Dynamics of microbial community during ensiling direct-cut alfalfa with and without LAB inoculant and sugar

Transcription

1 Journal of Applied Microbiology ISSN ORIGINAL ARTICLE Dynamics of microbial community during ensiling direct-cut alfalfa with and without LAB inoculant and sugar M.L. Zheng, D.Z. Niu, D. Jiang, S.S. Zuo and C.C. Xu College of Engineering, China Agricultural University, Beijing, China Keywords alfalfa silage, bacterial community, clostridial community, fermentation, high-throughput sequencing. Correspondence Chuncheng Xu, College of Engineering, China Agricultural University, No. 1 Qinghua East Road, Beijing 18, China. xucc@cau.edu.cn 216/2594: received 1 September 216, revised February 21 and accepted 5 March 21 doi:1.1111/jam. Abstract Aim: To gain deeper insights into the clostridial community dynamics and chemical transformations during the ensiling of alfalfa. Methods and Results: Direct-cut alfalfa silage (with the dry matter content of 24 g kg 1 ) was prepared with or without the addition of a lactic acid bacterial inoculant and sucrose. Silages were sampled at,,,, and days after ensiling and their bacterial community was determined using high-throughput sequencing with a special focus on the clostridial community. A clostridial fermentation occurred in the control silage, with high contents of acetic acid, butyric acid and ammonia nitrogen and Clostridia counts; while the inoculated silage was well preserved, with low ph and high lactic acid content. Lactic acid bacteria dominated the bacterial community during the ensiling process. In the control silage, Weissella confusa, Lactobacillus brevis, Enterococcus mundtii and Pediococcus acidilactici were identified at the beginning of the fermentation. Thereafter, W. confusa, Lactobacillus helsingborgensis and Bifidobacterium asteroides appeared and quickly prevailed. In the inoculated silage, Lactobacillus plantarum dominated the whole ensiling process. The genus Clostridium dominated the clostridial community, and was depressed with the inoculated treatment. Clostridium perfringens, Garciella sp. and Clostridium baratii were the main initiators of the clostridial fermentation of the control silage, while Clostridium tyrobutyricum became the most abundant Clostridia with prolonged ensiling. Overall in the inoculated silage, little changes in the clostridial community were observed throughout the ensiling period. Treating alfalfa silage with a homolactic acid-based bacterial inoculant prevented a clostridial fermentation resulting in more efficient fermentation. Conclusion: Distinct changes in the bacterial community with a special focus on the clostridial community were associated with the development of the clostridial fermentation during the ensiling of alfalfa. Significance and Impact of the Study: High-throughput sequencing based on a novel Clostridia-specific primer set proved a potentially useful tool to study the clostridial community dynamics, and could aid to elucidate the mechanism by which the clostridial fermentation develops during the ensiling of alfalfa. Introduction Ensiling is a preservation method used for moist crops and is based on naturally occurring lactic acid bacterial fermentation under anaerobic conditions. In this process, epiphytic lactic acid bacteria (LAB) convert water-soluble carbohydrates (WSC) into organic acids, mainly lactic acid, thereby reducing ph (McDonald et al. 1991). The conditions created are inhibitory to the subsequent activity of deleterious micro-organisms such as Clostridia and Enterobacteria. However, the epiphytic LAB of forages do not always reduce ph rapidly because their initials Journal of Applied Microbiology 122, The Society for Applied Microbiology

2 M.L. Zheng et al. Clostridial community of alfalfa silage numbers can be low, fast acidifying homofermentative species may be absent, there is an inadequate WSC content of forage, stage of maturation or wilting and conservation procedures may delay or hinder development (Rossi and Dellaglio 2). Accordingly, moist silage that is poorly preserved often contains high concentrations of butyric acid and ammonia nitrogen (NH -N), which is associated with clostridial activity (Pahlow et al. 2). In addition to causing reductions in nutritional value, the clostridial fermentation affects the hygienic quality of silage by creating conditions that permit an increase in the numbers of pathogenic organisms and their toxins (Flythe and Russell 24). Moreover, silage of poor microbiological quality has been identified as the main source of clostridia spore contamination in raw milk (Colombari et al. 21). Differences may exist in the ensiling characteristics between crop species. Unlike cereal crops, such as corn, alfalfa is viewed as a difficult crop to ensile, primarily because of its high buffering capacity, low concentration of WSC and vulnerability to the undesirable secondary clostridial fermentation, especially when ensiled at moisture concentration > g kg 1 dry matter (DM) (Coblentz and Muck 212). Despite the impact of silage feeding on animal health and productivity, the fermentation is not a fully controlled process and the clostridial fermentation was difficult to eliminate in large-scale ensiling of alfalfa (Rossi and Dellaglio 2). It is possible, however, to use chemicals, enzymes and bacterial inoculants to overcome these problems in order to enhance the preservation of alfalfa silage. For instance, LAB inoculants and sugars are often added in alfalfa silage preparation to facilitate the lactic acid bacterial fermentation (Zhang et al. 29; Wu and Nishino 216). In this regard, ensiling direct-cut alfalfa with or without the addition of LAB inoculant and sucrose can be a good approach to distinguish between the lactic acid bacterial and clostridial fermentation. Of the many factors that can affect the silage fermentation, the type of micro-organisms that dominate the process often dictates the quality of the silage produced (Ennahar et al. 2). Hence, monitoring of the ensiling process with respect to changes in the chemical and microbial compositions would be helpful for thoroughly understanding and improving the ensiling process (Namihira et al. 21). The microbial community dynamics that occur during ensiling have been described previously; however, the vast majority of these investigations utilized culture-based techniques, that while informative, do not give much information with regard to the level of microbial diversity. It has been estimated in microbial ecology studies that considerably <1% of the microbial diversity present is culturable (McCabe et al. 215). Recent advancements in molecular tools have enabled us to elucidate the entire structure of the complex microflora found in silage. For instance, random amplified polymorphic DNA (Rossi and Dellaglio 2), terminal restriction fragment length polymorphism (McEniry et al. ) and denaturing gradient gel electrophoresis (DGGE) (Zhang et al. 29; Wu and Nishino 216) have been used to study specific aspects of the ensiling process. Although these studies have expanded our knowledge of the ensiling processes, they only identified a few of the most abundant operational taxonomic units (OTUs) present, because of their poor limits of detection (McGarvey et al. 21). Additionally, microbial taxa composition is often not examined using these conventional molecular techniques and requires considerable efforts with downstream Sanger sequencing to identify dominant members of communities (Hamm et al. 216). Therefore, a comprehensive understanding of the silage microbial community remains far from complete largely. The introduction of high-throughput sequencing has opened new ways to explore complex microbial ecosystems. High-throughput sequencing is a rapid and sensitive molecular approach that can offer information related to the occurrence and abundance of the microbial community in a given ecosystem. It also gives much greater coverage of the microbial community than can be provided by other molecular techniques (e.g. fingerprints, cloning) (Zened et al. 21). This technique has been widely applied across a range of systems including soil (Zhang et al. 216), rumen (McCabe et al. 215), compost (Neher et al. 21) and biogas reactor (Jaenicke et al. 211). As for silage microbial community analysis, two published studies used this technique to investigate the bacterial community of Lactobacillus buchneri-inoculated grass silage (Eikmeyer et al. 21) and microalgaesupplemented Manyflower silvergrass silage (Li et al. 215). These studies provided vast amounts of new information about the potential metabolism, and the number of unclassified bacterial genera detected demonstrated the paucity of present knowledge on the microbial community. Sufficient data are not yet available on the bacterial community dynamics involved in alfalfa ensiling, and the clostridial community dynamics associated with the development of the clostridial fermentation have not yet been identified clearly. The objective of this study is to describe the bacterial community dynamics during the ensiling of alfalfa by high-throughput sequencing, with a special focus on the clostridial community using a novel Clostridia-specific primer set. We also provide data on the fermentation characteristics and microbial counts of alfalfa silage. The results of this study may aid to elucidate the mechanism by which the clostridial fermentation develops during the ensiling of alfalfa. Journal of Applied Microbiology 122, The Society for Applied Microbiology 5

3 Clostridial community of alfalfa silage M.L. Zheng et al. Materials and methods Plant material and laboratory silage preparation Alfalfa (Medicago sativa L. Sanditi ) was grown in experimental plots of the Beijing Academy of Agricultural and Forestry Sciences (9 4 N, 116 E), Beijing, China. Triplicate experimental plots were used, each plot with an area of 9 1 m. Approximately 6 kg per hectare of chicken manure was applied prior to sowing, with no additional fertilizer being applied during the crop growing. The average soil organic matter in the experimental plots was 111 gkg 1 ; effective N, P and K in the soil were 829, 19 and 115 mgkg 1 respectively. First-cut alfalfa was randomly harvested artificially at full-bloom stage on 25 May 215, from each of the three experimental plots, leaving a stubble of 1 cm. The fresh forage samples were then immediately taken into the laboratory for processing. The fresh forage samples were chopped to about 1 2 cm using a forage cutter, and inoculated with or without LAB inoculant (Lactobacillus plantarum strain L2JPL65, GenBank accession number: KM5162.1) at a concentration of 1 5 CFU per g fresh matter and sucrose (analytical reagent; Beijing Chemical Industries Ltd, Beijing, China) at a concentration of 2 g kg 1 fresh matter. Two hundred grams of the pre-ensiled material was packed into a plastic film bag silo (Hiryu KN type, mm; Asahikasei, Tokyo, Japan), and this was followed by air removal using a vacuum sealer (BH 95; Matsushita, Tokyo, Japan). Silos were prepared in triplicate and stored at ambient temperature for,,,, and days. Chemical analyses and microbial enumeration DM content was determined by drying the materials in an oven at 65 C for 48 h, and WSC content was determined using the method of Owens et al. (1999). Fermentation products of silage were determined from cold-water extracts. Wet silage (1 g) was homogenized with 9 ml of sterilized distilled water, and then filtered through four layers of medical gauze and a qualitative filter paper (pore size 15 2 lm; Hangzhou Special Paper Co., Ltd, Hangzhou, China). The ph was measured with a glass electrode ph meter (S2K; Mettler Toledo, Greifensee, Switzerland), and NH -N content was determined by the method of Broderick and Kang (198). The filtrate was further processed with a dialyser of 22 lm to determine organic acid contents, including lactic acid, acetic acid, propionic acid and butyric acid, by high-performance liquid chromatography (LC-1A; Shimadzu, Tokyo, Japan). The analytical conditions were as follows: column, Shodex RSpak KC-811S-DVB gel C (8 mm 9 cm; Shimadzu); oven temperature, 5 C; mobile phase, mmol l 1 HClO 4 ; flow rate, 1 mlmin 1 ; injection volume 5 ll; detector, SPD- M1AVP (Shimadzu). Wet silage samples (1 g) were blended with 9 ml of sterilized water and serially diluted (1 1 to 1 5 ) in sterilized water before microbial enumeration. The numbers of LAB were measured by plate count on de Man, Rogosa and Sharpe agar (Difco Laboratories, Detroit, MI) incubated at C for 48 h under anaerobic conditions (Anaerobic box, TE-HER Hard Anaerobox, ANX-1; Hirosawa Ltd, Tokyo, Japan). The numbers of Enterobacteria were determined on Blue Light Broth agar (Nissui Ltd, Tokyo, Japan) incubated at C for 48 h. The Clostridia were counted on Clostridia Count agar (Nissui Ltd) using the Hungate (1969) technique. Cold-water extracts were heated to 8 C for 1 min to inactivate the vegetative cells and to trigger the germination of spores, and 5-ll aliquots of the diluted extracts were inoculated into Hungate tubes. The tubes were rolled on ice and incubated at C for days. Black colonies were identified as Clostridia. Metagenomic DNA extraction of the microbial community For extraction of DNA from microbial community, cells were first harvested from silage. Ten grams of wet silage was homogenized with 4 ml of sterilized distilled water, the obtained extracts were then filtered through two layers of medical gauze and centrifuged at 1 g and 4 C for 15 min. The resulting cell pellets were kept at 8 C until DNA extraction. Genomic DNA was extracted using the FastDNA TM SPIN Kit for Soil and the FastPrep â Instrument (MP Biomedicals, Santa Ana, CA) according to the manufacturer s instructions. PCR amplification of target genes The universal primer pair 6-F (5 -GTACTCCTACGG GAGGCAGCA- ) and 86-R (5 -GTGGACTACHVGG GTWTCTAAT- ) was used to amplify the V and V4 regions of the bacterial 16S rrna gene. A new pair of Clostridia-specific primer SJ-F (5 -CGGTGAAATGCGTA GAKATTA- ) and SJ-R (5 -CGAATTAAACCACATGCT CCG- ) was used to amplify the half V4 and total V5 regions of the clostridial 16S rrna gene (Hu et al. 2). Both primer pairs contained 12-bp barcodes unique to each sample, in order to enable the pooling of all PCR products for sequencing and the subsequent assignation of sequence reads to their respective samples (Neher et al. 21). PCR reactions were performed in 5 ll volumes containing ng of template DNA, 2 ll of each primer (5 lmol l 1 ), 4 ll of dntps (25 mmol l 1 ), 5 ll of19 58 Journal of Applied Microbiology 122, The Society for Applied Microbiology

4 M.L. Zheng et al. Clostridial community of alfalfa silage Pyrobest Buffer, ll of Pyrobest DNA polymerase (25 Ull 1 ; Takara Bio Inc., Shiga, Japan), and doubledistilled water was added to obtain a final volume of 5 ll. Thermocycling conditions were as follows: 95 C for 5 min to denature the DNA, with amplification proceeding for 25 cycles at 95 C for s, C for s and 2 C for 4 s; a final extension of 1 min at 2 C was added to ensure complete amplification. PCR products were mixed in equidensity ratios. Then, a mixture of PCR products were purified with GeneJET Gel Extraction Kit (Thermo Fisher Scientific Inc., Carlsbad, CA). Library preparation and sequencing Sequencing libraries were generated using NEB Next â Ultra TM DNA Library Prep Kit for Illumina (New England Biolabs Inc., Ipswich, MA) following the manufacturer s recommendations and index codes were added. The library quality was assessed on the Qubit â 2. Fluorometer (Thermo Fisher Scientific Inc.) and Agilent Bioanalyzer 21 system (Agilent Technologies Deutschland GmbH, Waldbronn, Germany). At last, the library was sequenced on an Illumina MiSeq PE platform (Illumina Corporation, San Diego, CA) by the method of Neher et al. (21). Processing of sequencing data Quality control and assignment of sequences to samples based on their barcodes were done following the standard QIIME (ver. 1.., pipeline as described previously (Caporaso et al. 21; Bokulich et al. 21). A default similarity level of 9% was used to cluster sequences into individual OTUs using UPARSE software package (ver...11, (Edgar 21), and a single representative sequence from each clustered OTU was used to align to the Greengenes database ( (DeSantis et al. 26). Taxonomic classification for each OTU was determined with RDP Classifier (ver. 2.2, (Wang et al. 2) using a minimum confidence cut-off of 8. The Alpha diversities of samples, mainly of the Shannon index, Chao1 richness estimator and the Good s coverage, were created using MOTHUR software (ver. 1..1, (Schloss et al. 29). Phylogenetic and molecular evolutionary analyses of the prevalent OTUs were performed as described by McGarvey et al. (21), using MEGA6 software ( gasoftware.net/) (Tamura et al. 21). Prior to the redundancy analysis (RDA) on genus level, high-throughput sequencing data were analysed to confirm that the gradient lengths fit a linear model using detrended correspondence analyses. RDA was conducted by the R software package (ver..2.5, using Monte Carlo permutation (999 repetitions) (Zhang et al. 216). Statistical analysis Silage chemical compositions and microbial counts data were analysed by a two-way ANOVA for a (treatment 9 storage periods) factorial arrangement of treatments using the GLM procedures of SAS 9.1 (SAS Institute, Inc., Cary, NC), and the Tukey s test was used for comparisons at 5% significant level. Results Microbiological and chemical analysis of alfalfa silage Direct-cut alfalfa contained a low DM content of 24 g kg 1, and their WSC, crude protein, neutral detergent fibre and acid detergent fibre contents were 59, 195, and 291 g kg 1 DM respectively (data not shown). Adding LAB inoculant and sucrose increased WSC content from 59 to 1 g kg 1 DM and LAB counts from 9 1 to CFU per g fresh matter respectively (Table 1). Lactic acid content, lactic to acetic acid ratio and WSC content were higher (P < 1), while ph, contents of acetic acid, propionic acid, butyric acid and NH -N and microbial counts were all lower (P < 1) in the inoculated silage compared to the control. Silage ph and WSC content decreased (P < 1), while NH -N content increased (P < 1), over the course of the fermentation in both treatments. The interaction between treatments and storage periods significantly affected (P < 1) all of the studied parameters. The greatest increase in lactic acid, acetic acid, propionic acid and NH -N contents occurred during the first days of ensiling. Stage of ensiling had a significant effect on microbial composition as determined by enumeration using culture-based approaches. The numbers of LAB increased (P < 1) rapidly in both treatments after ensiling with the highest numbers observed at days of ensiling, and then decreasing through day. Clostridia numbers increased (P < 1), while Enterobacteria numbers decreased (P < 1) throughout the ensiling period. Metagenomic analysis of the bacterial community dynamics After removing unqualified sequences, the valid sequences of all 12 samples summed up to 25 4, with an average Journal of Applied Microbiology 122, The Society for Applied Microbiology 59

5 Clostridial community of alfalfa silage M.L. Zheng et al. Table 1 ph, fermentation products (g kg 1 dry matter, unless otherwise stated), water-soluble carbohydrates content (g kg 1 dry matter) and microbial counts (log 1 CFU per g fresh matter) of alfalfa silage Fermentation products Microbial counts Treatments Days ph LA AA PA BA LA/AA NH -N (g kg 1 TN) WSC LAB Clostridia ENT CK 6 a 55 e f d f 95 a 91 f 59 a 5 d 29 d 59 e 589 b 66 d 54 e 8 c 6 e 121 b e 1 b 85 a 29 c 9 b 5 c 4 b d 8 c 22 d 11 c 9 d 15 c 858 a 264 c 88 a 4 d 22 a 24 c c 4 c 84 d 115 c d 82 b 6 b 6 c 554 e 1 b 25 b 15 b 6 b 55 e 1546 b 1 e 694 c 52 b 544 d 541 f 122 c 9 a 22 a 94 a 4 f 18 a 11 f 685 c 4 a 44 f LS 665 A 61 C 15 D B 4 D 9 E 1 A 529 F 1 AB B 22 B 124 C B 581 BC 11 D 2 B 852 A 12 AB <2 46 C 81 A A B 2 C 1 C 22 C 82 B 166 AB <2 41 D 88 A 15 B A 61 B 124 B 16 D C 1 B <2 4 D 8 A 124 C 4 A 646 A 1 B E 589 D 21 A <2 4 CD 824 A 12 C A 65 A 1 A 12 F 5 E 1 B <2 SEM Levels of significance Treatments (T) *** *** *** *** *** *** *** *** *** *** *** Days (D) *** *** *** *** *** *** *** *** *** *** *** T 9 D *** *** *** *** *** *** *** *** *** *** *** CK, control; LS, with the addition of LAB inoculant (Lactobacillus plantarum, 1 5 CFU per g fresh matter) and sucrose (2 g kg 1 fresh matter); LA, lactic acid; AA, acetic acid; PA, propionic acid; BA, butyric acid; NH -N, ammonia nitrogen; TN, total nitrogen; LA/AA, lactic to acetic acid ratio; WSC, water-soluble carbohydrates; LAB, lactic acid bacteria; ENT, Enterobacteria; SEM, standard error of the mean; Values with different superscript letters (a f and A F) are significantly different by Tukey s multiple comparison method (P < 5); ***P < 1. length of 4 bp per sequence for bacteria (Table 2). And these reads were clustered into 62 core OTUs based on 9% sequence identity (equal to species level). Another bacterial community richness estimator, Chao1, was utilized to estimate the OTUs number. These two indexes showed the decrease in richness of the bacterial community once alfalfa was ensiled. The diversity index of the bacterial community, Shannon index, was negatively affected by the inoculated treatment. The sampling completeness was evaluated by using the Good s coverage estimator, which calculates the probability that a randomly selected amplicon sequence from a sample has already been sequenced. The Good s coverage values were around 99, indicated that the sampling depth had adequately captured most of the bacterial community. Changes in the bacterial community at the genus level during the ensiling process are shown in Fig. 1. Before ensiling, the most abundant 16S rrna gene sequences at the genus level were associated with the Pantoea (62%) and the Enterobacter (185%). The genera Lactococcus and Clostridium both accounted for as low as 2% of total sequences. During ensiling, although, the bacterial community differed for each treatment, there were four predominant genera common to both treatments (the Lactobacillus, Pantoea, Enterobacter and Weissella). In both treatments, the Pantoea decreased in relative abundance with prolonged ensiling. In the control silage, the Enterobacter, Weissella and Lactobacillus increased in relative abundance after ensiling, peaking at, and days respectively; while an increase in abundance was observed for the Bifidobacterium throughout the ensiling period. In the inoculated silage, an increase in abundance was also observed for the Lactobacillus and Weissella throughout the ensiling period, while the relative abundance of the Enterobacter peaked at days. At the end of the sampling period, the Lactobacillus (24%), Weissella (2%) and Bifidobacterium (2%) together dominated the bacterial community of the control silage in contrast to the Lactobacillus (862%) for the inoculated silage. Evaluation at finer taxonomical levels was carried out to determine the distribution of the different bacteria. The ten most commonly occurring OTUs isolated from the bacterial community are shown in Fig. 1 and their phylogenetic relationships are shown in Fig. 2. The bacterial community of alfalfa shifted significantly during the ensiling process, from predominantly Pantoea agglomerans and Enterobacter sp. to mostly LAB. In the control silage, Weissella confusa, Lactobacillus brevis, Pediococcus acidilactici and Enterococcus mundtii developed rapidly once the herbage was ensiled at days. Other LAB species, such as Lactobacillus helsingborgensis, Lact. plantarum and Bifidobacterium asteroides, appeared at days. Weissella confusa, Lact. helsingborgensis and Bif. asteroides became the predominant LAB during the remaining 6 Journal of Applied Microbiology 122, The Society for Applied Microbiology

6 M.L. Zheng et al. Clostridial community of alfalfa silage Table 2 Statistics of high-throughput sequencing data and the bacterial community diversity with a special focus on the clostridial community Treatments Days Reads Length OTUs Shannon Chao1 Coverage Bacterial community CK LS Clostridial community CK LS CK, control; LS, with the addition of LAB inoculant (Lactobacillus plantarum, 1 5 CFU per g fresh matter) and sucrose (2 g kg 1 fresh matter); length, average read length (base pair); OTUs, number of operational taxonomic units; coverage, Good s coverage. storage period. In the inoculated silage, Lact. plantarum and W. confusa were the prevalent LAB throughout the ensiling period, with higher relative abundance observed for Lact. plantarum. Metagenomic analysis of the clostridial community dynamics The recovered reads isolated from the clostridial community summed up to 5, with an average length of 2 bp per sequence, and these reads were clustered into a total of 1 OTUs at 9% sequence similarity. Compared to the control, the inoculated silage obtained lower OTU number during initial days of ensiling, thereafter, a converse tendency occurred. The Chao1 estimator varied within the range of 16 and for the control and the inoculated silage respectively. After days of ensiling, the inoculated treatment increased the Shannon indexes compared to the control. Most of the clostridial community was detected as indicated by the Good s coverage values of 99. Changes in the clostridial community at the family and genus level during the ensiling process are shown in Fig.. Before ensiling, 6% of the clostridial 16S rrna gene sequences were associated with the family Clostridiaceae, followed by the family Peptostreptococcaceae (8%). The Clostridiaceae were primarily composed of members of the genera Clostridium (4%) and lesser amounts of the Garciella (96%) and the Sporacetigenium (26%). Of the Peptostreptococcaceae-associated sequences, the genera Romboutsia and Terrisporobacter were the dominant members, accounting for 244 and 84% of total sequences respectively. During ensiling, the relative abundance of sequences associated with the family Peptostreptococcaceae dropped significantly, with a concomitant significant increase in the family Cliostridiaceae. Although the clostridial community differed for each treatment, there were four predominant genera common to both treatments (the Clostridium, Garciella, Romboutsia and Terrisporobacter). In both treatments, the Romboutsia and Terrisporobacter decreased in relative abundance with prolonged ensiling, with an exception for the inoculated silage at and days. In the control silage, the Clostridium and Garciella increased in relative abundance after ensiling, peaking at and days respectively. In the inoculated silage, the Clostridium and Garciella also increased in relative abundance after ensiling, peaking at and days Journal of Applied Microbiology 122, The Society for Applied Microbiology 61

7 Clostridial community of alfalfa silage M.L. Zheng et al. 1 1 Relative abundance (%) Genus Lactobacillus Pantoea Enterobacter Weissella Bifidobacterium Enterococcus Brevundimonas Pseudomonas Pediococcus Buchnera Hafnia Bacillus Clostridium Lactococcus Others Relative abundance (%) OTUs OTU-B62 (Pantoea agglomerans) OTU-B4 (Enterobacter sp.) OTU-B (Weissella confusa) OTU-B6 (Lact. brevis) OTU-B1 (Pediococcus acidilactici) OTU-B11 (Lact. helsingborgensis) OTU-B1 (Lact. plantarum) OTU-B111 (Lact. kullabergensis) OTU-B2 (Bifidobacterium asteroides) OTU-B9 (Enterococcus mundtii) Others CK LS CK LS Figure 1 Relative abundance of the bacterial community at the genus level and the 1 most commonly isolated operational taxonomic units (OTUs) from the bacterial community. Silage samples are labelled with Latin letters indicating the type of treatments (CK, control; LS, with the addition of LAB inoculant (Lactobacillus plantarum, 1 5 CFU per g fresh matter) and sucrose (2 g kg 1 fresh matter)) and Arabic numerals indicating the day of ensiling. Lact., Lactobacillus. 1 OTU-B Lact. plantarum (KX15888) OTU-B1 1 Pediococcus acidilactici (KT96848) OTU-B6 1 1 Lact. brevis (KT5592) OTU-B OTU-B9 Enterococcus mundtii (KT6588) OTU-B111 1 Lact. kullabergensis (EF18241) 54 OTU-B11 9 Lact. helsingborgensis (KP1) 1 Weissella confusa (KT222) 1 OTU-B2 Bifidobacterium asteroids (KP119) 99 OTU-B4 Enterobacter sp. (KX18) 1 OTU-B62 98 Pantoea agglomerans (KF82) Figure 2 Phylogenetic tree showing the genetic relatedness of the 1 most commonly isolated operational taxonomic units (OTUs) from the bacterial community. The bar indicates 2% sequence divergence. Lact., Lactobacillus. respectively. At the end of the sampling period, the Clostridium (445%), Garciella (119%), Romboutsia (2%) and Terrisporobacter (9%) together dominated the clostridial community of the inoculated silage in contrast to the Clostridium (8%) for the control. The 1 most commonly occurring OTUs isolated from the clostridial community are shown in Fig. 4 and their phylogenetic relationships are shown in Fig. 5. Romboutsia sp. (244%), Clostridium tyrobutyricum (125%), Clostridium perfringens (15%), Garciella sp. (9%) and Terrisporobacter mayombei (84%) were the prevalent Clostridia species epiphytic on alfalfa, others such as Clostridium baratii (1%) and Clostridium sporogenes (11%) were also observed with lower relative abundance. A significant difference in the clostridial community dynamics was observed between the two treatments during the ensiling process. For the control silage, the relative abundance of Cl. perfringens, Garciella sp. and 62 Journal of Applied Microbiology 122, The Society for Applied Microbiology

8 M.L. Zheng et al. Clostridial community of alfalfa silage 1 1 Relative abundance (%) Family Clostridiaceae Peptostreptococcaceae Lachnospiraceae Eubacteriaceae Peptococcaceae Heliobacteriaceae Syntrophomonadaceae Halanaerobiaceae Ruminococcaceae Others Relative abundance (%) Genus Clostridium Garciella Romboutsia Terrisporobacter Sedimentibacter Anaerosalibacter Hespellia Sporacetigenium Acetobacterium Soehngenia Tepidimicrobium Peptoniphilus Roseburia Desulfotomaculum Others CK LS CK LS Figure Relative abundance of the clostridial community at the family and genus level. Silage samples are labelled with Latin letters indicating the type of treatments (CK, control; LS, with the addition of LAB inoculant (Lactobacillus plantarum, 1 5 CFU per g fresh matter) and sucrose (2 g kg 1 fresh matter)) and Arabic numerals indicating the day of ensiling. 1 Figure 4 Relative abundance of the 1 most commonly isolated operational taxonomic units (OTUs) from the clostridial community. Silage samples are labelled with Latin letters indicating the type of treatments (CK, control; LS, with the addition of LAB inoculant (Lactobacillus plantarum, 1 5 CFU per g fresh matter) and sucrose (2 g kg 1 fresh matter)) and Arabic numerals indicating the day of ensiling. Cl., Clostridium. Relative abundance (%) CK LS OTUs OTU-C15 (Romboutsia sp.) OTU-C (CI. perfringens) OTU-C1 (Garciella sp.) OTU-C84 (Cl. baratii) OTU-C4 (Cl. tyrobutyricum) OTU-C9 (Cl. sporogenes) OTU-C2 (Clostridium sp.) OTU-C24 (Sporacetigenium mesophilum) OTU-C2 (Terrisporobacter mayombei) OTU-C8 (Anaerosalibacter bizertensis) Others Cl. baratii increased rapidly, while in contrast Romboutsia sp., Cl. tyrobutyricum and T. mayombei decreased once the herbage was ensiled at days. Clostridium perfringens and Cl. baratii were the predominant Clostridia species at days. With prolonged ensiling, Cl. tyrobutyricum increased rapidly and became the most abundant Clostridia species during the remaining storage period. For the inoculated silage, the relative abundance of Cl. perfringens, Garciella sp. and Cl. tyrobutyricum increased while Romboutsia sp. and T. mayombei decreased at days. The overall composition of the 1 most commonly occurring OTUs isolated from the clostridial community changed very little between days, and a similar clostridial community was observed between pre-ensiled alfalfa and days of ensiling. Redundancy analysis Relationships between high-throughput sequencing data and fermentation process parameters are visualized in a RDA ordination plot shown in Fig. 6. In the view of the bacterial community, these data indicated that the Journal of Applied Microbiology 122, The Society for Applied Microbiology 6

9 Clostridial community of alfalfa silage M.L. Zheng et al OTU-C 51 Cl. perfringens (KU66) OTU-C84 Cl. baratii (KP944152) OTU-C2 Clostridium sp. (KM24498) 98 OTU-C4 Cl. tyrobutyricum (KP546) OTU-C9 1 Cl. sporogenes (KU1968) 9 OTU-C24 Sporacetigenium mesophilum (GU5991) 99 OTU-C15 Romboutsia sp. (LN9984) OTU-C2 98 Terrisporobacter mayombei (KU88612) 1 OTU-C8 Anaerosalibacter bizertensis (HG9644) OTU-C1 95 Garciella sp. (KR5468) Figure 5 Phylogenetic tree showing the genetic relatedness of the 1 most commonly isolated operational taxonomic units (OTUs) from the clostridial community. The bar indicates 1% sequence divergence. Cl., Clostridium. contents of lactic acid, acetic acid, butyric acid, WSC and NH -N, lactic to acetic acid ratio and ph all influenced the bacterial community. The ordination plot also suggested that three potentially different groups of silage samples existed. The first group, including pre-ensiled alfalfa, was far away from the other two groups. The second group, including all of the control silage samples, was positively correlated with the increasing contents of acetic acid, butyric acid and NH -N. The third group, including all of the inoculated silage samples, was positively correlated with increasing lactic acid content, increasing lactic to acetic acid ratio and decreasing ph. Silage samples were also clustered into three groups based on the relationship between the clostridial community and fermentation process parameters. The first group, including both silage treatments at and days, appeared to be influenced strongly by all of the ensiling factors as their positions were located away from these factors on the ordination plot. The second group, including the control silage at and days, was positively correlated with increasing contents of acetic acid, butyric acid and NH -N. The third group, including the remaining silage samples, was positively related to increasing contents of lactic acid and WSC, ph and lactic to acetic acid ratio. Discussion It has been well documented that high LAB counts and WSC content are needed to ensure a rapid and vigorous fermentation that results in a rapid decline in ph and a large amount of accumulated lactic acid, which are the key factors to ensuring good crop quality during the ensiling process (Ennahar et al. 2). In this study, the lactic acid bacterial fermentation proceeded rapidly by adding LAB inoculant and sucrose, the function of LAB inoculant promoted rapid and efficient utilization of WSC, which resulted in the intensive production of lactic acid and a rapid decrease in ph, and lead to a suppression of undesirable micro-organisms. Direct-cut alfalfa used in this study had a typically low DM, low WSC and high CP contents. Therefore, the control silage resulted in unsatisfactory preservation for which the relative content of lactic and acetic acids, the high ph, the high content of butyric acid and the promotion in the breakdown of protein to NH -N were indicators, using the thresholds described by Haigh and Parker (1985). Undesirable fermentation quality together with high Clostridia counts indicated the clostridial fermentation (Pahlow et al. 2). The differences in initial WSC content and the LAB numbers may lead to different variation trends in some fermentation qualities, microbial and chemical compositions parameters between two types of silages during ensiling, which could be responsible for the significant interactions between days of ensiling and treatments observed in this study. The RDA plot showed that the bacterial community of the pre-ensiled alfalfa group was clearly separated from those of silage group, indicating that significant changes in the bacterial community as well as ensiling parameters occurred when alfalfa was ensiled. RDA also showed distinct clusters of alfalfa silage by fermentation status, 64 Journal of Applied Microbiology 122, The Society for Applied Microbiology

10 M.L. Zheng et al. Clostridial community of alfalfa silage 1 CK- Bacteria 6 LA/AA CK- LS- WSC LS- Clostridia WSC RDA2 5 5 LS- ph CK- CK- CK- CK- BA CK- NH -N AA LA/AA LS- LS- LS- LS- LA LS- RDA2 6 CK- CK- LA LS- LS- ph LS- LS- CK- CK- BA CK- NH -N AA RDA1 RDA1 5 Figure 6 Redundancy analysis (RDA) ordination diagram of high-throughput sequencing data related to the bacterial community with a special focus on the clostridial community, with explanatory (e.g. ph, lactic acid content) variables represented as arrows and silage samples as circles (, CK) or triangles (, LS). LA, lactic acid; AA, acetic acid; BA, butyric acid; NH -N, ammonia nitrogen; LA/AA, lactic to acetic acid ratio; WSC, water-soluble carbohydrates. Silage samples are labelled with Latin letters indicating the type of treatments (CK, control; LS, with the addition of LAB inoculant (Lactobacillus plantarum, 1 5 CFU per g fresh matter) and sucrose (2 g kg 1 fresh matter)) and Arabic numerals indicating the day of ensiling. namely, the lactic acid bacterial fermentation group was clearly separated from the clostridial fermentation group. For instance, samples from the lactic acid bacterial fermentation group were positively correlated with increasing lactic acid content, increasing lactic to acetic acid ratio and decreasing ph, perhaps being indicative of the decrease in acid sensitive, undesirable aerobic and anaerobic micro-organisms, with the development of the LAB (McEniry et al. ). Nishino et al. (212) reported the appearance of the genera Pantoea and Enterobacter on guinea grass by DGGE analysis, however, it was unclear if the Pantoea and Enterobacter were the predominance, as DGGE analysis is not quantitative. McGarvey et al. (21) observed a predominance of the Enterobacter (426%) on alfalfa by 16S rrna gene libraries analysis, however, the relative abundance of the Pantoea was as low as 11%. In this study, the prevalent 16S rrna gene sequences isolated from the bacterial community of alfalfa were associated with the Pantoea (62%) and Enterobacter (185%), which were different from aforementioned results. Environmental factors which affect the colonization of plant by bacteria, such as climate, geographical location and type of fertilizer used, could explain this difference. It is well established that LAB play an important role in silage fermentation. The different species and the characteristics of epiphytic LAB might change and influence fermentation process (Yang et al. 21). Several studies have reported LAB as the dominant microbial community on forage crops, and some isolates have been identified as the genera Lactobacillus, Enterococcus, Pediococcus, Weissella, Lactococcus and Leuconostocs (Yang et al. 21). In this study, however, only low relative abundance (2%) of the Lactococcus was detected on alfalfa. Therefore, the addition of LAB inoculant was an effective way to improve alfalfa silage quality. During ensiling, most of the 16S rrna gene sequences were related to LAB, which became rather different between the two silages. Lactobacillus plantarum dominated the bacterial community of the inoculated silage throughout the ensiling period, W. confusa was also observed with low relative abundance. The rapid acidification and antagonistic activity of Lact. plantarum may have led to a fast fall in the bacterial community diversity of the inoculated silage. Such elimination of indigenous bacteria was also seen in a previous study (Parvin et al. Journal of Applied Microbiology 122, The Society for Applied Microbiology 65

11 Clostridial community of alfalfa silage M.L. Zheng et al. 21; Ogiy et al. 216), in which wilted Italian ryegrass was inoculated with Lact. plantarum and Lact. brevis. While in contrast, in the control silage, W. confusa, Lact. brevis, Ent. mundtii and Ped. acidilactici developed rapidly once alfalfa was ensiled at days. With prolonged ensiling, other LAB species such as Bif. asteroides, Lact. plantarum, Lact. helsingborgensis and Lactobacillus kullabergensis appeared. Compared to the inoculated silage, lower lactic acid content and higher acetic acid content were found in the control, likely due to the property of heterofermentative LAB W. confusa and Lact. brevis. Some homofermentative LAB, such as Lact. plantarum, were also reported to be involved in the increase in acetic acid content because of prolonged ensiling under WSC-deficient conditions (Lindgren et al. 199). Although the role of Bif. asteroides, Lact. helsingborgensis and Lact. kullabergensis in the ensiling fermentation process is unknown, as they were firstly found in silage, the relatively high ph and high acetic acid environment may have facilitated their growth. Bifidobacteria are known as anaerobic/microaerophilic and fermentative micro-organisms, which commonly inhabit the gastrointestinal tract of various animals and insects, and phylogenomic-based analyses suggested that Bif. asteroides ferment glucose and fructose to lactic and acetic acids through the characteristic fructose-6-phosphate shunt (Bottacini et al. 212). Two novel species, Lact. helsingborgensis and Lact. kullabergensis, were isolated from the honey stomach of the honeybee recently, genetic and biochemical tests revealed that both of the two species were facultatively anaerobic, growth occurs at ph 12 and D-lactate is produced as the end product from hexoses (Olofsson et al. 2), however, further research is need to be performed to confirm that whether acetic acid is produced by these two species. Although there may be a progression of LAB from homofermentative to heterofermentative species (Li and Nishino 211), it is worth examining how Bifidobacterium asteroides, Lact. helsingborgensis and Lact. kullabergensis increased in late ensiling. Lactobacillus buchneri, which is known to be involved in anaerobic lactic acid degradation to acetic acid, was not detected during high-throughput sequencing analysis in this study. A higher content of acetic acid than lactic acid was also observed during the ensiling of direct-cut guinea grass which is of low WSC content (Nishino et al. 212), although the bacteria associated with enhancement of acetic acid production could not yet be clearly identified. Nevertheless, these findings suggested that low WSC content combined with low DM content of pre-ensiled materials enhanced acetic acid production, which decreased ph slowly, thereby stimulating the clostridial fermentation (Pahlow et al. 2). In this study, high-throughput sequencing data, which were based on the bacterial universal primer set 6-F and 86-R, revealed that the relative abundance of sequences associated with Clostridia was below 11% of total sequences. Furthermore, only the genus Clostridium was detected using this bacterial universal primer set, lots of functional Clostridia belonging to other genera could be underestimated. As reported in a previous study (Heilig et al. 22), universal primer sets would not work to access the clostridial community in various complex microbial ecosystems if the contents of some functional Clostridia were <1%. Therefore, the accurate clostridial community and their concrete potential functions in ensiling fermentation process were still not entirely clear. To target the higher diverse clostridial community as far as possible, designing a novel pair of specific and universal clostridial primers is the critical step. Recently, a new pair of specific primers (SJ-F and SJ-R) targeting the more diverse species within the class Clostridia was designed by Hu et al. (2). They also authenticated the specificity and universality of this novel primer set through theoretical and experimental evaluations, and the results suggested that the clostridial community in complex microbial ecosystems could be assessed using this primer set. Among all clostridial genera epiphytic on alfalfa, the genus Clostridium was the biggest group accounting for 4%, and others were Romboutsia (244%), Garciella (96%), Terrisporobacter (84%) and Sporacetigenium (26%). Nishino et al. (212) reported the appearance of the Clostridium on grass, while in contrast the others were found as unexpected clostridial genera present on fermenting plant material, although they have previously been detected in environmental samples including soil and faeces. It was in agreement with Hu et al. (2), who reported that this novel Clostridia-specific primer set combined with DGGE analysis revealed five unexpected clostridial genera present in fermented mud. This mainly because the novel specific primer pair can specifically amplify the 16S rrna gene from the Clostridia species, even they were non-dominant bacteria. The cell sap of plants is an ideal substrate for clostridial growth under the anaerobic conditions that are rapidly reached in adequately consolidated and effectively sealed silage. Hengeveld (198) found that Clostridia counts on grass at the time of ensiling had little influence on silage quality. Rammer (1996) reported that effect of some epiphytic Clostridia species on silage quality was unclear, namely, infecting grass with Cl. tyrobutyricum negatively affected the quality of grass silage in the first year while no effect was seen in the second year. Variance in the role of the epiphytic Clostridia on silage quality was observed in this study. The genera Clostridium and Garciella were mainly 66 Journal of Applied Microbiology 122, The Society for Applied Microbiology

12 M.L. Zheng et al. Clostridial community of alfalfa silage associated with the enhanced clostridial fermentation, as their summarized relative abundance increased with prolonged ensiling, while in contrast the rest genera made little contribution to the clostridial fermentation. Rossi and Dellaglio (2) reported that Cl. perfringens played the dominant role in the clostridial fermentation of Verona (Italian) farm-made alfalfa silage (seven samples) at, 4 or 5 days, while Clostridium saccharolyticum were exclusively isolated from Molise (Italian) farm-made alfalfa silage (two samples) at 6 days. In this study, the clostridial community dynamics over the course of ensiling was determined. Clostridium perfringens, Garciella sp. and Cl. baratii were the main initiators of the clostridial fermentation of alfalfa silage, as their relative abundance increased rapidly once the herbage was ensiled at days. With prolonged ensiling, Cl. perfringens and Cl. baratii increased, while in contrast Garciella sp. decreased. Clostridium tyrobutyricum became the most abundant Clostridia after days of ensiling, making great contribution to the enhanced clostridial fermentation. Our findings were inconsistent with aforementioned results. This was likely due to the differences in environmental factors which can influence fermentation products and the bacterial community, including crop species, ensiling condition, ensiling scale and storage period (Li and Nishino 211). Although it entails substantial work, research on more undesirable alfalfa silage samples is needed to have greater insight into the clostridial community dynamics during the clostridial fermentation. Nevertheless, our observations are promising as high-throughput sequencing based on a Clostridia-specific primer set is a more sensitive method ideal for characterization and comparison of the clostridial community compared to culture-based and some other molecular-based methods (Neher et al. 21). Although, there are some limitations in the use of highthroughput sequencing to compare the true diversity of the microbial community, due to species differences in 16S rrna gene copy number and biases resulting from DNA extraction and PCR amplification (Jaenicke et al. 211). However, as all samples in this study were subject to the same biases throughout all experimental procedures it is still possible to compare between samples on a relative basis. In this study, the genus Garciella was firstly found in silage. Garciella species have previously been detected in different anaerobic habitats, especially in a methanogenic bioreactor (Jaenicke et al. 211). Interestingly, a reference species of the Garciella, namely Garciella nitratireducens, was isolated from an oil well and represented an anaerobic, lactic acid-utilizing and nitrate-reducing bacterium (Miranda-Tello et al. 2). In addition to lactic acid, acetic acid and butyric acid were produced when glucose is fermented by G. nitratireducens. Furthermore, G. nitratireducens reduces nitrate to ammonium, the decrease in nitrate content probably stimulates the growth of Clostridia. As reported in previous studies (Spoelstra 198; Namihira et al. 21), nitrate, by the intermediates nitrite and nitric oxide, has a selective inhibition of clostridial growth under the reducing conditions of silage fermentation. The extent to which nitrite and nitric oxide accumulates depends on the initial nitrate content. If this is low, most of the nitrite and nitric oxide reduced soon after it has formed, thereby limiting the inhibitory effect on Clostridia. In this regard, G. nitratireducens can in part explain the role of the Garciella in the clostridial fermentation. However, bacterial characteristics differ among species (Borreani and Tabacco 215). Therefore, the specific function of the Garciella members in the clostridial fermentation are still unclear, it is worth isolating and evaluating the Garciella in alfalfa silage. Although high-throughput sequencing has great potential in the microbial community analysis, as with any community molecular-based methods, problems exist in providing quantitative data on the metabolic activity (e.g. live, inactive spore and dead) of the microbial community. Microbial composition is presented as relative abundance in extracted DNA and although partial fragmentation may occur during ensiling, DNA may be retained at a size sufficient for detection (Chiter et al. 2). Therefore, the specific functions of the Clostridia members dominating the clostridial community were not known. Further experiments, including isolation of Cl. perfringens, Garciella sp., Cl. baratii and Cl. tyrobutyricum and examination of the effect of inoculating alfalfa silage with these Clostridia species, are now in progress to complement the results of this study. Most studies in which Clostridia were isolated from silage were directed at problems commonly seen with late blowing of cheese (Garde et al. 211). Bacteriocin-producing LAB strains, such as Lactobacillus paracasei complex and Lactococcus lactis IFPL 59, have been screened for control of late blowing of cheese by their ability to inhibit the growth of Clostridia (Christiansen et al. 25; Carmen Martinez-Cuesta et al. 21). Similarly, in future research, we hope that bacteriocin-producing LAB strains can be screened and used as inoculants to improve alfalfa silage quality by suppressing the clostridial fermentation. The disappearance of epiphytic bacteria, which had lower adaptability to the anaerobic and acidic condition of silage, could explain the decrease in richness (OTU number, Chao1 estimator) of the bacterial community of alfalfa silage compared to pre-ensiled alfalfa. Lower Shannon diversity of the bacterial community was observed in treated silage compared to the control (Li et al. 215), which was consistent with the results of this study. With Journal of Applied Microbiology 122, The Society for Applied Microbiology 6