermentation characteristics, chemical composition and microorganism populations were analysed after 30, 60 and 65 days (60 days followed by exposure to air for flve additional days). In vitro digestibility was analysed for 30 and 60 days. Compared with the other treatments, selected LAB a214, a214 combined with either C or CB, and Chikuso-1 had the decreased (P＜0.001) pH and increased (P＜0.001) lactic acid concentrations during the ensiling process,and there were no differences (P＞0.05) among them. Fiber degradation was not signiflcant (P≥0.054) in any C or CB treatments. The a214 treatment showed the highest (P=0.009) in vitro digestibility of dry matter (595.0 g kg–1 DM) after ensiling and the highest abundance of Lactobacillus (69.42 and 79.81%, respectively) on days 60 and 65, compared to all of other treatments. Overall, the silage quality of alfalfa was improved with the addition of a214, which indicates its potential as an alfalfa silage inoculant.

Keywords: alfalfa silage, cellulase, fermentation quality, in vitro digestibility, lactic acid bacteria

1. lntroduction

Alfalfa (Medicago sativaL.) has been widely used for animal feed due to its high crude protein content. However,because the harvest of alfalfa is seasonal, a suitable conservation method is essential for year-round alfalfa available for livestock.

Ensiling is considered an excellent conservation method for storing green forage, as this method preserves nutritive components, as well as improves forage palatability for livestock (McDonaldet al. 1991). Well-preserved silage depends on sufflcient substrates of the ensiled materials and proper fermentation procedures. During the ensiling process, lactic acid bacteria (LAB) convert water-soluble carbohydrates (WSC) into lactic acid and decrease pH,thereby inhibiting harmful microbiological activity. However,the high buffering capacity, low epiphytic LAB populations,and insufflcient WSC content of alfalfa make it difflcult for successful fermentation (Wanget al. 2006).

Various additives have been applied to improve alfalfa silage quality, including LAB and cellulase, with cellulase acting to increase the release of plant cell wall carbohydrates. Theoretically, positive responses are most likely to be achieved using low-WSC forages, such as alfalfa. The liberation of additional soluble sugars caused by flber hydrolysis is likely to increase both the rate and extent of lactic acid production. In addition, this process may potentially enhance dry matter intake and ruminal digestibility of alfalfa silages (Sunet al. 2009; Khotaet al.2016). However, these positive results are not consistent in previous studies assessing cellulase treated alfalfa silages(Kozelovet al. 2008; Mucket al. 2018), possibly due to the selectivity of cellulases for speciflc flbre components(high lignin concentration) of certain lucerne materials(Wilsonet al. 1994). Previous studies have reported that the combined addition of LAB and cellulase may have a synergetic effect on silage quality (Zhanget al. 2011; Niet al. 2014; Tianet al. 2014). Alternatively, Stokeset al.(1992) found that combination of LAB and cellulase was antagonistic and, therefore, did not improve silage quality.Nadeau and Buxton (1997) also reported that the effect of this mixture on silage quality varied among plant species,with cocksfoot being generally more responsive to cellulase than alfalfa.

Bacillus(B) species have been investigated in recent years as novel silage inoculants. Laraet al. (2016) found some species from this genus could improve the aerobic stability of silage by producing bacteriocin, which inhibits yeasts and molds. The application of these species in animal feed also exhibited positive effects on feed utilization and livestock growth performance (Zhanget al. 2017). Recent studies have also found that numerousBacillusspecies(such asB.cereus,B.flexusandB.pumilus) are the primary sources of microbial enzymes in the degradation of plant structural carbohydrates during the silage fermentation (e.g.,soybean curd residue mixed with alfalfa hays) (Ninget al.2017). However, little published information is available on fermentation characteristics using cellulase-producingBacillusas an additive for alfalfa silages. We hypothesised that the presence of cellulase-producingBacillusspecies will effect cell wall degradation on alfalfa silages, and the combination of LAB with cellulase (or cellulase-producingBacillus) will improve silage quality andin vitrodigestibility.

Therefore, the main objective of this study was to evaluate effects of LAB, cellulase, cellulase-producingBacillusand their combinations on fermentation, bacterial community,nutritive value andin vitrodigestibility of alfalfa silage.

2. Materials and methods

2.1. Experimental design and treatments

A completely randomized design involving a 8 (silage additives)×3 or 2 (silage days) factorial arrangement of treatments was adopted with three replications for each treatment in the present study. The eight silage additive treatments were: untreated alfalfa (control); two commercial additives (GFJ and Chikuso-1); an originally selected LAB(Lactobacillus plantaruma214) isolated from alfalfa silage; a cellulase-producingBacillus(CB) isolated from fresh alfalfa;cellulase (C); and the combined additives (a214+C and a214+CB). Silage fermentation characteristics, chemical composition and microorganism population were analysed for periods of 30, 60 and 65 days (60 days followed by exposure to air for flve additional days).In vitrodigestibility was analysed for periods of 30 and 60 days.

2.2. Characteristics of the employed additives

TheLactobacillus plantaruma214 was isolated from alfalfa silage in our laboratory. This strain exhibits great acidiflcation activity and growth rate in de Man, Rogosa and Sharpe (MRS) broth, as well as a strong ability to inhibit pathogens, such asSalmonella enterica,Escherichia coli,Micrococcus luteusandListeria monocytogenes.

Cellulase-producingBacillus(CB) was isolated from fresh alfalfa material and identifled asBacillus pumilus, with carboxymethyl cellulase activity of 37.52 U mL–1.

The physiological properties, antimicrobial and cellulase activities of the two strains (a214 and CB) were determined following the methods of Ennaharet al. (2000) and Niet al.(2016b), as shown in Table 1.

The strains a214 and CB were cultivated overnight in 50 mL of MRS broth and nutrient broth, respectively, at 30°C and then centrifuged at 8 000×g for 15 min. Resulting cells were mixed with skimmed milk and lyophilized. The populations of viable cells were determined by plate cultivation in MRS and nutrient agar after serial dilutions. Cellulase(C; Sinopharm Chemical Reagent Co., Ltd., Shanghai,China) was obtained fromTrichoderma viridewith endo-β-1,4-glucanases activity ≥15 000 IU (international unit) g–1.

Two commercial inoculant strains, GFJ (Sichuan Gao Fu Ji Biological Technology Co., Ltd., Sichuan, China) and Chikuso-1 (Snow Brand Seed Co., Ltd., Sapporo, Japan),both containingL. plantarum, were included to compare known effectiveness with those of our novel additives.

Table 1 The characteristics of strain a214 and CB1)

2.3. Silage preparation

Alfalfa was harvested in Zhuozhou, Hebei Province, China(115°44′E, 39°21′N). The second cut of alfalfa herbage was hand-harvested with a sickle at the early blooming stage, August 5, 2016. The forage samples were cut to approximately 2 cm length using a forage cutter (Lingong Machinery, Shandong, China) and wilted for 12 h to obtain the recommended dry matter (DM) content of approximately 400 g kg–1fresh matter (FM) prior to ensiling (Nadeau and Buxton 1997). Baseline chemical and microbiological characteristics of the pre-ensiled alfalfa samples are shown in Table 2.

Silages were prepared using a small-scale system: 200 g of alfalfa material was packed into plastic bags (N-9, Asahi Kasei Co., Ltd., Japan), and air was removed using a vacuum sealer (BH950; Matsushita, Tokyo, Japan). The two commercial additives (GFJ and Chikuso-1) were added at their recommended dosages (both 5.0×10–3g kg–1FM). The additive strains (CB and a214) were applied at 6.0 log colony forming units (cfu) g–1FM (Guoet al. 2014).Cellulase additive was applied at 20 mg kg–1FM (Liet al.2017).

Table 2 Chemical and microbial compositions of alfalfa

Nine silage bags were included for each treatment, and all silage samples were stored at room temperature (25.0 to 30.0°C). Three bags from each treatment were opened for the evaluation of silage quality after ensiling for 30 and 60 days. In addition, samples were assessed for 65 days(60 days followed by exposure to air for 5 additional days).For this additional treatment, approximately 100 g samples were transferred from each silage bag to 0.5-L plastic jars capped with sterile gauze for aerobic exposure at room temperature (25.0 to 30.0°C) for 5 days (total 65 days).The design of ensiling periods was based on previous studies. Liet al. (2014) reported 30 days is required for silage fermentation, yet 60 days may be required for efflcient ensiling with insufflcient WSC content of alfalfa (Zhuanget al. 2012). For aerobic stability evaluation, aerobic deterioration typically occurs during 5-day exposure to air,following 60 days of ensiling (Zhanget al. 2016).

2.4. Sample collections and preparation

Silage bags from each treatment were opened following the speciflc time required (30, 60, or 65 days), contents mixed thoroughly, and divided into three silage samples.A silage sample (10 g) from each replicate silage bag was homogenized in 90 mL of sterilized distilled water, and used for pH, NH3-N, organic acid analysis and microbiological enumeration. The second sample (5 g) was frozen at –80°C for high-throughput sequencing of metagenomic DNA. The remaining silage sample was oven-dried at 65°C for 48 h,and used to determine nutritive values orin vitrodigestibility.

2.5. Microbiological analysis

Microbiological evaluation included enumeration of LAB(incubated anaerobically on MRS agar), coliform bacteria(violet red bile agar) and molds and yeasts (on potato dextrose agar) after incubation at 30°C for 48 h. These media were obtained from Beijing Aoboxing Bio-tech Co.,Ltd. (Beijing, China).

2.6. Analysis of microbial diversity through high-throughput sequencing of metagenomic DNA

Pre-ensiled alfalfa and silage samples were added to 20×volume of sterilized phosphate-buffered saline (pH 7.4), prior to DNA extraction (Niet al. 2016a). The V3–V4 region of the bacterial 16S rRNA gene was amplifled by PCR (95°C for 2 min followed by 25 cycles at 95°C for 30 s, 55°C for 30 s, 72°C for 30 s and a flnal extension at 72°C for 5 min)using the primers 338F (5′-ACTCCTACGGGAGGCAGC AG-3′) and 806R (5′-GGACTACHVGGGTWTCTAAT-3′).For the minimization of PCR bias, triplicate PCR reactions were conducted for each treatment. Then, three PCR products were mixed in equi-density ratios. Finally, mixture was purifled for DNA concentration and sequencing (Liet al. 2015; Zhenget al. 2017). Therefore, the data of high-throughput sequencing were not able to be further statistically analyzed. DNA samples were paired-end sequenced on an Illumina MiSeq platform at Majorbio Bio-Pharm Technology Co., Ltd. (Shanghai, China). Any sequences that contained mismatches and ambiguous reads in the primers were removed for quality-control purposes.

Operational Taxonomic Units (OTUs) were clustered with a similarity cut-off of 97% using UPARSE (version 7.1, http://drive5.com/uparse/), and chimeric sequences identifled and removed using UCHIME. The taxonomy of each 16S rRNA gene sequence was analysed using Ribosome Database Project (RDP) Classifler (http://rdp.cme.msu.edu/) against the SILVA (SSU115)16S rRNA database with a confldence threshold of 70% (Amatoet al. 2013). The α-diversities of the samples, mainly the Shannon index, Chao richness estimator, and Good’s coverage, were calculated using Mothur (version 1.30.1, http://www.mothur.org/wiki/Classify.seqs). Taxonomic classiflcation at the genus levels was performed using the RDP algorithm to classify the representative sequences of each OTU (Liet al. 2015).

2.7. Chemical analysis

DM and crude protein (CP) were analysed according to the guidelines provided by the Association of Offlcial Analytical Chemists (AOAC 1990). Acid detergent flber (ADF) and neutral detergent flber (NDF) were determined using the methods described by Van Soestet al. (1991). WSC content was determined using the anthrone method (Murphyet al.1958). DM recovery was estimated by measuring the differences of DM in silage before and after ensiling. pH was measured using a glass electrode pH metre (PHS-3C, INESA Scientiflc Instrument, Shanghai, China). The concentration of NH3-N was determined using the method described by Broderick and Kang (1980). The organic acid content was measured by HPLC (LC-20A, Shimadzu, Ltd.,Japan).

2.8. In vitro digestibility of the silages ensiled for 30 and 60 days

Thein vitrodry matter disappearance (IVDMD) from gas production was assessed according to previously described methods (Menke and Steingass 1988). Rumen fluid was collected from three rumen-flstulated dairy cows before the morning feeding (each dairy cow was fed 0.5 kg oat hay,2.5 kg alfalfa hay and 20.0 kg corn silage per day), and immediately transported to the laboratory. Rumen fluid was poured into a brown bottle with buffer solution (buffered rumen fluid, pH 6.85; Menke and Steingass 1988).

Oven-dried silage samples of 30 and 60 days were ground to pass through a 1-mm screen, and 200 mg of each samples weighed in triplicate. The samples were then incubated in serum bottles (120 mL) with 75 mL buffered rumen fluid at 39°C in a water bath. The bottles were purged with anaerobic N2for 5 s, sealed with a butyl rubber stopper and a Hungate’s screw cap, and connected to the gas inlets of an automated trace-gas recording system using medical plastic infusion pipes. This enabled continuous recording of cumulative gas production (GP). After incubation at 39°C for 72 h, the culture fluids in each bottle were individually flltered with pre-weighed nylon bags (8 cm×12 cm, 42-μm pore size). These bags were thoroughly rinsed and dried at 65°C for 48 h to a constant weight. The difference between the initial incubated DM and the residual DM, after correction using incubated blanks, was then calculated to determine the IVDMD (Zhang and Yang 2011).

2.9. Statistical analysis

The cumulative gas production (GP(t), mL g–1DM) at time(t) measured for each fermentation bottle was fltted to an exponential model (Franceet al. 2000):

Where,Arepresents the asymptotic GP generated at a constant fractional rate (c) per unit time,tis the gas recording time (h), and Lag represents the lag time phase(h) prior to the initiation of GP. The average gas production rate (AGPR, mL h–1) was deflned as the AGPR between the start of the incubation and the time at which the GP was half of its asymptotic value according to the following equation:

Data on silage fermentation characteristics, chemical composition and microorganism population were analysed as an 8 (silage additives)×3 (silage days) factorial arrangement of treatments using two-way ANOVA procedure in SAS 9.4 (SAS Institute, Cary, NC, USA, 2017). Forin vitrodigestibility, data were analysed as an 8 (silage additives)×2(silage days) factorial arrangement of treatments. The statistical models included silage additive, silage day and their interaction. Each silage bag treated as an experimental unit (n=3/treatment). Mean values were then compared for signiflcance using Duncan’s multiple range method.Differences were considered signiflcant whenP＜0.05.

3. Results

3.1. Effects of silage additives and silage days on pH and fermentation products of alfalfa silages

Silage additives, days and their interaction had effects(P＜0.001) on lactic acid concentration (Table 3). On day 30,lactic acid concentrations in silages of Chikuso-1, a214,a214+C and a214+CB increased (P＜0.001) compared to other silage treatments. There were no differences (P＞0.05)among Chikuso-1, a214, a214+C and a214+CB. Lactic acid concentrations were higher (P＜0.001) in control and GFJ than in CB, while there were no differences (P＞0.05)among control, GFJ and C or between C and CB. On day 60, lactic acid concentrations were also higher (P＜0.001)in silages of Chikuso-1, a214, a214+C and a214+CB than in other treatments. Lactic acid concentration was higher(P＜0.001) in GFJ than in control, C or CB treatment, and there were no differences (P＞0.05) among Chikuso-1, a214,a214+C and a214+CB or among control, C and CB. On day 65, lactic acid concentrations were higher (P＜0.001) in GFJ, Chikuso-1, a214, a214+C and a214+CB than in other treatments, and there were no differences (P＞0.05) among GFJ, Chikuso-1, a214, a214+C and a214+CB or among control, C and CB.

Silage days had no effect (P=0.772) on acetic acid concentration, but silage additives and their interaction had effects (P＜0.001) on acetic acid concentration. On day 30,acetic acid concentrations were higher (P＜0.001) in silages of a214, a214+CB and Chikuso-1 than in other treatments.Acetic acid concentrations were higher (P＜0.001) in GFJ and a214+C than in control, C and CB. C showed higher(P＜0.001) acetic acid concentration than control or CB. No differences (P＞0.05) were observed among GFJ, Chikuso-1 and a214+C, or between a214 and a214+CB, or between control and CB. On day 60, acetic acid concentrations were higher (P＜0.001) in silages of C and GFJ than in other silage treatments. Acetic acid concentration was higher(P＜0.001) in Chikuso-1 than in control, CB, a214, a214+C or a214+CB. Acetic acid concentration was higher (P＜0.001) in a214+CB than in control, CB, a214 or a214+C. Acetic acid concentrations were higher (P＜0.001) in a214 and a214+C than in control. There were no differences among CB, a214 and a214+C, or between control and CB. On day 65, acetic acid concentrations were higher (P＜0.001) in silages of GFJ,Chikuso-1 and a214+CB than in other silage treatments.Acetic acid concentrations were higher (P＜0.001) in control and a214+C than in C and CB. There were no differences(P＞0.05) among control, a214 and a214+C, or between C and a214. Acetic acid concentration was the lowest(P＜0.001) in CB among the different treatments.

Both silage additives and days affected (P＜0.001) the pH and NH3-N content, but their interaction had no effect(P≥0.325) on either pH or NH3-N content. The pH was lower(P＜0.001) in a214, a214+C, a214+CB or Chikuso-1 than in other silage treatments. GFJ or Chikuso-1 had lower(P＜0.001) pH than control, C or CB, and there were no differences (P＞0.05) between GFJ and Chikuso-1, or among control, C and CB. The pH was lower (P＜0.001) on day 60 or 65 than on day 30, and no differences (P＞0.05) were observed for pH between days 60 and 65. NH3-N contents were lower (P＜0.001) in a214, a214+C anda214+CB than in other treatments, and no differences (P＞0.05) were observed among a214, a214+C anda214+CB or among all the other treatments. NH3-N content was higher (P＜0.001)on day 65 than on day 30, with an intermittent content of NH3-N found on day 60.

Silage additives, days and their interaction had no effect(P≥0.112) on propionic acid or butyric acid content.

3.2. Effects of silage additives and silage days on chemical composition of alfalfa silages

Both silage additives and days affected (P＜0.001) the DM,but their interaction had no effect (P=0.327) on DM (Table 4).The DM was higher (P＜0.001) in a214 than in control, GFJ or Chikuso-1. There were no differences (P＞0.05) among C, CB, a214, a214+C and a214+CB, or among control,GFJ, Chikuso-1, C, a214+C and a214+CB. The DM was lower (P＜0.001) on day 30 or 60 than on day 65, and no differences (P＞0.05) were observed for DM between day 30 and 60.

Silage additives, days and their interaction affected(P＜0.001,P＜0.001,P=0.002, respectively) the DM recovery.On day 30, DM recovery was higher (P=0.002) in a214 than in other treatments, except C. There were no differences(P＞0.05) between a214 and C. DM recovery in C or a214+CB was similar (P＞0.05) to Chikuso-1, but was higher(P=0.002) than DM recovery in control, GFJ, CB or a214+C.There were no differences (P＞0.05) among control, GFJ,Chikuso-1, CB and a214+C, or among Chikuso-1, C and

a214+CB. On day 60, DM recovery was higher (P=0.002)in a214 or a214+CB than in other treatments. There were no differences (P＞0.05) between a214 and a214+CB. DM recovery was higher (P=0.002) in Chikuso-1 than in the control, GFJ, C, CB or a214+C. DM recovery was higher(P=0.002) in GFJ or a214+C than in control, C or CB, and there were no differences (P＞0.05) between GFJ and a214+C. DM recovery was higher (P=0.002) in CB than in C, with an intermittent value of DM recovery was found in control. On day 65, among the different treatments,the highest (P=0.002) DM recovery was found in a214 treatment. DM recovery of a214+CB was lower (P=0.002)than a214, but was higher (P=0.002) than other treatments.DM recovery was higher (P=0.002) in Chikuso-1 and a214+C than in control, GFJ, C and CB, and there were no differences (P＞0.05) between Chikuso-1 and a214+C. DM recovery was higher (P=0.002) in GFJ or CB than in control or C, and there were no differences (P＞0.05) between GFJ and CB. DM recovery was the lowest (P=0.002) in control among the treatments.

Table 3 Effects of silage additives(A) and silage days (D) on pH and fermentation products of alfalfa silages

Table 4 Effects of silage additives(A) and silage days (D) on chemical composition of alfalfa silages

Silage additives had no effect (P=0.467) on WSC content. But silage days and their interaction had effects(P＜0.001) on WSC content. On day 30, WSC contents were higher (P＜0.001) in silages of GFJ, a214, a214+CB,CB and a214+C than in other treatments. WSC content was higher (P＜0.001) in CB than in control or C. The lowest(P＜0.001) WSC content was found in C treatment. There were no differences (P＞0.05) among control, Chikuso-1 and a214+C. On day 60, WSC contents were higher(P＜0.001) in silages of a214+CB and a214+C than in other treatments. WSC contents were higher (P＜0.001) in C and a214 than in control and Chikuso-1, and there were no differences (P＞0.05) among C, a214, GFJ and CB, or between control and Chikuso-1. On day 65, WSC contents were higher (P＜0.001) in Chikuso-1 and GFJ than in other silage treatments. WSC contents were higher (P＜0.001)in C and a214+C than in control, CB, a214 and a214+CB,and there were no differences (P＞0.05) among GFJ, C and a214+C. The control and CB showed higher (P＜0.001) WSC contents than a214 and a214+CB. The lowest (P＜0.001)WSC content was found in a214+CB treatment.

Silage additives, days and their interaction did not affect (P≥0.105) the NDF content. Silage days had effect(P＜0.001) on ADF content, but silage additives and their interaction did not affect (P≥0.054) the ADF content. ADF was higher (P＜0.001) on day 65 than on days 30 or 60, and there were no differences (P＞0.05) between days 30 and 60.

Silage additives had effect (P＜0.001) on CP content, but silage days and their interaction did not affect (P≥0.052) the CP content. CP content was higher (P＜0.001) in silage of a214 than in control, C or CB, with an intermittent content of CP was found in GFJ, Chikuso-1, a214+C or a214+CB.

3.3. Effects of silage additives and silage days on microbial composition of alfalfa silages

No yeasts and molds were detected in any silage samples after ensiling (data not shown). Silage additives, days and their interaction had effects (P＜0.001) on LAB and coliform bacteria counts (Table 5). For LAB, on day 30, LAB counts were higher (P＜0.001) in silages of CB, control, GFJ and Chikuso-1 than in other silage treatments. LAB counts were higher (P＜0.001) in the silages of C and a214+CB than in a214+C. There were no differences (P＞0.05) among C,a214+CB and a214 or between a214 and a214+C. On day 60, LAB counts were higher (P＜0.001) in silages of CB and C than in other silage treatments. LAB counts were higher(P＜0.001) in silages of control, Chikuso-1 and a214+C than in GFJ, a214 and a214+CB. There were no differences(P＞0.05) among control, Chikuso-1, C and a214+C or among GFJ, a214 and a214+CB. On day 65, LAB counts were higher (P＜0.001) in control, C and CB than in other silage treatments. There were no differences (P＞0.05)among control, C and CB, or among GFJ, Chikuso-1,a214+C and a214+CB.

For coliform bacteria, on day 30, coliform bacteria counts were higher (P＜0.001) in Chikuso-1 and CB than in other treatments. Coliform bacteria count was higher (P＜0.001)in C than in GFJ, a214, a214+C or a214+CB. There were no differences (P＞0.05) between C and control or between a214+C and a214+CB. Coliform bacteria count was the lowest (P＜0.001) in a214 among the different treatments.On day 60, coliform bacteria counts were lower (P＜0.001) in a214 and a214+C than in other treatments. There were no differences (P＞0.05) between a214 and a214+C or among other treatments. On day 65, coliform bacteria counts were higher (P＜0.001) in a214+CB, GFJ, a214 and a214+C than in other treatments. Coliform bacteria count was the lowest(P＜0.001) in control among the different treatments. There were no differences (P＞0.05) among GFJ, Chikuso-1, C,CB, a214 and a214+C.

High-throughput sequencing of 16S rRNA gene amplicons was conducted to comprehensively characterize bacterial communities in silage samples. A total of 4 117 OTUs at 3% dissimilarity level were determined to further analyse the bacterial community. The OTUs ranged from 84 to 220 for each sample (Table 6). Chao index, utilized to estimate the OTUs number, showed a similar trend to OTUs. The coverage values in all silages ranged from 0.998 to 0.999, suggesting that most of bacteria were detected.For Shannon index (the diversity index of microorganic population), a214 treatment gradually decreased from 1.87 to 1.18, which was maintained as the lowest index among all the treatments across all ensiling days.

The dynamic variance of bacteria community structures of silages on genus level was demonstrated with the principle component analysis(PCA) in Fig. 1. Principle component 1 (PC1) and 2 (PC2) explained 38.85 and 26.35% of total variance, respectively. Most treated silage samples showed similar bacteria community structure after ensiling for 30, 60 and 65 days. This was especially true for the a214 treatment, as this silage additive displayed the closest distance to one another, across all three period samples.

The changes in the bacterial community structure at the genus level during the silage process were assessed using a threshold of ＞0.1% of the total (Fig. 2). The dominant genus in pre-ensiled alfalfa includedXanthomonasandCyanobacteria, which accounted for 73.96% of the sequences. After ensiling,Lactobacillusbecame the most abundant genus, particularly in the a214-treated samples: in fact the proportion ofLactobacillusin a214-treated alfalfa ranged from 57.45 to 79.81%during the ensiling. In contrast, the proportions ofXanthomonasandCyanobacteriadecreased to less than 6.00 and 0.10% after ensiling for all silages, respectively. However, the proportion ofEnterobacterincreased in most of the silages compared with the pre-ensiled alfalfa,especially for a214+C treatment, ranged from 46.92 to 67.18% during the ensiling.Enterococcusshowed relatively high abundance (＞20%)during ensiling for Chikuso-1, C and CB treatments. When C or CB was combined with a214, the proportion ofEnterococcusdecreased, whereas the amount ofLactobacillusandEnterobacterbothincreased. Certain amounts of other LAB genera likeWeissella,PediococcusandLactococcuswere also found among the silages.Marginallevels (＜0.50%) ofBacilluswere found only in the CB-treated silages after ensiling.

?

Table 6 Diversity statistics of bacterial community during ensilage process

Fig. 1 Principle component analysis (PCA) of samples on genus level. Pre, pre-ensiled; A, alfalfa; I, silages of 30 days; II, silages of 60 days; III, exposed to air for 5 days after 60 days; 1, control; 2, commercial additive (GFJ); 3, commercial additive (Chikuso-1);4, cellulase (C); 5, cellulase-producing Bacillus (CB); 6, Lactobacillus plantarum a214; 7, a214+C; 8, a214+CB.

3.4. Effects of silage additives and silage days on in vitro digestibility of alfalfa silages

Data are shown in Table 7. Silage additives had effect(P=0.009) on IVDMD, but silage days and their interaction had no effect (P≥0.066) on IVDMD. The highest (P＜0.009)IVDMD was found in a214 treatment. C treatment had the decreased (P＜0.009) IVDMD compared to the other silage treatments, except the a214+C treatment. Silage days had effect (P=0.018) on A, but silage additives and their interaction had no effect (P≥0.256) on A. A was higher(P=0.018) on day 60 than on day 30. Both silage additives and days had effects (P=0.028 andP=0.003, respectively) on c (the fractional rate for gas production of ‘A’), but their interaction had no effect(P=0.615) on c. The c was higher (P=0.028) in silage treatment of Chikuso-1 than in C. The c was higher(P=0.003) on day 60 than on day 30. There were no differences (P＞0.05)for c among control, GFJ,Chikuso-1, a214, a214+C and a214+CB, or between C and CB.

Silage additives, days and their interaction had no effect (P≥0.079) on GP72,Lag or AGPR.

4. Discussion

Previous studies have shown alfalfa is difflcult to be ensiled, due to its low WSC and high buffering capacity. Cellulase and LAB have been widely used to improve silage fermentation and enhance thein vitrodigestibility in many silage materials (McDonaldet al.1991; Santoset al. 2013;Khotaet al. 2016). In silage fermentation studies, the application of cellulase is often included to enhance flber degradation, providing LAB with more fermentable substrate (McDonaldet al.1991). Positive effects of the addition of cellulase have been observed for many silage materials,such as napier grass, king grass, guinea grass, maize stover, oil palm frond,sugarcane top and straw of naked barley(Sunet al.2009; Zhanget al. 2010;Ebrahimiet al. 2014; Liet al. 2014; Khotaet al.2016). However, in our study, cellulase did not have a signiflcant effect on the reduction of NDF or ADF content. In fact, many previous studies have reported enzymatic treatments may not be beneflcial. Seven out of eight flbrolytic enzymes did not improve flber degradation for maize silage (Eun and Beauchemin 2007) and addition of flbrolytic enzymes had no effect on cell wall degradation during the ensiling of corn, bermudagrass, or lucerne (Shepherd and Kung 1996;Mandebvuet al. 1999; Kozelovet al. 2008)

Table 7Effects of silage additives(A) andsilage days (D) on in vitrodry matter disappearanceand gas production kineticsinculture fluids after 72 hincubationof alfalfasilage

The capacity of cellulase to process silage quality depends on many factors, such as, environmental conditions, the method of enzyme preparation, the type and properties of the forage, the presence of epiphytic microflora, WSC, and speciflc ensiling technique utilized.It has been shown that the activity of cellulase could be inhibited by the presence of certain microorganisms in king grass silage (Liet al. 2014). It has also been shown that additions of lignocellulose degrading enzymes reduced IVDMD of guinea grass silage (Tianet al. 2014). The authors attributed this reduction to the high DM of the silage(401.1 g kg–1FM), inhibiting enzyme activity. Liuet al.(2016) reported that the poor activity of cellulase forin vitrodigestibility of silage was due to hydroxylation of the most digestible fraction of the structural polysaccharides, leaving a less digestible residue. The lack of response to enzyme treatment obtained in some studies has been attributed to enzyme instability (Moharreryet al. 2009), as temperature,pH, and ionic strength can affect the activity of enzymes.The high lignin concentration in lucerne has been reported to be responsible for its recalcitrance towards enzymatic attack by cellulase during the ensiling process (Nadeau and Buxton 1997; Moharreryet al. 2009). In general, the reasons for the above mentioned inefflcient results are complex and differed among experiments, which may explain among silages treated in this research, C alone treatment had no signiflcant effect for silage fermentation and IVDMD.

In our study, CB treatment did not result in signiflcant improvement in NDF or ADF degradation, which indirectly reflected similar WSC contents between CB and all other silage additive treatments. We speculated that the poor activity of CB additives was due to their low viability under the ensiling environment. Although CB was applied at a concentration of 6 log cfu g–1FM before ensiling, only a trace amount ofBacilluswas detected in the CB treatment at the end of the ensiling process. This phenomenon implied that the long duration of ensiling under anaerobic condition was unfavorable for the survival of CB, thus inhibiting its cellulase activity.

Silage inoculated with LAB (application rate ≥5 log cfu g–1FM as fed) has been reported to markedly enhance silage fermentation and increase DM recovery in alfalfa and other forages, dependent of the LAB species employed (Oliveiraet al. 2017). In our study, it was expected that addition of LAB to alfalfa silage would help increase lactic acid production, thereby accelerating fermentation, shortening the time to reach pH stability and inhibiting the activity of undesirable microorganisms. The higher DM and CP content, increased lactic acid concentration, decreased pH and NH3-N content observed in the presence of the LAB strain a214 supported this expectation.

Previous researches reported that LAB inoculation could reduce DM loss during the ensiling, thereby increase IVDMD (Caoet al. 2011). Importantly, a214 treatment had the highest DM recovery and IVDMD in this study. Similar results were reported by Nadeau and Buxton (1997), who found that cellulase alone behaved similarly to the control,while the silage quality enhanced when cellulase was combined with LAB inoculant for cocksfoot. Therefore, it is likely that the overall effect of these combinations are primarily caused by the LAB inoculant.

In this study, the a214 treatment showed similar pH and lactic acid concentration to the commercial additive Chikuso-1. But the decreased NH3-N content and increased IVDMD of a214 treatment compared to Chikuso-1 indicated that the strain a214 was more suitable for alfalfa silage. The possible explanation for the high efflciency of the LAB a214,that was originally isolated from alfalfa silage, is because of its strong ability to adapt to the speciflc environment (low epiphytic LAB, WSC concentrations and high buffering capacity) of alfalfa silage. This adaptability results in its rapid multiplication, enabling it to overwhelm the growth of undesirable microorganisms to eventually dominate the silage fermentation process and effectively preserve silage nutrients (Ogunadeet al. 2016; Silvaet al. 2016).

The aerobic exposure period has many risks, such as the multiplication of potentially pathogenic microorganisms,the production of mycotoxins, that aerobic exposure can cause spoilage with losses in nutritional value (Tabaccoet al. 2011). In our study, no signiflcant differences (P＞0.05)were found among the silages over the incubation period,except that the NH3-N content increased gradually over time. These results may be due to the inherent factors of alfalfa, such as its high buffering capacity and high acetic acid concentration which exhibits a strong inhibitory effect on many spoilage microorganisms. The acetic acid levels in all silages measured in this study ranged from 22.17 to 40.72 g kg–1DM, exceeding minimum inhibitory concentration level for 20 g kg–1DM (Danneret al. 2003).

In forage conservation, the nature of the bacteria that are prevalent in the fermentation process plays an important role during ensiling. In the present study, both culture-based isolation methods and DNA-based community analysis were used to acquire a comprehensive understanding of the complexity of the microbiota residing in the silage material.Sequence analysis conflrmed that the clear majority of the bacteria present in the silage was composed of the genusLactobacillus.

At the genus level of bacterial community structure,similarity was observed across all sampling days, among silage samples subjected to the same additive. This result indicated that the bacterial community structures were relatively stable after ensiling. Analysis of different treatments revealed that use of a214 as an additive resulted in a markedly higher content ofLactobacillus,when compared to other treatments. Fermentation analysis showed that the a214-treated silage was characterized by a relatively higher content of lactic acid and lower pH value.In contrast, C and CB treatments yielded a lower abundance ofLactobacilluswith relatively lower lactic acid contents and higher pH values. Similar results were reported by Niet al.(2016a) and Mendez-Garciaet al. (2015), who described that silages with extremely high relative abundances of LAB were characterized by low pH. Therefore, these results are possibly due to the extremely acidic conditions, restricted acid-sensitive bacteria and limit microbial diversity. Kraut-Cohenet al. (2016) reported that well-preserved silage has low taxonomic diversity, whereas spoiled silage showed a highly diverse microbiota with a low abundance of LAB. This phenomenon also implied that LAB inoculant can proliferate in silage and overcome indigenous microorganisms, whilst also affecting the fermentation characteristics of the silage(Eikmeyeret al. 2013).

Enterobactercan grow under anaerobic conditions and protects itself under adverse conditions, including low pH environments (Pereiraet al. 2007). In our study,Enterobacterwas shown to be the subdominant bacteria after ensiling, likely responsible for the large amounts of acetic acid produced in the ensiling process (McDonaldet al. 1991). Santoset al.(2015) reported that most of theEnterobacterdetected in silages are non-pathogenic. However, their development is undesirable because they create competition with LAB for WSC at the beginning of the fermentation. Silvaet al.(2016) also suggested thatEnterobacteris harmful to silage because it initiates the production of ammonia and slows down acidiflcation of silage.

5. Conclusion

The results of the present study reveal that LAB strain a214 was the most effective additive in improving fermentation characteristics, bacterial diversity andin vitrodigestibility of alfalfa silage. However, either C or CB did not improve the silage quality. Therefore, strain a214 isolated from alfalfa silage is recommended as a promising starter culture for alfalfa silage.

Acknowledgements

This work was supported by the National Key R&D Program of China (2017YFD0502102), the National Technology Leader“Ten Thousand People Plan” of China (201502510410040)and the National Key Technology R&D Program of China during the 12th Five-year Plan period of China(2011BAD17B02).