Effect of partially replacing a barley-based concentrate with flaxseed-based products on the rumen bacterial population of lactating Holstein dairy cows

Aims: The effects of partial replacement of a barley-based concentrate with ﬂaxseed-based products on the rumen bacterial population of lactating Holstein dairy cows were evaluated. Methods and Results: Treatments fed were CONT, a normal diet that included barley silage, alfalfa hay and a barley-based concentrate that contained no ﬂaxseed or faba beans; FLAX, inclusion of a nonextruded ﬂaxseed-based product containing 55 (cid:1) 0% ﬂaxseed, 37 (cid:1) 8% ﬁeld peas and 6 (cid:1) 9% alfalfa; EXT, similar to FLAX, but the product was extruded and EXTT, similar to FLAX, but product was extruded and ﬁeld gene Most abundant Paraprevotellaceae.


Introduction
Enriched unsaturated fatty acid animal products may offer beneficial effects to human health (Lee et al. 2005). Increasing the concentration of these fatty acids in ruminant products, however, is difficult due to intensive biohydrogenation by ruminal micro-organisms (Palmquist et al. 1993;Bauman and Griinari 2001;Jenkins et al. 2008). Thus, there has been significant interest to develop feeding strategies to decrease ruminal biohydrogenation of unsaturated fatty acids while ensuring high availability in the small intestine (Beam et al. 2000). Supplementation with raw flaxseed or extruded flaxseed has been suggested to be an effective strategy to increase the availability of polyunsaturated fatty acids in the small intestine (Litton 2008;He et al. 2012). For example, extruded flaxseed increased the content of alinolenic acid in blood serum or milk of dairy cattle (Kennelly 1996;Oeffner et al. 2013;Moats et al. 2015). Furthermore, the inclusion of dietary tannins in ruminant rations may be an effective approach for mitigating the biohydrogenation of polyunsaturated fatty acids (Vasta et al. 2009). However, dietary fat (Enjalbert et al. 2017) or tannins (Vasta et al. 2010) may have detrimental effects on bacterial taxa within the rumen or may negatively affect animal production performance (Vasta et al. 2010). Nonetheless, most studies evaluating these dietary strategies for mitigating fatty acid biohydrogenation have not determined effects on the broad bacterial community structure of the rumen of lactating Holstein cows. Consequently, the impact of dietary unsaturated fatty acids, feed extrusion and tannins on the broad bacterial population remains poorly understood in vivo. This research gap can now be addressed using molecular techniques in combination with bioinformatics, which have enabled researchers to evaluate the impact of diets on bacterial community structure (Krause et al. 2013;Chaucheyras-Durand and Ossa 2014;Castillo-Lopez et al. 2017). For example, high-throughput DNA sequencing provides new insights into the broad bacterial population of the rumen (Callaway et al. 2010;Aldai et al. 2012;Castillo-Lopez et al. 2014).
In addition, the bacterial population of the rumen influences production performance (Myer et al. 2015), ruminal fermentation (Fernando et al. 2010;Anderson et al. 2016), ruminal pH and fermentation efficiency (Callaway et al. 2010), metabolizable protein supply (Castillo-Lopez et al. 2013) and milk composition in dairy cattle (Jami et al. 2014). Consequently, investigating the effects of diet composition and ingredients being generated by the dairy feeding industry on the rumen bacterial community is essential not only for improving milk yield and composition, but also for preventing negative impacts on rumen function. Therefore, the objective of this study was to evaluate the effect of partial replacement of a barley-based concentrate with different flaxseedbased products on the rumen bacterial community structure of lactating Holstein dairy cows, assessed with highthroughput DNA sequencing. Our hypothesis is that inclusion of flaxseed-based products in dairy rations will shift the abundance of ruminal bacteria.

Animal care and housing, and experimental design
This experiment was conducted at the University of Saskatchewan Rayner Dairy Cattle Research and Teaching Facility (Saskatoon, Saskatchewan, Canada); it was performed in accordance with the guidelines published by the Canadian Council on Animal Care (1997). The protocols used in this study were preapproved by the University of Saskatchewan Animal Care and Use Committee (protocol number 20040048).
A total of eight multiparous, lactating Holstein cows from the University of Saskatchewan Greenbrae herd (mean and SD, 116Á5 AE 17Á5 DIM; 712Á7 AE 92Á3 kg BW) were used in a replicated 4 9 4 Latin square experimental design. Four of these cows were fitted with permanent ruminal cannulae to facilitate ruminal digesta sampling for microbial community analyses. Each experimental period comprised 28 days, which consisted of 26 days for dietary adaptation followed by 2 days for sample collection to provide enough time for animal adaptation to treatment change (Lillis et al. 2011;Boots et al. 2013). Cows were housed in individual tie stalls with continuous access to fresh water and feed except during milking. Animals were milked three times daily at 04:30, 12:30 and 19:00 h in a double six Herringbone parlour (DeLaval International, Peterborough, ON). The individual tie stalls were equipped with rubber mats. In addition, wood shavings were used for bedding and were replaced daily.

Experimental treatments, feed samples and feed chemical analysis
Rations were offered twice daily at 09:30 and 17:00 h for ad libitum access as total mixed rations. Each ration was mixed using a small-batch mixing cart (Data Ranger, American Calan, Northwood, NH). Treatments (DM basis; Table 1) were (i) CONT, a normal diet containing 28Á1% barley silage, 20Á0% alfalfa hay and 51Á8% of a barley-based concentrate that contained no flaxseed or faba beans; (ii) FLAX, inclusion of 11Á4% of a nonextruded flaxseed-based product which contained 55Á0% flaxseed, 37Á8% ground field peas and 6Á9% dehydrated alfalfa; (iii) EXT, inclusion of 11Á4% of an extruded flaxseed-based product which contained 55Á0% flaxseed, 37Á8% ground field peas and 6Á9% dehydrated alfalfa and (iv) EXTT, inclusion of 11Á4% of an extruded flaxseedbased product which contained 55Á0% flaxseed, 37Á8% ground high-tannin faba beans and 6Á9% dehydrated alfalfa.
The barley-based concentrate was partially substituted with the inclusion of the corresponding flaxseed-based product in FLAX, EXT and EXTT, and these products included 0Á4% of mould inhibitor plus vitamin E as antioxidant. High-tannin faba beans corresponded to variety Malik 9-4. Experimental diets were formulated based on two factors: (i) providing similar levels of net energy for lactation and (ii) achieving dietary ether extract levels approaching, but not exceeding 6% (DM basis) for the three flaxseed-containing treatments. All flaxseed-based products were manufactured and supplied by a local company (Oleet Processing Ltd., a division of O&T Farms Ltd., Regina, SK, Canada). Extruded flaxseed-based products were manufactured using a dry extrusion method with a single screw extruder (Model 2500; Insta-Pro International, Urbandale, IA) with barrel temperature averaging 120°C.
Samples of rations, barley silage, alfalfa hay, barleybased concentrate and flaxseed-based products were collected daily from day 21 to 28 of each period and pooled by treatment within each period. Feed samples were stored at À20°C pending analysis for chemical composition. In addition, barley silage samples were collected twice a week during the experiment for microwave DM determination (Valkeners et al. 2008). Briefly, a sample of approximately 100 g was heated in a microwave oven for 4 min. During the second step, drying time was decreased to 30 s, and the second step was repeated until obtaining a constant weight in two consecutive measurements. To avoid burning of the sample, a glass of water was also placed in the microwave. These DM data were used for adjusting the diet DM to ensure proper inclusion of ingredients in each treatment.
Alfalfa hay, barley silage and concentrate samples were dried at 55°C in a forced air oven for 48 h, and orts were freeze dried. Dried feed ingredients and orts were then ground to pass through a 1-mm screen (Christy-Norris mill, Christy and Norris Ltd., Chelmsford, UK) and analysed for chemical composition by an external laboratory (Cumberland Valley Analytical Services, Hagerstown, MD), which included DM (method no. 930.15;AOAC 2000), N (method no. 990.03; Leco FP-528 Nitrogen Combustion Analyzer; Leco Corp., St. Joseph, MI), NDF (Van Soest et al. 1991), starch (Hall 2009(Hall ), ether extract using diethyl ether (method no. 2003AOAC 2006) and ash (method no. 942.05;AOAC 2000). The nutrient composition of each total mixed ration (Table 2) was calculated based on analysis of individual ingredients, barley-based concentrate and flaxseed-based products and the rate of inclusion in the respective treatment. This method of reporting chemical composition of dairy diets is highly recommended, because when sampling total mixed rations for analysis of chemical composition results may be affected by sampling variation (Weiss et al. 2016). Feed fatty acid analysis was conducted at Lipid Analytical Services Ltd. (Guelph, ON, Canada); concentration of fatty acids was expressed as per cent of fatty acids methyl esters. The chemical analyses of individual feed ingredients were then used to calculate the chemical composition of experimental diets. In addition, feed samples were submitted to Lethbridge Research Centre (Lethbridge, AB) for determination of tannins using the acid-butanol assay (Porter et al. 1986).

Sampling of whole ruminal contents for bacterial community analysis
On days 27 and 28 of each experimental period, samples of intact, nonstrained ruminal contents (solid and liquid fractions) were taken using new palpation sleeves for each cow at each sampling time point. In order to obtain representative samples from the rumen of each cow, grab samples were taken from the caudal ventral sac, cranial ventral sac and two samples from the ruminal digesta mat in the dorsal rumen of each animal; samples were collected so that every 6-h interval in a 24-h period was represented. Specifically, these samples were collected at 10:00, 16:00 and 22:00 h on day 27, and 04:00 h on day 28. Within each time point, samples collected from the same cow were pooled, and a 10-ml subsample was Extruded flaxseed-based product with tannins 0 0 0 1 1 Á4 *CONT: a normal diet including barley silage, alfalfa hay and a barley-based concentrate with no flaxseed or faba beans; FLAX: inclusion of 11Á4% of a nonextruded flaxseed-based product containing flaxseed, field peas and alfalfa; EXT: similar to FLAX, but the product was extruded; EXTT: similar to FLAX, but product was extruded and field peas were replaced by high-tannin faba beans. placed in a sterile 15-ml vial and immediately snap frozen at À80°C. Thus, a total of 64 composited ruminal digesta content samples were collected during the trial (four cows 9 four time points 9 four experimental periods).
To obtain digesta samples representative of a 24-h period, at the end of the experiment, these samples were pooled to obtain one sample per cow within each of the four periods as previously outlined and conducted by other researchers for samples collected from cattle for microbial community evaluations (Lillis et al. 2011;Boots et al. 2013); these samples were used for DNA extraction, sequencing and bacterial phylogenetic analysis.  (Paz et al. 2016;Xie et al. 2016). The quality of the amplified DNA was verified by resolving on a 1Á5% agarose gel. Amplicons from each sample were pooled in equal amounts using the epMotion M5073 automated system (Eppendorf, Hauppauge, NY) and the resulting pooled library was purified using the Pippin Prep kit (Sage Science, Beverly, MA) according to the manufacturer's instructions, and analysed according to the Bio Analyzer High Sensitive DNA kit (Agilent Technologies, Santa Clara, CA); then, DNA concentration was measured with a Qubit 2Á0 fluorometer (Invitrogen, Carlsbad, CA) and the library was stored at À20°C for later analyses.

High-throughput DNA sequencing and bacterial phylogenetic analysis
The amplicon library was subjected to high-throughput DNA sequencing at the Department of Animal Science of University of Nebraska-Lincoln according to the protocol utilized by Paz et al. (2016) and Xie et al. (2016). Briefly, this method was conducted using the Ion Torrent Personal Genome Machine (PGM; Life Technologies, ND, not detected. *CONT: a normal diet including barley silage, alfalfa hay and a barley-based concentrate with no flaxseed or faba beans; FLAX: inclusion of 11Á4% of a nonextruded flaxseed-based product containing flaxseed, field peas and alfalfa; EXT: similar to FLAX, but the product was extruded; EXTT: similar to FLAX, but product was extruded and field peas were replaced by high-tannin faba beans. †Analysis conducted at Cumberland Valley Analytical Services, Hagerstown, MD. Carlsbad, CA), and applying the Sequencing Kit v2 on a 316 chip according to the manufacturer's instructions; then, the high-quality DNA sequences were binned into their respective samples based on their barcodes. Specifics for the methods used for emulsion PCR, bead deposition and sequencing on the PGM were as described by the manufacturer. The rationale behind the sequencing direction is to minimize sequencing errors. Because there is slightly more sequence variability towards the 518 end compared to the 341 end (Vasileiadis et al. 2012), and because the beginning of sequencing has less errors, sequencing started from the 518R end and moved towards the 341F end (which has less variability). Thus, there was less chance of sequencing errors on the region of the read which has more variability. Paired-end amplicons were not generated during this sequencing run; therefore, only single read sequences were generated and no contig assembly was performed. Sequenced data were deposited in the NCBI Sequence Read Archive under the accession no. SRR6023841. Sequence reads were analysed using published bioinformatics pipelines UPARSE (drive5.com/uparse/; Edgar et al. 2011), QIIME (qiime.org; Caporaso et al. 2010) and MOTHUR (Schloss et al. 2009). Initial quality control of the generated sequences was performed using the Torrent Suite Software ver. 3.6.2 as outlined by Paz et al. (2016), which included trimming of the 3 0 end of sequences that dropped below the average Q15 score over a 30-bp window and removing sequences with unidentified bases (N). Resulting sequences were downloaded from the Torrent Suite and demultiplexed using the QIIME software package (ver. 1.9.1) (Caporaso et al. 2010). During demultiplexing, sequences with an average quality score <25 were removed. Following demultiplexing, universal primers used for sequencing were removed, allowing one mismatch in the 5 0 (518R) primer and two in the 3 0 reverse primer (341F). Sequences shorter than 130 bp were removed and remaining sequences were trimmed to a fixed length of 130 bp (Paz et al. 2016). Quality trimmed sequences were then reverse complemented, screened for chimeric sequences using UCHIME (Edgar et al. 2011), and preclustered using the pseudo-single linkage-clustering algorithm to remove reads that resulted from sequencing errors (Huse et al. 2010). These sequences were then assigned to operational taxonomic units (OTUs) at 97% similarity using UPARSE pipeline (drive5.com/uparse/; Edgar 2011). Sequences from each OTU were then subjected to taxonomic classification using the latest version of the Greengenes taxonomy database (gg_13_5) (Wang et al. 2007). Based on the taxonomy information, any sequences associated with chloroplasts (from plant origin, and thus, most likely from feed) were filtered and discarded. In addition, representative OTU sequences were aligned to the bacterial 16S rRNA gene using the RDP aligner tool available at Michigan State University (https://rdp.cme.msu.edu/tu torials/aligner/RDPtutorial_ALIGNER.html); and those sequences that did not align with the sequenced region were filtered, thus removing OTU sequences that did not align within the expected region.
The OTU table was rarefied across samples to the lowest sample depth (6295 reads) using QIIME. All statistical analyses were performed with samples at an even depth. Furthermore, beta-diversity plots were generated in QIIME to evaluate differences based on sequence similarities and these plots were visualized with the Emperor visualization programme (Vazquez-Baeza et al. 2013). Moreover, alpha diversity estimators (Chao1 and observed species) and diversity index (Shannon) were evaluated for the overall community using QIIME (Caporaso et al. 2010). Good's coverage test was performed to evaluate if adequate sampling depth was achieved. A Venn diagram was constructed to illustrate the relationship and OTU distribution among treatments. To do so, the venn function in the gplots package of R was used (Warnes et al. 2015).
A core microbiome, which was defined as those OTUs present in all animals, was calculated. From the core microbiome, taxa were summarized and plots were generated at the taxonomic levels of phylum, class, order, family and genus, with emphasis on representative OTUs that represented at least 0Á1% of the microbial community in each sample. To minimize animal to animal variation and to represent the shared OTUs within each diet, the core microbiome was analysed. This allows identification of the microbial community influenced by the treatment sorting through animal to animal variation. The hypothesis is that, if the treatment affects the microbial community, this effect should be present across multiple animals on the same treatment. Therefore, the analysis of the core microbiome allows identification of the effects that might otherwise be hidden in the data (Benson et al. 2010;Castillo-Lopez et al. 2014. Compared to other sequencing platforms such as 454 Roche pyrosequencing (Castillo-Lopez et al. 2014) or Illumina (Castillo-Lopez et al. 2017), the use of ion torrent may have some limitations due to sequencing errors (Frey et al. 2014;Salipante et al. 2014). However, the quality control steps outlined by previous researchers such as initial quality control using the ion Torrent Suite Software, screening for chimeric sequences and preclustering using the pseudo-single linkage-clustering algorithm were specifically aimed at removing erroneous reads. Moreover, the core microbiota analysis where only OTUs that were present in all animals were considered should also filter out many OTUs generated due to random sequencing errors as these would not be expected to occur in all animals.

Statistical analysis
Data collected on the abundance of bacterial phyla, families and genera, and sample bacterial richness (Chao1 and observed species) as well as diversity index (Shannon) were analysed using the MIXED procedure of SAS (ver. 9.1; SAS Institute Inc., Cary, NC). Fixed effects included the treatment and period, with cow as the random effect. The statistical model for this experiment was as follows: where Y ijk represents observation ijk, l represents the overall mean, b i represents the random effect of cow i, q j represents the fixed effect of period j and a k represents the fixed effect of treatment k. The residual term e ijk was assumed to be normally, independently and identically distributed, with variance r e 2 . The comparison of treatment means was conducted using the PDIFF option in the LSMEANS statement. In addition, using the CON-TRAST statement, CONT was compared to FLAX+EXT +EXTT, and CONT was compared to each of the other treatments. Furthermore, FLAX was compared to EXT, and EXT was compared to EXTT. Treatment means are presented as least squares means. The largest standard error of the mean (SEM) is reported. Statistical significance was declared when P ≤ 0Á05 and tendencies were discussed when P > 0Á05 and ≤0Á10. The Spearman's correlation analysis was conducted to evaluate associations between dietary composition (content of fat, unsaturated fatty acids and tannins) and bacterial families and genera.
In addition, bacterial community composition differences were evaluated using the weighted UniFrac distance matrix as an input for a permutational multivariate analysis of variance (PERMANOVA) in R using the vegan package (adonis function) (Oksanen et al. 2015), where treatment was used as main effect.

Diets, milk yield and composition
The inclusion of the flaxseed-based products in diets resulted in significant changes (P < 0Á05) in chemical composition; the per cent of unsaturated fatty acids increased by approximately 4%; replacement of field peas with high tannin faba beans resulted in a content of condensed tannins of 1Á17 mg g À1 in EXTT. It is important to note that the level of condensed tannins decreased by 83% after extrusion (from 6Á87 to 1Á17 mg g À1 ).
Condensed tannins were not detected in CONT, FLAX or EXT, as was anticipated based on the ingredient composition of the flaxseed supplements. Production performance data and milk fatty acid composition have been reported (Moats et al. 2015). Briefly, there was a significant decrease (P < 0Á05) in DMI when EXT was fed compared to CONT (25Á9 and 23Á4 kg, respectively); fat corrected milk, however, was not affected and averaged 40Á5 kg. In addition, significant changes (P < 0Á05) were observed in milk fatty acid profile; for example, when the flaxseedbased products were fed, the proportion of total saturated fatty acids decreased by 10Á8%, the proportion of total polyunsaturated fatty acids increased by 0Á60% and the concentration of Omega-3 increased by 0Á51%. Nonetheless, no significant effects of treatment were observed on ruminal pH (6Á03), ruminal digestibility of dry matter (38Á4%), organic matter (40Á7%) and neutral detergent fibre (35Á7%).

Number of sequences, sample richness, diversity index and OTU distribution
Collectively, a total of 356 709 high-quality DNA sequences were obtained after initial quality control and filtering, and were used for downstream analysis. Diversity metric Chao1 was not affected by the inclusion of the flaxseed-based products (P ≥ 0Á38); observed species (P ≥ 0Á34) and the Shannon diversity index (P ≥ 0Á23) remained unaffected as well (Table S1). The Good's coverage test showed that sequencing depth was able to characterize >98% of the bacterial community.
The Venn diagram for OTU distribution revealed that each treatment showed a number of unique OTUs. However, there were 1285 OTUs shared by the four diets, which represented 69Á16% of total OTUs detected. In addition, according to the beta diversity for sequence similarities (Fig. S1) based on principal coordinate analysis, there appear to be two clusters and two sample outliers; however, no apparent clustering of microbial communities by treatment was found, indicating a similar spatial sample heterogeneity among the diets. The bacterial community analysis using PERMANOVA did not display a significant (P = 0Á20) effect on bacterial community composition.

Correlation coefficients
Correlation analysis between dietary components (content of fat, unsaturated fatty acids and tannins) and bacterial families and genera (Table 6) revealed that the content of dietary fat tended (P ≤ 0Á09) to be negatively correlated with the abundance of the families BS11, Christensenellaceae and Clostridiaceae, and was negatively correlated (P < 0Á05) with the genus Coprococcus, but was positively correlated with unclassified Veillonellaceae (P = 0Á02).
Dietary unsaturated fatty acid content tended (P = 0Á07) to be negatively correlated with the family Clostridiaceae, but was positively correlated (P ≤ 0Á04) with the genera unclassified Veillonellaceae and Schwartzia. Dietary tannin content tended (P ≤ 0Á08) to be negatively correlated with the families BS11, Paraprevotellaceae and Christensenellaceae, but was positively correlated (P < 0Á01) with WCHB1-25; in addition, tannin content tended to be negatively correlated (P = 0Á09) with the genus Prevotella, but was positively correlated (P ≤ 0Á02) with Oscillospira and Bulleidia.

Discussion
The gut microbial population influences physiology, metabolism, nutrition and immune function with disruption of this community being linked to gastrointestinal conditions (Guinane and Cotter 2013;Ridaura et al. 2013). In ruminants, gut microbes represent a source of metabolizable protein (Spicer et al. 1986;NRC 2000), they play an essential role in volatile fatty acid *CONT: a normal diet including barley silage, alfalfa hay and a barley-based concentrate with no flaxseed or faba beans; FLAX: inclusion of 11Á4% of a nonextruded flaxseed-based product containing flaxseed, field peas and alfalfa; EXT: similar to FLAX, but the product was extruded; EXTT: similar to FLAX, but product was extruded and field peas were replaced by high-tannin faba beans. †The largest standard error of the mean is reported. ‡1: CONT vs FLAX+EXT+EXTT; 2: CONT vs FLAX; 3: CONT vs EXT; 4: CONT vs EXTT; 5: FLAX vs EXT; 6: EXT vs EXTT.
production and feed digestion (McAllister et al. 1994) and milk composition (Jami et al. 2014). Moreover, the bacterial community is responsible for fatty acid biohydrogenation (Jenkins et al. 2008). Therefore, to effectively develop feeding strategies to enhance production performance or quality of dairy products, researchers must understand the effects of diet composition or biohydrogenation mitigating strategies on the broad ruminal microbiome in vivo. Given the laborious nature of studies involving the evaluation of ruminal fermentation and the rumen microbial community, some of the experiments have been conducted using small number of animals (Lillis et al. 2011;Boots et al. 2013;Mohammed et al. 2014;Denman et al. 2015) through the Latin square design, which is commonly used in cattle nutrition studies (Lillis et al. 2011;Boots et al. 2013), mostly because it is efficient as it generates replication with limited experimental units. However, it should be noted that the potential for carryover effects is one limitation of the design. Especially when evaluating the effects of plant secondary compounds on the microbial community because of their effect on animal physiology and metabolism (Dearing et al. 2005). In addition, once ruminal micro-organisms *CONT: a normal diet including barley silage, alfalfa hay and a barley-based concentrate with no flaxseed or faba beans; FLAX: inclusion of 11Á4% of a nonextruded flaxseed-based product containing flaxseed, field peas and alfalfa; EXT: similar to FLAX, but the product was extruded; EXTT: similar to FLAX, but product was extruded and field peas were replaced by high-tannin faba beans. †The largest standard error of the mean is reported. ‡1: CONT vs FLAX+EXT+EXTT; 2: CONT vs FLAX; 3: CONT vs EXT; 4: CONT vs EXTT; 5: FLAX vs EXT; 6: EXT vs EXTT.
have been treated with adverse plant dietary products, they may be no longer na€ ıve and their reaction may be damped on successive treatments (Kohl and Dearing 2016). In this study, the experiment was designed with 28-day periods as an attempt to minimize potential carryover effects, with the first 26 days serving as a 'washout' period and the final 2 days serving for collection of ruminal digesta for bacterial community analysis.
Increasing the size of the study would improve the experiential precision (Stroup 1999;Kononoff and Hanford 2006). However, despite being relatively small, the use of the Latin square experimental design in this study allowed detection of important and statistically significant differences in rumen fermentation and microbial analysis, as in previous reports using the same design (Lillis et al. 2011;Boots et al. 2013;Ramirez Ramirez et al. 2016a,b).
Feeding flaxseed or flaxseed-based products to dairy cows and effects on the overall rumen bacterial community Feeding flaxseed has been shown to improve milk fatty acid profile without affecting milk production (Oeffner et al. 2013). Current advances in the dairy feeding industry is spurring development of new flaxseed-based feed ingredients to enhance milk fatty acid profile; and the effects of those products on the broad rumen bacterial community of dairy cows must be clearly elucidated. Regardless of treatment, predominant ruminal bacteria agree with previous reports (Petri et al. 2012) showing that major bacterial phyla in the rumen of cattle are Bacteroidetes, Firmicutes and Proteobacteria. Collectively, these phyla represented approximately 96% of the rumen bacterial community in the current study. The influence of diet on the diversity and community composition of ruminal contents has long been recognized (Tajima et al. 2001;Fernando et al. 2010). In this experiment, when compared to the normal diet, the inclusion of raw flaxseed or extruded flaxseed with or without high-tannin faba beans did not cause drastic shifts on the abundance of major ruminal bacterial phyla. Contrasting these observations, Kong et al. (2010) used quantitative fluorescence in situ hybridization and found that inclusion of flaxseed reduced the total abundance of the phyla Bacteroidetes, Firmicutes and Proteobacteria in the rumen of cows fed silage-based diets. The discrepancies between these observations may be related to the chemical composition of the diets and the available substrates for microbial growth. In our study, the main dietary changes involved the increase in ether extract and polyunsaturated fatty acids in diets with flaxseed inclusion; in addition, there was a change in physical processing of flaxseed among diets, with a constant forage base. Thus, substrate availability for bacterial fermentation was relatively similar across treatments suggesting that bacterial phyla and other major taxa distribution may be resilient to changes in physical form of feeds when dietary fat does not exceed 6% in the diet.
Taxonomic analyses at the family and genus levels agree with previous findings using DNA pyrosequencing indicating that the ruminal microbiome is largely composed of the bacterial families Prevotellaceae, Lachnospiraceae and Ruminococcaceae and the genera Prevotella, Succiniclasticum and Ruminococcus (Castillo-Lopez et al. 2014). Prevotella, the largest bacterial genus detected, is composed of versatile organisms that can utilize a variety of nutrients including protein, starch, pectins and hemicellulose (Russell 2002), and has also been reported to predominate in the rumen of cattle being fed forage-based diets supplemented with corn distillers grains (Ramirez Ramirez et al. 2016a,b). In agreement with Kong et al. (2010), this experiment indicated that the fibre digesting genus Fibrobacter accounted for a minor fraction of the bacterial communities across diets. Overall, inclusion of flaxseed-based products did not affect predominant bacterial families and genera, and only affected taxa found in lower proportions, which could partially explain the lack of negative effects on *CONT: a normal diet including barley silage, alfalfa hay and a barley-based concentrate with no flaxseed or faba beans; FLAX: inclusion of 11Á4% of a nonextruded flaxseed-based product containing flaxseed, field peas and alfalfa; EXT: similar to FLAX, but the product was extruded; EXTT: similar to FLAX, but product was extruded and field peas were replaced by high-tannin faba beans.
ruminal digestibility of DM, organic matter and neutral detergent fibre.

Effect of unsaturated fatty acids, extrusion and dietary tannins on rumen bacterial community structure and function
Dietary strategies to improve fatty acid profile of ruminant products have included feeding unsaturated fatty acids, feed extrusion or supplementation with tannins. Although effects on some bacterial taxa (Vasta et al. 2010;Enjalbert et al. 2017) or production performance has been acknowledged, the impact of such strategies on the broad ruminal bacterial communities of dairy cows in vivo is yet to be clearly elucidated.
Ruminal bacteria, specially fibrolytic bacteria, may be negatively affected by dietary fat (Maia et al. 2006;Enjalbert et al. 2017). For example, Maia et al. (2010) reported that unsaturated fatty acids decreased the abundance of Butyrivibrio fibrisolvens in vitro. In addition, negative effects of dietary linseed oil have been reported on the genera Fibrobacter, Prevotella and Ruminococcus (Huws et al. 2014;Enjalbert et al. 2017). In the present experiment, the 4% increment in dietary unsaturated fatty acids was not accompanied by a decrease in the abundance of these genera. This may indicate that the increase in dietary fat and unsaturated fatty acids was not severe enough to exert negatively impacts on those taxa; in agreement with this observation, no negative effects were detected on ruminal fibre digestion across treatments (Moats et al. 2015).
Feed extrusion has been applied to decrease fatty acid saturation because heat denatures the protein matrix surrounding the fat droplets, consequently reducing the access of ruminal bacteria to dietary fat (Kennelly 1996). Moreover, Vasta et al. (2007) and Schofield et al. (2001) suggested that tannins may inhibit the activity of biohydrogenating bacteria because tannins can interfere with bacterial growth. Within the bacterial population residing in the rumen, a number of bacteria that participate in fatty acid saturation have been identified, which include bacteria belonging to the genera Pseudobutyrivibrio  et al. 1993) and Lactobacillus (Jenkins et al. 2008;Sakurama et al. 2014). In the present experiment, the abundance of the genera Selenomonas and Butyrivibrio were similar across treatments. However, Buccioni et al. (2014) and Vasta et al. (2010) reported an increase in B. fibrisolvens and a decrease in Butyrivibrio proteoclasticus in the rumen of sheep supplemented with quebracho tannins. This suggests that biohydrogenating bacterial species within the same genus show different degrees of sensitivity to dietary tannins (Nelson et al. 1997;Schofield et al. 2001). In this study, we did not evaluate bacterial species; however, it is possible that the lack of an effect of EXTT on most bacterial taxa may have been due to the lower tannin concentration compared to Vasta et al. (2010) who utilized quebracho-supplemented diets containing 6Á4% tannins. It is important to note that the content of tannins in the extruded product containing high-tannin faba beans was lower than expected, the 83% decrease in tannin content of the extruded product may have been caused by high temperature during the extrusion process (Iram et al. 2014). Thus, it may be beneficial to evaluate extrusion techniques to minimize the loss of tannins in supplements designed for ruminant diets.
A negative correlation does not necessarily indicate a direct cause-effect relationship; however, the negative association found between some bacterial taxa and the content of dietary fat, unsaturated fatty acids or tannins may indicate high sensitivity to these dietary components. For example, the negative correlation between the abundance BS11 and dietary fat may be due to toxic effects of fat on members of this bacterial family (van Lingen et al. 2017). Likewise, high sensitivity to tannins has been reported for Prevotella belonging to Paraprevotellaceae (Li et al. 2015).
Interestingly, when feeding extruded flaxseed to dairy cattle, the content of a-linolenic acid in blood serum and milk tended to increase (Kennelly 1996;Oeffner et al. 2013), and when cows consumed treatments containing the flaxseed-based products utilized in this experiment there was an increase in the concentration of omega-3 and total polyunsaturated fatty acids in omasal digesta and in milk (Moats et al. 2015). Bacterial species were not evaluated in this study; thus, we are unsure whether biohydrogenating micro-organisms were negatively affected. The family Christensenellaceae, which decreased with flaxseed inclusion, has been recently associated with low body mass index and reduced adiposity gain in nonruminants (Goodrich et al. 2014). When cows consumed the extruded flaxseed-based product, body weight was not affected (Moats et al. 2015). Further investigation elucidating the activity and role of members of this bacterial family in the rumen and how they may impact production performance or fatty tissue accretion in dairy cattle is warranted.
In this experiment, the microbial profile of diets fed was not determined. Reports have shown that bacteria found in the diet could potentially affect ruminal microorganisms (Ghorbani et al. 2002;Lettat et al. 2010), others have reported that the survival of some of these bacteria in the rumen is variable (Jeyanathan et al. 2016). More recently, Philippeau et al. (2017) reported that direct-fed microbials did not affect ruminal micro-organisms or volatile fatty acid production. Therefore, it was not possible to determine associations between microorganisms found in the diets, if any, and changes in the rumen microbial profile.
Overall, findings from this study indicate that flaxseedbased products tested were effective for replacing barleybased concentrate in lactating dairy rations without negative effects on predominant rumen bacterial taxa. However, the content of unsaturated fatty acids and tannins in the diets were negatively associated with some bacterial taxa found in lower proportions in the rumen such as Clostridiaceae, BS11, Paraprevotellaceae and Christensenellaceae; nonetheless, production performance and ruminal nutrient digestion were unaffected. The use of high-throughput DNA sequencing contributes to unravel the impact of diet composition on ruminal micro-organisms, strengthening our knowledge not only on dietary intervention methods to mitigate fatty acid biohydrogenation and improve milk quality, but also to prevent negative consequences on the ruminal bacterial population, feed digestion and rumen function.

Supporting Information
Additional Supporting Information may be found in the online version of this article: Figure S1 Beta diversity for bacterial communities in ruminal digesta samples for treatments CONT, a normal diet including barley silage, alfalfa hay and a barley-based concentrate with no flaxseed or faba beans; FLAX, inclusion of 11Á4% of a nonextruded flaxseed-based product containing flaxseed, field peas and alfalfa; EXT, similar to FLAX, but the product was extruded; EXTT, similar to FLAX, but product was extruded and field peas were replaced by high-tannin faba beans.
Table S1 Effect of partially replacing a barley-based concentrate with different flaxseed-based products on bacterial richness estimates and diversity index for ruminal digesta samples from lactating Holstein dairy cows.