FEMS Microbiology Ecology http://mc.manuscriptcentral.com/fems Temporal changes in microbial communities attached to forages with different lignocellulosic compositions in the cattle rumen Journal: FEMS Microbiology Ecology Manuscript ID FEMSEC-19-03-0119.R8 Manuscript Type: Research article Date Submitted by the Author: 10-Apr-2020 Complete List of Authors: Gharechahi, Javad; Agricultural Biotechnology Research Institute of Iran, Systems Biology Vahidi, Mohammad Farhad; Agricultural Biotechnology Research Institute of Iran, Syst ms Biology Ding, Xue-Zhi ; Chinese Academy of Agricultural Sciences Lanzhou Institute of Husbandry and Pharmaceutical Sciences, Key Laboratory of Yak Breeding Engineering Han, Jianlin; International Livestock Research Institute, Livestock Genetics Program Salekdeh, Ghasem Hosseini; Agricultural Biotechnology Research Institute of Iran, Systems Biology Keywords: microbiome, rumen, biomass degradation ScholarOne Support 1-434/964-4100 vie w e r R Pe e r Fo Page 1 of 50 FEMS Microbiology Ecology 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 ScholarOne Support 1-434/964-4100 iew ev er R e r P Fo FEMS Microbiology Ecology Page 4 of 50 1 2 3 4 Temporal changes in microbial communities attached to forages with different lignocellulosic compositions in 5 6 the cattle rumen 7 8 9 10 Javad Gharechahi111 , Mohammad Farhad Vahidi 2, Xue-Zhi Ding3, Jian-Lin Han4,5, Ghasem Hosseini Salekdeh2 12 13 14 15 1. Human Genetics Research Center, Baqiyatallah University of Medical Sciences, Tehran, Iran 16 17 2. Department of Systems Biology, Agricultural Biotechnology Research Institute of Iran, Agricultural Research, 18 Education, and Extension Organization, Karaj, Iran 19 20 3. Key Laboratory of Yak Breeding Engineering, Lanzhou Institute of Husbandry and Pharmaceutical Sciences, 21 Chinese Academy of Agricultural Sciences (CAAS), Lanzhou, China 22 23 4. CAAS-ILRI Joint Laboratory on Livestock and Forage Genetic Resources, Institute of Animal Science, Chinese 24 Academy of Agricultural Sciences (CAAS), Beijing, China 25 5. Livestock Genetics Program, International Livestock Research Institute (ILRI), Nairobi, Kenya 26 27 28 29 30 To whom correspondence should be addressed 31 32 33 Ghasem Hosseini Salekdeh 34 35 E-mail: hsalekdeh@yahoo.com 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 1 60 ScholarOne Support 1-434/964-4100 vie w er Re Pe Fo r Page 5 of 50 FEMS Microbiology Ecology 1 2 3 Abstract 4 5 6 The attachment of rumen microbes to feed particles is critical to feed fermentation, degradation and digestion. 7 However, the extent to which the physicochemical properties of feeds influence the colonization by rumen microbes 8 9 is still unclear. We hypothesized that rumen microbial communities may have differential preferences for attachments 10 to feeds with varying lignocellulose properties. To this end, the structure and composition of microbial communities 11 attached to six common forages with different lignocellulosic compositions were analyzed following in situ rumen 12 13 incubation in male Taleshi cattle. The results showed that differences in lignocellulosic compositions significantly 14 affected the inter-sample diversity of forage-attached microbial communities in the first 24 hours (h) of rumen 15 16 incubation, during which the highest dry matter degradation was achieved. However, extension of the incubation to 17 96 h resulted in the development of more uniform microbial communities across the forages. Fibrobacteres were 18 19 significantly overrepresented in the bacterial communities attached to the forages with the highest neutral detergent 20 fiber contents. Ruminococcus tended to attach to the forages with low acid detergent lignin contents. The extent of dry 21 matter fermentation was significantly correlated with the populations of Fibrobacteraceae, unclassified Bacteroidales, 22 23 Ruminococcaceae and Spirochaetacea. Our findings suggested that lignocellulosic compositions, more specifically 24 the cellulose components, significantly affected the microbial attachment to and thus the final digestion of the forages. 25 26 27 28 29 Keywords: microbiome, microbiota, rumen, rumen incubation, 16S rRNA gene sequencing, rumen fermentation, 30 31 biomass degradation 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 2 60 ScholarOne Support 1-434/964-4100 vie w Re ee r or P F FEMS Microbiology Ecology Page 6 of 50 1 2 3 Introduction 4 5 6 The gastrointestinal tracts (GIT) of ruminant animals have evolved to allow the colonization by a diverse community 7 of symbiotic microorganisms belonging to three taxonomic domains of life, i.e. Archaea, Bacteria and Eukarya. 8 9 Particularly, as the first and the largest compartment of ruminant stomach, rumen is the main site of microbial 10 colonization and fermentation. Without the aid of such microorganisms, the host ruminants cannot digest and convert 11 plant lignocellulosic biomasses into energy and other essential metabolites (Jami and Mizrahi 2012; Mackie 2002). It 12 13 is estimated that almost 70% of the energy requirement of ruminants is supplied by this microbial fermentation 14 (Bergman 1990). In addition to their crucial role in animal nutrition and production, the GIT microorganisms remain 15 16 vital to animal health, physiology and immunity against pathogenic microbes (Guarner and Malagelada 2003; Hungate 17 2013). 18 19 20 Bacteria are the major colonizers in the rumen and consequently make the greatest contribution to plant biomass 21 fermentation, degradation and digestion. Early studies on rumen microbial communities relied largely on culture- 22 based approaches, which were limited to the bacteria that can grow on culture medium. With the advances in next- 23 24 generation sequencing technologies, culture-independent approaches have gained preference, enabling researchers to 25 achieve a deeper insight into precise composition and structure of rumen bacterial communities. The latest high 26 27 throughput amplicon-based 16S ribosomal RNA gene (rRNA) sequencing studies suggest that rumen bacterial 28 communities of most ruminants are mainly affiliated to the phyla of Bacteroidetes, Firmicutes, Proteobacteria, 29 30 Fibrobacteres, Spirocheaetes, Actinobacteria, Tenericutes and Verrucomicrobia, which collectively account for 31 greater than 99% of the total rumen bacterial communities (Godoy-Vitorino et al. 2012; Jami et al. 2013; Jami and 32 33 Mizrahi 2012; Zened et al. 2013). Microbial communities are dynamic in the rumen and depend on the host species 34 and its diet, age, physiology and health status (Godoy-Vitorino et al. 2012; Jami et al. 2013; Jami and Mizrahi 2012; 35 Kocherginskaya et al. 2001; Kong et al. 2010; Petri et al. 2012). Under normal physiological conditions, diet is the 36 37 major driver to determine the composition and structure of rumen bacterial communities (Petri et al. 2012). 38 Interestingly, the amount of dietary fiber is a key factor shaping the growth and multiplication of cellulolytic bacteria 39 40 in the rumen. Changing dietary fiber content or substituting the fiber with easily fermentable carbohydrates has a 41 profound impact on the community of rumen microbiota and may result in metabolic diseases, such as subacute 42 43 ruminal acidosis (Khafipour et al. 2009; Petri et al. 2017). 44 45 Members of the rumen bacterial communities differ in their preferences for attachment to feed particles and rumen 46 wall, therefore they are accordingly categorized into particle-attached (tightly attached to feed particles), liquid borne 47 48 (freely available in liquid fraction) and epimural (attached to rumen epithelium) communities (De Mulder et al. 2017; 49 Gharechahi et al. 2015; Kong et al. 2010; Sadet et al. 2007). The attachment of rumen microbial communities to feed 50 51 particles is a key step in the process of rumen fermentation and digestion (McAllister et al. 1994) and it occurs shortly 52 after feed entry into the rumen. Analysis of microbial communities associated with perennial ryegrass following in 53 54 situ rumen incubation in Holstein–Friesian cows or steers shows that microbial attachment initiates within five minutes 55 (min) of its rumen entry and stabilizes between 15 and 30 min of its rumen incubation (Edwards et al. 2007; Huws et 56 57 58 59 3 60 ScholarOne Support 1-434/964-4100 vie w e er R or Pe F Page 7 of 50 FEMS Microbiology Ecology 1 2 3 al. 2013). A recent analysis of temporal changes in bacterial community attached to wheat straw in Holstein cow 4 5 rumen demonstrates that the first wave of microbial-based biomass degradation occurs within 30 min of its rumen 6 entry. The community of bacteria attached to feed particles during this time retains even after 72 hours (h) of the 7 8 rumen incubation (Jin et al. 2018). The degree of dry matter (DM) degradation differs among forages during the initial 9 hours of their rumen entry (Cheng et al. 2017; Liu et al. 2016). The community composition of particle-attached 10 microbes also varies among feeds and is likely influenced by the chemical compositions of feeds because cellulolytic 11 12 bacteria such as Fibrobacter, Ruminocuccus, Butyrivibrio and unclassified Treponema tend to attach to feeds with 13 relatively high neutral detergent fiber (NDF, Cheng et al. 2017; Liu et al. 2016). The community of feed particle- 14 15 attached bacteria also changes over the incubation time. For instance, rumen microbiota attachment to switchgrass in 16 Friesian cow occurs at two successive stages: The first wave takes place immediately after its rumen entry (within the 17 18 first 30 min) and is characterized by high abundances of Bacteroidia and Clostridia. The second wave, however, occurs 19 after 16 h of its rumen incubation, during which the populations of Spirochaeta and Fibrobacteria become dominant 20 (Piao et al. (2014). 21 22 23 There is limited knowledge on rumen bacteria diversity, their preference for attachment to and degradation of feeds 24 with extreme values of cellulose and/or hemicellulose. Understanding the dynamics of bacteria attached to feeds at 25 26 high level of lignification provides opportunities to improve the nutrient efficiency of low-quality forages through 27 pre-treatments or manipulation of rumen microbial communities. We hypothesized that rumen bacterial communities 28 29 differ in lignocellulose degrading capacities and thus show preference for attachment to feeds with different levels of 30 lignification. We therefore aimed to evaluate whether the rumen microbes have any preference for attachment to 31 32 forages with different cellulose and/or hemicellulose contents. We also explored the dynamic changes in microbial 33 communities attached to the forages under prolonged incubation in the cattle rumen (e.g. up to 96 h with 24 h sampling 34 intervals). Overall, our 16S rRNA genes-based diversity analysis revealed significant differences in the composition 35 36 and structure of microbial communities attached to different forages. 37 38 Materials and methods 39 40 41 In situ rumen incubation and sample collection of the forages 42 43 All experimental procedures relevant to animals were approved by the Ethics Committee for Animal Experiments of 44 45 the Animal Science Research Institute of Iran. Rumen cannulation was performed according to the American College 46 of Veterinary Surgeons (ACVS) using a two-stage rumen cannulation technique (Martineau et al. 2015). 47 48 Six common lignocellulosic forages, including camelthorn (Alhagi persarum, AP; both stem and leaves), common 49 50 reed (Phragmites australis, CR; stem and leaves), date palm (Phoenix dactylifera, DP; leaves), kochia (Kochia 51 scoparia, KS; both stem and leaves), rice straw (Oryza sativa; cultivar Hashemi, RS; both stem and leaves) and 52 53 salicornia (Salicornia persica, SC; both stem and leaves), were selected for in situ rumen incubation. Dried forage 54 materials were cut into pieces of approximately 2 mm in length and an equal amount of the pieces (5 ± 0.05 g) was 55 56 weighed into heat-sealed nylon bags (5 × 10 cm; 50 μm pore size). Three rumen-cannulated, purebred bulls (Taleshi 57 58 59 4 60 ScholarOne Support 1-434/964-4100 w ev ie r R Pe e Fo r FEMS Microbiology Ecology Page 8 of 50 1 2 3 cattle, an Iranian local breed with a historical origin of Bos taurus mixed with Bos indicus) aged between 2.5 and 3 4 5 years were used for this study. Forty-eight heat-sealed bags, eight per forage, were simultaneously placed into each of 6 the three rumens shortly after the cannulated bulls were offered the first meal in the morning. The bulls were housed 7 8 together in a stable, fed on a mixed diet containing 70% wheat straw and 30% concentrate and provided free access 9 to water. Two nylon bags from each of the six forages were retrieved from each rumen after 24, 48, 72 and 96 h of the 10 incubation, washed thoroughly with distilled water three times to remove liquid borne and loosely attached microbes, 11 12 which may not have a discriminating preference for attachment to forages with different lignocellulose properties, and 13 finally squeezed by hands with sterile gloves to remove excess water. The bags were then transferred in liquid nitrogen 14 15 to the laboratory where one replicate was stored at -70 °C for subsequent DNA extraction while the other was 16 processed for physicochemical analysis. 17 18 19 Physicochemical analysis of the incubated forage materials 20 21 The forage lignocellulosic biomasses were analyzed for dry matter (DM) and the contents of neutral detergent fiber 22 (NDF), acid detergent fiber (ADF) and acid detergent lignin (ADL) before and after the rumen incubation. DM was 23 24 determined following 48 h air-drying of the samples in a fan-assisted oven maintained at 55 °C. The dried material 25 was grounded to pass through a 1-mm sieve for the measurements of NDF, ADF and ADL according to the established 26 27 procedures (Goering 1970; Van Soest et al. 1991). Cellulose content was estimated by subtracting ADL from ADF 28 while hemicellulose content was measured by subtracting ADF from NDF. 29 30 31 Microbial cell recovery and metagenomic DNA extraction 32 33 Microbial cells firmly attached to the forages were subsequently stripped by incubating individual samples on ice in 34 35 a dissociation buffer containing 0.1% (v/v) Tween 80, 1% (v/v) methanol and 1% (v/v) tertiary butanol (adjusted at 36 pH 2), which has been particularly adapted for the dissociation of microbial cells from the rumen solid digesta. The 37 samples were vigorously vortexed every 1 min and this step was repeated at least 5 times. The forage materials were 38 39 sedimented by centrifugation at 500  g and the liquid supernatant containing microbial cells was transferred to a new 40 container. This step was repeated at least three times and the collected liquids were centrifuged at 12000  g for 10 41 42 min to sediment the detached microbial cells. Metagenomic DNA from the detached cells was extracted using the 43 QIAamp® DNA Stool Mini Kit (Qiagen, Hilden, Germany) according to the manufacturer’s protocol for the isolation 44 45 of DNA from stools for pathogen detection. The quality and quantity of the extracted DNA were evaluated by an 46 agarose gel electrophoresis (0.8%) and a Nanodrop spectrophotometer (Thermo Scientific, Wilmington, DE, USA). 47 48 49 PCR amplification and Illumina sequencing of 16S rRNA gene 50 51 The V3-V4 hypervariable region of 16S rRNA gene was amplified using the universally conserved primer set S-D- 52 53 Bact-0341-b-S-17 (5′-CCT ACG GGN GGC WGC AG-3′) and S-D-Bact-0785-a-A-21 (5′-GAC TAC HVG GGT 54 ATC TAA TCC-3′), which generated a fragment of 464 bp suitable for paired-end sequencing using the Illumina 55 MiSeq System (Illumina Inc. San Diego, CA, USA). PCR amplification was performed in triplicate in a 25 L reaction 56 57 58 59 5 60 ScholarOne Support 1-434/964-4100 vie w er Re Pe Fo r Page 9 of 50 FEMS Microbiology Ecology 1 2 3 containing 12.5 L 2 PCR master mix (Qiagen), 1 L (10 pM) of each primer, 30 ng of microbial DNA and 5-9 L 4 5 of double-distilled water. The PCR condition consisted of an initial denaturation at 94 °C for 4 min followed by 30 6 cycles of 94 °C for 30 s, 55 °C for 30 s and 72 °C for 30 s, and a final extension at 72 °C for 5 min. PCR products 7 8 were purified and 2 L of each reaction was used as a template for the second round of PCR, during which the Illumina 9 adaptors and barcode sequences were incorporated to the 5’-end of the amplified products. The second PCR was also 10 11 performed in triplicate under the same running condition except the number of cycles were limited to 15. The PCR 12 amplicons were recovered using the Qiaquick® Gel Extraction Kit (Qiagen), quantified fluorometrically, pooled in 13 equimolar quantities and paired-end sequenced (PE300) using the Illumina MiSeq System at Macrogen Inc. (Seoul, 14 15 South Korea). 16 17 Sequence analysis 18 19 20 Quality filtered paired-end sequences were joined using the Flash with --max-overlap option set to 200 (Magoc and 21 Salzberg 2011). Sequences failed to be assembled were discarded. The joined FASTQ files were processed using 22 23 split_libraries_fastq.py script in the QIIME pipeline v1.9.1 (Caporaso et al. 2010b). This script demultiplexed 24 sequences and filtered out sequences shorter than 200 bp or longer than 1000 bp, along with those containing 25 26 ambiguous bases with a mean quality score < 20, having runs of six or more of the same nucleotides, carrying a 27 missing quality score and including > two mismatches from the primer sequences. The multiplexed sequences were 28 searched against the latest Ribosomal Database Project (RDP release 11.5, containing 3,356,809 16S rRNA sequences) 29 30 to identify and discard chimeric sequences using the VSEARCH v2.8.3 operated under default setting (Rognes et al. 31 2016). Non-chimeric sequences were then used to pick operational taxonomic units (OTUs) using the 32 33 pick_de_novo_otus.py script in the QIIME pipeline as described in details in our previous paper (Gharechahi et al. 34 2017). OTUs were defined at 97% identity using the Uclust (Edgar 2010). The most abundant sequence in each OTU 35 36 cluster was selected as being a representative; and these sequences were then aligned against the Greengenes core set 37 (gg_13_8; (DeSantis et al. 2006)) using the PyNAST aligner with a minimum sequence identity of 75% (Caporaso et 38 al. 2010a). Taxonomies were assigned to each OTU using the Ribosomal Database Project naïve Bayesian classifier 39 40 (Wang et al. 2007) by applying a minimum confidence value of 0.8. The OTU table was filtered for low abundant 41 OTUs using the filter_otus_from_otu_table.py script with --min_count_fraction option set to 0.00001 (discarding 42 43 OTUs represented by < 0.001% of the sequences) and then rarefied to the sequencing depth at 4660 reads 44 corresponding to the number of reads in the sample with the smallest set of sequences. Rarefaction plots and alpha 45 46 diversity indices, including Shannon, Simpson, Good’s_coverage and Chao1, were calculated using the 47 core_diversity_analyses.py script in the QIIME pipeline. Beta diversity indices, including weighted and unweighted 48 49 Unifrac phylogenetic distance matrices, were constructed with the rarefied OTU table as input and visualized through 50 the principal coordinate analysis (PCoA) plots in the MicrobiomeAnalyst web server (Dhariwal et al. 2017). 51 52 Statistical analysis 53 54 55 56 57 58 59 6 60 ScholarOne Support 1-434/964-4100 w vie er Re or Pe F FEMS Microbiology Ecology Page 10 of 50 1 2 3 Statistically significant differences in physicochemical data, including DM, NDF, ADF, ADL, cellulose and 4 5 hemicellulose contents, were analyzed by one-way ANOVA using the general linear model (GLM) procedure in the 6 SAS software v9.3 (SAS Institute Inc., Cary, NC, USA). Permutational multivariate analysis of variance 7 8 (PERMANOVA) was performed using the adonis function of R-package vegan v2.5-5 to test for significant 9 differences between community compositions of forage-attached microbial communities. In addition, permutation 10 multivariate analysis of group dispersions (PERMDISP) based on the betadisper function of the vegan was used to 11 12 test for the homogeneity of dispersions (variances). Differences in taxa abundances among forages and sampling 13 intervals were estimated using analysis of composition of microbes (ANCOM) based on relative abundances of OTUs 14 15 summarized at various taxonomic levels (Mandal et al. 2015). Means were compared by Duncan's Multiple Range 16 test (DMRT) in PAST v3.26 given Bonferroni p-value cutoff < 0.05 (Hammer et al. 2001). Error correction was done 17 18 based on the number of groupwise comparisons performed at each taxonomic level. The Pearson’s correlation analysis 19 was performed using the corr.test function of the psych package v1.8.12 and p-values were corrected using Bonferroni 20 method based on the total number of correlations calculated for each variable separately. For all tests, p-values less 21 22 than 0.05 were considered statistically significant. 23 24 Results 25 26 27 Physicochemical properties of the rumen-incubated forages 28 29 The six forages were analyzed for the contents of NDF, ADF, ADL, cellulose and hemicellulose before their rumen 30 31 incubation (Figure 1). They showed different NDF contents (p < 0.05), being the highest in common reed (CR) but 32 the lowest in both camelthorn (AP) and rice straw (RS). Most of them also contained different ADF contents (p < 33 34 0.05), being the highest in AP but the lowest in both RS and salicornia (SC). The contents of ADL also differed 35 significantly between the forages (p < 0.05), with AP being the highest at 2 times that in date palm (DP), three times 36 those in SC, CR and kochia (KS; the latter two with similar contents at p > 0.05) and 6 times the lowest value in RS 37 38 (24.86% vs 4.06%). CR carried the highest cellulose followed by DP, RS, KS and SC while AP contained the lowest 39 cellulose (p < 0.05). SC possessed the highest amount of hemicellulose followed by CR, KS, RS, DP and AP. Overall, 40 41 all the six forages had different physicochemical properties in terms of their relative contents of hemicellulose, 42 cellulose and lignin. 43 44 45 Lignocellulosic biomass degradation following the rumen incubation 46 47 The six forages were monitored for their changes in the contents of DM, NDF, ADF, ADL, cellulose and hemicellulose 48 during the rumen incubation (Figure 1 and Figure S1). DM degradation was the fastest in AP but the slowest in CR 49 50 (42% vs 34%, p < 0.05) during the first 24 h of the incubation. DM degradation was completed in AP at 24 h but 51 continued in CR and KS until 48 h, and in DP, RS and SC up to 96 h of the incubation, with RS having the highest 52 53 degraded DM (66%) followed by SC (56%) and DP (53%). AP, DP and KS demonstrated similar trends of 54 significantly increased NDF, ADF, ADL and cellulose contents (p < 0.05), mirrored in their patterns of DM 55 56 degradations, mostly within the first 24 h, but their hemicelluloses decreased continuously (p < 0.05 in most cases) 57 58 59 7 60 ScholarOne Support 1-434/964-4100 w ev ie r R Pe e r Fo Page 11 of 50 FEMS Microbiology Ecology 1 2 3 following the incubation. However, the patterns of these five fiber-related parameters were significantly segregated 4 5 among CR, RS and SC during the incubation. Although ADF and ADL were significantly increased but celluloses 6 were remarkably declined in CR and SC (p < 0.05), their NDF and hemicelluloses showed contrasting patterns, being 7 8 significantly reduced in CR while steadily accumulated in SC (p < 0.05) throughout the incubation. ADF, cellulose 9 and hemicellulose in RS maintained stable levels while its NDF and ADL were slightly increased (p < 0.05) along the 10 incubation. It was apparent that the initial differences in fiber-related physicochemical properties significantly affected 11 12 the rumen digestion of the six forages. 13 14 16S rRNA gene sequencing 15 16 17 The paired-end sequencing of PCR amplicons from the V3-V4 region of 16S rRNA gene resulted in 5,945,300 pairs 18 of raw sequences (averaged at 82,573 sequences per sample) with an average length of 300 bp. Sequences were joined 19 20 into 4,421,604 full-length amplicons at an average of 61,411  38,275 per sample. The uneven sequencing depths 21 across samples may be due to the true differences in microbial abundance of samples or the technical variations 22 23 introduced during library preparation and sequencing. Although the numbers of sequences varied greatly among the 24 forages and across the lengths of the rumen incubation, there was a general pattern of a steady increase from the lowest 25 at 24 h (averaged at 44,100 per sample) to the highest at 72 h (averaged at 95,823 per sample). 26 27 28 Sequencing data analysis 29 30 Before processing the amplicon sequences for OTU picking, they were subjected to a single round of quality filtering, 31 32 resulting in 3,602,510 high quality sequences, of which 994,147 (27%) were identified to be chimeric and thus 33 discarded from further analyses. The qualified sequences (2,493,285) were clustered at 97% similarity level into 34 35 110,804 OTUs, of which 106,982 (96.4%, representing 202,403 sequences) were labeled as low abundant features 36 (e.g. those representing reads with frequencies less than 0.001% of the total sequences) and therefore filtered out from 37 the OTU table. Finally, 3,822 clean OTUs representing 2,290,882 sequences were subjected to further downstream 38 39 analyses. To assess whether our sequencing effort provided sufficient sequencing depths to describe the diversity of 40 forage-attached microbiotaes, rarefaction curves describing the numbers of observed OTUs, species richness (Chao1), 41 42 Shannon and Simpson diversity at various sequencing depths were generated for all samples (Figure S2a-d). 43 Rarefaction analysis based on observed species and species richness revealed incomplete sampling of microbiota 44 45 attached to the forages and thus indicated that highly diverse microbial communities attached to the forages with 46 different lignocellulose properties(Figure S2a and b). However, Shannon (Figure S2c) and Simpson indices (Figure 47 48 S2d) also showedreached plateaus, indicating that the majority of the diversity was explored that forage-attached 49 microbiota had not been evenly sampled. 50 51 Microbial diversity analysis 52 53 54 Differences in alpha diversity indices of forage-attached microbiota were mostly limited to the first 24 h of the rumen 55 incubation, during which the maximal differences in microbial attachment occurred (Figure S3). At this time, CR and 56 57 58 59 8 60 ScholarOne Support 1-434/964-4100 iew ev er R r P e Fo FEMS Microbiology Ecology Page 12 of 50 1 2 3 RS (two forages with the highest initial cellulose contents) had the lowest while AP and KS (two forages with the 4 5 lowest initial cellulose contents) had the highest average observed species and species richness. CR and RS also 6 showed the lowest average Shannon index, indicating their limited microbial diversity (p < 0.05). Diversity indices in 7 8 almost all forages were not affected by the incubation length (Figure S4) except that CR had very low alpha diversity 9 measures at 24 h of its incubation (p < 0.05). All samples showed a high Good’s coverage (> 0.91, data not shown) at 10 all sampling intervals, indicating that our sequencing effort figured out > 90% of the microbial diversity attached to 11 12 the forages. However, the uneven sequencing depths did not allow us to fully explore the diversity of forage-attached 13 microbiota. Considering potential random errors due to limited numbers of experimental animals and replicates in this 14 15 study, such differences in alpha diversity indices among the forages were not robust enough to be interpreted with any 16 biological significance. 17 18 19 Beta diversity analysis also showed limited variations among the forages as well as across the lengths of the rumen 20 incubation. Weighted Unifrac dissimilarity matrix, which considered taxa relative abundances and their phylogenetic 21 distances, explained 21% and 26% of the variations among the forages and across the lengths of the incubation, 22 23 respectively. PERMANOVA revealed that at least some forages (e.g. CR and SC) had different microbial communities 24 (Figure 2a, p < 0.001) while testing for homogeneity of group dispersions (sample distance from group centroid) 25 26 identified no significant difference between group dispersions (Figure 2b, PERMDISP p > 0.1). Nevertheless, 27 differences in microbial communities across the incubation lengths (Figure 2c, PERMANOVA p < 0.001) appeared 28 29 to be mainly affected by within-group dispersions, either by the forage types or inter-animal variations, particularly 30 during the first 24 h (Figure 2d, PERMDISP p < 0.05). At this time, CR-attached microbiota was well-separated from 31 32 those attached to other forages. The entire microbial community structure showed a clear shift among the forages 33 during the incubation because differences in the microbiota were apparent at 24 h but disappeared in later sampling 34 intervals (Figure 2c and d). This finding suggested the existence of a strong preference of rumen microbiota for 35 36 attachment to the forages of different digestibility while the rate and extent of such preference were quickly 37 compromised after the initial hours of rumen digestion (24 h). 38 39 40 Forage-attached microbial community 41 42 A total of 18 bacterial phyla and one archaeal phylum were identified from the forage-attached microbial communities 43 44 colonized in the Taleshi cattle rumen. The communities were dominated by phyla Firmicutes (45%) and Bacteroidetes 45 (41%) followed with Fibrobacteres (5%), Spirochaetes (3%) and Proteobacteria (2%). Variations in the abundances 46 of major bacterial phyla attached to the forages have been depicted in Figure 3a. The ANCOM analysis followed by 47 48 Duncan’s post hoc test revealed differential abundances of Bacteroidetes, Fibrobacteres, Lentisphaerae and 49 Spirochaetes among the forages (Figure 3b, ANCOM p < 0.05). Bacteroidetes were significantly overrepresented in 50 51 AP and CR compared with DP and SC (DMRT Bonferroni p < 0.001). Interestingly, fiber-utilizing bacteria belonging 52 to Fibrobacteres were observed in more than 9% of the reads of CR and SC (two forages with the highest initial NDF 53 54 contents) but in less than 2% of the reads of AP and KS (p < 0.05) and also in DP and RS (Figure 3a). CR contained 55 56 57 58 59 9 60 ScholarOne Support 1-434/964-4100 iewv r R e Pe e Fo r Page 13 of 50 FEMS Microbiology Ecology 1 2 3 more Lentisphaerae compared with other forages (p < 0.001) while SC carried more Spirochaetes relative to AP (p < 4 5 0.02). 6 7 At the family level (Figure S5), the forage-attached microbes were affiliated to 76 families, of which nine showed 8 9 differential abundances among the forages, including Bacteroidaceae, Clostridiaceae, Fibrobacteraceae, 10 Victivallaceae, Christensenellaceae, Lachnospiraceae, Spirochaetaceae, Oxalobacteraceae and RFP12 (ANCOM p < 11 0.05). Interestingly, Victivallaceae were significantly overrepresented in CR compared with other forages (p < 0.05), 12 13 while Bacteroidaceae was significantly enriched in CR and RS compared with AP, DP, KS and SC. Oxalobacteraceae 14 was more abundant in KS and SC than in AP, CR, DP and RS (p < 0.05). Members of Fibrobacteraceae also more 15 16 frequently appeared in CR and SC compared with AP, KS and RS with the lowest initial NDF contents (p < 0.02), 17 while those of Clostridiaceae and Lachnospiraceae were underrepresented in CR compared with other forages (p < 18 19 0.05). 20 21 Particle-attached microbiota were affiliated to 119 genera (taxonomic level 6), of which 12 displayed differential 22 abundances among the forages (ANCOM p < 0.05, Figure 4) most of which were among high abundant members of 23 24 rumen community which are known to play key roles in plant lignocellulose degradation, including Ruminococcus, 25 Fibrobacter, Prevotella, Treponema, Lachnospira, Succinivibrio, Pseudobutyrivibrio, Butyrivibrio, Oxalobacter, 26 27 Clostridium, BF311 and Succiniclasticum. Ruminococcus and members of BF311 were more dominant in RS than in 28 AP, KS and SC (p < 0.001). Fibrobacter were more highly represented in CR and SC. Species of Lachnospira were 29 30 present in 0.3% of the reads of AP and KS (p < 0.05) but only 0.11% of CR and 0.09% of SC. Succinivibrio were 31 overrepresented in AP and SC compared with CR (p < 0.05). Species of Prevotella were more abundant in CR (> 21% 32 33 of the reads) than in DP and SC (p < 0.008). Compared to other forages, CR carried less Butyrivibrio and 34 Pseudobutyrivibrio species (p < 0.007). 35 36 Changes in forage-attached microbes during the rumen incubation 37 38 39 In order to examine whether the community composition of forage-attached microbes changed during the rumen 40 incubation, the relative abundances of the microbes were tracked at 24 h intervals (Figure 5). The abundances of eight 41 42 out of the 119 bacterial genera showed significant differences among the incubation lengths (Bonferroni corrected p 43 < 0.05). Interestingly, the proportions of cellulolytic bacteria belonging to unclassified Ruminococcaceae linearly 44 45 increased with the incubation lengths in AP, KS and SC. The proportion of Fibrobacter sharply dropped after the first 46 24 h and reached to an average of 4% between 48 and 96 h of the incubation in CR. Members of Pseudobutyribibrio 47 linearly decreased in AP and KS while those of Butyrivibrio increased in CR with the incubation lengths. Members 48 49 of Clostridium increased in SC but those of Prevotella and unclassified Paraprevotellaceae decreased with the 50 incubation lengths, which were consistent with findings of Cheng et al. (2017). 51 52 53 Relationship between lignocellulose degradation and forage-attached microbial communities 54 55 56 57 58 59 10 60 ScholarOne Support 1-434/964-4100 w vie r R e e or Pe F FEMS Microbiology Ecology Page 14 of 50 1 2 3 The Pearson’s correlation between the initial physicochemical properties of the forages and the composition of the 4 5 forage-attached microbiota during initial hours (24 h) of the rumen incubation was performed to determine whether 6 rumen microbes preferred specific forages for attachment. Only correlations with p-values (Bonferroni-corrected) less 7 8 than 0.05 were considered to have significant biological terms. This analysis demonstrated that the prevalence of the 9 family Fibrobacteraceae (r = 0.77, p = 0.03) was positively correlated with NDF contents of the forages. At the genus 10 level, the abundance of Fibrobacter (r = 0.77, p = 0.05) was positively but an unclassified Erysipelotrichaceae genus 11 12 p-75-a5 (r = -0.79, p = 0.03) was negatively correlated with NDF contents in the forages. When hemicellulose contents 13 were considered, a negative correlation with members of the family Pirellulaceae (r = -0.79, p = 0.02) was detected. 14 15 16 We also correlated microbial profiles with physicochemical properties of the forages during the rumen incubation. 17 This analysis revealed that the prevalence of the families Fibrobacteraceae (r= 0.83, p = 0.03 in CR), 18 19 Anaeroplasmataceae (r = 0.82, p = 0.04 in CR), Prevotellaceae (r = 0.83, r = 0.04 in KS) and Paraprevotellaceae (r = 20 0.85, p = 0.02 in SC) were positively but Ruminococcaceae (r = < -0.85, p < 0.01 in CR and SC) and unclassified 21 Bacteroidales (r = -0.84, p = 0.05 in RS) were negatively corelated with DM contents of the forages. Particularly, 22 23 members of Ruminococcaceae were positively correlated with ADF, ADL and hemicellulose contents in CR, DP and 24 SC (r > 0.8 and p < 0.04), the forages with the highest initial NDF contents (Figure 1). The population of 25 26 Lachnospiraceae was also positively correlated (r = 0.83, and p = 0.04) with cellulose content in CR. 27 28 To ascertain whether there was any relationship between the rumen microbiota and lignocellulose degradation, an 29 30 additional Pearson’s correlation analysis was performed between DM loss and the relative abundance of forage- 31 attached microbial communities during the rumen incubation. Interestingly, DM degradation was positively correlated 32 33 with the prevalence of the families Fibrobacteraceae (r > 0.76, p < 0.001 in CR) and Spirochaetaceae (r = 0.91, p = 34 0.0004 in KS) but was negatively corelated with species belonging to unclassified Bacteroidales (r < -0.82, p < 0.05 35 in CR and DP), Ruminococcaceae (r = -0.83, p = 0.04 in CR), Mogibacteriaceae (r < -0.82, p < 0.05 in CR and RS) 36 37 and Erysiopelotrichaceae (r = -0.87, p = 0.01 in CR). The correlations of cellulose and hemicellulose degradations 38 with the abundances of forage-attached microbes also showed similar patterns. 39 40 41 Discussion 42 43 In this study, we investigated the relationship between biomass degradation of and microbial attachment to six 44 45 common lignocellulosic forages varying in their physicochemical properties, including percentages of NDF, ADF, 46 ADL and the contents of cellulose and hemicellulose. Forages containing the highest cellulose contents [(common 47 reed (CR) vs. camelthorn (AP)] were degraded to a limited extent during the initial hours of their rumen incubation. 48 49 The total DM degradation was mainly determined by NDF contents of the forages, because as rice straw (RS) with 50 the lowest initial NDF (77.4%) had the fastest (66%) while CR with the highest initial NDF (87.2%) had the lowest 51 52 (42%) DM loss over 96 h of the incubation. Variability in DM degradation of feeds reflected the differences in 53 lignocellulose composition of their cell walls (Bruno-Soares et al. 2000; Jančík et al. 2010). The rate and extent of 54 55 DM fermentation in the rumen determines the nutrition efficiency of feeds to ruminants (Jančík et al. 2010). 56 57 58 59 11 60 ScholarOne Support 1-434/964-4100 w ev ie er R or Pe F Page 15 of 50 FEMS Microbiology Ecology 1 2 3 The 16S rRNA gene sequencing data allowed taxonomic identification and quantification of the rumen microbiota 4 5 tightly attached to the forages. Rarefaction analysis based on the indices reflecting species richness and species relative 6 abundance, e.g. Shannon and Simpson (Kim et al. 2017), indicated that the majority of the diversity of rumen 7 8 microbiota attached to the forages was already sampled not sufficiently and evenly sampled. Alpha diversity analysis 9 also showed a limited biologically significant difference among the forages, which could likely be attributed to a high 10 microbial heterogeneity among animals included in the study. The changes in diversity measures were largely 11 12 restricted to the initial hours of rumen incubation when maximal DM degradation occurred. The differences in species 13 richness and evenness were linked to cellulose contents of the forages, where CR and RS with the highest cellulose 14 15 contents displayed limited species diversity. This result was in agreement with the data on rice straw and alfalfa hay 16 being fed to Holstein cows, in which more bacteria attached to alfalfa with lower NDF (Liu et al. 2016). These findings 17 18 suggested that only a limited fraction of rumen microbiota was capable of attachment to feeds with high lignocellulose 19 contents. 20 21 Analysis of community structure of forage-attached microbiota by Unifrac dissimilarity matrix revealed significant 22 23 differences among the forages and across the length of rumen incubation. Particularly, CR-attached microbiota was 24 well separated from those of other forages at the first 24 h of the incubation. In addition, microbiota attached to CR 25 26 and SC, the two forages with the highest initial NDF contents, were also distantly clustered while those of other 27 forages did not show any apparent separation, indicating the similarity of their communities. These variations could 28 29 likely be determined by the nature of the forages, e.g., their chemical compositions. Unifrac diversity measure also 30 showed that extension of the incubation time contributed to the increased similarity across the samples, as reflected 31 32 by the overlapped clustering in the PCoA plots for most samples collected at 72 and 96 h of the incubation. This could 33 be explained by the fact that initial DM degradation resulted in the reduction of digestible components but the 34 accumulation of indigestible residues which had quite similar properties across the forages, thus favoring the 35 36 attachment of structurally similar communities of rumen microbiota. Huws et al. (2013) also observed similar changes 37 in the microbial communities attached to perennial ryegrass following its rumen incubation. 38 39 40 Members of Firmicutes, Bacteroidetes, Fibrobacteres, Spirochaetes and Proteobacteria were dominant in all samples, 41 accounting for greater than 96% of the bacterial communities attached to the forages, consistent with the findings for 42 43 rice straw and alfalfa hay (Liu et al. 2016). The abundance of these bacterial phyla varied among the forages and 44 across the incubation lengths. Particularly, species of Bacteroidetes were more abundantly attached to the forages with 45 high ADF contents (AP and CR). Bacteroidetes were among highly abundant members of the rumen microbiota being 46 47 best recognized for their saccharolytic activities. The presence of a high number of pectinolytic and cellulolytic 48 enzymes in their genomes clustered with other lignocellulose degrading enzymes into polysaccharide utilization loci 49 50 (PUL) suggested that they are also actively involved in pectin, hemicellulose and cellulose degradation (Gharechahi 51 and Salekdeh 2018; Lapebie et al. 2019; Naas et al. 2014). Within this phylum, members of the families 52 53 Bacteroidaceae, Prevotellaceae and Paraprevotellaceae displayed significant differences among the forages. 54 Particularly, Bacteroidaceae were significantly overrepresented in the forages with the lowest initial ADL contents 55 56 (CR and RS). At the genus-level, this differential abundance was only affiliated to the BF311, an uncultured and 57 58 59 12 60 ScholarOne Support 1-434/964-4100 w ev ie r R Pe e Fo r FEMS Microbiology Ecology Page 16 of 50 1 2 3 unknown rumen bacterium. Prevotellaceae were particularly positively correlated with DM degradation, suggesting 4 5 their active role in lignocellulose degradation in the rumen. Prevotella spp. are abundant members of the rumen 6 microbiome that have a high genetic diversity and thus the ability to thrive on a wide range of substrates, including 7 8 cellulose, hemicellulose, pectin, proteins and peptides (Dodd et al. 2010; Golder et al. 2014). They are known for their 9 xylanolytic properties in the rumen and thus play an important role in fiber degradation (Dodd et al. 2010). 10 11 Fibrobacteres were predominantly represented in microbiota attached to the forages with the highest NDF contents 12 13 [e.g. CR and Salicornia (SC)]. They are known to produce a battery of cellulolytic enzymes capable of degrading 14 cellulose as a sole carbon source (Flint et al. 2008; Suen et al. 2011). The association of forage NDF content with the 15 16 prevalence of Fibrobacter is also recently observed in the cow rumen microbiota (Liu et al. 2016). In contrast to 17 Fibrobacteres, the genera Clostridium, Shuttleworthia and Ruminococcus tended to attach the forages with limited 18 19 NDF and cellulose contents. This is consistent with previous findings on the preference of Ruminococcus species for 20 attachment to high quality sugar-rich hays (Klevenhusen et al. 2017). Ruminococcaceae showed a strong negative 21 correlation with total DM contents in the forages as well. Shinkai et al. (2010) also reported that members of 22 23 Fibrobacteres, including F. succinogenes, abundantly attached to less digestible fibers while those of 24 Ruminococcaceae, specifically R. flavefaciens, preferred easily digestible fibers and thus were infrequently detected 25 26 in the stem parts of hays. The rate and extent of fiber degradation by F. succinogenes are also greater than R. albus 27 and R. flavefaciens (Kobayashi et al. 2008). Early microscopy analysis indicated that R. albus is less commonly 28 29 attached to plant cell walls while F. succinogenes forms extensive microcapsula enveloping the cell walls (Chesson 30 et al. 1986). The community of unclassified Ruminococcaceae tended to linearly expand in AP, KS and SC following 31 32 their rumen incubation. This finding suggested that the degradation of surface accessible fibers turned forage residues 33 into favored substrates for the attachment by species of Ruminococcaceae. A declined abundance of species of 34 Ruminococcus has been reported in the cow rumen when the starch content of diet was increased (Zened et al. 2013). 35 36 Ruminococcus spp. have been equipped with specialized mechanisms for fiber adhesion and degradation, in which 37 multiple carbohydrate degrading enzymes assemble into multienzyme cellulosome complexes capable of attachment 38 39 to and degradation of polysaccharides in various plant cell walls (Bayer et al. 2004; Doi and Kosugi 2004). 40 41 Recently, Huws et al. (2016) reported that microbial colonization of perennial ryegrass occurred at two successive 42 43 stages: The primary (the first 1-2 h after its rumen entry) and the secondary phases (4-8 h). It was the secondary phase, 44 during which the maximal DM degradation was achieved and the proportion of Succinivibrio spp. decreased while 45 those of Pseudobutyrivibrio, Roseburia and Ruminococcus spp. increased. Piao et al. (2014) also observed that the 46 47 populations of Pseudobutyrivibrio and Ruminococcus spp. increased during their secondary colonization to 48 switchgrass in the Friesian cow rumen. The microbial communities attached to feeds during the primary phase were 49 50 believed to utilize soluble and easily accessible nutrients while those colonized feeds during the secondary phase were 51 considered to be the true lignocellulose degraders (Huws et al. 2016; Liu et al. 2016). In this study, an increased 52 53 prevalence of Butyrivibrio was noted in CR. Butyrivibrio spp. are known for their proteolytic, hemicellulolytic and 54 biohydrogenating activities (Krause et al. 2003). Liu et al. (2016) reported a strong positive correlation between the 55 56 57 58 59 13 60 ScholarOne Support 1-434/964-4100 iew Re v ee r r P Fo Page 17 of 50 FEMS Microbiology Ecology 1 2 3 abundance of Butyrivibrio spp. and the crude protein contents in the rice straw and alfalfa hay, suggesting their 4 5 preference for the attachment to proteinaceous components of the feeds. 6 7 In summary, our results demonstrated that physicochemical compositions of the forages, more specifically their 8 9 cellulose contents, were among the major factors influencing microbial attachments and thus lignocellulose 10 degradation in the cattle rumen. No taxonomic lineage was found to be specific of a particular forage, suggesting that 11 most rumen microbes had an inherent tendency for attachment to lignocellulosic substrates. However, differential 12 13 attachments to the forages by rumen microbiota suggested that the physicochemical properties were the key factors 14 influencing the rate of microbial attachment. Our results also revealed that members of the rumen microbiota competed 15 16 for their attachments to the forages with different lignocellulose compositions mostly during the initial hours of the 17 rumen incubation. However, after the degradation of easily digestible components in the forages, relatively uniform 18 19 microbial communities developed on their surface. 20 21 It should be noted that the bacterial community composition detected in each forage at each sampling interval may 22 not necessarily reflect the actual abundance of microbes due to an inherent bias in sample collection, processing and 23 24 storage as well as DNA extraction and PCR amplification, which may cause over- or under-representation of some 25 taxa. For example, a recent comparative analysis of various DNA extraction methods used in rumen microbiome 26 27 analyses argued that DNA extraction method can have an adverse effect on the community composition of rumen 28 samples and may associate with an increased or decreased abundance of some specific taxa (Henderson et al. 2013). 29 30 Particularly, the inclusion of a physical lysis step using bead-beating method increased the efficacy of DNA extraction 31 from Gram-positive bacteria (Knudsen et al. 2016). The extent to which our DNA extraction method, in which the 32 33 physical lysis step was not included, affected community composition of the forage-attached microbiota was unknown. 34 In addition, the high number of PCR cycles may increase the number of PCR artifacts and bias the microbial 35 compositions. 36 37 38 Acknowledgements 39 40 This research was funded by the Agricultural Biotechnology Research Institute of Iran (ABRII), the international 41 42 cooperation and exchange program of the National Natural Science Foundation of China (No. 31461143020), and the 43 Chinese government contribution to CAAS-ILRI Joint Laboratory on Livestock and Forage Genetic Resources in 44 45 Beijing. The paper contributes to the CGIAR Research Program on Livestock. 46 47 Figure legends 48 49 Figure 1. Differences among the six forages in dry matter (DM) and percentages of neutral detergent fiber (NDF), 50 51 acid detergent fiber (ADF), acid detergent lignin (ADL), cellulose and hemicellulose before (0 h) and after 24, 48, 72, 52 and 96 h of the rumen incubation. The solid circle and error bar in each boxplot show mean and standard deviation, 53 54 respectively. Statistically significant differences were determined using one-way ANOVA. Means were compared 55 56 57 58 59 14 60 ScholarOne Support 1-434/964-4100 vie w er Re Pe Fo r FEMS Microbiology Ecology Page 18 of 50 1 2 3 using Duncan’s multiple range test and are labelled with different letters at each sampling interval for p < 0.05. AP, 4 5 camelthorn; CR, common reed; DP, date palm; KS, Kochia; RS, rice straw; and SC, Salicornia. 6 7 Figure 2. Beta diversity analysis of rumen microbial communities attached to the six forages during the rumen 8 9 incubation. PCoA plots show the distribution of samples based on weighted Unifrac distance matrix, in which samples 10 are grouped according to the forages (a) and the sampling intervals (b). Significant differences were tested using 11 PERMANOVA with a p-value cutoff 0.01. The percentage of variation explained by each principle coordinate is 12 13 indicated next to the corresponding axis. The homogeneity of dispersions was also tested for this diversity measure to 14 examine variance differences between the forages (c) and the sampling intervals (d). Significant differences were 15 16 determined using the betadisper function of R package vegan v2.5-5 at 999 permutations. P-values < 0.05 were 17 considered statistically significant. For each treatment, data in triplicate representing three separate bags, were 18 19 included. AP, camelthorn; CR, common reed; DP, date palm; KS, Kochia; RS, rice straw; and SC, Salicornia. 20 21 Figure 3. (a) The proportions of reads (mean ± SD) assigned to the major bacterial phyla (n = 7) detected in the forage- 22 attached microbial communities. The relative abundances were calculated based on the proportion of reads assigned 23 24 to each phylum in the rarefied OTU table at an even sequencing depth of 4660 reads. (b) Box plots show relative 25 abundance of phyla differentially attached to the forages. Statistically significant differences were calculated based on 26 27 the analysis of composition of microbes (ANCOM) using p < 0.05 following Duncan’s multiple range test for 28 comparison of the means. Center lines indicate the median value, boxes the interquartile range and red squares the 29 30 mean. Means with different letters differ at Bonferroni corrected p < 0.05. AP, camelthorn; CR, common reed; DP, 31 date palm; KS, Kochia; RS, rice straw; and SC, Salicornia. 32 33 34 Figure 4. The log of relative abundances of genera differentially attached to the six forages following the rumen 35 incubation. Differential abundances were statistically tested using ANCOM with a p-value cutoff < 0.05. Means were 36 compared using Duncan’s multiple range test only accepting Bonferroni corrected p-values < 0.05. Boxplots labeled 37 38 with different letters show statistically significant differences. Center line represents median value. AP, camelthorn; 39 CR, common reed; DP, date palm; KS, Kochia; RS, rice straw; and SC, Salicornia. 40 41 42 Figure 5. The average relative abundances of taxa (genus level) differentially represented in microbial communities 43 attached to the six forages during the rumen incubation (up to 96 h with 24 h intervals). Statistically significant 44 45 differences were calculated using ANCOM and means were compared using Duncan’s multiple range test. Error bars 46 represent standard deviations and their lengths are adjusted at 95% confidence interval. Means with different letters 47 are statistically significant at Bonferroni corrected p < 0.05. No taxa were found to be differentially represented in DP. 48 49 AP, camelthorn; CR, common reed; KS, Kochia; RS, rice straw; and SC, Salicornia. “Un” refers to unclassified. 50 51 Supplementary materials 52 53 Figure S1. 54 Changes in DM contents and percentages of NDF, ADF, ADL, cellulose and hemicellulose in six forages during the 55 56 rumen incubation. The solid circle in each boxplot shows mean and error bars represent standard deviation. 57 58 59 15 60 ScholarOne Support 1-434/964-4100 w ev ie r R Pe e Fo r Page 19 of 50 FEMS Microbiology Ecology 1 2 3 Statistically significant differences were determined using one-way ANOVA. Means were compared using Duncan’s 4 5 multiple range test and labelled with different letters for p < 0.05. 6 Figure S2. 7 8 Alpha diversity analysis through rarefaction plots. (a) Rarefaction curves showing the increase in number of observed 9 OTUs; (b) Species richness (number of observed OTUs + number of unobserved OTUs, Chao1); (c) Shannon; and (d) 10 Simpson diversity indices on Y-axis as a function of the number of reads sampled on X-axis. 11 12 Figure S3. 13 Alpha diversity indices of rumen microbiota attached to the six forages with different lignocellulosic compositions at 14 15 24, 48, 72 and 96 hours of the rumen incubation. Alpha diversity indices were measured based on OTUs present at an 16 even sequencing depth of 4,660 reads (corresponding to the sequencing depth of the sample with the lowest number 17 18 of reads) in all samples. Statistically significant differences were determined using one-way ANOVA and means were 19 compared by Duncan’s multiple range test. * p value < 0.05, ** p value < 0.01 and *** p value < 0.001. 20 Figure S4. 21 22 Alpha diversity analysis of rumen microbial communities attached to the six forages during the rumen incubation. The 23 OTU table for each forage was sampled at an even sequencing depth (n = 4660) and diversity measures were calculated 24 25 using plot_richness function of R package phyloseq and visualized using geom_boxplot function of R package 26 ggplot2. Statistically significant differences were calculated using one-way ANOVA at p-value cutoff of 0.05. 27 28 Differences between means were determined using Duncan’s multiple range test in PAST v3.26. * p value < 0.05, ** 29 p value < 0.01 and *** p value < 0.001. 30 31 Figure S5. 32 The proportions of reads affiliated to OTUs at the taxonomic level of family in the six forages during the rumen 33 incubation. Data for each sampling interval is the mean of three biological replicates corresponding to three bags 34 35 retrieved one each from the three cannulated bulls. For each forage, the OTU table is rarefied to the sequencing depth 36 of the sample with the lowest number of reads. 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 16 60 ScholarOne Support 1-434/964-4100 w ev ie er R e or P F FEMS Microbiology Ecology Page 20 of 50 1 2 3 References 4 5 6 Bayer EA, Belaich JP, Shoham Y et al. The cellulosomes: multienzyme machines for degradation of plant cell wall 7 polysaccharides. Annu Rev Microbiol 2004;58: 521-54. 8 Bergman EN. Energy contributions of volatile fatty acids from the gastrointestinal tract in various species. Physiol 9 Rev 1990;70: 567-90.Bruno-Soares AM, Abreu JMF, Guedes CVM et al. Chemical composition, DM and NDF degradation kinetics in 10 rumen of seven legume straws. Animal Feed Science and Technology 2000;83: 75-80. 11 Caporaso JG, Bittinger K, Bushman FD et al. PyNAST: a flexible tool for aligning sequences to a template alignment. 12 Bioinformatics 2010a;26: 266-7. 13 Caporaso JG, Kuczynski J, Stombaugh J et al. QIIME allows analysis of high-throughput community sequencing data. 14 Nat Methods 2010b;7: 335-6. 15 Cheng Y, Wang Y, Li Y et al. Progressive Colonization of Bacteria and Degradation of Rice Straw in the Rumen by 16 Illumina Sequencing. Front Microbiol 2017;8: 2165. 17 Chesson A, Stewart CS, Dalgarno K et al. Degradation of isolated grass mesophyll, epidermis and fibre cell walls in 18 the rumen and by cellulolytic rumen bacteria in axenic culture. Journal of Applied Bacteriology 1986;60: 19 327-36. 20 De Mulder T, Goossens K, Peiren N et al. Exploring the methanogen and bacterial communities of rumen 21 environments: solid adherent, fluid and epimural. FEMS Microbiol Ecol 2017;93. 22 DeSantis TZ, Hugenholtz P, Larsen N et al. Greengenes, a chimera-checked 16S rRNA gene database and workbench 23 compatible with ARB. Appl Environ Microbiol 2006;72: 5069-72. 24 Dhariwal A, Chong J, Habib S et al. MicrobiomeAnalyst: a web-based tool for comprehensive statistical, visual and 25 meta-analysis of microbiome data. Nucleic Acids Res 2017;45: W180-W88. 26 Dodd D, Moon YH, Swaminathan K et al. Transcriptomic analyses of xylan degradation by Prevotella bryantii and 27 insights into energy acquisition by xylanolytic bacteroidetes. J Biol Chem 2010;285: 30261-73. 28 Doi RH, Kosugi A. Cellulosomes: plant-cell-wall-degrading enzyme complexes. Nat Rev Microbiol 2004;2: 541-51. 29 Edgar RC. Search and clustering orders of magnitude faster than BLAST. Bioinformatics 2010;26: 2460-1. 30 Edwards JE, Huws SA, Kim EJ et al. Characterization of the dynamics of initial bacterial colonization of nonconserved 31 forage in the bovine rumen. FEMS Microbiol Ecol 2007;62: 323-35. 32 Flint HJ, Bayer EA, Rincon MT et al. Polysaccharide utilization by gut bacteria: potential for new insights from genomic analysis. Nat Rev Microbiol 2008;6: 121-31. 33 Gharechahi J, Kharazian ZA, Sarikhan S et al. The dynamics of the bacterial communities developed in maize silage. 34 Microb Biotechnol 2017;10: 1663-76. 35 Gharechahi J, Salekdeh GH. A metagenomic analysis of the camel rumen's microbiome identifies the major microbes 36 responsible for lignocellulose degradation and fermentation. Biotechnol Biofuels 2018;11: 216. 37 Gharechahi J, Zahiri HS, Noghabi KA et al. In-depth diversity analysis of the bacterial community resident in the 38 camel rumen. Syst Appl Microbiol 2015;38: 67-76. 39 Godoy-Vitorino F, Goldfarb KC, Karaoz U et al. Comparative analyses of foregut and hindgut bacterial communities 40 in hoatzins and cows. ISME J 2012;6: 531-41. 41 Goering HK. Forage fiber analyses (apparatus, reagents, procedures, and some applications). In: Van Soest PJ (ed.). 42 Washington, D.C. :: Agricultural Research Service, U.S. Dept. of Agriculture, 1970. 43 Golder HM, Denman SE, McSweeney C et al. Ruminal bacterial community shifts in grain-, sugar-, and histidine- 44 challenged dairy heifers. J Dairy Sci 2014;97: 5131-50. 45 Guarner F, Malagelada JR. Gut flora in health and disease. Lancet 2003;361: 512-9. 46 Hammer Ø, Harper DA, Ryan PD. PAST: paleontological statistics software package for education and data analysis. 47 2001. 48 Henderson G, Cox F, Kittelmann S et al. Effect of DNA extraction methods and sampling techniques on the apparent 49 structure of cow and sheep rumen microbial communities. PLoS One 2013;8: e74787. 50 Hungate RE. The rumen and its microbes: Elsevier, 2013. 51 Huws SA, Edwards JE, Creevey CJ et al. Temporal dynamics of the metabolically active rumen bacteria colonizing 52 fresh perennial ryegrass. FEMS Microbiol Ecol 2016;92. 53 Huws SA, Mayorga OL, Theodorou MK et al. Successional colonization of perennial ryegrass by rumen bacteria. Lett 54 Appl Microbiol 2013;56: 186-96. 55 Jami E, Israel A, Kotser A et al. Exploring the bovine rumen bacterial community from birth to adulthood. ISME J 2013;7: 1069-79. 56 57 58 59 17 60 ScholarOne Support 1-434/964-4100 iew Re v ee r or P F Page 21 of 50 FEMS Microbiology Ecology 1 2 3 Jami E, Mizrahi I. Composition and similarity of bovine rumen microbiota across individual animals. PLoS One 4 2012;7: e33306. 5 Jančík F, Koukolová V, Homolka P. Ruminal degradability of dry matter and neutral detergent fibre of grasses. Czech 6 J Anim Sci 2010;55: 359-71. 7 Jin W, Wang Y, Li Y et al. Temporal changes of the bacterial community colonizing wheat straw in the cow rumen. 8 Anaerobe 2018;50: 1-8. 9 Khafipour E, Li S, Plaizier JC et al. Rumen microbiome composition determined using two nutritional models of 10 subacute ruminal acidosis. Appl Environ Microbiol 2009;75: 7115-24. 11 Kim BR, Shin J, Guevarra R et al. Deciphering Diversity Indices for a Better Understanding of Microbial 12 Communities. J Microbiol Biotechnol 2017;27: 2089-93. 13 Klevenhusen F, Petri RM, Kleefisch MT et al. Changes in fibre-adherent and fluid-associated microbial communities 14 and fermentation profiles in the rumen of cattle fed diets differing in hay quality and concentrate amount. 15 FEMS Microbiol Ecol 2017;93. 16 Knudsen BE, Bergmark L, Munk P et al. Impact of Sample Type and DNA Isolation Procedure on Genomic Inference 17 of Microbiome Composition. mSystems 2016;1. 18 Kobayashi Y, Shinkai T, Koike S. Ecological and physiological characterization shows that Fibrobacter succinogenes 19 is important in rumen fiber digestion - review. Folia Microbiol (Praha) 2008;53: 195-200. 20 Kocherginskaya SA, Aminov RI, White BA. Analysis of the Rumen Bacterial Diversity under two Different Diet Conditions using Denaturing Gradient Gel Electrophoresis, Random Sequencing, and Statistical Ecology 21 Approaches. Anaerobe 2001;7: 119-34. 22 Kong Y, Teather R, Forster R. Composition, spatial distribution, and diversity of the bacterial communities in the 23 rumen of cows fed different forages. FEMS Microbiol Ecol 2010;74: 612-22. 24 Krause DO, Denman SE, Mackie RI et al. Opportunities to improve fiber degradation in the rumen: microbiology, 25 ecology, and genomics. FEMS Microbiol Rev 2003;27: 663-93. 26 Lapebie P, Lombard V, Drula E et al. Bacteroidetes use thousands of enzyme combinations to break down glycans. 27 Nat Commun 2019;10: 2043. 28 Liu J, Zhang M, Xue C et al. Characterization and comparison of the temporal dynamics of ruminal bacterial 29 microbiota colonizing rice straw and alfalfa hay within ruminants. J Dairy Sci 2016;99: 9668-81. 30 Mackie RI. Mutualistic fermentative digestion in the gastrointestinal tract: diversity and evolution. Integr Comp Biol 31 2002;42: 319-26. 32 Magoc T, Salzberg SL. FLASH: fast length adjustment of short reads to improve genome assemblies. Bioinformatics 33 2011;27: 2957-63. 34 Mandal S, Van Treuren W, White RA et al. Analysis of composition of microbiomes: a novel method for studying 35 microbial composition. Microb Ecol Health Dis 2015;26: 27663. 36 Martineau R, Proulx JG, Cortes C et al. Two-Stage Rumen Cannulation Technique in Dairy Cows. Vet Surg 2015;44: 37 551-6. 38 McAllister TA, Bae HD, Jones GA et al. Microbial attachment and feed digestion in the rumen. J Anim Sci 1994;72: 39 3004-18. 40 Naas AE, Mackenzie AK, Mravec J et al. Do rumen Bacteroidetes utilize an alternative mechanism for cellulose 41 degradation? mBio 2014;5: e01401-14. 42 Petri RM, Forster RJ, Yang W et al. Characterization of rumen bacterial diversity and fermentation parameters in 43 concentrate fed cattle with and without forage. J Appl Microbiol 2012;112: 1152-62.Petri RM, Pourazad P, Khiaosa-Ard R et al. Temporal dynamics of in-situ fiber-adherent bacterial community under 44 ruminal acidotic conditions determined by 16S rRNA gene profiling. PLoS One 2017;12: e0182271. 45 Piao H, Lachman M, Malfatti S et al. Temporal dynamics of fibrolytic and methanogenic rumen microorganisms 46 during in situ incubation of switchgrass determined by 16S rRNA gene profiling. Front Microbiol 2014;5: 47 307. 48 Rognes T, Flouri T, Nichols B et al. VSEARCH: a versatile open source tool for metagenomics. PeerJ 2016;4: e2584. 49 Sadet S, Martin C, Meunier B et al. PCR-DGGE analysis reveals a distinct diversity in the bacterial population 50 attached to the rumen epithelium. Animal 2007;1: 939-44. 51 Shinkai T, Ueki T, Kobayashi Y. Detection and identification of rumen bacteria constituting a fibrolytic consortium 52 dominated by Fibrobacter succinogenes. Anim Sci J 2010;81: 72-9. 53 Suen G, Weimer PJ, Stevenson DM et al. The complete genome sequence of Fibrobacter succinogenes S85 reveals a 54 cellulolytic and metabolic specialist. PLoS One 2011;6: e18814. 55 Van Soest PJ, Robertson JB, Lewis BA. Methods for dietary fiber, neutral detergent fiber, and nonstarch 56 polysaccharides in relation to animal nutrition. J Dairy Sci 1991;74: 3583-97. 57 58 59 18 60 ScholarOne Support 1-434/964-4100 w vie er Re e or P F FEMS Microbiology Ecology Page 22 of 50 1 2 3 Wang Q, Garrity GM, Tiedje JM et al. Naive Bayesian classifier for rapid assignment of rRNA sequences into the 4 new bacterial taxonomy. Appl Environ Microbiol 2007;73: 5261-7. 5 Zened A, Combes S, Cauquil L et al. Microbial ecology of the rumen evaluated by 454 GS FLX pyrosequencing is 6 affected by starch and oil supplementation of diets. FEMS Microbiol Ecol 2013;83: 504-14. 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 19 60 ScholarOne Support 1-434/964-4100 vie w er Re Pe Fo r AP DP RS Feed CR KS SC SamplingTime: 0 SamplingTime: 24 SamplingTime: 48 5 Page 23 of 50 FEMS Microbiology 4Ecology 3 AP DP RS 1 2 Feed CR KS SC AP DP RS 2 Feed CR KS SC 1 SamplingTime: 0 SamplingTime: 24 SamplingTime: 48 3 SamplingTime: 0 SamplingTime: 24 SamplingTime: 48 100 SamplingTime: 72 SamplingTime: 96 AP CR DP KS RS SCp = 1.5e-08 p = 1.5e-07 p = 1.6e-09 4 5 p = 0.2 p < 0.001 5 p < 0.001 p < 0.001p = 0.9 5 904 4 6 ab b b a b ab a3 b b 3 7 c 80 bc c b d A cP DP RS Feed c AP DP RS CR KS SC 8 2 Feed 2 dCR KS SC 70 SamplingTime: 0 SamplingTime: 24 SamplingTime: 48 9 1 SamplingTime: 0 SamplingTime: 24 SamplingTime: 48 1 SamplingTime: 72 SamplingTime: 96 AP CR DP KS RS SC 10 100 SampplingTime: 72 Samplin AP CR DP KS RS SC AP CR DP KS RS SC AP CR DP KS RS SC < 0.001 P < 0.g0T0i1me: 96 100 p < 0.001 p < 0.001 5 p = 1.5e-08 p = 1.5e-07 p = 1.6e-0 p9 < 0.001 70 p = 4.6e-05 a p = 1.4e-06 Feed 11 a a a ab ab90 a b 9600 c bc a a 12 4 b c bc c c bc c b c 13 d3 50 AP DP RS80 e d d Feed ce AP DP RS 80 d CR KS SC 14 Feed2 CR KS SC 40 SamplingTime: 0 SamplingTime: 24 SamplingTime: 48 15 70 SamplingTime: 0 SamplingTime: 24 SamplingTime: 48 740 SamplingTime: 72 SamplingTime: 96 AP CR DP KS RS SC 16 1 aSapm two mismatches from the primer sequences. The multiplexed sequences were 28 searched against the latest Ribosomal Database Project (RDP release 11.5, containing 3,356,809 16S rRNA sequences) 29 30 to identify and discard chimeric sequences using the VSEARCH v2.8.3 operated under default setting (Rognes et al. 31 2016). Non-chimeric sequences were then used to pick operational taxonomic units (OTUs) using the 32 33 pick_de_novo_otus.py script in the QIIME pipeline as described in details in our previous paper (Gharechahi et al. 34 2017). OTUs were defined at 97% identity using the Uclust (Edgar 2010). The most abundant sequence in each OTU 35 36 cluster was selected as being a representative; and these sequences were then aligned against the Greengenes core set 37 (gg_13_8; (DeSantis et al. 2006)) using the PyNAST aligner with a minimum sequence identity of 75% (Caporaso et 38 al. 2010a). Taxonomies were assigned to each OTU using the Ribosomal Database Project naïve Bayesian classifier 39 40 (Wang et al. 2007) by applying a minimum confidence value of 0.8. The OTU table was filtered for low abundant 41 OTUs using the filter_otus_from_otu_table.py script with --min_count_fraction option set to 0.00001 (discarding 42 43 OTUs represented by < 0.001% of the sequences) and then rarefied to the sequencing depth at 4660 reads 44 corresponding to the number of reads in the sample with the smallest set of sequences. Rarefaction plots and alpha 45 46 diversity indices, including Shannon, Simpson, Good’s_coverage and Chao1, were calculated using the 47 core_diversity_analyses.py script in the QIIME pipeline. Beta diversity indices, including weighted and unweighted 48 49 Unifrac phylogenetic distance matrices, were constructed with the rarefied OTU table as input and visualized through 50 the principal coordinate analysis (PCoA) plots in the MicrobiomeAnalyst web server (Dhariwal et al. 2017). 51 52 Statistical analysis 53 54 55 56 57 58 59 6 60 ScholarOne Support 1-434/964-4100 w vie r R e e or Pe F FEMS Microbiology Ecology Page 38 of 50 1 2 3 Statistically significant differences in physicochemical data, including DM, NDF, ADF, ADL, cellulose and 4 5 hemicellulose contents, were analyzed by one-way ANOVA using the general linear model (GLM) procedure in the 6 SAS software v9.3 (SAS Institute Inc., Cary, NC, USA). Permutational multivariate analysis of variance 7 8 (PERMANOVA) was performed using the adonis function of R-package vegan v2.5-5 to test for significant 9 differences between community compositions of forage-attached microbial communities. In addition, permutation 10 multivariate analysis of group dispersions (PERMDISP) based on the betadisper function of the vegan was used to 11 12 test for the homogeneity of dispersions (variances). Differences in taxa abundances among forages and sampling 13 intervals were estimated using analysis of composition of microbes (ANCOM) based on relative abundances of OTUs 14 15 summarized at various taxonomic levels (Mandal et al. 2015). Means were compared by Duncan's Multiple Range 16 test (DMRT) in PAST v3.26 given Bonferroni p-value cutoff < 0.05 (Hammer et al. 2001). Error correction was done 17 18 based on the number of groupwise comparisons performed at each taxonomic level. The Pearson’s correlation analysis 19 was performed using the corr.test function of the psych package v1.8.12 and p-values were corrected using Bonferroni 20 method based on the total number of correlations calculated for each variable separately. For all tests, p-values less 21 22 than 0.05 were considered statistically significant. 23 24 Results 25 26 27 Physicochemical properties of the rumen-incubated forages 28 29 The six forages were analyzed for the contents of NDF, ADF, ADL, cellulose and hemicellulose before their rumen 30 31 incubation (Figure 1). They showed different NDF contents (p < 0.05), being the highest in common reed (CR) but 32 the lowest in both camelthorn (AP) and rice straw (RS). Most of them also contained different ADF contents (p < 33 34 0.05), being the highest in AP but the lowest in both RS and salicornia (SC). The contents of ADL also differed 35 significantly between the forages (p < 0.05), with AP being the highest at 2 times that in date palm (DP), three times 36 those in SC, CR and kochia (KS; the latter two with similar contents at p > 0.05) and 6 times the lowest value in RS 37 38 (24.86% vs 4.06%). CR carried the highest cellulose followed by DP, RS, KS and SC while AP contained the lowest 39 cellulose (p < 0.05). SC possessed the highest amount of hemicellulose followed by CR, KS, RS, DP and AP. Overall, 40 41 all the six forages had different physicochemical properties in terms of their relative contents of hemicellulose, 42 cellulose and lignin. 43 44 45 Lignocellulosic biomass degradation following the rumen incubation 46 47 The six forages were monitored for their changes in the contents of DM, NDF, ADF, ADL, cellulose and hemicellulose 48 during the rumen incubation (Figure 1 and Figure S1). DM degradation was the fastest in AP but the slowest in CR 49 50 (42% vs 34%, p < 0.05) during the first 24 h of the incubation. DM degradation was completed in AP at 24 h but 51 continued in CR and KS until 48 h, and in DP, RS and SC up to 96 h of the incubation, with RS having the highest 52 53 degraded DM (66%) followed by SC (56%) and DP (53%). AP, DP and KS demonstrated similar trends of 54 significantly increased NDF, ADF, ADL and cellulose contents (p < 0.05), mirrored in their patterns of DM 55 56 degradations, mostly within the first 24 h, but their hemicelluloses decreased continuously (p < 0.05 in most cases) 57 58 59 7 60 ScholarOne Support 1-434/964-4100 w ev ie r R Pe e r Fo Page 39 of 50 FEMS Microbiology Ecology 1 2 3 following the incubation. However, the patterns of these five fiber-related parameters were significantly segregated 4 5 among CR, RS and SC during the incubation. Although ADF and ADL were significantly increased but celluloses 6 were remarkably declined in CR and SC (p < 0.05), their NDF and hemicelluloses showed contrasting patterns, being 7 8 significantly reduced in CR while steadily accumulated in SC (p < 0.05) throughout the incubation. ADF, cellulose 9 and hemicellulose in RS maintained stable levels while its NDF and ADL were slightly increased (p < 0.05) along the 10 incubation. It was apparent that the initial differences in fiber-related physicochemical properties significantly affected 11 12 the rumen digestion of the six forages. 13 14 16S rRNA gene sequencing 15 16 17 The paired-end sequencing of PCR amplicons from the V3-V4 region of 16S rRNA gene resulted in 5,945,300 pairs 18 of raw sequences (averaged at 82,573 sequences per sample) with an average length of 300 bp. Sequences were joined 19 20 into 4,421,604 full-length amplicons at an average of 61,411  38,275 per sample. The uneven sequencing depths 21 across samples may be due to the true differences in microbial abundance of samples or the technical variations 22 23 introduced during library preparation and sequencing. Although the numbers of sequences varied greatly among the 24 forages and across the lengths of the rumen incubation, there was a general pattern of a steady increase from the lowest 25 at 24 h (averaged at 44,100 per sample) to the highest at 72 h (averaged at 95,823 per sample). 26 27 28 Sequencing data analysis 29 30 Before processing the amplicon sequences for OTU picking, they were subjected to a single round of quality filtering, 31 32 resulting in 3,602,510 high quality sequences, of which 994,147 (27%) were identified to be chimeric and thus 33 discarded from further analyses. The qualified sequences (2,493,285) were clustered at 97% similarity level into 34 35 110,804 OTUs, of which 106,982 (96.4%, representing 202,403 sequences) were labeled as low abundant features 36 (e.g. those representing reads with frequencies less than 0.001% of the total sequences) and therefore filtered out from 37 the OTU table. Finally, 3,822 clean OTUs representing 2,290,882 sequences were subjected to further downstream 38 39 analyses. To assess whether our sequencing effort provided sufficient sequencing depths to describe the diversity of 40 forage-attached microbiota, rarefaction curves describing the numbers of observed OTUs, species richness (Chao1), 41 42 Shannon and Simpson diversity at various sequencing depths were generated for all samples (Figure S2a-d). 43 Rarefaction analysis based on observed species and species richness revealed incomplete sampling of microbiota and 44 45 thus indicated that highly diverse microbial communities attached to the forages (Figure S2a and b). However, 46 Shannon (Figure S2c) and Simpson indices (Figure S2d) reached plateaus, indicating that the majority of the diversity 47 48 was explored. 49 50 Microbial diversity analysis 51 52 Differences in alpha diversity indices of forage-attached microbiota were mostly limited to the first 24 h of the rumen 53 54 incubation, during which the maximal differences in microbial attachment occurred (Figure S3). At this time, CR and 55 RS (two forages with the highest initial cellulose contents) had the lowest while AP and KS (two forages with the 56 57 58 59 8 60 ScholarOne Support 1-434/964-4100 vie w er Re r P e Fo FEMS Microbiology Ecology Page 40 of 50 1 2 3 lowest initial cellulose contents) had the highest average observed species and species richness. CR and RS also 4 5 showed the lowest average Shannon index, indicating their limited microbial diversity (p < 0.05). Diversity indices in 6 almost all forages were not affected by the incubation length (Figure S4) except that CR had very low alpha diversity 7 8 measures at 24 h of its incubation (p < 0.05). All samples showed a high Good’s coverage (> 0.91, data not shown) at 9 all sampling intervals, indicating that our sequencing effort figured out > 90% of the microbial diversity attached to 10 the forages. However, the uneven sequencing depths did not allow us to fully explore the diversity of forage-attached 11 12 microbiota. Considering potential random errors due to limited numbers of experimental animals and replicates in this 13 study, such differences in alpha diversity indices among the forages were not robust enough to be interpreted with any 14 15 biological significance. 16 17 Beta diversity analysis also showed limited variations among the forages as well as across the lengths of the rumen 18 19 incubation. Weighted Unifrac dissimilarity matrix, which considered taxa relative abundances and their phylogenetic 20 distances, explained 21% and 26% of the variations among the forages and across the lengths of the incubation, 21 respectively. PERMANOVA revealed that at least some forages (e.g. CR and SC) had different microbial communities 22 23 (Figure 2a, p < 0.001) while testing for homogeneity of group dispersions (sample distance from group centroid) 24 identified no significant difference between group dispersions (Figure 2b, PERMDISP p > 0.1). Nevertheless, 25 26 differences in microbial communities across the incubation lengths (Figure 2c, PERMANOVA p < 0.001) appeared 27 to be mainly affected by within-group dispersions, either by the forage types or inter-animal variations, particularly 28 29 during the first 24 h (Figure 2d, PERMDISP p < 0.05). At this time, CR-attached microbiota was well-separated from 30 those attached to other forages. The entire microbial community structure showed a clear shift among the forages 31 32 during the incubation because differences in the microbiota were apparent at 24 h but disappeared in later sampling 33 intervals (Figure 2c and d). This finding suggested the existence of a strong preference of rumen microbiota for 34 attachment to the forages of different digestibility while the rate and extent of such preference were quickly 35 36 compromised after the initial hours of rumen digestion (24 h). 37 38 Forage-attached microbial community 39 40 41 A total of 18 bacterial phyla and one archaeal phylum were identified from the forage-attached microbial communities 42 colonized in the Taleshi cattle rumen. The communities were dominated by phyla Firmicutes (45%) and Bacteroidetes 43 44 (41%) followed with Fibrobacteres (5%), Spirochaetes (3%) and Proteobacteria (2%). Variations in the abundances 45 of major bacterial phyla attached to the forages have been depicted in Figure 3a. The ANCOM analysis followed by 46 Duncan’s post hoc test revealed differential abundances of Bacteroidetes, Fibrobacteres, Lentisphaerae and 47 48 Spirochaetes among the forages (Figure 3b, ANCOM p < 0.05). Bacteroidetes were significantly overrepresented in 49 AP and CR compared with DP and SC (DMRT Bonferroni p < 0.001). Interestingly, fiber-utilizing bacteria belonging 50 51 to Fibrobacteres were observed in more than 9% of the reads of CR and SC (two forages with the highest initial NDF 52 contents) but in less than 2% of the reads of AP and KS (p < 0.05) and also in DP and RS (Figure 3a). CR contained 53 54 more Lentisphaerae compared with other forages (p < 0.001) while SC carried more Spirochaetes relative to AP (p < 55 0.02). 56 57 58 59 9 60 ScholarOne Support 1-434/964-4100 iewv r R e Pe e Fo r Page 41 of 50 FEMS Microbiology Ecology 1 2 3 At the family level (Figure S5), the forage-attached microbes were affiliated to 76 families, of which nine showed 4 5 differential abundances among the forages, including Bacteroidaceae, Clostridiaceae, Fibrobacteraceae, 6 Victivallaceae, Christensenellaceae, Lachnospiraceae, Spirochaetaceae, Oxalobacteraceae and RFP12 (ANCOM p < 7 8 0.05). Interestingly, Victivallaceae were significantly overrepresented in CR compared with other forages (p < 0.05), 9 while Bacteroidaceae was significantly enriched in CR and RS compared with AP, DP, KS and SC. Oxalobacteraceae 10 was more abundant in KS and SC than in AP, CR, DP and RS (p < 0.05). Members of Fibrobacteraceae also more 11 12 frequently appeared in CR and SC compared with AP, KS and RS with the lowest initial NDF contents (p < 0.02), 13 while those of Clostridiaceae and Lachnospiraceae were underrepresented in CR compared with other forages (p < 14 15 0.05). 16 17 Particle-attached microbiota were affiliated to 119 genera (taxonomic level 6), of which 12 displayed differential 18 19 abundances among the forages (ANCOM p < 0.05, Figure 4) most of which were among high abundant members of 20 rumen community which are known to play key roles in plant lignocellulose degradation, including Ruminococcus, 21 Fibrobacter, Prevotella, Treponema, Lachnospira, Succinivibrio, Pseudobutyrivibrio, Butyrivibrio, Oxalobacter, 22 23 Clostridium, BF311 and Succiniclasticum. Ruminococcus and members of BF311 were more dominant in RS than in 24 AP, KS and SC (p < 0.001). Fibrobacter were more highly represented in CR and SC. Species of Lachnospira were 25 26 present in 0.3% of the reads of AP and KS (p < 0.05) but only 0.11% of CR and 0.09% of SC. Succinivibrio were 27 overrepresented in AP and SC compared with CR (p < 0.05). Species of Prevotella were more abundant in CR (> 21% 28 29 of the reads) than in DP and SC (p < 0.008). Compared to other forages, CR carried less Butyrivibrio and 30 Pseudobutyrivibrio species (p < 0.007). 31 32 33 Changes in forage-attached microbes during the rumen incubation 34 35 In order to examine whether the community composition of forage-attached microbes changed during the rumen 36 incubation, the relative abundances of the microbes were tracked at 24 h intervals (Figure 5). The abundances of eight 37 38 out of the 119 bacterial genera showed significant differences among the incubation lengths (Bonferroni corrected p 39 < 0.05). Interestingly, the proportions of cellulolytic bacteria belonging to unclassified Ruminococcaceae linearly 40 41 increased with the incubation lengths in AP, KS and SC. The proportion of Fibrobacter sharply dropped after the first 42 24 h and reached to an average of 4% between 48 and 96 h of the incubation in CR. Members of Pseudobutyribibrio 43 44 linearly decreased in AP and KS while those of Butyrivibrio increased in CR with the incubation lengths. Members 45 of Clostridium increased in SC but those of Prevotella and unclassified Paraprevotellaceae decreased with the 46 incubation lengths, which were consistent with findings of Cheng et al. (2017). 47 48 49 Relationship between lignocellulose degradation and forage-attached microbial communities 50 51 The Pearson’s correlation between the initial physicochemical properties of the forages and the composition of the 52 53 forage-attached microbiota during initial hours (24 h) of the rumen incubation was performed to determine whether 54 rumen microbes preferred specific forages for attachment. Only correlations with p-values (Bonferroni-corrected) less 55 56 than 0.05 were considered to have significant biological terms. This analysis demonstrated that the prevalence of the 57 58 59 10 60 ScholarOne Support 1-434/964-4100 vie w r R e e or Pe F FEMS Microbiology Ecology Page 42 of 50 1 2 3 family Fibrobacteraceae (r = 0.77, p = 0.03) was positively correlated with NDF contents of the forages. At the genus 4 5 level, the abundance of Fibrobacter (r = 0.77, p = 0.05) was positively but an unclassified Erysipelotrichaceae genus 6 p-75-a5 (r = -0.79, p = 0.03) was negatively correlated with NDF contents in the forages. When hemicellulose contents 7 8 were considered, a negative correlation with members of the family Pirellulaceae (r = -0.79, p = 0.02) was detected. 9 10 We also correlated microbial profiles with physicochemical properties of the forages during the rumen incubation. 11 This analysis revealed that the prevalence of the families Fibrobacteraceae (r= 0.83, p = 0.03 in CR), 12 13 Anaeroplasmataceae (r = 0.82, p = 0.04 in CR), Prevotellaceae (r = 0.83, r = 0.04 in KS) and Paraprevotellaceae (r = 14 0.85, p = 0.02 in SC) were positively but Ruminococcaceae (r = < -0.85, p < 0.01 in CR and SC) and unclassified 15 16 Bacteroidales (r = -0.84, p = 0.05 in RS) were negatively corelated with DM contents of the forages. Particularly, 17 members of Ruminococcaceae were positively correlated with ADF, ADL and hemicellulose contents in CR, DP and 18 19 SC (r > 0.8 and p < 0.04), the forages with the highest initial NDF contents (Figure 1). The population of 20 Lachnospiraceae was also positively correlated (r = 0.83, and p = 0.04) with cellulose content in CR. 21 22 To ascertain whether there was any relationship between the rumen microbiota and lignocellulose degradation, an 23 24 additional Pearson’s correlation analysis was performed between DM loss and the relative abundance of forage- 25 attached microbial communities during the rumen incubation. Interestingly, DM degradation was positively correlated 26 27 with the prevalence of the families Fibrobacteraceae (r > 0.76, p < 0.001 in CR) and Spirochaetaceae (r = 0.91, p = 28 0.0004 in KS) but was negatively corelated with species belonging to unclassified Bacteroidales (r < -0.82, p < 0.05 29 30 in CR and DP), Ruminococcaceae (r = -0.83, p = 0.04 in CR), Mogibacteriaceae (r < -0.82, p < 0.05 in CR and RS) 31 and Erysiopelotrichaceae (r = -0.87, p = 0.01 in CR). The correlations of cellulose and hemicellulose degradations 32 33 with the abundances of forage-attached microbes also showed similar patterns. 34 35 Discussion 36 37 In this study, we investigated the relationship between biomass degradation of and microbial attachment to six 38 39 common lignocellulosic forages varying in their physicochemical properties, including percentages of NDF, ADF, 40 ADL and the contents of cellulose and hemicellulose. Forages containing the highest cellulose contents [(common 41 42 reed (CR) vs. camelthorn (AP)] were degraded to a limited extent during the initial hours of their rumen incubation. 43 The total DM degradation was mainly determined by NDF contents of the forages, because as rice straw (RS) with 44 45 the lowest initial NDF (77.4%) had the fastest (66%) while CR with the highest initial NDF (87.2%) had the lowest 46 (42%) DM loss over 96 h of the incubation. Variability in DM degradation of feeds reflected the differences in 47 lignocellulose composition of their cell walls (Bruno-Soares et al. 2000; Jančík et al. 2010). The rate and extent of 48 49 DM fermentation in the rumen determines the nutrition efficiency of feeds to ruminants (Jančík et al. 2010). 50 51 The 16S rRNA gene sequencing data allowed taxonomic identification and quantification of the rumen microbiota 52 53 tightly attached to the forages. Rarefaction analysis based on the indices reflecting species richness and species relative 54 abundance, e.g. Shannon and Simpson (Kim et al. 2017), indicated that the majority of the diversity of rumen 55 56 microbiota attached to the forages was already sampled. Alpha diversity analysis also showed a limited biologically 57 58 59 11 60 ScholarOne Support 1-434/964-4100 w ev ie er R or Pe F Page 43 of 50 FEMS Microbiology Ecology 1 2 3 significant difference among the forages, which could likely be attributed to a high microbial heterogeneity among 4 5 animals included in the study. The changes in diversity measures were largely restricted to the initial hours of rumen 6 incubation when maximal DM degradation occurred. The differences in species richness and evenness were linked to 7 8 cellulose contents of the forages, where CR and RS with the highest cellulose contents displayed limited species 9 diversity. This result was in agreement with the data on rice straw and alfalfa hay being fed to Holstein cows, in which 10 more bacteria attached to alfalfa with lower NDF (Liu et al. 2016). These findings suggested that only a limited 11 12 fraction of rumen microbiota was capable of attachment to feeds with high lignocellulose contents. 13 14 Analysis of community structure of forage-attached microbiota by Unifrac dissimilarity matrix revealed significant 15 16 differences among the forages and across the length of rumen incubation. Particularly, CR-attached microbiota was 17 well separated from those of other forages at the first 24 h of the incubation. In addition, microbiota attached to CR 18 19 and SC, the two forages with the highest initial NDF contents, were also distantly clustered while those of other 20 forages did not show any apparent separation, indicating the similarity of their communities. These variations could 21 likely be determined by the nature of the forages, e.g., their chemical compositions. Unifrac diversity measure also 22 23 showed that extension of the incubation time contributed to the increased similarity across the samples, as reflected 24 by the overlapped clustering in the PCoA plots for most samples collected at 72 and 96 h of the incubation. This could 25 26 be explained by the fact that initial DM degradation resulted in the reduction of digestible components but the 27 accumulation of indigestible residues which had quite similar properties across the forages, thus favoring the 28 29 attachment of structurally similar communities of rumen microbiota. Huws et al. (2013) also observed similar changes 30 in the microbial communities attached to perennial ryegrass following its rumen incubation. 31 32 33 Members of Firmicutes, Bacteroidetes, Fibrobacteres, Spirochaetes and Proteobacteria were dominant in all samples, 34 accounting for greater than 96% of the bacterial communities attached to the forages, consistent with the findings for 35 rice straw and alfalfa hay (Liu et al. 2016). The abundance of these bacterial phyla varied among the forages and 36 37 across the incubation lengths. Particularly, species of Bacteroidetes were more abundantly attached to the forages with 38 high ADF contents (AP and CR). Bacteroidetes were among highly abundant members of the rumen microbiota being 39 40 best recognized for their saccharolytic activities. The presence of a high number of pectinolytic and cellulolytic 41 enzymes in their genomes clustered with other lignocellulose degrading enzymes into polysaccharide utilization loci 42 43 (PUL) suggested that they are also actively involved in pectin, hemicellulose and cellulose degradation (Gharechahi 44 and Salekdeh 2018; Lapebie et al. 2019; Naas et al. 2014). Within this phylum, members of the families 45 Bacteroidaceae, Prevotellaceae and Paraprevotellaceae displayed significant differences among the forages. 46 47 Particularly, Bacteroidaceae were significantly overrepresented in the forages with the lowest initial ADL contents 48 (CR and RS). At the genus-level, this differential abundance was only affiliated to the BF311, an uncultured and 49 50 unknown rumen bacterium. Prevotellaceae were particularly positively correlated with DM degradation, suggesting 51 their active role in lignocellulose degradation in the rumen. Prevotella spp. are abundant members of the rumen 52 53 microbiome that have a high genetic diversity and thus the ability to thrive on a wide range of substrates, including 54 cellulose, hemicellulose, pectin, proteins and peptides (Dodd et al. 2010; Golder et al. 2014). They are known for their 55 56 xylanolytic properties in the rumen and thus play an important role in fiber degradation (Dodd et al. 2010). 57 58 59 12 60 ScholarOne Support 1-434/964-4100 w ev ie er R Pe Fo r FEMS Microbiology Ecology Page 44 of 50 1 2 3 Fibrobacteres were predominantly represented in microbiota attached to the forages with the highest NDF contents 4 5 [e.g. CR and Salicornia (SC)]. They are known to produce a battery of cellulolytic enzymes capable of degrading 6 cellulose as a sole carbon source (Flint et al. 2008; Suen et al. 2011). The association of forage NDF content with the 7 8 prevalence of Fibrobacter is also recently observed in the cow rumen microbiota (Liu et al. 2016). In contrast to 9 Fibrobacteres, the genera Clostridium, Shuttleworthia and Ruminococcus tended to attach the forages with limited 10 NDF and cellulose contents. This is consistent with previous findings on the preference of Ruminococcus species for 11 12 attachment to high quality sugar-rich hays (Klevenhusen et al. 2017). Ruminococcaceae showed a strong negative 13 correlation with total DM contents in the forages as well. Shinkai et al. (2010) also reported that members of 14 15 Fibrobacteres, including F. succinogenes, abundantly attached to less digestible fibers while those of 16 Ruminococcaceae, specifically R. flavefaciens, preferred easily digestible fibers and thus were infrequently detected 17 18 in the stem parts of hays. The rate and extent of fiber degradation by F. succinogenes are also greater than R. albus 19 and R. flavefaciens (Kobayashi et al. 2008). Early microscopy analysis indicated that R. albus is less commonly 20 attached to plant cell walls while F. succinogenes forms extensive microcapsula enveloping the cell walls (Chesson 21 22 et al. 1986). The community of unclassified Ruminococcaceae tended to linearly expand in AP, KS and SC following 23 their rumen incubation. This finding suggested that the degradation of surface accessible fibers turned forage residues 24 25 into favored substrates for the attachment by species of Ruminococcaceae. A declined abundance of species of 26 Ruminococcus has been reported in the cow rumen when the starch content of diet was increased (Zened et al. 2013). 27 28 Ruminococcus spp. have been equipped with specialized mechanisms for fiber adhesion and degradation, in which 29 multiple carbohydrate degrading enzymes assemble into multienzyme cellulosome complexes capable of attachment 30 31 to and degradation of polysaccharides in various plant cell walls (Bayer et al. 2004; Doi and Kosugi 2004). 32 33 Recently, Huws et al. (2016) reported that microbial colonization of perennial ryegrass occurred at two successive 34 stages: The primary (the first 1-2 h after its rumen entry) and the secondary phases (4-8 h). It was the secondary phase, 35 36 during which the maximal DM degradation was achieved and the proportion of Succinivibrio spp. decreased while 37 those of Pseudobutyrivibrio, Roseburia and Ruminococcus spp. increased. Piao et al. (2014) also observed that the 38 39 populations of Pseudobutyrivibrio and Ruminococcus spp. increased during their secondary colonization to 40 switchgrass in the Friesian cow rumen. The microbial communities attached to feeds during the primary phase were 41 42 believed to utilize soluble and easily accessible nutrients while those colonized feeds during the secondary phase were 43 considered to be the true lignocellulose degraders (Huws et al. 2016; Liu et al. 2016). In this study, an increased 44 prevalence of Butyrivibrio was noted in CR. Butyrivibrio spp. are known for their proteolytic, hemicellulolytic and 45 46 biohydrogenating activities (Krause et al. 2003). Liu et al. (2016) reported a strong positive correlation between the 47 abundance of Butyrivibrio spp. and the crude protein contents in the rice straw and alfalfa hay, suggesting their 48 49 preference for the attachment to proteinaceous components of the feeds. 50 51 In summary, our results demonstrated that physicochemical compositions of the forages, more specifically their 52 53 cellulose contents, were among the major factors influencing microbial attachments and thus lignocellulose 54 degradation in the cattle rumen. No taxonomic lineage was found to be specific of a particular forage, suggesting that 55 56 most rumen microbes had an inherent tendency for attachment to lignocellulosic substrates. However, differential 57 58 59 13 60 ScholarOne Support 1-434/964-4100 w vie r R e e or Pe F Page 45 of 50 FEMS Microbiology Ecology 1 2 3 attachments to the forages by rumen microbiota suggested that the physicochemical properties were the key factors 4 5 influencing the rate of microbial attachment. Our results also revealed that members of the rumen microbiota competed 6 for their attachments to the forages with different lignocellulose compositions mostly during the initial hours of the 7 8 rumen incubation. However, after the degradation of easily digestible components in the forages, relatively uniform 9 microbial communities developed on their surface. 10 11 It should be noted that the bacterial community composition detected in each forage at each sampling interval may 12 13 not necessarily reflect the actual abundance of microbes due to an inherent bias in sample collection, processing and 14 storage as well as DNA extraction and PCR amplification, which may cause over- or under-representation of some 15 16 taxa. For example, a recent comparative analysis of various DNA extraction methods used in rumen microbiome 17 analyses argued that DNA extraction method can have an adverse effect on the community composition of rumen 18 19 samples and may associate with an increased or decreased abundance of some specific taxa (Henderson et al. 2013). 20 Particularly, the inclusion of a physical lysis step using bead-beating method increased the efficacy of DNA extraction 21 from Gram-positive bacteria (Knudsen et al. 2016). The extent to which our DNA extraction method, in which the 22 23 physical lysis step was not included, affected community composition of the forage-attached microbiota was unknown. 24 In addition, the high number of PCR cycles may increase the number of PCR artifacts and bias the microbial 25 26 compositions. 27 28 Acknowledgements 29 30 31 This research was funded by the Agricultural Biotechnology Research Institute of Iran (ABRII), the international 32 cooperation and exchange program of the National Natural Science Foundation of China (No. 31461143020), and the 33 34 Chinese government contribution to CAAS-ILRI Joint Laboratory on Livestock and Forage Genetic Resources in 35 Beijing. The paper contributes to the CGIAR Research Program on Livestock. 36 37 Figure legends 38 39 40 Figure 1. Differences among the six forages in dry matter (DM) and percentages of neutral detergent fiber (NDF), 41 acid detergent fiber (ADF), acid detergent lignin (ADL), cellulose and hemicellulose before (0 h) and after 24, 48, 72, 42 43 and 96 h of the rumen incubation. The solid circle and error bar in each boxplot show mean and standard deviation, 44 respectively. Statistically significant differences were determined using one-way ANOVA. Means were compared 45 46 using Duncan’s multiple range test and are labelled with different letters at each sampling interval for p < 0.05. AP, 47 camelthorn; CR, common reed; DP, date palm; KS, Kochia; RS, rice straw; and SC, Salicornia. 48 49 Figure 2. Beta diversity analysis of rumen microbial communities attached to the six forages during the rumen 50 51 incubation. PCoA plots show the distribution of samples based on weighted Unifrac distance matrix, in which samples 52 are grouped according to the forages (a) and the sampling intervals (b). Significant differences were tested using 53 54 PERMANOVA with a p-value cutoff 0.01. The percentage of variation explained by each principle coordinate is 55 indicated next to the corresponding axis. The homogeneity of dispersions was also tested for this diversity measure to 56 57 58 59 14 60 ScholarOne Support 1-434/964-4100 iewv r R e e or Pe F FEMS Microbiology Ecology Page 46 of 50 1 2 3 examine variance differences between the forages (c) and the sampling intervals (d). Significant differences were 4 5 determined using the betadisper function of R package vegan v2.5-5 at 999 permutations. P-values < 0.05 were 6 considered statistically significant. For each treatment, data in triplicate representing three separate bags, were 7 8 included. AP, camelthorn; CR, common reed; DP, date palm; KS, Kochia; RS, rice straw; and SC, Salicornia. 9 10 Figure 3. (a) The proportions of reads (mean ± SD) assigned to the major bacterial phyla (n = 7) detected in the forage- 11 attached microbial communities. The relative abundances were calculated based on the proportion of reads assigned 12 13 to each phylum in the rarefied OTU table at an even sequencing depth of 4660 reads. (b) Box plots show relative 14 abundance of phyla differentially attached to the forages. Statistically significant differences were calculated based on 15 16 the analysis of composition of microbes (ANCOM) using p < 0.05 following Duncan’s multiple range test for 17 comparison of the means. Center lines indicate the median value, boxes the interquartile range and red squares the 18 19 mean. Means with different letters differ at Bonferroni corrected p < 0.05. AP, camelthorn; CR, common reed; DP, 20 date palm; KS, Kochia; RS, rice straw; and SC, Salicornia. 21 22 Figure 4. The log of relative abundances of genera differentially attached to the six forages following the rumen 23 24 incubation. Differential abundances were statistically tested using ANCOM with a p-value cutoff < 0.05. Means were 25 compared using Duncan’s multiple range test only accepting Bonferroni corrected p-values < 0.05. Boxplots labeled 26 27 with different letters show statistically significant differences. Center line represents median value. AP, camelthorn; 28 CR, common reed; DP, date palm; KS, Kochia; RS, rice straw; and SC, Salicornia. 29 30 31 Figure 5. The average relative abundances of taxa (genus level) differentially represented in microbial communities 32 attached to the six forages during the rumen incubation (up to 96 h with 24 h intervals). Statistically significant 33 34 differences were calculated using ANCOM and means were compared using Duncan’s multiple range test. Error bars 35 represent standard deviations and their lengths are adjusted at 95% confidence interval. Means with different letters 36 are statistically significant at Bonferroni corrected p < 0.05. No taxa were found to be differentially represented in DP. 37 38 AP, camelthorn; CR, common reed; KS, Kochia; RS, rice straw; and SC, Salicornia. “Un” refers to unclassified. 39 40 Supplementary materials 41 42 Figure S1. 43 Changes in DM contents and percentages of NDF, ADF, ADL, cellulose and hemicellulose in six forages during the 44 45 rumen incubation. The solid circle in each boxplot shows mean and error bars represent standard deviation. 46 Statistically significant differences were determined using one-way ANOVA. Means were compared using Duncan’s 47 multiple range test and labelled with different letters for p < 0.05. 48 49 Figure S2. 50 Alpha diversity analysis through rarefaction plots. (a) Rarefaction curves showing the increase in number of observed 51 52 OTUs; (b) Species richness (number of observed OTUs + number of unobserved OTUs, Chao1); (c) Shannon; and (d) 53 Simpson diversity indices on Y-axis as a function of the number of reads sampled on X-axis. 54 55 Figure S3. 56 57 58 59 15 60 ScholarOne Support 1-434/964-4100 vie w e er R e or P F Page 47 of 50 FEMS Microbiology Ecology 1 2 3 Alpha diversity indices of rumen microbiota attached to the six forages with different lignocellulosic compositions at 4 5 24, 48, 72 and 96 hours of the rumen incubation. Alpha diversity indices were measured based on OTUs present at an 6 even sequencing depth of 4,660 reads (corresponding to the sequencing depth of the sample with the lowest number 7 8 of reads) in all samples. Statistically significant differences were determined using one-way ANOVA and means were 9 compared by Duncan’s multiple range test. * p value < 0.05, ** p value < 0.01 and *** p value < 0.001. 10 Figure S4. 11 12 Alpha diversity analysis of rumen microbial communities attached to the six forages during the rumen incubation. The 13 OTU table for each forage was sampled at an even sequencing depth (n = 4660) and diversity measures were calculated 14 15 using plot_richness function of R package phyloseq and visualized using geom_boxplot function of R package 16 ggplot2. Statistically significant differences were calculated using one-way ANOVA at p-value cutoff of 0.05. 17 18 Differences between means were determined using Duncan’s multiple range test in PAST v3.26. * p value < 0.05, ** 19 p value < 0.01 and *** p value < 0.001. 20 Figure S5. 21 22 The proportions of reads affiliated to OTUs at the taxonomic level of family in the six forages during the rumen 23 incubation. Data for each sampling interval is the mean of three biological replicates corresponding to three bags 24 25 retrieved one each from the three cannulated bulls. For each forage, the OTU table is rarefied to the sequencing depth 26 of the sample with the lowest number of reads. 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 16 60 ScholarOne Support 1-434/964-4100 vie w er Re or Pe F FEMS Microbiology Ecology Page 48 of 50 1 2 3 References 4 5 6 Bayer EA, Belaich JP, Shoham Y et al. The cellulosomes: multienzyme machines for degradation of plant cell wall 7 polysaccharides. Annu Rev Microbiol 2004;58: 521-54. 8 Bergman EN. Energy contributions of volatile fatty acids from the gastrointestinal tract in various species. Physiol 9 Rev 1990;70: 567-90.Bruno-Soares AM, Abreu JMF, Guedes CVM et al. Chemical composition, DM and NDF degradation kinetics in 10 rumen of seven legume straws. Animal Feed Science and Technology 2000;83: 75-80. 11 Caporaso JG, Bittinger K, Bushman FD et al. PyNAST: a flexible tool for aligning sequences to a template alignment. 12 Bioinformatics 2010a;26: 266-7. 13 Caporaso JG, Kuczynski J, Stombaugh J et al. QIIME allows analysis of high-throughput community sequencing data. 14 Nat Methods 2010b;7: 335-6. 15 Cheng Y, Wang Y, Li Y et al. Progressive Colonization of Bacteria and Degradation of Rice Straw in the Rumen by 16 Illumina Sequencing. Front Microbiol 2017;8: 2165. 17 Chesson A, Stewart CS, Dalgarno K et al. Degradation of isolated grass mesophyll, epidermis and fibre cell walls in 18 the rumen and by cellulolytic rumen bacteria in axenic culture. Journal of Applied Bacteriology 1986;60: 19 327-36. 20 De Mulder T, Goossens K, Peiren N et al. Exploring the methanogen and bacterial communities of rumen 21 environments: solid adherent, fluid and epimural. FEMS Microbiol Ecol 2017;93. 22 DeSantis TZ, Hugenholtz P, Larsen N et al. Greengenes, a chimera-checked 16S rRNA gene database and workbench 23 compatible with ARB. Appl Environ Microbiol 2006;72: 5069-72. 24 Dhariwal A, Chong J, Habib S et al. MicrobiomeAnalyst: a web-based tool for comprehensive statistical, visual and 25 meta-analysis of microbiome data. Nucleic Acids Res 2017;45: W180-W88. 26 Dodd D, Moon YH, Swaminathan K et al. Transcriptomic analyses of xylan degradation by Prevotella bryantii and 27 insights into energy acquisition by xylanolytic bacteroidetes. J Biol Chem 2010;285: 30261-73. 28 Doi RH, Kosugi A. Cellulosomes: plant-cell-wall-degrading enzyme complexes. Nat Rev Microbiol 2004;2: 541-51. 29 Edgar RC. Search and clustering orders of magnitude faster than BLAST. Bioinformatics 2010;26: 2460-1. 30 Edwards JE, Huws SA, Kim EJ et al. Characterization of the dynamics of initial bacterial colonization of nonconserved 31 forage in the bovine rumen. FEMS Microbiol Ecol 2007;62: 323-35. 32 Flint HJ, Bayer EA, Rincon MT et al. Polysaccharide utilization by gut bacteria: potential for new insights from genomic analysis. Nat Rev Microbiol 2008;6: 121-31. 33 Gharechahi J, Kharazian ZA, Sarikhan S et al. The dynamics of the bacterial communities developed in maize silage. 34 Microb Biotechnol 2017;10: 1663-76. 35 Gharechahi J, Salekdeh GH. A metagenomic analysis of the camel rumen's microbiome identifies the major microbes 36 responsible for lignocellulose degradation and fermentation. Biotechnol Biofuels 2018;11: 216. 37 Gharechahi J, Zahiri HS, Noghabi KA et al. In-depth diversity analysis of the bacterial community resident in the 38 camel rumen. Syst Appl Microbiol 2015;38: 67-76. 39 Godoy-Vitorino F, Goldfarb KC, Karaoz U et al. Comparative analyses of foregut and hindgut bacterial communities 40 in hoatzins and cows. ISME J 2012;6: 531-41. 41 Goering HK. Forage fiber analyses (apparatus, reagents, procedures, and some applications). In: Van Soest PJ (ed.). 42 Washington, D.C. :: Agricultural Research Service, U.S. Dept. of Agriculture, 1970. 43 Golder HM, Denman SE, McSweeney C et al. Ruminal bacterial community shifts in grain-, sugar-, and histidine- 44 challenged dairy heifers. J Dairy Sci 2014;97: 5131-50. 45 Guarner F, Malagelada JR. Gut flora in health and disease. Lancet 2003;361: 512-9. 46 Hammer Ø, Harper DA, Ryan PD. PAST: paleontological statistics software package for education and data analysis. 47 2001. 48 Henderson G, Cox F, Kittelmann S et al. Effect of DNA extraction methods and sampling techniques on the apparent 49 structure of cow and sheep rumen microbial communities. PLoS One 2013;8: e74787. 50 Hungate RE. The rumen and its microbes: Elsevier, 2013. 51 Huws SA, Edwards JE, Creevey CJ et al. Temporal dynamics of the metabolically active rumen bacteria colonizing 52 fresh perennial ryegrass. FEMS Microbiol Ecol 2016;92. 53 Huws SA, Mayorga OL, Theodorou MK et al. Successional colonization of perennial ryegrass by rumen bacteria. Lett 54 Appl Microbiol 2013;56: 186-96. 55 Jami E, Israel A, Kotser A et al. Exploring the bovine rumen bacterial community from birth to adulthood. ISME J 2013;7: 1069-79. 56 57 58 59 17 60 ScholarOne Support 1-434/964-4100 iew Re v ee r or P F Page 49 of 50 FEMS Microbiology Ecology 1 2 3 Jami E, Mizrahi I. Composition and similarity of bovine rumen microbiota across individual animals. PLoS One 4 2012;7: e33306. 5 Jančík F, Koukolová V, Homolka P. Ruminal degradability of dry matter and neutral detergent fibre of grasses. Czech 6 J Anim Sci 2010;55: 359-71. 7 Jin W, Wang Y, Li Y et al. Temporal changes of the bacterial community colonizing wheat straw in the cow rumen. 8 Anaerobe 2018;50: 1-8. 9 Khafipour E, Li S, Plaizier JC et al. Rumen microbiome composition determined using two nutritional models of 10 subacute ruminal acidosis. Appl Environ Microbiol 2009;75: 7115-24. 11 Kim BR, Shin J, Guevarra R et al. Deciphering Diversity Indices for a Better Understanding of Microbial 12 Communities. J Microbiol Biotechnol 2017;27: 2089-93. 13 Klevenhusen F, Petri RM, Kleefisch MT et al. Changes in fibre-adherent and fluid-associated microbial communities 14 and fermentation profiles in the rumen of cattle fed diets differing in hay quality and concentrate amount. 15 FEMS Microbiol Ecol 2017;93. 16 Knudsen BE, Bergmark L, Munk P et al. Impact of Sample Type and DNA Isolation Procedure on Genomic Inference 17 of Microbiome Composition. mSystems 2016;1. 18 Kobayashi Y, Shinkai T, Koike S. Ecological and physiological characterization shows that Fibrobacter succinogenes 19 is important in rumen fiber digestion - review. Folia Microbiol (Praha) 2008;53: 195-200. 20 Kocherginskaya SA, Aminov RI, White BA. Analysis of the Rumen Bacterial Diversity under two Different Diet Conditions using Denaturing Gradient Gel Electrophoresis, Random Sequencing, and Statistical Ecology 21 Approaches. Anaerobe 2001;7: 119-34. 22 Kong Y, Teather R, Forster R. Composition, spatial distribution, and diversity of the bacterial communities in the 23 rumen of cows fed different forages. FEMS Microbiol Ecol 2010;74: 612-22. 24 Krause DO, Denman SE, Mackie RI et al. Opportunities to improve fiber degradation in the rumen: microbiology, 25 ecology, and genomics. FEMS Microbiol Rev 2003;27: 663-93. 26 Lapebie P, Lombard V, Drula E et al. Bacteroidetes use thousands of enzyme combinations to break down glycans. 27 Nat Commun 2019;10: 2043. 28 Liu J, Zhang M, Xue C et al. Characterization and comparison of the temporal dynamics of ruminal bacterial 29 microbiota colonizing rice straw and alfalfa hay within ruminants. J Dairy Sci 2016;99: 9668-81. 30 Mackie RI. Mutualistic fermentative digestion in the gastrointestinal tract: diversity and evolution. Integr Comp Biol 31 2002;42: 319-26. 32 Magoc T, Salzberg SL. FLASH: fast length adjustment of short reads to improve genome assemblies. Bioinformatics 33 2011;27: 2957-63. 34 Mandal S, Van Treuren W, White RA et al. Analysis of composition of microbiomes: a novel method for studying 35 microbial composition. Microb Ecol Health Dis 2015;26: 27663. 36 Martineau R, Proulx JG, Cortes C et al. Two-Stage Rumen Cannulation Technique in Dairy Cows. Vet Surg 2015;44: 37 551-6. 38 McAllister TA, Bae HD, Jones GA et al. Microbial attachment and feed digestion in the rumen. J Anim Sci 1994;72: 39 3004-18. 40 Naas AE, Mackenzie AK, Mravec J et al. Do rumen Bacteroidetes utilize an alternative mechanism for cellulose 41 degradation? mBio 2014;5: e01401-14. 42 Petri RM, Forster RJ, Yang W et al. Characterization of rumen bacterial diversity and fermentation parameters in 43 concentrate fed cattle with and without forage. J Appl Microbiol 2012;112: 1152-62.Petri RM, Pourazad P, Khiaosa-Ard R et al. Temporal dynamics of in-situ fiber-adherent bacterial community under 44 ruminal acidotic conditions determined by 16S rRNA gene profiling. PLoS One 2017;12: e0182271. 45 Piao H, Lachman M, Malfatti S et al. Temporal dynamics of fibrolytic and methanogenic rumen microorganisms 46 during in situ incubation of switchgrass determined by 16S rRNA gene profiling. Front Microbiol 2014;5: 47 307. 48 Rognes T, Flouri T, Nichols B et al. VSEARCH: a versatile open source tool for metagenomics. PeerJ 2016;4: e2584. 49 Sadet S, Martin C, Meunier B et al. PCR-DGGE analysis reveals a distinct diversity in the bacterial population 50 attached to the rumen epithelium. Animal 2007;1: 939-44. 51 Shinkai T, Ueki T, Kobayashi Y. Detection and identification of rumen bacteria constituting a fibrolytic consortium 52 dominated by Fibrobacter succinogenes. Anim Sci J 2010;81: 72-9. 53 Suen G, Weimer PJ, Stevenson DM et al. The complete genome sequence of Fibrobacter succinogenes S85 reveals a 54 cellulolytic and metabolic specialist. PLoS One 2011;6: e18814. 55 Van Soest PJ, Robertson JB, Lewis BA. Methods for dietary fiber, neutral detergent fiber, and nonstarch 56 polysaccharides in relation to animal nutrition. J Dairy Sci 1991;74: 3583-97. 57 58 59 18 60 ScholarOne Support 1-434/964-4100 w vie er Re e or P F FEMS Microbiology Ecology Page 50 of 50 1 2 3 Wang Q, Garrity GM, Tiedje JM et al. Naive Bayesian classifier for rapid assignment of rRNA sequences into the 4 new bacterial taxonomy. Appl Environ Microbiol 2007;73: 5261-7. 5 Zened A, Combes S, Cauquil L et al. Microbial ecology of the rumen evaluated by 454 GS FLX pyrosequencing is 6 affected by starch and oil supplementation of diets. FEMS Microbiol Ecol 2013;83: 504-14. 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 19 60 ScholarOne Support 1-434/964-4100 vie w er Re Pe Fo r