Final Report Summary - RUMINOMICS (Connecting the animal genome, gastrointestinal microbiomes and nutrition to improve digestion efficiency and the environmental impacts of ruminant livestock production) Executive Summary:RuminOmics, so named to imply the application of –omics technologies in both the rumen and ruminant animal, involved 11 partners in a 4-year project 2012-2015. The project comprised one main study of phenotypic, genomic and metagenomic analyses in 1000 cows on-farm in the UK, Sweden, Finland and Italy, allied to several other studies of host-microbe interactions, nutrition and the development of tools, with the aim of lowering environmental emissions from dairy livestock. The 1000-cow study revealed the nature of the genetic background controlling methane and nitrogen emissions and milk fatty acid composition and the role of the ruminal microbiome in these outputs. Variations between farms were allowed for in data analysis. Unexpectedly, it was established beyond doubt that genetic selection of dairy cows for low methane emissions would also select for less efficient animals, thus making the case to breeders and farmers to aim for low-methane livestock more difficult. Selection for lower methane will select for lower ruminal digestibility. In contrast, higher-efficiency cows produced less methane per unit product, indicating that a strategy involving selection of high-efficiency ruminants needs to be followed rather than the reverse direction. Nutrition, rather than the ruminal microbiome, is the main driver of emissions, nevertheless identification of certain members of the community as being associated with low methane emissions indicated a route for future development. The higher incidence of Proteobacteria in low-CH4-emitting animals had been observed previously, but the high incidence of some anaerobic fungal species in low emitters was unexpected, as was different clusters of Prevotella being associated differently with high or low emitters. Metagenomic analysis of the highest and lowest CH4emitters showed differences in the abundance of many genes, several associated with methanogenic archaea. The abundance of certain genes in the metagenome might become useful as a selection tool. Changes in dietary protein content had only marginal effects the rumen microbial composition. Blood urea was demonstrated to have low predictive value for N retention. Forage quality was key in lowering concentrate requirement. Digesta transplants were carried out between cows and reindeer to investigate whether the host species had a dominant role in determining its own rumen microbial community. It emerged that archaea and fungi were affected little by the host, protozoa were entirely host-specific, whereas bacteria adapted slowly but incompletely to the new host. Twin cows and unrelated cows were investigated with the same aims, reasoning that animals with the same genetic background, against the same environmental background, should have more similar microbial communities than unrelated animals. No evidence of such an association was found. Thus, the host animal does not have control over the whole rumen microbiome but affects certain microbial taxa. Transcriptomics analysis of rumen wall papillae in the same experiments showed individual responses to digesta exchange, specifically correlated to functional differences between the ruminal digesta. Metaproteomics provided useful taxonomic and metabolic information about individual samples, but was unable to identify differences between high and low emitters. Milk fatty acid composition showed correlations with methane emissions and the microbiome, but was of low predictive value. A much more successful tool will stem from the finding that community analysis of oral samples is predictive of the corresponding ruminal community. This will enable large numbers of animals to be sampled for breeding purposes and will enhance animal welfare. The results of the project have been stored as a data warehouse and will become open access at EBI as papers are published by consortium members. Extensive dissemination activities were pursued during the course of the project, culminating in four highly successful stakeholder workshops held in Warsaw, Budapest, Lodi and Edinburgh, a training school for young researchers in Piacenza, two researcher workshops in Dublin and Aberdeen, and the final conference in Paris. Each was attended by 50-120 people. An open access ELearning course was established at http://www.ruminomics.eu/elearning.Project Context and Objectives:The concept of this project was that long-standing problems associated with ruminant livestock production could now be revisited using recent quantum-leap developments in –omics technologies and bioinformatics, with the aim of solving the environmental problems of methane emissions and low N retention. At the project beginning, advances in 16S rRNA gene and sequencing technology had made it possible to describe the community avoiding cultural bias, and to sequence all genes – the metagenome – in a microbial community. It was now possible to determine what happens to the whole spectrum of genes present in the microbial ecosystem and which organisms are present. Thus, if the analyses were accompanied by measurements of methane emissions, digestion efficiency, nitrogen emissions and product quality, it might be possible to link all of these to the ruminal microbiome.A key question of the project was – ‘How much does the host animal define its own gut microflora?’ Anecdotal observations that different animals retain a microbiome that differs from its neighbour could be examined rigorously using the new state-of-the-art technologies. If this could be proved, it might be possible to select against animals that encourage methane-producing or N- wasting microbes in the rumen. The consequence of such selection would also be beneficial to the ecological footprint of ruminant livestock production. Advances in microarrays meant that an experiment of the scale originally proposed – 1000 cows –should be able to identify specific genetic loci associated with emissions and efficiency, and most importantly in the long term the key features of the ruminal microbiome that define these production characteristics – a systems biology approach. Smaller experiments proposed in Ruminomics involved the metagenomic analysis of monozygotic twins and a comparison between cows and reindeer receiving the same diet. Similar analyses have been carried out with humans and their intestinal microbiota. In Ruminomics, the ‘experimental animal’ is much more amenable, and digesta-exchange will be carried out to tease out important host-microbe interactions.Although advances in the –omics enable researchers to pursue ever more sophisticated analyses, there is a need to translate that knowledge into tools that can be useful in a more widely useful analysis. The concept in Ruminomics was to use –omics to help develop new tools. The final component of the concept relate to dissemination and uptake of knowledge. The –omics technologies have been moving at bewildering speed, particularly next-generation sequencing. This consortium intended to address the need to keep industry abreast of developments, also academics in related subjects that could benefit from updating of knowledge. To this end, industry-focused workshops and a summer school were planned to address the issue, focussing on industry, particularly SMEs, and academia from countries in the enlarged EU, candidate countries and developing countries, where efficiency and environmental issues are as important as in other countries but receive less attention.Applicability to meat animals would be monitored by the European Association for Animal Production (EAAP, partner 9) and Quality Meat Scotland (QMS, consultant), the red meat levy board for Scotland, who have been involved with the project from its inception.Objective 1: Relating animal genome to microbiome, feed efficiency, and methane emissionsThis objective fulfilled the principal requirement of the call to generate information on genetic variation in the animal (‘livestock genome sequences, in particular that of cattle’) and its interaction with ‘GIT function and feed efficiency’ while providing corresponding information on the ‘diversity and efficiency of microorganisms in the GIT’. The ‘production and emission of greenhouse gases (notably CH4)’ is a measurement that would be made across WP3-5 and Objectives 1-3, but was one that was especially pertinent to this Objective, because a relation between host genome and methane production, via the microbiome, would enable breeders to relate specific host loci to methane emissions, key members of the ruminal microbiota and feed efficiency – ‘defining key indicator traits’. Methane emissions were to be measured, and samples of blood, milk, ruminal digesta, and faeces collected from up to 1000 dairy cows, 800 for finding associations, 200 for validation. Subsequent animal genome and metagenome analyses were to enable the association of marker genotypes to be compared with production efficiency, methane emissions and the microbiome. This information would form the basis for marker assisted and genomic selection programmes in animal breeding. Objective 2: Determining host-microbe interactions in genetically identical and genetically diverse animalsSamples from identical twins and non-identical cows on the same diet were to be used to compare ruminal microbiomes against identical and non-identical host genetics, providing a different kind of information on ‘host-microbe interactions’ from Objective 1. Digesta were to be exchanged to determine if the microbiome reverts to its original composition. Cows and reindeer were to be sampled and their digesta analysed at a greater level of detail by carrying out metagenomic sequencing. Digesta-exchange was to be part of this experiment as well, along with genotypes. Because methane and digestibility measurements would be made on all these animals, fundamental questions on the digestion efficiency of reindeer and cattle, and thereby ‘’nutritional efficiency and environmental footprint’ could be matched with the complete catalogue of the microbes that are present in the rumen and their gene complement.Objective 3: Emissions, the microbiome, nutrition and healthVariations in the three most abundant macro-nutrients, carbohydrate, protein and lipid, were to be made in order to relate changes in the nutrient supply of the cow with the composition and function of the ruminal microbiome, as assessed by methane and N emissions. This fulfilled the call for ‘the interplay (between) the efficiency of feed utilisation (energy and proteins), animal metabolism, product quality and the production of greenhouse gases’.Carbohydrate nutrition. The efficiency of utilisation of forages is key to successful and profitable dairy production. This project intended to analyse samples from cows on diets containing differing predominant carbohydrates for their ruminal microbiomes and products/emissions. Grass silage was to be the main form of forage fibre. Themes in this sub-objective also included the following: comparison of glycosyl hydrolase genes (polysaccharide-degrading enzymes) in metagenomes from reindeer and cattle – a novel comparison that might reveal genes in reindeer that explain their better use of plant fibre. The enzymes may have application in a number of other industries, including biofuel production from high-fibre substrates; sequencing genomes of anaerobic fungi – why are these organisms so much better than bacteria at digesting recalcitrant fibre?; six different Butyrivibrio spp., representatives across the broad phylogenetic grouping of Butyrivibrio, will be sequenced and interrogated for the genetic basis for different properties of fibre digestion. Then, only a human B. fibrisolvens genome sequence was available in public databases.Protein nutrition. Extreme commercial-type diets in terms of overall protein content and availability of nitrogen in the rumen was to be fed to dairy cows in order to determine the effects of protein nutrition on the ruminal microbiota. Samples were to be analysed for the total microbiome, but special attention would be paid to sub-populations, namely the ruminal protozoa and ‘hyper-ammonia-producing bacteria’ (HAB) species, linking these to N retention and feed efficiency. Themes in this sub-objective also included the following: three significant HAB bacteria, Peptostreptococcus anaerobius, Eubacterium pyruvativorans and Clostridium aminophilum, were to be sequenced and explanations sought for their non-saccharolytic, amino acid-fermenting, metabolic behaviour. [The sequence of C. aminophilum was published early in the project, so this was not done]; expression and analysis of peptidase genes of Prevotella ruminicola. Only one peptidase of Prevotella, dipeptidyl peptidase 4 (DPP-4), had been properly characterised. DPP-4 is not the main peptidase activity of these bacteria. The gene of the main one, DPP-1, had not been identified.Lipid nutrition. The combined effects of certain fatty acids on methane formation and milk composition were to be analysed in samples of ruminal digesta, faeces and milk, with accompanying methane and microbiome measurements. The microbiome will be analysed from the point of view of the whole microbiome, but specifically the archaea and Butyrivibrio spp., the latter which was thought to be the main species involved in biohydrogenation of fatty acids. A systems analysis approach was to be used to combine these measurements into a form that could be predictive in nature, relating nutrition, emissions and the ruminal microbiome, also to product quality in terms of the fatty acid composition of milk. Themes in this sub-objective also included the following:-Six different Butyrivibrio spp. genome sequences were also to be interrogated for the genetic basis for different properties of fatty acid metabolism. Only B. proteoclasticus is known to perform the last, vital step of fatty acid biohydrogenation, namely the conversion of vaccenic acid (trans-10-18:1) to stearic acid, but it is not clear why. Controlling the activity of this species would transform the fatty acid composition of milk, and therefore its perceived nutrition/health quality.Two specific metabolic health problems arising from the dysfunction of the ruminal microbiota were to be monitored. Some of the cows in WP4 particularly were expected to suffer from chronic sub-acute acidosis – they will be analysed particularly closely for changes in their microbiome. The other health issue that may arise in WP4 is milk fat depression, most probably associated with a high oil content of the diet. The microbial aetiology of this condition is controversial. Examination of the microbiomes of these animals and comparison with healthy animals will help to explain how the disorders occur and will suggest methods to avoid them.Objective 4: Tools and bioinformatics for rapid analysis of phenotypes, microbiomesThe intended tools comprising this objective impinged on different aspects of the project. They aimed to translate the genetics/metagenomics/emissions observations to a simpler form that is useful to scientists and livestock breeders:-4(i) The meta-barcoding techniques developed in the project will be a tool that others interested in rapid characterization of the ruminal microbiome will adopt.4(ii) The totality of the information generated in the project will lead to selection tools for breeders, based on emissions, microbiomes and animal genotype.4(iii) Recent publications had suggested that the concentrations of minor fatty acids in milk can predict methane emissions. How well does this work over the wide range of animals and samples collected here? Confirmation that this is a valid tool would enable proxy methane estimations to be made relatively easily and cost effectively.4(iv) Can buccal or faecal samples be proxies for ruminal samples to describe the ruminal metagenome and its relation to efficiency and emissions? Samples of oral regurgitated ruminal digesta, ruminal digesta and faeces were to be compared for their microbiome. If either oral or faecal samples were to give a useful indirect indication of the ruminal microbiome, others carrying out research in this field would be able to conduct their experiments without invasive rumen sampling, either by stomach tube or cannulation.4 (v) Can metaproteomics be a predictive tool for feed efficiency, methane emissions and N retention? Metagenomics only describes the genes that are present. Metaproteomics provides an analysis of how those genes are expressed. A metaproteome had not been published for the rumen. This project intended to evaluate the ruminal metaproteome, identifying the main spots by LC-MS and link those empirically to production/emission measurements. Specific gene products would be sought. The mrcA enzyme is specific to archaea that produce methane. 4(vi) Some long-chain alcohols such as archaeol in faeces are derived entirely from methanogenic archaea in the rumen. How well does the measurement of these polar lipid compounds correlate with emissions? A simple faecal extraction and GC analysis would avoid the need to use any of the invasive or environmentally damaging (SF6) measurements of methane emissions from cows. Objective 5: Public rumen metagenomics databaseIn this project all the metagenomics data produced would be processed and stored along with all the information generated in the different project activities. A central informatics platform would be used to manage and to connect the project datasets and the different information. During the project all the partners would have access to the data through a secure FTP site hosted by PTP. The metagenomics next generation sequencing data of the rumen microbiome would then be correlated with all the information on the animals analysed by combining together the data produced in the project. The stored data will provide the required background information to perform the statistical models and find associations between the animal genome and the rumen microbiome composition. The project data, including sequencing and animal phenotypes information, would be converted, prepared for submission and uploaded to EBI public databases at the end of the project. This will allow the researchers from the scientific community a long-term access to the data generated by the project and, at the same time, they will be able to upload and submit new rumen meta-genomics data and extend the RuminOmics datasets in the same public resource. The project would promote the use of this public central resource for rumen microbiome data management and information exchange and will fulfil the call request for “including the development of bioinformatics resources/tools for possible use by scientists beyond the beneficiaries in the project”.Objective 6: Dissemination of project technologies and resultsThe sub-objectives of this important area of work would cover communication between partners, of partners with stakeholders, exchange of workers between partners, and training of partners and stakeholders. We would organize workshops that will enable busy stakeholders to understand the new technologies and their implications; create a participatory framework that would allow structured and continuous dialogue between partners and the stakeholders to ensure that the project meets the needs of the end users; organize a final conference to enhance the relevance of the research to end users and to agree future research, technology and implementation strategies; disseminate results to the intended end users and secure exploitation of innovations from the project through the provision of innovative training elements (E-learning), dissemination material and demonstration activities; create a core group of interested industries in animal breeding and feed, farming cooperatives, and processors; consult industries and cooperatives via individual company visits on the needs and perspectives with regard to RuminOmics; organize a technical visit of the interested industries to the ruminant research site of one of the participating universities; organize national and regional advisory and scientific workshops to achieve two way interaction between researchers and end users and to agree future research, technology and implementation strategies.Project Results:3 Main Scientific and Technical Results3.1: OverviewThe RuminOmics project was organised into ten work packages with one managing, seven scientific and two disseminating the results. In the first year of the project WP2 optimised and harmonised and the methods to minimise variation between countries/farms. Once in place the methods and SOPs were circulated. WPs 3-5 were responsible for the animal studies generating all the samples for the project. The molecular analysis was carried out by WP6 with the data generated passed back to WPs 4 and 5 and on to 7 and 8 for further processing and data analysis. 3.2 Workpackage 2 - Methods development and SOPs 3.2.1 Calibration of methods for the large scale measurement of feed intake and digestibility. This task aimed to adapt the marker technique developed for estimating feed intake of grazing animals (Mayes et al., 1986) based on the ratio between faecal excretion of external markers (even number n-alkanes C32 and C30) fed at known daily amount, with diet digestibility estimated by the increase of one or more internal markers (odd number n-alkanes) from feeds to faeces (Unal and Garnsworthy, 1999). Three trials were carried out at UNICATT on 6 dairy cows (housed in tie stall or free stall conditions) to compare possible techniques to administer the external marker and select the best time for faecal collection. The aim was to reduce the collection to one time per day and then check the method in simulated field conditions. In Trial 1 marker administration by oral drench (DRENCH) or as dry concentrate (CONC) and time of faeces collection (7:00; 14:00 and 19:00) were compared. CONC led to more consistent estimates (R2 >0.85) vs. DRENCH, with the best correlations obtained for sampling at 7:00 (R2 = 0.91 P<0.001). Trial 2.1 and 2.2 were carried out to re-test the results of T1 (CONC). Collecting faeces at morning feeding time produced again the best predictive equations (R2 = 0.88 for C31 and 0.83 for C32) as in trial 1. Trial 3 used the same protocol on 6 free stall housed dairy cows and collected faeces only at morning feeding time. High correlations were found between predicted and observed intake, but absolute estimated intake differed from measured intake in all the trials. To correct for this systematic bias, predicted DMI values were multiplied by the ratio between group mean predicted and observed DMI, which reduced the maximum difference between measured and predicted to between 6 and 8% of actual intake. At UNOTT a refinement of the alkane technique developed by UNICATT was conducted using the C30 alkane naturally present in high concentration in the concentrate fed in the robot milking station, which was not present in concentrates fed on the Italian farms. A study was conducted on 40 cows with direct measurement of feed intake and administering the C30 marker 2-3 times/day through electronic feeders. Faeces were collected once-daily between 07:00 and 09:00. Partial least squares regression (PLSR) was used to predict intake from daily intake of C30 in concentrates plus concentrations and ratios of C27, C28, C29, C30, C31 and C33 in faeces and feed. Intake predicted using this method was highly correlated with observed intake: r = 0.85; P <0.001; MSPE= 1.58 kg DMI/day or 6% of mean intake. No intake effect on accuracy and no systematic bias were detected. Results demonstrated potential for estimating feed intake from faecal and diet alkane concentrations using alkanes as external/internal markers and collecting faeces once a day. 3.2.2 Protocol for measurement of methane emissions.The already validated method developed at UNOTT for measuring methane emissions from individual cows during robotic milking (Garnsworthy et al. 2012) could not be adopted by UNICATT due to the unavailability of farms with robot milking where intake measurement by the alkane technique was also possible. UNICATT tested an original method based on the SF6 technique to collect breath samples that failed to produce reliable results. Consequently, two GreenFeed devices (C-Lock Inc., US; Huhtanen et al., 2015) for methane emission measurements in WP3 were purchased, installed and tested at the experimental farm of UNICATT. 3.2.3: Definition of the best method for rumen liquor sampling, preserving animal welfare at the highest possible level. UNICATT compared four rumen sampling techniques, rumen fistula (RuFi) (reference method), nasogastric tube (NaTu), orogastric tube (OrTu) and rumenocentesis (RuPu), on three dairy cows and at two sampling times (before feeding and 6 h later). Rumen pH, VFA (gas-chromatography) and microbial composition (RT-PCR and DGGE) were checked. NaTu did not consistently allow withdrawal of rumen fluid and was immediately removed from trial. OrTu and RuPu both recorded a higher pH (P<0.05) and lower total VFA compared with RuFi. VFA molar proportions were not significantly affected by sampling technique. Enumeration of protozoa population revealed no significant difference between the two sampling techniques. Qualitative analysis of Eubacteria population revealed no apparent variation due to sampling technique. Quantitative analysis of Archaea seemed to indicate some significant differences in samples collected by OrTu compared to RuFi and RuPu. qPCR showed a reduction in bacteria content with OrTu. Though RuPu produced data similar to the RuFi technique, RuPu was considered too invasive and not in agreement with national rules on animal welfare in some countries. OrTu represented a valuable alternative and was adopted as the standard method for rumen sample collection in WP3. 3.2.4: Development of optimized DNA 16S rRNA amplicon sequencing markers for characterization and quantification of rumen microflora.The first objective was to obtain primer pairs (1) specific to the target group (i.e. eubacteria, archaea, fungi, and protozoa), (2) uniformly amplifiable across this group, and (3) that generated high-resolution barcodes for good taxonomic identification within this group. The second objective was to determine the extraction and storage protocols most suited to community analysis using 16S rRNA amplicon sequencing. CNRS designed primer pairs in silico based on sequences available in GenBank; they were compared with primers described in literature and further evaluated in vitro. The in vitro results were also used to choose the best DNA extraction and conservation method. In vitro and in silico results coincided, but the α-diversity in bacteria was maximal for BactA in vitro whereas ProkA recovered the most sequences in silico, and the resolution power for ciliates was highest with CilA in vitro, but with CilC in silico. There was virtually no difference between the intracellular extraction and storage methods tested. In WP3 BactB were used for bacteria, ArchA for Archaea, FungA for fungi, CilA for ciliate, ProkA for the relative abundance of bacteria vs. archaea. 3.2.5: Preparation of SOP and Model Agreements for commercial farms and harmonization of experimental procedures among partners Eleven Standard Operating Procedures were written for feeds (SOP 1-2), faeces (SOP 3), milk (SOP 4), blood (SOP 6) and rumen (SOP 9) sampling and manipulation. Methods of analysis were described in SOP 7 and 11. SOP 8 (Measurement of methane emissions) and 10 (Indirect measurement of feed intake) have been previously cited. A Model of Agreement was formulated by UNICATT and PTP and adopted with minor adaptations by UNOTT. 3.3 Work Package 3 – Animal Phenotypes The aim of WP3 was to collect measurements and samples from animals on a large international scale. The WP would provide data to test the association of the microbiome with efficiency of feed utilization, methane production and product quality. The specific objectives were to • Obtain standardized measurements and samples from Holstein and Swedish/Finnish Red cattle • Measure feed efficiency and methane production by applying state-of-the-art techniques • Analyse feed, blood, faeces and milk Task 3.3.1 Experimental Design It was confirmed that we would aim for 1,000 cows. To allow for missing values or outlying data, the final number of cows sampled was 1,016, which comprised 407 Holstein cows in UK (UNOTT), 409 Holstein cows in Italy (UNICATT), 100 Swedish Red cows in Sweden (SLU) and 100 Finnish Red cows in Finland (LUKE). All cows were offered total mixed rations (TMR) which were standardised within partner sites. At UNOTT and UNICATT, TMR were based on maize silage, grass silage or grass hay, and concentrates; at SLU and LUKE, TMR were based on grass silage and concentrates. Data and samples were collected over a 7-day period for each cow that had received the standard TMR for at least 14 days. All cows were in established lactation (90-180 days in milk) when methane emissions are relatively stable.Cows were phenotyped for the following:• Dry matter intake (DMI), milk yield, milk fat, crude protein, urea-nitrogen and lactose concentrations, milk fatty acid composition, live weight and body condition score• Methane production, diet digestibility, rumen pH, volatile fatty acids and ammonia nitrogen concentrations and the rumen microbiome. • Plasma non-esterfied fatty acids, beta-hydroxybutyrate, albumin, cholesterol, urea, creatinine and haptoglobulin concentrations.Task 3.3.2 Logistics and implementationEthical approval was granted before sampling at each centre. SOPs were agreed in WP2, sampling commenced at all centres in 2013 and was completed in 2015. Methodology varied according to facilities available on different farms. Methane was measured using breath sampling either during milking (UNOTT) or during feeding (SLU and UNICATT) and respiration chambers at LUKE; feed intake was measured directly at SLU and LUKE, and in 15% of cows at UNOTT; for the remaining cows at UNOTT and all cows at UNICATT intake was estimated using alkanes in feed and faeces. Rumen fluid was sampled at all centres with a ruminal probe designed for cattle.Task 3.3.3 On farm sample and data collection Data were recorded in a secure shared database. This database was used also to record transfer of samples from one partner to the next in the processing chain. Methods for handling and processing samples of rumen fluid, blood, milk and faeces were established in WP2 and the same SOPs were followed at each centre. Task 3.3.4 Laboratory analysesLaboratory analyses of blood, milk, feeds, faeces and rumen fluid were completed using SOPs established in WP2. Rumen samples were freeze-dried and then sent to IAPG for DNA extraction followed by 16S rRNA amplicon sequencing by CNRS. Blood samples were sent to PTP for genotyping. Milk lipids were extracted and transesterified to fatty acid methyl esters, which were then sent to LUKE for gas chromatography to determine milk fatty acid profiles.All results were checked first by the centre where they were generated, and then sent to UNOTT for collation, further checking and statistical analysis. ResultsAll phenotypes were within normal ranges encountered in previous studies and in the literature (Table 1). Methane emissions varied between farms (Figure 1), so a random model was applied in statistical analysis, which adjusted for Month of Sampling within Farm. Figure 1. Methane production (g/cow/day) at seven farms.Methane production (g/d) was positively correlated with DMI, live weight, milk yield, milk butterfat, urea nitrogen and saturated fatty acids (SFA), and rumen proportions of acetate, butyrate and iso-butyrate. Methane production was negatively correlated with polyunsaturated fatty acids (PUFA), and with rumen propionate and valerate.Variables correlated with methane production were entered into linear mixed models. Due to auto-correlations among many variables, dimensionality reduction was conducted by dropping terms individually that did not contribute significantly to the model. Dry matter intake (kg/d) accounted for 6.1% of methane variation, live weight (kg) 5.1%, milk SFA yield (kg/d) 10.0%, milk PUFA yield (kg/d) 2.4% and rumen propionate concentration (mmol/mol) 7.4%. Propionate concentration was negatively related to methane production, whereas all other variables were positively related to methane production.To investigate propionate variation further, factors that may drive propionate production were tested. Only microbial terms were retained in the model. The proportion of bacteria within the phylum Proteobacteria explained 34% of variation in rumen propionate, followed by the fungi genus Neocallimastix (14.2%) and bacteria phylum Bacteroidetes (13.1%), fungi genus Orpinomyces (6.1%), bacteria phylum Fibrobacteres (5.9%) and bacteria phylum Firmicutes (2.2%). Propionate was positively related to Proteobacteria and Neocallimastix, and negatively related to the other microbial terms in the model. Methane emission per kg of DMI was also examined with linear mixed models. Milk yield explained 23.3% of variation, rumen acetate 5%, milk PUFA 4.4%, and the fungi genus Neocallimastix 4.4%. Methane was positively related to milk yield and rumen acetate, and negatively related to milk PUFA and Neocallimastix.Table 1. Descriptive statistics for main phenotypes in 1016-cow sample population of cows mean sd min maxDry matter intake (kg/d) 22.8 4.5 13.0 41.8Milk yield (kg/d) 33.7 8.5 9.0 64.6Milk composition (g/kg): Butterfat 37.9 7.1 12.2 66.8 Crude Protein 33.0 2.9 25.9 47.6 Lactose 47.5 2.5 35.7 53.6 Urea-N (mg/100 ml) 29.5 7.7 9.7 54.0Milk fatty acids (% total FA): Saturated 69.5 3.2 56.1 77.3 Monounsaturated 26.4 3.0 18.4 39.4 Polyunsaturated 3.8 0.6 2.4 6.5Live weight (kg) 673 78.2 469 900Methane (g/d) 439 70.1 183 708Dry matter digestibility (g/kg) 703 53.4 525 861Rumen pH 6.6 0.40 5.4 8.0Rumen volatile fatty acids (mmol/mol): Acetate 598 49.7 448 735 Propionate 227 46.1 130 365 Butyrate 131 29.8 25 248 Iso-butyrate 9.2 3.3 3.0 28.0 Valerate 15.6 4.1 7.0 41.0 Iso-valerate 12.8 4.6 1.0 32.0Rumen Ammonia (mmol/l) 6.16 3.06 0.26 17.6Plasma: non-esterfied fatty acids (mmol/l) 0.12 0.08 0.02 0.68 beta-hydroxybutyrate (mmol/l) 0.82 0.38 0.15 4.03 albumin (g/l) 36.2 2.34 26.1 46.1 cholesterol (mmol/l) 6.47 1.36 2.16 10.7 urea (mmol/l) 5.51 1.26 0.02 11.1 creatinine (mmol/l) 85.6 10.8 46.2 140 haptoglobulin (g/l) 0.19 0.23 0.01 1.48Dry matter digestibility was negatively correlated with DMI, milk lactose concentration, yield of PUFA and trans fatty acids, and rumen propionate concentration; digestibility was positively correlated with methane per kg intake, live weight and milk urea-N, and rumen acetate, iso-butyrate, iso-valerate and ammonia.The most important correlation, in the context of the RuminOmics project, is the positive correlation between digestibility and methane production per unit of feed intake. This emphasises that cows with low methane emissions could be those that digest feed less efficiently than high emitters.The positive correlations between digestibility, acetate and milk fat emphasise the importance of forage digestion to overall digestibility; greater forage digestibility yields more rumen acetate, which increases precursors available for milk fat synthesis.Feed EfficiencyFeed conversion efficiency was calculated as both kg ECM per kg DMI and kg milk solids per kg DMI, and then examined for any relationship with methane production. There was no correlation between feed conversion efficiency and methane production per day (r= -0.004). Both measures of feed conversion efficiency were highly correlated with milk yield (positive) and dry matter intake (negative), as expected because these variables are used to calculate feed efficiency. Consequently, methane production per unit of feed intake was positively correlated with feed conversion efficiency on a milk solids basis (r= 0.339; P<0.001) and on an ECM basis (r= 0.370; P<0.001). This provides further support for the hypothesis that more efficient cows produce more methane per unit of feed intake.The implication of this finding is that selecting cows for low methane emissions could result in lower milk production efficiency. This would have an adverse effect on methane emissions at the herd and national levels because more cows and more total feed would be required per unit of milk produced. There are, however, opportunities to select cows which have lower methane emissions and higher efficiency than average; such cows are in the lower right quadrant of Figure 2.Figure 2. Relationship between milk production efficiency (kg energy-corrected milk yield per kg dry matter intake) and methane production (g methane per kg dry matter intake).Conclusions and implicationsThe findings demonstrate that relationships between methane emissions, digestion and efficiency are complex. Because methane represents wasted energy, it is commonly assumed that lowering methane emissions would divert more energy towards milk production, leading to improved efficiency. Results from analyses presented here do not support that assumption: cows with lower methane emissions are not more efficient. In fact, more efficient cows produce more methane per unit of feed intake. This is because more efficient cows digest feed to a greater extent, thereby generating more rumen acetate and hence methane.There are, however, opportunities to select cows which have both lower methane emissions and higher feed efficiency.There is evidence that methane emissions and feed efficiency are related to rumen microorganisms. These relationships are being explored further in other work packages.3.4: Host-microbiome interactionTask 3.4.1 Cow-reindeer rumen content exchangeRuminant livestock are capable of converting fibrous feed resources into high quality human foods. A major concern is that digestive processes results in the production of methane and excretion of nitrogen into the environment which contribute to greenhouse gas emissions. Different groups of rumen microbes have various roles in fibre digestion, starch fermentation and protein metabolism in ruminant animals. Much of what is known about these organisms is based on studies of specific strains that can be cultured ex-vivo. The advent of sequencing technologies has provided new tools to better understand the diversity of microbial communities in the rumen, and therefore their involvement in methane formation, extent of fibre digestion and the efficiency of nitrogen retention. Recent reports have demonstrated that diet, host species and geography influence the microbial community composition in ruminants, with diet thought to be the most important factor. However, what is much less clear is whether the host animal defines its own gut microflora. In mice, an association between host genetics and the intestinal microbiome has been reported, but it is not known if this also holds true for ruminants. To address this lack of knowledge an experiment involving the exchange of ruminal digesta from cows into reindeer was performed. The dairy cow (Bos taurus) and semi-domesticated reindeer (Rangifer tarandus tarandus) were chosen as animal models due to distinctive differences in anatomy and digestive physiology. Cows and reindeer were fed the same diet to meet maintenance energy requirements over a 15 week period. After 6 weeks a complete rumen evacuation was performed. The same amount of digesta removed from the rumen of reindeer was replaced with digesta recovered from cows. Changes in the rumen microbiome of reindeer were monitored for the next 9 weeks. Before digesta exchange and 5 weeks afterwards measurements of rumen fermentation, methane production, whole tract digestibility coefficients and nitrogen balance were also made. Whole tract nutrient digestion did not differ between reindeer and cows before or after digesta exchange. Compared with cows, reindeer excreted a higher proportion of dietary nitrogen in faeces and a lower proportion in urine. Transfer of digesta from the cow into reindeer did not influence these parameters. Rumen pH and the concentration of volatile fatty acids (VFA) and ammonia nitrogen (AM-N) did not differ between reindeer and cows. Concentrations of VFA and AM-N were unaffected by rumen evacuation or the exchange of rumen contents. Even though both ruminant species were offered a similar diet throughout the experiment there were distinct differences between cows and reindeer in the relative abundance of fermentation acids. Molar proportions of acetate were higher and those of propionate, butyrate, isobutyrate, valerate, isovalerate and caproate were lower in reindeer compared with cows. Rumen evacuation in the cow had no effect on rumen VFA proportions. In reindeer, the transfer of digesta from cows was accompanied by a decrease in acetate and an increase in butyrate, isobutyrate and valerate. Nevertheless, the rumen fermentation pattern in reindeer following digesta exchange was atypical of both species. Methane production was higher when expressed as mg/kg live weight but lower when expressed as a function of DM intake for reindeer than cows fed the same diet. Following the transfer of rumen contents, reindeer produced more methane, whereas methane production in cows was not altered by rumen evacuation. 16S rRNA amplicon sequencing sequencing was used to explore the diversity of bacteria, archaea, ciliate protozoa and anaerobic fungi in the rumen. Comparison of the microbial communities in the rumen of reindeer and cows before digesta exchange revealed that both species share a core microbiome. About 80% of identified bacterial taxa did not differ between species, with 10% being more closely associated with either the reindeer or cow host. Host specific taxa represented 22 and 13% of the total bacterial community in reindeer and cows, respectively. Following the transfer of digesta from the cow into reindeer, a gradual time dependent increase in bacterial communities specific to this species was detected. Such findings indicate a clear host effect influencing bacterial communities specific to reindeer. Ciliate protozoal communities in the reindeer and cow exhibited a similar range of diversity at the genus level. However, certain protozoan genus were detected at a high frequency in reindeer before the rumen evacuation, but found to be almost absent in the rumen content of cows or following digesta exchange. At the same time, protozoan of cow origin were found to survive in the reindeer rumen. Diversity of anaerobic fungi in the rumen of reindeer was found to be much greater than previously thought, and comprised of both known and still to be identified fungi. Following the exchange of digesta the abundance of fungi in the rumen was similar in the reindeer and cow. Differences between in the archaeal communities between species and following digesta exchange were much less profound with differences for only a few species being detected. In conclusion, the exchange of rumen contents from cows into reindeer, provided evidence that the host ruminant animals has a “stabilising influence” on its own rumen microflora allowing a new community structure to be rapidly established after inoculation with a different microbial population. Functioning of this new ecosystem leads to important and measurable changes in rumen fermentation, ruminal gas production and nutrient use efficiency in the host ruminant. Task 3.4.2: Twin cow-cow rumen exchange The primary objective of WP4 was to understand the role of the host-animal in controlling the composition and function of the rumen microbiome. The second objective was to monitor host genome expression responses in various tissues in relation to changes in the rumen microbiome or rumen function.Studies in mice and humans have indicated that the intestinal microbiome composition is shaped by host genetics. In ruminants, rumen microbial populations are commonly believed to be animal specific and variation in the composition between individuals and breeds has been observed. However, only a few studies have been conducted to demonstrate the host specificity of the rumen microbiome, and these have covered only the bacterial community.The second experiment in WP4 involved the exchange of ruminal digesta from identical twin cows into unrelated cows and vice versa in order to understand the genetic control influencing ruminal microbial ecology and function in the same species, and the response at host genome level to changes in the rumen composition. Measurements of rumen fermentation, methane production, whole tract digestibility coefficients and nitrogen balance were made before and 4 weeks after rumen evacuation and exchange of digesta among experimental animals. The identical cows were more similar than non-identical animals for ten production or fermentation characteristics prior to digesta exchange and only for one parameter at the end of the experiment. These observations provided tentative evidence that identical animals became less similar for most measured phenotypes after digesta exchange, an effect that had not reverted back to baseline by the end of the study.For the first objective we analysed the host effect on the rumen microbiome by metabarcode-sequencing and comparing the entire rumen microbiome before rumen exchange and at weekly intervals after the exchange. The hypotheses tested were: If the host animal controls its own microbiome i) Differences in the rumen microbiome composition are smaller between genetically similar animals compared with genetically more distant animals, and ii) After rumen exchange the rumen microbiome composition will revert back from the composition of the donor towards the original composition of the host recipient. The comparison of rumen microbiobial community composition between identical twins and unrelated cows showed that the rumen microbiome is not consistently more similar in genetically related animals than between unrelated cows. We did not get clear support for the hypothesis that the host had an effect on its microbiome. The rumen microbiome composition between twins did not follow expectations of being most different immediately after rumen content exchange, but gradually returning to a more similar composition thereafter. Instead we observed substantial between-animal variation in the response to rumen exchange and the response was specific to microbial group (bacteria, archaea, protozoa or fungi). After rumen exchange, bacteria appear to have acquired a new community structure that was different compared with the composition in the rumen before digesta exchange in all animals. The largest deviation in the community structure over the 6 week period was detected among archaea, with one individual returning back to the original composition, while others remained intermediate. In case of ciliate protozoa and anaerobic fungi some animals were consistently remaining similar to themselves before rumen exchange, while others acquired rumen composition of the donor which remained similar for the rest of the experimental period. These results agree with earlier observations indicating substantial variation in between-individual response to rumen digesta exchange and suggest that the source of the variation is not solely driven by genetic differences between individuals. Data were interrogated further to ask the question that if the host does not affect entire rumen microbial communities, it may influence specific taxa. Interestingly, individual microorganisms, including bacteria from Firmicutes phylum, some archaea, ciliates and fungi showed a pattern of gradual recovery to original status following the rumen exchange. Our observations are in line with recent findings in humans showing strongest heritabilities for Ruminococcaceae and Lachnospiraceae families from Firmicutes phylum. This study was designed to account for variation in age, stage of lactation, diet and environment allowing for a sensitive test of the host animal genetics. It appears that while there are some indications of host effects on the composition of certain taxonomic groups, there is a stochastic component that can lead to even bigger differences than originally introduced by the exchange of rumen contents. A much larger number of twins and less identical cows would be needed to confirm the findings. For the second objective we collected tissue samples from the liver and adipose tissues and rumen papillae. Adipose and liver biopsies were taken during the two periods of physiological measurements before and after rumen content exchange (at weeks 4 and 8 of the experiment). Rumen papilla samples were collected at the time of digesta exchange and at weekly intervals when rumen contents were sampled. The samples were deep sequenced for their transcriptome (all expressed mRNAs) and the papillae also for regulatory miRNAs.The hypotheses to be tested were: i) changes in rumen content may affect rumen epithelium gene function and thereby nutrient absorption kinetics, ii) these changes may be reflected in the associated gene functions and pathways in the liver and adipose tissue.The rumen papillae transcriptome has not been extensively studied and our results serve as a good starting point for further analyses. An interesting observation is that nearly half of the mapped sequences were located in regions of the bovine genome with no annotated (known) function. We are still working towards functional annotation of these regions by comparing them to other vertebrate genomes and functional elements.Differences in papillae gene expression within animals before and after rumen content exchange were used to identify subgroups of animals with similar responses. The animals formed two clusters that were shown to differ in 12 rumen functional characteristics before the exchange, indicating that some or all of these differences may have caused the differences in the papillae gene expression response. Differences in miRNA expression between the two groups were also revealed, indicating putative regulatory effects on specific target genes. We are still working on ways to combine the microbial composition data with the gene expression, miRNA expression and physiological data in a meaningful way.The liver and adipose tissue gene expression analysis showed that the twins were more similar to each other in gene expression profiles than the unrelated cows before and after the exchange of rumen content. Regarding liver gene expression, the twins became slightly less similar to each other after the rumen content exchange, but still all animals remained more closely correlated to themselves before exchange than to any other animal. Interestingly, in the adipose tissue the situation was quite different, with much less correlation between and after rumen exchange within each animal; i.e. the adipose gene expression was affected more by the rumen exchange in all animals. In fact, some animals’ gene expression profiles were more correlated with others’ after the exchange than before. The genes responding most strongly to the rumen content exchange in each exchange pair in the adipose were analysed further by pathway analysis. The top affected gene network was related with carbohydrate and lipid metabolism, including key genes like LPL, FASN, SCD, DGAT2, INSIG1 and RXR receptors.In conclusion, the gene expression pattern in rumen epithelium was affected by the rumen content exchange, and specifically correlated to functional differences of the rumen content. Gene expression in the adipose tissue was thereby altered more than gene expression in the liver, especially for genes involved in carbohydrate and lipid metabolism. Full understanding of the sequence of events in the different tissues will require more comprehensive modeling analyses including data at all levels simultaneously.3.5: Nutrition-microbiome-emissions/product quality interactions3.5.1: Impact of dietary carbohydrate on the composition and function of the rumen microbiome Increasing the proportion of concentrates in ruminant diets is one strategy to reduce methane production, but high levels of concentrates decreases fibre digestibility, can cause health problems and uses resources that can be used directly as human food or more efficiently in the diets of simple-stomached animals with much lower emissions. Two experiments were conducted to investigate if the high amount of grain in a diet based medium-quality silage can be decreased by improving forage quality without compromising milk production or environmental emissions. Sixteen cows were used in 4 x 4 Latin square design to study the effects of graded replacement of medium-quality silage and barley with high-quality silage. The four diets were formulated to produce same amount of energy-corrected milk. Feed intake decreased with increasing proportion of high-quality silage in the diet, but milk or energy corrected milk yield were not influenced by diet. Milk fat concentration increased and protein concentration decreased with the proportion of high-quality silage in the diet. Total methane production and methane per kg energy-corrected milk yield were not influenced by the diet, but methane emissions per kg dry matter intake increased with the proportion of high-quality silage in the diet. Efficiency of nitrogen utilisation was not influenced by the diet, but feed efficiency in terms of energy corrected milk yield per kg dry matter intake improved when the proportion of high-quality silage in the diet increased.The same diets were used in a 4 x 4 Latin square design to investigate the effects on rumen metabolism, rumen microbiome and digestibility. Differences in feed intake and milk production were consistent with those observed in production study. Rumen fermentation pattern was not influenced despite large differences in carbohydrate composition (replacement of barley starch with digestible fibre). Diet digestibility increased with the proportion of high-quality silage in the diet with the most evident differences in the digestibility of neutral detergent fibre and it’s potentially digestible fraction. Therefore, the amount of potentially fermentable substrate in manure could potentially be greater for diets based on high proportion of grain. The flow of nitrogen components into the omasum tended to decrease with the proportion of high-quality silage reflecting mainly reduced feed intake. Diet composition had only marginal effects the rumen microbial composition and the abundance of different microbial groups despite considerable differences in dietary carbohydrate composition. Most prevalent microorganisms in the rumen were unaffected, but changes in specific less abundant taxa were observed.3.5.2: Effect of dietary protein on the composition and function of the bovine rumen microbiome Protein supplements are commonly fed to dairy cows to increase milk production. However, incremental protein is used for milk protein synthesis with a low marginal efficiency, decreases efficiency of N utilisation and increases milk urea N (MUN) concentration. Less is known about the effects of protein supplementation on methane production and about between-cow variation in MUN and its association to nitrogen utilisation. A meta-analysis was conducted to evaluate within-cow variations in the concentrations of MUN and relationships between MUN and nitrogen utilisation. The effects of diet and period were removed in statistical analysis. MUN concentration displayed rather high between-cow variation and repeatability. MUN concentration was negatively related to nitrogen utilisation, but the effect per unit of MUN was much less than observed when MUN was influenced by diet composition. Overall, improvements in management and a closer control over diet composition relative to requirements appear to have greater potential to improve nitrogen utilisation of lactating cows than selection of cows with an inherently low MUN concentration. Rumen ammonia N concentration was positively associated to diet digestibility, but the concentration for optimal diet digestibility appears to be higher than that required for maximizing microbial protein supply.The effects of four incremental levels of soybean meal and rapeseed meal supplementation on methane production were investigated in a cyclic change-over study with 28 cows. Total methane production was not influenced by the level or type of protein supplementation, but some decreases in methane production per kg dry matter intake were observed. Methane production per kg energy corrected milk decreased more in cows fed rapeseed meal diets compared to those fed soybean meal diets. Efficiency of nitrogen utilisation decreased with increasing dietary protein concentration. Negative effects of increased protein feeding on nitrogen efficiency are most likely offset reduced methane emissions.Four cows were used in 4 x 4 Latin square design to study the effects of four different protein levels on the rumen microbiome, rumen metabolism and digestibility. Dietary protein concentration was increased by replacing crimped barley with graded levels of heat-treated rapeseed meal in a total mixed ration fed ad libitum. Increasing dietary protein concentration increased feed intake and milk production, but decreased efficiency of N utilisation. Omasal flow of total nitrogen and feed nitrogen increased with the level of protein supplementation, but microbial N flow tended to decrease despite of increased feed intake. The efficiency of microbial protein synthesis tended to decrease with increased dietary protein concentration. The estimates of dry matter flow to omasum and reticulum were strongly correlated suggesting that omasal sampling system might be replaced with simpler reticular sampling. Increases in dietary protein content had only marginal effects the rumen microbial composition and the abundance of different microbial groups. Most prevalent microorganisms in the rumen were unaffected, but changes in specific less abundant taxa were observed.5.3.3: Role of medium chain fatty acids and plant oils on the composition and function of the bovine rumen microbiomeFat supplements are typically used to increase energy content of the ruminant diet and can be used to alter the fatty acid composition of meat and milk. Depending on composition, lipid supplements may influence rumen function decreasing enteric methane production and nutrient digestion. Much less is known about how fat in the ruminant diet influences the rumen microbial ecosystem. An experiment with lactating cows was performed to address this lack of knowledge. Five cows were used in a 5 x 5 Latin square design to study the effects of four different fat supplements on the rumen microbiome, environmental emissions and product quality. Supplements rich in myristic acid, oleic acid, linoleic acid or linolenic acid were included (5% dry matter) in a total mixed ration based on grass silage offered ad libitium. Adding fat to the diet lowered intake, methane production and milk production. Methane production per kg of energy corrected milk was also decreased. Including extra fat in the diet did not influence the efficiency of nitrogen utilisation, however. All fat supplements lowered milk 4- to 12 and 16 carbon saturated fatty acid concentrations. Depending on source, fat supplements also altered the proportions of trans fatty acids in milk. Increases in dietary fat content altered the rumen microbial composition and the abundance of different microbial groups. Most prevalent microorganisms in the rumen were unaffected, but changes in specific less abundant taxa were observed. Even though changes in the bacterial and protozoal communities were detected, myristic acid supplement had a greater influence on the anaerobic fungi. Rather large variability between individual animals prevented a clear distinction of changes in enteric methane production and the rumen archaea population.3.6: Molecular Analysis3.6.1: Microbial genomicsThe genus Butyrivibrio and related strains have key roles in the ruminal digestion of plant fibre, which leads to H2 and therefore methane production. They also have a vital role in lipid metabolism. ‘Hyper-ammonia-producing bacteria’ (HAB) are major contributors to high N excretion in ruminants, because they deaminate amino acids extremely rapidly in the rumen.Eight bacterial genomes were sequenced by Illumina HiSeq 250-bp next-generation sequencing. Two separate sequencing runs were carried out on all bacterial DNA samples and then assembled into between 60-170 scaffolds each. They were annotated using Prokka and the data fed in to WP73.6.2: DNA 16S rRNA amplicon sequencing:During the course of the project 1437 rumen samples were collected and the DNA analysed for five DNA amplicon markers especially designed and targeting bacteria, archaea, prokaryotes, ciliate protozoa and Neocallimastigaceae fungi. Sequencing was performed using the Illumina MiSeq and HiSeq platforms and the raw sequence reads passed on to WP7 for analysis. 3.6.3: Metagenomes: High, low and intermediate methane emitting animals were identified from the 1000 cow population based on the relationship between methane output vs. dry matter intake weighted for the live weight of the animal. Rumen samples were processed to extract the DNA and deep shotgun sequencing was performed using the Illumina platform. On average 101,500,000 paired end reads (median 103,000,000 PE reads) were generated for each sample that passed Illumina standard filtering on the instrument, for an average of 20 GBases of sequence data per sample. The raw sequences were further processed and analysed in WP73.6.4 GenotypingBlood samples were collected from the 1000 cows in WP3 and sent to PTP for DNA extraction. In total, 1005 bovine whole blood samples were received in vacutainers, already labelled according to the projects samples identifiers and the country of origin. Once DNA was extracted and quantified, the samples were normalized at the concentration required by the Genotyping Service Company (NeoGen www.neogen.com) i.e. 50ng/μl at a final volume of 20μl for each sample.The extracted DNA was dried and the samples sent in plates to NeoGen company to perform the genotyping on the Bovine GGP HD (GeneSeek Genomic Profilers) in 3 different batches, according to samples collection progression in WP3. The 200 cows coming from Finland and Sweden were genotyped using the Bovine GGP HD chip v1 (80K) that included 76.883 SNPs, while the 800 samples from UK and Italy were genotyped in a second moment and were processed using the Bovine GGP HD chip v2 (150K) that included 138.892 SNPs, as the v1 of the chip was no longer available from the manufacturer. The v2 of the chip includes all the SNPs that were present in the previous v1 of the chip, while at the same time providing more markers for the same final processing cost.Neogen company received the samples prepared in plates by PTP and performed the DNA hybridisation, image scanning and data acquisition of the genotyping chips according to the manufacturer's protocols (Illumina Inc.).3.7: Data management and analysis 3.7.1: Data warehouse and information managementThe work conducted in the data warehouse for the public resources was to prepare an online web portal for the community working on rumen biology and metagenomics, based upon the Ruminomics experience in data and information management developed during the project. A website was built to create a central resource for rumen microbiome analysis and is available at the following address: http://bioinformatics.tecnoparco.org/ruminomics. The project data including the phenotypic measurements, the metagenomes and the 16S rRNA amplicon sequencing sequencing data will be submitted to EBI databases once published in peer-reviewed journals. The website developed was built to expose the information and to allow an easy access to the rumen data and experiments currently available in public databases. Specific sections have been built also to promote the data centralisation and submission of rumen experiments generated by the international community, following the same strategies used by Ruminomics project.The online resource developed allows facilitating the exploration and collection of rumen data from the scientific community, relying at the same time on public databases from EBI that will provide long term data accessibility and simplified data sharing and integration across rumen microbiome experiments. A section was also dedicated to describe the projects data analysis strategies and protocols for rumen microbiome and metagenomic data, as well as to provide ready to use templates for animal measurements and phenotypes data collections based upon the project experience.3.7.2: Analysis of 16S rRNA amplicon sequencing data16S rRNA amplicon sequencing was performed on the DNA of all the 1437 rumen samples collected within the project. Raw sequences generated in WP6 were assembled to pair-end reads and assigned to the corresponding marker and individual. The sequences were further processed to remove error or chimeras before taxonomic assignment.For each marker system, we report here only the taxa identified at the taxonomic level relevant for subsequent diversity analyses: phylum level for Bacteria, species level for Archaea, family level for Ciliates, genus level for Neocallimastigaceae fungi, and superkingdom level for prokaryotes.Bacteroidetes (47.19%) were the most abundant in the Bacteria phylum followed by Proteobacteria (23.68%), Firmicutes (15.81%), and Fibrobacteres (4.45%). Of the archaea species Methanobrevibacter smithii (50.83%) were in the majority with Methanobrevibacter ruminantium (29.57%), and Methsnosphaera sp. (8.45%). 83.1% of the ciliates came from the Ophryoscolecidae family, 16.71% fropmm the Isotrichidae family, 0.18% Blepharocorythidae and the rest unidentified. For fungus the Cyllamyces were the most abundant at 32.2% with Neocallimastix at 31.63%. The five marker systems produced data globally consistent with what is already known about rumen microbial composition. For example, about 5% of the reads obtained with the prokaryote marker were assigned to the archaea superkingdom, a value congruent with previous findings (Wallace et al. 2015). Moreover, the percentage of sequences with a taxonomic assignation at the taxonomic level relevant for subsequent analyses was satisfactory, since it ranged from 90.5% for the archaea dataset at the species level to 100.0% for the prokaryote dataset at the superkingdom level. 3.7.3: Metagenomic sequencing analysesThe raw sequencing data of the 60 samples from the high low and intermediate methane emitting cattle from task 6.3 was processed to remove low quality bases and sequencing adapters before being assembled into contigs. These were processed to remove further redundant information and predict gene and protein sequences. Ribosomal genes were extracted for taxonomic analysis against the M5rna database and the predicted genes were annotated.The analysis tools of MG-RAST were used to perform these tasks and explore the content of each group of samples in each country and data were also extracted from MG-RAST and imported in R (v3.2) for advanced inspection and graphical representation of the metagenomes functional analysis outputs. A series of information were extracted from the metagenomes of each country, to explore the correlation of the metagenome content with the samples groups defined by the methane emissions measurements.All the functional analysis were performed using the KEGG Orthologs sequences as reference for the annotation of the predicted genes and proteins on the metagenomes, using a sequence identity cut-off of 60% and a E-value cut-off of 1e-5 for the similarity searches. For each group of samples in each country, the relative abundances of key metabolic pathways have been calculated and plotted, to verify the prevalence of genes involved in methane pathways in the rumen microbiomes of the different methane emitters groups.The metagenomes analysis showed that the different methane emitters groups within each country have a consistent and distinct composition in terms of genes and biological function, especially for Italian, Swedish and UK samples. The in-depth analysis of the genes involved in key biological pathways such as the methane pathway (KO00680 http://www.genome.jp/kegg-bin/show_pathway?ko00680) showed, for all countries, a higher abundance of the microbial genes involved in the methane metabolism for the high methane emitters animals compared to the other groups. The analysis of the metagenomes of these selected animals in the 4 countries offers a unique overview of the role of the microbiome in the control of the methane emissions, highlighting the changes in the gene functions of the rumen microbiome in the selected conditions.3.7.4 Statistical modelsA standard genome-wide association study (GWAS) approach was followed with the full set of genotyped animals and one SNP at a time was fitted in a linear regression model for continuous variables.Population structure is a known source of bias in GWAS analyses, leading to possible false positive associations (e.g. reviewed by Goddard and Hayes, 2009; Hayes, 2013). The multidimensional scaling plot of the genetic distances between animals, clearly showed the presence of a strong population structure in our data. Such structure was attributable to the two breeds (Holstein Frisian and Nordic Red) from four countries (Italy, Sweden, UK, Finland) used to collect phenotypes and genotypes in the project. Consequently, a simple GWAS analysis without correcting for the population structure showed a large amount of background noise and a lambda inflation factor over 6.23 which indicates probable spurious associations (Amin et al., 2007). Including the breed effect in the model reduced inflation to lambda = 1.59 thereby improving the fit. A further improvement of the model was obtained by adding the kinship matrix and polygenic effect to the GWAS model, which reduced inflation to lambda = 1.004 (virtually no inflation). The goodness of fit with the latter model was also confirmed by the qq-plot, which showed good agreement between residuals of the model and theoretical residuals, assumed to be normally distributed. The final model including the breed effect and the polygenic effect (via the kinship matrix) was therefore selected and applied to GWAS for the traits recorded within the project. The results of the GWAS analysis for the methane emissions phenotypes is summarized in the tables below:Top Significant SNPs for Methane Emissions (gram / day)SNP Chromosome Position Pc1dfBovineHD0400026833 4 96264925 9.24E-09ARS-BFGL-NGS-23713 10 77423941 1.64E-07BovineHD2500003888 25 13942842 1.42E-05ARS-USDA-AGIL-chr7-10908243-000745 7 10908243 2.02E-05BovineHD2500003899 25 13982998 2.61E-05Top Significant SNPs for Methane Emissions (gram / Kg of dry matter intake)SNP Chromosome Position Pc1dfBovineHD1200014532 12 52684800 4.26E-05BovineHD0600010656 6 38540049 5.34E-05BovineHD1500003133 15 12282171 6.89E-05BovineHD1700002623 17 9298847 9.53E-05BovineHD0500031816 5 110342273 0.00010092Top Significant SNPs for Methane Emissions (gram / Kg of energy corrected milk)SNP Chromosome Position Pc1dfBovineHD1700002623 17 9298847 6.71E-06BTB-01377157 8 112758017 1.07E-05BovineHD0100045391 1 155468198 2.68E-05Hapmap51102-BTA-97964 6 55462479 3.01E-05ARS-BFGL-NGS-87397 15 24167440 4.05E-05Pathway analysis on the genes found in the associated regions on the cow genomeThe top SNPs identified in the genome wide association analysis (GWAS) performed for the methane emissions were used to identify the genes present in the chromosomal regions.The list of SNPs for each GWAS analysis were imported into R and the region under each SNP was inspected up to 500Kb upstream and downstream for the presence of annotated genes. The genes in the corresponding chromosomal regions were retrieved using the biomaRt library (Durinck S et al 2009) and querying the Ensembl database (Cunningham F et al. 2015) to retrieve the corresponding Gene IDs.The list of SNPs for each phenotype were then processed using Reactome (Fabregat A et al. 2015) tools to verify their presence within known biological pathways in Bos taurus and to assess the pathways enrichment on the gene lists generated. The top SNPs for each GWAS analysis were selected having a Pc1df value below 5e-5.The analysis of the genes found in the associated regions from the genome wide association study conducted highlighted a number of different biological pathways that are significantly enriched in the selected gene sets present in the regions associated with the methane emissions. The genes found are directly involved in pathways correlated with the signal transduction (Olfactory Signaling Pathway R-BTA-381753), cell interactions (Integrin cell surface interactions pathway R-BTA-216083) and apoptosis (Apoptotic execution phase pathway R-BTA-75153), which may play a role in controlling this trait in the host through biochemical signaling mechanisms.3.7.5: Environmental footprint and efficiencyFeed costs represent >70% of the total costs of milk production, so improvements in feed efficiency are generally associated with increased profits unless prices of diet ingredients change. Feed efficiency can be defined in several ways, all involving a measure of milk output divided by a measure of feed input. For RuminOmics we calculated energy-corrected milk yield per unit of dry matter intake, and milk solids yield (fat + protein + lactose yields) per unit of dry matter intake. Because feed efficiency is a ratio of milk yield and dry matter intake, which are themselves positively correlated, it is not possible to separate influences on the numerator, the denominator, or both. Thus, feed efficiency can be improved by increasing milk yield with no change in dry matter intake, by decreasing dry matter intake with no change in milk yield, or a combination of increased milk yield and decreased dry matter intake.Feed efficiency is influenced by variations in diet composition, including factors such as palatability, digestibility, ratio of forages to concentrates, energy density and nutrient concentrations. Even when fed on the same diet, however, individual cows vary in feed efficiency. Part of this variation can be explained by differences in how cows interact with the diet, e.g. differences in feeding behaviour, appetite, rumination and chewing behaviour, passage rate, rumen environment and microbial population; part can be explained by differences in post-absorption partitioning of nutrients between milk and body tissues, which is under genetic and hormonal control. In the 1000-cow study (WP3), after adjusting for diet effects, feed efficiency was positively correlated with milk yield and negatively correlated with dry matter intake (Table 1), as expected because these variables are used to calculate feed efficiency. Feed efficiency was also negatively correlated with live weight, which is a driver of dry matter intake. Feed conversion efficiency was positively correlated with total rumen volatile fatty acid (VFA) concentration, indicating that cows which produce more VFA from their feed are more efficient. Feed conversion efficiency was positively correlated with rumen propionate and negatively correlated with rumen acetate, indicating that cows which produce more propionate rather than acetate are more efficient; propionate is the main precursor of plasma glucose in ruminants, which is essential for synthesis of lactose, the main determinant of milk yield. The negative correlation between feed efficiency and acetate concurs with the negative correlation between feed efficiency and milk fat concentration; it also concurs with the negative correlation between feed efficiency and milk saturated fatty acids which, as already noted, are partly synthesised from acetate. The positive correlations between feed efficiency and unsaturated milk fatty acids (MUFA, PUFA and trans) could be simply the corollary of the negative correlation between feed efficiency and milk saturated fatty acids, or they could indicate that incorporation of preformed long-chain fatty acids into milk fat is energetically more efficient than de novo synthesis of short-chain fatty acids.The contribution of between-cow variation in methane emissions to milk production or feed conversion can be only marginal. Assuming 10% coefficient of variation in methane emissions of cows at the same level of intake (20 kg dry matter per day) ± 1 SD variation in methane emissions corresponds to about 4.8 MJ of metabolisable energy. When cows are at zero energy balance about 50% of incremental energy is partitioned to milk production, so ± 1 SD range in methane production corresponds to about 0.45 kg energy corrected milk provided that diet digestibility is not decreased. In the 4 x 4 production study (WP 5.1) investigating the effects of replacing medium-quality grass silage and barley gradually with high quality silage, feed efficiency (kg energy corrected milk per kg dry matter intake) was positively associated with methane production per unit of intake when the effects of period and diet were removed. This can at least partly be because diet digestibility was positively related to methane production. In the 1000-cow study (WP3), methane production per unit of dry matter intake was positively correlated with feed efficiency (Table 1) and with digestibility (r = 0.109 P<0.001). This provides further support for the hypothesis that more efficient cows produce more methane per unit of feed intake.Table 1. Correlations between feed efficiency (FE) calculated as kg energy-corrected milk (ECM) or kg milk solids per kg dry matter intake (DMI) and feed intake, live weight, milk variables and rumen volatile fatty acids in 1016-cow sample population of cows Correlationwith FE(kg ECM/kg DMI) Probability Correlationwith FE(kg milk solids/kg DMI) ProbabilityDry matter intake (kg/d) -0.445 <0.001 -0.455 <0.001Live weight (kg) -0.169 <0.001 -0.177 <0.001Methane (g/day) -0.087 0.034 -0.137 <0.001 (g/kg ECM) -0.515 <0.001 -0.530 <0.001 (g/kg DMI) 0.338 <0.001 0.307 <0.001Milk: yield (kg/d) 0.425 <0.001 0.466 <0.001 fat (g/kg) -0.060 0.106 -0.198 <0.001Milk fatty acids (% total FA): saturated -0.090 0.021 -0.168 <0.001 monounsaturated 0.093 0.016 0.167 <0.001 polyunsaturated 0.015 0.777 0.082 0.044 trans 0.100 0.012 0.180 <0.001Rumen VFA (mmol/l) 0.103 0.012 0.099 0.016Rumen VFA (mmol/mol): Acetate -0.172 <0.001 -0.218 <0.001 Propionate 0.180 <0.001 0.238 <0.001 Butyrate -0.061 0.132 -0.111 0.007 Iso-butyrate -0.040 0.333 -0.044 0.281 Valerate 0.071 0.083 0.122 0.003In summary, cows with greater feed efficiency are those that digest feed to a greater extent, thereby generating more rumen VFA including acetate and hence methane. The implication of this finding is that selecting cows for low methane emissions alone could result in lower feed efficiency. This would have an adverse effect on methane emissions at the herd and national levels because more cows and more total feed would be required per unit of milk produced. The recommendation, therefore, is to select cows with high feed efficiency and low methane emissions.Nitrogen efficiencyEfficiency of nitrogen (N) utilisation of milk production (NE) is usually expressed as milk N output per N intake. As for FE, NE can be improved by increasing milk protein yield at constant protein intake or maintaining protein yield while decreasing protein intake. In dairy cows NE is rather low, typically 25-30%. As N retention is very small, on average about 70% on N intake is excreted in manure, which contributes to both atmospheric and hydrospheric pollution. Urinary N has more adverse effects on environment than faecal N as is more susceptible for leaching and evaporation losses than faecal N.For the 1000-cow study NE was on average 29.4% but it varied considerably in the population (SD=5.7%). Feacal and urinary N output was 31.8 and 37.4% of N intake, respectively. NE was negatively related to N intake (r = - 0.38) and positively to protein yield (r = 0.42). Faecal and urinary N output were strongly and positively (r > 0.80) related to N intake. Nitrogen efficiency was correlated with three bacterial phyla and two fungal genera, although correlation coefficients were low. Nitrogen efficiency was negatively correlated with the proportions of rumen acetate (r = -0.15) butyrate (r = -0.10) and isovalerate (r = -0.14) and positively correlated with rumen propionate (r = 0.24) and valerate (0.08) but the correlation with rumen ammonia was not significant. The relatively strong correlation between rumen propionate and nitrogen efficiency concurs with correlations of rumen propionate with milk protein yield (r = 0.211; P < 0.001). In meta-analysis of data from production studies (WP5) rumen ammonia and milk urea concentrations were negatively related to NE and positively to urinary N output when the effects of diet and period were removed. However, the relationship between milk urea concentration and NE was much weaker than found when milk urea was influenced by dietary protein. It was concluded that between-cow variation is too small to rank the cows according to NE when fed the same diet.In summary, selecting cows for greater overall feed efficiency will also lead to improved N efficiency. As for FE, selecting for low methane emissions alone can result in reduced NE. Overall, improvements in management and a closer control over diet composition relative to requirements appear to have greater potential to improve NE of lactating cows than selection of cows with an inherently low methane emissions of low milk urea concentration.3.7.6: Product quality, health and safety The aim of this part of the project was to describe how the rumen microbiome composition is reflected in milk fatty acid composition and to determine if milk fatty acids have value in predicting methane emissions. The milk fatty acids composition of the 1000-cow samples, comprising up to 200 different fatty acids, were incorporated into a large database describing the 1000-cow experiment. The major features of the community were analysed by qPCR of the microbial ss rRNA gene and by extraction of phylum information from amplicon sequencing. The influence of the microbiome on product quality of milk in terms of human health was assessed by the proportions of saturated fatty acids (SFA), particularly 16:0, monounsaturated fatty acids (MUFA) and polyunsaturated fatty acids (PUFA). Individual components with recognized health-promoting properties are cis-9,trans-11-CLA (rumenic acid) and trans-11-18:1 (vaccenic acid). The key fatty acids for animal health are the corresponding di- and monoenoic acids, trans-10,trans-12-CLA and trans-10-18:1, which are associated with milk fat depression. Novel, surprising findings included (i) an inverse relationship between the degree of unsaturation and archaeal abundance, with certain ribotypes more strongly associated than others and (ii) the strongest links between health-associated fatty acids and ribotypes were for t10,c12-CLA and t10-18:1. Archaea have been largely ignored as participants in biohydrogenation; this view must now be revised in light of (i). In terms of t10,c12-CLA and t10-18:1, the suspected main species to be involved in their metabolism were rare in the microbiome; other candidates have now been identified.3.7.7 Systems biology and pathway analyses of the hostThe analysis of the genes found in the associated regions from the genome wide association study conducted in T7.4 highlighted a number of different biological pathways that are significantly enriched in the selected gene sets, in particular for the regions associated with the methane emissions and the relative abundances of specific bacteria phyla.The genes found in the regions associated with methane emissions measurements are directly involved in pathways correlated with the signal transduction, cell interactions and apoptosis, which may be an indication of a role of the host in controlling, through biochemical signalling mechanisms, this particular trait.The genes found in the associated regions for the abundances of the bacteria phyla were found involved in pathways related to immune response, cell migration, fatty acids and hormone transport, complement cascade and signal transduction. These findings may suggest a role of the host in the definition of the rumen microbiome composition, using innate biological mechanisms such as the immune response, signalling cascades and hormone controls.3.83.8.1: RuminOmics ring testSince the introduction of molecular methods, it has been possible to characterise the profile of the rumen microbiota without the need for in vitro culture. This is achieved by amplifying, by PCR, a target gene using primers universal to a selected taxonomic group. This is usually at the domain level, targeting the 16S rRNA gene for bacteria and archaea, 18S rRNA for ciliates and the 18S ITS region for fungi. The description of the ruminotype can vary depending on the method used. There are several examples of the effect of different methods at various stages in the process that can influence the result in terms of OTU richness, relative abundance and the taxonomic classification of the species of the microbiome. The aim of the RuminOmics ring test was to assess how closely microbiome analyses of the same samples of DNA by different laboratories around the world resembled each other an ultimatelye to standardise a methodology for NGS rumen microbiome characterisationEight international research institutes with an expertise in rumen microbiology participated. The choice of methods were left entirely to the discretion of the participants, who were free to select the primer set, PCR conditions, sequencing platform, QC and bioinformatic pipelines that they used routinely for their own research. Method summaries were obtained by questionnaire filled in by the respective participants with details of the steps considered most likely to affect the result (Table 1).The starting point for the analysis was genomic DNA from two each of three ruminant species, red deer, sheep and reindeer, extracted using a standard procedure It was left to each of the participants to produce an OTU table containing abundance and taxonomy information to describe the microbiome composition in each of the experimental animals.The results were different for each participant, particularly for archaea, as illustrated by Fig. 8x. Reasons for the differences were predominantly. The picture that has emerged is a weak clustering of ruminotypes by animal provided that the same primer choice, PCR conditions and taxonomic classification method were used. Harmonising the processing of the raw sequence data and running a common analysis pipeline including classification using a recognised rumen species reference database may help to establish stronger clusters within animal and reduce the methodological effects separating the participants’ data.3.8.2: Rumen metaproteomics The aim of Task 8.2 was to assess whether metaproteomics can be a useful tool to characterise the functions of the rumen microbial ecosystem. Metaproteomics can be considered as complementary to metagenomics, in the sense that metagenomics describes the genetic capability of the microbial community, while metaproteomics describes the proteins that are actually expressed. It was thought possible that, if good quality data could be produced, then certain proteins or patterns of proteins could be related to the status of the animal in terms of production and methane emissions. The benefits and drawbacks of the two main metagenomics techniques have been investigated, and the results and conclusions are reported here. 2-D PAGE has the advantage of a visible output from each sample, but it suffers from poor sample-to-sample variation, and contamination by plant materials. Shotgun metaproteomics provides a wealth of detail; it suffers, however, from limited database reliance to ruminal microorganisms and from greater expense. Both methods showed a remarkable and unexpected capability for predicting taxonomic composition. Abundant proteins and their role in cell functions were also identified to a satisfactory degree. Although the shotgun method performed better in other respects, there was a significant drawback when assessing eukaryotic proteins. Although plant proteins were rare on 2-D PAGE, which monitors intact proteins, the shotgun method includes protein breakdown products. Peptides deriving from the partial digestion of plant proteins swamped microbial eukaryotic proteins. Applying both methods to compare digesta from high and low methane emitters was unsuccessful in identifying proteins that differed between the two groups.3.8.3 Public bioinformatics resourcesThe data warehouse system created in WP 7 (http://bioinformatics.tecnoparco.org/ruminomics) was extended and opened to public access. The portal is organized into 3 main sections:Explore public databasesThis section allow an easy connection to the EBI European Nucleotide Database that stores sequence data, the EBI Metagenomics database that holds information about shotgun metagenomics and 16S rRNA amplicon sequencing projects and the EBI BioSamples database that stores information on phenotypes and samples collection. To facilitate the retrieval and the data mining of public rumen datasets, this section directly points to all submitted data in these databases that belongs to rumen experiments. This section is specifically dedicated for the community, to have a unique point of access to all these databases and datasets and facilitate the exploration of the information currently available on rumen metagenomics and related animal physiological data. The corresponding database interfaces allow for a number of queries to be performed over the stored datasets, including filtering by species and free text search over samples and projects descriptions, to allow the users to retrieve the relevant data for their researches.Submit data to public databasesThis section facilitate the submission of rumen data and experiments for the international community, by centralizing the access to BioSamples and ENA databases. The community will have a unique online resource that point to the corresponding sections of the public databases, with detailed instructions for data submission and data sharing.With these sections the Ruminomics project is facilitating and actively promoting the centralisation of rumen microbiome data and experiments by the international community. The use of public resources and database for data submission and data sharing avoid issues with long term data availability after research projects have ended and data is published, facilitating also the integration of new information and results with already existing public datasets submitted to the same online resources.Two sections point directly to the ENA and BioSamples submission forms:Ruminomics template and guidelines for data collectionThis section provides detailed documentation on the analysis strategies and tools used by Ruminomics to process the metagenomics and 16S rRNA amplicon sequencing data along with templates used by the project to collect the animal phenotypes and on farms measurements. This section provides useful guidelines for the community on how the project generated and collected the data. The templates for animal measurements made available on this section were built following the activities performed in WP3 and were carefully structured to account for all the possible measurements and the different data collections strategies on farms for methane emissions, milk production and rumen content analysis along with animal information as well as milk fatty acids and blood metabolites. These templates will be useful for the international community to allow other rumen microbiome projects to store and collect animal data in a similar way to the Ruminomics projects, facilitating the future data sharing and integration of information across different projects. 3.8.5: Lipidomics as a surrogate for methane production The aim of this part of the project was to investigate whether milk fatty acid composition and to determine if milk fatty acids have value in predicting methane emissions and the ruminal microbiome. The milk fatty acids composition of the 1000-cow samples, comprising up to 200 different fatty acids, were incorporated into a large database describing the 1000-cow experiment. The database also included methane emissions. The major features of the community were analysed by qPCR of the microbial ss rRNA gene and by extraction of phylum information from amplicon sequencing. The predictive value of individual fatty acid concentrations for methane emissions was weak. Multiple correlations were found, however, few of which had been observed previously. The Protobacteria phylum was inversely related to methanogenesis. When ribotypes were analysed at the genus level, two surprising findings emerged. The first was that the fungi were the group most strongly correlated with methane emissions, but contrary to what might have been predicted the relationship was inverse, i.e. high fungal abundance was associated with lower methane. As fungi produce abundant H2, the correlation might have been expected to be in the opposite direction. The other main novel finding was that there seem to be two taxonomically distinct groups of Prevotella that are associated with methane emissions, one that are more abundant in high emitting animals, the other more abundant in low emitting animals. The metabolic basis for these associations is unknown but could prove useful for future developments in mitigation of methane emissions.3.8.5: Oral Samples as Non-Invasive Proxies for Assessing the Composition of the Rumen Microbial Community Microbial community analysis was carried out on ruminal digesta obtained directly via rumen fistula and buccal fluid, regurgitated digesta (bolus) and faeces of dairy cattle to assess if non-invasive samples could be used as proxies for ruminal digesta. Samples were collected from five cows receiving grass silage based diets containing no additional lipid or four different lipid supplements in a 5 x 5 Latin square design. Extracted DNA was analysed by qPCR and by sequencing 16S and 18S rRNA genes or the fungal ITS1 amplicons. Faeces contained few protozoa, and bacterial, fungal and archaeal communities were substantially different to ruminal digesta. Buccal and bolus samples gave much more similar profiles to ruminal digesta, although fewer archaea were detected in buccal and bolus samples. Bolus samples overall were most similar to ruminal samples. The differences between both buccal and bolus samples and ruminal digesta were consistent across all treatments. It can be concluded that either proxy sample type could be used as a predictor of the rumen microbial community, thereby enabling more convenient large-scale animal sampling for phenotyping and possible use in future animal breeding programs aimed at selecting cattle with a lower environmental footprint.3.8.6: Estimated breeding valuesA linear mixed model approach was used for BLUP estimation of genetic parameters and genomic breeding values (GEBVs) for milk yield (kilograms per day) and methane emissions (grams per day) in dairy cattle. The covariance structure between observations was modelled through the genomic relationships between animals, which were based on genotypes at 115,816 SNP, and calculated according to Van Raden (2008).The matrix G of genomic relationships was then used to set up the variance and covariance structure of the mixed model for genetic analysis: a bivariate model for methane emissions and daily milk production was fitted. A bivariate model allows to estimate the genetic correlations between traits, and to better adjust methane emissions for the level of milk production. The systematic effects included in the model were country of origin, live weight of the animals, dry matter intake (kilograms per day), parity, herd, and the sampling period. From the variance components estimated with the bivariate model, the genetic were obtained. The heritability for methane emissions was 0.6744 (±0.0895) and the heritability for daily milk production was 0.5806 (± 0.1082). The genetic correlation between methane emissions g/d and daily milk yield kg was -0.3118 (± 0.2096); the phenotypic correlation was -0.1456 (± 0.1012). A negative genetic (and phenotypic) correlation between methane emissions and milk yield was therefore estimated from this experiment. It is known that methane emissions from the rumen are related to a loss of metabolic energy in cows (e.g. Pickering et al., 2015); cows that emit more methane subtract energy from milk production and therefore have lower milk yields and vice versa. A mild negative correlation appears therefore to make sense biologically.Genomic breeding values (GEBVs) for methane emissions and daily milk production were predicted for all animals from the bivariate model. The average accuracy of GEBVs from model was 0.986 (± 0.004) for CH4 and 0.903 (± 0.013). The higher accuracy for CH4 compared to milk yield in the bivariate model can be explained with the higher estimated heritability for methane emissions (0.67 vs 0.58).The software ASREML (Gilmour et al., 2009) was used to estimate variance components and breeding values in a restricted maximum likelihood (REML) framework.Potential Impact:4. Potential impact, main dissemination activities and exploitation of resultsThe ruminant livestock production industry is an important component of future food security. Ruminants are capable of converting non-competitive plant resources into high quality human foods through a symbiotic relationship with the gut microbiota. Nevertheless there remain concerns over the contribution of ruminant production to greenhouse gas emissions and the high proportion of saturated fatty acids in meat and milk. Animal selection for traits, such as lowered methane production and improved feed conversion efficiency is theoretically possible but requires a better understanding of the interplay between rumen microbial communities, diet and animal genotype. RuminOmics aimed to address those problems and thus decrease the environmental footprint of livestock production by using the latest –omics technologies and bioinformatics to understand how ruminant gastrointestinal microbiomes are controlled by the host animal and by the diet consumed and how it impacts on greenhouse gas emissions, product quality and feed efficiency.One of the challenges of large scale animal studies performed in multiple countries is harmonising techniques to introduce as little variation as possible. Standard methods and operating procedures were developed in the first year of the project to address the problem of inter-country variation and these are now an important reference points for future scientific and technical activities. A standard non-surgical method for collecting samples for the evaluation of the rumen microbiota was developed and this is now used routinely in the scientific community. A technique for monitoring the feed intake of animals at a farm level was refined and it is now possible to use it in breeding programmes either on farms or in genetic centres.Prior to RuminOmics the knowledge of the microbial ecology of the rumen was far behind other microbial ecosystems however by using the latest sequencing methods we have closed that gap and the methods developed in the project will also serve as a template for others to follow.The large-scale genetic association study involving 1000 cows in 4 different countries was completed to relate feed intake, digestion efficiency, milk production/composition and methane emissions to the ruminal microbiome and host genome. Extensive phenotyping was carried out on the 1000 cattle as well as genotyping with dense SNP panels. Metagenomic analysis to identify the microbial species present in the rumen was performed on 60 of the animals covering a range of low, intermediate and high methane emitters. The vast dataset established as a result of the activities of WP3 will have two main, very different impacts. The first is a strategic one. Intuitively it would seem that if the emission of an energy-rich molecule (methane) is decreased, the animal should retain more energy and therefore grow more efficiently. This is an argument that has long been used to, in part, justify research to lower methane emissions. RuminOmics WP3 has demonstrated that this is not the case, and that selecting animals for low methane output also selects for lower digestibility. Thus, a key argument in justifying lower methane emissions to farmers would be lost. In future, as a result of WP3, the alternative strategy of selecting for improved feed conversion efficiency, which in turn leads to lower methane emissions, will be the direction to take. The second impact will be that this is an unprecedented dataset, from which the partners have only scratched the surface of possible implications. Breeding companies, nutritionists and researchers will be able to mine the dataset for years to come once the results have all been posted on EBI.Experiments involving the exchange of digesta among twin cows and less genetically similar animals, and also between two ruminant species provided new insights into the role of the host in controlling the composition and function of the microbial community in the rumen. The outputs from WP4 demonstrated that (i) genetically identical cows do not necessarily share the same rumen microbiome; (ii) host animals do not have control over the whole rumen microbiome but affect certain microbial taxa and (iii) two different ruminant species (reindeer and cow) share a core bacterial microbiome with differences in species specific taxa abundance. These species specific taxa are under host animal control. In contrast, rumen ciliates and fungi are not controlled by the host animal. Between ruminant species comparisons indicated that differences in specific bacterial taxa were related to differences in fermentation pattern, methane emissions, digestion and nutrient utilization. Rumen content exchange also influenced gene expression in rumen epithelium and adipose. From these measurements it was possible to demonstrate a clear connection between changes in rumen function with alterations in the transcription of genes involved in adipose carbohydrate and lipid metabolism. By linking phenotypic differences among and within animals to specific microbial populations or genes in the rumen, it provided a proof of principle to support the collection of digesta from ruminants to inform breeding in the future. Selection for more sustainable and resilient livestock production requires the measurement of both rumen microbial and host animal traits. This information serves as a catalyst for future innovations for increasing the competitiveness of European ruminant livestock sector, maintaining its position as the major global milk producer while lowering impacts on the environment.Increasing the proportion of concentrates in ruminant diets is a well-known strategy to mitigate methane emissions. This is true especially for high concentrate diets fed to growing cattle in feed-lot. However, using high grain diets to ruminants can be questioned, since cereal grains can be directly as human food or more efficiently to simple-stomached animals such as poultry and pigs. We investigated, if the proportion of grain (barley) in a diet based on medium-quality grass silage can be reduced by increasing the proportion of high quality silage in the diet without compromising production performance and increasing methane intensity (emissions per kg milk). Our results indicated that the proportion of grain in the diet could be markedly reduced by 50% without compromising milk yield or increasing methane or nitrogen emissions. Taking into account possible differences in methane emissions from manure and carbon sequestration in soil forage-based milk production in more favourable in terms of greenhouse gas emissions than high grain diets.High quality protein supplements such as soybean meal are used in large quantities for dairy cows despite incremental protein is used with a low efficiency to milk protein. Increasing the level of protein supplementation decreased methane intensity, but the effect was rather small relative to large increases in nitrogen emissions. Feeding increased amounts of heat-treated rapeseed meal reduced the efficiency of microbial protein synthesis in the rumen. This is not taken into account in the feed evaluation and ration formulation systems. Overestimation of the protein value of heat-treated protein feeds can result in economic losses for dairy farmers. In the future feeding high quality protein supplements to ruminants cannot be considered sustainable due to the low efficiency of utilisation, increased nitrogen emissions to the environment and also because they can be used directly as human food (soybeans) or much more efficiently in the diets of poultry and pigs.Feeding fat supplements to ruminants is another common nutritional strategy to reduce methane emissions. When used at high levels fat supplements can reduce fibre digestibility and reduce feed intake and consequently production. On the other hand, fat supplements have positive effects on milk fat composition and consequently beneficial health effects in humans. However, fat supplements increase feed costs and unlikely to be used above optimal economical level unless farmers are paid premiums for improved milk fat quality. Like WP3, the impacts from WP6 are mainly twofold, and also similarly, one has strategic dimensions while the other forms a data legacy. The ring test exploring the validity of microbiome assessment in the world’s main rumen microbiology labs has exposed major technological deficiencies that prevent both the accurate description of the microbiome by amplicon sequencing and the direct comparison of data from different laboratories. Follow-up work will resolve these deficiencies and enhance the value of all similar research in the future. The data legacy is the genotyping of 1000 cows which, when put together with the phenotypic and microbiome measurements, forms an unprecedented basis for exploring the possibilities of genetic selection based on previously unknown traits.WP7 resulted in the data repository and datasets that will be the main legacy of the project for the breeding industry, nutritionists and research scientists. All the data generated by the project was processed and several analyses performed on the genetic, metagenomic and microbiome data, as well as on the phenotypic information collected from the 1000 cows sampled in Italy, UK, Finland and Sweden. This enabled the exploration of the complexity of the microbiome composition of the rumen as well as the gene content of the microbial genomes present. It was possible to identify patterns of microbial taxa and genes content present in the high and low methane emitting animals that highlighted a role of the rumen microbiome in the greenhouse gases emission from dairy cows. The genetic analysis on the cow genomes allowed also to identify chromosomal regions associated with phenotypic traits such as methane emissions but also the rumen microbiome composition. These outputs combined together potentially can have an impact in both animal breeding and animal nutrition, to generate supporting information for advanced breeding schemes to include the chromosomal mutations identified which have a role in controlling the methane emissions from the animals. The increased insights in the rumen microbiome composition from a large cohort of dairy cattle will also offer the possibility to identify correlations between the rumen composition, the nutrition efficiency and the milk production, which will have an impact in turn for the modulation of new diets and silage composition, possibly coupled with targeted prebiotics and probiotics.All these developments will have a potential deep impact in both the quality and quantity of animal productions and with an increased improvement in dairy farms management and reduced greenhouse gases emissions from livestock productions, with a positive impact on the climate changes effects for the coming years. In Task 8.1 resulting in D8.1 the remit of the DoW was expanded considerably, to the significant benefit of the scientific community studying microbiomes. Sample treatment before DNA extraction was investigated, and a protocol drawn up that ensures reliable storage of samples. Of even greater significance was the international ring test, which for the first time took a harshly critical look at the methodology leading from isolated metagenomics DNA to taxonomic assignment. The results from different laboratories were different, and the causes of the differences investigated. The RuminOmics ring test will lead to greater international harmonisation of methodology.Task 8.2 (D8.2) explored for the first time the possibility that the metaproteome can be useful as a complementary tool to metagenomics. The strengths and weaknesses of the various methods have been identified and disseminated. Once again, it will be the scientific community who will benefit.Task 8.3 (D8.7) was vital to the continuing legacy of RuminOmics. All the data will become available on a public database, providing a mine of information that will be a reference for mant years to come. No other project has the combination of nutritional, microbiome and genetic data on such a large number of dairy cows.It was discovered in Task 8.4 (D8.3) that certain fatty acids in milk are correlated with both methane emissions and the ruminal microbiome. However, the correlations were not strong enough to be predictive, in the sense that measuring the concentrations of several fatty acids cannot be an accurate proxy for methane emissions. The large dataset involved will convince others that this line of enquiry will ultimately be futile.Task 8.5 (D8.4) was highly successful. It will be possible to deduce the ruminal microbial community from oral (but not faecal) samples. This is already being implemented by one Austrian biotech company, and will hopefully be taken up by breeders.In Task 8.6 (D8.6) genomic breeding values (GEBVs) for methane emissions and daily milk production were predicted for all animals from a bivariate model. The average accuracy of GEBVs from model was 0.986 (± 0.004) for CH4 and 0.903 (± 0.013). The higher accuracy for CH4 compared to milk yield in the bivariate model can be explained with the higher estimated heritability for methane emissions (0.67 vs 0.58). These values are highly useful for breeding purposes.Main dissemination ActivitiesAt the start of RuminOmics a core group of industries and stakeholders were identified and the database updated throughout the project. This was used to create a participatory framework allowing structured and continuous dialogue between partners and the stakeholders to ensure that the project met the needs of the end users. Five ENewsletter were issued to the stakeholders and were posted on the project website to provide updates on the project, report seminars and industry days and advertise the Final Conference. Two project brochures were created within the project time frame. The first focussed on the objectives, partners, methods, project structure and expected outputs whereas the second provided progress towards the objectives. These were distributed at all the symposia and workshops organised by the project as well as conferences attended by members of the consortium.Technical VisitThe objective of the technical visit was to introduce and provide an update on the outcomes of the project and to receive feedback from industries, scientists and other stakeholders regarding the relevance of the tools and technologies being developed in the project. The visit was held on the 25th of June 2014 back to back with the Nottingham Feed Conference held on the University of Nottingham Sutton Bonnington Campus (UK).The programme of the technical visit aimed at sharing the insights of the RuminOmics project demonstrating the research of the project and to discuss the application of the research outcomes for end users. The feedback from industry was that they were interested in hearing about the state-of-the-art developments from the research fields especially the 1000-cow database being generated by the project. The farmers and representatives from the feeding and breeding industries were particularly interested in the idea of using genetics to select for reduced methane emissions and increased production efficiency as reducing the environmental impact of feed is a priority within the industry. The farmers stressed that they were only interested in making an effort if the tools and technologies developed to reduce methane emissions also improved performance, feed conversion and therefore production efficiency at the same time. Data from the 1000-cow study have since shown that although the relationships between methane emissions, digestion and efficiency are complex and that cows with lower methane emissions are not more efficient. Nevertheless there is opportunities to select cows which have lower methane emissions and higher efficiency and the evidence suggests this is related to the rumen microorganisms.Company VisitsAt the start of RuminOmics targeted industry representatives were visited and shown a presentation of the project and its aims. The project was also represented at numerous workshops and meetings throughout Europe. Participants were asked to provide feedback of the presentation and provide insight of what they would like to see from the results. In general, all visited organisations and attendants of the meetings and sessions appreciated the introduction to RuminOmics and the information and updates provided. The company visits provided valuable information on industry’s view on the expected outcomes of the project however there was a difference in the level of interest in the Western European countries and the eastern and Southern European countries. The former Soviet Eastern and Southern Europe countries have suffered enormous changes in their livestock sector since the end to the Russian federation. They reported that their main concern (apart form the Baltic states where there was more interest in breeding technologies) was overcoming poor industry structure for example low herd size and increasing competition from Eastern European milk producers. Overall it was apparent that the project would need to show some win-win technologies that reduce GHG and improve performance and efficiency to convince industry to change.One point raised in the visits was that the identification of proxies for methane emissions would be useful especially for the breeding sector. It had previously been reported that the concentration of certain fatty acids in milk fat are correlated, either positively or negatively, with methane emissions however RuminOmics found that none of the correlations were particularly strong. Farm and breed tended to be the root cause of any trends identified perhaps reflecting a dietary rather than microbiological effect. WorkshopsSix workshops were organised throughout the course of the project. Two were linked to scientific conferences (GGAA 2013, Dublin and Rowett/INRA 2014, Aberdeen) with the remit to raise awareness and expose the project to a wider international audience. The remaining four workshops were pitched at industry, policy makers and the scientific community with the aim to discuss implementation strategies for outputs and provide recommendations at a European regional level for future research and technology needs. In 2013 over 60 scientists converged on Dublin to discuss the harmonization of techniques associated with ruminal microbiome and metagenome analysis. The overall aim was to find out how the methods used in different laboratories could be optimised and harmonized with the intended endpoint that studies from different groups could be compared with confidence and not be compromised by methodological difficulties. The Aberdeen workshop concentrated on addressing the question of how the gut microbiota influenced feed efficiency and was attended by over 80 international scientists.The regional workshops were held in Warsaw (Poland), Budapest (Hungary), Lodi (Italy) and Edinburgh (Scotland). A standard programme was adopted for all the workshops.• Overview of the RuminOmics project.• Global view of environmental impact of ruminant livestock production.• Regional Livestock Sector, breeding industry and goals.• Nutrition, efficiency and emissions.• Rumen Microbial Ecology.• Does the host animal control the activity and composition of its gut microbes?• Tools for rapid analysis of animal phenotypes and the rumen microbiome.• Field-scale study of rumen function, efficiency and emissions in dairy cows ‘The 1000 cow study’.• Round table discussion-implementation of project outputs in the region and future research prioritiesAll workshops stressed that technologies to reduce methane would not be adopted unless they could be shown to increase profit or through legislation or market pressure. Animal breeding provided the most feasible option especially if projects such as RuminOmics provided improved selection methods to reduce greenhouse gases.Summer School/ELearningA summer school was held in July 2014 targeted at PhD students and young research scientists and attended by 36 students from throughout Europe. The lectures were given by members of the RuminOmics consortium and other leading scientists working in rumen microbial ecology and animal genetics. The lectures were filmed and those films combined with the presentations to provide an open access ELearning course. The courses can be accessed here http://www.ruminomics.eu/elearningFinal ConferenceThe aim of the RuminOmics final conference was to provide an overview of the ambitions and outputs from the project and offer a forum for discussing how these could be used to inform animal breeding, feeding and management to lower the environmental footprint of ruminant livestock production and improve feed conversion efficiency and product quality. The conference was titled “Ruminant livestock production: Improving efficiency and reducing environmental impact” and was held on 7th December 2015 at FIAP Jean Monnet, Paris, France. A total of 86 delegates registered for the conference with the project providing 20 scholarships for early career scientists and those currently working in a scientific field related to project activities.The conference reported on the most recent findings from the RuminOmics project and discussed how these could be used to improve the efficiency and sustainability of ruminant livestock production in Europe. The presentations were given by members of the RuminOmics consortium and can be accessed here http://www.ruminomics.eu/index.php/climate-smart-cattle-farming-breeding-presentations/The main conclusions arising from the conference wereNutrition – Carbohydrate, Nitrogen, Lipid studies• Potential to reduce CH4 emissions by nutrition rather limited • Increased proportion of concentrateso Increased costs, health problemso Grain can be used as human food or more efficiently by simple-stomached animals with much smaller CH4 emissionso Overall effects on GHG?• Economically optimal fat supplementationo Above optimum increased feed costs, reduced intake and fibre digestibility, reduced milk protein content• Increased protein supplementation may decrease CH4 per unit of product, but increase N emissionso Increased feed costso Is it ethical to feed high quality protein with 10% marginal efficiency to dairy cows Microbiome – methods and surrogates• (Unspoken) problems of community analysis are being resolved• Oral samples are useful proxies for ruminal community analysis• Metaproteomics interesting for microbiome, but practical usefulness is limited• Milk fatty acids vs. methane makes sense, but predictive value between farms is limited Host effects on rumen microbiome – lessons from reindeer and twin cow studiesEmissions and ruminant species• Rumen fermentation characteristics differ between ruminant species• Reindeer produce less methane per unit of digestible organic matter intake than cows • Reindeer excrete a higher proportion of dietary nitrogen in faeces and less in urine compared with cowsHost effects on rumen function and the microbiome• A proportion of the microbial community in the rumen was specific to cows or reindeer• A core microbiome was common to both species• After digesta exchange microbial communities in the rumen of reindeer were more similar to that of cows than the original populations• Evidence of changes in species specific to reindeer over time, consistent with an effect of the host animalTwin Cow Studies• Rumen microbial composition in identical twins was not identical• High between animal variation in response to rumen content exchange• Some microbes show interesting pattern of host genetic control (e.g. Firmicutes) Papillae transcriptome• 80% of mapped reads come from top 10% genes• Kegg pathways for these genes: Oxidative phosphorylation, nitrogen metabolism, PPAR signaling pathway, Ubiquitin mediated proteolysis, Protein processing in endoplasmic reticulum, Cholesterol biosynthesis• high number of expressed regions, where no bovine transcripts have been annotated (50% of reads)• checking sequences against other genomes in order to identify possible orthologs• characterizing these regions is highly interesting for understanding the responses in the papillae (and to the annotation of the bovine genome) Field-scale study of rumen function, efficiency and emissions in dairy cows‘The 1000 cow study’• 1000 cows was a big challenge• There is a good range in values for all phenotypes • Within countries, and overall, data are normally distributed• CH4 emissions (g/d and g/kg DMI) vary widely between cows• CH4 is not necessarily related to efficiency, so genetic selection for low emitters needs caution• Variation could be due to genetics, physiology, behaviour ... • ... is this reflected in the rumen microbiome or cow genome?Genotyping analysis: Ruminomics and the cow genome• First indications of cow genomic regions associated with o CH4 emissionso archaea and bacteria proportions• This information will be used as the basis to define genetic breeding values for these traits• What’s Next• Annotation of the genes found under the association peaks and exploration of the related pathways• The phenotypic data gathered by the project is massive and there is much more to explore to find interesting associations (e.g. Digestibility, VFA, Rumen composition etc.)Exploitation of resultsIn RuminOmics, the microbiome of more than 1000 rumen samples from cow and reindeer will be amplicon sequenced and the 16S regions analysed. Out of this 60 samples will be deep sequenced to produce complete meta-genomes and all of this data as well as the phenotypes and genotypes, methane production, N retention, the nutrition and diet, production quality and environmental interactions will be recorded and made available. This vast amount of information will uploaded to EBI public databases including the BioSamples database for the animal data and phenotypes, the European Nucleotide Archive and the EBI Metagenomics database for the sequencing data. These public databases allow complex queries to be run, at user request, and will support powerful data mining on all the project outputs. This will allow correlating rumen microbiome next generation sequencing data with all the animals’ traits, the experiments and measurements performed in the RuminOmics project. The data will initially be kept private in line with EBI privacy plans and then made public according to project policy. This will allow researchers from the scientific community long-term access to the data generated by the project and, at the same time, they will be able to upload and submit new metagenomics data and extend the RuminOmics datasets in the same public resource. The project will promote the use of this public central resource through it’s website. List of Websites:www.ruminomics.euCoordinatorProfessor John WallaceRowett Institute of Nutrition & HealthForesterhill CampusAberdeenAB25 2ZB Related documents final1-ruminomics-additional-information.pdf
