Ruminants are crucial for humans by producing milk and meat as major protein sources for human nutrition. They are characterized by their unique mode of digesting plants in their forestomach, the rumen. The rumen is densely populated by different microorganisms which are central for the breakdown of plant material which cannot be degraded by the ruminant itself. The symbiosis of ruminants and their microbiota is essential for ruminal function and enteric fermentation processes. The intimate association of ruminants and their symbiotic microbes is a result of evolutionary coexistence, and is central for energy harvest from otherwise indigestive plant material. The rumen wall, the interface between the rumen content and the host animal, has important physiological roles for the host in energy absorption, metabolism and transport of nutrients and might be associated with changes in feed efficiency (Remond et al., 1995; Kong et al., 2016). The rumen wall is covered by microorganisms that are directly attached to the rumen epithelium (Cheng et al., 1979). These bacteria are also known as “epimural” bacteria or microbiota (Mead and Jones, 1981). Many essential metabolites, such as VFA, ammonia, urea and minerals are exchanged across the rumen wall (Rémond et al., 1995), making the epimural microbiota being exposed to – and possibly participating in these nutrient exchanges.

In the last decade, a number of studies have yielded insights into the composition of the epimural microbiota in cattle and revealed that epimural bacteria are largely distinct from the microbial community in the rumen content (Malmuthuge et al., 2014; Mao et al., 2015; Liu et al., 2016). While these studies provided valuable insights into the composition of these microbial communities, our knowledge about the function of epimural bacteria and archaea is still very limited. It is known that epimural bacteria are involved in the hydrolysis of urea (Abdel Rahman, 1966; Wallace et al., 1979; Jin et al., 2017), the scavenging of oxygen (Cheng et al., 1979; Wallace et al., 1979) and in epithelial tissue recycling (Mccowan et al., 1978). It has been suggested that epimural bacteria might be involved in amino acid metabolism, using a function-predicting tool (Mao et al., 2015). Epimural microbiota may also compete with adhesive pathogenic microorganisms and they can form a protective layer on the ruminal epithelium (Hungate, 1966; Kamra, 2005).

High production dairy cows are challenged by feeding more concentrates and less forage to support high milk yields, which can result in a ruminal accumulation of VFA and a reduced buffering capacity (Kleen et al., 2003; Rustomo et al., 2006; Plaizier et al., 2008). In consequence, a depression of the rumen pH can lead to subacute ruminal acidosis (SARA), a critical metabolic disorder of cattle (Nocek, 1997; Kleen et al., 2003; AlZahal et al., 2007; Zebeli and Metzler-Zebeli, 2012).

Previously, we have performed a long-term SARA experiment with rumen-cannulated non-lactating Holstein cows fed forage (grass silage-hay mix) and increasing concentrate amounts. At the baseline, before concentrate was fed, and during the SARA challenges, rumen papillae biopsies were taken and the epimural bacterial microbiota were determined using Illumina MiSeq sequencing of 16S rRNA genes. SARA led to a decrease in diversity and species richness and induced shifts in the epimural bacterial community structure (Wetzels et al., 2016, 2017). Although this dataset revealed first insights into the composition and the complexity of the epimural bacterial microbiome, no conclusions about the metabolically active part or the functional potential of the epimural microbiome could be drawn. Transcriptome profiling of the host animal rumen epithelial tissue has gained attention by the scientific community in the last few years (Baldwin et al., 2012; Kong et al., 2016; Zhang et al., 2016; Wang et al., 2017), but functional data for the rumen wall microbiota is still missing.

Our metatranscriptome sequencing approach had two aims: the main aim was to obtain insights into the gene expression of epimural microbiota in general and secondly to determine changes in the gene expression of the epimural microbiota in response to a dietary shift from forage-based feeding to high concentrate diet induced SARA conditions.

In a recent companion study, belonging to the same feeding experiment as the current one, we analyzed the composition of the epimural bacteria using DNA-based 16S amplicon sequencing within the same SARA feeding model, using eight Holstein cows (Wetzels et al., 2017). A cross-over design was used; thus, each animal served as its own control. Overall, SARA induced major shifts in the bacterial communities (Wetzels et al., 2017). Here, our untargeted RNA-based metatranscriptome sequencing approach revealed that on phylum level, Proteobacteria, Firmicutes, and Bacteroidetes were most abundant and Bacteroidetes showed statistically significantly higher abundance in SARA conditions, which is similar to what we found in our previous DNA-based amplicon sequencing study using samples from the same time points and animals (Wetzels et al., 2017). A reduction of bacterial diversity when transitioning from a fiber-based to a starch-based diet – as found in the current study – has been reported previously (Belanche et al., 2012; Plaizier et al., 2017).

Bacteria belonging to the genera Clostridium, Campylobacter, Neisseria, Prevotella, Bacteroides, Treponema, Eubacterium, and Butyrivibrio were found to be the most abundant phylotypes on the rumen wall based on metatranscriptome sequencing. Among these, Campylobacter and Clostridium were also highly abundant in the epimural bacterial community in the 16S rRNA gene DNA-based datasets. Bacteroides, Desulfobulbus, and Desulfovibrio and Eubacterium were less abundant but among the 50 most abundant DNA-based phylotypes. Some genera with high abundance in the metatranscriptome dataset – such as Prevotella, Bacillus, Neisseria, Treponema, or Butyrivibrio – were not detected with a high abundance in our previous 16S amplicon sequencing study (Wetzels et al., 2017), which is most likely explained by the two fundamentally different methodologies used in the studies. Here, as we have applied an untargeted activity (RNA)-based approach, we identify only metabolically active phylotypes, whereas in the DNA-based amplicon sequencing, also inactive bacteria or DNA from dead cells can be amplified and detected. In addition, the phylogenetic assignment of reads is different between these two approaches. Nevertheless, overall these two different methods provide a similar overview on the epimural bacterial communities. In accordance with other studies (Shin et al., 2004b; Pei et al., 2010; De Mulder et al., 2016; Scharen et al., 2017), archaea were found to be present at the rumen wall, although at relatively low abundance (varying from 0.3 to 1% of all KEGG hits), as well as fungi (relative abundance 0.4%). Analysis of eukaryotic sequences is currently not supported by MG-RAST. Therefore, our results represent only an overview of the taxonomic distribution of fungal reads. Nevertheless, we found several fungal genera to be abundant and metabolically active at the rumen wall, of which only Aspergillus has been described in the rumen content before (Lund, 1974; Abrao et al., 2014, 2017). We found no anaerobic fungi affiliating to the phylum Neocallimastigomycota and no protozoa in our metatranscriptomics approach. The absence of protozoa on the rumen wall has already been described before (Shin et al., 2004a). In conclusion, we reveal only moderate changes in the composition of epimural communities during the switch from forage-only feeding to SARA conditions. Based on recent observations that the composition of rumen wall and rumen content microbial communities is different (Mao et al., 2015) and we did not analyze rumen content microbial gene expression, we want to emphasize that no conclusions regarding the microbiota in the rumen content can be drawn from our results.

In a next step, we were interested in analyzing the function of epimural microbes and possible changes of their gene expression from forage-only feeding to SARA conditions. A number of recent 16S rRNA gene-based sequencing studies have provided an extensive knowledge on the composition of rumen wall microbes in different ruminants under different management and feeding strategies (Petri et al., 2013; Malmuthuge et al., 2014; Mao et al., 2015; Scharen et al., 2017). Nevertheless, the knowledge about the function of rumen wall microbiota is still highly limited, as 16S rRNA gene based community surveys reveal only very limited functional insights into bacterial and archaeal communities. Recently, some studies have described the metatranscriptomes of rumen content (Kamke et al., 2016; Comtet-Marre et al., 2017; Li and Guan, 2017) and the transcriptomes of rumen epithelial tissues (Kong et al., 2016; Wang et al., 2016; Zhang et al., 2016); however, the latter was performed without analyzing microbial gene expression.

We found high levels of expression of genes related to coping with oxidative stress conditions: particularly thioredoxin reductase, glutathione peroxidase and superoxide dismutase were among the most highly transcribed genes. A role of rumen wall microbes in oxygen scavenging has been proposed by Cheng et al. (1979). An exposure of epithelial microbial communities to oxygen diffusing from tissues has been described for non-ruminant mammals (Espey, 2013) and might also occur in the rumen when ruminants swallow air. Sampling location and rumen motility might also possibly affect rumen wall exposure to oxygen. Being able to cope with reactive oxygen might be thus advantageous for epimural bacteria, as also high-grain diets can influence the redox potential in the rumen, which affects microbial communities (Friedman et al., 2017). Furthermore, also a decrease in pH as a result of high grain diets, as seen in SARA, can also lead to an increase in redox potential (Friedman et al., 2017). Our metatranscriptome sequencing data thus suggest that the rumen wall is not a completely anoxic ecosystem and that some epimural microorganisms might be facultative anaerobes as suggested in earlier studies (Pennington, 1954; Cheng et al., 1979; Wallace et al., 1979). As there were no significant differences in the expression of genes involved in oxidative stress response between baseline and SARA, we assume that the high expression of these genes might be a physiological response and not a response to SARA. To the best of our knowledge, the transcription of oxidative stress response genes has not been detected in the rumen content using a metatranscriptome approach. Here, we also provide in situ evidence for cellulose and cellobiose degradation by rumen wall bacteria. Enzymatic hydrolysis of cellulose by microbes is a key step in the degradation of fiber-rich feed in the rumen of cattle. Fibrobacter succinogenes, Ruminococcus albus, and R. flavefaciens are major cellulolytic rumen bacteria, primarily responsible for biodegradation of plant material. Even if they are primarily associated with the solid phase of rumen contents (Henderson et al., 2015), the genera Fibrobacter and Ruminococcus were also shown to be abundant at the rumen wall (Petri et al., 2013). Evidence for active cellulose degradation has already been found in metatranscriptome approaches with rumen content samples (Kamke et al., 2016; Li and Guan, 2017) which reflects the higher importance of rumen content microorganisms (particularly of the particle-associated) for cellulose degradation compared to the rumen wall microbes.

In addition to cellulose degradation, rumen wall bacteria might also be involved in breakdown of starch indicated by the expression of glycogen phosphorylase and of alpha amylase. The high expression levels of glycogen phosphorylase and (to a lesser degree) of alpha amylase suggests that starch degradation might be an important metabolic process in the epimural bacteria. It should be noted that glycogen phosphorylase might also be involved in glycogen degradation; however, based on the high energy content in both diets, we believe that degradation of glycogen is unlikely to occur. On a functional level, high levels of transcription of carbohydrate metabolism genes have also been detected in the rumen content (Kamke et al., 2016; Li and Guan, 2017).

In general, amino acid metabolism was highly expressed in rumen wall bacteria, suggesting an active role of many rumen wall bacteria in degradation of proteins, which is in line with a previous hypothesis by Mao et al. (2015). Ammonia and urea are key metabolites in the rumen which are exchanged across the rumen wall (Abdoun et al., 2006; Reynolds and Kristensen, 2008). The ruminants recycle large amounts of urea into the rumen. Ammonia is produced by degradation of urea or dietary amino acids or other non-protein nitrogen sources of the diet, being then absorbed across the rumen wall or used by microbes for microbial protein synthesis (Abdoun et al., 2006). Already a few decades ago, several studies have shown that bacteria on the rumen wall have urease activity, although the identity of these microbes remained largely unknown (Abdel Rahman, 1966; Cheng et al., 1979; Cheng and Wallace, 1979; Wallace et al., 1979). Recently, a DNA-based amplicon sequencing study provided further evidence for ureolytic potential of rumen wall microbes (Jin et al., 2017) revealing members of the Proteobacteria and Firmicutes to have ureolytic potential. Their study also showed that bacteria and archaea of the liquid and particle-associated fractions of the rumen content possess ureolytic activity. However, the rumen wall and rumen content harbor distinct ureolytic bacteria and archaea (Jin et al., 2017). Here, we provide first transcriptome-based evidence for high urease activity of rumen wall microbes in situ, suggesting that urease activity is a central function of rumen wall microbial communities. In line with the study by Jin et al. (2017), we also found members of the Firmicutes and Proteobacteria and in addition also Actinobacteria and Bacteroidetes to express urease genes. As a result of their urease activity the epimural bacteria thus may influence the rumen ecosystem by affecting urea exchange (Abdoun et al., 2006) across the rumen wall more efficiently than thought, thereby playing an important role in the nitrogen cycle. Similarly, the high expression level of other key enzymes in nitrogen metabolism such as glutamate dehydrogenase, glutamine synthase and glutamate synthase underpins the importance of rumen wall bacteria in nitrogen metabolism. Furthermore, the detection of nitrogenase gene expression at the rumen wall – although overall at relatively low level – adds to the importance of rumen wall bacteria in many different areas of nitrogen metabolism. However, the importance of the rumen content microorganisms in nitrogen metabolism should also be acknowledged and, similar to our results, a high transcription of genes involved in glutamate metabolism in the rumen content has been described recently by Kamke et al. (2016), Li and Guan (2017). Li and Guan (2017) suggested that differences in nitrogen metabolism in the rumen content may be correlated with feed efficiency. Nitrogen fixation has been shown to be present by bacteria in the rumen content (Granhall and Ciszuk, 1971; Hobson et al., 1973; Li Pun and Satter, 1975) and in fecal samples of humans (Igai et al., 2016) although at low levels. To the best of our knowledge, the results from our study describe nitrogen fixation in epimural bacteria at the rumen wall of cattle for the first time.

Putative sulfate-reducing bacteria have been found on the rumen wall in a number of recent 16S rRNA gene-based studies (Petri et al., 2013; Mao et al., 2015; De Mulder et al., 2016; Wetzels et al., 2017). Here, we provide evidence for dissimilatory sulfate reduction activity by rumen wall bacteria. Whether the production of sulfide by sulfate reducers occurs at a sufficiently high level to have a possible negative impact on the rumen epithelium is currently unclear. We assume that sulfate reduction activity in the rumen wall has a negative effect on the rumen epithelium due to the toxicity of H₂S (Drewnoski et al., 2014).

Enrichment analysis revealed only two statistically significantly differently expressed features on KO level 2 which were expressed at lower level during SARA conditions: nitrogen metabolism and oxidative stress response. On KO level function, we found only nine features to be statistically significantly differentially expressed between baseline and SARA conditions: only three features were higher expressed under SARA conditions. Furthermore, with the exception of the N-acetylneuraminate synthase, the expression level of the statistically significantly differently expressed genes was generally low. It was somewhat surprising that in spite of highly different diets and environmental conditions in the rumen such as a decreased pH during SARA conditions (Wetzels et al., 2017), our metatranscriptome sequencing results revealed only limited statistically significant changes in the gene expression pattern, as well as in community composition, of the epimural communities. This might be explained by the presence of functional guilds in the epimural bacteria: taxonomically different bacteria could fulfill the same metabolic function on the rumen wall. It has been shown that the epimural bacterial community is less sensitive to changes in environmental conditions than the microbial communities in the rumen content (Mao et al., 2015; Liu et al., 2016). This might also – at least to some degree – explain the lack of significant differences in gene expression patterns of the rumen wall microbial communities between baseline and SARA conditions. The response to SARA has been shown to be highly variable between animals (Humer et al., 2015). We can thus not exclude that our sample size of three animals might have been too low to determine an effect. Although we used only a small number of animals and due to our stringent quality control, we obtained only a comparatively low number of unique high-quality reads for our analyses. Based on our study design using a cross-over design, it should be highlighted that a control group (kept on the baseline diet) was not included, this might be helpful to reduce biases of possible time effects. Our results nevertheless reveal first insights into the gene expression of rumen wall microbiota. Future studies including more animals and higher sequencing depth will be needed to obtain a deeper understanding of rumen wall microbiota gene expression.

Our study provides the first metatranscriptome sequencing of the rumen wall microbial communities. Although we could detect only limited statistically significant differences in gene expression of the epimural microbial communities between forage-only feeding and SARA conditions, we provide the first insights into the functional potential of rumen wall microbial communities in situ. Our results indicate that urease activity, oxygen scavenging, degradation of starch and amino acids are important functions of the rumen wall microbial communities. Furthermore, we provide evidence for nitrogen fixation and sulfate reduction by bacterial communities of the rumen wall and show the presence of archaea and fungi on the rumen wall.

EM and SW performed the experiments. SW, EM, and SS-E analyzed and interpreted the data. EM, SW, MW, QZ, and SS-E drafted and wrote the paper. SS-E and EM designed the study. All authors revised and approved the final manuscript.

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.