The world’s population is estimated to reach 9.15 billion by 2050 (Opio et al., 2013). This will place increasing demands, in the region of an additional 70%, on food production systems necessitating additional crop availability for animal feed to support the ever growing human requirement for meat and milk production (Alexandratos and Bruinsma, 2012). Profitability of beef production systems relies on achieving maximum outputs using minimum inputs (Ahola and Hill, 2012). Given that feed accounts for the single largest variable cost in ruminant production systems (Finneran et al., 2010) global interest has focused on the identification of cattle that are more feed efficient (Ahola and Hill, 2012). Improving feed efficiency in the livestock sector will be crucial for sustainable animal production as it has the potential to amplify nutrient utilization from feed, improve profitability and reduce agricultural greenhouse gas emissions (Crowley et al., 2010). Residual feed intake (RFI) is a measure of feed efficiency, it is defined as the difference between observed and predicted feed intake calculated as the residuals from a multiple regression model of intake on the various energy expenditures (e.g., weight and growth rate) (Koch et al., 1963). It is the preferred measure of feed efficiency due to its independence from production traits, thus, inter-animal differences in RFI is most likely as a result of variation in component metabolic processes rather than variation due to differences at the production level (Nkrumah et al., 2006).

Ruminants rely on the microbiota (bacteria, archaea, ciliate protozoa, fungi, and viruses) present in the rumen to degrade their feed and produce volatile fatty acids (VFAs), microbial proteins and vitamins. The primary VFAs produced in the rumen, acetate, propionate, and butyrate, provide up to 70% of the host’s energy requirements (Flint et al., 2008). Animal performance is dependent on numerous factors including genotype, stage of development and the chemical composition and quantity of feed offered (Carberry et al., 2012; Keogh et al., 2015; McDonnell et al., 2016; Shabat et al., 2016), which also influence the rumen microbiome (Carberry et al., 2012; Jami et al., 2013; Shabat et al., 2016; McGovern et al., 2017). Therefore there is potential for a link between feed efficiency and microbiota present in the rumen (Ahola and Hill, 2012). Research from our own group and others have indeed identified links between rumen microbiome and inter-animal variation in feed efficiency phenotype (Carberry et al., 2012; Jewell et al., 2015; Myer et al., 2015; Roehe et al., 2016; Shabat et al., 2016; Li and Guan, 2017). Shabat et al. (2016) previously reported rumen microbiome genes and species that could predict (91% accuracy) of the variation in an animal’s feed efficiency phenotype and discerned that reduced rumen microbial community diversity may support a more feed efficient animal.

Although research investigating the role of the rumen microbiome in influencing feed efficiency phenotype exists, data are limited, especially for beef cattle. Furthermore, many of these studies are restricted to only investigating rumen liquid digesta and use alternate methods of measuring feed efficiency, such as feed conversion ratio (FCR), which can be problematic due to the unfavorable genetic correlations with between growth rate and mature body size (Schenkel et al., 2004). This implies that selection of cattle for improved FCR will lead to an increase in mature body size and therefore, maintenance requirements (Schenkel et al., 2004; Crews, 2005). This has negative economic and environmental ramifications for beef production (Arthur et al., 2001; Schenkel et al., 2004; Crews, 2005).

In the current study, RFI phenotype was determined for a combination of four cohorts of Simmental bulls over 4 years. Archaeal and bacterial populations present in both the rumen solid and liquid digesta of Simmental bulls divergent for RFI phenotype were determined using amplicon sequencing targeting the 16S rRNA gene. Rumen fluid from the bulls was further analyzed to establish its metabolomic composition. Differences in microbial populations and metabolic profiles were examined between divergent phenotypes. Archaeal and bacterial populations were correlated to RFI phenotype as well as inter-taxonomically, in order to investigate possible microbial drivers of feed efficiency. As data were derived from a combined population, there was increased strength in observed association due to the conservation of the relationship between the cohorts. Our data address gaps in knowledge surrounding the rumen microbial ecosystem and provides evidence of potential mechanisms relating components of the rumen microbiome to host feed efficiency phenotype.

All procedures involving animals from years 1–3, in this study were conducted under an experimental license from the Irish Department of Health and Children, while procedures involving animals from year four were conducted under an experimental license (AE19132/P011) from the Health Products Regulatory Authority. Both licenses where issued in accordance with the cruelty to Animals Act 1876 and the European Communities (Amendment of Cruelty to Animals Act 1876) Regulations 2002 and 2005.

In total, 30 Simmental bulls divergent in RFI ranking (n = 15 HRFI and n = 15 LRFI) were used in this study (P < 0.0001) (Table 1). There was no interaction between RFI and year or individual microbial taxa and RFI (P > 0.05). HRFI bulls had an increased DMI in comparison to LRFI animal and consumed 23.5% more than their LRFI counterparts, respectively (P < 0.001) (Table 1). ADG and MBW did not differ between the treatment groups (P > 0.05) (Table 1).

The Illumina Miseq amplicon sequencing generated 16,711,973 total reads giving an average of 168,807.8 ± 57,658 reads per sample. This reduced to 6,281,459 when sequences were merged and quality filtered. The average number of counts per sample that were assigned to an OTU (post filtering) was 123,729 ± 45,617.

This study examined the effect of RFI phenotype on ruminal bacteria and archaeal populations and VFA profiles of Simmental bulls offered an ad libitum concentrate diet. Analysis of population diversity in solid and liquid phases of the rumen digesta showed that these fractions contained similar community profiles. This is consistent with a study from Ji et al. (2017), which discerned similar bacterial diversity from unseparated rumen digesta in comparison to rumen samples fractionally separated into liquid and solid via cheese cloth. Schären et al. (2017) also found a high level of similarity between the solid and liquid ruminal fractions, despite a more stringent separation technique. This is likely due to the homogenous nature of the two fractions (Sadet et al., 2007). In order obtain a stringent separation of solid and liquid fractions require separation steps from which nucleic acid degradation could potentially ensue. Nucleic acid integrity was a priority for this experiment therefore; simple but efficient separation via cheesecloth was utilized (Fliegerova et al., 2014).

Prevotella and an uncultured Proteobacteria genus of bacteria, identified as a member of the Succinivibrionaceae family, were the dominant genera in both liquid and solid digesta present in the LRFI and HRFI animals of the current study. This is supported by global studies (Henderson et al., 2015) where these genera were found to be more abundant in cattle fed high-energy concentrate diets (Henderson et al., 2015). Prevotella and unclassified Succinivibrionaceae are the major producers of propionate and succinate in the rumen of animals fed concentrate rich diets (Murphy et al., 1982; Chen et al., 2017). Succinate does not accumulate in the rumen and is rapidly converted to propionate (Suen et al., 2011), which is transported out of the rumen via the rumen wall and converted to glucose through the gluconeogenesis pathway (McDonald et al., 2002). The high relative abundance of Prevotella and unclassified Succinivibrionaceae imply that the divergent cohorts should both be producing a relatively high concentration of propionate, thus, the lower molar proportion of propionate in the rumen of LRFI animals suggest alternative fermentative or absorptive patterns in these animals. Low propionate concentrations have previously been associated with feed inefficient cattle (Krueger et al., 2008). It has been suggested that inter-animal variation in ruminal VFA concentrations may contribute to some of the observed variation in RFI (Guan et al., 2008; Krueger et al., 2009; McDonnell et al., 2016; Shabat et al., 2016), however, to date there has been a lack of consistency in the results of VFA profiles of cattle divergent for feed efficiency (Guan et al., 2008; Hernandez-Sanabria et al., 2010; Fitzsimons et al., 2013; McDonnell et al., 2016; Shabat et al., 2016). There was an increase in A:P ratio in LRFI in comparison to HRFI bulls, this can be attributed to the reduced molar proportions of propionate in the rumen of LRFI animals and consistent molar proportions of acetate between LRFI and HRFI animals. A higher ruminal A:P ratio in LRFI heifers was previously observed by Krueger et al. (2009).

The phylum of Fibrobacteres showed a negative relationship with RFI in both ruminal fractions as did OTUs assigned as F. succinogenes. F. succinogenes is known for its importance as an efficient and prolific degrader of cellulose in the rumen, which produces three major end products; succinate, acetate, and formate (Miller, 1978). Unlike other rumen bacteria which derive energy from numerous polysaccharide sources, F. succinogenes is specialized only for the utilization of cellulose as a substrate (Suen et al., 2011). The genome of F. succinogenes encodes numerous enzymes capable of degrading a wide array of polysaccharides and it has been suggested it uses these enzymes to remove plant cell wall hemicelluloses in order to gain access to cellulose (Suen et al., 2011). It is hypothesized that these hemicellulose breakdown products provide growth substrates for the other bacteria present in the rumen (Suen et al., 2011) which is supported by our data by the strong relationships observed between the relative abundance of Fibrobacter and other genera present in the rumen. As mentioned above, succinate does not accumulate in the rumen, the decarboxylation of succinate, which is produced by F. succinogenes, is predicted to be facilitated by Selenomonas ruminantium in the rumen (Suen et al., 2011), which exhibits a positive relationship with F. succinogenes in the current study. The 16S rRNA results indicate that there is potentially an increase in the cellulolytic activity in the rumen of LRFI bulls and to further add evidence to this we note a positive correlation between F. succinogenes with acetate which is one of the main products of F. succinogenes’s metabolic pathway. It is proposed that the LRFI animals have a higher abundance of F. succinogenes and that the succinate produced by this bacterium is rapidly converted to propionate by S. ruminantium thereby providing a higher level of substrate available for gluconeogenesis. However, this hypothesis was based on 16S DNA, which may not represent metabolically active species and therefore will require further investigation.

Increased rumen butyrate has been associated with greater feed efficiency in cattle (Guan et al., 2008; Shabat et al., 2016) and the findings of the current study are in agreement with this result, with a higher concentration of butyric acid tending to be observed for the LRFI bulls. Correspondingly we noted increased in plasma βHB in LRFI animals. Butyric acid has been associated with increased mitotic activity in the rumen epithelial cells (Sakata and Tamate, 1978). In the rumen epithelium, butyrate is converted to βHB, which is an important energy source for many host tissues (McDonald et al., 2002). Efficient cattle have been shown to have increased rumen epithelium thickness compared with inefficient contemporaries (Lam et al., 2017). It is possible that greater butyrate availability in the rumen of LRFI animals may increase the rumen surface area for absorption, therefore proliferating absorption of high energy VFAs, such as propionate from the rumen to the liver (Daniel and de Resende Júnior, 2012; Keogh et al., 2017). Decreased absorption of VFAs across the rumen epithelium has been previously associated with low feed efficiency (Albornoz et al., 2013; Zhang et al., 2013). In addition, research from our own group has provided evidence of up regulation in transcriptional activity in the rumen epithelium of feed efficient cattle, suggesting a higher rate of nutrient absorption under this feeding regime (Keogh et al., 2017).

OTUs identified as Butyrivibrio, a butyrate producing genera, were negatively correlated with RFI in both the solid and liquid digesta. Myer et al. (2015) had previously identified a higher abundance of Butyrivibrio in efficient steers. Butyrivibrio are members of the Lachnospiraceae family. OTUs assigned to Coprococcus, another butyrate producing genus from the Lachnospiraceae, showed a negative relationship with RFI in the rumen liquid digesta in the current study. This association has been previously observed in the rumen of LRFI cattle (Jewell et al., 2015). Shabat et al. (2016) noted an enrichment of Coprococcus and the metabolic pathways it encodes in LRFI dairy cows. Increased abundance of Lachnospiraceae is noted in HRFI animals in the current study at genus level. In our data, OTUs only assigned as far as Lachnospiraceae exhibited both a positive and negative relationship with RFI. Lachnospiraceae has been positively correlated with feed efficiency (Rius et al., 2012; Shabat et al., 2016), however, certain members of this family have also been found to be more abundant in inefficient animals in the current study and in the literature (Shabat et al., 2016; Li and Guan, 2017). Ruminococcaceae and Ruminococcus OTUs were negatively correlated with RFI, indicative of a role in enhancing the efficiency in LRFI bulls. OTUs assigned as Ruminococcus flavefaciens held a negative relationship with RFI. R. flavefaciens are an active cellulolytic bacterial species which are succinate producers, again suggestive of alternate fermentation patterns in the rumen of LRFI bulls (Pettipher and Latham, 1979).

The phylum of Cyanobacteria exhibited a negative relationship with RFI. The OTUs driving this relationship were from the order of YS2. The YS2 order has previously been found in the ruminant gut (AlZahal et al., 2017; Neves et al., 2017) and has many functions, including obligate anaerobic fermentation, nitrogen fixation, syntrophic hydrogen production, and synthesis of vitamin B and K synthesis (Di Rienzi et al., 2013). The phylum of Cyanobacteria is known for its photosynthetic capability but a new candidate phylum has been suggested, “Melainabacteria” which would encompasses non-photosynthetic members currently assigned to this phylum, inclusive of YS2 (Di Rienzi et al., 2013). Despite YS2 being previously identified in cattle (AlZahal et al., 2017; Neves et al., 2017), no previous association with feed efficiency has been recorded. The positive correlation between RFI and YS2 observed in the current study may indicate a possible niche role for YS2 in nutrient supply (Zeng et al., 2015) in LRFI bulls.

Tenericutes was also negatively correlated with RFI, OTUs within this phylum, including RF39, were also negatively correlated with RFI. This contradicts Jami et al. (2014) who identified a positive tentative correlation between RFI and RF39 in dairy cows, while in humans lean BMI subjects were enriched with RF39 (Goodrich et al., 2014), suggesting humans with reduced energy harvest capacity harbor RF39.

YRC22 from the Paraprevotellaceae family showed a negative relationship with RFI as did OTUs assigned as YRC22. YRC22 has previously been shown to be more abundant in the rumen solid fraction of low RFI dairy cattle (Jewell et al., 2015) and in dairy cattle supplemented with tannins and are therefore deemed as more efficient than their supplemented controls (Díaz Carrasco et al., 2017). Myer et al. (2015) found no difference in abundance of YRC22 steers divergent for feed efficiency, however, an alternate measure of feed efficiency was utilized by Myer et al. (2015) classifying animals based on their feed intake and growth measurement, which may have led to confounding results. The literature indicates no conclusive result on the relationship between YRC22 and feed efficiency in cattle, however, Paraprevotellaceae has been found to be more abundant in obese rats, indicating this family may facilitate increased energy harvest in the rat gut (Hakkak et al., 2017).

Residual feed intake is also associated with the lactate producing genera with OTUs annotated as Sharpea p-3329-23G2 and p-3329-23G3 showing a positive relationship with RFI. Jewell et al. (2015) found an increase in abundance of Sharpea OTUs in HRFI dairy cows complementing the results of the current study. Additionally Sharpea is from the Erysipelotrichaceae family the abundance of which was reduced in LRFI porcine feces in comparison to HRFI (McCormack and Curiao, 2017). This family has also been associated with inflammation in the human gut (Kaakoush, 2015) indicating possible links with inefficiency and host inflammation.

Methanobrevibacter millerae YE315 is negatively correlated with RFI in the rumen fluid. This is a hydrogenotrophic species which is a member of the gottschalkii clade, and encodes only one version of the methyl coenzyme reductase (MCR) isoenzyme, mcr I (Li, 2016). This indicates it thrives in a low hydrogen environment (Reeve et al., 1997). Our group has previously observed a negative relationship between this genera and RFI in both Charolais and Holstein Friesian steers fed a grain based diet (McGovern et al., unpublished). The relative abundance of Fibrobacter is linked with that of Methanobrevibacter. It is possible that Fibrobacter directly influences proliferation of species within these genera through the production of hydrogenotrophic substrate formate or indirectly (Suen et al., 2011); through providing substrate to other species whose fermentation end products are substrate for members of Methanobrevibacter genera. This relationship is also evident with another methanogenic genera; vadinCA11 from the Methanomassiliicoccaceae family. This family is obligate H₂-dependent methylotrophs, utilizing methyl groups from methanol and methylamines (mono-, di-, and tri-methylamine) and methyl thiols for methanogenesis (Lambie et al., 2015). It is probable here that Fibrobacter indirectly provides substrates (Suen et al., 2011) to Methanomassiliicoccaceae. For example, Succinivibrionaceae degrade pectin a structural polysaccharide contained in certain plant cell wall to produce methanol (O’Herrin and Kenealy, 1993), which is a substrate required for the growth of Methanomassiliicoccaceae (Paul et al., 2012). Further functional analysis will be required to investigate the proposed relationships from these data.

In agreement with the hypothesis that LRFI phenotype is potentially related to increased capability of rumen microbiota to harvest nutrients from ingested feed, in the solid ruminal digesta phase F:B is negatively correlated with RFI. This indicates that an increased F:B ratio may support a more feed efficient animal. In dairy cattle F:B ratio has been associated with increased milk fat-yield (Jami et al., 2014). This ratio has also been used in human and murine studies with increased F:B ratio an indication of obesity (Ley et al., 2005; Turnbaugh et al., 2006) with Ley et al. (2006) suggesting that in obese mice the higher ratio was due to an effective release of energy through digestion. In a recent study Shabat et al. (2016) suggested efficient dairy cattle had a less diverse microbial community that produced relevant metabolites to promote growth. In the human gut a low diversity gut microbiome has been associated with obesity (Consortium, 2012). In relation to animal protein production, a low diversity microbial population in the rumen focused on promoting energy yield from feed is desirable (Shabat et al., 2016). However, there was no evidence of divergence in microbial diversity between cattle divergent for RFI in the current study.

In conclusion this study examined the effect of RFI phenotype on ruminal bacterial and archaeal populations present in the rumen of Simmental bulls offered a concentrate based finishing diet. This study suggests that RFI phenotype does not affect the overall diversity of the bacterial or archaeal communities present in the rumen of beef cattle. However, it also compliments previous research indicating that microbiota with high energy harvesting capacities are required for feed efficient weight gain with a relationship evident between F:B ratio and feed efficiency. The majority of the taxa identified as exhibiting a relationship between in their relative abundance with host feed efficiency were recognized to possess potent cellulolytic capacities or had previously been noted to cause metabolic shifts in their host. The data indicate that there is an increase in residual butyrate in the rumen of LRFI bulls which corresponded to an increase in βHB, perhaps suggestive of increased epithelial absorption. Fibrobacter succinogenes showed a negative relationship with RFI which was apparent from species to phylum level. The unique cellulolytic metabolism of Fibrobacter succinogenes suggests the bacterium may have contributed to host efficiency by providing substrates to other microbial population in the rumen. We provide evidence that abundance of certain bacterial genera and bacterial and archaeal OTUs exhibit relationships with RFI phenotype, therefore, phenotypic expression of RFI may potentially be influenced through the role of bacterial and archaeal in ruminal degradation of complex plant polysaccharides. The data also suggest that it is probable that microbial species or strains may impact the efficiency of the animal more than an overall ruminal community shift. Therefore, in order to investigate phenotypic impact of RFI on the microbiome, global high depth metatranscriptomic sequencing is required to deduce this complex symbiotic relationship.

DK, MM, AK, and SW conceived and designed the experiments. EM, MSM, and CF performed the experiments. EM, MSM, and CF analyzed the data. SW, DK, MM, AK, SW, MSM, and CF contributed reagents, materials, and analysis tools. EM, SW, DK, MG, AK, MSM, MM, and CF result interpretation and wrote paper.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.